article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
_ introduction._ in the main text , we illustrated key ideas by examining several examples : ( i ) the quantum one - dimensional ising matter coupled gauge theory ( the @xmath137 higgs model ) and its dual quantum spin chain in an external field , ( ii ) a two - dimensional quantum xy model and its solid - on - solid ( @xmath68 ) type gauge theory dual , and ( iii ) the quantum @xmath60 matter coupled ( or abelian higgs ) gauge theory and its @xmath68 matter field coupled gauge dual .
in what follows , we illustrate how our concepts can be similarly worked out in other examples : ( iv ) the two - dimensional transverse field ising model and its quantum ising gauge theory dual , ( v ) the self - dual two - dimensional quantum @xmath137 higgs model , ( vi ) the two - dimensional quantum @xmath138-clock model and its @xmath139 gauge theory dual , ( vii ) the two - dimensional @xmath139 higgs theory which we earlier illustrated to be self - dual for all @xmath138 , ( viii ) the two - dimensional @xmath139 higgs theory and its extended toric code dual , and ( ix ) the quantum xy model on the kagome lattice and its @xmath68 gauge theory dual . taken together , these examples demonstrate how holographic symmetries and generalized ops and associated correlators can be systematically derived in disparate theories .
_ the two - dimensional transverse field ising model and its quantum ising gauge theory dual._ as in the main text , we will perform dualities for finite size systems and explicitly label the cartesian coordinates @xmath140 of sites @xmath141 on an @xmath142 square lattice .
the planar transverse - field ising model for such any @xmath142 square lattice system is given by @xmath143 the first term denotes the transverse field along the @xmath144 direction while last two terms correspond to ising interactions ( along the internal @xmath145 direction ) associated with horizontal and vertical links of the lattice respectively .
this transverse field ising model is dual to ( the gauge - invariant sector of ) the @xmath9 ising gauge theory in the infinite size limit @xcite .
for open boundary conditions , we obtain via a bond - algebraic duality the _ exact _ dual of @xmath146 which reads @xmath147 following our earlier conventions , the link indices @xmath148 , that appear _ explicitly _ in @xmath149 labeling spins @xmath150 will be termed_allowed indices _ , with all other link indices being _
forbidden_. this convention will be invoked to define the plaquette product @xmath151 which will omit forbidden links .
the plaquette operators will be of a similar nature to those of fig .
[ dual_ising_fig ] of the main text . in the `` bulk ''
, the plaquette operator @xmath151 reads @xmath152 at the topmost left site ( see fig .
[ dual_ising_fig ] ) , @xmath153 hence @xmath154 involves one spin @xmath155 labelled by a forbidden link index ( see fig . [ dual_ising_fig ] ) . elsewhere on the boundary of the square lattice ,
the remaining boundary plaquettes are determined by eq . .
we now examine to the bond algebra- the set of algebraic relations between the interaction terms in the hamiltonian of eqs .
( [ hi ] , [ hig ] ) .
this will allow us to establish a duality mapping between the transverse field ising model and the ising gauge theory .
a natural partition of the transverse field ising hamiltonian @xmath146 of eq .
( [ hi ] ) is into three type of bonds : ( a ) the transverse fields @xmath156 , ( b ) horizontal ising interactions @xmath157 , and ( c ) vertical ising interactions @xmath158 .
the bond algebra of this system is exhausted by the following relations : \(2 ) single site bonds of type ( a ) anticommute with the two - site bonds of type ( b ) [ or ( c ) ] whenever any two such bonds share a common site on the lattice ; this is a trivial consequence of the anticommutation relation @xmath159 .
\(3 ) whenever bonds of type ( a ) share no common lattice sites with bonds of type ( b ) or ( c ) then any such bonds commute with one another .
that this is so is readily seen as @xmath160=0 $ ] whenever @xmath161 .
\(5 ) the product of any horizontal / vertical bonds of types ( b ) and type ( c ) around any closed plaquette is equal to the identity operator .
that is , as @xmath163 for any site @xmath4 , for all @xmath164 , @xmath165[\sigma^z_{(i-1,j)}\sigma^z_{(i-1,j-1 ) } ] \nonumber \\ \times [ \sigma^z_{(i-1,j-1)}\sigma^z_{(i , j-1 ) } ] [ \sigma^z_{(i , j-1)}\sigma^z_{(i , j)}]\end{aligned}\ ] ] upon removing gauge redundancies , an identical set of relations appears for the bonds of the ising gauge theory of eq .
( [ hig ] ) .
to underscore this , we partition eq .
( [ hig ] ) into three types of bonds as follows : ( a ) the plaquette operators @xmath166 , ( b ) gauge fields @xmath167 on vertical and ( c ) gauge fields @xmath168 on longitudinal links .
that similarity of algebraic relations underlies the duality mapping between the theories . that the two theories of eqs .
( [ hi ] , [ hig ] ) are indeed dual to each other on an @xmath142 square lattice ( with of an size arbitrary @xmath2 ) is readily seen by the mapping of the bonds ( a , b , c ) @xmath169 ( a,b,c ) , or @xmath170 once the local gauge symmetries of the ising gauge theories are removed and set to unity , @xmath171 with the constraint of eq .
( [ long_removal ] ) implemented , redundant gauge degrees of freedom are eliminated- applying a gauge transformation of the form of eq .
( [ long_removal ] ) does not lead to a new state . as the mapping of eqs .
( [ long_explanation_ising ] , [ long_removal ] ) preserves all of the algebraic relations amongst the bonds in both theories of eqs .
( [ hi ] , [ hig ] ) , this mapping implements an duality @xmath172 for any @xmath142 square lattice . similar to the main text , we briefly elaborate on the change of number of degrees of freedom once eq .
( [ long_removal ] ) is implemented and how this leads to an identical number of degrees of freedom in both the transverse field ising and ising gauge theory .
the counting of the number of degrees of freedom is identical to that of the @xmath60 system examined in the main text .
we repeat it anew here for the ising systems at hand .
the hamiltonian @xmath149 describes @xmath173 spins and possesses @xmath174 gauge symmetries which are removed by the constraint of eq .
( [ long_removal ] ) .
this leaves a total of @xmath175 independent spins .
this is precisely the number of spins in @xmath146 .
now , we turn to a quintessential feature seen by this duality- that of the holographic symmetry . as , by virtue of our definitions above of forbidden indices
, @xmath155 makes an appearance only in @xmath154 ( see fig .
[ dual_ising_fig ] ) and @xmath176 appears nowhere in the hamiltonian of eq .
( [ hig ] ) , the operator @xmath155 constitutes a _ boundary symmetry _ of @xmath149 .
what is this boundary symmetry s origin ?
the answer is afforded by a simple calculation , @xmath177 _ thus , the boundary symmetry @xmath178 is holographic .
_ as seen by eq .
( [ long_holography_ising ] ) , this symmetry is a `` hidden '' dual of the global @xmath9 symmetry of the ising model .
its recognition became possible with the specific choice of operators and use of bond algebras which suited to treat exactly for finite size systems and their boundaries .
we conclude our examination of this duality and its consequences by deriving the op of the ising gauge theory .
as we will establish now , this op will be given by string products- products of operators along connected links . to write this op
, we need to set a few preliminaries . towards this end , as in the main text , we let @xmath99 denote a path ( set of concatenated links ) in the ising model s lattice , starting at site @xmath179 and ending at @xmath180 so that @xmath181 we furthermore let @xmath182 denote that link in the lattice of the ising gauge theory such that @xmath183 . from these definitions , it follows that @xmath184 denote , respectively , the ground states of the ising gauge ( eq . ( [ hig ] ) ) and transverse field ising model ( eq .
( [ hi ] ) ) .
thus , putting all of the pieces together , the ( infinite separation limit of the ) string correlator @xmath185 is the generalized op of the ising gauge theory .
this op is the dual of the landau order as it appears in eq .
( [ landau_op ] ) .
_ the self - dual @xmath9 higgs model in two dimensions._ we next examine the self - dual quantum ising matter coupled gauge theory on the square lattice .
similar to its one - dimensional variant of eq .
( [ hh2 ] ) , the hamiltonian of this system is given by @xmath186 similar to the main text , in the @xmath9 higgs hamiltonian of eq .
( [ 2dz2higgs ] ) , @xmath187 are pauli operators that reside on the lattice site @xmath4 ( which play the role of quantum matter fields ) while the quantum gauge fields @xmath188 are pauli operators that lie on a square lattice link that connects the site @xmath4 with a neighboring site site @xmath189 along one of the two ( @xmath190 square lattice directions .
following all that we have stated thus far , the reader can verify that the correspondence of bonds @xmath191 along with analogous relations for the matter fields @xcite proves that @xmath192 is self - dual in its gauge - invariant sector .
it is straightforward to check that @xmath107 maps to the path @xmath99 just as before under the self - duality mapping of eqs .
( [ sd1],[sd2 ] ) . hence , @xmath193 where @xmath194 denotes the ground state of the higgs hamiltonian of eq .
( [ 2dz2higgs ] ) .
the gauge - invariant correlator on the right - hand side of eq .
( [ z2fm ] ) is related to known string ops for higgs theories @xcite .
the reader is referred to these works for details concerning asymptotic behaviors of string operators in the confined and deconfined phases . for completeness
, we briefly reiterate that in these ops @xcite , the expectation value of eq .
( [ z2fm ] ) is divided by string @xcite ( or wilson loop @xcite products which are derived from @xmath99 .
this division enables the extraction of the non - trivial asymptotic large @xmath99 behavior of the string correlator of eq .
( [ z2fm ] ) ; this factor is , equivalently , related to the normalization of the state generated , in an imaginary time representation , by the application of the string product on the ground state @xcite .
other relations to effective line tension resulting from combined bare line ( borne by matter coupling ) and surface ( generated by pure gauge fields ) tensions are discussed in @xcite . _ the @xmath61 gauge model._ we may place the @xmath9 theories of the previous sections and the @xmath60 theories in the main part of the paper at the ends of an infinite sequence of models with @xmath61 symmetry , @xmath195 .
a `` clock model '' spin is associated to the group @xmath61 is described by two unitary @xmath196 matrices @xmath197 satisfying @xmath198 and @xmath199 , with @xmath200 a @xmath138th root of unity @xcite .
when @xmath201 , these correspond to the pauli operators @xmath202 .
thus the hamiltonian for the quantum @xmath138-clock model can be written as @xmath203+h.c.\label{pclock_h}\ .\nonumber\end{aligned}\ ] ] similar to the @xmath9 theories , the @xmath138-clock is dual to ( the gauge - invariant sector of ) the @xmath61 gauge theory @xmath204+h.c.\ .\nonumber\end{aligned}\ ] ] following , verbatim , the same convention as earlier ( as also marked in fig .
[ dual_ising_fig ] ) ) yet now applied to these operators , the plaquette operator @xmath151 for plaquettes in the `` bulk '' is given by @xmath205 on the boundary , @xmath206 the remaining boundary plaquettes are determined by eq .
provided operators labelled by forbidden link indices ( those extend beyond the system boundaries ) are omitted , see fig .
[ dual_ising_fig ] ) ) .
the correspondence of bonds @xmath207 establishes the ( gauge - reducing ) duality @xmath208 .
the operator @xmath209 is an _ holographic symmetry _ of @xmath210 , since @xmath211 the operator on the right - hand side of eq .
is the _ global _ @xmath61 symmetry of the quantum @xmath138-clock model of eq .
( [ pc ] ) . as we see from the left - hand side of eq .
, this operator is dual to the _ boundary symmetry _ of the gauge theory , @xmath212 .
thus , this symmetry , @xmath212 , is holographic .
as we have in earlier examples , we now systematically devise an order parameter for the gauge theory .
this abelian gauge theory exhibits , as most others , both confined and deconfined phases . the program for methodically deriving a generalized op diagnostic will replicate our calculations in earlier examples .
first , we write down a landau correlator which directly captures the long range order in the landau type system ( in this case , the clock model of eq .
( [ pc ] ) ) . relying on the duality between the quantum clock model and the quantum @xmath213 gauge theory
, we then determine the corresponding correlator which serves as a generalized op for the gauge theory of eq .
( [ pg ] ) .
the landau theory of the quantum p - clock exhibits long range order as seen by the finite value of the two - point correlation function , in the limit of infinite separation , @xmath214 where @xmath215 denotes the ground state of the p - clock model of eq .
( [ pc ] ) .
we now map this two - point correlator onto a generalized op for the gauge theory by invoking the duality that we found . towards this end ,
our steps will replicate , yet once again , those that we invoked in earlier examples .
let @xmath99 be an _ oriented _ path from @xmath4 to @xmath100 , and @xmath101 its links oriented accordingly .
furthermore , we employ the convention that @xmath216 if @xmath103 points from @xmath4 to @xmath104 , and @xmath217 if @xmath103 points oppositely from @xmath104 to @xmath4 . with this conventions ,
@xmath218 if we define @xmath107 as the set of links mapping to the path @xmath99 under duality and @xmath219 , then we will find that @xmath220 where @xmath221 is the ground state of the gauge theory @xmath210 . in the confined phase of the gauge theory
, the generalized op @xmath222 assumes a non - zero value in the limit of infinite @xmath107 .
by contrast , in its deconfined phase , the gauge theory of eq .
( [ pg ] ) has a vanishing value of @xmath222 as @xmath107 is made asymptotically large .
this result constitutes a generalization of our earlier ising ( i.e. , @xmath201 ) results .
thus , similar to our earlier examples , putting all of the pieces together , the ( infinite separation limit of the ) string correlator @xmath222 is a generalized op for the @xmath61 gauge theory distinguishing confinement from deconfinement , and derived from simple considerations of spontaneous symmetry breakdown , duality , and holographic symmetries .
_ the self - dual @xmath61 higgs model._ we now extend the earlier ising results to the @xmath61 system , review the self - duality of the @xmath61 gauge theory when it is coupled to @xmath61 matter fields , and derive new string operators for this theory .
this matter coupled gauge theory is the quantum @xmath61 higgs theory .
that is , we will augment the ( pure ) gauge theory @xmath210 of eq .
( [ pg ] ) by terms that minimally couple the gauge fields on the lattice links @xmath223 to matter fields @xmath224 and @xmath225 at sites that form the endpoints of those links as well as by additional matter only terms @xmath226 , @xmath227+h.c .. \ ] ] from the minimal coupling term , where @xmath118 is the power of the gauge field on the link that is coupled to the matter field , similar to the @xmath60 example in the main text , it is evident that @xmath118 plays the role of a charge .
as @xmath228 , this charge can assume the values @xmath229 $ ] ( where @xmath230 $ ] denotes the integer par of its argument ) .
on augmenting the hamiltonian of eq .
( [ pg ] ) by the terms in eq .
( [ matterc ] ) , the resulting model hamiltonian @xmath231- the ( charge @xmath118 ) @xmath61 higgs model- is self - dual in its gauge invariant sector @xcite .
specifically , this self - duality is given by the mapping @xmath232 for the gauge fields , and the transformation @xmath233 for the matter fields .
the string operator in eq . , becomes , on invoking eqs .
( [ psd1 ] , [ psd2 ] ) , @xmath234 in eq .
( [ pfm ] ) , @xmath235 denotes the ground state of the higgs hamiltonian ( given by @xmath236)- the sum of the hamiltonians in eqs .
( [ pg ] , [ matterc ] ) ) .
similar to our earlier conventions , in eq .
( [ pfm ] ) we set @xmath237 if @xmath103 points from @xmath4 to @xmath104 , and @xmath238 otherwise . as in our discussions for the other examples of higgs theories concerning string ops and their normalization , in the infinite separation ( i.e. , @xmath99 ) limit , the gauge - invariant correlator on the right - hand side is related to a normalized generalized op proposed in ref .
@xcite . taken on a _ closed _ path , the right - hand side of eq . reduces to a wilson loop . _
the @xmath61 extended toric code model._ the toric code model @xcite and its extensions , e.g. , @xcite have gained much interest in recent years .
their original motivation was to serve as a caricature of toy models in which some of the basic concepts of topological quantum computing can be explained @xcite . since the model and its variants have been investigated heavily large on their own right .
the original toric code model had an ising symmetry . in the current section
, we will further employ its @xmath61 variant . in the current context
, our interest in this system derives from the fact that the @xmath61 higgs model of the previous section is dual to a generalization of the extended @xmath61 toric code model ( petc ) .
this model is given by @xmath239+h.c.\ , \end{aligned}\ ] ] on the square lattice , with @xmath240 .
the standard extended toric code is recovered for @xmath201 , in which case @xmath241 is the only possible value .
the @xmath61 toric code discussed in ref .
@xcite is recovered for @xmath241 .
a gauge - reducing duality connecting @xmath231 to @xmath242 is given by @xmath243 while leaving the gauge fields invariant , @xmath244 and @xmath245 .
this duality removes degrees of freedom that lie at lattice sites . while the hamiltonian @xmath242 has no global symmetries , the @xmath61 higgs model has one global symmetry .
similar to our earlier discussions , the duality @xmath246 maps a global symmetry of @xmath231 to a holographic symmetry of @xmath242 .
the generalized ops introduced in the previous section for the @xmath61 higgs model can be mapped to ( dual ) generalized ops for the ( self - dual ) extended toric code model . employing eqs .
( [ map_petc ] ) on the correlator of eq .
( [ pfm ] ) , @xmath247 and @xmath248 where @xmath249 is the ground state of the extended toric code model of eq .
( [ petca ] ) and , as in eq .
( [ pfm ] ) , @xmath250 denotes the ground state of the higgs model . as @xmath231 is self - dual in its gauge - invariant sector
, @xmath242 is indeed also self - dual .
the two string operators of eqs .
( [ ph+],[ph- ] ) transform into each other under this self - duality transformation . _ the quantum xy model on the kagome lattice._
we now examine a frustrated spin system with a continuous symmetry : the quantum xy model on the open kagome lattice [ see fig .
[ kagomes_fig],we will denote this lattice by @xmath251 , @xmath252 the xy rotor at any kagome lattice site @xmath4 has an orientation @xmath253 and @xmath254 .
this xy hamiltonian is invariant under global rotations generated by the total angular momentum @xmath255 .
a duality maps this model to a solid - on - solid like ( @xmath68 ) gauge theory ( similar to the one discussed in the main text ) which now appears on the dice lattice @xmath256 , @xmath257 here , @xmath103 are the oriented links of the dice lattice , @xmath258 its plaquettes , and @xmath259 , with @xmath260 only if @xmath103 viewed as a link in @xmath256 has opposite orientation to that assigned by @xmath261 ( @xmath262 oriented boundary of @xmath258 ) .
similar to the examples that we discussed earlier , the exact dual possesses incomplete plaquette at the boundaries ; this is seen in the lower panel of the figure . under the duality transformation ,
the generator of the global rotational symmetry of the xy model maps onto a holographic symmetry , @xmath263 here , @xmath264 is a degree of freedom placed at the boundary link marked with an open circle .
any other @xmath265 enters the twice sum , with opposite sign .
this is then an example of holography for a continuous symmetry in a highly frustrated system .
symmetric xy model on the kagome lattice is holographically dual to a solid - on - solid - like gauge theory on the dice lattice.,title="fig : " ] + symmetric xy model on the kagome lattice is holographically dual to a solid - on - solid - like gauge theory on the dice lattice.,title="fig : " ] | we introduce a universally applicable method , based on the bond - algebraic theory of dualities , to search for generalized order parameters in disparate systems including non - landau systems with topological order . a key notion that we advance
is that of _ holographic symmetry_. it reflects situations wherein global symmetries become , under a duality mapping , symmetries that act solely on the system s boundary .
holographic symmetries are naturally related to edge modes and localization .
the utility of our approach is illustrated by systematically deriving generalized order parameters for pure and matter - coupled abelian gauge theories , and for some models of topological matter . _
introduction._ landau s concept of an order parameter ( op ) and spontaneous symmetry breaking are central in physics @xcite . in systems with
long - range landau orders , two - point correlation functions of an op field @xmath0 , in their large distance limit , tend to a finite ( i.e. , non - zero ) value , @xmath1 , where @xmath2 is the linear size of the @xmath3-dimensional system , and @xmath0 is _ local _ in the ( spatial ) variable @xmath4 .
it is in landau s spirit to use the op as a macroscopic variable characterizing the ordered phase _ and _ as an indicator of a possible phase transition ( classical or quantum ) to a disordered state where the op becomes zero .
there is much experience , including systematic methods @xcite , for deriving landau ops and their effective field theories @xcite .
landau s ideas of a ( _ local _ ) op can not be extended to topological states of matter as , by definition @xcite , these lie beyond landau s paradigm . however , the notion of long - range order , or the design of a _ witness correlator _
( i.e. , a correlator discerning the existence of various phases and related transitions ) , can be extended to topological phases - phases that can only be meaningfully examined by _ non - local probes _ @xcite .
topological orders appear in gauge theories , quantum hall and spin liquid states ( when defined as deconfined phases of emergent gauge theories @xcite ) , including well - studied exactly solvable models @xcite . in this paper
we demonstrate that generalized _ non - local _ ops may diagnose topological phases of matter .
most importantly , we outline a method based on bond - algebraic duality mappings to search systematically for generalized ops .
dualities have the striking capability of mapping landau to topological orders and viceversa for essentially two reasons .
first , dualities in general represent non - local transformations of elementary degrees of freedom @xcite and may even perform transmutation of statistics @xcite .
second , bond - algebra techniques @xcite allow for the generation of dualities in finite and infinite size systems .
as we will show , in systems with a boundary , dualities realize a form of _ holography _ @xcite capable of transforming a _
global _ symmetry that may drive spontaneous symmetry breakdown into a _ boundary _ symmetry .
we term these distinguished boundary symmetries _
they are under suitable further conditions connected to _
edge ( boundary ) states_. to illustrate the method , we derive explicitly a ( non - local ) witness correlator and a generalized op , suited to diagnose the transition between deconfined and confined phases of matter - coupled gauge theories , undetectable by standard ops or wilson loops .
other examples are reported in ref @xcite . _
the search for generalized order parameters._ a natural mathematical _ language _ to describe a physical system is that for which the system s degrees of freedom couple _ locally_. this simple observation is key to understanding that topological order is a property of a state(s ) relative to the algebra of observables ( defining the language ) used to probe the system experimentally @xcite . in the language in which the system is topologically ordered , it is also robust ( at zero temperature @xcite ) against perturbations local in that language .
spectral properties are invariant under unitary transformations of the local hamiltonian @xmath5 governing the system , @xmath6 . if @xmath7 corresponds to a sensible local theory then the unitary transformation @xmath8 establishes a _ duality _ @xcite .
a duality may map a system that displays topological order to one that does not @xcite .
dualities for several of kitaev s models @xcite epitomize this idea @xcite .
since dualities are unitary transformations ( or , more generally , partial isometries ) @xcite they can not in general change a phase diagram , only its interpretation .
this leads to a central point of our work : _ a duality mapping a landau to a topologically ordered system must map the landau op to a generalized op characterizing the topological order . _
our method for searching for generalized ops , combines this observation with the advantages of the bond - algebraic theory of dualities @xcite . in this framework ,
dualities in arbitrary size ( finite or infinite ) systems can be systematically searched for as alternative local representations of _ bond algebras _ of interactions associated to a hamiltonian @xmath5 .
hence it is possible for _ any _ system possessing topological order to systematically search for a duality mapping it to a landau order .
when a dual landau theory is found , the dual system s op can be mapped back to obtain a generalized op for the topologically ordered system . in what follows and in ref .
@xcite , we study various quantum gauge and topologically ordered theories , and their duals , to illustrate our ideas . _ holographic symmetries and edge states : the gauged kitaev wire._ we next illustrate the concept of holographic symmetry and its relation to generalized ops and edge modes
. consider the kitaev wire hamiltonian @xcite with open boundary conditions , here generalized to include a @xmath9 gauge field ( termed the _ gauged kitaev wire _ ) , @xmath10,\ ] ] where @xmath11 , @xmath12 denote two majorana fermions ( @xmath13 , @xmath14 ) placed on each site of an open chain with @xmath2 sites .
the pauli matrices @xmath15 , placed on the links @xmath16 connecting sites @xmath17 represent a @xmath9 gauge field . for the gauged kitaev wire , fermionic parity
is obtained as the product of the local ( gauge ) @xmath9 symmetries @xmath18 , @xmath19 , and @xmath20 ( @xmath21 ) . just like the standard kitaev wire ,
@xmath22 $ ] has two free edge modes @xmath23 and @xmath24 .
the gauged kitaev wire holds two important dualities .
it is dual to the one - dimensional @xmath9 higgs model @xcite @xmath25,\ ] ] with pauli matrices @xmath26 placed on sites @xmath27 .
moreover , the gauge - reducing @xcite duality mapping @xmath28 @xmath29 transforms @xmath30 into a spin-1/2 system @xmath31 .
\label{hdh}\ ] ] defined on @xmath32 sites .
the fermionic parity @xmath33 maps to a holographic symmetry under this duality , since @xmath34 , i.e. , the product of two ( commuting ) boundary symmetries .
holography is a relational phenomenon ( see @xcite ) .
a duality that uncovers a holographic symmetry links a global ( higher - dimensional ) symmetry of a system to a boundary ( lower - dimensional ) symmetry of its dual .
boundary symmetries need not in general be duals of global symmetries . what is the physical consequence of having an holographic symmetry ?
consider the not uncommon situation in which the holographic symmetry is supplemented by an additional ( non - commuting ) boundary symmetry _ in some region of the phase diagram_. by definition , holographic symmetries are boundary symmetries which are dual to global symmetries .
thus , global symmetries linking degenerate states ( and properties in the broken symmetry phase ) in the dual system have imprints in their holographic counterparts .
then , the many - body level degeneracy of the ground state may be ascribed to boundary effects .
if the couplings are now changed , the ground state degeneracy may get removed , together with some boundary symmetries .
however , so long as the system remains in a topological phase dual to the ( broken symmetry ) ordered phase , the low energy state splitting will be _ exponentially small _ in the system size , so that in the thermodynamic limit ground state degeneracy is restored . the language providing the most local operator description of the ground - state manifold is the one realizing the edge modes , which are expected to be exponentially localized to the boundary .
thus , as long as the thermodynamic - limit degeneracy remains , a suitable local probe will detect localization on the boundary for those states .
conversely , non - commuting edge mode operators in a gapped phase reflect the existence of low - energy many - body states with energy splittings vanishing exponentially in the system size . many - body ( _ zero - energy _ ) edge states are thus simply a natural consequence of a degenerate ground state manifold in a gapped system .
they are witnesses of an ordered ( degenerate ) phase described in a most local language .
note that boundary operators that commute with the hamiltonian at special values of the coupling(s ) are a necessary but not sufficient condition to realize _ exact _ ( zero - energy ) edge modes .
the duality @xmath35 maps a global symmetry of @xmath36 $ ] to a boundary symmetry of @xmath37 $ ] , i.e. , @xmath38 , and one boundary symmetry to another , @xmath39 .
if we now turn on @xmath40 , keeping @xmath41 , the edge mode operators @xmath42 , evolve respectively into @xmath43 and @xmath44 .
these modes are exponentially localized as long as the system is in the ordered gapped phase within a gauge sector @xcite .
the majorana language affords a _ local _ boundary description of these ( partly non - local in the higgs language ) zero - energy modes . for @xmath45 , and/or @xmath46 , the ground state is unique , even in the thermodynamic limit , as we learn from the phase diagram of the one - dimensional higgs model @xcite .
hence the zero - energy modes disappear together with the ground - state degeneracy . for @xmath47 , they disappear despite the fact that fermionic parity remains an exact symmetry and can not be spontaneously broken @xcite .
consider now @xmath48 of eq . .
at @xmath49 , it has zero - energy edge mode operators @xmath50 . for @xmath51 , and @xmath41 , two of these remain unchanged , and the other two evolve into @xmath52 and @xmath53 .
these behave just as their majorana relatives , yet they are recognized as non - local .
the majorana language distinguishes itself as the _ most local one _ for zero modes . to obtain a generalized op for the gauged kitaev wire ,
notice that @xmath54 $ ] reduces to the transverse - field ising ( ti ) chain .
hence it exhibits a second - order phase transition at @xmath55 . for @xmath54
$ ] , this transition is witnessed by the landau op correlator @xmath56 .
( from now on @xmath57 represents the ground state of @xmath58 ) .
our duality maps this correlator back to a generalized op for the gauged kitaev wire , the string correlator @xmath59 .
_ generalized ops in higher - dimensional theories _ we next show how to systematically _ derive _ generalized ops in higher space dimensions .
our main goal is to illustrate the methodology in the challenging case of the abelian ( @xmath60 ) matter - coupled gauge ( higgs ) theory .
previous works @xcite _ conjectured _ generalized ops for matter - coupled gauge theories and were numerically implemented , for instance , in ref .
@xcite .
unfortunately , a systematic mathematical _ derivation _ was missing and this is what our work is about .
our ( non - local ) witness correlator for the higgs model turns out to be the one conjectured in ref .
@xcite . in ref .
@xcite , we study several other examples ( displaying also holographic symmetries ) , including ising and @xmath61 gauge and higgs theories , the @xmath61 extended toric code @xcite as an interesting example of topological order , and the xy model on the frustrated kagome lattice .
non - abelian extensions of our ideas based on ref .
@xcite are currently under investigation .
gauge theory exactly dual to the quantum xy model must satisfy special boundary conditions and possesses a boundary symmetry .
the lattice corresponding to the xy model is shown in thick lines , for @xmath62 . ] to derive the generalized op for the abelian higgs theory , our starting point is the xy model defined in terms of continuous @xmath60 degrees of freedom @xmath63 placed at sites @xmath64 of a square lattice .
the model s hamiltonian reads @xmath65 , \nonumber\end{aligned}\ ] ] with @xmath66 , and @xmath67 .
the xy model is dual to a @xmath68 ( solid - on - solid like ) gauge theory also defined on a square lattice , but with degrees of freedom @xmath69 and @xmath70 associated to links ( @xmath71 . ( in matter - coupled gauge theories we will also have operators acting on sites @xmath4 . )
these operators satisfy @xmath72 , @xmath73 , @xmath74 , with @xmath75 , and commute on different links ( and/or sites )
. then , the _ exact _ dual of @xmath76 for _ open boundary conditions _ reads @xmath77.\nonumber\end{aligned}\ ] ] we will call _ system indices _ the link indices @xmath78 labeling @xmath70 operators that explicitly appear in @xmath79 , and _ extra indices _ the remaining link indices . in the bulk , the plaquette operator @xmath80 reads @xmath81 on the lattice boundary , the plaquette operators are set by two rules : ( i ) @xmath82 thus , @xmath83 involves one degree of freedom @xmath84 labelled by an extra link index .
( ii ) the remaining boundary plaquettes are determined by eq . .
with these definitions in tow , the mapping of bonds @xmath85 implements the duality transformation @xmath86 . as the operators @xmath87 _ do not appear _ in @xmath79 , the operator @xmath84 constitutes a _ boundary symmetry _ of @xmath79 .
similar to the duality between the one - dimensional theories of eqs .
( [ hh2 ] ) and ( [ hdh ] ) , this is a _ gauge - reducing _ duality .
the gauge symmetries of @xmath79 , given by @xmath88 , @xmath89 , are removed by the mapping since @xmath90 . in the thermodynamic ( @xmath91 )
limit , the strongly - coupled ( @xmath92 ) phase of the xy model displays spontaneous symmetry breakdown of its global @xmath60 symmetry with generator @xmath93 , as evinced by a non - vanishing @xmath94 in the limit @xmath95 . by virtue of being dual to the xy system , the gauge theory displays a non - analyticity in its ground state energy as @xmath96 is varied and its symmetry is broken .
however , the phase transition in the gauge theory can not be characterized by a local op .
so , how can the duality connecting the two models bridge the drastic gap separating the physical interpretation of their common phase diagram ?
the answer lies in our notion of holography , since @xmath97 thus , the global symmetry of the xy model is holographically dual to the ( local ) boundary symmetry @xmath98 of its dual gauge theory and can not be spontaneously broken in this dual theory @xcite .
this is how holographic symmetries explain the non - landau nature of critical transitions in the @xmath68 gauge theory .
there are no edge modes nor localization associated with this holographic symmetry as the ordered phase of the xy model is gapless . we now derive a generalized op for the @xmath68 gauge theory .
let @xmath99 be an _ oriented _ path from @xmath4 to @xmath100 made of directed links @xmath101 , and we adopt the convention that @xmath102 if @xmath103 points from @xmath4 to @xmath104 , or @xmath105 if @xmath103 points
oppositely from @xmath104 to @xmath4 .
then @xmath106 . also let @xmath107 denote the set of links @xmath108 such that @xmath109 ( @xmath107 need not be continuous , see fig .
[ t_path_fig ] ) .
then @xmath110 and so _ the string correlator on the left - hand side is a generalized op for the @xmath68 gauge theory , displaying long - range order in the ordered phase_. on a _ closed _ path , @xmath111 reduces to a product of gauge symmetries . and @xmath99 .
] finally , we couple the @xmath68 gauge theory to a @xmath68 matter field ( defined on sites @xmath4 ) , @xmath112 , with @xmath113,\ ] ] and @xmath114 .
the resulting matter - coupled theory @xmath115 is dual to the abelian higgs model @xcite with hamiltonian @xmath116.\end{aligned}\ ] ] here @xmath117 includes a coupling with integer charge @xmath118 to the @xmath60 gauge field @xmath119 , @xmath120 , and @xmath121 .
the correspondence between the two models , established by the mapping of bonds @xmath122 which holds only on physical gauge - invariant states .
the reason is that @xmath123 preserves all commutation relations while `` trivializing '' all gauge symmetries .
more precisely , @xmath115 s gauge symmetries @xmath124 map to @xmath125 , while @xmath126 s gauge generators @xmath127 map to @xmath128 as follows from eqs .
( @xmath129 is the mapping obtained from eqs . by reversing all the arrows ) . if the @xmath68 matter field is weakly coupled to the @xmath68 gauge field , the string correlator of eq .
will still change analytic behavior across transitions . then from eqs .
@xmath130 we obtain a witness correlator for the abelian higgs model that reduces to a wilson loop on closed contours ( @xmath131 ) ( here @xmath132 if a link @xmath103 points from @xmath4 to @xmath104 and @xmath133 otherwise ) .
this non - local correlator is directly related to intuitively motivated generalized ops like @xmath134 conjectured in earlier work @xcite ( @xmath135 denotes a closed loop roughly twice as long as @xmath99 and containing it ) . _ outlook._ as demonstrated , holographic symmetries and generalized ops appear in numerous systems once boundary conditions are properly accounted for in the framework of bond - algebraic dualities . by providing a systematic methodology and many examples , our results might bring the theory of generalized ops and topological orders to a new level of development closer to that of landau s theory .
more key problems need to be tackled .
first , the sufficient conditions under which a given topological order may be mapped to a landau order and _ viceversa _ should be understood .
second , the problem of associating effective field theories to generalized ops should be studied systematically . _
acknowledgements .
_ this work was supported by the dutch science foundation nwo / fom and an erc advanced investigator grant , and , in part , under grants no .
nsf phy11 - 25915 and cmmt 1106293 .
h. nishimori and g. ortiz , _ elements of phase transitions and critical phenomena _ ( oxford university press , oxford , 2011 ) . c. d. batista , g. ortiz , and j. e. gubernatis , phys .
rev .
b * 65 * , 180402(r ) ( 2002 ) .
s. furukawa , g. misguich , and m. oshikawa , phys .
rev .
lett . * 96 * , 047211 ( 2006 ) .
x - g .
wen , int .
j. mod .
phys .
b * 4 * , 239 ( 1990 ) .
z. nussinov and g. ortiz , annals of physics * 324 * , 977 ( 2009 ) ; proceedings of the national academy of sciences * 106 * , 16944 ( 2009 ) .
k. gregor , d. a. huse , r. moessner , and s. l. sondhi , new j.phys .
* 13 * , 025009 ( 2011 ) .
a. yu kitaev , ann . phys . * 303 * , 2 ( 2003 ) .
a. yu kitaev , ann . phys . * 321 * , 2 ( 2006 ) .
f. j. wegner , j. math .
phys .
* 12 * , 2259 ( 1971 ) .
e. cobanera , g. ortiz , and z. nussinov , phys .
rev .
lett . * 104 * , 020402 ( 2010 ) ; adv .
phys . * 60 * , 679 ( 2011 ) .
e. cobanera , g. ortiz , and e. knill , arxiv:1206.1367v1 [ cond-mat.stat-mech ] ( 2012 ) .
z. nussinov and g. ortiz , phys .
rev .
b * 79 * , 214440 ( 2009 ) .
z. nussinov , g. ortiz , and e. cobanera , ann . phys .
* 327 * , 2491 ( 2012 ) .
j. bricmont and j. frlich , phys .
lett .
b * 122 * , 73 ( 1983 ) .
k. fredenhagen and m. marcu , phys .
rev .
lett .
* 56 * , 223 ( 1986 ) ; comm .
math .
phys . * 92 * , 81 ( 1983 ) .
z. nussinov and g. ortiz , phys .
rev .
b * 77 * , 064302 ( 2008 ) .
a. yu kitaev , phys .- usp . * 44 * , 131 ( 2001 ) .
supplemental material .
e. fradkin and s. h. shenker , phys .
rev .
d * 19 * , 3682 ( 1979 ) .
m. d. schulz , s. dusuel , r. orus , j. vidal , and k. p. schmidt , new j. phys .
14 , 025005 ( 2012 ) .
s. elitzur , phys .
rev .
d * 12 * , 3978 ( 1975 ) .
i. herbut , _ a modern introduction to critical phenomena _ ( cambridge university press , cambridge , 2010 ) .
a precise estimate follows from @xmath136,\gamma_1]=-2ij(-h / j)^nb_n\sigma^z_{(n-1;1)}\cdots\sigma^z_{(1;1)}$ ] , and @xmath136,\gamma_2]=-2ij(h / j)^na_1\sigma^z_{(1;1)}\cdots\sigma^z_{(n-1;1)}$ ] . |
mass transfer ; diffusion ; dissolution ; precipitation ; particle ; mathematical analysis
many technological applications involve mass transfer with respect to particles . for fundamental understanding in mathematical terms ,
the problem of mass transfer to and from a particle is typically treated as an isolated sphere with time - dependent radius in a continuous medium of infinite extent as a consequence of heat - mass transfer ( cf . * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
when the attention is focused on the dissolution ( or growth by precipitation ) of solid particles in liquids , as especially important in pharmaceutical dosage form development , the mass transfer problem may often be simplified by ignoring the effects of convection and phase - change heating such that the governing equations become linear with the `` quasi - stationary '' treatment . thus , the mathematical problem is tractable for deriving analytical solutions as usually desired for engineering evaluations .
moreover , when the mass transfer is mainly limited by the diffusion process rather than the rate of phase change , as often to be the case in several realistic applications , the solute concentration at the particle - medium interface can be assumed to take a constant value of the so - called solubility .
then the mathematical problem physically describes a diffusion - controlled mass transfer process , with all the boundary conditions given in the form of dirichlet type . despite the efforts of many authors over years
, the mathematical analyses of this relatively simplified diffusion - controlled quasi - stationary mass transfer problem have not been thoroughly satisfactory in terms of completeness and clarity .
basic understanding of the accuracy and validity of some approximation formulas seems to be lacking in the literature .
the purpose here is to first present a consolidated mathematical formulation of the spherically symmetric mass - transfer problem , then to derive the quasi - stationary approximating equations mainly based on a perturbation procedure for the leading - order effect , and to provide a complete set of exact analytical solutions for the entire parameter range . because the exact solutions can only be written in implicit forms , effort in semi - empirical construction of explicit formulas for the particle radius as a function of time is also made for convenience in engineering practice .
the diffusion - controlled mass transfer to and from a spherical solid particle of ( a time dependent ) radius @xmath0 in an incompressible continuous fluid medium with a constant density @xmath1 and a constant diffusion coefficient @xmath2 is governed by @xcite @xmath3 where @xmath4 denotes the mass concentration of the solute ( namely the dissolved solid from the particle ) , @xmath5 the time , and @xmath6 the radial distance from the center of the sphere . in an incompressible fluid with a spherically symmetric flow , the radial velocity is simply @xmath7 where @xmath8 denotes the ( constant ) solid particle density , to satisfy the equation of continuity and to account for the effect of volume change during the solute phase change ( e.g. , * ? ? ?
* ) . at the particle surface ,
the mass balance based on fick s first law of binary diffusion accounting for the bulk flow effect with the solvent flux being ignored @xcite leads to @xmath9_{\breve{r}=\breve{r } } \ , .\ ] ] in a diffusion - controlled process , the typical boundary conditions for @xmath10 are @xmath11 and initial conditions are @xmath12 where @xmath13 denotes the solubility ( or ` saturated mass concentration ' ) of the solute in the fluid medium is treated as a constant , implying that the particle size effect on solubility as may be observed for submicron particles ( often due to significant surface energy influence ) , is ignored for theoretical simplicity ] and @xmath14 the initial uniform solute concentration . if we consider @xmath15 as a dimensionless variable @xmath16 , measure length in units of @xmath17 and time in units of @xmath18 ,
the governing equations ( [ diffusion_eq])-([ic_eq ] ) can be written in a nondimensionalized form @xmath19 where @xmath20 , @xmath21 , @xmath22 , and @xmath23 $ ] . in genera
equationsl ( [ gov_eq])-(9 ) describe a free - boundary ( or moving - boundary ) nonlinear problem , intractable to exact analytical solutions .
but if @xmath24 can be regarded as a small parameter ( e.g. , @xmath25 ) , which is generally valid for dilute solutions with materials of low solubility when @xmath13 , @xmath26 and @xmath13 @xmath27 , the solutions to those nonlinear equations may be successively approximated by solutions of linear equations following a perturbation procedure .
usually with perturbation approximation , the leading - order solution describes the most significant part of the phenomenon under study . hence attention here
is restricted only to the leading - order solutions . for small @xmath24
, usually corresponding to the situation of relatively low solubility , ( 7 ) indicates that the time variation of @xmath28 is slow comparing to that of @xmath29 .
thus , the convection term in the convection - diffusion equation ( 6 ) can be neglected when considering the leading - order effect . in terms of perturbation solutions for small @xmath24
, @xmath29 can be written in the typical expansion form as @xmath30 where the zeroth - order solution @xmath31 is the base solution at @xmath32 to the zeroth - order equations @xmath33 obviously , at zeroth - order the particle radius @xmath34 is not changing with time ; it becomes a diffusion problem on a fixed domain @xmath35 .
the solution of @xmath31 to ( 11 ) , which can also be written as @xmath36 is given by @xmath37 noteworthy here is that treating @xmath34 as a time - independent `` variable '' in the zeroth - order diffusion equation is consistent with the so - called `` quasi - stationary '' approximation often used in engineering practice ( where the convection transport is neglected and the diffusion equation is solved with the particle surface being considered stationary ) . here
we ignore the initial condition @xmath38 and consider @xmath34 as a variable yet to be determined .
to evaluate the dissolution process of a particle , we need to consider the first - order effect of @xmath24 at least for @xmath28 in ( 7 ) where only the zeroth - order @xmath31 given by ( [ zero - order - c ] ) is involved , i.e. , @xmath39 as pointed out by @xcite , @xcite , and more recently @xcite ) obtained by @xcite was not noticed by @xcite and @xcite .
however , @xcite used their `` highly nonlinear '' implicit formula merely to iteratively determine the value of diffusion coefficient , while @xcite solved the same equation for an analytical solution with the application in drug particle dissolution in mind .
the recent work of @xcite again obtained `` an exact analytical solution '' by solving the same mathematical problem although with a slight change in the form of a parameter to account for the `` bulk flow effect '' as in ( 3 ) ) .
] , ( [ first - order - r_eq ] ) can be rearranged with some variable substitution to have a form of homogeneous ordinary differential equation as @xmath40 where @xmath41 , @xmath42 , @xmath43 , and @xmath44 ( which are mathematically valid for @xmath45 ) .
straightforward integration of ( [ hode ] ) incorporating the initial condition @xmath38 yields @xmath46 } { \epsilon ( 2 - \epsilon)(1+\hat{u}^2 ) } \quad ( \hat{u } \ge \hat{\epsilon } , 0 <
\epsilon < 2 ) \ , .\ ] ] this is a solution that can only be expressed in an implicit form of @xmath28 with @xmath47 , though a cleaner formula than those presented by previous authors @xcite .
the formula given by ( [ hode_solution ] ) suggests an easy way of generating curves for @xmath34 as a function of @xmath48 by first selecting a series of values of @xmath49 ( @xmath50 ) to calculate @xmath48 ( according to ( [ hode_solution ] ) ) , and then from the relationship @xmath51 to calculate the corresponding @xmath28 from the given @xmath49 and known @xmath48 .
furthermore , as shown by previous authors @xcite , the time to complete dissolution @xmath52 when @xmath53 happens at @xmath54 and is thus from ( [ hode_solution ] ) given by @xmath55 } { \epsilon ( 2 - \epsilon)(1+\hat{\epsilon}^2 ) } = \frac{\exp[\hat{\epsilon}(2 \tan^{-1}\hat{\epsilon}-\pi)]}{2 \epsilon } \ , .\ ] ] of mathematical interest , it would be worthwhile to mention that a seemingly different approach to the same mathematical problem was presented by @xcite who were able to obtain the leading - order quasi - stationary solution of @xmath28 in an explicit form . with their approach ,
the moving interface is immobilized by introducing a boundary - fitted coordinate mapping @xmath56 they also converted the equation system ( 6)-(9 ) into one in terms of a compound variable @xmath57 such that they arrived at @xmath58_{\hat{r } = 1 } = \frac{\epsilon}{r(t)}\left[\left(\frac{\partial \hat{c}^{<0>}}{\partial \hat{r}}\right)_{\hat{r}=1}-1\right ] \ , , \\
\hat{c}^{<0>}(1 , t ) = 1 \mbox { and } \hat{c}^{<0>}(\infty , t ) = 0 \ , , \\
\hat{c}^{<0>}(\hat{r } , 0 ) = 0 \mbox { and } r(0 ) = 1 \ , .\end{aligned}\ ] ] similar to ( 15 ) , the solution of @xmath59 for ( 22 ) is then @xmath60 therefore , ( 23 ) yields @xmath61 which is indeed a clean explicit formula . with this
, @xcite could also carry out derivations of subsequent higher - order perturbation solutions , which would be very difficulty , if not impossible , with the implicit formula ( [ hode_solution ] ) .
now the question is why we can have two apparently different solutions ( [ hode_solution ] ) and ( [ new_solution_r ] ) for the same order of approximation . to make sure ( [ hode_solution ] ) and ( [ new_solution_r ] ) are reasonably equivalent mathematical solutions , it is helpful to check the numerical values with each of the formulas . assuming @xmath62 ( then @xmath63 ) and @xmath64 , ( [ hode_solution ] ) yields @xmath65 and then @xmath66 but ( [ new_solution_r ] ) gives @xmath67 . at @xmath68 , ( [ tcd ] ) predicts the time to complete dissolution @xmath69 when @xmath70 , but the time to complete dissolution based on ( [ new_solution_r ] ) would be @xmath71 ^ 2 $ ] @xmath72 .
thus , we see that ( [ new_solution_r ] ) can not be the same as ( [ hode_solution ] ) at least for @xmath62 . a careful examination of the derivation of ( 22 ) , however , reveals an error ( which was somehow not corrected by those authors even in several follow - up publications , e.g. , * ? ? ?
* ; * ? ? ?
the correct expression of ( 22 ) should be @xmath73 therefore , we should have @xmath74 which leads to the same equation as ( [ first - order - r_eq ] ) and solution as ( [ hode_solution ] ) rather than ( [ new_solution_r ] ) .
thus , the same leading - order result can be obtained via seemingly different treatments . for describing the quasi - stationary dissolution process , ( [ hode_solution ] ) should be taken as the correct leading - order solution ( for @xmath45 ) .
it might be noted that so far consideration is only given to the case of @xmath45 , which describes the quasi - stationary dissolution process of a spherical particle .
mathematically , solution also exists for the case of @xmath75 as well as @xmath76 in ( [ first - order - r_eq ] ) . from a physical point of view
, the case of @xmath75 in ( [ first - order - r_eq ] ) describes the inverse process of precipitation growth of a spherical particle , i.e. , when @xmath77 corresponding to the situation of particle growth in a supersaturated solution .
somehow , the exact analytical solution to ( [ first - order - r_eq ] ) for @xmath75 does not seem to have been presented in published literature , unlike the case of @xmath78 .
here it is derived to complete the mathematical solution for ( [ first - order - r_eq ] ) .
with @xmath75 , ( [ hode ] ) must be replaced by @xmath79-(u+\epsilon)^2}{u } \mbox { or } \tau \frac{d\tilde{u}}{d\tau } = \frac{1-\tilde{u}^2}{\tilde{u}+\tilde{\epsilon } } \ , , \ ] ] where @xmath80 , and @xmath81 .
the solution to ( [ hode2 ] ) is then @xmath82 ( \tilde{u}^2 - 1)}\left(\frac{\tilde{u}+1}{\tilde{u}-1}\right)^{\tilde{\epsilon } } \quad ( \tilde{u } > 1 , \epsilon < 0 ) \ , , \ ] ] which appears to be quite different from ( [ hode_solution ] ) = 2 \mbox { coth}^{-1 } \tilde{u}$ ] is used , we can have ( [ hode2_solution ] ) in a similarly looking form to ( [ hode_solution ] ) as @xmath83 ( \tilde{u}^2 - 1 ) } \quad ( \tilde{u } > 1 , \epsilon < 0 ) \,\ ] ] where coth@xmath84 denotes the inverse hyperbolic cotangent function . ] .
as with the dissolution case , curves of @xmath28 can easily be generated by first selecting a series of values of @xmath85 ( @xmath86 ) to calculate @xmath48 from ( [ hode2_solution ] ) , and then from the relation @xmath87 to calculate @xmath28 , for the particle growth case .
although the quasi - stationary diffusion - controlled mass transfer problem can be treated as a leading - order perturbation problem , it is noteworthy that the perturbation procedure is only performed on the convection - diffusion equation ( 6 ) in terms of the small parameter @xmath24 such that the neglection of the convection term can be justified in the zeroth - order equation .
in fact , the convection term may naturally disappear in ( 6 ) when @xmath88 ( which might in fact be a quite reasonable assumption especially for a solid solute particle either dissolving in a liquid solvent or growing by precipitation in a supersaturated liquid solution , because the density of solid solute is usually not much different from that of liquid solution , unlike the situation of a liquid droplet in gas or a gas bubble in liquid with orders of magnitude of density differences ) .
in that case , the perturbation treatment in terms of @xmath24 becomes unnecessary ; the `` leading - order '' solution is the exact solution for any value of @xmath24 .
then , ( [ hode2 ] ) and ( [ hode2_solution ] ) with @xmath89 being replaced by @xmath90 are also valid for the case of @xmath91 but with @xmath92 which becomes zero when @xmath93 , where @xmath94 ( @xmath86 ) and @xmath95 , for @xmath96 .
special attention though should be paid at @xmath97 where @xmath98 which has the ( implicit ) solution @xmath99}{u+2 } \mbox { or } t = \frac{\exp[-4/(u+2)]}{(u+2)^2 } \quad ( 0 \le u < \infty ) \
, , \ ] ] with @xmath100 @xmath101 .
as expected , ( [ hode_solution ] ) indeed approaches ( [ hode3_solution ] ) at the limit as @xmath102 by applying lhspital s rule , and so does ( [ hode2_solution ] ) with @xmath89 being replaced by @xmath90 for @xmath91 by using the relationship @xmath103 .
thus , we have a complete set of solutions to ( [ first - order - r_eq ] ) with ( [ hode_solution ] ) for @xmath45 , ( [ hode2_solution ] ) for @xmath75 and @xmath91 ( with @xmath89 being replaced by @xmath90 ) , and ( [ hode3_solution ] ) for @xmath97 , covering all possible @xmath104 ( whenever ignoring the convection term in ( 6 ) is justifiable in practice ) .
moreover , the case of @xmath32 corresponds to a mathematically trivial solution @xmath105 , if desired to be included for completeness .
due to the awkwardness of practical usage of the implicit solution to ( [ first - order - r_eq ] ) , authors ( e.g. , * ? ? ?
* ; * ? ? ?
* ) who derived the analytical solution ( for @xmath45 ) often also suggested explicit `` approximate solution '' with the quasi - steady - state treatment ( when the time derivative term in the diffusion equation ( 11 ) is also neglected ) .
simple as it may appear , however , the explicit quasi - steady - state solution may be found not to offer a satisfactory approximation to ( [ hode_solution ] ) for @xmath106 ( depending on the intended application ) . to provide much improved approximate explicit formulas for practically accurate evaluation of @xmath28 ,
effort is made here via a semi - empirical approach . in view of the physical meanings , the two terms on the right side of ( [ first - order - r_eq ] ) represent two different aspects of the diffusion process : @xmath107 comes from the spherically symmetric steady - state solution of the laplace equation whereas @xmath108 describes transient diffusion from a planar surface .
initially when @xmath109 , the concentration gradient front has not propagated far enough from the particle surface to bring the curvature effect out yet and thus the effect of transient diffusion from a planar surface dominates . with increasing @xmath48 , at some point @xmath110 is expected to happen especially for @xmath25 ; then the steady - state diffusion term becomes dominant and ( 16 ) can be reduced to @xmath111 which is a so - called `` quasi - steady - state '' approximation commonly seen in the literature ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the quasi - steady - state solution to ( [ qss_ode ] ) is simply @xmath112 which yields noticeable difference from ( [ hode_solution ] ) even at @xmath113 . on the other hand , for @xmath114 we have @xmath115 intuitively then , one might believe that a constructed explicit formula like @xmath116 could provide an improved approximation from the quasi - steady - state result ( [ qss_ode_solution ] ) to the exact solution ( [ hode_solution ] ) . clearly , ( [ approx_solution ] ) approaches @xmath117 as @xmath118 ( where the ` higher - order ' term associated with @xmath119 becomes negligible ) . the approximate time to complete dissolution corresponding to ( [ approx_solution ] ) is @xmath120 for comparison at various values of @xmath24 , the values of time to complete dissolution
are tabulated in table 1 for @xmath52 from the exact ( quasi - stationary ) solution ( [ tcd ] ) , the quasi - steady - state result @xmath121 , and the intuitively constructed model result @xmath122 from ( [ approx_tcd ] ) .
the consistent improvement of ( [ approx_tcd ] ) over @xmath121 from ( [ qss_ode_solution ] ) in comparison with ( [ tcd ] ) is obvious , but not as significant as desired .
both ( [ qss_ode_solution ] ) and ( [ approx_solution ] ) underestimate the change of @xmath34 with @xmath48 .
0.75@ l c c c + @xmath24 & @xmath52 from ( [ tcd ] ) & @xmath121 from ( [ qss_ode_solution ] ) & @xmath122 from ( [ approx_tcd ] ) + 1 & 0.1039 & 0.5 ( @xmath123 ) & 0.1667 ( @xmath124 ) + 0.5 & 0.2984 & 1 ( @xmath125 ) & 0.5 ( @xmath126 ) + 0.1 & 2.6971 & 5 ( @xmath127 ) & 4.1667 ( @xmath128 ) + 0.05 & 6.3622 & 10 ( @xmath129 ) & 9.0909 ( @xmath130 ) + 0.01 & 40.421 & 50 ( @xmath131 ) & 49.020 ( @xmath132 ) + 0.005 & 85.876 & 100 ( @xmath133 ) & 99.010 ( @xmath134 ) + 0.001 & 466.54 & 500 ( @xmath135 ) & 499.00 ( @xmath136 ) + 0.0005 & 952.01 & 1000 ( @xmath137 ) & 999.00 ( @xmath138 ) + 0.0001 & 4890.6 & 5000 ( @xmath139 ) & 4999.0 ( @xmath140 ) + interestingly , ( [ new_solution_r ] ) obtained by @xcite is somehow found to offer a better approximation than ( [ approx_solution ] ) to the exact quasi - stationary solution especially for @xmath141 , although the its derivation has been shown not quite correct in the mathematical sense .
for example , the time to complete dissolution predicted with @xmath71 ^ 2 $ ] based on ( [ new_solution_r ] ) would be @xmath142 , @xmath143 , @xmath144 , and @xmath145 respectively for @xmath146 , @xmath147 , @xmath148 , and @xmath149 .
moreover , ( [ new_solution_r ] ) tends to predict faster dissolution whereas ( [ approx_solution ] ) slower .
this is because ( [ new_solution_r ] ) over estimated the flux term associated with @xmath108 in ( 26 ) by mistakenly replacing @xmath34 ( @xmath150 ) with @xmath151 .
however , the effect of @xmath108 usually only dominates for a short time when @xmath48 is small and @xmath34 is not too far from unity especially when @xmath152 . shown in fig . 1 is a comparison among the exact quasi - stationary solution ( 18 ) , the quasi - steady - state solution ( 35 ) , and the approximate formulas ( 27 ) of @xcite and ( 37 ) .
even at @xmath153 , the deviation of quasi - steady - state solution ( 35 ) from the exact solution is still quite significant due to the unaccounted initial effect from the flux term @xmath108 for small @xmath48 .
the overall improvement of ( 37 ) from the quasi - steady - state solution ( 35 ) is clear , especially for small @xmath48 ( or in general @xmath154 ) where the curve of ( 37 ) consistenly remains close to that of the exact quasi - stationary solution . .
] in view of fig .
1 , an explicit approximation formula may be constructed semi - empirically by combining ( [ new_solution_r ] ) and ( [ approx_solution ] ) as @xmath155 + [ 1-\alpha(\epsilon ) ] \left(\sqrt{1 - 2 \epsilon \ , t } - 2 \epsilon \sqrt{t}\right)^2 } \ , , \ ] ] which with @xmath156 \quad ( 0 < \epsilon
< 0.1 ) \ , \\ 0.0193 ( \log_{10}\epsilon)^2 - 0.2703 \log_{10}\epsilon + 0.095 \quad ( 0.1 \le \epsilon \le 0.5 ) \ , \end{array } \right . \ , \nonumber\end{aligned}\ ] ] can consistently produce the value of @xmath28 very close to that of the exact solution ( [ hode_solution ] ) for any @xmath157 in the entire range of @xmath158 . by the similar token , accurate formulas of @xmath159 for @xmath160
can also be constructed but is not attempted here because most practical situations , e.g. , in pharmaceutical dosage form development @xcite , typically concern with @xmath161 . for particle growth in a supersaturated environment , i.e. , @xmath77 as in the case of phase separation in a drying coating , the exact quasi - stationary solution is given by ( [ hode2_solution ] ) with @xmath75 .
2 for particle growth shows that comparing to ( [ hode2_solution ] ) , ( [ approx_solution ] ) seems to overestimate the particle growth whereas ( [ new_solution_r ] ) of @xcite and the quasi - steady - state solution ( [ qss_ode_solution ] ) both underestimate it . .
] based on this observation , a fairly accurate explicit approximate formula to the exact quasi - stationary solution ( [ hode2_solution ] ) may be semi - empirically constructed with the same form as ( [ new_approx_solution ] ) but having @xmath162 \quad ( -0.5 \le \epsilon < 0 ) \ , \ , , \ ] ] for evaluating the situation of particle growth .
starting from a consolidated mathematical formulation of the spherically symmetric mass - transfer problem , the quasi - stationary approximating equations can be derived based on a perturbation procedure for the leading - order effect . for the diffusion - controlled quasi - stationary process , a mathematically complete set of the exact analytical solutions
is obtained in implicit forms for consideration of both dissolution of a particle in a solvent and growth of it by precipitation in a supersaturated environment .
understanding the dissolution behavior of solid particles in liquid plays an important role in pharmaceutical dosage form development @xcite , and particle growth by precipitation in a supersaturated environment is relevant to the drug - polymer microsphere formation process @xcite as well as the observed phase separation process during solvent removal in drying of a coating with the drug - polymer mixture @xcite .
the commonly used explicit formula based on the solution with quasi - steady - state approximation is shown to provide unsatisfactory accuracy unless the solubility is restricted to very small values ( corresponding to @xmath163 ) .
therefore , accurate explicit formulas for the particle radius as a function of time are also constructed semi - empirically to extend the applicable range at least to @xmath164 for practical convenience .
the author is indebted to professor richard laugesen of the university of illinois for his helpful discussions and skillful illustration of mathematical manipulations .
the author also wants to thank yen - lane chen , scott fisher , ismail guler , cory hitzman , steve kangas , travis schauer , and maggie zeng of bsc for their consistent support .
barocas , v. , drasler ii , w. , girton , t. , guler , i. , knapp , d. , moeller , j. , parsonage , e. , 2009 . a dissolution - diffusion model for the taxus drug - eluting stent with surface burst estimated from continuum percolation .
_ j. biomed .
b : appl . biomater . _ 90(1 ) , 267 - 274 .
richard , r. , schwarz , m. , chan , k. , teigen , n. , boden , m. , 2009 .
controlled delivery of paclitaxel from stent coatings using novel styrene maleic anhydride copolymer formulations .
_ j. biomed .
res . a _ 90(2 ) , 522 - 532 .
wu , x .- s .
preparation , characterization , and drug delivery applications of microspheres based on biodegradable lactic / glycolic acid polymers .
in wise , d. l. , trantolo , d. j. , altobelli , d. e. , yaszemski , m. j. , gresser , j. d. , schwartz , e. r. ( eds ) , _ encyclopedic handbook of biomaterials and bioengineering part a : materials . _ marcel dekker , new york , vol .
1151 - 1200 . | a consolidated mathematical formulation of the spherically symmetric mass - transfer problem is presented , with the quasi - stationary approximating equations derived from a perturbation point of view for the leading - order effect . for the diffusion - controlled quasi - stationary process , a mathematically complete set of the exact analytical solutions
is obtained in implicit forms to cover the entire parameter range .
furthermore , accurate explicit formulas for the particle radius as a function of time are also constructed semi - empirically for convenience in engineering practice .
both dissolution of a particle in a solvent and growth of it by precipitation in a supersaturated environment are considered in the present work . |
turbulence is one of the most fascinating phenomena in nature and one of the biggest challenges for modern physics .
it is common knowledge that a flow of a simple , low molecular , newtonian fluid is likely to be turbulent , when velocity , @xmath2 , is high , kinematic viscosity of the fluid , @xmath3 , is low , and size of the tank , @xmath4 , is large .
all these three factors contribute to high value of the reynolds number , @xmath5 .
motion of the newtonian fluids is governed by the navier - stokes equation @xcite @xmath6 where @xmath7 is pressure and @xmath8 is density of the fluid . the equation has a non - linear term , @xmath9 , which is inertial in its nature .
the reynolds number defines the ratio of this non - linear term to the viscous dissipative term , @xmath10 .
when @xmath11 is high , the non - linear effects are strong , and the flow is likely to be turbulent .
therefore , turbulence in fluids at high @xmath12 is a paradigm for a strongly non - linear phenomenon in spatially extended systems@xcite .
solutions of flexible high molecular weight polymers are visco - elastic liquids , and they differ from newtonian fluids in many aspects @xcite .
the most striking elastic property of the polymer solutions is , probably , dependence of mechanical stresses in flow on history of the flow .
so , the stresses do not immediately become zero when fluid motion stops , but rather decay with some characteristic relaxation time , @xmath13 , which can be well above a second .
when a polymer solution is sufficiently dilute , its stress tensor , @xmath14 , can be divided into two parts , @xmath15 .
the first part , @xmath16 , is defined by viscosity of the newtonian solvent , @xmath17 , and rate of strain in the flow , @xmath18 $ ] .
so , the equation of motion for a dilute polymer solution becomes @xmath19 here the elastic stress tensor , @xmath20 , is due to the polymer molecules , which are stretched in the flow , and it depends on history of the flow .
one can see that @xmath21 enters the equation of motion linearly .
so the degree of non - linearity of the equation of motion can still be defined by the reynolds number , @xmath22 .
the simplest model incorporating the elastic nature of the polymer stress tensor , @xmath23 , is a maxwell type constitutive equation @xcite with a single relaxation time , @xmath13 , @xmath24.\ ] ] here @xmath25 is a material time derivative of the polymer stress .
an appropriate expression for the time derivative @xmath26 has to take into account that the stress is carried by fluid elements , which move , rotate and deform in the flow .
the translational motion implies an advection term @xmath27 in an appropriate expression for @xmath28 , while the rotation and deformation of the fluid particles should lead to contributions like @xmath29 or @xmath30 @xcite .
therefore , along with terms linear in @xmath31 and @xmath32 , some non - linear terms , in which @xmath23 is coupled to @xmath32 , should appear in the constitutive relation .
a simple model equation for @xmath33 , which is commonly used for description of dilute polymer solutions , is the upper convected time derivative , @xmath34 the equations ( 3,4 ) together with the expression for @xmath35 constitute the oldroyd - b model of polymer solution rheology @xcite .
one can see that non - linear terms in the constitutive equation ( eqs .
3,4 ) are all of the order @xmath36 .
the ratio of those non - linear terms to the linear relaxation term , @xmath20 , is given by a dimensionless expression @xmath37 , which is usually called weissenberg number , @xmath38 .
( the relaxation term @xmath20 is somewhat analogous to the dissipation term in the navier - stokes equation . )
one can expect mechanical properties of the polymer solutions to become notably non - linear at sufficiently large weissenberg numbers .
indeed , quite a few effects originating from the non - linear polymer stresses have been known for a long time @xcite .
so , in a simple shear flow of a polymer solution there is a difference between normal stresses along the direction of the flow and along the direction of velocity gradient . at low shear rates
this normal stress difference , @xmath39 , is proportional to the shear rate squared .
when flow lines are curvilinear , it gives rise to a volume force acting on the liquid in the direction of the curvature , the hoop stress .
therefore , if a rotating rod is inserted in an open vessel with a highly elastic polymer solution , the liquid starts to climb up on the rod , instead of being pushed outwards @xcite .
this phenomenon is known as rod climbing , or
weissenberg effect .
further , in an purely extensional flow resistance of a polymer solution depends on rate of extension in a strongly non - linear fashion .
there is a sharp growth in the elastic stresses , when the rate of extension exceeds @xmath40 , that is at @xmath41 . as a result apparent viscosity of a dilute polymer solution can increase by up to three orders of magnitude @xcite .
both the weissenberg effect and the growth of the extensional flow resistance have been most clearly observed in very viscous polymer solutions and in flows with quite low @xmath12 , when non - linear inertial effects were insignificant .
a natural question arising here is , whether there may exist some kind of turbulent flow produced by the non - linear polymer stresses alone , in the absence of any significant inertial effects , at low @xmath12 .
an important step in this direction was made about a decade and a half ago when purely elastic instabilities were experimentally identified in a rotational flow between two plates @xcite and in the couette - taylor ( ct ) flow between two cylinders @xcite .
the experiments were carried out with a boger fluid @xcite , a dilute solution of high molecular weight polymers in a viscous newtonian solvent .
the boger fluids are almost universally used as model viscoelastic fluids .
their relaxation times can be quite large , reaching seconds or even minutes , while their rheological properties are semi - quantitatively described by the simple oldroyd - b model .
the purely elastic instabilities occurred at @xmath38 of order unity and vanishingly small @xmath11 .
as a result of the instabilities , secondary vortex flows developed @xcite , and flow resistance somewhat increased @xcite .
the problem of linear stability of the ct flow was also treated theoretically using the oldroyd - b model@xcite .
the analysis showed that the non - linear mechanical properties of the polymer solution can indeed lead to a flow instability , and a simple mechanism of this purely elastic instability was proposed . during the past decade
the purely elastic instabilities in the boger fluids have been a subject of many theoretical and experimental studies , which are partially reviewed in ref.@xcite . after the pioneering work by larson , muller and shaqfeh @xcite ,
purely elastic instabilities were also found in other shear flows with curvilinear streamlines . those included the flow between a rotating cone and a plate and the taylor - dean flow@xcite .
the original theoretical analysis of ref .
@xcite was refined , and more elaborate experiments were carried out .
a few new mechanisms of flow instability driven by non - linear elastic stresses were suggested for cone - and - plate and for taylor - dean flows . the original mechanism proposed in ref .
@xcite was verified experimentally in ref.@xcite .
some flow patterns observed above the purely elastic instability threshold in the curvilinear flows had rather disordered appearance and exhibited chaotic spectra of velocities .
so , it was reasonable to suggest that at some conditions a truly turbulent flow might be excited by elastic stresses at vanishingly small @xmath12 .
this idea was explicitly stated in ref.@xcite , where analogy between elastic and inertial flow transitions was discussed .
in fact , irregular flow patterns and growth of flow resistance in elastic polymer solutions at low @xmath12 were observed even before the purely elastic instabilities were identified @xcite .
those flow phenomena were even sometimes coined elastic turbulence . in all those cases , however , the term turbulence was used in a rather loose fashion , without being given a proper definition .
more important , no quantitative data on either flow velocity field or spatial and temporal velocity spectra in those irregular flows were ever presented .
although the notion of turbulence is widely used in scientific and technical literature , there is no unique commonly accepted definition of it .
therefore , turbulent flow is usually identified by its main features .
turbulence implies fluid motion in a broad range of temporal and spatial scales , so that many degrees of freedom are excited in the system .
there are no characteristic scales of time and space in the flow , except for those restricting the excited temporal and spatial domains from above and below .
turbulent flow is also usually accompanied by a significant increase in momentum and mass transfer .
that is , flow resistance and rate of mixing in a turbulent flow become much higher than they would be in an imaginary laminar flow with the same @xmath12 . in our recent publication
@xcite we showed , how the first three of these features of turbulence appeared in a flow of a highly elastic polymer solution at low reynolds numbers .
the experiments were done in a swirling flow between two plates with a wide gap , and the phenomenon was called elastic turbulence . in the present manuscript we give a more complete account of our experiments on the elasticity driven turbulent flow .
we show additional velocity spectra , and distributions of probabilities of flow velocities and their derivatives .
we also show that the elastic turbulence leads to quite efficient mixing in the flow .
further , we present results of our experiments on couette - taylor ( ct ) flow and on an open flow in a curvilinear channel ( dean flow ) , where we also observed the elasticity driven turbulence .
the paper is organized as follows . in the next section
we discuss some practical problems concerning the experiment and the polymer solution . in sections
iii , iv and v the results on the swirling flow between two plates , the ct flow and the flow in the curvilinear channel , respectively , are presented . the results are briefly summarized and discussed in section vi .
as it was suggested in ref .
@xcite , where the ct flow was discussed , there is some analogy between flow transitions driven by elasticity and inertia .
so , the inertially driven taylor instability occurs at constant taylor number @xcite , @xmath42 , while the elastic instability is controlled by a parameter @xmath43 @xcite . here
@xmath44 is radius of the inner cylinder , @xmath45 is width of the gap between the cylinders in the ct set - up , @xmath46 is the total solution viscosity , and @xmath47 is polymer contribution to the viscosity , @xmath48 . the weissenberg number is defined here as @xmath49 , where @xmath50 is angular velocity of the rotating inner cylinder .
( it was termed as the deborah number in some of the original texts @xcite . ) the weissenberg number , which reflects non - linearity of the constitutive equation ( eqs .
3,4 ) , appears to be analogous to the reynolds number , which expresses non - linearity of the equation of motion .
the geometric parameter determining curvature , the gap ratio , @xmath51 , enters the expressions for both @xmath52 and @xmath53 .
( there is also a specific term @xmath54 , though , which shows polymer contribution to the solution viscosity and naturally appears only in the expression for the parameter @xmath53 for the elastic instability ) .
scales of time and velocity for the purely elastic flow transition are given by @xmath13 and @xmath55 , respectively . as it was shown in ref .
@xcite they are analogous to @xmath56 and @xmath57 , which define scales of time and velocity for the inertially driven flow transitions . here
@xmath56 is the viscous diffusion time defined as @xmath58 . nevertheless , along with all those analogies
, there are still some important differences between flow transitions driven by inertia and elasticity .
so , it is an experimental fact that _ any _ laminar flow of a newtonian fluid becomes unstable at sufficiently high @xmath12 , and all high reynolds number flows are turbulent .
that includes rectilinear shear flows , such as poiseille flow in a circular pipe and plane couette flow , which are supposed to be linearly stable at any @xmath12 .
in contrast to it , purely elastic flow instabilities in shear flows have only been observed so far in systems with curvilinear stream lines .
all these instabilities are supposed to be driven by the hoop stress , which originates from the normal stress differences .
elastic instabilities also occur in flows of polymer solutions with open surfaces and in extensional flows through contractions in channels @xcite .
a particular feature of the latter flows is that the rate of strain changes along the flow lines , so that even the basic flow states are not stationary in the langrangian sense .
instabilities in those flows are not discussed in this paper .
this difference between the inertial and elastic instabilities may originate , of course , from the distinct governing equations .
there are , however , some purely practical factors that can explain rather well the lack of observations of purely elastic flow transitions in rectilinear shear flows . inertial instabilities in rectilinear shear flows of newtonian fluids occur at quite high reynolds numbers
. those are typically about two orders of magnitude higher than @xmath12 at which curvilinear shear flows with large gap ratios , @xmath51 , become unstable .
a priory , one may suggest that rectilinear and curvilinear shear flows would have a similar relation between @xmath38 at thresholds of the purely elastic flow instabilities as well .
the problem is , however , that while it is rather easy to generate high @xmath12 flows with low viscosity newtonian fluids , it is usually impossible to reach the corresponding high values of @xmath38 in shear flows of elastic polymer solutions .
that is , there always exist rather severe practical limitations restricting non - linearity in elastic polymer stresses in shear flows .
their molecular mechanisms have been recently elucidated in a seminal paper by chu s group @xcite .
polymer molecules have finite extensibility , and their relaxation time decreases when they get stretched in a shear flow .
this thinning of the relaxation time at high @xmath38 is usually quite a strong and well recognized effect .
in fact , different variations of the basic oldroyd - b model have been specially developed to take into account the shear thinning @xcite .
significant decrease of @xmath13 with shear rate , @xmath59 , renders growth of the weissenberg number , @xmath60 , much slower than linear in @xmath59 .
substantial stretching of the polymer molecules in the primary shear flow also reduces their ability for further extension and susceptibility of the elastic stresses to flow perturbations , which is necessary for generation of the flow instabilities and secondary vortex flows@xcite . finally , high shear rates cause mechanical degradation of the polymer molecules .
it leads to permanent reduction of elasticity during experimental runs and decay of @xmath13 that can be very fast at high @xmath38 . because of all those reasons
it was found to be very difficult or even impossible to observe elastic instabilities , when expected values of @xmath38 at the instability threshold were high .
it was the case in curvilinear flows with small gap ratios , @xmath51 , @xcite and small viscosity ratios @xmath54 @xcite . therefore , in order to maximize the non - linear elastic effects and to get a better opportunity to observe the elastic turbulence we had to choose experimental conditions quite carefully @xcite .
first , it was important to obtain an elastic instability at a possibly low critical weissenberg number , @xmath61 . for that purpose the gap ratio and the viscosity ratio had to be possibly large .
therefore , we used a polymer solution with a rather large @xmath54 of about 1/4 .
( further increase of the polymer concentration and of @xmath54 was not very efficient , and would also complicate the solution rheology , including large shear thinning of the solution viscosity . ) in order to have a large gap ratio with a possibility to vary it , we carried out our experiments in a swirling flow between two plates .
the ratios @xmath51 which we used were 0.263 and 0.526 .
in fact , historically we first tested a ct flow with @xmath62 , but growth of flow resistance in it was significantly smaller than in the swirling flow set - up , which we studied later ( see sections iii - iv ) .
second , an appropriate polymer sample for the solution had to be chosen , that would not suffer major mechanical degradation under experimental flow conditions .
further , the limit for extensibility of the polymer molecules had to be possibly high compared with their typical conformations at the instability threshold .
we used polyacrylamide ( paam ) with a large average weight @xmath63 and a broad molecular weight distribution ( and low concentration of polyacrylic acid monomers ) supplied by polysciences .
this commercial polymer sample proved to be remarkably stable with respect to mechanical degradation , that allowed us to reach high values of @xmath38 and to explore strongly non - linear flow regimes .
the high molecular weight of the polymers resulted in a large characteristic relaxation time , @xmath13 , even with solvents of moderate viscosity , and in a small characteristic stress , @xmath64 .
one can learn from the eqs .
3,4 that @xmath65 sets a scale of the polymer stress , at which its non - linearity becomes significant .
therefore , the value of @xmath65 together with @xmath66 and @xmath51 determine the polymer stress @xmath67 in the primary shear flow at the instability threshold .
it is rather natural to suggest that when polymer molecules transduce less stress , they are also less subjected to mechanical degradation .
further , molecular interpretation of the weissenberg number in a shear flow is the degree of deformation of polymer molecules from their relaxed random coil conformations .
so , @xmath68 can be regarded as a characteristic value at which extension of polymer molecules becomes considerably larger than the size of relaxed coils . by the same token ,
extension of a polymer molecule at @xmath69 ( the elastic instability threshold ) is supposed to be a fixed multiple of the relaxed coil size ( in the first , linear approximation valid at moderate @xmath69@xcite ) .
high molecular weight and flexibility of a polymer suggests large number of the kuhn segments @xcite in the polymer chain , and a high ratio between its contour length ( size , when fully extended ) and the size ( radius of gyration , @xmath70 ) of a relaxed coil .
( for a polymer molecule in a good solvent , with @xmath71 , this ratio should increase as @xmath72 @xcite .
we would like to point out here that addition of 1% of nacl to the solution significantly reduced @xmath70 . ) therefore , using a higher @xmath73 polymer we increase `` reserve '' of extensibility starting from the typical conformation at @xmath69 .
that opens a way for reacher flow dynamics above the elastic instability threshold .
the experimental apparatus is schematically shown in fig.1 .
polymer solution was held in a stationary cylindrical cup with a flat bottom ( lower plate ) .
a coaxial rotating upper plate was just touching the surface of the fluid .
the cup was mounted on top of a commercial rheometer , ar-1000 of ta - instruments .
the upper plate was attached to the shaft of the rheometer , which allowed precise control of its rotation velocity , @xmath74 , and measurements of the torque , @xmath75 .
the average shear stress at the upper plate , @xmath76 , was calculated using the equation @xmath77 , that gave @xmath78 .
( the integration is over the upper plate surface . )
sidewalls of the cup were machined of a single piece of perspex , which was optically clear .
the cup was circular from the inside and square from the outside in horizontal cross - section
. that allowed measurements of flow velocity in the horizontal plane by a laser doppler velocimeter ( ldv ) with two crossing frequency shifted beams . by appropriate positioning and orientation of the beam crossing region , azimuthal ( longitudinal ) and radial ( spanwise ) velocity components , @xmath79 and @xmath80 , respectively , could be measured at different @xmath81 and @xmath82 . here
@xmath83 are cylindrical coordinates .
the bottom of the cup was machined of stainless steel and the temperature was stabilized at 12 @xmath84c by circulation of water below the bottom plate .
a slightly modified version of the set - up was designed to photograph the flow from below and to observe mixing in the flow . a special cup of the same shape but with a transparent bottom ( lower plate , made of perspex )
was attached to the rheometer concentrically with the shaft but above the rheometer base , and a mirror tilted by 45@xmath84 was placed under the cup , as it is schematically shown in fig.1 .
the mirror was used both to illuminate the fluid by diffuse light and to relay images of the flow to a ccd camera .
the camera was equipped with a regular video lens and mounted horizontally near the rheometer ( fig.1 ) .
the source of the diffuse light was an illuminated white screen around the camera .
the images were digitized by an 8 bit 512x512 frame grabber . in order to provide thermal stabilization , in this case
the whole rheometer was placed in a thermally isolated box with through flow of temperature controlled air . in the basic standard set - up ,
the radii of the upper plate and of the cup were @xmath85 mm and @xmath86 mm , respectively , and the distance between the plates was @xmath87 mm .
the configuration was similar to the devices with rotating flow between two plates used in experiments on purely elastic instability@xcite .
its gap ratio , @xmath51 , was significantly higher , though . in order to study dependence of flow conditions on size of the system , two smaller set - ups , a half - size and a quarter - size , with all the dimensions reduced by factors of two and four , respectively ,
were used .
every time , when dimensions of an experimental set - up are non - standard , it is stated explicitly in the text .
we prepared a stock solution of paam ( @xmath63 by polysciences ) and used it through the whole course of the experiments .
first we dissolved 0.9 g of paam powder and 3 g of nacl in 275 ml of deionized water by gentle shaking .
( addition of nacl was necessary to improve solubility of paam . )
next the solution was mixed for 3 hours in a commercial mixer with a propeller at a moderate speed .
the idea behind that procedure was to cause mechanical degradation paam molecules with the highest weights , and to `` cut '' a high @xmath88 tail of the broad molecular weight distribution of the paam sample . in a solution with a broad distribution of polymer molecular weights the heaviest molecules , which are most vulnerable to mechanical degradation ,
may also make a major contribution to the solution elasticity . a possible negative effect of that is significant degradation of elasticity during experimental runs , and inconsistency of results of the experiments .
we found empirically that the procedure of pre - degradation in the mixer leads to substantial reduction of the degradation during the experiments and to substantial improvement of their consistency@xcite .
finally , 9 g of iso - propanol was added to the solution ( to preserve it from aging ) and water was added up to 300 g. the final concentrations of paam , nacl and iso - propanol in the stock solution were 3000 ppm , 1% and 3% , respectively . as
a viscous newtonian solvent for paam , we used a solution of 65% sugar ( saccharose ) and 1% nacl in water .
the salt was added to fix the ionic contents .
viscosity and relaxation time were measured with the same ar-1000 rheometer in a temperature controlled narrow gap mooney - ewart geometry .
the solvent viscosity was @xmath89 pa@xmath90s at 12 @xmath84c .
the paam concentration was 80 ppm , fig.2 .
solution viscosity , @xmath91 , was carefully measured in a broad range of shear rates , @xmath59 , fig.2 .
viscosity , @xmath46 , was slowly decreasing with @xmath59 , so that its shear thinning was about 7% per a decade of @xmath59 . at a shear rate of @xmath92 s@xmath93 , corresponding to the onset of a purely elastic instability in the standard configuration ( see below ) , @xmath46 was @xmath94 pa@xmath90s , and the viscosity ratio was @xmath95 .
polymer relaxation time , @xmath13 , was measured in oscillatory tests with a shear rate amplitude of 1 s@xmath96 .
components of the apparent viscosity of the solution in phase and out of phase with the applied shear , @xmath97 and @xmath98 , respectively , were measured in long series of oscillations in a range of angular frequencies , @xmath99 .
( individual series were typically repeated about 10 times , and average viscosity values were calculated . ) the same procedure was applied then to the solvent , and its viscosity components , @xmath100 and @xmath101 , were calculated as well .
( the latter was nearly zero and it was an important test of virtual absence of inertial effects and of general applicability of the procedure . ) the values for the polymer in phase and out of phase viscosity were calculated then as , @xmath102 and @xmath103 . the ( frequency dependent ) relaxation time
was calculated as @xmath104 , and @xmath105 at @xmath106 , estimated as 3.4 s , was chosen as a representative relaxation time , @xmath107 , inset in fig.2 .
the diffusion coefficient for the saccharose molecules was about @xmath108 @xmath109/s @xcite . in a swirling flow between two plates shear
rate is quite inhomogeneous over the fluid bulk , even when the flow is laminar .
so , in the simplest case of a narrow gap , @xmath110 , when the shear rate is constant along the @xmath82-direction , it grows along the radius , from zero at @xmath111 to the maximal value of @xmath112 at @xmath113 .
shear rate averaged over the surface of the upper plate ( and over the fluid volume ) is then @xmath114 . in our case ,
when the gap is quite wide ( @xmath115 in the standard configuration ) and the sidewalls are close to the upper plate , the situation is even more complicated .
the shear rate becomes strongly non - homogeneous along @xmath82 in a large region of space corresponding to @xmath81 beyond @xmath116 .
so , measurements of the ratio @xmath117 in a laminar flow of a viscous newtonian liquid gave a value 1.68 times larger , than a value calculated suggesting homogeneous shear rate along @xmath82 .
( the latter suggestion is not realistic for the large gap ratio set - up , of course . )
this discrepancy is obviously due to a high shear rate layer near the upper plate at large @xmath81 .
one can see that choice of a representative shear rate becomes somewhat arbitrary under these circumstances .
we decided to consider the simple expression @xmath112 as a characteristic shear rate , and to define the weissenberg number as @xmath49 .
the reynolds number was defined as @xmath118 . in order to evaluate growth of flow resistance due to elastic instability and irregular secondary flow in the system , the average shear stress near the upper plate , @xmath119 , in an imaginary laminar shear flow at the same @xmath50 had to be estimated .
the stress , @xmath119 , would depend on an average shear rate , @xmath120 , at the upper plate , and on viscosity of the polymer solution , @xmath46 , at this shear rate .
an appropriate expression for it is @xmath121 .
the average shear rate , @xmath122 , was estimated from measurements of the ratio @xmath123 in laminar flow of a newtonian fluid with a large viscosity , @xmath124 , at low @xmath12 .
the shear rate in the laminar flow , calculated as @xmath125 , was proportional to @xmath50 , being @xmath126 in the standard configuration , and it gave a properly weighted average over the surface of the upper plate .
a suitable parameter characterizing growth of the flow resistance is the ratio of the average stress at the upper plate , @xmath127 ( defined as @xmath128 again ) , to the stress @xmath129 in the laminar flow at the same @xmath50 . in newtonian fluids this ratio generally increases with @xmath12 as
the flow becomes increasingly irregular , and magnitude of @xmath130 can be considered as a measure of strength of turbulence and of the turbulent resistance . in our standard
set - up , when a low viscosity newtonian fluid is used , @xmath127 becomes 30% higher than @xmath119 at @xmath131 .
therefore , @xmath131 can be regarded as a characteristic value , at which inertial effects in the flow become considerable .
the first indication of a strongly non - linear state in the swirling flow was significant growth of the flow resistance above the elastic instability threshold .
dependence of @xmath132 on the shear rate , @xmath133 , is shown in fig .
the shear rate was gradually increased in time , very slowly ( by about 10% h@xmath96 ) in the transition region , @xmath134 s@xmath96 , and faster below and above it .
one can see a sharp transition in flow of the polymer solution ( curve 1 , black line ) , that occurs at @xmath59 of about 1 s@xmath96 ( corresponding to @xmath135 ) , and that is manifested in a significant increase in @xmath127 compared to the laminar flow .
the reynolds number at the transition point is about 0.3 , so that the inertial effects are quite negligible there .
the transition has pronounced hysteresis ( gray line , corresponding to slow reduction of @xmath136 ) , which is rather typical for the purely elastic flow instability @xcite .
the ratio @xmath132 keeps growing with the shear rate and at the highest @xmath59 that has been reached the flow resistance is about 12 times larger than it would be in a laminar flow . in the same range of shear rates
flow of the pure solvent , curve 3 , is completely laminar and the ratio @xmath132 is unity within resolution of the rheometer ( about 1% ) .
growth of the resistance in the polymer solution flow becomes even larger , when the size of the gap is increased to 20 mm ( 2nd curve in fig.3 ) , and the gap ratio becomes 0.526 .
then the ratio @xmath132 reaches a value of 19 . for newtonian fluids in the same flow geometry such growth of the flow resistance
is found at @xmath12 of about 2@xmath9010@xmath137 .
for flow in a circular pipe this value of @xmath132 is reached at @xmath138 , which is usually considered as a region of rather developed turbulence @xcite . mechanical degradation of the polymers was quite small at shear rates below 1.5 s@xmath96 and 1 s@xmath96 , for @xmath87 mm and @xmath139 mm , respectively .
the dependencies of @xmath132 on @xmath59 were therefore reproducible in consecutive runs within about 1% . at higher shear rates degradation effects became appreciable .
thus , to ensure consistency of the results , time of measurements at high shear rates had to be made possibly short .
so , for @xmath59 above 1.5 s@xmath96 , curve 1 , the shear rate was raised by about 7% per minute , and the measurement time was not sufficient to average out fluctuations of @xmath140 produced by the flow .
irregular undulations in curves 1 and 2 in fig.3 at high @xmath59 are due to those fluctuations and the short time of averaging . in spite of the short time of the experiments , the solution relaxation time , @xmath141 ,
typically decreased by up to 10% after runs of the kind shown by the curves 1 and 2 .
nevertheless , a major part of the degradation occurred at the highest shear rates , above 5 s@xmath96 and above 3.5 s@xmath96 for the curves 1 and 2 , respectively .
therefore , the data at lower shear rates can be regarded as quite consistent .
resistance is an integral characteristic of a flow , and its fluctuations give some information about flow events on the scale of the whole system . in order to get information about characteristic frequencies of the flow on this integral scale
, we measured spectra of fluctuations of angular velocity of the upper plate at constant applied torque , @xmath75 .
( ar-1000 is a stress controlled rheometer ,
so it was more feasible to apply a constant torque and to monitor rotation velocity , than to do it the other way around . )
spectra of fluctuations of @xmath74 at a few @xmath75 corresponding to different average shear rates ( all above the transition point , @xmath142 s@xmath93 ) are presented in fig.4 .
root mean squares , rms , of fluctuations of the angular velocity , @xmath143 , were about 4% of its average values , @xmath144 , for all @xmath75 . because of the low rotation rates and low oscillation frequencies , inertial effects associated with acceleration of the shaft of the rheometer and of the liquid were always quite small .
( @xmath145 was less than 1% of @xmath146 , where @xmath147 is the moment of inertia of the system , and @xmath148 is rms of the time derivative of @xmath74 . )
so , the spectra in fig .
4 should reflect time scales of fluctuations of viscous and elastic stresses in the flow .
as the average shear rate is raised , the power of fluctuations , @xmath149 , increases and characteristic frequencies become higher , fig.4 .
the general form of the spectra remains very much the same , however .
there are no distinct peaks , except for instrumental peaks at @xmath150 , which are multiples of the average frequencies of the upper plate rotation , @xmath151 .
further , each spectrum has a broad region , spanning about a decade in frequencies , where dependence of @xmath149 on @xmath150 is close to a power law , @xmath152 , with @xmath153 .
( some increase in the power of fluctuations at higher @xmath150 is due to instrumental noise , and it was also measured in laminar flow of the pure solvent , curve 6 in fig.4 . ) the power law dependence implies that flow events on the integral scale of the system occur in this whole region of frequencies . as it was argued above , excitation of fluid motion in a broad continuous range of frequencies is one of the characteristic features of turbulence .
time spectra of fluctuations of flow velocity in a point at different constant shear rates ( all above the elastic instability threshold ) are shown in fig.5 .
one component of the velocity in the horizontal plane was measured with ldv in the center of the set - up , where its average value was zero .
the shear rates , @xmath59 , for the curves 1 - 5 were the same as the average shear rates for the curves 1 - 5 , respectively , in fig.4 .
the spectra in fig.5 have the same general features as those in fig.4 .
power of the fluctuations and their characteristic frequencies increase with @xmath59 , but the spectra remain very similar in their appearance .
in particular , the spectra do not have distinct peaks , and do have broad regions of frequencies , where the power of fluctuations decays according to a power law @xmath154 .
( flattening of the curves at high @xmath150 is due to instrumental noise ; see also discussion of fig.16 below . ) again , for each spectrum the power law decay region spans about an order of magnitude in @xmath150 , that implies excitation of the fluid motion in the whole range of the corresponding temporal scales .
one can see that characteristic frequencies of fluctuations in a point and at the integral scale(cf . fig.4 and fig.5 ) are rather close .
the exponent , @xmath155 , is about 3.5 , which is rather close to the exponent of 4.3 in fig.4 , and much larger than the kolmogorov exponent of 5/3 , found for velocity spectra of high @xmath12 inertial turbulence @xcite .
a few representative snapshots of the flow viewed from below are shown in fig.6 . in order to visualize the flow ,
the fluid was seeded with light reflecting flakes ( 1% of kalliroscope liquid ) .
the upper plate was black .
so , the bright regions are those , where the flakes are oriented parallel to the upper plate . the patterns of the polymer solution flow above the transition , at @xmath156 , fig.6a - b , and at @xmath157 , fig.6c - e , look quite irregular and exhibit structures of different sizes .
one can see , however , that the structures tend to have a spiral - like form , which is probably imposed by the average azimuthal flow and circular symmetry of the set - up .
further , there is a dark spot in the middle , which appears in most of the snapshots .
it corresponds to the center of a big persistent toroidal vortex , which has dimensions of the whole set - up ( see also below ) .
direction of the vortical motion was inwards near the upper plate , downwards near the center and outwards near the lower plate .
average flow velocity along the radial direction was measured by ldv in a few points , and the results agreed quite well with presence of the big persistent toroidal vortex .
flow of the pure solvent at the same shear rate was completely laminar that can be learned from the snapshot in fig.6f .
it looks quite uniform up to some variation along the radial direction due to varying shear rate and inhomogeneity along the circumference due to somewhat uneven ( and directed ) illumination .
the visual impression of spatial disorder in the flow at high @xmath38 is confirmed by a more careful analysis .
figure 7 shows average fourier spectra of the brightness profiles at @xmath157 along the diameter and along the circumference at @xmath158 .
the spectra were averaged over long series of flow pattern snapshots taken in consecutive moments of time .
the wavelength is measured in units of @xmath45 , so that the wave number , @xmath159 , of unity corresponds to a length of @xmath160 .
amplitude of fluctuations of the brightness along the circumference does not have any peaks and it exhibits a power law decay , @xmath161 , over almost a decade in the wave number domain .
this implies that there are no selected wave numbers in the azimuthal direction and the fluid motion is excited in a broad range of spatial scales .
( the plateau at high @xmath159 may be due to restricted spatial resolution of the method of visualization that was used . ) the spectrum taken along the diameter shows a general tendency of power law decay in even broader range of wave numbers .
however , it apparently differs from the azimuthal spectrum by a series of broad peaks , which may be due to the persistent toroidal vortex discussed above .
the average flow shear rate and the local gap ratio , @xmath162 , vary along the radius , so that the radial direction is not neutral . thus
, the flow is not structureless and homogeneous along the radial direction .
the visualization of the flow with the light reflecting flakes does not provide direct information about the fluid velocity field .
therefore , the specific value of the exponent in the power law fit in fig.7 does not have any special meaning . in order to obtain quantitative information about spatial structure of the fluctuating velocity field
, we explored velocity spectra in various off - center points , with non - zero average azimuthal velocity , @xmath163 . spectra of fluctuations of the radial component of velocity , @xmath164 , at @xmath165 s@xmath96 ( @xmath166 ) at four different radii are shown in fig.8 .
they all were measured at @xmath167 , where @xmath82 is the distance from the upper plate .
root mean squares of the fluctuations in all four points were rather close , varying between 0.88 and 0.99 mm / s .
the spectra in fig.8 bear the same general features of turbulence as the spectra in figs.4 , 5 and 7 discussed above .
they do not exhibit distinct peaks , and have broad regions of @xmath150 , where the power of fluctuations , @xmath149 decreases according to @xmath168 .
one can learn from fig.8 that as the point of measurement is moved away from the center , characteristic frequencies of the fluctuations become higher .
the most reasonable explanation of that is growth of the average azimuthal velocity , @xmath169 , which was 3.81 mm / s and 6.99 mm / s , at @xmath158 and @xmath170 , respectively .
so , the fluctuations of the velocity in time in these two points are mainly due to fluctuations in space , which are advected by the large mean flow velocity , @xmath171 . applying the taylor hypothesis
, we can view the spectra in time as spectra in space , with relation between the frequency and the wave number given by @xmath172 .
then the power law decay regions in curves 3 and 4 imply that the fluid motion is excited in the whole corresponding ranges of @xmath159 .
the ranges of the spatial scales , where the motion is excited , span about an order of magnitude for the both curves .
the exponents @xmath173 in the power laws @xmath174 ( and , so , @xmath175 ) are again quite large , about 3.6 for @xmath158 and about 3.3 for @xmath170 .
these large values of @xmath155 imply that the power of fluctuations decays very quickly as the size of eddies decreases .
the main contribution to fluctuations of both velocity and velocity gradients ( the power of the latter should scale as @xmath176 ) should therefore be due to the largest eddies .
mixing in the flow was observed using a droplet of black ink added to working fluid before rotation of the upper plate was started , fig .
9 . using a micropipette the droplet was carefully placed near the center ( at @xmath111 ) at about a half of the fluid depth .
the ink was dissolved in a concentrated sugar syrup , to match density of the droplet with the density of the working fluid .
we used the half - size set - up with @xmath177 mm , @xmath178 mm , @xmath179 mm .
the volume of the working fluid was about 9 ml , and the amount of the ink was 50 @xmath180l .
the upper plate was colored white and the fluid was viewed from below .
so , regions without the ink appeared bright , whereas dark regions corresponded to positions in the @xmath181 plane , where the total amount of the ink across the fluid layer was large .
consecutive stages of mixing in the polymer solution are shown in fig.9 .
rotation of the upper plate was abruptly started at @xmath182 at a rate of 1 turn per 4.26 seconds , which corresponded to a shear rate @xmath183 s@xmath96 .
it took about one minute for the irregular flow to develop , after the rotation of the upper plate was started .
( development of the irregular flow was judged by growth of the flow resistance , which was saturated after about one minute , cf.@xcite . )
so , no significant changes in the ink distribution occurred during the first 15 seconds ( fig.9 ) .
after 30 sec the ink became spread over the surface of the lower plate by the big toroidal vortex discussed above .
in the snapshots taken at later moments of time ( 60 , 90 and 120 sec in fig.9 ) one can see many fine scale structures
. those can be either due to excitation of the fluid motion at small spatial scales or significant stretching of fluid elements along their lagrangian trajectories by randomly fluctuating large scale eddies .
the contrast of the patterns gradually decreases with the time that indicates progressing mixing .
the pattern in the last snapshot , taken 8 minutes after the flow has been started , looks completely homogeneous . from the appearance of the mixing patterns in fig.9 , characteristic time of mixing
can be estimated as 120 seconds , corresponding to about 30 full turns of the upper plate .
the time required for mixing by molecular diffusion without macroscopic flow can be estimated as @xmath184 . substituting for @xmath185 the diffusion coefficient for the saccharose molecules , @xmath186 @xmath187/s
, we obtain a diffusion time of about @xmath188 seconds , that is four orders of magnitude larger than the mixing time in the flow .
( diffusion time for suspended solid particles of the black ink could be significantly larger . ) mixing in the flow of the pure solvent at the same shear rate is shown for comparison in fig.10 .
one can see that distribution of the ink remained inhomogeneous even after 9 hours , although the ink became somewhat spread with the time .
the reynolds number was about 0.5 , and there were some non - vanishing inertial effects in the flow .
in particular , centrifugal force gave rise to a slow fluid motion out of the @xmath189 plane .
this motion had a form of a big toroidal vortex with outflow near the upper plate and inflow near the lower plate .
( so , this vortex spun in the direction opposite to that of the toroidal vortex produced by the elastic stresses . )
the vortical motion stirred the fluid but it did not really mix it .
so , when the set - up was viewed from a side , one could see that the ink did not get to the toroidal core of the vortex
. summarizing the experimental results , we conclude that the flow of the elastic polymer solution at sufficiently high @xmath38 has indeed all the main features of developed turbulence that were stated above .
the fluid motion is excited in broad ranges of frequencies and wave numbers , both spanning about an order of magnitude .
the flow is accompanied by dramatic increase in the rate of transfer of momentum and mass . by the strength of the turbulent resistance and by the span of scales in space and time ,
where the fluid motion is excited , the observed flow can be compared to turbulence of a newtonian fluid in a pipe at @xmath12 of about 10@xmath190 .
this apparently turbulent fluid motion in the swirling flow between two plates arises at very low @xmath12 , where inertial effects are negligible , solely because of the non - linear mechanical properties of the elastic polymer solution .
we therefore call the phenomenon _ elastic turbulence_. distinctions between the elastic turbulence and the usual inertial turbulence , which is observed in newtonian fluids at high @xmath191 , are discussed in section iii.d below . in order to learn more about velocity field generated by the elastic turbulence , we measured average velocity and rms of the velocity fluctuations at different points .
profiles in @xmath82-direction of the average azimuthal velocity , @xmath192 , and of rms of its fluctuations , @xmath193 , at different flow conditions are shown in fig.11 .
( it was the standard set - up . )
the measurements are done at @xmath158 , that is rather far from the edge of the upper plate ( @xmath194 ) .
profile of @xmath192 in a low @xmath12 flow of the pure solvent is an almost straight line , curve 3 .
the elasticity driven turbulent flow significantly changes the distribution of @xmath192 .
it produces a high shear rate layer near the upper plate , curve 1 , and a low shear rate region near the middle of the gap ( at @xmath195 ) .
such a distribution of @xmath192 is reminiscent of average velocity profiles in usual high @xmath12 turbulence .
the perturbation of the @xmath192 profile becomes stronger when @xmath59 is increased , curve 2 .
in particular , the slope of the @xmath192-curve at small @xmath196 becomes larger that obviously corresponds to growth of @xmath130 with @xmath59 ( fig.3 ) .
fluctuations of the azimuthal velocity , curve 4 in fig.11 , are small near the upper plate , reach a maximum at @xmath197 , and start to decrease at larger @xmath82 . again
, such distribution of @xmath193 along @xmath82 is reminiscent of velocity fluctuations in turbulent flows of newtonian fluids @xcite .
rms of fluctuations reaches a value of about @xmath198 , so that rate of deformations produced by the fluctuating velocity field is on the order of @xmath199 .
that implies an essentially non - linear relation between the rate of deformations and the fluctuating elastic stress .
( it is discussed in more detail in the next section . )
the maximal rms of the velocity fluctuations at @xmath165 s@xmath96 was about 1.55 mm / s , that was about 7.5% of the upper plate velocity ( @xmath200 ) at @xmath158 and about 25% of the local value of @xmath192 .
as one can learn from curves 1 and 2 in fig.11 , there is significant asymmetry in distribution of the average shear rate , @xmath201 , along @xmath82 .
the main decrease of @xmath192 occurs near the upper plate .
so , at @xmath165 s@xmath96 , @xmath202 becomes 0.5 at @xmath203 and at @xmath195 the average velocity is only @xmath204 .
( velocity profile near the lower plate could not be measured because of instrumental restrictions . )
the uneven distribution of the shear rate along @xmath205-direction should create an essentially inhomogeneous profile of @xmath39 along @xmath82 .
indeed , in a laminar shear flow we have @xmath206 ( where @xmath67 is the polymer contribution to the shear stress ) .
therefore , the high shear rate near the upper plate should generate large @xmath39 and large hoop stress there .
a result of this large unbalanced hoop stress would be motion of the fluid near the upper plate inwards in the radial direction , which is quite consistent with our observation of the permanent toroidal vortex discussed above ( see fig.6 ) .
in fact the shape of the @xmath207 profiles in fig.11 , which is strongly asymmetric with respect to the mid - plane ( @xmath167 ) , is also consistent with the picture of a _ single _ large scale toroidal vortex .
indeed , even a possible increase in the slope of the @xmath207 profile near the lower plate still would not match the high shear at small @xmath82 , and the vortical motion in the @xmath208-plane should be mainly driven by the high hoop stress near the upper plate .
in contrast , a symmetric @xmath207 profile would imply equally high hoop stress near the upper and the lower plate ( and low hoop stress in the middle ) that would produce a pattern of two toroidal vortices on top of each other .
in fact the turbulent @xmath207-profiles of the kind shown by curves 1 and 2 in fig.11 were not always stable , especially at larger gap ratios .
so , in a system with @xmath209 at @xmath210 s@xmath96 the high shear rate layer randomly moved between a vicinity of the upper plate and a vicinity of the lower plate with a typical residence time of a few tens of minutes and a transition time of about a minute .
these fluctuations in the location of the high shear rate layer were accompanied by inversions of the spinning direction of the big toroidal vortex .
when the main drop of @xmath211 occurred near the lower plate , it was its vicinity where the radial flow was directed inwards .
therefore , the asymmetry of the @xmath207 profile with respect to the @xmath167 plane ( fig.11 ) is indeed directly related with the large scale vortical flow .
big toroidal vortices driven by the hoop stress are actually quite well known to appear in swirling flows of elastic fluids @xcite , and inhomogeneity of the shear rate profile in the primary laminar flow has long been recognized as their common origin . in our system
this vortex first arises as a stationary structure at low shear rates .
as it can be learned from fig.3 ( curve 1 at @xmath212 s@xmath96 and curve 2 at @xmath213 s@xmath93 ) the toroidal vortex leads to some increase in flow resistance even before the transition to the elastic turbulence . as it is suggested by the first three snapshots in fig.9 , the big toroidal vortex is also the first vortical motion to arise upon sudden inception of a shear rate , which is above the elastic instability threshold level .
( further redistribution of the average azimuthal velocity due to development of the turbulent flow should enhance this large scale vortical motion . )
so , we can conclude that transition to the elastic turbulence in the swirling flow between two plates is mediated by this vortex .
the vortex is always present in the flow , and it probably suffers permanent instability and is permanently fluctuating , that may be an important mechanism of production of fluid motion at smaller scales .
the toroidal vortex is providing a smooth , large scale velocity field , which is randomly fluctuating in time , and in which the liquid and the stress tensor imbedded in it are chaotically advected .
this type of advection can create variations of stress in a range of smaller scales , which may cause small scale fluid motion@xcite .
( this would be analogous to generation of small scale concentration variations in chaotic mixing by large fluctuating vortices . )
the large scale vortical flow is a probable reason for low @xmath38 at the elastic instability threshold , and for the very strong increase of resistance above the transition to elastic turbulence in the flow between two plates .
indeed , growth of the flow resistance in the ct flow at comparable @xmath38 ( fig.18 below ) is significantly smaller .
this situation is somewhat similar to inertial turbulence in newtonian fluids in the same systems . in the swirling flow between two plates
a big toroidal vortex , which is driven by centrifugal force , causes transition to turbulence at quite low @xmath12 and provides a route of efficient energy transfer to the turbulent flow .
many instructive pictures of the toroidal vortex , its evolution and instabilities in inertial , inertia - elastic , and elasticity dominated flow regimes in a swirling flow set - up with larger gap ratios can be found in ref.@xcite .
characteristic probability distribution functions ( pdf ) of azimuthal and radial components of the flow velocity in the regime of elastic turbulence are shown in fig.12a , b .
these distributions were obtained from ldv measurements taken at @xmath158 , @xmath214 at a shear rate of 4 s@xmath96 .
pdf s of the azimuthal and radial velocity can be fitted quite well by the functions @xmath215 $ ] and @xmath216 $ ] , respectively . here
@xmath217 mm / s and @xmath218 mm / s are the average velocities ; @xmath219 mm / s and @xmath220 mm / s are the rms of fluctuations for the azimuthal and radial components , respectively .
one can see that the both distributions have only minor skewness and are very close to gaussians .
( that skewness is probably due to the non - zero average velocity . )
probability distribution functions for gradients of velocity ( rates of deformation ) in the longitudinal and transversal directions , obtained from the same ldv time series , are shown in fig.13a and fig.13b , respectively .
the rates of deformation are multiplied by the relaxation time , @xmath13 , to make them dimensionless .
the velocity gradients were estimated using the taylor hypothesis , with smoothing over a distance of about 1.45 mm .
the difference , @xmath221 , between consecutive velocity readings with even time intervals of @xmath222 s was divided by @xmath223 and by the average velocity @xmath224 mm / s .
one can see that the distributions of @xmath225 and @xmath226 can not be fitted by gaussians .
in contrast to the velocity distributions in fig.12 , the both pdf s in fig.13 have well pronounced exponential tails , which imply significant intermittency of the velocity gradients .
the situation of nearly gaussian statistics of velocities and essentially non - gaussian , strongly intermittent distributions of velocity gradients is actually quite typical for high @xmath12 inertial turbulence@xcite .
hence , the elastic turbulence resembles high @xmath12 inertial turbulent flows in this respect .
along with the apparent similarity in phenomenology between the elastic and inertial turbulence , there are also quite a few very important distinctions .
an obvious reason for that is difference in the physical mechanisms , which underlie the two kinds of turbulent motion .
turbulence in newtonian fluids is observed at high @xmath12 , when the equation of motion ( eq .
1 ) is strongly nonlinear . the nonlinear term in eq .
1 is connected with kinematics of _ translational _ _ motion _ in a continuous medium and is due to fluid inertia .
the elastic turbulence occurs at low @xmath12 , when nonlinearity of the equation of motion ( eq .
2 ) is negligible .
the origin of the elastic turbulence is nonlinear dependence of the elastic stresses on rate of deformation in flow .
( this nonlinear dependence is very well seen in the model constitutive relation , eqs .
the latter nonlinearity is connected with kinematics of _ deformation _ in a continuous medium , and it is due to finite relaxation time of polymer molecules .
therefore , one can suggest that in the case of the elastic turbulence it would be more relevant to study the field of stresses and of rates of deformations rather than the velocity field .
the products of the randomly varying rates of deformation ( @xmath227 and @xmath226 in fig.13 ) and @xmath228 can actually be viewed as different estimates for weissenberg number , @xmath229 , of the fluctuating velocity field . the distributions in fig.13 show that @xmath229 is rather often above unity , and rms of its fluctuations estimated from fig.13a is about 0.83 .
fluctuations of the rate of deformation are directly connected with fluctuations of the elastic stresses by the constitutive relation ( see eqs . 3 and 4 ) .
since the fluctuating weissenberg number is rather high , this connection should be essentially nonlinear .
therefore , one can suggest that the field of @xmath230 in the elastic turbulence is highly complex and strongly intermittent
. it would certainly be quite instructive to explore spatial structure and temporal distribution of the elastic stresses , but there is currently no technique for local measurements of @xmath230 in a turbulent flow .
on the other hand , large scale properties of the @xmath230 field can be inferred from measurements of the total flow resistance . as it is well known , the high flow resistance in high @xmath12 inertial turbulence is due to large reynolds stresses .
the reynolds stress tensor is defined by the average values of @xmath231 , where @xmath232 and @xmath233 are different components of flow velocity . in the case of elastic turbulence
the reynolds stresses are quite small , since @xmath12 is low .
indeed , @xmath234 can be estimated as @xmath235 .
taking @xmath236 mm / s , which is the maximal value for rms of the azimuthal velocity fluctuations at @xmath158 and @xmath237 s@xmath96 , we obtain a reynolds stress of @xmath238 pa , while @xmath127 at the same conditions is about 16 pa .
therefore , the whole rise of flow resistance in the elastic turbulence is caused by growth of average elastic stresses @xcite . by the same token ,
the irregular fluctuations of the total flow resistance in the elastic turbulence , that were discussed above ( fig.4 ) , are due to fluctuations in the @xmath230 field at large spatial scales .
average growth of the polymer stresses can be estimated from the curves in fig.3 .
indeed , by momentum conservation the stress near the wall should be equal to its average value in the bulk , @xmath239 @xmath240 , @xmath241 @xmath242 ( we omitted signs of averaging for brevity ) .
the viscous stress due to newtonian solvent , @xmath243 , is linearly proportional to the shear rate , and its average value depends only on the boundary conditions and can not change as a result of a turbulent flow @xcite .
therefore , we can take @xmath244 @xmath245 , where @xmath46 is viscosity of the polymer solution measured in a laminar shear flow at a corresponding shear rate .
then for the ratio of the average polymer stresses in turbulent and laminar flow we obtain a relation @xmath246 . after plugging in numeric values from fig.3 ( curve 2 ) and from measurements of @xmath247 ( fig.2 )
we get ratios of @xmath248 as high as 65 at high shear rates . for a set - up with larger @xmath51 and a solution with
a smaller polymer concentration the ratio of the polymer stresses in turbulent and laminar flows estimated this way reached a factor of 170@xcite .
in molecular theory of polymer dynamics the tensor @xmath249 is found to be proportional to the polymer concentration and to the average tensor of inertia of the polymer molecules , @xmath250 ( where @xmath159 is an elasticity coefficient , which remains constant in a linear elasticity regime ) . since elastic energy density of the stretched polymer molecules
is given by a quadratic form of a similar kind , one can suggest linear dependence of the elastic energy on @xmath67 .
therefore , the density of the elastic energy should increase in the turbulent flow by about the same factor as @xmath67 .
the density of kinetic energy is proportional to the reynolds stress and although it certainly increases with development of the elastic turbulence , it always remains quite small compared to the elastic energy density .
so , we can suggest that the elastic stress , @xmath230 , should be the object of primary importance and interest in the elasticity driven turbulent flow , and it may be appropriate to view elastic turbulence as turbulence of the @xmath230 field .
the growth of the elastic stresses is also an evidence of significant extension of the polymer molecules in the flow .
the relation @xmath250 suggests that polymer extension should grow as @xmath251 .
taken together with the data on @xmath252 , it implies an eight - fold stretching of the polymer molecules in the turbulent flow beyond their extension in a laminar flow with the same average shear rate .
( for @xmath253 the corresponding extension factor is 13 . )
dramatic extension of the flexible polymer molecules in the turbulent flow environment , inferred here from the bulk measurements of the flow resistance , has been recently confirmed by direct visualization of individual polymer molecules in flow@xcite .
behavior of polymer molecules in chaotic and turbulent flows has been a subject of a long controversy@xcite , and it deserves a special discussion .
dynamics of polymer liquids have been most well studied in basic rheometric flows , such as simple shear and extension . in the former case @xmath254 is usually found to increase slower than linearly with the rate of deformation @xmath59 ( shear thinning ) . in
the latter case @xmath67 grows faster than linearly with the rate of deformation , @xmath255 , and at @xmath256 @xmath67 can grow by a few orders of magnitude as the total deformation ( measured by the henky strain , @xmath257 ) increases .
molecular basis of this behavior , stretching and tumbling in the case of a shear flow , and a coil - stretch transition in the case of an extensional flow , has been clearly demonstrated in the recent experiments by chu s group@xcite .
a turbulent flow involves complex three - dimensional velocity field , with a tensor of rate of deformation randomly varying in time and space .
therefore , an attempt to directly fit it into the context of rheometric flows may lead to frustration and confusion .
being stochastic phenomena , chaotic and turbulent flows are most appropriately treated statistically .
an object of particular interest for polymer dynamics is statistics of deformations of the fluid elements , in which the polymer molecules reside .
since size of the polymer molecules is almost always much smaller than the size of the smallest turbulent eddies , velocity field is smooth on the scale of the molecules . at every moment of time it is defined by some average velocity and a tensor of rate of deformation , @xmath258 .
statistics of deformations of a fluid element can be derived from statistics of the tensor @xmath259 in the lagrangian reference frame moving with the fluid element .
the tensor @xmath260 is three - dimensional and in all non - degenerate cases it has a positive eigenvalue , corresponding to extensional flow along some axis .
rate of the extension and direction of the axis vary randomly at a time scale , connected with characteristic time of variations in the flow .
a detailed statistical analysis was first made by lumley@xcite in early 70 s and was revised and refined recently@xcite . under quite general assumptions about statistics of @xmath259 , it was shown that a fluid element , which is initially a sphere , becomes an ellipsoid with its longest axis growing in average as @xmath261 .
the coefficient @xmath155 is a lyapunov exponent of divergence of two closely spaced points in the flow , and the value of @xmath155 is usually on the order of @xmath262 .
a somewhat surprising conclusion of the theory is that a generic random flow is in average an extensional flow in every point , with the rate of extension @xmath263 and unlimited henky strain .
therefore , one should expect the polymer molecules to become strongly stretched by the random ( turbulent ) flow , when the lyapunov exponent @xmath155 , as given by the statistics of @xmath259 , becomes larger than @xmath264 .
( more refined theory suggests that a significant fraction of the molecules should become substantially stretched even at lower @xmath155 due to intermittency in the statistics@xcite . )
one can see that the strong extension of the polymer molecules in the regime of elastic turbulence , taken together with the estimate @xmath265 are in good agreement with these theoretical predictions@xcite .
based on the discussion above , one can suggest the following scenario of development of the elastic turbulence .
the polymer molecules are stretched in the primary shear flow that leads to large elastic stresses , @xmath266 .
the elastic stresses render the primary shear flow unstable and cause an irregular secondary flow .
the flow acts back on the polymer molecules stretching them further and raising the elastic stresses even more@xcite .
that makes the flow increasingly turbulent , until a kind of a saturated dynamic state is finally reached .
this state implies some mutually consistent fields of average stresses and velocities , and their fluctuations .
those fields are connected with each other by the equation of motion and the constitutive relation .
we explored variation of different characteristics of the elasticity induced turbulent flow with viscosity of the polymer solution and with size of the set - up .
it was in sharp contradiction with the general principles of newtonian fluid mechanics , but quite in line with the concept of viscoelastic similarity suggested for purely elastic flow transitions @xcite .
so , velocity required for excitation of inertial turbulence in a newtonian fluid is proportional to viscosity of the fluid . this is connected with the fact that characteristic inertial time ( viscous diffusion time ) , @xmath267 , is inversely proportional to viscosity .
in contrast to it , polymer relaxation time , @xmath13 , usually grows proportionally to viscosity of the solvent@xcite .
when the system gap ratio and the viscosity ratio are both kept constant , purely elastic transitions should occur at a constant weissenberg number , @xmath268 . therefore , increase in viscosity of the solvent should lead to transition to elastic turbulence at lower velocity . in order to verify that we prepared paam solutions with the same @xmath269 and a relaxation times larger than 3.4 s using a few other sugar syrups with higher viscosities as solvents .
the transition to elastic turbulence in these solutions occurred at the same @xmath38 , indeed , while the angular velocity of the upper plate and reynolds number at the transition threshold were lower , scaling as @xmath270 and @xmath271 , respectively .
( in fact the situation was very similar to the case of purely elastic instabilities in ct flow @xcite . ) the dependencies of @xmath272 on @xmath38 for those solutions were close to curve 2 in fig.3 .
( in order to obtain a similar dependence , though , it was important to vary the shear rate slower , in proportion with the increase in @xmath107 . ) using a very viscous sugar syrup with @xmath273 pa@xmath274s , we prepared a polymer solution with a relaxation time of about 30 s , and observed a transition to elastic turbulence at @xmath275 and @xmath276 @xcite .
further , the viscous diffusion time , @xmath56 , is proportional to a square of the size of the system
. therefore , when size of the set - up is proportionally reduced , the linear and angular velocities required for excitation of inertial turbulence should increase as @xmath277 and @xmath278 , respectively . in contrast to it , polymer relaxation time , @xmath107 , does not depend on system size .
therefore , @xmath74 ( and @xmath59 ) at transition to the elastic turbulence should not depend on size of the set - up , as long as its proportions are preserved
. a natural scale of stress in the purely elastic flow regime is given by @xmath279 , and it also does not depend on size of the system .
( it may depend on @xmath280 , though . )
therefore transition to the elastic turbulence is expected to occur at the same characteristic stress , and dependence of @xmath132 on @xmath38 is supposed to be the same for all system sizes .
figure 14 shows dependence of @xmath132 on @xmath38 in a set - up with all the dimensions proportionally reduced by a factor of four compared with the standard configuration .
the plot looks quite similar to curve 1 in fig.3 .
one can see that the transition to the elastic turbulence indeed occurred at the same @xmath59 as in the standard set - up , although the reynolds number was 16 times lower . in order to compensate for lower sensitivity of the rheometer to shear stress due to smaller total torque
, the rate of variation of @xmath59 was reduced .
that may be the reason for a smaller region of hysteresis in fig.14 compared with fig.3 .
the plot in fig.14 also implies that the transition to elastic turbulence occurred at the same @xmath140 and @xmath38 , and dependence of @xmath140 on @xmath59 was also almost the same as in the standard configuration .
the parameters of the system relevant to the elastic turbulence are @xmath45 , @xmath44 , @xmath281 , @xmath282 , @xmath107 and @xmath74 .
( density of the fluid should be of no importance , since inertial effects are negligible . )
there seems to be no apparent way to construct out of these parameters a scale of length other than the size of the system ( with a possible pre - factor depending on @xmath280 and @xmath283 ) .
it implies that , when the dimensions of the system , @xmath45 and @xmath44 , are proportionally changed , while @xmath281 , @xmath282 , @xmath107 and @xmath74 are preserved , the turbulent velocity field should remain the same up to linear rescaling with size of the system .
this suggestion is qualitatively confirmed by the mixing patterns presented in fig.15 .
the snapshots in the left column are the same as in fig.9 .
they show consecutive stages of mixing of a droplet of ink in a polymer solution upon sudden inception of shear flow .
the dimensions of the system were two times smaller than those of the standard configuration .
the snapshots in the right column show mixing in the same polymer solution in a system with dimensions reduced four - fold compared with the standard configuration . in both cases
the area of the snapshots corresponds to the area of the upper plates .
so , the snapshots in the right column are magnified by a factor of two compared with the snapshots in the left column .
( the volume of the ink droplet added to the solution in the right column was 4 times smaller than for the left column in order to match the amount of ink per unit area . )
the angular velocities of the upper plates were the same , @xmath284 s@xmath96 , and so were the shear rates , @xmath183 s@xmath96 , and the weissenberg numbers .
however , the reynolds number for the right column was four times smaller , @xmath285 compared with @xmath286 for the left column .
the mixing patterns in fig.15 are irregular in the both columns , but the two sequences look rather similar and make an impression of two realizations of the same stochastic process .
so the transition to the elastic turbulence seems to go similarly in the two systems , and the mixing seems to progress with about the same characteristic time .
this implies that characteristic velocities of the irregular flow scale linearly with size of the system and with velocity of the primary flow .
patterns , which appear in the snapshots taken at equal elapsed times , seem to have similar sizes and similar contrast levels .
this suggests that in the elastic turbulence characteristic spatial scales of velocity field are indeed proportional to size of the system .
quantitative data on dependence of the turbulent velocity field on the system size are obtained from velocity spectra measured at a point with a large mean flow velocity . as it was argued above
, those measurements provide information about both power of the velocity fluctuations and spatial structure of the velocity field .
spectra of fluctuations of the azimuthal and radial velocity components measured at @xmath158 , @xmath287 in the standard and in the half - size set - ups are shown in fig.16 .
the shear rate was 4 s@xmath96 ( corresponding to @xmath166 ) in the both systems , the same as for the spectra in fig.8 and for pdf s in figs.12,13 . from the above similarity arguments fluctuating velocity
was expected to scale as @xmath55 .
rms of the fluctuations of @xmath288 and @xmath164 were 1.55 mm / s and 0.97 mm / s , respectively , for the standard configuration , and 0.78 mm / s and 0.48 mm / s , respectively , for the half - size system .
so the corresponding fluctuation amplitudes in the half - size system were indeed twice lower .
therefore , in order to match spectra in the two systems , the velocity measured in the half - size system was multiplied by a factor of 2 .
one can see , that upon this velocity rescaling , the spectra of both @xmath79 and @xmath80 become very close .
the apparent difference at high frequency is due to lower signal - to - noise ratio in the half - size system . indeed ,
one of the main sources of noise in the ldv measurements is finite size of the crossing region of the laser beams , over which the measured velocity is averaged .
characteristic velocity gradients ( scaling with @xmath74 and @xmath199 ) are expected to be equal in the two systems , and dimensions of the laser beam crossing region are the same as well .
therefore , the level of the gradient noise should not change , when the size of the system is reduced , whereas the level of the signal ( characteristic fluctuating velocity ) decreases . the average azimuthal velocities were 6.34 mm / s and 3.37 mm / s for the standard and the half - size systems , respectively , with their ratio being rather close to 2 again . therefore , applying the taylor hypothesis to the power spectra in fig.16 , we conclude that spatial structure of the velocity fluctuations in the two systems is statistically identical up to rescaling by a factor of 2 in the @xmath159-space .
thus , the experimental results in fig.16 provide further , more quantitative support to the suggestion of simple linear rescaling of the turbulent velocity field with size of the system at constant @xmath74 ( and @xmath281 , @xmath282 and @xmath107 ) .
structure of flow at small scales and near boundaries gives another example of substantial differences between the elastic and inertial turbulence .
it is instructive to summarize some relevant basic facts about the latter first . in high @xmath12 inertial turbulence
the range of excited spatial scales is restricted from below by the kolmogorov dissipation length , @xmath289 .
it can be estimated from the condition @xmath290 @xcite , with the local reynolds number , @xmath291 , defined as @xmath292 , where @xmath293 is a characteristic turbulent velocity difference at the scale @xmath294 .
since @xmath293 decreases as @xmath294 becomes smaller ( as @xmath295 in kolmogorov s picture of turbulence @xcite ) , @xmath291 decreases faster than linearly with decreasing @xmath294 .
so , the dissipation length @xmath289 necessarily exists in any inertial turbulent flow .
an issue related to the problem of the smallest excited length scale is stability of boundary layers near solid walls .
the latter can be most simply approached by introduction of a local reynolds number , @xmath296 , which is a growing function of distance , @xmath82 , from the wall .
( the usual convention is to use @xmath297 for distance from a wall ( transverse direction ) , and @xmath82 as a coordinate in the spanwise direction , but we would like to stay consistent with the notation used in fig.11 and cylindrical geometry . ) near a wall the mean flow velocity , @xmath207 , increases linearly with @xmath82 , giving @xmath298 at small @xmath82 . therefore , @xmath299 remains low in some vicinity of a wall and the flow there is a laminar shear .
this laminar shear layer near a wall ( usually called viscous sublayer ) ends at a distance @xmath82 ( decreasing with growth of global @xmath12 ) , where @xmath299 becomes sufficiently large , and the shear flow becomes unstable and turbulent .
analogously to the inertial turbulence , we can introduce a local weissenberg number , @xmath300 .
it does not decrease as @xmath294 becomes smaller , though .
indeed , the ratio @xmath301 is a characteristic gradient of the turbulent velocity smoothed over the length @xmath294 , and it can only grow as @xmath294 decreases . since @xmath302 does not get small at small @xmath294 , there appears to be no straightforward way to construct an analog of the dissipation length , @xmath289 , for the elastic turbulence .
so , there may be no macroscopic length setting a lower limit to the range of scales of the turbulent motion .
( something may change , though , at a scale comparable with size of the polymer molecules or with their characteristic diffusion length . ) this suggestion is consistent with the plots shown in fig.16 .
indeed , the spectra of both @xmath79 and @xmath80 in fig.16 have broad frequency ranges where the power of fluctuations decays as @xmath174 .
the value of @xmath155 is close to the values found for the other velocity spectra ( fig.5 , 8) .
the highest frequency in a power law decay region corresponds to the smallest spatial ( or temporal ) scale , at which the turbulent flow is excited .
the power law decay regions in the spectra in fig.16 are cut off at high frequencies by plateaus rather than by abrupt drops , however . as we argued before
, those plateaus appear solely due to the instrumental noise .
the power law decay range and the apparent range of excited spatial scales deduced from it become broader , when the signal - to - noise ratio becomes higher , fig.16 .
this lack of the cut - off frequency from above does not lead to any kind of high frequency divergence ( ultraviolet catastrophe ) , though . indeed , since @xmath303 , the total power of the velocity fluctuations is converging at high @xmath150 and @xmath159 , and the same is true for fluctuations of velocity gradients with their power scaling as @xmath304 in the spatial domain .
( we assume again that the taylor hypothesis is applicable . ) similarly to @xmath299 , one can introduce a local weissenberg number near a wall , @xmath305 ( @xmath207 here is the difference between the local mean flow velocity and the velocity at the wall at @xmath306 ) , and easily see that it does not become small at @xmath307 .
in fact , we can learn from the average velocity profiles in the elastic turbulence , fig.11 , that the region near the upper plate corresponds to the highest average shear rate and to the highest @xmath308 . on the other hand
, the velocity fluctuations clearly decline near the upper plate , fig.11 , and existence of such a stable high shear rate boundary layer in the elastic turbulence may appear rather puzzling .
a possible mechanism providing stability of the flow near the upper plate is significant shear thinning of the relaxation time , @xmath13 . as it was argued above , in the elastic regime flow stability
depends on the parameter @xmath43 . adopting it for the boundary layer
, we can suggest that stability should be defined by a local parameter @xmath309 , where @xmath310 , @xmath311 is the shear rate dependent relaxation time , and @xmath44 is the radial position of the region of interest . we can consider a case of a simple sigmoidal profile , where the whole growth of the mean flow velocity from zero to @xmath192 occurs across thin layers of widths @xmath312 near the boundaries . if @xmath192 , @xmath313 , and @xmath44 are all kept constant , we obtain @xmath314 , where @xmath315 . therefore ,
if the relaxation time decreases with the shear rate faster than @xmath316 , @xmath317 is a decreasing function , and the boundary layer remains stable as it becomes thinner and the slope , @xmath318 , increases
. such fast decay of @xmath13 with @xmath59 would also suggest that a more global ( although radius dependent ) parameter for the flow between two plates @xmath319}^{2}$ ] would be decreasing with @xmath44 , and the elastic instability would be limited to a region of sufficiently small @xmath44 .
this kind of behavior was indeed reported for a different polymer solution and a system with small @xmath51@xcite . in a more general case of shear thinning , with the relaxation time scaling as @xmath320 ( and @xmath321 ) , the condition @xmath322 ( where @xmath323 is a critical value for instability ) is satisfied for @xmath324 ^{1/(2 - 2\alpha ) } \cdot z^{\frac{1 - 2\alpha } { 2 - 2\alpha } }
$ ] , which is reminiscent of the @xmath207 profiles in fig.11 .
dependence of @xmath13 on @xmath59 in a shear flow can be inferred from measurements of normal stress difference , @xmath39 . at @xmath325
it is usually found that @xmath326 with constant @xmath13 and @xmath47 .
therefore , @xmath327 can be used as an estimate of shear thinned relaxation time at higher @xmath59 . measuring this way @xmath328 for solutions of a few lower molecular weight paam samples we usually found @xmath329 , with @xmath330 around 0.3 @xcite .
( resolution of our rheometer was not sufficient to measure normal stresses in the solution used in the experiments on the elastic turbulence at @xmath38 in the range of interest@xcite . )
therefore the shear thinning may indeed play an important role in stabilization of the boundary layer .
finally , it is worth noting that the whole concept of boundary layers is less straightforward in the case of elastic turbulence , and the condition of their stability may be less restrictive than for inertial turbulence . indeed , in a high @xmath12 turbulent flow characteristic fluctuating velocity , @xmath331 , generated by a flow instability originating at a distance @xmath82 from a wall scales as @xmath332 , and the boundary layer must become stable at some @xmath82 , since @xmath333 as @xmath307 .
in contrast , characteristic fluctuating velocity generated by an elastic instability should be proportional to @xmath334 and approaching zero at @xmath307 .
the experiments on the couette - taylor flow were carried out in a small set - up mounted vertically on the ar-1000 rheometer .
the stationary outer cylinder was made of glass , had a radius @xmath335 13.5 mm , was open from above and closed from below by a stainless steel plug with a flat top , fig.17 .
the rotating inner cylinder had a radius @xmath336 9 mm and was conical at the bottom with an angle of @xmath337 , so that the gap ratio in the cone - and - plate flow at the bottom was the same as in the annular gap , @xmath338 .
the cylindrical surface of the inner cylinder was 72 mm long .
the both cylinders were coaxial with the shaft of the rheometer , and the inner cylinder was attached to it .
the working liquid was filled above the upper rim of the inner cylinder . in order to reduce evaporation of the fluid from its open surface a special cover ( vapor trap )
the outer cylinder was enclosed in a thermal jacket with a square cross - section that was made out of plexiglass .
the temperature of the system was set by a fast stream of water pumped through the jacket from a refrigerated circulating bath .
the polymer solution was the same as in the experiments on the flow between two plates ( 80 ppm of paam in 65% sucrose , 1% nacl in water ) , and the experiments were carried our at the same temperature of 12 @xmath84c .
the characteristic shear rate in the flow was calculated as @xmath339 , and the weissenberg number was taken to be @xmath340 with @xmath341 s. first , we measured flow resistance in the polymer solution as a function of @xmath59 and @xmath38 .
the protocol was very similar to that described above for the flow between two plates . at the beginning ,
pure solvent ( with a viscosity @xmath282 ) was filled into the system and torque , @xmath342 , applied to the inner cylinder was measured as a function of angular velocity , @xmath343 .
the dependence was linear over the whole range of @xmath50 explored .
suggested dependence of torque for an ( imaginary ) laminar flow of the polymer solution was calculated from it as @xmath344 .
then the system was filled with the polymer solution , @xmath50 was gradually increased starting from a low point corresponding to @xmath345 , and the torque , @xmath75 , required to drive the actual polymer solution flow was measured .
the ratio of the two torques , @xmath346 , is equal to the ratio of the average stresses , @xmath347 , in the actual and the imaginary laminar flow of the polymer solution .
it is a quantitative expression of growth of the flow resistance compared with the laminar flow .
the dependence of @xmath348 on @xmath38 is shown in fig.18 ( thin black line ) for @xmath38 varying over almost two orders of magnitude .
one can see that the resistance ratio @xmath347 is equal to unity at low values of @xmath38 corresponding to laminar flow in the system .
then at @xmath38 of about 4 an abrupt transition occurs , and the flow resistance starts growing quickly and non - linearly .
reynolds number at the transition point is quite small , @xmath349 .
therefore , the transition is of a purely elastic nature .
the non - linear growth of the flow resistance is quite significant ; at the highest tested value of @xmath38 the ratio @xmath347 is more than 4 .
undulations of the curve in fig.18 reflect fluctuating nature of the flow resistance above the transition and are due to limited integration time during the measurements . in order to explore behavior of the flow resistance at decreasing shear rate
, we carried out another experiment , where @xmath38 was first set to 7 , well above the transition point , and gradually reduced afterwards .
dependence of @xmath347 on @xmath38 in this run is shown by a thick gray curve in fig.18 .
the two curves in fig.18 overlap at @xmath38 above 4.7 and below 1.8 , whereas in the region in between a well pronounced hysteresis occurs .
this broad region of hysteresis agrees quite well with previously reported results for couette - taylor flow in a purely elastic regime , which were obtained in a system with a smaller gap ratio and a different paam solution@xcite .
the appearance of the plot in fig.18 is quite similar to the dependencies of @xmath347 on @xmath38 for the flow between two plates shown in figs.3 and 14 .
the non - linear velocity transition was accompanied by emergence of a disordered and randomly fluctuating pattern of vortices on top of the basic shear flow .
when visualized with the light reflecting kalliroscope flakes , the fluctuating pattern looked similar to the disordered oscillations reported before at similar flow conditions in a larger ct system with @xmath350@xcite . in order to measure velocity associated with the fluctuating flow we used the ldv set - up again .
two horizontal laser beams were crossing inside the polymer solution seeded with polystyrene beads in the middle of the annular gap at @xmath351 and at about half height from the bottom .
the beams were adjusted in such a way that it was the radial component of the flow velocity , @xmath80 , perpendicular to the basic shear flow that was measured .
power spectra of fluctuations of @xmath80 at a few different @xmath38 above the non - linear transition are shown in fig.19 .
as @xmath38 is raised the total power of the velocity fluctuations increases and fluctuations at higher frequencies are excited .
nevertheless , all four curves share the same general appearance : there are two plateau regions at low and high @xmath150 ( the latter should be due to the instrumental noise ) and two contiguous regions of a power law decay in between , with two different exponents and a distinct inflection point separating them .
although characteristic frequencies increase with @xmath38 , those exponents are rather close for all four curves , around -1.1 at lower frequencies and around -2.2 in the high frequency power law decay regions .
apart from the inflection points , none of the curves exhibits any distinct peaks .
amplitude of the velocity fluctuations in all four cases was quite small compared with average flow velocity .
the latter was directed along the circumference ( azimuthally ) and equal to about @xmath352 .
( see also fig.20 below and discussion thereof ) .
therefore , the spectra in fig.19 should actually reflect spatial variation of @xmath80 in the azimuthal direction . in order to obtain additional information about the fluctuating flow velocity
, we used the same time series to calculate velocity statistics .
the probability distribution functions of @xmath80 at different @xmath38 are shown in fig.20 .
one immediately observes that , although the average velocity is always zero , the shapes of the distributions are essentially non - gaussian .
moreover , unlike the velocity distributions for the flow between two plates in fig.12 , pdf s in fig.20 can not be satisfactory fitted by skewed gaussians ( functions of the form @xmath353 $ ] ) .
one of the most notable features of the pdf s in fig.20 is relatively high probability for large negative values of @xmath80 ( long tails of the distributions at negative @xmath354 ) , which implies asymmetry for radial flows inwards and outwards .
although most striking in the curve for @xmath355 , this feature is still quite well expressed at @xmath356 .
in fact , the curve at @xmath355 is also quite similar to a pdf of @xmath80reported before in the ct flow with @xmath350 , ref.@xcite ( fig.12 therein ) , at comparable @xmath38 in the regime of disordered oscillations . in that latter case
the asymmetry in the radial velocity distribution was due to abundance of coherent structures having a shape of couples of vortices ( diwhirls ) with a narrow region of intensive radial flow inwards in the middle and slow radial flow directed outwards on the sides of a vortex couple .
such an asymmetric shape of the diwhirls is closely related to mechanism of generation of the elastic instability in ct flow , and it allows efficient pumping of energy from the primary shear flow into the secondary fluctuating velocity field @xcite .
therefore , the asymmetric shapes of pdf s in fig.20 suggest that the coherent structures of that diwhirl type are present in the ct flow even at high @xmath38 and may have significant influence on its organization and statistical properties .
the distribution functions in fig.20 also allow finding characteristic rates of extension in the flow in the radial direction , @xmath357 , and weissenberg numbers , @xmath358 , associated with them .
indeed , we have got boundary conditions of @xmath359 and @xmath360 at @xmath361 and @xmath362 .
therefore , @xmath363 in the bulk can be estimated as @xmath364 , where @xmath365 is rms of the radial velocity fluctuations .
the resulting values of @xmath363 calculated from pdf s in fig.20 are then 0.059 , 0.071 , 0.097 and 0.15 @xmath366 at @xmath38 of 5.4 , 13.6 , 21.8 and 35 , respectively .
those values of @xmath363 correspond to @xmath229 of 0.2 , 0.24 , 0.33 and 0.52 , respectively .
these estimated numbers are rather high and are generally consistent with the picture of strong elongation of the polymer molecules by elements of extensional flow in the fluctuating velocity field@xcite . summarizing the results shown in figs.18 - 20 ,
we conclude that in the purely elastic regime the ct flow behaves similarly to the flow between two plates described in section iii .
the purely elastic non - linear flow transition occurs at @xmath367 and ( arbitrarily ) small @xmath12 in the both cases ; the transition is rather sharp , has pronounced hysteresis and it is accompanied by significant growth of flow resistance .
the fluctuating velocity field above the transition is quite random for the both flows and its spectra do not exhibit any distinct peaks , cf .
fig.8 and fig.19 . therefore , although characterized much less extensively , the ct flow above the non - linear transition fits quite well the definition of elastic turbulence introduced above . nevertheless , the elasticity induced turbulent flow in the ct system differed from the turbulent flow between two plates quantitatively , and the turbulent flow effects in the ct system were less dramatic .
indeed , the non - linear transition in the ct system occurred at @xmath368 ( fig.18 ) , which is higher than critical @xmath38 in the flow between two plates , 3.2 and 2 , for @xmath369 and @xmath370 , respectively ( fig.3 ) .
the growth of flow resistance in the ct flow above the transition was notably smaller than in the flow between two plates - factor of 4.2 ( fig.18 ) vs. factors of 11 and 19 ( fig.3 ) at the highest @xmath38 explored .
( the ratio @xmath347 in the ct flow kept growing with @xmath38 , though , and it was the increasingly fast mechanical degradation of the polymers that prevented us from raising @xmath38 higher , fig.18 . ) characteristic amplitudes of fluctuating velocity in the ct flow ( fig.19 ) were significantly smaller than in the flow between two plates ( figs.11 , 12 ) , both in terms of @xmath55 ( 0.06 vs. 0.5 at @xmath371 ) and of the maximal azimuthal flow velocity ( 0.005 vs. 0.075 at @xmath371 ) .
this comparison is certainly not quite fair , since @xmath80 in the ct flow is a transversal velocity component .
hence , it should be best compared with @xmath372 in the flow between two plates , that was not measured .
the general tendency appears to be rather clear , though .
characteristic estimated rate of extension due to the fluctuating velocity field in the ct flow was also considerably smaller than in the flow between two plates , @xmath373 vs. @xmath374 at @xmath371 .
( we note here again , though , that these estimates were obtained from different types of data sets , fig.20 vs. fig.13 ) . nevertheless , even the relatively low value of the rate of extension in the ct flow is reasonably consistent with the general condition for it to be on the order of @xmath375 to cause significant elongation of the polymer molecules , growth of the elastic stresses and notable flow resistance increase .
as it was already suggested above , more favorable conditions for development of the elastic turbulence and stronger turbulent effects in the flow between two plates may be due to the large toroidal vortex .
this vortex , which is driven by the distribution of normal stresses in the basic shear flow , may provide a major channel for pumping of energy into the turbulent velocity field .
ct flow certainly has a more limiting geometry , without any coherent three - dimensional large scale flow of comparable magnitude and with fluid motion confined to the annular gap .
that latter confinement limits possible curvature of flow trajectories and may also make the general requirement of axial symmetry for excited vortex structures ( as dictated by the system geometry ) more severe and restrictive , somewhat impeding development of the turbulence .
in fact , axial symmetry of the vortex patterns is a quite pronounced feature in inertial turbulence in ct flow even at very high @xmath12 @xcite . among all the discrepancies between the elastic turbulence in the ct system and in the flow between two plates
, the most principal one appears to be the difference in the decay exponents in the power spectra of velocity fluctuations , cf .
figs.5 , 8 , 16 vs. fig.19 . in the case of the flow between two plates
the decay exponents were around -3.5 in the both spatial and temporal domains , while the power spectra in the ct flow had two distinct power law decay regions with significantly lower exponents of -1.1 and -2.2 . as it is shown in section v below ( fig.25 ) , velocity spectra in an elastic turbulent flow in a curvilinear channel
have regions of a power law decay as well .
the decay exponents are about -3.3 , suggesting that it is the ct flow that behaves abnormally
. a clue to the behavior of the ct flow may be the inflection points in the spectra in fig.19 , which appear at frequencies @xmath376 at all four @xmath38 .
these frequencies are close to average rates of rotation ( in revolutions per second ) of the fluid in the measurement point .
together with the non - gaussian and asymmetric shapes of pdf s in fig.20 , that brings us to a suggestion that the power spectra in fig.19 are significantly influenced by the axisymmetric coherent structures ( possibly of the diwhirl type ) discussed above .
those coherent structures may dominate the low frequency part of the spectra , @xmath377 , and also feed into the high frequency part , @xmath378 , ( where the original power of fluctuations should be rather low to begin with , figs.8 , 16 ) reducing the apparent absolute values of the decay exponents at @xmath378 .
the main motivation of the experiments on flow in a curvilinear channel ( dean flow ) was to carry out detailed quantitative study of mixing in the elastic turbulence @xcite .
it is an open flow that allows extended continuous experimental runs with reproducible and well controlled initial conditions , and easy gathering of extensive data at different stages of mixing .
the channel is schematically shown in fig.21 .
it had a uniform depth @xmath379 mm , was machined in a transparent bar of perspex , and it was sealed from above by a transparent window .
the channel consisted of a sequence of smoothly connected half - rings with the inner and outer radii @xmath380 mm and @xmath381 mm , respectively ; it was square in its cross - section , and had quite a high gap ratio , @xmath382 .
the high gap ratio was intended to facilitate development of an elastic instability at low @xmath38 and of intensive irregular flow above the instability threshold .
the channel had 30 repeating units , 18 mm in length each , fig.21 .
the liquids to be mixed were fed into the channel by two identical syringe pumps through two separate tubing lines , always at equal discharge rates .
chemical composition of the two liquids was always identical as well , with the only difference of a small concentration , @xmath383 ppm , of a fluorescent dye ( fluorescein ) added to one of them .
they were prepared from the same stock of a carefully filtered liquid , which was divided into two equal parts .
a small amount of a concentrated solution of the dye was added to one part , while the other part was diluted by an equal amount of pure water .
( each of the liquids was carefully mixed afterwards . )
this method of preparation provided very good matching of densities and refraction indices of the liquids .
the channel was illuminated from a side by an argon - ion laser beam converted by two cylindrical lenses to a broad sheet of light with a thickness of about 40 @xmath384 m in the region of observation .
it produced a thin cut of the three - dimensional mixing pattern , parallel to the top and the bottom of the channel at a half of the channel depth .
fluorescent light emitted by the liquid in the direction perpendicular to the beam and to the channel plane was projected onto a ccd array by a camera lens and digitized by a 8-bit , 512@xmath385512 frame grabber . using homogeneous solutions with different amounts of the dye
, we found intensity of the fluorescent light captured by the camera to be linearly proportional to the dye concentration .
therefore , concentration of the dye was evaluated from the fluorescent light intensity .
( the liquid without the dye appeared completely dark , fig .
22 , and signal form it was below noise level of the camera . )
the flow was always observed near the middle of a half - ring on the side , from which the laser beam was coming .
so , the number , @xmath386 , of a unit ( fig.21 ) counted from the inlet was a natural linear coordinate along the channel . since the total rate of liquid discharge , @xmath387 , was constant , @xmath386 was also proportional to the average time of mixing . in order to observe further stages of mixing ( corresponding to @xmath388 )
, we carried out a series of experiments , where the liquids were pre - mixed before they entered the channel .
a shorter channel of the same shape was designed for this purpose and put before the entrance to the original one . in multiple calibration experiments we found good
matching between states of mixing at @xmath389 with the pre - mixer and at @xmath390 without it@xcite .
therefore , for the experiments with the pre - mixing a number of 25 was added to the actual coordinate along the channel to calculate the effective @xmath386 .
flow velocity was measured directly with the aid of the ldv set - up . because of small width of the channel , a special effort was made to obtain high spatial resolution by reduction of the region of space , where the two laser beams crossed , and reduction of distance between the interference fringes .
focusing lenses with a small focal length ( about 25 mm ) were used , and the angle between the beams was raised to about 90@xmath84 in the air ( and about 60@xmath84 in the liquid ) . as a result
, the region of the beam crossing was brought down to @xmath391 @xmath384 m , and the distance between the fringes was 0.44 @xmath384 m .
we used the same polymer solution as in the experiments on the flow between two plates and on the ct flow .
( it was 80 ppm of paam , 65% sugar , 1% nacl in water . )
this time , however , the experiments were made at room temperature , @xmath392 @xmath84c .
so , the solvent viscosity was @xmath393 @xmath394 , and viscosity of the solution was @xmath395 @xmath394 at a shear rate @xmath165 s@xmath96 .
the relaxation time , @xmath13 , estimated from phase shift between the stress and the shear rate in oscillatory tests with a shear rate amplitude of 3 @xmath366 , was 1.4 s. an estimate for the diffusion coefficient of the dye was given by that for the saccharose molecules , @xmath108 @xmath109/s .
the characteristic shear rate and the weissenberg number in the flow were estimated as @xmath396 and @xmath397 , respectively .
the reynolds number was calculated as @xmath398 .
the reynolds number in the flow was always quite small , reaching only 0.6 for the highest @xmath387 that we explored .
therefore , flow of the pure solvent always remained laminar and no mixing occurred , fig.22a .
the boundary separating the liquid with and without the dye was smooth and parallel to direction of the flow and only became somewhat smeared due to molecular diffusion as the liquid advanced downstream .
flow of the polymer solution was laminar and stationary up to a value of @xmath387 corresponding to @xmath399 ( and @xmath400 ) , at which an elastic instability occurred .
this instability lead to irregular flow and fast mixing of the liquids .
a few typical mixing patterns at different @xmath386 in the polymer solution above the instability threshold are shown in photographs in fig.22b - d .
more insight about structure and evolution of the mixing patterns can be obtained from space - time diagrams .
representative diagrams taken at @xmath401 at four different @xmath386 are shown in fig.23 .
brightness profiles along a single line going perpendicularly to the channel near the middle of a half - ring ( a horizontal line going through the middle of a snapshot in fig.22 ) were captured at even time intervals of 80 ms and plotted from top to bottom .
the diagrams in fig.23 share the same chaotic appearance and show features at comparable scales , but they loose contrast as liquid advances downstream and gets progressively mixed . as it is illustrated by the space - time diagrams in fig.23 , mixing in the polymer solution flow above the instability threshold was a random process calling for statistical analysis@xcite .
a simple parameter characterizing homogeneity of the mixture is a root mean square of deviations of the dye concentration from its average value , @xmath402 , divided by the average value itself , @xmath403 .
small value of @xmath404 means high homogeneity and good mixing of the liquids . at the channel entrance , where the two injected liquids are perfectly separated , @xmath404 is unity , and it should become zero for a perfectly mixed liquid .
dependence of @xmath404 on @xmath38 near the exit of the channel , at @xmath405 , is presented in fig.24 .
statistics of the dye concentration was evaluated from space - time diagrams similar to those in fig.23 .
the regions near the walls of the channel with the widths of @xmath406 were excluded from the statistics , though , because of possible image aberrations . in a stationary flow regime ( @xmath38 below 3.2 ) , when the concentration profile did not change in time , the brightness profiles were measured over short time intervals ( about 100 s ) . in the regime of an irregular flow , however , the profile of concentration was strongly fluctuating .
so , in order to obtain representative statistics of @xmath407 , the measurements of the brightness profile were taken during quite long intervals of time ( about 20 - 30 min ) , that typically corresponded to the total liquid discharge of about @xmath408 . the plot in fig.24
is somewhat analogous to those in fig.3 and in fig.18 , which show dependence of the flow resistance on @xmath38 .
indeed , the decrease in @xmath404 is an integral result of mass transfer produced by the irregular flow in the channel , just as growth of the flow resistance is an integral characteristic of increase of momentum transfer in the elastic turbulence . at @xmath38 below 2 dye distribution patterns in the polymer solution
were similar to those in the solvent , fig.22a , with @xmath404 close to unity .
some decrease in @xmath404 at low @xmath38 is due to large residence time and diffusional smearing of the boundary between the liquids . at @xmath38 of about 2 reduction of @xmath404 from 1 to about 0.9 occurred , which was a result of a transition in the flow .
although this transition produced rather complex spatial distribution of the dye ( data not shown ) , the patterns were stationary and mixing was rather minor .
this transition was probably corresponding to onset of stationary vortices with vorticity directed along the mean flow , as it was predicted in ref.@xcite .
those vortices have recently been observed directly using particle image velocimetry@xcite ( see also below ) .
the most striking feature of the plot in fig.24 is certainly an abrupt drop in @xmath404 at @xmath399 , where the irregular motion of the liquid set in .
it is worth noting here that since the mixing and reduction in @xmath404 is a cumulative effect of stirring and diffusion in the flow starting from the inlet , the actual value of @xmath404 above the transition is defined by the distance from the inlet .
on the other hand , the onset of the fluctuating flow occurred rather consistently along the whole channel , and we expect @xmath61 to be virtually independent of the channel length , when the latter is sufficiently large .
we studied dependence of @xmath404 on @xmath386 at @xmath409 corresponding to highest homogeneity of the mixture near the channel exit ( fig.24 ) , and found @xmath404 to decay exponentially with the distance from the inlet with a characteristic decay length @xmath410 of about 15 segments@xcite .
one can learn from fig.24 that , if @xmath38 is raised above 6.7 , @xmath404 starts to increase again .
the most plausible explanation for this is saturation of growth of the velocity fluctuations together with reduction of the residence time in the flow at growing @xmath38 ( and average flow velocity , @xmath411 ) .
if the ratio between fluctuating and average flow velocities remains constant , while they both increase , the stirring in the flow remains the same , but there is less time available for molecular diffusion , and homogeneity is reduced as a result of it .
this situation can be qualitatively described by growth of the peclet number , @xmath412 .
it was recently found for a flow of a polymer solution in a channel of the same shape and at similar @xmath38 , that the characteristic length @xmath410 increases as @xmath413@xcite .
this suggests that @xmath404 should start increasing with @xmath38 once growth of the velocity fluctuations is slowed down .
typical time of mixing in the channel at @xmath401 was found to be 3 - 4 orders of magnitude shorter than diffusion time , @xmath414 , for the small molecules of fluorescein@xcite .
dependence of the efficiency of mixing at the optimal flow conditions ( for the 80 ppm solution it was @xmath401 , fig.24 ) on concentration of the polymers was surprisingly weak ( although @xmath61 grew fast , as the polymer concentration was decreasing ) .
so , for a solution with the polymer concentration of 10 ppm ( @xmath415 ) , @xmath404 of as low as 0.29 could be reached at @xmath405 .
( it was measured at @xmath416 , where inertial effects were still negligible ) .
excitation of irregular flow and active mixing was observed down to the polymer concentration of 7 ppm .
fig.25 shows power spectra of fluctuations of longitudinal and transversal components of the velocity in the polymer solution at @xmath401 .
the measurements were done at @xmath417 , near the middle of the half - ring in the middle of the channel .
a spectrum of the velocity fluctuations in the flow of the pure solvent at the same @xmath387 , giving just instrumental noise , is shown for comparison .
the mean velocity was @xmath418 mm / s ; the rms of fluctuations , @xmath193 , was @xmath419 and @xmath420 in the longitudinal and transversal directions , respectively .
measuring the transversal velocity component in a few off - center positions , we found non - zero averages , which typically persisted for a few minutes and changed their sign rather randomly in time .
this situation can be explained by the presence of persistent longitudinal vortical structures in the flow .
these vortices should be filling the whole channel cross - section , with their vorticity direction randomly jumping between parallel and anti - parallel to the mean flow .
those vortices may first appear at @xmath421 but remain stationary below @xmath422 . in order to measure resistance as a function of the flow rate
, we used a micro - fabricated channel with the same proportions but with a 30 times smaller segment size@xcite .
it had @xmath423 @xmath384 m , @xmath424 @xmath384 m , @xmath425 @xmath384 m and consisted of 46 segments .
the polymer solution had the same concentration of 80 ppm by weight of the same paam sample .
the newtonian solvent was significantly thinner , though , containing 35% of sucrose and 1% nacl , and having viscosity of @xmath426 @xmath394 at room temperature of 22 @xmath427c .
the polymer solution viscosity was @xmath428 @xmath394 at @xmath429 s@xmath96 .
the flow was driven by a difference of hydrostatic pressures , @xmath430 , applied between the inlet and outlet .
dependence of the flow rate , @xmath387 , on @xmath430 was measured using an in - situ compensation technique@xcite(see details in ref.@xcite ) .
fig.26 shows resistance ratio @xmath431 as a function of @xmath387 , where @xmath432 is the pressure difference expected in a laminar flow for a liquid with a viscosity , @xmath91 , at the same @xmath387 .
( average @xmath91 was calculated as @xmath433 again . )
one can see a distinct transition and an onset of non - linear growth of the flow resistance at @xmath434 nl / s , corresponding to @xmath435 s@xmath96 and @xmath436 .
the low value of re suggests that the transition is of a purely elastic nature .
relaxation time , @xmath13 , of the polymer solution was low and could not be measured directly . for a purpose of estimation
we can assume @xmath437 and plug into this relation the values for the more viscous solution used in the table - top channel , @xmath438 s at @xmath393 @xmath394 .
that gives @xmath439 s and @xmath440 at the instability threshold for the current solution , which is in reasonable agreement with the results in fig.24 .
as @xmath387 increases to 60 nl / s ( corresponding to @xmath441 ) , the flow resistance growth reaches a factor of 2.8 , which is comparable with the ct flow , although significantly lower than in the flow between two plates .
the experimental results in fig.24 and fig.26 show that above the elastic instability threshold the flow of the polymer solution in the channel exhibits two major features of turbulent flows : major increase in the rate of mixing and in the flow resistance .
further , more solid evidence for the turbulent character of the flow is given by the power spectra of fluctuating velocity in fig.25 .
the spectra of the both longitudinal and transversal velocity components do not exhibit any distinct peaks and have broad regions of a power law decay with a power of about -3.3 . since the power spectra in fig.25 were measured in a point with a high mean flow velocity ( 10 times higher than a characteristic fluctuating velocity )
, we can use the taylor hypothesis and argue that the spectra in fig.25 actually reflect spatial structure of the flow .
then the power law decay region can be transferred to the spatial domain , with the power of the velocity fluctuations scaling as @xmath442 with the wave number , @xmath159 .
the absence of peaks in the spectra in fig.25 and significant extension of the frequency region , where @xmath149 follows the power law , are strong evidence of fluid motion in a broad range of spatial scales and of turbulent character of the flow .
we also notice here that the exponent of -3.3 in fig.25 is very close to those measured in the flow between two plates , figs.5 , 8 , 16 , which varied from -3.3 to -3.6 depending on the position .
so , one can suggest that decay of the power of the velocity fluctuations with an exponent around -3.5 is a rather general feature of the elasticity induced turbulent flows .
it does not appear in the radial velocity spectrum of the ct flow ( fig .
19 ) , though , possibly because of abundance of the coherent axisymmetric structures ( see section 4b ) .
the functional form of the velocity power spectra , @xmath442 , deduced from fig.25 , suggests that the power of fluctuations of velocity gradients scales as @xmath443 .
an integral of @xmath443 diverges at @xmath444 and converges at @xmath445 .
it means that the main contribution to the fluctuations of the velocity gradients and the velocity differences at all scales comes from the biggest eddies , having dimensions of the whole system ( diameter of the channel ) .
this conclusion has an immediate implication for mixing in the flow : it should result in the same type of patterns and in functionally the same statistics as in the case of a completely homogeneous flow , @xmath446 , randomly varying in time .
such a flow is a realization of the so - called batchelor regime of mixing @xcite , and the problem of statistics of a tracer ( dye ) distribution in it has been solved analytically recently@xcite .
the batchelor regime occurs at small scales ( below the kolmogorov dissipation scale@xcite ) in the usual , high @xmath12 , turbulence , and it is rather difficult to realize in laboratory otherwise .
therefore , the elastic turbulent flow in the channel provided a very convenient experimental system for quantitative study of mixing in this regime @xcite .
the experimental results on correlation functions and pdf of dye concentration , and on their dependence on time of mixing agreed very well with the theoretical predictions @xcite .
a practical message of the experiments is that very viscous liquids can be efficiently mixed in curvilinear channels at very low flow rates by adding high molecular weight polymers at very low concentrations .
this method of mixing , we believe , can find some industrial and laboratory applications@xcite .
we studied flows of dilute solutions of a flexible high molecular weight polymer in three different flow set - ups , which shared the same feature of high curvature of the flow lines , quantitatively expressed in large gap ratios , @xmath51 .
high viscosity of the solvent and long polymer relaxation times ensured an elasticity dominated flow regime , where effects of the non - linear constitutive relation between polymer stress and rate of deformation prevailed over the inertia related non - linearity of the navier - stokes equation . as a result
, all flow transitions were induced by the non - linear elasticity , occurred at @xmath38 on the order of unity and at very low @xmath12 . in all three systems ,
a flow above the instability threshold was randomly fluctuating in time and space , and exhibited some or all of the main features of turbulence : fluid motion excited in a broad range of spatial and temporal scales , and significant increase in the rates of momentum and mass transfer ( flow resistance and mixing ) .
the extend of the flow resistance growth varied rather substantially between the systems ( factors of 20 , 4.2 and 2.8 for the flow between two plates , the ct flow , and the flow in the curvilinear channel , respectively ) and so did the exponents in the power law decay regions of the velocity spectra ( being from -3.3 to -3.6 for the flow between two plates , -1.1 and -2.2 for the ct flow , and -3.3 for the curvilinear channel ) .
we believe , however , that all the experimental findings are fitting rather well to the same general framework of a phenomenon , which we call elastic turbulence .
observation of the elastic turbulence and investigation of its properties have been greatly facilitated by the choice of the systems with large @xmath51 , where the elastic instability occurs at low @xmath38 , and by the choice of the high molecular weight polymer , which is only moderately extended ( compared to its full contour length ) at the instability threshold .
nevertheless , we believe that elements of the elastic turbulence should generally appear in three - dimensional flows of polymer solutions at high @xmath38 and low @xmath12 .
possible realizations of this regime may range from flows of polymer melts in industrial reactors to small scale flows in drag reducing aqueous polymer solutions .
dependence of the elastic turbulence onset conditions on parameters of the system and on properties of the polymers solution ( discussed in section iii.d.2 ) implies that using polymer solutions with sufficiently high elasticity , one can excite turbulent motion at arbitrary low velocities and in arbitrary small tanks .
( in some cases it may be the only way to produce a chaotic flow , which may be needed for fluid mixing . ) irregularly fluctuating flows showing multiple features of the elastic turbulence have been recently demonstrated in a rotating flow between two plates with a 300 @xmath384 m gap@xcite and in a 100 @xmath180 m thick curvilinear channel@xcite .
an apparent restriction is , however , that the size of the tank has still to be large compared to the size of the polymer coils .
recently there have been some serious advances in theory of random flows of polymer solutions in the regimes of both high and low re@xcite .
in particular , the following explanation of the velocity spectra shapes in figs.5 , 8 , 16 , 25 was proposed@xcite
. fast decay of the fluctuation power with @xmath159 implies a velocity field , where the main contribution to deformation and stirring ( stretching and folding ) at all scales comes from randomly fluctuating velocity field at the largest scale of the system .
therefore , it is suggested that the leading mechanism for generation of small scale ( high @xmath159 ) fluctuations in the elastic stress is advection of the fluid ( which carries the stress ) in this fluctuating large scale velocity field .
hence , the fluctuating velocity field and stress tensor can both be decomposed into large and small scale components , and the leading mechanism for generation of the small scale ( high @xmath159 ) portions is advection by the fluctuating large scale flow .
( the theory considers the elastic stress tensor being passively advected in a random velocity field , that is analogous to the concept of a passively advected vector in the magnetic dynamo theory @xcite . )
the small scale velocity fluctuations are a product of the small scale stress fields .
the elastic stresses are mainly pumped by the large scale flow field , and the smaller scale stress fields created by the advection are permanently decaying because of the polymer relaxation .
these simultaneously occurring advection and relaxational decay of @xmath447 result in a quick decay of @xmath447 fluctuations at large @xmath159 that should produce @xmath448 velocity spectra with @xmath303 .
the current understanding of the elastic turbulence is still remaining quite incomplete , however , and this phenomenon certainly awaits further detailed experiments and numerical simulations . * acknowledgement .
* we are grateful to m. chertkov , g. falkovich , and v. lebedev for many useful and illuminating discussions .
this work is partially supported by an israel science foundation grant , binational us - israel foundation , and by the minerva center for nonlinear physics of complex systems . | following our first report ( a. groisman and v. steinberg , @xmath0 @xmath1 , 53 ( 2000 ) ) we present an extended account of experimental observations of elasticity induced turbulence in three different systems : a swirling flow between two plates , a couette - taylor ( ct ) flow between two cylinders , and a flow in a curvilinear channel ( dean flow ) .
all three set - ups had high ratio of width of the region available for flow to radius of curvature of the streamlines .
the experiments were carried out with dilute solutions of high molecular weight polyacrylamide in concentrated sugar syrups .
high polymer relaxation time and solution viscosity ensured prevalence of non - linear elastic effects over inertial non - linearity , and development of purely elastic instabilities at low reynolds number ( re ) in all three flows . above the elastic instability threshold ,
flows in all three systems exhibit features of developed turbulence .
those include : ( i)randomly fluctuating fluid motion excited in a broad range of spatial and temporal scales ; ( ii ) significant increase in the rates of momentum and mass transfer ( compared to those expected for a steady flow with a smooth velocity profile ) .
phenomenology , driving mechanisms , and parameter dependence of the elastic turbulence are compared with those of the conventional high re hydrodynamic turbulence in newtonian fluids .
some similarities as well as multiple principal differences are found . in two out of three systems ( swirling flow between two plates and flow in the curvilinear channel ) power spectra of velocity fluctuations are decaying rather quickly , following power laws with exponents of about -3.5 .
it suggests that being random in time , the flow is rather smooth in space , in the sense that the main contribution to deformation and mixing ( and , possibly , elastic energy ) is coming from flow at the largest scale of the system .
this situation - random in time and smooth in space - is analogous to flows at small scales ( below kolmogorov dissipation scale ) in high re turbulence . 2 |
the spontaneous breakdown of chiral symmetry plays a central role in the spectrum of light hadrons .
since it is an intrinsically non - perturbative phenomenon , the only way to study it from the first principles of qcd is via the lattice regularization . yet
, already many years ago nielsen and ninomiya proved that a translationally invariant , local lattice formulation of the qcd dirac operator @xmath9 , retaining chiral symmetry in the massless limit , and with the correct number of physical fermionic degrees of freedom , is forbidden @xcite .
this no - go theorem can be circumvented , by constructing lattice fermions satisfying a _ modified _ form of chiral symmetry @xcite , and obeying the ginsparg- wilson relation @xcite .
although explicit formulations of lattice ginsparg - wilson fermions are known @xcite , currently their practical use in realistic , large - scale lattice qcd simulations is still limited , due to the high computational overhead .
the most widely - used lattice discretizations of the dirac operator are either based on the addition of a second - derivative term to the kinetic part of the quark action @xcite to remove ( or `` quench '' ) the unphysical `` doubler '' modes in the continuum limit by giving them a mass @xmath10 , or on a site - dependent spin diagonalization , which leads to the so - called staggered formulation @xcite .
the former approach introduces an explicit breaking of chiral symmetry , and , as a consequence , an additive renormalization of the quark mass , which has to be fine - tuned .
in contrast , the staggered operator preserves a remnant of chiral symmetry ( sufficient to forbid additive mass renormalization ) , and leads to a reduction of the matrix size .
however , the staggered formulation only removes part of the unphysical modes , reducing the number of quark species in four ( @xmath11 ) spacetime dimensions from 16 ( @xmath12 ) down to four ( @xmath13 ) `` tastes '' , which become degenerate @xcite ( and consistent with the properties related to the global symmetries of the continuum dirac operator @xcite ) in the @xmath14 limit . in order to simulate qcd with two light fermions ,
one then has to apply the so - called `` rooting trick '' , which has been a subject of debate for the last few years @xcite .
some recent works have discussed the idea of using a staggered kernel with a _ taste - dependent _ mass term to obtain two ( or one @xcite ) massless fermion species . such formulation , which is one of the various approaches aiming at minimally doubled fermions @xcite , could combine the advantages of the overlap construction with the computational efficiency of a staggered kernel .
furthermore , this formulation appears to be particularly attractive from the point of view of topological properties @xcite . using a staggered operator with a `` flavored '' mass term as
the kernel in an overlap construction is a very appealing idea , but the properties of such operators ( with various taste - dependent mass terms ) are interesting on their own .
in fact , while the overlap construction completely removes the need for fine tuning to achieve massless fermions , it still leads to a considerable computational overhead .
in contrast , using a staggered operator with taste - dependent mass _ la _ wilson requires fine tuning to obtain exactly massless modes , but , by virtue of the reduced size of the operator , may still be a computationally competitive alternative to the usual wilson discretization , while avoiding the rooting prescription
. this motivation led us to address a numerical investigation of different operators of this type that we present here ( preliminary results have appeared in @xcite ) . in the following , we present a systematic classification of the possible taste - dependent mass terms , discuss their analytical features in the free theory , and then move on to the interacting case that we study via numerical simulations .
we perform an elementary measurement of the pion mass on a set of quenched configurations , and verify the expected pcac behaviour as one approaches the chiral limit . in an appendix
, we also explore the properties of the staggered overlap operator proposed in @xcite , in comparison with the usual overlap based on the wilson kernel .
in particular , we compare the locality of the operators , and the computational cost of applying them to a vector and of solving for the quark propagator .
the structure of this paper is as follows .
first , in sec .
[ sec : theoretical ] we recall theoretical aspects of the construction of taste - dependent mass terms , and discuss their spectral structure in the free field case . then , we address the interacting case , presenting our numerical studies in sec .
[ sec : numerical ] .
we summarize our findings and discuss their implications for possible future , large - scale applications of these operators in sec .
[ sec : conclusions ] . finally , in the appendix
[ appendix ] , we report on our study of an overlap operator based on a staggered kernel , as proposed in ref . @xcite .
the staggered operator @xcite @xmath15 with @xmath16 and @xmath17 , is a computationally very efficient way to discretize the massless qcd dirac operator on a @xmath11-dimensional euclidean hypercubic lattice of spacing @xmath18 .
this operator is invariant under a global @xmath19 symmetry , which can be interpreted as a remnant of chiral symmetry : in fact , @xmath20 anticommutes with the operator @xmath21 defined by @xmath22 . in the free theory , one can easily see that in four dimensions the operator @xmath21 has @xmath23 structure in spin - taste space @xcite . the construction of @xmath20 is based on a local spin diagonalization , which , for the four - dimensional case , allows one to reduce the number of fermion components by a factor of @xmath24 with respect to the naive operator , and yields four tastes in the continuum limit .
the degeneracy between these four tastes is explicitly broken by gauge interactions at finite lattice spacing @xmath18 , but is recovered in the continuum limit @xmath14 .
recently , various works explored the idea of using staggered operators with taste - dependent mass terms @xcite .
following , e.g. , the discussion in the classic paper by golterman and smit @xcite , the possible matrix structures ( in taste space ) for a mass term can be classified as * @xmath25 ( `` 0-link '' ) , of the form @xmath26 * @xmath27 ( `` 1-link '' ) , involving a sum of terms , each containing 1 link @xmath28 * @xmath29 ( `` 2-link '' ) , involving a sum of terms , each containing 2 links @xmath30 * @xmath31 ( `` 3-link '' ) , involving a sum of terms , each containing 3 links @xmath32 * @xmath33 ( `` 4-link '' ) , involving a sum of terms , each containing 4 links @xmath34 it is highly desirable to preserve the symmetry @xmath35 , because it guarantees that @xmath36 is real , and non - negative ( in the absence of real negative eigenvalues ) , thus avoiding a `` sign problem '' in the measure @xcite .
this symmetry is satisfied only if the hermitian mass term connects sites of the same parity .
thus , we do not consider the 1-link or a 3-link mass terms further .
this leaves three possibilities : 0- , 2- and 4-link mass terms .
@xcite . in that case , @xmath33- hermitian mass terms are obtained in the 0- , 1- and 3-link cases .
however , we did not find a continuum - like dispersion relation for the real modes in any of these cases . ]
the 0-link mass term corresponds to the usual staggered operator , with a taste - independent bare mass @xmath37 the staggered operator with a 2-link mass term , which was discussed in refs .
@xcite , can be written in the form @xmath38 where ( following the notation of ref .
@xcite ) @xmath39 @xmath40 @xmath41 @xmath42 finally , a staggered operator featuring a mass term with @xmath33 structure in taste space @xcite can be written as @xmath43 with @xmath44 where @xmath45 while @xmath46 is the average of four - link parallel transporters joining sites at opposite corners of the elementary lattice hypercubes @xmath47 note that the mass term appearing on the r.h.s .
( [ fourlinkkernel ] ) is hermitean and commutes with @xmath21 . in the free limit .
central panel : free spectrum of operator @xmath48 ( eq.([twolinkkernel ] ) ) , which includes a taste - dependent mass term with tensor - like structure in taste space ( i.e. , a 2-link mass term ) .
right panel : free spectrum of operator @xmath49 ( eq.([fourlinkkernel ] ) ) , which includes a taste - dependent mass term with @xmath33 structure in taste space ( i.e. , a 4-link mass term).,title="fig:",scaledwidth=73.0% ] in the free limit .
central panel : free spectrum of operator @xmath48 ( eq.([twolinkkernel ] ) ) , which includes a taste - dependent mass term with tensor - like structure in taste space ( i.e. , a 2-link mass term ) .
right panel : free spectrum of operator @xmath49 ( eq.([fourlinkkernel ] ) ) , which includes a taste - dependent mass term with @xmath33 structure in taste space ( i.e. , a 4-link mass term).,title="fig:",scaledwidth=29.5% ] to understand the properties of these three different types of operators it is instructive to start by discussing their spectra in the free limit .
the three panels in fig .
[ fig : freespectra ] show the structure of the spectrum of eigenvalues for @xmath20 ( for @xmath50 , the spectrum is just trivially shifted by @xmath51 ) , for @xmath48 , and for @xmath49 in the non - interacting case .
in the free limit the eigenvalues of @xmath50 on a lattice with @xmath52 sites along the @xmath53 direction read @xmath54 with eight degenerate eigenvalues of both signs , and where @xmath55 ( @xmath56 ) if the fermionic field satisfies ( anti-)periodic boundary conditions along the @xmath53 direction .
for @xmath48 the free eigenvalues take the form ( for @xmath57 ) @xmath58 and @xmath59 in which the @xmath60 signs are chosen independently and the eigenvalues are doubly degenerate , and having defined @xmath61 @xmath62 @xmath63 where @xmath64 , and @xmath65 . expanding for small
momenta gives @xmath66 so that at low momenta , the eigenmodes corresponding to @xmath67 get a mass of @xmath68 , while the eigenmodes corresponding to @xmath69 are massless . finally , the free spectrum of @xmath49 reads : @xmath70 note that , in the continuum limit , the point where the spectrum of the @xmath20 operator intersects the real axis corresponds to four massless modes .
by contrast , @xmath48 leads to one mode in each of the two intersections away from the origin , and two at the origin .
finally , for @xmath49 one obtains two modes at each of the two intersections of the spectrum with the real axis .
the taste chirality of the eigenmodes @xmath71 of @xmath49 and @xmath48 , is given by @xmath72 , where @xmath73 exactly , and @xmath74 in spin @xmath75 taste .
the taste chirality of the eigenmodes corresponding to eigenvalues @xmath69 is @xmath76 , while that of the @xmath67-eigenvectors is @xmath77 .
this is also depicted in fig .
[ lip ] ; one can see that the taste chirality of the real eigenmodes is @xmath78 , and is the same ( @xmath79 or @xmath80 ) for all modes in a given branch of the @xmath49 or @xmath48 spectrum .
the implications of a well - defined taste chirality have been stressed in @xcite : if @xmath81 , then @xmath82 , so that the _ spin _ chirality of the real eigenmodes can be probed by @xmath21 .
this is the reason why the index theorem applies to @xmath48 and @xmath49 , while it does not for @xmath20 ( where both the @xmath78 taste chiralities lay on the same single branch . ) and @xmath49 eigenvectors , as a function of ( the component of minimum modulus of ) their momentum , in the free limit . on an infinite lattice ,
the eigenvectors associated with real eigenvalues have vanishing momentum and a well - defined taste chirality @xmath78 . for eigenmodes corresponding to eigenvalue @xmath69 ,
the taste chirality becomes @xmath79 , while for @xmath67-eigenmodes the taste chirality approaches @xmath80 .
( right ) the taste chiralities of the eigenmodes of the @xmath48 operator ; the size of the points corresponds to the magnitude of @xmath76 , while the color indicates the sign : blue for @xmath83 , red for @xmath77.,title="fig:",scaledwidth=50.0% ] and @xmath49 eigenvectors , as a function of ( the component of minimum modulus of ) their momentum , in the free limit . on an infinite lattice ,
the eigenvectors associated with real eigenvalues have vanishing momentum and a well - defined taste chirality @xmath78 . for eigenmodes corresponding to eigenvalue @xmath69 ,
the taste chirality becomes @xmath79 , while for @xmath67-eigenmodes the taste chirality approaches @xmath80 .
( right ) the taste chiralities of the eigenmodes of the @xmath48 operator ; the size of the points corresponds to the magnitude of @xmath76 , while the color indicates the sign : blue for @xmath83 , red for @xmath77.,title="fig:",scaledwidth=50.0% ] [ lip ] a shift of the spectra by a real value can thus lead to chiral low - momentum zero modes in each branch , and hence to the possibility of constructing an appropriate index . a common way to study the index consists of looking at the flow of eigenvalues @xmath84 of : @xmath85 in general ,
if @xmath86 has a zero - mode @xmath87 for @xmath88 , then , correspondingly , @xmath89 has a vanishing eigenvalue @xmath90 . with a small perturbation of @xmath51 away from @xmath91 ,
i.e. @xmath92 , at leading order the eigenvalues get displaced by an amount @xmath93 , namely one finds _ crossings _
@xmath94 , if @xmath87 is a chiral mode : @xmath95 . as pointed out in @xcite , the saturation of @xmath72 at value @xmath78 discussed above allows us to trade @xmath96 for @xmath21 and use the latter in eq.([standardflow ] ) .
an alternative way to look at the spectral flow was proposed in ref .
@xcite for the @xmath49 operator , by studying the eigenvalues of .
as recognized in @xcite , that operator is the same as eq.([adamsflow ] ) up to a redefinition of the @xmath97 phase factors . ]
@xmath98 fig .
3 displays a comparison of the two different ways to define the spectral flow for the @xmath49 operator ( see @xcite for a recent related study ) : the plots in the top row show the flow of eigenvalues of @xmath99 as a function of @xmath100 ( eq.([adamsflow ] ) ) , whereas those in the bottom row refer to the `` standard '' definition of the flow , using eq .
( [ standardflow ] ) . in each row , the left panel displays the results from a cold ( i.e. , free ) configuration on a lattice of size @xmath101 , while the central panel is obtained from a cooled configuration of topological charge @xmath102 on a lattice of size @xmath103 , and finally the right panel displays the results from a `` rough '' ( i.e. , non - cooled ) quenched @xmath104 configuration at @xmath105 , on a lattice of size @xmath106 . in the latter case
, the comparison of the two flow definitions shows that , with the standard definition , the region around the real axis is populated by a large number of eigenvalues , preventing one from identifying the crossing with accuracy .
operator as obtained from the eigenvalues of the operators defined in eq .
( [ adamsflow ] ) ( top row panels ) and eq .
( [ standardflow ] ) ( bottom panels ) .
the three plots ( from left to right ) in each row show , respectively , the eigenvalues of @xmath107 ( or @xmath89 ) as a function of @xmath108 ( or @xmath51 ) from a free configuration on a lattice of size @xmath109 , from a cooled configuration of topological charge @xmath102 on a lattice of size @xmath103 , and from a non - cooled @xmath104 gauge configuration at @xmath110 , on a lattice of size @xmath106.,title="fig:",scaledwidth=39.0% ] operator as obtained from the eigenvalues of the operators defined in eq .
( [ adamsflow ] ) ( top row panels ) and eq .
( [ standardflow ] ) ( bottom panels ) .
the three plots ( from left to right ) in each row show , respectively , the eigenvalues of @xmath107 ( or @xmath89 ) as a function of @xmath108 ( or @xmath51 ) from a free configuration on a lattice of size @xmath109 , from a cooled configuration of topological charge @xmath102 on a lattice of size @xmath103 , and from a non - cooled @xmath104 gauge configuration at @xmath110 , on a lattice of size @xmath106.,title="fig:",scaledwidth=39.0% ] operator as obtained from the eigenvalues of the operators defined in eq .
( [ adamsflow ] ) ( top row panels ) and eq .
( [ standardflow ] ) ( bottom panels ) .
the three plots ( from left to right ) in each row show , respectively , the eigenvalues of @xmath107 ( or @xmath89 ) as a function of @xmath108 ( or @xmath51 ) from a free configuration on a lattice of size @xmath109 , from a cooled configuration of topological charge @xmath102 on a lattice of size @xmath103 , and from a non - cooled @xmath104 gauge configuration at @xmath110 , on a lattice of size @xmath106.,title="fig:",scaledwidth=39.0% ] operator as obtained from the eigenvalues of the operators defined in eq .
( [ adamsflow ] ) ( top row panels ) and eq .
( [ standardflow ] ) ( bottom panels ) .
the three plots ( from left to right ) in each row show , respectively , the eigenvalues of @xmath107 ( or @xmath89 ) as a function of @xmath108 ( or @xmath51 ) from a free configuration on a lattice of size @xmath109 , from a cooled configuration of topological charge @xmath102 on a lattice of size @xmath103 , and from a non - cooled @xmath104 gauge configuration at @xmath110 , on a lattice of size @xmath106.,title="fig:",scaledwidth=39.0% ] operator as obtained from the eigenvalues of the operators defined in eq .
( [ adamsflow ] ) ( top row panels ) and eq .
( [ standardflow ] ) ( bottom panels ) .
the three plots ( from left to right ) in each row show , respectively , the eigenvalues of @xmath107 ( or @xmath89 ) as a function of @xmath108 ( or @xmath51 ) from a free configuration on a lattice of size @xmath109 , from a cooled configuration of topological charge @xmath102 on a lattice of size @xmath103 , and from a non - cooled @xmath104 gauge configuration at @xmath110 , on a lattice of size @xmath106.,title="fig:",scaledwidth=39.0% ] operator as obtained from the eigenvalues of the operators defined in eq .
( [ adamsflow ] ) ( top row panels ) and eq .
( [ standardflow ] ) ( bottom panels ) .
the three plots ( from left to right ) in each row show , respectively , the eigenvalues of @xmath107 ( or @xmath89 ) as a function of @xmath108 ( or @xmath51 ) from a free configuration on a lattice of size @xmath109 , from a cooled configuration of topological charge @xmath102 on a lattice of size @xmath103 , and from a non - cooled @xmath104 gauge configuration at @xmath110 , on a lattice of size @xmath106.,title="fig:",scaledwidth=39.0% ] [ adams_original_classic ] next , it is interesting to compare the identification of the index , using the spectral flow defined from eq .
( [ standardflow ] ) , for staggered fermions with a taste - dependent mass term , and for conventional wilson fermions .
this is shown in fig .
[ flowcomparison ] : the left , central and right plot in each row show the spectral flow for @xmath49 , @xmath48 and a standard wilson operator , respectively , while the three different rows , from top to bottom , refer to a cold configuration , to a cooled @xmath102 configuration , and to a non- cooled @xmath104 quenched configuration at @xmath110 .
it is interesting to observe that , as expected , the spectral flow on a cooled instanton configuration clearly reveals @xmath111 crossings .
however , one already sees that in the plots of the @xmath110 configuration the gap tends to close .
this is especially the case for the @xmath49 operator , and is related to the properties that will be discussed in section [ sec : numerical ] .
( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] ( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] ( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] ( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] ( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] ( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] ( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] ( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] ( left ) , @xmath48 ( center ) and a standard wilson operator ( right ) , on a cold configuration ( top ) , on a cooled @xmath102 configuration ( middle ) , and on a non - cooled @xmath104 quenched configuration at @xmath110 ( bottom ) .
the solid blue lines in the top row show analytic results.,title="fig:",scaledwidth=39.0% ] the overall message that can already be drawn from these observations ( before addressing a full - fledged numerical investigation ) is that the gauge field fluctuations in interacting configurations reduce the width of the gap in the spectrum , and blur the distinction between light modes and doublers .
as we showed in the previous section , the fluctuations in typical interacting configurations lead to a filling of the gap in the spectral flow for the various lattice dirac operators that we are considering , making a proper identification of the index difficult .
a related effect can also be seen directly in the spectra of the operators : the panels in fig . [ adams_vs_hoelbling ]
show a comparison of the spectra of @xmath49 ( top row ) and @xmath48 ( bottom row ) in the free case ( left ) , and in interacting configurations at @xmath110 ( central panels , in which different values of @xmath108 or @xmath51 are used ) and at @xmath112 ( right ) .
the figure shows evidence for the superior robustness of lattice fermions based on the @xmath48 operator , over @xmath49 : at @xmath112 for example , a gap remains clearly visible for @xmath48 , while it has all but disappeared for @xmath49 .
this can be understood from the fact that , since @xmath49 involves 4-link parallel transporters , it is more sensitive to the gauge field fluctuations in interacting configurations than @xmath48 which involves 2-link transporters only . however , for practical applications in large - scale simulations , it is important to remark that , as usual , the effect of gauge fluctuations can be considerably reduced through some suitably optimized smearing procedure .
next , we considered the effectiveness of these operators for spectroscopy calculations . to this end
, we performed a simple test , by studying the mass @xmath113 of the lightest meson in the pseudoscalar channel ( the pion ) .
we computed the quark propagator @xmath114 from a point source , on quenched configurations at @xmath110 on a lattice of size @xmath101 , then we evaluated the @xmath115 correlation function @xmath116 and extracted @xmath117 searching for the large - time plateau in the effective mass plot , as a function of @xmath118 .
monitoring the behavior of @xmath119 as a function of @xmath120 , one can study the partially conserved axial current and the issues related to mass renormalization .
[ pion_free_conf ] shows the correlators obtained on a free configuration , for @xmath50 ( left panel ) , for @xmath48 ( central panel ) and for @xmath49 ( right panel ) .
as expected , the @xmath48 operator leads to a massless pion for both @xmath121 and @xmath122 . for an interacting configuration ( at @xmath110 ) , the comparison between @xmath48 and @xmath49 shown in fig .
[ pion_interacting_conf ] reveals that for @xmath48 one obtains a massless pion at approximately @xmath123 , while for @xmath49 the same happens for @xmath124 . comparing these numbers with the values of the bare masses corresponding to a massless pseudoscalar state in the free limit ( 1 and 2 respectively ) , these results give an indication that the mass renormalization is more pronounced for @xmath49 than for @xmath48 .
quantitatively , one can observe that the renormalization factor grows exponentially with the length of the parallel transporters used : @xmath125 , in agreement with the fact that @xmath49 involves 4-link terms , as opposed to @xmath48 , in which the mass term is constructed from 2- link terms .
remarkably , with the @xmath48 operator , the pion mass shows a square - root behaviour of three different kinds : one can approach the critical bare quark mass @xmath126 from the left or from the right , i.e. from the inside or the outside of the @xmath48 spectrum ( the behaviour is square - root - like even though the theory describes one flavour only
it is caused by the approach to the aoki phase ) .
in addition , one can also approach the other critical quark mass @xmath127 , corresponding to the central branch of the spectrum , which remains zero as in the free case by symmetry of the average spectrum .
the transition from one branch to another seems rather abrupt , and the scaling of the pion mass can be observed over a broad range of quark masses approaching zero .
the lesson is that @xmath48 may provide a cost - effective way to simulate @xmath128 light quark species , _ without _ fine - tuning of the bare quark mass to approach the chiral limit .
in this work , we performed a numerical study of staggered dirac operators with a taste - dependent mass term .
we restricted our attention to operators including mass terms with tensor or pseudoscalar structure in taste space : their @xmath33-hermiticity properties are such , that their eigenvalues come in complex conjugate pairs ( as is the case for the usual staggered dirac operator ) , leading to a real fermionic determinant , which is non - negative in the absence of negative real eigenvalues .
such operators were proposed by adams @xcite and by hoelbling @xcite .
we compared their properties both in the free limit and on interacting configurations at typical values of the gauge coupling .
our results show that these operators can indeed be used to separate the low - lying modes and reduce the number of tastes , in a way characterized by well - defined topological properties .
our study of the spectral flow reveals that , for the 4-link operator ( with a taste - pseudoscalar mass term ) , the gap in the eigenvalue spectrum tends to close rather early , obstructing an easy identification of the eigenvalue crossings , which are related to the index .
as one might have expected , the 2-link operator shows markedly more robustness to gauge fluctuations .
we also performed an elementary study of pion propagators , which shows that the lightest meson is rather easy to isolate without explicitly disentangling spin and taste degrees of freedom .
approaching the chiral limit requires in general the fine - tuning of an additive mass term , as for wilson fermions .
one important exception occurs for the 2-link operator : if one chooses the middle branch of the spectrum , one can study a theory with two tastes , where the additive mass renormalization vanishes due to the symmetry of the spectrum .
therefore , no fine - tuning is needed .
although the 2-link operator was designed to produce a single taste ( with a fine - tuned additive mass ) , it may well be that its most promising use is to simulate two tastes without additive mass renormalization .
note that the heavy doubler modes do not completely decouple in that situation . in the background of a topological charge @xmath129 , they contribute real eigenvalues @xmath130 and @xmath131 , making the determinant negative when @xmath129 is odd .
the @xmath132-parameter is thus equal to @xmath133 .
this sign @xmath134 should be removed by hand ( or simply ignored ) in order to simulate the @xmath135 theory . finally , we studied the properties of an overlap operator with a @xmath49 kernel ( see appendix ) .
we found that its locality properties are similar to those of the operator based on a wilson kernel .
as it concerns the computational cost for a quark propagator calculation , we found that , in the free limit or on very smooth gauge configurations , the inversion of the operator based on a kernel with a four - link mass term is almost one order of magnitude faster than using an overlap with wilson kernel .
however , we also observed a significant loss of efficiency on interacting ( quenched ) configurations at @xmath110 , where the operator with the @xmath49 kernel is only approximately twice as fast as that with a wilson kernel .
the reason for this can probably be traced back to the fact that the four - link transporters in the mass term are more sensitive to the effect of the fluctuations in gauge configurations on coarser lattices .
our crude assessment indicates that this new , staggered , overlap operator does not bring a major computational advantage over a wilson kernel , while producing two degenerate flavors , but without the full @xmath136 flavor symmetry
. two copies of an overlap operator with a kernel based on hoelbling s 2-link operator would give more flexibility , e.g. that of simulating two flavors with unequal masses , for a similar computer effort .
* note added : * after this paper was completed , a difficulty with the hoelbling operator @xmath48 eq.([twolinkkernel ] ) was pointed out by steve sharpe , and clarified by david adams , during the yukawa institute workshop `` new types of fermions on the lattice '' .
it appears that the hoelbling operator lacks sufficient rotational symmetry , so that fine - tuned wilson loop counterterms will presumably be needed to maintain hypercubic rotational symmetry in unquenched simulations .
adams operator @xmath49 eq.([fourlinkkernel ] ) does not suffer from this problem .
this research was supported by the natural sciences and engineering research council of canada , by the academy of finland , project 1134018 , and in part by the national science foundation under grant no .
phy11 - 25915 .
de f. thanks the yukawa institute for theoretical physics , kyoto , japan , for hospitality .
ph . de f. and m. p. gratefully acknowledge the kavli institute for theoretical physics in santa barbara , usa , for support and hospitality during the `` novel numerical methods for strongly coupled quantum field theory and quantum gravity '' program , during which part of this work was done .
we thank d. h. adams , m. creutz , s. drr , c. hoelbling , s. kim , t. kimura , t. misumi , a. ohnishi , s. sharpe and all participants of the yukawa institute workshop `` new types of fermions on the lattice '' for discussions .
we also studied the properties of an overlap operator based on a staggered @xmath49 kernel , as originally proposed in @xcite .
the construction is completely standard : @xmath137 and leads to two exactly massless physical fermions in the continuum limit , without fine - tuning . as compared to a conventional overlap operator based on a wilson kernel , the potential advantages of this construction are related to the reduced kernel size ( @xmath49 is a matrix of size four times smaller than a wilson operator on the same lattice ) .
we take @xmath138 . to understand the effectiveness of an overlap operator with a @xmath49 kernel ,
the first important issue to be discussed is the locality of the operator . as is well - known , an overlap operator is not ultra - local @xcite .
its locality properties can be studied by looking at the decay of its matrix elements between source and sink at sites @xmath139 and @xmath140 ( which we denote as @xmath141 , where , for simplicity , we only show the indices corresponding to the site coordinates ) , as a function of the distance between @xmath139 and @xmath140 @xcite .
to this end , in the two plots at the top of fig .
8 we show the decay of @xmath142 against @xmath143 , the 1-norm distance ( or `` manhattan distance '' ) between the sites @xmath139 and @xmath140 , comparing the matrix elements of an overlap operator obtained using a @xmath49 kernel ( left panel ) or a conventional wilson kernel ( right panel ) . although the @xmath49 kernel is less local than the wilson kernel , the locality properties of the corresponding overlap operators are comparable .
this appears quite clearly in the plots displayed at the bottom of the figure , in which the results for the two operators are shown together , for a cold configuration ( left panel ) and for a configuration at @xmath110 ( right panel ) .
another important factor in the efficiency of a lattice dirac operator is the cost of applying the operator to a vector .
the multiplication by the kernel is about twice as fast , if one uses @xmath49 instead of a wilson kernel ( staggered fields do not have an explicit spinor index but @xmath49 has twice as many non - zero elements as the wilson operator ) . in the computation of the sign of @xmath144 , using the conjugate gradient ( cg ) method , and no deflation , the gain with respect to a conventional wilson kernel is a factor ranging from approximately 2 - 3 to about 8 .
however , these numbers are only gross estimates , and could be improved , e.g. , by optimizing the parameters of the @xmath49 kernel . similarly , an improved form for the kinetic operator , link smearing ( for the kinetic and/or the mass term ) , and standard tricks related to deflation , preconditioning , etc ... could be applied .
to discuss the computational cost of the inversion of the operator , we compared the two overlap operators on the same pure - glue , @xmath110 , background on a @xmath106 lattice , using the same , basic , inner / outer cg algorithm . in our computation , we evaluated the propagator as the solution of the equation : @xmath145 with @xmath146 , using a conjugate gradient ( cg ) iterative solver : at each iteration , @xmath147 is applied to a vector @xmath148 through a 2-pass lanczos process .
one builds a tridiagonal matrix @xmath149 and takes the signs of its eigenvalues ( which are representative of those of @xmath89 ) , then reconstructs @xmath150 , as described in ref .
the results are displayed at the top of fig .
8 : the three plots ( from left to right ) show the computational cost for the outer cg iteration , for the matrix - times - vector multiplication , and the total cpu cost . for comparison , we also show the analogous results in the free - field case , in the plots at the bottom of the figure .
this comparison shows that the computational advantages expected from elementary arguments , and observed in the free limit , turn out to be dramatically reduced in `` realistic '' interacting configurations .
again , this reduction points to a reduction in the eigenvalue gap of @xmath49 .
d. h. adams , phys .
* 104 * ( 2010 ) 141602 .
[ arxiv:0912.2850 [ hep - lat ] ] .
d. h. adams , phys .
b * 699 * ( 2011 ) 394 [ arxiv:1008.2833 [ hep - lat ] ] .
c. hoelbling , phys .
* b696 * ( 2011 ) 422 .
[ arxiv:1009.5362 [ hep - lat ] ] .
h. b. nielsen , m. ninomiya , phys . lett .
* b105 * ( 1981 ) 219 ; nucl .
* b185 * ( 1981 ) 20 ; [ erratum - ibid .
* b195 * ( 1982 ) 541 ] .
* b193 * ( 1981 ) 173 .
d. friedan , commun .
* 85 * ( 1982 ) 481 .
m. lscher , phys .
* b428 * ( 1998 ) 342 [ hep - lat/9802011 ] .
h. ginsparg , k. g. wilson , phys .
rev . * d25 * ( 1982 ) 2649 . d. b. kaplan , phys .
lett . * b288 * ( 1992 ) 342 [ hep - lat/9206013 ] . h. neuberger , phys .
* b417 * ( 1998 ) 141 [ hep - lat/9707022 ] .
p. hasenfratz , v. laliena , f. niedermayer , phys .
lett . *
b427 * ( 1998 ) 125 [ hep - lat/9801021 ] . k. g. wilson , in _ new phenomena in subnuclear physics _
a. zichichi ( plenum press , 1977 ) .
j. b. kogut , l. susskind , phys .
rev . * d11 * ( 1975 ) 395 .
e. follana _ et al .
_ [ hpqcd and ukqcd collaborations ] , phys .
d * 72 * ( 2005 ) 054501 [ hep - lat/0507011 ] .
s. durr and c. hoelbling , phys .
d * 69 * ( 2004 ) 034503 [ hep - lat/0311002 ] ; s. durr , c. hoelbling and u. wenger , phys .
d * 70 * ( 2004 ) 094502 [ hep - lat/0406027 ] ; f. bruckmann _ et al .
_ , phys .
d * 78 * ( 2008 ) 034503 [ arxiv:0804.3929 [ hep - lat ] ] ; pos lat * 2007 * ( 2007 ) 274 [ arxiv:0802.0662 [ hep - lat ] ] .
m. creutz , pos confinement * 8 * ( 2008 ) 016 [ arxiv:0810.4526 [ hep - lat ] ] . g. c. donald , c. t. h. davies , e. follana and a. s. kronfeld , phys .
d * 84 * ( 2011 ) 054504 [ arxiv:1106.2412 [ hep - lat ] ] .
de forcrand , a. kurkela , m. panero , pos * lattice2010 * ( 2010 ) 080 [ arxiv:1102.1000 [ hep - lat ] ] ; see also ` http://super.bu.edu/~brower/qcdna6/talks/deforcrand.pdf ` .
l. h. karsten , phys .
b * 104 * ( 1981 ) 315 .
f. wilczek , phys .
lett . * 59 * ( 1987 ) 2397
. m. creutz , jhep * 0804 * ( 2008 ) 017 [ arxiv:0712.1201 [ hep - lat ] ] .
a. borii , phys .
d * 78 * ( 2008 ) 074504 [ arxiv:0712.4401 [ hep - lat ] ] .
s. capitani _ et al .
_ , jhep * 1009 * ( 2010 ) 027 [ arxiv:1006.2009 [ hep - lat ] ] . m. creutz , t. kimura and t. misumi , jhep * 1012 * ( 2010 ) 041 [ arxiv:1011.0761 [ hep - lat ] ] ; pos lattice * 2011 * ( 2011 ) 106 [ arxiv:1110.2482 [ hep - lat ] ] ; t. kimura , s. komatsu , t. misumi , t. noumi , s. torii and s. aoki , jhep * 1201 * ( 2012 ) 048 [ arxiv:1111.0402 [ hep - lat ] ] .
e. follana , v. azcoiti , g. di carlo and a. vaquero , pos lattice * 2011 * ( 2011 ) 100 [ arxiv:1111.3502 [ hep - lat ] ] . h. s. sharatchandra , h. j. thun , p. weisz , nucl . phys .
* b192 * ( 1981 ) 205 .
f. gliozzi , nucl .
* b204 * ( 1982 ) 419 .
c. van den doel , j. smit , nucl .
* b228 * ( 1983 ) 122 .
h. kluberg - stern _ et al . _ ,
* b220 * ( 1983 ) 447 . m. f. l. golterman , j. smit , nucl .
phys . * b245 * ( 1984 ) 61 .
de forcrand , pos lat * 2009 * ( 2009 ) 010 [ arxiv:1005.0539 [ hep - lat ] ] .
d. h. adams , pos lattice * 2010 * ( 2010 ) 073 [ arxiv:1103.6191 [ hep - lat ] ] .
p. de forcrand , m. garca prez , j. e. hetrick , e. laermann , j. f. lagae , i. o. stamatescu , nucl .
* 73 * ( 1999 ) 578 .
[ hep - lat/9810033 ] .
i. horvth , phys .
* 81 * ( 1998 ) 4063 .
[ hep - lat/9808002 ] .
p. hernndez , k. jansen , m. lscher , nucl .
phys . *
b552 * ( 1999 ) 363 .
[ hep - lat/9808010 ] .
w. bietenholz , fortsch .
* 56 * ( 2008 ) 107 .
[ hep - lat/0611030 ] .
s. drr , g. koutsou , [ arxiv:1012.3615 [ hep - lat ] ] .
a. borii , phys .
* b453 * ( 1999 ) 46 .
[ hep - lat/9810064 ] . | the numerical properties of staggered dirac operators with a taste - dependent mass term proposed by adams @xcite and by hoelbling @xcite are compared with those of ordinary staggered and wilson dirac operators . in the free limit and on ( quenched ) interacting configurations ,
we consider their topological properties , their spectrum , and the resulting pion mass .
although we also consider the spectral structure , topological properties , locality , and computational cost of an overlap operator with a staggered kernel , we call attention to the possibility of using the adams and hoelbling operators without the overlap construction . in particular , the hoelbling operator could be used to simulate two degenerate flavors without additive mass renormalization , and thus without fine - tuning in the chiral limit .
cern - ph - th/2012 - 040 + hip-2012 - 06/th + nsf - kitp-12 - 017 + yitp-12 - 7 * numerical properties of staggered quarks + with a taste - dependent mass term * 1.3 cm philippe de forcrand,@xmath0 aleksi kurkela@xmath1 and marco panero@xmath2 1.5 cm _ @xmath3 institute for theoretical physics , eth zrich , ch-8093 zrich , switzerland _ 0.5 cm _ @xmath4 cern , physics department , th unit , ch-1211 genve 23 , switzerland _ 0.5 cm _ @xmath5 kavli institute for theoretical physics , university of california , santa barbara , ca 93106 , usa _ 0.5 cm _
@xmath6 yukawa institute for theoretical physics , kyoto university , kyoto 606 - 8502 , japan _ 0.5 cm _ @xmath7 department of physics , mcgill university , 3600 rue university , montral , qc h3a 2t8 , canada _ 0.5 cm _ @xmath8 department of physics and helsinki institute of physics , university of helsinki , fin-00014 helsinki , finland _ 0.5 cm _ e - mail : _ ` [email protected] , [email protected] , [email protected] ` 1.0 cm |
let @xmath10 : a\in { \mathbb q}/{\mathbb z}$ ] } be a basis for a free abelian group @xmath11 . then the ( dimension @xmath12 ) universal ordinary distribution @xmath2 of level @xmath0 for any positive integer @xmath0 is given by @xmath13 : a\in \frac{1}{r}{\mathbb z}/{\mathbb z}>}{<[a]-{\displaystyle \sum_{\ell b = a } } [ b ] : \
\ell |r\ \text{prime } ,
a\in \frac{\ell}{r}{\mathbb z}/{\mathbb z}>}.\ ] ] for any @xmath14 , if @xmath15 , set @xmath16)=[xa]$ ] . by this action , @xmath2 becomes a @xmath17-module .
the universal distribution is well known to be a free abelian group , moreover , for any integer @xmath7 , then the natural map from @xmath18 to @xmath2 is a split monomorphism and thus @xmath18 can be considered as a submodule of @xmath2 .
the theory of the universal distribution plays an important role in the theory of cyclotomic fields .
detailed information can be found in the well - known textbooks by lang @xcite and washington @xcite .
most notably , kubert @xcite and @xcite , and sinnott @xcite studied the @xmath19-cohomology of @xmath2 , from which , sinnott got his famous index formula about the cyclotomic units and the stickelberger elements .
recently , anderson @xcite found a brand new way to compute the @xmath19-cohomology of the universal distribution @xmath2 .
he discovered a cochain complex which is a resolution of the universal distribution . to study a certain group cohomology of @xmath2 ,
it is therefore translated to study a double complex related to this group cohomology . in this paper , we use anderson s resolution to construct a double complex related to the @xmath17-cohomology of @xmath2 and study the spectral sequence of the double complex .
suppose that @xmath0 is fixed odd squarefree integer , we prove the following theorem : the cohomology group @xmath20 where @xmath21number of prime factors of @xmath22 and @xmath23 we shall discuss the unabridged form in @xmath24 .
what s more , for any positive integer @xmath25 which is a common factor of @xmath3 over all prime factors @xmath4 of @xmath0 , let @xmath26 be a generator of the cyclic group @xmath27 and let @xmath28 then [ theob ] the image of the family @xmath29 : \forall\ r'|r \right \}\ ] ] in @xmath30 is a @xmath8-basis for @xmath9 .
theorem b has interesting application in arithmetic .
we follow the line given in rubin @xcite .
let @xmath31 be the maximal real subfield of @xmath32 , assume @xmath33 is a family of distinct odd primes which split completely in @xmath34 and are @xmath35 for fixed integer @xmath25 .
suppose that we have a @xmath17 -homomorphism @xmath36 from @xmath2 to @xmath37 .
then @xmath36 induces a map @xmath38 for each @xmath39 . in the case @xmath40 , since @xmath41 , we have the map @xmath42 in particular , let @xmath43 be the abelian closure of @xmath44 .
let @xmath45 be an injective homomorphism from @xmath46 to @xmath47 . put @xmath48)=(\mathbf e(a+\frac{1}{m})-1)(\mathbf e(a-\frac{1}{m})-1).\ ] ] then @xmath36 is a @xmath17-homomorphism from @xmath2 to @xmath37 .
the image @xmath49 ) $ ] is just the kolyvagin element @xmath6 as given in @xcite . from this point of view
, we can regard the euler system as a system in the cohomology group @xmath9 .
this is the initial motivation for this paper .
this paper is organized in the following order .
we give general notations in @xmath50 . in @xmath51 , we study anderson s resolution in detail . in @xmath52 , a special @xmath17-projective resolution @xmath53 of @xmath54 is constructed and the group cohomology of @xmath54 and of @xmath8 are given . with this projective resolution @xmath53 , we construct a double complex in @xmath55 whose total cohomology is the @xmath17-cohomology of @xmath2 .
the standard spectral sequence method is then used to compute the cohomology group @xmath56 . in @xmath24 , we study the lifting problem and prove theorem b. 0.3 cm * acknowledgment . * i thank my advisor professor greg w. anderson for introducing me to this interesting field , for his insight and insistence driving me to this paper .
the double complex method is the brain child of professor anderson .
i also thank dr .
hans uli walther for some useful discussion about lemma [ la2 ] .
fix a finite set @xmath57 of cardinality @xmath58 . fix a family @xmath59 of distinct odd prime numbers . fix a positive integer @xmath25 dividing @xmath60 for all @xmath61 . fix a total order @xmath62 of @xmath57 .
put * @xmath63 , * @xmath64 . for each @xmath61 , put * @xmath65:=the inertia subgroup of @xmath66 at @xmath67 , * @xmath68:=a fixed generator of @xmath65 , * @xmath69 , @xmath70 , * @xmath71:=the arithmetic frobenius automorphism at @xmath67 in @xmath72 . for each subset @xmath73 ,
put * @xmath74 , @xmath75 , * @xmath76 , * @xmath77 , @xmath78 .
put @xmath79 $ ] . for any element @xmath80 ,
put * @xmath81 , * @xmath82 .
* @xmath83 where @xmath84 . for any @xmath85 ,
put @xmath86 . for @xmath87 ,
the order of @xmath88(denoted by @xmath89 ) means its order in @xmath46 . for any set @xmath90
, the cardinality of @xmath90 is denoted by @xmath91 , the free abelian group generated by @xmath90 is denoted by @xmath92 , the free @xmath8-module generated by @xmath90 is denoted by @xmath93 .
the family of all subsets of @xmath90 is denoted by @xmath94 .
we call a subfamily @xmath95 of @xmath94 an @xmath96 of @xmath90 if for all @xmath97 , @xmath98 . for any pair of sets @xmath90 and @xmath99 ,
the difference of @xmath90 and @xmath99 is denoted by @xmath100 . for any complex @xmath101 ,
the complex @xmath102 $ ] is the complex with components @xmath103=c^{m+n}$ ] . for any complex @xmath101 of @xmath54-modules , @xmath104 .
let @xmath10 : a\in { \mathbb q}/{\mathbb z}\}$ ] be a basis of a free abelian group @xmath11 .
recall by kubert @xcite , the ( rank @xmath12 ) universal ordinary distribution @xmath105 is given by @xmath106 : a\in { \mathbb q}/{\mathbb z}>}{<[a]-{\displaystyle\sum_{n b = a } } [ b]:\ n\in { \mathbb n}>}.\ ] ] for any positive number @xmath107 , the universal ordinary distribution of level @xmath107 is given by @xmath108 : a\in \frac{1}{f}{\mathbb z}/{\mathbb z}>}{<[a]-{\displaystyle\sum_{p b = a } } [ b ] : p\mid f , a\in \frac{p}{f}{\mathbb z}/{\mathbb z}>}.\ ] ] for any @xmath109 , set @xmath16)=[xa]$ ] if @xmath110 sends each @xmath107-th root of unity to its @xmath111-th power . by this action , @xmath112 is a @xmath113-module .
much has been studied about the structures of @xmath105 and @xmath112 , we list some basic properties here(for detailed proof , see anderson @xcite , kubert @xcite and washington @xcite ) .
first recall for any @xmath114 , @xmath88 can be written uniquely @xmath115 then [ fact ] 1 ) .
the universal ordinary distribution @xmath112 is a free abelian group of rank @xmath116 , the set @xmath10 : a\in \frac{1}{f}{\mathbb z}/{\mathbb z } , \ a_{p1 } \neq p-1,\ \forall p\mid f\}$ ] is a @xmath54-basis for @xmath112 . 2 ) .
for any factor @xmath117 of @xmath107 , the natural map from @xmath118 to @xmath112 is a split monomorphism .
moreover , by this natural map , @xmath105 is the direct limit of @xmath112 for @xmath119 and thus @xmath105 is free . in the sequel , for our convenience
, the universal distribution @xmath2 will be written as @xmath120 and @xmath121 as @xmath122 .
now let @xmath123 : a\in \frac{r_t}{r_s}{\mathbb z}/{\mathbb z } , t\subseteq s>\ ] ] be the free abelian group generated by the symbols @xmath124 $ ] , and let @xmath125 : |t|=-p , a\in \frac{r_t}{r_s}{\mathbb z}/{\mathbb z } , t\subseteq s > , \ ] ] then @xmath126 is a bounded graded module . furthermore , for any @xmath127 , set @xmath128=[xa , t]$ ] if @xmath110 sends each @xmath0-th root of unity to its @xmath111-th power . by this action , @xmath126 becomes a @xmath66-module .
let @xmath129=\sum_{i\in t}\omega(i , t ) ( [ a , t\backslash \{i\}]- \sum_{\ell_i b = a}[b , t\backslash \{i\}]).\ ] ] where @xmath130 it is easy to check that @xmath131 and @xmath132 is @xmath66-equivariant . thus @xmath133 is a cochain complex .
note that the definition of @xmath132 depends on @xmath62 .
we ll write @xmath134 instead if we need to emphasize the order @xmath62 .
the following proposition is given by anderson : [ anderson ] the @xmath135-th cohomology of the complex @xmath133 is @xmath136 for @xmath137 and @xmath120 for @xmath40 , furthermore , the map from @xmath138 to @xmath120 is given by @xmath139\mapsto [ a]$ ] .
the above proposition(in a more general form suitable for a resolution for the distribution @xmath112 ) , though known by anderson for quite a while , has no published proof by now .
we put the proof in appendix a , traces of the idea behind the proof can be found in anderson @xcite .
\2 . for the sake of this proposition , we call @xmath140 anderson s resolution of the universal distribution @xmath120 .
this resolution has been used by das @xcite in his work about the algebraic @xmath141-monomials and double covering of cyclotomic fields . a remarkable fact about anderson s resolution
@xmath126 is that it possesses an even more delicate double complex structure , which in turn gives a natural filtration for the universal distribution @xmath120 .
we start with a more careful look at @xmath126 , which we ll denote by @xmath142 instead . for any @xmath73
, we always regard @xmath143 as a subcomplex of @xmath142 .
moreover , for any order ideal @xmath95 of @xmath57 , put @xmath144 in particular , let @xmath145 be the order ideal consisting of all subsets @xmath146 such that @xmath147 , and let @xmath148 and @xmath149 .
note that @xmath150 .
for any @xmath151 , let @xmath152 we see that @xmath153 : \operatorname{\rm supp}a\cap t=\emptyset>.\ ] ] then @xmath154 is the free abelian group generated by @xmath155 : t\cup \operatorname{\rm supp}a\in { \mathcal i } ,
\ t\cap \operatorname{\rm supp}a=\emptyset \right \},\ ] ] and @xmath156 is the free abelian group generated by @xmath157 : \operatorname{\rm supp}a\in { \mathcal i},\ a_{\ell_i 1}\neq \ell_i-1 \ \text{for all } \ i\in s\right \}.\ ] ] immediately we have [ intersection ] let @xmath158 and @xmath159 be two order ideals of @xmath57 , then 1 ) .
@xmath160 , @xmath161 . 2 ) .
@xmath162 , @xmath163 .
then [ exact ] the complex @xmath164 is acyclic with the @xmath136-cohomology @xmath165 .
we let @xmath166 be the complex @xmath167 hence it suffices to show that @xmath166 is exact .
let @xmath146 be a maximal element in the order ideal @xmath95 .
let @xmath168 be the order ideal whose maximal element set is the maximal element of @xmath95 excluding @xmath146 , then @xmath169 by proposition [ intersection ] , we have @xmath170 now we prove the proposition by induction on the cardinality of maximal elements of @xmath95 . if @xmath95 has only one maximal element , this is just proposition [ anderson ] .
in general , both @xmath168 and @xmath171 have less maximal elements than @xmath95 has .
thus the exactness of @xmath172 follows from the exactness of the three complexes @xmath173 , @xmath174 and @xmath175 .
now we can construct a double complex whose total single complex is @xmath176 . with abuse of notation ,
we ll write it as @xmath177 . for any pair of subsets @xmath178 , @xmath146 of @xmath57 such that @xmath179 , set @xmath180 : \operatorname{\rm supp}a={s\backslash t'}>,\ ] ] then @xmath181 is isomorphic to @xmath182
. moreover , for any @xmath183 , the map @xmath184\mapsto [ a , t\backslash \{i\}]\ ] ] defines a natural isomorphism between @xmath181 and @xmath185 .
now for any @xmath73 , @xmath186 and if let @xmath187 , then @xmath188 in general for any @xmath61 , define @xmath189\mapsto \chi_t(i ) [ a , t\backslash \{i\}]\ ] ] where @xmath190 is the characteristic function of @xmath146 .
let @xmath191 be the subgroup of @xmath192 generated by @xmath193 for all @xmath61 , inductively , let @xmath194 be the subgroup of @xmath195 generated by @xmath196 for all @xmath61 . by this setup
, there is a filtration of @xmath197 given by @xmath198 this filtration enables us to define the double complex structure of @xmath142 . for the element @xmath124\in { \mathbf l}^{\bullet}$ ] , we say @xmath199 $ ] is of bidegree @xmath200 if @xmath201\in \varphi^{p_2}(l^{p_1})\backslash
\varphi^{p_2 + 1}(l^{p_1 - 1})$ ] , more explicitly , if @xmath202 then we see that all elements of @xmath181 are of bidegree @xmath203 .
let @xmath204 be the subgroup of @xmath142 generated by all symbols @xmath124 $ ] with bidegree @xmath205 , then @xmath206 set @xmath207\mapsto -\sum_{i\in t } \omega(i , t ) n_i[fr^{-1}_i a + \frac{1}{\ell_i } , t\backslash \{i\ } ] , \ ] ] @xmath208\mapsto \sum_{i\in t } \omega(i , t)(1-fr^{-1}_i ) [ a , t\backslash \{i\ } ] .\ ] ] it is easy to check that @xmath209 . hence we construct a double complex @xmath210 .
note that @xmath211 and @xmath212 thus @xmath213 is the single total complex of the double complex @xmath210 , with the second filtration given by @xmath214 .
thus the total cohomology of @xmath210 is the cohomology of @xmath213 .
the @xmath215 term of the spectral sequence arising from the double complex @xmath210 by the first filtration(i.e . , @xmath216 ) is @xmath217 thus the spectral sequence for the first filtration degenerates at @xmath215 .
note that @xmath218 then it is easy to see that @xmath219 $ ] is nothing but the quotient complex @xmath220 .
the short exact sequence @xmath221 \longrightarrow 0\ ] ] induces a long exact sequence @xmath222)\rightarrow h^{i+1}({\mathbf l}^{\bullet } ( s-{p_2 - 1 } ) ) \rightarrow \cdots\ ] ] by proposition [ exact ] , for @xmath223 and @xmath224 , both @xmath225 and @xmath226 are @xmath136 , so is @xmath227)$ ] .
therefore the above long exact sequence is just the exact sequence @xmath228)\rightarrow u_{s } ( s-{p_2 - 1 } ) \rightarrow u_{s } ( s-{p_2 } ) \rightarrow h^{0}({\mathbf l}^{\bullet , p_2}[-p_2 ] ) \rightarrow 0,\ ] ] since the map from @xmath229 to @xmath230 is injective , the proposition follows immediately .
the following results follow immediately from proposition [ fact ] , proposition [ anderson ] and proposition [ exact ] : 1 ) .
one has @xmath231 2 ) .
moreover , for any order ideal @xmath95 of @xmath57 , one has @xmath232
we first have a convention here : let @xmath90 be a finite totally ordered set and @xmath235 .
suppose @xmath236 is a module related to @xmath111 .
we call @xmath237 the standard tensor product of @xmath236 over @xmath90 if @xmath238 and @xmath239 .
similarly , we can define the standard tensor product of elements @xmath240 and of complexes @xmath241 .
let @xmath242 with @xmath243 $ ] for any @xmath244 , @xmath245 is the multiplication by @xmath246 if @xmath247 is even and by @xmath248 if @xmath247 is odd .
it is well known that @xmath249 is a @xmath250$]-projective resolution of trivial module @xmath54 . for any @xmath73 ,
let @xmath251 be the standard tensor product of @xmath249 over @xmath183 .
it is well known by homological algebra that @xmath251 is a @xmath252$]-projective resolution of trivial module @xmath54 .
now for the collection @xmath253 , the standard product of @xmath254 over @xmath146 is a rank @xmath12 free @xmath252$]-module whose grade is @xmath255 .
now let @xmath256 be the element whose @xmath257-th component is @xmath258 if @xmath183 and @xmath136 if not , and write the standard product of @xmath254 over @xmath146 as @xmath252[e]$ ] , then @xmath259[e].\ ] ] for any @xmath260[e]$ ] , the differential is given by @xmath261 in particular , for @xmath262 .
let @xmath263 [ e].\ ] ] for any @xmath264 , we have a natural inclusion @xmath265[e ] \hookrightarrow { \mathbb z}[g_t][e]$ ] for any @xmath256 such that @xmath266 . by this inclusion
, @xmath267 becomes a subcomplex of @xmath251 .
now we define a diagonal map @xmath268 .
first set @xmath269 then the map @xmath270 given by @xmath271 is the diagonal map for the cyclic group @xmath65 ( see cartan - eilenberg @xcite , p250 - 252 ) . for any @xmath272 with support contained in @xmath146 , consider the standard product @xmath273 of @xmath274 over @xmath183 .
the isomorphism @xmath275 induces an isomorphism @xmath276[e]\otimes { \mathbb z}[g_t][e']$ ] by @xmath277 on the other hand , the standard product of the diagonal maps @xmath271 over @xmath183 defines a map @xmath278[e+e']\rightarrow p_{e , e'}$ ] .
we let @xmath279 and let @xmath280 then @xmath281 defines the diagonal map from @xmath251 to @xmath282 .
this map enables us to compute the cup product structures .
let @xmath285 , then @xmath286 is the complex @xmath287 with the initial term at degree @xmath136 .
we denote by @xmath288 the @xmath247-th term of @xmath286 . by the theory of group cohomology , @xmath289 now for any @xmath73 , let @xmath290 be the standard tensor product of @xmath286 for @xmath183 . if write @xmath291[e ] , { \mathbb z})={\mathbb z}[e],\ ] ] then @xmath292.\ ] ] and @xmath293 moreover , for any @xmath264 , the inclusion @xmath294 induces a map @xmath295 which is just the natural projection of @xmath296\longrightarrow { \displaystyle \bigoplus_{\operatorname{\rm supp}e\subseteq t ' } } { \mathbb z}[e].\ ] ] on the other hand , @xmath297 can also be considered naturally as a quotient group of @xmath298 , by this meaning , the inflation map is just the injection @xmath299 \hookrightarrow \bigoplus_{\operatorname{\rm supp}e\subseteq t } { \mathbb z}[e].\ ] ] now for any @xmath300 even , let @xmath301 for any @xmath302 , i.e. , @xmath258 even for all @xmath61 , we let @xmath303 be the standard product @xmath304 over @xmath61 .
if @xmath305 , then @xmath303 is a subcomplex of @xmath290 and @xmath306 figure 1 shows us how the decomposition looks like in the case @xmath307 .
denote by @xmath308 the cohomology group @xmath309 and @xmath310 its @xmath135-th component . then @xmath311 when @xmath307 . ]
we now study the abelian group @xmath308 .
first we need a lemma from linear algebra : [ la1 ] let @xmath312 be a @xmath135-dimensional column vector with integer entries @xmath313 , then the greatest common divisor of @xmath313 is @xmath12 if and only if there exists an @xmath314 matrix @xmath315 whose first column is @xmath316 .
now suppose @xmath317 and @xmath318 . if @xmath319 , then @xmath320 , it is easy to see that @xmath321 .
now if @xmath322 , we claim that @xmath323 $ ] is isomorphic to the exterior algebra @xmath324 with differential @xmath325 and @xmath326 .
this claim is easy to check : first if @xmath327 , let @xmath328 , then @xmath329 .
this case is trivial . in general ,
if @xmath330 $ ] is isomorphic to @xmath331 , the tensor product of @xmath330 $ ] is nothing but @xmath323 $ ] and the tensor product of @xmath331 is just @xmath324 , hence they are isomorphic to each other .
now let @xmath332 be the greatest common divisor of @xmath60 for @xmath183 , thus the greatest common divisor of @xmath333 is 1 , let @xmath334 be the matrix given by lemma [ la1 ] corresponding to the vector @xmath335 .
let @xmath336 .
then @xmath337 is a set of new generators for the above exterior algebra and we have @xmath338 .
we see easily that @xmath339 and @xmath340 combining the above analysis , we have [ propa ] there exists a family of complexes @xmath341 such that 1 ) . for each @xmath73 , we can identify @xmath342 with @xmath343 through the following splitting exact sequence @xmath344 2 ) .
the cohomology groups @xmath345 and @xmath346 are given by : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ( a ) . if @xmath347 , let @xmath348 be the greatest common divisor of @xmath60 for @xmath349 , then @xmath308 is the abelian group @xmath350 , and @xmath351 ( b ) .
if @xmath352 , then @xmath321 .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ for the case
@xmath353 , the situation is much easier .
we have [ propb ] there exists a family @xmath354\in h^{\ast}(g_s , { \mathbb z}/m{\mathbb z } ) : e\in r\right \}\ ] ] with the following properties : 1 ) . for each @xmath73 and @xmath39 , the restriction of the family @xmath355 : e\in r , \ \operatorname{\rm
supp}e\subseteq t,\ \deg e = n\right \}\ ] ] to @xmath356 is a @xmath8-basis of the latter . 2 ) .
for each @xmath73 and @xmath256 such that @xmath357 , the restriction of @xmath358 $ ] to @xmath234 vanishes . 3 ) .
one has the cup product structure in @xmath234 given by @xmath359\cup [ e']=(-1)^{\omega(e , e ' ) } \prod_{\substack{i\in s\\e_ie'_i\equiv 1(2 ) } } \left ( \frac{\ell_i-1}{2}\right ) [ e+e']\ ] ] for all @xmath85 .
the complex @xmath360 by definition , is a complex with @xmath361 for @xmath244 and the differential @xmath136 .
in general , @xmath362 is exactly the standard tensor product of @xmath363 for all @xmath183 .
write @xmath364.\ ] ] since now @xmath365 has differential @xmath136 , @xmath366 .
the restriction map is easy to see .
this finishes the proof of 1 ) and 2 ) . for the cup product , the diagonal map @xmath281 given above
naturally induces a map : @xmath367 which defines the cup product structure .
more specifically , the cup product map @xmath368 \times { \mathbb z}/m{\mathbb z}[e ' ] \longrightarrow { \mathbb z}/m{\mathbb z}[e+e']\ ] ] is induced from @xmath369 .
now the claim follows soon from the explicit expression of @xmath369 .
with the preparation in @xmath51 and @xmath52 , set @xmath372 let @xmath132 and @xmath373 be the induced differentials of @xmath132 and @xmath374 by @xmath142 and @xmath375 respectively . if we let @xmath376:=(e\mapsto [ a , t])\in \operatorname{\rm hom}_{g_s}(p_{e } , < [ a , t]>),\ ] ] then @xmath377 : a\in \frac{r_t}{r_s}{\mathbb z}/{\mathbb z } , |t|=-p , \deg e = q>;\\ & d[a , t , e]=\sum_{i\in t } \omega(i , t)([a , t\backslash \{i\ } , e]- \sum_{\ell_i b = a } [ b , t\backslash \{i\ } , e ] ) ; \\ & \delta[a , t , e]=(-1)^{|t| } \sum_{i\in s } ( -1)^{{\omega(e)}_i}\cdot \begin{cases } ( 1-\sigma_i)[a , t , e+\epsilon_i ] , \ & if \ e_i\ even;\\n_i [ a , t , e+\epsilon_i ] , \ & if\ e_i \ odd .
\end{cases } \end{split}\ ] ] for any @xmath73 , set @xmath378 : [ a , t']\in { \mathbf l}^{\bullet}_t , e\in r>\ ] ] and @xmath379 : [ a , t']\in { \mathbf l}^{\bullet}_t , e\in r , \operatorname{\rm supp}e\subseteq t > .\ ] ] furthermore , for any order ideal @xmath95 , set @xmath380 and set @xmath381 set @xmath382 : a\in \frac{1}{r_s}{\mathbb z}/{\mathbb z } , e\in r > } { < [ a , e]-\sum_{\ell_i b = a } [ b , e ] : a\in \frac{\ell_i}{r_s}{\mathbb z}/{\mathbb z } , e\in r>},\ ] ] with the induced differential @xmath373 from @xmath374 . correspondingly , @xmath383 : a\in \frac{1}{r_t}{\mathbb z}/{\mathbb z}\ \text{for some}\ t\in { \mathcal i } , e\in r > } { < [ a , e]-\sum_{\ell_i b = a } [ b , e ] : a\in \frac{\ell_i}{r_t}{\mathbb z}/{\mathbb z}\ \text{for some}\ t\in { \mathcal i } , e\in r>},\ ] ] which is a subcomplex of @xmath384 .
we consider @xmath384 as the double complex @xmath385 concentrated on the vertical axis . from proposition [ anderson
] , we have a map @xmath386\mapsto \begin{cases } [ a , e ] , \ & \text{if}\ t=\emptyset ; \\ 0,\ & \text{if}\ t\neq \emptyset . \end{cases}\ ] ] [ xin ] the map @xmath387(resp .
its restriction ) is a quasi - isomorphism between @xmath388 ( resp .
@xmath389 ) and @xmath390(resp .
@xmath391 ) . therefore 1 ) .
@xmath392 , @xmath393 . 2 ) .
@xmath394 , @xmath395 . immediately from proposition
[ anderson](resp .
proposition [ exact ] for @xmath95 ) , we see that @xmath396 is @xmath132-acyclic , by spectral sequence argument , it is thus @xmath397-acyclic . on the other hand , @xmath387 is surjective .
thus @xmath387 is a quasi - isomorphism . now
1 ) follows directly from the quasi - isomorphism . for 2 ) , just consider @xmath398 , which is also a quasi - isomorphism . from proposition [ xin ] ,
the @xmath66-cohomology of @xmath120 is isomorphic to the total cohomology of the double complex @xmath399 therefore we can use the spectral sequence of the double complex @xmath388 to study the @xmath66-cohomology of @xmath120 .
the spectral sequence of @xmath388 from the second filtration exactly gives us proposition [ xin ] .
now we study the spectral sequence from the first filtration .
then @xmath400 now since @xmath401 then @xmath402 for the double complex @xmath389 , recall @xmath403 in @xmath404 , we have @xmath405 suppose that for any @xmath73 , there is an abelian group @xmath406 related to @xmath146 , set @xmath407 then for any @xmath179 , there is a natural projection from @xmath408 to @xmath409 .
now let @xmath410 be the cochain complex with components given by @xmath411 and differential @xmath132 given by @xmath412 where @xmath413 is the projection of @xmath111 in @xmath414 .
it is easy to verify that @xmath410 is indeed a chain complex .
furthermore , we have [ la2 ] for any @xmath73 , @xmath415 let @xmath416 be the subcomplex of @xmath410 with the same components as @xmath410 except at degree @xmath136 , where @xmath417 we only need to show that @xmath416 is exact .
we show it by double induction to the cardinalities of @xmath57 and @xmath146 .
if @xmath320 , we get a trivial complex . if @xmath57 consists of only one element , or if @xmath146 consists only one element , it is also trivial to verify .
in general , suppose @xmath418 .
let @xmath419 and @xmath420 .
then we have the following commutative diagram which is exact on the columns : @xmath421 here @xmath422 means projection and @xmath257 means inclusion .
the differential @xmath423 is induced by the differential @xmath132 of the second row .
notice that the third row is a variation of the chain complex @xmath424 , the first row is the chain complex @xmath425 . by induction ,
the first row and and the third row are exact , so is the middle one . by @xmath427 , we know that @xmath428 now let s consider the induced differential @xmath429 of @xmath132 in the @xmath215 term . since @xmath211 , we can also write @xmath430 .
we first look @xmath431 , which is induced by @xmath432 & \longmapsto \sum_{i\in t}\omega(i , t ) ( 1-fr^{-1}_i)[a , t\backslash \{i\ } ] \end{split}\ ] ] since for any @xmath183 , @xmath433 and @xmath434 are @xmath66-isomorphic by the map @xmath435 , and since for any @xmath436 , @xmath437 is a trivial @xmath438-module , we have @xmath439 for the map @xmath440 , which is induced by @xmath441 & \longmapsto - \sum_{i\in t}\omega(i , t ) n_i[fr^{-1}_i a+ \frac{1}{\ell_i } , t\backslash \{i\ } ] \end{split}\ ] ] for any @xmath183 , let @xmath442&\longmapsto n_i[fr^{-1}_i a+\frac{1}{\ell_i } , t\backslash \{i\ } ] \end{split}\ ] ] @xmath443 is a @xmath66-homomorphism and therefore induces a map in the @xmath66-cohomology : @xmath444 we have the commutative diagram : @xmath445 where the top row are @xmath66-modules , the left @xmath54 is a trivial @xmath297-module and the right @xmath54 a trivial @xmath446-module , and @xmath447 is the homomorphism sending @xmath448 $ ] to @xmath12 and @xmath449 $ ] to @xmath136 if @xmath450 . then the above diagram induces the following commutative diagram : @xmath451 where @xmath452(and @xmath453 ) is the isomorphism given by shapiro s lemma(see serre @xcite , chap .
vii , @xmath55 , exercise ) .
we identify @xmath454 with @xmath455 , moreover , to keep track on @xmath146 , we ll write @xmath456 as @xmath457 .
then we see that @xmath458 is the restriction map from @xmath457 to @xmath459 .
the induced differential @xmath460 is exactly the map @xmath461 where @xmath462 is the restriction of @xmath111 in @xmath463 .
hence we have a cochain complex @xmath464 @xmath465 note that the complex @xmath466 is just the direct sum of @xmath464 over all subsets @xmath146 of @xmath57 .
moreover , the complex @xmath467 is the direct sum of @xmath468 over all subsets @xmath469 .
recall in proposition [ propa ] , we obtained @xmath470 if let @xmath471 then we have @xmath472 .
the complex @xmath473$]satisfies the conditions in lemma [ la2 ] , thus the @xmath135-th cohomology of the cochain complex @xmath474 is @xmath136 if @xmath475 and @xmath476 if @xmath477 .
we have the following proposition : [ e2 ] one has 1 ) .
@xmath478 . 2 ) .
@xmath479 .
finally we are in a stage to prove theorem a. put @xmath480\in { \mathbf k}^{\bullet,\bullet } , a\neq 0\ \text{if } \
\operatorname{\rm supp}e\supseteq t>.\ ] ] it is easy to verify that @xmath481 is a subcomplex of @xmath388 .
set @xmath482 : \operatorname{\rm supp}e\supseteq t>.\ ] ] note that the induced differential of @xmath132 at @xmath483 is @xmath136 .
moreover , set @xmath484 and @xmath485:\ t\in { \mathcal i } , \
\operatorname{\rm supp}e \supseteq t>.\ ] ] let @xmath107 be the corresponding quotient map , then we have a commutative diagram : @xmath486 we claim that [ theoq ] the quotient map @xmath487 is a quasi - isomorphism . moreover , the quotient map @xmath488 is a quasi - isomorphism .
let @xmath489 : e\in r>=\operatorname{\rm hom}_{g_s}({\mathbf p}_{\bullet } , l_{s , t})\subseteq { \mathbf k}^{\bullet , \bullet}\ ] ] and let @xmath490 : \operatorname{\rm supp}e\supseteq t>,\ { \mathcal l^ { '' \bullet}}_t:=<[0 , t , e ] : \operatorname{\rm supp}e\subseteq s\backslash\{i\},\ \text{for some}\ i\in t>\ ] ] through the map @xmath491\mapsto 1 $ ] , we have a commutative diagram @xmath492 where @xmath493 and @xmath494 are given in proposition [ propa ] . by this diagram
, we identify @xmath495 with @xmath493 . by proposition [ propa ]
, we have @xmath496 then by the proof of proposition [ e2 ] , @xmath497 where the second and the last identities we use the isomorphisms given in the above diagram .
hence we have @xmath498 on the other hand , @xmath499 since @xmath500 in @xmath483 , the spectral sequence of @xmath483 by the first filtration(i.e .
, by @xmath132 ) degenerates at @xmath215 .
we have @xmath501 since the projective map from @xmath495 to @xmath502 in the commutative diagram is nothing but the restriction of the quotient map @xmath107 at @xmath495 , by the above analysis , we get an isomorphism @xmath503 thus the spectral sequence of @xmath504 and @xmath483 are isomorphic at @xmath505 for @xmath506 . in our case ,
the first filtration is finite , thus strongly convergent , therefore @xmath107 is a quasi - isomorphism(see cartan - eilenberg @xcite , page 322 , theorem 3.2 ) .
the case @xmath95 is similar . in this case ,
@xmath507 and @xmath508 now follow the same analysis as above . for any subset @xmath146 of @xmath57 , set @xmath509
we see that @xmath510 by the identification of @xmath495 and @xmath493 .
the following theorem is the main result in the paper : [ theoa ] 1 ) .
the cohomology group @xmath370 is given by @xmath511 = \bigoplus_{t\subseteq s } \bigoplus_{\substack{e\in 2r \\
\operatorname{\rm supp}e\supseteq t } } a_e[\ , |t| \ , ] .\ ] ] where @xmath512 $ ] represents the cohomology group @xmath513)$ ] . 2 ) .
the cohomology group @xmath514 is given by @xmath515 = \bigoplus_{t\in { \mathcal i } } \bigoplus_{\substack{e\in 2r\\ \operatorname{\rm supp}e\supseteq t } } a_e[\ , |t| \ , ] .\ ] ] we only prove 1 ) .
the proof of 2 ) follows the same route . by proposition [ xin ] and
proposition [ theoq ] , we know that @xmath516 now @xmath517 part 1 ) follows immediately .
we can see that part 1 ) is actually a special case of part 2 ) when the order ideal @xmath95 is @xmath518 . in the case
@xmath8 , we have [ theom ] there exists a family @xmath519 with the following properties : 1 ) . for each @xmath39 , the subfamily @xmath520 is a @xmath8-basis for @xmath521 . 2 ) . for
any order ideal @xmath95 of @xmath57 , let @xmath522 .
by the inclusion @xmath523 , @xmath524 can be considered as a submodule of @xmath525 .
furthermore , the subfamily @xmath526 is a @xmath527 basis for @xmath528 . 3 ) .
one has cup product structure @xmath529\cup c_{\text{\tiny t},e}=(-1)^{\omega(e ' , e)}\prod_{\substack{i\in s\\e_ie'_i\equiv 1(2 ) } } \left ( \frac{\ell_i-1}{2}\right ) c_{\text{\tiny t } , e+e'}\ ] ] for all @xmath272 and @xmath530 . 1 ) .
by proposition [ theoq ] , we have induced quasi - isomorphism : @xmath531 now since the induced differentials of @xmath132 and @xmath373 in @xmath532 are @xmath136 .
consider the cocycle @xmath533 $ ] in @xmath534 , there exists a cocycle @xmath535(unique modulo boundary ) which is the lifting of @xmath533 $ ] by the quotient map @xmath536 .
hence @xmath537 is a cocycle in the complex @xmath538 .
let @xmath539 denote the cohomology element in @xmath540 represented by the cocycle @xmath537 .
then @xmath541 is a canonical @xmath8-basis for the cohomology group @xmath540 .
this finishes the proof of 1 ) . 2 ) .
similar to 1 ) , just consider the map @xmath542 .
3 ) . for the cup product ,
there is natural homomorphism @xmath543 therefore @xmath540(and also @xmath544 has a natural @xmath545-module structure . by the theory of spectral sequences(see , for example brown @xcite , chap .
@xmath546 , @xmath55 ) , we have the cochain cup product @xmath547 by using the diagonal map @xmath548 defined in @xmath51 , it is easy to check that : @xmath549 hence we can pass the cup product structure to the quotient and have @xmath550 now 3 ) follows immediately from the explicit expression of @xmath548 .
this concludes the proof .
in @xmath55 , we obtain a canonical basis @xmath552 for the cohomology group @xmath540 .
however , little is known yet for the explicit expression of the cocycle @xmath539 in the complex @xmath553 , which makes it necessary to study how to lift the cocycle @xmath533 $ ] in @xmath554 to the cocycle @xmath535 in @xmath555 .
unfortunately , we are unable to get a complete answer for this problem in this paper .
we obtain partial solution in the @xmath136-cocycles case , however , which is enough for us to prove theorem b. recall in @xmath51 , @xmath556 has a double complex structure , therefore we can make @xmath371 as a triple complex
. set @xmath557 : [ a , t]\in l^{p_1 , p_2 } , \deg e = q>\ ] ] and the differentials @xmath558 given by @xmath559=-\sum_{i\in t } \omega(i , t ) n_i[fr^{-1}_i a+\frac{1}{\ell_i } , t\backslash \{i\ } , e]\ ] ] @xmath560=\sum_{i\in t } \omega(i , t ) ( 1-fr^{-1}_i)[a , t\backslash \{i\ } , e]\ ] ] and @xmath373 as given in the double complex @xmath388 . in this setup
, we see that @xmath561 becomes a triple subcomplex of @xmath371 , moreover @xmath562 correspondingly , we define triple complex structures on @xmath563 , @xmath564 and @xmath565 .
this triple complex structure enables us to construct different double complex structures in @xmath371 and @xmath563 . by studying those double complexes ,
we can gather more information about @xmath371 .
this method will be illustrated on next subsection .
for fixed @xmath567 , let @xmath568 with differentials @xmath569 and @xmath373 , then we get a double complex @xmath570 .
similarly , we can get the double complex @xmath571 whose @xmath572-component is @xmath573 .
as before , for any @xmath95 , we have double complexes @xmath574 and @xmath575 which are subcomplexes of @xmath576 and @xmath577 respectively .
first we have [ p1q ] 1 ) .
@xmath578 is a free @xmath8-module generated by cocycles @xmath579 with leading term @xmath533 $ ] and the remainder with @xmath580-degree less than @xmath581 over all pairs @xmath582 satisfying @xmath583 and @xmath584 . 2 ) .
moreover , @xmath585 is a free @xmath8-module generated by cocycles @xmath579 with leading term @xmath533 $ ] and the remainder with @xmath580-degree less than @xmath581 over all pairs @xmath582 satisfying @xmath469 , @xmath583 and @xmath584 .
we only prove ( 1 ) .
the proof of ( 2 ) is similar .
first look the spectral sequence of @xmath576 with the second filtration(i.e . , the filtration given by @xmath580 ) , then @xmath586 next for the differential @xmath569 induced on @xmath215 , with the same analysis as in computing the @xmath426 terms of @xmath587(see @xmath52 , proposition [ e2 ] ) , we have @xmath588 furthermore , let @xmath589 be the double complex generated by all symbols @xmath590 $ ] satisfying @xmath583 and @xmath591 , which can be considered as a quotient complex of @xmath576 . similar to the proof of theorem
a , the quotient map induces an isomorphism between the cohomology groups of each other .
now let @xmath579 be the canonical lifting of the cocycle @xmath590 $ ] in @xmath592 , then @xmath579 is a cocycle in @xmath576 with the leading term @xmath590 $ ] and the remainder contained in the direct sum of @xmath593 where @xmath594 and @xmath595 . [ p1qp2 ]
the spectral sequence of the double complex @xmath596 with the first filtration , degenerates at @xmath215 .
the spectral sequence of the double complex @xmath597 with the first filtration , degenerates at @xmath215 .
we only prove the first part .
the @xmath215-terms of the spectral sequence are @xmath598 note that @xmath599 in general for any spectral sequence , then @xmath600 by theorem a and proposition [ p1q ] , the left hand side and the right hand side of the above inequality have the same number of elements .
hence the inequality is actually an identity .
therefore , the spectral sequence of @xmath555 with filtration given by @xmath601 degenerates at @xmath215 .
the advantage of studying the triple complex structure of the complex @xmath563 is that we can obtain the @xmath602-cocycles of @xmath603 rather quickly .
recall that @xmath604 now for the @xmath602-cocycles @xmath579 , the pair @xmath605 must satisfy @xmath606 and therefore @xmath607 . in this case , for any @xmath183 , we always have @xmath608 first @xmath609=0 ,
\ d_1 [ 0,t , e_t]=-\sum_{i\in t } \omega(i , t ) n_i [ \frac{r_{t\backslash \{i\ } } } { \ell_i } , t\backslash \{i\ } , e_t],\ ] ] then @xmath610)= ( -1)^{|t| } d_1 [ 0,t , e_t],\ ] ] continue this procedure , we have @xmath611.\ ] ] apparently , we see that if @xmath469 , then the cocycles @xmath612 are all contained in the subcomplex @xmath613 . combining the above results , we have [ lifting ] 1 ) .
the canonical basis @xmath614 of @xmath615 is given by @xmath611.\ ] ] 2 ) .
if we restrict our attention in the subcomplex @xmath613 , then @xmath616 has a canonical basis @xmath617 .
first we claim that @xmath618\in h^0(g_s , u_s / mu_s)=(u_s / mu_s)^{g_s}.\ ] ] we prove it by induction on @xmath619 . for @xmath620 , it is easy to see that @xmath621=0 $ ] for all @xmath61 .
now in general , for any @xmath622 , @xmath623=(fr_j-1 ) d_{t\backslash \{j\ } } [ \sum_{i\in t\backslash \{j\ } } \frac{1}{\ell_i}]\ ] ] which is @xmath136 by induction , for @xmath624 , it is obviously @xmath136 .
hence the claim holds .
now we consider the double complex @xmath625 . by proposition
[ p1qp2 ] , we know that @xmath625 degenerates at @xmath215 for the first filtration . by proposition [ lifting ] , @xmath626 is generated by @xmath614 .
we plan to lift @xmath612 to a @xmath136-cocycle in @xmath555 , which is guaranteed by the degeneration at @xmath215 .
moreover , we can study the lifting @xmath612 in @xmath627 .
therefore there exists a cocycle @xmath628 in @xmath627 with the leading term @xmath612 and the remainder contained in the direct product of @xmath629 where @xmath630 and @xmath631 .
hence the image @xmath632 is exactly of the form @xmath633 + re(t),\ ] ] the remainder @xmath634 $ ] where @xmath635 is a proper factor of @xmath636 . both @xmath632 and @xmath637 $ ]
are @xmath136-cocycles of @xmath638 , so is @xmath639 . in order to prove theorem b ,
it is sufficient to prove @xmath640\ ] ] for @xmath178 proper subsets of @xmath146 .
we show @xmath641 by induction to @xmath619 . if @xmath642 , this is trivial .
now in general , without loss of generality , we suppose that @xmath262 and for any @xmath643 , @xmath644 is a linear combination of @xmath645 $ ] for @xmath646 .
then @xmath647 for any @xmath648 is a linear combination of @xmath645 $ ] with @xmath649 . by proposition [ xin ] ,
proposition [ p1qp2 ] and theorem a , @xmath650 is generated by @xmath651 and hence by @xmath652 $ ] .
but obviously @xmath653 , so @xmath641 holds for @xmath654 .
theorem b is proved .
one natural question to ask is if the bases of @xmath551 obtained in theorem [ theom ] and in theorem b are the same .
unfortunately , they are not the same even in the case @xmath655 . right now
, we do nt know too much about the explicit expression of the cocycles @xmath539 .
a deep understanding of those cocycles should tell us more about the arithmetic of the cyclotomic fields .
greg w. anderson let @xmath656 be a free abelian group equipped with a basis @xmath657\}$ ] indexed by @xmath658 .
for all @xmath659 put @xmath660:=[\langle x\rangle]$ ] , where @xmath661 is the unique rational number in the interval @xmath662 congruent to @xmath111 modulo @xmath12 .
the _ universal ordinary distribution _
@xmath105 is defined to be the quotient of @xmath656 by the subgroup generated by all elements of the form @xmath663-\sum_{i=1}^f\left[\frac{x+i}{f}\right ] \;\;\;\left(f\in { \mathbb z}_{>0},\;\ ; x\in { \mathbb q}\right).\ ] ] fix a positive integer @xmath107 .
let @xmath664 be the subgroup of @xmath656 generated by the set @xmath665\left|x\in\frac{1}{f}{\mathbb z}\right.\right\}$ ] .
universal ordinary distribution _
@xmath666 of _ level @xmath107 _ is defined to be the quotient of @xmath664 by the subgroup generated by all elements of the form @xmath663-\sum_{i=1}^g\left[\frac{x+i}{g}\right ] \;\;\;\left(g\in { \mathbb z}_{>0 } , \;g\mid f,\;\;x\in \frac{g}{f}{\mathbb z}\right).\ ] ] the inclusions @xmath667 induce a natural isomorphism @xmath668 .
let @xmath669 be the polynomial ring over @xmath54 generated by a family @xmath670 of independent variables indexed by primes @xmath422 , and for each positive integer @xmath107 , put @xmath671 where @xmath672 is the prime factorization of @xmath107 .
each of the families @xmath673 and @xmath674 is a basis for @xmath669 as a free abelian group .
we equip @xmath656 with @xmath669-module structure by the rule @xmath675=\sum_{i=1}^p\left[\frac{x+i}{p}\right]\ ] ] for all primes @xmath422 and @xmath676 .
one has @xmath677 this last observation is the motivation for all the results to follow .
each @xmath676 has a unique partial fraction expansion @xmath678 where @xmath422 ranges over primes , @xmath257 ranges over positive integers , @xmath679 , @xmath680 , and all but finitely many of the coefficients @xmath681 vanish . for each nonnegative integer @xmath135 ,
put @xmath682 and let @xmath683 be the subgroup of @xmath656 generated by @xmath657\mid x\in { { \mathcal r}}_n\}$ ] . 1 . for all positive integers @xmath107 and @xmath135
, one has @xmath684 where the sum is extended over primes @xmath422 dividing @xmath107 .
2 . for each positive integer @xmath107 , the family @xmath685\right\}$ ] indexed by
the set @xmath686 is a basis for @xmath664 .
3 . for each positive integer @xmath107 , the family @xmath687\right\}$ ] indexed by
the set @xmath686 is a basis for @xmath664 .
the family @xmath688\right\}$ ] indexed by pairs @xmath689 is a basis for @xmath656 .
the family @xmath690\right\}$ ] indexed by pairs @xmath689 is a basis for @xmath656 .
the free abelian group @xmath656 is free as a @xmath669-module , and the family @xmath665\right\}$ ] indexed by @xmath691 is a @xmath669-basis for @xmath656 .
\1 . for each @xmath692
, there exists some prime @xmath422 dividing @xmath107 such that @xmath693 , and one has @xmath663=- \left(\sum_{i=1}^{p-1}\left[x+\frac{i}{p}\right]\right ) + x_p[px],\ ] ] whence the result . 1 . for each @xmath696 ,
the group @xmath666 is free abelian and the family @xmath657\}$ ] indexed by @xmath697 gives rise to a basis for @xmath666 .
the group @xmath105 is free abelian and the family @xmath657\}$ ] indexed by @xmath698 gives rise to a basis for @xmath105 .
3 . the natural map @xmath699 is a split monomorphism .
let @xmath701 be a free abelian group equipped with a basis @xmath702\}$ ] indexed by pairs @xmath703 with @xmath704 and @xmath117 a squarefree positive integer . for all @xmath676 and squarefree integers @xmath117
, put @xmath705:=[\langle x\rangle , g]$ ] . for all @xmath676 and increasing sequences @xmath706 of primes ,
we declare the symbol @xmath707 $ ] to be of degree @xmath708 and we set @xmath709\\\\ & : = & \displaystyle\sum_{i=1}^{m } ( -1)^{i-1 } \left([x , p_1\cdots p_{i-1}p_{i+1}\cdots p_m ] -\sum_{j=1}^{p_i } \left[\frac{x+j}{p_i},p_1\cdots p_{i-1}p_{i+1}\cdots p_m\right ] \right ) , \end{array}\ ] ] thereby equipping the group @xmath701 with a grading and a differential @xmath132 of degree @xmath12 .
the map @xmath710\mapsto [ x]$ ] induces an isomorphism @xmath711 . fix a positive integer @xmath107 .
we define @xmath713 to be the graded subgroup spanned by the symbols of the form @xmath705 $ ] where @xmath117 divides @xmath107 and @xmath714 .
it is clear that @xmath713 is @xmath132-stable .
the map @xmath710\mapsto [ x]$ ] induces an isomorphism @xmath715 .
let @xmath716 be the exterior algebra over @xmath669 generated by a family of symbols @xmath717 indexed by primes @xmath422 .
for each increasing sequence @xmath706 of prime numbers , put @xmath718 and declare @xmath719 to be of degree @xmath708 , thereby defining a @xmath669-basis @xmath720 for @xmath716 indexed by squarefree positive integers @xmath721 and equipping @xmath716 with a @xmath669-linear grading .
let @xmath132 be the unique @xmath669-linear derivation of @xmath716 of degree @xmath12 such that @xmath722 for all primes @xmath422 .
one then has @xmath723 for all increasing sequences @xmath706 of prime numbers .
fix a positive integer @xmath107 .
the graded subgroup @xmath725 generated by all elements of the form @xmath726 where @xmath727 divides @xmath107 is @xmath132-stable .
it is not difficult to verify that the complex @xmath724 is acyclic in nonzero degree , and that @xmath728 is a free abelian group of rank @xmath12 generated by the symbol @xmath729 .
we equip @xmath701 with graded left @xmath716-module structure by the rules @xmath730=\left\{\begin{array}{ll } ( -1)^{|\{i\mid p_i < p\}|}[x , pp_1\cdots p_m]&\mbox{if $ p\not\in \{p_1,\dots , p_m\}$}\\ 0&\mbox{if $ p\in \{p_1,\dots , p_m\}$ } \end{array}\right.\ ] ] and @xmath731=\sum_{i=1}^p \left[\frac{x+i}{p},p_1\cdots p_m\right]\ ] ] for all primes @xmath422 and increasing sequences @xmath706 of primes . by a straightforward calculation that we omit
, one can verify that @xmath732 for all homogeneous @xmath733 and @xmath734 .
we have only to prove the first statement . by theorem [ theorem : free ] and
a straightforward calculation that we omit , one has @xmath735\ ] ] where the direct sum is indexed by pairs @xmath703 with @xmath736 and @xmath117 is the largest positive integer such that @xmath714 .
each of the subcomplexes @xmath737,d)$ ] is an isomorphic copy of @xmath738 , and the latter we have already observed to be acyclic in nonzero degree .
the construction of @xmath712 presented here is cobbled together from ideas presented in the author s papers @xcite and @xcite , along with simplifications suggested by many conversations with pinaki das and yi ouyang on these topics .
brown , kenneth , _ cohomology of groups .
_ graduate texts in mathematics 87 , spring - verlag , 1982 .
cartan , h. and s. eilenberg , _ homological algebra .
_ princeton university press , 1956 .
das , pinaki , _ algebraic gamma monomials and double coverings of cyclotomic fields , _ trans .
soc . , to appear .
graduate texts in kubert , d.s .
, _ the universal ordinary distribution , _ bull .
soc . math .
107*(1979 ) , 179 - 202 .
kubert , d.s . , _ the @xmath739 cohomology of the universal ordinary distribution , _ bull .
107*(1979 ) , 103 - 224 . lang , serge , _ cyclotomic fields , i and ii , combined 2nd ed .. _ graduate texts in mathematics 121 , springer verlag , 1990 .
ouyang , yi , _
spectral sequences of universal distribution and sinnott s index formula .
_ preprint , 1999 .
rubin , karl , _ the main conjecture , _ appendix to s. lang s cyclotomic fields i and ii .
rubin , karl , _ euler systems .
_ serre , j - p . , _ local fields . _ graduate texts in mathematics 67 , springer verlag , new york , 1979 .
serre , j - p . , _ galois cohomology . _ springer verlag , berlin heidelberg , 1997 . | for any odd squarefree integer @xmath0 , we get complete description of the @xmath1 group cohomology of the universal ordinary distribution @xmath2 in this paper . moreover ,
if m is a fixed integer which divides @xmath3 for all prime factors @xmath4 of @xmath0 , we study the cohomology group @xmath5 .
in particular , we explain the mysterious construction of the elements @xmath6 for @xmath7 in rubin@xcite , which come exactly from a certain @xmath8-basis of the cohomology group @xmath9 through an evaluation map . |
we use a room - temperature type iia diamond with @xmath107 of @xmath1c grown by chemical vapor deposition ( element 6 ) . we apply a magnetic field of @xmath108 g along the nv symmetry axis ( @xmath30-axis ) , yielding a @xmath1c larmor frequency @xmath109 khz . the electronic dephasing time @xmath60 is @xmath110s .
the hyperfine interaction for nuclear spin @xmath7 is given by @xmath111 ( fig .
1a ) , with @xmath53 the component parallel to the magnetic field and @xmath112 the perpendicular component . @xmath113 khz and @xmath114 khz for nuclear spin 1 , and @xmath115 khz and @xmath116 khz for nuclear spin 2 .
because @xmath117 khz the nuclear spins are weakly coupled to the electron spin and the hyperfine splittings are unresolved . in a suitable rotating frame , the hamiltonian with a single nuclear spin can be written : @xmath118 with @xmath119 and @xmath120 and with @xmath35 and @xmath14 the @xmath34 and @xmath45 electron states , respectively .
nuclear spin gates are performed by applying sequences of the type @xmath121 on the electron spin ( rabi frequency @xmath122 mhz ) . because we set @xmath123
, sharp resonances occur at @xmath124 , with integer @xmath125 . at these values
a nuclear @xmath16-rotation is performed ( assuming @xmath126 ) . for odd
@xmath125 the direction of the rotation is conditional on the electron spin ( e.g. the @xmath15 gates ) , for even @xmath125 it is unconditional ( @xmath127 gates ) . for the conditional gates we use @xmath128s , @xmath129 for spin 1 and @xmath130s , @xmath131 for spin 2 . for the unconditional gates we use @xmath132s , @xmath133 for spin 1 and @xmath134s , @xmath135 for spin 2 .
z - rotations are implemented by choosing @xmath29 off - resonant .
detailed simulations of the nuclear spin dynamics are available in the supplementary information . the electron ramsey measurements in fig . 1d and 1e
are analyzed in two ways : ( 1 ) the measurements are separately fit to @xmath136 , in which @xmath60 is a measure for the dephasing time set by the entire spin bath . the external magnetic field stability of better than @xmath137 mg over the total integration time ( @xmath138 hours ) , required in these experiments ,
was achieved by post selecting from a larger measurement set .
( 2 ) we determine the nuclear spin initialization fidelities @xmath139 and @xmath140 by averaging over multiple measurement runs ( supplementary information ) and using the hyperfine components @xmath141 and @xmath142 together with : @xmath143 + f_1(1-f_2 ) \cos\bigg[\left(\omega+\frac{a_\parallel^1-a_\parallel^2}{2}\right)t\bigg]\ ] ] @xmath144+(1-f_1)(1-f_2)\cos\bigg[\left(\omega-\frac{a_\parallel^1 + a_\parallel^2}{2}\right)t\bigg]\bigg).\ ] ] here @xmath145s is the electronic dephasing due to the rest of the spin bath , i.e. not including the two spins under study . the applied errors realize the quantum map : @xmath146 in which @xmath147 is the initial density matrix ( error characterization in supplementary information ) .
we analyze the error - correction protocol by separating depolarization during the encoding , decoding and error - correction steps from the robustness of the encoded state to applied errors , which is characterized by the three probabilities @xmath86 that an error applied on qubit @xmath87 is successfully corrected ( derivation in supplementary information ) .
the process fidelity for a single - qubit error ( fig .
4d ) is then given by : @xmath148,\ ] ] where @xmath149 and @xmath150 characterize the additional depolarization and are given by the average fidelities without applied errors .
the equation contains a constant due to the @xmath69 states , which are unaffected by the applied error , and a sum of successful ( @xmath151 ) and unsuccessful ( @xmath152 ) error correction for the @xmath70 and @xmath71 states . for errors simultaneous on all three qubits ( fig .
4e ) , the process fidelity becomes : @xmath153,\ ] ] with @xmath154 the error probability .
in general this equation describes a third order polynomial . for ideal error correction ( @xmath96 ) the linear term vanishes , whereas without robustness to errors ( @xmath94 ) , the result is strictly linear .
the inversion symmetry about @xmath155 observed both theoretically and experimentally ensures that the nonlinear behavior is not due to spurious coherent rotations .
after submission of this manuscript we became aware of related work by waldherr et al .
, arxiv:1309.6424 in which 3-qubit quantum error correction is implemented using strongly coupled nuclear spins .
we thank l. childress , j. j. l. morton , o. moussa and l.m.k .
vandersypen for helpful discussions and comments .
t.h.t . acknowledges support by a marie curie intra european fellowship within the 7th european community framework programme .
work at the ames laboratory was supported by the u.s .
department of energy basic energy sciences under contract no .
de - ac02 - 07ch11358 .
we acknowledge support from the dutch organization for fundamental research on matter ( fom ) , the netherlands organization for scientific research ( nwo ) , the darpa quasar programme , the eu solid , and diamant programmes and the european research council through a starting grant .
[ 1 ] awschalom , d.d . ,
bassett , l.c . ,
dzurak , a.s . , hu ,
e.l . , & petta , j.r . , quantum spintronics : engineering and manipulating atom - like spins in semiconductors .
_ science _ * 339 * , 1174 - 1179 ( 2013 ) .
[ 2 ] koehl , w.f . ,
buckley , b.b . , heremans , f.j . , calusine , g. , & awschalom , d.d .
, room temperature coherent control of defect spin qubits in silicon carbide .
_ nature _ * 479 * , 84 - 87 ( 2011 ) .
[ 3 ] yin , c. et al . , optical addressing of an individual erbium ion in silicon .
_ nature _ * 497 * , 91 - 94 ( 2013 ) .
[ 4 ] gurudev dutt , m.v .
quantum register based on individual electronic and nuclear spin qubits in diamond .
_ science _ * 316 * , 1312 - 1316 ( 2007 ) .
[ 5 ] neumann , p. et al . ,
multipartite entanglement among single spins in diamond
. _ science _ * 320 * , 1326 - 1329 ( 2008 ) . [ 6 ]
fuchs , g.d .
, burkard , g. , klimov , p.v . , & awschalom , d.d . , a quantum memory intrinsic to single nitrogen - vacancy centres in diamond . _ nature phys . _ * 7 * , 789 - 793 ( 2011 ) . [ 7 ] van der sar , t. et al .
, decoherence - protected quantum gates for a hybrid solid - state spin register .
_ nature _ * 484 * , 82 - 86 ( 2012 ) .
[ 8 ] maurer , p.c .
et al . , room - temperature quantum bit memory exceeding one second .
_ science _ * 336 * , 1283 - 1286 ( 2012 ) .
[ 9 ] pfaff , w. et al . ,
demonstration of entanglement - by - measurement of solid - state qubits . _ nature phys . _
* 9 * , 29 - 33 ( 2013 ) .
[ 10 ] pla , j.j .
et al . , high - fidelity readout and control of a nuclear spin qubit in silicon .
_ nature _ * 496 * , 334 - 338 ( 2013 ) .
[ 11 ] dolde , f. et al .
, room - temperature entanglement between single defect spins in diamond .
_ nature phys . _
* 9 * , 139 - 143 ( 2013 ) .
[ 12 ] jiang , l. et al .
, repetitive readout of a single electronic spin via quantum logic with nuclear spin ancillae .
_ science _ * 326 * , 267 - 272 ( 2009 ) .
[ 13 ] lee , s .- y .
et al . , readout and control of a single nuclear spin with a metastable electron spin ancilla .
_ nature nanotech . _
* 8 * , 487 - 492 ( 2013 ) .
[ 14 ] cory , d.g .
et al . , experimental quantum error correction . _
_ * 81 * , 2152 - 2155 ( 1998 ) . [ 15 ] moussa , o. , baugh , j. , ryan , c.a . , &
laflamme , r. , demonstration of sufficient control for two rounds of quantum error correction in a solid state ensemble quantum information processor .
lett . _ * 107 * , 160501 ( 2011 ) .
[ 16 ] schindler , p. et al . , experimental repetitive quantum error correction .
_ science _ * 332 * , 1059 - 1061 ( 2011 ) .
[ 17 ] reed , m.d .
et al . , realization of three - qubit quantum error correction with superconducting circuits .
_ nature _ * 482 * , 382 - 385 ( 2012 ) .
[ 18 ] bernien , h. et al . , heralded entanglement between solid - state qubits separated by three metres .
_ nature _ * 497 * , 86 - 90 ( 2013 ) .
[ 19 ] nickerson , n.h . , li , y. , & benjamin , s.c . , topological quantum computing with a very noisy network and local error rates approaching one percent . _ nat .
commun . _ * 4 * , 1756 ( 2013 ) .
[ 20 ] neumann , p. et al . , single - shot readout of a single nuclear spin .
_ science _ * 329 * , 542 - 544 ( 2010 ) . [ 21 ] robledo , l. et al . , high - fidelity projective read - out of a solid - state spin quantum register .
_ nature _ * 477 * , 574 - 578 ( 2011 ) .
[ 22 ] dreau , a. , spinicelli , p. , maze , j.r . ,
roch , j.f .
, & jacques , v. , single - shot readout of multiple nuclear spin qubits in diamond under ambient conditions . _ phys . rev . lett . _ * 110 * , 060502 ( 2013 ) .
[ 23 ] taminiau , t.h .
et al . , detection and control of individual nuclear spins using a weakly coupled electron spin .
lett . _ * 109 * , 137602 ( 2012 ) .
[ 24 ] kolkowitz , s. , unterreithmeier , q.p . ,
bennett , s.d . , & lukin , m.d . , sensing distant nuclear spins with a single electron spin .
* 109 * , 137601 ( 2012 ) .
[ 25 ] zhao , n. et al . , sensing single remote nuclear spins . _ nature nanotech . _ * 7 * , 657 - 662 ( 2012 ) . [ 26 ] hodges , j.s .
, yang , j.c . , ramanathan , c. , & cory , d.g . ,
universal control of nuclear spins via anisotropic hyperfine interactions .
rev . a _ * 78 * , 010303 ( 2008 ) .
[ 27 ] zhang , y. , ryan , c.a . ,
laflamme , r. , & baugh , j. , coherent control of two nuclear spins using the anisotropic hyperfine interaction .
lett . _ * 107 * , 170503 ( 2011 ) .
[ 28 ] london , p. et al . , detecting and polarizing nuclear spins with double resonance on a single electron spin .
lett . _ * 111 * , 067601 ( 2013 ) .
[ 29 ] filidou , v. et al .
, ultrafast entangling gates between nuclear spins using photoexcited triplet states .
_ nature phys . _
* 8 * , 596 - 600 ( 2012 ) .
[ 30 ] bar - gill , n. , pham , l.m . ,
jarmola , a. , budker , d. , & walsworth , r.l .
, solid - state electronic spin coherence time approaching one second .
commun . _ * 4 * , 1743 ( 2013 ) .
the experimental setup and sample are described in detail in the supplementary information of van der sar et al .
@xcite we used a type - iia chemical vapour deposition grown diamond with a @xmath107 natural abundance of carbon-13 ( element 6 ) .
solid immersion lenses were fabricated on top of the nitrogen vacancy ( nv ) centres to enhance the collection efficiency @xcite .
the electron spin is controlled by microwaves through an on - chip stripline ( rabi frequency of 31.25 mhz ) .
a magnetic field of @xmath108 g was applied along the nv symmetry axis using three electromagnets . at this magnetic field
the intrinsic nv nitrogen-14 spin is polarized due to an excited - state anti - crossing @xcite .
this section discusses the electron spin initialization , re - initialization and readout .
in particular it analyzes how the imperfect spin and charge state initialization affect the outcomes of the different type of experiments performed .
the electron spin is initialized in the @xmath34 state by a @xmath156 nm ( @xmath157w ) laser pulse ( typically 4@xmath158s ) and read out through its spin - dependent time - resolved fluorescence . in all experiments we measure the difference signal @xmath159 between the fluorescence signal @xmath160 for the final state and the fluorescence signal @xmath161 for the final state with a pi - pulse applied just before readout ( @xmath34 to @xmath45 transition ) .
the obtained value is then normalized by dividing it by the same difference signal right after initialization : @xmath162 , where @xmath163 is without pi - pulse and @xmath164 with pi - pulse .
the final normalized contrast @xmath165 is : @xmath166 this method directly measures the contrast between @xmath34 and @xmath45 states .
note that @xmath167 and that the result is independent of the population in other states , such as @xmath168 , that are not affected by the microwave pi - pulse .
the reported expectation values directly correspond to @xmath165 , the measured fidelities are obtained from @xmath169 .
the electronic initialization involves both spin states ( @xmath170 ) and charge states ( @xmath171 and @xmath172 ) .
the initial state @xmath173 is : @xmath174 with @xmath175 , and in which @xmath176 is the desired @xmath34 state , @xmath177 is the completely mixed state of @xmath34 and @xmath45 , @xmath178 represents the other spins states ( here @xmath168 ) and @xmath179 other charge states ( here @xmath172 )
. the precise values for @xmath180 and @xmath181 are unknown . for this nv centre
the spin - state initialization fidelity was previously reported to be @xmath182 under similar conditions @xcite .
the nv@xmath183 charge - state initialization fidelity @xmath184 is unknown here , but values of @xmath185 have been reported for other nv centres @xcite .
the available initial population is given by @xmath186 .
ideally , measurements of @xmath187 directly reflect the actual polarization so that @xmath188 . because only the @xmath176 term in equation [ eq : electron init ] is affected by microwave pulses only this term yields signal ( non - zero @xmath189 ) , so that the normalization signal always is @xmath190 , with @xmath191 an unknown proportionality constant .
next we determine the obtainable final signal @xmath192 for two types of experiments : experiments that do not re - initialize the electron to create additional polarization in the nuclear spin register and those that do .
first consider experiments where only a single electron initialization step is used , i.e. experiments that do not transfer polarization to the nuclear spins before resetting the electron spin . in this case
the maximum value of @xmath192 simply is @xmath193 and the maximum contrast is @xmath194 . due to the calibration ,
the final measured contrast is independent of @xmath186 and therefore does not take into account the charge and spin initialization fidelities .
now consider experiments in which the electron spin polarization is transferred to a nuclear spin , the electron is re - initialized , and finally the electron is used to measure the nuclear spin state .
the result @xmath192 depends on the correlations of the spin ( charge ) state after the re - initialization step with the spin ( charge ) state before it .
we assume the spin states before and after re - initialization are uncorrelated , and derive the result @xmath192 for both uncorrelated ( no memory ) and maximally positive correlations ( ideal memory ) for the charge state .
the state of the initialized electron and a single nuclear spin in a completely mixed state is : @xmath195 swapping the electron and nuclear spin states gives : @xmath196 as the swap gate has no effect on the erroneous electron spin ( @xmath197 ) and charge ( @xmath198 ) states .
the electron spin initialization @xmath186 is thus directly transferred to the nuclear spin .
we re - initialize the electron spin and assume that electron spin initialization is independent of the nuclear spin state .
first consider the case of no correlations ( no memory ) for the charge state , so that the electron is completely re - initialized .
the state in equation [ eq : after_swap ] becomes : @xmath199 reading out the nuclear spin with the electron spin only yields non - zero signal for both the electron and nuclear spins in the pure state @xmath176 , so that : @xmath200 which shows that the maximum contrast @xmath201 is reduced by a factor @xmath186 and thus that the experiment faithfully reflects the actual nuclear spin state , including a reduced fidelity due to the imperfect electron spin and charge initialization .
if the electron re - initialization does not change the charge state , equation [ eq : after_swap ] after electron re - initialization becomes : @xmath202 again taking into account that no difference signal is obtained if either the electron or the nuclear spin is not in state @xmath176 : @xmath203 the result now accurately reflects the spin state initialization , but is independent of the charge state initialization .
the high nuclear initialization fidelity obtained here ( @xmath204 , figure 1 of the main text ) , indicates that the charge state initialization fidelity is high ( @xmath205 ) or that the measurements are not sensitive to it ( i.e. the re - initialization laser pulse has low probability to change the charge state ) .
the same value gives a lower limit of the electron spin initialization @xmath206 , as the swap gate for initialization and the nuclear spin readout have limited fidelities as well .
as in previous room temperature experiments , the charge state is thus not rigourously initialized nor proven to be fully reflected in the measurement outcomes . therefore the measured state fidelities do not give the actual purity of the states and no entanglement can be proven to be present .
nevertheless the ( entangling ) gates and protocols developed and studied in this work can be accurately investigated through their action on the prepared states .
note that methods to initialize the charge state have been developed at room temperature @xcite and that pure entangled states have been reported at cryogenic temperatures using simultaneous spin and charge initialization @xcite .
to demonstrate that harnessing weakly coupled spins makes multiple qubits available for each defect with high certainty , we have controlled three weakly coupled nuclear spins for each of the three nv centres studied .
this section contains the characterization of the nv centres and the nuclear - spin free - evolution experiments that demonstrate the initialization , control and readout of the nuclear spins . , nv@xmath207 and nv@xmath208
. the electron spin is prepared in a coherent superposition state ( @xmath209 ) and a dynamical decoupling sequence with 32 pi - pulses of the form @xmath210 is applied with variable interpulse delay @xmath9 before measuring @xmath211 .
sharp dips in the signal indicate an entangling operation of the electron spin with individual @xmath1c spins in the spin bath @xcite .
the arrows indicate the 9 different @xmath1c nuclear spins , and the values of @xmath29 used , for which we implemented initialization , control and readout ( see figures [ fig : ramseys_nv_a ] for nv@xmath212 , [ fig : ramseys_nv_b ] for nv@xmath207 and [ fig : ramseys_nv_c ] for nv@xmath208 ) .
the experiments in the main text use nuclear spin 1 and 2 of nv@xmath212 .
@xmath213 is the bare larmor period . ]
we use dynamical decoupling spectroscopy @xcite to characterize the nuclear spin environment of a total of three nv centres : nv@xmath212 , which is studied in the main text , and the two additional centres nv@xmath207 and nv@xmath208 ( fig .
[ fig : compare_nv ] ) .
the resulting curves provide characteristic fingerprints of the nuclear spin environments of the nv centre .
nv@xmath212 and nv@xmath207 show qualitatively similar behavior ( fig .
[ fig : compare_nv ] ) ; both curves display broad echo collapses due to the spin bath at @xmath214 with odd @xmath215 and show distinct sharp dips due to individual @xmath1c nuclear spins that become visible at larger @xmath29 @xcite .
however , the positions and depths of the different dips differ strongly due to the characteristic distribution of nuclear spins near each nv centre .
in addition to a bath of weakly - coupled @xmath1c spins nv@xmath208 shows a rapidly oscillating component in the signal due to the presence of a strongly coupled nuclear spin ( hyperfine interaction of @xmath216 khz ) . * .
experiments as in figure 2 of the main text , but with the electron spin in @xmath34 ( left ) or @xmath45 ( right ) .
the values for @xmath29 used for spin 1 ( orange ) , spin 2 ( green ) and spin 3 ( purple ) are marked in fig .
[ fig : compare_nv ] . note that nuclear spin 1 and nuclear spin 2 are the two spins studied in detail and used for implementing the quantum - error - correction protocol . ] * .
experiments as in figure 2 of the main text , but with the electron spin in @xmath34 ( left ) or @xmath45 ( right ) .
the values for @xmath29 used for spin 1 ( orange ) , spin 2 ( green ) and spin 3 ( purple ) are marked in fig .
[ fig : compare_nv ] . ] * .
experiments as in figure 2 of the main text , but with the electron spin in @xmath34 ( left ) or @xmath45 ( right ) .
the values for @xmath29 used for spin 1 ( orange ) , spin 2 ( green ) and spin 3 ( purple ) are marked in fig .
[ fig : compare_nv ] . ] for each of the three nv centres in figure [ fig : compare_nv ] , we select three nuclear spins ( marked in the figure ) and demonstrate initialization , control and direct readout by performing nuclear free precession experiments ( see figures 2a - d of the main text ) .
we prepare the electron spin in @xmath34 or @xmath45 .
the unique precession frequencies for @xmath45 confirm that in each case three different @xmath1c spins are controlled ( fig .
[ fig : ramseys_nv_a ] , [ fig : ramseys_nv_b ] and [ fig : ramseys_nv_c ] ) .
these results demonstrate the control of three weakly - coupled nuclear spins for each nv centre studied .
our decoherence - protected gates therefore make several nuclear spins available per defect centre with a high certainty , in stark contrast to the highly probabilistic nature of the presence of strongly coupled @xmath1c spins .
the fact that the gates can be applied in the presence of strongly coupled nuclear spins , including the intrinsic nitrogen and nearby @xmath1c nuclear spins , indicates that quantum registers with over @xmath217 nuclear spins are now well within reach ( see e.g. nv@xmath208 in figure [ fig : compare_nv ] ) .
the hyperfine parameters for the three nuclear spins for nv@xmath212 , the nv centre used in the main text , are given in table [ table : hyperfine_parameters ] .
the two nuclear spin qubits in the main text are spin 1 and 2 .
.*hyperfine parameters for 3 relevant @xmath1c spins .
* @xmath53 is the component parallel to the applied magnetic field ( along the nv symmetry axis ) .
@xmath112 is the perpendicular component .
this nv centre was studied previously @xcite . [
cols="^,^,^ " , ] for a variable strength error on one of the qubits the final fidelity for inputs @xmath70 and @xmath71 is given by a weighted sum of the two corresponding values in table [ table:8 signals ] : @xmath218 in which @xmath219 and @xmath220 identify the applied error combination .
for example , for the variable error applied to qubit 2 and no error to qubits 1 and 3 , we have @xmath221 and @xmath222 . for @xmath69
the signal is simply constant : @xmath223 in figure 4d of the main text two different types of errors are applied : ( 1 ) just a variable error on qubit @xmath87 and ( 2 ) a variable error on qubit 2 and a full flip on qubit 1 . for a variable error on qubit @xmath87 ( @xmath224 ) and input @xmath70 or @xmath71
equation [ fit_equation1 ] simplifies to : @xmath225 in agreement with the interpretation of the values @xmath86 as the probability that an error on qubit @xmath87 is successfully corrected . for the variable error on qubit 2 and a full flip on qubit 1
we find : @xmath226 which is of the same form as equation [ fit_equation3 ] and shows that for @xmath227 a cosine around fidelity @xmath228 is obtained ; the error correction is effectively switched of .
the process fidelity @xmath81 of the error correction process with the identity is : @xmath229 in which the @xmath230 are given by equations [ fit_equation1 ] and [ fit_equation2 ] . for a single applied error
this simplifies to : @xmath231 in which @xmath232 and @xmath233 .
note that all the different fidelities without error get grouped into two constants , one related to the average fidelity of the @xmath69 states and one related to the average fidelity of the @xmath70 and @xmath71 states without applied errors . as in figure 4d of the main text .
solid lines are a fit to the model including equation [ initialization_matrix ] .
the dashed lines are the expected results for ideal initialization of the two nuclear ancillas ( @xmath234).,title="fig : " ] + the above model can be modified to explicitly take the effect of imperfect initialization / polarization of the ancilla qubits ( qubit 1 and 3 ) into account .
the effect of the initialization fidelities @xmath235 and
@xmath236 of the two ancilla qubits is that the measured values @xmath237 are now combinations of the @xmath238 values in table [ table:8 signals ] following : @xmath239 this extended model separates the initialization imperfections from imperfections in the error correction process .
we take @xmath240 as an estimate for the initialization fidelities ( fig .
[ fig : nuclear_initialization ] ) .
the resulting fits are shown in figure [ selected_errors_process_corrected ] ( solid lines ) and yield @xmath241 and @xmath242 .
we calculate the expected result for ideal initialization by using the same value for @xmath243 but now setting @xmath234 ( dashed lines , fig .
[ selected_errors_process_corrected ] ) .
the imperfect initialization has two effects .
first it strongly affects the success probability of the error correction as double errors ( one initialization error + one applied error ) can not be corrected .
second , it lowers the overall maximum fidelity .
this is a weak effect because it requires an error in the preparation of both ancillas at the same time and is therefore proportional to @xmath244 .
input states @xmath245 and @xmath246 , for @xmath247 different combinations of errors and as a function of the error angle @xmath76 .
we apply @xmath248 to each of the three qubits separately and a combination of @xmath248 to the electron and @xmath249 to nucleus 1 .
lines are fits to equations [ fit_equation2]-[fit_equation4 ] . , title="fig : " ] + the process fidelity @xmath81 for simultaneous errors is given by : @xmath250 with @xmath251 the error probability and @xmath252 the process fidelity for applied error @xmath219 , i.e. : @xmath253 with @xmath254 as given in table [ table:8 signals ] we obtain : @xmath255,\ ] ] the experiment is completely described by just 3 parameters : the offset @xmath84 ( due to the average @xmath69 fidelity without applied errors ) , the amplitude @xmath256 ( due to the average @xmath257 fidelity without applied errors ) and the average error correction probability @xmath258 .
order polynomial fits to the data .
, title="fig : " ] + the complete set of state fidelities used to derive the process fidelities for errors applied to one of the qubits at a time ( main text figure 4d ) is shown in figure [ selected_errors ] .
the complete sets of state fidelities used to obtain the process fidelities for simultaneously applied errors ( fig .
4e of the main text ) are given in figure [ 6_curves_encdec]a ( without error correction ) and figure [ 6_curves_encdec]b ( with error correction ) .
in this section we analyze the different decoherence mechanisms in the three - qubit register .
the electronic depolarization ( longitudinal relaxation or t1-type process ) due to phonon interactions plays an important role at these room temperature experiments . to measure the depolarization rates we prepare one of the three states @xmath45 , @xmath34 and @xmath168 and
let the system relax for a time @xmath259 .
we then apply a pi - pulse on the @xmath34 transition ( for the state starting in @xmath34 nothing is done ) before reading out the electron .
the results are fit to a 3-level model that yields three rates between the different levels ( fig .
[ fig : t1_3level ] ) .
we find : @xmath260 s@xmath261 , @xmath262 s@xmath261 and @xmath263 s@xmath261 . in this three level system
no unique `` @xmath19 '' value can be defined .
nevertheless , a separate analysis of each of the curves gives @xmath264 times of @xmath265 ms ( @xmath45 ) , @xmath266 ms ( @xmath34 , which is often reported as the @xmath19 value ) and @xmath267 ms ( @xmath168 ) .
we verified that the same rates were obtained with a 4 times lower laser output power , indicating that transitions induced by background illumination are negligible .
the fits additionally use the measured fluorescence for @xmath45 and @xmath168 preparation . ] to measure the electronic coherence time under dynamical decoupling @xmath31 the electron spin is prepared along @xmath16 .
we then apply a decoupling sequence with @xmath268s and measure the spin projection along @xmath16 . the total time is varied by varying the number of pulses @xmath8 in the sequence .
the result is shown in figure [ fig : dd_superposition ] . and measure the final state along @xmath16 .
the interpulse delay @xmath29 is @xmath269 , in the same range as used for the nuclear gates .
purple : exponential fit to the data that gives @xmath12 .
green : expected value due to electron depolarization alone . ]
the green line marks the limit given by phonon - induced depolarization of the electron spin .
it is given by the total decay rate out of the @xmath34 and @xmath45 levels : @xmath270 .
the additional decoherence observed experimentally is consistent with previous reports @xcite and is likely due to phonon - induced dephasing , as much longer coherence times were reported at low temperatures @xcite .
the expected signal without phonon - induced dephasing and depolarization is given by : @xmath271 with the spin echo time @xmath272s .
with @xmath273 this gives an estimated decay time of @xmath274 ms , indicating that decoupling from the spin bath is not the limiting factor .
we measure the nuclear dephasing time @xmath275 by preparing the nuclear spin in a superposition and the electron spin in @xmath34 , and let the system evolve for variable time .
the electron spin is then reset to @xmath34 before the nuclear spin is measured along an axis that creates an effective detuning of approximately 1 khz .
we obtain @xmath276 ms for nuclear spin 1 and @xmath277 ms for nuclear spin 2 ( fig .
[ fig : t2star ] ) .
an electron free - precession ( ramsey - type ) measurement is performed during the experiments ( interleaved on a @xmath278 timescale ) , so that the electron and nuclear @xmath275 can be compared under the same magnetic field fluctuations . . * ( a ) for nuclear spin 1 .
@xmath276 ms , simultaneously measured electron @xmath279s .
( b ) for nuclear spin 2 .
@xmath277 ms , simultaneously measured electron @xmath280s .
fits are sine functions with a decaying envelope @xmath281 .
spin 1 : @xmath282 , spin 2 : @xmath283 . no readout correction . ]
we expect the nuclear dephasing time to be set by a combination of electron relaxation and magnetic field fluctuations ( including the nuclear spin bath ) .
electron relaxation gives a rate of @xmath284 s@xmath261 ( time constant of 8.2 ms ) . to estimate the intrinsic nuclear dephasing timescale @xmath285 we subtract the electron depolarization rate from the inverse of the measured dephasing time @xmath286 .
we find @xmath287 ms for spin 1 and @xmath288 ms for spin 2 .
the difference inthese values could originate from the differences in the nuclear spin s microscopic environments .
we stabilize the magnetic field through a feedback loop by periodically measuring the energy splitting of the nv centre .
this stabilization is required to counteract slow magnetic field drifts ( order of 0.1 g ) over the measurement time .
figure [ fig : b - field_stabilization ] characterizes the magnetic field stability during the quantum - error - correction measurements with simultaneous errors ( taken over a total of @xmath289 hours , spread out over 1 month ) .
these values are representative for the other measurements .
khz ( @xmath26 ) are observed , corresponding to @xmath290 mg .
( right ) the average measurement uncertainty in a single instance of the magnetic field measurement is @xmath291 khz ( @xmath137 mg ) . ]
the measured residual slow fluctuations of the magnetic field ( @xmath292 khz , @xmath293 g ) are small compared to the fast fluctuations due to the @xmath1c bath ( @xmath294 khz , @xmath295 g ) .
these slow fluctuations are expected to decrease the electron @xmath296 from the instantaneous value @xmath297s to @xmath298s ( and the nuclear @xmath299 with approximately the same factor ) .
part of the fluctuations are caused by the uncertainty in the measurements of the electron splitting .
figure [ fig : b - field_stabilization ] shows that this effect is small , because the measurement uncertainty of ( @xmath291 khz ) is small compared to the total drift observed ( @xmath292 khz ) .
the nuclear spin coherence times can be extended by decoupling from the spin bath .
figure [ fig : spin_echo ] shows the results of nuclear spin echo experiments .
the required pi - pulse is constructed in the same way as all nuclear gates in this work .
the electron is prepared in @xmath34 and re - initialized before the pi - pulse , which makes it possible to use a conditional gate , and re - initialized again to be used in the final measurement .
we find @xmath300 ms ( nuclear spin 1 ) and @xmath301 ms ( nuclear spin 2 ) .
( b ) for nuclear spin 2 : @xmath301 ms .
single exponential fits . no readout or initialization correction . ] being able to re - initialize the electron spin without depolarizing or dephasing the nuclear spins is essential for initializing the multiqubit register and for performing partial measurements within such registers . under illumination .
* ( a ) nuclear spin 1 , @xmath302 @xmath158s .
( b ) nuclear spin 2 , @xmath303 ms .
the laser power is the same as used in the initialization , re - initialization and readout steps .
no readout or initialization correction . ]
figure [ fig : t2star_laser ] shows a @xmath286 under illumination of @xmath304 @xmath158s for nuclear spin 1 and @xmath62 ms for nuclear spin 2 .
these times are long compared to the time required to re - initialize the electron spin ( @xmath305s ) .
for example , for nuclear spin 1 this predicts a contrast loss of approximately @xmath306 .
figure [ fig : t1_nuclei ] shows nuclear relaxation measurements for the both spins , with and without laser illumination .
the nuclear spin is prepared in @xmath35 and the electron spin in @xmath34 .
we let the system relax for a variable time during which the laser is either on or off .
for the experiment without laser illumination , the electron is reset by a short laser pulse ( @xmath307s ) so that it can be used to measure the nuclear spin state . without illumination , we find @xmath308 s and @xmath309 ms for spin 1 and 2 respectively . with illumination
, we find @xmath310 ms and @xmath311 ms for spin 1 and 2 respectively .
the nuclear depolarization during laser illumination is slow compared to the time it takes to re - initialize the electron spin ( @xmath305s ) , so that the electron can be re - initialized without depolarizing the nuclei .
note that the final signal approaches a fidelity of @xmath312 ; prolonged laser light does not create a preferential polarization for these nuclear spins .
the estimates for the final fidelities for the three - qubit quantum error correction protocol in the main text are obtained from the values above as follows .
we take the electron decoherence time ( @xmath313 ms ) and the two nuclear spin intrinsic dephasing times ( @xmath287 ms and @xmath314 ) . as a rough estimate we approximate the three processes for the three qubits by rates and add them to obtain a final decay time @xmath315 ms .
the typical fidelity for the @xmath68 ms quantum error correction protocol becomes @xmath316 .
this corresponds to estimated process fidelity @xmath317 , similar to the observed value .
the average gate fidelity for the 10 nuclear gates in the error correction protocol is estimated from @xmath318{2f_{est}-1 } + 1/2 = 0.94 $ ] . 99 van der sar , t. et al . ,
nature * 484 * , 82 ( 2012 ) .
bernien , h. et al .
lett . * 108 * , 043604 ( 2012 ) .
smeltzer , b. , mcintyre , j. , childress , l. , phys . rev . a 80 ( 5 ) , 050302 ( 2009 ) .
jacques , v. et al .
102 ( 5 ) , 057403 ( 2009 ) .
taminiau , t.h .
et al . , phys . rev .
lett . * 109 * , 137602 ( 2012 ) .
bar - gill , n. , pham , l.m . ,
jarmola , a. , budker , d. , walsworth , r.l . , nat
. commun . * 4 * , 1743 ( 2013 ) .
bernien , h. et al . ,
nature * 497 * , 7447 ( 2013 ) .
aslam , n. , waldherr , g. , neumann , p. , jelezko , f. , wrachtrup , j. , new j. phys . * 15 * , 013064 ( 2013 ) .
aslam , n. et al .
* 107 * , 090401 ( 2011 ) .
pfaff , w. et al .
, nature phys .
* 9 * , 29 ( 2013 ) . | quantum registers of nuclear spins coupled to electron spins of individual solid - state defects are a promising platform for quantum information processing [ 1 - 13 ] .
pioneering experiments selected defects with favourably located nuclear spins having particularly strong hyperfine couplings [ 4 - 10 ] . for progress towards large - scale applications , larger and deterministically available nuclear registers are highly desirable . here
we realize universal control over multi - qubit spin registers by harnessing abundant weakly coupled nuclear spins .
we use the electron spin of a nitrogen - vacancy centre in diamond to selectively initialize , control and read out carbon-13 spins in the surrounding spin bath and construct high - fidelity single- and two - qubit gates .
we exploit these new capabilities to implement a three - qubit quantum - error - correction protocol [ 14 - 17 ] and demonstrate the robustness of the encoded state against applied errors .
these results transform weakly coupled nuclear spins from a source of decoherence into a reliable resource , paving the way towards extended quantum networks and surface - code quantum computing based on multi - qubit nodes [ 11,18,19 ] .
electron and nuclear spins associated with defects in solids provide natural hybrid quantum registers [ 3 - 11 ] .
fully - controlled registers of multiple spins hold great promise as building blocks for quantum networks [ 18 ] and fault - tolerant quantum computing [ 19 ] .
the defect electron spin enables initialization and readout of the register and coupling to other ( distant ) electron spins [ 11,18 ] , whereas the nuclear spins provide well - isolated qubits and memories with long coherence times [ 8,9,11 ] .
previous experiments relied on selected defects having nuclear spins with strong hyperfine couplings that exceed the inverse of the electron spin dephasing time ( @xmath0 ) . with these strongly coupled spins , single - shot readout [ 9,10,20 - 22 ] and entanglement [ 9,11 ] were demonstrated .
however , the number of strongly coupled spins varies per defect and is intrinsically limited , so that universal control has so far been restricted to two - qubit registers [ 4,7 ] and the required control of multi - qubit registers has remained an open challenge . here we overcome this challenge by demonstrating universal control of weakly coupled nuclear spins ( unresolved hyperfine coupling @xmath0 ) .
we use the electron spin of single nitrogen - vacancy ( nv ) centres in room - temperature diamond to selectively control multiple carbon-13 ( @xmath1c ) nuclear spins in the surrounding spin bath ( fig .
1a ) .
with this new level of control we realize multi - qubit registers by constructing high - fidelity unconditional and electron - controlled gates , implementing initialization and readout , and creating nuclear - nuclear entangling gates through the electron spin . finally , we demonstrate the power of this approach by implementing the first quantum - error - correction protocol with individual solid - state spins . ;
@xmath2,@xmath3 ) and weakly coupled @xmath1c nuclear spins ( @xmath4 ; state @xmath5 and hyperfine interaction @xmath6 for nuclear spin @xmath7 , see methods for values ) .
all gates on nuclear spins are implemented by sequences of @xmath8 pi - pulses on the electron spin spaced by a time @xmath9 ( methods ) .
( b ) the electronic coherence as a function of the total sequence length .
the number of pi - pulses @xmath8 is increased for fixed @xmath10 , which is representative for our gates .
@xmath11 khz is the @xmath1c larmor frequency .
the 1/e time is @xmath12 ms .
( c ) nuclear spin initialization by swapping the electron state , @xmath13 or @xmath14 , onto the nuclear spin .
the controlled gates ( @xmath15 ) are @xmath16-rotations by @xmath17 with a direction conditional on the electron spin state ( methods ) .
the final electron spin re - initialization by a @xmath18s laser pulse ( labelled `` init . '' ) preserves the nuclear spin polarization ( @xmath19 values under illumination : @xmath20 ms for nuclear spin 1 and @xmath21 ms for nuclear spin 2 , supplementary information ) .
( d ) electron ramsey measurements without nuclear spin initialization and with nuclear spin 1 initialized in @xmath22 or @xmath23 , and ( e ) with nuclear spin 1 initialized in @xmath23 and nuclear spin 2 in @xmath24 or @xmath25 . all error bars and uncertainties in this work
are @xmath26 . ]
we have used dynamical decoupling spectroscopy [ 23 - 25 ] to characterize the nuclear spin environment of a total of three nv centres , including one with an additional strongly coupled @xmath1c spin ( supplementary information ) . to demonstrate the universality of our approach to create multi - qubit registers ,
we have realized initialization , control and readout of three weakly coupled @xmath1c spins for each nv centre studied ( supplementary information ) .
below we consider one of these nv centres in detail and use two of its weakly coupled @xmath1c spins to form a three - qubit register for quantum error correction ( fig .
1a ) .
our control method exploits the dependence of the nuclear spin quantization axis on the electron spin state due to the anisotropic hyperfine interaction ( see methods for hyperfine parameters ) , so that no radio - frequency driving of the nuclear spins is required [ 23 - 27 ] .
all nuclear gates are implemented by pulse sequences of the form ( @xmath27 where @xmath28 is a microwave pi - pulse on the electron spin , @xmath9 is the inter - pulse delay and @xmath8 is the total number of pulses in the sequence .
each nuclear spin is controlled by precisely choosing @xmath29 in resonance with that spin s particular hyperfine interaction .
the target spin , the type of gate ( conditional or unconditional ) and the rotation axis ( @xmath16- or @xmath30-rotation ) are determined by the value of @xmath29 ; the total rotation angle is determined by @xmath8 ( methods ) .
crucially , these sequences at the same time decouple the electron from the other nuclear qubits and the environment [ 7 ] ; these decoherence - protected gates are selective and allow the full electron coherence time @xmath31 to be exploited ( @xmath12 ms , fig .
1b ) .
the gates are thus not limited by the electron spin dephasing time @xmath32s or hahn echo time @xmath33 and do not require strong coupling . to initialize the nuclear spins we first prepare the electron spin in @xmath34 by optical pumping ( supplementary information ) , then swap the electron state onto the nuclear spin , and finally re - initialize the electron spin ( fig .
1c ) .
we characterize the nuclear initialization by preparing the electron spin in a superposition state and letting it evolve in a ramsey - type experiment . without initialization a single - frequency oscillation with a gaussian decaying envelope
is observed , confirming that the nv centre feels a decohering bath of weakly coupled spins ( fig .
1d ) .
initializing one of the nuclear spins in the @xmath35 ( @xmath14 ) state ( fig .
1d and 1e ) , we increase ( decrease ) the oscillation frequency because the magnetic field at the electron is enhanced ( reduced ) due to the hyperfine interaction .
the oscillations also persist longer as quasistatic fluctuations of the two nuclear spins are suppressed [ 28 ] , increasing the electronic dephasing time to @xmath36s . from this data , we obtain state initialization fidelities of @xmath37 and @xmath38 for nuclear spin 1 and 2 respectively ( see methods ) .
-rotation is implemented by an off - resonant sequence ( @xmath39 ) with a variable number of pulses @xmath8 .
( b ) nuclear spin state tomography is performed by mapping the @xmath40 , @xmath41 and @xmath42 expectation values onto the electron spin and reading out the electron ( shaded gates are optional basis rotations ) .
( c - d ) measurement of @xmath40 , @xmath41 and @xmath42 as function of the free - evolution time .
the oscillations in @xmath40 and @xmath41 confirm the selective control and readout of the targeted nuclear spins .
the amplitude yields a combined readout and initialization fidelity of @xmath43 for spin 1 in ( c ) and @xmath44 for spin 2 in ( d ) .
curves are sinusoidal fits .
see supplementary information for a complete data set with three nuclear spins for each of the three nv centres studied , demonstrating the universality of the control method .
( e ) characterization of the conditional gate for nuclear spin 1 .
the nuclear spin rotates about @xmath16 with opposite directions for @xmath34 ( without shaded gates ) and @xmath45 ( with shaded gates ) .
time for a @xmath46-rotation : @xmath47s .
( f ) unconditional gate for nucleus 1 ; the rotation is independent of the electron state .
time for a @xmath17-rotation : @xmath48s .
see supplementary information for gates on nuclear spin 2 .
results are not corrected for initialization or readout fidelities . ]
next we demonstrate the measurement of the individual nuclear spin states and verify that we observe two distinct @xmath1c spins by performing nuclear free - evolution experiments ( fig .
2a - d ) .
the oscillations in the expectation values @xmath40 and @xmath41 show that the nuclear spins states are successfully read out .
the precession frequencies , @xmath49 khz for nuclear spin 1 ( fig .
2c ) and @xmath50 khz for nuclear spin 2 ( fig .
2d ) , are different and agree with the average of @xmath51 ( for @xmath34 ) and @xmath52 ( for @xmath45 ) , as expected because the electron spin is continuously flipped .
@xmath53 is the parallel component of the hyperfine interaction ( methods ) and @xmath11 khz is the bare nuclear larmor frequency .
these results confirm that we selectively address the two targeted @xmath1c spins .
universal control requires both conditional and unconditional gates , while maintaining a high degree of coherence for all qubits in the register . to characterize our gates
, we initialize the nuclear spins , prepare the electron spin either in @xmath34 or in @xmath45 and apply a gate with a variable number of pulses . for the conditional gate
, @xmath41 oscillates in anti - phase for the two electron states : the nuclear spin rotates around @xmath16 in a direction that depends on the initial electron state ( fig .
2e ) .
in contrast , for the unconditional gate the rotation direction is independent of the electron state ( fig .
2f ) .
the slow decay of the oscillations indicates that high gate fidelities are possible ( @xmath54 for a nuclear @xmath28/2-rotation ) , enabling us to explore multi - gate sequences that implement nuclear - nuclear gates and quantum error correction . and two - qubit tomography is performed by mapping the @xmath55 combinations of pauli operators @xmath56 onto the electron spin ( supplementary information ) .
after correction for single - qubit initialization and readout fidelities , the state fidelity with the target state is @xmath57 , indicating that the sequential initialization and two - qubit readout are accurate .
( b ) entangling gate between nuclear spins by coherently swapping the state of the electron onto nuclear spin 1 .
the gate consists of @xmath58 electron operations over @xmath59s ( excluding initialization and tomography ) .
the nuclear spin coherence is preserved during electron spin re - initialization ( a @xmath18s laser pulse ) ; @xmath60 values under illumination are @xmath61s and @xmath62 ms for nuclear spin 1 and 2 respectively ( supplementary information ) .
the grey bars depict the target state . ]
to realize quantum gates between the nuclear spins [ 27,29 ] , whose mutual interaction is negligible , we use the electron spin as a quantum bus .
we first verify that both nuclear spins can be prepared and read out in the same experiment by initializing the spins in an eigenstate and performing state tomography by mapping the two - qubit correlations onto the electron spin ( fig .
3a ) .
we then implement entangling gates through an electron controlled gate on nuclear spin 2 and a subsequent coherent swap gate between the electron and nuclear spin 1 ( fig .
3b ) .
the tomography reveals strong correlations between the nuclear spins with near - zero single - qubit expectation values , a clear signature of an entangling gate .
the fidelity with the target state is @xmath63 ( initialization and readout corrected ) , demonstrating that the gate can take a pure input state into an entangled state of nuclear spins .
finally , we implement a quantum - error - correction protocol that protects a quantum state from bit - flip errors by encoding it in a @xmath64-qubit state and correcting errors through majority voting ( fig .
4a ) .
such protocols have been realized with nuclear magnetic resonance [ 14,15 ] , trapped ions [ 16 ] and superconducting qubits [ 17 ] , but have so far been out of reach for individual solid - state spins due to a lack of multi - qubit control .
we compose this protocol from one- and two - qubit gates ( fig .
4b ) and separately confirm that the constructed doubly - controlled gate flips the state around x only if the control qubits ( nuclear spins ) are in @xmath65 ( fig .
4c ) .
is encoded in an entangled state using two ancilla qubits .
potential errors @xmath66 are detected by decoding and are corrected based on a doubly - controlled not gate .
( b ) our implementation of the quantum - error - correction protocol in ( a ) .
the doubly - controlled gate ( blue ) is constructed using 4 controlled gates as the final ancilla states are irrelevant .
the experiment consists of @xmath67 electron operations in @xmath68 ms ( excluding initialization ) .
( c ) characterization of the doubly - controlled gate ( blue gates in ( b ) only ) .
the average output fidelities for @xmath69 , for @xmath70 and for @xmath71 are shown for the four ancilla basis states .
the average process fidelity with the targeted action is @xmath72 .
( d ) process fidelity for errors applied to nucleus 1 , to nucleus 2 , or to the electron spin ( with and without additional flip of nuclear spin 1 ) .
grey data and fit are @xmath73 which sets the average value for the expected oscillations if no errors are corrected .
( e ) process fidelity for errors simultaneously applied to all three qubits with error probability @xmath74 .
purple : without error correction .
blue : with error correction .
grey : for ideal robustness against errors .
error bars are given by the symbol size ( typical standard deviation @xmath75 ) .
inset : deviation of the error correction data from a linear curve .
all curves in ( d ) and ( e ) are fits to the model in the methods section . ]
we first characterize the effect of errors on each individual qubit .
the applied errors are rotations around the x - axis by an angle @xmath76 with a random sign ( @xmath77 clockwise , @xmath77 anticlockwise ) and therefore represent a decoherence - type process with a strength determined by @xmath76 .
we prepare @xmath78 input states @xmath79 and @xmath71 , measure the corresponding fidelities @xmath80 of the output states and calculate the process fidelity @xmath81 with the identity process : @xmath82 without error correction , errors on the data qubit ( electron spin ) are expected to result in an oscillation about @xmath83 because only the @xmath69 states are unaffected by the applied errors .
instead , the experimental process fidelity with error correction always remains above @xmath84 , even for a completely randomizing error ( @xmath85 ) , indicating that the state is partly recovered ( fig .
4d ) .
if one of the ancilla qubits ( nuclear spins ) is also flipped , an oscillation about @xmath84 is observed ; the error correction is effectively turned off because the protocol can not correct two - qubit errors . to quantitatively determine the effectiveness of the error correction we analyze it in terms of the three probabilities @xmath86 that an applied error on qubit @xmath87 is successfully corrected and a decoherence / depolarization process during the error - correction protocol itself ( methods ) .
the model accurately fits the data and gives @xmath88 , @xmath89 and @xmath90 for errors on the electron , nucleus 1 and nucleus 2 respectively .
crucially , the average probability @xmath91 is well above @xmath92 , demonstrating that the process is robust against applied single - qubit errors and that the entropy associated with the errors is successfully shuttled to the ancilla qubits .
we further demonstrate the robustness by applying errors simultaneously on all three qubits ( fig .
4e ) . without error correction ,
i.e. without doubly - controlled gate , a linear dependence is observed and a fit to the expected form gives @xmath93 in excellent agreement with @xmath94 expected for no robustness to errors . with error correction a markedly slower initial decay and a non - linear behaviour with @xmath95
is obtained .
this suppression of the linear dependence is a key characteristic of quantum error correction .
the deviation from @xmath96 is mainly due to imperfect nuclear initialization , which might be improved by repeated initialization steps ( supplementary information ) or projective measurements [ 9,21 ] .
we calculate @xmath97 for ideal initialization fidelity ( supplementary information ) . without applied errors , decoherence and depolarization during the protocol itself ( more than @xmath98 electron operations over @xmath68 ms ) result in a process fidelity of @xmath99 , corresponding to an average gate fidelity of @xmath100 for the @xmath101 nuclear spin gates .
the main source of infidelity is electron decoherence ( @xmath102 ms , fig .
1b ) , which is likely phonon - induced [ 30 ] and limits the average gate fidelity to @xmath103 .
nuclear spin dephasing further reduces the fidelity to @xmath104 , close to the observed value ( supplementary information ) .
the electronic coherence time is greatly increased at cryogenic temperatures , at which @xmath105 ms ( single nv ) [ 18 ] and @xmath106 s ( ensembles ) [ 30 ] have already been reported .
nuclear spin dephasing can be mitigated by decoupling nuclear - nuclear interactions ( @xmath33 measurements in supplementary information ) . with such future improvements ,
our results can be used to protect entangled states of solid - state spins . in conclusion , we have established universal control over weakly coupled nuclear spins that were previously regarded as a source of decoherence .
these results provide multiple qubits per defect with high certainty and are compatible with control of the intrinsic nitrogen spin and potential strongly coupled @xmath1c spins .
our techniques can be applied to a wide variety of other electron - nuclear spin systems [ 2,3,10,13 ] .
the resulting reliable multi - qubit registers can be combined with recently demonstrated coherent coupling between ( distant ) electron spins [ 11,18 ] to realize novel surface - code quantum - computation architectures that use four qubits per defect node [ 19 ] and extended quantum networks for long - distance quantum communication . |
galaxy modelling can be broadly separated into two categories : light profile fitting and kinematic dynamical modelling . in dynamical modelling ,
part or all of the 6-dimensional phase space structure is constrained by kinematic data and dynamical constraints such as the poisson and/or jeans equations . by contrast , profile fitting ( or profiling ) quantifies the _ projected _ structure of a galaxy on the sky , usually by fitting radially - varying density profiles to one - dimensional , azimuthally - averaged profiles or two - dimensional images .
both methods have been used to infer the presence of distinct structural components in galaxies , such as discs , bulges , bars and spiral arms ( so - called decomposition ) .
@xcite is credited as the first to develop a galaxy classification scheme including distinct features or components , including spiral arms and a concentrated central component .
as early as @xcite it had been quantitatively demonstrated , using photoelectric measurements along the major and minor axes , that m31 comprised of a distinct bulge and disc component . in the modern era
, large samples of thousands of galaxies have been decomposed quantitatively into distinct components @xcite .
such datasets have been used to infer distinct evolutionary pathways and mechanisms @xcite for galaxies of different types and compositions and to develop galaxy formation models @xcite . improving these theories
requires both better data ( by volume , depth and quality ) and analysis methods .
this paper offers a new route to reliably divide galaxies into their constituent parts by way of a new publicly released library and interface ( ` libprofit`and profitrespectively ) for galaxy profile modelling and decomposition .
these will be used by the authors for near future large - scale studies , and are being made available to the wider community along with the prospect of practical support and future updates .
the earliest efforts of galaxy structural characterisation concentrated on simple one - dimensional intensity profile fitting , leading to the early discovery of the de - vaucouleurs profile @xcite for describing early - type galaxies and the realisation that disc galaxies had close to exponential profiles @xcite .
@xcite generalised these profiles into the @xmath2 law that remains widely popular today , commonly referred to as the srsicprofile @xcite .
increasingly complicated one - dimensional component fitting came in work by @xcite and @xcite , where galaxies were decomposed into distinct components rather than treated as having a single light profile .
the earliest efforts at two - dimensional galaxy profiling came with the work of @xcite , @xcite , @xcite , @xcite and @xcite .
the latter four approaches are broadly similar in application , and are recognisably similar to modern two - dimensional galaxy fitting efforts .
the basic philosophy was the same as for previous one - dimensional work , the aim being to find the distinct light components of galaxies , but here the analysis was made using image pixels directly rather than by fitting azimuthally averaged ellipses of light .
a large number of key insights into galaxy properties have been uncovered through both one - dimensional and two - dimensional structural analysis .
early work found that the disc component of galaxies is very well represented by an exponential drop - off in light ( e.g. * ? ? ?
* ; * ? ? ?
of recent interest in the astronomy literature is the relationship between galaxy mass and size , both globally @xcite and for individual components @xcite .
there is also a large body of work investigating the luminosity - surface brightness relation for different classes of galaxy ( see reviews by * ? ? ?
* ; * ? ? ?
* ) , as well as investigating the distribution of mass for different types of structure ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . such global morphological properties of galaxies are finally being utilised and predicted by the newest theoretical work , both in the regime of semi - analytic models @xcite and cosmological hydrodynamic simulations @xcite
. given the constraints that galaxy morphology can offer theory , it will be increasingly important that well quantified structural measurements for galaxy properties are extracted from current and future large - area , deep , high - resolution photometric surveys ( e.g. kilo degree survey ( kids ) , dark energy survey ( des ) , hyper suprime camera ( hsc ) and large synoptic survey telescope ( lsst ) : * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* respectively ) . there remains some debate regarding the relative merits of one - dimensional and two - dimensional image analysis ( e.g. * ? ? ?
. this introduction will not serve as a thorough discussion of the various issues , but we will discuss a few key points .
one - dimensional fitting relies heavily on an additional step in the analysis : the galaxy must be collapsed into a one - dimensional profile in some manner , usually using software such as iraf s ellipse task .
it is impossible to do this in an entirely model independent manner when the galaxy is anything other than smoothly changing isophotes ; however , such an approach can easily cope with a smoothly varying orientation in the isophotes .
a caveat to this is that it is unclear formally how the atmospheric point - spread function ( psf ) should be treated when one - dimensional fitting ( note that there are analytic solutions for the specific case of face - on convolution , see * ? ? ?
even a galaxy constructed from perfect concentric ellipses will become artificially circular towards the profile centre when convolved with a circular psf . whilst this effect is captured from the one - dimensional profile information ,
it is non - trivial to properly propagate this information such that you measure the intrinsic two - dimensional profile properties of the galaxy . collapsing two - dimensional information into a one - dimensional form is almost always a lossy process ( it can be no better than lossless ) , and real azimuthal profile fluctuations may be lost entirely .
correctly propagating errors is also non - trivial , and in practice these are often not used at all ( an approach advocated in * ? ? ?
there are two other critical aspects of this collapsing process that can have a dramatic effect on galaxy profiling : the centre of the galaxy must be fixed exactly a - priori and the components must share the same centre . the freedom to fit for the galaxy centre is often important for capturing the steepest inner parts of a galaxy properly ( especially if the centre is sub - pixel ) and galaxy components sometimes do not share exactly the same centres @xcite .
two - dimensional fitting is closer in spirit to bayesian generative model fitting , e.g. creating a two - dimensional distribution of flux that is then propagated through the atmospheric psf . as such it is an attractive approach to quantifying image components
however , the isophotes of popular galaxy profile functions are poorly behaved when integrating over two - dimensional pixels , requiring relatively expensive numerical estimation , while complex geometry ( like rotating isophotes ) is more difficult to capture fully , with most codes ignoring this issue entirely . in the regime where the true galaxy is well represented by the two - dimensional model being used a two - dimensional fit _
should _ be preferable . in principle , effects like twisting isophotes can be incorporated into a two - dimensional generative model technique ( see * ? ? ?
* ) , but this is rarely attempted in practice . because of these relative advantages and disadvantages between one - dimensional and two - dimensional approaches , their application to galaxy profiling has pragmatically diverged .
one - dimensional codes are more popular in the regime of highly resolved galaxies where complex geometric effects ( like twisting isophotes ) are clearly visible and the issue of psf convolution is less significant . because of the more automated nature of two - dimensional codes ( they do not require a carefully controlled collapse of the data to one - dimensional ) , they have tended to be more popular for large samples of galaxies where the fitting is made with little user interaction ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
our long - term ambition is to decompose galaxy structures from a large number of imaging data sources for many millions of galaxies , so our use - case is squarely in the highly - automated regime .
for this reason in particular , we focussed our efforts on two - dimensional decomposition codes .
current publicly - available two - dimensional galaxy profile fitting codes include budda @xcite , galfit@xcite , galmorph @xcite , gim2d @xcite , and imfit @xcite - all of which support the commonly - used srsicprofile .
these codes share their origins in the early two - dimensional profiling efforts of @xcite and @xcite and have been utilised in a number of notable large - scale galaxy studies : @xcite .
notable alternatives include gasphot @xcite , which is a one - dimensional fitting code , and mge @xcite , which abandons traditional srsicprofile bulge - disc decomposition for more flexible ( but arguably less physical ) concentric gaussians . in all of the other codes ,
the model is treated generatively by numerically integrating a radially - varying profile over a rectangular pixel grid ; in practice , most codes improve integration accuracy by oversampling the native pixel grid , particularly near the centre of the galaxy .
once this idealised model is integrated , it can be convolved with a target psf , either via brute - force or fast fourier transform ( fft ) techniques .
these codes all have their positive features and weaknesses , and differ in fine aspects such as how the likelihood is computed and maximised .
however , no existing code achieves the desired design requirements that our new code , profit , aims to address , namely that it : * be legally open source under a gpl , bsd , mit license or similar , * be practically collaborative and hosted in a publicly available location such as github or bitbucket , * be written in a low - level modern object - oriented language , aiding modularity , extensions and robustness for complex software , * have the core image generation code clearly separated from the optimisation code , available as a shared library , and accessible via an easy - to - use api , * offer a useful range of built - in models and have a well documented mechanism for adding more , * support sub @xmath31% flux weighted error model image generation for built - in models , and be testable in this regard , * be fast at generating these model images , * have a range of options to define likelihoods , i.e. normal , poisson , chi - squared and student - t statistics , * be able to fit a large number of components in a flexible manner , e.g. components being fixed , free or locked together with other components , * be untied to a specific optimiser , i.e. able to use a range of ( user - selectable ) optimisation engines , * be able to accept prior distributions of an arbitrary type for model parameters ( vital for bayesian model evaluation ) , and * have the ability to fit parameters in linear or log space as deemed appropriate by the user for the problem ( vital for bayesian model evaluation ) , * be able to handle parameter limits and constraints in flexible manner ( e.g. prevent some parameters getting larger than other etc ) , * have a core interface that does not require the use of text file inputs , i.e. there is a completely functional interactive interface to the software ( though scripting with files should be allowable ) , * simple to parallelise . the reasons for desiring the above features are multi - faceted , but broadly it is to aid the structural analysis of large amounts of data obtained from new and future imaging surveys such as kids , des , hsc and lsst . also , the addition of optional priors and log or linear parameter scaling is paramount for proper bayesian model evaluation .
automating historic code is a complicated affair @xcite , and there are in practice many subtle issues with trying to bootstrap new code from legacy software .
starting from scratch with a new code offers full flexibility in language choice , design flexibility and low - level choice regarding pixel integration and likelihood calculations .
our estimation was it would be at least as much effort to adapt any available open - source code to meet the above requirements that profitdelivers .
this paper discusses the development and application of the core image generation code ( ` libprofit ` , section [ sec : imgen ] ) and the fitting code ( profit , section [ sec : fit ] ) .
the profitcode is then applied to various examples ( section [ sec : examples ] ) , spanning a detailed case - study ( section [ sec : example1 ] ) and the analysis of 10 exemplar bulge - disc galaxies that are included with the profitpackage ( section [ sec : example2 ] ) .
estimations of decomposition fidelity is then made for current and future datasets ( section [ sec : future ] ) .
finally , we discuss and summarise the results of this paper ( section [ sec : discuss ] ) .
the image generation code is contained in a shared library written in c++ called ` libprofit ` . `
libprofit ` enables the user to construct a _ model _ to which _ profiles _ are appended .
each profile can be fine - tuned , and fully describes a component of the resulting image .
after profiles are appended the model is evaluated , and the resulting image can be retrieved . ` libprofit ` cycles through each of the profiles generating individual images for each of them , which are optionally convolved and finally added .
hence it is trivial to describe a one- , two- or multi - component system .
this external interface , internal organisation and execution of the ` libprofit`library is depicted in figure [ fig : libprofit ] using the unified modelling language .
` libprofit ` includes a default set of popular profiles : srsic , core - srsic , broken - exponential , moffat , ferrer , modified empirical king , point spread function ( psf ) and sky ( see figure [ fig : profiles ] ) .
the first six share some key similarities : they can be fully described by an analytic radial profile which is then evaluated in two - dimensions over pseudo - elliptical isophotes .
they also share similarities in terms of how they must be evaluated in a given pixel : if the gradient of the model image is varying rapidly over the scale of a pixel then using a simple trapezoidal integration scheme ( effectively taking the flux value predicted at the centre and assuming this to reflect the average for the pixel ) can be error prone . for this reason
the srsic , core - srsic , broken - exponential , moffat , ferrer and king profiles all share the same core c++ code that evaluates the flux in a pixel and determines whether more accuracy is required .
this means new profiles can be defined quite simply within ` libprofit ` , effectively reducing to the one - dimensional form of the radial profile and a scheme to calculate the total flux within the profile . for the srsic , moffat and ferrer the total integration are known ( or have been calculated by the authors ) analytically , so this makes them particularly simple cases .
all eight profiles are discussed in more detail below , with the caveat that the detail of the integration scheme discussed with respect to the srsicprofile also applies to the core - srsic , broken - exponential , moffat , ferrer and king . for galaxy fitting purposes
the srsicprofile @xcite is a modern day work - horse , having enough flexibility in profile characteristics to fit a large variety of galaxy types ( see e.g. * ? ? ?
the intensity profile is usually parameterised as : @xmath4 \right\},\ ] ] where @xmath5 is the intensity of the profile at @xmath6 ( the radius containing half of the flux ) , @xmath0 is the srsicindex ( which has special cases of normal / gaussian with @xmath7 , exponential with @xmath8 and de - vaucouleurs with @xmath9 ) and @xmath10 is a derived parameter that ensures the correct integration properties at @xmath6 and is the quantile at which the gamma probability distribution function integrates to 0.5 given a shape parameter of @xmath11 .
this can be computed directly using high - level interfaces to statical distribution libraries ( e.g. in r@xmath12 qgamma(0.5 , 2*n ) ) .
@xmath13 is the modified radius where we want to make the evaluation , which in one dimension is simply the radius from the profile centre . in two dimensions
@xmath13 has a more complex form : @xmath14 where @xmath15 is the two - dimensional location where we wish to make the evaluation and @xmath16 is the location of the profile centre .
when the two - dimensional profile is non - circular the isophotal contours become elliptical annuli , so calculating @xmath13 becomes a series of computations that rotate and circularise the annuli .
when the major - axis angle @xmath17 is defined such that it is 0 degrees vertically and increases positively as the galaxy is rotated counter - clockwise ( as per galfit ) these steps become : @xmath18 where @xmath19 is the minor to major axis ratio ( so always a number between 0 and 1 , where 0 is an infinitely thin line and 1 is a circle or disc ) . the final modifier to the standard one - dimensional srsicprofile intensity is to allow for apparent ` boxiness ' .
boxiness is often used to create the more rectangular visual appearance of galaxy bulges or elliptical galaxies .
it effectively modifies the unit circle ( or ellipse ) away from the standard @xmath20 norm and either towards a diamond / kite appearance ( @xmath21 norm ) or a boxy appearance ( @xmath22 norm ) .
this is achieved though a small manipulation of the final @xmath13 calculation above : @xmath23^\frac{1}{2+b},\ ] ] where @xmath24 is the boxiness , which for practical ( as opposed to mathematical ) purposes is usually allowed to vary between -1 and 1 ( thus covering the full range of apparent visual boxiness ) .
figure [ fig : boxiness ] shows the effect of different negative and positive values of boxiness .
a consideration for the user is the relative expense of computing the boxiness .
powers that differ from the @xmath20 norm are much more expensive to compute , in particular non - integer values might be a factor of a couple slower when computing a model image . in most standard c math libraries ( libm )
it is faster to make successive calls to sqrt and csqrt rather than a single call to pow . `
libprofit ` acknowledges this fact and calls the most efficient set of functions as appropriate for certain exponents . whilst the above is an accurate description of the computations that need to be made to calculate an arbitrary two - dimensional srsicprofile , they are not optimal . ` libprofit ` uses rotation matrix arithmetic internally to achieve the above steps efficiently
, projecting pixel coordinates onto the major and minor axes of the ellipse .
this can be a factor of a few faster than explicitly computing rotations with expensive calls to trigonometric functions .
it is not necessary to parameterise the profile using @xmath5 to define the normalisation , as this is rarely intuitive . instead
, the user can specify the total flux contained within a profile in magnitudes ( @xmath25 ) given a magnitude zero - point ( @xmath26 ) , and calculations are made internally to convert this to the correct value of @xmath5 .
these include appropriate modifications for the axis ratio ( @xmath19 ) and boxiness ( @xmath24 ) : @xmath27 where @xmath28 is the beta function defined as @xmath29 , and @xmath30 is the standard gamma function .
if the profile has no boxiness then the above is simplified by noting that @xmath31 and therefore the denominator term for @xmath32 disappears . to aid flexibility
it is also possible in ` libprofit ` to define the profile normalisation using the mean surface brightness within @xmath6 ( @xmath33 ) , where : @xmath34 ` libprofit ` follows the convention that @xmath33 is determined for an elliptical radius containing half of the objects flux , i.e. it is not modified for boxiness .
this can be useful when parameterising a galaxy where the galaxy magnitude is entirely unknown since galaxies span a much smaller dynamic range in surface brightness than total magnitude @xcite . in conclusion ,
the srsicprofile is specified fully through the provision of eight parameters that are parsed into the ` libprofit ` image building routine as a list structure of equal length vectors .
below is an example specification of a two - component srsicmodel in the rimplementation of profitthat directly calls the ` libprofit`api to generate model images , where @xmath35xcen , @xmath36ycen , @xmath37mag , @xmath38re , @xmath39nser , @xmath40ang , @xmath41axrat and @xmath42box : this model is shown in figure [ fig : examplesersicmodel ] . as expected , the high srsicindex bulge component dominates the flux near the galaxy centre as well as at very large radii , due to the shallow slope of an @xmath43 srsicprofile .
this is visible as side lobes in the surface brightness contours jutting out from the visually dominant disc component running diagonally across the image .
closely related to the srsicprofile is the core - srsicprofile introduced by @xcite .
this was designed to parameterise the strong inner core ( i.e. suppression compared to pure srsic ) profile of well - resolved elliptical galaxies .
the inner components of such galaxies had previously been parameterised well with the nuker profile @xcite , at least in the inner parts of the profile .
however , the nuker profile fits poorly at large radii and does not have a well behaved integral to infinite radius for typical parameters .
the core - srsicaddresses these issue by allowing the inner component to behave in a nuker - like manner , with the outer parts being described by a standard srsicprofile .
the intensity profile can be parameterised as : @xmath44^\frac{b}{a } \exp \left[-b_n \left(\frac{r_m^a+r_b^a}{r_e^a}\right)^\frac{1}{n a}\right],\ ] ] where @xmath45 is the intensity at a parameter dependent reference point ( the centre if @xmath46 ) , @xmath47 is the break radius that controls the point where the outer srsic - like profile transitions to become an inner nuker - like profile , @xmath6 is the standard srsichalf light radius ( and @xmath10 has the same meaning as for the pure srsicprofile ) and @xmath48 controls the strength of transition ( large value is a sharper transition ) and @xmath49 controls the inner power - law .
@xmath13 is the modified radius where we want to make the evaluation , which in one dimension is simply the radius from the profile centre . in two dimensions
@xmath13 has a more complex form where , as with the srsicprofile , we can use this basic form and compute annuli that are modified from pure circular ellipses by introducing geometric transformations that compute an effective @xmath13 , as per equation [ eq : rmod ] .
accounting for the allowed geometric distortions of axial ratio and boxiness , the value of @xmath45 for a given magnitude can be calculated numerically with : @xmath50 the integral above has no simple analytic solution , so it is instead calculated numerically inside ` libprofit ` using either the ror gsl implementation of the quadpack library ( specifically adaptive gauss - kronrod quadrature integration ) .
this is a much better solution than attempting to re - normalise the generated image since much of the flux could exist beyond the confines of the image pixels , especially for large values of @xmath0 .
below is an example of a core - srsicmodel in the rimplementation of profitthat directly calls the ` libprofit`api to generate model images , where @xmath35xcen , @xmath36ycen , @xmath37mag , @xmath51rb , @xmath52re , @xmath39nser , @xmath53a , @xmath54b , @xmath40ang , @xmath41axrat and @xmath42box : the broken - exponential profile is a popular means to parameterise rolling or sharp truncations in the surface profiles of discs , which are usually close to exponential in profile until they enter the moderate- to low - surface brightness regime .
@xcite introduced a simple modification to the standard exponential profile ( which is itself a subset of possible srsicprofiles ) in order to capture the range of truncations observed .
the intensity profile can be parameterised as : @xmath55^{\frac{1}{a}\left(\frac{1}{h_1}-\frac{1}{h_2}\right)},\end{aligned}\ ] ] where @xmath56 is the intensity at the centre of the profile , @xmath57 is the scale length of the inner exponential profile , @xmath58 is the scale length of the outer exponential profile , @xmath47 is the radius at which the profile transitions from the inner to outer part and @xmath48 controls the sharpness of this transition .
@xmath13 is the modified radius where we want to make the evaluation , which in one dimension is simply the radius from the profile centre . in two dimensions
@xmath13 has a more complex form where , as with the srsicprofile , we can use this basic form and compute annuli that are modified from pure circular ellipses by introducing geometric transformations that compute an effective @xmath13 , as per equation [ eq : rmod ] .
accounting for the allowed geometric distortions of axial ratio and boxiness , the value of @xmath56 for a given magnitude can be calculated analytically with : @xmath59 where @xmath25 , @xmath26 and @xmath60 have the same meanings as per equation [ eq : sersnorm ] .
an example specification of a broken exponential model in the rimplementation of profitthat directly calls the ` libprofit`api to generate model images , where @xmath35xcen , @xmath36ycen , @xmath37mag , @xmath61h1 , @xmath62h2 , @xmath63rb , @xmath53a , @xmath40ang , @xmath41axrat and @xmath42box , is : the moffat function was designed as a means to parameterise a telescope point spread function @xcite , and is often used on occasions when the psf is predicable and when you might not have many point sources within the field - of - view to estimate it empirically ( e.g. for space telescopes with small fields - of - view like the hubble space telescope ) . whilst this was the purpose behind its design , in practice in modern applications it rare to see analytic approximations of the psf ( such as the moffat ) used , the modern preference being for empirically derived estimates .
regardless , for reasons of flexibility it is included as a profile option in ` libprofit ` .
the moffat is a re - parameterisation of the bivariate student - t distribution , meaning it creates a normal - like core with lorentzian - like wings .
the intensity profile can be parameterised as : @xmath64^{-c},\ ] ] where @xmath65 where @xmath56 is the intensity at the centre of the profile , @xmath66 is the profile concentration ( @xmath67 is pure lorentzian , @xmath68 is pure normal ) and @xmath69 is the full - width half max of the profile across the major - axis of the intensity profile .
@xmath13 is the modified radius where we want to make the evaluation , which in one dimension is simply the radius from the profile centre . in two dimensions @xmath13 has a more complex form where , as with the srsicprofile
, we can use this basic form and compute annuli that are modified from pure circular ellipses by introducing geometric transformations that compute an effective @xmath13 , as per equation [ eq : rmod ] .
thus in ` libprofit ` the moffat function need not be circular , which is often the case for the psf of a wide - field telescope where corner distortions tend to be radial or tangentially aligned with respect to the field centre .
accounting for the allowed geometric distortions of axial ratio and boxiness , the value of @xmath56 for a given magnitude can be calculated analytically with : @xmath70 where @xmath25 , @xmath26 and @xmath60 have the same meanings as per equation [ eq : sersnorm ] .
an example specification of three moffat models in the rimplementation of profitthat directly calls the ` libprofit`api to generate model images , where @xmath35xcen , @xmath36ycen , @xmath37mag , @xmath71fwhm , @xmath72con , @xmath40ang , @xmath41axrat and @xmath42box , is : since the moffat profile is almost exclusively used to model the image psf , ` libprofit ` is hard - coded to never convolve the profile with another psf , even if this is provided .
this is to help eliminate potential user error .
the modified ferrer profile ( a type of projected potential derived from ferrers functions , as discussed in @xcite ) is a useful parameterisation of galaxy features that have very strong drop - offs , most typically bar structures . whilst it should formally be called the ferrers profile
, it is very common for it to be written as ferrer ( the assumption is that authors have erroneously attributed the final ` s ' as possessive , but the name derives from norman macleod ferrers ) . for convenience
, the profile can be named as ferrer or ferrers in ` libprofit`and in higher - level interfaces .
the intensity can be parameterised as : @xmath73^{a},\ ] ] where @xmath56 is the intensity at the centre of the profile , @xmath74 is the outer truncation radius ( the profile is 0 beyond this radius ) , @xmath48 controls the global power - law slope to the profile centre , and @xmath49 controls the strength of truncation as @xmath75 approaches @xmath74 .
@xmath13 is the modified radius where we want to make the evaluation , which in one dimension is simply the radius from the profile centre . in two dimensions
@xmath13 has a more complex form where , as with the srsicprofile , we can use this basic form and compute annuli that are modified from pure circular ellipses by introducing geometric transformations that compute an effective @xmath13 , as per equation [ eq : rmod ] . for the ferrer profile it is common to introduce some positive boxiness to reflect the visually rectangular isophotes of galaxy bars .
accounting for the allowed geometric distortions of axial ratio and boxiness , the value of @xmath56 for a given magnitude can be calculated analytically with : @xmath76 below is a sample specification of a ferrer model in the rimplementation of profitthat directly calls the ` libprofit`api to generate model images , where @xmath35xcen , @xmath36ycen , @xmath37mag , @xmath77rout , @xmath53a , @xmath54b , @xmath40ang , @xmath41axrat and @xmath42box : the modified king profile in ` libprofit ` shares its origins with the empirically motivated two - dimensional profile presented in @xcite , and remains a popular function for parameterising globular cluster light profiles .
it should be emphasised that it is _ not _ the same as the equally popular profile presented in @xcite , the latter being parameterised in three - dimensions , theoretical in origin and non - trivial to project . ` libprofit ` uses the same basic modification to the @xcite as used in galfit , where the intensity can be parameterised as : @xmath78^\frac{1}{a}}-\frac{1}{\left[1+\left(\frac{r_t}{r_c}\right)^2\right]^\frac{1}{a}}\right\}^a,\ ] ] where @xmath45 is the intensity at a parameter dependent reference point ( the centre if @xmath79 ) , @xmath80 is the core radius , @xmath81 is the outer truncation radius ( the profile is 0 beyond this radius ) and @xmath48 controls the global power - law . when @xmath82 , this parameterisation is identical to the one presented originally in @xcite .
@xmath13 is the modified radius where we want to make the evaluation , which in one dimension is simply the radius from the profile centre . in two dimensions
@xmath13 has a more complex form where , as with the srsicprofile , we can use this basic form and compute annuli that are modified from pure circular ellipses by introducing geometric transformations that compute an effective @xmath13 , as per equation [ eq : rmod ] .
accounting for the allowed geometric distortions of axial ratio and boxiness , the value of @xmath56 for a given magnitude can be calculated analytically with : @xmath83 an example specification of a modified king profile in the rimplementation of profitthat directly calls the ` libprofit`api to generate model images , where @xmath35xcen , @xmath36ycen , @xmath37mag , @xmath84rc , @xmath85rt @xmath53a , @xmath40ang , @xmath41axrat and @xmath42box , is : the point spread function profile for an unresolved point source ( referred to as a pointsource within ` libprofit`and profitto distinguish it from the psf ) is a simple prescription requiring the @xmath16 location of the desired point source , the total flux contained in magnitudes @xmath25 and either an empirical psf image or analytic psf model . in the former case ,
the psf image is linearly interpolated onto the specified location in the image ( thus fractional pixel values can be provided ) and is renormalised to ensure the flux contained is correct given the magnitude zero - point ( @xmath26 ) ; otherwise , the point source is accurately integrated with the specified psf model using an analytic profile ( e.g. the moffat , as discussed in section [ subsubsec : moffat ] ) .
an example of a model containing three point sources in the rimplementation of profit , where @xmath35xcen , @xmath36ycen and @xmath37mag , is : numerical integration of an analytic psf is considerably more accurate than interpolation of an empirical psf image , so we generally recommend using the former approach if it is at all possible . as mentioned in section [ subsubsec : moffat ] , it is more typical to see astronomers make use of empirical psfs . since the intrinsic psf is unknown , using such a description will result in a simple linear interpolation to fractional pixel locations . to accommodate this fact ,
` libprofit ` and profitallow for the specification of an oversampled empirical psf .
if psf modelling is required , this offers both a pragmatic and accurate solution .
the simplest profile of all is the sky . in ` libprofit
` this is specified as the flux per pixel without any adjustment made for the magnitude zero - point ( @xmath26 ) , i.e. it is the flat sky level pedestal directly measured from the image in its native flux units ( any subsequent conversions must be made explicitly by the user ) . an example sky model in the rimplementation of profitrequires only the background level bg : currently only flat sky profiles are available in ` libprofit ` and profit , with the assumption that in typical profituse cases the sky is a pedestal term , with no gradient of curvature .
depending on community feedback additional complexity for the sky profile could be added in the future .
however , it is clear that the majority of modern survey image processing involves an explicit sky - subtraction stage which reduces the data to a pedestal - only sky profile ( e.g. see * ? ? ? * ) . like any other profile type it is possible to fit for the sky profile background within ` libprofit ` , however in many use cases the background will be known a - priori and should not be fit .
a common issue with the leaving the sky as a fitted parameter is how degenerately it behaves with srsic - index outer profiles , which also appear close to flat at many multiples of @xmath6 . to guard against this effect
it is wise to fit using pixels which are dominated by galaxy flux .
combining models in ` libprofit ` is simple given the nested list nature of object specification .
a structure containing the srsic , ferrer bar , point - source and sky models is given by : this is the list structure used in r profitto specify models , and in principle an unlimited number of profiles can be provided to create complex model images . in practice , for galaxy image fitting simple models are made of the region of direct interest .
this structure is also the mechanism for providing initial conditions for galaxy image fitting .
this structure can be easily manipulated from outside of the renvironment , or directly within r. figure [ fig : examplecombmodel ] shows the image generated by the example model above .
this is very similar to figure [ fig : examplesersicmodel ] but with the addition of a ferrer bar profile , three bright point sources and a sky background ( which is not visible due to the image scaling ) .
more details on how ` libprofit ` generates these images is provided in the following section . , but with the addition of a ferrer bar profile , three psfs and a sky background . ] the main focus of development effort in ` libprofit ` was optimizing model image generation for speed _ and _ accuracy , particularly for the srsicprofile ( since this is by some margin the most popular general purpose profile for galaxy modelling ) . the first requirement to achieve rapid image generation was that the core code should be written in a fast low - level language . c++
was chosen for these purposes , with the idea that ` libprofit ` should provide a generic low - level library that higher level languages ( e.g. rand python ) can access easily . with c++ chosen as the implementation language , the main considerations left were how to best approach the problem of model image generation algorithmically .
two - dimensional modelling routines are circumspect in how they achieve pixel integration , with most published references stating that inner pixels are ` oversampled ' in order to achieve reasonable accuracy in the pixel flux ( note this is different to the type of oversampling required to ensure accurate image convolution , as discussed in section [ subsec : convolution ] ) . depending on the profile being created the degree of oversampling can vary by many orders of magnitude . getting accurate solutions for the flux in a pixel containing the peak of a steep ( high @xmath0 )
srsicprofile is clearly a very different prospect to estimating the flux of a pixel a large distance from the centre , where the profile is very flat . to combat this problem , ` libprofit ` uses a unified oversampling scheme ( inspired by adaptive quadrature ) for all profiles that have an analytic radial intensity profile - initially the srsic , core - srsic , broken - exponential , moffat , ferrer and king profiles discussed above .
here we concentrate on the evaluation of the srsicprofile , but the basic principles hold true for all radially - varying profiles . in ` libprofit ` all such profiles inherit from a generic class of radial type profiles ( see figure [ fig : profiles ] ) , where the pixel integration scheme is shared in a common method .
the profiles are specified by their radial intensity and total luminosity ( as per section [ sec : profiles ] ) .
this minimises code duplication and simplifies the addition of new profiles to ` libprofit ` , where the user only needs to be able to express the one - dimensional form of the profile shape and a method to integrate it to get the total luminosity ( which can simply be a numerical integration if there is no convenient analytic solution ) .
the qualitative version of the pixel integration scheme is that an oversampling resolution , a recursion depth and an accuracy level is specified for a given pixel .
the grid is initially oversampled by the integer factor provided , for each sub - pixel the gradient along the minor axis ( this being the steepest descent ) is computed and when it is over a tolerance specified by the accuracy value this pixel is selected for recursive oversampling . in this situation the sub - pixel becomes redefined as the parent pixel and the scheme above is repeated .
figure [ fig : pixint ] is a simple schematic of the central @xmath86 pixels being recursively over - sampled by a factor @xmath87 to achieve an accurate numerical integration in the regime where the profile is rising steeply in the core of the galaxy .
the default allowed maximum depth for sub - pixel recursion is 2 , but it can be user - specified .
the default oversampling factor is 8 , meaning at most sub - regions of pixels would be oversampled by a factor @xmath88 .
a special case is made for the central pixel in the srsicprofile ( containing the flux peak ) . to ensure accurate results even for extreme examples of very steep profiles of barely - resolve galaxies , the maximum recursion depth
is increased to 10 , giving potential oversampling of @xmath89 . in extreme cases , where deep recursion is triggered on the very central pixel of each layer
, the practical worst case scenario is that @xmath90 calculations are actually required because only the sub - pixels containing the profile peak require deeper recursion .
this combination of accuracy and speed means ` libprofit ` performs well in model image generation bench - marking whilst maintaining user - definable accuracy with sensible defaults .
detailed comparisons were made with reference to galfit , where the target was to achieve similar or better pixel level accuracy whilst ensuring the model image computation time was no longer for typical values of the srsicindex ( @xmath0 ) .
for these comparisons a modified build of galfitwas used that did not read or write to disk and instead took arguments and produced outputs in memory ( ensuring this relatively large i / o bottleneck was not a factor ) .
the exact pixel fluxes were computed using the r2cuba rpackage that is an interface to the cuba c++ library that offer deterministic algorithms for multidimensional numerical integration . the algorithm used was cuhre , which offers a user definable level of integration accuracy .
for these tests it was set to be better than 0.1% pixel accuracy ( in practice achieving an order of magnitude better accuracy in most pixels ) , so at most this amount of inaccuracy may come from the reference _ exact _ pixel fluxes .
it is worth highlighting why this approach can not be used in general : to create a typical @xmath91 pixel image ( a very small image by galaxy fitting standards ) with this level of accuracy takes @xmath33 seconds ( and much longer for some profiles ) . in comparison ` libprofit ` and galfitcan both create similar image in less than 0.01 seconds , i.e. they are both at least 300 times faster than using a multidimensional integration library , even a fast one written in c++ .
it is not clear why using this specialised library is so slow , but it ruled out using it within ` libprofit`for general application .
figure [ fig : fluxweighterror ] compares the flux - weighted pixel error of ` libprofit ` and galfitas a function of srsicindex for srsicprofiles ( again , srsicprofiles have been high - lighted due to their popularity in the galaxy profiling community ) with @xmath92pixels , @xmath93 degrees and @xmath94 .
two things are immediately clear : ` libprofit ` has a lower flux - weighted error per pixel ( typically 0.1% ) , and this error does not vary significantly with srsicindex . galfitachieves a flux weighted error per pixel of 0.3% for low values of the srsicindex , but this inflates to > 1% for larger values of @xmath0 .
pixels , @xmath93 degrees and @xmath94 in a @xmath95 pixel image .
a subset of these results ( for @xmath96 ) are shown in detail in figure [ fig : pixelaccuracy ] . ]
the image generation is shown in more detail for these example profiles in figure [ fig : pixelaccuracy ] .
this shows the actual error as a function of pixel location on the two - dimensional image for the target model for both ` libprofit ` and galfit .
figure [ fig : pixelaccuracydiag ] shows the one - dimensional pixel error across the central portion of pixels that incorporate the peak pixel ( which is usually the hardest to integrate accurately for the srsicprofile )
. there are specific regions of pixels where galfithas smaller errors , but the general trend is that ` libprofit ` is more accurate over a large range of pixels , and never experiences large integration error in pixels that contain significant fractions of the profile flux . in the case of both
` libprofit ` and galfitouter pixels might often have large relative error , but in these cases the pixels contain almost no flux and do not drive the overall model image to a large flux - weighted error . exact image generated with cubic quadrature routine via cuba + + error in approximate ` libprofit
` image + + error in approximate galfitimage + + error across central diagonal pixels + + + as established in the design goals of ` libprofit ` , generating images accurately is important , but there is a trade - off to be made with the required computation time .
it is unavoidable that some pixels will be harder to evaluate due to the rapid change in flux within a pixel ( especially true for the central pixel containing the peak of the srsicprofile flux ) , and more computing time will need to be spent on correctly evaluating the pixel flux . `
libprofit ` uses adaptive recursion to adjust the integration resources as required in order to maintain a relatively constant level of pixel integration accuracy .
this is evident from figure [ fig : imagetime ] , which compares the image generation times for ` libprofit ` and galfitfor the same srsicprofiles as presented in figure [ fig : fluxweighterror ] .
it is clear that ` libprofit ` varies the integration time ( which is akin to more resources spent evaluating _ hard _ pixels ) more in order to maintain relatively constant flux weighted errors ( see figure [ fig : fluxweighterror ] ) , whereas galfitmaintains nearly constant integration times .
pixels , @xmath93 degrees and @xmath94 in a @xmath97 pixel image . ]
the computation times of ` libprofit ` and galfitfor srsicprofiles are investigated in more detail in figure [ fig : timevtime ] and figure [ fig : paramtime ] .
for all of these tests the same macbook pro running el - capitan with 2.6 ghz i7 processors and 16 gb of ram was used . in figure
[ fig : timevtime ] we see that ` libprofit ` is factors of a few faster compared to galfitfor most combinations of model parameters . in figure
[ fig : paramtime ] we see there is weak dependence of computation time on @xmath6 and @xmath0 , and a strong dependence on axial ratio ( in all three cases larger means longer computation time for ` libprofit ` ) . for reasons not immediately clear galfitactually takes longer to compute compact models - perhaps because its integration scheme is not sufficiently adaptive .
in general we find a wide range of potential computation times , spanning a factor of @xmath310 for ` libprofit ` versus @xmath33 for galfiteven though the target image is @xmath98 pixels in all cases .
this makes it hard to predict in general how long a particular fit might take given the image size . in a pathological situation of the target galaxy having large @xmath6 , large @xmath0 and being circular in projection the computation could be a factor 10 longer than for the case of a compact galaxy with low @xmath0 and fairly elongated projection .
pixel images .
galaxy size ( 10 @xmath6 ) , srsicindex ( 11 @xmath0 ) , rotation angle ( 10 @xmath17 ) and axial ratio ( 10 @xmath99 ) were sampled on a grid , with 11,000 model images generated in total .
the red lines show 1:1 , 1:2 , 1:3 , 1:4 and 1:5 relative speeds for the same target model image , ` libprofit ` being substantially faster in almost all situations . ]
time to create @xmath100 pixel images + + + it is possible to configure ` libprofit ` to exceed @xmath30.1% flux weighted image accuracy .
this default was deemed appropriate for typical ( and certainly our ) uses of ` libprofit ` since it equates to 0.001 mag accuracy , which in practice far exceeds the systematic and random measurement uncertainties for even relatively high fidelity photometric images .
it also allows for rapid image fitting since small gains in accuracy begin to take factors longer to compute .
however , a combination of increasing the oversampling and recursion depth allows for arbitrarily high levels of integration accuracy should this be required by the user . by including a simple interface to the cuba library in the rprofitpackage for all six radial profile types , it is possible to tune these parameters in order to achieve the desired level of accuracy and speed .
profit supports both brute - force convolution and fft - based convolution using the fftw library . in the rimplementation ,
both methods are benchmarked and the optimal method for the given image and psf size is chosen .
brute - force convolution is often significantly faster than fft - based methods but scales directly with image _ and _ psf size , being @xmath101 , whereas fft convolution is @xmath102 , accounting for necessary zero - padding .
testing shows that both methods give nearly identical answers , with fractional differences of less than @xmath103 .
one important consideration is that the accuracy of convolution depends on the resolution of the psf relative to the pixel scale .
a poorly - resolved psf with a fwhm @xmath104 pixels may give flux errors on the order of a few percent for the innermost pixels .
profitallows for model and psf oversampling for more accurate convolution , whereby the model pixel grid is subdivided by an integer factor ( preferably odd ) .
this is independent of the oversampling for the purposes of accurate model convolution , and requires either an analytic psf or prior interpolation of the empirical psf .
interpolating a noisy empirical psf offers a limited gain in accuracy , so for this reason ( and as mentioned in [ subsubsec : psf ] ) , analytic psfs are generally preferred . when this process is attempted
empirically it is usually via stacking or drizzling of the data ( e.g. with the use of a program like multidrizzle ) .
accounting for pixel covariance when drizzling is a complex problem , and best handled by fitting the native resolution images simultaneously rather than attempting to combine them in an information - lossy process .
simultaneous image fitting of multi - exposure data and covariance likelihoods are high on the list of future extensions to profit , but are not present in v1.0 ( although experimental covariance likelihoods are already being tested ) .
having demonstrated that ` libprofit ` integrates and convolves analytic profiles quickly and accurately , the remaining ingredients for model fitting are the likelihood function and optimisation and/or likelihood sampling method . this logic is implemented in the profit rpackage .
ris becoming increasingly widely used in astronomy and other fields of advanced data analysis , and since it features a large number of statistics packages and optimisation methods ( maximum - likelihood and bayesian ) , it is ideally suited for the purpose of bayesian galaxy image fitting . having easy access to a wealth of optimisers was a key design goal of profit .
a large subdivision of statistical data science and computer science has invested huge efforts into describing and implementing a multitude of optimisers that have a range of strengths and weaknesses . depending on the specific problem at hand
, one optimiser might be preferred over another for reasons that are very hard for us to intuit .
since this is such a key component of galaxy modelling , where a typical use case might involve fitting dozens of partly degenerate parameters , giving the user flexibility to experiment with optimisation engines was deemed critical . in practice
an initial global parameter search might be made using a fast but biased downhill gradient optimiser , or one which is known to get easily trapped in local minima . from this point a more expensive but robust markov chain mote carlo ( mcmc ) optimiser
could be used to accurately refine the parameter posteriors ( this is the typical mode of operation recommended by the authors and described in detailed vignettes included with the rpackage version of profit ) . in the rimplementation alone
the user has easy plug - and - play access to hundreds of optimisers accessible though the central r archive network ( cran ) with hundreds more served from online repositories ( e.g. github ) .
we do not wish to be overly prescriptive in how samplers are used to model data , but some typical applications will be discussed in the following sections . a number of inputs are required to meaningfully fit a profile to a target galaxy observation . for good quality data where the whole image is to be used for fitting , the bare minimum is strictly an image where pixels contain integer photo - electron counts and an analytical description of the point spread function . in practice ,
a more complete set of inputs can include the flux image ( @xmath105 , in whatever linear units are appropriate ) , a segmentation map ( that is pixel matched to @xmath105 , where distinct integer values represent different detected structures in the image as per sextractor ) , a binary mask ( that is pixel matched to @xmath105 , where true means mask out and ignore for analysis , and false means do not mask ) and the @xmath106 image ( @xmath106 , that is pixel matched to @xmath105 , where values represent normal errors in the same units as for @xmath105 ) .
the point spread function can be specified in two main ways : either via an analytical description using any of the available radial profiles ( in practice the moffat function is a good profile for this task ) , or with an image of the psf that is on the same pixel scale as @xmath105 .
profithas a high - level setup function ( profitsetupdata ) that can take these basic inputs and create an internal object that is appropriate for fitting a profitmodel .
there are a wealth of routes to compute meaningful image versus model likelihoods , depending on the data regime .
if many ( i.e. hundreds of ) photo - electrons are registered in pixels for the observed galaxy then the data is likely to be operating in the normal statistics regime , where profitlikelihoods can be computed with : @xmath107 here @xmath108 represents the observed pixel data at pixel @xmath109 , @xmath110 represents a computed profitmodel at pixel @xmath111 and @xmath112 represents the estimated normal uncertainty at pixel @xmath113 .
an alternative view is that cumulative sum of squared residuals should follow a @xmath114 distribution , with a profitlikelihood computed by : @xmath115 here @xmath116 represents the degrees of freedom of the @xmath114 distribution . for fitting of this type
this will usually be determined by the number of pixels containing galaxy flux ( here @xmath117 ) minus the number of parameters being used to fit the model to the data ( up to 8 in the case of a single component srsicmodel ) . in the regime of the statistics being truly normal and the model fitting well ( i.e. a reduced @xmath114 near unity ) , the normal and @xmath114 distribution likelihoods converge to similar results .
the chief distinction is that the @xmath114 statistic penalises overfitting , since extremely small residuals across all pixels are highly unlikely .
both statistics are sensitive to residuals significantly above the shot noise ( @xmath118 somewhat more so ) , such as non - axisymmetric features like bars , spiral arms and dust lanes .
nonetheless , if the dominant source of uncertainty in the image is truly shot noise rather than systematics , and the signal - to - noise is high , the @xmath114 statistic is appropriate .
however , if the count rates are relatively low , then the poisson statistics must be used . in this case , profitlikelihoods are computed with : @xmath119 here @xmath120 is the expectation for the number of photo - electrons counted in at pixel @xmath113 , and @xmath108 is the integer number of photo - electrons recorded . dropping the @xmath121 term ( since it depends only on the data ) applies a global offset to log - likelihood , which makes no difference to any posterior inference since this is all computed in a relative manner .
expressed in this form and multiplied by a factor of -2 we recover the `` cash '' statistic @xmath122 @xcite , which is similar to @xmath114 but for poisson statistics . in the same manner that @xmath123 for normal statistics , @xmath124 for poisson statistics .
it is worth highlighting that returning images to true photo - electron counts is non - trivial and in our experience many published data sets do not produce enough ancillary information to return the images to true counts to better than an accuracy of a few factors .
a particularly pernicious issue is that counts must be known within the ccd itself , not for photons at the top of the atmosphere .
the latter is relatively trivial to compute after the fact if we know intrinsic properties of the source spectral energy distribution ; however , the former requires estimates of atmospheric transmission , instrumental losses , detector gain and quantum efficiency , etc .
sometimes this information is easy to obtain at least approximately ( e.g. for the sloan digital sky survey , sdss : * ? ? ?
* ) , but this level of ancillary meta - data is in general rare
. added to this is the fact that optical survey data generally has copious counts per pixel : typically thousands of photo - electric counts .
for these reasons , poisson likelihoods should seldom be used for optical survey data , even if quite shallow .
it is more appropriate for fitting x - ray data , where the counts are low and have the required meaning for poisson statistics .
indeed , x - ray astronomy is the main subdivision that uses poisson ( usually in the guise of cash ) statistics .
if the data is nominally in the normal statistics regime ( i.e. many photo - electron counts per pixel ) , but in practice have a significant number of data points that are poorly represented by our attempted model ( typical when trying to fit smooth two - dimensional srsicprofiles to well - resolved galaxies with asymmetries ) then more robust student - t distribution statistics might be appropriate .
these distributions are approximately normal in the core but have broad lorentzian wings that puts more likelihood mass at large distances from the normal core .
this behaviour is controlled by the ` degrees - of - freedom ' , which in practice is not free and is instead estimated via maximum - likelihood from the data directly .
the profitstudent - t likelihoods is be computed by : @xmath125 profitgives the users easy access to all four types of commonly used likelihood statistics , and it is for the user to decide the most appropriate given the quality and type of data at hand . for high @xmath126 images of galaxies
the student - t statistic behaves well over a broad regime given its robustness to outlier flux values . for this reason it is selected as the default option when setting profitup for fitting . in the future
we will add covariance likelihoods ( more appropriate for for data that exhibits large amounts of pixel correlation , e.g. nir , see * ? ? ? * ) , but these are not present in v1.0 .
once the observational data has been correctly assembled and the most appropriate type of likelihood statistic has been chosen , the user is free to fit a model image .
profitrequires the user to suggest an initial parameterisation of the model , which need not be close to the optimal parameters with robust optimisers .
the user is free to fix some components and fit others ( e.g. they might wish to fix the disc srsicindex to 1 ) .
they are also free to provide prior distributions for each parameter ( important for formal bayesian analysis ) , hard limits and/or constraints on the allowed fitting region , to specify whether parameters are optimised in log or linear space and to specify additional constraints between parameters ( e.g. pair parameters together ) .
the detailed process for specifying these options is provided in the profitmanual and example vignettes , so we will not repeat the full description here .
once the fitting structure has been specified within profit(profitsetupdata , as mentioned above ) the user can interface with a number of popular optimisation routines .
the r profitpackage provides examples of using the base optim function , the open source laplacesdemon package which gives access to gradient optimisers via laplaceapproximation and a large suite of markov chain monte carlo ( mcmc ) samplers via laplacesdemon , and the genetic algorithm cma - es package .
these four routes to optimisation offer over a hundred distinct optimisers .
the central r archive network ( cran ) collates packages that tackle particular problems types : optimisation and bayesian analysis task views suggest over a hundred additional packages which themselves potentially contain multiple algorithms .
to the uninitiated this can seem daunting , but the reality is that a core few packages and optimisers give the user a solid foundation for a diverse range of fitting problems .
sample vignettes provided in the rprofitpackage goes through some example fitting problems in detail , but we will summarise the approach here . the basic strategy for preparing the inputs for galaxy
fitting is : * required : read in an * image * of an observed galaxy ( in principle any format , but e.g. fits ) , * optional : read in a * mask * with the same size , pixel scale and astrometry as the * image * , * optional : read in a * sigma * image with the same size , pixel scale and astrometry as the * image * , * optional : read in a * segmentation * map with the same size , pixel scale and astrometry as the * image * , * optional : read in a * psf * image with the same pixel scale as * image * , * required : read in an * initial * model list ( as detailed above ) , * optional : read in a list with the same structure as * initial * detailing which parameters profitwill * fit * ( the others will be fixed at their * initial * values ) , * optional : read in a list with the same structure as * initial * detailing which parameters profitwill fit in * log * space ( the others will be fitted in linear space ) , * optional : read in a function detailing the calculation of the * prior * log - likelihood , * optional : read in a list with the same structure as * initial * detailing the allowed * limits * for each parameter when fitting , * optional : read in a function detailing the calculation of additional * constraints * to apply to parameters , e.g. pairing parameters together or making them scale relative to each other in a fixed manner , * required : all of the above two - dimensional matrix images , list structures and functions are provided to the high level profitsetupdata function that organises the information into an object of class * profit.data * ready for fitting with third party optimisation engines .
once the user has created the * profit.data * object , which includes some additional options for choosing the type of likelihood and setting the verbosity of the fitting process , a basic strategy for fitting might look like : * initially find a maximum likelihood galaxy model solution using the base r optim function using the broyden , fletcher , goldfarb and shanno ( bfgs : * ? ? ?
* ) algorithm ( which is fairly robust to spurious minima ) , * take the solution from the bfgs optimisation as a starting point and use laplacesdemon component - wise hit - and - run metropolis ( charm ) to make mcmc samples of the likelihood space ( this is very robust to local minima , and can move large distances from poor solutions ; hence hit - and - run ) , * use diagnostic tools provided with the laplacesdemon package to test for the quality and convergence of the model fit , including tests for auto - correlation between parameters , the number of effective stationary samples and traditional parameter triangle plots , * assuming the fit is flagged as being well converged and well behaved the log marginal likelihood ( lml ) can be computed to aid comparisons between models .
the above fitting process can be carried out for models of varying complexity .
if they are well - converged then the lml computed can be used to calculate bayes factors between simpler and more complex models , offering an objective route for deciding how complex the model needs to be .
the profitpackage includes 10 example data sets with all the required inputs for fitting .
these examples use public sloan digital sky survey ( sdss ; * ? ? ?
* ) and the kilo degree survey ( kids ; * ? ? ?
* ) data for galaxies selected from the galaxy and mass assembly ( gama ; * ? ? ?
* ; * ? ? ?
* ) survey that are at reasonably low redshift and are deemed to have multiple components ( as found in * ? ? ?
the vignettes included with the software give examples of fitting this data using high - level rand pythoninterfaces .
bulge component of galaxy fit : + + disc component of galaxy fit : + + bulge+disc components of galaxy fit : + + as a case study , figure [ fig : modelfit ] shows the quality of mcmc fit and residuals that are possible using profitfor g266033 .
a qualitative assessment of the residuals shows that we are able to remove the major bulge and disc structural components , but that non - axisymmetric components remain ( the spiral arms in particular ) .
the three rows of panels show the bulge ( top ) , disc ( middle ) and combined ( bottom ) components .
the outputs of the charm mcmc sampling are shown in figure [ fig : modeltriangle ] . here
only stationary samples are shown ( i.e. there is no ` burn - in ' of the sampling ) , and whilst there is evidence of some weak covariance between some parameters ( e.g. disc magnitude ( dmag ) and effective radius ( dre ) ) , in general the posteriors look well sampled and consistent with multivariate normal distributions .
this latter characteristic is useful when estimating marginalised parameter errors . using the rinterface to profit ,
more quantitative information can be extracted from the fit .
figure [ fig : modelfitchisq ] shows the output of the profitmakeplots function , which as well as producing the standard residuals , also outputs how they compare to a reference normal and student - t distribution .
the @xmath114 excess is also demonstrated with respect to a @xmath114 distribution with one degree - of - freedom .
finally , a residual image scaled by the pixel error @xmath106 is provided .
this shows that there are some significant residual spiral arm structures that are highly unlikely from the point of view of the idealised generative model we are fitting . as discussed in the introduction
, it is common in the field of galaxy modelling to collapse two - dimensional data into a one - dimensional form . whilst profitdoes not evaluate likelihood using this one - dimensional form ,
it does provide a functions for producing such isophotal data and the associated plots ( profitellipse and profitellipseplot ) .
in brief these functions use the geometric parameters of the disk and bulge components of a simple two - component model to extract isophotal annuli , allowing for the standard geometric distortions such as ellipticity and boxiness ( if present ) . the one - dimensional profile of g266033 with overlaid model fits is shown in figure [ fig : modelfitoned ] .
error region for the data pixels , and ideally the total model ( green line ) should sum to follow this distribution quite closely . at a few radii
the data and model are in some tension ( at the 0.1 mag level in surface brightness ) , but for the majority of the profile they agree very well . ] given the residual flux not replicated by the model , the question remains how representative any posterior error distributions might be . to estimate the impact an imperfect model can have on the returned posterior error estimates ,
the residual is subtracted from the initial image .
figure [ fig : modelfitchisqmod ] shows the effect of fitting such an input where the image noise has been added back in via poisson sampling of the convolved model image .
the residuals almost entirely disappear , and the @xmath106 scaled residuals show no radial structure at all . comparing the posteriors to the original input data and the residual - removed input data ,
the former has typical marginalised posterior errors for each parameter that are a factor of @xmath32 _ larger _ ( spanning the range 1.33 ) with best fit values that agree to within 1% .
this is encouraging , since it suggests in this particular case , where profitis correctly identifying the major structural components , the presence of non - axisymmetric components creates posteriors with larger errors .
how such fitting behaves will depend on the data at hand ( i.e. this should not be assumed to be true in general ) , but such a test is easy to implement , and should be part of a standard profitworkflow .
as mentioned above , the r profitpackage comes with 10 fairly isolated example galaxies included that have imaging from sdss and kids .
all of these are well resolved galaxies taken from the gama survey where we already have well converged bulge and disc fits from @xcite . to create this sample 40 galaxies from @xcite
were randomly sampled and these were ranked in terms if fit - ability ( removing galaxies with more complex backgrounds and nearby sources ) .
of these the 10 largest were chosen as example galaxies .
as such this sample , whilst small , is broadly representative of how common different classes of bulge - disc systems are ( e.g. we are dominated by lower @xmath127 systems ) .
consistent sky subtractions were made using lambdar @xcite , and segmentation maps and psfs were created using an updated version of sigma @xcite .
we then proceeded to fit these 10 galaxies using both sdss and kids inputs , and using profitand galfit .
the free parameters used throughout were @xmath128 and @xmath129 centres , bulge mag , @xmath6 and @xmath0 ( @xmath99 for the bulge was fixed to be 1 ) , and disc mag , @xmath6 , @xmath99 and @xmath17 ( @xmath0 for the disc was fixed to be 1 , i.e. exponential ) .
the input parameters were taken from the fits in @xcite , although it should be noted that profit(used in full mcmc mode ) is largely insensitive to the inputs used , as long as the total input magnitude is approximately correct ( i.e. within a couple of magnitudes of the correct value ) . used in pure downhill gradient mode profitsuffers from similar local minima issue to galfit(see the discussion of galfitconvergence in * ? ? ?
the full range of comparison figures are included in appendix [ app : consist ] . here
we compare how well these different combinations of decomposition codes and data sources affect the returned values for bulge and disc magnitudes .
figure [ fig : compmagb ] compares the agreement between estimated bulge magnitudes .
the one major disagreement between codes is the faintest bulge measured . in profitthe decomposition preferred removing the bulge entirely , often hitting our specified lower limit of 30 mag . using galfit
, the code does not allow the solution to move huge distances from the initial estimates unless the likelihood terrain is very smooth , which occurs only when the model reproduces the data with no clear residuals .
figure [ fig : sdssdecomp ] shows the fit and residuals for the profitfit for this galaxy , suggesting there is little need for any bulge at all .
otherwise the main conclusion that can be drawn is that the codes correlate tightly , and so do the data sets , but in all cases the scatter is much larger than the errors suggested by galfitor profit . although the errors returned by profitwere consistently larger for these galaxies , they are still smaller than the points in the panels .
a general remark is that we find more consistency between codes than between data sets .
removing the outlier galaxy g266035 , we measured the intrinsic scatter using the hyper - fitcode of @xcite and find it to be reduced by more than a factor of two .
it is at a minimum for the sdss data ( 0.14 dex versus 0.20 dex scatter orthogonal to the 11 line ) , suggesting that some aspect of our data processing is more internally consistent for sdss data
. this might be the background subtraction or the psf determination .
figure [ fig : compmagd ] compares the agreement between estimated disc magnitudes . in general
we find more self - consistent fits for discs both between codes and between data . whilst the smallest intrinsic scatter is again found for sdss using profitand galfit , the results for profitusing sdss and kids are not far behind
. the intrinsic scatter for kids using profitand galfit , and galfitusing sdss and kids are notably worse ( the latter showing the most scatter ) .
for all these fits we find that the profitmcmc solution provides a _ more likely _
fit both when forcing the profitparameters through galfit , and when forcing the galfitparameters through profit . usually the increases in likelihood when using profitare substantial , suggesting the difference is due to local convergence issues with the levenberg - marquardt @xcite algorithm built into galfit , rather than issues to do with model accuracy that we explored in section [ sec : profit - acc ] .
a general conclusion is that we appear to be able to recover more self - consistent results for disc properties compared to bulge properties .
further evidence for this is shown in appendix [ app : consist ] , where disc @xmath6 is found to be more reliably measured than bulge @xmath6 .
bulge @xmath0 is notably the hardest property to measure consistently , a finding in agreement with recent work by @xcite .
we investigate in more detail how well we expect to be able to recover various bulge - disc properties for different quality and depth data in the next section .
having described and tested our new profitdecomposition code in some detail , we finally investigate how it performs on data of different depth and quality . for the purposes of these tests we create and fit bulge - disc systems in @xmath130 arc - second image stamps .
we investigate three sources of survey imaging that are indicative of the types of data that profitwill be used on en - masse in the future : sdss , kids and lsst . since these results are purely indicative we make some simplifying assumptions for the three surveys .
table [ tab : surveys ] presents the assumptions made when simulating and fitting the galaxies .
these are not precisely the values advertised for the various surveys , but are indicative given the inevitable variations in data quality .
for lsst these estimates are for the stacked five - year survey .
since we know the precise psf used to convolve the model and the images themselves are precisely background subtracted , the following results are the upper - limits on the decomposition performance we can hope to expect . the simulations themselves sample 10,000 profiles from a uniform grid of total magnitude ( @xmath131 ) , bulge fraction ( @xmath132 ) , logarithmic bulge effective radius ( @xmath133 ) , logarithmic disc effective radius ( @xmath133 ) , logarithmic bulge srsicindex ( @xmath134 ) and disc axial ratio ( @xmath135 ) .
the bulge axial ratio was assumed to have @xmath136 and the disc srsicindex was fixed to @xmath8 .
profitnever appears to struggle to find the correct disc angle @xmath17 , so we only simulated galaxies with the minor axis varying on the @xmath128-axis of the image .
note we do not impose internal correlations between parameters , so these results span an overly generous range of structural parameter space .
.assumed image quality parameters for different surveys .
[ cols="<,<,<,<",options="header " , ] [ tab : surveys ] rather than magnitude , i.e. @xmath137 ) .
the solid lines show the median of the results , and the dotted show the @xmath138 and @xmath139 percentile ranges ( akin to the @xmath140 range if the distributions are normal ) . in simple terms , larger value are better in the top panel and lower values are better in the bottom panel . ]
figure [ fig : converge_frac ] shows the results of these simulations in a compact manner , comparing the fraction of fits that are found to converge ( which means they find solutions that are not hard limits imposed within profit ) and the typical average separation between all input and output parameters in log - space .
these properties are plotted for all major observables in appendix [ app : sims ] , but here we show the results versus the most observable input parameter : the total magnitude of the galaxy . unsurprisingly , all the results are best when the galaxy is intrinsically bright and the source signal dominates over the sky background . of perhaps more surprise
is that we see a large improvement in decomposition performance when moving from sdss to kids quality data , but a relatively minor improvement when we move from kids to lsst .
this appears to be due to the major improvement being limited to surface brightness depth , the pixel scale and the typical psf fwhm is not expected to be much ( if at all ) improved between kids and lsst ( one caveat being the latter has a much larger field - of - view , so further un - simulated sky - subtraction gains could be expected ) . to briefly summarise the other trends presented in the appendix : lower bulge and disc @xmath6 , lower disc @xmath99 , lower bulge @xmath0 and a moderate @xmath141 all lead to improved fitting results . for @xmath6
this is simply because lower @xmath6 means high surface brightness for a given magnitude , and the components become easier to distinguish against the background .
we start our simulations at an @xmath6 of 1 arc second , so the caveat to make is that much below this ( @xmath6 less than half the psf fwhm ) , and the parameters become indistinguishable structurally . even in this regime
we can still expect the component magnitude to be well recovered , it is just hard to resolve structure inside the psf scale .
it is clear from these results that only in the most favourable regimes can we expect complete _ and _ robust bulge - disc decompositions , even using the best current and next generation deep survey data .
this is true even when we are operating in the regime of moderately low redshift galaxies extracted from surveys such as gama .
a future low - redshift focussed waves - wide survey @xcite , being inherently lower redshift due to its photo-@xmath142 pre - selection ( with typical redshift likely to be similar to sdss dr10 ) is the best prospect for such future studies given the target of high redshift completeness , and a guarantee it will have deep good quality imaging data ( since kids will act as the primary input source for targets ) .
in this paper we have presented our new publicly available galaxy modelling and decomposition code profit .
the core code comes in a standalone c++ library ( ` libprofit ` ) that allows for easy access to higher level languages ( e.g. rand python ) through a simple api .
the core advances that profitoffers over currently available software are that it : * is fully open - source with multiple active developers , * offers a standalone library ( ` libprofit ` ) for accurate and fast pixel integrations when generating a model , * can be extended with new profiles in a simple well - documented manner , * allows for simple or complex priors on parameters ( an important aspect of bayesian analysis ) , * offers a range of likelihood calculations , * is untied to any specific optimisers but has easy access to downhill minimisation , genetic algorithm and mcmc routines , * can fit parameters in log or linear space , * allows for simple or complex additional constraints between parameters , * offers brute - force and fft psf convolution options , with automatic benchmarking to select the fastest strategy .
initially profitcomes in three varieties : an fully featured rpackage ( profit , discussed in detail in this paper ) , a basic python wrapper ( pyprofit ) and a command line terminal interface ( profit - cli ) .
the rpackage is the most advanced in terms of features , and we have a longer term aim to bring the pythonpackage up to the same level of sophistication .
the command line interface only exists for the easy generation of model images via calling the underlying ` libprofit`library , so it does not provide any built - in optimisation . by separating the model image generation ( which should be an objective black - box task achieved through an api ) into a separate ` libprofit`library
, users are not tied to our high - level solution for the much more subjective problem of galaxy fitting ( with many caveats over masking , convolution , likelihoods etc ) .
we are aware of at least one community fork of ` libprofit`already that makes use of diffusive nested sampling ( dnest ; * ? ? ?
* ) in order to overcome sampling of highly multi - modal data ( huijser et al .
, in prep . ) .
extensions to other languages using the library api are encouraged , but simpler workarounds using the command line interface are also possible for less programming - savvy users . to avoid community confusion we request that any forks that are released append the name in some manner ( e.g. profit_stan ) .
the aim would be to pull any substantial and useful changes into the main profitbranch in the longer - term .
the first versions of the profitcode stemmed from preliminary work by asgr . following this ,
dst and rt have added many new features and have heavily modified and expanded the functionality of the software . from the early stages
profithas been designed in a highly modular manner .
this is important to prevent the user becoming limited to our default choice of profile , likelihood evaluation or optimisation method .
the core ` libprofit`library is designed to be fast and accurate at integrating and convolving target model images , where the likelihood function to compute , and optimisation engine to use , are largely choices for the user via higher level interfaces .
the rpackage version of profitcontains a large number of examples using simple downhill gradient schemes , more complex genetic algorithms and more computationally expensive mcmc techniques .
users are encourages to build from these examples to use more or less sophisticated engines as appropriate .
the various libraries and higher level interfaces are all available on github , and future support and functionality will be added through these repositories .
this paper necessarily refers to a static v1.0 of profit , as such specifics regarding operation and options should be derived from the public repositories and documentation found there , rather than this text .
the most reliable longterm location for the base ` libprofit`library is https://github.com/icrar/libprofit .
the higher - level rimplementation of profitis available at https://github.com/icrar/profit and the python variant is maintained at https://github.com/icrar/pyprofit .
this paper was written using the rimplementation exclusively .
community use and feedback is encouraged , especially via raising _ issues _ and _ pull _ requests through github .
in particular , the requirements outlined in section [ section_introduction ] are not an exhaustive list , so users are encouraged to submit feature proposals and requests .
currently planned extensions include built - in support for monte carlo image generation , simultaneous fitting of multiple images ( in the same band ) and/or multiple bands , zero - point calibration , covariance likelihoods and psf fitting .
indeed , the rpackage already includes a vignette with psf fitting examples , a full description of which is omitted here for brevity .
in this paper we have investigated the application of profitto both sdss and kids data , using galfitreference runs as a comparison . in pragmatic terms , profitand galfitachieve consistent results
given the same input data , with at least as much scatter being produced by changes between data source ( in this case sdss versus kids ) .
this is encouraging in that it validates the broad findings of recent work that made exclusive use of galfit , and lends credibility to our future applications of profitto a number of different data sources .
the main difference is a large degree of intervention was required when running galfit via running on a large grid of initial conditions and post - processing the results ( see * ? ? ?
* ) , whereas profitwas run by am with just a basic wrapper using one of the many available mcmc routines in rand converged to reasonable global solutions with little user intervention .
much of the work presented here was made possible by the free and open rsoftware environment @xcite .
all figures in this paper were made using the r magicaxis package @xcite .
data from the kids @xcite and sdss @xcite surveys were used for this work .
funding for sdss - iii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , and the u.s .
department of energy office of science .
the sdss - iii web site is http://www.sdss3.org/. parts of this research were conducted by the australian research council centre of excellence for all - sky astrophysics ( caastro ) , through project number ce110001020 .
credit to e. mannering for figure [ fig : pixint ] , she was inspired by the lamentable initial efforts of asgr . thank you to the anonymous referee , whose comments particularly assisted in clarifying complex parts of the paper . 1
to supplement figures [ fig : compmagb ] and [ fig : compmagd ] , below are the various other profit , galfit , sdss and kids comparisons for the 10 galaxies that are included in the standard rprofitinstallation for demonstration and testing purposes .
for the 10 well resolved galaxies included with the rprofitpackage .
see figure [ fig : compmagb ] for details . ] for the 10 well resolved galaxies included with the rprofitpackage .
see figure [ fig : compmagb ] for details . ] for the 10 well resolved galaxies included with the rprofitpackage .
see figure [ fig : compmagb ] for details . ] for the 10 well resolved galaxies included with the rprofitpackage .
see figure [ fig : compmagb ] for details . ] for the 10 well resolved galaxies included with the rprofitpackage .
see figure [ fig : compmagb ] for details . ] for the 10 well resolved galaxies included with the rprofitpackage .
see figure [ fig : compmagb ] for details . ]
but for bulge magnitude on the x - axis . ] | we present profit , a new code for bayesian two - dimensional photometric galaxy profile modelling .
profitconsists of a low - level c++ library ( ` libprofit ` ) , accessible via a command - line interface and documented api , along with high - level r(profit ) and python(pyprofit ) interfaces ( available at github.com/icrar/libprofit , github.com/icrar/profit , and github.com/icrar/pyprofit respectively ) .
r profitis also available pre - built from cran , however this version will be slightly behind the latest github version . `
libprofit`offers fast and accurate two - dimensional integration for a useful number of profiles , including srsic , core - srsic , broken - exponential , ferrer , moffat , empirical king , point - source and sky , with a simple mechanism for adding new profiles .
we show detailed comparisons between ` libprofit ` and galfit . `
libprofit`is both faster and more accurate than galfitat integrating the ubiquitous srsicprofile for the most common values of the srsicindex @xmath0 ( @xmath1 ) .
the high - level fitting code profitis tested on a sample of galaxies with both sdss and deeper kids imaging .
we find good agreement in the fit parameters , with larger scatter in best - fit parameters from fitting images from different sources ( sdss vs kids ) than from using different codes ( profit vs galfit ) .
a large suite of monte carlo - simulated images are used to assess prospects for automated bulge - disc decomposition with profit on sdss , kids and future lsst imaging .
we find that the biggest increases in fit quality come from moving from sdss- to kids - quality data , with less significant gains moving from kids to lsst .
[ firstpage ] galaxies : statistics galaxies : structure methods : statistical techniques : photometric galaxies : fundamental parameters methods : data analysis |
url # 1`#1`urlprefix[2]#2 [ 2][]#2 and . . _ _ * * , ( ) . , and .
_ _ * * , ( ) . .
_ _ * * , ( ) . and .
_ _ * * , ( ) . and .
_ _ * * , ( ) . , , , and .
_ _ * * , ( ) . .
_ _ * * , ( ) . , and .
_ _ * * , ( ) . and .
_ _ * * , ( ) . , , , , , , , , , and .
_ _ * * , ( ) . , , , , , , and .
_ _ * * , ( ) . , , and .
_ _ * * , ( ) . , , , and .
_ _ * * , ( ) . .
_ _ * * , ( ) . , , , and .
_ _ * * , ( ) . , , , , , , and .
_ _ * * , ( ) . , , , , , , , , , , , and .
_ _ * * , ( ) . , , , , , , , , and .
_ _ * * , ( ) . , , , , , , , , , , , , and .
_ _ * * , ( ) . , , , , , and .
_ _ * * , ( ) . and .
( , ) . and .
( , ) . and .
_ _ * * , ( ) . , , and .
_ _ * * , ( ) . , , and .
_ _ * * , ( ) . , , and .
_ _ * * , ( ) . .
_ _ * * , ( ) . and .
_ _ * * , ( ) . , and .
_ _ * * , ( ) . , and .
_ _ * * , ( ) . , , , and .
_ _ * * , ( ) . .
_ _ * * , ( ) . , , and .
_ _ * * , ( ) . , , , , , , and .
_ _ * * , ( ) . , , , , , , and .
_ _ * * , ( ) .
we acknowledge the financial support from the einstein stiftung berlin and we are thankful to the dfg for support through spp 1459 .
furthermore , we thank m. helm , s. winnerl , m. mittendorff ( helmholtz - zentrum dresden - rossendorf ) , t. winzer , g. berghaeuser , and eike verdenhalven ( tu berlin ) for fruitful discussions .
thanks sfb 787 and the bmbf ( nanopin ) for support . | carrier multiplication is a many - particle process giving rise to the generation of multiple electron - hole pairs .
this process holds the potential to increase the power conversion efficiency of photovoltaic devices . in graphene ,
carrier multiplication has been theoretically predicted and recently experimentally observed . however , due to the absence of a bandgap and competing phonon - induced electron - hole recombination , the extraction of charge carriers remains a substantial challenge .
here we present a new strategy to benefit from the gained charge carriers by introducing a landau quantization that offers a tunable bandgap . based on microscopic calculations within the framework of the density matrix formalism
, we report a significant carrier multiplication in graphene under landau quantization .
our calculations reveal a high tunability of the effect via externally accessible pump fluence , temperature , and the strength of the magnetic field .
* introduction . * in 1961 schockley and queisser predicted a fundamental limit of approximately @xmath0 for the power conversion efficiency of single junction solar cells @xcite .
their calculation is based on a simple model , in which the excess energy of absorbed photons is assumed to dissipate as heat .
this is a good assumption for most conventional semiconductors , whereas for low - dimensional nanostructures with strong coulomb interaction , the excess energy can be exploited to generate additional electron - hole pairs . in such structures with highly efficient auger processes , the schockley - queisser limit can be exceeded up to @xmath1 @xcite .
furthermore , the gained multiple electron - hole pairs can also increase the sensitivity of photodetectors and other optoelectronic devices .
carrier multiplication ( cm ) was first theoretically predicted @xcite and measured @xcite for semiconductor quantum dots exhibiting a strong spatial confinement in all three directions .
multiplication of photo - excited carriers has also been found in one - dimensional nanostructures , such as carbon nanotubes @xcite . in graphene ,
the carrier confinement combined with the linear bandstructure was predicted to account for highly efficient auger processes @xcite giving rise to a considerable carrier multiplication @xcite .
just recently , these predictions were confirmed by brida et al . within a joint experiment - theory study also including new insights into collinear scattering in graphene @xcite , which has been controversially discussed in literature .
however , to realize graphene - based photovoltaic devices , the key problem of charge carrier extraction in gapless nanostructures needs to be solved .
here we suggest a strategy based on landau quantization of graphene to address this substantial challenge , where a magnetic field is used to induce gaps between discrete landau levels .
graphene in high magnetic fields has already attracted an immense interest in recent years @xcite . unlike an ordinary two - dimensional electron gas , it exhibits an unique landau level structure , where the energy is proportional to the square root of the landau level index @xmath2 ( @xmath3 ) , cf .
fig .
[ fig : sketch ] .
this is a direct consequence of the linear density of states in the low - energy region and results in an anomalous quantum hall effect @xcite . and @xmath4 , cf . the yellow arrows .
the energy - conserving process of impact excitation ( ie ) involves the inter - landau level transition @xmath5 , which provides the energy to excite an electron from @xmath6 to @xmath7 resulting in an increase of the number of charge carriers , cf . the red arrows .
the inverse process of auger recombination ( ar ) is shown by the green arrows .
, width=321 ] in this article , we predict the occurrence of carrier multiplication in landau - quantized graphene , where the energy gap between the landau levels is of advantage for the extraction of gained charge carriers .
the magnetic field can be tuned so that the inter - landau level spacings do not fit the energy of the most relevant optical phonons resulting in a strongly suppressed carrier - phonon scattering .
although at first sight the non - equidistant landau level spectrum seems to suppress auger processes , a number of energetically degenerate inter - landau level transitions exists providing the basis for efficient auger scattering , cf .
fig [ fig : sketch ] .
this allows to exploit the advantage of graphene having a strong coulomb interaction while the acoustic phonon coupling is weak in comparison to an ordinary two - dimensional electron gas in a semiconductor heterostructure . as a result ,
graphene under landau quantization presents optimal conditions for the appearance of a significant carrier multiplication .
* theoretical approach .
* while the carrier dynamics in graphene has been intensively investigated over the last few years @xcite , there has been only a single study in the presence of a strong magnetic field by plochocka et al .
@xcite . in the latter study
, the relaxation dynamics between higher landau levels ( @xmath8 ) is investigated in a differential transmission experiment revealing the importance of auger scattering in a magnetic field and an overall suppression of scattering rates in comparison to the no - field case .
based on the density matrix formalism @xcite , we perform time - resolved microscopic calculations of the ultrafast carrier relaxation dynamics within the energetically lowest landau levels .
we exploit a correlation expansion in second order born - markov approximation @xcite to derive bloch equations for graphene under landau quantization : @xmath9+s_{n}^{\text{in}}(t)\left[1-\rho_{n}(t)\right]-s_{n}^{\text{out}}(t)\rho_{n}(t),\label{eq : bloch-1}\\ \dot{p}_{nn'}(t ) & = & i\triangle\omega_{nn'}p_{nn'}(t)+\omega_{nn'}(t)\,\left[\rho_{n}(t)-\rho_{n'}(t)\right]-\frac{\gamma(t)}{\hbar}p_{nn'}(t).\label{eq : bloch-2}\end{aligned}\ ] ] it is a coupled system of differential equations for the carrier occupation probability @xmath10 in the landau level @xmath11 and the microscopic polarization @xmath12 being a measure for optical inter - landau level transitions according to optical selection rules , cf . fig .
[ fig : sketch ] .
the equations explicitly include the carrier - light interaction , all energy conserving electron - electron scattering processes , as well as the coupling with most relevant optical phonons .
the carrier - light interaction enters the equations via the rabi frequency @xmath13 with the elementary charge @xmath14 , the electron mass @xmath15 , the optical matrix element @xmath16 , and the vector potential @xmath17 .
the energy difference @xmath18 between the involved landau levels defines the resonance condition .
the magnetic field is incorporated into the equations by exploiting the peierls substitution ( see supplementary note 1 ) , accounting for the change of electron momentum induced by the confinement into cyclotron orbits @xcite .
the appearing in- and out - scattering rates @xmath19 describe coulomb- and phonon - induced many - particle scattering processes .
the rates include pauli - blocking terms and the corresponding matrix elements , cf .
supplementary notes 2 - 4 for more details . due to the energy mismatch of the optical phonon modes with the inter - landau level transitions at the assumed magnetic field strength , optical phonon scattering is strongly suppressed .
processes involving acoustic phonons and two - phonon relaxation processes can occur @xcite , however , both are much slower compared to the coulomb - induced dynamics and are neglected in the following .
the scattering processes do not only change the carrier occupation @xmath10 , they also contribute to the dephasing @xmath20 of the microscopic polarization @xmath12 .
the main contribution of @xmath20 stems from electron - impurity scattering , which also induces a broadening of the landau levels .
this broadening is explicitly calculated in the self - consistent born approximation following the approach of ando @xcite , cf .
supplementary note 5 , where the electron - impurity scattering strength is chosen in agreement with previous theoretical studies @xcite .
the validity of the applied markov approximation has been tested by performing calculations taking also into account non - markovian effects .
the latter reflect the quantum - mechanical energy - time uncertainty principle giving rise to a weak temporal oscillation of carrier occupations .
however , the effect is small and does not qualitatively change the investigated inter - landau level carrier dynamics , cf .
supplementary note 6 . in order to account for momentum - dependent dynamical screening ,
we consider the dielectric function @xmath21 in the random phase approximation following the approach of refs . @xcite . while the real part of the dielectric function closely resembles the static limit ( cf .
ref .
@xcite ) , taking a finite energy transfer @xmath22 into account gives rise to a non - vanishing imaginary part , cf .
supplementary note 3 for more details .
then , the fourier transform of the coulomb potential @xmath23 appearing in the coulomb matrix element @xmath24 is substituted by the screened potential @xmath25 . here , @xmath26 and @xmath22 are the momentum and energy transfers of the scattering events and @xmath27 denotes the wave function of a state @xmath28 . since in the presence of a magnetic field ,
the momentum is not a well - defined quantity , the vanishing screening in the long - wavelength limit does not imply a divergence of the coulomb interaction .
furthermore , the singularities for collinear scattering processes appearing in graphene in the absence of a magnetic field @xcite do not occur in landau - quantized graphene due to the absence of a strictly defined momentum conservation in such a discrete system .
moreover , note that excitonic effects of the coulomb interaction can be neglected , since the landau level broadening is generally larger than the coulomb - induced finite excitonic dispersion , cf .
supplementary note 3 .
involved in the optical excitation characterized by an ultrashort pulse with a width of @xmath29{ps}$ ] and @xmath30 reflecting the importance of auger processes in the presence of a magnetic field with @xmath31{t}$ ] .
the dashed lines illustrate the dynamics in the absence of coulomb scattering .
the yellow shaded region denotes the width of the excitation pulse .
the occupation @xmath32 is not involved in the relaxation dynamics , @xmath33 remains constant , and @xmath34 shows an increase during pumping just like @xmath35 .
the study is performed for undoped graphene at room temperature .
, width=321 ] * carrier relaxation dynamics . * using tight - binding wave functions , we analytically determine the electronic bandstructure and the matrix elements appearing in the bloch equations .
then , we have all ingredients at hand to evaluate the equations and obtain a microscopic access to the time - resolved ultrafast relaxation dynamics of optically excited carriers .
the initial occupation probabilities @xmath36 before the optical excitation are given by the fermi function at room temperature .
we apply an external magnetic field of @xmath31{t}$ ] and an optical excitation pulse with a width of @xmath29{ps}$ ] , an applied pump fluence of @xmath37{\mu jcm^{-2}}$ ] , and an energy of @xmath38{mev}$ ] matching the inter - landau level transitions @xmath4 and @xmath39 , cf . fig .
[ fig : sketch ] . as illustrated in fig .
[ fig : occupations_0.1pump ] , the carrier occupation probability @xmath40 increases up to @xmath41 during the optical excitation ( blue shaded area ) , while @xmath42 decreases for the same amount ( not shown ) .
interestingly , the build up of occupation in the landau level @xmath43 is stopped during the optical excitation and @xmath44 starts to slightly decrease , cf .
the orange solid line in fig .
[ fig : occupations_0.1pump ] .
this effect can be unambiguously traced back to the coulomb - induced relaxation process , cf . the orange dashed line in fig .
[ fig : occupations_0.1pump ] illustrating the temporal evolution of @xmath44 without coulomb scattering . at the same time , the occupation in landau level @xmath7 that is not driven at all by the optical pulse , increases until saturation is reached .
the occupation @xmath45 remains constant , if the coulomb interaction is switched off , cf .
the blue dashed line in fig .
[ fig : occupations_0.1pump ] .
this behavior can be understood as following : first , the exciting laser pulse generates a non - equilibrium carrier distribution by transferring electrons from @xmath46 to @xmath43 ( yellow arrows in fig .
[ fig : sketch ] ) .
then , auger - type processes including impact excitation ( ie ) and auger recombination ( ar ) redistribute the carriers between the equidistant landau levels @xmath47 and @xmath48 ( red and green arrows in fig .
[ fig : sketch ] ) .
our calculations clearly reveal that ie is the predominant channel at the beginning of the relaxation dynamics due to the pauli blocking terms in the scattering rates @xmath19 in the bloch equations . during the optical excitation ,
the occupation of @xmath43 is enhanced , whereas @xmath7 contains only few electrons corresponding to the thermal distribution at room temperature . as a result
, the electron - electron interaction favors the transition @xmath5 over the inverse process and provides the energy to excite an additional electron into @xmath7 from the energetically equidistant zeroth landau level .
this explains the observed increase of @xmath45 and the decrease of @xmath40 , cf .
fig .
[ fig : occupations_0.1pump ] .
interestingly , the occupation @xmath49 does not change and remains @xmath50 during the entire relaxation dynamics ( not shown ) .
this is related to the peculiarity of the zeroth landau level being half valence half conduction band @xcite . due to the symmetry between electrons and holes in the considered undoped graphene ,
each transition evolving electrons can also occur with holes with exactly the same probability . as a result ,
the electron transition @xmath51 is always accompanied by a hole transition @xmath52 corresponding to the excitation of an electron into the zero landau level .
this results in an unconventional ie , where effectively only one charge carrier is created in every scattering event .
the result is a constant @xmath49 , while @xmath40 decreases and @xmath45 increases , just as our theory predicts .
including the thermal equilibrium distribution @xmath53 , and the optically excited carrier density @xmath54 .
the auger - induced carrier dynamics results in a carrier multiplication of @xmath55 at the considered pump fluence of @xmath56{\mu jcm^{-2}}$].,width=321 ] * carrier multiplication . * to examine the total impact of auger scattering processes
, we investigate the temporal evolution of the carrier density that is composed of electrons in the conduction and holes in the valence band landau levels .
the electron - hole symmetry allows a convenient definition of the carrier density @xmath57 with the factor 2 accounting for the electron - hole symmetry , the total degeneracy of each landau level @xmath58 and the area @xmath59 . under the investigated conditions , the zeroth landau level , being equally divided between both bands ,
has a temporally constant occupation and does not contribute .
due to the finite temperature , the initial value of the carrier density is non - zero and we start with an equilibrium value @xmath53 ( yellow line in fig . [
fig : carrier density_0.1pump ] ) .
the optical pulse excites carriers and therewith leads to an ascent of the carrier density on a picosecond time scale .
if no many- particle interactions are taken into account , @xmath54 reaches a value of approximately @xmath60{cm^{-2}}$ ] , cf . the orange line in fig .
[ fig : carrier density_0.1pump ]
. the full dynamics including coulomb - induced carrier scattering reveals a further increase of the carrier density up to @xmath61{cm^{-2}}$ ] reflecting the appearance of a significant carrier multiplication .
the cm factor is defined by @xmath62 and corresponds to @xmath63 at the given characteristics of the excitation pulse .
it is a direct consequence of the impact excitation , where more carriers are generated in @xmath7 than are lost in @xmath43 .
this is apparent when comparing the drop - off in @xmath40 ( @xmath64 ) with the increase of @xmath45 ( @xmath65 ) after two picoseconds in fig .
[ fig : occupations_0.1pump ] .
for the impact excitation ( ie ) and auger recombination ( ar ) .
the yellow area represents the excitation region .
the blue shaded area illustrates the asymmetry region in which ie prevails over ar resulting in the occurrence of carrier multiplication ( cm).,width=321 ] to further prove the occurrence of cm as direct implication of an efficient ie , we determine the time - resolved scattering rates @xmath66 of the two competing auger channels .
our calculations show a much more efficient ie as a consequence of the non - equilibrium distribution created by the optical excitation , cf .
fig .
[ fig : auger rates_0.1pump ] .
the ongoing impact excitation during and after the excitation leads to a more efficient auger recombination until both rates are equal after roughly @xmath67{ps}$ ] .
this is the time frame , in which cm takes place , cf .
fig .
[ fig : carrier density_0.1pump ] .
its value corresponds to the area between both curves , which is ultimately defined by the asymmetry between the two competing processes .
their efficiency as coulomb channels strongly depends on the pump fluence determining the number of available scattering partners .
furthermore , the electronic temperature and the strength of the magnetic field have an influence on the predicted asymmetry between ie and ar
. * discussion . * addressing the optimization of cm , we investigate its dependence on the externally controllable parameters , such as pump fluence , temperature , and magnetic field , cf . fig .
[ fig : cm(fluence , b , t ) ] .
although the initial asymmetry between impact excitation and auger recombination is larger at enhanced pump fluences , it also leads to a faster equilibration of both rates leaving a shorter time frame for cm to occur @xcite . as a result ,
the carrier multiplication decreases with the pump fluence until a saturation of pumping at higher fluences is reached , as illustrated in fig .
[ fig : cm(fluence , b , t)]a .
then , cm starts to slightly increase again reflecting the inefficient optical pumping at saturation conditions .
the dependence on temperature is governed by occupation effects . a higher temperature results in less efficient impact excitation , since an increased initial occupation of @xmath7 lowers the efficiency of the electronic transition @xmath51 .
the qualitative temperature trend resembles a fermi distribution ( cf .
fig .
[ fig : cm(fluence , b , t)]b ) resulting in the largest cm for temperatures lower than @xmath68 . finally , the dependence on the magnetic field is more complex as it is a combination of both effects discussed above , namely the excitation strength and occupation effects .
increasing the magnetic field , the landau levels ( except for @xmath6 ) shift to higher energies according to the relation @xmath69 .
therefore , the initial thermal occupation probability @xmath70 for @xmath71 is reduced with increasing magnetic field and ie becomes more efficient .
moreover , the effective excitation strength is determined by the amplitude of the laser pulse and is inverse proportional to its frequency .
-[fig : auger rates_0.1pump].,width=321 ] as the latter increases with the magnetic field , the pumping becomes less efficient giving rise to a more pronounced cm , as already shown in fig .
[ fig : cm(fluence , b , t)]a .
we observe a saturation value for the carrier multiplication of @xmath72 , cf .
fig .
[ fig : cm(fluence , b , t)]c .
this value is characteristic for the investigated undoped graphene within the energetically lowest landau levels . here
, ie can only occur once via scattering involving the energetically equidistant landau levels @xmath47 and @xmath48 .
furthermore , the peculiar nature of the zeroth landau level being a conduction and valence band at the same time allows only the generation of one additional charge carrier per excitation , as discussed above .
therefore , an optically excited electron - hole pair can yield in total up to three charge carriers resulting in a maximal cm of @xmath72 .
note that at higher magnetic fields , certain inter - landau level transitions can become resonant with optical phonon modes creating so called magneto - phonon resonances .
this is expected to reduce the observed cm , since the phonons open up a new relaxation channel directly competing with auger processes , and even the electron - phonon interaction is modified @xcite .
therefore , the high - magnetic field region of fig .
[ fig : cm(fluence , b , t)]c should be understood as a trend of the cm as a function of the magnetic field . at very low magnetic fields , the spacing between landau levels decreases , resulting in an overlapping of levels . here , the applied theoretical approach is not valid anymore .
furthermore , note that the carrier multiplication does not significantly depend on the landau level broadening , as illustrated in fig .
[ fig : cm(fluence , b , t)]d . on the one hand ,
the lorentzian appearing in the scattering rates decreases for larger broadenings . on the other hand ,
the dielectric function is reduced leading to an enhancement of the coulomb interaction . additionally , due to an increased dephasing of @xmath73 ,
the optical pumping is less efficient .
overall these effects nearly cancel out resulting in a cm that is to a large extent independent of landau level broadening . including energetically higher landau levels
will increase the cm , since the excess energy of a highly excited electron - hole pair can provide enough energy for more than one ie scattering event . in this very first study ,
we have provided clear evidence of cm in the low energy limit .
taking into account the findings of plochocka et al .
@xcite that experimentally prove the importance of auger scattering even for higher landau levels , we expect a pronounced cm to occur in a broad spectral range . although they suspect that auger scattering may be suppressed for lower landau levels , their findings are indeed compatible with our results .
a multi - color pump - probe experiment addressing the carrier dynamics between the discrete low - energy landau levels should confirm our theoretical predictions in a straightforward way . in summary
, we have gained new fundamental insights into the relaxation dynamics in graphene under landau quantization , in particular revealing the appearance of a considerable carrier multiplication that can be traced back to an efficient impact excitation .
we have shown that landau quantization offers a promising strategy to address the key challenge of carrier extraction in a gapless structure which might be beneficial for the design of graphene - based optoelectronic devices . |
since the advent of radar systems much of the efforts have been devoted to increasing radar range resolution .
the relationship between range resolution and signal bandwidth is given by @xmath0 where @xmath1 denotes range resolution , @xmath2 is the speed of light and @xmath3 is the bandwidth of the signal being used .
hence , wideband radar systems can achieve higher resolution than their narrow - band counterparts . however , wideband signals correspond to short pulses that experience low signal - to - noise ratio ( snr ) at the receiver .
further , they require high speed a / ds , and fast processors @xcite .
step - frequency radar ( sfr ) achieves high range resolution without sharing the disadvantages of wideband systems .
sfr transmits several narrowband pulses at different frequencies .
the frequency remains constant during each pulse but increases in steps of @xmath4 between consecutive pulses .
thus , while its instantaneous bandwidth is narrow , the sfr system has a large effective bandwidth .
conventional step - frequency radars obtain one sample from each received pulse and then apply an inverse discrete fourier transform ( idft ) on the phase detector output sequence for detection . since the idft resolution increases with the number of transmitted pulses
, sfr requires a large number of transmit pulses , or equivalently , large effective bandwidth . in this paper
, we propose a step - frequency radar with compressed sampling , that assuming the existence of a small number of targets exploits the sparseness of targets in the range - speed space .
the application of compressive sampling to narrow - band radar systems was recently investigated in @xcite-@xcite . a cs - based data acquisition and imaging method
was proposed in @xcite for stepped - frequency continuous - wave ground penetrating radars ( sfcw - gprs ) . in @xcite , cs - based step frequency
was applied on through - the - wall radar imaging ( twri ) . in @xcite,@xcite
the authors have assumed stationary targets and have shown that the the cs approach can provide a high - quality radar image using much fewer data samples than conventional methods . unlike @xcite,@xcite , the work in this paper explores joint target range and speed estimation based on compressive sampling , and proposes a radar system with reduced effective bandwidth as compared to traditional sfr systems .
compressive sampling ( cs ) @xcite-@xcite has received considerable attention recently , and has been applied successfully in diverse fields , e.g. , image processing @xcite and wireless communications @xcite@xcite .
the theory of cs states that a @xmath5-sparse signal @xmath6 of length @xmath7 can be recovered exactly with high probability from @xmath8 measurements via @xmath9-optimization .
let @xmath10 denote the basis matrix that spans this sparse space , and let @xmath11 denote a measurement matrix .
the convex optimization problem arising in cs is formulated as follows : @xmath12 where @xmath13 is a sparse vector with @xmath5 principal elements and the remaining elements can be ignored ; @xmath11 is an @xmath14 matrix with @xmath15 , that is incoherent with @xmath10 .
it has been shown that two properties govern the design of a stable measurement matrix : _ restricted isometry property _ and _ incoherence property _ @xcite .
a @xmath5-sparse signal @xmath16 of length-@xmath7 can be recovered from @xmath17 samples provided @xmath18 and @xmath19 satisfies @xmath20 where @xmath21 is an arbitrary @xmath5-sparse signal and @xmath22 .
this property is referred to as _ restricted isometry property _ ( rip ) .
the _ incoherence _ property suggests that the rows @xmath23 of @xmath24 should be incoherent with the columns @xmath25 of @xmath10 .
let us consider an sfr system that transmits @xmath26 pulses and waits for echoes of all pulses to return before it starts any processing .
the frequency of @xmath27 pulse equals @xmath28 where @xmath29 is the starting frequency and @xmath30 .
the @xmath27 transmit pulse is of the form @xmath31 .
the signal reflected by a target at distance @xmath32 moving with speed @xmath33 is @xmath34 where @xmath2 is the speed of light , @xmath35 is the pulse repetition interval ( pri ) .
here we assume that @xmath33 is small enough to be considered constant within the pulse interval . as we assume each pulse to be narrowband
, we can ignore the delay of @xmath36 in the signal envelope , and just consider the phase shift for extracting the information on targets .
therefore we have @xmath37 and the output of phase detector is of the form @xmath38 the exponent of ( [ phase ] ) can be written as @xmath39 the first term in ( [ phase_decomp ] ) represents a constant phase shift due to the starting frequency , while the second term represents the phase shift due to frequency offset of the @xmath40th pulse .
the maximum unambiguous range and range resolution for step - frequency radar are given by @xmath41 and @xmath42 respectively .
here @xmath43 is the total effective bandwidth of the signal over @xmath26 pulses .
targets which are at distance @xmath44 will be seen by the system to be at distance @xmath45 . the third term of ( [ phase_decomp ] )
gives the doppler frequency shift experienced by the signal due to the target speed @xmath33 .
the fourth term of ( [ phase_decomp ] ) represents the frequency spread due to target speed .
this has the effect of spreading the energy of the main lobe at the target position .
let us take the transmitter and receiver to be co - located and employ @xmath26 pulses for estimating the range and speed of targets .
let us discretize the range space as @xmath46 $ ] , and the speed space as @xmath47 $ ] .
the whole target scene can be described using @xmath48 grid points in the range - speed plane .
the range and speed spaces discretization steps are @xmath49 and @xmath50 , respectively .
we assume that the targets can be present only on the grid points . by representing the target scene as a matrix @xmath51 of size @xmath48 , equation ( [ phase ] ) becomes @xmath52 where @xmath53 and @xmath54 represents zero - mean white noise . putting the outputs of phase of the phase detector , i.e. , @xmath55 in vector @xmath56 , we get @xmath57 where @xmath58^t$ ] , @xmath59 represents white zero - mean measurement noise , and the elements of matrix @xmath60 equal @xmath61 for @xmath62 , @xmath63 .
we can think of the basis matrix @xmath64 as being a stack of column vectors @xmath65 , i.e. , @xmath66 where each @xmath67 is of size @xmath68 containing the phase detector outputs for all @xmath26 pulses corresponding to the phase shift due to a target located at the @xmath69 grid point .
thus , @xmath70 accounts for the phase shift of all possible combinations of range and speed .
taking the measurement matrix @xmath24 to be an @xmath71 identity matrix yields @xmath72 .
based on ( [ y ] ) we can recover @xmath73 by applying the dantzig selector to the convex problem ( @xcite ) @xmath74 according to @xcite , the sparse vector @xmath75 can be recovered with very high probability if @xmath76 , where @xmath77 is a positive scalar , @xmath78 is the maximum norm of columns in the sensing matrix @xmath79 and @xmath80 is the variance of the noise in ( [ matrixeqn ] ) . a lower bound is readily available , i.e. , @xmath81 .
also , @xmath82 should not be too large because in that case the trivial solution @xmath83 is obtained .
thus , we may set @xmath84 .
in conventional sfr systems , the idft algorithm requires the columns of the transform matrix to be orthogonal .
the range resolution in space depends on the frequency resolution in the fourier domain .
therefore , these systems require @xmath85 pulses in order to have a range resolution of @xmath86 . for the proposed approach we can use @xmath87 pulses and still achieve a range resolution of @xmath86
. for moving targets , the conventional idft method for estimating range and speed observes a shift in the target positions ( due to speed ) and a spreading effect around the shifted position ( due to the fourth term in equation ( [ phase_decomp ] ) ) .
these effects degrade the receiver performance causing erroneous range estimation and sometimes missing the target completely . in the proposed approach , since @xmath60 has columns corresponding to all the possible range - speed combinations , the estimated results are comparatively more accurate .
* stationary targets - * our simulations use the following parameter values : @xmath88mhz , @xmath89khz , number of grid points @xmath90 .
these values of @xmath4 give @xmath91 , and @xmath92 .
we assume that the stationary point targets are present on the grid points .
@xmath93 iterations of sparse target vectors are generated and the _ estimation accuracy _ is computed as the ratio of number of iterations for which the target ranges are correctly estimated to the total number of iterations .
the basis matrix @xmath94 of size @xmath14 is generated according to equation ( [ matrix ] ) .
the measurement matrix @xmath95 is an identity matrix @xmath96 of size @xmath97 .
the optimization algorithm used to solve equation ( [ dantzig ] ) was obtained from @xcite .
the number of pulses , @xmath26 , controls the column correlation of the measurement matrix for a given value of @xmath7 ( number of grid points ) . lowering the correlation between adjacent columns of @xmath98 increases the isolation among the columns , which results in better range estimation . for the measurement matrix generated by using equation ( [ matrix ] ) ,
the adjacent column cross - correlation equals @xmath99 figure [ 100_5 ] shows the effect of changing the column correlation on the estimation accuracy of the cs sensing matrix when the unambiguous range is divided into @xmath93 grid points .
figure [ acc_snr ] shows the accuracy of the cs detector in the presence of noise at different snr values for the case in which only @xmath100 targets are within the detectable range .
the noise signal added at the received signal was gaussian zero - mean with variance @xmath101 , where the variance @xmath101 changes with snr .
figure [ acc_comp ] compares the performance of the cs detector with the conventional idft detector for @xmath102 for a target scene containing @xmath100 stationary targets . as it can be seen , the cs detector performs better than the idft detector for all snrs .
figures [ acc_snr ] and [ acc_comp ] show that we can use @xmath87 pulses to obtain @xmath103 , provided that @xmath104 .
this proves that we can use lower bandwidths in cs compared to conventional techniques of idft and accurately estimate the target parameters , when the targets are sparsely present in the range - speed space . *
moving targets - * the number of grid points in the range domain is taken to be @xmath105 .
@xmath106 grid points were used for discretizing the speed axis .
the carrier frequency is @xmath107 and the step frequency @xmath108 .
the ranges and speeds of @xmath109 targets are generated randomly in each of @xmath93 monte carlo runs . out of @xmath109 targets ,
two targets are placed on adjacent grid points on the range - speed space in each run .
the accuracy of the detector is computed as the ratio of the number of runs for which all target ranges and speeds have been estimate accurately to the total number of runs . in fig .
[ n_100_different_snr ] , we show the detection accuracy of cs and idft methods for different values of snr . for moving targets , the idft method requires speed compensation before performing idft .
since the target speed is unknown , we compensate the received signal with all possible speed and choose the one with the highest and sharpest idft output .
as it can be seen , the proposed cs significantly improves the detection accuracy as compared to the idft method .
the advantage of the cs approach is more obvious at low snr .
figure [ snr_15_different_pulses ] compares the detection accuracy of the cs and the idft methods for different number of pulses for @xmath110 .
we can easily see that the proposed method requires much fewer pulses than the idft method to achieve the same accuracy level .
for example , the cs approach requires @xmath111 pulses to achieve detection accuracy of @xmath112 , while the idft method needs about @xmath113 pulses in this particular case considered in our simulations .
one trade - off that is not apparent from the simulations is computation time .
convex optimization techniques have much higher computation cost as compared to the idft .
the basis pursuit ( bp ) algorithm used in our simulations has a computation complexity of @xmath114 .
thus the processing speed of the receiver system may put a limit on the number of grid points @xmath115 that can be used and the range resolution @xmath1 .
however there have been other algorithms like orthogonal matching pursuit ( omp ) which have computation complexities of @xmath116 . a decoupled range - speed estimation approach along the lines of @xcite
could also be employed here to reduce complexity .
we have proposed a cs - based sfr system for joint range - speed estimation .
it has been shown by our simulation results that the proposed cs approach can achieve high resolution while employing lower effective bandwidth than traditional sfr systems .
unlike the idft method , the proposed approach does not suffer from range shift and range spreading around the shift positions caused by the movement of targets . | this paper proposes a novel radar system , namely step - frequency with compressive sampling ( sfr - cs ) , that achieves high target range and speed resolution using significantly smaller bandwidth than traditional step - frequency radar .
this bandwidth reduction is accomplished by employing compressive sampling ideas and exploiting the sparseness of targets in the range - speed space . |
the celebrated rogers - ramanujan identities and their generalizations ( see @xcite , @xcite ) have influenced current research in many areas of mathematics and physics ( see @xcite ) .
these identities can be expressed as : @xmath0 and @xmath1 identities ( [ rr1 ] ) and ( [ rr2 ] ) have a natural combinatorial interpretation in terms of partitions , which was generalized by gordon ( @xcite , @xcite , theorem 7.5 ) .
a partition of a positive integer @xmath2 is a finite , non - increasing sequence of positive integers , called parts , whose sum is @xmath2 .
[ gordon ] [ gordon ] for @xmath3 and @xmath4 , the number of partitions of @xmath2 of the form @xmath5 , where @xmath6 and at most @xmath7 of the parts are 1 is equal to the number of partitions of @xmath2 into parts not congruent to 0 , @xmath8 , or @xmath9 modulo @xmath10 . setting @xmath11 and @xmath12 in theorem [ gordon ]
gives ( [ rr1 ] ) and @xmath13 , @xmath12 gives ( [ rr2 ] ) .
about twenty years ago it was observed that these identities play an important role in the representation theory of affine lie algebras via its principal characters @xcite . in 1978
lepowsky and wilson @xcite gave the first explicit realization of the affine lie algebra @xmath14 .
this led to a new algebraic structure called the z - algebra @xcite which gave a formal foundation to study systematically the connection between affine lie algebras and combinatorial identities .
in particular , lepowsky and wilson @xcite used the z - algebra structure to construct the integrable highest weight representations of the affine lie algebra @xmath15 and gave a @xmath16-algebraic proof of the rogers - ramanujan identities and @xmath16-algebraic interpretation of the generalized rogers - ramanujan identities .
one can also use ( generalized ) rogers - ramanujan identities to construct explicitly integrable representations of other affine lie algebras .
see for example @xcite , @xcite , @xcite , @xcite , @xcite for these developments . in this connection
the @xmath16-operators still play an important role .
these operators act on a certain space @xmath17 called the vacuum space associated with the representation space @xmath18 @xcite and are parameterized by the set of roots of the associated simple lie algebra @xmath19 .
however , on @xmath18 many of these @xmath16-operators are scalar multiples of each other .
for example , let us consider the affine lie algebra @xmath20 and its integrable highest weight representation @xmath21 , with highest weight @xmath22 , where @xmath23 are the fundamental weights of the lie algebra @xmath24 .
the principal character of @xmath21 is @xmath25 where @xmath26 . in this case
there are two independent families of @xmath16-operators : @xmath27 for @xmath28 and @xmath29 , where @xmath30 are the simple roots of @xmath31 corresponding to the principal cartan subalgebra @xmath32 ( see @xcite ) . in @xcite @xmath33 has been constructed using only one set of @xmath16-operaters @xmath34 and gordon s generalization of the rogers - ramanujan identities with @xmath35 and @xmath36 : @xmath37 where @xmath38 denotes the number of partitions of @xmath2 such that the outer two of any three consecutive parts differ by at least 2 and at most two parts are 1
. however , from the representation theory point of view it would be more natural to construct the representation using both families of operators @xmath34 and @xmath39 .
it is expected that this would correspond to another expansion of the left - hand side of ( [ r3m7 ] ) , namely @xmath40 where we let @xmath41 .
the expansion ( [ genrr ] ) is a special case of andrews and bressoud s analytic generalization of the rogers - ramanujan identities : @xmath42 where @xmath43 .
andrews ( @xcite , theorem 7.8 ) derived this generalization for the case of odd modulus ( @xmath44 ) and bressoud @xcite for the case of even modulus ( @xmath45 ) . in @xcite feigin and stoyanovsky used the representation of @xcite in the homogeneous gradation to give certain combinatorial interpretations of the multisum side of the generalized rogers - ramanujan identities . later georgiev @xcite and also meurman and primc ( @xcite , @xcite , @xcite and @xcite ) related the sum sides of various generalized rogers - ramanujan type expressions to multi - color partitions by attaching colors to different roots in @xmath16-algebra type constructions of the homogeneous irreducible highest weight representations of certain affine lie algebras .
it is clear from this work that the language of multi - color partitions are suitable for z - algebraic constructions and interpretations of rogers - ramanujan type identities .
the z - operator constructions in the principal gradation motivated us to seek a new multi - color interpretation ( corresponding to the parameters @xmath46 in the sum side of ( [ verygenrr ] ) ) of the generalized rogers - ramanujan identities .
it is well - known that the product side of ( [ verygenrr ] ) can also be written as @xmath47/2}\ ] ] a result which follows from the jacobi triple product identity . in this paper
, we provide a combinatorial description of the sum side of ( [ other_sum ] ) in terms of multi - color partitions ( @xmath48 colors for identity ( [ verygenrr ] ) ) with the hope that this will give new insights into the z - operator constructions in the principal gradation .
our work builds on an interpretation of the sum side of ( [ other_sum ] ) in terms of partitions with bounds on successive ranks due to andrews @xcite and bressoud @xcite .
andrews and bressoud showed that the sum side of ( [ other_sum ] ) is the generating function for @xmath49 , where @xmath50 is the set of all partitions of @xmath2 whose successive ranks lie in the interval @xmath51 $ ] ( to be discussed in more detail in section 2 ) .
our main theorem establishes a bijection between @xmath50 and a family of multi - color partitions as described below .
[ d:2 ] for @xmath52 , a _ @xmath53-color partition _ of @xmath2 is a pair @xmath54 where @xmath55 is a partition of @xmath2 and @xmath56 is a function which assigns to each @xmath57 one of the colors @xmath58 so that if @xmath59 then @xmath60 . we say that @xmath61 is the _ color _ of the @xmath62-th part of @xmath63 . for example , @xmath64 is a 3-color partition of 38 , where the subscript of a part denotes its color . our main theorem is stated below and proved in section 2 as theorem [ main ] .
* main theorem * _ for integers @xmath8 , @xmath10 , and @xmath65 satisfying @xmath66 and @xmath67 , let @xmath68 be the set of @xmath48-color partitions @xmath69 of @xmath2 satisfying the following three conditions .
let @xmath55 .
_ \(i ) ( initial conditions ) for @xmath70 , @xmath71 \(ii ) ( color difference conditions ) for @xmath72 , @xmath73 \(iii ) ( parity condition on last color when @xmath10 is even ) for @xmath74 , if @xmath10 is even and @xmath75 , the last color , then @xmath76 then @xmath77/2}\ ] ] our results suggest several interesting problems .
our motivating problem is to make use of the multi - color partition interpretation to construct natural realizations of integrable representations of affine lie algebras , via a correspondence between the @xmath48 colors and the parameters @xmath78 of ( [ verygenrr ] ) .
however , another intriguing problem is to establish a direct bijection between the multi - color partitions and the partitions counted by the sum side of the rogers - ramanujan identity ( [ verygenrr ] ) , possibly through @xmath50 .
we note some related work for the sum side . in @xcite
, andrews gave a combinatorial interpretation of the sum side of ( [ verygenrr ] ) in terms of durfee dissection partitions .
another result , due to burge @xcite and formulated in terms of lattice paths by bressoud @xcite , interprets the sum side of ( [ verygenrr ] ) as the number of lattice paths of weight @xmath2 starting at @xmath79 which have no peak of height @xmath65 or greater .
( steps allowed in the lattice path are : @xmath80 ; @xmath81 ; and , if @xmath82 , @xmath83 .
the weight of a lattice path is the sum of the @xmath84-coordinates of its peaks . )
although we have never seen a multi - color interpretation of ( [ verygenrr ] ) , together with its conditions , made explicit , multi - color interpretations of other identities of the rogers - ramanujan type appear , for example , in @xcite , @xcite , @xcite , ( @xcite , @xcite , @xcite and @xcite ) . in section 3
we note that ideas implicit in the papers @xcite and @xcite give rise to an alternative multi - color interpretation of ( [ verygenrr ] ) . in section 4 ,
we give an example of a context in which the multi - color interpretation seems quite natural . in @xcite ,
foda and quano derive a finitization of a form of the generalized rogers - ramanujan identities .
we show that the corresponding refinement for our multi - color partitions is simply an additional constraint on the size of the largest part .
in this section , we give a new combinatorial interpretation of the product side of ( [ verygenrr ] ) .
our main tool will be a combinatorial generalization of the rogers - ramanujan identities , due to george andrews , which involves the _ successive ranks _ of a partition .
a partition @xmath85 can be visualized by its ferrers diagram , an array of dots , where @xmath86 is the number of left justified dots in the @xmath62th row .
the largest square subarray of dots in this diagram is the durfee square and the durfee square size , denoted by @xmath87 , is the length of a side .
flipping the diagram along its main diagonal , one obtains the dual diagram , associated with its dual partition @xmath88 , where @xmath89 is the number of indices @xmath90 with @xmath91 .
the sequence of successive ranks of @xmath92 is the sequence @xmath93 , where @xmath94 .
let @xmath92 be a partition of @xmath2 with successive ranks @xmath95 .
for @xmath96 , let @xmath97 i.e. , @xmath98 is the number of dots on the @xmath62-th `` angle '' of @xmath92 .
denote by @xmath99 the partition with parts : @xmath100 , @xmath101 , @xmath102 , @xmath103 .
note that @xmath99 is a partition of @xmath2 such that @xmath104 ( see figure 1 . ) for @xmath105 , @xmath106 .
[ partsize ] * proof .
* since @xmath107 , we have @xmath108 and @xmath109 , so @xmath110 and @xmath111 @xmath112 [ diff ] for @xmath113 , @xmath114 * proof . * if @xmath115 , then since @xmath116 , @xmath117 thus @xmath118 similarly , if @xmath119 , then since @xmath120 , @xmath121 thus @xmath122 @xmath112 [ d:1 ] call a partition _ type 1 _ if successive parts differ by at least 2 .
[ type1 ] for any partition @xmath92 , @xmath99 is a type 1 partition .
* by lemma [ diff ] , @xmath123 .
@xmath112 the number of partitions of @xmath2 with all successive ranks in @xmath124 $ ] is equal to the number of type 1 partitions of @xmath2 .
the map @xmath125 is a bijection .
the number of partitions of @xmath2 with all successive ranks in @xmath126 $ ] is equal to the number of type 1 partitions of @xmath2 in which every part is larger than 1 .
the map @xmath125 is again a bijection .
we can now state andrews generalization of the rogers - ramanujan identities .
the theorem below was established by andrews for odd moduli @xmath10 @xcite and was generalized to even moduli by bressoud @xcite .
[ ab ] [ andrews - bressoud ] for integers @xmath10 , @xmath8 , satisfying @xmath66 , let @xmath50 be the set of partitions of @xmath2 with all successive ranks in the interval @xmath127 $ ] . then @xmath49 is equal to the number of partitions of @xmath2 with no part congruent to 0 , @xmath8 , or @xmath9 modulo @xmath10 .
the smallest @xmath10 in the above theorem is @xmath128 , as when @xmath129 and @xmath13 , the interval @xmath127 $ ] and the set @xmath50 are both empty . to see that theorem [ ab ]
generalizes the rogers - ramanujan identities , note that when @xmath11 and @xmath12 , the theorem says that the number of partitions of @xmath2 with all ranks in @xmath124 $ ] is equal to the number of partitions of @xmath2 using no part congruent to 0 , 2 or 3 modulo 5 . by remark 1 and theorem [ gordon ] , this is the first rogers - ramanujan identity ( [ rr1 ] ) .
when @xmath130 , the andrews - bressoud theorem says that the number of partitions of @xmath2 with all ranks in @xmath126 $ ] is equal to the number of partitions of @xmath2 using no part congruent to 0 , 1 or 4 modulo 5 . by remark 2 and theorem [ gordon
] , this is the second rogers - ramanujan identity ( [ rr2 ] ) .
in fact , what andrews and bressoud prove is : @xmath131/2 } \ \
= \prod_{n\not \equiv 0 , \pm r ( \bmod m)}(1-q^n)^{-1}\ ] ] where the first equality ( the hard part ) follows by a sieve argument and the second equality by application of the jacobi triple product identity .
we now show that the partitions defined by andrews rank conditions are equivalent to certain classes of multi - color partitions . [ main ] for integers @xmath8 , @xmath10 , and @xmath65 satisfying @xmath66 and @xmath67 , let @xmath68 be the set of @xmath48-color partitions @xmath69 of @xmath2 satisfying the following three conditions .
let @xmath55 .
\(i ) ( initial conditions ) for @xmath70 , @xmath71
\(ii ) ( color difference conditions ) for @xmath72 , @xmath73 \(iii ) ( parity condition on last color when @xmath10 is even ) for @xmath74 , if @xmath10 is even and @xmath75 , the last color , then @xmath76 then the number of partitions in the two sets @xmath132 and @xmath133 are the same . *
* this follows from ( [ qinfform ] ) once we establish a bijection between @xmath50 and @xmath68 .
let @xmath92 be a partition of @xmath2 with ranks @xmath134 , and assume that all ranks lie in the interval @xmath51 $ ] .
for @xmath135 , let @xmath136 , as before , and let @xmath99 be the partition of @xmath2 defined by @xmath137 .
color the parts of @xmath63 as follows : for @xmath135 , @xmath138 we show that the @xmath48-color partition @xmath54 satisfies conditions ( i)-(iii ) of the theorem .
condition(i ) . by lemma [ partsize ] , @xmath106 and from ( [ assigncolor ] ) , @xmath139 condition(ii ) . by lemma [ diff ] , @xmath140 and the conditions follow from ( [ rankfromcolor ] ) .
condition(iii ) . if @xmath141 and @xmath142 , then by definition of @xmath56 , @xmath143 but @xmath144 since @xmath145 $ ] .
it follows then that @xmath146 and therefore that @xmath147 to show this is a bijection , given @xmath148 , and @xmath65 satisfying the conditions of the theorem , let @xmath69 be a @xmath48 color partition of @xmath2 , with @xmath149 , satisfying ( i ) , ( ii ) , and ( iii ) of the theorem .
we define an inverse map which sends @xmath63 to a partition @xmath92 of @xmath2 with @xmath150 angles , where the @xmath62th angle of @xmath92 has width @xmath151 defined by @xmath152 and height @xmath153 .
we must verify that @xmath92 is a partition , i.e. that @xmath154 and @xmath155 ( clearly it has weight @xmath2 ) , and that the @xmath62th rank @xmath156 lies in the interval @xmath51 $ ] and furthermore that it satisfies ( [ assigncolor ] ) .
we first verify that for @xmath74 , @xmath157 and @xmath158 . if @xmath159 , then using ( [ inverse ] ) and condition ( i ) , @xmath160 and @xmath161 similarly , if @xmath162 , @xmath163 and @xmath164 checking the rank of the @xmath62th angle @xmath165 which satisfies ( [ assigncolor ] ) .
now we verify that for @xmath72 , @xmath166 and @xmath167 .
we check each of three cases using ( [ inverse ] ) and condition ( ii ) of the theorem .
if @xmath168 , then @xmath169 @xmath170 if @xmath171 , then @xmath172 @xmath173 finally , if @xmath174 , then @xmath175 @xmath176 @xmath112 the following result is an immediate consequence of theorem [ main ] and ( [ qinfform ] ) .
@xmath177/2 } \ \
= \prod_{n\not \equiv 0 , \pm r ( \bmod m)}(1-q^n)^{-1}.\ ] ] @xmath112 see figure [ example1 ] for an example of the bijection when @xmath178 , and @xmath179 .
figure [ example2 ] shows the bijection when @xmath180 , and @xmath179 .
an alternative coloring is to color the part @xmath98 , derived from the @xmath62th angle with rank @xmath181 , by : @xmath182 and proceed to formulate the conditions required to make this mapping a bijection between @xmath183-color partitions of @xmath2 which satisfy the conditions and the set @xmath184 of partitions of @xmath2 with all ranks in @xmath185 $ ] .
this is the coloring which would follow by applying coloring ideas in agarwal and andrews ( for other families ) @xcite to the partition family @xmath50 .
this same multi - coloring , @xmath186 , arises by applying the coloring scheme in agarwal and bressoud @xcite ( where it was used on a different family ) to the lattice path interpretation of @xmath50 described in bressoud @xcite , namely , for each peak of height @xmath187 at location @xmath84 in the lattice path , create a part of size @xmath84 with color @xmath187 .
this can be shown to be equivalent to the coloring ( [ color2 ] ) using the bijection between the lattice path and rank interpretations described by bressoud in @xcite .
an advantage of our coloring is that the color depends only on @xmath8 and @xmath181 and _ not on @xmath65_. this may make it easier to establish a direct connection between the multi - sum and our multi - color interpretation .
the hope is that for our particular goals , this interpretation will be more fruitful in the lie algebra setting .
in view of ( [ verygenrr ] ) and the second equality in ( [ qinfform ] ) , the generalized rogers - ramanujan identities can be written as @xmath188/2 } \ \ = \sum_{n_1 \geq \ldots \geq n_{k-1}\geq 0 } \frac{q^{n_1 ^ 2+n_2 ^ 2 + \ldots + n_{k-1}^2 + n_r + n_{r+1 } + \ldots + n_{k-1 } } } { ( q)_{n_1-n_2 } \ldots ( q)_{n_{k-2}-n_{k-1}}(q^{2-s } ; q^{2-s})_{n_{k-1}}},\ ] ] where @xmath45 if @xmath10 is even and @xmath44 if @xmath10 is odd .
in fact , this is the form of interest in recent work relating rogers - ramanujan identities to applications in statistical mechanics and conformal field theory . in these applications ,
the left - hand side is the _ bosonic form _ and the right - hand side is the _
fermionic form_. in the general case , the bosonic form is associated with a character of the minimal model of a virasoro algebra . in @xcite and later in @xcite a generalized fermionic pauli exclusion principle was discovered and discussed in connection with higher level representations of the affine lie algebra @xmath189 . using similar ideas
the sum side of the generalized rogers - ramanujan identities has recently been studied in several papers , and it has been shown that under certain restrictions , every bosonic form has a fermionic form and , even more , there is a corresponding finitization @xcite . in @xcite , motivated by these connections to physics , foda and quano derived finite approximations of the identities ( [ physicsform ] ) .
we will show that the multi - color partition interpretation also provides an interpretation of the left - hand side for the finitization in this case .
we first make an observation .
for @xmath66 and @xmath190 , let @xmath191 be the set of partitions of @xmath2 with all ranks in the interval @xmath127 $ ] and whose ferrer diagrams are contained in a @xmath192 rectangle .
[ box ] for @xmath66 and @xmath190 , the set @xmath191 is in one - to - one correspondence with the set of @xmath48-color partitions @xmath69 of @xmath2 in @xmath68 with the following additional constraint @xmath193 * proof .
* under our map in the proof of theorem [ main ] , the partitions in @xmath191 are mapped into @xmath48 color partitions @xmath194 satisfying ( i ) , ( ii ) , and ( iii ) in theorem [ main ] . since
this map was shown to be a bijection , it suffices to show that its image is characterized by the extra condition ( iv ) .
so , for a @xmath48 color partition @xmath63 satisfying ( i)-(iii ) , we derive necessary and sufficient conditions on @xmath63 to guarantee that under the inverse mapping its first angle has width @xmath195 and height @xmath196 . the inverse map ( [ e : invmap ] )
has @xmath197 so @xmath198 is equivalent to @xmath199 which in turn is equivalent to @xmath200 on the other hand , the height @xmath201 must satisfy @xmath202 which is equivalent to @xmath203 , so we require @xmath204 which is equivalent to @xmath205 combining ( [ ubound ] ) and ( [ vbound ] ) gives the lemma .
@xmath112 in @xcite ,
foda and quano derive the following finitized rogers - ramanujan identity : @xmath206/2}\left[\begin{array}{c } n \\ \lfloor\frac{n - k+r-(2k+1)j}2\rfloor\end{array}\right]_q \\ & = & \sum q^{n_1 ^ 2+\cdots+n_{k-1}^2+n_{r}+\cdots+n_{k-1 } } \prod_{j=1}^{k-1 } \left[\begin{array}{c } n-2(n_1+\cdots+n_{j-1})-n_j - n_{j+1}-a_{rj}^{(k)}\\ n_j - n_{j+1}\end{array}\right]_q , \nonumber\end{aligned}\ ] ] where the sum runs over @xmath207 such that @xmath208 and @xmath209 is the @xmath210-entry of the @xmath211 matrix @xmath212 when @xmath213 , the identity reduces to ( [ physicsform ] ) in the case of odd @xmath10 and @xmath44 . foda and quano
prove that the left - hand side of ( [ finitizeodd ] ) is the generating function for the set of partitions @xmath214 in the special case that @xmath3 , @xmath215 , and @xmath216 . combining this with lemma ( [ box ] ) we get the following . for @xmath3 and @xmath217 ,
the left - hand side of ( [ finitizeodd ] ) is the generating function for the set of @xmath48-color partitions @xmath69 of @xmath2 in @xmath68 with the following additional constraint @xmath218 in particular , @xmath219 . in @xcite ,
foda and quano derive the following additional finitized rogers - ramanujan identity : @xmath220_q = \sum q^{n_1 ^ 2+\cdots+n_{k-1}^2+n_{r}+\cdots+n_{k-1 } } \times\\ \prod_{j=1}^{k-2 } \left[\begin{array}{c } 2n-2(n_1+\cdots+n_{j-1})-n_j - n_{j+1}-b_{rj}^{(k)}\\ n_j - n_{j+1}\end{array}\right]_q \left[\begin{array}{c } n-(n_1+\cdots+n_{k-2})\\ n_{k-1}\end{array}\right]_{q^2},\nonumber\end{aligned}\ ] ] where the sum runs over @xmath207 such that @xmath221 and @xmath222 is the @xmath210-entry of the @xmath223 matrix @xmath224 when @xmath213 , the identity reduces to ( [ physicsform ] ) in the case of even @xmath10 and @xmath45 .
foda and quano prove that the left - hand side of ( [ finitizeeven ] ) is the generating function for the set of partitions @xmath214 in the special case that @xmath141 , @xmath225 , and @xmath226 . combining this with lemma [ box ]
we get : for @xmath141 , the left - hand side of ( [ finitizeeven ] ) is the generating function for the set of @xmath48-color partitions @xmath69 of @xmath2 in @xmath68 and the following additional constraint @xmath227 clearly , our map in theorem [ main ] , when restricted to partitions in @xmath228 with durfee square of size @xmath150 , gives a bijection with those @xmath48 color partitions of @xmath2 in @xmath68 which have exactly @xmath150 parts and satisfy ( iv ) of of lemma [ box ] .
* acknowledgment * we would like to thank james lepowsky for helpful comments on an earlier version of this paper and christian krattenthaler for bringing the references @xcite and @xcite to our attention .
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx= partition @xmath92 rank vector the 2-color partition + of 10 with all of @xmath92 @xmath229 of 10 + ranks in @xmath230 $ ] + + @xmath231 @xmath232 $ ] @xmath233 + @xmath234 @xmath235 $ ] @xmath236 + @xmath237 @xmath238 $ ] @xmath239 + @xmath240 @xmath241 $ ] @xmath242 + @xmath243 @xmath244 $ ] @xmath245 + @xmath246 @xmath247 $ ] @xmath248 + @xmath249 @xmath250 $ ] @xmath251 + @xmath252 @xmath250 $ ] @xmath253 + xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx= partition @xmath92 rank vector the 3-color partition + of 10 with all of @xmath92 @xmath229 of 10 + ranks in @xmath254 $ ] + + @xmath255 @xmath256 $ ] @xmath257 + @xmath258 @xmath241 $ ] @xmath233 + @xmath246 @xmath247 $ ] @xmath259 + @xmath260 @xmath256 $ ] @xmath236 + @xmath249 @xmath250 $ ] @xmath261 + @xmath262 @xmath263 $ ] @xmath239 + @xmath264 @xmath265 $ ] @xmath266 + @xmath267 @xmath268 $ ] @xmath242 + @xmath252 @xmath250 $ ] @xmath245 + @xmath269 @xmath270 $ ] @xmath271 + @xmath272 @xmath273 $ ] @xmath274 + @xmath275 @xmath265 $ ] @xmath248 + @xmath276 @xmath277 $ ] @xmath278 + @xmath279 @xmath280 $ ] @xmath251 + @xmath281 @xmath282 $ ] @xmath283 + @xmath284 @xmath285 $ ] @xmath253 + @xmath231 @xmath232 $ ] @xmath286 + @xmath237 @xmath238 $ ] @xmath287 + @xmath288 @xmath289 $ ] @xmath290 + @xmath243 @xmath244 $ ] @xmath291 g. e. andrews , _ q - series : their development and application in analysis , number theory , combinatorics , physics , and computer algebras _ , cbms regional conf .
ser . in math . * 66 * , amer .
soc . , providence , ri , 1986 .
a. berkovich and b. m. mccoy , _ rogers - ramanujan identities : a century of progress from mathematics to physics _ , proceedings of the international congress of mathematicians , vol .
iii ( berlin , 1998 ) , doc . math .
1998 , extra vol .
iii , 163172 ( electronic ) .
b. feigin and a. stoyanovski , _ functional models of the representations of current algebras , and semi - infinite schubert cells _ , funct .
* 28 * ( 1994 ) , no .
1 , 5572 ; quasi - particles model for the representations of lie algebras and geometry of flag manifolds , ( hep - th/9308079 ) .
g. georgiev , _ combinatorial constructions of modules for infinite - dimensional lie algebras , * i*. principal subspace . _
j. pure appl .
algebra * 112 * ( 1996 ) , 247286 ; * ii*. _ parafermionic space _ , q - alg/9504024 .
j. lepowsky and r.l .
wilson , _ the structure of standard modules , * i * : universal algebras and the rogers - ramanujan identities _ , invent . math . * 77 * ( 1984 ) , 199 - 290 ; * ii * , _ the case @xmath297 , principal gradation _
* 79 * ( 1985 ) , 417442 . | basil gordon , in the sixties , and george andrews , in the seventies , generalized the rogers - ramanujan identities to higher moduli .
these identities arise in many areas of mathematics and mathematical physics .
one of these areas is representation theory of infinite dimensional lie algebras , where various known interpretations of these identities have led to interesting applications .
motivated by their connections with lie algebra representation theory , we give a new interpretation of a sum related to generalized rogers - ramanujan identities in terms of multi - color partitions . |
our object of study is determinantal representations @xmath8 for a @xmath9-variable polynomial @xmath10 , @xmath11 , with @xmath1 . here
@xmath12 is in the set @xmath13 of @xmath9-tuples of nonnegative integers , @xmath14 , @xmath15 , and @xmath4 is a complex square matrix .
it is of interest of how and to what extent , the algebraic and operator - theoretic properties of the polynomial correspond to the size and norm of the matrix @xmath4 of its representation .
various determinantal representations of polynomials have been studied , often for polynomials over the reals : see a recent overview article @xcite , together with bibliography , and also @xcite .
the particular form of has appeared before , for instance , in @xcite .
an important early result on two - variable polynomials was obtained by a. kummert ( * ? ? ?
* theorem 1 ) . given a @xmath9-variable polynomial @xmath10 , @xmath1
, we consider the question of whether it can be represented in the form for some @xmath16 and some @xmath17 complex matrix @xmath4 , possibly subject to a constraint .
it will be shown that the unconstrained version of this question can always be answered in the affirmative ( section [ sec : affine - det - repr ] ) .
the problem of minimizing the operator norm of @xmath4 over all representations of @xmath0 will be seen to be more involved ( section [ sec : constr - det - repr ] ) .
we will say that the _ multi - degree _ @xmath18 of a polynomial @xmath0 is @xmath19 if @xmath20 is the degree of @xmath0 as a polynomial of @xmath21 , @xmath22 . the _ total degree _
@xmath23 of @xmath0 is the largest @xmath24 over all monomials @xmath25 of @xmath0 . for @xmath26
, the inequality @xmath27 will be meant in the usual component - wise sense : @xmath28 , @xmath29 . for a matrix @xmath4 , the principal submatrix determined by an index set @xmath30 will be denoted by @xmath31 $ ] . given a collection of complex numbers @xmath32 , indexed by nonempty subsets @xmath30 of @xmath33 , the principal minor assignment problem ( see , e.g. , @xcite ) consists of finding a @xmath34 matrix @xmath4 such that @xmath35=c_\alpha$ ] for all @xmath30 .
this problem is , in general , overdetermined since the number of independent principal minors grows exponentially with the matrix size , @xmath9 , while the number of free parameters , the matrix entries , is @xmath36 .
it becomes well - posed under additional assumptions on @xmath4 or @xmath9 . for theoretical and computational advances , see @xcite .
a polynomial of multi - degree @xmath37 , is said to be _ multi - affine_. for such a polynomial , the problem of finding a representation with @xmath38 is equivalent to the principal minor assignment problem .
this follows by comparing the expansion @xmath39
\prod_{i\in \alpha } z_i,\ ] ] to the general form @xmath40 of a @xmath9-variable multi - affine polynomial @xmath0 , @xmath1 . for a general polynomial @xmath0 ,
finding a determinantal representation with @xmath41 may not be possible by the same dimension count as above .
it is clear that implies that @xmath42 .
if @xmath43 is prescribed , one may view as the principal minor relation problem formulated in section [ sec : affine - det - repr ] .
this paper is largely motivated by our study @xcite of the multivariable von neumann inequality and the discrepancy between the schur and schur agler norms of analytic functions on the unit polydisk @xmath44 the _ schur class _ , consisting of analytic @xmath45-valued functions @xmath46 on @xmath47 such that @xmath48 will be denoted by @xmath49 . here
@xmath45 is the banach space of bounded linear operators from a hilbert space @xmath50 to a hilbert space @xmath51 .
the _ schur
agler class _ , introduced in @xcite , will be denoted by @xmath52 .
it consists of analytic @xmath45-valued functions on @xmath53 such that @xmath54 where the supremum is taken over all @xmath9-tuples @xmath55 of commuting strict contractions on a common hilbert space . in the scalar case @xmath56 and in the case @xmath57 , we will use respective shortcuts @xmath58 , @xmath59 , and @xmath60 , @xmath61 . for a bounded analytic function @xmath62 , the _ von neumann inequality _ is the inequality between its schur and schur agler norms : @xmath63 it is valid when @xmath64 @xcite and @xmath65 @xcite , and not always valid when @xmath66 @xcite .
thus a schur function @xmath46 is schur
agler if and only if holds .
one has the inclusion @xmath67 .
the two classes coincide when @xmath64 and @xmath65 , and the inclusion is proper when @xmath66 .
see , e.g. , @xcite for details .
a @xmath9-variable polynomial is said to be _ stable _ if it has no zeros in @xmath68 , and _ semi - stable _ if it has no zeros in @xmath69 .
a rational function in @xmath70 is said to be _ inner _ if its radial limits are unitary ( unimodular , in the scalar case ) almost everywhere on the @xmath9-torus .
every scalar - valued rational inner function is necessarily of the form @xmath71 for some @xmath72 and a semi - stable polynomial @xmath0 ( * ? ? ?
* theorem 5.5.1 ) , where @xmath73 .
a rational inner function @xmath74 is said to have a _ transfer - function realization _ ( of order @xmath75 ) if there exists a unitary matrix @xmath76 \in { \mathbb c}^{(1+|m|)\times ( 1+|m|)}\ ] ] so that @xmath77 such a realization for a scalar - valued rational inner @xmath46 exists if and only if @xmath78 @xcite , ( * ? ? ?
* theorem 2.9 ) . in section [ wedge ] , we explore the schur agler class in the context of exterior products , proving , in particular , that if @xmath79 is a matrix - valued schur agler function , then so are its determinant @xmath80 and permanent @xmath81 .
following @xcite , we will say that a semi - stable polynomial @xmath0 is _ an agler denominator _ if the rational inner function @xmath82 is schur agler . extending this notion
, we will call a semi - stable polynomial @xmath0 an _ eventual agler denominator _ of order @xmath16 if @xmath83 is schur agler . representations
may allow for a fresh approach to the study of the multivariable von neumann inequality . in section [ ad
] , we examine the discrepancy between the schur and schur agler classes via ( eventual ) agler denominators .
it is shown that ( i ) not every ( semi-)stable polynomial is an agler denominator , ( ii ) if @xmath5 is a polynomial in @xmath84 with @xmath85 , then @xmath7 admits a representation for some @xmath86 and @xmath4 a contraction , and ( iii ) ( building on results in section [ wedge ] ) if @xmath0 is representable in the form for some @xmath86 and contractive @xmath4 , then @xmath0 is an eventual agler denominator of order @xmath43 . as a corollary
, we deduce that every semi - stable linear polynomial is an agler denominator , thus solving a problem suggested in @xcite . to illustrate a possible advantage of our approach
, we compare a minimal determinantal representation to a minimal transfer - function realization in remark [ rem : linear ] . in section [ sec6 ] , we revisit the kaijser varopoulous
holbrook example to build a family of polynomials in @xmath87 , for every odd @xmath66 .
if @xmath88 , this leads to a slightly improved bound for the von neumann constant .
[ thm : k - repr ] every @xmath89 $ ] , with @xmath1 , admits a representation for some @xmath90 and some @xmath91 . note that the @xmath9-tuple @xmath43 is not prescribed in the statement ,
although a bound on @xmath43 will be deduced in the proof .
if @xmath43 is specified , as in the principal minor assignment problem , theorem [ thm : k - repr ] ensures the solvability of the following problem for sone @xmath43 ( see also remark [ rem : pmrp ] ) . * the principal minor relation problem .
* let @xmath92 , and let @xmath93 , @xmath94 , be a collection of complex numbers . given @xmath95 , @xmath96 , find a matrix @xmath97 whose principal minors @xmath98 $ ] , indexed by @xmath99 , @xmath100 , , @xmath101 , satisfy the relations @xmath102=p_k,\quad 0\le k\le m.\ ] ] when @xmath103 , this is the classical principal minor assignment problem mentioned in section [ sec : intro ]
. the following standard result ( see , e.g. , ( * ? ? ? * theorem 3.1.1 ) ) will occasionally be used .
[ lem : schur - compl ] let @xmath104 be a block matrix with square matrices @xmath105 and @xmath106 . if @xmath107 , then @xmath108 .
similarly , if @xmath109 , then @xmath110 .
the proof of theorem [ thm : k - repr ] will be based on the next two lemmas .
[ lem : prod ] for every @xmath111 $ ] , there exist natural numbers @xmath112 , @xmath113 , , @xmath114 , @xmath115 , matrices @xmath116 , and diagonal @xmath117 matrix functions @xmath118 with the diagonal entries in @xmath119 , such that @xmath120 the factorization can be chosen so that @xmath121 .
we apply induction on @xmath122 . if @xmath123 , then holds trivially with @xmath124 .
suppose a representation exists for every matrix polynomial in @xmath125 ,
, @xmath126 of total degree @xmath127 .
then a polynomial @xmath111 $ ] of total degree @xmath122 can be represented in the form @xmath128 where @xmath129 $ ] is a polynomial of total degree @xmath127 . by assumption , we have @xmath130 which gives @xmath131 , with @xmath132 [ lem : det ] let @xmath133 , @xmath134 , and @xmath135 , where @xmath112 . then @xmath136 we apply induction on @xmath137 . for @xmath138 , lemma [ lem : schur - compl ]
gives @xmath139 suppose holds for @xmath127 in the place of @xmath122 .
then , again by lemma [ lem : schur - compl ] , @xmath140 applying lemma [ lem : prod ] to @xmath141 , we obtain @xmath142 ( here @xmath143 ) .
so , by lemma [ lem : det ] , @xmath144 where @xmath145 and @xmath146 since @xmath147 factors as @xmath148 , where @xmath149 it may be written in the form @xmath150 where @xmath151 and @xmath152 is a permutation matrix . representing @xmath153 as a @xmath154 block matrix , we obtain that @xmath155 hence @xmath156 and , in particular , the matrix @xmath157 is invertible .
therefore , by lemma [ lem : schur - compl ] , @xmath158 with @xmath159 .
we will now look into the existence of a determinantal representation with a norm constraint on the matrix @xmath4 .
first , we give norm - constrained versions of lemma [ lem : prod ] and theorem [ thm : k - repr ] .
[ lem : prod - constr ] for every polynomial @xmath160 , a factorization exists with constant contractive matrices @xmath161 , @xmath162 , where @xmath163 .
let @xmath164 be a polynomial . by ( * ?
* corollary 18.2 ) , @xmath5 can be written as a product of constant contractive matrices and diagonal matrices with monomials on the diagonal .
every such diagonal matrix is , in turn , a product of matrices @xmath118 as in ( interlacing with @xmath165 ) .
[ thm : q - matrix ] let @xmath0 be a polynomial of the form @xmath166 , where @xmath5 is a schur agler polynomial with matrix coefficients , i.e. , @xmath167 $ ] and @xmath168 . if @xmath1 , then holds with @xmath4 a contraction . by lemma [ lem : prod - constr ] , the matrices @xmath161 in the factorization can be chosen contractive .
then the matrix @xmath153 as in the proof of theorem [ thm : k - repr ] is also contractive , and by the standard closed - loop mapping argument , @xmath4 is contractive as well . for reader
s convenience , we include this argument . given @xmath169 , the vector equation @xmath170 in @xmath171 and @xmath172 has a unique solution @xmath173 , @xmath174 .
since @xmath153 is a contraction , we have @xmath175 , i.e. , @xmath176 .
since @xmath169 is arbitrary , @xmath4 is a contraction as claimed .
[ cor : q - scalar ] let @xmath89 $ ] , with @xmath1 , be such that @xmath177 , then holds with @xmath4 a contraction .
[ rem : converse - false ] the converse to corollary [ cor : q - scalar ] is false . indeed , if @xmath64 and @xmath178 , then @xmath179 satisfies with @xmath180 , obviously a contraction .
however , @xmath181 .
define the _ stability radius _ @xmath182 of a @xmath9-variable polynomial @xmath0 to be @xmath183 clearly , @xmath0 is semi - stable if @xmath184 , and stable if @xmath185 .
it is easy to see that @xmath186 whenever @xmath0 admits . with respect to a given subalgebra @xmath187
, the structured singular value @xmath188 of a matrix @xmath189 is defined to be @xmath190 the theory of structured singular values was introduced in @xcite to analyze linear systems with structured uncertainties ; for an overview , see for instance ( * ? ? ?
* chapter 10 ) . if @xmath0 satisfies and @xmath191 , we recognize that @xmath192 .
the next theorem gives a way of constructing a representation with a certain upper bound on the norm of @xmath4 .
[ thm : norm - bds ] given a polynomial @xmath193 , where @xmath194 and the coefficients @xmath195 are nonzero , let @xmath196 , @xmath197 , and @xmath198 .
then @xmath0 admits a representation with @xmath97 , and @xmath199 for some integer @xmath200 , @xmath201 .
[ rem : beta ] if @xmath202 and @xmath203 , which is the case for semi - stable linear polynomials , the norm bound asserted in the theorem is sharp . in general , it is not sharp , even in the univariate case .
[ rem : pmrp ] theorem [ thm : norm - bds ] implies that the principal minor relation problem with data @xmath204 is solvable for @xmath197 .
form the matrices @xmath205,\ c_1=\ldots = c_{t-1}=\beta i_{|s|},\ c_t=\beta^{\frac{1-t}{2}}\ \underset{k\in s}{\operatorname{col}}\bigg[\frac{p_k}{|p_k|^{\frac12}}\bigg],\ ] ] all of equal norm @xmath206 .
relative to the standard ordering of factors , @xmath207 write each monomial @xmath208 , @xmath209 , as an expanded product @xmath210 , where @xmath211 , for @xmath212 , and @xmath213 , for @xmath214 .
let @xmath215,\quad j=1,\ldots , t,\ ] ] and observe that @xmath216 contains no unit entries by construction .
it is then easy to check that holds for @xmath141 .
thus , by lemma [ lem : det ] , we obtain @xmath217 where @xmath218 and @xmath219 are as in . using an appropriate permutation @xmath152 , we can bubble - sort @xmath219 so that all diagonal ones are stacked in the left upper corner block : @xmath220 where @xmath221 .
then , partitioned accordingly , @xmath222 has the following structure : @xmath223 we observe that @xmath224 is nilpotent of index @xmath200 , where @xmath200 is the number of nonzero blocks @xmath225 ( counting @xmath226 ) . since , @xmath227 , we necessarily have @xmath201 .
thus , we obtain that @xmath228 where @xmath229 by the same argument as in the proof of theorem [ thm : q - matrix ] .
the norm bound on @xmath4 is obtained as follows .
for a fixed @xmath169 , the solution to the vector equation @xmath170 in @xmath230 and @xmath231 is given by @xmath232 observing that @xmath233 , we have @xmath234 if @xmath235 , then @xmath236 and thus @xmath237 .
if @xmath238 , then the estimate @xmath239 implies that @xmath240 which yields .
[ rem : question ] given a @xmath9-variable polynomial @xmath0 , @xmath1 , and a @xmath9-tuple @xmath42 such that holds , one may consider the set @xmath241 of @xmath242 matrices @xmath4 such that @xmath243 .
it is then of interest to determine the constant @xmath244 in particular , it is unclear whether @xmath245 ( @xmath246 ) for @xmath0 stable ( semi - stable ) .
we will now examine the schur
agler norm of tensor and exterior products of operator - valued functions .
the results are preceded by some definitions . for a background on tensor and exterior algebras
see , e.g. , @xcite .
let @xmath247 be the @xmath248-fold tensor power of a vector space @xmath249 .
the @xmath248-th antisymmetric tensor power @xmath250 of @xmath249 may be viewed as a subspace of @xmath247 , generated by elementary antisymmetric tensors @xmath251 where the summation is taken over all permutations @xmath252 of @xmath253 .
given a linear map @xmath254 of vector spaces , the linear operator @xmath255 , determined by the equalities @xmath256 is the compression @xmath257 of the tensor power @xmath258 .
here @xmath259 denotes the orthogonal projection onto a subspace @xmath260 . if @xmath261 form a basis for @xmath249 , then @xmath262 form a basis for @xmath250 of cardinality @xmath263 .
relative to a choice of bases for @xmath264 and @xmath249 , the matrix entry for @xmath265 in row - column position @xmath266 is the minor of the matrix of @xmath105 built from rows @xmath267 and columns @xmath268 .
if @xmath269 and @xmath270 are normed vector spaces and if @xmath271 is a power series with coefficients in @xmath272 , then , for any tuple @xmath273 of commuting operators on some normed vector space @xmath274 , we may consider the operator @xmath275 acting from @xmath276 to @xmath277 , provided the series converges .
more generally , starting with power series @xmath278 , @xmath279 with coefficients in @xmath272 , and @xmath280 as above , form the operator @xmath281 and its compression @xmath282 the objective of this section is to establish the following theorem .
let @xmath264 and @xmath249 be hilbert spaces and let @xmath283 belong to @xmath284 .
then @xmath285 belongs to @xmath286 and @xmath287 belongs to @xmath288 .
let @xmath280 be a tuple of commuting strict contractions on some hilbert space @xmath289 .
then the mapping @xmath290 is contractive as a product of contractive factors .
hence @xmath291 belongs to @xmath292 .
consequently , @xmath293 which gives the second assertion .
[ detcor ] let @xmath79 be a @xmath294 matrix - valued schur
agler function , i.e. , @xmath295 .
then , for every @xmath296 , the @xmath248-th compound matrix - valued function of @xmath79 is also schur agler .
in particular , @xmath297 is a schur agler function .
the matrix of @xmath298 is the @xmath248-th compound matrix of @xmath79 .
the case @xmath299 corresponds to @xmath297 .
similarly , in the setting of @xmath248-th symmetric tensor powers , one may consider the operators @xmath300 .
the proof of the following theorem is omitted as it parallels the preceding development .
let @xmath264 and @xmath249 be hilbert spaces and let @xmath283 belong to @xmath284 .
then @xmath301 belongs to @xmath302 .
let @xmath79 be a @xmath294 matrix - valued schur agler function , i.e. , @xmath295 .
then , for every @xmath296 , the @xmath248-th permanental compound matrix - valued function of @xmath79 is also schur agler . in particular , the permanent of a schur
agler function is also schur agler .
we note that a permanental analog of features in @xcite .
we are in a position to discuss ( eventual ) agler denominators and stability in relation to .
it will first be shown that there exist stable polynomials in three or more variables that are not agler denominators .
let @xmath10 be a @xmath9-variable polynomial , with @xmath303 and multi - degree @xmath304 , violating the von neumann inequality .
let there exist a tuple @xmath280 of commuting contractions such that @xmath305 , for some @xmath306 , and @xmath307 .
examples of such a scenario can be found in @xcite ; see also section [ sec6 ] . for @xmath308 ,
the polynomial @xmath309 is stable and so the rational function @xmath310 is inner .
however , since @xmath311 , @xmath46 does not belong to @xmath84 whenever @xmath312 .
in particular , if the multi - degree of @xmath313 is also @xmath304 , then @xmath314 , so that @xmath5 is not an agler denominator . to give a concrete example ,
we specialize to the kaijser varopoulos holbrook setting .
the polynomial @xmath315 satisfies @xmath303 , and there exist commuting contractions @xmath316 such that @xmath317 and @xmath318 .
the corresponding rational inner function @xmath319 is not schur
agler for @xmath320 . for these values of @xmath321 ,
the stable polynomial @xmath322 is not an agler denominator .
we now have the following result . [ aglerden ]
let a polynomial @xmath0 admit a representation for some @xmath72 and contractive @xmath4 .
then @xmath323 in particular , @xmath0 is an eventual agler denominator of order @xmath43 .
if @xmath324 , then @xmath0 is an agler denominator .
a lemma is needed ; see , e.g. , ( * ? ? ?
* theorem 3.1.2 ) .
[ commutingac ] let @xmath325 , and @xmath106 be square matrices of the same size , and suppose that @xmath326 .
then @xmath327 by lemma [ lem : schur - compl ] , the right hand side of equals @xmath328 let @xmath329 be a singular value decomposition of @xmath4 .
then the numerator of equals @xmath330 applying lemma [ commutingac ] , noting that @xmath331 and @xmath332 commute , we get that equals @xmath333 @xmath334 to prove it remains to observe that @xmath335 @xmath336 where in the last step we used that @xmath337 . as the julia operator @xmath338 is unitary ,
the multivariable rational inner matrix function @xmath339 is in the schur
agler class . by corollary [ detcor ]
, so is its determinant , and thus @xmath340 is in the schur
agler class .
[ cor : shrink ] for every @xmath89 $ ] with @xmath1 , there exists @xmath341 such that the polynomial @xmath342 is an eventual agler denominator .
in fact , if @xmath0 is given by , then one can choose any @xmath343 . since , by theorem [ thm : k - repr ] , every polynomial @xmath0 with @xmath1 admits a representation , the assertion follows from the identity @xmath344 and theorem [ aglerden ] .
[ rem : shrink ] the first statement of corollary [ cor : shrink ] can also be deduced from corollary [ cor : q - scalar ] : since @xmath345 , the inequality @xmath346 holds , by continuity , for a sufficiently small @xmath341 . for multi - affine symmetric
polynomials a stronger statement is true ( * ? ? ?
* theorem 1.5 ) : @xmath347 is an agler denominator for sufficiently small @xmath341 .
following @xcite , we call a semi - stable polynomial @xmath0 _ scattering schur _ if @xmath0 and @xmath348 have no factor in common . in (
* theorem 1 ) it was proven that every two - variable scattering schur polynomial @xmath0 of degree @xmath349 is of the form with @xmath4 an @xmath350 contraction .
thus every two - variable scattering schur polynomial @xmath0 is an agler denominator .
the following result provides a partial converse to theorem [ aglerden ] .
[ aglerdenrev ] let @xmath0 be a @xmath9-variable scattering schur polynomial with @xmath1 .
if , for some @xmath351 , the rational inner function @xmath352 has a transfer - function realization of order @xmath304 , then @xmath0 admits a representation with @xmath353 and @xmath4 a contraction .
taking the determinant of both sides of the equality @xmath354 and using lemma [ lem : schur - compl ] , we obtain @xmath355 note that both @xmath356 and @xmath357 are of degree at most @xmath304 .
we now obtain that @xmath358 as @xmath0 is scattering schur we must have that @xmath10 divides @xmath357 , say @xmath359 . dividing out @xmath10 in the above equation , we obtain that @xmath360 as the left hand side has degree @xmath361 and the right hand side degree at most @xmath304 , we obtain that @xmath5 must be a constant .
but then , using @xmath1 and @xmath362 , we obtain that @xmath363 , and thus @xmath364 with @xmath365 .
[ affine ] the polynomials @xmath366 with @xmath367 , are agler denominators .
let @xmath4 be a @xmath34 rank 1 contraction with diagonal entries @xmath368 .
one such choice is given by @xmath369_{j , k=1}^d.\ ] ] then @xmath370 and the result follows directly from theorem [ aglerden ] .
[ rem : min ] the matrix @xmath4 in the proof of corollary [ affine ] is clearly of minimal size .
it is also of minimal norm , @xmath371 , for otherwise @xmath372 would be stable .
it was shown in ( * ? ? ? * theorem 3.3 ) that a multi - affine symmetric polynomial is an agler denominator if and only if a certain matrix @xmath373 constructed from the christoffel
darboux equation is positive semidefinite .
subsequently , for @xmath374 , the positivity of the matrix @xmath375 was computationally checked up to @xmath376 . using our corollary [ affine ]
, we deduce this fact for all @xmath9 .
let @xmath377 , where @xmath378 , and let @xmath379 be the @xmath380 matrix indexed by subsets @xmath381 of @xmath382 , defined via : @xmath383 where @xmath384 , @xmath385 , @xmath386 , @xmath387 , and @xmath388 for @xmath389 not satisfying @xmath390 .
then @xmath391 is positive semidefinite . to illustrate , we choose @xmath122 real and @xmath88 : @xmath392 where @xmath393 [ rem : linear ] in the context of
, the question of whether a given polynomial is an ( eventual ) agler denominator is reduced to analyzing the matrix of its determinantal representation .
this provides a possible alternative to the transfer - function realization method .
for example , @xmath394 admits a representation with @xmath395 which is minimal both in size and in norm ; see remark [ rem : min ] . at the same time , the minimal order of a transfer - function realization of @xmath396 is @xmath397 @xcite .
for @xmath398 real , consider the multivariable polynomial @xmath399 the case of @xmath88 and @xmath400 corresponds to the polynomial from @xcite .
[ sharper ] let @xmath401 , @xmath402 , and @xmath0 be defined as above .
then @xmath403 , if @xmath9 is even , and @xmath404 , if @xmath9 is odd , while @xmath405 for all @xmath9 . in particular
, @xmath406 is not in @xmath84 for odd @xmath401 .
write @xmath407 , where @xmath408 observe that @xmath105 is symmetric with eigenvalues @xmath409 and @xmath410 , so @xmath411 hence @xmath412 .
if @xmath9 is even , one immediately has equality since @xmath413 .
if @xmath9 is odd , the inequality is strict .
indeed , otherwise @xmath414 would be maximized for some unimodular @xmath415 with zero sum , as @xmath416 would then lie in the eigenspace of @xmath105 corresponding to @xmath409 .
but then the equality @xmath417 would force @xmath418 in conflict with the zero sum condition .
next , for a tuple of commuting contractions @xmath419 , we have @xmath420 so that @xmath421 choose @xmath422 to be any unit vectors in @xmath423 with zero sum . then the matrices @xmath424 are such that @xmath425 @xmath426 @xmath427 , and @xmath428 where @xmath429 is the @xmath430th standard unit vector in @xmath431 . hence @xmath432 . in the case of @xmath88 , maximizing @xmath433 , the von neumann constant of @xmath0 , over @xmath398 , we find that the maximum possible ratio is @xmath434 ( occuring for @xmath435 ) .
the previously known lower bound for the von neumann constant was @xmath436 @xcite .
j. agler . on the representation of certain holomorphic functions defined on a polydisc . in _ topics in operator theory : ernst d. hellinger memorial volume _ ,
volume 48 of _ oper .
theory adv .
_ , pages 4766 .
birkhuser , basel , 1990 .
a. grinshpan , d. s. kaliuzhnyi - verbovetskyi , v. vinnikov , and h. j. woerdeman .
classes of tuples of commuting contractions satisfying the multivariable von neumann inequality . _ j. funct .
anal . _ 256 ( 2009 ) ,
9 , 3035 - 3054 .
j. a. holbrook .
schur norms and the multivariate von neumann inequality . in _ recent advances in operator theory and related topics ( szeged , 1999 ) _ , volume 127 of _ oper .
theory adv .
_ , pages 375386 .
birkhuser , basel , 2001 .
v. vinnikov .
lmi representations of convex semialgebraic sets and determinantal representations of algebraic hypersurfaces : past , present , and future . in mathematical methods in systems , optimization , and control : festschrift in honor of j. william helton " ( eds.harry dym , mauricio c. de oliveira , mihai putinar ) , _ operator theory : advances and applications _ ,
birkhuser , to appear . | for every multivariable polynomial @xmath0 , with @xmath1 , we construct a determinantal representation @xmath2 where @xmath3 is a diagonal matrix with coordinate variables on the diagonal and @xmath4 is a complex square matrix .
such a representation is equivalent to the existence of @xmath4 whose principal minors satisfy certain linear relations .
when norm constraints on @xmath4 are imposed , we give connections to the multivariable von neumann inequality , agler denominators , and stability .
we show that if a multivariable polynomial @xmath5 , @xmath6 satisfies the von neumann inequality , then @xmath7 admits a determinantal representation with @xmath4 a contraction . on the other hand , every determinantal representation with a contractive @xmath4
gives rise to a rational inner function in the schur
agler class . |
the origin of galactic disk warping , which is typically vertical bending of disks in a integral shape , is a long - standing problem . even after 40 year efforts to explain the phenomena since the discovery of the warp in the milky way @xcite
, there is still no generally acceptable theory .
a general review of the problem can be found in .
there are several observational facts that any convincing theory must explain .
the first point is the longevity of the warps .
observations of neutral hydrogen layers @xcite and stellar disks @xcite have revealed that about a half of spiral galaxies exhibit warps .
such high frequency of warps indicate that the warps are long - lived with about the same age as galaxies , otherwise some mechanism keeps exciting the warps . in both cases the warp excitation mechanism must be ubiquitous .
the second point is the warp amplitude .
for instance the milky way has a prominent warp and its amplitude is 2 kpc at galactocentric distances @xmath5 kpc@xcite , and 4 kpc at @xmath6 kpc @xcite .
the torque that can bend the disk in such a large amplitudes is so strong , which make it difficult to find the source of the torque @xcite .
the third point is alignment of the warp nodes .
most of spiral galaxies including our own have straight line of nodes within the holmberg radius , and outward the lines of nodes make leading spirals @xcite .
this fact presents a contrast to the spiral arms in the grand - design spiral .
simple analysis @xcite shows that the simplest interpretation of the warp as a vertical oscillation of matters without considering selfgravity , is not working , because the line of nodes wind up in a time scale of 1 or 2 gyr .
therefore self - consistent treatment is necessary .
there are several hopeful theories , all of which have , however , some shortcomings . the first step to understand long - lived warps is the normal mode analysis in steady state galaxies .
showed that there are many bending normal modes in their thin - disk models , but unless density of the disks are cut - off sharply at their edges the frequency spectrum is at least partly continuous . in such disks any realistic perturbation would excite propagating wave packets ,
thus the warping wave is dispersed .
a long - lived warp can correspond only to a discrete mode .
if the disk is truncated at some radius , there is only one discrete normal mode , which is called a _ modified tilt mode _ @xcite .
showed that if disk is embedded in a flattened halo there is at most one discrete mode .
shapes of the mode show good agreement with observed warps , and even though the disk is not in the right shape of the modified tilt mode initially , it dynamically evolves to fit the mode @xcite .
the normal mode theory works very well as long as the halo is treated as a rigid body .
however , once the self - consistent response of halo is included several problems emerge .
for example , dynamical friction acting on the disk from the halo dissipates the warping motion @xcite , and also the line of nodes often winds up tightly @xcite . to find the normal modes in fully self - consistent composite systems of a disk and a halo is extremely intractable .
one possible mechanism to keep the misalignment between a disk and a halo symmetric plane is continuous addition of matter with misaligned angular momentum to the halo .
inclined cosmic infall @xcite might support the mechanism .
have shown that when infall reorientates the outer part of a galactic halo by several degrees per gyr , a realistic warp is excited .
though the result seems appealing , their model of the infall is still too idealistic , and numerical studies are relatively short ( @xmath7 1 gyr ) to explain long - lived warps .
refined studies are necessary .
another mechanism , which we are examining in this paper , is interaction with satellites .
this mechanism is intuitively plausible because the most frequent @xmath8 warp is easily associated with a tidal field from a satellite . furthermore increasing
observational evidence suggests that most of the galaxies have satellite dwarfs .
in fact , recently a faint dwarf companion has been discovered around ngc 5907 , which was considered as a typical example of a warping galaxy without interacting companions @xcite .
also statistics show positive correlation between warping and existence of interacting companions @xcite the only but severe problem of the interaction scenario is that the tidal forces from satellites are usually too small .
for example , showed that the direct tidal force from the large magellanic cloud ( lmc ) of @xmath9 at 50 kpc can excite a warp with an amplitude at largest 100 pc at 16 kpc , while the real galactic warp is about 3 kpc . against the problem ,
a remedy was proposed by ( hereafter w1998 ) .
main point of his idea is to incorporate the halo response to the orbiting motion of a satellite into the source of the tidal forces . as the satellite moves , rotating density waves which have resonant frequencies to the satellite rotation are excited . in a nearly isothermal halo
the most dominant wake appears at the position of 2:1 resonance , which is at about half distance to the satellite orbit .
and the mass of the wake is about the same as the satellite , thus its tidal force is much larger than those from the satellite .
w1998 showed with linear analysis for infinitely thin disks , that the tidal forces from the large magellanic cloud ( lmc ) and its wake are large enough to excite the observable warp . in order to prove this idea , fully self - consistent numerical simulations of three - dimensional disks are necessary
this is , in fact , very difficult and expensive , because we have to treat the small vertical motions of a very thin disk embedded in a very massive halo , which is extended well beyond the lmc orbit .
w1998 estimated that 1 million particles are necessary to reduce numerical noise in order to distinguish meaningful dynamics of the disk .
this requirement is here accomplished by using a hybrid algorithm , tree - scf @xcite , which particularly suits our problem of warping disk dynamics .
furthermore , we make the models as close to the real milky way lmc system as possible , so that we can make precise evaluation of the warp amplitudes . in section [ sec : galaxy_models ] we first describe by a brief summary the method to construct the galaxy models using , then give some parameters of the models which we use in this study .
the precise input parameters in the codes are listed in the appendix
. section [ sec : numerical_methods ] explains algorithms and parameters of the tree - scf code .
the results of the numerical simulations are given in section [ sec : simulations ] .
section [ sec : conclusion ] is devoted to conclusion and discussions .
the aim of this study is to examine the influence of the lmc on excitation of the galactic warp . therefore we must first construct equilibrium galaxy models that have no warp , and then put a satellite in the lmc orbit . to construct equilibrium models for a multi - component system like
the milky way is actually a hard task .
many galactic models are constructed with the assumptions that the halo and bulge are static backgrounds , or assuming that velocity distribution is gaussian and solving only the jeans equations .
the deviation of these approximations from a true equilibrium causes initial transient behavior , in other words relaxation into the true equilibrium . that behavior changes the initial distribution especially in the disk . in the disk warping problem
it is necessary to avoid any disturbance on the disk other than that from the satellite . a suitable initial model is given by ( hereafter kd1995 ) , which is nearly an exact solution of the collisionless boltzmann and poisson s equations . in their method , the configurations of a halo , bulge , and disk are given by means of distribution functions which depend only on integrals of motion . by the jeans theorem such distribution functions are already steady state solutions of the collisionless boltzmann equation .
densities can be calculated from the distribution functions , which depend on the positions only through the potential function .
therefore the main task is to determine the functional form of the potential by solving poisson s equation numerically .
the halo distribution function takes the form @xmath10
\left [ \exp ( -e/\sigma_0 ^ 2 ) -1\right ] \quad & \mathrm{if } e<0 , \\ 0 & \mathrm{otherwise } , \\
\end{array } \right.\ ] ] where @xmath11 is the specific angular momentum about the axis of symmetry , and @xmath12 is the relative energy , which is defined so that @xmath13 at the edge of the distribution ( where @xmath14)@xcite .
this distribution function has five free parameters : the potential at the center @xmath15 , which appears implicitly through the definition of the relative energy , the velocity scale @xmath16 , and three factors @xmath17 , @xmath18 , and @xmath19 .
this distribution is based on evans s model , which has axisymmetric logarithmic potential , but is truncated at a finite radius by lowering the distribution function in the same manner as the lowered isothermal models .
the factors @xmath17 and @xmath18 control the system flattening ( @xmath20 ) and the core radius ( @xmath21 ) , respectively , and all the three factors are scaled by the density scale ( @xmath22 ) .
the bulge distribution function is the same as a king model @xcite and has the form @xmath23 & \{\exp[-(e-\psi_\mathrm{c})/\sigma_\mathrm{b}^2]-1\ } \\ &
\mbox{if } e<\psi_\mathrm{c } , \\ 0 & \mbox{otherwise}. \end{array } \right.\ ] ] this distribution has 3 parameters .
@xmath24 is the cutoff potential , @xmath25 the central bulge density , and @xmath26 the bulge velocity dispersion .
the disk distribution must depend on 3 integrals of motion , in order to keep the triaxial velocity ellipsoid in the disk @xcite .
the third integral is , however , not known as any analytical function , so that distribution function can not be written as an analytical function . for the sake of convenience ,
an approximated distribution function is employed which depends on the energy in the vertical oscillations , @xmath29 , and that in planer motion , @xmath30 , as well as @xmath11 .
@xmath31 is quite well conserved for stars in nearly circular orbits .
the form is where @xmath33 and @xmath34 are the radius and energy of a circular orbit with angular momentum @xmath11 , and @xmath35 and @xmath36 are the circular and epicyclic frequencies at radius @xmath33 .
the ` tilde ' functions @xmath37 , @xmath38 , and @xmath39 are free functions . therefore the density and the radial velocity dispersion are conveniently selected as @xmath40,\ ] ] and @xmath41 @xmath37 and @xmath39 are iteratively adjusted so that the densities on the mid - plane and at height @xmath42 will agree with those given by eq .
( [ eq : disk_density ] ) .
the distribution of the disk has 6 free parameters : the disk mass , @xmath43 , the radial scale length @xmath44 , the vertical scale height @xmath45 , the disk truncation radius @xmath46 , the truncation width @xmath47 , and the central velocity dispersion of the disk @xmath48 .
also give a numerical program to calculate the distribution functions and densities .
their algorithm consists of the following 4 steps .
( i ) a test potential function @xmath49 is given .
( ii ) the potential is substituted into the density functions which are calculated from the distribution functions .
( iii ) poisson s equation is solved by using s method to determine a new potential functions .
( iv ) the procedure 2 and 3 are repeated until the potential functions have converged .
* solar radius : @xmath50 = 8 kpc * circular velocity of the disk at the solar radius : @xmath51 km / s * total surface density within 1.1 kpc of the disk plane : + @xmath52 @xcite * contribution of the disk material to @xmath53 : + @xmath54 @xcite * total galaxy mass within 50 kpc : + @xmath55 @xcite * total galaxy mass within 170 kpc : + @xmath56 @xcite * the halo is spherical : @xmath57 , i.e. , @xmath58 @xcite * disk mass is @xmath59 , * disk scale length is 3.2 3.5 kpc , * disk scale height is 200 250 pc , * disk edge is 25 28 kpc , * bulge mass is 15% of that of the disk , * and bulge size is about 2 kpc . | multi - scale interaction between the lmc , the galactic halo , and the disk is examined with n - body simulations , and precise amplitudes of the galactic warp excitation are obtained .
the galactic models are constructed most realistically to satisfy available observational constraints on the local circular velocity , the mass , surface density and thickness of the disk , the mass and size of the bulge , the local density of the halo matter at the solar radius , and the mass and orbit of the lmc .
the mass of the halo within @xmath0 kpc is set to about @xmath1 .
since the observational estimate of the mass distributed in outer region has large ambiguity , two extreme cases are examined ; @xmath2 and @xmath3 .
lmc is orbiting in a ellipse with apocentric radii of 100 kpc , thus the main difference between our two models is the mass density in the satellite orbiting region , so that our study can clarify the role of the halo on excitation of the warp . by using hybrid algorithm ( scf - tree )
i have succeeded to follow the evolution with millions of particles .
the orbiting satellite excites density enhancement as a wake , and the wake exerts a tidal force on the disk . because of the additional torque from the wakes in the halo , the amplitudes of the induced warps are much larger than the classical estimate by , who considered only the direct torque from the lmc .
the obtained amplitudes of @xmath4 , 1 , 2 warps in the larger halo model show very good agreement with the observed amplitude in the milky way .
this result revives the lmc as a possible candidate of the origin of the galactic warp .
our smaller halo model , however , yield only weak warps in all the harmonic modes .
therefore the halo still has significant influence on excitation of warp even in the interaction scenario for excitation of warps .
galaxy : disk , galaxy : kinematics and dynamics , local group , methods : n - body simulations , galaxy : structure , galaxies : spiral 98.35.hj , 98.10.+z , 98.56.-p , 95.75.-z |
elemental lithium , with superconducting t@xmath5 15 k around@xcite 30 gpa and up to 20 k in strained samples@xcite at 50 gpa , joins mgb@xmath6 ( t@xmath0=40 k ) , metallic y ( t@xmath0=20 k)@xcite , and pucoga@xmath7 ( t@xmath0=19 k)@xcite as the big surprises in critical temperature in this century . transformation from a simple nearly - free - electron metal not superconducting@xcite above 100 @xmath8k to a strongly coupled electron - phonon superconductor at 20 - 50 gpa has been explained , at least semiquantitatively , in recent work by profeta _
et al._@xcite and by kasinathan _
et al._@xcite .
the evolution of t@xmath9 has been linked to the increase in electron - phonon coupling of specific phonon branches along intersections of kohn anomaly surfaces with pressure .
the coupling strength is quite strongly dependent on the phonon wave vector @xmath10 and phonon polarization . in particular
, the @xmath11 polarized phonon along the ( 110 ) @xmath12-k direction softens with increasing pressure and finally becomes unstable .
the nesting function @xmath13 which describes the phase space for the electron - hole scattering processes across the fermi surface has fine structure@xcite requiring a very dense mesh to obtain accurate values where its amplitude is large .
since the electron - phonon ( ep ) coupling involves these same scattering events weighted by ep matrix elements that normally vary smoothly with q , the fine structure in @xmath13 implies corresponding fine structure in ep coupling ( as contained for example in the ep contribution to the phonon linewidths @xmath14 ) .
such fine structure makes the sampling of the brillouin zone an issue if accurate numerical results for @xmath3 are desired .
our previous linear response phonon calculations @xcite were performed on 12@xmath2 mesh ( 72 @xmath10 points in the irreducible zone ) . in the present work ,
we focus on the 20 gpa case ( fcc lattice constant a@xmath15 = 6.80 bohr ) where strong phonon softening arises but structural instability has not yet occurred .
the phonon energies and the eliashberg spectral function constants have been presently calculated on a much finer 24@xmath2 mesh ( 413 @xmath10 points in the irreducible zone ) .
careful examination of the results lead to a revised picture of what coupling is important for t@xmath0 ( as opposed to what is important in increasing @xmath1 , which is a far simpler question ) .
the electronic wave functions are calculated using a full potential linear muffin tin orbital ( fp lmto ) method as implemented by savrasov @xcite .
we have used the vosko - wilk - nusair local exchange - correlation @xcite .
the phonon energies and the coupling constants have been calculated using linear response theory @xcite .
phonons were calculated for a dense 24@xmath2 mesh corresponding to 413 inequivalent @xmath10 points in the irreducible zone .
the electronic bz zone integration was performed with a 40@xmath2 k point grid .
the calculated phonon dispersion curves are displayed in fig . [ phonons ] .
the only noticeable softening is seen midway along the @xmath12-l direction ( transverse @xmath16 branches ) , for the @xmath11 polarized @xmath4 branch along the @xmath12-k direction emphasized previously,@xcite and possibly softening in the @xmath16 branches around the x point .
we had identified previously that the @xmath10 = @xmath17 @xmath4 phonon ( near k ) causes large band shifts with atomic displacement ( @xmath18 5 ev / ) near the fs necks , reflecting strong ep coupling .
the corresponding ` mode @xmath1 ' values @xmath19 are presented in fig .
[ modelambda ] .
the @xmath19 values for the @xmath4 branch near k , and also near the midpoint of the @xmath12-l line are larger compared to other areas in the bz , as expected from the phonon kinks . in addition , we notice that the @xmath19 values for all the three branches are large all along @xmath12-x ; this coupling is not immediately obvious from the phonon dispersion in fig .
[ phonons ] because it depends only weakly on q. + in fig .
[ linewidth ] , we display the corresponding linewidths @xmath14 , which are given by@xcite @xmath20}|^2 \delta(\varepsilon_k)\delta(\varepsilon_{k+q}).\end{aligned}\ ] ] while @xmath14 includes the same q - specific fermi surface average of the ep matrix elements with respect to the available phase space for scattering through wavevector q as does @xmath21 , it is proportional to @xmath22 rather than inversely proportional . as a result
, @xmath14 is a much better indicator of the importance of the coupling for t@xmath0 . coupling at high frequency is much more important than at low frequency for t@xmath0 , whereas the opposite is true for @xmath1 .
thus it is not surprising that the linewidth ` dispersion ' in fig .
[ linewidth ] provides a different viewpoint than does the @xmath21 dispersion .
there is still the large contribution from the region around k where there is an incipient instability , but it is very limited in q and very sharply structured , not being resolved even by the q mesh we have used .
the really impressive region however is the @xmath23 branch along ( 100 ) directions , which dominates the landscape ( at least along symmetry lines ) .
there are also important contributions from the @xmath23 branch along the @xmath12-k lines and from the @xmath16 branches along ( 100 ) .
even the small peak in the @xmath16 branches midway between @xmath12 and l ( where the kink in the @xmath16 branch occurs ) is small compared to the contribution from the @xmath23 branch along the same line .
present @xmath24 at v / v@xmath15 = 0.45 gpa ( @xmath25 50 gpa ) which have some features @xcite in common with fig .
[ linewidth ] , but can not be compared in detail because of the different volumes .
the phonon frequency distribution @xmath26 is shown in fig .
[ phonondoscomp ] where it is clear that the coarser q mesh had given a good picture of the density of states .
this insensitivity does not persist when coupling is included .
the eliashberg spectral function @xmath27 from both meshes is plotted in fig .
[ spectralfncomp ] . for this 20 gpa case
a large broad peak dominates the 15 - 35 mev range , with less intense coupling extending on up to 70 mev .
compared to its value from the coarser mesh , it is much smaller in the 7 - 20 mev range .
the intensity in the low frequency region is clearly from the transverse branch phonons , while the less impressive ( but still substantial ) intensity in the high frequency region is from the longitudinal phonon branch .
the frequency resolved coupling strength @xmath28 , also shown in fig .
[ spectralfncomp ] is dramatically changed by using the finer q mesh .
there is now a broad peak around 12 mev , still indicative of comparatively stronger coupling in the low frequency regime .
however , the overriding peak in the 7 - 16 mev range has disappeared , showing that is was an artifact of the sampling of the volume near the point k using only a 12@xmath2 mesh .
this feature is reminiscent of results calculated by yin _
et al . _ , @xcite who have found an even narrower peak in @xmath29 growing at lower frequencies ( 2 - 5 mev ) in elemental y under pressure .
this peak also was traced to incipient instability for a transverse branch near the zone boundary .
we now concentrate on the difference in t@xmath0 , and the interpretation of the distribution of the coupling , that we obtain from our denser q mesh of data ( we note that the quantities we discuss are still not fully converged ) .
the results of this much better zone sampling are displayed in table [ table ] .
startlingly , the value of @xmath1 has dropped from 3.1 to 1.0 , ( which was also obtained by profeta _ et al . _
@xcite ) reflecting the fact that the low energy region of @xmath3 was ( and still is ) reliant on large contributions ( from near k and perhaps elsewhere ) that vary sharply with q. for the determination of t@xmath0 , the strong reduction in @xmath1 is considerably compensated by increases in the frequency moments .
the logarithmic moment @xmath30 _ increases _ by a factor of 2.7 , and the second moment by 40% . at a similar volume , shi and papaconstantopoulos @xcite
obtain @xmath1 = 0.5 from rigid muffin tin calculations .
one source of difference is the value @xmath31 325 k ( scaled from the p = 0 debye frequency which is our calculated value of @xmath32 .
.from the calculated @xmath33 at 20 gpa for both @xmath10 meshes , this table provides : the logarithmic , first , and second moments of the frequency ; the value of @xmath1 ; t@xmath0 ( k ) calculated using @xmath34=0.13 and @xmath34=0.20 ; and two simple measures of @xmath3 ( see text ) . [ cols="^,^,^,^,^,^,^,^,^ " , ] [ table ] we have obtained t@xmath9 using the allen - dynes equation@xcite @xmath35 where @xmath36 are strong coupling corrections that depend on @xmath1 , @xmath34 , and the ratio @xmath37 , which is a measure of the shape of @xmath3 .
these @xmath38 factors are important for @xmath1=3.1 but have little effect for @xmath391 or less . for the commonly used value of @xmath34=0.13 ,
our more nearly converged results give t@xmath0=13 k vs. the earlier estimate of 20 k , only a 35% drop even though @xmath1 decreases by more than a factor of three . for @xmath34=0.20 , t@xmath0 = 3.4 k. these values should be compared with the nearly hydrostatic value t@xmath0 = 6 k at 20 gpa found by deemyad and schilling.@xcite
this apparent weak dependence of t@xmath0 on @xmath1 reflects the observation , made several times in the past,@xcite that , for providing an estimate of t@xmath0 with averages of @xmath3 , the combination of @xmath1 and some frequency moment separately is not the best choice .
this observation is connected with the observation that coupling at low frequency , which strongly enhances @xmath1 , has only a weak effect on t@xmath0 .
carbotte and collaborators@xcite argued that the product @xmath40 ( with @xmath41 being the first moment as defined by allen and dynes@xcite ) is nearly proportional to t@xmath0 for 1.2@xmath42 2.4 ; this quantity is proportional to the area under @xmath3 and corresponds to a zone integral of @xmath43 .
they were not including strong coupling corrections to t@xmath0,@xcite however , which were unknown at the time of the original work .
the strong - coupling limit@xcite gives t@xmath44 @xmath45 ; this product @xmath46 corresponds to a zone integral of @xmath14 .
we find that , for the fine mesh compared to the coarse ( see the table ) : @xmath1 decreases by @xmath47% ; t@xmath0 drops by 35% ( at @xmath34=0.13 ) ; the carbotte product decreases by 30% ; the strong - coupling limit factor drops by only 5% .
hence the carbotte ratio follows the change in t@xmath0 best , of these choices .
the behavior of @xmath48 does not faithfully represent the relative importance of modes for t@xmath0 except in the weak - coupling limit . in the strong coupling limit ( which we do _ not _
claim is the case for li ) it is @xmath14 that gives the better picture . for another perspective we display in fig .
[ isogamma ] an isosurface of @xmath14 for the @xmath23 modes . as expected from fig .
[ linewidth ] , it is dominated by contributions near the center of the ( 100 ) line .
the isocontours form a blunt jack - like shape although the true symmetry is only four - fold .
our corresponding plots for @xmath21 are almost like those of profeta _
et al._@xcite except that we seem to find somewhat larger contributions near the zone center . in this paper
we have clarified and supported our earlier suggestion@xcite that denser zone sampling would be necessary to obtain accurate numerical results .
we find in addition that careful sampling of the fine structure in the coupling is required even for obtaining a correct understanding of the microscopic origin of the pressure - induced increase in t@xmath0 .
in particular , we call into question the view ( easily surmised from the earlier work@xcite even if not so claimed ) that the impressive value of t@xmath0 in li under pressure arises either from soft phonons near k , or from a box - like set of regions in momenta along ( @xmath49 ) with @xmath50 . these questions , and the strong polarization dependence of matrix elements , will be addressed in further work .
we acknowledge important advice from s. y. savrasov on the linear response code , and interactions with a. lazicki , r. t. scalettar , and c. s. yoo . we thank u. nitzsche for assistance in the use of the ifw cluster to perform the calculations .
this work was supported by national science foundation grant nos .
dmr-0421810 and dmr-0312261 .
j.k . was supported by doe grant fg02 - 04er46111 , and h.r .
was supported by dfg ( emmy - noether - program ) .
g. profeta _
et al . _ ,
lett . * 96 * , 047000 ( 2006 ) .
d. kasinathan _ et al .
lett . * 96 * , 047004 ( 2006 ) .
p. b. allen , phys .
b * 6 * , 2577 ( 1972 ) ; p. b. allen and m. l. cohen , phys . rev
* 29 * , 1593 ( 1972 ) .
s. y. savrasov , phys .
b * 54 * , 16470 ( 1996 ) ; s. y. savrasov and d. y. savrasov , _ ibid .
_ * 54 * , 16487 ( 1996 ) .
s. h. vosko , k. wilk and n. nusair , can .
* 58 * , 1200 ( 1980 ) .
j. s. tse , y. ma , and h. m. tutuncu , j. phys . : condens .
* 17 * , s911 ( 2005 ) .
l. shi and d. a. papaconstantopoulos , phys .
b * 73 * , 184516 ( 2006 ) .
z. p. yin , s. y. savrasov , and w. e. pickett , private communication .
p. b. allen and r. c. dynes , phys .
b * 12 * , 905 ( 1975 ) .
see j. p. carbotte , rev .
phys . * 62 * , 1027 ( 1990 ) and references therein . | in an attempt to provide a clearer understanding of the impressive increase in t@xmath0 under pressure in elemental li , linear response calculation of the phonon dispersion curves , electron - phonon matrix elements , phonon linewidths and mode @xmath1 s have been carried out on a finer mesh ( 24@xmath2 in the brillouin zone ) than done previously , for the volume corresponding to 20 gpa pressure .
the result illustrates the great need for a fine mesh ( even finer than this ) for converged results of @xmath1 and the spectral function @xmath3 .
although the initial pressure - induced transverse @xmath4 phonon instability ( in harmonic approximation ) near the symmetry point k has dominated attention , the current results show that the high value of t@xmath0 gets strong contributions from elsewhere in the zone , particularly from the longitudinal mode along ( 100 ) .
superconductivity , lithium , high pressure , strong coupling . ... |
_ photons are light _ , namely they are massless and , consequently , do not possess a rest frame . therefore , although they are quanta of vector fields of spin @xmath0 , photons have only two independent spin projections along the direction of propagation ( _ helicity _ ) @xcite . in the classical optics realm
this means that at a fixed point in space , there are only two independent directions of oscillation ( _ polarization _ ) for the electric field carried by monochromatic light @xcite . for plane and paraxial waves propagating along a given axis , which here and hereafter will be identified with the @xmath1-axis of a cartesian reference frame ,
such oscillations are perpendicular to the direction of propagation @xmath1 .
if the two components of the electric field have equal amplitude and are @xmath2 out of phase , the light is said to be circularly polarized .
if the amplitudes are not equal , the light is elliptically polarized @xcite . however , monochromatic electromagnetic waves confined by cavities and waveguides @xcite , strongly focused beams @xcite , plasmonic fields and evanescent waves @xcite , exhibit complex spatial structures and the two independent directions of oscillation of the electric field are not necessarily perpendicular to the propagation direction @xmath1 . in these cases
there is a longitudinal electric field component oscillating along @xmath1 and the light can be elliptically polarized in a meridional plane , namely in a plane containing the @xmath1-axis @xcite .
therefore , the electric field vector can spin around an axis perpendicular to @xmath1 , as a `` photonic wheel '' @xcite , thus generating a transverse spin angular momentum ( am ) density @xcite .
light beams with a longitudinal electric field component are key to many relevant applications in physics and chemistry , such as particle acceleration @xcite , single - molecules manipulation and spectroscopy @xcite , high resolution near - field optical microscopy @xcite , strong focussing of light @xcite and nanophotonics @xcite . as recently reviewed by several authors @xcite , transverse spin am density manifests in many different forms depending on the specific wavefield configuration @xcite . here
, we establish a general theory of monochromatic light waves carrying transverse spin am density , which permits us to describe different phenomena in a consistent fashion .
we begin with by considering generic optical waves with longitudinal components of the electric field vector . from the study of the dynamical properties of these waves , such as energy , linear momentum and spin am densities
, we find that the last may be transverse in an either global or local sense ; the first or second case occurring when the spin density vector is perpendicular to either the mean direction of the linear momentum of the wave or to the linear momentum density . in both cases ,
we quantify transverse spin am density by means of suitably defined meridional stokes parameters @xcite .
an intriguing and previously unnoticed connection between the theory of complex functions ( see , e.g. , @xcite ) and the transverse spin am density of light is revealed by studying the _ maximal _ transverse spin am density occurring along the so - called c lines of the electric field , where the polarization of the electric field is purely circular @xcite .
finally , with the purpose of illustrating our findings , we work out two detailed examples involving non - diffracting bessel beams and evanescent waves .
in order to set the notation and the units used throughout this paper , we now briefly review some notable results from the works of nye & hajnal @xcite , nye @xcite and berry & dennis @xcite . the observable electric field @xmath3 of a monochromatic electromagnetic wave of angular frequency @xmath4 , can be written as the real part of the time - harmonic complex vector field @xmath5 , where @xmath6 and @xmath7 here , @xmath8 and @xmath9 are the conjugate radii of the so - called _ polarization ellipse _ centered at @xmath10 , swept out by @xmath3 as time elapses .
the vector @xmath11 normal to the polarization ellipse is parallel to the electric part of the cycle - averaged spin am density @xmath12 , that is @xmath13 where , from now on , a prefactor @xmath14 in front of energy , linear and angular momentum densities , will be omitted . in a homogeneous plane wave with real - valued wave vector @xmath15 ,
the propagation direction is naturally defined by @xmath15 itself .
such definition is _ global _ , in the sense that the wave vector @xmath15 is independent of the position vector @xmath10 .
however , for any wave that is not simply plane , one needs to specify a _
local _ , namely dependent on @xmath10 , propagation direction .
let us denote with @xmath16 the local wave vector giving the propagation direction of the electric field @xmath17 at @xmath10 .
it is defined as @xmath18 \nonumber \\ = & \ ; { \mathbf{p}}\cdot \left ( { \bm{\nabla}}\right ) { \mathbf{q}}- { \mathbf{q}}\cdot \left ( { \bm{\nabla}}\right ) { \mathbf{p}}\equiv { \mathbf{p}}\cdot \bigl ( { \overset{\text{\small$\bm\leftrightarrow$}}{{\bm{\nabla } } } } \bigr ) { \mathbf{q } } , \ ] ] where the suggestive notation @xmath19 has been used @xcite .
there is a simple relation between the local wave vector @xmath16 and the cycle - averaged linear momentum of the electric field @xmath20 .
this can be written as the sum of an orbital part , denoted @xmath21 , and a spin part , indicated with @xmath22 , that is @xmath23 .
it turns out that @xmath16 coincides ( apart from the prefactor @xmath14 ) with @xmath21 , while the spin part can be simply defined in terms of the spin am density @xmath12 as : @xmath24 another important dynamical property of the electric field @xmath17 , is its cycle - averaged energy density that can be written in terms of @xmath25 as @xmath26 from this equation and eq .
it follows that , given a fixed value of @xmath27 , @xmath28 is _ maximal _ when @xmath29 and @xmath30 . in this case @xmath31 and the wave is circularly polarized . in many cases of practical importance as , for example , in laser beams , light propagates along a well defined global direction , here and hereafter identified with the @xmath1-axis of a cartesian reference frame @xmath32 .
in such a frame , the _
mean _ value of a dynamical quantity like energy or linear momentum , is defined as the integral of the corresponding density over the cross - section of the beam .
for example , the mean wave vector is given by @xmath33 and , because of our choice of the reference frame , it is parallel to the @xmath1-axis .
a special situation occurs when @xmath34 . in this case
the local and global propagation directions coincide , although the wave is not plane .
we will see explicitly later in what circumstances this happens . in general , for optical waves that are not simply plane , the dynamical properties and polarization patterns of the electric and magnetic fields differ @xcite . in the remainder
, we will study explicitly the transverse spin am density of the electric field @xmath35 only .
however , all our conclusions keep their validity if we replace @xmath35 with the magnetic field @xmath36 .
this is possible because for monochromatic electromagnetic waves the electric and magnetic fields are `` uncoupled '' , in the sense that the electric and magnetic parts of the electrodynamic hamiltonian are separately time - independent . in vacuum , in absence of free charges and electric currents , the electric vector field @xmath37 is purely transverse , namely @xmath38 , and satisfies the wave equation @xmath39 .
a plane - wave solution of these equations takes the form @xmath40 , where @xmath41 is the wave vector . here
@xmath42 is the wave number and @xmath43 denotes an either real- or complex - valued unit vector such that @xmath44 , where the `` dot '' symbol between any two @xmath45d vectors @xmath46 and @xmath47 , stands for @xmath48 ( no complex - conjugation required ) . when @xmath49 is complex - valued , as in the evanescent waves considered in sec .
[ evanescentfield ] , it can be written as @xmath50 , where @xmath51 and @xmath52 denote the real and imaginary parts of the unit wave vector , respectively .
in this case the orthogonality condition @xmath44 entails @xmath53 and @xmath54 @xcite .
moreover , @xmath55 is the angular frequency of the wave and @xmath56 is a three - vector perpendicular to @xmath15 , namely @xmath57 . in our preferred cartesian reference frame
@xmath58 , the condition @xmath57 can be explicitly written as @xmath59 apart from the trivial solution @xmath60 , eq . also admits three elementary solutions of the form @xmath61 , @xmath62 and @xmath63 , where one of the three cartesian components of @xmath64 is chosen to be zero .
the first solution @xmath65 generates a plane - wave mode whose electric field is purely perpendicular to the @xmath1-axis @xcite .
the remaining two solutions @xmath66 generate plane waves with both longitudinal ( @xmath1 ) and transverse ( either @xmath67 or @xmath68 , respectively ) components of the electric field and are , therefore , interesting to us .
consider , for example , the solution @xmath69 with @xmath70 . in this case
the electric vector field can be evidently written as @xmath71 as desired , @xmath72 has transverse and longitudinal components of amplitudes @xmath73 and @xmath74 , respectively , but it does not possess yet the structure of a photonic wheel because from eq . it follows that the observable electric field @xmath75 simply oscillates along the direction @xmath76 and does not rotate .
moreover , @xmath3 propagates in the direction of @xmath15 which , in general , is not parallel to the @xmath1-axis . in fact , @xmath77 to remove this problem , we notice that propagation along the @xmath1-axis can be simply achieved by taking the sum of two plane waves @xcite that are _ mirror - symmetric _ with respect to the @xmath1-axis , namely @xmath78 \nonumber \\ & \ ; \times \exp(-i \omega t)\nonumber \\ = & \ ; \left ( \zeta \ , \hat{\mathbf{y } } - i \eta \ , \hat{\mathbf{z}}\right ) \exp(i k_z z - i \omega t ) , \ ] ] where @xmath79 and @xmath80 , @xmath81 . as byproduct of this choice , we automatically get a @xmath2 phase difference between the transverse and the longitudinal components of @xmath72 , which ensures the rotation of the field in the meridional plane @xmath82
. moreover , the exponential factor @xmath83 shows that @xmath72 effectively propagates along the @xmath1-axis .
therefore , both conditions required for the existence of a photonic wheel field configuration are satisfied .
a straightforward calculation reveals that , as expected , the spin am density carried by @xmath84 is transverse , @xmath85 and that the local wave vector is purely longitudinal , @xmath86 as desired .
a similar reasoning can be repeated for the field obtained by replacing @xmath87 with @xmath88 in eq . .
in this case we obtain @xmath89 where @xmath90 .
it is worth noticing that @xmath91 if @xmath92 ( the two plane waves lay on the @xmath93-plane ) and @xmath94 if @xmath95 ( the two plane waves lay on the @xmath82-plane ) . this is a direct consequence of the transverse nature of electromagnetic fields .
however , _
arbitrary _ transverse orientation of @xmath12 may be achieved by superimposing , for a given wave vector @xmath15 , the fields @xmath84 and @xmath96 , to obtain @xmath97 with @xmath98 .
it is not hard to show that in this case we have @xmath99 { \hat{\mathbf{z}}}. \end{split } \ ] ] from eq .
it follows that the polarization ellipse centered at @xmath10 and swept out by @xmath100 $ ] , is located on the @xmath82-plane .
this means that we can fully characterize such an ellipse by introducing the position - dependent _ meridional stokes parameters _ @xmath101 , defined in terms of the vectors @xmath102 , as @xcite , @xmath103 specifically , for @xmath84 we find @xmath104 from these expression and eqs . , we obtain after a straightforward calculation , @xmath105 at points @xmath10 where @xmath106 and @xmath107 , the wave is circularly polarized and the spin am density is _ maximal_.
this occurs when @xmath108 , that is when @xmath109 , with @xmath110 , irrespective of the values of @xmath1 . as we will see later
, the form for the stokes parameters is generic for propagation invariant light beams carrying transverse spin am density .
it should be noted that the expression in the last row of eq . reveals , when compared with eq .
, that @xmath111_x$ ] . in all elementary examples considered above ,
the spin am density and the local wave vector are perpendicular everywhere , irrespective of the position @xmath10 , @xmath112 therefore , we say that the spin am density is transverse in a _
local _ sense , which is a strong condition .
later , when we will study fields more general than @xmath113 , we will find waves with spin am density transverse in a _
global _ sense , namely fulfilling the weaker requirement @xmath114 where we have used the fact that according to our choice of the reference frame , @xmath115 is directed along @xmath116 .
in addition to the few basic properties illustrated in this section , the two plane - wave fields @xmath113 possess many other interesting characteristics , thoroughly investigated in @xcite . however , these are only particular solutions of maxwell s equations exhibiting transverse spin am density .
conversely , we are going now to consider general electromagnetic waves carrying transverse spin am density .
let us begin the study of perfectly general fields exhibiting photonic wheel structures , by noticing that eq . can be rewritten as @xcite @xmath117 where @xmath118 , etc . , and we have defined @xmath119 \nonumber \\ = & \ ; \cos \left ( { \bm{\kappa}}\cdot { \bm{\rho}}\right ) \exp(i z k_z ) , \ ] ] with @xmath120 and @xmath121 denoting the transverse parts of the wave and position vectors , respectively . instead of the particular two plane - wave field
@xmath122 , let us consider now an arbitrary solution @xmath123 of the monochromatic helmholtz equation , @xmath124 , and use it to build the more general electric vector field @xmath125 the @xmath25 vectors of this field can be calculated from eq . .
the result is @xmath126 where @xmath127 and @xmath128 denote , respectively , the real and the imaginary part of @xmath129 .
it is not hard to show from eq .
that the spin am density is either automatically transverse and parallel to the @xmath68-axis or null , depending on the shape of the functions @xmath130 and @xmath131 : @xmath132 a lengthy but straightforward calculation shows that the local wave vector @xmath133 possesses nonzero @xmath68- , @xmath67- and @xmath1-components ( we do not report their expressions here because they are cumbersome and not very illuminating ) and , therefore , is not longitudinal .
this means that , as anticipated , @xmath134 is not transverse in a local sense .
however , it can be still transverse in a global sense providing that @xmath135 is parallel to @xmath136 . in order to determine what conditions the generic field @xmath123 must satisfy in order to yield @xmath135 parallel to @xmath136 , we first rewrite it in the standard _ homogeneous angular spectrum _ representation , @xmath137 where @xmath138 , and @xmath139
is the @xmath140d fourier transform of @xmath141 , with @xmath142 for @xmath143 @xcite .
then , substituting eq . into eq . and using eqs . and
, we obtain @xmath144 { \mathrm{d}}k_x { \mathrm{d}}k_y . \ ] ] this equation shows that whenever the modulus square of the angular spectrum is mirror - symmetric about the @xmath1-axis , that is when @xmath145 , the mean wave vector becomes parallel to the @xmath1-axis and equal to @xmath146 where @xmath147 .
this is the main result of this section .
we make three remarks .
a. the condition @xmath145 is just the ( weaker ) version of the mirror - symmetry property @xmath148 exhibited by the two plane - wave field .
thus , the same geometric condition induces both local and global transverse spin am density .
b. from eqs . and
@xmath149 , it follows that @xmath150 where @xmath151 , @xmath152 etc .
if @xmath153 were an analytic ( or , holomorphic ) function of the complex variable @xmath154 , then its real and imaginary parts @xmath130 and @xmath131 , respectively , would satisfy the so - called cauchy - riemann conditions @xcite , @xmath155 substituting these expressions into eqs .
we obtain , after a little of algebra , @xmath156 @xmath157 means that the light is circularly polarized and the transverse spin am density is maximal .
a similar results for ordinary waves with _ longitudinal _ spin am density , was obtained by leckner @xcite . + if @xmath153 is not an analytic function ( as it is usually the case ) , the conditions can still be satisfied _ locally _ , namely at some isolated points @xmath158 etc . in this case eqs .
become the equations determining the transverse c lines of the electric field @xcite , along which the light is circularly polarized in the @xmath82-plane and the spin am density achieves its maximum value . in these points ,
the tangent to the c lines is perpendicular to the @xmath1-axis . in other words ,
transverse c lines lay on the @xmath159-plane by definition .
c. just as we did for the elementary two plane - wave fields , also in the general case we can consider both @xmath160 and @xmath161 fields and their linear combinations of the form @xmath162 the derivation of the formulas for the mean wave vector and the transverse spin am density is straightforward and , therefore , will not be reported here .
in the previous section we considered both particular and general scalar fields @xmath122 and @xmath123 , respectively , as the building blocks of electric vector fields carrying transverse spin am density . we have found that the spin density of the electric field generated by @xmath122 is transverse in a local sense , that is perpendicular to the linear momentum density ( or , local wave vector ) @xmath16 .
conversely , the spin density from @xmath123 is transverse in a global sense , namely perpendicular to the mean linear momentum density @xmath115 .
however , the existence of locally transverse spin density is not restricted to elementary plane - wave fields only .
in fact , suitably prepared propagation invariant beams as bessel and mathieu beams , ( see , e.g. , @xcite and references therein ) , also possess a locally transverse spin density . to show this ,
consider the two monochromatic scalar fields @xmath163 where @xmath164 and @xmath165 are independent solutions of the transverse helmholtz equation @xmath166 , with @xmath167 and @xmath168 .
these fields are dubbed propagation invariant ( or , non - diffracting ) because the `` intensity '' @xmath169 does not depend on @xmath1 .
the electric vector field built from eq . with the replacements @xmath170 and @xmath171 , is @xmath172\nonumber \\ & \times \exp(i k_z z -i \omega t ) . \ ] ] this field propagates along the @xmath1-axis and has a longitudinal component out of phase with respect to the transverse ones ; therefore , photonic wheel structures are possible .
the spin am density can be calculated from eq . after determining @xmath25 .
the result is @xmath173 the local wave vector is evaluated with the help of eq . and is equal to @xmath174 { \hat{\mathbf{z}}}\nonumber \\ = & \ ; k_z { \left\lvert{\mathbf{e}}({\mathbf{r}})\right\rvert}^2 { \hat{\mathbf{z } } } , \ ] ] with @xmath175 .
as anticipated in the previous section , in this case @xmath16 is factorable in the product between a position - dependent scalar function and a constant vector , namely @xmath176 , with @xmath177 .
since the generic functions @xmath178 and @xmath179 depend on the transverse coordinates @xmath180 only , we conclude that the densities @xmath181 and @xmath16 are orthogonal at all points on the @xmath159-plane , irrespective of the propagation distance @xmath1 .
a convenient illustration of the spin and linear momentum densities and is furnished by the fields @xmath182 where @xmath183 , @xmath184 , and @xmath185 , with @xmath186 , denotes a bessel function of the first kind @xcite .
the electric field is polarized in the meridional plane perpendicular to the azimuthal direction @xmath187 , that is @xmath188 \exp \left ( i k_z z\right ) , \ ] ] where @xmath189 .
the spin am density is purely azimuthal and equal to @xmath190 \hat{\bm{\varphi}}. \ ] ] this expression has the same form of @xmath191 in eq . with @xmath192 and @xmath193 .
the main difference between the two densities is that @xmath191 has a cartesian symmetry while @xmath194 manifests cylindrical symmetry .
the local wave vector is purely longitudinal and given by eq . .
electric fields polarized in the meridional plane perpendicular to the azimuthal direction , were considered in @xcite in the context of c lines in three dimensions . a paraxial version thereof
was previously studied in @xcite .
in this section we illustrate the theory developed above with the help of two clear cases .
first , we consider a propagating zeroth - order bessel beam with cartesian symmetry . then , we study an evanescent plane wave .
consider the scalar bessel field @xmath195 where @xmath196 and @xmath197 , with @xmath198 denoting the aperture of the bessel cone @xcite .
the electric field vector has the expression given in eq . with @xmath199 transverse spin
am density and local wave vector can be calculated substituting eqs . into eq . and
eq . , respectively . they are directed along the @xmath68- and @xmath1-axis , respectively .
the observable electric vector field is given by @xmath200 figure [ fig1 ] shows the temporal evolution of @xmath3 as a function of the scaled coordinates @xmath201 and @xmath202 at @xmath203 , where @xmath204 and @xmath205 denotes the first zero of @xmath206 . substituting the real and imaginary parts of @xmath123 into eq . , we find the implicit equation of the c lines for the field : @xmath207 this equation can be numerically inverted to find the curves ( c lines ) along which becomes an identity .
these are illustrated in fig .
[ fig2 ] .
consider now the evanescent field of an inhomogeneous plane wave exponentially decaying in the positive @xmath67-direction and propagating along the @xmath1-axis @xcite : @xmath208 where @xmath209 , @xmath210 and the real parameter @xmath211 fixes the scale of inhomogeneity .
for @xmath212 the wave is purely propagating .
exactly as in the previous case , the electric field vector is found by substituting @xmath213 into eq . .
transverse spin am density and local wave vector are given , as usual , by eq .
and eq . , respectively .
for the observable electric vector field we obtain from eq .
\ ] ] this equation clearly shows the photonic wheel features of evanescent waves @xcite . for the sake of completeness , we have calculated the stokes parameters in the meridional @xmath82-plane .
the result is @xmath215 + where @xmath216 .
this means that for @xmath217 the electric field has ( almost ) transverse circular polarization uniformly over the @xmath82-plane .
in conclusion , we have presented a perfectly general theory of light waves carrying an electric field rotating around an axis perpendicular to the propagation direction , thus producing nonzero transverse spin am density . to define longitudinal and transverse components of the spin angular density , we have assumed that a propagation direction can be unambiguously identified either locally or globally . in the first case ,
the direction of light at any point @xmath218 is identified by the local wave vector @xmath219 . in the second case ,
the propagation direction is fixed by the mean wave vector @xmath220 defined as the integral of the local wave vector over the cross - section of the beam .
the novelty of our approach resides in its general character , which provides for a conceptual unifying view of seemingly different polarization wave phenomena .
the success of such unification is made manifest in the several examples reported in this paper .
last but not least , our treatment reveals a somewhat hidden and intriguing connection between light with circular polarization and the theory of complex functions .
pb acknowledges financial support by the alexander von humboldt foundation ( feodor lynen fellowship ) and by the canada excellence research chair ( cerc ) in quantum nonlinear optics .
53ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1098/rspa.1909.0060 [ * * , ( ) ] @noop _ _ , vol .
( , , ) @noop _ _ , ed .
( , , ) @noop _ _ , ed .
( , , ) @noop _ _ , ed . , radiation laboratory series ( , ) @noop _ _ ( , , ) link:\doibase 10.1098/rspa.1959.0200 [ * * , ( ) ] link:\doibase 10.1103/physreva.11.1365 [ * * , ( ) ] @noop _ _ ( , ) \doibase http://dx.doi.org/10.1016/0030-4018(93)90095-m [ * * , ( ) ] link:\doibase 10.1364/josab.24.000624 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.033831 [ * * , ( ) ] link:\doibase 10.1103/physreva.89.033841 [ * * , ( ) ] link:\doibase 10.1103/physreva.90.023842 [ * * , ( ) ] http://www.jeos.org/index.php/jeos_rp/article/view/13032 [ * * ( ) ] link:\doibase 10.1103/physrevlett.103.100401 [ * * , ( ) ] link:\doibase 10.1103/physreva.81.053838 [ * * , ( ) ] link:\doibase 10.1103/physreva.85.061801 [ * * , ( ) ] link:\doibase 10.1103/physreva.41.3727 [ * * , ( ) ] link:\doibase 10.1103/physreva.44.2656 [ * * , ( ) ] link:\doibase 10.1126/science.272.5259.255 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.85.4482 [ * * , ( ) ] @noop * * , ( ) \doibase http://dx.doi.org/10.1016/s0030-4018(99)00729-4 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.91.233901 [ * * , ( ) ] @noop _ _ , ed .
( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevx.5.011039 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.114.063901 [ * * , ( ) ] @noop _ _ , ed .
, no . ( , , ) link:\doibase 10.1098/rspa.1987.0002 [ * * , ( ) ] @noop * * , ( ) @noop @noop * * , ( ) http://stacks.iop.org/1464-4258/11/i=9/a=094001 [ * * , ( ) ] @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop _ _ , edited by , ieee press series on electromagnetic wave theory ( , , ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physreva.71.033411 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1364/oe.18.005809 [ * * , ( ) ] @noop @noop * * , ( ) @noop * * , ( ) | a circularly polarized electromagnetic plane wave carries an electric field that rotates clockwise or counterclockwise around the propagation direction of the wave . according to the handedness of this rotation , its _ longitudinal _
spin angular momentum density is either parallel or antiparallel to the propagation of light .
however , there are also light waves that are not simply plane and carry an electric field that rotates around an axis perpendicular to the propagation direction , thus yielding _ transverse _ spin angular momentum density . electric field configurations of this kind have been suggestively dubbed `` photonic wheels '' .
it has been recently shown that photonic wheels are commonplace in optics as they occur in electromagnetic fields confined by waveguides , in strongly focused beams , in plasmonic and evanescent waves . in this work
we establish a general theory of electromagnetic waves propagating along a well defined direction , which carry transverse spin angular momentum density .
we show that depending on the shape of these waves , the spin density may be either perpendicular to the _ mean _ linear momentum ( globally transverse spin ) or to the linear momentum _ density _ ( locally transverse spin ) .
we find that the latter case generically occurs only for non - diffracting beams , such as the bessel beams .
moreover , we introduce the concept of _ meridional _ stokes parameters to operationally quantify the transverse spin density .
to illustrate our theory , we apply it to the exemplary cases of bessel beams and evanescent waves .
these results open a new and accessible route to the understanding , generation and manipulation of optical beams with transverse spin angular momentum density . |
for disk accretion to occur angular momentum must be removed from orbiting material .
it has long been thought that magnetic fields could play more than just a passive role in this situation @xcite . the magnetorotational instability ( mri - * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) has since emerged as a robust mechanism for destabilising a differentially rotating disk , placing the importance of magnetic fields on a firm footing .
subsequent numerical work has demonstrated that the non - linear motions induced by the mri lead to self - sustaining turbulence with the resulting stresses effectively transporting angular momentum ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
a large body of the numerical work on magnetorotational turbulence to - date has focused on a local patch of disk with the vertical component of gravity neglected - the unstratified shearing - box .
studies examining various magnetic field configurations have produced a wide range of values for the stress - to - gas - pressure ratio , @xmath3 ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* to mention but a few ) .
values of @xmath4 are , however , only achieved by models initialised with a net vertical magnetic field . by construction ,
the unstratified shearing - box preserves an initial net vertical field .
hence , in these cases power can be effectively injected by vertical mri modes on large spatial scales , feeding the turbulent cascade from the top down .
astrophysical accretion disks are , however , intrinsically global objects ( by which we mean they are not periodic in the radial and/or vertical directions ) .
recent studies by @xcite and @xcite have demonstrated that localised patches of global disks do not preserve their magnetic field configuration in the same manner as an unstratified shearing - box .
moreover , an initial vertical field is free to evolve in global models , and it is not clear whether global disks will achieve the large stresses implied by net - vertical flux unstratified shearing - box models .
what dependence of the turbulent stresses on initial field configuration should one expect for a stratified global disk ?
newtonian global disk models exhibit @xmath5 ( e.g. * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
larger values of @xmath6 are typically associated with the saturation stress level during the initial transient phase of the simulation , with some evidence for stronger stresses when the disk initially possesses a poloidal magnetic field .
there are mixed results from general relativistic global disk simulations .
for instance , @xcite find markedly similar disk properties irrespective of the initial field configuration .
in contrast , @xcite ( see also - * ? ? ? * ) quote @xmath7 with the lowest values for an initially toroidal field , whereas the largest values are for models where the disk initially contains a large quantity of poloidal magnetic flux . in the present study
we investigate the dependence of magnetorotational turbulence in a global disk on the initial field configuration .
the simulations include a mass source term which facilitates simulations with long run times .
thus , the late - time quasi - steady disk evolution can be studied without concerns about contamination from mass exhaustion .
this differs from previous global disk models , with the exception of @xcite and @xcite .
a key result is the ignorance of the quasi - steady state turbulent stress , and thus efficiency of angular momentum transport , to the initial field configuration ; given sufficient time the disk expels and/or digests the initial field , evolving to a statistically almost identical state .
the simulations produce @xmath8 during the quasi - steady state , which falls short of observationally inferred values of @xmath9 ( see * ? ? ?
* and references therein ) .
we perform supplementary simulations to address this quandary , finding that @xmath10 can be achieved by trapping field , and @xmath11 results when mass or vertical magnetic field are injected .
the former case is relevant for magnetically arrested @xcite and magnetically levitating @xcite disk models .
the remainder of this paper is organised as follows :
the simulation setup and initial conditions are described in [ sec : model ] .
a preparatory discussion of the evolution of mean magnetic fields subject to boundary conditions is given in
[ sec : mean_field ] , followed by the results of the simulations in [ sec : results ] .
a discussion of the implications of the results for astrophysical disks , and some potentially interesting avenues for future studies , are discussed in [ sec : discussion ] .
finally , we close with conclusions in [ sec : conclusions ] .
the global disk simulations are performed by solving the time - dependent equations of ideal magnetohydrodynamics ( mhd ) using the pluto code @xcite in a 3d spherical @xmath12 coordinate system .
the grid used for all simulations in this work has ( @xmath13 ) = ( 512,256,256 ) uniformly spaced cells , covering the spatial extent @xmath14 , @xmath15 ( where @xmath16 ) , and @xmath17 .
( the @xmath18-extent of the grid equates to @xmath19 for a constant aspect ratio disk . ) the simulated disks all have @xmath20 , where @xmath21 is the disk scale height . in terms of cells per scale height ,
the adopted grid has ( @xmath22 which is comfortably within the regime of convergence with resolution detailed by @xcite .
the simulations are performed in scaled units : @xmath23 , @xmath24 ( where @xmath25 is the speed of light ) , @xmath26 , @xmath27 , and the value of @xmath28 corresponds to the gravitational radius of a @xmath29 black hole .
the adopted numerical setup closely follows that of @xcite - a detailed description can be found there - in . in brief
, pluto code was configured to use the five - wave hlld riemann solver of @xcite , piece - wise parabolic reconstruction ( ppm - * ? ? ?
* ) , limiting during reconstruction on characteristic variables ( e.g. * ? ? ?
* ) , second - order runge - kutta time - stepping , the upwind constrained transport scheme @xcite , and the fargo - mhd module @xcite .
a mass source term is included which relaxes the gas density in the region @xmath30 , @xmath31 towards the initial density distribution over a timescale of an orbital period , where @xmath21 is the thermal disk scale - height .
this latter addition allows the total mass and energies in the disk to stabilise , permitting long run - time global disk simulations which avoid the restriction of exhausting the mass supply @xcite .
the adopted boundary conditions are identical to those used in @xcite , with the exception that for model vert - trap the radial velocity at the inner radial boundary is set to @xmath32 , where @xmath33 is the sound speed , @xmath34 is the angular velocity , and @xmath35 is the @xcite @xmath36-parameter ( see eq [ eqn : alphap ] ) .
this equates to fixing the radial velocity to a viscous outflow rate and has been used previously by @xcite and @xcite to reduce the rate at which mass exits the grid . in the present study
this boundary condition is utilised to deliberately trap magnetic field , by adopting @xmath37 in the equation for @xmath38 ( which is lower than the @xmath39 produced by the turbulent disk ) .
.summary of simulations .
@xmath40 is the plasma-@xmath41 of the initial magnetic field . [
cols="<,<,<,<",options="header " , ] in this section we begin our investigation by considering models with different initial field configurations and their respective evolution when allowed to evolve freely . in this
regard models tor-@xmath42 , tor-@xmath43 , vert , and loop represent isolated accretion disks with no external influence and a limited source of magnetic field . in [ subsec : vert - vbc ] , [ subsec : vert - bthadd ] , and [ subsec : tor - inj ] we consider models with modified boundary conditions , magnetic field injection , and mass injection .
these additional models are geared towards exploring possible external influences on the evolution of the disk turbulence .
the simulations performed in this work aim to explore the differences in the ensuing turbulent state arising from different initial field topology .
the models tor-@xmath43 , tor-@xmath42 , vert , and loop start from an almost identical disk ( in terms of density , temperature , and rotation velocity distributions ) but differ in the initial strength and topology of the magnetic field was previously presented by @xcite in the context of turbulent energetics and further detailed analysis of this simulation may be found therein . ] .
as the simulations evolve , the perturbations introduced into the initial conditions excite the mri . because of the different field topologies and field strengths in the models the respective initial mri growth rates differ .
consequently , the resultant stresses arising during the initial transient disk evolution are not the same between the simulations .
an indication of the differences in the turbulent stresses during the initial transient phase can be gained from the @xcite @xmath36-parameter , which we define as the @xmath44 component of the combined reynolds and maxwell stress normalised by the gas pressure , @xmath45 where primes denote a fluctuating ( turbulent ) deviation from an azimuthal mean , such that , @xmath46/[\rho ] , \end{array}\ ] ] and , @xmath47 , \end{array}\ ] ] where over - bars indicate a mean and square brackets indicate an azimuthal average , @xmath48 = \frac{\int q r \sin\theta d\phi } { \int r \sin \theta d\phi}.\ ] ] note that we use density - weighted averages to compute the mean velocities , consistent with the analytical approach of @xcite - see also @xcite .
the volume averages in eq ( [ eqn : alphap ] ) , and in all other instances in this work , are taken over the region @xmath49 , @xmath31 , and the entire azimuthal extent of the domain , @xmath50 . fig .
[ fig : alpha ] shows that over the first @xmath51 of the simulations ( where @xmath52 corresponds to the orbital period at a radius , @xmath53 ) model tor-@xmath42 has the highest @xmath54 , followed by vert , tor-@xmath43 , and then loop . the prominent spike in @xmath54 for model tor-@xmath42
is associated with the rapid development of large - scale , non - axisymmetric mri modes , powered by the strong toroidal field . in contrast , the relatively large transient stresses for model vert result from vertical mri mode growth , with an additional contribution from parasitic instabilities feeding on channel flows @xcite .
the dependence of the turbulent stress on initial field topology was recently examined by @xcite using high resolution global disk simulations . given the similarity in their goals to the present study , a comparison is warranted .
however , we note that the simulations presented by @xcite were only run until a time of @xmath55 ( indicated by the vertical dashed line in the upper panel of fig .
[ fig : alpha ] ) which , as noted by the authors , was the result of the limitations of a finite disk mass . within this time interval - which covers the transient evolution phase close to the beginning of the simulation - our results agree with @xcite in that there is a clear dependence of the peak saturation stress level on the initial field topology .
however , there are some differences in the details . for instance
, @xcite found that a higher saturation stress was achieved by a simulation with two poloidal loops compared to one with a toroidal field .
in contrast , we find the highest saturation stress for a toroidal field model with @xmath56 initially ( tor-@xmath42 ) , with the lowest stress level for a model with ten concentric poloidal loops ( model loop ) . focusing on the latter half of the simulations in fig .
[ fig : alpha ] , one sees that the turbulence evolves to be markedly similar in the different models . indeed
, time - averaged values are in the range @xmath57 ( noted in table [ tab : models ] ) which is consistent with the converged quasi - steady state found by @xcite for a toroidal field disk similar to tor-@xmath43 .
importantly , this demonstrates that given sufficient time an isolated turbulent accretion disk evolves to a state that is identical , irrespective of the initial magnetic field configuration .
thus , we concur with @xcite that the turbulent stress _ during the initial transient phase _ depends on the initial field configuration , whereas our results show that the late - time quasi - steady stresses ( i.e. at @xmath58 ) do not .
implications of this finding for explaining observations of astrophysical disks are discussed in
[ sec : discussion ] .
c + + + + c + further similarities between the late - time evolution of the models are found in the parameters @xmath59 and @xmath60 , which are defined to be the @xmath44 component of the maxwell stress normalised by the magnetic pressure , @xmath61 and the @xmath44 component of the reynolds stress normalised by the turbulent kinetic energy , @xmath62 where @xmath63 .
the importance of the parameter @xmath64 resides in turbulent kinetic energy generation and the significant role of the correlation between reynolds stresses with mean flow shear as an energy injection term @xcite . during the latter half of the simulation @xmath59 and @xmath60 reach quasi - steady values of @xmath65 and @xmath66 , respectively , with little difference between models tor-@xmath42 , tor-@xmath43 , vert , and loop ( table [ tab : models ] ) .
the values of @xmath67 from the models are consistent with , but slightly higher than , the value of 0.25 reported for stratified shearing - box models by @xcite .
we note that @xmath59 reaches a steady value faster than @xmath54 , allowing safe comparisons to be made against previous shorter run time global models .
as such , we find good agreement for time averaged values of @xmath59 with the studies by @xcite , @xcite , and @xcite .
the ratio of maxwell - to - reynolds stress , @xmath68 is also consistent with previous global disk studies @xcite .
following the development of the mri in the simulations , the disk structure in models tor-@xmath42 , tor-@xmath43 , vert , and loop evolves towards a quasi - steady state .
the time evolution of vertical profiles of the gas density , toroidal magnetic field , plasma-@xmath41 , and turbulent gas velocity are shown for model vert in fig .
[ fig : vert_time ] . during the initial transient phase at the start of the simulation
the rapid growth of accretion stresses drives mass out of the disk . at later times
the density distribution in the disk reaches a close equilibrium between mass inflow / outflow .
the heightened level of turbulent activity during the transient phase is evident from the turbulent gas velocity , which rises at the start of the run and then falls at later times .
there is a coincident rise and fall in magnetic field strength , illustrated by the plasma-@xmath41 parameter .
quasi - periodic oscillations indicative of dynamo cycles are apparent in the toroidal magnetic field ( upper middle panel in fig .
[ fig : vert_time ] ) .
the oscillation period of @xmath69 is in good agreement with that observed in previous shearing - box ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) and global models @xcite .
interestingly , much shorter period fluctuations can be seen in the plasma-@xmath41 and turbulent velocity , which may be associated with the intermittency of the turbulence , or a secondary cycle .
examining the vertical structure of the simulated disks ( time - averaged over the latter half of the run ) one sees markedly similar profiles for model tor-@xmath411 , tor-@xmath4120 , vert , and loop . the gas density ( fig .
[ fig : vert](a ) ) follows a roughly gaussian shape , consistent with previous global disk studies ( e.g. * ? ? ?
* ; * ? ? ?
the magnetic field strength is sub - thermal close to the mid plane , illustrated by @xmath70 in fig [ fig : vert](b ) at @xmath71 , and rises ( @xmath72 ) with increasing height in agreement with @xcite , @xcite , @xcite , and @xcite .
similarly , the turbulent gas velocities are lowest close to the mid plane , being subsonic at all heights , and increasing as one tends towards the upper / lower grid boundary . inspecting the separate components of the turbulent gas velocity , shown in fig [ fig : vert](d)-(f ) ,
the radial and azimuthal fluctuations are larger than those in the vertical direction .
similar behaviour for turbulent gas velocities was observed by @xcite , and @xcite , who found @xmath73 at the mid plane for the different velocity components .
c + c + the establishment of turbulence and associated angular momentum transport in the simulations leads to evolution of the total disk mass , as illustrated by fig .
[ fig : total_mass ] .
all of the models show a decrease in the total disk mass during the transient evolution phase , followed by a flattening as the turbulent activity subsides and a balance is reached between rates of mass leaving and entering the disk via accretion and mass injection , respectively .
the total disk mass during the quasi - steady state , @xmath74 , is noted in table [ tab : models ] .
reasonable agreement is seen in @xmath74 between models tor-@xmath42 , tor-@xmath43 , vert , and loop , despite tor-@xmath42 having a slightly higher initial disk mass of @xmath75 ( compared to @xmath76 for the other models ) . on average , the total disk mass during the quasi - steady state
is @xmath77 of the initial disk mass . surface integrated mass fluxes have been computed to assess the rates at which mass is leaving the disk in the radial , @xmath78 , and vertical , @xmath79 , directions .
it is clear from table [ tab : models ] that the radial mass flux dominates over the vertical mass flux for models tor-@xmath42 , tor-@xmath43 , vert , and loop .
models of stratified disks with a net vertical magnetic field , in both the shearing - box and * ? ? ?
* for further discussion ) .
therefore , it is the magnitude of the vertical component of the initial magnetic field , and not the field topology that is critical for shearing - box models . ] and global setting , suggest a strong correspondence between the vertical magnetic field strength and the rate of vertical mass - loss @xcite . to examine whether a similar trend is observed in the global disk models presented in this work , in fig .
[ fig : b_mdotvert ] we plot data - points for @xmath79 and @xmath80 . for all models except tor-@xmath411
there is a suggestion that larger @xmath79 values are coincident with a stronger vertical magnetic field strength ( @xmath81 ) .
this is most clearly seen at early times in the simulations ( top row of fig .
[ fig : b_mdotvert ] ) .
however , a contrasting trend is seen for model tor-@xmath411 , where the largest vertical mass - loss rates arise during the transient phase of the simulation when @xmath82 is relatively small .
interestingly , the large @xmath79 at early times ( @xmath83 ) in tor-@xmath411 correlates well with radial magnetic field ( @xmath84 ) , with a weaker correlation with the azimuthal magnetic field ( @xmath85 ) .
the data points for tor-@xmath411 in the top row of fig .
[ fig : b_mdotvert ] are consistent with vertical mass - loss resulting from stresses and magnetic pressure gradients associated with the disruption of the strong toroidal field during the early - time transient phase . before closing this section
it is pertinent to recall an important finding from the recent study by @xcite , namely that the vertical mass - loss rate depends on the vertical extent of the simulation domain .
this behaviour is linked to the position of the critical points of the wind launching , and whether the adopted vertical grid extent captures all of the relevant critical points . due to the large computational expense of performing global disk simulations at the resolution of the models presented in this work
, we must defer a comprehensive study of similar models with considerably larger vertical extent to a future study .
as such , we urge that the differences in vertical mass - loss rate between the models in this work should be taken as illustrative ( because the flow velocities are subsonic as the gas exits the vertical grid boundary ) .
further work will be needed to establish the exact quantitative details of magnetic flux transport in thin accretion disks possessing a large - scale vertical field , and whether inward or outward radial flux transport occurs ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . to demonstrate that the simulations are sufficiently well resolved , and that their common late - time evolution is not an artefact of under - resolved mri modes , we calculate a resolvability factor , @xmath86 , which is defined to be the fraction of cells in the disk body that resolve the wavelength of the fastest growing mri mode , @xmath87 , with at least 8 cells . defining ,
@xmath88 where @xmath89 , and @xmath90 is the alfv ' en speed .
time - averaged values for the resolvability factors ( computed for the quasi - steady turbulent state ) are noted in table [ tab : models ] .
the values are similar between the models , consistent with their broad similarity during the latter half of the runs .
moreover , the values are indicative of well - resolved mri - driven turbulence @xcite .
c + c + the behaviour of the magnetic energies in the different models is in keeping with their respective initial field topologies ( fig .
[ fig : beta ] ) .
for instance , defining the plasma-@xmath41 as @xmath91 , where @xmath89 or `` tot '' ( i.e. all components ) , models tor-@xmath42 and tor-@xmath43 are clearly dominated by toroidal magnetic field during the first few orbits , whereas model vert ( which started with a purely vertical field ) has a smaller @xmath92 initially .
in fact , after a few orbits , the conversion of poloidal magnetic field into toroidal field by the disk rotation results in @xmath93 clearly dominating @xmath94 in all of the models .
consistent with the diagnostics discussed in the previous sections , during the quasi - steady state ( @xmath95 ) , the evolution of @xmath41 in models tor-@xmath42
, tor-@xmath43 , vert , and loop closely matches one another .
time - averaged values are also in good agreement ( table [ tab : models ] ) .
curiously , the vertical magnetic field strength is weakest for model vert , despite it possessing the strongest vertical field in the initial conditions .
however , the indication from the plot of @xmath54 in fig .
[ fig : alpha ] is that the transient phase lasts the longest for this model .
hence , a prolonged phase with a higher accretion rate will drain the poloidal magnetic fields from the disk .
the evolution of the mean magnetic fields , discussed below , supports this explanation .
focusing next on the evolution of the mean ( volume - averaged ) magnetic fields shown in fig .
[ fig : bav ] , one sees both similarities and marked differences between the models .
firstly , they show different evolution for the mean vertical field , @xmath96 .
the initial vertical field decays throughout the simulation for model vert , as anticipated from the discussion of mean field evolution in
[ sec : mean_field ] .
second , all models show anti - correlated , quasi - periodic oscillations in @xmath97 and @xmath98
. however , these oscillations are most evident for model tor-@xmath43 , with a weaker indication for vert and loop .
thirdly , there is a suggestion of quasi - periodic oscillations in @xmath99 for models tor-@xmath42 , tor-@xmath43 , and loop , although of quite different period , being more erratic for tor-@xmath43 whilst more slowly varying for loop .
hence , although diagnostics of the turbulence ( such as @xmath54 and @xmath41 ) do not display considerable differences between the models during their late - time evolution , the mean magnetic fields do .
the intriguing implication is that the large - scale turbulent dynamo retains some memory of the initial conditions for a longer time than the turbulent stresses , and may exhibit different behaviour ( on very long timescales ) as a consequence . with regard to characterising the dynamo in the disk , it is noteworthy that the time - averaged radial and azimuthal fields are close to zero ( table [ tab : models ] ) , whereas the vertical field maintains a ( weak ) net value .
local ( shearing - box ) studies of magnetorotational turbulence indicate a correlation between the vertical field strength and the turbulent stresses @xcite . this correlation has also been cast as a relation between the wavelength of the fastest growing vertical mri mode and the stress @xcite . considering the ab initio vertical flux conservation inherent to the shearing - box ( see [ sec : mean_field ] and * ? ? ?
* ) , it is pertinent to investigate this relation using the global models in this work . to search for possible correlations between turbulent stresses and mean magnetic field components , in fig .
[ fig : alphap_btheta ] we plot the data - points for @xmath54 against those for @xmath100 ( left column ) , @xmath101 ( middle column ) , and @xmath85 ( right column ) .
the points are sampled every @xmath102 in time through the simulation , such that there is a relation between the clustering of points and the duration of time that a specific value was possessed by the model .
the upper panel of fig .
[ fig : alphap_btheta ] shows points for the time interval @xmath103 , which corresponds to the transient episode close to the start of the simulations . during this phase
the models are in the process of expelling , or turbulently corrupting , the initial field configuration . as such , the vertical field initially present in models vert and
loop has not been entirely removed and a general trend of higher @xmath54 with higher @xmath104 can be seen .
however , model tor-@xmath42 achieves the highest stresses during its transient phase ( @xmath105 ) , with a less apparent correlation between @xmath106 and any of @xmath100 , @xmath107 , or @xmath108 . by the time the quasi - steady state is reached , models tor-@xmath42 , tor-@xmath43 , vert , and loop have removed any prominent vertical field , showing similar vertical magnetic field strengths ( [ subsec : mag_fields ] ) . at these late - times ( @xmath109 )
the data - points are heavily clustered around @xmath110 and @xmath111 .
c + in summary , during the transient phase close to the beginning of the runs the models suggest a correlation between @xmath54 and @xmath96 .
however , at later times there is little sign of such a trend , mainly due to the effective removal of a mean vertical field by the evolution of the disk .
hence , it seems less a question of whether turbulent stresses , and thus the efficiency of accretion , correlate with vertical magnetic field strength , and more one of whether a global turbulent disk retains a vertical field in the face of said accretion .
the results of the previous section highlight the common late - time evolution for models that are initialised with different magnetic field configurations and for which the radial velocity at the inner radial edge of the disk ( the simulation boundary in this case ) is unconstrained .
there are , however , many examples of astrophysical scenarios in which the radial velocity may be limited by large scale effects .
for instance , clumping due to self - gravity in an extended disk @xcite . also , the excessive build up of magnetic flux close to a black hole leading to a magnetically arrested disk @xcite . in this section
we examine the impact of restricting the outflow of material at the inner radial disk edge using model vert - trap ( which adopts the viscous - outflow boundary condition described in
[ subsec : hydromodel ] ) .
the evolution of @xmath54 for model vert - trap is shown in fig . [ fig : alpha_tests ] . over the first @xmath112 of the run the turbulence develops in a very similar manner to model vert
; both models display transient accretion associated with the initiation of magnetorotational turbulence in the disk , which saturates at @xmath113 and gradually declines thereafter .
however , in model vert - trap the accretion flow builds - up at the inner radial disk edge , leading to a compression of magnetic field .
consequently , the magnetisation of the disk rises and with it the turbulent stress ( @xmath54 ) . during the latter half of the simulation ( @xmath114 ) @xmath106 and
@xmath115 reach time - averaged values of 0.21 and 1.4 , respectively ( table [ tab : models ] ) .
large amplitude fluctuations for @xmath106 bear similarities with those observed in strongly magnetized ( @xmath116 ) disk simulations by @xcite .
the vertical field becomes an order of magnitude more magnetized for vert - trap than for vert , with @xmath117 and 139 for the respective models .
strong magnetisation in the disk leads to a ratio of maxwell - to - reynolds stress , @xmath118 .
hence , @xmath54 is considerably dominated by magnetic stresses in the field - trapped state .
c + c + + + + c + the marked difference brought about by magnetic field trapping is highlighted by a comparison of the time evolution of the vertical disk structure for models vert and vert - trap in figs .
[ fig : vert_time ] and [ fig : vert_time_vt ] , respectively .
quasi - periodic oscillations are seen for model vert - trap , but only at relatively early times in the run ( @xmath119 ) . for the remainder of the run
the quasi - periodic oscillations cease , continuing with sporadic fluctuations .
a possible explanation is that the suppression of the dynamo cycles by the onset of a strong ( @xmath120 ) magnetic field , following which a mean toroidal magnetic field develops - for further discussion of this point see @xcite .
strong turbulent gas velocities arise at all heights with close - to - transonic streams of gas emanating from the disk mid plane and rising into the corona . as the turbulent activity in the disk rises , patches of the disk
are seen to rotate at a sub - keplerian rate . in fig .
[ fig : vphi_rad ] we show the radial profile of the fractional deviation from keplerian rotation as a function of time , where values have been vertically and azimuthally averaged .
one sees that the initial rotation profile ( @xmath121 ) is close to keplerian . as turbulence establishes in the disk its rotation deviates from keplerian , typically by @xmath122 during the first @xmath123 . at later times , however , considerably more sub - keplerian rotation is observed , and during the latter half of the simulation deviations of @xmath124 are seen in the radial range @xmath125 .
interestingly , this coincides with the build - up of a strong magnetic field in the disk ( see , e.g. , figs .
[ fig : beta ] and [ fig : vert_time_vt ] ) , indicating that the additional magnetic pressure support provided by the strong field alters the disk rotation rate .
consequently , the shear rate in the disk is modified , which has implications for the growth rate of the mri @xcite . also apparent from fig .
[ fig : vphi_rad ] is the development of sub - keplerian rotation from smaller to larger radii , consistent with the magnetic field piling - up at the inner radial disk edge in model vert - trap .
further differences between models vert and vert - trap are apparent from an inspection of the time - averaged vertical profiles of various quantities shown in fig .
[ fig : vert ] .
the density profile for model vert - trap shows a similar rounded shape close to the mid plane but the profile is much flatter between @xmath126 .
both the magnetic field strength and turbulent velocities are noticeably larger for vert - trap .
the magnetic field is dominated by the toroidal component , and the strength of the magnetic field drops off with increasing height away from the mid plane .
similarly , the turbulent velocities fall - off with height for vert - trap , however at a slower rate that in the other models . although the radial velocity at the inner radial disk edge is reduced in vert - trap compared to vert , the radial mass flux for vert - trap is notably larger , with @xmath127 and @xmath128 , respectively - see table [ tab : models ] .
this indicates that the density of the accretion flow has increased for vert - trap . at first sight
this result appears at odds with the lower normalisation of the vertical density profile for vert - trap compared to vert in fig .
[ fig : vert ] . on inspection of the simulation data
, one sees that mass builds up at the inner radial edge of the disk , which is outside of the region averaged - over to construct the vertical profile .
interestingly , model vert - trap displays a considerably larger vertical mass flux than model vert ( @xmath79 ) .
in fact , vertical and radial mass fluxes differ by less than a factor of two for model vert - trap .
interestingly , this suggests that the launching of a wind in a dynamically evolving turbulent disk could be instigated by a slowing of the radial accretion flow .
the evolution of mean magnetic fields proceeds differently for model vert - trap compared to vert ( see figs .
[ fig : bav ] and [ fig : bav_vert_vbc ] ) . as previously mentioned , the mean toroidal field grows considerably up until @xmath129 for vert - trap , at which point a sudden sharp field reversal occurs with a corresponding spike in @xmath54 . inspecting the correlation between vertical field ( @xmath101 ) and turbulent stresses ( @xmath54 ) in fig .
[ fig : alphap_btheta ] , one sees that during the initial transient phase of the simulation ( @xmath130 ) similar behaviour to model vert is observed .
however , at late times ( @xmath114 ) , a much stronger correlation between @xmath131 and @xmath54 is apparent . in summary , trapping of field in a turbulent disk leads to a highly magnetized state exhibiting large , highly variable accretion stresses .
in contrast , model tor-@xmath42 shows that an initially equipartition strength field will disperse if outflow is unconstrained . in this regard the results of model vert - trap and tor-@xmath42 are complementary to the study of efficiently accreting ( @xmath132 ) , strongly magnetized ( @xmath133 ) disks by @xcite and @xcite .
models vert - trap and tor-@xmath42 suggest that a strongly magnetized state can only arise , and be maintained , when radial flow is constrained .
this point is discussed further in [ subsec : environment ] .
c + c + a general trend from the models in
[ subsec : field_topology ] is that during the initial transient phase of disk evolution the field configuration is disrupted and not entirely conserved .
for example , the initially purely vertical magnetic field in model vert leaves the domain in the accretion flow and its subsequent evolution becomes similar to other models initialised with different field topologies ( e.g. tor-@xmath43 and loop ) . in this section
we use model vert - b+ to examine the disk evolution that occurs when a weak vertical magnetic field is steadily added to the simulation domain , replenishing the field that is advected away at early times .
the motivation for model vert - b+ comes from the possibility that in astrophysical disks additional magnetic field ( i.e. beyond that contained in the simulation initial conditions in this case ) may flow into the disk from larger radii or , alternatively , a field component on scales much larger than the simulation domain may pervade the disk .
the setup for model vert - b+ is identical to that of model vert , with the exception that a constant @xmath134 is added to the simulation domain between @xmath135 and for all @xmath18 and @xmath136 .
the @xmath18-component of the magnetic field is added at a constant rate such that it would generate a @xmath137 strength field in a time interval of @xmath138 .
note that the injected @xmath134 is divergence - free and is added to the staggered mesh component of @xmath134 .
the effect of injecting a weak vertical magnetic field on the turbulent stresses is shown in fig .
[ fig : alpha_tests ] .
up until a simulation time of @xmath139 the @xmath140 values for models vert - b+ and vert are quite similar . however , from @xmath141 onwards
the models diverge and the turbulent stress in model vert - b+ rises sharply . by @xmath142 the stress reaches
@xmath143 - we were unable to evolve model vert - b+ beyond this point as the highly turbulent state of the disk proved problematic for the simulation code .
as such , vert - b+ does not reach a quasi - steady state making an analysis of time - averaged quantities / profiles difficult . from the short duration of vert - b+
it is unclear whether additional vertical mass - loss in the form of a wind is triggered by the burst in activity at @xmath144 . during the simulation the disk rotation deviates from keplerian rotation ( fig .
[ fig : vphi_rad_vert - b+ ] ) .
patches of sub - keplerian rotation are observed between @xmath145 at times later than @xmath146 .
however , at @xmath147 regions of sub - keplerian rotation also begin to extend to smaller radii , coincident with the onset of the accretion burst . recalling the results of model vert - trap where sub - keplerian rotation was coincident with strong magnetic fields in the disk ,
a possible explanation for the emergence of patches of sub - keplerian rotation in model vert - b+ is that strong vertical field is dragged through the disk by the accretion flow .
this causes a back - reaction , with further field growth leading to an increase in magnetic field , as in the classic picture of field amplification via the mri @xcite .
the contrast in model evolution between vert - b+ and vert in the time interval @xmath148 does , however , highlight important differences resulting from vertical field replenishment .
firstly , it is worthwhile to note that the rate of injection of @xmath134 into the domain is not abrupt , yet the accumulation of additional field is sufficient to bring about a burst in accretion . during this burst phase the strength of the radial , vertical , and toroidal magnetic field components
are all seen to grow , illustrated by a decrease the respective plasma-@xmath41 values ( fig .
[ fig : beta ] ) .
in essence , only a relatively weak vertical magnetic field need be injected into the disk to cause a considerable change in the accretion activity of the disk .
local shearing - box models by @xcite display similar behaviour whereby the injection of a vertical magnetic field into an already turbulent disk causes considerable changes to the turbulent state .
model vert - b+ extends this picture to global disks , albeit with a simple prescription for magnetic field injection .
a detailed study of the parameter space relating to field injection ( e.g. , rate of field injection , topology of injection field , radius of field injection ) would require additional models that are beyond the scope of the current work .
the cause of the transient outburst clearly must be related to the injection of vertical magnetic field . however , on inspection , the exact details are subtle .
firstly , if allowed to accumulate , the injected magnetic field would equate to a @xmath137 vertical field in a time span of @xmath138 . at
@xmath149 one finds from eq ( [ eqn : lambdamri ] ) that @xmath150 , where @xmath151 and @xmath152 are the cell size and number of cells per scale - height in the @xmath153-direction , respectively . for @xmath154
one finds @xmath155 , indicating adequate resolution of the fastest growing mri mode for the initial field in models vert , vert - trap , and vert - b+ .
however , the transient arises after @xmath156 in model vert - b+ , after which time the accumulated field would have @xmath157 , which is borderline in terms of adequate resolution for the mri to develop .
the next clue comes from examining the mean field evolution and plasma-@xmath41 for model vert - b+ . the mean vertical field in model vert - b+ levels - off at @xmath158 ( see fig .
[ fig : bav_vert_b+ ] ) , in contrast to the continuing decline seen for vert ( fig .
[ fig : bav ] ) .
then , at @xmath159 , @xmath160 grows , which directly coincides with the burst in accretion activity ( fig .
[ fig : alpha_tests ] ) .
the magnitude of the vertical field , measured by the vertical field plasma-@xmath41 , also grows at this time ( illustrated by a decrease in @xmath161 in fig .
[ fig : beta ] ) .
therefore , although the accumulated field is relatively weak in terms of adequate resolution it counters the loss of field in the accretion flow to the extent that the disk evolution is altered .
this may arise from the injected field building - up in dormant areas of the disk ( e.g. at larger radii where the mri takes longer to set - in initially ) which , upon becoming actively turbulent , triggers a burst in accretion .
c + the simulated disks in models tor-@xmath42 , tor-@xmath43 , vert , and loop evolve to a similar quasi - steady turbulent state with no indication of a large amplitude recurrent burst of accretion will be subject to recurring accretion bursts , related to an artificial build - up of low azimuthal wavenumber energy triggering a secondary phase of linear non - axisymmetric mri growth .
however , the simulations presented in this work use an azimuthal extent of @xmath162 with no sign of a recurrent linear - mri phase , suggesting that run times longer than we have considered may be required for sufficient energy to build up . ] .
astrophysical disks , on the other hand , do exhibit recurring transient bursts .
this variability is often attributed to thermal instability @xcite , whereby the build - up of matter during the quiescent state leads to a sudden increase in disk opacity related to the ionisation of hydrogen .
subsequently , the disk heats up and the turbulent activity rises .
the compatibility of this scenario with magnetorotational turbulence has recently been demonstrated by @xcite and @xcite . from a slightly different tack ,
@xcite suggest that dwarf novae eruptions originate from a combination of magnetorotational turbulence with some secondary episodic ( global ) instability ( e.g. * ? ? ?
common to the above suggestions for triggering a transient burst is the build - up of mass in the disk . in this section
we examine the results of a supplementary simulation ( model tor - m+ ) which tests the response of a turbulent magnetic global disk to the rapid addition of zero angular momentum mass .
this model uses the state of tor-@xmath43 at @xmath163 for its initial conditions .
then , in the interval @xmath164 the mass source term ( described in [ subsec : hydromodel ] ) is modified to : i ) relax the density towards twice that of the initial conditions , and , ii ) conserve angular momentum in density increments , @xmath165 , such that the azimuthal velocity , @xmath166 , is respectively modified by @xmath167 .
we note that model tor - m+ is also applicable for understanding the subsequent disk evolution following stochastic accretion in the disk tearing scenario proposed by @xcite .
shortly after the enhanced mass injection commences the disk undergoes a transient burst of accretion activity and a sharp rise in @xmath6 for model tor - m+ at @xmath168 ( fig .
[ fig : alpha_tests ] ) .
@xmath54 subsequently decays , with a mild recurrence at @xmath169 , possibly related to clumps of mass compressing magnetic field as they advect through the disk .
this premise is supported by @xmath170 falling to @xmath171 at this time ( compared to @xmath172 for model tor-@xmath43 ) - see fig .
[ fig : beta ] and table [ tab : models ] . examining the vertical structure of the disk in model tor - m+ ( see fig . [
fig : vert ] ) it bears close similarity to model tor .
the most notable difference is in the strength of the magnetic field , which is clearly stronger in all components for model tor - m+ compared to tor , but more apparently so for the radial and vertical field .
the mechanism that triggers the outburst is two - fold .
firstly , increasing the density reduces the alfv ' en speed , which in turn reduces the wavelength of the fastest growing mri mode ( see eq [ eqn : lambdamri ] ) .
this opens up additional fast growing , short wavelength mri modes which can inject energy on the smallest scales . secondly , reducing the azimuthal velocity in regions where mass is added causes gravitational attraction to exceed the centrifugal force .
hence , over - dense regions move radially inwards , dragging field lines with them which are then sheared .
the result is a rise in radial field followed by conversion to toroidal field , both of which act to enhance the turbulent stresses .
the stresses produced by magnetorotational turbulence effectively transport angular momentum .
however questions have been raised about whether the level of angular momentum transport produced by mri - driven turbulence ( quantified by @xmath54 ) is actually sufficient to explain observations of astrophysical accretion disks .
for example , @xcite compiled results from shearing - box studies and highlighted a disparency between simulated @xmath54 values of @xmath173 and observationally inferred values of @xmath174 for post - outburst dwarf novae disks .
there are some important points to note about this discrepancy .
firstly , unstratified shearing - box studies with a net vertical field display @xmath54 values as high as 0.2 - 0.7 @xcite
. however , a typical consequence of a net vertical field is a disk wind , whereas a number of the objects discussed by @xcite show no evidence of a wind , jet , or large - scale vertical field of any description .
the evolution of a global disk threaded with a net vertical field has been explored in this work - see also @xcite and @xcite .
interestingly , during the transient phase of model tor-@xmath42 , @xmath54 reaches values as high as @xmath175 , with no associated disk wind ( i.e. the vertical - to - radial mass flux ratio is small ) . at late - times in the simulation
the net vertical field has been effectively removed , slipping away in the accretion flow , and irrespective of the initial field configuration one finds @xmath176 .
a similar range of values for @xmath106 is seen in previous global disk simulations .
for example , @xcite find @xmath177 during the late - time evolution , whereas the study by @xcite ( for which the models are not evolved beyond the initial transient phase ) reports @xmath106 as high as 0.15 .
taking stock of the above results , there is an obvious suggestion that magnetorotational turbulence can produce sufficiently large @xmath106 values to gain agreement with observation .
however , the disk must either be made to undergo a transient burst , or , a net vertical field must be retained / replenished over dynamically significant timescales .
we have demonstrated the viability of these channels with models tor - m+ , vert - trap , and vert - b+ , which produce @xmath178 .
an isolated disk will evolve to a common quasi - steady turbulent state , irrespective of the initial conditions and/or initial disturbance to the disk .
this is illustrated by models tor-@xmath43 , tor-@xmath42 , vert , and loop , for which statistical measures of the turbulence are similar during the late - time disk evolution ( see also * ? ? ?
* where the influence of different initial perturbations is examined ) .
as discussed in the previous section , astrophysical disks do exhibit transient evolution and in some cases changes in the accretion state .
model vert - trap shows that altering the rate that mass and magnetic field can leave the disk can cause abrupt changes to the accretion state .
in essence , events occurring exterior to the disk can directly influence the accretion state by regulating the inflow / outflow of material / field .
all of this points to the importance of the environment outside of the disk .
one must then give consideration to the timescale(s ) associated with the external process(es ) .
for example , if the disk expels magnetic flux in the accretion flow at a rate faster than it can be digested by the central object , be it a star or a black hole , then the build - up of magnetic flux may impinge on the disk ( e.g. * ? ? ?
* ; * ? ? ?
furthermore , accretion of magnetized gas onto the disk at a rate faster than it can be processed by the turbulence can lead to a strongly magnetized state @xcite .
the environment surrounding the accretion disk may enforce preservation of field topology , or provide a source of poloidal magnetic flux into the disk over a dynamically important timescale .
for example , the conservation of magnetic flux in collapse scenarios @xcite , accretion of magnetic field in a galactic nucleus @xcite , or wind accretion in a binary system @xcite .
global models similar to those presented in this study could be readily augmented to examine these scenarios .
launching a wind from a turbulent magnetized disk remains a challenging problem .
recent studies adopting shearing - box models have demonstrated that vertical outflows arise from a patch of a turbulent disk if initialised with a net flux vertical field @xcite . however , the radially periodic shearing - box boundary conditions preserve the net vertical field topology ( see [ sec : mean_field ] ) .
in contrast , the vertical field in model vert exits the disk in the accretion flow ( see also * ? ? ? * ) and during the quasi - steady state the ratio of vertical - to - radial mass flux is @xmath179 .
model vert - trap , on the other hand , shows that if the radial flow out of the disk is constrained ( so as to trap vertical and azimuthal field ) then the vertical - to - radial mass flux reaches @xmath180 .
hence , the results indicate that time - steady disks do not launch winds if the radial flow is unconstrained , whereas wind - launching may be triggered if the rate at which vertical / azimuthal field exits the disk is impeded ( compared to the steady - state rate ) .
the dependence of turbulent stresses on initial field configuration , magnetic field trapping due to inhibited radial outflow , magnetic field injection , and rapid mass injection has been investigated using global mhd disk simulations .
properties of the saturation stress level during the early - time transient phase of disk evolution show a dependence on the initial magnetic field .
this finding is consistent with models presented by @xcite which focused on the transient phase during roughly the first @xmath181 of the simulation .
however , a major new result in this work , and one which was achieved by utilising a mass source term to avoid the exhaustion of disk mass by ongoing accretion , is that irrespective of the strength and topology of the initial magnetic field an _ isolated _ magnetorotationally turbulent disk evolves to a statistically similar quasi - steady state that is characterised by : i ) a volume - averaged stress - to - gas - pressure ratio , @xmath182 , ii ) a gas - to - magnetic pressure ratio , @xmath183 , and , iii ) zero - net time - averaged radial and azimuthal magnetic fields .
simulations exploring the influence of trapping of magnetic field , vertical field injection , and of rapid mass injection , on the turbulence reveal that these mechanisms lead to relatively large stresses in the disk
. trapping of magnetic field produces a strongly magnetized disk with @xmath10 and considerable vertical mass flux in the form of an intermittent wind .
the injection of a vertical magnetic field , even a relatively weak one that equates to an accumulation of a @xmath137 field over a time interval of @xmath138 , is sufficient to trigger a burst in accretion with stresses reaching @xmath184 .
rapid mass injection leads to a sudden sharp increase in turbulent stresses , peaking at @xmath132 and subsequently decaying to the pre - outburst level .
the results of these simulations may be relevant for understanding accretion state changes , transient bursts , and large observationally inferred accretion efficiencies for astrophysical disks .
indeed limit cycle behaviour reminiscent of dwarf novae eruptions could be produced with the inclusion of a temperature ( and perhaps also surface density ) dependent resistivity , following similar suggestions by @xcite , @xcite , and @xcite .
i thank geoffrey bicknell , yuri levin , and daniel price for helpful discussions , the australian research council s discovery projects scheme ( project number dp1096417 ) for financial support , and the national computational infrastructure for computational resources through access to the raijin supercomputer . | the stresses produced by magnetorotational turbulence can provide effective angular momentum transport in accretion disks .
however , questions remain about the ability of simulated disks to reproduce observationally inferred stress - to - gas - pressure ratios . in this paper
we present a set of high resolution global magnetohydrodynamic disk simulations which are initialised with different field configurations : purely toroidal , vertical field lines , and nested poloidal loops .
a mass source term is included which allows the total disk mass to equilibrate in simulations with long run times , and also enables the impact of rapid mass injection to be explored .
notably different levels of angular momentum transport are observed during the early - time transient disk evolution .
however , given sufficient time to relax , the different models evolve to a statistically similar quasi - steady state with a stress - to - gas - pressure ratio , @xmath0 .
such behaviour is anticipated based on consideration of mean magnetic field evolution subject to our adopted simulation boundary conditions .
the indication from our results is that _ steady , isolated _ disks may be unable to maintain a large - scale magnetic field or produce values for the stress - to - gas - pressure ratio implied by some observations .
supplementary simulations exploring the influence of trapping magnetic field , injecting vertical field , and rapidly injecting additional mass into the disk show that large stresses can be induced by these mechanisms . in the first instance , a highly magnetized disk is produced with @xmath1 , whereas the latter cases lead to a transient burst of accretion with a peak @xmath2 . as a whole ,
the simulations highlight the common late - time evolution and characteristics of turbulent disks for which the magnetic field is allowed to evolve freely ( i.e. , without constraint / replenishment ) . in contrast , if the boundaries of the disk , the rate of injection of magnetic field , or the rate of mass replenishment are modified to mimic astrophysical disks , markedly different disk evolution occurs .
[ firstpage ] accretion , accretion disks - mhd - instabilities - turbulence |
consider a topological dynamical system @xmath0 , where @xmath8 is a continuous transformation on a compact metric space @xmath9 with metric @xmath10 .
for @xmath3 , let @xmath11 ( product of @xmath12 copies of @xmath9 ) and let @xmath13 be the space of continuous functions @xmath1 . for @xmath14 and @xmath15 ,
let @xmath16 and @xmath17 if the limit exists . for @xmath18
, define @xmath19 the problem treated in the present paper is to measure the sizes of the sets @xmath20 . to measure the sizes of the sets @xmath20 ,
we adopt the notion of topological entropy introduced by bowen ( @xcite ) , denoted by @xmath21 .
we denote by @xmath22 the set of all @xmath23-invariant probability borel measures on @xmath9 and by @xmath24 its subset of all ergodic measures .
the measure - theoretic entropy of @xmath25 in @xmath26 is denoted by @xmath27 . for @xmath28 ,
the set @xmath29 of _ @xmath25-generic points _ is defined by @xmath30 where @xmath31 stands for the weak star convergence of measures .
bowen ( @xcite ) proved that on any dynamical system , we have @xmath32 for any @xmath33 . for ergodic measure measure @xmath25 , we get equality .
but in general , the equality does nt hold . a dynamical system @xmath0 is said to be _ saturated _ if for any @xmath28 , we have @xmath34 .
it is proved in @xcite that systems of specification are saturated . in this paper
, we shall prove a variational principle which relates the topological entropy @xmath35 to the measure theoretic entropies of invariant measures in the following set , called @xmath36-fiber , @xmath37 where @xmath38 is the product of @xmath12 copies of @xmath25 .
[ vp ] suppose that the dynamical system @xmath0 is saturated .
let @xmath14 ( @xmath3 ) .
if @xmath39 , we have @xmath40 .
if @xmath41 , we have @xmath42 theorem [ vp ] is well known when @xmath6 ( see e.g. @xcite ) . in particular , it is known that for regular potential @xmath4 , @xmath43 is an analytic function ( see e.g. @xcite ) .
but as we shall see , when @xmath44 , this function can admit discontinuity even for `` very regular '' potentials .
the above consideration was motivated by the following problem .
recently the multiple ergodic limit @xmath45 have been studied by furstenberg ( @xcite ) , bergelson ( @xcite ) , bourgain ( @xcite ) , assani ( @xcite ) , host and kra ( @xcite ) , and others .
fan , liao and ma proposed in @xcite to give a multifractal analysis of the multiple ergodic average @xmath46 , in other words , to determine the hausdorff dimensions of the level sets @xmath47 this problem in its generality remains open .
however , there are two results for the shift dynamics on symbolic space and for some special potentials @xmath4 .
the first one concerns the case where @xmath48 , @xmath23 is the shift and @xmath49 ( @xmath50 being the first coordinate of @xmath51 ) . by using riesz products
, the authors in @xcite proved that for @xmath52 $ ] we have @xmath53 the second one concerns the case where @xmath54 , @xmath23 is the shift and @xmath55 is a function depending only on the first coordinates @xmath56 and @xmath57 of @xmath58 and @xmath59 .
the multifractal analysis of these double ergodic average was determined in @xcite .
a related work was done in @xcite answering a question in @xcite about the hausdorff dimension of a subset of @xmath60 for extremal values of @xmath61 .
as shown in @xcite , the dimension of the `` mixing part '' of @xmath60 which is defined by @xmath62 is equal to @xmath63 this equality is very similar to the variational principal stated in theorem [ vp ] .
in section [ v - stat ] , we recall some facts about v - statistics . in section [ top ent ]
, we recall some notions like topological entropy , generic points and specification property .
the main theorem , theorem [ vp ] , is proved in section [ proof ] . in section [ examples ] , we examine the special case of full shift together with some examples .
we will see that , even for very regular function @xmath4 , the function @xmath64 may admit discontinuity . to finish this introduction ,
we emphasise that the problem of multifractal analysis of multiple ergodic limits remains largely open .
v - statistics are tightly related to u - statistics which are well known in statistics .
let @xmath25 be a probability law on @xmath65 .
a u - parameter of @xmath25 is defined through a function called kernel @xmath66 by @xmath67 where @xmath68 is the product measure @xmath69 ( @xmath10 times ) on @xmath70 .
this @xmath71-statistics is well defined for all @xmath25 such that the integral exists . in statistics , u - parameters
are also called estimable parameters and they constitute the set of all parameters that can be estimated in an unbiased fashion . a fundamental problem in statistics
is the estimation of a parameter @xmath72 for an unknown probability law @xmath25 . to estimate a u - parameter @xmath73
, people employ the u - statistics for @xmath73 : @xmath74 where the sum is taken over all @xmath75 with @xmath76 s distinct and @xmath77 , where @xmath78 is a sequence of observations of @xmath25 . closely related to u - statistics
is the v - statistics ( von mises statistics ) : @xmath79 people expect that @xmath80 converges almost surely to @xmath73 .
this fact , if it holds , allows one to estimate @xmath73 using observations .
if it is the case , we say the u - parameter strong law of large numbers ( slln ) holds .
the u - statistics slln had been well studied for independent observations . in @xcite ,
the authors have studied the u - statistics slln for ergodic stationary process @xmath81
, i.e. @xmath82 where @xmath23 is ergodic measure - preserving transformation on a probability space @xmath83 , @xmath84 is a measurable function and @xmath85 admits @xmath25 as probability law . if @xmath86 is a kernel bounded by a integrable function and if @xmath81 is ergodic , it can be proved ( see @xcite ) that almost surely @xmath87 it is also proved in @xcite that the u - statistics slln holds if the kernel @xmath86 is continuous . in the following ,
we consider only v - statistics .
for any integer @xmath15 , the bowen metric @xmath88 on @xmath9 is defined by @xmath89 for any @xmath90 , we will denote by @xmath91 the open @xmath88-ball centered at @xmath92 of radius @xmath93 .
let @xmath94 be a subset of @xmath9 .
let @xmath90 .
a cover is a collection of bowen balls ( at most countable ) @xmath95 such that @xmath96 .
for such a cover @xmath97 , we put @xmath98 .
let @xmath99 .
define @xmath100 where the infimum is taken over all covers @xmath97 of @xmath101 with @xmath102 .
the quantity @xmath103 being a non - decreasing function of @xmath104 , the following limit exists @xmath105 consider the quantity @xmath106 as a function of @xmath107 , there exists a critical value , which we denote by @xmath108 , such that @xmath109 the following limit exists @xmath110 the limit @xmath111 is called the _ topological entropy _ of @xmath101 ( @xcite ) . for @xmath112 , we denote by @xmath113 the set of all weak limits of the sequence of probability measures @xmath114 . recall that @xmath9 is compact .
it is clear then that for any @xmath92 we have @xmath115 the following lemma is due to bowen ( @xcite ) .
[ bowen ] for @xmath116 , we have @xmath117 where @xmath118 the set @xmath119 of @xmath25-generic points is the set of all @xmath92 such that @xmath120 .
the bowen lemma implies that @xmath121 for any invariant measure @xmath25 .
it is simply because @xmath122 implies @xmath123 .
bowen also proved that the inequality becomes equality when @xmath25 is ergodic .
however , in general , we do not have the equality and it is even possible that @xmath124 .
in fact , @xmath125 according to whether @xmath25 is ergodic or not ( see @xcite ) .
the equality @xmath126 does hold for any invariant probability measure in any dynamical system with specification ( @xcite ) .
[ flp ] any dynamical system with specification @xmath0 is saturated . in other words , @xmath126 for any @xmath28 . a dynamical system @xmath127 is said to satisfy the _ specification property _ if for any @xmath90 there exists an integer @xmath128 having the property that for any integer @xmath129 , for any @xmath130 points @xmath131 in @xmath9 , and for any integers @xmath132 with @xmath133 there exists a point @xmath134 such that @xmath135 the specification property implies the topological mixing .
blokh ( @xcite ) proved that these two properties are equivalent for continuous interval transformations . mixing subshifts of finite type
satisfy the specification property . in general
, a subshift satisfies the specification if for any admissible words @xmath136 and @xmath137 there exists a word @xmath138 with @xmath139 ( some constant @xmath130 ) such that @xmath140 is admissible . for @xmath141-shifts defined by @xmath142 ,
there is only a countable number of @xmath141 s such that the @xmath141 shifts admit markov partition ( i.e. subshifts of finite type ) , but an uncountable number of @xmath141 s such that the @xmath141-shifts satisfy the specification property ( @xcite ) .
we finish this section by mentioning that continuous functions on @xmath143 can be uniformly approximated by tensor functions .
it is a consequence of the stone - weierstrass theorem .
[ sw ] let @xmath144 . for any @xmath145
, there exists a function of the form @xmath146 where @xmath147 , such that @xmath148
. we will write @xmath149
we can actually consider banach - valued @xmath150-statistics .
more than theorem [ vp ] can be proved .
let @xmath151 be a real banach space and @xmath152 its dual space .
the duality will be denoted by @xmath153 ( @xmath154 ) .
we consider @xmath152 as a locally convex topological space with the weak star topology @xmath155 . for any @xmath152-valued continuous function @xmath156
, we consider its @xmath150-statistics @xmath157 as before , formally in the same way .
fix a subset @xmath158 . for a sequence @xmath159 and a point @xmath160 ,
we denote by @xmath161 the fact @xmath162 it is clear that @xmath163 means @xmath164 converges to @xmath165 in the weak star topology @xmath155 . given @xmath166 and @xmath158 .
we define @xmath167 @xmath168 where @xmath169 denotes the vector - valued integral in pettis sense ( see @xcite ) and the inequality
@xmath170 " means @xmath171 [ vp2 ] suppose that the dynamical system @xmath0 is saturated .
if @xmath172 , we have @xmath173 .
if @xmath174 , we have @xmath175 we prove the first assertion by showing that @xmath176 implies @xmath177 .
let @xmath178 .
there exists a measure @xmath179 and a sequence of integers @xmath180 such that @xmath181 we are going to show that @xmath182 .
let @xmath183 .
then @xmath184 is a continuous function on @xmath9 . for an arbitrarily small @xmath90 , by the stone - weierstrass theorem ( see lemma[sw ] )
there exists a function @xmath185 of the form @xmath186 ( finite sum of tensor products ) such that @xmath187 notice that @xmath188 where @xmath189 denotes the ergodic sum for a given function @xmath190 .
according to ( [ limit_mu ] ) , we have @xmath191 on the other hand , we write @xmath192 where @xmath193
we have @xmath194 so , we get @xmath195 since @xmath93 is arbitrary , we have thus proved that @xmath196 .
the first assertion is then proved .
prove now the second assertion .
what we have just proved also implies @xmath197 where @xmath198 . by the bowen lemma ( lemma [ bowen ] )
, we get @xmath199 to finish the proof of the second assertion , it suffices to prove the reverse inequality of ( [ upperbound ] ) .
let @xmath200 .
let @xmath201 .
for any @xmath90 and any @xmath183 , consider @xmath185 as above .
we have @xmath202 it follows that @xmath203 letting @xmath204 we get @xmath205 in other words , we have proved @xmath206 for all @xmath196 .
so , @xmath207 by lemma [ flp ] , @xmath208 .
taking the supremum over @xmath196 leads to the reverse inequality of ( [ upperbound ] ) .
let @xmath209 with @xmath210 , where @xmath211 is the shift on the space @xmath212 .
let @xmath218 .
suppose @xmath219 in the weak star topology .
we have to show that @xmath220 since each fiber like @xmath221 is compact , there are maximizing measures @xmath222 and @xmath223 such that @xmath224 without loss of generality , we can assume that @xmath225 converge weakly , say to @xmath226 . since @xmath227 taking limit shows that @xmath228 .
it follows that @xmath229 on the other hand , recall that for the shift dynamics , the entropy function @xmath230 is upper semi - continuous .
so , @xmath231 we combine ( [ usc1]),([usc2 ] ) and ( [ usc3 ] ) to finish the proof .
[ usc ] assume that @xmath4 is a function defined on @xmath232 ( @xmath3 ) which depends only on the first @xmath130 coordinates of each of its variables ( @xmath233 ) .
then the suppremum in the variational principle ( [ variational - principle ] ) is attained by a @xmath234-markov measure .
this is just because the integral @xmath235 depends only on the values @xmath236)$ ] of the measure @xmath25 on cylinders @xmath237 $ ] and there exists a @xmath234-markov measure @xmath238 such that @xmath239 ) = \nu([a_1 , \cdots , a_k])\ ] ] for all cylinders @xmath237 $ ] and such that @xmath240 . in particular , if @xmath241 , maximizing measures are bernoulli measures .
for the bernoulli measure @xmath242 determined by a probability vector @xmath243 , we have @xmath244 where @xmath245 suppose that the function @xmath4 is a product of @xmath12 functions and each of its factor depends only on the first coordinate , i.e. @xmath246 let @xmath247 notice that @xmath248 iff @xmath249 for some probability vector @xmath243 .
the following result is a direct consequence of the last theorem . if @xmath255 , maximizing measures are markov measures .
a markov measure @xmath256 is determined by a probability vector @xmath253 and a transition matrix @xmath257 .
its entropy is equal to @xmath258 suppose @xmath259 is of the form @xmath260 let @xmath261 _ example 1 . _
consider the case @xmath272 , @xmath241 and @xmath273 .
let @xmath274 .
then @xmath275 and we have @xmath276[\phi_2(0)(1-x ) + \phi_2(1)x].\ ] ] for simplicity , we write @xmath277 for @xmath278 .
suppose that @xmath279 and @xmath280 .
otherwise , the question is trivial . by multiplying @xmath281 by a constant
we can suppose that @xmath277 is of the form @xmath282 let @xmath283 be the critical point of the quadratic function @xmath284 ( i.e. , @xmath285 ) .
_ case i. _ @xmath291 or @xmath292 ( see figure [ figure1 ] ) .
@xmath277 is strictly monotonic in the interval @xmath293 $ ] .
@xmath294 is the interval with end points @xmath295 and @xmath296 .
+ 3 . for any @xmath297
, @xmath298 admits a unique solution @xmath286 in @xmath293 $ ] . _ case ii .
_ @xmath299 ( see figure [ figure2 ] ) .
@xmath277 is strictly monotonic in the intervals @xmath300 $ ] .
@xmath294 is the interval with end points @xmath301 and @xmath296 .
+ 3 . for any @xmath297
, @xmath298 admits a unique solution @xmath286 in @xmath300 $ ] .
_ case iii .
_ @xmath302 ( see figure [ figure3 ] ) .
@xmath277 is strictly increasing in the interval @xmath303 $ ] .
@xmath294 is the interval with end points @xmath295 and @xmath301 .
+ 3 . for any @xmath297
, @xmath298 admits a unique solution @xmath286 in @xmath303 $ ] .
we can see in the case @xmath272 , @xmath241 and @xmath273 the spectrums are always continuous ( in fact , they are differentiable in the interior of @xmath308 ) . in the following examples we will see that this is no longer the case when @xmath272 , @xmath241 and @xmath309 .
this cubic polynomial function is either increasing or admit a local maximal point @xmath312 and a local minimal point @xmath313 and then we must have @xmath314 .
as we will see , the continuity of the spectrum depends on the location of @xmath312 and @xmath313 .
when @xmath284 is increasing or when @xmath315 , @xmath294 is the interval with @xmath316 and @xmath317 as end points . for any @xmath61 in the interval
, @xmath298 admits a unique solution @xmath286 in @xmath318 $ ] and @xmath319 . in this case
the spectrum is continuous ( and differentiable ) .
suppose now that @xmath277 admits a local maximal point @xmath312 and a local minimal point @xmath313 ( with @xmath314 ) . then there exist a unique @xmath320 and a unique @xmath321 such that @xmath322 we point out that there are three possible situations : the spectrum is continuous , admits one discontinuous point or admits two discontinuous points .
before present in detail these three situations we prove the following lemma which will be useful for our discussion .
let @xmath257 be a polynomial of degree 3 with positive leading coefficient .
suppose that @xmath257 admits a local maximal point @xmath312 and a local minimal point @xmath313 .
then @xmath314 and @xmath323 @xmath324 the fact @xmath314 follows from @xmath325 and @xmath326 . by the existence of the extremal points , we can write @xmath327 with @xmath328 .
it follows that @xmath329 this means that for two equidistant points from @xmath312 , the left point climbs quicker than the right point descents . by integration
, we get @xmath330 making the change of variable @xmath331 , we obtain @xmath332 the first equality holds since @xmath333 is positive in @xmath334 .
hence @xmath335 .
we prove @xmath336 in the same way . in the following we present three situations .
we use the last two lemmas .
in each situation , there is a unique point @xmath286 such that @xmath337 we call @xmath286 the maximizing point . for every @xmath338
, there could be one , two or three points @xmath92 such that @xmath290 .
the maximizing point @xmath288 is the one which is the nearest to 1/2 . in figures
[ figure4 ] , [ figure5 ] and [ figure6 ] , those parts of graph of @xmath284 corresponding to the maximizing points will be traced by solid lines , other parts will be traced by dotted lines . _ situation i. _ @xmath339 ( see figure [ figure4 ] ) . + let @xmath340 , @xmath341 , and @xmath342
. then @xmath343 and @xmath344 .
the spectrum is continuous .
the following hold : + 1 .
@xmath345 $ ] .
the maximizing points lie in @xmath346 $ ] .
@xmath277 is strictly monotonic in @xmath346 $ ] .
_ situation ii .
_ @xmath347 ( see figure [ figure5 ] ) .
+ let @xmath340 , @xmath348 , and @xmath349
. then @xmath350 , @xmath351 and @xmath352 .
the spectrum admits one discontinuous point .
the following hold : + 1 . @xmath353
the maximizing points lie in @xmath346\cup ( x',1]$ ] .
@xmath277 is strictly monotonic in each of above two intervals .
the spectrum has one discontinuous point at @xmath354 , the entropy jumps from @xmath355 to @xmath356 .
_ situation iii .
_ @xmath357 ( see figure [ figure6 ] ) .
+ let @xmath358 , @xmath348 , and @xmath349
. then @xmath359 , @xmath360 , @xmath361 and @xmath362 .
the spectrum admits two discontinuous points .
the following hold : + 1 .
@xmath353 $ ] .
the maximizing points lie in the intervals @xmath363\cup ( x',1]$ ] .
@xmath277 is strictly monotonic in each of above three intervals .
the spectrum has two discontinuity points .
one is @xmath364 , where the entropy jumps from @xmath365 to @xmath366 , the other is @xmath354 , where the entropy jumps from @xmath355 to @xmath356 .
d. ruelle , thermodynamic formalism .
the mathematical structures of classical equilibrium statistical mechanics , " encyclopedia of mathematics and its applications , * 5*. addison - wesley publishing co. , 1978 . | let @xmath0 be a topological dynamical system and let @xmath1 be a continuous function on the product space @xmath2 ( @xmath3 ) .
we are interested in the limit of v - statistics taking @xmath4 as kernel : @xmath5 the multifractal spectrum of topological entropy of the above limit is expressed by a variational principle when the system satisfies the specification property . unlike the classical case ( @xmath6 ) where the spectrum is an analytic function when @xmath4 is hlder continuous , the spectrum of the limit of higher order v - statistics ( @xmath7 ) may be discontinuous even for very nice kernel @xmath4 . |
the study of exact spherically symmetric perfect fluid solutions of einstein s equations has been reinvigorated over the last decade or so and we now have a remarkable collection of solution generating techniques @xcite . at the same time , interest in exact anisotropic fluid solutions of einstein s equations has grown @xcite . whereas a solution generating technique is known for anisotropic fluids @xcite , this requires the specification of two input functions and one is not guaranteed a reasonable output , say in the density @xmath2 .
a less general technique is given here , but the algorithm presented requires only one input function , the density itself .
the technique is demonstrated by the construction of an infinite number of regular solutions , some of which satisfy all the standard energy conditions .
let us write the geometry in the familiar form @xcite @xmath3 where @xmath4 is the metric of a unit 2-sphere ( @xmath5 ) .
the source of ( [ metric ] ) , by way of einstein s equations , is taken to be a comoving fluid described by the stress - energy tensor @xmath6 $ ] .
einstein s equations give the effective gravitational mass @xmath7 the source equation for @xmath8 @xmath9 and the generalized tolman - oppenheimer - volkoff equation which we write in the form @xmath10 where @xmath11 .
the fluid distribution is assumed to terminate at @xmath12 where @xmath13 and to join there , by way of a regular boundary surface , onto a schwarzschild vacuum of mass @xmath14 .
no further conditions need be specified for this junction .
a fluid is considered regular if it is free of singularities .
scalar polynomial singularities involve scalar invariants built out of the riemann tensor . due to the spherical symmetry assumed here
, there are only four independent invariants of this type @xcite .
a direct calculation shows that @xmath15 and @xmath16 do not enter these invariants and that the ricci invariants are regular as long as @xmath17 and @xmath18 are .
however , the second weyl invariant grows like @xmath19 as @xmath20 . as a consequence
, we have the following necessary conditions for regularity @xmath21 the latter of which is already obvious from ( [ peqn ] ) .
one could , of course , incorporate derivatives of the riemann tensor into the construction of invariants .
for example , a calculation shows that @xmath22 grows like @xmath23 as @xmath20 where @xmath24 is the ricci scalar and @xmath25 is the covariant dalembertian .
however , the background theory involves the riemann tensor , not its derivatives , and so we impose no a priori restrictions beyond ( [ regular ] ) .
the algorithm proposed here involves the use of the newtonian hydrostatic equilibrium equation for an isotropic fluid , @xmath26 in effect then @xmath2 and @xmath27 are related exactly as in the newtonian theory of isotropic fluids , e.g. @xcite . a solution in einstein s theory
is obtained simply by reading off @xmath18 from ( [ peqn ] ) . no isotropic solutions for einstein
s theory can be found in this way since if we set @xmath28 in ( [ peqn ] ) , and use ( [ hydro ] ) , the resultant quadratic equation for @xmath27 returns @xmath29 .
one can take the view that it is the tangential stress @xmath18 that holds these configurations together . from ( [ peqn ] ) and ( [ hydro ] ) , and assuming @xmath30 , we have @xmath31 explicitly , inserting ( [ phi ] ) and ( [ hydro ] ) into ( [ peqn ] ) we obtain @xmath32 as a result , the weak and classical strong energy conditions are automatically satisfied via the proposed algorithm , as long as @xmath30 .
the dominant energy condition is not @xcite .
however , in the present circumstance , the classical strong energy conditions are in fact weaker than the weak energy conditions and so the trace energy condition @xmath33 is usually applied .
subject to ( [ trace ] ) it is known that the buchdahl bound @xmath34 @xcite is robust @xcite . from ( [ trace ] )
we can write @xmath35 and consider the dominant energy condition as @xmath36 and the stronger trace energy condition as @xmath37 @xcite .
the procedure outlined above can be executed with the specification of @xmath2 alone .
we demonstrate this here with an infinite number of simple examples mostly derived from the ansatz @xmath38 where @xmath39 and @xmath40 are constants @xmath41 and @xmath42 is an integer @xmath43 .
first , let us consider the simplest case , that of constant density , @xmath44 @xcite . from ( [ mass ] ) and ( [ rhos ] ) it follows that @xmath45 and so @xmath46 . from ( [ hydro ] ) , and setting @xmath13 , we obtain @xmath47 and so @xmath48 .
this is a newtonian polytrope of zero index , e.g. @xcite .
we can now simply read off @xmath18 from ( [ peqn ] ) with ( [ phi ] ) .
we find @xmath49 to avoid singularities in @xmath18 then @xmath50 and so for non - singular solutions @xmath51 as expected . at @xmath52
we find @xmath53 along with @xmath54 at @xmath12 we find @xmath55 along with @xmath56 for solutions with @xmath57 , since @xmath58 , the dominant energy condition is also satisfied throughout the solution . from ( [ pgradboundaryconst ] ) for these solutions
we find @xmath59 where @xmath60 , and so in the limiting case @xmath61 somewhat larger that the the buchdahl bound @xmath62 @xcite .
some examples are shown in figure [ figure1 ] . in order to demonstrate that energy conditions of the type ( [ radtio ] ) can be satisfied
, it is convenient to define @xmath63 and @xmath64 so that @xmath65 a number of examples are shown in figure [ figure2 ] .
we set @xmath67 so that @xmath68 .
other boundary conditions are given below . from ( [ mass ] ) and ( [ rhos ] ) it follows that @xmath69 and so @xmath70 from ( [ hydro ] ) , ( [ rhos ] ) and ( [ masss ] ) we obtain @xmath71 we can now simply read off @xmath18 from ( [ peqn ] ) with ( [ phi ] ) .
we find @xmath72 where @xmath73 is a polynomial of degree @xmath74 , given explicitly in appendix a , and @xmath75 as shown in appendix b , ( [ poly ] ) has no roots for @xmath76 as long as @xmath77 subject to condition ( [ alimit ] ) then the solutions given here are free of singularities throughout the range @xmath78 .
whereas the configurations satisfy the weak and strong energy conditions , the dominant energy condition is satisfied for sufficiently large @xmath39 .
this is demonstrated below . at @xmath52
we find @xmath79 along with @xmath80 at @xmath12 we find @xmath81 along with @xmath82 from ( [ rs ] ) , ( [ ms ] ) and ( [ alimit ] ) we find @xmath83 since @xmath84 , the solutions given here allow an arbitrarily large degree of compactness , in principle .
however , smaller values of @xmath85 correspond to smaller values of @xmath39 which lead to a violation of the dominant ( and therefor trace ) energy condition . again
, this is demonstrated below . from ( [ tenuity ] )
it follows that the buchdahl bound @xmath86 is obtained for @xmath87 for @xmath88 there is no `` core '' in the density .
it follows from ( [ alimit ] ) that @xmath89 and from ( [ tenuity ] ) that @xmath90 .
the physical variables are shown in figure [ figure3 ] for @xmath91 and in figure [ figure4 ] for @xmath92 , the buchdahl limit . in the first case
all energy conditions ( excepting the trace ) are satisfied and @xmath93 .
( the trace energy condition can be satisfied with a slightly larger @xmath39 , for example , @xmath94 . ) in the second case the dominant energy condition fails . unlike the case @xmath44 , the minimum @xmath85 for which the dominant energy condition is satisfied must be found numerically for each @xmath42 . for @xmath88 the minimum @xmath95 , of course above the buchdahl limit . for @xmath96 it follows from ( [ alimit ] ) that @xmath97 and from ( [ tenuity ] ) that @xmath98 .
the physical variables are shown in figure [ figure5 ] for @xmath99 and in figure [ figure6 ] for @xmath100 , the buchdahl limit . in the first case
all energy conditions ( excepting the trace ) are satisfied and @xmath101 .
( the trace energy condition can be satisfied with a slightly larger @xmath39 , for example , @xmath102 . ) in the second case the dominant energy condition fails .
the solutions with @xmath103 are qualitatively similar , but with a flatter density profile for smaller @xmath104 , that is , a larger `` core '' .
an example is shown in figure [ figure7 ] .
the newtonian polytrope of zero index was discussed above , but it is well known that there are two finite analytical polytropic solutions , e.g. @xcite . in this section
we use a classical polytrope of index 1 to generate an anisotropic fluid .
let us take @xmath105 where @xmath39 is a constant , so that from ( [ hydro ] ) @xmath106 and @xmath107 from ( [ mass ] ) and ( [ poly1rho ] ) we have @xmath108 and so @xmath109 from which we have @xmath110 and so @xmath111 for the buchdahl bound to hold .
we now find @xmath112 where @xmath113 is a polynomial of degree @xmath114 , given explicitly in appendix c. at @xmath52 we find @xmath115 at @xmath12 we find @xmath116 along with @xmath117 where the last inequality assumes the buchdahl bound .
these models are qualitatively similar to the case @xmath96 discussed above . in a manner analogous to figure [ figure2 ] some details associated with these models are shown in figure [ figure8 ] .
an algorithm which converts isotropic newtonian static fluid spheres into anisotropic einsteinian static fluid spheres has been given which requires one input function , the density . an infinite number of explicit regular solutions have been generated , some of which satisfy all the standard energy conditions .
it is a pleasure to thank hkan andrasson and boyko ivanov for useful comments .
this work was supported by a grant from the natural sciences and engineering research council of canada .
portions of this work were made possible by use of _ grtensorii _ @xcite .
for @xmath66 and @xmath118 we find @xmath119 \\
\nonumber+{b}^{2}{r}^{2 } ( n+3 ) ^{2 } ( n+2 ) ^{2 } ( n+1 ) ^{2 } ( 4 \,{b}^{2}{r}^{4}{\pi } ^{2}-3 + 8\,b{r}^{2}\pi ) + 6\,a b{r}^{n+2 } ( n+3 ) ( n+2 ) ( n+6 ) ( n+ 1 ) ^{2}\\ \nonumber-2\,\pi \,{b}^{2}a{r}^{n+4 } ( n+36 ) ( n+3 ) ( n+2 ) ^{2 } ( n+1 ) ^{2}-16\,{\pi } ^{2}{b}^{3}a{r}^{n+6 } ( n+3 ) ( n+2 ) ( n+6 ) ( n+1 ) ^{2}\\ \nonumber-9\,{a}^{2}{r}^{2 + 2\,n } ( n+3 ) ( n+1 ) ( n+2 ) ^{2}+12\,{a}^{2}\pi \,b{r}^{4 + 2\,n } ( n+2 ) ( 3\,{n}^{2}+20\,n+36 ) ( n+1 ) ^{2}\\ \nonumber+8\,{\pi } ^{2}{b}^{2}{a}^{2}{r}^{6 + 2 \,n } ( n+1 ) ( 5\,{n}^{3}+47\,{n}^{2}+144\,n+108 ) -18\,{a}^{3}\pi \,{r}^{4 + 3\,n } ( 4 + 3\,n ) ( n+1 ) ( n+2 ) ^{2}\\ \nonumber-48\,{\pi } ^{2}{a}^{3}b{r}^{6 + 3\,n } ( n+6 ) ( n+2 ) ( n+1 ) + 36\ , { \pi } ^{2}{a}^{4}{r}^{6 + 4\,n } ( n+2 ) ^{2}.\end{aligned}\ ] ]
this follows immediately from the following elementary theorem @xcite : for @xmath120 @xmath121 real and @xmath122 integers , there are no positive roots as long as @xmath66 and @xmath123 where @xmath124 .
in terms of ( [ poly1rho ] ) and ( [ poly1mass ] ) we find [ sec : texbooks ] see , for example , s. rahman and m. visser , class . quant .
* 19 * , 935 ( 2002 ) [ arxiv : gr - qc/0103065 ] , k. lake , phys .
d * 67 * ( 2003 ) 104015 [ arxiv : gr - qc/0209104 ] , d. martin and m. visser , phys .
d * 69 * ( 2004 ) 104028 [ arxiv : gr - qc/0306109 ] , p. boonserm , m. visser , and s. weinfurtner , phys .
d * 71 * ( 2005 ) 124037 [ arxiv : gr - qc/0503007 ] , p. boonserm , m. visser , and s. weinfurtner , phys .
d * 76 * ( 2007 ) 044024 [ arxiv : gr - qc/0607001 ] , p. boonserm , `` some exact solutions in general relativity '' , msc thesis , victoria university of wellington , 2005 [ arxiv : gr - qc/0610149 ] , p. boonserm , m. visser and s. weinfurtner , journal of physics : conference series * 68 * ( 2007 ) 012055 , [ arxiv : gr - qc/0609088 ] and p. boonserm , m. visser and s. weinfurtner , `` solution generating theorems : perfect fluid spheres and the tov equation '' , [ arxiv : gr - qc/0609099 ] ( marcel grossmann 11 ) . for a review of these developments
see p. boonserm and m. visser , int .
d * 17 * , 135 ( 2008 ) [ arxiv:0707.0146v1 ] .
for further solutions , see , for example , j. loranger and k. lake , phys .
d * 78 * , 127501 ( 2008 ) , [ arxiv:0808.3515 ] , c. grenon , p. j. elahi and k. lake , phys .
d * 78 * , 044028 ( 2008 ) [ arxiv:0805.3329v2 ] , and k. lake , phys .
d * 77 * , 127502 ( 2008 ) , [ arxiv:0804.3092v2 ] .
see , for example , l. herrera and n. o. santos , phys .
rep . * 286 * , 53 ( 1997 ) , d. e. barraco , v. h. hamity and r. j. gleiser , phys .
d * 67 * , 064003 ( 2003 ) , l. herrera , a. di prisco , j. martn , j. ospino , n. o. santos and o. troconis , phys .
d * 69 * , 084026 ( 2004 ) , c. cattoen , t. faber and m. visser , class .
quantum grav .
* 22 * , 4189 ( 2005 ) , a. debenedictis , d. horvat , s. ilijic , s. kloster and k. viswanathan , class .
* 23 * , 2303 ( 2006 ) , g. bohmer and t. harko , class .
quantum grav .
* 23 * , 6479 ( 2006 ) , w. barreto , b. rodrguez , l. rosales and o. serrano , gen .
. grav . * 39 * , 23 ( 2007 ) , m. esculpi , m. malaver and e. aloma , gen .
grav . * 39 * , 633 ( 2007 ) , g. bohmer and t. harko , mon . not .
379 * , 393 ( 2007 ) , s. karmakar , s. mukherjee , r. sharma and s. maharaj , pramana j. * 68 * , 881 ( 2007 ) , h. abreu , h. hernndez and l. a. nez , class .
quantum grav .
4631 ( 2007 ) and s. viaggiu , int . j. mod .
d * 18 * , 275 ( 2009 ) [ arxiv:0810.2209v2 ] .
l. herrera , j. ospino and a. di prisco , phys .
d * 77 * , 027502 ( 2008 ) , [ arxiv:0712.0713v3 ] .
see also k. lake , phys .
* 92 * , 051101 ( 2004 ) , [ arxiv : gr - qc/0302067v5 ]
. we use geometrical units and usually designate functional dependence only on the first appearance of a function .
we assume that all functions are at least @xmath126 .
k. santosuosso , d. pollney , n. pelavas , p. musgrave and k. lake , computer physics communications * 115 * , 381 ( 1998 ) , [ arxiv : gr - qc/9809012 ] .
s. chandrasekhar , _ an introduction to the study of stellar structure _ ,
( university of chicago press , chicago , 1939 ) . in the present notation
, the weak energy condition requires @xmath30 , @xmath127 and @xmath128 ; the strong energy condition requires @xmath127 , @xmath128 and @xmath129 ; and the dominant energy condition requires @xmath30 , @xmath130 and @xmath131 .
see , for example , m. visser , _ lorentzian wormholes _
( springer - verlag , new york , 1996 ) and e. poisson , _ a relativist s toolkit : the mathematics of black - hole mechanics_(cambridge university press , cambridge , 2004 ) .
see , for example , j. guven and n. .
murchadha , phys .
d * 60 * , 084020 ( 1999 ) .
h. andrasson , j. diff eqns , * 245 * , 2243 ( 2008 ) [ arxiv : gr - qc/0702137 ] .
see also p. karageorgis and j. g. stalker , class .
* 25 * , 195021 ( 2008 ) [ arxiv:0707.3632 ] .
the resultant bounds on the surface redshift are given by b. v. ivanov , phys .
d * 65 * , 104011 ( 2002 ) .
it is important to note that the solution we are considering is not the most general solution to ( [ peqn ] ) with constant @xmath2 .
rather , it is the solution , with constant @xmath2 , that satisfies ( [ hydro ] ) .
compare , for example , r. l. bowers and e. p. t. liang , astrophys .
j. * 188 * , 657 ( 1974 ) .
a proof by s. a. levin can be found at ` http://sepwww.stanford.edu/oldsep/stew/descartes.pdf ` this is a package which runs within maple .
it is entirely distinct from packages distributed with maple and must be obtained independently .
the grtensorii software and documentation is distributed freely on the world - wide - web from the address ` http://grtensor.org ` | i use the newtonian equation of hydrostatic equilibrium for an isotropic fluid sphere to generate exact anisotropic solutions of einstein s equations .
the input function is simply the density .
an infinite number of regular solutions are constructed , some of which satisfy all the standard energy conditions .
two classes of these solutions generalize the newtonian polytropes of index @xmath0 and @xmath1 . |
the top quark is the heaviest known elementary particle , and it plays a fundamental role in testing the standard model ( sm ) and the extensions of the sm .
its production and decay channels are important probes of new physics , and because of its large coupling to the higgs , the top quark production processes provide a sensitive probe of electroweak symmetry breaking . the total cross - section for the top quark pair production has been calculated up to nnlo within the @xmath15-scheme in refs .
the sm estimates , especially those obtained by using the principle of maximum conformality ( pmc ) @xcite , agree well with the experimental result which has been measured with a precision @xmath16 at the tevatron @xcite and @xmath17 at the lhc @xcite .
the top quark forward - backward asymmetry which originates from charge asymmetry physics @xcite has also been studied at the tevatron and lhc .
two options for the asymmetry have been used for experimental analysis ; i.e. the @xmath0-rest frame asymmetry @xmath18 and the @xmath19-laboratory frame asymmetry @xmath20 where @xmath21 is the top quark rapidity in the @xmath0-rest frame and @xmath22 is the top quark rapidity in the @xmath19-laboratory frame ( or the @xmath19 center - of - mass frame ) .
the cdf and d0 collaborations have found comparable values in the @xmath0-rest frame : @xmath23 @xcite and @xmath24 @xcite , where the uncertainties are derived from a combination of statistical and systematic errors .
the asymmetry in the @xmath19-laboratory frame measured by cdf is @xmath25 @xcite .
the cdf collaboration has also measured the dependence of @xmath26 with respect to the @xmath0-invariant mass @xmath27 : the asymmetry increases with @xmath27 , and @xmath28 @xcite .
these measured top quark forward - backward asymmetries are much larger than the usual sm estimates .
for example , the nlo qcd contributions to the asymmetric @xmath0-production using conventional scale setting yield @xmath29 and @xmath30 ( see e.g. @xcite ) , which are about @xmath31-deviation from the above measurements .
for the case of @xmath32 , using the mcfm program @xcite , one obtains @xmath33 which is about @xmath34-deviation from the data .
these discrepancies have aroused great interest because of the possibility for probing new physics beyond the standard model .
a recent reevaluation of the electroweak correction raises the qcd asymmetries by at most @xmath35 : i.e. @xmath36 @xcite and @xmath37 @xcite .
it has been argued that the missing higher - order corrections can not be the reason for the significant discrepancy @xcite .
in fact , only a several percent increment has been observed in ref.@xcite using a next - to - next - to - leading - logarithmic ( nnll ) calculation .
it is for this reason that many new physics models beyond the sm have been suggested .
since the sm estimate on the total cross - section agrees well with the experimental data , it is hard to understand the large deviation of the asymmetry . before introducing any new physics ,
it is best to have a more precise estimation within the sm .
it should be noted that all the present sm estimations are based on the conventional scale setting , where the renormalization scale @xmath38 is set to be the typical momentum transfer @xmath39 of the process ; i.e. @xmath40 .
one then estimates the scale - uncertainty by varying @xmath41 $ ] , which will lead to a @xmath42 scale uncertainty to the asymmetry .
usually , it is argued that this scale uncertainty can be suppressed by including higher - order corrections in an order - by - order manner .
the conventional scale setting procedure is clearly problematic since the resulting fixed - order pqcd prediction will depend on the choice of renormalization scheme .
in fact , it gives the wrong result when applied to qed processes .
it should be recalled that there is no ambiguity in setting the renormalization scale in qed . in the standard gell - mann - low scheme for qed ,
the renormalization scale is the virtuality of the virtual photon @xcite .
for example , the renormalization scale for the electron - muon elastic scattering through one - photon exchange can be set as the virtuality of the exchanged photon , i.e. @xmath43 . but it is wrong to use @xmath39 directly as the scale for any other renormalization scheme . some displacement must be included in order to ensure scheme - independence .
for example , under the @xmath44 scheme we have @xmath45 @xcite .
this result shows that the effective scale of the @xmath15 scheme should generally be about half of the true momentum transfer occurring in the interaction .
the invariance under choice of renormalization scheme is a consequence of the transitivity property of the renormalization group @xcite .
of course , the question is more complicated in qcd due to its non - abelian nature .
recently , it has been suggested that one can systematically fix the renormalization scale at any fixed order by using the pmc @xcite .
the pmc provides the principle underlying the brodsky - lepage - mackenzie method @xcite , and they are consistent with each other through the pmc - blm correspondence principle @xcite .
the main idea is that , after proper procedures , all nonconformal @xmath46-terms in the perturbative expansion are summed into the running coupling so that the remaining terms in the perturbative series are identical to that of a conformal theory ; i.e. , the corresponding theory with @xmath47 . the underlying conformal symmetry is a useful principle for physics ; e.g. the ads / qcd theory @xcite , the conformal general relativity model @xcite and the canonical quantum gravity theory @xcite .
after pmc scale setting , the divergent renormalon series with @xmath48-growth does not appear in the conformal series .
this is consistent with the treatment done in ref.@xcite . since renormalon terms
are absent , one obtains a more convergent perturbative expansion series , and thus the full next - to - leading order ( nlo ) , or even the leading - order ( lo ) calculation , is often enough to achieve the required accuracy .
the pmc scale @xmath49 is unambiguous at any finite order .
we emphasize that the pmc is consistent with the renormalization group property that a physical result is independent of the renormalization scheme and the choice of the initial renormalization scale @xmath2 .
any residual dependence on @xmath2 for a finite - order calculation is highly suppressed since the unknown higher - order @xmath46-terms will be absorbed into the pmc scales higher - order perturbative terms .
as an application , we have previously applied the pmc procedure to obtain nnlo predictions for the @xmath0-pair hadroproduction cross - section at the tevatron and lhc colliders @xcite .
it is found that there is almost no dependence on the choice of initial renormalization scale ; i.e. the total cross - section remains almost unchanged even when taking very disparate initial scales @xmath2 equal to @xmath3 , @xmath4 , @xmath5 and @xmath6 , thus greatly improving the precision of the qcd prediction . by using the pmc scales ,
a larger @xmath1 is obtained in comparison to the conventional scale setting , which agrees well with the present tevatron and lhc data .
it is thus interesting to see whether the use of pmc scales , especially those of the dominant asymmetric @xmath7-channel at the tevatron , can improve our understanding on the top quark forward - backward symmetry ; this is the purpose of the present paper .
the remaining parts of this paper are organized as follows : in sec .
ii , we give the relevant formulae for the top quark forward - backward asymmetry .
the new properties of the predictions after pmc scale setting are presented . in sec .
iii , we present the numerical results and some discussions for the top quark forward - backward asymmetry at the tevatron .
iv provides a summary .
before discussing the top quark forward - backward asymmetry , we first review the total cross - sections for the top quark pair production at the tevatron up to nnlo . a comparison of the production cross - sections using conventional scale setting versus pmc scale setting will give us some useful information on how the pmc scale setting can improve our understanding of the top quark pair production : the relative importance of all the production channels , especially those which provide the asymmetries ; the convergence of the perturbative series for each production channel ; etc .
this information will be helpful for constructing a more precise perturbative expansion for calculating the top quark forward - backward asymmetry .
.total cross - sections ( in unit : pb ) for the top quark pair production at the tevatron with @xmath19-collision energy @xmath50 tev . for conventional scale
setting , we set the renormalization scale @xmath51 . for pmc scale
setting , we set the initial renormalization scale @xmath52 . here
we take @xmath53 gev and use the msrt 2004-qed parton distributions @xcite as the pdf . [
cols="^,^,^,^,^,^,^,^,^ " , ] analytical expressions up to nnlo have been provided in the literature , e.g. ref.@xcite ,- terms at the nnlo have been well set . a full nnlo calculation for the @xmath7-channel
has been recently presented by using the conventional scale setting which also shows a very small perturbative uncertainty between the nnlo and nnll calculation , i.e. @xmath54 @xcite .
] and the explicit calculation technology for pmc scale setting can be found in ref.@xcite , so we will not present them here .
numerical results for the top quark pair production at the tevatron with @xmath19-collision energy @xmath50 tev are presented in table [ tab1 ] .
the coulomb - type corrections will lead to sizable contributions in the threshold region @xcite which are enhanced by factors of @xmath55 .
thus the terms which are proportional to @xmath56 or @xmath57 ( @xmath58 is the top quark velocity in the partonic center - of - mass frame ; @xmath59 is the subprocess center - of - mass energy squared ) at the nlo or nnlo level should be treated separately @xcite . for this purpose ,
the results listed in the _ total_-column is not a simple summation of the corresponding lo , nlo and nnlo results ; the results are obtained by using the sommerfeld rescattering formula to treat the coulomb part . in doing the numerical calculation , we set @xmath60 gev @xcite and the factorization scale @xmath61 .
we set @xmath62 for conventional scale setting , and take the initial renormalization scale @xmath63 to initialize pmc scale setting . for the pdfs , we adopt msrt 2004-qed parton distributions @xcite to be consistent with the choice of ref.@xcite . by comparing with the total cross - sections derived from the pmc scale setting and the conventional scale setting listed in table [ tab1 ] , we observe the following points : * at the tevatron , the top quark pair cross - section is dominated by the @xmath7-channel which provides @xmath64 contribution to the total cross - section .
the @xmath7-channel is asymmetric at the nlo level , so it will lead to sizable top quark forward - backward asymmetry at the tevatron .
in contrast , one finds that the dominant channel at the lhc is the symmetric @xmath65-channel , c.f .
ref.@xcite , so the top quark forward - backward asymmetry from other channels will be greatly diluted at the lhc ; this asymmetry becomes small which agrees with the cms and atlas measurements @xcite .
accordingly , at present , we will concentrate on the top quark forward - backward asymmetry at the tevatron .
+ + * at the lowest order , the two channels @xmath66 and @xmath67 do not discriminate the final top quark and top - antiquark , so their differential distributions are symmetric for the hadronic production process . at the nlo level , either the virtual or real gluon emission will cause sizable differences between the differential top quark and top - antiquark production , thus leading to an observable top quark forward - backward asymmetry . at the tevatron ,
the asymmetric channels are @xmath7- , @xmath68- and @xmath69- channels accordingly .
table [ tab1 ] shows the total cross - sections of the @xmath68 and @xmath69 channels are quite small , less than @xmath70 of that of @xmath7-channel , so their contributions to the asymmetry can be safely neglected .
figure ( [ asy1 ] ) shows the dominant feynman diagrams for the qcd charge asymmetry at the nlo level and fig.([asy2 ] ) shows the less important flavor - excitation feynman diagrams for the qcd charge asymmetry at the nlo level . *
when using conventional scale setting , the nnlo cross - section for the @xmath7-channel is about @xmath71 of its nlo cross - section ; i.e. @xmath72 . thus in order to derive a consistent asymmetry up to nnlo , one must consider the asymmetric contribution from the nnlo @xmath7-channel , which may be sizable .
in contrast , a much more convergent pqcd series expansion is obtained after pmc scale setting , since all non - conformal @xmath46-terms in the perturbative expansion are summed into the running coupling .
for example for the asymmetric @xmath7-channel , the value of @xmath73 is lowered to be only @xmath74 .
this shows that after pmc scale setting , the change to the asymmetry from the nnlo is greatly suppressed . *
writing the numerator and the denominator of the two asymmetries @xmath75 defined by eqs.([afbtt],[afblab ] ) in powers of @xmath76 , we obtain + @xmath77 , \end{aligned}\ ] ] + where the @xmath78-terms stand for the total cross - sections at certain @xmath76-order and the @xmath79-terms stand for the asymmetric cross - sections at certain @xmath76-order .
the terms up to nlo ( @xmath80 ) have been calculated , whereas only parts of @xmath81 and @xmath82 are currently known @xcite . + as shown in table [ tab1 ] , using conventional scale setting , the relative importance of the denominator terms is @xmath83 $ ] , and the numerator terms for the asymmetric @xmath7-channel satisfy @xmath84 $ ] . is not available , as a first approximation , we treat these asymmetric terms to have the same relative importance as their total cross - sections ; i.e. @xmath85 . ]
thus , the @xmath86 term and the @xmath82 term have the same importance .
because the nnlo @xmath82 term is not available at the present , one has to use the lowest - order @xmath87 cross - section in the denominator and the @xmath88 term in the numerator ; i.e. dealing with only the so - called lo asymmetry @xcite : @xmath89 .
+ however , after pmc - scale setting , we have @xmath90 $ ] and the numerators for the asymmetric @xmath7-channel becomes @xmath91 $ ] .
it shows that , after pmc scale setting , the nnlo corrections for both the total cross - sections and the asymmetric part are lowered by about one order of magnitude .
therefore , the nnlo - terms @xmath92 and @xmath93 can be safely neglected in the calculation , and we can obtain the asymmetry at the so - called nnlo level : @xmath94 .\ ] ] furthermore , it is natural to assume that those higher - order terms @xmath79 and @xmath78 with @xmath95 after pmc scale setting will also give negligible contribution ;- dependent but renormalization scale independent terms should not be absorbed into the coupling constant .
an important example in qed case is the electron - loop light - by - light contribution to the sixth - order muon anomalous moment which is of order @xmath96 @xcite . ]
the above asymmetry can thus be resummed to a more convenient form : @xmath97 + .
the wave lines stand for the photon.,scaledwidth=48.0% ] * as argued by refs .
@xcite , the electromagnetic and weak interaction will provide an extra @xmath98 increment for the asymmetry .
this shows that the electromagnetic contribution provides a non - negligible fraction of the qcd - based antisymmetric cross - section with the same overall sign .
the asymmetry to be calculated thus changes to @xmath99 representative diagrams contributing to the qcd - qed interference term @xmath100 at the order @xmath101 are shown in fig.([asyqed ] ) .
the weak contributions to the asymmetry are obtained by changing the photon propagator to be a @xmath102-propagator .
the pure electroweak antisymmetric @xmath103 term @xmath104 arises from @xmath105 @xcite .
-channel versus the sub - process collision energy @xmath6 for the top quark pair production up to @xmath106 tev , where we have set the initial renormalization scale @xmath107 gev . ,
scaledwidth=45.0% ] based on the above considerations , the top quark forward - backward asymmetry after pmc scale setting can be written as @xmath108\end{aligned}\ ] ] and @xmath109,\end{aligned}\ ] ] where @xmath110 is total hadronic cross - section up to nlo .
the symbol @xmath111 stands for the asymmetric cross - section of the @xmath7-channel which includes the above mentioned @xmath112 , @xmath113 and @xmath103 terms . here
@xmath49 stands for the pmc scale . in the denominator for the total cross - section up to nlo , for each production channel , we need to introduce two lo pmc scales which are for the coulomb part and non - coulomb part accordingly , and one nlo pmc scale for the non - coulomb part .
are distinct and non - interfering , their pmc scales should be set separately @xcite .
] in the numerator , we only need the nlo pmc scale @xmath114 for the @xmath7-channel , since it is the only asymmetric component .
detailed processes for deriving these pmc scales can be found in ref.@xcite , which are obtained by using the cross - sections calculated within the @xmath15-scheme .
we present the behaviors of the pmc scales for the dominant asymmetric @xmath7-channel in fig.([pmcscale ] ) .
note that if the cross - sections are calculated within any other renormalization scheme , some proper scale - displacements to the present pmc scales will be automatically set by pmc scale setting so as to ensure the scheme - independence of the final estimation .
-channel versus the subprocess collision energy @xmath6 , which determine the dip behavior of the nlo pmc scale @xmath114 .
@xmath115 gev . , scaledwidth=40.0% ] it is interesting to observe that there is a dip for the nlo scale @xmath116 of the @xmath7-channel when @xmath117m_t \sim 563 $ ] gev , which is caused by the correlation among the pmc coefficients for nlo and nnlo terms .
more specifically , it is found that @xmath118 where the coefficients are defined through the standard pmc scale setting @xcite ; i.e. @xmath119_{\rm non - coulomb } & = & a_{0q\bar{q } } a^2_s({\mu^{\rm init}_r } ) + \left [ a_{1q\bar{q } } + b_{1q\bar{q } } n_f \right ] a^3_s({\mu^{\rm init}_r } ) + \left [ a_{2q\bar{q } } + b_{2q\bar{q } } n_f + c_{2q\bar{q } } n^2_f \right ] a^4_s({\mu^{\rm init}_r } ) \nonumber\\ & = & a_{0q\bar{q } } a^2_s({\mu^{\rm pmc , lo}_r } ) + \left[\tilde{a}_{1q\bar{q}}\right ] a^3_s({\mu^{\rm pmc , lo}_r } ) + \left[\tilde{a}_{2q\bar{q } } + \tilde{b}_{2q\bar{q } } n_f \right ] a^4_s({\mu^{\rm pmc , lo}_r } ) \nonumber\\ & = & a_{0q\bar{q } } a^2_s({\mu^{\rm pmc , lo}_r } ) + \left[\tilde{a}_{1q\bar{q}}\right ] a^3_s({\mu^{\rm pmc , nlo}_r } ) + \left[\tilde{\tilde{a}}_{2q\bar{q } } \right ] a^4_s({\mu^{\rm pmc , nlo}_r } ) .\end{aligned}\ ] ] here @xmath120 stands for the partonic cross - section .
as shown in fig.([qqcoe ] ) , the value of @xmath121 is always negative and @xmath122 has a minimum value at small @xmath6 . as a result
, there will be a dip for the nlo pmc scale @xmath114 as shown in fig.([pmcscale ] ) .
quantitatively , the nlo pmc scale @xmath114 for the @xmath7-channel is considerably smaller than @xmath3 in the small @xmath6-region ( corresponding to small momentum fraction of the incident partons which are favored by the parton luminosity @xmath123 @xcite ) .
the nlo cross - section of the @xmath7-channel will thus be greatly increased ; it is a factor of two times larger than its value derived under conventional scale setting , as shown by table [ tab1 ] . as a byproduct
, it is found that if fixing the calculation only at the nlo level , i.e. there is no higher - order terms in the lo pmc scale , so as to absorb all @xmath124-dependent terms properly @xcite . ] and setting the initial renormalization scale to be equal to the factorization sale @xmath125 , our present lo pmc scale @xmath126 for the @xmath7-channel returns to the normal choice which agrees with the qed case , @xmath127 note if @xmath128 , one can apply the renormalization group method to derive the full scale - dependent coefficients @xcite and then get the same result .
the new terms which involve the factor @xmath129 must be separated into two parts : one is proportional to @xmath130 which should be kept in its original form , and the other one is proportional to @xmath131 which should be absorbed into the lower - order @xmath76-terms through the standard pmc scale setting .
the pmc asymmetries @xmath132 and @xmath133 can be compared with the asymmetries calculated using conventional scale setting . for definiteness , we apply pmc scale setting to improve hollik and pagani s results @xcite , and we obtain @xmath134 here @xmath135 stands for the renormalization scale set by conventional scale setting and the symbol hp stands for the corresponding values of ref.@xcite ; i.e. for @xmath136 , it shows @xcite @xmath137 where * @xmath138 and @xmath139 stand for the pure qcd asymmetry at the @xmath140-order under the @xmath0-rest frame and the @xmath19 lab frame respectively . *
@xmath141 and @xmath142 stand for the combined qed and weak with the qcd asymmetry at the @xmath143-order under the @xmath0-rest frame and the @xmath19 lab frame respectively . * @xmath144 and @xmath145 stand for the pure electroweak asymmetry at the @xmath146-order under the @xmath0-rest frame and the @xmath19 lab frame respectively .
-terms at the @xmath147-order of the @xmath7-channel , which are responsible for the smaller effective nlo pmc scale @xmath148 , where the solid circles stand for the light quark loops .
, scaledwidth=45.0% ] in the formulae ( [ pmcr1],[pmcr2 ] ) , we have defined an effective coupling constant @xmath149 for the asymmetric part , which is the weighted average of the strong coupling constant for the @xmath7-channel ; i.e. in using the effective coupling constant @xmath149 , one obtains the same @xmath7-channel nlo cross - section as that of @xmath150-terms or the @xmath151-terms for both the symmetric and asymmetric parts at the nnlo level separately . ] .
it is noted that the nlo - level asymmetric part for @xmath7-channel only involves the nlo pmc scale for the non - coulomb part , so the effective coupling constant @xmath152 can be unambiguously determined .
we obtain a smaller effective nlo pmc scale @xmath153 which corresponds to @xmath154 it is larger than @xmath155 @xcite .
this effective nlo pmc scale is dominated by the non - coulomb @xmath124-terms at the @xmath147-order , which are shown in fig.([nloscale ] ) . in these diagrams ,
the momentum flow in the virtual gluons possess a large range of virtualities .
this effect for nlo pmc scale @xmath156 can be regarded as a weighted average of these different momentum flows in the gluons , so it can be small .
-pair forward - backward asymmetry for the whole phase space .
the upper diagram is for @xmath8 in the @xmath0-rest frame and the lower diagram is for @xmath9 in the laboratory frame .
the hollik and pagani s results ( hp ) @xcite using conventional scale setting are presented for a comparison . the result for
d0 data @xcite shows a similar behavior .
, title="fig:",scaledwidth=45.0% ] -pair forward - backward asymmetry for the whole phase space .
the upper diagram is for @xmath8 in the @xmath0-rest frame and the lower diagram is for @xmath9 in the laboratory frame .
the hollik and pagani s results ( hp ) @xcite using conventional scale setting are presented for a comparison . the result for
d0 data @xcite shows a similar behavior .
, title="fig:",scaledwidth=45.0% ] finally , we obtain @xmath157 thus , after pmc scale setting , the top quark asymmetry under the conventional scale setting is increased by @xmath158 for both the @xmath0-rest frame and the @xmath19-laboratory frame .
this large improvement is explicitly shown in fig.([pmcasy ] ) , where hollik and pagani s results which are derived under conventional scale setting @xcite are presented for comparison . in fig.([pmcasy ] ) , the upper diagram is for @xmath8 in the @xmath0-rest frame and the lower diagram is for @xmath9 in the laboratory frame . and the corresponding cdf data @xcite for the @xmath0-pair forward - backward asymmetry for @xmath160 gev .
the hollik and pagani s results ( hp ) @xcite using conventional scale setting are presented for a comparison .
, scaledwidth=45.0% ] the cdf collaboration has found that when the @xmath0-invariant mass , @xmath160 gev , the top quark forward - backward asymmetry @xmath161 is about @xmath162 standard deviations above the sm asymmetry prediction under the conventional scale setting @xcite . however , after applying pmc scale setting , with the help of the formulae ( [ pmcr1],[pmcr2 ] ) and the cross - sections derived by using conventional scale setting which are listed in ref.@xcite , we will obtain a much larger @xmath163 than the previous estimation @xcite .
for the present case , , which however agrees with that of @xmath0-rest frame within a high accuracy , since as shown in ref.@xcite , the events near the partonic threshold provide the dominant contributions to the cross - section at the tevatron .
] we have @xmath164 pb and @xmath165 with @xmath166 then , we obtain @xmath167 which is increased by about @xmath168 times of the previous one @xmath169 @xcite . our present prediction is only about @xmath14-deviation from the cdf data , which is shown in fig.([asycut ] ) .
we emphasize that the top quark asymmetry calculated under pmc scale setting is almost free of renormalization scale dependence . versus the sub - process collision energy @xmath6 up to @xmath106 tev for the @xmath7-channel , where @xmath170 and @xmath5 respectively . here
@xmath60 gev .
, scaledwidth=45.0% ] to show how the change of initial scale affects the pmc scales , we define the ratio @xmath171 where @xmath172 stands for the pmc scales determined under the condition of @xmath173 . in fig.([ratio ] ) , we show the ratio @xmath174 versus the sub - process collision energy @xmath6 up to @xmath106 tev for the @xmath7-channel , where @xmath170 and @xmath5 respectively .
the residual scale dependence for the pmc scales slightly increases with the subprocess collision energy @xmath6 ; i.e. for the interested non - coulomb nlo pmc scale , the value of @xmath174 is about @xmath175 for @xmath170 and @xmath176 for @xmath177 at @xmath178 tev .
the cross - section at high collision energies is strongly suppressed by the parton luminosities , so that the total cross - section at the tevatron remains almost unchanged even when taking disparate initial scales @xmath2 equal to @xmath3 , @xmath4 , @xmath5 . due to this fact ,
the top quark asymmetry after pmc scale setting is also almost free of initial renormalization scale dependence ; i.e. the residual scale uncertainty is less than @xmath179 by taking @xmath180 , @xmath181 , @xmath182 and @xmath6 respectively .
with the help of present known top quark pair production cross - sections up to nnlo , we have presented a new analysis on the top quark forward - backward asymmetry using pmc scale setting .
after pmc scale setting , a more convergent pqcd series expansion is obtained and the renormalization scale and scheme ambiguities are removed . in comparison to the previous sm values estimated under conventional scale setting ,
we have shown that after pmc scale setting , both the top quark forward - backward asymmetries @xmath8 and @xmath9 for @xmath0-rest frame and @xmath19-laboratory frame can be increased by @xmath158 ; i.e. @xmath183 moreover , the top quark asymmetry with certain kinematical cut , such as @xmath184 , can be raised by about @xmath168 times ; i.e. @xmath185 this shows that , after pmc scale setting , the top quark forward - backward asymmetries are close to the cdf and d0 measurements within only @xmath186-deviation .
the discrepancies between the sm estimate and the present cdf and d0 data are greatly reduced .
this greatly suppresses the parameter space for new physics .
it is clear that the previous large discrepancy between the sm estimation and the cdf and d0 data for the top quark forward - backward asymmetry is caused by the improper setting of the renormalization scale .
the pmc provides a systematic way to obtain optimal renormalization scales for the high energy process , whose theoretical predictions are essentially free of initial renormalization scale dependence even at fixed order .
as we have shown the top quark pair total cross - section and its forward - backward asymmetry are almost unaltered by taking very disparate initial renormalization scales at the nnlo level .
* acknowledgements * : we thank leonardo di giustino , benedict von harling , alexander mitov , michal czakon and andrei kataev for helpful conversations .
this work was supported in part by the program for new century excellent talents in university under grant no.ncet-10-0882 , natural science foundation of china under grant no.11075225 , and the department of energy contract de - ac02 - 76sf00515 .
slac - pub-15006 . | the renormalization scale uncertainty can be eliminated by the principle of maximum conformality ( pmc ) in a systematic scheme - independent way . applying the pmc for the @xmath0-pair hadroproduction at the nnlo level , we have found that the total cross - sections @xmath1 at both the tevatron and lhc remain almost unchanged when taking very disparate initial scales @xmath2 equal to @xmath3 , @xmath4 , @xmath5 and @xmath6 , which is consistent with renormalization group invariance . as an important new application ,
we apply pmc scale setting to study the top quark forward - backward asymmetry .
we observe that the more convergent perturbative series after pmc scale setting leads to a more accurate top quark forward - backward asymmetry .
the resulting pmc prediction on the asymmetry is also free from the initial renormalization scale - dependence . because the nlo pmc scale has a dip behavior for the @xmath7-channel at small subprocess collision energies , the importance of this channel to the asymmetry is increased .
we observe that the asymmetries @xmath8 and @xmath9 at the tevatron will be increased by @xmath10 in comparison to the previous estimates obtained by using conventional scale setting ; i.e. we obtain @xmath11 and @xmath12 .
moreover , we obtain @xmath13 .
these predictions have a @xmath14-deviation from the present cdf and d0 measurements ; the large discrepancies of the top quark forward - backward asymmetry between the standard model estimate and the cdf and d0 data are thus greatly reduced .
+ pacs numbers : : 12.38.aw , 14.65.ha , 11.15.bt , 11.10.gh keywords : : pmc , renormalization scale , top quark forward - backward asymmetry |
the merger of two galaxies results in the formation of a newly assembled galaxy containing a supermassive black hole binary in its nucleus @xcite .
models of hierarchical galaxy formation predict that the most common supermassive black hole binaries arise from minor galactic mergers , and have unequal - mass black holes @xcite .
the tidal interaction between the binary black hole , and the disc of gas and stars it is embedded in , shrinks the binary s orbit .
the hardening efficiency has a complex dependence on the binary s mass ratio @xcite , the gas disc - to - binary mass ratio @xcite , and stellar dynamical interactions @xcite .
provided the binary hardens to sufficiently small separations , the angular momentum extracted from the binary s orbit becomes progressively more dominated by gravitational radiation .
when the gravitational wave torque prevails over the disc torque , the disc s accretion rate becomes much smaller than the binary s hardening rate .
one - dimensional disc models @xcite showed that the density of the disc region located between the primary and the secondary black holes should significantly increase prior to coalescence , as the binary s hardening is dominated by gravitational radiation . in this scenario , the secondary rapidly drains the inner disc onto the primary , acting like a snow - plough .
the increase in the inner disc density would lead to a sudden increase in the disc s bolometric luminosity in the few days prior to merger .
this would constitute an electromagnetic precursor to a supermassive black hole merger , which could be identified in conjunction with the gravitational wave signal detectable with the laser interferometer space antenna ( lisa ) @xcite . in this letter
, we show that this prediction does not hold in two - dimensional ( 2d ) disc models .
we demonstrate with the help of 2d hydrodynamical simulations that when the binary s hardening is dominated by gravitational radiation , the inner disc is progressively funneled to the outer disc through horseshoe trajectories with respect to the secondary , with the consequence that there should be no significant increase in the disc luminosity prior to the merger .
the physical problem we consider is described in [ sec : pb ] , and the numerical setup of our simulations is detailed in
[ sec : setup ] .
our results of calculations are presented in [ sec : results ] , followed by some concluding remarks in [ sec : cl ] .
we examine the tidal interaction between a supermassive black hole binary with mass ratio @xmath0 and the gaseous disc it is embedded in , neglecting for simplicity the presence of other stars .
the torque responsible for the binary s hardening comprises the tidal torque @xmath1 exerted by the gaseous disc , and the torque @xmath2 due to the emission of gravitational waves ( gw ) .
the latter reads ( e.g. , * ? ? ?
* ) @xmath3^{1/2}m_1(m_1+m_2)m_2 ^ 2 \label{gammagw}\ ] ] for a circular binary , where @xmath4 and @xmath5 denote the mass of the primary and of the secondary , respectively , @xmath6 is the binary s semi - major axis , @xmath7 is the speed of light , and @xmath8 is the gravitational constant .
since @xmath9 strongly increases with decreasing @xmath6 , the binary s hardening is ultimately driven by the gravitational wave torque .
the binary s hardening timescale @xmath10 due to the gw torque is @xmath11 further denoting the schwarzschild radius @xmath12 , eq .
( [ taugw1 ] ) can be recast as @xmath13 where @xmath14 is the orbital period at the binary s semi - major axis @xmath6 . at the innermost stable circular orbit ( isco ) ,
located at @xmath15 for a non - rotating primary black hole , @xmath16 and 250 for @xmath17 , @xmath18 and @xmath19 , respectively . in this study , the secondary to primary mass ratio is fixed to @xmath20 , and this large value implies that the secondary is able to open a gap around its orbit even in a relatively thick and viscous disc @xcite .
it also makes the two - dimensional approximation for the disc applicable when discarding gas accretion onto the secondary @xcite .
we point out , as did @xcite , that apart from the presence of the gw torque , the physical problem we consider shares a number of analogies with the orbital evolution of a massive gap - opening planet in a protoplanetary disc .
we examine the evolution of the fluid elements in the inner disc close to the location of the inner separatrix of the secondary s horseshoe region .
these fluid elements are on approximately circulating streamlines with respect to the secondary , with a relative ( synodic ) period @xmath21 half the horseshoe libration period .
the horseshoe libration period reads @xmath22 , where @xmath23 is the secondary s angular frequency , and @xmath24 is the radial half - width of the horseshoe region . for gap - opening secondaries , @xmath25 @xcite , where @xmath26 is the hill radius of the secondary . during a synodic period ,
if the radial distance @xmath27 by which the binary hardens becomes comparable to , or greater than the radial distance @xmath28 by which fluid elements near the inner separatrix drift inward , these fluid elements embark on horseshoe trajectories and are funneled to the outer disc . the value of @xmath28 is essentially set by the viscous torque , and @xmath27 by @xmath29 when the binary shrinkage is dominated by gravitational radiation .
thus , in order of magnitude , @xmath30 when @xmath10 becomes shorter than the viscous drift timescale @xmath31 , where @xmath32 is the disc s turbulent viscosity ( @xmath33 denotes the alpha viscous parameter , h the disc s aspect ratio , and @xmath34 the keplerian angular velocity ) . using eq .
( [ taugw ] ) , the typical binary separation below which @xmath35 , often called decoupling radius , can thus be estimated as @xmath36 where @xmath33 and @xmath37 are to be evaluated at the secondary s location .
detailed modeling of the disc s thermal balance yields a decoupling radius that is a few times larger than that given in eq .
( [ acrit ] ) @xcite .
this illustrates that when the hardening of an unequal - mass binary black hole is dominated by gravitational radiation , a substantial funneling of the inner disc to the outer disc is possible beyond the isco location , which we confirm with 2d hydrodynamical simulations in
[ sec : results ] .
we investigate the tidal interaction between a black hole binary and the gas disc it is embedded in , assuming the binary s hardening is dominated by the emission of gravitational waves . for this purpose ,
2d hydrodynamical simulations were carried out with the fargo code ( * ? ? ?
* ` http://fargo.in2p3.fr ` ) .
a cylindrical coordinate system @xmath38 centred onto the primary black hole is adopted .
the disc extends from @xmath39 to @xmath40 ( that is , @xmath41 r_{\rm isco}$ ] with assuming a non - rotating primary ) .
+ _ binary parameters _ the mass of the primary and secondary black holes is @xmath42 and @xmath43 , respectively , as in @xcite .
the initial semi - major axis of the binary is @xmath44 pc ( or , @xmath45 ) .
the binary is assumed to be circular , and its orbital plane to be aligned with the disc .
each black hole is treated as a point mass potential .
relativistic effects are discarded for simplicity and to facilitate comparison with previous studies .
similarly , gas accretion onto the secondary is neglected . also , to avoid a large accumulation of mass inside its hill radius ,
the secondary s gravitational potential is smoothed over a softening length @xmath46 , where @xmath47 is the pressure scale height ( its value is specified below ) .
+ _ disc parameters _ the disc is initially axisymmetric and rotates at the angular frequency @xmath48 about the primary .
the expression for @xmath48 assumes a radial equilibrium between the centrifugal acceleration , the radial acceleration due to the pressure gradient , and the gravitational acceleration due to the primary alone .
the initial gas density is @xmath49 g@xmath50 , which corresponds to an initial mass @xmath51 . for simplicity , a locally isothermal equation of state
is considered where the initial radial profile of the disc temperature , which we take proportional to @xmath52 , remains stationary .
the disc temperature is conveniently related to the disc s aspect ratio @xmath53 , which we take to be @xmath54 .
the disc turbulence is modeled with a constant kinematic viscosity @xmath55 that corresponds to an alpha parameter @xmath56 uniformly throughout the disc .
+ _ numerical setup _ the grid has @xmath57 cells along the radial direction , and it covers the full @xmath58 range in azimuth with @xmath59 cells . a logarithmic spacing
is used along the radial direction to optimize the grid s resolution near the disc s inner edge .
the secondary s hill radius is resolved by about 30 grid cells along each direction at all times .
standard zero - gradient outflow boundary conditions are applied at the grid s inner and outer edges .
our simulations were performed in two steps . the secondary is first held on a fixed circular orbit about the binary s center of mass .
its mass is gradually increased over 100 orbits to avoid a violent disc relaxation following the introduction of the secondary .
the secondary progressively depletes a gap around its orbit , which attains a quasi steady - state density profile @xmath60 400 orbits after the insertion of the secondary .
simulations were then restarted with the binary feeling the gravitational wave torque only .
this second stage , which we refer to as the hardening stage in what follows , lasts @xmath61 orbits , as indicated by eq .
( [ taugw ] ) .
the disc torque on the secondary is discarded for simplicity , as @xmath62 , where we compare the time evolution of the binary s semi - major axis when only the gw torque is included ( black curve ) , and when only the disc torque is included ( red curve ) .
[ fig : fig2 ] shows a time sequence of the gas surface density during the hardening stage .
time is expressed in orbital periods at the binary s initial separation , @xmath63 .
streamlines in the frame rotating with the secondary are shown as white curves .
the circulating streamline of the inner disc closest to the secondary is depicted by a red curve .
this time sequence clearly shows no density increase in the inner disc , but underlines instead the progressive replenishment of the gap .
the gas density inside the gap becomes more and more asymmetric as the binary s hardening gets faster , with more gas behind the secondary ( @xmath64 ) than ahead of it .
this asymmetry results from fluid elements of the inner disc embarking onto horseshoe streamlines and being funneled to the outer disc or to the gap .
this gap asymmetry is reminiscent of the one occurring during type iii runaway migration ( a fast , generally inward migration regime driven by the disc torque ; see * ? ? ?
[ fig : fig3 ] displays the azimuthally - averaged disc density at different times prior to the binary s merger .
it further illustrates that the rapid shrinkage of the binary does not squeeze the inner disc in our 2d simulations , in contrast to previous height - integrated 1d disc models @xcite . to gain further insight into the time evolution of the mass in the inner disc
, we have performed simulations in which a small patch of the disc inside the inner separatrix of the horseshoe region is polluted initially with a passive contaminant that is advected with the flow .
we denote by @xmath65 the concentration per unit area of the contaminant , and evolve the equation @xmath66 writing @xmath67 , where @xmath68 is referred to as the specific concentration of the contaminant , and using the continuity equation , eq .
( [ eqn : scalar ] ) can be recast as @xmath69 .
the patch of contaminant is introduced 340 orbits after the beginning of the hardening stage , and the time evolution of its specific concentration , @xmath70 , is displayed as a time sequence in fig . [
fig : fig4 ] .
time increases by @xmath71 orbits moving from left to right through the panels .
the location of the secondary is shown by a vertical dashed line in each panel . in the left - hand panel ,
most of the contaminant is concentrated inside the inner separatrix of the horseshoe region , located at @xmath72 . at this time , a ( relatively ) tiny fraction of the contaminant is being funneled to the outer disc , and ends up sliding along the outer separatrix of the horseshoe region at @xmath73 .
the second panel highlights that more and more contaminant is being redirected to the outer disc . as the binary continues hardening , most of the contaminant now moving in the outer disc remains in the outer disc , only a small fraction of it undergoes inward horseshoe u - turns back to the inner disc . in the third panel , most of the contaminant is now concentrated in the outer disc .
the radial profile of the contaminant gets progressively thicker in the outer disc , because the outer separatrix of the horseshoe region is moving inwards faster than the contaminant may drift inward viscously .
this is even more clear in the fourth panel of fig .
[ fig : fig4 ] , where the contaminant in the outer disc is left well behind the secondary .
we also mention that in addition to the material passing to the outer disc , we never observe the contaminant being pushed inward : the region below @xmath74 remains dark . the time evolution of the gas surface density in figs .
[ fig : fig2 ] and [ fig : fig3 ] , and of the contaminant s specific concentration in fig .
[ fig : fig4 ] , suggest that most of the mass in the inner disc relative to the secondary is transferred to the outer disc through horseshoe streamlines . to help quantify this funneling mechanism , we display in fig . [
fig : fig5 ] the time evolution of the mass in the inner disc , in units of the binary mass , during the last @xmath75 orbits before the binary s merger .
the increase in the mass of the outer disc during the same time span ( also in units of the binary mass ) is overplotted by a red curve .
the difference between both curves accounts for the mass lost from the computational grid , which is shown by a dashed curve in fig .
[ fig : fig5 ] .
these curves clearly show that the fraction of the inner disc s mass funneled to the outer disc largely exceeds that drained onto the primary .
the minor merger of two galaxies leads to the formation of a supermassive black hole binary with unequal mass ratio embedded in a gaseous disc .
the tidal interaction between the binary and the disc hardens the binary s orbit . when the binary gets sufficiently tight , emission of gravitational waves becomes the main source of angular momentum extraction from the binary s orbit , and causes further rapid shrinkage until coalescence takes place .
we have focused in this letter on the evolution of the disc region located between the primary and the secondary black holes ( the inner disc ) , when the binary s hardening is dominated by the emission of gravitational waves . with the help of 2d hydrodynamical simulations
, we have shown that the rapid hardening of the binary does not lead to a significant squeezing of the inner disc .
the latter is redirected instead toward the disc region beyond the secondary s orbit ( the outer disc ) through horseshoe streamlines . when the binary s hardening timescale driven by gravitational radiation becomes shorter than the disc s viscous drift timescale , fluid elements in the inner disc embark on horseshoe trajectories with respect to the secondary , and are progressively funneled to the outer disc .
the funneling of the inner disc toward the outer disc implies that , in contrast to the predictions of 1d disc models , the accretion rate onto the primary is not dramatically increased just prior to merger , and , as a result , the disc emission before the binary merger should remain at about the same level .
after the merger , an electromagnetic afterglow could be detected as the disc gets accreted by the merged black hole @xcite .
the physical model we have considered is a straightforward 2d extension of the model considered in @xcite , @xcite , and @xcite , and as was already pointed out by these authors , this model has a number of simplifying assumptions .
we have considered an intermediate binary mass ratio ( @xmath20 ) . a smaller mass ratio would decrease the binary s separation below which a significant funneling occurs ( which we have checked with additional simulations , not reported here ) .
a mass ratio closer to unity would initially lead to a rapid depletion of the inner disc , and the binary would be surrounded by a circumbinary disc ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , making our funneling mechanism not applicable . also , different values of the disc aspect ratio and viscosity would change the structure of the gap opened by the secondary ( width , depth ) .
partial gap - opening could occur in very thick discs , and a three - dimensional disc modeling would be valuable . in some cases , the disc secondary interaction can make the outer disc eccentric , thereby driving the binary s eccentricity ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , although the latter should be partially damped during the fast inspiral driven by gravitational radiation @xcite .
still it remains to be clarified how the funneling mechanism would operate in the presence of an eccentric binary .
we thank p. chang , j. guilet , z. haiman , d. n. c. lin , k. menou , and e. quataert for stimulating discussions , and the referee for useful comments .
cb is supported by a herchel smith postdoctoral fellowship .
er - r acknowledges support from the david and lucille packard foundation and the nsf grant : ast-0847563 . | we present two - dimensional hydrodynamical simulations of the tidal interaction between a supermassive black hole binary with moderate mass ratio , and the fossil gas disc where it is embedded .
our study extends previous one - dimensional height - integrated disc models , which predicted that the density of the gas disc between the primary and the secondary black holes should rise significantly during the ultimate stages of the binary s hardening driven by the gravitational radiation torque .
this snow - plough mechanism , as we call it , would lead to an increase in the bolometric luminosity of the system prior to the binary merger , which could be detected in conjunction with the gravitational wave signal .
we argue here that the snow - plough mechanism is unlikely to occur . in two - dimensions ,
when the binary s hardening timescale driven by gravitational radiation becomes shorter than the disc s viscous drift timescale , fluid elements in the inner disc get funneled to the outer disc through horseshoe trajectories with respect to the secondary .
mass leakage across the secondary s gap is thus found to be effective and , as a result , the predicted accretion disc luminosity will remain at roughly the same level prior to merger .
[ firstpage ] accretion , accretion discs black hole physics gravitational waves hydrodynamics methods : numerical |
material interfaces in hybrid structures give rise to proximity effects , whereby the properties of one material can `` leak '' into the adjacent material , creating a region with properties derived from both materials . in superconductor / ferromagnet ( sf ) hybrid structures @xcite ,
the proximity effect causes superconducting correlations to penetrate into the ferromagnetic region and vice versa .
these correlations typically decay over short distances , which in diffusive systems is of the order @xmath2 , where @xmath3 is the diffusion coefficient of the ferromagnet and @xmath4 is the strength of the exchange field .
however , for certain field configurations , the singlet correlations from the superconductor may be converted into so - called long - range triplets ( lrts ) @xcite .
these triplet components have spin projection parallel to the exchange field , and decay over much longer distances .
this results in physical quantities like supercurrents decaying over the length scale @xmath5 , which is usually much larger than the ferromagnetic coherence length @xmath6 , where @xmath7 is the temperature .
this distance is independent of @xmath4 , and at low temperatures it becomes increasingly large , which allows the condensate to penetrate deep into the ferromagnet .
the isolation and enhancement of this feature has attracted much attention in recent years as it gives rise to novel physics and possible low - temperature applications by merging spintronics and superconductivity @xcite .
it is by now well - known that the conversion from singlet to long - range triplet components of the superconducting state can happen in the presence of magnetic inhomogeneities @xcite , _ i.e. _ a spatially varying exchange field , and until recently such inhomogeneities were believed to be the primary source of this conversion @xcite , although other proposals using _
e.g. _ non - equilibrium distribution functions and intrinsic triplet superconductors also exist @xcite .
however , it has recently been established that another possible source of lrt correlations is the presence of a finite spin - orbit ( so ) coupling , either in the superconducting region @xcite or on the ferromagnetic side @xcite .
in fact , it can be shown that an sf structure where the magnetic inhomogeneity is due to a bloch domain wall , as considered in _
e.g. _ refs .
, is gauge equivalent to one where the ferromagnet has a homogeneous exchange field and intrinsic so coupling@xcite .
it is known that so scattering can be caused by impurities@xcite , but this type of scattering results in purely isotropic spin - relaxation , and so does not permit the desired singlet - lrt conversion . to achieve such a conversion , one needs a rotation of the spin pair into the direction of the exchange field @xcite .
this can be achieved by using materials with an intrinsic so coupling , either due to the crystal structure in the case of noncentrosymmetric materials@xcite , or due to interfaces in thin - film hybrids@xcite , where the latter also modifies the fundamental process of andreev reflection @xcite .
the role of so coupling with respect to the supercurrent in ballistic hybrid structures has also been studied recently @xcite . in this paper
, we establish how the presence of spin - orbit coupling in sf structures manifests in two important experimental observables : the density of states @xmath8 probed via tunneling spectroscopy ( or conductance measurements ) , and the critical temperature @xmath1 .
a common consequence for both of these quantities is that neither becomes independent of the magnetization direction .
this is in contrast to the case without so coupling in conventional monodomain ferromagnets , where the results are invariant with respect to rotations of the magnetic exchange field .
this symmetry is now lifted due to so coupling : depending on the magnetization direction , lrt cooper pairs are created in the system which leave clear fingerprints both spectroscopically and in terms of the @xmath1 behavior . on the technical side
, we will present in this work for the first time a riccati parametrization of the usadel equation and its corresponding boundary conditions that include so coupling .
this is an important advance in terms of exploring the full physics of triplet pairing due to so coupling as it allows for a solution of the quasiclassical equations without any assumption of a weak proximity effect .
we will also demonstrate that the so coupling can actually protect the singlet superconducting correlations even in the presence of a strong exchange field , leading to the possibility of a minigap that is magnetically tunable via the orientation of the exchange field .
the remainder of the article will be organised as follows : in section [ sec : theory ] , we introduce the relevant theory and notation , starting from the quasiclassical usadel equation , which describes the diffusion of the superconducting condensate into the ferromagnet .
we also motivate our choice of intrinsic so coupling in this section , and propose a new notation for describing rashba
dresselhaus couplings .
the section goes on to discuss key analytic features of the equations in the limit of weak proximity , symmetries of the density of states at zero energy , and analytical results needed to calculate the critical temperature of hybrid systems .
we then present detailed numerical results in section [ sec : results ] : we analyze the density of states of an sf bilayer in [ subsec : sf ] [ see fig .
[ fig : model](a ) ] , with the case of pure rashba coupling considered in section [ subsec : purer ] , and we study the sfs josephson junction in [ subsec : sfs ] [ see fig . [
fig : model](b ) ] .
we consider different orientations and strengths of the exchange field and so coupling , and in the case of the josephson junction , the effect of altering the phase difference between the condensates .
then , in section [ sec : results - tc ] , we continue our treatment of the sf bilayer in the full proximity regime by including a self - consistent solution in the superconducting layer , and focus on how the presence of so coupling affects the critical temperature of the system .
we discover that the so coupling allows for spin - valve functionality with a single ferromagnetic layer , meaning that rotating the magnetic field by @xmath9 induces a large change in @xmath1 .
finally , we conclude in section [ sec : disc ] with a summary of the main results , a discussion of some additional consequences of the choices made in - text , as well as possibilities for further work .
the diffusion of the superconducting condensate into the ferromagnet can be described by the usadel equation , which is a second - order partial differential equation for the green s function of the system@xcite .
together with appropriate boundary conditions , the usadel equation establishes a system of coupled differential equations that can be solved in one dimension .
we will consider the case of diffusive equilibrium , where the retarded component @xmath10 of the green s function is sufficient to describe the behaviour of the system @xcite .
we start by examining the superconducting correlations in the ferromagnet , and use the standard bardeen cooper
schrieffer ( bcs ) bulk solution for the superconductors .
in particular , we will clarify the spectroscopic consequences of having so coupling in the ferromagnetic layer . in the absence of so coupling
, the usadel equation @xcite in the ferromagnet reads @xmath11=0,\end{aligned}\ ] ] where the matrix @xmath12 , and @xmath13 is the quasiparticle energy .
the magnetization matrix @xmath14 in the above equation is @xmath15 where @xmath16 is the ferromagnetic exchange field , @xmath17 denotes complex conjugation , @xmath18 is the pauli vector , and @xmath19 are the usual pauli matrices . the corresponding kupriyanov
lukichev boundary conditions are@xcite @xmath20 \ , , \end{aligned}\ ] ] where the subscripts refer to the different regions of the hybrid structure ; in the case of an sf bilayer as depicted in fig .
[ fig : model](a ) , @xmath21 denotes the superconductor , and @xmath22 the ferromagnet , while @xmath23 denotes the derivative along the junction @xmath24 .
the respective lengths of the materials are denoted @xmath25 , and the interface parameters @xmath26 describe the ratio of the barrier resistance @xmath27 to the bulk resistance @xmath28 of each material . , [ subsec : purer ] and [ sec : results - tc ] .
we take the thin - film layering direction along the @xmath29-axis , and assume an @xmath30-plane rashba dresselhaus coupling in the ferromagnetic layer .
( b ) the sfs trilayer in [ subsec : sfs ] . ]
we will use the riccati parameterisation@xcite for the quasiclassical green s function @xmath10 , @xmath31 where the normalisation matrices are @xmath32 and @xmath33 .
the tilde operation denotes a combination of complex conjugation @xmath34 and energy @xmath35 , with @xmath36 , @xmath37 .
the riccati parameterisation is particularly useful for numerical computation because the parameters are bounded @xmath38 $ ] , contrary to the multi - valued @xmath39-parameterisation @xcite . in practice
, this means that for certain parameter choices the numerical routines will only converge in the riccati formulation .
appendix [ app : riccati ] contains some further details on this parameterisation . to include intrinsic so coupling in the usadel equation
, we simply have to replace all the derivatives in eq . with their gauge covariant counterparts:@xcite @xmath40 \ , .\end{aligned}\
] ] this is valid for any so coupling linear in momentum .
we consider the leading contribution ; higher order terms , _
e.g. _ those responsible for the su(2 ) lorentz force , are neglected here . such higher order terms are required to produce so - called @xmath41 junctions
which have attracted interest of late@xcite , and consequently we will see no signature of the @xmath41 effect in the systems considered herein .
the object @xmath42 has both a vector structure in geometric space , and a @xmath43 matrix structure in spin nambu space , and can be written as @xmath44 in terms of the so field @xmath45 , which will be discussed in more detail in the next subsection .
so coupling in the context of quasiclassical theory has also been discussed in refs . .
when we include the so coupling as shown above , we derive the following form for the usadel equation ( see appendix [ app : riccati ] ) : @xmath46\nonumber\\ & & \,+2id_f\left[(\partial_k \gamma)\tilde{n}({a}^*_k+\tilde{\gamma}{a}_k\gamma)+({a}_k+\gamma { a}^*_k\tilde{\gamma})n(\partial_k \gamma)\right]\!,\end{aligned}\ ] ] where the index @xmath47 indicates an arbitrary choice of direction in cartesian coordinates .
the corresponding equation for @xmath48 is found by taking the tilde conjugate of eq .
( [ eqn : sousadel ] ) .
similarly , the boundary conditions in eq . become : @xmath49 and the @xmath48 counterparts are found in the same way as before . for the details of these derivations ,
see appendix [ app : riccati ] .
we will now discuss the definition of current in the presence of spin - orbit interactions .
since the hamiltonian including so coupling contains terms linear in momentum ( see below ) , the velocity operator @xmath50 is affected .
we stated above that the kupriyanov - lukichev boundary conditions are simply modified by replacing the derivative with its gauge covariant counterpart including the so interaction .
to make sure that current conservation is still satisfied , we must carefully examine the usadel equation . in the absence of so coupling , the quasiclassical expression for electric current
is given by @xmath51 where @xmath52 is the @xmath53 green s function matrix in keldysh space @xmath54 and @xmath55 is a constant that is not important for this discussion .
current conservation can now be proven from the usadel equation itself .
we show this for the case of equilibrium , which is relevant for the case of supercurrents in josephson junctions . in this case
@xmath56 and we get @xmath57 performing the operation tr@xmath58 on the usadel equation , we obtain @xmath59\ } = 0.\end{aligned}\ ] ] now , inserting the most general definition of the green s function @xmath10 , one finds that the second term in the above equation is always zero .
thus , we are left with @xmath60 which expresses precisely current conservation since the same analysis can be done for @xmath61 .
now , let us include the so coupling .
the current should then be given by @xmath62 so that the expression for the charge current is modified by the presence of so coupling , as is known .
the question is now if this current is conserved , as it has to be physically
. we can prove that it is from the usadel equation by rewriting it as @xmath63 + d[{\underline{a}},[{\underline{a}},\hat{g}^r ] ] - { i}[\varepsilon\rho_3 + \hat{m},\hat{g}^r ] \,,\end{aligned}\ ] ] and then performing the operation tr@xmath58 , one finds : @xmath64 so we recover the standard current conservation law @xmath65 .
the precise form of the generic so field @xmath66 is imposed by the experimental requirements and limitations .
as the name suggests , spin - orbit coupling couples a particle s spin with its motion , and more specifically its momentum .
as mentioned in the introduction , the so coupling in solids can originate from a lack of inversion symmetry in the crystal structure .
such spin - orbit coupling can be of both rashba and dresselhaus type and is determined by the point group symmetry of the crystal @xcite .
it is also known that the lack of inversion symmetry due to surfaces , either in the form of interfaces to other materials or to vacuum , will give rise to antisymmetric spin - orbit coupling of the rashba type . for sufficiently thin structures ,
the so coupling generated in this way can permeate the entire structure , but the question of precisely how far into adjacent materials such surface - so coupling may penetrate appears to be an open question in general .
intrinsic inversion asymmetry arises naturally due to interfaces between materials in thin - film hybrid structures such as the ones considered herein .
noncentrosymmetric crystalline structures provide an alternative source for intrinsic asymmetry , and are considered in ref . .
in thin - film hybrids , the rashba spin splitting derives from the cross product of the pauli vector @xmath67 with the momentum @xmath68 , @xmath69 where @xmath70 is called the rashba coefficient , and we have chosen a coordinate system with @xmath71 as the layering direction
. another well - known type of so coupling is the dresselhaus spin splitting , which can occur when the crystal structure lacks an inversion centre .
for a two - dimensional electron gas ( quantum well ) confined in the @xmath72-direction , then to first order @xmath73 , so the dresselhaus splitting becomes @xmath74 where @xmath75 is called the dresselhaus coefficient . in our structure , we consider a thin - film geometry with the confinement being strongest in the @xmath29-direction .
although there may certainly be other terms contributing to the dresselhaus so coupling in such a structure , since real thin - film structures will have three - dimensional quasiparticle diffusion and we use a @xmath76 form of the so coupling here , we consider the standard form eq .
( [ eq : dhaus ] ) as an approximation that captures the main physics in the problem .
this is a commonly used model in the literature to explore the effects originating from so coupling in a system .
when we combine both interactions , we obtain the hamiltonian for a general rashba
dresselhaus so coupling , @xmath77 in this work , we will restrict ourselves to this form of so coupling .
it should be noted that our setup may also be viewed as a simplified model for a scenario where the so coupling and ferromagnetism exist in separate , thin layers , in which case we expect qualitatively similar results to the ones reported in this manuscript . as explained in ref .
, the so coupling acts as a background su(2 ) field , _ i.e. _ an object with both a vector structure in geometric space , and a @xmath78 matrix structure in spin space .
we can therefore identify the interaction above with an effective vector potential @xmath66 which we will call the _ so field _
, @xmath79 from which we derive that @xmath80 at this point , it is convenient to introduce a new notation for describing rashba dresselhaus couplings , which will let us distinguish between the physical effects that derive from the strength of the coupling , and those that derive from the geometry . for this purpose ,
we employ polar notation defined by the relations @xmath81 where we will refer to @xmath82 as the _ so strength _ , and @xmath83 as the _
so angle_. rewritten in the polar notation , eq
. takes the form : @xmath84 from the definition , we can immediately conclude that @xmath85 corresponds to a pure dresselhaus coupling , while @xmath86 results in a pure rashba coupling , with the geometric interpretation of @xmath83 illustrated in fig . [
fig : soc - geometry ] . note that @xmath87 , which means that @xmath88 .
another useful property is that we can switch the components @xmath89 by letting @xmath90 .
( 2.5,2.0 ) node[above ] @xmath91 ( 0.0,0.0 ) ( 3.2,0.0 ) node[right ] @xmath92 ; ( 1.0,0.0 ) arc ( 0:20:1.0 ) node[right ] @xmath83 [ - > ] arc ( 20:35:1.0 ) ; ( 4.5,+2.5 ) node[above ] @xmath93
( 4.5,-0.7 )
( 6.5,+1.8 ) node[above ] @xmath94 ; ( 4.5,0.3 ) arc ( 90:70:1.0 ) node[above ] @xmath83 [ - > ] arc ( 70:55:1.0 ) ; the appearance of lrts in the system depends on the interplay between so coupling and the direction of the exchange field .
recall that the lrt components are defined as having spin projections parallel to the exchange field , as opposed to the short - ranged triplet ( srt ) component which appears as long as there is exchange splitting@xcite but has spin projection perpendicular to the field and is therefore subject to the same pair - breaking effect as the singlets@xcite , penetrating only a very short distance into strong ferromagnets .
thus if we have an so field component along the layering direction , _
e.g. _ if we had @xmath95 in figs .
[ fig : model](a ) and [ fig : model](b ) , achievable with a noncentrosymmetric crystal or in a nanowire setup , then a non - vanishing commutator @xmath96}$ ] creates the lrt .
however , we will from now only consider systems where @xmath97 , in which case the criterion for lrt is @xcite that @xmath98}]}$ ] must not be parallel to the exchange field @xmath99 . expanding , we have @xmath100 } ] } & = 4a^2 ( \,{\underline{h } } \cdot { \underline{\sigma}}\ , + h_z{\sigma_{z } } ) \nonumber\\ & \;- 4a^2 ( h_x{\sigma_{y } } + h_y { \sigma_{x } } ) \sin 2\chi \ ; , \label{eqn : cond}\end{aligned}\ ] ] from which it is clear that no lrts can be generated for a pure dresselhaus coupling @xmath85 or rashba coupling @xmath101 when the exchange field is in - plane .
however , the effect of so coupling becomes increasingly significant for angles close to @xmath102 ( see fig .
[ fig : dosangle ] in section [ subsec : sf ] ) .
we also see that no lrts can be generated for in - plane magnetization in the special case @xmath103 and @xmath104 , since @xmath105 can then be rewritten as @xmath106 , which is parallel to @xmath107 .
there is no lrt generation for the case @xmath108 and @xmath109 for similar reasons . in general
however , the lrt will appear for an in - plane magnetization as long as @xmath110 and the so coupling is not of pure dresselhaus or pure rashba type .
it is also important to note that the lrt can be created even for pure rashba type so coupling if the magnetization has both in- and out - of - plane magnetization components .
we will discuss precisely this situation in sec .
[ subsec : purer ] .
once the condition for long - range triplet generation is satisfied , increasing the corresponding exchange field will also increase the proportion of long - range triplets compared with short - range triplets .
whether or not the presence of long - range triplets can be observed in the system , _
i.e. _ if they retain a clear signature in measurable quantities such as the density of states when the criteria for their existence is fulfilled , depends on other aspects such as the strength of the spin - orbit coupling and will be discussed later in this paper .
_ thus , a main motivation for this work is to take a step further than discussing their existence @xcite and instead make predictions for when long - ranged triplet cooper pairs can actually be observed via spectroscopic or @xmath1 measurements in sf structures with spin - orbit coupling .
_ however , we will also demonstrate that the presence of so coupling offers additional opportunities besides the creation of lrt cooper pairs .
we will show both analytically and numerically that the so coupling can protect the singlet component even in the presence of an exchange field , which normally would suppress it .
this provides the possibility of tuning the well - known minigap _ magnetically _ , both in bilayer and josephson junctions , simply by altering the direction of the magnetization . in order to establish a better analytical understanding of the role played by so coupling in the system before presenting the spectroscopy and @xmath1 results
, we will now consider the limit of weak proximity effect , which means that @xmath111 , @xmath112 in the ferromagnet .
the anomalous green s function in general is given by the upper - right block of eq .
( [ eqn : gr ] ) , @xmath113 , which we see reduces to @xmath114 in this limit .
it will also prove prudent to express the anomalous green s function using a singlet / triplet decomposition , where the singlet component is described by a scalar function @xmath115 , and the triplet components encapsulated in the so - called @xmath116-vector@xcite , @xmath117 combining the above with the weak proximity identity @xmath114 , we see that the components of @xmath118 can be rewritten as : @xmath119 under spin rotations , the singlet component @xmath115 will then transform as a scalar , while the triplet component @xmath120 transforms as an ordinary vector .
another useful feature of this notation is that it becomes almost trivial to distinguish between short - range and long - range triplet components ; the projection @xmath121 along the exchange field corresponds to the srts , while the perpendicular part @xmath122 describes the lrts , where @xmath123 here denotes the unit vector of the exchange field . for a concrete example
, if the exchange field is oriented along the @xmath29-axis , then @xmath124 will be the short - range component , while both @xmath125 and @xmath126 are long - ranged components . in the coming sections
, we will demonstrate that the lrt component can be identified from its density of states signature , as measurable by tunneling spectroscopy . in the limit of weak proximity effect
, we may linearize both the usadel equation and kupriyanov lukichev boundary conditions . using the singlet / triplet decomposition in eq . , and the rashba dresselhaus coupling in eq .
, the linearized version of the usadel equation can be written : @xmath127 where we for brevity have defined an so interaction matrix @xmath128 we have now condensed the usadel equation down to two coupled differential equations for @xmath115 and @xmath129 , where the coupling is proportional to the exchange field and the so interaction term .
the latter has been written as a product of a factor @xmath130 , depending on the strength @xmath82 , and a factor @xmath131 , depending on the angle @xmath83 in the polar notation .
the matrix @xmath132 becomes diagonal for a dresselhaus coupling with @xmath85 or a rashba coupling with @xmath101 , which implies that there is no triplet mixing for such systems .
in contrast , the off - diagonal terms are maximal for @xmath133 , which suggests that the triplet mixing is maximal when the rashba and dresselhaus coefficients have the same magnitude .
in addition to the off - diagonal triplet mixing terms , we see that the diagonal terms of @xmath132 essentially result in imaginary energy contributions @xmath130 .
as we will see later , this can in some cases result in a suppression of all the triplet components in the ferromagnet
. we will now consider exchange fields in the @xmath30-plane , @xmath134 since the linearized usadel equations show that the presence of a singlet component @xmath115 only results in the generation of triplet components along @xmath107 , and the so interaction term only mixes the triplet components in the @xmath30-plane , the only nonzero triplet components will in this case be @xmath125 and @xmath126 .
the srt amplitude @xmath135 and lrt amplitude @xmath136 can therefore be written : @xmath137 by projecting the linearized usadel equation for @xmath129 along the unit vectors @xmath138 and @xmath139 , respectively , then we obtain coupled equations for the srts and lrts : @xmath140 \ , d_{{\mkern3mu\vphantom{\perp}\vrule depth 0pt\mkern2mu\vrule depth 0pt\mkern3mu}}\notag\\ & - 2id_fa^2 \cos 2\theta \ , \sin 2\chi \;d_\perp + hf_s \ ; , \label{eq : usadel - lin - pll}\\ \frac{i}{2 } d_f\partial^2_z d_\perp = \ , & [ \epsilon + 2id_fa^2 ( 1+\sin 2\theta \
, \sin 2\chi ) ] \ , d_\perp \notag\\ & - 2id_fa^2 \cos 2\theta \ , \sin 2\chi
\;d_{{\mkern3mu\vphantom{\perp}\vrule depth 0pt\mkern2mu\vrule depth 0pt\mkern3mu}}\ ; .
\label{eq : usadel - lin - perp}\end{aligned}\ ] ] these equations clearly show the interplay between the singlet component @xmath115 , srt component @xmath135 , and lrt component @xmath136 .
if we start with only a singlet component @xmath115 , then the presence of an exchange field @xmath4 results in the generation of the srt component @xmath135 .
the presence of an so field can then result in the generation of the lrt component @xmath136 , where the mixing term is proportional to @xmath141 .
this implies that in the weak proximity limit , lrt mixing is absent for an exchange field direction @xmath142 , corresponding to @xmath143 , while it is maximized if @xmath144 and at the same time @xmath133 . in other words ,
the requirement for maximal lrt mixing is therefore that the exchange field is aligned along either the @xmath145-axis or @xmath146-axis , while the rashba and dresselhaus coefficients should have the same magnitude .
it is important to note here that although the mixing between the triplet components is maximal at @xmath144 , this does not necessarily mean that the signature of the triplets in physical quantities is most clearly seen for these angles , as we shall discuss in detail later .
moreover , these equations show another interesting consequence of having an so field in the ferromagnet , which is unrelated to the lrt generation .
note that the effective quasiparticle energies coupling to the srts and lrts become @xmath147 when @xmath148 , then the srts are entirely unaffected by the presence of so coupling ; the triplet mixing term vanishes for these parameters , and @xmath149 is also clearly independent of @xmath82 .
however , when @xmath150 , the situation is drastically different .
there is still no possibility for lrt generation , however the srt energy @xmath151 will now obtain an imaginary energy contribution which destabilizes the srts .
in fact , numerical simulations show that this energy shift destroys the srt components as @xmath82 increases . as we will see in section [ sec : results - tc ] , this effect results in an increase in the critical temperature of the bilayer .
thus , switching between @xmath152 in a system with @xmath153 may suggest a novel method for creating a triplet spin valve .
when @xmath154 but @xmath155 , the triplet mixing term proportional to @xmath156 will no longer vanish , so we get lrt generation in the system .
we can then see from the effective triplet energies that as @xmath157 , the imaginary part of @xmath149 vanishes , while the imaginary part of @xmath158 increases .
this leads to a relative increase in the amount of srts compared to the amount of lrts in the system .
in contrast , as @xmath159 , the imaginary part of @xmath158 vanishes , and the imaginary part of @xmath149 increases .
this means that we would expect a larger lrt generation for these parameters , up until the point where the triplet mixing term @xmath156 becomes so small that almost no lrts are generated at all .
the ratio of effective energies coupling to the triplet component at the fermi level @xmath160 can be written as @xmath161 the density of states @xmath8 containing all spin components can be written in terms of the riccati matrices as @xmath162 /2 \ ; , \ ] ] which for the case of zero energy can be written concisely in terms of the singlet component @xmath115 and triplet components @xmath129 , @xmath163 the singlet and triplet components are therefore directly competing to lower and raise the density of states @xcite . furthermore , since we are primarily interested in the proximity effect in the ferromagnetic film , we will begin by using the known bcs bulk solution in the superconductor , @xmath164 where @xmath165 , and @xmath166 is the superconducting phase . using eq . and the definition of the tilde operation , and comparing @xmath10 in eq . with its standard expression in a bulk superconductor eq .
, we can see that at zero energy the singlet component @xmath167 must be purely imaginary and the asymmetric triplet @xmath168 must be purely real if the superconducting phase is @xmath169 . by inspection of eq .
, we can see that a transformation @xmath170 along with @xmath171 leaves the equations invariant .
the density of states will therefore be unaffected by such permutations , @xmath172 = d[{\underline{h}}=(b , a,0 ) ] \ ; , \ ] ] while in general @xmath173 \neq d[\underline{h}=(b,0,a ) ] \ ; .\ ] ] however , whenever one component of the planar field is exactly twice the value of the other component , one can confirm that the linearized equations remain invariant under a rotation of the exchange field @xmath174 with associated invariance in the density of states .
when superconducting correlations leak from a superconductor and into a ferromagnet in a hybrid structure , there will also be an inverse effect , where the ferromagnet effectively drains the superconductor of its superconducting properties due to tunneling of cooper pairs . physically , this effect is observable in the form of a reduction in the superconducting gap @xmath175 near the interface at all temperatures .
furthermore , if the temperature of the hybrid structure is somewhat close to the bulk critical temperature @xmath176 of the superconductor , this _ inverse proximity effect _ can be strong enough to make the superconducting correlations vanish entirely throughout the system .
thus , proximity - coupled hybrid structures will in practice always have a critical temperature @xmath1 that is lower than the critical temperature @xmath176 of a bulk superconductor .
depending on the exact parameters of the hybrid system , @xmath1 can sometimes be significantly smaller than @xmath176 , and in some cases it may even vanish ( @xmath177 ) . to quantify this effect , it is no longer sufficient to solve the usadel equation in the ferromagnet only
. we will now also have to solve the usadel equation in the superconductor , @xmath178 along with a self - consistency equation for the gap @xmath175 , @xmath179 where @xmath180 is the density of states per spin at the fermi level , and @xmath181 is the electron - electron coupling constant in the bcs theory of superconductivity . for a derivation of the gap equation ,
see appendix [ app : gap ] .
to study the effects of the so coupling on the critical temperature of an sf structure , we therefore have to find a self - consistent solution to eq . in the ferromagnet , eq . at the interface , and eqs . and in the superconductor .
in practice , this is done by successively solving one of the equations at a time numerically , and continuing the procedure until the system converges towards a self - consistent solution . to obtain accurate results
, we typically have to solve the usadel equation for 100150 positions in each material , around 500 energies in the range @xmath182 , and 100 more energies in the range @xmath183 , where the debye cutoff @xmath184 for the superconductors considered herein .
this procedure will then have to be repeated up to several hundred times before we obtain a self - consistent solution for any given temperature of the system .
furthermore , if we perform a conventional linear search for the critical temperature @xmath185 in the range @xmath186 with a precision of 0.0001 , it may require up to 10,000 such iterations to complete , which may take several days depending on the available hardware and efficiency of the implementation .
the speed of this procedure may , however , be significantly increased by performing a binary search instead . using this strategy
, the critical temperature can be determined to a precision of @xmath187 after only 12 iterations , which is a significant improvement .
the convergence can be further accelerated by exploiting the fact that @xmath175 from iteration to iteration should decrease monotonically to zero if @xmath188 ; however , the details will not be further discussed in this paper .
we consider the proximity effect in an sf bilayer in [ subsec : sf ] , using the bcs bulk solution for the superconductors . the case of pure rashba coupling
is discussed in [ subsec : purer ] , and the sfs josephson junction is treated in [ subsec : sfs ] . we take the thin - film layering direction to be oriented in the @xmath29-direction and fix the spin - orbit coupling to rashba dresselhaus type in the @xmath30-plane as given by eq . .
we set @xmath189 .
the coherence length for a diffusive bulk superconductor typically lies in the range @xmath190 nm .
we solve the equations using matlab with the boundary value differential equation package ` bvp6c ` and examine the density of states @xmath8 for energies normalised to the superconducting gap @xmath191 . for brevity of notation
, we include the normalization factor in the coefficients @xmath70 and @xmath75 in these sections .
this normalization is taken to be the length of the ferromagnetic region @xmath192 , so that for instance @xmath193 in the figure legends means @xmath194 .
finally , in section [ sec : results - tc ] , we calculate the dependence of the critical temperature of an sf bilayer as a function of the different system parameters . consider the sf bilayer depicted in fig .
[ fig : model](a ) . in section [ subsec : theory_so ]
we introduced the conditions for the lrt component to appear , and from eq .
( [ eqn : cond ] ) it is clear that no lrts will be generated if the exchange field is aligned with the layering direction , _
i.e. _ @xmath195 , since eq .
( [ eqn : cond ] ) will be parallel to the exchange field .
conversely , the general condition for lrt generation with in - plane magnetisation is both that @xmath110 and that the so coupling is not of pure rashba or pure dresselhaus form .
however , it became clear in section [ subsec : wpe ] that the triplet mixing was maximal for equal rashba and dresselhaus coupling strengths , and in fact the spectroscopic signature is quite sensitive to deviations from this . in ref .
, the density of states for an sf bilayer was shown to display oscillatory behavior as a function of distance penetrated into the ferromagnet .
the physical origin of this stems from the non - monotonic dependence of the superconducting order parameter inside the f layer , which oscillates and leads to an alternation of dominant singlet and dominant triplet correlations as a function of distance from the interface . when the triplet ones dominate , the proximity - induced change in the density of states is inverted compared to sn structures , giving rise to an enhancement of the density of states at low - energies in this so - called @xmath196-phase where the proximity - induced superconducting order parameter is negative . for sf bilayers without so coupling and a homogeneous exchange field ,
one expects to see a spectroscopic minigap whenever the thouless energy is much greater than the strength of the exchange field .
the minigap in sf structures closes when the resonant condition @xmath197 is fulfilled , where @xmath198 is the minigap occuring without an exchange field , and a zero - energy peak emerges instead@xcite .
the minigap @xmath198 depends on both the thouless energy and the resistance of the junction . for stronger fields we will have an essentially featureless density of states ( see _ e.g. _ ref . and references therein ) .
this is indeed what we observe for @xmath199 in fig .
[ tab : sfhz ] . with purely out - of - plane magnetisation @xmath195 , the effect of
so coupling is irrespective of type : rashba , dresselhaus or both will always create a minigap . with in - plane magnetisation
however , the observation of a minigap above the so - free resonant condition @xmath200 indicates that dominant rashba or dominant dresselhaus coupling is present .
the same is true for sfs trilayers , and thus to observe a signature of long - range triplets the rashba
dresselhaus coefficients must be similar in magnitude , and in the following we shall primarily focus on this regime . to clarify quantitatively
how much the rashba and dresselhaus coefficients can deviate from each other before destroying the low - energy enhancement of the density of states , which is the signature of triplet cooper pairs in this system , we have plotted in fig .
[ fig : zepbilayer ] the density of states at the fermi level ( @xmath201 ) as a function of the spin - orbit angle @xmath83 and the magnetization direction @xmath39 . for purely rashba or dresselhaus coupling @xmath202 ,
the deviation from the normal - state value is small .
however , as soon as both components are present a highly non - monotonic behavior is observed .
this is particularly pronounced for @xmath203 , although the conversion from dominant triplets to dominant singlets as one rotates the field by changing @xmath39 is seen to occur even away from @xmath204 . with either @xmath205 , or equivalently @xmath206 ,
lrts are generated provided @xmath207 , and in fig .
[ tab : sfhy ] we can see that the addition of so coupling introduces a peak in the density of states at zero energy , which saturates for a certain coupling strength .
this peak manifests as sharper around @xmath201 than the zero - energy peak associated with weak field strengths of the order of the gap ( _ i.e. _ as evident from @xmath199 in fig .
[ tab : sfhy ] ) , which occurs regardless of magnetisation direction or texture@xcite . by analysing the real components of the triplets , for a gauge where the superconducting phase is zero
, we can confirm that this zero - energy peak is due to the lrt component , in this case @xmath125 , also depicted in fig .
[ tab : sfhy ] , in agreement with the predictions for _ textured _ magnetisation _ without _ so coupling @xcite .
however , it is also evident from fig .
[ tab : sfhy ] that increasing the field strength rapidly suppresses the density of states towards that of the normal metal , making the effect more difficult to detect experimentally .
the way to ameliorate this situation is to remember that the introduction of so coupling means the _ direction _ of the exchange field is crucially important , as we see in fig .
[ fig : dosangle ] , and this allows for a dramatic spectroscopic signature for fields without full alignment with the @xmath145- or @xmath146-axes . as a function of the spin - orbit angle @xmath83 and magnetization angle @xmath39 .
we have used a ferromagnet of length @xmath189 with an exchange field @xmath208 and a spin - orbit magnitude @xmath209 . ]
[ fig : dosangle ] shows how the density of states at zero energy varies with the angle @xmath39 between @xmath210 and @xmath211 at zero energy ; with @xmath212 the field is aligned with @xmath210 , and with @xmath213 it is aligned with @xmath211 .
we see that the inclusion of so coupling introduces a nonmonotonic angular dependance in the density of states , with increasingly sharp features as the so coupling strength increases , although the optimal angle at approximately @xmath214 and @xmath215 varies minimally with increasing so coupling .
clearly the ability to extract maximum lrt conversion from the inclusion of so coupling is highly sensitive to the rotation angle , with near step - function behaviour delineating the regions of optimal peak in the density of states and an energy gap for strong so coupling .
it is remarkable to see how @xmath216 vs. @xmath39 formally bears a strong resemblance to the evolution of a fully gapped bcs @xcite density of states @xmath217 vs. @xmath218 to a flat density of states as the so coupling decreases .
these results can again be explained physically by the linearized equations . since the case @xmath219 corresponds to @xmath220 in the notation developed in the preceding sections , eq .
implies that @xmath221 when @xmath222 , while @xmath223 when @xmath224 . in other words , for negative @xmath39 ,
the so coupling suppresses the lrt components , and the exchange field suppresses the other components . since the singlet and srt components have opposite sign in eq .
, this renders the density of states essentially featureless .
however , for positive @xmath39 , both the so coupling and the exchange field suppress the srt components , meaning that lrt generation is energetically favoured . note that @xmath225 as @xmath226 , which explains why the lrt generation is maximized in this regime . since the triplet mixing term in eq .
is proportional to @xmath227 , the lrt component vanishes when the value of @xmath39 gets too close to @xmath228 .
furthermore , since @xmath149 has a large imaginary energy contribution in this case , the srts are also suppressed at @xmath229 .
thus , despite lrts being most energetically favored at this exact point , we end up with a system dominated by singlets due to the srt suppression and lack of lrt production pathway .
nevertheless , one would conventionally expect that exchange fields of a magnitude @xmath230 as depicted in fig.[fig : dosangle ] would suppress any features in the density of states , while we observe an obvious minigap .
_ thus , the singlet correlations become much more resilient against the pair - breaking effect of the exchange field when spin - orbit coupling is present .
_ to identify the physical origin of this effect , we solve the linearized equations along with their corresponding boundary conditions for the specific case @xmath201 , @xmath231 .
we consider a bulk superconductor occupying the space @xmath232 while the ferromagnet length @xmath192 is so large that one in practice only needs to keep the decaying parts of the anomalous green s function .
we then find the following expression for the singlet component at the sf interface in the absence of so coupling : @xmath233 with increasing @xmath4 , the singlet correlations are suppressed in the conventional manner .
however , we now incorporate so coupling in the problem . for more transparent analytical results ,
we focus on the case @xmath234 .
this condition can be rewritten as @xmath235 . in this case , a similar calculation gives the singlet component at the sf interface in the presence of so coupling : @xmath236 clearly , the so coupling enhances the singlet component in spite the presence of an exchange field since @xmath237 .
this explains the presence of the conventional zero energy gap for large so coupling even with a strong exchange field .
a consequence of this observation is that so coupling in fact provides a route to a _ magnetically tunable minigap_. fig .
[ fig : dosangle ] shows that when both an exchange field and so coupling is present , the direction of the field determines when a minigap appears .
this holds even for strong exchange fields @xmath0 as long as the so coupling is sufficiently large as well .
we recall that the lrt cooper pairs , defined as the components of @xmath238 perpendicular to @xmath239 , may be characterized by a quantity @xmath136 which is defined by the cross product of the two vectors : @xmath240 .
we saw above that the spectroscopic signature of lrt generation is strongly dependent on the angle of the field , and this angle is a tunable parameter for sufficiently weak magnetic anisotropy . in fig .
[ tab : sfhxhy ] we see an example of the effect this rotation can have on the spectroscopic signature of lrt generation : when the exchange field is changed from @xmath241 , _
i.e. _ changing the direction of the field , we see that a strong zero - energy peak emerges due to the presence of lrt in the system .
this large peak emerges despite the stronger exchange field that would ordinarily reduce the density of states towards the normal state , _
i.e. _ as in fig
. [ tab : sfhy ] for @xmath242 .
if one were to remove the so coupling , the low - energy density of states would thus have no trace of any superconducting proximity effect , which demonstrates the important role played by the so interactions here .
finally , for completeness we include an example of the effect of rotating the field to have a component along the junction in fig .
[ tab : sfhyhz ] .
comparing the case of @xmath243 in fig .
[ tab : sfhyhz ] with @xmath244 in fig .
[ tab : sfhxhy ] , we see that the two cases are identical , as predicted in the limit of weak proximity effect , and increasing the magnitude of the out - of - plane @xmath29 component of the field has no effect on the height of the zero - energy peak , which is instead governed by the in - plane @xmath146 component . between the @xmath145 and @xmath146 components of the magnetisation exchange field @xmath245 for increasing so coupling . as the strength of the so coupling increases we see increasingly sharp variations in the density of states from an optimal peak at around @xmath246 and @xmath247 to a gap around @xmath142 . ] for the sf bilayer with energies normalised to the superconducting gap @xmath191 and so coupling normalised to the inverse ferromagnet length @xmath248 .
the table shows the spectroscopic effect of increasing so coupling with @xmath219 when the magnetisation @xmath249 , _
i.e. _ with the field perpendicular to the interface , and the effect of increasing difference between the rashba and dresselhaus coefficients for both @xmath249 and @xmath250 .
although the conditions for lrt generation are fulfilled in the latter case , it is clear that no spectroscopic signature of this is present . ] for the sf bilayer with energies normalised to the superconducting gap @xmath191 and so coupling normalised to the inverse ferromagnet length @xmath248 .
the table shows the spectroscopic effect of equal rashba
dresselhaus coefficients when the magnetisation is oriented entirely in the @xmath146-direction , and also the correlation between the so - induced zero - energy peak with the long - range triplet component @xmath251 .
it is clear that the predominant effect of the lrt component , which appears only when the so coupling is included , is to increase the peak at zero energies .
increasing the field strength rapidly suppresses the density of states towards that of the normal metal . ] in the sf bilayer for energies normalised to the superconducting gap @xmath191 and so coupling normalised to the inverse ferromagnet length @xmath248 .
the table shows the spectroscopic features of the sf bilayer with rotated exchange field in the @xmath30-plane .
again we see a peak in the density of states at zero energy due to the lrt component , _ i.e. _ the component of @xmath238 perpendicular to @xmath239 , @xmath136 .
the height of this zero - energy peak is strongly dependent on the angle of the field vector in the plane , as shown in fig .
[ fig : dosangle ] . for near - optimal field orientations increasing the
so coupling leads to a dramatic increase in the peak of the density of states at zero energy . ] in the sf bilayer for energies normalised to the superconducting gap @xmath191 and so coupling normalised to the inverse ferromagnet length @xmath248 .
the table shows the spectroscopic features of the sf bilayer with a rotated exchange field in the @xmath252-plane .
note that when the field component along the junction is twice the component in the @xmath146-direction , here @xmath253 , the density of states is equivalent to the case @xmath244 illustrated in fig .
[ tab : sfhxhy ] , as predicted in the limit of weak proximity effect . ]
there exists another experimentally viable setup where the lrt can be created . in the case where pure rashba so coupling is present
, originating e.g. from interfacial asymmetry , the condition for the existence of lrt is that the exchange field has a component both in - plane and out - of - plane .
although the lrt formally is non - zero , it is desirable to clarify if and how it can be detected through spectroscopic signatures . from an experimental point of view , it is known that pdni and cuni @xcite can in general feature a canted magnetization orientation relative to the film - plane due to the competition between shape anisotropy and magnetocrystalline anisotropy .
this is precisely the situation required in order to have an exchange field with both an in - plane ( @xmath30-plane in our notation ) and out - of - plane ( @xmath29-direction ) component . in our model ,
the ferromagnetism coexists with the rashba so coupling , which may be taken as a simplified model of two separate layers where the so coupling is induced e.g. by a very thin heavy metal and pdni or cuni is deposited on top of it . to determine how the low - energy density of states is influenced by the triplet pairing , we plot in fig .
[ fig : purer](a ) @xmath216 as a function of the misalignment angle @xmath254 between the film - plane and its perpendicular axis [ see inset of fig . [
fig : purer](b ) for junction geometry ] . in order to correlate the spectroscopic features with the lrt , we plot in fig .
[ fig : purer](b ) the lrt green s function @xmath255 .
it is clear that the lrt vanishes when @xmath256 or @xmath257 .
this is consistent with the fact that for pure rashba coupling , purely in - plane or out - of - plane direction of the exchange field gives @xmath258 according to our previous analysis .
however , for @xmath259 the lrt exists .
its influence on @xmath216 is seen in fig .
[ fig : purer](a ) : an enhancement of the zero - energy density of states . for any particular
set of junction parameters there is an optimal value of the so coupling , and in approaching this value the density of states is correlated with re@xmath260 . beyond this optimal value , they are anticorrelated , as evident from fig .
[ fig : purer ] as the so coupling increases , but the angular correlation remains .
we note that the magnitude of the enhancement of the density of states is substantially smaller than what we obtained with both rashba and dresselhaus coupling . at the same time
, the magnitude of the enhancement is of precisely the same order as previous experimental works that have measured the density of states in s / f structures @xcite .
note that it is only the angle between the plane and the tunneling direction which is of importance : the density of states is invariant under a rotation in the film - plane of the exchange field .
the so - induced enhancement of the zero - energy density of states reaches an optimal peak before further increases in the magnitude of the rashba coupling results in a suppression of both the short- and long - ranged triplet components , causing the low - energy density of states enhancement to vanish .
the correlation with the lrt component @xmath255 correspondingly changes to anticorrelation , evident in fig .
[ fig : purer ] .
nevertheless , the strong angular variation with @xmath216 remains although @xmath261 for all @xmath254 [ see inset of fig . [
fig : purer](a ) ] . increasing the exchange field
@xmath4 further suppressed the proximity effect overall .
the main effect of the so coupling is that @xmath216 depends on the exchange field direction . as seen for the case of @xmath262 in fig .
[ fig : purer](a ) , there is no directional dependence without so coupling .
thus , depending on the exchange field angle between the in - plane and out - of plane direction , measuring an enhanced @xmath216 at low - energies is a signature of the presence of lrt cooper pairs in the ferromagnet . more generally , measuring a dependence on the exchange field direction @xmath254
would be a direct consequence of the presence of so coupling in the system , even in the regime of _
e.g. _ moderate to strong rashba coupling where the triplets are suppressed . in an s / f structure with pure rashba spin - orbit coupling .
we have set @xmath263 and @xmath264 .
inset : stronger so coupling @xmath265 , demonstrating that the angular variation of @xmath216 remains , although the enhancement due to triplets is absent .
( b ) plot of the magnitude of the lrt anomalous green s function @xmath255 at @xmath201 .
as seen , its enhancement correlates with an accompanying increase in the density of states for the same angle @xmath254 , and beyond an optimal so coupling value there is anticorrelation between the density of states peak and @xmath255 .
the only angle of importance is the angle @xmath254 between the out - of - plane and in - plane component of the exchange field , shown in the inset . ] by adding a superconducting region to the right interface of the sf bilayer we form an sfs josephson junction .
it is well known that the phase difference between the superconducting regions governs how much current can flow through the junction@xcite , and the density of states for a diffusive sns junction has been measured experimentally with extremely high precision@xcite . here
we consider such a transversal junction structure as depicted in fig.[fig : model](b ) , again with intrinsic so coupling in the @xmath30-plane ( eq . [ eqn : a ] ) in the ferromagnet and with bcs bulk values for each superconductor . in [ subsec : junc1 ] we consider single orientations along the principal axes of the system @xmath266 of the uniform exchange field and in [ subsec : juncrot ] we consider a rotated field .
experimentally , the density of states can be probed at the superconductor / ferromagnet interface if one of the superconductors is a superconducting island , and the scanning tunneling microscope approaches from the top , next to this superconductor island .
let us first recapitulate some known results .
we saw in section [ sec : theory ] that the spin - singlet , srt and lrt components compete to raise and lower the density of states at low energies .
their relative magnitude is affected by the magnitude and direction of both the exchange field and so coupling and results in three distinctive qualitative profiles : the zero - energy peak from the lrts , the singlet - dominated regime with a minigap , and the flat , featureless profile in the absence of superconducting correlations . in the josephson junction ,
the spectroscopic features are in addition sensitive to the phase difference @xmath166 between the superconductors . in junctions with an interstitial normal metal ,
the gap decreases as @xmath267 , closing entirely at @xmath268 such that the density of states is that of the isolated normal metal ; identically one@xcite . without an exchange field
the density of states is unaffected by the so coupling .
this is because without an exchange field the equations governing the singlet and triplet components are decoupled and thus no singlet - triplet conversion can occur . from a symmetry point of view
, it is reasonable that the time - reversal invariant spin - orbit coupling does not alter the singlet correlations . without so coupling and
as long as the exchange field is not too large , changing the phase difference can qualitatively alter the density of states from minigap to peak at zero energy ( see fig .
[ tab : sfs1 ] ) , a useful feature permitting external control of the quasiparticle current flowing through the junction .
the underlying reason is that the phase difference controls the relative ratio of the singlet and triplet correlations : when the singlets dominate , a minigap is induced which mirrors their origin in the bulk superconductor . as in the bilayer case , there is a resonant condition@xciteindicating an exchange field strength beyond which the minigap can no longer be sustained and increasing the phase difference simply lowers the density of states towards that of the normal metal . amongst the features we outline in the following subsections , one of the effects of adding
so coupling is to make this useful gap - to - peak effect accessible with stronger exchange fields , _
i.e. _ for a greater range of materials .
at the same time , the so coupling can not be _ too strong _ since the triplet correlations are suppressed in this regime leaving only the minigap and destroying the capability for qualitative change in the spectroscopic features .
consider first the case in which the exchange field is aligned in a single direction , meaning that we only consider an exchange field purely along the principal @xmath269 axes of the system .
if we again restrict the form of the so - vector to ( [ eqn : a ] ) , aligning @xmath239 in the @xmath29-direction will not result in any lrts . in this case
the spectroscopic effect of the so coupling is dictated by the singlet and short - range triplet features , much as in the sf bilayer case ( fig .
[ tab : sfhz ] ) .
this is demonstrated in fig .
[ tab : sfs1 ] , where again we see a qualitative change in the density of states as the exchange field increases , with the regions of minigap and zero - energy - peak separated by the resonant condition @xmath197 without so coupling .
we will now examine the effect of increasing the exchange field aligned in the @xmath145- or , equivalently , the @xmath146-direction . in this case , we have generation of lrt cooper pairs .
if @xmath239 is sufficiently weak to sustain a gap independently of so coupling , introducing weak so coupling will increase the gap at zero phase difference while maintaining a peak at zero energy for a phase difference of @xmath270 ( see fig .
[ tab : sfs1 ] ) . increasing the so coupling increases this peak at zero energy up to a saturation point .
as the exchange field increases sufficiently beyond the resonant condition to keep the gap closed , increasing the so coupling increases the zero - energy peak at all phases , again due to the lrt component , eventually also reaching a saturation point .
as the phase difference @xmath271 , the density of states reduces towards that of the normal metal , closing entirely at @xmath268 as expected@xcite .
as the value of the density of states at zero energy saturates for increasing so coupling , fixed phase differences yield the same drop at zero energy regardless of the strength of so coupling .
we note in passing that when the so coupling field has a component along the junction direction @xmath272 , it can qualitatively influence the nature of the superconducting proximity effect . as very recently shown in ref . , a giant triplet proximity effect develops at @xmath268 in this case , in complete contrast to the normal scenario of a vanishing proximity effect in @xmath196-biased junctions . with two components of the field @xmath239 ,
e.g. _ from rotation , it is again useful to separate the cases with and without a component along the junction direction .
when the exchange field lies in - plane ( the @xmath30-plane ) , and provided we satisfy the conditions @xmath110 and @xmath207 , increasing the so coupling drastically increases the zero energy peak as shown in fig .
[ tab : sfsrot ] , again due to the lrt component .
this is consistent with the bilayer behavior , where the maximal generation of lrt cooper pairs occurs at an angle @xmath273 .
as the phase difference approaches @xmath196 , the proximity - induced features are suppressed in the centre of the junction .
this can be understood intuitively as a consequence of the order parameter averaging to zero since it is positive in one superconductor and negative in the other . in the sfs junction with increasing
so coupling and exchange field in a single direction , with @xmath8 normalised to the superconducting gap @xmath191 and so coupling normalised to the inverse ferromagnet length @xmath248 . with no so coupling and very weak exchange field
we see a phase - dictated gap - to - peak qualitative change in the density of states at zero energy .
when the field is strong enough to destroy this gap , _ i.e. _ above the resonant condition , increasing the phase difference simply lowers the density of states towards that of the normal metal , which is achieved at a phase difference of @xmath268 . with the addition of so coupling
we see a clear difference in the density of states due to the long range triplet component , which is present when the field is oriented in @xmath146 but not in @xmath29 . when lrts are present with weak exchange fields , a phase - dictated gap - to - peak feature is retained and increased as the strength of so coupling increases the gap , with the peak shown here at a phase difference of @xmath270 . for stronger exchange fields , increasing
the so coupling produces the minigap when there is no lrt component , whereas the existence of an lrt component again introduces an increasing peak at zero energy when no minigap is present . ] in the sfs junction for energies normalised to the superconducting gap @xmath191 and so coupling normalised to the inverse ferromagnet length @xmath248 .
the table shows the spectroscopic effects of increasing so coupling in sfs with rotated exchange field . in the absence of so coupling ,
the density of states is flat and featureless at low energies .
increasing the so coupling again leads to a strong increase in the peak of the density of states at zero energy , while increasing the phase difference reduces the peak and shifts the density of states weight toward the gap edge for higher so coupling strengths . with a component of the field in the junction direction a qualitative change in the density of states from strongly suppressed to enhanced at zero energy can be achieved by altering the phase difference between the superconductors .
this change can occur in the presence of stronger exchange fields when so coupling is included .
increasing the exchange field destroys the ability to maintain a gap in the density of states and the lrt component of the so coupling increases the zero - energy peak as it did in the bilayer case . ]
the 2d plots in this paper of the local density of states are given for the centre of the junction ( @xmath274 ) , where one naturally expects the relative proportion of lrts to be greatest .
however , it is interesting to note that the large peak at zero energy the signature of the lrts is maintained throughout the ferromagnet .
this is shown in fig .
[ fig : sfsdos_spatial ] , for the case @xmath275 and @xmath276 , where the maximal peak for @xmath169 is almost twice the normal - state value . in comparison ,
the depletion of this peak is surprisingly small at the superconductor interfaces . throughout the ferromagnet of an sfs junction with phase difference @xmath277 , spin - orbit coupling @xmath275 and magnetisation @xmath276 . ] with one component of the exchange field along the junction and another along either @xmath145 or @xmath146 , a phase - dictated gap - to - peak transition at zero energy is possible with stronger fields than with the field aligned in a single direction , as shown in fig .
[ tab : sfsrot ] .
notice that in this case increasing the phase difference @xmath278 gives an increase in the peak at zero energy before reducing towards the normal metal state . for higher field strengths we find once again that increasing the so coupling increases the peak at zero energy , up to a system - specific threshold , and increasing phase difference reduces the density of states towards that of the normal metal .
it is also useful to consider how the zero - energy density of states depends simultaneously on the phase - difference and magnetization orientation . to this end
, we show in fig .
[ fig : zepjosephson ] a contour plot of the density of states at the fermi level @xmath279 as a function of the superconducting phase difference @xmath166 across the junction and the magnetization direction @xmath39 .
the proximity effect vanishes in the centre of the junction at @xmath268 for any value of the exchange field orientation , giving the normal - state value . just as in the bilayer case ( fig .
[ fig : zepbilayer ] ) , we see that the proximity effect is strongly suppressed for the range of angles @xmath224 . when rotating the field in the opposite direction , @xmath280 , strongly non - monotonic behavior emerges . for zero phase - difference ,
the physics is qualitatively similar to the bilayer situation . in this case
, we proved analytically that the lrt is not produced at all when @xmath281 .
accordingly , fig .
[ fig : zepjosephson ] shows a full minigap there . whether or not a clear zero - energy peak can be seen due to the lrt depends on the relative strength of the rashba and dresselhaus coupling . in the top panel
, we have dominant dresselhaus coupling in which case the low - energy density of states show either normal - state behavior or a minigap .
interestingly , we see that the same opportunity appears in the present case of a josephson setup as in the bilayer case : a magnetically tunable minigap appears .
this effect exists as long as the phase difference is not too close to @xmath196 , in which case the minigap closes . in the bottom panel corresponding to equal magnitude of rashba and dresselhaus ,
however , a strong zero - energy enhancement due to long - range triplets emerges as one moves away from @xmath282 . with increasing phase difference ,
the singlets are seen to be more strongly suppressed than the triplet correlations since the minigap region ( dark blue ) vanishes shortly after @xmath283 while the peaks due to triplets remain for larger phase differences . [ cols="^,^ " , ] in this section , we present numerical results for the critical temperature @xmath1 of an sf bilayer .
the theory behind these investigations is summarized in section [ sec : theory - tc ] , and discussed in more detail in appendix [ app : gap ] .
an overview of the physical system is given in fig . [
fig : model](a ) . in all of the simulations we performed , we used the material parameter @xmath284 for the superconductor , the exchange field @xmath285 for the ferromagnet , and the interface parameter @xmath286 for both materials .
the other physical parameters are expressed in a dimensionless form , with lengths measured relative to the superconducting correlation length @xmath287 , energies measured relative to the bulk zero - temperature gap @xmath288 , and temperatures measured relative to the bulk critical temperature @xmath176 .
this includes the so coupling strength @xmath82 , which is expressed in the dimensionless form @xmath289 .
the plots presented in this subsection were generated from 1236 data points per curve , where each data point has a numerical precision of 0.0001 in @xmath185 .
the results were smoothed with a loess algorithm . before we present the results with so coupling
, we will briefly investigate the effects of the ferromagnet length @xmath192 and superconductor length @xmath290 on the critical temperature , in order to identify the interesting parameter regimes .
the critical temperature as a function of the size of the superconductor is shown in fig .
[ fig : tc - ds ] . as a function of the length @xmath291 of the superconductor for @xmath292 . below a critical length @xmath290
, superconductivity can no longer be sustained and @xmath1 becomes zero . for larger thicknesses of the superconducting layer , @xmath1 reverts back to its bulk value.,title="fig : " ] as a function of the length @xmath291 of the superconductor for @xmath292 . below a critical length @xmath290
, superconductivity can no longer be sustained and @xmath1 becomes zero . for larger thicknesses of the superconducting layer , @xmath1 reverts back to its bulk value.,title="fig:",width=321 ] first of all , we see that the critical temperature drops to zero when @xmath293 .
this observation is hardly surprising ; since the superconducting correlation length is @xmath287 , the critical temperature is rapidly suppressed once the length of the junction goes below @xmath287 . after this , the critical temperature increases quickly , already reaching nearly 50% of the bulk value when @xmath294 , demonstrating that the superconductivity of the system is clearly very sensitive to small changes in parameters for this region .
the next step is then to observe how the behaviour of the system varies with the size of the ferromagnet , and these results are presented in fig .
[ fig : tc - df ] . as a function of the ferromagnet length @xmath295 for @xmath292 . increasing the thickness of the ferromagnet
gradually suppresses the @xmath1 of the superconductor , causing a stronger inverse proximity effect .
, title="fig : " ] as a function of the ferromagnet length @xmath295 for @xmath292 . increasing the thickness of the ferromagnet
gradually suppresses the @xmath1 of the superconductor , causing a stronger inverse proximity effect .
, title="fig:",width=321 ] we again observe that the critical temperature increases with the size of the superconductor , and decreases with the size of the ferromagnet . the critical temperature for a superconductor with @xmath296 drops to zero at @xmath297 , and
stays that way as the size of the ferromagnet increases .
thus we do not observe any strongly nonmonotonic behaviour , such as reentrant superconductivity , for our choice of parameters .
this is consistent with the results of fominov _ et al .
_ , who only reported such behaviour for systems where either the interface parameter or the exchange field is drastically smaller than for the bilayers considered herein@xcite .
we now turn to the effects of the antisymmetric so coupling on the critical temperature , which has not been studied before .
[ fig : tc - chi - z - long ] and [ fig : tc - chi - z - short ] show plots of the critical temperature as a function of the so angle @xmath83 for an exchange field in the @xmath29-direction .
the critical temperature is here independent of the so angle @xmath83 .
this result is reasonable , since the so coupling is in the @xmath30-plane , which is perpendicular to the exchange field for this geometry .
we also observe a noticeable increase in critical temperature for larger values of @xmath82 .
this behaviour can be explained using the linearized usadel equation . according to eq .
, the effective energy @xmath298 coupling to the triplet component in the @xmath29-direction becomes @xmath299 so in other words , the srts obtain an imaginary energy shift proportional to @xmath300 . however , as shown in eq . , there is no corresponding shift in the energy of the singlet component .
this effect reduces the triplet components relative to the singlet component in the ferromagnet , and as the triplet proximity channel is suppressed the critical temperature becomes restored to higher values . as a function of the so angle @xmath83 ,
when @xmath301 , @xmath302 , and @xmath303 .
increasing the so coupling causes @xmath1 to move closer to its bulk value , since the triplet proximity effect channel becomes suppressed.,title="fig : " ] as a function of the so angle @xmath83 , when @xmath301 , @xmath302 , and @xmath303 .
increasing the so coupling causes @xmath1 to move closer to its bulk value , since the triplet proximity effect channel becomes suppressed.,title="fig:",width=321 ] as a function of the so angle @xmath83 , when @xmath304 , @xmath302 , and @xmath303.,title="fig : " ] as a function of the so angle @xmath83 , when @xmath304 , @xmath302 , and @xmath303.,title="fig:",width=321 ] the same situation for an exchange field along the @xmath145-axis is shown in figs .
[ fig : tc - chi - x - long ] and [ fig : tc - chi - x - short ] . for this geometry
, we observe a somewhat smaller critical temperature for all @xmath305 and all @xmath83 compared to figs .
[ fig : tc - chi - z - long ] and [ fig : tc - chi - z - short ] .
this can again be explained by considering the linearized usadel equation in the ferromagnet , which suggests that the effective energy @xmath306 coupling to the @xmath145-component of the triplet vector should be @xmath307 which has a smaller imaginary part than the corresponding equation for @xmath298 .
furthermore , note the drop in critical temperature as @xmath308 . since the linearized equations contain a triplet mixing term proportional to @xmath309 , which is maximal precisely when @xmath154 , these are also the geometries for which we expect a maximal lrt generation .
thus , this decrease in critical temperature near @xmath133 can be explained by a net conversion of singlet components to lrts in the system , which has an adverse effect on the singlet amplitude in the superconductor , and therefore the critical temperature . as a function of the so angle @xmath83 , when @xmath301 , @xmath302 , and @xmath310 .
the critical temperature depends on the relative weight of the rashba and dresselhaus coefficients.,title="fig : " ] as a function of the so angle @xmath83 , when @xmath301 , @xmath302 , and @xmath310 .
the critical temperature depends on the relative weight of the rashba and dresselhaus coefficients.,title="fig:",width=321 ] as a function of the so angle @xmath83 , when @xmath304 , @xmath302 , and @xmath310 .
, title="fig : " ] as a function of the so angle @xmath83 , when @xmath304 , @xmath302 , and @xmath310 . , title="fig:",width=321 ] in figs .
[ fig : tc - theta - long ] and [ fig : tc - theta - short ] we present the results for a varying exchange field @xmath311 in the @xmath30-plane . in this case
, we observe particularly interesting behaviour : the critical temperature has extrema at @xmath312 , where the extremum is a maximum if @xmath39 and @xmath83 have the same sign , and a minimum if they have opposite signs . since @xmath313 is precisely the geometries for which we do not expect any lrt generation , triplet mixing can not be the source of this behaviour . for the choice of physical parameters chosen in fig . [
fig : tc - theta - short ] , this effect results in a difference between the minimal and maximal critical temperature of nearly @xmath314% as the magnetization direction is varied . as shown in fig .
[ fig : tc - theta - long ] , the effect persists qualitatively in larger structures as well , but is then weaker . as a function of the exchange field angle @xmath39 ,
when @xmath301 , @xmath302 , and @xmath209 .
in contrast to ferromagnets without so coupling , @xmath1 now depends strongly on the magnetization direction .
this gives rise to a spin - valve like functionality with a single ferromagnet featuring so coupling.,title="fig : " ] as a function of the exchange field angle @xmath39 , when @xmath301 , @xmath302 , and @xmath209 .
in contrast to ferromagnets without so coupling , @xmath1 now depends strongly on the magnetization direction .
this gives rise to a spin - valve like functionality with a single ferromagnet featuring so coupling.,title="fig:",width=321 ] as a function of the exchange field angle @xmath39 , when @xmath304 , @xmath302 , and @xmath209.,title="fig : " ] as a function of the exchange field angle @xmath39 , when @xmath304 , @xmath302 , and @xmath209.,title="fig:",width=321 ] instead , these observations may be explained using the theory developed in section [ sec : theory ] .
when we have a general exchange field and so field in the @xmath30-plane , eq .
reveals that the effective energy of the srt component is @xmath315 since the factor @xmath316 vanishes for @xmath148 , we get @xmath317 for this case .
this geometry is also one where we do not expect any lrt generation , since the triplet mixing factor @xmath318 , so the conclusion is that the so coupling has no effect on the behaviour of srts for these parameters at least according to the linearized equations .
however , since @xmath319 for @xmath320 , the situation is now dramatically different .
the srt effective energy is now @xmath151 , with an imaginary contribution which again destabilizes the srts , and increases the critical temperature of the system .
we emphasize that the variation of @xmath1 with the magnetization direction is present when @xmath321 as well , albeit with a magnitude of the variation that gradually decreases as one approaches pure rashba or pure dresselhais coupling .
the results discussed in the previous section show that the critical temperature can be controlled via the magnetization direction of one single ferromagnetic layer .
this is a new result originating from the presence of so coupling . in conventional sf structures
, @xmath1 is independent of the magnetization orientation of the f layer . by using a spin - valve setup such as fsf @xcite
, it has been shown that the relative magnetization configuration between the ferromagnetic layers will tune the @xmath1 of the system .
in contrast , in our case such a spin - valve effect can be obtained with a single ferromagnet ( see figs . [
fig : tc - theta - long ] and [ fig : tc - theta - short ] ) : by rotating the magnetization an angle @xmath9 , @xmath1 goes from a maximum to a minimum .
the fact that only a single ferromagnet is required to achieve this effect is of practical importance since it can be challenging to control the relative magnetization orientation in magnetic multilayered structures .
it was pointed out in ref .
that for the case of transversal structures as depicted in fig .
[ fig : model](b ) , pure rashba or pure dresselhaus coupling and arbitrary magnetisation direction are insufficient for long range triplets to exist .
however , although these layered structures are more restrictive in their conditions for lrt generation than lateral junctions they are nevertheless one of the most relevant for current experimental setups @xcite , and herein we consider the corresponding experimentally accessible effects of so coupling as a complement to the findings of ref . .
we have provided a detailed exposition of the density of states and critical temperature for both the sf bilayer and sfs junction with so coupling , highlighting in particular the signature of long range triplets .
we saw that the spectroscopic signature depends nonmonotonically on the angle of the magnetic exchange field , and that the lrt component can induce a strong peak in the density of states at zero energy for a range of magnetization directions .
in addition to the large enhancement at zero energy , we see that by carefully choosing the so coupling and exchange field strengths in the josephson junction it is again possible to control the qualitative features of the density of states by altering the phase difference between the two superconductors _
e.g. _ with a loop geometry @xcite .
the intrinsic so coupling present in the structures considered herein derives from their lack of inversion symmetry due to the _
e.g. _ junction interfaces , so - called interfacial asymmetry , and we restricted the form of this coupling to the experimentally common and , in some cases , tunable rashba - dresselhaus form .
a lack of inversion symmetry can also derive from intrinsic noncentrosymmmetry of a crystal .
this could in principle be utilised to provide a component of the so - field in the junction direction , but to date we are not aware of such materials having been explored in experiments with sf hybrid materials . however , analytic and numerical data suggest that such materials could have significant importance for spintronic applications making use of a large triplet cooper pair population@xcite .
it is also worth considering the possibility of separating the spin - orbit coupling and the ferromagnetic layer , which would arguably be easier to fabricate , and we are currently pursuing this line of investigation . in this case
, we would expect similar conclusions regarding when the long - range triplets leave clear spectroscopic signatures and also regarding the spin - valve effect with a single ferromagnet , as found when the so coupling and exchange field coexist in the same material .
one way to practically achieve such a setup would be to deposit a very thin layer of a heavy normal metal such as au or pt between a superconductor and a conventional homogeneous ferromagnet .
the combination of the large atomic number @xmath322 and the broken structural inversion symmetry at the interface region would then provide the required so coupling .
with a very thin normal metal layer ( of the order of a couple of nm ) , the proximity effect would be significantly stronger , and thus analysis of this regime is only possible with the full usadel equations in the riccati parameterisation developed herein .
the current analysis pertains to thin film ferromagnets . upon increasing the length of ferromagnetic film
one will increase the relative proportions of long - range to short - range triplets in the middle of the ferromagnet . for strong ferromagnets where the exchange field is a significant fraction of the fermi energy , the quasiclassical usadel formalism may no longer describe the system behaviour appropriately , since it assumes that the impurity scattering rate is much larger than the other energy scales involved , and the eilenberger equation should be used instead @xcite . in the previous section , we also observed that the presence of so coupling will in many cases _ increase _ the critical temperature of a hybrid structure .
this effect is explained through an increase in the effective energy coupled to the triplet component in the usadel equation , which destabilizes the triplet pairs and closes that proximity channel .
however , for the special geometry @xmath150 , the linearized equations suggest that the srts are unaffected by the presence of so coupling , and this is consistent with the numerical results
. we also note that for the geometries with a large lrt generation , such as @xmath323 and @xmath154 , the lrt generation reduces the critical temperature again .
thus , for the physical parameters considered herein , we see that there is a very slight increase in critical temperature for these geometries , but not as large as for the geometries without lrt generation .
one particularly striking result from the critical temperature calculations is that when the rashba and dresselhaus contribution to the so coupling is of similar magnitude , one observes that the critical temperature can change by as much as 60% upon changing @xmath282 to @xmath324 , _
i.e. _ by a 90@xmath325 rotation of the magnetic field .
this implies that it is possible to create a novel kind of triplet spin valve using an sf bilayer , where the ferromagnet has a homogeneous exchange field and rashba dresselhaus coupling .
this is in contrast to previous suggestions for triplet spin valves , such as the one described by fominov _
_ , which have required trilayers with different homogeneous ferromagnets@xcite .
the construction of such a device is likely to have possible applications in the emerging field of superconducting spintronics @xcite .
the authors thank angelo di bernardo , matthias eschrig , camilla espedal , and iryna kulagina for useful discussions and gratefully acknowledge support from the ` outstanding academic fellows ' programme at ntnu and cost action mp-1201 novel functionalities through optimized confinement of condensate and fields. j.l .
was supported by the research council of norway , grant no .
205591 ( frinat ) and grant no . 216700 .
the @xmath43 components of the retarded green s function @xmath326 are not entirely independent , but can be expressed as @xmath327 which suggests that the notation can be simplified by introducing the _ tilde conjugation _
@xmath328 moreover , the normalization condition @xmath329 further constrains the possible form of @xmath326 by relating the @xmath330 components to the @xmath331 components , @xmath332 remarkably , if we pick a suitable parametrization of @xmath326 , which automatically satisfies the symmetry and normalization requirements above , then both the usadel equation and the kupriyanov lukichev boundary conditions can be reduced from @xmath43 to @xmath78 matrix equations . in this paper , we employ the so - called _ riccati parametrization _ for this purpose , which is defined by @xmath333 where the normalization matrices are @xmath334 and @xmath335 .
solving the riccati parametrized equations for the function @xmath336 in spin space is then sufficient to uniquely construct the whole green s function @xmath337 .
it is noteworthy that @xmath338 when @xmath339 , while the elements of @xmath326 diverge to infinity when @xmath340 ; so we see that a finite range of variation in @xmath341 parametrizes an infinite range of variation in @xmath326 .
we begin by deriving some basic identities , starting with the inverses of the two matrix products @xmath342 and @xmath343 : @xmath344 by comparison of the results above , we see that @xmath345 .
similar calculations for other combinations of the riccati matrices reveal that we can always move normalization matrices past gamma matrices if we also perform a tilde conjugation in the process : @xmath346 since we intend to parametrize a differential equation , we should also try to relate the derivatives of the riccati matrices
. this can be done by differentiating the definition of @xmath347 using the matrix version of the chain rule : @xmath348 ( 1-{\gamma}{\tilde{\gamma}})^{-1 } \notag\\ & = ( 1-{\gamma}{\tilde{\gamma}})^{-1 } \left [ { ( \partial_z{\gamma})}{\tilde{\gamma}}+ { \gamma}{(\partial_z{\tilde{\gamma}})}\right ] ( 1-{\gamma}{\tilde{\gamma}})^{-1 } \notag\\ & = { n}\left [ { ( \partial_z{\gamma})}{\tilde{\gamma}}+ { \gamma}{(\partial_z{\tilde{\gamma}})}\right ] { n}\ ; .\end{aligned}\ ] ] performing a tilde conjugation of the equation above , we get a similar result for @xmath349 .
thus , the derivatives of the normalization matrices satisfy the following identities : @xmath350 { n}\ ; , \\ \partial_z { \tilde{n}}&= { \tilde{n}}\left [ { ( \partial_z{\tilde{\gamma}})}{\gamma}+ { \tilde{\gamma}}{(\partial_z{\gamma})}\right ] { \tilde{n}}\ ; .\end{aligned}\ ] ] in addition to the identities derived above , one should note that the definition of the normalization matrix @xmath351 can be rewritten in many forms which may be of use when simplifying riccati parametrized expressions ; examples of this include @xmath352 and @xmath353 .
now that the basic identities are in place , it is time to parametrize the usadel equation in the ferromagnet , @xmath354=0 \ , , \end{aligned}\ ] ] where we for simplicity will let @xmath355 in this appendix .
we begin by expanding the gauge covariant derivative @xmath356 , and then simplify the result using the normalization condition @xmath329 and its derivative @xmath357 , which yields the result @xmath358 } -{[\hat a,\,\hat g\hat a\hat g ] } \ ; .
\end{aligned}\ ] ] we then write @xmath326 in component form using eq . , and also write @xmath359 in the same form using @xmath360 . in the rest of this appendix
, we will for simplicity assume that @xmath66 is real , so that @xmath361 ; in practice , this implies that @xmath66 can only depend on the spin projections @xmath362 and @xmath363 .
the derivation for the more general case of a complex @xmath359 is almost identical .
the four terms in eq . may then be written as follows : @xmath364 & \partial_z(\hat g \hat a \hat g ) \nonumber\\ & = \begin{bmatrix } \phantom{-}\partial_z ( g a g + f a \tilde f ) & \phantom{-}\partial_z ( g a f + f a \tilde g ) \phantom{-}\\ - \partial_z(\tilde g a \tilde f + \tilde
f a g ) & - \partial_z(\tilde g a \tilde g + \tilde f a f ) \phantom{- } \end{bmatrix } \ ; ; \\[1em ] & { [ \hat a,\,\hat g \partial_z \hat g ] } \nonumber\\ & = \begin{bmatrix } \phantom{- } { [ a,\ , g \partial_z g - f \partial_z \tilde f\ , ] } & \phantom{-}{\{a,\ , g \partial_z f - f \partial_z \tilde g\ } } \phantom{-}\\ - { \{a,\,\tilde g \partial_z \tilde f - \tilde f \partial_z g\ } } & - { [ a,\,\tilde g \partial_z \tilde g - \tilde f \partial_z
f\ , ] } \phantom{- } \end{bmatrix } \ ; ; \\[1em ] & { [ \hat a,\,\,\hat g \hat a \hat g ] } \nonumber\\ & = \begin{bmatrix } \phantom{- } { [ a,\ , g a g + f a \tilde f\ , ] } & \phantom{-}{\{a,\ , g a f + f a \tilde g\ } } \phantom{-}\\ \phantom{-}{\{a,\,\tilde g a \tilde f + \tilde f a g\ } } & \phantom{- } { [ a,\,\tilde g a \tilde g + \tilde f a f\ , ] } \phantom{- } \end{bmatrix } \ ; .\end{aligned}\ ] ] substituting these results back into eq . , we can find the upper blocks of the covariant derivative @xmath365 , @xmath366^{(1,1 ) } \notag\\ & = \partial_z(g\partial_z g - f\partial_z \tilde f ) - i\partial_z(g a_z g + f a_z \tilde f ) \notag\\ & - i{[a_z,\,g\partial_z g - f\partial_z\tilde f ] } - { [ { \underline{a}},\,g{\underline{a } } g + f { \underline{a } } \tilde f ] } \ ; , \\[1em ] \label{dgdg-12 } & [ { \tilde{\nabla}}\cdot(\hat g{\tilde{\nabla}}\hat g)]^{(1,2 ) } \notag\\ & = \partial_z(g\partial_z f - f\partial_z \tilde g ) - i\partial_z(g a_z f + f
a_z \tilde g ) \notag \notag\\ & - i{\{a_z,\,g\partial_z f - f\partial_z\tilde g\ } } - { \{{\underline{a}},\,g{\underline{a } } f + f { \underline{a } } \tilde g\ } } \ ; .\end{aligned}\ ] ] in this context , the notation @xmath367 refers to the @xmath368th row and @xmath369th column in nambu space .
since the green s function @xmath326 and background field @xmath359 also have a structure in spin space , the @xmath370 element in nambu space is the upper - left @xmath78 block of the matrix , and the @xmath371 element is the upper - right one .
there are two kinds of expressions that recur in the equations above , namely the components of @xmath372 , and the components of @xmath373 .
after we substitute in the riccati parametrization @xmath374 and @xmath375 , these components take the form : @xmath376^{(1,1 ) } & = g \partial_z g - f \partial_z \tilde f \notag\\ & = 2{n}\left[{(\partial_z{\gamma})}{\tilde{\gamma}}- { \gamma}{(\partial_z{\tilde{\gamma}})}\right ] { n}\ ; ; \\[0.5em ] [ \hat g
\partial_z \hat g ] ^{(1,2 ) } & = g \partial_z f - f \partial_z \tilde g \notag\\ & = 2{n}\left[{(\partial_z{\gamma})}- { \gamma}{(\partial_z{\tilde{\gamma}})}{\gamma}\right ] { \tilde{n}}\ ; ; \\[0.5em ] [ \hat g \hat a \hat g]^{(1,1 ) } & = g { \underline{a } } g + f { \underline{a } } \tilde f \notag\\ & = 4{n}({\underline{a } } + { \gamma}{\underline{a } } { \tilde{\gamma}}){n}- 2{\{{\underline{a}},\,{n}\ } } + { \underline{a } } \ ; ; \\[0.5em ] [ \hat g \hat a \hat g]^{(1,2 ) } & = g { \underline{a } } f + f { \underline{a } } \tilde g \notag\\ & = 4{n}({\underline{a}}{\gamma}+ { \gamma}{\underline{a } } ) { \tilde{n}}- 2{\{{\underline{a}},\,{n}{\gamma}\ } } \ ; .\end{aligned}\ ] ]
if we explicitly calculate the commutators of @xmath359 with the two matrices @xmath372 and @xmath373 , then we find : @xmath377}^{(1,1 ) } & = { [ { \underline{a}},\,g\partial_z g - f \partial_z \tilde f ] } \notag\\ & = 2{n}(1-{\gamma}{\tilde{\gamma}}){\underline{a}}{n}\left[{(\partial_z{\gamma})}{\tilde{\gamma}}- { \gamma}{(\partial_z{\tilde{\gamma}})}\right ] { n}\notag\\ & - 2{n}\left [ { ( \partial_z{\gamma})}{\tilde{\gamma}}- { \gamma}{(\partial_z{\tilde{\gamma}})}\right ] { n}{\underline{a}}(1-{\gamma}{\tilde{\gamma}}){n}\ ; ; \\[0.5em ] { [ \hat a,\,\hat g\partial_z\hat g]}^{(1,2 ) } & = { \{{\underline{a}},\,g\partial_z f - f \partial_z \tilde g\ } } \notag\\ & = 2{n}(1-{\gamma}{\tilde{\gamma}}){\underline{a}}{n}\left[{(\partial_z{\gamma})}- { \gamma}{(\partial_z{\tilde{\gamma}})}{\gamma}\right ] { \tilde{n}}\notag\\ & + 2{n}\left [ { ( \partial_z{\gamma})}- { \gamma}{(\partial_z{\tilde{\gamma}})}{\gamma}\right ] { \tilde{n}}{\underline{a}}(1-{\tilde{\gamma}}{\gamma } ) { \tilde{n}}\ ; ; \\[0.5em ] { [ \hat a,\,\hat g\hat a\hat g]}^{(1,1 ) } & = { [ { \underline{a}},\,g{\underline{a}}g + f { \underline{a } } \tilde f ] } \notag\\ & = 4{\underline{a}}{n}({\underline{a } } + { \gamma}{\underline{a}}{\tilde{\gamma } } ) { n}\notag\\ & - 4{n}({\underline{a } } + { \gamma}{\underline{a } } { \tilde{\gamma } } ) { n}{\underline{a } } \notag\\ & - 2{[{\underline{a}}^2,\,{n } ] } \ ; ; \\[0.5em ] { [ \hat a,\,\hat g\hat a\hat g]}^{(1,2 ) } & = { \{{\underline{a}},\,g{\underline{a}}f + f { \underline{a } } \tilde g\ } } \notag\\ & = 4{\underline{a}}{n}({\underline{a}}{\gamma}+ { \gamma}{\underline{a}}){\tilde{n}}\notag\\ & + 4{n}({\underline{a}}{\gamma}+ { \gamma}{\underline{a}}){\tilde{n}}{\underline{a } } \notag\\ & - 4{\underline{a}}{n}{\gamma}{\underline{a } } - 2{\{{\underline{a}}^2,\,{n}{\gamma}\ } } \ ; .\end{aligned}\ ] ] if we instead differentiate the aforementioned matrices with respect to @xmath29 , we obtain : @xmath378^{(1,1 ) } & = \partial_z(g \partial_z
g - f \partial_z \tilde f ) \notag\\ & = 2{n } [ { ( \partial_z^2{\gamma})}+ 2{(\partial_z{\gamma})}{\tilde{n}}{\tilde{\gamma}}{(\partial_z{\gamma } ) } ] { \tilde{\gamma}}{n}\notag\\ & - 2{n}{\gamma } [ { ( \partial_z^2{\tilde{\gamma}})}+ 2{(\partial_z{\tilde{\gamma}})}{n}{\gamma}{(\partial_z{\tilde{\gamma } } ) } ] { n}\ ; ; \\[1em ] [ \partial_z(\hat g\partial_z \hat g)]^{(1,2 ) } & = \partial_z(g \partial_z f - f
\partial_z \tilde g ) \notag\\ & = 2{n } [ { ( \partial_z^2{\gamma})}+ 2{(\partial_z{\gamma})}{\tilde{n}}{\tilde{\gamma}}{(\partial_z{\gamma } ) } ] { \tilde{n}}\notag\\ & - 2{n}{\gamma } [ { ( \partial_z^2{\tilde{\gamma}})}+ 2{(\partial_z{\tilde{\gamma}})}{n}{\gamma}{(\partial_z{\tilde{\gamma } } ) } ] { \gamma}{\tilde{n}}\ ; ; \\[1em ] [ \partial_z(\hat g{\underline{a}}\hat g)]^{(1,1 ) } & = \partial_z(g{\underline{a}}g + f { \underline{a } } \tilde f ) \notag\\ & = 2{n}(1+{\gamma}{\tilde{\gamma}}){\underline{a}}{n}[{\gamma}{(\partial_z{\tilde{\gamma}})}+ { ( \partial_z{\gamma})}{\tilde{\gamma } } ] { n}\notag\\ & + 2{n}[{\gamma}{(\partial_z{\tilde{\gamma}})}+ { ( \partial_z{\gamma})}{\tilde{\gamma } } ] { n}{\underline{a}}(1+{\gamma}{\tilde{\gamma } } ) { n}\notag\\ & + 4{n}{\gamma}{\underline{a}}{\tilde{n}}[{(\partial_z{\tilde{\gamma}})}+{\tilde{\gamma}}{(\partial_z{\gamma})}{\tilde{\gamma } } ] { n}\notag\\ & + 4{n}[{(\partial_z{\gamma})}+ { \gamma}{(\partial_z{\tilde{\gamma}})}{\gamma}]{\tilde{n}}{\underline{a}}{\tilde{\gamma}}{n}\ ; ; \\[1em ] [ \partial_z(\hat g{\underline{a}}\hat g)]^{(1,2 ) } & = \partial_z(g{\underline{a}}f + f { \underline{a } } \tilde g ) \notag\\ & = 2{n}(1+{\gamma}{\tilde{\gamma}}){\underline{a}}{n}[{(\partial_z{\gamma})}+ { \gamma}{(\partial_z{\tilde{\gamma}})}{\gamma } ] { \tilde{n}}\notag\\ & + 2{n}[{(\partial_z{\gamma})}+ { \gamma}{(\partial_z{\tilde{\gamma}})}{\gamma } ] { \tilde{n}}{\underline{a}}(1+{\tilde{\gamma}}{\gamma } ) { \tilde{n}}\notag\\ & + 4{n}{\gamma}{\underline{a}}{\tilde{n}}[{\tilde{\gamma}}{(\partial_z{\gamma})}+{(\partial_z{\tilde{\gamma}})}{\gamma } ] { \tilde{n}}\notag\\ & + 4{n}[{\gamma}{(\partial_z{\tilde{\gamma}})}+ { ( \partial_z{\gamma})}{\tilde{\gamma}}]{n}{\underline{a}}{\tilde{\gamma}}{\tilde{n}}\ ; .\end{aligned}\ ] ] combining all of the equations above , we can express eqs . and using riccati matrices .
in order to isolate the second - order derivative @xmath379 from these , the trick is to multiply eq . by @xmath341 from the right , and
subsequently subtract the result from eq . :
@xmath380^{(1,2 ) } - [ { \tilde{\nabla}}\cdot(\hat g{\tilde{\nabla}}\hat g)]^{(1,1 ) } { \gamma}\big\ } \notag\\ & = \partial^2_z\gamma + 2{(\partial_z{\gamma})}{\tilde{n}}{\tilde{\gamma}}{(\partial_z{\gamma})}\notag\\ & -2i(a_z + { \gamma}a_z { \tilde{\gamma}}){n}{(\partial_z{\gamma})}-2i{(\partial_z{\gamma})}{\tilde{n } } ( a_z + { \tilde{\gamma}}a_z{\gamma } ) \notag\\ & -2({\underline{a}}{\gamma}+{\gamma}{\underline{a}}){\tilde{n}}({\underline{a}}+{\tilde{\gamma}}{\underline{a}}{\gamma } ) -{\underline{a}}^2{\gamma}+ { \gamma}{\underline{a}}^2 \ ; .\end{aligned}\ ] ] if we finally rewrite @xmath381^{(1,1)}$ ] and @xmath381^{(1,2)}$ ] in the equation above by substituting in the usadel equation , then we obtain the following equation for the riccati matrix @xmath341 : @xmath382 the corresponding equation for @xmath383 can be found by tilde conjugation of the above . after restoring the diffusion coefficient @xmath384 , and generalizing the derivation to a complex so field @xmath66
, the above result takes the form shown in eq . .
after parametrizing the usadel equation , the next step is to do the same to the kupriyanov lukichev boundary conditions . the gauge covariant version of eq .
may be written @xmath385 \
, , \end{aligned}\ ] ] which upon expanding the covariant derivative @xmath386 becomes @xmath387 } + i \hat g_n{[\hat a_z,\,\hat g_n ] } \ ; , \end{aligned}\ ] ] where we have introduced the notation @xmath388 for the interface parameter .
we will now restrict our attention to the ( 1,1 ) and ( 1,2 ) components of the above , @xmath389 } + if_n{\ { a_z,\,\tilde f_n\ } } \ ; , \\
g_n\partial_z f_n - f_n \partial_z \tilde g_n & = \frac12\omega_n(g_1f_2 - g_2f_1 - f_1 \tilde g_2 + f_2 \tilde g_1 ) \notag\\ & \;+ ig_n{\ { a_z,\,f_n\ } } + if_n { [ a_z,\,\tilde g_n ] } \ ; .\end{aligned}\ ] ] substituting the riccati parametrizations @xmath390 and @xmath391 in the above ,
we then obtain : @xmath392 { n_{n } } & = \omega_n { n_{1 } } ( 1-{{\gamma}_{1 } } { \tilde{\gamma}_{2 } } ) { n_{2 } } \notag\\ & \,- \omega_n { n_{2 } } ( 1-{{\gamma}_{2 } } { \tilde{\gamma}_{1 } } ) { n_{1 } } \notag\\ & \,- i { n_{n}}(1 - { { \gamma}_{n } } { \tilde{\gamma}_{n } } ) { \underline{a } } { n_{n } } \notag\\ & \,- i { n_{n}}{\underline{a}}(1 - { { \gamma}_{n } } { \tilde{\gamma}_{n } } ) { n_{n } } \notag\\ & \,+ 2i { n_{n } } ( { \underline{a } } + { { \gamma}_{n } } { \underline{a } } { \tilde{\gamma}_{n } } ) { n_{n } } \label{kl-11-comp } \ ; , \\[1em ] { n_{n}}[{(\partial_z{{\gamma}_{n } } ) } - { { \gamma}_{n } } { ( \partial_z{\tilde{\gamma}_{n } } ) } { { \gamma}_{n}}]{\tilde{n_{n } } } & = \omega_n { n_{1 } } ( 1-{{\gamma}_{1 } } { \tilde{\gamma}_{2 } } ) { { \gamma}_{2 } } { \tilde{n_{2 } } } \notag\\ & \,-
\omega_n { n_{2 } } ( 1-{{\gamma}_{2 } } { \tilde{\gamma}_{1 } } ) { { \gamma}_{1 } } { \tilde{n_{1 } } } \notag\\ & \,+i{n_{n}}(1 + { { \gamma}_{n } } { \tilde{\gamma}_{n}}){\underline{a } } { { \gamma}_{n } } { \tilde{n_{n } } } \notag\\ & \,+i{n_{n}}{{\gamma}_{n } } { \underline{a } } ( 1 + { \tilde{\gamma}_{n } } { { \gamma}_{n}}){\tilde{n_{n } } } \label{kl-12-comp } \ ; .\end{aligned}\ ] ] if we multiply eq . by @xmath393 from the right ,
subtract this from eq . , and divide by @xmath394 from the left , then we obtain the following boundary condition for @xmath393 : @xmath395 when we evaluate the above for @xmath396 and @xmath397 , then it simplifies to the following : @xmath398 the boundary conditions for @xmath399 and @xmath400 are found by tilde conjugating the above .
if we generalize the derivation to a complex so field @xmath66 , and substitute back @xmath388 in the result , then we arrive at eq . .
for completeness , we present here a detailed derivation of the self - consistency equation for the bcs order parameter @xcite in a quasiclassical framework .
similar derivations can also be found in refs . .
in this paper , we follow the convention where the keldysh component of the anomalous green s function is defined as @xmath401}\rangle } \ , , \ ] ] where @xmath402 is the spin - dependent fermion annihilation operator , and the superconducting gap is defined as @xmath403 where @xmath181 is the electron
electron coupling constant in the bcs theory . for the rest of this appendix
, we will also assume that we work in an electromagnetic gauge where @xmath191 is a purely real quantity .
comparing eqs . and , and using the fermionic anticommutation relation @xmath404 we see that the superconducting gap @xmath405 can be expressed in terms of the green s functions in two different ways , @xmath406 we may then perform a quasiclassical approximation by first switching to wigner mixed coordinates , then fourier transforming the relative coordinates , then integrating out the energy dependence , and finally averaging the result over the fermi surface to obtain the isotropic part .
the resulting equations for the superconducting gap are @xmath407 where @xmath408 is the quasiclassical counterpart to @xmath409 , @xmath13 is the quasiparticle energy , and @xmath180 is the density of states per spin at the fermi level . in the equilibrium case
, the keldysh component @xmath410 can be expressed in terms of the retarded and advanced components of the green s function , @xmath411 and the advanced green s function may again be expressed in terms of the retarded one , @xmath412 which implies that the keldysh component can be expressed entirely in terms of the retarded component , @xmath413 if we extract the relevant anomalous components @xmath414 and @xmath415 from the above , we obtain the results @xmath416 \tanh(\epsilon/2 t ) \ , , \label{eq : keldysh - ra-1 } \\
f^k_{{\downarrow}{\uparrow } } & = [ f^r_{{\downarrow}{\uparrow}}({\underline{r}},+\epsilon ) + f^r_{{\uparrow}{\downarrow}}({\underline{r}},-\epsilon ) ] \tanh(\epsilon/2 t ) \ , .
\label{eq : keldysh - ra-2}\end{aligned}\ ] ] we then switch to a singlet / triplet - decomposition of the retarded component @xmath417 , where the singlet component is described by a scalar function @xmath115 , and the triplet component by the so - called @xmath116-vector @xmath418 .
this parametrization is defined by the matrix equation @xmath419 or in component form , @xmath420 parametrizing eqs . and according to eq .
, we obtain @xmath421 \tanh(\epsilon/2 t ) \
, , \label{eq : kzs-1}\\ f^k_{{\uparrow}{\downarrow}}({\underline{r}},\epsilon ) = [ & d_z({\underline{r}},+\epsilon ) - f_s({\underline{r}},+\epsilon ) \nonumber\\ + & d_z({\underline{r}},-\epsilon ) + f_s({\underline{r}},-\epsilon ) ] \tanh(\epsilon/2 t ) \ , .
\label{eq : kzs-2}\end{aligned}\ ] ] the triplet component @xmath124 can clearly be eliminated from the above equations by subtracting eq . from eq . ,
@xmath422 \tanh(\epsilon/2 t ) \ , , \ ] ] and a matching expression for the superconducting gap can be acquired by adding eqs . and ,
@xmath423 \tanh(\epsilon/2 t ) \ , .\ ] ] by comparing the two results above , we finally arrive at an equation for the superconducting gap which only depends on the singlet component of the quasiclassical green s function : @xmath424 \tanh(\epsilon/2 t ) \ , .\ ] ] if the integral above is performed for all real values of @xmath13 , it turns out to be logarithmically divergent _
e.g. _ for a bulk superconductor .
however , physically , the range of energies that should be integrated over is restricted by the energy spectra of the phonons that mediate the attractive electron
electron interactions in the superconductor .
this issue may therefore be resolved by introducing a debye cutoff @xmath425 , such that we only integrate over the region where @xmath426 . including the integration range , the gap equation is therefore @xmath427 \tanh(\epsilon/2 t ) \ , .\ ] ]
the equation above can however be simplified even further .
first of all , both @xmath428 and @xmath429 are clearly antisymmetric functions of @xmath13 , which means that the product is a symmetric function , and so it is sufficient to perform an integral over positive values of @xmath13 ,
@xmath430 \tanh(\epsilon/2 t ) \ , .\ ] ] however , because of the term @xmath431 , we still need to know the green s function for negative values of @xmath13 before we can calculate the gap . on the other hand
, the singlet component of the quasiclassical green s functions also has a symmetry when the superconducting gauge is chosen as real @xmath432 which implies that @xmath433 substituting eq . into eq .
, the gap equation takes a particularly simple form , which only depends on the real part of the singlet component @xmath434 for positive energies @xmath13 : @xmath435 let us now consider the case of a bcs bulk superconductor , which has a singlet component given by the equation @xmath436 so that the gap equation may be written as @xmath437 the part in the curly braces is only real when @xmath438 , which means that the equation can be simplified by changing the lower integration limit to @xmath191 . after also dividing the equation by @xmath439 , we then obtain the self - consistency equation @xmath440 for the zero - temperature case , where @xmath441 and @xmath442 , performing the above integral and reordering the result yields @xmath443 using the above equation for @xmath425 , and the well - known result @xmath444 where @xmath445 is the euler mascheroni constant , we can finally rewrite eq .
as : @xmath446 this version of the gap equation is particularly well - suited for numerical simulations .
one advantage is that we only need to know the green s function for positive energies , which halves the number of energies that we need to solve the usadel equation for .
the equation also takes a particularly simple form if we use energy units where @xmath447 and temperature units where @xmath448 , which is common practice in such simulations .
buzdin , phys .
* 101 * , 107005 ( 2008 ) ; a.a .
et al . _ ,
b. * 86 * , 214519 ( 2012 ) ; i. kulagina and j. linder , phys .
b * 90 * , 054504 ( 2014 ) ; f.s .
bergeret and i.v .
tokatly , preprint , arxiv:1409.4563 . | we investigate theoretically how the proximity effect in superconductor / ferromagnet hybrid structures with intrinsic spin - orbit coupling manifests in two measurable quantities , namely the density of states and critical temperature . to describe a general scenario , we allow for both rashba and dresselhaus type spin - orbit coupling .
our results are obtained via the quasiclassical theory of superconductivity , extended to include spin - orbit coupling in the usadel equation and kupriyanov lukichev boundary conditions .
unlike previous works , we have derived a riccati parametrization of the usadel equation with spin - orbit coupling which allows us to address the full proximity regime and not only the linearized weak proximity regime .
first , we consider the density of states in both sf bilayers and sfs trilayers , where the spectroscopic features in the latter case are sensitive to the phase difference between the two superconductors . we find that the presence of spin - orbit coupling leaves clear spectroscopic fingerprints in the density of states due to its role in creating spin - triplet cooper pairs . unlike sf and sfs structures without spin - orbit coupling , the density of states in the present case depends strongly on the direction of magnetization
moreover , we show that the spin - orbit coupling can stabilize spin - singlet superconductivity even in the presence of a strong exchange field @xmath0 .
this leads to the possibility of a magnetically tunable minigap : changing the direction of the exchange field opens and closes the minigap .
we also determine how the critical temperature @xmath1 of an sf bilayer is affected by spin - orbit coupling and , interestingly , demonstrate that one can achieve a spin - valve effect with a single ferromagnet .
we find that @xmath1 displays highly non - monotonic behavior both as a function of the magnetization direction and the type and direction of the spin - orbit coupling , offering a new way to exert control over the superconductivity of proximity structures . |
( fiwi ) access networks , also referred to as wireless - optical broadband access networks ( wobans ) , combine the reliability , robustness , and high capacity of optical fiber networks and the flexibility , ubiquity , and cost savings of wireless networks @xcite . to deliver peak data
rates up to 200 mb / s per user and realize the vision of complete fixed - mobile convergence , it is crucial to replace today s legacy wireline and microwave backhaul technologies with integrated fiwi broadband access networks @xcite .
significant progress has been made on the design of advanced fiwi network architectures as well as access techniques and routing protocols / algorithms over the last few years @xcite . among others ,
the beneficial impact of advanced hierarchical frame aggregation techniques on the end - to - end throughput - delay performance of an integrated ethernet passive optical network ( epon)/wireless mesh network ( wmn)-based fiwi network was demonstrated by means of simulation and experiment for voice , video , and data traffic @xcite .
a linear programming based routing algorithm was proposed in @xcite with the objective of maximizing the throughput of a fiwi network based on a cascaded epon and single - radio single - channel wmn .
extensive simulations were conducted to study the throughput gain in fiwi networks under peer - to - peer traffic among wireless mesh clients and compare the achievable throughput gain with conventional wmns without any optical backhaul .
the presented simulation results show that fiwi and conventional wmn networks achieve the same throughput when all traffic is destined to the internet , i.e. , no peer - to - peer traffic , since the interference in the wireless front - end is the major bandwidth bottleneck .
however , with increasing peer - to - peer traffic the interferences in the wireless mesh front - end increase and the throughput of wmns decreases significantly , as opposed to their fiwi counterpart whose network throughput decreases to a much lesser extent for increasing peer - to - peer traffic . the design of routing algorithms for the wireless front - end only or for both the wireless and optical domains of fiwi access networks has received a great deal of attention , resulting in a large number of wireless , integrated optical - wireless , multipath , and energy - aware routing algorithms .
important examples of wireless routing algorithms for fiwi access networks are the so - called delay - aware routing algorithm ( dara ) @xcite , delay - differentiated routing algorithm ( ddra ) @xcite , capacity and delay aware routing ( cadar ) @xcite , and risk - and - delay aware routing ( radar ) algorithm @xcite .
recently proposed integrated routing algorithms for path computation across the optical - wireless interface include the so - called availability - aware routing @xcite , multipath routing @xcite , and energy - aware routing algorithms @xcite .
most of these previous studies formulated routing in fiwi access networks as an optimization problem and obtained results mainly by means of simulation . in this paper , we present to the best of our knowledge the first analytical framework that allows to evaluate the capacity and delay performance of a wide range of fiwi network routing algorithms and provides important design guidelines for novel fiwi network routing algorithms that leverage the different unique characteristics of disparate optical fiber and wireless technologies .
although a few fiwi architectural studies exist on the integration of epon with long - term evolution ( lte ) ( e.g. , @xcite ) or worldwide interoperability for microwave access ( wimax ) wireless front - end networks ( e.g. , @xcite ) , the vast majority of studies , including but not limited to those mentioned in the above paragraph , considered fiwi access networks consisting of a conventional ieee 802.3ah epon fiber backhaul network and an ieee 802.11b / g wireless local area network ( wlan)-based wireless mesh front - end network @xcite .
our framework encompasses not only legacy epon and wlan networks , but also emerging next - generation optical and wireless technologies , such as long - reach and multi - stage 10 + gb / s time and/or wavelength division multiplexing ( tdm / wdm ) pons as well as gigabit - class very high throughput ( vht ) wlan .
our contributions are threefold .
first , we develop a unified analytical framework that comprehensively accounts for both optical and wireless broadband access networking technologies .
we note that recent studies focused either on tdm / wdm pons only , e.g. , @xcite , or on wlans only , e.g. , @xcite .
however , there is a need for a comprehensive analytical framework that gives insights into the performance of bimodal fiwi access networks built from disparate yet complementary optical and wireless technologies . toward this end ,
our framework is flexibly designed such that it not only takes the capacity mismatch and bit error rate differences between optical and wireless networks into account , but also includes possible fiber cuts of optical ( wired ) infrastructures .
second , our analysis emphasizes future and emerging next - generation pon and wlan technologies , as opposed to many previous studies that assumed state - of - the - art solutions , e.g. , conventional ieee 802.11a wlan without frame aggregation @xcite .
our analytical approach in part builds on previous studies and includes significant original analysis components to achieve accurate throughput - delay modeling and cover the scope of fiwi networks .
specifically , we build on analytical models of the distributed coordination function in wlans , e.g. , @xcite , and wlan frame aggregation , e.g. , @xcite .
we develop an accurate delay model for multihop wireless front - ends under nonsaturated and stable conditions for traffic loads from both optical and wireless network nodes , as detailed in section [ sec : wireless ] .
third , we verify our analysis by means of simulations and present extensive numerical results to shed some light on the interplay between different next - generation optical and wireless access networking technologies and configurations for a variety of traffic scenarios .
we propose an _ optimized fiwi routing algorithm ( ofra ) _ based on our developed analytical framework .
the obtained results show that ofra outperforms previously proposed routing algorithms , such as dara @xcite , cadar @xcite , and radar @xcite .
they also illustrate that it is key to carefully select appropriate paths across the fiber backhaul in order to minimize link traffic intensities and thus help stabilize the entire fiwi access network . to our best knowledge
, the presented unified analytical framework is the first to allow capacity and delay evaluations of a wide range of fiwi network routing algorithms , both previously proposed and new ones .
our analytical framework covers not only legacy epon and wlan , but also next - generation high - speed long - reach wdm pon and emerging gigabit - class vht wlan technologies .
the remainder of the paper is structured as follows . in section
[ sec : related ] , we discuss related work and recent progress on fiwi access networks . section [ sec : fiwi ] describes fiwi access networks based on next - generation pon and gigabit - class wlan technologies in greater detail .
section [ sec : model ] outlines our network model as well as traffic and routing assumptions .
the capacity and delay of the constituent fiber backhaul and wireless front - end networks are analyzed in sections [ sec : fiber ] and [ sec : wireless ] , respectively , while the stability and end - to - end delay of the entire fiwi access network are evaluated in section [ sec : end - to - end ] .
section [ sec : results ] presents numerical and verifying simulation results .
section [ sec : conclusions ] concludes the paper .
the recent survey of hybrid optical - wireless access networks @xcite explains the key underlying photonic and wireless access technologies and describes important fiwi access network architectures .
energy - efficient fiwi network architectures as well as energy - efficient medium access control ( mac ) and routing protocols were reviewed in @xcite .
recent efforts on energy - efficient routing in fiwi access networks focused on routing algorithms for cloud - integrated fiwi networks that offload the wireless mesh front - end and the optical - wireless gateways by placing cloud components , such as storage and servers , closer to mobile end - users , while at the same time maintaining low average packet delays @xcite .
a delay - based admission control scheme for providing guaranteed quality - of - service ( qos ) in fiwi networks that deploy epon as backhaul for connecting multiple wimax base stations was studied in @xcite .
a promising approach to increase throughput , decrease delay , and achieve better load balancing and resilience is the use of multipath routing schemes in the wireless mesh front - end of fiwi networks .
however , due to different delays along multiple paths , packets may arrive at the destination out of order , which deteriorates the performance of the transmission control protocol ( tcp ) .
a centralized scheduling algorithm at the optical line terminal ( olt ) of an epon that resequences the in - transit packets of each flow to ensure in - order packet arrivals at the corresponding destination was examined in @xcite .
in addition , @xcite studied a dynamic bandwidth allocation ( dba ) algorithm that prioritizes flows that may trigger tcp s fast retransmit and fast recovery , thereby further improving tcp performance .
given the increasing traffic amounts on fiwi networks , their survivability has become increasingly important @xcite .
cost - effective protection schemes against link and node failures in the optical part of fiwi networks have been proposed and optimized in @xcite .
the survivability of fiwi networks based on multi - stage pons , taking not only partial optical protection but also protection through a wireless mesh network into account , was probabilistically analyzed in @xcite .
deployment of both back - up fibers and radios was examined in @xcite .
recent research efforts have focused on the integration of performance - enhancing network coding techniques to increase the throughput and decrease the delay of fiwi access networks for unicast and multicast traffic @xcite .
most previous fiwi access network studies considered a cascaded architecture consisting of a single - stage pon and a multihop wmn , as shown in fig .
[ fig : fig1 ] . typically , the pon is a conventional ieee 802.3ah compliant wavelength - broadcasting tdm epon based on a wavelength splitter / combiner at the remote node ( rn ) , using one time - shared wavelength channel for upstream ( onus to olt ) transmissions and another time - shared wavelength channel for downstream ( olt to onus ) transmissions , both operating at a data rate of 1 gb / s .
a subset of onus may be located at the premises of residential or business subscribers , whereby each onu provides fiber - to - the - home / business ( ftth / b ) services to a single or multiple attached wired subscribers .
some onus have a mesh portal point ( mpp ) to interface with the wmn .
the wmn consists of mesh access points ( maps ) that provide wireless fiwi network access to stations ( stas ) .
mesh points ( mps ) relay the traffic between mpps and maps through wireless transmissions .
most previous fiwi studies assumed a wmn based on ieee 802.11a / b / g wlan technologies , offering a maximum raw data rate of 54 mb / s at the physical layer .
future fiwi access networks will leverage next - generation pon and wlan technologies to meet the ever increasing bandwidth requirements . a variety of next - generation pon technologies
are currently investigated to enable short - term evolutionary and long - term revolutionary upgrades of coexistent gigabit - class tdm pons @xcite .
promising solutions for pon evolution toward higher bandwidth per user are ( @xmath0 ) data rate upgrades to 10 gb / s and higher , and ( @xmath1 ) multi - wavelength channel migration toward wavelength - routing or wavelength - broadcasting wdm pons with or without cascaded tdm pons @xcite .
similarly , to alleviate the bandwidth bottleneck of the wireless mesh front - end , future fiwi networks are expected to be based on next - generation ieee 802.11n wlans , which offer data rates of 100 mb / s or higher at the mac service access point , as well as emerging ieee 802.11ac vht wlan technologies that achieve raw data rates up to 6900 mb / s . as shown in fig .
[ fig : fig2 ] , current tdm pons may evolve into next - generation single- or multi - stage pons of extended reach by exploiting high - speed tdm and/or multichannel wdm technologies and replacing the splitter / combiner at the rn with a wavelength multiplexer / demultiplexer , giving rise to the following three types of next - generation pons : fig . [
fig : fig2](a ) depicts a high - speed tdm pon , which maintains the network architecture of conventional tdm pons except that both the time - shared upstream wavelength channel @xmath2 and downstream wavelength channel @xmath3 and attached olt and tdm onus operate at data rates of 10 gb / s or higher @xcite .
a wavelength - broadcasting wdm pon has a splitter / combiner at the rn and deploys multiple wavelength channels @xmath4 , as shown in fig .
[ fig : fig2](b ) .
each of these @xmath5 wavelength channels is broadcast to all connected wdm onus and is used for bidirectional transmission .
each wdm onu selects a wavelength with a tunable bandpass filter ( e.g. , fiber bragg grating ) and reuses the downstream modulated signal coming from the olt for upstream data transmission by means of remodulation techniques , e.g. , fsk for downstream and ook for upstream @xcite .
[ fig : fig2](c ) shows a wavelength - routing wdm pon , where the conventional splitter / combiner at the rn is replaced with a wavelength multiplexer / demultiplexer , e.g. , arrayed - waveguide grating ( awg ) , such that each of the @xmath5 wavelength channels on the common feeder fiber is routed to a different distribution fiber . a given wavelength channel may be dedicated to a single onu ( e.g. , business subscriber ) or be time shared by multiple onus ( e.g. , residential subscribers ) . in the latter case , the distribution fibers contain one or more additional stages , whereby each stage consists of a wavelength - broadcasting splitter / combiner and each wavelength channel serves a different sector , see fig .
[ fig : fig2](c ) . note that due to the wavelength - routing characteristics of the wavelength multiplexer / demultiplexer , onus can be made colorless ( i.e. , wavelength - independent ) by using , for example , low - cost reflective semiconductor optical amplifiers ( rsoas ) that are suitable for bidirectional transmission via remodulation @xcite .
wavelength - routing multi - stage wdm pons enable next - generation pons with an extended optical range of up to 100 km , thus giving rise to _ long - reach wdm pons _ at the expense of additional in - line optical amplifiers .
long - reach wdm pons promise major cost savings by consolidating optical access and metropolitan area networks @xcite .
ieee 802.11n specifies a number of phy and mac enhancements for next - generation wlans .
applying orthogonal frequency division multiplexing ( ofdm ) and multiple - input multiple - output ( mimo ) antennas in the phy layer of ieee 802.11n provides various capabilities , such as antenna diversity ( selection ) and spatial multiplexing .
using multiple antennas also provides multipath capability and increases both throughput and transmission range .
the enhanced phy layer applies two adaptive coding schemes : space time block coding ( stbc ) and low density parity check coding ( ldpc ) .
ieee 802.11n wlans are able to co - exist with ieee 802.11 legacy wlans , though in greenfield deployments it is possible to increase the channel bandwidth from 20 mhz to 40 mhz via channel bonding , resulting in significantly increased raw data rates of up to 600 mb / s at the phy layer .
a main mac enhancement of 802.11n is frame aggregation , which comes in two flavors , as shown in fig .
[ fig : fig3 ] .
* aggregate mac service data unit ( a - msdu ) : * multiple msdus , each up to 2304 octets long , are joined and encapsulated into a separate subframe , see fig .
[ fig:3a ] .
specifically , multiple msdus are packed into an a - msdu , which is encapsulated into a phy service data unit ( psdu ) .
all constituent msdus must have the same traffic identifier ( tid ) value ( i.e. , same qos level ) and the resultant a - msdu must not exceed the maximum size of 7935 octets . each psdu is prepended with a phy preamble and phy header .
although the fragmentation of msdus with the same destination address is allowed , a - msdus must not be fragmented .
* aggregate mac protocol data unit ( a - mpdu ) : * multiple mpdus , each up to 4095 octets long , are joined and inserted in a separate subframe , see fig .
[ fig:3b ] .
specifically , multiple mpdus are aggregated into one psdu of a maximum size 65535 octets .
aggregation of multiple mpdus with different tid values into one psdu is allowed by using multi - tid block acknowledgment ( mtba ) .
both a - msdu and a - mpdu require only a single phy preamble and phy header . in a - msdu , the psdu includes a single mac header and frame check sequence ( fcs ) , as opposed to a - mpdu where each mpdu contains its own mac header and fcs .
a - mpdu and a - msdu can be used separately or jointly .
future gigabit - class wmns may be upgraded with emerging ieee 802.11ac vht wlan technologies that exploit further phy enhancements to achieve raw data rates up to 6900 mb / s and provide an increased maximum a - msdu / a - mpdu size of 11406/1048575 octets @xcite .
we consider a pon consisting of one olt and @xmath6 attached onus .
the tdm pon carries one upstream wavelength channel and a separate downstream wavelength channel .
we suppose that both the wavelength - broadcasting and the wavelength - routing multi - stage wdm pons carry @xmath5 bidirectional wavelength channels @xmath7 . in the wavelength - routing multi - stage wdm pon , the @xmath6 onus
are divided into @xmath5 sectors .
we use @xmath8 to index the wavelength channel as well as the corresponding sector . in our model ,
sector @xmath9 , accommodates @xmath10 onus .
specifically , onus with indices @xmath11 between @xmath12 and @xmath13 belong to sector @xmath8 , i.e. , form the set of nodes @xmath14 thus , sector @xmath15 comprises onus @xmath16 , sector @xmath17 comprises onus @xmath18 , and so on , while we assign the index @xmath19 to the olt .
the one - way propagation delay between olt and onus of sector @xmath8 is @xmath20 ( in seconds ) and the data rate of the associated wavelength channel @xmath8 is denoted by @xmath21 ( in bit / s ) .
hence , each sector of the wavelength - routing multi - stage wdm pon is allowed to operate at a different data rate serving a subset of onus located at a different distance from the olt ( e.g. , business vs. residential service areas ) . for ease of exposition
, we assume that in the wavelength - broadcasting tdm and wdm pons all wavelength channels operate at the same data rate @xmath22 ( in bit / s ) and that all onus have the one - way propagation delay @xmath23 ( in seconds ) from the olt .
all or a subset of the @xmath6 onus are equipped with an mpp to interface with the wmn .
the wmn is composed of different zones @xmath24 , whereby each zone operates on a distinct frequency such that the frequencies of neighboring zones do not overlap .
frequencies may be spatially reused in nonadjacent zones .
a subset of mps are assumed to be equipped with multiple radios to enable them to send and receive data in more than one zone and thereby serve as relay nodes between adjacent zones .
we denote each radio operating in a given relay mp in a given zone @xmath24 by a unique @xmath25 .
the remaining mps as well as all mpps , maps , and stas are assumed to have only a single radio @xmath25 operating on the frequency of their corresponding zone .
all wireless nodes are assumed to be stationary ; incorporating mobility is left for future research . adopting the notation proposed in @xcite ,
we let @xmath26 denote the set of multi - radio relay mps and @xmath27 denote the set of single - radio mps , mpps , maps , and stas in zone @xmath24 .
note that set @xmath26 is empty if there are only single - radio mps in zone @xmath24 .
note that due to this set definition each multi - radio mp is designated by multiple @xmath25 ; one @xmath25 and corresponding set @xmath28 for each zone @xmath24 in which it can send and receive .
the wmn operates at a data rate @xmath29 ( in bit / s ) . in the wmn
, we assume that the bit error rate ( ber ) of the wireless channel is @xmath30 . on the contrary
, the ber of the pon is assumed to be negligible and is therefore set to zero .
however , individual fiber links may fail due to fiber cuts and become unavailable for routing traffic across the pon , as described next in more detail . throughout
, we neglect nodal processing delays . we denote @xmath31 for the set of fiwi network nodes that act as traffic sources and destinations . specifically , we consider @xmath31 to contain the olt , the @xmath6 onus ( whereby a given onu models the set of end users with wired access to the onu ) , and a given number @xmath32 of stas . in our model ,
mpps , mps , and maps forward in - transit traffic , without generating their own traffic .
hence , the number of traffic sources / destinations is given by @xmath33 .
furthermore , we define the traffic matrix @xmath34 , where @xmath35 represents the number of frames per second that are generated at fiwi network node @xmath0 and destined to fiwi network node @xmath36 ( note that @xmath37 for @xmath38 ) . we allow for any arbitrary distribution @xmath39 of the frame length @xmath40 ( in bit ) and denote @xmath41 and @xmath42 for the mean and variance of the length of a frame , respectively .
the traffic generation is assumed to be ergodic and stationary . our capacity and delay analysis flexibly accommodates any routing algorithm .
for each pair of fiwi network source node @xmath0 and destination node @xmath36 , a particular considered routing algorithm results in a specific traffic rate ( in frames / s ) @xmath43 sent in the fiber domain and traffic rate @xmath44 sent in the wireless domain .
a conventional onu @xmath11 without an additional mpp can not send in the wireless domain , i.e. , @xmath45 , and sends its entire generated traffic to the olt , i.e. , @xmath46 . on the other hand
, an onu @xmath11 equipped with an mpp can send in the wireless domain , i.e. , @xmath47 .
note that we allow for multipath routing in both the fiber and wireless domains , whereby traffic coming from or going to the olt may be sent across a single or multiple onus and their collocated mpps .
we consider throughout first - come - first - served service in each network node .
for the wavelength - routing multi - stage wdm pon , we define the normalized downstream traffic rate ( intensity ) in sector @xmath48 , as @xmath49 where the first term represents the traffic generated by the olt for sector @xmath8 and the second term accounts for the traffic from all onus sent to sector @xmath8 via the olt .
we define the upstream traffic rate ( in frames / s ) of onu @xmath11 as @xmath50 where the first term denotes traffic destined to the olt and the second term represents the traffic sent to other onus via the olt .
the normalized upstream traffic rate ( intensity ) of sector @xmath8 is @xmath51 for stability , the normalized downstream and upstream traffic rates have to satisfy @xmath52 in each sector @xmath53 , of the wavelength - routing multi - stage wdm pon . in the wavelength - broadcasting tdm pon ( @xmath54 ) and wdm pon ( @xmath55 ) ,
we define the upstream traffic intensity @xmath56 and downstream traffic intensity @xmath57 as : @xmath58 @xmath59 the tdm and wdm pons are stable if @xmath60 and @xmath61 .
the delay analysis of section [ delay_ana : sec ] applies only for a stable network , which can be ensured through admission control techniques . in the wavelength - routing multi - stage wdm pon
, the olt sends a downstream frame to an onu in sector @xmath8 by transmitting the frame on wavelength @xmath8 , which is received by all onus in the sector .
we model all downstream transmissions in sector @xmath8 to emanate from a single queue . for poisson frame traffic ,
the downstream queueing delay is thus modeled by an m / g/1 queue characterized by the pollaczek - khintchine formula @xcite @xmath62 giving the total downstream frame delay @xmath63 weighing the downstream delays @xmath64 in the sectors @xmath8 by the relative downstream traffic intensities @xmath65 in the sectors , gives the average downstream delay of the wavelength - routing multi - stage wdm pon @xmath66 for the upstream delay , we model each wavelength channel @xmath67 , as a single upstream wavelength channel of a conventional epon .
accordingly , from eq .
( 39 ) in @xcite , we obtain for the mean upstream delay of sector @xmath8 @xmath68 and the average upstream delay of the wavelength - routing multi - stage wdm pon equals @xmath69 to improve the accuracy of our delay analysis , we take into account that traffic coming from an onu @xmath11 in sector @xmath70 and destined to onu @xmath71 in sector @xmath8 is queued at the intermediate olt before being sent downstream to onu @xmath71 , i.e. , the olt acts like an insertion buffer between onus @xmath11 and @xmath71 .
consequently , to compensate for the queueing delay at the olt we apply the method proposed in @xcite by subtracting the correction term @xmath72 whereby for the setting that @xmath73 for all channels @xmath8 @xmath74 denotes the rate of upstream traffic in sector @xmath70 destined for sector @xmath8 , from the above calculated mean downstream delay .
thus , for sector @xmath75 , the corrected mean downstream delay @xmath76 is given by @xmath77 by replacing @xmath64 with @xmath76 in eq .
( [ eq : wdm_pon_dsdelay ] ) we obtain a more accurate calculation of the average downstream delay for the wavelength - routing multi - stage wdm pon , as examined in section [ sec : results ] .
next , we evaluate the average downstream and upstream delays for the wavelength - broadcasting tdm pon ( @xmath54 ) and wdm pon ( @xmath55 ) . with the aforementioned correction term
the average downstream and upstream delays are given by @xmath78 and @xmath79 respectively , whereby @xmath80
so far , we have analyzed only the optical fiber backhaul of the fiwi network .
next , we focus on the wireless front - end . in the following ,
we derive multiple relations between known parameter values and unknown variables .
afterwards , we outline how to obtain the unknowns numerically .
more specifically , in sections [ ftm : sec][sec : durframetra ] we build on and adapt existing models of distributed coordination @xcite and frame aggregation @xcite in wlans to formulate the basic frame aggregate transmission and collision probabilities as well as time slot duration in the distributed access system .
we note that these existing models have primarily focused on accurately representing the collision probabilities and system throughput ; we found that directly adapting these existing models gives delay characterizations that are reasonably accurate only for specific scenarios , such as single - hop networking , but are very coarse for multi - hop networking . in sections [ sertime_fa :
sec][dwmnp : sec ] we develop a general multihop delay model that is simple , yet accurate by considering the complete service time of a frame aggregate in the wireless front - end network carrying traffic streams from and to both wireless and optical network nodes . as defined in section [ sec : netarch ] , we denote the radio operating in a given sta or onu equipped with an mpp by a unique @xmath25 .
moreover , we denote each radio operating in a given relay mp in a unique zone @xmath24 by a unique @xmath25 . for ease of exposition
, we refer to `` radio @xmath25 '' henceforth as `` node @xmath25 . ''
similar to @xcite , we model time as being slotted and denote @xmath81 for the mean duration of a time slot at node @xmath25 .
the mean time slot duration @xmath81 corresponds to the average time period required for a successful frame transmission , a collided frame transmission , or an idle waiting slot at node @xmath25 and is evaluated in section [ sec : durframetra ] .
we let @xmath82 denote the probability that there is a frame waiting for transmission at node @xmath25 in a time slot . for an sta or onu with collocated mpp @xmath25
we denote @xmath83 for the traffic load that emanates from node @xmath25 , i.e. , @xmath84 for a relay mp we obtain for a given wireless mesh routing algorithm the frame arrival rate for each of the mp s radios @xmath85 associated with a different zone @xmath24 : @xmath86 whereby @xmath0 and @xmath36 denote any pair of sta or onu with collocated mpp that send traffic on a path via relay mp @xmath25 , as computed by the given routing algorithm for the wireless mesh front - end of the fiwi network . for exponentially distributed inter - frame arrival times with mean @xmath87 ( which occur for a poisson process with rate @xmath83 )
, @xmath82 is related to the offered frame load at node @xmath25 during mean time slot duration @xmath81 via @xmath88 in this section , we first characterize the sizes of the frame aggregates and then the frame aggregate error probability . for a prescribed distribution @xmath89 of the size ( in bit ) of a single frame ,
e.g. , the typical trimodal ip packet size distribution , the distribution @xmath90 of the size ( in bit ) of a transmitted a - msdu or a - mpdu can be obtained as the convolution of @xmath39 with itself , i.e. , @xmath91 the number of required convolutions equals the number of frames carried in the aggregate , which in turn depends on the minimum frame size , including the mac - layer overhead of the corresponding frame aggregation scheme , and the maximum size of an a - msdu / a - mpdu @xmath92 ( see fig . [
fig : fig3 ] ) . from the distribution @xmath90
we obtain the average frame aggregate sizes @xmath93 $ ] and @xmath94 $ ] .
correspondingly , we divide the traffic rate @xmath95 ( in frames / s ) by the average number of frames in an aggregate to obtain the traffic rate in frame aggregates per second .
moreover , as ground work for section [ sec : durframetra ] we obtain the average size of the longest a - msdu , @xmath96 $ ] , and longest a - mpdu , @xmath97 $ ] , involved in a collision with the simplifying assumption of neglecting the collision probability of more than two packets @xcite as @xmath98 = \int_0^{a_{\max}^{\text{a - msdu / a - mpdu}}}\left(1-a(x)^2\right ) dx.\ ] ] the probability @xmath99 of an erroneously transmitted frame aggregate , referred to henceforth as `` transmission error '' , can be evaluated in terms of bit error probability @xmath100 and size @xmath101 of a transmitted a - msdu ( with distribution @xmath90 ) with ( * ? ? ?
( 16 ) ) ; for a - mpdu , @xmath99 can be evaluated in terms of @xmath100 and the sizes @xmath102 of the aggregated frames with ( * ? ? ?
( 18 ) ) . in particular
, @xmath103 where @xmath101 is the size @xmath101 of a transmitted a - msdu ( with distribution @xmath90 ) , index @xmath0 runs from one to the total number of aggregated frames , and @xmath102 is the size of the @xmath0th frame in a transmitted a - mpdu .
following @xcite , we note that the transmission of any transmitting node @xmath104 in zone @xmath24 can not collide if none of the other nodes @xmath105 transmits , i.e. , we obtain the collision probability @xmath106 as @xmath107 where @xmath108 denotes the transmission probability of wmn node @xmath109 .
note that if the considered node is a relay mp , eq .
( [ eq : p_c ] ) holds for each associated zone @xmath24 ( and corresponding radio @xmath25 ) .
we define the probability of either a collision or transmission error @xmath110 , in brief collision / transmission error probability , as @xmath111 the transmission probability @xmath112 for any node @xmath113 can be evaluated as a function of the frame waiting probability @xmath114 , the frame collision / transmission error probability @xmath115 , the minimum contention window @xmath116 , and the maximum backoff stage @xmath117 by ( * ? ? ?
( 1 ) ) , as explained in @xcite . in particular , @xmath118}-\frac{q_\omega^2(1-p_\omega ) } { 1-q_\omega}\right),\ ] ] with @xmath119}\nonumber \\ & & \ \ \ + ( 1-q_\omega ) \nonumber \\ & & \ \ \
+ \frac{q_\omega(w_0 + 1)[p_\omega(1-q_\omega)-q_\omega(1-p_\omega)^2 ] } { 2(1-q_\omega)}\nonumber \\ & & \ \ \ + \frac{p_\omega q_\omega^2}{2(1-q_\omega)(1-p_\omega ) } \left(\frac{w_0}{1-(1-q_\omega)^{w_0}}-(1-p_\omega)^2\right ) \nonumber \\ & & \ \ \ \ \ \ \ \ \ \ \ \
\cdot \left ( \frac{2w_0 [ 1-p_\omega - p_\omega(2p_\omega)^{h-1 } ] } { 1 - 2p_\omega } + 1 \right),\end{aligned}\ ] ] where @xmath116 is node
@xmath25 s minimum contention window , @xmath120 is the node s maximum window size , and @xmath117 is the maximum backoff stage .
the probability that there is at least one transmission taking place in zone @xmath24 in a given time slot is given by @xmath121 a successful frame aggregate transmission occurs if exactly one node @xmath25 transmits ( and all other nodes @xmath122 are silent ) , given that there is a transmission , i.e. , @xmath123 we denote @xmath124 for the duration of an empty time slot without any data transmission on the wireless channel in zone @xmath24 , which occurs with probability @xmath125 . with probability
@xmath126 there is a transmission in a given time slot in zone @xmath24 , which is successful with probability @xmath127 and unsuccessful ( resulting in a collision ) with the complementary probability @xmath128 .
we denote @xmath129 for the mean duration of a successful frame aggregate transmission and @xmath130 is the mean duration of a frame aggregate transmission with collision in zone @xmath24 .
note that @xmath129 and @xmath130 depend on the frame aggregation technique ( a - msdu or a - mpdu ) and on the access mechanism @xmath131 ( basic access denoted by @xmath132 or rts / cts denoted by @xmath133 ) .
for the basic access mechanism , we define @xmath134 , where @xmath135 denotes the propagation delay and @xmath29 the wmn data rate . for the rts / cts access mechanism , we define @xmath136 .
( note that in ieee 802.11n the parameters ack , rts , and cts as well as the phy / mac header and fcs below are given in bits , while the other parameters are given in seconds . ) then , for a successful frame aggregate transmission we have : @xmath137 + \text{fcs})/r & \text{for a - msdu}\\ \\ \theta_s^{\alpha } + e[\text{a - mpdu}]/r\ & \text{for a - mpdu}. \end{array}\right.\ ] ] moreover , with @xmath138 , for a collided frame aggregate transmission we have : @xmath139 + \text{fcs})/r & \text{for a - msdu},\\ \\ \theta_c^{\rm basic } + e[\text{a - mpdu}^*]/r\ & \text{for a - mpdu } \end{array}\right.\ ] ] as well as for both a - msdu and a - mpdu , @xmath140 thus , we obtain the expected time slot duration @xmath81 at node @xmath25 in zone @xmath24 of our network model ( corresponding to ( * ? ? ?
( 13 ) ) ) as @xmath141.\end{aligned}\ ] ] equations ( [ eq : q_omega ] ) , ( [ eq : p_omega ] ) , ( * ? ? ?
( 1 ) ) , and ( [ eq : e_omega ] ) can be solved numerically for the unknown variables @xmath82 , @xmath110 , @xmath112 , and @xmath81 for each given set of values for the known network model parameters .
we use the obtained numerical solutions to evaluate the mean delay at node @xmath25 as analyzed in the following sections [ sertime_fa : sec ] and [ wmnnodedel : sec ] .
we proceed to evaluate the expected service ( transmission ) time for a frame aggregate , which may require several transmission attempts , at a given node @xmath25 . with the basic access mechanism ,
the transmission of the frame aggregate occurs without a collision ( @xmath142 ) or transmission error with probability @xmath143 ( [ eq : p_omega ] ) , requiring one @xmath144 . with probability @xmath145 ,
the frame aggregate suffers @xmath146 , collisions or transmission errors , requiring @xmath36 backoff procedures and re - transmissions .
thus , the expected service time for basic access is @xmath147 for the rts / cts access mechanism , collisions can occur only for the rts or cts frames ( which are short and have negligible probability of transmission errors ) , whereas transmission errors may occur for the frame aggregates .
collisions require only retransmissions of the rts frame , whereas transmission errors require retransmissions of the entire frame aggregate .
more specifically , only one frame transmission ( @xmath148 ) is required if no transmission error occurs ; this event has probability @xmath149 .
this transmission without transmission error may involve @xmath150 , collisions of the rts / cts frames . on the other hand , two frame transmissions ( @xmath151 )
are required if there is once a transmission error ; this event has probability @xmath152 .
this @xmath151 scenario requires twice an rts / cts reservation , which each time may experience @xmath153 collisions , as well as two full frame transmission delays @xmath129 .
generally , @xmath154 , frame transmissions are required if @xmath155 times there is a frame transmission error .
each of the @xmath156 frame transmission attempts requires an rts / cts reservation and a full frame transmission delay @xmath129 . in summary
, we evaluate the mean service delay for a frame aggregate with rts / cts access as @xmath157.\end{aligned}\ ] ] we first evaluate the overall service time @xmath158 from the time instant when a frame aggregate arrives at the head of the queue at node @xmath25 to the completion of its successful transmission .
subsequently , with @xmath158 characterizing the overall service time at node @xmath25 , we evaluate the queueing delay @xmath159 .
the overall service time @xmath158 is given by the service time @xmath160 required for transmitting a frame aggregate and the sensing delay @xmath161 required for the reception of frame aggregates by node @xmath25 from other nodes , i.e. , @xmath162 as a first step towards modeling the sensing delay at a node @xmath163 , we consider the service times @xmath164 at nodes @xmath165 and scale these service times linearly with the corresponding traffic intensities @xmath166 to obtain the sensing delay component @xmath167 as a second modeling step , we consider the service times plus sensing delay components scaled by the respective traffic intensities to obtain the sensing delay @xmath168 employed in the evaluation of the overall service delay ( [ delom_ovall : eqn ] ) .
we approximate the queue at node @xmath25 by an m / m/1 queue with mean arrival rate @xmath83 and mean service time @xmath158 .
this queue is stable if @xmath169 the total delay ( for queueing plus service ) at node @xmath25 is then given by @xmath170 if node @xmath25 is an onu with a collocated mpp the accuracy of the queueing delay calculation is improved by subtracting a correction term : @xmath171 for the wavelength - broadcasting tdm pon and wdm pon , or @xmath172 for the wavelength - routing multi - stage wdm pon , whose sector @xmath8 accommodates the onu with collocated mpp .
note that @xmath173 or @xmath174 accounts for the traffic of all pairs of source node @xmath0 and destination node @xmath36 traversing onu @xmath25 from the fiber backhaul towards the wireless front - end network . in order to obtain the delay in the wireless front - end of our fiwi network
, we have to average the sums of the nodal delays of all possible paths for all pairs of source node @xmath0 and destination node @xmath36 : @xmath175 with the queueing delay correction terms @xmath176 whereby @xmath177 is the traffic intensity at node @xmath25 due to traffic flowing from source node @xmath0 to destination node @xmath36 .
the entire fiwi access network is stable if and only if all of its optical and wireless subnetworks are stable .
if the optical backhaul consists of a wavelength - routing multi - stage wdm pon the stability conditions in eq .
( [ eq : wr_1 ] ) must be satisfied . in the case of the wavelength - broadcasting tdm and wdm pon ,
the optical backhaul is stable if both @xmath56 and @xmath57 defined in eqs .
( [ eq : rho_u ] ) and ( [ eq : rho_d ] ) , respectively , are smaller than one . the wireless mesh front - end is stable if the stability condition in eq .
( [ eq : rho_omega ] ) is satisfied for each wmn node .
we obtain the mean end - to - end delay of the entire bimodal fiwi access network as @xmath178
we set the parameters of the fiwi mesh front - end to the default values for next - generation wlans @xcite , see table [ tab:1 ] .
we consider a distance of 1 km between any pair of adjacent wmn nodes ( which is well within the maximum distance of presently available outdoor wireless access points ) , translating into a propagation delay of @xmath179 s. .fiwi network parameters [ cols="^,^",options="header " , ] [ default ] we first verify the accuracy of our probabilistic analysis by means of simulations . in our initial verifying simulations
, we consider the fiwi network configuration of fig .
[ fig : fig4 ] .
the fiber backhaul is a tdm pon , or a wavelength - broadcasting / routing wdm pon with @xmath180 bidirectional wavelength channels ( @xmath181 , @xmath182 ) , each operating at @xmath183 gb / s ( compliant with ieee 802.3ah ) . in the case of the wavelength - routing ( wr ) wdm
pon , the two sectors are defined as : @xmath181 : @xmath184 and @xmath182 : @xmath185 .
all four onus are located 20 km from the olt ( translating into a one - way propagation delay @xmath186 ms ) and are equipped with an mpp .
the wmn is composed of the aforementioned 4 mpps plus 16 stas and 4 mps , which are distributed over 11 wireless zones , as shown in fig .
[ fig : fig4 ] . for instance ,
the wmn zone containing @xmath187 comprises 1 mpp , 2 stas , and 1 mp .
mpps and stas use a single radio , whereas mps use 3 , 4 , 4 , 3 radios from left to right in fig .
[ fig : fig4 ] .
all wmn nodes apply the rts / cts access mechanism .
the wmn operates at @xmath188 mb / s ( compliant with ieee 802.11n ) with a bit error rate of @xmath189 .
we consider poisson traffic with fixed - size frames of 1500 bytes ( octets ) .
we use a - msdu for frame aggregation , whereby each a - msdu carries the maximum permissible payload of 5 frames , see fig .
[ fig : fig3](a ) . similar to @xcite , we consider two operation modes : ( @xmath0 ) _ wmn - only mode _ which has no fiber backhaul in place ; and wmn nodes apply minimum wireless hop routing ( @xmath1 ) _ wireless - optical - wireless mode _ which deploys the fiwi network configuration of fig .
[ fig : fig4 ] . for both modes ,
we consider the minimum interference routing algorithm @xcite , which selects the path with the minimum number of wireless hops .
we compare different routing algorithms in section [ fiwiroutalg : sec ] .
the simulation results presented in @xcite indicate that the throughput performance of wmns deteriorates much faster for increasing peer - to - peer traffic among stas than that of fiwi networks , while wmn and fiwi networks achieve the same throughput when all traffic is destined to the internet . for comparison with @xcite
, we consider _ peer - to - peer ( p2p ) traffic _ , where each frame generated by a given sta is destined to any other of the remaining 15 stas with equal probability 1/15 , and _ upstream traffic _ , where all frames generated by the stas are destined to the olt .
[ fig : result1 ] depicts the results of our probabilistic analysis for the mean delay as a function of the mean aggregate throughput of a stand - alone wmn network and a tdm pon based fiwi network for p2p and upstream traffic .
the figure also shows verifying simulation results and their 95% confidence intervals , whereby simulations were run 100 times for each considered traffic load .
we observe from fig .
[ fig : result1 ] that the mean delay of the wmn increases sharply as the mean aggregate throughput asymptotically approaches its maximum data rate of 300 mb / s .
we also confirm the findings of @xcite that under p2p traffic the mean aggregate throughput can be increased by using a tdm pon as fiber backhaul to offload the wireless mesh front - end at the expense of a slightly increased mean delay due to the introduced upstream and downstream pon delay to and from the olt . as opposed to @xcite ,
however , fig .
[ fig : result1 ] shows that the throughput - delay performance of the considered fiwi network is further improved significantly under upstream traffic .
these different observations are due to the fact that in @xcite the single - radio single - channel wmn based on legacy ieee 802.11a wlan with a limited data rate of 54 mb / s suffered from severe channel congestion close to the mpps , which is alleviated in the multi - radio multi - channel wmn based on next - generation high - throughput wlan technologies . next
, we verify different fiwi network architectures and their constituent subnetworks for _ uniform _ and _ nonuniform traffic _ for minimum ( wireless or optical ) hop routing @xcite .
[ fig : result2 ] depicts the throughput - delay performance of a stand - alone wmn front - end , stand - alone tdm pon , and a variety of integrated fiwi network architectures using different fiber backhaul solutions , including conventional tdm pon , wavelength - broadcasting wdm pon ( wdm pon ) , and wavelength - routing wdm pon ( wr pon ) . in the tdm pon
only ( wmn only ) scenario under uniform traffic , each onu ( sta ) generates the same amount of traffic and each generated frame is destined to any of the remaining onus ( stas ) with equal probability . as expected , the wmn and tdm pon saturate at roughly 300 mb / s and 1 gb / s , respectively , and the tdm pon is able to support much higher data rates per source node ( onu ) at lower delays than the wmn .
furthermore , we observe from fig . [
fig : result2 ] that under uniform traffic conditions , where stas and onus send unicast traffic randomly uniformly distributed among themselves , fiwi networks based on a wavelength - broadcasting wdm pon or a wr pon give the same throughput - delay performance , clearly outperforming their single - channel tdm pon based counterpart .
however , there is a clear difference between wdm pon and wr pon fiber backhauls when traffic becomes unbalanced . to see this ,
let us consider a nonuniform traffic scenario , where @xmath187 and @xmath190 and their 4 associated stas ( see fig . [
fig : fig4 ] ) generate 30% more traffic than the remaining onus and stas .
under such a nonuniform traffic scenario , a fiwi network based on a wavelength - broadcasting wdm pon performs better , as shown in fig .
[ fig : result2 ] .
this is due to the fact that the wdm pon provides the two heavily loaded @xmath187 and @xmath190 with access to both wavelength channels , as opposed to the wr pon , thus resulting in an improved throughput - delay performance .
overall , we note that the analysis and verifying simulation results presented in figs .
[ fig : result1 ] and [ fig : result2 ] match very well for a wide range of fiwi network architectures and traffic scenarios .
recall from section [ sec : traffic_model ] that our capacity and delay analysis flexibly accommodates any routing algorithm and allows for multipath routing in both the fiber and wireless domains . in this section ,
we study the impact of different routing algorithms on the throughput - delay performance of next - generation fiwi access networks in greater detail , including their sensitivity to key network parameters .
specifically , we examine the following single - path routing algorithms : * minimum hop routing : * conventional shortest path routing selects for each source - destination node pair the path minimizing the required number of wireless and/or optical hops .
* minimum interference routing * @xcite : the path with the minimum wireless hop count is selected .
the rationale behind this algorithm is that the maximum throughput of wireless networks is typically much lower compared to the throughput in optical networks .
thus , minimizing the wireless hop count tends to increase the maximum fiwi network throughput . *
minimum delay routing : * similar to the previously proposed wmn routing algorithms dara @xcite , cadar @xcite , and radar @xcite , we apply a slightly extended minimum delay routing algorithm , which aims at selecting the path that minimizes the end - to - end delay of eq .
( [ eq : delay ] ) across the entire bimodal fiwi access network .
the applied minimum delay routing algorithm is a greedy algorithm and proceeds in two steps . in the initialization step , paths are set to the minimum hop routes .
the second step computes for each source - destination node pair the path with the minimum end - to - end delay under given traffic demands .
* optimized fiwi routing algorithm ( ofra ) : * we propose the optimized fiwi routing algorithm ( ofra ) , which proceeds in two steps similar to minimum delay routing .
after the initialization step to minimum hop routes , the second step of ofra computes for each source - destination node pair the path @xmath191 with the minimization objective @xmath192 where @xmath193 represents the long - run traffic intensity at a generic fiwi network node @xmath194 , which may be either an optical node belonging to the fiber backhaul or a wireless node belonging to the wireless mesh front - end . based on a combination of historic traffic patterns as well as traffic measurements and estimations similar to @xcite ,
the traffic intensities @xmath193 used in ofra can be periodically updated with strategies similar to @xcite .
these long - run traffic intensities vary typically slowly , e.g. , with a diurnal pattern , allowing essentially offline computation of the ofra paths .
more precisely , for the wr pon we have @xmath195 ( see eq .
( [ eq : wr_olt ] ) ) if node @xmath194 is the olt and @xmath196 ( see eq .
( [ eq : wr_onu ] ) ) if node @xmath194 is an onu . for the wavelength - broadcasting tdm and wdm pon
, we have @xmath197 ( eq .
( [ eq : rho_d ] ) ) and @xmath198 ( eq . ( [ eq : rho_u ] ) ) if node @xmath194 is the olt or an onu , respectively . for a wireless node @xmath194
, @xmath193 is given by the left - hand side of ( [ eq : rho_omega ] ) .
ofra s path length measure includes the maximum traffic intensity @xmath199 along a path @xmath191 in order to penalize paths with a high traffic intensity at one or more fiwi network nodes . for a given set of traffic flows
, ofra minimizes the traffic intensities , particularly the high ones , at the fiwi network nodes . decreasing the traffic intensities tends to allow for a higher number of supported traffic flows and thus higher throughput . to allow for a larger number of possible paths for the following numerical investigations of the different considered routing algorithms , we double the fiwi network configuration of fig .
[ fig : fig4 ] .
we consider a wavelength - routing ( wr ) wdm pon with a total of 8 onu / mpps , 8 mps , and 32 stas in 22 wireless zones , whereby onu / mpps 1 - 4 and onu / mpps 5 - 8 are served on wavelength channel @xmath181 and @xmath182 , respectively .
furthermore , to evaluate different traffic loads in the optical and wireless domains , we consider the following traffic matrix for the olt , @xmath6 onus , and @xmath32 stas : + + , + + + where @xmath200 denotes the mean traffic rate ( in frames / second ) . the parameter @xmath201 can be used to test different traffic intensities in the pon , since the onus could be underutilized compared to the wmn in the considered topology .
recall from fig .
[ fig : fig1 ] that onus may serve multiple subscribers with wired onu access , whose aggregate traffic leads to an increased load at onus . . ] for a conventional wr wdm pon with a typical optical fiber range of 20 km , fig .
[ fig : result3 ] illustrates that ofra yields the best throughput - delay performance for @xmath202 , i.e. , every optical and wireless fiwi node generates the same amount of traffic .
minimum interference routing tends to overload the wireless mpp interfaces as it does not count the fiber backhaul as a hop , resulting in high delays .
the throughput - delay performance of the four considered fiwi routing algorithms largely depends on the given traffic loads and length of the fiber backhaul .
[ fig : result4 ] depicts their throughput - delay performance for ( @xmath0 ) a conventional 20 km range and ( @xmath1 ) a 100 km long - reach wr wdm pon , whereby in both configurations we set @xmath203 , i.e. , the amount of generated traffic among optical nodes ( olt and onus ) is 100 times higher than that between node pairs involving at least one wireless node ( sta ) .
more precisely , all the ( increased ) inter - onu / olt traffic is sent across the wdm pon only , thus creating a higher loaded fiber backhaul .
we observe from fig .
[ fig : result4 ] that in general all four routing algorithms achieve a higher maximum aggregate throughput due to the increased traffic load carried on the fiber backhaul .
we observe that for a conventional 20 km range wr wdm pon with small to medium traffic loads , ofra gives slightly higher delays than the other considered routing algorithms .
this observation is in contrast to fig .
[ fig : result3 ] , though in both figures ofra yields the highest maximum aggregate throughput .
we have measured the traffic on the optical and wireless network interfaces of each onu / mpp .
our measurements show that at low to medium traffic loads with @xmath204 , ofra routes significantly less traffic across the wdm pon than the other routing algorithms , but instead uses the less loaded wireless mesh front - end .
this is due to the objective of ofra to give preference to links with lower traffic intensities . as a consequence , for @xmath203 , ofra routes relatively more traffic over lightly loaded wireless links , even though this implies more wireless hops , resulting in a slightly increased mean delay compared to the other routing algorithms at low to medium loads . )
a conventional 20 km range and ( @xmath1 ) a 100 km long - reach wavelength - routing ( wr ) wdm pon and @xmath203 . ] fig .
[ fig : result4 ] also shows the impact of the increased propagation delay in a long - reach wdm pon with a fiber range of 100 km between olt and onus . aside from a
generally increased mean delay , we observe that minimum hop and minimum interference routing as well as ofra provide comparable delays at low to medium traffic loads , while the maximum achievable throughput differences at high traffic loads are more pronounced than for the 20 km range .
the favorable performance of ofra at high traffic loads is potentially of high practical relevance since access networks are the bottlenecks in many networking scenarios and thus experience relatively high loads while core networks operate at low to medium loads .
[ fig : result4 ] illustrates that minimum delay routing performs poorly in long - reach wdm pon based fiwi access networks .
our measurements indicate that minimum delay routing utilizes the huge bandwidth of the long - reach wdm pon much less than the other routing algorithms in order to avoid the increased propagation delay . as a consequence , with minimum delay routing most traffic
is sent across the wmn , which offers significantly lower data rates than the fiber backhaul , resulting in a congested wireless front - end and thereby an inferior throughput - delay performance . to highlight the flexibility of our analysis
, we note that it accommodates any type and number of fiber failures .
fiber failures represent one of the major differences between optical ( wired ) fiber and wireless networks and affect the availability of bimodal fiwi networks . in the event of one or more distribution fiber cuts , the corresponding disconnected onu / mpp(s )
turn(s ) into a conventional wireless mp without offering gateway functions to the fiber backhaul any longer .
fiwi access network survivability for arbitrary fiber failure probabilities has been analyzed in @xcite .
[ fig : result5 ] illustrates the detrimental impact of distribution fiber failures on the throughput - delay performance of a 20 km range wavelength - routing wdm pon , which is typically left unprotected due to the small number of cost - sharing subscribers and cost - sensitivity of access networks . .
] we also note that the analytical framework is able to account for other types of network failure , e.g. , onu / mpp failures . in this case , malfunctioning onu / mpps become unavailable for both optical and wireless routing . in principle ,
fiwi access networks can be made more robust against fiber failures through various optical redundancy strategies , such as onu dual homing , point - to - point interconnection fibers between pairs of onus , fiber protection rings to interconnect a group of closely located onus by a short - distance fiber ring , or meshed pon topologies @xcite .
these redundancy strategies in general imply major architectural and onu modifications of the fiwi access network of fig .
[ fig : fig1 ] under consideration . to incorporate such topological pon modifications ,
the fiber part of the capacity and delay analysis would need to be modified accordingly .
our analysis is also applicable to the emerging ieee standard 802.11ac for future vht wlans with raw data rates up to 6900 mb / s .
in addition to a number of phy layer enhancements , ieee 802.11ac will increase the maximum a - msdu size from 7935 to 11406 octets and the maximum a - mpdu size from 65535 octets to 1048575 octets .
both enhancements can be readily accommodated in our analysis by setting the parameters @xmath92 and @xmath29 accordingly .
( optical and wireless data rates , @xmath21 and @xmath29 , are given in gb / s and mb / s , respectively ) . ]
[ fig : result6 ] illustrates the fiwi network performance gain achieved with a wireless front - end based on vht wlan instead of ieee 802.11n wlan with maximum data rate of 600 mb / s , for minimum hop routing , an optical range of 20 km , and @xmath202 .
for a wavelength - routing wdm pon operating at a wavelength channel data rate of 1 gb / s , we observe from fig .
[ fig : result6 ] that vht wlan roughly triples the maximum mean aggregate throughput and clearly outperforms 600 mb / s 802.11n wlan in terms of both throughput and delay .
furthermore , the figure shows that replacing the 1 gb / s wavelength - routing wdm pon with its high - speed 10 gb / s counterpart ( compliant with the ieee 802.3av 10g - epon standard ) does not yield a higher maximum aggregate throughput , but it does lower the mean delay especially at medium traffic loads before wireless links at the optical - wireless interfaces get increasingly congested at higher traffic loads .
a variety of routing algorithms have recently been proposed for integrated fiwi access networks based on complementary epon and wlan - mesh networks . in this article
, we presented the first analytical framework to quantify the performance of fiwi network routing algorithms , validate previous simulation studies , and provide insightful guidelines for the design of novel integrated optical - wireless routing algorithms for future fiwi access networks leveraging next - generation pons , notably long - reach 10 + gb / s tdm / wdm pons , and emerging gigabit - class wlan technologies .
our analytical framework is very flexible and can be applied to any existing or new optical - wireless routing algorithm .
furthermore , it takes the different characteristics of disparate optical and wireless networking technologies into account . beside their capacity mismatch and bit error rate differences
, the framework also incorporates arbitrary frame size distributions , traffic matrices , optical / wireless propagation delays , data rates , and fiber cuts .
we investigated the performance of minimum hop , minimum interference ( wireless hop ) , minimum delay , and our proposed ofra routing algorithms .
the obtained results showed that ofra yields the highest maximum aggregate throughput for both conventional and long - reach wavelength - routing wdm pons under balanced and unbalanced traffic loads .
for a higher loaded fiber backhaul , however , ofra gives priority to lightly loaded wireless links , leading to an increased mean delay at small to medium wireless traffic loads .
we also observed that using vht wlan helps increase the maximum mean aggregate throughput significantly , while high - speed 10 gb / s wdm pon helps lower the mean delay especially at medium traffic loads .
there are several important directions for future research .
one direction is to examine mechanisms for providing qulity of service or supporting specific traffic types , see e.g. , @xcite .
further detailed study of the impact of different dynamic bandwidth allocation approaches for long - reach pons , e.g. , @xcite , and their effectiveness in integrated fiwi networks is of interest .
yet another direction is the examine the internetworking of fiwi networks with metropolitan area networks @xcite .
s. sarkar , h .- h .
yen , s. dixit , and b. mukherjee , `` hybrid wireless - optical broadband access network ( woban ) : network planning using lagrangean relaxation , '' _ ieee / acm transactions on networking _ , vol .
17 , no . 4 , pp .
10941105 , aug .
m. a. ali , g. ellinas , h. erkan , a. hadjiantonis , and r. dorsinville , `` on the vision of complete fixed - mobile convergence , '' _ ieee / osa j. lightwave technol .
_ , vol . 28 , no . 16 , pp .
23432357 , aug .
2010 .
a. reaz , v. ramamurthi , s. sarkar , d. ghosal , s. dixit , and b. mukherjee , `` cadar : an efficient routing algorithm for a wireless - optical broadband access network ( woban ) , '' _ ieee / osa journal of optical communications and networking _ , vol . 1 , no . 5 , pp . 392403 , oct .
2009 .
g. shen , r. s. tucker , and c .- j .
chae , `` fixed mobile convergence architectures for broadband access : integration of epon and wimax , '' _ ieee communications magazine _ , vol .
45 , no . 8 , pp . 4450 , aug .
2007 .
f. aurzada , m. scheutzow , m. reisslein , and m. maier , `` towards a fundamental understanding of the stability and delay of offline wdm epons , '' _ ieee / osa j. optical communications and networking _ , vol . 2 , no . 1 , pp . 5166 , jan .
f. aurzada , m. scheutzow , m. reisslein , n. ghazisaidi , and m. maier , `` capacity and delay analysis of next - generation passive optical networks ( ng - pons ) , '' _ ieee transactions on communications _ , vol .
59 , no . 5 , pp .
13781388 , may 2011 .
s. bharati and p. saengudomlert , `` analysis of mean packet delay for dynamic bandwidth allocation algorithms in epons , '' _ ieee / osa journal of lightwave technology _ , vol . 28 , no . 23 , pp . 34543462 , 2010 .
b. lannoo , l. verslegers , d. colle , m. pickavet , m. gagnaire , and p. demeester , `` analytical model for the ipact dynamic bandwidth allocation algorithm in epons , '' _ osa journal of optical networking _ , vol . 6 , no . 6 , pp . 677688 , jun .
m. mcgarry , m. reisslein , f. aurzada , and m. scheutzow , `` shortest propagation delay ( spd ) first scheduling for epons with heterogeneous propagation delays , '' _ ieee journal on selected areas in communications _
28 , no . 6 , pp . 849862 , aug . 2010 .
m. t. ngo , a. gravey , and d. bhadauria , `` a mean value analysis approach for evaluating the performance of epon with gated ipact , '' in _ proc . of int .
conference on optical network design and modeling ( ondm ) _ , mar .
2008 , pp . 16 .
a. reaz , v. ramamurthi , and m. tornatore , `` cloud - over - woban ( cow ) : an offloading - enabled access network design , '' in _ proc . , ieee international conference on communications ( icc ) _ , jun .
2011 , pp .
a. r. dhaini , p .- h . ho , and x. jiang , `` qos control for guaranteed service bundles over fiber - wireless ( fiwi ) broadband access networks , '' _
ieee j. lightw .
_ , vol . 29 , no . 10 , pp . 15001513 , may 2011 .
y. liu , l. guo , b. gong , r. ma , x. gong , l. zhang , and j. yang , `` green survivability in fiber - wireless ( fiwi ) broadband access network , '' _ optical fiber technology _ , vol .
18 , no . 2 ,
6880 , mar .
2012 .
y. liu , l. guo , r. ma , and w. hou , `` auxiliary graph based protection for survivable fiber - wireless ( fiwi ) access network considering different levels of failures , '' _ opt . fiber techn .
_ , vol . 18 , no . 6 , pp .
430439 , 2012 .
z. yubin , h. li , x. ruitao , q. yaojun , and j. yuefeng , `` wireless protection switching for video service in wireless - optical broadband access network , '' in _ proc .
conf . on broadband network multimedia technology ( ic - bnmt ) _
, 2009 , pp . 760764 .
n. ghazisaidi , m. scheutzow , and m. maier , `` survivability analysis of next - generation passive optical networks and fiber - wireless access networks , '' _ ieee trans .
_ , vol . 60 , no . 2 , pp
. 479492 , june 2011 .
y. liu , q. song , r. ma , b. li , and b. gong , `` protection based on backup radios and backup fibers for survivable fiber - wireless ( fiwi ) access network , '' _ j. network and computer appl .
, in print _ , 2013 .
j. zhang , w. xu , and x. wang , `` distributed online optimization of wireless optical networks with network coding , '' _ ieee / osa journal of lightwave technology _ , vol .
30 , no .
14 , pp . 22462255 , jul . 2012 .
m. d. andrade , g. kramer , l. wosinska , j. chen , s. sallent , and b. mukherjee , `` evaluating strategies for evolution of passive optical networks , '' _ ieee comm .
_ , vol .
49 , no .
7 , pp . 176184 , jul . 2011 .
f. aurzada , m. scheutzow , m. herzog , m. maier , and m. reisslein , `` delay analysis of ethernet passive optical networks with gated service , '' _ osa journal of optical networking _ , vol . 7 , no . 1 ,
2541 , jan . 2008 .
a. bianco , j. finochietto , g. giarratana , f. neri , and c. piglione , `` measurement - based reconfiguration in optical ring metro networks , '' _ ieee / osa j. lightwave tech .
_ , vol . 23 , no . 10 , pp . 31563166 , 2005 .
a. elwalid , d. mitra , i. saniee , and i. widjaja , `` routing and protection in gmpls networks : from shortest paths to optimized designs , '' _ ieee / osa j. lightwave tech .
_ , vol .
21 , no . 11 , pp . 28282838 , 2003 .
a. dixit , b. lannoo , g. das , d. colle , m. pickavet , and p. demeester , `` dynamic bandwidth allocation with sla awareness for qos in ethernet passive optical networks , '' _ ieee / osa journal of optical communications and networking _ , vol . 5 , no . 3 , pp .
240253 , 2013 .
j. ahmed , j. chen , l. wosinska , b. chen , and b. mukherjee , `` efficient inter - thread scheduling scheme for long - reach passive optical networks , '' _ ieee communications mag .
_ , vol .
51 , no . 2 , pp .
s35s43 , feb .
a. buttaboni , m. de andrade , and m. tornatore , `` a multi - threaded dynamic bandwidth and wavelength allocation scheme with void filling for long reach wdm / tdm pons , '' _ ieee / osa journal of lightwave technology _ , vol .
31 , no . 8 , pp . 11491157 , apr .
2013 .
t. jimenez , n. merayo , p. fernandez , r. duran , i. de miguel , r. lorenzo , and e. abril , `` implementation of a pid controller for the bandwidth assignment in long - reach pons , '' _ ieee / osa j. optical commun . and netw .
_ , vol . 4 , no . 5 , pp . 392401 , may 2012 .
a. mercian , m. mcgarry , and m. reisslein , `` offline and online multi - thread polling in long - reach pons : a critical evaluation , '' _ ieee / osa journal of lightwave technology _ , vol .
31 , no . 12 , pp .
20182228 , jun . 2013 .
n. merayo , t. jimenez , r. duran , p. fernandez , i. miguel , r. lorenzo , and e. abril , `` adaptive polling algorithm to provide subscriber and service differentiation in a long - reach epon , '' _ photonic network communications _
257264 , 2010 .
m. maier , m. reisslein , and a. wolisz , `` a hybrid mac protocol for a metro wdm network using multiple free spectral ranges of an arrayed - waveguide grating , '' _ computer networks _
41 , no . 4 , pp . 407433 , mar .
m. scheutzow , m. maier , m. reisslein , and a. wolisz , `` wavelength reuse for efficient packet - switched transport in an awg - based metro wdm network , '' _ ieee / osa journal of lightwave technology _ , vol .
21 , no . 6 , pp . 14351455 , jun .
yang , m. maier , m. reisslein , and w. m. carlyle , `` a genetic algorithm - based methodology for optimizing multiservice convergence in a metro wdm network , '' _ ieee / osa journal of lightwave technology _ , vol .
21 , no . 5 , pp .
11141133 , may 2003 .
m. yuang , i .-
f . chao , and b. lo , `` hopsman : an experimental optical packet - switched metro wdm ring network with high - performance medium access control , '' _ ieee / osa journal of optical communications and networking _ , vol . 2 , no . 2 , pp .
91101 , feb .
i. m. white , m. s. rogge , k. shrikhande , and l. g. kazovsky , `` a summary of the hornet project : a next - generation metropolitan area network , '' _ ieee journal on selected areas in communications _ , vol .
21 , no . 9 ,
14781494 , nov . | current gigabit - class passive optical networks ( pons ) evolve into next - generation pons , whereby high - speed 10 + gb / s time division multiplexing ( tdm ) and long - reach wavelength - broadcasting / routing wavelength division multiplexing ( wdm ) pons are promising near - term candidates . on the other hand , next - generation wireless local area networks ( wlans ) based on frame aggregation techniques will leverage physical layer enhancements , giving rise to gigabit - class very high throughput ( vht ) wlans . in this paper
, we develop an analytical framework for evaluating the capacity and delay performance of a wide range of routing algorithms in converged fiber - wireless ( fiwi ) broadband access networks based on different next - generation pons and a gigabit - class multi - radio multi - channel wlan - mesh front - end .
our framework is very flexible and incorporates arbitrary frame size distributions , traffic matrices , optical / wireless propagation delays , data rates , and fiber faults .
we verify the accuracy of our probabilistic analysis by means of simulation for the wireless and wireless - optical - wireless operation modes of various fiwi network architectures under peer - to - peer , upstream , uniform , and nonuniform traffic scenarios .
the results indicate that our proposed optimized fiwi routing algorithm ( ofra ) outperforms minimum ( wireless ) hop and delay routing in terms of throughput for balanced and unbalanced traffic loads , at the expense of a slightly increased mean delay at small to medium traffic loads .
availability , fiber - wireless ( fiwi ) access networks , frame aggregation , integrated routing algorithms , next - generation pons , vht wlan . |
one of the most important results from the @xmath64 collider hera is the observation that about @xmath65 of deep inelastic scattering ( dis ) events exhibit a large rapidity gap ( lrg ) between the direction of the proton beam and that of the first significant energy deposition in the detector @xcite .
these lrg events , @xmath66 , result predominantly from the diffractive dissociation of the virtual photon , @xmath67 , and have been interpreted in terms of the exchange of a colour - singlet object known as the pomeron ( @xmath68 ) , introduced to describe the energy dependence of total cross sections in hadron - hadron scattering @xcite .
the hera data @xcite have been analysed in terms of a diffractive structure function , @xmath69 @xcite , defined in analogy with the proton structure function , @xmath70 .
the results are consistent with a `` resolved pomeron '' model where @xmath69 factorises into a regge - inspired pomeron flux and a pomeron structure function .
the latter has been analysed @xcite using the dglap evolution equations .
the extracted parton densities are dominated by gluons .
alternatively , a number of qcd models @xcite has been proposed to describe diffraction in dis . in these models ,
the virtual photon dissociates into a @xmath71 or @xmath72 state that interacts with the proton by the exchange of a gluon ladder .
the diffractive cross section can then be formulated in terms of contributions from @xmath71 and @xmath72 final states .
the @xmath72 component is expected to dominate the diffractive cross section at high masses , @xmath73 @xcite .
an important feature of the three - parton final state in both classes of models is that , in the rest frame of the system @xmath6 ( @xmath74 centre - of - mass system for the resolved pomeron model ) , the @xmath71 system tends to populate the virtual - photon hemisphere , while the gluon is emitted in the opposite direction ( the pomeron direction in the resolved pomeron model ) . in the resolved pomeron model ,
the gluon is the remnant of the pomeron left after the boson - gluon fusion process has produced a @xmath71 pair ; as a remnant , it follows the original pomeron direction . in dipole models , to leading order , the gluon in the @xmath72 fluctuation from the virtual photon either couples directly to the exchanged gluon ladder or emerges from it , depending on the reference frame . in this case also , therefore , it follows the direction of the gluon ladder ( @xmath75 pomeron in resolved pomeron model ) and the @xmath71 pair carries on in the direction of the virtual photon .
studies of the hadronic final state and jet production in diffractive interactions have been presented by both h1 @xcite and zeus @xcite . in this paper , the jet - like structure of the hadronic final state
is studied with the aim of isolating three - parton final states by selecting three - jet events . as was already observed in @xmath76 experiments @xcite ,
three - jet topologies can be clearly separated only for centre - of - mass energies above about 20 gev .
the measurements are thus performed in a restricted phase - space region populated by high - mass diffractive systems .
such systems are a small fraction of the total diffractive cross section .
they are separated from a substantial background of non - diffractive processes by requiring a large rapidity gap .
a search is made for jets using an algorithm that allows the study of jet configurations aligned with respect to the @xmath74 axis and ensures good parton - hadron correlations .
the properties of the three - jet final states are used to test the expectation that the gluon from the dominant @xmath72 contribution is emitted in the pomeron direction .
cross sections are presented as a function of the jet pseudorapidity and jet transverse momentum with respect to the @xmath74 axis .
in addition , the internal structure of the jets is measured for jets found in either the photon or the pomeron directions , with the aim of distinguishing between quark- and gluon - initiated jets .
the different measurements are compared to models of diffraction in dis .
the data were collected with the zeus detector at hera during 1998 - 2000 when hera collided 27.6 gev electrons or positrons range studied , @xmath77 and @xmath78 scattering are assumed to give identical results since contributions from @xmath79 exchange are negligible . ] on 920 gev protons .
a total of @xmath80 pb@xmath1 of @xmath81 data and @xmath82 pb@xmath1 of @xmath83 data was used .
a detailed description of the zeus detector can be found elsewhere @xcite .
a brief outline of the most relevant components for this analysis is given below .
the zeus high - resolution uranium - scintillator calorimeter ( cal ) @xcite covers @xmath84 of the total solid angle .
it is divided into three parts with respect to the polar angle axis pointing in the proton beam direction , referred to as the `` forward direction '' , and the @xmath6 axis pointing left towards the centre of hera .
the coordinate origin is at the nominal interaction point .
the pseudorapidity is defined as @xmath85 , where the polar angle , @xmath86 , is measured with respect to the proton beam direction . ] as viewed from the nominal interaction point : forward ( fcal ) , rear ( rcal ) and barrel ( bcal ) calorimeters .
each part is subdivided longitudinally into electromagnetic ( emc ) and hadronic ( hac ) sections .
the cal energy resolution , as measured in test - beams , is @xmath87 for positrons and @xmath88 for hadrons .
charged particles were tracked by the central tracking detector ( ctd ) @xcite , which operates in a magnetic field of 1.43 t provided by a thin superconducting coil . the ctd is a drift chamber consisting of 72 cylindrical layers , organised in 9 superlayers , covering the region @xmath89 .
the transverse momentum resolution for full - length tracks is @xmath90 , with @xmath91 in gev . in 1998 ,
a forward plug calorimeter ( fpc ) @xcite was installed in the @xmath92 @xmath93 beam hole of the fcal with a small hole of radius @xmath94 cm in the centre of the fpc to accommodate the beam pipe .
this increased the forward calorimetric coverage by about 1 unit of pseudorapidity to @xmath95 .
the fpc consisted of a lead - scintillator sandwich calorimeter divided into electromagnetic and hadronic sections which were read out separately by wavelength - shifting fibres and photomultipliers .
the emc ( hac ) sections were segmented into cells of @xmath96 @xmath93 ( @xmath97 @xmath93 ) area and constituted a total of 76 read - out channels .
the energy resolution was @xmath98 and @xmath99 for positrons and pions , respectively , where the energies are measured in units of gev @xcite .
the small - angle rear tracking detector ( srtd ) @xcite was used to measure the position of those positrons scattered at a sufficiently small angle to strike it .
it consists of two planes of scintillator strips attached to the front face of the rcal .
the srtd signals resolve single minimum - ionizing particles and provide a position resolution of 0.3 cm .
the position of those positrons falling outside the srtd acceptance was measured in the cal .
the luminosity was measured via the positron - proton bremsstrahlung process , @xmath100 , using a lead - scintillator calorimeter ( lumi ) @xcite located at @xmath101 m in the hera tunnel .
the kinematics of inclusive deep inelastic scattering @xmath102 is described by any two of the following variables : @xmath103 where @xmath104 and @xmath105 are , respectively , the four - momenta of the initial and final positrons , @xmath106 is the initial proton four - momentum , @xmath107 is the fraction of the energy transferred to the proton in its rest frame , @xmath108 is the proton mass and @xmath109 is the @xmath110 centre - of - mass energy . for the description of the diffractive process @xmath111 , in addition to the invariant mass of the system @xmath6 , @xmath73 ,
two further variables are introduced : @xmath112 in resolved pomeron models , where the interaction is described as the exchange of a particle - like pomeron , @xmath113 is the fraction of the proton momentum carried by the pomeron and @xmath114 is the fraction of its momentum carried by the parton within it that is probed by the virtual photon .
three monte carlo ( mc ) models for the diffractive deep inelastic process @xmath111 have been considered in order to describe the diffractive hadronic final state . the rapgap 2.08/06 mc generator @xcite implements the ingelman - schlein factorisable model @xcite for the pomeron .
the matrix elements for the @xmath115 process ( @xmath116 ) and for the @xmath117 processes ( @xmath118 , @xmath119 ) are included . to avoid divergences in the matrix elements for the @xmath117 processes for massless quarks ,
a cut of @xmath120 gev@xmath121 was applied , where @xmath122 is the transverse momentum of any of the outgoing partons with respect to the photon direction in the centre - of - mass frame of the hard scattering .
the relative contribution of each component is determined by the quark and gluon densities within the pomeron , which evolve according to the dglap equations .
the h1 parameterisation @xcite for the pomeron parton densities with a ` hard gluon ' carrying @xmath123 of the momentum at @xmath124 gev@xmath121 was used .
two different rapgap mc samples were generated , in which the higher - order qcd corrections were treated in two different ways : either using the colour - dipole model ( cdm ) @xcite as implemented in ariadne 4.08 @xcite or using parton showers ( ps ) as implemented in lepto 6.1 @xcite . the lund string - fragmentation scheme @xcite as implemented in jetset 7.4 @xcite was used for hadronisation .
first - order electroweak corrections are taken into account via the heracles 4.6 program @xcite . the ridi 2.0 mc generator @xcite implements the diffractive dissociation of the virtual photon into @xmath71 and @xmath72 final states following the approach of ryskin @xcite , in the framework of the leading logarithmic approximation of pqcd .
contributions from transversely and longitudinally polarised photons are included .
the predicted cross section is proportional to the square of the gluon momentum density in the proton .
the cteq4 m parameterisation @xcite for the parton densities was used .
final - state parton showers and fragmentation were implemented using jetset 7.4 .
first - order radiative corrections were also included .
the satrap mc generator @xcite is based on the golec - biernat and wsthoff model @xcite . in this model
, the diffractive cross section is expressed as the convolution of the photon wave - function , calculated for @xmath71 and @xmath72 colour dipoles and including the contributions of both transversely and longitudinally polarised photons , with the dipole cross section for scattering on the proton .
the parameters of the model were tuned to describe the total @xmath125 cross section as measured at hera .
final - state parton showers and fragmentation were simulated via jetset 7.4 .
first - order radiative corrections were taken into account via the heracles 4.6 program . during the course of this analysis
, it became clear that neither ridi nor satrap gave a reasonable description of the data .
as discussed in section 8 , this was traced to inadequacies in the implementation of the modelling of higher - order qcd processes in these models .
most notably , no initial - state parton cascades were included , and the final - state qcd radiation from the gluon in the dominant @xmath72 contribution was suppressed .
a new implementation of higher - order qcd processes in satrap was carried out @xcite , in which the cdm approach was implemented in a similar fashion to that in rapgap .
this model is referred to as satrap - cdm .
processes where the proton dissociates into a system @xmath126 , @xmath127 , were generated using the epsoft 2.0 mc generator @xcite .
this simulates events based on the triple - regge formalism @xcite and an @xmath128 distribution as measured for diffractive dissociation in @xmath129 scattering
. samples with different parameters for the simulation of the system @xmath126 were considered .
non - diffractive dis background was generated including first - order qed radiative corrections using lepto 6.5 interfaced to heracles 4.5 via djangoh 1.1 @xcite .
the cteq4 m set of proton parton densities was used .
the colour - dipole model was used to simulate parton cascades and fragmentation was performed using jetset 7.4 .
the rapgap - cdm model provides a good description of most of the measured distributions and was used to study the accuracy of the kinematic variables and jet reconstruction , the efficiency for selecting events and the corrections for detector and resolution effects .
the detector simulation was based on the geant 3.13 program @xcite .
the variables @xmath130 and @xmath109 were determined from the information of the scattered positron . the hadronic final - state system , @xmath6 , resulting from the dissociation of the virtual photon ,
was reconstructed using `` energy flow objects '' , hereafter denoted as efos @xcite , assumed to have the mass of the pion .
these efos combine information from charged tracks , as measured in the ctd , and energy clusters measured in the cal and fpc . in the kinematic region
considered in this analysis , the invariant mass @xmath73 was reconstructed on average with a systematic shift of @xmath131 and a relative resolution of @xmath65 , while the hadronic three - vector was reconstructed with a resolution of about @xmath132 gev for the @xmath6 and @xmath133 components and @xmath134 gev for the @xmath135 component .
the four - momentum of the virtual photon , @xmath136 , was determined with no significant bias and with a relative resolution of about @xmath65 for the @xmath6 and @xmath133 components and @xmath137 for the @xmath135 and energy components .
the triggering and online event selections were similar to those used for the zeus measurement of the structure function @xmath70 @xcite .
the offline selection criteria , applied to select diffractive dis events , required : * a scattered positron candidate ( identified @xcite via its pattern of energy deposition in the calorimeter ) with energy , @xmath138 , greater than 10 gev and @xmath139 . for an incoming positron of energy @xmath140
, @xmath107 was estimated from the energy and the polar angle , @xmath141 , of the scattered positron by : @xmath142 the impact position of the scattered positron in the rcal was required to be outside a box of 26 @xmath143 20 @xmath93 around the beam pipe to ensure a fully contained positron candidate . in addition , candidates within the ctd acceptance region , but without a matched charged track reconstructed in the ctd , were rejected ; * @xmath144 gev , where @xmath145 and the sum runs over all calorimeter cells , i.e. including those belonging to the scattered positron .
this cut , together with the previous requirements on the positron candidate , eliminated background from photoproduction events and beam - gas interactions producing fake scattered positrons ; * a reconstructed vertex with @xmath146 cm and at least two associated tracks ; * the presence of a large rapidity gap in the outgoing proton direction , to select diffractive events .
events with @xmath147 were selected , where @xmath148 is the pseudorapidity of the most - forward cluster with energy above 400 mev , as measured in the cal and fpc ; * at least three reconstructed efos in the hadronic final state to ensure that the event plane could be determined .
the measurements were limited to the following kinematic range : * @xmath2 gev@xmath121 ; * @xmath4 gev ; * @xmath5 gev , where @xmath73 was corrected for energy losses ( see section 6.1 ) ; * @xmath149 .
in this region , the diffractive hadronic final state @xmath6 is well contained in the central detector , thus minimising energy losses through the forward and rear beam - pipe holes .
a good separation of the diffractive signal from the non - diffractive background was achieved : after the @xmath148 requirement , the background was typically around @xmath150 and at most @xmath65 , as estimated from mc studies .
after all cuts , 7175 events remained .
a search for jets was performed in the centre - of - mass system of the observed hadronic final state , @xmath6 , using the exclusive @xmath7-algorithm @xcite .
all efos in the event were used as input to the jet - finding procedure . for each pair of particles (
efos ) @xmath151 and @xmath152 in an event , the quantity @xmath153 was computed , where @xmath154 is the energy of the @xmath155 particle and @xmath156 is the angle between particles @xmath151 and @xmath152 .
the pair of particles with the smallest value of @xmath157 was replaced by a _
pseudo - particle _ or cluster .
the four - momentum of the cluster was determined using the `` e - recombination scheme '' : @xmath158 this clustering procedure was repeated until all @xmath157 values exceeded a given threshold , @xmath159 , and all the remaining clusters were then labelled as jets . for this recombination scheme ,
the jets become massive and the total invariant mass , @xmath73 , coincides with the invariant mass of the three - jet system .
as applied in the centre - of - mass system , this algorithm produces at least two jets in every event .
the same jet search procedure was applied to the final - state hadrons for simulated events .
+ figure [ fig1 ] shows the measured @xmath160 fractions for @xmath161 and @xmath162 as a function of the jet resolution parameter , @xmath159 , in the region @xmath163 .
the rate of three - jet production varies from @xmath164 at @xmath165 to @xmath166 at @xmath167 .
the measured jet fractions were compared with diffractive mc models .
all the models show trends similar to the data but only satrap - cdm provides a good description of the measurements .
the predicted @xmath160 fractions are sensitive to the particular implementation of higher - order qcd corrections and hadronisation processes in each mc model .
the differences observed between the predicted jet fractions from rapgap - cdm and rapgap - ps for @xmath168 indicate where the dependence on the model used for the simulation of higher - order qcd contributions starts to be sizeable .
+ good parton - hadron correlation in three - jet production was observed in the mc simulation for @xmath169 gev . in this mass range ,
values for @xmath159 around @xmath170 gave the smallest hadronisation corrections .
the parton - hadron jet correlation was studied using a mc sample with three jets found in the final state at both parton and hadron levels .
the angular distance between each parton and the corresponding jet of hadrons was such that the partons were always well contained within the size of the hadron jets , which have a typical angular radius larger than 0.5 radians .
the reconstruction of the jet variables was studied using mc events with three jets at both hadron and detector levels .
matched pairs of jets were selected and used to study the jet reconstruction in the detector .
the characteristics of the reconstructed jet quantities were found to be as follows : * the polar angle of the jet , @xmath171 , defined with respect to the @xmath74 axis direction .
] , was reconstructed with no significant systematic bias and a resolution of @xmath172 and @xmath173 for the two most energetic jets and the least energetic jet in the event , respectively ; * the measured invariant mass , @xmath73 , and jet energies , @xmath174 , were corrected for the energy loss by a common factor of @xmath175 , determined from mc studies . in the kinematic range
considered , the jet energy was reconstructed with a typical resolution of @xmath65 and the jet energy correction was independent of the jet direction ; * the transverse momentum of the jets , @xmath176 , measured with respect to the @xmath74 axis , was reconstructed without significant bias and with a resolution of 0.7 gev ; * the jet pseudorapidity , @xmath177 , measured in the laboratory frame ( see below ) , was reconstructed without significant bias and with a resolution of 0.1 to 0.2 units .
in addition , it was checked that , in the @xmath73 region considered and for a fixed value of @xmath178 , the jets , defined in the centre - of - mass frame , always had energy significantly above the noise level in the cal .
the jet energies in the laboratory frame , @xmath179 , computed by boosting the jet four - momenta in the centre - of - mass system back to the laboratory frame , were typically above 4 gev , thus ensuring a well understood jet - energy reconstruction for all jets .
+ the sample of three - jet final states was defined by using a fixed @xmath178 and the jets were required to be in the range @xmath180 .
the variable @xmath177 was computed in the same way as @xmath179 .
the effect of the inclusion of jet masses in the determination of @xmath177 was found to be negligible .
this procedure selected events with jets reconstructed in the central part of the detector , which are not affected by the @xmath148 cut applied in the forward region and for which a good correlation between the jet before and after detector effects was preserved .
after the @xmath177 cut , a final sample of 891 events with exactly three identified jets was obtained .
neglecting jet mass , the topology of a three - jet final state can be described using the fractional - momentum variables @xmath181 where @xmath182 denotes the @xmath183 jet three - momentum , determined in the centre - of - mass frame , and by construction @xmath184 .
the jets were sorted according to their momenta in such a way that @xmath185 two of the @xmath186 variables in eq .
( [ eq : xi ] ) are sufficient to describe the jet topology in the final state .
following the studies of three - jet production in @xmath76 collisions at petra @xcite , these quantities were chosen to be @xmath187 and @xmath188 , as suggested by ellis and karliner @xcite .
the internal structure of the jets was studied in terms of the differential jet shape , defined in the centre - of - mass system as the average of the fraction of the jet energy which lies inside an annulus of inner angular distance @xmath189 and outer angular distance @xmath190 from the jet axis .
efos were used to reconstruct the jet shape defined as @xmath191 where @xmath192 denotes the sum of the energies of the efos belonging to a given jet whose angular distance to the jet axis is within the range @xmath193 with @xmath194 radians ; @xmath195 is the number of jets . for mc events , the same jet - shape definition as used above for the efos
was applied to the final - state hadrons .
the measured distributions were corrected bin - by - bin for acceptance and migrations using rapgap - cdm .
the measurement was carried out in the region @xmath2 gev@xmath121 , @xmath4 gev , @xmath5 gev and @xmath196 , for diffractive events with @xmath197 and @xmath198 , where @xmath199 is the pseudorapidity of the most - forward final - state hadron with energy above 400 mev .
the measurements presented here thus cover a very restricted phase - space region . according to rapgap , in the kinematic region considered , less than @xmath65 of all diffractive events have @xmath197 and three jets in the final state with @xmath198 .
the fraction of three - jet events selected by the @xmath197 and @xmath198 requirements was estimated using rapgap to be of the order of @xmath200 . + a detailed study of the main sources contributing to the systematic uncertainties of the measurements was performed @xcite .
the different contributions are listed below , with a typical value quoted for each : * the uncertainty in the hadronic final - state energy scale was taken into account by varying the cal ( fpc ) energy by @xmath201 ( @xmath202 ) in the mc , after removing the energy deposits belonging to the scattered positron .
this introduced an uncertainty in the measured distributions of about @xmath203 ; * the selected window in the reconstructed @xmath73 was varied by @xmath202 , both in data and simulated events , according to the estimated resolution in the reconstruction of this quantity .
the measured cross sections varied by @xmath203 ; * the uncertainty in the positron energy scale was included by varying the positron energy by @xmath204 in the mc . the effect on the @xmath205 and @xmath206 distributions ( see section 8) was typically below @xmath207 ; * the box cut applied to the position of the scattered positron in the rcal was varied by @xmath208 mm ( both in data and simulated events ) to account for the uncertainty in simulating the measured positron position .
the effect was typically below @xmath209 ; * the selected range in the reconstructed jet pseudorapidity in the laboratory frame , @xmath210 , ( see section 6.1 ) was varied by 0.2 units around the nominal value , both in data and mc simulation , according to the estimated resolution in the reconstruction of the jet direction .
this introduced an uncertainty on the measured cross sections of about @xmath211 ; * the measured distributions were corrected to the hadron level using rapgap - ps instead of rapgap - cdm to account for the uncertainty in the simulation of higher - order qcd processes in the mc model .
the effect varied between @xmath203 and @xmath65 and was the dominant systematic uncertainty .
the final systematic uncertainty was computed by adding the different contributions in quadrature . the contribution from proton - dissociative processes , @xmath212 , with @xmath213 gev , where the system @xmath126 escaped through the forward beam hole in the fpc ,
was estimated using epsoft to be @xmath214 and was subtracted .
the overall normalisation uncertainty of @xmath215 in the measured cross sections , coming from the uncertainty on the proton - dissociative contribution and the @xmath216 error in the luminosity determination , is not included in the figures .
multi - jet production in the centre - of - mass system of the hadronic final state , @xmath6 , is presented in the kinematic region defined in section 7 using the exclusive @xmath7-algorithm with @xmath178 .
the measurements were corrected for electroweak radiative effects to the born level .
the topology of the three - jet final state was studied using @xmath186 , the fractional - momentum variables ( see section 6.2 ) .
figure [ fig2](a ) shows the distribution of the three - jet sample in the ( @xmath217 ) plane .
as indicated in the figure , different regions in the plane correspond to different three - jet topologies .
all configurations are present , including those for which the three jets have similar energies .
the distribution of the events shows that , for the given @xmath159 , the favoured three - jet topologies are those for which one leading jet carries approximately twice the energy of either of the other two jets @xmath218 .
the normalised differential cross sections for three - jet production as a function of @xmath187 and @xmath219 are presented in figs . [
fig2](b ) and [ fig2](c ) , compared to the mc predictions .
the measured @xmath187 distribution has a peak structure with a maximum at @xmath220 , which is determined by the @xmath159 value used in defining jets since the position of the peak approximately corresponds to @xmath221 .
the shapes of the measured cross sections are reproduced by all three mc models .
this indicates that they provide a good description of the dynamics of the three - jet emission , which dictates the population of the ( @xmath222 ) plane .
+ the energy flow , as measured in the event plane defined by the two most - energetic jets in the event , was studied as a function of the azimuthal angle , @xmath223 .
the origin of @xmath223 was defined to be along the direction of the most - energetic jet with positive @xmath223 increasing in the direction of the second most - energetic jet .
the energy flow was measured by projecting the particle momenta onto the event plane and summing , in each @xmath223 bin , the particle energies .
the sums were normalised by dividing by @xmath73 . in fig .
[ fig3 ] , the normalised energy flow is presented for different regions of the ( @xmath224 ) plane . a clear three - jet structure is observed in all regions .
the measurements are compared with the rapgap and satrap - cdm predictions .
rapgap - cdm , rapgap - ps ( not shown but similar to rapgap - cdm ) and satrap - cdm all provide a very good description of the measured flows for all ( @xmath224 ) regions and @xmath223 ranges .
figure [ fig4](a , b ) shows the differential cross section as a function of the jet pseudorapidity , @xmath225 , measured with respect to the @xmath74 axis .
all jets in each event were included .
the total cross section for three - jet production , obtained by integrating the differential measurement , is : @xmath226 for this measurement , the overall normalisation uncertainty of @xmath215 ( see section 7 ) is combined in quadrature in the quoted total systematic uncertainty .
the measured cross section corresponds to @xmath227 of the diffractive cross section for two or more jets .
this rate is consistent with being independent of @xmath130 and @xmath73 .
+ figure [ fig4](a ) shows the measured differential cross section compared to the rapgap - cdm and rapgap - ps predictions .
both implementations of rapgap provide a good description of the shape of the cross section but underestimate it by about @xmath228 .
the observed deficit in the absolute normalisation predicted by rapgap may be attributed to the absence of next - to - leading - order terms in the computed matrix elements , which are only approximated via parton cascades .
+ in fig .
[ fig4](b ) the measured differential cross section is compared to the satrap , satrap - cdm and ridi predictions . for the latter ,
the contribution from the hadron jet associated with the gluon from the @xmath72 final state is shown separately .
the satrap and ridi predictions are found to be more than a factor of two below the measured cross sections and show a double - peak structure not observed in the data .
this tendency is particularly pronounced in ridi .
this reflects the presence of dominant topologies in these models in which the jet in the pomeron direction is emitted almost collinear to the @xmath74 axis and for which the second and third jets are found in the photon hemisphere .
the observed shape of the satrap and ridi predictions is a consequence of the inadequacies of the implementation of the modelling of higher - order qcd processes in these mc models as discussed in section 4 .
satrap - cdm describes the shape of the measured distribution but also underestimates the normalisation by about @xmath228 , which , as for rapgap , may be attributed to the absence of next - to - leading - order and some leading - order terms in the computed matrix elements .
the differential cross section as a function of the jet transverse momentum , @xmath176 , measured with respect to the @xmath74 axis for the most - forward jet is presented in fig .
[ fig4](c , d ) .
the measurement was performed in the region @xmath229 gev , where a good parton jet - hadron jet correlation is preserved , and for which the measurement is well defined , given the estimated resolution in the reconstruction of @xmath176 in the detector ( see section 6 ) .
the measured cross section is dominated by configurations with @xmath176 below 3 gev but contains a long tail towards larger @xmath176 values , reflecting the presence of contributions from higher - order qcd processes .
+ figure [ fig4](c ) shows the measured cross section compared to rapgap - cdm and rapgap - ps .
the rapgap predictions underestimate the measured cross section and have a slightly different shape than the data .
rapgap - cdm provides a better description of the data than rapgap - ps .
the observed difference between these two predictions indicates the sensitivity of the @xmath176 distribution to the approach used for the simulation of higher - order qcd processes in the mc .
+ figure [ fig4](d ) shows the measured @xmath176 distribution compared to the predictions from satrap , satrap - cdm and ridi . the @xmath176 distribution predicted by satrap and ridi is steeper than that of the data , indicating that , in these mc models , the large @xmath176 configurations are missing .
the cross section in these models is dominated by configurations strongly aligned with respect to the @xmath74 axis in the forward region . as already mentioned , this is most probably a consequence of the inadequacies of the implementation of the modelling of higher - order qcd processes in these models . in satrap - cdm ,
the shape and normalisation of the @xmath176 distribution is in better agreement with the data .
+ in all models , the gluon in the @xmath72 final state is emitted with very small transverse momentum with respect to the @xmath74 axis .
the contributions from higher - order qcd processes are expected to be important and to influence the kinematics of the corresponding hadron jet which , due to string effects , tends to have smaller pseudorapidity ( larger transverse momentum ) compared to the initial parton . given the @xmath230 requirement applied in defining the measured cross sections , the absolute predictions from these models depend on an appropriate treatment of these higher - order qcd emissions and a proper description of the forward hadron - jet dynamics ( as shown in fig .
[ fig4](d ) ) .
this is particularly clear in the case of satrap , which fails to reproduce the three - jet data but provides a good description of inclusive diffractive dis @xcite .
the differential jet shapes , @xmath231 , were measured for the most - forward and most - backward jet in three - jet events , where the forward region is defined by the pomeron direction .
the measurements were performed in the @xmath74 centre - of - mass frame for jets with energy , @xmath174 , above 9 gev .
this requirement removes the contributions from the least - energetic jet in the event which , in the @xmath73-range considered and for the given @xmath159 , has energy below this threshold and for which hadronisation effects on the jet shape are expected to be important .
figure [ fig5 ] shows the measured jet shapes compared with the rapgap - cdm predictions ( the satrap - cdm predictions , not shown , are almost identical ) .
the jet in the pomeron direction is broader than the jet in the photon direction .
these measurements are described by models in which a gluon populates the pomeron hemisphere and a quark is found in the photon direction .
the energy distribution of the two jets is similar , indicating that the observed difference in the measured jet shapes can be ascribed to the different nature of the initial partons .
moreover , the difference between the jet shapes in the pomeron and photon directions is qualitatively similar to that between quark- and gluon - initiated jets as measured in @xmath76 experiments @xcite .
this measurement therefore supports the validity of the picture where the three - body final state is dominated by a @xmath72 configuration with the gluon preferentially emitted in the pomeron direction .
multi - jet production has been studied in the kinematic region @xmath2 gev@xmath121 , @xmath4 gev , @xmath5 gev and @xmath196 , for diffractive events with @xmath197 and @xmath198 .
a study of three - jet production , using the exclusive @xmath7-algorithm with @xmath178 in the centre - of - mass system of the hadronic final state , has been presented .
the rate of three - jet production has been measured to be @xmath232 of the diffractive jet cross section , and is consistent with being independent of @xmath130 and @xmath73 .
the topologies of the three - jet final states have been studied in terms of the jet fractional - momentum variables @xmath233 and @xmath187 ; all possible topologies have been observed .
the distribution of the events in the ( @xmath217 ) plane shows that , for a given @xmath159 , the favoured topologies are those for which one leading jet carries approximately twice the energy of either of the other two jets @xmath218 .
differential cross sections for three - jet production have been measured as a function of the jet pseudorapidity , @xmath205 , and jet transverse momentum , @xmath176 , with respect to the @xmath74 axis .
the total cross section for three - jet production , determined by integrating the @xmath205 distribution , is @xmath234 .
the @xmath176 distribution for the most - forward jet is dominated by configurations with @xmath176 below 3 gev but exhibits a long tail towards larger values , reflecting the presence of higher - order qcd contributions .
the differential jet shapes , @xmath235 , of the two most - energetic jets in three - jet events have been measured .
the jet in the pomeron direction is broader and consistent with the shape of gluon - initiated jets as predicted by diffractive mc models .
+ the data are broadly consistent with models in which the hadronic final - state is dominated by a @xmath72 system with the gluon preferentially emitted in the pomeron direction .
such configurations are predicted both by resolved pomeron models ( as implemented in rapgap ) with a pomeron dominated by gluons , and by models where the virtual photon dissociates in a @xmath72 system which interacts with the proton via the exchange of a gluon ladder ( as implemented in satrap and ridi ) .
rapgap provides a reasonable description of the measured distributions , although it underestimates the measured cross section by about @xmath228 .
this deficit may be attributed to the absence of next - to - leading - order corrections , which are only included approximately via parton cascades .
the predictions from satrap and ridi fail to describe the shape and normalisation of the @xmath205 and @xmath176 distributions .
however , it has been shown that an improved implementation of the modelling of higher - order qcd processes in satrap - cdm gives reasonable agreement with the data .
thus , the study of three jets in diffraction constitutes a very sensitive method to investigate the dynamics of diffractive deep inelastic scattering . the strong support and encouragement of the desy directorate
have been invaluable .
the experiment was made possible by the inventiveness and the diligent efforts of the hera machine group .
the design , construction and installation of the zeus detector have been made possible by the ingenuity and dedicated efforts of many people from inside desy and from the home institutes who are not listed as authors .
their contributions are acknowledged with great appreciation .
we would like to thank the staff support from the various institutes which collaborated in the construction of the fpc and in the setup of the test systems , in particular j. hauschildt and k. lffler ( desy ) , l. hervs ( cern ) , r. feller , e. mller and h. prause ( i. inst . for exp .
hamburg ) , a. maniatis ( ii .
inst . for exp . phys . ,
hamburg ) , and the members of the mechanical workshop of the faculty of physics from freiburg university .
g. ingelman and p. schlein , phys .
b 152 ( 1985 ) 256 . see e.g. _ proceedings of the workshop on future physics at hera _
, g. ingelman , a. de roeck and r. klanner ( eds . ) , vol . 2 , desy , hamburg , ( 1996 ) , and references therein .
ryskin , sov .
52 ( 1990 ) 529 .
nikolaev and b.g .
zakharov , z. phys .
c 5 ( 1992 ) 331 ; + m. wsthoff , phys .
d 56 ( 1997 ) 4311 ; + w. buchmller , a. hebecker and m.f .
mcdermott , nucl .
b 487 ( 1997 ) 283 ; + j. bartels et al . , eur .
j. c 7 ( 1999 ) 443 .
h1 collaboration , s. aid et al . , z. phys .
c 70 ( 1996 ) 609 ; + h1 collaboration , c. adloff et al . , phys .
b 428 ( 1998 ) 206 ; + h1 collaboration , c. adloff et al . , eur . phys .
j. c 1 ( 1998 ) 495 ; + h1 collaboration , c. adloff et al .
j. c 6 ( 1999 ) 421 ; + h1 collaboration , c. adloff et al .
j. c 20 ( 2001 ) 29 .
zeus collaboration , m. derrick et al .
b 332 ( 1994 ) 228 ; + zeus collaboration , m. derrick et al .
, z. phys .
c 67 ( 1995 ) 227 ; + zeus collaboration , j. breitweg et al . , phys .
b 421 ( 1998 ) 368 . for a review of petra results , see e.g. : + s.l .
wu , phys .
107 ( 1984 ) 59 ; + b. naroska , phys . rep . 148 ( 1987 )
zeus collaboration , m. derrick et al .
b 293 ( 1992 ) 465 ; + zeus collaboration , m. derrick et al . , z. phys .
c 63 ( 1994 ) 391 .
zeus collaboration , u. holm ( ed . ) , _ the zeus detector _ , status report ( unpublished ) , desy , 1993 , available on + http://www-zeus.desy.de/bluebook/bluebook.html .
m. derrick et al .
instr . and meth . a 309 ( 1991 ) 77 ; + a. andresen et al .
instr . and meth . a 309 ( 1991 ) 101 ; + a. caldwell et al .
instr . and meth . a 321 ( 1992 ) 356 ; + a. bernstein et al . ,
instr . and meth . a 336 ( 1993 ) 23 .
n. harnew et al . , nucl .
instr . and meth . a 279 ( 1989 ) 290 ; + b. foster et al .
b 32 ( 1993 ) 181 ; + b. foster et al . ,
instr . and meth . a 338 ( 1994 ) 254 .
a. bamberger et al . ,
instr . and meth . a 450 ( 2000 ) 235
a. bamberger et al . , nucl .
instr . and meth . a 401 ( 1997 ) 63 .
j. andruszkw et al . , desy 01 - 041 ( 2001 ) ; + zeus collaboration , m. derrick et al .
, z. phys .
c 63 ( 1994 ) 391 .
rapgap ( version 2.08/06 ) , h. jung , comp .
phys . comm .
86 ( 1995 ) 147 . y. azimov et al . , phys .
b 165 ( 1985 ) 147 ; + g. gustafson , phys .
b 175 ( 1986 ) 453 ; + g. gustafson and u. petersson , nucl
b 306 ( 1988 ) 746 ; + b. andersson , g. gustafson and l. lnnblad , z. phys .
c 43 ( 1989 ) 625 .
& & & + @xmath159 & rate & ( stat .
@xmath236 syst . ) & rate & ( stat .
@xmath236 syst . ) & rate & ( stat .
@xmath236 syst . ) + 0.01 & @xmath237&@xmath238 & @xmath239&@xmath240 & @xmath241&@xmath242 + 0.02 & @xmath243&@xmath244 & @xmath245&@xmath246 & @xmath247&@xmath248 + 0.03 & @xmath249&@xmath250 & @xmath251&@xmath252 & @xmath253&@xmath254 + 0.04 & @xmath255&@xmath256 & @xmath257&@xmath258 & @xmath259&@xmath260 + 0.05 & @xmath261&@xmath262 & @xmath263&@xmath264 & @xmath265&@xmath260 + 0.06 & @xmath266&@xmath267 & @xmath268&@xmath269 & @xmath270&@xmath271 + 0.07 & @xmath272&@xmath273 & @xmath274&@xmath275 & - & - + 0.08 & @xmath276&@xmath277 & @xmath278&@xmath279 & - & - + 0.09 & @xmath280&@xmath281 & @xmath282&@xmath283 & - & - + 0.10 & @xmath284&@xmath285 & @xmath286&@xmath248 & - & - + @xmath287 & @xmath288 & ( stat . ) & ( stat .
@xmath236 syst . )
+ [ 0.650,0.720 ] & 0.288 & @xmath289 & @xmath290 + [ 0.720,0.790 ] & 1.639 & @xmath291 & @xmath292 + [ 0.790,0.860 ] & 4.869 & @xmath293 & @xmath294 + [ 0.860,0.930 ] & 6.339 & @xmath295 & @xmath296 + [ 0.930,1.000 ] & 1.150 & @xmath297 & @xmath298 + @xmath299 & @xmath300 & ( stat . ) & ( stat . @xmath236 syst . )
+ [ 0.0,0.2 ] & 2.189 & @xmath301 & @xmath302 + [ 0.2,0.4 ] & 1.516 & @xmath303 & @xmath304 + [ 0.4,0.6 ] & 1.153 & @xmath305 & @xmath306 + [ 0.6,0.8 ] & 0.142 & @xmath307 & @xmath308 + -3.016 & @xmath311&@xmath312 & @xmath313&@xmath314 & @xmath315&@xmath316 & @xmath317&@xmath318 + -2.765 & @xmath319&@xmath264 & @xmath320&@xmath321 & @xmath322&@xmath323 & @xmath324&@xmath325 + -2.513 & @xmath326&@xmath327 & @xmath328&@xmath329 & @xmath330&@xmath331 & @xmath332&@xmath333 + -2.262 & @xmath334&@xmath335 & @xmath336&@xmath337 & @xmath338&@xmath339 & @xmath340&@xmath341 + -2.011 & @xmath342&@xmath343 & @xmath344&@xmath345 & @xmath346&@xmath347 & @xmath348&@xmath349 + -1.759 & @xmath350&@xmath351 & @xmath352&@xmath353 & @xmath354&@xmath355 & @xmath356&@xmath357 + -1.508 & @xmath358&@xmath359 & @xmath360&@xmath361 & @xmath362&@xmath363 & @xmath364&@xmath365 + -1.257 & @xmath366&@xmath327 & @xmath367&@xmath279 & @xmath368&@xmath369 & @xmath322&@xmath370 + -1.005 & @xmath371&@xmath372 & @xmath373&@xmath374 & @xmath375&@xmath376 & @xmath377&@xmath378 + -0.754 & @xmath286&@xmath379 & @xmath368&@xmath380 & @xmath381&@xmath382 & @xmath317&@xmath383 + -0.503 & @xmath384&@xmath359 & @xmath385&@xmath386 & @xmath387&@xmath388 & @xmath389&@xmath390 + -0.251 & @xmath391&@xmath392 & @xmath393&@xmath394 & @xmath395&@xmath396 & @xmath397&@xmath398 + 0.000 & @xmath399&@xmath400 & @xmath401&@xmath402 & @xmath403&@xmath404 & @xmath405&@xmath406 + 0.251 & @xmath407&@xmath408 & @xmath409&@xmath410 & @xmath411&@xmath412 & @xmath413&@xmath414 + 0.503 & @xmath415&@xmath416 & @xmath417&@xmath418 & @xmath419&@xmath420 & @xmath421&@xmath422 + 0.754 & @xmath423&@xmath424 & @xmath425&@xmath426 & @xmath427&@xmath428 & @xmath429&@xmath273 + 1.005 & @xmath430&@xmath275 & @xmath367&@xmath431 & @xmath432&@xmath433 & @xmath373&@xmath434 + 1.257 & @xmath435&@xmath374 & @xmath436&@xmath437 & @xmath438&@xmath353 & @xmath439&@xmath440 + 1.508 & @xmath441&@xmath279 & @xmath367&@xmath442 & @xmath377&@xmath361 & @xmath360&@xmath443 + 1.759 & @xmath444&@xmath445 & @xmath446&@xmath437 & @xmath447&@xmath448 & @xmath449&@xmath450 + 2.011 & @xmath451&@xmath452 & @xmath453&@xmath258 & @xmath454&@xmath455 & @xmath456&@xmath457 + 2.262 & @xmath458&@xmath459 & @xmath460&@xmath461 & @xmath462&@xmath463 & @xmath464&@xmath465 + 2.513 & @xmath466&@xmath467 & @xmath468&@xmath469 & @xmath470&@xmath471 & @xmath472&@xmath473 + 2.765 & @xmath474&@xmath475 & @xmath476&@xmath477 & @xmath478&@xmath479 & @xmath328&@xmath480 + 3.016 & @xmath481&@xmath482 & @xmath483&@xmath484 & @xmath485&@xmath258 & @xmath486&@xmath238 + @xmath487 & @xmath488 ( pb ) & ( stat . ) ( pb ) & ( stat .
@xmath236 syst . )
( pb ) + [ -3.6,-3.0 ] & 0.650 & @xmath489&@xmath490 + [ -3.0,-2.4 ] & 2.167 & @xmath491&@xmath492 + [ -2.4,-1.8 ] & 4.279 & @xmath493&@xmath494 + [ -1.8,-1.2 ] & 8.193 & @xmath495&@xmath496 + [ -1.2,-0.6 ] & 10.322 & @xmath497&@xmath498 + [ -0.6,0.0 ] & 9.384 & @xmath499&@xmath500 + [ 0.0,0.6 ] & 7.132 & @xmath501&@xmath502 + [ 0.6,1.2 ] & 8.359 & @xmath503&@xmath504 + [ 1.2,1.8 ] & 9.417 & @xmath505&@xmath506 + [ 1.8,2.4 ] & 9.543 & @xmath507&@xmath508 + [ 2.4,3.0 ] & 1.039 & @xmath509&@xmath510 + @xmath511 ( gev ) & @xmath512 ( pb / gev ) & ( stat . ) ( pb / gev ) & ( stat . @xmath236
syst . ) ( pb / gev ) + [ 1.0,3.0 ] & 3.970 & @xmath513&@xmath514 + [ 3.0,5.0 ] & 2.291 & @xmath515&@xmath516 + [ 5.0,7.0 ] & 0.818 & @xmath517&@xmath518 + [ 7.0,9.0 ] & 0.172 & @xmath519&@xmath520 + & & + @xmath521 ( radians ) & @xmath231 & ( stat . ) & ( stat .
@xmath236 syst . ) & @xmath231 & ( stat . ) & ( stat .
@xmath236 syst . ) + [ 0.0,0.2 ] & 1.524&@xmath522&@xmath523 & 2.188 & @xmath524&@xmath525 + [ 0.2,0.4 ] & 1.500&@xmath526&@xmath527 & 1.273 & @xmath528&@xmath529 + [ 0.4,0.6 ] & 0.877&@xmath530&@xmath531 & 0.721 & @xmath532&@xmath533 + [ 0.6,0.8 ] & 0.473&@xmath534&@xmath535 & 0.350 & @xmath536&@xmath414 + [ 0.8,1.0 ] & 0.300&@xmath537&@xmath538 & 0.219 & @xmath539&@xmath540 + [ 1.0,1.2 ] & 0.165&@xmath541&@xmath542 & 0.125 & @xmath543&@xmath325 + [ 1.2,1.4 ] & 0.103&@xmath544&@xmath545 & 0.069 & @xmath546&@xmath547 + [ 1.4,1.6 ] & 0.047&@xmath548&@xmath549 & 0.025 & @xmath550&@xmath551 + | three - jet production in the reaction @xmath0 has been studied with the zeus detector at hera using an integrated luminosity of 42.74 pb@xmath1 .
the data were measured in the kinematic region @xmath2 gev@xmath3 , @xmath4 gev and @xmath5 gev .
the diffractive signal was selected by requiring a large rapidity gap in the outgoing proton direction .
jets were reconstructed in the centre - of - mass system of @xmath6 using the exclusive @xmath7-algorithm .
a sample of three - jet events in diffraction has been identified .
differential cross sections were measured as a function of the jet pseudorapidity and jet transverse momentum with respect to the virtual photon - pomeron axis .
the jets going in the pomeron direction are broader than those going in the virtual - photon direction .
this is consistent with models predicting that gluons are predominantly produced in the pomeron direction and quarks in the virtual - photon direction . -3.0
cm 2gev@xmath3 2q^2 3o(_s^3 ) -0.3 cm
-0.3 cm 16.cm 0.cm 0.3 cm plus0.05 cm minus0.05 cm the zeus collaboration s. chekanov , m. derrick , d. krakauer , s. magill , b. musgrave , a. pellegrino , j. repond , r. yoshida + _ argonne national laboratory , argonne , il , usa _
@xmath8 m.c.k .
mattingly + _ andrews university , berrien springs , mi , usa _ p. antonioli , g. bari , m. basile , l. bellagamba , d. boscherini@xmath9 , a. bruni , g. bruni , g. cara romeo , l. cifarelli@xmath10 , f. cindolo , a. contin , m. corradi , s. de pasquale , p. giusti , g. iacobucci , g. levi , a. margotti , t. massam , r. nania , f. palmonari , a. pesci , g. sartorelli , a. zichichi + _ university and infn bologna , bologna , italy _ @xmath11 g. aghuzumtsyan , i. brock , s. goers , h. hartmann , e. hilger , p. irrgang , h .-
p .
jakob , a. kappes@xmath12 , u.f .
katz@xmath13 , r. kerger , o. kind , e. paul , j. rautenberg , h. schnurbusch , a. stifutkin , j. tandler , k.c .
voss , a. weber , h. wieber + _ physikalisches
institut der universitt bonn , bonn , germany _
@xmath14 d.s .
bailey@xmath15 , n.h .
brook@xmath15 , j.e .
cole , b. foster , g.p .
heath , h.f .
heath , s. robins , e. rodrigues@xmath16 , j. scott , r.j .
tapper , m. wing + _ h.h .
wills physics laboratory , university of bristol , bristol , u.k .
_ @xmath17 m. capua , a. mastroberardino , m. schioppa , g. susinno + _ calabria university , physics dept.and infn , cosenza , italy _ @xmath11 h.y .
jeoung , j.y .
kim , j.h .
lee , i.t .
lim , k.j .
ma , m.y . pac@xmath18 + _ chonnam national university , kwangju , korea _
@xmath19 a. caldwell , m. helbich , w. liu , x. liu , b. mellado , s. paganis , s. sampson , w.b .
schmidke , f. sciulli + _ columbia university , nevis labs .
, irvington on hudson , n.y . , usa _ @xmath20 j. chwastowski , a. eskreys , j. figiel , k. klimek@xmath21 , k. olkiewicz , m.b .
przybycie@xmath22 , p. stopa , l. zawiejski + _ inst . of nuclear physics ,
cracow , poland _
@xmath23 b. bednarek , k. jele , d. kisielewska , a.m. kowal@xmath24 , m. kowal , t. kowalski , b. mindur , m. przybycie , e. rulikowska - zarbska , l. suszycki , d. szuba + _ faculty of physics and nuclear techniques , academy of mining and metallurgy , cracow , poland _ @xmath23 a. kotaski + _ jagellonian univ . , dept . of physics ,
cracow , poland _ l.a.t . bauerdick@xmath25 , u. behrens , k. borras , v. chiochia , j. crittenden@xmath26 , d. dannheim , k. desler , g. drews , , u. fricke , a. geiser , f. goebel , p. gttlicher , r. graciani , t. haas , w. hain , g.f .
hartner , k. hebbel , s. hillert , w. koch@xmath27@xmath28 , u. ktz , h. kowalski , h. labes , b. lhr , r. mankel , j. martens , m. milite , m. moritz , d. notz , m.c .
petrucci , a. polini , , f. selonke , s. stonjek , g. wolf , u. wollmer , j.j .
whitmore@xmath29 , r. wichmann@xmath30 , c. youngman , + _ deutsches elektronen - synchrotron desy , hamburg , germany _ c. coldewey , , a. meyer , + _ desy zeuthen , zeuthen , germany _ g. barbagli , e. gallo , p. g. pelfer + _ university and infn , florence , italy _ @xmath11 a. bamberger , a. benen , n. coppola , p. markun , h. raach@xmath31 , s. wlfle + _ fakultt fr physik der universitt freiburg i.br .
, freiburg i.br .
, germany
_ @xmath14 m. bell , p.j .
bussey , a.t .
doyle , c. glasman , s.w .
lee@xmath32 , a. lupi , g.j .
mccance , d.h .
saxon , i.o .
skillicorn + _ dept . of physics and astronomy , university of glasgow , glasgow , u.k .
_ @xmath17 b. bodmann , n. gendner , u. holm , h. salehi , k. wick , a. yildirim , a. ziegler + _ hamburg university , i. institute of exp .
physics , hamburg , germany _
@xmath14 t. carli , a. garfagnini , i. gialas@xmath33 , e. lohrmann + _ hamburg university , ii .
institute of exp .
physics , hamburg , germany _
@xmath14 c. foudas , r. gonalo@xmath16 , k.r .
long , f. metlica , d.b .
miller , a.d .
tapper , r. walker + _ imperial college london , high energy nuclear physics group , london , u.k .
_ @xmath17 p. cloth , d. filges + _ forschungszentrum jlich , institut fr kernphysik , jlich , germany _ m. kuze , k. nagano , k. tokushuku@xmath34 , s. yamada , y. yamazaki + _ institute of particle and nuclear studies , kek , tsukuba , japan
_ @xmath35 a.n .
barakbaev , e.g. boos , n.s .
pokrovskiy , b.o .
zhautykov + _ institute of physics and technology of ministry of education and science of kazakhstan , almaty , kazakhstan _ s.h .
ahn , s.b .
lee , s.k .
park + _ korea university , seoul , korea _
@xmath19 h. lim@xmath32 , d. son + _ kyungpook national university , taegu , korea _
@xmath19 f. barreiro , g. garca , o. gonzlez , l. labarga , j. del peso , i. redondo@xmath36 , j. terrn , m. vzquez + _ univer .
autnoma madrid , depto de fsica terica , madrid , spain _
@xmath37 m. barbi , a. bertolin , f. corriveau , a. ochs , s. padhi , d.g .
stairs + _ mcgill university , dept . of physics , montral , qubec , canada _
@xmath38 t. tsurugai + _ meiji gakuin university , faculty of general education , yokohama , japan _
a. antonov , v. bashkirov@xmath39 , p. danilov , b.a .
dolgoshein , d. gladkov , v. sosnovtsev , s. suchkov + _ moscow engineering physics institute , moscow , russia _ @xmath40 r.k .
dementiev , p.f .
ermolov , yu.a .
golubkov , i.i .
katkov , l.a .
khein , n.a .
korotkova , i.a .
korzhavina , v.a .
kuzmin , b.b .
levchenko , o.yu .
lukina , a.s .
proskuryakov , l.m .
shcheglova , a.n .
solomin , n.n .
vlasov , s.a .
zotkin + _ moscow state university , institute of nuclear physics , moscow , russia _
@xmath40 c. bokel , j. engelen , s. grijpink , e. maddox , e. koffeman , p. kooijman , s. schagen , e. tassi , h. tiecke , n. tuning , j.j .
velthuis , l. wiggers , e. de wolf + _ nikhef and university of amsterdam , amsterdam , netherlands _ @xmath41 n. brmmer , b. bylsma , l.s .
durkin , j. gilmore , c.m .
ginsburg , c.l .
kim , t.y . ling + _ ohio state university , physics department , columbus , ohio , usa _ @xmath8 s. boogert , a.m. cooper - sarkar , r.c.e .
devenish , j. ferrando , j. groe - knetter@xmath42 , t. matsushita , m. rigby , o. ruske@xmath43 , m.r .
sutton , r. walczak + _ department of physics , university of oxford , oxford u.k .
_ @xmath17 r. brugnera , r. carlin , f. dal corso , s. dusini , s. limentani , a. longhin , a. parenti , m. posocco , l. stanco , m. turcato + _ dipartimento di fisica dell universit and infn , padova , italy _ @xmath11 l. adamczyk@xmath44 , l. iannotti@xmath44 , b.y .
oh , p.r.b .
saull@xmath44 , w.s .
toothacker@xmath27@xmath28 + _ pennsylvania state university , dept . of physics ,
university park , pa , usa _
@xmath20 y. iga + _ polytechnic university , sagamihara , japan
_ @xmath35 g. dagostini , g. marini , a. nigro + _ dipartimento di fisica , univ .
la sapienza and infn , rome , italy _ @xmath45 c. cormack , j.c .
hart , n.a .
mccubbin + _ rutherford appleton laboratory , chilton , didcot , oxon , u.k .
_ @xmath17 d. epperson , c. heusch , h.f .- w .
sadrozinski , a. seiden , d.c .
williams + _ university of california , santa cruz , ca , usa _
@xmath8 i.h .
park + _ seoul national university , seoul , korea _
n. pavel + _ fachbereich physik der universitt - gesamthochschule siegen , germany _
@xmath14 h. abramowicz , s. dagan , a. gabareen , s. kananov , a. kreisel , a. levy + _ raymond and beverly sackler faculty of exact sciences , school of physics , tel - aviv university , tel - aviv , israel _
@xmath46 t. abe , t. fusayasu , t. kohno , k. umemori , t. yamashita + _ department of physics , university of tokyo , tokyo , japan _
@xmath35 r. hamatsu , t. hirose , m. inuzuka , s. kitamura@xmath47 , k. matsuzawa , t. nishimura + _ tokyo metropolitan university , dept . of physics , tokyo , japan
_ @xmath35 m. arneodo@xmath48 , n. cartiglia , r. cirio , m. costa , m.i .
ferrero , s. maselli , v. monaco , c. peroni , m. ruspa , r. sacchi , a. solano , a. staiano + _ universit di torino , dipartimento di fisica sperimentale and infn , torino , italy _ @xmath11 d.c .
bailey , c .- p .
fagerstroem , r. galea , t. koop , g.m .
levman , j.f .
martin , a. mirea , a. sabetfakhri + _ university of toronto , dept . of physics , toronto , ont . ,
canada _ @xmath38 j.m .
butterworth , c. gwenlan , r. hall - wilton , m.e .
hayes@xmath42 , e.a .
heaphy , t.w .
jones , j.b .
lane , m.s .
lightwood , b.j .
west + _ university college london , physics and astronomy dept .
, london , u.k .
_ @xmath17 j. ciborowski@xmath49 , r. ciesielski , g. grzelak , r.j .
nowak , j.m .
pawlak , b. smalska@xmath50 , t. tymieniecka@xmath51 , j. ukleja@xmath51 , j.a .
zakrzewski , a.f .
arnecki + _ warsaw university , institute of experimental physics , warsaw , poland _ @xmath23 m. adamus , p. plucinski , j. sztuk + _ institute for nuclear studies , warsaw , poland _
@xmath23 o. deppe@xmath52 , y. eisenberg , l.k .
gladilin@xmath53 , d. hochman , u. karshon + _ weizmann institute , department of particle physics , rehovot , israel _ @xmath54 j. breitweg , d. chapin , r. cross , d. kira , s. lammers , d.d .
reeder , a.a .
savin , w.h .
smith + _ university of wisconsin , dept . of physics , madison ,
wi , usa _
@xmath8 a. deshpande , s. dhawan , v.w .
hughes p.b .
straub + _ yale university , department of physics , new haven , ct , usa _
@xmath8 s. bhadra , c.d .
catterall , w.r .
frisken , m. khakzad , s. menary + _ york university , dept . of physics , toronto , ont . ,
canada _
@xmath38 @xmath55 now visiting scientist at desy + @xmath56 now at univ . of salerno and infn napoli , italy + @xmath57 supported by the gif , contract i-523 - 13.7/97 + @xmath58 on leave of absence at university of erlangen - nrnberg ,
germany + @xmath59 pparc advanced fellow + @xmath60 supported by the portuguese foundation for science and technology ( fct ) + @xmath61 now at dongshin university , naju , korea + @xmath62 supported by the polish state committee for scientific research , grant no . 5 p-03b 08720 + @xmath63 now at northwestern univ .
, evaston / il , usa + @xmath24 supported by the polish state committee for scientific research , grant no .
5 p-03b 13720 + @xmath25 now at fermilab , batavia / il , usa + @xmath26 on leave of absence from bonn university + @xmath27 deceased + @xmath29 on leave from penn state university , usa + @xmath30 partly supported by penn state university and gif , contract i-523 - 013.07/97 + @xmath31 supported by desy + @xmath32 partly supported by an icsc - world laboratory bjrn h. wiik scholarship + @xmath33 univ . of the aegean , greece + @xmath34 also at university of tokyo + @xmath36 supported by the comunidad autonoma de madrid + @xmath39 now at loma linda university , loma linda , ca , usa + @xmath42 now at cern , geneva , switzerland + @xmath43 now at ibm global services , frankfurt / main , germany + @xmath44 partly supported by tel aviv university + @xmath47 present address : tokyo metropolitan university of health sciences , tokyo 116 - 8551 , japan + @xmath48 now also at universit del piemonte orientale , i-28100 novara , italy + @xmath49 and d university , poland + @xmath50 supported by the polish state committee for scientific research , grant no .
2 p-03b 00219 + @xmath51 supported by the polish state committee for scientific research , grant no .
5 p-03b 09820 + @xmath52 now at evotec biosystems ag , hamburg , germany + @xmath53 on leave from msu , partly supported by university of wisconsin via the u.s .- israel bsf + [ cols= " > , < " , ] -3.0 cm |
one of the fundamental goals of modern cosmology is to probe einstein s general relativity ( gr ) at any scale beyond the classical tests that confirmed such a theory in the weak field limit and at solar system level .
gr is assumed as the standard theory of gravity describing astrophysical structures up to the whole observed universe ; however there are some inconsistencies at ultraviolet scales ( e.g. the initial singularity , the quantum gravity issue ) and infrared scales ( e.g. cosmic acceleration , concordance problem , flatness problem , galaxy rotation curves , large scale structure , massive stars formation ) that strongly suggest that einstein s approach should be revised or at least extended .
furthermore , astrophysical observations of the last decades suggest that new ( dark ) ingredients are necessary to achieve a self - consistent cosmological model .
in particular , the observations suggest the hubble flow is currently accelerating , and the simplest way to explain the cosmic acceleration is to insert a cosmological constant ( @xmath5 ) in the friedmann - robertson - walker cosmology @xcite , representing about @xmath6 of the total amount of energy . on the other hand ,
the galaxy rotation curves and the large scale structure could be dynamically addressed by introducing huge amounts of dark matter ( about the 25@xmath7 of the total matter ) .
only 5@xmath7 of the cosmic budget is constituted by standard matter as stars , neutrinos , radiation , heavy elements and free cosmological hydrogen ad helium .
alternative approaches to gr could be pursued with the aim to explain the observed acceleration and missing matter without introducing new ingredients up to now not observed at fundamental scales .
the so - called @xmath0-gravity is considered as a possible , straightforward mechanism to explain the cosmic acceleration without inserting unknown elements as dark energy and dark matter , but extending the geometric part of the field equations by relaxing the strict hypothesis that the gravitational action has to be restricted to @xmath4 as in the hilbert - einstein one @xcite .
these theories have been investigated both at cosmological scales and in the weak field limit @xcite .
it has been shown that a late accelerating behavior can be easily recovered @xcite and it can be related to an early inflationary expansion @xcite . furthermore , modifying the gravity action by assuming non - linear lagrangians , one obtain corrections to the gravitational potential which can be useful for astrophysical phenomenology at galactic scales .
in particular , without the introduction of dark matter , the rotation curves of spiral galaxies and the haloes of galactic clusters can be dynamically addressed @xcite .
several of these extended models reproduce solar system tests so they are not in conflict with gr experimental results but simply extend them @xcite .
it is important to stress that @xmath0 gravity has interesting applications also in stellar astrophysics and could contribute to solve several puzzles related to observed peculiar objects ( e.g. magnetars , stars in the instability strips , protostars , etc .
@xcite ) , structure and star formation @xcite . here
we analyze the jeans instability for self - gravitating systems in @xmath0-gravity coupled with perfect - fluid matter .
the aim is to show that several self - gravitating systems , in particular those involved in star formation ( e.g. large molecular clouds or bok globules ) , can be exactly addressed in this framework by considering the corrections to the newtonian potential coming out from @xmath0-gravity .
this fact could constitute a remarkable signature to retain or rule out these theories at astrophysical level .
the paper is organized as follows . in section [ due ]
the classical theory of gravitational collapse for dust - dominated systems is summarized . in section [ tre ]
, we discuss the weak field limit of @xmath0-gravity obtaining corrections to the standard newtonian potential that can be figured out as two newtonian potentials contributing to the dynamics . in section [ quattro ]
we recover the dispersion relation and jeans mass limit while , in section [ cinque ] , some self - gravitating dust system are considered in this approach . the difference between gr and @xmath0-gravity
are put in evidence , in particular the jeans mass profiles with respect to the temperature .
we report a catalogue of observed molecular clouds in order to compare the classical jeans mass with the @xmath0 one . finally , in section [ sei ] , results are discussed .
the collapse of self - gravitational collisionless systems can be dealt with the introduction of coupled collisionless boltzmann and poisson equations ( for details , see @xcite ) : @xmath8 @xmath9 where @xmath10 and @xmath11 mean 3-dimensional vectors in the spatial manifold .
a self - gravitating system at equilibrium is described by a time - independent distribution function @xmath1 and a potential @xmath2 that are solutions of eq . and . considering a small perturbation to this equilibrium : @xmath12 where @xmath13 and by substituting in eq . and and by linearizing , one obtains : @xmath14 @xmath15 since the equilibrium state is assumed to be homogeneous and time - independent , one can set @xmath16 , and the so - called jeans `` swindle '' to set @xmath17 . in fourier components , eqs . and become : @xmath18 by combining these equations , the dispersion relation @xmath19 is obtained . in the case of stellar systems , by assuming a maxwellian distribution function for @xmath20
, we have @xmath21 imposing that @xmath22 and substituting in eq .
, one gets : @xmath23 by setting @xmath24 , the limit for instability is obtained : @xmath25 by which it is possible to define the jeans mass ( @xmath26 ) as the mass originally contained within a sphere of diameter @xmath27 : @xmath28 where @xmath29 is the jeans length . substituting eq . into eq . , we recover @xmath30 all perturbations with wavelengths @xmath31 are unstable in the stellar system . in order to evaluate the integral in the dispersion relation , we have to study the singolarity at @xmath32 .
to this end , it is useful to write the dispersion relation as @xmath33 defining @xmath34 where @xmath35 and @xmath36 .
we set also @xmath37 and @xmath38=0 $ ] , because we are interested in the unstable modes .
these modes appear when the imaginary part of @xmath39 is greater than zero and in this case the integral in the dispersion relation can be resolved just with previous prescriptions . in order to study unstable modes ( for details , see appendix b in @xcite ) we replace the following identities @xmath40\ , , \\ \\ { \rm{erf } } z = \dfrac{2}{{\sqrt \pi } } \int\limits_0^z { e^ { - t^2 } dt}.
\end{array } \right.\ ] ] into the dispersion relation obtaining : @xmath41 } \right\}}.\ ] ] this is the standard dispersion relation describing the criterion to collapse for infinite homogeneous fluid and stellar systems @xcite .
as discussed in the introduction , @xmath0-gravity is a straightforward extension of gr by which it is possible , in principle , to recover good results of gr without imposing _ a priori _ the form of gravitational lagrangian , chosen to be @xmath4 by hilbert and einstein .
this means that we do not impose a priori the gravitational action but it can be , in principle , re - constructed by generic curvature invariants and then matched with observations ( the simplest choice in this sense is to take into account an analytic function of the ricci scalar @xmath42 @xcite ) .
however , from a genuine mathematical viewpoint , the initial value problem of such theories has to be carefully addressed in order to achieve self - consistent results ( see for example @xcite ) .
let us start with a general class of higher order theories given by the action @xmath43\,,\end{aligned}\ ] ] where @xmath0 is an analytic function of curvature invariant @xmath42 and @xmath44 is the usual coupling of gravitational field equations @xcite .
the term @xmath45 is the minimally coupled ordinary matter contribution . in the metric approach ,
the field equations are obtained by varying ( [ fogaction ] ) with respect to @xmath46 .
we get : @xmath47 with the trace equation @xmath48 here , @xmath49 is the the energy - momentum tensor of matter , while @xmath50 is the trace , @xmath51 and @xmath52 .
the convention for ricci s tensor is @xmath53 while for the riemann tensor is @xmath54 .
the affinities are the usual christoffel symbols of the metric @xmath55 . ] .
the signature is @xmath56 @xcite ) . for our purposes
, we have to start by setting the right approximation in the metric tensor @xmath46 @xcite : @xmath57 where @xmath58 ( with @xmath59integer ) denotes the order of the expansion .
it is worth stressing that the expansion parameter is @xmath60 and , in the newtonian limit , we are assuming perturbations up to @xmath61 .
this means that in the above expression ( [ metric_tensor_ppn ] ) , we can discard terms of order @xmath62 and @xmath63 that have to be considered in further perturbation post - newtonian limit ( see @xcite and references therein ) .
the set of coordinates and @xmath64 ; the latin index between @xmath65 and @xmath64 .
] adopted is @xmath66 .
the ricci scalar becomes @xmath67 the @xmath68 derivative of ricci function can be developed as @xmath69 here @xmath70 denotes a quantity of order @xmath58 .
it is worth stressing that the symbol @xmath71 means the @xmath68 derivative of the analytic function @xmath0 . in the following
, we are going to use the numbers @xmath72 of the taylor series .
from lowest order of field equations ( [ fieldequationfog ] ) , we have @xmath73 which trivially follows from the above assumption ( [ metric_tensor_ppn ] ) that the space - time is asymptotically minkowskian .
eqs.([fieldequationfog ] ) and ( [ trace ] ) at @xmath74-order ( newtonian level ) become : @xmath75 where @xmath76 is the laplacian in the flat space , @xmath77 and for the sake of simplicity , we set @xmath78 .
we recall that the energy - momentum tensor for a perfect fluid is the @xmath79 where @xmath80 is the pressure and @xmath81 is the energy density . if we consider a perfect fluid of dust ( @xmath82 ) , we have @xmath83 @xcite
. then we have @xmath84 where @xmath85 is the mass density . ] . for @xmath86 , the standard poisson equation @xmath87
is recovered .
the solution for the gravitational potential @xmath88 has a yukawa - like behavior depending on a characteristic length .
then as it is evident , the gauss theorem is not valid since the force law is not @xmath89 . the equivalence between a spherically symmetric distribution and point - like distribution is not valid and how the matter is distributed in the space is very important . besides the birkhoff theorem at newtonian level
is modified : the solution can be only factorized with a space - depending function and an arbitrary time - depending function . furthermore , the correction to the gravitational potential is depending on the first two derivatives of @xmath0 in @xmath90 .
so different analytical models , up to the third derivative , admit the same newtonian general solution . field equations ( [ hoeq ] ) and ( [ hoeqtra ] ) give rise to the _ modified poisson equations _ for @xmath0-gravity .
we know that @xmath91 inserting in the above result the @xmath46 approximations we obtain @xmath92 where @xmath93 is the further gravitational potential related to the metric component @xmath94 . substituting in eqs .
( [ hoeq ] ) and ( [ hoeqtra ] ) , we obtain @xmath95 by eliminating the higher - order terms , the standard poisson equation is recovered .
our task is to check how the jeans instability occurs in @xmath0-gravity .
an important consideration is in order at this point . as we pointed out above ,
we are supposing that the space - time is asymptotically minkowski .
however , this is against the general idea of @xmath0-gravity which should mimic dark energy behavior .
this means that the space - time should be asymptotically de sitter .
so , in general , it is necessary that the @xmath0 function is expandable at @xmath96 , or even if it is , the interesting asymptotic space is nevertheless not minkowskian ( @xmath96 ) but @xmath97 .
this fact is connected with the assumption that the energy - density @xmath85 is homogenous and asymptotically constant in order to leads to de sitter space - time .
this is , at the very end , why the so called `` jeans swindle '' is needed in newtonian theory . in the present case , @xmath85 is explicitly written in eqs .
( [ hoeq ] ) and ( [ hoeqtra ] ) and has to converge asymptotically to zero in order to restore the asymptotic minkowskian behavior . in other words , the possible gravitational actions have to be chosen so that the condition @xmath98 holds .
our task is now to study the jeans instability in the framework of @xmath0 gravity .
let us assume the standard collisionless boltzmann equation : @xmath99 where , according to the newtonian theory , only the potential @xmath88 is present .
considering the @xmath0 poisson equations , given by eqs . and
, also the potential @xmath93 has to be considered so we obtain the coupled equations @xmath100 @xmath101 in the previous equations , we have replaced @xmath102 with the greek letter @xmath103 .
it is important to stress that while in the standard theory , the distribution function @xmath104 is related only to the potential @xmath88 , it is related to both @xmath88 and @xmath93 in the newtonian limit coming from @xmath0 gravity . as in standard case , we consider small perturbation to the equilibrium and linearize the equations .
after we write equations in fourier space so they became @xmath105 combining eqs . and
, we obtain a relation between @xmath106 and @xmath107 , @xmath108 inserting this relation in eq . and combining it with eq . , we obtain the dispersion relation @xmath109 if we assume , as in standard case , that @xmath20 is given by and @xmath110 , one can write @xmath111 = 0}.\ ] ] by eliminating the higher - order terms ( imposing @xmath112 ) , we obtain again the standard dispersion eq . .
in order to compute the integral in the dispersion relation , we consider the same approach used in the classical case , and finally we obtain : @xmath113 } \right ) } \right ] = 0 , \end{array}}\ ] ] where @xmath114 and @xmath115 . in order to evaluate eq .
comparing it with the classical one , given by eq . , it is very useful to normalize the equation to the classical jeans length showed in eq .
, by fixing the parameter of @xmath0-gravity , that is @xmath116 this parameterization is correct because the dimension @xmath103 ( an inverse of squared length ) allows us to parametrize as in standard case .
finally we write @xmath117 } \right ) } \right ] = 0 .
\end{array}\ ] ] the function is plotted in fig.[fig : jeans ] , where eq . and
the standard dispersion @xcite are confronted in order to see the difference between @xmath0 and newtonian gravity
. given by .
the thin line indicates the plot of the standard dispersion equation @xcite.,title="fig : " ] + as shown in fig .
[ fig : jeans ] , the effects of a different theory of gravity changes the limit of instability .
the limit is higher than the classical case and the curve has a greater slope .
this fact is important because the mass limit value of interstellar clouds decreases changing the initial conditions to start the collapse .
a numerical estimation of the @xmath0-instability length in terms of the standard newtonian one can be achieved . by solving numerically eq . with the condition @xmath24
, we obtain that the collapse occurs for @xmath118 however we can estimate also analytically the limit for the instability . in order to evaluate the jeans mass limit in @xmath0-gravity , we set @xmath24 in eq . and then @xmath119 it is worth stressing that the additional condition @xmath120 discriminates the class of viable @xmath0 models : in such a case we obtain stable cosmological solution and positively defined massive states @xcite .
in other words , this condition selects the physically viable models allowing to solve eq . for real values of k. in particular ,
the above numerical solution can be recast as @xmath121 the relation to the newtonian value of the jeans instability is @xmath122 now , we can define the new jeans mass as : @xmath123 that is proportional to the standard newtonian value
. we will confront this specific solutions with some observed structures .
star formation is one of the best settled problems of modern astrophysics .
however , some shortcomings emerge as soon as one faces dynamics of diffuse gas evolving into stars and star formation in galactic environment .
one can deal with the star formation problem in two ways : @xmath125 we can take into account the formation of individual stars and @xmath126 we can discuss the formation of the whole star system starting from interstellar clouds @xcite . to answer these problems
it is very important to study the interstellar medium ( ism ) and its properties.the ism physical conditions in the galaxies change in a very wide range , from hot x - ray emitting plasma to cold molecular gas , so it is very complicated to classify the ism by its properties .
however , we can distinguish , in the first approximation , between @xcite : * * diffuse hydrogen clouds*. the most powerful tool to measure the properties of these clouds is the 21 cm line emission of hi .
they are cold clouds so the temperature is in the range @xmath127 k , and their extension is up to @xmath128 kpc from galactic center . *
* diffuse molecular clouds * are generally self - gravitating , magnetized , turbulent fluids systems , observed in sub - mm .
the most of the molecular gas is @xmath129 , and the rest is co. here , the conditions are very similar to the hi clouds but in this case , the cloud can be more massive .
they have , typically , masses in the range @xmath130 , temperature in @xmath131k and particle density in @xmath132 m@xmath133 . *
* giant molecular clouds * are very large complexes of particles ( dust and gas ) , in which the range of the masses is typically @xmath134 but they are very cold .
the temperature is @xmath135k , and the number of particles is @xmath136 m@xmath133 @xcite .
however , there exist also small molecular clouds with masses @xmath137@xcite .
they are the best sites for star formation , despite the mechanism of formation does not recover the star formation rate that would be @xmath138 @xcite . * * hii regions*. they are ism regions with temperatures in the range @xmath139 k , emitting primarily in the radio and ir regions . at low frequencies , observations are associated to free - free electron transition ( thermal bremsstrahlung ) .
their densities range from over a million particles per @xmath140 in the ultra - compact h ii regions to only a few particles per @xmath140 in the largest and most extended regions .
this implies total masses between @xmath141 and @xmath142 m@xmath143 @xcite . *
* bok globules * are dark clouds of dense cosmic dust and gas in which star formation sometimes takes place .
bok globules are found within h ii regions , and typically have a mass of about @xmath144 to @xmath145 m@xmath143 contained within a region of about a light year .
using very general conditions @xcite , we want to show the difference in the jeans mass value between standard and @xmath0-gravity .
let us take into account eq . and eq .
: @xmath146 in which @xmath147 is the ism density and @xmath148 is the velocity dispersion of particles due to the temperature .
these two quantities are defined as @xmath149 where @xmath150 is the number of particles measured in @xmath151 , @xmath152 is the mean molecular weight , @xmath153 is the boltzmann constant and @xmath154 is the proton mass . by using these relations , we are able to compute the jeans mass for interstellar clouds and to plot its behavior against the temperature .
results are shown in tab.[table1 ] and fig.[fig : mass ] .
we have plotted the relation for gr and for @xmath0 gravity .
any astrophysical system reported in tab.[table1 ] is associated to a particular @xmath155-region .
differences between the two theories for any self - gravitating system are clear . -@xmath124 relation .
dashed - line indicates the newtonian jeans mass behavior with respect to the temperature .
continue - line indicates the same for @xmath0-gravity jeans mass . ] by using eq . and by referring to the catalog of molecular clouds in roman - duval j. et al .
@xcite , we have calculated the jeans mass in the newtonian and @xmath0 cases .
tab.[table2 ] shows the results . in all cases
we note a substantial difference between the classical and @xmath0 value . in @xmath0 scenario ,
molecular clouds become sites where star formation is strongly supported and more efficient because in each of them the limit for the gravitational collapse is lower than the one in gr .
@xmath0-gravity is an approach aimed to address some shortcomings of modern cosmology just assuming extensions of gr without invoking the presence of dark ingredients . in other words , dark energy and dark matter
could be effects related to curvature further degrees of freedom instead of new fundamental particles .
here we have analyzed the jeans instability mechanism , adopted for star formation , considering the newtonian approximation of @xmath0-gravity .
the related boltzmann - vlasov system leads to modified poisson equations depending on the @xmath0-model .
in particular , considering eqs . and , it is possible to get a new dispersion relation where instability criterion results modified ( see also @xcite ) .
the leading parameter is @xmath103 , i.e. the second derivative of the specific @xmath0-model .
standard newtonian jeans instability is immediately recovered for @xmath112 corresponding to the hilbert - einstein lagrangian of gr . in fig .
[ fig : jeans ] , dispersion relations for newtonian and a specific @xmath0-model are numerically compared .
the modified characteristic length van be given in terms of the classical one .
both in the classical and in @xmath0 analysis , the system damps the perturbation .
this damping is not associated to the collisions because we neglect them in our treatment , but it is linked to the so called landau damping @xcite .
a new condition for the gravitational instability is derived , showing unstable modes with faster growth rates .
finally we can observe the instability decrease in @xmath0-gravity : such decrease is related to a larger jeans length and then to a lower jeans mass .
we have also compared the behavior with the temperature of the jeans mass for various types of interstellar molecular clouds ( fig .
[ fig : mass ] ) . in tables
[ table1 ] and [ table2 ] we show the results given by this new limit of the jeans mass for a sample of giant molecular clouds . in our model
the limit ( in unit of mass ) to start the collapse of an interstellar cloud is lower than the classical one advantaging the structure formation .
real solutions for the jean mass can be achieved only for @xmath120 and this result is in agreement with cosmology @xcite . in particular , the condition @xmath120 is essentials to have a well - formulated and well - posed cauchy problem in @xmath0-gravity @xcite .
finally , it is worth noticing that the newtonian value is an upper limit for the jean mass coinciding with @xmath4 .
this work is intended to indicate the possibility to deal with ism collapsing clouds under different assumptions about gravity .
it is important to stress that we fully recover the standard collapse mechanisms but we could also describe proto - stellar systems that escape the standard collapse model . on the other hand , this is the first step to study star formation in alternative theories of gravity ( see also @xcite ) . from an observational point of view
, reliable constraints can be achieved from a careful analysis of the proto - stellar phase taking into account magnetic fields , turbulence and collisions .
finally , addressing stellar systems by this approach could be an extremely important to test observationally @xmath0-gravity .
moreover , the approach developed in this work admits direct generalizations for other modified gravities , like non - local gravity , modified gauss - bonnet theory , string - inspired gravity , etc . in these cases
, the constrained poisson equation may be even more complicated due to the presence of extra scalar(s ) in non - local or string - inspired gravity .
developing further this approach gives , in general , the possibility to confront the observable dynamics of astrophysical objects ( like stars ) with predictions of alternative gravities .
s. capozziello , m. de laurentis , v. faraoni , _ the open astr .
jour _ , * * 2**1874 , ( 2009 ) ; nojiri s. , odintsov s.d . ,
_ , * 505 * , 59 - 144 , ( 2011 ) ; s. capozziello , v. faraoni , _ beyond einstein gravity : a survey of gravitational theories for cosmology and astrophysics _ , springer , new york , ( 2010 ) ; t. p. sotiriou , v. faraoni , _ rev . mod .
phys . _ * 82 * , 451 ( 2010 ) ; s. capozziello , m. de laurentis , _ invariance principles and extended gravity : theory and probes _ , nova science publishers , new york , ( 2011 ) ; t. clifton , p. g. ferreira , a. padilla , c. skordis , arxiv:1106.2476 [ astro-ph.co ] ( 2011 ) .
s. capozziello , m. de laurentis , s. nojiri , s.d .
odintsov , _ gen .
* 41 * , 2313 , ( 2009 ) ; _ physical review d _ * 79 * , 124007 , ( 2009 ) ; m. de laurentis , s. capozziello , _ nuclear physics b ( proceedings supplements ) _ , 212 , ( 2009 ) ; m. de laurentis , s. capozziello , s. nojiri , s.d .
odintsov , _ annalen der physik _ * 19 * , 347 , ( 2009 ) ; s. capozziello , n. carlevaro , m. de laurentis , m. lattanzi , g. montani , arxiv:1104.2169v1 [ astro-ph.co ] ( 2011 ) ; o.m .
lecian , g. montani , _ class .
* 26 * , 045014 , ( 2009 ) .
m. de laurentis , s. capozziello , _ astropart .
phys . _ * 35 * 257 ( 2011 ) .
s. capozziello , r. cianci , m. de laurentis , s. vignolo s.,_astroph . and
space science _ * 70 * , 341 , ( 2010 ) ; s. capozziello , m. de laurentis , m. francaviglia _ astrop .
phys . _ * 29*,125 , ( 2008 ) ; s. capozziello , c. corda , m. de laurentis , _ phys . lett .
b _ * 669 * , 255 , ( 2008 ) . | dynamics and collapse of collisionless self - gravitating systems is described by the coupled collisionless boltzmann and poisson equations derived from @xmath0 gravity in the weak field approximation .
specifically , we describe a system at equilibrium by a time - independent distribution function @xmath1 and two potentials @xmath2 and @xmath3 solutions of the modified poisson and collisionless boltzmann equations . considering a small perturbation from the equilibrium and linearizing the field equations
, it can be obtained a dispersion relation .
a dispersion equation is achieved for neutral dust - particle systems where a generalized jeans wave - number is obtained .
this analysis gives rise to unstable modes not present in the standard jeans analysis ( derived assuming newtonian gravity as weak filed limit of @xmath4 ) . in this perspective
, we discuss several self - gravitating astrophysical systems whose dynamics could be fully addressed in the framework of @xmath0-gravity . |
the fundamental symmetry underlying quantum chromodynamics ( qcd ) , the theory of strong interactions , is the exact local @xmath3 color symmetry . as a consequence of the non - abelian nature of this symmetry the gauge fields of qcd , the gluons , are colored objects and therefore interact strongly with each other .
because of confinement , one expects that gluons can also form colorless , or ` white ' , states which are called glueballs .
the first calculations of glueball masses were based on the bag - model approach @xcite . later on , the rapid improvement of lattice qcd allowed for precise simulations of yang - mills theory , leading to a determination of the full glueball spectrum @xcite .
however , in full qcd ( i.e. , gluons plus quarks ) the mixing of glueball and quark - antiquark configurations with the same quantum number occurs , rendering the identification of the resonances listed in the particle data group ( pdg ) @xcite more difficult .
the search for states which are ( predominantly ) glueball represents an active experimental and theoretical area of research , see ref .
@xcite and refs . therein .
the reason for these efforts is that a better understanding of the glueball properties would represent an important step in the comprehension of the non - perturbative behavior of qcd .
however , although up to now some glueball candidates exist ( see below ) , no state which is ( predominantly ) glueball has been unambiguously identified .
in general , a glueball state should fulfill two properties regarding its decays : it exhibits ` flavor blindness ' , because the gluons couple with the same strength to all quark flavors , and it is narrow , because qcd in the large-@xmath4 limit shows that all glueball decay widths scale as @xmath5 which should be compared to the @xmath6 scaling law for a quark - antiquark state . the lightest glueball state predicted by lattice qcd simulations is a scalar - isoscalar state ( @xmath7 ) with a mass of about @xmath8 gev @xcite .
the resonance @xmath9 shows a flavor - blind decay pattern and is narrow , thus representing a good candidate for a state which is ( predominantly ) a scalar glueball . also the resonance @xmath10 is a glueball candidate because its mass is very close to lattice qcd predictions and it is copiously produced in the gluon - rich decay of the @xmath11 meson .
both scenarios have been investigated in a variety of works , e.g. refs .
@xcite and refs .
therein , in which mixing patterns involving the scalar resonances @xmath12 @xmath9 , and @xmath10 are considered . in particular , in ref .
@xcite the decays of the @xmath11 have been included in a phenomenological fit and both assignments turn out to be consistent , but slightly favour a predominant gluonic amount in @xmath9 .
the second lightest lattice - predicted glueball state has tensor quantum numbers ( @xmath13 ) and a mass of about @xmath14 gev ; a good candidate could be the very narrow resonance @xmath15 @xcite , if the total spin of the latter will be experimentally confirmed to be @xmath16 .
the third least massive glueball predicted by lattice qcd ( in the quenched approximation ) has pseudoscalar quantum numbers ( @xmath17 ) and a mass of about @xmath0 gev .
quite remarkably , most theoretical works investigating the pseudoscalar glueball did not take into account this prediction of yang - mills lattice studies , but concentrated their search around @xmath18 gev in connection with the isoscalar - pseudoscalar resonances @xmath19 , and @xmath20 . a candidate for a predominantly light pseudoscalar glueball
is the middle - lying state @xmath21 due to the fact that it is largely produced in ( gluon - rich ) @xmath11 radiative decays and is missing in @xmath22 reactions @xcite . in this framework
the resonances @xmath23 and @xmath20 represent radial excitations of the resonances @xmath24 and @xmath25 . indeed , in relation to @xmath24 and @xmath25 , a lot of work has been done in determining the gluonic amount of their wave functions .
the kloe collaboration found that the pseudoscalar glueball fraction in the mixing of the pseudoscalar - isoscalar states @xmath24 and @xmath25 can be large ( @xmath26% ) @xcite , but the theoretical work of ref .
@xcite found that the glueball amount in @xmath24 and @xmath25 is compatible with zero [ see , however , also ref .
@xcite]*. * in this work we study the decay properties of a pseudoscalar glueball state whose mass lies , in agreement with lattice qcd , between @xmath27 and @xmath28 gev .
following ref .
@xcite we write down an effective chiral lagrangian which couples the pseudoscalar glueball field ( denoted as @xmath29 ) to scalar and pseudoscalar mesons .
we can thus evaluate the widths for the decays @xmath30 and @xmath31 where @xmath32 and @xmath33 stand for pseudoscalar and scalar quark - antiquark states .
the pseudoscalar state @xmath32 refers to the well - known light pseudoscalars @xmath34 , while the scalar state @xmath33 refers to the quark - antiquark nonet of scalars above 1 gev : @xmath35 .
the reason for the latter assignment is a growing consensus that the chiral partners of the pseudoscalar states should not be identified with the resonances below 1 gev , see refs .
@xcite for results within the so - called extended linear sigma model and also other theoretical works in ref .
@xcite ( and refs .
therein ) .
the chiral lagrangian that we construct contains one unknown coupling constant which can not be determined without experimental data .
however , the branching ratios can be unambiguously calculated and may represent a useful guideline for experimental search of the pseudoscalar glueball in the energy region between @xmath27 to @xmath28 gev . in this respect , the planned panda experiment at the fair facility @xcite will be capable to scan the mass region above 2.5 gev . the experiment is based on proton - antiproton scattering , thus the pseudoscalar glueball @xmath29 can be directly produced as an intermediate state .
we shall therefore present our results for the branching ratios for a putative pseudoscalar glueball with a mass of 2.6 gev . on the other hand ,
it is also possible that the pseudoscalar glueball @xmath29 has a mass that is a bit lower than the lattice qcd prediction and that it has been already observed in the besiii experiment where pseudoscalar resonances have been investigated in @xmath11 decays @xcite . in particular ,
the resonance @xmath2 which has been clearly observed in the @xmath36 channel represents a good candidate , because it is quite narrow ( @xmath37 mev ) and its mass lies just below the lattice qcd prediction .
for this reason we repeat our calculation for a pseudoscalar glueball mass of @xmath1 gev , and thus make predictions for the resonance @xmath2 , which can be tested in the near future .
this paper is organized as follows . in sec .
ii we present the effective lagrangian coupling the pseudoscalar glueball to scalar and pseudoscalar quark - antiquark degrees of freedom , and we calculate the branching ratios for the decays into @xmath38 and @xmath39 . finally , in sec .
iii we present our conclusions and an outlook .
following ref .
@xcite we introduce a chiral lagrangian which couples the pseudoscalar glueball @xmath40 with quantum numbers @xmath17 to scalar and pseudoscalar mesons @xmath41 where @xmath42 is a coupling constant , @xmath43 represents the multiplet of scalar and pseudoscalar quark - antiquark states , and @xmath44 are the generators of the group @xmath45 . in this work we consider the case @xmath46 and
the explicit representation of the scalar and pseudoscalar mesons reads @xcite:@xmath47{ccc}% \frac{(\sigma_{n}+a_{0}^{0})+i(\eta_{n}+\pi^{0})}{\sqrt{2 } } & a_{0}^{+}% + i\pi^{+ } & k_{s}^{+}+ik^{+}\\ a_{0}^{-}+i\pi^{- } & \frac{(\sigma_{n}-a_{0}^{0})+i(\eta_{n}-\pi^{0})}% { \sqrt{2 } } & k_{s}^{0}+ik^{0}\\ k_{s}^{-}+ik^{- } & \bar{k}_{s}^{0}+i\bar{k}^{0 } & \sigma_{s}+i\eta_{s}% \end{array } \right ) \;. \label{phimatex}%\ ] ] under @xmath48 chiral transformations the multiplet @xmath49 transforms as @xmath50 where @xmath51 and @xmath52 are @xmath53 matrices .
the determinant of @xmath49 is invariant under @xmath54 , but not under @xmath55 . on the other hand ,
the pseudoscalar glueball field @xmath29 is invariant under @xmath56 transformations . under parity , @xmath57 and @xmath58 , thus the effective lagrangian of eq .
( [ intlag ] ) is invariant under @xmath54 and under parity .
notice that eq .
( [ intlag ] ) is not invariant under @xmath59 , in agreement with the so - called axial anomaly in the isoscalar - pseudoscalar sector .
the rest of the mesonic lagrangian which describes the interactions of @xmath49 and also includes ( axial-)vector degrees of freedom is presented in sec .
[ app1 ] of the appendix . for more details , see refs .
@xcite .
the assignment of the quark - antiquark fields in this paper is as follows : ( i ) in the pseudoscalar sector the fields @xmath60 and @xmath61 represent the pions or the kaons , respectively @xcite .
the bare fields @xmath62 and @xmath63 are the non - strange and strange contributions of the physical states @xmath24 and @xmath25 @xcite:@xmath64 where @xmath65 is the mixing angle @xcite .
using other values for the mixing angle , e.g. @xmath66 @xcite or @xmath67 , as determined by the kloe collaboration @xcite , affects the presented results only marginally .
( ii ) in the scalar sector we assign the field @xmath68 to the physical isotriplet state @xmath69 and the scalar kaon fields @xmath70 to the resonance @xmath71 as a first approximation , the non - strange bare field @xmath72 is assigned to the physical isoscalar resonance @xmath73 and the bare field @xmath74 is assigned either to @xmath10 or to @xmath75 in a more complete framework , @xmath76 @xmath77 and a bare scalar glueball field @xmath78 mix and generate the physical resonances @xmath12 @xmath9 , and @xmath79 see the discussion below . in order to evaluate the decays of the pseudoscalar glueball @xmath29 we have to take into account that the spontaneous breaking of chiral symmetry takes place , which implies the need of shifting the scalar - isoscalar fields by their vacuum expectation values @xmath80 and @xmath81 , @xmath82 in addition , when ( axial-)vector mesons are present in the lagrangian , one also has to ` shift ' the axial - vector fields and to define the wave - function renormalization constants of the pseudoscalar fields:@xmath83 where @xmath84 runs over the four kaonic fields and @xmath85 the numerical values of the renormalization constants are @xmath86 , @xmath87 @xmath88 @xmath89 @xcite .
moreover , the condensates @xmath80 and @xmath81 read@xmath90 where the standard values @xmath91 gev and @xmath92 gev have been used @xcite . once the operations in eqs .
( [ shift ] ) and ( [ psz ] ) have been performed , the lagrangian in eq .
( [ intlag ] ) contains the relevant tree - level vertices for the decay processes of @xmath29 , see appendix ( sec .
[ app2 ] ) .
the branching ratios of @xmath29 for the decays into three pseudoscalar mesons are reported in table i for both choices of the pseudoscalar masses , @xmath0 and @xmath1 gev ( relevant for panda and besiii experiments , respectively ) .
the branching ratios are presented relative to the total decay width of the pseudoscalar glueball @xmath93 .
( for details of the calculation of the three - body decay we refer to sec .
[ app3 ] of the appendix . )
[ c]|c|c|c|quantity & case ( i ) : @xmath94 gev & case ( ii ) : @xmath95 gev + @xmath96 & @xmath97 & @xmath98 + @xmath99 & @xmath100 & @xmath101 + @xmath102 & @xmath103 & @xmath104 + @xmath105 & @xmath106 & @xmath107 + @xmath108 & @xmath109 & @xmath110 + @xmath111 & @xmath112 & @xmath112 + @xmath113 & @xmath114 & @xmath115 + @xmath116 & @xmath117 & @xmath118 + next we turn to the decay process @xmath119 the results , for both choices of @xmath120 are reported in table ii for the cases in which the bare resonance @xmath77 is assigned to @xmath10 or to @xmath75 [ c]|c|c|c|quantity & case ( i ) : @xmath94 gev & case ( ii ) : @xmath95 gev + @xmath121 & @xmath122 & @xmath123 + @xmath124 & @xmath125 & @xmath126 + @xmath127 & @xmath128 & @xmath129 + @xmath130 & @xmath131 & @xmath132 + @xmath133 & @xmath134 @xmath135 & @xmath136 @xmath137 + @xmath138 & @xmath110 @xmath139 & @xmath110 @xmath140 + concerning the decays involving scalar - isoscalar mesons , one should go beyond the results of table ii by including the full mixing pattern above 1 gev , in which the resonances @xmath12 @xmath141 and @xmath10 are mixed states of the bare quark - antiquark contributions @xmath142 and @xmath77 and a bare scalar glueball field @xmath143 this mixing is described by an orthogonal @xmath144 matrix @xcite . in view of the fact that a complete evaluation of this mixing in the framework of our chiral approach has not yet been done
, we use the two solutions for the mixing matrix of ref .
@xcite and the solution of ref .
@xcite in order to evaluate the decays of the pseudoscalar glueball into the three scalar - isoscalar resonances @xmath12 @xmath141 and @xmath10 . in all three solutions @xmath73 is predominantly described by the bare configuration @xmath72 , but the assignments for the other resonances vary : in the first solution of ref . @xcite the resonance @xmath9 is predominantly gluonic , while in the second solution of ref .
@xcite and the solution of ref .
@xcite the resonance @xmath10 has the largest gluonic content . the results for the decay of the pseudoscalar glueball into scalar - isoscalar resonances are reported in table iii .
[ c]|c|c|c|c|quantity & sol . 1 of ref .
@xcite & sol . 2 of ref .
@xcite & sol . of ref .
@xcite + @xmath145 & & & + @xmath146 & & & + @xmath147 & & & + @xmath148 & & & + @xmath149 & & & + @xmath150 & & & + in fig .
[ fig1 ] we show the behavior of the total decay width @xmath151 as function of the coupling constant @xmath42 for both choices of the pseudoscalar glueball mass .
( we assume here that other decay channels , such as decays into vector mesons or baryons are negligible ) . in the case of @xmath94 gev ,
one expects from large-@xmath4 considerations that the total decay width @xmath152 mev .
in fact , as discussed in the introduction , the scalar glueball candidate @xmath9 is roughly @xmath153 mev broad and the tensor candidate @xmath154 is even narrower . in the present work , the condition @xmath155 mev implies that @xmath156 @xmath157 .
moreover , in the case of @xmath95 gev in which the identification @xmath158 has been made , we can indeed use the experimental knowledge on the full decay width [ @xmath159 mev @xcite ] to determine the coupling constant to be @xmath160 .
( however , we also refer to the recent work of ref .
@xcite , where the possibility of a broad pseudoscalar glueball is discussed . ) [ ptb ] app.eps some comments are in order : \(i ) the results depend only slightly on the glueball mass , thus the two columns of table i and ii are similar .
it turns out that the channel @xmath161 is the dominant one ( almost 50% ) .
also the @xmath162 and @xmath163 channels are sizable . on the contrary ,
the two - body decays are subdominant and reach only 20% of the full mesonic decay width .
\(ii ) the decay of the pseudoscalar glueball into three pions vanishes : @xmath164 this result represents a further testable prediction of our approach .
\(iii ) the decays of the pseudoscalar glueball into a scalar - isoscalar meson amount only to @xmath165 of the total decay width .
moreover , the mixing pattern in the scalar - isoscalar sector has a negligible influence on the total decay width of @xmath166 nevertheless , in the future it may represent an interesting and additional test for scalar - isoscalar states .
\(iv ) once the shifts of the scalar fields have been performed , there are also bilinear mixing terms of the form @xmath167 and @xmath168 which lead to a non - diagonal mass matrix . in principle , one should take these terms into account , in addition to the already mentioned @xmath169 mixing , and solve a three - state mixing problem in order to determine the masses of the pseudoscalar particles .
this will also affect the calculation of the decay widths .
however , due to the large mass difference of the bare glueball fields @xmath29 to the other quark - antiquark pseudoscalar fields , the mixing of @xmath29 turns out to be very small in the present work , and can be safely neglected .
for instance , it turns out that the mass of the mixed state which is predominantly glueball is ( at most ) just @xmath170 gev larger than the bare mass @xmath94 gev .
\(v ) if a standard linear sigma model without ( axial-)vector mesons is studied , the replacements @xmath171 need to be performed .
most of the results of the branching ratios for the three - body decay are qualitatively , but not quantitatively , similar to the values of table i ( variations of about @xmath172-@xmath173 ) .
however , the branching ratios for the two - body decay change sizably w.r.t . the results of table 2 .
this fact shows once more that the inclusion of ( axial-)vector degrees of freedom has sizable effects also concerning the decays of the pseudoscalar glueball .
\(vi ) in principle , the three - body final states for the decays shown in table i can also be reached through a sequential decay from the two - body final states shown in table ii , where the scalar particle @xmath33 further decays into @xmath174 , for instance , @xmath175 .
there are then two possible decay amplitudes , one from the direct three - body decay and one from the sequential decay , which have to be added coherently before taking the modulus square to obtain the total three - body decay width . summing the results shown in table i and
ii gives a first estimate ( which neglects interference terms ) for the magnitude of the total three - body decay width .
we have verified that the correction from the interference term to this total three - body decay width in a given channel is at most of the order of @xmath176% for @xmath94 gev and @xmath177 for @xmath95 gev . for a full understanding of the contribution of the various decay amplitudes to the final three - body state
, one needs to perform a detailed study of the dalitz plot for the three - body decay .
in this work we have presented a chirally invariant effective lagrangian describing the interaction of the pseudoscalar glueball with scalar and pseudoscalar mesons for the three - flavor case @xmath46 .
we have studied the decays of the pseudoscalar glueball into three pseudoscalar and into a scalar and pseudoscalar quark - antiquark fields .
the branching ratios are parameter - free once the mass of the glueball has been fixed .
we have considered two possibilities : ( i ) in agreement with lattice qcd in the quenched approximation we have chosen @xmath94 gev .
the existence and the decay properties of such a hypothetical pseudoscalar resonance can be tested in the upcoming panda experiment @xcite .
( ii ) we assumed that the resonance @xmath178 measured in the experiment besiii , is ( predominantly ) a pseudoscalar glueball state , and thus we have also used a mass of @xmath1 gev @xcite .
the results for both possibilities have been summarized in tables i and ii : we predict that @xmath161 is the dominant decay channel , followed by ( almost equally large ) @xmath162 and @xmath163 decay channels . on the contrary
, the decay into three pions is predicted to vanish . in the case of besiii , by measuring the branching ratio for other decay channels than the measured @xmath179 , one could ascertain if @xmath2 is ( predominantly ) a pseudoscalar glueball . in the case of panda , our results may represent a useful guideline for the search of the pseudoscalar glueball .
future studies should consider possible mixing of the pseudoscalar glueball with charmonia states * * and an improved description of the scalar - isoscalar sector .
new lattice results for the pseudoscalar glueball mass , which go beyond the quenched approximation and include the effect of dynamical fermions , would be very useful for model building .
moreover , the mechanism of the glueball production via proton - antiproton fusion using the so - called mirror assignment @xcite represents an interesting outlook @xcite .
the authors thank klaus neuschwander for computing the interference terms for the sequential decay into three - body final states .
they also thank diego bettoni , stephen olsen , denis parganlija , anja habersetzer , antje peters , klaus neuschwander , and marc wagner for useful discussions . w.e .
acknowledges support from daad and hgs - hire , s.j . acknowledges support from h - qm and hgs - hire .
thanks the foundation polytechnical society frankfurt am main for support through an educator fellowship .
the chirally invariant @xmath180 lagrangian for the low - lying mesonic states with ( pseudo)scalar and ( axial-)vector quantum numbers has the form @xmath181-m_{0}^{2}\mathrm{tr}(\phi^{\dagger}\phi)-\lambda_{1}[\mathrm{tr}% ( \phi^{\dagger}\phi)]^{2}-\lambda_{2}\mathrm{tr}(\phi^{\dagger}\phi ) ^{2}\nonumber\\ & -\frac{1}{4}\mathrm{tr}[(l^{\mu\nu})^{2}+(r^{\mu\nu})^{2}]+\mathrm{tr}% [ ( \frac{m_{1}^{2}}{2}+\delta)(l_{\mu}^{2}+r_{\mu}^{2})]+\mathrm{tr}% [ h(\phi+\phi^{\dagger})]\nonumber\\ & + c_{1}(\mathrm{det}\phi-\mathrm{det}\phi^{\dagger})^{2}+i\frac{g_{2}}% { 2}\{\mathrm{tr}(l_{\mu\nu}[l^{\mu},l^{\nu}])+\mathrm{tr}(r_{\mu\nu}[r^{\mu } , r^{\nu}])\}\nonumber\\ & + \frac{h_{1}}{2}\mathrm{tr}(\phi^{\dagger}\phi)\mathrm{tr}\left ( l_{\mu } ^{2}+r_{\mu}^{2}\right ) + h_{2}\mathrm{tr}[\left\vert l_{\mu}\phi\right\vert ^{2}+\left\vert \phi r_{\mu}\right\vert ^{2}]\nonumber\\ & + 2h_{3}\mathrm{tr}(l_{\mu}\phi r^{\mu}\phi^{\dagger } ) .
\label{fulllag}%\end{aligned}\ ] ] where @xmath182{ccc}% \frac{\omega_{n}^{\mu}+\rho^{\mu0}}{\sqrt{2}}+\frac{f_{1n}^{\mu}+a_{1}^{\mu0}% } { \sqrt{2 } } & \rho^{\mu+}+a_{1}^{\mu+ } & k^{\star\mu+}+k_{1}^{\mu+}\\ \rho^{\mu-}+a_{1}^{\mu- } & \frac{\omega_{n}^{\mu}-\rho^{\mu0}}{\sqrt{2}}% + \frac{f_{1n}^{\mu}-a_{1}^{\mu0}}{\sqrt{2 } } & k^{\star\mu0}+k_{1}^{\mu0}\\ k^{\star\mu-}+k_{1}^{\mu- } & \bar{k}^{\star\mu0}+\bar{k}_{1}^{\mu0 } & \omega_{s}^{\mu}+f_{1s}^{\mu}% \end{array } \right ) \;,\ ] ] and @xmath183{ccc}% \frac{\omega_{n}^{\mu}+\rho^{\mu0}}{\sqrt{2}}-\frac{f_{1n}^{\mu}+a_{1}^{\mu0}% } { \sqrt{2 } } & \rho^{\mu+}-a_{1}^{\mu+ } & k^{\star\mu+}-k_{1}^{\mu+}\\ \rho^{\mu-}-a_{1}^{\mu- } & \frac{\omega_{n}^{\mu}-\rho^{\mu0}}{\sqrt{2}}% -\frac{f_{1n}^{\mu}-a_{1}^{\mu0}}{\sqrt{2 } } & k^{\star\mu0}-k_{1}^{\mu0}\\ k^{\star\mu-}-k_{1}^{\mu- } & \bar{k}^{\star\mu0}-\bar{k}_{1}^{\mu0 } & \omega_{s}^{\mu}-f_{1s}^{\mu}% \end{array } \right ) \;,\ ] ] for details see refs . @xcite . in the present context
we are interested in the wave - function renormalization constants @xmath184 introduced in eq .
( [ psz ] ) .
their explicit expressions read @xcite:@xmath185@xmath186 for completeness we report the explicit expression for the three - body decay width for the process @xmath188 @xcite:@xmath189 where @xmath190 and@xmath191 the quantities @xmath192 @xmath193 @xmath194 refer to the masses of the three pseudoscalar states @xmath195 @xmath196 , and @xmath197 @xmath198 is the corresponding tree - level decay amplitude , and @xmath199 is a symmetrization factor ( it equals @xmath200 if all @xmath195 @xmath196 , and @xmath201 are different , it equals @xmath27 for two identical particles in the final state , and it equals @xmath202 for three identical particles in the final state ) .
for instance , in the case @xmath203 one has : @xmath204 , @xmath205 gev , @xmath206 gev , and @xmath94 gev .
then : @xmath207}\;. \label{4}%\ ] ] the full decay width into the channel @xmath161 results from the sum@xmath208 the other decay channels can be calculated in a similar way .
r. l. jaffe and k. johnson , phys .
b * 60 * , 201 ( 1976 ) ; r. konoplich and m. shchepkin , nuovo cim .
a * 67 * , 211 ( 1982 ) ; m. jezabek and j. szwed , acta phys .
b * 14 * , 599 ( 1983 ) ; r. l. jaffe , k. johnson and z. ryzak , annals phys . * 168 * , 344 ( 1986 ) ; m. strohmeier - presicek , t. gutsche , r. vinh mau and a. faessler , phys .
d * 60 * , 054010 ( 1999 ) [ arxiv : hep - ph/9904461 ] . c. morningstar and m. j. peardon , aip conf .
proc . * 688 * , 220 ( 2004 ) [ arxiv : nucl - th/0309068 ] ; c. j. morningstar and m. j. peardon , phys .
d * 60 * ( 1999 ) 034509 [ hep - lat/9901004 ] ; m. loan , x. q. luo and z. h. luo , int .
j. mod .
a * 21 * , 2905 ( 2006 ) [ arxiv : hep - lat/0503038 ] ; e. b. gregory , a. c. irving , c. c. mcneile , s. miller and z. sroczynski , pos * lat2005 * , 027 ( 2006 ) [ arxiv : hep - lat/0510066 ] ; y. chen _ et al .
_ , phys .
d * 73 * , 014516 ( 2006 ) [ arxiv : hep - lat/0510074 ] ; e. gregory , a. irving , b. lucini , c. mcneile , a. rago , c. richards and e. rinaldi , arxiv:1208.1858 [ hep - lat ] .
f. e. close , rept
. prog .
* 51 * , 833 ( 1988 ) ; s. godfrey and j. napolitano , rev .
* 71 * , 1411 ( 1999 ) [ arxiv : hep - ph/9811410 ] ; c. amsler and n. a. tornqvist , phys . rept . *
389 * , 61 ( 2004 ) ; e. klempt and a. zaitsev , phys .
rept . * 454 * , 1 ( 2007 ) [ arxiv:0708.4016 [ hep - ph ] ] . f. giacosa , phys .
d * 80 * ( 2009 ) 074028 .
[ arxiv:0903.4481 [ hep - ph ] ] . c. amsler and f. e. close , phys .
d * 53 * ( 1996 ) 295 [ arxiv : hep - ph/9507326 ] .
w. j. lee and d. weingarten , phys .
d * 61 * , 014015 ( 2000 ) .
[ arxiv : hep - lat/9910008 ] ; f. e. close and a. kirk , eur .
j. c * 21 * , 531 ( 2001 ) . [ arxiv : hep - ph/0103173 ] .
f. giacosa , t. gutsche , v. e. lyubovitskij and a. faessler , phys .
b * 622 * , 277 ( 2005 ) [ arxiv : hep - ph/0504033 ] .
f. giacosa , t. gutsche and a. faessler , phys .
c * 71 * , 025202 ( 2005 ) [ arxiv : hep - ph/0408085 ] .
v. mathieu , n. kochelev and v. vento , int .
j. mod .
e * 18 * ( 2009 ) 1 [ arxiv:0810.4453 [ hep - ph ] ] .
a. masoni , c. cicalo and g. l. usai , j. phys .
g : nucl . part .
phys . * 32 * , r293-r335 ( 2006 ) ; t. gutsche , v. e. lyubovitskij and m. c. tichy , phys .
d * 80 * , 014014 ( 2009 ) ; h. cheng , h. li and k. liu , phys .
d * 79 * , 014024 ( 2009 ) ; v. mathieu and v. vento , phys .
d * 81 * , 034004 ( 2010 ) ; c. di donato , g. ricciardi and i. bigi , phys
. rev .
d * 85 * ( 2012 ) 013016 [ arxiv:1105.3557 [ hep - ph ] ] ; b. a. li , phys . rev .
d * 81 * ( 2010 ) 114002 [ arxiv:0912.2323 [ hep - ph ] ] .
c. rosenzweig , a. salomone and j. schechter , phys .
d * 24 * ( 1981 ) 2545 .
c. rosenzweig , j. schechter and c. g. trahern , phys .
d * 21 * ( 1980 ) 3388 .
c. rosenzweig , a. salomone and j. schechter , nucl .
b * 206 * ( 1982 ) 12 [ erratum - ibid .
b * 207 * ( 1982 ) 546 ] ; k. kawarabayashi and n. ohta , nucl .
b * 175 * ( 1980 ) 477 .
d. parganlija , f. giacosa , p. kovacs , g. wolf , aip conf .
* 1343 * ( 2011 ) 328 - 330 .
[ arxiv:1011.6104 [ hep - ph ] ] .
d. parganlija , f. giacosa , d. h. rischke , p. kovacs , g. wolf , int .
j. mod .
a * 26 * ( 2011 ) 607 - 609 .
[ arxiv:1009.2250 [ hep - ph ] ] .
d. parganlija , acta phys .
supp . * 4 * ( 2011 ) 727 - 732 .
[ arxiv:1105.3647 [ hep - ph ] ] .
l. maiani , f. piccinini , a. d. polosa and v. riquer , phys .
* 93 * ( 2004 ) 212002 [ arxiv : hep - ph/0407017 ] .
f. giacosa , phys .
d * 74 * ( 2006 ) 014028 [ arxiv : hep - ph/0605191 ] .
a. h. fariborz , r. jora and j. schechter , phys .
d * 72 * ( 2005 ) 034001 [ arxiv : hep - ph/0506170 ] .
a. h. fariborz , int .
j. mod .
phys . a * 19 * ( 2004 ) 2095 .
[ arxiv : hep - ph/0302133 ] .
m. napsuciale and s. rodriguez , phys .
d * 70 * ( 2004 ) 094043 .
f. giacosa , g. pagliara , nucl .
a * 833 * ( 2010 ) 138 - 155 .
[ arxiv:0905.3706 [ hep - ph ] ] .
f. giacosa , phys .
d * 75 * ( 2007 ) 054007 [ arxiv : hep - ph/0611388 ] .
a. heinz , s. struber , f. giacosa and d. h. rischke , phys .
d * 79 * ( 2009 ) 037502 [ arxiv:0805.1134 [ hep - ph ] ] .
m. ablikim _
( bes collaboration ) , phys .
lett . * 95 * , 262001 ( 2005 ) ; n. kochelev and d. p. min , phys .
b * 633 * , 283 ( 2006 ) ; m. ablikim _
( bes iii collaboration ) , phys .
lett . 106.072002 ( 2011 ) . | we study a chiral lagrangian which describes the two- and three - body decays of a pseudoscalar glueball into scalar and pseudoscalar mesons .
the various branching ratios are a parameter - free prediction of our approach .
we compute the decay channels for a pseudoscalar glueball with a mass of @xmath0 gev , as predicted by lattice qcd in the quenched approximation , which is in the reach of the panda experiment at the upcoming fair facility . for completeness
, we also repeat the calculation for a glueball mass of @xmath1 gev which corresponds to the mass of the resonance @xmath2 measured in the besiii experiment . |
several years ago , we authored a paper in this journal entitled `` neutrino physics , '' hereafter called i@xcite , in order to encourage inclusion of material involving neutrinos into the introductory curriculum .
we noted at the time that neutrinos , with new experiments about to provide first data , might continue to be a popular press item .
this prediction has proven true : * ray davis , the founder of the field of experimental solar neutrino physics , shared the 2002 nobel prize in physics with mastoshi koshiba , who led the kamioka solar neutrino experiment , and riccardo giacconi @xcite . *
results from the sudbury neutrino observatory ( sno ) resolved the solar neutrino puzzle , showing that approximately two - thirds of these neutrinos oscillate into other flavors before reaching earth @xcite . *
the kamland experiment , in which antineutrinos from japanese power reactors were detected , confirmed the sno results and further narrow the allowed range of neutrino mass differences @xcite . * the wilkinson microwave anisotropy probe ( wmap )
measured subtle temperature differences within the oldest light in the universe , from the epoch when atoms first formed 380,000 years after the big bang @xcite . when combined with the results of large scale structure studies
@xcite , a new bound on the sums of the neutrino masses is obtained .
in addition there has been published ( and disputed ) evidence for the existence of neutrinoless double beta decay which , if confirmed , would show that one of the standard model s most important symmetries , the conservation of lepton number , is violated @xcite .
thus we decided to bring our early paper up to date by explaining here the importance and implications of the new results .
in section 2 we present a much abbreviated summary of the material presented in i. in section 3 we discuss the sno results and how they resolved the puzzling discrepancies ray davis first uncovered 30 years ago . in section 4
we discuss kamland , the first terrestrial experimental to achieve the sensitivity to the small neutrino mass differences relevant to solar neutrino experiments . in section 5
we describe marvelous new cosmological probes of large scale structure and of the time when atoms were first formed , and why the new data may soon challenge recent double beta decay claims .
we summarize where we stand in neutrino physics including the new discoveries that may be soon be within reach in section 6 .
in i we summarized the basics of neutrino physics , including neutrino history , properties , and implications for contemporary physics .
for the purposes of the present work , we note that in the so - called standard model of particle / nuclear physics@xcite , which is consistent with nearly all present experimental information , there exist three massless neutrino types@xmath0which are produced with purely left - handed helicity in weak interaction processes .
if the neutrino were shown to have a nonvanishing mass , it would be the first clear failure of this 30-year - old standard model and the first proof of the existence of the `` particle dark matter '' that appears necessary to explain the structure and expansion of our universe .
nonzero neutrino masses were suggested by the results of experiments measuring the flux of solar neutrinos , which are produced as a byproduct of the thermonuclear reactions occurring in the high - temperature core of our sun @xcite .
additional strong evidence comes from the study of atmospheric neutrinos , which are produced when high energy cosmic rays collide with the upper atmosphere , producing pions and other particles that then decay into neutrinos @xcite .
the largest of the various atmospheric neutrino experiments is superkamiokande , a detector in a mine in the japanese alps that contains 50,000 tons of ultra - pure water .
superkamiokande s precise data show that the flux of muon - type atmospheric neutrinos arriving from the opposite side of the earth , which have travelled a long distance to reach the detector , is depleted .
both the solar and atmospheric neutrino results can be explained quantitatively if neutrinos are massive and if the mass eigenstates are not coincident with the weak interaction eigenstates @xmath0 , i.e. , if the neutrinos produced in weak interactions are combinations of the various mass eigenstates .
this is exactly what is known to occur in the analogous case of the quarks@xcite .
the relation between the neutrino mass and weak - interaction eigenstates is described by a unitary matrix : for three neutrinos , the mass and weak - interaction eigenstates can be viewed as two distinct 3d coordinate systems .
the unitary matrix specifies the three rotations describing the orientation of one coordinate system relative to the other .
the reduced flux of atmospheric muon neutrinos ( see figure 1 ) is quantitatively explained by @xmath1 oscillations governed by maximal mixing , @xmath2 : the relationship between the mass eigenstate @xmath3 and @xmath4 and the flavor eigenstates @xmath5 and @xmath6 involves a rotation of 45 degrees .
the magnitude of the mass difference @xmath7 between the two equal components making up the flavor eigenstates is @xmath8 ev@xmath9 .
when i was written , the favored solution to the solar neutrino problem was also neutrino mixing , but was described by a much smaller mixing angle @xmath10 specifying the relationship between mass eigenstates @xmath11 and @xmath3 and the flavor eigenstates @xmath12 and @xmath5 .
the effects of this small mixing are magnified by matter effects in the sun .
this is the so - called msw effect and is described in i. but the results of the sudbury neutrino observatory provided a bit of a surprise .
because both charged and neutral currents contribute to the reaction important to superkamiokande , @xmath13 the experimentalists can not easily distinguish @xmath12s from the @xmath5s and @xmath6s : the detector records both fluxes , though with a reduced sensitivity ( 0.15 ) for the heavy - flavor types .
the reaction produces energetic recoil electrons which generate cerenkov radiation that is recorded in an array of phototubes surrounding the detector .
as the cross section is sharply forward peaked , the correlation with the position of the sun can be used to `` cut '' background contributions associated with cosmic rays and radioactivity in the rock walls surrounding the detector . because the threshold for electron detection is @xmath14 mev , only the high energy portion of the @xmath15b solar neutrino flux is measured .
these are the same neutrinos that dominate the radiochemical measurements of ray davis : superkamiokande confirmed that this flux was substantially below that predicted by the standard solar model ( ssm)@xcite @xmath16 when the davis and superkamiokande results were combined with those from the gallium experiments sage and gallex , the resulting three constraints on the three principal solar neutrino sources ( @xmath17b , @xmath18be , and the low - energy pp fluxes ) produced a surprising result . no combination of these fluxes could reproduce the combined data well . though circumstantial
, this evidence indicated that the solution to the solar neutrino problem would not be found in the ssm , but instead would require new particle physics .
the favored explanation became neutrino oscillations which , as we have just summarized , occur for massive neutrinos when the weak and mass eigenstates do not coincide .
the superkamiokande discrepancy , a solar neutrino rate less than half that expected from the ssm , would require that approximately two - thirds of the high energy electron neutrinos generated in the solar interior oscillate into other species @xmath19 before reaching earth .
low - energy @xmath5s and @xmath6s are invisible to the davis and sage / gallex detectors and scatter electrons in the superkamiokande detector with a reduced cross section , as we noted previously .
the key idea behind the sudbury neutrino observatory ( sno ) was construction of a detector that would have multiple detection channels , recording the @xmath12s by one reaction and the total flux of all neutrinos ( @xmath12s + @xmath5s + @xmath6s ) by another .
this was accomplished by replacing the ordinary water in a water cerenkov detector with heavy water
d@xmath20o instead of h@xmath20o . the charged - current ( cc ) channel that records the @xmath12s is analogous to the reaction used in the davis detector @xmath21 as the electron produced in this reaction carries off most of the neutrino energy , its detection in the sno detector ( by the cerenkov light it generates ) allows the experimentalists to determine the spectrum of solar @xmath12s , not just the flux . a second reaction , the neutral current ( nc ) breakup of deuterium , gives the total flux , independent of flavor ( the @xmath12 , @xmath5 , and @xmath6 cross sections are identical ) , @xmath22 the only signal for this reaction in a water cerenkov detector is the neutron , which can be observed as it captures via the @xmath23 reaction .
sno is currently operating with salt added to the water , as cl in the salt is an excellent @xmath23 target , producing about 8 mev in @xmath24s .
while the strategy may sound straightforward , producing such a detector was an enormous undertaking .
the needed heavy water worth about $ 300 m was available through the canadian government because of its candu reactor program .
the single - neutron detection required for the neutral current reaction is possible only if backgrounds are extremely low .
for this reason the detector had to be placed very deep underground , beneath approximately two kilometers of rock , so that cosmic - ray muon backgrounds would be reduced to less than 1% of that found in the superkamiokande detector .
the experimentalists found the needed site in an active nickel mine , the sudbury mine in ontario , canada , where they worked with the miners to carve out a 10-story - high cavity on the mine s 6800-ft level .
trace quantities of radioactivity were another background concern : if a thimblefull of dust were introduced into the massive cavity during construction , the resulting neutrons from u and th could cause the experiment to fail . thus , despite the mining activities that continued around them , the experimentalists constructed their detector to the strictest cleanroom standards .
the detector also provided a third detection channel , neutrino elastic scattering ( es ) off electrons ( eq .
( 1 ) ) , which we have noted is sensitive to @xmath12s and , with reduced sensitivity , @xmath5s and @xmath6s .
the es reaction , of course , provides sno a direct cross check against superkamiokande .
threshold for measuring these electrons is about 5 mev . assuming no oscillations , sno s detection rate is equivalent to a @xmath12 flux of @xmath25 a result in excellent accord with that from superkamiokande , @xmath26 the greater accuracy of the superkamiokande result reflects the larger mass ( 50 kilotons ) and longer running time of the japanese experiment .
( sno contains , in addition to the one kiloton of heavy water in its central acrylic vessel , an additional seven kilotons of ordinary water that surrounds the central vessel , helping to shield it . )
the crucial new information provided by sno comes from the two reactions on deuterium .
the cc current channel is only sensitive to @xmath12s . under the assumption of an undistorted @xmath17b neutrino flux , sno experimentalists deduced @xmath27 the cc flux is less than that deduced from the es rate , indicating that @xmath5s and @xmath6s must be contributing to the later . from the difference between the superkamiokande es and the sno cc results @xmath28 and recalling that the @xmath29 es cross section is only 0.15 that for the @xmath12 , one deduces the heavy - flavor contribution to the solar neutrino flux @xmath30 that is , approximately two - thirds of the solar neutrino flux is in these flavors .
while the first sno analysis was done in the manner described above , a second publication gave the long awaited nc results .
this allowed a direct and very accurate determination of the flavor content of solar neutrinos , without the need for combining results from two experiments .
the published nc results were obtained without the addition of salt to the detector : the neutron was identified by the 6.25 mev @xmath24 ray it produces by capturing on deuterium .
the resulting total flux , independent of flavor , is @xmath31 combining with the cc signal yields @xmath32 the presence of heavy - flavor solar neutrinos and thus neutrino oscillations is confirmed at the 5.3@xmath33 level !
furthermore the total flux is in excellent agreement with the predictions of the ssm eq .
[ eq : ssm ] , an important vindication of stellar evolution theory .
the sno analysis is summarized in figure 2 , which shows the three bands corresponding to the cc , nc , and es measurements coinciding in a single region .
these results can now be combined with other solar neutrino measurements to determine the parameters the mixing angle and mass - squared difference governing the oscillations . at the time
i was written , there were several contending solutions , though the data favored one characterized by a small mixing angle ( thus called the sma solution ) .
figure 3 shows that the sno result has determined an oscillation solution that , at 99% confidence level , is unique and as in the atmospheric neutrino case , it has a large mixing angle , @xmath34 degrees .
this lma oscillation is clearly distinct from that seen with atmospheric neutrinos , with @xmath35 centered on a region @xmath36 ev@xmath9 .
the discovery that the atmospheric and solar neutrino problems are both due to neutrino oscillations has provided the first evidence for physics beyond the standard model . that neutrinos provided
this evidence is perhaps not unexpected : if the standard model is viewed as an effective theory , one largely valid in our low - energy world but missing physics relevant to very high energies , beyond the reach of current accelerators , then a neutrino mass term is the lowest - order correction that can be added to that theory . but a surprise is the large mixing angles characterizing the neutrino oscillations which contradicts the simple prejudice that neutrino mixing angles might be similar to the small angles familiar from quark mixing . perhaps this simply reinforces something that should have been apparent at the outset : with their small masses and distinctive mixings , neutrinos likely have an underlying mechanism for mass generation that differs from that of the other standard model fermions .
flux of @xmath37b solar neutrinos is divided into @xmath29 and @xmath12 flavors by the sno analysis .
the diagonal bands show the total @xmath37b flux as predicted by the ssm ( dashed lines ) and that measured with the nc reaction in sno ( solid band ) .
the widths of these bands represent the @xmath38 errors .
the bands intersect in a single region for @xmath39 and @xmath40 , indicating that the combined flux results are consistent with neutrino flavor transformation assuming no distortion in the @xmath37b neutrino energy spectrum.,width=470 ] for two - flavor mixing , the values of the mass difference and mixing angle consistent with the world s data on solar neutrinos , post sno . at 99% confidence level
the addition of sno data isolates a unique , large - mixing angle solution.,width=374 ]
one remarkable aspect of the solar and atmospheric neutrino discoveries is that the derived oscillation parameters are within the reach of terrestrial experiments .
this did not have to be the case
solar neutrinos are sensitive to neutrino mass differences as small as 10@xmath41 ev@xmath9 , for which terrestrial experiments would be unthinkable .
the first terrestrial experiment to probe solar neutrino oscillation parameters , kamland , very recently reported first results .
the acronym kamland stands for kamioka liquid scintillator anti - neutrino detector .
the inner detector consists of one kiloton of liquid scintillator contained in a spherical balloon , 13 m in diameter .
the balloon is suspended in the old kamioka cavity ( where superkamiokande s predecessor was housed ) by kevlar ropes , with the region between the balloon and an 18m - diameter stainless steel containment vessel filled with additional scintillator ( serving to shield the target from external radiation ) .
several japanese power reactors are about 180 km from the kamioka site , and the electron antineutrinos emitted by nuclear reactions in the cores of these reactors can be detected in kamland via the inverse beta decay reaction @xmath42 where the @xmath43 is seen in coincidence with the delayed 2.2 mev @xmath24 ray produced by the capture of the accompanying neutron on a proton .
this coincidence allows the experimentalists to distinguish @xmath44 reactions from background .
> from the reactor operations records , which the power companies have made available , kamland experimentalists can calculate the resulting flux at kamiokande to a precision of @xmath45% , in the absence of oscillations .
thus , if a significant fraction of the reactor @xmath44s oscillate into @xmath46s or @xmath47s before reaching the detector , a low rate of @xmath43/capture-@xmath24-ray coincidences will be evident : this is an example of the `` disappearance '' oscillation technique we described in i. for the 162 ton / yr exposure so far reported by the kamland collaboration , the number of events expected in the absence of oscillations is @xmath48 .
but the number measured is 54
just 61% of the no - oscillation expectation . from the two - neutrino - flavors oscillation survival probability @xmath49
one obtains the oscillation parameters of figure 4 .
kamland confirms the lma solution and significantly narrows sno s allowed region ( the red area in figure 4 ) .
kamland has excellent sensitivity to @xmath50 but less sensitivity to @xmath51 ( due to uncertainties in the shape of the reactor @xmath44 spectrum ) .
the result is the separation of the sno lma allowed region into two parts , with the best - fit @xmath52 ev@xmath9 , but with a larger mass difference @xmath53 ev@xmath9 also fitting well .
kamland is an excellent example of complementary terrestrial and astrophysical measurements : solar neutrino experiments provide our best constraints on @xmath10 , but kamland places the tightest bounds on @xmath50 .
despite the wonderful recent discoveries in neutrino physics , there remain quite a number of open questions . several have to do with the pattern of neutrino masses .
all of the experiments described above probe mass _ differences _ , not absolute neutrino masses .
furthermore the atmospheric neutrino experiments only constrain the magnitude of @xmath54 , and not its sign . as a result ,
there exist several mass patterns fully consistent with all known data .
one choice would assign @xmath55 to be the heaviest neutrino , split by the atmospheric mass difference @xmath56 ev@xmath9 from a lighter , nearly degenerate pair of neutrinos responsible for solar neutrino oscillations .
( this pair is split by solar neutrino mass difference @xmath57 ev@xmath9 . )
however , as the sign of @xmath54 is not known , it is also possible that @xmath55 is the lightest neutrino , with the nearly degenerate @xmath58 and @xmath59 heavier .
finally , the best direct laboratory constraint on absolute neutrino masses comes from studies of tritium beta decay , as described in i. studies of the tritium spectrum near its endpoint energy places a bound of 2.2 ev @xcite on the @xmath44 mass ( or more properly , on the principal mass eigenstate contributing to the @xmath44 ) .
consequently , one can add an overall scale of up to 2.2 ev to the mass splittings described above .
that is , no terrestrial measurement rules out three nearly degenerate neutrinos , each with a mass @xmath60 2.2 ev , but split by requisite @xmath61 and @xmath62 .
as discussed in i , the absolute neutrino mass is crucial in cosmology , as a sea of neutrinos produced in the big bang pervades all of space .
if these neutrinos carry a significant mass , they would constitute an important component of particle dark matter , invisibly affecting the structure and expansion of our universe .
light neutrinos , such as those we have been discussing , decouple from the rest of the matter as relativistic particles , about one second after the big bang .
their number density and temperature can be calculated at the time of decoupling and today .
their contribution to the universe s mass - energy budget is now dominated by their masses , @xmath63 where @xmath64 is the critical density that will just close the universe .
a variety of cosmological probes suggest that our universe is very close to the critical density@xmath65 .
> from the discussion above , we know that at least one neutrino must have a mass of at least @xmath66 .
similarly , the tritium beta decay limit places an upper bound on the sum of the masses of 6.6 ev ( corresponding to three nearly generate neutrinos of mass 2.2 ev ) .
it follows that the neutrino contribution to dark matter is bounded above and below @xmath67 this broad range implies that the amount of mass in neutrinos could easily exceed all the familiar baryon matter
stars , dust , gas clouds , and us
visible or invisible : big - bang nucleosynthesis and precision measurements of the cosmic microwave background ( cmb ) both indicate that @xmath68 .
however , in the past few years a series of extraordinarily precise measurements have been made in cosmology .
one of these is the recent wmap full - sky map of the cmb and its subtle ( few millionths of a degree ) temperature anisotropies .
this is the oldest light in the universe , the photons that decoupled from matter at the time atoms formed , about 380,000 years after the big bang .
the cmb temperature anisotropies tell us about the structure of the universe its clumpiness at this very early epoch .
measurements of distant snia supernovae a sort of `` standard candle '' by which astronomers can measure cosmological distances have constrained the expansion rate and mass / energy budget of the universe .
large - scale surveys , such as the 2df galaxy redshift survey , have mapped the distribution of visible matter in the universe today and in recent times .
the result from combining these and other cosmological probes is a rather sharp constraint on the amount of hot dark matter in particular , the mass density in neutrinos that can be allowed , given that our universe has evolved to its present state .
one finds @xmath69 that is , an upper bound on the sum of neutrino masses of about 1.0 ev .
( some analyses claim even tighter upper bounds . )
this bound is significantly tighter than that of eq .
( 14 ) , derived from laboratory data only .
cosmology now demands that neutrino dark matter can make up no more than 2 - 3% of the universe s mass - energy budget and , in particular , is less important than other forms of matter we know about ( e.g. , nucleons ) .
( we will not go into the disconcerting fact that at least 93% of the universe s mass - energy budget appears to be dark energy and cold dark matter that we have not yet adequately characterized ! )
it also tells us that neutrino mass at the @xmath60 1 ev level now effects cosmological analyses : such analyses would constrain other cosmological parameters more tightly if neutrino masses were measured , rather than being free parameters that one must dial in cosmological models until unacceptable deviations are found . (
the one ev bound was derived in this way . )
this underscores how important it is to significantly improve laboratory mass limits .
one possibility is a new - generation tritium experiment : there is a serious effort underway to improve the current bound on the @xmath44 mass to about 0.3 ev , which would then place an upper bound on the sum of the masses of about 0.9 ev @xcite .
another possibility less definitive , perhaps , but with even greater reach is offered by new - generation neutrinoless double beta decay experiments .
the phenomenon of neutrinoless double beta decay , described in i , tests not only mass , but also whether a standard model symmetry called lepton number conservation is violated . in neutrinoless double beta decay a nucleus spontaneously decays by changing its charge by two units while emitting two electrons , @xmath70 where the intermediate nuclear state @xmath71 is virtual .
the emitted electrons carry off the entire nuclear energy release , allowing this process to be distinguished from the standard - model - allowed process of two - neutrino double beta decay ( where the energy is shared between two electrons and two @xmath44s in the final state ) .
the neutrinoless process clearly violates lepton number , as two leptons ( the electrons ) are spontaneously produced .
( by contrast , as the @xmath72 carries @xmath73 and the @xmath44 @xmath74 , two - neutrino double beta decay conserves lepton number . ) what conditions will lead to neutrinoless double beta decay ?
the necessary lepton number violation is present if the neutrino is a majorana particle_i.e .
_ , is identical to its antiparticle .
most theoretical models include majorana neutrinos : this is part of the mechanism that allows us to understand why neutrinos have masses much smaller than those of the other standard - model fermions , such as electrons and quarks .
but the existence of a majorana neutrino alone is not sufficient because of the exact handedness of neutrinos , which we discussed in i. in the neutrinoless double beta decay reaction written above , the @xmath44 appearing in the intermediate nuclear state was produced in the nucleus by the neutron @xmath75 decay reaction @xmath76 . to complete the decay
, the antineutrino must be reabsorbed on a second neutron , @xmath77 . at first glance , if @xmath78 , _
i.e. _ , if the neutrino is its own antiparticle , this reabsorption looks possible .
however , this conclusion overlooks the neutrino handedness . in the first step ,
the @xmath44 produced is righthanded , while the second reaction only proceeds if the @xmath12 is lefthanded .
thus it would appear that the neutrinoless process is forbidden , even if @xmath78 .
this argument , however , overlooks the effects of neutrino mass : a small neutrino mass breaks the exact neutrino handedness , allowing neutrinoless @xmath79 decay to proceed , though the amplitude is suppressed by the factor @xmath80 , where @xmath81 mev is the typical energy of the exchanged neutrino .
it follows that neutrinoless double beta decay measures the neutrino mass at least the majorana portion of that mass .
( making this statement more precise , unfortunately , takes us beyond the limits of this paper . ) in the simplest case a single majorana mass eigenstate dominating the @xmath82 decay
the neutrinoless amplitude is proportional to @xmath83 , where @xmath84 is the mixing probability of the @xmath85th mass eigenstate in the @xmath12 and @xmath86 is the mass . currently the best neutrinoless @xmath82 decay limits are those obtained by the heidelberg - moscow and igex enriched ( 86% ) @xmath87ge experiments , which are both probing lifetimes beyond @xmath88 years corresponding to roughly one decay per kg - year ! these experiments employ ge crystals
the ge is both source and detector containing about 10 kg of active material .
next - generation experiments , using a variety of double beta decay sources ( @xmath87ge , @xmath89xe , @xmath90mo ) , have been proposed at the one ton scale .
these have as their goals sensitivities to neutrino masses of 10 - 50 milli - ev , corresponding to lifetimes well in excess of @xmath91 y. one important motivation for these heroic proposals is the number @xmath92 55 milli - ev : in several scenarios accommodating the solar and atmospheric oscillation results , this scale plays a role in determining the level at which neutrinoless @xmath82 decay might be observed .
a few members of the heidelberg - moscow collaboration have claimed that their present results are not a limit , but rather a detection of neutrinoless @xmath82 decay , with a best value for the majorana neutrino mass of @xmath60 0.4 ev @xcite .
this claim has also been strongly criticized by a group that argues that the claimed peak , shown in figure 5 , is not statistically significant .
regardless , this claim will clearly be tested soon , in other @xmath82 decay experiments and in future cosmological tests , which promise to soon be probing masses @xmath60 0.3 ev .
in the three years since the publication of i , several very significant neutrino discoveries have been made : * sno has shown that approximately two - thirds of the @xmath17b neutrinos that arrive on earth have oscillated into @xmath5s or @xmath6s , thus demonstrating that new neutrino physics is responsible for the solar neutrino puzzle first uncovered by ray davis , jr .
together with the atmospheric neutrino discoveries of superkamiokande , this discovery of an effect requiring massive neutrinos and neutrino mixing is the first evidence for physics beyond the standard model .
the sno results for the total solar neutrino flux , independent of flavor , are in excellent agreement with the predictions of the standard model despite the challenge of calculating a flux that varies as @xmath93 ( ! ) , where @xmath94 is the solar core temperature .
the sno results , when added to other solar neutrino data , isolate a single oscillation scenario , the lma solution . *
the first terrestrial experiment to probe solar neutrino oscillation parameters , kamland , has confirmed the sno results and further narrowed the lma range of allowed @xmath50 . *
both the absolute scale of neutrino masses and the detailed pattern of the masses remain unknown , as the results in hand measure only mass differences ( and leave the sign of @xmath54 undetermined ) .
the most stringent current bound on the absolute scale of neutrino mass comes from recent precision tests of cosmology ( notably wmap and the 2df galaxy redshift survey ) .
this limit , a bound of about 1 ev for the sum of neutrino masses , is likely to improve as new surveys are done .
in addition , much improved tritium @xmath75 decay and neutrinoless @xmath82 decay experiments are being planned .
there is one controversial claim of an observation of neutrinoless @xmath82 decay that must be checked soon .
we stress , as we did in i , that this field is producing many new results that promise to impact physics broadly . the most common mechanisms for explaining neutrino mass suggest that current experiments are connected with phenomena far outside the standard model , residing near the grand unified energy scale of @xmath95 gev .
thus there is hope that , by fully determining the properties of neutrinos a few of the unresolved problems have been mentioned here we may equip theorists to begin constructing the next standard model .
neutrino physics is also crucial to astrophysics not just the standard solar model , but also in supernovae and in high - energy astrophysical environments and to cosmology .
we now have identified the first component of particle dark matter , though the significance of neutrino mass is still unclear due to our ignorance of the overall scale .
neutrinos could prove central to one of cosmology s deepest questions , why our universe is matter dominated , rather than matter - antimatter symmetric .
but this is a story for another paper and another time .
haxton and b.r .
holstein,``neutrino physics , '' am .
* 68 * , 15 - 32 ( 2000 ) .
see information , including the nobel lectures , at the website of the nobel foundation http://www.kva.se .
q.r . ahmad et al . , `` measurement of the rate of @xmath96 interactions produced by @xmath17b solar neutrinos at the sudbury neutrino observatory , '' phys .
87 * , 071301 ( 2001 ) pp 1 - 6 .
ahmad et al .
, `` direct evidence for neutrino flavor transformation from neutral - current interactions in the sudbury neutrino observatory , '' phys . rev . lett . * 89 * , 011301 ( 2002 ) pp 1 - 6 .
k. eguchi et al .
, `` first results from kamland : evidence for reactor antineutrino disappearance , '' phys . rev . lett . * 90 * , 021802 ( 2003 ) pp 1 - 6 .
d. n. spergel et al . , `` first year wilkinson microwave anisotropy probe ( wmap ) observations : determination of cosmological parameters , '' [ arxiv : astro - ph/0302213 ] ( submitted to ap .
j. peacock et al .
, `` a measurement of the cosmological mass density from clustering in the 2df galaxy redshift survey , '' nature * 410 * , 169 - 73 ( 2001 ) .
klapdor - kleingrothaus et al . , `` evidence for neutrinoless double beta decay , '' mod .
a*16 * , 2409 - 20 ( 2001 ) ; c. e. aaalseth et al . , `` comment on evidence for neutrinoless double beta decay , '' mod .
a*17 * , 1475 - 78 ( 2002 ) .
e.g. _ , j.f .
donoghue , e. golowich , and b.r .
holstein * dynamics of the standard model * , cambridge university press , new york ( 1992 ) .
r. davis , jr .
harmer , and k.c .
hoffman , `` search for neutrinos from the sun , '' phys .
* 20 * , 1205 - 12 ( 1968 ) ; j. n. bahcall and r. davis , jr . ,
`` solar neutrinos : a scientific puzzle , '' science * 191 * , 264 - 67 ( 1976 ) . y. fukuda et al . , `` evidence for the oscillation of atmospheric neutrinos , '' phys .
lett . * 81 * , 1562 - 65 ( 1998 ) .
bahcall , m.h .
pinsonneault , and s. basu , ap .
j. * 555 * , 990 - 1012 ( 2001 ) .
j. f. wilkerson and r. g. h. robertson , `` direct measurements of neutrino mass , '' in * current aspects of neutrino physics * ( springer - verlag , berlin , 2001 ) , ed .
d. o. caldwell , p. 39 - 64 . | we update our recent didactic survey of neutrino physics , including new results from the sudbury neutrino observatory and kamland experiments , and recent constraints from wmap and other cosmological probes . |
throughout this paper we work only over the complex numbers , dealing with @xmath6 characters .
let @xmath7 . in a natural way , @xmath8 as well .
let @xmath9 be the centralizer of @xmath1 in @xmath4 .
let @xmath10 be any linear character of @xmath11 . in @xcite
, hemmer showed that for @xmath12 the induced character @xmath13 becomes representation stable for @xmath14 .
therefore , these induced characters arise naturally when studying braid group cohomology .
( for more on representation stability and braid group cohomology , see @xcite . )
it was proposed that in general the decomposition of the induced character @xmath15 into irreducible characters of @xmath4 was an open problem .
+ however , the case when @xmath16 was studied in @xcite and @xcite . in this case , @xmath17
. then the linear characters of @xmath11 are precisely the irreducible characters , which are indexed by the numbers @xmath18 and take @xmath1 to @xmath19 .
it was shown that for an irreducible character @xmath20 of @xmath21 , the multiplicity of @xmath20 in the decomposition of @xmath22 is equal to the number of standard young tableaux of shape @xmath23 with major index congruent to @xmath24 mod @xmath2 .
once this is computed , one can use the littlewood - richardson rule or the branching rule to induce the resulting characters up to @xmath4 .
so , in theory , the decomposition of @xmath25 is known , however no explicit formula is available in general .
+ in this paper we will deal with the case when @xmath1 is an @xmath2-cycle of @xmath26 and @xmath27 ( i.e. @xmath28 ) , the trivial character .
we present a partial result toward an explicit formula as well as a formula for the multiplicities of certain irreducible @xmath29 characters appearing in the decomposition .
+ this paper was submitted to the university at buffalo as the author s undergraduate honors thesis .
the work was inspired by dr .
david hemmer and the question posed in the closing of @xcite .
hemmer also served as the author s advisor and oversaw the progress on the paper .
@xmath30 is a * _ partition _ * of @xmath2 , written @xmath31 , if @xmath32 , each @xmath33 , and @xmath34 .
we say each @xmath35 is a * _ part _ * of @xmath23 .
let @xmath36 .
then the * _ young diagram _ * , @xmath37 $ ] , of @xmath23 is the set @xmath38 = \{(i , j ) \in \mathbb{n } \times \mathbb{n } \ | \ j
\leq \lambda_i \}. \notag\ ] ] we say each @xmath39 $ ] is a * _ node _ * of @xmath40 $ ] . if @xmath31 , we represent @xmath40 $ ] by an array of boxes . as an example , consider the partition @xmath41 .
then we visualize @xmath40 $ ] as : ( 50,70)(0,0 ) ( 0,0 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 0,10 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 10,10 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 0,20 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 10,20 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 0,30 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 10,30 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 20,30 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 0,40 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 10,40 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 20,40 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 30,40 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 40,40 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) where the upper left box is defined to be the ordered pair @xmath42 , the upper right is @xmath43 , the lower left is @xmath44 , just like the entries of a matrix . + we will often drop the bracket notation and use @xmath23 and @xmath37 $ ] interchangeably , though it will be clear by context to which we are referring . if @xmath35 is a part of @xmath31 , then @xmath45 is the partition of @xmath46 formed by deleting @xmath35 from @xmath23 .
so @xmath47 . if @xmath48 is a node in the young diagram of @xmath23 , we will write @xmath49
. suppose @xmath50 .
returning to our previous example , it is easy to see that each node @xmath51 is also a node of @xmath23
. we will denote this in the obvious way , @xmath52 .
with this idea in mind , we make a definition .
let @xmath23 and @xmath53 be partitions such that @xmath52 . then the * _ skew diagram _ * @xmath54 is the set of nodes @xmath55 in the case of our example , the skew diagram @xmath54 in this case would be : ( 50,70)(0,0 ) ( 0,0 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 0,10 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 10,10 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 10,20 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 20,30 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 30,40 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) ( 40,40 ) ( 10,10)(0,0 ) ( 0,10)(1,0)10 ( 10,0)(0,1)10 ( 0,0)(1,0)10 ( 0,0)(0,1)10 ( 2.5,1.3 ) one important aspect of young diagrams that will be of great important in this paper are rim hooks . for a skew diagram @xmath56
, we say the unique node @xmath57 such that @xmath58 and @xmath59 for all @xmath60 is the * _ top node _ * of @xmath56 . a * _ rim hook _ * is a skew diagram @xmath56 such that if @xmath61 is not the top node of @xmath56 then either @xmath62 or @xmath63 , but not both .
we will say a rim @xmath24-hook or simply a @xmath24-hook is a rim hook consisting of @xmath24 nodes .
we will say that a partition @xmath23 _ has _ a @xmath24-hook if it is possible to remove a @xmath24-hook from @xmath23 and have the resulting diagram be the young diagram of some partition @xmath64 . to each rim hook @xmath56
is assigned the leg length of @xmath56 .
let @xmath56 be a rim hook .
the * _ leg length of @xmath56 _ * , denoted @xmath65 , is @xmath66 once again returning to our example where @xmath67 , we see that @xmath23 has three rim 4-hooks : in the first and third cases , the 4-hooks have leg length 2 , while in the second case the 4-hook has leg length 1 .
one can also see that @xmath23 does not have any rim 5-hooks , since it is not possible to remove a 5-hook from @xmath23 and have the resulting diagram be the young diagram of a partition
. the basics of representation and character theory will be assumed , and can be found in @xcite .
it is well known ( @xcite ) that there is a one - to - one correspondence between the set of partitions of @xmath2 and the set of irreducible characters of @xmath26 .
for example , @xmath68 corresponds to the trivial character , @xmath69 corresponds to the number of fixed points minus one , and @xmath70 corresponds to the sign character . also , the conjugacy classes of @xmath26 have a natural correspondence to the partitions of @xmath2
. if @xmath71 is of cycle type @xmath23 , @xmath31 , then we will denote the conjugacy class of @xmath72 by @xmath73 .
let @xmath74 .
suppose one wants to evaluate the character @xmath75 on the conjugacy class @xmath76 , which we will denote @xmath77 .
then we have the following theorem which allows one to recursively compute @xmath78 : @xcite [ mnrule ] let @xmath79 .
then @xmath80where the sum is taken over all rim hooks @xmath56 of @xmath23 containing @xmath81 nodes .
now , in a natural way , one can think of @xmath82 as a subgroup of @xmath21 .
suppose @xmath75 is the character of @xmath26 corresponding to @xmath23 and @xmath83 is the character of @xmath82 corresponding to @xmath53
. then one can easily compute the restricted character @xmath84 and the induced character @xmath85 using the branching rule .
let @xmath31 .
we say an * _ inner corner _ * of @xmath40 $ ] is a node @xmath39 $ ] such that @xmath40 - \{(i , j)\}$ ] is the young diagram of some partition of @xmath86 .
we denote any such partition by @xmath87 .
we say an * _ outer corner _
* is a node @xmath88 $ ] such that @xmath40 \cup \{(i , j)\}$ ] is the young diagram of some partition of @xmath89 .
we denote any such partition by @xmath90 .
@xcite [ branchingrule ] let @xmath91 .
then @xmath92 and @xmath93 as an example , suppose @xmath94 and @xmath95 .
using theorem [ branchingrule ] we calculate @xmath96
recall in the introduction of this paper we defined @xmath9 , with @xmath98 .
one can compute @xcite @xmath99 . keeping this in mind
we note the following notation : for @xmath100 , we will write @xmath101 for @xmath102 and @xmath103 . also , if @xmath104 then @xmath105 and @xmath106 . when evaluating any character @xmath107 on the conjugacy class of @xmath4 corresponding to @xmath108 or @xmath109 , we will write @xmath110 and @xmath111 respectively . for the remainder of this paper ,
we will write @xmath112 . [ p:1 ] let @xmath113 .
let @xmath114 be the irreducible character of @xmath4 corresponding to the partition @xmath115 . then @xmath116 using frobenius reciprocity , we have @xmath117 but since @xmath114 is the trivial character , @xmath118 , so we have : @xmath119 [ p:2 ] let @xmath120 .
let @xmath121 be the irreducible character of @xmath4 corresponding to @xmath122 .
then @xmath123 first note that this character records the number points fixed by a permutation and subtracts 1 .
using frobenius reciprocity , we expand the inner product as follows : @xmath124 by previous remarks made at the beginning of this section , the last term in ( [ 1 ] ) becomes : @xmath125 when @xmath28 , @xmath126 and @xmath127 fixes @xmath128 points .
when @xmath129 , @xmath130 fixes @xmath131 points , giving : @xmath132 \notag \\ & = \frac{1}{n \
n!}\left[\displaystyle\sum_{\pi \in s_n } n + \displaystyle\sum_{\pi \in s_n } \chi^{(n-1,1)}(\pi ) + ( n-1 ) \displaystyle\sum_{\pi
\in s_n } \chi^{(n-1,1)}(\pi)\right ] \notag \\ & = \frac{1}{n \ n!}\left[n \ n !
\langle \chi^{(n ) } , \chi^{(n-1,1 ) } \rangle_{s_n}\right ] . \label{2}\end{aligned}\ ] ] but since both @xmath68 and @xmath69 are irreducible , their inner product is 0 .
so ( [ 2 ] ) becomes : @xmath133 [ p:3 ] let @xmath120 .
let @xmath134 be the irreducible character of @xmath4 corresponding to @xmath135 .
then @xmath136 throughout , let @xmath137 .
using frobenius reciprocity , we write : @xmath138 we break the sum up into three pieces : one for @xmath28 , one for @xmath139 ( of which there are @xmath140 such @xmath24 , where @xmath141 denotes euler s totient function ) and one for @xmath142 : @xmath143 .
\intertext{in order to use theorem \ref{mnrule } , we sum over all partitions of $ n$ and rewrite the sum as : } \langle \phi , \chi^{(n , n ) } \rangle_{s_{2n } } & = \frac{1}{n \ n!}\left[n ! \langle \chi^{(n ) } , \chi^{(n , n)}\downarrow_{s_n } \rangle_{s_n } + \varphi(n ) \displaystyle\sum_{\lambda \vdash n } \chi^{(n , n)}_{(n,\lambda ) } |k_{\lambda}|+ \displaystyle\sum_{\begin{subarray}{1 } 1<k < n \\ d_k \not
= 1 \end{subarray}}\displaystyle\sum_{\lambda \vdash n } \chi^{(n , n)}_{((\frac{n}{d_k})^{d_k},\lambda ) } |k_{\lambda}|\right ] .
\label{3}\end{aligned}\ ] ] by theorem [ branchingrule ] , we write : @xmath144 where @xmath145 . then by linearity , we have @xmath146 since all the @xmath75 are irreducible . using theorem [ mnrule ] , @xmath147 so that @xmath148 now let @xmath142 , for some @xmath24 . again with theorem [ mnrule ] ,
we write @xmath149 where @xmath150 .
then @xmath151 we note that there are @xmath152 numbers @xmath24 strictly between 1 an @xmath2 so that @xmath142 , so substituting ( [ 4 ] ) , ( [ 5 ] ) , and([6 ] ) into ( [ 3 ] ) we have : @xmath153 = \frac{1}{n \ n ! } n
\ n ! = 1 .
\notag\ ] ] in the case of @xmath154 it turns out that propositions [ p:1 ] , [ p:2 ] , and [ p:3 ] give a full decomposition .
that is : @xmath155 we notice that our first three results all showed that there are certain irreducible characters appearing in the decomposition of @xmath97 that have constant or stable multiplicities , independent of @xmath2 .
our next result shows that this is not the case for all constituents , but a closed form formula for the multiplicity is known in some cases .
let @xmath120 .
let @xmath156 be the irreducible character of @xmath4 corresponding to @xmath157
. then @xmath158 throughout , @xmath137 .
using frobenius reciprocity we write : @xmath159
if @xmath154 , we are done , by proposition [ p:3 ] . throughout the rest of the proof
we assume @xmath160 . as in the proof of proposition [ p:3 ]
, we break the sum into three pieces : @xmath161 \notag \\ & = \frac{1}{n \ n ! }
\left [ n ! \langle \chi^{(n ) } , \chi^{(2n-2,2 ) } \rangle_{s_n } + \varphi(n ) \displaystyle\sum_{\lambda \vdash n } \chi^{(2n-2,2)}_{(n , \lambda ) } |k_\lambda| + \displaystyle\sum_{\begin{subarray}{1 }
d_k \not = 1 \end{subarray}}\displaystyle\sum_{\lambda \vdash n } \chi^{(2n-2,2)}_{((\frac{n}{d_k})^{d_k},\lambda ) } |k_\lambda| \right ] .
\label{7}\end{aligned}\ ] ] from theorem [ branchingrule ] , we have @xmath162 using theorem [ mnrule ] we write @xmath163 so that @xmath164 when @xmath2 is even , @xmath165 divides @xmath2
. then @xmath166 .
we can then remove the 2-hook from bottom row of @xmath157 , and then successively remove @xmath167 hooks of length 2 from the top row of @xmath157 .
there are @xmath168 ways to do this .
we combine this with theorem [ mnrule ] to see that @xmath169 with @xmath170 .
then @xmath171 so in the case when @xmath2 is even , substituting ( [ 8 ] ) , ( [ 9 ] ) , and ( [ 10 ] ) into ( [ 7 ] ) , we have @xmath172 = \frac{1}{n } \left [ \binom{n}{2 } + \frac{n}{2 } \right ] = \frac{1}{n } \left [ \frac{n(n-1)}{2 } + \frac{n}{2 } \right ] = \frac{n-1}{2 } + \frac{1}{2 } = \frac{n}{2 } \notag\ ] ] as desired .
now , when @xmath2 is odd , @xmath173 does not divide @xmath2 .
then @xmath165 is not an integer and thusly does not divide @xmath2 . as a result , we can not remove the hook of length 2 from the bottom row of @xmath157 .
so when we apply theorem [ mnrule ] , the trivial character does not appear in the decomposition and we have @xmath174 with @xmath175
. then @xmath176 so then substituting ( [ 8 ] ) , ( [ 9 ] ) , and ( [ 11 ] ) into ( [ 7 ] ) , we have @xmath177 giving the result .
we now present a theorem that generalizes the previous propositions and gives a formula for the multiplicities of a number of the irreducible characters of @xmath4 appearing in the decomposition of @xmath97 .
let @xmath179 .
let @xmath180 be the irreducible character of @xmath4 corresponding to @xmath178 .
for @xmath181 , let @xmath182 , and @xmath183
. then @xmath184 .
\notag\ ] ] with frobenius reciprocity , we write @xmath185 as usual , we split the sum into three pieces : @xmath186 \notag \\ & = \frac{1}{n \ n!}\left [ n!\langle \chi^{(n ) } , \chi^{(2n - k , k)}\downarrow_{s_n } \rangle_{s_n } + \varphi(n)\displaystyle\sum_{\lambda \vdash n } \chi^{(2n - k , k)}_{(n,\lambda ) } |k_\lambda| + \displaystyle\sum _ { \begin{subarray}{1 } 1<h < n \\
d_h \not = 1 \end{subarray } } \sum_{\lambda \vdash n } \chi^{(2n - k , k)}_{((\frac{n}{d_h})^{d_h } , \lambda ) } |k_\lambda| \right ] .
\label{12}\end{aligned}\ ] ] since @xmath179 , we can remove the @xmath24 blocks from the bottom row of @xmath178 and remove @xmath187 blocks from the top row of @xmath178 , which leaves @xmath2 blocks remaining .
we can do this removal in @xmath188 ways , so with @xmath189 , we have @xmath190 theorem [ mnrule ] gives @xmath191 since @xmath179
. then @xmath192 now suppose there is some @xmath193 so that @xmath194
. then @xmath195 is a product of @xmath196 @xmath197-cycles .
if @xmath198 is of cycle type @xmath23 then @xmath199 by theorem [ mnrule ] , in order for @xmath68 to have non - zero multiplicity in the decomposition of @xmath200 , we have to be able to remove the @xmath24-hook from the bottom row of @xmath178 .
so if @xmath197 does not divide @xmath24 then this is not possible .
then in this case @xmath201 with @xmath170 . now
, suppose that for some @xmath193 , @xmath194 and furthermore that @xmath197 divides @xmath24 . then we can successively remove the @xmath202 hooks of length @xmath197 from the bottom row of @xmath178 and remove the @xmath203 hooks of length @xmath197 from the top row of @xmath178 , which will result in @xmath68 having positive multiplicity in the aforementioned decomposition .
in fact , a simple counting argument via theorem [ branchingrule ] shows the exact multiplicity will be @xmath204 .
then in this case @xmath205 with @xmath175 . then
( [ 16 ] ) and ( [ 17 ] ) give @xmath206 substituting ( [ 13 ] ) , ( [ 14 ] ) , ( [ 18 ] ) into ( [ 12 ] ) we have @xmath207 = \frac{1}{n } \bigg [ \binom{n}{k } + \displaystyle\sum _ { \begin{subarray}{1 } 1<h < n \\ d_h \not = 1\\ \frac{n}{d_h } \big| k \end{subarray } } \binom{d_h}{l_k } \bigg]\ ] ] as claimed .
the preceding work is only the beginning of a large selection of problems to be worked out .
it is possible that there are more stable multiplicities ( independent of @xmath2 ) in this decomposition .
also , the multiplicities and formulas found here only cover a small number of partitions and therefore characters .
one may find that _ all _ characters have a stable or closed form formula for their multiplicities .
note that in this paper we only discuss the trivial character of @xmath11 , and much can be learned from studying the decomposition of the non - trivial characters of @xmath11 when induced up to @xmath4 , which arise in braid group cohomology
. it may be possible to learn more by first decomposing the character @xmath208 , studying this character , and then inducing the resulting constituents up to @xmath4 . | we give explicit multiplicities and formulas for multiplicities of characters appearing in the decomposition of the induced character @xmath0 , where @xmath1 is an @xmath2-cycle , @xmath3 is the centralizer of @xmath1 in @xmath4 , and @xmath5 denotes the trivial character on @xmath3 . |
the connection between random matrix theory and number theory began with the work of montgomery @xcite when he conjectured that the distribution of the complex zeros of the riemann zeta function follows the same statistics as the eigenvalues of a random matrix chosen from @xmath3 generated uniformly with respect to haar measure .
this conjecture is supported by numerical evidence @xcite and also by further work @xcite suggesting that the same conjecture is true for more general @xmath2-functions .
for all these @xmath2-functions there is a generalized riemann hypothesis that the non - trivial zeros lie on a vertical line in the complex plane .
the conjectures mentioned above concern the statistics of the zeros high on this critical line .
the philosophy of katz and sarnak @xcite extended the connection with random matrix theory by proposing that rather than averaging over many zeros of a given @xmath2-function , if the zeros near to the point where the critical line crosses the real axis are averaged over a family of naturally connected @xmath2-functions , then they will be found to follow the statistics of the eigenvalues of one of the three classical compact groups of random matrices : @xmath3 , @xmath4 or @xmath5 , where again the statistics are computed with respect to the probability measure given by haar measure .
there is numerical evidence for this conjecture as well @xcite , and strong support is given to it by the rigorous work of katz and sarnak in the case of function field zeta functions .
for a review of applications of random matrix theory to questions in number theory see , for example , @xcite or @xcite .
there has been a series of papers , starting with @xcite and continuing with @xcite , examining how random matrix theory can be used to predict the distribution of values of the riemann zeta function and other @xmath2-functions , either averaged over an interval high on the critical line , or over a family at the critical point where the critical line crosses the real axis . for large values of the natural asymptotic parameter , for example the variable ordering the @xmath2-functions within the family ,
the moments of @xmath2-functions are conjectured to split into a product of an arithmetic contribution , determined by the family being averaged over , and a component derived from a random matrix calculation - the corresponding moment of the characteristic polynomial of the matrices in one of the three matrix groups @xmath3 , @xmath4 or @xmath5 .
the asymptotic parameter on the random matrix side is the dimension of the matrix @xmath6 and a natural equivalence can be made between the two .
as will be reviewed in the following section , this conjecture for the leading order behaviour of moments of @xmath2-functions has been used by conrey , keating , rubinstein and snaith @xcite to predict the frequency , within a family of elliptic curves , of an @xmath2-function taking the value zero at the critical point .
this is done by using random matrix theory to predict the value distribution of the @xmath2-values at the critical point and then the discretization of these values @xcite to calculate a probability of the @xmath2-value being zero .
see also david , fearnley and kisilevsky @xcite for a similar use of random matrix theory to predict frequency of vanishing at the critical point amongst families of elliptic curve @xmath2-functions twisted by cubic characters .
in section [ sect : derivdist ] we calculate the value distribution of the first non - zero derivative of the characteristic polynomial at the point one ( corresponding to the value at the critical point in the analogy with @xmath2-functions ) when a given number of eigenvalues are conditioned to lie at the point one .
we find , at equation ( [ eq : delaunay ] ) , that moment of the @xmath1th derivative grows like @xmath7 where @xmath8 is the characteristic polynomial of @xmath9 , @xmath10 is the barnes double gamma function ( see [ eq : g ] ) and the angle brackets denote an average over the set of matrices from @xmath0 with @xmath1 eigenvalues lying at the point one ( @xmath11 ) .
we also show that the probability that @xmath12 over those @xmath13 with @xmath1 eigenvalues at 1 ( again , @xmath11 ) is , for small @xmath14 , given by ( see ( [ eq : nrzero ] ) and the sentence following ) @xmath15 for the function @xmath16 given at ( [ eq : f ] ) .
if a discretization of the values of derivatives of @xmath2-functions at the critical point were known , this would provide a way to predict the frequency of zeros of various orders occurring at this point .
this has been investigated in the case of a single zero at the critical point @xcite .
in this situation , @xmath17 , the results of @xcite give support to the validity of the model presented here . the result presented in section [
sect : derivdist ] on the moments of the first derivative of characteristic polynomials from @xmath18 ( equation [ eq : introdel ] with @xmath17 ) has already been applied by delaunay in @xcite in order to predict the moments of the orders of tate - shafarevich groups and regulators of elliptic curves with odd rank belonging to a family of quadratic twists . in section [
sect : evaluestats ] , we briefly discuss the eigenvalue statistics of random matrices with a certain number of eigenvalues conditioned to lie at one .
we note that while we believe that the model presented in section [ sect : derivdist ] should apply to the @xmath2-functions selected from families of quadratic twists of elliptic curves for the property that they have high order zeros at the critical point , there has been some very interesting theoretical work computing the one- and two - level densities for parametric families of elliptic curves that implies that the zeros of the associated @xmath2-functions follow a different model ( see @xcite and @xcite ) .
this does not seem to be a contradiction , as the zero statistics are examined over collections of @xmath2-functions selected in very different ways , but it certainly makes the question of the zero statistics of @xmath2-functions with high - order zeros at the critical point very intriguing . see section [ sect : discussion ] for further discussion on this issue .
we review here the results of @xcite which apply random matrix theory to predicting the frequency of vanishing at the critical point of the @xmath2-functions in the family of elliptic curves described below ; or equivalently , assuming the birch and swinnerton - dyer conjecture , the frequency of rank 2 or higher curves occurring in the family of elliptic curves .
the motivation for the random matrix calculations presented in section [ sect : derivdist ] is the expectation that they may be used similarly to examine curves of higher rank .
we consider an @xmath2-function ( defined by the dirichlet series and euler product below when @xmath19 ) @xmath20 that is associated to an elliptic curve @xmath21 over @xmath22 @xmath23 the coefficients @xmath24 , for prime @xmath25 , are determined by @xmath26 , where @xmath27 counts the number of pairs @xmath28 , with @xmath29 , such that @xmath30 , plus one for the point at infinity .
@xmath31 is the discriminant of the cubic @xmath32 . for an extremely clear introduction to elliptic curves in the context discussed here ,
see the review paper by rubin and silverberg @xcite .
a family of elliptic curves is formed by @xmath33 for integers @xmath34 that are fundamental discriminants .
( a fundamental discriminant is an integer other than 1 that is not divisible by the square of any prime other than two and that satisfies @xmath35 or @xmath36 . ) the corresponding family of @xmath2-functions , ordered by @xmath37 , are @xmath38 where the characters @xmath39 are kronecker s extension of the legendre symbol , defined for prime @xmath25 as @xmath40 the family of curves @xmath41 is called the family of quadratic twists of @xmath21 because the characters @xmath42 are real , quadratic dirichlet characters .
evidence ( see for example @xcite ) points to the zeros near the critical point ( where the critical line crosses the real axis ) of such a family of @xmath2-functions having statistics like the eigenvalues near 1 of matrices from @xmath4 with haar measure . to explain this correspondence more clearly
, about half of the @xmath2-functions in this family are expected to have an even functional equation that relates @xmath43 to @xmath44 , meaning that the zeros on the critical line ( @xmath45 ) appear in complex conjugate pairs and if there is a zero at the critical point then it must be of even order , while the other half have an odd functional equation , forcing a zero of order 1,3,5 , ... at the critical point and all other zeros appearing in complex conjugate pairs .
matrices in @xmath4 also have eigenvalues that appear in complex conjugate pairs , with the possible exception of unpaired zeros at 1 and -1 .
the low - lying zeros of elliptic curve @xmath2-functions from the family above that have even functional equation show the statistics that eigenvalues of @xmath46 display near the point one .
these eigenvalues occur in complex conjugate pairs and an eigenvalue at one must have even multiplicity . in contrast , the low - lying zeros of the @xmath2-functions above that have an odd functional equation display the same statistics as @xmath18 eigenvalues near one , since in this case there is always an eigenvalue at one and it has to have odd multiplicity .
( note that we could just as well have chosen @xmath47 matrices ( orthogonal with determinant -1 ) to model the @xmath2-functions with odd functional equation , since the statistics of these eigenvalues near to one are the same as those of @xmath18 , see for example @xcite . )
both numerical and analytical evidence has been given already @xcite that random matrix theory can be used to conjecture the frequency of @xmath2-functions with even functional equation vanishing at the critical point in a family such as that described above .
this is particularly important because of the birch and swinnerton - dyer conjecture which suggests that the order of the zero of an elliptic curve @xmath2-function at the critical point is the same as the rank of the elliptic curve itself .
the rank is a non - negative integer @xmath48 that characterizes the number of rational points on the elliptic curve .
the set of pairs @xmath49 that satisfy equation ( [ eq : elliptic ] ) , plus one point at infinity , form a commutative group ( with a standard law of addition defined on the curve ) that is isomorphic to @xmath50 .
@xmath51 is the subgroup of elements of finite order and @xmath48 is the rank of the curve .
the frequency of various ranks in a family such as the one defined at ( [ eq : elliptic_d ] ) , or even whether the rank is bounded in such a family of elliptic curves , are important unanswered questions . in @xcite the first step
is taken towards using random matrix theory to generate conjectural answers to these questions by addressing the frequency of @xmath2-functions from elliptic curve families having any zero at all at the critical point . the argument used
will be sketched below to illustrate the role played by random matrix theory .
the purpose of this paper is to perform the random matrix calculations needed to predict the frequency of a zero of a _ given order _ at the critical point . applying the random matrix results from section [ sect : derivdist ] to the order of vanishing of @xmath2-functions involves the knowledge of complicated arithmetic quantities and much more work is needed in this area . some preliminary numerical investigation has been done @xcite for the case of third order vanishing .
we now review the results in @xcite .
we define the family of @xmath2-functions discussed above with even functional equation : @xmath52 the zeros of these @xmath2-functions near the critical point , @xmath53 , show the same statistics as the eigenvalues of matrices from @xmath46 , and so the @xmath2-values at the critical point are modelled by the characteristic polynomial @xmath54 evaluated at the point @xmath55 . here
@xmath56 are the eigenvalues of the matrix @xmath57 .
the moments of @xmath58 are easily calculated using weyl s expression @xcite for haar measure on the conjugacy classes of @xmath46 and a form of selberg s integral to be @xcite @xmath59 the @xmath2-function moments are then conjectured to have the form @xcite @xmath60 for large @xmath61 .
here @xmath62 ( from equating the density of zeros near the critical point with the density of the matrix eigenvalues ) , the sum is over fundamental discriminants @xmath34 , @xmath63 is the number of terms in the sum and @xmath64 is an euler product that contains arithmetic information specific to the elliptic curve @xmath21 and the family of @xmath2-functions being averaged over . in practice
, it is often a subset of @xmath65 that is summed over .
if , for example , we select those @xmath2-functions @xmath43 in @xmath65 with @xmath66 and further restricted by a condition on @xmath67 , if @xmath68 is odd , and on @xmath69 , if @xmath68 is even , then the arithmetic factor would be @xmath70 see @xcite and @xcite , section 4.4 , for more examples .
next we consider the distribution of the values of the characteristic polynomials of @xmath46 matrices at the point 1 .
if @xmath71 is the probability that the characteristic polynomial of a matrix chosen from @xmath46 with haar measure has a value between @xmath72 and @xmath73 , then @xmath74 for @xmath75 , since for small @xmath72 the behaviour is dominated by the pole of @xmath76 at @xmath77 . here
@xmath78 denotes a path of integration along the vertical line from @xmath79 to @xmath80 .
for large @xmath6 , @xmath81 ( @xmath82 is the barnes double gamma function , defined as @xcite : @xmath83/2 } \prod_{n=1}^{\infty } \left [ ( 1+z / n)^n e^{-z+z^2/(2n)}\right],\ ] ] where @xmath84 is euler s constant .
see also ( [ eq : g ] ) for more properties of this function . ) since the probability that an element of @xmath46 has a characteristic polynomial whose value at 1 is @xmath14 or smaller is @xmath85 , we find that the the behaviour of this probability for small @xmath14 and large @xmath6 is @xmath86 we see from equation ( [ eq : elliptic_moment ] ) that for large @xmath34 , moments of @xmath2-functions are conjectured to be just @xmath64 ( the prime product ) times the random matrix moment @xmath76 .
if this is true , then we define @xmath87 as the probability , amongst members of @xmath65 , that @xmath88 , for @xmath37 around @xmath89 , will take a value between @xmath72 and @xmath73 , giving @xmath90 and an approximation for this probability for small @xmath72 should be @xmath91 here equating densities of zeros means @xmath92 .
so , roughly , random matrix theory predicts that the probability that an @xmath2-function @xmath93 , with @xmath37 close to @xmath94 , has a value at @xmath53 that is @xmath14 or smaller is @xmath95 but these @xmath2-functions are constrained to take only certain discretized values at the critical point @xmath53 .
the @xmath2-values have the form @xcite : @xmath96 where the @xmath97 are integers , the fourier coefficients of a half - integral weight form .
so , if @xmath98 then @xmath99 and the first thought would be to take @xmath100 in ( [ eq : discret ] ) and then to integrate @xmath94 from 0 to @xmath61 . however , there is arithmetical information encoded in the @xmath97 ( one is not always the smallest allowed non - zero value ) and so to avoid this problem , the conjecture stated in @xcite is restricted to prime discriminants .
( conrey , keating , rubinstein , snaith ) : [ conj:1 ] let @xmath21 be an elliptic curve defined over @xmath101 .
then there is a constant @xmath102 such that @xmath103 ( the conjecture was originally stated in @xcite with @xmath104 , but in fact numerics have revealed that the constant can be zero @xcite for certain families .
an explanation of such a case with @xmath105 was given by delaunay @xcite . ) with the birch and swinnerton - dyer conjecture , this suggests that out of all the elliptic curves associated with @xmath2-functions in @xmath65 with prime discriminant @xmath106 ( there are of order @xmath107 of them ) , a number of order @xmath108 of them should have rank two or greater .
the @xmath109 has been predicted previously by sarnak using different arguments , but random matrix theory adds more detailed information in the form of the power on the logarithm . for numerical evidence supporting the conjecture , see @xcite and @xcite . in the next section
we will calculate the distribution of values equivalent to ( [ eq : dist ] ) that is required to probe questions of elliptic curves of a given rank occurring in families of quadratic twists ( equation ( [ eq : elliptic_d ] ) ) .
in this section we will calculate the probability of the @xmath1th derivative of a characteristic polynomial ( [ eq : cp ] ) of a random @xmath0 matrix taking a value less then @xmath14 at @xmath55 when all lower derivatives are constrained to be zero at this point .
that is , we perform the random matrix average over the ( measure zero ) subset of @xmath0 of matrices with @xmath1 eigenvalues at the point 1 on the unit circle , and @xmath110 other eigenvalues occurring in complex conjugate pairs ( @xmath11 )
. we will also illustrate the @xmath1-level density of the @xmath110 eigenvalues _ not _ constrained to lie at 1 . for a matrix @xmath13 with @xmath1 eigenvalues equal to one ( @xmath11 ) ,
the characteristic polynomial looks like @xmath111 we consider the @xmath1th derivative @xmath112 \prod_{j=1}^m ( 1-e^{i\theta_j-\alpha})(1-e^{-i\theta_j-\alpha})\big|_{\alpha=0}\nonumber\end{aligned}\ ] ] and use this to define @xmath113 the matrices with @xmath1 eigenvalues equal to one form a set of measure zero within @xmath0 with haar measure , for @xmath114 . for even @xmath115 , haar measure on the conjugacy classes of @xmath116
is given by @xmath117 with @xmath118 a normalization constant , and we define the probability of having @xmath119 eigenvalues at 1 and the rest within infinitesimal neighbourhoods of @xmath120 as @xmath121 here @xmath122 is the normalization constant .
the same procedure for @xmath6 odd , starting from the measure on @xmath123 @xmath124 with normalization constant @xmath125 , leads to a general expression for the measure on the matrices in @xmath0 ( where now @xmath6 can be even or odd ) with @xmath1 eigenvalues equal to 1 ( @xmath11 ) : @xmath126 thus the quantity to calculate is @xmath127 this can be evaluated using a form of selberg s integral ( for details see @xcite ) : @xmath128 if @xmath129 , @xmath130 and @xmath131 .
we have @xmath132 , @xmath133 and @xmath134 , so the integral in equation ( [ eq : integral ] ) is @xmath135 but the normalization constant can also be evaluated by selberg s integral to be @xmath136 , giving us @xmath137 consider @xmath1 fixed and finite and @xmath138 large . if the probability that @xmath139 takes a value between @xmath72 and @xmath73 is given by @xmath140 then from a standard result in probability ( with @xmath78 denoting a path of integration along the vertical line from @xmath79 to @xmath80 ) @xmath141 from ( [ eq : dists ] ) ) at the point one of the @xmath1th derivative of the characteristic polynomial of the subset of matrices from @xmath142 that are conditioned to have @xmath1 eigenvalues at one .
figures c ) and e ) show in more detail the behaviour at the origin of figures b ) and d ) ( see equation ( [ eq : nrzero])).,title="fig : " ] + from ( [ eq : dists ] ) ) at the point one of the @xmath1th derivative of the characteristic polynomial of the subset of matrices from @xmath142 that are conditioned to have @xmath1 eigenvalues at one . figures
c ) and e ) show in more detail the behaviour at the origin of figures b ) and d ) ( see equation ( [ eq : nrzero])).,title="fig : " ] from ( [ eq : dists ] ) ) at the point one of the @xmath1th derivative of the characteristic polynomial of the subset of matrices from @xmath142 that are conditioned to have @xmath1 eigenvalues at one .
figures c ) and e ) show in more detail the behaviour at the origin of figures b ) and d ) ( see equation ( [ eq : nrzero])).,title="fig : " ] + from ( [ eq : dists ] ) ) at the point one of the @xmath1th derivative of the characteristic polynomial of the subset of matrices from @xmath142 that are conditioned to have @xmath1 eigenvalues at one . figures
c ) and e ) show in more detail the behaviour at the origin of figures b ) and d ) ( see equation ( [ eq : nrzero])).,title="fig : " ] from ( [ eq : dists ] ) ) at the point one of the @xmath1th derivative of the characteristic polynomial of the subset of matrices from @xmath142 that are conditioned to have @xmath1 eigenvalues at one .
figures c ) and e ) show in more detail the behaviour at the origin of figures b ) and d ) ( see equation ( [ eq : nrzero])).,title="fig : " ] we are interested in the behaviour at small @xmath72 , and this is dominated by the nearest pole to zero of the integrand : the pole at @xmath143 of @xmath144 . thus for small @xmath72 @xmath145 the probability that @xmath12 over those @xmath146 with @xmath1 eigenvalues at 1 is therefore @xmath147 for small @xmath14 and the dependency on @xmath1 and @xmath138 is determined by @xmath148 we can discover the behaviour of @xmath149 for large @xmath138 by introducing the barnes g - function @xcite : @xmath150/2 } \prod_{n=1}^{\infty } \left [ ( 1+z / n)^n e^{-z+z^2/(2n)}\right],\ ] ] which has zeros at the negative integers , @xmath151 , with multiplicity @xmath1 ( @xmath152 ) .
other properties useful to us are that @xmath153 and furthermore , for large @xmath154 @xmath155 thus @xmath156 so we can write @xmath157 and expanding the @xmath82-functions for large @xmath138 gives @xmath158 note that we can also use ( [ eq : gasymp ] ) to revisit @xmath159 ( defined at [ eq : integral ] ) and examine that asymptotically for large @xmath138 . this gives us the large @xmath138 behaviour of the moments at the point one of the @xmath1th derivative of characteristic polynomials for which all lower derivatives vanish .
we have @xmath160 to get an idea of what the eigenvalue statistics of the remaining eigenvalues look like when we condition @xmath1 eigenvalues to lie at one in the manner described in ( [ eq : condprob ] ) , we briefly review the @xmath161-level densities of the unconditioned eigenvalues .
these calculations are not new ; these same matrix ensembles were studied by nagao and wadati @xcite and also duenez @xcite and are discussed in the book by forrester @xcite . concurrently with the present work , this matrix model
has also been discussed in connection with the zero statistics of elliptic curve @xmath2-functions by miller @xcite ( with appendix by duenez ) .
however , we include a brief discussion of the statistics here to illustrate the effect that forced eigenvalues at one have on the remaining eigenvalues .
the terminology for these statistics varies ( for example , mehta @xcite refers to them as @xmath161-point correlation functions ) so we will write ( with @xmath162 the normalization constant ) @xmath163 and refer to @xmath164 as the @xmath161-level density ( see @xcite for further discussion of these densities ) . by standard techniques
( see for example @xcite , chapter 5 , @xcite , chapter 4 ) @xmath165 is defined in terms of the jacobi polynomials @xmath166 ( described in detail in @xcite ) with the orthogonality condition @xmath167 where @xmath168 scaling the eigenvalues and using asymptotic formulae for the jacobi polynomials ( see @xcite , page 197 ) we find that in the large - matrix limit ( where @xmath169 are bessel functions of the first kind ) @xmath170 the scaled 1-level density is therefore @xmath171 and can be seen to behave like a constant times @xmath172 for small @xmath173 , where we remember that @xmath1 is the number of degenerate eigenvalues
. this increasing repulsion of the first eigenvalue not located at 1 can be seen in figure [ fig : onedensity ] . ) for various values of @xmath1 , the degeneracy of the eigenvalue at one . ]
we have shown that the probability that @xmath12 over those @xmath13 with @xmath1 eigenvalues at 1 ( @xmath11 ) is , for small @xmath14 , given by @xmath174 and the mean value of the @xmath1th derivative grows like @xmath175 where the angle brackets denote an average over the set of matrices from @xmath0 with @xmath1 eigenvalues lying at the point one ( @xmath11 ) . the result ( [ eq : del ] ) , see also ( [ eq : delaunay ] ) , was applied by delaunay in @xcite to predict moments of the orders of tate - shafarevich groups and the regulators of elliptic curves belonging to a family of quadratic twists , see ( [ eq : elliptic_d ] ) .
the birch and swinnerton - dyer conjecture provides a formula for the first non - zero derivative of an @xmath2-function in terms of various quantities related to the associated elliptic curve .
one of these quantities is the order of the tate - shafarevich group , and another , in the case of the first derivative of @xmath2-functions with odd functional equation , is the regulator . for families of @xmath2-functions with even functional equation
the result ( [ eq : delaunay ] ) is used with @xmath176 by delaunay to predict the asymptotic form of moments of the order of the tate - shafarevich group for the associated family of elliptic curves , and for families with odd functional equation the case @xmath17 was used to conjecture the form of moments of the regulator . to predict , for example , the number of @xmath2-functions in @xmath177 ( the family defined like @xmath65 in ( [ eq : f+ ] ) but with odd functional equation ) that have a zero of order at least three at the point 1 we need the result ( [ eq : disc1 ] ) for @xmath17 plus some information analogous to ( [ eq : cutoff ] ) giving the smallest non - zero value that can be taken by the derivative of an @xmath2-function from this family . a formula like ( [ eq : cutoff ] ) is not known at the moment for derivatives of @xmath2-functions ,
but numerics suggest @xcite that there is a gap between zero and the smallest non - zero value of @xmath178 , which allows random matrix theory to be applied to the question of the number of @xmath2-functions with a zero of order at least three at the point 1 .
further work in this area is ongoing , but it is also worth noting that the numerics in @xcite support the random matrix model presented here , as the numerical cumulative distribution of values of @xmath178 for @xmath2-functions in @xmath177 shows the expected @xmath179 behaviour near @xmath180 , agreeing with the exponent @xmath181 in ( [ eq : disc1 ] ) when @xmath17 .
this leads us to believe that the measure in ( [ eq : measure ] ) should correctly model the zeros near the point 1 of @xmath2-functions with at least an @xmath1th order zero at 1 selected from the family @xmath177 or @xmath65 ( depending on whether @xmath1 is odd or even , respectively ) , but we note again that this is not the only model for @xmath2-functions with @xmath1 zeros at 1 .
the measure presented here is what miller @xcite calls the interaction model , and he contrasts this with the independent model where the @xmath1 eigenvalues at 1 have no effect on the statistics of the remaining eigenvalues .
this gives , for example for matrices from @xmath182 with @xmath48 eigenvalues at 1 , the set of matrices @xmath183 with a @xmath184 identity block in the upper left corner and with the joint probability density on the remaining eigenvalues being haar measure on @xmath185 , that is @xmath186 miller finds numerical evidence @xcite , and for theoretical results see @xcite and @xcite , that the independent model agrees with the zero statistics of @xmath2-functions of a parametric family of elliptic curves constructed so that each member of the family has rank @xmath48 . through the birch and swinnerton - dyer conjecture
, the associated @xmath2-functions are expected to have a zero of order @xmath48 at the point 1 , and miller finds evidence for this high - order zero , as well as for the fact that its presence does not affect the statistics of the other zeros in the way that the model presented here would ( for example as illustrated in figure 2 ) , but rather the zeros follow miller s independent model .
this is not a contradiction to the proposal that the model presented here ( the same as miller s interaction model ) is correct for the family of quadratic twists of elliptic curve @xmath2-functions .
it will be the subject of further work to produce convincing numerics for the zero statistics of the elliptic curve @xmath2-function families discussed in this paper , although this is a difficult task due to the small number of curves in these families with rank greater than or equal to two .
thanks to brian conrey for his number theoretical insight in recognizing the potential importance of these calculations from the start , to peter forrester for advice on orthogonal polynomials , to eduardo due~ nez , jon keating and steven miller for comments on early drafts of this paper and to steve gonek , chris hughes , michael stoltz and mark watkins for asking and answering a key question in the space of one lunchtime ! my gratitude also to the royal society for supporting my research with the dorothy hodgkin fellowship , to epsrc for an advanced research fellowship and to the isaac newton institute for mathematical sciences for their hospitality during the programme
random matrix approaches in number theory " during which this paper was prepared .
conrey , notes on eigenvalue distributions for the classical compact groups , in _
recent perspectives on random matrix theory and number theory _ , pages 11145 .
cambridge university press , cambridge , 2005 .
conrey , j.p .
keating , m.o .
rubinstein , and n.c .
snaith , on the frequency of vanishing of quadratic twists of modular @xmath187-functions , in _ number theory for the millennium i : proceedings of the millennial conference on number theory _ ; editor , m.a .
bennett et al .
, pages 301315 .
a k peters , ltd , natick , 2002 , arxiv : math.nt/0012043 . | the value distribution of derivatives of characteristic polynomials of matrices from @xmath0 is calculated at the point 1 , the symmetry point on the unit circle of the eigenvalues of these matrices .
we consider subsets of matrices from @xmath0 that are constrained to have @xmath1 eigenvalues equal to 1 , and investigate the first non - zero derivative of the characteristic polynomial at that point .
the connection between the values of random matrix characteristic polynomials and values of @xmath2-functions in families has been well - established .
the motivation for this work is the expectation that through this connection with @xmath2-functions derived from families of elliptic curves , and using the birch and swinnerton - dyer conjecture to relate values of the @xmath2-functions to the rank of elliptic curves , random matrix theory will be useful in probing important questions concerning these ranks . |
dark matter is omnipresent in the universe .
most of it is non - baryonic , and a favoured candidate is a weakly interacting massive particle ( wimp ) , often generically taken to be the lightest stable relic particle surviving from when the universe was supersymmetric .
the freeze - out of such a neutralino of mass @xmath0 occurs at @xmath1 , and the annihilation cross - section determines the current value of the cdm density @xmath2 for minimal susy , for example , one can derive the relation between annihilation cross - section and particle mass , and compute the various annihilation products , including continuum gamma rays from @xmath3 decays and line gamma rays from rare quark decays ( c.f .
bergstrm 2000 for a recent review ) .
while cosmological observations specify @xmath4 scanning over minimal susy parameter space results in an uncertainty in the gamma - ray emissivity of several orders of magnitude . our galactic halo is a logical place to look for evidence of annihilations .
unfortunately , for a uniform dark halo with a realistic density profile , even the most optimistic models fall short of the observed diffuse high - galactic - latitude gamma - ray flux , as measured by egret , by an order of magnitude or more ( ullio et al .
in fact , high - resolution numerical simulations show that the dark halo has considerable substructure ( klypin et al .
1999 ; moore et al .
this substructure may boost the annihilation flux substantially .
if this is indeed the case , then the isotropic diffuse background flux from the many small halos that merged in the past into our halo and others may also become significant .
there is considerable uncertainty among cosmological halo simulators , however , about the quantitative role of substructure in our dark halo , and of the concentration of the substructure and of the dark matter itself towards the centre of the galaxy .
no reliable estimates have been given up till now of the properties of halo substructure on very small scales , and in particular of their dependence on halo mass and their evolution with cosmological epoch .
we have developed a semi - analytical model of halo formation which is capable of following the key physics of tidal disruption of substructure during merging . in principle
this approach has arbitrarily high resolution .
hence we are able to provide robust calculations of the annihilation flux generated within our own halo , but also especially of the component generated during the evolution of structure at early times , and visible as an isotropic gamma - ray background at the present day . in this paper
, we calculate the flux produced by annihilations in simple cdm halos , relative to the flux produced in a uniform background .
we correct this result for halo substructure , and integrate over a large volume to determine the cosmological background from wimp annihilation as a function of redshift .
the outline of this paper is as follows . in section 2
, we define a dimensionless flux multiplier @xmath5 that accounts for the enhanced rate of two - body interactions produced by inhomogeneities in the dark matter distribution , and determine its value for simple virialised halos . in section 3 we calculate @xmath5 for cosmological volumes , using analytic estimates of the halo mass function and halo concentrations , and determine its redshift dependence .
finally , in section 4 we study the contribution to @xmath5 from substructure within virialised halos , and calculate @xmath5 for a set of realistic halos generated using a semi - analytic model of halo substructure . throughout this paper
we assume a lambda - cdm ( lcdm ) cosmology with a cosmological constant @xmath6 , a matter density @xmath7 and a hubble parameter @xmath8 , with @xmath9 .
in current hierarchical models , cold dark matter is expected to form centrally concentrated halos with a characteristic density profile .
dense substructure is abundant within these halos , as a relic from earlier stages of the hierarchical merging process .
since the annihilation flux is quadratic in the density , these inhomogeneities will increase the flux from a halo of a given mean density .
we begin by computing a dimensionless quantity that describes this enhancement . in the next section
we will then study the evolution of this quantity with epoch .
if dark matter consists of neutralinos with a mass @xmath0 and a velocity - averaged cross - section for annihilation @xmath10 , then the annihilation flux produced within a volume @xmath11 will be @xmath12[fluxm0 ] where @xmath13 is the local density of cdm . for non - relativistic particles ,
@xmath10 is approximately independent of @xmath14 so the flux will just be proportional to @xmath15 .
since the rate depends quadratically on the density , the total rate from a given mass within a given volume will be higher if the dark matter is distributed inhomogeneously .
we can study this enhancement by defining the dimensionless flux multiplier @xmath16 for a distribution within a volume @xmath11 , where @xmath17 is the average density within this volume .
this function can also be written as a mass - weighted density average : @xmath18 where @xmath19 is the total mass within @xmath11 .
clearly @xmath20 for a homogeneous distribution , while for a power - law density profile @xmath21 , @xmath22 provided @xmath23 . for @xmath24 , integrating equation ( [ fluxm1 ] ) from @xmath25 to @xmath26 gives : @xmath27 so @xmath5 diverges logarithmically as @xmath25 goes to zero .
two analytic density profiles are commonly used to fit the spherically averaged properties of dark matter halos , the nfw profile ( navarro , frenk & white 1996 , 1997 ) , and the moore profile ( moore et al .
we can specify these generically as @xmath28 with @xmath29 for the nfw profile and @xmath30 for the moore profile .
the total mass within radius @xmath31 is @xmath32 with @xmath33 for the nfw profile and @xmath34 for the moore profile , while the mean density within this radius is simply @xmath35 where @xmath36 . in a cosmological
setting , halos are virialised out to a radius @xmath37 corresponding to an overdensity @xmath38 of roughly 200 relative to the background
. the concentration @xmath39 of a halo describes the size of this radius relative to @xmath40 .. ] calculating the flux multiplier over the virialised region of a halo , we get a function which depends only on @xmath41 : @xmath42 thus for the nfw profile @xmath43 with a limiting value of @xmath44 for small @xmath41 , whereas for the moore profile , @xmath45 @xmath46^c_{x_{\rm min}}. \label{eq : moore_fc}\ ] ] this expression diverges logarithmically as @xmath47 goes to zero , reflecting the fact that the flux from a pure @xmath48 cusp is infinite . in practice , annihilation would reduce the density at the centre of the halo to a finite value even for pure cdm , while in real halos baryons will also affect halo structure on small scales .
furthermore , the simulations that indicate central cusps in halos can only probe down to @xmath49 , or @xmath50 ( power et al .
2002 ) ; below this scale , it is possible that the slope of the profile continues to change ( taylor & navarro 2001 ) .
thus in what follows we will consider moore profiles truncated at various inner cutoff radii , either @xmath51 ( profile m1 hereafter ) , @xmath52 ( profile m2 hereafter ) , @xmath53 ( profile m3 hereafter ) , or @xmath54 ( profile m4 hereafter ) , with cores of constant density within the cutoff radius .
( adding a core modifies equations ( [ eq : nfw_fc ] ) and ( [ eq : moore_fc ] ) slightly , but the difference is negligible for @xmath55 . )
the first cutoff corresponds the scale on which annihilation alone might truncate a pure power - law density profile ( clcneo - roldan & moore 2000 ) and the third cutoff corresponds to the largest core consistent with the results of current high - resolution simulations , while the fourth cutoff corresponds to the limit of the region within which baryons might plausibly have flattened out the core in the halos of massive galaxies . as mentioned above , the simulations that provide evidence for a universal density profile for cdm halos can only reliably resolve two decades of radius below the virial radius . while the logarithmic slope appears to converge to a value between @xmath56 and @xmath57 in the inner regions , it is clearly not a simple power - law in the outer regions , where it flattens substantially over a decade in radius .
thus , there remains the possibility that the inner slope of cdm halos , were it resolved to smaller radii , would continue to change in value . a shallower inner slope would help to explain the rotation curves of dark - matter dominated dwarf galaxies ( e.g. blais - ouellette , amram , & carignan 2001 ; de blok & bosma 2002 ) , as well as the mass distribution of the milky way within the solar circle ( binney & evans 2001 ) .
one possible form for such a profile was proposed by taylor and navarro ( 2001 , tn hereafter ) . in the outer regions
it resembles the nfw or moore profiles , while in the inner regions ( below about @xmath58 ) , its logarithmic slope decreases slowly to @xmath59 .
an analytic fit to the inner slope is : @xmath60 where @xmath61 , with @xmath62 .
we will calculate @xmath63 , which is convergent and well - behaved for this profile , from a direct numerical integration , although it can be computed to reasonable accuracy using the fitting formula above . from equation ( [ powlaw ] ) , the limiting value for small @xmath41 is @xmath64 .
note that in its outer regions , the equilibrium density profile derived by tn goes to zero at a finite radius .
this assumes an equilibrium that will never be achieved in cosmological setting , however , where halos will continue to accrete material onto their outer parts . thus , we assume that beyond @xmath65 , the profile drops off as @xmath66 , as in the nfw and moore fits .
figure [ fig:1 ] shows the density profiles ( top panel ) and @xmath67 ( bottom panel ) for the six profiles discussed above , including four versions of the moore profile with different constant - density cores ( within radii indicated by the horizontal dashed lines ) . in order to compare the different profiles ,
they have been plotted in terms of @xmath68 , the radius relative to the point @xmath69 at which the circular velocity peaks , and normalised so that the density is the same at @xmath69 . in the top panel ,
we see that all the density profiles aside from m4 are in excellent agreement for @xmath70 , the region probed by current simulations .
existing simulations are roughly compatible with any of the fits shown here . the differences in the inferred central behaviour of halos are based on different interpretations of the density profile very close to the current numerical resolution limit .
the bottom panel shows the effect of this uncertainty on @xmath63 , where @xmath41 is measured relative to @xmath69 , as in the top panel .
we see that in each case , the flux multiplier increases with concentration , reflecting the contribution from the dense central core .
to first approximation , the dependence on @xmath41 is independent of profile , while the inner slope of the halo profile determines the overall normalisation of @xmath63 . for the moore profile
it is 225 times larger than for an nfw profile , while for the shallower tn profile it is comparable to a moore profile with a core between 1% and 10% of @xmath40 in size .
the solid vertical lines on the plot indicate typical values of @xmath41 for present - day galaxies and clusters . for galaxy halos ,
@xmath63 is typically 50 for an nfw profile , 200 for the tn profile or profile m2 , and 1200 for the most extreme moore profile , m1 .
for all profiles it scales with concentration as roughly @xmath71 in this range of @xmath41 , as indicated by the solid line in the corner of the figure .
this figure indicates the importance of the inner region in determining the total flux from annihilating dark matter .
if recent simulations are correct , it seems likely that this flux may be 24 times larger than that predicted for a pure nfw profile ; in the extreme case of a moore profile with an inner cusp limited only by self - annihilation , the total flux is 25 times the nfw value .
since the dependence on @xmath72 is similar for all profiles , in what follows we will assume the nfw form for @xmath63 , and allow that the flux may be a constant 225 times greater than this .
to calculate the background flux from annihilations within a large volume , we need to add the relative contributions from regions of different density within this volume . following the press - schechter approximation ( press & schechter 1974 )
, we can consider all the mass in a given ( physical ) volume @xmath11 to be contained in virialised halos of some ( possibly very small ) mass .
we make the further approximation that these halos are spherical , with a fixed virial overdensity @xmath38 relative to the critical density , and have universal density profiles with a concentration @xmath73 that depends only on their mass .
if the volume contains @xmath74 halos in the mass range @xmath19 to @xmath75 , then from equation ( [ fluxm1 ] ) , the total flux multiplier for the volume will be @xmath76 where @xmath13 is the mean density of dark matter within @xmath11 , @xmath77 is the average density of bound halos , and @xmath78 is the flux multiplier for halos of concentration @xmath79 . since @xmath80 , we can rewrite the dimensionless flux multiplier for large volumes as : @xmath81 where @xmath82 is the fraction of the universe in virialised halos of mass @xmath19 or larger .
@xmath83 can be estimated using the press - schechter formalism : @xmath84 with @xmath85 , where @xmath86 is the critical overdensity and @xmath87 describes the power spectrum of density fluctuations .
recent simulations ( e.g. jenkins et al .
1998 ) have suggested , however , that this mass function may overestimate the number of halos near @xmath88 , the characteristic mass for which @xmath89 ( where @xmath90 is the linear growth factor at redshift @xmath91 ) .
an alternative mass function , proposed by sheth and tormen ( 1999 ) and based on the ellipsoidal collapse model , is : @xmath92 with @xmath93 and @xmath94 .
since both these mass functions have a power - law behaviour at low masses , where concentrations and thus @xmath63 are larger , it is not clear the integral in equation ( [ locvol ] ) will converge as we include the contribution from smaller and smaller halos .
( although the total mass per unit volume must converge , @xmath5 within a given volume need not converge , as in the case of a pure @xmath48 density profile . ) in practice , we expect baryonic phenomena to complicate structure formation on small mass scales , and annihilation itself will also limit the contribution from very dense material .
thus , we will truncate the integral at some limiting mass , and consider the behaviour of @xmath5 as a function of this mass limit .
the concentration of a halo should reflect the density of the universe at the time when it assembled the material now in its central core .
there are several predictions of the concentration - mass relation ( navarro , frenk & white 1997 ; bullock et al .
2001 ; eke , navarro , & steinmetz 2001 ; wechsler et al .
2002 ) , based on this interpretation .
we will use the most recent analytic model , that of eke , navarro , & steinmetz ( 2001 ens hereafter ) , to calculate the dimensionless flux multiplier , though we expect similar results from the other models .
we use the public code supplied by the authors to calculate concentrations , and integrate the expressions above numerically . note that one point of uncertainty in all these models is what minimum concentration to assign to halos that have just formed .
the ens code , for instance , predicts concentrations of 1 or less for very massive or very high - redshift halos , but the least concentrated halos actually found in their simulations have @xmath95 23 . for recently formed objects
the universal profile may not be well established , so any uncertainty produced by varying the minimum concentration reflects the limitations of the analytic model .
the top panel of figure 2 shows the dimensionless flux multiplier contributed by halos over some limiting mass , at various epochs in a lcdm cosmology , plotted as a function of the mass limit .
the solid lines show the results for the ps mass function , while the dashed lines show the results for the st mass function .
we see that in either case , the total flux is more or less convergent as we include smaller and smaller masses ; at @xmath96 , it only varies by a factor of 2 , for instance , for a limiting mass anywhere between @xmath97 and @xmath98 .
the value is also reasonably independent of the mass function used ; it is at most 2 times larger for the ps mass function .
the uncertainties in halo concentrations are only important at high redshift , when many halos have recently formed .
we have actually plotted two sets of curves in figure 2 , one with a minimum concentration of 1 ( the lower set of lines at each redshift ) and one with a minimum concentration of 2 ( the upper set ) , but the difference between them is negligible for @xmath99 . finally , the total flux multiplier increases with time , reflecting the progressive growth of structure . of course the density of the universe is decreasing at the same time ,
so the flux relative to that at the present day is larger at high redshift ; this is shown in the bottom panel , where we have used the present - day density to normalise @xmath5 .
this increased flux will partly compensate for the decreased volume element at high redshift , as discussed below . in previous section
we have derived the flux multiplier for a local volume , large enough to contain a representative sample of halos , but small enough that evolutionary effects are negligible . in a similar fashion
we can extend the definition of @xmath5 to entire observable universe . consider the expression for @xmath5 integrated over volume elements @xmath100 at different scale factors @xmath101 , to some maximum @xmath102 @xmath103 where we have chosen to normalise by the present - day density of the universe , @xmath104 , @xmath105
is given by equation ( [ locvol ] ) , and the last term corrects for the time dilation at a given epoch . since @xmath106 ,
this is equivalent to the following integral over redshift : @xmath107 for the lcdm cosmology considered here .
one quantity of interest is the relative contribution to the flux per interval of time , @xmath108 , which is plotted as a function of redshift in figure [ fig:3 ] .
the four sets of lines indicate the contribution from mass functions truncated at the different minimum masses indicated .
the line styles indicate the results for ps and st mass functions .
we see that the relative contribution peaks at quite early redshifts , roughly the epoch when the smallest halos considered have formed , as indicated by the arrows . at higher redshift
, it drops off sharply , because there are very few halos over the mass limit considered , while at lower redshift it drops off more gradually , as small , dense systems merge to form larger systems of lower density .
the latter drop - off may be overestimated , however , since the cores of dense halos may not be disrupted by merging , but may survive instead as distinct substructure within larger systems .
we consider this effect in the next section .
in cosmologies dominated by cdm , large halos assemble hierarchically through successive mergers between smaller objects .
this merging process is inefficient , and leaves many undigested remnants , or subhalos , orbiting within a galaxy or cluster halo .
the properties of these objects have been studied extensively in several recent , high - resolution scdm and lcdm simulations ( ghigna et al . 1998 ; klypin et al . 1999 ;
moore et al . 1999 ; springel et al . 2001 ) , and can also be understood using analytic and semi - analytic models ( bullock , kravtsov , & weinberg 2000 ; taylor 2001 ; benson et al .
2002 ; somerville 2002 ) .
the latter have the advantage that the properties of the underlying halos can be specified directly , and are independent of resolution and numerical effects . in this section
we use a recently developed semi - analytic model of halo formation ( taylor & babul in preparation ; see also taylor 2001 ) to determine the flux multiplier for a realistic halo with substructure , at various different epochs . from equation [ fluxm2 ] ,
the relative contribution to the flux multiplier from an individual subhalo within a larger halo is proportional to @xmath109 , where the subscript @xmath110 indicates the subhalo and the subscript @xmath111 indicates the average for the background system in which it resides .
this implies that if subhalos are substantially denser or more concentrated than the main system , they can dominate @xmath5 , even though they only contribute to a small fraction of the total mass of the system .
this is particularly true since tidal stripping will remove the lowest - density material from substructure , increasing the relative contribution from the material which remains bound .
more precisely , if we decompose the density of cdm within a halo into two components , a smooth background following the basic profiles given in section 2 , and a set of subhalos which have been stripped to varying degrees , then @xmath112 the contribution to @xmath5 then consists of several terms @xmath113 where @xmath114 , @xmath115 and @xmath116 are the mass , concentration and volume of the main system , and @xmath117 , @xmath118 and @xmath119 are the corresponding properties of each subhalo . in deriving the final result
, we have assumed that the subhalos do not overlap , and that each one is small compared to the main system , so that @xmath120 is more or less constant within the volume covered by a given subhalo .
the function @xmath121 is the dimensionless flux multiplier defined in section 1 , modified to account for the effects of tidal heating and stripping , which may change the density profile of the subhalos as they orbit within the main system .
we derive an expression for this function in terms of the original profile and the amount of mass lost below .
the density profiles considered in section ( [ sec : simple ] ) characterise isolated halos ; interactions between halos or mergers with larger systems will modify these profiles through a combination of tidal heating and tidal mass loss .
numerical simulations show that tidal effects tend to strip mass off the outer regions of a halo preferentially , preserving the central density of the halo until it has lost a large fraction of its original mass .
hayashi et al . ( 2002 ) propose the following modification of the density profile to describe this process : @xmath122 where @xmath123 is the tidal radius to which the halo is stripped , and @xmath124 is the reduction in the central density . for a halo with an nfw profile and an original concentration of 10 , they find that these parameters are related to the fraction @xmath125 of the original mass still bound to the halo by @xmath126 using this modified density profile , we can calculate @xmath127 , the flux multiplier for tidally stripped halos .
this should be a function of @xmath128 , the original concentration of the halo , and @xmath129 , the concentration of the stripped system . in practice
, we shall consider @xmath130 since this was the profile modeled by hayashi et al .
( 2002 ) . in defining @xmath127
, we also need to choose an appropriate volume over which to integrate .
since the modified density profile in equation ( [ hayashi1 ] ) is not truncated completely at @xmath123 , we could continue to integrate over the original volume ; in this case , @xmath121 would rise sharply ( roughly as @xmath131 ) as the mass of the halo is confined to a smaller and smaller region of the original volume . since @xmath123 roughly defines the region in which the subhalo density profile dominates over the background , the approximation made in deriving equation ( [ subst3 ] ) will be more accurate if we define @xmath127 over the volume within @xmath123 . in any case , in this section we will show the results of the calculation for several possible choices of volume , to clarify the effect on @xmath127 .
figure [ fig:4 ] shows @xmath130 for the volume within @xmath132 ( dashed line ) , @xmath133 ( dotted line ) , and @xmath134 ( dot - dashed line ) , with @xmath135 shown for comparison ( solid line ) .
the truncated functions differ from the original one at @xmath136 because the density profile given by equation ( [ hayashi1 ] ) is different from the unstripped one even when @xmath137 .
we see however that if we consider the volume within @xmath123 ( dotted curve ) , @xmath138 to good approximation .
we will assume that this result remains true for halos with larger or smaller original concentrations , provided @xmath139 , and use this to calculate the contribution to the total flux from evolved subhalos .
using equations ( [ subst3 ] ) ( [ hayashi3 ] ) , we can calculate the total flux multiplier for composite systems with substructure .
to do so , we use a semi - analytic model of halo formation ( taylor & babul in preparation ; see also taylor 2001 ) to determine the number , masses , density profiles and positions of subhalos within a set of hierarchically formed halos .
we have followed halo substructure down to @xmath140 times the mass of the main system .
since the spectrum of substructure is fairly steep at low masses , the total flux multiplier may depend on this cutoff , so we will consider our results as a function mass limit , as in section ( [ sec : cosmo ] ) , and extrapolate downwards to estimate the total flux multiplier for smaller limiting masses .
the top panel of figure [ fig:5 ] shows the relative contribution to the flux multiplier contributed by substructure over a minimum mass limit @xmath141 , as a function of @xmath141 , in models using nfw profiles .
the contribution has been normalised to the value @xmath142 expected for a smooth nfw halo without substructure .
the solid curves are average results for a set of semi - analytic galaxy halos with a total mass of @xmath143 , in a lcdm cosmology . the dashed vertical line indicates the resolution limit of the merger trees used ; below this , the spectrum of subhalos becomes increasingly incomplete , particularly below @xmath144 . at high redshift ,
where the halos are forming through major mergers around this mass scale , the curves are steep so the flux estimate will depend sensitively on the mass limit . at low redshift ,
the effect of the mass limit is less important , as most of the flux comes from more massive subhalos . in either case
, however , the total flux appears to have a power - law dependence on @xmath145 at low masses , so it is straightforward to extrapolate our estimates to lower mass limits .
overall , we see that substructure makes an important contribution to total flux right up to the present day , when it exceeds the contribution from the background halo .
the bottom panel shows the redshift dependence of the contribution more clearly , for two different mass limits , @xmath146 and @xmath147 .
as expected , at high redshift the contribution from substructure increases somewhat faster as we lower @xmath145 , reflecting the steep slope of the high - redshift curves in the top panel , while at low redshift the increase is slower .
if we extrapolate the power - law dependence seen in the top panel to smaller values of @xmath145 such as @xmath148@xmath149 , we expect a net contribution from substructure of anywhere from 10 to 20 times the smooth background value for the nfw profile .
we can determine the effect of substructure on the gamma - ray background by correcting the relative contribution calculated in section ( [ ssec : univ ] ) for the additional contribution from substructure in shown in figure [ fig:5 ] ; this is indicated by the dashed lines in figure [ fig:6 ] , for halos with nfw profiles , and mass cutoffs of @xmath146 and @xmath147 .
we see that when substructure is taken into account , the flux multiplier still peaks just after the epoch when @xmath150 and declines at lower redshift , but both the peak value and the value at @xmath96 are substantially enhanced .
we note that the peak and subsequent decline may partly compensate for the cosmological factor @xmath151 and for intervening absorption , which reduce the high - redshift contribution to the gamma - ray background .
we will investigate the detailed line and continuum spectra predicted by this model in a subsequent paper .
we have calculated the amount by which structure in dark matter on subhalo , halo and cosmological scales amplifies the expected signal from neutralino annihilation .
the overall magnitude of the effect remains uncertain at several levels , some more important than others .
uncertainty in the fundamental density profile which characterises halos , or more specifically in the slope of their density profiles at small radii , can change the flux multiplier for simple halos by a factor of 2025 , assuming concentrations typical of galaxy halos ( see figure [ fig:1 ] ) .
there is evidence in recent simulations for some excess mass in the inner regions , relative to the nfw profile ( power et al .
2002 ) , so a reasonable estimate for the flux multiplier may be 34 times the nfw value used for most of the calculations in this paper , although studies of the milky way ( e.g. binney & evans 2001 ) suggest less dark matter in its central regions than these profiles would imply . there is no direct evidence for a pure moore profile extending to very small scales , although if this is the case the resulting flux from simple halos will be 2025 times the nfw value used here .
the mass function of dark matter halos , averaged over large volumes , is also uncertain , particularly at the low mass end .
the mass function determined from simulations is better fit by the form proposed by sheth and tormen ( st ) than by the traditional press and schechter ( ps ) mass function . using the st mass function to predict the background flux reduces its amplitude by about 1.5 over ps .
we assume that the scale invariance of dark matter substructure will be broken at very small masses , perhaps close to jeans mass at recombination , for instance . at a minimum
, there should be some maximum density for dark matter halos , or equivalently some maximum for @xmath87 at early times .
varying the lower mass limits over a reasonable range ( @xmath152@xmath153 ) changes the flux by a factor of 2 or so for simple halos , but the contribution from substructure may increase this difference to as much as a factor of 40 .
overall , these uncertainties combine to produce a total uncertainty of more than an order of magnitude in the flux multiplier . a conservative model , at the bottom of this range , consists of an nfw profile , with a st mass function and a mass limit of @xmath154 . for this model ,
the present - day flux multiplier is @xmath155 .
we estimate that a more likely model has a profile with more mass in its inner regions , and lower limit to the mass function of @xmath156 . for this model ,
the present - day flux multiplier is @xmath157 . for an extreme model , with a moore profile limited only by annihilation , and a mass limit close to the jeans mass at recombination
, the present - day flux multiplier will be @xmath158 .
it is worth noting that with these large flux multipliers , some neutralino candidates can already be ruled out .
the 86 gev neutralino considered in bergstrm , edsj , & ullio ( 2001 ) , for instance , produced almost 10% of the gamma - ray background observed by egret at 1 gev , assuming a flux multiplier of @xmath159 , while a 166 gev neutralino produced more than 50% of the flux observed at 10 gev .
both these candidates are thus possible in our most conservative model , while the higher - energy candidate is marginally excluded in our favoured model , and both are ruled out in our extreme model .
the authors would like to thank e. hayashi and his collaborators for making their results available prior to publication .
we also thank a. babul and j. edsj for useful discussions .
jet gratefully acknowledges funding from the leverhulme trust during the course of this work . | we examine the expected signal from annihilation events in realistic cold dark matter halos . if the wimp is a neutralino , with an annihilation cross - section predicted in minimal susy models for the lightest stable relic particle , the central cusps and dense substructure seen in simulated halos may produce a substantial flux of energetic gamma rays .
we derive expressions for the relative flux from such events in simple halos with various density profiles , and use these to calculate the relative flux produced within a large volume as a function of redshift .
this flux peaks when the first halos collapse , but then declines as small halos merge into larger systems of lower density .
simulations show that halos contain a substantial amount of dense substructure , left over from the incomplete disruption of smaller halos as they merge together .
we calculate the contribution to the flux due to this substructure , and show that it can increase the annihilation signal substantially .
overall , the present - day flux from annihilation events may be an order of magnitude larger than predicted by previous calculations .
we discuss the implications of these results for current and future gamma - ray experiments .
= = = = = = = = # 1 # 1 # 1 # 1 @mathgroup@group @mathgroup@normal@groupeurmn @mathgroup@bold@groupeurbn @mathgroup@group @mathgroup@normal@groupmsamn @mathgroup@bold@groupmsamn = `` 019 = ' ' 016 = `` 040 = ' ' 336 = " 33e = = = = = = = = # 1 # 1 # 1 # 1 = = = = = = = = elementary particles
dark matter galaxies : structure gamma rays : observations gamma rays : theory |
observations of the galactic centre revealed one or two discs of @xmath2myr old stars orbiting the central super - massive black hole ( smbh ) at a distance of @xmath30.1pc @xcite .
until now , a large number of studies have been carried out to investigate the dynamical behaviour of such stellar discs around smbhs .
@xcite have shown that a single cold disc in the absence of a perturbing potential can not explain the observed large inclinations and eccentricities . @xcite
have provided upper mass limits to the discs masses by studying the amount of warping of two discs at large inclination . in @xcite
we have shown that a significant number of disc stars eventually achieve highly eccentric orbits due to kozai interaction , thus providing a mechanism to create the observed s - stars and hyper - velocity stars by disruption of close - passage binaries .
all these simulations have been carried out neglecting the stellar cusp present in the vicinity of the smbh ( e.g. , * ? ? ?
@xcite argued that a smooth background potential hinders the creation of high eccentricities in full kozai cycles since it causes orbital precession . on the other hand ,
@xcite have shown that relaxation processes in a grainy stellar cusp surrounding the smbh act on time scales comparable with the age of the stellar discs ( see also * ? ? ?
they demonstrate that due to warping and relaxation , two highly inclined stellar discs can not be recognised as flat circular discs after 5myr of interaction .
significant contributions arise from both two - body relaxation due to stellar encounters customary in star clusters , and _ resonant _ relaxation effects , an enhanced rate of relaxation of angular momentum in near - keplerian systems @xcite .
furthermore , both disc and cusp stars contribute significantly to the disc evolution . considering the apparent importance of the influence of a stellar cusp on the disc evolution , in this paper we challenge the results of earlier studies by including both analytic and @xmath1-body cusp potentials in corresponding models . in section [ sec : gcdci : cuspcomp ] , we discuss the structure of the stellar cusp as predicted by theory and derived from observations .
section [ sec : gcdci : models ] describes the numerical method underlying our integrations and our models of the stellar discs .
we present our results with respect to the studies mentioned above in section [ sec : gcdci : results ] and discuss them in section [ sec : gcdci : discussion ] .
the discs of young stars orbiting the smbh in the centre of the milky way are embedded in a population of old stars and stellar remnants .
theoretical arguments and @xmath1-body simulations show that a stellar system around an smbh evolves into a cusp with a @xmath4 power - law density distribution @xcite .
the density profile of the stellar cusp in the galactic centre is estimated from the observed luminosity profile and kinematics by @xcite as @xmath5 where @xmath6 is the distance from the smbh , @xmath7 inside 0.22pc and @xmath8 outside 0.22pc .
hence , the stellar cusp has a mass within 1.5pc comparable to the black hole mass .
we use this result ( omitting the uncertainty specified ) when modelling the cusp as an analytic potential .
however , as we will discuss with the results in section [ sec : gcdci : results ] , relaxation between stars can play an important role in the highly dense central parsec of our galaxy .
hence it is not sufficient to include such a smooth potential : to perform realistic simulations of disc dynamics in the galactic centre , it is required to consider stars and stellar remnants in the @xmath1-body calculations . a model of the stellar population in the central 100pc of our galaxy
is given by @xcite .
due to dynamical friction causing massive stars to sink towards the centre , it can be assumed that the central few 0.1pc , where we also find the young stellar discs , are mass - dominated by the most massive components , namely stellar - mass black holes @xcite . in this paper
we will analyse the dynamics of the stellar discs , which will be described in more detail in the next section .
this restricts our interest to the potential within 1pc from the smbh . at 2pc from the centre
@xcite found a rotating ring of gas of @xmath9 , the so - called circum - nuclear disc ( cnd ) . @xcite
have shown that the torque exerted by the cnd may cause ( additional ) warping of a stellar disc at @xmath10pc from the smbh .
furthermore , it may cause somewhat higher eccentricities due to kozai interaction ( cf . section [ sec : gcdci : sstars ] ) . throughout this paper
, we will ignore the effect of the cnd .
this leaves the stellar cusp as the only dynamical perturbation to the disc evolution .
an exhaustive study of the dynamics of stars in the galactic centre will have to include the effects of the cnd .
however , the general conclusions about the influence of a stellar cusp on the disc dynamics as discussed in this paper will remain unaffected .
to test the mutual interaction of two stellar discs orbiting the central smbh , we performed direct @xmath1-body integrations using our bhint code @xcite . bhint has been developed specifically to calculate the dynamics of stars orbiting around an smbh , avoiding the large secular error other @xmath1-body codes based on the hermite scheme @xcite accumulate in case of the highly eccentric orbits and extreme mass ratios present in the galactic centre .
bhint integrates the equation of motion of a star orbiting an smbh by moving it along the keplerian orbit and adding the forces from other components ( disc stars and cusp black holes ) as perturbations .
it solves kepler s equation analytically ( and thus without secular error ) to determine the orbital motion around the dominating smbh , which is assumed to rest at the origin of the coordinate system , while using the hermite scheme to integrate the perturbing forces .
bhint includes special treatment of close encounters and fast approaches to allow for high precision .
in addition , it comprises post - newtonian ( pn ) approximation of the effects of general relativity up to order 2.5 following @xcite , which is added to the perturbations .
however , since the pn terms need relatively high accuracy and thus a significant amount of computational time , we usually switch on these terms only in integrations of highly eccentric orbits , where pericentre - shift becomes important ( see table [ tab : params ] ) .
for example , @xcite have shown that in the regime of the stellar discs , pericentre shift due to the spherical cusp is at least one order of magnitude more effective than the relativistic one , as long as the eccentricity is below @xmath11 .
[ 1.5ex]model & [ 1.5ex]@xmath12 & [ 1.5ex]@xmath13 & [ 1.5ex]@xmath14 & [ 1.5ex]@xmath15 & [ 1.5ex]@xmath16 & [ 1.5ex]@xmath17 & [ 1.5ex]@xmath18 & [ 1.5ex]@xmath19 & [ 1.5ex]@xmath20 & [ 1.5ex]@xmath21 & [ 1.5ex]@xmath22 & cusp & [ 1.5ex]pn + ` 1a ` & 15000 & 0.050.5 & & & & 1120 & 1.35 & 30% & 5myr & 15 & 0.22 & & + ` 1b ` & 15000 & 0.050.5 & & & & 1120 & 1.35 & 30% & 5myr & & & yes &
+ ` 1c ` & 15000 & 0.050.5 & & & & 1120 & 1.35 & 30% & 5myr & & & & + ` 2a ` & 10000 & 0.050.5 & 5000 & 0.070.5 & 88@xmath23 & 1120 & 1.35 & 30% & 5myr & 15 & 0.22 & & + ` 2b ` & 10000 & 0.050.5 & 5000 & 0.070.5 & 88@xmath23 & 1120 & 1.35 & 30% & 5myr & & & yes & + ` 2c ` & 10000 & 0.050.5 & 5000 & 0.070.5 & 88@xmath23 & 1120 & 1.35 & 30% & 5myr & & & & + ` 3a ` & 18000 & 0.050.5 & 9000 & 0.070.5 & 88@xmath23 & 1120 & 2.3 & 40% & 5myr & 15 & 0.22 & & + ` 3b ` & 13000 & 0.050.5 & 8500 & 0.070.5 & 105@xmath23 & 1120 & 2.3 & 13% & 5myr & 15 & 0.22 & & + ` 3c ` & 11000 & 0.050.5 & 6500 & 0.070.5 & 88@xmath23 & 1120 & 2.3 & 13% & 4myr & 15 & 0.22 & & + ` 3d ` & 24000 & 0.050.5 & 12000 & 0.070.5 & 88@xmath23 & 0.01120 & canonical & 30% & 4myr & & & & + ` 4a ` & 20000 & 0.050.5 & 10000 & 0.070.5 & 125@xmath23 & 1120 & 1.35 & 30% & 6myr & 15 & 0.22 &
& if @xmath24 + ` 4b ` & 15000 & 0.050.5 & 7500 & 0.070.5 & 88@xmath23 & 1120 & 1.35 & 40% & 6myr & 15 & 0.22 & & + ` 4c ` & 20000 & 0.030.4 & 8000 & 0.050.4 & 100@xmath23 & 1120 & 2.3 & 40% & 6myr & 15 & 0.22 & & + ` 4d ` & 15000 & 0.050.5 & 7500 & 0.070.5 & 88@xmath23 & 1120 & 1.35 & 40% & 6myr & 40 & 0.22 & & yes + unless stated otherwise , our model parameters are as follows : we start with two cold stellar discs around a @xmath25 smbh at high relative inclination @xmath16 .
the discs have finite radial extents @xmath26 , a surface density @xmath27 , and a thickness ( standard deviation ) of 1.44@xmath23 .
the stellar orbits have a mean eccentricity of 0.03 .
we vary the initial values for disc masses , inclination , and radial extent within ranges compatible with the observations ( see table [ tab : params ] ) .
the underlying imf in most of our models has a slope of @xmath28 , which was derived by @xcite for the stellar discs in the galactic centre .
this is considerably flatter than the canonical initial mass function ( imf ) according to @xcite , @xmath29 , with @xmath30 @xmath31 , @xmath32 @xmath33 , and @xmath34 @xmath35 .
the number of stars in the mass interval @xmath17 to @xmath36 is @xmath37 , and the @xcite imf has @xmath38 .
since the disc mass for an imf as flat as @xmath28 is not very sensitive to the lower stellar mass limit , we use a mass range of @xmath39 . for most of our models ,
we assume a disc age of 5myr @xcite .
the discs can not be much older : according to standard stellar evolution models ( e.g. * ? ? ?
* ) , an age of 6myr implies a maximum zero - age main sequence ( zams ) mass of 32 for observed stars . the observed number of disc stars with minimum zams masses of 20 @xcite would thus imply a total initial disc mass of 16750 for the imf with @xmath40 ( 77 stars with @xmath41 ) , plus some binary fraction , and the discs are believed to have a total mass @xmath42 @xcite .
however , we vary the disc masses and ages consistently in sections [ sec : gcdci : imf ] and [ sec : gcdci : sstars ] . to test for disruption of binaries , we mark a certain fraction @xmath19 of stars in our models as binaries ( varying between the different models , see table [ tab : params ] ) , and double their mass ( assuming equal - mass binaries ) .
we track close encounters during our simulations to determine possible points of disruption .
disruption of a binary of mass @xmath43 and separation @xmath44 occurs if it passes another particle ( smbh , disc or cusp star ) of mass @xmath45 at a distance less than @xmath46{m / m_{\rm bin } } \label{eq : r - tid - bin}$ ] @xcite .
we accounted for the influence of a background cusp as discussed in section [ sec : gcdci : cuspcomp ] either by using an analytic potential , or by adding 14000 stellar black holes ( sbhs ) of 15 into the inner 0.22pc , consistent with the density profile derived by @xcite .
table [ tab : params ] lists the varying parameters of the models used throughout this paper . on 3ghz desktop computers equipped with grape6-pro/8 or microgrape boards , calculation of these models took from five hours ( top - heavy imf , no sbh cusp ) to a few 100 hours ( all other models ) per model .
@xcite have shown that a single cold disc in the absence of a perturbing potential can not explain the observed large inclinations and eccentricities .
the required perturbing potential may result from a massive circum - nuclear disc ( @xcite ; cf .
@xcite ) , an intermediate - mass black hole and/or star cluster ( probably irs 13 ; * ? ? ?
* ) , or a second disc ( @xcite ; cf .
@xcite ) . to test whether a spherical cusp alone significantly alters the evolution of a disc
, we integrated models of a single disc with the @xmath28 imf suggested by @xcite , restricted to @xmath39 .
for respective masses lead to similar results . for a discussion on the choice of underlying imf see section [ sec : gcdci : imf ] . ]
figure [ fig:1disc - e - i ] shows the time evolution of eccentricity and inclination of a single disc in a stellar cusp ( ` 1a ` ) . in line with the results of @xcite
, we find a faster increase of eccentricities and inclinations than in the absence of a stellar cusp ( ` 1c ` ) .
this still holds if we consider only the massive stars .
however , less than 20 per cent of the massive stars ( above @xmath47 ) have final eccentricities @xmath48 , and a mere two per cent of the massive stars reach an inclination larger than 45@xmath23 , while half of the disc stars observed have eccentricities @xmath48 , and a large number of stars have much higher inclinations to the observed clockwise disc @xcite .
in fact , the largest contribution to the high values of observed eccentricities and inclinations stems from the counter - clockwise stars .
we conclude that the stellar cusp does have a significant effect on the evolution of eccentricities and especially orbital inclinations , but agree with @xcite in that it can not account for the highly eccentric and counter - clockwise stars assuming an initial configuration of a single cold disc in the absence of any perturbing potential as discussed above .
@xcite have shown how the mutual interaction of two discs with large relative inclination leads to an increase of disc thickness .
the evolution of orbital eccentricities and inclination ( i.e. disc warping ) in the presence of a cusp has been studied in detail by @xcite .
they find that while the evolution of orbital inclinations is dominated by the discs potential ( precession and warping ) , the change in eccentricity and semi - major axis of the disc stars is driven by non - resonant as well as resonant relaxation with disc and cusp stars . here
we analyse the effect of the stellar cusp on the discs by computing the discs modelled by @xcite a ) in a cusp of 15 sbhs , b ) in an analytic cusp , and c ) in the absence of any cusp potential . ; the open ( red ) symbols show the mean angular momentum vector of the initial discs .
the direction and amount of warping does not depend significantly on the presence of a cusp .
[ fig : warp],width=313 ] figure [ fig : warp ] shows the orientation of the orbital planes of the disc stars after an integration time of 5myr . since precession is dominated by the disc potentials , direction and amount of warping do not depend significantly on the cusp . on the other hand , the distribution of eccentricities does depend on the surrounding cusp : figure [ fig : eacomp ] shows the final distribution of the disc stars eccentricities as a function of semi - major axis , for both single - disc and two - disc setup . while the overall eccentricity distribution is similar in all cases , eccentricities @xmath49
are only achieved in the presence of a stellar cusp , or in a two disc setup without any cusp potential ( at the very inner edge ) . as we will discuss in section
[ sec : gcdci : sstars ] , while high eccentricities reached by long - term kozai interaction with the two discs are impeded by an analytic cusp , relaxation processes in a stellar cusp ( cf .
* ) can lead to eccentricities high enough for short - term resonances to be effective despite the spherical potential , thus explaining the high eccentricities in three of our six cases ( two discs in the absence of a cusp , or one or two discs interacting with cusp remnants ) . as a direct result of the disc warping described in the previous section , @xcite have provided upper mass limits to the discs masses considering their apparent flatness as measured by @xcite after a few myr of interaction .
however , @xcite did not find convincing evidence for a counter - clockwise disc as flat as 19@xmath23 , which can also be interpreted as an indication for a higher mass of the massive disc in a two - disc model .
@xcite have extended the argument of warping as a mere measure of disc thickening by measuring the warping angle as a function of radial distance and found good agreement of a two - disc model with the observations , showing how the warping angle can reach 50@xmath23 already at 0.15pc from the smbh .
here we calculated different models with two discs of stars following a canonical imf from @xmath39 .
using binary fractions of 40 and 13 per cent implies total masses of 27000 ( ` 3a ` ) and 21500 ( ` 3b ` ) , respectively , to have the same number of massive stars .
figure [ fig : imf1 ] shows the orientation of orbital planes of model ` 3a ` after 5myr of integration , which shows the same trend as the results of the previous section ; the same holds for model ` 3b ` . assuming a disc age of 4myr ( the lower limit as estimated by * ? ? ?
* ) reduces the initial disc mass by 20 per cent ( ` 3c ` ) , since a larger fraction of massive stars survives the shorter disc age , and thus a smaller initial number of massive stars is needed to explain the observed 77 stars . in another model ( ` 3d ` ) , we choose the initial disc mass to be consistent with the canonical imf @xcite down to 0.01 , resulting in 80 per cent more mass than with a cut - off at 1 . to keep the total number of stars low
, we had to replace the low - mass stars ( @xmath50 ) by 1 stars of equal total mass .
again , the overall spiral trend of disc warping as plotted in fig .
[ fig : imf1 ] is similar to the earlier results .
figure [ fig : imf2 ] shows the angle between a disc fitted to massive stars at a certain radius and the outermost fitted disc . while the amount of warping of all clockwise ( massive ) discs modelled with a canonical imf is compatible with the observed values ( or even below ; cf .
discussion in * ? ? ?
* ) , the integration results of the counter - clockwise disc fit the observations well only outside @xmath51pc from the smbh .
inside this distance , the counter - clockwise disc dissolves , and thus the warping angle is no longer a smooth function of central distance .
however , it is essential to note that the observed stars are assigned to one of the two discs depending on whether they are moving clockwise or counter - clockwise in the sky , since there is no other way to tell their disc of origin .
this implies that none of the squares on the northern hemisphere of fig .
[ fig : imf1 ] would be considered for determining the warping angles of the counter - clockwise disc .
furthermore , more stars observed within 0.16pc from sgr a@xmath52 move on projected radial orbits ( and are thus completely dismissed ) than on projected counter - clockwise orbits . if these ( and even some of the clockwise - moving stars ) stem from the more strongly warped counter - clockwise disc , they might significantly increase the amount of warping of the observed counter - clockwise disc .
we waive a further analysis since assignment of stars to one of the two discs is rather uncertain in the inner 0.16pc with the currently available observational data .
in addition , the choice of the line of sight in our models is arbitrary . .
[ fig : imf1],width=313 ] -thickness of the fitted planes , respectively .
the corresponding error bars for the simulations are generally smaller and thus omitted for clarity ( cf . * ? ? ?
* ) . in none of our models
does the amount of warping of the clockwise disc exceed the observed value .
the modelled counter - clockwise discs dissolve towards the inner part , where in most cases warping is no longer a smooth function of central distance . [
fig : imf2],width=313 ] considering this and the fact that there is as yet no model explaining all of the observed features of the young stars and including all perturbations like e.g. the cnd , we conclude that these models are still consistent with the observations .
this seems to be in contrast to the results of @xcite , however they used stellar discs only out to 0.2pc from the smbh , while more than one third of the stars listed by @xcite are further away from sgr a@xmath52 .
in addition , they did not distinguish between stars in the outer parts of the discs which almost retain their orbital planes , and stars close to the centre which strongly precess .
our results show that a canonical imf over the range @xmath53 can not be excluded dynamically for the young discs in the galactic centre .
@xcite find a flat imf from analysing the @xmath54-band luminosity function of the disc stars ; however , they assume a lower mass limit of 20 for the observed stars , and their estimate of the disc age of @xmath2myr limits the initial masses of the now observed stars to @xmath55 , so this slope refers to a fairly small mass interval only .
@xcite find a best fit of observations in the central parsec of our galaxy with a model of continuous star formation with a top - heavy imf .
furthermore , @xcite argue that the x - ray luminosity of the sgr a@xmath52 field is one order of magnitude too low to account for the number of young @xmath56 stars expected from a canonical imf , considering the large number of o - stars observed in the discs .
this can also be explained by a higher low - mass cutoff of the imf at of the order of 1 .
@xcite find that eccentricities as high as @xmath57 can only be explained by a top - heavy imf when assuming a circular disc origin , however they neglect the influence of the cusp and the second disc as elaborated by @xcite ; both suggest an eccentric disc origin for the counter - clockwise system
. simulated star formation of gaseous discs suggests that stars form with a top - heavy imf only if the cooling times are just as large as allowed to still induce disc fragmentation ( @xcite and references therein ) .
unfortunately , theoretical imf predictions have failed in the past to correctly describe the observations near the galactic centre @xcite .
another argument for a top - heavy imf arises from the relation between disc and cusp mass . assuming that we do not live in a special time ,
the formation of stellar discs in the galactic centre must be a frequent event .
to explain the observed number of @xmath58 disc stars at an assumed age of 5myr with the @xmath28 imf suggested by @xcite , a total initial disc mass of @xmath59 is required , of which a mass of @xmath60 ends up in stellar remnants . using the canonical imf requires a total initial disc mass of @xmath61 , of which @xmath62 in remnants and low - mass stars are retained over a hubble time . assuming that most stars survive in the central region ( or are replaced by inspiralling massive remnants ) , the observed cusp mass of @xmath63 in the central 0.5pc @xcite allows for the formation of discs like those observed every 50myr for a top - heavy imf , and only every 500myr for a canonical imf .
the fact that we observe very young stellar discs thus seems much more unlikely if star formation in the galactic centre follows the canonical imf .
however , it may also be an indication of recently enhanced star formation processes .
for example , if the galactic bar is young , as an increased fraction of barred galaxies for lower redshifts suggests @xcite , bar - induced gas inflow may explain such an enhancement ( @xcite and references therein ) .
we conclude that while theoretical star formation arguments and some observations indicate a top - heavy imf in the centre of the milky way , our self - consistent models of stellar dynamics allow the existence of a canonical imf as well .
the observational evidence clearly needs to be improved by star counts , while the theory of star formation is currently not able to make conclusive statements on the imf of stars in the galactic centre ( see @xcite for a discussion of recent failures of imf theory ) .
in @xcite we have shown that a significant number of disc stars eventually achieve highly eccentric orbits due to kozai interaction , thus providing a natural mechanism to create the observed s - stars and hyper - velocity stars by disruption of close - passage binaries .
however , @xcite remarked that the effects of pericentre shift due to relativistic precession and a spherical cusp can destroy the kozai mechanism .
while the relativistic precession period decreases strongly with increasing orbital eccentricity , @xmath64 ( where @xmath65 , @xmath44 and @xmath66 are the orbital period , semi - major axis , and eccentricity , respectively ) , they showed that cusp precession at a given distance is most effective for near - circular orbits : @xmath67 ( both equations assuming @xmath68 and the cusp profile inside 0.22pc by @xcite ) . from this
we conclude that the very high eccentricities in kozai resonance are prevented by relativistic precession at high eccentricities and by cusp precession at low eccentricities . figure [ fig : kozai ] shows the eccentricity evolution during kozai resonance for integrations of a three - body problem with and without a cusp , and with and without pn treatment for relativistic effects .
it is clearly seen that the pn terms prevent higher eccentricities when a certain eccentricity threshold is reached ( in this case @xmath69 ) , while their effect on the rest of the kozai cycle is not dramatic .
this can be understood by general relativity being most effective for high eccentricities ( and thus small pericentre distances ) . on the other hand ,
the changes on both maximum eccentricity and kozai period exerted by a stellar cusp are quite low , until a certain mass threshold of the cusp is reached , which prevents not only the highest eccentricities but essentially any significant eccentricity increase and thus the complete kozai cycle . even for the innermost disc stars ( @xmath70pc )
, we find this threshold to be at only a few per cent of the cusp mass given by @xcite : while the 1 per cent curve reaches @xmath71 , the 5 per cent curve does not exceed @xmath72 .
smbh with semi - major axes of 0.04pc and 0.16pc , respectively , resulting in the star s orbital period of 400yr .
we calculate models without cusp ( solid black line ) , one per cent ( dashed line ) , and five per cent ( dotted line ) of the extended cusp mass observed in the galactic centre ( modelled as a smooth potential ) .
each curve is accompanied by a corresponding thick green curve which represents a respective integration including post - newtonian terms up to 2.5 pn to account for the effects of general relativity .
while relativistic effects damp the kozai effect at high eccentricities , the presence of a stellar cusp leaves it almost unchanged until a mass threshold of a few per cent of the observed value , where eccentricity growth is stopped at very small values so that the orbit remains almost circular ( straight dotted lines ) .
[ fig : kozai],width=313 ] as a result of this analysis , we can not expect any disc star in the galactic centre to undergo a full kozai cycle , which was also shown by @xcite .
however , if a star achieves high eccentricity by ( non-)resonant relaxation or close encounters @xcite , it may find itself in a point of kozai resonance where the time until maximum eccentricity is significantly shorter than the full kozai period , and also the cusp precession period is longer than for the circular case .
furthermore , in our system of two disc potentials and a large reservoir of perturbing disc and cusp stars , other forms of resonance may occur ( short - term kozai / resonant relaxation with changing sources of perturbation ) .
going beyond @xcite , we computed different models of two - disc systems in the presence of a stellar cusp . as a result of cusp precession
, we find only @xmath0 binaries in each model achieving eccentricities high enough for binary disruption to occur ( ` 4a ` , ` 4b ` ) .
when using a canonical imf down to 1 , 7 s - stars are created ( ` 4c ` ) . besides the smaller disc extent used for model ` 4c ` ( resulting in faster relaxation and shorter kozai periods ) , this is due to the number of b - stars being four times higher than in the previous cases ( and also the total mass increase ) , while the number of o - stars is still consistent with the observed value .
due to the increase of total disc mass , the number of s - stars created may increase even further when using a canonical imf down to 0.01 . for those stars achieving high eccentricities we can see the behaviour as predicted above : the eccentricity changes in a random fashion by relaxation processes until a certain value of high eccentricity is achieved . at this point
, the star ( or binary ) goes into kozai resonance for a time scale long enough to achieve extreme eccentricity , but too short to be prevented by precession effects .
figure [ fig : eccevol ] shows the evolution of eccentricity as well as one component of the normalised angular momentum vector , @xmath73 , of selected stars of model ` 4b ` achieving high eccentricity . while the eccentricity evolution seems at first sight compatible with the secular instability in flat eccentric discs as discovered by @xcite , the rapid change of the stars orbital planes as indicated by @xmath73 is revealing of kozai resonance . the former secular instability can only occur in a flat ( single , non - warping and non - thickening ) , eccentric disc in a cusp with a density profile @xmath74 .
component of the normalised angular momentum vector , @xmath73 ( bottom ) , of three selected stars of model ` 4b ` .
the cyclic nature of eccentricity evolution as well as the rapid change of the stars orbital planes as indicated by @xmath73 are revealing of the kozai mechanism .
[ fig : eccevol],width=313 ] interestingly , we find the orbital planes eventually flipping at very high eccentricities ( e.g. solid curve at maximum eccentricity , @xmath75myr : @xmath73 suddenly changes from @xmath76 to @xmath77 ) .
this was first discovered by @xcite who applied their softened version of gauss method to the two - disc problem .
however , in our case no softening is involved . at such high eccentricities , no large torque is needed to change the small angular momentum to the opposite direction ( note that this does not occur in an isolated three - body system undergoing kozai resonance ) . in another calculation ( ` 4d ` ; using the top - heavy imf with @xmath28 ) , we used the pn terms for all disc stars from the beginning , and a spherical cusp composed of @xmath78 sbhs , which resulted in the creation of 8 b - type and 1 o - type s - star .
almost all of the originating binaries had close encounters with one of the sbhs ( although not close enough to be disrupted ) , however none of these encounters lead to a sudden change in orbital eccentricity or semi - major axis which could explain the high eccentricity it achieved much later in the integration .
we ascribe the higher number of s - stars created to the fact that the effectiveness of resonant relaxation is proportional to the sbh mass @xcite .
apparently , pericentre shift due to relativistic effects does not prevent such high eccentricities .
we conclude that if the stellar cusp in the galactic centre is mass - dominated by massive sbhs ( as predicted by @xcite and @xcite ) , at least some of the s - stars observed could have originated from the young stellar discs . in a setup of an initially eccentric disc ( as suggested by @xcite for the counter - clockwise disc ) , we can expect a somewhat higher probability of high - eccentricity orbits required for the formation of s - stars . since the age of the s - stars is less well - determined than that of the stellar discs , some of them may stem from an earlier generation of discs , which would anyway be needed to explain the longer travel times of the hyper - velocity stars observed in the milky way halo ( @xcite and references therein ) , assuming the same disruption scenario as their origin @xcite .
furthermore , scattering of young binaries by massive perturbers may contribute to the observed number of s - stars @xcite .
in this paper , we have investigated the influence of the stellar cusp on the evolution of young stellar discs in the galactic centre .
our results show that the impact of stars and stellar remnants on the disc stars orbits must not be neglected in theoretical models and numerical simulations : the stellar cusp enhances evolution of orbital eccentricities and disc thickness by ( non-)resonant relaxation of angular momentum as described by @xcite .
we find that this is especially true for high eccentricities , which are prevented by a smooth cusp potential resulting in orbital precession , but are achieved in a grainy cusp as we expect to find it in the inner parsec of our galaxy .
a combination of two - body relaxation processes resulting in a random increase of eccentricity and short - term kozai resonances leads to eccentricities high enough for binary disruption , leaving behind an s - star on a tight orbit and ejecting a hyper - velocity star .
this process is not as efficient as it would be in the absence of a stellar cusp , but a few generations of stellar discs could account for the observed s - stars .
we conclude that the creation of s - stars is a natural consequence of the interaction of two stellar discs in a grainy cusp around an smbh , where stellar orbits are exposed to the effects of general relativity .
scattering of stars onto highly eccentric orbits is enhanced by a cusp composed of more massive particles , such as sbh multiples , ( remnants of ) star clusters or imbhs .
thus , more realistic modelling of stellar dynamics in the galactic centre requires a better understanding of the composition of the stellar cusp .
even more so , the composition of the stellar discs remains unclear .
the mere existence of the s - stars may be regarded as evidence for a steeper ( i.e. more standard - like ) imf , as this would significantly increase the number of b - stars ( and thus stars eventually reaching tight orbits ) given the observed number of o - stars .
a detailed survey of b - type stars in the central parsec of the milky way will not only help us to better understand the composition of the young stellar discs and thus the formation of the s - stars , but also increase the significance of the counter - clockwise disc as well as the observed disc warping .
however , if the formation of stellar discs in the galactic centre is a common scenario , most of the b - type stars ( including the known s - stars ) may stem from previous generations of discs , and their orbits are expected to be randomised by the influence of the cusp as well as the potential of current and previous discs . as a by - product of our models
, we can determine the merger rates of stellar black holes with the smbh .
these mergers can be caused by an sbh plunging into the smbh , or by gradual inspiral due to gravitational wave emission ( so - called extreme mass ratio inspirals , emris ) , yielding a promising scenario for the detection of gravitational wave signals by the planned laser interferometer space antenna ( e.g. * ? ? ?
since in our models we focus on the evolution of the stellar discs , we do not differentiate between emris and plunges of sbhs , but only identify the merger of an sbh with the smbh as the point where their distance is smaller than three schwarzschild radii .
our models including a cusp of 14000 sbhs of mass 15 showed merger rates from @xmath79myr@xmath80 , with a mean of 58myr@xmath80 .
note that these rates may be higher when using sbhs beyond 0.22pc or of smaller masses ( and thus a higher number ) . on the other hand ,
they need to be corrected for cusp depletion and loss - cone - refilling .
we would like to thank pau amaro - seoane for useful comments .
this work was supported by the german research foundation ( dfg ) through the priority program 1177 ` witnesses of cosmic history : formation and evolution of black holes , galaxies and their environment ' . | observations of the galactic centre show evidence of one or two disc - like structures of very young stars orbiting the central super - massive black hole within a distance of a few 0.1 pc .
a number of analyses have been carried out to investigate the dynamical behaviour and consequences of these discs , including disc thickness and eccentricity growth as well as mutual interaction and warping .
however , most of these studies have neglected the influence of the stellar cusp surrounding the black hole , which is believed to be @xmath0 orders of magnitude more massive than the disc(s ) . by means of @xmath1-body integrations using our bhint code
, we study the impact of stellar cusps of different compositions .
we find that although the presence of a cusp does have an important effect on the evolution of an otherwise isolated flat disc , its influence on the evolution of disc thickness and warping is rather mild in a two - disc configuration .
however , we show that the creation of highly eccentric orbits strongly depends on the graininess of the cusp ( i.e. the mean and maximum stellar masses ) : while @xcite recently found that full cycles of kozai resonance are prevented by the presence of an analytic cusp , we show that relaxation processes play an important role in such highly dense regions and support short - term resonances .
we thus find that young disc stars on initially circular orbits can achieve high eccentricities by resonant effects also in the presence of a cusp of stellar remnants , yielding a mechanism to create s - stars and hyper - velocity stars .
furthermore , we discuss the underlying initial mass function ( imf ) of the young stellar discs and find no definite evidence for a non - canonical imf .
[ firstpage ] black hole physics stellar dynamics
methods : @xmath1-body simulations galaxy : centre . |
dalitz plot analysis of three - body decays of neutral @xmath1 mesons is a useful tool in various measurements where coherent admixtures of @xmath5and @xmath6are observed .
this technique was initially proposed for the measurement of the unitarity triangle angle @xmath0 in @xmath2decays @xcite .
later it was applied to the measurement of charm mixing @xcite and to the resolution of the quadratic ambiguity in the measurement of the angle @xmath7 using a time - dependent analysis of the decay @xmath8
most of these measurements are based on the @xmath9decay which offers the best precision among three - body @xmath5 decays .
the technique is model - dependent it depends on the complex amplitude of the @xmath5decay which is obtained from the @xmath10sample using model assumptions .
the result of the measurement contains therefore model uncertainties . in the case of @xmath0 measurement , this uncertainty ( @xmath11 ) is already comparable to the statistical accuracy @xcite . however , a modification of the analysis is possible that allows to perform a completely model - independent measurement @xcite .
it requires the phase space of the three - body @xmath1 decay to be divided into bins .
information about the complex phase in each bin can be extracted from the quantum - correlated @xmath5 decays from @xmath12 process .
the measurement of the strong phase in bins of the @xmath9phase space was recently performed by the cleo collaboration @xcite .
this measurement should allow to reduce the error of @xmath0 related to the uncertainty in the @xmath9amplitude to @xmath13 .
recently , charm mixing was observed by the belle and babar experiments @xcite . with degree - level precision ,
the effect of charm mixing can become significant in the measurement of @xmath0 .
it was shown that mixing contributes only of second order in the @xmath14 and @xmath15 parameters to the ads , glw and model - dependent dalitz plot analysis methods , and thus can be safely neglected @xcite .
but the effect of mixing on the binned analysis with the phase terms extracted from quantum - correlated @xmath5 decays is of separate interest . in this paper , we investigate the impact of the charm mixing on the model - independent @xmath0 measurement , and show that ignoring the mixing at all stages of the analysis is safe up to a sub - degree level of precision .
we also find that in the coherent production of @xmath3 system in @xmath4 collisions , the effect of charm mixing is enhanced compared to the case of @xmath16 production , and propose a model - independent method to measure charm mixing parameters in time - integrated dalitz analysis at charm factories .
the method is sensitive to both mixing parameters , @xmath14 and @xmath15 , as well as to @xmath17violation parameters @xmath18 and @xmath19 .
the sensitivity of the proposed method can be improved by adding doubly cabibbo - suppressed three - body modes such as @xmath20 .
we estimate the sensitivity of the proposed method using monte - carlo ( mc ) simulation .
to introduce the notation we briefly recap the technique of model - independent binned dalitz plot analysis of @xmath2 , @xmath9decays used to extract the angle @xmath0 . as usually presented
, this does not take charm mixing effects into account . the amplitude of the @xmath2
, @xmath9decay can be written as @xmath22 where @xmath23 is the amplitude of the @xmath24decay , @xmath25 is the amplitude of the @xmath26decay ( @xmath27 in the case of @xmath17conservation in @xmath1 decay ) , @xmath28 is the ratio of the absolute values of the interfering @xmath29 and @xmath30 amplitudes , and @xmath31 is the strong phase difference between these amplitudes .
the density of the @xmath1 decay dalitz plot from @xmath2 decay is given by the absolute value squared of the amplitude @xmath32 where @xmath33 the functions @xmath34 and @xmath35 are the cosine and sine of the strong phase difference @xmath36 between the @xmath24and @xmath26 amplitudes decay amplitudes introduced in ref .
@xcite . ] : @xmath37 the equations for the charge - conjugate mode @xmath38 are obtained with the substitution @xmath39 .
using both @xmath40 charges , one can obtain @xmath0 and @xmath31 separately . in the binned model - independent approach , the dalitz plot is divided into @xmath41 bins symmetrically to the exchange @xmath42 .
the expected number of events in the bin `` @xmath43 '' of the dalitz plot of @xmath1 from @xmath2 is @xmath44 \ , , \label{n_b}\ ] ] where @xmath45 is the number of events in the corresponding bin of the dalitz plot of the @xmath1 meson in a flavor eigenstate ( obtained using @xmath46 samples ) and @xmath47 is a normalization constant .
the bin index `` @xmath43 '' ranges from @xmath48 to @xmath49 ( excluding 0 ) ; the exchange @xmath50 corresponds to the exchange @xmath51 .
the terms @xmath52 and @xmath53 include information about the cosine and sine of the phase difference averaged over the bin region : @xmath54 here @xmath55 represents the dalitz plot phase space and @xmath56 is the bin region over which the integration is performed .
the terms @xmath53 are defined similarly with cosine substituted by sine . the expected number of events in each dalitz plot bin ( [ n_b ] ) is trivially obtained from the probability density ( [ p_b ] ) by integrating over the bin area , which leads to the substitutions @xmath57 , @xmath58 , @xmath59 . in what follows ,
we only quote the number of events to save space .
normalization constants ( such as @xmath60 in ( [ n_b ] ) ) are also omitted .
the symmetry under @xmath61 requires @xmath62 and @xmath63 .
the values of @xmath52 and @xmath53 terms can be provided by charm - factory experiments operated at the threshold of @xmath64 pair production @xcite .
the wave function of the two mesons is antisymmetric , thus the four - dimensional density of two correlated @xmath9dalitz plots is @xmath65 where the indices `` 1 '' and `` 2 '' correspond to the two decaying @xmath1 mesons . in the case of a binned analysis , the number of events in the region of the @xmath66 phase space described by the indices `` @xmath43 '' and `` @xmath67 '' is @xmath68 once the values of the terms @xmath52 and @xmath53 are known from charm - factory data , the system of equations ( [ n_b ] ) contains only three free parameters ( @xmath69 , @xmath70 , and @xmath60 ) for each @xmath40 charge , and can be solved using maximum likelihood method to extract the value of @xmath0 . note that technically the system ( [ n_b ] ) can be solved without external constraints on @xmath52 and @xmath53 for @xmath71 . however , due to the small value of @xmath28 , there is very little sensitivity to the @xmath52 and @xmath53 parameters in @xmath2decays , which results in a reduction in the precision on @xmath0 that can be obtained @xcite .
in the case of @xmath17conservation , the mass eigenstates of the neutral @xmath1 system are given by @xmath73 charm mixing is described by two parameters , @xmath74 and @xmath75 , which are defined as @xmath76 where @xmath77 and @xmath78 are the mass and decay widths of the mass eigenstates .
we use notations @xmath74 and @xmath75 instead of the more common @xmath14 and @xmath15 in order not to confuse them with the @xmath17-violating parameters @xmath69 and @xmath70 introduced before .
the current world average values are : @xmath79 , @xmath80 @xcite .
@xmath17violation modifies the expression ( [ d_cpcons ] ) to @xmath81 where @xmath82 and @xmath83 satisfy @xmath84 .
@xmath17-violating mixing is thus described by two additional parameters @xmath18 and @xmath19 : @xmath85 below we present all the quantities that enter the model - independent analysis including the contribution of the @xmath17-conserving charm mixing ( the corresponding quantities are denoted with the prime mark ) .
the full formalism including @xmath17violation in mixing is given in appendix [ full_formalism ] .
the amplitude for the correlated @xmath88 decay , assuming that the particle denoted with the index `` 1 '' decayed first , equals @xmath89 before the particle `` 1 '' decays , the amplitude stays antisymmetric and the mixing does not affect it .
we can therefore assume that the particle `` 1 '' decays at the time @xmath90 and count time from the moment of its decay .
after the integration over time and taking the absolute value squared of the amplitude ( [ a_corr ] ) , we obtain the number of events @xmath91 in the bin `` @xmath92 '' of the phase space for the pair of @xmath1 mesons ( still assuming that the particle `` 1 '' denoted here with the index `` @xmath43 '' decayed first ) : @xmath93 however , measurement of the decay time in an experiment with symmetric beams is difficult . therefore one has to average over the decay order , which leads to the cancellation of all terms linear in mixing parameters : @xmath94 the real analysis performed at cleo uses the values @xmath91 ( which , as we have seen , are unaffected by mixing at first order ) , and values of @xmath45 obtained from correlated @xmath64 decays where one of the @xmath1 mesons serves as a flavor tag .
the @xmath45 values extracted this way are also unaffected by mixing . therefore , the values of @xmath52 and @xmath53 extracted in this analysis contain no linear mixing contribution .
as far as processes observed at @xmath40 factories are concerned , both the @xmath9decay from @xmath2and the flavor - tagged @xmath9decay contain mixing contributions , and therefore the observable numbers of events in the dalitz plot bins are @xmath95 and @xmath96 , respectively . clearly , if one uses the values @xmath45 , obtained from a charm factory , in the fit to obtain @xmath0 from @xmath95 described by eq .
[ n_mix ] , the resulting value contains contribution in first order in @xmath74 , @xmath75 .
if the values @xmath96 are used , eq . [ n_mix ] can be rewritten as @xmath97 _ i. e. _ it has the same form as eq .
[ n_b ] , without mixing contributions up to second - order in @xmath98 , but the phase terms @xmath99 contain first - order mixing corrections : @xmath100 thus , if the terms @xmath101 are left as free parameters in the fit to @xmath40 decay data , the mixing correction is only of second order ( the effective parameters @xmath102 , @xmath103 are measured in this case @xcite ) , but if these terms are obtained from correlated @xmath1 decays , first order mixing corrections to @xmath0 appear .
however , these corrections are additionally suppressed by factor @xmath104 , and the residual contribution of charm mixing to @xmath69 and @xmath70 is at a percent level . 1 . using @xmath45 ( unaffected by mixing ) from the coherent @xmath64 production , 2
. using @xmath96 ( eq .
[ k_mix ] ) measured in @xmath10 , @xmath9decays , 3 . using @xmath45 and applying a linear correction for the mixing contribution according to eq .
[ n_mix ] ( assuming that @xmath74 , @xmath75 are known ) .
the effect of mixing on the fitted value of @xmath0 depends on the value @xmath105 , the ratio @xmath106 , and the @xmath31 and @xmath0 values . in our study
, we use @xmath107 ( all biases are proportional to this quantity ) and scan over the other parameters . the results are shown in table [ tab : gamma_bias ] . clearly ,
if the @xmath96 values from @xmath10 , @xmath9decays are used in the @xmath0 fit , the mixing contribution can be neglected .
.estimates of the charm mixing effect on the @xmath0 value measured using the model - independent time - integrated dalitz plot analysis method , for three different analysis strategies .
the maximum @xmath0 biases for varying @xmath108 , @xmath31 and @xmath0 , as well as the values of these parameters at which the maximum is reached , are shown .
the estimation assumes @xmath107 . [ cols="<,^,^,^,^",options="header " , ] the observed numbers of events @xmath109 are obtained using the exact formulas with mixing ( see appendix [ full_formalism ] ) .
when the systematic effect of charm mixing on the @xmath0 measurement is studied , the values of @xmath109 are fixed ( which corresponds to infinite statistics ) , while in the study of the statistical sensitivity of the mixing measurement in sections [ mixing_tint ] and [ other_modes ] , we sample the @xmath109 values according to the poisson distribution .
the expected numbers of events @xmath110 include the free parameters of interest ( @xmath69 and @xmath70 in section [ gamma_mixing ] or mixing parameters in sections [ mixing_tint ] and [ other_modes ] ) and are expressed via the numbers of events in flavor - specific @xmath5 state @xmath45 and the phase terms @xmath52 , @xmath53 .
the values of @xmath45 , @xmath52 and @xmath53 are calculated from the @xmath9and @xmath111 decay amplitudes @xcite as @xmath112 @xmath113 @xmath114 the free parameters are extracted using minuit to minimize the likelihood function ( [ eq : lhfunc ] ) .
table [ tab : evnum_step ] shows the numbers of events that we use to obtain the results in sections [ mixing_tint ] and [ other_modes ] . to estimate the numbers we use the detection efficiency determined by cleo @xcite and the statistics corresponding to an integrated luminosity of @xmath115 fb@xmath116 at the @xmath117 resonance .
100 a. giri , y. grossman , a. soffer and j. zupan , phys .
d * 68 * ( 2003 ) 054018 [ arxiv : hep - ph/0303187 ] .
a. bondar .
proceedings of binp special analysis meeting on dalitz analysis , 24 - 26 sep . 2002 , unpublished . k. abe _ et al .
_ [ belle collaboration ] , phys .
lett . * 99 * , 131803 ( 2007 ) [ arxiv:0704.1000 [ hep - ex ] ] .
del amo sanchez _ et al .
_ [ the babar collaboration ] , arxiv:1004.5053 [ hep - ex ] .
a. bondar , t. gershon , p. krokovny , phys .
b * 624 * 1 - 10 ( 2005 ) [ arxiv : hep - ph/0503174 ] p. krokovny _ et al .
_ [ belle collaboration ] , phys .
lett . * 97 * , 081801 ( 2006 ) [ arxiv : hep - ex/0605023 ] . k. abe _ et al . _ [ belle collaboration ] , arxiv:0803.3375 [ hep - ex ] .
b. aubert _ et al . _ [ babar collaboration ] , phys .
d * 78 * ( 2008 ) 034023 [ arxiv:0804.2089 [ hep - ex ] ] .
r. a. briere _ et al . _
[ cleo collaboration ] , phys .
d * 80 * , 032002 ( 2009 ) arxiv:0903.1681 [ hep - ex ] . m. staric _ et al . _ [ belle collaboration ] , phys .
lett . * 98 * , 211803 ( 2007 ) [ arxiv : hep - ex/0703036 ] .
b. aubert _ et al . _ [ babar collaboration ] , phys .
lett . * 98 * , 211802 ( 2007 ) [ arxiv : hep - ex/0703020 ] .
y. grossman , a. soffer , j. zupan , phys .
d * 72 * , 031501 ( 2005 ) .
a. bondar and a. poluektov , eur .
j. c * 55 * ( 2008 ) 51 [ arxiv:0801.0840 [ hep - ex ] ] .
a. bondar and a. poluektov , eur .
j. c * 47 * ( 2006 ) 347 [ arxiv : hep - ph/0510246 ] .
hfag preliminary averages for fpcp2010 , http://www.slac.stanford.edu/xorg/hfag/charm/index.html j. p.
silva and a. soffer , phys .
d * 61 * , 112001 ( 2000 ) [ arxiv : hep - ph/9912242 ] . y. grossman , a. l. kagan and y. nir , phys .
d * 75 * , 036008 ( 2007 ) [ arxiv : hep - ph/0609178 ] .
e. golowich , j. hewett , s. pakvasa and a. a. petrov , phys .
d * 76 * , 095009 ( 2007 ) [ arxiv:0705.3650 [ hep - ph ] ] .
d. m. asner _
[ cleo collaboration ] , phys .
d * 78 * ( 2008 ) 012001 [ arxiv:0802.2268 [ hep - ex ] ] .
d. m. asner and w. m. sun , phys .
d * 73 * , 034024 ( 2006 ) [ erratum - ibid .
d * 77 * , 019902 ( 2008 ) ] [ arxiv : hep - ph/0507238 ] .
d. cronin - hennessy _ et al .
_ [ cleo collaboration ] , phys . rev .
d * 80 * , 072001 ( 2009 ) [ arxiv:0801.3418 [ hep - ex ] ] .
a. blinov _ et al .
_ icfa beam dyn .
48:268 - 279 , 2009 .
s. kopp _ et al .
_ [ cleo collaboration ] , phys . rev .
d * 63 * , 092001 ( 2001 ) [ arxiv : hep - ex/0011065 ] .
b. aubert _ et al . _ [ babar collaboration ] , phys .
lett . * 103 * , 211801 ( 2009 ) [ arxiv:0807.4544 [ hep - ex ] ] .
n. lowrey _ et al . _
[ cleo collaboration ] , phys .
d * 80 * , 031105 ( 2009 ) [ arxiv:0903.4853 [ hep - ex ] ] .
z. z. xing , phys .
d * 55 * , 196 ( 1997 ) [ arxiv : hep - ph/9606422 ] . | we investigate the impact of charm mixing on the model - independent @xmath0 measurement using dalitz plot analysis of the three - body @xmath1 decay from @xmath2process , and show that ignoring the mixing at all stages of the analysis is safe up to a sub - degree level of precision .
we also find that in the coherent production of the @xmath3 system in @xmath4 collisions , the effect of charm mixing is enhanced , and propose a model - independent method to measure charm mixing parameters in time - integrated dalitz plot analysis at charm factories . |
the hamiltonian for the quantum spin hall josephson junction takes the form @xmath64 with nambu spinor @xmath65 in terms of electron operators and the bogoliubov - de gennes hamiltonian @xmath66 where @xmath25 ( @xmath67 ) denotes the coordinate ( momentum ) along the quantum spin hall insulator edge , @xmath68 is the edge - mode velocity , and @xmath69 and @xmath70 are pauli matrices in spin and nambu ( particle - hole ) space , respectively .
the proximity - induced superconducting gap @xmath71 \nonumber \\ & = \delta\theta({\left| x \right|}-\frac{l}{2 } ) e^{i\varphi(x)\tau_z } \end{aligned}\ ] ] has strength @xmath72 and a phase difference @xmath6 across the junction region of length @xmath73 .
we have introduced a spatially dependent phase @xmath74 for convenience .
we introduce a local gauge transformation @xmath75 to eliminate the spatial dependence of the superconducting phase , and obtain the transformed hamiltonian @xmath76 where the prime denotes a derivative with respect to @xmath25 .
we will denote @xmath77 as @xmath78 in the following . to solve for the andreev bound states
, we follow the approach detailed in ref .
@xcite to rearrange the bogoliubov - de gennes equation @xmath79 as @xmath80 \psi.\ ] ] the solution can be written as @xmath81 in terms of the state at some reference point @xmath82 .
in particular , we have @xmath83 = \exp\left(\frac{iel}{v}\sigma_z\tau_z-\frac{\phi}{2}\tau_z\right ) \right\}\ ] ] which connects the states @xmath84 .
we match the properly decaying solutions of the bogoliubov de - gennes equation on the left and right of the junction , and obtain the bound state wave functions in the two spin sectors : @xmath85 and @xmath86 where @xmath87 with @xmath88 , and @xmath89 is the positive eigenvalue in each spin sector given by the relation @xmath90 and @xmath91 for @xmath6 around @xmath92 , solutions in both spin sectors can exist simultaneously for @xmath93 .
we will consider the situation when at most one solution in each spin sector exists , and write the subgap effective hamiltonian as @xmath94 where @xmath95 is the bogoliubov quasiparticle annihilation operator for the andreev bound state with spin @xmath96 .
when projected onto the subspace spanned by the subgap andreev bound states , the electron annihilation operators for both spins can be written approximately as @xmath97
consider the hamiltonian describing the coupling of the edge channel to a magnetic impurity with spin s @xmath98 in which the spin density of the helical edge can be written in term of the bogoliubov operators @xmath99 where @xmath100 note that we can also write @xmath101 then we have @xmath102 @xmath103\left(e^{i\frac{(\epsilon_\uparrow + \epsilon_\downarrow)l}{2v } } a_{\uparrow}^{*}a_{\downarrow}^{*}\gamma_{\downarrow}^{\dagger}\gamma_{\uparrow}^{\dagger}+e^{-i\frac{(\epsilon_\uparrow+\epsilon_\downarrow)l}{2v}}a_{\uparrow}a_{\downarrow } \gamma_{\downarrow}^{\dagger}\gamma_{\uparrow}^{\dagger } \right ) .
\end{aligned}\ ] ] generic many body spectrum for the quantum spin hall josephson junction coupled to a spin-@xmath104 impurity with @xmath105 .
the red solid and blue dashed curves indicate that the corresponding many - body states have even and odd fermion parity , respectively .
, scaledwidth=50.0% ]
in the case @xmath106 $ ] , and for @xmath6 close to @xmath107 we have @xmath108 now we consider @xmath112 and denote @xmath113 . by eq .
( [ eq : up_spin ] ) and ( [ eq : down_spin ] ) and condition ( [ eq : condition ] ) , we have @xmath114 \nonumber \\ & = \delta\cos\frac{\phi}{2}\left(\cos\frac{\epsilon_\uparrow l}{v } - \cos\frac{\epsilon_\downarrow l}{v } \right ) + \delta\sin\frac{\phi}{2}\left ( \sin\frac{\epsilon_\uparrow l}{v } + \sin\frac{\epsilon_\downarrow l}{v } \right ) \nonumber \\ & \simeq \delta \sin(\frac{\epsilon l}{v } ) \phi = \kappa v \phi , \end{aligned}\ ] ] where @xmath115 this is valid up to first order in @xmath6 . in this situation , the operators @xmath116 , @xmath117 and @xmath118 get simplified as @xmath119 a_{\uparrow}^{*}a_{\downarrow}\gamma_{\uparrow}^{\dagger}\gamma_{\downarrow } \\
\hat{\sigma}^- & = 2\exp\left[-i\left(\kappa l - 1\right)\frac{\phi}{2 } \right ] a_{\uparrow}a_{\downarrow}^*\gamma_{\uparrow}^{\dagger}\gamma_{\downarrow } \\
\hat{\sigma}^z & = { \left| a_{\uparrow } \right|}^{2}(2\gamma_{\uparrow}^{\dagger}\gamma_{\uparrow}-1)-{\left| a_{\downarrow } \right|}^{2}(2\gamma_{\downarrow}^{\dagger}\gamma_{\downarrow}-1).\end{aligned}\ ] ] because of this , the total occupation number @xmath120 becomes a good quantum number , namely @xmath121=0 $ ] where @xmath122 .
the many body hilbert space is spanned by the states @xmath123 with @xmath124 labeling the occupation number of the andreev bound state and @xmath125 labeling the eigenstates of @xmath126 of the impurity spin .
the subspace for @xmath127 is spanned by @xmath128 , and @xmath129 .
the hamiltonian @xmath46 in this subspace is represented by a @xmath130 by @xmath130 matrix @xmath131 where @xmath132 with @xmath133 are pauli matrices in this two dimensional subspaces .
the subspace with @xmath135 is spanned by @xmath136 , @xmath137 , @xmath138 , @xmath139 .
the hamiltonian in this case can be written as @xmath140a_{\uparrow}^{*}a_{\downarrow}(j_{z+}\tau_{z}+j_{++}\tau_{+}+j_{-+}\tau_{-})\rho_{+}+h.c.\ ] ] where @xmath141 and @xmath142 are pauli matrices in @xmath143 space , and we have neglected terms linear in @xmath144 .
an unitary transformation @xmath145 for some @xmath146 $ ] can always be chosen such that @xmath147 \end{aligned}\ ] ] which has the same spectrum as @xmath148 .
the four eigenvalues are generically nondegenerate . | we study the josephson effect in a quantum spin hall system coupled to a localized magnetic impurity . as a consequence of the fermion parity anomaly , the spin of the combined system of impurity and spin - hall edge alternates between half - integer and integer values when the superconducting phase difference across the junction advances by @xmath0 .
this leads to characteristic differences in the splittings of the spin multiplets by exchange coupling and single - ion anisotropy at phase differences , for which time - reserval symmetry is preserved .
we discuss the resulting @xmath1-periodic ( or @xmath2 ) fractional josephson effect in the context of recent experiments .
_ introduction._the fractional josephson effect @xcite constitutes one of the most striking effects heralding topological superconductivity @xcite . in a josephson junction made from conventional superconductors ,
the josephson current is carried by cooper pairs and is @xmath0 periodic in the phase difference applied to the junction .
when the junction connects topological superconductors @xcite , the coupling of majorana bound states across the junction allows a josephson current to flow by coherent transfer of single electrons , resulting in @xmath3 periodicity in the phase difference .
robust @xmath3 periodicity requires that time - reversal symmetry be broken through proximity coupling to a magnetic insulator or an applied magnetic field @xcite .
a fractional josephson effect can occur in time - reversal - symmetric junctions as a consequence of electron - electron interactions @xcite . in the limit of strong interactions ,
this @xmath1-periodic effect can be understood in terms of domain walls carrying @xmath2 parafermions , enabling tunneling of @xmath4 quasiparticles between the superconductors .
recent experiments on superconductor quantum spin hall
superconductor junctions show intriguing evidence for @xmath3-periodic josephson currents .
one experiment probes shapiro steps and shows that the first shapiro step is absent @xcite .
a second experiment reports that the josephson radiation emitted by a biased junction is also consistent with @xmath3 periodicity @xcite .
these results are surprising as both experiments were performed without explicitly breaking time - reversal symmetry so that basic theory would predict a dissipative @xmath0-periodic behavior when neglecting electron - electron interactions , or an @xmath1-periodic behavior when taking interactions into account .
these expectations are based on considering pristine quantum spin hall josephson junctions with a fully gapped bulk and a single helical channel propagating along its edges .
density modulations in actual quantum spin hall samples are widely believed to induce puddles of electrons in addition to the helical edge channels @xcite . when these puddles host an odd number of electrons , charging effects turn them into magnetic impurities which are exchange coupled to the helical edge channels . in this paper , we discuss the fractional josephson effect in realistic quantum spin hall josephson junctions which include such magnetic impurities .
the effects of magnetic impurities on quantum spin hall edge channels have been intensively studied in the absence of superconductivity @xcite . in the high - temperature limit
, a magnetic impurity induces backscattering between the kramers pair of helical edge channels and thus deviations from a quantized conductance in a two - terminal measurement . as the temperature is lowered , the impurity spin is increasingly kondo screened by the helical edge channel and perfect conductance quantization is recovered in the zero - temperature limit . in the presence of superconductivity ,
the kondo effect is quenched by the superconducting gap so that one may expect that magnetic impurities field more prominent consequences .
here , we assume that the kondo temperature is well below the superconducting gap so that we can safely neglect the effects of kondo screening .
we find that coupling to magnetic impurities alters the behavior of quantum spin hall josephson junctions qualitatively .
the josephson current generically becomes @xmath1 periodic , replacing the dissipative @xmath0-periodic effect in the absence of this coupling .
this can be viewed as a variant of the @xmath2 josephson effect .
indeed , unlike its classical counterpart , coupling to a quantum spin preserves time - reversal symmetry and interactions are effectively included through the local - moment formation , which is quite reminiscent of the ingredients of the @xmath2 fractional josephson effect .
thus , our results show that this remarkable effect is considerably more generic than one might have previously thought
. moreover , the present setting emphasizes a remarkable mechanism for producing an @xmath1-periodic fractional josephson effect . as a result of the fermion parity anomaly @xcite
, the spin of the helical edge effectively changes by @xmath5 when the superconducting phase difference is advanced by @xmath0 .
this adiabatically transmutes the combined spin of helical edge and magnetic impurity between half - integer and integer values , with their characteristically different behavior in the presence of time - reversal symmetry as described by the kramers theorem . _
quantum spin hall josephson junctions._we begin by reviewing the andreev spectrum of quantum spin hall josephson junctions in the absence of magnetic impurities @xcite .
consider a quantum spin hall edge with counterpropagating edge modes , placed in between two superconductors whose phases differ by @xmath6 .
the subgap spectrum of such a josephson junction as a function of @xmath6 is shown in fig . 1 . ; ( b ) @xmath7 in the presence of backscattering due to a zeeman field ; ( c ) @xmath8 ; ( d ) @xmath9 ; ( e ) @xmath10 .
the green curves correspond to andreev states consisting of a superposition of an up - spin electron and an andreev - reflected hole .
the orange curves are for the particle - hole conjugated states.,scaledwidth=48.0% ] in the short junction limit @xmath11 , the subgap spectrum contains a single particle - hole symmetric pair of andreev states [ see fig . [
fig : abs](a ) ] .
both andreev levels emanate from and merge into the quasiparticle continuum .
hence , when the phase advances proportional to an applied bias voltage @xmath12 , the junction exhibits an @xmath13 josephson effect with the conventional frequency @xmath14 and energy dissipation rate @xmath15 . the dissipative nature of the josephson effect is closely related to the helical nature of the edge modes and the ensuing absence of backscattering .
when introducing backscattering into the junction by breaking time - reversal symmetry through an applied magnetic field or proximity coupling to a magnetic insulator , the andreev levels no longer merge with the quasiparticle continuum [ see fig . [
fig : abs](b ) ] , quenching dissipation in the small - voltage limit @xcite .
moreover , the @xmath13 josephson effect occurs at half the conventional frequency , i.e. , at @xmath16 , as fermion number parity is conserved .
indeed , the level crossing at @xmath17 is protected by fermion number parity so that the individual andreev levels are @xmath3 periodic in the phase difference @xmath6 .
this can be viewed as a consequence of the fermion parity anomaly : as a result of the quantum spin hall effect , the fermion parity of the edge changes when the superconducting phase difference is advanced by @xmath0 , requiring a phase change of @xmath3 for a full period .
additional subgap levels appear for longer junctions , see figs .
[ fig : abs](c ) and ( d ) .
the structure of level crossings in these spectra is not only controlled by fermion parity , but also by time - reversal symmetry . while time reversal is generally broken by the phase difference across the junction ( resulting in a nonzero josephson current flowing in the junction ) , it remains unbroken at phase differences equal to integer multiples of @xmath18 .
_ coupling to magnetic impurity._we now consider the coupling of the edge channel to a magnetic impurity with spin s. generically , disorder in conjunction with the strong spin - orbit coupling will remove any symmetry other than time reversal which we assume to be broken only by the applied superconducting phase difference .
thus , we focus on the general hamiltonian @xmath19 for the impurity spin @xmath20 .
the first term describes the exchange coupling between the impurity spin and the helical edge , with @xmath21 denoting the local spin density of the helical edge at the position @xmath22 of the coupling to the impurity .
the operator @xmath23 annihilates an electron with spin projection @xmath24 at position @xmath25 , and @xmath26 denotes the @xmath27 pauli matrix .
the second term describes a general single - ion anisotropy of the impurity spin with strengths @xmath28 .
time reversal implies that the exchange couplings are real , but otherwise arbitrary . ) without impurity spin ( left ) and with coupling to the impurity spin ( right , @xmath29 is of order @xmath30 ) .
the red solid and blue dashed curves indicate even and odd fermion parity , respectively , with the discontinuity in parity at @xmath17 originating from the merging of andreev levels with the continuum , see fig .
[ fig : abs](d ) .
the crossings at and near @xmath31 indicated by black circles are between states of opposite fermion parity .
the crossings between states with even fermion parity at @xmath32 and @xmath0 , highlighted by red dashed circles , are protected by time reversal .
the arrows indicate the impurity - spin polarization along the @xmath33-axis .
( b ) fourier transforms of the many - body ground state energy or , equivalently , the josephson current ( upper panel ) and of the expectation value of the impurity spin polarization @xmath34 ( lower panel ) as a function of the phase difference @xmath6 .
the @xmath1-periodic harmonics are indicated by the vertical dashed lines.,scaledwidth=45.0% ] _ josephson effect._analyzing the josephson effect of the quantum spin hall edge channel coupled to the magnetic impurity is greatly simplified by the discrete nature of the subgap spectrum . for definiteness , consider an intermediate - length junction whose subgap spectrum has exactly two positive - energy subgap states @xmath35 ( @xmath36 ) at all values of the phase difference as in fig .
[ fig : abs](d ) .
( this convenient choice is used in our numerical illustrations but not essential for our results . )
then , we can analyze the low - energy ( many - body ) spectrum of the junction coupled to the impurity spin in the finite - dimensional basis of low - energy states spanned by the product of occupation states of the two subgap bogoliubov quasiparticles ( yielding four basis states ) and the @xmath37 spin states of the spin-@xmath38 impurity .
the low - energy many - body spectrum is effectively decoupled from the quasiparticle continuum as long as the exchange couplings @xmath39 are not too large .
the corresponding hamiltonian is readily derived by retaining only the contributions of the two positive - energy subgap bogoliubov operators @xmath40 to the edge - state electron operators ( see @xcite for details ) . in this limit
, the total hamiltonian can be approximated as @xmath41 with @xmath42 the hamiltonian of the bare edge .
consider coupling the quantum spin hall edge states to a spin-1/2 impurity .
figure [ fig : s = half](a ) shows the many - body spectrum of @xmath43 in eq .
( [ eq : hbare ] ) , i.e. , of the bare edge ( left panel ) , and of @xmath44 for a generic choice of exchange couplings @xmath39 ( right panel ) .
the spectrum of the coupled edge is best understood by analyzing the nature of the degeneracies at phase differences equal to integer multiples of @xmath18 .
the degeneracies at and near @xmath17 are protected by fermion parity . here
, level crossings occur between states with even and odd occupations of the bogoliubov quasiparticles of the edge .
in contrast , the level crossings at @xmath32 and @xmath45 occur between states of the same fermion parity and are kramers degeneracies reflecting time - reversal symmetry . in the present system ,
a kramers degeneracy appears when the bogoliubov quasiparticles @xmath40 of the edge are either both empty or both occupied , leading to a half - integer spin of the combined system of edge and impurity .
specifically , the lower ( higher ) energy crossing in fig .
[ fig : s = half](a ) corresponds to states in which the quasiparticle states are both empty ( occupied ) .
away from @xmath32 and @xmath0 , time reversal is broken and the kramers degeneracies are lifted .
this interpretation is corroborated by further restricting the hamiltonian @xmath46 for small @xmath6 to the low - energy subspace of empty quasiparticle states . in this limit
, the spin density @xmath47 of the edge only has a nonzero @xmath33 component @xmath48 $ ] and the hamiltonian simplifies to @xmath49 with the effective zeeman field @xmath50
\sum_\alpha j_{\alpha z } { \bf \hat e}_\alpha$ ] . here
, we use the subgap energy @xmath51 and @xmath52 .
the four nondegenerate states at intermediate energies for @xmath32 [ see fig . [
fig : s = half](a ) ] have overall single occupation of the quasiparticle states , leading to a combined edge - impurity system with integer spin .
unlike in the odd - integer spin case , time reversal does not enforce a degeneracy of the many - body spectrum in this case .
writing the hamiltonian for small @xmath6 in this subspace using the basis @xmath53 and @xmath54 ( with @xmath55 ) for the states of the edge ( with corresponding pauli matrices @xmath56 ) , we find the effective hamiltonian @xmath57\ ] ] generically , this hamiltonian has no degeneracies . with this understanding , the many - body spectrum in fig .
[ fig : s = half](a ) reveals a remarkable fact : the total spin of the edge - impurity system alternates between half - integer and integer spin when the superconducting phase difference advances adiabatically from @xmath32 to @xmath45 . this spin transmutation is a direct consequence of the parity anomaly . as the phase difference changes by @xmath0 , the fermion parity of the edge changes by virtue of the quantum spin hall effect .
consequently , also the spin of the edge changes by @xmath5 .
this change in spin has important consequences for the periodicity of the josephson effect .
indeed , adiabatically following the energy levels in fig .
[ fig : s = half](a ) , we find that they are @xmath1 periodic , corresponding to an @xmath13 josephson frequency of @xmath58 .
due to the spin transmutation , the system passes through successive kramers degeneracies only after advancing the superconducting phase difference by @xmath3 , requiring a phase change of @xmath1 for completing a full period .
note also that starting with the ground state at @xmath32 , the many - body state remains well below the quasiparticle continuum for all @xmath6 , so that the @xmath13 josephson effect remains nondissipative at a sufficiently small bias .
the polarization of the impurity spin varies with the superconducting phase difference in an @xmath1-periodic manner .
when adiabatically varying the superconducting phase difference , the spin orientation remains unchanged at the kramers crossings and flips in the vicinity of the avoided crossings where the edge - impurity system is in an integer - spin state .
this variation of the spin with @xmath6 is illustrated in fig .
[ fig : s = half](a ) . impurity .
the red solid and blue dashed curves correspond to many - body states with even and odd fermion parity , respectively .
spectra correspond to ( a ) vanishing single - ion anisotropy , ( b ) easy - plane anisotropy @xmath59 , ( c ) easy - axis anisotropy @xmath60 , and ( d ) generic single - ion anisotropy with @xmath61 .
the degeneracies at @xmath45 ( as well as their partners at @xmath32 ) highlighted by blue dashed circles are kramers degeneracies .
degeneracies highlighted by red circles are lifted by introducing generic single - ion anisotropy .
subsequent @xmath0 periods for an adiabatic evolution of @xmath6 are indicated by increasing numbers of arrows.,scaledwidth=45.0% ] these results for @xmath62 impurities are in fact generic and persist for higher - spin impurities . as an illlustration ,
consider the results for an @xmath63 impurity in fig .
[ fig : s=1 ] . first consider panel ( d ) which shows results for generic values of @xmath39 and @xmath28 .
unlike in the @xmath62 case , the low - energy states are now integer spin states and thus nondegenerate , while the intermediate - energy states have half - integer spin and are kramers degenerate at @xmath32 and @xmath0 .
nevertheless , the @xmath1 periodicity remains intact .
different periodicities can occur for nongeneric choices of the couplings @xmath39 and @xmath28 . in the absence of any single - ion anisotropy
[ see fig .
[ fig : s=1](a ) ] , the many - body spectrum does not decouple from the quasiparticle continuum and the josephson effect becomes dissipative and @xmath0 periodic .
the same result is obtained for easy - plane anisotropy , with one of the single - ion anisotropies being positive and the others equal to zero , see fig .
[ fig : s=1](b ) .
finally , easy - axis anisotropy makes the junction nondissipative and @xmath3 periodic as shown in fig .
[ fig : s=1](c ) . _
discussion._we find that generically , coupling to a magnetic impurity makes the josephson effect in quantum spin hall systems @xmath1 periodic .
the @xmath1 periodicity relies only on time - reversal symmetry , the parity anomaly , and the absence of fine tuning such as the absence of interactions or the presence of spin conservation .
this general conclusion requires two comments .
first , while the josephson current is @xmath1 periodic , with a frequency @xmath58 of the @xmath13 josephson effect , this may actually not be the dominant fourier component observable in experiment .
indeed , it is evident from fig .
[ fig : s = half ] that the @xmath1-periodic cycle consists of two rather similar @xmath3 sections .
the magnitude of the splitting of the two sections is controlled by the strength of the exchange coupling . as long as this
exchange splitting remains weak compared to the superconducting gap , the dominant fourier component of the josephson current is @xmath3 periodic .
this is shown in fig .
[ fig : s = half](b ) , together with the fourier analysis of the expectation value of the impurity spin which has a dominant @xmath1-periodic harmonic .
it is interesting to note that this result for the josephson current is different from the realization of the @xmath2 josephson effect discussed by zhang and kane @xcite which has a dominant @xmath1-periodic fourier component .
second , it is important to realize that so far , our results consider only the electronic system and neglect coupling to other degrees of freedom such as phonons or the electromagnetic environment . relaxing this restriction introduces inelastic relaxation processes which may be crucial for determining the experimentally observed periodicity . while relaxation between states of opposite fermion parity may be quite slow
, relaxation beween states of the same fermion parity should be considerably more efficient .
observation of the @xmath1 periodicity requires that the latter relaxation processes are slow compared to the period in which the @xmath1 cycle is traversed .
indeed , the two @xmath3 sections of the @xmath1 cycle involve states of the same fermion parity , so that one would always remain in the lower - energy state if the cycle is traversed slowly on the time scale of parity - conserving relaxation processes .
this would make the observed josephson effect @xmath3 rather than @xmath1 periodic .
it is intriguing to compare these results to the recent experiments on quantum spin hall junctions which observe shapiro steps and josephson radiation consistent with @xmath3 periodicity @xcite .
our results provide an interesting scenario that is in principle consistent with these observations . at the same time , this is by no means the only explanation of a @xmath3-periodic josephson effect in this system .
an alternative scenario considers relaxation processes in a pristine quantum spin hall junction .
consider an intermediate - length junction with at least two positive - energy andreev states for any phase difference .
when both of these andreev states are occupied , the two quasiparticles can relax inelastically by recombining into a cooper pair .
two positive - energy quasiparticles are created every time the phase difference advances by @xmath3 .
thus , if their recombination into a cooper pair is an efficient process , one would also observe a @xmath3-periodic josephson effect .
it is an interesting problem for future research to devise experimental probes which distinguish between these alternative scenarios for a @xmath3-periodic josephson effect in quantum spin hall josephson junctions . _
acknowledgments._we acknowledge financial support by the deutsche forschungsgemeinschaft ( crc 183 as well as priority program 1666 ) , the minerva foundation , the alexander - von - humboldt foundation , and nsf grant dmr-1603243 .
part of this work was performed at the aspen center for physics , which is supported by national science foundation grant phy-1066293 .
18ifxundefined [ 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 * * , ( ) link:\doibase 10.1140/epjb / e2004 - 00066 - 4 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.161408 [ * * , ( ) ] link:\doibase 10.1088/0034 - 4885/75/7/076501 [ * * ( ) , 10.1088/0034 - 4885/75/7/076501 ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.100.096407 [ * * , ( ) ]
link:\doibase 10.1103/physrevlett.105.077001 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.177002 [ * * , ( ) ] link:\doibase 10.1038/nphys1915 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.113.036401 [ * * , ( ) ] @noop ( ) @noop ( ) link:\doibase 10.1103/physrevlett.110.216402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.256803 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.106.236402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.206803 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.086401 [ * * , ( ) ] supplemental material available online .
link:\doibase 10.1103/physrevb.94.085409 [ * * , ( ) ] |
several years ago , linde and vilenkin pointed out the possibility that the core of a localized topological defect could inflate under appropriate conditions in a process that was aptly dubbed topological inflation @xcite .
see also @xcite .
a common characteristic of such defects is some non - linear scalar field ( the higgs field ) forced up in its core into a constant excited false vacuum state , falling through a transition layer to the true vacuum value remote from the core . without further refinement ,
a configuration of this type will always collapse , with or without gravity .
the dynamics of this basic configuration , a so called false vacuum bubble , was studied in detail in the eighties by various groups , perhaps most comprehensively by blau , guendelman and guth ( bgg ) in ref.@xcite where references to the earlier literature are provided .
their work was motivated , like linde and vilenkin s , by the possibility that a false vacuum bubble could provide a seed for an inflationary universe @xcite .
they discovered that the essential radial dynamics of the scalar field is captured extremely accurately by a simple one dimensional mechanical caricature of the field configuration .
this led them to consider the dynamics of a spherically symmetric region of false vacuum which is separated by a domain wall from an infinite true vacuum exterior ( with zero energy density ) .
the energy momentum tensor of false vacuum is inherently homogeneous and isotropic .
the birkhoff theorem then constrains the spacetime it occupies to coincide with some patch of de sitter space ; the exterior is simply the schwarzschild geometry truncated at the false vacuum boundary . in this model , it becomes clear that collapse does not necessarily spell the demise of the false vacuum interior .
when gravity is taken into account , there are ( at least ) two spherically symmetric configurations associated with each sufficiently low value of conserved arnowitt - deser - misner ( adm ) mass @xmath3 . while the smaller of the two , no matter what its initial radius , will always collapse to a vanishing radius , the other will inflate ; but unexpectedly from an euclidean perspective , not at the expense of the exterior .
this peculiar state of affairs is possible because the expansion occurs behind a minimal surface ( a wormhole ) in the schwarzschild geometry connecting the false vacuum core to the exterior .
while false vacuum is destroyed by the motion of the boundary , it is created exponentially faster in the interior .
the wormhole itself , however , always collapses into a black hole .
field theories admitting static configurations when gravity is turned on were first constructed in the eighties .
self - gravitating global monopoles were considered by barriola and vilenkin @xcite ( with additional insight provided in ref .
gauge monopoles were considered by gibbons @xcite , with the subsequent numerical solution of the static einstein - yang - mills - higgs equations by ortiz @xcite and breitenlohner , forgacs and maison @xcite . however , as gibbons himself realized the balance of forces is not always possible when gravity is acting . as linde and vilenkin were later to show , if the symmetry - breaking scale is increased towards the planck scale , at some point the interior radius will exceed the corresponding cosmological horizon , triggering the inflation of the interior . a bound exists beyond which static configurations necessarily become unstable .
this process was studied by sakai _
et al . _
@xcite by numerically solving the dynamical einstein - higgs and einstein - yang - mills higgs equations .
the inflating core was not unlike an inflating false vacuum bubble .
in ref.@xcite two of us showed how the model of a false vacuum bubble could be adapted to imitate the dynamics of a self - gravitating global monopole under these extreme conditions .
technically this was simple , involving the substitution of the barriola - vilenkin geometry describing the asymptotics of a global monopole for the exterior schwarzschild geometry of the former .
it was possible to capture , remarkably faithfully , the essential underlying physics of topological inflation found earlier by sakai _
the onset of topological inflation was very clearly indicated at @xmath7 ( in natural units ) .
the details of topological inflation in a gauge monopole are very different . here , the only long range field is the magnetic coulomb field and the total energy is finite . however , the source resides on the core boundary which suggests the same mechanical caricature : de sitter space inside , a domain wall , but with the reissner - nordstrm geometry outside .
the structure of this geometry is very different from schwarzschild .
if the conserved charge to mass ratio of the configuration @xmath8 , the analyically continued geometry possesses horizons , otherwise it contains a naked singularity . in this paper , we examine the above model in detail .
though simple in principle , a thorough analysis of the three - dimensional parameter space @xmath9 , @xmath2 , @xmath3 ) is complicated in practice .
we will focus on the identification of the regimes of parameter space admitting solutions which are either static , collapsing or inflating within the core and we will examine the fate of these solutions as the relevant boundaries in parameter space are crossed .
our results can be summarized as follows : 1 .
fix @xmath1 : for each non - vanishing value of @xmath2 up to some critical value @xmath10 there exists a unique stable and globally static configuration with a fixed core radius ( a monopole ) with mass @xmath11 .
stable radial oscillations of this configuration exist for all @xmath5 above @xmath12 .
these are the only solutions with @xmath5 .
+ now raise @xmath3 above @xmath2 but below some value @xmath13 : 2 .
\(i ) for non - vanishing values of @xmath2 up to some value @xmath14 lower than @xmath10 , this radially oscillating solution falls through an event horizon and is terminated by the collapse of the exterior into a reissner - nordstrm black hole ; ( ii ) for higher ( but bounded ) values of @xmath2 radial oscillations lie within the inner horizon and get isolated by the collapse of the exterior ( a monopole inside a black - hole ) ; an inflating bounce co - exists with these solutions ; ( iii ) if @xmath2 or @xmath3 lies outside these two regimes , but @xmath3 lying above some minimum value , there is a unique inflating bounce .
the bounce occuring in these three regimes can be characterized roughly as follows : ( i ) the course of the bounce lies within a single asymptotically flat region ( afr ) and it resembles closely the bounce exhibited by a false vacuum bubble ( with @xmath6 ) ; ( ii),(iii ) the course of the bounce spans two consecutive afrs . however , from the point of view of either region it resembles a monotonic false vacuum bubble ; in all cases , the expansion takes place behind an event horizon .
these configurations are analogues in this model of the topologically inflating solutions observed numerically by sakai @xcite for large @xmath1 .
4 . all collapsing and monotonic solutions are ruled out as either unphysical or inconsistent with asymptotically flat boundary conditions .
aspects of the model have been examined before . indeed in the sixties
, it received its first incarnation in dirac s proposal ( without spin ! ) of a model of the electron as a closed charged conducting membrane surrounding a vacuum interior@xcite .
tachizawa , maeda and torii focused on the stability of the monopole from the point of view of catastrophe theory @xcite , modeling the monopole core and exterior as we do but without an intermediate surface layer , a model originally proposed by lee , nair and weinberg , in ref . @xcite . in this limit
, the core radius exceeds the cosmological horizon when @xmath15 , signaling inflation .
however , without the domain wall to transmit energy from the false vacuum , all dynamical possibilities are not faithfully represented .
more recently , alberghi , lowe and trodden @xcite considered the model within the context of the anti- de sitter space / conformal field theory correspondence .
for this purpose they catalogued accurately the possible trajectories of the charged false vacuum bubble .
however , they did not consider how these trajectories depend on the values of @xmath2 , @xmath3 or @xmath1 and they did not consider the parameter regime @xmath16 corresponding to stable configurations .
the paper is organized as follows . in sec .
[ model ] we introduce the model . in sec .
[ ssfk ] , we determine all possible trajectories of the wall radius consistent with set values of @xmath2 , @xmath1 and @xmath3 . in sec .
[ ebwm ] we describe briefly the interior and exterior spacetimes and how the routing of trajectories is determined in each . in secs .
[ bwtst ] - [ bgg ] , we identify all physically interesting solutions and compare our results with earlier work . finally , in sec .
[ con ] we conclude with a few brief comments .
the configuration possesses a core in which the magnitude of the higgs field approximates its false vacuum value , @xmath17 ; in the core region , the potential energy of the higgs field dominates the gradient energy in the higgs and gauge fields .
we model this core by a spherically symmetric region of false vacuum , and for the mexican sombrero potential @xmath18 this energy density is given by @xmath19 .
the corresponding spacetime is then described by the de sitter line element @xmath20 where @xmath21 the hubble parameter @xmath22 appearing here is given by @xmath23 .
we will suppose that there is a charge @xmath2 localized on the boundary of this core .
the energy in the neighborhood of the core is dominated by field gradients .
this boundary layer can be modeled as a relativistic domain wall with a surface energy density ( tension ) @xmath24 , @xcite . outside the core
, the energy density in the massive fields falls off exponentially fast so that , to a good approximation , the energy in matter is dominated by the asymptotic magnetic coulomb field .
the spherically symmetric exterior spacetime can then be modeled as a region of the reissner - nordstrm geometry described by the line element @xmath25 where @xmath26 here @xmath3 is the conserved adm mass which represents the combined material and gravitational binding energy of the configuration .
@xmath3 must be positive .
appearing here is related to the magnetic charge of the monopole @xmath27 by @xmath28 where @xmath29 and @xmath30 is the gauge coupling strength . ] in this model , we attempt to capture the dynamics of the bubble wall in a single variable , the radius @xmath31 of the core boundary or wall .
following ref.@xcite , it is straightforward to cast the einstein equations at the wall in the form @xmath32 where we define @xmath33 , and the overdot represents a derivative with respect to proper time .
equation ( [ eq = dbeta ] ) can be exploited to express both @xmath34 and @xmath35 as functions of the wall radius : @xmath36\,,\label{eq = betadm}\ ] ] where we have rescaled variables as follows : @xmath37 now eq . (
[ eq = dbeta ] ) can be recast as @xmath38 where the overdot represents a derivative with respect to proper time rescaled by @xmath22 .
the potential @xmath39 appearing here can be expressed in either of two equivalent forms @xmath40 the einstein equations determine the local geometry in the neighborhood of the wall .
the sign of the functions @xmath41 encodes the boundary conditions required to construct the complete global geometry .
finally , we note that , in terms of the symmetry - breaking scale @xmath1 , the ratio @xmath0 is given by @xmath42 here , we have exploited the fact that @xmath43 and @xmath44 with constants of proportionality @xmath45 and @xmath46 of order unity . for a gut symmetry - breaking scale , @xmath47gev , @xmath48 . for planck scale @xmath1 , @xmath49 .
the potential @xmath50 given by eq.([eq = u ] ) is parametrized by three positive dimensionless parameters characterizing the mass , the charge , and the symmetry breaking scale @xmath51 , @xmath52 and @xmath0 respectively .
we will consider sections of constant @xmath0 and of constant @xmath2 of this three dimensional space . because both @xmath51 and @xmath52 have @xmath1 folded into their definition , when we vary @xmath0 , it is appropriate to undo the ` natural ' scaling depending on @xmath1 one exploits for calculational purposes . in general , the potential is always negative .
in addition , @xmath53 at @xmath54 and as @xmath55 so that it always possesses at least one maximum . to discuss the qualitative dependence of the potential on the values of @xmath3 , @xmath2 and @xmath0 , it is useful to identify the following boundaries on the parameter space : 1 . the location of the extremal exterior reissner - nordstrm geometry , @xmath56 : @xmath57 + if @xmath58 the complete reissner - nordstrm geometry possesses an ( outer ) event horizon at @xmath59 and an ( inner ) cauchy horizon at @xmath60 which are given by the two positive solutions of @xmath61 where @xmath62 is given by eq.([am ] ) .
when @xmath57 , the two horizons possess the same radius ( this does not mean that they coalesce ) . if , however , @xmath5 there are no horizons and the corresponding spacetime possesses a naked singularity at @xmath63 .
this criterion is independent of @xmath0 .
+ this boundary will play an important role in determining the limit of stability of a self - gravitating object .
+ we refer the reader to the @xmath64 and @xmath65 parameter planes represented in fig . [ regmq ] and fig . [ regmkambos ]
respectively . as a visual aid ,
the former is reproduced zoomed - in as fig .
[ regmqzi ] and zoomed - out as fig .
[ regmqzo ] . the corresponding potential in different regions of parameter space is plotted in fig . [ npot ] .
the lower bound on the mass providing a potential with a well , @xmath66 : + suppose now that we fix @xmath2 and @xmath0 .
consider the dependence of the potential on @xmath3 .
below some fixed value @xmath66 , @xmath50 possesses a single maximum ; there is no well . above @xmath66
, @xmath50 possesses a well : with minimum @xmath67 ( say ) , and maxima @xmath68 and @xmath69 on its left and its right respectively .
we note also that @xmath68 ( and never @xmath69 ) is always the absolute maximum of @xmath50 .
lowering @xmath51 through @xmath66 at a fixed values of @xmath2 and @xmath0 we find that @xmath67 and @xmath69 ( not @xmath68 ) coalesce when @xmath70 .
the value @xmath66 increases monotonically from zero ( as a function of both @xmath2 and @xmath0 ) . on the parameter plane
have been discussed elsewhere by two of the authors in the context of global monopoles and will be omitted here .
see ref . @xcite .
] + this completes the discussion of the topological form of the potential , as characterized by its critical points .
this topology is not , however , always relevant physically
. this will be the case if the well is not accessible physically .
+ the domains of @xmath71 which are physically accessible in the potential are determined by the mass shell condition eq.([masss ] ) . to locate these domains we examine where the critical points of the potential lie with respect to the fixed ` energy ' @xmath72 .
again we fix @xmath2 and @xmath0 .
these conditions will identify three values of @xmath3 .
3 . the upper limit on monotonic motion @xmath73 : +
if @xmath3 is below some value @xmath73 the absolute maximum of the potential will lie below the value @xmath72 .
all values of @xmath31 are then accessible and all candidate physical trajectories necessarily monotonic either expanding from zero radius or collapsing to it . , @xmath74 and an unstable equilibrium with the wall poised precariously at @xmath68 is , of course , possible . ] if @xmath75 there are no monotonic trajectories . for each such @xmath3
there are always at least two trajectories , each with a single turning point , one bounded and another unbounded .
when we refer to them below we will describe the trajectory initially at rest at the turning points : the former collapses from a finite maximum to zero radius ; the latter expands from a minimum to infinite radius .
+ whether the trajectories we have described translate into configurations which are compatible with the boundary conditions is a question which we address in the following section .
the einstein equations , as we will see , _ do _ admit spurious solutions which do not correspond to the isolated lump of energy we are interested in .
+ the value @xmath73 like @xmath66 increases monotonically from zero as a function of both @xmath2 and @xmath0 .
4 . the limits of oscillatory motion , @xmath12 and @xmath13 : + the analogue in our model of a radially deformed monopole corresponds to an oscillatory trajectory .
these are the only trajectories which should survive when gravity is turned off .
when is such motion possible ? to accomodate a stable oscillating trajectory in the potential , the well must be accessible on shell , @xmath76 , and confine the motion on the right , @xmath77 .
clearly these conditions will not be realized for every specification of @xmath2 and @xmath0 .
when they are they will limit @xmath3 to values within a finite band @xmath78 $ ] .
the values @xmath12 at which @xmath79 and @xmath13 at which @xmath80 are indicated in fig . [ regmq ] ) .
these two boundaries in the three dimensional parameter space coalesce on the boundary @xmath66 along some critical curve @xmath81 where they terminate . for fixed @xmath0 ,
we denote the limiting value of the charge on @xmath81 by @xmath82 .
we have plotted @xmath83 as a function of @xmath0 in fig .
[ qtfinal ] . note that @xmath83 decreases monotonically to zero as @xmath84 .
in particular , the relationship @xmath85 is invertible for the corresponding limiting value of @xmath0 at fixed @xmath2 , @xmath86 .
+ even without consulting spacetime diagrams , it is already possible to conclude the following : + for a given @xmath0 there exists a unique ` static ' trajectory for each @xmath2 up a limiting value , @xmath82 ; and that for a given @xmath2 , there is a corresponding limiting value of @xmath0 , @xmath87 .
as we will see when we examine the corresponding spacetimes , not all ` static ' trajectories correspond to static spacetimes so that the physical limiting values will be lower .
+ there exists , at best , a finite spectrum @xmath78 $ ] bounded below by @xmath12 , of stable oscillations about any static configuration .
finally , we comment that the boundary structure on parameter space is captured completely by either of the two sections we have considered .
the @xmath64 section contracts continuously towards the unique fixed point , @xmath88 , @xmath6 as @xmath0 is raised .
its topological structure is unchanged . in the following section
, we will consider the embedding of the wall trajectories in both the interior de sitter and the exterior reissner - nordstrm spacetime .
in the present context , de sitter space is represented most conveniently by a gibbons - hawking diagram . for details , in the present context the reader is referred to ref.@xcite . in this diagram
the center is placed at the ( north ) pole of a round sphere .
the evolution of this point is represented by the trajectory indicated @xmath63 on the left hand side of the spacetime diagram .
the diagonal running from the upper right to the lower left represents the cosmological horizon of this point .
the core interior is represented by the spacetime region to the left of the trajectory on this diagram .
it is clear that turning points of the motion of the wall must occur within the static regions i and iii with @xmath89 where the killing vector @xmath90 is timelike , and @xmath91 spacelike . in particular , oscillatory solutions
are necessarily confined to these regions ( one should not rule out , a priori , an oscillating core boundary in region i with an inflating interior ) .
a globally static core must , however , lie in the left hand quadrant iii .
any trajectory which crosses the horizon necessarily inflates inside .
see fig .
[ rnwh ] .
the nature of the reissner - nordstrm spacetime depends crucially on the charge to mass ratio , @xmath92 .
if @xmath5 there are no horizons in the reissner - nordstrm geometry and a penrose conformal diagram of its maximal analytic extension consists of a single asymptotically flat globally static spacetime with a naked timelike singularity at @xmath63 .
see fig.[rnwh ] . in our analysis , the reissner - nordstrm geometry will always be truncated at some finite radius within which it is replaced by a patch of de sitter space .
if this radius does not fall to zero , the singularity does not appear in the physical spacetime and there is no physical justification to limit ourselves to values of @xmath3 exceeding @xmath2 as one does in vacuum .
the maximal analytic extension of the reissner - nordstrm geometry when @xmath58 is represented on the penrose - carter(pc ) diagram , fig .
see ref.@xcite and also ref.@xcite for a recent pedagogical discussion .
this consists of an infinite tower of identical connected universes .
the singularities at @xmath63 are not visible at infinity in this geometry . within the regions
@xmath93 and @xmath94 , the killing vector @xmath95 is timelike : both of these regions are static . in the inter horizon region , @xmath96 , @xmath95 becomes spacelike and @xmath91 generates temporal evolution .
the spacetime in this region is dynamical no matter how one cares to look at it .
as in the interior de sitter space , any turning points of the motion must occur in the static regions
. this will be useful to remember when locating turning points in spacetime .
we remark that the cauchy horizon is unstable @xcite . under a small generic perturbation in the metric
, it has been shown to collapse into a schwarzschild type spacelike singularity limiting motion towards the future .
a consequence is that the exotic possibilities evoked by the reissner - nordstrm tower are irrelevant .
the life span of the physical system is limited to just one floor on this tower .
the exterior is represented by the spacetime region to the right of the trajectory on this spacetime diagram .
the topology of a regular spatial slice is @xmath97 with a disk removed .
it can be shown that the fugacities @xmath41 are proportional to the derivative of the corresponding coordinate time with respect to proper time , @xmath98 in the case of de sitter space and the @xmath58 reissner - nordstrm geometry , the right hand side of eq.([fugptime ] ) in turn relates these two functions to the course of the polar angle @xmath99 subtended by the trajectory about a fixed point in the corresponding spacetime diagram which permits the routing of the trajectory about this point to be determined . for @xmath34 one
has @xmath100 we note that @xmath101 indicates clockwise ( counter - clockwise ) motion about the origin .
unlike the de sitter geometry , the penrose - carter diagram for reissner - norsdtrm geometry with @xmath58 , possesses neither preferred origin , nor unique corresponding polar angle on the spacetime diagram .
we consider the routing of the motion about the bifurcation points of @xmath59 and @xmath102 .
we find @xmath103 where @xmath104 are the corresponding angles .
the interpretatinon of @xmath35 is different on the reissner - norsdtrm spacetime with @xmath5 . due to the absence of horizons
, @xmath35 necessarily possesses a fixed sign .
it is easily checked that @xmath35 must be positive for an isolated monopole with an infinite exterior .
a negative @xmath35 in this case corresponds to a finite exterior geometry with a naked singularity .
we are now in a position to describe the wall motion in spacetime which corresponds to any given set of parameters .
we begin with a discussion of trajectories which correspond to the intuitive notion of a monopole as a stable compact lump of energy .
as we have seen , such solutions must lie in the ` oscillatory ' regime of parameter space admitting bounded radial motion , with mass bounded below by @xmath12 and above by @xmath13 .
the boundary @xmath57 provides a natural partition of this region .
indeed , we note that for low values of @xmath2 , @xmath105 , with equality along @xmath106 .
this value is strictly lower than @xmath82 .
the boundary @xmath13 , on the other hand , lies strictly above @xmath2 except along @xmath107 where the two touch ( with a common tangent ) .
these two features are clearly indicated on the zoom - in of the @xmath108 parameter plane . in fig .
[ qtfinal ] we plot @xmath10 and @xmath14 as functions of @xmath0 .
they clearly converge as @xmath0 becomes large .
@xmath109 , @xmath110 and @xmath83 partition the oscillatory regime into three regions which we label _ s , qsi _ , and _ qsii _ on fig .
[ regmq ] .
the oscillatory motion compatible with each of these regions is different . _
s _ : there exist stable oscillating trajectories with both static interior and exterior : in the interior , @xmath111 along the trajectory so that it lies in region iii of a gibbons - hawking diagram the interior does not inflate ; the exterior reissner - nordstrm geometry with @xmath5 is globally static , the trajectory is indicated @xmath112 in fig .
[ rnwh ] . _ qsi , ii _ : stable oscillating trajectories would also appear to be admitted in these neighboring regions of parameter space .
however , whereas the interior geometry in both is essentially identical to that of an _ s _ trajectory , the exterior geometry necessarily contains a black hole .
if a genuine static trajectory with @xmath58 exists , it must do so along that section of the boundary @xmath12 where @xmath113 , which occurs within _ qsii_. because @xmath31 is constant , it must lie entirely within one of the static regions with @xmath93 or @xmath94 . outside this domain , @xmath114 is a timelike coordinate and a constant value of @xmath114 defines an impossible spacelike trajectory . if @xmath115 , the static trajectory lies within a black hole . only if @xmath116 ( with no horizon ) is the exterior spacetime geometry globally static , so that we can speak of a genuinely static solution .
there are , however , no solutions of this form : within _ qsii _ the turning points @xmath117 and @xmath118 of oscillatory motion both lie below @xmath102 .
in fact , the possibility @xmath119 , while consistent with the spacetime geometry , never occurs . stable static monopoles ( and stable radial oscillations about them ) , appear always to configure themselves so that @xmath5 .
for a given @xmath2 there exists an upper bound on @xmath1 admitting such a solution ( as for a given @xmath1 there exists an upper bound on @xmath2 ) , determined by the crossing of @xmath12 and @xmath57 , strictly below the ` naive ' bound on the ` oscillatory ' regime discussed in sec .
the existence of the limit on @xmath1 was predicted within the simplified zero surface tension model by tachizawa , maeda and torii in ref .
the existence of this limit was also noted by sakai in @xcite .
this value of @xmath1 signals the onset of topological inflation .
consider , now , the fate of an oscillating trajectory as @xmath3 is raised through @xmath56 from some initial value in _ s _ maintaining @xmath2 and @xmath0 constant . because @xmath120 in this regime , the trajectory must undergo finite oscillation in @xmath31 ( there are no static trajectories ) .
the surplus @xmath3 provides the wall with radial kinetic energy . as @xmath56 is crossed , two horizons with initially equal radii appear in the exterior geometry . where the turning points of the oscillatory motion , @xmath117 and @xmath118 say , lie with respect to the horizons at @xmath59 and @xmath102 will depend on the values of @xmath2 and @xmath0 . if @xmath121 , _ qsi _ is entered with @xmath122 and @xmath123 ; whereas if @xmath2 lies between @xmath124 and @xmath125 , _ qsii _ is entered with both @xmath117 and @xmath118 less than @xmath102 .
and @xmath0 within the range @xmath126 $ ] which can not be considered as excitations of any _ s _ static configuration . ]
when @xmath127 , @xmath118 coincides with the right maximum of the potential and @xmath128 .
the exterior spacetimes which correspond to ` oscillatory ' trajectories @xmath129 in _ qsi _ and @xmath130 in _ qsii _ are illustrated in figs.[rnh ] and [ rnh2 ] , respectively .
the ` oscillatory ' trajectory @xmath129 interpolates between a maximum in region i and a minimum in region @xmath131 .
its apparent subsequent oscillatory course up through the penrose - carter tower is an analytical accident without any observable consequences .
the physical solution clearly does not oscillate coming as it will to the unpleasant end described in the previous section as it crosses the cauchy horizon .
the exterior geometry collapses in a black hole .
the trajectory @xmath130 oscillate within region @xmath131 .
the exterior geometry again collapses in a black hole isolating the monopole inside .
the gauge monopole analogues of both solutions were observed numerically by sakai in @xcite .
their zero tension analogue was identified in ref .
@xcite by tachizawa , maeda and torii .
finally , we note that the penrose singularity theorem places no classical obstruction on the formation of any of the solutions we have described from non singular initial conditions , be they static or black hole.@xcite
the oscillatory solution in _ s _ described above is the only solution of the einstein equations satisfying the boundary conditions which corresponds to an isolated monopole in the parameter regime @xmath5 .
technically , this is because @xmath132 along the remaining trajectories , be they monotonic , collapsing or expanding bounces .
this is just as well : while gravity might be sufficiently strong to provoke the collapse of a charged object , one would not expect this to happen if the charge exceeds @xmath3 ; nor would one expect gravity to promote the explosion of a monopole .
a negative @xmath35 corresponds , in the regime @xmath5 , to an exterior which is a finite region of the reissner - nordstrm geometry with an unphysical naked singularity at the antipode .
the spatial geometry is a closed three sphere which is inconsistent with the asymptotically flat boundary conditions that we associate with an isolated monopole . because it contains a naked singularity we consider it unphysical .
an immediate corollary of the above observation is the existence of a lower bound on the mass of a physically realistic configuration , static or otherwise : ( i ) if @xmath133 , so that stable static solutions exist , this value is @xmath12@xcite ; ( ii ) if @xmath134 and there do not , a bound is provided by @xmath2 . the later bound will be sharpened below . for a constant @xmath0 , the mass of a static solution @xmath135 which has the same functional form as the minkowski space limit .
in sec . [ ssfk ] , bounce solutions were identified in the parameter regime bounded below by @xmath73 .
such solutions coexist with the quasi - static solutions we have described in each of _ qsi _ and _
qsii_. we again discard the collapsing solution as an unphysical closed universe with a naked singularity .
however , the expanding bounce trajectories are consistent with the boundary conditions .
the regime admitting bounces partitions naturally into three regions : _ qsi _ ( as before ) , @xmath137 ( which contains _ qsii _ ) indicated on fig .
[ regmq ] , and @xmath138 indicated on fig .
[ regmqzi ] .
the interior spacetime of an expanding bounce clearly inflates .
the trajectories are embedded on the reissner - nordstrm spacetime as @xmath139 on fig .
[ rnh ] for _ qsi _ ; @xmath140 on fig .
[ rnh2 ] for _ b _ ; and @xmath141 on fig .
[ rnh3 ] . for @xmath138 .
the expansion in all cases occurs behind a throat geometry which subsequently collapses into a reissner - nordstrm black hole ( in the same way as it does outside the oscillatory counterparts discussed in sec .
[ bwtst ] ) this expansion does not occur at the expense of the exterior geometry but ( with respect to a reasonable slicing of spacetime ) does get cut off from the exterior by the formation of a black hole .
qualitatively , the bounce occuring in _ qsi _ is very similar to the false vacuum bubble bounces described in ref.@xcite .
note that the penrose singularity theorem places an obstruction to its formation from non singular initial conditions .
accessible or not classically , this trajectory is of interest because of the possibility of tunneling into it from its bounded counterpart , @xcite .
the bounce occuring in _ b _ is very different , contracting from infinity in one asymptotically flat region of the reissner - nordstrm spacetime to a minimum on the left hand side of the penrose - carter tower before expanding to the corresponding asymptotically flat region on the next floor of the tower .
clearly , the full bounce is not a physically realizable configuration .
the physically relevant leg of any bounce is its expansion from a stationary minimum .
bounces correspond either to the thermodynamical or quantum mechanical materialization of a configuration .
interestingly , there is a narrow window in the neighborhood of this stationary initial configuration where the penrose theorem does not present any obstruction to the classical assembly of the _ b _ bounce from non - singular initial conditions .
these are the analogues of topologically inflating gauge monopoles . in his numerical simulations of the dynamics of gauge monopoles ,
sakai also observed inflating monopoles ( corresponding to our bounces ) to co - exist with collapsing monoples ( corresponding our @xmath129 ) .
bounces occuring in the narrow regime of parameter space indicated @xmath138 occur in a convex potential . whereas the asymptotics of such bounces are identical to those for _
b _ , their minimum occurs now on the right of the penrose - carter tower .
in contrast to _ b _ bounces , the penrose theorem implies that its formation is unphysical on the complete expanding leg . on its contracting leg , there is no asymptotically flat spatial slice containing the trajectory .
we must conclude that such trajectories are unphysical .
we have already discounted monotonic trajectories with @xmath5 . the boundary @xmath66 partitions the remainder of this regime . in the bounded regime @xmath142 the potential is convex and both @xmath34 and @xmath35 possess definite signs .
we have indicated the trajectory by @xmath143 on fig .
[ rnh4 ] . in the remaining unbounded region with @xmath144 , the effective potential develops a well and both @xmath34 and @xmath35 change sign in the course of their evolution .
apart from this single dynamical detail , motion is qualitatively identical in both regimes .
is this motion physical from a classical point of view ?
the part of the trajectory lying within @xmath115 necessarily contains a naked singularity in its exterior ; moreover , the interior initially contains a three - sphere s worth of de sitter space .
the trajectory is clearly unphysical in this regime .
in fact , the penrose theorem forbids the assembly of such a trajectory by classical means@xcite .
we dismiss this solution as unphysical .
it would appear that there are no physical monotonic trajectories in this model . if we take in account the elimination as unphysical of all possible trajectories in both @xmath143 and @xmath138 , the lower bound on the mass of any asymptotically flat configuration is raised .
as @xmath2 increases above @xmath109 , the lower bound follows the line @xmath57 , then @xmath145 , and finally @xmath146 .
we are finally in a position to examine the limit @xmath147 , where the model had better reproduce the `` false vacuum bubble '' investigated by blau , guendelman and guth and others .
briefly , for each value of @xmath3 below some critical mass @xmath148 , both collapsing and expanding bounce motion occur as we described in our introduction . for masses above @xmath148 all motion is monotonic : the core expands from a singular zero radius behind the schwarzschild horizon and like the bounce described in the introduction is connected to the asymptotically flat region by a throat .
though the throat collapses , the core expands forever with an inflating interior .
the reader is referred to @xcite for details .
both the expanding bounce and the monotonic solution violate the penrose theorem along their course @xcite .
this limit should be consistent with solutions lying on the @xmath3-axis on the @xmath64 section . at first sight
, however , the limit @xmath147 of our model appears to contradict their findings .
specifically , there does not appear to be any analog of @xmath148 at @xmath6 the monotonic trajectories we find do not even exist in this regime . to resolve this apparent contradiction , note
that , as @xmath147 in this regime , the left maximum of the potential @xmath50 occurs at ever decreasing radius ( @xmath149 ) while , simultaneously , the well depth becomes infinitely deep , @xmath150 .
we also note that , as @xmath147 , the inner horizon of the reissner - nordstrm geometry appoaches zero , @xmath151 , while the outer horizon at @xmath152 becomes the schwarzchild horizon .
the bounce trajectory occuring in @xmath140 thus approaches arbitrarily close to @xmath153 , the penrose window we discussed in sec .
[ bouncy ] closes and the trajectory on its expanding leg becomes indistinguishable from a monotonically growing false vacuum bubble .
it is clear that we should identify @xmath148 with @xmath13 at @xmath6 , not with @xmath73 .
we also note that below @xmath13 , in _ qsi _ the a quasi - oscillatory trajectory approaches @xmath54 arbitrarily closely and become indistinguishable from a collapsing bounce .
the expanding bounce , as we commented earlier does not suffer any signifacant local change .
we have examined in some detail the dynamics of a charged false vacuum bubble within the thin wall approximation .
we claim that , with the identification of @xmath2 with the magnetic charge @xmath27 ( related to the electric charge by @xmath154 ) the model mimics the radial dynamics of a spherically symmetric magnetic monopole .
in particular , the model provides a valuable guide to understanding the physics which underlies both the onset of instability of a static monopole as well as the conditions which need to be met to produce a topologically inflating object .
it would appear that inflation does not necessarily require dialling up the symmetry - breaking scale @xmath1 ; an inflating solution only requires that the adm mass be sufficiently large , which is possible in principle for arbitrarily low values of @xmath1 or @xmath2 . in this respect ,
the monopole we consider differs from the ` global ' monopole discussed in @xcite where inflation is only possible when @xmath1 is raised above the planck scale .
however , it should also be pointed out that in a field theory of monopoles the mass is itself a function of @xmath1 .
it is not clear if the high mass and low @xmath1 inflating solutions we find can be realized in practice .
we note that , for every monopole which collapses into a black hole in the parameter regimes _ qsi _ and _ qsii _ , there will be a corresponding expanding bounce configuration with identical values of the conserved mass and charge . on semiclassical grounds
, one would anticipate a finite amplitude for tunneling from the former to the latter .
the construction of the instanton mediating this passage should provide a valuable exercise in semi - classical quantum gravity .
we have considered a description of a spherically symmetric field theoretical monopole in which its core boundary is modeled as a relativistic membrane .
how robust is this description when spherical symmetry is relaxed ? in the seventies it was shown that , in dirac s extensible model of the electron , the static charged membrane is unstable to non - radial deformations@xcite .
the origin of this instability is similar to that which triggers fission of the atomic nucleus ( the boundary conditions differ ) . on
the other hand , in ref.@xcite goldhaber argued that a global monopole suffers from a cylindrical string - like instability .
superficially , this would appear to be analogous to the spike ( zero area ) instability of a nambu - goto membrane .
however , it is likely that higher curvature ( rigidity ) corrections to the nambu - goto action must be included to model the field theory when spherical symmetry is relaxed .
such additions would tend to moderate ( or eliminate ) the instabilities of nambu - goto membranes .
it would be interesting to explore the membrane - topological defect correspondence in greater detail .
thanks to gilberto tavares for technical assistence .
g.a . was supported by a conacyt graduate fellowship .
was supported in part by the institute of cosmology at tufts university .
the work of j.g . has received support from dgapa at unam , conacyt proyect 32307e and a conacyt - nsf collaboration . | the stability of a spherically symmetric self - gravitating magnetic monopole is examined in the thin wall approximation : modeling the interior false vacuum as a region of de sitter space ; the exterior as an asymptotically flat region of the reissner - nordstrm geometry ; and the boundary separating the two as a charged domain wall .
there remains only to determine how the wall gets embedded in these two geometries . in this approximation , the ratio @xmath0 of the false vacuum to surface energy densities is a measure of the symmetry breaking scale @xmath1 .
solutions are characterized by this ratio , the charge on the wall @xmath2 , and the value of the conserved total energy @xmath3 .
we find that for each fixed @xmath0 and @xmath2 up to some critical value , there exists a unique globally static solution , with @xmath4 ; any stable radial excitation has @xmath3 bounded above by @xmath2 , the value assumed in an extremal reissner - nordstrm geometry and these are the only solutions with @xmath5 . as @xmath3 is raised above @xmath2 a black hole forms in the exterior : ( i ) for low @xmath2 or @xmath0 , the wall is crushed ; ( ii ) for higher values , it oscillates inside the black hole . if the mass is not too high these ` collapsing ' solutions co - exist with an inflating bounce ; ( iii ) for @xmath0 , @xmath2 or @xmath3 outside the above regimes , there is a unique inflating solution . in case ( i )
the course of the bounce lies within a single asymptotically flat region ( afr ) and it resembles closely the bounce exhibited by a false vacuum bubble ( with @xmath6 ) . in cases ( ii ) and ( iii ) the course of the bounce spans two consecutive afrs .
however , from the point of view of either region it resembles a monotonic false vacuum bubble . 2 |
recently , the development of dilute gas bose - einstein condensation ( bec ) [ 1,2,3 ] has opened up the study of atom laser [ 4 - 7 ] , which is the matter wave analogs of optical lasers .
the ideal atom laser beam is a single frequency de broglie wave with well - defined intensity and phase .
the first realization of a pulsed atom laser was achieved with sodium atoms at mit by coupling a bec from a magnetically trapped state to an untrapped state using a rf pulse [ 8 ] .
then , it was repeated with long rf pulses [ 9 ] and raman transitions [ 10 ] .
+ in recent years , much attention has been focused on the problem of nonlinear atomic optics .
the four matter - wave mixing was realized in the remarkable experiment [ 11 ] by making use of the optical technique of bragg diffraction to the condensate .
the possibility of optical control on the quantum statistics of the output matter wave was pointed out in the framework of nonlinear atomic optical optics [ 12 ] .
several dynamical analysis concerning the nonclassical properties of the output of the trapped condensed atoms have been presented in refs.[13,14,15 ] .
+ on the other hand , there has been increased interest in the problem about decoherence of bec and atom laser [ 16 - 19 ] .
one fact causing decoherence comes from the nonlinear interaction between atoms . for a single - mode condensate
the significant effect of atomic collisions is to cause fluctuations in the energy and hence fluctuations in the frequency , thus causing increased phase uncertainty , eventually lead to decoherence [ 16 ] . in ref.[19 ] , the influence of the decoherence on quantum coherent atomic tunneling between two condensates is studied .
it is shown that the decoherence leads to the decay of the population difference and the suppression of the coherent atomic tunneling . in this paper
, we would like to investigate the influence of decoherence on the nonclassical properties of the output of a bose - einstein condensate by adopting milburn s model of intrinsic decoherence [ 20 ] .
the influence of decoherence on the nonclassical properties , such as sub - poisson distribution and quadrature squeezing of the atom laser beam is investigated .
it is shown that , under very special conditions , the atom laser beam may exhibit stationary quadrature squeezing in such a decoherence model .
+ the paper is organized as follows : in sec.ii we briefly outline the simple model describing the output coupling of the trapped dilute condensed atoms , in which the nonlinear interaction between the atoms and the quantized motion of atom center of mass in the inhomogeneous magnetic field has been ignored .
then , the quantum dynamical behavior of this system in the intrinsic decoherence model is discussed by making use of bogoliubov approximation . in sec.iii ,
the influence of decoherence on the nonclassical properties , such as sub - poisson distribution , quadrature squeezing effect is investigated . in sec.iv , there are some discussions .
in this section , we consider the output coupling of the trapped dilute condensed atoms . for simplicity , the atoms are assumed to have two states , @xmath0 and @xmath1 , with the initial condensation occurring in the trapped state @xmath0 .
state @xmath1 , which is typically unconfined by the magnetic trap , is coupled to @xmath0 by a one - mode squeezed optical field tuned near resonance with the @xmath2 transition .
the hamiltonian of this system is given as [ 14,15 ] ( @xmath3 ) @xmath4 where @xmath5(@xmath6 ) and @xmath7(@xmath8 ) are the annihilation ( creation ) operators of bosonic atoms for the untrapped state @xmath1 and the trapped state @xmath0 with transition frequency @xmath9 , respectively .
@xmath10(@xmath11 ) are the annihilation ( creation ) operators of the optical field with frequency @xmath12 . here
@xmath13 is the coupling constant , and @xmath14 is the effective mode volume and @xmath15 the vacuum permittivity . in the system ( 1 ) , the atom - atom coupling and the nonlinear interaction between the atoms and the quantization motion of atomic center of mass in the trapped state by an inhomogeneous magnetic field has been ignored . it was shown that the above system leads to an oscillation behavior of the quantum statistics between the optical field and the output atomic laser beam [ 15 ] . + in what follows , we outline the basic content of the milburn model of decoherence . based on an assumption that on sufficiently short time steps the quantum system does not evolves continuously under unitary evolution but rather in a stochastic sequence of identical unitary transformations , milburn has derived the equation for the time evolution density operator @xmath16 of the quantum system [ 20 ] , @xmath17 , \eqno{(2)}\ ] ] where @xmath18 is the mean frequency of the unitary time step . in the limit @xmath19 ,
eq.(2 ) reduces to the ordinary von neuman equation for the density operator .
it is easy to obtain the formal solution of eq.(2 ) as follows , @xmath20 where the kraus operator @xmath21 is given by @xmath22 obviously , @xmath23 . + we assume that the initial state of system ( 1 ) is described by @xmath24 . here ,
@xmath25 with @xmath26 the glauber coherent state of the operator @xmath7 characterizing the condensed atoms in the trapped state @xmath0 ; @xmath27 , @xmath28 represents that there is initially no occupying atoms in the untrapped state @xmath1 , and the optical field is in the squeezed vacuum state @xmath29 .
@xmath30 is the squeezed operator . +
substituting the @xmath31 into eq.(3 ) , we obtain @xmath32 where @xmath33 , \eqno{(6)}\ ] ] @xmath34 @xmath35 if @xmath36 , we can adopt the bogoliubov approximation [ 21 ] , i.e. , neglect @xmath37 in eq.(5 ) .
then in the bogoliubov approximation , the density operator @xmath16 can be reexpressed as follows , @xmath38 where @xmath39 is the reduced density operator describing the subsystem of the untrapped atoms and optical field .
+ now , we confine ourselves in the resonance case , i.e. , @xmath40 .
we define the operators @xmath41 ( @xmath42 ) and @xmath43 ( @xmath44 ) as @xmath45 @xmath46 @xmath47 @xmath48 it is easy to obtain that @xmath49 @xmath50 @xmath51 @xmath52 where @xmath53 and @xmath54 . by making use of eq.(12 ) , we can express average values of arbitrary operator functionals @xmath55 as following @xmath56=e^{-\gamma{t}}\sum^{\infty}_{k=0}\frac{(\gamma{t})^k}{k!}\textrm{tr}[g(a(k),a^{\dagger}(k);b(k),b^{\dagger}(k))|0\rangle_{ff}\langle0|\otimes|\xi\rangle\langle\xi| ] .
\eqno{(13)}\ ] ]
in this paper , we investigate nonclassical properties of the output of a bose - einstein condensate in milburn s model of intrinsic decoherence by making use of bogoliubov approximation .
it is shown that the squeezing properties of the atom laser field can be destroyed by the decoherence under most conditions . however , under some very special conditions , the squeezing properties of atom laser is robust against the decoherence .
this phenomenon is highly dependent on the particular modelling of decoherence .
+ in what follows , we briefly discuss the relevance of our theoretical results to the realistic experimental conditions . by making use of the preliminary atom laser experiments [ 8 ] , as a guide to realistic parameter values , we choose @xmath122khz , @xmath123khz , the initial trapped atom number @xmath124 , the mean frequency of the unitary time step @xmath125mhz , and @xmath126 as a example to illustrate our theoretical results . in such a condition ,
the rabi - like oscillation behaviors of the numbers of both the output atom and output photon become frozen after @xmath127 .
however , the atom laser persists in exhibiting oscillation of squeezing coefficient and its squeezing property is finally destroyed by decoherence till about @xmath128 .
+ milburn s model of intrinsic decoherence is currently a very active field of research .
nevertheless , so far little attention has been paid to its study in the context of ultracold quantum gases . however , due to their high sensitivity to decoherence , such systems could provide an interesting testing ground for the milburn model . in this paper , we provide a first step in this direction by studying the influence of intrinsic decoherence on the output properties of a simple atom laser model .
a thorough discussion of all other sources of decoherence that could be present in an experiment and comparison with the results presented in this work should be interesting and necessary in the future study .
it is also very interesting to investigate the entanglement between output atoms and output photons and discuss the influence of intrinsic decoherence on entanglement [ 25 - 28 ] .
moreover , it is worth to discuss the possible quantum chaotic dynamical behavior of atom laser and the quantum - classical corresponding by fully taking account of the atom - atom interaction .
this project was supported by the national natural science foundation of china ( project no .
10174066 ) .
99 m. h. anderson _
et al . _ ,
science * 269 * , 198 ( 1995 ) .
k. b. davis _
et al . _ ,
rev . lett . * 75 * , 3969 ( 1995 ) . c. c. bradley _
et al . _ ,
lett . * 75 * , 1687 ( 1995 ) .
m. olshanii , y. castin , and j. dalibard , in _ proceedings of the xii conference on lasers spectroscopy _ , edited by m. inguscio , m. allegrini , and a. sasso ( world scientific , singapore , 1995 ) .
r. j. c. spreeuw , t. pfau , u. janicke , and m. wilkens , europhys . lett . * 32 * , 469 ( 1995 ) .
h. m. wiseman and m. j. collett , phys .
a * 202 * , 246 ( 1995 ) .
a. m. guzman , m. moore , and p. meystre , phys .
rev . a * 53 * , 977 ( 1996 ) .
lett . * 78 * , 582 ( 1997 ) .
i. bloch , t. w. hnsch , and t. esslinger , phys .
* 82 * , 3008 ( 1999 ) .
e. w. hagley _
et al . _ , science * 283 * , 1706 ( 1999 ) .
et al . _ , nature * 398 * , 218 ( 1999 ) .
m. g. moore , o. zobay and p. meystre , e - print cond - mat/9902293 .
n. m. bogoliubov , _ et al .
_ , phys . rev
. lett . * 86 * , 4439 ( 2001 ) . c. sun , h. zhan , y. miao , j. li , commun .
* 29 * , 161 ( 1998 ) .
h. jing , j .- l . chen , m .-
ge , phys .
a * 63 * , 015601 ( 2001 ) .
l. k. thomsen and h. m. wiseman , phys .
a * 65 * , 063607 ( 2002 ) .
i. bouchoule and k. mlmer , phys .
a * 67 * , 011603(r ) ( 2003 ) .
l. pitaevskii and s. stringari , phys .
lett . * 87 * , 180402 ( 2001 ) .
et al . _ ,
a * 61 * , 013608 ( 1999 ) .
g. j. milburn , phys .
a * 44 * , 5401 ( 1991 ) .
f. dalfovo _
phys . * 71 * , 463 ( 1999 ) .
d. f. walls and g. j. milburn , _ quantum optics _ , ( springer - verlag , berlin , 1994 ) .
v. buzek , a. v. barranco and p. l. knight , phys .
a * 45 * , 6570 ( 1992 ) .
z. p. karkuszewski , c. jarzynski , and w. h. zurek , phys .
* 89 * , 170405 ( 2002 ) .
li , j .- b .
xu , phys .
a * 311 * , 313 ( 2003 ) .
li , j .- b .
xu , phys .
lett . a * 313 * , 175 ( 2003 ) .
b . li , j .- b .
xu , chin .
. lett . * 20 * , 985 ( 2003 ) .
li , s .- b .
li and j .- b .
xu , chin .
* 21 * , 774 ( 2004 ) .
: : the mandel q parameter @xmath73 of optical field is plotted as a function of time @xmath75 with @xmath129 and @xmath130 for two different values of the parameter @xmath18 , ( solid line ) @xmath131 , ( dot line ) @xmath132 .
: : the mandel q parameter @xmath74 of atom laser field is plotted as a function of time @xmath75 with @xmath129 and @xmath130 for two different values of the parameter @xmath18 , ( solid line ) @xmath131 , ( dot line ) @xmath132 .
: : the quadrature squeezing coefficient @xmath94 of the atom laser field is plotted as a function of time @xmath75 with @xmath133 , @xmath134 , @xmath135 , @xmath136 and @xmath137 for three different values of the parameter @xmath18 , ( a ) @xmath131 , ( b ) @xmath138 , ( c ) @xmath132 .
: : the quadrature squeezing coefficient @xmath94 of the atom laser field is plotted as a function of time @xmath75 with @xmath139 , @xmath134 , @xmath135 , @xmath140 , @xmath137 and @xmath132 .
: : the quadrature squeezing coefficient @xmath94 of the atom laser field is plotted as a function of time @xmath75 with @xmath139 , @xmath134 , @xmath135 , @xmath141 and @xmath132 for four different values of @xmath97 : ( solid line ) @xmath140 , ( dot line ) @xmath142 , ( dash dot line ) @xmath143 , ( dash dot dot line ) @xmath144 .
+ of the atom laser field is plotted as a function of time @xmath75 with @xmath133 , @xmath134 , @xmath135 , @xmath136 and @xmath137 for three different values of the parameter @xmath18 , ( a ) @xmath131 , ( b ) @xmath138 , ( c ) @xmath132 .
[ fig3 ] ] of the atom laser field is plotted as a function of time @xmath75 with @xmath139 , @xmath134 , @xmath135 , @xmath141 and @xmath132 for four different values of @xmath97 : ( solid line ) @xmath140 , ( dot line ) @xmath142 , ( dash dot line ) @xmath143 , ( dash dot dot line ) @xmath144 . [ fig5 ] ] | we investigate nonclassical properties of the output of a bose - einstein condensate in milburn s model of intrinsic decoherence .
it is shown that the squeezing property of the atom laser is suppressed due to decoherence . nevertheless ,
if some very special conditions were satisfied , the squeezing properties of atom laser could be robust against the decoherence .
+ pacs numbers : 03.75.gg , 03.75.pp , 03.65.ta + keywords : atom laser ; quadrature squeezing ; decoherence + |
it is usual to consider the famous experiment of fizeau ( 1851)@xcite on the drag of light by a uniformly moving medium as one of the crucial experiments which , just as the michelson - morley experiment , can not be correctly understood without profound modification of newtonian space - time concepts ( for a review of einstein s relativity as well as a discussion of several experiments the reader is invited to consult @xcite ) .
the result of this experiment which was predicted by fresnel@xcite , in the context of elastic theory , is indeed completely justified by well known arguments due to von laue ( 1907 ) @xcite .
he deduced the fresnel - fizeau result for the light velocity @xmath0 in a medium , corresponding to a relativistic first order expansion of the einstein velocity transformation formula : @xmath1 here , @xmath2 represents the optical index of refraction of the dielectric medium in its proper frame , and we suppose that the uniform medium motion with velocity @xmath3 is parallel to the path of the light and oriented in the same direction of propagation . in the context of electromagnetic theory @xcite
all derivations of this effect are finally based on the invariance property of the wave operator @xmath4 $ ] in a lorentz transformation .
it is easy to write the wave equation @xmath5\psi=0 $ ] in the co - moving frame r@xmath6 of the medium in covariant form@xcite @xmath7\psi=0 $ ] which is valid in all inertial frames and which for a plane waves , implies the result of eq .
in this calculation we obtain the result @xmath8 if we use the galilean transformation which proves the insufficiency of newtonian dynamics .
+ however , the question of the physical meaning of this phenomenon is not completely clear .
this fact is in part due to the existence of a derivation made by lorentz ( 1895 ) @xcite based on the mixing between the macroscopic maxwell s equations and a microscopic electronic oscillator model which is classical in the sense of the newtonian dynamics . in his derivation lorentz did not use the relativistic transformation between the two coordinate frames : laboratory and moving medium .
consequently , the relativistic nature of the reasoning does not appear explicitly .
following the point of view of einstein ( 1915 ) @xcite the lorentz demonstration must contain an implicit hypothesis of relativistic nature , however , this point has not been studied in the literature .
recent developments in optics of moving media @xcite allows us to consider this question as an important one to understand the relation between optics , relativity and newtonian dynamics .
this constitutes the subject of the present paper . here , we want to analyze the physical origin of the fresnel - fizeau effect .
in particular we want to show that this phenomenon is , in its major part , independent of relativistic dynamics . + the paper is organized as follows . in section
ii we present the generalized lorentz `` microscopic - macroscopic '' derivation of the fresnel formula and the principal defect of this treatment . in section iii
we show how to derive the fresnel result in a perturbation approach based on the lorentz oscillator model and finally in iv we justify this effect independently from all physical assumptions concerning the electronic structure of matter .
in this part , we are going to describe the essential contents of the lorentz model and of its relativistic extension .
let @xmath9 be the displacement of an electron from its equilibrium position at rest , written as an explicit function of the atomic position @xmath10 and of the time @xmath11 . in the continuum approximation we can write the equation of motion for the oscillator as @xmath12 where the supposed harmonic electric incident field appears and where the assumption of small velocity allows us to neglect the magnetic force term . in the case of a non relativistic uniformly moving medium we have @xmath13 which includes the magnetic field @xmath14 of the plane wave and the associated force due to the uniform motion with velocity @xmath15 .
the equation of propagation of the electromagnetic wave in the moving medium has an elementary solution when the velocity of the light and of the medium are parallel . if we refer to a cartesian frame @xmath16 , @xmath17 we have in this case @xmath18 , @xmath19 for the electromagnetic field and @xmath20 for the displacement vector parallel to the @xmath21 axis .
the relativistic extension of this model can be obtained directly putting @xmath22 in eq .
[ deux ] or [ trois ] and using a lorentz transformation between the moving frame and the laboratory one .
we deduce the displacement @xmath23 where @xmath24 .
we could alternatively obtain the same result considering the generalization of the newton dynamics i. e. by doing the substitutions @xmath25 and @xmath26 in eq .
the dispersion relation is then completely fixed by the maxwell equation @xmath27 , where the current density @xmath28 is given by the formula @xmath29 depending on the local number of atoms per unit volume @xmath30 supposed to be constant . using @xmath28 and eq .
[ trois ] or [ quatre ] we obtain a dispersion relation @xmath31 where the effective refractive index @xmath32 depends on the angular frequency @xmath33 and on the velocity @xmath3 . the more general index obtained using eq .
[ quatre ] is defined by the implicit relation @xmath34 [ 1-\frac{n\left(\omega\right)v_{e}}{c}]^{2}\label{cinq}.\ ] ] here @xmath35 , and @xmath36 is the classical lorentz index ( also called drude index ) which contains the local proper density which is defined in the frame where the medium is immobile by @xmath37 .
these relativistic equations imply directly the correct relativistic formula for the velocity of light in the medium : writing @xmath38 we deduce @xmath39 which can be easily transformed into @xmath40 it can be added that by combining these expressions we deduce the explicit formula @xmath41 } { [ 1+\frac{n_{0}\left(\omega'\right)v_{e}}{c}]^{2}}.\ ] ] the non relativistic case can be obtained directly from eq .
[ trois ] or by writting @xmath42 in eqs .
[ cinq],[six ] . this limit @xmath43 + o\left(\frac{v_{e}^{2}}{c^{2}}\right)\ ] ] is the fresnel - fizeau formula corrected by a frequency - dispersion " term due to lorentz@xcite .
for our purpose , it is important to note that in the non - relativistic limit of eq .
[ cinq ] we can always write the equality @xmath44 where @xmath45 is the index of refraction defined relatively to the moving medium .
we then can see directly that the association of maxwell s equation with newtonian dynamics implies a modification of the intuitive assumption @xmath46 " used in the old theory of emission .
in fact , the problem can be understood in the newtonian mechanics using the absolute time @xmath47 and the transformation @xmath48 . in the laboratory
frame the speed of light , which in vacuum is @xmath49 , becomes @xmath50 in a medium at rest . in the moving frame
the speed of light in vacuum is now @xmath51@xcite .
however , due to invariance of acceleration and resultant force in a galilean transformation we can interpret the presence of the magnetic term in eq . [ deux ] as a correction to the electric field in the moving frame .
this effective electric field affecting the oscillator in the moving frame is then transformed into @xmath52 .
it is this term which essentially implies the existence of the effective optical index @xmath53 and the light speed @xmath54 in the moving frame .
it can be observed that naturally maxwell s equations are not invariant in a galilean transformation .
the interpretation of @xmath52 as an effective electric field is in the context of newtonian dynamics only formal : this field is introduced as an analogy with the case @xmath22 only in order to show that @xmath55 must be different from @xmath2 .
the velocity of the slab is @xmath56 , and the three spatial regions in front , in and after the slab are denoted by 1 , 2 , and 3 , respectively .
we have plotted in addition a typical observation point p(@xmath57,0,0 ) . , title="fig:",width=288 ] +
the difficulty of the preceding model is that the lorentz derivation does not clarify the meaning of the fresnel - fizeau phenomenon .
indeed we justify eq .
[ un ] using a microscopical model which is in perfect agreement with the principle of relativity .
however we observe that at the limit @xmath58 the use of the non relativistic dynamics of newton ( see eq .
[ deux ] ) gives the same result .
more precisely one can see from eq .
[ deux ] that the introduction of the magnetic force @xmath59 in addition to the electric force is already sufficient to account for the fresnel - fizeau effect and this even if the classical force formula @xmath60 is conserved .
since the electromagnetic force contains the ratio @xmath61 and originates from maxwell s equations this is already a term of relativistic nature ( einstein used indeed this fact to modify the dynamical laws of newton@xcite ) .
the derivation of lorentz is then based on newton as well as on einstein dynamics .
it is well know in counterpart that the doppler - fizeau effect , which includes the same factor @xmath62 , can be understood without introducing einstein s relativity .
indeed this effect is just a consequence of the invariance of the phase associated with a plane wave when we apply a galilean transformation ( see @xcite , chap .
11 ) as well as a lorentz transformation .
we must then analyze further in detail the interaction of a plane wave with a moving dipole in order to see if the fresnel phenomenon can be understood independently of the specific lorentz dynamics .
+ we consider in this part a different calculation based on a perturbation method and inspired by a derivation of the optical theorem by feynman@xcite .
consider a thin slab of thickness @xmath63 perpendicular to the @xmath57 axis .
let this slab move along the positive @xmath57 direction with the constant velocity @xmath64 .
let in addition @xmath65 be the incident electric field of a plane wave which pursues the moving slab ( see fig .
[ slab ] ) .
therefore , the electric field after the slab can be formally written as @xmath66 where @xmath67 appear as a retardation time produced by the interaction of light with the slab and where all reflections are neglected ( @xmath68 ) . for a `` motionless '' slab ( i. e. , the case considered by feynman ) we can write the travel time of the light through the slab as @xmath69 and therefore @xmath70 where @xmath71 defines the proper length of the slab in the frame where it is at rest . for the general case of a moving slab of reduced length @xmath72 we find for the travel time : @xmath73 and
therefore the perturbation time is @xmath74 we can obtain this result more rigorously by using maxwell s boundary conditions at the two moving interfaces separating the matter of the slab and the air ( see appendix a ) . in order to evaluate the diffracted field which is @xmath75 we can limit our calculation to a first order approximation .
thereby , each dipole of the lorentz model as discussed above can be considered as being excited directly by the incident electromagnetic wave and where we can neglect all phenomena implying multiple interactions between light and matter . in this limit
[ caput ] reduces to @xmath76 if the distance between the slab and an observation point is much larger than @xmath63 we can consider the slab as a 2d continuous distribution of radiating point dipoles .
the vector potential @xmath77 radiated by a relativistically moving point charge @xmath78 is in according with the lienard - wiechert s formula given by @xcite : @xmath79 here @xmath80 is the distance separating the observation point @xmath81 ( denoted by p ) and the point charge position located at @xmath82 at the time @xmath11 ; additionally @xmath83 is the velocity of the point charge and @xmath84 is the unit vector @xmath85 . in this formula , in agreement with causality , all point charge variables are evaluated at the retarded time @xmath86 . + in the present case the motion of the point charge can be decomposed into a uniform longitudinal component @xmath87 oriented along the positive @xmath57 direction and into a transversal oscillating part @xmath88 obeying the condition @xmath89 . owing to this condition we can identify @xmath90 with @xmath91 .
consequently , in the far - field the contribution of the electron uniform velocity is cancelled by the similar but opposite contribution associated with the nucleus of the atomic dipole : only the vibrating contribution of the electron survives at a long distance from the diffraction source .
if we add the contribution of each dipole of the slab acting on the observation point p at the time @xmath11 we obtain then the total diffracted vector potential @xmath92 produced by the moving medium : @xmath93 here @xmath94 is the radial coordinate in a cylindrical coordinate system using the direction @xmath57 as a revolution axis , and the quantity @xmath95 is the number of dipoles contained in the cylindrical volume of length @xmath63 and of radius varying between @xmath94 and @xmath96 if we consider a local dipole density given by @xmath97 . in this formula the retarded distance @xmath98 is a function of @xmath94 and we have ( see the textbook of jackson @xcite ) @xmath99 this expression shows that the minimum @xmath100 is obtained for a point charge on the @xmath57 axis , and that : @xmath101 in order to evaluate the integral in eq .
[ field ] we must use in addition the following relation ( see appendix c ) : @xmath102 hence , we obtain the following integral : @xmath103 where we have used the relations eq .
[ cacatoes ] , eq .
[ cactus ] in the denominator and in the exponential argument of the right hand side of eq .
[ field ] , respectively .
the diffracted field is therefore directly calculable by using the variable @xmath104 .
we obtain the result @xmath105 the total diffracted electric field @xmath106 is obtained using maxwell s formula @xmath107 , which gives : @xmath108 the final result is given substituting eq .
[ quatre ] in eq .
[ chose ] and implies by comparison with eq .
[ trucmuch]@xmath109 this equation constitutes the explicit limit @xmath110 of eq .
[ cinq ] and implies the correct velocity formula eq .
[ six ] when we neglect terms of @xmath111 $ ] .
it can again be observed that the present calculation can be reproduced in the non relativistic case by neglecting all terms of order @xmath112 .
the central fact in this reasoning is the travel condition " given by eqs .
[ ultime],[ultime2 ] . indeed , of the same order in power of @xmath113 we can deduce the relation @xmath114 and consequently the condition eq . [ ultime ] reads @xmath115 if we call @xmath116 the time during which the energy contained in a plane of light moving in the positive x direction is absorbed by the slab at rest in the laboratory , @xmath117 is evidently the enlarged time for the moving case . during the period where this plane of light is absorbed by the slab its energy moves at the velocity @xmath3
this fact can be directly deduced of the energy and momentum conservation laws .
indeed , let @xmath118 be the momentum of the slab of mass @xmath119 before the collision and @xmath120 the energy of the plane of light , then during the interaction the slab is in a excited state and its energy is now @xmath121 and its momentum @xmath122 .
the velocity of the excited slab is defined by @xmath123 and we can see that in the approximation @xmath124 used here @xmath125 ( we neglect the recoil of the slab ) . during @xmath117
the slab moves along a path length equal to @xmath126 and thus the travel condition of the plane of energy in the moving slab can be written @xmath127 which is an other form for eq .
[ ultime3 ] . now eliminating directly @xmath128 in eq .
[ condition ] give us the velocity @xmath0 of the wave : @xmath129 i. e. @xmath130 which depends on the optical index @xmath131 . after straightforward manipulations
this formula becomes @xmath132 which is the einstein formula containing the fresnel result as the limit behavior for small @xmath3 .
+ it can be observed that this reasoning is even more natural if we think in term of particles .
a photon moving along the axis @xmath57 and pursuing an atom moving at the velocity @xmath3 constitutes a good analogy to understand the fresnel phenomenon .
this analogy is evidently not limited to the special case of the plane wave @xmath133 . if for example we consider a small wave packet which before the interaction with the slab has the form @xmath134 where @xmath135 is a small interval centered on @xmath136 , then after the interaction we must have : @xmath137 where @xmath138 is given by eq .
[ ultime2 ] . after some manipulation
we can write these two wave packets in the usual approximative form : @xmath139 } \nonumber\\ = e^{i\left(k_{m}x-\omega_{m}t\right ) } f\left(t - x / v_{g}\right)\nonumber\end{aligned}\ ] ] @xmath140\right ) } f\left(t - x / v_{g}-\delta t_{g}\right).\end{aligned}\ ] ] here , @xmath141 is the group velocity of the pulse in vacuum and @xmath142 is the perturbation time associated with this group motion .
this equation for @xmath143 possesses the same form as eq .
[ caput ] and then the same analogy which implies eq .
[ ultime3 ] is possible .
this can be seen from the fact that we have @xmath144 with @xmath145 .
we deduce indeed @xmath146 where we have @xmath147 i. e. @xmath148 .
since eq . [ phase ] and eq .
[ groupe ] have the same form the fresnel law must be true for the group velocity . + it is important to remark that all this reasoning conserves its validity if we put @xmath42 and if we think only in the context of newtonian dynamics .
since the reasoning with the travel time does not explicitly use the structure of the medium involved ( and no more the magnetic force @xmath59 ) it must be very general and applicable in other topics of physics concerning for example elasticity or sound .
+ consider as an illustration the case of a cylindrical wave guide with revolution axis @xmath57 and of constant length @xmath63 pursued by a wave packet of sound .
we suppose that the scalar wave @xmath149 obeys the equation @xmath150\psi=0 $ ] where @xmath49 is the constant sound velocity .
the propagative modes in the cylinder considered at rest in the laboratory are characterized by the classical dispersion relation @xmath151 where the cut off wave vector @xmath152 depend only of the two `` quantum '' numbers @xmath153 and of the cross section area @xmath154 of the guide ( @xmath155 ) .
the group velocity @xmath156 of the wave in the guide is defined by @xmath157 $ ] and the travel time @xmath158 by @xmath159/c$ ] which implies @xmath160 . in the moving case where the cylinder possesses the velocity
@xmath3 we can directly obtain the condition given by eq . [ ultime3 ] ( with @xmath42 ) and then we can deduce the group velocity of the sound in the guide with the formula @xmath161 this last equation give us the fresnel result if we put the effective sound index @xmath162 . we can control the self consistency of this calculation by observing that the dispersion relation eq .
[ dispersion ] allows the definition of a phase index @xmath163 which is equivalent to eq .
[ ultime3 ] when @xmath164 and @xmath165 .
this reveals a perfect analogy between the sound wave propagating in a moving cylinder and the light wave propagating in a moving slab .
it is then not surprising that the fresnel result is correct in the two cases .
+ the principal limitation of our deduction is contained in the assumption expressed above for the slab example : @xmath166 i. e. the condition of no reflection supposing the perturbation on the motion of the wave to be small .
nevertheless , the principal origin of the fresnel effect is justified in our scheme without the use of the einstein relativity principle .
+ we can naturally ask if the simple analogy proposed can not be extended to a dense medium i. e. without the approximation of a weak density @xmath113 or of a low reflectivity . in order to see that it is indeed true we return to the electromagnetic theory and we suppose an infinite moving medium like the one considered in the second section . in the rest frame of the medium we can define a slab of length @xmath71 .
the unique difference with the section 3 is that now this slab is not bounded by two interfaces separating the atoms from the vacuum but is surrounded by a continuous medium having the same properties and moving at the same velocity @xmath3 . in the laboratory frame
the length of the moving slab is @xmath72 .
we can write the time @xmath128 taken by a signal like a wave packet , a wave front or a plane of constant phase to travel through the moving slab : @xmath167the optical index @xmath168 can be the one defined in section 2 for the case of the drude model but the result is very general .
we can now introduce a time @xmath169 such that eq .
[ ultime3 ] , and consequently eq . [ fin ] , are true _
by definition_. we conclude that this last equation eq .
[ fin ] is equivalent to the relativistic eq .
[ relativist ] if , and only if , we define the time @xmath116 by the formula @xmath170 in other terms we can always use the analogy with a photon pursuing an atom since the general formula eq .
[ relativist ] is true whatever the microscopic and electrodynamics model considered . in this
model - based on a retardation effect- the absorbtion time @xmath116 is always given by eq .
[ pursue ] .
+ this opens new perspectives when we consider the problem of a sound wave propagating in an effective moving medium .
indeed there are several situations where we can develop a deep analogy between the propagation of sound and the propagation of light .
this implies that the conclusions obtained for the fresnel effect for light must to a large part be valid for sound as well .
this is in particular true if we consider an effective meta material like the one that is going to be described now : + we consider a system of mirrors as represented in fig . 2a , at rest in the laboratory . a beam of light propagates along the zigzag trajectory @xmath171 .
the length @xmath172 is given by @xmath173 where the distance @xmath71 and @xmath174 are represented on the figure .
the time @xmath175 spent by a particle of light to move along @xmath172 is then @xmath176 .
we can equivalently define an effective optical index @xmath2 such that we have @xmath177 this implies @xmath178we consider now the same problem for a system of mirrors moving with the velocity @xmath3 . in order to be consistent with relativity
we introduce the reduced length @xmath72 .
the beam propagating along the path @xmath171 must pursue the set of mirrors .
we then define the travel time @xmath128 along an elementary path @xmath172 by @xmath179 where @xmath168 is the effective optical index for the moving medium . from this equation
we deduce first @xmath180 and then @xmath181 which finally give us the formula @xmath182 we can again justify the fresnel formula at the limit @xmath183 . .
a ) when the mirrors are at rest in the laboratory the travel time @xmath184 is dependent only on the distances @xmath71 and @xmath174 .
b ) when the mirrors move at the velocity @xmath3 relatively to the laboratory the travel time @xmath185 is affected by the motion and depends on @xmath3 as well as on @xmath174 and @xmath186.,title="fig:",width=288 ] + the simplicity of this model is such that it does not depend on the physical properties of atoms , electrons and photons but only on geometrical parameters .
clearly we can make the same reasoning for a sound wave by putting @xmath42 .
this still gives us the fresnel formula when we neglect terms equal or smaller than @xmath187 .
in addition this model allows us to conclude that the essential element justifying the fresnel - fizeau result is the emergence of a delay time a retardation effect when we consider the propagation of the signal at a microscopic or internal level . the index @xmath168 which characterizes the macroscopic or external approach is then just a way to define an effective velocity without looking for a causal explanation of the retardation . +
the essential message of our analysis is that by taking explicitly into account the physical origin of the delay we can justify the essence of the fresnel - fizeau effect in a non relativistic way .
the fresnel - fizeau effect is then a very general phenomenon .
it is a consequence of the conservation of energy and momentum and of the constant value of the wave velocity in vacuum or in the considered medium .
the so called travel condition ( eq . [ condition ] ) which is a combination of these two points can be compared to the usual demonstration for the doppler effect . in these two cases of light pulses pursuing a moving particle
the perturbation time @xmath188 is a manifestation of the doppler phenomenon .
it should be emphasized that the analogy between sound and electromagnetic waves discussed in this article could be compared to the similarities between sound wave and gravitational waves discussed in particular by unruh . on this subject and
some connected discussions concerning the acoustic aharonov - bohm effect ( that is related to the optical aharonov - bohm effect that follows from the fizeau effect ) the reader should consult @xcite .
we have obtain the fresnel - fizeau formula using a perturbation method based on the optical theorem and in a more general way by considering the physical origin of the refractive index .
the modification of the speed of light in the medium appears then as a result of a retardation effect due to the duration of the interaction or absorbtion of light by the medium , and the fresnel - fizeau effect , as a direct consequence of the medium s flight in front of the light .
these facts rely on the same origin as the doppler - fizeau effect .
we finally have shown that it is not correct to assume , as frequently done in the past , that a coherent and `` newtonian interpretation '' of these phenomena would be impossible .
on the contrary , the results do not invalidate the derivation of the fresnel - fizeau effect based on the principle of relativity but clarify it .
we observe indeed than all reasoning is in perfect agreement with the principle of relativity .
we must emphasize that even if the fizeau / fresnel effect is conceptually divorced from relativity it strongly motivated einstein s work ( more even than the michelson and morley result ) .
the fact that the fizeau as well as the michelson - morley experiment can be justified so easily with special relativity clearly show the advantages of einstein s principle to obtain quickly the correct results .
nevertheless , if we look from a dynamical point of view , as it is the case here , this principle plays a role only for effects of order @xmath189 which however are not necessary to justify the fresnel formula .
maxwell s equations impose the continuity of the electric field on each interface of the slab .
more precisely these boundary relations impose : @xmath190 where @xmath191 is one of the two moving interfaces separating vacuum and matter .
hence we obtain an equality condition between the two phases @xmath192 and @xmath193 valid for all times at the interface .
let @xmath194 be the phase of the plane wave before the slab .
in a similar way let @xmath195 and @xmath196 be the phases in the slab and in vacuum after traversing the slab , respectively . in these expressions there appear two retardation constants , @xmath197 and the optical index of the slab . on the first interface denoted by ( i - ii ) we have @xmath198 and consequently @xmath199 which is valid for each time and possesses the unique solution : @xmath200 considering the second interface ( ii - iii ) in a similar way we obtain the following conditions @xmath201 where the @xmath202 equality is eq . [ ultime2 ] .
and p is the observation point at the time @xmath11 .
the particle moves uniformly on the the @xmath57 line qr following the trajectory @xmath203 and q is the position of the particle at the time @xmath11 separated of p by the distance @xmath204 .
in addition , if we call r the projection of p on qr , then @xmath57 and @xmath94 are the coordinates of the observation point in the plane of the figure.,title="fig:",width=341 ] + using geometrical considerations ( see fig . [ retard ] ) we can deduce the relation @xmath205 which is equivalent after manipulations to the other : @xmath206 we can in a second step rewrite this equality as follows : @xmath207which is eq .
[ cactus ] .
h. fizeau , c. r. acad .
( paris ) * 33 * , 349 ( 1851 ) .
j. d. jackson , _ classical electrodynamics , third edition _ ( j. wiley and sons , new york , 1999 ) .
j. van bladel , _ relativity and engineering _ ( springer - verlag , berlin , 1984 ) .
j. h. smith , _ introduction to special relativity _
( benjamin , new york , 1965 ) .
a. fresnel , ann .
* 9 * , 57 ( 1818 ) . h. von laue , ann .
( leipzig ) * 23 * , 989 ( 1907 ) .
h. minkowsky , nachr .
d. k. ges .
d. wiss .
zu gott . , math .
* 53 * ( 1908 ) .
w. pauli , _ theory of relativity _
( pergamon press , london , 1958 ) .
h. a. lorentz , _ the theory of electrons and its application to the phenomena of light and radiant heat _
, 2nd edn .
, chapter 5 ( b. g. teubner , stutters , 1916 ) .
a. einstein , ber die spezielle und die allgemeine relativittstheorie gemeinverstnlich ( vieweg , brunswick , 1917 ) .
u. leonhardt and p. piwinicki , phys . rev . *
a60 * , 4301 ( 1999 ) ; j. fiurazek , u. leonhardt and r. parentani , phys .
rev . * a65 * , 011802 ( 2002 ) ; u. leonhardt and p. piwinicki , phys .
lett . * 84 * , 822 ( 2000 ) ; u. leonhardt , phys .
rev . * a62 * , 012111 ( 2000 ) ; u. leonhardt and p. ohberg , phys .
rev . * a67 * , 053616 ( 2003 ) .
m. wilkens , phys .
rev . lett .
* 72 * , 5 ( 1994 ) ; h. wei , r. han and x. wei , _ ibid .
_ * 75 * 2071 , ( 1995 ) ; g. spavieri , _ ibid . _
* 82 * , 3932 ( 1999 ) .
i. carusotto , m. artoni , g. c. la rocca , and f. bassani , phys .
rev . lett .
* 86 * , 2549 ( 2001 ) ; m. artoni and i. carusotto , phys .
rev . * a68 * , 011602 ( 2003 ) ; i. carusotto , m. artoni , g. c. la rocca , and f. bassani , phys .
* a68 * , 063819 ( 2003 ) .
d. strekalov , a. b. matsko .
n. yu . and l. maleki , phys .
lett . * 93 * , 023601 ( 2004 ) .
a. einstein , ann .
( leipzig ) * 17 * , 891 ( 1905 ) .
r. p. feynman , r. b. leighton and m. l. sands , _ the feynman lectures on physics _ , chapters 30 - 31 ( addison - wesley , massachusetts , 1963 ) . h. c. van de hulst , _ light scattering by small particles _ , chapter 4 ( john wiley and sons inc .
, new york , 1957 ) .
w. g. unruh , phys .
lett . * 46 * , 1351 ( 1981 ) ; m. visser , class .
grav . * 15 * , 1767 ( 1998 ) . s. e. p. bergliaffa , k. hibberd , m. stone , and m. visser , physica d * 191 * , 121 ( 2004 ) ; p. roux , j. de rosny , m. tanter , and m. fink , phys .
lett . * 79 * , 3170 ( 1997 ) . | we present a new derivation of the fresnel - fizeau formula for the drag of light by a moving medium using a simple perturbation approach .
we focus particulary on the physical origin of the phenomenon and we show that it is very similar to the doppler - fizeau effect . we prove that this effect is , in its essential part , independent of the theory of relativity . the possibility of applications in other domains of physics is considered . |
adaptive synchronization of networked dynamical systems has attracted a growing interest during recent years @xcite .
it is motivated by a broad area of potential applications : formation control , cooperative control , control of power networks , communication networks , production networks , etc .
existing works @xcite and others are dealing with full state feedback and linear interconnections .
the solutions are based on lyapunov functions formed as sum of lyapunov functions for local subsystems . as for adaptive control algorithms
they are based on either local ( decentralized @xcite ) or nearest neighbor ( described by an information graph @xcite ) strategies . despite a great interest in control of network ,
only a restricted class of them is currently solved .
e.g. in existing papers mainly linear models of subsystems are considered @xcite . in nonlinear case
only passive or passifiable systems are studied and control is organized according to information graph , i.e. not completely decentralized @xcite .
availability of the whole state vector for measurement as well as appearance of control in all equations for all nodes is assumed in decentralized stability and synchronization problems @xcite .
powerful passivity based approaches are not developed for adaptive synchronization problems . in this paper
we consider the problem of master - slave ( leader - follower ) synchronization in a network of nonidentical systems in lurie form where system models can be split into linear and nonlinear parts .
case of identical nodes is studied in @xcite .
linearity of interconnections is not assumed ; links between subsystems can also be nonlinear . in the contrary to known works on adaptive synchronization of networks ,
see @xcite , only some output function is available and control appears only in a part of the system equations .
it is also assumed that some plant parameters are unknown.the leader subsystem is assumed to be isolated and the control objective is to approach the trajectory of the leader subsystem by all other ones under conditions of uncertainty .
interconnection functions are assumed to be lipschitz continuous .
the results of @xcite are employed to solve the posed problem .
adaptation algorithm is designed by the speed - gradient method .
it is shown that the control goal is achieved under leader passivity condition , if the interconnection strengths satisfy some inequalities .
the results are illustrated by example of synchronization in network of nonidentical chua curcuits .
we need yakubovich - kalman lemma in following form , see @xcite .
[ yakub_lemma ] let @xmath0 be @xmath1 real matrices and @xmath2 @xmath3 + @xmath4 then the following statements are equivalent : \1 ) there exists matrix @xmath5 such that @xmath6 \2 ) polinomial @xmath7 is hurwitz and following frequency domain conditions hold @xmath8 for all @xmath9 @xmath10 in order to present the main syncronization result of this paper we need to formulate problem statement of decentralized control and theorem @xmath11 from @xcite which can also be derived from theorem @xmath12 from @xcite . consider stands for column vector with components consisting of components of @xmath13 a system @xmath14 consisting of @xmath15 interconnected subsystems @xmath16 dynamics of each being described by the following equation : @xmath17 where @xmath18 state vector , @xmath19 - vector of inputs ( tunable parameters ) of subsystem , @xmath20 @xmath21 - aggregate state and input vectors of system s , @xmath22 @xmath23 vector - function @xmath24 describes local dynamics of subsystem @xmath25 and vectors @xmath26 describe interconnection between subsystems .
let @xmath27 be local goal functions and let the control goal be : @xmath28 for all @xmath29 we assume existence of smooth vector functions @xmath30 such that @xmath31 i.e. @xmath32 decentralized speed - gradient algorithm is introduced as follows : @xmath33 where @xmath34 @xmath35 @xmath36 - matrix . [ th_2_18 ]
_ suppose the following assumptions hold for the system @xmath14 : _ 1 .
functions @xmath24 are continuous in @xmath37 @xmath38 continuously differentiable in @xmath39 and locally bounded in @xmath40 functions @xmath41 are uniformly continuous in second argument for all @xmath42 in bounded set , functions @xmath43 are convex in @xmath44 there exist constant vectors @xmath45 and scalar monotonically increasing functions @xmath46 such that @xmath47 @xmath48 and @xmath49 2 .
functions @xmath50 are continuous and satisfy the following inequalities @xmath51 is hurwitz , @xmath52 @xmath53 @xmath54 is identity matrix .
then system , is globally asymptotically stable in variables @xmath55 all trajectories are bounded on @xmath56 and satisfy .
let the leader subsystem be described by the equation @xmath57 where @xmath58 state , @xmath59 measurement , @xmath60 is control that specified in advance , @xmath61 internal nonlinearity .
let @xmath62 and @xmath63 be known and not depending on the vector of unknown parameters @xmath64 where @xmath65 is known set . consider a network @xmath66 of @xmath15 interconnected subsystems @xmath67 let subsystem @xmath68 be described by the following equation @xmath69 where @xmath70 functions @xmath71 @xmath72 describe interconnections between subsystems .
we assume @xmath73 let matrices @xmath74 and functions @xmath71 @xmath75 depend on the vector of unknown parameters @xmath76 network model can describe , for example , interconnected electrical generators @xcite .
let the control goal be specified as convergence of all subsystems and the leader trajectories : @xmath77 the adaptive synchronization problem is to find a decentralized controller + @xmath78 ensuring the goal for all values of unknown plant parameters .
denote @xmath79 let the main loop of the adaptive system be specified as set of linear tunable local control laws : @xmath80 where @xmath81 are tunable parameters . by applying speed - gradient method
@xcite it is easy to derive the following adaptation law : @xmath82 where @xmath83
@xmath84 matrices , @xmath85 introduce the following definition .
[ g_monot ] _ let @xmath86 function @xmath87 is called g - monotonically decreasing if inequality @xmath88 holds for all @xmath89 . _ _ remark 1 .
_ apparently , for @xmath90 @xmath91-monotonical decrease of the function @xmath92 is equivalent to incremental passivity @xcite of the static system with characteristics @xmath93 definition [ g_monot ] is easily extended to dynamical systems with the state vector @xmath94 input @xmath95 and output @xmath96 it corresponds to existence of a smooth function @xmath97 satisfying an integral inequality @xmath98 the corresponding property can be called incremental @xmath91-passivity by analogy with @xcite .
consider real matrices @xmath99 of size @xmath100 correspondingly and a number @xmath101 such that : @xmath102 denote @xmath103 condition number of matrix @xmath104 where @xmath105 are maximum and minimum eigenvalues of matrix @xmath106 .
for analysis of the system dynamics the following assumptions are made .
a1 ) the functions @xmath71@xmath107 @xmath108 are globally lipschitz : @xmath109 the function @xmath63 is such that the unique existence of solutions of holds . a2)(matching conditions , @xcite ) for each @xmath110 there exist vectors @xmath111 and numbers @xmath112 such that for @xmath29 @xmath113 denote @xmath114 for the case when matrix @xmath115 is hurwitz introduce notation @xmath116 for stability degree of the function s @xmath117 denominator , i.e. @xmath118 where @xmath119 are eigenvalues of @xmath115 . [ th_noident ] _ let @xmath120 matrix @xmath115 be hurwitz and for some @xmath121 the following frequency domain conditions hold : @xmath122 for all @xmath123 then there exist @xmath124 such that relations hold .
_ let for all @xmath110 assumptions a1 , a2 hold , function @xmath63 be @xmath125-monotonically decreasing , and following inequalities hold @xmath126 where @xmath127 , @xmath128 is condition number of matrix @xmath106
. then for all @xmath129 adaptive controller , ensures achievement of the goal and boundedness of functions @xmath130 on @xmath131 for all solutions of the closed - loop system , , , .
_ let s apply lemma [ yakub_lemma ] .
note that in our case @xmath132 i. e. @xmath133 is scalar .
let s choose @xmath134 instead of @xmath135 in .
then statement of the lemma [ yakub_lemma ] and conditions of theorem [ th_noident ] ensure existence of matrix @xmath5 such that @xmath136 now we can conclude that there exists number @xmath101 such that the following is true : @xmath137 denoting @xmath138 introduce auxiliary error subsystems : @xmath139 here we choose @xmath140 same as in .
let us choose following goal functions @xmath141 and apply theorem [ th_2_18 ] .
we need to evaluate the derivative trajectories of @xmath142 along trajectories of isolated ( i.e. without interconnections ) auxiliary subsystems : @xmath143 .
\end{aligned}\ ] ] denote @xmath144 by taking @xmath145 we obtain @xmath146=\\ & z_i{^{\scriptscriptstyle { \rm t}}}h[a_l x_i+b_l\overline{u}+b_l(\psi_0(y_i)-\psi_0(\overline{y}))-a_l x_i - b_l\overline{u}]=\\ & z_i{^{\scriptscriptstyle { \rm t}}}h[a_l z_i+b_l(\psi_0(y_i)-\psi_0(\overline{y}))].\\ \end{aligned}\ ] ] further , for @xmath29 @xmath147 the last inequality holds because @xmath63 is @xmath125-monotonically decreasing .
so @xmath148 taking into account we conclude @xmath149 by taking @xmath150 we ensure that holds for @xmath29 .
other conditions from the first part of theorem [ th_2_18 ] hold , since the right hand side of the system and function @xmath142 are continuous in @xmath151 functions not depending in @xmath152 for any @xmath29 .
convexity condition is valid since the right hand side of is linear in @xmath39 .
the interconnection condition in our case reads : @xmath153 and matrix @xmath154 should be hurwitz ( @xmath52 @xmath155 ) .
for the case @xmath156 we can take @xmath157 and last inequality will be satisfied .
let s consider case @xmath158 for @xmath29 rewrite as follows : @xmath159 for @xmath160 then for @xmath29 evaluate lower bound of the right - hand side of : @xmath161 it is seen that for @xmath29 to ensure it is sufficient to impose an inequality @xmath162 or @xmath163 denote @xmath164 where @xmath165 noting that @xmath166 in can be chosen arbitrarily close to @xmath116 and taking into account we can conclude that @xmath167 the left - hand side of : @xmath168 thus , if following inequality holds then is ensured : @xmath169 introduce matrix @xmath170 as follows @xmath171 such choice of @xmath172 ensures .
note that @xmath172 is symmetric .
if matrix @xmath173 is positive definite then @xmath154 is hurwitz .
diagonal elements of @xmath173 are positive since @xmath174 and @xmath175 by taking into account that @xmath176 and applying gershgorin circle theorem we conclude that @xmath154 is positive definite .
thus , statement of the theorem [ th_noident ] follows from theorem [ th_2_18 ] .
_ remark 2 .
_ the value of @xmath177 can be evaluated by solving lmi by means of one of existing software package .
_ remark 3 .
_ by interconnections graph of network @xmath66 we can consider directed graph which is a pair of two sets : a set of nodes and a set of arcs .
cardinality of a set of nodes is @xmath178 @xmath179-th node is associated with subsystem @xmath68 for any @xmath160 we say that arc from @xmath179-th node to @xmath180-th node belongs to the set of arcs if @xmath181 is not zero function . by weighted in - degree of @xmath179-th node
we define following number : @xmath182 if each nonzero addend from last sum is equal to 1 then introduced definition of weighted in - degree of the node coincides with the definition of in - degree of digraph s node .
thus the inequality can be interpreted as follows : weighted in - degree of each node of interconnections graph must be less than @xmath183
chua circuit is a well known example of simple nonlinear system possessing complex chaotic behavior @xcite .
its trajectories are unstable and it is represented in the lurie form .
let us apply our results to synchronization with leader subsystem in the network of five interconnected nonidentical chua systems .
let @xmath184 and @xmath185 let the leader subsystem be described by the equation @xmath186 where @xmath187 is state vector of the system , @xmath188 is output available for measurement , @xmath189 is scalar control variable , @xmath190 where @xmath191 further , let @xmath192 @xmath193 transfer function @xmath194 it is seen from the nyquist plot of @xmath195 presented on fig .
[ nyquist ] , that first frequency domain inequality of holds .
the second frequency domain inequality of also holds since relative degree of @xmath196 is equal to one and highest coefficient of its numerator is positive .
obviously @xmath63 is @xmath125-monotonically decreasing .
.,width=288,height=172 ] let subsystem @xmath68 for @xmath197 be described by with @xmath198 by choosing @xmath199 and using we obtain matrices @xmath74 for @xmath200 which are not equal , i.e. nodes are nonidentical .
denote @xmath201 let @xmath202 be equal to @xmath203 further , let @xmath204 lipschitz constants of all @xmath205 are equal to @xmath206 it follows from theorem [ th_noident ] that decentralized adaptive control provides synchronization goal if for all @xmath197 inequality @xmath207 holds , i.e. if interconnections are sufficiently weak .
consider following control of leader subsystem @xmath208.$ ] such @xmath189 ensures chaotic behavior of leader subsystem .
let us put @xmath209 where @xmath54 identity matrix , and @xmath210 denote by @xmath211 @xmath212 matrix with element @xmath213 lying in the @xmath179-th row and the @xmath180-th column , @xmath214 and @xmath215 let us choose adaptive control @xmath216 as in and apply theorem [ th_noident ] .
if we take @xmath217 then simulation shows that @xmath218 i.e. synchronization is achieved : all state vectors of nonidentical nodes converge to the state vector of the leader subsystem , see fig .
[ fig_sim]-(b )
. phase portrait of the leader subsystem , @xmath219 found by 40 sec .
simulation are shown on fig .
[ fig_sim ] .
in contrast to a large number of previous results , we obtained synchronization conditions for networks consisting of nonidentical nonlinear systems with incomplete measurement , incomplete control , incomplete information about system parameters and coupling .
the design of the control algorithm providing synchronization property is based on speed - gradient method @xcite , while derivation of synchronizability conditions is based on yakubovich - kalman lemma and result presented in @xcite .
99 j. lu , g. chen , `` a time - varying complex dynamical network model and its controlled synchronization criteria '' , _ ieee trans .
autom.control_ , vol .
50(6 ) , pp . 841 - 846 , 2005 .
j. yao , d. j. hill , z .- h .
guan , h. o. wang , `` synchronization of complex dynamical networks with switching topology via adaptive control '' , in _ proc .
45th ieee conf .
2819 - 2824 , 2006 .
j. zhou , j. lu , j. lu , `` adaptive synchronization of an uncertain complex dynamical network '' , _ ieee trans .
control _ , vol.51(4 ) ,
652 - 656 , 2006 .
w. s. zhong , g. m. dimirovski , j. zhao , `` decentralized synchronization of an uncertain complex dynamical network '' , in _ proc .
2007 amer .
_ , pp . 1437 - 1442 , 2007 .
a. ioannou , `` decentralized adaptive control of interconnected systems '' , _ ieee trans . on autom .
31(4 ) , pp . 310 - 314 , 1986 .
b. m. mirkin , `` adaptive decentralized control with model coordination '' , _ automation and remote control _ , vol .
60(1 ) , pp . 73 - 81 , 1999 .
d. t. gavel , d. d. siljak , `` decentralized adaptive control : structural conditions for stability '' , _ ieee trans .
34(4 ) , pp . 413 - 426 , 1989 . c. wen ,
y. c. soh , `` decentralized model reference adaptive control without restriction on subsystem relative degrees '' , _ ieee trans . on autom .
44(7 ) , pp . 1464 - 1469 , 1999 .
s. jain , f. khorrami , `` decentralized adaptive control of a class of large - scale nonlinear systems '' , _ ieee trans .
control _ ,
42(2 ) , pp . 136 - 154 , 1997 .
p. jiang , `` decentralized and adaptive nonlinear tracking of large - scale systems via output feedback '' , _ ieee trans .
45(11 ) , pp . 2122 - 2128 , 2000 .
a. l. fradkov , _ adaptive control in complex systems _ , moscow : nauka , 1990 ( in russian ) .
a. l. fradkov , i. v. miroshnik , v. o. nikiforov , _ nonlinear and adaptive control of complex systems _ , kluwer academic publishers , dordrecht , 1999 . j. r. fax , r. m. murray , `` information flow and cooperative control of vehicle formations '' , _ ieee trans . on autom .
49(9 ) , pp . 1465 - 1476 , 2004 . c. yoshioka , t. namerikawa , `` observer - based consensus control strategy for multi - agent system with communication time delay '' , in _ proc .
17th ieee intern .
conf . on control applications _ ,
1037 - 1042 , 2008
. n. chopra , m. w. spong , `` output synchronization of nonlinear systems with time delay in communication '' , in _ proc .
45th ieee conf .
4986 - 4992 , 2006
. n. chopra , m. w. spong , `` output synchronization of nonlinear systems with relative degree one '' in _ recent advances in learning and control _ , vol .
371 , springer - verlag , pp .
51 - 64 , 2008 . i. a. dzhunusov and a. l. fradkov , `` adaptive synchronization of a network of interconnected nonlinear lure systems '' , _ automation and remote control _ ,
70(7 ) , pp . 1190 - 1205 , 2009
. v. a. yakubovich , g. a. leonov , a. kh .
gelig _ stability of stationary sets in control systems with discontinuous nonlinearities _
, singapore : world scientific , 2004 . j. l. willems , `` a partial stability approach to the problem of transient power system stability '' , _ int .
j. of control _ , vol .
19(1 ) , pp . 1 - 14 , 1974 .
a. l. fradkov , `` passification of nonsquare linear systems and feedback yakubovich - kalman - popov lemma '' , europ .
j. of contr .
9(6 ) , pp . 577 - 586 , 2003 .
a. pavlov , l. marconi , `` incremental passivity and output regulation '' , systems and control letters , vol .
57(5 ) , pp . 400 - 409 , 2008 .
e. skafidas , a. l. fradkov , r. j. evans , i. m. mareels , `` trajectory approximation based adaptive control for nonlinear systems under matching conditions '' , _ automatica _ , vol .
34(3 ) , pp . 287 - 299 , 1998 . c. w. wu ,
l. o. chua , `` synchronization in an array of linearly coupled dynamical systems '' , _ ieee trans .
circuits and systems - i . _ vol .
42(8 ) , pp . 430 - 447 , 1995 . | for a network of interconnected nonlinear dynamical systems an adaptive leader - follower output feedback synchronization problem is considered .
the proposed structure of decentralized controller and adaptation algorithm is based on speed - gradient and passivity .
sufficient conditions of synchronization for nonidentical nodes are established .
an example of synchronization of the network of nonidentical chua systems is analyzed .
the main contribution of the paper is adaptive controller design and analysis under conditions of incomplete measurements , incomplete control and uncertainty . |
the combination of cosmic magnetic fields and charged particles , both of which are ubiquitous in the universe , results in the emission of synchrotron radiation from radio sources ( e.g. * ? ? ? * ) .
this radiation is fundamentally linearly polarized , and both the fractional polarization and electric vector polarization angle ( evpa ) of this emission show significant frequency - dependence ( e.g. * ? ? ?
this frequency - dependence is due to faraday rotation and depolarization which , for extragalactic sources , are typically considered to be caused by magnetoionic material that either intervenes between us and the observed emitting region , or is intermixed with the emitting region itself .
understanding of faraday rotation and astrophysical depolarization requires broadband radio measurements , and a large number of facilities are currently available or planned that will have the necessary bandwidth to study cosmic magnetic fields . for example , the australian square kilometre array pathfinder ( askap ) will observe at frequencies between 700 mhz and 1.8 ghz , @xcite , the australia telescope compact array ( atca ) at @xmath1 ghz ( e.g. * ? ? ?
* ) , the giant metrewave radio telescope ( gmrt ) at frequencies @xmath21.4 ghz ( e.g. * ? ? ?
* ) , the galfa continuum transit survey ( galfacts ) with arecibo between 1.2 and 1.5 ghz @xcite , the karl g. jansky very large array ( jvla ) at @xmath31.2 ghz ( e.g. * ? ? ?
* ) , the low - frequency array ( lofar ) at @xmath2230 mhz , and the murchison widefield array ( mwa ) between 80 and 300 mhz @xcite . many of these facilities are pathfinders towards the square kilometre array ( ska ) cosmic magnetism project @xcite
. the ska will detect up to @xmath4 polarized extragalactic sources on the sky at a mean spacing of @xmath5 arcsec @xcite , and provide broadband measurements of both the polarized fraction and faraday rotation .
the polarimetric measurements that result will be used to construct a densely - sampled ` rotation measure grid ' @xcite that allow for analysis and reconstruction of the galactic magnetic field , for investigation of the evolution of magnetic fields over cosmic time ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and can reveal physical properties of the central engines in radio sources ( e.g. * ? ? ?
making sense of such data will require two capabilities : fast cross - matching algorithms for identification of counterpart sources with known redshifts at complementary wavelengths , and more importantly ` @xmath6-corrections ' that will allow for the polarization properties to be determined in the emitting frame of a source .
these @xmath6-corrections will be necessary for the polarized fractions , the rotation measures , and possibly also for faraday rotators that have a non - linear relationship between the evpa and @xmath7 . in order to @xmath6-correct the polarized fraction to the emitting frame
, we will require well - defined polarized spectral energy distributions ( seds ) .
such seds can also be used to investigate the predominant causes of depolarization , and to classify individual sources . as continuous broadband polarization data for a large number of sources
will not be available until the next generation of radio telescopes , we can instead attempt to construct seds from existing data , to enable @xmath6-corrections of polarized sources .
polarized seds can be reconstructed using the considerable amount of archival polarimetric radio data , but require careful consideration of numerous systematics including resolution effects , beam depolarization , multiple source components , time - variability , rician bias , differing uncertainties , outliers , and the use of either an interferometer or a single dish for data collection .
the polarized fractions and rotation measures of 37,543 sources have been determined previously at 1.4 ghz using the nrao vla sky survey ( nvss ) @xcite .
polarized fractions and evpas have also been presented in a large number of other radio catalogs .
in addition , surveys that provide total intensity data allow for measurement of the spectral index and any spectral curvature .
furthermore , the nvss rotation measures ( rms ) have recently been matched against optical catalogs , providing spectroscopic redshifts for more than 4,000 polarized radio sources @xcite .
this conglomeration of data not only provides the evpas , polarized fractions , and redshifts needed to construct and then @xmath6-correct a polarized sed , but also accumulates a large number of complementary parameters that are relevant to efforts for understanding cosmic magnetism . in this paper , we cross - correlate the aforementioned data using a k - dimensional tree ( as described by * ? ? ?
such a technique is able to rapidly eliminate large numbers of sources from a list of possible cross - matches .
consequently , the algorithm allows for a computationally inexpensive nearest - neighbor search reducing the problem from @xmath8 to @xmath9 . by accumulating data obtained at different frequencies ,
the algorithm allows us to construct a polarized radio sed with measurements between 0.4 ghz to 100 ghz .
we then present a catalog of model fits to the polarized seds .
the catalog also contains narrow and broadband rms , total intensity spectral indices , estimates of the depolarization , and spectroscopic redshifts .
we carefully detail the systematics and limitations of the catalog , including resolution effects , and the effect of multiple source components .
this catalog constitutes the most comprehensive database of linear radio polarization currently publicly available , and increases the number of well - defined polarized seds by over an order of magnitude . in this paper , we present the catalog of measurements of compact polarized radio sources , and define the possible uses and limits of our sample . thorough scientific exploitation of this catalog is beyond the scope of this paper and will be the subject of future studies . throughout this paper , we characterize faraday rotation as a rotation of the observed evpa as a function of wavelength , so that @xmath10 where @xmath11 is the observing wavelength , @xmath12 and @xmath13 are the measured and intrinsic evpa respectively , and the factor of proportionality rm , the rotation measure , is the gradient of the evpa with @xmath7 . a more generalised quantity to parameterize faraday rotation is the fundamental physical quantity , the faraday depth . as a special case , the faraday depth at which all polarized emission is produced is equal to the rm if there is only one emitting source along the line of sight , which has no internal faraday rotation , and is not affected by beam depolarization , and there are only faraday screens along the sight line . in such a case @xmath14 where @xmath15 is generally
the electron number density of the plasma and @xmath16 is the magnetic field strength .
the constants @xmath17 , @xmath18 , and @xmath19 are the electronic charge , the mass of the electron , and the speed of electromagnetic radiation in a vacuum respectively .
the integral is performed along the line of sight from the source ( at distance @xmath20 ) to the observer .
the paper is structured as follows : section [ accumulating ] details the collection and preparation of the data used from various radio facilities . the cross - matching and fitting of each sed are explained in sections [ crossmatching ] and [ sourcefitting ] respectively .
the catalog itself is presented in section [ catalog ] , and possible systematic effects on the catalog are considered in section [ systematics ] .
the results are presented in section [ results ] .
possible future applications and a general discussion are presented in section [ discussion ] .
the total intensity spectral index , @xmath0 , is defined such that @xmath21 , where @xmath22 is the radio flux density and @xmath23 is the observing frequency .
the polarized seds are fit using a number of applicable models , which include a polarization spectral index ; @xmath24 is defined such that @xmath25 , where @xmath26 is the polarized fraction and @xmath11 is the observing wavelength . note that @xmath24 is defined in the opposite sense to the total intensity spectral index , @xmath0 , in that it is the exponent of observing frequency rather than wavelength .
we refer to ` polarization ' on multiple occasions , in all cases we are referring to linear radio polarization
circular polarization is beyond the scope of this work .
we have collected data from throughout the literature at many different observing frequencies .
the various parameters of the data incorporated into the catalog are displayed in table [ tab : surveyparameters ] , along with the abbreviations with which we will refer to them .
many of these data are available in an electronic format , while for others we manually converted them into a digitally accessible form .
ccccccc nvss rm & 37543 & 45@xmath27 & 1400 & yes & no & 1 + nvss & @xmath28 & 45@xmath27 & 1400 & not used & no & 2 + h12 & 4003 & 45@xmath27 & 1400 & not used & yes & 3 + ti80 & 1510 & various ( coarsest of @xmath29 ) & 400100000 & yes & not used & 4 + at20 g & 5890 & @xmath30 & 4860 & yes & no & 5 + & & @xmath31 & 8640 & yes & no & + & & 10.8@xmath27 & 20000 & yes & no & + z99 & 154 & 3@xmath32 & 4700 & yes & no & 6 + t03 & 185 & 7@xmath27 & 1400 & yes & no & 7 + & & 4.5@xmath27 & 2496 & yes & no & + & & 2@xmath27 & 4800 & yes & no & + & & 1@xmath27 & 8640 & yes & no & + b3-vla & 192 & 261@xmath27 & 2695 & yes & no & 8 + & & 147@xmath27 & 4850 & yes & no & + & & 69@xmath27 & 10500 & yes & no & + sn80 & 103 & 6.5@xmath32 & 1580 & yes & no & 9 + & & 6.0@xmath32 & 1720 & yes & no & + sn81 & 185 & 6.5@xmath32 & 1580 & yes & no & 10 + & & 5.9@xmath32 & 1760 & yes & no & + & & 0.7@xmath32 & 14750 & yes & no & + sn82 & 68 & 2.1@xmath32 & 10500 & yes & no & 11 + gb6 & 75162 & 3.6@xmath32 & 4850 & no & no & 12 + wenss & 229420 & 54@xmath33 & 326 & no & no & 13 + texas & 66841 & @xmath34 & 365 & no & no & 14 + north6 cm & 53522 & 3.5@xmath32 & 4850 & no & no & 15 [ tab : surveyparameters ] in all of the various catalogs , sources flagged as extended or unreliable in the original data were removed .
many source fluxes and polarized fractions in the literature were stated as an upper limit based on the detection threshold of a particular set of observations
these measurements were also excluded . on occasion , some data displayed unusual values , with a flux density of zero janskys , negative values for positive - definite quantities , and polarized fractions of @xmath35% such values were also removed .
furthermore , the evpas were defined using differing conventions , with values between either @xmath36 to @xmath37 or between @xmath38 to @xmath39 .
all input data were adjusted so that the evpa in the final catalog is consistently defined between @xmath36 to @xmath40 from north through east .
the coordinates of each input catalog were all converted to equatorial j2000 ( in decimal degrees ) . note that not all polarization catalogs also contain the corresponding total intensity data .
the final catalog contains data measured with a variety of different instruments , including that obtained during many large - scale surveys .
the nvss @xcite was a 1.4 ghz survey with the very large array ( vla ) covering the entire sky north of @xmath41 declination at a resolution of 45 arcsec .
the rms brightness fluctuations are approximately uniform across the sky at @xmath42 mjy beam@xmath43 in stokes @xmath44 and @xmath45 mjy beam@xmath43 in stokes @xmath46 and @xmath47 .
the astrometry is accurate to within @xmath48 arcsec for point sources with flux densities @xmath49 mjy , and to @xmath50 arcsec for the faintest ( @xmath51 mjy ) detectable sources .
the survey has a completeness limit of 2.5 mjy , and resulted in a catalog of over 1.8 million discrete sources in stokes @xmath44 .
@xcite reanalyzed the nvss in order to derive rms towards the radio sources with signal - to - noise ( s / n ) greater than @xmath52 in polarized intensity , leading to rms for 37,543 radio sources an average density of approximately one rm per square degree .
the rms are based on just two narrow adjacent frequency bands , resulting in the possibility of significant uncertainty in each measurement .
an attempt to overcome ambiguities in the rm was made using the depolarization of each radio source .
the overwhelming majority of these sources are presumed to be extragalactic .
we now detail other previous observations that we have included into our catalog .
we complement our catalog with measurements from the b3-vla sample , which consists of 1,049 radio sources in five flux - limited subsamples stronger than 0.1 jy at 408 mhz .
follow - up measurements with the effelsberg 100 m telescope extended the frequency range to 151 mhz through to 10.5 ghz , which allowed detailed spectral studies of a sample exceeding 1,000 sources in total intensity .
for the 192 sources with detected polarization at 10.5 ghz , additional measurements were taken at 2.695 ghz and 4.85 ghz .
we also include data from the at20 g survey , which was a blind 20 ghz survey of the entire southern sky at galactic latitudes , @xmath53@xmath54 , carried out using the australia telescope compact array ( atca ) @xcite .
the at20 g consisted of 5,890 sources above a flux density limit of @xmath55 mjy , and included near - simultaneous observations at 4.8 ghz and 8.6 ghz for most sources south of @xmath56 in declination .
the at20 g includes linear polarization measurements at all frequencies .
our catalog also includes additional linear polarization measurements obtained at 4.7 ghz for 154 extragalactic radio sources using the effelsberg 100 m telescope , at 1.4 ghz , 2.5 ghz , 4.8 ghz , and 8.6 ghz for 202 sources using the atca @xcite , between 1.58 ghz and 1.76 ghz and at 14.75 ghz for 141 sources , and at 1.58 ghz and 1.72 ghz for 91 sources using the 100 m telescope at the max - planck - institut fr radioastronomie , and at 10.5 ghz for 68 sources using the algonquin radio observatory ( aro ) 46 m telescope .
we also include the accumulated data for 1,510 radio sources published prior to 1978 .
although containing no linear polarization information , there are also a significant number of total intensity catalogs .
we include data into our catalog from the gb6 survey at 4.85 ghz , which was made with the nrao seven - beam receiver on the green bank 91 m telescope and detected 75,162 discrete sources @xcite .
we also include the north6 cm survey at 4.85 ghz , which detected 53,522 sources @xcite .
furthermore , we include data at lower frequencies : from the westerbork northern sky survey ( wenss ) at 326 mhz which detected 11,299 sources to a limiting flux density of @xmath57 mjy , and from the texas interferometer at the university of texas radio astronomy observatory which detected 66,841 sources at 365 mhz and is 80% complete at @xmath58 jy @xcite .
efforts have previously been made to identify complementary sources at other wavelengths in total intensity ( e.g. * ? ? ?
* ; * ? ? ?
the nvss has also been used to identify optical counterparts and spectroscopic redshifts for 4,003 linearly polarized radio sources , using various resources including the sloan digital sky survey ( sdss ) @xcite .
the varying sky coverage , observational frequencies , and other systematics have complicated the interpretation of these previous catalogs , as it is not particularly clear which source population is being studied .
a sample selected at a single observational frequency , and with as uniform sky coverage as possible , is therefore beneficial in minimizing the selection effects that are intrinsically abundant when producing such a catalog .
as shall be detailed in section [ crossmatching ] , our catalog constitutes a polarized 1.4 ghz flux - limited sample , based on the nvss .
cross - matching is the process of identifying counterpart sources that are coincident in position on the sky .
this process is typically used for total intensity measurements , but we seek to expand the method for cases with a corresponding polarization measurement .
we wish to use such a technique to group measurements at different observing frequencies that correspond to the same physical source .
various techniques have been previously implemented to cross - match radio catalogs , and any such method must provide quantitative limits on the probability of false catalog associations .
previously used techniques have included nearest - neighbor searches ( e.g. * ? ? ?
* ; * ? ? ?
* ) and bayesian approaches ( e.g. * ? ? ?
such techniques constitute a form of coincidence - assessment that can be used to determine the probability of finding false matches , i.e. the false - detection rate ( fdr ) .
a cross - matching algorithm needs to be computationally efficient .
we are limited here by our largest data set , the nvss , which consists of 1.8 million sources across 75% of the sky .
we therefore implement the use of a k - dimensional tree @xcite .
such an algorithm is a computationally inexpensive nearest - neighbor search reducing the problem from @xmath8 to @xmath9 and returns all of the matches within a pre - defined spatial radius around a reference source .
the k - dimensional tree is thereby able to rapidly eliminate large numbers of sources from a list of possible cross - matches .
it is necessary to choose a _ reference catalog _ to which source associations can be concatenated .
we define this reference catalog to be the nvss rm data at 1.4 ghz @xcite . each _ secondary catalog _
was then cross - matched with the reference catalog .
there are two possible outcomes to this process : ( i ) a cross - match exists and the relevant secondary catalog data are attached to the corresponding nvss rm source , or ( ii ) a cross - match does not exist and the secondary catalog data are discarded for that nvss rm source . in this manner , we produce a flux - limited sample that is selected on the basis of being polarized at 1.4 ghz .
further constraints are also required in order to control the fdr , as the various secondary catalogs all have different resolutions , sensitivities , and unevenly - spaced coverage on the sky .
the fdr is further affected by the combined astrometric errors of both the reference and secondary catalogs , which limit the accuracy of any given source position .
the various sensitivities and sky coverage will only affect the completeness of our final catalog . on the other hand
, the differing resolutions can introduce a number of systematics ( see section [ systematics ] for further details on catalog systematics and limitations ) .
however , we only consider sources that are unresolved at all wavelengths : ensuring that any measured beam depolarization is not affected by the differing parameter - space of each input catalog .
the presence of beam depolarization , or otherwise , is one of the many physical inferences that may be drawn from a resulting catalog .
furthermore , while the effects of multiple source components or internal source structure could likely only be overcome by a broadband and high - resolution all - sky polarization survey , such as that to be performed by askap or the ska other measures of a source ( such as the total intensity spectral index and optical counterpart properties ) can also allow for such effects to be studied . in this way
, differing resolutions do not affect a resulting catalog ; each source being consistently unresolved is rather an aspect that will be useful for future investigations .
these effects are therefore beyond the scope of this current paper .
importantly however , the differing resolutions could lead to potentially cross - matching independent sources that are unresolved in a given individual catalog thereby increasing the fdr .
we therefore need a matching criterion that defines a suitable astrometric matching radius , and that considers whether each source is appropriately isolated within the observing beam .
we shall consider matching criteria in section [ thematchingradii ] . as the nvss rm data do not contain the angular size , evpa , or redshift of each source
, we add this information to the catalog by cross - matching to the original nvss @xcite and to a catalog of nvss redshifts @xcite .
as all three catalogs use the nvss source positions , an arbitrarily small cross - matching radius was used to combine the data . as
the fdr is a measure of the probability of coincidental associations , this must be primarily related to the density of sources on the sky .
following @xcite , such a relation is given by @xmath59 where the probability of the nearest unrelated source , @xmath60 , lying within an area @xmath61 is dependent on the number of sources per square degree , @xmath62 . by adjusting the maximum threshold radius at which cross - matches are confirmed or rejected , it is therefore possible to control the fdr to be below some subjective limit that we deem to be acceptable .
such methods are useful for checking if optical sources emit in the radio , as the coordinates to optical sources typically have good positional accuracy .
the situation becomes more complicated when comparing two sets of radio data , in which there may be significant inaccuracies in both surveys to be matched .
furthermore , while such cross - matching evaluates the positional accuracy ( or astrometry ) of each measurement , it does not take resolution effects into account .
it is possible that unresolved sources as seen with a larger observing beam are resolved into separate components in a high resolution survey .
consequently , we define two radii for our matching criterion an astrometric radius , @xmath63 , that accommodates the accuracy of a measured source position , and an isolation radius , @xmath64 , that handles resolution effects . in choosing an astrometric radius , we need to properly consider that the astrometric errors are not dominated by those in a single survey , and that there are typically significant contributions from both the reference and secondary catalogs .
we assume that the effects of s / n on the astrometry are negligible ; while the positional uncertainty typically depends on the s / n , all of our sources are of sufficient s / n to be detected in linear polarization which tends to select the sources of highest total intensity flux density within a given survey .
we approximate the rms uncertainty in the source positions to be given by the combined rms of both catalogs , such that @xmath65 , where @xmath66 and @xmath67 are the contributions from the reference and secondary catalogs respectively .
we estimate that a suitable @xmath63 could be @xmath68 , such that we would allow a match if a source is within the @xmath69% confidence interval .
nevertheless , values of @xmath63 smaller than @xmath70 would also be acceptable , as this only serves to place a tighter constraint on the identified cross - matches .
moreover , due to the low source density of our catalogs , coincidental matches can be expected to be low even for substantially larger @xmath63 ( see equation [ probfdr ] ) .
of course , the most reliable method would be to independently measure the positional errors in each catalog . to balance these considerations , the exact chosen values of @xmath63 are determined empirically from an assessment of the fdr in section [ section : fdr ] .
the estimated positional errors for each survey are stated in table [ tab : astrometry ] . in cases where the astrometry is not provided for a secondary catalog , we make the conservative estimate that the rms uncertainty in each position is given by @xmath71 , where @xmath72 is the fwhm of the resolution element . in terms of an isolation radius , we always take @xmath64 to be equal to the fwhm of the catalog with the larger observing beam regardless of whether the secondary catalog to be cross - matched has an observing beam that is smaller or larger than the nvss beam of fwhm @xmath73 arcsec . as
@xmath64 is a radial distance , the source will always be isolated within @xmath74fwhm of the catalog with the larger observing beam .
the isolation radius used is listed in table [ tab : astrometry ] .
this provides a limited number of possible isolation scenarios while cross - matching an individual source , as follows : no sources are detected within @xmath64 .
only one source is detected within @xmath64 .
more than one source is detected within @xmath64 . in case ( i ) , nothing is done and we simply iterate to the next source . in case ( ii ) , the detected source is appended to the reference catalog if the astrometric radius does not denote this as a positional coincidence . in case ( iii ) , the source consists of multiple components and neither of these components are appended to the final catalog .
consequently , the catalog selects isolated sources , which are unresolved at all frequencies .
ccccccc nvss & @xmath75 for @xmath76 mjy & n / a & n / a & n / a & n / a & n / a + & @xmath77 for @xmath78 mjy & & & & & + ti80 & @xmath79 & 390 & 150 & 38 & 767 & 5.0 + at20 g & 0.9@xmath27 in right ascension & 45 & 7 & 4 & 1272 & 0.3 + & 1.0@xmath27 in declination & & & & & + z99 & @xmath80 & 180 & 25 & 0 & 54 & 0.0 + t03 & @xmath75 & 45 & 5 & 1 & 103 & 1.0 + b3-vla & @xmath81 & 69 & 10 & 1 & 132 & 0.8 + sn80 & @xmath79 & 390 & 140 & 7 & 144 & 4.9 + sn81 & @xmath79 & 390 & & & & + sn82 & @xmath79 & 390 & & & & + gb6 & @xmath82 in right ascension & 216 & 90 & 715 & 16700 & 4.3 + & @xmath83 in declination & & & & & + wenss & @xmath84 & 54 & 17 & 133 & 11893 & 1.1 + texas & @xmath85 & 204 & 10 & 363 & 5789 & 6.3 + north6 cm & @xmath86 & 210 & 90 & 463 & 14166 & 3.3 [ tab : astrometry ] in order to assess the reliability of a particular cross - match , we need to measure the combined positional errors of each of our catalogs relative to the nvss rm reference catalog .
this also allows us to measure the most suitable astrometric cross - matching radius , @xmath63 .
in addition , we would also like to estimate the fdr by placing quantitative limits on the number of false cross - matches that would arise if a given survey was cross - matched to a collection of sources that were distributed randomly across the sky .
following the methodology of @xcite , we run the cross - matching as normal , and then a second time with the position of each source being offset by @xmath87 arcmin in both right ascension and declination .
for both runs , we record the radial distance between every source and its identified counterpart .
the histogram of the number of matches as a function of the radial distance between matches can then be created for both the true and offset sky positions .
the separation between matched sources is a measure of the combined positional errors in both catalogs , and allows us to estimate the positional errors and the number of coincidental matches .
these histograms are shown in fig .
[ histograms ] for every cross - matched catalog .
the number of matches for the true and offset sources becomes approximately equal at some specific separation , and this distance was estimated and chosen as the astrometric matching radius .
if this astrometric radius is underestimated , a substantial number of real matches will be rejected .
conversely , if overestimated , we will be accepting an increasing number of false matches with minimal return in terms of an increased sample size .
our estimates therefore provide a balance between maximizing our sample size and maintaining a low fdr .
the estimated values for @xmath63 are typically proportional to the estimated uncertainties in the source positions as shown in table [ tab : astrometry ] . for cross - matching , the set of secondary measurements were only appended to the reference catalog if the listed source positions agreed to within @xmath63 . as an example , we consider the case of the at20 g . based on our matching criteria , the offset sky positions result in four matches ,
compared to 1,272 matches when no offset is added .
we therefore estimate the fdr of the at20 g ( when combined with the nvss rm catalog ) to be @xmath88% . in table
[ tab : astrometry ] , we list the measured astrometric radii that we used for the cross - matching of each catalog , together with the number of matches for each catalog , and the estimated fdrs .
histograms of the number of cross - matched measurements for each source are shown in fig .
[ polhisto ] .
the histograms show that most polarized sources have @xmath89 measurements , while a few sources have up to 56 measurements .
similarly , most total intensity sources have @xmath90 measurements , while a few sources have up to 12 measurements .
note that some measurements are at similar or closely - spaced wavelengths .
-axes are shown as a log - scale .
sources with just one datum ( i.e. no cross - matches ) , are not shown.,title="fig:",width=359 ] -axes are shown as a log - scale .
sources with just one datum ( i.e. no cross - matches ) , are not shown.,title="fig:",width=359 ]
the cross - matching in section [ crossmatching ] provides 951 seds , each of which yields measurements of the polarized fraction as a function of observing wavelength ( as shown in fig .
[ seds ] ) , and of the total intensity as a function of frequency ( as shown in fig .
[ seds - i ] ) .
the seds show a significant number of different behaviors , with sources showing a monotonic decline , distinct maxima , or even oscillatory behavior with varying wavelength .
we now wish to fit a depolarization model to each sed and here focus on the fitting and model selection procedure the physics that motivates each model is detailed in appendix [ appendix - b ] . across the wavelength range of 0.4 ghz to 100 ghz
, we only use sources that have three or more data often providing minimal constraints to our fits . fitting to such empirical data is one of the most ubiquitous and notorious problems in astronomy @xcite .
model fitting can assist in further reducing the fdr ( the fdrs were measured in section [ section : fdr ] ) .
consider the case where the data exhibit ` regular depolarization ' , which we define to be an unambiguous and monotonic decline in fractional polarization as a function of increasing wavelength , e.g. as shown by the sources in the top - left and middle - left of fig .
we make the assumption that the distribution of @xmath26 values we here neglect any correction for the effect of rician bias see section [ ricianbias ] for further details .
] can be very approximately modeled as being drawn from a rayleigh distribution
such an assumption may be reasonable based on previous studies which show a distribution that is positive - definite , with a defined peak , and an asymmetric tail ( e.g. * ? ? ?
* ; * ? ? ?
* ) , although such analyses require careful consideration of numerous systematics ( e.g. * ? ? ?
* ) . in the simplest case ,
all values are drawn from distributions with a similar mean , @xmath91 , and mode , @xmath92 , for the population at each frequency .
the probability of drawing three values from these distributions that appear to monotonically decline as a function of wavelength was estimated using monte carlo techniques to be 0.16 .
consequently , our estimated typical fdr of @xmath93% ( see table [ tab : astrometry ] ) can be modified to @xmath94% ( i.e. @xmath95%@xmath96 ) for cases where an sed with three measurements is found to be a regular depolarizer .
the fdr continues to decrease as a function of the number of measurements in the sed : with four data our estimated fdr can be modified to @xmath97% , while with five data our fdr becomes @xmath98% .
note that our catalog contains up to 56 measurements in a single sed ( see table [ tab : surveyparameters ] ) .
these fdrs are best case scenarios , depending on our assumption that a rayleigh distribution can be used to model the polarized fractions . at the very minimum
, we estimate that at least 578 of the 951 sources in our catalog are regular depolarizers , and can be considered to have a reduced fdr . for many more sources
this becomes complicated to assess due to the uncertainty of each datum .
it should be noted that such an argument does not exclude that we may be probing different parts of the source at each observing frequency , especially as the recorded @xmath26 in each input catalog is biased towards the most highly polarized region of a given source .
consequently , any measurement of different source components will affect our estimate of a reduced fdr we therefore maintain our worst - case assumption of a fdr of @xmath93% ( see section [ systematics ] for a discussion of systematic effects in the catalog ) .
we are therefore faced with the issue of both obtaining the parameters that define the fitted sed , while also attempting to determine the physical model that best describes each source . for this purpose of model selection ,
we therefore need to pick a number of models a - priori that may describe the physics of each source .
the functional forms of the chosen models are described in detail in section [ fittingpolarizedsed ] , alongside our justification for their use .
the physics that motivates each of these models is explained in appendix [ appendix - b ] . distinguishing which of these is the _ correct _ model has been well - described in the case of univariate data @xcite , while in the case of bivariate data such as that presented here , linear regression is the most appropriate tool ( as described by e.g. * ? ? ?
* ) . as each sed consists of independent data with differing @xmath99 uncertainties , we use a combination of weighted least squares ( wls ) and the non - linear weighted levenberg
marquadt algorithm ( lma ) to perform the fitting of each model .
we use wls for cases where the model to be fit can be linearized , and use the lma for non - linearized cases .
the reported errors in the fits for the lma are fundamentally approximate .
furthermore , the lma only provides the parameters of a local rather than a global minima . for a significantly large sample of fitted seds
, we argue that we will still be dominated by random , rather than systematic errors . for the large data volumes expected with the next generation of telescopes
, automated and robust classification algorithms will be essential .
such automated classification is complicated not just by an assessment of the quality of fit , but also by different physical models having varying numbers of degrees of freedom . in order to perform model selection in an automated fashion
, we therefore require a methodology to distinguish between better fits , while simultaneously penalizing increasing model complexity . to do this
, we assessed the bayesian information criterion or bic @xcite .
the bic evaluates @xmath100 where @xmath101 is the log - likelihood of the data given the model , and @xmath102 is a parsimony term with @xmath6 giving the number of free parameters in the model , and @xmath103 giving the sample size . for each physical scenario
, we select as the ` true model ' that which has the lowest value of the bic .
the model we select serves as a parameterization of the sed . despite the challenges presented for seds with a small number of measurements
, we can determine the model of greatest likelihood given the data .
it is possible that the ` true ' sed has deviations from the model that our algorithm selects , and broadband polarization observations will be necessary to understand such rapid fluctuations as a function of @xmath11 .
deviations from our assumed models are beyond the scope of this paper , but can in principle be investigated using the statistical tests we used to assess the quality of an sed model fit . beyond a method for model selection
, we also require a measurement of the quality of the model selection process . in the catalog we provide both the reduced-@xmath104 and the kolmogorov
smirnov statistic for assessment of the quality of the model that is selected by our classification algorithm .
we also provide the associated @xmath105-value for each test .
these statistical tests are important for revealing cases where no model describes the data well .
note that reduced-@xmath104 is not appropriate for many seds , as @xmath104 statistics are derived in the limit of a sample size that tends towards infinity .
however , as least squares methods are derived from likelihood theory and assume normally distributed residuals ( e.g. * ? ? ?
* ) the kolmogorov smirnov statistic ( ks - test ) can be used as an alternative method to assess the normality of the residuals between our data and selected model . in our particular case
, we require comparison of the residuals to a normal distribution of unknown mean and variance the ks - test can be modified to deal with such a case ( as described by * ? ? ?
furthermore , the ks - test assumes that the data and model are independent , while in this case the data have been used to derive the model .
this does not affect the value of the ks - statistic , but causes all associated @xmath105-values to be in error .
we therefore perform ` bootstrap resampling ' .
we evaluate the ks - test @xcite for an sed , and then find the correct @xmath105-value by combining our data and model datasets , and randomly sampling from them ( with replacement ) to form two new samples of size @xmath106 ( where @xmath106 is the number of sed datapoints ) .
we then compute the ks - test statistic for these new self - generated samples .
this process was repeated 10,000 times .
the calculated @xmath105-value is the proportion of test statistics that are as extreme , or more so , than our original ks - test statistic .
we find that the values correspond well with a @xmath105-value derived simply using tables and formulae @xcite .
we do not indicate @xmath105-values @xmath107 or @xmath108 ; such values in the catalog indicate a lower or upper limit respectively .
unlike for @xmath104 , a high sample size is not fundamental to the derivation of the ks - test . nevertheless , the ks - test still has sensitivity to sample size , and in seds with few data the residuals will have to deviate substantially from normality in order for a low @xmath105-value to be output .
the @xmath105-values indicate the probability of getting a result as extreme as the one obtained , if the null hypothesis is true ( that the residuals are normally distributed or that the data are described by the assumed model , for the ks and @xmath104 tests respectively ) .
note that the @xmath105-value only specifies the probability with which one would reject the null - hypothesis , if it _
were _ correct it unfortunately provides no information on the probability that the null hypothesis _ is _ correct , i.e. we have calculated @xmath109 and not @xmath110 . for this frequentist approach , in cases where the @xmath105-value is lower than some threshold ( for example , @xmath111 ) , we take this to indicate a low probability of the two samples being as different as they are ( or more so ) , if drawn from the same distribution ( e.g. * ? ? ?
since each test has its limitations and so as to avoid overreliance on a single measure , we combine the output of all tests to produce a data quality flag . the flag value is evaluated based on a subjective combination of the @xmath104 of the sed model , the @xmath105-value of the @xmath104 , and the @xmath105-value of the ks - test
the flag has a value of 1 , 2 , or 3 and is provided as an indicative measure only ; these correspond to ` accept ' , ` caution ' , and ` poor ' , respectively .
the extensive number of depolarization models , as discussed at length in appendix [ appendix - b ] , can be primarily summarised by the ` burn ' @xcite , ` tribble ' @xcite , ` rossetti mantovani ' , ` repolarizer ' , and ` spectral depolarizer ' @xcite forms .
however , fitting these physical models to our data would arguably be misguided , as many of them assume an optically - thin emitting region . furthermore ,
all of these models make either the critical assumption that we detect the same emitting region at each frequency , or that the polarized fraction is a meaningful quantity with the measured peak in polarized intensity on the sky emanating from the same emission region as the total intensity peak .
as there is no way to verify this , and due to the large frequency range used , it is possible that any attempt at source classification will be corrupted .
for example , a source that _ truly _ depolarizes following @xcite may appear as a peaked spectral depolarizer as different emission regions are probed at different frequencies , and at high frequencies the sed may be altered as the result of the combined effects of faraday depolarization together with beam depolarization due to multiple components , and disordered magnetic fields ( e.g. * ? ? ?
we therefore argue that there is no reason to believe that any attempt to derive @xmath112 ( the rm dispersion of the faraday screen within a single beam see appendix [ appendix - b ] ) would be , or ever is , a reasonable physical probe without high resolution observations .
in addition , our main goal is to use the sed to derive the rest - frame source properties , allowing us some flexibility in our fitting procedures .
note that we are merely seeking a smoothly interpolating fit that is functionally similar to a physical model . for the purposes of fitting models to the data ,
we therefore ignore the underlying physics involved , and choose three models that serve as mathematical edifices in order to fit to the data . for the quality and sampling of data typically found in our catalog ,
we find that these three simple , generalized functional forms can produce seds that broadly mimic the wavelength - dependence of polarization of the various physical models described in appendix [ appendix - b ] .
for our fits , we consider three possibilities : a gaussian , a power law , and a gaussian with a constant term , as shown in equations [ gaussian ] , [ powerlaw ] , and [ gauss+const ] respectively . given the limited number of data in each sed , these models serve as empirical analogues to the physical models in which we are ultimately interested , and are given by @xmath113 \ , , \label{gaussian}\ ] ] @xmath114 and @xmath113 + c_{4 } \ , , \label{gauss+const}\ ] ] where @xmath11 is in units of centimetres , and the @xmath115 are coefficients for which we solve during the fitting process .
the units of @xmath115 are given in table [ tab : modelsunits ] .
with appropriately chosen coefficients , the gaussian model of equation [ gaussian ] shows almost indistinguishable wavelength - dependence of fractional polarization to a ` burn ' law ( see equation [ burnbabyburn ] ; * ? ? ? * ) or to a spectral depolarizer ( see appendix [ spectraldepolrepol ] ; * ? ? ?
the power law model of equation [ powerlaw ] provides a fit for a ` tribble ' law ( see equation [ tribbling2 ] ; * ? ? ?
* ) , and is flexible enough to fit a ` repolarizer ' . similarly to equation [ gaussian ] , with appropriately chosen coefficients the model for a gaussian with a constant term of equation [ gauss+const ] behaves similarly to the ` rossetti
mantovani ' law .
these models are all discussed in detail in appendix [ appendix - b ] .
we emphasize that for two of the functional forms for which we fit to the data , namely equations [ gaussian ] and [ gauss+const ] , the fit parameters @xmath116 , @xmath117 , @xmath118 , and @xmath119 should not be trivially equated with the physical coefficients included in the equations in the appendices .
this is due to the fact that we use a dependence on @xmath7 , rather than on @xmath120 as per the physical models of appendix [ appendix - b ] .
as both functions look similar , distinguishing between such models would require data of exceptional quality .
we also again emphasize that our goal is to obtain the rest - frame source properties using a smoothly interpolating predictive sed that covers a broad range of wavelengths , and that can be used for the purpose of a @xmath6-correction .
in addition , we also highlight that that there is no reason to believe that any attempt to derive @xmath112 would be , or ever is , a reasonable physical probe without high resolution observations .
should a user of the catalog wish to obtain a quantity other than a source classification or @xmath6-correction from the sed , they are able to refit the sed using the raw data available in the catalog . in all cases , fitting was only performed on sources for which data were available at three or more wavelengths and when the minimum wavelength separation between the three measurements was greater than 5 cm . due to the differing number of degrees of freedom for the various models ,
attempts to solve equations [ gaussian ] and [ powerlaw ] were made only when there were three or more data , and attempts to solve equation [ gauss+const ] were made only when there were five or more data .
various constraints were placed on the fitting process to ensure that only values of @xmath115 corresponding to physical solutions could be obtained .
this is particularly important when fitting equation [ gaussian ] , which has three degrees of freedom and can be fit to three data .
nevertheless , this is necessary in order to attempt to classify peaked sources , i.e. spectral depolarizers .
the addition of term @xmath117 to equations [ gaussian ] and [ gauss+const ] allows an opportunity to identify spectral depolarizers , or other ` peaked ' sources . in cases where this offset was found to be low ( @xmath121 cm ) , the data were refit without the inclusion of an offset .
repolarizing sources can be identified via the flexibility of equation [ powerlaw ] , which provides a polarization spectral index , @xmath122 in such cases , where @xmath24 is defined such that @xmath25 .
however , such sources should be treated with caution and may represent an undersampled peaked source particularly if the sed lacks multiple measurements at lower frequencies .
example fitted model seds from the catalog are shown in fig .
the seds in the catalog show a number of different forms , most of which are well described by our chosen models , but with a few unusual and possibly oscillatory cases also present .
the model selected by the bic is indicated in the catalog for each sed , alongside the coefficients for this model , and the corresponding statistical tests . however , the best fit to the sed with a polarization spectral index , @xmath24 , is always returned in the catalog unless the source was found to be peaked in which case , only measurements in the tail region ( i.e. those with a wavelength larger than the peak value ) were used to calculate @xmath24 .
a best fit power law is calculated whenever there are @xmath123 polarization measurements .
for some sources , the power law does not provide an accurate representation of the full sed ; the model fits provide more detailed information .
we choose two mathematical models to fit to our total intensity data .
these are the power law , and the curved power law respectively , as given by @xmath124 and @xmath125 } \
, , \label{powerlawicurved}\ ] ] where @xmath126 is a reference frequency that we define as 1.4 ghz , @xmath44 is the total intensity flux density in units of millijansky , @xmath23 is the observing frequency in hertz , and the @xmath127 are some coefficients for which we solve .
the units of @xmath127 are given in table [ tab : modelsunits ] .
in all cases , we only attempt to estimate curvature in spectra with three or more data at different wavelengths and when the minimum distance between the three measurements is greater than 5 cm of wavelength . for sources with just two data points ,
we calculate a two - point spectral index based on equation [ powerlawi ] and make no attempt to estimate the presence of curvature .
example total intensity seds from the catalog are shown in fig .
[ seds - i ] .
a histogram of the number of sources with a given spectral index is shown in fig .
[ i_histo ] .
the histogram shows a peaked distribution with a maximum at @xmath128 .
there is an extended tail in the @xmath0-distribution towards flatter - spectral indices , caused by the flat - spectrum population .
nevertheless , our sample is clearly dominated by steep - spectrum objects .
calculating the spectral index also allows us to estimate the non - thermal rest - frame luminosity of the source at 1.4 ghz .
we do this using @xmath129 , where @xmath130 is the luminosity distance .
the calculation assumes that the dominant emission mechanism is the synchrotron process , and that this process follows a power law such that @xmath21 .
we also assume a @xmath131cdm model and a flat cosmology , using the planck cosmological parameters such that @xmath132 and @xmath133 km s@xmath43 mpc@xmath43 @xcite .
the fr - i / fr - ii luminosity divide is typically taken to be @xmath134 w hz@xmath43 . .
the sample is dominated by steep - spectrum sources.,width=321 ] under some circumstances , it can be advantageous to calculate the peak faraday depth using the technique of rm synthesis .
this technique has three main advantages in comparison to fitting a straight line to a plot of the evpa versus @xmath7 : ( i ) no ` @xmath135 ' ambiguities , ( ii ) recognition of several faraday depth components in faraday - space within the source ( which can also cause the depolarization to be non - monotonic ) , and ( iii ) recognition of internal faraday rotation , which can lead to deviations of the evpa from a simple @xmath7 law ( see equation [ rotationmeasureequation ] ) and can cause broad structures in faraday - space .
however , as our catalog covers such a broad range in @xmath11 , and as there are large gaps in our @xmath7-coverage , it is problematic to use rm synthesis to calculate the rm . instead
, we fit a straight line to a plot of the evpa versus @xmath7 , where the evpa is defined between @xmath136 and @xmath137 from north through east .
such a method assumes that the rm is constant as a function of @xmath11 . while deviations from a @xmath7-law ( see equation [ rotationmeasureequation ] ) are beyond the scope of this paper , such effects can in principle be explored using the raw data available in the catalog .
the fitting process is complicated by the ` @xmath138 ambiguity ' , with each datum having undergone wrapping by some integer multiple of @xmath139 .
we therefore need to maintain goodness of fit while carefully unwrapping these ambiguities .
various approaches to this have been attempted previously .
we undertake a ` brute force ' approach , and fit to the data for every possible combination of positive and negative wrapping per datum , with any number of wraps between zero and five being allowed .
this is computationally expensive , particularly for sources with large numbers of measurements .
for example , a source with @xmath103 data across the squared - wavelength coverage has @xmath140 wrap combinations , each of which ( with the exception of zero wrap , which has no sign ambiguity ) has a further @xmath141 sign combinations providing a total number of @xmath142 fitting attempts for a single source .
however , only a fraction of these possibilities are physical ; for a source that demonstrates a linear relationship between evpa and @xmath7 the sign of a wrap at higher frequency can not change towards low frequencies . furthermore ,
if there is a wrap at high frequency then all lower frequency measurements must also be unwrapped by at least an equal integer amount .
this property was exploited to reduce the number of brute force attempts ; for a source with seven data this lowers the number of trial fits from 35,831,808 to just 10,836 .
such a technique is biased towards providing high rms , as with a sufficient number of wraps the goodness of fit can be improved to arbitrary levels . to counteract this ,
we utilize a modified version of the bic discussed in section [ sourcefitting ] , taking @xmath6 to be the square of the maximum number of wraps .
such a method penalizes high numbers of wraps substantially , unless they provide a substantial increase in fitting quality .
we only allow for a maximum of five possible wraps in our data , allowing for a maximum |rm|@xmath143370 rad m@xmath144 ( depending on the frequency coverage ) .
such high rms are expected to be rare on the sky , and exist mostly near the galactic plane ( e.g. * ? ? ?
furthermore , we use the nvss rms as our reference catalog which suffers from significant bandwidth depolarization for sources with cases , we only fit rms to sources with three or more data at different wavelengths and when the minimum wavelength separation between the three measurements is greater than 5 cm . for each fit
, we also attempt to calculate the intrinsic evpa , i.e. the evpa at infinite frequency . due to the significant number of wraps required at low observing frequencies ,
all measurements at wavelengths greater than @xmath145 cm were excluded from the fitting .
all sources have evpa measurements at 1.4 ghz from the nvss catalog ( see section [ crossmatching ] ) , with the majority of other measurements being at higher frequencies . due to the distribution of these measurements in frequency - space ,
our broadband rm fits are biased towards providing the faraday structure at high frequency .
it is difficult to define the exact frequency at which rms are measured , and we take them to be a high - frequency rm at @xmath93 ghz .
these broadband rms are compared against the nvss rms in fig .
[ rmmatch ] .
the distributions of both the broadband and nvss rms are also shown for sources that have both values . while there are clearly outliers , many sources are also in reasonable agreement and cluster about the line of unity .
the outliers appear associated with low nvss rms , which have a high corresponding broadband rm .
similar effects have been seen in other similar samples @xcite .
the differences could occur due to errors in the unwrapping of our data , uncertainty in the calculation of the rm in the nvss data , time - variability of the rm towards a radio source , or ` faraday complexity ' ( i.e. a non - linear relationship between the evpa and @xmath7 ) .
data for all sources and the associated fitted and calculated properties are provided as machine - readable tables .
these tables are published in their entirety in the electronic edition of the astrophysical journal supplement series .
a portion is shown here for guidance regarding its form and content .
there are two versions of the catalog available , the ` full - catalog ' and the ` sed - catalog ' respectively .
the full - catalog contains 37,543 rows , providing an expanded version of the @xcite data and contains all of the derived total intensity and polarized fraction spectral indices and other incorporated cross - matched and derived quantities .
note that the catalog contains no upper limits on polarized or total intensity measurements .
the sed - catalog contains only the 951 sources for which a model has been fitted to the polarized sed .
the column headings are identical in both catalog versions , and the sed - catalog is just a subset of rows from our primary data product , the full - catalog .
the column headings of both catalogs are described in appendix [ appendix - a ] , including calculated quantities and ancilliary data provided during the cross - matching process .
we emphasize the various statistical tests that are provided within the catalog , which are included so that a user may assess the quality of a particular model fit .
selected columns for the first 45 sources in the sed - catalog are presented in tables [ tab : thedata ] to [ tab : thedata5 ] .
all errors are the @xmath99 uncertainties and have been calculated using standard error propagation .
c c c c c c c c c c c 104 & 000322 - 172711 & 00h03m22.00s & -17d27m11.40 & 71.5253 & -75.2777 & 4 & 3 & [ 21.41 , ,
, 1.51 , @xmath146 & [ 1.47 , , , 3.70 , @xmath146 & [ 0.11 , , , 0.37 , @xmath146 + 157 & 000457 + 124818 & 00h04m57.12s & + 12d48m18.90 & 105.686 & -48.5038 & 5 & 2 & [ 21.41 , , , ,
@xmath146 & [ 19.89 , , , , @xmath146 & [ 0.84 , , , , @xmath146 + 186 & 000559 + 160946 & 00h05m59.41s & + 16d09m46.70 & 107.3192 & -45.3266 & 7 & 4 & [ 21.41 , , , , @xmath146 & [ 2.10 , , , , @xmath146 & [ 0.13 , , , , @xmath146 + 195 & 000613 - 062335 & 00h06m13.87s & -06d23m35.20 & 93.5053 & -66.6465 & 9 & 10 & [ 21.41 ,
, , 1.51 , @xmath146 & [ 3.19 , , , 4.60 , @xmath146 & [ 0.14 ,
, , 0.46 , @xmath146 + 200 & 000622 - 000425 & 00h06m22.60s & -00d04m25.10 & 99.2806 & -60.8593 & 14 & 3 & [ 21.41 , ,
, 1.51 , @xmath146 & [ 0.71 , , , 2.50 , @xmath146 & [ 0.10 , , , 0.25 , @xmath146 + 246 & 000826 - 255912 & 00h08m26.21s & -25d59m12.30 & 37.2729 & -80.3196 & 4 & 5 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 1.47 , 2.90 , 4.20 , 6.0 , @xmath146 & [ 0.14 , 0.29 , 0.42 , 0.60 , @xmath146 + 274 & 000935 - 321636 & 00h09m35.76s & -32d16m36.90 & 0.8704 & -79.5645 & 3 & 5 & [ 21.41 , 6.25 , 3.47 ,
, @xmath146 & [ 4.36 , 4.40 , 3.10 , , @xmath146 & [ 0.23 , 0.44 , 0.31 , , @xmath146 + 366 & 001259 - 395425 & 00h12m59.95s & -39d54m25.70 & -27.5425 & -74.9393 & 4 & 4 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 2.74 , 2.20 , 3.30 , 4.10 , @xmath146 & [ 0.16 , 0.22 , 0.33 , 0.41 , @xmath146 + 383 & 001331 + 405136 & 00h13m31.09s & + 40d51m36.00 & 115.2409 & -21.4448 & 5 & 7 & [ 21.41 , , , , @xmath146 & [ 2.31 , , , , @xmath146 & [ 0.14 , , , , @xmath146 + 391 & 001354 + 204522 & 00h13m54.84s & + 20d45m22.90 & 111.2333 & -41.2615 & 3 & 3 & [ 21.41 , , , , @xmath146 & [ 5.41 , , , , @xmath146 & [ 0.21 , , , , @xmath146 + 449 & 001559 + 390027 & 00h15m59.98s & + 39d00m27.20 & 115.442 & -23.3483 & 4 & 7 & [ 21.41 , , , , @xmath146 & [ 4.59 , , , , @xmath146 & [ 0.21 , , , , @xmath146 + 460 & 001631 + 791651 & 00h16m31.61s & + 79d16m51.70 & 121.245 & 16.5238 & 9 & 2 & [ 21.41 , , , , @xmath146 & [ 6.51 , , , , @xmath146 & [ 0.25 , , , , @xmath146 + 519 & 001851 - 124233 & 00h18m51.38s & -12d42m33.50 & 93.4551 & -73.6872 & 5 & 3 & [ 21.41 , , , 1.51 , @xmath146 & [ 1.17 , , , 5.40 , @xmath146 & [ 0.11 , , , 0.54 , @xmath146 + 571 & 002025 + 154052 & 00h20m25.32s & + 15d40m52.70 & 112.0476 & -46.5339 & 16 & 3 & [ 21.41 , , , , @xmath146 & [ 5.72 , , , , @xmath146 & [ 0.20 , , , , @xmath146 + 696 & 002430 - 292848 & 00h24m30.12s & -29d28m48.90 & 9.7191 & -83.6231 & 6 & 3 & [ 21.41 , , , ,
@xmath146 & [ 2.03 , , , , @xmath146 & [ 0.12 , , , , @xmath146 + 725 & 002526 + 391935 & 00h25m26.15s & + 39d19m35.70 & 117.4617 & -23.2689 & 4 & 7 & [ 21.41 , , , , @xmath146 & [ 6.02 , , , , @xmath146 & [ 0.21 , , , , @xmath146 + 730 & 002549 - 260212 & 00h25m49.17s & -26d02m12.80 & 42.2744 & -84.1701 & 12 & 3 & [ 21.41 , , , ,
@xmath146 & [ 0.79 , , , , @xmath146 & [ 0.10 , , , , @xmath146 + 813 & 002831 - 002540 & 00h28m31.35s & -00d25m40.80 & 110.3451 & -62.7384 & 3 & 1 & [ 21.41 , , , , @xmath146 & [ 8.06 , , , , @xmath146 & [ 0.38 , , , , @xmath146 + 896 & 003133 - 010121 & 00h31m33.63s & -01d01m21.10 & 111.7545 & -63.4621 & 3 & 1 & [ 21.41 , , , , @xmath146 & [ 14.19 , , , , @xmath146 & [ 0.49 , , , , @xmath146 + 924 & 003235 + 394215 & 00h32m35.51s & + 39d42m15.50 & 118.9947 & -23.0234 & 4 & 7 & [ 21.41 , , , , @xmath146 & [ 1.78 , , , , @xmath146 & [ 0.14 , , , , @xmath146 + 998 & 003508 - 200359 & 00h35m08.80s & -20d03m59.20 & 94.233 & -82.0144 & 3 & 5 & [ 21.41 , , , , @xmath146 & [ 0.63 , , , , @xmath146 & [ 0.10 , , , , @xmath146 + 1030 & 003606 + 183759 & 00h36m06.55s & + 18d37m59.00 & 117.8735 & -44.0881 & 7 & 3 & [ 21.41 , , , , @xmath146 & [ 1.94 , , , , @xmath146 & [ 0.12 , , , , @xmath146 + 1031 & 003607 + 585548 & 00h36m07.89s & + 58d55m48.90 & 120.9536 & -3.8827 & 4 & 4 & [ 21.41 , , , , @xmath146 & [ 1.94 , , , , @xmath146 & [ 0.12 , , , , @xmath146 + 1062 & 003704 - 010909 & 00h37m04.06s & -01d09m09.40 & 114.7753 & -63.7966 & 7 & 3 & [ 21.41 , , , 1.51 , @xmath146 & [ 4.68 , , , 3.70 , @xmath146 & [ 0.19 , , , 0.37 , @xmath146 + 1098 & 003758 + 240711 & 00h37m58.31s & + 24d07m11.60 & 118.9968 & -38.649 & 5 & 5 & [ 21.41 , , , , @xmath146 & [ 2.40 , , , , @xmath146 & [ 0.13 , , , , @xmath146 + 1115 & 003820 - 020740 & 00h38m20.41s & -02d07m40.40 & 115.2268 & -64.8035 & 14 & 4 & [ 21.41 , , , , @xmath146 & [ 0.61 , , , , @xmath146 & [ 0.10 , , , , @xmath146 + 1144 & 003916 - 185405 & 00h39m16.95s & -18d54m05.30 & 103.5398 & -81.312 & 3 & 4 & [ 21.41 , 6.25 , 3.47 , ,
@xmath146 & [ 2.57 , 7.10 , 6.40 , , @xmath146 & [ 0.19 , 0.71 , 0.64 , ,
@xmath146 + 1190 & 004057 - 014631 & 00h40m57.62s & -01d46m31.60 & 116.8359 & -64.5231 & 6 & 2 & [ 21.41 , , , 1.51 , @xmath146 & [ 3.48 , , , 2.70 , @xmath146 & [ 0.15 , , , 0.27 , @xmath146 + 1268 & 004308 + 520334 & 00h43m08.79s & + 52d03m34.40 & 121.6347 & -10.7913 & 13 & 5 & [ 21.41 , , , , @xmath146 & [ 2.07 , , , , @xmath146 & [ 0.12 , , , , @xmath146 + 1282 & 004349 + 383010 & 00h43m49.42s & + 38d30m10.10 & 121.2969 & -24.3448 & 4 & 7 & [ 21.41 , , , , @xmath146 & [ 4.23 , , , , @xmath146 & [ 0.21 , , , , @xmath146 + 1289 & 004358 + 393756 & 00h43m58.38s & + 39d37m56.60 & 121.3681 & -23.2167 & 4 & 7 & [ 21.41 , , , , @xmath146 & [ 8.79 , , ,
, @xmath146 & [ 0.35 , , , , @xmath146 + 1315 & 004441 - 353032 & 00h44m41.47s & -35d30m32.80 & -47.7431 & -81.4966 & 8 & 5 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 1.46 , 6.30 , 7.80 , 7.70 , @xmath146 & [ 0.11 , 0.63 , 0.78 , 0.77 , @xmath146 + 1421 & 004843 + 404458 & 00h48m43.83s & + 40d44m58.50 & 122.3784 & -22.1193 & 4 & 7 & [ 21.41 , , , , @xmath146 & [ 13.83 , , , , @xmath146 & [ 0.52 , , , , @xmath146 + 1476 & 005041 - 092906 & 00h50m41.27s & -09d29m06.20 & 122.3217 & -72.356 & 9 & 3 & [ 21.41 , , , 1.51 , @xmath146 & [ 3.16 , , , 1.80 , @xmath146 & [ 0.14 , , , 0.18 , @xmath146 + 1561 & 005408 - 033355 & 00h54m08.40s & -03d33m55.90 & 124.6181 & -66.4285 & 4 & 4 & [ 21.41 , , , , @xmath146 & [ 0.52 , , , , @xmath146 & [ 0.10 , , , , @xmath146 + 1566 & 005427 + 404211 & 00h54m27.31s & + 40d42m11.80 & 123.5494 & -22.1649 & 4 & 7 & [ 21.41 , , , , @xmath146 & [ 1.80 , , , , @xmath146 & [ 0.12 , , , , @xmath146 + 1606 & 005550 + 262436 & 00h55m50.38s & + 26d24m36.00 & 124.1572 & -36.4513 & 6 & 3 & [ 21.41 , , , , @xmath146 & [ 0.82 , , , , @xmath146 & [ 0.11 , , , , @xmath146 + 1664 & 005802 - 323418 & 00h58m02.26s & -32d34m18.30 & -71.3925 & -84.3718 & 4 & 4 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 4.12 , 8.90 , 10.10 , 11.40 , @xmath146 & [ 0.25 , 0.89 , 1.01 , 1.14 , @xmath146 + 1698 & 005905 + 000651 & 00h59m05.53s & + 00d06m51.50 & 127.1063 & -62.6954 & 14 & 5 & [ 21.41 , , , , @xmath146 & [ 3.91 , , , , @xmath146 & [ 0.15 , , , , @xmath146 + 1738 & 010005 + 394935 & 01h00m05.28s & + 39d49m35.70 & 124.7362 & -23.0149 & 4 & 6 & [ 21.41 , , , , @xmath146 & [ 16.67 , , , , @xmath146 & [ 0.65 , , , , @xmath146 + 1817 & 010246 + 462438 & 01h02m46.64s & + 46d24m38.10 & 124.9692 & -16.4163 & 4 & 6 & [ 21.41 , , , , @xmath146 & [ 4.56 , , , , @xmath146 & [ 0.18 , , , , @xmath146 + 1820 & 010250 + 255216 & 01h02m50.22s & + 25d52m16.60 & 126.14 & -36.9295 & 4 & 4 & [ 21.41 , , , , @xmath146 & [ 0.72 , , , , @xmath146 & [ 0.11 , , ,
, @xmath146 + 1852 & 010422 - 123515 & 01h04m22.49s & -12d35m15.60 & 135.3358 & -75.147 & 5 & 3 & [ 21.41 , , , , @xmath146 & [ 5.88 , , , , @xmath146 & [ 0.23 , , , , @xmath146 + 1870 & 010509 - 381057 & 01h05m09.33s & -38d10m57.20 & -70.8025 & -78.5774 & 3 & 1 & [ 21.41 , , , , @xmath146 & [ 8.13 , , , , @xmath146 & [ 0.29 , , , , @xmath146 + 1900 & 010622 - 015538 & 01h06m22.98s & -01d55m38.30 & 131.6477 & -64.5459 & 4 & 3 & [ 21.41 , , , 1.51 , @xmath146 & [ 7.62 , , , 4.30 , @xmath146 & [ 0.25 ,
, 0.43 , @xmath146 [ tab : thedata ] c c c c c c c c 104 & [ 126.6 , , , 18.0 , @xmath146 & [ 0.20 , , , 2.57 , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 2414.80 , , , 386.0 , @xmath146 & [ 70.0 , ,
, 40.0 , @xmath146 & power & tribble + 157 & [ 158.3 , , , , @xmath146 & [ 0.10 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 1069.50 , , , , @xmath146 & [ 40.0 , , , , @xmath146 & power & increase + 186 & [ 131.1 , , ,
, @xmath146 & [ 0.60 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 804.40 , , , , @xmath146 & [ 27.0 , , , , @xmath146 & gauss & burn + 195 & [ 36.4 , , , 32.0 , @xmath146 & [ 0.20 , , , 0.99 , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 2050.70 , , , 2084.0 , @xmath146 & [ 60.0 , , , 100.0 , @xmath146 & power & increase + 200 & [ 157.9 , , , 127.0 , @xmath146 & [ 0.30 , , , 14.45 , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 3897.60 , , , 425.0 , @xmath146 & [ 120.0 , , , 17.0 , @xmath146 & gau+t & mantovani + 246 & [ 107.6 , 50.0 , 138.0 , 154.0 , @xmath146 & [ 1.90 , 10.0 , 11.45 , 7.43 , @xmath146 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 468.20 , 360.0 , 259.0 , 120.0 , @xmath146 & [ 14.0 , 18.0 , 14.0 , 8.0 , @xmath146 & power & tribble + 274 & [ 95.0 , 64.0 , 62.0 , , @xmath146 & [ 0.50 , 4.57 , 6.89 ,
, @xmath146 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 388.30 , 313.0 , 275.0 , 221.0 , @xmath146 & [ 12.0 , 16.0 , 14.0 , 10.0 , @xmath146 & gauss & peaked + 366 & [ 113.0 , 89.0 , 100.0 , 91.0 , @xmath146 & [ 0.70 , 1.98 , 2.42 , 2.70 , @xmath146 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 493.60 , 2015.0 , 2009.0 , 1609.0 , @xmath146 & [ 17.0 , 100.0 , 100.0 , 70.0 , @xmath146 & power & tribble + 383 & [ 86.3 , , , , @xmath146 & [ 0.30 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 1650.90 , , , , @xmath146 & [ 60.0 , , , , @xmath146 & gauss & peaked + 391 & [ 73.0 , , , , @xmath146 & [ 0.30 , , , , @xmath146 & [ 21.41 , , , ,
@xmath146 & [ 624.80 , , , , @xmath146 & [ 21.0 , , , , @xmath146 & gauss & peaked + 449 & [ 168.3 , , , , @xmath146 & [ 0.90 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 263.60 , , , , @xmath146 & [ 8.0 , , , , @xmath146 & power & tribble + 460 & [ 62.2 , , , , @xmath146 & [ 0.40 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 3650.30 , , , , @xmath146 & [ 120.0 , , , , @xmath146 & power & tribble + 519 & [ 155.2 , , , 83.0 , @xmath146 & [ 0.40 , , , 20.75 , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 2122.30 , , , 141.0 , @xmath146 & [ 60.0 , , , 7.0 , @xmath146 & power & tribble + 571 & [ 161.1 , , , , @xmath146 & [ 0.10 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 2180.0 , , , , @xmath146 & [ 70.0 , , , , @xmath146 & gau+t & mantovani + 696 & [ 151.9 , , , , @xmath146 & [ 0.20 , , , , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 2922.80 , , , 126.0 , @xmath146 & [ 100.0 , , , 8.0 , @xmath146 & power & tribble + 725 & [ 13.8 , , , , @xmath146 & [ 0.20 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 707.90 , , , , @xmath146 & [ 21.0 , , , , @xmath146 & power & increase + 730 & [ 140.4 , , ,
, @xmath146 & [ 0.90 , , , , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 8753.30 , , , 984.0 , @xmath146 & [ 260.0 , , , 60.0 , @xmath146 & power & tribble + 813 & [ 107.7 , , , , @xmath146 & [ 0.30 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 645.10 , , , , @xmath146 & [ 22.0 , , , , @xmath146 & power & tribble + 896 & [ 24.8 , , , , @xmath146 & [ 0.10 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 463.40 , , , , @xmath146 & [ 16.0 , , , , @xmath146 & power & increase + 924 & [ 95.9 , , ,
, @xmath146 & [ 1.50 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 387.20 , , , , @xmath146 & [ 14.0 , , , , @xmath146 & power & tribble + 998 & [ 164.7 , , , , @xmath146 & [ 3.80 , , , , @xmath146 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 1957.10 , 560.0 , 282.0 , 87.0 , @xmath146 & [ 60.0 , 28.0 , 15.0 , 6.0 , @xmath146 & power & tribble + 1030 & [ 10.7 , , , , @xmath146 & [ 0.50 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 1984.10 , , , , @xmath146 & [ 70.0 , , , , @xmath146 & power & tribble + 1031 & [ 76.3 , , , , @xmath146 & [ 0.20 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 2541.90 , , , , @xmath146 & [ 90.0 , , , , @xmath146 & power & increase + 1062 & [ 20.7 , ,
, 45.0 , @xmath146 & [ 0.10 , , , 6.92 , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 4067.10 , ,
, 404.0 , @xmath146 & [ 140.0 , , , 17.0 , @xmath146 & gau+t & mantovani + 1098 & [ 55.2 , , , , @xmath146 & [ 0.40 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 996.10 , , , , @xmath146 & [ 30.0 , , , , @xmath146 & power & tribble + 1115 & [ 6.7 , , , , @xmath146 & [ 0.80 , , , , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 6187.80 , , , 885.0 , @xmath146 & [ 240.0 , , , 40.0 , @xmath146 & power & tribble + 1144 & [ 148.2 , 117.0 , 111.0 ,
, @xmath146 & [ 1.30 , 4.85 , 6.90 , , @xmath146 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 236.60 , 182.0 , 151.0 , 105.0 , @xmath146 & [ 7.0 , 9.0 , 8.0 , 7.0 , @xmath146 & gauss & peaked + 1190 & [ 8.8 , ,
, 44.0 , @xmath146 & [ 0.50 , , , 3.38 , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 573.20 , , , 467.0 , @xmath146 & [ 17.0 , , , 22.0 , @xmath146 & power & increase + 1268 & [ 116.3 , , ,
, @xmath146 & [ 0.10 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 11534.0 , , , , @xmath146 & [ 400.0 , , , , @xmath146 & power & tribble + 1282 & [ 161.8 , , , , @xmath146 & [ 1.0 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 220.80 , , , , @xmath146 & [ 7.0 , , , , @xmath146 & power & tribble + 1289 & [ 177.8 , , , , @xmath146 & [ 0.70 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 146.90 , , , , @xmath146 & [ 4.0 , , , , @xmath146 & power & tribble + 1315 & [ 9.7 , 74.0 , 73.0 , 70.0 , @xmath146 & [ 0.30 , 1.19 , 1.74 , 7.78 , @xmath146 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 2560.10 , 991.0 , 532.0 , 240.0 , @xmath146 & [ 80.0 , 50.0 , 27.0 , 12.0 , @xmath146 & gau+t & mantovani + 1421 & [ 5.6 , , , , @xmath146 & [ 0.40 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 223.0 , , , , @xmath146 & [ 8.0 , , , , @xmath146 & power & increase + 1476 & [ 74.6 , ,
, 26.0 , @xmath146 & [ 0.40 , , , 8.67 , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 814.0 , , , 511.0 , @xmath146 & [ 24.0 , , , 24.0 , @xmath146 & power & increase + 1561 & [ 32.5 , , ,
, @xmath146 & [ 3.20 , , , , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 2210.80 , , , 119.0 , @xmath146 & [ 70.0 , , , 6.0 , @xmath146 & power & tribble + 1566 & [ 92.9 , , , , @xmath146 & [ 0.50 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 888.10 , , , , @xmath146 & [ 30.0 , , , , @xmath146 & power & tribble + 1606 & [ 72.9 , , , , @xmath146 & [ 1.20 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 1361.70 , , , , @xmath146 & [ 50.0 , , , , @xmath146 & power & tribble + 1664 & [ 157.4 , 145.0 , 143.0 , 140.0 , @xmath146 & [ 1.20 , 1.46 , 1.19 , 3.24 , @xmath146 & [ 21.41 , 6.25 , 3.47 , 1.51 , @xmath146 & [ 185.90 , 267.0 , 309.0 , 327.0 , @xmath146 & [ 6.0 , 14.0 , 16.0 , 14.0 , @xmath146 & power & tribble + 1698 & [ 72.7 , , , , @xmath146 & [ 0.10 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 2508.80 , , , , @xmath146 & [ 80.0 , , , , @xmath146 & gau+t & mantovani + 1738 & [ 68.4 , , , , @xmath146 & [ 0.80 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 71.50 , , , , @xmath146 & [ 2.20 , , , , @xmath146 & power & tribble + 1817 & [ 37.9 , , , , @xmath146 & [ 0.20 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 1322.0 , , , , @xmath146 & [ 40.0 , , , , @xmath146 & power & tribble + 1820 & [ 160.3 , , , , @xmath146 & [ 0.70 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 1093.30 , , , , @xmath146 & [ 30.0 , , , , @xmath146 & gauss & burn + 1852 & [ 154.8 , , , , @xmath146 & [ 0.10 , , , , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 2125.10 , , , 152.0 , @xmath146 & [ 80.0 , , , 7.0 , @xmath146 & power & tribble + 1870 & [ 62.8 , , , , @xmath146 & [ 0.10 , , , , @xmath146 & [ 21.41 , , , , @xmath146 & [ 1238.70 , , , , @xmath146 & [ 40.0 , , , , @xmath146 & power & increase + 1900 & [ 158.7 , , , 179.0 , @xmath146 & [ 0.20 , , , 0.18 , @xmath146 & [ 21.41 , , , 1.51 , @xmath146 & [ 602.60 , , , 261.0 , @xmath146 & [ 18.0 , ,
, 13.0 , @xmath146 & gauss & peaked [ tab : thedata2 ] c c c c c c c c c c c c c c c c c c c c c 104 & 0.61 & 0.12 & -0.33 & 0.07 & & & & & 0.191 & 0.200 & 0.597 & 0.550 & 1 & 0.609 & -0.33 & 0.07 & 0.597 & 0.550 & 0.191 & 0.200 + 157 & 0.6 & 0.7 & 0.4 & 0.6 & & & & & 0.218 & 0.200 & 25.513 & 0.000 & 3 & 0.601 & 0.4 & 0.6 & 25.513 & 0.000 & 0.218 & 0.200 + 186 & 2.5 & 0.3 & 0.0 & 0.0 & 41.6 & 11.0 & & & 0.206 & 0.200 & 2.330 & 0.040 & 1 & 2.153 & -1.4 & 0.5 & 7.067 & 0.000 & 0.291 & 0.074 + 195 & 0.5 & 0.17 & 0.03 & 0.16 & & & & & 0.211 & 0.200 & 31.766 & 0.000 & 3 & 0.501 & 0.03 & 0.16 & 31.766 & 0.000 & 0.211 & 0.200 + 200 & 3.3 & 0.8 & 8.7 & 0.9 & 3.5 & 1.4 & 0.8 & 0.25 & 0.222 & 0.060 & 3.385 & 0.000 & 2 & 0.000 & -0.0 & 0.0005 & 25.171 & 0.000 & 0.294 & 0.010 + 246 & 0.9 & 0.18 & -0.55 & 0.11 & & & & & 0.306 & 0.200 & 0.263 & 0.769 & 1 & 0.899 & -0.55 & 0.11 & 0.263 & 0.769 & 0.306 & 0.200 + 274 & 6.6 & 0.7 & 13.8 & 1.4 & 8.4 & 0.8 & & & 0.385 & 0.108 & 0.000 & 1.000 & 1 & & & & & & 0.329 & 0.200 + 366 & 0.5 & 0.1 & -0.076 & 0.015 & & & & & 0.223 & 0.200 & 7.865 & 0.000 & 2 & 0.511 & -0.076 & 0.015 & 7.865 & 0.000 & 0.223 & 0.200 + 383 & 2.5 & 0.5 & 17.9 & 6.0 & 12.0 & 8.0 & & & 0.260 & 0.200 & 9.138 & 0.000 & 2 & & & & & & 0.211 & 0.200 + 391 & 11.1 & 1.1 & 14.0 & 1.4 & 6.2 & 0.6 & & & 0.237 & 0.200 & 0.000 & 1.000 & 1 & & & & & & 0.336 & 0.200 + 449 & 0.77 & 0.15 & -0.072 & 0.014 & & & & & 0.228 & 0.200 & 7.772 & 0.000 & 2 & 0.770 & -0.072 & 0.014 & 7.772 & 0.000 & 0.228 & 0.200 + 460 & 0.7 & 0.5 & -0.5 & 0.4 & & & & & 0.269 & 0.059 & 77.742 & 0.000 & 3 & 0.717 & -0.5 & 0.4 & 77.742 & 0.000 & 0.269 & 0.059 + 519 & 1.0 & 0.4 & -0.69 & 0.29 & & & & & 0.256 & 0.200 & 8.855 & 0.000 & 2 & 0.985 & -0.69 & 0.29 & 8.855 & 0.000 & 0.256 & 0.200 + 571 & 16.3 & 5.0 & 0.0 & 0.0 & 12.3 & 9.0 & 1.4 & 5.0 & 0.134 & 0.200 & 5.585 & 0.000 & 2 & 2.300 & -1.25 & 0.15 & 12.858 & 0.000 & 0.265 & 0.010 + 696 & 0.8 & 1.2 & -0.4 & 0.9 & & & & & 0.132 & 0.200 & 4.316 & 0.002 & 1 & 0.812 & -0.4 & 0.9 & 4.316 & 0.002 & 0.132 & 0.200 + 725 & 0.31 & 0.06 & 0.31 & 0.06 & & & & & 0.252 & 0.200 & 18.989 & 0.000 & 3 & 0.310 & 0.31 & 0.06 & 18.989 & 0.000 & 0.252 & 0.200 + 730 & 0.24 & 0.06 & -0.26 & 0.06 & & & & & 0.168 & 0.200 & 132.132 & 0.000 & 3 & 0.245 & -0.26 & 0.06 & 132.132 & 0.000 & 0.168 & 0.200 + 813 & 1.0 & 0.2 & -0.089 & 0.018 & & & & & 0.278 & 0.200 & 1.499 & 0.221 & 1 & 1.018 & -0.089 & 0.018 & 1.499 & 0.221 & 0.278 & 0.200 + 896 & 0.17 & 0.03 & 0.73 & 0.15 & & & & & 0.196 & 0.200 & 7.418 & 0.006 & 1 & 0.165 & 0.73 & 0.15 & 7.418 & 0.006 & 0.196 & 0.200 + 924 & 1.8 & 0.4 & -1.15 & 0.23 & & & & & 0.377 & 0.048 & 12.966 & 0.000 & 2 & 1.771 & -1.15 & 0.23 & 12.966 & 0.000 & 0.377 & 0.048 + 998 & 0.38 & 0.08 & -0.44 & 0.09 & & & & & 0.384 & 0.109 & 4.539 & 0.033 & 1 & 0.381 & -0.44 & 0.09 & 4.539 & 0.033 & 0.384 & 0.109 + 1030 & 1.1 & 0.3 & -0.59 & 0.26 & & & & & 0.166 & 0.200 & 2.635 & 0.022 & 1 & 1.055 & -0.59 & 0.26 & 2.635 & 0.022 & 0.166 & 0.200 + 1031 & -0.21 & 0.04 & 0.38 & 0.08 & & & & & 0.304 & 0.200 & 2.120 & 0.120 & 1 & -0.206 & 0.38 & 0.08 & 2.120 & 0.120 & 0.304 & 0.200 + 1062 & 3.7 & 0.8 & 11.5 & 1.2 & 8.5 & 2.1 & 2.3 & 0.7 & 0.202 & 0.200 & 2.427 & 0.033 & 1 & 1.771 & -0.83 & 0.17 & 42903.144 & 0.000 & 0.357 & 0.010 + 1098 & 1.9 & 0.3 & -1.15 & 0.25 & & & & & 0.440 & 0.010 & 12.231 & 0.000 & 3 & 1.907 & -1.15 & 0.25 & 12.231 & 0.000 & 0.440 & 0.010 + 1115 & 0.047 & 0.014 & -0.028 & 0.008 & & & & & 0.238 & 0.030 & 3.121 & 0.000 & 3 & 0.047 & -0.028 & 0.008 & 3.121 & 0.000 & 0.238 & 0.030 + 1144 & 7.2 & 0.7 & 8.1 & 0.8 & 9.3 & 0.9 & & & 0.253 & 0.200 & 0.000 & 1.000 & 1 & & & & & & 0.221 & 0.200 + 1190 & 0.1 & 0.4 & 0.2 & 0.4 & & & & & 0.161 & 0.200 & 21.713 & 0.000 & 3 & 0.117 & 0.2 & 0.4 & 21.713 & 0.000 & 0.161 & 0.200 + 1268 & 0.3 & 0.2 & -0.22 & 0.14 & & & & & 0.224 & 0.074 & 23.464 & 0.000 & 3 & 0.323 & -0.22 & 0.14 & 23.464 & 0.000 & 0.224 & 0.074 + 1282 & 1.7 & 0.3 & -0.81 & 0.16 & & & & & 0.375 & 0.051 & 0.124 & 0.883 & 1 & 1.700 & -0.81 & 0.16 & 0.124 & 0.883 & 0.375 & 0.051 + 1289 & 1.2 & 0.24 & -0.19 & 0.04 & & & & & 0.255 & 0.200 & 3.116 & 0.044 & 1 & 1.200 & -0.19 & 0.04 & 3.116 & 0.044 & 0.255 & 0.200 + 1315 & 7.3 & 0.6 & 0.0 & 0.0 & 6.1 & 0.4 & 1.4 & 0.7 & 0.251 & 0.156 & 3.984 & 0.001 & 2 & 0.977 & -0.6 & 0.3 & 15.823 & 0.000 & 0.274 & 0.079 + 1421 & 1.12 & 0.22 & 0.0103 & 0.0021 & & & & & 0.419 & 0.014 & 1.593 & 0.203 & 1 & 1.121 & 0.0103 & 0.0021 & 1.593 & 0.203 & 0.419 & 0.014 + 1476 & 0.3 & 0.15 & 0.17 & 0.15 & & & & & 0.282 & 0.037 & 27.938 & 0.000 & 3 & 0.305 & 0.17 & 0.15 & 27.938 & 0.000 & 0.282 & 0.037 + 1561 & 0.46 & 0.09 & -0.54 & 0.11 & & & & & 0.331 & 0.159 & 0.133 & 0.875 & 1 & 0.456 & -0.54 & 0.11 & 0.133 & 0.875 & 0.331 & 0.159 + 1566 & 1.5 & 0.3 & -0.96 & 0.19 & & & & & 0.422 & 0.012 & 11.255 & 0.000 & 3 & 1.537 & -0.96 & 0.19 & 11.255 & 0.000 & 0.422 & 0.012 + 1606 & 0.18 & 0.18 & -0.17 & 0.17 & & & & & 0.469 & 0.010 & 0.790 & 0.531 & 1 & 0.180 & -0.17 & 0.17 & 0.790 & 0.531 & 0.469 & 0.010 + 1664 & 1.21 & 0.24 & -0.44 & 0.09 & & & & & 0.222 & 0.200 & 3.843 & 0.021 & 1 & 1.211 & -0.44 & 0.09 & 3.843 & 0.021 & 0.222 & 0.200 + 1698 & 3.9 & 1.0 & 9.0 & 2.6 & 10.1 & 4.0 & 1.8 & 1.0 & 0.193 & 0.175 & 8.846 & 0.000 & 2 & 1.924 & -1.1 & 0.24 & 111270.044 & 0.000 & 0.323 & 0.010 + 1738 & 1.7 & 0.3 & -0.32 & 0.06 & & & & & 0.380 & 0.044 & 11.261 & 0.000 & 3 & 1.654 & -0.32 & 0.06 & 11.261 & 0.000 & 0.380 & 0.044 + 1817 & 0.9 & 0.18 & -0.17 & 0.03 & & & & & 0.410 & 0.018 & 1.624 & 0.197 & 1 & 0.897 & -0.17 & 0.03 & 1.624 & 0.197 & 0.410 & 0.018 + 1820 & 9.14 & 0.22 & 0.0 & 0.0 & 9.92 & 0.22 & & & 0.238 & 0.200 & 1.235 & 0.291 & 1 & 2.486 & -1.9 & 0.4 & 2.032 & 0.131 & 0.314 & 0.200 + 1852 & 0.96 & 0.23 & -0.16 & 0.19 & & & & & 0.308 & 0.144 & 1.959 & 0.118 & 1 & 0.960 & -0.16 & 0.19 & 1.959 & 0.118 & 0.308 & 0.144 + 1870 & 0.67 & 0.13 & 0.18 & 0.04 & & & & & 0.261 & 0.200 & 0.212 & 0.645 & 1 & 0.666 & 0.18 & 0.04 & 0.212 & 0.645 & 0.261 & 0.200 + 1900 & 11.82 & 0.29 & 13.4 & 0.2 & 8.35 & 0.25 & & & 0.235 & 0.200 & 0.303 & 0.738 & 1 & & & & & & 0.431 & 0.010 [ tab : thedata3 ] c c c c c c c c c c c c c c c c c c c c c 104 & 1.465 & 0.003 & qso & r & 9.8 & -0.71 & & 2.0 & 0.14 & & 1.033 & & 0.309 & & -26.9 & 3.0 & 17.2 & 2.0 & 1.057 & 0.348 + 157 & & & & 2 & 11.0 & -0.88 & & 1.3 & 0.18 & & & & & & -5.3 & 5.0 & 172.27 & 0.24 & 14.129 & 0.000 + 186 & 0.4509 & 0.001 & qso & r & 9.3 & -0.7 & & 1.9 & 0.14 & & 0.211 & 85.798 & 0.810 & 0.000 & -33.4 & 1.7 & 38.94 & 0.08 & 0.293 & 0.830 + 195 & 0.3466 & 0.0 & & r & 4.1 & -0.09 & & 0.7 & 0.07 & & 6.138 & 6.972 & 0.000 & 0.000 & -409.2 & 7.0 & 31.5 & 0.3 & 134.518 & 0.000 + 200 & 1.0365 & 0.0002 & qso & r & 11.2 & -0.83 & & 2.2 & 0.17 & & 0.911 & & 0.340 & & 20.3 & 1.9 & 104.55 & 0.09 & 5.166 & 0.000 + 246 & & & & r & 6.3 & -0.4 & & 1.7 & 0.17 & & 19.863 & 53.460 & 0.000 & 0.000 & -430.2 & 50.0 & 157.8 & 18.0 & 1.080 & 0.340 + 274 & 0.0254 & 0.0 & & r & 4.8 & -0.23 & & 0.7 & 0.07 & & 4.084 & 15.658 & 0.007 & 0.000 & 12.9 & 1.3 & 61.1 & 6.0 & 0.000 & 0.999 + 366 & & & & r & -2.1 & 0.53 & & 0.4 & 0.11 & & 25.123 & 206.815 & 0.000 & 0.000 & 8.0 & 0.9 & 91.9 & 11.0 & 7.730 & 0.000 + 383 & 0.255 & 0.002 & g & r & 6.4 & -0.35 & & 0.8 & 0.08 & & 3.331 & 26.047 & 0.005 & 0.000 & -55.2 & 4.0 & 51.19 & 0.18 & 10.098 & 0.000 + 391 & & & & r & 11.4 & -0.94 & & 2.3 & 0.19 & & 0.740 & & 0.390 & & -31.2 & 3.0 & 155.1 & 16.0 & 0.640 & 0.424 + 449 & 1.721 & 0.007 & qso & r & 11.0 & -0.94 & & 0.3 & 0.04 & & 0.739 & 0.880 & 0.594 & 0.475 & -123.9 & 14.0 & 133.8 & 15.0 & 2.599 & 0.074 + 460 & 0.8404 & 0.0004 & g & 2 & 12.5 & -0.98 & & 1.5 & 0.06 & & & & & & 112.7 & 2.7 & 126.17 & 0.12 & 11.003 & 0.000 + 519 & 1.589 & 0.001 & g & r & 12.6 & -1.0 & & 2.5 & 0.2 & & 1.469 & & 0.226 & & 8.7 & 5.0 & 132.36 & 0.21 & 4.295 & 0.005 + 571 & 2.0225 & 0.0003 & qso & r & 14.2 & -1.19 & & 2.8 & 0.24 & & 0.119 & & 0.730 & & -20.3 & 0.7 & 34.4 & 0.03 & 1.785 & 0.051 + 696 & 0.4061 & 0.0 & qso & r & 14.1 & -1.16 & & 2.8 & 0.23 & & 20.450 & & 0.000 & & 18.9 & 4.0 & 102.31 & 0.18 & 4.246 & 0.005 + 725 & 1.946 & 0.002 & qso & r & 4.6 & -0.19 & & 0.4 & 0.04 & & 1.592 & 1.506 & 0.158 & 0.197 & -98.5 & 11.0 & 92.6 & 11.0 & 1.675 & 0.187 + 730 & 0.3216 & 3e-05 & g & r & 11.2 & -0.8 & & 2.2 & 0.16 & & 15.003 & & 0.000 & & 47.6 & 6.0 & 14.66 & 0.26 & 12.417 & 0.000 + 813 & & & & n & & & & & & & & & & & 287.3 & 29.0 & 72.8 & 7.0 & 7.347 & 0.007 + 896 & & & & n & & & & & & & & & & & -8.5 & 0.9 & 47.3 & 5.0 & 0.740 & 0.390 + 924 & & & & r & 11.9 & -1.02 & & 0.7 & 0.07 & & 4.222 & 11.187 & 0.001 & 0.000 & -101.2 & 12.0 & 1.7 & 0.2 & 0.107 & 0.898 + 998 & 0.518 & 0.0 & qso & r & 13.7 & -1.13 & & 0.9 & 0.09 & & 3.530 & 24.427 & 0.014 & 0.000 & -51.5 & 5.0 & 119.9 & 12.0 & 0.076 & 0.782 + 1030 & 1.469 & 0.001 & qso & r & 12.9 & -1.05 & & 2.6 & 0.21 & & 0.084 & & 0.772 & & 124.6 & 5.0 & 43.37 & 0.22 & 4.974 & 0.000 + 1031 & & & & r & 12.8 & -1.0 & & 2.6 & 0.2 & & 0.373 & 1.463 & 0.689 & 0.226 & 31.7 & 4.0 & 173.1 & 20.0 & 1.505 & 0.222 + 1062 & 0.0733 & 0.0002 & g & r & 11.5 & -0.86 & & 2.3 & 0.17 & & 1.319 & & 0.251 & & -10.3 & 1.2 & 47.72 & 0.05 & 1.353 & 0.248 + 1098 & 2.2627 & 9.3e-05 & qso & r & 8.9 & -0.64 & & 0.4 & 0.05 & & 0.198 & 6.717 & 0.898 & 0.001 & 3.3 & 2.9 & 46.6 & 0.13 & 3.213 & 0.022 + 1115 & 0.2196 & 0.0 & g & r & 10.4 & -0.73 & & 2.1 & 0.15 & & 0.436 & 0.508 & 0.647 & 0.476 & 86.6 & 5.0 & 138.79 & 0.24 & 21.714 & 0.000 + 1144 & & & & r & 5.0 & -0.28 & & 1.0 & 0.06 & & 0.700 & 29.100 & 0.497 & 0.000 & 13.5 & 1.4 & 112.6 & 11.0 & 0.215 & 0.643 + 1190 & 1.178 & 0.0 & qso & 2 & 3.5 & -0.08 & & 0.4 & 0.05 & & & & & & -15.0 & 2.7 & 48.16 & 0.12 & 8.165 & 0.000 + 1268 & 0.174 & 0.001 & g & r & 10.9 & -0.76 & & 1.2 & 0.12 & & 1.597 & 41.617 & 0.188 & 0.000 & 90.5 & 2.3 & 58.51 & 0.11 & 4.281 & 0.000 + 1282 & & & & r & 13.3 & -1.2 & & 0.7 & 0.08 & & 7.080 & 11.788 & 0.000 & 0.000 & -103.5 & 12.0 & 73.7 & 8.0 & 0.735 & 0.480 + 1289 & & & & r & 9.8 & -0.83 & & 0.4 & 0.04 & & 0.804 & 2.670 & 0.547 & 0.030 & -96.1 & 11.0 & 70.2 & 8.0 & 0.920 & 0.398 + 1315 & & & & r & 11.5 & -0.88 & & 0.5 & 0.05 & & 1.057 & 8.916 & 0.366 & 0.000 & 47.1 & 1.4 & 65.86 & 0.06 & 9.713 & 0.000 + 1421 & & & & r & 10.8 & -0.92 & & 0.9 & 0.09 & & 9.667 & 11.889 & 0.000 & 0.000 & -85.5 & 10.0 & 50.3 & 6.0 & 0.297 & 0.743 + 1476 & 1.5338 & 0.0019 & bllac & r & 4.7 & -0.19 & & 0.9 & 0.04 & & 3.626 & & 0.057 & & 283.4 & 1.3 & 50.13 & 0.06 & 6.806 & 0.000 + 1561 & 0.2104 & 0.0 & g & r & 13.3 & -1.09 & & 2.7 & 0.22 & & 5.759 & 11.652 & 0.003 & 0.001 & 151.3 & 17.0 & 175.1 & 20.0 & 0.516 & 0.597 + 1566 & & & & c & 2.94 & -0.962 & -0.0769 & 0.27 & 0.027 & 0.0022 & 2.628 & 0.838 & 0.022 & 0.501 & -109.1 & 13.0 & 19.4 & 2.2 & 3.269 & 0.038 + 1606 & 0.1952 & 0.0002 & g & r & 19.6 & -1.8 & & 4.0 & 0.4 & & 0.042 & & 0.838 & & -198.1 & 10.0 & 53.7 & 0.4 & 5.625 & 0.000 + 1664 & & & & r & 0.19 & 0.23 & & 0.04 & 0.05 & & 0.652 & 28.141 & 0.521 & 0.000 & 5.6 & 0.6 & 142.8 & 16.0 & 0.619 & 0.539 + 1698 & 0.7186 & 4e-05 & qso & r & 7.85 & -0.49 & & 0.28 & 0.03 & & 0.094 & 0.141 & 0.964 & 0.868 & -6.3 & 0.8 & 89.25 & 0.04 & 5.063 & 0.000 + 1738 & & & & r & 8.94 & -0.77 & & 0.09 & 0.01 & & 0.085 & 0.100 & 0.987 & 0.960 & -94.3 & 11.0 & 136.1 & 16.0 & 0.212 & 0.809 + 1817 & & & & r & 14.5 & -1.24 & & 0.6 & 0.07 & & 0.247 & 8.609 & 0.912 & 0.000 & -62.9 & 7.0 & 23.1 & 2.7 & 1.379 & 0.252 + 1820 & 0.0784 & 0.0 & g & r & 12.6 & -1.05 & & 2.5 & 0.21 & & 0.272 & 2.098 & 0.762 & 0.147 & -33.6 & 4.0 & 68.5 & 8.0 & 0.135 & 0.874 + 1852 & 0.387 & 0.001 & g & r & 12.2 & -1.0 & & 2.4 & 0.2 & & 9.087 & & 0.003 & & 3.5 & 1.0 & 145.72 & 0.04 & 0.539 & 0.655 + 1870 & & & & n & & & & & & & & & & & 1.31 & 0.13 & 59.4 & 6.0 & 2.888 & 0.089 + 1900 & 2.201 & 0.0 & qso & r & 5.8 & -0.33 & & 1.2 & 0.07 & & 2.437 & & 0.119 & & -7.8 & 0.9 & 179.1 & 21.0 & 0.081 & 0.922 [ tab : thedata4 ] c c c c c c c c c c c c c c c c c c c 104 & -33.2 & 2.4 & 2414.8 & 72.4 & 32.64 & 0.35 & 1.47 & 0.11 & < & 15.6 & & < & 13.9 & & & & 28.4 & 6.0 + 157 & -11.6 & 6.0 & 1069.5 & 35.1 & 103.2 & 2.73 & 19.9 & 0.8 & & 86.7 & 1.0 & & 20.4 & 2.2 & -81.1 & 0.0 & & + 186 & -28.5 & 5.0 & 804.4 & 27.0 & 14.25 & 0.25 & 2.1 & 0.13 & & 30.8 & 1.6 & < & 17.5 & & -63.6 & 0.2 & 26.8 & 5.0 + 195 & 12.0 & 1.2 & 2050.7 & 61.5 & 61.84 & 0.32 & 3.19 & 0.14 & < & 19.6 & & < & 16.6 & & & & 26.8 & 27.0 + 200 & 22.5 & 3.0 & 3897.6 & 116.9 & 26.17 & 0.35 & 0.7 & 0.1 & < & 17.1 & & < & 15.5 & & & & 28.3 & 6.0 + 246 & 13.6 & 12.0 & 468.2 & 14.1 & 6.46 & 0.37 & 1.47 & 0.14 & < & 16.9 & & < & 13.9 & & & & & + 274 & -2.9 & 8.0 & 388.3 & 11.7 & 17.65 & 0.67 & 4.36 & 0.23 & < & 18.8 & & < & 15.3 & & & & 23.8 & 23.0 + 366 & 10.2 & 6.0 & 493.6 & 17.4 & 13.16 & 0.34 & 2.74 & 0.16 & & 14.4 & 3.0 & < & 17.0 & & -33.7 & 1.6 & & + 383 & -39.3 & 2.0 & 1650.9 & 64.2 & 35.06 & 0.36 & 2.31 & 0.14 & & 18.4 & 2.4 & & 15.3 & 2.8 & -76.1 & 0.6 & 26.5 & 26.0 + 391 & -36.9 & 2.3 & 624.8 & 20.5 & 27.16 & 0.3 & 5.41 & 0.21 & & 38.1 & 1.4 & < & 16.3 & & 77.0 & 0.1 & & + 449 & -108.3 & 5.0 & 263.6 & 7.9 & 11.3 & 0.31 & 4.59 & 0.21 & < & 16.7 & & < & 14.2 & & & & 27.7 & 26.0 + 460 & -22.1 & 1.9 & 3650.3 & 122.1 & 180.33 & 1.74 & 6.51 & 0.25 & & 31.7 & 1.6 & < & 16.8 & & 25.3 & 0.0 & 28.1 & 27.0 + 519 & -34.6 & 4.0 & 2122.3 & 63.7 & 24.6 & 0.33 & 1.17 & 0.11 & < & 17.3 & & < & 15.3 & & & & 28.6 & 6.0 + 571 & -16.8 & 1.1 & 2180.0 & 65.4 & 130.67 & 0.64 & 5.7 & 0.2 & < & 19.6 & & < & 16.1 & & & & 28.9 & 6.0 + 696 & 40.9 & 1.4 & 2922.8 & 95.0 & 50.91 & 0.31 & 2.03 & 0.12 & & 42.0 & 1.3 & < & 16.7 & & -11.4 & 0.0 & 27.3 & 5.0 + 725 & -101.6 & 1.5 & 707.9 & 21.2 & 40.7 & 0.27 & 6.02 & 0.21 & < & 15.9 & & < & 15.7 & & & & 27.9 & 27.0 + 730 & 55.5 & 1.8 & 8753.3 & 262.6 & 67.24 & 0.7 & 0.8 & 0.1 & < & 17.8 & & < & 16.7 & & & & 27.5 & 5.0 + 813 & -17.3 & 8.0 & 645.1 & 21.9 & 31.21 & 0.97 & 8.1 & 0.4 & & 54.6 & 1.2 & & 27.7 & 1.7 & 76.3 & 0.2 & & + 896 & -10.4 & 1.0 & 463.4 & 15.6 & 60.49 & 0.29 & 14.2 & 0.5 & & 30.3 & 1.6 & < & 17.3 & & -32.6 & 0.3 & & + 924 & -91.6 & 10.0 & 387.2 & 13.7 & 6.47 & 0.29 & 1.78 & 0.14 & & 15.0 & 2.9 & < & 18.1 & & 46.5 & 2.4 & & + 998 & -91.6 & 6.0 & 1957.1 & 58.7 & 11.79 & 0.31 & 0.6 & 0.1 & < & 18.6 & & < & 16.3 & & & & 27.4 & 26.0 + 1030 & 1.9 & 2.1 & 1984.1 & 68.0 & 33.1 & 0.29 & 1.94 & 0.12 & & 24.2 & 1.9 & < & 18.8 & & -16.7 & 0.1 & 28.5 & 6.0 + 1031 & 6.5 & 1.8 & 2541.9 & 85.0 & 41.26 & 0.35 & 1.94 & 0.12 & & 31.9 & 1.6 & < & 15.7 & & 51.6 & 0.0 & & + 1062 & -13.8 & 0.5 & 4067.1 & 138.5 & 167.96 & 0.35 & 4.68 & 0.19 & & 26.1 & 1.8 & < & 16.2 & & -27.4 & 0.0 & 25.8 & 5.0 + 1098 & 33.4 & 2.5 & 996.1 & 29.9 & 22.74 & 0.26 & 2.4 & 0.13 & < & 19.8 & & < & 17.3 & & & & 28.4 & 27.0 + 1115 & -4.3 & 2.8 & 6187.8 & 239.1 & 33.58 & 0.41 & 0.6 & 0.1 & & 19.9 & 2.2 & & 16.1 & 2.7 & -70.1 & 0.1 & 27.0 & 5.0 + 1144 & 1.4 & 13.0 & 236.6 & 7.1 & 5.86 & 0.32 & 2.57 & 0.19 & < & 16.5 & & < & 15.3 & & & & & + 1190 & 4.8 & 4.0 & 573.2 & 17.2 & 19.61 & 0.31 & 3.48 & 0.15 & < & 19.0 & & < & 18.0 & & & & 27.4 & 27.0 + 1268 & -78.3 & 0.6 & 11534.0 & 373.7 & 191.62 & 0.55 & 2.07 & 0.12 & & 43.3 & 1.3 & < & 16.6 & & -71.0 & 0.0 & 27.0 & 26.0 + 1282 & -94.6 & 7.0 & 220.8 & 6.6 & 8.51 & 0.28 & 4.23 & 0.21 & < & 17.1 & & < & 15.8 & & & & & + 1289 & -92.7 & 5.0 & 146.9 & 4.4 & 12.4 & 0.29 & 8.8 & 0.3 & < & 19.8 & & < & 16.6 & & & & & + 1315 & 18.9 & 4.0 & 2560.1 & 76.8 & 31.79 & 0.51 & 1.46 & 0.11 & < & 15.7 & & < & 14.1 & & & & & + 1421 & -81.9 & 1.6 & 223.0 & 8.0 & 23.46 & 0.18 & 13.8 & 0.5 & & 33.4 & 1.6 & & 29.6 & 1.7 & 57.1 & 3.3 & & + 1476 & -19.0 & 2.7 & 814.0 & 24.4 & 24.79 & 0.3 & 3.16 & 0.14 & < & 18.3 & & < & 16.4 & & & & 27.8 & 6.0 + 1561 & -15.4 & 7.0 & 2210.8 & 66.3 & 11.03 & 0.33 & 0.5 & 0.1 & < & 18.8 & & < & 15.2 & & & & 26.5 & 5.0 + 1566 & -65.2 & 5.0 & 888.1 & 31.3 & 15.13 & 0.3 & 1.8 & 0.12 & & 15.1 & 2.9 & < & 17.7 & & -36.5 & 0.8 & & + 1606 & 33.6 & 10.0 & 1361.7 & 45.4 & 9.51 & 0.46 & 0.82 & 0.11 & & 33.1 & 1.5 & < & 16.5 & & -31.9 & 0.1 & 26.3 & 5.0 + 1664 & 23.8 & 10.0 & 185.9 & 5.6 & 7.1 & 0.34 & 4.12 & 0.25 & < & 16.8 & & < & 16.2 & & & & & + 1698 & -10.1 & 0.8 & 2508.8 & 75.3 & 86.79 & 0.34 & 3.91 & 0.15 & < & 17.0 & & < & 16.0 & & & & 27.7 & 27.0 + 1738 & -94.2 & 6.0 & 71.5 & 2.2 & 11.48 & 0.27 & 16.7 & 0.7 & < & 18.9 & & < & 17.3 & & & & & + 1817 & -76.5 & 1.3 & 1322.0 & 42.7 & 45.98 & 0.29 & 4.56 & 0.18 & & 44.3 & 1.3 & < & 15.4 & & -57.7 & 0.1 & & + 1820 & -80.1 & 12.0 & 1093.3 & 32.8 & 7.54 & 0.38 & 0.72 & 0.11 & < & 18.5 & & < & 17.4 & & & & 25.3 & 5.0 + 1852 & 6.9 & 0.6 & 2125.1 & 75.0 & 118.61 & 0.33 & 5.88 & 0.23 & & 14.5 & 3.0 & < & 17.4 & & 29.8 & 0.4 & 27.1 & 5.0 + 1870 & 0.6 & 0.9 & 1238.7 & 41.1 & 85.22 & 0.37 & 8.13 & 0.29 & & 34.1 & 1.5 & < & 15.8 & & -15.2 & 0.1 & & + 1900 & -9.5 & 1.5 & 602.6 & 18.1 & 44.59 & 0.3 & 7.62 & 0.25 & < & 16.5 & & < & 16.1 & & & & 28.0 & 6.0 [ tab : thedata5 ]
we here seek to present a collection of systematic effects that may guide the interpretation of data within the catalog . in order to ensure that the catalog is providing an accurate measure of astrophysical effects
, we discuss these systematics , limitations , and other possible issues that could affect the quality of the catalog and of which a user of the catalog needs to be aware .
note that each source in the catalog is considered unresolved at all wavelengths , and is still unresolved at the 45 arcsec resolution of the nvss .
the angular size of each source in the nvss is included in the catalog .
many of the sources included via the cross - matching process are likely affected by beam depolarization .
this occurs as all observations have an intrinsic angular resolution that is determined by the size of the resolving beam .
consider an extended source that is not uniformly polarized : if the radio telescope has a resolution that is coarser ( i.e. an observing beam size that is larger ) than the angular scale over which the source polarization is coherent , then a given synthesized beam contains regions with different evpas .
the synthesized beam therefore averages out the polarization of the source , and the measured polarization will be less than the true source polarization ( see appendix [ externaldispersion ] ) .
this ` beam depolarization ' can not be corrected for , even using a polarized sed with data at multiple wavelengths , which can not separate the astrophysical and beam - induced contributions .
the only way to obtain the true source polarization would be to observe with a telescope that has better angular resolution relative to the original observations .
complications can arise due to the spectral index and the multiple components of a radio source .
an unresolved source may contain many sub - structures , so that within the resolving beam is a flat - spectrum compact nucleus and two steep - spectrum lobes / jets .
these components become fainter or brighter with observing frequency , such that it is not trivial to confirm that we are observing the same dominant source component at all frequencies .
again , observations with increased angular resolution would be useful in separating out these various source contributions .
furthermore , for future studies , rm synthesis can in principle distinguish different emitting regions via their separation in the faraday spectrum , even if the components are not angularly resolved .
the polarized fraction can also help in separating these components , as @xmath26 is typically higher at frequencies where steep - spectrum jets / lobes dominate .
similarly , measurements of total intensity spectra may also help in resolving this issue .
not only is there an expected difference in the spectral index from different emitting regions , but sources that contain multiple components within an individual resolution element may also show spectral curvature although a number of other mechanisms can also cause such curvature .
the data in the catalog are obtained over many different epochs , with the time between observations spanning up to @xmath147 years .
both flux- and angle- variability in the sample may be significant .
nevertheless , it is feasible that time - variable sources will experience similar amounts of positive and negative perturbations in total intensity and/or polarized fraction , i.e. increases and decreases will occur in equal measure .
in addition , it is also feasible that the time - variations themselves will be simultaneously similar for both the total and polarized intensity components leading to an unchanged polarized fraction , i.e. an increase in total intensity is directly proportional to the increase in polarized intensity .
together with the large number of compiled measurements , this will serve such that any time - variability will just appear as increased variance in the estimates of depolarization and spectral index . furthermore , the catalog has also been cross - matched with optical identifications for each source @xcite
. it should be possible to distinguish between source types with differing variability , and a user of the catalog is able to study sources by classification e.g. galaxies and blazars .
for the purposes of statistical studies , any derived results are again only affected in terms of increased sample variance . to assist the user in distinguishing potentially time - variable sources
, we provide a data quality flag that is calculated based on statistical tests during fitting of the sed ( see section [ sourcefitting ] ) .
consider a source that has no intrinsic polarization , so that @xmath148 . as the measurements will be derived from images of @xmath46 and @xmath47 that contain gaussian noise ,
it is clear that a measurement of @xmath149 from an image will always yield @xmath150 .
this over - estimation of polarized intensity due to @xmath149 being positive - definite is known as ` rician bias ' . for higher s / n ratios ,
the amount of this bias becomes increasingly small .
nevertheless , at lower s / n ratios an estimator of the bias is essential to retrieve the true polarized intensity .
the effects of rician bias , and the bias estimator have often changed in their use over the period in which the measurements have been collected . in many cases ,
the exact method used has not been stated in the literature .
as all of our sources are detected at @xmath151 at 1.4 ghz @xcite , and these sources tend to increase in polarized fraction at higher frequency , we make the assumption that each source is of sufficient s / n that the rician statistics are approximately gaussian under such circumstances the bias in measuring a positive - definite quantity can be considered to be negligible .
additional systematics in our catalog may also result from the different ways in which source properties have been measured and parameterized , i.e. the source - finding technique used .
we assume that such effects behave as random errors on each datum , and have negligible effects on an sed .
modern single - dish measurements are typically done in ` on - off ' or ` dual - beam ' mode , so that any smooth galactic background can in principle be subtracted .
nevertheless , the processing and observational details have not been stated in the literature for the single dish measurements that are included in the catalog some of which were taken @xmath147 years ago .
these measurements may not have removed a contribution from surrounding diffuse background components .
such a systematic would increasingly affect our measurements in proximity to the galactic plane .
any subtle correlations in the data at low galactic latitudes should be treated with caution .
many individual catalogs also ignore the total intensity contribution to the error in polarized fraction . following standard error propagation
, the fractional uncertainties in the polarized and total intensity measurements should be summed in quadrature as given by @xmath152 where @xmath26 is the polarized fraction , @xmath149 is the polarized intensity ( equal to @xmath153 , assuming that rician bias is negligible ) , @xmath44 is the total intensity , and @xmath154 is the one sigma uncertainty in these quantities . for sources that are weak in total intensity , or equivalently have a large fractional error in @xmath155 , neglecting this total intensity contribution can significantly underestimate the error in the polarized fraction . for our catalog , this additional contribution has been recalculated when enough information has been available to do so . in practice ,
antenna feeds never have a perfect response to polarization .
these imperfections can be modeled as leakage from stokes @xmath44 into polarization , and is known as ` instrumental polarization ' or ` polarization leakage ' .
error contributions from this instrumental polarization are by convention also typically ignored when calculating the errors in the polarized fraction . however
, if leakages are calculated accurately to within @xmath156% then we would expect an additional error in each measurement .
such an error can be trivially handled as an additive component to the uncertainty in each datum .
the seds that we fit to the data in section [ sourcefitting ] are potentially affected by the stated precision of each datum , which we use for weighting our data .
the effects of outliers in our data are also of interest .
there are two types of outliers : ( i ) those that are inaccurate , and ( ii ) those that are both inaccurate and imprecise .
case ( i ) relates to occasional false cross - matches , or poor measurements in the original surveys that occur due to calibration or source fitting issues .
these cause systematic errors in our derived source parameters .
our low fdr , combined with the large number of measurements that are believed to be accurate , ensures that we still get reasonable values .
this was tested on a simulated distribution of evpa measurements as a function of @xmath7 , and we find that a single outlier affects the calculated rm by less than 10 rad m@xmath144 for sources with five measurements .
case ( ii ) is more simply dealt with , as it occurs due to large measurement errors that in all cases are appropriately weighted during our fitting procedure .
the effects of outliers therefore tend to be reduced by the fitting and weighting procedure , but still have the property of increasing the variance in any estimated parameters .
the catalog therefore works well for statistical measurements of polarized source properties our fdr suggests that the number of inaccurate matches represents @xmath157% of the total sample .
nevertheless , we recommend that appropriate caution is used if attempting to derive physical properties from the polarization sed of a single source .
we have produced a new , broadband radio polarization catalog .
our catalog uses the nvss rm catalog @xcite as a reference , with our ` full - catalog ' incorporating 37,543 radio sources , of which 25,649 have measurements of both the rotation measure ( rm ) and total intensity spectral index .
the catalog is expected to consist of mostly classical radio galaxies and quasars .
normal galaxies are only expected to constitute a significant fraction of the 1.4 ghz polarized source counts at @xmath1431 mjy in total intensity , and at 1 to @xmath158 jy in polarized intensity ( e.g. * ? ? ?
* ; * ? ? ?
* ) approximately two orders of magnitude below the nvss sensitivity limit . in this catalog ,
we have provided power - law depolarization measurements for 1,171 radio sources with @xmath123 polarization measurements at independent frequencies , and a full , modeled sed in both polarized and total intensity for 951 radio sources with @xmath159 multiwavelength polarization data . these 951 sources with modeled seds are available as a secondary ` sed - catalog ' , constituting a subset of our primary data product .
the catalog includes spectroscopic redshifts for 620 of these polarized seds .
based on the model selection criteria and various statistical tests described in the main text , we consider 533 of these seds to be excellent an improvement of an order of magnitude over any other published sample of polarized seds .
the breakdown of seds in our catalog by different depolarization models suggests that 42.2% of sources are best described by a tribble - law , 20.6% by repolarization , 19.4% by spectral depolarization , 9.0% by a rossetti mantovani law , and 8.8% by a burn - law .
see section [ sourcefitting ] and table [ tab : models ] for further detail on how these models are derived .
note that these add up to 100% , as we have not included a ` no best fit ' category . even if the data quality flag is taken into account , and ` poor ' fitting models are excluded from our analysis , the percentages are changed such that 42.5% of sources are best described by a tribble - law , 16.0% by repolarization , 21.9% by spectral depolarization , 11.0% by a rossetti mantovani law , and 8.6% by a burn - law .
note that the fraction of repolarizers is substantially affected this is unsurprising given that we use a power law to fit the repolarizers , and that this is based on the flexibility of the power law , rather than an exact physical model of what could be expected from different emission regions .
note that such depolarization models are only applicable if the same emitting region is probed at all frequencies as we shall discuss in this section , a key physical inference from our catalog is that this is likely not the case . a density plot showing the number of sources with given total intensity and polarized spectral indices is shown in fig . [ i_versus_p_alphas ] .
there is a clear divide in the depolarization of radio sources that occurs at @xmath160 , with steep - spectrum sources showing larger depolarization ( @xmath161 ) , and flat - spectrum sources tending to retain an essentially constant polarized fraction as a function of @xmath11 ( @xmath162 ) .
we split the sample into two subsets of @xmath24 values , consisting of either flat - spectrum ( @xmath163 ) or steep - spectrum ( @xmath164 ) sources .
sources with @xmath165 are not included in order to minimize cross - contamination between our samples , although our conclusions are not affected by changing this @xmath0-threshold . the two - sample kolmogorov
smirnov test on @xmath24 provides a @xmath105-value of @xmath166 , which indicates a very low probability of the two samples being this different , or more so , if drawn from the same distribution .
this is consistent with a population of both core- and jet - dominated sources , in which steep - spectrum sources measure the optically thin lobes and jets ejected by an agn , while flat - spectrum sources measure the optically thick region surrounding the central black hole of a radio galaxy .
as the total intensity spectral index is an intrinsic source property , this implies that depolarization must mainly occur within the local source environment , rather than being due to intervening faraday screens .
we argue that such an effect is unlikely to be due to confounding variables , with independent samples selected at 20 ghz ( i.e. core - dominated sources ) also finding no statistically significant evidence of a relationship between the fraction of polarization and frequency @xcite .
one may argue that this could be caused by beam depolarization , with the flat - spectrum central engine being more compact , and tending to have smaller rm dispersion than the steep - spectrum sources .
in such a case , the compact sources would show _ some _ depolarization ; instead , the distribution of @xmath24 for flat - spectrum sources is approximately symmetric about a mean value of zero ( see the red histogram in fig . [ i_versus_p_alphas ] ) .
in fact , the distribution appears very slightly skewed with a few more sources showing @xmath167 rather than @xmath161 .
for pure beam depolarization , @xmath24 must _ always _ be @xmath168 .
consequently , while there are possible alternative interpretations of our results , any such hypotheses are restricted in that they must be able to explain this approximately symmetric distribution of the flat - spectrum sources around @xmath162 . as the most consistent explanation for the repolarizing ( @xmath167 ) sources
is that this is a consequence of different emitting regions being probed at different frequencies , the symmetrical distribution appears to indicate that different emitting regions are also giving rise to the depolarizing ( @xmath161 ) sources .
previous studies have suggested that intervening galaxies with a clumpy interstellar medium are important for explaining depolarization of background sources ( e.g. * ? ? ?
* ; * ? ? ?
as the likelihood of intersecting an intervenor increases for sources at high @xmath169 , one could hypothesise from fig .
[ i_versus_p_alphas ] that steep - spectrum sources are more distant .
however , such a scenario can not explain why the flat - spectrum sources have an approximately symmetrical distribution that is centred about @xmath162 .
similarly , it also does not explain the presence of repolarizing sources . as a further point in favour of our interpretation , due to relativistic beaming
, one expects the flat - spectrum ( core - dominated ) sources to be the more distant for a flux - limited selected sample such as the nvss .
furthermore , there is evidence from the catalog that allows one to completely rule out the currently proposed models of ` partial coverage ' ( farnes et al . in preparation ) .
our data also show that repolarization can occur in some sources , although importantly this is predominantly associated with flat - spectrum ( i.e. presumably core - dominated ) sources .
our classification algorithm finds that 203 sources ( 21% ) of our sample show repolarization .
[ i_versus_p_alphas ] shows that the lobe - dominated sources are mostly located with @xmath170 , however @xmath147% of the population of core - dominated sources have @xmath167 . together with the flat - spectrum sources tending to retain a relatively constant polarized fraction as a function of @xmath11 , we interpret this as a consequence of the core s optical thickness , with synchrotron self - absorption leading to the observation of different emitting regions within the source .
this results in the observation that a source is repolarizing ( and even depolarizing ) , when we are instead likely detecting regions of increasingly ordered magnetic field that are at greater distances from the central engine .
optically thick emission , although rarely considered , can result in more complicated behavior in a polarized sed including oscillations and even constant polarized fractions at increasing wavelengths ( e.g. * ? ? ? * ) .
an example of a possible oscillating source , for which our algorithm fits a flat value of @xmath24 , is shown in the sed for nvss j003820020740 in fig .
our data therefore support the hypothesis that flat - spectrum radio polarization measurements probe polarization structure in the optically thick quasar core , with the sources tending to provide a constant polarized fraction over a large range in wavelength merely due to the sparse sampling in @xmath11-space which undersamples the oscillatory behaviour .
as repolarization is also present for some lobe - dominated sources , it appears to indicate that a combination of spectral index and depolarization effects can always cause one to probe different polarized emitting regions at different frequencies .
nevertheless , there are only a relatively small number of repolarizing steep - spectrum objects , unlike the symmetric flat - spectrum distribution about @xmath171 .
the steep - spectrum sample _ can not _ therefore be similar to the flat - spectrum sample but with some additional depolarizing component .
we argue that the most consistent physical description is of a divide based on the local source environment , with the large number of repolarizers showing that we do not sample similar emitting regions at different frequencies . as a substantial number of sources in our sample
are repolarizing , this indicates that the run of polarized fraction with observing wavelength is substantially affected by different emitting regions within a source . from an observational perspective
this is problematic , as attempts to derive any particular depolarization law from a run of polarized fraction with @xmath11 must attempt to separate different emitting regions .
this is normally not considered , as the depolarization is typically considered to be most strongly affected by screens of material along the line of sight .
nevertheless , the results from our new and larger sample suggest that faraday depolarization in extragalactic sources is largely an ` internal ' effect that occurs within the local source environment .
this has implications for the physical interpretation of polarized seds , and complicates the interpretation of broadband polarization observations for the purposes of measuring seds and rms .
in such a scenario , a balance must be obtained between broadband measurements ( which provide a larger lever in @xmath7-space ) , and narrow - band observations ( which are more likely to probe the same source component across a frequency range of interest ) .
this has implications for future broadband polarimetric surveys .
high - resolution follow - up observations will be required to further understand the effects of different emitting regions on polarized seds . for the steep - spectrum objects in our catalog ,
there is weak evidence of a bimodal distribution in @xmath24 ( see fig .
[ i_versus_p_alphas ] ) , with two distinct peaks .
if true , this would suggest two populations of depolarizer , possibly consistent with previous studies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
stronger conclusions on the garrington laing and liu pooley effects will require extensive further study of these data . this will require a cleaner sample that is separated by source type ( e.g. by using the optical identifications that are included in the catalog see appendix [ appendix - a ] ) , and through high - resolution follow - up observations to identify the morphology of both the source and the polarized emitting region . such studies , albeit with smaller sample size , have recently identified relations between rm , depolarization and the complexity of the faraday spectrum for polarized sources concluding that the depolarization originated within the sources , for instance in their radio lobes , or in intervening galaxies on the line of sight . as a point of further interest ,
plots of the distribution of parameters in our catalog for the various models selected by the bic are shown in fig .
[ c_plot ] .
this shows that the majority of sources have a maximum polarized fraction of @xmath93% , with some sources having polarized fractions up to @xmath172% .
while the majority of sources show regular depolarization , those sources that are peaked tend to have a maxima in polarized fraction at a wavelength of @xmath158 to @xmath173 cm .
sources that maintain a constant polarized fraction out to large wavelengths ( a rossetti mantovani law , i.e. that include a coefficient @xmath119 ) , tend to remain polarized at only the @xmath174 to @xmath175% level . the physical model that is selected as corresponding to the coefficient values can be determined using table [ tab : models ] .
we have presented a multiwavelength catalog of radio polarization that increases the number of well - defined polarized seds by over an order of magnitude .
the resulting polarized seds have a frequency range of 0.4 ghz to 100 ghz , and our best sources are constrained by up to 56 independent polarization measurements .
we have used a k - dimensional tree for the cross - matching , reducing the computational expense by a factor of @xmath176 , and have used an automated classification algorithm based on the bayesian information criterion to distinguish between different models for the faraday depolarization .
the catalog also contains constraints on the total intensity spectral index and curvature , the broadband rm , spectroscopic redshift , angular size , and estimated non - thermal rest - frame luminosity at 1.4 ghz .
the catalog will allow a number of parameters to be explored in the effort to understand cosmic magnetic fields . in this paper
, we have found that our sample is consistent with two populations of core- and jet - dominated sources based on the clustering in the plane of the polarized fraction versus total intensity spectral indices .
this is consistent with the optically - thin jet / lobe - dominated sources undergoing significantly more depolarization relative to the optically - thick core - dominated sources .
such a connection implies that radio source depolarization predominantly occurs within the local source environment , rather than being due to intervening faraday screens .
importantly , the catalog will be of particular use in @xmath6-correcting polarized seds into the source rest - frame .
these @xmath6-corrections can be performed by using the statistical tests provided in the catalog to select good quality seds , and to use the depolarization model selected by the bic .
the @xmath6-corrected polarized fraction , i.e. in the source rest - frame , for equivalent emission to that at 1.4 ghz and @xmath177 is then the corresponding polarized fraction at @xmath178 cm .
the @xmath6-corrected faraday depth can be obtained using a multiplicative factor of @xmath179 to correct for the effects of cosmological expansion , however this assumes that all of the faraday rotation is occurring at the source .
such a correction will break down if there is any galactic or intervening contribution .
the @xmath179 correction to the faraday depth also assumes a linear relationship between the evpa and @xmath7 if the relationship is non - linear and the faraday depth measured using a narrowband , then it will be necessary to @xmath6-correct the polarization angle sed in a similar way to @xmath6-correcting the polarized fraction , i.e. by sliding the sed and then reestimating the faraday depth .
this catalog is an enabling step for such @xmath6-corrected studies , as an increased sample of polarized seds will not be available until the advent of the ska and other next generation facilities such as askap which will yield the necessary broadband polarization data in combination with a redshift .
the results of the @xmath6-corrections themselves constitute an extensive additional study , as such studies of magnetic field evolution are hindered by the ability to classify different source types ( e.g. normal versus active galaxies , by viewing angle , core- versus lobe- dominated ) , by luminosity effects / malmquist bias , and possibly by other evolutionary effects such as changes in the bulk lorentz factor of radio jets .
furthermore , as @xmath6-correction of the rm into the rest - frame by a factor of @xmath179 assumes that the faraday rotating medium is all local to the source reliable removal of galactic contributions , together with measures of the magnetoionic content along typical lines of sight , are essential to probe the evolution of cosmic magnetism .
other parameters such as the total intensity spectral indices will assist in the process of source classification and correcting for luminosity effects , although higher resolution data are necessary to reveal how reliably this separates core- and lobe - dominated sources .
in addition to @xmath6-corrections , there are also many subsidiary applications , including studies of where faraday rotation occurs along the line of sight , the spectral index versus redshift ( e.g. * ? ? ?
* ) , measurements of the galactic magnetic field ( e.g. * ? ? ?
* ) , and investigations of intervening mg ii absorption systems and their effects on sed type and depolarization ( e.g. * ? ? ?
* ; * ? ? ?
previous similar catalogs have not focused on polarization measurements ( e.g. * ? ? ?
this therefore is the most comprehensive catalog of polarization measurements available to date .
our catalog is reliable , with more than 95% of polarization measurements being correctly associated , and more than 93% of total intensity measurements .
the reliability is further improved by the source fitting procedures . nevertheless , while we have selected our catalog as a 1.4 ghz flux - limited polarized sample , selection effects are still likely considerable and are hard to constrain .
the sky coverage of the accumulated data is patchy , and typically taken via shallow observations on the brightest known sources .
as the brightest sources have typically been observed repeatedly , we obtain many more measurements for the brightest sources , and consequently the seds of the fainter sources are less well - constrained . despite these inevitable effects
, the listed polarized seds can reveal the physical form of depolarization , and allow for physical models of the intervening magnetic field structure to be ascertained .
the predominant depolarizing mechanism is important for modelling polarized source counts at low radio frequencies , which will have a substantial impact on the number of polarized sources detectable by facilities such as lofar , mwa , and the ska @xcite . alongside the multiple possible analyses of the catalog we have presented
, additional future work will allow for expansion of these data .
numerous other polarization measurements are available in the literature .
furthermore , our catalog currently uses the nvss rm catalog of @xcite as the reference data ; if a source is not listed as polarized in the @xcite catalog at 1.4 ghz , then polarization measurements at other frequencies are not accumulated into our catalog .
furthermore , polarization measurements in the southern sky ( at declinations @xmath180 ) are currently not included in our catalog .
future work could therefore use a larger reference catalog .
for example , the sydney university molonglo sky survey ( sumss ) @xcite explores a similar parameter space to the nvss in the southern sky ( albeit without polarization information ) .
a combination of the full total intensity nvss measurements @xcite , together with the total intensity measurements from sumss , would therefore greatly expand our reference catalog .
using an nvss+sumss reference catalog for cross - matching with additional polarization measurements accumulated from across the literature , would lead to a much larger sample of polarized seds including those sources that undergo the most extreme depolarization .
more recent polarization surveys , such as the s - band polarization all sky survey ( s - pass ) at 2.3 ghz , will provide @xmath181 polarized sources in the southern sky @xcite corresponding to a significant increase in sample size .
in addition , possible upcoming polarization surveys will also be able to make substantial contributions to such an effort detecting up to 2.2@xmath18210@xmath183 sources in polarized intensity at 2 to 4 ghz ( e.g. * ? ? ? * ) . such cross - matched catalogs ,
particularly when combined with large samples of redshifts , will continue to constitute useful resources for probing magnetic fields and other astrophysical phenomena .
we thank paul hancock for useful discussions on cross - matching , phil kronberg for helpful advice on fitting broadband rotation measures , shane osullivan for insightful conversations on radio source depolarization , and xiaohui sun for helpful comments on other existing polarization measurements .
we are also grateful to the anonymous referee for helpful comments on the original version of the manuscript .
j.s.f . & b.m.g .
acknowledge the support of the australian research council through grants dp0986386 & fl100100114 .
parts of this research were conducted by the australian research council centre of excellence for all - sky astrophysics ( caastro ) , through project number ce110001020 .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .
the australia telescope compact array is part of the australia telescope national facility which is funded by the commonwealth of australia for operation as a national facility managed by csiro .
the westerbork synthesis radio telescope is operated by the astron ( netherlands institute for radio astronomy ) with support from the netherlands foundation for scientific research ( nwo ) .
this research has made use of : the simbad database , operated at centre de donnes astronomiques de strasbourg , france ; the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with nasa ; nasa s astrophysics data system abstract service ; the sdss , for which funding has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , and the u.s .
department of energy office of science ; the 2df/6df qso redshift surveys ( 2qz/6qz ) , which were compiled by the 2qz/6qz teams from observations made with the 2-degree field and 6-degree field on the anglo - australian telescope .
abazajian , k. n. , adelman mccarthy , j. k. , ageros , m. a. , et al . , 2009 , , 182 , 543 .
akritas , m. g. , & bershady , m. a. , 1996 , , 470 , 706 .
arshakian , t. g. , & beck , r. , 2011 , , 418 , 2336 .
battye , r. a. , browne , i. w. a. , peel , m. w. , jackson , n. j. , & dickinson , c. , 2011 , , 413 , 132 .
beck , r. , shukurov , a. , sokoloff , d. , wielebinski , r. , 2003 , , 411 , 99 .
beck , r. , & gaensler , b. m. , 2004 , new astronomy reviews , 48 , 1289 .
beck , r. , 2011 , in magnetic fields in the universe : from laboratory and stars to primordial structures , eds .
m. soida , k. otmianowska - mazur , e. m. de gouveia dal pina & a. lazarian , preprint ( arxiv:1111.5802 ) .
beck , r. , frick , p. , stepanov , r. , sokoloff , d. , 2012 , , 543 , 113 .
beck , a. m. , dolag , k. , lesch , h. , kronberg , p. p. , 2013 ,
mnras , 435 , 3575 .
becker , r. h. , white , r. l. , & edwards , a. l. , 1991 , , 75 , 1 .
bentley , j. l. , 1975 , communications of the association for computing machinery , 18 , 509 .
bernet , m. l. , miniati , f. , lilly , s. j. , kronberg , p. p. , & dessauges - zavadsky , m. , 2008 , , 454 , 302 .
bernet , m. l. , miniati , f. , & lilly , s. j. , 2012 , , 761 , 144 .
bock , d. c. j. , large , m. i. , sadler , e. m. , 1999 , aj , 117 , 1578 .
brentjens , m. a. , de bruyn , a. g. , 2005 , a&a , 441 , 1217 .
budavri , t. , & szalay , a. s. , 2008 , , 679 , 301 .
burn , b. j. , 1966 , , 133 , 67 .
carretti , e. , 2011 , japa , 32 , 457 .
carretti , e. , crocker , r. m. , staveley - smith , l. , et al . , 2013 , , 493 , 66 .
cioffi , d. f. , & jones , t. w. , 1980 , , 85 , 368 .
clauset , a. , rohilla shalizi , c. , & newman , m. e. j. , 2009 , society for industrial and applied mathematics review , 51 , 661 .
condon , j. j. , 1992 , , 30 , 575 .
condon , j. j. , cotton , w. d. , greisen , e. w. , et al .
, 1998 , , 115 , 1693 .
conway , r. g. , haves , p. , kronberg , p.
p. , et al .
, 1974 , , 168 , 137 .
dallal , g. e. , wilkinson , l. , 1986 , journal of the american statistical association , 40 , 294 .
douglas , j. n. , bash , f. n. , bozyan , f. a. , torrence , g. w. , & wolfe , c. , 1996 , , 111 , 1945 .
eichendorf , w. , & reinhardt , m. , 1979 , , 61 , 153 .
fanti , c. , branchesi , m. , cotton , w. d. , et al . , 2004 , , 427 , 465 .
farnes , j. s. , green , d. a. , kantharia , n. g. , 2014 , , 437 , 3236 .
farnsworth , d. , rudnick , l. , & brown , s. , 2011 , , 141 , 191 .
feigelson , e. d. , & babu , g. j. , 1992 , , 397 , 55 .
fisher , r. a. , 1922 , phil .
lond . , 222 , 309 . frick , p. , sokoloff , d. , stepanov , r. , beck , r. , 2011 , , 414 , 2540 .
gaensler , b. m. , beck , r. , & feretti , l. , 2004 , new astronomy reviews .
, 48 , 1003 .
gaensler , b. m. , beck , r. , & feretti , l. , 2005 , baas , 37 , 137.03 .
garn , t. , green , d. a. , riley , j. m. , & alexander , p. , 2009
, , 397 , 1101 .
garrington , s. t. , leahy , j. p. , conway , r. g. , & laing , r. a. , 1988 , , 331 , 147 .
garrington , s. t. , conway , r. g. , 1991 , , 250 , 198 .
giebel , r. , heald , g. , beck , r. , arshakian , t. g. , 2013 , , 559 , 27 .
goldstein , s. j. , jr . , & reed , j. a. , 1984 , , 283 , 540 .
goodlet , j. a. , 2004 , ph.d .
thesis , university of southampton . grant , j. k. , taylor , a. r. , stil , j. m. , et al .
, 2010 , , 714 , 1689 .
gregory , p. c. , scott , w. k. , douglas , k. , & condon , j. j. , 1996 , , 103 , 427 .
hales , c. a. , gaensler , b. m. , norris , r. p. , & middelberg , e. , 2011 , baas , 43 , 142.32 .
hales , c. a. , 2013 , ph.d .
thesis , university of sydney .
hammond , a. m. , robishaw , t. , & gaensler , b. m. , 2012 , preprint ( arxiv:1209.1438v3 ) .
homan , d. c. , ojha , r. , wardle , j. f. c. , et al .
, 2002 , , 568 , 99 .
homan , d. c. , 2012 , , 747 , l24 .
hovatta , t. , lister , m. l. , aller , m. f. , et al . , 2012 , , 144 , 105 .
isobe , t. , feigelson , e. d. , akritas , m. g. , & babu , g. j. , 1990 , , 364 , 104 .
ivezi , . ,
menou , k. , knapp , g. r. , et al . , 2002 , , 124 , 2364 .
jackson , n. , browne , i. w. a. , battye , r. a. , gabuzda , d. , & taylor , a. c. , 2010 , , 401 , 1388 .
jansson , r. , farrar , g. r. , 2012 , , 757 , 14 .
johnston , s. , taylor , r. , bailes , m. , et al .
, 2008 , exp .
astron . , 22 , 151 . kass , r. e. , raftery , a. e. , 1995 , journal of the american statistical association , 90 , 430 .
kelly , b. c. , 2007 , , 665 , 1489 .
kimball , a. e. , & ivezi , . , 2008 , , 136 , 684 .
klamer , i. j. , ekers , r. d. , et al .
, 2006 , , 371 , 852 .
klein , u. , mack , k. h. , gregorini , l. , & vigotti , m. , 2003 , , 406 , 579 .
konigl , a. , 1982 , ` relativistic jets as radio and x - ray sources ' , extragalactic radio sources ; proceedings of the symposium , albuquerque , nm , reidel publishing co. , p. 363 .
edited by dordrecht , d. kronberg , p. p. , bernet , m. l. , miniati , f. , et al . , 2008 , , 676 , 70 .
laing , r. a. , 1988 , , 331 , 149 .
lilliefors , h. w. , 1967 , journal of the american statistical association , 62 , 399 .
liu , r. , pooley , g. , 1991 , , 249 , 343 .
longair , m. s. , 2011 , high energy astrophysics , cambridge , uk : cambridge university press .
mantovani , f. , mack , k. h. , montenegro - montes , f. m. , rossetti , a. , & kraus , a. , 2009 , , 502 , 61 .
mao , s. a. , banfield , j. , gaensler , b. , et al .
, 2014 , preprint ( arxiv:1401.1875 ) .
massardi , m. , burke - spolaor , s. g. , murphy , t. , et al . , 2013 , mnras , 436 , 2915
. mesa , d. , baccigalupi , c. , de zotti , g. , et al .
, 2002 , , 396 , 463 .
murphy , t. , sadler , e. m. , ekers , r. d. , et al .
, 2010 , , 402 , 2403 .
noutsos , a. , johnston , s. , kramer , m. , & karastergiou , a. , 2008 , , 386 , 1881 .
osullivan s. , stil j. , taylor a. r. , ricci r. , grant j. k. , shorten k. , 2008 , in pos , proc .
vlbi network symp . for radio astron . , p. 107 .
osullivan , s. p. , brown , s. , robishaw , t. , et al . , 2012 , , 421 , 3300 .
oppermann , n. , et al .
, 2012 , a&a , 542 , 93 .
pacholczyk , a. g. , & swihart , t. l. , 1967 , apj , 150 , 647 .
perley , r. , napier , p. , jackson , j. , et al .
, 2009 , ieee proceedings , 97 , 1448 .
planck collaboration , ade , p. a. r. , aghanim , n. , et al .
, 2013 , preprint ( arxiv:1303.5076 ) .
raftery , a. e. , 1995 , sociol .
methodol . , 25 , 111 .
reich , w. , frst , e. , reich , p. , et al . , 2000 , , 363 , 141 .
rengelink , r. b. , tang , y. , de bruyn , a. g. , et al .
, 1997 , , 124 , 259 .
ricci , r. , et al .
, 2013 , mnras , 435 , 2793 .
rossetti , a. , dallacasa , d. , fanti , c. , fanti , r. , & mack , k. h. , 2008 , , 487 , 865 .
roy , s. , rao , a. p. , & subrahmanyan , r. , 2005 , , 360 , 1305 .
rudnick , l. , & owen , f. , 2014 , preprint ( arxiv:1402.3637 ) .
sajina , a. , partridge , b. , evans , t. , et al . , 2011 , , 732 , 45 .
sazonov , v. n. , 1970 , radiophysics and quantum electronics , 13 , 162 .
schwarz , g. , 1978 , ann .
, 6 , 461 .
simard - normandin , m. , kronberg , p. p. , & neidhoefer , j. , 1980 , , 40 , 319 .
simard - normandin , m. , kronberg , p. p. , & button , s. , 1981 , , 45 , 97 .
simard - normandin , m. , kronberg , p. p. , & neidhoefer , j. , 1981 , , 43 , 19 .
simard - normandin , m. , kronberg , p. p. , & button , s. , 1982 , , 48 , 137 .
simmons , j. f. l. , stewart , b. g. , 1985 , , 142 , 100 .
sokoloff , d. d. , bykov , a. a. , shukurov , a. , et al . , 1998 , , 299 , 189 .
spiegelhalter , d. j. , best , n. g. , carlin , b. p. , van der linde , a. , 2002 , journal of the royal statistical society : series b ( statistical methodology ) , 64 , 583 .
sprent , p. , 1969
, ` models in regression and related topics ' , methuen , london .
stil , j. , taylor , a. r. , krause , m. , beck , r. , 2007 , ` polarization of mjy radio sources ' , from planets to dark energy : the modern radio universe , published online at sissa , proceedings of science , p. 69 . edited by beswick r. et al .
strom , r. g. , & jaegers , w. j. , 1988 , , 194 , 79 .
tabara , h. , & inoue , m. , 1980 , , 39 , 379 .
taylor , a. r. , stil , j. m. , & sunstrum , c. , 2009 , , 702 , 1230 .
taylor , a. r. , & salter , c. j. , ` galfacts : the g - alfa continuum transit survey ' , the dynamic interstellar medium : a celebration of the canadian galactic plane survey .
proceedings of the conference , held at the naramata centre , naramata , british columbia , canada , june 610 , 2010 . edited by r. kothes , t. l. landecker , and a. g. willis .
san francisco : astronomical society of the pacific , 2010 , p.402 .
tingay , s. j. , jauncey , d. l. , king , e. a. , et al . , 2003 , , 55 , 351 .
tingay , s. j. , goeke , r. , bowman , j. d. , et al . , 2013 ,
pasa , 30 , 7 .
tribble , p. c. , 1991 , , 250 , 726 .
van haarlem , m. p. , et al .
, 2013 , a&a , 556 , 2 .
warton , d. i. , wright , i. j. , falster , d. s. , westoby , m. , 2006 , biological reviews , 81 , 259 .
whittam , i. h. , riley , j. m. , green , d. a. , et al . , 2013 , , 429 , 2080 .
wilson w. e. , ferris r. h. , axtens p. , et al . , 2011 , mnras , 416 , 832 .
zukowski , e. l. h. , kronberg , p. p. , forkert , t. , & wielebinski , r. , 1999 , , 135 , 571 .
the polarization properties of sources at radio wavelengths are dominated by various forms of astrophysical depolarization that can be described by a number of models . for the purpose of model selection ,
we need to pick a number of models a - priori that may describe the physics of each source . this depolarization can be a reduction in the polarized fraction that occurs either internally to the radio source or in its foreground i.e. externally somewhere along the line of sight between us and the source environment .
when the depolarization occurs externally , we consider this to be the result of a _ foreground depolarizing screen_. the effects of external screens must be removed if we are to understand the internal structure of radio sources , where thermal plasma is mixed with the magnetoionic medium responsible for the radio emission .
this separation of internal and foreground effects is complicated as both have a similar wavelength dependence ( e.g. * ? ? ?
consequently , the relationship between polarized fraction , @xmath26 , and wavelength , @xmath11 , can be an insensitive test of the rm structure , unless information in stokes @xmath46/@xmath47 is considered simultaneously ( e.g. * ? ? ?
nevertheless , statistical measures of sed properties can still be used to ascertain the predominant form of depolarization .
we now explain the various forms of depolarization in detail .
unless otherwise stated , we focus on depolarization in optically thin emission ( e.g. * ? ? ?
optically thick emission could result in more complicated behavior in a polarized sed including oscillations and even constant polarized fractions at increasing wavelengths ( e.g. * ? ? ?
. in the case of optically thin synchrotron radiation , random fluctuations from a disordered magnetic field within a source can lead to a reduced fractional polarization given by @xmath184 where @xmath185 and @xmath186 are the uniform and random field strengths respectively ( e.g. * ? ? ?
* ) , and the total intensity spectral index , @xmath0 , is defined such that @xmath21 .
extensions to equation [ internal ] have been proposed .
as the sources in our catalog are always unresolved , this effect is independent of wavelength we consider it no further in this paper .
internal faraday depolarization ( also called ` depth- ' or ` front back- ' depolarization ) takes places when the emitting and faraday rotating regions within a source are intermixed .
this internal depolarization can be separated into two types : ( i ) differential faraday rotation , and ( ii ) internal faraday dispersion . for ( i ) , in the presence of a regular magnetic field , the plane of polarization of emission from the far side of the region undergoes a different amount of faraday rotation compared to emission from the near side .
the process is known as ` differential faraday rotation ' and the sum through the entire region ( along the line of sight ) results in depolarization ( e.g. * ? ? ?
* ; * ? ? ?
while this process leads to depolarization , it is characterized by ` bobbing ' of the polarized fraction as a function of increasing wavelength .
this bobbing is difficult to identify in seds with sparse frequency sampling , and could likely only be measured using a wide and continuous observing bandwidth . for (
ii ) , in the presence of a turbulent magnetic field , the plane of polarization experiences random fluctuations as it propagates through the region .
this process is known as ` internal faraday dispersion ' and this addition of random anisotropic magnetic fields results in further depolarization ( e.g. * ? ? ?
* ; * ? ? ?
however , observational evidence suggests that internal faraday effects are likely not the dominant cause of depolarization in extragalactic radio sources we therefore consider it no further in this paper .
external faraday depolarization occurs if the faraday rotation in front of the emitting region varies significantly across the source over a solid angle smaller than the beam ( e.g. * ? ? ?
* ; * ? ? ?
such a faraday screen consists of a magnetoionic region that is devoid of relativistic particles and that exists somewhere along the line of sight between the observer and the source ( e.g. * ? ? ?
if this faraday screen contains a constant regular field then the region causes faraday rotation of the polarized emission from background sources , but does not cause any depolarization .
nevertheless , any deviation from a constant field within the synthesized beam will create a rm gradient and subsequently cause depolarization .
such anisotropy of the @xmath187 field is caused by either turbulent , or systematically varying regular fields .
for turbulent magnetic fields or _
external faraday dispersion _ , depolarization occurs when there are many turbulent regions within the synthesized beam . for regular magnetic fields or _
beam depolarization _
, depolarization will occur if there are any variations in the strength or orientation of the field within the synthesized beam ( e.g. * ? ? ?
the two situations are analogous and for the case of turbulent fields with the values of rm having a large spread over the beam , and for the case of a systematically varying regular field with a gradient in the external rm across the beam , the polarization is rotated through different angles in each direction at longer @xmath11 , resulting in a reduction in @xmath26 ( see e.g. * ? ? ? * for further detail ) .
the effect of external faraday depolarization was initially described by @xcite with an equation of the form @xmath188 where @xmath112 is the rm dispersion of the faraday screen within a single beam .
a smoothly interpolating simple approximation to a burn law can be provided using a gaussian model , see e.g. equation [ gaussian ] in section [ fittingpolarizedsed ] .
nevertheless , observations show that significant polarized emission continues to be detectable at relatively large wavelengths ( e.g. * ? ? ?
* ) , showing that the relationship between polarized fraction and wavelength can not be as strong as a function of @xmath120 for all sources . evidence for deviations from a burn law have also been found from diffuse polarized emission .
the equations of @xcite were extended by @xcite to show that at long wavelengths ( @xmath189 , where @xmath190 is the wavelength at which the polarized fraction is half its maximum value ) the polarized fraction can decrease via a power law such that @xmath191 \lambda^{-2 } \
, , \label{tribbling}\ ] ] where @xmath192 is a measure of the resolution , with @xmath193 being related to the beam fwhm@xmath194 and @xmath195 being the size of the rm fluctuations .
importantly , @xcite shows that depending on the precise form of the structure function of the rm fluctuations , the polarization can actually fall off such that @xmath196 ( and not only as @xmath197 ) , where @xmath198 has been shown observationally to approximately vary between @xmath174 and @xmath199 @xcite .
furthermore , the tribble models are all calculated in the source rest - frame , such that the regime @xmath189 is not well - defined without extensive prior measurements of a radio source .
a more general solution is therefore given by @xmath200 where @xmath201 is some constant and @xmath24 is a polarization spectral index .
note that @xmath24 is defined in the opposite sense to the total intensity spectral index , @xmath0 , in that it is the exponent of observing frequency rather than wavelength .
such power laws have previously been used to define polarization spectral energy distributions .
a smoothly interpolating fit for a tribble law can be provided using a power law model , see e.g. equation [ powerlaw ] in section [ fittingpolarizedsed ] .
more recently , it has also been shown that the polarization can decrease as a function of @xmath120 at smaller wavelengths as proposed by @xcite but then remains unexpectedly constant out to longer wavelengths .
there have been attempts to explain these seds as a result of ` partial coverage ' , with the explanation that only a fraction of the source is covered by an inhomogeneous faraday screen . in an effort to derive @xmath112 for these sources , make an empirical modification to equation [ burnbabyburn ] so that @xmath202 \ , , \label{partialcoverage}\ ] ] where @xmath203 is interpreted as the covered ( depolarizing ) fraction of the source , with the uncovered fraction @xmath204 retaining a constant @xmath26 out to arbitrarily long wavelengths .
this model was found to be more successful in reproducing the form of some polarized seds . a smoothly interpolating simple approximation for a rossetti
mantovani law can be provided using a gaussian with a constant term , see e.g. equation [ gauss+const ] in section [ fittingpolarizedsed ] . as a point of further interest
, @xcite also considered a partial coverage model . under the assumption that depolarization could be originating from discrete clouds in the galaxy
, there would be @xmath106 average clouds along the line of sight .
following , if @xmath205 the depolarization is similar to equation [ burnbabyburn ] , while several lines of sight will not intersect any cloud if @xmath206 .
consequently , wavelength - independent polarization may emerge through gaps in this ` faraday web ' .
theoretically @xmath207 } ) ] } \ , , \label{faradayweb}\ ] ] where @xmath208 is the rm of a single cloud .
such clouds may reasonably exist in the environment surrounding a radio source .
such a functional form also follows a similar decline to that of equation [ partialcoverage ] , and high - quality data would be needed to distinguish between these similar models .
differences in the total intensity spectral index of components in an unresolved source can result in different regions of a source being probed at different observing frequencies .
the effect of a typical radio galaxy which has a flat - spectrum , weakly polarized core combined with steep - spectrum , more - highly polarized lobes / jets , can be observed in a significant number of sources ( e.g. * ? ? ?
this ` spectral depolarization ' tends to give rise to a prominent peaked structure in a plot of @xmath26 versus @xmath11 . in some cases
, this may also give rise to behavior analogous to beating between two oscillatory components @xcite .
a smoothly interpolating simple approximation for a spectral depolarizer can be provided using a gaussian model , see e.g. equation [ gaussian ] in section [ fittingpolarizedsed ] .
some sources also exhibit ` inverse depolarization ' or ` repolarization ' , with the polarized fraction increasing at longer wavelengths .
this is again likely the consequence of a multiple unresolved components within a compact source , and the probing of different emitting regions at different frequencies .
such repolarization can also be explained as increased ordering of the magnetic field in the component of the source nearest to the observer along the line of sight . at lower frequencies ,
the emitting region further into the source depolarizes effectively constituting a measurement of different source components at different observational frequencies .
increasing @xmath26 could also arise in the unresolved , optically - thick core region where one is observing different parts of the jet at different wavelengths ( e.g. * ? ? ?
a smoothly interpolating simple approximation for a repolarizer can be provided using a power law model , see e.g. equation [ powerlaw ] in section [ fittingpolarizedsed ] .
other explanations have also been proposed , and repolarization could occur in jets that are well separated from the core in the optically thin part of the jet , and are also reasonably isolated from other strong jet features @xcite .
other suggested physical models propose that internal faraday rotation acts to align the polarization from the far and near sides of a jet , leading to increased @xmath26 at longer wavelengths .
helical and randomly tangled magnetic fields have also been suggested as being responsible ( e.g. * ? ? ?
* ; * ? ? ?
data for all sources and the associated fitted and calculated properties are provided as machine - readable tables .
there are two versions of the catalog available , the ` full - catalog ' and the ` sed - catalog ' respectively .
the column headings are identical in both catalog versions , and the sed - catalog is just a subset of rows from our primary data product , the full - catalog .
the column headings of both catalogs are described in table [ tab : catkey ] .
selected columns for the first 45 sources in the sed - catalog are presented in tables [ tab : thedata ] to [ tab : thedata5 ] in section [ catalog ] .
ccclc 1 & i5 & source_num & & + 2 & a13 & source_name & & + 3 & a12 & ra_str & right ascension , from the nvss ( j2000 ) . &
+ 4 & a12 & dec_str & declination , from the nvss ( j2000 ) . & + 5 & f9.5 & ra & right ascension , from the nvss ( j2000 ) . & @xmath209 + 6 & f8.5 & dec & declination , from the nvss ( j2000 ) . & @xmath209 + 7 & f8.4 & g_lon & galactic longitude , from the nvss . & @xmath209 + 8 & f8.4 & g_lat & galactic latitude , from the nvss . & @xmath209 + 9 & i2 & num_data & number of polarized data . & + 10 & i2 & num_data_i & number of total intensity data . &
+ 1179 & f6.3 & lambda & & cm + 80148 & f5.2 & pol_frac & & % + 149217 & f5.2 & pol_frac_err & & % + 218286 & f6.2 & pol_angle & & @xmath209 + 287355 & f6.2 & pol_angle_err & & @xmath209 + 356376 & f6.3 & i_lambda & & cm + 377397 & f7.2 & i_flux & & mjy + 398418 & f7.2 & i_flux_err & & mjy + 419 & a5 & depol_type & & + 420 & a9 & depol_physics & & + 421 & f7.3 & c1 & & see table [ tab : modelsunits ] + 422 & f7.3 & c2 & & see table [ tab : modelsunits ] + 423 & f7.3 & c3 & & see table [ tab : modelsunits ] + 424 & f7.3 & c4 & & see table [ tab : modelsunits ] + 425 & f7.3 & c1_err & & see table [ tab : modelsunits ] + 426 & f7.3 & c2_err & & see table [ tab : modelsunits ] + 427 & f7.3 & c3_err & & see table [ tab : modelsunits ] + 428 & f7.3 & c4_err & & see table [ tab : modelsunits ] + 429 & f7.3 & depol_ks & & + 430 & f7.3 & depol_ks_pval & & + 431 & f7.3 & depol_chi2 & & + 432 & f7.3 & depol_chi2_pval & & + 433 & i1 & data_flag & & + 434 & f7.3 & pol_beta_c1 & & + 435 & f7.3 & pol_beta & & + 436 & f7.3 & pol_beta_err & & + 437 & f7.3 & pol_beta_chi2 & & + 438 & f7.3 & pol_beta_chi2_pval & & + 439 & f7.3 & pol_beta_ks & & + 440 & f7.3 & pol_beta_ks_pval & & + 441 & f7.3 & z & & + 442 & f7.3 & z_err & & + 443 & a8 & selected_obj & & + 444 & a1 & i_class & & + 445 & f7.3 & i_alpha_d1 & & see table [ tab : modelsunits ] + 446 & f7.3 & i_alpha & & see table [ tab : modelsunits ] + 447 & f7.3 & i_alpha_d3 & & see table [ tab : modelsunits ] + 448 & f7.3 & i_alpha_d1_err & & see table [ tab : modelsunits ] + 449 & f7.3 & i_alpha_err & & see table [ tab : modelsunits ] + 450 & f7.3 & i_alpha_d3_err & & see table [ tab : modelsunits ] + 451 & f7.3 & i_chi2 & & + 452 & f7.3 & i_chi2_curved & & + 453 & f7.3 & i_pval & & + 454 & f7.3 & i_pval_curved & & + 455 & f7.2 & rm_broad & & rad m@xmath144 + 456 & f7.2 & rm_broad_err & & rad m@xmath144 + 457 & f7.2 & evpa_broad & & @xmath209 + 458 & f7.2 & evpa_broad_err & & @xmath209 + 459 & f7.3 & rm_broad_chi2 & & + 460 & f7.3 & rm_broad_pval & & + 461 & f6.1 & nvss_rm & & rad m@xmath144 + 462 & f6.1 & nvss_rm_err & & rad m@xmath144 + 463 & f5.1 & i & & mjy + 464 & f5.1 & i_err & & mjy + 465 & f5.1 & p & & mjy + 466 & f5.1 & p_err & & mjy + 467 & f4.1 & pi & & % + 468 & f4.1 & pi_err & & % + 469 & a1 & major_axis_limit & & + 470 & f4.1 & major_axis & & arcsec + 471 & f4.1 & major_axis_err & & arcsec + 472 & a1 & minor_axis_limit & & + 473 & f4.1 & minor_axis & & arcsec + 474 & f4.1 & minor_axis_err & & arcsec + 475 & f5.1 & nvss_pa & & @xmath209 + 476 & f5.1 & nvss_pa_err & & @xmath209 + 477 & f5.2 & l & & @xmath210
/w hz@xmath211 + 478 & f5.2 & l_err & & @xmath212 /w
[ tab : catkey ] cccc 1 & nvss / nvss rm & 1400 & 21.41 + 2 & at20 g & 4860 & 6.169 + 3 & & 8640 & 3.470 + 4 & & 20000 & 1.499 + 5 & t03 & 1400 & 21.41 + 6 & & 2496 & 12.01 + 7 & & 4800 & 6.246 + 8 & & 8640 & 3.470 + 9 & z99 & 4700 & 6.379 + 10 & b3-vla & 2695 & 11.12 + 11 & & 4850 & 6.181 + 12 & & 10500 & 2.855 + 13 & sn80/sn81/sn82 & 158014750 & 18.97 2.032 + @xmath213 & @xmath213 & @xmath213 + 17 & & 158014750 & 18.972.032 + 18 & ti80 & 40499930 & 74.210.300 + @xmath213 & @xmath213 & @xmath213 + 69 & & 40499930 & 74.210.300 [ tab : polelements ] cccc 1 & nvss / nvss rm & 1400 & 21.414 + 2 & at20 g & 4860 & 6.169 + 3 & & 8640 & 3.470 + 4 & & 20000 & 1.499 + 5 & t03 & 1400 & 21.414 + 6 & & 2496 & 12.011 + 7 & & 4800 & 6.246 + 8 & & 8640 & 3.470 + 9 & z99 & 4700 & 6.379 + 10 & b3-vla & 2695 & 11.124 + 11 & & 4850 & 6.181 + 12 & & 10500 & 2.855 + 13 & sn80/sn81/sn82 & 158014750 & 18.9742.032 + @xmath213 & @xmath213 & @xmath213 + 17 & & 158014750 & 18.9742.032 + 18 & wenss & 326 & 91.961 + 19 & texas & 365 & 82.135 + 20 & gb6 & 4850 & 6.181 + 21 & north6 cm & 4850 & 6.181 [ tab : ielements ] ccccccc gauss & [ gaussian ] & & @xmath214 & & & burn + gauss & [ gaussian ] & & @xmath215 & & & peaked + power & [ powerlaw ] & & @xmath216 & & & increase + power & [ powerlaw ] & & @xmath217 & & & tribble + gau+t & [ gauss+const ] & & & & & mantovani [ tab : models ] cccccccc [ gaussian ] & per cent ( % ) & centimetres ( cm ) & centimetres ( cm ) & n / a & & & + [ powerlaw ] & @xmath218 $ ] & unitless & n / a & n / a & & & + [ gauss+const ] & per cent ( % ) & centimetres ( cm ) & centimetres ( cm ) & per cent ( % ) & & & + [ powerlawi ] & & & & & millijansky ( mjy ) & unitless & n / a + [ powerlawicurved ] & & & & & millijansky ( mjy ) & unitless & unitless [ tab : modelsunits ] | an understanding of cosmic magnetism requires converting the polarization properties of extragalactic radio sources into the rest - frame in which the corresponding polarized emission or faraday rotation is produced .
motivated by this requirement , we present a catalog of multiwavelength linear polarization and total intensity radio data for polarized sources from the nrao vla sky survey ( nvss ) .
we cross - match these sources with a number of complementary measurements combining data from major radio polarization and total intensity surveys such as at20 g , b3-vla , gb6 , north6 cm , texas , and wenss , together with other polarization data published over the last 50 years . for 951 sources , we present spectral energy distributions ( seds ) in both fractional polarization and total intensity , each containing between 3 and 56 independent measurements from 400 mhz to 100 ghz .
we physically model these seds , and where available provide the redshift of the optical counterpart . for a superset of 25,649 sources we provide the total intensity spectral index , @xmath0 .
objects with steep versus flat @xmath0 generally have different polarization seds : steep - spectrum sources exhibit depolarization , while flat - spectrum sources maintain constant polarized fractions over large ranges in wavelength .
this suggests the run of polarized fraction with wavelength is predominantly affected by the local source environment , rather than by unrelated foreground magnetoionic material .
in addition , a significant fraction ( 21% ) of sources exhibit ` repolarization ' , which further suggests that polarized seds are affected by different emitting regions within the source , rather than by a particular depolarization law .
this has implications for the physical interpretation of future broadband polarimetric surveys . |
i am grateful to s. mathur , k. olum and e. weinberg for the interesting discussions .
e. e. flanagan , d. marolf and r. m. wald , phys .
d * 62 * , 084035 ( 2000 ) [ hep - th/9908070 ] .
r. bousso , e. e. flanagan and d. marolf , phys .
d * 68 * , 064001 ( 2003 ) [ hep - th/0305149 ] .
a. masoumi and s. d. mathur , phys .
d * 90 * , no .
8 , 084052 ( 2014 ) [ arxiv:1406.5798 [ hep - th ] ] .
a. masoumi and s. d. mathur , arxiv:1412.2618 [ hep - th ] . | entropy of all systems that we understand well is proportional to their volumes except for black holes given by their horizon area .
this makes the microstates of any quantum theory of gravity drastically different from the ordinary matter .
because of the assumption that black holes are the maximum entropy states there have been many conjectures that put the area , defined one way or another , as a bound on the entropy in a given region of spacetime .
here we construct a simple model with entropy proportional to volume which exceeds the entropy of a single black hole .
we show that a homogeneous cosmology filled with this gas exceeds one of the tightest entropy bounds , the covariant entropy bound and discuss the implications .
bekenstein s discovery @xcite of black hole entropy and hawking s calculation @xcite of the temperature are among the most important clues we have on a quantum theory of gravity .
these show a significant difference between the microstates of quantum gravity and ordinary matter .
quantum field theory , the way we understand it , implies that the density of states in momentum space is proportional to the volume and so is the entropy . the assumption , or the bias , that black holes are the highest entropy states brought about many conjectures on the maximum entropy confined in a region .
in the absence of gravity , one can put arbitrary entropy in a region by inserting enough energy in it .
gravity changes this picture .
energy distorts the geometry and volume in such a way that the entropy is no longer arbitrarily large .
bekenstein @xcite conjectured that in a system where gravity is not the dominant interaction , the entropy in a region is bounded by the entropy of a black hole of the same size .
this conjecture fails in well known examples .
consider a cosmology with flat spatial slices and ordinary matter , like radiation .
the entropy is proportional to the volume which grows much faster than the area and exceeds it for large enough regions .
later , several more refined versions of this conjecture were stated @xcite that survive in more general circumstances .
all of these conjectures assume some constraint on the energy - momentum tensor like the weak or dominant energy conditions .
it is not possible to put any bound on entropy if one allows an arbitrary energy - momentum tensor .
for example , consider a combination of positive and negative energy states with a zero net energy - momentum .
these states would not have any feedback on geometry and hence , can violate any entropy bound .
therefore , in a sense , the entropy bounds are limitations on the form of energy - momentum tensor rather than restricting the entropy . putting enough restrictions on the energy - momentum tensor , one can prove some of the bounds in weak gravity limit @xcite .
however , we should be cautious about putting too much constraint on the states of a theory ( quantum gravity ) that we do not understand well . here
we motivate a heuristic model which satisfies the dominant energy condition and breaks one of the tightest entropy bounds , the covariant ( bousso ) bound @xcite .
the details and the subtleties of this model is discussed in @xcite .
let s investigate the maximum entropy one can put in a cube of size @xmath0 . as depicted in fig.[fig : entorpy ] , at small energy densities we expect a gas of relativistic particles to have the highest entropy .
increasing the energy in the box makes it collapse into a black hole which indeed has a much higher entropy .
can we put more entropy and energy in this box ?
the answer is yes and a simple configuration is motivated in fig.[fig : singlevslattice ] .
each of the 27 smaller holes on the right panel have 1/9 of the area and entropy of the hole on the left .
hence , the net entropy of the right panel is 3 times of the single hole configuration .
using @xmath1 holes of size @xmath2 , we get @xmath3 times more entropy than a single black hole .
if this box is put in empty space , it is entropically favorable for the smaller ones to merge into a single black hole .
the result is a black hole of size @xmath4 and @xmath5 times the entropy of the original lattice .
however , if the whole universe is filled with this black hole lattice , there is no room for them to merge .
it is entropically favorable to have this lattice until the universe expands enough and they can merge then .
we may worry that this system is completely unstable to gravitational collapse .
however , to surpass the single black hole entropy we do not necessarily create strong gravitational fields .
one can build a gas of small black holes where each is well outside the schwarzschild radius of the others .
suppose the separation between the black holes is @xmath6 times larger than their schwarzschild radius . to exceed the entropy of the single black hole configuration , we need small black holes of size @xmath7 .
consider a solar system size box in which the black holes are separated by 10000 times of their horizon radius .
we will need small black holes of size 5 meters to surpass the single black hole entropy .
the big separation of the black holes prevents an immediate coalescence . contains a single black hole .
right , the same box , filled with 27 smaller black holes of size @xmath8 .
the combined horizon area of smaller black holes is 3 times larger than the left panel.,title="fig:",width=211 ] contains a single black hole .
right , the same box , filled with 27 smaller black holes of size @xmath8 .
the combined horizon area of smaller black holes is 3 times larger than the left panel.,title="fig:",width=211 ] this example shows that even using the simple ingredients of classical gravity we can get entropies larger than the one for a single black hole . in more realistic scenarios obtained form string theory
we replace this lattice of black holes with a set of intersecting branes .
these branes interact with each other and have pressure and thermodynamics .
let s consider a gas of these black holes with typical sizes @xmath9 and separation of the same order @xmath9 .
each black hole has entropy @xmath10 and energy @xmath11 and there are @xmath12 of them in a volume @xmath13 .
the total energy and entropy is given by @xmath14 eliminating @xmath9 between these two equations we get s~1r_s vg~ ( veg)^-vg ~ introducing a constant @xmath15 of order unity and rewriting in terms of energy and entropy densities @xmath16 and @xmath17 we get [ entropyformul ] s = k .
this gives an extensive equation for entropy .
using the first law of thermodynamics @xmath18 we get @xmath19 this corresponds to the equation of state @xmath20 which satisfies the dominant energy condition .
although we derived eq . using a heuristic model , this equation has been derived from many different approaches including : 1 .
veneziano @xcite obtained it assuming that the entropy is given by the area of the apparent horizon with radius @xmath21 .
2 .
banks and fishcler @xcite derived @xmath20 using coalescing horizon size black holes .
3 .
horowitzand and polchinski @xcite derived it for a gas of string states near the point where the strings are large enough to collapse into a black hole .
4 .
sasakura @xcite conjectured it by demanding a ` spacetime uncertainty relation ' .
5 .
verlinde @xcite argued that a similar relation correspond to the cardy formula for the density of states .
6 . brustein and veneziano @xcite obtained it by using the notion of a causal connection .
7 .
masoumi and mathur @xcite showed that eq .
is the only entropy formula which is manifestly invariant under s and t - dualities and approaches the black hole entropy at small densities .
now we show that this high - entropy matter can violate one of the most beautiful entropy bounds , the covariant ( bousso ) bound .
this bound is based on the entropy in the light sheet and has a very different philosophy and formulation .
the question of interest in the earlier bounds was : how much is the maximum entropy one can put in a given region ?
the answers were typically the area of that region in planck units .
the covariant entropy bound changes the question : given a two - dimensional surface @xmath22 of area @xmath23 , what is the region which has a maximum entropy given by @xmath23 ?
the conjecture was expressed in terms of non - expanding null geodesic congruences .
take a bundle of null rays perpendicular to @xmath22 and follow them until they reach a caustic or start diverging .
the region covered by this bundle is called the light - sheet .
the conjecture expresses that the entropy of this light - sheet is bounded by @xmath23 .
we show that a homogeneous cosmology with entropy density given in eq .
exceeds this bound .
assume a bianchi type i metric @xmath24 this is a homogenous anisotropic universe .
the friedmann equations for this universe filled with the matter discussed in eq .
have a power law solution in the form @xmath25 where @xmath26 .
let s choose a surface @xmath27 in the @xmath28 plane .
if @xmath29 is the maximum entropy allowed by the covariant bound , we can show @xmath30 this ratio can get bigger than one for _ any _ value of @xmath15 if @xmath31 is close enough to one .
for example if @xmath32 we need @xmath33 which is a mild anisotropy .
the ratio remains the same even if we do not follow the light sheet all the way to the initial singularity .
this is in contrast with the conditions for ordinary matter like radiation . to break the bound for radiation
we must follow the light sheet to planckian densities where we have no good understanding .
here we constructed a model which violates the tightest of entropy bounds .
there is no simple reason to assume this model is not allowed on first principles . if a shear viscosity prevents large anisotropy we may be able to find a deeper connection between the viscosity and entropy bounds .
if this violation stands as it is , it has important consequences . the simplest would be tightening the assumptions of the area - based entropy bounds
this would render these conjectures less useful . as mentioned above ,
the entropy bounds are really constraints on the type of matter allowed by the theory .
the bounds would not be very impressive under too much limitation .
a deeper consequence would be returning to the more natural bounds based on the volume instead of area .
this will bring the number counting for the states allowed by gravity to the same footing as other quantum theories where we have a much better understanding and control . |
in physics , the standard way to achieve _ spontaneous symmetry breaking _ ( ssb ) is by introducing a scalar fieldu(1 ) gauge symmetry spontaneously . ] with a potential like @xmath3 both the negative ( mass)@xmath4 @xmath5 term and @xmath6 term play a crucial role to make the potential behaving like figure [ 2d higgs potential plot ] or figure [ 3d higgs potential plot ] and possessing multiple true vacua . in the process of choosing one among these true vacua , ssb is achieved and , in the meanwhile , some topological defects like domain walls or cosmic strings is a single real scalar field , domain walls can form along with ssb . on the other hand , if @xmath7 denotes a complex scalar field or n @xmath8 real scalar fields , ( cosmic ) strings can form . ] could form .
the _ non - minimal coupling _ between scalar fields and gravity has been introduced in various topics in quantum field theories and cosmology . in quantum field theories ( for
a comprehensive review and a list of references , see the book @xcite ) , the conformal invariance of the massless scalar field with the non - minimal coupling constant @xmath9 was first noted by penrose @xcite ( see also @xcite ) . in the framework of quantum field theories in curved space - time
, the non - minimal coupling term can also be introduced by the requirement of renormalizability @xcite ( for a review , see @xcite ) . in cosmology
( see @xcite @xcite and references therein ) , the usage of the non - minimal coupling in the inflation models @xcite and models for the cosmological constant problem @xcite @xcite ( first proposed by dolgov @xcite ) has been widely explored .
in addition , the non - minimal coupling is also involved in an interesting model called `` induced gravity '' which was first proposed by zee in 1979 @xcite and is still explored nowadays ( e.g. a series of work about induced gravity and inflation by kao @xcite ) . the role of the _ non - minimal coupling _ in ssb and phase transitions in 4d space - time has been widely discussed ( for a review , see @xcite ) .
it has been pointed out that the non - minimal coupling with the ` external ' gravitational field may lead to ssb .
it has also been noted that ssb and phase transitions can be induced by curvature via the non - minimal coupling with the ` external ' gravitational field @xcite .
accompanying the quest of symmetry breaking and vacuum stability in curved space - time , that the curvature can be a symmetry - breaking factor has been shown .
note that , in the papers mentioned above , the effect of the curved space - time is taken into account by introducing an ` external ' gravitational field , which means that the metric tensor of the space - time is treated as a background , and the _ ricci scalar _
@xmath10 in the non - minimal coupling term is regarded as an ` external ' parameter ( like the temperature of a macroscopic system in thermodynamics ) .
this is different from our usage of the gravitational field to be discussed below . in this paper , we propose an alternative way to achieve ssb in 4d and arbitrary higher ( @xmath2 ) dimensional space - time .
we will show that , by introducing the _ non - minimal coupling _ between the scalar field and gravity and keeping the negative ( mass)@xmath4 @xmath5 term , we can build up a new scenario to implement ssb without using @xmath11 term .
avoiding the usage of @xmath11 term could be a benefit when we consider ssb in the higher dimensional space - time , because @xmath12 term is non - renormalizable in the formalism of quantum field theories in @xmath2 dimensional space - time .
the gravitational field introduced in this paper is not an ` external ' field or a background field , but strongly depends on the scalar field through the einstein equations .
such a strong dependence plays a crucial role in implementing ssb in our scenario .
this is one of the essential differences between the previous work mentioned in last paragraph and our work in this paper .
to illustrate our idea specifically , we consider a simple case in the third part of this paper : a world with an arbitrary space - time dimension which is dominated by the ` cosmological constant ' , the ` vacuum energy ' , and the ` potential energy ' in the classical framework also contributes a energy - momentum tensor which has the same cosmological - constant form when @xmath7 is constant in space - time.[origin of effective lambda ] ] .
( in fact , the space - time dimension @xmath13 should be larger than two in our discussion . ) in this simple case , we will see that the potential of the scalar field entails multiple true vacua , which is the key element to implement ssb .
to introduce the non - minimal coupling between the scalar field and gravity , we consider the lagrangian density of the scalar field @xmath7 in @xmath13 dimensional space - time , @xmath14 where @xmath15 is the potential term for the scalar field @xmath7 which includes the non - minimal coupling to gravity via the _ ricci scalar _ @xmath10 with a coupling constant @xmath16 which is _ _ positive _ _ ] which is different from that in some of the papers mentioned in ` * introduction * ' so that the sign of the non - minimal coupling constant @xmath16 in this paper is opposite to that in them . ] in our consideration , and @xmath17 is the metric tensor of the @xmath13 dimensional space - time . to discuss the behavior of the potential @xmath18(@xmath19 ) , we need to know the behavior of the _ ricci scalar _
@xmath10 , which can be obtained from the @xmath13 dimensional _ einstein equations _ : @xmath20 where @xmath21 is the @xmath13 dimensional einstein tensor , @xmath22 is the cosmological constant , @xmath23 is the energy - momentum tensor , @xmath24 is the effective energy - momentum tensor can be absorbed into @xmath23 ( as shown in eq.([effective energy momentum ] ) ) to form an effective energy - momentum tensor @xmath24 in which the contribution from @xmath25 is included.[effective lambda ] ] , and @xmath26 is a parameter ( to be called `` @xmath13 dimensional gravitational constant '' ) which is positive and has the same role with the newton s constant @xmath27 in four dimensional space - time . for simplicity , we consider the case for the perfect fluid with the effective energy - momentum tensor@xmath28 by taking trace of the einstein equations , it is straightforward to get a relation between the _ ricci scalar _ @xmath10 and @xmath29 : @xmath30 consider a simple case that the effective energy - momentum tensor is dominated by the effective cosmological constant @xmath31 which includes the original cosmological constant @xmath32 and the possible contribution from the vacuum energy or the potential energy ( see footnote [ origin of effective lambda ] ) : @xmath33 and hence @xmath34 consequently , in this simple case , @xmath35 which is positive(negative ) for a negative(positive ) effective cosmological constant @xmath36 and proportional to @xmath36 .
the above result is essential in the following discussion about implementing ssb in the third part of this paper .
we will see that , if we treat the _ ricci scalar _ @xmath10 as a parameter and write the potential as @xmath37 the `` coefficient '' @xmath38 can be negative for small @xmath5 and positive for large @xmath39 in some situation .
and hence the potential @xmath40 behaves similarly to the well - known potential ( eq . (
[ higgs potential ] ) ) for higgs mechanism , and entails multiple true vacua . to see it more specifically ,
let us consider a simple case .
we consider the action in which only gravity and the scalar field are introduced , @xmath41 where @xmath42 @xmath43 @xmath26 is the ( positive ) @xmath13 dimensional gravitational constant , @xmath44 is the absolute value of the determinant of the metric tensor @xmath45 , @xmath22 is the cosmological constant , and @xmath16 is a _ positive _ coupling constant . the variation of the above action with respect to the scalar field @xmath7 and the metric tensor @xmath46 yields the field equation of @xmath7 : @xmath47 and the einstein equations : @xmath48 where @xmath49 ( note that there is no non - minimal coupling term in @xmath50 . )
the semicolon ` ; ' denotes the ` covariant derivative ' . in the brace in eq.([einstein equation ] ) , the term @xmath51 , as well as the terms @xmath52 and @xmath53 , is derived from the variation of the term @xmath54 in the action .
this term modifies the einstein equations , the gravitational constant @xmath26 and the cosmological constant @xmath22 as follows : @xmath55 where the ` modified cosmological constant ' @xmath56 and the ` effective gravitational constant ' @xmath57 are defined as @xmath58 note that both @xmath56 and @xmath57 are not constant parameters , but have the dependence on the square of the scalar field .
taking trace of both sides of the _ modified _
einstein equations ( [ modified einstein equation ] ) gives us the _ ricci scalar _
@xmath10 as a function of the scalar field @xmath7 and its covariant derivatives : @xmath59 for exploring the existence of multiple true vacua entailed by the action ( eq .
( [ action ] ) ) or the potential @xmath40 ( eq . ( [ potential including non - minimal coupling ] ) ) , we need to find out the constant solutions for the scalar field @xmath7 and explore their stability . for constant @xmath7 , the field equation of @xmath7 ( [ field equation of phi ] ) and the _ modified _ einstein equations ( [ modified einstein equation ] )
become @xmath60 @xmath61 where the ` _ _ effective _ _ potential energy ( or vacuum energy ) ' @xmath62 and the ` _ _ effective _ _ cosmological constant ' @xmath36 are defined as @xmath63 @xmath64 from eq .
( [ lambda dominated einstein equation ] ) , we can see that the constant-@xmath7 solution is corresponding to a @xmath32-dominated world , and hence the situation here becomes as simple as the case mentioned in last section . taking trace of both sides of eq .
( [ lambda dominated einstein equation ] ) ( or applying the condition @xmath65 constant to eq.([ricci scalar ] ) ) gives us the _ ricci scalar _
@xmath10 as a function of the scalar field @xmath7 : @xmath66 using the above equation , eq .
( [ field equation for constant phi ] ) becomes @xmath67 \phi = 0 .\ ] ] then we can obtain the constant solutions for the scalar field @xmath7 : @xmath68 where @xmath69 is defined by @xmath70 the corresponding metric tensor @xmath71 ( the solution(s ) of the modified einstein equations corresponding to @xmath72 ( eq.([lambda dominated einstein equation ] ) ) ) will be the metric tensor of de sitter or anti - de sitter space - time , depending on the ` _ _ effective _ _ cosmological constant ' @xmath36 is positive or negative .
now we need to explore the stability of these three constant-@xmath19 solutions ( @xmath72 , @xmath73 ) .
considering the small variations around these solutions : @xmath74 where @xmath75 is the _ ricci scalar _ for @xmath76 , the field equation of @xmath7 ( [ field equation of phi ] ) becomes ( up to @xmath77 ) : @xmath78 while the equation for the _ ricci scalar _ @xmath10 ( [ ricci scalar ] ) becomes ( up to @xmath79 ) : @xmath80 using above equations and eq .
( [ ricci scalar for constant phi ] ) , we can obtain the field equations of @xmath81 from eq .
( [ field equation of delta phi with r ] ) : for @xmath82 , @xmath83 \delta \phi = 0 , % \label{field equation of delta phi at zero phi without r}\ ] ] where @xmath84 so the solution @xmath85 is unstable or stable for @xmath86 is greater or smaller than @xmath87 .
( note that @xmath88 is the same condition with the one for @xmath89 ) . on the other hand , for @xmath90 , @xmath91 ^{-1 } % \left ( \frac{\kappa \xi \mu ^{2 } v^{2}}{1+\kappa \xi v^{2 } } \right ) \delta \phi = 0 , \ ] ] where @xmath92 ^{-1 } % \left ( \frac{\kappa \xi \mu ^{2 } v^{2}}{1+\kappa \xi v^{2 } } \right ) > 0 \quad \mbox{for } \quad d > 2 .
% \label{condition for vev - phi field equation}\ ] ] so the solutions @xmath93 are stable for @xmath0 .
( remember that , from eq .
( [ vev ] ) , the condition @xmath94 should be fulfilled in order to ensure the existence of the solutions @xmath93 . )
consequently we can conclude that , considering the space - time dimension @xmath0 , for @xmath95 , there is only one constant solution @xmath96 , which is stable and corresponding to one true vacuum ; while for @xmath97 , there are one unstable constant solution @xmath96 and two stable constant solutions @xmath93 which are corresponding to two true vacua .
we have shown that the potential @xmath98 does entail multiple true vacua and hence be able to result in spontaneous symmetry breaking under suitable conditions . in particular , we can see , from eq .
( [ vev])([conditions for zero - phi field equation])([condition for vev - phi field equation ] ) , that the conditions @xmath99 should be fulfilled in order to produce spontaneous symmetry breaking in our scenario . in our scenario , the gravitational constant @xmath26 will undergo a `` phase transition '' accompanying the spontaneous symmetry breaking : @xmath100 when the scalar field @xmath7 rolls down to one of the true vacua from the origin .
it is interesting to see the large @xmath86 limit of @xmath57 : @xmath101 accordingly , even though the original gravitational constant @xmath26 could be arbitrarily large , we still can get a small effective gravitational constant @xmath57 by choosing ( or tuning ) the parameters : @xmath102 , @xmath16 , and @xmath22 such that the fraction @xmath103/@xmath86 is small enough .
this phenomenon is similar to the result of the `` induced gravity '' model @xcite .
it would be of interest to find , numerically , the solutions in our model , that is , the solutions of the coupled equations the field equation of the scalar field @xmath7 ( [ field equation of phi ] ) and the modified einstein equations ( [ modified einstein equation ] ) . except the trivial solutions : @xmath104 , @xmath105 , which have been discussed in last section ,
we want to find out non - trivial solutions , especially the domain - wall solution(s ) for studying the domain - wall formation under our new ssb scenario and the relationship between such kind of domain wall and the randall - sundrum scenario @xcite @xcite in 5d space - time . in our preliminary sight on it
, we do see that the configuration of this domain wall is similar to the configuration of the brane world in the randall - sundrum scenario @xcite , in which the appropriate `` friedmann equation(s ) '' ( describing the expansion or contraction of the brane world ) can be obtained successfully @xcite .
we also wish to explore more details on it .
this work is supported in part by the national science council , taiwan , r.o.c .
( nsc 89 - 2112-m-002 062 ) and by the cospa project of the ministry of education ( moe 89-n - fa01 - 1 - 4 - 3 ) . p. w. higgs , phys .
* 12 * ( 1964 ) 132 ; p. w. higgs , phys .
* 145 * ( 1966 ) 1156 ; t. w. kibble , phys .
* 155 * ( 1967 ) 1554 .
i. l. buchbinder , s. d. odintsov and i. l. shapiro , _ effective action in quantum gravity _ ( iop pub . ,
bristol , uk , 1992 ) .
r. penrose , in _ relativity , groups and topology _ , edited by c. dewitt and b. s. dewitt ( gordon and breach , new york , 1964 ) .
r. fakir and w. g. unruh , phys .
d * 41 * ( 1990 ) 1783 ; d. s. salopek , j. r. bond and j. m. bardeen , phys .
d * 40 * ( 1989 ) 1753 ; e. w. kolb , d. s. salopek and m. s. turner , phys .
d * 42 * ( 1990 ) 3925 ; r. fakir , s. habib and w. unruh , astrophys .
j. * 394 * ( 1992 ) 396 ; r. fakir and s. habib , mod .
lett . a * 8 * ( 1993 ) 2827 ; m. s. turner and l. m. widrow , phys .
d * 37 * ( 1988 ) 3428 ; r. fakir and w. g. unruh , phys . rev .
d * 41 * ( 1990 ) 1792 .
g. m. shore , ann .
* 128 * ( 1980 ) 376 ; b. allen , nucl .
b * 226 * ( 1983 ) 228 ; k. ishikawa , phys .
d * 28 * ( 1983 ) 2445 ; i. l. buchbinder and s. d. odintsov , class . quant . grav .
* 2 * ( 1985 ) 721 ; yad .
* 42 * ( 1985 ) 1268 ; j. perez - mercader , phys .
b * 223 * ( 1989 ) 300 .
d. j. toms , phys .
d * 25 * ( 1982 ) 2536 ; l. h. ford and d. j. toms , phys .
d * 25 * ( 1982 ) 1510 ; a. vilenkin and l. h. ford , phys .
d * 26 * ( 1982 ) 1231 ; a. vilenkin , nucl .
b * 226 * ( 1983 ) 504 ; v. n. melnikov and s. v. orlov , phys .
lett . a * 70 * ( 1979 ) 263 .
l. randall and r. sundrum , phys .
* 83 * ( 1999 ) 4690 [ hep - th/9906064 ] .
l. randall and r. sundrum , phys . rev .
* 83 * ( 1999 ) 3370 [ hep - ph/9905221 ] .
e. e. flanagan , s. h. tye and i. wasserman , phys .
d * 62 * ( 2000 ) 044039 [ hep - ph/9910498 ] ; p. kraus , jhep*9912 * ( 1999 ) 011 [ hep - th/9910149 ] ; c. csaki , m. graesser , c. kolda and j. terning , phys .
b * 462 * ( 1999 ) 34 [ hep - ph/9906513 ] . | we propose a new scenario to implement spontaneous symmetry breaking in the space - time of an arbitrary dimension ( @xmath0 ) by introducing the non - minimal coupling between the scalar field and the gravity . in this scenario ,
the usage of the familiar @xmath1 term , which is non - renormalizable for @xmath2 , can be avoided altogether . _
pacs : _ 11.10.kk ; 04.50.+h ; 11.15.ex + _ keywords : _ spontaneous symmetry breaking ; non - minimal coupling ; arbitrary dimension |
in quantum field theories based on operator formalism , the creation and annihilation field operators @xmath0 and @xmath1 are the fundamental objects creating and annihilating resp .
particles in quantum states denoted by quantum numbers @xmath2 , as for example , by the momentum @xmath3 and the spin projection @xmath4 . in any quantum field theory
the number of the operator @xmath5 pairs is infinite .
these operators satisfy the canonical commutation or anticommutation relations @xmath6 the operators @xmath7 act on state vectors @xmath8 which span a hilbert space @xmath9 . in order to achieve a unique specification of the commutator or anticommutator ring of the operators ( [ 1.1 ] ) , in addition to ( [ 1.1 ] ) one requires the existence of a vacuum state @xmath10 in the hilbert space @xmath9 for which @xmath11 for all @xmath2 . in this case , the hilbert space @xmath9 is a space for a representation of the commutator or anticommutator ring ( [ 1.1 ] ) with the auxiliary condition ( [ 1.2 ] ) .
as long as the number of the operators @xmath7 entering the algebraic structure ( [ 1.1 ] ) and ( [ 1.2 ] ) is finite , there exists only one inequivalent representation for the algebraic relations ( [ 1.1 ] ) and ( [ 1.2 ] ) .
however , in quantum field theories describing systems with an infinite number of degrees of freedom , the algebraic structure ( [ 1.1 ] ) has infinitely many inequivalent representations @xcite .
intuitively speaking , one can say that there exist infinitely many different and inequivalent matrix realizations of the operators @xmath1 and @xmath0 satisfying the same algebraic structure as ( [ 1.1 ] ) and ( [ 1.2 ] ) .
the situation reminds us very distantly of a lie algebra of a non - compact group which has infinitely many unitary irreducible representations for its generators realized in forms of infinitely dimensional matrices .
in contrast to the aforementioned lie algebra with a finite number of its generators , the canonical ring ( [ 1.1 ] ) involves infinite number of the elements @xmath1 and @xmath0 which can be realized by infinitely many different and inequivalent representations in forms of infinitely dimensional matrices . the operators @xmath1 and @xmath0 entering the ring ( [ 1.1 ] ) are assumed to form a complete set of operators which means that every operator in a quantum field theory can be built up out of them .
thus a grand canonical hamiltonian @xmath12 governing the dynamics of a given physical system is expressed as a given function of @xmath1 and @xmath0 , i.e. , @xmath13 the grand canonical partition function @xmath14 is expressed as the density matrix trace @xmath15 where @xmath16 is the inverse temperature .
all thermodynamic properties of the system are determined by the grand canonical potential @xmath17 the statistical average values corresponding to physical observables associated with operators @xmath18 are determined by the relations @xmath19 the grand canonical potential ( [ 1.5 ] ) and statistical average values ( [ 1.6 ] ) specify the so - called gibbs state of the system governed by the given hamiltonian ( [ 1.3 ] ) .
since the commutator or anticommutator ring ( [ 1.1 ] ) admits infinitely many inequivalent representations for the operators @xmath1 and @xmath0 , it implies that one has to associate the corresponding inequivalent representations to the hamiltonian @xmath12 , partition function @xmath14 and grand canonical potential @xmath20 .
again , intuitively speaking , one can say that the matrix form of the hamiltonian ( [ 1.3 ] ) is distinct for each inequivalent representation of the ring ( [ 1.1 ] ) . this implies that the partition function ( [ 1.4 ] ) , grand canonical potential ( [ 1.5 ] ) and statistical average values of physical observables ( [ 1.6 ] ) are distinct for each inequivalent representation of ( [ 1.1 ] ) . in other words ,
the same hamiltonian @xmath12 gives rise to different results for @xmath14 , @xmath20 and @xmath21 corresponding to chosen inequivalent representations of the canonical ring ( [ 1.1 ] ) .
it implies that in each quantum field theory a given hamiltonian can give rise to infinitely many gibbs states .
this theoretical conclusion seems to be in a conflict with the experience because every physical system resides always in a single gibbs state for given values of thermodynamic variables , like temperature @xmath22 , volume @xmath23 and particle number @xmath24 .
the single gibbs state corresponds to a single inequivalent representation of the commutator or anticommutator ring ( [ 1.1 ] ) .
the answer how one should select an appropriate single inequivalent representation for a system at given values @xmath22 , @xmath23 and @xmath24 out of infinitively many representations is uniquely given by the second law of thermodynamics .
the second law of thermodynamics requires , e.g. , the helmholtz free energy @xmath25 at given values of the thermodynamic variables @xmath22 , @xmath23 and @xmath24 to be minimal with respect to any free parameters entering @xmath25 . theoretically , it means that one should evaluate the helmholtz free energies @xmath25 for all inequivalent representations of ( [ 1.1 ] ) and select that single one which corresponds to their infimum at the given range of the thermodynamic variables @xmath22 , @xmath23 and @xmath24 . a concrete example will demonstrate how it is done in practice . for detailed understanding of inequivalent representations of the commutator or anticommutator ring ( [ 1.1 ] ) of field operators role in practical applications we will study them from three different aspects .
first , we will explicitly construct a certain class of inequivalent representations of the anticommutator ring ( [ 1.1 ] ) of field operators .
second , we will show how one , in practical calculations , tacitly selects a single inequivalent representation by choosing an appropriate perturbation theory . at third
, theoretical implications of inequivalent representations of electron field operator anticommutator ring will be explicitly demonstrated on the bcs model hamiltonian of superconductivity @xcite .
more specifically , a new class of inequivalent representations of ( [ 1.1 ] ) , which has not been known till now and leads to a new unexpected superconducting state , will be constructed .
for the purpose of practical applications to the bcs model hamiltonian @xcite and for the sake of simplicity , we consider a complete set of annihilation and creation operators @xmath26 and @xmath27 of the fermion type .
let the index @xmath28 run over integer numbers over the interval @xmath29 and @xmath4 denotes spin @xmath30 projection of a fermion , i.e. , @xmath31 . in order to have a quantum field theory
, we take the limit @xmath32 .
the field operators @xmath26 and @xmath27 satisfy the anticommutator ring @xmath33 with the subsidiary condition @xmath34 on the vacuum state @xmath10 for all @xmath35 .
the representation space for the anticommutator ring ( [ 2.1 ] ) with the subsidiary condition ( [ 2.2 ] ) can be chosen to be the hilbert space @xmath9 spanned by the basis vectors @xmath36 defined by the formula @xmath37 where @xmath38 are the occupation numbers of fermions in states @xmath35 , and @xmath39 denotes an infinite number array of items @xmath40 and @xmath41 .
each such infinite array specifies one of the basis vectors of the hilbert space @xmath9 .
next we construct a class of inequivalent representations of the anticommutator ring ( [ 2.1 ] ) by adopting the same approach as outlined in haag s work @xcite , however , for boson operators .
we start from the operators @xmath26 and @xmath27 obeying ( [ 2.1 ] ) and introduce the `` unitary '' transformations @xmath42 where @xmath43 is the hermitian `` operator '' @xmath44 and @xmath45 are arbitrary real parameters . the anticommutation relations for the transformed operators @xmath46 and @xmath47 are , of course , the same as given by ( [ 2.1 ] ) .
the operator @xmath48 can be expressed as the following infinite product @xmath49\end{aligned}\ ] ] where @xmath50 the transformations ( [ 2.4 ] ) , if evaluated through ( [ 2.4 ] ) , are similar to the well - known bogoliubov - valatin transformations @xcite @xmath51 where @xmath52 in the limit @xmath32 , the operator @xmath48 given by ( [ 2.4 ] ) is not a proper operator but transforms every vector @xmath8 of the hilbert space @xmath9 into one @xmath53 of a hilbert space @xmath54 with unexpected properties explained as follows .
let us denote by @xmath55 any basis vector of @xmath9 given by the formula @xmath56 where @xmath57 .
all the basis vectors @xmath55 form a subspace of @xmath9 .
the transformation @xmath48 transforms every basis vector @xmath55 into one @xmath58 of @xmath54 , given by the formula @xmath59\sin \alpha_{k}+\big(a^{\dagger}_{k,+}a^{\dagger}_{-k,-}\big)^{n_{k}}\cos \alpha_{k}\bigg\}\phi_0\;.\end{aligned}\ ] ] the result is that the scalar product @xmath60 for every basis vector @xmath8 of @xmath9 given by ( [ 2.3 ] ) is either identically equal to zero or equal to the infinite product @xmath61 where @xmath62 however , the infinite product ( [ 2.12 ] ) diverges also to zero in the limit @xmath32 .
this is so because by noting the properties of the coefficients @xmath63 and @xmath64 given by ( [ 2.13 ] ) we see that the scalar product ( [ 2.12 ] ) reduces to the infinite product of the following type @xmath65 the last relation implies that the hilbert space @xmath54 spanned by the transformed state vectors @xmath53 contains a subspace of the state vectors @xmath66 which are orthogonal to every vector @xmath8 of @xmath9 .
we make the same conclusion as in haag s work @xcite : @xmath46 and @xmath47 given by ( [ 2.4 ] ) are operators satisfying the same canonical ring as ( [ 2.1 ] ) , i.e. , @xmath67 but there is no proper unitary transformation connecting these two operator systems
. the operators @xmath68 and the operators @xmath69 act in two different hilbert spaces @xmath9 and @xmath54 respectively . in haag
s terminology , they belong to inequivalent representations of the same anticommutator ring ( [ 2.1 ] ) or ( [ 2.15 ] ) of the field operators . it is a simple matter to show that among all transformed vectors @xmath53 there is no one such as @xmath70 for which @xmath71 for all @xmath35 .
next we supplement the new hilbert space @xmath54 spanned by all state vectors @xmath53 with the vacuum state @xmath70 satisfying the auxiliary condition ( [ 2.16 ] ) . each inequivalent representation of the anticommutator ring ( [ 2.15 ] ) with the subsidiary condition ( [ 2.16 ] ) is specified by a chosen infinite set of parameters @xmath45 entering the transformations ( [ 2.4])-([2.9 ] ) .
thus , the number of the inequivalent representations of ( [ 2.15 ] ) is indeed infinite . in contradistinction to inequivalent representations of lie algebras , where the representations are specified by a finite number of casimir operator eigenvalues , the inequivalent representations of the anticommutator ring ( [ 2.15 ] ) of the field operators @xmath46 and @xmath47 are specified by infinite set of parameters . our analysis of inequivalent representations of the anticommutator ring of the field operators ( [ 2.1 ] ) , as presented above , can be generalized in a straightforward way to any set of quantum numbers @xmath35 and to many other classes of inequivalent representations both for fermions and bosons .
next we consider a hamiltonian @xmath12 governing a physical system with an infinite number of degrees of freedom as a given function of the operators @xmath26 and @xmath27 , i.e. , @xmath72 the field operators @xmath26 and @xmath27 and the corresponding hamiltonian ( [ 2.17 ] ) act in the representation space which is the hilbert space @xmath9 spanned by the basis vectors given by ( [ 2.3 ] ) . in this representation
we have the corresponding partition function @xmath73 and statistical average values @xmath21 of physical observables associated with operators @xmath18 given by the relations @xmath74 for each inequivalent representation of the anticommutator ring ( [ 2.15 ] ) we can construct the transformed hamiltonian @xmath75 in its normal form by employing the canonical anticommutator relations ( [ 2.15 ] ) .
the corresponding partition function @xmath76 can be different from that given by ( [ 2.18 ] ) because the operators @xmath77 and @xmath78 act in two different hilbert spaces @xmath9 and @xmath54 respectively . by the same way ,
the statistical average value @xmath21 corresponding to a physical observable @xmath79 can be different from that given by ( [ 2.19 ] ) .
these conclusions may seem to be paradoxes residing in facts as if physical observables were dependent on our will how we select a single inequivalent representation for the anticommutator ring of the field operators ( [ 2.15 ] ) .
there is , in fact , no freedom in selecting an appropriate inequivalent representation of ( [ 2.15 ] ) .
as we have mentioned in the introduction , the second law of thermodynamics dictates uniquely which inequivalent representation is relevant for describing the physical system at given values of thermodynamic variables .
one may even incorrectly believe that the conclusion concerning the different results for the same physical quantities given by the relations ( [ 2.18 ] ) and ( [ 2.21 ] ) or by the relations ( [ 2.19 ] ) and ( [ 2.22 ] ) are wrong .
such an incorrect belief is supported by the formal appearance of the transformation ( [ 2.20 ] ) which seems to be a similarity transformation .
the unpermited application of the cyclic properties for the trace of infinitely dimensional matrices product would lead to incorrect conclusions that the results of the relations ( [ 2.18 ] ) , ( [ 2.20 ] ) or ( [ 2.19 ] ) , ( [ 2.21 ] ) are identical .
we conclude this section by stating that in quantum field theory a given hamiltonian @xmath12 of a system leads to distinct results for the partition function @xmath14 and for the statistical average values of physical observables @xmath21 depending on a selected inequivalent representation of the anticommutator ring of field operators .
these conclusions will be explicitly demonstrated on the bcs model hamiltonian of superconductivity .
in this section we elucidate how one tacitly selects a single inequivalent representation out of infinitely many representations of the commutator or anticommutator ring ( [ 1.1 ] ) of field operators in a practical application of quantum field theory . in quantum field theories with interactions between fields there is not known even one physical example with an exact solution . in all practical applications one divides hamiltonian @xmath12 of a system into the sum @xmath80 where @xmath81 is called the unperturbed hamiltonian and the remaining term @xmath82 is called the perturbative part .
the unperturbed hamiltonian @xmath81 is chosen in a way to be exactly diagonalized and by this fact its effects are treated exactly .
its eigenvalues @xmath83 , where @xmath84 denotes an array with an infinite number of items form a complete basis of a hilbert space @xmath9 .
the hilbert space @xmath9 is the representation space for a single inequivalent representation of the commutator or anticommutator ring ( [ 2.1 ] ) of the field operators @xmath26 and @xmath27 entering the hamiltonian ( [ 3.1 ] ) .
the unperturbed partition function @xmath85 can be exactly evaluated and is typical for the chosen inequivalent representation .
the total partition function @xmath14 is expressed by the perturbation series @xmath86 where the symbol @xmath22 stands for the time - ordered product , @xmath87 and @xmath88 denotes the statistical average value of the operator inside the brackets @xmath89 with respect to the unperturbed hamiltonian @xmath81 .
it is needless to say that the statistical average values ( [ 3.5 ] ) are evaluated in the chosen inequivalent representation .
however , the splitting of the total hamiltonian as given by ( [ 3.1 ] ) is not unique .
one can equally well divide the same hamiltonian as @xmath90 where the new unperturbed hamiltonian @xmath91 is by definition different from @xmath81 and is not related to @xmath81 by any proper unitary transformation .
the new unperturbed hamiltonian @xmath91 is assumed to be diagonalized by such transformations of the field operators @xmath26 and @xmath27 like ( [ 2.4 ] ) in order to achieve its diagonal form @xmath92 the requirement to have @xmath93 in its diagonal form puts certain constraints on the transformation parameters @xmath45 entering the transformations ( [ 2.4])-([2.9 ] ) .
in other words , all transformation parameter @xmath45 values are determined by a finite set of physical parameters present in the chosen unperturbed hamiltonian @xmath94 .
since the infinite set of parameters @xmath45 specifies a single inequivalent representation of the commutator or anticommutator ring of the field operators ( [ 2.15 ] ) , with new chosen unperturbed hamiltonian @xmath91 one again tacitly selects another single inequivalent representation .
the eigenvalues @xmath95 of @xmath96 form again a complete basis of a new hilbert space @xmath54 for the new selected inequivalent representation . in this inequivalent representation one gets the unperturbed partition function @xmath97 which is , of course , different from ( [ 3.2 ] ) by the definitions ( [ 3.1 ] ) and ( [ 3.6 ] ) .
the corresponding total partition function @xmath98 is also distinct from that given by ( [ 3.3 ] ) because the same hamiltonian @xmath12 has two different , so to speak , `` matrix '' realizations ( [ 3.1 ] ) and ( [ 3.3 ] ) corresponding to two different inequivalent representations of the anticommutator ring of field operators ( [ 2.1 ] ) or ( [ 2.15 ] ) . by the same approach as outlined above , one can continue to study a series of inequivalent representations associated with a given hamiltonian @xmath12 . by selecting a series @xmath99 of unperturbed hamiltonians @xmath100
, @xmath101 , @xmath102 one can explore physical properties of the corresponding series of different gibbs states associated with the same hamiltonian @xmath12 .
the aforementioned methods will be explicitly demonstrated for the investigation of physical properties of several distinct gibbs states associated with the bcs model hamiltonian in theory of superconductivity @xcite .
the bcs hamiltonian has been deliberately chosen from three different reasons . at first , the bcs hamiltonian is generally very well - known , extraordinary simple and despite of its simplicity it has been very successful in explaining properties of a large class of the so - called low temperature superconductors ( lts ) on terms of only two phenomenological material parameters . on the other hand , properties of high temperature superconductors ( hts ) having anisotropic layered structures are claimed to be unexplainable on the basis of the bcs theory . at second , the bcs theory as a theory with a nontrivial interactions between electrons has provided us with two different solutions for the gibbs states which are asymptotically exact in the thermodynamic limit . at third , the bcs theory has been studied from various aspects for many years .
that is why there is a generally spread belief that its consequences are completely exhausted and therefore there are no motivations for its further exploration .
however , its consequences has not been studied from the point of view of inequivalent representations of the anticommutator ring of electron field operators ( [ 2.1 ] ) or ( [ 2.15 ] ) yet . since the ring has infinitely many inequivalent representations the bcs hamiltonian should have infinitely many gibbs states .
it is , therefore , worthwhile to undertake the task to find at least one new gibbs state associated with the bcs hamiltonian .
the bcs theory is based on the following well - known grand canonical hamiltonian @xmath103 where @xmath104 is the kinetic energy of an electron ( in the state specified by the wave vector @xmath3 ) counted from the chemical potential @xmath84 which can be approximated by the fermi energy @xmath105 , i.e. , @xmath106 .
the index @xmath107 denotes the spin @xmath30 projection of an electron , the symbols @xmath1 and @xmath108 are annihilation and creation operators of electrons in the states @xmath109 resp .
finally , @xmath110 is the squared electron - phonon coupling constant , @xmath111 is the debye frequency , @xmath23 is the volume of the system .
@xmath112 denotes the electron kinetic energy operator and @xmath82 is the interaction term .
the sum over @xmath3 and @xmath113 in @xmath82 is restricted by the conditions @xmath114 , @xmath115 , as indicated by the appropriate step functions @xmath116 in the relation ( [ 4.1 ] ) .
all thermodynamic properties of the system governed by the hamiltonian ( [ 4.1 ] ) are determined by the grand canonical partition function @xmath117 the grand canonical potential @xmath118 specifies the gibbs state of the system and provides us with complete information on thermodynamic properties of the system described by the hamiltonian ( [ 4.1 ] ) .
@xmath20 is a function of the thermodynamic variables @xmath22 , @xmath23 and @xmath84 , and in addition to them , it is also a function of the coupling constant @xmath110 , i.e. , @xmath119 . from the definitions ( [ 4.2 ] ) and ( [ 4.3 ] ) it follows @xmath120 where the statistical average @xmath121 is , of course , a non trivial function of @xmath110 , i.e. , @xmath122 .
suppose that one has calculated @xmath123 , then he can integrate the equation ( [ 4.4 ] ) to get the exact relation @xmath124 where @xmath119 is the grand canonical potential as defined by ( [ 4.2 ] ) and ( [ 4.3 ] ) and @xmath125 is the grand canonical potential corresponding to the ideal electron gas .
the grand canonical potential difference ( [ 4.5 ] ) is particularly convenient for two reasons .
first , it expresses physical anomalies associated with the interaction term @xmath82 above the background corresponding to properties of the ideal electron gas .
second , it exhibits manifestly the dependence of @xmath20 on a chosen inequivalent representation of the anticommutator ring ( [ 2.1 ] ) of the electron field operators entering the hamiltonian @xmath12 . in practical calculations one can not evaluate the exact average value @xmath121 and must resort to a perturbation theory by selecting an unperturbed hamiltonian @xmath81 which can be exactly diagonalized . with a chosen unperturbed hamiltonian @xmath81 one calculates the unperturbed grand canonical partition function ( [ 3.2 ] ) and the corresponding statistical averages @xmath126 of operators @xmath127 defined by the relations @xmath128 the average value @xmath123 entering the relations ( [ 4.4 ] ) and ( [ 4.5 ] ) can be expressed in the form @xmath129 where @xmath130 represents symbolically all contributions coming from remaining perturbative terms .
first we choose @xmath81 to be the electron kinetic energy operator in ( [ 4.1 ] ) , i.e. , @xmath131 in this selection , the unperturbed hamiltonian @xmath100 is already diagonalized and the complete operator system is given by the electron annihilation and creation operators @xmath1 and @xmath108 resp .
in the quantum states @xmath132 .
these operators obey the canonical anticommutator ring ( [ 2.1 ] ) with the subsidiary condition ( [ 2.2 ] ) . in this case
the hilbert space @xmath133 is the space of a single inequivalent representation of the anticommutator ring ( [ 2.1 ] ) with the auxiliary condition ( [ 2.2 ] ) .
each basis vector @xmath8 is then specified by an infinite array @xmath134 of the occupation numbers @xmath135 for each one particle state @xmath109 , i.e. , @xmath136 with this choice of the unperturbed hamiltonian @xmath100 one gets the spectrum @xmath137 of elementary excitations @xmath138 and the statistical average values @xmath139 valid in all orders of the perturbation theory .
the statistical average values valid in the first order of the perturbation theory have the following forms @xmath140 @xmath141 -\tanh \frac{\beta \hbar \omega_d}{2 } \bigg)\;,\end{aligned}\ ] ] where @xmath142 is the density of electron states for one spin projection at the fermi surface . in evaluating @xmath143
the relations @xmath144 and @xmath145 have been used . from the relation ( [ 4.13 ] )
one sees that @xmath143 does not scale with the volume @xmath23 and the same is true for all remaining terms @xmath130 in ( [ 4.7 ] ) , as was formally proven in @xcite and demonstrated by explicit calculations in @xcite .
the result ( [ 4.13 ] ) together with ( [ 4.7 ] ) implies the relation @xmath146 the right - hand side of the last equation is not proportional to the volume @xmath23 and by this fact the density of the grand canonical potential @xmath147 approaches the density @xmath148 corresponding to the ideal electron gas in the thermodynamic limit @xmath149 .
in other words , the interaction term @xmath82 in ( [ 4.1 ] ) has no macroscopic effects on @xmath150 in the thermodynamic limit @xmath151 provided that one has chosen the unperturbed hamiltonian @xmath152 as given by ( [ 4.8 ] ) , i.e. , @xmath153 for the chosen inequivalent representation in the thermodynamic limit .
next we choose the unperturbed hamiltonian @xmath101 , to be the one corresponding to the standard bcs theory of superconductivity , i.e. in the form @xmath154 where @xmath155 and @xmath156 are the following average values @xmath157 called the gap functions .
@xmath101 is the hamiltonian ( [ 4.1 ] ) in the so called mean field approximation .
the unperturbed hamiltonian @xmath101 is diagonalized by means of transformations ( [ 2.8 ] ) , ( [ 2.9 ] ) with the transformation parameters @xmath158 determined by the formula @xmath159 all the transformation parameters @xmath158 are determined in terms of the physical parameters entering @xmath101 and specify the single inequivalent representation of the anticommutator ring ( [ 2.15 ] ) with the auxiliary condition ( [ 2.16 ] ) .
its representation space is the hilbert space denoted by @xmath160 .
the unperturbed hamiltonian @xmath101 has the following diagonal form @xmath161\;,\end{aligned}\ ] ] where @xmath162 is the energy spectrum of elementary excitations given by the formulae @xmath163 the energy spectrum @xmath162 as the function of @xmath164 is schematically depicted on fig .
[ fig1 ] . as is seen from the fig .
[ fig1 ] the energy spectrum @xmath162 has discontinuities at two fixed points @xmath165 which are specified by the material parameter @xmath111 .
the unperturbed hamiltonian @xmath101 as given by ( [ 4.17 ] ) or ( [ 4.20 ] ) both determines and is determined by the average values ( [ 4.18 ] ) .
thus the relation ( [ 4.18 ] ) is , in fact , the self - consistency condition known in all text books @xcite as the gap equation @xmath166 the gap function @xmath155 is different from zero , @xmath167 , for @xmath168 , where @xmath169 is the critical temperature given by the formula @xcite @xmath170 \doteq 1.13\;t_d \exp{\bigg[-\frac{1}{gn(0)}}\bigg ] \;.\end{aligned}\ ] ] here @xmath171 is euler constant and @xmath172 is the debye temperature .
the critical temperature @xmath169 , as seen from ( [ 4.23 ] ) , is proportional to @xmath173 and is a singular function of @xmath110 at @xmath174 .
the solution to the gap equation ( [ 4.22 ] ) provides @xmath175 as a function of @xmath22 and @xmath110 , which can not be expressed in an analytic form , but only numerically . the only contribution to @xmath123 which survives the thermodynamic limit @xmath151 has the form @xmath176 because all remaining terms denoted by @xmath130 in ( [ 4.7 ] ) become negligible in this limit as formally proven in @xcite and demonstrated by explicit calculations in @xcite . two different results ( [ 4.13 ] ) and ( [ 4.24 ] ) for the average value of the same physical abservable @xmath121 exhibit clearly and evidently the important role of the inequivalent representations in practical applications .
namely , @xmath121 has no macroscopic effects if evaluated in the representation hilbert space @xmath133 , however , @xmath121 has a very relevant macroscopic contribution if evaluated in the representation hilbert space @xmath160 .
the result ( [ 4.24 ] ) with the relation ( [ 4.5 ] ) gives the expression for the grand canonical potential @xmath150 in the form @xmath177 where @xmath178 is the thermodynamic critical magnetic field defined by the relation @xmath179 the solution to the bcs hamiltonian ( [ 4.1 ] ) , as represented by the relations ( [ 4.20])-([4.26 ] ) , describes , as is well - known , the superconducting state of the system governed by the hamiltonian ( [ 4.1 ] ) .
this solution will be referred as to the standard solution of the bcs theory of superconductivity .
we have discussed it briefly in order to demonstrate the role of inequivalent representations of the canonical anticommutator ring ( [ 2.1 ] ) or ( [ 2.15 ] ) of electron field operators on solutions which are generally very well - known .
the main purpose of this section is to explore the physical implications of an additional inequivalent representation of the anticommutator ring ( [ 2.15 ] ) associated with the hamiltonian ( [ 4.1 ] ) of the bcs theory . without any physical motivation
we choose the third form of the unperturbed hamiltonian @xmath180 , as given by @xmath181 where @xmath182 are gap functions which are dependent on the wave vector @xmath3 .
here , @xmath183 is chosen to be the absolute value of the wave vector corresponding to the minimal energy of an electron confined in a box with the edges @xmath184 and @xmath185 , i.e. , @xmath186 \;.\end{aligned}\ ] ] one may say that @xmath180 is a kind of an interpolation between the hamiltonians @xmath100 and @xmath101 .
indeed , @xmath180 contains only those terms from @xmath101 which partially conserve the energy in microscopic scattering processes between electrons .
this fact is represented by the presence of the step function @xmath187 in ( [ 4.27 ] ) and ( [ 4.28 ] ) .
one may regard @xmath180 as a mathematical toy in order to search for an additional inequivalent representation of the anticommutator ring ( [ 2.15 ] ) within the framework of the same hamiltonian ( [ 4.1 ] ) .
the hamiltonian @xmath180 is again diagonalized by the well - known bogoliubov - valatin transformations ( [ 2.8])-([2.9 ] ) .
the parameters @xmath158 of the transformations ( [ 2.4])-([2.9 ] ) are defined by the relation @xmath188 with the gap functions @xmath189 depending on @xmath3 , which specify the chosen inequivalent representation of the anticommutator ring ( [ 2.15 ] ) . in this representation
the unperturbed hamiltonian @xmath180 gets its diagonal form @xmath190\;,\quad\end{aligned}\ ] ] where @xmath191 is the energy spectrum of quasiparticles represented by the field operators @xmath192 and @xmath193 . by evaluating the statistical average value
@xmath194 we get the following relation @xmath195 the last relation , when inserted into the definitions ( [ 4.28 ] ) , gives us the self - consistency condition for the gap function @xmath196 the equations ( [ 4.33 ] ) and ( [ 4.34 ] ) determining @xmath194 and the gap function @xmath189 resp .
are , of course , closely related to those ones in the standard solution to the bcs theory of the superconductivity except for the @xmath3 dependence of @xmath189 and the presence of the step function @xmath197 in ( [ 4.34 ] ) by which a partial energy conservation rule is required even for microscopic scattering processes between electrons .
the solution to the gap equation ( [ 4.34 ] ) is analyzed similarly as in the work @xcite .
the assumption that the gap function @xmath189 is a function of the magnitude @xmath198 only , simplifies the relation ( [ 4.34 ] ) substantially .
we employ the inequality @xmath199 which is valid for every vector @xmath113 in the sum ( [ 4.34 ] ) . by this fact
, we can replace the equation ( [ 4.34 ] ) by the relation @xmath200 where @xmath201 we emphasize that the equation ( [ 5.1 ] ) takes into account only linear terms in the tiny quantity @xmath183 .
the quantity @xmath202 , as given by the equation ( [ 5.2 ] ) , is , in fact , the number of wave vectors @xmath113 with the lengths @xmath203 from the interval @xmath204 . in order to evaluate the number @xmath202
we remind ourselves that the vector @xmath113 have the components given below @xmath205 where @xmath206 , @xmath207 , @xmath208 are integers and @xmath209 , @xmath210 , @xmath185 are edge lengths of a box into which the electron system is enclosed .
the given magnitude @xmath203 defines the surface of an ellipsoid with semiaxes @xmath211 , @xmath212 and @xmath213 in an euclidean space with coordinates @xmath214 and @xmath208 , i.e. , @xmath215 thus the number of the states @xmath202 is approximately equal to the volume enclosed between two surface areas of the ellipsoids corresponding to the minimal and maximal semiaxes @xmath216 and @xmath217 , i.e. , @xmath218 we analyze the number @xmath202 for two limiting cases , namely , when the ellipsoid ( [ 5.4 ] ) degenerates either into a sphere or into a thin circular disc .
the sphere corresponds to the case when the electron system under consideration is enclosed in a cube with the edges @xmath219 , i.e. , it has isotropic bulk properties .
if the ellipsoid ( [ 5.4 ] ) degenerates into a thin circular disc with @xmath220 , then the system of electrons has a form of a thin film of the given thickness @xmath221 , i.e. , it has anisotropic properties . both of these limiting cases represent objects of the physical interest .
we first analyze the case corresponding to the isotropic bulk material . in this case
we get @xmath222 by inserting the last result with @xmath223 into the equation ( [ 5.1 ] ) we get the relation @xmath224 which in the thermodynamic limit @xmath225 has the only solution @xmath226 .
thus this type of superconductivity can not exist in three dimensional isotropic materials .
we next analyze the second limiting case when the system has a form of a thin film with a given thickness @xmath221 . in this case
@xmath227 , @xmath228 and the number of the states @xmath202 is given by the relation @xmath229 by inserting the last result into equation ( [ 5.1 ] ) we get the relation @xmath230 which can have a nontrivial solution @xmath231 in the thermodynamic limit @xmath225 .
since the vectors @xmath3 in the relation ( [ 4.28 ] ) are restricted by the inequalities below @xmath232 we can replace the quantity @xmath233 in ( [ 5.9 ] ) by @xmath234 .
then from the equation ( [ 5.9 ] ) it follows that the energy spectrum @xmath162 of the quasiparticles in the superconducting state with @xmath231 must satisfy the following relation @xmath235 where @xmath236 is the dimensionless effective coupling constant and @xmath237 is the minimal energy of an electron confined in a thin layer of the thickness @xmath221 .
the equation ( [ 5.11 ] ) determines the quasiparticle energy spectrum @xmath162 as a certain function of temperature @xmath22 .
it can be satisfied only if the temperature @xmath22 is below the critical temperature @xmath169 , @xmath238 where @xmath239 is the fermi temperature .
it is interesting to point out that the relation ( [ 5.14 ] ) between the critical temperature @xmath169 and the fermi temperature @xmath239 has accidently the same form as the one found from the analysis of experimental data by uemura et al .
@xcite and discussed in @xcite on a phenomenological basis . in the phenomenological formula of the form ( [ 5.14 ] ) , proposed in @xcite and called as the uemura plot in @xcite , the phenomenological parameter @xmath240 is the same for a unique group of hts
. the universal correlation between @xmath169 and @xmath239 as given by ( [ 5.14 ] ) has been claimed @xcite to exist in all hts with planar structures .
the coincidence between the theoretical formula ( [ 5.14 ] ) and the phenomenological formula @xcite seems to be promising and stimulates us for further theoretical investigation of this type of superconductivity to be compared with the experimental data of planar hts .
the relation ( [ 5.14 ] ) determining the critical temperature @xmath169 differs qualitatively from the relation ( [ 4.23 ] ) corresponding to the standard solution of the bcs theory .
@xmath169 is proportional to @xmath239 in contrast to the debye temperature @xmath173 in ( [ 4.23 ] ) , i.e. , @xmath169 does not depend on ionic mass .
thus this type of superconductivity , despite of the fact that it is due to electron - phonon interaction , does not exhibit the isotope effect .
@xmath169 is an analytic function of @xmath110 at @xmath174 .
the proportionality coefficient @xmath240 depends on the lowest bound energy @xmath241 for the energy spectrum of electrons confined in a layer of the thickness @xmath221 , i.e. , for the thinner layer one has the higher critical temperature @xmath169 . for an isotropic bulk material @xmath242 and in the same way @xmath243 .
this property seems to represent the reason why this type of superconductivity can not exist in isotropic systems , but it can appear in systems with electrons residing on quasi two dimensional planar structures .
all these features seem to be promising that this type of superconductivity may have some relevance for the properties of hts observed in experiments . the solution to equation ( [ 5.11 ] ) for the energy spectrum @xmath162 , in the range of @xmath3
for which @xmath231 , is a quantity @xmath244 which is independent of @xmath3 .
however , it is a function of @xmath22 , i.e. , @xmath245 one can not obtain an explicit expression for the solution @xmath244 to the equation ( [ 5.11 ] ) in the form of elementary functions , but has to resort to numerical methods given below . the energy spectrum @xmath162 as a function of @xmath164 is depicted on the fig .
[ fig2 ] .
the energy spectrum @xmath162 as given by the formulae ( [ 5.15 ] ) differs quantitatively from that one given by the relation ( [ 4.21 ] ) and depicted on the fig .
[ fig1 ] corresponding to the standard solution of the bcs theory . the energy spectrum ( [ 5.15 ] )
has no discontinuities at fixed points , at @xmath246 , in contrast to the standard solution of the bcs theory .
it has a single discontinuity , at the point @xmath247 , which is a function of the temperature @xmath22 .
the function @xmath244 determined by the transcendental equation @xmath248 has the following behavior for the limiting cases @xmath249 and @xmath250 where @xmath251 thus the ratio @xmath252 is a universal constant , independent on material parameters of the thin layer under consideration . for the numerical analysis of the solution @xmath244 to the equation ( [ 5.16 ] )
it is convenient to introduce the following dimensionless variables @xmath253 where @xmath254 is the so called reduced temperature . in this notation
the function @xmath255 is expressed in the implicit form @xmath256 and its numerical representation is shown on the fig .
[ fig3 ] . from the relations ( [ 4.33 ] ) and ( [ 5.15 ] ) we get the explicit expression for the following average value @xmath257 which is of great importance for the calculation of thermodynamic properties of the novel superconducting state .
all thermodynamic properties of the superconducting state are extracted from the difference ( [ 4.5 ] ) between the grand canonical potential @xmath258 corresponding to the superconducting state for @xmath168 and @xmath259 corresponding to the normal state .
for this reason we first evaluate the average value @xmath123 . by using the result ( [ 5.22 ] )
we get the expression @xmath260 the above sum over @xmath3 and @xmath113 is evaluated as an integral by the standard substitution @xmath261 and by using the constraints ( [ 5.10 ] ) we get the result @xmath262 where @xmath130 denotes all remaining terms in the perturbation series which are negligible in the thermodynamic limit @xmath149 similarly as for the perturbation series with unperturbed hamiltonians @xmath100 and @xmath101 .
now , we are ready to express the difference ( [ 4.5 ] ) as the following integral @xmath263 where @xmath240 is the dimensionless coupling constant defined by equation ( [ 5.12 ] ) .
the last integral is luckily in a particularly convenient form because the relation ( [ 5.16 ] ) expresses @xmath240 as the following function of @xmath264 , @xmath265 the direct substitution of the last relation into the integral ( [ 6.3 ] ) leads to the formula @xmath266 two per partes integrations of the last integral and the use of the reduced variables @xmath267 and @xmath268 give us the result @xmath269\;,\end{aligned}\ ] ] where @xmath270 is a function of @xmath271 defined by the integral @xmath272 by passing from the grand canonical potential @xmath273 to the helmholtz free energy @xmath25 we get the difference @xmath274 where @xmath178 is the thermodynamic critical magnetic field given by the formula @xmath275^{\frac{1}{2 } } \;.\end{aligned}\ ] ] here , @xmath276 is the critical magnetic field at @xmath277 , @xmath278 the temperature dependence of the ratio @xmath279 is shown on the fig .
[ fig4 ] . + the theoretical curve on the fig .
[ fig4 ] is very similar to the experimental data of the upper critical magnetic field for hts @xmath280 @xcite .
the behavior of @xmath178 at @xmath281 and @xmath282 is given by the following formulae @xmath283 and @xmath284 where @xmath285 is the riemann dzeta function , @xmath286 .
now we compare the results ( [ 6.10])-([6.12 ] ) with the behavior of the critical magnetic field @xmath287 corresponding to the results of the standard solution of the bcs theory @xcite @xmath288^{\frac{1}{2}}k_bt^{'}_c \doteq 1.76[4\pi n(0)]^{\frac{1}{2}}k_bt^{'}_c \;,\end{aligned}\ ] ] @xmath289^{\frac{1}{2}}(1-\tau')\doteq 1.74h_c^{'}(0)(1-\tau')\;,\ ; t^{'}_c - t \ll t^{'}_c\end{aligned}\ ] ] and @xmath290\doteq h_c^{'}(0)\big[1 - 1.06\;\tau'^{2 } \big ] \;,\;t \ll t^{'}_c \;.\end{aligned}\ ] ] from the last formulae one sees qualitative differences between the critical magnetic fields @xmath178 and @xmath287 corresponding to the unconventional solution and to the standard solution resp . of the bcs theory .
the curve @xmath178 is convex for @xmath291 and concave for @xmath292 while @xmath287 of the standard solution is a convex curve for the full temperature interval @xmath293 .
we now calculate the specific heat anomaly from the difference ( [ 6.6 ] ) to get the result @xmath294 where @xmath295 is the specific heat of an ideal electron gas .
the temperature behavior of the specific heat anomaly ( [ 6.16 ] ) plotted as the ratio @xmath296 is shown on the fig .
[ fig5 ] .
+ the specific heat anomaly ( [ 6.16 ] ) has the following behavior in the limits @xmath297 and @xmath298 this behavior of the specific heat anomaly is qualitatively completely different from that corresponding to the standard solution @xcite of the bcs theory @xmath299\doteq 1.43+o\big[(1-\tau)^2\big ] \;,\;t_c - t \ll t_c\end{aligned}\ ] ] and @xmath300 the specific heat @xmath301 corresponding to the unconventional superconductivity ( [ 6.16 ] ) is a continuous function of @xmath22 at @xmath302 , as is seen from ( [ 6.19 ] ) , however , with a discontinuous derivative at @xmath302 in a sharp contradistinction to the standard case .
thus the unconventional solution considered here gives rise to a superconducting phase transition of the third order .
from the results ( [ 6.8])-([6.20 ] ) it is evident that the properties of superconducting states associated with two different inequivalent representations of the canonical anticommutator ring of the electron field operators ( [ 2.15 ] ) are qualitatively completely different . despite of the fact that we did not have any ambitions to explain some experimental data , but only to study the theoretical consequences coming from different inequivalent representations of the canonical anticommutator ring ( [ 2.15 ] ) within the framework of the bcs theory
, we nevertheless mention a resemblance of the theoretical results ( [ 6.16])-([6.20 ] ) to existing experimental data . the specific heat anomaly on the fig .
[ fig5 ] for @xmath303 is similar to the specific heat anomalies of cuprate hts @xcite .
the shape of the specific heat anomaly as @xmath22 approaches to zero , as given by ( [ 6.20 ] ) , tells us that the specific heat @xmath301 in this novel superconducting state behaves as a polynomial of @xmath22 , i.e. , similarly as electronic components of ordinary metals in the normal state .
this behavior is again , perhaps accidentally , consistent with experimental data of cuprates hts @xcite and is completely different from the behavior of lts described by the standard solution of the bcs theory .
we also recall that the specific heat jump at @xmath169 for the standard solution of the bcs theory ( [ 6.21 ] ) is given by a universal number which is fairly consistent with experimental data for lts
. however , the novel superconducting state discussed here relates the specific heat anomaly ( [ 6.16 ] ) to the normal specific heat @xmath304 by a material parameter @xmath305 .
this feature seems to be again in a correspondence with experimental data of hts @xcite .
for example , the specific heat anomaly of @xmath306 which ranks highest among hts , does not exceed @xmath307 of the normal specific heat .
the situation for @xmath308 is about three times worse .
thus it seems that the properties of the unconventional solution to the bcs hamiltonian investigated in this paper may have some relevance for understanding of the superconductivity mechanism in hts . the system of electrons under consideration , however , stabilizes in a phase corresponding to the minimal helmholtz free energy .
thus , for completeness , we analyze the stability of the novel superconducting phase resulting from the unperturbed hamiltonian @xmath180 with respect to the standard bcs phase arising from the unperturbed hamiltonian @xmath101 .
the helmholtz free energy in each of the superconducting phases is determined by the critical magnetic fields @xmath178 and @xmath309 as given by the relation ( [ 6.8 ] ) .
the critical magnetic field @xmath178 for the novel superconducting phase is given by the relations ( [ 6.9])-([6.10 ] ) .
the temperature behavior of the critical magnetic field @xmath309 corresponding to the standard superconducting phase can be approximated with a high accuracy by the formula @xmath310 where @xmath311 and @xmath312 is given by ( [ 6.13 ] ) .
it is convenient to introduce the `` reduced '' helmholtz free energy difference @xmath313 defined by @xmath314 which has the following expression @xmath315\end{aligned}\ ] ] for the novel superconducting phase and @xmath316 for the standard bcs result .
the novel superconducting phase is preferable over the standard phase provided that the following inequality is satisfied : @xmath317 the last inequality is always satisfied if @xmath318 , i.e. , if the material parameters @xmath319 and the thickness of the layers @xmath221 satisfy the following condition @xmath320 where @xmath169 and @xmath321 should be expressed in terms of material parameters as given by ( [ 5.12])-([5.14 ] ) and ( [ 4.23 ] ) respectively .
since we have the same hamiltonian responsible for both the novel superconducting phase and for the standard bcs solution in two inequivalent representations of the anticommutator ring of electron field operators we may assume that the values of the coupling constant @xmath110 and the density of electron states @xmath322 at the fermi surface are the same as those experimentally found for lts @xcite . therefore , the values @xmath323 @xcite . for the reasonably chosen values , @xmath324 k and @xmath325 the numerical results of temperature dependence of the reduced helmholtz free energies @xmath313 are shown on the fig .
[ fig6 ] .
[ fig6 ] clearly shows that the novel superconducting phase associated with the third inequivalent representation of the anticommutator ring of electron field operators of the same bcs hamiltonian is energetically preferable with respect to its standard superconducting phase .
in applications of quantum field theory of many particle systems one has intuitively believed that a given hamiltonian @xmath12 determines uniquely a single gibbs state of a system at given values of thermodynamic variables like temperature @xmath22 , volume @xmath23 and the number @xmath24 of particles .
such a belief would be completely correct provided that the canonical commutation or anticommutation relations of field operators had a unique representation . in the thermodynamic limit @xmath151 , @xmath32 , @xmath326 , one deals , in fact , with systems having the infinite number of degrees of freedom . in this case , as it has been pointed out by haag @xcite , the canonical commutation or anticommutation relations of field operators have no longer unique solutions , i.e. , they have several different inequivalent representations .
this fact has to be taken into account in all discussions concerning the quantum field theory of many particle systems .
namely , with each inequivalent representation of commutation or anticommutation relations one has to associate the corresponding `` matrix form '' of the hamiltonian @xmath12 leading to the corresponding grand canonical partition function @xmath14 and the grand canonical potential @xmath20 which fully specify the gibbs state of the system .
this argument indicates the fact that the system can have as many gibbs states as many inequivalent representations of the canonical commutator or anticommutator ring exist for the given hamiltonian @xmath12 . to our best knowledge
there is no systematic method for a classification of all inequivalent representations of the canonical commutator or anticommutator ring of field operators entering a given hamiltonian @xmath12 , e.g. , as the method one has for the classification of the irreducible representations of lie algebras .
the only known way how one selects an inequivalent representation of the canonical commutator or anticommutator ring of field operators entering a given hamiltonian @xmath12 is to select a suitable unperturbed hamiltonian @xmath81 which can be exactly diagonalized . in the present work we have explicitly constructed three inequivalent representations of the canonical anticommutator ring of electron field operators entering the hamiltonian of the bcs theory of superconductivity .
each inequivalent representation has been constructed by selecting the unperturbed hamiltonian @xmath100 , @xmath101 or @xmath180 .
each inequivalent representation specifies the corresponding gibbs state of the system .
the physical system , however , stabilizes in a gibbs state having the minimal helmholtz free energy at given values of the thermodynamic variables like temperature @xmath22 , volume @xmath23 and the particle number @xmath24 . from this reason it is important to search for all relevant inequivalent representations of the canonical anticommutator ring of field operators entering the hamiltonian of the system .
the normal state and the standard superconducting state of the bcs theory corresponding to two inequivalent representations have been generally known long time ago .
we have constructed the third inequivalent representation which has not been known till now .
this representation can , perhaps , be relevant for describing the superconducting properties of hts .
the authors are very grateful for many elucidating discussions with professors c. cronstrm , i. huba and r. hlubina .
r. haag , on quantum field theories , _ dan .
* 29 * , no 12 ( 1955 ) .
j. bardeen , l. n. cooper and j. r. schrieffer , _ phys .
rev . _ * 108 * , 1175 ( 1957 ) .
n. n. bogoliubov , _ physics _ * 26 * , s1 ( 1960 ) .
p. kalinay , m. noga , _ czech .
j. phys . _
* 48 * , 1615 ( 1998 ) .
n. n. bogoliubov , _ sov .
- jetf _ * 41 * , ( 1958 ) ; + j. g. valatin , _
nuovo cimento _ * 7 * , 843 ( 1958 ) . see e.g. a. l. fetter and j. d. walecka , quantum theory of many - particle systems , mcgraw - hill inc .
new york , 1971 , p. 396 .
c. cronstrm and m. noga , _ czech .
j. phys . _ * 51 * , 175 ( 2001 ) .
y. uemura et al , _ phys .
_ * 62 * , 2317 ( 1989 ) ; * 66 * , 2665 ( 1991 ) .
f. pistolesi and g. c. strinati , _ phys .
b _ * 49 * , 6356 ( 1994 )
. a. s. alexandrov et al , _ phys .
lett . _ * 76 * , 983 ( 1996 ) .
a. junod et al , _ physica b _ * 149 - 196 * , 1497 ( 1994 ) ; + a. mirmelstein et al , _ physica c _ * 248 * , 335 ( 1995 ) ; + c. meingast et al , _ physica c _ * 272 * , 106 ( 1996 ) .
a. junod et al , _ physica c _ * 229 * , 209 ( 1994 ) .
see e.g. superconductivity , marcel dekker inc .
new york , 1969 , vol . 1 , p. 122 . | [ abstr ] hamiltonian of a system in quantum field theory can give rise to infinitely many partition functions which correspond to infinitely many inequivalent representations of the canonical commutator or anticommutator rings of field operators .
this implies that the system can theoretically exist in infinitely many gibbs states .
the system resides in the gibbs state which corresponds to its minimal helmholtz free energy at a given range of the thermodynamic variables .
individual inequivalent representations are associated with different thermodynamic phases of the system .
the bcs hamiltonian of superconductivity is chosen to be an explicit example for the demonstration of the important role of inequivalent representations in practical applications .
its analysis from the inequivalent representations point of view has led to a recognition of a novel type of the superconducting phase transition .
pacs numbers : 03.70.+k , 05.30.-d , 11.10.-z , 74.20.fg , 74.25.bt , 74.78.bz |
the painlev first equation @xcite @xmath4 is the simplest of the six classical equations of painlev
gambier @xcite and can be derived from any other painlev equation using certain scaling reductions @xcite .
the recent interest to this equation is due to its significant role in various physical models .
for instance , equation @xmath5 describes certain solutions to kdv and bussinesq equations @xcite , bifurcations in some non - integrable nonlinear models @xcite , continuous limits in matrix models of quantum gravity @xcite ; @xmath0-function associated with @xmath5 appears in @xmath6-large asymptotics of semi - classical orthogonal and bi - orthogonal polynomials @xcite and thus becomes a primary object in the problem of laplacian growth @xcite . in context of the string theory ,
physical " solutions of @xmath5 are distinguished from non - physical " ones by the _ monotonic _
asymptotic behavior as @xmath7 @xcite .
there are two kinds of such monotonic boundary conditions , i.e. @xmath8 . using elementary perturbation analysis , the solution @xmath9 is unique as being a background to a 2-parametric family of oscillating solutions .
solutions approaching a positive branch of the square root as @xmath7 , i.e. @xmath10 , form a @xmath11-parametric family parameterized by an amplitude of an exponentially small perturbation to a power - like dominant solution .
these solutions have the asymptotic expansion @xmath12 , whose coefficients @xmath13 , admit a combinatorial interpretation @xcite . in the problem of laplacian growth without surface tension ( hele - shaw problem , quantum hall effect etc .
) , the shape of a growing droplet is described using certain @xmath0-function , see @xcite .
if the droplet develops a cusp singularity , this @xmath0-function can be approximated by a @xmath0-function associated to the first painlev transcendent , @xcite .
physical " asymptotic conditions imposed on this @xmath0-function determine the relevant stokes multipliers @xmath14 . in turn , these @xmath14 pinch out the monotonic as @xmath7 painlev function @xmath10 .
equation @xmath5 has unexpectedly rich asymptotic properties in the complex @xmath15-plane .
boutroux @xcite has shown that , generically , asymptotics of the painlev first transcendent as @xmath3 is described by the modulated weierstra elliptic function whose module is a transcendent function of @xmath16 .
furthermore , the module function is such that the elliptic asymptotic ansatz degenerates along the directions @xmath17 , @xmath18 .
p. boutroux called the corresponding trigonometric expansions _ `` truncated '' solutions_. their @xmath11- and @xmath19-parameter reductions ,
if they admit analytic continuation into one or two neighboring sectors of the complex @xmath15-plane , were called by boutroux _
`` bi - truncated '' _ and _ `` tri - truncated '' solutions_. all bi- and tri - truncated solutions have the algebraic leading order behavior , @xmath20 , perturbed by exponential terms .
we call a discontinuity in the asymptotic form of an analytic function the _ stokes phenomenon_. in the case of @xmath5 , a jump in the phase shift of a modulated elliptic ansatz across the rays @xmath17 is called the _ nonlinear _ stokes phenomenon . for bi- and tri - truncated solutions , a jump in the exponentially small perturbation of a dominant solution resembles the well known stokes phenomenon in the linear theory and thus is called the _ quasi - linear _ stokes phenomenon . in @xcite , equation @xmath5 was studied further using classical tools like the perturbation approach and the method of nonlinear integral equations . mainly , these articles discuss the behavior of the painlev transcendents on the real line .
the recent paper @xcite adopts the same approach carefully studying the behavior of the tri - truncated solution on the negative part of the real line . in @xcite ,
the multiple scale analysis was applied to @xmath5 ( and @xmath21 ) to find a precise form of the phase shift in the elliptic asymptotic ansatz within complex sectors between the indicated rays . in @xcite
, the witham averaging method was used to describe the elliptic tail of the monotonic at @xmath22 solution of @xmath5 .
the isomonodromy deformation approach to painlev equations , see @xcite , was applied to equation @xmath5 in @xcite . in this way , the asymptotics of the painlev functions is expressed in terms of the stokes multipliers of an associated linear system .
then the equation of a monodromy surface yields connection formulas for the asymptotic parameters along different directions of the complex @xmath15-plane .
a complete description of the nonlinear stokes phenomenon in @xmath5 is given in @xcite .
a heuristic description of the quasi - linear stokes phenomenon in @xmath5 can be found in @xcite .
using the borel transform technique and some assumptions on the analytic properties of the relevant borel transforms , as well as the isomonodromy deformation approach based on the so - called exact wkb analysis , y. takei @xcite has re - derived the latter result ( look for more discussion in @xcite ) . in the present paper
, we construct the @xmath0-function associated with the monotonic as @xmath3 solution of @xmath5 and rigorously describe the relevant quasi - linear stokes phenomenon .
our main tool is the riemann - hilbert ( rh ) problem .
we observe that the jump graph for our rh problem can be decomposed into a disjoint union of two branches , one of which is responsible for the background @xmath23 while another one produces the exponentially small perturbation of the dominant solution ( look @xcite for similar observation in @xmath21 case ) . using the steepest descent approach of deift and zhou @xcite , we prove the unique solubility of this problem and compute the asymptotics of the painlev transcendent . applying a rotational symmetry ,
we prove the existence of five solutions , @xmath24 , @xmath18 , asymptotic to @xmath25 as @xmath3 in the respective overlapping sectors @xmath26 , see ( [ y4n ] ) , ( [ y4n_ext ] ) , and find the exponentially small differences @xmath27 , @xmath18 , see ( [ yk_hat_yk ] ) .
the latter constitute the quasi - linear stokes phenomenon .
a collection of the functions @xmath24 , @xmath18 , forms a piece - wise meromorphic function @xmath28 as @xmath3 .
the moments of this function immediately yield the asymptotics as @xmath29 for the coefficients @xmath13 ( [ an_fin ] ) of the @xmath15-series expansion to the dominant solution ( [ y+formal ] ) .
for the first time , the formula for the coefficient asymptotics was found in @xcite .
the authors of @xcite studied the recurrence relations for the same coefficients by direct means and prove a similar asymptotic formula modulo a common factor ( an advanced version of the direct approach to a generalized recurrence relation which contains one for @xmath5 as a special case can be found in @xcite ) .
the exact value of this common factor was announced in @xcite with the reference to the method of @xcite based on the borel transform formula .
in contrast , we do not use the borel transform technique at any stage of our investigation .
the paper is organized as follows . in section [ rh_p1 ] ,
we recall the lax pair for @xmath5 , formulate the relevant rh problem and solve it asymptotically . using the approximate @xmath0-function
, we find the asymptotics of the bi- and tri - truncated painlev transcendents and of the relevant hamiltonian functions . in section [ as_coeffs ] , we find the coefficient asymptotics in the power - like expansion to the formal solution of @xmath5 .
introduce generators of @xmath30 , @xmath31 , @xmath32 , @xmath33 and the pauli matrices @xmath34 and @xmath35 together and consider the system of matrix differential equations for @xmath0 , see @xcite , [ lax_pair ] @xmath36 compatibility of ( [ a ] ) and ( [ u ] ) implies that the coefficients @xmath37 and @xmath38 depend on the deformation parameter @xmath15 in accord with the nonlinear differential system @xmath39 which is equivalent to the classical painlev first equation [ p1 ] .
following @xcite , see also @xcite , linear equation ( [ a ] ) has the only one ( irregular ) singular point at infinity , and there exist solutions @xmath40 of ( [ lax_pair ] ) with the asymptotics @xmath41 as @xmath42 solutions @xmath40 , @xmath43 , are called the _
canonical _ solutions , while sectors @xmath44 are called the _
canonical _ sectors .
canonical solutions @xmath40 are uniquely determined by ( [ psi_k])([omega_k ] ) and solve both equations ( [ lax_pair ] ) .
they differ from each other in constant right matrix multipliers @xmath45 called the _ stokes matrices _ , @xmath46 observing that all solutions of ( [ a ] ) are entire functions , thus @xmath47 and using the relation @xmath48 which follows from the definition of the canonical solutions and the asymptotics ( [ psi_k ] ) , ( [ omega_k ] ) , we readily find the constraints for the stokes matrices @xcite , @xmath49 or , in the scalar form , @xmath50 thus , generically , two of the stokes multipliers @xmath14 , @xmath51 , determine all others .
the inverse monodromy problem consists of reconstruction of @xmath40 using the known values of the stokes multipliers @xmath14 .
it can be equivalently formulated as a riemann - hilbert ( rh ) problem . with this aim ,
introduce the union of rays @xmath52 , where @xmath53 , @xmath54 , and @xmath55 , all oriented toward infinity .
denote the sectors between the rays @xmath56 and @xmath57 by @xmath58 , between @xmath59 and @xmath60 , @xmath61 , by @xmath62 , and between @xmath63 and @xmath56 by @xmath64 .
all the sectors @xmath62 are in one - to - one correspondence with the canonical sectors @xmath44 ( [ omega_k ] ) , see figure [ f1 ] .
let each of the sectors @xmath62 , @xmath65 , be a domain for a holomorphic @xmath66 matrix function @xmath40 .
denote the collection of @xmath40 by @xmath67 , @xmath68 let @xmath69 and @xmath70 be the limits of @xmath67 on @xmath71 to the left and to the right , respectively .
let @xmath72 be defined on the complex @xmath73-plane cut along the negative part of the real axis .
the rh problem we talk about is the following one : 1 .
find a piece - wise holomorphic @xmath66 matrix function @xmath67 such that @xmath74 2 . on the contour @xmath71 ,
the jump condition holds @xmath75 where the piece - wise constant matrix @xmath76 is given by equations : + [ jump_matrices ] @xmath77 + with the constants @xmath14 satisfying the constraints ( [ scalar_relations ] ) ; because @xmath67 satisfies the asymptotic condition @xmath78 where @xmath79 the solution @xmath80 of the painlev equation can be found from the residue " of @xmath81 at infinity , @xmath82 [ rem1 ] it is easy to see that @xmath83 is nothing but the hamiltonian for the painlev first equation with the canonical variables @xmath84 and @xmath85 . equation ( [ y_from_y ] ) specifies the painlev transcendent as a function @xmath86 of the deformation parameter @xmath15 and of the stokes multipliers @xmath14 . using the solution @xmath86 and the symmetries of the stokes multipliers described in @xcite , we obtain further solutions of [ p1 ] : [ p_symmetries ] @xmath87 where the bar means the complex conjugation . by technical reason , to find the asymptotics of @xmath80 , we use below not @xmath81 but related auxiliary functions @xmath88 and @xmath89 with expansions ( [ chi_expansion ] ) and ( [ x_expansion ] ) , respectively .
the latter involve differences @xmath90 , @xmath91 are known , which can be estimated using singular integral equations with contracting operators .
let us consider the rh problem above where @xmath92 assuming that @xmath3 within the sector @xmath93 $ ] .
equations ( [ p23 ] ) imply that @xmath67 has no jump across the ray @xmath94
. the constraints ( [ scalar_relations ] ) reduce to the following system of equations , @xmath95 our first step in the rh problem analysis consists of introduction an auxiliary @xmath96-function , @xmath97 defined on the complex @xmath73-plane cut along the ray @xmath98 $ ] .
the asymptotics of @xmath96-function as @xmath99 coincides with the canonical one , @xmath100 let us formulate an equivalent rh problem for the piece - wise holomorphic function @xmath101 , @xmath102 @xmath103 if @xmath104 then @xmath105 .
our next goal is to transform the jump contour @xmath71 to the contour of the steepest descent for the matrix @xmath106 .
we denote by @xmath107 the level line @xmath108 passing through the stationary phase point @xmath109 and asymptotic to the rays @xmath110 .
this is the steepest descent line for @xmath111 .
let @xmath112 be the union of the level lines @xmath113 , @xmath114 , @xmath115 , and @xmath116 , @xmath117 , all emanating from the critical point @xmath118 . among them ,
the level line @xmath119 approaching the ray @xmath120 ( if @xmath121 , the level line @xmath119 is the segment @xmath122 $ ] ) is the steepest descent line for @xmath111 , while the level lines @xmath123 and @xmath124 approaching the rays @xmath125 and @xmath126 , respectively , are the steepest descent lines for @xmath127 .
the level line @xmath116 , @xmath117 , approaches the ray @xmath128 .
since the stokes matrix @xmath129 can be factorized , @xmath130 it is convenient to consider the following equivalent rh problems for @xmath67 : for @xmath131 $ ] , the jump contour is the union of @xmath107 oriented from up to down and @xmath132 whose components are oriented toward infinity , see figure [ f2 ] .
the jump matrices are as follows : @xmath133 [ rem2 ] the jump contour for the rh problem ( [ + jumps ] ) is decomposed into the disjoint union of the line @xmath107 and the graph @xmath132 , see figure [ f2 ] . for the boundary value @xmath134 , the level line @xmath119 emanating from @xmath118 passes through the stationary phase point @xmath135 and partially merges with the upper half of the level line @xmath107 . to construct the rh problem however , it is not necessary to transform the jump contour precisely to the steepest descent graph .
it is enough to ensure that the jump matrices approach the unit matrix uniformly with respect to @xmath73 and fast enough as @xmath136 or @xmath99 .
as @xmath131 $ ] , introduce the reduced rh problem ( @xmath137 ) for the piece - wise holomorphic function @xmath138 discontinuous across @xmath132 only : @xmath139 the jump matrix @xmath76 here is defined in ( [ + jumps ] ) .
[ theorem1 ] if @xmath131 $ ] and @xmath140 is large enough , then there exists a unique solution of the rh problem ( [ phi_rh ] ) .
the painlev function @xmath141 corresponding to @xmath137 has the asymptotics @xmath142 as @xmath136 in the above sector .
uniqueness . since @xmath143
, we have @xmath144 , and hence @xmath145 is an entire function . furthermore , because of normalization of @xmath138 at infinity , @xmath146 .
let @xmath147 and @xmath148 be two solutions of ( [ phi_rh ] ) . taking into account the cyclic relation in ( [ stokes_relations ] ) which implies the continuity of the rh problem for @xmath138 at @xmath118 , the
ratio " @xmath149 is an entire function of @xmath73 . using the liouville theorem and normalization of @xmath148 and @xmath147 at infinity , we find @xmath150 , i.e. @xmath151 . existence .
introduce an auxiliary function @xmath152 where the prime means differentiation w.r.t .
@xmath37 and @xmath153 with @xmath154 standing for the classical airy function which can be defined using the taylor expansion @xcite , @xmath155 asymptotics at infinity of this function and its derivative are as follows , @xmath156 it is worth to note that the function @xmath157 satisfies the linear differential equation @xmath158 using the properties of the airy functions , we find that the products @xmath159 @xmath160 , have the asymptotic expansion @xmath161 as @xmath162 , @xmath163 , where @xmath164 let @xmath165 be the union of the rays @xmath166 , @xmath114 , and @xmath167 all oriented toward infinity .
this graph divides the complex @xmath37-plane into four regions : @xmath168 which is the sector between @xmath169 and @xmath170 , the sectors @xmath171 , @xmath172 , between the rays @xmath173 and @xmath174 , and the sector @xmath175 between the rays @xmath176 and @xmath169 .
define a piece - wise holomorphic function @xmath177 , @xmath178 by construction , this function solves the following rh problem , see figure [ f3 ] : @xmath179 therefore the function @xmath177 has precisely the jump properties of the function @xmath138 . to find @xmath138 with the correct asymptotic behavior at infinity ,
let us use the mapping @xmath180 within the disk @xmath181 , the mapping ( [ z_lambda ] ) yields a holomorphic change of the independent variable .
introduce a piece - wise holomorphic function @xmath182 , @xmath183 where @xmath184 is defined on the plane cut along the level line @xmath116 asymptotic to the ray @xmath128 .
note that @xmath185 is holomorphic in the interior of the above disk @xmath186 and thus does not affect the jump properties of @xmath187 .
we look for the solution of the rh problem ( [ phi_rh ] ) in the form of the product @xmath188 consider the rh problem for the correction function @xmath88 . by construction
, it is a piece - wise holomorphic function discontinuous across the clockwise oriented circle @xmath189 of the radius @xmath190 centered at @xmath191 and across the part of @xmath132 located outside the above circle ( in fact , @xmath88 is continuous across @xmath116 , see ( [ chi_rh ] ) below ) .
the latter is divided by @xmath132 in four arcs : @xmath192 between @xmath116 and @xmath123 , @xmath193 , @xmath172 , between @xmath194 and @xmath195 , and @xmath196 between @xmath124 and @xmath116 , see figure [ f4 ] .
to simplify our notations , let us put @xmath197 then the rh problem for @xmath88 is as follows : @xmath198 taking into account the equations ( [ hat_jumps ] ) , it is easy to check the continuity of the rh problem at the node points . observing that ,
on the circle @xmath199 , @xmath200 , we have @xmath201 is large , we immediately see that @xmath202 where the precise value of the positive constant @xmath203 is not important for us . taking into account
that , by the above reason , on the circle @xmath204 , we may use for @xmath205 its asymptotics ( [ hat_phi_k_as ] ) , the jump matrix @xmath206 has the asymptotic expansion @xmath207 where the precise value of the positive constants @xmath208 is not important for us .
now , the solubility of the rh problem ( [ chi_rh ] ) and therefore of ( [ phi_rh ] ) for large enough @xmath140 is straightforward . indeed , consider the equivalent system of the non - homogeneous singular integral equations for the limiting value @xmath209 , i.e.@xmath210 or , in the symbolic form , @xmath211 . here
@xmath212 means the left limit of @xmath73 on @xmath213 ( recall , that the circle @xmath204 is clock - wise oriented ) , and @xmath214 is the composition of the operator of the right multiplication in @xmath215 and of the cauchy operator @xmath216 .
an equivalent singular integral equation for @xmath217 differs from ( [ chi_sing ] ) in the inhomogeneous term only , @xmath218 consider the integral equation ( [ psi_sing ] ) in the space @xmath219 . since @xmath220 is small in @xmath219 for large enough @xmath140 , and
@xmath216 is bounded in @xmath219 , then @xmath221 with some positive constant @xmath203 , thus @xmath214 is contracting and @xmath222 is invertible in @xmath219 for large enough @xmath140 . because @xmath223 , equation ( [ psi_sing ] ) for @xmath224 is solvable in @xmath219 , and the solution @xmath88 of the rh problem ( [ chi_rh ] ) is determined by @xmath225 using the equation @xmath226 .
let us find the asymptotics of the painlev function .
using ( [ y_expansion ] ) and definition of @xmath182 ( [ tilde_phi ] ) , the asymptotics of @xmath88 as @xmath99 in terms of @xmath38 and @xmath83 is as follows , @xmath227 on the other hand , in accord with the said above , the function @xmath228 is given by the converging iterative series , @xmath229 . to compute the term @xmath230
, we observe that the contribution of the infinite branches @xmath195 is exponentially small in @xmath15 due to estimate ( [ hk_estim ] ) .
using the expansion ( [ h_on_circle ] ) , we reduce the evaluation of the integral along the circle @xmath204 to the residue theorem . omitting this elementary computation , we present the final result : for large enough @xmath140 , @xmath131 $ ]
, the asymptotics of @xmath88 as @xmath99 is given by @xmath231 comparing entries @xmath232 in ( [ chi_expansion ] ) and ( [ chi_as ] ) , we see that the hamiltonian function @xmath233 corresponding to the stokes multipliers @xmath137 is given by @xmath234 next comparing entries @xmath235 in ( [ chi_expansion ] ) and ( [ chi_as ] ) and using ( [ ham0_as ] ) , we find the asymptotics of the painlev function @xmath236 , @xmath237 recall that @xmath238 where the main branch of the root is taken .
let us go to the case of the nontrivial @xmath239 described by the rh problem ( [ + jumps ] ) .
we look for the solution @xmath67 in the form of the product @xmath240 where @xmath138 is the solution of the reduced rh problem ( [ phi_rh ] ) and @xmath241 ( [ ham0_as ] ) is the hamiltonian function ( [ ham_def ] ) corresponding to the painlev transcendent @xmath141 ( [ y0_as ] ) . using ( [ y_expansion ] ) , we find the asymptotics of @xmath89 as @xmath99 , @xmath242 thus we arrive at the rh problem for the correction function @xmath89 on the steepest descent line @xmath107 , @xmath243 note , @xmath138 is continuous across @xmath107 and therefore holomorphic in some neighborhood of @xmath107 as @xmath93 $ ] , @xmath140 is large enough .
the jump matrix on @xmath107 can be estimated as follows , @xmath244 here @xmath203 is some positive constant whose precise value is not important for us , and @xmath245 is the stationary phase point for @xmath246 , see ( [ g_function ] ) .
estimate ( [ x_h_estim ] ) yields the estimate for the norm of the singular integral operator @xmath247 in the equivalent system of singular integral equations , @xmath248 , @xmath249 if @xmath140 is large enough and @xmath250 $ ] , @xmath251 , then the operator @xmath247 is contracting and the system @xmath248 is solvable by iterations in @xmath252 , i.e. @xmath253 . however , to incorporate the oscillating direction @xmath254 in the general scheme , we use some more refined procedure .
[ theorem2 ] if @xmath92 , @xmath131 $ ] and @xmath140 is large enough , then there exists a unique solution of the rh problem ( [ p28])([jump_matrices ] ) .
the corresponding painlev function has the asymptotics @xmath255 where @xmath256 is the solution of the painlev equation for @xmath137 , @xmath257 .
it is enough to prove the solubility of the rh problem ( [ psi_x_rh ] ) . using for @xmath138 the expressions ( [ chi_def ] ) with ( [ tilde_phi ] ) and the estimate ( [ chi_as ] ) together
, we find the asymptotics of the jump matrix @xmath258 , @xmath259 consider the following model rh problem , @xmath260 this problem is solvable by the following quadrature , @xmath261 we look for the solution @xmath89 of the rh problem ( [ psi_x_rh ] ) in the form of the product , @xmath262 the correction function @xmath263 satisfies the rh problem @xmath264 using ( [ h_as])([p_explicit ] ) , we find the estimate for the jump matrix @xmath265 on @xmath107 , @xmath266 our next steps are similar to presented in the proof of theorem [ theorem1 ] .
consider the system of the singular integral equations for @xmath267 equivalent to the rh problem ( [ q_rh ] ) , @xmath268 . here
the singular integral operator @xmath269 is the superposition of the multiplication operator in @xmath270 and of the cauchy operator @xmath216 . because the cauchy operator is bounded in @xmath252 , the singular integral operator @xmath269 for large enough @xmath140 , @xmath131 $ ] , satisfies the estimate @xmath271 with some positive constant @xmath203 whose precise value is not important for us .
thus equation @xmath272 for the difference @xmath273 is solvable by iterations in the space @xmath252 for large enough @xmath140 .
solution of the rh problem ( [ q_rh ] ) is given by the integral @xmath274 .
this implies the asymptotics of @xmath263 as @xmath99 , @xmath275 now let us find the asymptotics of the painlev function @xmath80 . using ( [ x = qp ] ) , ( [ p_explicit ] ) and the estimate ( [ q_as ] ) , we find @xmath276 comparing ( [ x_as ] ) and ( [ x_expansion ] ) , we conclude that the hamiltonian function for @xmath92 is as follows , @xmath277 while the painlev function is given by @xmath278 where @xmath279 and @xmath141 are the hamiltonian and the painlev functions , respectively , corresponding to @xmath137 . applying the symmetry ( [ p_symmetries_bar ] ) to the solution ( [ y+_as ] ) and changing the argument of @xmath15 in @xmath280 , we obtain [ theorem3 ] if @xmath92 and @xmath3 , @xmath281 $ ] , then the asymptotics of the painlev first transcendent is given by @xmath282 where @xmath283 is the solution of the painlev equation for @xmath284 , @xmath285 .
the solutions @xmath141 and @xmath286 are meromorphic functions of @xmath287 and thus can be continued beyond the sectors indicated in theorems [ theorem2 ] and [ theorem3 ] . to find the asymptotics of @xmath288 in the interior of the sector
@xmath131 $ ] , we apply ( [ y+_as ] ) .
similarly , we find the asymptotics of the solution @xmath141 in the interior of the sector @xmath281 $ ] using ( [ y-_as ] ) .
either expression implies [ corollary1 ] if @xmath3 , @xmath289 $ ] , then @xmath290 applying symmetries ( [ p_symmetries ] ) to @xmath291 , @xmath292 , we find the solutions @xmath291 corresponding to the stokes multipliers @xmath293 , @xmath294 since there is one - to - one correspondence between the points of the monodromy surface and the painlev functions , the first of the identities ( [ scalar_relations ] ) , @xmath295 , implies that @xmath296 . using theorems [ theorem2 ] and [ theorem3 ]
, we find that @xmath297 , \notag \\\label{y4n-2 } y_{4n-2}(x)=y_{4n+3}(x)&=-\sqrt{e^{-i\pi}\tfrac{x}{6 } } + { \cal o}(x^{-2 } ) , \\ & \quad |x|\to\infty,\quad \arg x\in [ \tfrac{3\pi}{5}-\tfrac{4\pi}{5}n,\tfrac{7\pi}{5}-\tfrac{4\pi}{5}n ] .
\notag\end{aligned}\ ] ] the symmetry ( [ p_symmetries ] ) with the definition ( [ yk_def ] ) applied to ( [ y0_hat_y0 ] ) yields [ corollary2 ] if @xmath3 and @xmath298 $ ] , then @xmath299 on the one hand , equations ( [ y0_hat_y0 ] ) , ( [ yk_hat_yk ] ) constitute the quasi - linear stokes phenomenon for the painlev first equation . on the other hand ,
these equations give the asymptotic description of the degenerate painlev functions beyond the sectors in ( [ y4n ] ) and ( [ y4n-2 ] ) . observing that the difference ( [ yk_hat_yk ] ) is exponentially small in the interior of the indicated sector ,
we conclude that the asymptotics ( [ y4n ] ) and ( [ y4n-2 ] ) as @xmath3 continue to wider open sectors , @xmath300 where @xmath251 is an arbitrary small constant . [ rem3 ] the solutions @xmath301 ( [ yk_def ] ) corresponding to the trivial values of two stokes multipliers @xmath302 are the most degenerate among the painlev transcendent since they behave algebraically in four of five sectors @xmath303 , @xmath304 , see ( [ y4n_ext ] ) , ( [ y4n-2_ext ] ) .
nevertheless these solutions are transcendent , since their asymptotics as @xmath3 within the remaining fifth sector involves the elliptic function of weierstra , look for more details in @xcite .
moreover , the fact that the asymptotics of @xmath301 is not elliptic in four sectors uniquely determines the values of all the stokes multipliers @xmath14 .
thus the asymptotics ( [ y4n_ext ] ) , ( [ y4n-2_ext ] ) uniquely determine the degenerate solutions @xmath301 .
[ rem4 ] the asymptotics of less degenerate solutions corresponding to @xmath305 and @xmath306 can be found applying the symmetries ( [ p_symmetries_rotate ] ) to equations ( [ y+_as ] ) and ( [ y-_as ] ) .
using the steepest descent approach , cf .
@xcite , we can show the existence of the asymptotic expansion of @xmath301 , @xmath307 , in the negative degrees of @xmath308 .
further elementary investigation of the recursion relation for the coefficients of the series allows us to claim that the asymptotic expansion for @xmath301 in ( [ y4n_ext ] ) , ( [ y4n-2 ] ) has the following form : @xmath309 where coefficients @xmath13 are determined uniquely by the recurrence relation @xmath310 several initial terms of the expansion are given by @xmath311 our next goal is to determine the asymptotics of the coefficients @xmath13 in ( [ y+formal ] ) as @xmath29 . with this purpose ,
let us construct a sectorial analytic function @xmath312 , @xmath313\colon \quad \hat y(t)=y_{4n}(e^{i\pi}t^2),\quad n=-2,-1,0,1,2.\ ] ] the function @xmath312 has a finite number of poles all contained in a circle @xmath314 and is characterized by the uniform asymptotic expansion near infinity , @xmath315 let @xmath316 be a partial sum @xmath317 and @xmath318 be a product @xmath319 because @xmath320 is polynomial , the integral of @xmath318 along the counter - clock - wise oriented circle of the radius @xmath321 satisfies the estimate @xmath322 with some positive constant @xmath323 whose precise value is not important for us .
on the other hand , inflating the sectorial arcs of the circle @xmath321 , we find that @xmath324 because @xmath325 , the first of the integrals in the r.h.s . of ( [ vn_inflation ] )
is computed as follows : @xmath326 remaining integrals in ( [ vn_inflation ] ) are computed using definitions ( [ hat_y])([vn ] ) and ( [ yk_def ] ) with the identification @xmath327 and the formula ( [ y0_hat_y0 ] ) together , @xmath328 thus , letting @xmath329 , we find the asymptotics of the coefficient @xmath330 in ( [ y+formal ] ) as @xmath331 , @xmath332 [ rem5 ] the presented asymptotic formula shows a remarkable accuracy : neglecting in ( [ an_fin ] ) error terms , we find an approximation to @xmath330 with the relative error not exceeding 2% for @xmath333 and 1% for @xmath334 .
furthermore , for the initial set of @xmath335 , the relative error decreases approximately as @xmath336 which is significantly better than estimated .
r. haberman , slowly varying jump and transition phenomena associated with algebraic bifurcation problems , _
j. appl .
_ * 51 * ( 1979 ) 69106 ; slow passage through the nonhyperbolic homoclinic orbit associated with a subcritical pitchfork bifurcation for hamiltonian systems and the change in action , _
siam j. appl .
_ * 62 * ( 2001 ) no . 2 , 488 - 513 . a.s .
fokas , a.r .
its , a.v .
kitaev , discrete painlev equations and their appearance in quantum gravity , _ comm .
math.phys._ * 142 * ( 1991 ) , no .
2 , 313344 ; the isomonodromy approach to matrix models in @xmath337d quantum gravity , _ comm . math .
* 147 * ( 1992 ) , no .
2 , 395430 . a.s .
fokas , a.r .
its and a.v .
kitaev , matrix models of two - dimensional quantum gravity , and isomonodromic solutions of painlev `` discrete equations '' , _ zap .
nauchn . sem .
lomi _ * 187 * ( 1991 ) 330 ; translation in : _ j. math .
_ * 73 * ( 1995 ) , no .
4 , 415429 .
p. boutroux , recherches sur les transcendantes de m. painlev et letude asymptotique des quations diffrentielles du second ordre _ ann .
norm.supr._ * 30 * ( 1913 ) 255376 ; _ ann .
ecol . norm.supr._
* 31 * ( 1914 ) 99159 .
r. garnier , sur les quations diffrentielles du troisime ordre do nt lintgrale gnrale est uniforme et sur une classe dquations nouvelles dordre suprieur do nt lintgrale gnrale a ses points critique fixes , _ ann .
sci . cole norm .
_ ( 4 ) * 29 * ( 1912 ) , 1126 . | using the riemann - hilbert approach , the @xmath0-function corresponding to the solution of the first painlev equation @xmath1 with the asymptotic behavior @xmath2 as @xmath3 is constructed . the exponentially small jump in the dominant solution and the coefficient asymptotics in the power - like expansion to the latter
are found . |
the goal of this paper is to construct solutions of a scalar , fractional ginzburg - landau ( or allen - cahn ) equation in a periodic medium , whose interface stays in a prescribed slab and whose energy is minimal among compact perturbations .
the simplest case that we have in mind is the non - local equation @xmath0 in which @xmath1 is a fractional parameter and @xmath2 is a smooth function , bounded and bounded away from zero , and such that @xmath3 the operator @xmath4 in is a fractional power of the laplacian , see e.g. @xcite for an introduction to this topic . in the framework of equation , the solution @xmath5 $ ] represents a state parameter in a model of phase coexistence ( the two `` pure phases '' being represented by @xmath6 and @xmath7 ) .
the presence of a fractional exponent @xmath1 is motivated by models which try to take into account long - range particle interactions ( as a matter of fact , these models may produce either a local or non - local tension effect , depending on the value of @xmath8 , see @xcite ) .
we also recall that equations of this type naturally occur in other areas of applied mathematics , such as the peierls - nabarro model for crystal dislocations when @xmath9 , and for generalizations of this model when @xmath1 ( see e.g. @xcite ) .
related problems also arise in models for diffusion of biological species ( see e.g. @xcite ) .
the periodicity condition in takes into account a possible geometric ( or crystalline ) structure of the medium in which the phase transition takes place .
the level sets of the solution @xmath10 have particular physical importance , since they correspond , at a large scale , to the interface between the two phases of system .
the question that we address in this paper is then to find solutions of whose level sets lie in any given strip of universal size .
the direction of this strip will be arbitrary and the size of the strip is bounded independently on the direction .
in addition to this geometric constraint on the level sets of the solution , we will also prescribe an energy condition .
namely , equation is variational . though the associated energy functional diverges ( i.e. nontrivial solutions have infinite total energy in the whole of the space ) , it is possible to `` localize '' the non - local energy density in any fixed domain of interest and require that the solution has a minimal property with respect to any perturbation supported in this domain .
the existence of minimal solutions of phase transition equations whose level sets are confined in a strip goes back to @xcite , where equation was taken into account for @xmath11 and it is strictly related to the construction , performed in @xcite , of minimal surfaces which stay at a bounded distance from a plane ( see also @xcite ) .
furthermore , these types of results may be seen as the analogue in partial differential equations ( or pseudo - differential equations ) of the classical aubry - mather theory for dynamical systems , see @xcite ( a more detailed discussion about the existence literature will follow ) . as a matter of fact
, we will consider here a more general equation than .
indeed , we will deal with operators that are more general than the fractional laplacian , which can be also spatially heterogeneous and periodic , and also with more general forcing terms , which may possess different growths from the pure phases other than the classical quadratic growth .
the details of the mathematical framework in which we work are the following . for @xmath12 , we consider the formal energy functional @xmath13 the term @xmath14 $ ] is supposed to be a measurable and symmetric function , comparable to the kernel of the fractional laplacian .
that is , @xmath15 and for large values of @xmath16 - see e.g. hypothesis ( 1.3 ) in @xcite or ( 2.2b ) in @xcite - we prefer to adopt the more restrictive condition in order to simplify the exposition .
requirements and nonetheless allow for a great variety of space - dependent , possibly truncated kernels .
in particular , we stress that no regularity is asked on @xmath17 . ]
@xmath18}(|x - y|)}{|x - y|^{n + 2 s } } { \leqslant}k(x , y ) { \leqslant}\frac{\lambda}{|x - y|^{n + 2 s } } \quad \mbox{for a.a .
} x , y \in { \mathbb{r}}^n,\ ] ] for some @xmath19 and @xmath20 .
the mapping @xmath21 is a double - well potential , with zeros in @xmath6 and @xmath22 .
more specifically , we assume @xmath23 to be a bounded measurable function for which @xmath24 and , for any @xmath25 , @xmath26 where @xmath27 is a non - increasing positive function of the interval @xmath28 . moreover , we require @xmath21 to be differentiable in the second component , with partial derivative locally bounded in @xmath29 , uniformly in @xmath30 . accordingly , we let @xmath31,\ ] ] for some @xmath32 . since we are interested in modelling a periodic environment , we require both @xmath17 and @xmath21 to be periodic under integer translations .
that is , @xmath33 and @xmath34 for any fixed @xmath29 .
the assumptions listed above allow us to comprise a very general class of kernels and potentials . as possible choices for @xmath17
, we could indeed think of heterogeneous , isotropic kernels of the type @xmath35 for a measurable @xmath36 $ ] , or instead consider a translation invariant , but anisotropic @xmath17 , as given by @xmath37 with @xmath38 a measurable norm in @xmath39 .
furthermore , one can combine both heterogeneity and anisotropy to obtain , for instance , kernels of the form @xmath40 where @xmath41 is a symmetric , uniformly elliptic @xmath42 matrix with bounded entries .
of course , the functions @xmath43 and @xmath41 should satisfy appropriate symmetry and periodicity conditions , in order that hypotheses and could be fulfilled by the resulting @xmath17 s .
also , such functions may exhibit a degenerate behavior when @xmath44 and @xmath45 are far from each other ( compare this with the left - hand side of ) .
important examples of admissible potentials @xmath21 are given by @xmath46 with @xmath47 and @xmath2 a positive periodic function.[multiblock footnote omitted ] by taking @xmath48 and @xmath49 , one obtains that the critical points of the energy functional satisfy the model equation in ( up to normalization constants ) .
in the present work we look for minimizers of the functional @xmath50 which connects the two pure phases @xmath6 and @xmath22 , which are the zeroes of the potential @xmath21 .
in particular , given any vector @xmath51 , we address the existence of minimizers for which , roughly speaking , _ most _ of the transition between the pure states occurs in a strip orthogonal to @xmath52 and of universal width .
moreover , when @xmath52 is a rational vector , we want our minimizers to exhibit some kind of periodic behavior , consistent with that of the ambient space .
note that we will often call a quantity _ universal _ if it depends at most on @xmath53 , @xmath8 , @xmath54 , @xmath55 , @xmath56 and on the function @xmath27 introduced in . in order to formulate an exact statement ,
we introduce the following terminology . for a given @xmath57
, we consider in @xmath39 the relation @xmath58 defined by setting @xmath59 notice that @xmath58 is an equivalence relation and that the associated quotient space @xmath60 is topologically the cartesian product of an @xmath61-dimensional torus and a line .
we say that a function @xmath62 is _ periodic with respect to @xmath58 _ , or simply _
@xmath58-periodic _ , if @xmath10 respects the equivalence relation @xmath58 , i.e. if @xmath63 when no confusion may arise , we will indicate the relation @xmath58 just by @xmath64 and the resulting quotient space by @xmath65 . to specify the notion of minimizers that we take into consideration ,
we need to introduce an appropriate localized energy functional . given a set @xmath66 and a function @xmath62
, we define the _ total energy _ @xmath50 of @xmath10 in @xmath67 as @xmath68 where @xmath69 notice that when @xmath67 is the whole space @xmath39 , then the energy coincides with that anticipated in . sometimes ,
a more flexible notation for this functional will turn out to be useful . to this aim , recalling our symmetry assumption on @xmath17 , we will refer to @xmath70 as the sum of the _ _ kinetic part _ _ is typically used to describe non - local interactions and elastic forces . however , we adopt this slight abuse of terminology in conformity with the classical jargon used for local dirichlet energies in particle mechanics . ] @xmath71 with @xmath72 for any @xmath73 , and the _ potential part _
@xmath74 with this in hand , the notion of minimization inside a bounded set is described by the following [ mindef ] let @xmath67 be a bounded subset of @xmath39 .
a function @xmath10 is said to be a _ local minimizer _ of @xmath50 in @xmath67 if @xmath75 and @xmath76 for any @xmath77 which coincides with @xmath10 in @xmath78 .
for simplicity , in definition [ mindef ] and throughout the paper we assume every set and every function to be measurable , even if it is not explicitly stated .
[ restrmk ] we point out that a minimizer @xmath10 on @xmath67 is also a minimizer on every subset of @xmath67 .
though not obvious , this property is easily justified as follows .
let @xmath79 be measurable sets and @xmath77 be a function coinciding with @xmath10 outside @xmath80 . recalling the notation introduced in ,
it is immediate to check that @xmath81 and @xmath82 therefore , it follows that the integrands of the kinetic parts of @xmath70 and @xmath83 coincide on @xmath84 . since also
the respective arguments of the potential terms are equal on @xmath85 , by we conclude that @xmath86 thus , @xmath10 is a minimizer on @xmath80 . up to now
we only discussed about local minimizers . since we plan to construct functions which exhibit minimizing properties on the full space , we need to be precise on how we mean to extend definition [ mindef ] to the whole of @xmath39 ( where the total energy functional may be divergent ) .
[ classadef ] a function @xmath10 is said to be a _ class a minimizer _ of the functional @xmath50 if it is a minimizer of @xmath50 in @xmath67 , for any bounded set @xmath87 . now that all the main ingredients have been introduced , we are ready to state formally the main result of the paper .
[ mainthm ] let @xmath12 and @xmath20 .
assume that the kernel @xmath17 and the potential @xmath21 satisfy , , and , , , , respectively .
+ for any fixed @xmath88 , there is a constant @xmath89 , depending only on @xmath90 and on universal quantities , such that , given any @xmath51 , there exists a class a minimizer @xmath91 of the energy @xmath50 for which the level set @xmath92 is contained in the strip @xmath93 \right\}.\ ] ] moreover , 1 .
if @xmath57 , then @xmath91 is periodic with respect to @xmath58 , while 2 . if @xmath94 , then @xmath91 is the uniform limit on compact subsets of @xmath39 of a sequence of periodic class a minimizers .
we remark that theorem [ mainthm ] is new even in the model case in which @xmath48 and @xmath49 . in this case
, theorem [ mainthm ] provides solutions of equation ( up to normalizing constants ) . in the local case - which formally corresponds to taking @xmath95 and can be effectively realized by replacing our kinetic term with the dirichlet - type energy @xmath96 where @xmath41 is a bounded , uniformly elliptic matrix - the result contained in theorem [ mainthm ] was proved by the second author in @xcite .
after this , several generalizations were obtained , extending such result in many directions .
see , for instance , @xcite and @xcite .
we also mention the pioneering work @xcite of caffarelli and de la llave , where the two authors proved the existence of plane - like minimal surfaces with respect to periodic metrics of @xmath39 .
the proof of theorem [ mainthm ] makes use of a geometric and variational technique developed in @xcite and @xcite , suitably adapted in order to deal with non - local interactions . for a given rational direction @xmath57 and a fixed strip @xmath97 \right\},\ ] ] with @xmath98 ,
one takes advantage of the identifications of the quotient space @xmath65 to gain the compactness needed to obtain a minimizer @xmath99 with respect to periodic perturbations supported inside @xmath100 . by construction ,
this minimizer is such that its interface @xmath101 is contained in the strip @xmath100 . with the aid of some geometrical arguments
, one then shows that @xmath99 becomes a class a minimizer for @xmath50 , provided @xmath102 is larger than some universal parameter @xmath103 . the fact that the threshold @xmath103 is universal and that , in particular , it does not depend on the fixed direction @xmath52 is of key importance here and it allows , as a byproduct , to obtain the result for an irrational vector @xmath94 , by taking the limit of rational directions .
we remark that the non - local character of the energy @xmath50 introduces several challenging difficulties into the above scheme .
first of all , the way the compactness is used to construct the minimizer @xmath99 is somehow not as straightforward as in the local case . to have a glimpse of this difference
, consider that in @xcite the candidate @xmath99 is by definition a minimizer with respect to @xmath64-periodic perturbations occurring in @xmath100 .
that is , one really considers the energy @xmath50 driven by as defined on the cylinder @xmath65 viewed as a manifold and obtain @xmath99 as the absolute minimizer of @xmath50 within a particular class of functions defined on @xmath65 .
however , since the restriction of the local kinetic term to a fundamental domain of @xmath65 only _ sees _ what happens inside that domain , it is clear that one is allowed in the local case to identify periodic perturbations and perturbations which are compactly supported inside @xmath65 . as a result ,
@xmath99 is automatically a local minimizer for @xmath50 in the strip @xmath100 .
as it is , this technique can not work in the non - local setting .
indeed , let @xmath10 be any @xmath64-periodic function and @xmath104 be compactly supported in a fixed fundamental region @xmath105 of @xmath65 : if we denote by @xmath106 the @xmath64-periodic extension of @xmath107 to @xmath39 , then the two quantities @xmath108 and @xmath109 , as defined in , are not equal in general . in order to overcome this difficulty
, we introduce an appropriate auxiliary functional @xmath110 that is used to define the periodic minimizer @xmath99 .
then , it happens that @xmath99 is a local minimizer for the original energy @xmath50 , since @xmath110 couples with @xmath50 in a favorable way .
an additional difficulty comes from the different asymptotic properties of the energy in terms of the fractional parameter @xmath8 . as a matter of fact
, the threshold @xmath9 distinguishes the local and non - local behavior of the functional at a large scale ( see @xcite ) and it reflects into the finiteness or infiniteness of the energy of the one - dimensional transition layer . in our setting , this feature implies that not all the kernels @xmath17 satisfying can be dealt with at the same time .
more precisely , when @xmath111 the behavior at infinity dictated by causes infinite contributions coming from far .
for this reason , at least at a first glance , it may seem necessary to restrict the class of admissible kernels by imposing some additional requirements on the decay of @xmath17 at infinity .
however , we will be able to remove this limitation by an appropriate limit procedure .
namely , we will first assume a fast decay property of the kernel to obtain the existence of a class a minimizer , but the estimates obtained will be independent of this additional assumption .
consequently , we will be able to extend the result to general kernels by treating them as limits of truncated ones .
finally , we want to point out a possibly interesting difference between the proof displayed here and that of e.g. @xcite and @xcite . in the existing literature , the technique that is typically adopted to show that @xmath99 is a class a minimizer relies on the so - called energy and density estimates .
these estimates respectively deal with the growth of the energy @xmath50 of a local minimizer @xmath10 inside _ large _ balls and the fractions of such balls occupied by a fixed level set of @xmath10 .
the latter , in particular , is a powerful tool first introduced by caffarelli and crdoba in @xcite to study the uniform convergence of the level sets of a family of _ scaled _ minimizers .
although such density estimates have been established in @xcite in a non - local setting very close to ours , for some technical reasons we decided not to incorporate them into our argument ( roughly speaking , the periodic setting is not immediately compatible with large balls in euclidean spaces ) . in their place
, we take advantage of some @xmath112 bounds satisfied by local minimizers of @xmath50 , along with a suitable version of the energy estimates .
the above mentioned hlder continuity result is essentially the regularity theory for bounded weak solutions to integro - differential equations developed by kassmann in @xcite . on the other hand , energy estimates for minimizers of non - local energies
have been independently obtained in @xcite and @xcite ( in different settings ) . since both
these two results were set in a slightly different framework than ours , we provide their proofs in full details in sections [ regsec ] and [ enestsec ] , respectively . the paper is organized as follows .
sections [ regsec ] and [ enestsec ] are devoted to the hlder regularity of the minimizers and the energy estimates .
we stress that in these two sections both @xmath17 and @xmath21 are subjected to slightly more general requirements than those listed in the introduction ( the statements of the results proved in these sections will contain the precise hypotheses needed for their proofs ) .
section [ s>1/2sec ] is occupied by the main construction leading to the proof of theorem [ mainthm ] . for the reader
s ease , this section is in turn divided into seven short subsections . in each of these subsections
, we will consider , respectively : * the minimization arguments by compactness , * the notion of minimal minimizer ( i.e. the pointwise infimum of all the possible minimizers , which satisfy additional geometric and functional features ) , * the doubling property ( roughly , doubling the period does not change the minimal minimizer ) , * the notion of minimization under compact perturbations , * the birkhoff property ( namely , the level sets of the minimal minimizers are ordered by integer translations ) , * the passage from constrained to unconstrained minimization ( for large strips , we show that the constraint is irrelevant ) , * the passage from rational to irrational slopes . the argument displayed in section [ s>1/2sec ] only works under an additional assumption on the decay rate of the kernel @xmath17 at infinity . in the subsequent section [ s<1/2sec ]
we will show that this hypothesis can be in fact removed by a limit procedure .
the proof of theorem [ mainthm ] will therefore be completed .
we conclude this paper with two appendices which contain some auxiliary material needed for the technical steps in the proofs of our main results .
in this introductory section we show that the local minimizers of @xmath50 are hlder continuous functions . in order to do this ,
we prove a general regularity result for bounded solutions to non - local equations driven by measurable kernels comparable to that of the fractional laplacian . in this regard
, we stress that the main result of this section - namely , theorem [ holderthm ] - is stated in a broader setting , in comparison with the rest of the paper .
the periodicity of the medium does not play any role here and it is therefore not assumed .
we point out that , while we do not obtain uniform estimates as @xmath113 , our result is still independent of @xmath8 , as long as @xmath8 is far from @xmath114 and @xmath22 .
let @xmath115 and @xmath67 be a bounded open set of @xmath39 .
let @xmath17 be a measurable kernel satisfying and .
we now introduce the space of solutions @xmath116 . given a measurable function @xmath62 , we say that @xmath117 if and only if @xmath118 it is not difficult to see that implies that @xmath119 .
we also denote by @xmath120 the subspace of @xmath116 made up by the functions which vanish a.e . outside @xmath67
. then @xmath121 , if @xmath122 .
we refer the reader to ( * ? ? ?
* section 5 ) , where some useful properties of these spaces are discussed .
we consider the non - local dirichlet form @xmath123 observe that @xmath124 is well - defined for instance when @xmath117 and @xmath125 .
let now @xmath126 .
we say that @xmath117 is a _ supersolution _ of the equation @xmath127 if @xmath128 analogously , one defines _
subsolutions _ of by reverting the inequality in and , as well , _
solutions _ by asking to be an identity and neglecting the sign assumption on @xmath104 .
it is almost immediate to check that a function @xmath10 is a solution of if and only if it is at the same time a super- and a subsolution .
the main result of the section is given by the following [ holderthm ] let @xmath67 be a bounded open set of @xmath39 , with @xmath12 , and @xmath129 be a fixed parameter .
let @xmath130 $ ] and @xmath17 be a measurable kernel satisfying and .
if @xmath131 and @xmath132 is a solution of in @xmath67 , then there exists an exponent @xmath133 , only depending on @xmath53 , @xmath134 , @xmath54 and @xmath55 , such that @xmath135 in particular , there exists a number @xmath136 , depending only on @xmath53 , @xmath134 , @xmath54 and @xmath55 , such that , for any point @xmath137 and any radius @xmath138 for which @xmath139 , it holds @xmath140,\ ] ] for any @xmath141 .
theorem [ holderthm ] is an extension to non - local equations of the classical de giorgi - nash - moser regularity theory . in recent years a great number of papers dealt with interior hlder estimates for solutions of elliptic integro - differential equations , as for instance @xcite and @xcite .
however , since we have not been able to find a satisfactory reference for theorem [ holderthm ] in our exact setting , we provide here all the details of its proof . before advancing to the arguments that lead to theorem [ holderthm ] , we point out how the regularity of the minimizers of @xmath50 can be recovered from it .
[ minminholdercor ] fix @xmath129 and let @xmath130 $ ] .
let @xmath10 be a bounded local minimizer of @xmath50 in a bounded open subset @xmath67 of @xmath39 .
then , @xmath142 , for some @xmath133 .
the exponent @xmath143 only depends on @xmath53 , @xmath134 , @xmath54 and @xmath55 , while the @xmath112 norm of @xmath10 on any @xmath144 may also depend on @xmath145 , @xmath146 and @xmath147 .
let @xmath10 be a bounded local minimizer of @xmath50 in @xmath67 . by taking the first variation of @xmath50 ,
it is easy to see that @xmath10 is a solution of the euler - lagrange equation in @xmath67 , with @xmath148 .
notice that @xmath117 , since @xmath70 is finite .
moreover , being @xmath149 and @xmath150 locally bounded , we obtain that @xmath151 is also a bounded function in @xmath67 .
thence , theorem [ holderthm ] applies and yields the @xmath112 regularity of @xmath10 . the quantitative estimate of the hlder norm of @xmath10 on compact subsets of @xmath67 follows by applying along with a standard covering argument .
the remaining part of the section is devoted to the proof of theorem [ holderthm ] , which is based on the moser s iteration technique and some arguments in @xcite .
we begin with a lemma dealing with non - negative supersolutions of .
[ jnlem ] let @xmath152 and @xmath153 be a non - negative supersolution of in @xmath154 .
suppose that @xmath155 for some @xmath156 .
then , @xmath157 for some constant @xmath158 and exponent @xmath159 which depend only on @xmath53 , @xmath134 , @xmath54 and @xmath55 .
we plan to show that @xmath160 . to this aim
, we claim that there exists a constant @xmath161 , depending only on @xmath53 , @xmath134 , @xmath54 and @xmath55 , such that @xmath162_{h^s(b_r(z ) ) } { \leqslant}c_1 r^{- s + n / 2},\ ] ] holds true for any @xmath163 and @xmath164 for which @xmath165 . in order to prove , we take a cut - off function @xmath166 satisfying @xmath167 in @xmath39 , @xmath168 , @xmath169 in @xmath170 and @xmath171 in @xmath39 .
we test formulation with @xmath172 .
note that @xmath173 and @xmath174 thanks to the definition of @xmath175 and condition .
recalling , inequality becomes @xmath176 for any @xmath177 we compute @xmath178 hence , using together with the numerical inequality @xmath179 that holds for any @xmath180 , we get is used to denote the volume of the unit ball of @xmath39 .
that is , @xmath181 accordingly , the @xmath61-dimensional hausdorff measure of the sphere @xmath182 is then given by @xmath183 . ]
@xmath184 k(x , y ) \ , dx dy \\ & { \leqslant}\frac{16 \lambda}{r^2 } \int_{b_{2 r}(z ) } \int_{b_{2 r}(z ) } \frac{dx dy}{|x - y|^{n - 2 + 2 s } } - \lambda \int_{b_r(z ) } \int_{b_r(z ) } \frac{\left| u(x ) - u(y ) \right|^2}{u(x ) u(y ) } \ , \frac{dx dy}{|x - y|^{n + 2 s } } \\ & { \leqslant}2^{n + 4 } n \alpha_n^2 \lambda r^{n - 2 } \int_0^{4 r } \rho^{1 - 2 s } \ , d\rho - \lambda \int_{b_r(z ) } \int_{b_r(z ) } \frac{\left| \log u(x ) - \log u(y ) \right|^2}{|x - y|^{n + 2 s } } \ , dx dy \\ & { \leqslant}\frac{2^{n + 7 } n \alpha_n^2 \lambda}{s_0 } \ , r^{n - 2 s } - \lambda [ \log u]_{h^s(b_r(z))}^2 .
\end{aligned}\ ] ] on the other hand , by the non - negativity of @xmath10 and again we estimate @xmath185 finally , using we have @xmath186 since @xmath187 .
claim then follows by combining this last equation with , and .
we are now ready to show that @xmath160 . for a bounded @xmath87 and @xmath188 , write @xmath189 applying both hlder s and fractional poincar s inequality , from we obtain @xmath190_{h^s(b_r(z ) ) } \\ & { \leqslant}c_3 r^n,\end{aligned}\ ] ] for some @xmath191 which may depend on @xmath53 , @xmath134 , @xmath54 and @xmath55 . since the above inequality holds for any @xmath165 ,
we conclude that @xmath160 .
estimate then follows by the john - nirenberg embedding in one of its equivalent forms ( see , for instance , theorem 6.25 of @xcite ) .
observe that the exponent @xmath192 given by such result is of the form of a dimensional constant divided by the @xmath193 semi - norm of @xmath194 .
this norm being bounded from above by @xmath195 and since we are free to make @xmath192 smaller if necessary , it turns out that we can choose @xmath159 to depend only on @xmath53 , @xmath134 , @xmath54 and @xmath55 .
next is the step of the proof in which the iterative argument really comes into play .
[ moserlem ] let @xmath152 and @xmath153 be a supersolution of in @xmath154 .
assume that @xmath10 satisfies , for some @xmath156 .
then , for any @xmath196 , @xmath197 for some constant @xmath198 which may depend on @xmath53 , @xmath134 , @xmath54 , @xmath55 and @xmath199 .
fix @xmath88 .
we claim that , for any @xmath200 $ ] and @xmath201 , it holds @xmath202 for some constant @xmath161 depending on @xmath53 , @xmath134 , @xmath54 and @xmath55 . to prove ,
consider a cut - off @xmath166 such that @xmath167 in @xmath39 , @xmath203 , @xmath169 in @xmath204 and @xmath205 in @xmath39 , and plug @xmath206 into .
inequality then follows by arguing as in lemma 3.5 of @xcite and noticing that , by , @xmath207 where we also used the fact that @xmath187 . by using in combination with the fractional sobolev inequality
, we then deduce @xmath208 for some @xmath209 which depends only on @xmath53 , @xmath134 , @xmath54 and @xmath55 .
we are now in position to run the iterative scheme , which is based on the fundamental estimate .
for any @xmath210 , define @xmath211 so that @xmath212 we apply with @xmath213 , @xmath214 and @xmath215 , to get @xmath216 for any @xmath210 , where @xmath217^{1 / p_k}.\ ] ] from it then follows that @xmath218 now we observe that @xmath219 therefore , recalling that @xmath220 , @xmath221 and hence @xmath222 { \leqslant}\frac{1}{p_k } \log \left [ c_3 \left ( \frac{2 n}{n - 2 s } \right)^{2 k } \right ] { \leqslant}c_4 \left ( \frac{n - 2 s_0}{n } \right)^k k,\ ] ] for some @xmath223 that may also depend on @xmath199 .
this implies that the product of the @xmath224 s converges , as @xmath225 .
thence , follows from , since @xmath226 by putting together lemmata [ jnlem ] and [ moserlem ] , we easily obtain the following _ weak harnack inequality_. [ weakharnackcor ] let @xmath227 $ ] and @xmath228 .
assume that @xmath229 is a non - negative supersolution of in @xmath230 .
then , @xmath231 for some @xmath232 depending only on @xmath53 , @xmath134 , @xmath54 and @xmath55 .
assume for the moment @xmath233 .
let then @xmath156 be a small parameter and define @xmath234 .
note that @xmath235 is still a non - negative supersolution of in @xmath154 and that it satisfies .
thus , we are free to apply lemmata [ jnlem ] and [ moserlem ] to @xmath235 and obtain that @xmath236 letting @xmath237 we obtain when @xmath233 . for a general radius @xmath238
the result follows by a simple scaling argument . with the aid of corollary [ weakharnackcor ]
, we can prove the following proposition , which will be the fundamental step in the conclusive inductive argument . in the literature ,
results of this kind are often known as _ growth lemmata_. [ keyprop ] there exist @xmath239 and @xmath240 , depending only on @xmath53 , @xmath134 , @xmath54 and @xmath55 , such that for any @xmath227 $ ] , @xmath228 and @xmath229 supersolution of in @xmath230 , for which @xmath241 @xmath242 and @xmath243 hold true , then @xmath244 write @xmath245 . using and , it is easy to see that @xmath246 is a supersolution of @xmath247 where @xmath248 applying corollary [ weakharnackcor ] we get that @xmath249 using then hypotheses and , this yields @xmath250 now we turn our attention to the @xmath251 norm of @xmath252 .
first , we notice that implies that @xmath253 as the right hand side of is negative . moreover , given @xmath254 and @xmath255 , it holds @xmath256 consequently , recalling we compute @xmath257 \\ & = 2^{n + 1 }
n \alpha_n \lambda \left [ \frac{8^\gamma}{2 s - \gamma } - \frac{1}{2 s } \right ] r^{- 2 s},\end{aligned}\ ] ] if @xmath258 .
notice that the term in brackets on the last line of the above formula converges to @xmath114 as @xmath259 , uniformly in @xmath260 .
therefore , we can find @xmath261 , in dependence of @xmath53 , @xmath134 , @xmath54 and @xmath55 , such that @xmath262 inequality then follows by combining this with .
we are now ready to move to the actual we focus on the proof of , as the hlder continuity of @xmath10 inside @xmath67 would then easily follow .
furthermore , we may assume without loss of generality @xmath263 to be the origin .
set @xmath264 with @xmath175 as in proposition [ keyprop ] , and take @xmath265 $ ] .
we claim that there exist a constant @xmath133 , depending only on @xmath53 , @xmath8 , @xmath54 and @xmath55 , a non - decreasing sequence @xmath266 and a non - increasing sequence @xmath267 of real numbers such that for any @xmath268 @xmath269 with @xmath270 we prove this by induction .
set @xmath271 and @xmath272 . with this choice
, property clearly holds true for @xmath273 .
then , for a fixed @xmath274 , we assume to have constructed the two sequences @xmath266 and @xmath267 up to @xmath275 in such a way that is satisfied and show that we can also build @xmath276 and @xmath277 . for any @xmath30 ,
define @xmath278 with @xmath279 and @xmath280 as in proposition [ keyprop ] .
since @xmath10 is a solution of in @xmath67 , we deduce that @xmath77 satisfies @xmath281 moreover , @xmath282 , there exists a unique @xmath283 for which @xmath284 writing @xmath285 and @xmath286 for every @xmath287 , we compute @xmath288 analogously , one checks that @xmath289 for a.a . @xmath290 .
we distinguish between the two mutually exclusive possibilities a. @xmath291 , and b. @xmath292 . in case ( a ) , set @xmath293 . from
we deduce in particular that @xmath294 in view of and we apply proposition [ keyprop ] to @xmath295 , with @xmath296 , and obtain that @xmath297 from which , by and , it follows @xmath298 note that we took advantage of the fact that @xmath299 , by .
if we translate this estimate back to @xmath10 , applying once again we finally get @xmath300 accordingly , is satisfied by setting @xmath301 and @xmath302 .
if on the other hand ( b ) holds we define @xmath303 . with a completely analogous argument using in place of
, we end up estimating @xmath304 so that again follows with @xmath305 and @xmath306 .
the proof of the theorem is therefore complete , as the bound in is an immediate consequence of claim .
we include here a result which addresses the growth of the energy @xmath50 of local minimizers inside large balls .
we point out that , as in section [ regsec ] , this estimate is set in a general framework .
in particular , the periodicity of @xmath17 and @xmath21 encoded in and is not significant here . writing @xmath307 we can state the following [ enestprop ] let @xmath308 , @xmath20 , @xmath309 and @xmath310 . assume that @xmath17 and @xmath21 satisfy encoded in is relevant here .
thus , no assumption on the derivative @xmath150 is necessary .
see in particular the the proof of proposition [ enestprop ] .
] , and , , respectively . if @xmath311 $ ] is a local minimizer of @xmath50 in @xmath312 , then @xmath313 for some constant @xmath314 which depends on @xmath53 , @xmath8 , @xmath55 and @xmath56
. the above proposition will play an important role later in subsection [ unconstrsubsec ] , as it will imply that the interface region of a minimizer can not be too wide .
estimate has first been proved in @xcite and @xcite for the fractional laplacian .
while in the first paper the authors use the harmonic extension of @xmath10 to @xmath315 to prove , in the latter work the result is obtained by explicitly computing the energy @xmath50 of a suitable competitor of @xmath10 .
it turns out that this strategy is flexible enough to be adapted to our framework and the proof of proposition [ enestprop ] is actually an appropriate and careful modification of that of ( * ? ? ?
* theorem 1.3 ) . before heading to the proof of proposition [ enestprop ] ,
we first need the following auxiliary result that will be also widely used in the following section [ s>1/2sec ] .
[ uvmmlem ] let @xmath316 be two measurable subsets of @xmath39 and @xmath317 .
then , @xmath318 and @xmath319 since the derivation of identity is quite straightforward , we focus on only .
we write for simplicity @xmath320 and @xmath321 .
observe that we may assume the right hand side of to be finite , the result being otherwise obvious . in order to show , we actually prove the stronger pointwise relation @xmath322 .
let then @xmath44 and @xmath45 be two fixed points in @xmath39 .
in order to check that is true , we consider separately the two possibilities a. @xmath323 and @xmath324 , or @xmath325 and @xmath326 ; b. @xmath323 and @xmath326 , or @xmath325 and @xmath324 . in the first situation it is immediate to see that holds as an identity .
suppose then that point ii ) occurs .
if this is the case , we compute @xmath327 which is .
the proof of the lemma is thus complete . without loss of generality ,
we assume @xmath263 to be the origin . in the course of the proof we will denote as @xmath328 any positive constant which depends at most on @xmath53 , @xmath8 , @xmath55 and @xmath56 .
let @xmath329 be the radially symmetric function defined by @xmath330 we claim that @xmath329 satisfies in @xmath331 , that is @xmath332 indeed , let @xmath333 and set @xmath334 .
it is easy to see that @xmath335 consequently , applying we compute @xmath336 \\ & { \leqslant}c d(x)^{- 2 s}.\end{aligned}\ ] ] furthermore , using polar coordinates we get @xmath337 hence , @xmath338 since by and we also have @xmath339 it is clear that estimate follows .
now , set @xmath340 and @xmath341 . by the definition of @xmath329 and the fact that @xmath342 , we observe that @xmath343 and @xmath344 by virtue of , @xmath345 on the other hand , we claim that @xmath346 indeed , using , and the fact that @xmath347 a.e . in @xmath39
, we compute @xmath348 k(x , y ) \ , dy \right ) dx \\ & \hspace{50pt } { \leqslant}4 \lambda \int_{b_r } \left ( \int_{{\mathbb{r}}^n \setminus b_{r + 1 } } |x - y|^{- n - 2 s } \ , dy \right )
dx { \leqslant}c \int_{b_r } d(x)^{- 2 s } \ , dx,\end{aligned}\ ] ] and
claim then follows from .
accordingly , by and we obtain that @xmath349 we now take advantage of the minimality of @xmath10 and to deduce @xmath350 then , from this and lemma [ uvmmlem ] it follows immediately that @xmath351 note that the first inequality above is true as a consequence of the inclusion @xmath352 ( see remark [ restrmk ] ) . by applying in sequence , and
, we finally get .
the present section contains the proof of theorem [ mainthm ] under the additional assumption that @xmath17 satisfies @xmath353 for some constants @xmath354 . we stress that this hypothesis is merely technical and in fact it will be removed later in section [ s<1/2sec ] .
however , we need the fast decay of the kernel @xmath17 at infinity - ensured by the fact that @xmath355 - in order to perform a delicate construction at some point ( roughly speaking , the decay assumed in is needed to ensure the existence of a competitor with finite energy in the large , but the geometric estimates will be independent of the quantities in and this will allow us to perform a limit procedure ) . hence , we assume to hold in the whole section . notice that if @xmath356 , then is automatically fulfilled in view of .
the argument leading to the proof of theorem [ mainthm ] is long and articulated .
therefore , we divide the section into several subsections which we hope will make the reading easier .
we first deal with the case of a rational direction @xmath52 . under this assumption
, we can take advantage of the equivalence relation @xmath58 defined in to build the minimizer .
this construction occupies subsections [ minpersubsec]-[unconstrsubsec ] .
irrational directions - i.e. @xmath94 - are then treated in subsection [ irrsubsec ] as limiting cases . for simplicity of exposition
, we restrict ourselves to consider @xmath357 .
the general case is in no way different .
of course , the choice @xmath358 is made in order to represent a value of @xmath90 close to @xmath22 .
let @xmath57 be fixed .
given a measurable function @xmath62 , we say that @xmath359 if @xmath360 and @xmath10 is periodic with respect to @xmath64 . given @xmath361 ,
let @xmath362 be the set of admissible functions .
we introduce the auxiliary functional @xmath363 note that in the integrals above , @xmath65 stands for any fundamental domain of the relation @xmath64 . in the following
, we will often identify quotients with any of their respective fundamental domains .
the aim of this subsection is to prove the existence of an _ absolute minimizer _ of @xmath110 within the class @xmath364 , that is a function @xmath365 such that @xmath366 , for any @xmath367 .
such minimizers are the building blocks of our construction , as will become clear in the sequel . as a first step toward this goal
, we show that @xmath110 is not identically infinite on @xmath364 .
[ finfunlem ] let @xmath368 be defined by setting @xmath369 , where @xmath370 is the piecewise linear function given by @xmath371 then , @xmath372 . since @xmath373 vanishes at @xmath374 , for a.a .
@xmath30 , it is clear that the potential term of @xmath110 evaluated at @xmath375 is finite .
thus , we only need to estimate the kinetic term . to do this , by and , it is in turn sufficient to show that @xmath376 notice that , up to an affine transformation , we may take @xmath377 . moreover
, we assume for simplicity that @xmath378 and @xmath379 . in this setting
, we have @xmath380^{n - 1 } \times { \mathbb{r}}$ ] and , consequently , is equivalent to @xmath381^{n - 1 } \times { \mathbb{r } } } \left ( \int_{b_{\bar{r}}(x ) } \frac{|\bar{u}(x ) - \bar{u}(y)|^2}{|x - y|^{n + 2 s } } \ , dy \right ) dx < + \infty,\ ] ] and @xmath382^{n - 1 } \times { \mathbb{r } } } \left ( \int_{{\mathbb{r}}^n \setminus b_{\bar{r}}(x ) } \frac{|\bar{u}(x ) - \bar{u}(y)|^2}{|x - y|^{n + \beta } } \ , dy \right ) dx < + \infty.\ ] ] by the definition of @xmath375 , it is clear that @xmath383^{n - 1 } \times [ - \bar{r } , \bar{r } + 1 ] } \left ( \int_{b_{\bar{r}}(x ) } \frac{|\bar{u}(x ) - \bar{u}(y)|^2}{|x - y|^{n + 2 s } } \ , dy \right ) dx.\ ] ] then , we take advantage of @xmath375 being lipschitz to compute , using polar coordinates , @xmath384^{n - 1 } \times [ - \bar{r } , \bar{r } + 1 ] } \left ( \int_{b_{\bar{r}}(x ) } \frac{dy}{|x - y|^{n + 2 s - 2 } } \right ) dx = \frac{2 n \alpha_n}{1 - s } ( 2 \bar{r } + 1 ) \bar{r}^{2 - 2 s},\ ] ] which implies . on the other hand , to prove we first write @xmath385 , where @xmath386^{n - 1 } \times [ 2 , + \infty ) } \left ( \int_{{\mathbb{r}}^n \setminus b_{\bar{r}}(x ) } \frac{|\bar{u}(x ) - \bar{u}(y)|^2}{|x - y|^{n + \beta } } \ , dy \right ) dx , \\
j_2 & : = \int_{[0 , 1]^{n - 1 } \times ( - \infty , - 1 ] } \left ( \int_{{\mathbb{r}}^n \setminus b_{\bar{r}}(x ) } \frac{|\bar{u}(x ) - \bar{u}(y)|^2}{|x - y|^{n + \beta } } \ , dy \right ) dx , \\
j_3 & : = \int_{[0 , 1]^{n - 1 } \times [ - 1 , 2 ] } \left ( \int_{{\mathbb{r}}^n \setminus b_{\bar{r}}(x ) } \frac{|\bar{u}(x ) - \bar{u}(y)|^2}{|x - y|^{n + \beta } } \ , dy \right ) dx.\end{aligned}\ ] ] using the definition of @xmath375 , we observe that @xmath387^{n - 1 } \times [ 2 , + \infty ) } \left ( \int_{{\mathbb{r}}^{n - 1 } \times ( - \infty , 1 ] } \frac{|- 1 - \bar{\mu}(y_n)|^2}{|x - y|^{n + \beta } } \ , dy \right ) dx \\ & { \leqslant}4 \int_{[0 , 1]^{n - 1 } \times [ 2 , + \infty ) } \left ( \int_{{\mathbb{r}}^{n - 1 } \times ( - \infty , 1 ] } \frac{dy}{|x - y|^{n + \beta } } \right ) dx.\end{aligned}\ ] ] making the substitution @xmath388 , we have @xmath389 } \frac{dy}{|x - y|^{n + \beta } } & = \int_{- \infty}^1 |x_n - y_n|^{- n - \beta } \left [ \int_{{\mathbb{r}}^{n - 1 } } \left ( 1 + \frac{|x ' - y'|^2}{|x_n - y_n|^2 } \right)^{- \frac{n + \beta}{2 } } dy ' \right ] dy_n \\ & = \int_{- \infty}^1 |x_n - y_n|^{- 1 - \beta } \left [ \int_{{\mathbb{r}}^{n - 1 } } \left ( 1 + |z'|^2 \right)^{- \frac{n + \beta}{2 } } dz ' \right ] dy_n \\ & = \frac{\xi}{\beta } ( x_n - 1)^{- \beta},\end{aligned}\ ] ] where we denoted with @xmath390 the finite quantity @xmath391 accordingly , @xmath392 since @xmath355 .
similarly , one checks that @xmath393 is finite too .
the computation of @xmath394 is simpler . by taking advantage of the fact that @xmath375 is a bounded function and switching to polar coordinates
, we get @xmath395^{n - 1 } \times [ - 1 , 2 ] } \left ( \int_{{\mathbb{r}}^n \setminus b_{\bar{r}}(x ) } \frac{dy}{|x - y|^{n + \beta } } \right ) dx = \frac{12 n \alpha_n}{\beta } \bar{r}^{- \beta}.\ ] ] hence , follows .
we want to highlight how crucial condition has been in the proof of the above lemma . indeed ,
if the kernel @xmath17 has a slower decay at infinity , the result is no longer true .
lemma [ rapidneclem ] in appendix [ addcompapp ] shows that , under this assumption , the functional @xmath110 is nowhere finite on the whole class of admissible functions @xmath364 .
we also point out that this is the only part of the section in which we need the additional hypothesis and future computations will involve neither @xmath396 , nor @xmath397 , nor @xmath398 . with the aid of the finiteness result yielded by lemma [ finfunlem ] , we can now prove the existence of minimizers .
[ perminprop ] there exists an absolute minimizer of the functional @xmath110 within the class @xmath364 .
our argumentation follows the lines of the standard direct method of the calculus of variations . by lemma [ finfunlem ] and
the fact that @xmath110 is non - negative , we know that @xmath399 let then @xmath400 be a minimizing sequence .
observe that we may assume without loss of generality that latexmath:[\[\label{ujle1 } restriction only makes the energy @xmath110 decrease .
moreover , we fix an integer @xmath402 and consider the lipschitz domains @xmath403 by and we have @xmath404_{h^s(\omega_k)}^2 & { \leqslant}\int_{\omega_k } \left ( \int_{b_1(x ) } \frac{\left| u_j(x ) - u_j(y ) \right|^2}{|x - y|^{n + 2 s } } \ , dy \right ) dx + 4 \int_{\omega_k } \left ( \int_{{\mathbb{r}}^n \setminus b_1(x ) } \frac{dy}{|x - y|^{n + 2 s } } \right ) dx \\ & { \leqslant}\frac{2}{\lambda } { { \mathscr{f}_\omega}}(u_j ) + \frac{2 n \alpha_n |\omega_k|}{s},\end{aligned}\ ] ] so that @xmath405 is bounded in @xmath406 , uniformly in @xmath407 . by the compact embedding of @xmath406 into @xmath408 ( see e.g. theorem 7.1 of @xcite ) , we then deduce that a subsequence of @xmath405 converges to some function @xmath10 in @xmath408 and , thus , a.e . in @xmath409 . using a diagonal argument ( on @xmath407 and @xmath410 ) , we may indeed find a subsequence @xmath411 of @xmath405 which converges to @xmath10 a.e . in @xmath65 .
furthermore , we may identify the @xmath412 s and @xmath10 with their @xmath64-periodic extensions to @xmath39 and thus obtain that such convergence is a.e . in the whole space @xmath39 .
accordingly , @xmath365 and an application of fatou s lemma shows that @xmath413 .
this concludes the proof .
denote by @xmath414 the set composed by the absolute minimizers of @xmath110 in @xmath364 , i.e. @xmath415 clearly , @xmath414 is not empty , as shown by proposition [ perminprop ] .
here below we introduce a particular element of the class @xmath414 , that will turn out to be of central interest in the remainder of the paper .
[ minmindef ] we define the _ minimal minimizer _ @xmath416 as the infimum of @xmath414 as a subset of the partially ordered set @xmath417 .
more specifically , @xmath416 is the unique function of @xmath364 for which @xmath418 and @xmath419 of course , the existence of the minimal minimizer is far from being established .
aim of the subsection is to prove that such function is in fact well - defined and that it belongs to @xmath414 itself .
in order to construct @xmath416 we first need to show that the set @xmath414 is closed with respect to the operation of taking the minimum between two of its elements . to do this
, we actually prove a stronger fact , which will be needed , in its full generality , only later in subsection [ birksubsec ] .
[ min2lem ] let @xmath420 and @xmath421 , with @xmath361 and @xmath422 . if @xmath423 and @xmath424 , then @xmath425 .
first , notice that @xmath426 and @xmath427 .
moreover , employing lemma [ uvmmlem ] we deduce @xmath428 taking advantage of this inequality , together with the fact that @xmath424 , we get @xmath429 which in turn implies that @xmath430 consequently , @xmath425 . by choosing @xmath431 and @xmath432
, we obtain the desired [ min2cor ] let @xmath433 . then , @xmath425 .
now that we know that the minimum between two - and , consequently , any finite number of - minimizers is still a minimizer , we can show that also the infimum over a _
countable _ family of elements of @xmath414 belongs to @xmath414 .
[ minnumlem ] let @xmath434 be a sequence of elements of @xmath414
. then , @xmath435 .
write @xmath436 .
we define inductively the auxiliary sequence @xmath437 by corollary [ min2cor ] , we know that @xmath438
. moreover , @xmath439 converges to @xmath440 a.e . in @xmath39 .
an application of fatou s lemma then yields that @xmath441 and @xmath442 for any @xmath274 .
therefore , @xmath443 .
finally , we are in position to prove the main result of the present subsection .
[ minminismin ] the minimal minimizer @xmath416 , as given by definition [ minmindef ] , exists and belongs to @xmath414 .
the set @xmath414 is separable with respect to convergence a.e .
, i.e. there exists a sequence @xmath444 such that for any @xmath423 we may pick a subsequence @xmath445 which converges to @xmath10 a.e . in @xmath39 .
a rigorous proof of this fact can be found in proposition [ aesepprop ] of appendix [ sepapp ] .
set @xmath446 by lemma [ minnumlem ] , we already know that @xmath447 .
we claim that @xmath416 is the minimal minimizer , i.e. that satisfies the properties and listed in definition [ minmindef ] .
take @xmath423 and let @xmath448 be a subsequence of @xmath405 converging to @xmath10 a.e . in @xmath39 . by definition ,
@xmath449 in @xmath39 , for any @xmath274 .
hence , taking the limit as @xmath225 , condition follows .
now we turn our attention to and we assume that there exists @xmath367 such that @xmath450 , for any @xmath451 . then , in particular , we have @xmath452 , for any @xmath287 which implies @xmath453 .
thus , follows and the proof of the proposition is complete .
an important feature of the minimal minimizer is the so - called _ doubling property _ ( or _ no - symmetry - breaking property _ ) .
namely , we prove in this subsection that @xmath416 is still the minimal minimizer with respect to functions having periodicity multiple of @xmath64 . in order to formulate precisely this result
, we need a few more notation .
let @xmath454 denote some vectors spanning the @xmath61-dimensional lattice induced by @xmath64 .
thus , any @xmath455 such that @xmath456 may be written as @xmath457 for some @xmath458 . for a fixed @xmath459
, we introduce the equivalence relation @xmath460 , defined by setting @xmath461 also , set @xmath462 and denote by @xmath463 the space of @xmath460-periodic functions which belong to @xmath464 .
note that @xmath465 contains exactly @xmath466 copies of @xmath65 .
indeed , the relation @xmath460 is weaker than @xmath64 and @xmath467 .
we consider the space of admissible functions @xmath468 related to this new equivalence relation , together with the set of absolute minimizers @xmath469 of the functional @xmath470 we indicate with @xmath471 the minimal minimizer of the class @xmath472 .
of course , its existence is granted by the same arguments of subsection [ minminsubsec ] . finally , given a function @xmath62 and a vector @xmath473
, we denote the translation of @xmath10 in the direction @xmath474 as @xmath475 after this preliminary work , we can now prove that the minimal minimizer in a class of larger period coincides with the one in a class of smaller period : [ u = umprop ] for any @xmath459 , it holds @xmath476 . for simplicity of exposition we restrict ourselves to the case in which @xmath477 and @xmath478 , for every @xmath479 .
the approach in the general case would be analogous , but much heavier in notation .
we begin by showing that @xmath480 .
notice that the inequality follows if we prove that @xmath481 . to see this
, we consider the translation @xmath482 of @xmath471 in the _ doubled _ direction @xmath483 . clearly , @xmath484 .
then , we define @xmath485 observe that @xmath486 is @xmath64-periodic and hence belongs to @xmath364 . then , @xmath487 where the last inequality follows by lemma [ uvmmlem ] , arguing as in the proof of lemma [ min2lem ] .
accordingly , we deduce that @xmath481 and so @xmath480 , since @xmath471 is the minimal minimizer of @xmath472 . on the other hand , being @xmath488 and @xmath489
, we have @xmath490 which implies that @xmath491 .
consequently , @xmath492 , and the proposition is therefore proved . in the previous subsections we have been concerned with functionals of the type @xmath493 .
we proved that absolute minimizers for such functionals exist in particular classes of @xmath460-periodic functions .
since our ultimate goal is the construction of class a minimizers for the energy @xmath50 , we now need to show that the elements of @xmath414 are also minimizers of @xmath50 with respect to compact perturbations occurring within the strip @xmath494 \right\}.\ ] ] in what follows , it will also be useful to introduce the quotient @xmath495 the first result of the subsection addresses a general relationship intervening between the two functionals @xmath50 and @xmath493 . [ efrellem ] let @xmath496 be a bounded function with finite @xmath493 energy .
given an open set @xmath67 compactly contained in @xmath497 , is meant to be compactly contained in a fundamental domain of @xmath497 , and not only in the quotient set itself .
the difference is that we do not allow @xmath67 to touch the _
lateral _ boundary of the domain .
[ cptfoot ] ] let @xmath77 be another bounded function such that @xmath498 outside @xmath67 and set @xmath499 . denoting with @xmath500 and @xmath106 the @xmath460-periodic extensions to @xmath39 of @xmath501 and @xmath502 , respectively , it then holds @xmath503 in particular , if @xmath504 , then @xmath505 note that the integral written on the right - hand sides of and is finite , since @xmath104 is compactly supported on @xmath497 and bounded . for a justification of this fact , see lemma [ intfinlem ] in appendix [ addcompapp ] . for simplicity ,
we restrict ourselves to consider @xmath506 , the general case being completely analogous .
moreover , it is enough to prove formula , as then easily follows by noticing that @xmath507 . recalling definition , we first inspect the term @xmath508 . to this aim ,
let @xmath509 and @xmath510 .
we compute @xmath511 and thus @xmath512 notice now that @xmath513 so that we may write the integral on the second line of as @xmath514 by changing variables as @xmath515 , @xmath516 , recalling and taking advantage of the periodicity of @xmath10 and @xmath106 , we find that @xmath517 by summing up on @xmath410 this identity , becomes @xmath518 the thesis then follows by noticing that @xmath519 and recalling the definitions of @xmath50 and @xmath110 . with this in hand , we may state the following proposition , where we prove that the absolute minimizers of @xmath493 in the class @xmath520 also minimizes @xmath50 with respect to compact perturbations occurring inside @xmath497 [ cptminprop ] let @xmath504 .
then , @xmath10 is a local minimizer of @xmath50 in every open set @xmath67 compactly contained in @xmath497 , that is @xmath521 for any @xmath77 which coincides with @xmath10 outside @xmath67 .
first of all , we assume without loss of generality that @xmath522 and @xmath523 a.e . in @xmath39 .
set @xmath499 and observe that @xmath104 is supported on @xmath67 .
we will show that inequality holds on the larger region @xmath465 , in place of @xmath67 , i.e. @xmath524 this will imply , in light of remark [ restrmk ] . to prove
, we first notice that if @xmath104 is either non - negative or non - positive , then follows as a direct consequence of inequality . on the other hand , if @xmath104 is sign - changing , we consider the minimum and the maximum between @xmath10 and @xmath525 .
recalling lemma [ uvmmlem ] it is immediate to see that @xmath526 moreover , since it holds @xmath527 we may apply and get @xmath528 this leads to . from this proposition and the results of subsection [ doubsubsec ] , we immediately deduce the following [ minmincptcor ] the minimal minimizer @xmath416 is a local minimizer of @xmath50 in every bounded open set @xmath67 compactly contained in the strip @xmath529 . given @xmath67 , we take @xmath459 large enough in order to have @xmath530 . in view of proposition [ u = umprop ] , @xmath416 is the minimal minimizer with respect to @xmath472 .
but then , by proposition [ cptminprop ] , @xmath416 is a local minimizer of @xmath50 in @xmath67 . in this subsection
we introduce an interesting geometric feature shared by the level sets of the minimal minimizer : the _ birkhoff property _ ( also known in the literature as `` non - self - intersection property '' ) .
namely , the level sets of the minimal minimizers are ordered under translations . in order to give a formal definition of this property
, the following notation will be useful .
similarly to what we did in for functions , we consider the translation of a set @xmath531 with respect to a vector @xmath473 @xmath532 notice that , with this notation , the translation of a sublevel set then is given by @xmath533 and analogously for the superlevel sets .
let @xmath534 be a subset of @xmath39 .
we say that @xmath534 has the _ birkhoff property _ with respect to a vector @xmath535 if : 1 .
@xmath536 , for any @xmath455 such that @xmath537 , and 2 .
@xmath538 , for any @xmath455 such that @xmath539 . before exploring the connection between the minimal minimizer and the birkhoff property
, we present a proposition which addresses birkhoff sets from an abstract point of view and displays a rigidity feature of those of such sets that have _ fat _ interior .
[ fatbirk ] let @xmath531 be a set satisfying the birkhoff property with respect to a vector @xmath540 . if @xmath534 contains a ball of radius @xmath541 , then it also contains a half - space which includes the center of the ball , has delimiting hyperplane orthogonal to @xmath542 and is such that @xmath542 points outside of it .
let @xmath543 be the ball of radius @xmath541 and center @xmath263 contained in @xmath534 . by the birkhoff property
, it holds @xmath544 the thesis now follows by observing that the set on the left - hand side above contains the half - space @xmath545 , for some @xmath546 .
now we show that the level sets of the minimal minimizer are birkhoff sets .
recalling the relation between translations and level sets established in , we have [ ubirk ] let @xmath547 .
then , the superlevel set @xmath548 has the birkhoff property with respect to @xmath52 . explicitly , 1 .
@xmath549 , for any @xmath455 such that @xmath550 , and 2 .
@xmath551 , for any @xmath455 such that @xmath552 .
analogously , the sublevel set @xmath553 has the birkhoff property with respect to @xmath554 .
+ the same statements still hold if we replace strict level sets with broad ones .
let @xmath555 and observe that @xmath556 is the minimal minimizer with respect to the strip @xmath557 . if @xmath550 then by lemma [ min2lem ] it follows that @xmath558 .
thus , @xmath559 and hence @xmath560 on the other hand , if @xmath552 then @xmath561 and therefore @xmath562 the conclusion for the sublevel set @xmath563 follows observing that a set @xmath531 is birkhoff with respect to a vector @xmath535 if and only if @xmath564 is birkhoff with respect to @xmath565 . finally , by writing @xmath566 and noticing that the union of a family of sets that are birkhoff with respect to a mutual vector is itself birkhoff with respect to the same vector
, we deduce that @xmath553 has the birkhoff property with respect to @xmath567 . in a similar way one
checks that the superlevel set @xmath568 is birkhoff with respect to @xmath52 . from now on we mainly restrict our attention to strips of the form @xmath569 \right\}.\ ] ] we simply write @xmath570 for the space @xmath571 of admissible functions , @xmath572 for the absolute minimizers and @xmath99 for the minimal minimizer .
we also assume @xmath573 , in order to avoid degeneracies caused by too narrow strips .
the main purpose of this subsection is to show that the minimal minimizer @xmath99 becomes unconstrained for large , universal values of @xmath102 . by _
unconstrained _ we mean that @xmath99 no longer _ feels _ the boundary data prescribed outside the strip @xmath100 and gains additional minimizing properties in the whole space @xmath39 .
of course , we will be more precise on this later in proposition [ unconstrprop ] .
we begin by adapting the results of sections [ regsec ] and [ enestsec ] to the minimal minimizer @xmath99 .
recall that @xmath99 is a local minimizer for @xmath50 inside the strip @xmath100 , thanks to corollary [ minmincptcor ] . in view of corollary [ minminholdercor ]
, we deduce that there exist universal quantities @xmath133 and @xmath574 for which @xmath575 for any open set @xmath576 such that @xmath577 .
on the other hand , proposition [ enestprop ] tells that , given @xmath578 and @xmath310 in such a way that @xmath579 , it holds @xmath580 for a universal constant @xmath581 . recall that @xmath582 was defined in . now that and are established , we may proceed to the core proposition of the present subsection .
[ dist1prop ] there exists a universal @xmath89 such that if @xmath583 , then the superlevel set @xmath584 is at least at distance @xmath22 from the upper constraint @xmath585 delimiting @xmath100 .
in the course of this proof we will often indicate balls and cubes without any explicit mention of their center .
thus , @xmath586 will be for instance used to denote a ball not necessarily centered at the origin , in contrast with the notation adopted in the rest of the paper .
we claim that @xmath587 let @xmath588 be given and suppose that for any ball @xmath589 of radius @xmath541 compactly contained in @xmath100 , there exists a point @xmath590 such that @xmath591 .
if we show that @xmath102 is less or equal to a universal value @xmath103 , claim would then be true .
let @xmath592 be the only integer for which @xmath593 take a point @xmath578 lying on the hyperplane @xmath594 and consider the ball @xmath595 .
by , we have that @xmath596 , with @xmath597 consequently , we may apply the bound in to deduce that @xmath598 let now @xmath2 be a cube of sides @xmath599 , centered at @xmath263 .
of course , @xmath600 .
it is easy to see that @xmath2 may be partitioned ( up to a negligible set ) into a collection @xmath601 of cubes with sides of length @xmath602 , parallel to those of @xmath2 .
moreover , we denote with @xmath603 the ball of radius @xmath541 having the same center of @xmath604 .
see figure [ cubepart ] . into the subcubes @xmath604 s and the concentric balls @xmath605 s .
] in view of our starting assumption , for any @xmath606 there exists a point @xmath607 at which @xmath608 .
we claim that @xmath609 . indeed , setting @xmath610 , by we get @xmath611 for any @xmath612 . hence , is established . furthermore , since @xmath613 , we have @xmath614 by combining and , recalling we compute @xmath615 with @xmath616 universal . on the other hand ,
implies that @xmath617 for some universal @xmath618 .
note that the energy estimate may be applied to the ball @xmath586 thanks to . comparing the last two inequalities and recalling ,
we find out that @xmath410 can not be greater than a universal constant . by
, the same holds true for the quotient @xmath102 and hence follows .
now , we want to rule out the possibility of @xmath99 being greater or equal to @xmath358 on @xmath619 , thus showing that @xmath620 in @xmath619 . by contradiction , assume that @xmath621 in view of proposition [ ubirk ] the set @xmath622 has the birkhoff property with respect to @xmath52 .
hence , thanks to and proposition [ fatbirk ] , this superlevel set contains the half - space @xmath623 . since @xmath624
, we then deduce that the distance of @xmath625 from the lower constraint @xmath626 is at least @xmath22 . accordingly
, if we assume without loss of generality that @xmath627 , then the translation @xmath628 belongs to @xmath570 ( recall definition ) .
but then , the periodicity assumptions - imply that @xmath629 and thus @xmath630 .
being @xmath99 the minimal minimizer , we conclude that @xmath631 by iterating this inequality we then find that @xmath632 i.e. , @xmath633 a.e . in @xmath39 , in contradiction with the fact that , by construction , @xmath620 in @xmath634 . as a result ,
@xmath620 on the ball @xmath619 .
the proof then finishes by applying once again propositions [ ubirk ] and [ fatbirk ] to the sublevel set @xmath635 .
[ upperunconst ] if @xmath583 , then @xmath636 , for any @xmath637 .
fix @xmath583 and @xmath638 $ ] . by applying proposition [ dist1prop ] to the minimal minimizer @xmath639
, we find that @xmath640 a.e . in the half - space @xmath634 .
hence , @xmath641 and @xmath642 , by the minimization properties of @xmath99 . on the other hand , clearly @xmath643 , so that we also have @xmath644 .
thus , both @xmath99 and @xmath639 belong to @xmath645 and , consequently , they define the same function . by iteration
, the arguments extends to any @xmath637 .
this corollary essentially tells that when @xmath102 is greater than the universal constant @xmath103 found in proposition [ dist1prop ] , then the upper constraint @xmath585 becomes immaterial for the minimal minimizer @xmath99 , which starts attaining values below the threshold @xmath646 well before touching that constraint .
the next result shows that a similar behavior also occurs with the lower constraint @xmath626 , thus proving that the minimal minimizer is unconstrained .
recalling the notation introduced right above lemma [ min2lem ] , we state the following [ unconstrprop ] if @xmath583 , then @xmath99 is unconstrained , that is @xmath647 , for any @xmath637 .
let @xmath455 be such that @xmath648 .
given @xmath649 , we consider its translation @xmath650 . by corollary [ upperunconst ] , it then holds @xmath651 .
the thesis then follows , as @xmath652 by - . to conclude the subsection
, we combine the previous proposition with the results of subsection [ cptsubsec ] and obtain that @xmath99 is indeed a class a minimizer .
if @xmath583 , then @xmath99 is a class a minimizer of the functional @xmath50 .
let @xmath67 be any given bounded subset of @xmath39 .
take @xmath637 and @xmath653 large enough to have @xmath67 compactly contained in the quotient @xmath654 ( recall notation ) . by virtue of proposition [ u = umprop ] we know that @xmath655 is the minimal minimizer of the class @xmath656 .
on the other hand , proposition [ unconstrprop ] yields @xmath657 . recalling the terminology introduced in subsection [ doubsubsec ]
, we then have @xmath658 with @xmath659 .
hence , @xmath660 and proposition [ cptminprop ] implies that @xmath99 is a local minimizer of @xmath50 in @xmath67 .
here we finish the proof of theorem [ mainthm ] for kernels satisfying hypothesis , by extending the results obtained in the previous subsections to irrational vectors @xmath52 .
this task is accomplished by means of an approximation argument , whose most technical steps are inspired by ( * ? ? ?
* section 7 ) .
fix @xmath94 and consider a sequence @xmath661 converging to @xmath52 . denote with @xmath662 the class a minimizer corresponding to @xmath663 , given by our construction .
we recall that @xmath664 , with @xmath665 in @xmath39 , and that @xmath666 \right\},\ ] ] for any @xmath287 .
moreover , by corollary [ minminholdercor ] , the @xmath662 s are uniformly bounded in @xmath667 , for some universal @xmath133 . hence , by ascoli - arzel theorem there exists a subsequence of @xmath405 - which , without loss of generality , we will assume to be @xmath405 itself - converging to some continuous function @xmath10 , uniformly on compact subsets of @xmath39 . of course , @xmath668 in @xmath39 .
also , passes to the limit , so that the same inclusion holds with @xmath10 and @xmath52 replacing @xmath662 and @xmath663 .
in order to finish the proof of theorem [ mainthm ] we therefore only need to show that @xmath10 is a class a minimizer of @xmath50 . to do this ,
let @xmath669 be a fixed number : we claim that @xmath10 is a local minimizer of @xmath50 in @xmath670 , that is @xmath671 and @xmath672 observe that , going back to remark [ restrmk ] , this implies that @xmath10 is a class a minimizer . to see that is true , we first apply proposition [ enestprop ] to @xmath662 and
obtain that @xmath673 for some constant @xmath674 independent of @xmath407 .
furthermore , by fatou s lemma , we know that @xmath675 for any @xmath676 $ ] , and thus , in particular , @xmath677 recall that @xmath678 is monotone non - decreasing with respect to set inclusion
. now , we deal with the limit on the right - hand side of .
let @xmath679 be the sequence of positive real numbers given by @xmath680 clearly , @xmath681 converges to @xmath114 and we may also assume @xmath682 for any @xmath407 . take @xmath683 to be a cut - off function satisfying @xmath684 in @xmath39 , @xmath685 in @xmath670 , @xmath686 and @xmath687 in @xmath39 .
let @xmath104 be as in and suppose without loss of generality that @xmath688 .
we are also allowed to assume @xmath689 , formula being trivially satisfied otherwise . as a consequence of this , , and the boundedness of @xmath10 and @xmath104 , we have that @xmath690 .
we define @xmath691 and @xmath692 notice that @xmath693 in @xmath670 and @xmath694 in @xmath695 .
accordingly , @xmath439 is an admissible competitor for @xmath662 in @xmath696 and thus @xmath697 in view of the minimizing property of @xmath662 .
furthermore , @xmath439 converges to @xmath77 uniformly on compact subsets of @xmath39 and , in particular , @xmath698 fix a number @xmath699 and take @xmath407 big enough to have @xmath700 .
we address the right - hand side of .
concerning its kinetic part , we decompose the domain of integration @xmath701 as @xmath702 where , up to sets of measure zero , @xmath703 see figure [ defdecomp ] .
also set @xmath704 and observe that , analogously to , it holds @xmath705 first , we deal with the tail term of @xmath50 , which corresponds to @xmath706 .
note that @xmath706 may be written as the union of @xmath707 and @xmath708 . by , it is clearly enough to study what happens inside the first set of this union . given @xmath709 and @xmath710 , we have @xmath711 moreover , @xmath712 |y|$ ] and thus @xmath713 using , for any @xmath714 and @xmath710 we get @xmath715 for some constant @xmath314 independent of @xmath407 . recalling that @xmath439 converges pointwise to @xmath77 in @xmath39 , by the dominated convergence theorem
we conclude that @xmath716 now , we focus on @xmath717 . by the triangle inequality , for any @xmath718
we write @xmath719 where we also used and that @xmath720 . hence , taking advantage of and the regularity of @xmath721 , @xmath722^{\frac{1}{2 } } \\ & \hspace{20pt } { \leqslant}\left [ 4 \lambda \iint_{e_{j , \delta } } \frac{dx dy}{|x - y|^{n - 2 + 2 s } } \right]^{\frac{1}{2 } } + \left [ \iint_{e_{j , \delta } } \left| u(x ) - u(y ) \right|^2 k(x , y ) \ , dx dy \right]^{\frac{1}{2 } } \\ & \hspace{20pt } \quad + \left [ \iint_{e_{j , \delta } } \left| u_j(x ) - u_j(y ) \right|^2 k(x , y ) \ , dx dy \right]^{\frac{1}{2 } } + \left [ \iint_{e_{j , \delta } } \left| \varphi(x ) - \varphi(y ) \right|^2 k(x , y ) \ , dx dy \right]^{\frac{1}{2}}.\end{aligned}\ ] ] note that the arguments of the first , second and fourth integrals on the right - hand side above are integrable on the set @xmath723 , which contains @xmath717 . thus , by the absolute continuity of the lebesgue measure in @xmath724 , it follows that those integrals go to zero , as @xmath725 ( observe in this regard that @xmath726 ) .
moreover , in view of , we conclude that @xmath727 for some sequence @xmath728 of positive real numbers such that @xmath729 we are left with the term involving @xmath730 .
we recall that @xmath693 in @xmath670 , so that @xmath731 therefore , we just need to examine the complement @xmath732 and thus , by symmetry , the region @xmath733 only . letting @xmath734 and @xmath735 , by we have @xmath736 then , by the definition of @xmath721 and @xmath737 we get @xmath738^{\frac{1}{2 } } \\ & \hspace{30pt } { \leqslant}\left [ \int_{b_r } \int_{b_{r + \delta } \setminus b_r } \left| v(x ) - v(y ) \right|^2 k(x , y ) \ , dx dy \right]^{\frac{1}{2 } } + \left [ \int_{b_r } \int_{b_{r + \delta } \setminus b_r } \frac{4 \lambda \ , dx dy}{|x - y|^{n - 2 + 2 s } } \right]^{\frac{1}{2 } } \\ & \hspace{30pt } { \leqslant}\left [ \int_{b_r } \int_{b_{r + \delta } \setminus b_r } \left| v(x ) - v(y ) \right|^2 k(x , y ) \ , dx dy \right]^{\frac{1}{2 } } + c \left| b_{r + \delta } \setminus b_r \right|^{\frac{1}{2}},\end{aligned}\ ] ] for some constant @xmath314 independent of @xmath407 and @xmath739 . recalling , we may thence conclude that there exists a function @xmath740 for which @xmath741 and @xmath742 for any @xmath407 big enough .
observe now that for the potential term of @xmath50 we may simply estimate @xmath743 taking advantage of decomposition on both sides of and using inequalities , , we write @xmath744 which in turn simplifies to @xmath745 if we exploit the fact that @xmath746 and recall , , , by taking the limit in @xmath407 in the previous formula we find @xmath747 putting together this last inequality with , we finally obtain @xmath748 then , follows from the arbitrariness of @xmath739 and .
we conclude that @xmath10 is a class a minimizer of @xmath50 .
in this section we complete the proof of theorem [ mainthm ] , by extending the results of section [ s>1/2sec ] to kernels which do not necessarily satisfy condition .
this can be done in consequence of the fact that none of the estimates established there involve any of the parameters appearing in .
this enables us to perform a limit argument analogous to that of subsection [ irrsubsec ] .
let @xmath17 be a kernel satisfying , and only .
given any monotone increasing sequence @xmath749 which diverges to @xmath750 , we set @xmath751}(|x - y| ) \quad \mbox{for any } x , y \in { \mathbb{r}}^n.\ ] ] notice that the new truncated kernel @xmath752 still satisfies hypotheses , and .
moreover , @xmath752 clearly fulfills the additional requirement with @xmath753 .
let @xmath754 be the energy functional corresponding to @xmath752 .
for a fixed direction @xmath51 , let @xmath662 be the plane - like class a minimizer for @xmath754 directed along @xmath52 .
the existence of such minimizers is a consequence of section [ s>1/2sec ] , as @xmath752 verifies .
it holds @xmath755 \right\},\ ] ] for a universal value @xmath89 .
furthermore , @xmath665 in @xmath39 and , in view of corollary [ minminholdercor ] , @xmath756 , for some @xmath757 $ ] and @xmath314 .
we highlight the fact that we can choose @xmath103 , @xmath143 and @xmath758 to be independent of @xmath407 , since each @xmath752 satisfies with the same structural constants .
accordingly , by ascoli - arzel theorem @xmath405 converges , up to a subsequence , to a continuous function @xmath10 , uniformly on compact subset of @xmath39 .
observe that @xmath10 satisfies .
also , if @xmath52 is rational then each @xmath662 is @xmath64-periodic and , consequently , so is @xmath10 . to prove that @xmath10 is a class a minimizer , fix @xmath669 and consider a perturbation @xmath104 , with @xmath759 .
we know that @xmath760 on the one hand , a simple application of fatou s lemma implies that @xmath761 on the other hand , following the strategy presented in subsection [ irrsubsec ] it is not hard to see that we also have @xmath762 it follows that @xmath10 is a class a minimizer of @xmath50 and the proof of theorem [ mainthm ] is therefore complete .
in this first appendix we enclose a couple of lemmata which cover some technical aspects that we faced throughout the paper . we begin with an observation on the necessity of hypothesis for the validity of the computations of section [ s>1/2sec ] .
we refer to subsection [ minpersubsec ] , in particular , for the notation employed in the statement
. [ rapidneclem ] assume that @xmath17 is a measurable kernel satisfying @xmath763,\ ] ] for some @xmath764 .
then , given any two real numbers @xmath361 , it holds @xmath765 for any @xmath365 .
consequently , @xmath766 on @xmath364 . of course
, we may take @xmath377 , @xmath378 and @xmath379 .
then , @xmath767 \quad \mbox{and } \quad { { \widetilde{{\mathbb{r}}}^n}}\cap \ { \omega \cdot x { \geqslant}b \ } = [ 0 , 1]^{n - 1 } \times [ 1 , + \infty).\ ] ] under these conditions , the left hand side of is controlled from below by @xmath768^{n - 1 } \times [ 1 , +
\infty ) } \left ( \int_{\left ( { \mathbb{r}}^{n - 1 } \times ( -\infty , 0 ] \right ) \setminus b_{\bar{r}}(x ) } \frac{\left| u(x ) - u(y ) \right|^2}{|x - y|^{n + \beta } } \ , dy \right ) dx.\ ] ] since @xmath769 , it follows that for any @xmath770 such that @xmath771 and @xmath772 , @xmath773 hence , @xmath774^{n - 1 } \times [ \bar{r } + 1 , + \infty ) } \left ( \int_{{\mathbb{r}}^{n - 1 } \times ( -\infty , 0 ] } \frac{dy}{|x - y|^{n + \beta } } \right ) dx.\ ] ] arguing as in the proof of lemma , it is easy to check that @xmath775 } \frac{dy}{|x - y|^{n + \beta } } = c x_n^{- \beta},\ ] ] for some constant @xmath776 independent of @xmath44 .
accordingly , @xmath777 since @xmath778 .
the thesis then follows .
next is a lemma that ensures the finiteness of the integral appearing on the right - hand side of , in subsection [ cptsubsec ] .
[ intfinlem ] let @xmath688 have support compactly contained in @xmath497 , in the sense of footnote [ cptfoot ] at page
. denote with @xmath106 the @xmath460-periodic extension to @xmath39 of @xmath502 .
then , the integral @xmath779 is finite .
assume for simplicity that @xmath377 and @xmath780 . with this choices ,
we identify @xmath65 with its fundamental region @xmath781 .
we split the domain of integration of as @xmath782 with @xmath783 we first deal with the integral involving the region @xmath784 . in view of the hypothesis on the support of @xmath104
, we have @xmath785 for some @xmath156 .
therefore , we estimate @xmath786 \ } } \frac{dx dy}{|x - y|^{n + 2 s } } \\ & { \leqslant}\| \varphi \|_{l^\infty({\mathbb{r}}^n)}^2 \delta^{-
n - 2 s } \left [ 2 \sqrt{n - 1 } \right]^{n - 1 } ( b - a)^2,\end{aligned}\ ] ] where we also used the fact that @xmath787 is contained in the strip @xmath788 $ ] . on the other hand ,
if @xmath509 and @xmath789 , then @xmath790 and @xmath791 . hence , @xmath792 and thus @xmath793 for some dimensional constant @xmath794 .
this concludes the proof .
we discuss here some separability properties of the subsets of the space @xmath796 of locally @xmath797-summable functions , for @xmath798 .
while the literature on the standard lebesgue spaces @xmath799 is large and exhaustive , @xmath796 classes are somehow rarely considered as functional spaces . as we have not been able to find precise references for the few facts about @xmath796 that we took advantage of in proposition [ minminismin ] ,
we provide directly here a proof of such results .
first , with the aid of the following proposition , we endow @xmath796 with a separable metric made up on the exhaustion of balls @xmath800 of @xmath39 .
[ lplocsepprop ] let @xmath798 and define @xmath801 for any @xmath802 .
then , @xmath803 is a separable metric space .
it is straightforward to check that @xmath804 is a metric .
thus , we only focus on the proof of the separability .
since @xmath799 is separable , we may select a sequence @xmath434 which is dense in this space .
we claim that @xmath405 is dense in @xmath803 , too . for a general function @xmath805 and any @xmath274 , write @xmath806 thus , @xmath807 .
fix now @xmath808 . for any @xmath274 ,
let @xmath809 be such that @xmath810 of course , such @xmath809 exists in view of the density of @xmath405 in @xmath799 .
moreover , we can choose @xmath811 to be increasing in @xmath410 , so that @xmath812 is a subsequence of @xmath405 . for any @xmath410
, we then have @xmath813 and hence @xmath814 as @xmath225 .
it follows that @xmath405 is dense in @xmath803 .
now that we have established this property , we can proceed to the kind of separability we are most interested in .
[ aesepprop ] let @xmath798 .
then , any subset @xmath815 of @xmath796 is separable with respect to pointwise a.e . convergence .
that is , there exists a sequence @xmath816 such that , for any @xmath817 , a subsequence @xmath812 of @xmath405 converges to @xmath10 a.e . in @xmath39 .
first of all , we point out that if @xmath818 in @xmath803 , then @xmath439 also converges to @xmath77 in @xmath819 , for any @xmath274 .
indeed , @xmath820 as @xmath821 and thence the claim follows by noticing that , given a sequence of non - negative real numbers @xmath822 and @xmath823 , @xmath824 as @xmath725 . after this preliminary observation
, we can now head to the actual proof of the proposition .
note that , since it is a subset of @xmath796 , @xmath815 is itself a separable metric space with respect to @xmath804 .
this follows by applying proposition [ lplocsepprop ] and , for instance , proposition 3.25 of @xcite .
let then @xmath816 be a dense sequence .
fixed an element @xmath817 , by the initial remark we know that there exists a subsequence @xmath825 of @xmath405 such that @xmath826 in @xmath819 , for any @xmath274 .
we perform a diagonal argument in order to extract a further subsequence @xmath827 from @xmath825 which converges to @xmath10 a.e . in @xmath39 .
since @xmath825 converges to @xmath10 in @xmath828 , we may select a subsequence @xmath829 from @xmath825 which converges to @xmath10 a.e . in @xmath154 . then
, @xmath830 still converges to @xmath10 in @xmath831 , as it is a subsequence of @xmath825 , and hence there exists another subsequence @xmath832 of @xmath829 converging to @xmath10 a.e . in @xmath833 .
we keep extracting nested subsequences and obtain , for any @xmath410 , a subsequence @xmath834 converging to @xmath10 a.e . in @xmath835 .
set @xmath836 for any @xmath287 .
this new sequence @xmath827 is eventually a subsequence of each of the previous sequences .
thus , it converges to @xmath10 a.e . in @xmath835 , for any @xmath274 , that is a.e . in @xmath39
the authors are indebted to professor ovidiu savin for offering several valuable insights and professor hans triebel for some keen comments on a previous version of the paper .
m. giaquinta , l. martinazzi , _ an introduction to the regularity theory for elliptic systems , harmonic maps and minimal graphs _ ,
appunti , scuola normale superiore di pisa ( nuova serie ) [ lecture notes .
scuola normale superiore di pisa ( new series ) ] , vol .
11 , edizioni della normale , pisa , 2012 . | we consider a non - local phase transition equation set in a periodic medium and we construct solutions whose interface stays in a slab of prescribed direction and universal width .
the solutions constructed also enjoy a local minimality property with respect to a suitable non - local energy functional .
elliptic pde s and symmetry of interfaces and layers for odd nonlinearities _ ) and the prin grant 201274fyk7 ( _ critical point theory and perturbative methods for nonlinear differential equations _ ) ] \(1 ) dipartimento di matematica `` federigo enriques '' + universit degli studi di milano , + via saldini 50 , i-20133 milano ( italy ) .
+ \(2 ) laboratoire aminois
de mathmatique fondamentale et applique + umr cnrs 7352 , universit picardie `` jules verne '' + 33 rue st leu , f-80039 amiens ( france ) .
+ \(3 ) weierstra institut fr angewandte analysis und stochastik + mohrenstrae 39 , d-10117 berlin ( germany ) .
e - mail addresses : [email protected] , [email protected] |
let @xmath0 be a compact hyperbolic riemann surface of genus @xmath1 .
a systole of @xmath0 is a shortest simple closed geodesic .
we denote by @xmath2 its length .
let @xmath3 be the value @xmath9 due to mumford s generalization of mahler s compactness theorem in @xcite , this supremum is a maximum .
the exact value of @xmath3 is only known for @xmath10 .
we also have the following estimates from @xcite : + there exists a universal , but unknown , constant @xmath11 such that for all genera @xmath1 @xmath12 here the upper bound follows from a simple area argument ( see @xcite ) .
the first account of a lower bound is due to buser in @xcite . here
an infinite sequence of surfaces with lower bound of order @xmath13 is constructed .
it was then shown in @xcite that there is a infinite sequence of genera @xmath14 with @xmath15 where @xmath16 is constant .
more examples of families of compact hyperbolic riemann surfaces satisfying the above inequality can be found in @xcite and @xcite .
+ the study of surfaces whose systole length is a global or local maximum in the moduli space @xmath17 of compact hyperbolic riemann surfaces of genus @xmath1 was initiated by schmutz ( see @xcite , @xcite and @xcite ) . here
he also provides a number of interesting properties of these surfaces .
the characterization of maximal surfaces was continued in @xcite , @xcite , @xcite and @xcite .
here it was shown that * a ( locally ) maximal surface of genus @xmath5 has at least @xmath18 systoles @xcite , @xcite .
* there is only a finite number of maximal surfaces of genus @xmath5 @xcite , @xcite . *
the systole function @xmath4 is a topological morse function on the moduli space @xcite . *
all systoles of maximal surfaces are non - separating ( see @xcite , * claim * on p. 336 ) .
an open question is , whether @xmath7 is a monotonously increasing function with respect to the genus .
though we can not prove or disprove this result , we can at least show the following intersystolic inequalities : let @xmath3 be the maximal value that @xmath4 can attain among the compact hyperbolic riemann surfaces of genus @xmath1 . 1 .
@xmath19 for @xmath20 .
2 . @xmath21 3 . if @xmath22 , then @xmath23 .
[ thm : intermsys ] this result is obtained by cutting and pasting maximal surfaces to construct compact hyperbolic riemann surfaces with large systoles . here
the main tool is * lemma [ thm : sys_coll ] * , a collar lemma for systoles . as a result
we obtain from * theorem [ thm : intermsys]-1 * : if @xmath0 is a compact riemann surface of genus @xmath5 , such that @xmath24 , then @xmath25 furthermore , by construction , we obtain a continuous parameter family @xmath26^k}$ ] of compact hyperbolic riemann surfaces of genus @xmath27 , such that @xmath28 this shows that though the systoles of these surfaces are large , these surfaces can not be maximal , as this would be a contradiction to the finiteness of the number of these surfaces .
though * theorem [ thm : intermsys]-1 * without a sharp inequality could be obtained by constructing a normal cover of a maximal surface in a similar fashion ( see * note [ thm : kcover_note ] * ) this shows that there exist non - trivial examples of continuous families of surfaces with large systoles .
+ in the following table tab .
[ tab : rs_large_sys ] we give a summary of compact hyperbolic riemann surfaces of genus @xmath29 with maximal known systoles thus providing a reference and benchmark for further studies .
most of these are constructed from the examples presented in @xcite,@xcite , @xcite and @xcite using * theorem [ thm : intermsys]*. .compact hyperbolic riemann surfaces with large systoles of genus @xmath29 . [ cols="^,^,^,^,^,^ " , ] it follows furthermore from the known examples in genus @xmath30 and @xmath31 and * theorem [ thm : intermsys]-1 * : let @xmath3 be the maximal value that @xmath4 can attain among the compact hyperbolic riemann surfaces of genus @xmath1 . then
: @xmath32 in section 3
we use the same methods to derive similar inequalities for non - compact hyperbolic surfaces with _ cusps _ ( see @xcite , * example 1.6.8 * for an exact definition ) .
let @xmath0 be a hyperbolic riemann surface of signature @xmath33 , i.e. of genus @xmath5 and with @xmath34 cusps .
let @xmath35 be the value @xmath36 as in the case of the compact hyperbolic surfaces this supremum is a maximum due to mahler s compactness theorem .
the known bounds for the value of @xmath35 are the following : for @xmath37 and @xmath38 @xmath39 the upper bound follows again from an area argument ( see @xcite ) , whereas the lower bound is due to the collar lemma for riemann surfaces ( see @xcite , * theorem 4.1.1 * ) .
more refined estimates have been obtained recently in @xcite .
+ heuristically speaking , if we fix the genus @xmath5 and increase the number @xmath34 of cusps continuously , then some of the cusps must move closer together and there will be a simple closed geodesic of bounded length surrounding two or several cusps . such a geodesic is separating the surface into two parts . indeed it can be shown that : * let @xmath6 be a maximal hyperbolic riemann surface of signature @xmath33 , where @xmath40 . then all systoles of @xmath6 are separating .
( @xcite , * proposition 3.1 * ) .
this will be important in the statement of the following theorem .
let @xmath6 be a maximal non - compact hyperbolic riemann surface of signature @xmath41 , where @xmath42 and @xmath38 that has a separating systole .
let @xmath43 be its systole length . then @xmath44 . furthermore
: 1 . @xmath45 .
2 . @xmath46 3 .
@xmath47 [ thm : intermsys2 ] here the lower bound @xmath44 is due to @xmath48 .
we furthermore construct normal covers of maximal non - compact hyperbolic surfaces of signature @xmath33 and show that the systole length does not decrease in these surfaces .
this enables us to prove : let @xmath35 be the maximal value that @xmath4 can attain among the non - compact hyperbolic riemann surfaces of signature @xmath33 , where @xmath42 and @xmath49
. then @xmath50 .
[ thm : intermsys3 ]
the second author has been supported by the alexander von humboldt foundation .
we would like to thank guillaume bulteau for helpful discussions .
by abuse of notation we will denote the length of a geodesic arc by the same letter as the arc itself in the following sections . in this section
we treat the case of compact hyperbolic surfaces and a surface of signature @xmath33 is a compact surface of genus @xmath5 with @xmath34 disjoint boundary components , each of which is a smooth simple closed geodesic .
we will first prove a collar lemma for systoles .
then we will present the construction of a certain normal cover of riemann surfaces . in this cover
the injectivity radius does not decrease .
finally this construction will be used to show * theorem [ thm : intermsys]*. + + let @xmath51 be a simple closed geodesic on @xmath0 .
let @xmath52 be the supremum of all @xmath53 , such that the geodesic arcs of length @xmath53 emanating perpendicularly from @xmath51 are pairwise disjoint .
we define a _ collar _ @xmath54 around @xmath51 of width @xmath55 by @xmath56 we call a collar @xmath54 of width @xmath57 the _ maximal collar _ of @xmath51 .
we call a _ half - collar _ @xmath58 of width @xmath53 one of the two parts of a collar @xmath54 that we obtain by cutting @xmath54 along @xmath51 .
+ we first show the following collar lemma for systoles of compact surfaces .
let @xmath59 be a systole of a compact hyperbolic riemann surface @xmath0 of genus @xmath1 .
then the maximal collar @xmath60 of @xmath59 has width @xmath61 [ thm : sys_coll ] let @xmath59 be a systole of a maximal compact hyperbolic riemann surface @xmath6 of genus @xmath1 .
the closure @xmath62 of the maximal collar of @xmath59 self - intersects in a point @xmath63 .
there exist two geodesic arcs @xmath64 and @xmath65 of length @xmath66 emanating from @xmath59 and perpendicular to @xmath59 having the endpoint @xmath63 in common .
these two arcs form a smooth geodesic arc @xmath67 .
the endpoints of @xmath67 on @xmath59 divide @xmath59 into two parts .
we denote these two arcs by @xmath68 and @xmath69 .
let without loss of generality @xmath68 be the shorter arc of these two .
we have that @xmath70 let @xmath71 be the simple closed geodesic in the free homotopy class of @xmath72 .
we have that @xmath73 now if @xmath74 then it follows from this inequality that @xmath75 . a contradiction to the minimality of @xmath59 .
hence each systole @xmath59 in a compact hyperbolic riemann surface @xmath0 has a collar @xmath76 of width @xmath77 which is embedded in @xmath0 . from this fact we also obtain an upper bound for the length of a systole via an area argument .
however , this estimate is not better than the one given in the introduction ( see inequality ( [ eq : log_bound ] ) ) .
+ the following construction of a normal cover can be used to provide a version of * theorem [ thm : intermsys]-1 * , where the inequality is not sharp .
let @xmath0 be a compact hyperbolic riemann surface of genus @xmath1 .
there is a normal cover @xmath78 of order @xmath79 and of genus @xmath27 , such that @xmath80 [ thm : kcover_note ] (.30*.86 ) @xmath81 + (.68*.86 ) @xmath82 + (.68*.11 ) @xmath83 + (.30*.11 ) @xmath84 + (.25*.49 ) @xmath85 + (.46*.84 ) @xmath86 + (.67*.49 ) @xmath87 + (.46*.12 ) @xmath88 + let @xmath0 be a compact hyperbolic riemann surface of genus @xmath1 and let @xmath59 be a non - separating simple closed geodesic of @xmath0 .
let @xmath89 be a partition of @xmath0 into surfaces of signature @xmath90 , or _
y - pieces _ , such that @xmath59 is the boundary curve of a y - piece .
let @xmath91 be a set of _ fenchel - nielsen coordinates _ corresponding to @xmath89 ( see @xcite , p. 27
- 30 ) and let @xmath92 be the _ twist parameter _ at @xmath59 .
+ to prove our note we construct a new surface @xmath78 of genus @xmath27 from @xmath0 such that @xmath80 to this end we cut open @xmath0 along @xmath59 and call the surface obtained this way @xmath93 .
the signature of @xmath93 is @xmath94 .
let @xmath95 and @xmath96 be the two boundary geodesics of @xmath93 .
we recall that by gluing @xmath93 with twist parameter @xmath92 along @xmath95 and @xmath96 we obtain @xmath0 .
+ take @xmath79 copies @xmath97 of the surface @xmath93 and let @xmath98 and @xmath99 be the boundary geodesics of @xmath100 , such that @xmath98 is the copy of @xmath95 and @xmath99 is the copy of @xmath96 .
we identify the boundaries of the different @xmath97 in the following way @xmath101 choosing the @xmath79 twist parameters @xmath102 , such that @xmath103 we denote the surface of genus @xmath27 obtained according to this pasting scheme as @xmath104 we denote by @xmath105 the image of @xmath106 in @xmath78 . for @xmath107 ,
let @xmath108 be the natural isometry on the interior of the @xmath100 .
as all twist parameters in the pasting are chosen to be @xmath92 the @xmath109 can be extended to an isometry @xmath110 and we have that @xmath111 then @xmath112 is the cyclic group of order @xmath79 generated by @xmath113 and @xmath78 is a normal cover of @xmath0 .
+ now it is well - known that the injectivity radius @xmath114 of a riemannian manifold @xmath115 does not decrease in a normal cover .
but in the case of a compact hyperbolic riemann surface @xmath116 we have by @xcite , * theorem 4.1.5 * that @xmath117 .
hence @xmath118 this concludes the proof of our note .
* proof of theorem [ thm : intermsys ] * + + _ 1 .
_ + + let @xmath6 be a globally maximal compact hyperbolic riemann surface of genus @xmath1 . to prove the first part of the theorem we construct a new surface @xmath116 of genus @xmath27 from @xmath6 such that @xmath119 let @xmath59 be a systole of @xmath6 .
as all systoles of maximal surfaces are non - separating ( see @xcite , * claim * on p. 336 ) , @xmath59 is a non - separating simple closed geodesic . we proceed as in the proof of * note [ thm : kcover_note ] * , setting @xmath120 and @xmath121 .
however , this time we choose the @xmath79 twist parameters @xmath122\ ] ] in the pasting ( [ eq : paste1 ] ) freely to obtain the closed surface @xmath116 of genus @xmath27 . recall that @xmath105 is the image of @xmath106 in @xmath116 . due to our construction and * lemma [ thm : sys_coll ] * each @xmath105 has an embedded collar @xmath123 of width @xmath77 ( see fig .
[ fig : paste ] ) .
+ we now show that any simple closed geodesic @xmath51 in @xmath116 has length bigger than or equal to @xmath59 .
therefore we distinguish two cases : either @xmath51 intersects at least one of the @xmath124 transversally or not .
+ consider the first case .
if @xmath51 intersects one of the @xmath124 , say @xmath125 , then , due to the geometry of hyperbolic cylinders , @xmath51 traverses @xmath126 .
now , to be a closed curve @xmath51 has to traverse at least two times the same cylinder @xmath126 or has to traverse @xmath126 and a different cylinder @xmath127 .
therefore its length is bigger than @xmath128 . in this case
we have that @xmath129 which implies that @xmath130 . + in the second case , a simple closed geodesic that intersects none of the @xmath124 transversally
is either contained in the interior of one of the @xmath97 or is one of the @xmath124 . in any case
we have that @xmath129 thus @xmath131 .
+ hence in any case we have that @xmath132 .
+ setting the twist parameters @xmath133 arbitrarily , we see that @xmath116 belongs to a continuous family @xmath26^k}$ ] of compact hyperbolic riemann surfaces of genus @xmath27 , such that @xmath28 this shows that , though the systoles of the surfaces in this family are large , these can not be maximal , as this would be a contradiction to the finiteness of the number of these surfaces . hence @xmath134 . + + _ 2 .
@xmath135 . _
+ + we call a surface of signature @xmath136 a _ f - piece_. let @xmath137 and @xmath138 be the boundary geodesics of a f - piece .
let @xmath139 be a surface whose interior systole has maximal length among all f - pieces with boundaries @xmath137 and @xmath138 of equal length @xmath140 denote by @xmath141 the length of the interior systole of @xmath139 .
it was shown in @xcite , * theorem 5.1 . *
that @xmath142 this implies that @xmath143 .
+ to prove the second part of the theorem we construct a new surface @xmath144 of genus @xmath145 from a globally maximal surface @xmath6 of genus @xmath1 and @xmath139 such that @xmath146 to construct @xmath144 , we take the surface @xmath93 of signature @xmath94 obtained by cutting open @xmath6 along a systole @xmath59 and paste the surface @xmath147 with boundary geodesics of length @xmath59 along the two boundary geodesics of @xmath93 .
+ we denote by @xmath148 and @xmath149 the image of the two boundary geodesics of the embedded f - piece in @xmath144 which is isometric to @xmath147 . due to our construction
each @xmath105 has an embedded half - collar @xmath150 of width @xmath77 .
+ we now show that any simple closed geodesic @xmath51 in @xmath144 has length bigger than @xmath151 .
+ therefore we distinguish two cases : either @xmath51 intersects either @xmath148 or @xmath149 transversally or not .
consider the first case .
if @xmath51 intersects one of the @xmath152 , say
@xmath148 , transversally , then , due to the geometry of the half - collar , @xmath51 traverses @xmath153 .
furthermore @xmath51 has to traverse at least two times the same half - collar @xmath153 or has to traverse @xmath153 and @xmath154 to be a closed curve . in this case
its length is bigger than or equal to @xmath155 .
hence in this case we obtain @xmath156 and therefore @xmath157 .
( in fact this inequality is strict , because the part of @xmath51 in the f - piece must have strictly positive length . ) + we now consider the second case . any simple closed geodesic that intersects neither @xmath148 nor @xmath149 transversally is either contained in the interior of @xmath93 , the interior of @xmath147 or is one of the @xmath152 .
here we obtain the lower bound for @xmath51 from the lower bound on @xmath158 .
it follows from equation ( [ eq : sys_f ] ) that @xmath159 thus @xmath160 .
+ in any case we obtain that @xmath161 .
+ as in the previous part , the inequality in * theorem [ thm : intermsys]-2 * is strict due to the fact that the construction does not depend on the twist parameters .
+ + _ 3 . if @xmath22 , then @xmath23 . _
+ + to prove the final statement we take two maximal surfaces @xmath162 and @xmath163 of genus @xmath164 and @xmath165 , respectively . for @xmath166
we cut @xmath167 open along a systole @xmath168 and call the surface obtained in this way @xmath169 .
as @xmath170 we have that @xmath171 .
+ now , we can not directly paste these surfaces together , as the boundary length is different . however , by @xcite , * theorem 1.1 . * , we can construct a comparison surface @xmath172 for @xmath173 of signature @xmath174 , such that * all interior geodesics of @xmath172 are longer than @xmath175 . *
the boundary geodesics @xmath176 and @xmath177 of @xmath172 have length @xmath178 .
we identify the open boundaries of @xmath172 and @xmath179 to obtain the surface @xmath180 of genus @xmath181 .
furthermore the two boundary geodesics @xmath182 and @xmath183 of @xmath179 have both an embedded half - collar of width @xmath184 in @xmath179 . due to the properties of @xmath180 we can apply similar arguments as in the previous case of the surface @xmath144 to show that @xmath185 this concludes the proof of * theorem [ thm : intermsys]*. @xmath186
in this section we denote by a hyperbolic ( riemann ) surface @xmath0 of signature @xmath33 a non - compact hyperbolic riemann surface of genus @xmath5 with @xmath34 cusps and assume that @xmath187 . a surface with a cusp
can also be interpreted as a surface with a degenerated boundary of length zero . in view of this
we will denote in this section by a hyperbolic surface @xmath0 of signature @xmath188 a surface of genus @xmath5 with @xmath34 disjoint boundary components , of which @xmath79 are cusps and of which @xmath189 are smooth simple closed geodesics .
+ + to prove * theorem [ thm : intermsys2 ] * , we will first show a collar lemma for separating systoles of hyperbolic surfaces with cusps .
more specifically , we show that a systole with a large collar exists if a separating systole is intersected by another systole .
contrary to the case of the compact hyperbolic surfaces , this collar lemma applies only in this special situation . with the help of this lemma
we then prove * theorem [ thm : intermsys2]*. then we explain shortly why no such strong collar lemma should hold if the systole of hyperbolic surface with cusps is non - separating .
finally we prove * theorem [ thm : intermsys3 ] * using the construction of a normal cover from section 2 .
let @xmath0 be a non - compact hyperbolic riemann surface of signature @xmath190 , where @xmath187 and @xmath38 .
if @xmath0 has a separating systole that is intersected by another systole , then there exists a separating systole @xmath59 that bounds a surface @xmath191 of signature @xmath192 .
the maximal half - collar @xmath193 of @xmath59 on the side opposite of @xmath191 has width @xmath194 [ thm : sys_coll_cusp ] let @xmath0 be a hyperbolic surface of signature @xmath33 , where @xmath33 satisfies the conditions of the lemma .
we first show that the surface @xmath0 has a separating systole that bounds a y - piece @xmath191 of signature @xmath192 .
therefore we will make use of the following facts : * if @xmath116 is a hyperbolic surface of signature @xmath33 then two systoles intersect at most twice ( see @xcite , * proposition 3.2 * ) . * if @xmath116 is a hyperbolic surface of signature @xmath33 and two systoles intersect twice , then one of them bounds two cusps ( see @xcite , * proposition 3.3 * ) . *
if two systoles in a hyperbolic surface @xmath116 of signature @xmath33 intersect twice , then @xmath195 let @xmath59 be a separating systole in @xmath0 that is intersected by another systole @xmath71 .
now suppose @xmath59 and @xmath71 intersect once .
then by surface topology both are homologically non - trivial and both must be non - separating , which is a contradiction to the fact that @xmath59 is separating . hence @xmath59 and @xmath71 intersect twice .
it follows from @xcite , * proposition 3.3 * that either @xmath59 or @xmath71 bounds a surface @xmath191 of signature @xmath192 .
let without loss of generality @xmath59 be the boundary geodesic of @xmath191 .
we now prove the existence of the half - collar .
+ to this end we cut open @xmath0 along the systole @xmath59 .
as @xmath59 is separating , @xmath0 decomposes into two parts .
let @xmath196 be the part which is a surface of signature @xmath197 .
consider the closure @xmath198 of the half - collar @xmath199 in @xmath196 .
this maximal half - collar of @xmath59 self - intersects in a point @xmath63 .
there exist two geodesic arcs @xmath200 and @xmath201 of length @xmath66 emanating from @xmath59 and perpendicular to @xmath59 having the endpoint @xmath63 in common .
these two arcs form a smooth geodesic arc @xmath202 ( see fig .
[ fig : x12 ] ) .
the endpoints of @xmath202 on @xmath59 divide @xmath59 into two parts .
we denote these two arcs by @xmath68 and @xmath69 .
let without loss of generality @xmath68 be the shorter arc of these two . then @xmath70 let @xmath203 be the simple closed geodesic in the free homotopy class of @xmath204 .
we have that @xmath205 now two cases can occur : either @xmath203 is a simple closed geodesic of non - zero length or @xmath203 has length zero i.e. defines a cusp .
+ + _ case 1 : @xmath203 has non - zero length _ + + in this case , we conclude that @xmath61 otherwise it would follow from equation ( [ eq : omega_cusp_sep ] ) that @xmath203 is a simple closed geodesic of length smaller than the systole . a contradiction .
this settles our claim in the first case .
+ + _ case 2 : @xmath203 has length zero _
+ + it remains the second case , where @xmath203 defines a cusp .
in this case we have to work a bit harder .
let @xmath51 be the simple closed geodesic in the free homotopy class of @xmath206 .
now @xmath51 can not have length zero , because then @xmath0 would be a surface of signature @xmath207 , which we excluded .
therefore let @xmath208 be the surface of signature @xmath209 , whose boundary curves are the cusp defined by @xmath203 , @xmath51 and @xmath59 .
then @xmath210 defines a surface @xmath211 of signature @xmath212 ( see fig [ fig : x12 ] ) .
(.27*.90 ) @xmath211 + (.09*.60 ) @xmath191 + (.09*.36 ) @xmath208 + (.44*.49 ) @xmath59 + (.14*.52 ) @xmath68 + (.32*.45 ) @xmath69 + (.35*.18 ) @xmath51 + (.27*.60 ) @xmath213 + (.35*.59 ) @xmath214 + (.15*.40 ) @xmath215 + (.12*.25 ) @xmath216 + (.30*.30 ) @xmath89 + (.22*.46 ) @xmath217 + (.21*.53 ) @xmath218 + (.20*.60 ) @xmath219 + (.71*.90 ) @xmath220 + (.89*.60 ) @xmath221 + (.89*.36 ) @xmath222 + (.88*.49 ) @xmath59 + (.63*.18 ) @xmath223 + (.85*.30 ) @xmath202 + (.65*.46 ) @xmath224 + (.85*.46 ) @xmath225 + (.72*.45 ) @xmath226 + (.80*.80 ) @xmath227 + (.56*.70 ) @xmath228 + (.59*.53 ) @xmath229 + (.67*.53 ) @xmath230 + (.75*.53 ) @xmath231 + (.70*.33 ) @xmath232 + let @xmath213 be the shortest geodesic arc in @xmath191 connecting @xmath59 with itself . by symmetry
the two endpoints of @xmath213 on @xmath59 divide @xmath59 into two arcs , both of length @xmath151 .
let @xmath233 be a geodesic hyperbolic quadrilateral in @xmath191 with three angles of value @xmath234 that has one side on @xmath213 of length @xmath235 , one side on @xmath59 of length @xmath77 and two sides of infinite length meeting in the point at infinity defined by one of the cusps .
@xmath233 can be interpreted as a degenerated right - angled hyperbolic pentagon .
it follows from the geometry of right - angled pentagons ( see @xcite , p. 454 ) that @xmath236 we have furthermore that @xmath237 as otherwise @xmath238 . + let @xmath239 be a geodesic hyperbolic quadrilateral in @xmath208 with three angles of value @xmath234 that has one side on @xmath202 of length @xmath240 , one side on @xmath59 of length @xmath241 and two sides of infinite length meeting in the point at infinity defined by the cusp .
it follows as for the quadrilateral @xmath233 that @xmath242 now let @xmath243 and @xmath244 be two arcs on @xmath59 that connect the endpoints of @xmath202 and @xmath213 on @xmath59 , such that @xmath245 the existence of these arcs follows from the pigeonhole principle .
let @xmath232 be the shortest simple closed geodesic in @xmath211 in the free homotopy class of @xmath246 .
as @xmath247 , we obtain from ( [ eq : delta2_collar ] ) , ( [ eq : delta1_alpha ] ) and the upper bound on @xmath248 : @xmath249 this is close to our desired result . using more refined estimates for @xmath202 in @xmath211
, one can show that indeed @xmath250 to this end let @xmath89 be a geodesic right - angled pentagon in @xmath208 that has one side on @xmath202 of length @xmath240 , one side on @xmath59 of length @xmath251 and a side on @xmath51 of length @xmath252 .
it follows from the geometry of @xmath89 that @xmath253 with the help of this equation , the equivalent relation in @xmath239 from equation ( [ eq : delta1_alpha ] ) and the fact that @xmath254 , we find the following expression for @xmath202 : @xmath255 this can be deduced using the addition formulas for hyperbolic functions . as @xmath256 is a monotonically increasing function with respect to @xmath51 and as @xmath257
, we obtain : @xmath258 hence in total we obtain ( [ eq : delta1_lemma ] ) by combining ( [ eq : delta1_collar_simp ] ) and ( [ eq : delta1_eta ] ) using the fact that @xmath259 ( see @xcite ) . combining inequality ( [ eq : delta1_lemma ] ) with the inequality from _
case 1 _ we obtain our lemma .
let @xmath6 be a maximal non - compact hyperbolic surface of signature @xmath41 , where @xmath187 and @xmath38 .
if @xmath6 has a separating systole , then there exists a separating systole @xmath59 that bounds a surface @xmath191 of signature @xmath192 .
the maximal half - collar @xmath199 of @xmath59 on the side opposite of @xmath191 has width @xmath194 [ thm : sep_sys_smax ] let @xmath6 be a maximal hyperbolic surface of signature @xmath33 that has a separating systole @xmath59 and whose signature satisfies the conditions of the lemma
. now @xmath59 is intersected by another systole @xmath71 .
this follows from the fact that there is only a discrete number of maximal surfaces in the moduli space @xmath260 of hyperbolic surfaces of signature @xmath33 .
otherwise we could generate a smooth family of maximal surfaces by applying a small twist around @xmath59 , a contradiction .
the remainder of * lemma [ thm : sep_sys_smax ] * then follows from * lemma [ thm : sys_coll_cusp]*. * proof of theorem [ thm : intermsys2 ] * + + let @xmath6 be a maximal surface of signature @xmath33 that has a separating systole and such that @xmath33 satisfies the conditions of the theorem .
we cut open @xmath6 along the systole @xmath59 that bounds a y - piece @xmath191 of signature @xmath192 ( see * lemma [ thm : sep_sys_smax ] * ) .
let @xmath196 be the remainder of @xmath6 of signature @xmath197 .
take two copies , @xmath261 and @xmath262 , of @xmath196 and let @xmath148 be the boundary of @xmath261 and @xmath149 be the boundary of @xmath262 .
+ + _ 1 .
@xmath263 . _
+ + to obtain this inequality , we paste @xmath261 and @xmath262 together with arbitrary twist parameter along the boundaries to obtain a surface @xmath264 of signature @xmath265 . due to the half - collar at @xmath148 and @xmath149 we obtain by similar arguments as in the proof of * theorem [ thm : intermsys ]
* that @xmath266 + _ 2 .
@xmath267 . _ + + as in the proof of * theorem [ thm : intermsys]-2 * , let @xmath139 be a surface of signature @xmath268 whose interior systole @xmath269 has maximal length among all f - pieces with two boundary geodesics of equal length @xmath270 . then @xmath271 let @xmath147 be such a surface where @xmath272 .
we paste together @xmath273 and @xmath262 along the boundaries to obtain a surface @xmath274 of signature @xmath275 .
it follows from the properties of this surface by similar arguments as in the proof of * theorem [ thm : intermsys]-2 * that @xmath276 .
+ + _ 3 . @xmath277 .
_ + + we call a surface of signature @xmath278 a _ q - piece_. let @xmath270 be the boundary geodesic of a q - piece
. let @xmath279 be a surface whose interior systole has maximal length among all q - pieces with a boundary of length @xmath270
. denote by @xmath280 the length of the interior systole of @xmath279 .
it was shown in @xcite and @xcite , * proposition 5.4 * that @xmath281
let @xmath282 be such a surface with boundary length @xmath59 . to prove the final part of the theorem , we paste together @xmath196 and @xmath282 along the boundaries to obtain a surface @xmath283 of signature @xmath284 .
it follows from the properties of this surface by similar arguments as in the proof of * theorem [ thm : intermsys ] * that @xmath285 .
this concludes the proof of * theorem [ thm : intermsys2]*. @xmath186 + + we now explain shortly , why no strong collar lemma should hold for non - separating systoles of non - compact hyperbolic riemann surfaces . to this end
we start by trying to prove such a collar lemma and then show where the arguments fail .
+ let @xmath0 be a hyperbolic surface of signature @xmath33 that has a non - separating systole @xmath59 .
the closure @xmath62 of the maximal collar of @xmath59 self - intersects in a point @xmath286 .
there exist two geodesic arcs @xmath200 and @xmath201 of length @xmath66 emanating from @xmath59 and perpendicular to @xmath59 having the endpoint @xmath286 in common .
these two arcs form a smooth geodesic arc @xmath202 ( see fig .
[ fig : x12 ] ) .
the endpoints of @xmath202 on @xmath59 divide @xmath59 into two parts .
we denote these two arcs by @xmath225 and @xmath226 .
let without loss of generality @xmath225 be the shorter arc of these two .
then @xmath287 suppose that @xmath200 and @xmath201 emanate from the same side of @xmath59 .
+ let @xmath176 be the simple closed geodesic in the free homotopy class of @xmath288 and suppose that @xmath176 is of zero length , i.e. defines a cusp .
let @xmath223 be the curve in the free homotopy class of @xmath289
. then @xmath223 must be a non - separating curve , because otherwise @xmath59 would be separating .
we have that @xmath290 let @xmath222 be the surface of signature @xmath209 defined by the boundary geodesics @xmath223 , @xmath59 and the cusp defined by @xmath176 .
now we assume that @xmath291 . in this case
it can be shown from the hyperbolic geometry of the y - piece @xmath222 ( see equation ( [ eq : delta1_eta0 ] ) and ( [ eq : delta1_eta ] ) replacing @xmath51 by @xmath223 ) that @xmath292 where @xmath293 and @xmath294 if @xmath295 .
+ now we might like to try to show the existence of a large half - collar on the opposite side of @xmath59 .
however , we can not exclude that the situation is symmetric .
more precisely , suppose that there is a y - piece @xmath296 isometric to @xmath222 embedded in @xmath0 and that @xmath222 and @xmath296 have common boundary @xmath59 .
let @xmath297 be the isometry between these two surfaces and set @xmath298 in this case the closure @xmath62 of the maximal collar of @xmath59 on the opposite side of @xmath296 also self - intersects in a point @xmath299 , such that the two geodesic arcs of length @xmath66 emanating from @xmath59 and perpendicular to @xmath59 having the endpoint @xmath300 in common are @xmath301 and @xmath302 . + now let @xmath243 and @xmath244 be two arcs on @xmath59 that connect the endpoints of @xmath202 and @xmath213 on @xmath59 , such that @xmath303 is minimal .
let @xmath232 be the curve in the free homotopy class of @xmath246 .
now @xmath304 one would like to argue that @xmath305 is in fact large due to this inequality , which would lead to a contradiction to inequality ( [ eq : small_collar ] ) .
+ but then again , if the twist parameter in the pasting of @xmath306 is @xmath307 i.e for any point @xmath308 on @xmath59 , @xmath309 these arguments fail .
this can be shown by calculating the length of @xmath232 explicitly . however , for the sake of brevity , we omit the details .
+ though this example does not provide a rigorous proof that a non - separating systole of a non - compact maximal surface does not have a large collar , it leads us to conjecture that this can indeed occur .
+ + we now prove the following lemma , from which follows * theorem [ thm : intermsys3 ] * by passing to a maximal surface .
let @xmath0 be a non - compact hyperbolic riemann surface of signature @xmath33 , where @xmath42 and @xmath310 .
there is a normal cover @xmath78 of order @xmath79 and of signature @xmath311 , such that @xmath80 [ thm : kcover_lem2 ] let @xmath0 be a hyperbolic riemann surface of signature @xmath33 , where @xmath42 and @xmath310 . to prove the theorem we construct a new surface @xmath78 of signature @xmath312 from @xmath0 such that @xmath80 let @xmath59 be a non - separating simple closed geodesic of @xmath0 .
we use the same construction from the proof of * note [ thm : kcover_note ] * to construct a covering surface @xmath78 by cutting open @xmath0 along @xmath59 and pasting @xmath79 copies of the cut surface as described there . recall that @xmath113 is the isometry , such that @xmath313 be the corresponding covering map which is a local isometry .
now , to prove the intersystolic inequality between @xmath314 and @xmath2 we can not use the argument with the injectivity radius . in the non - compact case with cusps we have that @xmath315 , whereas @xmath316 .
+ therefore we will prove the inequality from * lemma [ thm : kcover_lem2 ] * in another way . to this end
we remark the following : * a systole in @xmath0 or @xmath78 is a simple closed geodesic . * for any simple closed geodesic @xmath317 , we have that @xmath318 is a smooth closed geodesic .
the length of @xmath203 is equal to the length of @xmath319 .
now let @xmath320 be a systole of @xmath78 of length @xmath314 .
then @xmath321 is a smooth closed geodesic in @xmath0 of length @xmath322 .
as it is smooth it is the shortest curve in its free homotopy class . now
if @xmath68 defines a cusp , it would follow that @xmath68 has zero length . a contradiction .
hence @xmath68 has non - zero length and it follows that @xmath323 this concludes our proof .
hugo akrout + department of mathematics , universit montpellier 2 + place eugne bataillon , 34095 montpellier cedex 5 , france + e - mail : [email protected]_ + + bjoern muetzel + institute for algebra and geometry , karlsruhe institute of technology + kaiserstrasse 89 - 93 , office 4b-03 ( allianz building ) + 76133 karlsruhe , germany + e - mail : [email protected]_ | let @xmath0 be a compact hyperbolic riemann surface of genus @xmath1 .
we call a _ systole _
a shortest simple closed geodesic in @xmath0 and denote by @xmath2 its length .
let @xmath3 be the maximal value that @xmath4 can attain among the compact riemann surfaces of genus @xmath5 .
we call a _ ( globally ) maximal _ surface @xmath6 a compact riemann surface of genus @xmath5 whose systole has length @xmath3 . in section 2
we use cutting and pasting techniques to construct compact hyperbolic riemann surfaces with large systoles from maximal surfaces .
this enables us to prove several inequalities relating @xmath7 of different genera . as a consequence
we greatly enlarge the number of genera @xmath5 for which the bound @xmath8 is valid . in section 3
we derive similar intersystolic inequalities for non - compact hyperbolic riemann surfaces with cusps .
+ keywords : riemann surfaces , systoles , intersystolic inequalities , maximal surfaces .
+ mathematics subject classification ( 2010 ) : 30f10 , 32g15 and 53c22 . |
since early 90 s many papers have been published with an unusual combination of two keywords - _ blazars _ and _ cosmic - infrared - background _ ( cib ) radiation .
formally , one may argue that there is no apparent link between these two topics .
@xmath0 the blazars constitute a sub - class of agn dominated by highly variable ( several hours or less ) components of broadband ( from radio to gamma - rays ) _ non - thermal emission _ produced in relativistic jets pointing close to the line of sight .
@xmath0 cib is a part of the overall diffuse extragalactic background radiation ( debra ) dominated by _
thermal emission _
components produced by stars and dust , and accumulated over the entire history of the universe .
while the blazars may serve as ideal laboratories for study of mhd structures and particle acceleration processes in relativistic jets , cib carries crucial cosmological information about the formation epochs and history of evolution of galaxies . to a large extent , these two topics are relevant to quite independent areas of modern astrophysics and cosmology . yet
, the current studies of cib and blazars , more specifically the sub - population of blazars emitting tev gamma - rays ( _ tev blazars _ ) , are tightly coupled through the intergalactic ( ig ) absorption of tev radiation by infrared photons of debra . the astrophysical / cosmological importance of this interesting effect ( nikishov , 1962 ; gould and schreder , 1966 ; jelly , 1966 ; stecker et al . , 1992
) was clearly recognized after the discovery of tev @xmath1-rays from two bl lac objects
mkn 421 and mkn 501 ( for review see e.g. vassiliev 2000 ) .
cib basically consists of two emission components produced by stars and partly absorbed / re - emitted by dust during the entire history of evolution of galaxies .
consequently , two distinct bumps in the spectral energy distribution ( sed ) of red - shifted radiation at near infrared ( nir ) @xmath2 1 - 2 @xmath3 and far infrared ( fir ) @xmath2 100 - 200 @xmath4 wavelengths , and a mid infrared ( mir ) `` valley '' between these bumps are expected .
because of the heavy contamination caused by foregrounds of different origin , predominantly by the zodiacal ( interplanetary dust ) light , the measurements of cib contain large uncertainties .
moreover , these results only conditionally can be treated as _ direct measurements _ , because their interpretation primarily depends on the modeling and removal of these foregrounds .
therefore , the direct observations of cib generally allow derivation of the flux _
upper limits _ rather than detection of positive residual signals . in fig . 1 we show the cib fluxes based on the latest reports , and refer the reader to the review article by hauser and dwek ( 2001 ) on the current status of direct observations of cib .
presently , the most reliable results are obtained from the cobe observations at nir and fir domains where the contribution of the zodiacal light becomes comparable with the cib flux .
this concerns , first of all , the fir wavelengths at 140 and 240 @xmath4 ( hauser et al . , 1998 , schlegel et al . , 1998 , lagache et al . , 1999 ) , and perhaps also at 100 @xmath4 ( lagache et al . , 1999 ; finkbeiner et al . , 2000 ) where the results of 3 different groups are in a reasonable agreement with each other , but somewhat higher compared to the theoretical predictions .
if the reference of these fluxes to the truly extragalactic background is correct , this would imply that most of the star formation in the early universe must occur in highly obscured , dusty environments .
( stars ) are from hauser et al .
( 1998 ) , at 100 @xmath4 ( filled triangle ) from lagache et al .
( 1999 ) , at 60 and 100 @xmath4 ( open circles ) from finkbeiner et al .
the point at 15 @xmath4 ( open square ) is derived from the isocam source counts ( franceschini et al . , 2001 ) .
the fluxes reported at nir and optical wavelengths are shown by filled squares ( wright and johnson , 2001 ) , open circles ( matsumoto , 2000 ) , open diamonds ( cambresy et al .
, 2001 ) , and filled circles ( bernstein , 1998 ) .
the open triangles from pozzetti et al .
( 1998 ) correspond to lower limits .
while the reference models 2 and 3 should be considered as lower and upper limits , the curve 1 may be treated as the `` model of choice '' . ]
detections of cib are claimed also at nir - at 2.2 @xmath4 and 3.5 @xmath4 wavelengths ( dwek and arendt , 1998 ; gorjian et al . , 2000 ; wright and johnson , 2001 ) .
these fluxes together with the j - band upper limit at @xmath5 ( wright and johnson , 2001 ) and the fluxes of the optical light derived from the hst data below @xmath6 ( bernstein , 1998 ) , agree with the recent theoretical calculations by primack et al .
an independent analysis of the cobe data by cambresy et al .
( 2001 ) , as well as the results of the japanese irts satellite ( matsumoto et al .
2001 ) are in agreement with the fluxes at @xmath7 and @xmath8 reported by wright and johnson ( 2001 ) , but at shorter wavelengths the claimed fluxes are noticeably higher compared to other measurements ( see fig . 1 ) .
the `` best guess '' estimate of the sed at optical / nir wavelengths of about 20 - 50 @xmath9 is comparable with the fir flux of about 40 - 50 @xmath9 ( madau and pozzetti , 2000 ) .
this indicates that an essential part of the energy radiated by stars is absorbed and re - emitted by cold dust .
our current knowledge of mir , which carries information about the warm dust component , is quite limited .
the only flux estimates in this band derived from the isocam source counts at 6 and 15 @xmath10 ( franceschini et al . , 2001 )
apparently should be taken as lower limits .
-rays in the intergalactic medium calculated for 3 cib models presented in fig . 1 .
] these measurements stimulated new calculations of cib based on different phenomenological and theoretical ( backward , forward , cosmic chemical evolution , semi - analytical , _ etc .
_ ) approaches . in spite of certain achievements ,
the ability of current ( most successful ) models to reproduce the cib measurements nevertheless should not be overstated , because all these models contain a number of adjustable parameters and in fact are `` primarily designed for that purpose '' ( hauser and dwek , 2001 ) . in fig .
1 three cib models are shown which _ provisionally _ may be assigned as `` nominal '' ( 1 ) , `` low '' ( 2 ) , and `` extreme '' ( 3 ) descriptions of cib .
the curve 2 corresponds to the so - called `` lcdm - salpeter '' model of primack et al .
( 1999 ) , but assuming twice larger contribution of the dust component .
similar nir spectra with a flux of about @xmath11 at 1 @xmath4 have been assumed by de jager and stecker ( 2001 ) and franceschini et al .
the curve 1 is close to the prediction by primack et al .
( 2001 ) based on the so - called `` kennicut '' stellar initial mass function ( imf ) . and finally , the curve 3 is `` designed '' to match the extreme fluxes reported by matsumoto ( 2000 ) and cambresy et al .
( 2001 ) at nir and finkbeiner et al .
( 2000 ) at fir .
to calculate the mean free path of @xmath1-rays @xmath12 in the igm one must convolve the cib photon number distribution @xmath13 with the pair production cross - section .
because of the narrowness of the latter , for broad - band photon spectra over half the interactions of a @xmath1-ray photon of energy @xmath14 occur with a quite narrow interval of target photons @xmath15 centered on @xmath16 .
this gives a convenient approximation for the optical depth @xmath17 for @xmath1-rays emitted by a source at a distance @xmath18 : @xmath19 where @xmath20 is the sed of cib , @xmath21 is the source redshift , @xmath22 is the hubble constant normalized to @xmath23 .
no deviation of the observed @xmath1-ray spectrum from the intrinsic ( source ) spectrum , e.g. by a factor of @xmath24 , would imply @xmath25 , and consequently provide an upper limit on the cib flux at @xmath26 .
for example , this condition for 1 tev @xmath1-rays from mkn 421 or mkn 501 ( @xmath27 ) gives @xmath28 around 1 - 2 @xmath4 . since this estimate is quite close to the recent nir measurements ( see fig .
1 ) , we may conclude that already at 1 tev we detect significantly absorbed @xmath1-radiation from mkn 421 and mkn 501 .
this is demonstrated in fig . 2 where we present accurate numerical calculations of the mean free path of @xmath1-rays for three cib reference models shown in fig . 1 .
the horizontal lines indicate the distances to the reported tev blazars ( see below ) , as well as to 2 nearby prominent extragalactic objects the radiogalaxy m 87 and the ultraluminous starburst galaxy arp 220 . the curve 2 calculated for the cib lower limit ( model 2 in fig .
2 ) , should be treated as an upper limit for the mean free path .
this implies that we can not ignore the ig - absorption of tev @xmath1-rays from sources located beyond 100 mpc . on the other hand , this gives a unique chance to extract information about cib by detecting absorption features in tev spectra of extragalactic objects .
( aharonian et al . , 2001a ) . the vertical line at 17 tev indicates the edge of the spectrum measured with high statistical significance
] very often such a possibility is reduced to the search for sharp cutoffs in the energy spectra of extragalactic tev sources ( see e.g. stecker et al . , 1992 ) . however , in many cases this could be a misleading recommendation . indeed , the _ cutoff _ in a @xmath1-ray spectrum does not yet imply an _ evidence for the ig - absorption _ and vice versa , the _ lack of the cutoff _ can not be interpreted as _ absence of ig - absorption_. it is interesting to note that all cib models with conventional spectral shape between 1 and 10 @xmath4 , predict almost constant ( energy - independent ) mean free path of @xmath1-rays , and correspondingly insignificant spectral deformation at energies between 1 and several tev ( primack et al . , 2001 ) .
the explanation of this effect is straightforward . if the spectrum of background photons in a certain energy interval has a power - law dependence , @xmath29 , the mean free path in the corresponding @xmath1-ray energy interval @xmath30 . within @xmath31 the cib spectrum
is approximately described by a power - law @xmath32 or @xmath33 ( see fig . 1 )
, therefore in the interval between 1 and several tev the @xmath1-ray mean free path only slightly depends on energy ( fig . 2 ) .
in order to demonstrate the impact of the ig - absorption , in fig .
3 we show the sed of mkn 501 together with the absorption - corrected spectra reconstructed for 3 cib models shown in fig . 1 .
many nonthermal extragalactic objects representing different classes of agn and located within 1 gpc are potential tev sources .
first of all this concerns the bl lac population of blazars of which two nearby representatives , mkn 421 ( @xmath34 ) and mkn 501 ( @xmath35 ) are firmly established as tev @xmath1-ray emitters .
the current list of extragalactic tev sources contains 4 more bl lac objects with reported signals at 4 to 6 sigma level : 1es 2344 + 51 [ whipple ( catanese et al . , 1998 ) ] at @xmath36 , pks 2155 - 304 [ durham ( chadwick et al . , 1999 ) ] at @xmath37 , 1es 1959 + 650 [ telescope array ( nishiyama et al . , 1999 ) and hegra ( gtting et al . , 2001 ) ] at @xmath38 , and 1es 1426 + 428 [ veritas ( horan et al . , 2001 ) and hegra ( gtting et al . , 2001 ) ] at @xmath39 . since the discovery of tev @xmath1-radiation of mkn 421 ( punch et al .
1992 ) , this object has been subject of intensive studies through multiwavelength observations . the tev flux of the source is variable with typical average value between @xmath40 to @xmath41 that of the crab nebula , but with strong and rapid , as short as 0.5 h , flares ( gaidos et al . , 1996 ) which correlate with the source activity at other wavelengths . until the spectacular high state of the source in 2001
, the spectral studies were based mainly on the data taken during a quiescent or moderately high states . in particular , the spectrum measured by hegra during the 1998 `` asca '' campaign ( takahashi et al . , 2000
) is fitted by a steep power - law with photon index @xmath42 ( aharonian et al . , 1999b ) .
the exceptionally bright and long - lasting activity of mkn 421 in 2001 allowed the veritas ( krennrich et al . ,
2001 ) and hegra groups ( horns et al . , 2001 )
to derive the time - averaged gamma - ray spectrum of the source in the high state which up to @xmath43 tev is described as @xmath44 , i.e. by the same canonical `` power - law with exponential cutoff '' function found earlier for the high - state , time - averaged spectrum of mkn 501 ( aharonian et al .
, 1999a ) .
the spectra of these sources are not , however , identical . in fig .
4 we show the energy spectra based on approximately 40,000 and 60,000 tev @xmath1-ray events detected by hegra during the exceptionally high states of mkn 501 in 1997 and mkn 421 in 2001 , respectively . they are described by different [ @xmath45 combinations : @xmath46 and @xmath47 for mkn 501 ( aharonian et al . , 1999a ) and @xmath48 and @xmath49 ( horns et al . , 2001 ) .
the power - law part of the spectrum of mkn 421 is steeper , and the exponential cutoff starts earlier .
since both sources are located at approximately same distances , the difference in the cutoff energies may be interpreted as an evidence against the hypothesis which attributes the cutoffs to the pure ig - absorption effect .
for the typical energy resolution of cherenkov telescopes of about @xmath50 or larger , one has to be careful with conclusions concerning the spectral shape at energies beyond @xmath51 . obviously for determination of spectra with sharper ( super - exponential ) cutoffs ,
a significantly better energy resolution is required .
motivated by this , hegra collaboration recently re - analyzed the `` 1997 high - state '' spectrum of mkn 501 using an improved method for energy reconstruction with @xmath52 resolution achievable for the stereoscopic mode of observation ( hofmann , 2000 ) .
the new analysis confirms the result of the previous study up to 17 tev , but constrains stronger the flux at 21 tev ( aharonian et al .
, 2001a ) with this new point , the mkn 501 spectrum at very high energies is better described by a super - exponential cutoff ( compare curves ( a ) and ( b ) in fig .
-ray observations of the 2001 march 19 flare by rxte ( squares ) and veritas ( circles ) .
the bottom panel shows the rxte / pca `` 6 - 11 kev/3.5 - 4.5 kev '' hardness ratio ( from fossati et al . , 2001 ) . ] despite the huge overall photon statistics , the spectra of both sources above 10 tev can be determined using data accumulated over long periods of observations .
obviously , the time - averaged spectra of variable sources obtained in such way can be considered as _ astrophysically meaningful _ provided that the spectral shape is essentially time - independent , i.e. does not correlate with the absolute flux . remarkably , it turned out that during the high state in 1997 the shapes of daily spectra of mkn 501remained essentially stable despite dramatic flux variations ( aharonian et al . , 1999a ) , although for some specific flares significant spectral variations can not be excluded .
for example , the cat group ( djannati - atai et al . , 1999 ) found an evidence for a non - negligible hardness - intensity correlation for the 1997 april 7 and 16 flares .
the sensitivity of current ground - based detectors is not sufficient to study the spectral variability of tev radiation of mkn 421 and mkn 501 in quiescent states on timescales less than several days .
however , even the time - averaged spectra of these sources in low states are of certain interest .
the hegra observations of mkn 501 in 1998 and 1999 , when the source was at a @xmath53 times lower flux level compared to the average flux in 1997 , revealed a noticeable steepening ( by @xmath54 in photon index ) of the energy spectrum ( aharonian et al . 2001b ) .
this seems to be true also for mkn 421 , al least at low energies . while the spectrum of this source in a quiescent state has a steep power - law spectrum with photon index @xmath42 , the time averaged spectrum of the 2001 outburst is significantly flatter , with @xmath55 around 1 tev .
the seds of bl lacs are expected to be very hard up to tev energies , therefore the imaging atmospheric cherenkov telescopes ( iacts ) are nicely suited to search for short signals from these objects . in particular , the hegra , cat and veritas iacts can follow @xmath56 ( at 1 tev ) flares of mkn 421 and mkn 501 on timescales less than several hours , and thus are well - matched to the sensitivity and spectral coverage of x - ray satellites like rxte , bepposax and xmm for multiwavelength monitoring of flux variations .
this is a key condition which makes the simultaneous x- and @xmath1-ray observations meaningful and very important ( e.g. pian et al . , 1998 , sambruna et al .
, 2000 ; maraschi et al . , 1999 ; kataoka et al . , 1999 ;
krawczynski et al . , 2000 ;
takahashi et al . , 2000 ) .
the best so far results in this regard became available recently , after the well coordinated multiwavelength campaigns of spectacular flares of mkn 421 in 2001 .
on several occasions , truly simultaneous observations by rxte and tev instruments with durations up to 6 hours per night were carried out . a nice sample of such events , the 2001 march 22 flare detected by rxte and veritas ( fossati et al
. , 2001 ) shows a clear kev / tev correlation on sub - hour timescales ( see fig . 5 ) . for deep understanding of acceleration and radiation processes in jets ,
it is crucial to search for _ spectral _ variability on timescales comparable to the characteristic dynamical times of about 1 h. the hegra iact system is able to perform such studies , if the flux at 1 tev exceeds 1 - 2 crab units , @xmath57 .
the strongest flares of mkn 421 in 2001 provide us with such unique data .
the preliminary spectral analysis of observations of these flares ( horns et al . , 2001 ) shows a complex picture . if the spectra of some of these flares do not show noticeable spectral changes , the others , in particular the 21/22 march 2001 flare demonstrates an impressive correlation of the hardness ratio with the absolute flux ( see fig .
the studies of spectral evolution of tev @xmath1-rays and their correlations with x - rays on sub - hour timescales not only imply a new standard in @xmath1-ray astronomy , but also indicate that the quality of the tev data is approaching to the level which would allow us to follow and resolve simultaneously the predicted fluctuations in the putative synchrotron and ic emission components on timescales close to the shortest ones likely in these objects .
this may have two important consequences ( coppi and aharonian , 1999a ) : @xmath0 matching the observed x - ray / tev light curves ( as opposed to simply fitting snapshot spectra obtained many hours and days apart ) should provide a very stringent test of the so - called synchrotron - self compton ( ssc ) model since we have two detailed handles on the _ single electron _
distribution responsible for both emission components .
this test can rule out alternative _ hadronic _ models which are less attractive but still viable options for explanation of tev emission .
@xmath0 if the ssc model works , it will be possible to fix the key model parameters and calculate the blazar s intrinsic spectrum , comparing the observed variations of _ absolute fluxes _ and _ spectral shapes _ with the predictions of self - consistent , time - dependent numerical codes .
this is a crucial point because with an estimate of the intrinsic tev spectrum , then , and _ only then _
, we can attempt to estimate the ig - absorption effect by comparing the intrinsic and observed spectra , and thus to get information about cib . the observed x / tev correlations are often interpreted as a strong argument in favor of ssc model , the simplified one - zone version of which generally gives satisfactory fits to the observed x - ray and tev spectra of mkn 421 and mkn 501 ( e.g. mastichiadis and kirk , 1997 ; bednarek and protheroe , 1999 ; kataoka et al . , 1999 ; krawczynski et al .
, 2000 ; tavecchio et al . , 2001 , _ etc .
an example of very good spectral fits for the 1997 april flares of mkn 501 , obtained for fixed `` standard '' model parameters @xmath58 and @xmath59 and changing only the maximum energy of accelerated electrons , is shown in fig .
7 ( tavecchio et al . , 2001 ) .
it should be noticed , however , that the calculation in fig .
7 do not take into account the ig - absorption of @xmath1-rays which otherwise results in rather flat intrinsic @xmath1-ray spectra ( coppi and aharonian , 1999b ; konopelko et al .
, 1999 ; guy et al , 2000 ) .
in particular , any cib model that predicts nir fluxes close or higher the cib reference model no.1 would require unusually hard intrinsic tev spectrum ( guy et al . , 2000 ) with photon index @xmath60 ( see fig .
such spectra can not be easily explained by the standard one - zone ssc model without invoking extreme jet parameters like doppler factors @xmath61 , and very small b - field @xmath62 , which however reduce the radiation efficiency to an almost unacceptably low level ( krawczynski et al . , 2001 ) .
the situation is even more difficult for mkn 421 , the x - ray synchrotron cutoff of which never extends beyond 10 - 20 kev .
this significantly limits the freedom for assuming parameters which would allow hard multi - tev ic spectra . despite these difficulties ,
the leptonic models ( perhaps in their more sophisticated versions including the multi - zone ssc and/or external compton components ) remain the generally preferred concept for tev blazars .
these models have two attractive features : _ ( i )
_ the capability of the ( relatively ) well understood shocks to accelerate electrons to tev energies ( sikora and madejski , 2001 ; pelletier , 2001 ) and _ ( ii ) _ effective production of tightly correlated x - ray and tev @xmath1-ray emission components via the synchrotron and inverse compton radiation channels ( e.g. ulrich et al .
, 1997 ) .
the so - called hadronic models are generally lacking in these virtues .
these models assume that the observed @xmath1-ray emission is initiated by accelerated protons interacting with ambient mater ( bednarek , 1993 ; dar and laor , 1997 ; pohl and schlickeser , 2000 ) , photon fields ( mannheim , 1996 ) , magnetic fields ( aharonian , 2000 ) or both ( mcke and protheroe , 2001 ) .
the models involving interactions of protons with photon and b - fields require particle acceleration to extreme energies exceeding @xmath63 which is possible if the acceleration time is close to @xmath64 with @xmath65 ( @xmath66 and @xmath67 are the so - called gyro - radius , and gyro - factor , respectively , @xmath68 is the speed of light ) .
this corresponds ( independent of specific acceleration mechanism ) to the maximum ( theoretically possible ) acceleration rates which hardly can be achieved by the conventional diffusive shock acceleration mechanism ( e.g. pelletier , 2001 , sikora and madejski , 2001 ) . on the other hand , the undisputed ( for any tev blazar model ) condition of high ( close to @xmath69 ) efficiency of radiative cooling of accelerated particles
requires extreme parameters characterizing the sub - parsec jets and their environments , in particular very high ( often , almost unrealistic ) densities of the thermal plasma , radiation and/or b - fields .
but of course , these problems should not prevent us from considering the hadronic models as a viable alternative to the leptonic models .
it should be noticed in this regard that the very fact of strong x / tev correlations do not yet exclude the hadronic models , because the _ secondary _ electrons the products of interactions of accelerated protons and primary @xmath1-rays can be readily responsible for the synchrotron x - radiation . also , some of the hadronic models , for example the _ synchrotron - proton _ model can provide not only effective radiative ( synchrotron ) cooling but also quite good tev spectral fits .
the proton - synchrotron model allows very hard intrinsic tev spectra and can rather naturally explain the stable spectral shape of tev radiation of mkn 501 during strong flares by the self - regulated synchrotron cutoff at @xmath70 ( aharonian , 2000 ) .
8 demonstrates the ability of the synchrotron - proton model to explain the spectra of tev radiation of mkn 501 in both the low and high states of the source . in the proton - synchrotron model we deal with highly magnetized ( 100 g or so ) condensations of @xmath1-ray emitting clouds of ehe protons , where the magnetic pressure dominates over the pressure of relativistic protons . in the ssc models
the situation is exactly opposite .
the pressure of relativistic electrons is much larger than the pressure of the b - field ( aharonian , 2000 ; kino et al . ,
surprisingly , for tev blazars we do not have an intermediate arrangement
! both the leptonic and hadronic models of tev blazars pose a number of questions .
the answer to some of these question may become possible in the near future .
such an optimistic view is based on the recent progress in the theoretical ( time - dependent ) studies of characteristics of nonthermal radiation produced in agn jets , and , of course , on the unique data obtained by rxte and ground - based @xmath1-ray detectors during the spectacular flares of mkn 421 in 2001 .
importantly , on some occasions positive signals from mkn 421 were detected also by low - energy air cherenkov instruments stacee ( ong et al . , 2001 ) and
celeste ( le gallou et al . , 2001 ) .
these results are obtained at energies around 100 gev at which the ig - absorption could be safely neglected , and therefore the reported fluxes can be used as `` calibration '' points when comparing the theoretical predictions with experimental data . because mkn 421 and mkn 501 are located at almost _ same _ distances
, we can require that any spectral feature attributed to the ig - absorption be _ exactly the same _ for both sources .
this is an important circumstance which can help to disentangle the local features in the intrinsic spectra from the contribution induced by interactions with the cib photons . at the same time it is crucial to have tev blazars at larger distances ,
so we can exploit the differences in the spectral modification factors .
fortunately , in addition to mkn 421 and mkn 501 , presently we do have 4 more tev blazar candidates
1es 2344 + 5 , 1es 1959 + 65 , pks 2155 - 304 and 1es 1426 + 428 located at larger distances ( see fig . 2 ) .
because of the relatively large redshift of 1es 1426 + 428 ( @xmath39 ) , the tev radiation from this extreme bl lac object suffers severe ig - absorption .
therefore the discovery of tev @xmath1-rays from this source by two independent ( veritas and hegra ) groups came , to a certain extent , as a pleasant surprise .
the spectrum deformation factors calculated for 3 different cib models show that above 300 gev the ig - absorption leads to significant steepening of the @xmath1-ray spectrum , but starting from 1 tev to several tev the spectrum is deformed only slightly , although the suppression of the absolute flux may be as large as a factor of 100 ( fig .
this effect has a simple explanation discussed in sec .
-ray spectra from 1es 1426 + 428 ( see the text ) . ]
remarkably , the _ preliminary _ spectrum of 1es 1426 + 428 derived from the hegra ct system data taken in 1999 and 2000 ( gtting et al . , 2001 ) , does show a steep slope at low energies with a tendency for a flattening above 1.5 tev .
the large statistical errors in flux measurements as well as the lack of information about the intrinsic @xmath1-ray spectrum prevent us from definite conclusions . nevertheless , in fig .
10 we show the expected spectra of @xmath1-rays _ assuming _ a single power - law intrinsic spectrum with photon index identical to the index observed by _
bepposax _ in the energy region up to 100 kev , @xmath71 ( costamante et al . , 2001 ) .
the solid curve represents the absorbed spectrum of @xmath1-rays calculated for the cib model 1 and for the hubble constant @xmath72 . in order to demonstrate the sensitivity of the intergalactic absorption effect to the ratio of the cib flux and the hubble constant , @xmath73 ( for the given redshift @xmath21 , and the shape of the cib spectrum , this ratio is the only parameter which defines the optical depth ) , in fig .
10 we show the @xmath1-ray spectra calculated for two other values of the ratio @xmath73 which are @xmath40 less ( dashed curve ) and @xmath40 more ( dot - dashed curve ) compared to the the nominal value ( solid curve ) .
hopefully , the future multiwavelength studies may allow us to distinguish between these three curves , thus we could be able to `` measure '' the absolute cib flux at nir , taking into account that the current uncertainty of the the hubble constant ( presumably ) does not exceed @xmath40 .
independent of details , the pure fact of detection of tev @xmath1-rays from 1es 1426 + 428 is an indicative of a need for revision of the current conceptual view of tev blazars , according to which the synchrotron ( x - ray ) peak in the sed dominates over the inverse compton ( tev ) peak ( see e.g. fossati et al .
1998 ) . indeed , although the _ detected _ flux of @xmath1-rays is only few times @xmath74 , corrected for ig - absorption this flux may well exceed @xmath75 . for comparison , the x - ray measurements ( costamante et al . , 2001 ;
t. takahashi , private communication ) show fluxes at the level of @xmath76 .
the tev luminosity significantly exceeds the x - ray luminosity , and apparently this should have an impact on the revision of the current tev - blazar models . in particular , for the inverse compton models ( for review see sikora and madejski , 2001 ) this would require that the radiation of electrons occurs in the regime dominated by compton losses .
interestingly , this could serve as a good argument for justification of the above assumption concerning the single power - law _ intrinsic _ tev spectrum .
indeed , in this regime a pronounced feature ( spectral flattening ) formed in the steady - state electron distribution ( caused by the structure in the klein - nishina cross - section ) is largely compensated by the same klein - nishina effect in the radiated @xmath1-ray spectrum ; thus the @xmath77 type _ acceleration _
spectrum of electrons is ( almost exactly ) transfered to the @xmath1-ray spectrum ( see e.g. zdziarski and krolik 1993 ) . of course ,
in this case we may expect , within the one - zone ssc model , quite hard synchrotron x - ray spectrum with spectral index @xmath78 .
this actually does not contradict to the _ bepposax _ data for which the broken power - law model gives very flat spectrum with an index above 10 kev , @xmath790.3 - 0.4 ( costamante et al .
, 2001 ) .
the hard intrinsic tev spectra of 1es 1426 + 428 with constant power - law slope and high @xmath80 ratio can be explained also by the _ proton - synchrotron _ model , assuming that a small , @xmath81 fraction of the proton - synchrotron tev @xmath1-rays suffer internal absorption .
if so , the hard x - ray emission could be attributed to synchrotron radiation of the secondary pair - produced electrons ( aharonian , 2000 ) .
the weak point of this model is in the requirement of very large , 100 g or so b - field and extremely high acceleration rate of protons . .
the radiation component associated with comptonization of ambient optical photons is shown by dashed curve i. the dashed curve ii is the residual of the total tev source emission ( after subtraction of the unshocked wind component i ) , and formally could be attributed either to the comptonization of the bulk motion on ambient fir photons or to the radiation of blobs in shocked jet .
the heavy solid line represents the superposition of these two components .
fits to the observed flux of mkn 501 are shown by thin solid lines .
curve 1 corresponds to the fit given by equation ( b ) in fig . 3 , and curve 2 to the steepest possible spectrum above 17 tev based on the recent reanalysis of mkn 501 hegra data ( aharonian et al . , 2001a ) .
the intrinsic ( absorption - corrected ) spectra of mkn 501 , corresponding to the fits to the observed points `` 1 '' and `` 2 '' , are shown by dotted lines `` @xmath82 '' and `` @xmath83 '' , respectively ( from aharonian et al . ,
from the discussion of the previous section we may summarize that the basic temporal and spectral characteristics of tev blazars can be described , at least qualitatively , by the the current leptonic , and perhaps also hadronic models . at the same time
all these models would fail to explain the sharp pile - up which may appear at the end of the `` reconstructed '' spectrum of mkn 501 , if the reported fir fluxes correctly describe the level of cib ( see fig .
3 ) . motivated by such a non - standard spectral shape ,
recently several extreme hypotheses have been proposed to overcome the `` ir background - tev gamma - ray crisis '' ( protheroe and meyer 2000 ) . in particular , harwit et al .
( 2000 ) suggested an interesting idea that the hegra highest energy events are due to bose - einstein condensations interacting with the air atmosphere .
another , even more dramatic hypothesis
violation of the lorentz invariance has been proposed to solve this problem ( e.g. kifune , 1999 ; stecker and glashow , 2001 ) .
two more proposals could be added to the list of `` exotic '' solutions : ( i ) assuming that mkn 501 is located at a distance @xmath84 mpc ; or ( ii ) assuming that tev @xmath1-rays from mkn 501 are not direct representatives of primary radiation of the source , but are formed during the development of high energy electron - photon cascades in the intergalactic medium ( aharonian et al .
2001b ) . at first glance , both ideas seem relatively `` neutral '' .
but in fact they do contain dramatic assumptions .
while first hypothesis implies non - cosmological origin of the redshift of mkn 501 , the second hypothesis requires extremely low intergalactic magnetic field of order of @xmath85 g or less .
although very fascinating , the appeal for revisions of essentials of modern physics and astrophysics seems to be too premature .
the nature of the fir isotropic emission detected by cobe is not yet firmly established , and it is quite possible that the bulk of the reported flux , especially below 100 @xmath4 , is a result of superposition of different local backgrounds .
it should be noticed , however , that not only the flux at @xmath86 , but also the cib fluxes at longer , @xmath87 wavelengths are responsible ( albeit in a less distinct form ) for the appearance of the pile - up , unless we assume very specific cib spectrum in the mir - to - fir transition region ( aharonian et al . ,
1999a ; coppi and aharonian , 1999b ; guy et al .
, 2000 ; renault et al . ,
therefore it is important to explore other possibilities for formation of pile - ups in the spectra of tev blazars .
motivated by this , recently we have proposed a new , _ non - acceleration _ scenario which postulates that the @xmath88 radiation of mkn 501 is a result of the _ bulk motion comptonization _ of ambient low - frequency photons by a cold ultrarelativistic conical wind ( aharonian et al . , 2002 ) .
the relativistically moving plasma outflows in forms of jets or winds , are common for many astrophysical phenomena on both galactic or extragalactic scales .
independent of the origin of these relativistic outflows , the concept of the jet seems to be the only successful approach to understand the complex features of nonthermal radiation of _ blazars , microquasars _ and _ grbs_. the lorentz factor of such outflows could be extremely large . in particular in the crab nebula the lorentz - factor of the mhd wind is estimated between @xmath89 and @xmath90 ( rees and gun , 1974 ; kennel and coroniti , 1984 ) .
meszaros and rees ( 1997 ) have shown that in the context of cosmological grbs the magnetically dominated jet - like outflows from stellar mass black holes may attain extreme lorentz factors exceeding @xmath89 .
the conventional lorentz factors of jets in the inverse compton models of @xmath1-ray blazars , are rather modest , @xmath91 .
however there are no apparent theoretical or observation arguments against the bulk motion with much larger lorentz - factors ( see e.g. discussion by celotti et al . , 1998 ) . because of existence of dense photons fields in the inner sub - parsec region of sources like mkn 421 and mkn 501 , the compton optical depth @xmath92 could be as large as 1 . obviously , for our model the most favorable value for @xmath92 lies between 0.1 ( in order to avoid huge energy requirements to the outflow ) and 1 ( in order to avoid the compton drag ) . in particular , we have to assume that in the proximity of the black hole the outflow should be poynting flux dominated , and only at large distances from the central object , where the photon density is significantly reduced , the major part of the electromagnetic energy is transfered to the kinetic energy of bulk motion . due to the extremely large lorentz factors exceeding @xmath90 , the compton scattering on the ambient nir / optical photons with energy more than 1 ev proceeds in deep klein - nishina regime ;
therefore the @xmath1-radiation should have a very narrow distribution with energy @xmath93 .
meanwhile , the ic scattering on ambient far ir photons surrounding the central source , still takes place in the thomson regime , and thus results in a smooth broad - band spectrum .
11 demonstrates that the overall _ absorption - corrected _ ( for the cib at fir close to the extreme cib model no.3 ) spectrum of mkn 501 can be satisfactorily explained in the terms of the inverse compton emission of the cold jet with bulk lorentz factor @xmath94 , assuming an ambient radiation field with a narrow ( planckian ) type radiation with temperature @xmath95 ( dashed curve i ) .
the dashed curve ii which formally is the residual from the subtraction of the unshocked wind component radiation from the intrinsic ( reconstructed ) tev emission ( solid heavy line ) , can be attributed to the bulk comptonization on ambient far ir photons produced e.g. by cold clouds surrounding the source .
alternatively , the `` residual component '' ( curve ii ) can be referred to the ssc ( or any other ) radiation component of blobs in shocked jet .
this two - stage ( _ pre - shock _ plus _ post shock _ ) scenario of formation of tev @xmath1-ray emission is schematically illustrated in fig .
12 the possibility to disentangle the multi - tev emission with a characteristic sharp pile - up at the very end of the spectrum , @xmath96 , from the sub-10 tev emission associated with the shocked structures ( e.g. blobs ) in the jet , not only solves the possible `` ir background
tev gamma - ray crisis '' but also allows more relaxed parameter space for interpretation of x - rays and the _ remaining _ `` low '' energy ( @xmath97 ) @xmath1-rays within the conventional ssc scenario .
consequently , this offers more options for interpretation of x - ray / gamma - ray correlations .
if the overall tev radiation of mkn 501 indeed consists of two , _ unshocked _ and _ shocked _ jet emission components , we may expect essentially different time behaviors of these components . in particular , the `` unshocked jet '' ( @xmath98 )
component should arrive earlier than the ssc components consisting of synchrotron x- and sub-10 tev ic @xmath1-rays .
therefore , an important test of the suggested two stage scenario of the tev radiation of jets would be the search for correlations ( or lack of such correlations ) of @xmath88 radiation with both the low energy ( e.g. 1 - 3 tev ) @xmath1-rays and synchrotron x - rays .
the low statistics of ( heavily absorbed ) @xmath1-rays above 10 tev makes the search for such correlations rather difficult , and requires ground - based instruments with huge , @xmath99 detection areas in this energy domain .
the new generation iact arrays like cangaroo-3 , h.e.s.s . and
veritas should be able to perform such correlation studies .
it should be noticed , however , that the study of the signatures of ig - absorption of @xmath1-rays from mkn 421 and mkn 501 does not contain adequate information about cib at @xmath100 .
this information is essentially lost due to the heavily absorbed @xmath1-rays above 20 tev .
this objective is rather contingent on discovery of nearby extragalactic tev sources within @xmath101 50 mpc , dictated by the condition that the mean free path of @xmath1-rays at these energies should not significantly exceed the distance to the source ( see fig . 2 ) .
despite the lack of very close blazars , other potential extragalactic @xmath1-ray sources like the nearby radiogalaxies centaurus a and m 87 , and perhaps also the startburst galaxies arp 220 , m 82 and ngc 253 may ( hopefully ) provide us with multi - tev @xmath1-rays for such important studies .
yet , the distant tev sources may offer an additional , and perhaps even more informative channel for derivation of cib fluxes at fir wavelengths .
such a possibility is based on the study of angular and spectral characteristics of hypothetical giant _ pair halos _ expected around powerful nonthermal extragalactic sources ( aharonian et al . , 1994 ) .
because of ig - absorption , the bulk of primary tev @xmath1-rays from distant extragalactic sources is not visible .
the energy of these photons however is not lost the electromagnetic cascade initiated by the absorbed @xmath1-rays can be observed ! the cascade in the ig medium ( e.g. berezinsky and kudriavtsev , 1990 ; protheroe and stanev , 1993 ; coppi and aharonian , 1997 ) is supported by 2 processes : ( i ) by inverse compton scattering on the photons of 2.7 k mbr and ( ii ) by photon - photon pair production on cib ( for energies above 100 tev - also on 2.7 k mbr ) . when the magnetic field near the source is sufficiently large , @xmath102 g ( generally on the scales of tens of mpc , but if the spectrum of primary @xmath1-rays extends into pev region - within tens of kpc ) , i.e. when the mean free path of cascade electrons ( at least at the initial stages ) is large compared to larmor radii , the cascade will generate , even for a highly beamed primary source , an extended isotropic pair halo ( see fig .
while the formation of pair halos is unavoidable for _ all _ extragalactic objects with spectra extending beyond 10 gev , only relatively compact pair halos formed around ( multi ) tev sources could be detectable .
the superposition of the cascade radiation of all pair halos formed in the universe may contribute significantly to the diffuse gev @xmath1-ray background ( coppi and aharonian , 1997 ) .
the halo radiation can be distinguished by its characteristic variation in spectrum and intensity with angular distance from the central source , which weakly depends on the details of the source model , in particular on the energy spectrum and the geometry of the primary source , especially when a hard primary @xmath1-ray spectrum extends to pev energies .
the energy @xmath103 and the arrival angle @xmath104 of the detected photon from a source at a distance @xmath18 are determined , on average , by the energy of the parent electron @xmath105 ( @xmath106 ) , and the distance @xmath107 from the central source where this electron is produced . because the mean free path of @xmath1-rays @xmath108 decreases with energy ( see fig .
2 ) , the energy and site of production of the secondary electron are defined , on average , by the `` last @xmath1-cib interaction '' , @xmath109 , therefore the energy @xmath103 and arrival angle @xmath104 of the _ detected _ @xmath1-rays contain information about their `` grandparents '' - @xmath1-rays with energy @xmath110 , and mean free path @xmath111 . correspondingly , the _ measured _ mean free path @xmath112 should tell us about the density of cib at @xmath113 .
thus , detections of pair halos surrounding mkn 421 and mkn 501 at 1 tev should provide information on cib around 50 @xmath4 , provided that the primary spectrum extends beyond 30 tev .
recently the hegra collaboration made the first attempt to probe the existence of a pair halo in direction of mkn 501 , using the data base accumulated during the period of observations from 1997 to 1999 ( aharonian et al .
, 2001c ) .
the derived limit on the halo flux @xmath114 within @xmath115 is unfortunately above the theoretical expectations based on the energy budget constraints . indeed ,
mkn 501 actually is not an ideal candidate for a halo source given its relatively modest luminosity and the fact that the preferable distance to observe halo is in the range of several 100 mpc to 1 gpc ; consequently , the most effective searches can be performed by detectors with energy threshold around 100 gev ( aharonian et al . , 1995 ) .
in this regard , the bl lac objects 1es 1426 + 428 at @xmath39 and pks 2155 - 30 at @xmath37 with @xmath1-ray luminosities exceeding the `` quiescent '' tev luminosity of mkn 501 al least by a factor of @xmath116 , seem to be more suitable objects for producing _ detectable _ pair halos .
the forthcoming h.e.s.s .
array of 10 m iacts equipped with optimal for this purpose @xmath117 fov high resolution cameras , will be the first instrument for a sensitive search for pair halos at @xmath118 .
generally , the blazars are are relatively weak sources
they seem very bright because of the doppler boosting ( proportional to @xmath119 ) .
the distant powerful radiogalaxies and quasars with large - scale jets ( as potential accelerators of ultrahigh energy cosmic rays ) may appear more promising sites for production of detectable _ strong _ and _ compact _ pair halos ( aharonian , 2002 ) . the detection and identification of such halos from cosmologically distant objects beyond @xmath120 is possible at energies of @xmath1-rays of about 10 gev .
while the spectral cutoffs in the energy spectra of the direct and halo emission components are caused by absorption in the diffuse uv background radiation , the angular size of the halo contains information about the cib at several to 10 @xmath4 wavelengths at the epoch corresponding to the redshift of the central source @xmath21 .
this is a unique ( not available by other means ) channel which provides us with model - independent information about the cosmological evolution of cib .
such studies can be effectively performed by glast ( e.g. gehrels and michelson 1999 ) and by future sub-10 gev threshold high - altitude iact arrays like 5@5 ( aharonian et al .
the energy - dependent mean free path of @xmath1-rays in the intergalactic medium at tev energies does not exceed several 100 mpc
. therefore vhe @xmath1-rays from tev blazars arrive with significantly distorted spectra .
our limited knowledge of cib results in large uncertainties in the reconstructed ( corrected for intergalactic absorption ) intrinsic @xmath1-ray spectra .
consequently , in spite of good quality of spectral measurements of two strongest tev blazars , mkn 421 and mkn 501 , nowadays we are faced with a challenge - our understanding of radiation processes in relativistic jets is neither complete nor conclusive .
nevertheless , the gamma - ray astronomers believe that eventually ( soon ? ) they will learn to identify confidently the radiation mechanisms , to fix / constrain the relevant model parameter space , and calculate robustly the intrinsic @xmath1-ray spectra based on the multiwavelength studies of spectral and temporal characteristics of blazars obtained simultaneously at x - ray and tev bands on sub - hour timescales
. then it will be possible to estimate unambiguously the effect of the intergalactic @xmath1-ray absorption , and thus to infer robust information about the cib fluxes .
moreover , the studies of angular and spectral properties of giant pair halos formed around powerful extragalactic multi - tev sources may provide us with a unique tool for derivation of the spectra and absolute fluxes of cib at _ different cosmological epochs z _ , and thus to probe the evolution of galaxies in past .
aharonian f.a . , 2002 , mnras , to appear in mnras aharonian f.a . , timokhin a. , plyasheshnikov a.v . , 2002 , to appear in a&a aharonian f.a .
et al . , 2001 ,
physics , 15 , 335 aharonian f.a .
et al.(hegra collaboration),2001a , a&a 366,62 aharonian f.a .
et al.(hegra collaboration),2001b , apj,546,898 aharonian f.a .
et al.(hegra collaboration),2001c , a&a,366,746 aharonian f.a . , 2000 , new astronomy , 5 , 377 aharonian f.a .
et al.(hegra collaboration),1999a , a&a,349,11 aharonian f.a .
et al.(hegra collaboration),1999b , a&a,350,757 aharonian f.a .
et al . , 1995 ,
24th icrc , rome , vol . 2 , 2622 aharonian f.a . , coppi p.s . and
vlk h.j , 1994 , apj , 423 , l5 bednarek w.,1993 , apj , 402 , l29 bednarek w. and protheroe j.r.,1999 , mnras , 310 , 577 berezinsky v. s. and kudriavtsev v. a. , 1990 , apj , 349 , 620 , cambresy l. et al . , 2001 ,
apj , 555 , 563 catanese m. et al . , 1998 , apj , 501 , 616 celotti a. , fabian a.c . ,
rees m.j . , 1998 , mnras , 293 , 239 chadwick p. et al . , 1999 ,
physics , 11 , 145 coppi p.s . and aharonian f.a . , 1999a , apj , 521 , l33 coppi p.s . and aharonian f.a . ,
1999b , astropart .
physics , 11 , 35 coppi p.s . and aharonian f.a . , 1997 , apj , 487 , l9 costamante l. et al . ,
a&a , 371 , 512 dar a. and laor a. , 1997 , apj , 478 , l5 de jager o.c . and
stecker f.w . , 2001 , to appear in apj djannati - atai a. et al . , 1999 ,
a&a , 350 , 17 dwek e. and arendt r.g . , 1998 , apj , 508 , l9 finkbeiner d.p .
, devis m. and schlegel d.j . , 2000 , apj , 544 , 81 fossati g. , jordan m. and buckley j. , 2001 , in : `` gamma - ray astrophysics 2001 '' , aip conf .
series , in press fossati g. et al .
, 1998 , mnras , 299 , 433 franceschini a. et al . , 2001 , a&a 378 , 1 gaidos j.a .
et al . , 1996 , nature , 383 , 319 gehrels n. and michelson p. , 1999
, astropart .
11 , 277 gorjian v. , wright e.l . ,
chary r.r .
, 2000 , apj , 536 , 550 gould r.j . and schreder , g.p . , 1966 , phys .
letters , 16 , 252 gtting n. et al .
( hegra collaboration ) , talk presented at 27th icrc , hamburg guy j. et al . , 2000 ,
a&a , 359 , 419 hauser m.g and dwek e. , 2001 , ara&a , 39 , 249 hauser m.g .
et al . , 1998 ,
apj , 508 , 25 hofmann w. , 2000 , astropart .
, 12 , 207 horan d. et al . , 2001 ,
27th icrc , hamburg , vol .
7 , 2622 horns d. et al .
( hegra collaboration ) , talk presented at 27th icrc , hamburg jelley j.v .
, 1966 , phys .
letters , 16 , 479 kataoka j. et al . , 1999 ,
apj , 514 , 138 kennel c.f . and
coroniti f.v . , 1984 ,
apj , 283 , 694 kifune t. , 1999 , apj , 518 , l21 kino m. , takahara f. and kusunose m. , 2001 , to appear in apj konopelko a. k. et al .
, 1999 , apj , 518 , l13 krawczynski h. et al . , 2000 ,
a&a , 353 , 97 krawczynski h. , et al . , 2001 , in preparation krennrich f. et al
. , 2001 , apj , 560 , l45 lagache g et al . , 1999 ,
a&a , 344 , 322 le gallou r. et al . , 2001 ,
talk presented at 27th icrc , hamburg madau p. and pozzetti l. , 2000 , mnras , 312 , l9 mannheim k. , 1998 , science 279 , 684 maraschi l. et al .
, 1999 , astropart .
physics , 11 , 189 mastichiadis a. and kirk j.g . , 1997 ,
a&a , 320 , 19 matsumoto t. , 2000 , the isas science report sp no .
14 , p. 179
meszaros p. and rees m.j .
, 1999 , apj , 482 , l29 mcke a. and protheroe r. j. , 2001 , astropart .
physics , 15 , 121 nikishov a.i .
, 1962 , sov .
jetp , 14 , 393 nishiyama t. et al.,1999 , 26th icrc ( salt lake city ) , vol .
3 , 370 ong r. et al . , 2001 ,
talk presented at 27th icrc , hamburg pian e. et al .
, 1998 , apj , 492 , l17 pohl m. and schlickeiser r. , 2000 , a&a , 354 , 395 pozzetti l. et al . , 1998 ,
mnras , 298 , 1133 primack j.r .
et al . , 2001 , in : high energy gamma - ray astronomy ( eds .
aharonian and h.j .
vlk ) , aip conf ser .
558 , 463 primack j.r .
et al . , 1999 ,
physics , 11 , 93 protheroe r.j . and meyer h. , 2000 , physics letters b , 493 , 1 protheroe r.j . and stanev t. , 1993 , mnras , 264 , 191 punch m. et al . , 1992 , nature , 358 , 477 rees m.j . and gun j.e . , 1974 , mnras , 167 , 1 renault c. et al , 2001 , a&a , 371 , 771 sambruna r. et al , 2000 , apj , 538 , 127 schlegel d.j . , finkbeiner d.p . and devis m. , 1998 , apj , 500 , 525 sikora m. and madejski g , 2001 , in : high energy gamma - ray astronomy aip conf ser .
558 , 275 stecker f.w . and
glashow s. l. , 2001 , astropart .
physics , 16 , 97 stecker f.w .
, de jager o.c . and
salamon m.h.,1992 , apj , 390 , l49 takahashi t. et al . , 2000 , apj , 542 , l105 tavecchio f. et al . ,
apj , 2001 , 554 , 725 ulrich m.h . , maraschi l. and urry c.m . , 1997 , ara&a , 35 , 445 vassiliev v. v. , 2000 , astropart .
physics , 12 , 217 wright e.l . , and johnson d.b .
, 2001 , to appear in apj zdziarski a. k. and krolik j.h . , 1993 , apj , 409 , l33 | the recent developments in studies of tev radiation from _ blazars _ are highlighted and the implications of these results for derivation of cosmologically important information about the _ cosmic infrared background radiation _ are discussed . |
vo@xmath0 belongs to transition metal compounds which exhibit metal - insulator transitions @xcite . at @xmath5=340
k , vo@xmath0 undergoes a phase transition from a semiconductor with monoclinic structure to a metal with the tetragonal rutile structure .
the nature of the ground - state semiconducting phase is still rather uncertain . in the semiconducting phase ,
the v atoms are paired which was a reason for the suggestion that the electron - phonon interaction is responsible for the splitting of the @xmath6-band and the opening of a band gap @xcite .
this idea is supported by band structure calculations according to which the crystallographic phase transition can be explained by the formation of a charge - density wave accompanied by a lattice distortion and a subsequent condensation of phonons @xcite . according to ref .
@xcite , there is enough energy gain to account for the metal - insulator transition through structural distortions that permits a strengthening of the vanadium @xmath6@xmath6 bonds and a reorganization of the states near the fermi level . on the other hand ,
some calculations indicate that a crystallographic distortion is not sufficient to open up a gap , and that the electron - correlation effects play an important role in the transition @xcite .
it was concluded in ref .
@xcite that the energy gap in vo@xmath0 is of the charge - transfer type rather than of the mott
hubbard type as in the late transition metal compounds , like nio and cuo @xcite .
in addition to earlier spectroscopic studies @xcite , the photoelectron spectroscopy measurements and low - energy electron diffraction studies on vo@xmath0 have been carried out recently in ref .
the present work aims at a combined experimental study of the electronic structure of a vo@xmath0 single crystal at room temperature by the use of high - energy spectroscopies .
a high - resolution x - ray photoelectron spectroscopy ( xps ) provides information about the total density of states ( dos ) in the valence band ( vb ) ; v@xmath1 ( the @xmath7 transition ) , v@xmath2 ( the @xmath8 transition ) , o@xmath3 ( the @xmath9 transition ) x - ray emission ( xes ) vb spectra ( excited by both electrons and photons ) probe the v@xmath10 , v@xmath11 and o@xmath12 partial dos in the valence band ; the o@xmath4 total electron yield spectrum probes the o@xmath12 unoccupied states .
band structure calculations are performed and compared with the experimental spectra leading to the conclusion that the electronic structure of vo@xmath0 is more bandlike than correlated .
the xps measurements have been carried out with a phi 5600 ci multitechnique spectrometer using monochromatized aluminium @xmath3 radiation with a full width at half - maximum ( fwhm ) of 0.3 ev .
the energy resolution of the analyzer was 1.5% of the pass energy .
we estimate an energy resolution of about 0.35 ev for the xps measurements on vo@xmath0 . the base pressure in the vacuum chamber during the measurements was @xmath13 torr .
all of the experiments presented in this paper have been performed at room temperature with the same single crystal of vo@xmath0 ( @xmath14 mm ) .
initial measurements were performed without cleaving the crystal , and hence a high contamination of carbon was detected on the surface . therefore the final measurements were done on a surface that was cleaved _ in vacuo_.
thus an excellent surface with a relatively small amount of defects and contaminations could be obtained , and hence the intrinsic properties of the samples could be studied . for the comparison
, xps measurements of pure v metal ( single crystal ) were also performed .
all xps spectra were calibrated using the au @xmath15 signal from an au foil [ @xmath16 = 84.0 ev ] .
electron excited v@xmath1-emission spectra ( the @xmath17 transition ) of vo@xmath18 were recorded using an rsm-500 spectrometer with a diffraction grating ( @xmath19=600 lines / mm ; @xmath20=6 m ) .
the accelerating voltage and current on the x - ray tube were @xmath21=4.4 kev and @xmath22=0.3 ma .
the energy resolution was 0.4 ev .
v@xmath23-spectra ( the @xmath8 transition ) of vo@xmath0 were measured using a fluorescent johan - type vacuum spectrometer with a position - sensitive detector @xcite .
cu@xmath3 x - ray radiation from the sealed x - ray tube was used for excitation for the fluorescent v@xmath23 xes .
a quartz crystal ( rhombohedral plane , second order reflection ) curved to @xmath20=1.8 m was used as an analyzer .
the spectra were measured with an energy resolution @xmath24=0.22 ev .
energy - selective excited o@xmath3- ( the @xmath9 transition ) , v@xmath1-spectra ( the @xmath17 transition ) were measured as well as x - ray absorption spectra ( the v@xmath12- and o@xmath4-edges ) in the sample drain - current mode .
these measurements were performed at the undulator beam line bw3 at hasylab hamburg , germany @xcite , equipped with a modified sx-700 monochromator .
the soft x - ray emission spectra were recorded at various excitation energies in the first order of diffraction with a resolution of about 0.7 ev .
we used a grazing - incidence grating spectrometer @xcite with an @xmath20=5 m spherical grating with 1200 lines / mm in a rowland circle geometry .
the resolution of the excitation radiation was set to about 1 ev by opening the exit slit of the monochromator to 400 @xmath25 m .
the spectrometer had solid angle acceptance of about @xmath26 sr , so that with a spot size of some @xmath27 @xmath25 m at a distance of 5 cm from the entrance slit of the spectrometer we registered a maximum of about 300 counts / minute in the v@xmath1 line at 100 ma ring current . in order to obtain a reasonable statistical accuracy
, we had to acquire for 120 to 240 minutes per single spectrum , depending on the excitation energy .
the angle between the incident x - ray beam and the detection direction was 90@xmath28 .
this minimized the elastic scattering into the spectrometer , because the polarization of the beam coincided with the detection direction .
the v@xmath12 and o@xmath4 absorption spectra had energy resolution of 0.2 ev for an exit slit width of 80 @xmath25 m .
no special surface preparation was undertaken for these measurements .
the band - structure calculations were carried out by the tight - binding linear muffin - tin orbital method ( tb - lmto ) in the atomic sphere approximation ( asa ) @xcite , with the use of the exchange - correlation potential as parametrized by von barth and hedin @xcite and gradient corrections as proposed by langreth and mehl @xcite on top of the local density approximation ( lda ) .
since both rutile and monoclinic structures of vo@xmath0 are relatively loosely packed , it was necessary to include empty spheres in the asa calculations . for the tetragonal rutile phase , we used the crystal structure parameters @xmath29=4.5546 , @xmath30=0.626 and @xmath31=0.300 as given in ref .
besides two formula units of vo@xmath0 , our unit cell included eight equivalent empty spheres , which form long chains parallel to the linear o - v - o fragments , separating the latter .
due to the difference between the v and o atomic sphere radii , the chains are slightly bent .
our choice of the sphere radii @xmath32 , based on the attainability of good matching between the potential at the v and o atomic spheres compatible with a radial sphere overlap below 25% of all interatomic distances , was : @xmath32(v)=2.702 a.u . ;
@xmath32(o)=1.762 a.u . ; @xmath32(empty spheres)=1.619 a.u . the calculated band structure and the total dos are rather close to those obtained earlier in ref .
@xcite by different methods .
we performed our own calculation because no data on the partial dos , that are necessary for the discussion of the xes , were available from the previous calculations .
our calculated total dos for the rutile phase of vo@xmath0 is shown in fig .
[ fig_1 ] , upper panel . as is usual in oxides , the valence band is formed by hybridized transition metal @xmath10 and o@xmath12 states .
the band gap of 0.9 ev within the valence band separates the regions where o@xmath12 states ( below the gap ) and v@xmath10 states ( above the gap ) dominate .
the fermi level crosses the upper subband , revealing a metallic behaviour of vo@xmath0 , as consistent with the results of other calculations done for the rutile - type structure @xcite . the o@xmath33-related subband ( which experiences some hybridization with the v@xmath10 and v@xmath11 states ) lies separately at about 20 ev below the fermi level .
the overall shape of the band structure and of the total dos obtained in our calculation is in agreement with earlier results @xcite ; the width of the gap within the valence band is somewhat larger than 0.62 ev as obtained in the full potential linear augmented plane wave calculation of ref .
the gap width of 4.6 ev reported in ref .
@xcite seems to be too large ( probably due to the use of the slater exchange potential ) and not in agreement with the experimental positioning of individual subbands , as discussed in ref .
@xcite or found in the present paper .
since the room - temperature phase ( for @xmath34c ) of vo@xmath0 is monoclinic , the relevant comparison with experiment needs the calculation data for the latter structure .
the monoclinic unit cell contains four formula units of vo@xmath0 , and , because of the large number of atoms and low symmetry , only few calculations have been done by now .
an earlier non - self - consistent calculation @xcite reproduces an experimentally observable semiconductor band gap , but otherwise seems to be very inaccurate in describing the overall structure of the valence band . the analysis of the structure transformation by means of _ ab initio _ molecular dynamics in ref .
@xcite shows that the monoclinic phase has lower energy than the rutile phase , and the equilibrium positions of the atoms are in good agreement with the experimental determination .
however , the strength of the tendency for opening a gap within the valence band is underestimated within the lda , and the flat bands in the vicinity of the fermi level do not separate completely ( see , e.g. , fig . 3 of ref .
this is also the case in our calculation , which we have carried out for the experimental monoclinic structure as specified in ref .
the change from rutile to monoclinic structure gives rise to a broadening of o@xmath12 and v@xmath10 bands by @xmath350.2 ev and to more pronounced splitting in the v@xmath10 states of @xmath36 symmetry .
in ref.@xcite it was shown that the inclusion of electron - phonon interaction ( in the periodic shell model , based on the results of discrete - variational x@xmath37 cluster calculation ) leads to the opening of the band gap .
according to our calculation , the distortion of the nested bands due to the displacement of the atoms in the doubled - cell monoclinic structure gives rise to only small changes in the partial dos , as compared to the rutile structure .
technically , the calculation deals with two inequivalent oxygen species and three inequivalent types of lattice - packing empty spheres of different sizes making a total of 24 sites .
our calculated total dos for the monoclinic structure is shown in fig .
[ fig_1 ] , lower panel , and some partial dos are shown in fig .
[ fig_2 ] . for comparison with xes
, we did not distinguish the data for two oxygen species and show the averaged o@xmath33 and o@xmath12 dos over all sites in fig .
[ fig_2 ] , lowest panel .
the densities of states are plotted for the @xmath38 mesh over the brillouin zone .
= 12.5 cm the intensity distribution of the vb xps spectra reflects the total dos of the vb , up to the deviations due to different atomic photoionization cross - sections .
the results of the xps measurements of single crystal vo@xmath0 are shown in fig .
[ fig_3 ] and in the table 1 .
= 12.0 cm the spectral measurements on the uncleaved crystal show a large contamination with carbon .
this leads to a smearing of the fine structure in the entire valence band , a broadening of subbands and the appearance of an additional subband around 1214 ev which we attribute to transitions from c@xmath33 states . in the xps spectra of the cleaved vo@xmath0 crystal ,
a distinct narrow peak is observed around 1 ev below the fermi level , which , according to the results of our band structure calculations ( see fig . [ fig_2 ] ) , has v@xmath10 character .
the fwhm of this peak is about 1 ev which is less than obtained in ref .
@xcite . core level ( band ) & binding energy ( ev ) & fwhm ( ev ) + v@xmath33 & 630.02 & 6.10 + o@xmath4 & 529.75 & 1.30 + v@xmath39 & 523.48 & 2.56 + v@xmath40 & 515.95 & 2.04 + v@xmath41 & 68.95 & 4.54 + v@xmath42 & 40.53 & 4.24 + o@xmath33 & 21.73 & 2.5 + o@xmath12 & & + v@xmath10 & 1.03 & + the next subband has a two - peak structure at 39 ev below the fermi level which are mainly due to the o@xmath12 states ( _ cf . _
[ fig_2 ] ) .
another reason for reaching this conclusion is that the energy difference between the centre of gravity of this band and the next one ( located at 22 ev ) is about 15 ev , which is in good agreement with the energy separation between o@xmath12 and o@xmath33 bands found in all vanadium oxides @xcite .
however , since the o@xmath12 and v@xmath10 atomic photoionization cross - sections for al@xmath3 excitation @xcite have a ratio of about 1:2 , one may expect some contribution of v@xmath10 states in this band .
this is also confirmed by our band structure calculation and by the results from the v@xmath1-emission measurements ( see below ) .
we point out , however , that there is a contradiction between the xps and the ultraviolet photoelectron spectra ( ups ) of vo@xmath0 given in refs .
@xcite with respect to energy resolution and the intensity ratio of the two peaks .
the peaks are well - resolved in our measurements , and both their energy separation and the intensity ratio are in good agreement with results of our band structure calculations . according to ref .
@xcite , charge - transfer satellites are found in the ups of scf@xmath43 , tio@xmath0 and v@xmath0o@xmath44 in the region from 11 to 17 ev below the top of the valence band . from the resonant photoemission measurements
it was concluded that the charge - transfer - type configuration manifests itself in such a way as to form a valence state in the ground state in addition to the originally filled @xmath12 state ligand .
on the basis of this experimental evidence , cluster calculations of the valence photoemission and bremsstrahlung isochromat spectra of vo@xmath0 were performed in ref .
@xcite , and it was concluded that vo@xmath0 belongs to the group of charge - transfer insulators .
however , we point out that our xps measurements of the uncleaved and cleaved vo@xmath0 single crystal ( fig . [ fig_3 ] ) strongly suggest that the appearance of this structure ( 14 - 15 ev below the fermi level ) is connected with carbon contamination and that its origin is c@xmath33 states .
a similar problem was discussed in ref .
@xcite in connection with the analysis of ups spectra of superconducting cuprates .
there , it was concluded that the satellites with binding energies of about 10 ev are generically connected with contaminations of light elements and disappear after cleaning .
therefore we have found a clear indication that the structure in question is not an intrinsic feature of the electronic structure of pure vo@xmath0 and can not be considered as an evidence of correlation effects in vo@xmath0 .
= 11.0 cm the xps core level spectra of vo@xmath0 are shown in fig .
[ fig_4 ] .
the energy positions of the o@xmath4 , v@xmath12 and v@xmath42 spectra of vo@xmath0 are close to those of ref .
@xcite , but the fwhm s of the lines in these spectra ( see table 1 ) are much smaller than it was found in refs .
the lineshapes are clearly asymmetric due to a structure seen at the high - binding energy side . in case of @xmath12 and @xmath42 levels this may be related to @xmath45 multiplets , as was shown in ref .
@xcite for the v@xmath42 spectrum
. it should be noted , however , that in our measurements the v@xmath42 lineshape is somewhat different from that obtained in ref .
@xcite . in case of v@xmath41 level
, the additional structure may be due to the exchange interaction of a ( spin - up or spin - down ) 3@xmath46 electron left in the final state with the ( v@xmath10 ) electrons in the valence band@xcite . in . fig .
[ fig_4 ] , the v@xmath12 , v@xmath41 and v@xmath42 xps core level spectra of vo@xmath0 are compared with those of pure vanadium metal .
evidently , there is a chemical shift of the core level spectra of vo@xmath0 with respect to that of pure metal not only for the v@xmath12 xps - line , but also for the v@xmath41 and v@xmath42 xps - lines .
the fact that the formal charge of the vanadium atom affects v@xmath41 and v@xmath42 lines in an uniform way can be used for an estimation of the oxidation state of the v atom in compounds .
recently , the same conclusion was found for the mn@xmath41 xps - spectra of manganese complexes in ref .
@xcite .
earlier , we have shown that the analysis of @xmath41 xps - spectra can be used to draw conclusions about the electronic structure of @xmath10 metal oxides @xcite .
it has been found that @xmath41 xps - spectra of nio have a very complicated fine structure due to charge - transfer processes . in this case
, an electron may be transferred from the ligand to a metal @xmath10-level after the emission process , and both states , the screened and unscreened one , are visible in the @xmath41 spectrum .
the simpler fine structure of the v@xmath41 xps - spectrum in vo@xmath0 can be considered as an evidence for negligible charge transfer , again indicating that the electronic structure of this compound is more bandlike than correlated .
x - ray emission valence spectra result from electron transitions between the valence - band and a core hole .
since the wave function of a core electron is strongly localized and its angular momentum symmetry is well defined , these spectra reflect the site - projected and symmetry - restricted ( in accordance with the dipole selection rules ) partial dos . in the case of vo@xmath0
, we have investigated v@xmath1- ( @xmath47 transition ) , vk@xmath48- ( @xmath8 transition ) and o@xmath3-emission spectra ( @xmath9 transition ) which reflect the distribution of v@xmath49 , v@xmath11 and o@xmath12 partial dos . by xps
we can measure the binding energies of the v@xmath12 and the o@xmath4 core levels so that we can determine the position of the fermi level in the x - ray emission spectra .
the o@xmath3 and v@xmath1 spectra , adjusted in such a way as to have a a common energy scale with the x - ray photoemission spectrum , are shown in fig .
[ fig_5 ] .
the v@xmath2 spectrum is positioned with respect to xps by matching the position of the peak which results from the hybridization with o@xmath33 states . in the same figure , the partial dos as calculated for the monoclinic phase and broadened with an energy - dependent lorentzian linewidth according to ref .
@xcite are shown .
calculated and broadened partial dos have been rigidly shifted with respect to the fermi energy ( arbitrarily , but by the same value for all spectra ) to account for a systematic error in the energy matching of the measured spectra and calculated dos due to , e.g. , the absense of the band gap in the calculation .
as is seen , the main maximum of the o@xmath3 spectrum ( o@xmath12 states ) is placed near the top of the valence band , that is quite common in oxides , and is essentially nonbonding .
a corresponding feature is also seen in the xps but merely as a shoulder at 4 ev , due to a smaller value of o@xmath12 photoionization cross section as compared to v@xmath10 @xcite .
xps vb fine sructure with two peaks at 5.4 and 7.5 ev reveals the o@xmath12v@xmath10 bonding band , which overlaps in energy with the maximum of the v@xmath1 spectrum and with a high - binding energy hump of the o@xmath3 spectrum .
= 12.0 cm we have excited o@xmath3 emission at two excitation energies ( @xmath50=530.8 and 532.2 ev ) which correspond to the maxima of o@xmath4 x - ray absorption spectrum ( xas ) , but found only little difference . the v@xmath1-emission spectrum has been obtained with both electron and photon excitation , that resulted in different intensity distributions ( curves 1 and 2 in fig .
[ fig_5 ] ) . with electron excitation at @xmath50=4.4 kev ( curve 1 )
, we have simultaneous excitation of v@xmath1 ( @xmath17 ) and v@xmath51-emission ( @xmath52 ) which are separated by only 7.53 ev
. therefore there is an overlap of the high - energy subband of v@xmath1 xes with the main peak of the v@xmath51-emission spectrum . for an excitation energy @xmath50=519.0 ev ( curve 2 )
, we selectively excite v@xmath1 which follows electron transitions from the occupied part of the v@xmath10 band . from these data
one can conclude that there is a considerable admixture of v@xmath10 states in the o@xmath12-like bands , so that the maximum emission comes from the v@xmath10o@xmath12 hybridized band and not from an only slightly populated v@xmath10 band .
we find that the fine structure and energy position of the subbands of the v@xmath1-emission spectrum are in reasonable agreement with the calculated v@xmath10 partial dos distribution given in fig .
[ fig_2 ] .
the v@xmath2-emission spectrum has two main subbands whose energy positions are very close to the o@xmath12 and the o@xmath33 bands due to v@xmath11o@xmath12 and v@xmath11o@xmath33 hybridization ( see fig . [ fig_2 ] ) .
the splitting of the main subband of the v@xmath2-emission spectrum also follows the calculated v@xmath11 dos distribution in this energy region .
it is seen that the admixture of the v@xmath11 states to the v@xmath10 band is too small to be detected in the v@xmath2 xes of vo@xmath0 .
the disagreement by @xmath352 ev between spectra ( xps ; v@xmath2 ) and calculated dos in what regards the position of the o@xmath33 states is known in oxides .
it is related to the fact that the hole relaxation , which effectively increases the binding energy of an electron leaving a comparatively localized state such as o@xmath33 , is not taken into account in our band structure calculations which describe the ground state but not excitations .
it should be mentioned that soft x - ray fluorescence spectra of vo@xmath0 ( v@xmath53 , o@xmath3 xes ) were also recently reported in ref .
@xcite and are compared with an ups spectrum ( measured at @xmath50=501.1 ev ) : however , no xps measurements of the v@xmath12 and the o@xmath4 binding energies are included there .
the o@xmath3 xes of ref .
@xcite does not show the two - peak structure found in our measurements .
o@xmath4 xas of vo@xmath0 probes the o@xmath12 unoccupied states .
such spectra , with good energy resolution , have been published earlier , by abbate _
_ @xcite . in order to compare the xes obtained on the same samples which were used in the xps , we include in fig .
[ fig_5 ] our own o@xmath4 spectrum obtained ( for the monoclinic phase ) in the sample drain - current mode .
the fermi level has been determined with the help of the xps binding energy of the o@xmath4 level given in table 1 .
the o@xmath4 spectrum was measured in a rather restricted energy range , so that we can only compare the spectra in the vicinity of the fermi level .
the fine structure of the o@xmath4 spectrum is in agreement with the shape of the o@xmath12 conduction band in our electronic structure calculations ( fig .
[ fig_2 ] ) . in the high - temperature rutile phase ,
the calculated conduction band exhibits a pronounced two - peak structure , also discussed in ref .
@xcite . in the low - temperature monoclinic phase ,
a further splitting of the conduction band is seen from the calculations , in agreement with a more pronounced structure in the o@xmath4 spectrum of ref .
the results of measurements of high - resolution x - ray photoelectron spectra , v@xmath1 , v@xmath2 and o@xmath3 x - ray emission spectra ( obtained by using both electron and x - ray excitation ) and o@xmath4-absorption spectra of a vo@xmath0 single crystal are presented .
they are compared with first - principles lmto band structure calculations of vo@xmath0 in monoclinic and tetragonal rutile phases .
it is concluded that the electronic structure of vo@xmath0 is more bandlike than correlated .
financial support by the deutsche forschungsgemeinschaft ( sfb 225 ) , the nato ( grant no .
htech.lg971222 ) , and the russian foundation for fundamental research ( projects no 96 - 03 - 32092 and 96 - 15 - 96598 ) is gratefully acknowledged .
one of us ( e.z.k ) wants to thank the university of osnabrck for generous hospitality during his stay .
the uppsala group gratefully acknowledges financial support by the swedish natural science research council ( nfr ) and the gran gustafsson foundation and we are indebted to hasylab - desy for the excellent facilities put to our disposal .
99 adler d 1968 _ solid state phys . _ * 21 * 1 goodenough j b 1960 _ phys
. rev . _ * 117 * 1442 gupta m , freeman a j and ellis d e 1977 _ phys .
b _ * 16 * 3338 wentzcovitch r m , schulz w w and allen p b 1994 _ phys
. rev .
lett . _ * 72 * 3389 sommers c , de groot r , kaplan d and zylbersztein a 1975 _ j. phys .
( paris ) _ * 36 * l157 uozumi t , okada k and kotani a 1993 _ j. phys . soc .
jpn . _ * 62 * 2595 zaanen j , sawatzky g a and allen j w 1985 _ phys .
* 55 * 418 blaauw c , leenhouts f , van der woude f and sawatzky g a 1975 _ j. phys . c _ * 8 * , 459 sawatzky g a and post d 1979 _ phys . rev .
b _ * 20 * , 1546 shin s , suga s , taniguchi m , fujisawa f , kanzaki h , fujimori a , daimon h , ueda y , kosuge k and kachi s 1990 _ phys . rev .
b _ * 41 * , 4993 shin s , agui a , watanabe m , fujisawa m , tezuka y , and ishii t 1995 _ techn .
report of issp , ser . a , no . 3038 _ goering e ,
schramme m , mller o , barth r , paulin h , klemm m , denboer m l and horn s 1997 _ phys .
b _ * 55 * , 4225 goering e , schramme m , mller o , paulin h , klemm m , denboer m l and horn s 1997 _ physica b _ * 230 - 232 * , 996 dolgih v e , cherkashenko v m , kurmaev e z , goganov d a , ovchinnikov e k and yarmoshenko yu m 1984 _ nucl . instrum . methods .
* 224 * , 117 pfugler j and gurtler p 1993 _ ibid .
_ a*287 * 628 nordgren j , bray g , gramm s , nyholm r , rubensson j e and wassdahl n 1989 _ rev .
_ * 60 * 1690 andersen o k and jepsen o 1984 _ phys .
lett . _ * 53 * 2571 ; andersen o k , pawlowska z and jepsen o 1986 _ phys .
b _ * 34 * 5253 von barth u and hedin l 1972 _ j. phys
. c _ * 5 * 1629 langreth d c and mehl m j 1981 _ phys . rev .
lett . _ * 47 * 446 ; langreth d c and mehl m j 1983 _ phys . rev
. b _ * 28 * 131 mcwhan d b , marezio m , remeika j p and dernier p d 1974 _ phys .
b _ * 10 * 490 caruthers e , kleinman l and zhang h i 1973 _ phys .
b _ * 7 * 3753 nikolaev a v , kostrubov yu
n and andreev b n 1992 _ fiz .
tela ( st .
petersburg ) _ * 34 * 3011 [ _ sov .
phys . solid state _ * 34 * 1614 ] .
caruthers e and kleinman l 1973 _ phys .
b _ * 7 * 3760 andersen g 1956 _ acta chem . scand . _ * 10 * , 623 nakatsugawa h and iguchi e 1997 _ phys . rev . b _ * 55 * 2157 cherkashenko v m , kurmaev e z , fotiev a a and volkov v l 1975 _ sov . phys . solid state _ * 17 * , 167 yeh j j and lindau i 1985 _ atomic data and nuclear data tables _ * 32 * 1 shin s , tezuka y , ishii t and ueda y 1993 _ solid state commun . _
* 87 * 1051 takahashi t , maeda f , katayama - yoshida h , okabe y , suzuki t , fujimori a , hosoya s , shamoto s and sato m 1988 _ phys .
b _ * 37 * 9788 van vleck j h 1934 _ phys .
rev . _ * 15 * 405 fujiwara m , matsushita t and ikeda s 1995 _ j. electr .
* 74 * , 201 uhlenbrock st , bartkowski st , postnikov a v , mayer b , neumann m , galakhov v r , finkelstein l d , kurmaev e z and leonyuk l i 1998 unpublished blokhin m a and sachenko v p 1960 _ izv .
nauk sssr , ser .
* 24 * , 397 abbate m , pen h , de groot f m f , fuggle j c , ma y j , chen c t , sette f , fujimori a , ueda y and kosuge k 1991 _ phys
b _ * 43 * , 7263 | a vo@xmath0 single - crystal has been subject of a combined investigation by high resolution x - ray photoelectron spectroscopy ( xps ) , x - ray emission spectroscopy ( xes ) with both electron and energy - selective x - ray excitation ( v@xmath1- , v@xmath2- and o@xmath3-emission ) and x - ray absorption spectroscopy ( xas ) ( o@xmath4 ) .
we performed first principles tight - binding lmto band structure calculations of vo@xmath0 in both monoclinic and tetragonal rutile structures and compare the densities of states ( dos ) with the experimental data . from this
we conclude that the electronic structure of vo@xmath0 is more bandlike than correlated . |
by the farkas - minkowski - weyl theorem a convex polyhedron in @xmath0 has two representations .
it can either be described by a finite set of linear inequalities ( facets ) or by a finite set of generators ( vertices and rays ) .
precise definitions are given in section [ sec : basic_notations ] .
one of the most fundamental problems in the theory of polyhedra and its applications , such as combinatorial optimization or computational geometry , is the conversion between two different descriptions .
many algorithms for this representation conversion have been proposed ( see for example @xcite , @xcite , @xcite , @xcite ) . for certain classes of polyhedra efficient methods
are known , but there is no approach known which efficiently solves the problem in general .
programs like cdd @xcite , lrs @xcite , pd @xcite , porta @xcite ( or polymake @xcite either relying on some of the others or using its own method closely related to ` cdd ` ) allow conversion of the representation of a polyhedron .
since the programs are implementations of quite different methods , their efficiency may vary tremendously on a given example .
many interesting polyhedra both pose difficulties for the standard representation conversion approaches and have many symmetries that could potentially be exploited . in many applications
it is sufficient ( or at least necessary ) to obtain a list of inequalities or generators up to symmetries . in the present paper
we give a brief survey of approaches that can be used for the representation conversion problem up to symmetries .
we do not discuss their asymptotic complexity ( which , in the worst case , is not encouraging ) , but rather refer to previous papers where the methods have proven themselves on difficult instances that could not otherwise be solved . for the new approach discussed in section [ sec : sym - pivoting ] we provide some experimental data ourselves .
the paper is organized as follows . in section [ sec : basic_notations ] we give some basic notations and facts from the theory of convex polyhedra . in section [ sec : symmetries ] we consider different notions of symmetries and describe how they can practically be obtained using a graph automorphism computation . in section
[ sec : orbits ] we describe the group theoretical notions used in the representation conversion methods discussed in the remainder of the paper . in section
[ sec : subpolytope ] we consider decomposition methods which reduce the given problem to a number of smaller problems .
these approaches have been used quite successfully by different authors . in section [ sec : cascade ] we describe the incremental _ cascade algorithm _ and in section [ sec : sym - pivoting ] we show how symmetry can be exploited in a simplex pivot based algorithm .
in this section , we give a brief introduction to some basic concepts and terminology of ( convex ) polyhedra . for more details on polyhedra
, we refer to the books @xcite , @xcite , @xcite . given the vector space @xmath0 , denote by @xmath1 its dual vector space , i.e. the vector space of linear functionals on @xmath2 .
a _ convex polyhedron _
@xmath3 can be defined by a finite set of linear inequalities @xmath4 with @xmath5 and @xmath6 for @xmath7 .
if the number of inequalities @xmath8 in the description is minimum , we speak of a non - redundant description .
the dimension @xmath9 of @xmath10 is the dimension of the smallest affine subspace containing it . under the assumption that @xmath10 is full - dimensional , i.e. @xmath11
, every inequality @xmath12 of a non - redundant description defines a _ facet _ @xmath13 of @xmath10 , which is a @xmath14-dimensional convex polyhedron contained in the boundary of @xmath10 . by the farkas - minkowski - weyl theorem ( see e.g. @xcite , corollary 7.1a ) , @xmath10 can also be described by a finite set of generators : @xmath15 where @xmath16 for @xmath17 .
if the number of generators is minimum , the description is again called _ non - redundant_. in the non - redundant case , the generators @xmath18 , @xmath19 , are called _ vertices _ and @xmath20 , @xmath21 , are the _ extreme rays _ of @xmath10 . in case @xmath10 is bounded we have @xmath22 and we speak of a _
convex polytope_. the _ representation conversion _ from a minimal set of generators into a minimal set of linear functionals ( or vice versa ) is called the _ dual description problem_. by using homogeneous coordinates , the general inhomogeneous problem stated above can be reduced to the homogeneous one where @xmath23 and @xmath24 .
for example , we embed @xmath25 in the hyperplane @xmath26 in @xmath27 and consider the closure of its conic hull .
the so - obtained polyhedron @xmath28 , with @xmath29 for @xmath30 and @xmath31 for @xmath32 , is referred to as _ polyhedral cone_. by duality , the problem of converting a description by homogeneous inequalities into a description by extreme rays is equivalent to the opposite conversion problem .
so for simplicity we assume from now on that @xmath3 is a _
polyhedral cone _ given by a minimal ( non - redundant ) set of generators ( extreme rays ) .
if @xmath33 we say that @xmath10 is generated by @xmath34 .
we want to find a minimal set @xmath35 with @xmath36 by choosing a suitable projection , it is possible to reduce the problem further to the case where @xmath10 is full - dimensional and does not contain any non - trivial linear subspaces .
for example , if @xmath34 span a @xmath37-dimensional linear subspace , we may just choose ( project onto ) @xmath37 independent coordinates . the appropriate projection ( i.e. equations of the linearity space ) can be found efficiently via gaussian elimination .
in other words , without loss of generality we may assume the @xmath18 span @xmath0 ( @xmath10 is full - dimensional ) and the linear inequalities @xmath38 span @xmath1 ( @xmath10 is pointed ) .
a _ face _ of @xmath10 is a set @xmath39 where @xmath40 is an element of @xmath41 the polyhedral cone _
dual _ to @xmath10 .
note that @xmath42 .
the faces of a pointed polyhedral cone are themselves pointed polyhedral cones .
we speak of a @xmath37-face , if its dimension is @xmath37 .
the faces form a ( combinatorial ) lattice ordered by inclusion , the _ face lattice _ of @xmath10 .
the rank of a face in the lattice is given by its dimension .
each face is generated by a subset of the generators of @xmath10 and therefore it is uniquely identified by some subset of @xmath43 . in particular , the @xmath44-dimensional face is identified with the empty set @xmath45 and @xmath10 itself with the full index set @xmath43 .
all other faces of @xmath10 are identified with some strict , non - empty subset @xmath46 .
we write @xmath47 for two faces of @xmath10 with @xmath48 and @xmath49 .
two @xmath37-faces of @xmath10 are said to be _ adjacent _ , if they contain a common @xmath50-face and are contained in a common @xmath51-face .
in particular , two extreme rays ( @xmath52-faces ) are adjacent , if they generate a common @xmath53-face and two facets are adjacent , if they share a common @xmath54-face ( a _ ridge _ ) . by the properties of a lattice , for two faces @xmath55 and @xmath56 , not necessarily of the same dimension , there is always a unique largest common face @xmath57 contained in them , and a unique smallest face @xmath58 , containing both .
any sub - lattice @xmath59 $ ] of the face lattice consisting of all faces containing @xmath60 and contained in @xmath58 is known to be isomorphic to the face lattice of some pointed polyhedral cone of dimension @xmath61 .
note that this is a special feature of face lattices of polyhedra , which is not true for general lattices .
in particular the _ diamond property _ holds for face lattices : every sub - lattice of rank @xmath53 carries the combinatorics of a @xmath53-dimensional polyhedral cone , namely a face @xmath60 of rank @xmath62 , a face @xmath58 of rank @xmath63 and two faces @xmath55 and @xmath56 of rank @xmath37 . the rank @xmath53 sub - lattices are in one - to - one correspondence to pairs of adjacent faces .
a _ polyhedral complex _
@xmath64 is a set of polyhedral cones ( the _ cells _ of @xmath64 ) satisfying the following two properties : 1 . if @xmath65 then every face of @xmath10 is also in @xmath64 .
2 . for all @xmath66 , @xmath67 is a face of both @xmath68 and @xmath69 . the facets of a polyhedral cone
@xmath10 form a polyhedral complex , the _ boundary complex _ of @xmath10 .
polyhedral complex @xmath70 is a _ subdivision _ of polyhedral complex @xmath64 if every cell of @xmath64 is the union of cells in @xmath70 and every cell of @xmath70 is contained in some cell of @xmath64 .
a subdivision is called a _ triangulation _
if every cell is a simplicial cone , i.e. is the conic hull of exactly @xmath71-extreme rays . by the homogeneous embedding of polytopes
discussed above , we may equally discuss boundary complexes , subdivisions , and triangulations for polytopes .
two faces ( cells ) of a polytopal complex have empty intersection exactly when the corresponding cones intersect only at the origin .
in this paper we are especially interested in the case where some non - trivial group acts on the given polyhedron @xmath10 .
the _ combinatorial automorphism group _ of @xmath10 , generated by @xmath72 extreme rays , is the subgroup of @xmath73 of all permutations ( acting on ray indices @xmath43 ) which preserve the whole face lattice of @xmath10 .
thus the combinatorial automorphism group acts not only on the generating rays , but also on the set of facets ( inequalities ) , and more generally on the sets of faces of a given dimension .
moreover , it preserves the inclusion relation among faces .
it is known that the combinatorial automorphism group of @xmath10 can be computed from the incidence relations between extreme rays and facets ( see @xcite ) .
given generators @xmath74 of a polyhedral cone @xmath10 and a subgroup @xmath75 of the combinatorial automorphism group of @xmath10 , we want to obtain the facets of @xmath10 up to symmetry , that is , one representative @xmath46 from each orbit under the action of @xmath75 . if no group or only a small group is given , we usually still want to exploit as much symmetries of the given polyhedron as possible in the representation conversion .
that is , ideally we would like to compute the full combinatorial automorphism group .
however , we do not know how to compute it without computing the facets , which is precisely the problem we want to solve .
so we have to settle for a compromise and work with a more restricted type of automorphism .
note though , that one may obtain the full combinatorial automorphism group of @xmath10 , after the facets have been computed in many cases the combinatorial automorphism group ( or some nontrivial subgroup ) may reflect geometric symmetries , for example if it has a representation as a subgroup of @xmath76 acting naturally on @xmath0 and the polyhedral cone @xmath10 .
the group of all matrices in @xmath76 preserving @xmath10 is called _ linear automorphism group _ of @xmath10 .
since a linear automorphism permutes the set of extreme rays @xmath77 , we naturally obtain a representation as a permutation group @xmath78 .
it is important to note here that although the induced permutation group is finite , the linear automorphism group of @xmath10 is not necessarily so , and it can be quite awkward .
think for example of the quadrant in @xmath79 generated by the two non - negative coordinate axes .
the induced permutation group is @xmath80 , its linear automorphism group however is @xmath81 as explained in section [ sec : basic_notations ] we may limit our discussion to the special case of a full - dimensional , pointed polyhedral cone @xmath10 generated by @xmath82 .
the elements @xmath83 of the linear automorphism group of @xmath10 satisfy @xmath84 , where @xmath85 is an induced permutation and @xmath86 .
note , in case @xmath10 is the homogenization of a @xmath14-dimensional polyhedron @xmath87 , the linear automorphism group of @xmath10 corresponds to the so called _ projective automorphism group _ of @xmath87 , that is , the set of all projective maps preserving @xmath87 .
we are not aware of any practical algorithm to compute linear ( or projective ) automorphism groups or to decide if two polyhedral cones are linear ( or projective ) isomorphic .
a _ restricted isomorphism _ of two full - dimensional vector families @xmath88 and @xmath89 in @xmath0 is given by a matrix @xmath90 such that there exists a permutation @xmath91 satisfying @xmath92 for @xmath93 . a _
restricted automorphism _ of a vector family is a restricted isomorphism of @xmath94 with itself .
we speak of a restricted isomorphism between two polyhedral cones generated by the two vector families @xmath94 and @xmath95 , if it is a restricted isomorphism between @xmath94 and @xmath95 .
note though that the definition of restricted isomorphisms for polyhedral cones depends strongly on the choice of generators .
since such generators are only unique up to any positive factor , the choice of generators is very crucial . in practice
the situation is usually not so bad as there is often a natural choice for the generators .
the big advantage of restricted isomorphisms is that we can compute them by obtaining the graph isomorphisms for an edge colored graph . given a vector family @xmath96 spanning @xmath0 we define the positive definite matrix @xmath97 let the graph @xmath98 be the complete graph with vertices @xmath18 and edge colours @xmath99 . then the following holds : [ prop : isomorphism - equivalence ] let @xmath100 be two finite vector families .
then every isomorphism of the edge coloured graphs @xmath98 and @xmath101 yields a restricted isomorphism of @xmath94 and @xmath95 and vice versa .
in practice the popular and very nice program nauty @xcite by mckay can be used to check for graph isomorphisms or compute the group of automorphisms .
note however , that the current version only takes vertex coloured graphs as input .
we therefore have to transform our edge coloured graph to a somewhat larger vertex coloured graph that preserves the automorphism group ( see for example @xcite , p.25 ) .
let @xmath102 be the matrix obtained from @xmath94 .
denote by @xmath103 the unique square root of @xmath104 , that is , the positive definite @xmath105 matrix @xmath103 with @xmath106 .
let @xmath107 for @xmath17 .
then the edge colours @xmath108 of the graph @xmath98 are exactly the inner products @xmath109 of the transformed vectors @xmath110 and @xmath111 . in the same way we obtain @xmath112 , @xmath113 and @xmath114 for @xmath95 .
now let @xmath83 be a restricted isomorphism for @xmath94 and @xmath95 with associated permutation @xmath85 .
then we have @xmath115 .
moreover , the matrix @xmath116 is orthogonal and satisfies @xmath117 .
this implies @xmath118 , i.e. the restricted isomorphism of the vector families @xmath94 and @xmath95 corresponds to an isomorphism between the edge coloured graphs @xmath98 and @xmath101 .
suppose on the other hand @xmath119 be an isomorphism between the edge coloured graphs @xmath98 and @xmath101 . by reordering elements
, we may simply assume @xmath120 .
instead of looking for a solution of @xmath121 we consider the equivalent equations @xmath122 .
since the @xmath18 generate @xmath0 , we find a basis @xmath123 of @xmath0 .
if @xmath124 , respectively @xmath125 is the @xmath105 matrix formed by @xmath126 , respectively @xmath127 , then the matrix equation @xmath128 takes the form @xmath129 .
so , the matrix @xmath130 is orthogonal and we have for any @xmath131 and @xmath132 : @xmath133 this yields @xmath134 and since the @xmath114 form a basis of @xmath0 the relation @xmath135 are implied .
thus we obtain a restricted isomorphism between @xmath94 and @xmath95 .
in order to generate facets ( or more generally faces ) of a polyhedron up to symmetries , it is necessary to deal with orbits of faces . in this section
we briefly indicate what the basic tasks are that we have to accomplish and how these can be approached . as before
, we assume that a polyhedral cone @xmath10 is given by a set of generators @xmath74 which define the extreme rays of @xmath10 .
each face is represented by a subset of @xmath136 which corresponds to the indices of generators incident to the face .
we assume @xmath78 is some subgroup of the combinatorial automorphism group of @xmath10 , hence a permutation group acting not only on the set of indices @xmath136 ( respectively rays ) , but also on the whole face lattice of @xmath10 .
in particular , the dimension of a face and inclusion between faces are preserved by every group element . in the most general form ,
the problem we have to solve is the following : given two subgroups @xmath137 and @xmath138 of @xmath73 and a list @xmath139 of @xmath137-inequivalent faces , we need to be able to obtain a list @xmath140 of @xmath138-inequivalent faces .
for example , if @xmath141 is trivial , the list @xmath142 simply is a list of pairwise unequal faces and we may obtain @xmath140 by testing for @xmath138-equivalence : starting with @xmath143 and then subsequently adding @xmath144 for @xmath145 to @xmath140 , if it is not @xmath138-equivalent to any element in @xmath140 . clearly , this _ orbit fusion _ can be applied whenever @xmath146 . in case @xmath147 it is necessary to `` break some symmetry '' and _ split ( factorize ) orbits_. this can be done with the _
double coset decomposition _
, described in section [ sec : canonical - rep ] .
note , if neither @xmath146 nor @xmath147 , one could in principle convert the @xmath137-orbits into @xmath138-orbits in two steps , by either using the intersection group @xmath148 or the group @xmath149 generated by @xmath137 and @xmath138 and obtaining @xmath75-orbits in an intermediate step .
orbit fusing and splitting is a typical and essential task when generating discrete structures up to isomorphism ( see for example @xcite , @xcite , @xcite , @xcite ) .
a common task is to decide whether or not two faces are @xmath75-equivalent , that is , whether or not the corresponding subsets of @xmath136 lie in the same orbit under the action of @xmath75 .
the dimension of a face ( rank in the face lattice ) and its cardinality are quickly testable invariants .
other invariants can easily be obtained , for example by looking at the action of @xmath75 on pairs , triples , or other @xmath37-tuples of indices ( generators ) , respectively on lower dimensional faces .
the number of elements from each such orbit included in a face is a @xmath75-invariant .
unfortunately , there is no clear rule how many such sets have to be chosen , so one has to rely on heuristics .
if we only consider restricted isomorphisms , metric invariants can be used as well , for example the set of pairwise inner products between generators , discussed further in section [ sec : metric - invariants ] . when all invariants are satisfied , then group computations must be done . for small groups
it is possible to simply generate the whole orbit of a face .
if the size of available memory is a problem , we may just keep a _
canonical representative _ for the orbit , for example the lexicographical minimum ( when viewed as a subset of @xmath136 ) . these can be found for example by a backtrack method ( see section [ sec : implementing_orbits ] ) .
assume @xmath150 and that a list of @xmath137-orbits is given . in order to obtain a list of @xmath138-orbits , we may split each orbit @xmath151 , with a representative face @xmath60 , by the well known _
double coset decomposition _
for example in @xcite and @xcite ) : the group @xmath137 can be decomposed into _ double cosets _
@xmath152 where @xmath153 are elements of @xmath137
. then the orbit @xmath151 is decomposed into @xmath154 hence into @xmath155 orbits @xmath156 with @xmath157 .
note that the more straightforward algorithm of generating the full orbit or of computing a decomposition of @xmath137 into cosets @xmath158 is slower and/or requires more memory ( see @xcite ) .
the fundamental data structures used to work with permutation groups are _ bases and strong generating sets _ ( bsgs ) ( see for example @xcite , @xcite ) . based on them a backtrack search on cosets of a point stabilizer chain can be used to obtain canonical representatives , to decide on ( non-)equivalence as well as to obtain stabilizers of subsets of @xmath136 ( faces ) .
for details we refer to @xcite and @xcite .
an elaborate version is the _ partition backtrack _ introduced by leon @xcite ( cf .
@xcite , @xcite ) .
these methods are known to work quite well in practice , although from a complexity point of view the problems are known to be difficult .
that is , there is no ( worst - case ) polynomial time algorithm known to solve these problems .
even worse , it is somewhat unlikely that there exist polynomial time algorithms , since the _ graph isomorphism problem _ , not known to be in @xmath159 , is reducible to them in polynomial time ( cf .
@xcite ) .
the computer algebra system gap @xcite provides functions for generation of full orbits ( orbit ) stabilizer computations ( stabilizer ) and equivalence tests ( representativeaction ) . in the case where we have special knowledge of the group @xmath75 or its representation
, it might be much easier to obtain canonical representatives or to compute stabilizers of faces .
for example , we may have a situation where the symmetric group @xmath73 acts on @xmath72 elements ( see @xcite and @xcite ) .
another example where the action of the symmetric group is used is described in @xcite ( see @xcite for corresponding computer code ) . typically , polyhedra arising in combinatorial optimization are convex hulls of @xmath160-vectors , where each coordinate ( variable ) represents an edge of a complete directed or undirected graph with @xmath72 vertices on which @xmath73 acts . in @xcite a method for obtaining canonical representatives in this situation is described .
in this section we describe two methods which reduce the facet generation problem under symmetries to a number of smaller instances of facet generation problems .
in contrast to the original problem , solving the smaller problems ( for sub - cones ) may be feasible for available software , such as cdd or lrs .
these techniques have been proven to be successful in practice , in cases where standard approaches failed .
we mainly distinguish two approaches , the _ incidence decomposition method _
( see section [ sec : incident ] ) and the _ adjacency decomposition method _ ( see section [ sec : adjacency ] ) .
both methods are reasonably natural and have been used separately by different authors .
the incidence decomposition method reduces the problem of facet generation to a number of smaller problems , in which we generate facets that are incident to some extreme rays . as before let @xmath10 be a polyhedral cone in @xmath0 , generated by @xmath161 .
assume @xmath78 is some permutation group acting on the face lattice of @xmath10 .
the set of extreme rays ( indices ) falls into orbits under the action of @xmath75 . for each orbit
we consider a representative @xmath162 ( index @xmath12 ) and generate a list of @xmath75-inequivalent facets of @xmath10 incident to it .
then , in a post - processing step the lists of facets obtained in this way are merged to a list of @xmath75-inequivalent facets of @xmath10 ( see section [ sec : orbits ] ) . since every @xmath75-orbit of facets of @xmath10 contains a facet which is incident to one of the chosen representatives , the resulting list is complete
. * input : * @xmath72 extreme rays of a polyhedral cone @xmath10 and a group @xmath163 acting on @xmath10 s face lattice . + * output : * complete set @xmath164 of @xmath75-inequivalent facets of @xmath10 .
+ @xmath165 .
+ @xmath166 .
+ * for * @xmath167 * do * + @xmath168 .
+ * for * @xmath169 * do * + * if * @xmath60 is @xmath75-inequivalent to facets in @xmath164 * then * + @xmath170 . + * end if * + * end for * + * end for * + the main computational gain comes from the following : when we compute the facets that are incident to a given ray @xmath162 , we may not have to consider all @xmath72 extreme rays of @xmath10 , because some of the rays may not be incident to facets which are incident to @xmath162 .
these are exactly the extreme rays @xmath171 which give redundant inequalities of the polyhedron @xmath172 so , for each of the @xmath173 rays with index in @xmath174 we may solve a linear program in @xmath175 to decide redundancy . as an outcome we obtain a list of @xmath176 rays needed in the definition of @xmath177 .
the smaller @xmath178 , the bigger the computational gain .
note , that @xmath177 is of dimension @xmath179 .
its dual contains the line @xmath180 and we may project it ( or at least think of it as projected ) down along this line to a @xmath179 dimensional polyhedral cone .
so the problem of enumerating facets incident to a given ray is a facet enumeration problem , but in one lower dimension and with fewer extreme rays
. some of the lower dimensional subproblems may still be too difficult ; in this case we may apply the method recursively .
we come back to this in section [ sec : recursion ] .
the incidence decomposition method has been used in @xcite for finding the vertices of the metric polytope @xmath181 .
the method was also introduced in @xcite along with the adjacency decomposition method but it was found to be less competitive for their application .
the incidence decomposition method was also discussed by fukuda and prodon @xcite . as with the incidence decomposition method ,
the adjacency decomposition method is a reasonably natural method for computing the facets of a polytope up to symmetries .
so it is no wonder that the method was discovered several times , for example , in @xcite as `` algorithm de lexplorateur '' , in @xcite as `` adjacency decomposition method '' and in @xcite as `` subpolytope algorithm '' .
other example of applications are in @xcite , @xcite or @xcite .
the adjacency decomposition scheme is also implicit in polyhedral decomposition schemes , i.e. when a space is decomposed as an union of polyhedral cones .
examples are in @xcite , @xcite and @xcite ( cf .
@xcite and @xcite ) . rather than focusing on the incidence of ( orbits of ) facets to extreme rays as the incidence decomposition method does ,
the adjacency decomposition method focusses on the incidence of facets with other facets . starting from a ( set of ) initial @xmath75-inequivalent facets
, it traverses the adjacency graph of facet orbits .
the initial facet(s ) may be obtained by suitable linear programs , or in strongly polynomial time using the methods described in @xcite .
* input : * @xmath72 extreme rays of a polyhedral cone @xmath10 and a group @xmath163 acting on @xmath10 s face lattice . + * output : * complete set @xmath164 of @xmath75-inequivalent facets of @xmath10 .
+ @xmath182 with @xmath60 a facet of @xmath10 .
+ @xmath165 . +
* while * there is an @xmath183 * do * + @xmath170 .
+ @xmath184 .
+ @xmath185 .
+ * for * @xmath186 * do * + @xmath187 facet of @xmath10 adjacent to @xmath60 along @xmath188 . + * if * @xmath58 is @xmath75-inequivalent to all facets in @xmath189 * then * + @xmath190 . +
* end if * + * end for * + * end while * + the facet @xmath58 of @xmath10 with @xmath191 for a given ridge @xmath192 can be found by a _ gift - wrapping step _ ( cf .
@xcite , @xcite ) .
let @xmath193 be the generators of @xmath10 s extreme rays . the defining inequality @xmath194 of the facet @xmath58 should satisfy @xmath195 for all generators @xmath196 .
the vector space of such functions has dimension @xmath53 .
let us select a basis @xmath197 of it . if @xmath198 is the defining inequality of @xmath60 , @xmath188 or @xmath58 , then @xmath199 for all @xmath200 .
this translates into a set of linear inequalities on @xmath201 defining a @xmath53-dimensional pointed polyhedral cone .
one finds easily its two generators @xmath202 . the corresponding inequalities @xmath203 define the two adjacent facets @xmath60 and @xmath58 of @xmath10 .
in the special case where no @xmath204 extreme rays lie on a hyperplane , the gift wrapping step corresponds to a simplex pivot ( see @xcite ) .
a nice feature of the adjacency decomposition method is that the adjacencies of the most symmetric facets , which are usually the most difficult to treat , may not have to be considered .
this is due to the well known _
balinksi s theoreom _
( see e.g. @xcite , theorem 3.14 ) : [ balinski](@xcite ) let @xmath10 be a @xmath71-dimensional , pointed polyhedral cone .
let @xmath75 be the undirected graph whose vertices are the facets of @xmath10 and whose edges are the ridges of @xmath10 .
two vertices @xmath205 are connected by an edge @xmath206 if @xmath207 .
then , the graph @xmath75 is @xmath208-connected , i.e. removal of any @xmath209 vertices leaves it connected . using balinski s theorem
, we know that if the number of facets in unfinished orbits ( i.e. those whose neighbors are not known ) is less than @xmath179 , then any @xmath75-inequivalent facet in the unfinished set must be adjacent to some already completely treated facet .
this simple criterion has proven @xcite , @xcite to be extremely useful in dealing with examples arising in the geometry of numbers . both , the incidence decomposition method and the adjacency decomposition method reduce the facet enumeration problem for a @xmath71-dimensional polyhedral cone @xmath10 to a number of facet enumeration problems for cones in dimension @xmath179
. these lower dimensional problems may be too difficult to treat with a standard method as well and therefore we might apply the incidence or adjacency decomposition method to these lower dimensional problems recursively .
so we may speak of the _ recursive incidence decomposition method _ , the _ recursive adjacency decomposition method _ and the _ recursive decomposition method _
, if a mixture of both is applied .
the recursive incidence decomposition method has successfully been used for computing the vertices of the metric polytope @xmath210 in @xcite , whereas the recursive adjacency decomposition method has successfully been used in @xcite and @xcite . to the best of our knowledge , and to our surprise , a recursive mixture of both has not been used so far .
the crucial steps for the recursion are `` @xmath211 '' ( where @xmath155 is some extreme ray ) for the incidence method ( see section [ sec : incident ] ) and `` @xmath212 '' ( where @xmath60 is some facet ) for the adjacency decomposition method ( see section [ sec : adjacency ] ) . in both cases the problem is to obtain a list of facets for a @xmath14-dimensional polyhedral cone @xmath60 , whose extreme rays are given .
again we can exploit symmetries . as a result of a call of the incidence or adjacency decomposition method for @xmath60
, we obtain a list of @xmath213-inequivalent facets of @xmath60 ( ridges of @xmath10 ) , where @xmath213 is some group acting on the face lattice of @xmath60 which we have to provide . in a post processing step
we then have to obtain a list of @xmath214-inequivalent ( respectively @xmath215-inequivalent ) facets of @xmath60 ( ridges of @xmath10 ) out of it .
note though , that the groups @xmath213 and @xmath216 may be unrelated , that is , the elements of @xmath213 do not have to be elements of @xmath216 and vice versa .
think of examples where @xmath10 is a polyhedral cone without any symmetries , but with a facet having some symmetries .
vice versa , since we did not assume that @xmath213 is the full ( combinatorial ) symmetry group , not even the stabilizer @xmath217 of @xmath60 in @xmath214 has to be a subgroup of @xmath213 . in the latter case we may simply replace @xmath213 by the group generated by @xmath217 and @xmath213 , so that we assume @xmath218 .
in this way we possibly enlarge the group and speed up the computations .
moreover , we are able to use the double coset decomposition ( see section [ sec : double - cosets ] ) to split each @xmath213-orbit @xmath219 of facets of @xmath60 into a finite number of @xmath215-orbits @xmath220 , where the @xmath221 represent the double cosets .
using the described decomposition methods recursively it might happen that we compute the facets of some sub - cones several times .
for example if a face @xmath60 satisfies @xmath222 for a facet @xmath55 of a polyhedral cone @xmath10 , then there is exactly one other facet @xmath56 such that @xmath223 .
hence , if we apply the adjacency decomposition method to @xmath55 and @xmath56 , then we have to compute the dual description of @xmath60 two times . clearly , the number of such repetitions increases as the recursion depth increases .
moreover , equivalent sub - cones may occur in different parts of the face lattice .
( recall that any sub - lattice of the face lattice is the face lattice of some polyhedral cone . ) to handle this , we propose to use a _ banking system_. that is , given a difficult sub - cone @xmath10 of which the facets have been computed up to symmetries , we store with @xmath10 ( represented by generators ) its group of restricted automorphisms and a representative for each facet orbit . what about the recursion depth needed to practically solve a given polyhedral conversion problem ? if we split the problem into subproblems with either decomposition method then some of the subproblems may be easy to solve and may not require a recursive treatment , others might be impossible to treat without such . the implementation @xcite allows one to provide a heuristic function to choose whether to recursively apply the adjacency decomposition method , but in choosing the right level of recursion requires some trial and error . from our computational experience with this code ( for example in @xcite , @xcite , @xcite , @xcite , @xcite ) , the _ incidence number _ of a face ,
that is , the number of extreme rays contained in it , gives a good measure for how difficult the representation conversion is for it .
we therefore propose to treat the faces with low incidence numbers first and as much as possible without recursion . for subproblems to be potentially solved with pivoting methods ( possibly up to symmetry as below ) , the `` probing feature '' of lrs
can be used to test for expected difficulties of a subproblem . finally , let us remark , that parallelization is a possible way to speed up the ( recursive ) decomposition methods as well ( see @xcite and @xcite ) .
we briefly sketch here the _ cascade algorithm _ by jaquet @xcite , which is a symmetry exploiting version of _ fourier motzkin elimination_. for this observe first that a @xmath71-dimensional polyhedral cone @xmath10 generated by @xmath82 can be obtained as the projection of an @xmath72-dimensional polyhedral cone with @xmath72 linear independent generators @xmath224 . without loss of generality , we may assume that @xmath225 is a basis of @xmath0 .
we set @xmath226 for @xmath227 and @xmath228 for @xmath229 .
let @xmath230 be the orthogonal projection of @xmath231 onto @xmath232 .
then for the cone @xmath233 we have @xmath234 .
the cone @xmath87 is simplicial , since the family @xmath235 is a basis of @xmath231 .
take the @xmath72-dimensional polyhedral cone @xmath87 and the projection @xmath236 of @xmath231 on a @xmath237-dimensional hyperplane .
the facets of the projection @xmath238 are either projections of facets of @xmath87 or projections of intersections of facets of @xmath87 . in the fourier
motzkin elimination we first compute the facets of @xmath87 and then successively obtain the facets of @xmath239 from one of these two cases .
note , since @xmath87 is a simplex , the facets of @xmath87 are simply given by the @xmath237-subsets of extreme rays of @xmath87 . in the cascade algorithm
we also consider symmetries in each step : if @xmath75 is a group of symmetries of the polyhedral cone @xmath10 , then the induced symmetry group on @xmath240 is the stabilizer of the set of vertices @xmath241 under @xmath75 , which we denote by @xmath242 .
so , in order to compute the orbits of facets of @xmath240 under @xmath242 we need to compute the orbits of facets and ridges of @xmath243 , first under @xmath244 and then under @xmath245 using the double coset method .
it is well known that the fourier
motzkin method depends in a critical way on the ordering of generators @xmath193 of @xmath10 ( see @xcite and @xcite ) .
the cascade algorithm introduces another such dependency as the size of the set - stabilizers vary enormously according to the chosen ordering . among the many possible orders of the generators , it is not clear which ordering is best . in practice fourier
motzkin elimination suffers from generating many redundant inequalities .
this problem can be eliminated by using the related _
double description method _ @xcite . for a description of how to use the double description method to perform the fourier
motzkin projection steps without introducing redundant inequalities , see @xcite , section 4.1 .
since for certain classes of input , the most successful methods for the polyhedral representation transformation problem ( without taking symmetries into account ) are based on the pivot operation of the simplex method , it is natural to consider whether pivoting techniques can be adapted to the symmetric setting . for the purposes of this discussion , by _ @xmath37-basis _
, we mean a set of @xmath37 extreme rays generating the linear span of a @xmath37-face of a pointed polyhedral cone .
two @xmath37-bases are _ adjacent _ if they share @xmath62 extreme rays .
when @xmath37 is not specified , we refer to @xmath14-bases of a @xmath71-dimensional polyhedral cone .
note that the bases of a polyhedral cone depend not only on the combinatorial information contained in the face lattice , but also on the linear dependencies among extreme rays incident to a given face .
when discussing @xmath14-dimensional polytopes or their @xmath71-dimensional homogenizations , we use basis to refer interchangeably to a @xmath14-basis of the homogeneous cone or a @xmath54-basis of the polytope ( i.e. , @xmath179 vertices spanning a facet of the polytope ) .
the _ basis graph _ has as nodes the bases , and as edges the pairs of adjacent bases . in the
so - called _ non - degenerate _ case each facet of a @xmath71-dimensional cone is a @xmath14-dimensional simplicial cone . by a perturbation argument , for any cone there exists a triangulation of the boundary which is combinatorially equivalent to the boundary of a polyhedral cone with simplicial facets ; it follows by balinski s theorem that the basis graph is at least @xmath14-connected .
one of the earliest published methods of polyhedral representation transformation @xcite is based on an exhaustive exploration of the basis graph . using the pivot operation of the simplex method
, adjacent @xmath14-bases can be found in time proportional to the number of input generators and the dimension ( for details , see e.g. @xcite ) .
pivoting methods based on reverse search @xcite have memory usage independent of the output size . in the present work we consider only more direct methods based on generation ( and storage ) of the basis graph up to symmetry .
in the typical case , generating the entire basis graph is impractical , even if it is not stored .
this is because of the enormous number of bases that correspond to each facet in the degenerate ( i.e.non-simplicial facet ) case . in the non - symmetric
setting , perturbation has been widely used to reduce the size of the basis graph under consideration , in particular via the lexicographic ( symbolic ) perturbation discussed further in section [ sec : symbolic ] .
we consider one way in which perturbation can be applied in the symmetric case in section [ sec : orb - perturb ] .
there is some tension between the notions of symmetry and perturbation though : whereas perturbation allows us to reduce the size of the basis graph , usually some of the given symmetries are lost .
so there is a tradeoff in which ideally the quotient graph of the obtained new basis graph with respect to its remaining symmetries is as small as possible .
the performance of pivoting based methods under symmetry is determined not by the total size of the basis graph , but by the number of orbits of bases .
this number is not determined by the number of orbits of extreme rays and the number of orbits of facets .
we define the _ basis automorphism group _ to be the subgroup of the combinatorial automorphism group that acts on the basis graph . to see that this is a non - trivial restriction , consider the cone generated by the following ( row ) vectors in @xmath246 @xmath247 combinatorially ,
it is ( the homogenization of ) a pyramid over an octahedron , and its combinatorial automorphism group is the @xmath248 element octahedral group .
the basis automorphism group has only @xmath249 elements ; in particular the orbit of the first @xmath71 generators ( as a basis ) is of size @xmath52 .
[ ex : crosspolytope ] the regular @xmath71-cross polytope @xmath250 has one orbit of @xmath71-bases and one orbit of @xmath14-bases with respect to its automorphism group
. since the @xmath71-cross polytope is simplicial , ( i.e. , its facets are all simplices ) its @xmath14-bases are exactly its facets .
observe that at most one pair of vertices in a @xmath71-basis must consist of an opposite pair @xmath251 , since any two pairs are co - planar .
it thus follows from the pigeonhole principle that a @xmath71-basis consists of a facet , along with one vertex not on that facet .
all such simplices are equivalent under the automorphism group of the cross polytope . for a slightly more involved example , let @xmath252 and @xmath253 .
the _ wreath product _ of @xmath10 with @xmath254 is defined as @xmath255 so loosely speaking , the wreath product of @xmath10 and @xmath254 is obtained by attaching to the vertices of @xmath254 pairwise orthogonal subspaces , each containing a copy of @xmath10 .
the wreath product @xmath256 of a regular @xmath71-cross polytope @xmath10 with a regular @xmath257-cross polytope @xmath254 has dimension @xmath258 , @xmath259 vertices , @xmath260 facets and one orbit of vertices , facets , and @xmath261-bases with respect to its linear automorphism group .
[ ex : wreath ] the proof is based on the results of section 2.2 in @xcite
. the number of vertices of the cross polytopes @xmath10 and @xmath254 are @xmath262 and @xmath263 .
the assertion on the dimension and the number of vertices of @xmath256 follows immediately from the definition of wreath products . the number of facets and the number of bases orbits can be derived from proposition 2.2 in @xcite . by it ,
the facets of @xmath256 are in one - to - one correspondence with choices @xmath264 of a facet @xmath60 of @xmath254 and facets @xmath144 of @xmath10 ; which gives the count of facets .
assume w.l.o.g .
that @xmath60 contains the vertices @xmath265 of @xmath254 .
then the rows of the following matrix describe the vertices contained in a facet of @xmath256 : @xmath266 the entries @xmath10 and @xmath144 stand for all choices of vertices from @xmath10 , respectively @xmath144 .
so a facet of @xmath256 contains @xmath267 vertices .
the linear automorphism group of @xmath256 contains the _ semidirect product _
@xmath268 ( where @xmath269 and @xmath270 denote the automorphism groups of @xmath10 and @xmath254 and @xmath271 denotes an @xmath72-fold direct product ) .
in particular , @xmath270 is isomorphic to a subgroup of @xmath75 , which permutes not only @xmath272 in the last @xmath257 coordinates , but also the @xmath72 copies of @xmath10 attached to them . in contrast to this action interchanging the @xmath72 @xmath71-dimensional subspaces , the elements of the subgroup @xmath271 of @xmath75 act only within these subspaces . by this ,
all of the vertices and facets lie in one orbit under the action of @xmath75 .
the assertion on the bases follows as well , after noting that a base within a facet ( as described above ) is determined by @xmath257 choices of @xmath71-bases ( in the @xmath257 copies of @xmath10 ) .
all of them are equivalent with respect to @xmath75 by example [ ex : crosspolytope ] .
we have implemented a prototype called _ symbal _ ( `` symmetry , bases , and lexicography '' ) for facet enumeration up to symmetries via pivoting .
the code is mainly in gap @xcite , with an external server based on ` lrs ` doing the pivoting .
the main algorithm is a depth first search of the pivot graph ; several options are provided for symmetry testing and perturbation ( discussed further below ) .
all symmetry groups are represented internally as permutation groups , and the backtracking algorithm of gap s representativeaction is used to test bases and facets for equivalence .
our main goal with this implementation is not to be able to attack large problems , but to provide a test - bed for better understanding of the main issues involved with pivoting under symmetry . in general
we thus are not too concerned with a constant multiplicative slowdown in the runtime of our code versus more polished software ; we are more interested in what kinds of problems can be solved in a reasonable amount of time . at least for certain special cases ,
our prototype is able to solve quite large problems .
figure [ fig : wreath - compare ] compares the run time of ` symbal ` with ` cdd ` and ` lrs ` on the wreath products of cross polytopes discussed in example [ ex : wreath ] .
of course both of ` cdd ` and ` lrs ` compute all of the facets rather than orbits .
the times reported here are on a 3ghz pentium iv with 1 g of memory .
we call a cone ( or polytope ) _ basis - simplicial _ if it has as few basis orbits as facet orbits .
evidently every simplicial polytope is basis - simplicial . for perhaps less contrived examples of basis - simplicial polytopes than example [ ex : wreath ]
we mention the dirchlet - voronoi - cells ( dv - cells ) of the root lattices @xmath273 , @xmath274 , and @xmath275 ( convex hull of their shortest non - zero vectors ) and some related _ universally optimal _ spherical polytopes ( see @xcite ) .
their boundaries consist of regular cross polytopes and regular simplices only and they thus have at most two orbits of basis with respect to their symmetry group ( cf .
example [ ex : crosspolytope ] ) . as with the non - symmetric case , for cones with a large number of ( orbits of ) bases , pivoting is not a good approach , at least not without some kind of perturbation .
one benchmark of how _ orbitwise degenerate _ a polytope is , is whether the number of orbits of bases is larger than the number obtained by triangulating the boundary . in this case , one is almost certainly better off applying either ` cdd ` or ` lrs ` . a familiar example where this level of degeneracy occurs
is in the @xmath71-dimensional cubes ( see table [ tab : cube - orbits ] ) .
.basis orbits of cubes [ cols="^,^,^,^,^",options="header " , ] figure [ fig : cube - tri ] illustrates our two heuristics applied to the three dimensional cube . on the left we have the triangulation induced by a subgroup of order @xmath276 having two generator orbits , in the center
, the triangulation is induced by the stabilizer of a pair of opposite vertices . on the right , we take the same stabilizer , but push the first orbit instead of pulling . the triangulation induced by pulling opposite vertices of a @xmath277-cube turns out to be a special case of a triangulation of the @xmath71-cube that has only one basis orbit in any dimension .
let @xmath278^d$ ] denote the centrally symmetric @xmath71-cube .
define @xmath279 for each permutation @xmath280 , let @xmath281 denote the simplex with vertices : @xmath282 let @xmath64 denote the union of all @xmath281 and let @xmath283 denote the set of @xmath14-simplices formed by intersecting the simplices of @xmath64 with the boundary of @xmath284 .
it is known @xcite that @xmath64 forms a triangulation of @xmath284 ; consequently @xmath283 forms a triangulation of the boundary , since every triangulation of a polytope induces a triangulation of its boundary .
we call @xmath64 ( respectively @xmath283 ) the _ linear ordering triangulation _ of @xmath284 ( respectively of the boundary of @xmath284 . ) since there is a bijection between permutations in @xmath285 and simplices in @xmath64 , @xmath285 acts transitively on @xmath64 by permuting coordinates
. let @xmath94 denote the vertices of @xmath284 , i.e. @xmath286 .
let @xmath188 denote the stabilizer of the automorphism group of @xmath284 on @xmath287 .
define @xmath288 .
let @xmath289 denote the @xmath188-orbitwise pulling of @xmath94 in the order induced by @xmath290 .
then the following holds : 1 .
@xmath188 acts transitively on @xmath283 .
@xmath291 is combinatorially equivalent to @xmath283 .
let us first remark that the order of orbits induced by @xmath290 is well defined .
the group @xmath292 is generated by a set of generators permuting coordinates , along with a _ switching permutation _
@xmath91 that maps @xmath293 to @xmath294 .
it follows that the @xmath188-orbits of @xmath94 are the equivalence classes of @xmath94 with respect to @xmath290 . to see ( a ) , consider the @xmath71-simplex ( corresponding to the identity permutation ) @xmath295 with vertices @xmath296
let @xmath297 denote the permutation in @xmath188 that first applies @xmath91 ( i.e. switching ) followed by reversing the order of coordinates .
the permutation @xmath297 is an automorphism of @xmath295 which carries the @xmath14-simplex @xmath298 to @xmath299 .
these two @xmath14 simplices are precisely the contribution of @xmath295 to the linear ordering triangulation @xmath283 .
thus any simplex of @xmath283 can be mapped to any other by an action of @xmath285 on @xmath64 ( i.e. permuting coordinates ) , followed by possibly applying @xmath297 .
we now consider ( b ) .
we argue that @xmath289 induces a linear ordering triangulation of each @xmath37-face of @xmath284 , @xmath300 .
each @xmath54-face will receive the same ( linear ordering ) triangulation from the two facets that contain it , hence the triangulations of the facets form a triangulation of the boundary . for @xmath301
, there is nothing to prove .
let @xmath60 be a @xmath37-face of @xmath284 , @xmath302 .
let @xmath303 ( resp .
@xmath304 ) be the vertex of @xmath60 with the most positive ( resp .
negative ) coordinates .
recall that we will first pull the @xmath188-orbit with smallest @xmath290 value .
if the functional @xmath290 is minimized uniquely at @xmath305 then the perturbation corresponds locally to a standard @xcite pulling and the corresponding subdivision is into pyramids @xmath306 with apex @xmath307 and bases corresponding to all of the @xmath62 faces of @xmath60 that do not contain @xmath307 .
otherwise @xmath290 is minimized at both @xmath304 and @xmath303 .
the perturbation thus takes @xmath303 to @xmath308 and @xmath304 to some @xmath309 , @xmath310 .
this turns out to induce a subdivision of @xmath60 into polytopes @xmath311 with vertices @xmath312 where @xmath313 is the vertex set of a @xmath314-face of @xmath60 containing neither @xmath304 nor @xmath303 .
the polytope @xmath311 is a _
@xmath53-fold pyramid _ , since @xmath315 and @xmath316 .
it follows that @xmath317 .
that the @xmath311 are induced by the perturbation can be seen by exhibiting a supporting hyperplane of @xmath318 . without loss of generality ,
let @xmath60 be defined by equations @xmath319 , @xmath320 .
the @xmath314-face @xmath321 must be defined by further equations @xmath322 , @xmath323 , @xmath324 .
let @xmath325 .
consider the hyperplane @xmath326 , where @xmath327 it can be verified that @xmath328 supports @xmath318 and @xmath329 .
it remains to see that the @xmath311 cover @xmath60 , i.e. that there are no other cells in the induced subdivision . consider an arbitrary relative interior point @xmath330 of @xmath60 .
let @xmath155 be the ray from @xmath331 through @xmath330 .
let @xmath332 be the first intersection of @xmath155 with @xmath333 after @xmath330 .
for each @xmath50-face of @xmath60 , the double pulling of @xmath334 acts like a single pulling decomposing the boundary of @xmath60 into pyramids with apex either @xmath331 or @xmath308 ; hence @xmath335 for some @xmath336 .
it follows that @xmath330 is in @xmath311 .
now suppose for all @xmath337 , @xmath289 induces a linear ordering triangulation of the @xmath336-faces of @xmath284 .
from the refinement property proposition [ prop : valid1 ] , we know that @xmath289 induces a decomposition of the pyramids @xmath338 ( respectively of the @xmath53-fold pyramids @xmath339 ) . in both cases the resulting @xmath37-simplices correspond to the coordinatewise - monotone paths from @xmath304 to @xmath303 in @xmath60 .
suppose we are given @xmath340 which are homogeneous coordinates for the generators @xmath94 of some polyhedron @xmath341 .
we consider @xmath94 here as a matrix ( with @xmath18 as rows ) , and let @xmath342 denote the column vector of corresponding @xmath343st coordinates in @xmath95 . following the conventions of section [ sec : basic_notations ] , we suppose @xmath344 is @xmath52 if @xmath345 is a vertex , and @xmath44 if it is an extreme ray .
we consider the polyhedron @xmath346 @xmath347 the polyhedron @xmath346 may be thought of as @xmath348 intersected with the hyperplane @xmath26 . by duality , to find the generators of @xmath346 is equivalent to finding the facets of @xmath10 .
let @xmath75 be the restricted automorphism group of @xmath94 .
let @xmath349 be the induced group acting on @xmath95 .
we will assume without loss of generality that the origin is the centroid of v and thus a fixed point of @xmath75 .
it follows that @xmath350 is a fixed point of @xmath349 .
applying proposition [ prop : fixedpoint ] to our dual representation , we see that orbitwise perturbing the right - hand side vector according @xmath351 will preserve the symmetries of @xmath188 .
the perturbed system thus has the form @xmath352 where @xmath353 for @xmath354 and @xmath336 is the index of the orbit containing @xmath18 . to ensure an orbitwise lexicographic perturbation
, we will insist @xmath355 where by @xmath356 we mean that @xmath332 is much smaller than @xmath330 , i.e. it is not possible to combinatorially change the polyhedron defined by by choosing @xmath357 smaller . to implement this symbolically , we need a modification of the standard lexicographic pivot rule ( see @xcite or @xcite for more details ) .
let @xmath358 and @xmath359 $ ] .
after adding _ slack variables _ to , we are left with a system of the form @xmath360 with @xmath72 rows and @xmath361 columns , where the first @xmath71 columns are the _ decision variables_. a _ feasible basis _ consists of a partition @xmath362 of the column indices such that @xmath363 ( columns of @xmath364 indexed by @xmath365 ) is non - singular and any slack variables in @xmath366 are non - negative . in order to move from one feasible basis to another , we need to perform a _
pivot_. we start by choosing a column index @xmath336 to leave @xmath365 . to find a column index to replace @xmath336
, we need to find @xmath367 where @xmath368 and @xmath369 . in our case ,
where @xmath370 , we may decompose @xmath371 , and thus the ratio test into two parts @xmath372 and @xmath373 . because the values of the @xmath374 are chosen very small ,
the second part is considered only to break ties .
we write @xmath375 where column @xmath336 of @xmath376 is defined by summing the columns of @xmath377 corresponding to orbit @xmath336 of generators , and multiplying by @xmath378 . because of the ordering , in order to evaluate @xmath379 we proceed column by column in @xmath376 , reducing the set of ties at each iteration . consider facets @xmath380 and @xmath55 that are equivalent under some symmetry of the basis automorphism group . this same symmetry acts as an isomorphism between the corresponding basis graphs .
it follows that when we discover a basis @xmath381 defining a new orbit , but the facet @xmath60 spanned by @xmath381 is known , we do not need to explore the neighbours of @xmath381 since they will be explored in our canonical ( i.e. discovered first ) facet in the orbit of @xmath60 . in order to ensure orbits
are not discarded , we are careful not to mark @xmath381 as known until its canonical discovery .
although this pruning does not reduce the number of orbits of bases explored , it can reduce the number of actual bases visited ( and tested for isomorphism ) , since bases of a given orbit are not revisited in every copy of the facet @xmath60 . as an example , consider the @xmath277 quadrilateral facets illustrated in figure [ fig : rotate ] , with rotational symmetry yielding @xmath382 orbits of bases .
facets with a @xmath277-fold rotational symmetry .
basis @xmath383 has vertices @xmath384 around the boundary of the facet.,height=144 ] without pruning , a depth first search visits all of the bases ; with pruning only @xmath385 of @xmath386 bases are visited ( see figure [ fig : prune ] ) .
the speedup obtained depends roughly on the number of facets visited by the unpruned search , which is bounded by the number of basis orbits .
c@c & in the case where our symmetry group preserves the inner product between pairs of vectors , as is the case for the restricted automorphisms discussed in this paper , we may take advantage of this in several ways . for any face or basis @xmath387 to be tested for isomorphism , we may construct a graph ( analogous to that constructed in proposition [ prop : isomorphism - equivalence ] ) whose nodes are the vectors of @xmath387 and whose edges are the angles between them .
this graph contains geometric information not present in the index sets representing @xmath387 , which can help to speed up an algorithm to find an isomorphism . a simpler observation , and equally widely applicable , is that the set of pairwise inner products of two isomorphic faces or bases must be equal .
this allows us to store orbit representatives in a data structure such as a hash table or a balanced tree , with the key to the data structure being the set of inner products .
this permits more efficient isometry testing by retrieving exactly those orbit representatives which pass the inner product invariant .
it is computationally easy to test whether a given linear transformation @xmath388 is in the restricted automorphism group @xmath389 of vector family @xmath94 .
since we are interested in restricted automorphisms carrying basis @xmath387 to basis y , we can additionally test if @xmath390 for some @xmath391 ( where @xmath392 can be computed by the same techniques as proposition [ prop : isomorphism - equivalence ] ) .
we have only implemented an exhaustive search of @xmath392 , and this is naturally only effective when @xmath392 is quite small . in principle
it should be possible to integrate the test for @xmath388 being a restricted isomorphism into a backtracking procedure to search for @xmath393 .
much as in the case of polyhedral representation conversion without symmetries , a certain amount of trial and error seems to be necessary to decide on the the best method to attack a given conversion problem up to symmetries .
currently decomposition methods have the best record of solving interesting problems ; on the other hand current software requires a certain amount of user intervention in the form of choosing how to treat subproblems .
it would be helpful to automate this process . in this context
, a virtue of the pivoting methods is that good methods to estimate their running time exist @xcite .
it would be beneficial , not just when working with symmetry , to have effective methods ( or at least heuristics ) for estimating the running time of incremental methods .
the authors would like to acknowledge fruitful discussions with david avis , antoine deza , komei fukuda , alexander hulpke , michael joswig , jesus de loera , brendan mckay , hugh thomas , and frank vallentin .
they would also like to thank the centre de recherche mathmatiques for making possible the workshop at which many of these discussions took place .
anzin , _ on the density of a lattice covering for @xmath394 and @xmath395 _ , tr . mat .
steklova * 239 * ( 2002 ) , diskret .
chisel , 2051 ; translation in proc .
* 239 - 4 * ( 2002 ) , 1344 .
b. ballinger , g. blekherman , h. cohn , n. giansiracusa , e. kelly and a. schrmann , _ experimental study of energy - minimizing point configurations on spheres _ , preprint at http://arxiv.org/abs/math.mg/0611451 .
m. dutour sikiri , a. schrmann and f. vallentin , _ classification of eight dimensional perfect forms _ , electron .
, to appear , preprint at http://arxiv.org/abs/math.nt/0609388 .
g. voronoi , _ nouvelles applications des paramtres continues la thorie des formes quadratiques 1 : sur quelques proprits des formes quadratiques positives parfaites _ , j. reine angew .
math * 133 * ( 1908 ) , 97178 .
voronoi , _ nouvelles applications des parametres continus l thorie des formes quadratiques , deuxime mmoire , recherches sur les paralllloedres primitifs _ , j. reine angew .
( 1909 ) * 134 * , 198287 and * 136 * , 67181 . | we give a short survey on computational techniques which can be used to solve the representation conversion problem for polyhedra up to symmetries .
we in particular discuss decomposition methods , which reduce the problem to a number of lower dimensional subproblems .
these methods have been successfully used by different authors in special contexts . moreover
, we sketch an incremental method , which is a generalization of fourier motzkin elimination , and we give some ideas how symmetry can be exploited using pivots . |
after electroweak symmetry breaking , we diagonalize the flavor eigenstates to the mass eigenstates @xmath11 with obvious notations .
is the ( 11)-component of @xmath12 and @xmath13 is ( 11)-component of the matrix @xmath14 , etc . ] in sm , we _ define _ the electron as the lightest charged lepton , thus its flavor and mass eigenstates are identical . at the same time we independently define the up quark as the lightest quark , because it is not connected to leptons by any interactions .
one may have equal right to define the muon as the lightest lepton , in sm this is just renaming . however in gut , the quarks and leptons are related , for instance @xmath15 is related to @xmath16 in @xmath17 of @xmath7 . once we define the electron by fixing @xmath15 and @xmath18 in the entire multiplets @xmath19 , the up - type quarks follow from the relative mass eigenstates @xmath20 by @xmath21 to form the flavor multiplet . therefore defining the muon as the lightest lepton modifies the identity of the up quark as the lightest , since the new definition is now distinguished .
thus in general , even the flavor eigenstates of up - type quarks are not guaranteed to be mass eigenstates .
also it means that we can distinguish contorted flavors , as in ( [ twistgen ] ) and ( [ twistlr ] ) . on the other hand , this information is not relevant in the low energy physics , since there is no interaction connecting like ( [ eu ] ) in sm .
even if a low level theory predicts the lighter mass for the muon than the electron belonging to @xmath22 , misidentification of the lightest as electron _ does not _ change sm .
after fixing the electron , the relative fermions in @xmath7 are completely connected by the relative basis change matrices ( rbcms ) , among them are ckm matrix @xmath23 @xcite and pmns matrix @xmath24 @xcite , up to renormalization effects .
for example , the emission of @xmath25 boson from down quark to _
quark accompanies the rbcm @xmath26 , which is to be distinguished from the ckm element @xmath27 . in the minimal @xmath7
, we have two yukawa couplings @xmath28 and @xmath29 diagonalized by four matrices @xmath30 .
we can fix @xmath31 and @xmath32 , where @xmath33 is a diagonal @xmath34 matrix .
thus , besides known quark and lepton masses and ckm matrix , we have two new phases @xcite .
since @xmath35 gut completely relates all the fields in sm , we can in principle completely observe the yukawa coupling itself , from interactions in gut .
this means , it is not possible to arbitrarily align the flavor and the mass eigenstates .
in the minimal @xmath7 , the rbcm between down - type quarks and conjugate leptons is @xmath36 , without loss of generality .
it means that , for example by charged @xmath37 boson exchange , @xmath38 always transits to @xmath18 , @xmath39 , but there is no transition between @xmath38 and @xmath40 , @xmath41 .
it is because there is no difference in mass diagonalization between down - type quarks and leptons @xmath42 .
however the minimal models lead to bad mass relations @xmath43 , which is renormalization group invariant .
one should introduce more higgs fields and as many complex yukawa matrices .
usual fermion textures in non - minimal gut imply similar rbcms between quarks and leptons , with slightly modified diagonality , but in general there emerges many new phases . to correct the relations , the main paradigm after ref .
@xcite has been to introduce 45 and and higher dimensional higgses . for illustration
, we adopt the most general renormalizable yukawa coupling including @xmath44 : @xmath45 where the coefficients are in general complex matrices in the flavor basis . with some assumptions on the form of yukawa couplings and vev
, we can have the following form of mass matrices for isospin @xmath46 fermions @xmath47 which is also a typical texture from ` minimal renormalizable ' @xmath35 , making use of @xmath48 and @xmath49 higgses @xcite . the elements in the same letters come from the same vevs of @xmath50 and @xmath44 , with some clebsch - gordan coefficients .
if we take @xmath51 , in the parameter range @xmath52 after diagonalization @xcite .
however in the parameter range @xmath53 as before , we have _ inverted lepton mass hierarchy _ with respect to quarks .
@xmath54 however we note that , although lepton mass hierarchy is inverted , still _ we identify the electron as the lightest _ charged lepton in sm
. therefore it will be more appropriate to collect the multiplet as @xmath55 and @xmath56 .
if the parameters are aligned to fit the observed fermion masses @xmath57 mev@xmath58 , @xmath59 mev@xmath58 ( neglecting renormalization corrections ) , we can not see the change in sm .
of course , this has distinguished consequences in gut , in particular proton decay that we see shortly . because we have inverted hierarchy of the charged leptons
, we have _ mostly off - diagonal @xmath60 and @xmath61 _ with respect to the one with normal hierarchy , @xmath62 now there is no transition between @xmath38 and @xmath18 , @xmath63 by @xmath37 boson exchange , but there is one between @xmath38 and @xmath40 , @xmath64 . in more general contortions , @xmath65 can be any element of permutation symmetry of order 3 , @xmath66 , which is a discrete subgroup of @xmath67 . in string compactifications , gauge bosons and matter fermions
have similar embedding as gut @xcite .
however we do not break them by higgs mechanism , but by projection conditions associated with symmetries of internal manifold , just leaving footprints of unified multiplets .
the fermion mass hierarchy is determined by geometry @xcite , from distributions of fields in the internal spaces .
thus we can have large suppression for down - type quarks while having small suppression for leptons in the same multiplet and vice versa .
thus the contortion naturally arises in string theory . because of symmetric yukawa coupling , we can not contort @xmath68 sector with respect to @xmath69 in @xmath7 .
also @xmath68 and @xmath38 sectors are related by ckm matrix , so we have a similar hierarchy between up and down type quarks .
we can further generalize the idea , as in ( [ twistlr ] ) , by exchanging _ only the right - handed _ fermions between different generations , @xmath70 or @xmath71 then we have more freedom for quark and lepton masses . in particular in @xmath7 ,
highly asymmetric texture is possible , thus the ` minimal ' @xmath7 can be different .
for example if we exchange @xmath69 and @xmath72 to form a contorted family , the original symmetry constraint from @xmath73 on @xmath74 becomes @xmath75
for the association of quark and lepton generations , the only direct experiment have been the proton decay @xcite .
first we consider the effect of dimension six operators .
the typical proton decay amplitude is @xmath76 mediated by @xmath77 bosons @xcite , @xmath78 here @xmath79 is the low energy contribution from hadronic matrix element , can be calculated by lattice qcd @xcite .
@xmath80 are respectively the proton mass , the pion decay constant , the unified gauge coupling , the renormalization factor . @xmath81 and @xmath82 are interaction strength between baryons and mesons obtained from chiral perturbation theory . neglecting the last factor ,
the partial lifetime for susy case is @xmath83 the gauge boson mass is guided by running gauge coupling : @xmath84 gev for non - susy and @xmath85 gev for susy cases @xcite . considering a typical decay channel like @xmath76
, we have the partial mean lifetime , in non - susy case @xmath86 years , depending on extensions .
the observed bound for the proton lifetime is @xmath87 years @xcite .
the last factor contains rbcms , which are respectively the components of @xmath88 , all of which lie outside ckm and pmns .
the amplitude is highly flavor dependent @xcite . in the minimal @xmath7 ,
all the elements are set to identity as above , except @xmath89 so the flavor factor is nearly 5 .
the flipped @xmath7 @xcite contains only the first term in the last factor @xcite .
if we contort the first two generations as in ( [ rediag ] ) , then we have @xmath90 with @xmath91 .
thus the transition from quark to lepton is dominated by off - diagonal elements
. eventually the proton decay is dominated by muon .
consequently from ( [ pdecaywid ] ) there can be no @xmath76 , but _ opens proton decay channels to other leptons , _ like @xmath92 with the same lifetime as the original case .
the bound for this process is @xmath93 years , is almost same order of magnitude to proton decay into positron .
thus large off - diagonal rbcm element does not help much .
there have been only water ( e.g. super - kamiokande ) and earthly metal ( e.g. iron in soudan ii ) based experiments , thus are able to identify decay from protons and neutrons , consisting of up and down quarks as initial states .
so if the electron is connected with other quarks than @xmath94 , we can not observe the decay from other sources .
the pairing @xmath95 is achieved if we invert the first and the third generation leptons .
the outgoing state would be @xmath96 , instead of @xmath97 . in this case
the proton can not decay into tauon , since it is heavier ( @xmath98 mev@xmath58 ) than proton ( @xmath99 mev@xmath58 ) .
thus if the tauon component of the mixing is high , we will lose decay information for @xmath100 .
however still there are decay channels @xmath101 where we can not distinguish the flavor of neutrino , since we only seek missing energy .
notably this decay does not exist in the flipped @xmath7 . in some model , with contortion like ( [ twistlr ] ) , we can suppress some nucleon decay by dimension six operators , however it is very hard to overcome the constraint from ( [ ptonupi ] ) and ( [ ntonupi ] ) .
it is because , exchanging right - handed quarks and leptons , there is always a pair @xmath102 in almost all known unification models .
thus a nucleon decay into neutrino is very important verification of all kind of gut interactions @xcite .
we may make theory with just simple exchange @xmath103 to have a generation @xmath104 , without mass inversion
. then we can easily see that , since the rbcm does not change @xmath105 , so that proton always decays into the lightest lepton , whatever name it has .
thus we have no change .
the most dominant dimension five proton decay operators of @xmath106 type in susy @xmath7 gut.,height=132 ] although the dimension six operators are not so harmful in susy gut , there are relevant dimension five operators from integrating out the triplet higgs pair , known as @xmath107 and @xmath108 : @xmath109 where @xmath110 at the unification scale . here
@xmath111 is the triplet higgs mass , obtained by unification condition .
because of @xmath112 contractions , @xmath113 vanishes if the family indices satisfy @xmath114 and @xmath115 vanishes if @xmath116 .
considering all the contribution , the dominant decay channels are @xmath117 for @xmath107 , a ` box ' diagram dressed with wino loop , and @xmath118 for @xmath108 with higgsino loop , depicted in fig . [
fig : r4 ] .
the latter has amplitude @xmath119_{l^4 } + [ a_\tau(\tilde t_r ) + a_\tau(\tilde c_r)]_{r^4},\ ] ] where @xmath120 where @xmath121 and @xmath122 are weak coupling , wino mass and mu parameter , respectively .
the experimental bound is @xmath123 years , which is blind to neutrino species , so it includes all the species . in the minimal @xmath7 ,
@xmath124 and @xmath125 contain the same phase matrix @xmath33 which belongs to a diagonal @xmath34 .
it is well known that the @xmath107 operators can be suppressed by destructive interference between two amplitudes mediated by scharm @xmath126 and stop @xmath127 , adjusting phases @xmath128 @xcite .
also it is calculated that the @xmath108 operators for the tau neutrino decay contributes much larger , enhanced by tau and top yukawa couplings and the mu parameter @xcite .
there the term ( [ 5rscharm ] ) with right - scharm loop was neglected , due to yukawa suppression by @xmath129 in the amplitude .
in fact the suppression power is closer to 1 , since we have additional ckm factors to have @xmath130 . in the minimal @xmath7 ,
both @xmath131 and @xmath132 have the same phase @xmath133 .
we note that in the minimal @xmath7 it is implicitly used that @xmath134 . in a contorted model ,
if it became small , the stop exchange amplitude could be comparable to that with scharm loop .
this can be done by exchanging tauon with muon or electron , as in ( [ twistgen ] ) , because we then have @xmath135 where @xmath65 is now mostly off - diagonal in ( 23 ) , or ( 13 ) components , as in ( [ rediag ] ) .
considering the former case @xmath136 , from @xmath137 we have @xmath138 so that @xmath139 then an additional phase in the rbcm @xmath140 would help to cancel @xmath108 part . besides this effect
, a contortion on the lepton sector does not affect other vertices of the proton decay operator , except the common part @xmath141 .
in fact this part quite insensitive to our contortion because of large mixing of pmns matrix . in the latter case @xmath142 , we have @xmath143 and @xmath144 thus the stop and scharm loop contributions are smaller and comparable . in any case
there must be enhancement in operators including the electron or the muon , since now either @xmath145 or @xmath146 .
thus we can not suppress all the @xmath108 operator contribution simultaneously . in the case of twisting between electron and tauon , with the dominant channel @xmath147 ,
we can enhance the proton lifetime by the factor @xmath148 with respect to ( [ 5rstop ] ) .
this restores minimum peak of the @xmath149 back to @xmath150 , while pushing @xmath151 to higher @xmath152 .
there are no sizable enhancements in the other box diagrams due to small yukawa couplings @xmath153 or @xmath154 .
the author is grateful to mitsuru kakizaki , bumseok kyae for discussions .
this work is supported by the dfg cluster of excellence origin and structure of the universe , the european union 6th framework program mrtn - ct-2004 - 503069
quest for unification " , mrtn - ct-2004 - 005104 forcesuniverse " , mrtn - ct-2006 - 035863 universenet " and sfb - transregios 27 neutrinos and beyond " and 33 the dark universe " by deutsche forschungsgemeinschaft ( dfg ) . 99 h. georgi and s. l. glashow , phys .
* 32 * , 438 ( 1974 ) .
j. c. pati and a. salam , phys .
lett . * 31 * , 661 ( 1973 ) .
h. georgi , in _ proceedings of the aps div . of particles and fields ,
_ ed . c. carlson , p. 575
( 1975 ) .
h. georgi , h. r. quinn and s. weinberg , phys .
* 33 * , 451 ( 1974 ) ; s. dimopoulos , s. raby and f. wilczek , phys .
d * 24 * , 1681 ( 1981 ) .
see , e.g. h. p. nilles , s. ramos - sanchez , m. ratz and p. k. s. vaudrevange , arxiv:0806.3905 [ hep - th ] ; j. e. kim and b. kyae , phys .
d * 77 * , 106008 ( 2008 ) ; k. s. choi , int .
j. mod .
a * 22 * , 3169 ( 2007 ) , phys .
d * 74 * , 066002 ( 2006 ) ; c. beasley , j. j. heckman and c. vafa , arxiv:0806.0102 [ hep - th ] , and references therein . c. jarlskog , phys .
b * 82 * , 401 ( 1979 ) ; r. n. mohapatra , phys .
rev . lett . * 43 * , 893 ( 1979 ) .
n. cabibbo , phys .
* 10 * , 531 ( 1963 ) ; m. kobayashi and t. maskawa , prog .
phys . * 49 * , 652 ( 1973 ) .
z. maki , m. nakagawa and s. sakata , prog .
* 28 * , 870 ( 1962 ) ; b. pontecorvo , sov .
jetp * 26 * , 984 ( 1968 ) [ zh . eksp .
* 53 * , 1717 ( 1967 ) ] .
j. hisano , h. murayama and t. yanagida , nucl .
b * 402 * , 46 ( 1993 ) ; s. raby , arxiv : hep - ph/0211024 ; e. kearns , talk at snowmass 2001 , http://hep.bu.edu/~kearns/pub/kearns-pdk-snowmass.pdf , and references therein .
j. hisano , arxiv : hep - ph/0004266 .
p. fileviez perez , phys .
b * 595 * , 476 ( 2004 ) .
s. m. barr , phys .
b * 112 * , 219 ( 1982 ) ; j. p. derendinger , j. e. kim and d. v. nanopoulos , phys .
b * 139 * , 170 ( 1984 ) .
see , e.g. s. raby in w .-
yao _ et al _ , j. phys .
g * 33 * , 1 ( 2006 ) , and references therein .
y. aoki _ et al . _
[ rbc - ukqcd collaboration ] , arxiv:0806.1031 [ hep - lat ] ; y. kuramashi [ jlqcd collaboration ] , arxiv : hep - ph/0103264 .
h. georgi and c. jarlskog , phys .
b * 86 * , 297 ( 1979 ) .
for example , k. s. babu and r. n. mohapatra , phys .
lett . * 70 * , 2845 ( 1993 ) .
k. s. choi , nucl .
b * 708 * , 194 ( 2005 ) .
k. s. choi and t. kobayashi , nucl .
b * 797 * , 295 ( 2008 ) .
i. dorsner , p. fileviez perez and g. rodrigo , theories phys .
b * 649 * , 197 ( 2007 ) .
i. dorsner and p. fileviez perez , phys .
b * 625 * , 88 ( 2005 ) .
p. nath , a. h. chamseddine and r. arnowitt , phys .
d * 32 * , 2348 ( 1985 ) ; j. hisano , t. moroi , k. tobe and t. yanagida , mod .
lett . a * 10 * , 2267 ( 1995 ) . | conventionally we collect electron and up / down - quarks @xmath0 to form unified generations , and collect the left - handed and the right - handed fermions of the same flavors @xmath1 .
we can alternatively have a contorted multiplet , made by pairing different quarks and leptons , like @xmath2 or @xmath3 , and/or between different left and right handed fermions to @xmath4 , etc .
these can suppress proton decay , due to its high flavor dependence , while having the correct fermion masses .
unification of apparently different forces of nature into a simpler one has driven the history of physics . with the unification of weak and the electromagnetic forces , the establishment of the standard model of particle physics ( sm )
is based on the gauge theory of the strong and the electroweak interactions described by the group @xmath5 .
a further unification was considered and resulted in grand unified theory ( gut ) @xcite , possibly with the aid of the supersymmetry ( susy ) , which have been the best hint for the possible single nature of strong and electroweak forces .
being gauge interaction , extending the gauge groups into a ( semi-)simple embracing group naturally describes a unified force . for the size of unification scale ,
the running of gauge coupling is the most important hint on the gauge sector @xcite . besides the grand unification scheme at the field theory level ,
every known top - down approach , for instance string theory , has the same structure of a larger simple group and larger representations @xcite .
thus the embeddability of quantum numbers are quite compelling . consequently the transforming matters are also unified to yield larger one(s ) .
thus the extension fixes relative identities between the quark and the lepton families . upon forming unified multiplets ,
conventionally the following is assumed : 1 .
quarks ( color triplets ) and leptons ( color singlets ) form the same unified generations according to increasing order of masses .
2 . left - handed ( weak doublet ) and
right - handed ( weak singlet ) fermions of the same flavor form the same unified generations .
although conventional , there is _ no _ a priori _ reason to follow these assumptions .
_ by relaxing the first assumption we can consider @xmath6 and so on . for example , in one ` contorted ' version of @xmath7 model may contain @xmath8 is color index . ] . by relaxing the second assumption we can have @xmath9 and so on . in an extreme case , one unified generation can mix all the quarks and leptons @xmath10 , for instance .
no physical principle restricted such pairing , such as anomaly cancellation , showing each generation is completely closed even if we consider different pairing . for a formation as in ( [ twistgen ] ) , we will see that the mass hierarchy can be a guideline . |
automatic structures were introduced in @xcite and later again in @xcite along the lines of the bchi - rabin equivalence of automata and monadic second - order logic .
the idea is to encode elements of a structure @xmath4 via words or labelled trees ( the codes need not be unique ) and to represent the relations of @xmath4 via synchronised automata . this way we reduce the first - order theory of @xmath4 to the monadic second - order theory of one or two successors . in particular , the encoding of relations defined in @xmath4 by first order formulas are also regular , and automata for them can be computed from the original automata .
thus we have the fundamental fact that the first - order theory of an automatic structure is decidable .
depending on the type of elements encoding the structure , the following natural classes of structures appear : automatic ( finite words ) , @xmath3-automatic ( infinite words ) , tree - automatic ( finite trees ) , and @xmath3-tree automatic ( infinite trees ) .
besides the obvious inclusions , for instance that automatic structures are also @xmath3-automatic , there are still some some outstanding problems .
for instance , a presentation over finite words or over finite trees can be transformed into one where each element has a unique representative .
kuske and lohrey @xcite point out an @xmath3-regular equivalence relation ( namely @xmath5 stating that two infinite words are position - wise eventually equal ) with no @xmath3-regular set of representatives .
thus , unlike the finite - word case , injectivity can not generally be achieved by selecting a regular set of representatives from a given presentation .
in fact , using topological methods it has recently been shown @xcite that there are omega - automatic structures having no injective presentation .
however , we are able to prove that every omega - regular equivalence relation having only countably many classes does allow to select an omega - regular set of unique representants .
therefore , every countable omega - automatic structure does have an injective presentation . a related question raised by blumensath
@xcite is whether every countable @xmath3-automatic structure is also automatic . in corollary [ coroll_cnt ]
we confirm this by transforming the given presentation into an injective one , and then noting that an injective @xmath3-automatic presentation of a countable structure can be `` packed '' into one over finite words .
all these results rest on our main contribution : a characterisation of when there exist countably many words @xmath6 satisfying a given formula with parameters in a given @xmath3-automatic structure @xmath4 ( with no restriction on the cardinality of the domain of @xmath4 or the injectivity of the presentation ) . the characterisation is first - order expressible in an @xmath3-automatic presentation of an extension of @xmath4 by @xmath5 .
hence we obtain an extension of the fundamental fact for @xmath3-automatic structures to include cardinality and counting quantifiers such as there exists ( un)countably many , there exists finitely many , and there exists @xmath1 modulo @xmath2 many. this generalises results of kuske and lohrey @xcite who achieve this for structures with _ injective _ @xmath3-automatic presentations . -0.3
by countable we mean finite or countably infinite . let @xmath7 be a finite alphabet . with @xmath8 and @xmath9
we denote the set of finite , respectively @xmath3-words over @xmath7 .
the length of a word @xmath10 is denoted by @xmath11 , the empty word by @xmath12 , and for each @xmath13 the @xmath14th symbol of @xmath15 is written as @xmath16 $ ] . similarly @xmath17
$ ] is the factor @xmath18w[n+1]\cdots w[m]$ ] and @xmath19 is defined by @xmath20 $ ] .
note that we start indexing with @xmath21 and that for @xmath22 we denote by @xmath23 the concatenation of @xmath24 number of @xmath25s , in particular @xmath26 .
we consider relations on finite and @xmath3-words recognised by multi - tape finite automata operating in a synchronised letter - to - letter fashion .
formally , _
@xmath3-regular relations _ are those accepted by some finite non - deterministic automaton @xmath27 with bchi , parity or muller acceptance conditions , collectively known as @xmath3-automata , and having transitions labelled by @xmath2-tuples of symbols of @xmath7 .
equivalently , @xmath27 is a usual one - tape @xmath3-automaton over the alphabet @xmath28 accepting the _ convolution _ @xmath29 of @xmath3-words @xmath30 defined by @xmath31 = ( w_1[i ] , \ldots , w_m[i])$ ] for all @xmath14 .
words @xmath32 have _ equal ends _ , written @xmath33 , if for almost all @xmath34 , @xmath35 = v[n]$ ] .
this is an important @xmath3-regular equivalence relation .
we overload notation so that for @xmath36 we write @xmath37 to mean for almost all @xmath34 , @xmath38 .
[ ex_1 ] the non - deterministic bchi automaton depicted in fig .
[ fig_ex1 ] accepts the equal - ends relation on alphabet @xmath39 .
@xmath40^{\binom{0}{0},\binom{1}{1 } } & * { } & 2 \ar@(ul , ur)^{\binom{0}{0},\binom{1}{1 } } & f = \{2\ } , i = \{1\ } \\ } \ ] ] in the case of finite words one needs to introduce a padding end - of - word symbol @xmath41 to formally define convolution of words of different length . for simplicity , we shall identify each finite word @xmath10 with its infinite padding @xmath42 where @xmath43 . to avoid repeating the definition of automata for finite words ,
we say that a @xmath2-ary relation @xmath44 is _ regular _ ( _ synchronised rational _ ) whenever it is @xmath3-regular over @xmath45 . -0.3
cm we now define what it means for a relational structure ( we implicitly replace any structure with its relational counterpart ) to have an ( @xmath3-)automatic presentation .
[ def_as ] + consider a relational structure @xmath46 with universe @xmath47 and relations @xmath48 .
a tuple of @xmath3-automata @xmath49 together with a surjective naming function @xmath50 constitutes an _
( @xmath3-)automatic presentation _ of @xmath27 if the following criteria are met : a. the equivalence , denoted @xmath51 , and defined by @xmath52 is recognised by @xmath53 , b. every @xmath54 has the same arity as @xmath48 , c. @xmath55 is an isomorphism between @xmath56 and @xmath4 .
the presentation is said to be _ injective _ whenever @xmath55 is , in which case @xmath53 can be omitted . the relation @xmath51 needs to be a congruence of the structure @xmath57 for item ( iii ) to make sense . in case
@xmath58 only consists of words of the form @xmath59 where @xmath10 , we say that the presentation is _
automatic_. call a structure _ @xmath60automatic _ if it has an ( @xmath3-)automatic presentation .
the advantage of having an ( @xmath3-)automatic presentation of a structure lies in the fact that first - order ( @xmath61 ) formulas can be effectively evaluated using classical automata constructions .
this is expressed by the following fundamental theorem .
[ thrm_fo ] ( cf .
@xcite , @xcite , @xcite . )
+ a. there is an effective procedure that given an @xmath60automatic presentation @xmath62 of a structure @xmath4 , and given a @xmath61-formula @xmath63 with parameters @xmath64 from @xmath4 ( defining a @xmath1-ary relation @xmath65 over @xmath4 ) , constructs a @xmath1-tape synchronous @xmath60automaton recognising @xmath66 .
b. the @xmath61-theory of every @xmath60automatic structure is decidable . c. the class of @xmath60automatic structures
is closed under @xmath61-interpretations let @xmath67 denote the extension of first - order logic with all quantifiers of the form * @xmath68 meaning that the number of @xmath6 satisfying @xmath69 is finite and is congruent to @xmath70 mod @xmath2 ; * @xmath71 meaning that there are infinitely many @xmath6 satisfying @xmath69 ; * @xmath72 and @xmath73 meaning that the cardinality of the set of all @xmath6 satisfying @xmath69 is countable , or uncountable , respectively . it has been observed that for _ injective _ ( @xmath3-)automatic presentations theorem [ thrm_fo ] can be extended from @xmath61 to @xmath67 @xcite .
moreover , kuske and lohrey show that the cardinality of any set definable in @xmath67 is either countable or equal to that of the continuum .
our main contribution is the following generalisation of their result .
[ thrm_foc ] the statements of theorem [ thrm_fo ] hold true for @xmath67 over all ( not necessarily injective ) @xmath3-automatic presentations .
it is easily seen that finite - word automatic presentations can be assumed to be injective .
this is achieved by restricting the domain of the presentation to a regular set of representatives of the equivalence involved .
this can be done effectively , e.g. by selecting the length - lexicographically least word of every class .
this brings us to the question which @xmath3-automatic structures allow an injective @xmath3-automatic presentation . in @xcite kuske and lohrey
have pointed out that not every @xmath3-regular equivalence has an @xmath3-regular set of representatives .
in particular , the following lemma shows that the _ equal - ends relation @xmath5 _ of example [ ex_1 ] is a counterexample .
[ lemma_kl ] let @xmath27 be a bchi automaton with @xmath24 states over @xmath74 and let @xmath75 be given . consider the set @xmath76 .
then @xmath77 is uncountable if and only if @xmath78 , otherwise it is finite or countable .
the lemma implies that an @xmath3-regular set is countable if and only if it meets only finitely many equal - ends - classes . in this case
each of its members is ultimately periodic with one of finitely many periods .
[ coroll_countable ] an @xmath3-regular set is countable iff it can be written as a finite union of sets of the form @xmath79 with each @xmath80 a regular set of finite words and each @xmath81 a finite word . a related question raised by blumensath @xcite is whether every _ countable
_ @xmath3-automatic structure is also automatic .
it is easy to see that every _ injective _
@xmath3-automatic presentation of a countable structure can be `` packed '' into an automatic presentation .
[ prop_cnt_inj ] ( ( * ? ? ?
* theorem 5.32 ) ) let @xmath82 be an injective @xmath3-automatic presentation of a countable structure @xmath27 . then , an ( injective ) automatic presentation @xmath83 of @xmath27 can be effectively constructed . in our proof of theorem [ thrm_foc ]
we identify a property of finite semigroups that recognise transitive relations ( lemma [ lemma_ab ] item ( 3 ) ) that allows us to drop the assumption of injectivity in the previous statement .
we are thus able to answer the question of blumensath .
[ coroll_cnt ] a countable structure is @xmath3-automatic if and only if it is automatic . transforming a presentation of one type into the other can be done effectively . -0.3
cm the fundamental correspondence between recognisability by finite automata and by finite semigroups has been extended to @xmath3-regular sets .
this is based on the notion of _ @xmath3-semigroups_. rudimentary facts on @xmath3-semigroups are well presented in @xcite .
we only mention what is most necessary .
an @xmath3-semigroup @xmath84 is a two - sorted algebra , where @xmath85 is a semigroup , @xmath86 is the _ mixed product _ satisfying for every @xmath87 and every @xmath88 the equality @xmath89 and where @xmath90 is the _ infinite product
_ satisfying @xmath91 as well as the associativity rule @xmath92 for every sequence @xmath93 of elements of @xmath94 and every strictly increasing sequence @xmath95 of indices .
for @xmath96 we denote @xmath97 .
morphisms of @xmath3-semigroups are defined to preserve all three products as expected .
there is a natural way to extend finite semigroups and their morphisms to @xmath3-semigroups . as in semigroup theory ,
idempotents play a central role in this extension .
an _ idempotent _ is a semigroup element @xmath98 satisfying @xmath99 . for every element @xmath100 in a finite semigroup
the sub - semigroup generated by @xmath100 contains a unique idempotent @xmath101 .
the least @xmath102 such that @xmath101 is idempotent for every @xmath96 is called the _ exponent _ of the semigroup @xmath94 and is denoted by @xmath103 .
another useful notion is absorption of semigroup elements : say that _ @xmath100 absorbs @xmath104 ( on the right ) _ if @xmath105 .
there is also a natural extension of the free semigroup @xmath106 to the @xmath3-semigroup @xmath107 with @xmath108 and @xmath103 determined by concatenation .
an @xmath3-semigroup @xmath109 _ recognises _ a language @xmath110 via a morphism @xmath111 if @xmath112 .
this notion of recognisability coincides , as for finite words , with that by non - deterministic bchi automata . in @xcite constructions from bchi automata to @xmath3-semigroups and back are also presented .
[ thrm_omega_semigroups ] + a language @xmath110 is @xmath3-regular iff it is recognised by a finite @xmath3-semigroup .
we note that this correspondence allows one to engage in an algebraic study of varieties of @xmath3-regular languages , and also has the advantage of hiding complications of cutting apart and stitching together runs of bchi automata as we shall do .
this is precisely the reason that we use this algebraic framework .
most remarkably , one does not need to understand the exact relationship between automata and @xmath3-semigroups and the technical details of the constructions behind theorem [ thrm_omega_semigroups ] to comprehend our proof .
an alternative approach , though likely less advantageous , would be to use the composition method , which is closer in spirit to @xmath3-semigroups than to automata .
@xmath113 as the sets of bounded ( in terms of quantifier rank ) theories of finite , respectively , of @xmath3-words .
the composition theorem ensures that @xmath114 , @xmath108 , @xmath103 can naturally be defined on bounded theories . ]
this section is devoted to establishing the key to theorem [ thrm_foc ] announced earlier .
we characterise when there exist countably many words @xmath6 satisfying a given formula _ with parameters _
@xmath115 in some @xmath3-automatic structure @xmath4 .
the characterisation is first - order expressible in an @xmath3-automatic extension of @xmath4 by the equal - ends relation @xmath5 .
so , fix an @xmath3-automatic presentation of some @xmath4 with congruence @xmath51 , and a first - order formula @xmath115 in the language of @xmath4 with @xmath6 and @xmath116 free variables .
[ prop_count ] there is a constant @xmath117 , computable from the presentation @xmath82 , so that for all tuples @xmath116 of infinite words the following are equivalent : 1 .
@xmath118 is satisfiable and @xmath51 restricted to the domain @xmath118 has countably many equivalence classes . 2 .
there exist @xmath117-many words @xmath119 each satisfying @xmath118 , so that every @xmath6 satisfying @xmath118 is @xmath51-equivalent to some @xmath120 .
formally , the structure @xmath121 models the sentence below .
@xmath122 suppose @xmath82 , @xmath4 , and @xmath69 are given . define @xmath117 to be @xmath123 , where @xmath124 is the size of the largest @xmath3-semigroup corresponding to any of the given automata ( from the presentation or corresponding to @xmath69 ) .
now fix parameters @xmath116 . from now on
, @xmath51 denotes the equivalence relation @xmath51 restricted to domain @xmath118 .
@xmath125 : condition @xmath126 and the fact that every @xmath5-class is countable imply that all words satisfying @xmath118 are contained in a countable number of @xmath51-classes .
@xmath127 : we prove the contra - positive in three steps .
if @xmath118 is satisfiable then the negation of condition @xmath126 implies that there are @xmath128 many words @xmath129 each satisfying @xmath118 , and so that for @xmath130 , @xmath131 , the @xmath51-class of @xmath132 does not meet the @xmath5-class of @xmath133 . in particular
, the @xmath133s are pairwise @xmath134 .
the plan is to produce uncountably many pairwise non-@xmath51 words that satisfy @xmath118 . in the first ramsey step ,
similar to what is done in @xcite , we find two words from the given @xmath117 many , say @xmath135 , and a factorisation @xmath136 so that both words behave the same way along the factored sub - words with respect to the @xmath51- and @xmath69-semigroups . in the second coarsening step
we identify a technical property of finite semigroups recognising transitive relations .
this allows us to produce an altered factorisation @xmath137 and new , well - behaving words @xmath138 . in the final step
, the new words are shuffled along @xmath137 to produce continuum many pairwise non-@xmath51 words , each satisfying @xmath118 . -0.3
cm this step effectively allows us to discard the parameters @xmath116 . before we use ramsey s theorem
, we introduce a convenient notation to talk about factorisations of words .
let @xmath139 be any subset of @xmath140 and @xmath141 be a morphism into a finite semigroup @xmath142 .
for an @xmath3-word @xmath143 , and element @xmath98 , say that _
@xmath144 is an @xmath145-homogeneous factorisation of @xmath146 _ if for all @xmath147 , @xmath148 .
observe that 1 .
if @xmath144 is an @xmath149-homogeneous factorisation of @xmath146 and @xmath150 then the set @xmath151 is an @xmath152-homogeneous factorisation of @xmath146 .
if @xmath144 is an @xmath145-homogeneous factorisation of @xmath146 and @xmath153 is idempotent , then every infinite @xmath154 is also an @xmath145-homogeneous factorisation of @xmath146 . in the following we write @xmath155 and @xmath156 to denote the image of @xmath15 under the semigroup morphism into the finite semigroup associated to @xmath69 and @xmath51 , respectively , as determined by the presentation .
accordingly , we will speak of e.g. @xmath157-homogeneous factorisations .
let us now colour every @xmath158 ^ 2 $ ] , say @xmath159 , by the tuple of @xmath3-semigroup elements @xmath160 by ramsey s theorem there exists infinite @xmath136 and a tuple of @xmath3-semigroup elements @xmath161 so that for all @xmath162 , * @xmath163 is a @xmath157-homogeneous factorisation of the word @xmath164 , * @xmath163 is @xmath165-homogeneous factorisation of the word @xmath166 .
note that by virtue of additivity of our colouring and ramsey s theorem each of the @xmath167 and @xmath168 above are idempotents
. note that since there are at most @xmath124-many @xmath167s and @xmath124-many @xmath169s there are at most @xmath123 many pairs @xmath170 and so there must be two indices , we may suppose @xmath171 and @xmath126 , with @xmath172 and @xmath173 .
-0.3 cm for technical reasons we now refine @xmath163 and alter @xmath174 so that the semigroup elements have certain additional properties . to start with , using the fact that @xmath175 and our observation on coarsenings , we assume without loss of generality that @xmath163 is coarse enough so that @xmath176 for all @xmath34 . [ lemma_ab ] there exists a subset @xmath177 , listed as @xmath178 , and @xmath3-words @xmath138 with the following properties : 1 .
the words @xmath179 and @xmath180 are neither @xmath51-equivalent nor @xmath5-equivalent , and each satisfies @xmath118 .
2 . there exists an idempotent @xmath69-semigroup element @xmath100 such that @xmath137 is a @xmath181-homogeneous factorisation for each of @xmath182 and @xmath183 .
3 . there exist idempotent @xmath51-semigroup elements @xmath184 so that for @xmath185 * both @xmath186 and @xmath187 absorb @xmath104 * @xmath188 absorbs @xmath104 * @xmath137 is an @xmath189-homogeneous factorisation of @xmath190 * @xmath137 is an @xmath191-homogeneous factorisation of @xmath192 * @xmath137 is an @xmath193-homogeneous factorisation of @xmath194 .
define @xmath3-words @xmath195 , and @xmath180 by @xmath196 _ item 1 . _ clearly , @xmath197 and each @xmath185 satisfies @xmath198 since by homogeneity and @xmath172 @xmath199 and similarly @xmath200 next we check that @xmath201 .
@xmath202 thus , if @xmath203 then also @xmath204 and so * by transitivity * @xmath205 .
but since @xmath206 , the @xmath51-class of @xmath207 meets the @xmath5-class of @xmath208 , contradicting the initial choice of the @xmath133s .
+ _ items 2 and 3 . _
define intermediate semigroup elements @xmath209 , @xmath210 , @xmath211 and @xmath212
. then 1 .
both @xmath213 and @xmath214 absorb @xmath70 , since @xmath215 is idempotent ; 2 .
@xmath216 and thus absorbs @xmath70 ( for @xmath185 ) . in this notation , for all @xmath217 and @xmath185 , * @xmath218 is @xmath219 , * @xmath220 is @xmath221 , * @xmath222 is @xmath223 , * @xmath224 is @xmath225 .
+ finally , define the set @xmath226 , i.e. @xmath227 , and the semigroup elements @xmath228 , @xmath229 , @xmath230 and @xmath231 .
the extra multiple of @xmath1 ( defined as the product of the exponents of the give semigroups for @xmath5 and @xmath51 ) ensures all these semigroup elements ( in particular @xmath186 and @xmath187 ) are idempotent .
we now verify the absorption properties : @xmath232 similarly , @xmath233 absorbs @xmath104 .
further , since @xmath234 , we have @xmath235 and thus absorbs @xmath104 .
finally we verify the homogeneity properties : @xmath137 is an @xmath193-homogeneous factorisation of @xmath194 since for @xmath217 @xmath236 the other cases are similar . -0.3 cm we continue the proof of proposition [ prop_count ] by shuffling the words @xmath179 and @xmath180 along @xmath137 resulting in continuum many pairwise distinct words that are pairwise not @xmath51-equivalent , each satisfying @xmath118 . to this end , define for @xmath237 the characteristic word @xmath238 by @xmath239 first note that @xmath240 . indeed , by lemma [ lemma_ab ] item 2 @xmath241 and @xmath242 by lemma [ lemma_ab ] item 1 .
moreover , for @xmath243 the construction gives that @xmath244 .
this is due our initial choice of @xmath245 and the assumption that the factorisation @xmath246 is coarse enough so that @xmath176 and therefore also @xmath247 for all @xmath24 .
the following two lemmas establish that if @xmath243 then @xmath248 .
write @xmath249 for the word @xmath250 , and @xmath251 for @xmath252 and let @xmath253 denote @xmath254 .
[ lemma_shuffle ] for all @xmath243 , @xmath255 define semigroup - elements @xmath256 for @xmath34 by @xmath257 let @xmath2 be the smallest number in @xmath258 .
suppose that @xmath259 . because both @xmath186 and @xmath187 are idempotent and since @xmath104 is absorbed by both @xmath253 , @xmath186 and @xmath187 we have @xmath260 and the case that @xmath261 similarly results in @xmath262 . @xmath263 .
define an intermediate word @xmath264 . by computations similar to the above we
find that @xmath265 and @xmath266 therefore , if @xmath267 then also @xmath268 and so * by symmetry * and * by transitivity * @xmath269 . but in this case also @xmath270 , contradicting lemma [ lemma_ab ] item 1 .
there are continuum many classes in @xmath271 , thus there is a continuum of pairwise not @xmath51-equivalent words @xmath238 each satisfying @xmath118 .
this completes the proof of proposition [ prop_count ] . -0.7
* theorem [ thrm_foc ] * _ the statements of theorem [ thrm_fo ] hold true for @xmath67 over all ( not necessarily injective ) @xmath3-automatic presentations . _ * @xmath275 saying that @xmath276 and the last position where @xmath277 differs from @xmath124 is no larger than the last position where @xmath278 differs from @xmath124 , and * @xmath279 saying that @xmath275 and @xmath280 and , writing @xmath1 for this common position , the word @xmath281 $ ] is lexicographically smaller than the word @xmath282 $ ] .
@xmath284 where @xmath285 expresses that @xmath286 satisfy @xmath118 and there exists a position , say @xmath287 , so that every @xmath51-class contains a word satisfying @xmath118 that coincides with one of the @xmath133 from position @xmath1 onwards .
this additional condition can be expressed by @xmath288 consequently , @xmath289 can be eliminated since we can pick out unique representatives of the @xmath51-classes as those @xmath6 so that , writing @xmath290 for the smallest index @xmath14 for which @xmath291 , for every @xmath292 in the same @xmath51-class as @xmath6 , either the relation defined by @xmath300 is @xmath3-regular since it is a first order formula over @xmath3-regular relations . by assumption
it is non - empty .
thus it contains an ultimately periodic word of the form @xmath301 ) .
thus each of these @xmath302s is ultimately periodic ; say @xmath303 .
it is easy to construct an @xmath3-regular well - founded linear order on @xmath305 . for every @xmath306 ,
let @xmath307 be the length - lexicographically smallest word such that @xmath15 has period @xmath308 .
also let @xmath309 be the length - lexicographically smallest word so that @xmath310 .
define an order @xmath311 on @xmath305 by @xmath312 if @xmath308 is length - lexicographically smaller than @xmath313 , or otherwise if @xmath314 and @xmath315 is length - lexicographically smaller than @xmath316 .
the ordering @xmath311 is @xmath3-regular since it is @xmath61-definable in terms of @xmath3-regular relations .
finally , the required set of representatives may be defined as the set of @xmath311-minimal elements of every @xmath51-class ; and an automaton for this set can be constructed from an automaton for @xmath51 .
this immediately yields an _ injective _
@xmath3-automatic presentation from a given @xmath3-automatic presentation which by proposition [ prop_cnt_inj ] can be transformed into an automatic presentation of the structure .
thus we conclude that every countable @xmath3-automatic structure is already automatic . + * corollary [ coroll_cnt ] * _ a countable structure is @xmath3-automatic if and only if it is automatic . transforming a presentation of one type into the other
can be done effectively . _
+ note that some of our technical results , in particular lemmas [ lemma_ab ] and [ lemma_shuffle ] , only require transitivity of the relation @xmath51 and do not use symmetry .
applying them to an @xmath3-automatic linear order @xmath311 we get an interesting uncountable set of words of the form @xmath317 .
for any two such words with @xmath243 , whether @xmath318 or not depends only on the first position @xmath319 .
thus , @xmath311 behaves like the lexicographic order on such words .
[ [ failure - of - lwenheim - skolem - theorem - for - omega - automatic - structures ] ] failure of lwenheim - skolem theorem for @xmath3-automatic structures ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ while so far the area of automatic structures has mainly focused on individual structures , it is interesting to look at their theories as well .
we note a consequence of our work for automatic model theory. an automatic version of the downward lwenheim - skolem theorem would say that every uncountable @xmath3-automatic structure has a countable elementary substructure that is also @xmath3-automatic .
unfortunately this is false since there is a first - order theory with an @xmath3-automatic model but no countable @xmath3-automatic model .
indeed , consider the first - order theory of atomless boolean algebras .
kuske and lohrey @xcite have observed that it has an uncountable @xmath3-automatic model , namely @xmath320 .
however , khoussainov et al .
@xcite show that the countable atomless boolean algebra is not automatic and so , by corollary [ coroll_cnt ] , neither @xmath3-automatic .
[ prop_ls_omegaas ] let @xmath321 be an omega - automatic presentation of @xmath4 and let @xmath322 be its restriction to the ultimately periodic words of @xmath323 .
then @xmath322 is a countable elementary substructure of @xmath4 .
relying on the tarski - vaught criterion for elementary substructures we only need to show that for all first - order formulas @xmath324 and elements @xmath325 of @xmath322 @xmath326 by theorem [ thrm_fo ] @xmath324 defines an omega - regular relation and , similarly , since the parameters @xmath325 are all ultimately periodic the set defined by @xmath327 is omega - regular .
therefore , if it is non - empty , then it also contains an ultimately periodic word , which is precisely what we needed .
this proof can be viewed as a model construction akin to a classical compactness proof . indeed , starting with ultimately constant words and throwing in witnesses for all existential formulas satisfied in @xmath4 in each round
one constructs an increasing sequence of substructures comprising ultimately periodic words of increasing period lengths .
the union of these is closed under witnesses by construction .
the argument is valid for relational structures with constants assuming that every constant is represented by an ultimately periodic word . + * future work * it remains to be seen whether statements analogous to theorem [ thrm_foc ] and corollary [ coroll_cnt ] also hold for automatic presentations over infinite trees .
+ * acknowledgment * we thank the referees for detailed technical remarks and corrections . | we investigate structures that can be represented by omega - automata , so called omega - automatic structures , and prove that relations defined over such structures in first - order logic expanded by the first - order quantifiers ` there exist at most @xmath0 many ' , there exist finitely many and there exist @xmath1 modulo @xmath2 many are omega - regular .
the proof identifies certain algebraic properties of omega - semigroups . as a consequence an omega - regular equivalence relation of countable index
has an omega - regular set of representatives .
this implies blumensath s conjecture that a countable structure with an @xmath3-automatic presentation can be represented using automata on finite words .
this also complements a very recent result of hjrth , khoussainov , montalban and nies showing that there is an omega - automatic structure which has no injective presentation .
ukasz kaiser sasha rubin vince brny -0.3 cm |
in game theory , a fundamental notion is _ nash equilibrium _ ( ne ) , which is a state that is _ stable _ against deviations of any individual participants ( known as agents ) of the game in the sense that any such deviation will not bring about additional benefit to the deviator .
much stronger stability is exhibited by a _
strong nash equilibrium _ ( sne ) , a notion introduced by aumann @xcite , at which no coalition of agents exists such that each member of the coalition can benefit from coordinated deviations by the members of the coalition .
evidentally selfish individual agents stand to benefit from cooperation and hence snes are much more preferred to nes for stability .
however , snes do not necessarily exist @xcite and , even if they do , they are much more difficult to identify and to compute @xcite .
it is therefore very much desirable to have the advantages of both computational efficiency and strong stability , which motivates our study in this paper .
we establish that , for general ne job assignments in load balancing games , which exist and are easy to compute , their loss of strong stability possessed by snes is at most 25% . in a load balancing game , there are @xmath7 selfish agents , each representing one of a set @xmath8 of @xmath7 jobs . in the absence of a coordinating authority ,
each agent must choose one of @xmath0 identical servers , @xmath9 , to assign his job to in order to complete the job as soon as possible .
all jobs assigned to the same server will finish at the same time , which is determined by the workload of the server , defined to be the total processing time of the jobs assigned to the server .
let job @xmath10 have a processing time @xmath11 ( @xmath12 ) and let @xmath13 denote the set of jobs assigned to server @xmath14 ( @xmath15 ) . for convenience , we will use agent " and `` job '' interchangeably , and consider job processing times also as their `` lengths '' .
the completion time @xmath16 of job @xmath17 is the _ workload _ of its server : @xmath18 .
the notions of ne and sne can be stated more specifically for the load balancing game . a job assignment @xmath19 is said to be an ne if no individual job @xmath20 can reduce its completion time @xmath21 by unilaterally migrating from server @xmath14 to another server . a job assignment @xmath19 is said to be an sne if no subset @xmath22 of jobs can each reduce their job completion times by forming a coalition and making coordinated migrations from their own current servers
. nes in the load balancing game have been widely studied ( see , e.g. , @xcite ) with the main focus of quantifying their loss of global optimality in terms of the price of anarchy , a term coined by koutsoupias and papadimitriou @xcite , as largely summarized in @xcite . in this paper , we study nes in load balancing games from a different perspective by quantifying their loss of strong stability .
we focus on pure nes , those corresponding to deterministic job assignments in load balancing games .
while high - quality nes are easily computed , identification of an sne is strongly np - hard @xcite .
given any job assignment in a load balancing game , the _ improvement ratio _ ( ir ) of a deviation of a job is defined as the ratio between the pre- and post - deviation costs .
an ne is said to be a @xmath2-approximate sne ( @xmath3 ) ( which is called @xmath2-se in @xcite ) if there is no coalition of jobs such that each job of the coalition will have an ir more than @xmath2 from coordinated deviations of the coalition .
clearly , the stability of ne improves with a decreasing value of @xmath2 and a @xmath23-approximate sne is in fact an sne itself . for the load balancing game of two servers
, one can easily verify that every ne is also an sne @xcite .
if there are three or four servers in the game , then it is proved in @xcite and @xcite , respectively , that any ne assignment is a @xmath24-approximate sne , and the bound is tight .
furthermore , it is a @xmath25-approximate sne if the game has @xmath0 servers for @xmath26 @xcite .
we establish in this paper that , in the @xmath0-server load balancing game ( @xmath5 ) , any ne is a @xmath24-approximate sne , which is tight and hence closes the final gap in the literature on the study of ne approximation of sne in load balancing games . to establish our approximation bound
, we make a novel use of a powerful graph - theoretic tool .
we start with an example to help the reader get some intuition of the problem under consideration .
the example also provides a lower bound of @xmath27 for any ne assignment to approximate sne .
the left panel of fig .
[ fig : lower_bound ] below shows an ne assignment of six jobs to three identical machines with job completions @xmath28 , @xmath28 and @xmath29 , respectively , for the three pairs of jobs .
if the four jobs of lengths @xmath4 and @xmath28 form a coalition and make a coordinated deviation as shown in the figure , then in the resulting assignment , each of the four jobs in the coalition achieves an improvement ratio of @xmath27 . + as a tool of our analysis , we start with the minimal deviation graph introduced by chen @xcite . for convenience
we collect into this subsection some basic results on minimal deviation graphs from @xcite . given an ne job assignment @xmath30 , as an ne - based coalitional deviation or simply _ coalitional deviation _
@xmath31 , we refer to a collective action of a subset @xmath32 of jobs in which each job of @xmath33 migrates from its server in the assignment @xmath34 so that its completion time is decreased after the migration .
accordingly , @xmath35 is called the corresponding _ coalition_. we introduce deviation graphs to characterize coalitional deviations . in a coalitional deviation
, a server @xmath14 is said to be _ participating _ or _ involved _ if its job set changes after the deviation . given a coalitional deviation @xmath31 with the corresponding coalition @xmath36 , we define the corresponding ( directed ) deviation graph @xmath37 as follows : @xmath38 in what follows , without loss of generality we consider coalitional deviations with @xmath39 .
given a coalitional deviation @xmath31 , we denote by @xmath40 the workload of server @xmath14 after deviation @xmath31 , and by ir(@xmath31 ) the minimum of the improvement ratios of all jobs taking part in @xmath31 . then we have the following definition and lemmas from @xcite : [ lem : out - degree ] the out - degree @xmath41 of any node @xmath14 of a deviation graph is at least 1 , and hence @xmath42 .
[ lem : no - cycles ] if all @xmath0 servers are involved in a coalitional deviation , then the deviation graph does not contain a set of node - disjoint directed cycles such that each node of the graph is in one of the directed cycles .
[ def : min - deviation ] let @xmath31 be a coalitional deviation and @xmath35 be the corresponding coalition .
deviation graph @xmath43 is said to be _ minimal _ if @xmath44 for any coalitional deviation @xmath45 such that the corresponding coalition @xmath46 is a proper subset of @xmath33 .
[ lem : in - degree ] the in - degree @xmath47 of any node @xmath14 of a minimal deviation graph is at least 1 .
[ lem : strong - connectivity ] a minimal deviation graph is strongly connected . in our study of bounding ne approximation of sne
, we can apparently focus on those coalitional deviations that correspond to minimal deviation graphs .
we start with several observations on any ne - based coalitional deviation @xmath31 involving @xmath0 servers for @xmath5 .
let @xmath48 denote the corresponding minimal deviation graph .
if two jobs assigned to server @xmath49 in the ne assignment migrate to server @xmath50 ( @xmath51 ) together , or both stay on the server , then we can treat them as one single job without loss of generality in our study of the minimal deviation graph . with this understanding ,
if we let @xmath52 ( @xmath49 ) denote the number of jobs assigned to server @xmath14 in the ne assignment , then the following is immediate .
[ obs : out - degrees ] for any @xmath49 , we have @xmath53 . @xmath54 or @xmath55 . as a result of the above observation ,
the node set @xmath56 can be partitioned into two , @xmath57 and @xmath58 , as follows : @xmath59 by applying a data scaling if necessary , we assume without loss of generality that @xmath60 [ obs : bound - of - li ] for any @xmath61 , we have @xmath62 .
suppose to the contrary that @xmath63 , which implies that @xmath64 .
let @xmath65 denote the length of the shortest job assigned to server @xmath14 in the ne assignment .
we have @xmath66 , which leads to @xmath67 , that is , @xmath68 , which implies that the shortest job assigned to server @xmath14 in the ne assignment can have the benefit of reducing its job completion time by unilaterally migrating to the server of which the workload is 1 , contradicting the ne property .
the following observation states that , if all jobs on a server participate in the migration , then none of the servers they migrate to will have _ all _ its jobs migrate out .
[ obs : out - degree - relation ] if @xmath69 and @xmath70 , then @xmath71 .
suppose to the contrary that @xmath72 . according to observation [ obs : out - degrees ]
, we have @xmath73 , which implies that all the jobs assigned to server @xmath14 and server @xmath74 in the ne assignment belong to coalition @xmath33 . since @xmath69 , there is a job @xmath75 that migrates from server @xmath14 to server @xmath74 .
consider the new coalition @xmath76 formed by all members of @xmath33 except @xmath77 .
then we have @xmath78 .
let @xmath45 be such a coalitional deviation of @xmath76 that is the same as @xmath31 except without the involvement of @xmath77 and the job(s ) that migrate(s ) to @xmath14 ( resp .
@xmath74 ) in @xmath31 will migrate to @xmath74 ( resp .
@xmath14 ) in @xmath45 .
then we have @xmath79 , contradicting the minimality of the deviation graph @xmath80 according to definition [ def : min - deviation ] .
the following observation is a direct consequence of observation [ obs : out - degree - relation ] : [ obs : equiv - min - deviation ] assume @xmath81 . hence @xmath82 according to observation [ obs : out - degree - relation ] .
let @xmath45 be the same as @xmath31 except that any job that migrates to @xmath14 ( resp .
@xmath74 ) in @xmath31 will migrate to @xmath74 ( resp .
@xmath14 ) in @xmath45 .
then @xmath79 , and @xmath83 is also minimal .
to help our analysis , we will introduce in this section a special arc set @xmath84 in the minimal deviation graph @xmath48 . for any node @xmath49 , denote @xmath86 and @xmath87 . for notational convenience , for any node set @xmath88 , we denote @xmath89 and @xmath90 . with @xmath91
replaced by @xmath85 above , we similarly define @xmath92 , @xmath93 , @xmath94 and @xmath95 .
let us define @xmath85 as follows . according to lemma [ lem : in - degree ] , @xmath96 for any @xmath61 .
for each @xmath61 we pick up an arc from the non - empty set @xmath97 to form an @xmath0-element subset @xmath98 .
then @xmath85 possesses the following property : @xmath99 denote @xmath100 for any @xmath61 .
then it is clear that @xmath101 if node set @xmath34 is a singleton , then we will also use @xmath34 to denote the singleton if no confusion can arise .
hence , due to ( [ eqn : in - degree ] ) we will also use @xmath102 to denote the single _ element _ of the corresponding set .
any arc set @xmath98 that possesses property ( [ eqn : in - degree ] ) is said to be _ tilde - valid_. our main result is stated in the following theorem .
[ thm : main ] for any minimal deviation graph @xmath103 involving @xmath0 servers , its improvement ratio @xmath104 .
let us perform some initial investigation to see what we need to do to prove the theorem .
recall that , for any @xmath61 , @xmath52 is the number of jobs assigned to server @xmath14 in the ne assignment and @xmath105 for a fixed arc set @xmath85 defined in section [ sec : auxiliary_arc_set ] for the minimal deviation graph @xmath103 . for
a pair of integers @xmath106 and @xmath107 with @xmath108 and @xmath109 , let @xmath110 .
then it is clear that @xmath111 denote @xmath112 for all possible pairs @xmath106 and @xmath107 : @xmath113 and @xmath114 .
let @xmath115 . then according to ( [ eqn :
in - degree ] ) and ( [ eqn : def - of - mab ] ) , we have @xmath116 according to the definition of ir , we have @xmath117 for @xmath118 .
summing up these inequalities over all @xmath0 arcs in @xmath85 leads to @xmath119 which implies that @xmath120 according to observation [ obs : bound - of - li ] , we have @xmath121 , which implies that the right - hand side of ( [ eqn : bound - of - r ] ) , which we denote by @xmath122 , is at most 2 , since @xmath122 is a convex combination of @xmath123 and @xmath124 with the corresponding combination coefficients @xmath125 ( @xmath126 ) and @xmath127 due to ( [ eq : scalling ] ) and ( [ eqn : out - degree ] ) . on the other hand , since @xmath128 according to the definition , we conclude that @xmath129 , which implies that @xmath122 is a decreasing function of @xmath130 for which @xmath131 or @xmath132 , and an increasing function of @xmath130 for which @xmath133
. therefore , we increase @xmath122 by increasing @xmath130 to @xmath134 for @xmath14 such that @xmath133 , and by decreasing @xmath130 to @xmath23 for @xmath14 such that @xmath131 or @xmath132 . noticing that @xmath135 according to observation [ obs : out - degrees ]
, we obtain @xmath136 which together with ( [ eqn : equations - for - dab ] ) implies that @xmath137 in order to prove theorem [ thm : main ] , we need to show @xmath138 .
then it suffices to show @xmath139 which is equivalent to @xmath140 by replacing the right - hand side @xmath0 of the above inequality with the left - hand side of the second equality in ( [ eqn : equations - for - dab ] ) , we have @xmath141 that is @xmath142 in what follows , we are to prove ( [ eqn : ultimate ] ) and thereby theorem [ thm : main ] through a series of lower bounds established in section [ sec : bounding ] on different terms of the right - hand side of inequality ( [ eqn : ultimate ] ) .
we introduce an auxiliary node set @xmath143 in addition to the auxiliary arc set @xmath85 introduced earlier .
let @xmath144 then we immediately have [ lem : w0 ] @xmath145 .
suppose to the contrary that @xmath146 .
then any @xmath147 in ( [ eqn : out - degree ] ) , which implies that @xmath132 for any @xmath49 , so that @xmath85 forms some node - disjoint directed cycles that span all nodes , contradicting lemma [ lem : no - cycles ] . note that , from the formation of arc set @xmath85 , it is clear that @xmath85 as a tilde - valid arc set may not be unique . however , among all possible choices of a tilde - valid arc set @xmath98 , we choose one that has some additional properties in terms of minimum cardinalities of some combinatorial structures , which we shall define in due course .
these additional properties will be presented in a sequence of three assumptions , which are made without loss of generality due to the finiteness of the total number of tilde - valid arc sets . with the same reason
, we assume that our coalitional deviation @xmath31 is chosen in such a way that it has a certain property ( see assumption [ ass : delta ] ) .
[ ass : mimimum - w0 ] arc set @xmath85 is tilde - valid and it minimizes @xmath148 . let @xmath149 . then @xmath150 according to lemmas [ lem : w0 ] and [ lem : out - degree ] .
a node @xmath61 is said to be _ associated _ with @xmath151 if it is _ linked _ to an element of @xmath152 through a sequence of arcs ( but not a directed path ) in @xmath85 and @xmath91 in alternation ( see fig . [
fig : w_1 ] for an illustration ) .
more formally , @xmath61 is associated with @xmath151 if and only if , for some integer @xmath153 , there are nodes @xmath154 with @xmath155 and @xmath156 , such that @xmath157 nb : the solid arcs belong to @xmath85 and dotted arcs to @xmath91 + , title="fig : " ] note that in the above definition , if @xmath158 is associated with @xmath151 , then @xmath159 , , @xmath160 used in ( [ eqn : def - of - w1 ] ) are each associated with @xmath151 .
define @xmath161 immediately we have @xmath162 , which implies that @xmath163 on the other hand , since @xmath164 according to ( [ eqn : in - degree ] ) , we have @xmath165 . [ lem : w1 ] for any @xmath166 , @xmath167 .
furthermore , @xmath168 .
it is clear from the definition that @xmath169 .
hence @xmath147 for any @xmath166 .
assume for contradiction that @xmath170 for some @xmath166 .
since @xmath14 is associated with @xmath151 , in addition to nodes @xmath171 satisfying ( [ eqn : def - of - w1 ] ) , we have a node @xmath172 ( hence @xmath173 ) such that @xmath174 according to the definition of @xmath152 .
now we remove @xmath175 arcs @xmath176 from @xmath85 and add @xmath175 new arcs @xmath177 to @xmath85 .
it is easy to see that the new set @xmath85 still has property ( [ eqn : in - degree ] ) .
additionally , under the new @xmath85 , all @xmath178 remain the same except two of them : @xmath179 and @xmath180 , with the former increased by 1 and the latter decreased by 1 . since @xmath170 under the original @xmath85 , then @xmath181 under the new @xmath85 .
consequently , the new @xmath151 determined by the new @xmath85 contains a smaller number of elements , contradicting assumption [ ass : mimimum - w0 ] about the original @xmath85 . to prove the second part of the lemma ,
let us first prove @xmath182 .
let @xmath183 and @xmath118 .
we show that @xmath184 .
in fact , since @xmath185 according to ( [ eqn : in - degree ] ) , we have a node @xmath186 such that @xmath187 .
now since @xmath14 is associated with @xmath151 , we conclude that @xmath188 is also associated with @xmath151 , which implies that @xmath184 .
therefore , with ( [ eqn : tilde - w0-versus - tilde - w1 ] ) we have proved that @xmath189 .
the other direction of the inclusion is apparent .
it follows from lemma [ lem : w1 ] and ( [ eqn : in - degree ] ) that the mapping @xmath190 from @xmath191 onto @xmath192 is a one - to - one correspondence and hence @xmath193 let @xmath194 as we can see from inequality ( [ eqn : ultimate ] ) , bounding the sizes @xmath195 and @xmath196 of the respective sets @xmath197 and @xmath198 is vital in our establishment of the desired bound .
we therefore take a close look at the two sets by partitioning @xmath199 into a number of subsets , so that different bounding arguments can be applied to different subsets .
we assemble our notations here in one place for easy reference and the reader is advised to conceptualize each _ only _ when it is needed in an analysis at a later point .
let @xmath200 . for convenience ,
we reserve letter @xmath201 to exclusively index elements of @xmath202 and let @xmath203 with the understanding that it is always the case that @xmath204 . for any @xmath205
, @xmath204 implies @xmath206 since @xmath207 . on the other hand , since arc @xmath208 , we have @xmath209 according to observation [ obs : out - degree - relation ] , which implies that @xmath210 must belong to one of the following three mutually disjoint node sets : @xmath211 therefore , if we define @xmath212 then we have @xmath213 and @xmath214 for any @xmath215 .
in other words , for any element @xmath216 , the two - element set @xmath217 has exactly one element in @xmath143 .
now let @xmath218 clearly , @xmath219 ( @xmath220 ) and @xmath221 .
to bound from below the right - hand side of the key inequality ( [ eqn : ultimate ] ) , or equivalently , to bound from above the left - hand side of ( [ eqn : ultimate ] ) , we establish through a series of five lemmas and a corollary that the number of nodes in @xmath222 is at most @xmath223 , where @xmath224 and @xmath225 are defined below for @xmath226 with @xmath227 mutually disjoint node sets .
we divide our proofs into two parts with the second part on bounding @xmath228 .
let us start with some straightforward upper bounds .
since @xmath229 for any @xmath230 according to the definition of @xmath231 , we immediately have the following lemma thanks to observation [ obs : out - degree - relation ] . [
lem : x2 ] let @xmath232
. then @xmath233 and @xmath234 .
@xmath235 note that @xmath236 for any @xmath237 and @xmath238 ( @xmath239 ) due to ( [ eqn : in - degree ] ) , which lead to the following lemma .
. then @xmath241 and @xmath242 .
@xmath235 the following lemma follows directly from the definition of @xmath243 : let @xmath244
. then @xmath245 and @xmath246 .
for any @xmath247 , @xmath248 and @xmath249 , unless @xmath250 .
@xmath235 at this point , we introduce our second additional assumption about @xmath85 without loss of generality .
[ ass : m-2 - 2 ] arc set @xmath85 is such that it first satisfies assumption [ ass : mimimum - w0 ] and then minimizes @xmath251 . for any @xmath252 ,
since @xmath253 according to the definition of @xmath254 , there is @xmath255 . then @xmath256 ( otherwise we would have @xmath257 according to lemma [ lem : w1 ] ) .
in fact , node @xmath74 has the following property : @xmath258 to see this , consider replacing @xmath259 with @xmath260 in @xmath85 to form a new tilde - valid arc set @xmath261 .
it is easy to see that @xmath261 satisfies assumption [ ass : mimimum - w0 ] .
however , with the new arc set @xmath261 , @xmath201 is no longer a node in the new @xmath262 , which implies that @xmath74 has to become a node in @xmath262 in order not to contradict assumption [ ass : m-2 - 2 ] with the original choice of @xmath85 , which in turn implies properties ( [ eqn:2nd - special - node - property ] ) .
furthermore , since @xmath263 and @xmath264 , there is no @xmath265 such that @xmath266 , which implies that @xmath267 .
consequently , we have the following lemma . [
lem : x5 ] let @xmath268
. then @xmath269 and @xmath270 .
@xmath235 now let us establish an upper bound on @xmath271 in the following lemma with the minimality of our deviation graph @xmath48 .
[ lem : x1 ] let @xmath272 . then @xmath273 .
suppose to the contrary that @xmath274 , that is , @xmath275 according to ( [ eqn : cadinality - of - w1 ] ) .
let @xmath276 be a proper subset of @xmath277 elements .
define @xmath278 where @xmath279 ( see fig . [
fig : lemma_proof ] for an illustration with explanations to follow ) .
, title="fig : " ] + then @xmath280 since @xmath281 .
we claim @xmath76 is a proper subset of @xmath33 .
to see this , let @xmath282 .
since @xmath283 ( lemma [ lem : w1 ] ) , we have @xmath284 . observation [ obs : out - degree - relation ] implies @xmath285 .
therefore , we have @xmath286 , i.e. , @xmath287 , but @xmath288 . with the same arguments we note that the three constituent subsets of @xmath289 are mutually disjoint . in fig .
[ fig : lemma_proof ] , the set @xmath290 is a subset of @xmath291 according to the definition of @xmath292 and the mapping between @xmath290 and @xmath293 is a one - to - one correspondence due to equation ( [ eqn : bipartition ] ) .
since @xmath294 , we can assume there is a one - to - one correspondence @xmath295 between the nodes ( i.e. , servers ) of the two sets @xmath296 and @xmath192 .
now let us define a new coalitional deviation @xmath45 with @xmath46 , which is the same as @xmath31 restricted on @xmath76 except that , if @xmath297 migrates in @xmath31 to a server of @xmath192 , then let @xmath10 migrate in @xmath45 to the corresponding ( under @xmath295 ) server of @xmath296 .
we show that the improvement ratio of any job deviation in @xmath45 is at least the same as that in @xmath31 , which then implies that @xmath298 , contradicting the minimality of @xmath43 according to definition [ def : min - deviation ] . to this end
, we only need to show that the new coalitional deviation @xmath45 takes place among the servers assigned with jobs of the coalition @xmath76 , that is , @xmath299
so that benefit of any job deviation will not decrease due to the fact that _ all _ jobs on servers of @xmath300 migrate out in @xmath31 and hence in @xmath45 as well , leaving empty space for deviational jobs under @xmath45 , which originally migrate to servers of @xmath192 under @xmath31 .
first we have @xmath301 according to lemma [ lem : w1 ] . on the other hand
, it can be easily verified that @xmath302 according to the definition of @xmath293 .
now we show @xmath303 , which then implies ( [ eqn : grand - inclusion ] ) .
in fact , for any @xmath304 , noticing that @xmath305 , according to the definitions of @xmath293 and @xmath292 , we have @xmath306 , which implies that @xmath307 according to the definition of @xmath291 .
to prove our final upper bound , we need to introduce the following two structures in graph @xmath48 with tilde - valid arc set @xmath85 : @xmath308 note that each element in @xmath309 represents a directed @xmath4-cycles of both arcs in @xmath85 and each element in @xmath310 is a directed 2-path of both arcs in @xmath85 . in both cases of @xmath309 and @xmath310 ,
the interior node @xmath311 has an in - degree @xmath312 and all its out - arcs are in @xmath85 .
our next result is based on the following further refinement of the tilde - valid arc set @xmath85 .
[ lem : omega ] if @xmath313 for some arc set @xmath85 satisfying assumption [ ass : m-2 - 2 ] , then there exists an arc set @xmath261 such that , while it also satisfies assumption [ ass : m-2 - 2 ] , additionally , @xmath314 is a proper subset of @xmath309 .
assume @xmath315 and let @xmath316 be as in the definition of @xmath309 . then there must be a node @xmath317 with @xmath318 , since otherwise @xmath319 , which implies that there would be no directed path from any other nodes in @xmath48 to nodes @xmath14 or @xmath311 , contradicting lemma [ lem : strong - connectivity ] .
therefore , the following set is not empty : @xmath320 let @xmath321 .
we define a new tilde - valid arc set @xmath322 it is easily seen that @xmath323 and @xmath324 . on the other hand ,
@xmath261 still satisfies assumption [ ass : mimimum - w0 ] due to @xmath325 , and hence also satisfies assumption [ ass : m-2 - 2 ] since @xmath326 ( which implies that @xmath327 ) according to observation [ obs : out - degree - relation ] ( as no other node not in @xmath328 can possibly become a member of @xmath329 ) . as a result of lemma [ lem : omega ]
, we can further refine our initial choice of @xmath85 so that it satisfies the following assumption , where the benefit of minimizing @xmath330 will be seen in the proof of lemma [ lem : company ] ( see inequality ( [ eqn : pi ] ) ) .
[ ass : omega - pi ] arc set @xmath85 is such that it first satisfies assumption [ ass : m-2 - 2 ] and then lexicographically minimizes @xmath331 .
[ cor : omega ] any arc set @xmath85 satisfying assumption [ ass : omega - pi ] must satisfy @xmath332 .
@xmath235 an arc set @xmath85 in graph @xmath48 that satisfies assumption [ ass : omega - pi ] is said to be _ derived _ from @xmath31 . without loss of generality ,
our coalitional deviation @xmath31 is considered to have been chosen so that it satisfies the following assumption .
[ ass : delta ] coalitional deviation @xmath31 defining minimal deviation graph @xmath48 is such that the arc set @xmath85 derived from @xmath31 gives lexicographical minimum @xmath333 let minimal deviation graph @xmath48 with @xmath31 satisfying assumption [ ass : delta ] be given . for any @xmath334
, there is @xmath335 , such that @xmath336 ( note ( [ eqn : bipartition ] ) ) and @xmath337 . given @xmath334 and @xmath338 .
since @xmath339 according to the definition of @xmath340 , we have @xmath341 and hence @xmath342 since @xmath343 according to observation [ obs : out - degree - relation ] . since @xmath344 and @xmath345 ( again according to the definition of @xmath340 ) ,
we let @xmath346 .
then @xmath335 since otherwise we would have @xmath347 , contracting corollary [ cor : omega ] with our assumption [ ass : omega - pi ] ( see fig . [
fig : company ] for an illustration with more explanations to follow ) . of @xmath348 or new coalitional deviation @xmath45,title="fig : " ] + we claim @xmath71 and hence are done .
let us assume for a contradiction that @xmath349 .
note that with @xmath350 replacing @xmath351 in the definition of @xmath310 , we conclude that @xmath352 .
now let us define a new coalitional deviation @xmath45 so that its derived arc set @xmath261 gives a @xmath353 .
in fact , let @xmath45 be defined as in observation [ obs : equiv - min - deviation ] after node @xmath14 has been replaced by @xmath201 in the statement of observation [ obs : equiv - min - deviation ] .
denote @xmath354 as the arc set of the resulting minimal deviation graph @xmath83 .
let @xmath261 be the natural result of @xmath85 after the re - orientation from @xmath31 and @xmath45 , i.e. , an arc in @xmath85 pointing to @xmath201 ( resp .
@xmath74 ) will become an arc in @xmath261 pointing to @xmath74 ( resp .
@xmath201 ) .
other arcs are the same for @xmath85 and @xmath261 .
apparently , @xmath355 has increased from @xmath356 , then clearly it must be the result of @xmath201 and/or @xmath74 becoming element(s ) of @xmath314 . in any such case ( say , the former case for the sake of argument ) , based on the definition of @xmath314
, we can use the approach in lemma [ lem : omega ] to find @xmath357 as defined in ( [ eqn : h ] ) and perform an arc - swap as in ( [ eqn : one - arc - swap ] ) with @xmath14 and @xmath311 replaced by @xmath201 and @xmath210 , respectively , to reduce @xmath358 while maintaining the values of @xmath359 and @xmath360 . for convenience
, we still use @xmath261 to denote the tilde - valid arc set after such arc - swap(s ) if needed .
consequently , we have @xmath361 however , we claim @xmath362 a desired contradiction . to see inequality ( [ eqn : pi ] ) , we first note that ( i ) any 2-path in @xmath310 starting at @xmath363 is also a 2-path in @xmath364 , and vice versa , and ( ii ) any 2-path in @xmath310 ( resp .
@xmath364 ) starting at @xmath201 ( resp .
@xmath74 ) must have the first arc @xmath259 ( resp .
@xmath365 , since @xmath366 due to @xmath367 according to ( [ eqn : bipartition ] ) ) . on the other hand ,
the following can be easily observed : 1 . if @xmath368 ( @xmath369 ) , then @xmath370 , and vice versa .
2 . if @xmath371 ( @xmath372 ) , then @xmath373 , and vice versa .
@xmath374 , since @xmath375 would imply @xmath376 by definition of @xmath364 and hence @xmath377 by definition of @xmath354 , which in turn implies that @xmath378 since @xmath379 under @xmath85 .
consequently , we obtain @xmath380 , contradicting corollary [ cor : omega ] .
4 . with similar reasons for @xmath381
, we have @xmath382 .
therefore , overall @xmath364 contains at least one element less than @xmath310 as indicated in points 3 and 4 above .
we call @xmath337 identified in the above lemma a _ company _ of @xmath348 .
clearly , any @xmath337 can not be a company of two different elements of @xmath340 according to the statement of the lemma , which leads us to the following corollary .
[ cor : x6 ] denote @xmath383:= \{j\in m\backslash x:\ \textrm{node $ j$ is a company of $ \ell$}\}$ ] for any @xmath348 and let @xmath384)$ ] .
then @xmath385 and @xmath386 .
@xmath235 we have used the cardinalities of the six sets @xmath387 to bound @xmath271 , @xmath388 , @xmath389 , @xmath390 , @xmath391 and @xmath392 , respectively .
let us make sure these sets do not overlap with @xmath393 and are mutually disjoint . according to lemmas [ lem : x2][lem : x5 ] and corollary [ cor : x6 ] , we have @xmath394 and @xmath395 hence @xmath396 and @xmath397 ( @xmath398 ) . since @xmath399 for any @xmath400 ( definition of @xmath85 ) and @xmath401 for any @xmath14 ( observation [ obs : out - degree - relation ] ) , noticing that @xmath402 ( lemma [ lem : w1 ] ) , we conclude that @xmath403
now we are ready to go back to proving ( [ eqn : ultimate ] ) and hence theorem [ thm : main ]
. since @xmath404 , the left - hand side of inequality ( [ eqn : ultimate ] ) is at most @xmath405 on the other hand , if we let @xmath406 which imply @xmath407 then noticing the properties ( [ eqn : disjoint - y ] ) and that @xmath408 we see that the right - hand side of inequality ( [ eqn : ultimate ] ) is at least @xmath409 according to lemmas [ lem : x2][lem : x1 ] and corollary [ cor : x6 ] , the right - hand side of inequality ( [ eqn : lhs ] ) is at most that of ( [ eqn : middle ] ) , which in turn ultimately leads to inequality ( [ eqn : ultimate ] ) .
consequently , theorem [ thm : main ] is established . from theorem [ thm : main ] and the lower bound demonstrated in section [ sec
: lower_bound ] , the following theorem follows . in the m - server load balancing game ( @xmath5 ) , any ne is a @xmath6-approximate sne and the bound is tight .
by establishing a tight bound of @xmath27 for the approximation of general nes to snes in the @xmath0-server load balancing game for @xmath5 , we have closed the final gap for the study of approximation of general nes to snes . however , as demonstrated by feldman & tamir @xcite and by chen @xcite , a special subset of nes known as lpt assignments , which can be easily identified as nes @xcite , do approximate snes better than general nes .
it is still a challenge to provide a tight approximation bound for this subset of nes .
research by the second and third author was partially supported by the national natural science foundation of china ( grant no .
11071142 ) and the natural science foundation of shandong province , china ( grant no .
zr2010am034 ) .
fotakis d. , kontogiannis s. , mavronicolas m. , and spiraklis p. , the structure and complexity of nash equilibria for a selfish routing game .
_ proc . of the 29th international colloquium on automata , languages and programming _ , 2002 , 510519 . | we study strong stability of nash equilibria in load balancing games of @xmath0 ( @xmath1 ) identical servers , in which every job chooses one of the @xmath0 servers and each job wishes to minimize its cost , given by the workload of the server it chooses . a nash equilibrium ( ne ) is a strategy profile that is resilient to unilateral deviations . finding an ne in such a game is simple .
however , an ne assignment is not stable against coordinated deviations of several jobs , while a strong nash equilibrium ( sne ) is . we study
how well an ne approximates an sne . given any job assignment in a load balancing game , the improvement ratio ( ir ) of a deviation of a job is defined as the ratio between the pre- and post - deviation costs .
an ne is said to be a @xmath2-approximate sne ( @xmath3 ) if there is no coalition of jobs such that each job of the coalition will have an ir more than @xmath2 from coordinated deviations of the coalition . while it is already known that nes are the same as snes in the @xmath4-server load balancing game , we prove that , in the @xmath0-server load balancing game for any given @xmath5 , any ne is a @xmath6-approximate sne , which together with the lower bound already established in the literature yields a tight approximation bound .
this closes the final gap in the literature on the study of approximation of general nes to snes in load balancing games . to establish our upper bound
, we make a novel use of a graph - theoretic tool .
* keywords : * load balancing game , nash equilibrium , strong nash equilibrium , approximate strong nash equilibrium |
we consider programs containing high security inputs and low security outputs . informally , the quantitative information flow problem concerns the amount of information that an attacker can learn about the high security input by executing the program and observing the low security output .
the problem is motivated by applications in information security .
we refer to the classic by denning @xcite for an overview .
in essence , quantitative information flow measures _ how _ secure , or insecure , a program is .
thus , unlike non - interference @xcite , that only tells whether a program is completely secure or not completely secure , a definition of quantitative information flow must be able to distinguish two programs that are both interferent but have different degrees of `` secureness . ''
for example , consider the following two programs : @xmath0 in both programs , @xmath1 is a high security input and @xmath2 is a low security output .
viewing @xmath1 as a password , @xmath3 is a prototypical login program that checks if the guess @xmath4 matches the password .
is a program constant .
see section [ sec : prelim ] for modeling attacker / user ( i.e. , low security ) inputs . ] by executing @xmath3 , an attacker only learns whether @xmath1 is equal to @xmath4 , whereas she would be able to learn the entire content of @xmath1 by executing @xmath5 .
hence , a reasonable definition of quantitative information flow should assign a higher quantity to @xmath5 than to @xmath3 , whereas non - interference would merely say that @xmath3 and @xmath5 are both interferent , assuming that there are more than one possible value of @xmath1 .
researchers have attempted to formalize the definition of quantitative information flow by appealing to information theory .
this has resulted in definitions based on the shannon entropy @xcite , the min entropy @xcite , the guessing entropy @xcite , and channel capacity @xcite .
much of the previous research has focused on information theoretic properties of the definitions and approximate ( i.e. , incomplete and/or unsound ) algorithms for checking and inferring quantitative information flow according to such definitions . in this paper
, we give a verification theoretic and complexity theoretic analysis of quantitative information flow and investigate precise methods for checking quantitative information flow . in particular , we study the following _ comparison problem _ : given two programs @xmath3 and @xmath5 , decide if @xmath6 . here
@xmath7 denotes the information flow quantity of the program @xmath8 according to the quantitative information flow definition @xmath9 where @xmath9 is either @xmath10 $ ] ( shannon - entropy based with distribution @xmath11 ) , @xmath12 $ ] ( min - entropy based with distribution @xmath11 ) , @xmath13 $ ] ( guessing - entropy based with distribution @xmath11 ) , or @xmath14 ( channel - capacity based ) .
note that , obviously , the comparison problem is no harder than actually computing the quantitative information flow as we can compare the two numbers once we have computed @xmath15 and @xmath16 . concretely , we show the following negative results , where @xmath9 is @xmath17 , @xmath18 $ ] , @xmath19 $ ] , or @xmath13 $ ] with @xmath11 uniform . *
checking if @xmath20 is not a @xmath21-safety property @xcite for any @xmath21 .
* restricted to loop - free boolean programs , checking if @xmath20 is # p - hard .
the results are in stark contrast to non - interference which is known to be a @xmath22-safety property in general @xcite ( technically , for the termination - insensitive case ) and can be shown to be conp - complete for loop - free boolean programs ( proved in section [ sec : complex ] ) . ( # p is known to be as hard as the entire polynomial hierarchy @xcite . )
the results suggest that precisely inferring ( i.e. , computing ) quantitative information flow according to these definitions would be harder than checking non - interference and may require a very different approach ( i.e. , not self composition @xcite ) .
we also give the following positive results which show checking if the quantitative information flow of one program is larger than the other for all distributions according to the entropy - based definitions is easier .
below , @xmath23 is @xmath24 , @xmath25 , or @xmath26 . *
checking if @xmath27(m_1 ) \leq
\mathcal{y}[\mu](m_2)$ ] is a @xmath22-safety property .
* restricted to loop - free boolean programs , checking if @xmath27(m_1 ) \leq \mathcal{y}[\mu](m_2)$ ] is conp - complete .
these results are proven by showing that the problems @xmath28(m_1 ) \leq { \it se}[\mu](m_2)$ ] , @xmath29(m_1 ) \leq { \it me}[\mu](m_2)$ ] , and @xmath30(m_1 ) \leq { \it ge}[\mu](m_2)$ ] are all actually equivalent to a simple @xmath22-safety relation @xmath31 .
we also show that this relation refines the channel - capacity based quantitative information flow , that is , if @xmath31 then @xmath32 .
the fact that @xmath31 is a @xmath22-safety property implies that it can be reduced to a safety problem via self composition .
this leads to a new approach to precisely checking quantitative information flow that leverages recent advances in automated software verification @xcite .
briefly , given @xmath3 and @xmath5 , @xmath31 means that @xmath3 is at least as secure as @xmath5 for all distributions while @xmath33 means that there must be a distribution in which @xmath3 is less secure than @xmath5 , according to the entropy - based definitions of quantitative information flow .
therefore , by deciding @xmath31 , we can measure the security of the program @xmath3 relative to another _ specification _ program @xmath5 .
note that this is useful even when @xmath3 and @xmath5 are `` incomparable '' by @xmath34 , that is , when @xmath35 and @xmath36 .
see section [ sec : qbysc ] for the details .
the rest of the paper is organized as follows .
section [ sec : prelim ] reviews the existing information - theoretic definitions of quantitative information flow .
section [ sec : qifnksafe ] proves the hardness of their comparison problems and thus shows the hardness of precisely inferring quantitative information flow according to these definitions .
section [ sec : qshan ] introduces the relation @xmath34 and proves it equivalent to the comparison problems for the entropy - based definitions with their distributions universally quantified .
the section also shows that this is a @xmath22-safety property and is easier to decide than the non - universally - quantified comparison problems , and suggests a self - composition based method for precisely checking quantitative information flow .
section [ sec : related ] discusses related work , and section [ sec : concl ] concludes .
appendix [ sec : lemdefs ] contains the supporting lemmas and definitions for the proofs appearing in the main text .
the omitted proofs appear in appendix [ sec : proofs ] .
we introduce the information theoretic definitions of quantitative information flow that have been proposed in literature .
first , we review the notion of the _ shannon entropy _ @xcite , @xmath37(x)$ ] , which is the average of the information content , and intuitively , denotes the uncertainty of the random variable @xmath38 .
let @xmath38 be a random variable with sample space @xmath39 and @xmath11 be a probability distribution associated with @xmath38 ( we write @xmath11 explicitly for clarity ) .
the shannon entropy of @xmath38 is defined as @xmath40(x)=\sum_{x\in\mathbb{x } } \mu(x = x)\log\frac{1}{\mu(x = x)}\ ] ] ( the logarithm is in base 2 . ) next , we define _
conditional entropy_. informally , the conditional entropy of @xmath38 given @xmath41 denotes the uncertainty of @xmath38 after knowing @xmath41 .
let @xmath38 and @xmath41 be random variables with sample spaces @xmath39 and @xmath42 , respectively , and @xmath11 be a probability distribution associated with @xmath38 and @xmath41 .
then , the conditional entropy of @xmath38 given @xmath41 , written @xmath37(x|y)$ ] is defined as @xmath40(x|y)=\sum_{y\in\mathbb y } \mu(y = y ) \mathcal{h}[\mu](x|y = y)\ ] ] where @xmath43(x|y = y ) \\ \hspace{2em}=\sum_{x\in\mathbb x } \mu(x = x|y = y)\log\frac{1}{\mu(x = x|y = y ) } \\
\mu(x = x|y = y)=\frac{\mu(x = x , y = y)}{\mu(y = y ) } \end{array}\ ] ] next , we define ( conditional ) mutual information . intuitively , the conditional mutual information of @xmath38 and @xmath41 given @xmath44 represents the mutual dependence of @xmath38 and @xmath41 after knowing @xmath44 .
let @xmath45 and @xmath44 be random variables and @xmath11 be an associated probability distribution .
then , the conditional mutual information of @xmath38 and @xmath41 given @xmath44 is defined as @xmath46(x;y|z)&=&\mathcal{h}[\mu](x|z)-\mathcal{h}[\mu](x|y , z)\\ & = & \mathcal{h}[\mu](y|z)-\mathcal{h}[\mu](y|x , z ) \end{array}\ ] ] let @xmath8 be a program that takes a high security input @xmath1 and a low security input @xmath47 , and gives the low security output @xmath2 . for simplicity , we restrict to programs with just one variable of each kind , but it is trivial to extend the formalism to multiple variables ( e.g. , by letting the variables range over tuples ) .
also , for the purpose of the paper , unobservable ( i.e. , high security ) outputs are irrelevant , and so we assume that the only program output is the low security output .
let @xmath11 be a probability distribution over the values of @xmath1 and @xmath47 .
then , the semantics of @xmath8 can be defined by the following probability equation .
( we restrict to terminating deterministic programs in this paper . )
@xmath48 note that we write @xmath49 to denote the low security output of the program @xmath8 given inputs @xmath50 and @xmath51 .
now , we are ready to introduce the shannon - entropy based definition of quantitative information flow ( qif ) @xcite .
[ def : se ] let @xmath8 be a program with high security input @xmath1 , low security input @xmath47 , and low security output @xmath2 .
let @xmath11 be a distribution over @xmath1 and @xmath47 .
then , the shannon - entropy - based quantitative information flow is defined @xmath52(m ) & = & \mathcal{i}[\mu](o;h|l ) \\ & = & \mathcal{h}[\mu](h|l)-\mathcal{h}[\mu](h|o , l ) \end{array}\ ] ] intuitively , @xmath37(h|l)$ ] denotes the initial uncertainty knowing the low security input and @xmath37(h|o , l)$ ] denotes the remaining uncertainty after knowing the low security output . as an example , consider the programs @xmath3 and @xmath5 from section [ sec : introduction ] .
for concreteness , assume that @xmath4 is the value @xmath53 and @xmath1 ranges over the space @xmath54 .
let @xmath55 be the uniform distribution over @xmath54 , that is , @xmath56 for all @xmath57 .
the results are as follows .
@xmath58(m_1)&=\mathcal{h}[u](h)-\mathcal{h}[u](h|o)\\ & = \log 4-\frac{3}{4}\log{3}\\ & \approx .81128\\&\\ { \it se}[u](m_2)&=\mathcal{h}[u](h)-\mathcal{h}[u](h|o)\\ & = \log 4-\log 1\\ & = 2 \end{array}\ ] ] consequently , we have that @xmath59(m_1 ) \leq { \it se}[u](m_2)$ ] , but @xmath59(m_2 ) \not\leq { \it se}[u](m_1)$ ] .
that is , @xmath3 is more secure than @xmath5 ( according to the shannon - entropy based definition with uniformly distributed inputs ) , which agrees with our intuition .
let us recall the notion of non - interference @xcite .
a program @xmath8 is said to be non - interferent iff for any @xmath60 and @xmath61 , @xmath62 .
it is worth noting that non - interference can be formalized as a special case of the shannon - entropy based quantitative information flow where the flow quantity is zero .
[ thm : reqni ] let @xmath8 be a program that takes high - security input @xmath1 , low - security input @xmath47 , and returns low - security output @xmath2 . then , @xmath8 is non - interferent if and only if @xmath63(m)=0 $ ] .
the above theorem is complementary to the one proven by clark et al .
@xcite which states that for any @xmath11 such that @xmath64 for all @xmath65 and @xmath66 , @xmath18(m)=0 $ ] iff @xmath8 is non - interferent .
next , we introduce the _ min entropy _ , which smith @xcite recently suggested as an alternative measure for quantitative information flow .
let @xmath38 and @xmath41 be random variables , and @xmath11 be an associated probability distribution .
then , the min entropy of @xmath38 is defined @xmath67(x)=\log\frac{1}{\mathcal{v}[\mu](x)}\ ] ] and the conditional min entropy of @xmath38 given @xmath41 is defined @xmath67(x|y)=\log\frac{1}{\mathcal{v}[\mu](x|y)}\ ] ] where @xmath68(x)&=\max_{x\in\mathbb x } \mu(x = x)\\ \mathcal{v}[\mu](x|y = y)&=\max_{x\in\mathbb x } \mu(x = x|y = y)\\ \mathcal{v}[\mu](x|y)&=\sum_{y\in\mathbb y } \mu(y = y ) \mathcal{v}[\mu](x|y = y ) \end{array}\ ] ] intuitively , @xmath69(x)$ ] represents the highest probability that an attacker guesses @xmath38 in a single try .
we now define the min - entropy - based definition of quantitative information flow .
[ def : me ] let @xmath8 be a program with high security input @xmath1 , low security input @xmath47 , and low security output @xmath2 .
let @xmath11 be a distribution over @xmath1 and @xmath47 .
then , the min - entropy - based quantitative information flow is defined @xmath70(m)=\mathcal{h}_\infty[\mu](h|l)-\mathcal{h}_\infty[\mu](h|o , l)\ ] ] whereas smith @xcite focused on programs lacking low security inputs , we extend the definition to programs with low security inputs in the definition above .
it is easy to see that our definition coincides with smith s for programs without low security inputs .
also , the extension is arguably natural in the sense that we simply take the conditional entropy with respect to the distribution over the low security inputs .
computing the min - entropy based quantitative information flow for our running example programs @xmath3 and @xmath5 from section [ sec : introduction ] with the uniform distribution , we obtain , @xmath71(m_1)&=\mathcal{h}_\infty[u](h)-\mathcal{h}_\infty[u](h|o)\\ & = \log 4-\log 2\\ & = 1\\&\\ { \it me}[u](m_2)&=\mathcal{h}_\infty[u](h)-\mathcal{h}_\infty[u](h|o)\\ & = \log 4 -\log 1\\ & = 2 \end{array}\ ] ] again , we have that @xmath72(m_1 ) \leq { \it me}[u](m_2)$ ] and @xmath73(m_2 ) \not\leq { \it me}[u](m_1)$ ] , and so @xmath5 is deemed less secure than @xmath3 .
the third definition of quantitative information flow treated in this paper is the one based on the guessing entropy @xcite , that is also recently proposed in literature @xcite .
let @xmath38 and @xmath41 be random variables , and @xmath11 be an associated probability distribution .
then , the guessing entropy of @xmath38 is defined @xmath74(x)=\sum_{1\le i\le m}i\times\mu(x = x_i)\ ] ] where @xmath75 and @xmath76 .
the conditional guessing entropy of @xmath38 given @xmath41 is defined @xmath74(x|y)=\sum_{y\in{\mathbb y}}\mu(y = y)\sum_{1\le i\le m}i\times\mu(x = x_i|y = y)\ ] ] where @xmath75 and @xmath77 .
intuitively , @xmath78(x)$ ] represents the average number of times required for the attacker to guess the value of @xmath38 .
we now define the guessing - entropy - based quantitative information flow .
[ def : ge ]
let @xmath8 be a program with high security input @xmath1 , low security input @xmath47 , and low security output @xmath2 .
let @xmath11 be a distribution over @xmath1 and @xmath47 .
then , the guessing - entropy - based quantitative information flow is defined @xmath79(m)=\mathcal{g}[\mu](h|l)-\mathcal{g}[\mu](h|o , l)\ ] ] like with the min - entropy - based definition , the previous research on guessing - entropy - based quantitative information flow only considered programs without low security inputs @xcite .
but , it is easy to see that our definition with low security inputs coincides with the previous definitions for programs without low security inputs .
also , as with the extension for the min - entropy - based definition , it simply takes the conditional entropy over the low security inputs .
we test _ ge _ on the running example from section [ sec : introduction ] by calculating the quantities for the programs @xmath3 and @xmath5 with the uniform distribution .
@xmath80(m_1 ) & = \mathcal{g}[u](h)-\mathcal{g}[u](h|o)\\ & = \frac{5}{2 } - \frac{7}{4}\\ & = 0.75\\\\ { \it ge}[u](m_2 ) & = \mathcal{g}[u](h)-\mathcal{g}[u](h|o)\\ & = \frac{5}{2 } - 1\\ & = 1.5 \end{array}\ ] ] therefore , we again have that @xmath81(m_1 ) \leq { \it ge}[u](m_2)$ ] and @xmath81(m_2 ) \not\leq { \it ge}[u](m_1)$ ] , and so @xmath5 is considered less secure than @xmath3 , even with the guessing - entropy based definition with the uniform distribution .
the fourth and the final existing definition of quantitative information flow that we introduce in this paper is the one based on _ channel capacity _ @xcite , which is simply defined to be the maximum of the shannon - entropy based quantitative information flow over the distribution . let @xmath8 be a program with high security input @xmath1 , low security input @xmath47 , and low security output @xmath2 .
then , the channel - capacity - based quantitative information flow is defined @xmath82(o;h|l)\ ] ] unlike the shannon - entropy based , the min - entropy based , and the guessing - entropy based definitions , the channel - capacity based definition of quantitative information flow is not parameterized by a distribution over the inputs . as with the other definitions ,
let us test the definition on the running example from section [ sec : introduction ] by calculating the quantities for the programs @xmath3 and @xmath5 : @xmath83(o;h)\\ & = 1\\&\\ { \it cc}(m_2)&=\max_\mu\mathcal{i}[\mu](o;h)\\ & = 2 \end{array}\ ] ] as with the entropy - based definitions ( with the uniform distribution ) , we have that @xmath84 and @xmath85 , that is , the channel - capacity based quantitative information flow also says that @xmath5 is less secure than @xmath3 .
we investigate the hardness of deciding the following _ comparison problem _
@xmath86 $ ] : given programs @xmath3 and @xmath5 having the same input domain , decide if @xmath18(m_1 ) \leq { \it se}[\mu](m_2)$ ] . because we are interested in hardness , we focus on the case where @xmath11 is the uniform distribution @xmath55 . that is , the results we prove for the specific case applies to the general case .
also note that the comparison problem is no harder than actually computing the quantitative information flow because we can compare @xmath18(m_1)$ ] and @xmath18(m_2)$ ] if we know their actual values .
likewise , we study the hardness of the comparison problem @xmath87 $ ] , defined to be the problem @xmath19(m_1 ) \leq { \it me}[\mu](m_2)$ ] , @xmath88 $ ] , defined to be the problem @xmath13(m_1 ) \leq { \it ge}[\mu](m_2)$ ] , and @xmath89 , defined to be the problem @xmath84 . as with @xmath90
$ ] , we require the two programs to share the same input domain for these problems . we show that none of these comparison problems are @xmath21-safety problems for any @xmath21 .
informally , a program property is said to be a _
property @xcite if it can be refuted by observing @xmath21 number of ( finite ) execution traces .
a @xmath21-safety problem is the problem of checking a @xmath21-safety property .
note that the standard safety property is a @xmath91-safety property .
an important property of a @xmath21-safety problem is that it can be reduced to a standard safety ( i.e. , @xmath91-safety ) problem , such as the unreachability problem , via a simple program transformation called _ self composition _ @xcite .
it is well - known that non - interference is a @xmath22-safety property ,- safety property @xcite . ] and this has enabled its precise checking via a reduction to a safety problem via self composition and piggybacking on advances in automated safety verification methods @xcite .
unfortunately , the results in this section imply that quantitative information flow inference problem is unlikely to receive the same benefits . because we are concerned with properties about pairs of programs ( i.e. , comparison problems ) ,
we extend the notion of @xmath21-safety to properties refutable by observing @xmath21 traces from each of the two programs .
more formally , we say that the comparison problem @xmath92 is a @xmath21-safety property if @xmath93 implies that there exists @xmath94\hspace*{-1.2pt}]}$ ] and @xmath95\hspace*{-1.2pt}]}$ ] such that * @xmath96 * @xmath97 * @xmath98\hspace*{-1.2pt } ] } \wedge t_2 \subseteq { [ \hspace*{-1.2pt}[m_2']\hspace*{-1.2pt } ] } \rightarrow ( m_1 ' , m_2 ' ) \not\in c$ ] in the above , @xmath99\hspace*{-1.2pt}]}$ ] denotes the semantics ( i.e. , traces ) of @xmath8 , represented by the set of input / output pairs @xmath100 .
we now state the main results of the section .
( recall that @xmath55 denotes the uniform distribution . )
we sketch the main idea of the proofs .
all proofs are by contradiction .
let @xmath92 be the comparison problem in the statement and suppose @xmath92 is @xmath21-safety .
let @xmath101 .
then , we have @xmath94\hspace*{-1.2pt}]}$ ] and @xmath102\hspace*{-1.2pt}]}$ ] satisfying the properties ( 1 ) , ( 2 ) , and ( 3 ) above . from this , we construct @xmath103 and @xmath104 such that @xmath105\hspace*{-1.2pt}]}$ ] and @xmath106\hspace*{-1.2pt}]}$ ] and @xmath107 to obtain the contradiction . @xmath108 $ ] is not a @xmath21-safety property for any @xmath109 .
[ thm : seks ] @xmath110 $ ] is not a @xmath21-safety property for any @xmath109 .
[ thm : meks ] @xmath111 $ ] is not a @xmath21-safety property for any @xmath112 .
[ thm : geks ] @xmath89 is not a @xmath21-safety property for any @xmath109 .
[ thm : ccks ] the notion of @xmath21-safety property , like the notion of safety property from where it extends , is defined over all programs regardless of their size .
( for example , non - interference is a @xmath22-safety property for all programs and unreachability is a safety property for all programs . ) but , it is easy to show that the comparison problems would become `` @xmath21-safety '' properties if we constrained and bounded the input domains because then the size of the semantics ( i.e. , the input / output pairs ) of such programs would be bounded by @xmath113 . in this case , the problems are at most @xmath113-safety .
however , these bounds are high for all but very small domains , and are unlikely to lead to a practical verification method .
we discuss the details of the proof of theorem [ thm : seks ] .
the proofs of theorems [ thm : meks ] , [ thm : geks ] , [ thm : ccks ] are deferred to appendix [ sec : proofs ] . for contradiction
, suppose @xmath108 $ ] is a @xmath21-safety property .
let @xmath8 and @xmath114 be programs having the same input domain such that @xmath115 $ ] .
then , it must be the case that there exist @xmath116\hspace*{-1.2pt}]}$ ] and @xmath117\hspace*{-1.2pt}]}$ ] such that @xmath118 , @xmath119 , and @xmath120\hspace*{-1.2pt}]}\wedge t ' \subseteq { [ \hspace*{-1.2pt}[m_c']\hspace*{-1.2pt } ] } \rightarrow ( m_c , m_c ' )
\not\in c_{\it se}[u]$ ] .
let @xmath121 where @xmath122 .
now , we construct new programs @xmath123 and @xmath124 as follows .
@xmath125 where * @xmath126 , * @xmath127 , * @xmath128 , @xmath129 , @xmath130 , @xmath131 , and @xmath132 are distinct , * @xmath133 , * @xmath134 , and * @xmath135 . then ,
comparing the shannon - entropy - based quantitative information flow of @xmath123 and @xmath124 , we have , @xmath136(\bar{m'})-{\it se}[u](\bar{m})\\ \hspace{2em}=\sum_{o_x'\in{\{{o_1',\dots , o_i'}\}}}u(o_x')\log\frac{1}{u(o_x')}\\ \hspace{3em}+u(o')\log\frac{1}{u(o')}+u(o_r')\log\frac{1}{u(o_r')}\\ \hspace{4em}-(\sum_{o_x\in{\{{o_1,\dots , o_j}\}}}u(o_x)\log\frac{1}{u(o_x)}\\ \hspace{5em}+\sum_{o_y\in{\{{o_{j+1},\dots , o_{j+i}}\}}}u(o_y)\log\frac{1}{u(o_y)}\\ \hspace{6em}+u(o_r)\log\frac{1}{u(o_r)})\\ \end{array}\ ] ] ( note the abbreviations from appendix [ sec : lemdefs ] . ) by lemma [ lem : a7 ] , we have @xmath137 and @xmath138 trivially , we have @xmath139 as a result , we have @xmath140(\bar{m'})-{\it se}[u](\bar{m})\ge 0\ ] ] note that @xmath123 and @xmath124 have the same counterexamples @xmath141 and @xmath142 , that is , @xmath143\hspace*{-1.2pt}]}$ ] and @xmath144\hspace*{-1.2pt}]}$ ] . however , we have @xmath145 $ ] .
this leads to a contradiction .
@xmath146 @xmath147\\ { \textit{wp}({{\sf if}\;\psi\;{\sf then}\;m_0\;{\sf else}\;m_1 } , { \phi})}\\ \qquad=(\psi\rightarrow { \textit{wp}({m_0 } , { \phi})})\wedge(\neg \psi\rightarrow { \textit{wp}({m_1 } , { \phi})})\\ { \textit{wp}({m_0;m_1 } , { \phi})}={\textit{wp}({m_0 } , { { \textit{wp}({m_1 } , { \phi } ) } } ) } \end{array}\ ] ] the purpose of this section is to show a complexity theoretic gap between non - interference and quantitative information flow .
the results strengthen the hypothesis that quantitative information flow is quite hard to compute precisely , and also suggest an interesting connection to counting problems .
we focus on loop - free boolean programs whose syntax is given in figure [ syntax ] .
we assume the usual derived formulas @xmath148 , @xmath149 , @xmath150 , and @xmath151 . we give the usual weakest precondition semantics in figure [ wpsemantics ] .
to adapt the information flow framework to boolean programs , we make each information flow variable @xmath1 , @xmath47 , and @xmath2 range over functions mapping boolean variables of its kind to boolean values .
so , for example , if @xmath152 and @xmath153 are low security boolean variables and @xmath154 is a high security boolean variable , then @xmath47 ranges over the functions @xmath155 , and @xmath1 and @xmath2 range over @xmath156 .
( every boolean variable is either a low security boolean variable or a high security boolean variable . )
we write @xmath157 for an input @xmath158 and an output @xmath159 if @xmath160 for a boolean formula @xmath161 such that @xmath162 and @xmath163 for all output @xmath164 . here
, @xmath165 is the usual logical satisfaction relation , using @xmath166 , etc . to look up the values of the boolean variables .
( note that this incurs two levels of lookup . ) as an example , consider the following program .
@xmath167 let @xmath152 , @xmath153 and @xmath168 be high security variables and @xmath154 be a low security variable . then , @xmath169(m ) & = & 1.5\\ { \it me}[u](m ) & = & \log 3\\ & \approx & 1.5849625\\ { \it ge}[u](m)&= & 1.25\\ { \it cc}(m ) & = & \log 3\\ & \approx & 1.5849625 \\ \end{array}\ ] ] we prove the following hardness results .
these results are proven by a reduction from # sat , which is the problem of counting the number of solutions to a quantifier - free boolean formula .
# sat is known to be # p - complete . because # sat is a function problem and the comparison problems are decision problems , a step in the proofs makes binary search queries to the comparison problem oracle a polynomial number of times .
( recall that the notation @xmath170 means the complexity class of function problems solvable in polynomial time with an oracle for the problem @xmath171 . )
[ thm : secomp ] @xmath172}$ ] [ thm : mecomp ] @xmath173}$ ] [ thm : gecomp ] @xmath174}$ ] [ thm : cccomp ]
@xmath175 we remind that the above results apply ( even ) when the comparison problems @xmath108 $ ] , @xmath110 $ ] , @xmath111 $ ] , and @xmath89 are restricted to loop - free boolean programs . in summary , each comparison problem @xmath108 $ ] , @xmath110 $ ] , @xmath111 $ ] , and @xmath89 can be used a polynomial number of times to solve a # p - complete problem . because toda s theorem @xcite implies that the entire polynomial hierarchy can be solved by using a # p - complete oracle a polynomial number of times ,
our results show that the comparison problems for quantitative information flow can also be used a polynomial number of times to solve the entire polynomial hierarchy , for the case of loop - free boolean programs .
as shown below , this presents a gap from non - interference , which is only conp - complete for loop - free boolean programs .
[ thm : nicomp ] checking non - interference is conp - complete for loop - free boolean programs .
the above is an instance of the general observation that , by solving quantitative information flow problems , one is able to solve the class of problems known as _ counting problems _ , which coincides with # sat for the case of loop - free boolean programs .
we discuss the details of the proof of theorem [ thm : secomp ] .
the proofs of theorems [ thm : mecomp ] , [ thm : gecomp ] , [ thm : cccomp ] are deferred to appendix [ sec : proofs ] .
first , we prove the following lemma which states that we can compare the number of solutions to boolean formulas by computing @xmath59 $ ] .
( for convenience , we use large letters @xmath1 , @xmath47 , @xmath2 , etc . to range over boolean variables as well as generic random variables . )
[ lem : a3 ] let @xmath176 and @xmath177 be distinct boolean random variables .
let @xmath178 and @xmath179 be any non - negative integers such that @xmath180 and @xmath181 .
let @xmath182 ( resp .
@xmath183 ) be a formula over @xmath176 having @xmath178 ( resp .
@xmath179 ) assignments
. then , @xmath184 iff @xmath59(m_j)\le{\it se}[u](m_i)$ ] where @xmath185 and @xmath186 .
let @xmath187 and @xmath188 .
we have @xmath189(m_j ) & = \frac{j}{2^{|h|+1}}\log\frac{2^{|h|+1}}{j } + \frac{2^{|h|+1}-j}{2^{|h|+1}}\log\frac{2^{|h|+1}}{2^{|h|+1}-j}\\ \hspace{5.3em } & = p\log p + ( 1-p)\log \frac{1}{1-p}\\ { \it se}[u](m_i ) & = \frac{i}{2^{|h|+1}}\log\frac{2^{|h|+1}}{i } + \frac{2^{|h|+1}-i}{2^{|h|+1}}\log\frac{2^{|h|+1}}{2^{|h|+1}-i}\\ \hspace{5.3em } & = q\log q + ( 1-q)\log \frac{1}{1-q } \end{array}\ ] ] * * only if * + suppose @xmath184 . then
, @xmath136(m_i)-{\it se}[u](m_j ) \\ \hspace{3em } = p\log\frac{1}{p }
+ ( 1-p)\log\frac{1}{1-p}\\ \hspace{4em } - q\log\frac{1}{q } - ( 1-q)\log\frac{1}{1-q}\\ \hspace{3em } = \log(\frac{1-p}{p})^p \frac{1-q}{1-p } ( \frac{q}{1-q})^q \end{array}\ ] ] then , from @xmath190 and @xmath191 , we have @xmath58(m_i)-{\it se}[u](m_j ) & \geq \log(\frac{1-p}{p})^p ( \frac{q}{1-q})^q\\ & \ge\log(\frac{1-p}{p})^q ( \frac{q}{1-q})^q\\ & = \log(\frac{(1-p)q}{p(1-q)})^q\\ & = \log(\frac{q - pq}{p - pq})^q\\ & = \log(\frac{pq - q}{pq - p})^q\\ & = \log(\frac{1-\frac{1}{p}}{1-\frac{1}{q}})^q\\ & \geq 0 \end{array}\ ] ] the last line follows from @xmath192 . * * if * + we prove the contraposition .
suppose @xmath193 . then , @xmath194(m_j ) - { \it
se}[u](m_i ) \\ \qquad = q\log\frac{1}{q } + ( 1-q)\log\frac{1}{1-q } \\ \qquad\qquad- p\log\frac{1}{p } - ( 1-p)\log\frac{1}{1-p } \\ \qquad > 0 \end{array}\ ] ]
the last line follows from the fact that @xmath195 .
therefore , @xmath59(m_j ) \not\leq { \it se}[u](m_i)$ ] . then , using lemma [ lem : a3 ] , we prove the following lemma which is crucial to proving theorem [ thm : secomp ] . [
lem : a4 ] let @xmath176 be distinct variables and @xmath161 be a boolean formula over @xmath176 . then , the number of assignments for @xmath161 can be computed by executing an oracle that decides whether programs are in @xmath108 $ ] at most @xmath196 times .
first , we define a procedure that returns the number of solutions of @xmath161 .
let @xmath197 where @xmath198 is a formula over @xmath176 having @xmath179 assignments and @xmath177 be a boolean variable such that @xmath199 .
note that , by lemma [ lem : a2 ] , such @xmath198 can be generated in linear time .
then , we invoke the following procedure where @xmath200 .
@xmath201(f(n),m ' ) \vee \neg c_{\it se}[u](m',f(n))\\ \qquad{\sf if}\;c_{\it se}[u](f(n),m')\\ \qquad\qquad{\sf then}\;\{\ell = n;n=(\ell+r)/2;\}\\ \qquad\qquad{\sf else}\;\{r = n;n=(\ell+r)/2;\}\\ { \sf return}\;n \end{array}\ ] ] note that when the procedure terminates , we have @xmath59(f(n ) ) = { \it se}[u](m')$ ] , and so by lemma [ lem : a3 ] , @xmath202 is the number of satisfying assignments to @xmath161 .
we show that the procedure iterates at most @xmath203 times . to see this , every iteration in the procedure
narrows the range between @xmath204 and @xmath51 by one half .
because @xmath205 is bounded by @xmath206 , it follows that the procedure iterates at most @xmath203 times .
hence , the oracle @xmath108 $ ] is accessed @xmath196 times , and this proves the lemma .
finally , theorem [ thm : secomp ] follows from lemma [ lem : a4 ] and the fact that # sat , the problem of counting the number of solutions to a boolean formula , is # p - complete .
as proved in section [ sec : qifnksafe ] , precisely computing quantitative information flow is quite difficult . indeed
, we have shown that even just comparing two programs on which has the larger flow is difficult ( i.e. , @xmath207 , @xmath208 , @xmath209 , and @xmath89 ) . in this section ,
we show that universally quantifying the shannon - entropy based comparison problem @xmath86 $ ] , the min - entropy based problem @xmath210 $ ] , or the guessing - entropy based problem @xmath88 $ ] over the distribution @xmath11 is equivalent to a simple relation @xmath34 enjoying the following properties .
* @xmath34 is a @xmath22-safety property .
* @xmath34 is conp - complete for loop - free boolean programs .
note that ( 1 ) implies that we can actually check if @xmath211 $ ] for all @xmath11 via self composition ( and likewise for @xmath210 $ ] and @xmath88 $ ] ) .
we actually show in section [ sec : qbysc ] that we can even use the security - type - based approach suggested by terauchi and aiken @xcite to minimize code duplication during self composition ( i.e. , do _ interleaved _ self composition ) . we remind that except for the conp - completeness result ( theorem [ thm : rcomp ] ) , the results in this section apply to any ( deterministic and terminating ) programs and not just to loop - free boolean programs .
[ def : r ] we define @xmath34 to be the relation such that @xmath212 iff for all
@xmath66 and @xmath213 , if @xmath214 then @xmath215 .
note that @xmath31 essentially says that if an attacker can distinguish a pair of high security inputs by executing @xmath3 , then she could do the same by executing @xmath5 .
hence , @xmath34 naturally expresses that @xmath3 is at least as secure as @xmath5 .
have appeared in literature ( often in somewhat different representations ) @xcite . in particular ,
clark et al .
@xcite have shown a result analogous to the @xmath216 direction of theorem [ thm : reqse ] below .
but , @xmath34 s properties have not been fully investigated . ]
it may be somewhat surprising that this simple relation is actually equivalent to the rather complex entropy - based quantitative information flow definitions when they are cast as comparison problems and the distributions are universally quantified , as stated in the following theorems .
first , we show that @xmath34 coincides exactly with @xmath207 with its distribution universally quantified .
[ thm : reqse ] @xmath217(m_1,m_2)}\}}$ ] the proof is detailed in section [ sec : reqseproof ] .
the next two theorems show that @xmath34 also coincides with @xmath208 and @xmath218 with their distribution universally quantified .
[ thm : reqme ] @xmath219(m_1,m_2)}\}}$ ] [ thm : reqge ] @xmath220(m_1,m_2)}\}}$ ] the first half of the @xmath216 direction of the proofs for the theorems above is much like the that of theorem [ thm : reqse ] , that is , it makes the observation that @xmath5 disambiguates the high security inputs at least as fine as does @xmath3 .
then , the proof concludes by utilizing the particular mathematical properties relevant to the respective definitions .
the proof for the @xmath221 direction is also similar to the argument used in theorem [ thm : reqse ] .
the details of the proofs appear in appendix [ sec : proofs ] .
next , we show that @xmath34 refines @xmath89 in the sense that if @xmath212 then @xmath222 .
[ thm : rimpcc ] @xmath223 note that , the other direction , @xmath224 , does not hold as @xmath34 is not always a total order , whereas @xmath89 is .
we also show that @xmath34 is compatible with the notion of non - interference .
[ thm : rni ] let @xmath5 be a non - interferent program . then
, @xmath212 iff @xmath3 is also non - interferent and @xmath3 has the same input domain as @xmath5 .
next , we show that @xmath34 is easier to decide than the non - universally - quantified versions of the comparison problems .
first , it is trivial to see from definition [ def : r ] that @xmath34 is a @xmath22-safety property .
[ thm : r2safe ] @xmath34 is a @xmath22-safety property .
it can be shown that , restricted to loop - free boolean programs , @xmath34 is conp - complete .
this follows directly from the observation that we can decide @xmath34 by self composition thanks to its @xmath22-safety property and the fact that , for loop - free boolean programs , self composition reduces the problem to an unsat instance .. ] [ thm : rcomp ] restricted to loop - free boolean programs , @xmath34 is conp - complete .
we discuss the details of the proof of theorem [ thm : reqse ] .
the proofs of theorems [ thm : reqme ] , [ thm : reqge ] , [ thm : rimpcc ] are deferred to appendix [ sec : proofs ] .
first , we prove the following lemma which says that , if @xmath225 then @xmath59(m')$ ] is at least as large as @xmath59(m)$ ] per each low security input @xmath66 .
[ lem : a8 ] suppose @xmath225 , that is , for all @xmath226 , @xmath227 in @xmath228 and @xmath51 in @xmath229 , @xmath230 .
let @xmath231 be the set of the outputs of @xmath8 , and @xmath232 be the set of the outputs of @xmath114 . then , for any @xmath51 , we have @xmath233 .
( recall the notational convention from definition [ def : distabrv ] . )
first , we prove for any output @xmath159 of @xmath8 , there exist corresponding outputs @xmath234 of @xmath114 such that @xmath235 let @xmath236 be the set such that @xmath237 .
let @xmath238 .
let @xmath239, and , @xmath240 .
for any @xmath241 such that @xmath242 and @xmath243 , we have @xmath244 since @xmath225 .
then , we have @xmath245 . by lemma [ lem : a7 ]
, we have @xmath246 now to prove the lemma , it suffices to show that each @xmath247 constructed above are disjoint .
that is , for @xmath248 and @xmath249 outputs of @xmath8 such that @xmath250 , @xmath251 . for contradiction ,
suppose @xmath252 .
then , there exist @xmath226 and @xmath227 such that @xmath253 , @xmath254 , @xmath255 , and @xmath256 . since @xmath225
, we have @xmath257 , and it leads to a contradiction .
hence , we have @xmath258 we now prove theorem [ thm : reqse ] . *
@xmath216 + suppose @xmath259 . by lemma [ lem : detse ] , @xmath260(m)&=\mathcal h[\mu](o|l)\\ & = \sum_\ell\sum_o \mu(o,\ell)\log\frac{\mu(\ell)}{\mu(o,\ell)}\\ \end{array}\ ] ] and @xmath260(m')&=\mathcal h[\mu](o'|l)\\ &
= \sum_\ell\sum_{o ' } \mu(o',\ell)\log\frac{\mu(\ell)}{\mu(o',\ell)}\\ \end{array}\ ] ] by lemma [ lem : a8 ] and the fact that @xmath259 , we obtain for any @xmath51 @xmath261 hence , @xmath262 * @xmath221 + we prove the contraposition .
suppose @xmath263 .
then , there exist @xmath264 such that @xmath265 and @xmath266 .
pick a probability function @xmath11 such that @xmath267 .
+ then , we have @xmath268(o'|l)&=\sum_\ell\sum_o \mu(o,\ell)\log\frac{\mu(\ell)}{\mu(o,\ell)}\\ & = \mu(o',\ell')\log\frac{\mu(\ell')}{\mu(o',\ell')}\\ & = 1\log\frac{1}{1}\\ & = 0 \end{array}\ ] ] let @xmath269 and @xmath248 be output variables such that @xmath270 , @xmath271 , and @xmath272 .
@xmath273(o|l)&=\sum_{o\in{\{{o_0,o_1}\ } } } \mu(o,\ell')\log\frac{\mu(\ell')}{\mu(o,\ell')}\\ & = \frac{1}{2}\log\frac{1}{\frac{1}{2}}+\frac{1}{2}\log\frac{1}{\frac{1}{2}}\\ & = 1\\ \end{array}\ ] ] therefore , @xmath18(m)\not\le{\it se}[\mu](m')$ ] , that is , @xmath274}\}}$ ] .
theorems [ thm : reqse ] , [ thm : reqme ] , [ thm : reqge ] , and [ thm : r2safe ] imply that we can check if the entropy - based quantitative information flow of a program ( i.e. , _ se _ , _ me _ , and _ ge _ ) is bounded by that of another for all distributions via self composition @xcite .
this suggests a novel approach to precisely checking quantitative information flow .
that is , given a _
program @xmath3 , the user would construct a _ specification _
program @xmath5 with the same input domain as @xmath3 having the desired level of security .
then , she would check @xmath212 via self composition .
if so , then @xmath3 is guaranteed to be at least as secure as @xmath5 according to the shannon - entropy based , the min - entropy based , and the guessing - entropy based definition of quantitative information flow for all distributions ( and also channel - capacity based definition ) , and otherwise , there must be a distribution in which @xmath3 is less secure than @xmath5 according to the entropy - based definitions .
note that deciding @xmath31 is useful even when @xmath3 and @xmath5 are @xmath34-incomparable , that is , when neither @xmath212 nor @xmath275 .
this is because @xmath276 implies that @xmath3 is less secure than @xmath5 on some distribution .
for example , suppose @xmath3 is some complex login program with the high security input @xmath1 and the low security input @xmath47 .
and we would like to verify that @xmath3 is at least as secure as the prototypical login program @xmath5 below .
@xmath277 then , using this framework , it suffices to just query if @xmath212 is true .
( note that the output domains of @xmath3 and @xmath5 need not to match . )
we now describe how to actually check @xmath212 via self composition . from @xmath3 and @xmath5
, we construct the self - composed program @xmath114 shown below .
@xmath278 note that @xmath212 is true iff @xmath114 does not cause an assertion failure .
the latter can be checked via a software safety verifier such as slam and blast @xcite . as an aside
, we note that this kind of construction could be easily generalized to reduce any @xmath21-safety problem ( cf . section [ sec : qifnksafe ] ) to a safety problem , as shown by clarkson and schneider @xcite .
note that the line @xmath279 ( resp .
@xmath280 ) of the pseudo code above is @xmath3 ( resp .
@xmath5 ) sequentially composed with a copy of itself , which is from where the name `` self composition '' comes .
therefore , technically , @xmath114 is a composition of two self compositions .
@xmath279 ( and @xmath280 ) are actually exactly the original self composition proposed for non - interference @xcite .
terauchi and aiken @xcite noted that only the parts of @xmath3 ( and @xmath5 ) that depend on the high security inputs @xmath1 and @xmath177 need to be duplicated and self composed , with the rest of the program left intact and `` interleaved '' with the self - composed parts .
the resulting program tends to be verified easier than the naive self composition by modern software safety verifiers .
they proposed a set of transformation rules that translates a while program annotated with security types @xcite ( or dependency analysis results ) to an interleaved self - composed program .
this was subsequently improved by a number of researchers to support a richer set of language features and transformation patterns @xcite .
these transformation methods can be used in place of the naive self compositions at @xmath281 and @xmath280 in building @xmath114 .
that is , we apply a security type inference ( or a dependency analysis ) to @xmath3 and @xmath5 to infer program parts that depend on the high security inputs @xmath1 and @xmath177 so as to only duplicate and self compose those parts of @xmath3 and @xmath5 .
we recall the ideal login program below . @xmath282
we check the following four programs using the above as the specification .
@xmath283 here , @xmath1 and @xmath47 are 64-bit values , @xmath284 is the bit - wise and operator , and @xmath285 is the left shift operator . @xmath3 leaks the entire password .
@xmath5 checks the password against the user guess but then leaks the first bit when the check fails .
@xmath286 only checks the first 32 bits of the password . and , @xmath287 implements password checking correctly via a while loop .
we verify that only @xmath287 satisfies the specification , that is , @xmath288 . to see that @xmath289 , note that for any @xmath290 such that @xmath291 , @xmath292 and @xmath293 , we have that @xmath294 but @xmath295 . to see that @xmath296 , note that for @xmath290 such that @xmath297 , @xmath292 , @xmath298 and @xmath299 , we have that @xmath300 but @xmath295 .
to see that @xmath301 , let @xmath290 be such that @xmath302 , @xmath303 , and @xmath291 , then , @xmath304 but @xmath305 . and
@xmath306 , that is , @xmath5 and @xmath286 are @xmath34-incomparable with @xmath307 . ]
( here , @xmath308 denotes @xmath309 , i.e. , the first 32 bits of @xmath152 . )
the results imply that for @xmath3 , @xmath5 , and @xmath286 , there must be a distribution where the program is less secure than @xmath307 according to each of the entropy - based definition of quantitative information flow .
for instance , for the shannon - entropy based definition , we have for the uniform distribution @xmath55 , @xmath58(m_{\it spec})&=\frac{1}{2^{58 } } + \frac{2^{64}-1}{2^{64}}\log\frac{2^{64}}{2^{64}-1}\\ & \approx 3.46944695 \times 10^{-18}\\ { \it se}[u](m_1)&=64\\ { \it se}[u](m_2)&= \frac{1}{2 } + \frac{1 + 2^{63}}{2^{65}}\log\frac{2^{64}}{1 + 2^{63}}+\frac{2^{63}-1}{2^{65}}\log\frac{2^{64}}{2^{63}-1}\\ & \approx 1.0\\ { \it se}[u](m_3)&=\frac{1}{2^{27}}+\frac{2^{64}-2^{32}}{2^{64}}\log\frac{2^{64}}{2^{64}-2^{32}}\\ & \approx 7.78648\times10^{-9}\\ \end{array}\ ] ] that is , @xmath59(m_1 ) \not\leq { \it se}[u](m_{\it spec})$ ] , @xmath310(m_2 ) \not\leq { \it se}[u](m_{\it spec})$ ] , and @xmath310(m_3 ) \not\leq { \it se}[u](m_{\it spec})$ ] .
finally , we have that @xmath288 , and so @xmath287 is at least as secure as @xmath307 according to all of the definitions of quantitative information flow considered in this paper .
in fact , it can be also shown that @xmath311 .
( however , note that @xmath287 and @xmath307 are not semantically equivalent , i.e. , their outputs are reversed . )
this work builds on previous work that proposed information theoretic notions of quantitative information flow @xcite .
the previous research has mostly focused on information theoretic properties of the definitions and proposed approximate ( i.e. , incomplete and/or unsound ) methods for checking and inferring them . in contrast , this paper investigates the verification theoretic and complexity theoretic hardness of precisely inferring quantitative information flow according to the definitions and also proposes a precise method for checking quantitative information flow .
our method checks the quantitative information flow of a program against that of a specification program having the desired level of security via self composition for all distributions according to the entropy - based definitions .
it is quite interesting that the relation @xmath34 unifies the different proposals for the definition of quantitative information flow when they are cast as comparison problems and their distributions are universally quantified . as remarked in section [ sec : qshan ] , @xmath34 naturally expresses the fact that one program is more secure than the other , and it could be argued that it is the essence of quantitative information flow .
researchers have also proposed definitions of quantitative information flow that do not fit the models studied in this paper .
these include the definition based on the notion of _ belief _ @xcite , and the ones that take the maximum over the low security inputs @xcite .
refines these notions in the same sense as theorem [ thm : rimpcc ] , but the other direction is not guaranteed to hold . ] despite the staggering complexity made apparent in this paper , recent attempts have been made to ( more ) precisely infer quantitative information flow ( without universally quantifying over the distribution as in our approach ) .
these methods are based on the idea of _ counting_. as remarked in section [ sec : complex ] , quantitative information flow is closely related to counting problems , and several attempts have been made to reduce quantitative information flow problems to them .
for instance , newsome et al .
@xcite reduce the inference problem to the # sat problem and apply off - the - shelf # sat solvers . to achieve scalability , they sacrifice both soundness and completeness by only computing information flow from one execution path .
backes et al .
@xcite also propose a counting - based approach that involves self composition .
however , unlike our method , they use self composition repeatedly to find a new solution ( i.e. , more than a bounded number of times ) , and so their results do not contradict the negative results of this paper .
we have investigated the hardness and possibilities of precisely checking and inferring quantitative information flow according to the various definitions proposed in literature .
specifically , we have considered the definitions based on the shannon entropy , the min entropy , the guessing entropy , and channel capacity .
we have shown that comparing two programs on which has the larger flow according to these definitions is not a @xmath21-safety problem for any @xmath21 , and therefore that it is not possible to reduce the problem to a safety problem via self composition .
the result is in contrast to non - interference which is a @xmath22-safety problem .
we have also shown a complexity theoretic gap with non - interference by proving the # p - hardness of the comparison problems and conp - completeness of non - interference , when restricted to loop - free boolean programs .
we have also shown a positive result that checking if the entropy - based quantitative information flow of one program is larger than that of another for all distributions is a @xmath22-safety problem , and that it is also conp - complete when restricted to loop - free boolean programs .
we have done this by proving a surprising result that universally quantifying the distribution in the comparison problem for the entropy - based definitions is equivalent to a simple @xmath22-safety relation .
motivated by the result , we have proposed a novel approach to precisely checking quantitative information flow that reduces the problem to a safety problem via self composition .
our method checks the quantitative information flow of a program for all distributions against that of a specification program having the desired level of security .
we would like to thank takeshi tsukada for important insights and useful discussions that motivated this work .
we also thank the anonymous reviewers for their useful comments .
this work was supported by mext kakenhi 20700019 and 20240001 , and global coe program `` ceries . ''
10 m. backes , b. kpf , and a. rybalchenko . automatic discovery and quantification of information leaks . in _
ieee symposium on security and privacy _ , pages 141153 .
ieee computer society , 2009 .
t. ball and s. k. rajamani .
the slam project : debugging system software via static analysis . in _ popl _ , pages 13 , 2002 .
g. barthe , p. r. dargenio , and t. rezk .
secure information flow by self - composition . in _ csfw _ , pages 100114 .
ieee computer society , 2004 .
d. beyer , t. a. henzinger , r. jhala , and r. majumdar .
the software model checker blast .
, 9(5 - 6):505525 , 2007 .
d. clark , s. hunt , and p. malacaria .
quantified interference for a while language .
, 112:149166 , 2005 .
d. clark , s. hunt , and p. malacaria .
quantitative information flow , relations and polymorphic types .
, 15(2):181199 , 2005 .
d. clark , s. hunt , and p. malacaria . a static analysis for quantifying information flow in a simple imperative language .
, 15(3):321371 , 2007 .
m. r. clarkson , a. c. myers , and f. b. schneider .
belief in information flow . in _
pages 3145 .
ieee computer society , 2005 .
m. r. clarkson and f. b. schneider
. hyperproperties . in _ csf _
, pages 5165 .
ieee computer society , 2008 .
e. s. cohen .
information transmission in computational systems . in _
sosp _ , pages 133139 , 1977 . .
darvas , r. hhnle , and d. sands . a theorem proving approach to analysis of secure information flow . in d.
hutter and m. ullmann , editors , _ spc _ , volume 3450 of _ lecture notes in computer science _ , pages 193209 .
springer , 2005 .
d. e. r. denning . .
addison - wesley longman publishing co. , inc . ,
boston , ma , usa , 1982 .
c. flanagan and j. b. saxe . avoiding exponential explosion : generating compact verification conditions . in _
pages 193205 , 2001 .
j. a. goguen and j. meseguer .
security policies and security models . in _ ieee symposium on security and privacy _ , pages 1120 , 1982 .
t. a. henzinger , r. jhala , r. majumdar , and g. sutre .
lazy abstraction . in _ popl _ , pages 5870 , 2002 .
b. kpf and d. basin . an information - theoretic model for adaptive side - channel attacks . in _
ccs 07 : proceedings of the 14th acm conference on computer and communications security _ , pages 286296 , new york , ny , usa , 2007 .
k. r. m. leino .
efficient weakest preconditions . , 93(6):281288 , 2005 .
p. li and s. zdancewic .
downgrading policies and relaxed noninterference . in j.
palsberg and m. abadi , editors , _ popl _ , pages 158170 .
acm , 2005 .
p. malacaria .
assessing security threats of looping constructs . in m.
hofmann and m. felleisen , editors , _ popl _ , pages 225235 .
acm , 2007 .
p. malacaria and h. chen .
lagrange multipliers and maximum information leakage in different observational models . in _
plas 08 : proceedings of the third acm sigplan workshop on programming languages and analysis for security _ , pages 135146 , new york , ny , usa , 2008 .
j. l. massey .
guessing and entropy . in _
isit 94 : proceedings of the 1994 ieee international symposium on information theory _
, page 204 , 1994 .
s. mccamant and m. d. ernst .
quantitative information flow as network flow capacity . in r.
gupta and s. p. amarasinghe , editors , _ pldi _ , pages 193205 .
acm , 2008 .
j. mclean . a general theory of composition for trace sets closed under selective interleaving functions . in _
sp 94 : proceedings of the 1994 ieee symposium on security and privacy _
, page 79 , washington , dc , usa , 1994 .
ieee computer society .
k. l. mcmillan .
lazy abstraction with interpolants . in t. ball and r. b. jones , editors , _ cav _ , volume 4144 of _ lecture notes in computer science _ , pages 123136 .
springer , 2006 .
d. a. naumann . from coupling relations to mated invariants for checking information flow . in _
computer security - esorics 2006 , 11th european symposium on research in computer security , proceedings _
, pages 279296 , hamburg , germany , sept . 2006 .
j. newsome , s. mccamant , and d. song .
measuring channel capacity to distinguish undue influence . in _ proceedings of the fourth acm sigplan workshop on programming languages and analysis for security ( plas ) _ , dublin , ireland ,
june 2009 .
a. sabelfeld and a. c. myers . a model for delimited information release .
in k. futatsugi , f. mizoguchi , and n. yonezaki , editors , _ isss _ , volume 3233 of _ lecture notes in computer science _ , pages 174191 .
springer , 2003 .
c. shannon . a mathematical theory of communication .
, 27:379423 , 623656 , 1948 .
g. smith .
on the foundations of quantitative information flow . in _
fossacs 09 : proceedings of the 12th international conference on foundations of software science and computational structures _ ,
pages 288302 , berlin , heidelberg , 2009 .
springer - verlag .
t. terauchi and a. aiken .
secure information flow as a safety problem . in c.
hankin and i. siveroni , editors , _ sas _ , volume 3672 of _ lecture notes in computer science _ , pages 352367 .
springer , 2005 .
s. toda .
is as hard as the polynomial - time hierarchy .
, 20(5):865877 , 1991 .
h. unno , n. kobayashi , and a. yonezawa .
combining type - based analysis and model checking for finding counterexamples against non - interference . in v. c. sreedhar and s. zdancewic , editors , _ plas _ , pages 1726 .
acm , 2006 .
d. volpano , g. smith , and c. irvine . a sound type system for secure flow analysis .
, 4(3):167187 , 1996 .
we define some abbreviations . [ def : distabrv ]
@xmath312 we use this notation whenever the correspondences between random variables and their values are clear . for convenience , we sometimes use large letters @xmath1 , @xmath47 , @xmath2 , etc . to range over boolean variables as well as generic random variables . for simplicity
, we often compute the shannon - entropy based quantitative information flow for programs that do not have low security inputs . for such programs ,
the equation _ se _ from definition [ def : se ] can be simplified as follows
. @xmath313(m)&=\mathcal{i}[\mu](o;h)\\ & = \mathcal{h}[\mu](o ) \end{array}\ ] ] we note the following property of deterministic programs @xcite .
[ lem : detse ] for @xmath8 deterministic , @xmath314(m)=\mathcal{i}[\mu](o;h|l ) = \mathcal{h}[\mu](o|l)\ ] ] the following lemma is used to show that we can generate a boolean formula that has exactly the desired number of solutions in polynomial ( actually , linear ) time . [
lem : a2 ] let @xmath21 be an integer such that @xmath315 .
then , a boolean formula that has exactly @xmath21 assignments over the variables @xmath316 can be computed in time linear in @xmath317 .
we define a procedure that returns the boolean formula .
below , @xmath318 , i.e. , @xmath319 is the @xmath178th variable .
@xmath320 here , @xmath321 is an empty string .
let @xmath322 be a @xmath317-bit binary representation of @xmath21 .
we prove that @xmath323 returns a boolean formula that has exactly k assignments by induction on the number of variables , that is , @xmath317 .
* @xmath324 * * @xmath325 + @xmath326 returns @xmath327 , that is , @xmath328 .
@xmath151 has no satisfying assignment .
* * @xmath329 + @xmath330 returns @xmath331 , that is , @xmath332 .
@xmath332 has only one satisfying assignment . *
@xmath333 * * @xmath334 + let @xmath335 be a binary representation of @xmath21 . @xmath336
returns @xmath337 . by induction hypothesis , @xmath338 has @xmath21 satisfying assignments for @xmath316 .
it follows that @xmath337 has just @xmath21 satisfying assignments , because @xmath339 has no assignment and @xmath340 has just @xmath21 assignments . *
* @xmath341 + let @xmath342 be a binary representation of @xmath21 .
@xmath343 returns @xmath344 .
@xmath51 is a binary representation of @xmath345 . by induction hypothesis
, @xmath338 has @xmath345 satisfying assignments for @xmath316 .
it follows that @xmath344 has just @xmath21 satisfying assignments , because @xmath346 has just @xmath345 assignments and when @xmath347 , @xmath344 has just @xmath348 assignments .
we frequent the following property of logarithmic arithmetic when proving statements concerning the shannon entropy .
[ lem : a7 ] let @xmath349 and @xmath350 be numbers such that @xmath351 $ ] .
then , we have @xmath352 . because @xmath353 and @xmath354 , it follows that , @xmath355
[ thm : reqni ] let @xmath8 be a program that takes high - security input @xmath1 , low - security input @xmath47 , and returns low - security output @xmath2 .
then , @xmath8 is non - interferent if and only if @xmath356(m)=0 $ ] . * ( @xmath358 ) suppose that @xmath8 is non - interferent .
then , by lemma [ lem : detse ] , @xmath52(m)&=&\mathcal{i}[\mu](o;h|l)\\ & = & \mathcal{h}[\mu](o|l)\\ & = & \sum_{o}\sum_{\ell}\mu(o,\ell)\log\frac{\mu(\ell)}{\mu(o,\ell)}\\ & = & \sum_{o}\sum_{\ell}\mu(o,\ell)\log\frac{\mu(\ell)}{\mu(\ell)}\\ & = & 0 \end{array}\ ] ] the last step follows from the fact that non - interference implies @xmath359 . *
( @xmath360 ) suppose that @xmath8 is interferent .
then , there must be @xmath361 and @xmath226 such that @xmath362 , @xmath363 , and @xmath272 .
pick a probability function @xmath11 such that @xmath267 . then , by lemma [ lem : detse ] , @xmath52(m)&=&\mathcal{i}[\mu](o;h|l)\\ & = & \mathcal{h}[\mu](o|l)\\ & = & \sum_{o}\sum_{\ell}\mu(o,\ell)\log\frac{\mu(\ell)}{\mu(o,\ell)}\\ & = & \mu(o_0,\ell')\log\frac{\mu(\ell')}{\mu(o_0,\ell')}\\ & & \qquad+\mu(o_1,\ell')\log\frac{\mu(\ell')}{\mu(o_1,\ell')}\\ & = & \frac{1}{2}\log 2 + \frac{1}{2}\log 2\\ & = & 1 \end{array}\ ] ] therefore , there exists @xmath11 such that @xmath18(m ) \neq 0 $ ] , and we have the conclusion .
for contradiction , suppose @xmath110 $ ] is a @xmath21-safety property .
let @xmath8 and @xmath114 be programs having same input domain such that @xmath364 $ ] .
then , it must be the case that there exist @xmath365\hspace*{-1.2pt}]}$ ] and @xmath366\hspace*{-1.2pt}]}$ ] such that @xmath118 , @xmath119 , and @xmath367\hspace*{-1.2pt } ] } \wedge t ' \subseteq { [ \hspace*{-1.2pt}[m_c']\hspace*{-1.2pt } ] } \rightarrow ( m_c , m_c ' )
\not\in c_{\it me}[u]$ ] . the number of outputs of the program @xmath124 is greater than or equal to the number of the outputs of the program @xmath123 .
hence , by lemma [ lem : melog ] , we have @xmath371 $ ] .
but , @xmath143\hspace*{-1.2pt}]}$ ] and @xmath144\hspace*{-1.2pt}]}$ ] .
this leads to a contradiction . for contradiction ,
suppose @xmath111 $ ] is a @xmath21-safety property .
let @xmath8 and @xmath114 be programs having the same input domain such that @xmath385 $ ] .
then , it must be the case that there exist @xmath365\hspace*{-1.2pt}]}$ ] and @xmath366\hspace*{-1.2pt}]}$ ] such that @xmath118 , @xmath119 , and @xmath367\hspace*{-1.2pt } ] } \wedge t ' \subseteq { [ \hspace*{-1.2pt}[m_c']\hspace*{-1.2pt } ] } \rightarrow ( m_c , m_c ' )
\not\in c_{\it ge}[u]$ ] .
we compare the guessing - entropy - based quantitative information flow of the two programs .
@xmath386(\bar{m'})-{\it ge}[u](\bar{m})\\ \quad=\frac{|\mathbb{h}|}{2}-\frac{1}{2|\mathbb{h}|}\sum_{o'\in m'(\mathbb{h})}|m'^{-1}(o')|^2\\ \qquad-\frac{|\mathbb{h}|}{2}+\frac{1}{2|\mathbb{h}|}\sum_{o\in m(\mathbb{h})}|m^{-1}(o)|^2\\ \quad=\frac{1}{2|\mathbb{h}|}\sum_{o\in m(\mathbb{h})}|m^{-1}(o)|^2\\ \qquad-\frac{1}{2|\mathbb{h}|}\sum_{o'\in m'(\mathbb{h})}|m'^{-1}(o')|^2\\ \quad=\frac{1}{2|\mathbb{h}|}(\sum_{o_x\in{\{{o_1,\dots , o_i}\}}}|m^{-1}(o_x)|^2\\ \qquad\qquad+|m^{-1}(o)|^2+|m^{-1}(o_r)|^2)\\ \qquad-\frac{1}{2|\mathbb{h}|}(\sum_{o_x'\in{\{{o_1',\dots , o_j'}\}}}|m'^{-1}(o_x')|^2\\ \qquad\qquad+\sum_{o_y'\in{\{{o_{j+1}',\dots , o_{j+i}'}\}}}|m'^{-1}(o_y')|^2\\ \qquad\qquad+|m'^{-1}(o_r')|^2)\\ \end{array}\ ] ] by lemma [ lem : gelem ] , we have @xmath387 trivially , we have @xmath388 as a result , we have @xmath389(\bar{m'})-{\it ge}[u](\bar{m})\ge 0\ ] ] recall that @xmath123 and @xmath124 have the same counterexamples @xmath141 and @xmath142 , that is , @xmath143\hspace*{-1.2pt}]}$ ] and @xmath144\hspace*{-1.2pt}]}$ ] .
however , we have @xmath390 $ ] .
this leads to a contradiction .
[ lem : a5 ] let @xmath176 be distinct boolean variables , @xmath161 be a boolean formula over @xmath176 , and @xmath202 be the number of satisfying assignments for @xmath161 . if @xmath202 is less than @xmath394 , then the number of the outputs of the boolean program @xmath395 defined in figure [ boolenc ] is equal to @xmath396 .
[ lem : a6 ] let @xmath176 be distinct variables and @xmath161 be a boolean formula over @xmath176 .
then , the number of assignments for @xmath161 can be computed by executing an oracle that decides whether programs are in @xmath110 $ ] at most @xmath196 times .
let @xmath397 where @xmath198 is a formula over @xmath176 having @xmath179 assignments and @xmath177 is a boolean variable such that @xmath398 .
note that by lemma [ lem : a2 ] , such @xmath198 can be generated in linear time .
then , we invoke the following procedure where @xmath141 is defined in figure [ boolenc ] .
@xmath399}\\ \qquad\qquad{\sf and}\;(t(b(n)),t(\phi\wedge h'))\in
c_{\it me}[u])\\ \qquad{\sf if}\;(t(\phi\wedge h'),t(b(n)))\in c_{\it me}[u]\\ \qquad\qquad{\sf then}\;\{\ell = n;n=(\ell+r)/2;\}\\ \qquad\qquad{\sf else}\;\{r = n;n=(\ell+r)/2;\}\\ { \sf return}\;n \end{array}\ ] ] note that when the procedure terminates , we have @xmath400(t(b(n))={\it me}[u](t(\phi\wedge h'))$ ] , and so by lemma [ lem : melog ] and lemma [ lem : a5 ] , @xmath202 is the number of satisfying assignments to @xmath161 . we show that the procedure iterates at most @xmath203 times . to see this , note that every iteration in the procedure narrows the range between @xmath204 and @xmath51 by one half .
because @xmath205 is bounded by @xmath206 , it follows that the procedure iterates at most @xmath203 times .
hence , the oracle @xmath110 $ ] is accessed @xmath196 times , and this proves the lemma .
[ lem : gemono ] let @xmath176 and @xmath177 be distinct variables and @xmath161 and @xmath401 be boolean formulas over @xmath176 .
let @xmath402 and @xmath403 .
then , we have @xmath404 iff @xmath81(m)\le{\it ge}[u](m')$ ] . by the definition , @xmath405(m)&=\mathcal{g}(h)-\mathcal{g}(h|o)\\ & = \frac{1}{2}(|{\overrightarrow{{h}}}|)+\frac{1}{2}-\sum_o\sum_{1\le i\le |{\overrightarrow{{h}}}|}i u(h_i , o)\\ & = \frac{|{\overrightarrow{{h}}}|}{2}\\ & \quad -\frac{1}{2|{\overrightarrow{{h}}}|}(|m^{-1}({\sf true})|^2+|m^{-1}({\sf false})|^2 )
\end{array}\ ] ] therefore , @xmath389(m ) \leq { \it ge}[u](m')\ ] ] iff @xmath406 but , trivially , the latter holds iff @xmath407 [ lem : gecomp ] let @xmath176 and @xmath177 be distinct variables and @xmath161 be a boolean formula over @xmath176 . then , the number of assignments for @xmath161 can be computed by executing an oracle that decides whether programs are in @xmath111 $ ] at most @xmath196 times .
let @xmath397 where @xmath198 is a formula over @xmath176 having @xmath179 assignments and @xmath177 is a boolean variable such that @xmath398 .
note that by lemma [ lem : a2 ] , such @xmath198 can be generated in linear time .
@xmath408}\\ \qquad\qquad{\sf and}\;(o:=b(n),o:=\phi\wedge h')\in
c_{\it ge}[u])\\ \qquad{\sf if}\;(o:=\phi\wedge h',o:=b(n))\in c_{\it ge}[u]\\ \qquad\qquad{\sf then}\;\{\ell = n;n=(\ell+r)/2;\}\\ \qquad\qquad{\sf else}\;\{r = n;n=(\ell+r)/2;\}\\ { \sf return}\;n \end{array}\ ] ] we show that the procedure iterates at most @xmath203 times . to see this , every iteration in the procedure narrows the range between @xmath204 and @xmath51 by one half . because @xmath205 is bounded by @xmath206 , it follows that the procedure iterates at most @xmath203 times .
hence , the oracle @xmath111 $ ] is accessed @xmath196 times , and this proves the lemma .
* @xmath411 + we reduce via self composition @xcite .
let @xmath8 be a boolean program that we want to know if it is non - interferent .
first , we make a copy of @xmath8 , with each variable @xmath152 in @xmath8 replaced by a fresh ( primed ) variable @xmath412 .
call this copy @xmath114 .
let @xmath413 , where @xmath414 is the boolean formula encoding the conjunction of equalities @xmath415 , @xmath416 , , @xmath417 , where @xmath418 are the low security output variables of @xmath8 .
note that @xmath161 can be obtained in time polynomial in the size of @xmath8 . here , instead of the rules in figure [ wpsemantics ] , we use the optimized weakest precondition generation technique @xcite that generates a formula quadratic in the size of @xmath419 . then
, @xmath8 is non - interferent if and only if @xmath161 is valid , that is , if and only if @xmath420 is unsatisfiable .
* @xmath421 + let @xmath161 be a formula that we want to know if it is unsatisfiable .
we prove that the following programs is non - interferent iff @xmath161 is unsatisfiable . here , all variables that appear in @xmath161 are high security input variables and @xmath1 is a high security input variable that is distinct from variables appearing in @xmath161 , and @xmath2 is the low security output variable .
@xmath422 trivially , if @xmath161 is unsatisfiable , then this program returns only @xmath151 , that is , this program is non - interferent .
if this program is non - interferent , then this program returns only @xmath423 for any input , or returns only @xmath151 for any input .
however , this program can not return only @xmath424 , because if @xmath425 then @xmath426 .
therefore , this program only returns @xmath151 , when this program is non - interferent .
that means @xmath161 is unsatisfiable when the program is non - interferent .
[ lem : a10 ] let @xmath430 be a set , and @xmath8 and @xmath114 be functions whose domains contain @xmath430 .
suppose that we have @xmath431 , for all @xmath432 in @xmath430 .
then , for all @xmath433 , we have @xmath434 .
[ lem : a12 ] let @xmath1 , @xmath2 , @xmath435 , and @xmath47 be distinct random variables .
let @xmath8 and @xmath114 be programs .
we have @xmath259 iff for any distribution @xmath11 , @xmath436(h|o',l)\le \mathcal{h}_\infty[\mu](h|o , l)$ ] where @xmath437 and @xmath438 . * ( @xmath358 ) + suppose @xmath225 .
we have @xmath439(h|o',l)\le \mathcal{h}_\infty[\mu](h|o , l)\\ \qquad\textrm { iff } \mathcal{v}[\mu](h|o , l)\le \mathcal{v}[\mu](h|o',l ) \end{array}\ ] ] by the definition of min entropy , and + @xmath440(h|o , l)\\ \quad=\sum_{o\in{\mathbb o},\ell\in{\mathbb l } } \mu(o,\ell)\max_{h\in\mathbb h } \mu(h|o,\ell)\\ \quad = \sum_{o\in{\mathbb o},\ell\in{\mathbb l } } \mu(o,\ell ) \max_{h\in\mathbb h } \frac{\mu(h , o,\ell)}{\mu(o,\ell)}\\ \quad = \sum_{o\in{\mathbb o},\ell\in{\mathbb l } } \max_{h\in\mathbb h } \mu(o,\ell)\frac{\mu(h , o,\ell)}{\mu(o,\ell)}\\ \quad = \sum_{o\in{\mathbb o},\ell\in{\mathbb l}}\max_{h\in\mathbb h } \mu(h , o,\ell)\\ \quad = \sum_{o\in{\mathbb o},\ell\in{\mathbb l}}\max_{h\in { \{{h'\mid o = m(h',\ell)}\ } } } \mu(h,\ell)\\ \end{array}\ ] ] where @xmath441 $ ] , and @xmath442 and @xmath443 are sample spaces of low - security input and high - security input , respectively .
therefore , it suffices to show that @xmath440(h|o',l)-\mathcal{v}[\mu](h|o , l)\\ \quad=\sum_{o'\in{\mathbb o'},\ell\in{\mathbb l}}\max_{h\in{\{{h'\mid o'=m'(h',\ell)}\ } } } \mu(h,\ell)\\ \qquad-\sum_{o\in\mathbb o,\ell\in{\mathbb l}}\max_{h\in{\{{h'\mid o = m(h',\ell)}\ } } } \mu(h,\ell)\\ \quad\ge 0 \end{array}\ ] ] where @xmath444 $ ] .
+ for any @xmath445 and @xmath66 , there exists @xmath446 such that @xmath447 . because @xmath225 , by lemma [ lem : a10 ] , we have @xmath448 + therefore , @xmath449 for some @xmath450 . hence , each summand in @xmath451 also appears in @xmath452 . and
, we have the above proposition .
* ( @xmath360 ) + we prove the contraposition .
. then , there exist @xmath453 such that @xmath454 , @xmath455 , @xmath253 , and @xmath272 .
pick a probability distribution @xmath11 such that @xmath456 . then
, we have @xmath440(h|o',l)\\ \quad = \sum_{o'\in{\mathbb o'},\ell\in{\mathbb l}}\max_{h\in{\{{h'\mid o'=m(h',\ell)}\ } } } \mu(h,\ell)\\ \quad= \frac{1}{2 } \end{array}\ ] ] and @xmath440(h|o , l)\\ \quad=\sum_{o\in{\mathbb o},\ell\in{\mathbb l}}\max_{h\in{\{{h'\mid o = m(h',\ell)}\ } } } \mu(h,\ell)\\ \quad= \frac{1}{2}+\frac{1}{2}\\ \quad= 1 \end{array}\ ] ] therefore , @xmath436(h|o',l)\not\le \mathcal{h}_\infty[\mu](h|o , l)$ ] .
straightforward from lemma [ lem : a12 ] and the fact that @xmath436(h|l)-\mathcal{h}_\infty[\mu](h|o , l)\le \mathcal{h}_\infty[\mu](h|l)-\mathcal{h}_\infty[\mu](h|o',l)$ ] iff @xmath436(h|o , l)\ge\mathcal{h}_\infty[\mu](h|o',l)$ ] . * @xmath216 + suppose @xmath458 . by the definition ,
@xmath459(m)=\\ \sum_{\ell\in \mathbb{l},h\in\mathbb{h}}in(\lambda h'.\mu(h',\ell),{\mathbb h},h)\mu(h,\ell)\\ -\sum_{o\in\mathbb{o},\ell\in\mathbb{l},h\in\mathbb{h } } in(\lambda h'.\mu(h',o,\ell),\mathbb{h},h)\mu(h , o,\ell ) \end{array}\ ] ] and @xmath460(m')=\\ \sum_{\ell\in \mathbb{l},h\in\mathbb{h}}in(\lambda
h'.\mu(h',\ell),{\mathbb h},h)\mu(h,\ell)\\ -\sum_{o'\in\mathbb{o'},\ell\in\mathbb{l},h\in\mathbb{h } } in(\lambda h'.\mu(h',o',\ell),\mathbb{h},h)\mu(h , o',\ell ) \end{array}\ ] ] where @xmath461 $ ] and @xmath462 $ ] .
+ it suffices to show that @xmath463 + let @xmath464 and @xmath66 .
let @xmath465 , and let @xmath466 . because @xmath225 , for any @xmath241 such that @xmath467 , we have @xmath468
. then , by lemma [ lem : gelem ] , we have @xmath469 where @xmath470 + now we prove each @xmath247 constructed above are disjoint .
that is , for @xmath248 and @xmath249 outputs of @xmath8 such that @xmath250 , @xmath471 . for a contradiction , suppose @xmath252
then , there exist @xmath226 and @xmath227 such that @xmath253 , @xmath254 , @xmath255 , and @xmath256 . since @xmath225 , we have @xmath257 , and it leads to a contradiction .
hence , we have for any @xmath61 , @xmath472 + therefore , it follows that @xmath463 * @xmath221 + we prove the contraposition .
. then , there exist @xmath473 such that * * @xmath474 , @xmath475 , and @xmath476 * * @xmath477 + then , we can pick @xmath11 such that @xmath478 .
we have @xmath79(m)=1.5 - 1=0.5\ ] ] and @xmath79(m')=1.5 - 1.5=0\ ] ] therefore ,
we have @xmath479 $ ] . by theorem [ thm : reqse ]
, we have @xmath481(m)\le{\it se}[\mu](m')\ ] ] now , there exists @xmath482 such that @xmath483(m)\ ] ] therefore , @xmath484(m)\le{\it se}[\mu'](m')\ ] ] trivially , @xmath484(m')\le { \it cc}(m')\ ] ] therefore , we have the conclusion .
* @xmath485 + we prove by reducing @xmath34 to unsat , which is conp - complete .
we reduce via self composition @xcite .
let @xmath8 and @xmath114 be boolean programs that we want to know if they are in @xmath34 .
first , we make copies of @xmath8 and @xmath114 , with all variables in @xmath8 and @xmath114 replaced by fresh ( primed ) variables .
call these copies @xmath486 and @xmath487 .
let @xmath488 where @xmath393,@xmath489,@xmath490 , and @xmath491 are the low security outputs of @xmath8,@xmath486,@xmath114 , and @xmath487 , respectively .
note that @xmath161 can be obtained in time polynomial in the size of @xmath8 and @xmath114 . here , like in theorem [ thm : nicomp ] , we use the optimized weakest precondition generation technique @xcite to generate a formula quadratic in the size of @xmath492 . then , @xmath259 if and only if @xmath161 is valid , that is , if and only if @xmath420 is unsatisfiable . * @xmath493 + we prove by reducing ni to @xmath34 , because ni is conp - complete by theorem [ thm : nicomp ] .
we can check the non - interference of @xmath8 by solving @xmath225 where @xmath114 is non - interferent and have the same input domain as @xmath8 by theorem [ thm : rni ] . note that such @xmath114 can be constructed in polynomial time .
therefore , we have @xmath493 . | researchers have proposed formal definitions of quantitative information flow based on information theoretic notions such as the shannon entropy , the min entropy , the guessing entropy , and channel capacity .
this paper investigates the hardness and possibilities of precisely checking and inferring quantitative information flow according to such definitions .
we prove that , even for just comparing two programs on which has the larger flow , none of the definitions is a k - safety property for any k , and therefore is not amenable to the self - composition technique that has been successfully applied to precisely checking non - interference .
we also show a complexity theoretic gap with non - interference by proving that , for loop - free boolean programs whose non - interference is conp - complete , the comparison problem is # p - hard for all of the definitions . for positive results , we show that universally quantifying the distribution in the comparison problem , that is , comparing two programs according to the entropy based definitions on which has the larger flow for all distributions , is a 2-safety problem in general and is conp - complete when restricted for loop - free boolean programs .
we prove this by showing that the problem is equivalent to a simple relation naturally expressing the fact that one program is more secure than the other .
we prove that the relation also refines the channel - capacity based definition , and that it can be precisely checked via the self - composition as well as the `` interleaved '' self - composition technique . |
quantum hamiltonian complexity ( qhc ) has blossomed into an incredibly active field of research over the past few years ( see @xcite for an introduction ) .
a key player in qhc is the local hamiltonian problem ( @xmath10 ) @xcite , whose study had led to various unexpected discoveries , ranging from the universality of adiabatic computation @xcite , to the @xmath0 hardness of one dimensional systems @xcite , to quantum gadgets @xcite , to computational complexity lower bounds on notorious open problems in physics @xcite , to the hardness of translationally invariant systems @xcite , and more . in this note
we focus on a special case of the @xmath10 problem , called the commuting local hamiltonian problem ( @xmath2 ) , which has attracted considerable attention in the past few years ( e.g.,@xcite ) . before we explain our result ,
let us define the problem more precisely .
a @xmath6-local hamiltonian is a hermitian matrix @xmath11 operating on the hilbert space of @xmath12 @xmath7-dimensional particles ( _ qudits _ ) , where each term @xmath13 ( sometimes called constraint ) acts non - trivially on at most @xmath6 particles .
kitaev defined in @xmath14 the local hamiltonian ( @xmath10 ) problem @xcite , where one is given a local hamiltonian on @xmath12 qudits , and two real numbers @xmath15 , with @xmath16 , and is asked whether the lowest eigenvalue ( _ ground energy _ ) is at most @xmath17 or at least @xmath18 .
kitaev showed @xcite that this problem is complete for the class @xmath0 , the quantum analogue of @xmath1 .
this provided a quantum counterpart of the celebrated cook - levin theorem @xcite . to restrict the discussion to _ commuting _
local hamiltonians , we require in addition that every two terms @xmath19 commute .
w.l.o.g we can assume in this case that @xmath13 s are projections .
we denote the family of such hamiltonians acting on @xmath7 dimensional particles and where each term is @xmath6-local , by @xmath20 ; we can now consider the @xmath2 problem ( introduced by bravyi and vyalyi in @xmath21 @xcite ) , which is the analogue of the @xmath10 problem for the commuting case .
note that since we are concerned with @xmath13 s which are commuting projections , the eigenvalues of @xmath22 are natural numbers .
the problem is now to distinguish between the ground energy being @xmath23 or at least @xmath24 .
the commutation restriction might seem at first sight to devoid the @xmath10 problem of its quantum nature , and place it in @xmath1 , since all local terms can be diagonalized simultaneously .
moreover , one usually attributes the interesting features of quantum mechanics to its non - commutative nature , cf . the heisenberg uncertainty principle .
this intuition is importantly wrong .
the main source for the interest in @xmath2s lies in the beautiful and striking fact that the groundspaces of @xmath2s are capable of possessing intriguing multi - particle entanglement phenomena , such as topological order @xcite , and quantum error correction @xcite ; the most well known example is kitaev s toric code @xcite . due to this intriguing phenomenon
, the study of @xmath2s has become highly influential in the physics as well as the quantum complexity communities , see e.g. the study of levin - wen models @xcite , quantum double models @xcite , as well as that of stabilizer error correcting codes @xcite , which are special cases of @xmath2 s . in their @xmath21 paper , bravyi and vyalyi
@xcite proved a central result regarding the @xmath2 problem : the @xmath9-local case ( namely , the case of @xmath25 ) lies in @xmath1 .
the proof uses a clever application of the theory of representations of @xmath26-algebras , with which bravyi and vyalyi showed that for such systems , there is always a ground state which can be generated by a constant depth quantum circuit ; in particular , the description of the constant depth quantum circuit can be given as a classical witness , from which the energy of the state can be computed efficiently by a classical computer ( implying the containment in @xmath1 ) .
the @xmath2 question is thus interesting ( from the quantum complexity point of view ) only with locality @xmath27 .
the bravyi - vyalyi result was extended later in various special cases @xcite , but the complexity of the general @xmath2 problem remains open , and as far as we currently know it may lie anywhere between @xmath1 and the full power of @xmath0 . - which is defined like @xmath0 except in case of yes , there exists a state which is accepted with probability exactly @xmath28 . here
we will not distinguish between @xmath0 and @xmath29 since we are not using those terms technically , but see @xcite and @xcite . ] in this paper , we study the hardness of _ approximation _ versions of the @xmath2 problem . our motivation is two - fold : one , we view this as an important stepping stone towards the clarification of the complexity of the exact @xmath2 problem , which has resisted resolution so far , despite the effort described above .
two , we view it as an important special case of a major conjecture in qhc , namely , the quantum @xmath8 ( @xmath3 ) conjecture , to be discussed below . our main result states that the approximation problem of @xmath2 lies within @xmath1 , whenever the underlying bi - partite interaction graph of the @xmath2 instance is a good enough _ locally - expanding graph _
( for definition see below ) .
this means that good local expansion of the underlying interaction graph does not help but in fact _ disturbs _ in generating `` hard '' quantum instances , in sharp contrast to the classical analogue in which the hard instances are excellent locally - expanding graphs ( see @xcite and theorem [ thm : nphard ] below ) .
the result implies severe constraints on the structure of possible @xmath3s for @xmath2s , as the output of those can not be good local expanders .
before we state the result and its implications more formally , we provide some further background and context .
the classical @xmath8 theorem @xcite is arguably the most important discovery in classical theoretical computer science over the past quarter century ; to state it , recall the constraint satisfaction problem ( @xmath30 ) : given is a collection of @xmath6-local constraints on @xmath12 boolean or @xmath7-state variables .
the problem is to decide whether the instance is satisfiable or not ; this problem is @xmath1 complete .
the @xmath8 theorem states that there exists an efficient mapping ( known as a @xmath8 reduction ) which takes satisfiable instances to satisfiable instances and non - satisfiable ones to non - satisfiable ones , such that the @xmath1 witness for the new instances can be tested , with constant probability of success , by reading only a constant number of randomly chosen locations in the witness ( ! )
. stated differently , the @xmath8 theorem says that it is @xmath1 hard to decide whether a given instance is satisfiable , or at most some constant fraction , say @xmath31 , of its constraints can be satisfied .
one may ask @xcite does an analogous statement hold in the quantum setting as well ?
a quantum @xmath8 conjecture @xcite can be phrased as follows : * quantum @xmath8 ( @xmath3 ) * there exists @xmath32 , such that it is @xmath0-hard to decide whether a given @xmath6-local hamiltonian @xmath33 , with @xmath34 whose terms are all positive semi - definite , is satisfiable ( namely , has ground value @xmath35 ) or its minimal eigenvalue is at least @xmath36 .
this conjecture is one of the major open problems in qhc today ( see the recent survey @xcite ) .
it is important for several reasons .
first of all , the question is related to the question of whether quantum entanglement can be made robust at room temperature . secondly , the question of whether or not @xmath3s exist is tightly related to fundamental questions regarding local and global multi - particle entanglement .
the resolution of the conjecture , in either direction , would constitute a major advance in qhc and in our understand of quantum matter .
however , despite considerable effort , so far all attempts to extend the classical proofs of the @xmath8 theorem to the quantum settings , or to dispute the conjecture , have encountered severe obstacles @xcite .
we currently do not even have a good guess whether the conjecture should hold or not .
just like in the classical case , the @xmath3 conjecture is equivalent to the following statement : the problem of approximating the ground energy of local hamiltonians to within a constant fraction is @xmath0 hard .
this means that the clarification of the complexity of approximating the local hamiltonian problem is of major interest .
given the difficulty of this question , it is natural to study hardness of approximation of restricted classes hamiltonians ; we focus here on @xmath2s , which are a particularly attractive special case , since they are both much easier to analyze than the general case , and in addition , they possess a simplifying structure as introduced in @xcite , while still giving rise to extremely complex phenomenon such as topological order , and used as a substrate for almost all known quantum codes . the geometry of the underlying interaction graph is well known to play a crucial role for the complexity of constraint satisfaction problem . for a start ,
when a @xmath30 is defined between variables set a one dimensional lattice , with constraints between nearest neighbors only , the problem lies in @xmath37 ( using dynamic programming ) , whereas for @xmath9 or more dimensions , the problem is @xmath1-hard .
when one is interested in approximations ( to within any constant fraction of the number of constraints ) the situation changes : approximating @xmath30s , as well as @xmath10s , to within any constant , lies in @xmath37 when the interaction graph is a lattice of _ any _ constant dimension . to see this , partition the lattice to constant size cubes , and remove the constraints that connect different cubes ; we arrive at a collection of disconnected constant - size subproblems , all of which can be optimized together in @xmath37 .
the number of constraints violated by this solution is at most the number of removed constraints , which can be made to be an arbitrarily small fraction .
this argument works similarly both in the quantum and classical cases , and implies that to make the approximation problem hard , one has to consider @xmath30s , @xmath2s or @xmath10s whose underlying graphs have certain expansion properties . in light of the above , and motivated by the important role that expanders play in the classical theory of hardness of approximation and @xmath8s , we turn to the study of approximating @xmath2 systems on expanders . we need our definition of expansion to work for @xmath6-local hamiltonians , for @xmath27 , since @xmath9-local @xmath2s are already known to be in @xmath1 even without approximation @xcite
. one might consider working with the naturally induced hypergraphs , but there is no consensus on the right definition of expansion for hypergraphs in the literature .
moreover , whereas one is perhaps used to intuitively think of standard graph expansion as the important property for hardness in classical complexity , perhaps due to the strong usage of expansion in dinur s @xmath8 proof @xcite , it is in fact an open question to relate expansion , and in particular _ small - set _ expansion , ( namely expansion of sets of constant fractional size ) to computational hardness @xcite .
here we consider , for a given @xmath2 , the bi - partite graph in which variables are on the left side and constraints are on the right , and a variable is connected to the constraints that it appears in .
for this bi - partite graph , we consider a very _ weak _ notion of expansion , which we call _ local expansion _ : we say that the bi - partite graph induced by a @xmath6-local @xmath2 instance is @xmath4-locally expanding , if for any set @xmath38 of at most @xmath6 particles , the number of local terms incident on these particles is at least @xmath39 , where @xmath40 is the left degree of the graph ( here we assume all degrees are equal , see definition [ def : expbi ] for the general case ) .
observe that @xmath41 is the maximal possible number of constraints acting on those particles , so @xmath4 can be viewed a `` correction '' to this number , or the _ expansion error_. we note that this definition of local expansion cares only about the expansion of constant - size sets , whereas in complexity theory , one is often interested in _ small - set _ bi - partite expanders , in which all sets of size up to some _ constant fraction _ of @xmath42 are required to expand @xcite ; our requirements on bi - partite expanders are thus significantly weaker.s are stronger if they hold for graphs which satisfy weaker requirements ] it turns out that the notion of local expansion can indeed be related , in the classical world , to computational hardness ; as we show in theorem [ thm : nphard ] sufficiently good local expanders are computationally hard for classical computers . as we will see ( in theorem [ thm : approxbpinnp ] ) this is not the case in the quantum world . ] .
we can now state our main result , theorem [ thm : approxbpinnp ] , in which we show that the problem of approximating @xmath2s on locally - expanding graphs lies inside @xmath1 , and thus can not be @xmath0 hard ( under standard complexity assumptions ) : [ thm : approxbpinnp ] ( the approximation of @xmath2 on good locally - expanding bi - partite graphs is in @xmath1 ) let @xmath43 .
let @xmath44 , be an instance of @xmath20 for constants @xmath45 , with bi - partite interaction graph which is @xmath4-locally - expanding , for @xmath46 .
then the @xmath47-approximation problem of @xmath20 on such @xmath4 locally - expanding bi - partite graph graphs is in @xmath1 .
moreover , for any eigenvalue @xmath48 of @xmath22 there exists a state @xmath49 such that @xmath50 , such that @xmath49 can be generated by a constant depth quantum circuit .
we rule out the possibility that @xmath47 is large enough so that by removing a @xmath47 fraction of the terms in @xmath22 we can simply trivialize the problem and place it in @xmath37 ( e.g. , by disconnecting the graph into small components , as is done in the approximation of @xmath2s on lattices ) .
we do this by showing that the problem remains @xmath1-hard even with this approximation factor .
[ thm : nphard ] fix @xmath51 .
let @xmath4 be such that @xmath52 where @xmath53 is a universal constant .
then it is @xmath1-hard to approximate @xmath20 whose bi - partite interaction graph is @xmath4 locally - expanding , to within a factor @xmath54 .
@xmath55 in the above statement is some universal constant which appears in the @xmath8 proof of dinur @xcite ; we delay its exact definition to subsection [ sec : nphard ] but it is not important . theorem [ thm : approxbpinnp ] implies that if indeed one could construct a quantum @xmath8 whose output is a @xmath2 , then in contrast to classical @xmath8 ( see the proof of theorem [ thm : nphard ] ) , the output can not have arbitrarily high local expansion .
the question of approximating the ground energy of local hamiltonians was broadly studied in the physics literature for specific hamiltonians of particular interest in physics .
not much is known regarding the hardness of approximation of the general case .
kempe and gharibian @xcite derived a non - trivial result for a related task : given a @xmath6-local hamiltonian on @xmath7 dimensional qudits , and we are asked to find the state with _ maximal _
energy @xmath56 .
they showed that there always exists a product state with energy at least @xmath57 , and so the problem of approximating max-@xmath6-local hamiltonian to within approximation factor @xmath58 lies in @xmath1 ( see also a follow up paper by the same authors on approximation versions of problems which are computationally even harder @xcite ) .
the approach of examining commuting hamiltonians as a precursor for the general case was taken before , e.g. in @xcite , and by arad @xcite who considered the commuting case as a base for perturbations to derive a partial no - go theorem for quantum @xmath8 .
notably , this approach is the underlying motivation in the recent works of hastings @xcite and hastings and freedman @xcite , which are tightly related to the current paper , as they both attempt to make progress on the @xmath3 conjecture by studying approximations of @xmath2s .
freedman and hastings are searching for hamiltonians whose low energy states are all highly entangled , and in particular can not be generated by constant depth quantum circuits ( such hamiltonians are called @xmath59 for no low - energy trivial states ) .
the existence of such hamiltonians is essential for the @xmath3 conjecture to hold , for otherwise it would be possible to provide a classical witness for the fact that the energy of the hamiltonian is low , placing the problem in @xmath1 .
freedman and hastings @xcite construct a family of @xmath2s which is not @xmath59 but only `` one - sided '' @xmath59 ( see also @xcite for more on this ) ; they leave as an open problem whether true @xmath59 hamiltonians exist .
theorem [ thm : approxbpinnp ] can be applied here to derive an energy _ upper bound _ above which such a construction _ can not _
be @xmath59 , since it shows the existence of trivial states below a certain threshold constant fraction of energy .
we note that this upper bound tends to @xmath35 as the local expansion of the graphs underlying the construction improves , and so one can not hope to achieve better @xmath59 constructions by improving the local expansion of the underlying graph .
we finally mention an important recent work discovered independently and around the same time as ours , by brando and harrow @xcite .
brando and harrow showed that the approximation of @xmath9-local hamiltonians on graphs whose degree is very large lies in @xmath1 ( the factor of approximation depends inverse polynomially on the degree ) .
this also implies that when the underlying graph of the @xmath9-local hamiltonian is an excellent expander ( which means that its degree must be very large ) the approximation problem ( to within a factor which depends inverse polynomially on the degree again ) lies in @xmath1 . both the result of @xcite and our main theorem [ thm : approxbpinnp ] have the same flavor : too good expanders are an obstruction to quantum hardness , contrary to what one might expect given what we know in the classical case ( @xcite , and theorem [ thm : nphard ] ) .
the comparison between the two results , however , is not straight - forward , and in particular none of the results implies the other .
first , theorem [ thm : approxbpinnp ] holds for any constant locality @xmath60 , while the result of @xcite holds only for @xmath25 .
one might ask what does theorem [ thm : approxbpinnp ] give if we substitute @xmath25 ; this turns out to be non interesting from the approximation point of view : it is exactly the result @xcite , which states that the _ exact _
@xmath9-local @xmath2 problem lies in @xmath1 . on the other hand ,
it is unclear whether the results of @xcite can be extended to @xmath6-local hamiltonians , for @xmath27 .
the straight forward attempt to do this , using quantum gadgets @xcite , that transform @xmath6-local hamiltonians @xmath61 into @xmath9-local hamiltonians encounters an immediate obstacle : these gadgets ( e.g. the best known gadgets so far @xcite ) significantly change the geometry of the graph , and in particular , may modify its expansion properties considerably .
specifically , they introduce large sets of vertices with no edges between them , with some vertices having very small degree ; this implies that the results of @xcite will no longer be applicable after such transformations .
our main result implies that surprisingly , good local expansion is an obstruction to quantum hardness of approximation , in the case of commuting local hamiltonians .
it implies that if @xmath3 were to hold for @xmath2s , then the output of the @xmath3 reductions can not be good local expanders .
our result also directly implies that @xmath59 hamiltonians ( see discussion in the previous subsection as well as @xcite ) can not be constructed on too good bi - partite local - expanders . the main insight underlying this work is that local hamiltonian systems made of commuting check terms , are severely limited by a phenomenon called _ monogamy of entanglement_. essentially
, the monogamy of entanglement limits the amount of entanglement that a qudit with @xmath62 quantum levels can `` handle '' .
our study of commuting system thus reveals that monogamy of entanglement is a significant limiting factor , whose influence increases together with expansion .
it is intriguing that a similar phenomenon has been identified by completely different methods , by @xcite in the context of general ( non - commuting ) local hamiltonians of locality @xmath25 .
a natural open question is to generalize both our result and that of @xcite into one coherent stronger result , which would state roughly that for general , not necessarily commuting , @xmath6-local hamiltonians ( for general @xmath6 ) , the expansion of the underlying interaction graph ( where expansion is defined suitably ) is an obstruction against quantum hardness .
another problem , is to generalize the results here and those of @xcite to richer forms of entanglement .
in particular the limitation on quantum hardness need not necessarily come in the form of constant - depth circuit , but rather any form of entanglement which has an efficient description .
we mention another interesting direction : one might hope that theorem [ thm : approxbpinnp ] could be extended to show that the approximation of the @xmath2 problem on a _ general _ graph is in @xmath1 , by `` bridging '' between our results on locally - expanding graphs and the easy case of approximation on lattices .
we note that a hint about the difficulty of such a result comes from the status of the unique - games - conjecture @xcite which is known to be easy on both extreme cases @xcite ( albeit for standard expander graphs rather than for locally - expanding graphs ) , though conjectured ( or rather , speculated ) to be @xmath1-hard in general .
still , it may be possible to achieve some weaker approximation for general graphs using `` bridging '' between the case of lattices and expanders , perhaps using sub - exponential rather than polynomial witnesses , following the works of @xcite and @xcite .
* organization of paper * in section [ sec : bg ] we provide some required background .
section [ sec : approxsec ] proves theorem [ thm : approxbpinnp ] as well as theorem [ thm : nphard ] .
a @xmath63-local hamiltonian on @xmath12 qudits is a hermitian matrix @xmath11 operating on a hilbert space @xmath64 of @xmath12 @xmath7-dimensional particles , where we assume for all @xmath65 , @xmath66 and @xmath13 can be written as : @xmath67 , where @xmath68 is a hermitian operator acting on at most @xmath6 particles .
when we say an operator acts on some particle , we mean it acts non - trivially , namely , it can not be written as an identity on that particle tensor with some operator on the rest . * the local hamiltonian @xmath10 and commuting local hamiltonian ( @xmath2 ) problems * in the @xmath63-local hamiltonian problem on @xmath12 @xmath7-dimensional qudits
we are given a @xmath63-local hamiltonian @xmath69 and two constants @xmath15 , @xmath70 .
we are asked to decide whether the lowest eigenvalue of @xmath22 is at least @xmath18 or at most @xmath17 , and we are promised that one of the two occurs . in the commuting case ,
we are guaranteed in addition that @xmath71=0 $ ] for all @xmath72 , and w.l.o.g , we also assume that @xmath13 s are projections . @xmath17 and @xmath18 can be taken to be @xmath35 and @xmath28 in this case , respectively .
an instance to this problem is called a @xmath20 instance .
we define bi - partite expanders , similar to @xcite , @xcite , who used them to construct locally - testable classical codes .
note that we require expansion to hold only for sets of constant size @xmath6 .
[ def : bipgraph ] * bi - partite interaction graph * for a @xmath63-local hamiltonian @xmath73 we define the bi - partite interaction graph @xmath74 as follows : @xmath75 , the nodes on the left , correspond to the @xmath12 particles of @xmath76 and @xmath77 corresponds to the set of local terms @xmath78 .
an edge exists between a constraint @xmath79 and a particle @xmath80 if @xmath81 acts non - trivially ( namely , not as the identity ) on @xmath82 .
note that the right degree is at most @xmath6 .
[ def : expbi ] * bi - partite local expansion * given a bi - partite graph @xmath83 , we say that a subset @xmath38 is @xmath4-expanding if @xmath84 , where @xmath85 is the degree of @xmath86 , and @xmath87 is the set of neighbors ( in @xmath77 ) of @xmath88 .
a bi - partite graph @xmath74 for a @xmath63-local @xmath10 is said to be @xmath4-locally - expanding , if every subset of particles @xmath38 of size @xmath89 is @xmath4-expanding .
we recall the main lemma of @xcite : [ lem : bvsum ] * adapted from @xcite * consider two commuting local terms @xmath90 , which intersect on some subset of qudits , whose hilbert space is @xmath91 .
then @xmath91 can be written as a direct sum of subspaces @xmath92 where @xmath13 and @xmath93 preserve the @xmath94 subspaces , and moreover , when @xmath90 are restricted to @xmath94 , then @xmath95 acts non trivially only on subsystem @xmath96 ( @xmath97 ) .
in particular , in each summand @xmath98 , both @xmath99 have dimension strictly less than @xmath91 .
thus , under an isometry on @xmath91 and a restriction to one of the @xmath94 s , the terms act on disjoint systems . see figure [ fig : ex2 ] .
in this paper we apply a simple corollary of lemma ( [ lem : bvsum ] ) : [ cor : bv ] let @xmath22 be a @xmath20 instance , and let @xmath100 be a local term such that for any @xmath101 that intersects @xmath100 , the intersection is at most on one qudit .
then for each qudit @xmath102 participating in @xmath100 there exists a direct - sum decomposition of its hilbert space @xmath103 , such that each of these subspaces is preserved by all local terms of @xmath22 , and such that the restriction of all terms in @xmath22 to any subspace @xmath104 results in @xmath105 acting on a disjoint subsystems , respectively , @xmath106 .
in particular , for any @xmath107 , we have that @xmath108 are strictly less than @xmath109 . for each particle
@xmath102 examined by @xmath100 , let the term @xmath110 be the sum of all local terms acting on @xmath102 except @xmath100 .
the lemma can now we applied on @xmath102 .
since each term other than @xmath100 intersects @xmath100 at only one location , we can apply the lemma simultaneously for all qudits in @xmath100 .
the result follows .
note , that the above corollary does not require that any two local terms acting on a qudit in @xmath100 , intersect only on that qudit .
@xmath7 will denote the dimensionality of particles , capital @xmath111 s will denote degrees in a graph , in particular @xmath85 is the degree of a node @xmath86 . given @xmath112 in a bi - partite graph , @xmath87 denotes the neighbor set of @xmath88 in @xmath113 .
@xmath4 will be used to denote the expansion error for bi - partite graphs ( as in definition [ def : expbi ] ) .
@xmath114 will be used to denote the promise gap , or alternatively the approximation error we are allowed ; often this means the fraction of terms we are allowed to throw away given a @xmath2 instance .
our proof makes use of a very simple property of locally - expanding graphs , namely that most of the neighboring hamiltonian terms of a set of qudits of size @xmath6 only touch this set at one point .
we first state a simple fact quantifying this statement ( similar observations were used also in @xcite ) and then explain how it is used in the proof .
[ fact : essence ] consider @xmath38 in a bi - partite graph @xmath83 and let @xmath88 be @xmath4-expanding , for @xmath46 .
then a fraction at most @xmath115 of all vertices of @xmath87 have degree strictly larger than @xmath28 in @xmath88 . put @xmath116 .
the average degree of a vertex in @xmath87 w.r.t .
@xmath117 is at most @xmath118 .
let @xmath119 denote the fraction @xmath120 , where @xmath121 is the set of neighbors of @xmath88 with degree exactly @xmath28 with respect to @xmath88 .
then @xmath122 where @xmath123 is the average degree of a vertex with at least two neighbors in @xmath88 .
then by simple algebra @xmath124 so @xmath125 is a monotonously increasing function of @xmath123 , and since @xmath126 , then @xmath119 is minimized for @xmath127 . hence , @xmath128 and
since @xmath129 we have : @xmath130 the idea of the proof is that if a constraint @xmath13 satisfies a simple condition , namely , that it shares at most one particle with any other constraint , then corollary ( [ cor : bv ] ) can be applied ( with @xmath13 playing the role of @xmath100 ) to `` disentangle '' the action of @xmath13 from the rest of the terms ( from there we can proceed to other terms iteratively ) .
the problem is that @xmath13 does not necessarily satisfy this condition .
the idea is that because the graph is a good locally - expanding graph , we can remove only a small number of terms to achieve the condition .
more formally , we define : [ def : isolate ] * isolated constraints and particles * let @xmath131 be a bi - partite graph .
a constraint @xmath132 is called * isolated * if for any constraint @xmath133 except @xmath134 , we have @xmath135 . we define the * isolation penalty * of @xmath134 , to be the minimal number of constraints in @xmath77 that we need to remove so that @xmath134 is isolated .
we thus can `` isolate '' @xmath13 by removing all terms that share at least @xmath9 vertices with @xmath13 .
this is where we use the local - expansion property : by fact [ fact : essence ] the set of constraints we need to remove constitutes just a small portion of the terms intersecting @xmath13 .
the remaining terms that intersect @xmath13 , intersect it at only one particle .
corollary ( [ cor : bv ] ) can now be applied almost directly on @xmath13 .
the final result is that the term @xmath13 can be separated from the remaining terms and can be viewed ( up to the relevant isometries , and after a restriction to one of the subspaces in each qudit ) as acting on a separated subsystem . to approximate the @xmath2 problem in @xmath1 ,
it is not enough to remove for each term , all terms that make it isolated ; this will require removing way too many terms .
instead , we use the above idea iteratively , and count how many terms we need to remove more carefully . at each iteration
@xmath136 we choose a local term @xmath137 and `` isolate '' it ( see figure [ fig : ex1 ] ) .
corollary ( [ cor : bv ] ) can then be applied , and @xmath137 can be separated from the rest , after restricting each of its qudits to its relevant subspaces .
some qudits may interact , following this process , with only a single local term of the hamiltonian , thereby allowing us to remove them .
we can now iterate this process , picking another local term ; this way we gradually `` tear away '' local subspaces of particles .
the analysis of the upper bound on the number of terms that need to be removed altogether is non - trivial , since the number of particles does not necessarily decrease through one iteration ; it is only their _ dimensionality _ that decreases .
we resort to an _ amortized _ analysis counting the total number of _ local dimensions _ ( namely , the sum over the particles of the local dimension of each particle ) that are removed in total . to the best of our knowledge this amortized counting of local dimensions is novel . note that removing terms here is done not in order to _ disconnect _ parts of a graph , but only to isolate terms ; loosely speaking , to make the interaction pattern more sparse .
this allows a much more economic removal of terms , and enables us to achieve the desired result .
we are now ready to prove our main theorem : ( of theorem [ thm : approxbpinnp ] ) let @xmath22 be an instance of @xmath20 , and let @xmath131 be its bi - partite interaction graph .
we shall find a state whose energy approximates the ground energy of @xmath22 , ( we note that the same procedure can be applied to the approximation of any eigenvalue of @xmath22 ) .
we perform an iterative process . at step @xmath136
, we have a set of remaining terms @xmath138 acting on the remaining hilbert space @xmath139 .
we also have a set @xmath140 which are terms we have collected so far that we want to throw away .
we initialize @xmath141 , i.e. the set of all local terms of @xmath22 , and @xmath142 to be the empty set .
we also initialize @xmath143 .
repeat the following : 1 .
* isolate * pick an arbitrary local term @xmath86 in @xmath138 . and isolate it ( definition [ def : isolate ] ) by removing as few as possible terms from @xmath138 , placing them in @xmath140 .
* isometries * now that @xmath86 is isolated , the conditions of corollary [ cor : bv ] hold : there is a tensor product of @xmath28-local unitaries on the qudits of @xmath86 : @xmath144 , such that applying @xmath145 allows to restrict the hilbert space to some strict subspace containing the zero eigenspace of @xmath138 .
moreover , inside this subspace , the term @xmath86 acts on a disjoint subsystem than the rest of @xmath138 .
we conjugate @xmath86 , and each term of @xmath138 by @xmath145
. we then _ prune _ : remove any triviality we encounter , be it removing qudits from hamiltonian terms or entire terms altogether . after pruning , each local term acts non - trivially on all the qudits that are attached to it .
* update * we set @xmath146 as the set of remaining local terms of @xmath138 after this pruning , and then update @xmath147 as the support of @xmath146 .
we then set @xmath148 as the union of @xmath149 and the term @xmath150 restricted to the appropriate subspaces of its qudits , as in corollary [ cor : bv ] .
we terminate when there is no longer any intersection between two different terms of @xmath138 .
clearly , this ends after at most polynomially many iterations , since the number of terms in @xmath138 decreases by at least @xmath28 each iteration .
let the number of iterations be @xmath151 .
we claim that a state whose energy is within @xmath152 from the ground energy of the original hamiltonian , can be recovered from this procedure by finding the ground state of all terms in @xmath153 , and applying the inverse of the isometries applied along the way .
this is true as long as the subspaces in the direct sum that are chosen at step @xmath136 , contain the groundspace of the current hamiltonian @xmath138 ; the @xmath1 prover can provide the indices of the subspaces so that this holds at each step , and the verification procedure is simply to check that those subspaces contain a non - zero kernel . to see that this implies a constant - depth quantum circuit that generates a groundstate ,
observe that all isometries on a given qudit @xmath102 , produced along the way as a result of isolating all local terms from @xmath154 on @xmath102 , commute by definition , and so we can apply them in any order , and in particular , simultaneously .
their application causes the interaction graph to break into linearly many connected components of size at most @xmath6 each , and so each component can be trivially diagonalized .
the composition of these local diagonalizing unitaries , and the tensor - product isometries results in a circuit of depth @xmath9 ( just as in @xcite ) .
[ [ bounding - the - approximation - error ] ] bounding the approximation error + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we now provide an upper bound on the size of the error @xmath152 . for a given @xmath155 , let @xmath156 denote its degree in @xmath157 .
consider the @xmath136-th `` isolation '' step , and denote @xmath137 as the constraint isolated at step @xmath136 , @xmath158 , and @xmath159 as the bi - partite interaction graph of @xmath160 .
we define for each @xmath161 , the quantity @xmath162 which is the number of check terms removed to @xmath163 during the @xmath136-th step , multiplied by the ratio @xmath164 , where those degrees are taken from @xmath157 .
this accounts for the `` relative penalty '' shared by @xmath102 , w.r.t .
@xmath157 - note that for any iteration @xmath136 , @xmath165 is exactly the number of check terms removed at the @xmath136-th iteration .
hence @xmath166 for any @xmath167 , and any @xmath161 , we have @xmath168 . we know
that in @xmath157 we had @xmath169 , and so by fact [ fact : essence ] in order to isolate @xmath170 , it suffices to remove @xmath171 terms . since @xmath159 is derived from @xmath157 by removing vertices and edges , this bound still holds the isolation penalty can not increase by removing elements from the graph .
so , at each step , when isolating a term acting on a set of particles @xmath170 , we remove at most @xmath171 constraints . by definition ,
@xmath172 .
on the other hand , let us calculate how many qudit dimensions we remove at such an iteration , out of the total number of qudit dimensions ( namely , the sum of dimensions over the particles).-level qudit by @xmath28 , the dimension of the total hilbert - space is reduced by a factor of @xmath173 ; here we are interested however not in the standard dimension of the entire hilbert space but in the _ sum _ of local dimensions over all particles .
] for each qudit @xmath161 , @xmath102 is either removed altogether , and thus its local dimension is reduced by @xmath109 , or it is restricted to some non - trivial subspace and maybe further divided to two sub - particles , one of which is removed from @xmath139 ; in any of these cases , the local dimension decreases by at least @xmath28 .
thus , the total number of local dimensions removed by one application of " isolate is at least @xmath174 .
we observe by the above that every qudit @xmath155 participates in at most @xmath7 iterations .
we would like to sum its relative penalties @xmath162 over those iterations .
by fact ( [ fact : penalty ] ) this sum is at most @xmath175 .
we would now like to sum over all @xmath155 ; by equation [ eq : totalpenalty ] this will upper bound @xmath152 .
we have @xmath176 therefore , the relative penalty satisfies @xmath177 as required .
the proof of theorem [ thm : nphard ] relies on the fact that the output of the @xmath8 reductions in dinur s proof @xcite can be made into an excellent locally - expanding graph without harming the promise gap , by a small modification of her construction .
we provide here only a sketch of the proof , assuming the reader is familiar with @xcite .
providing a self contained proof will require essentially repeating dinur s @xmath8 proof .
( of theorem [ thm : nphard ] ) let us denote by @xmath178 @xmath179-local @xmath30s acting on bits .
note that instances of @xmath178 are special cases of @xmath20 for @xmath180 and @xmath181 .
suppose that for given parameters @xmath182 we are also given @xmath4 , for which @xmath183 .
the theorem will follow for these parameters @xmath184 if we show that it is even @xmath1-hard to approximate @xmath178 whose bipartite interaction graphs are @xmath4-locally - expanding , to within @xmath55 . to do this ,
recall that dinur provides in @xcite a @xmath8 reduction which maps any @xmath9-local @xmath30 instance on @xmath12 variables of alphabet @xmath185 , @xmath186 , to a @xmath30 instance of @xmath187 size , such that the new instance has constant promise gap , say , @xmath188 .
we modify her construction so that the output @xmath30 is @xmath179-local , acting on bits , has local expansion error @xmath4 , and the promise gap is larger than @xmath55 where @xmath55 is @xmath188 divided by the loss in the promise gap introduced by the alphabet reduction step in dinur s construction ; this factor is bounded by a universal constant independent of the parameters of the problem @xcite . to achieve this
we make a small modification of dinur s gap amplification procedure .
recall that this procedure consists of iterating three steps : the _ preprocessing _ step in which the graph is made into a @xmath7-regular expander graph , _ gap amplification _ in which the @xmath30 is translated into another @xmath30 , whose new @xmath9-local constraints are defined using conjunctions of constraints in the original @xmath30 along collections of @xmath136-walks ( walks of length @xmath136 ) , thereby significantly amplifying the promise gap of the instance , but increasing the size of the alphabet , @xmath189 , to be doubly exponential in @xmath136 ; and finally _ the alphabet reduction _ in which a @xmath9-csp instance alphabet @xmath190 , is translated into a @xmath179-local on bits , with some small constant factor loss in the promise gap .
we denote @xmath188 divided by this loss to be @xmath55 . as a final sub - step , the output @xmath179-local @xmath30
is translated into a @xmath9-local @xmath30 on alphabet of size @xmath191 . by every such @xmath179-step iteration
, the gap is amplified by some constant larger than @xmath28 .
we apply dinur s procedure , until the output is a @xmath192 with promise gap at least @xmath188 .
now we add one additional iteration , in which we make two modifications .
first , we choose the parameter @xmath136 , defined to be the length of walks in the gap amplification step , in such a way that @xmath189 , the size of the new alphabet , which is doubly exponential in @xmath136 , satisfies @xmath193 where @xmath194 , and @xmath195 is some combinatorial constant to be determined below .
second , we skip the very final step where @xmath179-local constraints are translated into @xmath9-local constraints .
finally , regardless of how the alphabet reduction is defined in previous iterations ( there are several versions of dinur s proof ) apply at the final iteration alphabet reduction using dinur s variant @xcite of the long code test ( @xcite ) .
we claim that the resulting @xmath30 is a @xmath179-local @xmath30 whose underlying bi - partite graph has @xmath4 local expansion , and whose promise gap is at least @xmath55 .
the fact that the promise gap is at least @xmath55 follows from the fact that we have applied dinur s iteration to achieve promise gap at least @xmath188 , and then the alphabet reduction reduces it to at least @xmath55 regardless of the alphabet size ( this is how we have defined @xmath55 ) .
let us now show why the local expansion error is at most @xmath4 . let @xmath88 be some set of @xmath179 bits .
the local expansion error of @xmath88 can be bounded from above by @xmath196 , where @xmath197 as in the proof of fact ( [ fact : essence ] ) .
hence , it is sufficient to lower bound @xmath119 and show it is at least @xmath198 .
we consider in more detail the long - code construction used in the alphabet reduction .
this step associates with each @xmath9-local constraint acting on two large alphabet variables ( each is of dimension @xmath189 ) with @xmath194 bits , which are supposed to be the long - code encoding of the assignments to those two large alphabet variables .
those bits are called @xmath199 bits ; for each such @xmath75-tuples of bits , the long - code constraints ( called @xmath199 constraints ) are @xmath179-local constraints , one for each triplet among those @xmath75 bits .
in addition , we add for each @xmath189-size variable @xmath189 bits that are responsible for the _ consistency _ of this variable in all constraints it appears in .
these @xmath189 bits are thus called @xmath200-type ; the consistency constraints ( @xmath200-constraints ) , act on one @xmath200-type bit , and two @xmath199-bits from the long - code encoding of a given constraint .
we note that choosing any two bits of the @xmath200 constraint determines the third bit .
we first observe , that the only interesting @xmath88 s are those that have at least @xmath9 @xmath199-type bits in some long - code encoding of the _ same _ constraint .
this is because in all the other cases , either no constraint can intersect @xmath88 in more than one bit ( this is the case when @xmath88 contains no @xmath199 bits , or it contains @xmath179 @xmath199 bits , all from the long - code encoding of different constraints ) or the number of constraints that can intersect @xmath88 in two bits is at most @xmath9 ( this is the case if @xmath88 contains one @xmath199 bit only , or if @xmath88 contains @xmath9 @xmath199 bits , but from _ different _ long - code encodings ) . since all other constraints intersect @xmath88 in exactly one bit , @xmath119 is close enough to @xmath28 .
so we are left with the case that @xmath88 has at least @xmath9 bits in the same long - code encoding .
the fraction of @xmath199 constraints that intersect @xmath88 in @xmath9 places , out of all possible @xmath199 constraints that intersect it , is at most @xmath201 ; the number of @xmath200 constraints that intersect @xmath88 in @xmath9 places , is at most @xmath179 since every two bits determine the third in a @xmath200 constraint .
thus , the fraction of @xmath200 constraints that intersect @xmath88 in two bits , out of all @xmath200 constraints that intersect @xmath88 , is @xmath201 .
thus , the expansion error in this case is at most @xmath201 , which , by choosing @xmath195 correctly in the choice of @xmath136 above , is @xmath202 .
the authors would like to thank irit dinur and gil kalai for insightful discussions and prasad raghavendra , luca trevisan and umesh vazirani for helpful comments .
the research leading to these results has received funding from the european research council under the european union s seventh framework programme ( fp7/2007 - 2013 ) / erc grant agreement n 280157 , and from isf grant no . | the local hamiltonian problem is famously complete for the class @xmath0 , the quantum analogue of @xmath1 @xcite .
the complexity of its semi - classical version , in which the terms of the hamiltonian are required to commute ( the @xmath2 problem ) , has attracted considerable attention recently @xcite due to its intriguing nature , as well as in relation to growing interest in the @xmath3 conjecture @xcite .
we show here that if the underlying bipartite interaction graph of the @xmath2 instance is a good locally - expanding graph , namely , the expansion of any constant - size set is @xmath4-close to optimal , then approximating its ground energy to within additive factor @xmath5 lies in @xmath1 .
the proof holds for @xmath6-local hamiltonians for any constant @xmath6 and any constant dimensionality of particles @xmath7 .
we also show that the approximation problem of @xmath2 on such good local expanders is @xmath1-_hard_. this implies that too good local expansion of the interaction graph constitutes an obstacle against _ quantum _ hardness of the approximation problem , though it retains its classical hardness .
the result highlights new difficulties in trying to mimic classical proofs ( in particular dinur s @xmath8 proof @xcite ) in an attempt to prove the quantum @xmath8 conjecture .
a related result was discovered recently independently by brando and harrow @xcite , for @xmath9-local general hamiltonians , bounding the quantum hardness of the approximation problem on good expanders , though no @xmath1-hardness is known in that case . |
we have grown @xcite high quality single crystals of la@xmath0ba@xmath1cuo@xmath2 with x=0.095 and x=0.08 using floating zone image furnace techniques with a four - mirror optical furnace .
a small single crystal of la@xmath9cuo@xmath12 was employed as a seed for the growth , which was performed under enclosed pressures of 165 and 182 kpa of o@xmath9 gas for x=0.095 and x=0.08 samples respectively .
the resulting @xmath27 gram single crystals of la@xmath0ba@xmath1cuo@xmath2 were cylindrical in shape and cut to dimensions 25 mm in length by 5 mm in diameter ( x=0.095 ) and 38 mm in length by 5 mm diameter ( x=0.08 ) .
we have determined the ba concentration from the htt to mto transition temperature .
the la@xmath0ba@xmath1cuo@xmath2(x=0.095 ) and ( x=0.08 ) single crystals displayed htt to mto structural phase transitions at t@xmath28@xmath3272 k and t@xmath28@xmath3305 k , respectively , and mto to ltt transitions at t@xmath29@xmath345 k and t@xmath29@xmath3 35 k , respectively @xcite .
the bulk superconducting transition temperatures t@xmath8=27 k ( x=0.095 ) and t@xmath8=29 k ( x=0.08 ) are identified using the onset of the zero field cooled diamagnetic response of the crystals , as measured using superconducting quantum interference device ( squid ) magnetometry and shown in fig .
the various structural and superconducting phase transition temperatures are summarised in table 1 , along with the corresponding values for the la@xmath0ba@xmath1cuo@xmath12 ( x=0.125 ) sample @xcite .
oxygen stoichiometry in la@xmath0ba@xmath1cuo@xmath10 is more difficult to quantify , especially at the low levels relevant here
. experience with la@xmath0sr@xmath1cuo@xmath10 suggests the oxygen stoichiometry , y , is negative in as - grown samples , giving rise to crystals which possess an effective doping level which is lower than that given by the sr concentration alone ; the effective doping level is @xmath30 @xcite .
stoichiometric samples at optimal and underdoped sr concentrations display a maximum superconducting t@xmath8 which can not be increased by controlled annealing in o@xmath9 gas .
excess oxygen can be incorporated into lacuo@xmath10 crystals , but only through electrochemical doping methods , in which case y can be as high as 0.11 and the interstitial oxygen organizes itself into staged structures @xcite .
.[tab : table1 ] summary of la@xmath0ba@xmath1cuo@xmath2 structural ( t@xmath28 , t@xmath29 ) , superconducting ( t@xmath8 ) and magnetic ( t@xmath5 ) phase transition temperatures @xcite [ cols="^,^,^,^,^",options="header " , ] neutron scattering experiments were undertaken on the c5 and n5 triple axis spectrometers at the canadian neutron beam centre , chalk river .
all experiments were performed with pyrolytic graphite ( 002 ) planes as monochromator and analyser , with constant e@xmath31 = 14.7 mev .
a graphite filter was placed in the scattered beam to reduce contamination from higher order neutrons .
the single crystals were oriented with ( h , k,0 ) in the horizontal scattering plane .
crystallographic indices are denoted using tetragonal notation , where the basal plane lattice constant a=3.78 at low temperatures .
ba@xmath1cuo@xmath2 ( x=0.095 ) at ( 0.612,0.5,0 ) and ( x=0.08 ) at ( 0.5,0.607,0 ) , as well as that of b ) the ( 0,1,0 ) structural bragg peak , which marks the orthorhombic to low temperature tetragonal structural phase transition . note that a constant background has been subtracted in both cases .
the superconducting and structural phase transition temperatures are indicated by dashed lines for both samples .
all the data for x=0.08 have been scaled to the volume of the x=0.095 sample by phonon normalization .
the counting times refer to the unscaled x=0.095 data .
c ) temperature dependence of the elastic incommensurate magnetic scattering in la@xmath0ba@xmath1cuo@xmath2 ( x=0.095 ) in 0 and h=7 t @xmath32 c. ] we discuss the more extensive measurements on the la@xmath0ba@xmath1cuo@xmath2 sample with x=0.095 first .
elastic scattering scans at t=3.8 k are shown in fig 2a ) , in which bragg peaks occur at ( 0.5@xmath33,0.5,0 ) @xmath34 , indicating static incommensurate spin order .
analogous magnetic bragg peaks are observed at ( 0.5,[email protected],0 ) .
all peak widths are resolution limited with full widths at half maximum , fwhm=0.011 @xmath35 , indicating static spin correlations within the basal plane exceeding 180 .
by contrast , the spin correlations between planes are very short . to observe this ,
the crystal was reoriented in the ( h , h , l ) scattering plane and then tilted @xmath36 at constant l to intersect the incommensurate peak position for h=0.39 .
measurements along h of the form ( h , h/(1 - 2@xmath7),l ) at fixed l for l=2.5 and 3 are shown in fig .
2c at t=3.5 k and at l=3 for t=50 k to extract a background . since the peak intensity is independent of l , the scattering taking the form of an elastic rod along the l direction , the static spin order at low temperatures is two dimensional .
note that the intensity in fig .
2a ) is greater relative to that in fig .
2c ) for the sample oriented in the ( h , k,0 ) plane . this arises because the neutron spectrometer has a broad vertical resolution which integrates the signal in the l direction , perpendicular to the scattering plane .
the temperature dependence of the incommensurate magnetic elastic scattering is illustrated in fig .
the intensity of the magnetic bragg peak is proportional to the volume average of the square of the ordered staggered moment .
the spin order at ( 0.612,0.5,0 ) develops continuously with temperature below t@[email protected] k. the temperature dependence of the ( 0,1,0 ) structural bragg peak indicates the orthorhombic to low temperature tetragonal phase transition occurs at t@xmath29@xmath345 k , a transition which is discontinuous in nature @xcite . for reference , the superconducting transition at t@xmath8@xmath3 27 k ( see fig .
1 ) is also indicated on this plot as a dashed line .
the onset of spin ordering , t@xmath5 correlates most strongly with the completion of the transition to the low temperature tetragonal phase , and the incommensurate spin order coexists with the superconductivity below t@xmath8 . associated incommensurate charge ordering has not been observed .
the temperature dependence of the spin ordering is qualitatively similar to that observed in the x=1/8 compound @xcite , where the superlattice peak intensity becomes non - zero below @xmath38 k. similarly , no anomaly has been observed at t@xmath8 in ybco@xmath39 in the spin order , which has been attributed to robust spin correlations @xcite .
the most surprising result of this study is that the incommensurate spin structure shows no magnetic field dependence up to 7 t , applied vertically along the c * axis .
neither cooling nor warming the sample in a magnetic field has an effect on either the temperature dependence of the spin ordering , or the bragg intensity in la@xmath0ba@xmath1cuo@xmath2 ( x=0.095 ) , as shown in fig .
this result is in marked contrast with the behaviour of underdoped and optimally doped la@xmath0sr@xmath1cuo@xmath2 where pronounced field dependent effects are observed . for the optimally doped la@xmath0sr@xmath1cuo@xmath2
compound ( x=0.163 ) , the application of a magnetic field enhances the dynamical spin susceptibility but does not induce static order @xcite .
most dramatically , in a slightly underdoped sample ( x=0.144 ) , khaykovich _ et al _
@xcite report the development of a static incommensurate spin structure above a critical field of 2.7 t. the authors therefore argue that la@xmath0sr@xmath1cuo@xmath2 ( x=0.144 ) may be tuned through a quantum critical point , at which there is a magnetic field induced transition between magnetically disordered and ordered phases . their results are interpreted in terms of a ginzburg - landau model due to demler _
@xcite , which assumes a microscopic competition between spin and superconducting order parameters .
the predicted magnetic intensity increases as @xmath40 @xcite , which is consistent with experiments on la@xmath0sr@xmath1cuo@xmath2 .
note that the intensity changes most rapidly with magnetic field at low fields on a scale set by @xmath41 . in the la@xmath0ba@xmath1cuo@xmath2 family of compounds ,
the lower critical field is h@xmath42 t , while the upper critical field h@xmath43 is in excess of 40 t @xcite .
the upper critical field is of the same order of magnitude in optimally doped la@xmath0sr@xmath1cuo@xmath2 @xcite .
thus an applied magnetic field of 7 t should be sufficiently large to see an effect in la@xmath0ba@xmath1cuo@xmath2 . for sufficiently underdoped la@xmath0sr@xmath1cuo@xmath2
( x=0.12,0.10 ) , the ordered magnetic moment associated with preexisting static spin order is enhanced on application of a magnetic field @xcite .
the spin order within the vortex state of la@xmath0sr@xmath1cuo@xmath2 ( x=0.10 ) @xcite indicates long in - plane correlation lengths , greater than both the superconducting coherence length and the intervortex spacing at 14.5 t. as the coherence length is a measure of the size of the vortices , lake _
et al _ argue the static magnetism must therefore reside beyond the extent of the vortices themselves @xcite .
whereas the la@xmath0ba@xmath1cuo@xmath2 ( x-0.095 ) correlation length for static spin order is similarly long , the underlying physics is clearly different and the spins appear to order independent of vortex creation .
( * q * , @xmath26=2.07 and 3.1 mev ) ; ( middle panel ) the incommensurability @xmath17 ; and ( lower panel ) the inverse correlation length @xmath44 .
the dashed lines indicate the superconducting ( t@xmath45 k ) and magnetic ( t@xmath5=39.5 @xmath60.3 k ) transition temperatures . ]
the magnetic excitations were studied in constant energy transfer scans performed through the incommensurate ordering wavevectors .
horizontal collimation sequences of 0.54@xmath16 - [email protected]@xmath16 - 1.2@xmath46 and 0.54@xmath16 - [email protected]@xmath16 - 2.4@xmath16 were used at energy transfers of 2.07 and 3.1 mev respectively , yielding corresponding energy resolutions of @xmath47 and @xmath48 mev fwhm . the representative scan along ( h,0.5,0 ) and @xmath26=2.07 mev at t=30 k in fig .
2b , shows that the low energy dynamic spin response peaks up at the same wavevector , ( [email protected],0.5,0 ) , as the static spin structure . at higher energy transfers
the signal declines rapidly .
the measured dynamic structure factor @xmath49 is related to the imaginary part of the dynamical susceptibility @xmath50 through the fluctuation - dissipation theorem . for quantitative analysis , the data have been fit to the resolution convolution of @xmath51^{-1}$ ] where the susceptibility @xcite is : @xmath52 and @xmath53 represents the four incommensurate wave vectors @xmath54 and @xmath55 .
this assumes the magnetic excitations consist of four rods of scattering running along the c@xmath56 axis .
the extracted temperature dependences of @xmath57 , @xmath17 and @xmath44 are plotted in figs .
4a , b and c respectively .
@xmath58 is proportional to the integral of @xmath50 over @xmath59 in the ( h , k,0 ) scattering plane ; @xmath17 is the incommensurability , while @xmath44 is the inverse of the static correlation length in the basal plane , defined as the peak half width at half maximum . for reference , both the spin ordering transition at t@xmath5@xmath3 [email protected] k , and the superconducting transition near t@xmath8@xmath3
27 k are indicated on this plot . at both 2.07 mev and 3.1 mev , the dynamical susceptibility , @xmath58 ,
increases continuously as the temperature is reduced below @xmath3 60 k , becoming roughly constant and non - zero below t@xmath8 @xmath3 27 k. this is similar to measurements in both overdoped la@xmath9cuo@xmath10 , where a levelling off of the dynamic incommensurate spin response has been reported below t@xmath8@xmath3 42 k @xcite and also in la@xmath0ba@xmath1cuo@xmath2 ( x=0.125 ) in the normal state @xcite . in the latter compound , as a function of frequency there is relatively little change in @xmath58 at low temperature ( 8 k ) , whereas it drops rapidly in the present x=0.095 sample .
as the temperature is raised , @xmath58 varies linearly with frequency at lower energy transfers below 10 mev in the x=0.125 sample for t@xmath60 k , whereas it declines with increasing @xmath61 in x=0.095 for all t@xmath62 k. note that the signal intensity has been corrected for the monitor sensitivity to higher order incident neutrons , as described in ref .
these low energy excitations have some characteristics of the spin waves observed in the parent compound la@xmath9cuo@xmath12 @xcite as one warms through the nel temperature , where instantaneous spin correlations with the character of the nel state persist into the paramagnetic regime @xcite .
the form of @xmath58 varies dramatically as a function of doping in the related la@xmath0sr@xmath1cuo@xmath2compounds . in optimally and
slightly overdoped la@xmath0sr@xmath1cuo@xmath2(x=0.15,0.18 )
@xcite there is a characteristic energy of @xmath63 mev below which the dynamic susceptibility is dramatically reduced in the superconducting state - the opening up of a spin gap . however , on the underdoped side of the superconducting dome there is finite spectral weight in the spin response at all low energy transfers @xcite .
the bottom two panels of fig 4 show that the incommensurability , @xmath17 and the inverse correlation length , @xmath44 , correlate most strongly with the disappearance of the static spin order near t@xmath5@xmath339 k , which is not surprising .
above t@xmath5 the increase in @xmath64 may indicate that stripe correlations are weakened by thermal fluctuations that broaden the hole distribution about antiphase domain boundaries .
= 0.107 in la@xmath65ba@xmath66cuo@xmath12 at t=8 k , along ( 0.5,k,0 ) . ]
qualitatively , the magnetic and superconducting properties of la@xmath0ba@xmath1cuo@xmath2 ( x=0.08 ) ( t@xmath8=29 k ) are very similar to those of the higher doped x=0.095 ( t@xmath8=27 k ) sample .
elastic neutron scattering measurements were carried out under the same conditions as described earlier , with the single crystal oriented with ( h , k , o ) in the horizontal scattering plane . the elastic scattering scans at t=8 k ( see fig .
5 ) show that the incommensurate wave vector has decreased from @xmath67 in the x=0.095 sample to 0.107(3 ) .
surprisingly , the magnetic scattering is roughly a factor of eight less intense ( fig .
3a ) . in figs .
3a ) and b ) the intensities have been scaled to the sample volume using the integrated intensity of an acoustic phonon measured near a strong nuclear bragg peak at ( 2,0.15,0 ) .
we do not understand the reduced intensity for x=0.08 , since extinction does not play a role .
the temperature dependence of the elastic magnetic signals at the incommensurate wavevectors ( 0.5,0.607,0 ) for la@xmath0ba@xmath1cuo@xmath2 ( x=0.08 ) and at ( 0.612,0.5,0 ) for la@xmath0ba@xmath1cuo@xmath2 ( x=0.095 ) are reproduced in fig .
6 , where the intensities have been scaled so that their functional form may be directly compared . as can be seen , both the temperature dependence of the order parameter and the phase transition temperatures are very similar despite the difference in the strength of the elastic magnetic bragg scattering .
we therefore conclude the two electronic energy scales for these crystals at x=0.08 and x=0.095 , set by the superconducting t@xmath8 and t@xmath5 , are surprisingly similar .
it is not clear why a @xmath68 decrease in @xmath7 should produce an eight - fold decrease in the spin bragg intensity .
a search revealed no additional magnetic intensity in diagonal directions . as in x=0.095 , no incommensurate peaks due to charge ordering were observed in the x=0.08 sample .
ba@xmath1cuo@xmath2 ( x=0.095 ) at ( 0.612,0.5,0 ) and ( x=0.08 ) at ( 0.5,0.607,0 ) .
the net intensity for x=0.08 has been multiplied by a factor of 7.7 to match that of x=0.095 . ]
whether magnetism and superconductivity coexist in the same microscopic regions of the cuo@xmath9 planes or are phase separated is a topical subject of research . the issue of microscopic spatial segregation has been examined using a combination of neutron scattering @xcite and @xmath69sr @xcite techniques in lacuo@xmath10 . as a local probe
@xmath69sr is sensitive to heterogeneous structures .
the magnetic ordering in lacuo@xmath10 ( y=0.11 ) and la@xmath0sr@xmath1cuo@xmath2 ( x=0.12 ) has been reported to occur in reduced magnetic volume fractions of 40 and 18@xmath70 respectively @xcite .
et al _ argue an applied magnetic field enhances spin ordering primarily in the nonmagnetic regions @xcite , consistent with the above observations .
by contrast , the magnetic volume fraction in la@xmath0ba@xmath1cuo@xmath2 ( x=0.095 ) is much larger , approaching 100@xmath71 @xcite .
we speculate that no magnetic field dependence has been observed in la@xmath0ba@xmath1cuo@xmath2 x=0.095 because the non - magnetic volume fraction is too low .
a systematic study of the variation of the spin ordering with magnetic field is therefore of interest , with emphasis on the correlations between this effect and the magnetic volume fraction .
an interesting difference between the x=0.08 and the 0.095 la@xmath0ba@xmath1cuo@xmath2 samples is that the spin ordered state in the x=0.095 sample grows within a fully developed ltt structure , as t@[email protected] k and t@xmath29@xmath3 45 k. by contrast , in the x=0.08 la@xmath0ba@xmath1cuo@xmath2 sample , the mto to ltt structural phase transition begins near t@xmath5 on decreasing temperature and is only completed at temperatures below @xmath3 20 k. the situation for x=0.125 la@xmath0ba@xmath1cuo@xmath2 is similar to the x=0.095 case , as t@xmath5@xmath3 50 k , at which temperature the mto - ltt transition for x=0.125 is largely complete .
the first order nature of the mto - ltt structural phase transition implies co - existing structures over the temperature regime at which the spin order forms for x=0.08 la@xmath0ba@xmath1cuo@xmath2 .
it is then possible that the resulting structural heterogeneity interferes with the full development of spin order , giving rise to a substantially reduced magnetic bragg intensity as compared with the x=0.095 sample .
however , we also note that variability@xcite in the elastic magnetic bragg intensity has been reported from la@xmath0sr@xmath1cuo@xmath2 sample to sample with similar nominal doping levels of x@xmath3 0.1 and the la@xmath0sr@xmath1cuo@xmath2 system does not display the ltt phase at low temperatures .
is plotted vs elastic incommensuration , @xmath17 , for the x=0.08 , 0.095 , and 0.125 la@xmath0ba@xmath1cuo@xmath2 samples , compared with those measured in related la@xmath0sr@xmath1cuo@xmath2 samples .
b ) incommensurability @xmath17 is plotted vs ba - content , x , for our x=0.08 , x=0.095 la@xmath0ba@xmath1cuo@xmath2 samples , with previous x=0.125 la@xmath0ba@xmath1cuo@xmath2 results @xcite and those relevant to several underdoped la@xmath0sr@xmath1cuo@xmath2 samples @xcite .
the dotted line is a guide to the eye . ]
it is also interesting to examine the correlation between ba - content , x , incommensurability , @xmath17 , and superconducting t@xmath8 in la@xmath0ba@xmath1cuo@xmath2 and compare these relationships to those reported for la@xmath0sr@xmath1cuo@xmath2 .
the top panel of fig .
7 shows t@xmath8 vs @xmath17 for the x=0.08 , 0.095 , and 0.125 la@xmath0ba@xmath1cuo@xmath2 samples , compared with those measured in related la@xmath0sr@xmath1cuo@xmath2 samples .
as can be seen , our results for the x=0.08 and x=0.095 samples give the same t@xmath8 as la@xmath0sr@xmath1cuo@xmath2 for the same incommensuration , @xmath17 .
of course , la@xmath0ba@xmath1cuo@xmath2 displays the pronounced x=1/8 anomaly and consequently the x=0.125 point on this t@xmath8 vs @xmath17 plot lies far below the linear curve which is an excellent descriptor for the remainder of the underdoped la@xmath0ba@xmath1cuo@xmath2 and la@xmath0sr@xmath1cuo@xmath2 systems .
the abrupt nature of the x=1/8 anomaly in the lbco system is clear . the bottom panel of fig .
7 shows the incommensurability @xmath17 vs ba - content , x and once again we compare our new results on x=0.08 and x=0.095 la@xmath0ba@xmath1cuo@xmath2 samples with the x=0.125 la@xmath0ba@xmath1cuo@xmath2 results and those of several underdoped la@xmath0sr@xmath1cuo@xmath2 samples at sr concentrations above x=0.05 , where the incommensurate spin ordering is consistent with a picture of parallel ( as opposed to diagonal ) stripe ordering . in this sr - content regime , @xmath17 tracks x well , assuming stoichiometric oxygen content @xcite .
one can see that the incommensuration in x=0.08 and x=0.095 la@xmath0ba@xmath1cuo@xmath2 shows relatively little x - dependence .
indeed , we observe @xmath17 values which are only @xmath3 9 @xmath70 less than that displayed by x=0.125 la@xmath0ba@xmath1cuo@xmath2 , thereby departing significantly from the approximately linear @xmath17 vs x relation characterizing the underdoped la@xmath0sr@xmath1cuo@xmath2 studies .
it is possible that this difference between underdoped la@xmath0ba@xmath1cuo@xmath2 and la@xmath0sr@xmath1cuo@xmath2 also arises due to the mto - ltt structural phase transition , occurring in la@xmath0ba@xmath1cuo@xmath2 but absent in la@xmath0sr@xmath1cuo@xmath2 .
it is also conceivable that it arises due to some small oxygen off - stoichiometry , such that our samples have the composition la@xmath0ba@xmath1cuo@xmath10 , with oxygen stoichiometry greater than 4 .
such excess oxygen would give rise to an effective hole doping given by x@xmath72 = x+2y . to bring the @xmath17 values for x=0.08 and 0.095 back onto the linear relationship between @xmath17 and x@xmath72 seen in la@xmath0sr@xmath1cuo@xmath2 , small , but positive values of y : 0.013 and 0.0075 for the x=0.08 and x=0.095 la@xmath0ba@xmath1cuo@xmath10
samples respectively would be required .
this is too small to be detectable and runs counter to what is concluded in la@xmath0sr@xmath1cuo@xmath10 . in underdoped la@xmath0sr@xmath1cuo@xmath10
, the superconducting t@xmath8 is maximized by annealing in oxygen , at which point the measured @xmath17 vs sr concentration , x , lie on the straight line @xcite . consequently ,
as grown la@xmath0sr@xmath1cuo@xmath10 tends to be oxygen deficient ( y@xmath730 ) and annealing in oxygen results in stoichiometric la@xmath0sr@xmath1cuo@xmath2 .
this is also expected to be true for underdoped la@xmath0ba@xmath1cuo@xmath10 , which would imply that the deviation of @xmath17 vs x from a linear relationship is intrinsic to stoichiometric la@xmath0ba@xmath1cuo@xmath2 , a surprising result .
we have observed the coexistence of static , two dimensional incommensurate spin order and superconductivity in la@xmath0ba@xmath1cuo@xmath12 with x=0.095 and x=0.08 .
this result is in broad agreement with other well studied high temperature la-214 cuprate superconductors , such as la@xmath0sr@xmath74cuo@xmath12 or la@xmath9cuo@xmath10 , which show signatures of either incommensurate static spin ordering or dynamic spin correlations .
one significant finding of this study is the field _ independence _ of the incommensurate magnetic order in the x=0.095 sample , in marked contrast with other superconducting la-214 cuprates .
studies of the spin ordering as a function of magnetic field in other superconducting systems with large magnetic volume fractions should prove illuminating .
in addition , while the dependence of the superconducting t@xmath8 on the incommensuration of the magnetic structure , @xmath17 , in la@xmath0ba@xmath1cuo@xmath2(x=0.08 , 0.095 ) is the same as that observed in la@xmath0sr@xmath1cuo@xmath2 , the x - dependence appears to be substantially weaker than that seen in la@xmath0sr@xmath1cuo@xmath2 , where a linear relationship is observed over this range of concentration . now that crystal growth breakthroughs have resulted in the availability of large , high quality single crystals of la@xmath0ba@xmath1cuo@xmath2 , fuller experimental characterization of the original family of high temperature superconductors is clearly warranted .
this work was supported by nserc of canada .
it is a pleasure to acknowledge the contributions of a. kallin , a. dabkowski and e. mazurek at mcmaster , who were involved in the single crystal growth and characterization and the expert technical support of l. mcewan and r. sammon at nrc , chalk river .
4 m. a. kastner _
et al _ , rev mod phys * 70 * , 897 ( 1998 ) . j. m. tranquada , cond - mat/0512115 ( 2005 ) .
birgeneau , c.stock , j.m .
tranquada , k.yamada , j phys soc jpn ( 2006 ) . a. h. castro neto and c. morais smith , in _ strong interactions in low dimensions _ , edited by d. baeriswyl and l. degiorgi ( kluwer academic publishing , 2004 ) .
s. mori , c.h .
chen , s.w .
cheong , nature * 392 * , 473 ( 1998 ) ; phys .
lett . * 81 * , 3972 ( 1998 ) ; c. renner _
el al _ , nature * 416 * , 518 ( 2002 ) . j. m. tranquada , d.j .
buttrey , v. sachan , and j.e .
lorenzo , phys .
lett . * 73 * , 1003 ( 1994 ) ; s - h .
lee and s - w .
cheong , phys .
lett . * 79 * , 2514 ( 1997 ) ; h. yoshizawa _ et al _ , phys .
b * 61 * , r854 ( 2000 ) .
et al _ , phys .
lett . * 84 * , 3911 ( 2000 ) .
y. yoshinari , p.c .
hammel , and s.w .
cheong , phys .
* 82 * , 3536 ( 1999 ) .
t. katsufuji , t. tanabe , t. ishikawa , y. fukuda , t. arima , and y. tokura , phys .
b * 54 * , r14230 ( 1996 ) ; g. blumberg , m.v . klein , and s.w .
cheong , phys .
lett . * 80 * , 564 ( 1998 ) ; k. yamamoto , t. katsufuji , t. tanabe , and y. tokura , phys .
* 80 * , 1493 ( 1998 ) .
n. bulut , d. hone , d. j. scalapino , n .
e. bickers , phys rev lett * 64 * , 2723 ( 1990 ) .
n. bulut , d. j. scalapino , phys rev b * 53 * , 5149 ( 1996 ) .
m. r. norman , phys rev b * 61 * , 14751 ( 2000 ) .
s. a. kivelson _
et al _ , rev mod phys * 75 * , 1201 ( 2003 ) .
d. vaknin _ et al _ , phys rev lett * 58 * , 2802 ( 1987 ) .
et al _ , phys rev lett * 60 * , 156 ( 1988 ) k. yamada _
et al _ , phys .
b * 57 * , 6165 ( 1998 ) ; h. kimura _ et al _ , phys . rev .
b * 59 * , 6517 ( 1999 ) . c. stock , w.j.l . buyers , r.liang , d. peets , z. tun , d. bonn , w.n .
hardy , and r.j .
birgeneau , phys .
b , * 69 * , 014502 ( 2004 ) . c. stock ,
w. j. l. buyers , z. yamani , c. l. broholm , j. -h .
chung , z. tun , r. liang , d. bonn , w. n. hardy , r. j. birgeneau , phys .
b * 73 * , 100504(r ) ( 2006 ) . m. matsuda _
et al _ , phys .
b * 62 * , 9148 ( 2000 ) ; s. wakimoto _ et al _ , phys
b * 61 * , 3699 ( 2000 ) .
t. suzuki , t. goto , k. chiba , t. shinoda , t. fukase , h. kimura , k. yamada , m. ohashi , y. yamaguchi , phys rev b * 57 * , r3229 ( 1998 ) .
h. kimura , k. hirota , h. matsushita , k. yamada , y. endoh , s - h lee , c. f. majkrzak , r. erwin , g. shirane , m. greven , y. s. lee , m. a. kastner and r. j. birgeneau , phys rev b * 59 * , 6517 ( 1999 ) . s. wakimoto _
et al _ , phys rev lett , * 92 * 217004 ( 2004 ) .
a. lucarelli _
et al _ , phys .
lett . * 90 * , 037002 ( 2003 ) .
j. m. tranquada , b. j. sternlieb , j. d. axe , y. nakamura , s. uchida , nature * 375 * , 561 ( 1995 ) .
j. m. tranquada , j. d. axe , n. ichikawa , a. r. moodenbaugh , y. nakamura , s. uchida , phys rev lett * 78 * , 338 ( 1997 ) . m. fujita , h. goka , k. yamada , m. matsuda , phys rev b , * 66 * 184503 ( 2002 ) . j.g . bednorz and k.a .
muller , z. phys b * 64 * , 189 ( 1986 ) .
t. adachi , t. noji , and y. koike , phys rev b * 64 * , 144524 ( 2001 ) .
j.m tranquada , h. woo , t.g .
perring , h. goka , g.d .
gu , g. xu , m. fujita , and k. yamada , nature , * 429 * , 534 ( 2004 ) .
m. fujita , h. goka , k. yamada , j. m. tranquada , l. p. regnault , phys rev b * 70 * , 104517 ( 2004 ) .
a. r. moodenbaugh , y. xu , m. suenaga , t. folkerts , and r. shelton , phys rev b * 38 * , 4596 ( 1988 ) . y. zhao _
et al _ , arxiv:0707.2799 ( 2007 ) .
k. yamada , c. h. lee , k. kurahashi , j. wada , s. wakimoto , s. ueki , h. kimura , y. endoh , s .
hosoya , g. shirane , r. j. birgeneau , m. greven , m. a. kastner , and y. j. kim , phys rev b * 57 * , 6165 ( 1998 ) . y. s. lee , f. c. chou , a. tewary , m. a. kastner , s. h. lee and r. j. birgeneau , phys . rev .
b * 69 * , 020502(r ) ( 2004 ) j. d. axe _ et al _ , phys rev lett * 62 * , 2751 ( 1989 ) . b. lake et al , science * 291 * , 1759 ( 2001 ) .
b. khaykovich , s. wakimoto , r. j. birgeneau , m. a. kastner , y. s. lee , p. smeibidl , p. vorderwisch , k. yamada , phys rev b * 71 * , 220508 ( 2005 ) .
e. demler , s. sachdev , y. zhang , phys rev lett * 87 * , 067202 ( 2001 ) .
h. takagi et al , j. j. applied phys . * 26 * , l434 ( 1987 ) . g. s. boebinger , y. ando , a. passner , t. kimura , m. okuya , j. shimoyama , k. kishio , k. tamasaku , n. ichikawa and s. uchida , phys .
lett . * 77 * , 5417 ( 1996 ) .
s. katano , m. sato , k. yamada , t. suzuki , t. fukase , phys rev b * 62 * , r14677 ( 2000 ) .
b. lake et al , nature * 415 * , 299 ( 2002 ) .
b. lake , k. lefmann , n. b. christensen , g. aeppli , d. f. mcmorrow , h. m. ronnow , p. vorderwisch , p. smeibidl , n. mangkorntong , t. sasagawa , m. nohara , h. takagi , nature materials * 4 * , 658 ( 2005 ) .
b. khaykovich , y. s. lee , r. w. erwin , s .- h .
lee , s. wakimoto , k. j. thomas , m. a. kastner and r. j. birgeneau , phys rev b * 66 * , 014528 ( 2002 ) . y. s. lee , r. j. birgeneau , m. a. kastner , y. endoh , s. wakimoto , k. yamada , r. w. erwin , s .- h .
lee , g. shirane , phys . rev .
b * 60 * , 3643 ( 1999 ) .
k. yamada , k. kakurai , y. endoh , t. r. thurston , m. a. kastner , r. j. birgeneau , g. shirane , y. hidaka , t. murakami , phys rev b * 40 * , 4557 ( 1989 ) .
g. shirane , y. endoh , r. j. birgeneau , m. a. kastner , y. hidaka , m. oda , m. suzuki , and t. murakami , phys .
* 59 * , 1613 ( 1987 ) .
k. yamada , s. wakimoto , g. shirane , c. h. lee , m. a. kastner , s. hosoya , m. greven , y. endoh , r. j. birgeneau , phys rev lett * 75 * , 1626 ( 1995 ) . k. yamada , c. h. lee , y. endoh , g. shirane , r. j. birgeneau , m. a. kastner , physica c * 282 - 287 * , 85 ( 1997 ) chul - ho lee , kazuyoshi yamada , yasuo endoh , gen shirane , r. j. birgeneau , m. a. kastner , m. greven , y - j .
kim , j. phys .
soc jpn * 69 * , 1170 ( 2000 ) .
haruhiro hiraka , yasuo endoh , masaki fujita , young s. lee , jiri kulda , alexandre ivanov , robert j. birgeneau , j. phys soc jpn * 70 * , 853 ( 2001 ) .
m. fujita , k. yamada , h. hiraka , p. m. gehring , s. h. lee , s. wakimoto and g. shirane , phys rev b * 65 * , 064505 ( 2002 ) .
a. t. savici , y. fudamoto , i. m. gat , t. ito , m. i. larkin , y. j. uemura , g. m. luke , k. m. kojima , y. s. lee , m. a. kastner , r. j. birgeneau , k. yamada , phys rev b * 66 * , 014524 ( 2002 ) . s. r. dunsiger _ et al _ , ( unpublished ) . | using neutron scattering techniques , we have studied incommensurate spin ordering as well as low energy spin dynamics in single crystal underdoped la@xmath0ba@xmath1cuo@xmath2 with [email protected] and 0.08 ; high temperature superconductors with t@xmath4 27 k and 29 k respectively .
static two dimensional incommensurate magnetic order appears below t@xmath5=39.5 @xmath6 0.3 k in la@xmath0ba@xmath1cuo@xmath2(x=0.095 ) and a similar temperature for x=0.08 within the low temperature tetragonal phase .
the spin order is unaffected by either the onset of superconductivity or the application of magnetic fields of up to 7 tesla applied along the c - axis in the x=0.095 sample . such magnetic field _
independent _ behaviour is in marked contrast with the field induced enhancement of the staggered magnetisation observed in the related la@xmath0sr@xmath1cuo@xmath2 system , indicating this phenomenon is not a universal property of cuprate superconductors .
surprisingly , we find that incommensurability @xmath7 is only weakly dependent on doping relative to la@xmath0sr@xmath1cuo@xmath2 .
dispersive excitations in la@xmath0ba@xmath1cuo@xmath2 ( x=0.095 ) at the same incommensurate wavevector persist up to at least 60 k. the dynamical spin susceptibility of the low energy spin excitations saturates below t@xmath8 , in a similar manner to that seen in the superconducting state of la@xmath9cuo@xmath10 .
the roles of spin , charge and lattice degrees of freedom have been central to the rich behaviour brought to light over the last 20 years in superconducting lamellar copper oxides @xcite .
in particular , the cuprates exhibit phenomena which are a sensitive function of doping , evolving from an antiferromagnetic insulating parent compound into a superconducting phase with increasing hole density . a heterogeneous electronic phase composed of stripes of itinerant charges now appears to be a generic feature of hole doped ternary transition metal oxides @xcite such as manganites @xcite and nickelates @xcite , as well as cuprates .
the explanation for these incommensurate spin ordered states is the subject of ongoing debate . in an itinerant picture ,
the spin dynamics are described in terms of electron - hole pair excitations about an underlying fermi surface @xcite .
alternatively , within the stripe picture of doped , two dimensional mott insulators , the non - magnetic holes in these materials organize into quasi - one dimensional stripes which separate antiferromagnetic insulating antiphase domains @xcite .
adjacent antiferromagnetic regions are @xmath11 out of phase with each other giving rise to a magnetic structure with incommensurate periodicity , where the supercell dimension is twice the hole stripe periodicity .
the static spin structures in the undoped , parent compounds such as la@xmath9cuo@xmath12 @xcite or yba@xmath9cu@xmath13o@xmath14 @xcite have been determined by neutron scattering to be relatively simple two sublattice antiferromagnets characterized by a commensurate ordering wavevector of ( 0.5,0.5 ) in reciprocal lattice units within the tetragonal basal plane . on hole doping with either sr substituting for la in la@xmath0sr@xmath1cuo@xmath12 @xcite or by adding additional oxygen in yba@xmath9cu@xmath13o@xmath15 @xcite , the magnetic scattering moves out to incommensurate wavevectors , consistent with the stripe ordering picture described above .
this incommensurate magnetism can be either static or dynamic , as evidenced by either elastic or inelastic peaks in the neutron scattering respectively and now appears to be a common feature of the la@xmath0sr@xmath1cuo@xmath2 family of compounds .
specifically , for lightly doped la@xmath0sr@xmath1cuo@xmath12 , elastic incommensurate magnetic bragg features first appear split off from the ( 0.5,0.5 ) position in diagonal directions relative to a tetragonal unit cell @xcite . at higher doping in the underdoped superconducting regime ,
the peaks rotate by 45@xmath16 to lie along directions parallel to the tetragonal axes or cu - o - cu bonds , such that elastic magnetic scattering appears at ( 0.5@xmath6@xmath17,0.5,0 ) and ( 0.5,0.5@xmath6@xmath17,0 ) @xcite . for optimal and higher doping
the static order disappears , but dynamic incommensurate correlations nevertheless persist@xcite . within the stripe picture ,
one expects charge ordering associated with the holes to occur at an incommensurate wavevector @xmath18 , twice that describing the spin order .
neutron scattering is not directly sensitive to charge ordering _ per se _ , but it is sensitive to atomic displacements , such as those associated with oxygen , which arise from charge ordering . an incommensurate nuclear scattering signature is therefore expected to appear at ( 2@xmath62@xmath17,0,0 ) or ( 2,@xmath19,0 ) and related wavevectors . despite extensive efforts , such incommensurate charge related scattering has not been observed in la@xmath0sr@xmath1cuo@xmath12 by either neutron or x - ray scattering techniques , although there is indirect evidence of charge stripe excitations from optical measurements @xcite .
such scattering _ has _ been observed in la@xmath20nd@xmath21sr@xmath1cuo@xmath12 @xcite , as well as la@xmath22ba@xmath23sr@xmath1cuo@xmath12 ( x=0,0.05,0.06,0.075,0.085 ) @xcite , motivating discussion as to whether such static charge stripes compete with , rather than underlie , high temperature superconductivity .
ba@xmath1cuo@xmath2(x=0.095 and 0.08 ) single crystals measured at 0.001 t. dashed lines indicate the onset of superconductivity at t@xmath8=27 and 29 k for x=0.095 and x=0.08 samples respectively . ]
= 0.112 in la@xmath24ba@xmath25cuo@xmath12 at t=3.8 k , along ( h,0.5,0 ) .
b ) representative inelastic scans at t=30 k , also along ( h , 0.5 , 0 ) and at @xmath26= 2.07 mev . parameters characterizing this inelastic scattering
are shown in fig .
4 , while the solid line is discussed in the text .
c ) elastic scans of the form ( h , h/(1 - 2@xmath7 ) , l ) , which demonstrate the rod - like , two dimensional nature of the elastic magnetic scattering , as described in the text .
the scan at l=2.5 has been displaced by 50 counts upwards for clarity . ]
surprisingly , la@xmath0ba@xmath1cuo@xmath12 , the first high temperature superconductor ( htsc ) to be discovered @xcite , has been much less extensively studied than either la@xmath0sr@xmath1cuo@xmath2 or yba@xmath9cu@xmath13o@xmath15 due to the difficulty of growing single crystals , which has only been achieved@xcite recently . in this paper , we report neutron scattering signatures of static incommensurate spin order in single crystal la@xmath0ba@xmath1cuo@xmath2 ( x=0.095 , 0.08 ) , consistent with the stripe picture described above , but with interesting complexity not accounted for within present theoretical models . our results clearly indicate that the spin ordering is insensitive to both the onset of superconductivity and surprisingly the application of a magnetic field . tranquada _
et al _
@xcite and fujita _ et al _ @xcite have reported neutron scattering measurements on an x=0.125 sample of la@xmath0ba@xmath1cuo@xmath12 . in both
la@xmath0ba@xmath1cuo@xmath2 and la@xmath0sr@xmath1cuo@xmath2 this concentration corresponds to a suppression of t@xmath8 as a function of doping , known as the 1/8 " anomaly . in la@xmath0ba@xmath1cuo@xmath2
the suppression is almost complete @xcite and is associated with a structural phase transition at low temperature , from orthorhombic to tetragonal @xcite which gives rise to a superlattice peak at ( 0,1,0 ) and symmetry related reflections .
samples of la@xmath0ba@xmath1cuo@xmath2 near x=0.125 display a sequence of structures on lowering the temperature , going progressively from high temperature tetragonal ( htt , i4/mmm symmetry ) to orthorhombic ( mto , bmab symmetry ) to low temperature tetragonal ( ltt , p4@xmath9/ncm symmetry ) @xcite .
the high temperature tetragonal to orthorhombic transition in particular , and to a lesser extent , the orthorhombic to low temperature tetragonal transition are sensitive indicators for the precise ba doping level in the material . |
the motivation for ultra - relativistic heavy - ion experiments at the relativistic heavy ion collider ( rhic ) at brookhaven national laboratory is the study of nuclear matter at extremely high temperature and energy density with the hope of creating and detecting deconfined matter consisting of quarks and gluons the quark gluon plasma ( qgp ) .
lattice qcd calculations @xcite predict that the transition to a deconfined state occurs at a critical temperature @xmath14 170 mev and an energy density @xmath15 2 gev/@xmath16 . based on the bjorken estimation @xcite and the measurement of transverse energy ( @xmath17 ) in au+au collisions at @xmath4 = 130 gev @xcite and 200 gev , the spatial energy density in central au+au collisions at rhic is believed to be high enough to create such deconfined matter in a laboratory @xcite .
the hot and dense matter produced in relativistic heavy ion collisions may evolve through the following scenario : pre - equilibrium , thermal ( or chemical ) equilibrium of partons , possible formation of qgp or a qgp - hadron gas mixed state , a gas of hot interacting hadrons , and finally , a freeze - out state when the produced hadrons no longer strongly interact with each other .
since produced hadrons carry information about the collision dynamics and the entire space - time evolution of the system from the initial to the final stage of collisions , a precise measure of the transverse momentum ( @xmath7 ) distributions and yields of identified hadrons as a function of collision geometry is essential for the understanding of the dynamics and properties of the created matter . in the low @xmath7 region ( @xmath18 2 gev/@xmath6 ) , hydrodynamic models @xcite that include radial flow
successfully describe the measured @xmath7 distributions in au+au collisions at @xmath4 = 130 gev @xcite .
the @xmath7 spectra of identified charged hadrons below @xmath19 2 gev/@xmath6 in central collisions have been well reproduced by two simple parameters : transverse flow velocity @xmath20 and freeze - out temperature @xmath21 @xcite under the assumption of thermalization with longitudinal and transverse flow @xcite . the particle production in this @xmath7 region
is considered to be dominated by secondary interactions among produced hadrons and participating nucleons in the reaction zone .
another model which successfully describes the particle abundances at low @xmath7 is the statistical thermal model @xcite .
particle ratios have been shown to be well reproduced by two parameters : a baryon chemical potential @xmath22 and a chemical freeze - out temperature @xmath23 .
it is found that there is an overall good agreement between measured particle ratios at @xmath4 = 130 gev au+au and the thermal model calculations @xcite . on the other hand , at high @xmath7 ( @xmath24 4 gev/@xmath6 )
the dominant particle production mechanism is the hard scattering described by perturbative quantum chromodynamics ( pqcd ) , which produces particles from the fragmentation of energetic partons .
one of the most interesting observations at rhic is that the yield of high @xmath7 neutral pions and non - identified charged hadrons in central au+au collisions at rhic are below the expectation of the scaling with the number of nucleon - nucleon binary collisions , @xmath25 @xcite .
this effect could be a consequence of the energy loss suffered by partons moving through deconfined matter @xcite .
it has also been observed that the yield of neutral pions is more strongly suppressed than that for non - identified charged hadrons @xcite in central au+au collisions at rhic .
another interesting feature is that the proton and anti - proton yields in central events are comparable to that of pions at @xmath19 2 gev/@xmath6 @xcite , differing from the expectation of pqcd .
these observations suggest that a detailed study of particle composition at intermediate @xmath7 ( @xmath26 2 4 gev/@xmath6 ) is very important to understand hadron production and collision dynamics at rhic .
the phenix experiment @xcite has a unique hadron identification capability in a broad momentum range .
pions and kaons are identified up to 3 gev/@xmath6 and 2 gev/@xmath6 in @xmath7 , respectively , and protons and anti - protons can be identified up to 4.5 gev/@xmath6 by using a high resolution time - of - flight detector @xcite .
neutral pions are reconstructed via @xmath27 up to @xmath19 10 gev/@xmath6 through an invariant mass analysis of @xmath28 pairs detected in an electro - magnetic calorimeter ( emcal ) @xcite with wide azimuthal coverage . during the measurements of au+au collisions at @xmath4 = 200 gev in year 2001 at rhic ,
the phenix experiment accumulated enough events to address the above issues at intermediate @xmath7 as well as the particle production at low @xmath7 with precise centrality dependences . in this paper , we present the centrality dependence of @xmath7 spectra , @xmath29 , yields , and ratios for @xmath0 , @xmath1 , @xmath2 and @xmath3 in au+au collisions at @xmath4 = 200 gev at mid - rapidity measured by the phenix experiment
. we also present results on the scaling behavior of charged hadrons compared with results of @xmath30 measurements @xcite , which have been published separately .
the paper is organized as follows .
section [ sec : phenix ] describes the phenix detector used in this analysis . in section [ sec : analysis ] the analysis details including event selection , track selection , particle identification , and corrections applied to the data are described .
the systematic errors on the measurements are also discussed in this section . for the experimental results ,
centrality dependence of @xmath7 spectra for identified charged particles are presented in section [ sec : pt_spectra ] , and transverse mass spectra are given in section [ sec : mt_spectra ] .
particle yields and mean transverse momenta as a function of centrality are presented in section [ sec : meanpt_dndy ] . in section [ sec :
ratio ] the systematic study of particle ratios as a function @xmath7 and centrality are presented .
section [ sec : ncoll ] studies the scaling behavior of identified charged hadrons .
a summary is given in section [ sec : summary ] .
the phenix experiment is composed of two central arms , two forward muon arms , and three global detectors .
the east and west central arms are placed at zero rapidity and designed to detect electrons , photons and charged hadrons .
the north and south forward muon arms have full azimuthal coverage and are designed to detect muons .
the global detectors measure the start time , vertex , and multiplicity of the interactions .
the following sections describe the parts of the detector that are used in the present analysis . a detailed description of the complete detector can be found elsewhere @xcite . in order to characterize the centrality of au+au collisions , zero - degree calorimeters ( zdc )
@xcite and beam - beam counters ( bbc ) @xcite are employed .
the zero - degree calorimeters are small hadronic calorimeters which measure the energy carried by spectator neutrons .
they are placed 18 m up- and downstream of the interaction point along the beam line .
each zdc consists of three modules .
each module has a depth of 2 hadronic interaction lengths and is read out by a single photo - multiplier tube ( pmt ) .
both time and amplitude are digitized for each pmt along with the analog sum of the three pmt signals for each zdc .
two sets of beam - beam counters are placed 1.44 m from the nominal interaction point along the beam line ( one on each side ) .
each counter consists of 64 @xmath31erenkov telescopes , arranged radially around the beam line .
the bbc measures the number of charged particles in the pseudo - rapidity region 3.0 @xmath18 @xmath32 @xmath18 3.9 .
the correlation between bbc charge sum and zdc total energy is used for centrality determination .
the bbc also provides a collision vertex position and start time information for time - of - flight measurement .
charged particles are tracked using the central arm spectrometers @xcite .
the spectrometer on the east side of the phenix detector ( east arm ) contains the following subsystems used in this analysis : drift chamber ( dc ) , pad chamber ( pc ) and time - of - flight ( tof ) .
the drift chambers are the closest tracking detectors to the beam line at a radial distance of 2.2 m. they measure charged particle trajectories in the azimuthal direction to determine the transverse momentum of each particle . by combining the polar angle information from the first layer of the pc with the transverse momentum ,
the total momentum @xmath2 is determined .
the momentum resolution is @xmath33 ( gev/@xmath6 ) , where the first term is due to the multiple scattering before the dc and the second term is the angular resolution of the dc .
the momentum scale is known to 0.7% , from the reconstructed proton mass using the tof .
the pad chambers are multi - wire proportional chambers that form three separate layers of the central tracking system . the first pad chamber layer ( pc1 )
is located at the radial outer edge of each drift chamber at a distance of 2.49 m , while the third layer ( pc3 ) is 4.98 m from the interaction point . the second layer ( pc2 )
is located at a radial distance of 4.19 m in the west arm only .
pc1 and the dc , along with the vertex position measured by the bbc , are used in the global track reconstruction to determine the polar angle of each charged track .
the time - of - flight detector serves as the primary particle identification device for charged hadrons by measuring the stop time . the start time is given by the bbc .
the tof wall is located at a radial distance of 5.06 m from the interaction point in the east central arm .
this contains 960 scintillator slats oriented along the azimuthal direction .
it is designed to cover @xmath34 and @xmath35 in azimuthal angle .
the intrinsic timing resolution is @xmath36 ps , which allows for a 3@xmath37 @xmath38 separation up to @xmath39 gev/@xmath6 , and 3@xmath37 @xmath40 separation up to @xmath41 gev/@xmath6 .
in this section , we describe the event and track selection , charged particle identification and various corrections , including geometrical acceptance , particle decay , multiple scattering and absorption effects , detector occupancy corrections and weak decay contributions from @xmath42 and @xmath43 to proton and anti - proton spectra .
the estimations of systematic uncertainties on the measurements are addressed at the end of this section . for the present analysis
, we use the phenix minimum bias trigger events , which are determined by a coincidence between north and south bbc signals .
we also require a collision vertex within @xmath44 30 cm from the center of the spectrometer .
the collision vertex resolution determined by the bbc is about 6 mm in au+au collisions in minimum bias events @xcite .
the phenix minimum bias trigger events include @xmath45% of the 6.9 barn au+au total inelastic cross section @xcite .
figure [ fig : bbc_zdc ] shows the correlation between the bbc charge sum and zdc total energy for au+au at @xmath4 = 200 gev .
the lines on the plot indicate the centrality definition in the analysis . for the centrality determination ,
these events are subdivided into 11 bins using the bbc and zdc correlation : 05% , 510% , 1015% , 1520% , 2030% , ... , 7080% and 8092% .
due to the statistical limitations in the peripheral events , we also use the 6092% centrality bin as the most peripheral bin .
after event selection , we analyze 2.02@xmath46 minimum bias events , which represents @xmath5 140 times more events than used in our published au+au data at 130 gev @xcite . based on a glauber model calculation
@xcite we use two global quantities to characterize the event centrality : the average number of participants @xmath47 and the average number of collisions @xmath48 associated with each centrality bin ( table [ tab : ncoll ] ) .
.the average nuclear overlap function ( @xmath49 ) , the number of nucleon - nucleon binary collisions ( @xmath48 ) , and the number of participant nucleons ( @xmath47 ) obtained from a glauber monte carlo @xcite correlated with the bbc and zdc response for au+au at @xmath4 = 200 gev as a function of centrality .
centrality is expressed as percentiles of @xmath50 = 6.9 barn with 0% representing the most central collisions .
the last line refers to minimum bias collisions . [ cols="^,^,^,^",options="header " , ] | the centrality dependence of transverse momentum distributions and yields for @xmath0 , @xmath1 , @xmath2 and @xmath3 in au+au collisions at @xmath4 = 200 gev at mid - rapidity are measured by the phenix experiment at rhic .
we observe a clear particle mass dependence of the shapes of transverse momentum spectra in central collisions below @xmath5 2 gev/@xmath6 in @xmath7 .
both mean transverse momenta and particle yields per participant pair increase from peripheral to mid - central and saturate at the most central collisions for all particle species .
we also measure particle ratios of @xmath8 , @xmath9 , @xmath10 , @xmath11 , @xmath12 and @xmath13 as a function of @xmath7 and collision centrality .
the ratios of equal mass particle yields are independent of @xmath7 and centrality within the experimental uncertainties . in central collisions at intermediate transverse momenta @xmath5 1.5 4.5 gev/@xmath6 ,
proton and anti - proton yields constitute a significant fraction of the charged hadron production and show a scaling behavior different from that of pions . |
a majority of computational homogenization algorithms rely on the solution of the unit cell problem , which concerns the determination of local fields in a representative sample of a heterogeneous material under periodic boundary conditions . currently , the most efficient numerical solvers of this problem are based on discretization of integral equations . in the case of particulate composites with smooth
bounded inclusions embedded in a matrix phase , the problem can be reduced to internal interfaces and solved with remarkable accuracy and efficiency by the fast multipole method , see ( * ? ? ? * and references therein ) .
an alternative method has been proposed by suquet and moulinec @xcite to treat problems with general microstructures supplied in the form of digital images .
the algorithm is based on the neumann series expansion of the inverse to an operator arising in the associated lippmann - schwinger equation and exploits the fast fourier transform ( fft ) to evaluate the action of the operator efficiently .
the major disadvantage of the fft - based method consists in its poor convergence for composites exhibiting large jumps in material coefficients . to overcome this difficulty , eyre and milton proposed in @xcite an accelerated scheme derived from a modified integral equation treated by means of the series expansion approach .
in addition , michel et al .
@xcite introduced an equivalent saddle - point formulation solved by the augmented lagrangian method .
as clearly demonstrated in a numerical study by moulinec and suquet @xcite , both methods converge considerably faster than the original variant ; the number of iterations is proportional to the square root of the phase contrast instead of the linear increase for the basic scheme .
however , this comes at the expense of increased computational cost per iteration and the sensitivity of the augmented lagrangian algorithm to the setting of its internal parameters . in this short note
, we introduce yet another approach to improve the convergence of the original fft - based scheme @xcite based on the trigonometric collocation method @xcite and its application to the helmholtz equation as introduced by vainikko @xcite .
we observe that the discretization results in a system of linear equations with a structured dense matrix , for which a matrix - vector product can be computed efficiently using fft , cf .
section [ sec : methodology ] .
it is then natural to treat the resulting system by standard iterative solvers , such as the krylov subspace methods , instead of the series expansion technique . in section [ sec : results ]
, the potential of such approach is demonstrated by means of a numerical study comparing the performance of the original scheme and the conjugate- and biconjugate - gradient methods for two - dimensional scalar electrostatics .
in this section , we briefly summarize the essential steps of the trigonometric collocation - based solution to the unit cell problem by adapting the original exposition by vainikko @xcite to the setting of electrical conduction in periodic composites . in what follows ,
@xmath0 , @xmath1 and @xmath2 denote scalar , vector and second - order tensor quantities with greek subscripts used when referring to the corresponding components , e.g. @xmath3 .
matrices are denoted by a serif font ( e.g. @xmath4 ) and a multi - index notation is employed , in which @xmath5 with @xmath6 represents @xmath7 and @xmath8 stands for the @xmath9-th element of the matrix @xmath10 .
we consider a composite material represented by a periodic unit cell @xmath11 . in the context of linear electrostatics ,
the associated unit cell problem reads as @xmath12 where @xmath13 is a @xmath14-periodic vectorial electric field , @xmath15 denotes the corresponding vector of electric current and @xmath16 is a second - order positive - definite tensor of electric conductivity . in addition ,
the field @xmath13 is subject to a constraint @xmath17 where @xmath18 denotes a prescribed macroscopic electric field and @xmath19 represents the @xmath20-dimensional measure of @xmath14 .
next , we introduce a homogeneous reference medium with constant conductivity @xmath21 , leading to a decomposition of the electric current field in the form @xmath22 the original problem is then equivalent to the periodic lippmann - schwinger integral equation , formally written as @xmath23 where the @xmath24 operator is derived from the green s function of the problem with @xmath25 and @xmath26 .
making use of the convolution theorem , eq .
attains a local form in the fourier space : @xmath27 where @xmath28 denotes the fourier coefficient of @xmath29 for the @xmath30-th frequency given by @xmath31 " @xmath32 is the imaginary unit and @xmath33 here , we refer to @xcite for additional details . numerical solution of the lippmann - schwinger equation is based on a discretization of a unit cell @xmath14 into a regular periodic grid with @xmath34 nodal points and grid spacings @xmath35 .
the searched field @xmath13 in is approximated by a trigonometric polynomial @xmath36 in the form ( cf .
@xcite ) @xmath37 where @xmath38 , @xmath39 designates the fourier coefficients defined in and @xmath40 we recall , e.g. from @xcite , that the @xmath41-th component of the trigonometric polynomial expansion @xmath42 admits two equivalent finite - dimensional representations . the first one is based on a matrix @xmath43 of the fourier coefficients of the @xmath41-th component and equation with @xmath44 .
second , the data can be entirely determined by interpolation of nodal values @xmath45 where @xmath46 is a matrix storing electric field values at grid points , @xmath47 is the corresponding value at the @xmath30-th node with coordinates @xmath48 and basis functions @xmath49 satisfy the dirac delta property @xmath50 with @xmath51 . both representations can be directly related to each other by @xmath52 where the vandermonde matrices @xmath53 and @xmath54 @xmath55 implement the forward and inverse fourier transform , respectively , e.g. ( * ? ? ? * section 4.6 ) .
the trigonometric collocation method is based on the projection of the lippmann - schwinger equation to the space of the trigonometric polynomials of the form @xmath56 , cf .
@xcite . in view of eq .
, this is equivalent to the collocation at grid points , with the action of @xmath24 operator evaluated from the fourier space expression converted to the nodal representation by @xmath57 .
the resulting system of collocation equations reads @xmath58 where @xmath59 and @xmath60 store the corresponding solution and of the macroscopic field , respectively .
furthermore , @xmath61 is the @xmath62 unit matrix and the non - symmetric matrix @xmath63 can be expressed , for the two - dimensional setting , in the partitioned format as @xmath64 } { \left [ \begin{array}{cc } { \widehat{{{\mathsf{\gamma}}}}}^0_{11 } & { \widehat{{{\mathsf{\gamma}}}}}^0_{12 } \\ { \widehat{{{\mathsf{\gamma}}}}}^0_{21 } & { \widehat{{{\mathsf{\gamma}}}}}^0_{22 } \end{array}\right ] } { \left [ \begin{array}{cc } { { \mathsf{f } } } & { \mathsf{0 } } \\ { \mathsf{0 } } & { { \mathsf{f}}}\end{array}\right ] } { \left [ \begin{array}{cc } { { \mathsf{\delta{\mathsf{l}}}}}_{11 } & { { \mathsf{\delta{\mathsf{l}}}}}_{12 } \\ { { \mathsf{\delta{\mathsf{l}}}}}_{21 } & { { \mathsf{\delta{\mathsf{l}}}}}_{22 } \end{array}\right ] } , \ ] ] with an obvious generalization to an arbitrary dimension . here ,
@xmath65 and @xmath66 are diagonal matrices storing the corresponding grid values , for which it holds @xmath67 it follows from eq . that the cost of the multiplication by @xmath63 or by @xmath68 is driven by the forward and inverse fourier transforms , which can be performed in @xmath69 operations by fft techniques .
this makes the resulting system ideally suited for iterative solvers .
in particular , the original fast fourier transform - based homogenization ( ffth ) scheme formulated by moulinec and suquet in @xcite is based on the neumann expansion of the matrix inverse @xmath70 , so as to yield the @xmath71-th iterate in the form @xmath72 convergence of the series was comprehensively studied in @xcite , where it was shown that the optimal rate of convergence is achieved for @xmath73 with @xmath74 and @xmath75 denoting the minimum and maximum eigenvalues of @xmath76 on @xmath14 and @xmath77 being the identity tensor . here , we propose to solve the non - symmetric system by well - established krylov subspace methods , in particular , exploiting the classical conjugate gradient ( @xmath78 ) method @xcite and the biconjugate gradient ( @xmath79 ) algorithm @xcite . even though that @xmath78 algorithm is generally applicable to symmetric and positive - definite systems
only , its convergence in the one - dimensional setting has been proven by vondejc ( * ? ? ?
* section 6.2 ) .
a successful application of @xmath78 method to a generalized eshelby inhomogeneity problem has also been recently reported by novk @xcite and kanaun @xcite .
to assess the performance of the conjugate gradient algorithms , we consider a model problem of the transverse electric conduction in a square array of identical circular particles with @xmath80 volume fraction .
a uniform macroscopic field @xmath81 is imposed on the corresponding single - particle unit cell , discretized by @xmath82 nodes and the phases are considered to be isotropic with the conductivities set to @xmath83 for the matrix phase and to @xmath84 for the particle .
the conductivity of the homogeneous reference medium is parameterized as @xmath85 where @xmath86 corresponds to the optimal convergence of @xmath87 algorithm .
all conjugate gradient - related results have been obtained using the implementations according to @xcite and referred to as algorithm 6.18 ( @xmath78 method ) and algorithm 7.3 ( @xmath79 scheme ) .
two termination criteria are considered .
the first one is defined for the @xmath71-th iteration as @xcite @xmath88 and provides the test of the equilibrium condition @xmath57 in the fourier space .
an alternative expression , related to the standard residual norm for iterative solvers , has been proposed by vinogradov and milton in @xcite and admits the form @xmath89 with the additional @xmath90 term ensuring the proportionality to at convergence . from the numerical point of view , the latter criterion is more efficient than the equilibrium variant , which requires additional operations per iteration . from the theoretical point of view , its usage is justified only when supported by a convergence result for the iterative algorithm . in the opposite case , the equilibrium norm appears to be more appropriate , in order to avoid spurious non - physical solutions .
since no results for the optimal choice of the reference medium are known for @xmath91-based solvers , we first estimate their sensitivity to this aspect numerically .
the results appear in fig .
[ fig : example1a ] , plotting the relative number of iterations for @xmath78 and @xmath79 solvers against the conductivity of the reference medium parameterized by @xmath92 , recall eq . . as expected , both @xmath78 and @xmath79 solvers achieve a significant improvement over @xmath87 method in terms of the number of iterations , ranging from @xmath80 for a mildly - contrasted composite down to @xmath93 for @xmath94 .
moreover , contrary to all other available methods , the number of iterations is almost independent of the choice of the reference medium .
we also observe , in agreement with results by ( * ? ? ?
* section 6.2 ) for the one - dimensional setting , that @xmath78 and @xmath79 algorithms generate identical sequences of iterates ; the minor differences visible for @xmath95 or @xmath94 can be therefore attributed to accumulation of round - off errors .
these conclusions hold for both equilibrium- and residual - based norms , which appear to be roughly proportional for the considered range of the phase contrasts , cf .
[ fig : example1b ] .
therefore , the residual criterion will mostly be used in what follows . of @xmath78 and @xmath79
solvers plotted against the conductivity parameter @xmath92 for @xmath96-based termination condition with tolerance @xmath97 . ] in fig .
[ fig : example1c ] , we supplement the comparison by considering the total cpu time required to achieve a convergence .
the data indicate that the cost of one iteration is governed by the matrix - vector multiplication , recall eq .
: the overhead of @xmath78 scheme is about @xmath98 with respect to @xmath87 method , while the application of @xmath79 algorithm , which involves @xmath63 and @xmath68 products per iteration @xcite , is about twice as demanding . as a result , cg algorithm significantly reduces the overall computational time in the whole range of contrasts , whereas a similar effect has been reported for the candidate schemes only for @xmath99 , cf .
@xcite . as confirmed by all previous works ,
the phase contrast @xmath100 is the critical parameter influencing the convergence of fft - based iterative solvers . in fig .
[ fig : example2 ] , we compare the scaling of the total number of iterations with respect to phase contrast for cg and @xmath87 methods , respectively . the results clearly show that the number of iterations grows as @xmath101 instead of the linear increase for @xmath87 method .
this follows from error bounds @xmath102 the first estimate was proven in @xcite , whereas the second expression is a direct consequence of the condition number of matrix @xmath63 being proportional to @xmath100 and a well - known result for the conjugate gradient method , e.g. ( * ? ? ?
* section 6.11.3 ) .
the cg - based method , however , failed to converge for the infinite contrast limit .
such behavior is equivalent to the eyre - milton scheme @xcite .
it is , however , inferior to the augmented lagrangian algorithm , for which the convergence rate improves with increasing @xmath103 and the method converges even as @xmath104 .
nonetheless , such results are obtained for optimal , but not always straightforward , choice of the parameters @xcite .
plotted against phase contrast @xmath105 ; the reference medium corresponds to for @xmath86 and tolerance @xmath106 is related to @xmath96 norm . ]
the final illustration of the @xmath78-based algorithm is provided by fig .
[ fig : example3 ] , displaying a detailed convergence behavior for both low- and high - contrast cases . the results in fig .
[ fig : example3a ] correspond well with estimates for both residual and equilibrium - based norms .
influence of a higher phase contrast is visible from fig .
[ fig : example3b ] , plotted in the full logarithmic scale .
for @xmath87 algorithm , two regimes can be clearly distinguished . in the first few iterations , the residual error rapidly decreases , but
the iterates tend to deviate from equilibrium . then , both residuals are simultaneously reduced . for @xmath78 scheme
, the increase of the equilibrium residual appears only in the first iteration and then the method rapidly converges to the correct solution .
however , its convergence curve is irregular and
the algorithm repeatedly stagnates in two consecutive iterations .
further analysis of this phenomenon remains a subject of future work .
in this short note , we have presented a conjugate gradient - based acceleration of the fft - based homogenization solver originally proposed by moulinec and suquet @xcite and illustrated its performance on a problem of electric conduction in a periodic two - phase composite with isotropic phases .
on the basis of obtained results , we conjecture that : * the non - symmetric system of linear equations , arising from discretization by the trigonometric collocation method @xcite , can be solved using the standard conjugate gradient algorithm , * the convergence rate of the method is proportional to the square root of the phase contrast , * the methods fails to converge in the infinite contrast limit , * contrary to available improvements of the original fft - solver @xcite , the cost of one iteration remains comparable to the basic scheme and the method is insensitive to the choice of auxiliary reference medium .
the presented computational experiments provide the first step towards further improvements of the method , including a rigorous analysis of its convergence properties , acceleration by multi - grid solvers and preconditioning and the extension to non - linear problems .
the authors thank milan jirsek ( czech technical university in prague ) and christopher quince ( university of glasgow ) for helpful comments on the manuscript .
this research was supported by the czech science foundation , through projects no .
gar 103/09/1748 , no .
gar 103/09/p490 and no .
gar 201/09/1544 , and by the grant agency of the czech technical university in prague through project
sgs ohk1 - 064/10 .
h. moulinec , p. suquet , a fast numerical method for computing the linear and nonlinear mechanical properties of composites , comptes rendus de lacadmie des sciences .
srie ii , mcanique , physique , chimie , astronomie 318 ( 11 ) ( 1994 ) 14171423 .
g. vainikko , fast solvers of the lippmann - schwinger equation , in : r. p. gilbert , j. kajiwara , y. s. xu ( eds . ) , direct and inverse problems of mathematical physics , vol . 5 of international society for analysis , applications and computation , kluwer academic publishers , dordrecht , the netherlands , 2000 , pp . 423440 .
j. c. michel , h. moulinec , p. suquet , a computational method based on augmented lagrangians and fast fourier transforms for composites with high contrast , cmes - computer modeling in engineering & sciences 1 ( 2 ) ( 2000 ) 7988 .
r. fletcher , conjugate gradient methods for indefinite systems , in : g. watson ( ed . ) , numerical analysis , proceedings of the dundee conference on numerical analysis , 1975 , vol .
506 of lecture notes in mathematics , springer - verlag , new york , 1976 , pp .
j. vondejc , analysis of heterogeneous materials using efficient meshless algorithms : one - dimensional study , master s thesis , czech technical university in prague ( 2009 ) .
j. novk , calculation of elastic stresses and strains inside a medium with multiple isolated inclusions , in : m. papadrakakis , b. topping ( eds . ) , proceedings of the sixth international conference on engineering computational technology , stirlingshire , uk , 2008 , p. 16
pp , paper 127 .
doi:10.4203/ccp.89.127 h. moulinec , p. suquet , a numerical method for computing the overall response of nonlinear composites with complex microstructure , computer methods in applied mechanics and engineering 157 ( 12 ) ( 1998 ) 6994 . | in this short note , we present a new technique to accelerate the convergence of a fft - based solver for numerical homogenization of complex periodic media proposed by moulinec and suquet @xcite .
the approach proceeds from discretization of the governing integral equation by the trigonometric collocation method due to vainikko @xcite , to give a linear system which can be efficiently solved by conjugate gradient methods .
computational experiments confirm robustness of the algorithm with respect to its internal parameters and demonstrate significant increase of the convergence rate for problems with high - contrast coefficients at a low overhead per iteration . numerical homogenization ; fft - based solvers ; trigonometric collocation method ; conjugate gradient solvers |
over the last decades the @xmath0-calculus has been connecting mathematics and physics in applications that span from quantum theory and statistical mechanics , to number theory and combinatorics ( see @xcite and references therein ) .
its history dates back to the beginnings of the last century when , based on pioneering works of euler and heine , the english reverend frank hilton jackson developed the @xmath0-calculus in a systematic way @xcite .
his work gave rise to generalizations of series , functions and special numbers within the context of the @xmath0-calculus @xcite .
more important , he reintroduced the concepts of the @xmath0-derivative @xcite ( also known as jackson s derivative ) and introduced the @xmath0-integral @xcite .
the @xmath0-derivative of a function @xmath1 of one variable is defined as @xmath2 where @xmath0 is a real number different from @xmath3 and @xmath4 is different from @xmath5 . in the limit of @xmath6 ( or @xmath7 ) , the @xmath0-derivative reduces to the classical derivative .
let @xmath8 , for example . in this case , the classical derivative of @xmath9 is @xmath10 and the @xmath0-derivative is @xmath11x^{n-1}$ ] , where @xmath11 $ ] is the @xmath0-analogue of @xmath12 given by @xmath13= \frac{q^n -1}{q-1}.\ ] ] as @xmath6 , @xmath11 $ ] tends to @xmath12 .
this definition is used to calculate the @xmath0-binomial and establish a @xmath0-analogue of taylor s formula that encompasses many results such as the euler s identities for @xmath0-exponential functions , gauss s @xmath0-binomial formula , heine s formula for a @xmath0-hypergeometric function , among others mathematical results @xcite .
considering arbitrary functions @xmath1 and @xmath14 , the @xmath0-derivative operator satisfy the following properties @xcite : 1 .
the @xmath0-derivative is a linear operator for any constants @xmath15 and @xmath16 @xmath17 2 .
the @xmath0-derivative of the product of @xmath1 and @xmath14 is given by @xmath18 that , by symmetry , is equivalent to @xmath19 3 .
the @xmath0-derivative of the quotient of @xmath1 and @xmath14 is calculated as @xmath20 or equivalently @xmath21 the chain rule for @xmath0-derivatives does not exist , except for a function of the form @xmath22 , where @xmath23 , with @xmath24 being constants .
more details on the properties of @xmath0-derivatives can be found in @xcite .
let a general nonlinear unconstrained optimization problem be defined as @xmath25 where @xmath26 is the vector of the independent variables and @xmath27 is the objective function .
the steepest descent method ( also known as the gradient descent method ) uses information on the gradient of the objective function in seeking the optimum .
the search direction is given by the negative of the gradient of @xmath28 .
this search strategy is an obvious choice since along this direction the objective function decreases most rapidly . requiring only information about first - derivatives ,
the steepest descent method is attractive because of its limited computational cost and storage requirements @xcite .
however , for multimodal functions , unless one knows in advance where to start from , the search procedure frequently gets stuck in one of the local minima .
consequently , the steepest descent method is not recommended for real - world optimization problems that are usually multimodal .
nevertheless , because of its inherent simplicity , it represents a good starting point for the development of more advanced optimization methods . here
we propose a generalization of the steepest descent method in which the gradient of the objective function is replaced by its @xmath0-analogue .
accordingly , the search direction is taken as the negative of the @xmath0-gradient of @xmath28 . for @xmath29 ,
the here called @xmath0-gradient method reduces to the classical steepest descent method . in order to evaluate the performance of the @xmath0-gradient method we consider three unimodal and three multimodal test functions commonly used as benchmarks
we compare our results with those obtained with the genetic algorithms ( gas ) g3-pcx developed by deb et al .
@xcite , and the spc - vsbx and spc - pnx developed by ballester and carter @xcite , which previous studies have shown to be effective in minimizing multidimensional unimodal and multimodal functions .
the rest of the paper is organized as follows . in section [ sec : qgrad ] the @xmath0-gradient vector is defined . in section [ sec : qgradmethod ] the strategies to obtain the parameter @xmath0 and the step length are described .
section [ sec : description ] shows the computational experiments and section [ sec : results ] discusses the results . finally , in section [
sec : conclusions ] some conclusions and future work are presented .
given a differentiable function of @xmath12 variables @xmath30 , the gradient of @xmath28 is the vector of the @xmath12 first - order partial derivatives of @xmath28 .
similarly , the @xmath0-gradient is the vector of the @xmath12 first - order partial @xmath0-derivatives of @xmath28 .
thus , let the parameter @xmath0 be a vector @xmath31 , where @xmath32 , the first - order partial @xmath0-derivative with respect to the variable @xmath33 is given by @xmath34 with @xmath35 and @xmath36 this framework can be extended to define the @xmath0-gradient of a function of @xmath12 variables as @xmath37\ ] ] with the classical gradient being recovered in the limit of @xmath38 , for all @xmath39 . the fig .
[ fig : geometrico ] illustrates the geometric interpretation of the classical gradient and the @xmath0-gradient of a function of one variable @xmath40 . in this case , the gradient is simply the slope of the tangent line at @xmath4 .
similarly , the @xmath0-gradient is the slope of the secant line passing through the points @xmath41 and @xmath42 . if the slope of the secant line is positive ( negative ) , the @xmath0-gradient points to the right ( left ) direction .
-derivative of @xmath1 at @xmath43 and different values of the parameter @xmath0.,width=336 ] since the slope of the tangent line ( dotted line ) at @xmath44 is positive , the steepest descent method at this point will move necessarily to the left and , thus , will be trapped by the local minimum at @xmath45 .
the slope of the secant line passing through the points @xmath41 and @xmath42 can be positive or negative depending on the value of the parameter @xmath0 .
for instance , if @xmath46 the @xmath0-derivative is negative at @xmath44 ( see the secant line passing through @xmath41 and @xmath47 in fig . [
fig : geometrico ] ) , which potentially allows a minimization strategy based on the value of the @xmath0-gradient to take a leap to the right , towards the global minimum of @xmath9 .
note that there is a value @xmath48 for which @xmath44 is a stationary point of the @xmath0-gradient ( @xmath49 ) but that does not correspond to any minimum or maximum of @xmath9 .
the stationary points of the @xmath0-gradient are avoided in the method by generating a new the parameter @xmath0 .
finally , for @xmath50 or @xmath51 the slope of the secant line is positive and the @xmath0-gradient method will move to the left as the steepest descent method .
this simple example above shows that the use of the @xmath0-gradient offers a new mechanism to escape from local minima .
moreover , the transition from global to local search might be controlled by the parameter @xmath0 , provided a suitable strategy for generating @xmath0-values is incorporated into the minimization algorithm .
a general optimization strategy is to consider an iterative procedure that , starting from @xmath52 , generates a sequence @xmath53 given by @xcite @xmath54 where @xmath55 is the search direction and @xmath56 is the step length or the distance moved along @xmath55 in the iteration @xmath57 .
optimization methods can be characterized according to the direction and step length used in eq .
( [ eq : iterprocess ] ) .
the steepest descent method sets @xmath58 and the step length @xmath56 is usually determined by a line - search technique that minimizes the objective function along the direction @xmath55 . in the @xmath0-gradient method , as here proposed , the search direction is the negative of the @xmath0-gradient of the objective function @xmath59 .
thus the optimization procedure defined by eq .
( [ eq : iterprocess ] ) becomes @xmath60 key to the performance of the @xmath0-gradient method , the strategies used to specify the parameter @xmath0 and the step length @xmath61 are described below . considering a function of @xmath12 variables @xmath30 , a set of @xmath12 different parameters @xmath63 ( @xmath39 ) are needed to compute the @xmath0-gradient vector of @xmath28 .
the overall strategy adopted here is to draw the values of @xmath64 ( or some variable related to them ) from some suitable probability density function ( pdf ) , and with a standard deviation that decreases as the iterative search proceeds . in this sense
, the role of the standard deviation here is reminiscent of the one played by the temperature in a simulated annealing ( sa ) algorithm , that is , to make the algorithm go from a very random ( at the beginning ) to a very deterministic search ( at the end ) . in the current implementation we opted to first draw the values of @xmath65 from a gaussian pdf given by @xmath66,\ ] ] with @xmath67 and @xmath68 ; then , we computed the values of @xmath69 .
starting from @xmath70 , the standard deviation of the pdf is decreased by the following `` cooling '' schedule , @xmath71 , where @xmath72 is the reduction factor .
as @xmath73 approaches zero , the values of @xmath69 tend to unity , the algorithm reduces to the steepest descent method , and the search process becomes essentially local . as in a sa algorithm , the performance of the minimization algorithm depends crucially on the choice of parameters @xmath70 and @xmath74 .
a too rapid decrease of @xmath73 , for example , may cause the algorithm to be trapped in a local minimum .
the calculation of the step length @xmath61 is a tradeoff . on the one hand
, @xmath61 should give a considerable reduction of the objective function .
on the other hand , its calculation should not take too many evaluations of @xmath28 @xcite .
steepest descent algorithms generally use line - search techniques to determine the step length @xmath75 along the steepest descent direction @xmath76 at the iteration @xmath57 . a first version of our algorithm @xcite applied the golden section for step length determination .
however , traditional line - search algorithms , like the golden section , ensure that the condition @xmath77 is always satisfied , what obviously is a poor strategy when dealing with multimodal minimization problems .
in addition , depending on the value of @xmath0 , the negative of the @xmath0-gradient may not point to the local descent direction .
one way to circumvent these problems is to use a diminishing step length @xmath75 , i.e. , the initial step length @xmath78 is reduced by @xmath79 , where , for the sake of simplicity , @xmath74 is the same reduction factor used to compute @xmath73 .
as the step length decreases ( and the values of @xmath69 , in parallel , tend to unity ) , a smooth transition to an increasingly local search process occurs . + the main idea behind the @xmath0-gradient method is to use the negative of the @xmath0-gradient of @xmath28 , instead of the negative of the classical gradient of @xmath28 , as the search direction .
strategies for generating the parameter @xmath80 and the step length @xmath75 , at each iteration , complement this very simple algorithm .
note that there are three free parameters to be specified , namely , @xmath70 , @xmath78 and @xmath74 .
the initial standard deviation @xmath70 determines how stochastic is the search . for multimodal functions ,
it must be high enough to allow the method to properly sample the search space .
the reduction factor @xmath74 controls the speed of the transition from stochastic to deterministic search . a @xmath74 close to
@xmath3 reduces the risk of being trapped in a local minimum .
the last free parameter , the initial step length @xmath78 , depends heavily on the topology of the search space and , thus , requires some empirical exploration . in the end , as with the choice of the cooling schedule in a sa algorithm @xcite , an appropriate specification of the three free parameters is strictly dependent on the objective function .
although a bad choice may lead to some deterioration in its performance , the @xmath0-gradient method has shown to be sufficiently robust to still be capable of reaching the global minimum .
the performance of the @xmath0-gradient method was evaluated over six @xmath81-variable test functions ( @xmath82 ) commonly employed in the literature .
we use the same experimental setup and stopping criteria as described in @xcite and @xcite in order to allow a direct comparison with their results .
the stopping criteria are : maximum of @xmath83 function evaluations or @xmath84 . as in @xcite and @xcite
, we set the three free parameters for the @xmath0-gradient method after preliminary exploratory runs .
the results presented here are for those which yielded the best performance .
the benchmark consists of the following analytical functions : 1 .
ellipsoidal function ( @xmath85 ) @xmath86 although the ellipsoidal function is convex and unimodal , it is an example of a poorly scaled function . 2 .
schwefel s function ( @xmath87 ) @xmath88 the schwefel s function is an extension of the ellipsoidal function and it is also a unimodal and poorly scaled function .
4 . 5 .
generalized rosenbrock s function ( @xmath89 ) @xmath90.\ ] ] although the rosenbrock s function is a well - known unimodal function for @xmath91 , numerical experiments have shown that for @xmath92 the function has two minima , the global one at @xmath93 and a local minimum that changes with the dimensionality @xmath12 @xcite .
the rosenbrock s function is considered a test case for premature convergence once the global minimum lays inside a long , narrow , and parabolic shaped flat valley .
6 . 7 .
ackley s function ( @xmath94 ) @xmath95 the ackley s function is highly multimodal and the basin of the local minima increase in size as one moves away from the global minimum @xcite .
rastrigin s function ( @xmath96 ) @xmath97 the rastrigin s function has a parabolic landscape away from the global minimum , but as we move towards the global minimum , the size of the basins increase .
the function is highly multimodal and its characteristics are known to be difficult for many optimization algorithms to achieve the global minimum @xcite .
. 11 . rotated rastrigin s function ( @xmath98 ) . @xmath99
the rotated rastrigin s is a highly multimodal function without local minima arranged along the axis @xcite .
12 . for all these functions the global minimum is @xmath100 at @xmath101 , except the generalized rosenbrock s function where @xmath102 .
the initial point set @xmath52 for each function is generated by a uniform distribution within @xmath103 $ ] , as used in @xcite and @xcite .
extensive comparisons between the gas g3-pcx ( results obtained from @xcite for ellipsoidal , schwefel , rosenbrock and rastrigin functions ; and from @xcite for ackley and rotated rastrigin ) , spc - vsbx and spc - pnx ( results obtained from @xcite ) and the @xmath0-gradient method are presented in tables [ tab : unimodal ] and [ tab : multimodal ] . as in @xcite and @xcite ,
the `` best '' , `` median '' and `` worst '' columns refer to the number of function evaluations required to reach the accuracy @xmath104 .
when this condition is not achieved , the best value found so far for the test function after @xmath83 evaluations is given in column `` @xmath105 '' .
the column `` success '' refers to how many runs reached the target accuracy , for unimodal functions , or ended up within the global minimum basin , for multimodal ones .
the best performances are highlighted in bold in each table .
the corresponding values of the best parameters @xmath70 , @xmath78 and @xmath74 used in each test function are given in table [ tab : parameters ] .
llll functions & @xmath106 & @xmath107 & @xmath108 + ellipsoidal & @xmath109 & @xmath110 & @xmath111 + schwefel & @xmath112 & @xmath3 & @xmath113 + rosenbrock & @xmath112 & @xmath112 & @xmath114 + ackley & @xmath81 & @xmath115 & @xmath116 + rastrigin & @xmath117 & @xmath118 & @xmath114 + rotated rastrigin & @xmath119 & @xmath120 & @xmath121 + lllllll function & method & best & median & worst & @xmath105 & success + & * g3-pcx * & @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 + & spc - vsbx & @xmath127 & @xmath128 & @xmath129 & @xmath104 & @xmath130 + & spc - pnx & @xmath131 & @xmath132 & @xmath133 & @xmath104 & @xmath130 + & * @xmath0-gradient * & @xmath134 & @xmath135 & @xmath136 & @xmath125 & @xmath137 + & * g3-pcx * & @xmath138 & @xmath139 & @xmath140 & @xmath125 & @xmath126 + & spc - vsbx & @xmath141 & @xmath142 & @xmath143 & @xmath104 & @xmath130 + & spc - pnx & @xmath144 & @xmath145 & @xmath146 & @xmath104 & @xmath130 + & @xmath0-gradient & @xmath147 & @xmath148 & @xmath149 & @xmath104 & @xmath150 + & * g3-pcx * & @xmath151 & @xmath152 & @xmath153 & @xmath125 & @xmath154 + & spc - vsbx & @xmath83 & - & - & @xmath155 & @xmath156 + & spc - pnx & @xmath83 & - & - & @xmath157 & @xmath158 + & @xmath0-gradient & @xmath83 & - & - & @xmath157 & @xmath150 + lllllll function & method & best & median & worst & @xmath105 & success + & g3-pcx & @xmath83 & - & - & @xmath159 & @xmath5 + & spc - vsbx & @xmath160 & @xmath161 & @xmath162 & @xmath157 & 10/10 + & spc - pnx & @xmath163 & @xmath164 & @xmath165 & @xmath157 & 10/10 + & * @xmath0-gradient * & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath137 + & g3-pcx & @xmath83 & - & - & @xmath170 & @xmath5 + & spc - vsbx & @xmath171 & @xmath172 & @xmath173 & @xmath104 & 6/10 + & spc - pnx & @xmath83 & - & - & @xmath174 & 0 + & * @xmath0-gradient * & @xmath175 & @xmath176 & @xmath177 & @xmath125 & @xmath178 + & g3-pcx & @xmath83 & - & - & @xmath179 & @xmath5 + rotated & spc - vsbx & @xmath83 & - & - & @xmath180 & @xmath5 + rastrigin & spc - pnx & @xmath83 & - & - & @xmath181 & @xmath5 + & * @xmath0-gradient * & @xmath182 & @xmath183 & @xmath184 & @xmath125 & @xmath185 + in table [ tab : unimodal ] , for the ellipsoidal function , the @xmath0-gradient method achieved the required accuracy @xmath104 for all 50 runs , with an overall performance similar to the one displayed by the g3-pcx , the best algorithm among the gas .
as for the schwefel s function , the @xmath0-gradient method again attained the required accuracy for all runs but was outperformed by the g3-pcx in terms of the number of function evaluations .
finally , for the rosenbrock s function , the @xmath0-gradient was beaten by the g3-pcx ( the only to achieve the required accuracy ) but performed better then the two other gas .
the overall evaluation of the @xmath0-gradient method performance in these numerical experiments with unimodal ( or quasi - unimodal ) test functions indicates that it reaches the required accuracy ( or the minimum global basin ) in @xmath186 of the runs , but it is not faster than the g3-pcx .
this picture improves a lot when it comes to tackle the multimodal ackley s and rastringin s functions . in table
[ tab : multimodal ] , due to limited computing precision the required accuracy for the ackley s function was set to @xmath157 for the gas and @xmath187 in our simulations with double precision is equal to @xmath188e@xmath189 and not zero . ] .
the @xmath0-gradient method was here clearly better than the gas , reaching the required accuracy in more runs or in less functions evaluations .
for the rastrigin s function , the g3-pcx and the spc - pnx were unable to attain the global minimum basin .
the other two algorithms reached the required accuracy @xmath104 , but the @xmath0-gradient method was the only to do it in @xmath190 of the runs ( 48 over 50 ) .
finally , in the case of the rotated rastrigin s function , the @xmath0-gradient was the only algorithm to reach the minimum , attaining the required accuracy in @xmath81 out of @xmath191 independent runs .
summarizing the results with multimodal functions , we may say that the @xmath0-gradient method outperformed the gas in all the three test cases considered , reaching the minimum with less function evaluations or in more independent runs .
the main idea behind the @xmath0-gradient method is the use of the negative of the @xmath0-gradient of the objective function a generalization of the classical gradient based on the jackson s derivative as the search direction .
the use of jackson s derivative provides an effective mechanism for escaping from local minima .
the method has strategies for generating the parameter @xmath0 and the step length that makes the search process gradually shifts from global in the beginning to almost local search in the end . for testing this new approach , we considered six commonly used 20-variable test functions . these functions display features of real - world optimization problems ( multimodality , for example ) and are notoriously difficult for optimization algorithms to handle .
we compared the @xmath0-gradient method with gas developed by deb et al .
@xcite , and ballester and carter @xcite with promising results .
overall , the @xmath0-gradient method clearly beat the competition in the hardest test cases , those dealing with the multimodal functions .
it comes without suprise the ( relatively ) poor results of the @xmath0-gradient method with the rosenbrock s function , a unimodal test function specially difficult to be solved by the steepest descent method .
this result highlights the need for the development of a @xmath0-generalization of the well - known conjugate - gradient method , a research line currently being explored .
+ ballester , p.j . ,
carter , j.n . : an effective real - parameter genetic algorithm with parent centric normal crossover for multimodal optimisation . in : proceedings of the genetic and evolutionary computation conference ,
. 901913 .
seattle , wa , usa ( 2004 ) soterroni , a. c. , galski , r. l. , ramos , f. m. : the @xmath0-gradient vector for unconstrained continuous optimization problems . in : operations
research proceddings 2010 , pp . 365370 .
springer - verlag berlin heidelberg ( 2011 ) locatelli , m. : simulated annealing algorithms for continuous global optimization . in : pardalos ,
j.m . and romeijn , h. e. ( eds . ) handbook of global optimization ii , pp .
kluwer academic publishers , dordrecht ( 2002 ) | the @xmath0-gradient is an extension of the classical gradient vector based on the concept of jackson s derivative . here
we introduce a preliminary version of the @xmath0-gradient method for unconstrained global optimization .
the main idea behind our approach is the use of the negative of the @xmath0-gradient of the objective function as the search direction . in this sense , the method here proposed is a generalization of the well - known steepest descent method .
the use of jackson s derivative has shown to be an effective mechanism for escaping from local minima .
the @xmath0-gradient method is complemented with strategies to generate the parameter @xmath0 and to compute the step length in a way that the search process gradually shifts from global in the beginning to almost local search in the end . for testing this new approach
, we considered six commonly used test functions and compared our results with three genetic algorithms ( gas ) considered effective in optimizing multidimensional unimodal and multimodal functions . for the multimodal test functions , the @xmath0-gradient method outperformed the gas , reaching the minimum with a better accuracy and with less function evaluations . |
for over three decades , it has been well established that ( some ) supernova explosions give rise to strongly magnetized ( @xmath5 g ) , rapidly rotating ( @xmath630 ms ) neutron stars ( nss ) , as in the young radio pulsars found in the crab nebula and nearly a dozen other supernova remnants ( snrs ) . in some cases ,
a synchrotron nebula ( or `` plerion '' ) has been detected around the pulsar , powered by non - thermal emission from the ns .
emission ( in some cases pulsed ) has been detected at other wavelengths ( optical , x - ray , gamma - ray ) arising from thermal and non - thermal processes .
however , several clues have recently emerged suggesting that this paradigm is incomplete ( see kaspi 2000 and gotthelf & vasisht 2000 for recent reviews ) .
first , there are the six slowly - rotating ( @xmath7 s ) `` anomalous x - ray pulsars '' ( axps ) , which seem to be young isolated nss and may have extremely strong ( @xmath8@xmath9 g ) surface magnetic fields ( mereghetti 2000 ) .
half of the axps are associated with snrs .
possibly related are the four known soft gamma - ray repeaters ( sgrs ) , which in quiescence share many properties with axps and may also be associated with snrs ( hurley 2000 ) . also intriguing
has been the identification of at least three radio - quiet non - plerionic x - ray point sources near the centers of snrs ( pup a , rcw 103 , and pks 120952 ; see brazier & johnston 1999 and references therein )
. these objects have x - ray spectra roughly consistent with young , cooling nss , but show no evidence for either x - ray pulsations or emission at other wavelengths ( in contrast to `` normal '' young nss ) . finally , the ongoing failure to detect clear evidence for a young ns in the remnant of sn 1987a in the large magellanic cloud ( lmc ) has renewed theoretical interest in alternative models for the aftermath of a sn explosion , especially with respect to fallback of ejected material onto a newborn ns .
several groups have concluded that , under some circumstances , a newborn ns might collapse into a black hole ( bh ) shortly after birth ( brown & bethe 1994 ; woosley & timmes 1996 ; zampieri et al . 1998 ; fryer , colgate , & pinto 1999 ) .
nearby snrs without known stellar remnants are thus obvious targets for further study .
after sn 1987a , the youngest known snr in our galaxy or the satellite magellanic clouds is cassiopeia a. its parent supernova was evidently noticed ( though misunderstood ) by flamsteed in 1680 ( ashworth 1980 ) .
this @xmath10 yr historical age for cas a is consistent with its optical expansion time scale ( van den bergh & kamper 1983 ) , though somewhat shorter than its x - ray ( @xmath11 yr ; koralesky et al .
1998 , vink et al . 1998 ) and radio ( @xmath12750870 yr ; anderson & rudnick 1995 ) expansion time scales .
the progenitor of this oxygen - rich snr was probably a very massive ( zero - age main sequence mass @xmath13 ) late wn - type wolf - rayet star which underwent prodigious mass loss via a stellar wind and eventually exploded as a type ii supernova ( fesen , becker , & blair 1987 ) . the inferred distance to cas a is @xmath14 kpc ( reed et al . 1995 ) .
the remnant subtends 4 arcmin in the sky , is one of the brightest non - thermal radio sources , and has been extensively studied in the radio , optical , and x - ray bands .
the _ chandra x - ray observatory _ `` first light '' observation on 1999 august 20 revealed the presence of a compact x - ray source near the geometric center of cas a ( tananbaum 1999 ) .
the source morphology is point - like , with no obvious evidence for extension or a surrounding nebula ( e.g. , a plerion ) .
the discovery announcement notes that no obvious counterparts were detected within a 5 arcsec radius of the point source position on 20 cm radio maps or optical images .
previous x - ray missions lacked sufficient spatial resolution to separate the point source from the diffuse snr emisison .
however , armed with the _ chandra _ position , aschenbach ( 1999 ) detected the point source in archival 0.12.4 kev x - ray images taken with the _
rosat_/hri in 1995 - 1996 , and pavlov & zavlin ( 1999 ) recovered the point source in archival 0.54 kev x - ray images taken with the _
einstein_/hri in 1979 and 1981 .
the _ einstein _ , _ rosat _ , and _
count rates were consistent with a constant x - ray source flux over all the observations ( pavlov & zavlin 1999 ; pavlov et al .
they also noted that the observed spectrum appeared to be inconsistent with pure blackbody radiation from the entire surface of a cooling ns .
umeda et al .
( 2000 ) speculated on some possible scenarios for the nature of the point source , based on these early results . in this paper , we present a detailed analysis of the x - ray spectral and timing features of the central point source in cas a , based on the available public _
chandra _ data . in 2 , we give a detailed description of the observation and our data analysis , including our efforts to verify the instrumental calibration . in 3 , we discuss our results in the context of various models for the nature of the point source .
we summarize our finding in 4 .
while completing our manuscript , we learned of another paper presenting an independent analysis of the spectral data by pavlov et al .
they used a subset of the data that we discuss in our paper , and their spectral results are consistent with ours within the uncertainties .
we include a brief discussion of their preferred interpretation in 3 .
_ chandra _ ( formerly _ axaf _ ; weisskopf et al .
2000 ) was launched on 1999 july 23 .
its high resolution mirror assembly ( hrma ) , which consists of four pairs of nested grazing - incidence wolter type i mirrors with a 10 m focal length , focuses x - rays in the 0.110 kev range .
the on - axis point - spread function of the hrma has a 50% energy width of 0.25 arcsec at 0.3 kev and 0.6 arcsec at 9.7 kev .
numerous imaging observations of cas a have been made by _
chandra _ as part of the mission s orbital activation and checkout ( oac ) calibration program .
all oac data are immediately in the public domain . most of the cas a observations were made using the advanced ccd imaging spectrometer ( acis ; burke et al .
1997 ) , which records both the sky position ( 0.49 arcsec / pixel ) and the energy ( @xmath15 50200 ev ) of each detected photon in the 0.110 kev range , with a time resolution of 3.2 s. a few of the observations were made with the high resolution camera ( hrc ; zombeck et al .
1995 ; murray et al .
1997 ) , which precisely records the sky position ( 0.13 arcsec / pix ) and arrival time4 ms for the data presented in this paper ( s. s. murray 2000 , private communication ; seward 2000 ) .
see 2.3 for further discussion .
] ( @xmath16=16 @xmath17s ) of each detected photon , but with modest ( @xmath18 ) energy resolution .
no diffraction gratings were in place for any of the acis or hrc observations .
a summary of the observations used in our analysis is given in table 1 .
ccccccccc 214&1999 aug 20 , 00:07&acis / s3&2.3 & 3.3 & 6107 & 2803 & 420 & @xmath19 + 220&1999 aug 22 , 23:27 & acis / s3 & 2.8 & 2.0 & 4279 & 1212 & 150 & @xmath20 + 221&1999 aug 23 , 00:52 & acis / s3 & 3.9 & 4.5 & 2103 & 1060 & 199 & @xmath21 + 222&1999 aug 23 , 01:38 & acis / s3 & 2.8 & 3.8 & 2080 & 1044 & 179 & @xmath22 + + 172 & 1999 sep 05 , 18:45 & hrc - s & 1.3 & 3.0 & 9485 & 9485 & 411 & @xmath23 + 1409 & 1999 oct 23 , 18:31 &
hrc - i & 1.4 & 3.0 & 12770 & 12770 & 479 & @xmath24 + as the analysis tools and calibration for _ chandra _ are still under development at this early stage of the mission , we will describe our data reduction , analysis , and verification steps in detail . in using the acis data to derive a spectrum for the point source in cas a
, we have restricted our analysis to the four exposures obtained during 1999 august 2023 with cas a placed on the back - illuminated ( bi ) acis s3 chip , for three reasons .
first , preliminary analysis showed that the point source spectrum was relatively soft , and the two bi chips ( s1 and s3 ) have superior low energy response compared to the front - illuminated ( fi ) chips .
second , we wished to optimize the _
chandra _ point spread function by minimizing the off - axis angle of the point source , and the _ chandra _ aim points for acis lie on either the s3 or i3 chips .
inspection of the cas a images obtained on the other acis chips shows that the point source is spread out over many more pixels , making it more difficult to separate from the background ; the effective area of the mirror and detector combination at these angles is also reduced .
finally , the bi chips have not suffered the radiation damage that degraded the energy resolution of the fi chips soon after launch , and so are better calibrated at this early stage in the mission .
a summary of the observations we used is given in table 1 .
we obtained the fully processed ( level 2 ) acis event data for these observations from the _ chandra _ data archive .
the data were acquired in timed - exposure / faint mode , with a frame read out every 3.2 s. a substantial number of frames was lost to telemetry saturation due to the high total count rate from cas a , resulting in an overall observation duty cycle of 42% .
we filtered the events from the surviving frames , accepting only those which fell within the `` standard '' event grade set ( grades 0 + 2 + 3 + 4 + 6 ) in order to maximize the ratio of x - ray to non x - ray events , and additionally discarding events with very large pulse heights as due to cosmic rays .
for all the observations , the acis focal plane temperature was @xmath25 c. we did not attempt to improve upon the spacecraft aspect solutions from the standard processing , but instead proceeded from the assumption that the central source in cas a is indeed point - like in morphology , as determined by tananbaum ( 1999 ) .
for each observation , we extracted all events located within a given radius of the centroid position of the point source .
ideally , we would use the calibrated angular response of the hrma to choose an extraction radius which encircled some fixed fraction ( e.g. , 95% ) of the flux from a point source for a given off - axis angle .
however , both the focus position of the detector and the quality of the spacecraft aspect solution were not necessarily optimal in these early observations . instead , we measured the radial distribution of source photons for each observation from the data and estimated a 95% extraction radius .
table 1 lists the aperture sizes .
the background in the point source region is dominated by the diffuse emission of the supernova remnant and has strong contributions from both continuum and line emission ( e.g. , holt et al .
the point source is located within a relatively low surface brightness region of the remnant .
still , the background must be estimated with care , as there are significant compositional gradients in this part of the remnant ( and perhaps variations in the underlying plasma conditions ) , resulting in varying line strengths with position ( see also hughes et al .
after investigating the line strengths in a number of nearby regions , we selected an off - centered @xmath26 arcsec@xmath27 rectangular region around the point source ( with the point source itself and a small surrounding buffer region excluded ) to estimate the background surface brightness near the point source . the point source and the background region
are shown in figure 1
. the acis instrumental response may be parameterized by a photon energy redistribution matrix ( hereafter rmf ) , which maps incident photon energy to detected pulse height amplitudes ( phas ) . every acis chip ( @xmath28 )
is divided into four parallel ( @xmath29 ) quadrants , each read out separately by four individual amplifiers .
each quadrant requires a unique rmf , as each readout amplifier has a different gain .
additionally , the parallel and serial charge transfer inefficiency in the bi chips leads to differences in energy resolution and detection efficiency versus position within a quadrant as well .
consequently , the acis instrument team has supplied multiple position - dependent rmfs for each quadrant .
we used the 1999 october 28 version of rmfs developed for observations taken at a focal plane temperature of @xmath25 c. fortuitously , the cas a point source ( and nearly all of the selected background region ) lies entirely in quadrant 2 for all four of the s3 observations .
furthermore , the four observations were located within 250 pixels of each other within the chip quadrant , minimizing the differences in ccd response . thus , the four rmfs appropriate for each observation may be combined without any serious loss of accuracy .
this allows us to sum the four individual s3 count spectra into a single `` grand total '' s3 count spectrum with improved statistics .
in addition to the rmfs , an ancillary response function ( hereafter arf ) is needed to characterize the effective area of the hrma and the quantum efficiency of the acis detector as a function of incident photon energy and off - axis angle .
we computed the appropriate arf for each observation using software created by members of the acis and high energy transmission grating instrument teams at mit .
the overall instrumental response is given by the product of the rmf and the arf .
we computed this product for each observation , and then combined these products by an exposure - weighted average for use with our summed count spectrum .
the summed , background - subtracted acis count spectrum was analyzed using the xspec v10.00 software package ( arnaud 1996 ) .
we rebinned the data at 0.15 kev resolution ( 32 adu / channel ) and restricted our analysis to 21 bins in the 0.74.0 kev range , resulting in at least 10 count bin@xmath30 in all but 3 bins at high energy ( with a minimum of 5 count bin@xmath30 and a maximum of 71 count bin@xmath30 ) , and a total of 630 counts .
the rebinned count spectrum is shown in figure 2 . in light of the relatively low count rate ,
our model fitting weighted each bin by @xmath31 rather than @xmath32 ( where @xmath33 is the number of counts in bin @xmath34 ) .
this is a superior weighting scheme when @xmath33 is small and asymptotes to the usual weighting when @xmath33 is large ( gehrels 1986 ) .
we fit several different spectral models to these data : a simple power law , thermal bremsstrahlung , an ideal blackbody ( bb ) , and two modified bb models .
the modified bb models assume that the emitting object is a ns and account for the effect of a light - element stellar atmosphere on the emergent spectrum .
the strong ns surface gravity essentially guarantees that the photosphere will be dominated by the lightest element present ( alcock & illarionov 1980 ; romani 1987 ) . for such atmospheres with @xmath35
k , the @xmath36 dependence of free - free absorption will shift the peak of the emission blueward from that of an ideal bb ( romani 1987 ; zampieri et al .
previous authors have shown that neglecting this effect and fitting the wien tails of such spectra with an ideal bb model significantly overestimates the effective temperature and underestimates the emitting area ( rajagopal & romani 1996 ; zavlin , pavlov , & shibanov 1996 ; rutledge et al .
1999 ) . to consider this possibility , we employed two models : the simple analytic spectrum emerging from the power - law atmosphere of heyl & hernquist ( 1998a ; hereafter hh98a ) , with @xmath37 as appropriate for a light - element atmosphere ; and the more detailed h atmosphere model of zavlin , pavlov , & shibanov ( 1996 ; hereafter zps96 ) , developed for a weakly magnetized ns .
for all these models , we included the effect of photoelectric absorption by neutral gas along the line of sight ( morrison & mccammon 1983 ) , which is a significant effect at low ( @xmath38 kev ) energies . with the relatively small number of low - energy x - ray photons detected from the cas a point source , it is difficult to constrain @xmath39 precisely .
moreover , its value in this situation is highly spectral - model dependent and strongly covariant with the overall flux normalization .
however , @xmath40 may also be determined from radio measurements of the atomic and molecular hydrogen column densities using the 21 cm and 18 cm oh absorption lines , respectively .
keohane , rudnick , & anderson ( 1996 ) have used such data to derive a spatially - resolved column density map of the cas a snr at 30 arcsec resolution . using this map
, we estimate that the column density towards the cas a point source is @xmath41 @xmath42 .
however , this estimate must be used with caution , as the radio maps give no information on small angular - scale variations in the column . for each spectral model we employed ,
we fit the data both with @xmath40 held fixed at this value and with @xmath39 as a free parameter .
our spectral fitting results are summarized in table 2 , and a typical model fit is shown in figure 2 .
all of the models gave formally acceptable fits values are less than unity in some cases does not imply that the measurement uncertainties ( given by poisson statistics ) are underestimated .
the probability of obtaining a reduced @xmath43 as small as 0.7 with only 19 degrees of freedom ( assuming the true model is being fit ) is 18% , which is thus within the 1@xmath0 range for an acceptable fit . ] ; we are unable to reject any of them on the basis of these data . however , the bb and ns atmosphere models fit slightly better to the data than the other models , although with inferred radii considerably smaller than the 10 km typically assumed for a ns radius .
llcccccc power law & @xmath44 & @xmath45 & @xmath46 & & & @xmath47 & 1.12/18 + & 1.1 ( fixed ) & @xmath48 & @xmath49 & & & @xmath50 & 1.44/19 + + thermal brems . &
@xmath51 & & @xmath52 & @xmath53 & & @xmath54 & 0.92/18 + & 1.1 ( fixed ) & & @xmath55 & @xmath56 & & @xmath57 & 1.00/19 + + ideal blackbody & @xmath58 & & & @xmath59 & @xmath60 & @xmath61 & 0.69/18 + & 1.1 ( fixed ) & & & @xmath62 & @xmath63 & @xmath64 & 0.77/19 + + ns atm .
( hh98a ) & @xmath65 & & & @xmath66 & @xmath67 & @xmath68 & 0.70/18 + & 1.1 ( fixed ) & & & @xmath69 & @xmath70 & @xmath71 & 0.76/19 + + ns atm .
( zps96 ) & @xmath72 & & & @xmath73 & @xmath74 & @xmath75 & 0.74/18 + & 1.1 ( fixed ) & & & @xmath76 & @xmath77 & @xmath78 & 0.73/19 + interpretation of our results depends critically upon the reliability of the energy and effective area calibrations of _ chandra _ , which are still being evaluated by the instrument teams . to verify the robustness of our conclusions
, we made a rough check of these calibrations by fitting the x - ray spectral data from the 1999 aug 23 acis / s3 observation of snr e0102@xmath7972 ( obsid 1231 ) , the brightest supernova remnant in the small magellanic cloud ( smc ; hayashi et al .
1994 ) . for comparison
, we also analyzed archival spectral data for the same source as observed by _
asca _ on 1993 may 1213 . for both data sets , we restricted our analysis to 0.62.6 kev ( an energy band that both contains the majority of the snr flux and spans a range similar to that of the cas a point source ) and fit the same plasma model , with fixed column density and non - solar abundances .
the derived plasma temperatures agree to better than 2% , and the overall normalizations to within 13% , using the preliminary calibrations .
we examined two on - axis observations of cas a made with the hrc as part of the oac program , using the hrc - s and hrc - i detectors respectively .
these observations are summarized in table 1 .
we obtained the processed level 2 event data from the _ chandra _ data archive . in both cases ,
the overall count rate ( @xmath80 count s@xmath30 ) from cas a was below the telemetry threshold ( @xmath81 184 count s@xmath30 ) where deadtime effects become significant . for each observation
, we extracted all events within 3 arcsec of the point source centroid to conservatively ensure that all source photons were included .
the photon arrival times , provided in terrestrial time ( tt ) at the spacecraft , were corrected to barycentric dynamical time ( tdb ) at the solar system barycenter using the jpl de200 solar system ephemeris ( standish et al . 1992 ) and a geocentric spacecraft ephemeris .
we binned the events into 0.5 ms time bins and computed a fourier power spectrum of the resulting time series .
no significant pulsations were detected ; the highest peak in the power spectrum had a significance of only 2@xmath0 when the number of trials is accounted for . to improve our statistics , we also computed an incoherent power spectral sum of 9 ks segments of each observation .
again , no significant pulsations were found , with the highest peak having a significance of only 0.8@xmath0 including the number of trials .
a large pulse frequency derivative might spread a coherent pulsation over multiple power spectral bins , reducing our sensitivity .
however , for the @xmath82 ks observations used here , a pulse frequency derivative of @xmath83 hz s@xmath30 would be required .
this is 25 times larger than @xmath84 for the crab pulsar , and @xmath85 larger than the typical value for an axp or sgr . to enhance our sensitivity to non - sinusoidal pulse shapes
, we also performed harmonic folds of the power spectrum ( see , e.g. , de jager , swanepoel , & raubenheimer 1989 ) .
no significant pulsations were detected .
after our initial analysis was completed , we learned of a wiring problem in the hrc on - board electronics that causes hrc events to be time - tagged incorrectly ( s. s. murray 2000 , private communication ; seward 2000 ) .
the net effect of the problem is that the timing accuracy of the hrc is degraded from 16 @xmath17s to @xmath864 ms .
consequently , we conservatively conclude that the hrc analysis presented here has no sensitivity to pulsations with frequency above 50 hz .
we used our incoherent power spectral sum to estimate an upper limit for the sinusoidal pulsed fraction of the cas a point source , accounting for the suppression of power spectral sensitivity at higher frequencies due to binning of the data ( e.g. , van der klis 1989 ) .
we find that the 3@xmath0 upper limit on the sinusoidal pulsed fraction is @xmath135% for @xmath87 hz .
our sensitivity to rapid pulsations depends upon an accurate correction for the motion of the spacecraft with respect to the barycenter . as a check of our barycenter corrections and of the spacecraft ephemeris
, we analyzed a 1999 august 31 hrc - i observation ( obsid 132 ) of psr b054069 , a young 50 ms pulsar associated with the supernova remnant n158a in the large magellanic cloud , using the same data analysis procedure as for cas a. ( detailed independent analyses of this observation are presented by gotthelf & wang 2000 and kaaret et al .
we compared this measurement with a 1999 september 1 observation of an overlapping field with the _ rossi x - ray timing explorer _ ( these data were generously made available by f. e. marshall of nasa goddard space flight center ) .
the frequencies measured in the two data sets were consistent within the uncertainties , and also agreed with the value extrapolated from a timing model based on a 1996 _ bepposax _ observation ( mineo et al .
we have shown that the x - ray point source in cas a has a spectrum well - described by either an absorbed power law with photon index 2.83.6 and unabsorbed 0.110 kev luminosity ( 7160)@xmath88 erg s@xmath30 , or an absorbed bb or modified bb with @xmath890.5 kev and bolometric luminosity @xmath905)@xmath88 erg s@xmath30 .
our _ chandra _ spectral and timing measurements , combined with pre - existing limits at other wavelengths , severely constrain plausible models for the nature of the x - ray point source in cas a. we begin by pointing out , for completeness , that our steep power law spectral fit essentially rules out the possibility that the point source is a background galaxy , as active galactic nuclei ( agn ) typically lie within a range of photon indices @xmath912.2 ( turner & pounds 1986 ) .
the agn scenario is also extremely implausible given that the point source is located within a few arcsec of the expansion center of the snr ( tananbaum 1999 ) . moreover , given the surface density of agns on the sky at this flux level ( boyle et al .
1993 ) , the probability of a chance coincidence is negligible . there is thus extremely strong inferential evidence that the point source is indeed associated with the supernova remnant .
we now summarize the implications of our results for several other interpretations .
if the cas a point source is a classical young pulsar ( the conventional product expected for a type ii sn explosion ) , then the x - ray radiation should be predominantly non - thermal power - law emission from relativistic acceleration of @xmath92 pairs in the corotating ns magnetosphere ( see romani 1996 ) . in table 3 ,
the properties of the point source are compared with the six young ( @xmath93 yr ) classical pulsars whose x - ray spectra are well measured .
all six of these pulsars have power law x - ray spectra with @xmath941.7 .
the clearest distinction is that the spectral shape of the cas a point source is considerably steeper than that of the young pulsars , although its x - ray luminosity is marginally consistent with the lower end of the young pulsar range .
the upper limit on the point source s x - ray pulse fraction ( @xmath95% ) is lower than at least three of the measured pulsed fractions in the young pulsars . in the other three pulsars ,
the _ lower _ limit on the pulsed fraction is consistent with the point source s upper limit , but it is likely that these pulsar lower limits are a drastic underestimate due to the presence of plerionic emission .
radio pulsations have not been detected from cas a , with an upper limit of @xmath96 mjy kpc@xmath27 for the 600 mhz radio luminosity ( lorimer , lyne & camilo 1998 ) , in contrast to five of the six young pulsars .
this non - detection in cas a could be explained by beaming , judging from the observed radio luminosities in young pulsars ( @xmath97 mjykpc@xmath27 for the crab and psr b054069 , but only @xmath98 mjykpc@xmath27 for psr b150958 , psr j11196127 , and psr j16175055 ; taylor , manchester , & lyne 1993 ; kaspi et al .
1998 , 2000 ) .
images also show no obvious evidence for a plerion surrounding the point source .
plerions , powered by synchrotron emission , have been detected around five of the six classical young pulsars in table 3 ( the exception is psr j16175055 ; kaspi et al .
1998 ) , and even around some x - ray point sources in snrs which are not known pulsars . based principally on the x - ray spectral shape , with some support from the other properties ,
we conclude that the point source in cas a is not a classical young pulsar .
llccccccccccc + point source & cas a & 0.0 & & /2.5 & 3.4 & 2.23.6 & 33.834.6 & 0.5 & 0.5 & 33.3 & @xmath95 & + + + 1e 0820@xmath794247 & pup a & 0.1 & 0.075 ? & 3.9/3.6 & 2.0 & & & 0.28 & 2 & 33.6 & 20 ? &
13 + 1e 1614@xmath795055 & rcw 103 & 0.0 & & /3.1 & 3.1 & & & 0.56 & 0.81 & 33.9 & & 4,5 + 1e 1207@xmath795209 & pks 1209@xmath7952 & 0.2 & 0.424 & /3.8 & 1.5 & & & 0.25 & 1.1 & 33.1 & & 68 + + + psr j1846@xmath790258 & kes 75 & @xmath99 & 0.325 & 2.9/3.0 & 19 & 1.1 & @xmath100 & & & & @xmath101 & 9 + psr b0535@xmath10221 & crab & 0.0 & 0.033 & 3.1/3.0 & 2.0 & 1.74 & 36.2 & & & & @xmath103 & 10 + psr b1509@xmath7958 & msh 15@xmath795_2 _ & 0.2 & 0.150 & 3.2/3.2 & 5.2 & 1.36 & 35.3 & & & & 65 & 11,12 + psr b0540@xmath7969 & n158a&0.2 & 0.050 & 3.2/3.1 & 47 & 1.3 & 36.6 & & & & @xmath104 & 13,14 + psr j0537@xmath796910 & n157b&0.5 & 0.016 & 3.7/3.7 & 47 & 1.6 & 35.5 & & & & @xmath105 & 1517 + psr j16175055 & & & 0.069 & 3.9/ & 6.5 & 1.6 & @xmath106 & & & & @xmath8650 & 18,19 + + 1e 1841@xmath79045 & kes 73 & 0.2 & 11.8 & 3.6/3.3 & 7.0 & 3.4 & 36.9 & & & & 15 & 20,21 + 1e 1048.1@xmath795937 & & & 6.45 & 3.7?/ & 3.0 & 2.5 & 34.4 & 0.64 & 0.6 & 33.9 & 70 & 22,23 + ax j1845@xmath790258 & g29.6 + 0.1 & 0.2 & 6.97 & ?
/@xmath13.9 & 8.5 & 4.6 & 38.6 & 0.72 & 1.6 & 34.9 & 50 & 24,25 + rxs j1708@xmath7940 & & & 11.0 & 4.0/ & 10.0 & 2.9 & 36.9 & 0.41 & 8.9 & 35.5 & 50 & 2628 + 4u 0142@xmath10261 & & & 8.69 & 4.8?/ & 1 & 3.7 & 35.9 & 0.39 & 2.4 & 33.5 & 10 & 29,30 + 1e 2259@xmath102586 & ctb 109 & 0.3 & 6.98 & 5.3/4.0 & 4.0 & 4.0 & 36.9 & 0.43 & 2.2 & 34.3 & 30 & 28,31 + + + sgr 1900@xmath10214 & g42.8 + 0.6 ?
& 1.2 & 5.16 & 2.8?/4.0 & 5.0 & 1.1 & 34.5 & 0.51 & 1.4 & 34.2 & 11 & 32,33 + sgr 1806@xmath7920 & & 0.0 & 7.47 & 3.1?/ & 14.5 & 2.2 & 36.0 & & & & 23 & 3436 + sgr 0526@xmath7966 & n49 & 0.6 & 8 & ?
/3.7 & 47 & & & & & & @xmath166 & 37,38 + sgr 1627@xmath7941 & [email protected] ? & 2 & 6.41 ? & ?
/3.7 & 11.0 & 2.5 & 35.8 & & & & 10 & 3941 + if the cas a point source is a ns , the absence of both detectable radio pulsations and a synchrotron nebula may indicate that it lies beyond the so - called pulsar `` death line '' , an empirical boundary on the spin - period magnetic - field plane beyond which radio pulsars have generally not been detected , presumably because the ns does not generate enough @xmath92 pairs to power significant non - thermal emission s , @xmath107 g ) indicates that this argument must be used cautiously ( young , manchester , & johnston 1999 ) . ]
( chen & ruderman 1993 ) . in this case
, we would expect a strongly magnetized ( @xmath108 g ) ns to have a spin period of order at least a few seconds .
conversely , a rapidly spinning ns ( @xmath109 ms ) would have a very weak magnetic field ( @xmath110 g ) , perhaps consistent with delayed field growth ( blandford , applegate & hernquist 1983 ) .
an alternative interpretation is that the x - rays from the cas a point source arise from thermal emission from a cooling ns .
a 300 yr old ns cools primarily via neutrino emission ; standard cooling models predict thermal photon emission from the surface with @xmath1110.25 kev ( page & sarmiento 1996 ; page 1998 ) .
all of our bb and modified bb fits for the cas a point source yield somewhat higher temperatures ( @xmath1120.5 kev ) , as well as much smaller bb radii ( @xmath1132.6 km ) than expected for a 10 km ns , even when accounting for a light - element atmosphere .
the ns atmosphere models ( hh98a , zps96 ) that we fit to the data were computed assuming a weak ( @xmath114 g ) magnetic field , which may be a poor assumption for a young ns .
qualitatively , however , the presence of a magnetic field of order @xmath115 g will shift the peak of the emission in a light - element atmosphere _ redward _ towards the ideal bb case ( pavlov et al . 1995 ) , thus exacerbating the discrepancy between the data and standard ns cooling curves .
we note , however , that the behavior of spectral shifts in ultrastrong magnetic fields ( @xmath8@xmath9 g ) has not yet been calculated .
while our inferred temperature is marginally consistent with standard ns cooling curves , the small emitting area remains problematic , especially given our limits on the x - ray pulsed fraction .
strong ( @xmath116@xmath117 g ) magnetic fields will produce a non - uniform temperature distribution on the surface of a ns , owing to anisotropic electron conduction through the star s outer envelope ( e.g. greenstein & hartke 1983 ; heyl & hernquist 1998b , 2000 ) .
however , the resulting temperature variation across the surface does not produce small hot spots , but is instead smoothly varying with a local flux roughly @xmath118 , where @xmath119 is the polar angle from the magnetic axis .
this would reduce the effective area of the emitting surface by a factor @xmath120 , far less than the factor of 30100 required by our spectral fits .
similar conclusions motivated pavlov et al .
( 2000 ) to consider a model in which the magnetic polar caps are intrinsically hotter than the bulk of the stellar surface , as a result of horizontal chemical abundance gradients through the conductive envelope .
light element envelopes transmit heat more readily than ones made of heavy elements ( chabrier , potekhin & yakovlev 1997 ; heyl & hernquist 1997a ) , so hot polar caps consisting of hydrogen embedded in a cooler iron crust can , in principle , yield an emitting area consistent with the spectral fits ( pavlov et al .
however , such a model predicts that the emission should be pulsed at the rotation period of the star , unless either the magnetic and rotation axes are nearly aligned , or the line of sight nearly coincides with the rotation axis .
pulsed fractions at the level of 1070% required to account for the putative thermal emission from middle aged radio pulsars ( e.g. becker & trumper 1997 ) and axps ( e.g. , mereghetti 2000 ) can be produced from _
smoothly _ varying properties of ns envelopes , such as anisotropic heat conduction ( heyl & hernquist 1998b ) or directionally dependent opacities ( e.g. pavlov et al .
1994 , zavlin et al .
1995 , shibanov et al .
1995 ) , even when gravitational bending of light is included ( e.g. page 1995 , heyl & hernquist 1998b ) . an even larger pulsed fraction will result , in general , from the hot spot model .
for example , in the case of the orthogonal rotator model of pavlov et al . ( 2000 ) , we estimate typical maximum to minimum flux variations @xmath121 for @xmath122 nss with radii @xmath123 km .
this is in severe disagreement with the upper limits we obtain for the pulsed fraction .
observationally , we can not yet exclude the case of a nearly aligned rotator , but it is not clear that the horizontal abundance gradients required by the pavlov et al .
model will be stable for long times in the liquid crust .
we now consider the possibility that the point source in cas a is powered by accretion onto a ns or a bh .
this possibility was also raised by umeda et al .
( 2000 ) and pavlov et al .
we assume that the accretion is fed by fallback material left over after the original supernova explosion ( e.g. chevalier 1989 ) .
we prefer such a model to a binary accretion model since there is no optical / ir evidence for a such binary companion star ( van den bergh & pritchet 1986 ) .
a very low - mass dwarf companion might have evaded detection , but such a companion would be unlikely to remain bound in the binary following the supernova explosion , given the high mass of the cas a progenitor ( brandt & podsiadlowski 1995 ; kalogera 1996 ) .
we begin with the possibility of accretion onto a ns . if the accretion occurs via a thin disk extending down to the marginally stable orbit at @xmath124 , or the ns surface ( whichever is larger ) , then we expect significant emission from an equatorial boundary layer where the accreting material meets the star .
the emitting zone is expected to have a radial extent roughly equal to the local scale height of the disk ( narayan & popham 1993 ; see also inogamov & sunyaev 1999 for a recent discussion of boundary layer models ) .
if the boundary layer is optically thick , then it will emit bb radiation .
for an accretion luminosity @xmath125 , the scale height is @xmath126 km , and the effective area of the radiating zone is @xmath86few km@xmath27 .
this is in reasonable agreement with the effective area @xmath86km@xmath27 determined from fitting the chandra data ( table 2 ) .
we consider this a viable model , although we note that the optical thickness of such boundary layers is not well understood .
this boundary layer model requires that the ns have a very weak magnetic field .
specifically , the magnetospheric radius @xmath127 has to be smaller than the stellar radius , which implies that @xmath128 g for the estimated x - ray luminosity .
if the magnetic field is somewhat stronger , then the disk would be terminated at @xmath127 . even in this case
, we might expect a boundary layer to develop at @xmath127 , and the model would be consistent with the observations for fields @xmath129 g. we note that for sufficiently large @xmath130 , accreting material is centrifugally expelled from the system at @xmath127 ( illarionov & sunyaev 1975 ) . in this case ,
very little material would reach the neutron star and there would be negligible x - ray emission ( e.g. , chatterjee , hernquist , & narayan 2000 ; alpar 1999 , 2000 ) .
if the ns instead has a magnetic field @xmath131 g typical of young radio pulsars , then boundary layer emission is unlikely to explain the observed x - rays . in this case
, we would have to assume that the accreting material is able to reach the surface of the neutron star , which would require that the neutron star be spinning quite slowly ( in order to avoid centrifugal expulsion of matter ) .
the effective area of the radiating zone would again be small , and could thus be consistent with the observations .
however , we would predict strong coherent pulsations in the x - ray signal .
our upper limit of 25% on the pulsed fraction from the cas a point source does not rule out pulsations at the level observed in many known @xmath5 g accreting pulsars .
more sensitive searches for pulsation would be very worthwhile .
a third kind of accreting ns model is one in which the accretion disk is truncated at a large transition radius , and the flow farther in occurs via an adaf ( as in the narayan et al .
1997 model for low - luminosity bh binaries ) .
boundary layer emission in such a model is not well understood , so we are not in a position to predict the emission spectrum .
however , if the adaf is terminated at @xmath127 and the material then flows onto the magnetic poles , the spectrum would be similar to the case with a thin disk which was discussed above .
accreting bh models face considerably greater difficulty in fitting the observations .
this is because neither a boundary layer nor channeled accretion onto magnetic poles is expected .
therefore , any blackbody emission should be primarily from the inner accretion disk , with an effective area of several times @xmath132 , where @xmath133 is the schwarzschild radius of the black hole . for a @xmath134 bh ,
the area is @xmath135 km@xmath27 , which is clearly inconsistent with the x - ray data .
this means that the x - ray emission from a bh point source in cas a would have to originate either via compton scattering in an optically thin corona over a thin disk , or via optically thin bremsstrahlung emission from a hot adaf .
we have explored models of this kind , using the modeling techniques described in narayan et al .
( 1997 ) and quataert & narayan ( 1999 ) , but we find that we need to fine - tune the models to an uncomfortable degree to fit the observations .
it is important to note that accretion models are severely constrained by the observed flux ratio between x - ray and optical bands .
the optical limits on a stellar remnant in cas a are @xmath136 and @xmath137 ( van den bergh & pritchet 1986 ) . applying the extinction corrections
estimated by these authors , the x - ray to optical flux ratio is @xmath138 .
although this is typical for bright x - ray binaries , it is considerably larger than the ratios observed in quiescent accreting bhs and nss with luminosities comparable to the cas a point source , e.g. @xmath139 for the bh system a062000 and @xmath140 for the ns system cen x-4 ( mcclintock & remillard 2000 ) .
both sources have an x - ray luminosity lower than the point source in cas a , and both have optical luminosities significantly higher than the upper limit for the cas a source .
if the spectra of these sources are characteristic of accreting bhs and nss at low luminosities , then accretion models for cas a are ruled out with high confidence .
it is possible , however , to evade this conclusion by arguing that these low - mass x - ray binaries ( lmxbs ) are a poor comparison to the cas a point source .
first of all , the quiescent optical emission in some bh binaries has been modeled as synchrotron emission from a hot advection - dominated accretion flow ( adaf ) close to the bh ( narayan , mcclintock & yi 1996 ; narayan , barret & mcclintock 1997 ) .
however , if the emission instead comes from a `` hot spot '' ( e.g. , where the accretion stream from the mass donor hits the thin accretion disk ) , then it is specific to binary systems and should be absent in systems with a fallback disk .
this possibility has not been modeled in any detail .
also , x - ray reprocessing in the accretion disk is a significant source of optical emission in lmxbs ( van paradijs & mcclintock 1995 ) , especially since the outer disks evidently subtend a large semi - angle ( @xmath141 ) at the central object ( de jong , van paradijs , & augusteijn 1996 ) .
the reason for such large angles may be warping of the outer disk ( see dubus et al .
1999 ; esin et al .
but if such warps are induced primarily by binary effects , then a fallback disk in cas a might not be similarly warped ( although it has been argued that irradiation alone can lead to significant warping ; see petterson 1977 ; pringle 1996 ; maloney et al .
for an unwarped disk , the subtended semi - angle would be set by the disk s scale height at the relevant radius ( @xmath142 cm ) . by employing an analysis similar to that in perna , hernquist & narayan ( 1999 )
, we estimate that the re - emitted optical flux in such a model would be ( just ) below the optical flux limits for the cas a source .
thus , the optical limits strongly constrain accretion models , but do not yet conclusively eliminate them . by comparing the predictions of the models with the x - ray observations , specifically the small effective area @xmath143 of the emission
, we conclude that models with accreting nss are more promising than those with bhs .
it is interesting to compare the properties of the cas a point source with three classes of objects whose nature remains puzzling : axps , sgrs , and the radio - quiet non - plerionic x - ray point sources in snrs . both axps and sgrs pulse with slow spin periods despite being young objects ( based on their association with snrs ) .
we summarize the x - ray properties of these objects in table 3 .
the x - ray spectrum of most axps is best characterized by a two - component spectrum consisting of a @xmath144 kev bb and a steep ( @xmath1454 ) power law extending out to 10 kev , with comparable luminosity in both components .
the bb components have @xmath1464 km and @xmath147@xmath148 erg s@xmath30 , quite similar to the bb fit for the cas a point source .
( we note that , due to poor photon statistics in the present data , we would be unable to detect the presence of an additional power - law component with this bb fit if the cas a point source has an axp - like two - component spectrum . )
while four of the axps have large pulsed fractions ( 3070% ) , two have much lower pulsed fractions ( 1015% ) which are consistent with the non - detection of pulsations in cas a. however , the x - ray luminosity of the cas a source is at least a factor of three to ten lower than that of the axps , in spite of its relative youth .
overall , the properties of the cas a point source are roughly consistent with being an underluminous axp
. this possibility can be tested by a deeper search for long - period x - ray pulsations .
the quiescent emission of sgrs may be fit with a power law with @xmath942.5 and a 210 kev luminosity of ( 3100)@xmath149 erg s@xmath30 .
these spectral parameters are also consistent with those measured for cas a. however , we note that the proposed associations between sgrs and snrs are somewhat tenuous , in contrast to those proposed for axps ( gaensler 2000 ) .
moreover , no soft gamma - ray bursts have ever been detected from the direction of cas a. finally , there are the three radio - quiet x - ray point sources that each lie near the center of an snr and have no evidence for a plerion . the x - ray properties of these sources are summarized in table 3 . the x - ray emission from these objects
is well characterized by a bb spectrum with @xmath1500.6 kev and @xmath1513 km , comparable to what we measure in cas a. from a spectral point of view , these four sources form a remarkably homogeneous group .
pulsations ( @xmath152 ms ) have recently been detected from the source in pks 120952 ( zavlin et al .
2000 ) , and possibly from the source in pup a ( pavlov , zavlin , & trmper 1999 ) ; in both cases , the measured pulsed fraction is below the limit placed in cas a. however , the point source in rcw 103 has shown order of magnitude variability in its flux ( gotthelf , petre , & vasisht 1999a ) , with a possible 6 hr periodicity recently reported ( garmire et al .
this is in stark contrast to the cas a point source , which shows no evidence for strong variability on either short ( @xmath86hours weeks ; see table 1 ) or long ( @xmath86years ; see pavlov & zavlin 1999 ; pavlov et al .
2000 ) time scales .
at present , we do not have a unique model to account for the observed properties of the x - ray point source in cas a. thermal emission from an isolated , cooling ns can account for the data , but only if unconventional modifications are incorporated into spectral models in order to satisfy the requirement that the emitting area be small
. localized hot spots can , in principle , be produced on the surface of a ns by horizontal chemical abundance gradients in the liquid envelope ( pavlov et al .
however , such hot spots should give rise to a significant pulsed component to the emission , in general , and such models will likewise be constrained by future limits or detections of x - ray modulation .
accretion onto a bh does not appear promising , but ns accretion provides a viable mechanism for explaining the characteristics of the x - ray emission .
if the ns is very weakly magnetized , the observed emission could arise from a boundary layer .
otherwise , the x - rays could be produced by magnetically - channeled accretion onto the ns . in the latter case
, we expect pulsed emission , so again , future timing limits test this interpretation . the point source in cas
a is similar in many respects to the axps and quiescent sgrs , which have been interpreted as ultramagnetized neutron stars ( `` magnetars '' ; duncan & thompson 1992 ; thompson & duncan 1996 ; heyl & hernquist 1997a , 1997b ) or as neutron stars with normal magnetic fields of order @xmath153@xmath117 g accreting from fallback disks ( chatterjee et al . 2000 ; alpar 1999 , 2000 ) .
it is important to note the detailed thermal emission from neutron stars with magnetic fields @xmath8@xmath9 has yet to be calculated .
conceivably , x - ray spectra in this regime could account for the small emitting areas of the cas a point source and the axps without requiring small polar hot spots .
regardless of the eventual resolution to the puzzling aspects of the point source in cas a , it is already clear that the mere existence of this and similar objects demands a dramatic revision of our generally accepted notions of the nature of compact objects found in supernova remnants .
the discovery of radio pulsars in the 1960s led to a paradigm in which supernovae generally leave behind strongly magnetized , rapidly rotating neutron stars . the subsequent failure to locate radio pulsars in the majority of snrs
has been a long - standing problem for this point of view ( kaspi 2000 , gotthelf & vasisht 2000 ) .
the unanticipated properties of the cas a point source and the relative youth of the cas a snr imply that the birthrate of `` unusual '' compact objects is likely to be at least roughly comparable to that of radio pulsars , potentially resolving the difficulties posed by snr / pulsar associations . in this sense
, the identification of the point source in cas a may be as significant to our understanding of neutron stars as was the original discovery of radio pulsars .
shibanov , y. a. , pavlov , g. g. , zavlin , v. e. & tsuruta , s. 1995 , in 17th texas symposium on relativistic astrophysics and cosmology , ed .
h. bhringer , g. e. morfill & j. e. trmper ( new york : ny acad . of sci . ) , 291 | the spectacular `` first light '' observation by the _ chandra x - ray observatory _ revealed an x - ray point source near the center of the 300 yr old cas a supernova remnant .
we present an analysis of the public x - ray spectral and timing data .
no coherent pulsations were detected in the _
chandra_/hrc data .
the 3@xmath0 upper limit on the pulsed fraction is @xmath135% for @xmath2 ms .
the _ chandra_/acis spectrum of the point source may be fit with an ideal blackbody ( @xmath3=0.5 kev ) , or with blackbody models modified by the presence of a neutron star atmosphere ( @xmath3=0.250.35 kev ) , but the temperature is higher and the inferred emitting area lower than expected for a 300 yr old neutron star according to standard cooling models .
the spectrum may also be fit with a power law model ( photon index @xmath43.6 ) .
both the spectral properties and the timing limits of the point source are inconsistent with a young crab - like pulsar , but are quite similar to the properties of the anomalous x - ray pulsars .
the spectral parameters are also very similar to those of the other radio - quiet x - ray point sources in the supernova remnants pup a , rcw 103 , and pks 120952 .
current limits on an optical counterpart for the cas a point source rule out models that invoke fallback accretion onto a compact object if fallback disk properties are similar to those in quiescent low - mass x - ray binaries . however , the optical limits are marginally consistent with plausible alternative assumptions for a fallback disk . in this case
, accreting neutron star models can explain the x - ray data , but an accreting black hole model is not promising . |
the regge - newton integral equation of the cox - thompson method @xcite for the transformation kernel reads as @xmath0 with the input symmetrical kernel defined as @xmath1 here @xmath2 and @xmath3 means , respectively , the regular and irregular riccati - bessel functions defined as @xmath4 , @xmath5 , and the explicit expression holds for the @xmath6 numbers : @xmath7}{\prod_{l'\in s , l'\neq l}[l(l+1)-l'(l'+1)]},\qquad l\in s,\ ] ] with @xmath8 , @xmath9 , @xmath10 and @xmath11 @xcite . let @xmath12 denote the set of zeros of the determinant @xmath13 with @xmath14_{ll}=\frac{u_l(x)v'_l(x)-u'_l(x)v_l(x)}{l(l+1)-l(l+1)}.\ ] ] in ref .
@xcite it is proved that equation ( [ gl ] ) is uniquely solvable for @xmath15 and the elements of @xmath12 are isolated points .
therefore the continuous solution of equation ( [ gl ] ) ( if it exists ) is unique . in ref .
@xcite it has been shown that @xmath16 is not integrable near @xmath17 therefore the potential @xmath18 corresponding to the schrdinger equation has poles of order ( at least ) 2 at these isolated points @xmath19 . such potentials are not in @xmath20 and we call them singular potentials
. to get non - singular potentials by the cox - thompson method is thus in an intimate connection with the uniqueness of solution of equation ( [ gl ] ) . from now on the treatment is restricted to the one - term limit , i.e. , to the case when @xmath21 .
such a case represents a natural first step compared to the uniqueness solution of the ct method with finite number of phase shifts .
also , the one phase shift case is closely related to the phenomenon of quantum resonance scattering ( when the resonance - like enhancement of the total cross section is mainly determined by a single partial wave ) or the ramsauer - townsend effect ( when the electron - atom interaction is governed mainly by the @xmath22-wave phase shift ) @xcite . in the one - term limit , the numerator of equation ( [ determ ] ) becomes the wronskian @xmath23 to ensure a unique solution of the regge - newton integral equation ( [ gl ] ) , we shall establish a condition for @xmath24 .
this is also the condition for constructing a non - singular potential @xmath25 at the one - term level @xmath21 .
let @xmath26 and @xmath27 , @xmath28 . in order to get a potential that belongs to the class @xmath20 we shall prove the next statement . @xmath29 .
first we prove that there exists @xmath30 such that @xmath31 if @xmath32 .
let @xmath33 with @xmath34 .
then the different sign of the wronskian at the origin @xmath35>0 $ ] and at the infinity @xmath36 < 0 $ ] clearly signals the existence of at least one zero position @xmath19 for which @xmath37 because of the continuity of @xmath38 . for the uncovered region of @xmath39 with @xmath40 we shall use the standard notation for the @xmath41th zeros @xmath42 of the bessel functions @xmath43
let now @xmath44 .
we term _ regular _ sequence of zeros if the following interlacing holds for the @xmath41th and @xmath45th zeros : @xmath46 it is a simple matter to see that the local extrema of @xmath38 within the interval @xmath47 possess the same sign in case of regular sequence interlacing .
this is because at the extremum positions @xmath48 and @xmath49 of @xmath38 the wronskian simplifies to @xmath50 now , in case of any deviation from this regular sequence , e.g. , when an _ irregular _ sequence @xmath51 is first encountered at a particular @xmath52 , one gets different signs for the two consecutive extrema of the wronskian at @xmath53 and @xmath54 , respectively .
this assumes the appearance of a zero position of @xmath38 within the region @xmath55 . in summary , observing regular sequences of interlacing for all @xmath56 is equivalent to the absence of roots of @xmath38 . to see that in the considered region such deviation from the regular sequence interlacing happens we present the following argument
let @xmath57 such that @xmath58 .
for @xmath59 the first deviation from the regular sequence takes place at some @xmath60 .
it is easy to see that by increasing @xmath61 to @xmath62 one can not get a regular sequence and the first deviation will occur at some @xmath63 .
note that the case @xmath64 can be similarly treated .
turning now to the most interesting domain of @xmath65 , we consider again the case @xmath44 and the regular sequence of zero interlacing , @xmath66 .
as indicated above , its fulfillment ensures lack of root of the wronskian : @xmath67 . by noting that any @xmath41th zero of a bessel function is a strictly growing function of the order it is sufficient to prove that @xmath68 , holds for @xmath69 and @xmath70 .
the only unknown inequality here is that of @xmath71 . to prove its validity we use the known intermediate relation @xmath72
therefore , proving @xmath73 will suffice . consider the known relation @xmath74 which means that @xmath75 and @xmath76 have the same sign at @xmath77 .
now because of the interlacing property @xmath78 and the limit @xmath79 @xmath80 , this implies that the @xmath41th zero of @xmath81 precedes the @xmath45th zero of @xmath82 , i.e. @xmath83 which had to be proven .
note that the case @xmath64 can be similarly treated . in case of @xmath84 ,
the cox - thompson inverse scattering scheme yields a potential of the class @xmath20 iff the condition @xmath85 holds .
in the course of the proof we obtained the following result of its own right : denoting the @xmath41th root of the bessel functions @xmath86 , @xmath87 , @xmath88 , respectively , by @xmath89 , @xmath90 , @xmath91 then the following inequality is valid for @xmath92 : @xmath93 .
this proposition adds two new inequality sequences to the known ones ( see e.g. ref .
@xcite ) : @xmath94 and @xmath95
one can construct a potential that possesses one specified phase shift @xmath96 ( @xmath84 ) by using the inversion scheme of cox and thompson @xcite : @xmath97 @xmath98 @xmath99 relation ( [ tan ] ) gives @xmath100 , @xmath101 . for @xmath102 $ ] the corollary results in the choice of @xmath103 .
therefore , for any @xmath96 , there is only one , easily identifiable non - singular potential and an infinite number of singular potentials that the cox - thompson method can produce .
for an example let us choose @xmath104 and @xmath105 . in this case equation ( [ tan ] ) yields
@xmath106 , @xmath101 .
@xmath107 ( @xmath103 ) is not permitted by the assumption @xmath11 , however in order to get a non - singular potential one may replace this @xmath107 by @xmath108 with @xmath109 . using equation ( [ keq ] )
one gets at @xmath104 and @xmath110 @xmath111 since @xmath112 and @xmath113 are continuous in @xmath62 and @xmath114 , @xmath115 holds .
thus @xmath116 for @xmath30 .
this is the physical solution .
( see fig . 1 . ) with @xmath117 , and @xmath118 . ]
[ fig1 ] now let @xmath104 and @xmath118 . by the corollary we can not get an integrable potential in this case because @xmath119 ( see fig .
1 ) . in ref .
@xcite it has been shown explicitly that equation ( [ gl ] ) is not uniquely solvable at some @xmath120 for this case . however , while ref .
@xcite suggests that this fact makes the cox - thompson scheme useless , in this paper we have shown that in order to get an integrable potential , the choice @xmath118 is not permitted because the set @xmath12 is not empty . on the other hand , equation ( [ tan ] ) and the corollary
provide a one - to - one correspondence between the phase shift and the @xmath62 parameter of the cox - thompson method at the one - term level .
this correspondence has the property that the potential constructed from @xmath62 belongs to @xmath20 and possesses the specified phase shift . in fig .
2 we add examples of the construction of an unique potential in @xmath121 in the case of non - trivial phase shifts .
subfigure fig .
2a shows the potentials for zero angular momentum with @xmath122 corresponding to @xmath123 which is permitted by the corollary ( nonsingular case , full line ) and @xmath124 which violates the corollary ( one - singularity case , dotted line ) .
subfigure fig .
2b shows the potentials for the @xmath22-wave phase shift @xmath125 corresponding to @xmath126 [ permitted by the corollary ( nonsingular case , full line ) ] and @xmath127 [ violating the corollary ( two - singularity case , dotted line ) ] .
we note that the second singularity of the dotted curve in fig .
2b lies out of the region shown and , as expected form the _ proof _ , the singular potentials in fig . 2a and 2b have one and two locally non - integrable region(s ) corresponding to @xmath128 and @xmath129 , respectively .
+ [ fig2 ] applications ( see refs .
@xcite ) of the cox - thompson scheme for @xmath130 suggest the existence of a connection similar to the corollary of section 2 that specifies one nonsingular potential out of the possible infinite singular solutions .
however such a theorem has , as yet , not been proven .
the authors thank professors mikls horvth and dniel schumayer for reading the manuscript . | we establish a condition for obtaining nonsingular potentials using the cox thompson inverse scattering method with one phase shift .
the anomalous singularities of the potentials are avoided by maintaining unique solutions of the underlying regge
newton integral equation for the transformation kernel . as a by - product , new inequality sequences of zeros of bessel functions are discovered . + pacs : 03.65.nk , 03.65.ge , 02.30.gp , 02.30.rz keywords : bessel functions , integral equations , potential scattering , schrodinger equation |
ongoing doppler planet searches have discovered more than 200 extrasolar planets with various characteristics .
most of them are quite different from those in our solar - system : the planets have minimum masses of 5 @xmath615 @xmath3 and are distributed in the range of orbital radii from 0.02 to 6 au with orbital eccentricities of 00.9 ( e.g. butler et al .
2006 ) . distribution and correlation between these parameters
are now used to calibrate planet formation theories by comparing with predictions from numerical simulations taking account of key processes of planet formation such as migration , disk lifetime , and variation of disk mass ( ida & lin 2004 ; alibert et al . 2005 ) .
properties of host stars also play an important role to constrain planet formation mechanisms .
for example , frequency of planets is well correlated with metallicity of the host stars : metal - rich stars tend to harbor more planets than metal - poor ones do ( fischer & valenti 2005 ; santos et al . 2005 ) , which supports the core - accretion scenario for the mechanism of giant planet formation . the host stars mass can be another essential parameter .
recently , johnson et al . ( 2007a ) showed that frequency of planets around stars with @xmath7 is as high as about 9% , compared to about 4% for solar - mass ( @xmath8 ) stars and about 2% for low - mass ( @xmath9 ) k m stars .
it suggests that giant planets are more abundant in more massive stars probably because of their larger surface density of dust in proto - planetary disks ( ida & lin 2005 ; laughlin et al .
however , in higher - mass stars such as b a dwarfs ( @xmath10 ) , the shorter lifetime of the disk and the paucity of solid materials close to the central star could reduce the abundance of giant planets as a whole .
the frequency of planets around such massive stars has not been established yet .
doppler planet searches around intermediate - mass ( @xmath11 ) stars have gradually expanded during these five years . since massive stars on the main - sequence ( early - type stars )
are unsuitable for precise radial velocity measurements due to few absorption lines in their spectra , which are often rotationally broadened , major teams have targeted cool and slowly - rotating g and k giants and subgiants , that is , massive stars in evolved stages ( setiawan et al .
2005 ; sato et al .
2003 , 2007 ; hatzes et al . 2005 ; johnson et al . 2007b ; lovis & mayor 2007 ;
niedzielski et al . 2007 ; liu et al .
they have succeeded in discovering 12 substellar companions so far around stars with masses of 1.63.9@xmath5 including clump giants and subgiants , and about a half of the host stars have masses greater than 2@xmath5 .
although the number of planets is still small , the planets begin to show different properties from those around low - mass stars : high frequency of massive planets ( lovis & mayor 2007 ) , lack of inner planets ( johnson et al . 2007b ) , and low metallicity of host stars ( pasquini et al . 2007 ) .
these properties must reflect the history of formation and evolution of planetary systems , which are not necessarily the same as those for solar - type stars .
planets around low - mass ( @xmath12 ) giants have been also discovered from precise radial velocity surveys ( frink et al . 2002 ; setiawan et al . 2003 ; setiawan 2003 ; hatzes et al . 2003 ; d@xmath13llinger et al .
comparing orbital distribution of the planets with those around low - mass dwarfs can give insight into understanding of evolution of planetary systems .
in this paper , we report on the detection of 3 new extrasolar planets around intermediate - mass g and k giants ( 18 del , @xmath0 aql , hd 81688 ) from the okayama planet search program ( sato et al . 2005 ) .
we also update orbital parameters of hd 104985 b , the first planet discovered around g giants from our survey ( sato et al .
2003 ) , by using the data collected during the past six years .
based on the extended sample , we discuss orbital properties of planets around evolved stars taking account of evolution of central stars .
since 2001 , we have been conducting a precise doppler survey of about 300 g and k giants ( sato et al . 2005 ) using a 1.88 m telescope , the high dispersion echelle spectrograph ( hides ; izumiura 1999 ) , and an iodine absorption cell ( i@xmath14 cell ; kambe et al .
2002 ) at okayama astrophysical observatory ( oao ) . for precise radial velocity measurements
, we set a wavelength range to 50006100@xmath15 , in which many deep and sharp i@xmath14 lines exist , and a slit width to 200 @xmath16 m ( @xmath17 ) giving a spectral resolution ( @xmath18 ) of 67000 by about 3.3 pixels sampling .
we can typically obtain a signal - to - noise ratio s / n@xmath19200 pix@xmath20 for a @xmath21 star with an exposure time shorter than 30 min .
we have achieved a doppler precision of about 6 m s@xmath20 over a time span of 6 years using our own analysis software for modeling an i@xmath14-superposed stellar spectrum ( sato et al . 2002 , 2005 ) . for abundance analysis
, we take a pure ( i@xmath14-free ) stellar spectrum with the same wavelength range and spectral resolution as those for radial velocity measurements .
we also take a spectrum covering ca ii h k lines in order to check the chromospheric activity ( not simultaneously obtained with the radial velocity data ) for stars showing large radial velocity variations . in this case
, we set the wavelength range to 38004500 @xmath15 and the slit width to 250 @xmath16 m giving a wavelength resolution of 50000 .
we can typically obtain s / n@xmath2220 pix@xmath20 at the ca ii h k line cores for a @xmath23 star with a 30 min exposure .
the reduction of echelle data ( i.e. bias subtraction , flat - fielding , scattered - light subtraction , and spectrum extraction ) is performed using the iraf software package in the standard way .
18 del ( hr 8030 , hd 199665 , hip 103527 ) is listed in the hipparcos catalog ( esa 1997 ) as a g6 iii : giant star with a @xmath24 magnitude @xmath25 , a color index @xmath26 , and the hipparcos parallax @xmath27 mas , corresponding to a distance of [email protected] pc and an absolute magnitude @xmath29 taking account of correction of interstellar extinction @xmath30 based on the arenou et al s ( 1992 ) table .
_ hipparcos _ made a total of 175 observations of the star , revealing a photometric stability down to @xmath31 mag .
figure [ fig - cah ] shows a ca ii h line for the star obtained with hides revealing slight core reversal in the line .
x - ray luminosity for the star was derived to @xmath32 erg s@xmath20 from the @xmath33 measurements ( h@xmath34nsch et al .
1998 ) , suggesting that the star is slightly chromospherically active . however , the reversal is not significant compared to those in other chromospherically active stars in our sample such as hd 120048 ( figure [ fig - cah ] ) , which shows velocity scatter of about 30 m s@xmath20 .
thus , although the correlation between chromospheric activity and intrinsic radial velocity `` jitter '' for giants have not been well established yet , the jitter of 18 del is probably expected to be no larger than that of hd 120048 .
further discussions are presented in section 4 .
the atmospheric parameters and the fe abundance of the star were determined based on the spectroscopic approach using the equivalent widths of well - behaved fe i and fe ii lines ( cf .
takeda et al .
2002 for a detailed description of this method ) .
we obtained @xmath35 = 4979 k , @xmath36 = 2.82 cm s@xmath37 , @xmath38 = 1.22 km s@xmath20 , and [ fe / h ] = @xmath390.05 for the star .
the bolometric correction was estimated to @xmath40 based on the kurucz ( 1993 ) s theoretical calculation .
with use of these parameters and theoretical evolutionary tracks of lejeune & schaerer ( 2001 ) , we derived the fundamental stellar parameters , @xmath41 , @xmath42 , and @xmath43 , as summarized in table [ tbl - stars ] .
the procedure described here is the same as that adopted in takeda et al .
( 2005 ) ( see subsection 3.2 of takeda et al .
( 2005 ) and note of table [ tbl - stars ] for uncertainties involved in the stellar parameters ) .
the position of the star on the hr diagram is shown in figure [ fig - hrd ] together with other planet - harboring stars in this paper . da silva et al .
( 2006 ) obtained 2.13 @xmath28 0.13 @xmath5 for the mass of the star based on @xmath35 = 5089 @xmath28 70 k , @xmath36 = 2.93 @xmath28 0.08 cm s@xmath37 , and [ fe / h ] = [email protected] .
although the @xmath35 is about 100 k higher than that we obtained , the mass for the star reasonably agrees with our estimate .
we collected a total of 51 radial velocity data of 18 del between 2002 august and 2007 june , with a typical s / n of 200 pix@xmath20 for an exposure time of about 900 s. the observed radial velocities are shown in figure [ fig - hd199665 ] and are listed in table [ tbl - hd199665 ] together with their estimated uncertainties , which were derived from an ensemble of velocities from each of @xmath44200 spectral regions ( each 45@xmath15 long ) in every exposure .
the sinusoidal variability in the radial velocities is visible to the eye and it can be well fitted by a keplerian orbit with a period @xmath45 days , a velocity semiamplitude @xmath46 m s@xmath20 , and an eccentricity @xmath47 .
the resulting model is shown in figure [ fig - hd199665 ] overplotted on the velocities , and its parameters are listed in table [ tbl - planets ] .
the uncertainty of each parameter was estimated using a monte carlo approach .
we generated 100 fake datasets by adding random gaussian noise corresponding to velocity measurement errors to the observed radial velocities in each set , then found the best - fit keplerian parameters for each , and examined the distribution of each of the parameters .
the rms scatter of the residuals to the keplerian fit is 15.4 m s@xmath20 , which is comparable to the scatters of giants with the same @xmath48 as 18 del in our sample ( sato et al . 2005 ) . adopting a stellar mass of 2.3 @xmath5
, we obtain a minimum mass for the companion @xmath49 @xmath3 and a semimajor axis @xmath50 au .
the companion has the longest orbital period ever discovered around evolved stars .
we found that the residuals showed a decreasing trend with a slope of about @xmath394 m s@xmath20 yr@xmath20 , suggesting the existence of an outer companion ( figure [ fig - hd199665-res ] ) .
keplerian orbital fit including a linear trend slightly improves the quality of fit , decreasing the rms scatter from 15.4 m s@xmath20 to 13.6 m s@xmath20 and the reduced @xmath51 from @xmath52 to @xmath53 . to assess the significance of this trend , we evaluated a false - alarm probability ( @xmath54 ) , the probability that noise mimics the observed trend , by using a bootstrap analysis which is the same as adopted in wright et al .
we scrambled the residuals in a random manner while keeping fixed the observation time , and created a mock set of radial velocities by adding the residuals back to the best - fit keplerian radial velocity curve .
we created 100 such mock data sets , and obtained @xmath55 for each set .
eight of the 100 data sets showed @xmath56 less than that for the original data set , which means the @xmath54 is 8% .
considering the @xmath54 , we can not say that the observed trend is significant at this stage .
@xmath0 aql ( hr 7595 , hd 188310 , hip 97938 ) is a k0 iii giant star with a @xmath24 magnitude @xmath57 , a color index @xmath58 , and a trigonometric parallax @xmath59 mas ( esa 1997 ) , placing the star at a distance of [email protected] pc . the distance and an estimated interstellar extinction @xmath60 ( arenou et al .
1992 ) yield an absolute magnitude for the star @xmath61 .
_ hipparcos _ made a total of 98 observations of the star , revealing a photometric stability down to @xmath62 mag .
ca ii h k lines of the star show no significant emission in the line cores as shown in the figure [ fig - cah ] , suggesting that the star is chromospherically inactive .
the atmospheric parameters , fe abundance , and other fundamental parameters of the star are listed in table [ tbl - stars ] .
the [ fe / h ] of @xmath63 is consistent with the value of @xmath64 from taylor ( 1999 ) within the error .
gray and brown ( 2001 ) obtained @xmath65 k for the star based on line - depth - ratio analysis , which is @xmath66 k lower than our estimate ( 4780 k ) .
uncertainties in the mass and the radius are considered to be similar to those for 18 del ( see note of table [ tbl - stars ] ) .
we collected a total of 26 radial velocity data of @xmath0 aql between 2004 april and 2007 june , with a typical s / n of 200 pix@xmath20 for an exposure time of about 300 s. the observed radial velocities are shown in figure [ fig - hd188310 ] and are listed in table [ tbl - hd188310 ] together with their estimated uncertainties .
lomb - scargle periodogram ( scargle 1982 ) of the data exhibits a dominant peak at a period of 137 days . to assess the significance of this periodicity , we estimated @xmath54 , using a bootstrap randomization method in which the observed radial velocities were randomly redistributed , keeping fixed the observation time .
we generated 10@xmath67 fake datasets in this way , and applied the same periodogram analysis to them .
only 2 fake datasets exhibited a periodogram power higher than the observed dataset .
therefore , the @xmath54 is @xmath68 .
the observed radial velocities can be well fitted by a circular orbit with a period @xmath69 days and a velocity semiamplitude @xmath70 m s@xmath20 .
the resulting model is shown in figure [ fig - hd188310 ] , and its parameters are listed in table [ tbl - planets ] .
the uncertainty of each parameter was estimated using a monte carlo approach as described in section 3.1 .
the rms scatter of the residuals to the keplerian fit was 22.3 m s@xmath20 , which is slightly larger than the typical scatter of giants with @xmath71 in our sample ( sato et al .
2005 ) . adopting a stellar mass of 2.2 @xmath5
, we obtain a minimum mass for the companion of @xmath72 and a semimajor axis of @xmath73 au .
the companion has the shortest orbital period ever discovered among evolved stars .
hd 81688 ( hr 3743 , hip 46471 ) is classified in the hipparcos catalog ( esa 1997 ) as a k0 iii iv star with a @xmath24 magnitude @xmath74 , a color index @xmath75 .
the hipparcos parallax @xmath76 mas corresponds to a distance of [email protected] pc and yields an absolute magnitude @xmath77 corrected by interstellar extinction @xmath60 ( arenou et al .
_ hipparcos _ made a total of 107 observations of the star , revealing a photometric stability down to @xmath78 mag .
ca ii h k lines of the star show no significant emission in the line cores , suggesting that the star is chromospherically inactive ( figure [ fig - cah ] ) .
the atmospheric parameters , fe abundance , and other fundamental parameters of the star are listed in table [ tbl - stars ] .
mishenina et al .
( 2006 ) obtained @xmath79 k ( from line - depth - ratio analysis ) , @xmath36 = 2.3 cm s@xmath37 , @xmath38 = 1.3 km s@xmath20 , and [ fe / h ] = @xmath390.23 for the star .
while the [ fe / h ] is @xmath80 dex higher than our estimate , other parameters reasonably agree with those we obtained .
although uncertainties in the mass and the radius are considered to be similar to those for 18 del , systematic error up to @xmath81 could exist depending on adopted stellar evolutionary models ( see note of table [ tbl - stars ] ) .
we collected a total of 81 radial velocity data of hd 81688 between 2003 march and 2007 april , with a typical s / n of 200 pix@xmath20 for an exposure time of about 900 s. the observed radial velocities are shown in figure [ fig - hd81688 ] and are listed in table [ tbl - hd81688 ] together with their estimated uncertainties .
lomb - scargle periodogram ( scargle 1982 ) of the data exhibits a dominant peak at a period of 182 days with a @xmath82 , which was estimated by the same method as described in section 3.1 .
the observed radial velocities can be well fitted by a circular orbit with a period @xmath83 days and a velocity semiamplitude @xmath84 m s@xmath20 .
the resulting model is shown in figure [ fig - hd81688 ] , and its parameters are listed in table [ tbl - planets ] .
the uncertainty of each parameter was estimated using a monte carlo approach as described in section 3.1 . adopting a stellar mass of 2.1 @xmath5
, we obtain a minimum mass for the companion of @xmath85 @xmath3 and a semimajor axis of @xmath86 au .
the residuals to the keplerian fit exhibit non - random variations in some periods of time , which may be due to stellar activity or additional companions .
we performed periodogram analysis ( scargle 1982 ) to the residuals and found peaks at periods around 13 , 25 , and 55 days . however , the @xmath54 s for the peaks are larger than 0.5 , which is not considered to be significant at this stage ( some of them my be affected by aliasing )
. more dense sampling of data will help discriminate if the periods are real or not .
hd 104985 ( hr 4609 , hip 58952 ) is the first planet - harboring star discovered from our survey ( sato et al .
it is classified in the hipparcos catalog ( esa 1997 ) as a g9 iii giant star with a @xmath24 magnitude @xmath87 , a color index @xmath88 , and a parallax @xmath89 mas , corresponding to a distance of [email protected] pc and an absolute magnitude @xmath90 .
the atmospheric parameters of the star were updated by takeda et al .
( 2005 ) from those listed in the discovery paper by sato et al .
( 2003 ) to : @xmath35 = 4877 k , @xmath36 = 2.85 ( cm s@xmath37 ) , @xmath38 = 1.31 ( km s@xmath20 ) , and [ fe / h ] = @xmath64 . based on these parameters and the bolometric correction of @xmath91 ( kurucz 1993 ) , takeda et al .
( 2005 ) obtained the fundamental stellar parameters of @xmath92 , @xmath93 , and @xmath43 .
based on the metallicity [ fe / h]@xmath94 and evolutionary tracks from girardi et al .
the tracks tend to give @xmath95 lower mass compared to those from lejeune and schaerer ( 2001 ) for @xmath96 giants with @xmath97 ( [ fe / h]@xmath98 ) . ] after the discovery of the planet around hd 104985 in 2003 , we have continued observations of the star and collected a total of 52 data points between 2001 march and 2007 april .
the observed radial velocities are shown in figure [ fig - hd104985 ] and are listed in table [ tbl - hd104985 ] together with their estimated uncertainties .
based on the extended data set , we updated the orbital parameters of the planet : @xmath99 days , @xmath100 m s@xmath20 , and @xmath101 .
the resulting keplerian model is shown in figure [ fig - hd81688 ] , and the parameters are listed in table [ tbl - planets ] . adopting a stellar mass of @xmath102 , we obtain a minimum mass for the companion of @xmath103 and a semimajor axis of @xmath104 au .
we can not find any additional periodic signals or long - term trend in their radial velocities for now .
to investigate other causes producing apparent radial velocity variations such as pulsation and rotational modulation rather than orbital motion , spectral line shape analysis was performed with use of high resolution stellar templates followed by the technique of sato et al .
( 2007 ) . in our technique
, we extract a high resolution iodine - free stellar template from several stellar spectra contaminated by iodine lines ( sato et al .
basic procedure of the technique is as follows ; first , we model observed star+i@xmath14 spectrum in a standard manner but using the initial guess of the intrinsic stellar template spectrum .
next we take the difference between the observed star+i@xmath14 spectrum and the modeled one . since the difference is mainly considered to be due to an imperfection of the initial guess of the stellar template spectrum , we revise the initial guess taking account of the difference and model the observed star+i@xmath14 spectrum using the revised guess of the template .
we repeat this process until we obtain sufficient agreement between observed and modeled spectrum .
we take average of thus obtained stellar templates from several observed star+i@xmath14 spectra to increase s / n ratio of the template .
details of this technique are described in sato et al .
( 2002 ) . for spectral line shape analysis
, we extracted two stellar templates from several star+i@xmath14 spectra at the peak and valley phases of observed radial velocities for each star .
then , cross correlation profiles of the two templates were calculated for 3050 spectral segments ( 45@xmath15 width each ) in which severely blended lines or broad lines were not included .
three bisector quantities were calculated for the cross correlation profile of each segment : the velocity span ( bvs ) , which is the velocity difference between two flux levels of the bisector ; the velocity curvature ( bvc ) , which is the difference of the velocity span of the upper half and lower half of the bisector ; and the velocity displacement ( bvd ) , which is the average of the bisector at three different flux levels .
we used flux levels of 25% , 50% , and 75% of the cross correlation profile to calculate the above quantities .
resulting bisector quantities for 18 del , @xmath0 aql , hd 81688 , and hd 104985 are listed in table [ tbl - bisector ] . as expected from the planetary hypothesis , both of the bvs and the bvc are identical to zero , which means that the cross correlation profiles are symmetric , and the average bvd is consistent with the velocity difference between the two templates at the peak and valley phases of observed radial velocities ( @xmath105 ) .
the bvs for hd 104985 is slightly large compared to those for other stars , suggesting the higher intrinsic variability for hd 104985
. it may be consistent with the large rms scatters of the residuals to the keplerian fit ( @xmath106 m s@xmath20 ) for the star .
however , the bvs value is only one thirtieth of the bvd and thus it is unlikely that the observed radial velocity variations are produced by the intrinsic activity such as pulsation or rotational modulation .
based on these results , we conclude that the radial velocity variability observed in these 4 stars are best explained by orbital motion .
we discovered a total of 6 substellar companions around g and k giants so far from our okayama planet search program ( sato et al .
2003 , 2007 ; liu et al . 2007 ; this work ) .
the host stars are located at the clump region on the hr diagram and their masses are estimated to 2.12.7 @xmath5 .
when we include 6 more planets discovered around possible clump giants by other teams ( table [ tbl - pllist ] ) , the mass of the host stars ranges from 1.7 to 3.9 @xmath5 .
these discoveries definitely indicate that planets can form around intermediate - mass stars , such as b a dwarfs , as well as around low - mass ones .
the extended sample enables us to clarify the properties of the planets around intermediate - mass clump giants ; the planets have minimum masses of 2.319.8@xmath3 , semimajor axes of 0.682.6 au , and eccentricities of 00.4 . in figure
[ fig - avsm ] , we plotted the minimum mass against the semimajor axis . since intrinsic variability in radial velocity of clump giants is typically 1020 m s@xmath20 ( sato et al .
2005 ) , it is normally difficult to detect planets with @xmath107 at @xmath221 au around a 2@xmath5 star , which can produce radial velocity semiamplitude of 40 m s@xmath20 at most .
such lower - mass planets can be detected around stars with small intrinsic variability ( @xmath10810 m s@xmath20 ) like subgiants ( johnson et al .
while the largest semimajor axis of 2.6 au is limited by the time baseline of the current surveys , the lack of short - period planets with @xmath1090.7 au appears to be real at least for relatively massive planets because stellar radius of clump giants is typically 1020@xmath110 , which correspond to 0.050.1 au , and thus we should be able to find planets with [email protected] au if they exist . in figure [ fig - avse ] , we plotted the eccentricity against the semimajor axis together with lines expressing different periastron distances ( @xmath112 ) . from the view point of orbital evolution of planets ,
periastron distance is more essential rather than semimajor axis because tidal interaction between a planet and a central star strongly depends on distance between them .
as shown in the figure , all the companions around clump giants have @xmath113 au , while those around intermediate - mass subgiants ( 1.61.9@xmath5 ) and low - mass k giants ( @xmath12 ) have @xmath114 au and @xmath115 au , respectively .
since a lot of planets with @xmath1160.3 au have been found around solar - type dwarfs ( open circles ) , the lack of inner planets around low - mass k giants can be due to engulfment by the central stars .
we here examined whether the lack of short - period planets around clump giants could be reproduced by evolutionary effect of the central stars based on available stellar evolutionary models .
if the clump giants are post - rgb ( core helium burning ) stars , short - period planets around them might have been engulfed by the central stars at the tip of rgb due to tidal torque from the expanding stellar surface .
we numerically traced tidal evolution of a planetary orbit ( @xmath117 ) based on equations from zahn ( 1989 ) and stellar evolutionary tracks from lejeune & schaerer ( 2001 ) ( ls01 ) and claret ( 2004,2006,2007 ) .
we assumed a circular orbit but the result can be applied to the case of an eccentric orbit by replacing the semimajor axis with periastron distance .
we also took account of orbital evolution due to mass loss as @xmath118 , where @xmath119 is a mass of the central star .
mass loss of the central star makes planets move outward because of their weakened gravitational pull on the planets ( e.g. , sackmann et al . 1993 ; duncan & lissauer 1998 ) .
thus , net change of orbital radius of a planet is expressed as @xmath120 .
we finally found out , however , that orbital change due to mass loss is negligible in rgb phase for planets around 23 @xmath5 stars because the mass loss of those stars in rgb phase is negligible based on the adopted evolutionary tracks .
it excludes the scenario that inner planets were pushed out to 0.7 au resulting in the lack of planets within the radius .
the mass loss may be important in the case of lower - mass stars ( silvotti et al . 2007 ) . from the orbital calculations
, we found out that jupiter - mass planets within about 34 @xmath121 ( radius of a central star ) can be engulfed by the central stars due to the tidal torque during rgb phase .
our results predict that , around 23 @xmath5 stars , only planets within 0.2 au are engulfed by the central stars at the bottom of rgb ( @xmath122 ) , but those within about 0.5 au can be done at the tip of rgb ( @xmath123 ) . the critical orbital radius at the tip of rgb for 2@xmath5 can be larger than 1 au in the cases of @xmath124 for ls01 and @xmath125 and 0.04 for claret s tracks .
when we assume that most of the clump giants are post - rgb stars , it might be natural that we could not find short - period planets around them even if they had originally existed .
typically small orbital eccentricities ( @xmath126 ) of the planets discovered around clump giants compared to those around dwarfs may favor this scenario , in which the planetary orbits could be tidally circularized during rgb phase
. there can be another possibility , however , that short - period planets are primordially rare around intermediate - mass stars .
lack of short - period planets around less evolved subgiants with 1.61.9@xmath5 ( figure [ fig - avse ] ; johnson et al .
2007b ) , which are considered to be first ascent on rgb , may favor this scenario .
such dependence of orbital distribution of planets on host star s mass is predicted by burkert & ida ( 2007 ) .
they pointed out that in observed data for f
k dwarfs , there may be a paucity of planets in @xmath1270.10.6 au around @xmath128 stars ( f dwarfs ) while the semimajor axis distribution is more uniform around g and k dwarfs .
they showed that the gap could be produced by shorter viscous diffusion timescale of disks possibly due to smaller disk size for more massive stars , which can limit the efficiency of type ii migration of giant planets and keep planets residing close to initial formation locations beyond snow lines at several au .
the lack of inner planets around subgiants with 1.61.9@xmath5 might be consistent with this prediction .
it is not clear whether this is also the case for more massive stars with @xmath129 . around such stars , due to distant snow line ( @xmath13010 au ) and averagely large disk mass ,
the main region for giant planet formation could be closer to the central stars compared to around low - mass stars .
on the other hand , kennedy and kenyon ( 2007 ) recently showed that the snow line distance changes weakly with stellar mass if they take account of disk and pre - main - sequence evolution .
more detailed theoretical modeling is required for planet formation around stars with @xmath129 . in order to further investigate formation and evolution of planetary systems around massive stars via evolved giants , it is important to derive their accurate mass and evolutionary status .
it is normally difficult , however , because stars with different mass and evolutionary status can occupy similar position in the giant branch on the hr diagram . to overcome this difficulty
, asteroseismology will be a powerful tool , which can probe stellar interior by using tiny stellar oscillations .
such oscillations were actually detected in some g and k giants ( e.g. , frandsen et al .
2002 ; hatzes & zechmeister 2007 ; ando et al .
2007 ) . applying this technique to the planet - harboring evolved stars
is highly encouraged . + this research is based on data collected at okayama astrophysical observatory ( oao ) , which is operated by national astronomical observatory of japan ( naoj ) .
we are grateful to all the staff members of oao for their support during the observations .
m.i . thanks dr .
a. claret for his helpful comments on his evolutionary tracks .
data analysis was in part carried out on `` sb '' computer system operated by the astronomical data analysis center ( adac ) and subaru telescope of naoj .
we thank the national institute of information and communications technology for their support on high - speed network connection for data transfer and analysis .
is supported by grant - in - aid for young scientists ( b ) no.17740106 , and h.i .
, h.a . , and m.y
are supported by grant - in - aid for scientific research ( c ) no.13640247 , ( b ) no.17340056 , ( b ) no.18340055 , respectively , from the japan society for the promotion of science ( jsps ) .
e.t , d.m , and y.i . are supported by `` the 21st century coe program : the origin and evolution of planetary systems '' in ministry of education , culture , sports , science and technology ( mext ) .
e.k . and s.i .
are partially supported by mext , japan , the grant - in - aid for scientific research on priority areas , `` development of extra - solar planetary science . ''
this research has made use of the simbad database , operated at cds , strasbourg , france .
alibert , y. , mordasini , c. , benz , w. , winisdoerffer , c. 2005 , , 434 , 343 ando , h. , tan , k. , kambe , e. , sato , b. , & zhao , g. 2008 , pasj , in press arenou , f. , grenon , m. , & gomez , a. 1992 , a&a , 258 , 104 burkert , a. & ida , s. 2007 , apj , 660 , 845 butler , r.p .
et al . 2006 , apj , 646 , 505 claret , a. 2007 , a&a , 467 , 1389 claret , a. 2006 , a&a , 453 , 769 claret , a. 2004 , a&a , 424 , 919 da silva , l. et al .
2006 , a&a , 458 , 609 de medeiros , j. r. , & mayor , m. 1999 , , 139 , 433 d@xmath13llinger , m.p . ,
hatzes , a.p . , pasquini , l. , guenther , e.w . ,
hartmann , m. , girardi , l. , & esposito , m. 2007 , a&a , 472 , 649 duncan , m.j .
& lissauer , j.j .
1998 , icarus , 134 , 303 esa .
1997 , the _ hipparcos _ and tycho catalogues ( esa sp-1200 ; noordwijk : esa ) fischer , d.a .
& valenti , j. 2005 , , 622 , 1102 frandsen , s. et al .
2002 , a&a , 394 , 5 frink , s. , mitchell , d.s . , quirrenbach , a. , fischer , d. , marcy , g.w . , & butler , r.p .
2002 , apj , 576 , 478 girardi , l. , bressan , a. , bertelli , g. , & chiosi , c. 2000 , a&as , 141 , 371 gray , d.f . &
brown , k. 2001 , pasp , 113 , 723 hatzes , a.p .
2003 , apj , 599 , 1383 hatzes , a.p . ,
guenther , e.w . ,
endl , m. , cochran , w.d . ,
d@xmath13llinger , m.p . , & bedalov , a. 2005 , , 437 , 743 hatzes , a.p . &
zechmeister , m. 2007 , apj , 670 , l37 h@xmath34nsch , m. , schmitt , j. h. m. m. , & voges , w. 1998 , , 127 , 251 ida , s. & lin , d.n.c .
2004 , , 604 , 388 ida , s. & lin , d.n.c .
2005 , , 626 , 1045 izumiura , h. 1999 , in proc .
4th east asian meeting on astronomy , ed .
chen ( kunming : yunnan observatory ) , 77 johnson , j.a . , butler , r.p . , marcy , g.w . , fischer , d.a .
, vogt , s.s . ,
wright , j.t .
, & peek , k.m.g .
2007a , apj , 670 , 833 johnson , j.a .
et al . 2007b , , 665 ,
785 kambe , e. et al .
2002 , pasj , 54 , 865 kennedy , g.m . and
kenyon , s.j .
2007 , apj in press kurucz , r. l. 1993 , kurucz cd - rom , no . 13
( harvard - smithsonian center for astrophysics ) laughlin , g. , bodenheimer , p. & adams , f.c .
2004 , apj , 612 , 73 lejeune , t. & schaerer , d. 2001 , a&a , 366 , 538 liu , y.j .
2007 , apj , 672 , 553 lovis , c. & mayor , m. 2007 , a&a , 472 , 657 mishenina , t.v .
2006 , a&a , 456 , 1109 niedzielski , a. et al .
2007 , apj , 669 , 1354 pasquini , l. et al .
2007 , a&a , 473 , 979 sackmann , i.j . ,
boothroyd , a.i . ,
& kraemer , k.e .
1993 , , 418 , 457 santos , n.c . ,
israelian , g. , mayor , m. , bento , j.p . , almeida , p.c . ,
sousa , s.g .
& ecuvillon , a. 2005 , a&a , 437 , 1127 sato , b. et al .
2003 , apj , 597 , l157 sato , b. , kambe , e. , takeda , y. , izumiura , h. , & ando , h. 2002 , pasj , 54 , 873 sato , b. , kambe , e. , takeda , y. , izumiura , h. , masuda , s. & ando , h. 2005 , , 57,97 sato , b. et al .
2007 , apj , 661 , 527 scargle , j. d. 1982 , , 263 , 835 setiawan , j. et al .
2003 , a&a , 398 , l19 setiawan , j. darwin / tpf and the search for extrasolar terrestrial planets , 22 - 25 april 2003 , heidelberg , germany . edited by m. fridlund , t. henning , compiled by h. lacoste .
esa sp-539 , noordwijk , netherlands : esa publications division , 2003 , p. 595
- 598 setiawan , j. et al .
2005 , a&a , 437 , 31 silvotti , r. et al .
2007 , nature , 449 , 189 takeda , y. , ohkubo , m. , & sadakane , k. 2002 , , 54 , 451 takeda , y , sato , b. , kambe , e. , izumiura , h. , masuda , s. , & ando , h. 2005 , pasj , 57 , 109 taylor , b.j . 1999 ,
a&as , 134 , 523 wright , j.t .
2007 , apj , 657 , 533 zahn , j .-
1989 , , 220 , 112 references. ( 1 ) lovis & mayor ( 2007 ) ; ( 2 ) hatzes et al .
( 2005 ) ; ( 3 ) sato et al .
( 2007 ) ; ( 4 ) liu et al .
( 2007 ) ; ( 5 ) this work ; ( 6 ) sato et al .
( 2003 ) ; ( 7 ) niedzielski et al .
( 2007 ) ; ( 8) setiawan et al .
( 2005 ) ; ( 9 ) hatzes et al .
( 2006 ) ; ( 10 ) johnson et al .
( 2007 ) ; ( 11 ) hatzes et al .
( 2003 ) ; ( 12 ) setiawan ( 2003 ) ; ( 13 ) d@xmath13llinger et al .
( 2007 ) ; ( 14 ) setiawan et al .
( 2003 ) ; ( 15 ) frink et al . | we report the detection of 3 new extrasolar planets from the precise doppler survey of g and k giants at okayama astrophysical observatory . the host stars , namely , 18 del ( g6 iii ) , @xmath0 aql ( k0 iii ) and hd 81688 ( k0 iii
iv ) , are located at the clump region on the hr diagram with estimated masses of @xmath1 .
18 del b has a minimum mass of @xmath2 and resides in a nearly circular orbit with period of 993 days , which is the longest one ever discovered around evolved stars .
@xmath0 aql b and hd 81688 b have minimum masses of 2.8 and 2.7 @xmath3 , and reside in nearly circular orbits with periods of 137 and 184 days , respectively , which are the shortest ones among planets around evolved stars .
all of the substellar companions ever discovered around possible intermediate - mass ( @xmath4 ) clump giants have semimajor axes larger than 0.68 au , suggesting the lack of short - period planets .
our numerical calculations suggest that jupiter - mass planets within about 0.5 au ( even up to 1 au depending on the metallicity and adopted models ) around 23@xmath5 stars could be engulfed by the central stars at the tip of rgb due to tidal torque from the central stars . assuming that most of the clump giants are post - rgb stars
, we can not distinguish whether the lack of short - period planets is primordial or due to engulfment by central stars .
deriving reliable mass and evolutionary status for evolved stars is highly required for further investigation of formation and evolution of planetary systems around intermediate - mass stars . |
amidst the rich phenomenology of @xmath1-mesons lies a severe problem so old that new solutions are rarely attempted associated with the strangeness - changing @xmath17=1 , nonleptonic decays of the short- and long - lived states @xmath18 experimentally , there is a large enhancement of the isospin-@xmath19 decays .
this phenomenon is particularly striking in the @xmath20-wave @xmath21 mode , where the measured rates @xcite exhibit the ratios @xmath22 which are in enormous disagreement with the naive expectations @xmath23 from perturbative electroweak calculations .
it is useful to translate the above in terms of isospin amplitudes @xmath24 for the final @xmath21 state .
the @xmath25 state is forbidden by bose symmetry and thus the transition amplitudes can be parametrized as @xcite @xmath26 where the @xmath21-scattering phase shifts @xmath27 arise as a consequence of watson s theorem .
comparison with the data in ( [ eqn : rates ] ) leads to the @xmath0 rule for kaons @xmath28 whose origin remains a mystery despite five decades of theoretical investigation .
the difficulty presumably arises from the nonperturbative nature of quantum chromodynamics ( qcd ) at low energies @xmath29 , where confinement reigns and chiral symmetry for the light quarks @xmath4 is believed to be spontaneously broken by the formation of a quark condensate @xmath30 below the chiral symmetry - breaking scale @xmath31 gev , great progress has been made in charting the low - energy structure of qcd through the use of chiral @xmath32 perturbation theory @xmath33 .
the method relies on an alternative expansion parameter to the strong coupling @xmath34 , _
viz_. low - energy scattering amplitudes and matrix elements can be described by an asymptotic series @xmath35 in powers of @xmath36 momentum and quark masses @xmath37 .
the amplitudes of the series are then calculated in a systematic manner by constructing the most general effective lagrangian consistent with the underlying symmetries of qcd .
for example , the strong interactions of @xmath7 mesons are described in the leading order ( lo ) of @xmath33 by @xmath38 where @xmath39 mev is the pion decay constant , and @xmath40 is proportional to the @xmath4 quark mass matrix . here
@xmath41 is an @xmath42 field , where the octet of pseudoscalar nambu - goldstone ( ng ) bosons @xmath43 parametrize the coset space @xmath44 with group action @xmath45 the method can likewise be applied to the weak interactions and used to simulate nonleptonic decays . the chiral structure of the weak currents @xmath46 constrains the number of allowed operators .
for instance , in the leading order of @xmath47pt@xmath48 , two octets can be constructed with the required transformation properties , _
viz_. the derivative operator @xcite @xmath49 and a weak mass operator @xcite @xmath50 as an @xmath42 octet , @xmath51 is necessarily isospin-@xmath19 , however the requirement of a stable vacuum ensures that this operator does not contribute to physical processes . in this regard
, it can be shown @xcite that the effective potential is minimized by a chiral rotation such that @xmath52 and @xmath40 is diagonal and positive . by aligning the vacuum in this way ,
all trace of @xmath51 is removed from the effective lagrangian , and thus by including the @xmath53-spin triplet component @xcite of a * 27 * operator @xmath54 the nonleptonic decays are described by the compact expression @xmath55 in general , the low - energy coefficients @xmath56 are not fixed by symmetry arguments alone .
this constitutes the essence of the @xmath0 puzzle : why is @xmath57 unreasonably large ( @xmath58 22 ) when compared with simple quark - model estimates for @xmath59 @xmath1-decay amplitudes ? explaining this ` octet dominance ' has been tackled with a variety of techniques including the many color @xmath60 limit @xcite , qcd sum rules @xcite , and direct evaluation on the lattice @xcite , each with varying degrees of success .
the goal of this talk is to present an alternative line of investigation @xcite and argue that the @xmath0 rule for @xmath1-decays ( eq . ( [ eqn : ratio ] ) ) is intimately linked to the behavior of @xmath3 in the _ asymptotic _ infrared limit .
as is well known , the running of @xmath3 with scale @xmath61 is governed by the qcd @xmath62-function . at low energies
@xmath63 , heavy quarks @xmath64 decouple from the theory , and thus there are two logical possibilities for the resulting three - flavor theory ( fig . [
fig : beta ] ) : -function with three light quarks @xmath4 .
the dashed line shows the case when the strong running coupling @xmath3 undergoes continued growth with decreasing scale @xmath61 ( item ( [ unbound ] ) in text ) , while the solid line shows @xmath3 flowing to an infrared fixed point @xmath2 ( item ( [ ir fixed pt ] ) ) . for completeness
, we also include the case where @xmath3 diverges at a finite value of @xmath61 ( dotted line ) , but we emphasize that this scenario is of _ no physical relevance _ since it produces poles in green s functions in the spacelike momentum region ( i.e. a tachyon or
landau ghost " ) .
[ fig : beta ] ] 1 .
* growth without bound .
* if the integral @xmath65 is divergent , then the solution to the renormalization group equation @xmath66 implies that as @xmath61 decreases , @xmath3 grows in value , becoming infinite in the infrared limit @xmath67 .
analyses based on lattice simulations @xcite suggest that this scenario occurs for @xmath68 quark flavors ( pure yang - mills).[unbound ] 2 . * infrared fixed point at finite coupling . * in this scenario , @xmath69 diverges because of a simple zero in @xmath70 at an infrared fixed point @xmath71 . from a purely phenomenological point of view , this situation is perhaps the most interesting for it implies that the gluonic term @xmath72 in the trace anomaly of the energy - momentum tensor @xmath73 , @xmath74 vanishes at the fixed point @xmath2 and thus @xmath73 becomes traceless in the chiral limit @xmath75 naively , one might consider this scenario a phenomenological disaster : does a traceless @xmath76 not imply invariance under scaling transformations @xmath77 and hence a continuous mass spectrum ? in qcd , the answer is in the negative since the attendant strong gluon fields still drive quark condensation @xmath78 , from which one concludes that the scale and of course , chiral symmetry is in fact _
spontaneously broken_. may act as a chiral - scale condensate was considered prior to the advent of qcd in refs .
goldstone s theorem and approximate chiral - scale symmetry then implies that there are nine pseudo - ng bosons in the low - lying spectrum : the usual pseudoscalar octet @xmath7 and a @xmath79 qcd dilaton @xmath8 , whose mass is ( mostly ) set by the explicit breaking term @xmath80 in ( [ scale]).[ir fixed pt ] it is worth emphasizing that it is unclear from the literature which scenario is actually realized in qcd , and in particular , how sensitive the results are to the number of active quark flavors .
part of the problem resides in the fact that for each calculation , a nonperturbative _ definition _ for @xmath3 must be chosen , and thus comparing results from different approaches is rarely straightforward .
popular choices include the schrdinger functional method on the lattice , @xcite solutions to dyson - schwinger equations @xcite for three - gluon and ghost- or quark - gluon vertices , the method of effective charges @xcite , and ads / cft inspired models @xcite . despite the lack of consensus
, we take the view that an infrared fixed point in qcd ( item ( [ ir fixed pt ] ) ) should be taken seriously , and the remainder of this talk concerns the phenomenological implications which arise from this scenario .
while it may seem dramatic to introduce a new ng boson into the low - energy spectrum of qcd , there is in fact a very natural candidate for @xmath8 : the @xmath9 resonance , whose mass and width have been determined with remarkable precision through an analytic continuation of the roy equations @xcite , @xmath81 since this pioneering work , many studies ( see e.g. fig . 8 in ref .
@xcite ) have extracted @xmath82 pole parameters consistent with ( [ eqn : f0 pole ] ) , to the extent that the particle data group @xcite has recently updated their listing to account for the greatly reduced uncertainties . in all determinations of the kind given by ref .
@xcite , the real part of the @xmath82 pole is found to be @xmath83 .
this places @xmath82 right in the middle of the ng sector @xmath7 and encourages us to revisit the observation @xcite that @xmath33 works well except for @xmath79 amplitudes with @xmath36 extrapolations in momentum . since @xmath82 is a broad flavor singlet coupled strongly to @xmath7 , it should dominate low - energy scattering .
however , this leads to the following problem : @xmath82 can not contribute to the lo terms @xmath84 in ( [ eqn : asy exp ] ) so the dominant effect must be generated at next to leading order ( nlo ) .
how does this square with the expectation that nlo terms are @xmath85 the lo prediction ? one alternative ( and the one we shall pursue ) is to note that if @xmath86 is treated a pseudo - ng boson ( dilaton ) , then all the convergence issues in the @xmath79 channel disappear .
to that end , we replace @xmath33 with a model - independent chiral - scale theory @xmath10 based on expansions in @xmath3 about @xmath2 .
our task is to construct an effective field theory of approximate scale and chiral @xmath6 symmetry . for the strong interactions we seek an effective lagrangian of the form @xmath87 = : { \cal l}^{d=4}_{\mathrm{inv } } + { \cal l}^{d>4}_{\mathrm{anom } } + { \cal l}^{d<4}_{\mathrm{mass}}:\,,\ ] ] where each term is distinguished by the scaling dimension @xmath88
: @xmath89 the rules for constructing lagrangians of this type were worked out long ago and we refer to refs .
@xcite for explicit details .
the salient features are as follows . for expansions in @xmath3 about @xmath2
, we have @xmath90 while the gluonic operator insertion @xmath91 ( obtained by acting @xmath92 on the callan - symanzik equation ) implies @xmath93 to realize scale invariance as a ng symmetry , a dilaton field @xmath8 is introduced with nonlinear scaling property @xmath94 so that @xmath95 is covariant and has @xmath96 : @xmath97 here @xmath98 is the dilaton decay constant , defined by the coupling of @xmath73 to the vacuum @xmath99 lagrangian operators of the required scale dimension are then obtained by multiplying operators such as @xmath100 by appropriate powers of @xmath95 . in the lo of @xmath10 the most general effective lagrangian for strongly interacting @xmath7 and @xmath8 mesons
is given by @xmath101 the low - energy constants @xmath102 are not fixed by symmetry arguments , however , the requirement of a stable vacuum in the @xmath8 direction implies that @xmath103 and @xmath104 are @xmath105 and not independent . for expansions about @xmath106
, the absence of tadpoles implies that all terms linear in @xmath8 must cancel : @xmath107 the relationship between @xmath98 and @xmath108 is deduced by an analogue of the goldberger - treiman relation ( fig .
[ fig : g_signn ] ) : @xmath109 -pole diagram in @xmath47pt@xmath110 for @xmath111 . ]
one - boson exchange combined with large-@xmath60 arguments @xcite leads to the estimate @xmath112 with a large model - dependent uncertainty .
we take the central value and obtain @xmath113 100 mev . from @xmath114
we obtain the dilaton mass @xmath115 and the effective @xmath116 coupling @xmath117 note the key feature of ( [ eqn : gspipi ] ) : it is mostly derivative so it has a small effect on @xmath21-scattering in the @xmath118 limit where @xmath119 and @xmath120 ( and thus @xmath121 ) remains fixed .
the vertex for an on - shell amplitude for @xmath122 is readily obtained , @xmath123 and generates part of the broad width for @xmath8 @xmath124 note that @xmath125 and thus nlo in the chiral - scale expansion . in @xmath47pt@xmath110 ,
the anomalous dimension @xmath127 of @xmath51 differs from that of @xmath128 .
consequently , the chiral - scale lagrangian includes a term @xmath129 whose @xmath8 dependence can not be eliminated by a chiral rotation .
instead , after vacuum alignment , the weak interactions are described in the lo of @xmath10 by @xmath130 noting that @xmath131 represents quark - gluon operators with differing dimensions at @xmath132 . as a result , @xmath133 and @xmath8 mix through the interaction @xmath134 , where the effective coupling @xmath135 produces a @xmath136 amplitude @xmath137 ( fig .
[ fig : k_pipi ] ) . for @xmath126 .
the dominant @xmath8-pole implies that the couplings @xmath138 and @xmath139 are allowed to be of similar magnitudes . ]
the value of this coupling can be fixed by data on @xmath140 and @xmath141 , with the result @xmath142 accurate to about 30% precision .
subject to the uncertainty on @xmath143 , we estimate @xmath144 thereby accounting for the large @xmath145 amplitude @xmath146 in ( [ eqn : ratio ] ) .
it follows that @xmath138 and @xmath139 are allowed to have similar magnitudes ( with the latter fixed precisely by @xmath147 ) and thus the octet dominance hypothesis is no longer necessary to explain the @xmath16 rule .
another welcome feature of @xmath10 is that is makes a _ nonperturbative _ prediction for the drell - yan ratio @xmath148 at @xmath149 .
this is so because the electromagnetic trace anomaly @xcite @xmath150 implies an effective @xmath151 coupling @xmath152 here @xmath153 and @xmath154 are the electromagnetic field strength and fine - structure constant . through direct calculation , ( [ eqn : l_sig2gam ] ) can be shown to be @xcite @xmath155 dispersive analyses @xcite of @xmath13 indicate that the residue of the @xmath82 pole can be unambiguously extracted from data . within the uncertainty in the value of @xmath98 ,
we use the updated value @xcite @xmath156 to obtain @xmath157
we have seen that an infrared fixed point in the three flavor @xmath62-function of qcd leads to an extended ng sector @xmath158 , where the @xmath9 resonance is identified with @xmath8 as the dilaton of spontaneously broken scale symmetry . despite a seemingly drastic change to the accepted low - energy structure of qcd ,
the resulting chiral - scale perturbation theory @xmath10 is rather conservative : @xmath82 pole terms are promoted to leading order in @xmath10 ( thereby evading @xmath33 s convergence problems in the @xmath79 channel ) , yet the successful leading order predictions of @xmath33 are preserved . while our key result is a simple explanation for the @xmath0 rule in @xmath1-decays , @xmath10 is certainly a general framework and it would be interesting to examine the consequences of the effective theory for other well studied conundrums such as @xmath159-violation , rare kaon decays , and @xmath160 .
l. c. t. thanks the organizers and c. dominguez in particular for a most pleasant workshop and for the opportunity to present this work .
l. c. t. has benefited from discussions with profs .
h. leutwyler , m. jamin , and p. minkowski , and thanks them for their useful comments . | in this talk , we present an explanation for the @xmath0 rule in @xmath1-decays based on the premise of an infrared fixed point @xmath2 in the running coupling @xmath3 of quantum chromodynamics ( qcd ) for three light quarks @xmath4 . at the fixed point ,
the quark condensate @xmath5 spontaneously breaks scale _ and _ chiral @xmath6 symmetry .
consequently , the low - lying spectrum contains _ nine _ nambu - goldstone bosons : @xmath7 and a qcd dilaton @xmath8 .
we identify @xmath8 as the @xmath9 resonance and construct a chiral - scale perturbation theory @xmath10 for low - energy amplitudes expanded in @xmath3 about @xmath2 .
the @xmath0 rule emerges in the leading order of @xmath10 through a @xmath8-pole term @xmath11 , with a @xmath12 coupling fixed by data on @xmath13 and @xmath14 .
we also determine @xmath15 for the nonperturbative drell - yan ratio at @xmath2 . |
quantum entanglement has become one of the most important issues since the rapid developments in quantum information science @xcite .
much research has been devoted to studying entanglement as induced by a direct interaction between the individual subsystems @xcite .
very recently , a lot of attention has been focused on _ reservoir - induced _
entanglement @xcite with the purpose to shed light on the generation of entanglement , and to better understand quantum decoherence
. furthermore , shot noise @xcite has been identified as a valuable indicator of particle entanglement in transport experiments.@xcite a well studied example is the doubling of the full shot noise in * s - i - n * tunnel junctions@xcite , where * n * is a normal metal , * s * stands for a superconductor , and * i * is an insulating barrier .
the origin of this enhancement comes from the break - up of the spin - singlet state , which results in a quick transfer of two electrical charges . in this paper
, we demonstrate how the dynamics of entangled excitons formed by superradiance can be revealed from the observations of current fluctuations .
a doubled zero - frequency shot noise is found for the case of zero subradiant decay rate .
we relate the particle noise to photon noise by calculating the first order photon coherence function .
furthermore , strong reservoir coupling acts like a continuous measurement , which is shown to suppress the formation of the entanglement , except for the state of maximum entanglement .
these novel features imply that our model provides a new way to examine both the bunching behavior and a zeno - like effect of the reservoir induced entanglement .
the effect appears in double quantum dots embedded inside a _ p - i - n _ junction@xcite .
it involves superradiant and subradiant decay through two singlet and triplet entangled states , @xmath0 and @xmath1 , and one ground state @xmath2 , where @xmath3 ( @xmath4 ) represents one exciton in dot 1 ( 2 ) .
electron and hole reservoirs coupled to both dots have chemical potentials such that electrons and holes can tunnel into the dot . for the physical phenomena we are interested in ,
the current is assumed to be conducted through dot 1 only ( fig .
therefore , the exciton states @xmath5 ( in dot 2 ) can only be created via the exciton - photon interactions .
the exciton - photon coupling is described by an interaction hamiltonian @xmath6+h.c.\},\end{aligned}\ ] ] where @xmath7 is the photon operator , @xmath8 is the coupling strength , @xmath9 is the position vector between the two quantum dots . here , @xmath10 is a constant with the unit of the tunneling rate .
the dipole approximation is not used in our calculation since we keep the full @xmath11 terms in the hamiltonian .
the coupling of the dot states to the electron and hole reservoirs is described by the standard tunnel hamiltonian @xmath12 where @xmath13 and @xmath14 are the electron operators in the left and right reservoirs , respectively , and @xmath15 denotes one - hole state in dot 1 .
@xmath16 and @xmath17 couple the channels @xmath18 of the electron and the hole reservoirs .
here we have neglected the state @xmath19 for convenience .
this can be justified by fabricating a thicker barrier on the electron side so that there is little chance for an electron to tunnel in advance @xcite .
the rates @xmath20 ( electron reservoir ) and @xmath21 ( hole reservoir ) for tunneling between the dot and the connected reservoirs can be calculated from @xmath22 by perturbation theory . in double quantum dots , decay of the excited levels
is governed by collective behavior , i.e. superradiance and subradiance .
the corresponding decay rates for the state @xmath23 and @xmath24 can be obtained from @xmath25 and are denoted by @xmath26 and @xmath27 , respectively .
we are then in the position to set up the equations of motion for the time - dependent occupation probabilities
@xmath28 @xmath29 @xmath30 , @xmath31 @xmath32 , of the double dot states .
together with the normalization condition @xmath33 , the equations of motion are laplace - transformed into @xmath34space @xcite for convenience and read @xmath35+\gamma _ { u}\left[\frac{1}{z}% -\hat{n}_{s_{0}}(z)-\hat{n}_{t_{0}}(z)-\hat{n}_{d}(z)\right ] , \notag \\ z \hat{n}_{t_{0}}(z ) & = & -ig\left[\hat{p}_{t_{0},d}(z)-\hat{p}_{d , t_{0}}(z)\right]+\gamma _ { u}\left[\frac{1}{z}% -\hat{n}_{s_{0}}(z)-\hat{n}_{t_{0}}(z)-\hat{n}_{d}(z)\right ] , \notag \\ z \hat{n}_{d}(z ) & = & ig\left[\hat{p}_{s_{0},d}(z)-\hat{p}_{d , s_{0}}(z)+\hat{p}_{t_{0},d}(z)-\hat{p}_{d , t_{0}}(z ) \right]-\frac{2\gamma _ { d}}{z}\hat{n}_{d}(z).\end{aligned}\ ] ] here , @xmath36 and @xmath37 are off - diagonal matrix elements of the reduced density operator of the double dots , whose laplace - transformed equations of motion close the set ( 3 ) : @xmath38 note that in getting above equations , one has to do a decoupling approximation of dot operators and photon operators .
this means we are interested in small coupling parameters here , and a decoupling of the reduced density matrix @xmath39 is used : @xmath40 .
@xcite the stationary tunnel current @xmath41 can be defined as the change of the occupation of @xmath42 for large times @xmath43 and is given by @xmath44,\]]where we have set the electron charge @xmath45 for convenience . in a quantum conductor in nonequilibrium ,
electronic current noise originates from the dynamical fluctuations of the current being away from its average . to study correlations between carriers ,
we relate the qubit dynamics with the hole reservoir operators by introducing the degree of freedom @xmath46 as the number of holes that have tunneled through the hole - side barrier and write @xmath47eqs .
( 6 ) allow us to calculate the particle current and the noise spectrum from @xmath48 which gives the total probability of finding @xmath46 electrons in the collector by time @xmath43 .
in particular , the noise spectrum @xmath49 can be calculated via the macdonald formula @xcite .
@xmath50,\ ] ] where @xmath51 .
solving eqs .
( 6 ) and ( 3 ) , we obtain @xmath52\}.\ ] ] in the zero - frequency limit , eq . ( 8) reduces to @xmath53_{z=0}\},\ ] ] which is analogous to a recent calculation of noise in dissipative , open two - level systems @xcite .
to display the dependence of carrier correlations on the dot distance @xmath54 , fig .
2 shows the result for zero - frequency noise @xmath55 as a function of the inter - dot distance . in plotting the figure ,
the tunneling rates , @xmath20 and @xmath21 , are assumed to be equal to @xmath56 and @xmath57 respectively . here , a value of 1/1.3ns for the free - space quantum dot decay rate @xmath58 is used in our calculations @xcite . as shown in fig .
2 , the fano factor @xmath59is enhanced by a factor of 2 as the dot distance @xmath54 is much smaller than the wavelength ( @xmath60 ) of the emitted photon . to explain this enhancement , we approximate the fano factor in the limit of small subradiant decay rate , i.e. @xmath61where we obtain @xmath62.\]]this is analogous to the case of the single electron transistor near a cooper pair resonance as discussed recently by choi and co - workers @xcite . in their calculations , the fano factor is expressed as @xmath63 in the strong dephasing limit ( @xmath64 , where @xmath65 is the josephson coupling energy )
, the zero - frequency shot noise is also enhanced by a factor of 2 .
since the doubled shot noise in josephson junction is attributed to the bunching behavior of cooper pairs ( in singlet state ) , we then conclude the enhancement in our system is also due to the entanglement induced by the photon reservoir . and @xmath66 , respectively .
the inset shows the value of @xmath67 is equal to that of one - dot limit for @xmath68 ( dashed line ) , while it approaches _ zero noise _ as @xmath69 ( red line ) . ]
it is worthwhile to compare the current noise with the _ photon noise _ generated by the collective decay of the double dot excitons . in order to do so
, we have calculated the power spectrum of the fluorescence spectrum13 , which can be expressed as @xmath70e^{i\omega \tau } d\tau , \ ] ] where @xmath71 $ ] is the first order coherence function and reads @xmath72 & \propto & \left| 1+e^{i2\pi d/\lambda } \right| ^{2}\left\langle p_{s_{0},d}(0)p_{d , s_{0}}(\tau ) \right\rangle
\notag \\ & & + \left| 1-e^{i2\pi d/\lambda } \right| ^{2}\left\langle p_{t_{0},d}(0)p_{d , t_{0}}(\tau ) \right\rangle .\end{aligned}\ ] ] the two time - dependent correlation functions in the above equation can be calculated from the quantum regression theorem , and the numerical result of @xmath73 is shown explicitly in the inset of fig .
2 . as can be seen , the value of @xmath74 is equal to that of the one - dot case for @xmath75 ( dashed line ) , while it approaches zero as @xmath76 ( red line ) . in the limit of @xmath77 ,
one observes no photon emission from the double dot system since the exciton is now in its maximum entangled state and does not decay .
this feature implies that photon noise is suppressed by the bunching of excitons , and its behavior is opposite to that for the electronic case @xcite .
now we investigate how the measurement affects the shot noise spectrum . in the usual zeno paradox @xcite , a two - level system ( qubit )
is completely frozen under a series of measurements , whose time interval @xmath78 is much smaller than the memory time of the reservoir . in our model , the presence of the exciton state can be viewed as the excited state , and whether or not the next hole can tunnel in is determined by the occupation of this state .
similar to the quantum zeno effect , the tunneling of holes at the hole - side tunneling rate @xmath21 can be thought as a continuous measurements .
the interval time
@xmath78 is then inversely proportional to @xmath79 fig .
3 represents the effects of measurements on the frequency - dependent shot noise spectrum .
the numerical results for the tunneling rate @xmath80 @xmath58 and @xmath81 @xmath82 are demonstrated by solid and dashed curves , respectively . if the subradiant decay rate is set to zero , one obtains the doubled shot noise as mentioned above . without superradiance ,
the values of the fano factor are always below unity as shown by the right inset of fig .
an interesting feature is that the half - width of the spectrum is narrowed for strong measurements ( @xmath80 @xmath58 ) .
if one increases the electron - side tunneling rate @xmath20 , there exists no such behavior .
this implies that the effective decay rate is reduced in the presence of strong measurements .
( maximum superradiance , @xmath83 , @xmath84 ) .
solid and dashed lines correspond to @xmath85 @xmath86 and @xmath81 @xmath87 , respectively .
right inset : the case of no superradiance . left inset : expectation value of the excited states @xmath88 and @xmath89 as a function of @xmath79 ] to investigate thoroughly the underlying physics , we plot the expectation value of the excited states @xmath88 and @xmath91 as a function of @xmath21 in the left inset of fig .
one clearly finds the occupation probabilities grow with increasing @xmath21 , and both of them approaches the value of @xmath92 .
this not only means the measurements tend to localize the exciton in its excited state @xcite , but also tells us the entanglement is destroyed under the strong measurements .
however , in the limit of no subradiance ( @xmath93 ) , the occupation probability of the singlet state is always equal to one , i.e. maximum entanglement is robust against strong measurements .
this is because once the maximum entangled state is formed , the total probability in the excited states is also maximum .
strong measurements on the ground state @xmath94 have no influence on the singlet entangled state .
a few remarks about experimental realizations of the present model should be mentioned here .
one should note that biexciton and charged - exciton effects are not included in our present model .
inclusion of these additional states is expected to suppress the enhancement of the shot noise , i.e. degrees of entanglement .
however , this can be controlled well by limiting the value of bias voltage so that only the ground - state exciton is present @xcite . to produce the maximum entangled state
, one can also incorporate the device inside a microcavity @xcite .
there are two advantages of this design : the maximum entanglement can be generated even for remote separation of the two dots , and forster process @xcite is avoided at this distance .
as for the problem of decoherence due to interactions with phonons , recent experimental data have shown that the exciton - phonon dephasing rate is smaller than the radiative decay one in a quantum dot .
this means that due to the discrete energy level scheme in a quantum dot , the effect of _ phonon - bottleneck _ tends to suppress the exciton - phonon interaction @xcite .
although the present model describes tunneling of electrons and holes into semiconductor quantum dots , the whole theory can be applied to electron tunneling through coupled quantum dots which are interacting via a common phonon environment @xcite . in conclusion
, we have demonstrated that the shot noise of superradiant entangled excitons is enhanced by a factor of two as compared to the poissonian value .
this enhancement was attributed to exciton entanglement , induced by the electromagnetic field ( common photon reservoir ) , and an analogy to the cooper pair box was made .
second , we found the relaxation behavior of the qubits in the presence of strong measurements , and the zeno - like effect tends to destroy the entanglement and localize the qubits in the excited states .
99 c. h. bennett and d. p. divincenzo , nature ( london ) * 404 * , 247 ( 2000 ) ; t. pellizzari _
et al . _ ,
lett . * 75 * , 3788 ( 1995 ) ; j. i. cirac and p. zoller , phys .
lett . * 74 * , 4091 ( 1995 ) ; k. molmer and a. sorensen , phys .
* 82 * , 1835 ( 1999 ) .
m. s. kim _
et al . _ ,
phys . rev . * a * 65 , 040101 ( 2002 ) ; daniel braun , phys .
rev . lett . * 89 * , 277901 ( 2002 ) ; fabio benatti _ et al .
lett . * 91 * , 070402 ( 2003 ) ; r. ruskov and a. n. korotkov , phys .
b * 67 , 241305 ( 2003 ) . | the shot noise produced by tunneling of electrons and holes into a double dot system incorporated inside a p - i - n junction is investigated theoretically .
the enhancement of the shot noise is shown to originate from the entangled electron - hole pair created by superradiance .
the analogy to the superconducting cooper pair box is pointed out .
a series of zeno - like measurements is shown to destroy the entanglement , except for the case of maximum entanglement . |
the sign change of the hall effect observed in the superconducting state of many high-@xmath0 superconductors is one of the most puzzling problems in the electrodynamics of these materials @xcite .
one may ask then what is the doping dependence@xcite of this hall anomaly and how the vortex - lattice melting @xcite affects the hall behavior ? alas , due to the complexity of the vortex matter many related problems are still not answered satisfactory if at all .
it is quite possible that the sign reversal of the temperature dependence of the hall effect could be closely related to charging of the vortices @xcite .
there is no doubt @xcite that the experimental solution of this enigma would provide the key towards understanding the various electromagnetic phenomena . on the other hand ,
the currently existing theoretical models often lead to conflicting results thus making it difficult to discriminate between all those competing explanations .
in such a situation we feel it appealing to accelerate the selection by looking for simplicity in experiments with artificial structures where many of the complications typical for the real systems are avoided .
the aim of the present paper is to propose an experiment for determination of the vortex charge employing transport measurement in a layered metal - insulator - superconductor ( mis ) system .
we shall require that the quality of the insulator - superconductor interface be extremely high and the insulator layer be very thin .
such a layered mis structure incorporating a high-@xmath0 film can be manufactured by the contemporary technology of atomic - level engineering of superconducting oxide multilayers and superlattices @xcite .
in fact , structures of the kind are now being in use for purposes of the fundamental research@xcite in the physics of high-@xmath0 superconductors , therefore the vortex charge problem can find its solution thanks to the technological progress .
the simplest possible idea behind the search for the vortex charge is to study the electrostatic effect related to charged vortices @xcite , which is analogous to electrostatic effects originating , in turn , in the bernoulli effect @xcite due to circulating currents in a thin superconducting film in a vortex - free state ; the numerical value of the angular momentum @xmath1 is irrelevant for the current - induced contact - potential difference .
the paper is organized as follow : in sec .
[ sect : model ] we derive the formula for the vortex charge @xmath2 expressed via the effective mass of cooper pairs @xmath3 as well as the expression for the interface hall conductivity @xmath4 in sec . [
sect : set - up ] we will analyze our proposed experimental set - up for determination of the vortex charge by measuring the hall resistance of the vortex charge currents . an overview is made in sec .
[ sect : effective mass ] of different experimental methods for determination of the cooper pair mass : the surface hall current @xcite , subsection [ sect : surface hall ] ; the bernoulli effect @xcite , subsection [ sect : cpd ] , and the electrostatic charge modulation@xcite , subsection [ sect : charge modulation ] .
it is finally concluded in sec .
[ sect : conclusions ] that the vortex charge @xmath2 and the effective mass @xmath3 of fluctuation cooper pairs fall into the class of the last unresolved problems in the physics of superconductivity .
these important parameters enter the theories of a number of phenomena related to electrodynamics of superconductors and can be simultaneously determined by standard electronic measurements .
contemporary layer - by - layer growth of layered oxide structures gives the unique chance for finding @xmath2 and @xmath3 for high-@xmath0 materials but some of the proposed experiments can be realized also in mis structures with conventional superconductors .
this section gives an account of the vortex charging due to the bernoulli effect within the framework of london electrodynamics .
for a superconductor in thermodynamic equilibrium the electrochemical potential @xmath5 is constant and the space distribution of the electric potential @xmath6 is determined by the bernoulli - torricelli theorem @xmath7 formally , this equation can be derived within the framework of the bcs theory using the statistical mechanics methods , but its physical meaning is very simple it is a consequence of the energy conservation .
we shall further stick to the standard notations for the effective mass of cooper pairs in the @xmath8-plane @xmath9 the superfluid velocity @xmath10 related to the current density @xmath11 the mass density of the superfluid @xmath12 and the total charge of the conduction band @xmath13 at zero temperature all charge carriers are superfluid and according to the bcs theory @xmath14 the temperature dependence of @xmath15 can be extracted from that of the london penetration depth for screening currents flowing in the cuo@xmath16 plane , @xmath17 where the use of si units is implied , @xmath18 although the temperature dependence of the superfluid ratio @xmath19 is related to the gap anisotropy , the hydrodynamic relation eq .
( [ eq : bernoulli ] ) remains invariant .
consider now a thin cuprate film thread by a perpendicular magnetic field @xmath20 as a first step we determine the distribution of the electric potential as a function of the distance to the vortex line @xmath21 for @xmath22 larger than the ginzburg - landau ( gl ) coherence length in the @xmath8-plane but smaller than the penetration depth , @xmath23 one can use the bohr - zommerfeld relation @xmath24 substituting @xmath25 from the above equation into the bernoulli theorem eq .
( [ eq : bernoulli ] ) we derive the current - induced change of the electric potential @xmath26 this equation is applicable not only to the volume of the superconductor @xmath27 but even to the superconducting surface @xmath28 which is supposed to be clean enough as well as to expose the properties of the bulk material .
the superconductor is capped by a thin insulating layer of thickness much smaller than the penetration depth , @xmath29 on top of the latter a thin - normal - metal layer is evaporated , hence a plane capacitor configuration is achieved , being in fact realized as a metal - insulator - superconductor ( mis ) layered structure . for definiteness the electric potential of the normal plate is set to zero .
far from the vortex core , for @xmath30 the electric field @xmath31 of such a plane capacitor can be considered as being homogeneous , @xmath32 which is employed to express the induced on the normal plate surface charge density @xmath33 via the bernoulli potential @xmath34 @xmath35 where @xmath36 @xmath37 being the speed of light , and @xmath38 is the relative dielectric constant of the insulator .
we notice that @xmath33 has the same sign as the charge of the cooper pairs in the superconductor . on the other hand ,
the bernoulli potential keeps the cooper pairs on a circular orbits inside the vortex .
the radial electric force is then equal to the centrifugal force @xmath39 the electric potential attracts the cooper pairs and the charges with the same sign on the normal plate of the plane capacitor . in order to derive the total charge related to the vortex we have to integrate the charge density up to some maximum radius , @xmath40 corresponding to the screening length @xmath41 or the typical intervortex distance in case of high area density of vortex lines @xmath42 where @xmath43 ftm@xmath44 is the flux quantum . supposing that the insulator layer is thin enough , @xmath45 the integration of the surface density gives for the total vortex charge @xmath46 where @xmath47 is the bohr radius , @xmath48 kg is the mass of a free electron , and @xmath49 is a quantity analogous to the ginzburg - landau parameter @xmath50 according to our model the charge related to vortices is localized not in the vortex core but in the adjacent conducting layers : superconducting cuo@xmath16 planes in a real high-@xmath0 crystal or the normal layer in the model mis system . with this
we close the electrostatic consideration of the vortex charge , but rhe reader is reffered to a number of ingenious experiments related to electrostatics of vortices which are suggested in ref .
we believe , however , that the standard transport measurement have some advantage even if they are related to observations of pa - range and below .
the next important step is to address the vortex flow regime of the superconducting film when a strong enough dc current density @xmath51 is applied through the superconducting film .
this condition will create small dissipation and give rise to an electric field @xmath52 parallel to the current density . the electric field , in turn , creates a drift of the vortices with mean drift velocity in @xmath53-direction @xmath54 in a coordinate system moving with the vortex drift velocity @xmath55 the electric field is zero .
we suppose that @xmath56 is much smaller than the critical depairing velocity @xmath57 and the bernoulli potential is nearly the same as in the dissipation - free static regime .
along this line let us recall the fact that airplanes fly thanks to the bernoulli theorem that holds true for a unviscous dissipationless fluid , but the significant part of the ticket price covers the dissipated energy .
by the same token , for @xmath58 the vortex - induced charge has nearly the static value @xmath59 since the charge images will follow the vortices as shadows , the vortex flow will create a two dimensional ( 2d ) current density on the surface of the normal metal @xmath60 the electric field @xmath52 resides the superconducting layer , whereas the current @xmath61 exists in the normal slab .
the 2d hall conductivity directly gives the vortex charge @xmath62 for @xmath63 rectangular shape of the mis structure the voltage drop in the superconducting layer is @xmath64 the total current in the normal layer is @xmath65 and the interface hall resistivity is size - independent , @xmath66 where @xmath67 is the fundamental resistance determined by the quantum hall effect ( qhe ) . since the vortex charge @xmath68
the experiment would face the problem of measuring huge hall resistances .
this sets the first technological requirement regarding to the quality of the insulating layer in order to avoid the leakage currents the resistance @xmath69 of the plane capacitor should satisfy the relation @xmath70 in the present model we used the hydrodynamic approach applicable for extreme type - ii superconductors and completely neglected the influence of the geometrically small vortex core .
however the states in vortex core can have some influence in the total charge of vortex core@xcite . in order
qualitatively to `` interpolate '' a real situation with moderate ginzburg - landau parameter @xmath71 let us analyze the interface hall current for a type i superconductor . in this case the normal `` cores '' are domains of normal metal surrounded by circulating superconducting currents .
this problem , certainly , is only of an academic interest and is irrelevant for the oxide superconductors .
if the superconducting layer of a mis structure is of type - i superconductor , in a perpendicular magnetic field @xmath72 the magnetic field in the normal domains is equal to the thermodynamic one @xmath73 and is zero in the superconducting domains .
the relative area of the normal regions is @xmath74 , correspondingly the part of the superconducting area is @xmath75 thus @xmath76 and the external field is equal to the mean field @xmath77 the contact potential difference between the normal and the superconducting phase ( see eq .
( [ eq : bulk bernoulli ] ) bellow ) is @xmath78 this contact potential difference creates , in turn , a difference in the charge density at the surface of the normal layer in front of the normal domain @xmath79 where a plane capacitor configuration is implied .
when an electric field @xmath52 is applied in the superconducting layer the normal domains acquire a drift velocity in @xmath53-direction @xmath54 again , in the mobile coordinate system the domain structure is static and the mean electric field is zero .
the extra charges induced in the normal layer follow the moving normal domains and for the 2d current @xmath80 at the surface of the normal plate we obtain @xmath81 this very small interface hall conductivity vanishes at @xmath0 and its detection requires fa sensitivity .
for comparison with eq .
( [ eq : vortexcharge ] ) here we also give the expression for the induced charges per flux quantum @xmath82 of course , around every normal domain in a type - i superconductor @xmath83 having derived the formulae , eq . ( [ eq : conductivity ] ) and eq .
( [ eq : typei sigma ] ) , concerning the new predicted effect we proceed with more detailed discussion and description of the proposed new experiment in the next section .
to begin with , we have sketched a `` _ _ gedanken _ _ '' experimental set - up in fig .
[ fig : setup ] .
the contemporary technology of layer - by - layer growth of oxide superconductors opens the possibility for realization of such a layered structure a superconducting film protected by an insulating plate .
moreover , we consider that a mis plane capacitor is one of the simplest possible systems employed in the fundamental research towards further technical applications .
therefore we believe that the suggested experiment could become a standard tool in studying the quality of the insulator - superconductor interface . in order to check whether this idea is another case of a science fiction or , vice versa ,
is a smoking gun we provide below a numerical example involving an acceptable set of parameters which have been collected from various references : @xmath84 ( ref.@xcite ) , @xmath85 nm , @xmath86 nm , @xmath87 nm ( ref.@xcite ) , @xmath88 nm ( ref.@xcite ) . for an illustration we take as well : @xmath89 mt , @xmath90 v / cm , and @xmath91 mm . the value of @xmath72 we chose imply for the following parameters : @xmath92 @xmath93 and @xmath94 for a model estimate we also take @xmath95 nm . it is now straightforward to work out the vortex charge at liquid - helium temperature , _
i.e. _ in the temperature range far below @xmath96 in this case the substitution of the above mentioned set of parameters in eq .
( [ eq : vortexcharge ] ) gives @xmath97 the so estimated @xmath98 is in agreement with another model evaluation due to khomskii and freimuth@xcite .
further , eq . ( [ eq : hall resistance ] ) gives @xmath99 m@xmath100 and the electric field chosen gives for the voltage @xmath101 mv , therefore for the hall current we have @xmath102 pa .
lastly , the vortex drift velocity @xmath103 km / s , which is one order of magnitude smaller than the depairing velocity at @xmath104 @xmath105 km / s .
we note that the resistance of the capacitor should be thus at least @xmath106 m@xmath107 for conventional superconductors similar evaluations show that effect is less but still observable .
one can consider , for example , a thin nb metal film grown by molecular beam epitaxy , and an al layer after oxidation in natural condition could give a good insulator layer .
all technologies for planar josephson junctions provide as a rule metal - insulator interface of sufficient quality .
only the insulator layer should be thick enough to prevent leakage tunneling .
the example analyzed above shows that the proposed experiment is in principle possible to be carried out but we find it difficult to anticipate all problems that could arise in the course of it .
for instance , due to a good capacitance cross - talk the noise created by the vortex motion in the superconducting layer will be transmitted to the normal layer thus disturbing the detection of the small hall current .
we believe , however , that similar problems could be surmounted , given the challenge of the novel physics underlying the vortex charge .
furthermore , it is quite possible that the charge , concentrated in the vortex core , is comparable to the charge outside , so only a detailed analysis within the microscopic theory can shine a light on the latter point . in order to verify whether a hydrodynamic approach based upon the bernoulli effect suffices to quantitatively describe the predicted vortex - charge interface current
one needs independent methods to determine the effective mass of the cooper pairs . in the next section
we will analyze similar experiments employing artificial mis structures .
before addressing the problem of measuring the cooper pair effective mass @xmath3 let us analyze a parallel between the latter issue and the civil engineering , where in a static approximation only the weight @xmath108 is essential for a construction .
in this approximation the masses could reach colossal values if we renormalize the earth acceleration @xmath109 .
the uncertainty , however , immediately disappears during the first earthquake when a dynamical problem should be solved .
just the same is the situation with the superconducting order parameter @xmath110in the static gl theory the superfluid density @xmath111 and the effective mass @xmath3 are inaccessible separately .
they are contained in the experimental parameters , such as the penetration depth eq .
( [ eq : depth ] ) , only via the ratio @xmath112 in order to determine the effective mass one has to investigate some dynamic phenomenon , which is time - dependent . due to phase invariance , however , the time @xmath113 could participate only in the gauge invariant derivative @xmath114 that is why electric field effects in superconductors are to be studied .
the subtle point is that the latter are already dynamic effects even if the electric fields are static .
one therefore needs to perturb the thermodynamic equilibrium of the superconductor as slightly as possible and all methods for determination of the effective mass @xmath3 of cooper pairs thus become effectively ac methods , based on the electrostatic effects in superconductors .
the set - up proposed to determine the vortex charge , fig .
[ fig : setup ] , is a mis device having four terminals .
probably the most simple method to accomplish the task would be to use the same mis structure without making any contacts on the superconducting layer and to investigate the surface hall current@xcite as described in the next subsection .
this physical effect reffers to the 2d surface currents @xmath115 at the surfaces of a thin ( @xmath116 ) superconducting film induced by a normal - to - the - layer electric induction @xmath117 and parallel - to - the - layer magnetic field @xmath118 @xmath119 where the cooper pair mass @xmath3 is the material constant of the effect .
this is an electrostatic effect and the superconducting film is in vortex - free state .
the dissipation is zero and the superconductor is in thermodynamic equilibrium .
a symmetric layered structure is grown by capping of the superconducting film with an insulator layer .
two normal metal layers are evaporated on the protecting insulator layer and on the back side of the substrate thus achieving a plane capacitor configuration .
the normal - metal electrodes are circles with radius @xmath120 and a cartoon of the experimental set - up in corbino geometry is shown in fig .
[ fig : surface_hall ] .
exploiting the axial symmetry of the geometry eq .
( [ eq : surface hall ] ) reads as @xmath121 and for the total magnetic moment of the circulating currents we have @xmath122 this small magnetic moment could be difficult to detect against the large background due to the dc magnets creating @xmath123 we derive a static magnetic moment and the next natural step is to consider in a quasistatic approximation the electric induction @xmath124 as being time - dependent , @xmath125 .
the ac magnetic moment can be detected by the electromotive voltage @xmath126 induced in the solenoid having @xmath127 turns per unite length .
the total charge of the capacitor is @xmath128 and the time derivative of the electric induction , @xmath129 can be expressed by the current @xmath130 charging the capacitor .
for the electromotive voltage we finally obtain the equation @xmath131 where @xmath132 is the effective resistance describing this new electrodynamic effect created by the effective mass @xmath133 the experimental difficulties might be related with the careful compensation of the mutual inductance @xmath134 between the solenoid and the ac generator charging the mis plane capacitor . the rigorous analysis of the experiment requires the knowledge of the break - through voltages of the mis structure and the noise induced in the detecting coil , but in any case this auxiliary experiment would be easier to perform than the detection of vortex charge currents . in the following
we will also provide an elementary derivation of the formula for the surface hall current eq .
( [ eq : surface hall ] ) using the london electrodynamics .
let us trace the trajectory of a london _ superconducting electron _
( _ i.e. _ a cooper pair ) crossing the circular superconducting film during the charging of the mis plane capacitor , fig .
[ fig : surface_hall ] .
the _ superconducting electron _ leaves the inital surface of the film with zero velocity @xmath135 , experiences the lorentz force while traveling across the film @xmath136 and arrives at the opposite surface of the film at the @xmath137 _ i.e. _ @xmath138 with an additional azimuthal velocity component @xmath139 for @xmath140 all charges are superfluid and the electric induction determines the surface ( or 2d ) excess charge density @xmath141 for the surface current density of these polarization charges we therefore have @xmath142 for non - zero temperatures one has to take into account the thermal dissociation of the _ superconducting electrons _ , @xmath143 and the appearance of a normal fluid .
thus , taking into account the superfluid part , @xmath144 we recover the basic equation eq .
( [ eq : surface hall ] ) .
the bcs treatment certainly gives the same result because the london electrodynamics is not a mere , naive phenomenological alternative to the microscopic bcs theory , instead it should be viewed as an efficient tool to apply the bcs theory to low frequencies @xmath145 and small wave - vectors @xmath146 analogous experiment could be performed with a bulk crystal or thick film @xmath147 in this case in the initial eq .
( [ eq : surface hall ] ) and the final result , eq .
( [ eq : reff ] ) , the thickness of the film @xmath148 should be replaced with the penetration depth @xmath149 and the formula for the surface current then reads as @xmath150 the investigation of the temperature dependence of this effect can give a new method for determination of the temperature dependence of the penetration depth @xmath151 a srtio@xmath152 layer should be grown on the fresh cleaved surface of bi@xmath16sr@xmath16cacu@xmath16o@xmath153 crystal and a circular au electrode needs to be overgrown on the protecting layer .
one plate of the capacitor is the bulk high-@xmath0 crystal and the other one is the au layer . in order to avoid frozen vortices the constant magnetic field of the dc drive coil
must be applied after cooling down to low temperatures .
an ac voltage should be applied to the plane capacitor , a lock - in ammeter will measure the polarization current , and the induced due to the effect ac magnetic moment can be detected by a lock - in voltmeter connected to the detector coil .
for derivation of the above formula eq .
( [ eq : surface hall bulk crystal ] ) we have to use : ( i ) the distribution of the vector - potential at depth @xmath154 in the superconductor and some fixed radius @xmath155 @xmath156 where @xmath157 is the tangential magnetic field at the superconducting surface ; ( ii ) the london - bcs formula for the current response of the superconductor ( the polarization operator ) , @xmath158 and ( iii ) the formula for the bulk ( 3d ) density of the superfluid polarization charges @xmath159 where @xmath160 stands for the dirac @xmath160-function .
the effective mass @xmath3 can be determined not only by the surface hall effect but also from the bernoulli effect for which the bcs theory was developed by omelyanchuk and beloborodko @xcite as well as from all other predictions of the london theory . the existence of bernoulli effect for conventional superconductors is experimentally confirmed ; some references can be found , for example , in ref .
@xcite . in the next subsection
we give a brief account of the suggested here cooper pair mass spectroscopy .
the experimental set - up for a current - induced cooper pair mass spectroscopy is presented in fig .
[ fig : bernoulli ac ] .
the bernoulli effect is related to a current - induced contact - potential difference that can be measured by the electrostatic polarization of a normal metal electrode which covers the surface of the superconductor , forming a plane capacitor .
for the averaged change of the electric potential beneath the electrode the bernoulli theorem eq .
( [ eq : bernoulli ] ) gives @xmath161 _ i.e. _ the bernoulli potential is proportional to the averaged kinetic energy of cooper pairs beneath the electrode .
for thin films , @xmath162 the current across the layer is more or less homogeneous @xmath163 and we have to substitute in this equation @xmath164 then the formula for the bernoulli potential takes the form @xmath165 where @xmath166 is the kinetic inductance which can be measured directly by means of the mutual inductance method @xcite , @xmath167 is the area density of cooper pairs and @xmath168 is the averaged square of the 2d supercurrent beneath the electrode whose distribution has to be found by solving a magnetostatic problem . if two electrodes were grown on the superconductor surface , the bernoulli voltage @xmath169 can be considered as a voltage generator sequentially connected to two capacitors @xmath170 and @xmath171 as depicted in fig .
[ fig : bernoulli ac ] ( c ) .
the currents induced in the superconductor film are proportional to the current through the drive coil @xmath172 @xmath173 fig .
[ fig : bernoulli ac ] ( b , c ) .
the coefficient @xmath174 of this proportion @xmath175 has dimension of area @xcite . according to eq .
( [ eq : potential 2d ] ) an ac current @xmath176 will create an ac bernoulli voltage of doubled frequency @xmath177 initially , in the switched - off regime , when the detector contour resonates at frequency @xmath178 where @xmath179 the parasite mutual inductance between the drive coil contour and the detector contour must be carefully annulled by a small tunable mutual inductance @xmath180 after that taking @xmath181 in switched - on regime the detecting contour will resonate at doubled frequency @xmath182 @xmath183 in resonance conditions the bernoulli voltage can be directly detected by a lock - in voltmeter with a low noise preamplifier . if we know the penetration depth @xmath149 the measured bernoulli voltage , according to the eq .
( [ eq : potential 2d ] ) , gives the effective mass of cooper pairs @xmath133 if thick films , @xmath184 or bulk single crystals are to be used for such experiment we have to substitute in eq .
( [ eq : averaged potential ] ) the london formula for the velocity @xmath185 which is a trivial consequence of the newton equation @xmath186 for a nearly homogeneous electric field @xmath187 combining with eq .
( [ eq : ochsenfeld ] ) we obtain @xmath188 where @xmath189 is the pressure of the tangential to the superconducting surface magnetic field , and the @xmath190 is the temperature independent@xcite london - hall constant expressed via the total volume density of conduction band @xmath191 we consider the greiter , wilczek and witten s @xcite prediction for a temperature dependence of the london - hall constant as being erroneous and the problem still waits for its experimental solution . for type - i superconductors the eq .
( [ eq : bulk bernoulli ] ) can be applied up to @xmath73 obtaining in this way the contact potential difference eq .
( [ eq : cpd ] ) .
it is still questionable whether the thermal - induced contact - potential difference @xmath192 may be measured , but if the answer is positive this effect can be used to determine the thermodynamic critical field @xmath73 even for type - ii superconductors . in any case the fluctuation of the temperature should be taken into account in the experiments aiming to observe the bernoulli effect .
the realistic experiment proposed in ref .
@xcite can be substantially simplified ( cf . ref .
@xcite ) : the ring electrode capacitor can be substituted by a short circuit , and the central one could cover the whole facet .
we stress that at least one capacitive connection is indispensable .
the voltmeters do not measure any voltage difference but just the _ difference in the electrochemical potential _
( even nowadays almost 99% of the experimentalists are unaware of what a voltmeter really measures ) !
an error of the kind has prevented lewis @xcite during his pioneer investigations in the period 19531955 from observing the bernoulli effect in superconductors soon after it has been predicted by london@xcite .
lewis did not use the capacitive connection but he introduced all other necessary ingredients : lock - in voltmeter with nv sensitivity , ac magnetic field and doubling of the frequency .
now it is worthwhile measuring both the bernoulli effect and the surface hall effect in the same sample . at known total charge density @xmath193 eq . ( [ eq : bulk bernoulli ] ) , and penetration dept @xmath194 eq .
( [ eq : surface hall bulk crystal ] ) , the cooper pair mass can be determined as @xmath195 despite the @xmath196 papers published on high-@xmath0 superconductivity , without the cooper pair mass the physics of superconductivity remains hamlet without the prince , with only the role of ophelia performed by _ _
onnagata__. in the next subsection we briefly describe the only , to the best of our knowledge , reliable experiment for determination of effective mass @xmath133 when an electric voltage is applied to a mis plane capacitor the charging of the superconducting surface will create a change of the 2d superfluid charge density @xmath197 it is then easily worked out from eq .
( [ eq : kinetic inductance ] ) that this creates a modulation of the kinetic inductance and the derivative determines @xcite the effective mass @xmath198 this simple picture gets complicated due to @xmath0-changing upon electrostatic doping of the material , but below the critical region this experiment confirms @xcite a temperature independent effective mass @xmath3 .
when @xmath3 and all other parameters of the superconductor are already determined we can turn to the vortex charge problem .
the preceding analysis demonstrates that the proposed electronic measurements are feasible and the suggested experimental programme could be soon realized .
the appearance of the first good samples would immediately lead to the solution of the problem concerning the vortex charge and cooper pair mass .
these two parameters , @xmath2 and @xmath9 might fall in the lime - light of the physics of superconductivity in the nearest future . as a by - product
the cooper pair mass spectroscopy could become a standard tool for testing the quality of the superconducting films for future superconductor electronics . even in the present paper we suggested two or three new effects
thus there is no doubt that new physics will emerge from the development of the layer - by - layer oxide technology .
let us also list some of the main results of this study : the formulae for vortex charge eqs .
( [ eq : vortexcharge ] ) and ( [ eq : charge evaluation ] ) , vortex conductivity eqs .
( [ eq : conductivity ] ) and ( [ eq : hall resistance ] ) , surface hall current for bulk crystals eq .
( [ eq : surface hall bulk crystal ] ) , interface hall conductivity for type - i superconductors eq .
( [ eq : typei sigma ] ) , thermal - induced contact - potential difference eq .
( [ eq : thermal ] ) , etc .. finding a solution to the vortex charge problem by employing a model system , where the superconducting and the polarized layers are separated , will immediately trigger the answer to the question about what is the charge induced in the adjacent cuo@xmath16 layers by a pancake vortex .
one may further ask about the fate of the charge cloud when the pancake vortices `` polymerize '' in a vortex line , and what is the influence of the vortex charge in the vortex - vortex interaction and correlation . according to our analysis of the bernoulli effect
the charge will concentrate at the end of vortex lines , at kinks and sharp turns of stacks of pancake vortices .
needless to say , the clear solution of some model problems is always useful in the search for solution to the complex problems in material science .
the problem of determining the vortex charge by a transport measurement brings us back to one of the first ideas of the electron physics .
only two months after the discovery of the electron @xcite francis mott made the first attempt to observe the influence of the electric fields and surface charges on the conductivity of pt .
likewise , the vortex charge current has led us to another immortal idea of the xix century the kelvin vortex model of the `` atom '' .
starting from a hydrodynamic approach , we were able to realize that the hydrodynamic excitations can propagate as particles and that the charge related to vortex `` atoms '' gives a measurable electric current .
the author is thankful to prof .
y. bruynseraede and prof .
j. indekeu for the hospitality and the interest in this investigation .
the author is very much indebted to i. bozovic , j. clayhold , v. geshkenbein , d. i. khomskii , a. i. larkin , y. matsuda , m. simkin , k. temst and a. volodin for the correspondence , papers provided , comments and the clarifying discussions .
it is a pleasure to thank e. penev for critically reading the manuscript , providing the electronic versions of the figures and a many suggested improvements .
the author appreciates the collaboration with m. mishonov in deriving the results in subsection [ typei ] and the indispensable help of v. mishonova .
this paper was supported by the belgian dwtc , the flemish government programme vis/97/01 , the iuap and the goa .
99 t. nagaoka , y. matsuda , h. obara , a. sawa , t. terashima , i. chong , m. takano , m. suzuki , `` hall anomaly in the superconducting state og high-@xmath0 cuprates : universality in doping dependence '' , _ phys .
lett . _ * 80 * , pp . 35943597 , 1998 . g. danna , v. berseth , l. forr , a. erb , and e. walker , `` hall anomaly and vortex - lattice melting in superconducting single crystal yba@xmath16cu@xmath152o@xmath199 , '' _ phys .
lett . _ * 81 * , pp .
25302533 , 1998 .
d. i. khomskii and a. freimuth , `` charged vortices in high temperature superconductors , '' _ phys .
* 75 * , pp . 13841386 , 1995 .
g. blatter , m. feigelman , v. geshkenbein , a. larkin , and a. van otterlo , `` electrostatics of vortices in type - ii superconductors , '' _ phys .
lett . _ * 77 * , pp . 566569 , 1996 .
n. hayashi , m. ichioka , and k. machida , `` relation between vortex core charge and vortex bound states , ''
_ j. phys . soc .
of japan _ * 67 * , pp . 33683371 , 1998 . y. kato , `` charging effect on hall conductivity of single vortex in type ii superconductors,''_j .
soc . of japan
_ * 68 * , pp . 37983801 , ( 1999 ) .
i. bozovic and j. n. eckstein , `` atomic - level engineering of cuprates and manganites , '' _ applied surface science _ *
113**114 * , pp .
189197 , 1997 ; i. bozovic , j. n. eckstein , and g. f. virshup , `` superconducting oxide multilayers and superlattices : physics , chemistry , and nano - engineering , '' _ physica c _ * 235**240 * , pp .
178181 , 1994 ; i. bozovic and j. n. eckstein , `` superconducting superlattices '' in _ physical properties of superconductors _ vol . 5 , d. m. ginsberg , ed .
, pp . 99207 , world scientific , singapore , 1996 . j. corson _ et al .
_ , `` vanishing of phase coherence in underdoped bi@xmath16sr@xmath16cacu@xmath16o@xmath200 , '' letters to nature , _ nature _ * 398 * , pp .
221 - 223 , 1999 [ see also j. millis , `` research enters a new phase '' , news and views , _ nature _ * 398 * , pp . 193194 , 1999 ] ; r. mallozi , j. orenstein , j. n. eckstein , and i. bozovic , `` high - frequency electrodynamics of bi@xmath16sr@xmath16cacu@xmath201o@xmath202 : nonlinear response in the vortex state , '' _ phys . rev .
letters _ * 81 * , pp . 1485 - 1488 , 1998 ; f. ronning , c. kim , d. l. feng , d. s. marshall , a. g. loeser , l. l. miller , j. n. eckstein , i. bozovic , and z .- x .
shen photoemission evidence for a remnant fermi surface and a d - wave - like dispersion in insulating ca@xmath16cuo@xmath16cl@xmath16 ,
_ science _ * 282 * , pp . 20672072 , 1998 . t. mishonov , `` theory of cooper - pair mass spectroscopy by current induced contact potential difference , '' _ phys .
b _ * 50 * , pp .
4009 - 4012 , 1994 . t. mishonov and n. zahariev , `` predicted new electrodynamic effect for thin films of layered cuprates
surface hall currents induced by the electric induction @xmath203 , '' _ superlattices and microstructures _ * 26 * , pp . 5760 , 1999 . t. mishonov , `` comment on ` metallic and superconducting surfaces of yba@xmath16cu@xmath152o@xmath204 probed by electrostatic charge modulation of epitaxial films ' , '' _ phys .
lett . _ * 67 * , p. 3195
r. a. mckee , f. j. walker , and m. f. chisholm , `` crystalline oxides on silicon : the first five monolayers , '' _ phys .
lett . _ * 81 * , pp . 30143017 , 1998 .
a. n. omelyanchuk and s. i. beloborodko , `` bernoulli effect and contact potential difference in superconductors , '' _ sov . j. low temp
. phys . _ * 9 * , pp .
572574 , 1983 [ _ fiz .
nizk . temp _
* 9 * , pp . 11051108 , 1983 ] a. t. fiory , a. f. hebard , p. m. mankievich , and r. e. howard , `` renormalization of the mean - field superconducting penetration dept in epitaxial yba@xmath16cu@xmath152o@xmath204 films , '' _ phys .
_ * 61 * , pp .
14191422 , 1998 . c. rogers , k. luther - myers , j. n. eckstein , and i. bozovic , `` brownian motion of vortex - antivortex excitations in very thin films of bi@xmath16sr@xmath16cacu@xmath16o@xmath153 , '' _ phys
letters _ * 69 * , pp . 160163 , 1992 .
m. greiter , f. wilczek , and e. witten , _ mod .
_ b * 3 * , 903 , 1989 .
m. v. simkin , `` bernoulli effect in superconductors and cooper - pair mass spectroscopy , '' _ condensed matter _
, cond - mat/9605036 , 5 kb , 6 may 1996 . h. w. lewis , `` search for the hall effect in a superconductor .
i. experiment ii .
theory , '' _ phys .
rev . _ * 92 * , p. 11491954
, 1953 ; _ ibid . _ * 100 * , p. 641645
f. london , _ suprefluids _ , john .
wiley and sons .
, new york , 1950 , p. 70 .
a. t. fiory , a. f. hebard , r. h. eick , p. m. mankiewich , r. e. howard , and m. l. omalley,``metallic and superconducting surfaces of yba@xmath16cu@xmath152o@xmath204 probed by electrostatic charge modulation of epitaxial films , '' _ phys
_ * 65 * , pp . 3441 - 3444 , 1990 ; erratum , _ ibid . _
* 66 * , 845 ( 1991 ) ; reply to comment , _ ibid .
_ * 67 * , p. 3196
sir n. f. mott , private communication , 1982 . | a novel experiment for determination of the charge related to vortices in thin superconducting film is proposed and a number of related experimental set - ups are also theoretically considered .
the methods are based on the torricelli - bernoulli effect in superconductors and the phenomenology of the effect is briefly discussed .
the vortex charge is expressed via the effective mass of the cooper pairs , thus both parameters , inaccessible by now , could be simultaneously determined .
the experiment would require layered metal - insulator - superconductor structures and standard electronics employed in kinetic measurements .
the quality of the insulator - superconductor interface should be high enough as to allow for observation of electric field effects similar to those investigated in superconducting field - effect transistors .
the development of layer - by - layer growth technology of oxide superconductors provides unique possibility for investigation of new fundamental effects in these materials .
in particular , the structures necessary for determination of the vortex charge could be used to study the superconducting surface hall effect , bernoulli effect , the superfluid density , etc .. in conclusion , the systematic investigation of new effects in oxide superconductors is envisaged as an important part of the material science underlying the oxide electronics . |
in order to extract properties of halo wimps ( weakly interacting massive particles ) by using data from direct dark matter detection experiments as model independently as possible , we have developed a series of data analysis method for reconstructing the one dimensional wimp velocity distribution function ( drees shan @xcite ) as well as determining the wimp mass ( drees shan @xcite ) , the spin independent ( si ) wimp coupling on nucleons ( shan @xcite ) and the ratios between different wimp couplings / cross sections ( shan @xcite ) .
moreover , in collaboration with the damned ( dark matter network exclusion diagram ) dark matter online tool ( @xcite ) , part of the ilias project ( @xcite ) , the `` '' ( a model independent data analysis system ) website for online simulation / data analysis has also been established ( @xcite ; shan @xcite ) . in this article , in order to demonstrate the usefulness and powerfulness as well as the model independence of the package for direct dark matter detection experiments , i will analyze _ blindly _ some _ pseudodata _ sets generated for different detector materials and present the reconstructed wimp properties .
this means that i will simply upload these data sets onto the website and follow the instructions to reconstruct different wimp properties _ without _ using any information about the input setup used for generating the pseudodata . for cases in which some information about wimps ( e.g. , the mass @xmath2 ) and/or galactic halo ( e.g. , the local dark matter density @xmath3 ) is required
, i will naively use the commonly used / favorite values for the data analyses .
after that i show the blindly reconstructed properties of halo wimps in sec . 2 , in sec .
3 i will reveal the input setup used for generating the analyzed data and compare the reconstructed results to them .
finally , i conclude in sec . 4 .
in this section , i present the reconstructed wimp properties analyzed by the website .
while in each uploaded file there are _ exactly _ 50 data sets , in each data set there are _ on average _ 50 recorded events ( i.e. , 50 measured recoil energies ) ; the exact number of total events is poisson distributed . for simplicity ,
the experimental minimal and maximal cut off
energies have been set as 0 and 100 kev for all data sets . in order to check the effect of using a `` wrong '' elastic nuclear form factor ,
two forms have been considered for the si wimp nucleus cross section in our analyses .
one is the simple exponential form : f_ex^2(q ) = e^-q / q_0 .
[ eqn : fq_ex ] here @xmath4 is the recoil energy transferred from the incident wimp to the target nucleus , @xmath5 is the nuclear coherence energy given by @xmath6 , where @xmath7 is the radius of the nucleus and @xmath8 is the mass of the target nucleus .
meanwhile , we used also a more realistic analytic form for the elastic nuclear form factor : f_si^2(q ) = ^2 e^-(q s)^2 .
[ eqn : fq_ws ] here @xmath9 is a spherical bessel function , @xmath10 is the transferred 3-momentum , for the effective nuclear radius we use @xmath11 with @xmath12 and a nuclear skin thickness @xmath13 . for the sd wimp
nucleus cross section , we only used the `` thin
shell '' nuclear form factor : f_ts^2(q ) = [ eqn : fq_ts ] as one of the most important properties of halo wimps as well as the basic information for reconstructing other quantities in our model
independent analysis methods , i consider at first the determination of the wimp mass @xmath2 by means of the method introduced in drees shan ( @xcite ) . in figs .
[ fig : mchi ] i show the reconstructed wimp masses and the upper and lower bounds of their 1@xmath15 statistical uncertainties .
the usual target combination of nuclei has been used for this reconstruction , whereas two forms of the elastic nuclear form factor given in eqs .
( [ eqn : fq_ex ] ) and ( [ eqn : fq_ws ] ) have been used for determining @xmath2 in the upper and lower frames , respectively . while @xmath16 with @xmath17 , 1 , 2 and @xmath18 have been estimated by eqs .
( 34 ) and ( 40 ) of drees shan ( @xcite ) , respectively , @xmath19 has been estimated by the @xmath20fitting defined in eq .
( 51 ) of drees shan ( @xcite ) , which combines the estimators for @xmath16 and @xmath18 with each other .
the reconstructed wimp mass @xmath19 as well as @xmath16 and @xmath18 shown here have been corrected by the iterative @xmath21matching procedure described in drees shan ( @xcite ) .
it can be found here that , although all single estimators ( @xmath16 with @xmath17 , 1 , 2 and @xmath18 ) give generally a ( relatively lighter ) wimp mass of or even lighter and a 1@xmath15 upper bound of @xmath22 gev , the _ mean _ values of the combined ( in principle , more reliable ) results ( the second column in two tables ) of the reconstructed wimp mass give @xmath23 gev with a _ rough _ 1@xmath15 upper ( lower ) bound of @xmath24 ( 80 ) gev , or , equivalently , 120_-40^+70 gev .
[ eqn : mchi_rec ] moreover , the combined results with two different form factors show not only a large overlap between @xmath25 gev and @xmath26 gev , but also a good coincidence : comparing to the @xmath27 gev 1@xmath15 statistical uncertainty and the @xmath28 gev overlap , the difference between two median values is @xmath29 gev ! this indicates that , for the first approximation of giving / constraining the most plausible range of the wimp mass , the uncertainty on the nuclear form factor could be safely neglected . following the wimp mass determination
, i consider now the reconstruction of the si wimp coupling on nucleons @xmath32 ( shan @xcite ) with a @xmath33 target has been adopted in the package . ] . in figs .
[ fig : fp2 ] i show the reconstructed squared si wimp - nucleon couplings and the lower and upper bounds of their 1@xmath15 statistical uncertainties estimated by eqs .
( 17 ) and ( 18 ) of shan ( @xcite ) with an _ assumed _ ( 100@xmath3410 gev , labeled with the subscript `` input '' ) and the reconstructed ( from sec .
2.1 , labeled with `` recon '' ) wimp masses .
the commonly used value of the local dark matter density @xmath35 and a larger value of @xmath36 ( catena ullio @xcite ; salucci _ et al . _
@xcite ; pato _ et al . _
@xcite ) as well as the elastic nuclear form factors given in eqs .
( [ eqn : fq_ex ] ) and ( [ eqn : fq_ws ] ) have been used for estimating @xmath32 in the upper and lower frames , respectively . among these results ,
mean _ value and the _ overlap _ of two most plausible results ( estimated by using the reconstructed wimp mass ) give _ roughly _ ( and somehow _ naively _ ) a 1@xmath15 range of reconstructed wimp mass given in sec .
2.1 is @xmath23 gev , one can simply use the proton mass @xmath37 to approximate the wimp proton reduced mass @xmath38 and give a reconstructed si wimp nucleon cross section as ( eq .
( 17 ) of shan ( @xcite ) ) is a function of the ( reconstructed ) wimp mass , for light wimp mass , one has to use = ^2 , [ eqn : sigmapsi_mchi ] where @xmath39 is the atomic mass number of the target nucleus , @xmath40 is the experimental exposure
. then one has ( cf .
18 ) of shan ( @xcite ) ) & = & + & & ^1/2 .
[ eqn : sigma_sigmapsi_mchi ] here i have used ( drees shan @xcite ) _ m = ^-1 , [ eqn : calnm ] and defined ( ) = 2 . [
eqn : delta_mchi ] definitions and estimations of @xmath41 and @xmath42 can be found in e.g. , drees shan ( @xcite ) .
] = ^2 |f_p|^2 m_p^2 ~ |f_p|^2 4.31_-0.69^+1.01 10 ^ -9 pb . [ eqn : sigmapsi_rec ] in figs .
[ fig : ranap ] i show the reconstructed @xmath45 ratios and the lower and upper bounds of their 1@xmath15 statistical uncertainties estimated by eqs .
( 2.7 ) and ( 2.12 ) of shan ( @xcite ) with @xmath46 as well as by eqs . ( 3.16 ) and ( 3.20 ) of shan ( @xcite ) at the shifted energy points ( drees shan @xcite ; shan @xcite ) .
a combination of @xmath47 + @xmath48 targets has been used for the reconstruction of @xmath45 under the assumption that the sd wimp
nucleus interaction dominates over the si one ( labeled with the superscript `` sd '' ) , whereas a third target of @xmath49 has been combined with @xmath47 and @xmath48 for the case of the general combination of both si and sd wimp interactions ( labeled with the superscript `` si + sd '' ) .
it can be found that , firstly , the `` @xmath50 ( plus ) '' solutions of the @xmath45 ratios given here are obviously too large to be the reasonable choice for @xmath45 and the `` @xmath51 ( minus ) '' solutions should be the correct ones '' and `` @xmath51 '' solutions can be decided directly by the values of the group spins of protons and neutrons of the used target nuclei , @xmath52 . ] .
secondly , although the reconstructed result under the assumption of the sd dominant wimp interaction is in general _ larger _ than the ( in principle more plausible ) result obtained without such a prior assumption , one could still use the _ mean _ value and the _ overlap _ of these two results to _ roughly _ ( and somehow _ naively _ ) give a 1@xmath15 range of 0.89_-0.30^+0.26 .
[ eqn : ranap_rec ] in figs .
[ fig : rsigmasdsi ] i show the reconstructed @xmath54 ratios and the lower and upper bounds of their 1@xmath15 statistical uncertainties estimated by eqs .
( 3.9 ) , ( 3.10 ) and ( 3.21 ) of shan ( @xcite ) ( with @xmath45 estimated by eq . ( 3.16 ) of shan ( @xcite ) ) as well as by eqs . ( 3.25 ) and ( 3.29 ) of shan ( @xcite ) at the shifted energy points . by using the data sets of @xmath47 , @xmath48 and @xmath49 targets ( labeled with the superscript `` xyz '' ) or combining that of @xmath55 or @xmath56 with the ( common ) data set of @xmath33 ( labeled with the superscript `` xy '' )
, one can use the _ mean _ value and the _ overlap _ of these two results to _ roughly _ ( and somehow _ naively _ ) give a 1@xmath15 range of 9.61_-3.28^+2.55 10 ^ 5 , 5.45_-2.75^+1.56 10 ^ 5 .
[ eqn : rsigmasdsi_rec ] then , firstly , from these results one can further obtain that and neglected the correlation term in the bracket : -2 cov / .
, [ eqn : sigma_ranap_2_cov ] since the 1@xmath15 uncertainties given in eq .
( [ eqn : rsigmasdsi_rec ] ) are not the exact but only rough estimates from the overlaps of two results given in figs .
[ fig : rsigmasdsi].]@xmath57 are reconstructed with the @xmath45 ratio given in the last columns of the tables in figs . [ fig : ranap ] . ] 0.75_-0.23^+0.15 .
[ eqn : ranap_rec_2 ] secondly , combining the results in eq .
( [ eqn : rsigmasdsi_rec ] ) with @xmath58 given in eq .
( [ eqn : sigmapsi_rec ] ) , one can also obtain that and neglected the correlation term in the bracket : 2 cov / .
[ eqn : sigma_sigmapsd_cov ] by assuming that two independent data sets with the @xmath33 target and other two independent data sets with the @xmath49 target have been used for determining @xmath58 and @xmath59 . ]
4.14_-1.56^+1.47 10 ^ -3 pb , 2.35_-1.31^+0.87 10 ^ -3 pb .
[ eqn : rsigmasd_rec ] these results give in turn that and @xmath23 gev , one has |a_(p , n)| = , [ eqn : an / p ] and = .
[ eqn : sigma_an / p ] ] || 0.108_-0.020^+0.019 , || 0.081_-0.023^+0.015 .
[ eqn : a_rec ] on the other hand , one can also use the reconstructed @xmath45 ratio given in eq .
( [ eqn : ranap_rec ] ) and _ one _ of the two results given in eq .
( [ eqn : rsigmasd_rec ] ) to obtain that and neglected the correlation term in the bracket : 4 cov / . , [ eqn : sigma_sigmapsd_2_cov ] since the 1@xmath15 uncertainties given in eq .
( [ eqn : ranap_rec ] ) as well as in eq .
( [ eqn : rsigmasd_rec ] ) are not the exact but only rough estimates from the overlaps of the results given in figs .
[ fig : fp2 ] , [ fig : ranap ] and [ fig : rsigmasdsi ] .
the `` @xmath51 ( @xmath50 ) '' sign in eq .
( [ eqn : sigma_sigmapsd_2_cov ] ) is for the case with protons ( neutrons ) . ]
2.97_-2.60^+2.05 10 ^ -3 pb , 3.28_-2.53^+2.24 10 ^ -3 pb . [ eqn : rsigmasd_rec_2 ] these results can also give that || 0.091_-0.040^+0.032 , || 0.096_-0.037^+0.033 .
[ eqn : a_rec_2 ] it can be found that , not surprisingly , the statistical uncertainties on the reconstructed @xmath60 given in eq .
( [ eqn : rsigmasd_rec_2 ] ) are @xmath0 2 or 3 times larger than those given in eq .
( [ eqn : rsigmasd_rec ] ) : since @xmath54 reconstructed with the f + i + si combination involve already the reconstructed @xmath45 ratio given in eq .
( [ eqn : ranap_rec ] ) , the uncertainties on @xmath60 given in eq .
( [ eqn : rsigmasd_rec_2 ] ) are thus _
overestimated_. secondly , although the reconstructed @xmath61 and @xmath62 given in eqs .
( [ eqn : rsigmasd_rec ] ) and ( [ eqn : rsigmasd_rec_2 ] ) have overlaps , these results seem _ not _ to match to each other very well ; @xmath62 given in eq .
( [ eqn : rsigmasd_rec_2 ] ) is even larger than @xmath61 there although the @xmath45 ratios given in eqs .
( [ eqn : ranap_rec ] ) and ( [ eqn : ranap_rec_2 ] ) are @xmath63 1 .
one possible explanation is that the @xmath45 ratio given in eq .
( [ eqn : ranap_rec ] ) would be _
overestimated_. this can be seen by comparing the @xmath45 ratio given in eq .
( [ eqn : ranap_rec ] ) to that given in eq .
( [ eqn : ranap_rec_2 ] ) estimated ( somehow _ independently _ ) by the results given in eq .
( [ eqn : rsigmasdsi_rec ] ) .
nevertheless , the analyses given here show that , firstly , once one can estimate the si wimp nucleon coupling / cross section , @xmath64 or @xmath58 , and ( one of ) the ratios between the sd and si wimp nucleon cross sections , and/or the ratio between two sd wimp
nucleon couplings , the other couplings / cross sections could in principle be estimated . secondly , although the method under the assumption of the sd dominant wimp interaction would overestimate ( or underestimate , depending on the combination of the used targets ( shan @xcite ) ) the @xmath45 ratio , the reconstructed result(s ) could still be useful for at least determining the correct sign of @xmath45 .
moreover , the wimp couplings / cross sections estimated in different way would be self cross checks to each other and the ( in)compatibility between the reconstructed results would also help us to check the usefulness of the analyzed data sets offered from different experiments with different detector materials .
in table [ tab : setup ] i give finally the input setup for generating the pseudodata sets used in the analyses demonstrated in the previous section . for comparison ,
the reconstructed results shown in the previous section are also summarized here .
it can be found that , firstly , not only the wimp mass given in eq .
( [ eqn : mchi_rec ] ) and the result reconstructed with the input nuclear form factor ( lower frame of figs .
[ fig : mchi ] ) , but even the mass reconstructed with the `` wrong '' form factor ( upper frame ) can match the input wimp mass very well : the deviations between the input and the reconstructed values are only @xmath0 13% ( with the wrong nuclear form factor ) or even only @xmath0 6% ( with the input one ) . as discussed earlier , this indicates that , for the first approximation of giving / constraining the most plausible range of the wimp mass , the uncertainty on the nuclear form factor could be safely neglected . .
the input setup for generating the pseudodata sets used in the analyses demonstrated in this article .
the theoretically estimated values and the reconstructed results shown in the previous section are also given .
@xmath65 : estimated for 130 gev @xmath2 .
[ cols="^,^,^ , < " , ] secondly , all wimp nucleon couplings / cross sections as well as the ratios between them have also been reconstructed with only @xmath0 5% to @xmath1 40% deviations from the input / theoretically estimated values . although the si wimp coupling @xmath64 estimated with the input ( larger ) local dark matter density ( lower frame of figs .
[ fig : fp2 ] ) is _ underestimated _ ( shan @xcite ) , one can at least give an upper bound on @xmath64 . meanwhile , although the @xmath45 ratio given in eq .
( [ eqn : ranap_rec ] ) is overestimated , in sec .
2.4 we have demonstrated that by combining different methods for estimating different ( ratios between the ) wimp couplings / cross sections , one could in principle observe / confirm the ( in)compatibility between these results and probably correct the reconstructed values .
moreover , for _ generating _ pseudodata , we have used the shifted maxwellian velocity distribution : f_1 , ( v ) = , [ eqn : f1v_sh ] with the sun s galactic orbital velocity @xmath66 km / s ; @xmath67 is the _ time dependent _
earth s velocity in the galactic frame : ( t ) = v_0 , [ eqn : ve ] the date on which the earth s velocity relative to the wimp halo is maximal has been set as @xmath68 d. although these values for the astronomical setup are _ non standard _ , we would like to stress that , firstly , for using the package and website to analyze ( real ) data sets , one needs only the form factors for si and/or sd wimp
nucleaus cross sections , prior knowledge / assumptions about the wimp velocity distribution @xmath69 and local density @xmath3 ( except the estimation of the si wimp nucleon coupling @xmath32 ) are _ not required_. secondly , as shown in the previous section , such non standard values would not affect the reconstructed results .
in this article i demonstrated the data analysis procedures for extrating wimp properties by using theoretically generated pseudodata for different target nuclei .
as an extension as well as the complementarity of our earlier theoretical works , i combined reconstructed results of the ( ratios between different ) wimp couplings / cross sections on nucleons to estimate each _ individual _ coupling / cross section . hopefully , the package and website as well as this demonstration can help our experimental colleagues to analyze their _ real _ direct detection data in the near future and to determine ( at least rough ranges of ) properties of halo dark matter particles .
the author appreciates the ilias project and the physikalisches
institut der universitt tbingen for kindly providing the opportunity of the collaboration and the technical support of the website .
the author would also like to thank the friendly hospitality of the national institute for nuclear and high energy physics ( nikhef ) where part of this work was completed .
this work was partially supported by the national science council of r.o.c . under contract no .
nsc-99 - 2811-m-006 - 031 as well as by the national center of theoretical sciences ( south ) , r.o.c .. | in this talk i have presented the data analysis results of extracting properties of halo wimps : the mass and the ( ratios between the ) spin independent and spin
dependent couplings / cross sections on nucleons by the website .
although _ non standard _ astronomical setup has been used to generate _ pseudodata _ sets for our analyses , it has been found that , _ without _ prior information / assumption about the local density and velocity distribution of halo dark matter , these wimp properties have been reconstructed with @xmath0 5% to @xmath1 40% deviations from the input values . |
the aim of this talk is to review some of the atomic precision measurements in atomic physics leading to precious informations in the realm of high - energy physics .
the idea of atomic physics bringing light on the high - energy physics world requires some qualification due to the very different scales of energy involved in the two cases .
in fact , typically one has a separation of about six or seven order of magnitude between the two scales and one expects the two physics being almost decoupled . in fact , if we look at some observable , @xmath0 , at a scale @xmath1 , we expect that the observable can be represented in the form @xmath2 in order to be able to derive informations about the physics at the scale @xmath3 , being at the scale @xmath4 , one starts considering a combination of observables corresponding to the corrections coming from the higher scale @xmath5 in order to measure @xmath6 one needs either the coefficient @xmath7 being very large in such a way to partially compensate the scale factor , or having an extremely good experimental sensitivity . in this talk
i will consider two particular examples of situations where atomic physics can be relevant to high - energy physics , namely atomic violation of parity ( apv ) and possible violations of the discrete symmetry cpt , that is the product of charge - conjugation , parity and time - reversal .
in fact , using heavy atoms like cesium in apv measurements one can get a good enhancement factor . on the other hand
, cpt symmetry can be tested by using the extraordinary opportunities offered by the atomic traps in order to obtain a very accurate determination of frequencies .
in this section i will discuss mainly the latest determination of the weak charge in atomic cesium and some of its implications in models of physics beyond the standard model ( sm ) .
the sm has been tested very precisely at machines such as lep and slc , where , working at an energy around the @xmath8 mass , one is mainly testing the property of the @xmath8 itself . therefore , the physics beyond the sm that can be looked for at these machines is the one giving corrections to the @xmath8-propagator and/or to the couplings of the @xmath8 with fermion - antifermion pairs .
namely , new massive vector bosons , @xmath9 , which mix to the @xmath8 , or new particles running in loops and contributing to the @xmath8 self - energy or to vertex corrections . but consider , for instance , the case of a massive vector boson which does not mix to the @xmath8 , and therefore invisible at lep ( except for tiny radiative corrections ) . if the @xmath9 is coupled to fermions , in the low - energy limit it gives rise to an effective four - fermi interaction .
therefore , low - energy experiments are complementary to the high - energy ones , and furthermore they are able to measure directly the couplings of the @xmath8 to light quarks ; something that at lep and slc can be done only in an indirect way . among the low - energy experiments a particular role is played by the apv experiments , due to the precision almost at the level of the one reached at lep / slc .
let us now recall some feature of apv in atoms .
first of all , within the sm the four - fermi parity violating hamiltonian density for nucleons is given by @xmath10\ ] ] where c_ip=-2c_iu - c_id , c_in =- c_iu-2c_id , i=1,2and c_1q=-8a_ev_q=-(t_3^q-2s_^2q^q ) , c_2q=-8v_ea_q= -t_3^q(1 - 4s_^2 ) , q = u , dhere @xmath11 , @xmath12 , @xmath13 and @xmath14 are the vector and vector - axial couplings of the @xmath8 to the electrons and quarks .
for a point - like nucleus with @xmath8 protons and @xmath15 neutrons , the hamiltonian density , in the non - relativistic limit , is given by _
pv&=&where @xmath16 is the momentum of the electron , @xmath17 the total spin of the protons ( neutrons ) and @xmath18 the electron mass .
i have also defined the
_ weak charge _ of the atom asq_w(z , n)=2notice that for a heavy atom ( large values of @xmath8 ) the matrix element of the first term in @xmath19 is roughly proportional to @xmath20 , one factor coming from @xmath21 , one from the momentum of the electron and the third one from the wave function evaluated at the origin .
this coherence effect was noticed by bouchiat and bouchiat @xcite and it provides , in the case of cesium ( @xmath22 ) an enhancement factor of about @xmath23 , more or less what is necessary in order to compensate for the decoupling factor from the scales mentioned in the introduction . in order to get a rough idea of the bounds on new physics that can be obtained by a measurement of @xmath21 with a given sensitivity , we parametrize the new physics contribution to @xmath21 by a four - fermi effective interaction @xcite_np^pv= |e _ _ 5 e_q = u , dh_1q|q ^qif we assume @xmath24 , for a sensitivity @xmath25 one gets a bound ( 5 g_np ) tevif new physics is strongly interacting ( @xmath26 ) , then @xmath27 , whereas in the weakly interacting case ( @xmath28 ) we get @xmath29 . in any case
we see that at 1% level of sensitivity , @xmath21 is able to test new physics for scales greater than 1 @xmath30 . in apv measurements
one looks at optical transitions between a pair of states @xmath31 mixed by @xmath19 and a state @xmath32 of the same nominal parity as @xmath33 .
the mixing of the two eigenstates of parity is given by = where @xmath34 is the splitting between the two levels .
if i denote by @xmath35 and @xmath36 the amplitudes for the two unperturbed transitions @xmath37 and @xmath38 , the transition probability , after the mixing , is given by w = m_1 ^ 2+|e_1^pv|^22 im(e_1^pv)m_1the choice of the sign depends on the helicity of the photon which is emitted or absorbed in the transition . in the actual experiment on cesium
one measures the _ circular dichroism _
, that is the asymmetry for the absorption cross - section = 2 of course , the pv amplitude @xmath36 is proportional to the mixing parameter @xmath39 and therefore measuring @xmath40 one can get the matrix element of the pv hamiltonian .
these ideas have been applied in particular to the transition @xmath41 in atomic cesium @xmath42 @xcite , but also to other atoms as thallium @xcite .
the typical value of @xmath40 is @xmath43 , but there is a strong background which can be overcomed by letting the pv amplitude to interfere with a large electro - induced ( stark ) transition .
eventually one extracts from the experiment the matrix element of @xmath44 which is proportional to @xmath21 times an atomic form factor @xmath45 which must be evaluated theoretically in order to extract the value of the weak charge .
therefore the measurement must be coupled with theoretical calculations of similar accuracy in order to get a precise determination of @xmath21 . in the case of atomic cesium the calculation of @xmath45
was performed independently by two groups @xcite .
this calculation is not an easy task , as one has to use many - body perturbation theory coupled with hartree - fock techniques .
the theoretical errors are quite difficult to estimate .
the authors of refs .
@xcite did their estimate by looking at the differences between the theoretical and the experimental values of parity conserving quantities as dipole matrix elements and hyperfine splittings for the @xmath46 , @xmath47 , @xmath48 and @xmath49 states . in this way
the error @xmath50 was obtained .
after the new measurement of the weak charge of the cesium by the boulder group @xcite , which improved the accuracy of the previous experiment @xcite by more than a factor five , bennett and wieman @xcite re - examined the theoretical errors on @xmath45 .
in fact , since the time of the previous estimate there have been a number of new and more precise measurements of the quantities of interest .
the result is that now the agreement is much better than before , and as a consequence bennett and wieman got the estimate @xmath51 .
it should be noticed that there is a third element which contributes to the extraction of @xmath21 from the data .
this is the stark mixing - induced electric dipole moment amplitude , @xmath52 . the experiments in refs .
@xcite were using a theoretical determination of @xmath52 .
in @xcite the ratio @xmath53 has been measured .
the off - diagonal magnetic dipole moment induced by the hyperfine interaction is well known empirically and it is possible to extract a precise value for @xmath52 . however , in a contribution to this conference @xcite , the matrix element @xmath54 has been accurately calculated with the result that the empirical formula for it should be corrected by a factor of 0.24% increasing the discrepancy with the sm ( see later ) .
i would like also to comment about some possible neglected contribution in the evaluation of the atomic form factor .
it has been pointed out in ref .
@xcite that there could be a contribution arising from the difference of neutron and proton spatial distributions inside the nucleus .
this contribution turns out to be very difficult to estimate , in fact it is quite model dependent .
most probably it could introduce a further error on @xmath55 of about 0.3 . this would not change the conclusions in a very significant way .
another point has been raised recently in ref .
this author argues that the contribution from the breit interaction ( exchange of a transverse photon between two electrons ) could have been underestimated .
the breit interaction contribution to the atomic form factor was estimated in @xcite and it was found to be very small . however in ref . @xcite it is found that the total effect , taking into account also second and third order contributions , is about twice the first order effect . as a consequence ,
if `` all '' the higher order contributions could be shown to be negligible , the experimental measure would reconcile with the sm expectation for @xmath55 . however , see also ref .
@xcite . to conclude this analysis
i think that an evaluation of the atomic form factor by taking into account the next order in the many - body perturbative theory is highly desirable in order to settle the question . in any case
i find of some interest to that the theoretical error is indeed at the level of @xmath56 in order to see which are the possible implications of the apv in high - energy physics .
i can start now to discuss the experimental results on @xmath55 .
it is interesting to recall the value obtained in @xcite combined with the theoretical determination of @xmath45 @xcite q_w(cs)=-71.04(1.58)_exp(0.88)_ththe total error of these measurement on @xmath55 is at 2.5% level of accuracy that , at that time , was comparable with the sensitivity obtained at lep1 .
in fact , this determination of @xmath55 lead to the first indication that technicolor models , in their most simple version obtained from scaling of qcd , could not possibly fit the data . the new experimental result on @xmath55 @xcite combined with the new determination of the theoretical error @xcite gives q_w(cs)^exp = -72.06(0.28)_exp(0.34)_th[qwexp]a result at 0.6% level of accuracy .
on the theoretical side , @xmath21 can be expressed as @xcite q_w(cs)^th=-72.720.13 - 102_3^rad+_nq_w [ qw]including hadronic - loop uncertainty .
i use here the variables @xmath57 ( i=1,2,3 ) of ref .
@xcite , which include the radiative corrections , in place of the set of variables @xmath58 , @xmath59 and @xmath60 originally introduced in ref .
@xcite . in the above definition of @xmath61
i have explicitly included only the standard model ( sm ) contribution to the radiative corrections . new physics ( that is physics beyond the sm ) contributions are represented by the term @xmath62 .
also , i have neglected a correction proportional to @xmath63 .
in fact , as well known @xcite , due to the particular values of the number of neutrons ( @xmath64 ) and protons ( @xmath22 ) in cesium , the dependence on @xmath65 almost cancels out .
for a top mass of 175 @xmath66 and @xmath67 the value of @xmath68 is given by @xcite_3^rad=5.110(6.115)10 ^ -3for @xmath69 , corresponding roughly to the lower experimental bound from direct search at lep2 @xcite , one gets q_w(cs)^exp - q_w(cs)^sm=1.180.46 [ bounds]giving rise to a deviation of about @xmath70 .
furthermore , for increasing mass of the higgs the discrepancy increases . therefore ,
if we assume as being correct the experimental result , the theoretical evaluation of @xmath45 and the evaluation of the theoretical errors , we are forced to conclude that the .
we can draw another conclusion , that is , that in order to explain the data on @xmath55 we need new physics not constrained by the lep and slc data . in fact , as an example let me consider a type of new physics visible at lep as , for instance , contributing to the self - energy of the @xmath8 , the so called oblique corrections .
in such a case one can write @xmath71 , and in order to compensate for the discrepancy on @xmath55 one needs _ 3n=(-11.64.5)10 ^ -3whereas from lep and slc data one can determine the sum _
3^exp=_3^rad+_3n=(4.191)10 ^ -3therefore one gets @xmath72 , one order of magnitude too small to explain the data on @xmath55 .
i would like also recall the experimental result of apv on thallium @xcite q_w(tl)^exp=-114.8(1.2 ) _
exp(3.4)_ththis result is not as precise as the one on @xmath73 , and in fact the total error is about 3% . at this level
it is perfectly compatible with the sm prediction q_w(tl)^sm=-116.70.1 a new experiment on cesium is being planned in paris but the experimental sensitivity is going to be lower than the one obtained in boulder .
in berkeley and seattle there are plans for isotope ratio measurements . in this case
the dependence on the atomic form factor would go away eliminating the theoretical error .
however these ratios depend on the variation of the neutron density along the isotope chain .
this would introduce errors at least twice as big as the experimental ones @xcite .
we are now in the position of discussing the implications of eq .
( [ bounds ] ) on new physics . assuming that the contribution of new physics , @xmath74 , is such to reproduce the experimental results
, we can make use of eqs .
( [ qwexp ] ) and ( [ qw ] ) to write @xciteq_w(cs)^exp - q_w(cs)^th(m_h)=0.66 + 102_3^rad(m_h)-_nq_w0.46for @xmath69 at a 95% cl we find 0.28_nq_w2.08notice that the lower positive bound arises since the sm ( corresponding to @xmath75 ) does not fit the experimental value of @xmath55 at this cl value .
this is quite important since it implies an on the scale of new physics .
for the same reason new physics with a contribution @xmath76 is not allowed .
also notice that lower and upper bounds both increase for increasing higgs mass .
[ [ contact - interactions - from - compositness . ] ] contact interactions from compositness .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + a typical four - fermi operator in composite models contributing to the pv lagrangian is @xcite |e _ 2 e|q^2q [ contact]the effect of this interaction is to modify the coefficients @xmath77 c_1u,1dc_1u,1dwhere , since composite models correspond to strongly interacting new physics , we have assumed @xmath78 . from q_w=-2[(2z+n)c_1u+(z+2n)c_1d]we
see that the negative sign for the operator ( [ contact ] ) is excluded . for the positive sign
we get the bounds 12.1(tev)32.9the typical lower bound from high energy physics is about 3.5 @xmath30 @xcite . [
[ extra - dimension - models . ] ] extra - dimension models .
+ + + + + + + + + + + + + + + + + + + + + + + in ref .
@xcite a minimal extension to higher dimensions of the sm , with extra dimensions compactified , was considered . in this model
the fermions live in a 4-dimensional subspace , the wall , whereas the gauge bosons live in the full d - dimensional space , the bulk .
in general , there might be two higgs fields , one living in the bulk , @xmath79 , and the other living on the wall , @xmath80 .
the propagation of the gauge fields in the bulk is equivalent to the exchange of an infinite tower of kaluza - klein ( kk ) excitations with increasing mass .
for example , for @xmath81 , @xmath82 , @xmath83 , with @xmath84 the compactification radius . if only the higgs field @xmath80 is present , the ordinary gauge bosons do not mix with the kk resonances and it is easy to see that the contribution of these modes to @xmath21 is negative @xcite .
therefore the model does not fit the data on @xmath55 . for the more general case of both higgs fields present it has been shown @xcite that the lep / slc and @xmath55 experimental data are not compatible among them at 95% cl .
[ [ extra - z - models . ] ] extra @xmath9 models .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the implications of models with an extra neutral vector boson @xmath9 for apv have been considered in the literature for quite a long time @xcite .
the @xmath9 has couplings comparable to the ones of the @xmath8 in the sm and therefore this is an example of weakly interacting new physics .
there is a continuum of such models characterized by an angle @xmath85 . to any value of @xmath86
it corresponds a different model .
the 95% cl regions allowed by @xmath21 , in the plane @xmath87 , for different values of the higgs mass , are shown in fig .
[ zetap ] . in deriving these figures
the assumption of zero mixing between @xmath8 and @xmath9 has been made . in the figure
are also shown three popular models : @xmath39 ( @xmath88 ) , @xmath89 ( @xmath90 ) , @xmath91 ( @xmath92 ) .
we see that the @xmath39 and the @xmath91 models are not allowed by the data .
the direct search at the tevatron for a @xmath9 within the @xmath89 model gives a direct lower bound at 95% cl , @xmath93 ( a similar bound holds for all these models ) .
therefore this model is compatible with the data .
a recent best fit to all the data ( including apv ) gives for the @xmath89 model the following results @xcite , @xmath94 and a mixing angle compatible with zero , @xmath95 .
the cpt theorem is one of the fundamental results in local relativistic field theories .
therefore the idea of possible violations of this theorem implies that some of the axioms of these theories should be reviewed .
let me recall here the exact statement of the theorem @xcite : _ in a field theory satisfying _ 1 .
locality 2 .
lorentz invariance 3 .
analiticity of the lorentz group representations in the boost parameters the cpt transformation is a symmetry of the theory itself .
the first two conditions say that one is dealing with a local relativistic field theory , whereas the third one is satisfied in any finite - dimensional representation of the lorentz group .
it is interesting to notice that unitary representations fail to be analytic and as a consequence the cpt theorem can be violated in this case .
the first example of this situation dates back to majorana @xcite when he formulated a first order wave equation without negative - energy solutions .
he was able to do that by making use of a unitary infinite - dimensional representation of the lorentz group .
since this theory does not contain antiparticles the cpt symmetry is broken .
however , the quarks and leptons described by the sm belong to finite - dimensional representation of the lorentz group and therefore this does not seem a possible way to break the theorem .
it seems also very hard to give up locality , since it guarantees the microcausality of the theory .
therefore , the only sensible way to avoid the consequences of the cpt theorem in a local field theory seems to break lorentz invariance .
a situation of this type could arise at a more fundamental level as in string theory , where it is possible that lorentz invariance is spontaneously broken around the planck mass , @xmath96 @xcite .
one can take into account these effects by writing down a local effective lagrangian with lorentz and cpt breaking terms .
these terms can be written as an expansion in derivatives over the planck mass .
for instance , considering a single fermion , the violating term can be written as _
v=_n t|(i)^ni have used a somewhat symbolic notation where @xmath97 stays for a generic combination of dirac matrices and @xmath59 is a constant tensor and i take the mass dimensions of @xmath98 as @xmath99=1 $ ] .
furthermore i will assume the same internal symmetries as in the sm , that is @xmath100 @xcite . since the breaking terms should vanish in the limit @xmath101 also for @xmath102
, i will require g_0=c_owhere @xmath103 is some low - energy mass scale parameter .
we see that the relevant terms are the ones with @xmath102 and @xmath104 , and therefore the resulting theory preserves the renormalizability property .
let me now consider a single fermion interacting with the electromagnetic field .
one adds to the standard qed lagrangian the following two terms _ v^(n=0)=|[l0]and _
v^(n=1)=|[l1]where @xmath105 , with @xmath106 the electric charge of the fermion .
there are other possible terms with @xmath104 , but they are not compatible with the symmetries of the sm and therefore they should be suppressed . the following orders of magnitude are expected a _ , b _ , h_(m^2/m_p ) ,
c _ , d_(m / m_p)[tensors]the terms in @xmath107 violate lorentz invariance , since all the tensors in eq .
( [ tensors ] ) are constant ones .
however only the terms proportional to @xmath108 and @xmath109 violate cpt symmetry since @xmath110 , @xmath111 and @xmath112 are cpt odd , whereas the other covariant terms are cpt even .
therefore , in the following i will take into consideration only @xmath113
. notice also that when dealing with a single fermion the term in @xmath108 does not have physical meaning since we can write @xmath114 , showing that @xmath108 is a trivial gauge background field . of course , the situation changes when dealing with different fermions having different @xmath108 s . from eq .
( [ tensors ] ) we expect that the order of magnitude of the cpt and lorentz breaking terms is given by @xmath115 for @xmath116 , where @xmath18 is the electron mass and @xmath117 is the electroweak symmetry breaking scale
. lorentz and cpt breaking terms could appear also in the photon part of the total lagrangian .
this instance is discussed thoroughly in the second paper of ref .
@xcite , but i will not consider it in this talk . here
i want to illustrate some atomic physics experiment about cpt violation . but
before doing that let me just give a list of other existing or planned experiments about the violation of this fundamental symmetry * @xmath118 mass difference .
this experiment gives the best high - energy result @xcite 10 ^ -18[meson ] * experiments on neutral meson oscillations to be done at meson factories @xcite .
* experiments on muons @xcite .
* experiments with spin - polarized solids @xcite .
* experiments from clock - comparison @xcite . cpt violation may have also some relevance for baryogenesis and this subject has been discussed in ref .
@xcite .
let me now consider atomic physics experiments for testing cpt using atomic traps .
several of these experiments have been performed by confining single particles or antiparticles in a penning trap for a long time .
these experiments have a very high precision , of order @xmath119 or better , whereas the precision in experiments about mesons ( see eq .
( [ meson ] ) ) is much lower , of order @xmath120 .
i recall here the comparison of the electron and positron gyromagnetic ratios , @xmath121 , obtained measuring their cyclotron and anomaly frequencies ( see later ) , which gives the figure of merit @xcite||210 ^ -12measuring the proton and antiproton cyclotron frequencies , one can get their charge - to - mass ratios .
@xmath122 @xcite ||910 ^ -11analogously , from the charge - to - mass ratio for electron and positron @xcite||1.310 ^ -7as we see the relevant figures of merit are much bigger than the one for the mass difference @xmath118 , although , as noticed , these measurements are about six order of magnitude more sensitivity than the one leading to ( [ meson ] ) . in ref .
@xcite it has been argued that these figures of merit could not be the relevant ones in testing cpt breaking .
in fact , within the approach presented here , at the lowest order in the cpt violating parameters , one has @xmath123 , and similarly the charge - to - mass ratios do not depend on these parameters @xcite .
to review this point , let me start by the dirac equation for an electron or a proton including the breaking terms contained in @xmath124 ( of course , the breaking parameters may depend on the type of particle one is considering ) ( i^d_-m - b__5 ^ -1 2 h_^)=0 in a penning trap the radial confinement is obtained through a strong axial magnetic field , whereas the axial confinement is obtained by a quadrupole electric field .
the main corrections due to the cpt and lorentz breaking parameters are obtained by taking @xmath125 as the four - potential for a constant magnetic field .
then , to obtain the energy shifts generated by the breaking parameters one makes use of the relativistic landau levels wave functions and the expressions containing the full qed corrections for the unperturbed levels @xcite .
however , the underlying physics can be understood quite simply recalling the expression for the non - relativistic landau levels e_n,=(n+1 2+g 2 ) m , = 1 2the cyclotron and anomalous frequencies are obtained comparing two landau levels with different quantum number @xmath126 and with the same and opposite spin configurations respectively _
c&=&e_1,-1/2-e_0,-1/2=m + _ a&=&e_0,+1/2-e_1,-1/2=2mthe relevant cpt and lorentz breaking corrections to the energy levels are given by @xcite e^e^-_n,1/2 = b_3h_12 , e^e^+_n,1/2 = b_3h_12where we have taken the third axis along the magnetic field of the trap
. the frequencies for the antiparticles that we need according to the cpt theorem are the ones with inverted spin , therefore _
c^e^-=_c^e^+=_c , to conclude thisomega_a^e^=_a2 b_3 + 2h_12we get _ c_c^e^--_c^e^+=0 , _ a_a^e^- -_a^e^+=-4b_3we recall that these equations hold only at the first order in the breaking parameters and also that the usual relation @xmath127 does not hold here since , as noted before , the gyromagnetic ratios do not change at the lowest order .
since the observables that are measured in a penning trap are the anomalous and cyclotron frequencies , it seems natural to introduce figures of merit related to these observables .
a such figure of merit for cpt violation is @xcite r^e__a== where @xmath128 . for a weak magnetic field one
gets r^e__a==2ma new analysis of the 1987 experiment by dehmelt et al .
@xcite has been done recently in ref .
@xcite obtaining the following bound r^e__a1.210
^ -21however , the vector @xmath109 is absolutely constant and as such it rotates with a diurnal period of 23 h and 56 m , when seen in the laboratory frame wich is fixed with respect to the earth .
this effect might have given rise to non favorable situations during the observation , and therefore the bound has been a bit relaxed @xcite r^e__a310 ^ -21210 ^ -20 in the case of proton and atiproton there is no experiment at the moment . assuming an experimental sensitivity analogous to the electron positron case ( meaning @xmath129 ) one gets @xcite r^p__a=2m_p10 ^ -23
the last case i consider is the spectroscopy of free or magnetically trapped hydrogen ( @xmath130 ) and antihydrogen ( @xmath131 ) .
this is interesting since the two - photon @xmath132 transition has been measured with a precision of @xmath133 @xcite in a cold atomic beam of @xmath130 and with a precision of @xmath134 in trapped @xmath130 @xcite .
however for the free case the dependence of the @xmath132 transition on the cpt and lorentz breaking parameters is suppressed by a factor @xmath135 , since the @xmath136 and @xmath137 levels shift by the same amount at the leading order in the breaking @xcite .
consider now the spectroscopy of @xmath130 and @xmath131 in a magnetic field @xmath6 . in the basis
@xmath138 the four @xmath136 and @xmath137 hyperfine zeeman levels are , for @xmath139 & & |b_n=|-1/2,-1/2 , |d_n=|1/2,1/2 + & & |a_n=_n|-1/2,1/2-_n |1/2,-1/2 + & & |c_n=_n|-1/2,1/2+_n|1/2,-1/2with @xmath140 .
transitions of the type @xmath141 have leading - order sensitivity to lorentz and cpt violation , but they are field - dependent . as a consequence there is a problem connected with the broadening of the lines due to trapping field inhomogeneities .
consider now hyperfine transitions in the ground state .
again there is the problem of the zeeman broadening .
however one can try to eliminate the frequency dependence on @xmath6 ( at lowest order ) by choosing a field independent transition point @xcite . for @xmath142 the state @xmath143
is highly polarized ( @xmath144 ) .
then the effect on the transition @xmath145 of the cpt and lorentz violating parameters is @xmath146 .
therefore by putting @xmath147 the corresponding figure of merit can be defined as r_cd^h==4attaining a resolution of @xmath148 , one would get @xcite r_cd^h510 ^ -27
in this talk i have reviewed some important consequences of atomic physics measurements in the domain of high - energy physics .
in particular apv in cesium could be the first real indication of new physics beyond the sm .
the atomic physics tests of the cpt symmetry are already at a spectacular level of sensitivity , and the future experiments on @xmath130 and @xmath131 could give bounds well below the one expected from string theory .
vetter p.a . ,
meekhof d.m . ,
majumder p.k .
, lamoreaux s.k . and fortson e.n . , _ phys . rev .
lett . _ * 71 * , 3442 ( 1993 ) ; edwards n.h .
, phipp s.j . , baird e.g. , nakayama s. , _ phys .
* 74 * , 2654 ( 1995 ) .
marciano w.j . and rosner j.l . , _ phys .
lett . _ * 65 * , 2963 ( 1990 ) ; altarelli g. , lectures given at the _
les houches summer school : particles in the nineties _ , 30 jun - 26 jul 1991 , les houches , france .
the actual experimental lower bound on the higgs mass is 107.9 @xmath66 at 95% cl , see : aleph , delphi , l3 and opal collaborations . the lep working group for higgs boson searches , cern - ep-2000 - 055 , april 2000 .
amaldi u. _ et al .
_ , _ phys .
rev . _ * d36 * , 1385 ( 1987 ) ; marciano w.j . and rosner j.l . , _ phys .
lett . _ * 65 * , 2963 ( 1990 ) ; altarelli g. _ et al . _
* b261 * , 146 ( 1991 ) ; mahantappa k.t . and mohapatra p.k . , _ phys
. rev _ * d43 * 3093 ( 1991 ) ; rosner j.l .
_ * d61 * , 016006 ( 2000 ) . in these papers
it is also possible to find a complete list of references to the @xmath9 models .
kosteleck v.a . and potting r. , _ nucl .
_ * b359 * , 545 , ( 1991 ) ; _ ibidem _ , _ phys .
lett . _ * b381 * , 389 , ( 1996 ) ; kosteleck v.a . and samuel s. , _ phys . rev . lett . _ * 63 * , 224 , ( 1989 ) ; _ ibidem _ , _ phys .
lett . _ * 66 * , 1811 , ( 1991 ) ; _ ibidem _ , _ phys .
* d39 * , 683 , ( 1989 ) ; _ ibidem _ , _ phys .
rev . _ * d40 * , 1886 , ( 1989 ) .
van dyck r.s
, schwinberg p.b . and dehmelt h.g .
lett . _ * 59 * , 26 ( 1987 ) ; _ ibidem _ _ phys .
rev . _ * d34 * , 722 ( 1986 ) . for a review of the principles of the penning trap , see : brown l.s . and gabrielse g. _ rev .
_ * 58 * , 233 ( 1986 ) . | in this talk i discuss the relevance of atomic physics in understanding some important questions about elementary particle physics . a particular attention is devoted to atomic parity violation measurements which seem to suggest new physics beyond the standard model .
atomic physics might also be relevant in discovering possible violations of the cpt symmetry . |
the local density approximation ( lda ) ( ref . ) in density functional theory ( dft ) ( ref . ) has shown fantastic performance in understanding and even predicting material properties@xcite in spite of its relatively simple treatment of the exchange - correlation energy @xmath2 $ ] as a functional of the electron density @xmath3 ; e.g. , for many materials , lattice and elastic constants are generally reproduced .
the deviations from experimental values are within less than 1 - 2 % and several percent , respectively , in lda .
yet lda fails to describe some of properties including ground - state magnetic orderings even for bulk iron@xcite and for some transition - metal oxides.@xcite it also tends to overestimate the bonding strength , leading to an absolute error of molecular atomization energies.@xcite some of limitations of lda are remedied by the generalized gradient approximation ( gga ) in which the exchange - correlation energy is expressed in terms of not only the electron density but also its gradient .
the molecular atomization energies are calculated with the error of several tenths of ev,@xcite and the ground state of the bulk iron is correctly predicted to be a ferromagnetic body - center phase.@xcite the prevailed functional form of gga ( pbe ) ( ref . ) generally provides better accuracy for structural properties of variety of solids and activation energies in chemical reactions than lda does . the local ( lda ) and semilocal ( gga ) approximations are still insufficient to describe some of important properties , however .
the ground states of strongly correlated materials are incorrectly predicted and the energy band gaps of most semiconductors and insulators are substantially underestimated .
the meta - gga scheme@xcite extending the exchange - correlation functionals with including the kinetic energy density further improves the lda and gga results for molecular systems@xcite but not succeed to remedy above failures in condensed matters . the failure of lda and gga is occasionally discussed in terms of the self - interaction error ( sie).@xcite an electron is under the electrostatic potential due to other electrons . yet the expression of the electrostatic potential in the ( semi)local approximations includes the spurious interaction with the electron itself .
when we consider the hartree - fock ( hf ) exchange potential with kohn - sham orbitals , this spurious self - interaction is cancelled by a term in the exchange potential . in the ( semi)local expression of the exchange potential , however , this cancellation is incomplete so that each electron is affected by the self - interaction .
this sie causes delocalization of the electron , predicting the incorrect fractional - charged ground state of , e.g. , h@xmath4 with large nucleus separation.@xcite the sie affects the band gaps substantially .
the band gap @xmath5 is formally defined as the ionization energy subtracted by the electron affinity so that @xmath6 where @xmath7 is the total energy of the @xmath8-electron system . in dft with the exact exchange - correlation energy , the band gap is expressed as the difference between the highest occupied kohn - sham level @xmath9 of the @xmath10-electron system and its counterpart of the @xmath8-electron system @xmath11 : i.e. , @xmath12.@xcite when we introduce a fractional electron system with @xmath13 electrons as a mixed state of real integer - electron systems , then the total energy @xmath14 becomes linear for @xmath15 and shows discontinuity at the integer value @xmath8 for finite - gap systems .
using janak theorem@xcite which relates the kohn - sham level to the derivative of the total energy as @xmath16 , the linearity of @xmath17 leads to the constant @xmath18 as a function of @xmath19 . in the ( semi)local approximations , however , the kohn - sham level @xmath20 [ @xmath21 increases ( decreases ) with increasing @xmath19 due to the self - interaction , leading to the concave shape of @xmath14
this may cause an underestimate of the energy gap.@xcite the hf approximation ( hfa ) is free from the self - interaction . yet
the calculated band gaps in hfa are substantially overestimated due to the lack of the correlation energy .
an approach called the optimized effective potential@xcite which is incorporated in dft ( refs . ) intending to remedy the issue is still in an immature stage in a view of applications to polyatomic systems .
hence the hybrid functionals combining lda or gga with hfa may be effective to break the limitation of the semilocal approximations .
the hybrid approach has begun in empirical ways : the hf - exchange energy was mixed with the lda exchange - correlation energy in the half and half way@xcite and then three mixing parameters were introduced@xcite to mix the lda , gga , and hfa energies ; the latter scheme is called b3lyp and has widely been used to clarify thermochemical properties of molecules.@xcite a rationale for the hybrid functional is provided@xcite in the light of the adiabatic - connection theorem,@xcite @xmath22 = \int_0 ^ 1 d\lambda \ e_{\rm xc,\lambda } \ , \label{adconn}\ ] ] where @xmath23 is the energy of the exchange and correlation in a system , where the electron - electron interaction @xmath24 is reduced by the factor @xmath25 but the external potential @xmath26 is added to reproduce the electron density @xmath27 of the real system ( @xmath28 ) . here @xmath29 is the ground - state many - body wave function . by assuming that @xmath30 is the 4th polynomial of @xmath25 with particular asymptotic forms for @xmath31 and @xmath28 , perdew , ernzerhof , and becke have proposed a parameter - free hybrid functional called pbe0@xcite in which the hf and pbe exchange energies are mixed with the ratio of 1:3 .
its applicability has been examined for molecular systems.@xcite screening of the coulomb potential is effective in polyatomic systems .
hence it may be effective to apply the non - local hf - exchange operator only to the short - range part of the coulomb potential.@xcite this is conveniently done by introducing the error function splitting the coulomb potential into short - range and long - range components .
the hybrid functional which is constructed in this way from the pbe0 functional is proposed by heyd , sucseria , and ernzerhof ( hse).@xcite this treatment reduces computational cost substantially and opens a possibility to apply the hybrid functionals to condensed matters .
the structural properties as well as the band gaps of several solids have been calculated and significant improvements on semi - local functionals are achieved.@xcite on the other hand , effects of the exchange interaction for the long - range component of the coulomb potential are certainly important@xcite in a view of reducing sie .
hirao and his collaborators have proposed a long - range corrected ( lc ) functional in which the long - range component is treated by the hf exchange energy and the short - range component is by the lda exchange.@xcite they have applied to various molecular systems and obtained relatively successful results.@xcite further application of the lc functional to molecular systems and its comparison with other functionals have been done and the applicability of the lc functional have been recognized.@xcite the lc functional has also been applied to structural properties and band gaps of several condensed matters and the results are compared with those obtained from other functionals.@xcite at the present stage , several hybrid functionals have been implemented in different packages and the assessment of the validity of each functional has been done mainly to molecular systems .
applicability of the pbe0 and hse functionals to condensed matters have been examined.@xcite the structural and electronic properties such as lattice constants , bulk moduli , and also the band gaps of condensed matters are obtained only after careful examinations of various calculation parameters .
obviously , numerical precision should not be neglected .
in order to assess the validity of each hybrid functional , it is thus imperative to perform the computation in a single reliable calculation scheme . furthermore , the split of the coulomb potential into the short - range and long - range parts requires another parameter @xmath32 being the exponent of the error function .
the @xmath32 dependence of the results should certainly be examined for better understanding and further improvements on the hybrid functionals .
the aim of the present paper is to implement several important hybrid functionals in the well - established plane - wave - basis total - energy band - structure calculation code and examine validity and limitation of each functional .
a plane - wave code we adopt in this work is _ tokyo ab initio program package _
( tapp).@xcite we have calculated lattice constants , bulk moduli , band gaps , and band widths of various semiconductors and insulators with the pbe0 , hse , and lc hybrid functionals as well as ( semi)local gga functional . comparison of the obtained results unequivocally elucidates the validity and the limitation of the hybrid functionals . in section [ hybrid
] , we briefly describe each of the hybrid functionals used in the present paper .
section [ computation ] presents details of our computational scheme .
the calculated results are shown in section [ result ] , and our finding is summarized in section [ conclusion ] .
in this section , we briefly describe the three hybrid functionals , pbe0 , hse , and lc , which we examine their applicability in this paper .
perdew , burke , and ernzerhof have proposed@xcite a polynomial form for @xmath33 as @xmath34 where @xmath35 and @xmath36 are the exchange energies obtained by hfa and a certain ( semi)local approximation in dft , respectively . here
@xmath37 is the energy of the exchange and correlation defined as eq .
( [ exclam ] ) and obtained by the ( semi)local approximation in dft .
this formula infers that the ( semi)local approximation in dft is a good approximation to @xmath38 for @xmath39 .
when @xmath40 , this formula equals to @xmath41 since @xmath42 for @xmath40 . from eqs
( [ adconn ] ) and ( [ eq : exc_lambda ] ) , we obtain @xmath43 relying on the fourth - order mller - plesset perturbation theory applied for molecular systems , it is argued that @xmath44 is the best choice.@xcite using the pbe functional@xcite as the approximation in dft in eq .
( [ pbe0_formula ] ) , the pbe0 hybrid functional is given by @xmath45 leading to the mixing of 25 % hf exchange and 75 % pbe exchange .
heyd , sucseria , and ernzerhof have proposed@xcite a different hybrid functional in which the long - range part of the hf - exchange energy is treated by the semilocal approximation in dft and the short - range part is calculated exactly .
the actual procedure is conveniently done by splitting the coulomb potential as @xmath46 and applying the first term only , i.e. , the screened coulomb potential , to the hf - exchange energy . the second term to the exchange energy is calculated with gga .
adopting the mixing ratio in pbe0 , then the hse hybrid functional becomes @xmath47 @xmath48 is the fock - type double integral with the screened coulomb potential .
there is some complexity to divide the pbe exchange energy @xmath49 into the short - range part @xmath50 and the long - range part @xmath51 .
the dividing procedure will be shown in subsection [ srlr ] .
another ambiguous factor is the parameter @xmath32 in eq .
( [ split ] ) which defines the short - range and the long - range parts of the coulomb potential .
the optimum value of @xmath32 = 0.15 @xmath52 ( @xmath53 : bohr radius ) is proposed by examining the calculated results for molecular systems.@xcite examining the @xmath32 dependence of the calculated results for condensed matters is one of our aims in this paper .
the hse functional partly removes sie by incorporating the hf - exchange energy in the pbe functional .
yet the cancellation of the hartree potential and the exchange potential is absent in the long - range part .
this may cause erroneous description of , e.g. , the rydberg states in isolated polyatomic systems or properties of charge - transfer systems . to remedy this point , application of the long - range part of the coulomb potential to the hf - exchange energy is necessary .
the long - range corrected ( lc ) functional has been proposed based on this viewpoint,@xcite being expressed as @xmath54 with @xmath55 being the fock - type double integral with the long - range part of the coulomb potential [ the second term of eq .
( [ split ] ) ] . for the dft part , several approximations including lda,@xcite , pbe,@xcite and other gga ( ref . ) or meta - gga ( ref . )
forms are adopted in eq .
( [ lc_formula ] ) and their validities are examined . it is argued that pbe combined in the lc hybrid scheme provides good accuracy for molecular properties.@xcite as in the hse scheme , there is some complexity to extract the long - range component of the exchange energy of the dft part , @xmath56 .
the parameter @xmath32 in eq .
( [ split ] ) affects the results substantially . by examining the results for molecular systems , the values ranging @xmath32 = 0.25 - 0.5
are argued to be optimum.@xcite by applying lda in the lc - hybrid scheme to structural properties of solids , the value @xmath32 = 0.5 is found to produce reasonable results.@xcite it is of interest to investigate the appropriate value of @xmath32 in the application of pbe in the lc - hybrid scheme to structural properties and band gaps of condensed matters .
in this section , we describe our implementation of the hybrid functionals in the plane - wave - basis - set total - energy band - structure calculation code , tapp.@xcite nuclei and core electrons are simulated by either norm - conserving@xcite or ultrasoft@xcite pseudopotentials in the tapp code .
non - locality of the hf - exchange potential generally increases computational cost tremendously in the application to condensed matters .
we have circumvented this problem using the fast - fourier transform ( fft ) , as explained below .
the long - range nature of the coulomb potential leads to singularity of its fourier transform at the origin .
this causes difficulty in numerical integration over the brillouin zone ( bz ) to obtain the hf - exchange energy in the pbe0 and lc schemes .
we here adopt a simple truncation scheme to overcome the problem .
finally we explain how to divide the exchange energy in the pbe functional to the short - range and long - range components .
the hf - exchange energy @xmath57 in condensed matters ( crystal ) are written as @xmath58 where @xmath59 is the occupation number , and @xmath60 is given by @xmath61 here @xmath62 is the bloch - state orbital with the band index @xmath63 and the wavevector @xmath64 .
the orbital @xmath62 is obtained by solving selfconsistently the euler equation ( kohn - sham equation ) in which the exchange - correlation potential is given by the functional derivative of the hybrid exchange - correlation energy .
the sum over @xmath64 and @xmath63 are taken for the occupied states .
our algorithm to compute the integral @xmath60 in the plane - wave - basis code is as follows : eq .
( [ jij ] ) is written as @xmath65 with @xmath66 . in the plane - wave - basis - set scheme , the reciprocal lattice vectors *
g * are used to represent the periodic wave function as @xmath67 .
we then obtain @xmath68 when we compute eq .
( [ jij ] ) directly , the calculation costs of @xmath69 and @xmath70 are @xmath71 and @xmath72 , respectively , where @xmath73 is the total number of the reciprocal vectors , @xmath74 the total number of sampling @xmath75 points in bz , @xmath76 the total number of the occupied bands .
the number of @xmath73 is much bigger than either @xmath77 or @xmath78 .
hence the order @xmath79 is computationally demanding .
we reduce this computational cost by using fft .
we first define the overlap density between states ( @xmath80 ) and ( @xmath81 ) as @xmath82 using this quantity , we obtain @xmath83 we note that , since @xmath84 has unit - cell periodicity , the overlap density also has the same periodicity . using the fourier transformation of the overlap density @xmath85 , eq .
( [ jijk ] ) is rewritten as @xmath86 since the @xmath87 can be calculated outside the summation loop in eq .
( [ jijkl ] ) , the total calculation cost for @xmath88 is @xmath89 .
the computational cost of @xmath90 becomes @xmath91 . with considering the scaling as @xmath92 , @xmath93 , and @xmath94 with @xmath95
the number of atoms in the unitcell , the computational cost above is proportional to @xmath96 .
the fourier transform @xmath97 of the coulomb potential @xmath98 diverges at the long - wave length limit @xmath99 .
calculations of electrostatic energies thus require careful treatment and the well - known ewald summation is the typical example . in calculating non - local exchange energies in condensed matters
, more careful treatment is necessary .
we need to perform the bz integration in evaluation of the exchange energy in eq .
( [ hfx_solid ] ) .
the integration is usually performed by the summation with weighting factors of the integrand at finite discrete @xmath100 points , and thus encounter a difficulty to evaluate the integral accurately by picking up the singular behavior of the fourier transform of the coulomb potential .
there are several ways to overcome the difficulty .
one is the auxiliary - function approach : an auxiliary function which has the same singular behavior but is integrable is subtracted so that the summation can be done properly and the remaining term is obtained analytically.@xcite an alternative way which we adopt in the present paper is simpler : we make the coulomb potential truncated at @xmath101 and then examine the convergence by numerically increasing @xmath102.@xcite namely , we replace coulomb potential with a truncated potential , @xmath103 with a cutoff radius @xmath101 .
what we need is of course converged quantities with @xmath104 .
the truncated potential produces its non - divergent fourier transform , @xmath105 \ .\ ] ] convergence of required quantities with respect to @xmath101 is combined with the number of sampling @xmath100 points in the bz integration .
the number @xmath106 of sampling @xmath100 points required should increase with increasing @xmath101 .
we need to know the converged values with increasing both @xmath101 and @xmath106 .
the process to check this convergence in the two - parameter space can be done conveniently by introducing a certain relation between @xmath101 and @xmath106 .
we set up a relation , @xmath107 , where @xmath108 is the unit - cell volume , and examine the convergence of the exchange energy by increasing @xmath106 and equivalently @xmath101 .
we have found that the values of @xmath101 which are 4 - 6 times the dimension of the primitive unit cell are enough to assure the converged exchange energies in ten materials calculated in the present paper .
we next describe how to obtain the short - range and long - range parts in the pbe - exchange energy , following ref . . the original expression for the pbe - exchange energy is given by , @xmath111 n({\bf r } ) f_{\rm{x}}^{\rm{pbe}}(s )
, \label{tot_pbe_x}\end{aligned}\ ] ] where @xmath112 $ ] is the exchange energy of the homogeneous electron gas with the electron density @xmath63 and @xmath113 is an enhanced factor due to a density gradient @xmath114 with the fermi wavenumber @xmath115 .
the enhanced factor is written in an integral form @xmath116 where @xmath117 is a dimensionless quantity and @xmath118 describes an exchange - hole density at the distance @xmath119 .
heyde , scuseria , and ernzerhof@xcite proposed an expression for the short - range pbe - exchange functional , where the original coulomb interaction @xmath120 is modified to a screened form @xmath121 .
this modification leads to an enhanced factor somewhat different from the original expression of eq .
( [ tot_pbe_f ] ) as @xmath122 the short - range exchange energy is given with this enhanced factor as @xmath123 n({\bf r } ) f_{{\rm x}}^{{\rm pbe , sr}}(s,\omega ) .
\label{sr_pbe_x } \end{aligned}\ ] ] the long - range pbe - exchange term is defined by subtracting the short - range part in eq .
( [ sr_pbe_x ] ) from the original one in eq .
( [ tot_pbe_x ] ) as @xmath124 implementation details for these calculations can also be found in ref . .
we generate norm - conserving pseudopotential to simulate nuclei and core electrons , following a recipe by troullier and matins.@xcite the core radius @xmath125 is an essential parameter to determine transferability of the generated pseudopotential . we have examined @xmath125 dependence of the calculated structural properties of benchmark materials and adopted the pseudopotentials generated with the following core radii in this paper : 0.85 for si 3@xmath126 , and 1.16 for si 3@xmath127 , 1.06 for ge 4@xmath126 and 4@xmath127 , 1.06 for ga 4@xmath126 and 4@xmath127 , and 1.48 for ga 4@xmath128 , 0.64 for n 2@xmath126 and 2@xmath127 , 0.85 for c 2@xmath126 and 2@xmath127 , 0.79 for o 2@xmath126 and 2@xmath127 , 1.38 for na 2@xmath126 and 2@xmath127 , 1.16 for cl 3@xmath126 and 3@xmath127 , 0.95 for li 2@xmath126 , 0.64 for f 2@xmath126 and 2@xmath127 , 1.59 for ba 5@xmath126 , 5@xmath127 , and 5@xmath128 , 1.32 for mg 2@xmath126 and 2@xmath127 , and 1.38 for ti 3@xmath128 and 4@xmath126 , and 1.43 for ti 4@xmath127 , 1.48 for kr 4@xmath126 and 1.37 for kr 4@xmath127 .
the pseudopotentials are generated by the ( semi)local approximations in dft .
this means that the hf - exchange energy between core and valence states are neglected .
yet we have found that the calculated energy bands obtained with the pseudopotentials generated in lda and gga , are essentially identical to each other , implying that the treatment of the exchange - correlation energy in generating pseudopotentials has minor effects . the partial core correction@xcite is not included in our calculations .
this is partly because magnetic properties are not considered in the present work .
however , to assure the accuracy of structural properties , we regard some of core orbitals as valence orbitals and include them explicitly in the pseudopotential generation : such orbitals included as valence states are @xmath129 and @xmath130 orbitals of na and @xmath129 and @xmath130 orbitals of mg
. appropriate choice of cutoff energies @xmath131 in the plane - wave - basis set , which is related to hardness of the adopted norm - conserving pseudopotentials , is a principal ingredient to assure the accuracy of the results .
we have examined convergence of structural properties and band gaps with respect to @xmath131 and reached the following well converged values with @xmath131 for each material : 25 ryd for si and ge , 36 ryd for kr , 64 ryd for batio@xmath0 , and 100 ryd for diamond , gan , mgo , nacl , licl , and lif .
the remaining important ingredient to assure our assessment of each hybrid functional is the sampling @xmath100 points for the bz integration .
we have adopted the scheme by monkhorst and pack in which bz is divided by equally spaced mesh .
after careful examination , we have found that 4@xmath1324@xmath1324 sampling @xmath100 points are enough to assure the accuracy of the total energies and energy bands in the ten materials .
the results are confirmed by repeating the calculations with 6@xmath1326@xmath1326 sampling @xmath100 points .
l@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r + [ 2pt ] + [ -4pt ] & pbe & hf & pbe0 & & & expt .
+ [ 2pt ] + [ -4pt ] & & & & @xmath32=0.1 & @xmath32=0.2 & @xmath32=0.3 & @xmath32=0.4 & @xmath32=0.1 & @xmath32=0.2 & @xmath32=0.3 & @xmath32=0.4 & + + ge & 0&4.75&1.00&0.76 & 0.54 & 0.43 & 0.27&1.05&2.05&2.66&3.69&0.74 + si & 0.61&6.03&1.72 & 1.20 & 0.94 & 0.80&0.73&2.24&3.85&4.19&4.63&1.17 + batio@xmath0&2.14&11.62 & 4.21 & 3.57 & 3.12&2.83&2.73&4.59&6.26&7.45&8.01&3.2 + @xmath1-gan&2.06&8.84 & 3.51 & 3.02 & 2.65&2.42&2.26&3.95&5.58&6.55&7.36&3.30 + c & 4.01&12.44&5.87 & 5.28 & 4.88 & 4.62&4.45&5.06&6.80&7.95&8.73&5.48 + mgo & 4.95&14.21&6.99 & 6.62 & 6.14 & 5.79&5.52&6.38&8.33&9.88&11.12&7.7 + nacl & 5.13&13.38&6.95 & 6.46 & 6.00 & 5.69&5.49&7.03&8.90&10.22&11.12&8.5 + licl & 6.33&14.85&8.60 & 8.14 & 7.67 & 7.36&7.15&8.53&10.45&11.75&12.66&9.4 + lif & 9.70&21.57&12.51&12.02&11.45&11.02&10.68&11.59&13.87&15.65&17.06&14.30 + kr & 7.09&15.22&9.14 & 8.46 & 7.98 & 7.68 & 7.48 & 9.85&11.75&13.00&13.75 & 11.65 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 42.5&179.7&19.7&12.4&19.9&26.4&31.5&28.2&62.3&88.9&117.4 + mre ( % ) & @xmath13342.5&179.7&5.7&@xmath1339.0&@xmath13319.9&@xmath13326.4&@xmath13331.5&11.1&61.7&88.9&117.4 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 49.1&303.1&25.4&5.8&16.0&25.5&33.2&40.8&119.0&158.8&205.4 + mre ( % ) & @xmath13349.1&303.1&25.4&0.9&@xmath13316.0&@xmath13325.5&@xmath13333.2&37.8&119.0&158.8&205.4 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 35.9&56.3&14.0&19.0&23.9&27.3&29.8&15.6&5.6&18.9&29.4 + mae ( % ) & @xmath13335.9&56.3&@xmath13314.0&@xmath13319.0&@xmath13323.9&@xmath13327.3&@xmath13329.8&@xmath13315.6&4.4&18.9&29.4
+ we have performed total - energy electronic - structure calculations using pbe , hf , pbe0 , hse , lc exchange - correlation functionals for ten materials , including covalent semiconductors , ionic insulators , dielectric compounds , and a rare - gas solid : i.e. , si , ge , gan , diamond , mgo , nacl , licl , lif , batio@xmath0 , and kr .
the calculated results elucidate capability and limitation of each functional in discussing electronic and structural properties of these prototype materials .
we first present calculated electron states of these materials and compare them with experimental results in the subsection [ ele_state ] .
then we present the results of structural optimization in the subsection [ structure ] .
l@ r@ r@ r @ r@ r@ r @ r@ r@ r @ r + [ -2pt ] + [ 2pt ] + [ -4pt ] & pbe&hf&pbe0 & & & expt .
+ [ 2pt ] + [ -4pt ] & & & & @xmath32=0.1 & @xmath32=0.2 & @xmath32=0.3 & @xmath32=0.1 & @xmath32=0.2 & @xmath32=0.3 & + [ 2pt ] + [ -4pt ] ge&12.85&18.22&14.50&13.80&13.47&13.11&14.23&15.54&16.36&[email protected] + si&11.95&16.90&13.37&13.28&12.99&12.65&12.19&13.45&14.52&[email protected] + batio@xmath0 & 4.53&6.37&5.08&5.03&4.95&4.85&4.85&5.34&5.72 & + @xmath1-gan&6.73&8.27&7.20&7.11&7.02&6.93&7.11&7.44&7.91 & + c&21.65&30.09&23.64&23.51&23.28&22.96&24.15&25.00&26.19 & 24.2@xmath1341 , 21@xmath1341 + mgo & 4.43&6.16&4.97 & 4.80 & 4.70 & 4.60&5.21&5.56&5.92 & + nacl&1.65&2.24&1.84 & 1.77 & 1.71 & 1.67&2.04&2.25&2.38 & + licl&2.82&4.33&3.39 & 3.16 & 3.16 & 3.09&3.59&3.91&4.17 & + kr & 1.50&1.93&1.61 & 1.59 & 1.55 & 1.52&1.57&1.74&1.85 & + lif & 2.83 & 3.66&3.09 & 2.96 & 2.92 & 2.87 & 2.94&3.09&3.27 & + figure [ si_band ] shows calculated band structures of si with five exchange - correlation functionals . for the hse and lc functionals , the band structures with different choices of @xmath32 [ 0.1 , 0.2 , and 0.3 ( @xmath52 ) ] in eq .
( [ split ] ) are shown .
the overall features of the band structures obtained by the five functionals are similar to each other . yet the bandwidths and the fundamental gaps are different quantitatively . when the ratio of the hf exchange to the total - exchange functional is large , the resulting bandwidth and gap become large ; these quantities become larger in the order of pbe , pbe0 , hse , lc , and hf . notice that the band gaps obtained by the lc functional are always larger than those by the hse functional , indicating that the correction to the long - range part of the exchange potential tends to make the band gap large .
we also show the calculated energy bands of the dielectric compound batio@xmath135 and the ionic insulator nacl in figs .
[ batio3_band ] and [ nacl_band ] , respectively .
we find the general tendency similar to that in si : i.e. , the overall features of the energy bands are insensitive to the difference in the functionals ; the lc functional provides larger energy gaps compared with the hse functional .
table [ bandgap ] and [ bandwidth ] summarize the calculated bandwidths and band gaps for the ten materials .
for assessment of validity of each functional , it is convenient to introduce two quantities , the mean relative error ( mre ) which is the mean of the calculated value minus the experimental value over the ten materials , and the mean absolute relative error ( mare ) which is the mean of the absolute value of the difference between the calculated and the experimental values over the ten materials .
the mre is a measure of under- or over - estimates of the experimental values since each functional predicts either smaller or larger values than the experimental values for most of the ten materials . on the other hand , mare is a measure of closeness between the calculated and experimental values . in the discussion below , we categorize the ten materials into group i and group ii : group i , i.e. , diamond , gan , batio@xmath0 , si and ge has covalent characters in which the band gaps are less than 7 ev ; group ii , i.e. , lif , kr , nacl , licl and mgo , consists of ionic solids and a rare - gas solid in which the experimental band gaps are larger than 7 ev .
the calculated band gap by the pbe functional for each material is substantially smaller than the corresponding experimental value , as is reported in literature .
the calculated mre for the ten materials is @xmath13342.5 % . on the other hand ,
hfa largely overestimates the band gaps for all the ten materials : mre is 179.7 % .
the pbe0 functional provides better values .
the calculated mre by the pbe0 for the materials in group i is 25.4 % , whereas it gives @xmath13314.0 % for the materials in group ii .
the hse and lc functional also provide better agreement of the band gap with the experiments than the pbe does .
further when we choose optimum values of @xmath32 , the hse and lc results show nice agreement with the experimental values . in the hse functional ,
a general trend is the decrease in the band gap with increasing @xmath32 .
we have found that the value @xmath32 = 0.1 @xmath52 produces the band gaps close to the experimental values within mre = 0.9 % for the materials in group i. it produces worse values for the materials in group ii .
the calculated mre is @xmath13319.0 % for the materials in group ii , but still hse being the better approximation than pbe .
the lc functional provides good agreement with the experimental values for the materials in group ii : when we choose the optimum value of @xmath136 , the calculated mre is nicely small , 4.4 % for the materials in group ii .
yet the lc provides worse values for the materials in group i , showing its limitation as an universally valid approximation .
figure [ theory_vs_expt ] is a summary of our calculated band gaps by the pbe , hse and lc functionals .
for hse and lc , we show the calculated results with the optimum @xmath32 values : @xmath32 = 0.1 @xmath52 for hse and @xmath32 = 0.2 @xmath52 for lc .
it is clearly shown that the calculated band gaps by hybrid functionals , hse and lc , are in better agreement with the experimental values than the pbe ( gga ) approximation , indicating the promising possibility of the hybrid functionals .
the degree of the agreement is close to the greens - function - based gw approximation in which there are several ambiguities in theoretical treatments.@xcite however , this figure also shows the limitation ; hse is reasonably good for only the group i materials , whereas lc is good for only the group ii materials .
our finding here is that hse is a good approximation for relatively small - gap materials and that lc is a good approximation for the relatively large - gap materials .
screening of the coulomb interaction depends on the materials .
hence the best treatment of the short - range and long - range parts of the coulomb interactions should change material by material .
what we have found here is natural in the sense .
however , another point we have shown here is that the appropriate choice of @xmath32 with each exchange - correlation functional provides reasonable agreement with the calculated band gaps for rather wide range of materials ; hse with @xmath32 @xmath137 0.1 @xmath52 for the materials with their band gaps smaller than 7 ev and lc with @xmath32 @xmath137 0.2 @xmath52 for the materials with their band gaps
larger than 7 ev .
this information certainly provides a practical recipe to obtain reliable band gaps for various materials . the calculated bandwidths by the pbe functional are in good agreements with the experimental values available ( table [ bandwidth ] ) .
other hybrid functionals also provide reasonable agreement , in particular , with appropriate choices of the @xmath32 parameter for the hse and lc functionals . on the other hand ,
hfa shows substantial overestimate for each material .
l@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r + [ -4pt ] + [ 2pt ] + [ -4pt ] & pbe & hf & pbe0 & & & expt .
+ [ 2pt ] + [ -4pt ] & & & & @xmath32=0.1 & @xmath32=0.2 & @xmath32=0.3 & @xmath32=0.4 & @xmath32=0.1 & @xmath32=0.2 & @xmath32=0.3 & @xmath32=0.4 & + [ 2pt ] + [ -4pt ] ge&5.589&5.574&5.615&5.546&5.556&5.566&5.571&5.579&5.534&5.508&5.490&5.652 + si&5.463&5.387&5.431&5.435&5.441&5.446&5.452&5.459&5.434&5.416&5.403&5.430 + batio@xmath0&4.139&4.048&4.068&4.102&4.106&4.111&4.116&4.137&4.120&4.102&4.088&4.000 + @xmath1-gan&4.539&4.380&4.434&4.455&4.457&4.441&4.461&4.502&4.511&4.491&4.452&4.520 + c&3.563&3.485&3.506&3.522&3.540&3.543&3.547&3.560&3.556&3.543&3.531&3.567 + mgo&4.202&4.198&4.198&4.117&4.136&4.141&4.146&4.178&4.175&4.157&4.138&4.207 + nacl&5.541&5.416&5.444&5.500&5.505&5.509&5.514&5.525&5.503&5.487&5.472&5.595 + licl&5.175&5.107&5.133&5.145&5.147&5.155&5.158&5.165&5.140&5.117&5.115&5.106 + kr & 10.86&10.06&10.74&10.79&10.81&10.86&10.86&10.69&10.49&10.05&10.05&9.94 + lif&4.115&4.006&4.030&4.041&4.071&4.073&4.076&4.103&4.098&4.091&4.080&4.010 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 1.20&1.37&1.10&1.40&1.37&1.41&1.35&1.25&1.21&1.34&1.54 + mre ( % ) & 0.68&@xmath1331.10&@xmath1330.49&@xmath1330.47&@xmath1330.22&@xmath1330.16&@xmath1330.03&0.40&0.10&@xmath1330.27&@xmath1330.62 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 1.15&1.75&1.20&1.44&1.34&1.40&1.32&1.17&1.13&1.33&1.62 + mre ( % ) & 0.66&@xmath1331.27&@xmath1330.51&@xmath1330.39&@xmath1330.20&@xmath1330.17&0.00&0.41&0.10&@xmath1330.31&@xmath1330.74 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 1.26&0.88&0.99&1.34&1.41&1.41&1.39&1.35&1.32&1.34&1.44 + mre ( % ) & 0.72&@xmath1330.87&@xmath1330.47&@xmath1330.58&@xmath1330.24&@xmath1330.14&@xmath1330.06&0.38&0.11&@xmath1330.22&@xmath1330.48 + [ 2pt ] we next examine performance of the hybrid functionals in describing structural properties .
table [ lattice ] and [ bm ] show calculated lattice constants @xmath138 and bulk moduli @xmath139 , respectively , of the ten materials .
the @xmath138 and @xmath139 values are determined by fitting parameters in the murnaghan equation of state to the calculated total energy as a function of the volume . in mare and mre presented in table [ lattice ] ,
the calculated value for solid kr is not included .
as shown in this table , the calculated lattice constant of kr is substantially larger than the experimental value : it is overestimated by 1.1 - 9.2 % in gga and in the hybrid approximations , and by 1.2 % in hfa .
this is because van der waals interaction which is unable to be treated in the approximations examined in this paper plays an essential role in solid kr .
we thus exclude the value for kr from the statistical assessment .
the results obtained by pbe reasonably agree with the experimental values ( mre = 0.68 % and mare = 1.20 % for @xmath138 and mre = @xmath1337.1 % and mare = 8.3 % for @xmath139 ) . the accuracy of hfa is slightly inferior to that of pbe ( mre = @xmath1331.10 % and mare = 1.37 % for @xmath138 and mre = 13.7 % and mare = 13.7 % for @xmath139 ) . the pbe0 functional shows better accuracy with the mre = @xmath1330.49 % and mare = 1.10 % for @xmath138 , and mre = @xmath1331.7 % and mare = 5.1 % for @xmath139 .
the calculated lattice constants using the hse and lc functionals are insensitive to the choice of @xmath32 : the difference obtained from different @xmath32 values is within 1 % .
the difference in the calculated bulk modulus from different @xmath32 values are not so small as in the lattice constant partly because the fitting by the murnaghan equation is incomplete .
for the hse functional , the value @xmath32 = 0.1 @xmath52 produces best agreement with the experiments : mre = @xmath1330.47 % and mare = 1.40 % for @xmath138 , and mre = 1.7 % and mare = 3.6 % for @xmath139 . for the lc functional
, we have found that @xmath32 = 0.2 @xmath52 produces best results : mre = 0.10 % and mare = 1.21 % for @xmath138 , and mre = 2.0 % and mare = 5.5 % for @xmath139 .
these optimum values for @xmath32 for the hse and lc functionals are identical to the optimum values determined from the calculated mre and mare for band gaps , corroborating the appropriate choice of @xmath32 .
l@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r@ r + [ -4pt ] + [ 2pt ] + [ -4pt ] & pbe & hf & pbe0 & & & expt .
+ [ 2pt ] + [ -4pt ] & & & & @xmath32=0.1 & @xmath32=0.2 & @xmath32=0.3 & @xmath32=0.4 & @xmath32=0.1 & @xmath32=0.2 & @xmath32=0.3 & @xmath32=0.4 & + [ 2pt ] + [ -4pt ] ge&69.0&76.0&68.8&76.3&75.8&73.3&72.7&71.8&85.3&89.0&95.8&75.8 + si&91.1&116.3&98.1&95.9&93.8&92.0&90.4&90.4&98.0&104.3&109.0&99.2 + batio@xmath0&146.0&184.0&167.0&164.0&162.0&159.0&156.0&150.0&167.0&166.0&175.0&162.0 + @xmath1-gan&173.0&274.6&220.3&215.5&213.3&203.5&189.1&192.2&195.0&218.1&249.3&210.0 + c & 456.0&534.0&488.0&485.0&483.0&478.0&473.0&469.0&472.0&490.0&508.0&443.0 + mgo & 152.0&181.0&169.0&166.0&164.0&162.0&160.0&152.0&154.0&161.0&169.0&165.0 + nacl&27.3&32.7&29.3&29.0&28.9&28.7&28.4&28.6&29.1&29.9&30.9&26.6 + licl&30.7&37.3&33.6&33.6&33.5&33.3&33.1&34.2&36.3&35.1&36.0&35.4 + kr & 2.4 & 6.4 & 3.8 & 3.8 & 3.5 & 3.1 & 2.7 & 3.2 & 3.5 & 4.7 & 5.1 & 3.4 + lif & 67.2&72.0&69.9&69.8&69.8&69.6&69.2&73.1&69.3&67.9&72.0&69.8 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 8.3&13.7&5.1&3.6&3.4&4.4&5.6&6.6&5.5&6.4&11.2 + mre ( % ) & @xmath1337.1&13.7&1.7&1.7&0.9&@xmath1330.9&@xmath1332.6&@xmath1332.6&2.0&5.1&11.2 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 9.5&16.5&5.7&3.5&3.2&4.7&6.7&7.2&6.1&7.9&15.5 + mre ( % ) & @xmath1338.3&16.5&1.6&2.1&1.0&@xmath1331.5&@xmath1334.0&@xmath1334.8&2.8&7.9&15.5 + [ 2pt ] + [ -4pt ] + [ 2pt ] mare ( % ) & 6.9&10.3&4.5&3.7&3.7&4.0&4.3&5.9&4.8&4.6&5.9 + mre ( % ) & @xmath1335.6&10.3&1.9&1.1&0.7&0.0&@xmath1330.9&0.2&1.1&1.6&5.9 + [ 2pt ]
we have studied validity of hybrid exchange - correlation functionals in density functional theory by implementing three hybrid functionals in a well - established plane - wave - basis - set code named tapp and by calculating structural properties and electron states of representative ten materials where the experimental energy gaps range from 0.67 ev to 14.20 ev .
the three hybrid exchange - correlation functionals examined in this paper are pbe0 proposed by perdew , burke , and ernzerhoff , hse proposed by heyd , scuseria , and ernzerhoff , and lc originally proposed by savin and by hirao and his collaborators . for comparison ,
results from the generalized gradient approximation , i.e. , pbe , and hfa have been presented .
the ten materials we have examined are ge , si , gan , batio@xmath0 , diamond , mgo , licl , nacl , kr , and lif which are representatives of covalent , ionic , and rare - gas solids .
we have found that the structural properties such as lattice constants are already well reproduced by the pbe functional and also by hfa and that the hybrid functionals show better agreement with the experimental values .
we have determined appropriate values of @xmath32 in the separation of the short - range and long - range parts in the coulomb interaction : the optimum value is @xmath32 = 0.1 @xmath52 for the hse functional and @xmath32 = 0.2 @xmath52 for the lc functional . by choosing the appropriate value of @xmath32 in the hse and lc functionals , we have achieved better agreement in the lattice constants and further substantial improvement in description of elastic constants such as bulk moduli for the ten materials .
dramatic success of the hybrid functionals are observed in the calculated band gaps .
we have found that the calculated band gaps by the lc functional for the wide band - gap materials satisfactorily agree with the experimental values with mean relative error ( mre ) of 3.0 % , whereas the band gaps by the hse functional for the small band - gap materials agree well with the experimental values with mre = 0.7 % .
this good description of the band gaps is unprecedented in density functional theory where lda and gga produce the value of mre @xmath137 40 - 50 % , and is comparable with or better than what the gw approximation produces . the @xmath32 value leading to the best agreement with the experiments is 0.1 @xmath52 for the hse functional and 0.2 @xmath52 for the lc functional .
these @xmath32 values are identical to the optimum values determined from the examination of the structural properties .
the calculated valence - band widths by the hybrid functionals also agree satisfactorily with the experimental values .
it is now established that the hse and lc functionals with appropriate choice of the @xmath32 parameter are useful to describe structural and electronic properties of various materials .
rigorous justification of the choice of the form of the hybrid functionals along with a guiding principle of choice of @xmath32 would offer further developments in the first - principles calculations . the work is partly supported by a grant - in - aid project from mext , japan , scientific research on innovative areas : materials design through computics - complex correlation and non - equilibrium dynamics - " under the contract number 22104005 .
computations were done at supercomputer center , institute for solid state physics , university of tokyo , and at research center for computational science , national institutes of natural sciences .
w. kohn and l. j. sham , phys . rev . *
140 * , a1133 ( 1965 ) .
p. hohenberg and w. kohn , phys . rev . *
136 * , b864 ( 1964 ) . for a review ,
_ theory of the inhomogeneous electron gas _ , edited by s. lundqvist and n. h. march , ( plenum pres ny , 1983 ) .
t. asada and k. terakura , phys . rev .
b 46 , 13 599 ( 1992 ) , and references therein .
a. oshiyama , n. shima , t. nakayama , k. shiraishi and h. kamimura , in _ mechanisms of high temperature superconductivity _ , edited by h. kamimura and a. oshiyama , ( springer - verlag , 1989 ) p111 .
j. p. perdew and s. kurth , in _ a premier in density functional theory _ , edited by c. fiolhais , f. nogueira and m. a. l. marques ( springer , berlin 2003 ) . for a review , s. kmmel and l. kronik , rev .
phys . * 80 * , 3 ( 2008 ) .
j. p. perdew , k. burke and m. ernzerhof , phys .
lett . * 77 * , 3865 ; _ ibid . _ * 78 * , 1396(e ) ( 1997 ) .
j. p. perdew , s. kurth , a. zupan and p. blaha , phsy rev . lett . * 82 * , 2544 ( 1999 ) ; * 82 * , 5179(e ) ( 1999 ) .
t. van voorhis and g. e. scuseria , j. chem . phys . * 109 * , 400 ( 1998 ) .
j. tao , j. p. perdew , v. n. staroverov and g. e. scuseria , phys .
lett . * 91 * , 146401 ( 2003 ) .
v. n. staroverov , g. e. scuseria , j. tao and j. p. perdew , j. chem . phys . * 119 * , 12129 ( 2003 ) ; * 121 * , 11507(e ) ( 2004 ) .
j. p. perdew and a. zunger , phys .
rev . b*23 * , 5048 ( 1981 ) .
a. j. cohen , p. mori - snchez , w. yang , science * 321 * , 721 ( 2008 )
. j. p. perdew , r. g. parr , m. levy and j. l balduz jr .
lett . * 49 * , 1691 ( 1982 ) .
j. p. perdew , m. levy , phys .
lett . * 51 * , 1884 ( 1983 ) .
l. j. sham and m. schlter , phys .
* 51 * , 1888 ( 1983 ) .
l. j. sham and m. schlter , phys .
* , 3883 ( 1985 ) .
j. f. janak , phys .
rev . b*18 * , 7165 ( 1978 ) .
p. mori - snchez , a. j. cohen and w. yang , phys .
lett . * 100 * , 146401 ( 2008 ) .
a. j. cohen , p. mori - snchez and w. yang , phys .
b * 77 * , 115123 ( 2008 ) .
j. d. talman and w. f. shadwick , phys .
a*14 * , 36 ( 1976 ) s. ivanov , s. hirata and r. j. bartlett , phys .
* 83 * , 5455 ( 1999 ) .
a. grling , phys .
lett . * 83 * , 5459 ( 1999 ) .
w. yang and q. wu , phys .
* 89 * , 143002 ( 2002 ) .
a. becke , j. chem . phys . * 98 * , 1372 ( 1993 ) .
a. becke , j. chem . phys . * 98 * , 5648 ( 1993 ) .
see , e.g. , p. j. stephens , f. j. devlin , c. f. chabalowski and m. j. frisch , j. phys .
chem . * 98 * , 11623 ( 1994 ) ; r. h. hertwig and w. koch , chem .
phys . lett . * 268 * , 345 ( 1997 ) ; s. f. sousa , p. a. fernandes and m. j. ramos , j. phys .
a*111 * , 10439 ( 2007 ) .
j. p. perdew , m. ernzerhof and k. burke , j. chem .
phys . * 105 * , 9982 ( 1996 ) .
o. gunnarsson and b. i. lundqvist , phys .
b*13 * , 4274 ( 1976 ) .
c. adamo and v. barone , j. chem .
phys . * 110 * , 6158 ( 1999 ) .
d. m. bylander and l. kleinman , phys .
rev . b*41 * , 7868 ( 1990 ) .
j. heyd , g. e. scuseria , and m. ernzerhof , j. chem . phys . * 118 * , 8207 ( 2003 ) .
j. paier , m. marsman , k. hummer , g. kresse , i. c. gerber and j. g. ngyn , j. chem . phys .
* 124 * , 154709 ( 2006 ) .
m. marsman , j. paier , a. stroppa , and g. kresse , j. phys . : condens .
matter * 20 * , 064201 ( 2008 ) .
e. r. batista , j. heyd , r. g. hennig , b. p. uberuaga , r. l. martin , g. e. scuseria , c. j. umrigar and j. w. wilkins , phys .
b * 74 * , 121102 ( r ) ( 2006 ) .
f. fuchs , j. furthmller , f. bechstedt , m. shishkin and g. kresse , phys .
b*76 * , 115109 ( 2007 ) .
a. stroppa and g. kresse , new j. phys .
* 10 * , 063020 ( 2008 ) .
h. leininger , h. stoll , h .- j .
werner and a. savin , chem .
. lett . * 275 * , 151 ( 1997 ) .
a. savin , _ recent developmennts and applications of modern density functional theory _
( elsevier , 1996 ) pp327 - 357 .
h. iikura , t. tsuneda , t. yanai , and k. hirao , j. chem . phys .
* 115 * , 3540 ( 2001 ) .
m. kamiya , t. tsuneda , and k. hirao , j. chem . phys .
* 117 * , 6010 ( 2002 ) .
y. tawada , t. tsuneda , s. yanagisawa , t. yanai and k. hirao , j. chem .
* 120 * , 8425 ( 2004 ) .
m. chiba , t. tsuneda , and k. hirao , j. chem .
phys . * 124 * , 144106 ( 2006 ) .
h. sekino , y. maeda , m. kamiya , and k. hirao , j. chem . phys .
* 126 * , 014107 ( 2007 ) .
o. a. vydrov , j. heyd , a. v. krukau and g. e. scuseria , j. chem .
125 * , 074106 ( 2006 ) .
o. a. vydrov and g. e. scuseria , j. chem
125 * , 234109 ( 2006 ) .
i. c. gerber , j. g. angyan , m. marsman and g. kresse , j. chem .
phys . * 127 * , 054101 ( 2007 ) .
tokyo ab - initio program package ( tapp ) has been developed by a consortium initiated at the university of tokyo .
o. sugino and a. oshiyama , phys .
rev . lett . * 68 * , 1858 ( 1992 ) .
j. yamauchi , m. tsukada , s. watanabe and o. sugino , phys .
b * 54 * , 5586 ( 1996 ) .
h. kageshima and k. shiraishi , phys .
b * 56 * , 14985 ( 1997 ) .
j. heyd and g. e. scuseria , j. chem . phys .
* 120 * , 7274 ( 2004 ) .
j. heyd and g. e. scuseria , j. chem .
phys . * 121 * , 1187 ( 2004 ) .
j. heyd , g. e. scuseria and m. ernzerhof , j. chem . phys . * 124 * , 219906 ( 2006 ) . n. troullier and j. l. martins , phys . rev .
b * 43 * , 1993 ( 1991 ) .
d. vanderbilt , phys .
b * 41 * , 7892 ( 1990 ) . f. gygi and a. baldereschi , phtys . rev .
b * 34 * , 4405 ( 1986 ) .
b. wenzien , g. cappellini and f. bechstedt , phys .
b * 51 * , 14701 ( 1995 ) . p. carrier and g. a. voth , j. chem .
phys . * 108 * , 4697 ( 1998 ) .
j. spencer and a. alavi , phys . rev .
b * 77 * , 193110 ( 2008 ) .
m. ernzerhof and j. p. perdew , j. chem .
* 109 * , 3313 ( 1998 ) j. heyd , thesis , rice university , 2004 .
s. g. louie , s. froyen and m. l. cohen , phys .
rev . b*26 * , 1738 ( 1982 ) .
m. s. hybertsen and s. g. louie , phys .
b * 34 * , 5390 ( 1986 ) .
f. aryasetiawan and o. gunnarsson , rep .
phys . * 61 * , 237 ( 1998 ) . for discussion on several ambiguities in the gw approximation ,
see m. l. tiago , s. ismail - beigi and s. g. louie , phys .
b * 69 * , 125212 ( 2004 ) ; m. van schilfgaarde , t. kotani and s.v .
faleev , phys .
b * 74 * , 245125 ( 2006 ) , and references therein .
landolt - bornstein , vol .
iii , ( springer , new york , 1982 ) j. heyd , j. e. peralta , g. e. scuseria , and r. l. martin , j. chem
123 * , 174101 ( 2005 ) .
s. h. wemple , phys .
b * 2 * , 2679 ( 1970 ) .
s. adachi , optical properties of crystalline and amorphous semiconductors : numerical data and graphical information ( kluwer academic , dordrecht , 1999 ) .
r. t. poole , j. liesegang , r. c. g. leckey , and j. g. jenkin , phys .
b * 11 * , 5190 ( 1975 ) .
r. j. magyar , a. fleszar and e. k. u. gross , phys .
b * 69 * , 045111 ( 2004 ) .
m. rohlfing and s. g. louie , phys .
b * 62 * , 4927 ( 2000 ) .
a. l. wachs , t. miller , t. c. hsieh , a. p. shapiro , t. c. chiang , phys .
b * 32 * , 2326 ( 1985 ) .
f. r. mcfeely et al .
b * 9 * , 5268 ( 1974 ) .
f. j. himpsel , j. f. van der veen and d. e. eastman , phys .
b * 22 * , 1967 ( 1980 ) .
s. piskunov , , a , e. heifetsb , r. i. eglitisa and g. borstel , compt . mater .
* 29 * , 165 ( 2004 ) .
k.h . hellwege and a.m. hellwege , editors , ferroelectrics and related substancesnew series vol .
3 , landolt - bornstein , springer verlag , berlin ( 1969 ) group iii .
j. heyd and g. e. scuseria , j. chem .
phys . * 121 * , 1187 ( 2004 ) .
p. varotsos , k. alexopoulos , phys .
b * 15 * , 4111 ( 1977 ) .
a. r. jivani , p. n. gajjar , and a.r .
jani , semicond .
phys , quantum .
bf 5 , 243 ( 2002 ) . | we present a systematic study that clarifies validity and limitation of current hybrid functionals in density functional theory for structural and electronic properties of various semiconductors and insulators .
the three hybrid functionals , pbe0 by perdew , ernzerhof , and becke , hse by heyd , sucseria , and ernzerhof , and a long - range corrected ( lc ) functional , are implemented in a well - established plane - wave - basis - set scheme combined with norm - conserving pseudopotentials , thus enabling us to assess applicability of each functional on equal footing to the properties of the materials .
the materials we have examined in this paper range from covalent to ionic materials as well as a rare - gas solid whose energy gaps determined by experiments are in the range of 0.6 ev - 14.2 ev : i.e. , ge , si , batio@xmath0 , @xmath1-gan , diamond , mgo , nacl , licl , kr , and lif . we find that the calculated bulk moduli by the hybrid functionals show better agreement with the experiments than the generalized gradient approximation ( gga ) provides , whereas the calculated lattice constants by the hybrid functionals and gga show comparable accuracy .
the calculated energy band gaps and the valence - band widths for the ten prototype materials show substantial improvement using the hybrid functional compared with gga .
in particular , it is found that the band gaps of the ionic materials as well as the rare - gas solid are well reproduced by the lc - hybrid functional , whereas those of covalent materials are well described by the hse functional .
we also examine exchange effects due to short - range and long - range components of the coulomb interaction and propose an optimum recipe to the short - range and long - range separation in treating the exchange energy .
= 0.0em |
let @xmath3 be the @xmath0-cube graph , with vertex set @xmath4 and two vertices joined if they differ in one component .
in the language of association schemes , @xmath3 is the distance 1 graph of the binary hamming scheme .
it is of interest to compute linear algebraic invariants of a graph , such as its eigenvalues and the invariant factors of an adjacency matrix or laplacian matrix . in the case of @xmath3 ,
previous work includes @xcite and @xcite , where many of these invariants have been computed and some conjectures made about others .
here we shall consider the smith group .
if @xmath5 is an @xmath6 integral matrix , then the _ smith group _ of @xmath5 is the abelian group defined as the quotient of @xmath7 by the subgroup spanned by the columns @xmath5 ; that is , the abelian group whose invariant factor decomposition is given by the smith normal form of @xmath5 . if @xmath8 is the adjacency matrix ( with respect to any ordering of the vertices ) of a graph , then the smith group of the graph is defined as the smith group of @xmath8 , and does not depend on the ordering on the vertices .
we recall that two integral matrices @xmath5 and @xmath9 are _ integrally equivalent _ if there exist unimodular integral matrices @xmath10 and @xmath11 such that @xmath12 as is well known , @xmath5 and @xmath9 are integrally equivalent if and only if @xmath9 can be obtained from @xmath5 by a finite sequence of integral unimodular row and column operations . a _ diagonal form _ for @xmath5 is a matrix integrally equivalent to @xmath5 that has nonzero entries only on the leading diagonal .
the smith normal form of @xmath5 is one such diagonal form .
another way to describe the smith group is in terms of the @xmath13-elementary divisors of @xmath5 with respect to primes @xmath13 .
any diagonal form for @xmath5 gives a cyclic decomposition of the smith group , so in a certain sense , the various diagonal forms carry the same information as the list of @xmath13-elementary divisors as @xmath13 varies over all primes .
the notion of integral equivalence can be generalized to @xmath14-equivalence , where @xmath14 is the ring of @xmath13-local integers , by requiring that the matrices @xmath10 and @xmath11 appearing in ( [ uxv ] ) be invertible over @xmath14 .
we can also consider the @xmath13-elementary divisors of any matrix @xmath5 with entries in @xmath14 .
if @xmath5 happens to have integer entries then its @xmath13-elementary divisors are the same whether it is considered as a matrix over @xmath15 or @xmath14 let @xmath8 be an adjacency matrix for @xmath3 .
it was proved in @xcite that for every odd prime @xmath13 , @xmath8 is @xmath14-equivalent to the diagonal matrix of the eigenvalues ( all of which are integers ) .
when @xmath0 is odd , all the eigenvalues are odd integers , so @xmath16 is an odd integer .
thus , the eigenvalue matrix is a diagonal form when @xmath0 is odd .
when @xmath0 is even , there remains the problem of finding the @xmath17-elementary divisors of @xmath8 .
a conjecture for the multiplicity of each power @xmath18 as a @xmath17-elementary divisor was stated in @xcite .
( see conjecture [ conjdj ] below . )
the purpose of this paper is to give a proof of this conjecture . as a consequence of the conjecture , we obtain the following description of the smith group of @xmath3 when @xmath0 is even . [ qdiag ]
suppose that @xmath19 is even .
then the adjacency matrix @xmath8 of the @xmath0-cube @xmath3 is integrally equivalent to a diagonal matrix with @xmath20 diagonal entries equal to zero and whose nonzero diagonal entries are the integers @xmath21,@xmath17, @xmath22 , in which the multiplicity of @xmath23 is @xmath24 .
let @xmath0 be a positive integer and @xmath25 . for brevity , we shall use the term _
@xmath23-subsets _ for the subsets of @xmath5 of size @xmath23 . for @xmath26 , let @xmath27 denote the free @xmath15-module on the set of @xmath23-subsets and for @xmath28 , @xmath26 let @xmath29 be the incidence map , induced by inclusion . thus , if @xmath30 a @xmath28-subset is mapped to the sum of all @xmath23-subsets containing it , while if @xmath31 the image of a @xmath28-subset is the sum of all @xmath23-subsets which it contains . for each @xmath26 , if we fix ordering on the @xmath23-subsets , we can think of elements of @xmath27 as row vectors .
let @xmath32 denote the @xmath33 matrix of @xmath34 with respect to these ordered bases of @xmath35 and @xmath27 .
the notion of the _ rank _ of a subset was introduced by frankl @xcite .
we shall only need the concept of a @xmath28-subset of rank @xmath28 , for @xmath36 .
let @xmath37 , with the elements in increasing order .
then @xmath38 has rank @xmath28 if and only if @xmath39 for all @xmath40, ,@xmath28 .
a @xmath28-subset is of rank @xmath28 if an only if it is the set of entries in the second row of a standard young tableau of shape @xmath41 $ ] .
this is one way to see that the number of @xmath28-subsets of rank @xmath28 is @xmath42 .
assume @xmath43 .
let @xmath44 denote the @xmath45\times \tbinom{n}{k}$ ] submatrix of @xmath46 formed from the the rows labeled by @xmath47-subsets of rank @xmath47 , and let @xmath48 be the @xmath49 matrix formed by stacking the @xmath44 , @xmath50 , with @xmath47 increasing as we move down the matrix @xmath48 .
wilson @xcite found a diagonal form for @xmath32 .
we shall state his result for @xmath51 .
there exist unimodular matrices @xmath52 and @xmath53 such that @xmath54 where the diagonal form @xmath55 has diagonal entries @xmath56 , with multiplicity @xmath57 , for @xmath58 , , @xmath28 .
bier @xcite refined wilson s results , showing that we can take @xmath59 for all @xmath23 and @xmath60 for all @xmath28 .
the additional uniformity will be important for us .
[ bier]@xcite assume @xmath61
. then the matrix @xmath48 is unimodular .
furthermore , for @xmath30 , we have @xmath62 where @xmath55 is wilson s diagonal form
. we shall refer to the basis of @xmath27 corresponding to the rows of @xmath48 as the _ canonical basis _ of @xmath27 .
it consists of all vectors of the form @xmath63 , where @xmath50 and @xmath64 is a @xmath47-subset of rank @xmath47 .
let @xmath3 denote the @xmath0-cube graph .
the vertex set of @xmath3 is @xmath4 and @xmath65 is adjacent to @xmath66 if and only if there is exactly one index @xmath47 with @xmath67 .
there is clearly a bijection of the vertex set with the set of subsets of @xmath68 , under which a vertex corresponds to the subset of indices where the vertex has entry @xmath2 .
we use this bijection and our fixed ordering of @xmath23-subsets for @xmath26 to order the vertices of @xmath3 , taking the subsets in order of increasing size .
let @xmath8 denote the adjacency matrix , with respect to this ordering .
next we review the results of @xcite . by viewing the vertex set of @xmath3 as @xmath69 , and transforming @xmath8 by the character table of the additive group @xmath69 the eigenvalues of @xmath3
are found to be @xmath70 , with multiplicity @xmath71 for @xmath72 .
when @xmath0 is odd , we see that @xmath73 is odd ; hence all elementary divisors are odd .
then since @xmath69 is an abelian @xmath17-group , the same discrete fourier transform method yields the elementary divisors .
when @xmath0 is even , one still obtains the @xmath13-elementary divisors for all odd @xmath13 .
the @xmath17-elementary divisors were not computed in @xcite , but the following conjecture was stated : [ conjdj]@xcite suppose @xmath0 is even .
then the multiplicity of @xmath74 as a @xmath17-elementary divisor of @xmath8 is equal to the number of eigenvalues of @xmath8 whose exact @xmath17-power divisor is @xmath75 .
let @xmath76 denote the free @xmath15-module on the set of vertices of @xmath3 .
the matrix @xmath8 can be viewed as an endomorphism of @xmath76 , sending a vertex to the sum of all adjacent vertices .
it is important for us to adopt a slightly different point of view .
we can think of @xmath76 as the ring of @xmath15-valued functions on the set of vertices of @xmath3 .
then the matrix @xmath8 defines the map @xmath77 such that for any function @xmath78 we have @xmath79 for @xmath80 .
if we further regard the set @xmath4 of vertices of @xmath3 as a subset of @xmath81 , then functions are restrictions of polynomials and we have a ring isomorphism of @xmath76 with @xmath82/({x_i}^2-x_i,\ , 1\leq i\leq n).\ ] ] now a different natural basis becomes evident , namely the set of monomials @xmath83 , for @xmath84 .
with respect to the monomial basis we have @xmath85 therefore , if we order monomials in the same way as we ordered subsets , the matrix of @xmath77 with respect to this basis has the form @xmath86 ni&w_{0,1}&0&0&0&\hdots&0&0\cr\- 0&(n-2)i&w_{1,2}&0&0&\hdots&0&0\cr\- 0&0&(n-4)i&w_{2,3}&0&\hdots&0&0\cr\- 0&0&0&(n-6)i&w_{3,4}&\hdots&0&0\cr\- \vdots&\vdots&\ddots&\ddots&\ddots&\ddots&\vdots&\vdots\cr\- 0&0&\hdots&0&0&-(n-4)i&w_{n-2,n-1}&0\cr\- 0&0&\hdots&0&0&0&-(n-2)i&w_{n-1,n}\cr\- 0&0&\hdots&0&0&0&0&-ni\cr \end{pmat}.\ ] ] @xmath88 ni&w_{0,1}&0&0&0&\hdots&0&0\cr\- 0&(n-2)i&w_{1,2}&0&0&\hdots&0&0\cr\- 0&0&(n-4)i&w_{2,3}&0&\hdots&0&0\cr\- 0&0&0&(n-6)i&w_{3,4}&\hdots&0&0\cr\- \vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\vdots&\vdots\cr\- 0&\hdots&\hdots&\hdots&\ddots&4i&w_{m-2,m-1}&0\cr\- 0&\hdots&\hdots&\hdots&\hdots&0&2i&w_{m-1,m}\cr \end{pmat}.\ ] ] @xmath89 w_{m , m+1}&0&\hdots&0&0&0&0\cr\- -2i&w_{m+1,m+2}&\hdots&0&0&0&0\cr\- 0&-4i&\ddots&\vdots&\vdots&\vdots&\vdots\cr\- \vdots&\ddots&\ddots&w_{n-4,n-3}&0&0&0\cr\- \vdots&\vdots&\ddots&-(n-6)i&w_{n-3,n-2}&0&0\cr\- \vdots&\vdots&\hdots&0&(-(n-4)i&w_{n-2,n-1}&0\cr\- \vdots&\vdots&\hdots&0&0&-(n-2)i&w_{n-1,n}\cr\- 0&0&\hdots&0&0&0&-ni\cr \end{pmat}. \ ] ] up to now the choice of orderings on the @xmath47-susbsets , @xmath93 used in the definition of the inclusion matrices @xmath32 has been an arbitrary ( but fixed ) one .
any choice would result in matrices of the form @xmath91 and @xmath92 as above , but the rows and columns of the submatrices @xmath32 would be permuted .
now we shall specify these orderings more carefully .
the matrix @xmath91 involves only the matrices @xmath32 with @xmath94 , while the matrix @xmath92 involves only the matrices @xmath32 with @xmath95 .
we start from the arbitrary but fixed ordering on the @xmath47-subsets with @xmath96 that led to the matrix @xmath91 .
then for @xmath97 we choose the ordering of @xmath98-subsets to be the order induced by the complementation map . in this way we have specified an ordering on the @xmath47-subsets , for all @xmath47 .
finally we wish to consider a _
second _ ordering on @xmath22-subsets , namely , the ordering defined from the given ordering by complementation .
we use the first ordering on @xmath22-sets to define the submatrix @xmath99 of @xmath91 and the second ordering to define the submatrix @xmath100 of @xmath92 . from the block form ( [ mnmatrix ] )
, we see that the matrix @xmath90 thus constructed differs from the one in which the same ordering on @xmath22-sets is used for both @xmath99 and @xmath100 only by a permutation of the rows in the first row - block of @xmath92 , so the two matrices would be integrally equivalent .
@xmath102 -ni&w^t_{n-1,n}&0&0&0&\hdots&0&0\cr\- 0&-(n-2)i&w^t_{n-2,n-1}&0&0&\hdots&0&0\cr\- 0&0&-(n-4)i&w^t_{n-3,n-2}&0&\hdots&0&0\cr\- 0&0&0&-(n-6)i&w^t_{n-4,n-3}&\hdots&0&0\cr\- \vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\vdots&\vdots\cr\- 0&\hdots&\hdots&\hdots&\ddots&-4i&w^t_{m+1,m+2}&0\cr\- 0&\hdots&\hdots&\hdots&\hdots&0&-2i&w^t_{m , m+1}\cr \end{pmat}\\ & = \begin{pmat}[{||||||| } ] -ni&w_{0,1}&0&0&0&\hdots&0&0\cr\- 0&-(n-2)i&w_{1,2}&0&0&\hdots&0&0\cr\- 0&0&-(n-4)i&w_{2,3}&0&\hdots&0&0\cr\- 0&0&0&-(n-6)i&w_{3,4}&\hdots&0&0\cr\- \vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\vdots&\vdots\cr\- 0&\hdots&\hdots&\hdots&\ddots&-4i&{w_{m-2,m-1}}&0\cr\- 0&\hdots&\hdots&\hdots&\hdots&0&-2i&w_{m-1,m}\cr \end{pmat}. \end{aligned}\ ] ] thus , @xmath103 differs from @xmath91 only by the sign of the diagonal entries .
in fact , to see that @xmath103 is integrally equivalent to @xmath91 , we perform the following simple sequence of unimodular operations . first mutliply the first block - column by -1 , then multiply the second block - row by -1 , then the third block - column , etc . , until we reach the bottom - right of the matrix , at which point @xmath103 has been converted to @xmath91 .
[ half ] let @xmath91 and @xmath92 be the matrices in ( [ mnmatrix ] ) . then @xmath91 and @xmath104 are integrally equivalent . in particular the multiplicity of an elementary divisor of @xmath90 ( and
hence of @xmath8 ) is twice its multplicity as an elementary divisor of @xmath91 .
@xmath108 ni&d_{0,1}&0&0&0&\hdots&0&0\cr\- 0&(n-2)i&d_{1,2}&0&0&\hdots&0&0\cr\- 0&0&(n-4)i&d_{2,3}&0&\hdots&0&0\cr\- 0&0&0&(n-6)i&d_{3,4}&\hdots&0&0\cr\- \vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\vdots&\vdots\cr\- 0&\hdots&\hdots&\hdots&\ddots&4i&d_{m-2,m-1}&0\cr\- 0&\hdots&\hdots&\hdots&\hdots&0&2i&d_{m-1,m}\cr \end{pmat}.\ ] ] @xmath112 1&0&0&0&0&0&0&0&0&0&0\cr\- 0&1&1&1&1&0&0&0&0&0&0\cr 0&0&1&0&0&0&0&0&0&0&0\cr 0&0&0&1&0&0&0&0&0&0&0\cr 0&0&0&0&1&0&0&0&0&0&0\cr\- 0&0&0&0&0&1&1&1&1&1&1\cr 0&0&0&0&0&1&0&1&0&1&0\cr 0&0&0&0&0&0&1&1&0&0&1\cr 0&0&0&0&0&0&0&0&1&1&1\cr 0&0&0&0&0&0&0&0&0&1&0\cr 0&0&0&0&0&0&0&0&0&0&1\cr \end{pmat}\ ] ] @xmath116 0&d_{0,1}&0&0&0&0&\hdots\cr\- 0&0&d_{1,2}&0&0&0&\hdots\cr\- 0&0&0&d_{2,3}&0&0&\hdots\cr\- 0&0&0&0&d_{3,4}&0&\hdots\cr\- \vdots&\vdots&\vdots&\vdots&\ddots&\ddots&\ddots\cr 0&\hdots&0&0&0&0&d_{m-1,m}\cr \end{pmat}.\ ] ] @xmath120 \mathbf{i}&\mathbf{0}&\mathbf{0}&\hdots&\mathbf{0}&\mathbf{0}&\mathbf{0}\cr\- \mathbf{0}&\mathbf{i-1}&\mathbf{0}&\hdots&\mathbf{0}&\mathbf{0}&\mathbf{0}\cr\- \mathbf{0}&\mathbf{0}&\mathbf{i-2}&\hdots&\mathbf{0}&\mathbf{0}&\mathbf{0}\cr\- \vdots&\vdots&\vdots&\ddots&\vdots&\vdots&\vdots\cr\- \mathbf{0}&\mathbf{0}&\mathbf{0}&\hdots&\mathbf{2}&\mathbf{0}&\mathbf{0}\cr\- \mathbf{0}&\mathbf{0}&\mathbf{0}&\hdots&\mathbf{0}&\mathbf{1}&\mathbf{0}\cr \end{pmat},\ ] ] where a bold number @xmath121 , @xmath122 represents a scalar matrix @xmath123 of the appropriate size and @xmath124 denotes a zero block of the appropriate size .
the block sizes are readily found ; if @xmath125 , then @xmath117 has @xmath126 block - columns and @xmath127 block - rows , and the @xmath23-th block - column of @xmath117 contains @xmath128 columns , for @xmath129 , while for @xmath130 , the @xmath131-th block - row contains @xmath132 rows .
also , let @xmath133 denote a scalar matrix ( more precisely the class of scalar matrices ) whose scalar is a multiple of @xmath134 .
( we introduce this notation because all that we shall use about the diagonal entries is that they are even , and working with the entire class of such matrices will facilitate the use of mathematical induction . )
then we can write @xmath114 in the following form ( more precisely , @xmath114 lies in the given class of matrices )
. @xmath135 { \mathbf{(2)}}&{\mathbf{1}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}\cr\- { \mathbf{0}}&{\mathbf{(2)}}&{\mathbf{0}}&{\mathbf{2}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}\cr { \mathbf{0}}&{\mathbf{0}}&{\mathbf{(2)}}&{\mathbf{0}}&{\mathbf{1}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}\cr\- { \mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{(2)}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{3}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}\cr { \mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{(2)}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{2}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}\cr { \mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{(2)}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{1}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}\cr\- \vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\ddots&\ddots&\ddots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots\cr \vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\ddots&\ddots&\ddots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots\cr \vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\ddots&\ddots&\ddots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots\cr\- \vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\ddots&\ddots&\ddots&\ddots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots\cr \vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\ddots&\ddots&\ddots&\ddots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots\cr\- { \mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{(2)}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{m}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}\cr { \mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{(2)}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{m-1}}&\hdots&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}\cr { \mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&\vdots&\vdots&\ddots&\vdots&\vdots&\vdots&\vdots&\ddots&\vdots&\vdots&\vdots\cr { \mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{(2)}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{2}}&{\mathbf{0}}&{\mathbf{0}}\cr { \mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&\hdots&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{(2)}}&{\mathbf{0}}&{\mathbf{0}}&\hdots&{\mathbf{0}}&{\mathbf{1}}&{\mathbf{0}}\cr \end{pmat}.\ ] ] let @xmath137 denote a matrix ( actually the class of matrices ) in which all the blocks ( bold numbers ) in @xmath114 are replaced by the plain numbers , and each @xmath138 is replaced by @xmath139 , representing a multiple of the @xmath17-local integer @xmath134 .
it suffices to show that any matrix of the form @xmath140 is @xmath17-locally equivalent to the matrix obtained by zeroing out its diagonal .
indeed , if there is a sequence of @xmath136-unimodular row and column operations which zero out the diagonal of @xmath137 , then the same sequence of the blockwise versions of these operations will kill the diagonal of @xmath114 .
so we may discard @xmath114 and work only with @xmath137 from now on .
@xmath141 ( 2)i&\overline d_1&0&0&0&\hdots&0&0\cr\- 0&(2)i&\overline d_2&0&0&\hdots&0&0\cr\- 0&0&(2)i&\overline d_3&0&\hdots&0&0\cr\- 0&0&0&(2)i&\overline d_4&\hdots&0&0\cr\- \vdots&\vdots&\vdots&\ddots&\ddots&\ddots&\vdots&\vdots\cr\- 0&\hdots&\hdots&\hdots&\ddots&(2)i&\overline d_{m-1}&0\cr\- 0&\hdots&\hdots&\hdots&\hdots&0&(2)i&\overline d_{m}\cr \end{pmat},\ ] ] where @xmath142 { i}&{0}&{0}&\hdots&{0}&{0}&{0}\cr { 0}&{i-1}&{0}&\hdots&{0}&{0}&{0}\cr { 0}&{0}&{i-2}&\hdots&{0}&{0}&{0}\cr \vdots&\vdots&\vdots&\ddots&\vdots&\vdots&\vdots\cr { 0}&{0}&{0}&\hdots&{2}&{0}&{0}\cr { 0}&{0}&{0}&\hdots&{0}&{1}&{0}\cr \end{pmat}\ ] ] is the `` condensed '' version of @xmath117 , consisting of a diagonal @xmath143 matrix augmented by a column of zeros . from now on it will be convenient to refer to the entries of @xmath137 ( and matrices derived from it ) by their positions relative to the submatrices @xmath144 .
note this is a change in the way we are indexing blocks , compared to how we did it in the original matrix @xmath91 .
the row of the main matrix containing the @xmath23-th row of @xmath144 is assigned the label @xmath145 $ ] and the column containing the @xmath131-th column of @xmath146 is assigned the label @xmath147 $ ] , while the first column is labeled @xmath148 $ ] .
thus , in a row index @xmath145 $ ] we have @xmath149 and @xmath150 , while a column index @xmath147 $ ] has @xmath96 and @xmath151 .
we shall perform some row and column operations using the odd entries on the main diagonals of the @xmath144s to kill diagonal entries of the main matrix .
each odd entry of @xmath144 for @xmath152 will be used to kill the two diagonal entries of the main matrix , one in the same row as the odd entry and one in the same column .
the odd entries of @xmath153 will be used to kill the diagonal entry in the same row .
this procedure will create new entries @xmath154 at locations where there previously were zeroes . to be precise ,
if the odd entry is at @xmath155,[i , k])$ ] then the diagonal entry of the main matrix in the same row is at @xmath155,[i-1,k])$ ] .
the diagonal entry of the main matrix in the same column as @xmath155,[i , k])$ ] is at @xmath156,[i , k])$ ] ( with no such entry if @xmath157 ) . by multiplying column @xmath145 $ ] by a suitable element of @xmath158 and subtracting it from column @xmath159 $ ] , we kill off the entry at @xmath155,[i-1,k])$ ] and create a new entry @xmath154 at @xmath156,[i-1,k)$ ] , if @xmath160 . the new entry is @xmath154 because the entry being subtracted from zero is a @xmath158-multiple of the @xmath158 at @xmath156,[i , k])$ ] on the main diagonal .
no new entry is created if @xmath157 .
then , we can subtract a @xmath158-multiple of row @xmath145 $ ] from row @xmath161 $ ] to kill off the diagonal entry at @xmath156,[i , k])$ ] , without creating any new nonzero entries , if @xmath160 , while there is nothing to be done when @xmath157 .
the following figure ( for @xmath162 ) shows the resulting matrix after performing these operations with the entry 1 at @xmath163,[3,3])$ ] .
the block indices are specified by numbered braces .
@xmath164 ( 2)&1&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0\cr\- 0&(2)&0&2&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0\cr 0&0&(2)&0&1&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0\cr\- 0&0&0&(2)&0&0&3&0&0&0&0&0&0&0&0&0&0&0&0&0&0\cr 0&0&0&0&(2)&0&0&2&0&0&0&0&0&0&0&0&0&0&0&0&0\cr 0&0&0&0&0&0&0&0&1&0&0&0&0&0&0&0&0&0&0&0&0\cr\- 0&0&0&0&0&0&(2)&0&0&0&4&0&0&0&0&0&0&0&0&0&0\cr 0&0&0&0&0&0&0&(2)&0&0&0&3&0&0&0&0&0&0&0&0&0\cr 0&0&0&0&0&(4)&0&0&0&0&0&0&2&0&0&0&0&0&0&0&0\cr 0&0&0&0&0&0&0&0&0&(2)&0&0&0&1&0&0&0&0&0&0&0\cr\- 0&0&0&0&0&0&0&0&0&0&(2)&0&0&0&0&5&0&0&0&0&0\cr 0&0&0&0&0&0&0&0&0&0&0&(2)&0&0&0&0&4&0&0&0&0\cr 0&0&0&0&0&0&0&0&0&0&0&0&(2)&0&0&0&0&3&0&0&0\cr
0&0&0&0&0&0&0&0&0&0&0&0&0&(2)&0&0&0&0&2&0&0\cr 0&0&0&0&0&0&0&0&0&0&0&0&0&0&(2)&0&0&0&0&1&0\cr \end{pmat } \end{array}\ ] ] we do this for all odd entries of all the @xmath144 . in our @xmath162 example , after the operations the matrix looks as follows .
@xmath165 0&1&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0\cr\- ( 4)&0&0&2&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0\cr 0&0&0&0&1&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0\cr\- 0&0&0&0&0&0&3&0&0&0&0&0&0&0&0&0&0&0&0&0&0\cr 0&0&(4)&0&0&0&0&2&0&0&0&0&0&0&0&0&0&0&0&0&0\cr 0&0&0&0&0&0&0&0&1&0&0&0&0&0&0&0&0&0&0&0&0\cr\- 0&0&0&(4)&0&0&0&0&0&0&4&0&0&0&0&0&0&0&0&0&0\cr 0&0&0&0&0&0&0&0&0&0&0&3&0&0&0&0&0&0&0&0&0\cr 0&0&0&0&0&(4)&0&0&0&0&0&0&2&0&0&0&0&0&0&0&0\cr 0&0&0&0&0&0&0&0&0&0&0&0&0&1&0&0&0&0&0&0&0\cr\- 0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&5&0&0&0&0&0\cr 0&0&0&0&0&0&0&(4)&0&0&0&0&0&0&0&0&4&0&0&0&0\cr 0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&3&0&0&0\cr 0&0&0&0&0&0&0&0&0&(4)&0&0&0&0&0&0&0&0&2&0&0\cr 0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&0&1&0\cr \end{pmat}\ ] ] at this point , the rows and columns of the main matrix that correspond to an odd entry of some @xmath144 have no other nonzero entries .
the submatrix @xmath166 formed from these rows and columns is a square diagonal matrix with odd entries , and the main matrix is the block sum of this diagonal matrix with the submatrix @xmath167 formed from the remaining rows and columns .
( since permuting rows and columns results in an integrally equivalent matrix , the reader may find it helpful , for easier visualization of this block sum decomposition , to renumber the rows and columns in order to put @xmath166 at the bottom right of the main matrix . )
we observe that @xmath167 will have entries @xmath154 all along its main diagonal , and each row of @xmath167 has one nonzero entry off the main diagonal ( coming from the even entries on the main diagonals of the @xmath144 . ) in the picture below , we show @xmath168 .
@xmath169 ( 4)&0&2&0&0&0&0&0&0&0&0&0\cr\- 0&(4)&0&0&2&0&0&0&0&0&0&0\cr\- 0&0&(4)&0&0&0&4&0&0&0&0&0\cr 0&0&0&(4)&0&0&0&2&0&0&0&0\cr\- 0&0&0&0&(4)&0&0&0&0&4&0&0\cr 0&0&0&0&0&(4)&0&0&0&0&2&0\cr \end{pmat}\ ] ] we next note that if @xmath127 and @xmath47 have different parity then all entries at locations @xmath155,[j,\ell])$ ] are zero , so @xmath167 is the block sum of the submatrix @xmath170 corresponding to rows @xmath145 $ ] and columns @xmath147 $ ] for odd indices @xmath127 and @xmath47 , and the submatrix @xmath171 formed from the rows and columns corresponding to the even indices . for better visualization , we can reorder the rows and columns .
the block decomposition of @xmath168 after reording rows and columns is shown below .
@xmath172 ( 4)&2&0&0&0&0&0&0&0&0&0&0\cr\- 0&(4)&0&4&0&0&0&0&0&0&0&0\cr 0&0&(4)&0&2&0&0&0&0&0&0&0\cr\- 0&0&0&0&0&0&(4)&2&0&0&0&0\cr\- 0&0&0&0&0&0&0&(4)&0&4&0&0\cr 0&0&0&0&0&0&0&0&(4)&0&2&0\cr \end{pmat}\ ] ] we can see from this picture that @xmath173 . in general , it is easy to see that if we delete the rows and columns of @xmath144 which contain odd entries , the resulting matrix is @xmath174 where @xmath175 ( with the understanding that @xmath176 is the empty matrix ) .
it follows that @xmath177 , where @xmath178 , and @xmath179 , and @xmath180 .
( more formally , @xmath171 belongs to the class @xmath181 and @xmath170 belongs to the class @xmath182 . )
thus , we have shown that @xmath137 is @xmath17-locally equivalent to a matrix @xmath183 which is the diagonal block sum of @xmath184 and @xmath185 and the diagonal matrix @xmath166 .
arguing by induction on @xmath22 , there are @xmath136-unimodular row and column operations on @xmath183 that kill the diagonals of @xmath171 and of @xmath170 while leaving all other entries of @xmath183 unchanged .
the resulting matrix is equal , up to reordering rows and columns , to the matrix obtained from @xmath137 by zeroing out its diagonal .
[ adiag ] suppose that @xmath19 is even .
then the adjacency matrix @xmath8 of the @xmath0-cube @xmath3 is @xmath136-equivalent to a diagonal form with @xmath20 diagonal entries equal to zero and whose nonzero diagonal entries are the integers @xmath21,@xmath17, @xmath22 , in which the multiplicity of @xmath23 is @xmath24 . by lemma [ half ]
, it suffices to find a diagonal form for @xmath91 , hence for @xmath115 , with which we have shown @xmath91 to be integrally equivalent .
the nonzero entries of @xmath115 are the integers @xmath2 to @xmath22 .
the integer @xmath23 occurs in @xmath117 only when @xmath186 and then its multiplicity in @xmath117 is @xmath187 .
therefore , the multiplicity of @xmath23 in @xmath115 is @xmath188 by lemma [ half ] , the multiplicity of the prime power @xmath189 as an @xmath13-elementary divisor of @xmath90 , and hence also of @xmath8 , is twice the sum of the binomial coefficients @xmath190 , taken over those @xmath23 with @xmath191 that are exactly divisible by @xmath189 . on the other hand , we know ( [ ncube ] ) that the nonzero eigenvalues of @xmath8 are the integers @xmath192 for @xmath191 , and the multiplicity of @xmath193 is @xmath190 . comparing these numbers for @xmath194 , we see that conjecture [ conjdj ] is true . by theorem [ adiag ]
, it suffices to show , for every odd prime @xmath13 , that @xmath8 is @xmath14-equivalent to a diagonal matrix whose nonzero entries are @xmath21, ,@xmath22 , where @xmath23 has multiplicity @xmath24 .
let @xmath13 be given .
we know from @xcite that @xmath8 is @xmath14-equivalent to a diagonal matrix whose nonzero are @xmath70 with multiplicity @xmath71 , for @xmath195 . the latter is easily seen to be integrally equivalent to a diagonal matrix whose nonzero entries are @xmath196, ,@xmath22 , where @xmath23 has multiplicity @xmath24 , and hence @xmath14-equivalent to the diagonal form given in theorem [ qdiag ] .
it would be of interest to find a diagonal form for the laplacian matrix @xmath197 of @xmath3 .
the smith group of this matrix is called the _ critical group _ of @xmath3 . by the results of @xcite ,
only the @xmath17-sylow subgroup of the critical group remains to be determined , for both odd and even @xmath0 .
we do not have any conjecture about its exact structure .
however , we note that if two integral matrices are equal modulo @xmath198 then for @xmath199 the multiplicity of @xmath200 as a @xmath13-elementary divisor is the same for both matrices .
thus , for example , when @xmath201 , theorem [ adiag ] gives part of the cyclic decomposition of the @xmath17-sylow subgroup of the critical group . | the @xmath0-cube graph is the graph on the vertex set of @xmath0-tuples of @xmath1s and @xmath2s , with two vertices joined by an edge if and only if the @xmath0-tuples differ in exactly one component .
we compute the smith group of this graph , or , equivalently , the elementary divisors of an adjacency matrix of the graph .
this work was partially supported by a grant from the simons foundation ( # 204181 to peter sin ) ] |
the idea that our @xmath2-dimensional observed universe may be a hypersurface or `` brane '' in a higher dimensional spacetime is motivated by string- and @xmath3-theory @xcite . in particular , @xmath0-dimensional braneworld scenarios , in which our universe represents the boundary of a @xmath0-dimensional spacetime , have recently received considerable attention @xcite .
the case in which the bulk spacetime is anti de sitter space and orbifold compactification is realized with the brane as fixed point has been particularly studied . in this situation
, it has been shown that @xmath2-dimensional gravity is recovered on the brane @xcite on energy scales much lower than the brane tension and/or bulk curvature , and late time cosmology is not changed if the brane tension is sufficiently high @xcite . in an attempt to solve the fine - tuning problem between the bulk cosmological constant and the brane tension , more complex models in which the bulk or the brane are filled with several species ( such as scalar fields )
have recently been proposed ( see for example @xcite and references therein ) . in these models @xmath1 symmetry is often assumed , and this is particularly convenient when considering boundary conditions on the brane . if @xmath1 symmetry is dropped , brane motion in the bulk must be taken into account and involved calculations are required in order to determine the boundary conditions on the brane .
whilst @xmath1 symmetry is motivated by @xmath3-theory and is required for a supersymmetric brane configuration , such as a bps state @xcite , there exist situations in which @xmath1 symmetry is broken .
this occurs , for example , when the brane is charged and couples to a @xmath2-form field in the bulk @xcite .
cosmological asymmetric brane models have been studied in @xcite .
these developments have prompted us to derive gauge invariant perturbation theory for brane cosmology with one codimension .
our aim is develop a formalism which may then be applied to any situation of cosmological interest .
previously , perturbations in braneworld cosmology have been extensively studied in the literature mostly for the case of @xmath1 symmetry @xcite . here
we consider the most general situation in which the spatial background geometry on the brane has maximal symmetry and thus represents a space of constant curvature @xmath4 .
we do not assume @xmath1 symmetry , and the boundary conditions on the brane are discussed .
also , no particular gauge choice for the metric component @xmath5 is made .
the perturbation equations in the bulk and on the brane are derived for general bulk and brane stress - energy tensors .
this makes our formalism particularly convenient when analyzing situations in which different bulk components ( such as several scalar fields ) are also considered .
the formalism can be used to study phenomena which have important observational consequences , the most important of them being the calculation of the anisotropies of the cosmic microwave background @xcite .
since one must in general first determine the behaviour of perturbations in the bulk before being able to determine their behaviour on the brane @xcite , we pay particular attention to the relation between bulk and brane gauge invariant perturbation variables .
these become more subtle when the position of the brane is displaced .
indeed we define a set of gauge invariant variables in which the perturbation equations on the brane become similar to the usual @xmath2-dimensional equations .
we then study the new terms arising in braneworlds .
since we assume very general background spacetimes and no @xmath1 symmetry , some of our equations are extremely cumbersome . in order to guide the reader through the rest of the paper , we now give a general overview of the methods we use , the variables we introduce , and the equations we derive in this paper .
the basic setup is one of a @xmath6-dimensional brane where the @xmath7-space of constant time is maximally symmetric ( a space of constant curvature ) , embedded in a @xmath8-dimensional bulk .
as @xmath1 symmetry is not assumed , the bulk spacetimes on each side of the brane will generally differ .
both the brane and the bulk may contain arbitrary matter .
our notation is as follows : * @xmath9 , @xmath10 , @xmath11 , @xmath7 , @xmath12 : spacetime coordinates ( greek indices ) , with metric @xmath13 and covariant derivative @xmath14 , * @xmath15 , @xmath11 , @xmath7 : coordinates on the maximally symmetric @xmath7-space ( second part of latin alphabet ) with metric @xmath16 and covariant derivative @xmath17 , * @xmath18 , @xmath10 , @xmath11 , @xmath7 : brane - worldsheet coordinates ( first part of latin alphabet ) , * @xmath19 : brane position in target - space .
* a roman subscript @xmath20 indicates `` brane '' whilst @xmath21 denotes `` bulk '' . * certain variables such as the brane matter content ( @xmath22 , @xmath23 , etc ) are only defined _ on _ the brane .
other variables such as the normal vector to the brane @xmath24 or the extrinsic curvature @xmath25 are also defined at the brane position , but since they describe the embedding of the brane in the bulk , they may take different values _ on either side _ of the brane .
all these brane - related variables are underlined .
the action for the system is @xmath26 here @xmath27 is the gibbons - hawking boundary term required to consistently derive the israel junction conditions @xcite , and @xmath28 is the fundamental @xmath0-dimensional newton constant ( related to the @xmath0-dimensional planck mass @xmath29 by @xmath30 ) .
furthermore , @xmath31 is the bulk scalar curvature , @xmath32 and @xmath33 are the bulk metric and the induced metric on the brane respectively , and @xmath34 and @xmath35 are respectively the lagrangians for arbitrary matter in the bulk and matter confined on the brane .
they may also contain a cosmological constant or brane tension .
the induced metric on the brane is ( see for example @xcite ) @xmath36 and the einstein equations resulting from action ( [ action ] ) are @xmath37 where @xmath38 is a covariant dirac @xmath39-function specifying the position of the brane ( see section [ ssec_ee ] ) , and @xmath40 as noted above , we underline @xmath41 and @xmath42 to emphasize that they are only defined on the brane ( see section [ seciii ] ) .
we consider these einstein equations ( [ einstein ] ) for a homogeneous and isotropic brane and bulk background with first order perturbations . as is summarised schematically in the left hand panels of fig .
[ fig1 ] , these equations contain three parts : one is continuous ; the second is discontinuous across the brane ; and the third part is singular at the brane position ( proportional to @xmath38 ) .
the coefficients of each of the individual parts must be equated , leading to a number of different equations .
the continuous part gives the einstein equations in the bulk and , via the gauss - codacci equation , they also determine the @xmath2-dimensional einstein tensor on the brane ( see section [ seciv ] ) .
the discontinuous ( but non singular ) part is only non - trivial when @xmath1 symmetry is not assumed .
it then gives equations for the continuous part of the extrinsic curvature , and it describes the energy and momentum exchange between the brane and bulk , leading to the equation of motion for the brane the so - called `` sail equation '' @xcite .
finally , the singular part represents the second junction condition which relates the bulk geometries on each side of the brane through the brane geometry and matter content . * 0.4 cm einstein equations conservation equations
* we also discuss the so - called conservation equations for the stress - energy tensors given in eqns ( [ t1],[t2 ] ) .
again , these contain a discontinuous , continuous and singular part . as is summarized schematically in the right hand panels of fig .
[ fig1 ] , the continuous part gives the bulk energy momentum conservation , the discontinuity simply describes the conservation of the jump of the bulk stress - energy on the brane , and the singular part leads to energy momentum conservation of the brane with a possible contribution from the bulk , and to the sail equation .
when discussing the perturbations of these equations , we will make use of the maximal symmetry of the @xmath7-dimensional subspaces parallel to the brane .
our geometrical quantities will be decomposed into scalar , vector and tensor degrees of freedom ( with respect to these @xmath7-spaces ) .
this decomposition is not identical to the more physical one containing density modes , vorticity modes , and @xmath0-dimensional gravitational waves .
the relationship between these two approaches is given in section [ ssecsvtdvg ] .
finally , in order to set up a consistent gauge - invariant formalism for the evolution of these perturbations , we will see that it is crucial to take fully into account the perturbed brane motion ( which can be written in a gauge invariant manner ) .
this degree of freedom will be central to our analysis .
the outline of the paper is the following . in the next section ( section [ secii ] ) we discuss the unperturbed ( or background ) @xmath0-dimensional bulk : we allow a foliation ( with two codimensions ) into maximally symmetric @xmath7-spaces , and do not specify the presence of the brane .
the einstein and conservation equations for the bulk background are derived . in section [ seciii ]
we introduce the brane and we discuss the boundary conditions at the brane position for the unperturbed spacetime without imposing @xmath1 symmetry . in section [ seciv ]
we derive the background equations for an observer on the brane . in section [ secvi ] , we perturb the background .
we introduce gauge invariant variables and derive the perturbed einstein and conservation equations in terms of these variables .
the perturbed brane including the perturbation of the brane position is discussed in section [ secvii ] . in section [ secix ]
we reformulate the perturbation theory from the point of view of an observer confined to the brane , and in the last section we draw some conclusions .
finally , we also provide an extensive and highly technical appendix where we present all the relevant intermediate steps required to obtain the results presented in the text .
( examples are , for instance , the perturbed christoffel symbols and the components of the perturbed riemann and weyl tensors . ) the appendix is , in fact , more general than the main text since there we consider an @xmath43-dimensional brane ( with an @xmath44-dimensional maximally symmetric subspace ) embedded in a @xmath45-dimensional bulk : in the text we have set @xmath46 .
furthermore , whilst the text presents the perturbation equations in full generality , some specific examples such as a bulk scalar field are discussed briefly in the appendix .
in this section we describe the bulk background geometry and energy content without introducing a brane .
we assume that the space orthogonal to the fifth dimension is maximally symmetric so that a homogeneous and isotropic brane can be accommodated , as discussed in the next section .
we consider the most general stress - energy tensor which satisfies these symmetry conditions , and then derive the einstein equations and the conservation equations .
we consider a @xmath0-dimensional spacetime with one timelike coordinate @xmath47 , and four spacelike coordinates @xmath48 , where @xmath49 .
we assume that the constant time hypersurfaces are locally of the form @xmath50 , where @xmath51 is a @xmath7-dimensional maximally symmetric space , i.e. , a @xmath7-space of constant curvature , parameterized by the coordinates @xmath52 , with spatial metric @xmath53 .
the curvature of this space will be denoted by @xmath4 .
for example , we may choose the coordinates @xmath54 such that @xmath55 where @xmath56 is the kronecker symbol .
the last spacelike coordinate @xmath57 ( the `` extra dimension '' ) is orthogonal to the maximally symmetric space .
the metric has the signature @xmath58 .
the line element of the metric can therefore be written as @xmath59 an overdot will denote derivation with respect to @xmath60 , and a prime derivation with respect to @xmath57 .
in addition , we shall define @xmath61 covariant derivatives with respect to the full metric will be denoted by @xmath14 , and those with respect to @xmath16 by @xmath17 . for convenience
we also define @xmath62 as well as @xmath63 notice that @xmath64 and @xmath65 do not commute since @xmath66 .
we now consider the bulk stress - energy tensor @xmath67 whose energy flux need not be at rest with respect to our @xmath68 coordinates .
let @xmath69 be the normalized timelike eigenvector of @xmath70 with eigenvalue @xmath71 .
correspondingly let @xmath72 be the normalized eigenvector orthogonal to both @xmath69 and to the maximally symmetric @xmath7-spaces , with eigenvalue @xmath73 .
finally , let @xmath74 be the eigenvalue of the three eigenvectors parallel to the maximally symmetric @xmath7-dimensional slices .
note that any symmetric tensor can be decomposed in this way , and that the symmetry requires that the eigenvectors parallel to the symmetric @xmath7-spaces are degenerate . in terms of these variables ,
the bulk stress - energy tensor can be written as @xmath75 where the vectors @xmath69 and @xmath72 are given by @xmath76 here @xmath77 represents the lorentz boost which must be performed along the @xmath57 axis in order to be in the fluid s rest frame . when one is not in the rest frame of the fluid , both its energy density and its pressure along the extra dimension are modified , and the fluid exhibits a flux through an @xmath78 hypersurface .
as usual , @xmath79 . below it will be more convenient to use a different definition for the stress - energy tensor components a definition which is less adapted to the fluid , but better adapted to our coordinates . to derive it ,
let us denote by @xmath80 the @xmath0-velocity of a bulk observer who is at rest with respect to our coordinate system , @xmath81 and by @xmath82 the spacelike unit vector orthogonal to both @xmath80 and @xmath51 , @xmath83 note that neither @xmath80 nor @xmath82 are geodesic vector fields , but their orthogonality and normalization is conserved .
the most general form of the bulk stress - energy tensor satisfying the required symmetry with respect to translations and rotations in @xmath84 can be written as @xmath85 where @xmath86 denotes symmetrization . in components
this gives @xmath87 thus @xmath88 is the bulk energy density as measured by an observer with @xmath0-velocity @xmath89 , @xmath90 is the energy flux transverse to @xmath51 , and @xmath91 , @xmath92 are the pressure along the directions @xmath93 , @xmath57 , respectively . the new variables @xmath94 , @xmath92 , @xmath91 and @xmath95 are related to the old ones by @xmath96 the last relation again shows that @xmath95 represents the energy flux in @xmath57 direction .
this flux , as well as the energy density @xmath94 and pressure @xmath92 along the @xmath57 direction measured by an observer at rest with respect to the coordinate system , are obtained from the components of the stress - energy tensor in a frame at rest with respect to the fluid simply by a lorentz transformation .
somewhat more complicated but equally straightforward expressions express the old variables in terms of the new ones ( see appendix [ ssec_app_set ] ) .
note that in ( [ udef ] ) the bulk velocity of the fluid is @xmath97 .
when @xmath98 , @xmath99 , we recover the case in which @xmath100 and @xmath101 , so that the rest frame of the bulk matter and the coordinate system coincide .
the christoffel symbols , the riemann , ricci , einstein and weyl tensors for the metric ( [ metricback ] ) are given in appendices [ ssec_app_christ ] , [ ssec_app_ric ] , [ ssec_app_ein ] , [ ssec_app_riem ] and [ ssec_app_weyl ] respectively , and the background bulk einstein equations are @xmath102 with the stress - energy tensor ( [ emt ] ) and the einstein tensor from appendix [ ssec_app_ein ] , eq .
( [ eback ] ) becomes @xmath103 where we have indicated in braces on the right hand side from which component of the einstein tensor these bulk einstein equations are derived .
equations ( [ einback00q],[einback55q ] ) were first discussed in @xcite , and integrated with respect to the fifth dimension in @xcite , for the case of a negative bulk cosmological constant , @xmath104 .
the first and the third of these equations ( [ einback00q],[einback05q ] ) are constraints ( i.e. , they do not involve second derivatives with respect to time ) .
the other two are dynamical equations . in fact
, there are only two independent dynamical variables which can be written as a combination of the scale factors @xmath105 , @xmath106 , and @xmath107 .
one can choose coordinates to remove this ambiguity : for example , in gaussian coordinates @xmath108 as in @xcite , and in conformal coordinates @xmath109 @xcite .
of course other choices of coordinates are also allowed .
we shall , however , keep @xmath107 undetermined , so that any useful choice for @xmath107 can be made at the end .
the bianchi identities lead to the so - called conservation equations for the stress - energy tensor , @xmath110 only for @xmath111 and @xmath112 are there non - trivial relations , @xmath113 these are the conservation equations for the energy density and the energy flux of the bulk components , respectively .
the generalisation to several components is straightforward ( see appendix [ ssec_app_cons ] ) . written in term of the intrinsic fluid quantities , they give an equation of evolution for the energy density @xmath71 and for fluid bulk velocity @xmath77 .
we now consider a homogeneous and isotropic @xmath7-brane orthogonal to @xmath57 ( lying in the space of maximal symmetry ) as a singular source , with intrinsic stress - energy tensor @xmath114 .
let us choose the intrinsic brane coordinates @xmath115 , and embed the brane according to @xmath116 note that it is always possible to choose the background coordinate @xmath57 such that the unperturbed brane is at rest : this is the only coordinate choice made in this paper .
as we shall see , the presence of the brane will introduce discontinuities at @xmath117 in several variables .
for that reason , it is useful to decompose a given function @xmath118 as @xmath119 where @xmath120 is the heaviside function .
this equation defines the continuous function @xmath121 , whilst the discontinuity or jump of @xmath118 when going from one side to the other side of the brane is given by @xmath122 notice that we have the two product relations @xmath123 for later convenience , and when considering a continuous function @xmath124 , we will also define the continuous part and the jump of its derivative by @xmath125 sometimes we shall also need @xmath126 for variables @xmath127 describing the embedding of the brane , and thus which may take different values , @xmath128 , @xmath129 , on either side of the brane .
the quantities @xmath126 and @xmath130 are defined by @xmath131 the normal unit vector to the brane , @xmath132 , is given by @xmath133 one obtains ( up to an overall sign ) @xmath134 as we shall see , @xmath107 can be discontinuous on the brane and @xmath24 can have different values _ on either side _ of the brane .
from the induced metric one can define the first fundamental form @xcite @xmath135 , where @xmath13 is again evaluated on ( either side of ) the brane , and we have @xmath136 . on the brane
, @xmath137 is related to @xmath138 by @xmath139 we can decompose the stress - energy tensor _ on _ the brane , @xmath140 , as @xmath141 where @xmath142 is the 4-vector of the energy flux on the brane matter , @xmath143 note that @xmath144 .
this is the most generic stress - energy tensor compatible with a homogeneous and isotropic brane . in the presence of the brane ,
the @xmath0-dimensional einstein equations become @xmath145 where , from eqn ( [ action ] ) , the `` covariant dirac function '' @xmath38 is @xmath146 here @xmath147 and @xmath148 are the determinants of the metric @xmath32 and first fundamental form @xmath137 respectively , evaluated at the brane position . written in components , the einstein equations with the brane , eq .
( [ einsteinbulk ] ) , become @xmath149 a global solution to these equations has been derived in @xcite with the assumption of a pure negative cosmological constant in the bulk , and using gaussian coordinates .
the right hand sides of eqns ( [ backein00],[backeinij ] ) contain a singular term proportional to @xmath38 due to the presence of the brane .
as we will see below , although the first fundamental form is continuous on the brane , its first derivative with respect to the fifth dimension @xmath57 ( i.e. , the terms @xmath150 and @xmath151 ) may jump and its second derivative ( @xmath152 and @xmath153 ) can be singular . thus the einstein tensor contains a singular part which must be matched with the singular part of the stress - energy tensor .
we now turn to the problem of relating these terms to the brane matter content .
the extrinsic curvature formalism is a useful tool in the analysis of junction conditions on a singular surface @xcite .
the _ first israel condition _ @xcite imposes the continuity of the first fundamental form , @xmath154 hence @xmath137 is well - defined _ on _ the brane .
since we have @xmath155 and @xmath156 , this condition implies the continuity of the scale factors @xmath105 and @xmath106 : @xmath157 ( see figure [ fig2 ] ) .
note that the continuity of the metric function @xmath107 is not required by the junction conditions and will not be assumed in what follows to be discontinuous makes the covariant dirac function @xmath38 ill - defined .
this is not a serious problem , as all the terms involving this function can be grouped together to give the second israel condition .
therefore , we shall continue to use the notation @xmath38 and suppose that when @xmath107 is not continuous , it corresponds to a regularized and mathematically consistent expression . ]
( see also appendix [ sec_app_pert_bulk_geom ] ) .
nevertheless , the first derivative with respect to @xmath57 of @xmath106 and @xmath105 ( which are proportional to @xmath151 and @xmath150 ) , are allowed to jump . in order to study the behaviour of these quantities on the brane we consider the extrinsic curvature tensor ( or second fundamental form ) with respect to the brane , namely @xmath158 for the background metric
, the components of the extrinsic curvature are @xmath159 let us define the surface `` stress tensor '' @xmath160 on the brane by @xmath161 then the _ second israel condition
_ @xcite relates the jump in the extrinsic curvature with the energy content on the brane and requires that @xmath162 ( note that the choice of the sign here is consistent with our choice for the sign of @xmath132 in eq .
( [ sign_perp ] ) . ) for our background this condition can be written as ( see appendix [ ssec_app_isr2 ] ) @xmath163 ( see also figure [ fig3 ] . ) alternatively , eqns ( [ bjcn],[bjca ] ) can be obtained directly from the singular part of the einstein equations ( [ backein00],[backeinij ] ) .
we now comment briefly on the question of boundary conditions at the brane .
consider first the bulk einstein equations ( [ backein00][backein55 ] ) .
they form a system of second order partial differential equations in @xmath60 and @xmath57 , and in order to solve them we must specify initial conditions on a spacelike cauchy hypersurface , boundary conditions far from our braneworld ( at infinity in a one brane scenario , or on another brane ) , and boundary conditions at our brane .
the israel junction conditions impose the continuity of @xmath106 and @xmath105 , and fix the jump in their normal derivatives at the brane .
since the einstein equations represent a set of second order partial differential equations , these junction conditions are sufficient to allow us to solve the einstein equations everywhere in the bulk .
we now turn to the einstein equations on the brane .
an observer on the brane will not see @xmath2-dimensional einstein gravity
. this may , however , be recovered in particular situations at low energy .
the @xmath2-dimensional einstein tensor in general depends on bulk quantities and is quadratic in the brane stress - energy tensor .
here we discuss the @xmath2-dimensional `` einstein equations '' which lead to the modified friedmann equations , and also the conservation equation on the brane .
finally we interpret the deviation from the @xmath2-dimensional theory in terms of the @xmath0-dimensional one . through the gauss - codacci equations , we can write the @xmath2-dimensional einstein tensor @xmath164 in terms of the bulk stress - energy tensor @xmath67 , the extrinsic curvature @xmath25 and the projected weyl tensor @xmath165 .
the details of the calculation can be found in @xcite , and the result is @xmath166 the projected weyl tensor @xmath167 on the brane is obtained from the bulk weyl tensor @xmath168 as follows @xmath169 \beta } + g_{\beta [ \gamma } r_{\delta ] \alpha } ) + \frac{1}{6 } r g_{\alpha [ \gamma } g_{\delta ] \beta } , \\
\wn_{\alpha \beta } & = & c_{\alpha \mu \beta \nu } \normvec^\mu \normvec^\nu .\end{aligned}\ ] ] here @xmath170 } \equiv \tdemi ( f_\alpha g_\beta - g_\alpha f_\beta)$ ] denotes antisymmetrization .
this projection represents the contribution of the free gravity in the bulk to the gravity on the brane . in components
, we have @xmath171 there is only one independent component in the weyl tensor ( as well as in its projection on the brane ) .
this is related to the fact that this spacetime is the @xmath0-dimensional analog of a @xmath2-dimensional type-@xmath172 spacetime in petrov s classification @xcite .
since the tensor @xmath164 contains only derivatives of the continuous first fundamental form with respect to @xmath60 and @xmath93 , it is continuous .
hence , on taking the continuous part of the right hand side of eq .
( [ genein ] ) ( and applying the product relation ( [ pr1 ] ) ) , we find the projected @xmath2-dimensional einstein equation on the brane , @xmath173 the right hand side of equation ( [ generale ] ) can be split into four parts .
the first depends on the average of the bulk stress - energy tensor , @xmath174 .
the second is given by four terms quadratic in the average of the extrinsic curvature @xmath175 .
these terms are known once the bulk einstein equations have been solved .
they vanish when @xmath1 symmetry is assumed we return to this point below .
then there is a third part which contains four terms quadratic in the jump of the extrinsic curvature @xmath176 .
we have already determined these through the second junction condition ( [ bjcn],[bjca ] ) , and they are related to the brane stress - energy tensor : these terms will be responsible for the non - standard @xmath177 contribution in the brane friedmann equations .
finally there is a fourth part , the average of the projected bulk weyl tensor on the brane , describing the effect from the free gravity in the bulk . in components ,
we obtain the modified friedmann equations which contain a dynamical equation and a constraint : @xmath178 where we have isolated the continuous part of the projection of the bulk weyl tensor , @xmath179 ( these equations could alternatively have been obtained from the continuous part of the einstein equations , eqns ( [ backein00][backein55 ] ) . )
the cosmological consequences of these equations have been studied in @xcite with assumption of @xmath1 symmetry , in which case @xmath180 .
these authors considered a negative cosmological constant in the bulk and assumed that the brane stress - energy tensor consists of a rigid part the brane tension and a fluid , @xmath181 in this case the bulk stress - energy tensor can be tuned to the brane tension in such a way that deviations from standard friedmann equations are effective only at energies of order @xmath182 and higher .
if we assume @xmath183 , integrate the sum of eq .
( [ backein05 ] ) and ( [ backein55 ] ) once with respect to time , and compare the result with ( [ bf1 ] ) , one obtains an expression for the sum of the continuous part of the projected bulk weyl tensor and the continuous part of the bulk energy density and pressure : this behaves as a radiation term , @xmath184 where @xmath185 is an integration constant ( see appendix [ ssec_app_newfried ] ) . in
@xcite this is discussed for the case @xmath186 so that @xmath187 .
furthermore , notice that eq .
( [ brnlambda ] ) is not a necessary requirement in order for the @xmath0-dimensional friedmann equations on the brane to reduce to the standard @xmath2-dimensional friedmann equations at low energy .
it suffices that the brane stress - energy tensor is dominated by a term which is almost a cosmological constant today ( i.e. , it can be a slow - rolling scalar field , see for example @xcite ) . in this case , one has @xmath188 where @xmath189 , @xmath190 are the energy density and pressure of the ordinary matter content on the brane . now ,
if in addition @xmath191 the first two terms of the right hand side of the above friedmann equations on the brane are proportional to @xmath192 , @xmath193 , with @xmath194 and one re - obtains a term linear in the brane matter content .
note also that there are no extra @xmath195 appearing in the projected weyl tensor ( [ contweyl ] ) since with these conditions @xmath196 is @xmath182-independent .
this was also noted in ref .
@xcite .
when solving the einstein equations on the brane ( [ generale ] ) , we need the continuous part of the extrinsic curvature @xmath175 .
if @xmath1 symmetry is assumed as motivated by @xmath3-theory the evolution of the bulk is the same on both the sides of the brane . in this case
the israel conditions determine @xmath25 entirely : it is always possible to choose a coordinate system in which @xmath197 and @xmath198 , i.e. , where @xmath105 and @xmath106 are even .
thus @xmath151 and @xmath150 are odd and the continuous value of @xmath25 across the brane vanishes , @xmath199 this implies @xmath200 if @xmath1 symmetry is not assumed , as in this paper , the evolution on either side will in general be different and @xmath201 no longer vanishes , @xmath202 one can , however , obtain a condition for the continuous part of the extrinsic curvature by considering the jump of eq .
( [ genein ] ) .
we obtain @xmath203 ( notice that @xmath204 . )
this becomes , in components , @xmath205 where , using eq .
( [ weylback ] ) and the junction conditions ( [ bjcn ] ) and ( [ bjca ] ) , the jump of @xmath206 on the brane can be expressed as @xmath207 ( note that eqns ( [ grrr1],[grrr2 ] ) could alternatively have been obtained from the discontinuous part of the einstein equations ( [ backein00][backein55 ] ) . )
equations ( [ grrr1],[grrr2 ] ) allow one to fix the unknown quantities @xmath208 , @xmath209 , _ provided _ the jumps of both the bulk matter content and the weyl tensor are known .
thus the continuous part of the extrinsic curvature depends not only on the brane matter content but also on the discontinuity of the bulk stress - energy and the projected weyl tensors .
if both vanish , eqns ( [ grrr1],[grrr2 ] ) allow in particular the trivial solution @xmath210 , which holds with @xmath1 symmetry .
the jump of the weyl tensor , eq .
( [ wwweyl ] ) , contains first derivatives of the extrinsic curvature with respect to @xmath57 , and so it is not possible in general to determine @xmath209 and @xmath208 without first solving the einstein equations in the bulk . nonetheless ,
if the projection ( [ wwweyl ] ) of the @xmath0-dimensional bulk weyl tensor on the brane is known _ a priori _ ( as in the case of a schwarzschild - anti de sitter bulk with a known black hole mass on both side of the brane ) , then @xmath209 and @xmath208 can be determined _ directly _ from eqns ( [ grrr1],[grrr2 ] ) ( see @xcite for a more detailed discussion ) . on contracting eq .
( [ jumpbrnee ] ) with the first fundamental form @xmath211 and using the second junction condition ( [ juncgen ] ) , one obtains @xmath212 this equation is known as the `` sail equation '' @xcite .
the right hand side is an external force density on the brane due to the asymmetry of the bulk stress - energy tensor on the two sides . in analogy with newton s second law ( here
the force is due to a pressure difference between the two sides of the brane ) , @xmath41 , @xmath201 , and @xmath213 play the role of mass , acceleration , and force , respectively .
notice from ( [ contk ] ) that when @xmath1 symmetry is assumed , this equation vanishes identically . when @xmath1 symmetry is broken , the `` acceleration '' @xmath214 is non - zero . in this paper
we do not assume @xmath1 symmetry , but recall that we have chosen a coordinate system in which the brane is at rest : eq .
( [ newcons2 ] ) must therefore be understood as dictating the condition that must be satisfied by @xmath208 and @xmath209 ( and therefore by the coordinate system itself ) for the brane to remain at a fixed position @xmath215 .
later , however , we will see that eq .
( [ sail ] ) does indeed give a more intuitive equation of motion for the perturbed brane position or brane displacement ( see section [ ssec_sail ] ) . in components the sail equation leads to @xmath216 which can also be obtained by taking the discontinuous part of the @xmath217 component of the einstein equation , eq .
( [ backein55 ] ) or , of course , by a linear combination of eqns ( [ grrr1],[grrr2 ] ) .
this is the only combination of equations ( [ grrr1],[grrr2 ] ) which does not involve the weyl tensor .
the singular part of the @xmath0-dimensional energy conservation equation ( [ backbianchi1 ] ) yields the stress - energy conservation equation on the brane : we find @xmath218 ( again the generalisation to several interacting components may be found in appendix d5 . )
notice that the jump in the bulk energy flux transverse to the brane enters in the conservation equation , meaning that the brane matter content can act as a source or a sink to the energy flux along the fifth dimension .
when this energy flux is continuous , the conservation equation on the brane reduces to the usual one , as discussed in @xcite .
another conservation equation appears in brane cosmology : by considering the singular part of eq .
( [ backbianchi2 ] ) we obtain again the sail equation ( [ newcons2 ] ) .
both equations were first found in @xcite for the case a bulk cosmological constant . in the previous sections we have derived the bulk perturbation equations and their boundary conditions on the brane .
this allows us in principle to solve the full system of perturbation equations in the bulk for given initial conditions . from these
one can determine also the perturbed weyl tensor and the second fundamental form . in order to make contact with @xmath2-dimensional cosmology in this section , we want to write the perturbed version of the @xmath2-dimensional einstein equations on the brane .
as for the background , this can either be done directly from the perturbed bulk einstein equations ( [ einpert00][einpertijt ] ) , or using the gauss - codacci equation . the full expression of the perturbed weyl tensor @xmath466 is given in appendix [ ssec_app_pert_weyl ] .
here we write only the components of the perturbed projected weyl tensor , @xmath467 , on the spacelike direction @xmath353 , written in terms of the underlined gauge invariant variables .
we have @xmath468 where we have set @xmath469 since , by solving the bulk equations , we in principle obtain the non underlined variables , it is useful to express the above components also in terms of these , @xmath470 for the vector and tensor part of the projected weyl tensor we have defined @xmath471 @xmath472 with these definitions , we can now write the projected perturbed einstein equations on the brane . they split into four scalar equations , @xmath473 two vector equations , @xmath474 and one tensor equation @xmath475 where we have set , for the bulk matter quantities evaluated at the brane position , @xmath476 these corrections follow from the fact that we have to go in a coordinate system which follows the brane .
the terms proportional to @xmath477 are here because we consider the bulk matter content at @xmath371 rather than at @xmath117 , the terms proportional to @xmath478 come from the fact that we also perform a lorentz boost in order to follow the brane motion , and the term @xmath436 in the last equation comes from the fact that the brane is bent . as for the unperturbed case
, the continuous parts of the bulk stress - energy tensor and of the projected weyl tensor appear on the right hand side of these equations , as well as the components of the continuous part of the perturbed extrinsic curvature .
these are related through the discontinuous part of the einstein equations to the discontinuity of the perturbed weyl tensor and of the bulk perturbed matter content .
the corresponding equations can be found in appendix [ ssec_app_pffffff ] .
the brane matter conservation equations follow from the singular part of @xmath479 or from the bianchi identities .
one obtains ( see appendix [ ssec_app_pert_cons ] ) @xmath480 again , when there are no discontinuities in the bulk matter perturbations , one obtains the usual conservation equations .
we now turn to perturbed quantities , and begin in this section by analysing the properties of the perturbed bulk : the perturbed brane itself will be introduced in section [ secvii ] .
we work throughout with gauge independent perturbation variables , which are inspired from a generalisation of the newtonian ( or longitudinal ) gauge to the @xmath0-dimensional case .
first we introduce the bulk metric perturbation variables using the standard scalar , vector , tensor decomposition .
we study their gauge transformation properties and define gauge invariant combinations .
then , in section [ ssec_pert_se ] , the perturbations of the bulk stress - energy tensor are considered , leading , in section [ ssec_pert_ee ] , to the gauge invariant perturbed bulk einstein equations .
finally we write down the perturbed conservation equations ( bianchi identities ) .
let us consider the perturbations of a spacetime with one timelike and @xmath219 spacelike coordinates .
the perturbed metric of this spacetime possesses @xmath220 different components . amongst these , a coordinate transformation allows @xmath221 of them to be fixed , so that there are @xmath222 independent metric coefficients .
for example , in synchronous gauge , the @xmath223 are set to zero .
when solving perturbation equations about a given spacetime , one is naturally led to classify perturbations .
two classifications are of particular relevance .
firstly , the perturbations may be classified according to their physical meaning , and this is done by looking at the spin of the perturbation in a locally minkowskian frame .
the different perturbations are density ( spin @xmath224 ) modes , vorticity ( spin @xmath10 ) modes , and gravitational ( spin @xmath11 ) waves . secondly , the perturbations may be classified more geometrically in terms of irreducible components under the group of isometries of the unperturbed spacetime .
this leads to scalar , vector and tensor perturbations . under some circumstances , these two classifications are identical .
in particular , this is true for a friedmann - lematre - robertson - walker ( flrw ) spacetime , which can be foliated by a set of maximally symmetric spacelike hypersurfaces . in brane
cosmology , however , the bulk is not as symmetric as in the flrw case , and the two classifications are different . components which transform irreducibly under symmetries of the background spacetime evolve independently ( to linear order ) while the physical spin components mix .
as explained above , metric perturbations can be decomposed according to their spin with respect to a local rotation of the coordinate system .
this leads to density modes , vorticity modes , and gravitational waves .
gravitational ( spin @xmath11 ) waves are `` true '' degrees of freedom of the gravitational field in the sense that they can exist even in vacuum .
the number of gravitational wave modes is given by the dimension of the vector space spanned by symmetric , transverse , traceless rank @xmath11 tensors in an @xmath219-dimensional space : this is @xmath225 .
in addition , when there is a non trivial matter content , there may be vorticity ( or spin @xmath10 ) modes arising from rotational velocity fields , which have @xmath226 independent components . finally , there remain @xmath227 possible density ( spin @xmath224 ) modes , which are usually represented by the two bardeen potentials @xmath228 and @xmath229 @xcite .
more schematically , let us consider the metric perturbation around a locally inertial frame , written in synchronous gauge and in fourier space considering the wave vector @xmath230 : @xmath231 the quantities @xmath232 and @xmath233 describe the density modes ( with the standard definition of the bardeen potentials , one has @xmath234 and @xmath235 ) , the @xmath236 ( @xmath237 ) represent the vorticity modes , and the @xmath238 ( @xmath239 ) and @xmath240 ( @xmath241 ) represent the gravitational waves ( when @xmath242 , these notations agree with the standard definition of the @xmath243 and @xmath244 modes ) .
the three above types of perturbation generally do not evolve independently : even at linear order , they are coupled if the unperturbed spacetime does not possess any symmetries .
however , for most cosmological models ( including the ones considered in this paper ) , spacetime possesses some symmetries , being invariant under a certain group of global transformations .
we consider the symmetry group @xmath245 with @xmath246 , which is of course relevant when there exists a coordinate system in which @xmath44 coordinates span a maximally symmetric space .
when this is the case , perturbations may be decomposed into components which transform irreducibly under @xmath245-rotations of the coordinate system .
this leads to what we call scalar , vector and tensor perturbations which are perturbations whose spin with respect to @xmath245 is @xmath224 , @xmath10 and @xmath11 respectively .
the main advantage of this decomposition is that the three new types of perturbation are now decoupled from each other , and hence are convenient when studying the evolution of cosmological perturbations .
for example , consider an @xmath219-dimensional space with @xmath44 coordinates ( labelled by @xmath247 , @xmath248 , etc ) spanning an @xmath44-dimensional , maximally symmetric sub - space , with metric @xmath16 , and associated covariant derivative @xmath17 .
the @xmath249 remaining coordinates will be labelled by @xmath250 , @xmath251 , etc . in this case
, the metric perturbations can be decomposed as @xmath252 where barred quantities are divergenceless @xmath44-vectors , and double barred quantities are divergenceless , traceless @xmath44-tensors of rank 2 ( with respect to the covariant derivative @xmath17 and metric @xmath16 respectively ) . with our definitions , it is clear that @xmath233 , @xmath232 , @xmath253 , @xmath254 are scalars , @xmath255 , @xmath256 are vectors , and @xmath257 are tensors under @xmath245 rotations .
the perturbed metric components can then be written as @xmath258 with the @xmath259 describing @xmath257 . obviously , one has * @xmath260 scalar degrees of freedom , * @xmath261 vector degrees of freedom and , * @xmath262 tensor degrees of freedom . by definition ,
the tensor components are spin @xmath11 quantities and represent gravitational waves .
it is clear that when @xmath263 , not all the gravitational waves are tensor perturbations ( with respect to @xmath245 ) : @xmath264 of them are actually scalar or vector perturbations .
in fact , the spin of the second decomposition can be understood as the projection of the spin of the first decomposition on the maximally symmetric space .
therefore , density modes are always scalar perturbations , vorticity modes can be either scalar of vector perturbations , and gravitational waves can be any of the three . by comparing eqns ( [ dgphys ] ) and ( [ dggeom ] ) , it is clear that : * the @xmath260 scalars decompose as @xmath11 density modes , @xmath249 vorticity modes , and @xmath265 gravitational waves , * the @xmath261 vectors represent @xmath266 vorticity modes and @xmath267 gravitational waves , * the @xmath262 tensors all represent gravitational waves . for our purpose ( @xmath268 ) , this reduces to * @xmath2 scalar degrees of freedom which split into the @xmath11 density modes , @xmath10 vorticity mode , and @xmath10 gravitational wave , * @xmath2 vector degrees of freedom which go into @xmath11 vorticity modes and @xmath11 gravitational waves , * @xmath11 tensor degrees of freedom which all represent gravitational waves .
as expected , we have @xmath269 degrees of freedom @xmath0 of which are gravitational waves .
this decomposition ensures that even in the vacuum , the scalar and vector parts of the einstein equation will allow non trivial solutions .
these are usually called `` graviscalars '' and `` graviphotons '' @xcite .
this effect represents the most striking change to the physics of brane cosmological perturbations as compared to that of the standard flrw case since it can occur at arbitrary low energy as long as the corresponding gravitational waves exist in the bulk .
the brane is , by definition , described by @xmath270 coordinates : one timelike and the @xmath44 spacelike coordinates spanning an @xmath44-dimensional maximally symmetric space . for the case of one codimension , we have @xmath271 .
the perturbed induced metric of the maximally symmetric space then has @xmath272 independent components .
an important question is how these perturbations can interact with the bulk perturbations .
it is clear that whatever the bulk matter content , there are at least @xmath273 degrees of freedom which arise from the gravitational waves in the bulk .
therefore , one can expect that @xmath225 of the brane perturbations can interact with the bulk .
we will see that this is indeed the case : the second israel condition essentially states that the discontinuity of some bulk perturbations which can exist even in the vacuum describe the matter content of the brane .
but this also suggests that one additional scalar degree of freedom of the brane is likely not to be directly related with the bulk perturbations .
it happens , indeed , that this extra degree of freedom physically corresponds to the perturbation of the brane position in the bulk , which is independent of the gravitational waves .
for example , if the bulk is pure minkowski space , one can consider a fixed coordinate system ( as , e.g. , newtonian gauge , which is unambiguously fixed ) .
the position of the brane in this coordinate system is defined independently of the metric perturbations .
this extra degree of freedom ensures that in any situation all the brane perturbations can be related to bulk perturbations ( see also the discussion in ref .
one of the aims of this paper is to make the link between these two sets of perturbations .
we now make use of maximal symmetry on @xmath51 .
due to rotational invariance , we can split the perturbations into components which transform irreducibly under rotations , i.e. , into different @xmath274-spin components , which evolve independently to first order perturbation theory .
one could then go on and split these into irreducible components under translations , corresponding to the expansion in terms of eigenvectors of the laplacian on @xmath275 ( which is the fourier transform in the case @xmath276 ) @xcite .
following the discussion of the last paragraph , the perturbed line element can be generally written as @xmath277 here , the @xmath278 indices of @xmath279 , @xmath280 are labels .
the quantities @xmath281 and @xmath279 are vectors on @xmath51 which can be respectively decomposed into scalar ( spin 0 ) components @xmath251 , @xmath282 , and divergenceless vector ( spin 1 ) components @xmath283 , @xmath284 , such that @xmath285 .
equivalently , the tensor on @xmath51 , @xmath286 , can be decomposed into two scalars , @xmath233 and @xmath232 , a divergenceless vector , @xmath255 , and divergenceless , traceless , tensor ( spin @xmath11 ) component , @xmath257 , such that @xmath287 .
this decomposition is @xmath288 the indices of these @xmath51-quantities are raised and lowered with the metric @xmath16 .
the symmetries of the metric ensure that the scalar ( @xmath250 , @xmath251 , @xmath233 , @xmath232 , @xmath289 , @xmath282 , @xmath280 ) , vector ( @xmath283 , @xmath255 , @xmath290 ) and tensor ( @xmath257 ) quantities evolve independently .
let us consider an infinitesimal coordinate transformation @xmath291 with @xmath292 under this coordinate change the geometrical perturbations transform in the following way : @xmath293 ( recall that @xmath294 and that @xmath295 .
in this section we will use both this notation and the @xmath296 defined in eqns ( [ defperpu],[defperpn ] ) : we aim to do so in such a way as to keep the equations as simple as possible . )
we can therefore define the following four scalar and two vector perturbation variables , which are invariant under infinitesimal coordinate transformations , also called gauge transformations in this context : @xmath297 the two vector variables possess two independent components ( hence four degrees of freedom ) .
the tensor variable @xmath257 is gauge invariant since there are no tensor type gauge transformations , and possesses two independent components .
all these quantities represent a generalisation of the newtonian gauge often used in flrw cosmologies ( we can no longer call it a `` longitudinal gauge '' , as @xmath298 ) .
it is completely fixed by setting @xmath299 , @xmath300 , @xmath301 to @xmath224 and in this case one has @xmath302 this gauge is perfectly well - suited for describing the bulk perturbation _ without _ a brane . in the presence of a brane ,
however , things are more complicated since some of these quantities involve first or second derivatives with respect to the fifth dimension , and hence they are not always regular at the brane position ( see section [ ssec_reg_pert_geom ] ) . for this reason
, other gauge choices are often preferred , but not essential .
we now perturb the unit vectors @xmath69 and @xmath72 , defined in eqns ( [ tdef][ndef ] ) , which are the timelike and spacelike eigenvectors normal to the maximally symmetric @xmath7-spaces .
it follows from the normalization conditions , eqns ( [ udef],[ndef ] ) , that each vector has only four independent components .
furthermore , as @xmath69 and @xmath72 are eigenvectors of a symmetric tensor , they are normal to each other , @xmath303 .
hence at perturbed order there are only seven independent components which we denote by @xmath304 , @xmath305 , @xmath306 .
they are defined by @xmath307 neglecting the metric perturbations , the quantity @xmath306 represents the perturbation of the lorentz boost @xmath77 , @xmath308 as usual , we will decompose @xmath304 , @xmath305 , into scalar and vector components , @xmath309 finally , in order to write down the stress - energy tensor , it is also useful to introduce the variables @xmath310 which have a decomposition into scalar and vector components similar to ( [ decom1],[decom2 ] ) . with these definitions ,
general perturbations of the bulk stress - energy tensor , @xmath311 , are @xmath312 here we have defined , according to ( [ cordtointdeb][cordtointfin],[wperp ] ) : @xmath313 and we have introduced the anisotropic stress tensor @xmath314 , which again may be decomposed into a scalar , ( divergenceless ) vector , and ( divergenceless , traceless ) tensor components according to @xmath315 on investigation of the behaviour of these variables under the infinitesimal coordinate transformations ( [ infi ] ) ( see appendix [ sec_app_pert_bulk_matt ] ) , we find the following scalar gauge invariant variables @xmath316 where @xmath317 is any scalar quantity ( density @xmath71 , pressure @xmath74 , etc ) .
the anisotropic stress tensor @xmath314 is gauge invariant by itself due to the stewart - walker lemma @xcite .
notice that @xmath94 , @xmath92 , @xmath95 and @xmath306 are not scalars ( since they depend explicitly on the choice of the coordinate system via the vector fields @xmath80 and @xmath82 ) , but we can , however , define the following gauge invariant variables , @xmath318 as an example , the perturbed stress - energy tensor for a scalar field is given in appendix [ ssec_app_sf ] .
the explicit forms of the perturbed christoffel symbols , the perturbed riemann , ricci and einstein tensors are all given in appendices [ ssec_app_pert_christ ] , [ ssec_app_pert_ric ] , [ ssec_app_pert_riem ] , [ ssec_app_pert_ein ] , where they are expressed in terms of the gauge invariant variables introduced above .
we now write down the full perturbed bulk einstein equations also in terms of gauge invariant variables .
they split into seven scalar , three vector ( each with two independent components ) , and one tensor ( with two independent components ) equations , adding up to the required @xmath319 components of a symmetric @xmath320 tensor .
these equations are given below , where we indicate on the right hand side of each equation from which component of the einstein equations they were derived and , when necessary , the term to which they are proportional .
the seven scalar equations are @xmath321 the three vector equations are : @xmath322 and the tensor equation is @xmath323 as a small aside , it is interesting to check our analysis of section [ ssecsvtdvg ] .
we shall take for simplicity an empty , minkowski bulk , so that the terms proportional to @xmath324 , @xmath325 , @xmath326 , @xmath150 and @xmath151 vanish .
then the above equations reduce to @xmath327 in the vacuum , in addition to the usual two tensor modes , there are one scalar and two vector degrees of freedom which satisfy wave equations and represent the graviscalar and graviphoton ( for a total of five gravitons , as expected ) .
the remaining degrees of freedom can only exist if matter is present .
they describe either density or vorticity modes .
we now compute the perturbed energy momentum conservation equations .
even though they do not contain new information , they can provide useful evolution equations for the matter content of the bulk .
here we write them down just for the total bulk matter .
the generalisation to several components is straightforward and is given in appendix [ ssec_app_pert_cons ] .
written in terms of gauge invariant variables there are three scalar conservation equations , @xmath328 and one vector conservation equation , @xmath329 finally , in order to close the system , we must specify an equation of state for @xmath330 , @xmath331 , @xmath332 and @xmath314 , as functions either of @xmath333 or of some other non dynamical variables ( such as the entropy ) .
for example , all these quantities vanish for a bulk containing non relativistic matter . for a scalar field ,
most of them are also set to zero , as is discussed in appendix [ ssec_app_sf ] .
we can interpret the three scalar equations ( [ pert_cons_0][pert_cons_5 ] ) as the conservation equations for @xmath333 , @xmath334 , and @xmath335 .
notice that with these bianchi or conservation equations , three scalar and one vector einstein equations are redundant and can be dropped .
formally , the seven scalar einstein equations can be split into four dynamical equations for the four scalar metric perturbations @xmath228 , @xmath229 , @xmath336 and @xmath337 , and three constraint equations .
it happens , however , that with our choice of variables , the splitting is not completely straightforward .
for example , the @xmath338 einstein equations are the constraint equations for the metric components @xmath339 , @xmath340 , @xmath341 , the rest being the evolution equations
. however , in terms of our gauge invariant variables , eq .
( [ einpertijsst ] ) is obviously a constraint equation .
this is because @xmath229 involves first and second time derivatives of the metric perturbation @xmath232 .
equivalently , the three vector einstein equations can be split into one constraint equation and two dynamical equations for the variables @xmath342 and @xmath343 .
in the previous section we have considered the most general perturbed @xmath0-dimensional bulk spacetime for which there is a perturbed maximally symmetric space orthogonal to the fifth direction .
we have seen that its dynamics can be described in terms of four geometrical scalar perturbation variables governed by four evolution equations and three constraints , two geometrical vector perturbation variables governed by two evolution equations and a constraint , and one geometrical tensor variable governed by one tensor evolution equation . in this section
we add a brane to this system that is we assume , as was the case for the background , that the bulk contains a perturbed homogeneous and isotropic brane as a singular source .
this will introduce one new geometrical degree of freedom , the brane displacement , whose dynamics has to be considered in order to fully describe the perturbations on the brane .
the perturbed brane embedding is given by @xmath344 here @xmath345 is the displacement of the brane from its background position @xmath346 , and it is a function on the brane worldsheet . it is a true new degree of freedom which sometimes also called `` radion '' @xcite . on the contrary , as we will soon see , the perturbations in the @xmath347 and @xmath348 directions can be set to zero without loss of generality , as they do not lead to any physical consequences ( e.g. , to a physical `` deformation '' of the brane ) @xcite .
it was first noticed in @xcite that , when studying brane perturbations of a randall - sundrum background ( of ref .
@xcite ) , using the transverse and traceless gauge in gaussian normal coordinates , the brane position is no more at constant @xmath57 in presence of matter sources .
the presence of an @xmath349 can be interpreted as a bending of the brane due to the presence of matter or gravitational waves .
the bending @xmath345 can in principle be set to zero by choosing a convenient set of bulk coordinates such that @xmath350 , since by the infinitesimal coordinate transformation defined in the previous section , eq .
( [ cico ] ) , one has @xmath351 this is not , however , the most general possibility . as was noted in @xcite
, such a gauge choice will also fix the gauge of some other perturbation variables .
in fact , if we choose @xmath352 such that @xmath345 is zero , @xmath301 is fixed ( see eq .
( [ transmetfin ] ) ) . in a gauge invariant
approach one must keep @xmath345 arbitrary . we now can define the perturbed vector orthogonal to the brane , @xmath353 .
one easily obtains ( see appendix [ ssec_app_pert_brane2 ] ) @xmath354 note that no vector perturbations enter in the above formula .
this is just a consequence of frobenius theorem @xcite .
also , the fact that the perturbations @xmath355 do not enter in the above expression illustrates that they do not correspond to any physical deformation of the brane ( see further comments below ) . in this subsection
we calculate the perturbed first fundamental form which will be used in the perturbed first israel junction condition in the following subsections .
we shall first look at the perturbation of the induced metric @xmath356 . in doing so , it is important to recall that the brane embedding in the unperturbed and perturbed bulks ( respectively given by ( [ embdeb][embfin ] ) , ( [ embpertdeb][embpertfin ] ) ) are different . therefore , each brane variable has two contribution to its perturbation : one coming from the perturbation of the variable and a second contribution due to the fact that we have to evaluate it at the perturbed brane position .
we obtain @xmath357 where , to first order in perturbation theory , the right hand side of these equations are evaluated at @xmath358 ( corresponding to the unperturbed embedding ) . as usual
, we can decompose the perturbation @xmath359 into its scalar and vector parts using the metric @xmath16 evaluated at the unperturbed brane position : @xmath360 .
this is possible since , for perturbations , the time derivative @xmath361 and spatial derivatives @xmath362 are equivalent to derivation with respect to @xmath363 and @xmath364 respectively .
a word of caution is in order here .
recall that the quantities @xmath250 , @xmath233 , @xmath151 , @xmath150 , etc .
, are defined in the bulk . however , due to the presence of the brane , they may be ( and in fact , are see below ) discontinuous at the brane position . for this reason
, they may only be evaluated on each side of the brane .
therefore we must check that the above expressions are consistent in the sense that they have the same value when evaluated _ on both sides _ of the brane only if that is the case , they can be considered well defined _ on _ the brane . as we shall soon see , this consistency in fact results from the first israel condition . anticipating this result ,
the above equations allow us to define in the standard way the brane perturbations @xmath365 , @xmath366 , etc . , @xmath367 using the standard @xmath2-dimensional perturbation theory , we construct the two bardeen potentials , as well as the brane vector and tensor metric perturbations , @xmath368 ( the two first equations are equivalent to eqns ( 5.21 ) of ref . @xcite . ) finally , using eq .
( [ def_q ] ) , the perturbed first fundamental form is @xmath369 notice that the @xmath359 do not appear in equations ( [ corina][corine ] ) : this is again related to the fact that they do not represent physical degrees of freedom .
these above expressions can also be obtained by starting from the definition @xmath370 paying attention that in the perturbed and unperturbed cases , the bulk metric is not evaluated at the same position ( @xmath371 and @xmath117 respectively ) @xcite . using eqns ( [ transmetdeb][transmetfin],[upstrans ] ) ,
the coordinate transformations of the following variables are obviously continuous as they do not involve derivatives with respect to @xmath57 : @xmath372 given the first fundamental form ( [ gina][corine ] ) , the first israel conditions imply that these quantities , which are linear combinations of the components ( [ gina ] ) to ( [ rona ] ) of the perturbed first fundamental form @xmath373 , are continuous , @xmath374 or , equivalently @xmath375 hence tensor perturbations are continuous and only the two scalar quantities @xmath336 , @xmath337 and the vector quantity @xmath343 may jump .
the first two equations are equivalent to @xmath376 which ensures that the brane bardeen potentials @xmath377 and @xmath378 are well defined .
we also see that the bulk vector and tensor perturbation @xmath342 and @xmath257 may be defined on the brane where they reduce to the standard vector and tensor metric perturbations of a @xmath2-dimensional spacetime with maximally symmetric spacelike hypersurfaces .
equivalently , the corresponding first fundamental form can be rewritten as @xmath379 which indeed reduce to @xmath380 , @xmath381 and @xmath382 , in longitudinal gauge ( @xmath383 ) for scalar perturbations and in the gauge @xmath384 for vector perturbations .
so far we have given the relationship between the intrinsic brane metric perturbations , the brane displacement @xmath345 , and some of the bulk metric perturbations .
things become a little bit more involved when we consider the other bulk metric perturbations that appear in @xmath385 .
as we already noticed , the first israel condition does not constrain @xmath280 ( see eq .
( [ corine ] ) ) .
nevertheless , since the transformation law for @xmath280 can be written as @xmath386 and since all the metric coefficients must remain finite , it follows that @xmath387 ( see also appendix [ sec_app_pert_bulk_geom ] ) . furthermore , as the coordinate transformation @xmath388 must be invertible , we also require @xmath389 . thus if @xmath107 is not continuous , then @xmath390 : @xmath391 this is the only additional requirement that one must impose on the coordinate system in the vicinity of the brane .
note that if @xmath1 symmetry is assumed , @xmath392 must be imposed even though @xmath107 is continuous @xcite . as mentioned in section [ ssec_israel_and ] ,
the only place where discontinuities or singularities are allowed is the brane position .
when @xmath393 , the brane is at @xmath117 .
however , if the brane position is perturbed , @xmath349 , the above requirement implies @xmath394 and hence @xmath107 is not allowed to jump if the brane position is perturbed .
if the unperturbed metric has a zeroth order discontinuity in the coefficient @xmath107 , the brane position must remain at @xmath393 to first order perturbation theory ( see figure [ fig4 ] ) .
this statement is in fact valid even in the absence of metric perturbations ( see appendix [ sec_app_pert_bulk_geom ] ) .
let us now consider the coordinate transformations of the variables @xmath395 and @xmath396 given in eqns ( [ rona],[brigitta ] ) , @xmath397 the first israel condition also states that @xmath398 therefore , in order for the transformations ( [ t0perp],[tiperp ] ) to be continuous we have to imply that a valid coordinate change satisfies @xmath399 these conditions ensure that @xmath400 and @xmath401 admit second derivatives .
this standard requirement for any valid coordinate transformation is therefore preserved even in the presence of a brane . in particular ( see eqns [ transabbe],[transe ] ) ) , this means that _ if _ @xmath402 or @xmath403 are discontinuous ( which is allowed ) , then @xmath299 and @xmath300 _ can not _ be transformed identically to zero by coordinate changes .
this does not prevent the quantities @xmath228 , @xmath229 , @xmath336 , @xmath337 from being well defined .
however , it does mean that there may not be a coordinate system in which eqns ( [ newtdeb][newtfin ] ) are valid ( which we never needed to suppose ) .
the first israel condition does not require the continuity of @xmath336 and @xmath343 , see eqns ( [ isrvp],[isrhi ] ) below . finally , from @xmath404 and using the fact that @xmath387 , the jump @xmath405 is gauge invariant .
therefore the gauge invariant quantity @xmath337 given in eqn ( [ vhpp ] ) may contain a singular part , @xmath406 which again shows that @xmath337 can not always be a component of the perturbed metric tensor in the vicinity of the brane .
of course , this does not invalidate the results found previously , but simply suggests that other variables may be more suitable to describe the metric perturbations in the vicinity of the brane .
after these remarks on the regularity requirements of gauge transformations in a bulk - brane system , we can now define some further gauge invariant scalar variables in terms of which we will express the second israel condition .
they will also be used to write the einstein equations for an observer on the brane when we want to compare our results with the usual @xmath2-dimensional cosmological perturbation theory .
let us first define the gauge invariant combination @xmath407 since @xmath301 can be discontinuous , @xmath408 is defined on each side of the brane .
note that we have @xmath409 furthermore , we set @xmath410 when @xmath107 is continuous , @xmath411 can be interpreted as the `` gauge invariant brane position '' that is , the position of the brane unambiguously defined when @xmath412 is set to zero by a suitable coordinate change ( which always exists _ if _
@xmath107 is continuous ) . in principle , derivatives normal to the brane are not defined for brane variables .
but in what follows we will also use @xmath413 which we simply define as @xmath414 in other words , the operator @xmath65 acts on every metric perturbation defined in the bulk but not on @xmath345 ( i.e. , we define @xmath415 ) . along similar lines
, one can also define @xmath416 and @xmath417 .
the quantity @xmath413 can contain a singular term because of the discontinuous part of @xmath418 .
therefore , it is useful to define @xmath419 by @xmath420 this new quantity can take different values on each side of the brane , so that we can define @xmath421 and @xmath422 following eqns ( [ discbrn],[contbrn ] ) .
we may not simply continue @xmath345 into the bulk as a variable @xmath345 is independent of @xmath57 .
this is a gauge dependent continuation and the definitions for @xmath423 and @xmath416 given above would be valid only in the gauge where @xmath345 is independent of @xmath57 . from the above expressions it is also clear that the variables @xmath423 and @xmath416 and all brane perturbation variables which contain these derivatives , like e.g. @xmath424 below , are gauge invariant only with respect to gauge transformations parallel to the brane .
therefore , it is important to keep in mind that these quantities are defined only on ( each side of ) the brane , although the notation @xmath413 , @xmath419 may be slightly misleading .
they simply refer to eqns ( [ upnor1],[upnor2 ] ) .
using equation ( [ defups ] ) , one can define several gauge invariant quantities which can also only be evaluated on either side of the brane ( that is either at @xmath425 or at @xmath426 ) : @xmath427 by comparing the last equation to eq .
( [ singhpp ] ) , it appears that @xmath424 does not contain a singular term ( and hence it is a quantity that has a meaning on each side of the brane ) .
using eqns ( [ bardeeepsi],[bardeeephi ] ) we have @xmath428 the derivatives @xmath429 and @xmath430 are defined via eq .
( [ upnor1],[upnor2 ] ) above .
the above equations will become very useful when considering the second israel conditions and writing down the perturbed weyl tensor .
the first israel condition states that @xmath377 and @xmath378 are defined on the brane , and that @xmath430 , @xmath429 , @xmath431 , and @xmath424 are well - defined on both sides of the brane , but it does not imply their continuity . in fact , it is their discontinuity which will enter into the perturbed second israel condition . using the above definitions we have the following relations for the discontinuous and the continuous parts of the gauge invariant scalar perturbation variables : @xmath432 and @xmath433 if @xmath1 symmetry is assumed , the first of these relations reduce to @xmath434 and @xmath435 these relations are very important .
they allow us to move freely between the bulk ( non underlined ) perturbation variables , which are defined everywhere , and the brane - related ( underlined ) variables , which are well defined only on the brane , i.e. they are either defined _ on _ the brane , like @xmath377 and @xmath378 , or _ on both sides _ of the brane , like @xmath436 , @xmath430 , @xmath437 , @xmath431 , or @xmath424 , etc .
the main difference between the brane and bulk perturbation variables is @xmath408 which appears in the former .
the brane displacement , however , is not unrelated to the bulk metric perturbations : a displacement of the brane induces metric perturbation in the bulk ( as one could have guessed by making an analogy with a charged surface in electromagnetism ) .
we define the perturbed stress - energy tensor on the brane as @xmath438 where @xmath439 is the perturbation of the energy velocity on the brane which is given by @xmath440 ( the @xmath441 component is determined by the condition @xmath442 . )
the variable @xmath443 is the anisotropic stress tensor and it is gauge invariant by itself .
as discussed in appendix [ ssec_app_pert_stt ] , we define the gauge invariant perturbations for the energy density and the pressure on the brane by @xmath444 similarly we define the gauge invariant perturbation variables for the velocity on the brane , @xmath445 to impose the second israel junction condition , we need to compute @xmath446 which is the difference between the perturbed value of @xmath25 at the brane position @xmath447 , and the background value of @xmath25 .
the results are given in appendix [ ssec_app_pert_cex ] .
the extrinsic curvature has to be compared to the perturbation of the surface stress tensor , @xmath448 whose components are given in appendix [ ssec_app_pert_stt ] .
the perturbation of the second israel condition , @xmath449 yields four discontinuity conditions for the scalar perturbation variables @xmath431 , @xmath424 , @xmath450 , and the first derivatives @xmath451 and @xmath452 , @xmath453 in terms of the gauge invariant bulk variables these conditions read @xmath454 notice that when @xmath1 symmetry is imposed @xmath455 never appears in these equations . for the vector perturbation variables we obtain two discontinuity conditions for @xmath343 and the first derivative @xmath456 , @xmath457 finally , there is also a discontinuity condition for the normal derivative of the tensor perturbation variable @xmath257 : @xmath458 note that the israel conditions do not give any constraint on the @xmath459 components of eq .
( [ pertks ] ) .
the above constraints can also be found directly from the singular part of einstein s equations ( [ einpert00][einpertijt ] ) .
this is relatively straightforward for the vector and tensor modes , but much more involved for the scalar part , as one must rewrite the equations using the underlined quantities defined above , and also because one has to consider the perturbation of the covariant dirac function @xmath38 .
however , for completeness , this has been undertaken in appendix [ ssec_app_pert_newee ] , and we have checked that both approaches lead to the same result . as we have seen , the junction conditions are conveniently written using the underlined variables @xmath377 , etc . in order to use them , we must know @xmath436 . the jump of this quantity , @xmath450 , is given by eq .
( [ isrpi ] ) . as
its continuous part @xmath455 represents the brane displacement , it is natural to seek an equation describing the brane motion . as for the unperturbed case ,
such an equation is found by taking the discontinuous part of the @xmath460 component of einstein s equations .
this yields @xmath461 where @xmath462 corresponds to the pressure perturbation along the extra dimension as measured by an observer at rest with respect to the brane .
the relationship between @xmath463 and @xmath462 is given in eq .
( [ udtpresy ] ) below .
equation ( [ sailpert ] ) is the typical equation for the displacement of a membrane ( it involves the laplacian of the displacement @xmath455 ) . when going back to the bulk ( non underlined ) perturbations , eq .
( [ sailpert ] ) becomes , as expected , a wave equation for @xmath455 : @xmath464 note that there is nothing which guarantees _ a priori _ that the motion of the brane is stable . even in the simplest case ( @xmath1 symmetry , @xmath276 , no bulk perturbation contributing to the right hand side of the above equation and brane stress - energy tensor dominated by a constant tension term ) , this equation becomes @xmath465 and
the mass term becomes negative for sufficiently fast expansion rate of the brane !
in this paper we have derived gauge invariant cosmological perturbation theory in braneworld scenarios with one codimension .
the unperturbed background system we considered ( sections [ secii][seciv ] ) consists of a @xmath0-dimensional bulk spacetime with a maximally symmetric @xmath7-dimensional subspace of curvature @xmath4 , containing arbitrary ( possibly interacting ) matter with energy - momentum tensor @xmath67 , and a homogeneous and isotropic @xmath7-brane again with arbitrary stress energy tensor @xmath114 .
we have not assumed @xmath1 symmetry across the brane . as such , our work generalises that of previous authors who have considered perturbation theory mainly in the @xmath1-symmetric case , and with specific bulk ( and brane ) matter ( e.g. , a bulk cosmological constant @xcite or scalar field @xcite ) .
we believe that the general setup considered here is a necessary component of any serious attempt which may be made to tackle such important questions as the cosmic microwave background anisotropies in braneworlds .
the only coordinate choice we have made is to fix the unperturbed brane to be at a given position @xmath215 in the extra dimension .
the bulk metric is explicitly time - dependent .
when the bulk contains only a cosmological constant , this is not the most natural coordinate system : there one would work with ( static ) schwarzschild - ads@xmath481 and a dynamical brane @xcite .
however , in the case of arbitrary bulk matter and especially for the study of perturbations , we have found it more convenient to work in a coordinate system in which the brane is at rest . in sections [ secii][seciv ] we derived all the relevant background equations , ending with the brane friedmann equation ( [ bf1][bf2 ] ) . as discussed in section [ seciv ] , when @xmath1 symmetry is not assumed , one has additional contributions to the @xmath2-dimensional einstein tensor on the brane .
in order to study these terms one has to include equations for the extrinsic curvature . in the remainder of the paper we studied perturbations of this system , setting up a completely gauge invariant formalism .
section [ secvi ] contains a general discussion of the classification of perturbations in an @xmath482-dimensional space time , as well as the interplay between bulk and brane perturbations .
an important point which we note there is the existence of one extra scalar degree of freedom on the brane which is not a metric perturbation ( although it interacts with some bulk metric perturbations ) : this is the brane displacement . in section [ secvii ] we have derived an equation of motion for the gauge invariant perturbation variable describing this quantity . in section [ secvi ]
we introduced the perturbed @xmath0-dimensional bulk spacetime .
this led to the definition of four scalar , two vector and one tensor gauge invariant bulk perturbation variables given in equations ( [ vpsi][vhci ] ) . following the definition of gauge invariant variables for the perturbations of the bulk matter
, we were able to write down the perturbed bulk einstein equations in a gauge invariant manner .
the perturbed brane was then introduced in section [ secvii ] . in analogy with usual @xmath2-dimensional cosmological perturbation theory ,
our aim was to introduce two scalar gauge invariant brane perturbation variables ( the bardeen potentials ) , one vector and one tensor metric perturbation .
the correct definition of these variables can only be given once the perturbed brane metric and israel junction conditions are used determine the brane variables in terms of the continuous part and the jump of the bulk perturbations .
the brane variables are defined in equations ( [ corina][corina2 ] ) .
the gauge invariant brane displacement also enters in these definitions .
finally the perturbed einstein equations on the brane were derived in section [ secix ] . as in the unperturbed case
, they contain a contribution from the projection of the perturbed bulk weyl tensor which in general have to be determined by solving the bulk equations . despite the fact that we have tried to present our results as clearly as possible , the formalism presented in this paper is technically rather complicated .
this reflects the fact that we have considered a very general scenario .
the corollary is , however , that our results should be applicable to a whole variety of different ( and possibly simpler ) situations of interest in braneworld scenarios . in a forthcoming paper ,
we plan to apply this formalism to a specific model and solve some of the perturbation equations presented here .
we would like to thank francis bernardeau , timon boehm , philippe brax , david langlois , roy maartens and jean - philippe uzan for enlightening discussions .
thanks chlo riazuelo for pointing out an error in the perturbed weyl tensor ( [ cloclo ] ) . r.d . and f.v .
thank the intitute for advanced study for hospitality .
f.v . and d.s .
thank the swiss foundation ernst et lucie schmidheiny for travel support .
is grateful to the rockefeller foundation for wonderful hospitality in bellagio , italy , during the final stages of this work .
this work is supported by the swiss national science foundation .
100 j. polchinski , _ string theory _ ( cambridge universty press , cambridge , england , 1998 ) , volumes 1 and 2 .
p. horava and e. witten , nucl .
* b640 * , 506 ( 1996 ) .
p. horava and e. witten , nucl .
* b675 * , 94 ( 1996 ) .
k. akama , _ pregeometry _ , lecture notes in physics * 176 * , edited by k. kikkawa , n. nakanishi , and h. nariai ( springer - verlag , berlin , 1982 ) , p. 267 .
rubakov and m.e .
shaposhnikov , b * 125 * , 136 ( 1983 ) .
m. visser , b * 159 * , 22 ( 1985 ) .
squires , b * 167 * , 286 ( 1986 ) .
n. arkani - hamed , s. dimopolous , and g. dvali , b * 429 * , 263 ( 1998 ) .
i. antoniadis , n. arkani - hamed , s. dimopolous , and g. dvali , b * 436 * , 257 ( 1998 ) .
l. randall and r. sundrum , , 3370 ( 1999 ) .
l. randall and r. sundrum , , 4670 ( 1999 ) .
p. bintruy , c. deffayet , and d. langlois , nucl .
* b565 * , 269 ( 2000 ) .
p. bintruy , c. deffayet , u. ellwanger , and d. langlois , b * 477 * , 285 ( 1999 ) .
s. mukohyama , b * 473 * , 241 - 245 ( 2000 ) .
cline , c. grojean , and g. servant , , 4245 ( 1999 ) .
c. csaki , m. graesser , c. kolda , and j. terning , b * 462 * , 34 ( 1999 ) .
flannagan , s .- h .
tye , and i. wasserman , , 044039 ( 2000 ) .
j. lesgourgues , s. pastor , m. peloso , and l. sorbo , b * 489 * , 411 ( 2000 ) .
n. deruelle , t. doleel , and j. katz , , 083513 ( 2001 ) .
n. deruelle and j. katz , , 083515 ( 2001 ) .
n. deruelle and t. doleel , , 103506 ( 2001 ) .
j. garriga and t. tanaka , , 2778 - 2781 ( 2000 ) .
giddings , e. katz , and l. randall , jhep * 0003 * , 023 ( 2000 ) .
s. kachru , m. schulz , e. silverstein , , 045021 ( 2000 ) .
gonzalez - diaz b * 481 * , 353 ( 2000 ) .
davis , i.r .
vernon , s. davis , and w.b .
perkins , b * 504 * , 254 ( 2001 ) .
perkins , b * 504 * , 28 ( 2001 ) .
b. carter , int .
j. theor .
phys . * 40 * , 2099 ( 2001 ) .
battye and b. carter , b * b50 * , 331 ( 2001 ) .
b. carter and j .- p .
uzan , nucl .
* b606 * , 45 ( 2001 ) .
battye , b. carter , a. mennim , and j .-
uzan , , 124007 ( 2001 ) .
b. carter , j .-
uzan , r.a .
battye , and a. mennim , class .
quantum grav .
* 18 * , 4871 ( 2001 ) . c. van de bruck , m. dorca , c.j.a.p .
martins , and m. parry , b * 495 * , 183 ( 2000 ) .
t. shiromizu , k. maeda , and m. sasaki , , 024012 ( 2000 ) .
r. maartens , , 084023 ( 2000 ) . c. van de bruck , m. dorca , r. brandenberger , and a. lukas , , 123515 ( 2000 ) .
h. kodama , a. ishibashi , and o. seto , , 064022 ( 2000 ) .
s. mukohyama , , 084015 ( 2000 ) .
s. mukohyama , class .
quantum grav .
* 17 * , 4777 ( 2000 ) .
s. mukohyama , , 064006 ( 2001 ) .
d. langlois , , 126012 ( 2000 ) .
d. langlois , , 2212 ( 2001 ) .
a. neronov and i. sachs , b * 513 * , 173 ( 2001 ) . c. van de bruck and m. dorca , hep - th/0012073 .
h. bridgman , k. malik , and d. wands , , 043502 ( 2002 ) .
k. koyama and j. soda , , 123502 ( 2000 ) .
k. koyama and j. soda , , 023514 ( 2002 ) .
v. bozza , m. gasperini , and g. veneziano , nucl .
* b619 * , 191 ( 2001 ) . o. seto and h. kodama , , 123506 ( 2001 ) .
m. giovannini , , 064008 ( 2002 ) .
barrow and r. maartens , gr - qc/0108073 . j. khoury and r
zhang , hep - th/0203274 .
m. dorca and c. van de bruck , nucl .
* b605 * , 215 ( 2001 ) .
r. dick , b * 491 * , 333 ( 2000 ) .
p. bowcock , c. charmousis , and r. gregory , class .
quantum grav .
* 17 * , 4745 ( 2000 ) .
p. kraus , jhep * 9912 * , 011 ( 1999 ) .
d. ida , jhep * 0009 * , 014 ( 2000 ) .
s. mukohyama , t. shiromizu , and k. maeda , , 024028 ( 2000 ) . c. charmousis , r. gregory , and v. rubakov , , 067505 ( 2000 ) .
bardeen , , 1882 ( 1980 ) .
h. kodama and m. sasaki , prog .
phys . * 78 * , 1 ( 1984 ) .
stewart and m. walker , proc .
london , * a341 * ( 1974 ) . h.a .
bridgman , k.a .
malik , and d. wands , , 084012 ( 2001 ) .
myung , hep - th/0009117 w. israel , nuovo cimento
* b44 * , 1 ( 1966 ) ; erratum : _ ibid . _ * b48 * , 463 ( 1967 ) .
wald , _ general relativity _ , ( the university of chicago press , chicago , 1984 ) .
b. carter and p. peter , b 466 , 41 ( 1999 ) .
einstein spaces _ , ( pergamon press , oxford , 1969 ) . b. carter , b * 228 * , 446 ( 1989 ) . w. misner , k.s .
thorne , and j.a .
wheeler , _ gravitation _ , ( freeman , new york , 1970 ) .
in this appendix , we give all the necessary formulae that were used to obtain the results presented in the text . here we will consider an @xmath483-dimensional bulk with , @xmath44-dimensional maximally symmetric , spacelike hypersurfaces of constant curvature @xmath4 .
therefore , here @xmath111 , @xmath10 , @xmath484 , @xmath485 , where @xmath486 , and @xmath487 , @xmath11 , @xmath484 , @xmath44 .
we will consider both the unperturbed ( section [ sec_app_bulk_geom][sec_app_brane_pov ] ) and perturbed cases ( section [ sec_app_pert_bulk_geom][sec_app_pert_brane_pov ] ) .
both the bulk and the brane matter content are arbitrary as well as the global geometry of the bulk .
furthermore , we do not assume @xmath1 symmetry .
following the definitions of @xcite , we use the sign convention @xmath488 , that is the signature of the metric is @xmath489 , and the riemann and ricci tensors are respectively defined by @xmath490 the weyl tensor is defined by @xmath491 the induced metric is defined by @xmath492 the metric can be projected back in the bulk to give the first fundamental form @xmath493 more generally , any tensor @xmath494 defined for the brane can be projected back in the bulk using @xmath495 in particular , for the brane riemann tensor , one has @xmath496 one can also define the normal spacelike unit vector @xmath132 to the brane according to @xmath497 and the bulk metric evaluated at the brane position can be split into @xmath498 with @xmath499 note that eq .
( [ app_norm_vec_def ] ) implies @xmath500 one can also define the extrinsic curvature according to @xmath501 which obeys the following relations @xmath502 for the bulk matter , it is more convenient to introduce the unit vector @xmath518 which represents the bulk @xmath45-velocity of the fluid : @xmath519 here @xmath77 represents the lorentz boost which must be performed along the @xmath57 axis in order to be in the rest frame of the bulk matter .
as usual @xmath79 . due to the symmetries of spacetime , the stress energy tensor of any component possesses @xmath44 identical eigenvalues @xmath74 .
the other eigenvalues are @xmath71 ( associated to the timelike eigenvector @xmath69 ) and @xmath73 ( associated to the spacelike eigenvector @xmath72 ) .
one has @xmath520
@xmath521 one first builds the unit vector orthogonal to the brane : @xmath532 the components of this vector have the same functional form but possibly different numerical values when evaluated at @xmath533 and @xmath534 .
then the induced metric is given by : @xmath535 @xmath536 for any quantity @xmath118 , we define @xmath537 where @xmath538 is the discontinuity of @xmath118 , @xmath124 is the continuous part of @xmath118 and @xmath120 is the heaviside function .
the first israel condition states that @xmath539 : @xmath540 in particular , this means that @xmath107 is allowed to be discontinuous at the brane position .
also , the derivatives of @xmath106 and @xmath105 with respect to @xmath57 can be discontinuous .
formally , one can take a stress - energy tensor of the above form to describe the brane content , provided that we have @xmath543 with @xmath544 the condition @xmath99 and @xmath98 has not necessarily to be satisfied but is a consequence of the coordinate choice to put the brane at rest with respect to the coordinate system . because of the dirac term ( [ def_beta ] ) , @xmath545 and @xmath546 depend on @xmath547 only .
since the stress energy tensor of the brane is strictly zero elsewhere , its eigenvectors are not defined outside this hypersurface .
the vector @xmath24 appears as the analog of the vector @xmath72 as an eigenvector associated to the eigenvalue @xmath548 .
equivalently one can define an @xmath45-velocity @xmath142 which corresponds to the eigenvector associated with the eigenvalue @xmath71 .
therefore , @xmath549 taking the continuous part of the einstein equation at the brane position , we get @xmath558 as such , these equations are not yet very useful because they involve many terms which are not explicit ` brane variables ' .
these can be found either by taking the singular part of ( [ back_cons_bulk ] ) , or by considering the discontinuity of ( [ back_ee_05 ] ) . a bulk energy exchange term @xmath565 can have a singular component @xmath566 so that @xmath567 .
@xmath568 one also has the sail equation @xmath569 all @xmath7-vectors indices are raised and lowered using metric @xmath572 .
@xmath17 represents its associated covariant derivative and @xmath573 .
barred vectors are divergenceless , double barred tensors are divergenceless and traceless with respect to @xmath16 and @xmath17 .
there is one subtlety due to the fact that @xmath107 may be discontinuous .
we consider the above infinitesimal coordinate transformation . for the @xmath578 component
, we have @xmath579 for the coordinate change to be valid , the metric components must remain finite , therefore one _ must _ have @xmath580 = 0 .\ ] ] if @xmath107 is continuous , then @xmath352 can be an arbitrary ( continuous ) coordinate transformation , but if @xmath107 is discontinuous , then @xmath581 .
geometrically , this is related to the fact that the coordinate system is allowed to exhibit some pathologies only at the brane position , but not in its vicinity . from now on ,
we shall write any perturbation variables mostly with the gauge invariant quantities defined above . the non - gauge invariant terms will be put under the form of some factors involving components of the corresponding unperturbed tensor multiplied by @xmath299 , @xmath300 and @xmath301 .
@xmath592 \nonumber \\ & & + ( r_{0 i 0 k } \nabla_j + r_{0 i j 0 } \nabla_k ) \abbe - ( r_{0
i \finn k } \nabla_j + r_{0 i j \finn } \nabla_k ) \anee \nonumber \\ & & + r_{l i j k } \left(\nabla^l \dot e - \ibza \vi^l \right ) , \\
\delta r_{0 i 0 \finn } & = & - \ibb ( \pp + \iperp - \hperp ) \nabla_i \psi - \ibb \iperp \nabla_i \hpp + \ddemi \ibn ( \pe + 2 \uconf - \hconf ) \nabla_i \vp \nonumber \\ & & + \ddemi \iab ( \pe + 2 \hconf - \uconf ) \left ( ( \pp + \uperp - \hperp ) \vi_i \right ) - \ddemi \ina ( \pe + 2 \hconf - \iconf ) \left ( ( \pe + \iconf - \hconf ) \hci_i \right ) \nonumber \\ & & - r_{0 \finn 0 \finn } \nabla_i \anee + r_{0 i j \finn } \left(\dot e^j - \ibza \vi^j \right ) + r_{0 i 0 j } \left(e^j { } ' - \inza \hic^j \right ) , \\ \delta r_{0
i \finn j } & = & \iaa \gamma_{i j } ( \hperp \dot \hpp - \hconf \psi ' ) - \ddemi \left ( \ibn \nabla_{i j } - \inaazb \gamma_{i j } \hconf \uconf + \iaabzn \gamma_{i j } \hperp \iperp \right ) \vp \nonumber \\ & & + \iaa ( \pe \pp + ( \hconf - \uconf ) \pp + ( \hperp - \iperp ) \pe ) ( \tt{e}_{i j } - \gamma_{i j } \phi ) \nonumber \\ & & + \ddemi \iab ( \pp + \hperp - \iperp ) \nabla_{(i } \vi_{j ) } + \ddemi \ina ( \pe + \hconf - \uconf ) \nabla_{(i } \hci_{j ) } \nonumber \\ & & + ( \dot r_{0
i \finn j } + r_{0 i \finn j }
\pe + r_{0 i 0 j } \pp ) \abbe \nonumber \\ & & - ( r'_{0 i \finn j } + r_{0 i \finn j }
\pp + r_{\finn i \finn j } \pe ) \anee \nonumber \\ & & + r_{0 k \finn j } ( e^k_i - \delta^k_i \phi ) + r_{0 i \finn k } ( e^k_j - \delta^k_j \phi ) , \\
\delta r_{i j k l } & = & \left ( \iaaaazbb \hconf^2 \psi + \iaaaaznn \hperp^2 \hpp - \iaaaazbn \hconf \hperp \vp \right ) \gamma_{i k } \gamma_{j l } - \gamma_{k i } \left ( \iaaazb \hconf \nabla_{(j } \vi_{l ) } - \iaaazn \hperp \nabla_{(j } \hci_{l ) } \right ) \nonumber \\ & & + \left ( \iaa ( \nabla_{k i } + k \gamma_{k i } ) - \iaaaazbb \gamma_{k i } \hconf \pe + \iaaaaznn \gamma_{k i } \hperp \pp \right ) ( \tt{e}_{j l } - \gamma_{j l } \phi ) \nonumber \\ & & - [ i \leftrightarrow j ] + [ ik \leftrightarrow jl ] - [ k \leftrightarrow l ] \nonumber \\ & & + r_{m j k l } ( e^m_i - \delta^m_i \phi ) + r_{i m k l } ( e^m_j - \delta^m_j \phi ) + r_{i j
m l } ( e^m_k - \delta^m_k \phi ) + r_{i j k m } ( e^m_l - \delta^m_l \phi ) \nonumber \\ & & + \dot r_{i j k l } \abbe - r'_{i j k l } \anee , \\ \delta r_{i j 0 \finn } & = & \nabla_i \left ( \iab ( \pp + \iperp - \hperp ) \vi_j - \ina ( \pe + \uconf - \hconf ) \hci_j \right ) - [ i \leftrightarrow j ] , \\
\delta r_{\finn i j k } & = & \iaa \nabla_j \pp ( \tt{e}_{i j } - \gamma_{i j } \phi ) + \iaa \hperp \gamma_{i k } \nabla_j \hpp + \ina \nabla_j \nabla_{(i } \hci_{k ) } \nonumber \\ & & - \ddemi \iaa \hconf \gamma_{i k } \left ( \inzb \nabla_j \vp + \iazb ( \pp + \iperp - \hperp ) \vi_j - \inazbb ( \pe + \uconf - \hconf ) \hci_j \right ) \nonumber \\ & & - [ j \leftrightarrow k ] \nonumber \\ & & + ( r_{\finn i 0 k }
\nabla_j + r_{\finn i j 0 } \nabla_k ) \abbe - ( r_{\finn i \finn k } \nabla_j + r_{\finn i j \finn } \nabla_k ) \anee \nonumber \\ & & + r_{l i j k } \left(\nabla^l e ' - \inza \hic^l \right ) , \\
\delta r_{0 \finn 0 \finn } & = & - \ibb ( \pp - \uperp ) \pp \psi - 2 \ibb ( \pp + \iperp - \uperp ) ( \iperp \psi ) - \ibb \iperp \pp \hpp \nonumber \\ & & - \inn ( \pe - \iconf ) \pe \hpp - 2 \inn ( \pe + \uconf - \iconf ) ( \uconf \hpp ) - \inn \uconf \pe \psi \nonumber \\ & & + \ibn \left(\pe ( \tdemi \pp + \iperp ) + \pp ( \tdemi \pe + \uconf ) \right ) \vp \nonumber \\ & & + ( \dot r_{0 \finn 0 \finn } + 2 r_{0 \finn 0 \finn } \pe ) \abbe - ( r'_{0 \finn 0 \finn } + 2 r_{0 \finn 0 \finn } \pp ) \anee , \\
\delta r_{\finn 0 \finn i } & = & - \inn ( \pe + \uconf - \hconf ) \nabla_i \hpp - \inn \uconf \nabla_i \psi + \ddemi \ibn ( \pp + 2 \iperp - \hperp )
\nabla_i \vp \nonumber \\ & & - \ddemi \iab ( \pp + 2 \hperp - \uperp ) \left ( ( \pp + \iperp - \hperp ) \vi_i \right ) + \ddemi \ina ( \pp + 2 \hperp - \iperp ) \left ( ( \pe + \uconf - \hconf ) \hci_i \right ) \nonumber \\ & & + r_{\finn 0 \finn 0 } \nabla_i \abbe + r_{j 0 \finn i } \left(e^j { } ' - \inza \hic^j \right ) + r_{\finn j
\finn i } \left(\dot e^j - \ibza \vi^j \right ) , \\
\delta r_{\finn i \finn j } & = & \left(- \inn \nabla_{i j } + \iaa \hperp \gamma_{i j } \pp + \innaazbb \hconf \gamma_{i j } \pe \right ) \hpp + 2 \innaazbb \hconf \uconf \gamma_{i j } ( \psi + \hpp ) \nonumber \\ & & - \inaazb \gamma_{i j } ( \hconf \pp + \hconf \iperp + \uconf \hperp ) \vp - \innazb \uconf \nabla_{(i } \vi_{j ) } + \ina ( \pp + \hperp ) \nabla_{(i } \hci_{j ) } \nonumber \\ & & + \left(\iaa ( \pp^2 + ( 2 \hperp - \uperp ) \pp ) -
\innaazbb \uconf \pe \right ) ( \tt{e}_{i j } - \gamma_{i j } \phi ) \nonumber \\ & & ( \dot r_{\finn
i \finn j } + 2 r_{0 i \finn j } \pe + 2 r_{\finn i 0 j } \pe ) \abbe \nonumber \\ & & - ( r'_{\finn
i \finn j } + 2 r_{\finn i \finn j }
\pp ) \anee \nonumber \\ & & + r_{\finn k \finn j } ( e^k_i - \delta^k_i ) + r_{\finn
i \finn k } ( e^k_j - \delta^k_j ) .\end{aligned}\ ] ] defining @xmath593 @xmath594 \nonumber \\ & & + ( c_{0 i 0 k }
\nabla_j + c_{0 i j 0 } \nabla_k ) \abbe - ( c_{0 i \finn k } \nabla_j + c_{0 i j \finn } \nabla_k ) \anee \nonumber \\ & & + c_{l i j k } \left(\nabla^l \dot e - \ibza \vi^l \right ) , \\
\delta c_{0 i 0 \finn } & = & - \frac{n - 1}{n } \ibb \nabla_i \left(\pp ( \phi + \psi ) + ( \iperp - \hperp ) ( \hpp + \psi ) \right ) + \frac{n - 1}{n } \ibn \left(\tdemi \pezb + \uconf - \hconf \right )
\nabla_i \vp \nonumber \\ & & - \ddemi \frac{1}{n } \ibbnza ( \nabla^2 + k(n - 1 ) ) \hci_i \nonumber \\ & & + \ddemi \frac{n - 1}{n } \left ( \iab ( \pe + \hconf - \uconf ) \left ( ( \pp + \iperp - \hperp ) \vi_i \right ) - \ina ( \pe + \hconf - \iconf ) \left ( ( \pe + \uconf - \hconf ) \hci_i \right ) \right ) \nonumber \\ & & - c_{0 \finn 0 \finn } \nabla_i \anee + c_{j \finn 0 i } \left(\dot e^j - \ibza \vi^j \right ) + c_{0 j 0 i } \left(e^j { } ' - \inza \hic^j
\right ) , \\ \delta
c_{0 i\finn j } & = & - \ddemi \ibn \left(\nabla_{i j } - \frac{1}{n } \nabla^2 \gamma_{i j } \right ) \vp - \frac{1}{n } \iaazbn c_{0 \finn 0 \finn } \gamma_{i j } \vp \nonumber \\ & & + \ddemi \iab ( \pp + \hperp - \iperp ) \nabla_{(i } \vi_{j ) } + \ddemi \ina ( \pe + \hconf - \uconf ) \nabla_{(i } \hci_{j ) } \nonumber \\ & & + \iaa \left ( ( \tdemi \pp + \hperp - \iperp ) \pe + ( \tdemi \pe + \hconf - \uconf ) \pp \right ) \tt{e}_{i j } \nonumber \\ & & + ( \dot c_{0
i \finn j } + c_{0 i \finn j } \pe + c_{0
i 0 j } \pp ) \abbe - ( c'_{0
i \finn j } + c_{0 i \finn j } \pp + c_{\finn i \finn j } \pe ) \anee \nonumber \\ & & + c_{0 k \finn j } ( e^k_i - \delta^k_i \phi ) + c_{0 i \finn k } ( e^k_j - \delta^k_j \phi ) , \\ \delta c_{i j k l } & = & - \frac{1}{n } \iaa \left(\nabla_{i k } - \frac{1}{n } \nabla^2 \gamma_{i k } \right ) \gamma_{j l } ( 2 \phi + \psi - \hpp ) \nonumber \\ & & - \frac{1}{n^2 ( n + 1 ) } \iaa \gamma_{i k } \gamma_{j l } \left(\nabla^2 + k n \right ) ( 2 \phi + \psi - \hpp ) \nonumber \\ & & + c_{i j k l } ( \hpp - \psi ) + \iaa \left ( \nabla_{k i } + k \gamma_{k i } - \frac{1}{n } \gamma_{k i } ( \nabla^2 - 2 k ) \right ) \tt{e}_{j l } \nonumber \\ & & - \frac{1}{n ( n + 1 ) } \gamma_{i k } \gamma_{j l } \left(\iaaaazbb \xz + \iaaaaznn \xp + \iaaaazbn \vp \right ) \nonumber \\ & & + \iaaaazbb \frac{1}{n } \gamma_{i k } \eza_{j l } - \iaaaaznn \frac{1}{n } \gamma_{i k } \epa_{j l }
\nonumber \\ & & - [ i \leftrightarrow j ] + [ ik \leftrightarrow jl ] - [ k \leftrightarrow l ] \nonumber \\ & & + c_{m j k l } ( e^m_i - \delta^m_i \phi ) + c_{i m k l } ( e^m_j - \delta^m_j \phi ) + c_{i j
m l } ( e^m_k - \delta^m_k \phi ) + c_{i j k m } ( e^m_l - \delta^m_l \phi ) \nonumber \\ & & + \dot c_{i j k l } \abbe - c'_{i j k l } \anee , \\ \delta c_{i j 0 \finn } & = & \nabla_i \left ( \iab ( \pp + \iperp - \hperp ) \vi_j - \ina ( \pe + \uconf - \hconf ) \hci_j \right ) - [ i \leftrightarrow j ] , \\ \delta
c_{\finn i j k } & = & - \frac{1}{n } \gamma_{i k } \nabla_j \left ( \iaa \pp ( \phi + \psi ) + \iaa ( \iperp - \hperp ) ( \psi + \hpp ) - \inaazb ( \tdemi \pe + \uconf - \hconf ) \vp
\right ) \nonumber \\ & & + \ddemi \frac{1}{n } \gamma_{i k } \left ( \iaaazb ( \pe + \hconf - \uconf ) \left ( ( \pp + \iperp - \hperp ) \vi_j \right ) - \inaaazbb ( \pe + \hconf - \iconf ) \left ( ( \pe + \uconf - \hconf ) \hci_j \right ) \right ) \nonumber \\ & & + \ddemi \frac{1}{n } \gamma_{i k } \ina ( \nabla^2 + k ( n - 1 ) ) \hci_j + \ina \nabla_j \nabla_{(i } \hci_{k ) } + \iaa \nabla_j \tt{e}{}'_{i k } \nonumber \\ & & - [ j \leftrightarrow k ] \nonumber \\ & & + ( c_{\finn i 0 k }
\nabla_j + c_{\finn i j 0 } \nabla_k ) \abbe - ( c_{\finn i \finn k } \nabla_j + c_{\finn i j \finn } \nabla_k ) \anee \nonumber \\ & & + c_{l i j k } \left(\nabla^l e ' - \inza \hic^l \right ) , \\
\delta c_{0 \finn 0 \finn } & = & \ibbnnzaa \frac{n - 1}{n ( n + 1 ) } \left(\nabla^2 + k n \right ) ( 2 \phi + \psi - \hpp ) \nonumber \\ & & + \frac{n - 1}{n + 1 } \left(\inn \xz + \ibb \xp + \ibn \vp \right ) \nonumber \\ & & + ( \dot c_{0 \finn 0 \finn } + 2 c_{0 \finn 0 \finn } \pe ) \abbe - ( c'_{0 \finn 0 \finn } + 2 c_{0 \finn 0 \finn } \pp ) \anee , \\ \label{cloclo } \delta c_{\finn 0 \finn i } & = & \frac{n - 1}{n } \inn \nabla_i \left ( \pe(\phi - \hpp ) + ( \hconf - \uconf ) ( \psi + \hpp ) + ( \tdemi \pp + \iperp - \hperp ) \vp \right ) \nonumber \\ & & + \ddemi \frac{1}{n } \ibnnza ( \nabla^2 + k ( n - 1 ) ) \vi_i \nonumber \\ & & - \ddemi \frac{n - 1}{n } \left ( \iab ( \pp + \hperp - \uperp ) \left ( ( \pp +
\iperp -\hperp ) \vi_i \right ) - \ina ( \pp + \hperp - \iperp ) \left ( ( \pe + \uconf - \hconf )
\hci_i \right ) \right ) \nonumber \\ & & + c_{\finn 0 \finn 0 } \nabla_i \abbe + c_{j 0 \finn i } \left(e^j { } ' - \inza \hic^j \right ) + c_{\finn
j \finn i } \left(\dot e^j - \ibza \vi^j \right ) , \\ \delta c_{\finn i \finn j } & = & - \inn \left(\nabla_{i j } - \frac{1}{n } \nabla^2 \gamma_{i j } \right ) \left(\frac{n - 1}{n } ( \hpp - \phi ) + \frac{1}{n } ( \phi + \psi )
\right ) \nonumber \\ & & + \inn \frac{n - 1}{n^2 ( n + 1 ) } \gamma_{i j } \left(\nabla^2 + k n \right ) ( 2 \phi + \psi - \hpp ) \nonumber \\ & & - c_{\finn i \finn j } ( \psi + \hpp ) - \inn \frac{1}{n } ( \nabla^2 - 2 k ) \tt{e}_{i j } \nonumber \\ & & + \frac{n - 1}{n ( n + 1 ) } \gamma_{i j } \left(\innaazbb \xz + \iaa \xp + \inaazb \vp \right ) \nonumber \\ & & + \innaazbb \left(\frac{1}{n } \eza_{i j } + \ezb_{i j } \right ) + \iaa \left(\frac{n - 1}{n } \epa_{i j } + \epb_{i j } \right ) \nonumber \\ & & + ( \dot c_{\finn
i \finn j } + 2 c_{\finn i \finn j } \pe ) \abbe - ( c'_{\finn
i \finn j } + 2 c_{\finn i \finn j } \pp ) \anee \nonumber \\ & & + c_{\finn k \finn j } ( e^k_i - \delta^k_i \phi ) + c_{\finn i \finn k } ( e^k_j - \delta^k_j \phi ) .\end{aligned}\ ] ] where @xmath317 is any ( @xmath45)-scalar quantity ( density @xmath71 , pressure @xmath74 , etc ) .
note : @xmath602 , @xmath603 , @xmath94 , @xmath92 , @xmath95 are _ not _ scalars as they are defined through the vector fields @xmath80 , @xmath82 which specifically depend on a coordinate choice . with the @xmath611 sign determined by the condition @xmath612 .
@xmath613 @xmath614 @xmath615 here , the components @xmath305 are arbitrary as the eigenvalues @xmath74 and @xmath73 are degenerate .
the expression chosen here is purely for convenience in order to simplify the following untilded components .
since @xmath24 plays the same role for the brane as @xmath72 for a bulk component , this means that formally , the quantities @xmath306 , @xmath627 can be defined for the brane ( we will note them @xmath628 and @xmath629 respectively ) at @xmath350 , @xmath630 or , equivalently , @xmath631 the perturbed stress - energy tensor on the brane is given in eqns ( [ tensorbranebeg][tensorbraneend ] ) .
formally , for any quantity @xmath648 defined on the brane , the quantity @xmath649 is a bulk scalar .
the quantity @xmath38 is also a scalar quantity of the bulk since one is allowed to consider the case @xmath650 .
its perturbation is @xmath651 note that this derivation is a bit formal : since @xmath280 and @xmath424 can be discontinuous , this expression and the next one are ill - defined , even in the case where @xmath107 is continuous . but again , all the pathological terms cancel each other when one writes the einstein equations , so that we consider that this is not a serious problem . using the formula ( [ lie_scal_bulk ] ) ,
it is possible to build the gauge invariant counterparts of both @xmath652 and @xmath653 : @xmath654 this last quantity is invariant under any infinitesimal reparametrization of the @xmath530 .
equivalently , one has , using ( [ gi_v_scal],[gi_v_vect ] ) , as well as the fact that @xmath99 for the brane , @xmath655 with these definitions , @xmath656 as this quantity is defined on the brane , it is more convenient to express it in term of the brane - related ( underlined ) metric perturbations instead of the bulk - related ( non underlined ) metric perturbations as it was the case for the weyl tensor .
@xmath660 @xmath661 @xmath662 @xmath663 we rewrite the perturbed einstein near the brane , that is near @xmath371 .
we first define some new bulk matter content perturbations by @xmath664 the terms proportional to @xmath408 come from the fact that we are considering the bulk perturbations at @xmath371 instead of @xmath117 .
the other terms come from the fact that the brane is not at rest with respect to the bulk coordinate system , and are a mere consequence of a lorentz boost with velocity @xmath665 along the @xmath57 axis .
going from the non underlined ( bulk ) metric perturbation to the underlined ( brane - related ) metric perturbations , the einstein equation simplify a little bit to @xmath666 @xmath667 @xmath668 ( a sum on all the brane and bulk species is implicitly assumed on the right hand side of these equations . ) one can easily check that the singular part of the above equation reduces to the second israel condition .
they transform into @xmath671 @xmath672 for the brane components , they are obtained by considering the discontinuity of the @xmath673 , @xmath674 components of einstein equation or by taking the singular part of the above equations . | we derive the gauge invariant perturbation equations for a @xmath0-dimensional bulk spacetime in the presence of a brane .
the equations are derived in full generality , without specifying a particular energy content of the bulk or the brane .
we do not assume @xmath1 symmetry , and show that the degree of freedom associated with brane motion plays a crucial role .
our formalism may also be used in the @xmath1 symmetric case where it simplifies considerably .
preprint numbers : spht - saclay t02/040 ; orsay - lpt-02 - 36 ; hep - th/0205 ? ? ? |
how long did the epoch of primordial star formation last ? this is an important and timely question , as a primary goal of next generation telescopes , such as the _ james webb space telescope _
( jwst ; e.g. gardner et al .
2006 ; windhorst et al .
2006 ) and the _
european extremely large telescope _
( e - elt ) , is to detect primordial galaxies and star clusters in the early universe . while these telescopes may allow to detect individual supernovae ( sne ) exploding within the first population ( pop ) iii star clusters and galaxies ( see e.g. wise & abel 2005 ; haiman 2008 ) , the emission from the stellar populations within the first dwarf galaxies at @xmath0 @xmath1 10 is likely too dim for even the jwst to detect ( e.g. ricotti et al . 2008 ; johnson et al .
instead , only very rare , massive primordial galaxies at such high redshifts , or dwarf galaxies and star clusters forming at lower redshift may be bright enough for detection ( see e.g. ciardi & ferrara 2001 ; scannapieco et al .
therefore , whether or not pop iii star clusters can be found by next generation telescopes critically depends on the amount of primordial star formation that takes place at relatively low redshift .
the hierarchical nature of structure formation dictates that more massive dark matter ( dm ) haloes are built up from smaller ones , and that more massive haloes generally have progenitors in which star formation may take place at earlier times . in order for a pop iii star cluster to form in a dm halo , star formation in its progenitor haloes may have to be completely suppressed , lest sne enrich the primordial gas with heavy elements ( but see e.g. jimenez & haiman 2006 ; pan & scalo 2007 ; dijkstra & wyithe 2007 ) .
exposure to a persistant photoionizing background , such as takes place in reionized regions of the universe at high redshifts , can suppress star formation in dm haloes with masses substantially higher than those in which the first stars ( e.g. abel et al .
2002 ; bromm et al .
2002 ; yoshida et al . 2003 ) and galaxies ( e.g. greif et al . 2008 ; bromm et al . 2009 ) form , completely preventing the infall of gas into haloes with circular velocities up to @xmath3 30 km s@xmath9 , corresponding to virial masses of @xmath1 3 @xmath10 10@xmath11 m@xmath12 at @xmath0 @xmath2 10 ( e.g. efstathiou 1992 ; thoul & weinberg 1996 ; see also kitayama & ikeuchi 2000 ; gnedin 2000 ; ciardi & ferrara 2005 ) .
if star formation in the progenitors of these more massive haloes can be suppressed by the photoionizing background during reionization , then these haloes can remain unenriched until the primordial gas collapses into them to form clusters of pop iii stars at relatively low redshifts , e.g. at @xmath0 @xmath2 6 .
the formation of pop iii stars at such low redshifts has been a possibility recently studied through the use of large - scale cosmological simulations . in their high - resolution simulations of cosmic reionization ,
shin et al .
( 2008 ) ( also trac & cen 2007 ) found that haloes with masses @xmath3 10@xmath11 m@xmath5 may host pop iii star formation at z @xmath2 6 , but their simulation did not resolve the collapse of minihaloes , i.e. haloes with virial temperature @xmath13 @xmath2 10@xmath14 k , which could have led to the enrichment of the gas at higher redshifts ( see also schneider et al .
2006 ; tornatore et al .
2007 ) . in another recent study ,
trenti et al . (
2009 ) model the enrichment of haloes with @xmath13 @xmath1 10@xmath14 k by their progenitors , and find that significant pop iii star formation may take place at @xmath0 @xmath2 6 , as well .
however , their study did not explicitly account for the suppression of star formation by photoionization , although , especially at redshifts @xmath0 @xmath1 10 ( dijkstra et al . 2004 ) , it is possible that star formation takes place in haloes with @xmath13 @xmath3 10@xmath14 k even in the presence of a photoionizing background . in related analytical work ,
wyithe & cen ( 2007 ) estimated the contribution of pop iii stars to the reionization of the universe , arguing that primordial star formation in neutral minihaloes may be suppressed until these haloes grow to be sufficiently massive , perhaps allowing for the formation of a significant fraction of pop iii stars up to the end of reionization at @xmath0 @xmath3 6 . in this work
, it was assumed that star formation in minihaloes was prevented by the photodissociation of molecules at very high redshifts , e.g. @xmath0 @xmath1 20 ( haiman et al . 1997 ) .
however , more recent work has shown that molecular cooling can operate in minihaloes despite the build - up of the photodissociating background radiation field ( e.g. wise & abel 2007 ; oshea & norman 2008 ; johnson et al .
2008 ; trenti & stiavelli 2009 ) , providing motivation to relax this assumption and consider alternate scenarios for the formation of pop iii stars at low redshifts . in the present work
, we account for star formation in minihaloes at z @xmath1 20 , as well as for the suppression of star formation in dm haloes during reionization , with the aim of evaluating the abundance and detectability of pop iii star clusters in reionized regions of the universe . in section 2 ,
we outline the requirements for pop iii star cluster formation in the post - reionization universe . in section 3
, we calculate an upper limit for the abundance of such clusters and estimate the effect that large - scale metal - enriched galactic winds have in preventing their formation , while in section 4 we discuss the prospects for their detection in future surveys .
we conclude with a discussion of our results in section 5 . for our calculations
, we assume the latest cosmological parameters reported by the _ wilkinson microwave anisotropy probe _ ( wmap ) collaboration ( komatsu et al .
the first stars are expected to have formed in dm minihaloes with virial temperature @xmath3 10@xmath15 k at redshifts @xmath0 @xmath1 20 ( see e.g. bromm & larson 2004 ; oshea & norman 2007 ) .
these stars were likely very massive , with a fraction of them ending their lives as powerful supernovae which enriched the intergalactic medium ( igm ) with the first heavy chemical elements ( e.g. bromm et al .
2003 ; kitayama et al . 2005 ; greif et al . 2007
; but see also whalen et al .
for a pop iii star cluster to form in a given dm halo , the formation of stars which explode as sne and enrich the gas may have to be completely prevented in its minihalo progenitors .
we take this as a first condition for pop iii star formation in the reionized universe . for simplicity
, we make the conservative assumption that no pop iii stars can form in the progenitors of a given halo in order for it to remain pristine , although in principle primordial stars which collapse to form black holes , instead of exploding as sne , may not enrich the gas in their host haloes ( see fryer et al .
2001 ; heger et al . 2003 ; schneider et al .
we discuss the related issue of the metal enrichment of haloes by large - scale winds from neighboring galaxies in section 3.2 .
a molecule - dissociating , so - called lyman - werner ( lw ) , background radiation field can delay star formation in dark matter ( dm ) minihaloes , but the gas eventually collapses when the dm halo becomes massive enough , of the order of 10@xmath16 m@xmath5 for very high lw flux ( shang et al .
2009 ; see also wise & abel 2007a ; oshea & norman 2008 ; mesinger et al . 2006 , 2009 ) . furthermore , as we show below , star formation in minihalo progenitors at @xmath0 @xmath1 19 , on average , must be suppressed for haloes forming in the reionized universe to avoid self - enrichment by these progenitors .
recent modeling of the build - up of the lw background suggests that the lw flux may remain too low to prevent star formation in all but the rarest minihalos at @xmath0 @xmath1 15 ( johnson et al .
2008 ; dijkstra et al . 2008 ; ahn et al .
2009 ; trenti & stiavelli 2009 ) .
therefore , we expect that lw feedback alone is not sufficient to prevent the collapse of gas into the progenitors of dm haloes with masses @xmath1 10@xmath11 m@xmath5 , the mass range for haloes in which star formation can take after reionization is complete at @xmath0 @xmath3 6 .
a more potent form of radiative feedback at high redshifts is the photoionizing background radiation that builds up during reionization . in regions which are reionized early , at e.g. @xmath0 @xmath1 15
, the temperature of the photoheated gas is too high , at @xmath1 10@xmath14 k , for minihaloes to retain gas , and star formation can be suppressed .
we note that for star formation to be suppressed by the photoionizing background during reionization , the minihalo must be subjected to the radiation before it reaches advanced stages of collapse , as an ionization front passing through a dense minihalo may act to induce star formation , not to prevent it ( e.g. ahn & shapiro 2007 ; whalen et al .
2008b ; susa et al .
therefore , we make the simple , conservative assumption that the gas collapsing into minihalos with masses higher than @xmath17 = 10@xmath18 , corresponding to virial temperature @xmath13 @xmath3 2 @xmath10 10@xmath15 k ( see e.g. yoshida et al .
2003 ) , must be reionized before the virialization of the minihalo , in order for the collapse of the gas and star formation to be prevented
. this assumption could be relaxed slightly if we account for the photoevaporation of the gas inside minihaloes that is already somewhat collapsed , but we do not expect this to dramatically affect our results ( see e.g. shapiro et al .
2004 ; iliev et al . 2005 ; wyithe & cen 2007 ) .
a second condition for pop iii star formation in unenriched dm haloes after reionization is that the dm haloes must be massive enough to allow the photoheated gas to cool and collapse .
numerous studies have found that dm haloes must have circular velocities @xmath1 30 km s@xmath9 in order for photoionized gas to collapse into them ( e.g. thoul & weinberg 1996 ; quinn et al .
1996 ; kepner et al . 1997 ; kitayama et al .
this limit may be lower at redshifts @xmath0 @xmath1 10 , due to the lower average intensity of the ionizing background and to the higher densities to which baryons may have already collapsed when the ionizing radiation turns on ( kitayama & ikeuchi 2000 ; dijkstra et al .
2004 ; see also chiba & nath 1994 ) .
however , there is no concensus on what this limiting mass is in cosmological haloes ( wyithe & cen 2007 ) , and many previous calculations have been done assuming spherical symmetry ( but see the cosmological simulations of hoeft et al .
2006 ) , which may underestimate the minimum halo mass , as asphericity can allow ionizing photons to penetrate deeper into haloes and reduce the effect of self - shielding ( see haiman 2008 ) .
however , this is not necessarily so clear either , as dense filaments from which gas can accrete into cosmological haloes may also be self - shielding , in which case calculations in sperical symmetry would lead to an overestimate of the minimum mass . to account for some spread in the minimum mass required for gas to collapse into dm haloes ,
we consider two possibilities for this minimum mass , corresponding to minimum circular velocities of 20 and 30 km s@xmath9 , given below : @xmath19 @xmath20 using the extended press - schechter formalism ( lacey & cole 1993 ) , we calculate , as a function of redshift , the maximum mass @xmath21(@xmath22 ) of virialized dm haloes which have , on average , no minihalo progenitors with masses @xmath1 @xmath17 prior to redshift @xmath22 ; more precisely , we calculate the maximum mass of haloes which have @xmath23/@xmath24ln@xmath25 @xmath26 1 , where @xmath27 is the number of progenitor haloes at @xmath22 ( see e.g. bromm & clarke 2002 ) .
this nomenclature is chosen to denote the redshift by which the region in which a halo forms must be reionized in order to prevent star formation in these minihalo progenitors .
we compare this maximum mass , for five different choices of @xmath22 , to the minimum mass for star formation in figure 1 . only in haloes with masses below the maximum halo mass @xmath21(@xmath22 ) and above the minimum mass , @xmath28 or @xmath29 , at a given redshift @xmath0 can the reionized gas remain pristine and unenriched until it collapses to form stars . accordingly , in regions reionized at @xmath22 @xmath3 25 , primordial star formation may take place at @xmath0 @xmath2 12.5 , on average , if the minimum halo mass is @xmath28 .
if , instead , the minimum halo mass is @xmath29 then primordial star formation can only take place at @xmath0 @xmath2 8 . also , the most massive haloes that can host metal - free star formation form at @xmath0 @xmath2 4 in regions that are reionized at @xmath22 = 19 and 23 , for a minimum halo mass of @xmath28 or @xmath29 , respectively .
indeed , due to the increase in the minimum mass for star formation with decreasing redshift , it is only in regions reionized at @xmath0 @xmath1 19 in which pop iii star formation can take place in reionized regions of the universe , unless the minimum halo mass for star formation is considerably smaller than @xmath28 . to summarize ,
two minimum requirements for the formation of a pop iii star cluster in an unenriched dm halo after reionization , i.e. at redshifts @xmath0 @xmath2 6 , are the following ( see also fig .
2 ) : \(1 ) the halo must not have any progenitors which host star formation .
this implies a maximum halo mass @xmath21(@xmath22 ) , above which star formation occurs in progenitor minihaloes that form prior to the redshift @xmath22 at which the gas is reionized .
after @xmath22 , photoheating of the gas prevents star formation , and the gas remains unenriched .
\(2 ) the halo must be massive enough for photoheated gas to collapse into it and form stars .
this implies a minimum halo mass @xmath30 , below which stars are unable to form .
we thus take it that only haloes with masses @xmath1 @xmath30 and @xmath31 @xmath21 can host the formation of primordial stars in reionized regions of the universe .
a third requirement is that the haloes are not enriched by galactic winds emanating from neighboring star - forming haloes , as shown schematically in figure 2 .
such external chemical enrichment will limit the abundance of unenriched haloes ; this is discussed further in the next section .
now that we have the redshift and mass distribution of dm haloes which may potentially host pop iii star formation in reionized regions of the universe , we would like to estimate the abundance of such haloes as a function of redshift , thereby finding an upper limit for the abundance of pop iii star clusters . as shown in section 2
, these haloes can only form in the relatively rare regions of the universe that were reionized at very early times , i.e. at @xmath22 @xmath1 19 , meaning that these haloes are relatively rare objects . while the early reionization history is only poorly constrained by observational data ,
there have recently been conducted a number of large - scale high - resolution simulations of this process which yield useful results for our purposes ( e.g. iliev et al . 2007 ; shin et al . 2008 ) .
shin et al .
( 2008 ) report their result for the volume fraction @xmath32(@xmath22 ) of the universe that is reionized , as a function of redshift , up to @xmath22 @xmath3 22 . for our calculation
, we use the following formula , which provides a satisfactory fit to their results over the range of redshifts @xmath22 that we consider ( i.e. 19 @xmath2 @xmath22 @xmath2 25 ) : @xmath33 the rate of change of the collapse fraction @xmath34(@xmath22 , @xmath0 ) of mass in haloes which may host pop iii star formation at redshift @xmath0 , if forming in regions that are reionized at @xmath22 , is given by the rate at which mass becomes incorporated in haloes with the minimum mass for star formation , as follows : @xmath35 here @xmath36(@xmath37,@xmath0 ) is the collapse fraction in dm haloes with masses @xmath38 @xmath37 at redshift @xmath0 , as given by the sheth - tormen formalism ( sheth & tormen 1999 ) .
@xmath30 is the minimum mass for star formation , given either by equation ( 1 ) or ( 2 ) .
as the sheth - tormen formalism gives the fraction of mass that is collapsed , it is more appropriate to use the fraction of mass @xmath39 that is reionized than the volume - filling fraction @xmath32 for our calculation .
we shall assume that the mass fraction of the universe that is reionized is a factor of ten higher than the volume - filling fraction @xmath32 given in equation ( 3 ) at 19 @xmath2 @xmath0 @xmath2 25 , due to the relatively large overdensities of the rare regions that are reionized so early .
this is roughly consistent with the results of shin et al .
( 2008 ) , who report that at @xmath0 @xmath3 20 the volume - filling fraction of h ii regions is of the order of 10@xmath40 while the fraction of the gas that is ionized is of the order of 10@xmath41 , implying an average overdensity of the h ii regions of the order of ten ( see also furlanetto et al .
accounting for this fraction @xmath42 of the baryonic mass in the universe that was reionized before @xmath22 , we integrate over @xmath22 to find the total number density of haloes virializing at redshift @xmath0 which are unenriched by previous star formation . assuming that a single pop iii star cluster with a lifetime of @xmath43 forms in every unenriched halo , we thus estimate the comoving number density of pop iii clusters as @xmath44 where @xmath45 is the critical density of the universe , assuming a hubble constant h@xmath46 = 70.5 km s@xmath9 mpc@xmath9 ( komatsu et al . 2009 ) , and @xmath47 is the fraction of this density in dark matter .
the integral is taken from @xmath22 = 19 to 25 , because unenriched haloes with masses @xmath1 @xmath30 only form in regions reionized before @xmath22 = 19 , as found in the previous section , and because before @xmath22 = 25 such a small fraction of the universe is reionized that little contribution would be made to the integral for higher @xmath22 . finally , we divide by the minimum halo mass to find the number density of haloes .
we plot two curves for the number density @xmath48 of pop iii star clusters in the top panel of figure 3 , one for each of the choices of @xmath30 , given by equation ( 1 ) and ( 2 ) .
the more massive haloes with mass @xmath29 are much rarer than those with mass @xmath28 for two reasons .
the first is that the overall abundance of dm haloes generally decreases with halo mass .
the second reason is that more massive haloes have minihalo progenitors at higher redshift , meaning that on average they form only in the regions in which star formation is suppressed by reionization at higher @xmath22 , and which are correspondingly more rare . for reasons discussed in dijkstra
( 2004 ) , the lower minimum mass may be more accurate at higher redshift , @xmath0 @xmath2 10 , while at redshifts @xmath0 @xmath2 3 the higher minimum mass is likely more accurate .
if this is so , then the comoving number density of pop iii star clusters may remain below @xmath3 10@xmath14 gpc@xmath41 at all redshifts .
this would be consistent with the upper limit on the abundance of pop iii clusters at @xmath0 @xmath3 4 found by nagao et al .
( 2008 ) , shown in the top panel of fig .
these authors report the non - detection of strong emitters of both ly@xmath6 and he ii @xmath71640 within a comoving volume of 4.03 @xmath10 10@xmath49 mpc@xmath15 in the subaru deep field at 3.93 @xmath2 @xmath0 @xmath2 4.01 and 4.57 @xmath2 @xmath0 @xmath2 4.65 .
this corresponds to an upper limit on the number density of pop iii clusters of @xmath48 @xmath3 2500 gpc@xmath41 , between the upper limits plotted in fig . 3 for minimum halo masses of m@xmath50 and m@xmath51 .
while this upper limit suggests that either the minimum halo mass for star formation is higher than @xmath52 at low redshifts or that metal enrichment by galactic winds is relatively rapid ( see section 3.2 ) , we note that the limiting flux of the survey would have allowed to detect only the brightest clusters ( see section 4 ) . in the bottom panel of fig .
3 , we plot the number density of pop iii star clusters per degree observed on the sky per redshift interval @xmath53@xmath0 . we also plot the inverse of the 100 arcmin@xmath54 area of the planned deep - wide survey ( dws ) to be conducted by the jwst , in which pop iii star clusters may be detected ( gardner et al . 2006 ) ; this corresponds to the number density of clusters that is required for there to be at least one cluster within this survey area , on average , per unit redshift .
the number of clusters that the jwst may detect is clearly very sensitive to the minimum halo mass for star formation ; most optimistically , for @xmath28 , there may be up to @xmath3 100 clusters out to @xmath0 @xmath3 6 per dws area .
however , for a minimum halo mass of @xmath29 the total number density is too low for the detection of even a single cluster in such a survey . a larger survey area of , for example , 17 @xmath10 17 arcmin@xmath54 for the e - elt ( eso elt science working group ) would , of course , provide a better chance to detect pop iii clusters .
as shown in equation ( 5 ) , the number density of pop iii clusters is linearly proportional to the lifetime @xmath43 of an individual cluster .
thus one impediment to the detection of pop iii clusters is their brief lifetime . because a halo becomes enriched with metals once the first sne explode , the lifetime of a purely primordial cluster is in part determined by the lifetime of the massive stars which explode as sne , which is , at most , of the order of 10@xmath16 yrs ( schaerer 2002 )
however , the timescale for the formation of second - generation , pop ii stars from the enriched gas is also dependent on the timescale for the mixing of metal - enriched sne ejecta with the primordial gas , which could be @xmath1 10@xmath11 yr ( see e.g. de avillez & mac low 2002 ) . in this case , the timescale for the formation of enriched stars would be similarly long ; accordingly , the abundance of dm haloes hosting only pop iii stars would be @xmath1 50 times higher than in the fiducial case of @xmath43 = 2 @xmath10 10@xmath55 yr shown in fig .
we emphasize , though , that it is the most massive stars which are the most luminous and hence it is during the lifetime of these stars , a few times 10@xmath55 yrs , that a cluster is most easily detectable ( see section 4 ) , meaning that the brightest clusters are also likely to be the rarest .
while in figure 3 we have only plotted the pop iii cluster abundances for one value of the minimum halo mass @xmath17 = 10@xmath18 required for star formation in minihalos , this minimum halo mass is an uncertain quantity and may be higher if a persistant lw background radiation field is present ( e.g. oshea & norman 2008 ; trenti & stiavelli 2009 ) .
if @xmath17 is indeed higher , then the abundance of pop iii clusters would be higher as well , since these potentially star - forming progenitor minihaloes would form at lower redshifts when @xmath39 is considerably higher . for a minimum minihalo mass an order of magnitude higher , @xmath17 = 10@xmath56 , we find that the upper limits on the abundance of pop iii star clusters may be roughly an order of magnitude higher than those shown in fig .
3 .
we have so far considered the formation of star clusters in dm haloes which are unenriched by star formation in their progenitors .
however , metal enrichment can also take place via galactic winds launched by neighboring star - forming galaxies ( e.g. nath & trentham 1997 ; mac low & ferrara 1999 ; cen & bryan 2001 ; madau et al .
2001 ; mori et al . 2002 ; scannapieco et al . 2003 ; bertone et al . 2005 ; tornatore et al .
as such external metal enrichment was ignored in the last section , our results so far for the number density of pop iii clusters are only upper limits .
to estimate the extent to which galactic winds enrich the igm and thereby prevent pop iii star formation , we compute the volume - filling fraction @xmath57 of the igm that is reionized at @xmath0 @xmath1 19 which these winds sweep up , as a function of redshift .
as we are interested in the metal enrichment of reionized regions , we assume that winds only originate from dm haloes with masses greater than the minimum mass for star formation in reionized regions . while haloes with masses @xmath3 @xmath30 will be the much more abundant than more massive haloes ( see e.g. reed et al .
2007 ) , the winds emanating from the galaxies which form in these haloes are likely to be rather weak , as the star formation efficiency in these haloes is low ( e.g. furlanetto & loeb 2003 ; samui et al .
2008 ) , due to the haloes retaining only a rather small fraction ( @xmath3 10 - 20 percent ) of their gas due to photoheating ( e.g. thoul & wienberg 1996 ; see also shen et al .
. however , the regions of interest to us are in close proximity to the descendants of the first galaxies which reionized their surroundings at @xmath22 @xmath1 20 . as these first galaxies formed in haloes with masses of @xmath1 5 @xmath10 10@xmath16 m@xmath5 at @xmath0 @xmath1 20 ( trac & cen 2007 ) , their descendants , on average , are rare haloes with masses @xmath1 10@xmath58 m@xmath5 at @xmath0 @xmath2 10 , as found using the extended press - schechter formalism .
these galaxies likely began enriching the igm already at @xmath0 @xmath3 20 , and because star formation is likely to proceed with a high efficiency within such massive haloes , we expect that these galaxies launch the metal - enriched winds which are principally responsible for enriching the otherwise pristine haloes we are considering . thus , under this assumption , as described schematically in figure 2 , it is the same galaxies which reionize the primordial gas at early times , thereby suppressing star formation and metal enrichment , that are also the sources of metal - enriched winds which prevent pop iii star formation at later times .
to model this enrichment process in a simple way , we assume that a single galaxy launches a wind from the center of each h ii region at redshift @xmath0 @xmath3 20 , and we take it that each h ii region encloses a comoving volume @xmath59 = 1 mpc@xmath15 @xmath60 at this redshift . while shin et al .
( 2008 ) find that the abundance of cosmological h ii regions having comoving volumes of @xmath2 0.03 mpc@xmath15 @xmath60 is @xmath3 10 times higher than these larger h
ii regions at @xmath0 @xmath3 21 , overall the larger h ii regions contain the majority of the reionized volume ; thus , we take the larger volume as a typical value .
we assume that the metal - enriched wind propagates outward from the central galaxy at a velocity @xmath61 , and we compute the comoving distance @xmath62 that it travels , as a function of redshift @xmath0 : @xmath63 finally , we estimate the fraction @xmath57 of the h ii regions that are enriched by the winds as @xmath64 where we assume that the h ii regions each enclose a comoving volume of @xmath59 = 1 mpc@xmath15 @xmath60 , as explained above . to calculate the number density of pop iii star clusters taking into account this metal enrichment by winds ,
we take it that a fraction @xmath57 of haloes that are not self - enriched are instead enriched by winds ; thus , the number density of metal - free star clusters forming in reionized regions @xmath65 , as defined in equation ( 5 ) , is reduced by this factor .
note that for simplicity we have assumed that each individual h ii region has the same size and launches a wind starting at the same redshift ; thus the metal - enriched wind volume - filling fraction of an individual h ii region , given by equation ( 7 ) , is equal to the global fraction in this calculation .
the resulting number densities are shown in figure 4 .
the velocity at which metal - enriched winds propagate into the igm at high redshift is not well - constrained , and so we plot the number density of pop iii clusters that we obtain for several choices , @xmath61 = 30 , 60 , and 100 km s@xmath9 . for wind velocities of @xmath3 30 km s@xmath9 ,
the total number of clusters within the dws area is only slightly lower than in the case of no external metal enrichment . for higher velocities the expected number of clusters which may be detected drops dramatically , with the formation of pop iii clusters being prevented altogether below redshifts @xmath0 @xmath3 3 and 5 , for wind velocities of 60 and 100 km s@xmath9 , respectively .
to take a fiducial case , furlanetto & loeb ( 2003 ) find that the average wind velocity from the relatively massive galaxies that we are considering may be @xmath3 60 km s@xmath9 , in which case pop iii cluster formation is unlikely to take place at @xmath0 @xmath2 3 , although the abundance of pop iii clusters above this redshift is only a factor @xmath2 2 below the upper limits presented in section 3.1 .
we emphasize that the simple estimates presented here do not take into account processes such as the mixing of metals with the primordial gas in the complex cosmological density field or the anisotropic expansion of early cosmological h ii regions .
in particular , we note that the mixing of metals with the gas contained in pristine haloes may be incomplete , leaving the star - forming gas in the centers of the haloes pristine , even if the halo as a whole is overtaken by a metal - enriched wind ( see cen & riquelme 2008 ; see also wyithe & cen 2007 ) .
furthermore , the metallicity of the gas in primordial haloes that is mixed with metals may not exceed the threshold necessary for the transition to low - mass , pop ii star formation , generally taken to be between 10@xmath66 and 10@xmath67 z@xmath5 ( e.g. bromm & loeb 2003 ;
omukai et al . 2005 ; frebel et al .
2007 ) . in this case
, the stellar initial mass function ( imf ) in these clusters may be very similar to the purely metal - free case .
however , observations of high redshift quasars suggest that much of the enriched igm at @xmath0 @xmath2 5 has metallicity @xmath68 @xmath1 10@xmath40 z@xmath5 ( songaila 2001 ; pettini et al .
2003 ) , such that chemical enrichment by galactic winds may indeed effect the transition from a pop iii to a pop ii imf ( see also ferrara et al .
2000 ; scannapieco et al .
2003 ) . in principle , though , star formation could also be prevented entirely in haloes with masses @xmath3 m@xmath69 due to the stripping of gas by winds with velocities exceeding the escape velocity of the halo ( see scannapieco et al . 2000 ) .
while we have shown that there may exist a population of metal - free haloes in which pop iii star clusters form even after reioinization , the detectability of these clusters depends on their luminosities and spectral properties .
we consider the luminosities of the stellar clusters in two emission lines , ly@xmath6 and he ii @xmath71640 , prominent recombination lines emitted by ionized hydrogen and doubly ionized helium , respectively .
the ratio of the luminosities of these lines , as well as their equivalent widths , are sensitive to the stellar imf and metallicity .
in particular , for a given stellar mass , metal - free stars are expected to have higher surface temperatures than metal - enriched stars , leading to a relatively high ratio of he ii - ionizing photons to h i - ionizing photons being emitted . in turn , this leads to relatively strong nebular emission in he ii @xmath71640 ( e.g. tumlinson & shull 2000 ; bromm et al .
2001 ; oh et al . 2001 ; schaerer 2002 , 2003 ) .
detection of both ly@xmath6 and he ii @xmath71640 can thus , in principle , facilitate the identification of pop iii star clusters ( e.g. tumlinson et al .
2001 ; nagao et al .
2008 ; prescott et al . 2009 ) .
more massive stars also emit higher numbers of he ii - ionizing photons , meaning that detection of these lines can also allow to place constraints on the stellar imf . while in neutral regions of the igm ly@xmath6 is easily scattered ( e.g. loeb & rybicki 1999 ) , the star clusters that we consider reside in regions which are already reionized , in principle yielding the igm optically thin to ly@xmath6 , although even a small neutral fraction can result in substantial absorption ( e.g. gunn & peterson 1965 ) . in this case , other emission lines from ionized hydrogen will still be directly detectable , h@xmath6 being the strongest of these lines with a luminosity @xmath3 10 times lower than that of ly@xmath6 ( e.g. osterbrock & ferland 2006 ) , and these lines could alternatively be used to constrain the metallicity and imf ( see e.g. oh et al .
2001 ) . for our purposes
, we will assume that the igm is optically thin to ly@xmath6 when evaluating the detectability of this emission line .
we consider pop iii star clusters that are formed promptly , within the @xmath3 2 myr lifetime of the most massive individual stars ( see also oh et al .
2001 ) , and we similarly evaluate the detectability of these clusters within @xmath3 2 myr of their formation , as it is during this time that a cluster is at its most luminous . to estimate the mass of the stellar clusters we take it that a fraction @xmath70 of the gas that is able to collapse into a dm halo is converted into stars ; this quantity is not well - constrained , but is likely of the order of a few percent ( e.g. krumholz & tan 2007 ) . in turn , we assume that only a fraction @xmath71 @xmath3 0.1 of the gas associated with a dm halo is able to cool and collapse ( i.e. that the collapsed baryon fraction in the halo is 10 percent of its value in the absence of photoheating ) , as the star - forming haloes under consideration have just grown massive enough to allow the photoionized gas to collapse for the first time , meaning that the majority of the gas stays photoionized and at low densities ( e.g. thoul & weinberg 1996 ; dijkstra et al .
we thus estimate the luminosity emitted by a pop iii cluster in ly@xmath6 as @xmath72 where @xmath73 is the fraction of hydrogen - ionizing photons that escape the halo and so do not contribute to the nebular emission from the cluster .
the number of hydrogen - ionizing photons emitted per second per unit mass in stars , @xmath74 , depends on the imf , and we consider two possible values : @xmath74 = @xmath75 and @xmath76 , for a salpeter pop iii imf and a top - heavy pop iii imf with a characteristic stellar mass @xmath1 100 m@xmath5 , respectively ( tumlinson & shull 2000 ; bromm et al .
as well , we have accounted for the fact that only a fraction of @xmath3 2/3 of the ionizing photons that are absorbed are converted into ly@xmath6 photons ( e.g. spitzer 1978 ) . following the same steps
, we have for the luminosity in he ii @xmath71640 @xmath77 where @xmath78 is the escape fraction of he ii - ionizing photons
. for the number of he ii - ionizing photons emitted per second per unit mass in stars , we have @xmath79 = @xmath80 and @xmath81 , again for a salpeter pop iii imf and a top - heavy pop iii imf , respectively ( tumlinson & shull 2000 ; bromm et al .
the flux in ly@xmath6 observed at @xmath0 = 0 is then given by @xmath82 where @xmath83 is the luminosity distance to redshift @xmath0 .
the flux in he ii @xmath71640 is computed similarly , by replacing the ly@xmath6 luminosity with the he ii @xmath71640 luminosity . in figure 5
we show the fluxes in ly@xmath6 and he ii @xmath71640 from clusters forming in haloes with mass @xmath3 @xmath30 , for each of our choices of this minimum halo mass , and for our two choices for the imf .
we also plot the flux limits for 3 @xmath84 detection of these emission lines with 100 hours of observation by the near - infrared spectrograph ( nirspec ) that will be aboard the jwst .
nirspec will operate at observed wavelengths from 0.6 to 5 @xmath85 m ( e.g. gardner et al .
2006 ) . the lower limit of this wavelength range yields nirspec sensitive only to ly@xmath6 and he ii @xmath71640 at redshifts @xmath0 @xmath1 4 and @xmath0 @xmath1 2.7 , respectively . for the most luminous clusters that we consider , the case of a large fraction @xmath70 = 10@xmath9 of the collapsed gas going into stars with a top - heavy imf , both ly@xmath6 and he ii @xmath71640 would be detectable by nirspec out to beyond redshift @xmath0 = 10 , as shown in the panels on the right in fig .
, the detection of these emission lines is much more feasible for the case of a top - heavy imf , for all combinations of minimum halo mass and star formation efficiency @xmath70 . in particular , because even a 3 @xmath84 detection of he ii @xmath71640 is not feasible if clusters have a salpeter imf , as shown in the bottom panels of fig
. 5 , detection of this emission line would suggest at least a somewhat top - heavy imf .
we note , however , that the escape fractions of both h i- and he ii - ionizing photons are likely to be higher for the case of a top - heavy imf , owing to the strong hydrodynamic response of the gas to photoheating ( e.g. johnson et al .
2009 ; see also wise & cen 2009 ; razoumov & sommer - larsen 2009 ) , with a corresponding decrease in the emitted luminosity .
even so , the flux in he ii @xmath71640 will likely still , in general , be higher for a more top - heavy imf ( johnson et al .
2009 ) . in the bottom panels of fig .
5 we also plot the approximate flux limit of the survey for he ii @xmath71640 emitters at @xmath0 @xmath3 4 carried out by nagao et al .
while this flux limit is an order of magnitude higher than that we find for nirspec , the non - detection of he ii @xmath71640 emitters in this survey already places some constraint on the abundance of pop iii clusters with a top - heavy imf and a high star formation efficiency ( see section 3 ) .
the jwst dws will place stronger constraints on both the abundance and luminosity of pop iii clusters . as the dws will be conducted using the near - infrared imaging instrument nircam
, we would also like to evaluate the detectability of the clusters by comparing their expected fluxes to the flux limits with which nircam will operate for the dws . given the rest - frame spectral energy distribution ( sed ) of the cluster , the flux that nircam would detect at @xmath0 = 0 can be computed .
for the expected sed of a cluster of young pop iii stars the luminosity in ly@xmath6 is roughly an order of magnitude greater than the stellar and nebular continuum luminosity in the wavelength range covered by a given nircam filter ( assuming a small escape fraction @xmath86)(e.g .
bromm et al .
2001 ; schaerer 2002 ) . for simplicity
, we thus calculate the flux that would be detected by nircam assuming that it is entirely in ly@xmath6 .
similar to the calculation in equation ( 10 ) , we thus find the monochromatic ly@xmath6 flux to be @xmath87 where @xmath88 is the spectral resolution ; for the wide filters on the nircam which operate at the wavelength of ly@xmath6 , @xmath88 = 4 . both the monochromatic ly@xmath6 flux and the flux limits for the planned dws are shown in figure 6 . while nircam has an exquisite sensitivity which will allow for detection of fluxes as low as @xmath3 1 njy , pop iii clusters may only be detected in the dws if their imf is top - heavy or if the star formation efficiency is relatively high , i.e. of the order of @xmath70 @xmath3 10@xmath9 .
nonetheless , that these clusters would form at lower redshift and in more massive dm haloes suggests that their detection may be still more feasible than that of pop iii stars in the first dwarf galaxies at @xmath0 @xmath1 10 ( see e.g. johnson et al .
2009 ) . as discussed in section 3.2
, the fact that these clusters would only form within the h ii regions surrounding the first galaxies at @xmath0 @xmath1 20 ( which reside in haloes with masses @xmath1 5 @xmath10 10@xmath16 m@xmath5 ; trac & cen 2007 ) implies that they will be found preferentially in the vicinity of the descendants of these first galaxies , which the extended press - schechter formalism dictates would typically reside in dm haloes with masses of @xmath3 10@xmath4 m@xmath5 at 3 @xmath2 @xmath0 @xmath2 6 . in particular , according to shin et al .
( 2008 ) the comoving radius of the largest h ii regions at @xmath0 @xmath1 20 is @xmath2 1 mpc @xmath89 .
therefore , a natural prediction of our model for pop iii star formation after reionization is that pop iii clusters may be found within a comoving distance of @xmath3 1 mpc @xmath89 , corresponding to an angular separation of @xmath2 40 arcsec on the sky , of the galaxies residing in @xmath3 10@xmath4 m@xmath5 haloes at 3 @xmath2 @xmath0 @xmath2 6 .
we have shown that the suppression of star formation in dm minihaloes by photoheating during the early stages of reionization can lead to the existence of unenriched dm haloes with masses @xmath1 10@xmath11 m@xmath5 at redshifts @xmath0 @xmath2 10 which may host pop iii star clusters .
while the abundance of such haloes is sensitively dependent on the degree to which galactic winds enrich the igm with metals , as well as on the minimum halo mass for star formation in the reionized regions of the universe , there may be up to thousands of bright pop iii star clusters per square degree at redshifts 3 @xmath2 @xmath0 @xmath2 6 , corresponding to @xmath3 100 clusters within the area of the dws to be carried out by the jwst .
this is , however , likely a strong upper limit to the abundance , as metal - enriched winds expelled by the same galaxies that reionize the primordial gas at early times may easily enrich it by these redshifts , depending on the speed of the winds and on how quickly the metals are mixed into the pristine haloes .
if this enrichment is sufficiently rapid , pop iii star formation in pristine dm haloes may be completely prevented after reionization is complete at @xmath0 @xmath3 6 . although the stellar mass , and so the luminosity , of pop iii clusters formed in reionized regions of the universe is limited by the fraction of the primordial gas that is first able to collapse into the dm haloes , we estimate that ly@xmath6 emission from these star clusters could be detected by the jwst if the stellar imf is top - heavy or if the star formation efficiency @xmath70 is high , i.e. of the order of 10 percent .
the he ii @xmath71640 emission line , which is an indicator of stellar imf and metallicity , may be detectable out to at least @xmath0 @xmath3 3 for the case of a top - heavy imf , perhaps allowing to place constraints on the pop iii imf , at least at low redshift . furthermore ,
in reionized regions pop iii clusters may form preferentially within @xmath3 40 arcsec of more luminous galaxies at 3 @xmath2 @xmath0 @xmath2 6 , the descendants of the first galaxies which began reionization at @xmath0 @xmath1 20 , perhaps aiding in the identification of these clusters in future surveys .
current observations already yield some constraints on the abundance of pop iii star clusters at @xmath0 @xmath3 4 ( dawson et al .
2004 ; nagao et al .
2008 ; wang et al .
2009 ; see also e.g. fosbury et al .
2003 ; shapley et al . 2003
; nagao et al . 2005 ; jimenez & haiman 2006 ; prescott et al . 2009 ;
bouwens et al . 2009 ) .
the non - detection of strong he ii @xmath71640 emitters at @xmath0 @xmath3 4 reported by nagao et al .
( 2008 ) yields an upper limit on the number density of pop iii clusters of @xmath48 @xmath3 2500 gpc@xmath41 , between the upper limits plotted in fig .
however , the limiting flux in this survey is only low enough to rule out the presence of the most luminous pop iii clusters .
nonetheless , this imposes a useful constraint , and suggests that if clusters of massive pop iii stars do form after reionization , they may in fact form in relatively low - mass haloes in which star formation is inefficient .
we note that the pop iii imf in star clusters formed after reionization is likely different than that of the first stars formed in minihaloes at @xmath0 @xmath1 20 ( so - called pop iii.1 stars ) , due to the onset of turbulence in the more massive haloes in which the clusters form ( wise et al .
2007b ; greif et al .
2008 ) , to the possible enhanced abundances of the coolants h@xmath90 and deuterium hydride ( hd ) in the reionized primordial gas ( e.g. nagakura & omukai 2005 ; johnson & bromm 2006 ; yoshida et al . 2007 ; mcgreer & bryan 2008 ) , and to the lower temperature of the cosmic microwave background ( e.g. larson 1998 ; tumlinson 2007 ; smith et al .
2009 ) . instead of the characteristic stellar mass of @xmath3 100 m@xmath5 expected for pop iii.1 stars formed in the earliest minihaloes ,
pop iii.2 stars with a characteristic mass of the order of 10 m@xmath5 may instead be detected in star clusters formed from reionized primordial gas .
such clusters would have spectral properties intermediate between the two imfs , salpeter and top - heavy , discussed in section 4 .
an important next step to better estimate the abundance of pop iii star clusters will be to carry out cosmological simulations which capture the earliest stages of the inhomogeneous reionization of the universe , as well as the chemical enrichment from both the first stars formed in minihaloes and large - scale galactic winds .
we note that the cosmological metal - enrichment simulations presented in tornatore et al .
( 2007 ) already provide some indication that pop iii clusters may form in rare regions after reionization .
in addition , cosmological radiative transfer simulations may be necessary to estimate the masses , and so the luminosities , of stellar clusters that form in the dm haloes into which the reionized primordial gas is first able collapse in the course of hierarchical structure formation .
related to this , magnetic fields pervading the reionized gas may have an important dynamical effect on the formation of pop iii star clusters at low redshifts ( see schleicher et al .
2009a ) ; simulations tracking the growth and evolution of magnetic fields in the igm would thus also be of interest .
furthermore , in the present work we have only considered the average merger history of a potential pop iii star - forming halo , ignoring the range of possible merger histories that such haloes may have . this assumption would clearly be relaxed within the context of a cosmological simulation .
although we have accounted for the higher average density of regions which are reionized at early times with our choice of @xmath39 in equation ( 5 ) , we have otherwise assumed that the haloes in which pop iii star clusters form are not biased , i.e. that these haloes are otherwise located randomly in space . in principle
this could lead to an overestimate of @xmath48(@xmath0 ) , for example , if haloes with mass @xmath30(@xmath0 ) typically form in underdense regions , while it is only overdense regions which are reionized at early times . however , haloes with mass @xmath30(@xmath0 ) correspond to @xmath1 1-@xmath84 fluctuations in the cosmological density field at @xmath0 @xmath1 3 ( e.g. barkana & loeb 2001 ) , typically form at 3 @xmath2 @xmath0 @xmath2 4 ( see e.g. boylan - kolchin et al .
2009 ) , and so it does not appear likely that they form preferentially in underdense regions at @xmath0 @xmath1 3 , the redshifts at which we find pop iii star clusters are mostly likely to form in them .
therefore , we expect that accounting for the detailed clustering properties of these haloes around the descendants of the sources of reionization at @xmath0 @xmath3 20 would not change our results substantially , although the effect of bias should be considered in more depth in future studies .
another implicit assumption throughout this work has been that stellar clusters will form in the pristine dm haloes which are massive enough for the gas to collapse into them after reionization .
there is the possibility that this gas could instead collapse to form a @xmath1 10@xmath14 m@xmath5 black hole directly , if molecular cooling is suppressed ( e.g. bromm & loeb 2003 ; begelman et al .
2006 ; lodato & natarajan 2006 ; spaans & silk 2006 ; regan & haehnelt 2009 ; see also schleicher et al . 2009b ) .
shang et al .
( 2009 ) find that the lw background at @xmath0 @xmath2 8 ( ahn et al .
2009 ) may be high enough to suppress the formation of h@xmath90 molecules and allow for the formation of such a massive single object .
however , it is not clear that the relatively small fraction @xmath71 of the gas which is able to collapse is sufficiently high for this to occur .
also , despite the presence of a persistant lw background , the enhanced free electron fraction in the collapsing reionized gas will boost the formation rate of h@xmath90 to some extent , perhaps aiding to prevent direct collapse to a black hole .
although we have explored the impact on the abundance of pop iii clusters of different choices for the speed of metal enrichment by galactic winds , and for the minimum dm halo mass for star formation after reionization , we have considered only one choice for the early reionization history , based on the simulations of shin et al .
as the abundance of unenriched haloes is dependent on the fraction of the universe reionized at redshifts @xmath0 @xmath1 20 , as well as on the typical size of cosmological h ii regions at these redshifts , our results are in principle quite sensitive to this choice .
observational constraints on the early stages of reionization , for example from observations of 21-cm emission by instruments such as the _ square kilometer array _
( e.g. carilli et al .
2004 ) , may thus provide indirect constraints on the abundance of unenriched haloes at lower redshifts . in the upcoming decade ,
next generation telescopes such as the jwst and the e - elt , however , may first provide direct constraints on the amount of pop iii star formation taking place in the early universe , with the exciting possibility of detecting the emission from individual pop iii star clusters , perhaps even well after the epoch of reionization .
the author is grateful to volker bromm for the use of his extended press - schechter routine , and to an anonymous referee for comments which led to significantly improved clarity in the presentation of this work .
the author also thanks thomas greif for valuable comments on an early draft , as well as sadegh khochfar and the members of the theoretical modeling of cosmic structures ( tmox ) group at mpe for helpful discussions .
99 abel t. , bryan g. l. , norman m. l. 2002 , sci , 295 , 93 ahn k. , shapiro p. r. 2007 , mnras , 375 , 881 ahn k. , shapiro p. r. , iliev i. t. , mellema g. , pen u .-
2009 , apj , 695 , 1430 barkana r. , loeb a. 2001 , phr , 349 , 125 bertone s. , stoehr f. , white s. d. m. 2005 , mnras , 359 , 1201 begelman m. c. , volonteri m. , rees m. j. 2006 , mnras , 370 , 289 bouwens r. j. , et al .
2009 , apj , submitted ( arxiv:0910.0001 ) boylan - kolchin m. , springel v. , white s. d. m. , jenkins a. , lemson g. 2009 , mnras , 398 , 1150 bromm v. , kudritzki r. p. , loeb a. 2001 , apj , 552 , 464 bromm v. , coppi p. s. , larson r. b. 2002 , apj , 564 , 23 bromm v. , clarke c. j. 2002 , apj , 566 , l1 bromm v. , larson r. b. 2004 , ara&a , 42 , 79 v. , loeb a. 2003 , nat , 425 , 812 bromm v. , loeb a. 2003 , apj , 596 , 34 bromm v. , yoshida n. , hernquist l. 2003 , apj , 596 , l135 bromm v. , yoshida n. , hernquist l. , mckee c. f. 2009 , nat , 459 , 49 carilli c. l. , furlanetto s. , briggs f. , jarvis m. , rawlings s. , falcke h. 2004 , newar , 48 , 1029 cen r. , bryan g. l. 2001 , apj , 546 , l81 cen r. , riquelme m. a. 2008 , apj , 674 , 644 chiba m. , nath b. b. 1994 , apj , 436 , 618 ciardi b. , ferrara a. , 2001 , mnras , 324 , 648 ciardi b. , ferrara a. 2005 , ssrv , 116 , 625 dawson s. , et al .
2004 , apj , 617 , 707 de avillez m. a. , mac low m. m. 2002 , apj , 581 , 1047 dijkstra m. , haiman z. , rees m. j. , weinberg d. h. 2004 , apj , 601 , 666 dijkstra m. , haiman z. , mesinger a. , wyithe j. s. b. 2008 , mnras , 391 , 1961 dijkstra m. , wyithe j. s. b. 2007 , mnras , 379 , 1589 efstathiou g. 1992 , mnras , 256 , 43 ferrara a. , pettini m. , shchekinov y. 2000 , mnras , 319 , 539 fosbury r. a. e. , et al .
2003 , apj , 596 , 797 frebel a. , johnson j. l. , bromm v. 2007 , mnras , 380 , l40 furlanetto s. r. , loeb a. 2003 , apj , 588 , 18 furlanetto s. r. , zaldarriaga m. , hernquist l. 2004 , apj , 613 , 1 fryer c. l. , woosley s. e. , heger a. 2001 , apj , 550 , 372 gardner j. p. , et al .
2006 , ssrv , 123 , 485 gnedin n. y. 2000 , apj , 542 , 535 greif t. h. , johnson j. l. , bromm v. , klessen r. s. 2007 , apj , 670 , 1 greif t. h. , johnson j. l. , klessen r. s. , bromm v. 2008 , mnras , 387 , 1021 gunn j. e. , peterson b. a. , 1965 , apj , 142 , 1633 haiman z. , rees m. j. , loeb a. 1997 , apj , 476 , 458 haiman z. 2008 , astrophysics in the next decade : jwst and concurrent facilities " , astrophysics & space science library , eds . h. thronson , a. tielens , m. stiavelli ( arxiv:0809.3926 ) heger a. , fryer c. l. , woosley s. e. , langer n. , hatmann d. h. 2003 , apj , 591 , 288 hoeft m. , yepes g. , gottlber s. , springel v. 2006 , mnras , 371 , 401 iliev i. t. , shapiro p. r. , raga a. c. 2005 , mnras , 361 , 405 iliev i. t. , mellema g. , shapiro p. r. , pen u. l. 2007 , mnras , 376 , 534 jimenez r. , haiman z. 2006 , nat , 440 , 501 johnson j. l. , bromm v. 2006 , mnras , 366 , 247 johnson j. l. , greif t. h. , bromm v. 2008 , mnras , 388 , 26 johnson j. l. , greif t. h. , bromm v. , klessen r. s. , ippolito j. 2009 , mnras , 399 , 37 kepner j. v. , babul a. , spergel d. n. 1997 , apj , 487 , 61 kitayama t. , ikeuchi s. 2000 , apj , 529 , 615 kitayama t. , susa h. , umemura m. , ikeuchi s. 2001 , mnras , 326 , 1353 kitayama t. , yoshida n. 2005 , apj , 630 , 675 komatsu e. , et al .
2009 , apjs , 180 , 330 krumholz m. r. , tan j. c. 2007 , apj , 654 , 304 lacey c. , cole s. 1993 , mnras , 262 , 627 larson r. b. 1998 , mnras , 301 , 569 lodato g. , natarajan p. 2006
, mnras , 371 , 1813 loeb a. , rybicki g. b. , 1999 , apj , 524 , 527 mac low m. m. , ferrara a. 1999 , 513 , 142 madau p. , ferrara a. , rees m. j. 2001 , apj , 555 , 92 mcgreer i. d. , bryan g. l. 2008 , apj , 685 , 8 mesinger a. , bryan g. l. , haiman z. 2006 , apj , 648 , 835 mesinger a. , bryan g. l. , haiman z. 2009 , mnras , 399 , 1650 mori m. , ferrara a. , madau p. 2002
, apj , 571 , 40 nagakura t. , omukai k. 2005 , mnras , 364 , 1378 nagao t. , et al .
2005 , apj , 631 , l5 nagao t. , et al .
2008 , apj , 680 , 100 nath b. b. , trentham n. 1997 , mnras , 291 , 505 oh p. , haiman z. , rees m. j. 2001 , apj , 553 , 73 omukai k. , tsuribe t. , schneider r. , ferrara a. 2005 , apj , 626 , 627 oshea b. w. , norman m. l. 2007 , 654 , 66 oshea b. w. , norman m. l. 2008 , apj , 673 , 14 osterbrock d. , ferland g. 2006 , astrophysics of gaseous nebulae and active galactic nuclei . university science books : sausalito .
pan l. , scalo j. 2007 , apj , 654 , 29 pettini m. , madau p. , bolte m. , prochaska j. x. , ellison s. l. , fan x. 2003 , apj , 594 , 695 prescott m. k. m. , dey a. , jannuzi b. t. 2009 , apj , 702 , 554 quinn t. , katz n. , efstathiou g. 1996 , mnras , 278 , l49 razoumov a. o. , sommer - larsen j. 2009 , apj , submitted ( arxiv:0903.2045 ) reed d. s. , bower r. , frenk c. s. , jenkins a. , theuns t. 2007 , mnras , 374 , 2 regan j. a. , haehnelt m. g. 2009 , mnras , 396 , 343 ricotti m. , gnedin n. y. , shull j. m. 2008 , apj , 685 , 21 samui s. , subramanian k. , srianand r. 2008 , mnras , 385 , 783 scannapieco e. , ferrara a. , broadhurst t. 2000 , apj , 536 , l11 scannapieco e. , schneider r. , ferrara a. 2003 , apj , 589 , 35 schaerer d. 2002 , a&a , 382 , 28 schaerer d. 2003 , a&a , 397 , 527 schleicher d. r. g. , galli d. , glover s. c. o. , banerjee r. , palla f. , schneider r. , klessen r. s. 2009a , apj , 703 , 1096 schleicher d. r. g. , spaans m. , glover s. c. o. 2009b , apj , submitted schneider r. , salvaterra r. , ferrara a. , ciardi b. 2006 , mnras , 369 , 825 shang c. , bryan g. l. , haiman z. 2009 , mnras , submitted ( arxiv:0906.4773 ) shapiro p. r. , iliev i. t. , raga a. c. 2004 , mnras , 348 , 753 shapley a. e. , steidel c. c. , pettini m. , adelberger k. l. 2003 , apj , 588 , 65 shen s. , wadsley j. , stinson g. 2009 , mnras , submitted ( arxiv:0910.5956 ) sheth r. k. , tormen g. , 1999 , mnras , 308 , 119 shin m. s. , trac h. , cen r. 2008 , apj , 681 , 756 smith b. d. , turk m. j. , sigurdsson s. , oshea b. w. , norman m. l. 2009 , apj , 691 , 441 songaila a. 2001 , apj , 561 , l153 spaans m. , silk j. 2006 , apj , 652 , 902 spitzer l. 1978 , physical processes in the interstellar medium ( wiley , new york ) susa h. , umemura m. , hasegawa k. 2009 , apj , 702 , 480 thoul a. a. , weinberg d. h. 1996 , apj , 465 , 608 tornatore l. , ferrara a. , schneider r. 2007 , mnras , 382 , 945 tornatore l. , borgani s. , viel m. , springel v. 2009 , mnras , submitted ( arxiv:0911.0699 ) trac h. , cen r. 2007 , apj , 671 , 1 trenti m. , stiavelli m. 2009 , apj , 694 , 879 trenti m. , stiavelli m. , shull j. m. 2009 , apj , 700 , 1672 tumlinson j. , shull j. m. 2000 , apj , 528 , l65 tumlinson j. , giroux m. l. , shull j. m. 2001 , apj , 550 , l1 tumlinson j. 2007 , apj , 664 , l63 wang j. x. , malhotra s. , rhoads j. e. , zhang h. t. , finkelstein s. l. 2009 , apj , 706 , 762 whalen d. , van veelen b. , oshea b. w. , norman m. l. 2008a , apj , 682 , 49 whalen d. , oshea b. w. , smidt j. , norman m. l. 2008b , apj , 679 , 925 windhorst r. a. , cohen s. h. , jansen r. a. , conselice c. , yan h. 2006 , newar , 50 , 113 wise j. h. , abel t. 2005 , apj , 629 , 615 wise j. h. , abel t. 2007a , apj , 671 , 1559 wise j. h. , abel t. 2007b , apj , 665 , 899 wise j. h. , cen r. 2009 , apj , 693 , 984 wyithe j. s. b. , cen r. 2007 , apj , 659 , 890 yoshida n. , abel t. , hernquist l. , sugiyama n. 2003 , apj , 592 , 645 yoshida n. , omukai k. , hernquist l. 2007 , apj , 667 , l117 | in reionized regions of the universe , gas can only collapse to form stars in dark matter ( dm ) haloes which grow to be sufficiently massive .
if star formation is prevented in the minihalo progenitors of such dm haloes at redshifts @xmath0 @xmath1 20 , then these haloes will not be self - enriched with metals and so may host population ( pop ) iii star formation .
we estimate an upper limit for the abundance of pop iii star clusters which thus form in the reionized universe , as a function of redshift . depending on the minimum dm halo mass for star formation , between of the order of one and of the order of a thousand pop iii star clusters per square degree may be observable at 2 @xmath2 @xmath0 @xmath2 7 .
thus , there may be a sufficient number density of pop iii star clusters for detection in surveys such as the deep - wide survey ( dws ) to be conducted by the _ james webb space telescope _ ( jwst ) .
we predict that pop iii clusters formed after reionization are most likely to be found at @xmath0 @xmath1 3 and within @xmath3 40 arcsec ( @xmath3 1 mpc comoving ) of dm haloes with masses of @xmath3 10@xmath4 m@xmath5 , the descendants of the haloes at @xmath0 @xmath3 20 which host the first galaxies that begin reionization .
however , if star formation is inefficient in the haloes hosting pop iii clusters due to the photoionizing background radiation , these clusters may not be bright enough for detection by the near - infrared camera which will conduct the dws .
nonetheless , if the stellar initial mass function ( imf ) is top - heavy the clusters may have sufficiently high luminosities in both ly@xmath6 and he ii @xmath71640 to be detected and for constraints to be placed on the pop iii imf . while a small fraction of dm haloes with masses as high as @xmath3 10@xmath8 m@xmath5 at redshifts @xmath0 @xmath2 4 are not enriched due to star formation in their progenitors ,
external metal enrichment due to galactic winds is likely to preclude pop iii star formation in a large fraction of otherwise unenriched haloes , perhaps even preventing star formation in pristine haloes altogether after reionization is complete at @xmath0 @xmath3 6 .
-1 cm cosmology : theory early universe galaxies : formation high - redshift haloes intergalactic medium |
in a previous paper @xcite , we have proposed a methodology that enables monte carlo ( mc ) methods based on the gibbs canonical ensemble : @xmath7to account for the existence of an anomalous region with negative heat capacities @xmath1 @xcite and to overcome the so - called _ supercritical slowing down _ observed near of first - order phase transition @xcite .
this development is inspired on the consideration of a recently obtained fluctuation relation vel - jpa , vel - jstat , vel - geo : @xmath8which appears as a suitable extension of the known canonical identity @xmath9 involving the heat capacity @xmath10 and the energy fluctuations @xmath11 .
this last expression accounts for the realistic perturbation provoked on the internal state of a certain environment as a consequence of the underlying thermodynamic interaction with the system under study , which is characterized here in terms of the correlation function @xmath12 between the system internal energy @xmath13 and the environment inverse temperature @xmath14 . while the canonical ensemble ( [ can ] ) dismisses the existence of this feedback effect ( due to the constancy of the inverse temperature @xmath15 associated with this ensemble ) and it is only compatible with positive heat capacities , the general case possibilities to access anomalous macrostates with negative heat capacities @xmath1 as long as the condition @xmath16 is obeyed .
the incidence of non - vanishing correlated fluctuations @xmath17 can be easily implemented in mc simulations .
roughly speaking , the extension of canonical mc methods is achieved by replacing the canonical inverse temperature @xmath15 with a _
variable inverse temperature _ @xmath18 .
the resulting framework constitutes a suitable extension of the gerling and hller methodology on the basis of the so - called _ dynamic ensemble _ @xcite , where the microcanonical curve @xmath19 and the heat capacity @xmath20 can be estimated in terms of the energy and temperature expectation values @xmath21 and @xmath22 as well as their fluctuating behavior described by eq.([fdr ] ) .
this method successfully reduces the exponential divergence of decorrelation time @xmath23 with the increase of the system size @xmath4 of canonical mc methods to a weak power - law divergence @xmath24 , with a typical exponent @xmath25.@xmath26 for the case of @xmath27 ten - state potts model @xcite . by combining this type of argument with cluster algorithms ,
one obtains very efficient mc schemes that constitute attractive alternatives of the known multicanonical method and its variants @xcite . in this work
, we shall improve the present methodology to consider the existence of finite size effects that reduces the precision of a direct determination of the microcanonical caloric curve @xmath28 , as well as a better implementation of mc schemes .
the simplest way to implement the existence of non - vanishing correlations @xmath29 corresponds to a linear coupling of the environment inverse temperature @xmath30 with the thermal fluctuations of the system energy : @xmath31where @xmath32 is a coupling constant that appears here as an additional control parameter . while the case with @xmath33 corresponds to the canonical ensemble ( [ can ] ) , where @xmath34 , in general , the constancy of the inverse temperature @xmath35 can be only ensured in the average sense , @xmath36 . eq.([linear.ansatz ] )
can be substituted into eq.([fdr ] ) to obtain the following results : @xmath37where @xmath38 denotes the thermal dispersion of a given quantity @xmath39 .
since @xmath40 and @xmath41 should be nonnegative , one arrives at the following stability condition : @xmath42for @xmath33 , one derives the ordinary constraint @xmath43 that emphasizes the unstable character of macrostates with negative heat capacities @xmath1 within the canonical description
. however , these anomalous macrostates can be observed in a stable way in a general situation with @xmath44 when this control parameter satisfies eq.(stab.cond ) . by assuming an extensive character of the heat capacity @xmath45 in short - range interacting systems , the energy dispersion @xmath41 grows with the increase of the system size @xmath4 as @xmath46 , so that , the dispersion of the energy per particle @xmath47 behaves as @xmath48 . since @xmath49 in the case of the ansatz ( [ linear.ansatz ] )
, the dispersion of the inverse temperature also behaves as @xmath50 .
thus , the present equilibrium situation constitutes a physical scenario that ensures the stability of macrostates with negative heat capacities @xmath1 with the incidence of small thermal fluctuations .
this type of equilibrium situation is schematically represented in fig.caricature.eps . here
, it is shown the typical microcanonical caloric curve @xmath51 corresponding to a finite short - range interacting system that undergoes a first - order phase transition , as well as the energy distribution function @xmath52 associated with the thermal coupling of this system with an environment with inverse temperature @xmath18 .
the intersection points @xmath53 derived from the condition of thermal equilibrium : @xmath54determine the position of maxima and minima of the distribution function @xmath55 .
note that the linear ansatz ( [ linear.ansatz ] ) can be considered as the first - order approximation of the power expansion of a general dependence @xmath56 : @xmath57 where @xmath58 .
while in the canonical ensemble there could exist three intersections points ( @xmath59 , @xmath60 and @xmath61 ) because of the constancy of the inverse temperature , it is possible to ensure the existence of only one intersection point @xmath53 by appropriately choosing the inverse temperature dependence @xmath62 .
if the system size @xmath4 is sufficiently large , the energy distribution function @xmath52 adopts a bell shape , which is approximately described with the gaussian distribution:@xmath63 , \label{gd}\]]where @xmath64 .
the expectation values @xmath65 and @xmath22 provide a direct estimation of the energy @xmath53 and inverse temperature @xmath66 at the intersection point illustrated in fig.[caricature.eps ] : @xmath67 the procedure previously described is equivalent to the one employed by gerling and hller in the framework of the dynamical ensemble @xcite : @xmath68which accounts for the thermal coupling of the system with a bath exhibiting a constant heat capacity , e.g. , an ideal gas , whose inverse temperature obeys the following dependence on the system energy @xmath13 : @xmath69our proposal considers some improvements to the gerling and huller s methodology .
in fact , the energy dependence ( [ ghb ] ) is less convenient for calculations than the linear ansatz ( [ linear.ansatz ] ) , which is a feature particularly useful to perform the analysis of _ finite size effects _
( see subsection [ finite ] below ) .
moreover , one can also obtain the value of the heat capacity @xmath70 , or more exactly , the so - called _ curvature curve _ @xmath71:@xmath72at the intersection point @xmath53 , @xmath73 , through the energy dispersion @xmath41 in analogous way of mc calculation by using the canonical ensemble ( [ can ] ) : @xmath74this last equation was obtained from eq.([curvatura ] ) by rewriting the first relation of eq.([dispersions ] ) .
since the energy and the inverse temperature dispersions in eq.(dispersions ) are controlled by the coupling constant @xmath32 , it is desirable to reduce them as low as possible .
one can verify that the energy dispersion @xmath41 decreases with the increase of the coupling constant @xmath75 .
however , the value of this parameter should not be excessively large .
its increment leads to the increase of the inverse temperature dispersion @xmath76 , which affects the precision of the inverse temperature @xmath35 of the system indirectly derived from the expectation value @xmath22 .
a thermodynamic criterion to provide an optimal value for the coupling constant @xmath32 can be obtained by minimizing the total dispersion @xmath77 : @xmath78which talks about the precision in the determination of the intersection point @xmath79 .
this analysis leads to the following result : @xmath80 the energy distribution function corresponding to the thermal coupling of the system with an environment can be expressed by the following equation : @xmath81here , @xmath82 is a probabilistic weight that characterizes such a thermodynamic influence , which is related to the environment inverse temperature @xmath83 as follows : @xmath84a direct integration allows us to verify that the probabilistic weight associated with the linear ansatz ( [ linear.ansatz ] ) is simply the _ gaussian ensemble _
@xcite : @xmath85 \label{gens}\]]introduced by hetherington @xcite , which approaches in the limit @xmath86 to the microcanonical ensemble:@xmath87 as already explained , the estimation of the microcanonical caloric curve @xmath88 , as well as the fluctuation relation ( [ fdr ] ) , are based on the consideration of a gaussian shape for the energy distribution function @xmath52 .
such an approximation naturally arises as the asymptotic distribution as long as the system size @xmath89 be sufficiently large .
if the system size @xmath4 is not so large , small deviations from the gaussian profile ( [ gd ] ) are naturally expected .
fortunately , the particular mathematical form of the gaussian ensemble ( gens ) possibilities to consider some simple corrections formulae to deal with the existence of these finite size effects and to improve the precision of the present methodology .
let us denote the energy distribution function in terms of the energy per particle @xmath47 as follows : @xmath90 , \]]where the function @xmath91 , with @xmath92 being the entropy per particle , and @xmath35:@xmath93the system inverse temperature at the stationary point @xmath94 .
let us now develop a power - expansion : @xmath95with @xmath96 being the curvature .
the gaussian approximation is developed by dismissing those terms with @xmath97 . here
, the energy deviation @xmath98 obeys a gaussian distribution:@xmath99with standard deviation @xmath100 : @xmath101and the partition function @xmath102 can be approximated as : @xmath103note that eq.([sd ] ) is fully equivalent to eq.([curvature.fluc ] ) since the standard deviation @xmath104 .
eq.([z.lambda ] ) can be rewritten as follows : @xmath105where @xmath106 can be referred to as the plank thermodynamic potential corresponding to the gaussian ensemble ( [ gens ] ) . by performing the thermodynamic limit @xmath107
, this thermodynamic function can be related to the known legendre transformation with the microcanonical entropy : @xmath108clearly , the gaussian ensemble ( [ gens ] ) provides a suitable extension of the gibbs canonical ensemble ( [ can ] ) that is able to deal with the existence of macrostates with negative heat capacities @xmath1 .
in fact , it is a particular example of the so - called _ generalized canonical ensembles _ that preserve some of its more relevant features vel - geo , hugo.gce , toral .
the first correction of the gaussian approximation ( [ gaussian ] ) is obtained by dismissing terms with @xmath109 in the power expansion ( [ power ] ) : @xmath110where : @xmath111it is convenient to introduce the dimensionless variable @xmath112 . by considering the size dependencies @xmath113 and @xmath114 , it is possible to verify that the cubic term @xmath115 decreases as @xmath116 with the increase of the system size @xmath4 .
thus , one arrives at the distribution function : @xmath117this last result leads to the following expectation values : @xmath118which allow to express the expectation value of the energy deviation @xmath119 as well as its second and third order dispersions , @xmath120 and @xmath121 , as follows : @xmath122the previous results can be combined with the linear ansatz ( linear.ansatz ) to obtain the following first - order correction of eq.(work1 ) : @xmath123 the second - order correction of the gaussian constributions is carried out by dismissing terms with @xmath124 in power expansion ( [ power ] ) : @xmath125 where : @xmath126 these terms lead to the following correction of the distribution function : @xmath127 whose third and fourth terms account for finite size effects of order @xmath128 . by denoting the auxiliary constants
@xmath129 and @xmath130 as follows:@xmath131 direct calculations allow to obtain the normalization constant @xmath132 : @xmath133 as well as the following expectation values : @xmath134 while the expressions ( [ mean ] ) and ( [ fluct1 ] ) for the expectation values @xmath135 and @xmath136 remain invariable under the second - order approximation , the second and fourth order dispersions @xmath137 and @xmath138 exhibit the following corrections : @xmath139 by introducing the cumulants @xmath140 and @xmath141:@xmath142 the auxiliary constants @xmath143 and @xmath130 can be expressed as follows:@xmath144 thus , the main work equations ( [ work1 ] ) and ( [ curvature.fluc ] ) can be expressed in this second - order approximation as follows : @xmath145 where @xmath146 is a second - order term defined by the cumulants @xmath147 and @xmath141 as : @xmath148 it is easy to verify that the surviving finite size effects of the above correction for the caloric curve @xmath149 in terms of the energy per particle @xmath150 are of order @xmath151 , while the ones corresponding to the curvature curve @xmath152 _ versus _ @xmath150 are of order @xmath153 . as a by - product of the previous analysis , one can obtain the third and fourth derivatives of the entropy per particle @xmath154 as follows : @xmath155where @xmath156 is another second - order term defined by cumulants @xmath157 and @xmath158 as : @xmath159the underlying finite size effects of the dependence @xmath160 _ versus _
@xmath161 estimated by eq.([z3 ] ) are of order @xmath162 , while the ones corresponding to the dependence @xmath163 _ versus _ @xmath161 estimated by eq.([z4 ] ) are of order @xmath128 .
as already commented in the introductory section , a general way to extend a particular canonical mc algorithm with transition probability @xmath164 by using the present methodology is achieved by replacing the inverse temperature @xmath15 of the canonical ensemble ( [ can ] ) with a variable inverse temperature , @xmath165 .
the fulfilment of the detailed balance condition : @xmath166demands to use a certain value @xmath167 of the environment inverse temperature @xmath15 for both the direct and reverse process defined from the condition : @xmath168 which is hereafter referred to as the _ transition inverse temperature_. here , @xmath169 $ ] represents the distribution function associated with the environment inverse temperature @xmath18 , and @xmath170 , the energy varying of the system during the transition , where @xmath171 and @xmath172 .
the case of the linear ansatz ( [ linear.ansatz ] ) is special , since its transition inverse temperature @xmath167 is simply given by : @xmath173where @xmath174 and @xmath175 are the bath inverse temperatures at the initial and the final configurations respectively , @xmath176 and @xmath177 .
the direct applicability of the above result is restricted due to the final configuration @xmath178 must be previously known in order to obtain the exact value of the transition inverse temperature @xmath167 .
while such a requirement can be always satisfied in a local mc such as metropolis importance sample @xcite , the final configuration @xmath178 is _ a priori _ unknown in non - local mc methods such as clusters algorithms sw , pottsm , wolf , edwards , niedermayer , evertz , hasenbusch , dress , liu . for these cases ,
one is forced to employ an approximated value of the transition inverse temperature @xmath167 , e.g. : the inverse temperature @xmath174 at the initial configuration , @xmath179 .
although the resulting mc algorithm does not obey the detailed balance , we have shown that the deviation of the asymptotic distribution function @xmath180 from the exact distribution function @xmath181 can be disregarded for @xmath4 sufficiently large .
this is the reason why this method is particularly useful to overcome slow sampling problems in large scale mc simulations .
the use of an approximated value for the transition inverse temperature @xmath182 is not longer appropriated when one is also interested in the study of systems with size @xmath4 relatively small .
such an approximation introduces uncontrollable finite size effects that can not be dealt by using work equations ( [ ee]-[ec ] ) . in this case , it is necessary to fulfil the detailed balance condition ( [ detailed.balance ] ) to obtain the gaussian profile ( [ gens ] ) as asymptotic distribution function .
the most general way to achieve this aim is to introduce _ a posteriori _ acceptance probability @xmath183 : @xmath184to accept or reject the final configuration @xmath178 . here ,
the terms : @xmath185% , w_{f\rightarrow i}=w\left [ x_{f}\rightarrow x_{i};\beta _ { \omega } ^{f}% \right]\ ] ] represent the transition probabilities of the direct and the reverse process , respectively , which should be calculated for the given canonical clusters mc algorithm .
given a certain dependence of the environment inverse temperature @xmath186 , one can obtain from a mc run a punctual estimation of the system inverse temperature @xmath187 , the curvature @xmath188 , as well as the third and fourth derivatives of the entropy per particle @xmath189 and @xmath190 at the @xmath191-th intersection point @xmath192 .
these values can be considered to provide the next dependence @xmath193 .
the linear ansatz ( [ linear.ansatz ] ) can be rewritten in terms of the energy per particle as follows : @xmath194where @xmath195 and @xmath196 are roughly estimations of the correct values @xmath192 and @xmath187 , which are employed here as seed parameters .
the coupling constant @xmath197 is provided by its optimal dependence ( [ optimal ] ) : @xmath198by using an estimation @xmath199 of the curvature at the intersection point @xmath192 .
the seed values @xmath200 can be obtained from the previous estimated values @xmath201 by using the power expansions:@xmath202where @xmath203 is the energy step . here , it is convenient to consider the power - expansion up to the third derivative of the entropy per particle @xmath204 , since the calculation of the fourth derivative @xmath190 can be only obtained with a sufficient precision after performing a very large mc run . the efficiency of mc methods is commonly characterized in terms of the so - called _ decorrelation time _
@xmath205 , that is , the minimum number of mc steps needed to generate effectively independent , identically distributed samples in the markov chain @xcite .
its calculation in this approach can be performed by using the expression : @xmath206where @xmath207 is the variance of @xmath208 , which is defined as the arithmetic mean of the energy per particle @xmath161 over @xmath209 samples ( consecutive mc steps):@xmath210however , the decorrelation time @xmath205 only provides a partial view of the efficiency in the case of the extended canonical mc methods discussed in this work . in general , the efficiency is more appropriately characterized by the number of mc steps @xmath211 needed to achieve the convergence of a given run .
the question is that the number of mc steps @xmath211 needed to achieve a convergence of the expectation value @xmath135 of a given observable @xmath39 also depends on its thermal dispersion @xmath212 .
for example , to obtain an estimation of the expectation value @xmath213 with a statistical error @xmath214 , the number of mc steps @xmath211 should obey the following inequality : @xmath215while the fluctuating behavior of a given observable @xmath39 is an intrinsic system feature in canonical mc methods , this is not longer valid in the present framework . in this case , the fluctuating behavior crucially depends on the nature of the external influence acting on the system , e.g. : the thermal dispersion of the system energy @xmath216 depends on the coupling constant @xmath32 in eq.([dispersions ] ) .
in particular , the number of mc steps @xmath211 needed to obtain a point of the caloric curve @xmath217 with a precision @xmath218 should be evaluated in terms of the total dispersion @xmath219 introduced in eq.([min.deltat ] ) as follows : @xmath220we refer to the quantity @xmath221 as the _ efficiency factor_. clearly , an extended canonical mc method is more efficient as smaller is its efficiency factor @xmath222 .
for convenience , let us reconsider again the model system studied in our previous paper @xcite : the @xmath223-state potts model @xcite : @xmath224defined on the square lattice @xmath225 with periodic boundary conditions .
the sum @xmath226 is over nearest neighbor sites , with @xmath227 being the spin variable on the @xmath191-th site .
this model undergoes a continuous phase transition when @xmath228 , which turns discontinuous for @xmath229 .
the direct way to implement an extended canonical mc simulation of this model is by using the _ metropolis importance sample _ metro , hastings , whose transition probability is given by : @xmath230besides this last local algorithm , the mc study of the model ( [ potts ] ) can be also carried out by using non - local mc methods such as the known swendsen - wang @xcite or wolff s @xcite clusters algorithms . as discussed elsewhere @xcite , such clusters algorithms
are based on the consideration of the _ fortuin kasteleyn theorem _
@xcite:@xmath231which possibilities a mapping of this model system to a random clusters model of percolation . here
, @xmath232 is the acceptance probability of bonds , @xmath233 is the number of clusters , @xmath234 is the number of bonds , and @xmath235 is the total number of possible bonds . in order to implement the extended versions of these clusters algorithms , let us denote by @xmath236 the acceptance probability of bonds by starting from the initial configuration @xmath237 .
the transition probability of the direct process @xmath238 can be expressed as follows:@xmath239where @xmath240 is the number of inspected bonds which have been accepted , while @xmath241 is the number of inspected bonds which have been rejected . at this point , it is also important to identify the number of rejected bonds @xmath242 which have been destroyed in the final configuration @xmath243 , as well as the number @xmath244 of created bonds .
note that these bonds are responsible of the energy varying @xmath245 after the transition , @xmath246 . in the reverse process , the accepted bonds @xmath247 of the direct process
are also accepted with probability @xmath248 , while the created bonds @xmath249 as well as the rejected bonds @xmath250 that have not been destroyed in the final configuration @xmath178 are now rejected with probability @xmath251 .
thus , the transition probability of the reverse process @xmath252 can be expressed as follows : @xmath253within the canonical ensemble , where @xmath254 , it is easy to see that transition probability obeys the detailed balance condition:@xmath255for a general case with @xmath256 , one should introduce the _ a posteriori _ acceptance probability ( [ wif ] ) in order to fulfil the detailed balance , which can be expressed as follows:@xmath257where the argument @xmath258 depends on the integer numbers @xmath259 : @xmath260 the results derived from the extended versions of the metropolis importance sample , as well as the swendsen - wang and wolff s clusters algorithms are shown in fig.[finite.eps ] for the particular case of the @xmath27 seven - state potts model with @xmath261 .
each point of these dependencies have been obtained from mc runs with @xmath262 steps . for comparison purposes ,
we have also carried out the calculation of the caloric @xmath263 and the curvature @xmath264 curves by performing a direct numerical differentiation of the entropy per particles @xmath265 @xmath266 obtained from the wang - landau sampling method @xcite , which is shown in the inset panel ( @xmath267 ) .
although there exist a good agreement among all these mc results , the ones obtained from the wang - landau method seem to be less significant , overall , the results corresponding to the curvature curve @xmath268 .
the previous results constitute a clear illustration that the finite size corrections formulae introduced in subsection [ finite ] provide a significant improvement in the precision of this kind of mc calculations . in this particular example with @xmath261 , the contribution of first - order correction of finite size effects in the caloric curve @xmath269 have a typical order of @xmath270 , while the one corresponding to the second - order correction are of order @xmath271 .
nevertheless , the most significative correction of finite size effects in non - local canonical mc methods comes from the fulfilment of the detailed balance condition ( [ detailed.balance ] ) obtained after the introduction of _ a posteriori _ acceptance probability ( [ wif ] ) , whose correction has a typical order of @xmath272 .
this fact can be appreciated in fig.[corrections.eps ] for the case of the swendsen - wang clusters algorithm .
although the acceptance probability @xmath183 for clusters flipping is lower than the unity , its expectation value @xmath273 is significantly high for most of energy region ( see also in fig.[corrections.eps ] ) .
since the consideration of such an _ a posteriori _ acceptance probability @xmath183 corrects a finite size effect error @xmath274 associated with the estimation of the transition inverse temperature @xmath275 , one should observe a growth of the expectation value @xmath276 as @xmath4 increases .
such a behavior is indeed appreciated in fig.[sizeaccept.eps ] for the extended clusters algorithms . while all these dependencies have a similar qualitative behavior , the expectation value of the acceptance probability @xmath277 of the extended wolff s clusters algorithm in the paramagnetic region is larger than the one corresponding to the extended swendsen - wang method .
this is a quite expected qualitative result .
the acceptance probability @xmath183 should correct in the first case the finite size error @xmath278 related to the flipping of only one cluster , while the finite size error @xmath279 related to the extended swendsen - wang method is larger due to this algorithm involves the flipping of all system clusters . at low energies or in the ferromagnetic region ,
these methods have practically the same performance , since the number of spins belonging to the cluster is comparable to the system size @xmath4 ( see a more detailed comparison in the inset panel of fig.sizeaccept.eps ) . the qualitative behavior of the dependence @xmath280 _ versus _
@xmath161 finds a simple explanation in terms of the explicit dependence of the acceptance probability @xmath281 on the coupling constant @xmath32 . while the acceptance probability @xmath282 within the canonical ensemble where @xmath33
, this quantity undergoes a reduction with the increase of the coupling constant @xmath32 .
moreover , the optimal value of coupling constant @xmath283 employed in the present mc simulations increases with the reduction of the system curvature @xmath284 . indeed , the lowest values of @xmath277 are observed in the energy region where the system curvature also exhibits its lower values , that is , the anomalous region with negative heat capacities @xmath1 .
the size dependencies of the decorrelation time @xmath205 and the efficiency factor @xmath221 within the anomalous region with @xmath1 are shown in the main panel of fig.[decorrelation.eps ] . due to computational limitations
, the data were obtained from mc simulations with lattice sizes @xmath285 .
since the absolute values of the curvature @xmath284 are close to zero @xmath286 , the total dispersion is approximately given by the constant value @xmath287 , in accordance with eq.([optimal ] ) .
this is the reason why the dependencies @xmath288 and @xmath289 are almost displaced in a constant value along the vertical direction of the log - log graph shown in fig.[decorrelation.eps ] .
as expected , the extended version of the metropolis importance sample exhibits the largest values of the decorrelation time @xmath205 for the lattice sizes @xmath290 studied in this work .
curiously , the size dependence of its decorrelation time @xmath291 shows an abrupt transition from a power - law regime with exponent @xmath292.@xmath293 to other one with exponent @xmath294.@xmath295 close to @xmath296 ( @xmath297 ) . despite of the extended metropolis importance sample is a local mc method , the effective exponent @xmath298 for the larger system sizes is comparable to the ones associated with the extended clusters mc methods , @xmath299.@xmath300 ( swendsen - wang ) and @xmath301.@xmath302 ( wolff ) .
note also that the efficiency of these extended clusters methods are still significant despite of the consideration of the _ a posteriori _ acceptance probability ( [ wif.potts ] ) employed here to ensure the detailed balance condition ( [ detailed.balance ] ) .
the above examples confirm that the efficiency achieved with the application of the present methodology is more significant than the improvement considered by the application of re - weighting techniques such as multicanonical method and its variant , whose typical values of the exponent @xmath303 ranges from 2 to 2.5 in the case of potts models @xcite .
as already evidenced in fig.[finite.eps ] , while the wang - landau sampling method provides a good estimation of the entropy per particle @xmath304 , its underlying statistical errors are still appreciable in the curvature curve @xmath305 regardless the simulation was extended until the modifying factor fulfills the condition @xmath306 .
such an observation evidences that the results obtained from this last method are not sufficiently relaxed to provide a precise estimation of the curvature curve @xmath307 . by considering the cpu time - cost needed to achieve the convergence ,
it is more convenient to perform a punctual estimation of the caloric @xmath308 and curvature @xmath309 curves with any extended canonical mc method instead of carrying out a numerical differentiation of the microcanonical entropy @xmath304 over an energy range obtained from a re - weighting technics such as the wang - landau sampling method .
in this work , we have shown that the methodology to extend canonical mc methods inspired on the consideration of the recently obtained fluctuation relation ( [ fdr ] ) can be improved to account for the existence of finite size effects and to fulfil of the detailed balance condition ( [ detailed.balance ] ) . remarkably , despite of the consideration of the _ a posteriori _ acceptance probability ( [ wif ] ) reduces the efficiency of the clusters mc methods , it has been shown that the relaxation times needed to ensure the convergence are more significant than the ones achieved with re - weighting technics such as multicanonical methods and its variants .
for the particular case of the seven - states potts model , the consideration of any extended canonical mc algorithm enables a suppression of the super - critical slowing down associated with the occurrence of the temperature driven first - order phase transition of this model from an exponential growth to a very weak power - law dependence with exponent @xmath6 .
there is still some open questions in regard to the potentialities of the present methodology .
for example , although this method has been specially conceived to overcome the slow relaxations of canonical mc simulations near to a region with a first - order phase transition , in principle , there is no limitation that the same one can be also employed to improve canonical mc simulations near to a critical point of a continuous phase transition .
moreover , a similar extension is possible to carry out for those mc methods based on the consideration of the boltzmann - gibbs distributions : @xmath310dedx\ ] ] to account for the existence of anomalous values in other response functions besides the heat capacity .
the analysis of these questions deserve a special attention in future works . | previously , we have presented a methodology to extend canonical monte carlo methods inspired on a suitable extension of the canonical fluctuation relation @xmath0 compatible with negative heat capacities @xmath1 .
now , we improve this methodology by introducing a better treatment of finite size effects affecting the precision of a direct determination of the microcanonical caloric curve @xmath2 , as well as a better implementation of mc schemes . we shall show that despite the modifications considered , the extended canonical mc methods possibility an impressive overcome of the so - called _ super - critical slowing down _ observed close to the region of a temperature driven first - order phase transition . in this case
, the dependence of the decorrelation time @xmath3 with the system size @xmath4 is reduced from an exponential growth to a weak power - law behavior @xmath5 , which is shown in the particular case of the 2d seven - state potts model where the exponent @xmath6 . |
hd 24712 ( hr 1217 , do eri ) is a prototype of the rapidly oscillating ap ( roap ) stars that consist of about 40 members .
the oscillations , with periods ranging from @xmath9 to @xmath10 min , are high radial - order p - modes affected by strong magnetic fields of 1 kg to 25 kg . the 6.15 minute light variation of hd 24712 was discovered by @xcite . @xcite
first discovered radial velocity ( rv ) variations with an amplitude of @xmath11ms@xmath12 . a wet ( whole earth telescope ) campaign found eight frequencies with rotational side - lobes as presented in @xcite ; the paper also gives a thorough review of preceding research on this star . from accurate rv measurements
@xcite obtained two additional frequencies .
@xcite ( paper i ) studied oscillation amplitudes and phases corresponding to the two highest amplitude modes for @xmath13600 unblended spectral lines of different elements / ions , and found a diversity of these pulsational characteristics in two thirds of them .
the magnetic field of hd 24712 has been studied by many authors as reviewed in @xcite .
@xcite derived a polar magnetic strength of @xmath14 kg , a rotational inclination ( angle between line - of - sight and rotation axis ) of @xmath15 ( or @xmath16 ) and a magnetic obliquity ( the angle between the rotation and magnetic axes ) of @xmath17 ( or @xmath18 ) .
@xcite found that a slightly higher strength of @xmath19 kg is more consistent with their polarimetric observations .
recently , by inverting rotationally modulated polarized spectra , @xcite found that hd 24712 has a nearly dipole magnetic field with the strength varying between 2.2 kg and 4.4 kg , depending on the rotation phase .
one of the reasons why oscillations of roap stars are important lies in the fact that many ( often regularly spaced ) oscillation frequencies are excited simultaneously giving great potential for asteroseismic studies .
one difficulty in applying asteroseismology to roap stars is that strong magnetic fields affect the high - order p - modes in complex ways .
although the frequency of an oscillation mode generally increases with increasing magnetic field strength , it occasionally jumps down by @xmath20 to @xmath21hz and then starts increasing again @xcite . at present , there are three independent methods to calculate magnetic effects on high - order p - modes : the method of @xcite is based on a variational principle , the other two use truncated expansions with spherical harmonics to present the angular dependences of eigenfunctions @xcite .
the results from those three methods roughly agree with each other as discussed by @xcite and by @xcite .
a further complexity arises if the rotation effect is taken into account .
@xcite investigated the oscillation properties taking into account both effects of rotation and magnetic field . in our present investigation , however , we disregard the effect of rotation on the frequency and eigenfunctions for simplicity , hoping that the effect is small because the pulsation periods , @xmath22 min , are very much shorter than the @xmath23-d rotation period of hd 24712 . previously , theoretical pulsation models based on the method of @xcite were compared with the observed frequencies for @xmath24 equ @xcite , 10 aql @xcite and hd 101065 @xcite .
those attempts were more - or - less successful , although the required strength of the magnetic fields in models tended to be larger than that measured by the zeeman splittings of spectral lines .
in this paper we provide the same modelling of hd 24712 and for the first time we compare not only oscillation frequencies but also the amplitude and phase variations in the atmosphere of hd 24712 .
.unperturbed model parameters [ cols="<,^,^,^,^,^,^",options="header " , ] @xcite obtained phase lags of rv variations with respect to the light variations for various spectral lines .
these phase lags are always negative which means that luminosity maximum occurs after the rv maxima . for two other roap stars , 10 aql @xcite and hd 101065 @xcite ,
the observed phase lags are positive , in the range @xmath25 ( 10 aql ) and @xmath26 ( hd 101065 ) , respectively . allowing one cycle for the phase uncertainty , the negative phase lags in hd 24712 may be converted into the positive ones .
table [ phaselag ] lists some results for the lines formed in the most superficial layers , where optical depth of the formation of each line has been obtained by using a nlte model of @xcite .
generally the phase lag @xmath27 decreases with depth , indicating oscillations propagates outward as discussed below .
table [ phaselag ] shows that the phase lag around @xmath28 is about @xmath29 , roughly consistent with the theoretical prediction for the main frequency .
@xcite obtained amplitudes and phases of the rv variations in various spectral lines for the two highest amplitude modes ( f4 and f2 ) from time - resolved spectroscopic observations of hd 24712 . in the present paper
we consider the results at a rotational phase of 0.94 ( uves observations close to the magnetic maximum ) .
for a representative sample of , , and lines , as well as for the core , depths of formation were calculated in nlte approximation @xcite using a model atmosphere from @xcite . in these iterative model calculations
empirical stratifications of the chemical elements si , ca , cr , fe , sr , ba , pr and nd were derived for each iteration .
nlte calculations were performed for pr and nd only , because these elements are concentrated in the uppermost atmospheric layers , while other elements have a tendency to be accumulated close to photosphere where nlte effects may be neglected .
final element distributions were then used for depth of formation calculations .
the results have been converted to the relations between optical depth and rv amplitude / phase . for lines
no stratification analysis was performed , therefore , depth of line formation was calculated with a uniform yttrium distribution using an yttrium abundance @xmath30=8.60 derived from abundance analysis .
figs[fig : tauampd ] and [ fig : tauampd2nd ] compare rv amplitude / phase @xmath31 relations for f4 and f2 , respectively , with the corresponding theoretical relations of @xmath32 modes for he - depleted best models ( table [ bestmodels ] ) . for and
lines with very low rv amplitudes pulsational phases are plotted twice : as defined originally in @xcite , and shifted downward by a pulsation period taking into account the one - period uncertainty in the phase relation .
similar comparisons for models without he - depletion are given in figs[fig : tauamph ] and [ fig : tauamph2nd ] .
in comparing the observed and theoretical relations , the theoretical amplitude is multiplied by an arbitrary factor , and the theoretical phase delay is fitted at the outer boundary ( @xmath33 ) by applying an arbitrary shift . in the outermost layers oscillation phase gradually delays toward the outer boundary , indicating the presence of running waves propagating outward .
this is consistent with the fact that all the observed oscillation frequencies of hd 24712 are larger than the acoustic cut - off frequencies of theoretical models .
( if a reflective outer boundary condition is imposed , the phase in the outermost layers is nearly constant . )
the gradient of the phase variation in the outermost layers is well reproduced by theoretical relations .
we note that if the @xmath34 term in equation ( [ eq : edd ] ) is dropped , the gradient becomes steeper , indicating that the @xcite theory is consistent with the observations .
the phase distribution in hd 24712 has a jump between @xmath35 and @xmath36 for both f4 and f2 , indicating the presence of a quasi - node there .
this property agrees with the theoretical phase variations for a - models with the @xmath8 relation having a small temperature inversion at @xmath37 ( fig .
[ fig : ttau ] ) . a rapid phase change in s - models with the standard @xmath8 relation
, however , occurs between @xmath38 and @xmath39 , deeper than the phase jump derived from observations .
thus , the position of the phase jump supports the @xmath8 relation with a temperature inversion at @xmath37 as obtained by @xcite . whether the phase of theoretical models jump downward or upward depends on the assumption of helium depletion , and sometimes on the strength of the magnetic field .
generally , phase tends to decrease inward at the quasi - node in helium - depleted models , and to increase inward in models without helium depletion . in the best model of the ah165 series with @xmath40 kg , the phase of rv variations for the main frequency f4 decreases steeply inward at @xmath41 ,
while it _ increases
_ steeply at the same layer if @xmath42 is slightly increased to @xmath43 kg ( fig .
[ fig : tauamph ] ) , although for the secondary frequency f2 the phase increases steeply inward for both cases ( fig .
[ fig : tauamph2nd ] ) .
however , such a strong dependence of the direction of the phase jump on @xmath42 does not occur in the other cases . to obtain information on the helium depletion in the atmosphere of hd 24712 ,
further observations are needed to fill the gap between @xmath44 and @xmath45 . towards deep interior ,
theoretical pulsation phase approaches a constant ; i.e. , nearly a standing wave is realized in the deep atmosphere .
however , the innermost fe ii data for the main pulsation f4 deviate considerably from the theoretical curves .
a possible cause may be related to the surface element distribution ; fe is concentrated around the equator , while y ( as all rare earth elements , including pr and nd ) is concentrated near magnetic poles @xcite .
oscillation amplitude increases rapidly in the outermost layers due to a rapid decrease in the gas density .
in the layers of @xmath46 , theoretical amplitude variations roughly agree with the observed ones for both f4 and f2 . in the outermost layers with @xmath47 , however , theoretical amplitudes deviate significantly from the observations .
observed amplitudes level off in the most superficial layers , the cause of which is not clear ; it could be the effect of nonlinear dissipation , or the density stratification could differ significantly from our simple models .
we discussed theoretical models for the oscillations of the roap star hd 24712 ( hr 1217 ) . observed frequencies are fitted well with theoretical ones for models whose positions on the hr diagram are close to the position of hd 24712 determined by spectroscopy along with the hipparcos parallax .
the observed main frequency f4 is identified as a quasi - quadrupole ( @xmath32 ) mode , whose amplitude on the surface is strongly concentrated in the polar regions and is suppressed significantly around the equator . the theoretical amplitude distribution predicts a rotational modulation of the pulsation amplitude which is consistent with the observations .
we modelled for the first time the distributions of the phase and amplitude of rv variations as a function of atmospheric height and compared these with the observed distributions .
the gradual outward increase of phase lag in the outermost layers is well reproduced by theoretical results obtained with a running - wave outer boundary condition .
the presence of a steep phase change between @xmath48 and @xmath45 favours a @xmath8 relation with a temperature inversion at @xmath49 , rather than a standard @xmath8 relation .
although our models agree with most of the observed properties of the oscillations in the atmosphere of hd 24712 , we have recognized three problems to be solved in the future . 1 . the observed amplitude of velocity variation levels off in the outermost layers with @xmath50 , while in all the models we have calculated amplitude increases steeply outward in the layers with @xmath51 .
2 . if the observed main frequencies are identified with @xmath32 and @xmath52 modes , to explain their nearly equal spacings the small frequency spacing defined as @xmath53 $ ] must be as small as @xmath54@xmath55hz .
for all the models calculated , however , the small spacing was always larger than @xmath56@xmath55hz .
the problem goes away if we assume the observed equal spacings of frequencies are produced by @xmath32 and @xmath57 modes rather than @xmath32 and @xmath52 modes .
however , this solution has a problem ; @xmath57 modes have rotational amplitude modulations different from the modulation seen in the light variations of hd 24712 .
we have found no excited oscillation modes with frequencies appropriate for hd 24712 ; in other words all the modes examined are damped ones .
high - order p - modes in roap stars are generally thought to be excited by the @xmath58-mechanism in the hydrogen ionization zone @xcite .
however , the @xmath58-mechanism does not seem to be strong enough to excite the super - critical high - order p - modes in hd 24712 .
it is interesting to note that the inefficiency of the @xmath58-mechanism is in common with the case of another cool roap star hd 101065 having many regularly spaced frequencies @xcite .
probably we need to find a new excitation mechanism for these coolest roap stars .
we thank dr .
l. mashonkina for providing us with the numerical data on the nlte pr - nd line depth formation .
hs thanks chris cameron for stimulating discussions .
this work was supported by research grants from the rfbi ( 08 - 02 - 00469a and 09 - 02 - 00002a ) and from leading scientific school ( 4354.2008.2 ) to tr and ms .
huber d. , saio h. , gruberbauer m. , weiss w. w. , rowe j. f. , hareter , m. , kallinger t. , reegen p. , matthews j. m. , kuschnig r. , guenther d. b. , moffat a. f. j. , rucinski s. , sasselov d. , walker g. a. h. , 2008 , a&a , 483 , 239 | we discuss pulsations of the rapidly oscillating ap ( roap ) star hd 24712 ( hr 1217 ) based on nonadiabatic analyses taking into account the effect of dipole magnetic fields .
we have found that all the pulsation modes appropriate for hd 24712 are damped ; i.e. , the kappa - mechanism excitation in the hydrogen ionization layers is not strong enough to excite high - order p - modes with periods consistent with observed ones , all of which are found to be above the acoustic cut - off frequencies of our models .
the main ( 2.721 mhz ) and the highest ( 2.806 mhz ) frequencies are matched with modified @xmath0 and @xmath1 modes , respectively .
the large frequency separation ( @xmath2hz ) is reproduced by models which lay within the error box of hd 24712 on the hr diagram .
the nearly equally spaced frequencies of hd 24712 indicate the small frequency separation to be as small as @xmath3hz . however , the small separation derived from theoretical @xmath4 and 2 modes are found to be larger than @xmath5hz .
the problem of equal spacings could be resolved by assuming that the spacings correspond to pairs of @xmath0 and @xmath6 modes ; this is possible because magnetic fields significantly modify the frequencies of @xmath6 modes .
the amplitude distribution on the stellar surface is strongly affected by the magnetic field resulting in the predominant concentration at the polar regions .
the modified amplitude distribution of a quasi - quadrapole mode predicts a rotational amplitude modulation consistent with the observed one .
amplitudes and phases of radial - velocity variations for various spectral lines are converted to relations of amplitude / phase versus optical depth in the atmosphere .
oscillation phase delays gradually outward in the outermost layers indicating the presence of waves propagating outward .
the phase changes steeply around @xmath7 , which supports a @xmath8 relation having a small temperature inversion there .
[ firstpage ] stars : individual : hd 24712 stars : magnetic fields stars : oscillations |
recently , due to experimental advances in coupling a dilute gas of bosons to a single mode of an optical cavity , many theoretical and experimental works have been performed in order to explore and explain the physics of such complex systems .
bose - einstein condensate ( bec ) itself is a very rich platform which allows us to examine different properties of quantum systems such as quantum turbulence @xcite , quantum chaos @xcite and entanglement @xcite .
a bose - einstein condensate can also be used to study nonequilibrium dynamics and decoherence in finite quantum systems @xcite as well as calculation of fluctuation indices for atomic systems @xcite . moreover ,
if confined in an optical lattice @xcite , bose - einstein condensates provide the possibility to observe solid - state physics processes , such as localization @xcite , in these atomic systems .
further uses of becs are exemplified in transport problems @xcite , interaction driven instabilities @xcite , stability of boson - fermion gaseous mixtures @xcite , and raman pump - probe experiments @xcite . in a more complex setup , where a dilute condensate of bosons is confined inside a high - finesse optical cavity subject to external laser fields ,
if the cavity mode and laser fields are far - detuned from the atomic transition frequency of the condensate atoms , system is in dispersive regime . under the dispersive regime conditions
, atom - field interaction provides an optical lattice for the condensate atoms which affects their mechanical motion .
on the other hand , the atoms cause a position - dependent phase - shift of the cavity mode . as a result ,
the condensate - cavity system is highly nonlinear with nonlocal nonlinearities which give rise to a series of interesting physical phenomena such as self - organization of condensate atoms , cavity enhanced superradiant scattering , dicke quantum phase transition @xcite and optical bistability .
the nonlinearity caused by atom - filed interaction can be more effective @xcite than the nonlinearity from atom - atom interaction , which itself can play crucial role in scattering dynamics of condensates @xcite .
the optical bistability , which is the focus of this article , has been studied in spinor becs @xcite and two - component bec with two modes coupled by a classical field @xcite . in a system consisting of single - mode bec in an optical cavity
, a transverse laser pump has also been used to control the bistability of cavity photons induced by a parallel pump @xcite . in this work
we consider a two - mode bec in a one - dimensional optical cavity , where two laser fields are applied to the system , one parallel to the cavity axis and the other one perpendicular to it .
specifically we assume that the transverse pump is scattered to the cavity mode by condensate atoms in the raman scheme @xcite . under this condition
, we examine the effect of transverse pump strength on the bistability of both mean cavity photon number and normalized population difference ( magnetization ) of the two modes .
stationary state of cavity field and condensate wavefunction are obtained under discrete mode approximation ( dma ) which has been shown to be reliable for similar systems @xcite .
this paper is organized as follows . in sec.[model ] first the model for our system is introduced , then the hamiltonian of the system and equations of motion of cavity field and condensate are derived . using dma , we solve the equations of motion for the steady state of the system in sec.[results ] , where we show how mutual bistability of the mean cavity photon number and magnetization take place under certain conditions for laser field intensities and cavity - atom coupling strength .
finally , we summarize our work in sec.[summary ] .
we consider a condensate of @xmath0 atoms , each with two internal degrees of freedom , shown as states @xmath1 and @xmath2 and an excited state @xmath3 , in a one dimensional cavity along the @xmath4 axis .
the cavity has a single mode with frequency @xmath5 and is subjected to laser field with frequency @xmath6 , which is far detuned from the atomic transition , in directions parallel and perpendicular to its axis . if the transverse laser field interacts with matter through raman scattering ( fig . 1 )
, then the full hamiltonian of the system will have following form @xmath7 where @xmath8 is the frequency of transition between modes @xmath1 and @xmath2 and @xmath9 are the interaction strengths of atoms in modes @xmath10 and @xmath11 .
the parallel laser field intensity is shown by @xmath12 and @xmath13 and @xmath14 are the annihilation and creation operators of the cavity mode .
the interaction of atoms with the transverse pump is shown by the raman scattering hamiltonian ( @xmath15 ) and has the following form @xmath16 where @xmath17 and @xmath18 are the atom - pump and atom - cavity coupling strengths , respectively , and the hamiltonian is written without rotating wave approximation .
after adiabatically eliminating the excited state @xmath19 and introducing @xmath20 and @xmath21 , with @xmath22 being the pump detuning from the atomic transition , one obtains @xmath23 .
right : two internal modes ( @xmath1 and @xmath2 ) of the bec atoms are coupled by cavity field and laser field via the atomic excited state @xmath3 in raman scattering manner .
both laser field and cavity field are detuned from the atomic transition frequency.,title="fig:",scaledwidth=80.0% ] [ fig : raman ] substituting ( [ eq : hraman2 ] ) into ( [ eq : h1 ] ) , moving to a rotating frame defined by the unitary operator @xmath24 , and ignoring two - photon processes result in @xmath25 for the hamiltonian of the system where @xmath26 here @xmath27 and @xmath28 are the ascending and descending operators in the two - mode manifold and @xmath29 is the pump - cavity detuning . from equations ( [ eq : h2 ] ) and ( [ eq : hsingle ] ) one can derive the heisenberg equations of motion for condensate modes and cavity field as follow @xmath30 @xmath31 and @xmath32 where we have introduced a decay rate @xmath33 for the cavity in ( [ eq : heisen - alpha ] ) . in order to obtain the equations of motion for the expectation values of field operators , we further simplify our system .
since we work with a condensate of large number of atoms , we can safely treat the condensate as a coherent state and disentangle it from photons .
we also neglect higher order correlations within the photon subsystem and use @xmath34 and @xmath35 instead of @xmath36 and @xmath37 . at the end , as far as we only deal with the intensities , we treat the photon as a coherent state as well . as a result , following equations of motion are obtained @xmath38 @xmath39 @xmath40 where @xmath41 is the real part of the cavity field @xmath42 .
to solve the equations of motion derived in sec.[model ] for the steady state of the system , we employ discrete mode approximation .
the ground state of the condensate , without any laser field , is a homogeneous macroscopic state with zero momentum , which we refer to by @xmath43 . above superradiance threshold
one can assume that the condensate is fragmented to a symmetric superposition of the states with @xmath44 momentum due to the transverse laser field .
on the other hand , since the cavity mode is excited by the parallel pump , absorption and emission of the cavity photons can excite the condensate to the superposition of states with @xmath45 momentum .
therefore if we consider first order perturbation on the homogeneous wavefunction @xmath43 , following functions can be used as the basis for dma : @xmath46 with @xmath47 being the length of the cavity ( condensate ) .
now the wavefunctions of the two modes of the condensate can be expanded in this basis as follow @xmath48 if we substitute these wavefunctions into the equations of motion of the two modes , by ignoring external potentials and atom - atom interaction , the equation of motion for the condensate wavefunction can be written in a compact form : @xmath49x,\end{aligned}\ ] ] where @xmath50 and @xmath51 @xmath52 @xmath53 @xmath54 @xmath55 here @xmath56 is the recoil frequency . to examine the equilibrium properties of the system , we set @xmath57 in ( [ eq : eom - alpha ] ) which results in @xmath58 on the other hand , one can easily obtain the following alternative expressions for the integrals in the above equation @xmath59 @xmath60 therefore cavity field @xmath42 and averaged photon number @xmath61 will read as @xmath62 @xmath63 ( left ) and normalized magnetization @xmath64 ( right ) as functions of @xmath65 for different values of parallel pump strength @xmath12 .
we remind that @xmath65 is proportional to transverse pump strength and atom - cavity coupling strength . in all plots
@xmath66 , @xmath67 , @xmath68 , @xmath69 , @xmath70 , and @xmath71.,scaledwidth=100.0% ] ( left ) and normalized magnetization @xmath64 ( right ) as functions of @xmath65 for different values of atom - cavity coupling strength @xmath72 . in all plots
@xmath73 , @xmath67 , @xmath68 , @xmath69 , @xmath70 , and @xmath71.,scaledwidth=100.0% ] here one should notice that if there is no transverse laser field ( @xmath74 ) then @xmath75 and @xmath76 .
as a result , there will be no transition between the two modes and if atoms are initially in the ground state @xmath1 , they will stay there .
therefore expectation values of both @xmath77 and @xmath78 in ( [ eq : steadyn ] ) or , equivalently , the integrals in ( [ eq : steadyalpha1 ] ) would be zero and no bistability in the system is expected .
the importance of the transverse laser pump could be predicted from the crucial role it plays in raman scattering hamiltonian .
on the other hand to better understand the contribution of parallel laser field in bistability , we remind that due to dependence of @xmath79 on @xmath42 , expectation values in numerator and denominator of ( [ eq : steadyn ] ) are functions of @xmath80 .
to lowest order , we can assume that these terms depend on @xmath80 linearly .
therefore ( [ eq : steadyn ] ) is a cubic equation of @xmath80 which in the case of @xmath81 has always a zero root .
therefore , to avoid a zero root in ( [ eq : steadyn ] ) , we will consider cases where @xmath82 . to obtain the mean cavity field @xmath80 for the steady state of the system
we need to find the wavefunction @xmath83 in steady state .
since @xmath84 is nonlinear due to the dependence of @xmath79 on @xmath42 , we first solve it for @xmath85 and @xmath83 with a guess for the value of @xmath42 .
then by substituting the resulting @xmath83 into ( [ eq : steadyalpha2 ] ) , a new @xmath42 is obtained .
if this new @xmath42 is equal to the guessed value then the steady state is reached , otherwise we repeat the procedure , while using this new @xmath42 as the guess , until the steady state is attained . in addition to the average photon number , in a system consisting of two - mode condensate , there is another quantity which reveals the nonlinear effects of matter - field interaction and that is the normalized population difference ( magnetization @xmath64 ) of two modes . in our system magnetization
is defined as @xmath86 , where @xmath87 is the total number of atoms and is fixed .
2 shows mean cavity field @xmath80 and normalized magnetization @xmath64 as functions of @xmath65 for different values of parallel pump strength @xmath12 . we have considered a condensate of @xmath88 atoms in a cavity with decay rate @xmath68 .
the laser field is detuned from cavity mode by @xmath69 and from atomic transition by @xmath67 . in all plots ,
atom - cavity coupling is assumed to be @xmath66 . as one can observe from these plots , the effect of parallel pump on widening the area of bistability is not monotonic . while increasing @xmath12 from @xmath89 to @xmath90 results in a wider interval of @xmath65 in which bistability happens , further increasing it to @xmath91 has the opposite effect .
moreover , @xmath12 has larger impact on the width of the area in which bistability takes place than on the values of @xmath80 and @xmath64 in bistable region .
the interesting point about this system is the fact that nonlinear effects of matter - field coupling result in bistable behavior of both matter and field , such that , bistability happens for magnetization of condensate atoms exactly at the same region of @xmath65 where @xmath80 has shown bistable behavior .
moreover , it can be seen in fig . 2 that , in some points , two states not only with different values of magnetization but also with different signs of magnetization are stable .
in the next step , to better understand the role of atom - cavity coupling on bistable behavior of the system , we keep the value of parallel pump strength @xmath12 constant and increase the strength of atom - cavity coupling .
3 shows the average photon number @xmath80 and normalized magnetization @xmath64 as functions of @xmath65 for the same values of parameters used in fig . 2 but this time with @xmath73 and for three different values of @xmath72 .
clearly , atom - cavity coupling has a significant effect on the width of the area in which bistability happens .
more importantly , @xmath72 can change the distance between the two stable branches .
it is worth mentioning that @xmath65 itself is proportional to @xmath18 and therefore for a fixed value of @xmath65 , larger @xmath72 means smaller value for @xmath17 . as a result
, increasing @xmath72 makes it possible to achieve bistability with smaller values and wider range of atom - laser coupling @xmath17 .
in this work we developed an effective hamiltonian and equations of motion for a cavity - condensate system consisting of a two - mode bec in a one dimensional cavity , while parallel and transverse laser fields are applied to the system . under dma , simultaneous and mutual bistability ( multistability ) of cavity field and
population difference ( magnetization ) of the two modes has been observed for different values of transverse and parallel pump strength .
the system shows bistable behavior for quite wide range of parameters .
moreover it has been shown that , with strong enough cavity - matter coupling strength , bistability happens between two states with different sign of magnetization .
the authors would like to thank y. castin and m. . oktel for fruitful discussions and s. safaei would like to acknowledge discussions with p. elahi .
this work was supported by tbitak ( grant . no .
109t267 ) . v. i. yukalov , laser phys .
lett . * 7 * , 467 ( 2010 ) .
r. f. shiozaki , g. d. telles , v. i. yukalov , and v. s. bagnato , laser phys .
* 8 * , 393 ( 2011 ) .
j. a. seman , e. a. l. henn , r. f. shiozaki , g. roati , f. j. poveda - cuevas , k. m. f. magalhes , v. i. yukalov , m. tsubota , m. kobayashi , k. kasamatsu , and v. s. bagnato , laser phys .
* 8 * , 691 ( 2011 ) .
h. w. xiong and b. wu , laser phys .
* 8 * , 398 ( 2011 ) .
m. e. tagin , m. . oktel , l. you , and . e. mstecapliolu , laser phys .
* 20 * , 700 ( 2010 ) .
v. i. yukalov , laser phys .
lett . * 8 * , 485 ( 2011 ) .
v. i. yukalov , laser phys .
* 7 * , 831 ( 2010 ) . v. i. yukalov , laser phys . *
19 * , 1 ( 2009 ) .
y. cheng and s. k. adhikari , laser phys .
* 7 * , 824 ( 2010 ) .
a. zenesini , h. lignier , c. sias , o. morsch , d. ciampini , and e. arimondo , laser phys .
* 20 * , 1182 ( 2010 ) . v. o. nesterenko , a. n. novikov , and e. suraud , laser phys .
* 20 * , 1149 ( 2010 ) .
b. ole , p. zi , j. chwedeczuk , k. sacha , and m. trippenbach , laser phys .
* 20 * , 671 ( 2010 ) .
a. l. subai , s. sevinli , p. vignolo , and b. tanatar , laser phys .
* 20 * , 683 ( 2010 ) . n. souther , r. wagner , p. harnish , m. briel , and s. bali , laser phys .
* 7 * , 321 ( 2010 ) .
d. nagy , g. knya , g. szirmai , and p. domokos , * 104 * , 130401 ( 2010 ) .
k. baumann , c. guerlin , f. brennecke , and t. esslinger , nature * 464 * , 1301 - 1306 , 29 april ( 2010 ) . s. yang , m. al - amri , j. evers , and m. s. zubairy , * 83 * , 053821 ( 2011 ) . v. p. kraynov and h. a. ishkhanyan , laser phys .
* 20 * , 1856 ( 2010 ) .
l. zhou , h. pu , h. y. ling , and w. zhang , * 103 * , 160403 ( 2009 ) .
l. zhou , h. pu , h. y. ling , k. zhang , and w. zhang , * 81 * , 063641 ( 2010 ) .
y. dong , j. ye , and h. pu , * 83 * , 031608(r ) ( 2011 ) . m. m. cola and n. piovella , * 70 * , 045601 ( 2004 ) . h. uys and p. meystre , * 75 * , 033805 ( 2007 ) . j. m. zhang , f. c. cui , d. l. zhou , and w. m. liu , * 79 * , 033401 ( 2009 ) . | we consider a two - component bose - einstein condensate in a one - dimensional optical cavity .
specifically , the condensate atoms are taken to be in two degenerate modes due to their internal hyperfine spin degrees of freedom and they are coupled to the cavity field and an external transverse laser field in a raman scheme .
a parallel laser is also exciting the cavity mode .
when the pump laser is far detuned from its resonance atomic transition frequency , an effective nonlinear optical model of the cavity - condensate system is developed under discrete mode approximation ( dma ) , while matter - field coupling has been considered beyond the rotating wave approximation . by analytical and numerical solutions of the nonlinear dynamical equations ,
we examine the mean cavity field and population difference ( magnetization ) of the condensate modes .
the stationary solutions of both the mean cavity field and normalized magnetization demonstrate bistable behavior under certain conditions for the laser pump intensity and matter - field coupling strength . + + topic : physics of cold trapped atoms ( report number : 6.6.4 ) |
a variety of phenomena can be modeled by the one dimensional asymmetric simple exclusion process ( asep ) and its generalizations ( see @xcite and references therein ) .
the model has a natural interpretation as a description of traffic flow and constitutes a basis for more realistic ones @xcite . in traffic flow theories ,
the formation of traffic jams due to `` impurities '' is one of the fundamental problems . in the asep models ,
two kinds of impurities are discussed in the literature .
the first one is `` dynamic '' impurities , i.e. , defective particles which jump with a rate lower than others @xcite . in the traffic terminology
, such moving defects can be visualized as slow cars on a road .
the other kind is `` static '' impurities such as imperfect links where the hopping rate is lower than in other links @xcite .
both types of impurities can produce shocks in a system .
presently a limited amount of exact results is available for the shock formation and most of them are for models with the random sequential update @xcite .
the implementation of the update is an essential part of the definition of a model and it is of prime interest to determine whether distinct updating schemes can produce different behaviors . the aim of this work is to investigate consequences of changing the updating scheme of the model . here , we study an exactly solvable traffic model with two types of ordered sequential updates , which is identical to the model studied in ref .
@xcite except for the updating schemes .
consider two parallel one dimensional chains , each with @xmath0 sites and the periodic boundary condition .
there are @xmath1 cars and @xmath2 trucks in the first and the second chain , respectively , and cars move to the right while trucks move to the left .
we introduce inter - chain interaction that forbids a car and a truck to occupy two parallel sites simultaneously .
then the state of the system can be described by a single set of occupation numbers @xmath3 where @xmath4 ( empty site ) , @xmath5 ( occupied by a car ) , or @xmath6 ( occupied by a truck ) .
there are three possible hopping processes : @xmath7 the reduction factor @xmath8 ( @xmath9 ) is related to the width of roads : @xmath10 corresponds to a very wide road or a highway with a lane divider , and @xmath11 corresponds to a one lane road .
thus the value @xmath12 can be used as a measure of the road narrowness .
recently the model is studied with the random sequential update ( rsu ) . in an infinitesimal time interval @xmath13 , each link is updated with the probability that is the product of the relevant rate and @xmath13 [ since @xmath14 does not affect the dynamics at all in the rsu scheme , we choose @xmath15 for the rsu scheme ] . alternatively the ordered sequential update ( osu ) can be used .
one first chooses a particular site , i.e. the site @xmath0 , and updates the state of the links consecutively either in the backward direction @xmath16 [ backward sequential update ( bsu ) ] or in the forward direction @xmath17 [ forward sequential update ( fsu ) ] .
in contrast to the rsu update , the time is discrete in the bsu and fsu schemes , and in each time step , hopping occurs in links with the probabilities that are equal to the rates in eq .
( [ eq : dynamics ] ) .
the two - way traffic model is equivalent to a two - species asep and it can be solved exactly by the method of the matrix product state ( mps ) . in the rsu scheme ,
the steady state weight @xmath18 of a given configuration @xmath3 is proportional to the trace of the normal product of some matrices @xcite : p_s ( _ 1,_2,,_n ) ~tr(x_1x_2x_n ) [ mpsforrsu ] where @xmath19 for @xmath20 ; @xmath21 for @xmath22 ; and @xmath23 for @xmath24 , and these matrices satisfy the quadratic algebra de= d+e , ae = a , da= a ( ) .
[ for1 ] in a similar way , @xmath18 in the bsu ( fsu ) scheme can be written as follows : p_s(_1,_2,,_n ) ~tr(x_1x_2_n ) [ for3 ] where @xmath25 , or @xmath23 ( @xmath26 or @xmath27 ) depending on @xmath28 ( @xmath29 ) .
note that the matrices at the site @xmath0 are different from those at other sites @xcite .
the presence of the hatted matrix in eq .
( [ for3 ] ) breaks the translational invariance of the problem . in ref .
@xcite , it is shown that with a simple assumption , = a+a , = d+d , = e + e , [ for8 ] where @xmath30 , and @xmath31 are carefully chosen real numbers , the relevant matrix algebra for @xmath32 becomes identical to the one in eq .
( [ for1 ] ) except for the renormalizations of @xmath33 and @xmath34 to @xmath35 and @xmath36 . in the bsu scheme ,
the choice @xmath37 gives @xmath38 , @xmath39 , and in the fsu scheme , the choice @xmath40 results in @xmath41 , @xmath42 .
in this section we consider the special case where there is only one truck and evaluate average velocities .
the mps method allows exact evaluations of velocities .
we first present the car velocity in the thermodynamic limit . in the bsu scheme , v_car= \ { ll ( 1-n ) & if n1 + ' '' ''
-1-n n & if n1 , .
[ v_car_bsu ] and in the fsu scheme , v_car= \ { ll & if n1 + ' '' '' ( 1-n ) n & if n1 . .
[ v_car_fsu ] and different values of @xmath14 .
( b ) average velocity of cars in the fsu and rsu schemes for @xmath43 and different values of @xmath14 .
( c ) average velocity of the truck in the bsu and rsu schemes for @xmath43 and different values of @xmath14 .
( d ) average velocity of the truck in the fsu and rsu schemes for @xmath43 and different values of @xmath14 .
after ref .
, title="fig : " ] and different values of @xmath14 .
( b ) average velocity of cars in the fsu and rsu schemes for @xmath43 and different values of @xmath14 .
( c ) average velocity of the truck in the bsu and rsu schemes for @xmath43 and different values of @xmath14 .
( d ) average velocity of the truck in the fsu and rsu schemes for @xmath43 and different values of @xmath14 .
after ref .
, title="fig : " ] + ( a)(b ) + and different values of @xmath14 .
( b ) average velocity of cars in the fsu and rsu schemes for @xmath43 and different values of @xmath14 .
( c ) average velocity of the truck in the bsu and rsu schemes for @xmath43 and different values of @xmath14 .
( d ) average velocity of the truck in the fsu and rsu schemes for @xmath43 and different values of @xmath14 .
after ref .
, title="fig : " ] and different values of @xmath14 .
( b ) average velocity of cars in the fsu and rsu schemes for @xmath43 and different values of @xmath14 .
( c ) average velocity of the truck in the bsu and rsu schemes for @xmath43 and different values of @xmath14 .
( d ) average velocity of the truck in the fsu and rsu schemes for @xmath43 and different values of @xmath14 . after ref .
, title="fig : " ] + ( c)(d ) + figs .
1(a ) and ( b ) show the behaviors of @xmath44 in the two updating schemes as a function of the road narrowness @xmath45 while the values of @xmath46 and @xmath47 are fixed .
in both schemes , a continuous phase transition is evident : while @xmath48 is constant below the critical narrowness , it begins to drop suddenly at the critical narrowness ( @xmath49 ) , generating a cusp .
thus above the critical narrowness , a single truck results in global effects .
the functional dependence of the decrease differs in the two updating schemes .
the exact expressions for the truck velocity , on the other hand , are rather lengthy especially below the critical narrowness and we present the functional dependence on the road narrowness only through figures [ figs .
1(c ) and ( d ) ] . for exact expressions ,
see ref .
the appearance of a continuous transition at the critical narrowness ( @xmath49 ) is clear and the functional dependence of the truck velocity again differs in the two updating schemes .
in the rsu scheme , the probability to find a car at a site depends only on its relative distance @xmath50 to the truck due to the translational invariance . in ref .
@xcite , it is found that in the high density phase @xmath51 , the probability or density profile @xmath52 becomes n(x)= \ { cl 1 & n-1-1 + ' '' '' 1 & otherwise . .
[ nhighrsu ] note that the system consists of two regions , a traffic jam region in front of the truck and a free flow region behind it . in the low density phase @xmath53 , on the other hand , the presence of the truck has only local effects . in the thermodynamic limit
, the car density becomes n(x ) = n \ { 1 + ( n)^x } [ nlowrsu ] which shows that the disturbance by the truck decays exponentially with a characteristic length scale @xmath54 in the ordered sequential updates , complications occur in the definition of the probability itself .
since the choice of a particular site as a starting point of the update breaks the translational invariance , the probability to find a car at @xmath50 sites in front of the truck depends not only on the relative distance @xmath50 but also on the truck location .
this unnecessary complication can be avoided by choosing the starting point in an even way . in ref .
@xcite , it is found that with this `` homogeneous '' choice , the expressions for the probability in the bsu and the fsu scheme become _ identical _ to eqs .
( [ nhighrsu],[nlowrsu ] ) except for the replacement of the bare parameters @xmath33 and @xmath34 with the renormalized ones @xmath35 and @xmath36 .
characteristics of an exactly solvable two - way traffic model are investigated with the osu scheme and both qualitative and quantitative differences are found compared to those with the rsu scheme @xcite .
our approach is based on the so - called matrix product formalism which allows analytic solutions . in the osu schemes ,
the choice of a particular site as a starting point of the update breaks the translational invariance of the steady state measure , which is also evident in the form of the mps .
thus an averaging over the different choices of the update starting point is necessary to restore the translational invariance to the system .
performing the translationally invariant averaging , it is found that for characteristics such as the density profile of cars and the density - density correlation function , the differences in the updating schemes can be fully taken into account by the proper renormalization of the parameters @xmath55 and @xmath8 .
however this is not the case with average velocities .
changing the update scheme affects velocities in a more complicated manner and the renormalization of the parameters is not sufficient to account for different behaviors of @xmath56 and @xmath57 in different updating schemes . especially the dependence of @xmath48 and @xmath58 on the road narrowness @xmath59 can vary qualitatively depending on the updating schemes . *
acknowledgments * + m.e.f . would like to thank v. karimipour for fruitful comments and d. kim , r. asgari for useful helps .
thanks d. kim for bringing his attention to this problem .
w.l . was supported by the korea science and engineering foundation through the fellowship program and the src program at snu - ctp . | within the formalism of the matrix product ansatz , we study a two - species asymmetric exclusion process with backward and forward site - ordered sequential updates .
this model describes a two - way traffic flow with a _ dynamic impurity _ and shows a phase transition between the free flow and the traffic jam .
we investigate characteristics of this jamming and examine similarities and differences between our results and those with the random sequential update @xcite . |
since the standard model ( sm ) can not give an explanation for dark matter , new physics is expected . in this talk , we review the signatures of dark matter in two models which contain @xmath2 triplet scalars with the hypercharges y=0 and 2 under @xmath3 , respectively @xcite . in these models ,
the triplets are odd under an @xmath4 symmetry so that they neither directly couple to the sm fermions nor develop vacuum expectation values ( vevs ) , while the neutral components of the triplets are the dark particles .
we will refer to the models as the inert triplet models ( itms ) .
the number of new parameters in the y=0 itm is three compared to the sm , which is the same as those in the inert singlet model @xcite .
clearly , the y=0 itm is one of the minimal inert models .
similarly , the y=2 itm like the inert doublet model @xcite@xmath5@xcite has five new parameters .
the relic abundance of the cold dark matter in the universe is determined to be @xcite @xmath6 where @xmath7 is the scaled current hubble parameter in units of @xmath8 . on the other hand , the direct searches also provide constraints on dark matter .
for example , the spin - independent ( si ) dark matter cross section has to satisfy @xcite @xmath9 note that to have the above constraint , we have assumed that dark matter is the weakly interacting massive particle ( wimp ) with its mass smaller than @xmath10 @xcite .
we extend the sm by adding a zero hypercharge @xmath2 triplet scalar with an unbroken @xmath4 symmetry .
the relevant lagrangian is given by @xmath11 + \lambda_1 |h^\dag h|^2 + \lambda_2 \left({{\rm tr}}[t^2]\right)^2 + \lambda_3 h^\dag h\ { { \rm tr}}[t^2]\ , , \label{eq:3vertex}\end{aligned}\ ] ] where @xmath12 is the covariant derivative including the sm gauge bosons , and the sm higgs doublet @xmath13 and the triplet @xmath14 scalars are defined as @xmath15 respectively , with @xmath16 and @xmath17 .
the stability condition of the higgs potential requires @xmath18 the potential in eq .
( [ eq:3vertex ] ) becomes a local minimum if and only if @xmath19 where @xmath20 .
after @xmath21 acquires the vev , the scalars gain the following masses : @xmath22 at tree level . due to the radiative corrections @xcite ,
the masses of @xmath23 and @xmath24 are split as @xmath25 hence , @xmath26 turns out to be the lightest component of the triplet scalar with its stability protected by the @xmath4 symmetry .
since the triplet scalar is added to the sm , one may expect that it affects the so - called oblique ( s and t ) parameters .
in general , however , an @xmath27 triplet has no contribution to the s parameter , while the contribution to the t parameter is also vanishing in the limit of @xmath28 @xcite . even if we consider the small mass splitting in eq .
( [ eq : gap ] ) , its effect is negligibly small , such that @xmath29 \simeq 0,\end{aligned}\ ] ] where @xmath30 is the weak mixing angle and @xmath31 is the @xmath32 boson mass .
therefore , the constraint on the higgs boson mass ( @xmath33 ) from the precision electroweak measurements is the same as that in the sm . in our calculation
, we restrict @xmath33 to be within the range of @xmath34 as estimated in ref .
@xcite with the excluded region of @xmath35 reported by the tevatron @xcite . in the model with the inert triplet scalar of @xmath36 , the @xmath4 invariant scalar potential
is given by @xmath37 + \lambda_1 |h^\dag h|^2 + \lambda_2 { { \rm tr}}[t^\dag t t^\dag t ] + \lambda_3 \left ( { { \rm tr}}[t^\dag t ] \right)^2 \nonumber \\ & & + \lambda_4 h^\dag h\ { { \rm tr}}[t^\dag t ] + \lambda_5 h^\dag t t^\dag h\ , \end{aligned}\ ] ] where @xmath38 the masses of the scalars are found to be @xmath39 it is clear that @xmath40 can be the lightest @xmath4 odd particle if @xmath41 . since dark matter must not be a charged particle , we will concentrate on @xmath41 afterward .
in the model , since the masses of dark matter ( @xmath26 ) and charged components ( @xmath24 ) are almost degenerate , the coannihilation effects of @xmath42 and @xmath43 should be included in the calculation of the relic abundance of @xmath23 @xcite . with the vertical axis of @xmath44 $ ] , where black , gray , and white regions show the parameter regions larger than , agrees with , and smaller than the wmap constraint , respectively.,width=396 ] with the vertical axis of @xmath44 $ ] , where black , gray , and white regions show the parameter regions larger than , agrees with , and smaller than the wmap constraint , respectively.,width=207 ] in figs .
[ fig : relic_abundance ] and [ fig : relic_abundance2 ] , we show the relic abundance of @xmath23 , where we have used _ _ micromegas2.4__@xcite . for small couplings , i.e. @xmath45 , the dark matter annihilation is governed by the weak interaction .
so the annihilation cross section does not decrease so much . on the other hand , in the large coupling region ( i.e. @xmath46 ) , the main annihilation modes are higgs interactions . since the trilinear coupling of @xmath21 involves only @xmath47 , the cross sections are enhanced by @xmath48 . from the figure , we find that for @xmath49 , the relic abundance agrees with the wmap data in eq .
( [ eq : wmap_relic ] ) . and nucleus particles , where numbers on lines represent the cross sections in @xmath50 unit , while the light gray region is allowed by direct searches of dark matter . , width=226 ]
the si cross section of the y=0 itm is shown in fig .
[ fig : direct_detection ] . from the figure
, we can see that in most of the region , the model escapes the constraint from the direct search .
in this model , the allowed processes are the @xmath23-quark ( u , d ) scatterings at tree level with the small yukawa coupling , and @xmath23-gluon scatterings at loop level . as a result , the si cross section is clearly suppressed . with the vertical axis of @xmath51 $ ] , where the light gray region is allowed by wmap . , width=396 ] [ cols="^,^ " , ] the total relic abundance of @xmath40 and @xmath52 is shown in fig .
[ fig : relic_abundance3 ] . note that the masses of @xmath24 and @xmath53 are automatically fixed if @xmath54 and @xmath55 are known .
since the degeneracy of triplet masses is lifted at tree level , coannihilations of the triplet scalars are not so effective compared to that in the @xmath27 case .
thus , the relic abundance gets enhanced . in the large @xmath56 region , as the mass degeneracy of the triplet components is lifted , the coannihilation effect becomes weaker than that in the small @xmath56 , which enhances the relic abundance .
however , the annihilation cross section becomes large due to the large couplings of @xmath57 and @xmath56 , which suppresses the relic abundance more effective than the coannihilation effect . from fig .
[ fig : relic_abundance100 ] , one can see that in the small mass region , the relic abundance drastically changes due to the resonance effect as well as the opening of new annihilation final states . in the figure ,
we have fixed @xmath58 . as @xmath57 approaches to @xmath56 ,
the figures become similar to those with small @xmath56 since the main interactions enhanced by the large higgs coupling are proportional to @xmath59 . in the region with @xmath60 , where @xmath61 is most effective , the relic abundance is reduced . in the case of @xmath62 ,
the tri - higgs couplings proportional to @xmath59 are canceled to @xmath63 .
thus , the effective couplings of the higgs bosons are very weak and the relic abundance is mainly determined by gauge interactions .
we comment on the direct detection of the y=2 case .
unlike those in y=0 , there are additional scattering processes in the y=2 model , which are the @xmath23-quark scatterings through the gauge coupling of @xmath23 to z. they have larger cross sections due to the gauge coupling .
because of these large cross sections , almost all region is excluded by the direct detection constraint in eq .
( [ eq : dd_constraint ] ) .
note that since the scattering process does not depend on the higgs coupling , this tendency is same even if the ratio of @xmath57 and @xmath56 changes .
we have studied dark matter in the two inert triplet models and we have found that there are allowed regions which agree with the wmap result in tevscale for both models .
explicitly , we have shown that in the y=0 model , the dark matter mass of the neutral scalar around @xmath64 is favored by wmap , which is also allowed by the direct detection , while in the y=2 model , @xmath65 is preferred in terms of the relic abundance , but most of the region is excluded by the direct detection constraint since the @xmath23-quark scattering cross section mediated by z enhanced the si cross section ,
we are grateful to g. b@xmath66langer and a. pukhov for their kind help for _ micromegas_. the work of t.a . was supported in part by the national natural science foundation of china under grant no .
10425522 and no .
c.q.g . and k.i.n were partially supported by the national science council of taiwan under grant no .
nsc-98 - 2112-m-007 - 008-my3 and the national tsing hua university under the boost program no . 97n2309f1 .
99 t. araki , c. q. geng and k. i. nagao , phys .
d * 83 * , 075014 ( 2011 ) [ arxiv:1102.4906 [ hep - ph ] ] .
e. komatsu _ et al .
_ , arxiv:1001.4538 [ astro-ph.co ] . v. silveira and a. zee , phys .
b * 161 * , 136 ( 1985 ) ; c. p. burgess , m. pospelov and t. ter veldhuis , nucl .
b * 619 * , 709 ( 2001 ) ; w. l. guo and y. l. wu , jhep * 1010 * , 083 ( 2010 ) .
e. ma , phys .
d * 73 * , 077301 ( 2006 ) ; r. barbieri , l. j. hall and v. s. rychkov , phys .
d * 74 * , 015007 ( 2006 ) ; l. lopez honorez , e. nezri , j. f. oliver and m. h. g. tytgat , jcap * 0702 * , 028 ( 2007 ) ; e. lundstrom , m. gustafsson and j. edsjo , phys . rev .
d * 79 * , 035013 ( 2009 ) ; p. agrawal , e. m. dolle and c. a. krenke , phys .
d * 79 * , 015015 ( 2009 ) ; s. andreas , m. h. g. tytgat and q. swillens , jcap * 0904 * , 004 ( 2009 ) ; e. nezri , m. h. g. tytgat and g. vertongen , jcap * 0904 * , 014 ( 2009 ) ; e. m. dolle and s. su , phys .
d * 80 * , 055012 ( 2009 ) ; c. arina , f. s. ling and m. h. g. tytgat , jcap * 0910 * , 018 ( 2009 ) ; e. dolle , x. miao , s. su and b. thomas , phys . rev .
d * 81 * , 035003 ( 2010 ) ; x. miao , s. su and b. thomas , phys .
d * 82 * , 035009 ( 2010 ) ; l. lopez - honorez and c. e. yaguna , arxiv:1011.1411 [ hep - ph ] .
m. cirelli and a. strumia , new j. phys .
* 11 * , 105005 ( 2009 ) .
t. hambye , f. s. ling , l. lopez honorez and j. rocher , jhep * 0907 * , 090 ( 2009 ) [ erratum - ibid .
066 ( 2010 ) ] . j. angle _ et al . _ [ xenon collaboration ] , phys .
* 100 * , 021303 ( 2008 ) ; z. ahmed _ et al . _ [ the cdms - ii collaboration ] , science * 327 * , 1619 ( 2010 ) . h. h. zhang , w. b. yan and x. s. li , _ mod .
* 23 * , 637 ( 2008 ) .
k. griest and d. seckel , phys . rev .
d * 43 * , 3191 ( 1991 ) ; s. mizuta and m. yamaguchi , phys . lett .
b * 298 * , 120 ( 1993 ) .
g. belanger , f. boudjema , p. brun , a. pukhov , s. rosier - lees , p. salati and a. semenov , arxiv:1004.1092 [ hep - ph ] .
e. j. chun , jhep * 0912 * , 055 ( 2009 ) . | in this talk , we will review the signatures of dark matter in two inert triplet models .
for the first model with the hypercharge y=0 , the dark matter mass around @xmath0 is favored by both the wmap data and the direct detection .
in contrast , for the second model of y=2 , it is excluded by the direct detection experiments although dark matter with its mass around @xmath1 is allowed by wmap . |
to obtain superconductivity in a layered copper - oxide compound , it is necessary to introduce charge carriers into the antiferromagnetic cuo@xmath0 planes .
recent neutron diffraction studies of the system la@xmath6nd@xmath7sr@xmath8cuo@xmath3 @xcite provide evidence that the dopant - induced holes choose to segregate into periodically - spaced stripes which separate antiferromagnetic domains , in a manner similar to that found in hole - doped la@xmath0nio@xmath3 @xcite . the charge and spin stripe modulations
are identified by the appearance of scattering at incommensurate positions . in the nd - doped system , elastic scattering
is observed , corresponding to static stripes . in pure la@xmath9sr@xmath8cuo@xmath3 ,
the magnetic scattering that is observed is purely inelastic @xcite . where samples with and without nd , but with the same sr concentration , have been measured , the incommensurate ( ic ) splittings of the magnetic signal are found to be essentially identical @xcite .
it has been proposed that the spin correlations in the two systems are fundamentally the same , thus implying similar charge correlations .
the static nature of the stripes in the nd - doped system is attributed @xcite to pinning of the otherwise dynamic correlations by a special distortion of the lattice @xcite . that distortion is driven by purely ionic interactions and
is stabilized by the smaller ionic radius of the substituted nd . to strengthen the case for the charge - stripe interpretation , it is necessary go beyond a simple comparison of incommensurabilities . in this paper
i will show that the neutron scattering measurements of spin fluctuations in superconducting la@xmath1sr@xmath2cuo@xmath3 and yba@xmath0cu@xmath4o@xmath10 are quantitatively consistent with what one should expect based on the stripe picture .
the general expectations are described in section 2 .
specific analyses for the two different superconductors are given in sections 3 and 4 , respectively .
section 5 contains some further discussion .
first recall the situation for the 2d heisenberg model @xcite . at @xmath11 ,
where long - range order exists , the spin - wave dispersion is linear at small @xmath12 ( @xmath13 ) , and can be expressed as @xmath14 . at finite temperatures ,
thermal fluctuations result in a finite spin - spin correlation length @xmath15 .
spin waves are overdamped for @xmath16 , but spin - wave dispersion should still be detectable for @xmath17 ( as long as @xmath18 is less than the superexchange energy @xmath19 ) .
the characteristic energy @xmath20 separating the over- and underdamped regimes is then given by @xmath21 as for the dynamical susceptibility , @xmath22 , @xmath20 acts somewhat like a gap energy @xcite .
if @xmath23 is integrated over @xmath24 to give @xmath25 , then there is a peak at @xmath26 , with @xmath25 going to zero as @xmath27 goes to zero .
next consider an ordered stripe phase .
the charge stripes need not affect the superexchange energy within the antiferromagnetic domains , but they certainly will weaken the magnetic coupling between domains .
castro neto and hone @xcite have considered the effect of the anisotropic coupling on the spin wave velocity . for a large anisotropy ,
the velocity @xmath28 for propagation parallel to the stripes is equal to @xmath29 ; it is more difficult to estimate an effective velocity perpendicular to the stripes because the stripe modulation should result in several optical branches as well as the acoustic mode .
an experimental test of this situation is given by recent measurements of spin waves in stripe - ordered la@xmath0nio@xmath30 @xcite .
that study found @xmath31 for propagation parallel to the stripes .
dispersion was less well defined but comparable in the perpendicular direction .
( note that the spin waves disperse out of the ic points in reciprocal space that characterize the magnetic order . )
thus , it seems reasonable to take @xmath32 as an estimate of the effective spin - wave velocity in an ordered stripe phase . finally , consider a stripe phase with a finite spin - spin correlation length induced by fluctuations of the charge stripes .
the characteristic energy separating over- and underdamped modes should now be @xmath33 excitations at @xmath34 should look spin - wave - like . on the other hand ,
a scan at a fixed excitation energy smaller than @xmath35 through the ic points characterizing the stripe correlations should show a broad peak centered at each ic point , with the width varying inversely as @xmath15 . to obtain an analytic formula describing the @xmath36-dependence of the scattering , consider a model in which the spins have a uniform correlation within each domain , but where the correlations between domains fall off exponentially with distance .
if the charge stripes are separated by @xmath37 lattice spacings , then the intensity along @xmath38 should be proportional to @xmath39 where @xmath40 , and @xmath41 is the structure factor for a single domain .
( the factor of @xmath42 included in @xmath43 takes care of the antiphase relationship between neighboring domains . ) for the calculations below , it will be assumed that the spin correlations also fall off exponentially parallel to the stripes , with the same correlation length @xmath15 .
recently , hayden _ et al . _
@xcite reported inelastic neutron scattering measurements of @xmath22 in la@xmath44sr@xmath45cuo@xmath3 , covering a very large energy range .
they found that @xmath23 at @xmath46 mev could be modelled reasonably well using spin wave formulas , with only a slight reduction in @xmath19 . at lower energies
, they showed that @xmath25 has a peak at @xmath47 mev . if this latter result is combined with the value @xmath48 mev - @xcite , then by inverting eq .
( 2 ) one obtains the estimate @xmath49 . figure [ fg : lsco ] shows an inelastic scan , at a fixed energy transfer of 3 mev , through the ic peaks found in a single crystal of la@xmath1sr@xmath2cuo@xmath3 @xcite .
the peak positions are consistent with a charge stripe period of @xmath50 .
the solid line through the data points is a calculation using eq .
( 3 ) and the estimate @xmath51only the amplitude and the linear background have been adjusted to fit the data .
( the calculated intensity , taking into account signal from all 4 ic peaks , was also convolved with the experimental resolution function . )
the calculation clearly gives an excellent description of the data .
in contrast to the 214 system , measurements of @xmath23 in yba@xmath0cu@xmath4o@xmath10 do not show well resolved ic peaks @xcite . as shown in fig . 2
, the scattered intensity is peaked at @xmath52 , but with a peak shape that has a rather flat top and steep sides . despite the lack of clear evidence for incommensurability , it is possible nevertheless to describe the data with the same stripe model discussed above . to give the stripe picture an honest test , @xmath15
will again be estimated using eq .
the value of @xmath35 is approximately @xmath53 mev @xcite , and @xmath54 mev - @xcite , which together yield @xmath55 . to estimate the stripe spacing , assuming that the hole density within the charge stripes is the same as in the 214 system @xcite , it is first necessary evaluate the hole concentration within a plane .
inverting the formula @xcite @xmath56 yields @xmath57 for the @xmath58-k sample , which corresponds closely to @xmath59 . the solid line through the data in fig .
2 represents a calculation using the same stripe model as above , with the parameter values @xmath59 and @xmath60 .
the calculated function was convolved with the spectrometer resolution function , and only the amplitude and background were adjusted to fit the data . for the calculated curve ,
the filling in of weight between the ic peaks occurs because @xmath15 is less than the stripe spacing @xmath61 . in this case
the weak correlation between neighboring antiphase domains results in little interference . at @xmath52 the scattering from an individual domain does not get cancelled by the weak contributions from neighboring domains .
in contrast , the situation for la@xmath1sr@xmath2cuo@xmath3 is @xmath62 .
that is the necessary for condition for obtaining well resolved ic peaks .
although the correlations between antiphase domains are perhaps less well defined in 123 , the smaller value of @xmath15 suggests greater stripe fluctuations , a condition that appears to correlate with an increased @xmath63 @xcite .
it has been shown that the @xmath24 dependence of scattered intensity from low - energy spin fluctuations in la@xmath1sr@xmath2cuo@xmath3 and yba@xmath0cu@xmath4o@xmath10 can be consistently interpreted on the basis of dynamic antiphase antiferromagnetic domains .
the correlation lengths @xmath15 used to model the scattering measurements were calculated from experimental values of @xmath35 and @xmath64 .
it was pointed out that the ic peaks become unresolvable when @xmath65 , where @xmath61 is the charge - stripe spacing .
there are , of course , other neutron scattering results that are consistent with the stripe picture .
a comparison of @xmath25 in superconducting yba@xmath0cu@xmath4o@xmath10 and antiferromagnetic yba@xmath0cu@xmath4o@xmath66 shows that , despite a difference in @xmath27 dependence , the overall weight within the range 5 to 50 mev is comparable @xcite .
similar results were found in a comparison of la@xmath1sr@xmath2cuo@xmath3 and la@xmath0cuo@xmath3 @xcite , although the spectral weight for the superconductor is somewhat reduced when the comparison is extended up to 300 mev . with a strong segregation of holes of density @xmath67 ,
the density of cu moments contributing to the low - energy ( @xmath68 mev ) spin fluctuations is roughly @xmath69 , so one may expect to see relatively little change in overall spectral weight compared to the undoped antiferromagnet .
measurements on zn - doped la@xmath44sr@xmath45cuo@xmath3 have shown that the positions and @xmath70-widths of the ic peaks are essentially unaffected by the presence of the zn @xcite . similarly , doping zn into yba@xmath0cu@xmath4o@xmath10
does not change the @xmath24 dependence of the scattering , but it does shift spectral weight down to low - energy , so that @xmath25 looks more like that of an antiferromagnet @xcite .
this suggests that doping zn into the cuo@xmath0 planes destroys superconductivity by pinning the stripes .
recently there has been considerable excitement over observations of resonant enhancements of the magnetic scattering below @xmath63 @xcite .
if the enhanced scattering is assumed to come from intinerant electrons , then it is certainly quite remarkable . on the other hand ,
if the magnetic scattering all comes from antiferromagnetic domains between charge stripes , the change in the scattering might in some way be associated with a reduction in damping of the spin fluctuations due to the onset of superconductivity . in any case , it will be interesting to see how models of these effects evolve . | recent neutron diffraction studies have yielded evidence that , in a particular cuprate family , holes doped into the cuo@xmath0 planes segregate into stripes that separate antiferromagnetic domains . here
it is shown that such a picture provides a quantitatively consistent interpretation of the spin fluctuations measured by neutron diffraction in la@xmath1sr@xmath2cuo@xmath3 and yba@xmath0cu@xmath4o@xmath5 . |
a central topic in graph theory is that of giving conditions under which a graph is hamiltonian .
one such result is the classical theorem of dirac @xcite , which states that any graph on @xmath2 vertices with minimum degree at least @xmath3 contains a hamilton cycle . for an analogue in directed graphs
it is natural to consider the _ minimum semi - degree @xmath4 _ of a digraph @xmath0 , which is the minimum of its minimum outdegree @xmath5 and its minimum indegree @xmath6 .
the corresponding result is a theorem of ghouila - houri @xcite , which states that any digraph on @xmath7 vertices with minimum semi - degree at least @xmath3 contains a hamilton cycle .
( when referring to paths and cycles in directed graphs we always mean that these are directed , without mentioning this explicitly . )
both of these results are best possible . in 1979
thomassen @xcite raised the natural corresponding question of determining the minimum semi - degree that forces a hamilton cycle in an _ oriented graph _
( i.e. in a directed graph that can be obtained from a ( simple ) undirected graph by orienting its edges ) . over the years
since the question was posed , a series of improving bounds were obtained in @xcite .
recently , kelly , khn and osthus @xcite were able to obtain an approximate solution .
they proved that an oriented graph on @xmath7 vertices with minimum semi - degree at least @xmath8 has a hamilton cycle , which asymptotically matches a lower bound given by hggkvist @xcite . in this paper
we obtain the following result which exactly matches the lower bound of hggkvist , thus answering thomassen s question for large oriented graphs .
[ main ] there exists a number @xmath9 so that any oriented graph @xmath0 on @xmath10 vertices with minimum semi - degree @xmath11 contains a hamilton cycle .
note that theorem [ main ] implies that every sufficiently large regular tournament on @xmath7 vertices contains at least @xmath12 edge - disjoint hamilton cycles .
( to verify this , note that in a regular tournament , all in- and outdegrees are equal to @xmath13 .
we can then greedily remove hamilton cycles as long as the degrees satisfy the condition in theorem [ main ] . ) this is a slight improvement on the @xmath14 bound obtained in @xcite by the same argument .
it is the best bound so far towards the classical conjecture of kelly ( see e.g. @xcite ) , which states that every regular tournament on @xmath7 vertices can be partitioned into @xmath13 edge - disjoint hamilton cycles .
hggkvist @xcite also made the following conjecture which is closely related to theorem [ main ] .
given an oriented graph @xmath0 , let @xmath15 denote the minimum degree of @xmath0 ( i.e. the minimum number of edges incident to a vertex ) and set @xmath16 .
hggkvist conjectured that if @xmath17 , then @xmath0 has a hamilton cycle .
( note that this conjecture does not quite imply theorem [ main ] as it results in a marginally greater minimum semi - degree condition . ) in @xcite , this conjecture was verified approximately , i.e. if @xmath18 , then @xmath0 has a hamilton cycle .
it seems possible that our approach can be extended to obtain an exact solution to this problem , but this would certainly require some additional ideas beyond those applied here .
our argument can be extended to find a cycle of any length @xmath19 , with @xmath20 ( say ) through any given vertex .
we indicate the necessary modifications to the proof in the final paragraph , the details can be found in @xcite .
this result is used in @xcite to obtain the following pancyclicity result : any sufficiently large oriented graph @xmath0 with @xmath21 contains a cycle of length @xmath19 for all @xmath22 . for @xmath23 , it even contains a cycle of length @xmath19 through a given vertex .
the rest of this paper is organized as follows .
the next section contains some basic notation . in section 3
we describe the extremal example showing that theorem [ main ] is best possible .
we set up our main tools in section 4 , these being a digraph form of the regularity lemma due to alon and shapira @xcite and the blow - up lemma , both in the original form of komls , srkzy and szemerdi @xcite and in a recent stronger and more technical form due to csaba @xcite .
our argument uses the stability method , and so falls naturally into two cases , according to whether or not our given oriented graph @xmath0 is structurally similar to the extremal example described in section 3 . in section 5
we prove a lemma that enables us to find a hamilton cycle when @xmath0 is not structurally similar to the extremal example .
the argument in this case is based on that of @xcite .
then we prove our main theorem in the final section .
given two vertices @xmath24 and @xmath25 of a directed graph @xmath0 , we write @xmath26 for the edge directed from @xmath24 to @xmath25 .
the order @xmath27 of @xmath0 is the number of its vertices .
we write @xmath28 for the outneighbourhood of a vertex @xmath24 and @xmath29 for its outdegree .
similarly , we write @xmath30 for the inneighbourhood of @xmath24 and @xmath31 for its indegree .
we write @xmath32 for the neighbourhood of @xmath24 and @xmath33 for its degree .
we use @xmath34 etc . whenever this is unambiguous .
we write @xmath35 for the maximum of @xmath36 over all vertices @xmath37 .
given a set @xmath38 of vertices of @xmath0 , we write @xmath39 for the set of all outneighbours of vertices in @xmath38 .
so @xmath39 is the union of @xmath40 over all @xmath41 .
@xmath42 is defined similarly .
the directed subgraph of @xmath0 induced by @xmath38 is denoted by @xmath43 $ ] and we write @xmath44 for the set of its edges and put @xmath45 .
if @xmath46 is a subset of the vertex set of @xmath0 then @xmath47 denotes the digraph obtained from @xmath0 by deleting @xmath46 and all edges incident to @xmath46 .
recall that when referring to paths and cycles in directed graphs we always mean that they are directed without mentioning this explicitly . given two vertices @xmath24 and @xmath25 on a directed cycle @xmath48
, we write @xmath49 for the subpath of @xmath48 from @xmath24 to @xmath25 . similarly , given two vertices @xmath24 and @xmath25 on a directed path @xmath50 such that @xmath24 precedes @xmath25
, we write @xmath51 for the subpath of @xmath50 from @xmath24 to @xmath25 . a _ walk _ in a directed graph
@xmath0 is a sequence of ( not necessarily distinct ) vertices @xmath52 where @xmath53 is an edge for all @xmath54 .
the walk is _ closed _ if @xmath55 .
a _ @xmath56-factor _ of @xmath0 is a collection of disjoint cycles which cover all the vertices of @xmath0 .
we define things similarly for undirected graphs . given two vertices @xmath57 of @xmath0 , an _
@xmath24-@xmath25 path _ is a directed path which joins @xmath24 to @xmath25 .
given two disjoint subsets @xmath38 and @xmath58 of vertices of @xmath0 , an @xmath38-@xmath58 edge is an edge @xmath59 where @xmath41 and @xmath60 .
we write @xmath61 for the set of all these edges and put @xmath62 .
we sometimes also write @xmath63 for @xmath64 .
we denote by @xmath65 the oriented bipartite subgraph of @xmath0 whose vertex classes are @xmath38 and @xmath58 and whose edge set is @xmath61 .
we call an orientation of a complete graph a _ tournament _ and an orientation of a complete bipartite graph a _ bipartite tournament_. throughout the paper we omit floors and ceilings whenever this does not affect the argument . for a positive integer @xmath66 we write @xmath67:=\{1,\dots , k\}$ ] .
in this section , we show that the bound in theorem [ main ] is best possible for every @xmath7 . the construction is due to hggkvist @xcite .
however , he only gives a description for the case when @xmath7 is of the form @xmath68 .
below , we also add the remaining cases to give the complete picture .
[ extremal_example ] for any @xmath2 there is an oriented graph @xmath0 on @xmath7 vertices with minimum semi - degree @xmath69 which does not contain a @xmath56-factor , and so does not contain a hamilton cycle .
we construct @xmath0 as follows .
it has @xmath7 vertices partitioned into @xmath70 parts @xmath71 , with @xmath72 .
each of @xmath38 and @xmath48 spans a tournament , @xmath58 and @xmath73 are joined by a bipartite tournament , and we add all edges from @xmath38 to @xmath58 , from @xmath58 to @xmath48 , from @xmath48 to @xmath73 and from @xmath73 to @xmath38 ( see figure 1 ) . since every path which joins two vertices in @xmath58 has to pass through @xmath73 , it follows that every cycle contains at least as many vertices from @xmath73 as it contains from @xmath58 .
as @xmath72 this means that one can not cover all the vertices of @xmath0 by disjoint cycles , i.e. @xmath0 does not contain a 1-factor .
[ fig : extremal ] [ ] [ ] @xmath38 [ ] [ ] @xmath58 [ ] [ ] @xmath48 [ ] [ ] @xmath73 .,title="fig : " ] it remains to show that the sizes of @xmath71 and the tournaments can be chosen to give minimum semi - degree @xmath74 .
the following table gives possible values , according to the value of @xmath7 mod @xmath75 : @xmath76 we choose the tournaments inside @xmath38 and @xmath48 to be as regular as possible ( meaning that the indegree and outdegree of every vertex differ by at most @xmath56 ) : this can be achieved ( e.g. ) by arranging the vertices evenly around a circle and directing edges so that the larger part of the circle lies to the right of each edge , breaking ties arbitrarily .
for the bipartite tournament between @xmath58 and @xmath73 there are two cases .
firstly , when @xmath77 and @xmath78 we choose it so that @xmath79 for every @xmath80 and @xmath81 ( in either order ) for every @xmath82 : this can be achieved by the blow - up of a @xmath70-cycle , i.e. , write @xmath83 , @xmath84 with @xmath85 , @xmath86 and direct edges from @xmath87 to @xmath88 , from @xmath88 to @xmath89 , from @xmath89 to @xmath90 and from @xmath90 to @xmath87 . secondly , when @xmath91 and @xmath92 we choose the bipartite tournament between @xmath58 and @xmath73 so that @xmath93 and @xmath94 for every @xmath80 and so that for every @xmath95 either @xmath96 or @xmath97 and @xmath98 .
this can be achieved by labelling @xmath58 as @xmath99 $ ] , @xmath73 as @xmath100 and setting @xmath101 for @xmath80 .
indeed , to check that the vertices in @xmath73 have the correct indegrees note that @xmath102 is a segment of consecutive vertices of @xmath73 for each @xmath60 and the segment @xmath103 starts immediately after @xmath102 and ends with the first vertex of @xmath102 . to verify the minimum semi - degree condition we introduce the notation @xmath104 for the minimum in- and outdegrees of vertices in @xmath105 .
we can eliminate some of the checking by noting that we have always chosen @xmath106 and @xmath107 , so we have inequalities @xmath108 and @xmath109 as well as @xmath110 . in the cases
@xmath111 we also have @xmath112 and @xmath113 , so @xmath114 in the case @xmath115 we have @xmath116 , @xmath117 , @xmath118 and @xmath119 , so @xmath120 . the constructions for the cases @xmath121 are obtained from @xmath115 by increasing @xmath122 and/or @xmath123 by @xmath56 , and we still have @xmath124 . when @xmath125 we have @xmath126 , @xmath127 , @xmath128 and @xmath129 , so @xmath130
. finally , the construction for the case @xmath131 is obtained from @xmath125 by increasing @xmath122 by @xmath56 and we still have @xmath132 . in all cases we have @xmath133 , as required
. * remark .
* one may add any number of edges that either go from @xmath38 to @xmath48 or lie within @xmath73 without creating a @xmath56-factor ( although this does not increase the minimum semi - degree ) .
in this section we collect all the information we need about the diregularity lemma and the blow - up lemma .
see @xcite for a survey on the regularity lemma and @xcite for a survey on the blow - up lemma .
we start with some more notation .
the density of a bipartite graph @xmath134 with vertex classes @xmath38 and @xmath58 is defined to be @xmath135 we often write @xmath136
if this is unambiguous . given @xmath137 , we say that @xmath0 is _ @xmath138-regular _ if for all subsets @xmath139 and @xmath140 with @xmath141 and @xmath142 we have that @xmath143 .
given @xmath144 $ ] we say that @xmath0 is @xmath145-_super - regular _ if it is @xmath138-regular and furthermore @xmath146 for all @xmath41 and @xmath147 for all @xmath60 .
( this is a slight variation of the standard definition of @xmath145-super - regularity where one requires @xmath148 and @xmath149 . ) given partitions @xmath150 and @xmath151 of the vertex set of some graph , we say that @xmath150 _ refines _
@xmath151 if for all @xmath152 with @xmath153 there is some @xmath154 with @xmath155 which contains @xmath152 .
note that @xmath156 need not be contained in any @xmath154 , so this is weaker than the usual notion of refinement of partitions .
the diregularity lemma is a version of the regularity lemma for digraphs due to alon and shapira @xcite .
its proof is quite similar to the undirected version .
we will use the degree form of the diregularity lemma which can be easily derived ( see e.g. @xcite for an explicit proof of all but the ` refinement property ' ) from the standard version , in exactly the same manner as the undirected degree form .
[ direg ] for every @xmath157 and all numbers @xmath158 , @xmath159 there are numbers @xmath160 and @xmath9 such that if * @xmath0 is a digraph on @xmath161 vertices , * @xmath162 is a partition of the vertices of @xmath0 , * @xmath163 $ ] is any real number , then there is a partition of the vertices of @xmath0 into @xmath150 and a spanning subdigraph @xmath164 of @xmath0 such that the following holds : * @xmath165 , * @xmath166 , * @xmath167 , * @xmath150 refines the partition @xmath168 , * @xmath169 for all vertices @xmath37 , * @xmath170 for all vertices @xmath37 , * for all @xmath171 the digraph @xmath172 $ ] is empty , * for all @xmath173 with @xmath174 the bipartite graph whose vertex classes are @xmath152 and @xmath175 and whose edges are all the @xmath152-@xmath175 edges in @xmath164 is @xmath138-regular and has density either @xmath176 or density at least @xmath177 .
@xmath178 are called _
, @xmath156 is called the _
exceptional set _ and the vertices in @xmath156 are called _
exceptional vertices_. the last condition of the lemma says that all pairs of clusters are @xmath138-regular in both directions ( but possibly with different densities ) .
we call the spanning digraph @xmath179 given by the diregularity lemma the _
pure digraph_. given clusters @xmath178 and a digraph @xmath164 , the _ reduced digraph @xmath180 with parameters @xmath145 _ is the digraph whose vertex set is @xmath67 $ ] and in which @xmath181 is an edge if and only if the bipartite graph whose vertex classes are @xmath152 and @xmath175 and whose edges are all the @xmath152-@xmath175 edges in @xmath164 is @xmath138-regular and has density at least @xmath177 .
( so if @xmath164 is the pure digraph , then @xmath181 is an edge in @xmath180 if and only if there is a @xmath152-@xmath175 edge in @xmath164 . )
it is easy to see that the reduced digraph @xmath180 obtained from the regularity lemma ` inherits ' the minimum degree of @xmath0 , in that @xmath182 and @xmath183 .
however , @xmath180 is not necessarily oriented even if the original digraph @xmath0 is .
the next lemma ( which is essentially from @xcite ) shows that by discarding edges with appropriate probabilities one can go over to a reduced oriented graph @xmath184 which still inherits the minimum degree and density of @xmath0 .
[ red - orient ] for every @xmath157 and there exist numbers @xmath185 and @xmath186 such that the following holds .
let @xmath144 $ ] with @xmath187 , let @xmath0 be an oriented graph of order @xmath10 and let @xmath180 be the reduced digraph with parameters @xmath145 obtained by applying the diregularity lemma to @xmath0 with @xmath158 as the lower bound on the number of clusters .
then @xmath180 has a spanning oriented subgraph @xmath188 such that * @xmath189 , * @xmath190 , * for all disjoint sets @xmath191 with @xmath192 we have @xmath193 , where @xmath194 and @xmath195 .
* for every set @xmath196 with @xmath197 we have @xmath198 , where @xmath194 . statements ( a ) and ( b ) are proven in @xcite by considering the following probabilistic argument .
let @xmath188 be the spanning oriented subgraph obtained from @xmath180 by deleting edges randomly as follows .
for every unordered pair @xmath199 of clusters we delete the edge @xmath181 ( if it lies in @xmath180 ) with probability @xmath200 otherwise we delete @xmath201 ( if it lies in @xmath180 ) .
we interpret ( [ eq : probdelete ] ) as 0 if @xmath202 .
so if @xmath180 contains at most one of the edges @xmath203 then we do nothing .
we do this for all unordered pairs of clusters independently . in
@xcite it was shown that @xmath188 satisfies ( a ) and ( b ) with probability at least @xmath204 .
for ( c ) , note that @xmath205 a chernoff - type bound ( see e.g. ( * ? ? ? * cor .
a.14 ) ) now implies that there exists an absolute constant @xmath206 such that @xmath207 but the number of pairs @xmath208 of disjoint subsets of @xmath209 is at most @xmath210 , so the probability that @xmath188 does not satisfy ( c ) is at most @xmath211 .
( here we used that @xmath212 is sufficiently large . )
the proof for ( d ) is similar .
altogether , this implies that there must be some outcome which satisfies ( a)(d ) .
the oriented graph @xmath188 given by lemma [ red - orient ] is called the _ reduced oriented graph with parameters @xmath145_. the spanning oriented subgraph @xmath213 of the pure digraph @xmath164 obtained by deleting all the @xmath152-@xmath175 edges whenever @xmath214 is called the _ pure oriented graph_. given an oriented subgraph @xmath215 , the _ oriented subgraph of @xmath213 corresponding to @xmath46 _ is the oriented subgraph obtained from @xmath213 by deleting all those vertices that lie in clusters not belonging to @xmath46 as well as deleting all the @xmath152-@xmath175 edges for all pairs @xmath199 with @xmath216 . in our proof of theorem [ main ]
we will also need the blow - up lemma of komls , srkzy and szemerdi @xcite .
it implies that dense regular pairs behave like complete bipartite graphs with respect to containing bounded degree graphs as subgraphs .
[ standardblowup ] given a graph @xmath217 on @xmath67 $ ] and positive numbers @xmath218 , there is a positive real @xmath219 such that the following holds for all positive numbers @xmath220 and all @xmath221 .
let @xmath222 be the graph obtained from @xmath217 by replacing each vertex @xmath223 with a set @xmath152 of @xmath224 new vertices and joining all vertices in @xmath152 to all vertices in @xmath175 whenever @xmath181 is an edge of @xmath217 .
let @xmath164 be a spanning subgraph of @xmath222 such that for every edge @xmath225 the graph @xmath226 is @xmath227-super - regular .
then @xmath164 contains a copy of every subgraph @xmath228 of @xmath222 with @xmath229 .
moreover , this copy of @xmath228 in @xmath164 maps the vertices of @xmath228 to the same sets @xmath152 as the copy of @xmath228 in @xmath222 ,
i.e. if @xmath230 is mapped to @xmath152 by the copy of @xmath228 in @xmath222 , then it is also mapped to @xmath152 by the copy of @xmath228 in @xmath164 .
the ` moreover ' part of the statement is not part of the usual statement of the blow - up lemma but is stated explicitly in its proof .
we will also need to use the following stronger and more technical version due to csaba @xcite .
( the case when @xmath231 of this is implicit in @xcite . )
it removes the dependency of the regularity constant @xmath232 on the number @xmath66 of clusters .
also , it does not demand super - regularity .
the latter is replaced by conditions ( c6),(c8 ) and ( c9 ) below .
moreover , it is explicitly formulated to allow for a set @xmath156 of exceptional vertices in the host graph @xmath164 which are not part of any regular pairs . for this to work ,
one first has to find a suitable partition @xmath233 with @xmath234 of the graph @xmath228 which we are aiming to embed and a suitable bijection @xmath235 .
the embedding of @xmath228 into @xmath164 guaranteed by the blow - up lemma is then found by a randomized algorithm which first embeds each vertex @xmath236 to @xmath237 and then successively embeds the remaining vertices of @xmath228 .
condition ( c1 ) requires that there are not too many exceptional vertices and ( c2 ) ensures that we can embed the vertices in @xmath238 without affecting the neighbourhood of other such vertices .
@xmath239 will be embedded into the cluster @xmath152 of @xmath164 , so we clearly need to assume ( c3 ) .
condition ( c5 ) gives us a reasonably large set @xmath73 of ` buffer vertices ' which will be embedded last by the randomized algorithm .
( c7 ) ensures that the exceptional vertices have large degree in all ` neighbouring clusters ' .
( c8 ) and ( c9 ) allow us to embed those vertices whose set of candidate images in @xmath164 has grown very small at some point of the algorithm . in the statement of lemma [ blowup ] and
later on we write @xmath240 to mean that we can choose the constants @xmath241 from right to left .
more precisely , there are increasing functions @xmath242 and @xmath243 such that , given @xmath244 , whenever we choose some @xmath245 and @xmath246 , all calculations needed in the proof of lemma [ blowup ] are valid .
hierarchies with more constants are defined in the obvious way .
[ blowup ] for all numbers @xmath247 and every positive constant @xmath206 there exists an number @xmath248 such that whenever @xmath249 are positive constants with @xmath250 the following holds .
suppose that @xmath164 is a graph of order @xmath251 and @xmath252 is a partition of @xmath253 such that the bipartite graph @xmath226 is @xmath138-regular with density either @xmath176 or @xmath177 for all @xmath254 .
let @xmath228 be a graph on @xmath7 vertices with @xmath229 and let @xmath255 be a partition of @xmath256 with @xmath257 for every @xmath258 . furthermore ,
suppose that there exists a bijection @xmath259 and a set @xmath260 of vertices at distance at least @xmath70 from each other such that the following conditions hold : * @xmath261 .
* @xmath262 .
* @xmath239 is independent for every @xmath258 .
* @xmath263 for every @xmath258 .
* for each @xmath258 there exists @xmath264 with @xmath265 and such that for @xmath266 and all @xmath267 @xmath268 * if @xmath269 and @xmath270 then @xmath226 is @xmath138-regular with density @xmath177 . * if @xmath269 and @xmath271 then @xmath272 . * for each @xmath258 , given any @xmath273 with @xmath274 there exists a set @xmath275 and a bijection @xmath276 such that @xmath277 whenever @xmath278 ( for all @xmath279 and all @xmath280 ) . *
writing @xmath281 we have that @xmath282 .
then @xmath164 contains a copy of @xmath228 such that the image of @xmath239 is @xmath152 for all @xmath171 and the image of each @xmath236 is @xmath283 .
the additional properties of the copy of @xmath228 in @xmath164 are not included in the statement of the lemma in @xcite but are stated explicitly in the proof .
we aim to apply the blow - up lemma with @xmath164 being obtained from the underlying graph of the pure oriented graph . in order to satisfy ( c8 ) , it will suffice to ensure that all the edges of a suitable 1-factor in the reduced oriented graph @xmath188 correspond to @xmath145-super - regular pairs of clusters .
a simple calculation implies that this can be ensured by removing a small proportion of vertices from each cluster @xmath152 , and so ( c8 ) will be satisfied .
however , ( c6 ) requires all the edges of @xmath188 to correspond to @xmath138-regular pairs of density precisely @xmath177 and not just at least @xmath177 .
( although , as remarked by csaba @xcite , it actually suffices that the densities are close to @xmath177 in terms of @xmath138 . )
this is guaranteed by the second part of the following proposition from @xcite .
[ superreg ] let @xmath284 be positive numbers and let @xmath285 be positive reals such that @xmath286 .
let @xmath0 be an oriented graph of order at least @xmath9 .
let @xmath188 be the reduced oriented graph with parameters @xmath145 and let @xmath213 be the pure oriented graph obtained by successively applying first the diregularity lemma with @xmath138 , @xmath177 and @xmath158 to @xmath0 and then lemma [ red - orient ] .
let @xmath46 be an oriented subgraph of @xmath188 with @xmath287 .
let @xmath164 be the underlying graph of @xmath213
. then one can delete @xmath288 vertices from each cluster @xmath152 to obtain subclusters @xmath289 in such a way that @xmath164 contains a subgraph @xmath290 whose vertex set is the union of all the @xmath291 and such that * @xmath292 is @xmath293-super - regular whenever @xmath294 , * @xmath292 is @xmath295-regular and has density @xmath296 whenever @xmath297 .
this section covers the main part of the argument for the case when the original oriented graph @xmath0 is not close to being the ` extremal graph ' described in section [ extremal ] .
the following well - known fact will be very useful .
[ 1-factor ] suppose that @xmath0 is a digraph such that @xmath298 for every @xmath299 .
then @xmath0 has a @xmath56-factor .
the result follows immediately by applying hall s matching theorem to the following bipartite graph @xmath228 : the vertex classes @xmath300 of @xmath228 are both copies of the vertex set of the original digraph @xmath0 and we connect a vertex @xmath301 to @xmath80 in @xmath228 if there is a directed edge from @xmath302 to @xmath303 in @xmath0 .
a perfect matching in @xmath228 corresponds to a 1-factor in @xmath0 .
[ find - cycle ] suppose @xmath304 , and that @xmath188 is an oriented graph on @xmath67 $ ] such that * @xmath305 , * @xmath306 for any @xmath307 $ ] with @xmath308 , and * @xmath309 is a @xmath56-factor in @xmath188 . also , suppose that @xmath310 is an oriented graph obtained from @xmath188 by adding a set @xmath311 of at most @xmath312 vertices and some edges so that every @xmath313 has both an inneighbour and an outneighbour in @xmath67 $ ] .
suppose @xmath213 is an oriented graph on @xmath7 vertices with vertex partition @xmath314 , such that * each @xmath315 is independent ; * if @xmath181 is an edge on some cycle from @xmath309 , then the bipartite graph @xmath316 consisting of all the @xmath315-@xmath154 edges in @xmath213 forms an @xmath145-super - regular pair , while if @xmath317 , the bipartite graph @xmath316 forms an @xmath138-regular pair of density @xmath177 in @xmath213 , and finally @xmath316 is empty whenever @xmath318 ; * for all @xmath313 and @xmath319 $ ] , we have @xmath320 whenever @xmath321 and @xmath322 whenever @xmath323 ; * @xmath324 .
then @xmath213 has a hamilton cycle . throughout the proof of the lemma , given a cycle @xmath325 and a vertex @xmath326
, we will denote the predecessor of @xmath24 on @xmath48 by @xmath327 and its successor by @xmath328 .
we extend this convention to sets of vertices in the natural way .
we say that a walk @xmath329 in @xmath310 is _ balanced _ with respect to @xmath309 if for each cycle @xmath325 every vertex on @xmath48 is visited the same number of times by @xmath329
. we will first construct a closed walk @xmath329 in @xmath310 such that * @xmath329 is balanced with respect to @xmath309 ; * @xmath329 visits every vertex in @xmath67 $ ] at least once and at most @xmath330 times ; * @xmath329 visits every vertex in @xmath311 exactly once ; * any two vertices in @xmath311 are at distance at least @xmath70 along @xmath329 . to achieve the balance property ( a )
, we will build up @xmath329 from certain special walks ( which are themselves balanced with respect to @xmath309 ) : a _ @xmath309-shifted walk joining @xmath24 to @xmath25 and traversing @xmath331 cycles from @xmath332 _ is a walk of the form @xmath333 , where @xmath334 , @xmath335 and the @xmath336 are ( not necessarily distinct ) cycles from @xmath309 .
( note that this definition is slightly different from the corresponding one in @xcite . )
we claim that for any two vertices @xmath337 $ ] there is a @xmath309-shifted walk in @xmath188 joining @xmath24 to @xmath25 which traverses at most @xmath338 cycles from @xmath332 . to see this ,
let @xmath339 $ ] be the set of vertices that can be reached from @xmath24 by a @xmath332-shifted walk traversing at most @xmath340 cycles from @xmath332 .
then @xmath341 and @xmath342 for @xmath343 .
by assumption on @xmath188 , if @xmath344 then we have @xmath345 .
so if @xmath346 we have @xmath347 .
the minimum semi - degree condition on @xmath188 implies that @xmath348 .
so there is a vertex @xmath349 and we can reach @xmath25 by following a @xmath332-shifted walk to @xmath350 which traverses at most @xmath340 cycles from @xmath332 and then adding @xmath351 ( where @xmath48 is the cycle from @xmath352 containing @xmath25 . ) for each pair @xmath57 let @xmath353 denote such a @xmath332-shifted walk . to construct @xmath329 ,
we start at vertex @xmath354 $ ] and follow the walks @xmath355 .
note that altogether these walks yield a single ( balanced ) @xmath309-shifted walk joining @xmath56 to @xmath356 which visits every vertex of @xmath188 at least once .
next we aim to extend this walk to incorporate the vertices in @xmath311 . for this
, suppose that @xmath357 , say .
for each @xmath358 $ ] , we now select vertices @xmath359 , @xmath360 and follow the walks @xmath361 , @xmath362 , @xmath363 , @xmath364 , @xmath365 , @xmath366 , @xmath367 .
finally we close the walk by following @xmath368 and the edge @xmath369 .
note that the resulting closed walk @xmath329 is balanced with respect to @xmath332 since this holds for each of the walks @xmath370 .
moreover , @xmath329 traverses at least one cycle of @xmath309 between each visit to @xmath311 ( by definition ) , so any two vertices in @xmath311 are at distance at least @xmath70 along @xmath329 .
every vertex in @xmath311 is visited exactly once , and vertices in @xmath67 $ ] are visited at most @xmath371 times , so @xmath329 has the required properties ( a)(d ) .
we will use the blow - up lemma ( lemma [ blowup ] ) to ` transform ' @xmath329 into a hamilton cycle of @xmath213 . as a preliminary step
, it will be useful to consider an auxiliary ( undirected ) graph @xmath372 which it is obtained from @xmath213 by replacing every @xmath138-regular pair @xmath316 of nonzero density ( i.e. those which correspond to edges of @xmath188 ) by a complete bipartite graph .
the neighbourhoods of the vertices in @xmath311 are not affected .
we can find a hamilton cycle @xmath373 in @xmath372 as follows .
first we find a one - to - one mapping of @xmath329 to a cycle @xmath374 in @xmath372 such that each vertex of @xmath311 is mapped to itself and for each @xmath319 $ ] visits of @xmath329 to a vertex @xmath375 are mapped to distinct vertices in @xmath315 ; to do this we first greedily select distinct images for @xmath376 and @xmath377 for all @xmath358 $ ] , which is possible by assumption ( iii ) and the inequality @xmath378 .
we can then complete the rest of the mapping arbitrarily since @xmath329 visits every vertex @xmath375 at most @xmath379 times .
next we will extend @xmath374 to a hamilton cycle .
for each cycle @xmath380 , let @xmath381 denote the number of times that @xmath329 visits a vertex @xmath382 ( note that this number is the same for all vertices @xmath382 ) .
fix one particular occasion on which @xmath329 ` winds around ' @xmath48 and replace the corresponding path in @xmath374 by a path @xmath383 with the same endpoints that winds around the ` blow - up ' @xmath384 of @xmath48 in @xmath372 exactly @xmath385 times , so that it exhausts all the vertices in @xmath384 . since @xmath352 is a @xmath56-factor the resulting cycle @xmath373 uses all vertices in @xmath372 ,
i.e. it is hamiltonian . for future reference
we note that the number of times that @xmath383 winds around @xmath384 is @xmath386 .
( here we used property ( b ) and @xmath387 . )
now we use the version of the blow - up lemma by csaba ( lemma [ blowup ] ) to show that @xmath373 corresponds to a hamilton cycle in @xmath213 .
we fix additional constants @xmath388 with @xmath389 .
we will apply the blow - up lemma with @xmath228 being the underlying graph of @xmath373 and @xmath164 being obtained from the underlying graph of @xmath213 as follows : if @xmath321 with @xmath313 and @xmath319 $ ] , then delete all those edges which correspond to edges oriented from @xmath315 to @xmath24 ; similarly , if @xmath323 with @xmath313 and @xmath319 $ ] , then delete all those edges which correspond to edges oriented from @xmath24 to @xmath315 .
this deletion ensures that when we embed @xmath373 as an undirected cycle in @xmath164 the directions of the available edges will in fact make it a directed cycle in @xmath213 . for all @xmath390
, @xmath315 will play the role of @xmath152 in the blow - up lemma and we take @xmath391 to be the partition of @xmath228 induced by @xmath392 .
@xmath393 will be the obvious bijection ( i.e. the identity ) . to define the set @xmath260 of vertices of distance at least @xmath70 from each other which is used in the blow - up lemma , for each @xmath394 let @xmath395 be the subpath of @xmath228 corresponding to the path @xmath383 ( defined in the construction of @xmath373 ) .
note that @xmath396 . for each @xmath171 ,
let @xmath397 denote the cycle containing @xmath340 and let @xmath398 consist of all those vertices in @xmath399 which have distance at least 4 from the endvertices of @xmath400 .
thus in the graph @xmath228 each vertex @xmath401 has one of its neighbours in the set @xmath402 corresponding to the predecessor of @xmath340 on @xmath336 and its other neighbour in the set @xmath403 corresponding to the successor of @xmath340 on @xmath336 .
moreover , all the vertices in @xmath404 have distance at least 4 from all the vertices in @xmath238 and the lower bound on @xmath405 implies that @xmath406 .
it is easy to see that one can greedily choose a set @xmath407 of size @xmath408 such that the vertices in @xmath409 have distance at least 4 from each other .
let @xmath410 .
let us now check conditions ( c1)(c9 ) .
( c1 ) holds with @xmath411 since @xmath412 .
( c2 ) holds by definition of @xmath413 .
( c3 ) holds since @xmath228 is a hamilton cycle in @xmath372 and @xmath372 inherits the independence property ( i ) in the statement of the lemma from @xmath213 .
this also implies that for every edge @xmath414 with @xmath270 ( @xmath415 ) we must have that @xmath297 ( provided that the direction of @xmath26 was chosen such that @xmath416 ) . thus ( c6 ) holds as every edge of @xmath188 corresponds to an @xmath138-regular pair of clusters having density @xmath177 .
( c4 ) holds with @xmath417 because @xmath418 ( the final inequality follows from the fact that @xmath419 . ) for ( c5 ) we need to find a set @xmath420 of buffer vertices .
pick any set @xmath421 with @xmath422 and let @xmath266 .
since @xmath407 we have that @xmath423 for all @xmath280 .
hence @xmath424 for all @xmath267 and so ( c5 ) holds .
( c7 ) holds by assumption ( iii ) on @xmath310 and the fact that our walk @xmath329 used only edges of @xmath310 .
( c8 ) and ( c9 ) are now the only conditions we need to check .
given a set @xmath425 of size at most @xmath426 , we wish to find @xmath427 and a bijection @xmath276 such that every @xmath279 has a large number of neighbours in every cluster @xmath175 for which @xmath428 contains a neighbour of @xmath429 .
pick any set @xmath430 of size @xmath431 .
( this can be done since @xmath432 and so @xmath433 . )
let @xmath276 be an arbitrary bijection .
to see that ( c8 ) holds with these choices , consider any vertex @xmath434 and let @xmath435 be such that @xmath428 contains a neighbour of @xmath429 in @xmath228 .
since @xmath436 , this means that @xmath435 must be a neighbour of @xmath340 on the cycle @xmath397 containing @xmath340 . but
this implies that @xmath437 since each edge of the union @xmath438 of all the cycles from @xmath332 corresponds to an @xmath439-super - regular pair in @xmath164 .
finally , writing @xmath440 we have @xmath441 ( since @xmath442 for each @xmath280 ) and so ( c9 ) is satisfied with @xmath443 .
hence ( c1)(c9 ) hold and so we can apply the blow - up lemma to obtain a hamilton cycle in @xmath164 such that the image of @xmath239 is @xmath152 for all @xmath171 and the image of each @xmath236 is @xmath444 . by construction of @xmath164
this undirected hamilton cycle corresponds to a directed hamilton cycle in @xmath213 .
define @xmath445 and additional constants so that @xmath446 suppose @xmath0 is an oriented graph on @xmath10 vertices with minimum semi - degree @xmath447 and no hamilton cycle .
apply the diregularity lemma ( lemma [ direg ] ) to @xmath0 with parameters @xmath448 to obtain a partition @xmath449 of @xmath450 with @xmath451 .
let @xmath188 be the reduced oriented graph with parameters @xmath452 given by applying lemma [ red - orient ] and let @xmath213 be the pure oriented graph .
lemma [ red - orient](a ) and ( b ) implies @xmath453 note that if @xmath457 then @xmath458 ( say ) , and if @xmath459 then @xmath460 for any @xmath319 $ ] , i.e. @xmath461 , and @xmath462 $ ] .
so @xmath463 for every @xmath307 $ ] , and proposition [ 1-factor ] implies that there is a @xmath56-factor @xmath309 in @xmath188 .
in fact , @xmath464 .
we would now like to apply lemma [ find - cycle ] to @xmath188 , @xmath309 and @xmath213 to find a hamilton cycle .
however , the size of the exceptional set @xmath156 may be larger than the number of exceptional vertices allowed for by the lemma . to deal with this problem , we first partition @xmath465 into sets @xmath38 and @xmath58
. then we further refine the above ` regularity partition ' within @xmath466 to obtain a very small ` exceptional set ' @xmath467 .
it then turns out that we can use lemma [ find - cycle ] to find a hamilton path in @xmath468 and a hamilton cycle in @xmath469 which together form a hamilton cycle in @xmath213 . for each cluster @xmath152 consider a random partition of @xmath152 into two sets @xmath470 and @xmath471 obtained by assigning a vertex @xmath472 to @xmath470 with probability @xmath473 independently of all other vertices in @xmath152 .
so @xmath474 and @xmath475 .
then the probability that @xmath476 is at least @xmath477 ( see e.g. @xcite ) .
also , standard chernoff bounds imply that the probability that there is a vertex @xmath37 whose degree in @xmath470 is too large or too small is exponentially small in @xmath478 .
so with non - zero probability we obtain a partition as desired . to see the third property , note that the definition of regularity implies that the pair @xmath479 consisting of all the @xmath470-@xmath480 edges in @xmath213 is @xmath481-regular and has density at least @xmath482 whenever @xmath297 . on the other hand ,
@xmath479 is empty whenever @xmath318 since @xmath483 is empty in this case .
apply proposition [ superreg ] , with @xmath484 $ ] playing the role of @xmath213 , @xmath309 playing the role of @xmath46 ( so @xmath485 ) , the @xmath471 playing the role of the @xmath152 and @xmath285 replaced by @xmath486 .
we obtain an ( oriented ) subgraph @xmath487 of @xmath484 $ ] and sets @xmath488 of size @xmath489 so that a pair @xmath490 is @xmath138-regular of density @xmath491 whenever @xmath317 , and so that @xmath490 is @xmath492-super - regular whenever @xmath181 is an edge of @xmath332 .
let @xmath493 .
so @xmath494 .
let @xmath495 be obtained from @xmath156 by adding the vertices in @xmath496 for all @xmath497 $ ] . since @xmath498 and @xmath499 we have @xmath500 consider any cluster @xmath470 .
claim 1.1 implies that for all @xmath504 at most @xmath505 vertices in @xmath470 have less than @xmath506 outneighbours in @xmath480 .
call these vertices _
useless for @xmath435_. so on average any vertex of @xmath470 is useless for at most @xmath507 indices @xmath504 .
thus at most @xmath508 vertices in @xmath470 are useless for more than @xmath509 indices @xmath504 .
let @xmath510 be a set of size @xmath508 which contains all these vertices and some extra vertices from @xmath470 if necessary .
similarly , there is a set @xmath511 of size @xmath508 such that every vertex @xmath512 has at least @xmath506 inneighbours in @xmath480 for all but at most @xmath513 indices @xmath514 .
let @xmath515 and @xmath503 .
thus @xmath516 and every vertex @xmath517 satisfies @xmath518}(v)\ge \frac{d}{2}(1-{\varepsilon})|n^+_r(i)||a_i|-|u| \stackrel{(\ref{eq : minr})}{\ge } \frac{dn}{8}.\ ] ] the same bound holds for @xmath519}(v)$ ] and thus @xmath502)\ge dn/8 $ ] .
add all the vertices in @xmath520 to @xmath521 to obtain a new set @xmath522 with @xmath523 let @xmath524 be the subgraph of @xmath0 obtained from @xmath525 $ ] by adding the set @xmath522 and setting the out- and inneighbourhoods of any @xmath526 in @xmath524 equal to @xmath527 and @xmath528 .
thus @xmath529 by claims 1.1 and 1.2 . in what follows , we will still write @xmath470 for @xmath530 and @xmath38 for @xmath531 . then in @xmath524 each pair @xmath532 with @xmath297
is still @xmath138-regular and has density at least @xmath533 .
apply the diregularity lemma ( lemma [ direg ] ) to @xmath524 with parameters @xmath534 , @xmath535 , @xmath536 and initial partition @xmath537 to obtain a reduced digraph @xmath538 and a partition @xmath539 ( where the exceptional set is @xmath540 ) which refines @xmath537 .
apply lemma [ red - orient ] to @xmath524 ( playing the role of @xmath0 ) , @xmath538 ( playing the role of @xmath180 ) and @xmath541 ( playing the roles of @xmath542 ) .
we obtain a reduced oriented graph @xmath543 with parameters @xmath544 such that @xmath545 ( this is what we need claim 1.2 for ) .
let @xmath546 denote the corresponding pure oriented graph .
let @xmath550 , for @xmath551 , and @xmath552 , for @xmath553 . then @xmath554 .
since @xmath555 we get @xmath556 , and since @xmath557 we get @xmath558 . for @xmath319 $ ] let @xmath559 \colon a_x ' \subset a_i \}$ ] . then @xmath560 let @xmath561 : |l_i| < ( 1 - 10{\varepsilon})\ell / k\}$ ] .
we will show that @xmath562 is comparatively small .
for this , note that if @xmath563 , then @xmath564 since @xmath565 , this in turn shows that @xmath566 .
so @xmath567 would imply that @xmath568 , contradiction .
thus we must have @xmath569 .
let @xmath570 , @xmath571 \colon |t_i| > d^3\ell / k \}$ ] and @xmath572 .
then @xmath573 and @xmath574 : a'_x\subset v_0''\}| < ( d^3 + 7{\varepsilon } ) \ell$ ] imply that @xmath575 and so ( [ eq : exps ] ) implies that either @xmath576 or @xmath577 .
suppose first that the latter holds .
we will now show that for all @xmath578 and all @xmath579 we have @xmath580 and @xmath581 would span a bipartite subgraph @xmath582 of @xmath532 with vertex class sizes @xmath583 .
note that @xmath584 .
so lemma [ red - orient ] ( c ) applied to @xmath585 ( playing the role of @xmath46 ) and @xmath586 ( playing the role of @xmath587 ) shows that @xmath588 where the last inequality holds since @xmath589 and thus also @xmath590 .
this contradicts the fact that @xmath532 is an @xmath138-regular pair of density at least @xmath533 , which proves ( [ eq : lj ] ) .
therefore @xmath591 the argument in the case when @xmath592 is almost the same : as mentioned at the beginning of case 1 we now have @xmath593 $ ] and so @xmath594 .
now the same argument that we used to find the @xmath56-factor @xmath352 in @xmath188 gives us a @xmath56-factor @xmath595 in @xmath543 .
apply proposition [ superreg ] , with @xmath546 playing the role of @xmath213 , @xmath595 playing the role of @xmath46 ( so @xmath485 ) , the @xmath596 playing the role of the @xmath152 and @xmath285 replaced by @xmath597 .
we obtain an ( oriented ) subgraph @xmath598 of @xmath546 and sets @xmath599 so that a pair @xmath600 is @xmath601-regular of density @xmath602 whenever @xmath603 , and so that @xmath600 is @xmath604-super - regular whenever @xmath181 is an edge of @xmath605 .
let @xmath606 and let @xmath467 be the set obtained from @xmath540 by adding all the vertices in @xmath607 for all @xmath608 $ ] .
then @xmath609 and @xmath610 let @xmath611 be the subgraph of @xmath0 obtained from @xmath612 by adding the set @xmath613 and setting the out- and inneighbourhoods of any @xmath614 in @xmath611 equal to @xmath615 and @xmath616 .
we will apply lemma [ find - cycle ] to find a hamilton cycle in @xmath617 . for this , note that the reduced oriented graph @xmath188 defined above satisfies the initial two conditions of lemma [ find - cycle ] ( c.f .
the discussion just before the statement of claim 1.1 ) .
moreover , we will let the vertex partition @xmath618 of @xmath619 play the role of @xmath620 in the lemma .
the role of @xmath311 will be played by @xmath467 .
now define an oriented graph @xmath621 with vertex set @xmath67 \cup a_0''$ ] as follows .
the restriction of @xmath621 to @xmath67 $ ] is @xmath188 . for each @xmath622 add an edge from @xmath24 to @xmath319 $ ] if @xmath623 and an edge from @xmath319 $ ] to @xmath24 if @xmath624 . using claim 1.1 ,
it is easy to see that @xmath24 has indegree and outdegree at least @xmath625
. however there may be double edges , so delete edges to obtain an oriented graph ( which we still call @xmath621 ) in which each @xmath626 has indegree and outdegree at least @xmath627 ( say , although we only need these degrees to be non - zero ) . now applying lemma [ find - cycle ] to the oriented graphs @xmath188
, @xmath621 ( playing the role of @xmath310 ) and @xmath617 ( playing the role of @xmath213 ) gives a directed hamilton cycle @xmath628 in @xmath617 .
choose any vertex @xmath629 of @xmath617 and let @xmath630 be the vertex that succeeds @xmath629 on @xmath628 .
let @xmath631 be the digraph obtained from @xmath598 by adding a new vertex @xmath632 which has the following neighbours : the outneighbourhood of @xmath632 is @xmath633 and the inneighbourhood is @xmath634 .
let @xmath635 be the digraph obtained from @xmath543 by adding a new vertex @xmath632 and edges as follows .
add an edge from @xmath632 to @xmath636 $ ] if @xmath637 and an edge from @xmath636 $ ] to @xmath632 if @xmath638 .
using claim 1.1 it is easy to see that @xmath632 has indegree and outdegree at least @xmath639 .
again , we can delete edges to arrive at an oriented graph @xmath635 in which @xmath632 has indegree and outdegree at least @xmath640 .
now lemma [ find - cycle ] applied to @xmath543 ( playing the role of @xmath188 ) , @xmath595 ( playing the role of @xmath352 ) , @xmath641 ( playing the role of @xmath310 ) , @xmath642 ( playing the role of @xmath213 ) and with @xmath643 playing the roles of @xmath644 gives a directed hamilton cycle @xmath645 in @xmath642 .
note that this yields a hamilton path in @xmath598 which starts in @xmath646 and ends in @xmath647 , where @xmath648 and @xmath649 are the successor and predecessor of @xmath632 on @xmath645 .
now @xmath650 is a hamilton cycle in @xmath213 and thus in @xmath0 .
this contradiction completes the analysis of case 1 . note
that the argument in the proof of case 1 implies the following lemma , which will be used in @xcite to prove an approximate analogue of chvtal s theorem on hamiltonian degree sequences for digraphs .
[ cyclelemma ] let @xmath651 be positive numbers and let @xmath652 be positive constants such that @xmath653 .
let @xmath0 be an oriented graph on @xmath161 vertices such that @xmath654 .
let @xmath180 be the reduced digraph of @xmath0 with parameters @xmath655 and such that @xmath656 .
suppose that there exists a spanning oriented subgraph @xmath188 of @xmath180 with @xmath657 and such that @xmath658 for all sets @xmath659 with @xmath660 .
then @xmath0 contains a hamilton cycle .
the strategy in this case is as follows : by exploiting the minimum semi - degree condition , we will first show that @xmath0 has roughly a similar structure as the extremal example described in section [ extremal ] . in a sequence of further claims
, we will then either find a hamilton cycle or obtain further structural information on @xmath0 which means that it must be even more similar to the extremal example .
eventually , we arrive at a contradiction , since being ( almost ) exactly like the extremal example is incompatible with the minimum semi - degree condition . unless stated otherwise , all neighbourhoods , degrees and numbers of edges refer to the oriented graph @xmath0 from now on .
let @xmath661 { \setminus}(s \cup n^+_r(s ) ) , \ d_r : = s { \setminus}n^+_r(s).\ ] ] let @xmath662 and define @xmath663 similarly .
by definition we have @xmath664 .
since @xmath188 has parameters @xmath665 , lemma [ red - orient](c , d ) implies that we have @xmath666 from now on we will not calculate explicit constants multiplying @xmath206 , and just write @xmath667 .
the constants implicit in the @xmath668 notation will always be absolute .
first we show that @xmath670 , @xmath671 , @xmath672 and @xmath673 are non - empty .
since the average value of @xmath674 over all @xmath675 is less than @xmath676 , we have @xmath677 also @xmath678 .
since @xmath679 , for any @xmath680 we have @xmath681 .
thus @xmath682 and also @xmath683 .
pick a vertex @xmath684 whose degree in @xmath188 is minimal , a vertex @xmath685 whose outdegree in @xmath188 is minimal and a vertex @xmath686 of whose indegree in @xmath188 is minimal . since the minima are at most the averages , inequality ( [ eq : minr ] ) implies that @xmath687 , @xmath688 and @xmath689 .
we also have the inequality @xmath690 .
thus we may define positive reals @xmath691 , @xmath692 , @xmath693 , @xmath694 by since @xmath713 by ( [ lowdens ] ) we have @xmath714 on the other hand , @xmath715 .
so we must have @xmath716 and @xmath717 .
also , since @xmath718 we have @xmath719 so @xmath720 and @xmath707 .
next , writing @xmath721 and using the inequalities @xmath722 gives @xmath723 so @xmath724 and @xmath725 . finally , since @xmath726 we have @xmath727 and so @xmath728 .
since @xmath729 we have @xmath730 and so @xmath731 .
henceforth we will only use claims 2.1 and 2.2 and make no further use of the information in .
this has the advantage of making our picture invariant under the relabelling @xmath732 , @xmath733 .
also , we add @xmath156 to @xmath38 . since @xmath734 and @xmath735 , this does not affect the assertions of claims 2.1 and 2.2 ( except as usual the constant in the @xmath667-notation ) .
it will sometimes be convenient to use the notation @xmath736 , @xmath737 , @xmath738 , @xmath739 . given a vertex @xmath37 , we will use the compact notation _
@xmath24 has property @xmath740 _ as follows .
the notation starts with some @xmath741 , namely the set that @xmath24 belongs to .
next , for each of @xmath71 its superscript symbol refers to the intersection with @xmath34 and its subscript to the intersection with @xmath742 .
a symbol ` @xmath743 ' describes an intersection of size at least @xmath744 and a symbol ` @xmath745 ' describes an intersection of size at most @xmath746 .
the absence of a symbol means that no statement is made about that particular intersection , and we can omit any of @xmath71 if it has no superscript or subscript .
for example , to say @xmath24 has property @xmath747 means that @xmath748 , @xmath749 , @xmath750 and @xmath751 .
we say that a vertex @xmath24 is _ cyclic _ if it satisfies @xmath752 for some @xmath753 ( counting modulo 4 ) .
claim 2.2 implies that at most @xmath754 vertices are not cyclic .
we say that a vertex is _ acceptable _ if it has one of the following properties in other words , a vertex is acceptable if it has a significant outneighbourhood in one of its two _ out - classes _ and a significant inneighbourhood in one of its two _ in - classes _ , where ` out - classes ' and ` in - classes ' are to be understood with reference to the extremal oriented graph in section [ extremal ] .
so for example @xmath38 has out - classes @xmath38 and @xmath58 and in - classes @xmath38 and @xmath73 .
we will also call an edge _ acceptable _ if it is of the type allowed in the extremal oriented graph ( so for example an edge from @xmath38 to @xmath58 is acceptable but an edge from @xmath58 to @xmath38 is not ) .
note that a cyclic vertex is also acceptable . in what follows
, we will carry out @xmath754 reassignments of vertices between the sets @xmath38 , @xmath58 , @xmath48 and @xmath73 ( and a similar number of ` path contractions ' as well ) .
after each sequence of such reassignments it is understood that the hidden constant in the @xmath771-notation of the definition of an acceptable / cyclic vertex is increased .
* claim 2.3 . * _ by reassigning vertices that are not cyclic to @xmath38 , @xmath58 , @xmath48 or @xmath73 we can arrange that every vertex of @xmath0 is acceptable .
we can also arrange that there is no vertex that is not cyclic but would become so if it was reassigned .
_ we start by making any reassignments necessary to satisfy the second statement of the claim .
( to do this , we reassign all the vertices in question in one step .
as we are reassigning @xmath754 vertices we can then change the hidden constant in the @xmath771-notation to make sure that the vertices which were cyclic before are still cyclic . ) to satisfy the first statement of the claim , for any vertex @xmath37 we let @xmath772 , @xmath773 , and @xmath774 . by the minimum semi - degree condition @xmath775 , @xmath776 and @xmath777 .
if there is some @xmath340 such that @xmath778 and @xmath779 ( where we use addition and subtraction mod @xmath70 ) then we can put @xmath24 into @xmath780 and it will have property @xmath781 , i.e. @xmath24 will become acceptable . otherwise we must have either @xmath782 and @xmath783 , or @xmath784 and @xmath785 . in either case
we can put @xmath24 into @xmath786 : in the first case it will have property @xmath756 and in the second case property @xmath757 .
so in both cases @xmath24 will become acceptable .
as before , by increasing the hidden constant in the @xmath771-notation if necessary , we can ensure that the properties of all other vertices are maintained
. in the proof of the following claims , contractions of paths will play an important role . given a path @xmath50 in @xmath0
whose initial vertex @xmath787 and final vertex @xmath788 lie in the same class @xmath780 , the _ contraction of @xmath50 _ yields the following oriented graph @xmath228 : we add a new vertex @xmath789 to the class @xmath780 and remove ( the vertices of ) the path @xmath50 from @xmath0 .
@xmath790 will be @xmath791 and @xmath792 will be @xmath793 .
note that any cycle in @xmath228 which includes @xmath789 corresponds to a cycle in @xmath0 which includes @xmath50 .
the paths @xmath50 that we use in the proof of claim 2.4 will have initial and final vertices in the same class and will be _ acceptable _ , meaning that every edge on @xmath50 is acceptable .
note that each such acceptable path @xmath50 must be _
@xmath794-balanced _ , meaning that if we delete the initial vertex of @xmath50 we are left with a path that meets @xmath58 and @xmath73 the same number of times .
this may be seen from the observations that visits of @xmath50 to @xmath795 alternate between @xmath58 and @xmath73 , and if the path is in @xmath38 and then leaves it must visit @xmath58 and @xmath73 an equal number of times before returning to @xmath38 ( and similarly for @xmath48 ) .
note that if we have @xmath796 and contract a @xmath794-balanced path , then the resulting digraph will still satisfy @xmath796 .
the ` moreover ' part of the following claim is used later in the proof to turn a graph with @xmath797 into one with @xmath796 under certain circumstances .
* claim 2.4 . * _ if @xmath796 and every vertex is acceptable then @xmath0 has a hamilton cycle . moreover ,
the assertion also holds under the following slightly weaker assumption : @xmath796 and there exists some vertex @xmath24 such that every vertex except @xmath24 is acceptable , there is at least one acceptable edge going into @xmath24 and at least one acceptable edge coming out of @xmath24 .
_ we will use the ` standard ' version of the blow - up lemma ( lemma [ standardblowup ] ) to prove the first assertion . for this
, the idea is to first find suitable paths which together contain all the non - cyclic vertices .
we will contract these paths into vertices so that the resulting oriented graph @xmath798 consists entirely of cyclic vertices .
then we will find suitable paths whose contraction results in an oriented graph @xmath799 which satisfies @xmath800 and all of whose vertices are cyclic .
we can then apply the blow - up lemma to its underlying graph to find a directed hamilton cycle in @xmath799 which ` winds around ' @xmath71 .
the choice of our paths will then imply that this hamilton cycle corresponds to one in @xmath0 .
let @xmath801 be the vertices which are acceptable but not cyclic . for each @xmath802
, choose a cyclic outneighbour @xmath803 and a cyclic inneighbour @xmath804 so that all of these vertices are distinct and so that the edges @xmath805 and @xmath806 are acceptable . note that this can be done since @xmath807 .
let @xmath808 be a path of length at most 3 starting at @xmath803 and ending at a cyclic vertex which lies in the same class as @xmath804 , and where the successive vertices lie in successive classes , i.e. the successor of a vertex @xmath809 lies in @xmath810 .
( so if for example @xmath802 has the first of the acceptable properties of a vertex in @xmath38 , then we can choose @xmath811 , @xmath812 and so @xmath813 would have its final vertex in @xmath73 .
also , if @xmath804 and @xmath803 lie in the same class then @xmath808 consists of the single vertex @xmath803 . ) again , the paths @xmath808 can be chosen to be disjoint .
let @xmath814 .
then the @xmath815 are acceptable and thus @xmath794-balanced .
let @xmath798 be the oriented graph obtained from @xmath0 by contracting the paths @xmath815 .
then every vertex of @xmath798 is cyclic ( by changing the constant involved in the @xmath771 notation in the definition of a cyclic vertex if necessary ) .
moreover , the sets @xmath71 in @xmath798 still satisfy @xmath796 and we still have that the sizes of the other pairs of sets differ by at most @xmath754 .
now suppose that @xmath816 and let @xmath817 .
greedily find a path @xmath383 of the form @xmath818 consisting entirely of cyclic vertices .
so @xmath383 starts with an edge between two cyclic vertices in @xmath48 .
( claim 2.2 implies that almost all ( unordered ) pairs of ( cyclic ) vertices in @xmath48 are joined by an edge . )
then the path @xmath383 uses one cyclic vertex in @xmath73 , one in @xmath38 etc .
let @xmath799 be the digraph obtained by contracting @xmath383 . then in @xmath799
, we have @xmath819 and @xmath796 .
if @xmath820 we can achieve equality in the similar way by contracting a path @xmath821 of the form @xmath822 .
note that since @xmath823 , all vertices of @xmath799 are still cyclic .
now suppose that in @xmath799 we have @xmath824 .
let @xmath825 .
this time , we greedily find a path @xmath826 of the form @xmath827 consisting of cyclic vertices .
to see that such a path exists , note that claim 2.2 implies that there are at least @xmath828 ( say ) cyclic vertices in @xmath58 with at least @xmath829 cyclic outneighbours in @xmath73 .
so @xmath826 starts in such a cyclic vertex in @xmath58 and then uses cyclic vertices in @xmath73 , @xmath38 , @xmath58 and @xmath48 .
since there are at least @xmath829 cyclic vertices in @xmath73 having at least @xmath828 cyclic outneighbours in @xmath58 , @xmath383 can then move to such a cyclic vertex in @xmath73 and use cyclic vertices in @xmath58 , @xmath48 , @xmath73 and @xmath38 etc .
note that @xmath826 is @xmath794-balanced . by contracting @xmath826
, we obtain an oriented graph ( which we still call @xmath799 ) with @xmath800 and all of whose vertices are still cyclic .
finally , suppose that in @xmath799 we have @xmath830 . in this case
we can equalize the sets by contracting two paths @xmath821 and @xmath383 as above .
thus we have arranged that @xmath800 in @xmath799 .
let @xmath831 be the underlying graph corresponding to the set of edges oriented from @xmath780 to @xmath810 , for @xmath832 .
since all vertices of @xmath799 are cyclic and we chose @xmath833 , each pair @xmath834 is @xmath835-super - regular in @xmath836 . also , @xmath831 is simple , i.e. there are no multiple edges .
let @xmath222 be the @xmath70-partite graph with vertex classes @xmath837 where the 4 bipartite graphs induced by @xmath834 are all complete .
clearly @xmath222 has a hamilton cycle , so we can apply lemma [ standardblowup ] with @xmath838 , @xmath839 to find a hamilton cycle @xmath840 in @xmath831 .
the ` moreover ' part of lemma [ standardblowup ] implies that we can assume that @xmath840 continually ` winds around ' @xmath71 , i.e. one neighbour on @xmath840 of a vertex @xmath841 lies in @xmath810 and the other in @xmath842 .
@xmath840 corresponds to a directed hamilton cycle in @xmath799 and thus in turn to a directed hamilton cycle in @xmath0 .
now , we deduce the ` moreover ' part from the first part of claim 2.4 .
similarly as in the first part , the approach is to find a suitable path @xmath50 containing @xmath24 which we can contract into a single vertex so that the resulting oriented graph still satisfies @xmath796 and now all of its vertices are acceptable .
choose @xmath327 and @xmath328 so that @xmath843 and @xmath844 are acceptable edges .
since @xmath327 is acceptable , it has a cyclic inneighbour @xmath845 so that the edge @xmath846 is acceptable
. let @xmath780 be the class which contains @xmath845 .
let @xmath847 be a path of length at most 3 starting at @xmath328 and ending at an cyclic vertex in @xmath780 so that successive vertices lie in successive classes .
let @xmath848 .
then @xmath50 is acceptable and thus @xmath794-balanced .
let @xmath228 be the oriented graph obtained from @xmath0 by contracting @xmath50 .
then in @xmath228 we still have @xmath796 .
all vertices that were previously acceptable / cyclic are still so ( possibly with a larger error term @xmath771 ) .
since @xmath845 and the terminal vertex of @xmath50 are both cyclic ( and thus acceptable ) , this means that the new vertex resulting from the contraction of @xmath50 is still acceptable .
so we can apply the first part of the claim to obtain a hamilton cycle in @xmath228 which clearly corresponds to one in the original oriented graph @xmath0 .
suppose to the contrary that for example we can find @xmath856 vertices @xmath857 in @xmath38 having the property @xmath851 .
select distinct cyclic vertices @xmath858 in @xmath48 with @xmath859 for @xmath860 .
this can be achieved greedily , as @xmath861 , say . since @xmath857 are acceptable vertices in @xmath38 we can greedily select distinct cyclic vertices @xmath862 .
for each @xmath340 such that @xmath863 let @xmath864 .
for each @xmath340 such that @xmath865 select @xmath866 to be a cyclic vertex in @xmath867 ; we can ensure that @xmath868 are distinct .
similarly we can select cyclic vertices @xmath869 which are distinct from each other and from @xmath868 .
thus we have constructed @xmath870 vertex disjoint paths @xmath871 where @xmath815 starts at @xmath872 in @xmath58 , goes through @xmath48 , then it uses @xmath56 or @xmath873 vertices of @xmath38 , and it ends at @xmath866 in @xmath58 . consider a new oriented graph @xmath228 , obtained from @xmath0 by contracting these paths .
note that the vertices of @xmath228 correponding to the paths @xmath815 are acceptable and the analogues of @xmath58 and @xmath73 in @xmath228 now have equal size .
so we can apply claim 2.4 to find a hamilton cycle in @xmath228 .
this corresponds to a hamilton cycle in @xmath0 , which contradicts our assumption
. therefore there are less than @xmath855 vertices with property @xmath851 .
the statement for property @xmath853 follows in a similar way , as we can again start by finding a matching of size @xmath855 consisting of edges directed from @xmath48 to @xmath38 .
the arguments for the other two properties are also similar .
for instance , if we have @xmath855 vertices @xmath801 in @xmath38 having property @xmath852 we can find a matching @xmath874 of edges directed from @xmath58 to @xmath38 such that the @xmath875 are cyclic .
we can extend this to @xmath855 vertex disjoint paths @xmath871 , where @xmath815 starts with the edge @xmath876 , it then goes either directly to a cyclic vertex in @xmath58 or it uses one more vertex from @xmath38 before it ends at a cyclic vertex in @xmath58 .
now we construct a new oriented graph @xmath228 similarly as before and find a hamilton cycle in @xmath228 and then in @xmath0 .
we now say that a vertex is _ good _ if it is acceptable , and also has one of the properties @xmath877 ( the last option means that every acceptable vertex in @xmath73 is automatically good . ) note that a cyclic vertex is not necessarily good . while @xmath72 we reassign vertices that are bad ( i.e. not good ) as follows .
suppose there is a bad vertex @xmath878 with @xmath879 .
if we also have @xmath880 or @xmath881 , then by reassigning @xmath24 to @xmath73 it will become a good vertex .
( recall that for a vertex in @xmath73 , being acceptable and good is the same . ) if not ,
then @xmath882 , and so @xmath883 . in this case
we can make @xmath24 good by reassigning it to @xmath38 .
exactly the same argument works if there is a bad vertex @xmath878 with @xmath884 .
if @xmath24 was in @xmath58 we have decreased @xmath855 by @xmath56 or @xmath873 , whereas by claim 2.5 we will make at most @xmath754 such reassignments of vertices @xmath885 .
similarly , suppose there is a bad vertex @xmath886 with @xmath881 or @xmath880 .
if we also have @xmath879 or @xmath884 , then by reassigning @xmath24 to @xmath73 it becomes good .
if not , then @xmath887 , and so @xmath888 . in this case
we can reassign @xmath24 to @xmath48 .
again , if @xmath24 was in @xmath58 we have decreased @xmath855 by @xmath56 or @xmath873 , whereas by claim 2.5 we will make at most @xmath754 such reassignments of vertices @xmath889 .
the above cases cover all possibilities of a vertex being bad . if during this process @xmath890 and @xmath891 become equal we can apply claim 2.4 to find a hamilton cycle . alternatively
, it may happen that @xmath855 goes from @xmath892 to @xmath893 if a vertex @xmath24 is moved from @xmath58 to @xmath73 . in this case
we claim that we can put @xmath24 into @xmath38 or @xmath48 to achieve the following property : every vertex except @xmath24 is acceptable , there is at least one acceptable edge going into @xmath24 and at least one acceptable edge coming out of @xmath24 . to see this , suppose first that one of @xmath894 or @xmath895 is non - empty .
if one of @xmath896 or @xmath897 is also non - empty we can put @xmath24 into @xmath38 .
otherwise @xmath898 and @xmath899 are both non - empty .
now we can put @xmath24 into @xmath48 unless @xmath900 and @xmath901 are both empty , which can not happen , as then @xmath24 would not have been a bad vertex of @xmath58 .
now suppose that @xmath894 and @xmath895 are both empty , so @xmath900 and @xmath901 are both non - empty .
we can put @xmath24 into @xmath48 unless @xmath898 and @xmath899 are both empty .
but this implies that @xmath24 satisfies the property @xmath902 .
this can not happen , as then @xmath24 would have been reassigned to @xmath73 as a cyclic vertex in claim 2.3 .
therefore we can apply the ` moreover ' part of claim 2.4 to find a hamilton cycle .
but since we are assuming there is no hamilton cycle the process must terminate with an assignment of vertices where all vertices are good .
let @xmath160 be a maximum matching consisting of edges in @xmath903 .
say that @xmath904 matches @xmath905 with @xmath906 , that @xmath907 is a matching on @xmath908 , that @xmath909 matches @xmath910 with @xmath911 and that @xmath912 matches @xmath913 with @xmath914 .
note that @xmath915 .
we must have @xmath916 .
otherwise , by a similar argument to that in claim 2.5 we could extend @xmath917 edges @xmath918 of @xmath160 to @xmath855 vertex disjoint paths @xmath871 , where @xmath815 includes @xmath919 , starts and ends at cyclic vertices @xmath872 , @xmath866 in @xmath58 , and uses two more vertices from @xmath58 than from @xmath73
. then we would find a hamilton cycle as in claim 2.5 , which would be a contradiction .
we will first prove that @xmath920 .
assume to the contrary that @xmath922 .
let @xmath923 , @xmath924 and @xmath925 . by the maximality of @xmath160
there are no edges from @xmath926 to @xmath927 or from @xmath926 to @xmath928 .
since all vertices are good this implies @xmath929 similarly @xmath930 .
so by considering the average indegree of the vertices in @xmath38 we can find a vertex @xmath41 such that @xmath931 also , we can obtain @xmath932 in a similar way as ( [ eac ] ) .
so by averaging , we can find a vertex @xmath933 with @xmath934 by the maximality of @xmath160 , every edge in @xmath58 is incident with a vertex of @xmath935 .
together with the fact that every vertex is good , this implies that @xmath936 so we can find @xmath80 with @xmath937 now we use the minimum semi - degree condition and the above inequalities to get @xmath938 substituting @xmath939 and simplifying gives @xmath940 .
combining this gives @xmath941 , which is only possible if @xmath920 .
now ( [ b - d ] ) immediately implies the claim .
note that @xmath920 implies that @xmath942 .
so by averaging , there is a vertex @xmath301 with @xmath943 . since @xmath920 also implies that @xmath944 and @xmath945 there is vertex @xmath933 with @xmath946 and a vertex @xmath80 with @xmath947 .
so as in ( [ dadcdb ] ) , using @xmath948 we obtain that @xmath949 which is impossible . this contradiction to our initial assumption on @xmath0 completes the proof .
as remarked in the introduction , the proof can be modified to find a cycle of any given length @xmath19 through any given vertex @xmath629 of @xmath0 , where @xmath20 , say .
we give an outline of the necessary modifications here , the details can be found in @xcite . in case 1 , we find the cycle as follows : consider a random subset @xmath950 of @xmath951 vertices of @xmath952 obtained by choosing @xmath953 vertices from each cluster @xmath152 . with high probability ,
the oriented subgraph @xmath954 induced by @xmath955 also has @xmath188 as a reduced oriented graph ( with slightly worse parameters ) .
thus lemma [ cyclelemma ] implies that @xmath954 contains a hamilton cycle .
so suppose we are in case 2 .
the only significant difference in the argument is in the application of the blow - up lemma in the proof of claim 2.4 .
after contracting paths so that all vertices are now acceptable and @xmath800 , we then find a single path @xmath956 which contains all the vertices which correspond to contracted paths and also contains @xmath629 .
we then apply the blow - up lemma to find a path @xmath957 which ( i ) contains all but @xmath958 of the remaining vertices , ( ii ) whose initial vertex lies in the outneighbourhood of the final vertex of @xmath956 and ( iii ) whose final
vertex lies in the inneighbourhood of the initial vertex of @xmath956
. then @xmath959 corresponds to a cycle of length @xmath19 through @xmath629 in @xmath0 .
n. alon and a. shapira , testing subgraphs in directed graphs , _ journal of computer and system sciences _
* 69 * ( 2004 ) , 354382 . n. alon and j. spencer , _ the probabilistic method _ ( 2nd edition ) , wiley - interscience 2000 .
j. komls and m. simonovits , szemerdi s regularity lemma and its applications in graph theory , _ bolyai society mathematical studies 2 , combinatorics , paul erds is eighty ( vol .
2 ) _ ( d. mikls , v. t. ss and t. sznyi eds . ) , budapest ( 1996 ) , 295352 . c. thomassen , long cycles in digraphs with constraints on the degrees , in _ surveys in combinatorics _ ( b. bollobs ed . ) , _ london math . soc . lecture notes _ * 38 * , 211228 , cambridge university press , 1979 . | we show that every sufficiently large oriented graph @xmath0 with @xmath1 contains a hamilton cycle .
this is best possible and solves a problem of thomassen from 1979 . |
let @xmath1 be finite group and @xmath2 a subgroup of @xmath1 .
a subset @xmath3 of @xmath1 is a _ relative difference set _ with parameter @xmath4 if the list of nonzero differences of @xmath3 comprises every element in @xmath5 exactly @xmath6 times .
the subgroup @xmath2 is called the _ forbidden subgroup_. in this paper , we are interested in relative difference sets with parameters @xmath7 and we write @xmath7-rds for short . when @xmath1 is abelian , @xmath3 is called an _ abelian _ @xmath7-rds .
in @xcite , ganley and spence showed that , for every given @xmath7-rds @xmath3 in @xmath1 , we can construct an affine plane of order @xmath8 and group @xmath1 acts regularly on its affine points .
therefore we may use the elements in @xmath1 to denote all the affine points .
all the affine lines are @xmath9 and @xmath10 where @xmath11 and @xmath12 forms a transversal of @xmath2 in @xmath1 .
clearly this affine plane can be uniquely extended to a projective plane .
the extra line is @xmath13 and all @xmath10 s meet at the point @xmath14 .
it is not difficult to see that @xmath1 fix the flag @xmath15 . when @xmath1 is abelian , it is proved that @xmath8 has to be a power of prime . for the proofs , see @xcite for the @xmath8 even case and @xcite for the @xmath8 odd case
furthermore , this abelian group @xmath1 and such a plane are called a _
shift group _ and a _ shift plane _ respectively @xcite .
most of the known shift planes can be coordinatized by commutative semifields .
when @xmath8 is odd , all known abelian @xmath7-rdss are subsets of the group @xmath16 .
such a @xmath7 rds is equivalent to a function @xmath17 , such that @xmath18 is always a bijection for each nonzero @xmath19 .
this type of functions are called _ planar functions _ on @xmath20 , which were first investigated by dembowski and ostrom in @xcite . as the counterpart , when @xmath21 , abelian @xmath7-rdss only exist in @xmath22 where @xmath23 is the cyclic group of order @xmath24 .
these rdss can also be equivalently illustrated by functions over @xmath25 , which can be found in .
let @xmath26 be an integer larger than or equal to @xmath27 .
a _ unital _ of order @xmath26 is a @xmath28-@xmath29 design , i.e. a set of @xmath30 points arranged into subsets of size @xmath31 such that each pair of distinct points are contained in exactly one of these subsets .
most of the known unitals can be embedded in a projective plane @xmath32 of order @xmath0 . in such cases ,
the _ embedded unital _ is a set @xmath33 of @xmath34 points such that each line of @xmath32 intersects @xmath33 in @xmath35 or @xmath36 points .
when @xmath32 is the desarguesian projective plane @xmath37 , the set of absolute points of a unitary polarity , or equivalently speaking , the rational points on a nondegenerate hermitian curve form a _ classical _ unital .
there are also non - classical unitals in @xmath37 , for instance the buekenhout - metz unitals @xcite .
there also exist unitals which can not be embedded in a projective plane , such as the ree unitals @xcite .
similarly as in desarguesian planes , unitals can be derived from unitary polarities in shift planes ; see @xcite for their constructions and related research problems . as a special type of shift planes ,
commutative semifield planes also contain the unitals which are analogous to the buekenhout - metz ones in desarguesian planes ; see @xcite . in this paper , we consider the unitals in the shift planes of order @xmath0 . we restrict ourselves to the @xmath8 odd case , because most of the calculations and constructions in the @xmath8 even case are quite different .
the organization of this paper is as follows : in section [ sec : construction ] , after a brief introduction to shift planes @xmath38 derived from planar functions @xmath39 , we present a generic construction of unitals in shift planes of order @xmath0 . in section [ sec : properties ] , we investigate various geometric and combinatorial properties of them , such as the self - duality , the existences of onan configurations , the wilbrink s conditions , the designs formed by circles and so on . in section
[ sec : inequivalence ] , we consider the unitals derived from the unitary polarities of the shift planes and show that they are not equivalent to the unitals constructed in section [ sec : construction ] . as designs ,
our unitals are also not isomorphic to the classical unitals ( the hermitian curves ) .
let @xmath40 be a finite field of odd order and @xmath39 a planar function on @xmath40 .
the projective plane @xmath38 derived from @xmath39 is defined as follows : * * points : * @xmath41 and @xmath42 for @xmath43 ; * * lines : * @xmath44 for all @xmath45 , @xmath46 and @xmath47 . the points except for those on @xmath13 are called the _ affine points _ of @xmath38 . by removing the line @xmath13 and the points on it , we get an affine plane . it is routine to verify that the set of maps @xmath48 induces a collineation group on @xmath38 and this group is abelian and acts regularly on the affine points and all lines @xmath49 .
it is also transitive on the line set @xmath50 and on the points of @xmath51 .
hence there are three orbits of all the points ( lines ) in @xmath38 under this group .
we call it the _ shift group _ of @xmath38 . when @xmath39 can be written as a dembowski - ostrom polynomial , i.e. @xmath52 where @xmath53 , the plane @xmath38 is also a commutative semifield plane . using the corresponding semifield multiplication
, we can label the points and lines of @xmath38 in a different way .
the intersection of the translation group and the shift group of @xmath38 is @xmath54
. see ( * ? ? ?
* section 4 ) for details .
we refer to @xcite and @xcite for recent surveys on semifields and relative difference sets respectively . up to equivalence , all known planar functions @xmath39 on finite fields @xmath20 of odd characteristics can be written as a dembowski - ostrom polynomial except for the coulter - matthews ones which are power maps defined by @xmath55 on @xmath56 for certain @xmath57 ; see @xcite .
both the dembowski - ostrom planar functions and the coulter - matthews ones satisfy that * @xmath58 and * for arbitrary @xmath59 , @xmath60 if and only if @xmath61 . for a proof of the dembowski - ostrom polynomials case , we refer to @xcite ; for the coulter - matthews functions @xmath62 on @xmath56 , it can be verified directly from the fact @xmath63 . actually for a function @xmath39 defined by a dembowski - ostrom polynomial ,
the above conditions are necessary and sufficient for @xmath39 to be planar ; see @xcite .
if a planar function satisfies the aforementioned two conditions , then we call it a _ normal planar function_. [ le : basic ] let @xmath39 be a planar function on @xmath64 , where @xmath0 is odd .
let @xmath65 be a set of injections from @xmath20 to @xmath64 .
the set of points @xmath66 is a unital in @xmath38 if for each @xmath67 , there are @xmath35 or @xmath36 pairs @xmath68 such that @xmath69 it is not difficult to verify that @xmath70 satisfies the definition of a unital in the following steps .
first , every line through @xmath14 meet @xmath70 at @xmath36 points .
second , @xmath13 meet it at @xmath14 .
finally , the number of common points of @xmath71 and @xmath70 are exactly the number of pairs of @xmath68 such that @xmath72 .
let @xmath73 be an element in @xmath74 .
then every element @xmath75 of @xmath64 can be written as @xmath76 where @xmath77 .
similarly , every function @xmath78 can be written as @xmath79 where @xmath80 are maps from @xmath20 to itself . throughout this paper
, we frequently switch between the element @xmath81 and its two dimensional representation @xmath82 .
if a special assumption on @xmath73 is needed , we will point it out explicitly .
[ prop : general ] let @xmath39 be a planar function on @xmath64 and @xmath83 , where @xmath8 is odd .
assume that @xmath84 then the set of points @xmath85 is a unital in @xmath38 .
furthermore , @xmath71 is a tangent line to @xmath86 if and only if @xmath87 .
first , is equivalent to the following two equations @xmath88 as @xmath89 , the system of equations above is equivalent to @xmath90 ( if @xmath91 , then we replace the second equation by @xmath92 . ) noting that @xmath93 ranges through all the elements in @xmath20 , we see that the cardinality of @xmath68 satisfying is equivalent to the cardinality of @xmath75 such that holds .
according to the assumption , @xmath86 is a unital in @xmath38 .
the last statement of the proposition follows from directly .
next we consider several special cases of proposition [ prop : general ] .
let @xmath94 be the quadratic character on @xmath20 , i.e.@xmath95 the integer - valued function @xmath96 on @xmath20 is defined by @xmath97 for @xmath98 and @xmath99 . to prove the existence of unitals @xmath86 in many shift planes , we need the following well - known result . [ le : quadratic ] for odd prime power @xmath8 , let @xmath100 , @xmath101 , @xmath102 and @xmath103 .
let @xmath104 its _ discriminant _ is defined by @xmath105 .
assume that @xmath106 .
then the number of solution of @xmath107 is @xmath108 in particular , when @xmath109 is a nonsquare , @xmath110 [ th : many.planar.functions.1 ] let @xmath39 be a planar function on @xmath64 , where @xmath111 , @xmath112 is an odd prime and @xmath113 is a positive integer .
assume that @xmath114 is an element in @xmath115 such that @xmath116 is a nonsquare in @xmath20 and @xmath39 satisfies that @xmath117 for each @xmath118 .
then the set of points @xmath119 is a unital in @xmath38 .
furthermore , @xmath86 is a unital for the following planar functions : 1 .
@xmath120 , i.e. @xmath38 is a desarguesian plane .
2 . @xmath121 where @xmath122 is an integer satisfying that @xmath123 and @xmath124 is odd , i.e. @xmath38 is an albert s commutative twisted field plane @xcite .
@xmath125 where @xmath126 ( now @xmath127 ) , i.e. @xmath38 is a coulter - matthews plane which is not a translation plane @xcite . under the assumption , it is clear that @xmath128 now assume that @xmath129 satisfying @xmath130 .
similarly as in proposition [ prop : general ] , we denote @xmath131 and @xmath132 where @xmath133 , @xmath134 and @xmath135
. then @xmath136 by in proposition [ prop : general ] and lemma [ le : quadratic ] , we only have to show that @xmath137 is an irreducible quadratic form .
its discriminant equals @xmath138 . from the assumption and lemma [ le : quadratic ]
, it follows that @xmath86 is a unital in @xmath38 .
as the three families of planar functions are all power maps , by considering the greatest common divisors of the exponent and @xmath139 , it is not difficult to show that they all satisfy . therefore , @xmath86 is a unital in anyone of these shift planes .
[ th : many.planar.functions ] let @xmath39 be a planar function on @xmath64 , where @xmath111 , @xmath112 is an odd prime and @xmath113 is a positive integer .
let @xmath73 be an element in @xmath74 .
let @xmath140 be the set of points defined by .
when @xmath141 and @xmath142 is a nonsquare in @xmath20 , @xmath140 is a unital in the following commutative ( pre)semifield planes , besides those appeared in theorem [ th : many.planar.functions.1 ] : 1 .
dickson s semifield planes @xcite and the corresponding planar functions are @xmath143 where @xmath144 is an integer satisfying @xmath145 .
the semifield planes constructed by pott and the last author in @xcite and the corresponding planar functions are @xmath146 where @xmath147 are integers such that @xmath148 and @xmath149 is odd . when @xmath127 , @xmath140 is a unital in 1 .
ganley s semifields planes where @xmath113 is defined to be odd @xcite .
the corresponding planar functions are @xmath150 2 .
penttila - williams semifield planes where @xmath151 @xcite .
the corresponding planar functions are @xmath152 when @xmath153 , @xmath140 is a unital in 1 .
budaghyan - helleseth semifield planes @xcite .
the corresponding planar functions are @xmath154 where integer @xmath122 is such that @xmath155 and @xmath156 , and @xmath100 is a nonsquare in @xmath115 .
the proof for the first two cases is straightforward : as the first component of the corresponding planar functions are @xmath157 and @xmath158 , clearly we can choose @xmath159 for these two cases and @xmath160 is irreducible if and only if @xmath161 is a nonsquare in @xmath162 .
it is equivalent to that @xmath163 is a square , which holds exactly when @xmath164 . the cases ( c ) and ( d ) are also not difficult to verify
: we can replace @xmath165 in @xmath166 by @xmath167 and replace @xmath168 in @xmath169 by @xmath167 .
then the results follow from the fact that @xmath163 is a nonsquare in @xmath170 where @xmath113 is odd . in ( e ) , noting that @xmath79 where @xmath171 and @xmath172 for @xmath173 , by proposition [ prop : general ] we only have to concentrate on @xmath174 . as @xmath175
, we see that @xmath176 for each @xmath177 .
now we choose @xmath178 satisfying that @xmath179 and write @xmath180 where @xmath181 .
it is clear that @xmath182 . then @xmath183 its discriminant equals @xmath184 .
clearly @xmath185 is a nonsquare in @xmath186 , otherwise @xmath187 leading to a contradiction . as @xmath100 is a nonsquare in @xmath188
, @xmath189 has to be a nonsquare in @xmath186 .
therefore @xmath190 is a nonsquare in @xmath186 if and only if @xmath163 is a nonsquare in it .
this happens exactly when @xmath191 . from proposition
[ prop : general ] and lemma [ le : quadratic ] , it follows that @xmath140 is a unital .
in this section , we proceed to investigate several common properties of @xmath86 obtained in proposition [ prop : general ] .
several important subgroups of their automorphism groups and self - duality are considered in subsection [ subsec : basic ] .
then we look at the projections of the blocks in the unital , which are called circles , and the wilbrink condition . in subsection
[ subsec : onan ] , we investigate the existence of onan configurations . to conclude this section
, we present a conjecture on the automorphism groups of our unitals and several comments on this conjecture .
an _ oval _ @xmath192 in a projective plane @xmath32 of odd order @xmath8 is a set of @xmath36 points such that every line in @xmath32 meets @xmath192 in @xmath193 , @xmath35 or @xmath28 points . according to the famous result by segre in @xcite , all ovals in desarguesian planes of odd orders are nondegenerate conics .
[ prop : unital = ovals ] let @xmath39 be a normal planar function on @xmath64 and @xmath86 be a unital in @xmath38 constructed in proposition [ prop : general ] .
then for each @xmath194 , the set @xmath195 is an oval in @xmath38 and @xmath86 is a union of ovals . as @xmath39 is normal , for each given @xmath196 , there are at most two solution such that @xmath197 . by checking the cardinalities of the points in @xmath198 , @xmath199 and @xmath200 , we see that @xmath201 is an oval in @xmath38 .
therefore @xmath202 . 1 .
a family of unitals in @xmath203 , each of which is a union of ovals , were independently discovered by hirschfeld and sznyi @xcite and by baker and ebert @xcite .
actually , baker and ebert @xcite also showed that this family of unitals is a special subclass of the buekenhout - metz ones in @xmath203 . noting that @xmath38 is desarguesian @xcite
, we can readily verify that our unitals @xmath86 in @xmath204 ( see theorem [ th : many.planar.functions.1 ] ( a ) ) are exactly the family of unitals obtained in @xcite .
this result can also be derived from proposition [ prop : unital = ovals ] and ( * ? ? ?
* theorem 1.1 ) .
+ an analogous result for albert s twisted field planes ( theorem [ th : many.planar.functions.1 ] ( b ) ) can be found in @xcite by abatangelo , korchmros and larato .
bukenhout - metz type of unitals can be also obtained for dickson s semifield planes and the ones constructed in @xcite .
we refer to @xcite , in which the unitals obtained in theorem [ th : many.planar.functions ] ( a ) and ( b ) appear as a subclass of a family of bukenhout - metz type of unitals . given two unitals @xmath205 and @xmath206
, we say that they are _ isomorphic _ if there is a design isomorphism between them , i.e. there is a bijection between their point sets which maps the blocks of @xmath205 to the blocks of @xmath206 .
when @xmath205 and @xmath206 can be embedded into the same projective plane @xmath32 , we say they are _ equivalent _ if there is a collineation of @xmath32 mapping @xmath205 to @xmath206 . for a unital @xmath33
, we use @xmath207 to denote its automorphism group .
when @xmath33 can be embedded in a projective plane @xmath32 , we denote the group of all the collineation of @xmath32 fixing @xmath33 by @xmath208 .
[ prop : unital_powerplanar_orbit ] let @xmath57 be a positive integer and @xmath62 a planar function on @xmath64 . let @xmath209 . for @xmath210 and @xmath211 , @xmath212 on the points of @xmath38 is defined by @xmath213 then all @xmath212 together form a group @xmath214 as a collineation group of @xmath38 . in @xmath38 , all the unitals @xmath215
constructed in theorem [ th : many.planar.functions.1 ] are equivalent .
first it is straightforward to check that under @xmath212 , the line @xmath71 is mapped to @xmath216 , the line @xmath217 is mapped to @xmath218 and @xmath13 is fixed .
hence @xmath212 is a collineation .
it is also not difficult to see that under the composition of maps , the elements in @xmath214 form a group .
let @xmath114 be an element in @xmath115 satisfying that @xmath116 is a nonsquare in @xmath186 ( see theorem [ th : many.planar.functions.1 ] ) , i.e. @xmath114 is an odd power of a primitive element of @xmath64 . applying @xmath212 on @xmath86 , we get @xmath219 . as @xmath220 and
the set @xmath221 covers all the nonzero squares in @xmath64 , we see that all the unitals @xmath222 for @xmath223 are in one orbit under @xmath212 .
hence all the unitals @xmath224 , where @xmath225 satisfies the condition in theorem [ th : many.planar.functions.1 ] , are equivalent .
let @xmath33 be a unital of order @xmath113 embedded in a projective plane @xmath32 .
we can obtain a new design @xmath226 in the dual plane @xmath227 by taking the tangent lines to @xmath33 as the points of @xmath226 and the points of @xmath228 as the blocks of @xmath226 .
the incidence of @xmath226 is given by reverse containment .
it is easy to see that @xmath226 is another @xmath28-@xmath229 design , which is called the _ dual unital _ of @xmath33 .
if @xmath33 and @xmath226 are isomorphic as designs , then @xmath33 is called _ self - dual_. let @xmath39 be a planar function and @xmath38 the corresponding shift plane .
let @xmath230 be its dual plane .
the points in @xmath230 are the three types of lines @xmath71 , @xmath217 and @xmath13 in @xmath38 . if @xmath231 , then clearly @xmath232 .
that means the line @xmath233 in @xmath38 contains @xmath234 if and only if the point @xmath71 in @xmath230 is on the line @xmath235 . for @xmath217 s and @xmath13
, we can also get similar results .
that means if we switch the notations of points and lines in @xmath38 , i.e. @xmath236 then we get @xmath230 . let @xmath86 be the unital of order @xmath8 in @xmath38 defined in proposition [ prop : general ] where @xmath39 is a planar function .
then @xmath86 is self - dual . by proposition [ prop : general ] , we know that @xmath71 is a tangent line if and only if @xmath237 , i.e. @xmath238 as the constant @xmath239 , @xmath240 for certain @xmath241 . together with the tangent line @xmath13
, we know that the dual of @xmath86 in @xmath230 can be written as @xmath242 switching the notations of points and lines in @xmath243 and @xmath230 , we get @xmath86 and @xmath38 .
therefore @xmath86 is self - dual .
let @xmath39 be a planar function on @xmath64 and @xmath38 the corresponding projective plane .
let @xmath86 be a unital in @xmath38 defined by . as a design ,
its point set is @xmath244 and all of its blocks are @xmath245 for each @xmath246 and @xmath247 for each @xmath248 where @xmath249 .
there are totally @xmath250 blocks and each of them contains @xmath36 points . for each pair of points , there is exactly one block containing them both . hence @xmath86 is a @xmath28-@xmath251-design . we denote @xmath252 where @xmath253 . clearly , @xmath254 is the set of the elements appearing in the first coordinate of the elements in @xmath255 ; in other words , @xmath254 can be viewed as a projection of @xmath255 . when the context is clear , we omit @xmath100 and
write @xmath254 as @xmath256 . inspired by the approach of onan in @xcite , we call @xmath256 a _ circle _ for @xmath246 and @xmath257 . by setting @xmath258 ,
we can shortly write a circle as @xmath259 by choosing appropriate @xmath260 , we can also write @xmath261 [ le : circles ] let @xmath262 denote the set of circles derived from @xmath86 .
then following statements hold .
1 . @xmath263 , for arbitrary @xmath246 and @xmath257 .
2 . for each @xmath246 , the set @xmath264 forms a partition of @xmath265 .
all the blocks @xmath255 which intersect @xmath266 are projected to @xmath267 .
4 . @xmath268 if and only if @xmath269 and @xmath270 .
5 . @xmath271 .
\(a ) and ( b ) follows directly from the definition of @xmath256 .
when @xmath255 intersects @xmath266 , that means @xmath272 for certain @xmath241 .
together with , we have @xmath273 hence @xmath255 is projected to @xmath267 .
assume that @xmath274 and @xmath275 are both not @xmath193 , @xmath268 and @xmath276 .
it follows that there are at least @xmath36 elements @xmath75 in @xmath64 such that @xmath277 which equals @xmath278 as there are totally @xmath8 elements @xmath279 in @xmath64 such that @xmath280 and the map defined by @xmath281 is a permutation , there are exactly @xmath8 elements such that holds .
it is a contradiction .
hence @xmath269 . from ( b ) , we immediately get that @xmath270 .
\(e ) follows from ( d ) using a simple counting argument .
@xmath282 is a @xmath283-design .
we only have to prove that for each two difference elements @xmath284 , there are exactly @xmath8 circles containing them . by lemma [ le : circles ] ( c ) , the circle @xmath256 contains @xmath285 and @xmath286 if and only if @xmath287 , which is equivalent to @xmath288 by the same argument in the proof of lemma [ le : circles ] ( d ) , there are exactly @xmath8 elements @xmath246 such that the above equation holds
. therefore there are exactly @xmath8 circles containing @xmath285 and @xmath286 . in @xcite
, wilbrink characterized the classical unital by three intrinsic conditions .
the wilbrink s condition on a unital @xmath33 is as follows .
[ de : wilbrink ] let @xmath286 be a point of a unital @xmath33 .
let @xmath289 be a block such that @xmath290 .
let @xmath291 be a block which contains @xmath286 and meets @xmath289 , and @xmath292 a point in @xmath291 which is distinct from @xmath286 and @xmath293 .
if there exists a block @xmath294 such that @xmath295 and @xmath296 meets all blocks which contain @xmath286 and meet @xmath289 , then we call @xmath286 a _ vertex of wilbrink s condition _ ; see figure [ fig : wilbrink ] .
if @xmath286 is such a vertex for all the blocks @xmath291 , @xmath289 and point @xmath292 in @xmath291 satisfying aforementioned conditions , then @xmath286 is called a _
vertex of wilbrink s condition in strong form _
@xcite .
a point @xmath286 in @xmath33 is a vertex of wilbrink s condition in strong form together with the nonexistence of an onan configuration ( see subsection [ subsec : onan ] ) containing @xmath286 imply that @xmath289 and @xmath296 in above definition have no common points .
hence there is a parallelism defined on the blocks of @xmath33 not containing @xmath286 , i.e. two blocks @xmath289 and @xmath296 of @xmath33 are parallel if they meet the same blocks containing @xmath286 .
it can be proved that all the points in a desarguesian plane are vertices of wilbrink s condition in strong form ; see ( * ? ? ?
* lemma 7.42 ) .
we can prove the following result for our unitals .
[ prop : wilbrinkii ] let @xmath39 be a normal planar function on @xmath64 and @xmath86 a unital in @xmath38 defined in proposition [ prop : general ] . in @xmath86 , @xmath14 is a vertex of wilbrink s condition in strong form .
let @xmath248 and @xmath297 .
let @xmath291 be an arbitrary block which contains @xmath14 and intersects @xmath289 , namely @xmath298 for certain @xmath194 such that @xmath299 .
let @xmath292 be an arbitrary point on @xmath300 which is different from the intersection point @xmath301 of @xmath300 and @xmath255 .
hence @xmath302 for certain @xmath303 and @xmath304
. then we can take @xmath305 . from @xmath299
, we see that for any @xmath81 , @xmath306 if and only if @xmath307 . hence @xmath14 satisfies the wilbrink s condition for @xmath289 , @xmath291 and @xmath292 . from the arbitrariness of @xmath289 , @xmath291 and @xmath292 , we derive that the point @xmath14 is a vertex of wilbrink s condition in strong form . for different shift planes @xmath38
, we can not find a generic way to investigate whether the affine points in @xmath86 are vertices of wilbrink s condition or not .
for @xmath120 , we get the following result . [
th : wilbrink_x^2 ] let @xmath8 be a prime power larger than @xmath27 .
let @xmath120 be defined on @xmath64 and @xmath86 a unital in @xmath38 defined in theorem [ th : many.planar.functions.1 ] ( a ) . in @xmath86 , @xmath14 is the unique vertex of wilbrink s condition in strong form . from proposition [ prop : wilbrinkii ]
, we already know that @xmath14 is a vertex of wilbrink s condition in strong form .
assume that @xmath286 distinct from @xmath14 is a vertex of wilbrink s condition in strong form .
as the shift group is transitive on the affine points in @xmath86 , we only have to consider the point @xmath308 .
we choose the point @xmath292 , the blocks @xmath289 and @xmath291 as follows : @xmath309 clearly @xmath286 and @xmath292 are both on @xmath291 , and @xmath291 meets @xmath289 in point @xmath14 . according to definition [ de : wilbrink ] , there is a block @xmath296 such that * @xmath296 contains @xmath292 , * @xmath296 meets all the blocks which contain @xmath286 and meet @xmath289 .
the first condition implies that there exists @xmath246 such that @xmath310 .
the set @xmath311 of all the blocks containing @xmath286 , meeting @xmath289 and distinct from @xmath291 is @xmath312 in other words , for each point @xmath313 on @xmath289 , we have a block @xmath314 containing @xmath313 and @xmath286 .
the second condition means that for each @xmath315 , there is unique @xmath241 such that @xmath316 obviously @xmath317 .
according to the assumption , @xmath120 .
together with the definition of @xmath314 , and we obtain that @xmath318 plugging them back into , we have @xmath319 simplifying it , we have @xmath320 as @xmath241 , the above equation means @xmath321 multiplying both hand sides by @xmath322 and simplifying then yields @xmath323 letting @xmath324 and plugging @xmath325 into the above equation , we obtain @xmath326 if we view @xmath93 as an indeterminate and the above equation as an element @xmath327 in the polynomial ring @xmath328 $ ] , then @xmath327 vanishes on the elements in @xmath20 .
that means the polynomial @xmath329 divides @xmath327 . through straightforward calculation
, we see that the coefficient of @xmath330 in @xmath327 is @xmath193 and that of @xmath331 is @xmath332 this coefficient must be @xmath193 , because @xmath333 and @xmath334 . dividing it by @xmath116 and simplifying , we get @xmath335 plugging @xmath324 back into it , we deduce that @xmath336 which means @xmath337 for certain @xmath338 . noting that @xmath93 can be any element in @xmath20 , we take @xmath339 .
that means @xmath340 , and from we have @xmath341 it implies that @xmath342 , from which it follows that @xmath343 .
now we look at the constant term of @xmath327 , which is @xmath344 as we showed previously , @xmath334 .
hence the constant term must be zero .
plugging @xmath345 back into it , we have @xmath346 this can be view as a polynomial in @xmath114 of degree @xmath347 . by proposition [ prop : unital_powerplanar_orbit ]
, we know that all the @xmath348 unitals @xmath349 are equivalent , where @xmath350 . as @xmath351
, we may replace @xmath114 by some @xmath352 such that does not hold anymore .
this replacement does not affect the previous part of the proof , because we never used the value of @xmath114 there .
therefore , in the unital @xmath349 we can not find a block @xmath353 ( because fails ) such that the wilbrink condition holds . for @xmath354 , we can use magma @xcite to show that theorem [ th : wilbrink_x^2 ] also holds . for other planar functions ,
it is also possible to investigate the same question .
however , if we follow the steps of the proof of theorem [ th : wilbrink_x^2 ] , we will see the difficulties in solving and as well as more complicated calculations .
onan configuration _ is a collection of four lines intersecting in six points . in @xcite ,
onan first considered this configuration , and he proved that it does not exist in the classical unitals . in @xcite , piper conjectured that the nonexistence of onan configuration is also a sufficient condition for a unital to be classical . in @xcite
, wilbrink investigated this conjecture and obtained a weaker version characterization .
see @xcite for a recent progress on this conjecture . in this subsection
, we consider the existence of onan configuration in the unitals @xmath355 in @xmath38 for various planar functions @xmath39 .
first , we look at onan configurations containing the point @xmath14 . [ le : onan_infty ] let @xmath39 be a planar function from @xmath64 to itself and @xmath356 such that @xmath86 is a unital in @xmath38 .
there exists an onan configuration with @xmath14 as a vertex if and only if there exist @xmath223 and @xmath357 such that @xmath358 , where @xmath256 is the circle defined by
. @xmath359 by the definition of circles , there exist blocks @xmath255 and @xmath360 corresponding to @xmath256 and @xmath361 respectively for certain @xmath362 , such that @xmath363 assume that @xmath364 , which means that @xmath365 for @xmath366 .
let @xmath367 .
it follows that @xmath368 .
together with , we have @xmath369 hence @xmath255 and @xmath370 meets at the point @xmath371 and both @xmath370 and @xmath360 are projected to the same circle @xmath361 . from the assumption , we see that @xmath372 and @xmath373 intersect both @xmath255 and @xmath370 . therefore , these four blocks intersect totally in six points and form an onan configuration .
@xmath374 assume that there is an onan configuration containing @xmath14 .
it implies that there are two blocks @xmath289 and @xmath296 intersecting in a point @xmath375 and not containing @xmath14 , and they both intersect @xmath300 and @xmath376 for certain @xmath377 . under the automorphism group of the unital
, we can always assume that @xmath378 and @xmath379 , which means that the corresponding circle is @xmath361 . denoting the circle derived from @xmath296 by @xmath380
, we see that @xmath381 . according to (
* page 80 ) , there is no onan configuration with @xmath14 in the buekenhout - metz unitals of odd order .
it implies that there is no onan configuration containing @xmath14 in @xmath86 defined over @xmath38 , where @xmath120 .
it is also not difficult to see this result from lemma [ le : onan_infty ] and . for two arbitrary circles @xmath256 and @xmath382
, we look at the common solution of @xmath383 and @xmath384 . subtracting the second equation from the first one , we get @xmath385 . hence if @xmath276 , then @xmath386 equals a polynomial of degree one in @xmath387 $ ] . plugging it back into @xmath383 , we get a quadratic equation , which means there are at most two solutions .
the same approach also works for the unital defined in theorem [ th : many.planar.functions.1 ] when @xmath39 is derived from a dickson s semifield . for @xmath121 and @xmath388
, we conjecture that there do exist onan configuration containing the point @xmath14 .
this is confirmed for several small @xmath8 by using magma @xcite .
however we can not find a proof .
next , let us look for onan configurations without @xmath14 .
[ le : onan_without_infty ] let @xmath39 be a planar function from @xmath64 to itself and @xmath356 such that @xmath86 is a unital in @xmath38 .
let @xmath389 .
let @xmath390 , @xmath391 and @xmath392 and @xmath393 , @xmath394 , @xmath395 which are pairwise distinct . assume that there exist three distinct elements @xmath396 , @xmath397 and @xmath398 such that for each @xmath399 , 1 .
@xmath400 for pairwise distinct @xmath144 , @xmath401 , where @xmath402 , and 2 .
@xmath403 for @xmath404 .
then there is an onan configuration without @xmath14 in @xmath86 . under the assumption , we show that there exists an onan configuration containing the points @xmath405 , @xmath406 , @xmath407 .
let @xmath408 for @xmath409 .
clearly @xmath410 , @xmath411 and @xmath412 are the lines in @xmath38 containing @xmath413 , @xmath414 and @xmath415 respectively . the first coordinate @xmath416 of the intersection point of @xmath410 and @xmath411
is the solution of @xmath417 from which it follows that @xmath418 .
according to the definition of @xmath86 , this point is in the unital @xmath86 if and only if @xmath419 .
similar results can be obtained for @xmath420 and @xmath421 . noting that @xmath396 , @xmath397 and @xmath416 are pairwise distinct , @xmath410 , @xmath411 and @xmath412 do not intersect in the same point .
therefore these three intersecting points together with @xmath413 , @xmath414 and @xmath415 form a onan configuration . in lemma
[ le : onan_without_infty ] , the onan configuration obtained contains a line through @xmath14 . without loss of generality
, we can assume that @xmath422 . from its proof
, it is not difficult to see that for given @xmath122 , we only have to take one @xmath404 to check whether ( b ) holds . using lemma [ le : onan_without_infty ]
, we can prove the existence of onan configurations in @xmath86 for several different planar functions @xmath39 .
[ th : onan_in_x^d ] let @xmath113 be a positive integer , @xmath423 and @xmath122 an integer satisfying @xmath424 and @xmath425
. for the planar functions @xmath120 and @xmath121 on @xmath64 , the unital @xmath86 defined in theorem [ th : many.planar.functions.1 ] contains an onan configuration . to unify the proof for these two classes of planar functions , we set @xmath426 when @xmath120 .
let @xmath427 be an element in @xmath64 such that @xmath428 .
since @xmath429 , @xmath427 is also in @xmath430 .
let @xmath394 and @xmath431 be two elements in @xmath432 such that @xmath433 and @xmath434 .
define @xmath435 as @xmath427 is also in @xmath436 , we have that @xmath437 . from @xmath433 and @xmath438 ,
it is readily to deduce that @xmath393 is also distinct from @xmath193 , @xmath394 and @xmath431 .
as @xmath429 , we have @xmath439 , which guarantees the distinct values of @xmath193 , @xmath394 , @xmath431 and @xmath393 . by lemma [ le : onan_without_infty ] , now @xmath440 and @xmath441 hence @xmath442 where the last equality comes from the fact that @xmath428
. as @xmath443 and the mapping @xmath444 is a bijection on @xmath64 , we have @xmath445 .
thus @xmath446 .
similarly , we have @xmath447 and @xmath448 .
thus @xmath449 . therefore , the two conditions in lemma [ le : onan_without_infty ] are satisfied and there is an onan configuration in @xmath86 .
[ th : onan_in_two_dim ] let @xmath450 be an odd prime power satisfying that @xmath451 and @xmath452 .
let @xmath140 be a unital defined in theorem [ th : many.planar.functions ] for dickson s semifields or for the semifields constructed in @xcite . if there is @xmath453 such that @xmath428 , i.e. @xmath454 , then there is an onan configuration in @xmath86 . to unify the proof for these two semifields , we write @xmath455 , where @xmath456 and @xmath457 is odd .
it implies that the multiplication @xmath458 defined in lemma [ le : onan_without_infty ] is @xmath459 for @xmath460 .
similarly as in theorem [ th : onan_in_x^d ] , we take @xmath394 and @xmath461 such that @xmath433 and @xmath434 .
it is not difficult to see that @xmath462 .
thus we can always find @xmath394 and @xmath431 satisfying all the assumptions .
define @xmath463 it is straightforward to verify the conditions in lemma [ le : onan_without_infty ] . to conclude this section , we propose an open problem on the automorphism group @xmath464 of @xmath86 . [
conjecture : automorphism.main ] let @xmath78 be a normal planar function and @xmath86 a unital in @xmath38 defined in proposition [ prop : general ] .
then @xmath465 in general , it is quite difficult to determine the automorphism group of a unital as a design . for a classical unital in @xmath37 , onan
@xcite proved that its automorphism group is @xmath466 . in @xcite , hui ,
law , tai and wong obtained a similar result for the unitals defined by polarities in the dickson s semifield planes .
a natural approach , which was used in @xcite , is to consider the automorphism group of the design formed by the set of all circles derived / projected from the blocks of the corresponding unital . in @xcite , one extra point was added to this design and a miquelian inversive plane was obtained . as the automorphism group of a miquelian inversive plane is @xmath467 ( see @xcite ) ,
it is possible to further determine the automorphism group of the original unital .
unfortunately , this approach can not be generally applied on the unitals @xmath86 constructed in this paper .
for example , when @xmath468 is a coulter - matthews planar function , it is possible that there exists an onan configuration containing @xmath14 ( it can be verified by using magma for @xmath469 and @xmath470 ) . from lemma [ le : onan_infty ] , it implies that there exist three elements @xmath471 belonging to two cycles .
that means we can not extend the design formed by all circles into an inversive plane .
a _ correlation _ @xmath472 of a projective plane @xmath32 is a one - to - one map of the points onto the lines and the lines onto the points such that the point @xmath473 is on the line @xmath474 if and only if @xmath475 is on @xmath476 .
a _ polarity _ is a correlation of order two . for each polarity @xmath472 on a projective plane @xmath32 , a point @xmath473
is called _ absolute _ if @xmath473 is on the line @xmath476 .
when the plane @xmath32 is of order @xmath0 , if the polarity @xmath472 has @xmath34 absolute points , then @xmath472 is a _ unitary _ polarity . for each unitary polarity of @xmath32 ,
the set of absolute points and non - absolute lines forms a unital ; see ( * ? ? ?
* theorem 12.12 ) .
[ le : polarity ] let @xmath39 be a planar function on @xmath64 and @xmath477 ( also denoted by @xmath478 for convenience ) an additive map on @xmath64 such that 1 .
@xmath477 is involutionary , i.e. @xmath479 , 2 .
@xmath39 and @xmath477 are commutative under the composition , i.e. @xmath480 , 3 .
@xmath481 for each @xmath81 .
let @xmath472 be a map from the points of @xmath38 to its line sets defined by : @xmath482 then @xmath472 induces a unitary polarity on @xmath38 and the corresponding unital is @xmath483 first we show that @xmath472 induces a polarity .
let us look at the points on @xmath484 , which are @xmath485 . under @xmath472 ,
they are mapped to @xmath486 under the assumption ( a ) and ( b ) , the line @xmath487 .
all of these lines intersect in the point @xmath235 .
hence @xmath488 .
similarly , we can prove that @xmath489 and @xmath490 .
therefore , @xmath472 defines a polarity on @xmath38 .
it is not difficult to check that @xmath14 is an absolute point of @xmath472 and for each @xmath246 the point @xmath42 is not absolute . for each affine point @xmath235
, it is absolute if and only if @xmath235 is on @xmath491 , i.e. @xmath492 hence the set defined by is the set of absolute points of @xmath472 . from the assumption ( c ) , we know that there are totally @xmath493 points in @xmath33 .
therefore , @xmath33 is a unital .
when @xmath38 is a commutative semifield plane of order @xmath0 , i.e. @xmath39 can be written as a dembowski - ostrom polynomial over @xmath64 , the unitals derived from unitary polarities on @xmath38 are intensively investigated by ganley in @xcite . for the coulter - matthews planes ,
the unitary polarities and derived unitals are considered by knarr and stroppel in @xcite .
for the power planar functions ( see theorem [ th : many.planar.functions.1 ] ) , we take @xmath494 for @xmath81 ; for the dickson s semifields and the one constructed in @xcite ( see theorem [ th : many.planar.functions ] ( a ) and ( b ) ) , we set @xmath495 for @xmath496 .
it is readily to verify that for these two maps @xmath477 , the three assumptions in lemma [ le : polarity ] are satisfied .
recall that two unitals @xmath205 and @xmath206 in a projective plane @xmath32 are equivalent if there is a collineation of @xmath32 mapping @xmath205 to @xmath206 .
one of the goals of this section is to prove the following theorem .
[ th : inequivalence_unitals_semifields ] let @xmath39 be a dembowski - ostrom polynomial over @xmath64 such that @xmath497 is a planar function on @xmath64 .
let @xmath477 be a map on @xmath64 satisfying all the assumptions in lemma [ le : polarity ] and @xmath33 the unital derived from @xmath477 .
let @xmath86 be a unital defined in proposition [ prop : general ] .
then @xmath33 and @xmath86 are not equivalent .
to prove theorem [ th : inequivalence_unitals_semifields ] , we proceed to consider the collineation group of @xmath38 . slightly different from lemma
[ le : onan_without_infty ] , we define @xmath498 .
it is readily to verify that @xmath499 if @xmath39 is a dembowski - ostrom polynomial .
[ le : translation+shear_group ] let @xmath500 .
let @xmath39 be a dembowski - ostrom polynomial over @xmath64 which defines a planar function .
let the map @xmath501 on the points of @xmath38 be defined by @xmath502 for each @xmath460 . then the following statements hold . 1 .
all @xmath501 form a collineation group @xmath503 of order @xmath504 . for @xmath505 ,
@xmath506 2 .
the shift group @xmath507 .
3 . @xmath508 is the translation group of @xmath38 .
@xmath509 is the group of elations with the axis @xmath510 and the center @xmath14 . for any @xmath511 and @xmath81 , @xmath512 which equals @xmath513 . together with @xmath514
, we see that the line @xmath71 is mapped to @xmath515 under @xmath501 .
similarly , we can show that @xmath13 is fixed and @xmath217 is mapped to @xmath516 .
hence @xmath501 is a collineation .
moreover , it is routine to verify and that all @xmath501 form a collineation group @xmath503 . from ( a ) and @xmath517
, it is not difficult to derive ( b ) , ( c ) and ( d ) .
the following lemma can be found in ( * ? ? ?
* lemma 8.5 ) .
[ le : normal_sigma ] with the notation in lemma [ le : translation+shear_group ] , @xmath503 is a normal subgroup of @xmath518 .
now let us prove theorem [ th : inequivalence_unitals_semifields ] .
let us first look at the elements in @xmath503 fixing @xmath86 . if @xmath501 fixes @xmath86 , then @xmath519 .
otherwise , for arbitrary @xmath520 and @xmath521 , there is always a unique solution @xmath75 such that @xmath522 it implies that @xmath523 when we choose @xmath524 .
furthermore , it is also readily to verify that @xmath525 fixes @xmath86 if and only if @xmath526 .
therefore , the subgroup in @xmath503 fixing @xmath86 is @xmath527 which is abelian .
second , we consider the elements in @xmath503 fixing @xmath33 .
as there is only the point @xmath14 on @xmath13 which is fixed by @xmath503 , we only have to consider the affine points in @xmath33 .
let @xmath235 be an arbitrary point in @xmath33 , which means that @xmath528 .
the point @xmath529 is still in @xmath33 , which implies that @xmath530 from the above equation , as well as @xmath531 and @xmath532 we deduce that @xmath533 as holds for arbitrary @xmath81 ( see the assumption ( c ) in lemma [ le : polarity ] ) , we have @xmath534 hence the subgroup in @xmath503 fixing @xmath33 is @xmath535 from the assumption ( c ) in lemma [ le : polarity ] , there are totally @xmath536 elements in @xmath537 . by lemma [ le : translation+shear_group ] ( a ) , we see that @xmath537 is not abelian
. assume to the contrary that @xmath86 is equivalent to @xmath33 , i.e. there exists @xmath538 such that @xmath539 . as @xmath503 is normal in @xmath518 , the two subgroups @xmath540 and @xmath537 are conjugate , which contradicts the fact that @xmath540 is abelian but @xmath537 is not .
clearly theorem [ th : inequivalence_unitals_semifields ] does not cover the coulter - matthews planar functions . to prove a similar result for the inequivalence between the two unitals @xmath33 and @xmath86 , we need the following result by dempwolff and rder in @xcite .
[ th : aut.cm.plane ] let @xmath541 be a coulter - matthews plane of order @xmath542 .
then its automorphism group is @xmath543 where @xmath544 and @xmath545 . in theorem
[ th : aut.cm.plane ] , the group @xmath546 is the shift group ( see section [ sec : construction ] )
. the definition of @xmath214 can be found in proposition [ prop : unital_powerplanar_orbit ] .
using theorem [ th : aut.cm.plane ] , we can prove the following results .
[ th : inequivalence_unitals_cm ] let @xmath547 and @xmath541 a coulter - matthews plane of order @xmath542 .
the subgroups of @xmath546 respectively fixing @xmath33 and @xmath86 are of size @xmath548 and @xmath549
. therefore @xmath33 and @xmath86 are inequivalent .
similarly as in the proof of theorem [ th : inequivalence_unitals_semifields ] , @xmath550 fixes @xmath86 if and only if @xmath551 for some @xmath552 .
we denote the group of all these @xmath549 maps @xmath550 by @xmath540 , which is also abelian .
it is routine to show that @xmath550 fixes @xmath33 if and only if @xmath553 and @xmath554 .
hence there are totally @xmath548 such elements in @xmath546 .
they form a group denoted by @xmath537 .
as @xmath546 is a normal subgroup of @xmath555 , if @xmath33 can be mapped to @xmath86 under a collineation of @xmath541 , then @xmath537 must be conjugate to @xmath540 which leads to a contradiction . to conclude this section
, we proceed to consider the isomorphism between the classical unitals ( hermitian curves ) @xmath556 in @xmath37 and @xmath86 in any shift plane @xmath38 .
we use @xmath557 to denote the automorphism groups of the classical unitals @xmath556 as designs . in @xcite
, onan proved that @xmath558 which is transitive on the point sets of @xmath556 .
let @xmath209 and @xmath113 is such that @xmath450 .
the order of @xmath557 is @xmath559 ( ( * ? ? ?
* page 496 and 503 ) ) . in (
* theorem 1 ) , baker and ebert proved that for a given point @xmath14 in @xmath556 , the collineation group @xmath1 of @xmath203 fixing @xmath556 and @xmath14 is of order @xmath560 .
hence @xmath1 is exactly the stabilizer of @xmath14 in @xmath557 . in @xmath1
, there is a non - abelian subgroup @xmath561 of order @xmath536 and a subgroup @xmath562 corresponding to the field automorphism group of @xmath563 , @xmath562 and @xmath561 intersect at the identity element of @xmath1 and @xmath562 normalize @xmath561 ; see @xcite and ( * ? ? ?
* page 70 ) . from the proofs of theorems [
th : inequivalence_unitals_semifields ] and [ th : inequivalence_unitals_cm ] , we know that in the stabilizer of @xmath14 in the group @xmath464 , there is an abelian subgroup @xmath540 of order @xmath536 .
that means if there is no abelian subgroup of order @xmath536 in @xmath1 , then @xmath556 and @xmath86 can never be isomorphic as designs .
when @xmath564 , it is clear that the non - abelian group @xmath561 is a sylow @xmath112-subgroup of @xmath1 . by sylow theorems , all subgroups of order @xmath536 in @xmath1
are non - abelian .
when @xmath565 , it is also not difficult to follow the calculations in ( * ? ?
* page 70 ) to verify that there is no abelian subgroups of order @xmath536 in @xmath566 .
hence there is also no abelian subgroups of order @xmath536 in @xmath1 .
therefore , we have proved the following result .
let @xmath39 be a planar function on @xmath64 . if @xmath39 can be written as a dembowski - ostrom polynomial over @xmath64 or @xmath39 is the coulter - matthews function .
as designs , the unitals @xmath86 defined by in the plane @xmath38 and the classical unitals @xmath556 are not isomorphic .
the authors would like to thank alexander pott for his valuable comments and suggestions .
this work is supported by the research project of miur ( italian office for university and research ) strutture geometriche , combinatoria e loro applicazioni " 2012 .
l. budaghyan and t. helleseth .
new perfect nonlinear multinomials over @xmath568 for any odd prime @xmath112 . in _
seta 08 : proceedings of the 5th international conference on sequences and their applications _
, page 403414 , berlin , heidelberg , 2008 .
springer - verlag .
d. ghinelli and d. jungnickel .
finite projective planes with a large abelian group . in
_ surveys in combinatorics , 2003 ( bangor ) _ , volume 307 of _ london math .
lecture note ser .
_ , page 175237 .
cambridge univ . press ,
cambridge , 2003 .
a. pott , k .- u .
schmidt , and y. zhou .
semifields , relative difference sets , and bent functions . in h.
niederreiter , a. ostafe , d. panario , and a. winterhof , editors , _ algebraic curves and finite fields , cryptography and other applications_. de gruyter , 2014 . | a finite shift plane can be equivalently defined via abelian relative difference sets as well as planar functions . in this paper , we present a generic way to construct unitals in finite shift planes of odd orders @xmath0 .
we investigate various geometric and combinatorial properties of them , such as the self - duality , the existences of onan configurations , the wilbrink s conditions , the designs formed by circles and so on .
we also show that our unitals are inequivalent to the unitals derived from unitary polarities in the same shift planes . as designs ,
our unitals are also not isomorphic to the classical unitals ( the hermitian curves ) . |
understanding the form and evolution of the relationship between galaxy stellar mass , galaxy color , and dark matter halo mass has become a critical topic in galaxy formation . in @xcite ( hereafter l12 )
we combined measurements of the galaxy stellar mass function ( smf ) , galaxy clustering , and galaxy - galaxy lensing in the cosmos survey ( @xcite ) to place constraints on the stellar - to - halo mass relation ( shmr ) at @xmath7 using a halo occupation analysis ( hod ) . in this paper , we focus on the shmr for massive galaxies , @xmath8 , within group - scale halos , @xmath9 , across this same redshift range .
updating the l12 results , we now separately constrain the shmr s for galaxies that are actively star - forming and those that are passively evolving .
we compare these results with a sample of central galaxies identified in an x - ray selected cosmos group catalog ( @xcite ) .
we define a dark matter halo with as having an overdensity 200 times the mean cosmic density .
all calculations assume a flat cosmology of ( @xmath10,@xmath11,@xmath12,@xmath13,@xmath14 .
the cosmos sample that we use for clustering , lensing , and smfs has already been described in detail in l12 .
hod anlaysis is performed on these measurements .
the main difference with respect to l12 is that we now divide the sample into star - forming ( sf ) and passive subsamples using the uvj color - color cuts of @xcite .
we use the same stellar masses as l12 .
these have been estimated using the bayesian code of @xcite using a @xcite initial mass function . in our redshift range , there are 12,573 passive and 41,682 sf galaxies in the cosmos sample above our completeness limits .
we also use a cosmos x - ray selected group catalog to select and study central galaxies .
details regarding this group catalog can be found in @xcite and @xcite .
halo masses for these groups were determined in @xcite by calibrating the @xmath15-@xmath16 relation from weak lensing . to ensure a clean sample of groups and centrals , we exclude potentially merging systems , and groups near masked regions or with very few members ( flag_include=1 in @xcite ) .
this sample contains 129 groups out of 211 extended x - ray detections .
we further remove 18 groups with ambiguous identification of a central galaxy , i.e. , when the most massive group galaxy within the nfw - scale radius ( @xcite ) of the halo is not the most massive galaxy within the virial radius ( mmgg_scale_mstar @xmath17 mmgg_r200_mstar ) . at fixed redshift ,
the group catalog constitutes a roughly halo - mass limited sample of dark matter halos .
we divide the data into three redshift bins that span @xmath3 $ ] .
the specific redshift bins are the same as in l12 and are shown in fig .
[ shmr_groups ] .
the mean logarithmic halo mass in each redshift bin is 13.47 , 13.59 , and 13.75 .
we note that the mass calibration of @xcite assumes a halo mass definition of 200 times the critical density .
we have converted these values to our fiducial halo definition by assuming the nfw density profile with a concentration - mass relation given by @xcite .
we then rescale the masses from the 200-critical definition to the 200-mean ( e.g. , @xcite ) .
the central galaxies in our sample are well above the completeness limit for cosmos ( fig . 1 in l12 ) , even for passive galaxies .
we also check for agn contamination , which we will discuss subsequently .
in @xcite , we presented a theoretical framework for modeling combined measurements of the smf , galaxy clustering , and galaxy - galaxy lensing .
this method is a more generalized version of the traditional halo occupation distribution ( see , e.g. , @xcite for a review ) .
our hod method utilizes these three statistical measures to infer the number of galaxies within halos as both a function of halo and galaxy mass . in l12
, we implemented this formalism on stellar - mass defined samples within cosmos .
our analysis constrains the halo occupation of both central and satellite galaxies , but in this paper we focus exclusively on central galaxies within group - scale halos .
we constrain a shmr for both passive and sf central galaxies such that the total number of central galaxies per halo is unity .
this result is obtained independent of the shmr constrained from the group catalog .
in l12 we assumed that every halo has one central galaxy ; here we require that the sum of mean occupation of passive and sf central galaxies is unity . in a companion paper
( j. tinker et al .
, in preparation ) , we present full details of our measurements and our model fits .
our results focus on the relative clustering of groups at @xmath18 mpc . due to the small area of cosmos
, the integral constraint can affect the clustering of objects at our scale of interest ( l12 ) .
however , it will not alter the relative clustering of two samples in the same volume ( l12 ) , which is the quantity of interest here .
fig . [ shmr_groups ] shows the constraints on the shmr for passive and sf galaxies for each redshift bin .
results from l12 for the full stellar - mass limited samples are shown for comparison . at @xmath19
, there is a clear difference between the stellar masses of sf and passive central galaxies in groups of similar halo mass . at @xmath20 ,
the difference is 0.4 dex .
this is qualitatively consistent with the trends seen in aegis groups at lower halo mass ( @xcite ) . at lower redshift
, however , this difference gradually goes away . in the lowest redshift bin ,
the shmr for sf and passive galaxies cross at fixed halo mass .
this evolution is confirmed in the galaxy group sample : at high redshift , there is a 3-@xmath21 difference in the median central galaxy mass between passive and sf centrals .
errors for this quantity are calculated by bootstrap resampling of the stellar masses within each subsample .
the median galaxy masses are also in good agreement with those derived from the halo occupation analysis . at lower redshifts ,
the difference in the passive and sf galaxy masses gets monotonically smaller .
the median masses do not cross over , as they do in the shmrs ; there is a discrepancy between the results for the passive subsample in the lowest redshift bin .
however , results from sdss demonstrate that this cross over has indeed occurred by @xmath5 ( @xcite ) .
the results from the group catalog are qualitatively similar if one breaks the catalog up by morphology ( as shown in figure [ shmr_groups ] ) .
the galaxies with x - ray agn activity , either in the xmm or chandra observations are indicated on the plot .
the low number of such objects , and the ( lack of ) correlation with star formation estimates indicates that agn contamination is not playing any role in the observations .
removing these objects from the sample does not shift the medians beyond their 1-@xmath21 errors . at @xmath22 $ ]
, the discrepancy in the shmr values and those obtained from the groups for passive galaxies is a 2.4-@xmath21 difference based upon creating monte carlo samples of halos using all elements in the mcmc chain but with the same mass distribution as the groups sample .
the large - scale clustering amplitude of all structure in the low-@xmath23 bin is below average ( l12 ) , which could drive systematic errors in the hod results .
it is also possible that the halo mass function assumed in the hod analysis ( @xcite ) is not the same as the true mass function in that patch of sky , also resulting in systematic biases .
however , we note that the groups catalog is a subset of all x - ray groups within cosmos , while the shmrs are derived from statistics on the full sample of galaxies .
also , @xmath5 measurements of the shmr using lensing ( @xcite ) and satellite kinematics ( @xcite ) find that the halo masses for massive red galaxies are higher than those of massive sf central galaxies , following the evolutionary trend seen in the hod results .
[ wtheta_groups ] shows the cross - correlation between groups and all galaxies .
we split the groups into samples with sf and passive centrals , and cross - correlate each set of centrals with all galaxies .
both samples are in the z=[0.74 , 1.00 ] redshift bin , with a magnitude cut of @xmath24 .
galaxies brighter than this threshold have reliable photometric redshifts with errors @xmath25 ( see figure 2 in @xcite ) .
because most of the redshifts of the central group galaxies are known with spectroscopic precision ( most are sampled within the zcosmos survey ) , we can measure the real - space projected clustering , which has higher signal - to - noise relative to a simple angular cross - correlation .
we denote this clustering statistic @xmath26 .
details are given in @xcite .
briefly , we calculate the projected comoving separation between each group - galaxy pair from the angular separation and the redshift of the group .
we restrict all pairs to lie within a redshift interval of @xmath27 , or 3-@xmath21 of the photo - z error . to properly normalize @xmath26 requires detailed information of the photo - z error distribution function , but since we are only concerned with the relative clustering between two spectroscopic samples cross - correlated with the same photometric sample , this step is unnecessary .
we measure the angular cross - correlation , @xmath28 , for the same samples as a cross - check on our results .
errors are obtained by bootstrap resampling of the groups and recalculating @xmath26 or @xmath29 for each bootstrap sample .
a scale of importance is the 1 mpc scale ( comoving ) , roughly the virial radius of the groups . inside this scale ,
the cross - correlation probes the number of satellite galaxies within the groups . outside this scale , the cross - correlation probes the large - scale bias of the groups , which is an indicator of their environment . in the measurements of fig .
[ wtheta_groups ] , this scale marks the bifurcation in the two correlation functions . outside this scale ,
the passive - central groups have a lower large - scale bias , indicating that these halos have formed in lower - density environments . inside this scale ,
the correlation functions for the groups differ at the @xmath30 level , with the sf - centered groups having lower clustering , but the large errors prevent meaningful interpretation . for each redshift bin , we calculate the bias relative to an ( arbitrarily normalized ) non - linear matter correlation function calculated using the @xcite fitting function .
we calculate bias using bins at @xmath31 mpc or @xmath32 arcsec .
we estimate the covariance matrix by bootstrap resampling of the groups with replacement .
we use 200 bootstrap samples . due to the low number of groups ,
the clustering signal around each group can be considered independent .
we use the full covariance matrix to obtain the bias and its error .
the relative bias of sf - centered and passive - centered groups is shown as a function of redshift in fig .
[ relative_bias ] .
we show bias measurements from both @xmath26 and @xmath28 . for the former ,
a bias measurement is not possible at @xmath33 due to noise in the @xmath26 measurements for both subsamples . at @xmath34 ,
both measures indicate that the sf - centered groups have significantly enhanced clustering . at @xmath35 ,
the relative bias is above unity , but this detection is not significant given the errors . at @xmath33 ,
the angular clustering yields a 1-@xmath21 detection of elevated clustering in the sf - centered groups .
proposed quenching mechanisms for massive galaxies , such as major mergers ( i.e. , @xcite ) or agn ( i.e. , @xcite ) essentially remove galaxies from the star - forming sequence and place them in the red sequence .
the efficiency of such processes is correlated with halo mass .
some research suggests that cooling flows may lead to episodic star formation in central galaxies ( @xcite ) . the substantial difference in the shmrs for sf and passive galaxies at the group - scale halo masses at @xmath4
has many implications .
it implies that star formation is not stochastic in these objects : if massive central galaxies underwent periodic episodes of star formation followed by longer - term quiescence , the galaxies at fixed halo mass would have the same mass regardless of color .
the results also imply that massive quenched galaxies had far different growth histories than those that are forming stars at @xmath4 .
a scenario in which galaxies at fixed halo mass grow on a common star - forming sequence , with a quenching mechanism that removes these galaxies from this sequence , would make passive central galaxies _ less _ massive than sf central galaxies .
this is the opposite of what is observed at @xmath4 . to be consistent with our observations , passive central galaxies at @xmath4 form their stars rapidly at high redshift , essentially getting ` ahead of the growth curve ' relative to central galaxies that are still forming stars by @xmath4 . at high redshift
, central galaxies essentially `` knew '' they would be quenched by @xmath4 .
[ wtheta_groups ] indicates that color - selected groups represent special subsets of objects at this halo mass scale .
the current growth rate ( indicated by galaxy color ) and growth history ( probed by total stellar mass ) of the central galaxy is correlated with large - scale environment .
a similar effect is seen in dark matter halos in n - body simulations , an effect called assembly bias . for massive systems ,
younger halos exist in more dense environments ( @xcite ) .
the environment ( and formation history ) of massive halos is also correlated with angular momentum of dark matter halos ( @xcite ) such that high - spin halos are more clustered than low - spin halos .
this effect goes away below @xmath36 .
the shaded region in fig .
[ relative_bias ] is the numerical result from @xcite .
the lower limit is the bias of the top 20% of halos , ranked by angular momentum , relative to all halos .
the upper limit is the bias of the top 20% of halos relative to the lowest 20% of halos ( see their fig .
20 , thus we convert their @xmath5 results , which are plotted as a function of @xmath37 , to @xmath38 and interpolate the assembly bias at the values of @xmath38 for the groups samples at each redshift . ] ) .
the proper comparison to our measurements will lie in between .
recent hydrodynamic simulations of galaxy formation indicate that galaxy morphology is correlated with the angular momentum gained from the larger - scale environment around the halo at early epochs ( @xcite ) .
if the angular momentum of dark matter and baryons are connected , massive halos with central disk galaxies should have enhanced clustering , in agreement with the results in fig .
[ wtheta_groups ] .
@xcite also find that the disk galaxy masses are lower than their spheroidal counterparts .
there are , however , substantive differences between the sales simulations and our results : they show results for less - massive halos at @xmath5 , rather than group - scale systems at @xmath4 .
they conclude that there is little correlation between morphology and @xmath5 halo spin , but there is little correlation between halo spin at early and late epochs for their halo masses ( @xcite ) .
however , the existence of assembly bias implies that more massive halos retain memory of the angular momentum at the epoch of galaxy formation .
further investigation is required at higher masses and redshifts .
however , one need not invoke angular momentum to achieve both the relative clustering and relative masses of passive and sf centrals .
@xcite demonstrate that stellar mass growth peaks at a halo mass of @xmath36 , weakly dependent on redshift , but the star formation efficiency at that peak decreases with cosmic time . in this scenario , central galaxies within late - forming halos would lag behind those in early - forming halos , and have enhanced clustering .
this toy model does not explain the morphology dependence , or the difference in instantaneous sf rates at @xmath4 , but does provide a connection between halo formation history and galaxy properties .
there have been many attempts to find assembly bias in the @xmath5 galaxy distribution . @xcite and @xcite find no evidence for assembly bias for galaxies below the knee in the smf or luminosity function .
the assembly biases in the low mass and high mass halo populations are driven by disparate physical mechanisms .
younger halos form in denser environments at high mass through the statistics of gaussian random fields . at low mass ,
_ older _ halos form in denser environments due to tidal forces and interactions with nearby massive objects ( @xcite ) .
it is plausible that these two mechanisms may have different levels of impact on galaxy formation .
@xcite find that @xmath5 group - mass halos in sdss with redder total galaxy content ( centrals and satellites combined ) are more clustered than groups with bluer galaxies .
it is not clear how the clustering signal in @xmath4 cosmos data could reverse if the most - clustered halos at one redshift remain the most clustered at a lower redshift .
the @xcite detection is mitigated by the lack of independent constraints on the halo mass ; in their group - finding algorithm , the halo mass is estimated statistically by assuming a 1:1 correspondence between total group stellar mass and halo mass , with no scatter .
@xcite , using a different group - finding algorithm ( but the same data set ) , find the opposite signal : groups with bluer central galaxies are the ones that are more clustered .
both these methods rely on inferring halo mass statistically from the galaxies within them ; our x - ray detections and lensing masses are more legitimate for detecting assembly bias . from @xmath4 to @xmath5 , the shmrs evolve quite differently depending on star formation activity . by @xmath33 ,
the mean relations have crossed and passive central galaxies live in higher mass halos than sf central galaxies at fixed mass .
this inversion is also consistent with results from @xmath5 studies ( @xcite ) .
star forming galaxies grow by a factor of @xmath12 using the star formation rates of @xcite from @xmath34 to @xmath33 .
group - mass halos also grow by a factor of @xmath12 , thus central galaxies grow as fast as their host halos . for quenched galaxies ,
their growth rates are slower than that of their host halos , plausibly causing the inversion of the shmr seen in fig .
[ shmr_groups ] . at @xmath39 ,
our results imply that the process that shuts down star formation in massive galaxies can not be explained by a stochastic process that is a function of halo mass .
rather , the interplay between the dark matter halo and the surrounding environment , including the tidal fielde , strongly influences the fate of the galaxy forming within it .
we thank charlie conroy and tom theuns for useful discussions .
this work was supported by world premier international research center initiative ( wpi initiative ) , mext , japan .
the initial hst - cosmos treasury program was supported through nasa grant hst - go-09822 .
we gratefully acknowledge contributions of the entire cosmos collaboration consisting of more than 140 scientists .
more information on cosmos is available at * http://cosmos.astro.caltech.edu/ * and the data archive is at ipac / irsa .
jr was supported by jpl , run under contract for nasa by caltech . , k. , scarlata , c. , carollo , c. m. , ellis , r. s. , drory , n. , hopkins , p. , salvato , m. , leauthaud , a. , koekemoer , a. m. , murray , n. , ilbert , o. , oesch , p. , ma , c .- p . , capak , p. , pozzetti , l. , & scoville , n. 2010 , , 719 , 1969 , a. , guzzo , l. , hasinger , g. , scoville , n. z. , aussel , h. , bhringer , h. , brusa , m. , capak , p. , cappelluti , n. , comastri , a. , giodini , s. , griffiths , r. e. , impey , c. , koekemoer , a. m. , kneib , j .-
p . , leauthaud , a. , le fvre , o. , lilly , s. , mainieri , v. , massey , r. , mccracken , h. j. , mobasher , b. , murayama , t. , peacock , j. a. , sakelliou , i. , schinnerer , e. , silverman , j. d. , smoli , v. , taniguchi , y. , tasca , l. , taylor , j. e. , trump , j. r. , & zamorani , g. 2007 , , 172 , 182 , m. r. , leauthaud , a. , bundy , k. , finoguenov , a. , tinker , j. , lin , y .- t . , mei , s. , kneib , j .-
p . , aussel , h. , behroozi , p. s. , busha , m. t. , capak , p. , coccato , l. , covone , g. , faure , c. , fiorenza , s. l. , ilbert , o. , le floch , e. , koekemoer , a. m. , tanaka , m. , wechsler , r. h. , & wolk , m. 2011 , , 742 , 125 , a. , finoguenov , a. , kneib , j .-
p . , taylor , j. e. , massey , r. , rhodes , j. , ilbert , o. , bundy , k. , tinker , j. , george , m. r. , capak , p. , koekemoer , a. m. , johnston , d. e. , zhang , y .- y . , cappelluti , n. , ellis , r. s. , elvis , m. , giodini , s. , heymans , c. , le fvre , o. , lilly , s. , mccracken , h. j. , mellier , y. , rfrgier , a. , salvato , m. , scoville , n. , smoot , g. , tanaka , m. , van waerbeke , l. , & wolk , m. 2010 , , 709 , 97 , a. , tinker , j. , bundy , k. , behroozi , p. s. , massey , r. , rhodes , j. , george , m. r. , kneib , j .- p . , benson , a. , wechsler , r. h. , busha , m. t. , capak , p. , corts , m. , ilbert , o. , koekemoer , a. m. , le fvre , o. , lilly , s. , mccracken , h. j. , salvato , m. , schrabback , t. , scoville , n. , smith , t. , & taylor , j. e. 2012 , , 744 , 159 , k. g. , faber , s. m. , weiner , b. j. , koo , d. c. , primack , j. r. , dekel , a. , papovich , c. , conselice , c. j. , le floch , e. , rieke , g. h. , coil , a. l. , lotz , j. m. , somerville , r. s. , & bundy , k. 2007 , , 660 , l47 , n. , aussel , h. , brusa , m. , capak , p. , carollo , c. m. , elvis , m. , giavalisco , m. , guzzo , l. , hasinger , g. , impey , c. , kneib , j .-
p . , lefevre , o. , lilly , s. j. , mobasher , b. , renzini , a. , rich , r. m. , sanders , d. b. , schinnerer , e. , schminovich , d. , shopbell , p. , taniguchi , y. , & tyson , n. d. 2007 , , 172 , 1 , j. , dekel , a. , faber , s. m. , noeske , k. , koo , d. c. , gerke , b. f. , cooper , m. c. , salim , s. , dutton , a. a. , newman , j. , weiner , b. j. , bundy , k. , willmer , c. n. a. , davis , m. , & yan , r. 2012 , , submitted ( arxiv:1203.1625 ) | using observations in the cosmos field , we report an intriguing correlation between the star formation activity of massive ( @xmath0 ) central galaxies , their stellar masses , and the large - scale ( @xmath110 mpc ) environments of their group - mass ( @xmath2 ) dark matter halos . probing the redshift range @xmath3 $ ] , our measurements come from two independent sources : an x - ray detected group catalog and constraints on the stellar - to - halo mass relation derived from a combination of clustering and weak lensing statistics . at @xmath4 , we find that the stellar mass in star - forming centrals is a factor of two less than in passive centrals at the same halo mass .
this implies that the presence or lack of star formation in group - scale centrals can not be a stochastic process .
by @xmath5 , the offset reverses , probably as a result of the different growth rates of these objects .
a similar but weaker trend is observed when dividing the sample by morphology rather than star formation .
remarkably , we find that star - forming centrals at @xmath6 live in groups that are significantly more clustered on 10 mpc scales than similar mass groups hosting passive centrals .
we discuss this signal in the context of halo assembly and recent simulations , suggesting that star - forming centrals prefer halos with higher angular momentum and/or formation histories with more recent growth ; such halos are known to evolve in denser large - scale environments .
if confirmed , this would be evidence of an early established link between the assembly history of halos on large scales and the future properties of the galaxies that form inside them . |
blazars represent the most active side of active galactic nuclei due to a priviledged aspect angle enhancing the emission from the jet with respect to that from the accretion disk .
the strongly variable non - thermal continuum , extending from the radio to the high - energy @xmath7rays , produced within the jet ( with bulk lorentz factors @xmath8 ) is strongly amplified in the forward direction by relativistic beaming .
the short variability timescale ( down to hours ) indicates that the emission region must be located quite close to the central engine , at distances of the order of 0.010.1 pc .
therefore , studying emission from blazars we can probe the physical state of the jet close to its base , allowing important insights on the physical mechanism able to produce such extreme structures . the double - humped spectral energy distribution displayed by blazars ( sambruna et al .
1996 , fossati et al .
1998 ) is generally described in terms of synchrotron ( contributing from the radio to the uv - x - ray range ) and inverse compton ( extending from x - rays to @xmath6-rays ) emission from a single population of relativistic electrons ( e.g. , ghisellini et al .
the peaks of the two components show a clear trend with the total emitted power ( fossati et al .
1998 ) . for low power sources
( bolometric luminosity @xmath9 erg / s ) the synchrotron peak falls in the uv - x - rays and the ic peak in the multi gev - tev range . in high power sources ( bolometric luminosity @xmath10 erg / s ) ,
the synchrotron peak is located in the submm - fir band , while the ic component , usually dominating the total radiative output , peaks in the mev - gev region . thus in low - luminosity blazars ( mostly bl lac objects ) x - rays probe the high energy end of the synchrotron component allowing the investigation of the dynamics of high - energy electrons , while for powerful blazars ( belonging to the fsrq class ) the x - ray emission is produced via the ic process by relatively low energy electrons .
the study of the x - ray portion of the seds of fsrq is crucial for different reasons .
first of all , the hard x - ray spectrum of fsrq is generally attributed to the ic radiation produced by low - energy electrons scattering ambient photons ( external compton emission ) .
its investigation offers the only possibility to probe directly the low energy portion of the electron distribution which is essential for estimating the total power of the jet ( e.g. , maraschi & tavecchio 2003 ) and is unaccessible at radio frequencies because of self - absorption .
moreover , the slope of the x - ray continuum , directly connected to the slope of the underlying population of low - energy electrons , can provide important constraints on the acceleration mechanisms acting on the relativistic electrons .
a good description of the continuum requires broad band observations ( 0.1 - 100 kev ) .
unfortunately , few fsrqs have high statistics above 10 kev , most of them obtained in the past decade with the pds instruments onboard @xmath11sax ( e.g. , tavecchio et al . 2000 , 2002 ) .
the joint fit of the medium ( 2 - 10 kev ) and the hard ( above 10 kev ) x - ray data revealed in same cases the presence of very hard spectra , with slopes @xmath12 above 23 kev ( here and in the following @xmath13 is defined as @xmath14 ) .
these results have been recently extended using observations with _ swift _ ( sambruna et al . 2006a , 2007 ) ,
although in these cases the data in the hard x - rays represent averages over months .
the presence of such hard spectra is difficult to reconcile with the standard picture considering the formation of power law distributions through shock acceleration or cooling of high energy electrons ( e.g. , sikora et al .
2002 ) .
the medium - soft x - ray spectrum may reveal the very low end of the ec component and therefore of the relativistic electron spectrum . around these energies
a `` break '' is expected in the ec emission which is the `` footprint '' of the thermal photon population . while above the break the slope of the spectrum is connected to the slope of the electron distribution , below the break the slope of the spectrum
should instead reveal the spectral distribution of the ambient photons .
similar `` convex '' spectra have in fact been observed in some intermediate redshift fsrqs ( e.g. , tavecchio et al .
2000 , 2002 ) .
an important consequence of the detection of such a break is the determination of the mimimum lorentz factor of the emitting electrons , a key parameter for the determination of the kinetic power carried by the jet ( e.g. , celotti & fabian 1993 ) .
however , the situation can be more complex , with the simultaneous presence in this band of different components , such as the synchrotron self - compton emission and , as recently pointed out by celotti , ghisellini & fabian ( 2007 ) , the spectral components originating from the `` bulk comptonization '' of ambient photons by _ cold _ electrons in the jet .
the interpretation of the shape of the soft x - ray continuum is not unique .
in fact , in several high - redshift blazars ( @xmath15 ; e.g. , yuan et al .
2006 and references therein ) the steepening of the soft x - ray spectrum is interpreted in terms of absorption by a dense region of ( warm ) plasma ( @xmath2 @xmath3 ) present in the region surrounding the qso ( expected to be in the form of a wind / outflow , e.g. , fabian 1999 ) .
if confirmed , the presence of this large amount of absorbing gas in the vicinity of the jet , would have strong consequences for the structure of the jet and its interaction with the ism .
an excellent laboratory to address most of the issues discussed above is the fsrqs rbs 315 ( from the rosat bright survey , schwope et al .
2000 ; also known as gb6 j0225 + 1846 , @xmath0 ) , for which a previous _ xmm - newton _ observation revealed the presence of an extremely hard x - ray continuum ( photon index @xmath16 , page et al .
2005 , piconcelli & guainazzi 2005 ) and a strong deficit of soft photons . with the aim of obtaining a better description of the broad band x - ray continuum we asked for a observation . in the same days of the observation we obtained a target of opportunity short pointing ( @xmath17 s ) with _
swift_. in the following we report the data analysis ( sec.2 ) , the results ( sec.3 ) and the modelling of the derived seds ( sec.4 ) .
the discussion is given in sec.5 , where we explore in particular the possibility that the convex spectrum reflects the _ intrinsic _ curvature of the ic component .
conclusions are reported in sec.6 . throughout this work we use the following cosmological parameters : @xmath18 , @xmath19 , @xmath20 .
the payload ( mitsuda et al . 2007 ) carries four x - ray telescopes sensitive in the 0.3 - 12 kev band ( xis , koyama et al .
2007 ) , with ccd cameras in the focal plane , together with a non - imaging instrument ( hxd , takahashi et al . 2007 ) , sensitive in the 10 - 600 kev band , composed by a si - pin photo - diodes detector ( probing the 10 - 60 kev band ) and a gso scintillator detector ( sensitive above 30 kev ) .
three xis units ( xis0 , 2 and 3 ) have front - illuminated ccds , while xis1 uses a back - illuminated ccd , more sensitive at low energies .
rbs 315 was observed with from 2006 july 25 19:36:03 ut to 2006 july 27 00:58:19 ut ( sequence number 701077010 ) .
the total on - source time was 49.3 ksec . to increase the sensitivity of the hxd
the observation was performed at the hxd aim - point .
hxd / gso data are not used in the following analysis , since the performances and the background of the gso at the flux level expected for rbs 315 are still under study .
the analysis have been performed with the data obtained through the last version of the processing ( v1.2 ) and the last release of the heasoft sofware ( v6.1.2 ) and calibrations .
a more extended discussion of the procedure used can be found in kataoka et al .
no significant variability is detected within the observations ( see fig.[lcurve ] ) , therefore spectra were extracted over the whole exposure .
the reduction followed the prescriptions suggested in `` the suzaku data reduction guide '' . using the heasoft tool xselect we select good time intervals , excluding epochs of high background ( when the satellite crosses the south atlantic anomaly or the object is too close to the rim of the earth ) .
after screening the net exposure time is 35.4 ksec .
events are then extracted in a circle centered on the source with a radius of 4.6. background events are extracted in a similar circle centered in a region devoid of sources .
we checked that the use of different source and background regions do not significantly affect the resulting spectra .
response ( rmf ) and auxiliary ( arf ) files are produced using the tools developed by the team ( xisrmfgen and xissimarfgen ) distributed with the last version of heasoft .
arfs are already corrected for the degradation of the xis response using the tool xiscontamicalc .
hxd / pin data are reduced following the procedure suggested by the team .
the hxd / pin spectrum is extracted after the selection of good time intervals ( analogously to the xis procedure ) . to the extracted spectrum ( obtained through xselect )
we applied the suggested dead time correction ( of the order of 5% ) . the net exposure time after
screening is 40.2 ksec .
response and non x - ray background ( nxb ) files are directly provided by the team . for technical reasons , after 2006 may 24 the response of the pin has changed , therefore we used response and background files calculated for this epoch .
note also that , since the background level of hxd / pin is extremely low , the background event files are generated with a ten times scaled level than the actual background to avoid introducing a large statistical error .
the exposure keyword in the background file has to be changed before the analysis .
an important issue in the analysis of the hxd / pin data concerns the estimate of the cosmic x - ray background , whose spectrum peaks just in this band .
we followed the procedure suggested by the team ( see also kataoka et al .
2007 ) , simulating the expected contribution of the cxb from the entire pin field of view ( @xmath21 ) , assuming the _ heao-1 _ spectrum between 3 and 60 kev ( boldt 1987 , gruber et al . 1999 ) : @xmath22 it is worth noting that the normalization of the cxb is still under debate , since the level measured by heao-1 in the 2 - 10 kev band appears to be below ( with a difference of about 20% ) that measured by more recent missions . however , a recent analysis of the data of the pds instrument ( in the band 15 - 50 kev ) onboard _
beppo_sax seems to confirm the heao-1 spectrum , but with a slightly larger normalization ( frontera et al .
for comparison in fig.([bkg ] ) we report the total pin spectrum ( open circles ) , the total background ( cxb+nxb , filled circles ) , the nxb ( asterisks ) and the net pin spectrum resulting from the background subtraction ( open squares ) .
the net counts represent about 10% of the total counts .
roughly , the cxb flux account for 5% for the hxd / pin background .
we have binned the hxd / pin spectra so that each bin contains almost same number of signal photons . with this choice each bin has a s / n ratio of about 4 .
_ swift _ ( gehrels et al . 2004 ) observed rbs 315 two times : on 28/7/2006 from 18:42:40 ut to 20:45:10 ut ( for a net exposure of 4152 s ) and 31/7/2006 from 15:48:12 ut to 16:21:23 ut ( for a net exposure of 1983 ) .
_ xrt _ and _ uvot _ data have been analized as detailed in foschini et al .
( 2007 ) , using the same version of the software and calibration . for _ xrt
_ we use the data in photon counting mode , selecting events with grades 0 - 12 .
we reanalized the _ xmm - newton _ data of the observation of rbs 315 performed on 2003 july 25 , previously reported by page et al .
( 2005 ) and piconcelli & guainazzi ( 2005 ) . for processing , screening and analysis of the data we used the same procedure described in foschini et al .
( 2006a ) , with the latest version ( v.7.0.0 ) of xmm sas and updated calibration files ( 20/7/2006 ) .
pn , mos1 and mos2 instruments were operated in the full - frame mode .
no soft - protons flares occurred during the exposure : the total exposure time was about 20 ksec on each detector ( 18 for pn and 22 for mos ) .
we fit the data with the xspec package ( v.11.3.2 ) .
xis spectra were rebinned with a minimum of 50 counts per bin .
we checked that all the xis units provide consistent results ( differences in the cross - normalizations are of the order of 5% ) : therefore we decided to perform joint fits of all the xis0,1,2 and 3 spectra . to avoid problems with the calibration at low energy , we limit the fit to the range 0.7 - 10 kev for xis0,2,3 and to 0.5 - 10 kev for the xis1 .
we also limit the analysis of the pin spectrum to the range 12 - 60 kev ( see kataoka et al . 2007 ) .
all the errors in the parameters are intended at the 90% for one parameter of interest ( @xmath23 ) . a power law model with absorption fixed to the galactic value ( @xmath24 @xmath3 , dickey & lockman 1990 )
clearly does not provide an adequate fit to the xis data ( tab.[fitxis ] , the data / model ratio is reported in fig.[xispowabsgal ] ) .
if we allow the absorption to vary we obtain a great improvement of the fit ( @xmath25 ) , but the value of @xmath26 is clearly in excess to the galactic one ( @xmath27 @xmath3 ) .
then we try to fit the spectrum assuming galactic absorption and a broken power law model .
the fit is good ( @xmath28 ) and it requires a rather hard spectrum below 1 kev ( @xmath29 ) .
an alternative possibility is to assume that the intrinsic spectrum is well reproduced by a power - law with the required extra - absorption provided by ( neutral ) absorbing material in the quasar environment .
therefore we model the data assuming the zwabs model in xspec ( plus the galactic absorption ) .
again , the best fit is good , with the value of the @xmath30 ( @xmath31 ) almost coincident with that obtained with the broken power - law model .
the value of the required intrinsic absorption is rather large , @xmath32 @xmath3 ( assuming solar abundance ) , close to the value found by _ xmm - newton _
( page et al .
2005 , piconcelli & guainazzi 2005 ) .
we perform joint xis and pin fits ( results in tab.[fitxispin ] ) , allowing a free normalization factor between the xiss and the pin to take into account the uncertainty in the xis / pin cross - normalization ( of the order of 15% at the hxd nominal position ) .
all the best fit parameters are consistent with those obtained from the fit of the xis alone .
the required cross normalization factor is around 20% , compatible with the instrumental uncertainty . in fig.([xispinbknpowabsgal ] ) we show the spectrum and the data / model ratio for the fit with the broken power - law and galactic absorption . due to the short exposure
, the xrt data do not provide strong constraints on the spectral parameters .
fitting the data obtained summing the two observations with a power - law absorbed in the quasar rest frame we derive values consistent with those of : @xmath33 , @xmath34 @xmath3 , @xmath35 erg @xmath3 s@xmath36 with @xmath37 .
we fitted the data in the 0.2 - 10 kev range , after rebinning so that each energy bin contained a minimum of 50 counts .
results are presented in tab.[xmm ] . as already noted by page et al .
( 2005 ) and piconcelli & guainazzi ( 2005 ) , a single power - law and galactic absorption does not provide an adequate fit to the data .
the fit is satisfactory if we allow intrinsic absorption or a broken - power law shape for the continuum ( tab.[xmm ] ) . consistently with the analysis of page et al .
( 2005 ) and piconcelli & guainazzi ( 2005 ) the broken power law model provides a worst fit than the absorbed power - law model . in our case , however , the broken power - law model is still accettable , while piconcelli & guainazzi ( 2005 ) concluded that the broken power law model yelds an unaccettable fit ( with @xmath38 ) .
the difference with our analysis is probably due to the refined calibrations used in the more recent version of the software .
note , moreover , that piconcelli & guainazzi ( 2005 ) used the mos data in the restricted range 0.8 - 10 kev , while the improved calibration allows us to use of the whole range 0.2 - 10 kev . in fig.([contours ] ) we report the confidence levels ( at 68% , 90% and 99% ) for the fit with the power - law absorbed by gas in the quasar rest frame and for the broken power - law model , calculated for our observation and for the _ xmm - newton _ data .
although the required amount of absorbing gas is almost consistent between the two observations , the slope of the power - law is clearly different in the two cases . during the observation
the power law is clearly softer ( @xmath39 ) than the previous _ xmm - newton _ observation ( @xmath40 ) . on the other hand ,
the low energy slope and the break energy of the broken power law are marginally consistent .
in fig.([sedcomp ] ) ( upper panel ) we report the sed of rbs 315 constructed using archival radio data from ned , the optical - uv fluxes obtained with _ uvot _ and data deconvolved for the case of an intrinsically curved x - ray spectrum .
optical data have been corrected for galactic absorption using the maps of schlegel et al .
1998 reporting @xmath41 mag .
if absorption is important in the quasar rest frame , the points should be considered as lower limits to the real flux . for comparison , in the lower panel
we report the sed of 0836 + 710 ( from tavecchio et al .
2000 ) , a well known @xmath6-ray loud blazar ( also characterized by an extremely flat x - ray continuum and showing indications of absorption / hardening at low energy ) , located at comparable distance ( @xmath42 ) .
the similarity between the two seds is remarkable , strongly supporting the classification of rbs 315 as a blazar .
both sources lies at the top of the so - called blazar sequence ( fossati et al .
1998 ) with apparent luminosities in excess of @xmath43 erg / s .
the comparison also suggests that rbs 315 should be quite bright in the @xmath7-ray band , although it was not not detected by _
egret / cgro_. we expect that _ glast _ , with its wide field of view , allowing the complete monitoring of the sky in few hours and higher sensitivity , should easily detect rbs 315 . given the ambiguity on the shape of the intrinsic x - ray continuum , in the following we will discuss the seds for rbs 315 separately for the case of intrinsic curvature ( _ curved , c _ ) , and for the absorbed power law ( _ absorbed , a _ ) .
the x - ray data , unfolded with the two models , are shown in figs.([sedall])-([sedallabs ] ) . in the second case the intrinsic x - ray spectrum emitted by
the jet is a straight power - law we consider both the x - ray state corresponding to the _ xmm - newton _ observation in 2003 and that observed by in 2006 .
the optical - uv points probably trace the so - called `` blue bump '' , i.e. the direct emission from the accretion disk , whose luminosity is then rather large , @xmath44 erg / s .
we reproduced the sed with the standard model for blazars , assuming synchrotron and ic ( both ssc and ec ) emission from relativistic electrons in the jet .
specifically ( see maraschi & tavecchio 2003 for a full description ) we assume a spherical region with radius @xmath45 , moving with bulk lorentz factor @xmath46 at an angle of view @xmath47 with respect to the line of sight ( under these conditions the relativistic doppler factor is @xmath48 ) .
the region is filled by a tangled magnetic field @xmath49 and by relativistic electrons , assumed to follow a smoothed broken power - law energy distribution between the lorentz factors @xmath50 and @xmath51 , with slopes @xmath52 and @xmath53 before and after the break at @xmath54 .
purely phenomenological _ form has been assumed to reproduce the observed shape of the blazar seds , without any specific assumption on the acceleration / cooling mechanism acting on the particles . with this choice
we are allowed to assume extreme low - energy slopes ( @xmath55 ) such as those required for rbs 315 in 2003 , which can not be obtained under standard conditions . if the radiative cooling timescale is shorter than the light crossing time of the source we expect @xmath56 .
we remind that the model can not reproduce the emission in the radio band since the source is opaque in this band due to synchrotron self - absorption .
the radio emission is generally assumed to originate from larger regions of the jet , downstream to the blazar region .
the ic process involves both the synchrotron photons ( synchrotron self - compton , maraschi et al .
1992 ) and ambient photons ( from the disk [ dermer & schlickeiser 1993 ] and/or reprocessed by the broad line region [ sikora et al .
1994 ] ) entering the jet ( external compton ) . in sources like rbs 315 the ec emission dominate the high - energy component , while the ssc component could contribute in the soft x - ray band . in the case in which the ambient photons are dominated by the emission from the blr
clouds it is customary to model the external photon spectrum with a black body shape peaking around @xmath57 hz ( e.g. , ghisellini et al .
although rather crude , the ic spectra derived from this approximation reproduces quite well more refined spectra calculated taking into account the exact shape of the blr spectrum ( tavecchio et al .
, in prep . ) . to calculate the ec spectra we have also to assume the energy density of the blr radiation , @xmath58 , given by : @xmath59 .
unfortunately we could not found useful information on @xmath60 , the luminosity of the emission lines of this source .
therefore we indirectly fix @xmath60 assuming that optical - uv emission directly measure the disk luminosity , @xmath61 , and that a fraction @xmath62 of the disk emission is reprocessed by the blr ( this is consistent with the results obtained for the few sources in which a measure of the disk and line luminosity is simultaneously available , e.g. , sambruna et al .
the luminosity - radius relations for the blr ( kaspi et al .
2005 , bentz et al .
2006 ) extrapolated at these large luminosities would imply quite large blrs , with radii @xmath63 cm .
these large radii would imply small @xmath58 .
however , the validity of these relations in these extreme regimes is not tested . moreover , besides the blr radiation , we may have the radiation coming directly from the accretion disk and/or isotropized by intercloud scattering material .
therefore we assume @xmath64 cm , which implies a relatively large radiation energy density .
the models are reported in figs.([sedall])-([sedallabs ] ) and the model parameters are reported in tab.([parameters ] ) , together with the derived jet radiative luminosity and jet power ( see the discussion ) .
we report the models calculated both in the case of the x - ray spectrum measured by _ xmm - newton _ in 2003 and that derived by in 2006 . in fig .
( [ sedall ] ) we report the models for the _ c _ case ( the insert shows a zoom on the x - ray band ) . the high - energy emission , from x - rays to @xmath7-rays , is dominated by the ec component .
this can be approximately described by three power laws : 1 ) an extremely hard power law below a `` break frequency '' @xmath65 hz ( where @xmath66 hz , @xmath67 , @xmath68 , values typically derived in these sources , e.g. , tavecchio et al .
2000 ) ; 2 ) a power law with spectral index @xmath69 between @xmath70 and @xmath71 ( marking the position of the ec peak ) ;
3 ) a steeper power law with slope @xmath72 up to the frequency @xmath73 .
we remark that the steepening of the spectrum below @xmath70 is _ required _ by the standard emission models for this kind of sources , although the precise shape and position of the break depends on the jet parameters and on the assumed underlying ambient photon population ( tavecchio et al .
, in prep ) . in particular , the slope below @xmath70 reflects the shape of the underlying soft photons . using the black - body approximation
the resulting spectrum is @xmath74 .
the slope of the power law above @xmath70 , @xmath75 , is fixed by the measured slope of the power - law derived from the fitting of the x - ray data .
note that the presence of a strong depression between the uv - optical and the soft - x - ray band requires a minor contribution from the ssc component , which peaks just in this region . without any information on the @xmath6-ray emission , the values of @xmath54 and @xmath76 are basically unconstrained . for illustration ,
we report two realizations of the model reproducing the x - ray spectrum with two different values of @xmath77 , the lorentz factor of the electrons emitting at the peak ( @xmath78 , solid , and @xmath79 , long dashed ) . clearly
, the two cases mainly differ for the expected power output in the @xmath7-ray band , the first being above and the second below the _ egret _ limit of sensitivity ( around @xmath80 erg @xmath3 s@xmath36 , e.g. , thompson et al . 1993 ) . only simultaneous @xmath7-ray observations , providing a constraints on the position of the ec peak , can allow us to fix the value of @xmath54 . for comparison
we report the predicted _ glast _ 5@xmath81 sensitivity ( long - dashed curve ) calculated for 1 year all - sky survey .
the models for the case of intrinsic absorption are reported in fig.([sedallabs ] ) . in this case
the x - ray spectrum of the source is a power - law extending down to lower energy . to correctly reproduce this spectrum we decrease the bulk lorentz factor with respect to the previous case ( to @xmath82 ) , so that the hard tail falls outside the region probed by the data .
we compensated the lower degree of beaming with a larger electron density , with the effect to increase the relative importance of the ssc emission , which fills the region between the optical - uv and the soft x - ray band . in both cases the x - ray spectrum
can be well fitted by assuming a standard slope @xmath83 for the low energy branch of the electron energy distribution .
on the other hand , the extremely hard x - ray continuum measured by _ xmm - newton _ requires a rather flat distribution ( @xmath84 ) .
apart for the change of @xmath52 , the transition between the state of 2003 to that of 2006 can be reproduced with small changes in the other parameters , in particular a slightly larger value of the magnetic field .
both the and the data of rbs 315 show a clear hardening of the x - ray spectrum below 1 kev , equally well fitted by a power - law model with absorption in excess to the galactic value or by a broken - power law continuum with galactic absorption .
this feature , observed in several high-@xmath85 quasars , is commonly ascribed to the presence of absorbing material in the quasar environment ( e.g. , elvis et al .
1994 , cappi et al .
1997 , reeves et al .
1997 , fiore et al .
1998 , fabian et al . 2001a , b , bassett et al . 2004 , sambruna et al .
a debated issue is the possible existence of a trend for the absorption to increase with redshift ( elvis et al .
1994 , cappi et al .
1997 , fiore et al .
1998 , page et al .
2005 , yuan et al .
this would support scenarios ( e.g. silk & rees 1998 , fabian et al .
1999 ) proposing that in the earliest phases of their evolution qsos are substantially obscured by gas , subsequently expelled from the host galaxy by powerful winds . the observational evidence , however , is still far from being completely clear .
for instance , it seems that absorption is rather more common in radio - loud than in radio - quiet quasars ( e.g. page et al .
2005 , grupe et al .
2006 ) , suggesting that absorption is linked to the presence of a relativistic jet . however , as pointed out by page et al .
( 2005 ) , if the @xmath86 correlation is real , the fact that radio - loud sources appear to be more absorbed could simply be due to a selection effect , since radio - loud quasars are brighter in x - rays and can be observed at larger distances .
we note that a substantial number of the high @xmath85 radio loud quasars are in fact blazars , i.e. observed at small angle to the jet axis as is the case for rbs 315 and the very hard x - ray spectra indicate that their emission is dominated by the jet .
a potentially severe problem of the absorption interpretation is the apparent absence of strong extinction in the optical - uv band .
indeed , with the column densities inferred from the x - ray spectral fits , in some cases reaching values of @xmath87 @xmath3 ( fabian et al .
2001a ) , the corresponding obscuration would be huge ( @xmath88 ) , in contrast with the optical observations requiring rather small reddening .
a solution to this discrepancy is to invoke an extreme gas to dust ratio in these sources , possibly due to the high - ionization state of the gas ( as the warm absorber found in seyfert galaxies ) .
however , the evidence of the high - ionization state from the fits of the x - ray spectra is not conclusive , even with good quality data ( e.g. worsley et al .
2004a , b ) . in fact , due to the redshift , the most important features that can be used for the diagnostics move below 0.1 kev and can not be observed .
the discussion above is based on the assumption that the hardening at low energies is due to ( intrinsic ) absorption in both radio - quiet and radio - loud sources .
however in radio - loud sources , and in particular in blazars , the observed spectral shape could be _ naturally _ accounted for by the intrinsic curvature of the ic emission from the jet , as shown here for the case of rbs 315 ( see also fabian et al .
2001a ; sambruna et al . 2006a , 2007 ) .
indeed , for parameters typically derived in reproducing the sed of powerful blazars , the model predicts a smooth hardening of the spectrum below a few kev ( in the rest frame ) . to better illustrate this point
, we show in fig.([compa ] ) the comparison between the ec spectrum ( solid line ) and an absorbed power law spectrum ( dashed ) for an intrinsic column density @xmath89 @xmath3 ( calculated using the zwabs model of xspec ) , for a source located at @xmath90 ( energies are in the observer frame ) . as already discussed , the ec spectrum present a break at a frequency approximately given by @xmath91 , where @xmath92 is the peak frequency of the external radiation field . to mimic an effective column density one has therefore to adjust the parameters to obtain the right value for the break frequency .
for instance , the curve reported in fig.([compa ] ) have been calculated assuming @xmath93 hz and fixing the low - energy end of the electron distribution to @xmath94 .
therefore , @xmath70 is fixed only by the value of the bulk lorentz factor .
as can be seen , the shapes of the two kinds of spectra are substantially similar down to energies of 0.3 - 0.4 kev ( but this energy is lower for sources at higher @xmath85 ) . below that energy
, the ec curve stays systematically above the absorbed power - law , since it asymptotically reaches a spectrum @xmath95 ( reproducing the slope of the underlying population of soft photons , assumed to follow a black - body shape ) , while the absorption imprints an exponential cut - off .
however , with data covering a limited energy range the two models can give equally acceptable fits .
the hardening of the ec spectrum at low energies could be masked by the possible presence of the softer ssc component , which typically peaks between the uv and the soft x - ray band .
however , from the observational point of view ( e.g. fossati et al .
1998 ) the relative importance of the ssc component seems to decrease with the power of the source ( maybe due to an increasing importance of the accretion disk and then of the external radiation field : e.g. , celotti et al .
2007 ) , leaving an almost `` naked '' ec component in the most powerful sources . in this case , the apparent effect of the absorption should be more important for the most powerful sources , accounting for the apparent trend of the @xmath26 with @xmath85 ( since for high redshift we select the most powerful sources ) .
in conclusion , in radio - loud , blazar - like , quasars , the intrinsic curvature of the ec component is a viable alternative to explain the cut - off in the soft x - ray band .
clearly , more observational effort should be devoted to study this effect .
the most direct observation able to discriminate between the two possible explanations would be the direct detection of features ( i.e. absorption edges ) directly imprinted on the x - ray continuum by the absorbing material .
spectra with sufficient signal can possibly be obtained with long exposures with the present instrumentation . in any case
, this can be one of the goals of the next generation of x - ray instruments .
moreover , it is mandatory to investigate the detailed shape of soft the x - ray continuum expected by considering realistic spectra for the soft ambient photons scattered in the ec component ( tavecchio et al .
, in prep ) .
besides the problems related to the cut - off at low energies , the quite hard x - ray spectrum of rbs 315 above 1 kev ( similar to that of other powerful blazars ) poses a great challenge to the scenarios currently accepted for particle acceleration ( ghisellini 1996 , sikora et al .
2002 ) . indeed , to correctly reproduce the x - ray continuum we have to assume @xmath55 ( especially for the observations ) .
however , a slope @xmath96 for the electron energy distribution ( implying a spectral index @xmath97 ) is predicted in scenarios considering either shock accelerated electrons or a population of cooled electrons . particularly difficult to be reconciled with the standard models
is the exceptionally hard slope measured by in 2003 .
note , moreover , that such hard spectra are not uncommon .
for instance , in the sample of 16 radio - loud quasars at @xmath98 considered by page et al .
( 2005 ) , 4 have hard spectra with @xmath99 .
this evidence could suggest either that shocks can produce relativistic electrons with a distribution much harder than @xmath96 or that another mechanism energizes the electrons , at least those with lorentz factors below @xmath100 ( which typically emits through ec at energies below 100 kev ) .
this possibility has been discussed by sikora et al .
( 2002 ) which proposed a two - step acceleration process . in this scenario
, electrons are initially accelerated from thermal energies to relativistic ones by a stochastic process ( involving , for instance plasma instabilities / turbulence or following magnetic reconnection events ) , thought to provide electron distributions harder than @xmath96 .
only those particles with energies above a threshold can effectively be accelerated by the fermi process at the shock front at energies at which they can emit @xmath7-rays .
however these speculations need to be assessed by improved theoretical analysis and calculations .
another interesting point concerns the variability .
although extremely variable in the @xmath7-rays , fsrq show little variability in the x - ray band on small timescales ( days ) . on longer timescales ( @xmath101months ) variability
is observed , but the variations are usually small , both in flux and slope ( see , for instance , the comparison reported in sambruna et al .
2007 ) . only in few cases extreme
variability has been observed both on long ( e.g. , pian et al . 2006 ) and short ( e.g. , foschini et al .
2006b ) timescales . rbs 315 changes the slope of the x - ray spectrum from @xmath102 to 1.5 between the two observations , separated by 3 years .
an interesting point to note is that the spectrum seems to change `` pivoting '' around 1 kev ( see fig.[sedall],[sedallabs ] ) .
sources displaying a behaviour similar to rbs 315 are rx j1028.6 - 0844 ( @xmath103 ) , for which two observations separated by one year show a change of the photon index from @xmath104 to 1.5 ( yuan et al .
2005 ) , and 0836 + 710 , for which the hard x - ray spectrum ( 20 - 200 kev ) softens from @xmath105 ( as measured by _ bepposax _ ) to @xmath106 ( _ bat / swift _ ) ( sambruna et al .
2007 ) . in tab.[parameters ]
we also report the radiative luminosity , @xmath107 , and the power carried by the jet , @xmath108 , calculated assuming the presence of 1 proton for each relativistic electron .
this choice is dictated by the condition that the jet carries enough power to support the large radiative output of these sources ( e.g. maraschi & tavecchio 2003 ) .
the derived values are around @xmath109 erg / s , quite large , but consistent with the values inferred for sources of similar radiative luminosity ( e.g. tavecchio et al . 2000 ) . the ratio @xmath110 , indicating the radiative efficiency of the jet , spans the range 10 - 50 , again , typical for this kind of sources ( e.g. sambruna et al .
2006b ) . in the hypothesis that the disk substantially contributes to the optical - uv luminosity , one can derive the disk luminosity , @xmath44 erg / s .
as already discussed for other powerful blazars , in the case of the standard value of @xmath111 for the accretion efficiency , this would imply that the jet carries away a power comparable to that carried inward by the accretion flow ( rawlings & saunders 1991 , celotti et al .
1997 , maraschi & tavecchio 2003 ) .
finally we note that even if the black hole is accreting at the eddington limit , the large luminosity requires an extreme value for its mass , @xmath112 @xmath113 .
@xmath114 the analysis of the spectrum of rbs 315 confirms the finding of page et al .
( 2005 ) and piconcelli & guainazzi ( 2005 ) made using data on the presence of a cut - off in the soft - x - ray spectrum . this can be fitted equally well either using a power - law with absorption in excess of the galactic value or a broken power - law with galactic absorption .
@xmath114 the very hard spectrum measured by ( photon index @xmath115 ) poses a difficult challenge to the models usually considered for the acceleration of relativistic electrons . during the observation , however , the slope was steeper , @xmath4 , closer to more `` standard '' values .
@xmath114 the sed , constructed with historical radio data , optical - uv data obtained with _
uvot _ and the x - ray data , is consistent with those usually derived for other intermediate / high-@xmath85 powerful blazars , usually shining in the @xmath7-ray band .
it is thus likely that rbs 315 will be detected by the upcoming _ glast _ satellite .
@xmath114 through the modeling of the sed with the standard synchrotron+ic emission model for blazars we derive parameters typical for powerful blazars .
however , the lack of the @xmath7-ray detection prevent a firm constraint of all the parameters .
@xmath114 we showed that the cut - off in the soft x - ray band , commonly ascribed to absorption by material in the quasar environment , can be also naturally reproduced by the hard tail of the ic component .
further studies are required to distinguish between the two alternatives .
we thank giancarlo ghirlanda for useful discussions and the anonymous referee for constructive comments .
we are grateful to niel gehrels and the _ swift _ team for the too observation of rbs 315 .
we acknowledge financial support from asi . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
lccccc + model & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 + ( 1)&(2)&(3)&(4)&(5)&(6 ) + & & & & & + pl+ga & @xmath121 & & & 1252.9/1061 & 8.5@xmath122 + pl+ga+za & @xmath123 & 2.65@xmath124 & & 1110.4/1060 & 8.4@xmath122 + bpl+ga & 0.68@xmath125 & 1.42@xmath126 & 1.23@xmath127 & 1105.2/1059 & 8.4@xmath122 + & & & & & + lccccccc + model & @xmath116 & @xmath117 & @xmath118 & @xmath128&@xmath119 & @xmath120 & @xmath129 + ( 1)&(2)&(3)&(4)&(5)&(6)&(7)&(8 ) + & & & & & & & + pl+ga+za & @xmath123 & 2.7@xmath124 & & @xmath130 & 1123.0/1073 & 8.3@xmath122 & 25@xmath131 + bpl+ga & 0.68@xmath132 & 1.42@xmath126 & 1.22@xmath133 & @xmath134 & 1118.1/1072 & 8.4@xmath122 & 25@xmath131 + & & & & & & & + lccccc + model & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 + ( 1)&(2)&(3)&(4)&(5)&(6 ) + & & & & & + pl+ga+za & @xmath135 & 1.65@xmath136 & & 1033.3/1068 & 15.7 + bpl+ga & 0.26@xmath137 & 1.21@xmath138 & 1.07@xmath139 & 1066.3/1067 & 15.7 + & & & & & + .model parameters used to calculate the seds reported in fig.([sedall])-([sedallabs ] ) . in all the models we assume a radius @xmath140 cm , a minimum electron lorentz factor @xmath68 , a disk luminosity @xmath44 erg / s and a blr radius @xmath64 cm . [ cols="<,^,^,^,^,^,^,^,^,^ " , ] | we present the results from the observation of the powerful radio - loud quasar rbs 315 ( @xmath0 ) , for which a previous _ xmm - newton _ observation showed an extremely flat x - ray continuum up to 10 kev ( photon index @xmath1 ) and indications of strong intrinsic absorption ( @xmath2 @xmath3 assuming neutral gas ) .
the instrument for hard x - rays hxd / pin allows us a detection of the source up to 50 kev .
the broad - band continuum ( 0.5 - 50 kev ) can be well modeled with a power - law with slope @xmath4 ( definitively softer than the continuum measured by _ xmm - newton _ ) above 1 kev with strong deficit of soft photons .
the low - energy cut - off can be well fitted either with intrinsic absorption ( with column density @xmath2 @xmath3 in the quasar rest frame ) or with a break in the continuum , with an extremely hard ( @xmath5 ) power - law below 1 kev .
we construct the spectral energy distribution of the source , using also optical - uv measurements obtained through a quasi - simultaneous _
uvot / swift _ observation .
the shape of the sed is similar to that of other flat spectrum radio quasars ( fsrqs ) with similar power , making this source an excellent candidate for the detection in @xmath6-rays by _
glast_. we model the sed with the synchrotron - inverse compton model usually applied to fsrqs , showing that the deficit of soft photons can be naturally interpreted as due to an _ intrinsic curvature _ of the spectrum near the low energy end of the ic component rather than to intrinsic absorption , although the latter possibility can not be ruled out .
we propose that in at least a fraction of the radio - loud qsos at high redshift the cut - off in the soft x - ray band can be explained in a similar way .
further studies are required to distinguish between the two alternatives . |
graphene nanostructures@xcite attract increasing attention mainly due to their potential applications in high mobility electronics@xcite and solid state quantum information processing.@xcite in particular , low nuclear spin concentrations expected in graphene promise long spin lifetimes@xcite and make graphene quantum dots ( qds)@xcite interesting for spin - qubit operations.@xcite moreover , graphene nanostructures may allow to investigate phenomena related to massless dirac fermions in reduced dimensions.@xcite intensive research has been triggered by the unique electronic properties of graphene @xcite including the gapless linear dispersion , and the landau level ( ll ) spectrum.@xcite recent advances in fabricating width - modulated graphene nanoribbons have helped to overcome intrinsic difficulties in creating tunneling barriers and confining electrons in graphene , where transport is dominated by klein tunneling - related phenomena.@xcite graphene qds have been fabricated and coulomb blockade,@xcite quantum confinement@xcite and charge detection@xcite have been demonstrated . in this article , we focus on the eigenenergies of graphene quantum dots ( see fig . [
fig : geom ] ) as a function of a perpendicular magnetic field . in graphene
, the linear band crossing at the so - called dirac point suggests a close connection between the dynamics of electrons and free , ultrarelativistic dirac particles.@xcite one might therefore expect a magnetic - field dependence of quantum dot eigenenergies that closely mirrors that of massless dirac particles .
indeed , this connection has been used recently to discuss the spectrum of ideal , circular graphene dots with smooth confinement.@xcite however , in more realistic models of finite graphene nanostructures , quantum confinement , edge effects , and lattice defects introduce a host of competing length scales absent from the simple dirac picture .
much progress has been made in understanding the unique ll spectrum , and the resulting hall effect , in graphene.@xcite the magnetic - field dependence of the addition spectrum has been exploited in recent work to ( approximately ) pin down the electron - hole crossover point.@xcite in the present paper we report on a systematic study of the @xmath0-field dependence of electronic eigenstates of graphene quantum dots of experimentally realizable size ( diameter @xmath3 nm ) .
we highlight the interplay of different length scales controlling the break - down of the valley symmetry by @xmath2 scattering .
the latter is found to be key to the understanding of the diamagnetic spectrum .
we find the @xmath0-field dependence of the level variance to be a sensitive measure for the strength of @xmath2 scattering and obtain good agreement with experimental coulomb blockade data .
the paper is organized as follows : we first briefly summarize the dirac picture of landau level formation for massless charged dirac particles and discuss the length scales relevant to its applicability to finite - size graphene quantum dots ( sec .
ii ) . in sec .
iii we present realistic simulations for graphene quantum dots with zigzag and armchair edges , with edge roughness as well as with bulk disorder .
a comparison between the calculated @xmath0-field dependence of the level variance and experimental data is given in sec .
iv , followed by a short summary ( sec .
the remarkable similarity of the low - energy band structure of graphene with the dispersion relation of a massless dirac particle in two dimensions has been widely exploited in a variety of theoretical models for graphene.@xcite however , the applicability of such models requires careful consideration of competing effects that go beyond the simple , yet intriguing dirac picture.@xcite a case in point is the diamagnetism , i.e. , the magnetic response of a finite - size graphene quantum dot .
it is of considerable interest to inquire into the applicability as well as the limitations of the well - known diamagnetic theory of charged massless dirac fermions .
the magnetic - field ( @xmath0 ) dependence of the spectrum of free dirac particles was first solved in an early paper by rabi@xcite shortly after the dirac equation was proposed .
the dirac equation for a massless particle with charge @xmath4 in the presence of a potential @xmath5 with time - like coupling and a perpendicular , homogeneous magnetic field @xmath6 reads @xmath7 with @xmath8 the pauli matrices and @xmath9 . in the limiting case of strong magnetic field where @xmath10 we explicitly label this reference spectrum with the superscript `` d '' ( for dirac equation ) .
equation ( [ b_d_l ] ) contains several remarkable features absent from non - relativistic diamagnetism : a ground state landau level @xmath11 the energy of which does not depend on @xmath0 at all .
higher landau levels @xmath12 , are distributed symmetrically around @xmath11 , and feature a @xmath13 rather than a linear dependence on @xmath0 known from non - relativistic diamagnetism .
the high - field regime [ eq .
( [ b_d_l ] ) ] is controlled by just two length scales , the ( energy dependent ) de broglie wavelength @xmath14 and the magnetic length @xmath15 .
the strong ( weak ) field regime is characterized by @xmath16 ( @xmath17 ) . in the limit of weak magnetic fields , eq .
( [ eq : hdirac ] ) predicts the lowest - order energy corrections to scale linearly with @xmath0 , unlike the conventional non - relativistic behavior ( @xmath18 ) .
eigenstates of eq .
( [ eq : hdirac ] ) form two - spinors with definite helicity ( or `` chirality '' ) @xmath19 in the absence of external fields .
the ideal , infinitely extended graphene sheet featuring a honeycomb lattice made up by two interleafed triangular sublattices ( a and b ) , can be described in nearest neighbor tight - binding approximation by the hamiltonian@xcite @xmath20 where the sum @xmath21 extends over pairs of adjacent lattice sites , @xmath22 is the tight - binding orbital at lattice site @xmath23 , @xmath24 is a locally varying potential , and @xmath25 ( of the order of @xmath26 ev ) is the nearest neighbor hopping matrix element . [ in the numerical calculations we take into account second and third nearest neighbor coupling @xcite in addition to eq .
( [ h_graph_tb ] ) in order to quantitatively account for the realistic band structure .
] close to the fermi energy , the band structure of eq .
( [ h_graph_tb ] ) can be approximated ( assuming that @xmath27 ) by a conical dispersion relation around the @xmath28 point,@xcite @xmath29 where we have set @xmath30 .
note that the above expansion ignores both the length scale of the graphene lattice constant @xmath31 and preferred directions of the lattice : due to the discrete lattice symmetry , the cone structure becomes squeezed along the @xmath2 directions , an effect known as triangular warping.@xcite more importantly , the band structure features _ two _ non - equivalent cones ( `` valleys '' ) at the @xmath28 and @xmath32 points in the reciprocal lattice .
this additional degeneracy allows to formaly represent the low - energy band - structure near @xmath33 in terms of dirac - like four - spinors @xmath34 with amplitudes for the a - b sublattice in real space and for the k - k valleys in reciprocal space .
operators in the four - spinor space can be represented by tensor products of @xmath35 matrices acting on a - b sublattice amplitudes and analogous @xmath36 pauli - matrices acting on @xmath2 amplitudes .
choosing the origin in @xmath37-space such that the connecting line between @xmath28 and @xmath32 is along @xmath38 , the effective dirac hamiltonian in the absence of external scalar potentials becomes@xcite @xmath39 where @xmath40 .
in addition to chirality , the valley - pseudospin projection @xmath41 associated with the valley degree of freedom is conserved .
the upper ( `` particle - like '' , @xmath42 ) and lower ( `` hole - like '' , @xmath43 ) cones touching each other at @xmath28 and @xmath32 with @xmath33 are related to each other by a particle - hole transformation @xmath44 in the presence of a time - like scalar potential @xmath45 , the hamiltonian is invariant under an anti - unitary transformation ( `` time reversal '' ) , @xmath46 , where @xmath47 denotes complex conjugation.@xcite the wavefunctions at @xmath28 and @xmath32 are related by time reversal symmetry . this symplectic symmetry ( @xmath48 )
is broken in the presence of a magnetic field @xmath49 lifting the two - fold kramers - like degeneracy ( note that physical spin is not included in the present analysis . ) .
we now consider a finite - size system of linear dimension @xmath50 , where @xmath51 takes on the role of a confinement potential .
with this additional length scale present , the landau - level solution [ eq .
( [ b_d_l ] ) ] is only valid in the strong magnetic field regime with @xmath52 , while in the weak field regime , @xmath53 , the spectrum will be determined by @xmath51 . for zero magnetic field , eigenstates @xmath54 and @xmath55 localized at the @xmath28 and @xmath32 points are degenerate .
turning on a magnetic field lifts this degeneracy without ( to lowest order ) introducing couplings between @xmath28 and @xmath32 .
following first - order degenerate rayleigh - schrdinger perturbation theory , the perturbation matrix @xmath56 describing the lowest order correction to the field - free spectrum of eq .
( [ e_graph ] ) takes the form [ for the gauge @xmath57 @xmath58 where @xmath59 is real , and linear in @xmath0 .
the perturbation preserves the valley symmetry .
the absence of @xmath2 coupling and the linear magnetic field dependence of @xmath60 for each decoupled dirac cone implies that energy eigenvalues _
linearly _ cross the line @xmath1 in pairs of two , forming an @xmath61-shaped intersection ( see fig .
[ fig : ev_clean ] ) .
this is in contrast to the nonrelativistic diamagnetic response @xmath18 in the perturbative limit .
it rather resembles the paramagnetic level splitting in conventional quantum dots when the magnetic field lifts a degeneracy .
examples of the latter are lifting of kramers degeneracy @xcite or a symmetry induced degeneracy as in circular quantum dots.@xcite in order to quantitatively simulate the diamagnetic response of a finite - size graphene quantum dot we first consider a smooth confinement potential that is slowly varying on a length scale of the lattice constant such as to approximately conserve the valley - pseudospin projection @xmath62 ( or valley symmetry ) . moreover , it conserves particle - hole ( anti-)symmetry [ eq . ( [ eq : partholesymm ] ) ] in order to prevent klein tunneling .
such a potential was first proposed by berry and mondragon@xcite in the context of neutrino billiards , @xmath63 where @xmath64 is the outward distance from the quantum dot boundary and @xmath65 introduces an additional length scale controlling the preservation of valley symmetry .
we choose @xmath66 [ see fig . [
fig : geom](a ) ] much larger than the lattice spacing ( @xmath67 ) .
consequently , eq . ( [ eq : pot ] ) varies slowly on the scale of the lattice constant , conserves valley symmetry [ to order @xmath68 and provides a realization of the ( approximately ) @xmath2 decoupled diamagnetic perturbation [ eq .
( [ e_pert ] ) ] .
we note that realizations of potentials of the form of eq .
( [ eq : pot ] ) are , to our knowledge , currently experimentally not available .
we employ a third - nearest neighbor@xcite tight - binding approximation ( to correctly describe triangular warping ) and simulate a 50@xmath6950 nm graphene qd containing @xmath70 100.000 carbon atoms .
the magnetic field is included by a peierls phase factor .
we use a lanczos diagonalization in conjunction with an lu factorization to efficiently calculate the 500 eigenvalues closest to the fermi edge@xcite [ see fig . [
fig : ev_clean ] ] . in the limit of weak magnetic fields
, we find that our numerical results , indeed , follow the linear @xmath0-field dependence of the energy eigenvalues as predicted by perturbation theory [ eq .
( [ e_pert ] ) ] [ see fig . [
fig : ev_clean](b ) ] .
residual deviations from the perfect lattice symmetry ( due to the finite width @xmath65 of the confinement ) and , thus , weak non - conservation of @xmath62 appear as minute energy splittings between near - degenerate levels in the @xmath71 limit .
the resulting level splitting at @xmath1 is , however , very small ( @xmath72 ev ) , i.e. , two orders of magnitude below the mean level spacing ( @xmath73 mev ) .
turning now to the high - field limit , @xmath52 , the influence of confinement effects should be diminished and the formation of landau levels following the dirac picture [ eq .
( [ eq : hdirac ] ) ] is expected .
the transition from low to high magnetic fields drastically changes the density of states ( dos ) .
the depletion of the dos near @xmath33 at low fields , @xmath74 is replaced , for increasing @xmath0 , by an increasing number of eigenstates moving towards the landau level at @xmath75 , which is located at the dirac point [ see fig . [
fig : ev_clean](a ) ] .
more specifically , all graphene levels with energies in between the two first landau levels , @xmath76 , adiabatically converge to the level at @xmath77 . as we have shown recently,@xcite this unique feature can be used to pin down the energetic position of the dirac point in the experiment and thus of the electron - hole crossover region in real graphene quantum dot devices .
while , at low fields , the valley symmetry is approximately preserved by the potential in eq .
( [ eq : pot ] ) , a large number of sizeable avoided crossings appear at higher magnetic fields as the magnetic length is reduced to @xmath52 .
edge states that couple to bulk states or to other edge states become prevalent .
the complicated pattern in fig .
[ fig : ev_clean](a),(c ) of many avoided crossings near the first landau energy @xmath78 reflects this interplay between magnetic bulk and edge states .
levels with eigenenergies that follow the predicted values for the landau levels , @xmath79 , are localized in the interior of the quantum dot and well separated from the edges .
conversely , the states with energies in between the values @xmath79 should be strongly influenced by the spatial confinement in the quantum dot .
wave functions corresponding to energy levels close to @xmath11 and @xmath80 landau levels [ fig .
[ fig : es_clean](a , c ) ] as well as those in between @xmath11 and @xmath81 [ fig . [
fig : es_clean](b ) ] and between @xmath80 and @xmath82 [ fig .
[ fig : es_clean](d ) ] confirm these expectations .
apparently , the typical level splittings at the avoided crossings [ fig .
[ fig : ev_clean](b , c ) ] are dramatically enhanced in the high - field regime .
this is due to the fact that , as compared to the low - field case , the amplitudes of wave functions are enhanced at the dot boundary [ see fig . [
fig : es_clean](b ) ] . since
, in addition , the lattice symmetry is broken at the boundary , these edge states do have an increased coupling strength to all other states in the sample . following the wigner - von neumann non - crossing rule@xcite
this increased coupling strength leads to increased level splittings at the crossing point .
we turn now to realistic graphene quantum dots where the nanostructures are terminated by atomically sharp edges , either of armchair or zigzag shape [ fig . [
fig : geom](b ) ] . following recent estimates for passivation of the dangling carbon bonds at the edges of graphene samples ( e.g. , by attached hydrogen),@xcite we set the potential of the out - most carbon atoms to 1.7 ev .
although in this case many of the surface states present in a perfect zizag boundary remain suppressed , such a confining potential leads to substantial changes in the energy level spectrum [ see fig . [
fig : ev_hard](a ) ] as compared to the model potential [ eq . ( [ eq : pot ] ) ] .
most importantly , in the low - field regime [ see fig . [
fig : ev_hard](b ) ] the linear @xmath0-field dependence is replaced by a quadratic dependence ( @xmath18 ) of the level splitting resembling the nonrelativistic diamagnetic response .
this is due to the presence of sizable avoided crossings near @xmath1 as a result of broken valley symmetry caused by the edges .
the valley symmetry is , thus , broken upon reflection at atomically sharp `` clean '' zigzag edges and @xmath62 is no longer conserved . in terms of perturbation theory ,
the confining potential @xmath83 now includes off - diagonal components in the pseudospin degree of freedom @xmath84 with coupling matrix elements between the valleys @xmath85 and eigenvalues @xmath86 the coupling between @xmath28 and @xmath32 cones thus leads , according to the wigner - von neumann non - crossing rule,@xcite to avoided crossing with level splittings @xmath87 [ see fig .
[ fig : ev_hard](b ) ] proportional to the coupling strength @xmath88 between the two dirac cones .
conversely , a fit to eq .
( [ ev_pert_kk ] ) yields a sensitive indicator for the amount of @xmath2 scattering in the quantum dot . for high magnetic fields [ see fig .
[ fig : ev_hard](c ) ] , the presence of @xmath89 coupling lead to a large number of correlated avoided crossings when the edge states move towards the zeroth bulk landau level .
in other words avoided crossings appear when the energy of eigenstates evolving towards the @xmath90 level `` pass '' through the energy @xmath91 of the first landau level [ fig .
[ fig : ev_hard](c ) ] . due to a large number of avoided crossings , there are no states continuously following the first landau level .
we rather observe a bundle of states sequentially moving along the characteristic energy of the first landau level , much like in a relay [ see fig . [
fig : ev_hard](c ) ] .
such an interrelated sequence of avoided crossings is well - known from atomic physics as `` diabatic ridge '' - riding states localized on potential barriers.@xcite a direct consequence is that the evolution of eigenstates for an increasing magnetic field [ see green highlighted line in fig .
[ fig : ev_hard](a ) ] features sharp `` kinks '' when crossing the ridge following the first landau level [ see arrow in fig . [
fig : ev_hard](c ) ] . as the state
is transiently trapped by the ridge , it moves away from the dirac point , and continues again monotonically towards the dirac point once clear of the ridge .
these kinks due to the ridge riding mechanism have been observed in the experiment@xcite serving as an additional indicator for the position of the lowest landau level and the electron - hole crossover .
the present results show that the atomically sharp edges destroy the linear @xmath0-dependence of the energy levels at weak fields , but they do preserve the square - root @xmath0-dependence at very high fields .
the linear @xmath0 dependence results from the fragile suppression of @xmath2 scattering ( @xmath92 ) between the dirac cones while the square - root dependence results from the much more robust dispersion relation of the individual cone . therefore , landau levels survive the introduction of sharp boundaries much better than the energy levels at weak fields . in turn , even when in the experiment a dirac - like landau level spectrum is recorded , many other features of the same graphene sample may very well show large deviations from predictions based on dirac theory . to further elucidate the role of lattice symmetry breaking in graphene quantum dots
, we now consider isolated lattice defects in the bulk .
the preceding results suggest that disorder realizations that break the @xmath2 symmetry are of crucial importance .
we therefore consider first single lattice vacancies [ see inset in fig . [
fig : ev_disorder](a ) ] with defect densities of @xmath93 to @xmath94 impurities per carbon atom . to isolate the effect of such bulk defects from the @xmath2 scattering of the edges , we first use the smooth symmetry preserving boundary potential
[ see eq .
( [ eq : pot ] ) ] .
overall , the diamagnetic spectrum closely resembles that of clean samples with atomically sharp edges ( see fig .
[ fig : ev_disorder ] ) .
in particular , avoided crossings near @xmath1 with a quadratic field dependence as well as the formation of sequences of avoided crossings along the ridges of bulk landau levels are found .
wave functions of eigenstates at these energies display patterns which are very similar to those for the clean system [ compare fig .
[ fig : es_clean](b ) and fig .
[ fig : es_disorder](a ) ]
. likewise , the kink pattern at the crossing of the ridges appears robust against disorder [ see , e.g. , fig .
[ fig : ev_disorder](c ) ] .
we do , however , observe new ridges between the landau level energies which were absent for edge scattering [ see dotted lines in fig . [
fig : ev_disorder](a , c ) ] .
the corresponding eigenstates near these new ridges [ see fig . [
fig : es_disorder](b ) , marked by @xmath95 in fig .
[ fig : ev_disorder](a ) ] are pinned to a single defect , where the lattice periodicity and the a - b sublattice symmetry are broken , and the partitioning of the wavefunction in four components according to the hamiltonian in eq .
( [ e_graph ] ) fails .
such localized states can therefore be expected to behave differently from the bulk landau levels .
the resulting ridges feature a very weak quasi - linear magnetic field dependence .
we therefore conjecture that these structures are due to avoided crossings with such localized defect states .
recent analysis@xcite has shown that @xmath2 scattering at impurities in graphene , i.e. the strength of @xmath89 , tends to be strongly energy dependent .
bound states due to adsorbates ( leading to an enhanced local density of states at the adsorption site ) frequently have energies close to the dirac point.@xcite since the localized defect states with energy @xmath96 feature a very weak explicit linear magnetic field dependence , @xmath97 \approx e_n[v_{kk ' } ] + \alpha \left|b\right|,\quad |\alpha| < 1\frac{\mathrm{mev}}{\mathrm{t}},\ ] ] the implicit magnetic field dependence of @xmath98)$ ] may be neglected .
while the detailed energy dependence of @xmath89 may be connected to the specific defect present ( e.g. , stone - wales defects , or attached nitrogen molecules ) , we still qualitatively expect an analogous linear @xmath0-field dependence . such localized states with weak magnetic field dependence
were also observed experimentally in coulomb blockade measurements .
@xcite in the low field regime , we observe a significant change in the @xmath0-evolution of eigenenergies : avoided crossings become asymmetric [ see fig .
[ fig : ev_disorder](b ) ] .
the reason is that the @xmath2 splitting introduced by the lattice vacancies is strong enough to yield different matrix elements for @xmath99 and @xmath100 : consequently , the slope of both eigenvalues of eq .
( [ e_pert_kk ] ) is different [ see fig . [
fig : ev_disorder](b ) ] . to illustrate that the diamagnetic spectrum , in particular the avoided crossing distribution , is due to the breaking of the a - b sublattice and , in turn , to the breaking of valley pseudospin symmetry induced by the defects , we present as counter example the spectrum for double vacancies .
we introduce such a vacancy in accordance to our soft - wall potential , eq .
( [ eq : pot ] ) . as such double vacancies act on an entire unit cell in the hexagonal lattice
, they approximately conserve the electron - hole and @xmath2 symmetry .
we find , indeed , that for the same number of defects as of single vacancies ( fig .
[ fig : ev_disorder ] ) , only avoided crossings with comparatively small energy splittings appear in the energy level diagram ( fig .
[ fig : ev_impurities ] ) , resembling much more closely fig .
[ fig : es_clean ] .
this clearly indicates that it is the breaking of the a - b symmetry and not the presence of defects per se which is responsible for the break - down of the dirac picture for graphene quantum dots .
we consider now a clean graphene quantum dot with atomically sharp but disordered edges .
we connect short straight edge segments ( with a random length between 0.5 - 3 nm ) to obtain a polygon - shaped boundary [ see fig . [
fig : geom](c ) ] with a disorder amplitude of @xmath101 nm .
@xmath102 is thus ( at energies close to the dirac point ) smaller than the wavelength @xmath103 of the confined particles as well as the magnetic length , but larger than the lattice constant . since rough boundaries , just like bulk defects , localize states,@xcite we expect similar signatures of these two types of disorder
indeed , in the low field regime , dots with rough edges feature a similar pattern of fluctuating energy levels as dots with single - lattice defects ( compare fig .
[ fig : ev_rough ] and fig .
[ fig : ev_disorder ] ) .
the spectra are so similar that it is difficult to distinguish between bulk and edge disorder breaking the @xmath2 symmetry .
also in the high - field regime , the evolution towards the landau levels features the correlated sequence of avoided crossings reflecting the diabatic ridges [ see fig .
[ fig : ev_rough](c ) ]
. wave functions of eigenstates at these energies display patterns which are very similar to those for the clean system [ compare fig . [
fig : es_clean](a ) and fig .
[ fig : es_disorder](d ) ] . in particular , states
near landau levels localize in the interior of the dot , and thus are not influenced by edge disorder .
likewise , the kink pattern at the crossing of the ridges is robust against edge disorder . while the inclusion of edge disorder does not give rise to qualitatively new effects on the eigenenergy spectrum in the high - field regime , differentiating between edge and bulk disorder
might become possible by probing the different scaling behavior for bulk and edge disorder with the size of the graphene quantum dot .
it is instructive to directly contrast the level splitting @xmath87 of an avoided crossing [ eq . ( [ ev_pert_kk ] ) ] due to finite @xmath2 coupling for different scenarios : ( i ) valley - symmetry preserving confinement [ fig .
[ fig : summary](a ) ] , ( ii ) clean graphene quantum dots with atomically sharp edges [ fig . [
fig : summary](b ) ] and ( iii ) disordered graphene quantum dots [ figs .
[ fig : summary](c , d ) ] .
we observe an increase in the size of the average level splitting @xmath104 due to @xmath2 coupling at the edges and impurities . in scenario ( i )
@xmath104 is at least one order of magnitude smaller than the mean level spacing @xmath105 , @xmath106 . in ( ii ) where @xmath107
a first - order perturbative treatment of @xmath2 coupling correctly describes the level repulsion .
for localization at defects ( iii ) where @xmath108 level pairs at @xmath28 and @xmath32 feature different quadratic dependences on @xmath0 , i.e. , they can no longer be parametrized by eqs .
( [ e_pert_kk ] ) and ( [ ev_pert_kk ] ) .
for a comparison with the experimental data for the magnetic response of graphene quantum dots we pursue two strategies : in a direct approach we compare our models with the observed parametric @xmath0-field evolution of individual coulomb blockade peaks .
alternatively , we identify the @xmath0-field dependence of the level spacing fluctuations ( variance ) as a robust measure for the degree of disorder in a graphene quantum dot .
specifically , we determine the rescaled ( or unfolded ) variance @xmath109 of the distribution of neighboring energy level spacings @xmath110 @xmath111 since switching on a magnetic field @xmath0 leaves the number of states unchanged , @xmath112 is ( approximately ) independent of @xmath0 , while higher moments of the level distribution are drastically affected . at @xmath113 ,
pairs of energy levels are split by the characteristic energy @xmath87 of the avoided level crossings .
however , as long as the mean width of the avoided crossing @xmath104 [ eq . ( [ ev_pert_kk ] ) ] is small compared to the mean level spacing @xmath105 , the level sequence fluctuates between small and large spacings , while spacings of the order of @xmath110 are unlikely .
we thus expect for magnetic field @xmath1 a comparatively large variance @xmath109 . for increasing @xmath114
the levels become more equally spaced , leading to a decrease in @xmath109 . correspondingly
, the variance @xmath109 of the level spacings should feature a peak at @xmath1 .
the numerical results for the dependence of @xmath109 on @xmath0 for different disorder strength ( i.e. , different @xmath115 ) are shown in fig .
[ fig : ev_theor ] .
our data display a pronounced peak of @xmath109 for the clean flake slowly decreasing for increasing number of single - vacancy defects ( i.e. , for increasing @xmath2 scattering ) [ see fig . [
fig : ev_theor](a ) ] .
note that both the peak height at @xmath1 and the overall value of @xmath109 decrease with increasing disorder .
the latter can be explained by the emerging localized states that feature a regular spacing ( and hence a suppressed variance @xmath109 ) .
consequently , if a given spectral region is more prone to feature localized states due to adsorbates than others,@xcite we expect an energy - dependence on @xmath116 depending on the specific type of adsorbate .
such an energy dependence could be exploited to directly measure the energy dependence of @xmath2scattering by determining the statistics of coulomb blockade resonances at different back - gate voltages . for comparison
we also show @xmath109 for double vacancies preserving a - b symmetry .
[ see fig .
[ fig : ev_theor](b ) , note the different scales ] .
accordingly , @xmath109 is , indeed , strongly dependent on the amount of @xmath2 scatterers , not on the overall number of defects .
the decrease in peak height with increasing @xmath2 scattering should thus provide a robust and sensitive measure for @xmath2 scattering present in the experiment . to test this conjecture ,
we have measured the evolution of 42 coulomb blockade peaks for varying magnetic field.@xcite we follow the parametric motion of the peak positions , and hence the eigenstate energies , as a function of magnetic field @xmath117$]t . to compare with our numerical results , we take into account a charging energy of 13 mev ( determined independently@xcite ) as well as spin ( by zeeman splitting ) .
we observe , indeed , a quadratic @xmath0-field dependence rather than the linear dependence predicted for conserved @xmath2 symmetry of the dirac equation .
our experimental data can be well described by eq .
( [ ev_pert_kk ] ) [ see fig . [
fig : experiment](a ) ] . for pairs of consecutive coulomb blockade resonances belonging to
the same avoided crossing , we find a mismatch in slopes , in agreement with our numerical findings for the rough - edged quantum dot [ compare fig .
[ fig : experiment](a ) and fig .
[ fig : summary](c , d ) ] .
this has important consequences for the interpretation of experimental data : the roughness present in the experimental dot does not allow to disentangle @xmath28 and @xmath32 states . to be more quantitative , we compare our simulations for the level variance with experimental data .
we indeed find good agreement as confirmed by a noticeable peak also in the experimental data for @xmath118 [ see fig . [
fig : experiment](b ) ] .
the offset between the two data sets in fig .
[ fig : experiment](b ) is attributed to a possible energy dependence of @xmath119 scattering as well as statistical fluctuations in charging energy and in the number of localized states for different values of the back gate voltage . by using the edge roughness @xmath102 as the only adjustable parameter
, we can match the measured @xmath118 very well with our numerical simulation [ see lines in fig . [
fig : experiment](b ) ] .
good agreement is found for an edge roughness of about @xmath120 nm , or , equivalently , a @xmath2 scatterer concentration in the bulk @xmath121 , both of which are well within expectation .
we emphasize that , although we can quantify the resulting overall strength of @xmath2 coupling in our experimental quantum dot , we can not disentangle whether the observed @xmath2 scattering comes from edge roughness , lattice defects , or disorder through flake - substrate interactions with a length scale comparable to the lattice constant .
we have investigated the evolution of eigenstates in graphene quantum dots with increasing magnetic field . concentrating on the energy regime around the dirac point ,
we observe a smooth transition from a linear density of states to the emergence of landau levels . at high field strength , we find that landau levels follow the square - root dependence of the dirac equation , manifested in the energy level diagram by sequences of correlated avoided crossings along `` diabatic ridges ''
. these ridges lead to characteristic kinks in the evolution of energy states that cross a landau level .
appearing also in coulomb blockade measurements , these kinks can be used to experimentally pin down the electron - hole crossover point.@xcite in the perturbative regime of small magnetic fields , we find that the linear dependence on @xmath0 predicted by the model of massless dirac fermions disappears when the valley symmetry is broken .
even perfect armchair and zigzag edges are sufficient to break the sublattice symmetry giving rise to avoided crossings with a quadratic dependence on @xmath0 instead .
a similar effect is observed for lattice defects : single lattice vacancies break the @xmath2 symmetry and thus result in avoided crossings with substantial level splittings ( even for defect concentrations as low as 1 in 20.000 ) . by comparison with double vacancies which conserve the sublattice symmetry
we show that it is not the presence of disorder per se which leads to deviations from predictions by the dirac equation , but the breaking of valley symmetry .
we compare our theoretical predictions with experimental results on the parametric @xmath0-field evolution of coulomb blockade peaks . as a quantitative indicator for the strength of @xmath2 scattering
, we identify the variance @xmath109 of the level spacing distribution .
we observe a peak in the variance at @xmath1 due to level correlations near avoided crossings .
we find quantitative agreement between the measured and the calculated data for @xmath109 which enables us to pinpoint the amount of @xmath2 scattering present in our experimental flake .
the present results provide a sensitive indicator for the quality of the graphene dot and demonstrate the limits of the dirac picture in describing the experiment .
we thank k. ensslin , t. ihn , and m. wimmer for valuable discussions .
f.l . and j.b
acknowledge support from the fwf - sfb 016 and fwf - sfb 041 , j.g . from the swiss national science foundation , c.s .
support from nccr .
computational resources from the vienna scientific cluster ( vsc ) are gratefully acknowledged .
c. stampfer , j. gttinger , f. molitor , d. graf , t. ihn , and k. ensslin , appl .
* 92 * , 012102 ( 2008 ) , c. stampfer , e. schurtenberger , f. molitor , j. gttinger , t. ihn , and k. ensslin , nano lett . * 8 * , 2378 ( 2008 ) f. d. m. haldane , phys .
lett . * 61 * , 2015 ( 1988 ) y. zheng and t. ando , phys .
b * 65 * , 245420 ( 2002 ) e. mccann and v. i. falko , phys . rev .
96 , 086805 ( 2006 ) l. sheng , d. n. sheng , f. d. m. haldane , and l. balents , phys .
* 99 * 196802 ( 2007 ) p.
m. ostrovsky , i. v. gornyi , and a. d. mirlin , phys .
b * 77 * , 195430 d. a. bahamon , a. l. c. pereira , and p. a. schulz , phys . rev .
b * 79 * , 125414 ( 2009 ) p. h. rivera , a. l. c. pereira , and p. a. schulz , phys . rev .
b * 79 * , 205406 ( 2009 ) c. s. lent , phys . rev .
b * 43 * , 4179 ( 1991 ) .
c. lanczos , journal of research of the national bureau of standards * 45 * , 255 , ( 1950 ) p. r. amestoy , i. s. duff , j. koster , and j .- y .
lexcellent , siam journal of matrix analysis and applications * 23 * , 15 - 41 ( 2001 ) .
p. r. amestoy , a. guermouche , j .- y .
lexcellent and s. pralet , parallel computing * 32 * , 136 - 156 ( 2006 ) .
p. s. krsti , c. o. reinhold , and j. burgdrfer , phys .
a * 63 * , 052702 ( 2001 ) .
see also j. burgdrfer , n. rohringer , p. s. krsti , and c. o. reinhold , nonadiabatic processes near barriers in nonadiabatic transition in quantum systems edited by v.i .
osherov , l.i .
ponomarev , chernogolovka , russia p. 205
institute of chemical physics ras isbn 5 - 901675 - 48 - 7 yu . v. skrypnyk and v. m. loktev , prys .
b * 73 * , 241402(r ) ( 2006 ) j. p. robinson , h. schomerus , l. oroszlny , and v. i. falko , phys . rev .
* 101 * , 196803 ( 2008 ) t. o. wehling , m. i. katsnelson , and a. i. lichtenstein , phys .
b * 80 * , 085428 ( 2009 ) the measured graphene quantum dot had a similar geometry as the dot described in refs . .
by changing the potential of the back gate electrode ( @xmath122 ) a region of suppressed current between @xmath123 v and 45 v was found . from measurements of coulomb blockade peaks in this region ( with @xmath70 150 coulomb peaks ) a charging energy of @xmath7013 mev was extracted . for more details about the measurements
see refs . . | we investigate the electronic eigenstates of graphene quantum dots of realistic size ( up to 80 nm diameter ) in the presence of a perpendicular magnetic field @xmath0 .
numerical tight - binding calculations and coulomb - blockade measurements performed near the dirac point exhibit the transition from the linear density of states at @xmath1 to the landau level regime at high fields . details of this transition sensitively depend on the underlying graphene lattice structure , bulk defects , and localization effects at the edges .
key to the understanding of the parametric evolution of the levels is the strength of the valley - symmetry breaking @xmath2 scattering .
we show that the parametric variation of the level variance provides a quantitative measure for this scattering mechanism .
we perform measurements of the parametric motion of coulomb blockade peaks as a function of magnetic field and find good agreement .
we demonstrate that the magnetic - field dependence of graphene energy levels may serve as a sensitive indicator for the properties of graphene quantum dots and , in further consequence , for the validity of the dirac - picture . |
clusters of galaxies are the largest gravitationally bound systems in the universe , with typical masses of about 10@xmath0 m@xmath1 , and volumes of about 100 mpc@xmath2 .
most of the gravitating matter in any cluster is in the form of dark matter ( @xmath3 80% ) .
some of the luminous matter is in galaxies ( @xmath3 3% - 5% ) , the rest is in diffuse hot gas ( @xmath3 15% - 17% ) , detected in x - ray through its thermal bremsstrahlung emission .
this thermal plasma , consisting of particles of energies of several kev , is commonly referred to as intracluster medium ( icm ) .
most of the detailed knowledge of galaxy clusters has been obtained in recent years from the study of icm through x - ray astronomy .
clusters are formed by hierarchical structure formation processes . in this scenario ,
smaller units ( galaxies , groups and small clusters ) formed first and merged under gravitational pull to larger and larger units in the course of time .
cluster mergers are the mechanism by which clusters are assembled .
denser regions form a filamentary structure in the universe , and clusters are formed within filaments , often at their intersection , by a combination of large and small mergers .
major cluster mergers are among the most energetic events in the universe since the big bang @xcite . during mergers ,
shocks are driven into the icm , with the subsequent injection of turbulence .
the merger activity , which has characterized much of the history of the universe , appears to be continuing at the present time and explains the relative abundance of substructure and temperature gradients detected in clusters of galaxies by optical and x - ray observations .
eventually , clusters reach a relaxed state , characterized by a giant galaxy at the center , and enhanced x - ray surface brightness peak in the core .
the hot gas in the centre has a high density , which implies short radiative cooling times , typically one or two order of magnitudes smaller than the hubble time .
therefore energy losses due to x - ray emission are dramatic and produce a temperature drop towards the center .
it was formerly suggested that in these conditions , the icm plasma in the cluster core should cool and condense , giving rise to a steady , pressure - driven inward flow caused by the compression of the hot surrounding gas .
relaxed clusters were then classified as `` cooling flow '' clusters @xcite , with predicted gas mass deposition rates in the cluster center up to 1000 m@xmath1 year@xmath4 .
this model was the subject of much debate , when xmm - newton spectral results failed to confirm the lines and features expected as a product of a steady state cooling flow @xcite .
x - ray and optical studies showed that the gas cooling rates were overestimated by an order of magnitude or more @xcite .
the classical `` cooling - flow '' model has finally been replaced by the paradigm of `` cool - core '' .
observationally cool - cores are characterized by a strong peak in the surface brightness , a significant drop in the temperature and a peak in the metal distribution ( e.g. @xcite ) .
the cooling time is often much shorter than 1 gyr , therefore some source of heating is necessary to balance the radiation losses . at present
, it is widely accepted that the source of heating in cool - core clusters is the agn activity of the brightest cluster galaxy at the center ( see @xcite for recent reviews ) .
galaxy clusters are also characterized by emission in the radio band .
obvious radio sources are the individual galaxies , whose emission has been observed in recent decades with sensitive radio telescopes .
it often extends well beyond the galaxy optical boundaries , out to hundreds of kiloparsec , and hence it is expected that the radio emitting regions interact with the icm .
this interaction is indeed observed in tailed radio galaxies , and radio sources filling x - ray cavities at the centre of cool - core clusters @xcite .
more puzzling are diffuse extended radio sources , which can not be obviously ascribed to individual galaxies , but are instead associated with the icm .
this radio emission represents a striking feature of clusters , since it demonstrates that the thermal icm plasma is mixed with non - thermal components .
such components are large - scale magnetic fields with a population of relativistic electrons in the cluster volume .
further demonstration of the existence of magnetic fields in the icm is obtained by studies of the faraday rotation of polarized radio galaxies , lying in the background or embedded within the magnetized intracluster medium .
the number density of relativistic particles is of the order of 10@xmath5 @xmath6 , while their lorentz factors are @xmath7 1000 .
magnetic field intensities are around @xmath3 0.1 - 1 @xmath8 g , and the energy density of the relativistic plasma is globally @xmath9 1% than that of the thermal gas .
non - thermal components are nevertheless important for a comprehensive physical description of the intracluster medium in galaxy clusters , and play a major role in the evolution of large - scale structures in the universe .
they can have dynamic and thermodynamic effects : magnetic fields affect the heat conduction in the icm and the gas dynamics , relativistic particles are sources of additional pressure and undergo acceleration processes that can modify the details of the icm heating process .
realistic cosmological simulations of galaxy cluster formation also include non - thermal components , thus a deep knowledge of the properties of these components and of their interplay with the thermal gas are needed to properly constrain the large - scale structure formation scenario .
the discovery of diffuse cluster radio emission has represented an important step in the understanding of the physical processes in clusters of galaxies .
diffuse synchrotron sources are sensitive to the turbulence and shock structures of large - scale environments and provide essential complements to studies at other wavebands .
studies in the radio domain will fill essential gaps in both cluster astrophysics and in the growth of structure in the universe , especially where the signatures of shocks and turbulence , or even the thermal plasma itself , may be otherwise undetectable .
the aim of this review is to present the observational results obtained in recent years in the radio domain related to the diffuse radio sources , in order to give an overview of the state of the art of the current knowledge of non - thermal cluster components .
we will show how the radio properties can be linked to the x - ray properties and to the cluster evolutionary state , and will discuss the physical implications .
cluster non - thermal emission of inverse - compton ( ic ) origin is expected in the hard x - ray domain , due to scattering of the cosmic microwave background photons by the synchrotron relativistic electrons @xcite . to present days ,
the detection of a non - thermal hard tail in the x - ray spectrum of clusters of galaxies has been reported in the literature ( e.g. @xcite ) , and it is still debated in some cases .
the presence of a high frequency non - thermal emission is clearly confirmed in the ophiuchus cluster @xcite .
the discussion of this topic is beyond the scope of this review , which is focussed on the results in the radio band .
the organization of this paper is as follows : in sect .
[ sec:2 ] the general properties of diffuse cluster radio sources and their classification are presented ; in sect . [ sec:3 ] the problems related to the detection of diffuse emission are outlined ; sects .
[ sec:4 ] , [ sec:5 ] , [ sec:6 ] give the observational properties of radio halos , relics and mini - halo , respectively .
in sect . [ sec:7 ] the current results about magnetic fields are presented .
[ sec:8 ] presents the evidence for radio emission and magnetic fields beyond clusters of galaxies .
[ sec:9 ] briefly summarizes the current models , and sect .
[ sec:10 ] depicts the future prospects .
the intrinsic parameters quoted in this paper are computed with @xmath10cdm cosmology with @xmath11 = 71 km s@xmath4mpc@xmath4 , @xmath12 = 0.27 , and @xmath13 = 0.73 .
values taken from the literature have been scaled to this cosmology .
synchrotron emission is produced by the spiraling motion of relativistic electrons in a magnetic field .
the power emitted by a relativistic electron depends on its energy ( lorentz factor ) and on the magnetic field strength .
the higher the magnetic field strength , the lower the electron energy needed to produce emission at a given frequency .
the case of astrophysical interest is that of a homogeneous and isotropic population of electrons with a power - law energy distribution with index @xmath14 . in this case , if the plasma is optically thin , as in diffuse cluster radio sources , the diagnostics of synchrotron radiation are the following : \i ) the emissivity is related to the number density of the relativistic electrons and to the intensity of the magnetic field , and the monochromatic emissivity follows a power - law with spectral index related to the index of the electron energy distribution @xmath15 ; typical observed radio spectra are around 0.7 - 0.8 , consistent with typical values of @xmath14 around 2.5 .
\ii ) energy losses of the radio emitting particles produce , with passing time , a change in the overall particle energy distribution , which in turn produces a modification of the emitted radio spectrum .
in particular , the spectrum shows a cutoff at frequencies higher than a critical frequency @xmath16 , related to the particle lifetime . therefore aged radio sources show curved spectra and/or spectra steeper that the values 0.7 - 0.8 indicated above ; \iii ) the synchrotron radiation from a population of relativistic electrons in a uniform magnetic field is linearly polarized , with the electric ( polarization ) vector perpendicular to the projection of the magnetic field onto the plane of the sky .
the intrinsic polarization degree depends on the particle energy distribution , and equals @xmath3 75% - 80% for typical values of the spectral index .
complex and disordered magnetic field structures decrease the observed degree of polarization ; \iv ) the total energy content in a synchrotron radio source depends on the contributions from relativistic particles ( electrons plus protons ) and from magnetic fields ( considering the fraction of the source volume occupied by magnetic fields , i.e. the filling factor ) . from radio data , it is possible to derive , the minimum value of the total energy , which occurs when the contributions of magnetic fields and relativistic particles are approximately equal ( equipartition condition ) . the corresponding magnetic field is commonly referred to as equipartition value b@xmath17 .
rigorous treatments of the synchrotron theory and radiation processes are in @xcite .
useful formulas of common use can be found in @xcite . in recent years
, there has been growing evidence for the existence of cluster large - scale diffuse radio sources , of synchrotron origin , which have no optical counterpart and no obvious connection to the cluster galaxies , and are therefore associated with the icm @xcite .
the first diffuse radio source detected in a cluster of galaxies is the giant radio halo at the center of the coma cluster @xcite .
several years later , extended diffuse emission was also detected at the periphery of the coma cluster @xcite and at the center of the perseus cluster @xcite .
nowadays , diffuse radio emission with surface brightness down to 0.1 @xmath8jy arcsec@xmath18 at 1.4 ghz is known in about 80 clusters , under several cluster evolutionary conditions ( merging and relaxed clusters ) , at different cluster locations ( center , periphery , intermediate distance ) , and on very different size scales : 100 kpc to @xmath19mpc , ( see fig . [
fig : clcollection ] for several examples ) .
diffuse sources are typically grouped in halos , relics and mini - halos @xcite , according to their location in the cluster and to the cluster type ( merging or cool - core ) .
the three classes will be extensively discussed in the next sections .
we anticipate that halos are hosted in clusters showing merging processes @xcite and are located at the cluster center .
relics are located in cluster peripheral regions of both merging and relaxed clusters .
prototypical examples of these two classes are both in the coma cluster : the radio halo coma c and the relic 1253 + 275 .
mini - halos are hosted in relaxed cool - core clusters , are centrally located , and usually surround a powerful radio galaxy ; the protoype of this class is in the perseus cluster .
it should be remarked here that non - thermal diffuse emission can also be present beyond clusters , indicating the presence of relativistic particles and magnetic fields in very low density environments .
all diffuse radio sources in clusters have very steep radio spectra in common . a clear difference between halos and relics
is in polarization property , with peripheral sources being more strongly polarized than the central ones . a possible problem in the classification of halos and relics could be represented by projection effects . in other words , halos could be actually located at cluster periphery , and simply projected onto the cluster center .
this issue has been analyzed several years ago @xcite , using data at the time on about 30 clusters . from the distribution of projected distances of halos and relics from the cluster center
, it was shown that there is a large excess of sources at the center , thus diffuse sources are not located at random position in clusters .
this indicates that central halos are truly at the cluster center and not simply projected onto it .
moreover , detailed radio images show different morphologies of halos and relics .
there are 42 halos currently known , while 39 clusters show at least 1 relic source ( for a total of 50 relics ) .
this confirms the findings above .
a powerful way to distinguish between the two classes is through the polarization , this will become easy with new generation radio telescopes . giovannini et al .
@xcite showed that diffuse cluster sources are not present in all galaxy clusters , but are not a phenomenon as rare as formerly believed .
they first showed that diffuse sources with the surface brightness limit of the nrao vla sky survey ( nvss @xcite ) are present in 6%-9% of clusters with l@xmath20 @xmath21 5 @xmath22 10@xmath23 erg / s , but are present in 27% 44% of clusters with l@xmath20 @xmath19 5 @xmath22 10@xmath23 erg / s .
this trend is confirmed for the occurrence of halos and relics , as reported in sect .
[ sec:4 ] and sect .
[ sec:5 ] , respectively .
for the detection of diffuse radio sources , the sensitivity to low surface brightness emission is the critical parameter .
single dish observations are in principle best suited to detect extended emission , however they are generally limited by the high confusion , because of the large beam .
moreover , low angular resolution may prevent the separation of embedded unrelated discrete sources . on the other hand ,
an interferometer is characterized by a smaller observing beam , and thus by lower confusion , but its brightness sensitivity , set primarily by the aperture filling factor and the shortest spacings , may be insufficient .
a proper sampling of the short spacings is vital not only for the detection of an extended source , but more crucially for the determination of the full size and the measurement of the source flux density .
multi - frequency images with a similar uv - coverage are needed to derive spectral index distributions , which can provide information on the age and energy of radio emitting particles
. the total spectral index can be easily obtained , but in many sources confusion problems do not allow a proper determination of low frequency fluxes . to summarize , the instrument best suited to detect cluster diffuse radio sources should provide a compromise between high sensitivity to surface brightness , and high angular resolution . moreover , since diffuse emission is generally characterized by steep radio spectra , the observations should be carried out at low frequencies ( @xmath9 1.4 ghz ) .
the very large array ( vla ) at 1.4 ghz has been demonstrated to be a very good instrument to detect and study radio halos , owing to its superb sensitivity .
however , it should be noted , as a caveat , that the coma radio halo ( see fig .
[ fig : coma327 ] ) is not detected by westerbork synthesis radio telescope ( wsrt ) and vla observations at 1.4 ghz because of the missing short spacings .
after the discovery of the coma radio halo at 408 mhz with the 250-ft radio telescope at jodrell bank @xcite , confirmed from interferometric cambridge one - mile telescope radio data @xcite , observations carried out mostly with single dish radio telescopes ( arecibo , green bank 300-ft , effelsberg and others ) and the wsrt found about 10 other clusters with a diffuse halo - type radio emission ( see @xcite ) .
jaffe & rudnick @xcite , in their search for radio halos , found an extended emission region in the peripheral region of the coma cluster near the strong radio source coma a ( 3c 277.3 ) .
this was suggested to be a feature of the coma cluster , in the following classified as a relic , thanks to observations at 408 mhz with the northern cross radio telescope at better angular resolution @xcite .
a significant breakthrough in the study of radio halos and other diffuse cluster radio sources was obtained thanks to the all - sky surveys with the vla and wsrt .
giovannini et al .
@xcite , combining radio data from the nvss at 1.4 ghz and x - ray catalogues , detected 18 new halo and relic candidates , in addition to the 11 already known , owing to the large beam and the good surface brightness sensitivity of the vla d configuration .
all new candidates were confirmed by more sensitive targeted observations , mostly performed with the vla .
kempner & sarazin @xcite presented 7 new candidates from a search in the westerbork northern sky survey ( wenss @xcite ) at 327 mhz ; rudnick & lemmerman @xcite searched for mpc - scale radio emission reprocessing wenss radio images . recently , other extensive radio observations have been published .
we note the giant metrewave radio telescope ( gmrt ) survey of massive galaxy clusters at z = 0.2 - 0.4 by venturi and collaborators @xcite , who found diffuse sources in 10 clusters ( out of which 3 new halos , 2 new core - halo sources and 3 new halo or relic candidates ) , the gmrt follow up at low frequencies of all clusters known to contain radio halos and relics up to z = 0.4 @xcite , and the observing campaign carried out with the wsrt , vla and gmrt by van weeren et al .
@xcite , who discovery 6 new radio relics , including a probable double system , and 2 radio halos .
the above mentioned studies and their follow - ups have produced rich information in a short time in this field .
radio halos are diffuse radio sources of low surface brightness ( @xmath3 1 0.1 @xmath8jy arcsec@xmath18 at 1.4 ghz ) permeating the central volume of a cluster .
they are typically extended with sizes of @xmath24 1 mpc , regular in morphology , and are unpolarized down to a few percent level , probably because of internal or beam depolarization . the prototype of this class , coma c at the center of the coma cluster ( fig .
[ fig : coma327 ] ) has been studied in detail by many authors ( see e.g. @xcite ) .
thanks to the improvements in observations and data reduction procedures , radio halos with a smaller size and irregular morphology have also been detected in rich galaxy clusters .
their properties in most cases are similar to those of giant radio halos .
[ fig : hacollection ] presents images of well known radio halos of different sizes .
a spectacular case is represented by the double merging system , a399 and a401 , which both contain a radio halo and can be considered the only case so far of a double radio halo system ( @xcite , fig .
[ fig : hadouble ] ) .
lclrrrrcl name & z & kpc/@xmath25 & s(1.4 ) & log p(1.4 ) & lls & lx(10@xmath23 ) & ref & notes + & & & mjy & w / hz & mpc & erg / sec & & + a209 & 0.2060 & 3.34 & 16.9 & 24.31 & 1.40 & 6.17 & @xcite & a relic could be present + a399 & 0.0718 & 1.35 & 16.0 & 23.30 & 0.57 & 3.80 & @xcite & double with a401 + a401 & 0.0737 & 1.38 & 17.0 & 23.34 & 0.52 & 6.52 & @xcite & double with a399 + a520 & 0.1990 & 3.25 & 34.4 & 24.58 & 1.11 & 8.30 & @xcite & + a521 & 0.2533 & 3.91 & 5.9 & 24.05 & 1.17 & 8.47 & @xcite & a relic is also present + a523 & 0.1036 & 1.88 & 59.0 & 24.17 & 1.3 & 1.07 & @xcite & + a545 & 0.1540 & 2.64 & 23.0 & 24.16 & 0.89 & 5.55 & @xcite & + a665 & 0.1819 & 3.03 & 43.1 & 24.59 & 1.82 & 9.65 & @xcite & + a697 & 0.2820 & 4.23 & 5.2 & 24.11 & 0.65 & 10.40 & @xcite & + a746 & 0.232 & 3.67 & 18.0 & 24.58 & 0.85 & 3.68 & @xcite & a relic is also present + a754 & 0.0542 & 1.04 & 86.0 & 23.77 & 0.99 & 2.21 & @xcite & a relic is also present + a773 & 0.2170 & 3.48 & 12.7 & 24.23 & 1.25 & 7.95 & @xcite & + a781 & 0.3004 & 4.42 & 20.5 & 24.77 & 1.60 & 4.6 & @xcite & a relic is also present + a851 & 0.4069 & 5.40 & 3.7 & 24.33 & 1.08 & 5.04 & @xcite & + a1213 & 0.0469 & 0.91 & 72.2 & 23.56 & 0.22 & 0.10 & @xcite & + a1300 & 0.3072 & 4.49 & 20.0 & 24.78 & 1.3 & 13.73 & @xcite & a relic is also present + a1351 & 0.3224 & 4.64 & 39.6 & 25.12 & 0.84 & 5.47 & @xcite & + a1656 & 0.0231 & 0.46 & 530.0 & 23.80 & 0.83 & 3.99 & @xcite & a relic is also present + a1689 & 0.1832 & 3.05 & 9.9 & 23.96 & 0.73 & 12.00 & @xcite & + a1758a & 0.2790 & 4.20 & 16.7 & 24.60 & 1.51 & 7.09 & @xcite & total diffuse emission + a1914 & 0.1712 & 2.88 & 64.0 & 24.71 & 1.28 & 10.42 & @xcite & + a1995 & 0.3186 & 4.61 & 4.1 & 24.13 & 0.83 & 8.83 & @xcite & + a2034 & 0.1130 & 2.03 & 13.6 & 23.64 & 0.61 & 3.81 & @xcite & a relic is also present + a2163 & 0.2030 & 3.31 & 155.0 & 25.26 & 2.28 & 22.73 & @xcite & a relic is also present + a2218 & 0.1756 & 2.94 & 4.7 & 23.60 & 0.38 & 5.77 & @xcite & + a2219 & 0.2256 & 3.59 & 81.0 & 25.08 & 1.72 & 12.19 & @xcite & + a2254 & 0.1780 & 2.98 & 33.7 & 24.47 & 0.89 & 4.55 & @xcite & + a2255 & 0.0806 & 1.50 & 56.0 & 23.94 & 0.90 & 2.64 & @xcite & a relic is also present + a2256 & 0.0581 & 1.11 & 103.4 & 23.91 & 0.81 & 3.75 & @xcite & a relic is also present + a2294 & 0.1780 & 2.98 & 5.8 & 23.71 & 0.54 & 3.90 & @xcite & + a2319 & 0.0557 & 1.07 & 153.0 & 24.04 & 1.02 & 8.46 & @xcite & + a2744 & 0.3080 & 4.50 & 57.1 & 25.24 & 1.89 & 12.86 & @xcite & a relic is also present + a3562 & 0.0490 & 0.95 & 20.0 & 23.04 & 0.28 & 1.57 & @xcite & + rx j0107.7 + 5408&0.1066&1.93&55 & 24.26 & 1.10 & 5.42 & @xcite & + cl0016 + 16 & 0.5456 & 6.37 & 5.5 & 24.81 & 0.96 & 19.6 & @xcite & macsj0018.5 + 1626 + cl0217 + 70 & 0.0655 & 1.24 & 58.6 & 23.74 & 0.74 & 0.63 & @xcite & double relics also present + 1e0657 - 56&0.2960 & 4.38 & 78.0 & 25.33 & 2.1 & 22.59 & @xcite & bullet cluster + macsj0717.5 + 3745&0.5458&6.37&118.0 & 26.20 & 1.5&24.46 & @xcite + rxcj1314.4 - 2515&0.2439&3.81&20.3 & 24.55 & 1.60 & 10.75 & @xcite & double relics are present + rxcj1514.9 - 1523&0.22&3.52 & 10.0 & 24.23 & 1.5 & 7.2 & @xcite & 1.4 ghz flux density from nvss + rxcj2003.5 - 2323&0.3171&4.59&35.0 & 25.09 & 1.40 & 9.12 & @xcite & + cizaj2242.8 + 5301&0.1921&3.17& & & 3.1 & 6.8 & @xcite & double relics also present + + + + tab .
[ tab - halo ] reports all clusters known so far to host a radio halo as of september , 2011 .
this sample will be henceforth referred to as september2011-halo collection .
all halo clusters are undergoing merger processes , as demonstrated by the dynamical activity detected from x - ray and optical data ( see sect .
[ sec:432 ] ) .
the occurrence of halos in the nvss sample @xcite increases with the x - ray luminosity of the host clusters , reaching about 25% in clusters with l@xmath20 @xmath19 5 @xmath22 10@xmath23 erg / s @xcite .
this trend is confirmed by the recent work of cassano et al .
@xcite , who , with improved statistics , determined that radio halos are present in @xmath3 30% of clusters with l@xmath20 @xmath19 5 @xmath22 10@xmath23 erg / s . fig .
[ fig : haz ] reports the redshift distribution of clusters with a radio halo .
this distribution is rather homogeneous up to z = 0.35 .
the scarce number of halos beyond this value is likely due to sensitivity limitations , and/or selection effects @xcite , although a real lack of high redshift radio halos can not be excluded .
this might possibly be related with the cluster evolution , or with strong inverse compton losses with the cosmic microwave background radiation .
there are currently three high redshift ( z @xmath19 0.4 ) clusters known to host radio halos : a851 at z = 0.4069 @xcite , cl0016 + 16 @xcite , and macsj0717.5 + 3745 @xcite , both at z = 0.55 .
the latter shows also significant polarization , as also found in a2255 @xcite , unlike most radio halos which are unpolarized .
the physical parameters in radio halos can be estimated assuming equipartition conditions , and further assuming equal energy in relativistic protons and electrons , a volume filling factor of 1 , a low frequency cut - off of 10 mhz , and a high frequency cut - off of 10 ghz .
the derived minimum energy densities are of the order of 10@xmath26 - 10@xmath27 erg @xmath6 , i.e. much lower than the energy density in the thermal gas .
the corresponding equipartition magnetic field strengths range from 0.1 to 1 @xmath8 g .
the integrated radio spectra of halo sources are still poorly known .
the difficulty of spectral studies are that i ) only in a few cases the spectrum is obtained with more than three flux density measurements at different frequencies , ii ) for most sources the highest available frequency is 1.4 ghz , therefore it is difficult to determine the presence of a spectral steepening , crucial to discriminate between different reacceleration models .
spectral data available on the objects of the september2011-halo collection are reported in tab .
[ tab - spec ] , where we arrange the information according to the number of available frequency measurements .
halos always show a steep spectrum ( @xmath28 ) @xmath29 @xmath30 is assumed throughout the paper ] @xmath31 1 ) .
the best studied integrated spectrum is that of the coma cluster where clear evidence of a high frequency steepening is present ( see @xcite for a detailed discussion ) .
indications of high frequency spectral steepening are reported in 4 more cases .
a very steep spectrum is shown by the radio halo in a1914 @xcite where 9 different points show a straight spectrum with @xmath28 = 1.88 , although a possible high frequency curvature has been suggested @xcite .
halo spectra are typical of aged radio sources ( see sect .
[ sec:2 ] ) . in general
, it is estimated that the radiative lifetime of relativistic electrons from synchrotron and inverse compton energy losses is of the order of @xmath3 10@xmath32 yr @xcite .
since the expected diffusion velocity of the electron population is of the order of the alfvn speed ( @xmath3 100 km s@xmath4 ) , the radiative electron lifetime is too short to allow the particle diffusion throughout the cluster volume .
thus , the radiating electrons can not have been produced at some localized points of the cluster .
they must undergo _ in situ _ energization , with an efficiency comparable to the energy loss processes @xcite , or be continuously injected into the cluster volume @xcite . in sect . [ sec:432 ] , we will highlight that recent cluster mergers are likely to supply energy to the halos , through the reacceleration of radio emitting electrons . in the framework of the link between halos and clusters mergers ,
the existence of a possible correlation between radio spectral index and cluster temperature has been investigated in the literature @xcite , and a marginal evidence that clusters at higher temperature tend to host halos with a flatter spectrum has been found . by using the most recent results ( september2011-halo collection ) , and considering the average spectral index value in the frequency range 0.3 - 1.4 ghz , it is obtained that : * radio halos in clusters with an average temperature less than 8 kev show an average spectral index of [email protected] ; * radio halos in clusters with a temperature in the range from 8 to 10 kev show an average spectral index of 1.4@xmath34 ; * halos in clusters with a temperature greater than 10 kev show an average spectral index of [email protected] .
this trend indicates that hotter clusters are likely to host halos with flatter spectra , i.e. halos where energy gains are more significant .
unfortunately , the statistics is still poor , because of the low number of data and highly inhomogeneous spectral index measurements .
this result , if confirmed , is a very important clue in favor of reacceleration models : indeed hottest clusters are more massive and undergoing more violent mergers , thus supplying more energy to the radio emitting particles @xcite .
lcccc name & @xmath281 & @xmath282 & temperature & ref + & & & kev & + a1656 & @xmath35 = 1.16 & @xmath36 = 2.28 & 8.4&@xcite + a1914 & @xmath37 = 1.88 & & 7.9&@xcite + a2256 & @xmath38 = 1.61 & & 6.6&@xcite + a3562 & @xmath39 = 1.3 & @xmath40 = 2.1 & 5.2&@xcite + 1e0657 - 56 & @xmath41 = 1.3 & & 10.6&@xcite + a521 & @xmath42 = 1.8 & & 5.9&@xcite and ref therein + a754 & @xmath43 = 1.1 & @xmath44 = 1.5 & 9.5&@xcite + a2163 & @xmath45 = 1.1 & @xmath46 = 1.5 & 13.3&@xcite + a2319 & @xmath47 = 0.9 & @xmath48 = 2.2 & 8.8&@xcite + rxcj2003.5 - 2323 & @xmath49 = 1.3 & & - & @xcite + a545 & @xmath46 @xmath19 1.4 & & 5.5&@xcite + a665 & @xmath44 = 1.0 & & 9.0&@xcite + a697 & @xmath44 = 1.5 & & 10.2&@xcite + a1300 & @xmath44 = 1.3 & & 9.2&@xcite + a1758 & @xmath44 = 1.5 & & 7.1&@xcite + a2218 & @xmath50 = 1.6 & & 7.1&@xcite + a2219 & @xmath44 = 0.9 & & 12.4&@xcite + a2255 & @xmath44 = 1.7 & & 6.9&@xcite + a2744 & @xmath44 = 1.0 & & 10.1&@xcite + macsj0717.5 + 3745 & @xmath51 = 1.3 & & 11 & @xcite , and ref therein + cl0217 + 70 & @xmath44 = 1.2 & & & @xcite + rxcj1514.9 - 1523 & @xmath44 = 1.6 & & & @xcite + + + + a spectral index image is a powerful tool to understand the physical properties of radio halos , since it reflects the variations of the shape of the electron energy distribution and of the magnetic field in which they emit . the first spectral index map of a radio halo was obtained for the coma cluster @xcite , where the high sensitivity of images allowed the computation of spectral index up to @xmath3 30 from the cluster center .
the spectral distribution is smooth with a steepening from the center to the peripheral regions , where the spectral index is @xmath28 @xmath3 2 .
the spectral index images of giant radio halos in a665 and a2163 , with an angular resolution @xmath3 1 , show flattening and patches , and radial steepening in the undisturbed cluster regions @xcite . in the bullet cluster 1e0657 - 56 ,
no evidence of spectral steepening in the outer regions of the radio halo is found @xcite .
the spectral index distribution in the a3562 radio halo shows a very complex structure , with an average value of @xmath28 @xmath3 1.5 and a number of knots steepening up to @xmath28 @xmath3 2 @xcite .
the spectral index images of a2744 and a2219 @xcite show the presence of localized regions in which the radio spectrum is significantly different from the average . from the comparison of the spectral index map of a2744 ( fig .
[ fig : alfa2744 ] , left panel ) and the x - ray data from chandra , it is found for the first time that the flat spectrum regions of the radio halo tend to have higher temperature ( fig .
[ fig : alfa2744 ] , right panel ) . in the radio halo of a2255 ,
the spectrum is steeper at the center and flatter at the locations of the radio filaments @xcite , indicating either that we are looking at a superposition of different structures ( filaments in the foreground plus real halo in the background ) seen in projection across the cluster center , or that the halo is intrinsically peculiar .
the complex cluster a2256 shows flat spectral index in the region nw of the cluster center , and steep spectral index regions toward the se of the cluster center @xcite .
the authors interpret this spectral behavior as resulting from two populations of relativistic electrons created at two epochs ( two mergers ) .
spectral variations are likely to reflect the energy losses / gains of the radiating electrons .
it can be shown @xcite that , in regions of identical volume and identical radio brightness , a flattening of the spectral index from 1.3 to 0.8 requires an amount of energy injected into the electron population larger by a factor of @xmath3 2.5 .
alternatively , if electrons are no longer gaining energy , a flatter spectrum would indicate that the last reacceleration is more recent .
therefore , it can be argued that the radio spectrum is flatter in the regions influenced by merger processes .
the data available so far are in support of electron reacceleration models , with energy supplied by cluster merger @xcite .
in particular the spectral radial steepening is interpreted as due to the combined effect of a radial decrease of the cluster magnetic field strength and of the presence of a high energy break in the energy distribution of the reaccelerated electron population @xcite . in recent years
thanks to the improvement in low frequency radio observations , a large attention has been devoted to the search for and discussion of ultra - steep spectrum radio halos ( see e.g. @xcite ) . from the synchrotron emission theory , it is derived that strong energy losses in the radiating electrons produce a cutoff at high frequency , while the radio spectrum at low frequency reflects the original energy distribution of electrons
. the existence of radio halos showing very steep spectra also in the low frequency range would indicate objects characterized by dramatic energy losses .
cassano @xcite argues that radio halos with spectral index @xmath28 @xmath3 1.7 , and steeper , are likely missed by present observations at ghz frequencies .
this radio halo population can only be revealed with highly sensitive observations at very low radio frequencies . tab .
[ tab - spec ] shows that only a few ultra - steep spectrum radio halos are currently known : a1914 @xcite , a521 @xcite , and a2255 @xcite .
the cluster a697 was previously classified among ultra - steep spectrum halos @xcite , however according to recent flux measurements @xcite the spectral index is around 1.5 .
these objects may be the tip of the iceberg of a population which should be detected by future instruments at low frequencies ( lofar , lwa , ska ) .
we note that the few ultra - steep radio halos are in relatively low temperature clusters , in agreement with the correlation between radio halo spectral index and cluster temperature ( see sect .
[ sec:411 ] ) .
classical halos , i.e. the first which have been detected , are giant radio sources of about 1 mpc , or more , in size . in recent years , however , several small size halos have been revealed down to few hundred kpc , owing to the improvement in sensitivity and angular resolution of instruments .
giovannini et al .
@xcite reported the correlation between the radio halo linear size and the total radio power , and discussed that this correlation is continuous from small halos up to giant mpc - scale radio halos .
the correlation obtained for the september2011-halo collection is presented in fig .
[ fig : hasize ] .
the slope of the correlation is in agreement with results obtained using only giant radio halos by cassano et al .
murgia et al .
@xcite fitting the azimuthally averaged brightness profile with an exponential , found that radio halo emissivity is remarkably similar from one halo to the other , despite quite different length - scales , in agreement with the correlation in which larger halos show a higher radio power ( see the discussion in sect .
[ sec:43 ] ) .
possible evolutionary effects with redshift can not be analyzed with the current data , because of poor statistics .
there are 3 outliers with respect to the general correlation : one is macs j0717.5 + 3745 @xcite , which is the most powerful radio halo , but not the most extended ; the second is the halo in a1351 , which shows a size smaller than expected from its radio power .
it is an irregular radio halo @xcite and could be related to a recent merger .
the third one is the so far smallest radio halo , located in the poor abell cluster a1213 ( see also sect .
[ sec:431 ] ) .
as discussed in @xcite , more data are necessary to better understand its properties . except for the few cases outlined above , the correlation of radio size versus radio power shows a small dispersion , confirming that giant and smaller radio halos belong to the same class of sources : halos as small as a few hundred kpc show the same properties as mpc size giant halos .
this is in support of a common origin and physical mechanism for giant and small - size radio halos .
the radio properties of halos are linked to the cluster properties .
the probability of detecting radio halos is highest in the clusters with the highest x - ray luminosity ( see sect . [ sec:4 ] ) .
moreover , the monochromatic radio power of a halo at 1.4 ghz correlates with the cluster x - ray luminosity , and mass and temperature . in fig . [
fig : haradiox ] we present the correlation between the halo radio power at 1.4 ghz and the x ray luminosity between 0.1 - 2.4 kev .
we note that the correlation is obtained for merging clusters with radio halos and can not be generalized to all clusters ( see sect .
[ sec:432 ] ) .
the most powerful radio halo is in the cluster macs j0717.5 + 3745 .
it is remarkable that its radio power is about an order of magnitude higher than that of the second most powerful known halo . in the low x - ray luminosity regime
, there are a few clusters well above the extrapolation of the correlation , which will be discussed in the next sub - section . if we exclude the three outliers named in fig .
[ fig : haradiox ] we find p@xmath52 . on the other hand , ultra - steep spectrum radio halos , discussed in sect .
[ sec:413 ] , are expected to be less luminous than predicted by the radio - x - ray correlation , according to cassano @xcite .
an extrapolation of the above correlation to low radio and x - ray luminosities indicates that clusters with l@xmath53 @xmath9 10@xmath23 erg s@xmath4 should host halos of power @xmath21 10@xmath54 w hz@xmath4 . with a typical size of 1 mpc , they would have a radio surface brightness lower than current limits obtained in the literature .
it should be stressed , however , that the less powerful radio halos are also smaller in size ( fig .
[ fig : hasize ] ) , implying that actually it may not be appropriate to assume a typical size @xmath24 1 mpc for low power radio halos .
brunetti et al .
@xcite report the same correlation using all clusters searched for radio halos with the gmrt and derived upper limits to the radio power of undetected halos , by adopting a halo size of 1 mpc .
they found that upper limits are all in the region of high x - ray luminosity , and are all about one order of magnitude below the correlation , indicating a difference between clusters with and without halos .
their conclusion is that the distribution is likely related to the dynamical state of the cluster ( see sect .
[ sec:432 ] ) : radio halos are associated with dynamically disturbed clusters , while upper limits refer to relaxed clusters .
this is in agreement with the result that halos are only present in merging clusters . in fig .
[ fig : haradiox ] only the clusters with halos are reported , thus no upper limits are shown . since cluster x - ray luminosity and mass are correlated @xcite , the correlation between radio power ( @xmath55 ) and x - ray luminosity could reflect a dependence of the radio power on the cluster mass @xmath56 .
a correlation of the type @xmath55 @xmath29 @xmath57 has been derived @xcite , where @xmath56 is the total gravitational mass within a radius of 3(@xmath11/50)@xmath4 mpc .
using the cluster mass within the viral radius , the correlation is steeper @xcite .
another link between radio and x - ray properties is suggested by the close similarity of the radio and x - ray structure found in a number of well resolved clusters .
the similarity was quantitatively confirmed ( see e.g. @xcite ) , by comparing the point - to - point surface brightness of the radio and x - ray emission .
a nearly linear relationship was found in three clusters ( a2255 , a2744 and a3562 ) , while a power - law relation with index @xmath211 is present in coma , a2319 and a2163 .
as discussed in @xcite , the correlation suggests a direct connection between the thermal x - ray plasma and the non - thermal radio plasma ; its slope is that expected by electron reacceleration models .
the radio - x - ray brightness relation discussed above is valid for giant and regular halos .
actually , we have noticed that , owing to recent deep observations , more irregular and asymmetric halos have been found . in these halos ,
the radio emission may show significant displacement from the x - ray emission .
this has been investigated by comparing the center of the radio halo with that of the x - ray gas distribution in a sample of clusters with good radio and x - ray data @xcite .
the left panel of fig .
[ fig : deltarx ] shows that both giant and small radio halos can be significantly shifted , up to hundreds kpc , with respect to the centroid of the host cluster . to highlight radio halos with the most pronounced asymmetric distribution ,
the ratio between the radio - x - ray offset and the halo size is plotted in the middle and right panels of fig .
[ fig : deltarx ] .
it is deduced that halos can be quite asymmetric with respect to the x - ray gas distribution , and this becomes more relevant when halos of smaller size are considered .
a possibility is that the asymmetry in the structure is caused by magnetic field fluctuations as large as hundreds of kpc , as suggested by @xcite on the basis of magnetic field modeling .
giovannini et al .
@xcite were the first to discuss the possible presence of a radio halo in a1213 , an under - luminous x - ray cluster i.e. well outside the correlation between radio power and x - ray luminosity .
a similar case is found in the cluster cl0217 + 70 by brown et al .
@xcite . clear evidence of a radio halo more luminous than predicted by the radio - x - ray correlation is detected in a523 @xcite .
the number of these cases is still low , and some observational uncertainties may be present ( see @xcite for a more detailed discussion ) , however there is observational evidence that in a few cases giant radio halos can be associated with relatively low luminosity x - ray clusters , i.e. low mass clusters and low density environments .
these halos are over - luminous in radio by at least an order of magnitude with respect to what is expected from the extrapolation of the observed radio power - x - ray luminosity .
these objects are likely to represent a new class of halos that are difficult to explain by classical radio halo models .
we note that ultra - steep spectrum halos , presented in sect .
[ sec:413 ] and predicted by reacceleration models , show the opposite behaviour , i.e. they should be less luminous in radio than predicted by the radio - x - ray correlation .
giovannini et al @xcite suggest that these powerful radio halos associated with low density environments could be either young halos or related to clusters at a special time of the merger event , when particle acceleration processes have a higher efficiency ( see e.g. @xcite ) .
another possibility is that the x - ray luminosity might not be a good indicator of the previous history of cluster merging activity in these cases .
circumstantial evidence of the link between cluster merging and radio halos has been outlined in early works on this topic ( see e.g. @xcite ) , on the basis of the presence of radio halos only in clusters showing x - ray and substructure , x - ray temperature gradients and weak ( or absent ) cooling flow . this has been strongly supported by optical data
( see e.g. @xcite ) .
the first quantitative analysis of the cluster dynamical state was performed by buote @xcite , who obtained the cluster dipole ratio from x - ray images and found that the strongest radio halos appear only in those clusters experiencing the largest departures from a virialized state .
cassano et al .
@xcite recently confirmed the connection between radio halos and cluster mergers , from a quantitative study on a larger sample , using chandra x - ray data and adopting three methods to characterize the cluster state , i.e. the power ratio , the centroid shift and the x - ray brightness concentration parameter .
the link between radio halos and cluster merger is well supported by several arguments , and recent findings obtained with improved statistics support the early suggestion that there is a clear separation between relaxed and non relaxed clusters , in the radio behaviour .
the bi - modality between halo and non halo clusters , discussed by @xcite , reflects the dichotomy existing in the evolutionary state of clusters , between non relaxed and relaxed clusters , as also discussed by @xcite
. what is intriguing is the existence of dynamically disturbed clusters which do not show any evidence of radio halo .
cassano et al .
@xcite find 4 such clusters : a141 , a781 , a2631 and macs j2228.5 + 2036 .
a 781 has been found to host a radio halo @xcite , but there are still 3 outliers ( out of them one cluster is at z @xmath19 0.32 ) .
some other well known merging clusters do not host a radio halo , e.g. a119 @xcite and a2146 @xcite .
the existence of merging clusters with halos , and merging clusters without halo , indicates that there is a dichotomy in merging clusters , which is not currently understood .
it may be possible that giant halos are only present above a threshold of mass or temperature .
future radio data with next generation instruments ( lofar , lwa , ska ) will allow the detection of low brightness / low power large halos , in order to clarify if halos are present in all merging clusters or only in the most massive ones .
the dichotomy has to be solved to get a full and comprehensive understanding of these phenomena .
the properties of relics have been found to be related to the properties of their parent clusters . indeed , similarly to radio halos , a correlation is found between the relic radio power at 1.4 ghz and the x - ray luminosity of clusters ( fig . [
fig : relradiox ] ) .
the correlation holds for both elongated and roundish relics , although with somewhat large dispersion ; no systematic difference is visible between the two classes of relics .
we find p@xmath63 .
the similarity between the properties of halos and relics has been already noted in the past ( e.g. @xcite ) . in fig .
[ fig : harel ] we show the halo and relic radio power as a function of the x - ray cluster luminosity ( left panel ) , and as a function of their linear size ( right panel ) . for clarity
, we use different symbols for halos , elongated relics and roundish relics . the most powerful object in the plots is the radio halo in the cluster macs j0717.5 + 3745 .
it turns out that halos and relics show a very similar behaviour in both plots , supporting the connection between halos and relics , and their link to cluster mergers . from the left panel of fig .
[ fig : harel ] , we note that the most powerful halos and relics show similar values in radio power and cluster x - ray luminosity ( with the exception of macs j0717.5 + 3745 ) , with similar dispersion . at lower radio powers
, there is a large dispersion in both classes , and in particular it is interesting to note that , while halos in poor x - ray clusters are only a few ( sect .
[ sec:431 ] ) , relics are easily detected also in clusters of low x - ray luminosity . indeed , elongated and roundish relics fill the region of radio power p@xmath64 10@xmath65 and cluster x - ray luminosity l@xmath66 3 @xmath22 10@xmath23 .
the flatter correlation found for relics with respect to halos likely reflects this behaviour .
halos and relics show similar behaviour also in the plot of radio power versus largest linear size ( right panel of fig .
[ fig : harel ] ) . as for the previous correlation
, halos show a lower dispersion in the data , and at a given radio power they tend to be more extended than relics .
we finally note the peculiarity of macs j0717.5 + 3745 @xcite ( see sects .
[ sec:42 ] and [ sec:43 ] ) , which , even with the addition of relics , remains the most powerful diffuse radio source , with a radio power about an order of magnitude higher than that of other known diffuse sources .
relic sources are diffuse extended sources similar to radio halos in their low surface brightness , large size ( @xmath24 1 mpc ) and steep spectrum ( @xmath28 @xmath24 1 ) , but , unlike halos , they are located in cluster peripheral regions and are strongly polarized ( @xmath3 20 - 30% ) .
observations of relics provide the best indications for the presence of @xmath8 g level magnetic fields and relativistic particles in cluster outskirts . they provide evidence for the acceleration of relativistic particles at shock fronts at large distance ( mpc scale ) from the cluster centers ( see e.g. @xcite ) .
the prototype source of this class is the relic 1253 + 275 detected in the coma cluster ( see @xcite and references therein ) , shown in fig . [
fig : comarelic ] . the coma relic and
most relics show an elongated radio structure , with the major axis roughly perpendicular to the direction of the cluster center . recently
, new spectacular relics have been detected and studied , increasing our knowledge of these diffuse sources .
we quote as one of the most interesting cases , the giant narrow relic in ciza j2242.8 + 5301 ( sect .
[ sec:511 ] ) .
in addition , sources with a more roundish structure , as well as sources of small size , have been detected and classified as cluster relics .
different relic morphologies and related properties will be discussed in the next sub - section .
relics are found in both merging and cool - core clusters , suggesting that they may be related to minor or off - axis mergers , as well as to major mergers .
theoretical models propose that they are tracers of shock waves in merger events .
this is consistent with their elongated structure , almost perpendicular to the merger axis , and is confirmed by observational results ( see e.g. @xcite , and the review by brggen et al .
@xcite ) .
lclrrrrrcl name & z & kpc/@xmath25 & s(1.4 ) & log p(1.4 ) & rcc & lls & lx(10@xmath23 ) & ref & notes + & & & mjy & w / hz & mpc & mpc & erg / sec & & + a13 & 0.0943 & 1.67 & 31.0 & 23.80 & 0.20 & 0.25 & 1.18 & @xcite & 2 + a85 & 0.0551 & 1.01 & 43.0 & 23.44 & 0.43 & 0.35 & 4.10 & @xcite & 2 + a115 & 0.1971 & 3.13 & 147.0 & 25.18 & 1.32 & 2.44 & 8.91 & @xcite & 1 ; average rcc + a521 & 0.2533 & 3.81 & 15.0 & 24.41 & 0.93 & 1.00 & 8.47 & @xcite & 1 + halo + a548b - nw & 0.0424 & 0.81 & 61.0 & 23.38 & 0.43 & 0.26 & 0.15 & @xcite & 2 + a548b - n & & & 60.0 & 23.37 & 0.50 & 0.31 & 0.15 & @xcite & 2 + a610 & 0.0954 & 1.71 & 18.6 & 23.60 & 0.31 & 0.33 & &
@xcite & 2 + a746 & 0.232 & 3.67 & 24.5 & 24.83 & 1.61 & 1.10 & 3.68 & @xcite & 1 + halo + as753 & 0.014 & 0.29 & 460.0 & 23.32 & 0.41 & 0.35 & & @xcite & 2 + a754 & 0.0542 & 1.03 & 69.0 & 23.67 & 0.50 & 0.80 & 2.21 & @xcite & 1 ; complex + halo + a781 & 0.3004 & 4.42 & 15.5 & 24.65 & 0.66 & 0.44 & 4.6 & @xcite & 1 + halo + a1240-n & 0.1590 & 2.65 & 6.0 & 23.59 & 0.70 & 0.65 & 1.00 & @xcite & 1 + a1240-s & & & 10.1 & 23.81 & 1.10 & 1.25 & 1.00 & @xcite & 1 + a1300 & 0.3072 & 4.39 & 20.0 & 24.76 & 0.79 & 0.70 & 13.73 & @xcite & 1 + halo + a1367 & 0.0220 & 0.45 & 35.0 & 22.59 & 0.62 & 0.22 & 0.64 & @xcite & 2 + a1612 & 0.179 & 2.99 & 62.8 & 24.90 & 0.89 & 0.78 & 2.41 & @xcite & 1 + a1656 & 0.0231 & 0.47 & 260.0 & 23.49 & 2.20 & 0.85 & 3.99 & @xcite & 1 + halo + a1664 & 0.1283 & 2.22 & 107.0 & 24.64 & 1.03 & 1.07 & 3.09 & @xcite & 2 + a2034 & 0.1130 & 2.03 & 24.0 & 23.95 & 0.73 & 0.22 & 3.81 & @xcite & 1 + halo + a2048 & 0.0972 & 1.74 & 19.0 & 23.63 & 0.33 & 0.31 & 1.91 & @xcite & 2 + a2061 & 0.0784 & 1.46 & 27.6 & 23.65 & 1.56 & 0.68 & 3.95 & @xcite & 1 + a2063 & 0.0349 & 0.68 & 67.0 & 23.26 & 0.04 & 0.04 & 0.98 & @xcite & 2 ( 3c318.1 ) + a2163 & 0.2030 & 3.22 & 18.7 & 24.32 & 1.17 & 0.48 & 22.73 & @xcite & 1 + a2255 & 0.0806 & 1.46 & 12.0 & 23.25 & 0.90 & 0.70 & 2.64 & @xcite & 1 + halo + a2256 & 0.0581 & 1.08 & 462 & 24.56 & 0.44 & 1.13 & 3.75 & @xcite & 2 + halo + a2345-w & 0.1765 & 2.86 & 30.0 & 24.38 & 1.00 & 1.15 & 5.90 & @xcite & 2 + a2345-e & & & 29.0 & 24.36 & 0.89 & 1.50 & 5.90 & @xcite & 1 + a2443 & 0.1080 & 1.88 & 6.5 & 23.25 & 0.23 & 0.43 & 1.90 & @xcite & 2 ; complex cluster + a2744 & 0.3080 & 4.37 & 18.2 & 24.71 & 1.56 & 1.62 & 12.86 & @xcite & 1 + halo + a3365-e & 0.0926 & 1.66 & 42.6 & 24.04 & 1.08 & 0.56 & 0.86 & @xcite & 1 + a3365-w & & & 5.3 & 23.15 & 0.5 & 0.24 & 0.86 & @xcite & 1 + a3376-w & 0.0456 & 0.87 & 166.0 & 23.88 & 1.43 & 0.80 & 1.08 & @xcite & 1 + a3376-e & & & 136.0 & 23.79 & 0.52 & 0.95 & 1.08 & @xcite & 1 + a3667-nw & 0.0556 & 1.03 & 2400.0 & 25.21 & 2.05 & 1.86 & 4.73 & @xcite & 1 + a3667-se & & & & & 1.36 & 1.30 & 1.37 & @xcite & 1 + a4038 & 0.0300 & 0.56 & 49.0 & 22.94 & 0.04 & 0.13 & 1.97 & @xcite & 2 + zwcl0008.8 + 5215-w&0.1032&1.82&11.0&23.44&0.85 & 0.29 & 0.50 & @xcite & 1 + zwcl0008.8 + 5215-e & & & 56.0&24.15&0.85 & 1.40 & 0.50 & @xcite & 1 + s1081 & 0.2200 & 3.52 & 2.4 & 23.17 & 2.00 & 0.28 &
& @xcite & 1 + cl0217 + 70-nw&0.0655 & 1.26 & & & 1.00 & 0.95 & 0.63 & @xcite & 1 ; confused + cl0217 + 70-se & & & & & 1.00 & 0.38 & 0.63 & @xcite & 1 ; confused + cizaj0649.3 + 1801&0.064&1.22&125&24.08 & 0.80 & 0.8 & 2.38 & @xcite & 1 + rxcj1053.7 + 5452&0.0704&1.33&15.0&23.30 & 0.98 & 0.60 & 3.69 & @xcite & 1 + rxcj1314.4 - 2515-e&0.2474&3.76&10.1&24.27&0.90 & 0.56 & 10.75 & @xcite & 1 + halo + rxcj1314.4 - 2515-w & & & 20.2&24.57&0.55 & 1.01 & & @xcite & 1 + halo + cl1446 + 26 & 0.3700 & 5.08 & 9.2 & 24.61 & 0.66 & 0.36 & 3.42 & @xcite & 1 + cizaj2242.8 + 5301-n&0.1921&3.07&& & 1.57 & 1.70 & 6.80 & @xcite & 1 + halo + cizaj2242.8 + 5301-s & & & & & 1.06 & 1.45 & 6.80 & @xcite & 1 + halo + plck g287.0 + 32.9-n&0.39&5.26&33.0&25.24 & 1.58 & 1.4 & 17.2 & @xcite & 1 + plck g287.0 + 32.9-s & & & 25.0&25.12 & 3.00 & 1.1 & 17.2 & @xcite & 1 + + + + + currently we know 50 relics hosted in 39 clusters of galaxies : they are reported in tab .
[ table - rel ] .
this collection , which is updated to september , 2011 , will be referred to henceforth as september2011-relic collection .
most clusters are rich abell clusters , but relics have also been detected in x - ray selected clusters ( e.g. rxs j131423.6 - 251521 , @xcite ) , and in poor clusters ( e.g. s0753 , @xcite ) .
the classification of relics may be troublesome in some cases .
this may cause small differences from other lists in the literature ( see e.g. @xcite ) .
in particular , a133 is excluded from the september2011-relic collection of tab .
[ table - rel ] after the discussion given in @xcite , who suggest that the source formerly classified as a relic is one of the two lobes of a double radio galaxy , identified with the central cd galaxy .
moreover , we do not consider the relics in zwcl 2341.1 + 0000 , since the diffuse radio emission in this cluster is likely originating from a large - scale filament @xcite ( see sect .
[ sec:8 ] and fig .
[ fig : filament ] ) , rather than from double relics @xcite .
we also note that diffuse sources which arise from agn , in which the nuclear activity has ceased , may be sometimes classified as cluster relics .
actually , these sources are not originating from the icm , and should be more properly considered as _ dying sources _ or _ agn relics _ ( see @xcite ) .
they will not be considered here because they are not originated by the icm .
the occurrence of relics in the nvss sample @xcite is higher in clusters showing high x - ray luminosity , reaching about 25% in clusters with l@xmath20 @xmath19 5 @xmath22 10@xmath23 erg / s @xcite .
the redshift distribution of clusters of galaxies , hosting at least one relic , is given in fig .
[ fig : relz ] , where different colors refer to different structures ( sect . [ sec:51 ] ) . as discussed by @xcite
, relic sources at very low redshift could have a large angular extension , and therefore may be missed in interferometric observations because of the poor coverage of short spacings .
however , owing to the irregular and elongated structure of most relics , this observational problem is not as crucial as in the more regular and diffuse halo sources . relics are detected up to z @xmath3 0.4 , and their distribution with redshift is similar to that of radio halos . the low number of high redshift relics could be due to selection effects , but could also be a real trend , due e.g. to the evolution of cluster mergers , or to the inverse compton losses which become more relevant at increasing redshift .
deeper radio surveys and cluster catalogues are necessary to investigate this point . after the discovery of the prototypical relic 1253 + 275 in the coma cluster ( fig .
[ fig : comarelic ] ) , different morphologies have been found to be associated with relics ( see e.g. @xcite ) : classical sources are elongated and located in the periphery of galaxy clusters , but more regular and roundish structures have also been detected , which are located nearer the cluster center and/or in peripheral regions .
kempner et al .
@xcite , in a classification of cluster radio sources , suggested that radio relics could be distinguished in two classes : those associated with a previous agn activity , called radio phoenix , and those related to the icm , named radio gischt .
this classification has been used somewhat in the literature .
however , it implies that the origin of a relic source has to be established , in order to properly classify it . to be conservative
, we prefer to rely on a classification that is based only on the morphology , which is an observable .
therefore we divide relics in _ elongated _ and _ roundish _ , and obtain the properties of these two classes , to find out whether they may be indications of different origin and physical parameters . in all the following plots , as done in fig .
[ fig : relz ] , we will use different symbols for the elongated and roundish relics .
elongated relics are the classical extended objects , located in cluster peripheral regions , characterized by an elongated shape roughly perpendicular to the cluster center direction , as the relic 1253 + 275 in the coma cluster ( @xcite ; fig .
[ fig : comarelic ] ) .
they do not show any evident substructures , and in some case their transverse size is very small .
when observed with high angular resolution , they show an asymmetric transverse profile , with a sharp edge usually on the side toward the cluster outer edge and the radio emission is usually highly polarized .
a spectacular example , detected very recently , is represented by the northern relic in cizaj2242.8 + 5301 @xcite .
this giant relic is very narrow and slightly curved following the cluster boundary .
it is strongly polarized , with highly regular polarization vectors , and shows steep spectrum ( see fig .
[ fig : sausage ] ) . to the south of the cluster , at about 2.8 mpc distance ,
a second ( fainter ) relic structure is present .
we note that projection effects may be important in these elongated structures .
an extreme , and unique , case in the double cluster a115 @xcite , where the relic apparently starts at the center of the northern cluster and is elongated towards the cluster periphery ( fig .
[ fig : relstructures ] , left panel ) .
these morphologies are in very good agreement with models predicting that these sources are related to large - scale shocks generated during cluster merger events .
indeed , these shocks expanding with high velocity ( mach number @xmath3 1 - 3 ) can accelerate electrons to high energies , and compress magnetic fields , giving rise to large regions emitting synchrotron radiation .
the accelerated particles will have a power - law energy distribution , and magnetic fields aligned parallel to the shock front .
this is also consistent with the observed spectral index ( see sect . [ sec:53 ] and fig .
[ fig : sausage ] , right panel ) and polarization properties , and explain the shape and location of these sources ( see e.g. @xcite , and references therein ) .
in addition to elongated relics , diffuse extended radio sources with a more regular and roundish structure have been detected off - center in clusters : an example is in a1664 @xcite , shown in the right panel of fig .
[ fig : relstructures ] . despite their different structure
, these relics share with the elongated relics the location at the cluster periphery and the lack of an optical identification .
the main properties of roundish relics are the diffuse morphology resolved in filamentary sub - structures and a very steep curved spectrum ( see e.g. @xcite ) .
the detection of roundish relics might lead to the obvious interpretation that these sources would be elongated relics seen face - on .
this is however very unlikely , indeed the number of relics with elongated shape is too high to be consistent with simple projection effects .
moreover , we will show in the following and in the next subsections that the properties of elongated and roundish relics may be different .
the class of roundish relics defined here includes the sources named radio phoenix by @xcite , or relic sources near the first ranked galaxy ( frg ) by @xcite .
these objects are located near the central frg , usually a cd galaxy , but not coincident with it .
their size is in a large range of values , i.e. from @xmath9 100 kpc ( as in a2063 and a4038 ) to @xmath3 350 kpc ( as in a85 ) .
these sources , when observed at higher resolution , show evident filamentary substructures ( e.g. @xcite ) . owing to the proximity of these sources to an agn
, an obvious interpretation is that these sources consist of old radio lobes originated by previous agn activity , which have become no longer visible because of strong radiation losses , and have been revived by the energy supplied by shock waves .
a caveat to this interpretation is that these relics are always located only on one side with respect to the frg .
we note , however , the case of a133 , which was classified in the past as a relic ( radio phoenix ) but has been recently shown to be an old radio galaxy @xcite
. another possible explanation of the nature of these sources has been supplied by mathews & brighenti @xcite , who , in their study of the connection between x - ray cavities and radio lobes , suggest that radio bubbles , blown into the cluster gas by a radio galaxy will rise buoyantly and may eventually expand to the cluster outskirts , where the cosmic rays would impact into the surrounding medium , giving rise to small relics located near the agn . among relics with a roundish shape , there are also some objects which are located at large distance from the frg or any bright galaxy , therefore difficult to reconcile with any model involving a previous radio activity : we already noted a1664 shown in the right panel of fig .
[ fig : relstructures ] ; another case is in a548b , where two roundish relics are present , on the same side of the cluster @xcite ( see fig . [
fig : clcollection ] ) .
more data are necessary to understand the origin and nature of these objects .
interestingly , we note that roundish relics are peaked at lower redshift with respect to the total relic redshift distribution and that no roundish relics at z @xmath19 0.2 are known ( fig .
[ fig : relz ] ) . in the framework that relics are witnessing the presence of shock waves originated by mergers between clusters with approximately equal masses and low impact parameter
, it is expected that relics should often come in pairs and be located on opposite sides of the cluster along the axis merger , with the extended radio structures elongated perpendicularly to this axis ( e.g. @xcite ) .
the first cluster where two almost symmetric relics have been detected to be located on opposite sides with respect to the cluster center is a3667 @xcite .
several other objects have been found recently , so these cases are getting more and more common , see e.g. a3376 ( fig .
[ fig : a3376 ] , @xcite ) , rxcj1314.4 - 2515 @xcite , a1240 and a2345 @xcite , cizaj2242.8 + 5301 @xcite , zwcl 0008.8 + 5215 @xcite , plck g287.0 + 32.9 @xcite and cl0217 + 70 ( see fig.[fig : cl0217 ] and @xcite ) . in the case of cizaj2242.8 + 5301 , van weeren at el .
@xcite discuss , from hydrodynamical simulations and comparison with the data , that the relic morphology arises naturally from shocks produced by a head - on merger in the plane of the sky of two roughly equal mass clusters . in the clusters a1240 and a2345 , optical data support the outgoing merger shocks models @xcite . in the case of a3376 ( fig .
[ fig : a3376 ] ) , bagchi and collaborators @xcite suggested the possibility that the two relics may be tracing shocks induced by the accretion flows of the intergalactic medium during the large - scale structure formation ( accretion shocks @xcite ) .
this interpretation is under debate for this cluster ( e.g. @xcite ) , however it may be invoked to explain the origin of other sources as e.g. the outermost relics in a2255 @xcite and possibly the peripheral emission of the coma cluster @xcite . as the coma cluster
, several other clusters have been found to show a peripheral relic radio emission in addition to a central halo . according to the september2011-relic collection , at present 11 clusters
are known to show this feature .
sometimes , the relic is connected to the radio halo through a low brightness bridge of radio emission ( e.g. coma cluster , a2255 and a2744 ) , other cases are characterized by complex structures , as in a754 ( fig .
[ fig : clcollection ] ) and a2256 . in a few case
, clusters may host a central radio halo and double elongated relics . at present we know 3 clusters showing these features : cl0217 + 70 ( fig .
[ fig : cl0217 ] ) , rxcj1314.4 - 2515 ( fig .
[ fig : clcollection ] ) , and cizaj2242.8 + 5301 .
these structures are the most extreme examples of the connection between halo sources and cluster mergers ( see sect .
[ sec:432 ] ) .
indeed the merger that gives rise to the radio halo , could also be the origin of shock waves which supply the energy to the peripheral relics . in this scenario ,
the origin of the bridge of radio emission connecting the halo to the relic source in some clusters has still to be clarified .
we note that in most clusters showing at least a relic source , no central halo is present .
this could suggest that relics could be created also by minor mergers , which are not strong enough to produce a radio halo at the cluster center .
this is in agreement with the presence of relics in a few cool - core clusters , since a minor and off - center merger could create a peripheral shock wave , without destroying the central cool - core . as discussed by van weeren et al .
@xcite , double relics should be located in merging systems with a small mass ratio , while single relics should be found in mergers with larger mass ratios . to better understand the properties of relics , we derived correlations between their observational radio properties and their location in the cluster , using the data of the september2011-relic collection .
double relics in the same clusters are included separately . in the plots , we use different symbols for the elongated and roundish relics , and find that in general elongated and roundish relics are rather separated , with the exception of a few sources . in the left panel of fig .
[ fig : reldist ] , the distribution of projected distance of relics from the cluster center is presented .
we note that roundish sources are more concentrated toward the cluster center , as expected since this class includes the relics located near the frg ( see sect .
[ sec:512 ] ) .
their distribution is different from that of elongated relics , with an average projected distance from the cluster center r@xmath58 @xmath3 0.4 mpc suggesting that most of them are really at a small distance from the center .
elongated relics are mostly located between 0.5 and 1.5 mpc from the center , up to a distance of 3 mpc .
they avoid cluster centers and show a distribution peaked at the cluster periphery . from the figure
it can be deduced that projection effects are unlikely to play a crucial role , i.e. elongated relics are actually located in the cluster outskirts .
this may be due to the fact that high mach number shocks needed to supply energy to the radio emitting particles are only present in cluster peripheral regions @xcite .
this point has been recently discussed in detail by vazza et al .
@xcite .
the monochromatic radio power of relics as a function of the projected distance from the cluster center is given in the right panel of fig .
[ fig : reldist ] .
no clear correlation is present .
roundish relics are less powerful than elongated relics , except for a few cases .
it is derived that relics , both elongated and roundish , located at distances @xmath24 1 mpc from the cluster center , show powers in the range from 10@xmath54 to 10@xmath59 w / hz , whereas there is a lack of powerful ( @xmath19 10@xmath60 w / hz ) relics near the cluster center ( @xmath21 0.4 mpc ) , which could be related to the different relic morphology and to the low efficiency of shocks in cluster central regions ( see e.g. @xcite ) . the left panel of fig .
[ fig : relsize ] gives the largest linear size of relics as a function of their projected distance from the cluster center ( see also @xcite ) .
as in the plot of monochromatic radio power versus projected distance ( fig .
[ fig : reldist ] , right panel ) , elongated and roundish relics are well separated , except for a few roundish objects , the largest ones , which are in the same region of the plot as the elongated relics . the monochromatic radio power of relics at 1.4 ghz versus the relic largest linear size ( lls ) is presented in the right panel of fig .
[ fig : relsize ] .
elongated relics show a similar behaviour to the roundish relics , although they are somewhat separated on the plot , being on average more powerful and more extended . in particular , relics with low radio powers ( @xmath21 10@xmath60 w / hz ) can be either roundish or elongated , but the latter have larger size .
the most powerful relics are mostly large elongated relics . by considering elongated and roundish relics together
, a marginal correlation could be present between lls and projected distance from the cluster center , in the sense that larger relics are at larger distance from the cluster center .
the summary of the above plots is as follows : \1 ) roundish relics are more often located near the cluster center , while elongated relics are more distributed in peripheral regions .
\2 ) the properties of roundish and elongated relics are generally rather different . while roundish relics are on average of low power and small size , elongated relics cover a large interval of radio powers and sizes , and are on average more powerful and more extended .
only a few roundish relics show sizes and radio powers similar to those of elongated relics : these objects are the relics in a1664 , a2256 and a2345 ( western ) .
\3 ) in central cluster regions ( @xmath21 0.4 mpc ) , high power ( @xmath19 10@xmath60 w / hz ) relics are missing . in peripheral regions , on the contrary , there are both small and large size relics , as well as low and high power objects . in general relic properties
are in agreement with current models , where shocks are more efficient in low density peripheral regions , where they would give rise to larger and more powerful relics .
spectral data for relics are available for several objects .
indeed , the elongated structure of most relic favours their detection with interferometric observations , moreover their position at the cluster periphery , where the number of confusing sources is lower , allows more easily observations at different frequencies . in tab .
[ rel - alpha ] we present the available spectral index information for the objects of the september2011-relic collection . for homogeneity with radio halos ,
we report an average spectral index value in the range @xmath3 327 - 1415 mhz , and refer to single papers on the cluster of galaxies hosting a relic ( see tab .
[ table - rel ] ) for more detailed information .
the note in the last column of the table indicates the spectral shape , in the case that flux density measurements are available at more than 2 frequencies . similarly to radio halos ,
relics show quite steep spectra .
this is evident from the spectral index distribution presented in the left panel of fig .
[ fig : alpha-1 ] . in this plot , as in the following plots , the lower limit for a548b - b ( @xmath61 2 ) is not shown .
elongated relics show spectral index @xmath28 in the range 1 1.6 , with an average value of 1.3 .
spectral indices of roundish relics are in a much larger range of values : @xmath28 = 1.1 2.9 , with an average value of 2.0 .
therefore , there is a remarkable difference in spectral index properties between elongated and roundish relics , indeed spectra of roundish relics are steeper than those of elongated relics , with very few exceptions .
it is interesting to note that among the roundish relics with @xmath28 @xmath21 1.6 , there are the 3 objects , whose properties have been outlined in sect .
[ sec:52 ] , characterized by power , size and distance from the cluster center similar to those of elongated relics , i.e. a1664 ( @xmath28 = 1.1 ) , a2256 ( @xmath28 = 1.2 ) , and a2345w ( @xmath28 = 1.5 ) . for objects with measurements at more than two frequencies , the radio spectrum is straight for elongated relics , while it shows a high frequency steepening for roundish relics .
lrcl name & @xmath28 & type & notes + a13 & 2.3 & roundish & steepening + a85 & 2.9 & roundish & steepening + a115 & 1.2 & elongated & + a521 & 1.5 & elongated & straight + a548b - a & 2.0 & roundish & + a548b - b & @xmath192 & roundish & + a610 & 1.4 & roundish & + as753 & 2.1 & roundish & steepening + a754 & 1.5 & elongated & complex + a781 & 1.2 & elongated & + a1240n & 1.2 & elongated & + a1240s & 1.3 & elongated & + a1300 & 1.1 & elongated & + a1367 & 1.9 & roundish & + a1612 & 1.4 & elongated & straight + a1664 & 1.1 & roundish & + a1656 & 1.2 & elongated & straight + a2048 & 1.7 & roundish & steepening + a2061 & 1.0 & elongated & + a2063 & 2.9 & roundish & steepening + a2163 & 1.0 & elongated & + a2255 & 1.4 & elongated & + a2256 & 1.2 & roundish & complex + a2345w & 1.5 & roundish & + a2345e & 1.3 & elongated & + a2443 & 2.8 & roundish & steepening + a2744 & 1.1 & elongated & + a3667nw & 1.1 & elongated & straight + a4038 & 2.2 & roundish & steepening + zwcl0008-w & 1.5&elongated & straight + zwcl0008-e & 1.6&elongated & straight + cl0217 + 70-nw & 1.4&elongated & + cl0217 + 70-se & 1.5&elongated & + rxcj1314-e & 1.2&elongated & + rxcj1314-w & 1.4&elongated & + cizaj2242-n&1.1&elongated &
straight + plck g287.0 + 32.9-n&1.3&elongated & + plck g287.0 + 32.9-s&1.5&elongated & + + + in the right panel of fig .
[ fig : alpha-1 ] , we show the relic spectral index versus the relic projected distance from the cluster center .
we note that relics with steepest spectra are at small distances from the cluster center . from the plots of the spectral index versus the relic power , and of the spectral index versus the lls , presented in the 2 panels of fig .
[ fig : alpha-2 ] , it is derived that most powerful ( p@xmath62 @xmath19 10@xmath60 w / hz ) and most extended ( lls @xmath19 0.5 mpc ) relics are characterized by less steep spectra ( @xmath28 @xmath21 1.6 ) .
smaller / low power relics have a spectral index in the range @xmath3 1.0 2.8 ( see also @xcite ) .
these results are in agreement with the expected origin of all relic sources from strong shock waves : in peripheral regions they should have a larger mach number and therefore they should produce flatter spectra ( see e.g. @xcite ) .
spectral index maps have been obtained so far only for a few relics .
elongated relics show a well defined spectral index distribution , with a transverse steepening toward the cluster center .
this steepening is clearly detected e.g. in a2744 @xcite ( fig .
[ fig : alfa2744 ] , left panel ) , and in cizaj2242.8 + 5301 @xcite ( fig .
[ fig : sausage ] , right panel ) .
this trend is in agreement with the relic brightness profile and strongly supports the shock model ( see e.g. @xcite for a recent discussion ) . among roundish sources ,
a spectral index map has been obtained for a2256 @xcite , which shows spectral steepening from nw to se . in other clusters ,
available images of the spectral index do not show any obvious distribution .
the radio morphology at high resolution is filamentary and the spectral index distribution is irregular . as outlined in sect .
[ sec:531 ] , roundish relics may exhibit extremely steep spectra ( @xmath28 @xmath24 2 ) .
examples of these sources are in a13 , a85 , a4038 ( see @xcite ) , and a2443 ( see @xcite ) .
the extremely steep spectra also show a high frequency cutoff .
these objects have different properties from the giant elongated relics , being on average less powerful and less extended .
the number of these sources is very small
. a reason could be that they are the tip of the iceberg of a population of ultra - steep relics that may be detected at low frequency , similarly to the ultra - steep halos illustrated in sect .
[ sec:413 ] .
alternatively , we could also consider possible selection effects : indeed small size relics which are marginally resolved may be often classified as background discrete radio galaxies .
we can argue that the nature of roundish , ultra - steep relics is different from that of other relics , either elongated or roundish , and that the spectral index is likely a key parameter for the classification of relics .
some relaxed , cool - core , galaxy clusters exhibit signs of diffuse synchrotron radio emission that extend far from the dominant radio galaxy at the cluster center , forming a so - called mini - halo .
mini - halos are extended on a moderate scale ( @xmath67 500 kpc ) and in common with halos and relics they have a low surface brightness and a steep spectrum .
the diffuse source in the ophiucus cluster @xcite , reported in fig .
[ fig : ophiucus ] , is an example of a mini - halo .
mini - halos are sometimes considered small versions of radio halos , but their classification is more complex and fuzzy .
their emission originates from relativistic particles and magnetic fields which are believed to be deeply mixed with the thermal intracluster gas . as a general guideline
, this physical characteristic can be used to distinguish mini - halos from other kind of steep spectrum radio sources in clusters such as radio bubbles related to agn activity .
we shall not define as mini - halos those radio sources in which the ambient thermal gas is clearly separated by the non - thermal plasma , as in the case of agn radio lobes whose expansion has created cavities or holes in the intracluster x - ray emission .
an example of this situation is given by the radio source 3c317 at the center of a2052 ( e.g. @xcite ) .
3c317 is composed of a compact core surrounded by an amorphous halo
. this kind of radio morphology is also referred to as a _ core - halo _ ( see e.g. @xcite ) .
although the radio emission at the center of a2052 resembles on a smaller scale the typical mini - halo , almost all of the radio emission detected so far is contained within cavities devoid of x - ray emitting gas .
therefore , this type of radio sources will not be considered here .
the perseus cluster is a different case .
the radio source 3c84 at the center of perseus shows x - ray cavities in the inner region where the agn radio activity interacts with the thermal gas ( e.g. @xcite ) , but on larger scale the cluster exhibits a diffuse radio emission mixed with the thermal intracluster gas .
indeed perseus cluster represents the prototype example of a mini - halo ( e.g. @xcite ) .
when the cluster x - ray information is lacking , as for the diffuse emission found in the system mrc 0116 + 111 @xcite , the mini - halo classification can be particularly troublesome .
in addition to the above mentioned difficulties in the mini - halo classification , the relatively small angular size of mini - halos in combination with a possibly strong emission of the central radio galaxy , complicates their detection .
thus , our current observational knowledge on mini - halos is limited to a few clusters .
continued searches have so far revealed no more than 11 mini - halos .
a list of mini - halos is given in tab .
[ tab - minih ] .
this collection , which is updated to september , 2011 , will be referred to henceforth as september2011-mini - halo collection .
other mini - halos candidates have been found in a1068 , and a1413 ( @xcite ) , although to be classified as mini - halos they would require further investigation .
the spectra of mini - halos are steep , as well as for halos , with some evidence of radial spectral steepening ( e.g. perseus @xcite , a2626 @xcite , ophiucus @xcite ) .
the origin of mini - halos is still poorly known .
gitti et al .
@xcite argued that the radio emitting particles in mini - halos can not be connected to the central radio galaxy in terms of particle diffusion .
they proposed that mini - halos result from a relic population of relativistic electrons reaccelerated by mhd turbulence via fermi - like processes , the necessary energetics being supplied by the cool - core region .
this is supported by the correlation observed between the mini - halo radio power and the cooling rate power @xcite .
the analysis of the mini - halo present in the most x - ray luminous cluster rxj1347.5 - 1145 @xcite , suggests that additional energy for electron reacceleration in mini - halos might be provided by sub - cluster mergers that have not been able to destroy the central cluster cool core . in this cluster , the spatial correspondence between a radio excess of the mini - halo emission and a hot region , detected through both sunyaev - zeldovich effect ( see fig . [
fig : minihalo ] ) and x - ray observations , indicates that electron reacceleration in this location is most likely related to a shock front propagating into the intra - cluster medium @xcite .
in addition to rxj1347.5 - 1145 , other clusters of galaxies like ophiuchus @xcite and a2142 @xcite present a radio mini - halo , as well as a cool core that has likely survived a possible recent merging event . indeed , although cool - core clusters are generally considered relaxed systems , when analyzed in detail they sometimes reveal peculiar x - ray features in the cluster center which may be indicative of a link between the mini - halo emission and some minor merger activity .
finally , cold fronts associated to the core gas sloshing have been observed in some clusters hosting a mini - halo .
thus , the possibility has been advanced that gas sloshing may generate turbulence in the core , which in turn may reaccelerate the relativistic electrons necessary to form a mini - halo @xcite .
lllrrcl cluster & z & kpc/@xmath25 & s(1.4 ) & log p(1.4 ) & lx[0.1 - 2.4kev ] & ref + & & & mjy & w / hz & @xmath68 erg / sec & + perseus & 0.0179 & 0.36 & 3020.0 & 24.33 & 7.88 & @xcite + ophiuchus & 0.028 & 0.55 & 106.4 & 23.27 & 6.11 & @xcite + a1835 & 0.2532 & 3.91 & 6.0 & 24.06 & 24.15 & @xcite + a2029 & 0.0765 & 1.43 & 18.8 & 23.42 & 9.24 & @xcite + a2142 & 0.0894 & 1.65 & 18.3 & 23.55 & 11.47 & @xcite + a2390 & 0.228 & 3.62 & 63.0 & 24.98 & 5.33 & @xcite + a2626 & 0.0604 & 1.15 & 29.0 & 23.39 & 1.18 & @xcite + mrc0116 + 111 & 0.131 & 2.31 & 140.0 & 24.79 & @xmath69 & @xcite + rbs 797 & 0.35 & 4.91 & @xmath69 & @xmath69 & 16.91 & @xcite + rxj1347.5@xmath691145 & 0.451 & 5.74 & 25.2 & 25.27 & 51.95 & @xcite + rxcj1504.1@xmath690248 & 0.2153 & 3.46 & 20.0 & 24.42 & 27.55 & @xcite + + + + cassano et al .
@xcite found that the synchrotron emissivity of mini - halos is about a factor of 50 higher than that of radio halos . in the framework of the particle reacceleration scenario
, they suggested that , an extra amount of relativistic electrons would be necessary to explain the higher radio emissivity of mini - halos .
these electrons could be provided by the central radio galaxy ( e.g. @xcite ) or be of secondary origin as previously suggested in the literature by some authors ( e.g. @xcite ) . in this latter case relativistic electrons in mini - halos
are expected to be continuously produced by the interaction of cosmic ray protons with the ambient thermal protons .
murgia et al .
@xcite modeled the radio brightness profile @xmath70 of radio halos and mini - halos , with an exponential of the form @xmath71 . the proposed method to derive the central radio brightness , the length - scale , and hence the radio emissivity of diffuse sources
is relatively independent of the sensitivity of the radio observation .
[ fig : comparison ] shows the best fit central brightness @xmath72 versus the length - scale @xmath73 for a representative set of radio halos and mini - halos .
these authors found that radio halos can have quite different length - scales but their emissivity is remarkably similar from one halo to another .
in contrast , mini - halos span a wide range of radio emissivity .
some of them , like perseus , rxj1347.5 - 1145 and a2390 are characterized by a radio emissivity which is more than two order of magnitude greater than that of radio halos .
this finding seems to suggest that part of the mini - halo radio emission is related to properties of the local intergalactic medium and part is correlated with the agn activity of the central brightest galaxy , as suggested by the faint correlation discussed by govoni et al .
@xcite , between the mini - halo and the cd radio power .
on the other hand there are also mini - halos like a2029 , ophiuchus , and a1835 that have a radio emissivity which is much more typical of halos in merging clusters rather than of the other mini - halos .
thanks to new radio observations and to the improvement of interpretative models , our understanding of magnetic fields associated with the intracluster medium in clusters of galaxies has advanced significantly in recent years and the presence of @xmath74-level magnetic fields in clusters of galaxies is now widely acknowledged .
most of what we know about intracluster magnetic fields derives from the study of cluster diffuse radio sources ( halos , relics and mini - halos ) and faraday rotation measures of polarized radio galaxies located inside or behind galaxy clusters .
one of the first comprehensive reviews on extragalactic magnetic field was presented by kronberg @xcite , followed by more recent works @xcite .
large - scale magnetic fields are studied through radio halos , which reveal magnetic fields spread over extremely large volumes .
sometimes , the intracluster magnetic fields can be ordered on scales of hundreds of kpc , as revealed in a2255 @xcite , and macs j0717.5 + 3745 @xcite , where a polarized signal has been detected .
the brightness fluctuations and polarization level of radio halos are strictly related to the intracluster magnetic field structure .
for example , lack of polarization and a smooth and regular surface brightness may indicate that the cluster magnetic field is ordered on small scales , while a disturbed radio morphology and presence of polarization could be related to a magnetic field ordered on large scales @xcite .
currently , only in the radio halos of a2255 and a665 , brightness fluctuations and polarization information have been used to put some constraints on the magnetic field power spectrum of a galaxy cluster @xcite .
the most promising technique to derive a detailed view of the intracluster magnetic fields on small spatial scales is the analysis of the faraday rotation of radio galaxies located inside and beyond clusters .
the faraday effect is the rotation suffered by linearly polarized radiation traveling through a magnetized plasma .
the intrinsic polarization angle @xmath75 is rotated by an amount : @xmath76 where @xmath77 is the radiation wavelength while @xmath78 is the so - called rotation measure .
the rotation measure is defined by : @xmath79 where @xmath80 is the thermal electron density in @xmath6 , @xmath81 is the magnetic field component along the line - of - sight in @xmath8 g , and @xmath82 is the path length through the plasma in kpc @xcite . if we perform multi - frequency observations of a radio source embedded in or behind a galaxy cluster , then by a linear fit of the @xmath83-law in eq.[lambda2 ] we can estimate the rm along that direction . by knowing the rm and given a model for the distribution of the thermal electrons ( e.g. from x - ray observations ) we can infer the intracluster magnetic fields
however , because of the random and turbulent nature of intracluster magnetic fields , inverting eq.[rm ] is not straightforward even in the cases of simple electron density distributions . a technique to analyze and interpret the rm data is the rm synthesis @xcite , which uses the rm transfer function to solve the n@xmath84 ambiguity related to the rm computation , and allows distinguishing the emission as a function of faraday depth .
detailed rm images of extended radio sources have been obtained ( e.g. @xcite ) .
the rm distributions seen across these sources present patchy structures whose statistics is , in first approximation , gaussian - like with dispersions up to several hundred ( and even thousand ) rad / m@xmath85 .
the observed rm fluctuations indicate that the intracluster magnetic field is not regularly ordered but turbulent on scales ranging from tens of kpc to @xmath9 100 pc .
the early interpretation of this finding was that the intracluster magnetic field is composed by uniform cells of size @xmath86 with random orientation in space . in this `` single - scale '' model ( e.g. @xcite )
, the rm distribution is expected to be a gaussian with zero mean and dispersion given by : @xmath87 thus , by measuring the dispersion of the rm we can estimate the cluster magnetic field provided that the cell size @xmath88 is known .
the question is then what exactly is @xmath88 and how can we find it .
this issue was investigated theoretically and in numerical tests by different authors @xcite , who determined that the correct value for @xmath86 is the magnetic field auto - correlation length , @xmath89 , which can be calculated if the power spectrum of the magnetic field fluctuation is known . hence , if we want to determine the strength of the magnetic field we ultimately need to determine the power spectrum of its fluctuations .
murgia et al .
@xcite and laing et al .
@xcite developed dedicated software tools to attempt to constraint the magnetic field power spectrum parameters .
these codes have been designed to produce , starting from 3d magnetic field models , synthetic polarization images that can be directly compared with the observations . to limit the degrees of freedom and the computational burden , in many of these simulations the magnetic field power spectrum
is assumed to be a simple power - law of the form @xmath90 .
however , in some cases high - order terms in curvature have been also explored @xcite .
another more general method , relying on a semi - analytical approach and bayesian inference has been also developed by enlin & vogt @xcite . the power spectrum shape is decomposed with multiple spectral basis components and it is free to assume , in principle , any form required by the data @xcite .
in all the mentioned modeling , it is assumed that the power spectrum normalization scales as a function of the gas density such that @xmath91 , where the index @xmath92 is in the range [email protected] @xcite .
indeed there are several indications that the magnetic field intensity declines with radius with a rough dependence on the thermal gas density .
this is predicted by both cosmological simulations e.g @xcite and by the comparison between thermal and radio brightness profiles @xcite in clusters hosting radio halos .
we note that in agreement with the absence of a preferred direction in most of the rm images , the results of the modeling are based on the assumption that the magnetic field is statistically isotropic and that the faraday rotation is occurring entirely in the intracluster medium with a negligible local rm enhancement occurring at the interface between the radio lobes and the surrounding medium ( but see @xcite for a contrary viewpoint ) .
however , recent rm images @xcite of four radio galaxies ( 0206 + 35 , 3c270 , 3c353 , m84 ) located in environments ranging from poor groups to rich clusters show clearly anisotropic `` banded '' patterns . in some cases , these rm patterns coexist with regions of isotropic random variations .
these rm bands may be interpreted as a consequence of interactions between the sources and their surroundings and are likely be caused by magnetized plasma which has been compressed by the expanding radio lobes .
the faraday rotation modeling described above has been applied to interpret the rm images of single extended radio galaxies located at the center of galaxy clusters .
some of them present strong rm asymmetries between the two lobes .
this is the laing - garrington effect : the farthest lobe suffers a higher faraday rotation due to the largest amount of intervening material @xcite . indeed , at least for these sources , we know that the observed rm must originate mostly in the foreground magnetized intracluster medium .
radio sources at the center of dense cool - core clusters , like hydra - a , probe magnetic fields as high as 10@xmath6930@xmath93 @xcite .
the rm analysis of radio sources in the less dense environments of galaxy groups , such us 3c31 or 3c449 @xcite , indicates lower central magnetic field strengths in the range 3@xmath6910 @xmath74 .
the intracluster magnetic field auto - correlation length is generally found to be of the order of a few kpc . determining the detailed shape of the magnetic field power spectrum it is not trivial because of the degeneracy existing between its slope and the outer scale of the magnetic field fluctuations , @xmath94 @xcite .
in fact , the same rm images can be explained either by a flat power spectrum with a large @xmath94 or by a steeper power spectrum with a lower @xmath94 .
given the limited spatial dynamic range of the current rm images , it s still not easy to break the degeneracy . according to @xcite ,
the slope magnetic field power spectrum is close to the kolmogorov index in hydra - a while other authors report a flatter slope @xcite . to get information on the radial trend of the cluster magnetic field , one has to take into account that in the previously described modeling , another important degeneracy exists between the central magnetic field strength @xmath95 and the index @xmath92 , which describes the scaling of @xmath96 with the gas density .
the same rm image can be explained either by a higher @xmath95 with a rapid radial decrease ( high @xmath92 ) or by a lower @xmath95 with a flatter radial profile .
the degeneracy can be solved if rm images of different radio sources at different impact parameters from the cluster center are available . in a2382 the rm of two tailed sources
have been investigated @xcite . in this cluster ,
the power spectrum of the rm fluctuation is compatible with the kolmogorov index .
the observed @xmath97 decreases in the cluster outskirts and can be explained if @xmath98 @xmath74 and @xmath99 .
but it is also marginally compatible with a flat magnetic field profile with @xmath100 @xmath74 or with a higher central magnetic field strength of @xmath101 @xmath74 which decreases with radius following the thermal gas density , @xmath102 . in a few other abell clusters ,
the rm of radio sources at different impact parameters have been obtained , in a119 @xcite , the rm analysis indicates a central magnetic field strength of @xmath103 @xmath74 which decreases with radius roughly following the gas density .
the case of a2255 is particularly interesting since this cluster hosts a mpc - scale radio halo . by analyzing the rm of three different radio sources and the radio halo properties , a central magnetic field strength of about 2 @xmath74 for @xmath99
has ben obtained @xcite .
these results would imply that the magnetic field strength in clusters with or without a diffuse halo is indeed very similar .
the rm of radio sources in the coma cluster , are probably the best studied so far @xcite .
bonafede et al .
@xcite , by comparing the rm observed in seven radio sources ( see fig .
[ fig : coma ] ) with simulations , found a cluster magnetic field in the range @xmath95=3.9 @xmath8 g ; @xmath104 and @xmath95=5.4 @xmath8 g ; @xmath105 .
these values correspond to models where the magnetic field energy density scales as the gas energy density , or the magnetic field is frozen into the gas , respectively .
this is expected since the energy in the magnetic component of the intracluster medium is a tiny fraction of the thermal energy .
it is important to note that the average magnetic field intensity over a volume of @xmath31 mpc@xmath2 is @xmath32 @xmath8 g .
indeed , the model derived from rm analysis gives an average estimate that is compatible with the minimum energy condition in the radio halo @xcite .
a direct comparison with the magnetic field estimate derived from the inverse compton emission is more difficult since the hard - x detection is debated .
the field strength derived from rm analysis gives a magnetic field estimate that is consistent with the present lower limits obtained from hard x - ray observations @xcite while it is still a bit higher than the estimate obtained on the basis of the non - thermal hard x - ray detection from rxte @xcite and bepposax @xcite .
in addition to detailed rm studies focused on single clusters , interesting analysis have been carried out also by obtaining rm of radio galaxies through a sample of clusters .
clarke et al . @xcite analyzed the average rm values as a function of source impact parameter for a sample of 16 abell clusters .
they found a clear broadening of the rm distribution toward small projected distances from the cluster center clearly indicating , as suggested also by other statistical investigations ( e.g. @xcite ) , that most of the rm contribution comes from the intracluster medium and proving that magnetic fields are present in all galaxy clusters .
a similar correlation has been recently found by govoni et al .
they analyzed a sample of 12 galaxy clusters in which they included only those clusters hosting radio sources for which high - quality , high - resolution , extended rm images are available .
although the statistic is still limited , they found that the rm of radio galaxies in clusters containing a radio halo seem to show a similar behavior to that of clusters without diffuse synchrotron emission . increasing attention
is given in the literature to the connection between the magnetic field strength and the cluster temperature .
these studies are based on cosmological simulations e.g. @xcite or plasma physical considerations @xcite . the comparison between clusters with high and low temperatures @xcite does not yields significant differences in the rm data , indicating that a possible connection between the magnetic field strength and the gas temperature , if present , is very weak .
very recently bonafede et al .
@xcite selected a sample of 39 massive galaxy clusters from the highest x - ray flux galaxy cluster sample @xcite , and used nvss data @xcite to analyze the trend of the fractional polarization of radio sources lying at different projected distances from the cluster center .
they detected a clear trend of the fractional polarization , being smaller for sources close to the cluster center and increasing with increasing distance form the cluster central regions .
the higher rm suffered by sources in the inner regions , where both magnetic field strength and gas density are higher , causes in fact a higher beam depolarization .
sources at larger radii , instead , are subject to less depolarization , since their emission is affected by lower rm .
such a trend can be reproduced by a magnetic field model with central value of few @xmath8 g .
in addition there seems no differences in the depolarization trend observed in cluster with and without radio halo , indicating that magnetic fields in galaxy clusters are then likely to share the same properties regardless of the presence of radio emission from the intracluster medium .
cosmological theories and simulations predict that galaxy clusters are connected by intergalactic filaments along which they accrete mass .
shocks from infall into and along the filaments are expected to accelerate particles .
these accelerated particles can emit synchrotron radiation if cosmic magnetic fields are present .
attempts to detect diffuse radio emission beyond clusters , i.e. in very rarefied regions of the intergalactic space , have shown recent promise in imaging diffuse synchrotron radiation of very low level .
bagchi et al.@xcite suggested the presence of radio emission from the filament surrounding the cluster zwcl2341.1 + 0000 .
this finding has been recently confirmed with deeper observations by @xcite ( fig .
[ fig : filament ] ) , thus proving that intergalactic filaments can be magnetized .
another filamentary structure extending over 3.3 mpc is centered on the cd galaxy of the cluster a3444 @xcite .
the large size of this radio feature , as well as its morphology , rule out a direct connection with the central galaxy and invite comparison with the filamentary structures found in zwcl2341.1 + 0000 ( see also @xcite ) .
the presence of large - scale radio emission in regions in between galaxy clusters , ( e.g. 0917 + 75 @xcite ) , and giant ring - like features ( e.g. a3376 @xcite ) seems to be related to the large - scale structure formation process , rather than to the cluster .
we note also that diffuse bridges of radio emission connecting the halo to the relic have been observed in a few clusters , including coma @xcite and a2255 @xcite . in the coma cluster kronberg
@xcite , confirmed by brown & rudnick @xcite , detected a radio cloud surrounding both the halo and the relic source .
the radio cloud size corresponds to about 4 mpc if it is associated with the coma cluster , and is the most extensive complex of synchrotron emitting plasma yet seen in cosmic large - scale structure . in a2255 , in addition to the radio bridge , low brightness emission located far away from the cluster center has been detected @xcite .
all the mentioned large - scale diffuse radio emissions might trace the process of a large - scale structure formation , where cosmic shocks that originated in complex merger events are able to amplify magnetic fields and accelerate synchrotron electrons along cluster filaments . in many cases
the physical nature of these structures remains , however , ambiguous ; candidates include cosmological filaments , unevolved protoclusters in the process of formation , and or large - scale debris near rich , merging clusters . in any case
, all these data support the existence of an intergalactic magnetic field more widespread and somewhat lower than that in the intracluster medium within clusters .
this field may represent the seed field for galaxies and clusters , and may play an important role in the formation of large - scale structure .
these are important results since they allow expanding our knowledge of magnetic fields from clusters to filaments , and are a fundamental step to understanding the origin and properties of cosmological scale magnetic fields .
we are moving from magnetic field structures up to 1 mpc scale to structures 3 - 5 times more extended .
attempts to probe magnetic field strength on larger local - universe scales have also been done via faraday rotation studies @xcite .
the measurement of extremely weak magnetic fields in the voids and large scale structure is a challenging task and up to now only upper bounds have been derived using various techniques .
the tightest upper bounds come from the search for the faraday rotation of polarization of the radio emission from distant quasars ( e.g. @xcite ) , and from the effect of magnetic fields on the anisotropy of cosmic microwave background radiation ( e.g. @xcite ) .
a recent approach , using the cascade emission from blazars has been found to be promising ( e.g. @xcite ) . the multi - tev gamma - ray flux from distant blazars
is strongly attenuated by pair production of the infrared / optical extragalactic background light , initiating electromagnetic cascades in the intergalactic space , whose charged component is deflected by the extragalactic magnetic fields .
observable effects of these cascades include the delayed `` echoes '' of multi - tev gamma ray flares or gamma - ray bursts and the appearance of extended emission around point - like gamma - ray sources . from fermi observations ,
lower bounds are at levels of 10@xmath106 to 10@xmath107 g , depending on the assumptions .
dolag et al .
@xcite derive that these magnetic fields are likely to fill at least 60% of space , thus imposing constraints on the magnetic field origin : either the magnetic fields are produced by a volume filling process ( e.g. primordially ) or very efficient transport mechanisms must be present to distribute the locally generated magnetic fields into the space .
the connection between extended diffuse radio sources and cluster evolution , in particular the link between merger processes and halos / relics on one hand , and the link between mini - halos and relaxed clusters on the other hand , is well established as reported in this review .
major cluster mergers can supply energy to the radio emitting particles , as well as amplify magnetic fields @xcite .
it has been shown that the magnetic field strength can grow during the process of cluster formation , from initial fields of the order of 10@xmath108 g at z = 15 to final fields of the order of @xmath8 g at z = 0 .
this magnetic field amplification is spread over the entire history of cluster formation , and related to the various merger processes , therefore it is not expected to directly affect the formation of diffuse radio sources . since magnetic fields
have been found to be ubiquitous in galaxy clusters ( see sect . [ sec:74 ] ) , the crucial ingredient for the existence of diffuse synchrotron radio sources is the presence of relativistic particles .
radio halos are most difficult to explain , because of their very large size .
radiating electrons can not travel such large distances within their lifetime , because of strong radiative losses by synchrotron and inverse compton emission .
two main classes of models have been suggested for the origin of relativistic electrons present in the cluster volume and responsible for the diffuse radio emission : the _ primary electron _ model which predict that relativistic particles are continuously accelerated in the cluster volume , and the _ hadronic ( secondary electrons ) _ model , in which relativistic electrons are produced through the cluster volume by pp collisions .
we introduce here these models and summarize their expectations , to draw a picture of the formation scenario of diffuse radio sources . for a detailed analysis of the theoretical studies on this topic we refer to recent papers ( e.g. @xcite and references therein ) .
primary relativistic electrons are present in the cluster volume , because they were injected by agn activity ( quasars , radio galaxies , etc . ) , or by star formation in normal galaxies ( supernovae , galactic winds , etc . ) during the cluster dynamical history ( see @xcite for a review ) .
this population of electrons suffers strong radiation losses mainly because of synchrotron and inverse compton emission , thus reacceleration is needed to maintain their energy to the level necessary to produce the observed synchrotron radio emission in relatively weak magnetic fields ( e.g. @xcite ) .
the possibility that electrons are accelerated from the sub - relativistic electrons has also been considered in the literature ( e.g. @xcite and references therein ) . for the above reasons , primary electron models
can also be referred to as reacceleration models .
there are two main ways that entail the transfer of energy from the cluster icm to the radiating particles : these are cluster turbulence and cluster shocks .
simulations show that during cluster mergers , turbulence is generated throughout the cluster over @xmath3 mpc scales .
energy can thus be transferred from the icm into the non - thermal component through resonant or non - resonant interaction of electrons with mhd turbulence ( see e.g. @xcite ) .
the emerging scenario is that turbulence reacceleration is likely the major mechanism responsible for the supply of energy to the electrons radiating in radio halos .
turbulent acceleration is like a second - order fermi process , i.e. related with a random processes , and consequently not quite efficient .
the time during which the process is effective is relatively short ( a few 10@xmath32 years ) , so that the radio emission is expected to correlate with ongoing or most recent merger events ( e.g. @xcite ) .
these models predict that reaccelerated electrons will have a maximum energy ( @xmath109 10@xmath110 ) which produces a high frequency cut - off in the resulting synchrotron spectrum @xcite .
thus a high frequency steepening of the integrated spectrum is expected , as well as a radial steepening and/or a complex spatial distribution of the spectral index between two frequencies , the latter due to different reacceleration processes in different cluster regions . the evidence of the presence of faint diffuse magnetic fields in all galaxy clusters ( see sect . [ sec:74 ] ) , the behaviour of radio spectra ( see sect . [ sec:41 ] ) , the existence of ultra - steep radio halos ( sect .
[ sec:413 ] ) , the correlation between radio spectra and temperature ( sect .
[ sec:412 ] ) , the association between radio halos and cluster mergers ( sect .
[ sec:432 ] ) , and the fact that halos are not common in galaxy clusters @xcite , are supportive of primary electron models for radio halos . the radio power - x - ray luminosity correlation ( sect .
[ sec:43 ] ) is easily reproduced by reacceleration models ( see e.g. @xcite ) . in mini - halos ,
the mhd turbulence associated with the cool - core region has been also suggested to be responsible for reacceleration of radiating particles @xcite .
this is consistent with the spectral properties and the observed correlation between the mini - halo radio power and the cooling rate power @xcite .
another possibility , recently suggested by mazzotta & giacintucci @xcite and zuhone et al .
@xcite , is that turbulence in cool - core clusters could arise from the sloshing motions of the cluster core gas .
this hypothesis is supported by the detection of spiral - shaped cold - fronts in the x - ray emission of cool - core clusters , suggested to be the signature of gas sloshing ( see e.g. @xcite ) .
shock acceleration is a first - order fermi process of great importance in radio astronomy , as it is recognized as the mechanism responsible for particle acceleration in supernova remnants .
the acceleration occurs diffusively , in that particles scatter back and forth across the shock , gaining at each crossing and recrossing an amount of energy proportional to the energy itself .
the acceleration efficiency is mostly determined by the shock mach number .
current observations globally favour the scenario that cluster shocks are strictly linked to relics @xcite ) .
the production of outgoing shock waves at the cluster periphery with mach number @xmath3 2 - 3 is indeed observed in numerical simulations of cluster merger events and more generally in the large - scale structure formation @xcite .
accelerated electrons have been suggested either to be thermal icm electrons @xcite , or relativistic electrons released by a former active radio galaxy @xcite .
because of the short electron radiative lifetimes , radio emission is produced close to the location of the shock waves .
these models also predict that the magnetic field within the relic is aligned with the shock front , and that the radio spectrum is flatter at the shock edge , where the radio brightness is expected to decline sharply .
these expectations are consistent with the classic elongated structure of relics , almost perpendicular to the merger axis , and their polarization properties .
vazza et al .
@xcite show that the detection frequency of radio relics and the fact that they are also detected in cluster peripheries is consistent with their link to shocks .
the detection of shocks in the cluster outskirts is difficult with current instruments , because of the very low gas density and thus x - ray brightness in these regions ( for a recent review see @xcite ) . in the coma cluster , feretti & neumann @xcite , analyzing newton - xmm data ,
did not detect the temperature jump implied by the presence of a shock at the relic location , although the brightness trend may be consistent with the shock .
solovyeva et al .
@xcite presented evidence of a shock at the location of the relics in a548b . a weak shock has been detected by macario et al .
@xcite in a754 .
recently , shocks have been detected with suzaku in a3376 @xcite and a3667 @xcite .
it can not be excluded that shock acceleration may also be efficient in some particular regions of a halo ( e.g. @xcite ) , however , the radio emission of halos can be very extended up to large scales , thus it is hardly associated with localized shocks .
moreover , some clusters exhibit a spatial correlation between the radio halo emission and the hot gas regions @xcite , however , this is not a general feature , and in some cases the hottest gas regions do not exhibit radio emission . in the hadronic model ,
secondary electrons are injected as secondary particles by inelastic nuclear collisions between the relativistic protons and the nuclei of the thermal ambient intracluster medium ( e.g. @xcite ) .
the protons diffuse on large scale because their energy losses are negligible .
they can continuously produce _ in situ _ electrons , distributed through the cluster volume .
the possibility that the high - energy electrons , responsible for the synchrotron emission , arise from the decay of secondary products of the neutralino annihilation in the dark matter halos of galaxy clusters has also been suggested @xcite .
secondary electron models have been proposed for the emission of radio halos ( see e.g. @xcite ) and of mini - halos @xcite .
they can not work for relics , since peripheral cluster regions do not host a sufficiently dense thermal proton population required as targets for the efficient production of secondary electrons .
secondary electron models can reproduce the basic properties of the radio halos provided that the strength of the magnetic field averaged over the emitting volume is larger than a few @xmath8 g . in this case , they predict synchrotron power - law spectra , which are independent of cluster location , i.e. do not show any features and/or radial steepening , and the spectral index values are flatter than @xmath28 @xmath3 1.5 ( e.g. @xcite ) .
the profiles of the radio emission should be steeper than those of the x - ray gas ( e.g. @xcite ) . since the radio emitting electrons originate from protons accumulated during the cluster formation history , no correlation to recent mergers
is expected , but halos should be present in virtually all clusters .
moreover , emission of gamma - rays and of neutrinos is predicted .
recently , a universal linear correlation between radio and x - ray surface brightness in halos and mini - halos was found @xcite , assuming secondary electron models and a relatively strong magnetized igm ( @xmath19 3 @xmath8 g ) .
this model reproduces the observed radio spectral steepening , but requires a very high central magnetic field ( @xmath19 10 @xmath8 g ) .
the intracluster medium is a complex mixture of thermal and non - thermal protons and electrons .
it has been suggested in that halos could arise from both primary and secondary electrons .
relativistic protons are an important component of relativistic icm particles , and they are an important origin of relativistic electrons .
mhd turbulence can reaccelerate both primary and secondary particles , producing radio emission in agreement with observational results ( see e.g. @xcite ) .
the relative contribution of these two components is constrained by the observational properties .
constraints on the energy budget of the various components have been derived by brunetti & blasi @xcite and more recently by brunetti & lazarian @xcite .
these authors discuss also that secondary electrons could give rise to mpc scale halos characterized by very low surface brightness and radio power , not yet observed because of observational limits of present radio telescopes , but which should be detected by the incoming new radio telescopes .
the collisions producing secondary electrons also lead to gamma - ray emission through the production and subsequent decays of neutral pions .
the implications of gamma - ray detections from galaxy clusters are largely discussed in literature ( see e.g. blasi et al . @xcite and references therein ) . however ,
diffuse gamma - ray emission has not yet been detected from galaxy clusters @xcite .
these non detections constrain the possible energy density in cosmic - ray protons and , therefore , the density of relativistic electrons that may be produced from these .
the fermi - lat gamma - ray space telescope greatly improved the sensitivity of observations at mev / gev energies and reported gamma - ray upper limits for a large sample of clusters @xcite .
jeltema & profumo @xcite examined the secondary electron models for the origin of halos in the light of these new fermi results .
the non detection of gamma - ray emission from clusters puts constraints to the possible density of secondary relativistic electrons and imply a minimum magnetic field strength to reproduce the observed emission from radio halos . for clusters such as a1914 and a2256 , average magnetic fields at least of the order of several @xmath8gauss
are needed to avoid an overproduction of gamma - rays , not seen by fermi @xcite .
the implied magnetic fields in hadronic models are in excess of , or close to , the largest cluster magnetic field values obtained with faraday rotation measures ( see sect .
[ sec:7 ] ) .
this seems problematic for the hadronic origin of radio halos .
however , at present , due to the uncertainties , secondary electron models are not ruled out by fermi gamma - ray data even if required magnetic fields are in some clusters higher than values obtained with other observations .
this result show the importance of high energy data in understanding non - thermal emission in galaxy clusters .
our knowledge of non - thermal phenomena in galaxy clusters has significantly improved in the past years , owing to good and deep observations and advanced modeling .
nevertheless , new observations will be necessary to solve the outstanding issues faced in determining what is the dominating mechanism in the formation and maintenance of the diffuse radio emission , in particular : 1 ) search for new sources of the different classes , in particular at low powers and at large redshifts , to improve the information on the statistical properties of cluster diffuse sources ; 2 ) obtain polarization information on halos , and perform rotation measure studies of embedded and background radio sources , to get a detailed knowledge of the magnetic field strength , structure , and degree of ordering ; 3 ) derive accurate integrated spectra on a large frequency range and detailed spectral index distributions at high resolution . owing to the new capabilities of the evla , in particular sensitivity , frequency agility , and spectral capability , it will be possible to detect faint radio emission and trace the magnetic fields in galaxy clusters .
the new radio telescope lofar will explore the universe at low frequencies , as shown e.g. by the first results on the cluster a2256 @xcite .
as diffuse cluster radio sources are bright at low frequency , lofar will be able to detect thousands of such sources up to the epoch of cluster formation , in particular halos and relics with ultra - steep spectra , which are missed at cm frequencies
. this will allow the investigation of the occurrence of diffuse cluster radio sources .
their properties will be derived as a function of cluster properties up to the epoch at which the first massive clusters assembled ( z @xmath3 1 ) , testing predictions that cluster merging is rampant at high redshift .
the detection of synchrotron radiation at the lowest possible levels will allow the measurement of magnetic fields in clusters and in even more rarefied regions of the intergalactic space , and the investigation of the relation between the formation of magnetic fields and the formation of the large - scale structure in the universe .
with lofar , owing to the long wavelengths involved , it will also be possible to obtain faraday rotation down to 0.1 - 1 rad m@xmath111 , which are very difficult to measure with the current instruments at cm wavelengths .
the detection of very low rotation measures implies in turn the detection of very faint magnetic fields therefore lofar will become the radio telescope to measure the weakest cosmic magnetic fields so far .
two projects planned with the australian square kilometre array pathfinder ( askap ) , i.e. the deep radio sky survey project emu ( evolutionary map of the universe ) and the polarization survey possum ( polarisation sky survey of the universe s magnetism ) will allow the detection of diffuse radio sources and the improvement of faraday rotation information in clusters and in the overall intergalactic medium . the first science with the meerkat array in south africa region , in particular the mightee ( meerkat international gigahertz tiered extragalactic exploration survey ) project will cover a wide range of astrophysics , including non - thermal phenomena in clusters . apertif phased arrays that are being installed in wsrt will dramatically increase the survey speed for wsrt .
this will enable the wodan ( westerbork observations of the deep apertif northern - sky ) project , a survey complementary to emu .
the most innovative studies will be performed with ska : they are summarized in science with the square kilometre array
@xcite ( for cluster and intergalactic medium studies see e.g. @xcite ) . recently ,
krause et al .
@xcite investigate the potential of the square kilometer array ( ska ) for measuring the magnetic fields in clusters of galaxies via faraday rotation of background polarized sources .
they derive the distribution of rm due to clusters , by using an average gas density distribution and adopting models for the population of cluster radio sources , for the cluster mass distribution , and for the source polarization .
they calculate the statistics for background rm measurements in clusters at different redshift and different ska exposures , arrays and completion levels in the mid frequency range .
they conclude that the full ska will be able to make very detailed magnetic field structure measurements of clusters with more than 100 background sources in clusters up to z = 0.5 , and more than 1000 background radio sources in nearby clusters .
bogdanovic et al .
@xcite analyze different physical mechanisms operating in the icm , i.e. conduction - driven mhd instabilities and turbulence - dominated instabilities , and derive that these deeply affect the structure of the cluster magnetic field producing different field topologies .
they evaluate the effects of the two mechanisms on the faraday rm of background radio sources and conclude that future spectropolarimetrimetric measurements with evla and ska will have sufficient sensitivity to discriminate between them .
observations in the radio domain will need to be complemented by data at all other wavelenghts , in particular optical , x - ray and gamma - ray , to establish the cluster conditions , the merger evolutionary stage , the presence and properties of shocks , the signatures of cluster turbulence , and to derive firm correlations on large samples and over a large range of parameters between radio properties and cluster parameters .
finally , the study of the non - thermal inverse compton emission in the hard x - ray and extreme ultraviolet domains needs to be pursued on several clusters , with and without diffuse emission , to get independent information on the existence of relativistic particles .
beck , r. , proc . of magnetic fields in the universe : from laboratory and stars to primordial structures , eds .
m. soida , k. otmianowska - mazur , e.m
. de gouveia , a. lazarian , in press ( arxiv:1111.5802 ) ( 2011 ) kempner , j.c . ,
blanton , e.l . ,
clarke , t.e . ,
enlin , t. a. et al . , proceedings of the riddle of cooling flows in galaxies and clusters of galaxies , eds .
t. reiprich , j. kempner , and n. soker http://www.astro.virginia.edu/coolflow ( 2004 ) | clusters of galaxies , as the largest virialized systems in the universe , are ideal laboratories to study the formation and evolution of cosmic structures .
the luminous matter of clusters consists of galaxies and of an embedding intracluster medium ( icm ) , which has been heated to temperatures of tens of millions degrees , and thus is detected through its thermal emission in the soft x - ray regime .
most of the detailed knowledge of galaxy clusters has been obtained in recent years from the study of icm through x - ray astronomy . at the same time
, radio observations have proved that the icm is mixed with non - thermal components , i.e. highly relativistic particles and large - scale magnetic fields , detected through their synchrotron emission .
the knowledge of the properties of these non - thermal icm components has increased significantly , owing to sensitive radio images and to the development of theoretical models .
diffuse synchrotron radio emission in the central and peripheral cluster regions has been found in many clusters .
moreover large - scale magnetic fields appear to be present in all galaxy clusters , as derived from rotation measure ( rm ) studies .
non - thermal components are linked to the cluster x - ray properties , and to the cluster evolutionary stage , and are cucial for a comprehensive physical description of the intracluster medium .
they play an important role in the cluster formation and evolution .
we review here the observational properties of diffuse non - thermal sources detected in galaxy clusters : halos , relics and mini - halos .
we discuss their classification and properties .
we report published results up to date and obtain and discuss statistical properties .
we present the properties of large - scale magnetic fields in clusters and in even larger structures : filaments connecting galaxy clusters .
we summarize the current models of the origin of these cluster components , and outline the improvements that are expected in this area from future developments thanks to the new generation of radio telescopes .
example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore |
alternative formulations for the study of electroweak symmetry breaking that have the property of canceling quadratic divergences are the so called little higgs models ( lhm ) @xcite .
these models are based on dimensional deconstruction @xcite , where the quadratic divergence induced at the one - loop level by the standard model gauge bosons are canceled via the quadratic divergence introduced by heavy gauge bosons at the same perturbative level .
also , it is proposed the existence of heavy - mass fermions interacting with the higgs field in such a way that the one - loop quadratic divergence induced in the yukawa sector of the standard model ( sm ) due to top quark coupling with the higgs boson is canceled @xcite .
furthermore , the higgs fields acquire mass becoming pseudo - goldstone bosons via an approximate global symmetry breaking , where a massless higgs appears .
quadratically divergent corrections to the higgs mass arise at loop level , therefore , this naturally ensure a light higgs . as far as the littlest higgs ( lthm ) is concerned
, a remarkable feature is that there is no new degrees of freedom beyond the sm below tev scale .
moreover , above few tev s the lthm needs a very small new degrees of freedom to stabilize the higgs boson mass . at the tev energy scale , the arising new particles are a set of four gauge bosons with the same quantum numbers as the electroweak sm gauge bosons , namely , @xmath7 , @xmath4 , and @xmath8 , an exotic quark with the same charge as the top quark , and a scalar triplet @xcite .
the construction details of the model can be found in refs .
@xcite . in general
, these extensions of the sm predict new particles emerging at the tev scale and the new physics that could appear at these energies that soon will be tested at the large hadron collider ( lhc ) @xcite .
in particular , little higgs models predict the existence of a new neutral massive gauge boson , known as @xmath4 , which could offer another theoretical framework to justify the experimental scrutiny about the possible existence of heavy - mass ( at the tev scale ) particles like the @xmath9 gauge boson of the sm .
on the other hand , there are several models that predict the existence of a neutral massive gauge boson , identified as @xmath10 gauge boson , such as the 331 model @xcite or grand unified models @xcite .
these type of particles are under exhaustive search at the lhc @xcite , where the atlas and cms collaborations have imposed experimental bounds over the mass of a new particle related to @xmath10 gauge boson , their results indicate that the mass of the @xmath10 gauge boson must be greater than @xmath11 tev and @xmath12 tev , respectively .
in this work we are interested in the physics of the @xmath4 gauge boson , specifically , the main concern of this paper is to study the @xmath13 decay in the context of the linearized theory of the littlest higgs model @xcite .
the relevance of this process brings the possibility of testing the lthm , since the parameters space has been severely constrained by the higgs discovery channels and electroweak precision observables @xcite .
another remarkable feature of the @xmath13 decay consists in the fact the sm background is naturally suppressed , namely , the @xmath14 reaction , since it is highly suppressed for its electroweak origin @xcite .
thus , the higgs - photon associated production opens a new window to test the gauge sector of the sm and higgs physics @xcite . to support our analysis ,
it is calculated the higgs - photon associated production at lhc coming from sm background , by using current kinematical cuts employed by atlas collaboration @xcite .
previous studies on the @xmath15 decay have been performed in the context of left - right symmetric models @xcite , where the associated branching ratio is estimated , however , the used parameters such as the @xmath16 mass are below the present bounds established by the experimental measurements @xcite .
the paper is organized as follows . in section [ secmf ] , we briefly describe the theoretical framework of the lthm . in sec .
[ secd ] , we outline the analytical results for the @xmath13 decay in the lthm . in sec . [ nr ] , it is presented the numerical analysis .
finally , the conclusions appear in sec .
the littlest higgs model is based on a nonlinear sigma model with @xmath17 global symmetry and the gauged subgroup @xmath18\otimes [ su(2)_2\otimes u(1)_2]$ ] @xcite . the global symmetry of the @xmath17 group
is spontaneously broken down @xmath19 at the energy scale @xmath20 , where @xmath20 is constrained to be of the order of @xmath21 tev @xcite .
simultaneously , the @xmath18\otimes [ su(2)_2\otimes u(1)_2]$ ] group is also broken to its subgroup @xmath22 , which results to be the sm electroweak gauge group .
the global symmetry breaking pattern leaves 14 goldstone bosons which transform under the @xmath22 group as a real singlet @xmath23 , a real triplet @xmath24 , a complex doublet @xmath25 , and a complex triplet @xmath26 @xcite .
the spontaneous global symmetry breaking of the @xmath17 group is generated by the vacuum expectation value ( vev ) of the @xmath27 field , denoted as @xmath28 @xcite , at the scale @xmath20 , which is parametrized by @xmath29 with @xmath30 and @xmath31 being the goldstone boson matrix given by @xmath32 here , @xmath33 is a doublet and @xmath34 is a triplet under the @xmath22 sm gauge group @xcite . by the spontaneous symmetry breaking ( ssb ) ,
both the real singlet and the real triplet are absorbed by the longitudinal components of the gauge bosons at the energy scale @xmath20 . at this scale ,
the complex doublet and the complex triplet remain massless .
the complex triplet acquires a mass of the order of @xmath20 by means of the coleman - weinberg type potential when the global symmetry of the group @xmath35 breaks down .
the complex doublet is identified as the sm higgs field .
the effective lagrangian invariant under the @xmath18\otimes [ su(2)_2\otimes u(1)_2]$ ] group is @xcite @xmath36 where @xmath37 represents the gauge bosons kinetic contributions , @xmath38 the fermion kinetic contributions , @xmath39 the non - linear sigma model contributions of the lthm , @xmath40 the yukawa couplings of fermions and pseudo - goldstone bosons , and the last term symbolizes the coleman - weinberg potential .
the standard form of the lagrangian of the non - linear sigma model is @xmath41 where the covariant derivative is written as @xmath42.\ ] ] here , @xmath43 are the @xmath44 gauge fields , @xmath45 are the @xmath46 gauge fields , @xmath47 are the @xmath44 gauge group generators , @xmath48 are the @xmath46 gauge group generators , @xmath49 are the coupling constants of the @xmath44 group , and @xmath50 are the coupling constants of the @xmath46 group @xcite .
after ssb around @xmath28 , it is generated the mass eigenstates of order @xmath20 for the gauge bosons @xcite @xmath51 where @xmath52 and @xmath53 for @xmath54 ; @xmath55 , @xmath56 , @xmath57 , and @xmath58 .
notice that @xmath27 field has been expanded around @xmath28 holding dominant terms in @xmath39 @xcite . at this stage of ssb the @xmath59 and @xmath60 fields remain massless .
the ssb at the fermi scale provides mass to the sm gauge bosons ( @xmath61 and @xmath62 ) and induces mixing between heavy and light gauge bosons .
the arising masses at the leading order ( neglecting terms of order @xmath63 , with @xmath64 being the vacuum expectation value at the fermi scale ) are @xcite @xmath65 as it is known @xmath66 and takes the value equals to one at the leading order , we may assume that the @xmath67 parameter ranges from @xmath68 to @xmath69 @xcite , in order to have values for the masses of the weak gauge bosons not very different , as it occurs in the electroweak sector of the sm .
the lthm incorporate new heavy fermions which couple to higgs field in a such way that the quadratic divergence of the top quark is canceled @xcite .
in particular , this model introduces a new set of heavy fermions arranged as a vector - like pair ( @xmath70 ) with quantum numbers @xmath71 and @xmath72 , respectively .
the new yukawa interactions are proposed to be @xmath73 where @xmath74 ; @xmath75 and
@xmath76 are antisymmetric tensors for @xmath77 and @xmath78 @xcite . here , @xmath79 and @xmath80 are free parameters , where the @xmath80 parameter can be fixed such that , for given @xmath81 , the top quark mass adjust to its experimental value @xcite . expanding the @xmath27 field and retaining terms up to @xmath82 after diagonalizing the mass matrix
, it can be obtained the mass states @xmath83 , @xmath84 , @xmath85 , and @xmath86 , which correspond to sm top quark and the heavy top quark , respectively @xcite .
the explicit remaining terms of the lagrangian @xmath87 as well as the complete set of new feynman rules can be found in ref .
we now turn our attention to obtain the analytical expression for the amplitude and decay width of the @xmath88 process .
the amplitude was calculated in the unitary gauge . in fig .
[ feyn - diagrams - contrib ] we show the contributions at one - loop level to the @xmath88 coming from fermions , gauge boson and scalars . in the fermion loops we include sm and lthm fermion contributions .
the analysis is based on three sets of feynman diagrams : the set ( a ) contains the triangle loop contributions mediated by fermions ; the set ( b ) includes triangle and bubble loop contributions mediated by sm charged gauge bosons and new heavy charged gauge bosons , where also it is include the mixing effects between these two types of charged gauge bosons ; for the set ( c ) , we take into account the bubble loop contributions induced by scalars and scalars plus gauge bosons , together with triangle loop contributions mediated by gauge bosons and scalars .
decay at one - loop level . here ,
@xmath89 , @xmath90 and @xmath91.,title="fig : " ] + decay at one - loop level .
here , @xmath89 , @xmath90 and @xmath91.,title="fig : " ] the respective decay amplitude is given by @xmath92 where @xmath93 .
here , @xmath94 represents the contribution of the set ( a ) , @xmath95 contains the contribution of the set ( b ) , and @xmath96 includes the contribution of the set ( c ) .
moreover , @xmath97 and @xmath98 are the polarization vectors associated to @xmath4 boson and photon , respectively . after tedious algebraic manipulations
we can write down a generic expression for the total decay amplitude as follows @xmath99 where @xmath100 and @xmath101 .
the @xmath102 , @xmath103 and @xmath104 coefficients are in terms of passarino - veltman scalar functions . in specific , @xmath105 , @xmath106 and @xmath107 . here
, @xmath20 runs over all charged fermions , @xmath108 represents charged gauge bosons ( @xmath62 , @xmath109 ) , and @xmath110 symbolizes charged scalars ( @xmath111 , @xmath112 , @xmath113 , and @xmath114 ) .
we found that the total contribution arising from tadpole and self - energies diagrams vanishes .
the explicit form for the @xmath115 , @xmath116 , @xmath117 , @xmath118 , @xmath119 , @xmath120 and @xmath121 coefficients are presented below @xmath122 where @xmath123 , @xmath124 and @xmath125 are the known passarino - veltman scalar functions @xcite .
also , we used @xmath126 and @xmath127 . notice that for this particular process @xmath128 and @xmath129 , where , as usual @xmath130 represents the third component of isospin being @xmath131 for fermions up ( down ) type @xcite .
the @xmath116 and @xmath119 coefficients contain the contributions of @xmath62 and @xmath109 bosons as follows @xmath132 where @xmath133 , @xmath134 , and @xmath135 .
moreover , @xmath136 and @xmath137 .
the @xmath138 and @xmath139 coefficients can be obtained from @xmath140 and @xmath141 by replacing : @xmath142 and @xmath143 , where @xmath144 .
the @xmath145 and @xmath146 coefficients are given by @xmath147 - c_{g_{3a } } ( y_h-1 ) y_w \big[y_h ( 1-y_w-5y_{w_h})\nonumber\\ & + & y_w^2 + 10 y_w y_{w_h}+y_w+y_{w_h}^2 + 5 y_{w_h}-2\big ] - c_{g_{3b } } ( y_h-1)y_{w_h}\nonumber\\ & \times&\big[y_h ( 1 - 5 y_w - y_{w_h})+y_w^2 + 5 y_w ( 2 y_{w_h}+1)+y_{w_h}^2+y_{w_h}-2\big]\nonumber\\ & - & ( y_h-1 ) \left(y_{w_h } ( y_h+10 y_w-1)+(y_w-1)(y_h+y_w)+y_{w_h}^2\right ) \big)\label{intga}\\ \mathcal{b}_{g_3 } & = & \frac{2}{(y_h-1)}\,\mathcal{a}_{g_{3}}\label{intgb},\end{aligned}\ ] ] where @xmath148 , @xmath149 , @xmath150 , and + @xmath151 .
in addition , @xmath152 and @xmath153 .
notice that eqs .
( [ intga ] ) and ( [ intgb ] ) reflect the mixing between @xmath62 and @xmath109 gauge bosons .
the @xmath154 and @xmath155 coefficients are @xmath156 where @xmath157 , @xmath158 , @xmath159 , and @xmath160 . here , @xmath161 and @xmath162 .
the @xmath163 and @xmath164 coefficients can be get by replacing : @xmath142 and @xmath165 , where @xmath166 .
the form factor @xmath167 is given by @xmath168 therefore , the expression for the decay amplitude of the @xmath88 can be written as @xmath169\epsilon_\mu(q)\epsilon_\nu(k_1).\label{amplitudef}\ ] ] finally , the decay width for the process reads as @xmath170}{96\ , \pi\ , m_{z_h}}.\ ] ] it should be recalled that all the @xmath171 and @xmath167 terms in @xmath172 are free of ultraviolet divergences and the lorentz structure in eq .
( [ amplitudef ] ) satisfies the ward identity @xmath173 .
in this part of our work , we present the production cross section of the extra neutral @xmath4 gauge boson at lhc in the context of the lthm , where it is assumed a search for a mass resonance in the dilepton channel @xmath174 @xcite . here , we used version 2.1.1 of the whizard event generator , which is a program designed for the calculation of multi - particle scattering cross sections and simulated event samples @xcite to perform our calculations . as a test of the whizard package
, we carried out the calculation of @xmath175 cross section as a function of @xmath176 for @xmath177 and our results coincided with previous ones reported in ref .
@xcite ; to do that we employed cteq5 parton distribution function @xcite .
we simulate @xmath5 collisions with a center of mass energy of 14 tev . in fig .
[ zhprod ] , it can be appreciated @xmath175 as a function of the @xmath4 boson mass throughout the interval @xmath178 tev@xmath179 tev , where we have employed three different values for @xmath20 , namely , @xmath180 tev .
it should be noted that the mass range for @xmath4 comes from the discrete choice of the energy scale @xmath20 and allowed values for the @xmath67 parameter , which will be justified below .
let us analyze the branching ratio behavior of the @xmath0 decay . recall that the @xmath176 is a function of two model - dependent parameters @xmath67 and @xmath20 , where @xmath67 and @xmath20 are the mixing angle of the @xmath181 extended gauge group and the energy scale at which the @xmath17 gauge group breaks into @xmath35 group , respectively .
it is known that experimental data constrain the symmetry breaking scale to be in the interval @xmath182 , for @xmath183 and @xmath67 between @xmath184 $ ] @xcite . to make predictions
, we will take three distinct values for @xmath20 , namely , @xmath2 tev and will carry out an exhaustive study at the @xmath67 parameter region given above .
this analysis will provide us crucial information to test the experimental possibility for @xmath0 decay in the lthm context .
moreover , it should be recalled that the recent results reported by atlas and cms collaborations , established lower mass limits for a new neutral massive gauge boson , identified as @xmath10 .
atlas collaboration reports that a sequential @xmath10 gauge boson is excluded at 95@xmath185 c.l . for masses below 2.39 tev in the electron channel , 2.19 tev in the muon channel , and 2.49 tev in the two channels combined @xcite ; @xmath10 bosons coming from @xmath186-motivated models
are excluded at 95@xmath185 c.l . for masses below 2.09 - 2.24 tev @xcite . in accordance with cms results , in the context of the sequential @xmath9 model and the superstring - inspired model , the lower mass limits at 95@xmath185 c.l . for the @xmath10 gauge boson
correspond to 2.59 tev and 2.26 tev , respectively @xcite .
motivated by the above results , we will take a lower mass limit for the @xmath4 gauge boson to be 2.6 tev in order to explore the physical possibilities for the @xmath0 decay . in a previous work
it has been studied the dominant decays of the @xmath4 boson @xcite in the context of the lthm .
we employ this information to compute the total decay width of the @xmath4 boson for different values of the energy scale @xmath20 proposed above . from fig .
[ brc](a ) , for @xmath187 , it can be observed that the permitted region corresponds to @xmath188 for a @xmath4 mass interval @xmath189 , where the branching ratio ranges from @xmath190 to @xmath191 ; the maximum value of the branching ratio is @xmath191 for @xmath192 and @xmath193 tev . in fig .
[ brc](b ) , for @xmath194 , we can observe a permitted region for the @xmath67 parameter ranging between @xmath195 , due to @xmath4 mass interval @xmath196 , being the associated branching ratio around @xmath197 ; in this case , the maximum value of the branching ratio is @xmath198 for @xmath199 and @xmath200 tev . finally in fig . [ brc](c ) , we show the branching ratio as a function of @xmath67 and @xmath176 for @xmath201 tev .
this figure tell us that the whole interval for the @xmath67 parameter @xmath202 is permitted , accordingly with the interval @xmath203 , where the related branching ratio is as high as @xmath204 for @xmath205 and @xmath206 tev .
even when predicted branching ratio is small ( of the order of @xmath1 ) , the lhc luminosity at final stage of operation ( 14 tev at the center of mass energy ) is expected to be around 3000 @xmath207 @xcite which could counteract this situation . in fact , to explore the predictability of the lthm , let us consider different values of the @xmath4 mass for which it is produced few events . for @xmath208 tev and around @xmath209 or equivalently @xmath210 tev , we estimate 2 events . for @xmath211 tev and @xmath212 or @xmath213 tev , we found around 1 event .
for @xmath201 tev and @xmath214 or @xmath215 tev , we calculated less than 1 event .
moreover , the maximum number of events estimated for @xmath2 tev are 16 ( @xmath216 tev ) , 4 ( @xmath217 tev ) and 0.76 ( @xmath218 tev ) , respectively . at the lhc
, the higgs boson production in association with a photon may be produced from quark - antiquark annihilation . in gluon - gluon annihilation
, the higgs - photon associated production is forbidden by furry s theorem because of properties of color singlet state of gluons @xcite .
different results were reported where the @xmath220 final states are produced in @xmath5 collisions via @xmath221 subprocess or via weak - boson fusion and @xmath222 colliders @xcite . for the process
@xmath221 the cross section was obtained at @xmath223 tev using values of transverse momentum , @xmath224 , ranging from @xmath225 gev to @xmath226 gev . for our analysis
we take @xmath5 collisions with a center - of - mass energy of @xmath227 tev and introduce kinematical cuts based on experimental values employed by atlas collaboration for the transverse momentum @xmath228 and transverse energy @xmath229 from @xmath226 gev to @xmath230 gev for the photon and pseudorapity region @xmath231 @xcite . taking in account this
, we simulated the @xmath232 process corresponding to sm background by using whizard event generator package @xcite along with parton distribution functions ( pdfs ) cteq5 , cteq6 @xcite and mstw2008nlo @xcite .
we note that the pdfs used produce similar results for the production cross section in question .
in addition , we would like to mention that with the used cuts , the sm background results ( see fig .
[ sm - background ] ) are suppressed in comparison with the corresponding lthm results . concretely , at center of mass energy of 14 tev we found a sm background cross section of @xmath233 pb ( for cteq5 ) , while in the less optimistic scenario , our calculated cross section via lthm is @xmath234 pb ( for @xmath201 tev and @xmath215 tev ) . for @xmath208 ( @xmath211 )
tev and @xmath210 ( @xmath235 ) tev respectively , we obtain a lthm production cross section @xmath236 of the order of @xmath237 pb . at @xmath238 tev for @xmath239 tev . ] to distinguish the lthm signal from sm background we compute the higgs gamma drell - yan production ( background ) at 95% of confidence level , where systematic uncertainties has been neglected ( see fig .
[ sm - background ] ) . from this figure
, we observe that all the points corresponding to @xmath240 process at 14 tev are above the confidence level band of sm background .
these lthm results could be attractive for the search of new physics , however , they must be taken carefully since we have not considered their respective error bars .
moreover , it may be recalled that the purpose of this work relies on the lthm predictability for the physical parameter space @xcite by using the @xmath88 decay , to this end we present at the most a thorough estimation . finally , it is essential to point out that it has been shown the important influence of k factors on qcd nlo corrections to drell - yan production in both tevatron and lhc @xcite , where it is found that corrections to the one - loop level under certain scenarios are important . in this sense ,
one of the goals of this work has been calculating the @xmath4 production cross section at the lhc via @xmath4 drell - yan production , which has been made possible through the implementation of pdfs such as mstw2008nlo @xcite in the whizard event generator .
these pdfs already take into account corrections for nlo drell - yan production at the lhc @xcite .
indeed , as already mentioned before , while for the @xmath4 production and the sm background analysis it has been used the cteq5 pdf , we stress that the numerical simulation also was performed using mstw2008nlo and cteq6 and no significant changes between the three pdfs are appreciated , as the results provided by these pdfs hardly vary even in the most significant digit . for consistency ,
we use cteq5 in accordance with previous experimental analyses that consider this pdf for studies of @xmath241 drell - yan production @xcite .
the lthm resides on a nonlinear sigma model with a @xmath17 global symmetry and the gauged subgroup @xmath18\otimes [ su(2)_2\otimes u(1)_2]$ ] , where it is predicted the existence of heavy gauge bosons , particularly , a new neutral massive boson known as @xmath4 .
this gauge boson is another @xmath10 type gauge boson which at present is under experimental scrutiny at the lhc . although the parameters space of the lthm has been severely constrained , yet there is room left to test the predictability of the model . in specific ,
the @xmath88 decay was used to explore the current parameters space of the lthm , where we have analyzed physical regions according with experimental bounds and results ; specifically we have taken the following parameters : @xmath2 tev for @xmath3 .
it is found that for @xmath208 tev there is a permitted region @xmath242 corresponding to @xmath189 . in particular , for a @xmath210 tev it is calculated around 2 events for the @xmath88 decay at lhc operating at @xmath227 tev .
similarly , for @xmath211 tev the permitted region is @xmath195 or a @xmath4 mass interval @xmath196 . in this case , for @xmath213 tev it is estimated 1 event for the process in question .
finally , for the same process and taking @xmath201 tev we have found the permitted region in the interval of masses @xmath203 . here , it is computed less than 1 event for @xmath215 tev .
although we have chosen specific values of @xmath176 to get few events , our numerical results tell us that there are several intervals in which the number of events are larger than the previous ones . for instance , for @xmath208 and @xmath192 we could obtain tens of events for the @xmath88 decay . to explore future experimental possibilities of the decay in question
, we have performed a careful estimation of the sm background for the @xmath14 reaction , by using the cteq5 , cteq6 , and mstw2008nlo pdfs included in the whizard event generator .
taking into account current experimental kinematical cuts , the sm - background cross section results below the computed lthm cross section for the @xmath243 .
this work has been partially supported by conacyt and cic - umsnh .
i. corts - maldonado thanks to j. j. toscano for fruitful discussions . + n. arkani - hamed , a. g. cohen , and h. georgi , phys .
b513 * , 232 ( 2001 ) ; n. arkani - hamed , a. g. cohen , t. gregoire , and j. g. wacker , jhep * 08 * , 020 ( 2002 ) ; n. arkani - hamed , a. g. cohen , e. katz , a. e. nelson , t. gregoire , and j. g. wacker , jhep * 08 * , 021 ( 2002 ) ; i. low , w. skiba , and d. smith , phys . rev . *
d66 * , 072001 ( 2002 ) .
cms collaboration , phys .
b714 * , 158 ( 2012 ) ; atlas collaboration , epj web of conferences * 49 * , 15004 ( 2013 ) ; s. chatrchyan _
_ [ cms collaboration ] , phys .
* d87 * , 072002 ( 2013 ) ; g. aad _
[ atlas collaboration ] , phys . rev . *
d90 * , 052005 ( 2014 ) .
m. cveti , p. langacker , and b. kayser , phys . rev
. lett . * 68 * , 2871 ( 1992 ) ; m. cveti and p. langacker , phys . rev . *
d54 * , 3570 ( 1996 ) ; m. cveti _
et al . _ ,
rev . * d56 * , 2861 ( 1997 ) ; _ ibid_. * 58 * , 119905(e ) ( 1998 ) ; m. masip and a. pomarol , phys
. rev . * d60 * , 096005 ( 1999 ) ; c. t. hill and e. h. simmons , phys . rept .
* 381 * , 235 ( 2003 ) ; _ ibid_. * 390 * , 553 ( 2004 ) ; j. kang and p. langacker , phys . rev . * d71 * , 035014 ( 2005 ) ; b. fuks _
et al . _ ,
b797 * , 322 ( 2008 ) ; j. erler _ et al .
_ , jhep * 08 * , 017 ( 2009 ) ; m. goodsell _ et al .
_ , jhep * 11 * , 027 ( 2009 ) ; p. langacker , aip conf . proc .
* 1200 * , 55 ( 2010 ) ; p. langacker , rev .
phys . * 81 * , 1199 ( 2009 ) ; a. leike , phys . rept . * 317 * , 143 ( 1999 ) . | we present the calculation of the @xmath0 decay in the context of the littlest higgs model at one - loop level .
our calculations include the contributions of fermions , scalars and gauge bosons in accordance with the most recent experimental constraints on the parameters space of the model .
we find branching ratios of the order of @xmath1 for the energy scale @xmath2 tev on the @xmath3 region . in order to provide a complementary study we calculated the production cross section of the @xmath4 boson in @xmath5 collisions at large hadron collider with a center of mass energy of 14 tev . by using the integrated luminosity projected for the large hadron collider in the last stage of operation , we estimated the number of events for this process
. moreover , we analyze the sm background for the higgs - photon associated production and found that the @xmath6 production is above the sm background .
= = # 1tempboxa= tempdima=0.5tempdima 0.5tempboxa -tempdima tempboxa # 1tempboxa= tempdima=0.5tempdima 0.5tempboxa -tempdima tempboxa # 1 |
there exist numerous examples for the production of particles in heavy ion collisions at bombarding energies well below the single nucleon - nucleon threshold @xcite .
this phenomenon indicates collective interactions among the many participating nucleons and thus is expected to give information about the hot and dense matter formed in these collisions . at beam energies around 1 gev per nucleon
the most extreme of these subthreshold particles is the antiproton .
therefore , it is believed to be a very sensitive probe to collective behaviour in nucleus - nucleus collisions .
however , presently neither the production mechanism nor the final state interactions of antiprotons in dense nucleonic matter are well understood .
the antiproton yields measured at gsi and bevalac @xcite seem to be described equally well by various microscopic models using different assumptions about the production mechanism and particle properties in dense nuclear matter @xcite .
this ambiguity raises the question which kind of information can be really deduced from subthreshold @xmath0 spectra . in this paper
we use a simple thermodynamic framework as a background on wich we can systematically study the influence of different assumptions on the final @xmath0 spectrum . in the next section
we will focus on the production process . following
a discussion of the final state interactions of the antiproton in dense hadronic matter in section [ sec3 ] , we use in section [ sec4 ] a one - dimensional hydrodynamic model for the exploding fireball to clarify which features of the production and reabsorption mechanisms should survive in the final spectra in a dynamical environment .
we summarize our results in section [ sec5 ] .
unfortunately very little is known about the production mechanism for antiprotons in dense nuclear matter .
therefore , we are forced to use intuitive arguments to obtain a plausible expression for the production rate . as commonly done in microscopic models
@xcite we consider only two body collisions and take the experimentally measured cross sections for @xmath0 production in free nn collisions as input .
the problem can then be split into two parts : the distribution of the two colliding nucleons in momentum space and the elementary cross sections for antiproton production .
the procedure is later generalized to collisions among other types of particles ( section [ sec2c ] ) using phase space arguments .
formally , the antiproton production rate @xmath1 , i.e. the number of antiprotons produced in the space - time cell @xmath2 and momentum space element @xmath3 , is given by @xcite @xmath4 in this expression the summation includes all incoming particle species @xmath5 and @xmath6 with 4-momenta @xmath7 and @xmath8 and phase space distributions @xmath9 and @xmath10 , respectively .
the integration is over the invariant collision energy @xmath11 and the phase space of the two colliding particles : @xmath12 finally , @xmath13 is the well known flux factor , and @xmath14 the differential @xmath0 production cross section .
the influence of the shape of the phase space distributions and of the differential production cross sections will now be investigated in more detail , beginning with the case of @xmath0 production by @xmath15 collisions .
assuming that the momentum distribution of the nucleons gradually evolves from @xmath16-peaks at the momenta of the colliding nuclei to a relativistic maxwell - boltzmann distribution , an analytical expression for the phase space distribution @xmath17 was derived by schrmann _
_ in ref .
we use this model as the basis for a numerical evaluation of the probability @xmath18 of finding two nucleons at a center - of - mass energy @xmath19 at time @xmath20 ( the second integral on the right hand side of eq .
( [ p ] ) ) .
[ bild1 ] shows @xmath18 at @xmath21 fm / c after the nuclei have started overlapping , assuming a bombarding energy of 1 gev / a .
although the initial @xmath16-like distribution at @xmath22 widens quite rapidly , the extreme states needed for antiproton production ( @xmath23 ) are populated substantially only after about 5 fm / c . at this moment
the distribution is already nearly a relativistic maxwell - boltzmann distribution ( dashed line ) which maximizes the @xmath0 production rate in our approach .
this fact motivates us to neglect pre - equilibrium production and to switch on @xmath0-production only when the system has reached local thermal equilibrium .
similiar conclusions have been drawn from the results of a microscopic kinetic model for the phase space evolution @xcite .
there it was demonstrated that only nucleons which have suffered at least several collisions are effective in @xmath0 production .
there are no experimental data for the nucleonic processes @xmath24 @xmath25 close to the production threshold . as already pointed out in @xcite , different extrapolations of the data at higher energies down to threshold change the @xmath0-yield by orders of magnitude .
however , we will show that not only the total yield , but also the shape of the spectrum is very sensitive to the threshold behaviour .
we choose for the production cross section the parametrization @xmath26 where the parameter @xmath27 models the shape of the cross section near threshold .
the normalization @xmath28 is determined by the experimentally measured point closest to threshold .
if the behaviour of @xmath29 were dominated by the available phase space for the outgoing particles ( this is the fundamental assumption in fermi s statistical model @xcite ) , then @xmath30 . in order to obtain from the total cross section ( [ sigtot ] ) a formula for the differential cross section , we assume like others that the momentum distribution of the produced particles is mainly governed by phase space
this leads to the simple relationship @xmath31 here @xmath32 is the volume the @xmath33-particle phase space , which can be given analytically in the non - relativistic limit @xcite .
@xmath1 is the total 4-momentum of the 4-particle final state , which reduces to @xmath34 in the cms of the four particles . inserting ( [ sigtot ] ) and ( [ sigdiff ] ) into ( [ p ] ) and using thermal phase space distributions
, the production spectrum can be evaluated numerically . as seen in fig .
[ spektrum ] , the sensitivity on the threshold power @xmath27 is particularly pronounced for low - energy antiprotons .
the strong deviation at low @xmath0 energies from a thermal shape ( dashed lines ) for small values of @xmath27 is most remarkable . using the fermi value @xmath35 for the cross section , an analytical expression for the production rate
can be obtained : @xmath36 here @xmath37 are the densities of the incoming particles , @xmath38 is the threshold for producing an antiproton with momentum @xmath39 , and @xmath40 is the local temperature of the environment . clearly the rate @xmath1 is proportional to the product of the densities of the two colliding particle species @xmath5 and @xmath6 .
more importantly , this formula shows the extreme sensitivity on the production threshold and on the medium temperature which both appear in the exponent of the exponential function .
the production of resonances as intermediate energy storage has been pointed out as an important mechanism for @xmath0 production in @xcite .
because no experimental data are available for antiproton production by baryonic resonances , we follow common strategy @xcite and assume that the cross section @xmath41 is independent of the internal state of excitation of the colliding baryons in our thermal picture .
the consequences of this assumption are quite obvious . while the distance to the @xmath0-threshold is reduced by the larger rest mass of the resonances , the mean velocity of a heavy resonance state in a thermal system is smaller than that of a nucleon .
both factors counteract each other , and indeed we found that the total rate @xmath1 is not strongly changed by the inclusion of resonances . the role of pionic intermediate states for @xmath0-production in @xmath42-collisions was pointed out by feldman @xcite .
as mesons are created numerously in the course of a heavy ion collision , mesonic states gain even more importance in this case .
in fact , ko and ge @xcite claimed that @xmath43 should be the dominant production channel . relating the @xmath44-production channel to the @xmath45 annihilation channel @xcite by @xmath46 where @xmath47 is the spin factor for the @xmath48 ,
the production rate can be calculated straightforwardly from eq .
( [ p ] ) : @xmath49 the spin - isospin factor of the @xmath48 is @xmath50 , and @xmath51 is the energy of the produced antiproton .
the modified bessel function @xmath52 results from the assumption of local thermal equilibration for the @xmath48-distribution .
expanding the bessel function for large values of @xmath53 we see that the `` temperature '' @xmath54 of the @xmath0-spectrum is only half the medium temperature : @xmath55 . finally , the meson - baryon collision process @xmath56 remains as the last prototype of collisions with two incoming particles .
we choose a parametrisation for the cross section which reflects its qualitative behaviour @xcite : @xmath57 comparing this form with measured data on @xmath58 collisions @xcite , a value of @xmath59 mb is obtained .
due to the threshold behaviour of eq .
( [ sigpi ] ) and the rather large value of @xmath60 , it turns out @xcite that this process is by far the most important one in a _ chemically _ equilibrated system .
however , this chemical equilibration
if achieved at all is reached only in the final stages of the heavy ion collision when cooling has already started .
so it is by no means clear whether the dominance of the meson - baryon channel remains valid in a realistic collision scenario .
this point will be further discussed in section [ sec4 ] .
once an antiproton is created in the hot and dense hadronic medium , its state will be modified by interactions with the surrounding particles .
two fundamentally different cases have to be distinguished : elastic scattering , which leads to a reconfiguration in phase space , driving the momentum distribution towards a thermal one with the temperature of the surrounding medium , and annihilation .
each process will be considered in turn .
the time evolution of the distribution function @xmath61 is generally described by the equation @xcite @xmath62 where @xmath63 is the transition probability from momentum state @xmath64 at time @xmath65 to state @xmath39 at @xmath66 . because the number density of antiprotons is negligible compared to the total particle density in the system
, the evolution of @xmath61 can be viewed as a markov process .
assuming furthermore that the duration of a single scattering process @xmath67 and the mean free path @xmath68 are small compared to the typical time scale @xmath69 and length scale @xmath70 which measure the variation of the thermodynamic properties of the system , @xmath71 eq . ( [ f1 ] ) can be transformed into a master equation . considering the structure of the differential @xmath45 cross section one
notices that in the interesting energy range it is strongly peaked in the forward direction @xcite .
therefore , the master equation can be approximated by a fokker - planck equation @xcite : @xmath72 for the evaluation of the friction coefficient @xmath73 and the diffusion coefficient @xmath74 we follow the treatment described by svetitsky @xcite .
for the differential cross section we took a form suggested in @xcite : @xmath75 } { |t|}\,,\ ] ] where the constants @xmath76 are fitted to experimental data .
@xmath68 is the wave number of the relative motion , @xmath20 the transferred four momentum and @xmath77 a spherical bessel function .
the essential result @xcite is that the coefficients @xmath73 and @xmath78 are nearly independent of the @xmath0 momentum and satisfactorily fulfill the einstein relation @xmath79 the effect of these coefficients can be visualized by examining the time evolution of a non - relativistic maxwell - boltzmann distribution whose initial slope @xmath80 is different from the temperature of the medium : @xmath81 in a medium at fixed temperature @xmath82 mev and normal nuclear density @xmath83 the solution @xmath61 of the fokker - planck equation can be calculated analytically : @xmath84 \equiv \exp{\left [ - p^2/2 m t_{\rm eff}(t)\right ] } \,.\ ] ] this shows that the exponential shape of the distribution function is maintained throughout the time evolution , but that the slope @xmath85 gradually evolves from @xmath80 to the value @xmath86 which , according to the einstein relation ( [ einstein ] ) , is the medium temperature @xmath40 .
looking at fig .
[ abb33 ] it is clear that after about 10 fm / c the spectrum is practically thermalized
. therefore initial structures of the production spectrum ( like the ones seen in fig .
[ spektrum ] ) are washed out quite rapidely , and their experimental observation will be very difficult
. the annihilation of antiprotons with baryons is dominated by multi - meson final states @xmath87 . for the parametrisation of the annihilation cross section
@xmath88 we take the form given in @xcite for the process @xmath89 using the same philosophy as for the calculation of the production rate , a simple differential equation for the decrease of the antiproton density in phase space can be written down : @xmath90 this equation shows that the absorption rate is given by the thermal average of the annihilation cross section @xmath91 = \langle v \sigma^{\rm ann } \rangle$ ] where @xmath92 is the @xmath0 velocity .
although @xmath93 is strongly peaked for low - energy antiprotons , the thermal average @xmath94 is practically independent of the @xmath0 velocity in the medium .
therefore , in a stationary medium the shape of the @xmath0 spectrum is maintained ; the momentum independence of @xmath73 leads only to a renormalisation of the spectrum by annihlation .
any modifications of the spectral shape must be due to the dynamical evolution of the hadronic medium .
the absolute value of @xmath73 decreases slightly with increasing temperature ; it is of the order of 0.5 ( fm/@xmath95)@xmath96 at normal nuclear matter density .
this means that 99% of the antiprotons are absorbed within 9 fm / c .
of course , due to the rather large value of the absorption rate @xmath73 which is of the order of the size of the antiproton , the classical approach chosen in our model becomes questionable .
more reliable results should be based on a quantum field theoretic calculation which is beyond the scope of this paper .
in order to compare the results of the two previous sections with experimental data we connect them through a dynamical model for the heavy ion reaction . in the spirit of our thermodynamic approach the so - called hadrochemical model of montvay and zimanyi @xcite
is applied for the simulation of the heavy ion collision . in this picture
the reaction is split into two phases , an ignition and an explosion phase , and particles which have at least scattered once are assumed to follow a local maxwell - boltzmann distribution .
in addition , a spherically symmetric geometry is assumed for the explosion phase .
the included particle species are nucleons , @xmath97-resonances , pions and @xmath48-mesons . as initial condition
a fermi - type density distribution is taken for the nucleons of the incoming nuclei , @xmath98 with the nuclear radius @xmath99 and a surface diffuseness @xmath100 fm . because the fermi motion is small compared to the typical bombarding energy of 1 gev / a
, it can be neglected , and the initial phase space distribution therefore reads @xmath101 here @xmath102 describes the path of the center of the nucleus .
due to elastic and inelastic scattering hot and dense matter is formed in the central region .
its time - dependent chemical composition is governed by the kinetic equations of the hadrochemical model given in ref .
their particular form needed here can be found in @xcite .
some of the resulting distributions are visualized in fig .
[ vert90 ] for a central @xmath103ca-@xmath103ca collision with momentum @xmath104 mev in the c.m . system .
one clearly sees that at the moment of full overlap of the two nuclei a dense zone with hot nucleons , resonances and mesons ( not shown ) has been formed . only in the peripheral regions `` cold '' target and projectile nucleons can still be found .
on the other hand , chemical equilibrium of the hot collision zone is not reached in the short available time before the explosion phase sets in ; pions and in particular @xmath48 mesons remain far below their equilibrium abundances @xcite .
it is important to note that due to the arguments given in section [ sec2 ] @xmath0 production is strongly suppressed in this initial stage of the reaction . in our simple model
we have in fact neglected this early @xmath0 production completely .
the ignition phase is only needed to obtain the chemical composition of the hot fireball which is expected to be the dominant source for the creation of antiprotons .
for the subsequent expansion of the spherically symmetric fireball into the surrounding vacuum analytical solutions can be given if the equation of state of an ideal gas is taken as input @xcite .
if excited states are included in the model , an exact analytical solution is no longer possible .
because a small admixture of resonaces is not expected to fundamentally change the dynamics of the system , we can account for their effect in first order by adjusting only the thermodynamic parameters of the exploding fireball , but not the expansion velocity profiles .
there is one free parameter in the model @xcite , @xmath27 , which controls the density and the temperature profiles , respectively .
small values @xmath105 represent @xmath16-function like density profiles whereas @xmath106 corresponds to a homogeneous density distribution throughout the fireball ( square well profile ) .
the time - dependent temperature profiles for two representative values of @xmath27 are shown in fig .
[ abb47 ] for different times @xmath20 starting at the time @xmath107 of full overlap of the nuclei . clearly , a small value of @xmath27 leads to an unreasonably high temperature ( @xmath108 mev ) in the core of the fireball at the beginning of the explosion phase , and should thus be considered unphysical .
based on the time - dependent chemical composition of this hadrochemical model we can calculate the spectrum of the antiprotons created in a heavy ion collision .
let us first concentrate on the influence of the density distribution in the fireball characterized by the shape parameter @xmath27 .
due to different temperature profiles connected with different @xmath27 values ( see fig .
[ abb47 ] ) the absolute normalization varies substantially when the density profile is changed . for more @xmath16-like shapes an extremely hot fireball core
is generated , whereas for increasing @xmath27 both density and temperature are more and more diffuse and spread uniformly over a wider area . because of the exponential dependence upon temperature a small , but hot core raises the production rate drastically .
this fact is illustrated in fig .
[ abb54 ] for three values of @xmath27 . not only the total normalization , but also the asymptotic slope of the spectrum is modified due to the variation of the core temperature with @xmath27 as indicated in the figure . comparing the dotted lines , which give the pure production spectrum , to the solid lines representing the asymptotic spectrum at decoupling , the tremendous effect of antiproton absorption in heavy ion collisions is obvious .
as one intuitively expects absorption is more pronounced for low - energetic antiprotons than for the high - energetic ones which have the opportunity to escape the high density zone earlier .
therefore , the finally observed spectrum is flatter than the original production spectrum . interestingly , while the baryon - baryon and the @xmath44 channels are comparable in their contribution to @xmath0 production , the pion - baryon channel turned out to be much more effective for all reasonable sets of parameters .
this fact is indeed remarkable , because here , contrary to the discussion in section [ sec2 ] , the pions are not in chemical equilibrium ; in our hadrochemical model the total time of the ignition phase is too short to saturate the pion channel .
the meson - baryon channel is thus crucial for understanding @xmath0 spectra . only by including all channels reliable predictions about the antiprotons
can be drawn .
we did not mention so far that in our calculations we followed common practice and assumed a finite @xmath0 formation time of @xmath109 fm / c ; this means that during this time interval after a @xmath0-producing collision the antiproton is assumed to be not yet fully developed and thus can not annihilate . however , there are some ( although controversial ) experimental indications of an extremely long mean free path for antiprotons before final state interactions set in @xcite .
we have tested the influence of different values for the formation time @xmath67 on the @xmath0 spectrum .
[ abb55 ] shows that this highly phenomenological and poorly established parameter has a very strong influence in particular on the absolute normalization of the spectra , i.e. the total production yield . in the light of this uncertainty it appears difficult to argue for or against the necessity for medium effects on the antiproton production threshold based on a comparison between theoretical and experimental total yields only . in all the calculations shown above a bombarding energy of 1 gev / a has been assumed .
experimental data are , however , only available at around 2 gev / a . at these higher energies thermalization becomes more questionable @xcite , and our simple model may be stretching its limits .
especially , the temperature in the fireball core becomes extremely high . in order to avoid such an unrealistic situation and in recognition of results from kinetic simulations @xcite we thus assume that only part of the incoming energy is thermalized in the following a fraction of 70% was taken .
[ bild111 ] shows calculations for the antiproton spectrum from na - na and ni - ni collisions at a kinetic beam energy of 2 gev / nucleon .
the calculation assumes a @xmath0 formation time of @xmath110 fm/@xmath95 , and takes for the density and temperature profiles the parameter value @xmath111 which corresponds to an upside - down parabola for the density profile . comparing with the gsi data @xcite we see our model features too weak a dependence on the size of the collision system ; the absolute order of magnitude of the antiproton spectrum is , however , correctly reproduced by our simple hadrochemical model , without adjusting any other parameters .
no exotic processes for @xmath112 production are assumed . as mentioned in the previous subsection ,
the pion - baryon channel is responsible for getting enough antiprotons in our model , without any need for a reduced effective @xmath0 mass in the hot and dense medium @xcite .
the existing data do not yet allow for a definite conclusion about the shape of the spectrum , and we hope that future experiments @xcite will provide additional contraints for the model .
heavy ion collisions at typical bevalac and sis energies are far below the @xmath45-production threshold . as a consequence , pre - equilibrium antiproton production in such collisions
is strongly suppressed relative to production from the thermalized medium produced in the later stages of the collision .
therefore , @xmath0 production becomes important only when the heavy ion reaction is sufficiently far progressed , in accordance with microscopic simulations @xcite . by assuming a local maxwell - boltzmann distribution for the scattered and produced particles forming the medium in the collision zone one maximizes the @xmath0 production rate ( see fig . [ bild1 ] ) .
if , contrary to the assumptions made in this work , the extreme states in phase space described by the tails of the thermal boltzmann distribution are not populated , the antiproton yield could be reduced substantially .
we also found that the threshold behaviour of the @xmath0 production cross section is not only crucial for the total @xmath0 yield , but also introduces structures into the initial @xmath0 spectrum .
this might give rise to the hope that by measuring the @xmath0 momentum spectrum one may obtain further insight into the @xmath0 production mechanism .
on the other hand we saw here , using a fokker - plank description for the later evolution of the distribution function @xmath113 in a hot environment , that these structures are largely washed out by subsequent elastic scattering of the @xmath0 with the hadrons in the medium .
in addition , the large annihilation rate reduces the number of observable antiprotons by roughly two orders of magnitude relative to the initial production spectrum ; the exact magnitude of the absorption effect was found to depend sensitively on the choice of the @xmath0 formation time @xmath67 .
we have also shown that meson ( in particular pion ) induced production channels contribute significantly to the final @xmath0 yield and should thus not be neglected .
we were thus able to reproduce the total yield of the measured antiprotons in a simple model for the reaction dynamics without including , for example , medium effects on the hadron masses and cross sections @xcite .
however , we must stress the strong sensitivity of the @xmath0 yield on various unknown parameters ( e.g. the @xmath0 formation time ) and on poorly controlled approximations ( e.g. the degree of population of extreme corners in phase space by the particles in the collision region ) , and emphasize the rapidly thermalizing effects of elastic final state interactions on the @xmath0 momentum spectrum . we conclude that turning subthreshold antiproton production in heavy ion collisions into a quantitative probe for medium properties and collective dynamics in hot and dense nuclear matter remains a serious challenge .
u. mosel , annu .
nucl . part .
1991 , vol .
41 , 29 a. schrter _
et al . _ , z. phys .
a350 ( 1994 ) 101 a. shor _
et al . _ ,
63 ( 1989 ) 2192 s. teis , w. cassing , t. maruyama , and u. mosel , phys .
c50 ( 1994 ) 388 g.q .
li and c.m .
ko , phys .
c50 ( 1994 ) 1725 g. batko _
et al . _ , j. phys .
g20 ( 1994 ) 461 c. spieles _ et al .
a27 ( 1993 ) 2547 g.f .
bertsch and s. das gupta , phys .
160 ( 1988 ) 189 p. koch , b. mller , and j. rafelski , phys .
42 ( 1986 ) 167 b. schrmann , k. hartmann , and h. pirner , nucl .
a360 ( 1981 ) 435 g. batko _
et al . _ , phys .
b256 ( 1991 ) 331 r.h .
milburn , rev .
( 1955 ) 1 , and references therein c.m .
ko and x. ge , phys .
b 205 ( 1988 ) 195 p. koch and c. dover , phys .
c 40 ( 1989 ) 145 g. feldman , phys . rev . 95 ( 1954 ) 1697 landolt - brnstein , _ numerical data and functional relationships in science and technology _ , vol .
12a und vol . 12b , springer - verlag berlin , 1988 r. wittmann , ph .
d. thesis , univ .
regensburg , feb .
van kampen , _
stochastic processes in physics and chemistry _ , north - holland publishing company , amsterdam 1983 b. conforto _
_ , nouvo cim .
54 a ( 1968 ) 441 w. brckner _
b 166 ( 1986 ) 113 e. eisenhandler _ et al .
b 113 ( 1976 ) 1 s. chandrasekhar , rev . mod . phys . 15
( 1943 ) 1 b. svetitsky , phys .
d 37 ( 1988 ) 2484 i. montvay and j. zimanyi , nucl . phys .
a316 ( 1979 ) 490 j.p .
bondorf , s.i.a .
garpman , and j. zimnyi , nucl .
a296 ( 1978 ) 320 a.o .
et al . _ , jetp lett . 29 ( 1979 ) 661 a. lang _
b245 ( 1990 ) 147 a. gillitzer _
_ , talk presented at the xxxiii international winter meeting on nuclear physics , 23.-28 .
january 1995 , bormio ( italy ) | we study the structure of antiproton spectra at extreme subthreshold bombarding energies using a thermodynamic picture .
antiproton production processes and final state interactions are discussed in detail in order to find out what can be learned about these processes from the observed spectra . |
inclusive @xmath8 decays are a precise probe of the underlying @xmath3- to @xmath4-quark transition because hadronisation effects are small and have a simple structure . these effects are suppressed by powers of the heavy - quark mass and given in terms of a small number of non - perturbative parameters . in the heavy - quark limit ,
the hadronic decay rate becomes equal to the partonic decay rate .
the leading corrections are of order @xmath9 and are given in terms of two non - perturbative heavy - quark parameters , @xmath10 and @xmath11 which are the @xmath12-meson matrix elements of the kinetic and chromo - magnetic operator respectively .
schematically , the decay rate takes the form @xcite @xmath13 where @xmath14 .
the coefficients @xmath15 , @xmath16 and @xmath17 can be calculated in perturbation theory : @xmath18 the general structure of the expansion is the same for other observables , such as partial rates or moments of the decay spectrum , but the calculable coefficients @xmath15 , @xmath16 and @xmath17 are different . for the total rate the kinetic corrections have the same coefficient as the leading order , @xmath19 . also for other observables , such as partial rates and moments , the kinetic corrections can be obtained from the leading - power differential rate , but the relations are more complicated and are evaluated to @xmath20 for the first time in this paper .
to turn ( [ eq : rate ] ) into a precision determination of @xmath2 one needs the values of @xmath21 , @xmath22 and the heavy - quark parameters . since the same parameters enter moments of the decay spectrum , one can determine these parameters by measuring not only the rate , but also a number of moments . to this end , lepton energy moments and hadronic invariant mass and energy moments are measured @xcite . using the results of these measurements ,
several groups have performed fits of the theoretical expressions to the experimental data @xcite .
the theoretical expressions that are used in the fit include one - loop corrections to the leading - power coefficients @xmath23 @xcite as well as the @xmath24-part of the two - loop corrections @xcite , while the coefficients @xmath25 and @xmath26 of the power corrections are known only at the tree level .
in addition to the second - order power corrections proportional to @xmath10 and @xmath11 , the fits also include the third - order power corrections , which involve two additional hadronic parameters , @xmath27 and @xmath28 @xcite ( the fourth order corrections are now available as well @xcite ) .
this technique yields the most precise determination of @xmath2 together with very precise determinations of the heavy - quark masses .
already now , the estimated theoretical uncertainties are somewhat larger than the experimental ones @xcite . in the future
the experimental uncertainty will decrease further : the babar moment measurements which were used in @xcite were published in 2004 and are based on @xmath29 of data @xcite @xcite .
] , and the recently published belle measurements on @xmath30 @xcite , but combined the two experiments have already collected more than @xmath31 of data .
also , based on the convergence of the perturbative series of the rate for @xmath32-decay and based on the size of the two - loop contributions that arise when converting the theoretical expressions between different schemes used in the literature , it has been suggested that the theoretical uncertainties in the results of the moment fits might be underestimated @xcite .
whether the uncertainties are reliable is an important question because the value of @xmath21 extracted from the fit is a crucial ingredient for the determination of @xmath33 from inclusive decays . after imposing the severe cuts necessary to eliminate the charm background
, the prediction for the @xmath34 rate behaves as @xmath35 with @xmath36 @xcite .
the value and uncertainty of the extracted @xmath33 are thus strongly correlated with the value and uncertainty of @xmath21 .
it is clearly desirable to increase the precision of the theoretical predictions . to achieve this goal ,
two ingredients are needed : the leading - power moments have to be evaluated to two - loop accuracy , and the coefficients of the power corrections proportional to @xmath10 and @xmath11 need to be evaluated to one loop . in this paper
we take the first and simplest step in this direction by evaluating the coefficient of the kinetic operator to one - loop accuracy .
let us stress that , while it is demanding , also the @xmath37 calculation of the leading - power moments is doable .
a few years ago such a calculation looked prohibitively difficult , but in the meantime the necessary methods to perform it numerically have been developed @xcite .
indeed , the two - loop correction for muon decay @xmath38 , the qed equivalent of @xmath0 , has been evaluated recently using this method @xcite .
we use the same numerical approach for our one - loop calculation , because the size of the expressions involved is such that an analytic calculation does not look feasible .
since the method is gauge invariant ( we perform the calculation without introducing a gluon mass ) , it is also suited for the calculation of the corrections to the coefficient of the chromomagnetic operator .
the kinetic corrections are obtained by expanding the leading - power @xmath20 expressions up to second order in the small residual momentum of the @xmath3-quark inside the @xmath12-meson . to have a check of our results , we perform the calculation in two different ways .
a straightforward and tedious way of obtaining the kinetic corrections is to expand the leading - power feynman diagrams in the residual momentum before performing the loop and phase - space integrations .
the resulting expressions are long and involve terms which are individually strongly infrared divergent .
another complication arises because the expansion produces not only standard phase - space integrals , but also derivatives of such phase - space integrals , which arise from cutting higher powers of propagators .
a much more elegant and efficient way to perform the calculation is to expand the result for the leading - order differential rate in the residual momentum . in this way one obtains results for the moments in terms of integrals over the leading - power rate and its derivative .
in fact , without experimental cuts one obtains simple algebraic relations between the kinetic corrections and the leading - power moments . as a byproduct of our analysis , we obtain the @xmath20 corrections to the leading - power moments and reproduce the numerical results of @xcite .
as the two - loop corrections to the leading - power rate and the one - loop corrections to the @xmath11 terms are not yet available , it is too early to perform a detailed phenomenological analysis .
instead , we present numerical values for a few reference values of the electron energy cut . for the moments which do not vanish , the corrections we calculate turn out to be small , below @xmath39 as long as the cut on the lepton energy is not too strong . for the moments of the partonic invariant mass @xmath40 , on the other hand , which vanish at the tree - level and leading power , the corrections are larger , of order @xmath41 .
we expect the corrections proportional to @xmath11 to be more important than the kinetic corrections . for the tree - level rate
, they are roughly a factor ten larger than the kinetic corrections . this paper is organized as follows . in the next section ,
we explain how the kinetic corrections are calculated using the operator product expansion ( ope ) . in section [ phasespace ]
we give parameterizations for the phase - space and loop integrals which are needed to perform the calculation .
these parameterizations map the integration onto the unit hypercube and are such that infrared divergences appear only in a single variable and can be isolated before performing the numerical integration .
we present our numerical results in section [ results ] . in the same section , we give formulae for the conversion of partonic to hadronic moments .
we also sketch discuss how to convert our results into different schemes , such as the kinetic @xcite , potential - subtracted @xcite , @xmath42 @xcite or the shape - function scheme @xcite .
in this section we briefly explain how we evaluate the @xmath8 decay rate using the ope .
the application of the ope to inclusive @xmath12 decays has been worked out quite some time ago and the reader interested in more details should consult the original references @xcite or the textbook @xcite .
our goal in this section is to outline the necessary steps to obtain the result and to discuss some of the technicalities which are encountered in the course of the one - loop calculation .
the reader solely interested in the numerical results can skip to section [ results ] .
the @xmath8 decay is mediated by the effective hamiltonian @xmath43 the decay rate factors into a leptonic tensor @xmath44 and a hadronic tensor @xmath45 @xmath46 where @xmath47 and @xmath48 denotes the phase space @xmath49 in @xmath50 dimensions . since the differential rate is a finite quantity , we could set @xmath51
however , individual contributions to the hadronic tensor contain ultra - violet ( uv ) as well as infrared ( ir ) divergences which we regulate by keeping @xmath52 throughout .
the spin - averaged leptonic tensor is @xmath53\ , .\ ] ] the hadronic tensor is obtained by taking the imaginary part of the time - ordered products of currents @xmath54 where @xmath55 |\bar b(p_b)\rangle\,.\ ] ] we work in the kinematics @xmath56 and our states are canonically normalized . in analytic calculations the hadronic tensor is usually decomposed into five form factors , but we prefer to directly evaluate the relevant product @xmath57 . since we use dimensional regularization we need to specify how we treat @xmath58 in @xmath59 dimensions .
a definition of the axial current in @xmath59 dimensions suitable for our purposes has been given by larin @xcite and we adopt it for our calculation . the operator product appearing in ( [ tmunu ] ) is expanded in a series of local operators which corresponds to an expansion of the rate in inverse powers of the @xmath3-quark mass . to perform the expansion ,
one first removes a rapidly oscillating factor from the @xmath3-quark field by writing it as @xmath60 . using heavy - quark effective theory ( hqet ) @xcite ,
all the matrix elements necessary to second order in the expansion can be reduced to @xmath61 the hqet parameter @xmath10 is often denoted by @xmath62 and is not renormalized . up to terms suppressed by three powers of the heavy - quark mass
, the decay rate thus takes the form @xmath63 \ , .\nonumber\ ] ] the wilson coefficients @xmath64 of the three operators are independent of the external states and can be calculated using partonic initial and final states . to extract the coefficient @xmath65 of the kinetic operator , it is simplest to use an on - shell @xmath3-quark with momentum @xmath66 , which amounts to calculating the partonic decay rate @xmath67 . to find the coefficient of the operator @xmath68 with two derivatives , we expand the partonic rate to second order in the residual momentum @xmath69 . the result takes the form @xmath70 at the loop level
, the question arises whether to expand the diagrams before or after the loop integration . in general , either choice is valid as long as one evaluates the loop corrections to the operator product and to the matrix elements of the local operators @xmath71 in the same way . in our case the situation is especially simple : since we perform the matching calculation on - shell , the one - loop corrections to the hqet matrix elements of the operators @xmath71 vanish and the loop integration commutes with the expansion in the residual momentum . to have a check of our results , we evaluate the corrections in both ways .
we can further simplify the calculation by averaging over the direction of the transverse momentum @xmath72 .
the component parallel to @xmath73 is fixed by the on - shell condition @xmath74 .
taking the average we have @xmath75 to obtain the hadronic rate , we first bring the leading - power partonic matrix element into the form ( [ matrixelements ] ) by rewriting @xmath76 we then replace the partonic matrix elements by the corresponding hadronic matrix elements : @xmath77 \,\frac{1}{2m_b } \langle \bar b(p_b)|\ , { \bar b}_v\ , ( id)^2 b_v\ , | \bar b(p_b)\rangle + \dots \nonumber \\ = & a-\frac{\mu_\pi^2}{2m_b^2}\left[a - a_\mu\ , v^\mu + a_{\mu\nu } \frac{2}{d-1 } ( g_{\mu\nu}-v_\mu v_\nu)\right ] + \dots\,.\end{aligned}\ ] ] the ellipsis denotes terms which are suppressed by @xmath78 or proportional to @xmath11 .
the operator @xmath79 has a vanishing @xmath3-quark matrix element and its coefficient is therefore not determined by our matching calculation .
[ b]@xmath80 [ b]@xmath81 [ b]@xmath82 , title="fig:",scaledwidth=35.0% ] to illustrate the structure of the result , we now calculate the kinetic corrections at tree level .
the tree - level contribution to the hadronic tensor is shown in figure [ wmunutree ] .
taking its imaginary part and contracting with the leptonic tensor , the partonic tree - level rate is found to be @xmath83 we now expand up to second order in the residual momentum @xmath69 , average over the @xmath84-direction and replace the partonic by the hadronic matrix elements to obtain the result @xmath85\,,\end{gathered}\ ] ] where @xmath86 and @xmath87 \,,\\
f_1&=\frac{\mu_\pi^2}{3 } \big [ v\cdot p_c v\cdot p_\nu\ , ( 2 m_b v+5 p_c)\cdot p_\ell - p_c\cdot p_\ell \,(5 m_b v+2 p_c)\cdot p_\nu \big ] \
f_2&=2\,\mu_\pi^2\ , m_b\ , v\cdot p_\nu p_c\cdot p_\ell\ , \left[(v\cdot p_c)^2- p_c^2 \right]\ , .
\end{aligned}\ ] ] the one - loop results for the rate have a similar structure , also in this case the result contains @xmath88 with @xmath89 . to calculate moments of the decay spectrum
, we introduce an integration over the partonic phase space @xmath90 the result for the decay rate then takes the form @xmath91 f_0\right|_{m_x^2=m_c^2 } -\left .
\frac{\rm d}{{\rm d } m_x^2 } \int \left [ { \rm d } \pi \right ] f_1\right|_{m_x^2=m_c^2 } + \left .
\frac{\rm d^2}{{\rm d } ( m_x^2)^2 } \int \left [ { \rm d } \pi \right ] f_2\right|_{m_x^2=m_c^2}\ , , \ ] ] where we have used the notation @xmath92
\equiv \int \left [ { \rm d } \pi_{b\rightarrow x+\ell+\bar\nu}\right ] = \int\ ! { \rm d}\mu(p_x )
\int\ ! { \rm d}\mu(p_\ell )
\int\ ! { \rm
d}\mu(p_\nu ) \ , ( 2\pi)^d \delta^d(m_b v - p_x - p_\ell - p_\nu)\ , .\ ] ] it turns out that for the total rate the derivative terms in ( [ treeratediff ] ) do not contribute
. however , for partial rates or spectral moments these terms do give non - vanishing contributions . to evaluate ( [ treeratediff ] ) numerically ,
we need a suitable parameterization for the phase - space integral ( [ treephasespace ] ) .
the necessary parameterizations , both for the tree - level phase space and the phase space with the emission of an additional gluon , needed for the @xmath20 corrections to the rate , are given in the next section .
an alternative , more elegant and efficient way of evaluating the kinetic corrections was described in @xcite . instead of expanding the diagrams which contribute to the hadronic tensor ,
one takes the result for the differential partonic rate and expands in the residual momentum . to derive this result ,
it is convenient to introduce the dimensionless variables @xmath93 , @xmath94 and @xmath95 . when expanding in the residual momentum , one has @xmath96 expanding to second order and averaging over the @xmath84-direction , the hadronic differential rate is equal to @xcite @xmath97 \frac{{\rm d}\gamma^{\rm partonic } } { { \rm d}x\ , { \rm d}y \,{\rm d}\hat{q}^2}\ , .\end{gathered}\ ] ] the @xmath10-term without derivatives comes from expanding the matrix elements , see ( [ eq : matrix ] ) .
what makes this result particularly useful is that we can explicitly evaluate the derivatives using integration by parts when calculating moments . for the moments with a cut on the lepton energy @xmath98_{y_0 } = \int \ ! { \rm d}x\,{\rm d}y \,{\rm d}\hat{q}^2 \ , \frac{{\rm d}\gamma } { { \rm d}x\ , { \rm d}y\,{\rm d}{\hat q}^2 } \,x^n\ , y^m\ , ( { \hat q}^2)^l \theta(y - y_0 ) \ , , \ ] ] one finds @xmath99_{y_0 } = \left [ x^n\ , y^m\ , ( { \hat q}^2)^{l } \right]_{y_0}^{\rm partonic } \\ + \frac{\mu_\pi^2}{6m_b^2 } \left [ \left((n+m)^2 + 2m+2 n-3\right ) x^n\ , y^m\ , ( { \hat q}^2)^l-4\ , m\ , n \ , x^{n-1}\ , y^{m-1}\ , ( { \hat q}^2)^{l+1 } \right]_{y_0}^{\rm partonic } \\ + \frac{\mu_\pi^2}{6m_b^2 } \left [ \left((m+2n+1 ) x y_0 -4 n \hat{q}^2 \right ) x^{n-1 } y_0^{m } ( { \hat q}^2)^{l } \ , \delta(y - y_0 ) + x^{n } y_0^{m+2 } ( { \hat q}^2)^{l } \delta'(y - y_0 ) \right]^{\rm partonic } \ , . \end{gathered}\ ] ] the terms in the third line are boundary terms and vanish when setting @xmath100 . in this case
the kinetic corrections to the moments follow via simple algebraic relations from the leading - power term .
the explicit relations for the moments we are interested in are given in appendix [ app : momrel ] .
in the general case with a cut on the lepton energy @xmath101 , one needs to also evaluate moments of the partial rate and its first derivative .
to evaluate these boundary terms , it is important to keep in mind that @xmath102 , @xmath103 and @xmath104 are not completely independent : the rate includes a factor @xmath105 and at tree level the variables fulfill @xmath106 .
more generally , when choosing a phase - space parameterization to evaluate ( [ eq : momrel ] ) , the variables @xmath107 and @xmath103 become functions of @xmath104 and the derivative in the third line of ( [ eq : momrel ] ) then acts not only on the rate but also on the factors @xmath108 .
[ cols="^,^,^,^ " , ] ccccccc & @xmath109 & @xmath110 & @xmath111 & @xmath112 & @xmath113 & @xmath114 + @xmath109 & @xmath115 & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 + @xmath121 & @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 + @xmath128 & @xmath129 & @xmath130 & @xmath131 & @xmath132 & @xmath133 & @xmath134 + @xmath135 & @xmath136 & @xmath137 & @xmath129 & @xmath138 & @xmath139 & @xmath140 + @xmath141 & @xmath142 & @xmath143 & @xmath144 & @xmath145 & @xmath146 & @xmath147 + @xmath148 & @xmath149 & @xmath150 & @xmath151 & @xmath152 & @xmath153 & @xmath154 + @xmath155 & @xmath156 & @xmath154 & @xmath157 & @xmath158 & @xmath159 & @xmath160 + @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 & @xmath166 & @xmath167 + @xmath168 & @xmath169 & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 + @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 & @xmath181 + @xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & @xmath188 + @xmath189 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 + @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath181 & @xmath201 +
when doing the calculation , it is simplest to evaluate moments using partonic variables . to distinguish partonic and hadronic quantities , we denote the partonic energy and invariant mass by @xmath202 and @xmath203 , while writing @xmath204 and @xmath205 in the hadronic case . for the tree - level diagrams
@xmath202 is the energy of the charm quark , while in the diagrams where a gluon is emitted @xmath206 .
the partonic moments for which we present results in tables are defined as @xmath207 = \int_{e_0}^{e_{\rm max}}\!\!\ !
de_l \int\!\ !
de_x\ , dp_x^2 \,\frac{d\gamma}{de_x\ , dp_x^2\ , de_l } w(e_l , e_x , p_x^2)\,.\ ] ] we consider lepton energy moments @xmath208 with @xmath209 and the partonic energy and invariant mass moments @xmath210 with @xmath211 and @xmath14 .
note that we do not normalize the partonic moments to the rate .
numerical results for the moments without a cut and with a cut @xmath212 ( corresponding to @xmath213 for @xmath214 ) are shown in tables [ tab : res00 ] and [ tab : res04 ] for @xmath215 . in the last column
, we indicate the relative size of the kinetic @xmath20 corrections . to estimate the relative size we use @xmath216 and @xmath217 . as explained in section [ kinetic ] , we calculate the moments with two different methods and check that the results agree within numerical precision . in the tables
, we also include numerical results for the tree - level moments . using ( [ eq : momrel ] ) it would be simple to evaluate them analytically .
the one - loop kinetic corrections are small for the moments which get contributions at leading power but sizable for the moments of @xmath218 .
for example , the moment of @xmath219 , gets a correction of @xmath220 .
compared to the tree - level contributions of the kinetic operator , the one - loop terms are typically suppressed by a few times @xmath221 .
we thus expect that the extracted value of @xmath10 will be shifted by about @xmath222 by their presence .
given the size of the kinetic corrections to the rate , we do not expect that the @xmath20 corrections will affect the extracted value of @xmath2 . in the fit of ref .
@xcite the value of @xmath10 is varied by @xmath223 to estimate the theoretical uncertainty .
the corrections we calculate are indeed of this size , except that varying the value of @xmath10 correlates the change in all moments , while the perturbative corrections are different in each case . in the fit of @xcite ,
the corrections are underestimated to be @xmath224 : the contributions we find are roughly ten times larger . in table
[ interpolation ] we give the result for the one - loop kinetic corrections as a function of the cut energy @xmath225 and the ratio @xmath14 . to this end
, we perform a quadratic fit around default values @xmath226 and @xmath227 . the accuracy of the quadratic fit is a few per cent except in cases where the corrections become very small .
tables with precise numerical results for arbitrary cut energies and charm - mass values can be obtained from the authors . instead of the partonic moments , experimental papers present results for the normalized hadronic moments @xmath228 to translate the results to hadronic kinematics we note that leptonic quantities are identical on the hadronic and partonic level .
using that the @xmath12-meson momentum is @xmath229 , it follows that @xmath230 with these two equations , it is straightforward to translate our partonic results into hadronic language .
for example , the prediction for the lowest moments are obtained from the relations @xmath231 } m_b \left [ \hat { e}_\ell \right ] \ , , \nonumber\\ \langle e_x \rangle & = \frac{1}{\left [ 1 \right ] } \left ( m_b \left[{\hat e}_x \right ] + ( m_b - m_b)\,\left [ 1 \right]\right)\,,\\ \langle p_x^2-m_d^2 \rangle & = \frac{1}{\left [ 1 \right ] } \left ( m_b^2\,\left[\hat{p}_x^2-\rho\right ] + 2 m_b ( m_b - m_b)\,\left[{\hat e}_x \right ] + ( ( m_b - m_b)^2+(m_c^2-m_d^2))\,\left [ 1 \right]\right)\ , . \nonumber\end{aligned}\ ] ] the moment with unit weight function is the rate : @xmath232=\gamma(e_\ell > e_0)$ ] , see ( [ momentsdef ] ) . while the moments of @xmath233 vanish at tree level in the heavy - quark limit , the hadronic moments @xmath234 are nonzero . using the above conversion formulae together with the results in table [ tab : res04 ] we obtain for example @xmath235 \,{\rm gev}^2\ , , \\
\langle ( p_x^2-m_d^2)^2 \rangle & = \left [ 0.939 + 7.00 \frac{\alpha_s}{\pi}+\left(117.4 -178.2 \frac{\alpha_s } { \pi}\right)\frac{\mu_\pi^2}{2m_b^2}\right ] \,{\rm gev}^4 \,.\end{aligned}\ ] ] finally , let us note that we performed our calculation in the pole scheme , but it is simple to convert our result into different schemes .
the pole scheme is calculationally most convenient , but plagued by large higher - order corrections .
the problem arises because the definition of the parameters in this scheme relies on on - shell quark states , a concept not meaningful beyond perturbation theory .
the resulting bad perturbative behavior can be improved by using parameter definitions with less infrared sensitivity such as @xmath236 quark masses .
this is appropriate for the charm quark which we treat as light , however , the @xmath236 mass definition is not suitable for the @xmath3-quark , because it is not consistent with hqet power counting .
a number of alternative schemes , appropriate for heavy - quark processes , are available : they include the kinetic @xcite , the potential - subtracted @xcite , the @xmath42 @xcite and the shape - function scheme @xcite . to one - loop order ,
the scheme changes from the pole into the new schemes have the form @xmath237 where coefficients @xmath238 depend on the scheme .
for example , in the kinetic scheme @xmath239 , @xmath240 , @xmath241 , @xmath242 and @xmath243 .
the exact choice of the factorization scale @xmath244 is a matter of convention . when performing the moment fit , it is worthwhile to check to what extent the fit results are independent of this choice .
a compendium of two - loop scheme conversion formulae can be found in @xcite .
let us note that only the kinetic and the shape - function scheme provide improved definitions for the parameter @xmath10 . to obtain the scheme change
to @xmath20 one simply replaces the pole - scheme parameters by the redefined ones in our one - loop results .
additional @xmath20 corrections are generated when performing the scheme change in the tree - level results .
we refrain here from explicitly changing the scheme , but the default values of our parameters are chosen such that they correspond to values which are typical for the improved schemes .
we have evaluated the one - loop perturbative corrections to the coefficient of the kinetic operator in the operator product expansion of the decay @xmath245 .
the corrections are typically @xmath246 times the leading kinetic power correction .
we thus expect that these corrections will change the extracted value of @xmath10 from the moment fit by about @xmath1 .
whether this in turn has an effect on the extracted @xmath21 and @xmath22 values is hard to estimate without performing the global fit .
since the kinetic corrections are very small for the total rate , the value of @xmath2 will likely not be affected . with the same numerical methods used here the one - loop chromo - magnetic and the two - loop leading - power corrections
can be calculated as well .
once these are known , the theoretical precision on the predictions for the @xmath8 decay will be superior to the experimental accuracy .
these results will increase the precision of the extracted parameters and provide a nontrivial consistency check on the experimental data used in the fit .
the calculation will also answer the question whether the currently used theoretical error estimates on the extracted parameters are realistic .
this is of particular importance because the value and uncertainty of @xmath21 is a crucial input in the determination of @xmath33 from the inclusive @xmath247 decay .
we thank r. feger , t. feldmann , t. mannel , k. melnikov , m. neubert and g. paz for discussions .
h. b. thanks the fermilab theoretical physics department for hospitality .
fermilab is operated by fermi research alliance , llc under contract no .
de - ac02 - 07ch11359 with the united states department of energy .
without cuts on the available phase space , the kinetic corrections to the moments are directly related to the leading - power moments .
the relations can be derived from the general result ( [ eq : momrel ] ) . for convenience , we list here the explicit form of the relations for the moments we are interested in .
we write the relations as @xmath248 which should be read as @xmath249 for the energy moments the relations read @xmath250 and for the partonic invariant mass moments the relations are @xmath251 j. chay , h. georgi and b. grinstein , phys .
b * 247 * , 399 ( 1990 ) .
i. i. y. bigi , m. a. shifman , n. g. uraltsev and a. i. vainshtein , phys .
lett . * 71 * , 496 ( 1993 ) [ arxiv : hep - ph/9304225 ] .
b. blok , l. koyrakh , m. a. shifman and a. i. vainshtein , phys .
d * 49 * , 3356 ( 1994 ) [ erratum - ibid .
d * 50 * , 3572 ( 1994 ) ] [ arxiv : hep - ph/9307247 ] .
a. v. manohar and m. b. wise , phys .
d * 49 * , 1310 ( 1994 ) [ arxiv : hep - ph/9308246 ] .
b. aubert _ et al .
_ [ babar collaboration ] , phys .
d * 69 * , 111104 ( 2004 ) [ arxiv : hep - ex/0403030 ] .
b. aubert _ et al . _ [ babar collaboration ] , phys .
d * 69 * , 111103 ( 2004 ) [ arxiv : hep - ex/0403031 ]
. s. e. csorna _ et al . _
[ cleo collaboration ] , phys . rev .
d * 70 * , 032002 ( 2004 ) [ arxiv : hep - ex/0403052 ] .
a. h. mahmood _ et al .
_ [ cleo collaboration ] , phys . rev .
d * 70 * , 032003 ( 2004 ) [ arxiv : hep - ex/0403053 ] .
j. abdallah _ et al .
_ [ delphi collaboration ] , eur .
j. c * 45 * , 35 ( 2006 ) [ arxiv : hep - ex/0510024 ] .
d. acosta _ et al .
_ [ cdf collaboration ] , phys . rev .
d * 71 * , 051103 ( 2005 ) [ arxiv : hep - ex/0502003 ] .
urquijo _ et al . _
[ belle collaboration ] , phys .
d * 75 * , 032001 ( 2007 ) [ arxiv : hep - ex/0610012 ] . c. schwanda
_ et al . _ [ belle collaboration ] , phys .
d * 75 * , 032005 ( 2007 ) [ arxiv : hep - ex/0611044 ] . c. w. bauer , z. ligeti , m. luke and a. v. manohar , phys .
d * 67 * , 054012 ( 2003 ) [ arxiv : hep - ph/0210027 ] . c. w. bauer , z. ligeti , m. luke , a. v. manohar and m. trott , phys .
d * 70 * , 094017 ( 2004 ) [ arxiv : hep - ph/0408002 ] . o. buchmller and h. flcher , phys .
d * 73 * , 073008 ( 2006 ) [ arxiv : hep - ph/0507253 ] .
k. abe _ et al . _ [ belle collaboration ] , arxiv : hep - ex/0611047 .
m. jezabek and j. h. khn , nucl .
b * 314 * , 1 ( 1989 ) .
m. jezabek and j. h. khn , nucl .
b * 320 * , 20 ( 1989 ) .
a. czarnecki , m. jezabek and j. h. khn , acta phys .
b * 20 * , 961 ( 1989 ) .
a. czarnecki and m. jezabek , nucl .
b * 427 * , 3 ( 1994 ) [ arxiv : hep - ph/9402326 ] .
m. b. voloshin , phys .
d * 51 * , 4934 ( 1995 ) [ arxiv : hep - ph/9411296 ] .
a. f. falk , m. e. luke and m. j. savage , phys .
d * 53 * , 2491 ( 1996 ) [ arxiv : hep - ph/9507284 ] .
p. gambino and n. uraltsev , eur .
j. c * 34 * , 181 ( 2004 ) [ arxiv : hep - ph/0401063 ] .
m. trott , phys .
d * 70 * , 073003 ( 2004 ) [ arxiv : hep - ph/0402120 ] .
n. uraltsev , int .
j. mod .
a * 20 * , 2099 ( 2005 ) [ arxiv : hep - ph/0403166 ] . v. aquila , p. gambino , g. ridolfi and n. uraltsev , nucl .
b * 719 * , 77 ( 2005 ) [ arxiv : hep - ph/0503083 ] .
m. e. luke , m. j. savage and m. b. wise , phys .
b * 343 * , 329 ( 1995 ) [ arxiv : hep - ph/9409287 ] .
p. ball , m. beneke and v. m. braun , phys .
d * 52 * , 3929 ( 1995 ) [ arxiv : hep - ph/9503492 ] .
m. gremm and i. stewart , phys .
d * 55 * , 1226 ( 1997 ) [ arxiv : hep - ph/9609341 ] .
a. f. falk and m. e. luke , phys .
d * 57 * , 424 ( 1998 ) [ arxiv : hep - ph/9708327 ] .
m. gremm and a. kapustin , phys .
d * 55 * , 6924 ( 1997 ) [ arxiv : hep - ph/9603448 ] .
b. m. dassinger , t. mannel and s. turczyk , jhep * 0703 * , 087 ( 2007 ) [ arxiv : hep - ph/0611168 ]
. b. aubert , _ et al . _ [ babar collaboration ] , arxiv:0707.2670 [ hep - ex ] .
m. neubert , talk at fpcp 2007 , may 12 - 16 , 2007 , bled , slowenia .
t. becher and m. neubert , phys .
b * 535 * , 127 ( 2002 ) [ arxiv : hep - ph/0105217 ] .
b. o. lange , m. neubert and g. paz , phys .
d * 72 * , 073006 ( 2005 ) [ arxiv : hep - ph/0504071 ] .
t. binoth and g. heinrich , nucl . phys .
b * 585 * , 741 ( 2000 ) [ arxiv : hep - ph/0004013 ] . c. anastasiou , k. melnikov and f. petriello , phys .
d * 69 * , 076010 ( 2004 ) [ arxiv : hep - ph/0311311 ] ; c. anastasiou , k. melnikov and f. petriello , arxiv : hep - ph/0505069 .
n. uraltsev , nucl .
b * 491 * , 303 ( 1997 ) [ arxiv : hep - ph/9610425 ] .
m. beneke , phys .
b * 434 * , 115 ( 1998 ) [ arxiv : hep - ph/9804241 ] .
a. h. hoang , z. ligeti and a. v. manohar , phys .
lett . * 82 * , 277 ( 1999 ) [ arxiv : hep - ph/9809423 ] ; phys .
d * 59 * , 074017 ( 1999 ) [ arxiv : hep - ph/9811239 ] .
s. w. bosch , b. o. lange , m. neubert and g. paz , nucl .
b * 699 * , 335 ( 2004 ) [ arxiv : hep - ph/0402094 ] .
s. a. larin , phys .
b * 303 * , 113 ( 1993 ) [ arxiv : hep - ph/9302240 ] .
a. v. manohar and m. b. wise , camb .
monogr . part .
cosmol .
* 10 * , 1 ( 2000 ) .
m. neubert , phys .
rept . * 245 * , 259 ( 1994 ) [ arxiv : hep - ph/9306320 ] .
m. neubert , phys .
b * 612 * , 13 ( 2005 ) [ arxiv : hep - ph/0412241 ] . | we evaluate the one - loop corrections to the wilson coefficient of the kinetic operator in the operator product expansion of the differential @xmath0 decay rate . with a moderate cut on the lepton energy ,
the one - loop terms change the kinetic operator contributions to spectral moments by about @xmath1 .
this amounts to a small correction for leptonic and hadronic moments , except for those which vanish at the lowest order , where the effect can be sizable .
together with a two - loop calculation of the leading - power rate and an evaluation of the one - loop corrections to the wilson coefficient of the chromo - magnetic operator , our results will allow for a high - precision determination of @xmath2 and the @xmath3- and @xmath4-quark masses .
fermilab - pub-07 - 406-t + si - hep-2007 - 11 + august 6 , 2007 * kinetic corrections to @xmath5 at one loop * thomas becher@xmath6 , heike boos@xmath7 and enrico lunghi@xmath6 + |
while dust makes up only a very small fraction of the baryonic mass in galaxies @xcite , it leaves a large signature on their spectral energy distributions ( seds ) .
dust extinguishes light in a wavelength - dependent way , and therefore distorts the intrinsic sed of galaxies .
this distortion , or net dust attenuation , may depend on the properties of the dust , the dust - to - star geometry , or both quantities .
therefore , recovering the intrinsic stellar seds from observations requires a thorough understanding of both factors . dust properties and geometry
are studied using observations of both dust emission and attenuation .
studies of dust emission at mid- and far - infrared wavelengths have placed constraints on the composition , distribution , and mass of dust in nearby galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
measurements of dust attenuation are also needed to completely characterize the nature of dust , including the dust - to - star geometry .
one method of constraining the dust - to - star geometry is by comparing the integrated dust attenuation with the attenuation towards star - forming ( sf ) regions .
dust attenuation affecting the stellar continuum , @xmath10 , has been measured using a number of methods .
these include _
( i ) _ line of sight measurements ( e.g. , the mw , smc , lmc ) , _ ( ii ) _ matching attenuated galaxies with unattenuated galaxies with similar intrinsic stellar populations ( i.e. * ? ? ? * ; * ? ? ?
* ) , _ ( iii ) _ fitting the sed with stellar population synthesis models , including a prescription for dust , and _
( iv ) _ the @xmath11 ratio ( also known as irx ) , which probes dust attenuation using energy conservation .
the latter ratio is directly related to the uv continuum slope @xmath12 @xcite , which may be used to infer the dust content for galaxies for which no ir data is available .
( see @xcite for more discussion on this topic ) .
dust attenuation towards sf regions , @xmath13 , has also been extensively studied in low - redshift galaxies . @xmath14is
most directly probed with recombination line flux ratios , often using the balmer decrement h@xmath15/@xmath16 .
the intrinsic line ratio can be calculated given reasonable environmental parameters .
as dust attenuation is wavelength dependent , the measured line ratio compared with the intrinsic ratio combined with an assumed dust law yields a measure of the amount of dust attenuation towards sf regions .
this method was used by a number of studies to measure attenuation towards sf regions in nearby galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . by comparing @xmath0and @xmath1 ,
@xcite find that there is extra dust attenuation towards star - forming regions relative to the integrated dust content for local starburst galaxies .
@xcite expand on this work by finding that the amount of extra attenuation increases with the axial ratio and decreases with ssfr .
this implies that the dust content of galaxies might have two components ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) : a component associated with the short - lived birth clouds in sf regions and a diffuse component distributed throughout the ism . in this model ,
the diffuse dust attenuates light from all stars , while the birth cloud dust component only attenuates light originating from the sf region . at higher redshifts ,
the nature of dust attenuation is much more poorly understood .
much of the work on dust attenuation in high-@xmath17 galaxies has focused on the uv slope ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* also see @xcite for a comprehensive review ) , as it is relatively easy to observe .
however , deviations from the @xcite irx-@xmath12 relation have been found for various galaxy samples ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
additional methods of measuring attenuation in high-@xmath17 galaxies include sed modeling of photometric or spectroscopic observations ( e.g. , * ? ? ?
* ; * ? ? ?
* ) andcomparison of star formation rate ( sfr ) indicators ( e.g. , * ? ? ?
direct measurements of dust attenuation toward hii regions using balmer decrements are very challenging for @xmath18 as both @xmath3and @xmath16are shifted to the less - accessible near - infrared window .
careful survey design and instrument improvements have made measurements of the balmer decrement possible for larger samples of intermediate redshift galaxies , e.g. , between @xmath19 @xcite , @xmath20 @xcite , and @xmath21 @xcite . however , until recently balmer decrements have been measured for only a small number of more distant galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
interestingly , current studies of dust properties in distant star - forming galaxies yield contrasting results . @xcite and @xcite find that extra attenuation towards hii regions leads to an overestimate of the @xmath3sfr relative to the uv slope sfr .
however , @xcite compare measured and predicted @xmath22 and @xmath3equivalent widths and find the best agreement when @xmath0includes extra attenuation relative to @xmath1 .
a comparison of overlapping objects with @xcite shows that the previous aperture corrections might be overestimated , which could have masked some extra attenuation .
@xcite compare @xmath1from sed fitting and @xmath0from balmer decrements for a small sample and find that the high-@xmath17 objects are consistent with the local universe @xcite prescription for extra dust attenuation .
additionally , @xcite @xcite , @xcite and @xcite find the best agreement between @xmath3sfrs and uv+ir / sed sfrs when extra attenuation is adopted , either the same as the @xcite relation @xcite or a slightly lower ratio @xcite .
@xcite also measure the balmer decrement and find the amount of extra attenuation is lower than the @xcite relation .
these contrasting results may not be surprising , given the different and indirect methods and/or small and biased samples of most studies .
direct measurements of a statistical sample of distant galaxies are required to clarify these dust properties .
this is now possible , as new nir instruments with multiple object spectroscopy capabilities are able to measure the balmer decrement for larger and more complete samples of high - redshift objects . in particular , the _ hubble space telescope _
wfc3/g141 grism filter provides slit - less spectra , allowing for a non - targeted survey of a large number of high - redshift galaxies .
the _ hst _
grisms also avoid atmospheric near - ir absorption .
a number of surveys , including the 3d - hst survey @xcite , candels @xcite , and the wisp survey @xcite , have taken advantage of the _ hst _ grism capabilities to survey high redshift galaxies .
@xcite were the first to use wfc3 grism data to make measurements of the balmer decrement on a large , non - targeted sample .
however , as their sample was not drawn from regions of the sky with existing deep photometric coverage , they were unable to examine trends of dust versus integrated galaxy properties .
we present a statistical study of dust attenuation measured using the balmer decrement for a large , non - targeted sample of galaxies at @xmath23 .
both rest - frame optical spectra and deep photometry are available , allowing us to compare attenuation towards hii regions with the total integrated dust attenuation and other galaxy properties , including stellar mass and sfr . throughout this paper
we adopt a @xmath24cdm universe with @xmath25 , @xmath26 , and @xmath27 .
our sample is drawn from the 3d - hst survey @xcite , a _ hubble space telescope ( hst ) _ treasury program adding acs and wfc3 slit - less grism observations to the well - covered candels @xcite fields : aegis , cosmos , goods - s , and uds .
the 3d - hst data also include observations of the goods - n field from program go-11600 ( pi : b. weiner ) . in this work we use the 3d - hst wfc3/g141 grism spectra , which cover @xmath28 .
the raw grism dispersion is @xmath29 , but interlacing during data reduction improves the dispersion to @xmath30 ( corresponding to about 10 restframe @xmath31 for @xmath23 ) .
the g141 grism has a maximum resolution of @xmath32 , corresponding to @xmath33 in the middle of our wavelength range .
the 3d - hst survey makes use of existing deep photometric coverage in each of the survey fields , combining the grism spectra with multi - wavelength photometric data .
the 3d - hst photometric catalogs are discussed in detail in @xcite
. this work uses version 2.1 of the photometric and grism catalogs .
a modified version of the eazy code @xcite is used on the combined grism spectra + photometry to measure the redshifts , emission line fluxes , and rest - frame u , v , j fluxes of the individual 3d - hst galaxies .
stellar masses , integrated dust attenuation , sfrs , and specific star formation rates ( ssfrs ) are determined by fitting stellar population synthesis models to the photometric data using the fast code @xcite .
we use a separate set of parameters than those used by @xcite , for reasons discussed in section [ sec : sfrha_sfrsed ] .
we use the @xcite stellar population synthesis models , assuming a @xcite stellar initial mass function , solar metallicity , an exponentially declining star formation history with a minimum _ e_-folding time of @xmath34 , a minimum age of 40 myr , and an integrated dust attenuation @xmath35between 0 and 4 assuming the dust attenuation law by @xcite .
@xcite provide complete details on the 3d - hst survey data reduction and parameter measurement procedure .
we select galaxies in the redshift range @xmath36 , for which @xmath3and @xmath16are generously covered by the g141 grism .
in addition we impose a signal - to - noise ratio ( snr ) cut for @xmath3of snr @xmath37 , to measure a decent line signal .
we make no @xmath16snr cut , to avoid biasing our sample against the dustiest galaxies .
we have a number of additional selection criteria , to ensure high quality of the spectra .
first , to avoid cases of line misidentification , the photometric - only and grism+photometry redshifts ( hereafter referred to as the grism redshifts ) must have good agreement , i.e. @xmath38 .
second , the contamination from other sources ( which is an issue because of the slit - less nature of the grism spectra ) may not exceed 15% .
third , there must be grism coverage of at least 95% and must include the @xmath3and @xmath16lines .
fourth , no more than 50% of the spectrum may be flagged as problematic ( due to bad pixels or cosmic rays ) during reduction . to study dust attenuation towards star forming regions ,
we do not want agn to contaminate our emission lines . to reject agn , we exclude any objects that have a detected x - ray luminosity @xmath39 @xcite by matching against the chandra deep field north and south surveys @xcite and the xmm - newton serendipitous survey @xcite .
furthermore we reject all objects that fall within the @xcite irac agn region , or that fall within the quiescent box in the uvj diagram defined in @xcite ( as the line emission likely originates from an agn ) .
finally , we visually inspect the grism spectra and photometry of the preliminary sample to reject problematic objects ( i.e. objects with incorrect line identification or poor quality broadband photometry ) .
our final sample includes 163 galaxies .
figure [ fig : samp_char ] shows how our sample compares to the full galaxy distribution at a similar redshift .
the selected galaxies all have relatively high sfrs , and lie along the `` star - forming main sequence '' ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) .
the spectra of individual galaxies in this sample are too noisy to yield a clear measurement of the balmer decrement , the ratio of the flux of @xmath3to @xmath16(@xmath40 ) .
thus , we bin galaxies by parameter ( sed @xmath35 , ssfr , sfr , stellar mass ) and stack the spectra before measuring line fluxes . prior to stacking , we scale the spectra to match the photometry , as the spectra and photometry have a slightly different slope for most galaxies . a linear correction for this effect is calculated during the reduction stage , and we apply this linear correction to the individual grism spectra .
individual spectra are also corrected for contamination from other sources during the reduction process .
we adopt a uniform methodology for stacking spectra within a bin .
first , we only use the portion of the grism spectra that falls between @xmath41 and @xmath42 ( observed wavelength ) to avoid noise at the edge of the grism coverage
. then the individual spectra are continuum normalized by scaling the biweighted mean value of the flux between @xmath43 ( rest - frame ) to unity .
the individual spectra are then interpolated onto a common rest - frame wavelength grid .
the normalized , interpolated spectra are stacked at each wavelength .
for each bin , simulated spectra ( @xmath44 ) are generated by perturbing the individual spectra , assuming that the flux errors are normally distributed , and then stacking the perturbed spectra using the same procedure as above .
the simulated spectra are used in determining the errors of the emission lines ( see section [ sec : line_measurement ] ) .
the best - fit stellar population models are sampled over the same wavelength regime as the original spectra and used as the continua .
the best fits are stacked in the same way : they are continuum normalized and interpolated onto a common rest - frame wavelength grid , then averaged together .
we then convolve the stacked continua models with the stacked line profiles ( discussed in section [ sec : line_measurement ] ) to match the resolution of the grism spectra . to estimate the error in the continua , we perturb the photometry of an object and determine the fast best - fit model for each perturbation .
this procedure is repeated a number of times for each object . within each bin ,
we randomly select a best - fit model to the perturbed photometry for each object and stack to construct a simulated continuum .
this procedure is repeated 500 times , in order to create a continuum model for each simulated spectrum , as discussed above .
the average photometry and stacked best - fit models with errors are shown in the appendix .
finally , we apply a second - order polynomial correction between the stacked grism spectra and the stacked best - fit stellar population continua , to further correct for possible mismatches .
this small change corrects for otherwise uncharacterized differences across the grism spectra .
the corrected , stacked spectra and continua for the different bins are shown in figure [ fig : spec ] .
the adopted continuum normalization scheme leads objects with higher scaled @xmath45+[nii ] ( see section [ sec : line_measurement ] ) fluxes to have more weight in our line stack . to correctly compare the parameters from sed fitting ( @xmath1 , stellar mass , sfr , ssfr ) with values calculated from the stacked lines , we compute the weighted average of each parameter .
we use the scaled @xmath45+[nii ] fluxes as the individual objects weights .
errors on the average parameters are estimated with bootstrapping . to measure line fluxes for the stacked spectra
, we subtract the convolved , stacked continua from the tilt - corrected , stacked spectra .
we then fit the emission lines in the resulting spectra using least - squares minimization . for our sample s redshift range ,
the grism spectra have rest - frame coverage of the following spectral lines : @xmath16 , [ oiii]@xmath46 , @xmath3 , [ nii]@xmath47 , and [ sii]@xmath48 . however , the resolution of the grism data is insufficient to separate the @xmath3and [ nii ] lines , so we measure the blended h@xmath15+[nii]@xmath49 line .
the grism line shapes are not well described by gaussian profiles . because the spatial resolution of the wfc3 detector ( @xmath50 ) is much greater than the spectral resolution ,
the spectral line profiles are dominated by the object shapes .
thus , for each object the line profile is measured by summing the direct image from f140w or f160w over the spatial direction , which is perpendicular to the dispersion direction .
the composite line profiles are created by flux - normalizing the individual profiles , multiplying each profile by the object s scaled @xmath45+[nii ] flux ( described in section [ sec : stacking ] ) , and finally averaging .
this method yields a composite profile with the same effective weighting of the objects as results from the spectrum stacking method . finally , we scale the profile width to match the width of @xmath45+[nii ] , yielding the line profile template for each stack .
the grism spectra have roughly constant spectral resolution .
thus for each line , we scale the line profile width by @xmath51 . lines in a spectrum may not have the same profile , possibly due to dust , age gradients , or agn contribution ( e.g. , * ? ? ?
. however , the line fits obtained while using the same line profile ( with appropriate width and amplitude scaling ) match the data very well , suggesting that assuming a single profile for a stack is a reasonable approximation .
because of the low spectral resolution of the grism spectra , we simultaneously fit the [ oiii ] doublet and @xmath16 , and similarly the blended h@xmath15+[nii ] line and the [ sii ] doublet .
we fix the line ratio between [ oiii]@xmath52 and [ oiii]@xmath53 to 3:1 to reduce the number of degrees of freedom in our fit , and we fix the redshift of all lines to the value measured for @xmath3 .
we compute the emission line fluxes and ratios from the best - fit line profile parameters .
the errors on the line fluxes and ratios are estimated using the simulated stacked spectra and continua . for each simulation
we measure the best - fit line fluxes and ratios in a similar fashion as for the real stacked spectrum .
the errors on the line fluxes and ratios are calculated from the resulting distributions .
the best - fit line measurements for our stacks are shown in figure [ fig : spec ] . to measure the balmer decrement
, we need to correct the blended h@xmath15+[nii ] line for the [ nii ] contribution .
we use the stellar mass versus [ nii]@xmath54/@xmath3 relation measured in @xcite for galaxies at @xmath55 , as our sample covers a similar range of masses and sfrs , and is close in redshift .
in @xcite the stellar masses are calculated using the integrated sfh , while we use the current stellar mass . for the galaxies in our sample , which are all reasonably young
, the mass from the integrated sfh is about 10% higher than the current stellar mass .
we use this estimate to scale down the masses given by @xcite to match our stellar mass definition .
we interpolate the values @xcite report in table 2 to estimate the ratio of [ nii]@xmath54/@xmath3given the weighted average stellar mass in each bin .
we assume an intrinsic line ratio of 3:1 between [ nii]@xmath54 and [ nii]@xmath56 to scale this ratio to include the second [ nii ] line .
we use the resulting ratio to calculate the @xmath3flux in each stack .
we do not include the systematic errors in the [ nii]/@xmath3ratio in our flux errors .
the balmer decrement , @xmath57 , lets us determine the amount of dust attenuation towards star - forming regions by comparing the measured ratio with the expected line ratio given the physical conditions of the region .
we assume that the hii region has a temperature @xmath58 , an electron density of @xmath59 , and that the ions undergo case b recombination .
these assumptions result in an intrinsic ratio of @xmath60 @xcite .
we assume the reddening curve @xmath61 of @xcite , which gives us @xmath62.\ ] ] we combine this expression with @xmath63 , assuming the value @xmath64 from @xcite to calculate the attenuation @xmath0from the @xmath3and @xmath16flux measured for each stack . vs. @xmath0measured from the spectra stacked in bins of @xmath1 , shown in the top panel of fig [ fig : spec ] , both corrected ( black circles ) and uncorrected ( open grey circles ) for [ nii ] contamination .
the size of the data points corresponds to the fraction of the scaled @xmath45+[nii ] flux in each bin .
bins without a significant detection ( @xmath65 ) of @xmath16 are shown as @xmath66 lower limits .
the @xmath1errors shown are the @xmath67 scatter within the bins .
the black line shows the best - fit line to our [ nii ] corrected values , which has a slope of 1.86 .
the fit error is shown with the shaded grey region .
the best - fit indicates there is extra attenuation towards emission line regions .
the data are consistent with the ratio of @xmath0to @xmath1from @xcite ( blue dashed line ) , @xcite ( purple dotted line ) , and both relations found by @xcite ( green long dash , dash - dot lines ) .
the data are inconsistent with the assumption of no extra dust attenuation towards emission line regions ( orange dash - dot - dot line ) .
, scaledwidth=48.0% ] we first investigate @xmath0 in bins of @xmath1 , to better constrain the currently contested relationship between the integrated dust content and the dust associated with sf regions for high - redshift galaxies .
we choose bins of @xmath1to probe the full range of integrated stellar dust attenuation in our sample , from low to medium to high attenuation .
we stack the spectra in these bins and measure @xmath0on the stack using the relations given in section [ sec : measuring_dust_attenuation ] .
the results are shown in figure [ fig : avplot ] .
we perform a least - squares ratio fit to the data .
the best - fit relation , assuming @xmath68 is the same for the stellar continuum and the hii regions , is @xmath69 indicating that on average @xmath0is 1.86 times higher than @xmath1 in star - forming galaxies at @xmath23 .
this is a slightly lower amount of extra attenuation than the ratio of 2.27 which @xcite find for low redshift star - forming galaxies , but our data are not statistically different from the @xcite relation .
the data are inconsistent with the assumption of no extra dust attenuation towards star - forming regions and are inconsistent with a constant value at the @xmath70 levels , respectively .
the data are consistent with the results by @xcite , who find an average relation between @xmath1and @xmath71 for galaxies at @xmath72 .
their relation ( shown in figure [ fig : avplot ] by the dotted purple line ) is the dust attenuation required for agreement between @xmath3sfrs and uv+ir sfrs , or if there was no ir detection , sed sfrs .
our data are also consistent with the relations @xcite find by comparing uv and @xmath3sfr indicators ( green dash - dot line ) and measured from the balmer decrement ( green long dash line ) .
we also show the results without correcting the @xmath3flux for [ nii ] ( open grey circles ) .
these points overestimate the amount of @xmath0relative to @xmath1 , demonstrating the necessity of correcting for [ nii ] when measuring the attenuation of the star - forming regions using grism data . however , it is important to note that there may be considerable scatter in the @xmath1and @xmath0values for individual galaxies , so this result only holds on average for a collection of galaxies . in this section
we probe the change in dust properties over bins of stellar mass , ssfr , and sfr .
we select bin boundaries for each of these properties to distribute the number of galaxies as equally as possible between the bins .
the left panels of figure [ fig : avratio ] show the results for the stacks in stellar mass , including the measured @xmath0and weighted average @xmath1(top ) and the difference between @xmath0and @xmath1(@xmath73 , middle ) as a function of stellar mass .
we also plot the ratio @xmath74 ( bottom ) to facilitate comparison with previous studies .
we see an increase in both @xmath1and @xmath0with increasing stellar mass , inconsistent with a constant value at the @xmath75 levels , respectively . the plot of @xmath73shows a roughly constant amount of extra attenuation with mass .
the data are consistent with a constant value of @xmath73 .
the ratio @xmath74 is consistent with the value found by @xcite for all mass bins .
the results for stacks in sfr are shown in the center panel of figure [ fig : avratio ] . both @xmath1and @xmath0show an increase with sfr , at the @xmath76 levels , respectively . @xmath73is
consistent with no trend with sfr .
as before , we show the ratio @xmath74 for direct comparison with past work . on average
, the ratio is below the value found in @xcite .
the results for stacks in ssfr are shown in the right panel of figure [ fig : avratio ] . both @xmath1and @xmath0decrease with increasing ssfr , and
are inconsistent with a constant value at the @xmath77 levels , respectively .
we also find a slight decreasing trend in @xmath73with ssfr , however this trend is not significant as the difference between the data and a constant value is only @xmath78 .
again , we show the ratio @xmath74 for comparison .
our ratio is consistent with the @xcite value for the lowest two ssfr bins , while our ratio for the highest ssfr bin is lower . interestingly , @xmath73is
most strongly correlated with ssfr , rather than stellar mass or sfr .
we perform a least - squares linear fit to @xmath73vs .
log ssfr , using an offset in log ssfr to avoid correlated intercept and slope errors .
we find a best - fit relation of @xmath79.\ ] ] this possible trend could be explained by the two - component dust model , as we discuss in section [ sec : interpretation ] .
the observed extra attenuation towards emission - line regions , and the decrease in the amount of extra attenuation with increasing ssfr , are consistent with a two - component dust model ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this model assumes there is a diffuse ( but possibly clumpy ) dust component in the ism that affects both the older stellar populations and star - forming regions , as well as a component associated with the short - lived stellar birth clouds that only affects the stars within those regions . for galaxies with the highest ssfrs
, the continuum light is dominated by young , massive stars .
these massive stars would predominately still reside in the birth clouds .
so for the two - component model , both the emission lines and the continuum features would be attenuated by both the birth cloud and the diffuse dust components , resulting in @xmath80 .
galaxies with lower ssfrs would have a smaller continuum contribution from massive stars , so more of the continuum light would come from stars only attenuated by the diffuse dust , resulting in @xmath0greater than @xmath1 .
these different cases are illustrated in figure [ fig : illustration ] .
we compare our results with those of @xcite for local galaxies .
they also find increasing amounts of extra attenuation with decreasing ssfrs , in agreement with the trend we observe .
@xcite find higher amounts of extra attenuation than we do , but the average ssfrs of their sample are lower than for our sample .
the two - component dust model naturally explains this difference , based on the dependence of extra attenuation on ssfr as discussed above .
this model may also explain the higher amount of extra attenuation found by @xcite if their sample has higher average ssfrs than our sample .
the two - component dust model could also explain the discrepancies found between different high - redshift studies , if the samples consist of galaxies with different ranges in specific sfr .
for example , the @xcite sfr - m@xmath81 relation implies a higher average ssfr than our sample , and they find a lower amount of extra attenuation .
@xcite find evidence for no extra attenuation , but the average ssfr is higher than for the @xcite sample .
our explanation for the trend between extra @xmath35and ssfr was previously mentioned by @xcite .
they suggest that the trend may alternatively be caused by a decline in diffuse dust with decreasing ssfr .
however , we observe that @xmath1increases slightly with decreasing ssfr , so we do not expect a decline in diffuse dust with decreasing ssfr . in absolute terms , we find that @xmath0increases with mass and sfr , and decreases with ssfr . as stellar mass and sfr
are correlated , the trend of increasing dustiness with sfr and mass could share the same cause .
the trend of increasing @xmath1and @xmath0with decreasing ssfr could also share the same cause , as the ssfr decreases slightly with increasing mass both in the local universe @xcite and at higher redshifts @xcite .
as the trends of @xmath1and @xmath0with increasing mass are the strongest , it is likely that mass is the key property .
this finding may be explained by the fact that more massive star - forming galaxies have higher gas - phase metallicities @xcite .
we do note that we assume a fixed attenuation law in our work .
recent work ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) shows that the dust attenuation curve varies with ssfr . as the origin of these observed trends
are not well understood , and this variation may be the consequence of age - dependent extinction in a two - component dust model , we have decided to use the same dust law for the derivation of the two dust measurements .
however , we can not rule out the possibility that variations in the dust attenuation law may have impacted the trends found in this work .
the two - component dust model also predicts a dependence of dust attenuation properties on the axial ratio . under this model ,
the amount of dust attenuation from the stellar birth clouds is similar in face - on or edge - on systems , while the longer path length in edge - on systems results in a larger overall @xmath1and a smaller amount of extra attenuation towards the star - forming regions .
@xcite find evidence for the two - component model from the trends of attenuation with axial ratio for objects in the local universe .
the spatially resolved wfc3 images yield excellent axial ratio measurements .
however , the axial ratio distribution for our sample is heavily biased towards face - on systems .
the more edge - on systems may be dustier , so our selection criteria likely introduce this bias against edge - on systems .
it might also be that our sample objects are not disk - like galaxies . because of sample incompleteness and the small range in axial ratios , we are unable to test the two - component model using the inclinations of the galaxies . vs. stellar mass comparison between this work and past studies . the filled data ( this work ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) indicate direct measurements of @xmath0using balmer decrements , while the open data indicate @xmath0measured with a different method that is calibrated using balmer decrements .
the dashed line gives the median relation derived by @xcite using sdss star - forming galaxies . with the exception of the single object from @xcite ( combined with information from
* ) , all other data are the result of stacks ( using various stacking schemes , both mean and median ) or are mean ( * ? ? ?
* combined with stacking ) or median values @xcite from samples .
, scaledwidth=48.0% ] a number of past studies have measured balmer decrements ( and often @xmath0 ) versus stellar mass @xcite .
other studies have employed different methods of measuring @xmath0versus stellar mass .
these methods include comparing @xmath82 to @xmath83 as by @xcite , or calibrating [ oii]/@xmath3ratio as an alternate to balmer decrements as by @xcite .
we compare our results with their findings in figure [ fig : avemis_mass ] . as necessary
, we convert stellar masses derived using a salpeter or kroupa imf to match our assumption of a chabrier imf using the relations given in equations 12 and 13 of @xcite .
the collection of results over a range of redshifts makes it tempting to speculate that there is no redshift evolution in the mass-@xmath0relation .
however , the current data are not conclusive .
first , there is incompleteness at all masses .
second , the measurements of @xmath0are not necessarily equivalent .
the samples of @xcite and @xcite are similar to ours in redshift and method ( stacking ) for measuring @xmath0 .
concerning the work by @xcite , we first note that their sample contains fewer objects than our sample .
second , we have a broad range of deep photometric coverage for our objects , which reduces the errors in our masses taken from sed modeling .
we also do not combine measurements between different grisms , to avoid possible normalization mismatches affecting line fluxes .
compared with @xcite , our sample is larger and on average has slightly lower ssfrs .
the trend we observe between @xmath0and ssfr shows that the average @xmath0decreases with increasing ssfr , which explains why their values of @xmath0are lower ( though consistent within the errors ) than what we measure for similar masses .
sfrs @xcite and the sed sfrs .
the two sets of sed sfrs differ only in the choice of the minimum star formation _
e_-folding timescale , @xmath84 : @xmath85 ( open red squares ) and @xmath86 ( closed black circles ) .
( the dashed black line shows equal @xmath3sfrs and sed sfrs . )
the @xmath3sfrs agree better with the sed sfrs calculated with the higher @xmath87 , which is similar to previous findings that setting @xmath88 yields the most reasonable sed fits for star - forming galaxies .
( see section [ sec : sfrha_sfrsed ] for a full discussion . )
the errors in log sed sfr are the @xmath89 scatter within the bins . , scaledwidth=48.0% ] with our measurements of dust attenuation towards hii regions , we can for the first time directly measure @xmath3sfrs for a large sample of @xmath23 galaxies .
we calculate the @xmath3sfrs using the relation from @xcite converted to a @xcite imf .
@xmath3fluxes are taken from the stacks in sed sfr , and are converted to physical units by comparing the flux of the stack with the weighted average ( see section [ sec : line_measurement ] ) of the individual objects @xmath3fluxes , each corrected for [ nii ] . the luminosity distance is taken as the weighted average of the individual objects luminosity distances .
we then use the @xmath3sfrs to test the much more debated sed sfrs .
the comparison of sfr indicators is shown in figure [ fig : sfr_comp ] .
the @xmath3sfrs are compared to the weighted average of sed sfrs derived assuming an exponentially decaying sfh and a minimum _ e_-folding time of @xmath34 ( shown as filled black circles ) . for these parameters ,
the sed sfr values underestimate the @xmath3sfrs , and are inconsistent at the @xmath90 level . however , if a shorter decay time of @xmath91 ( open red squares ) is adopted , the sed sfrs are low and inconsistent with the @xmath3sfrs at the @xmath92 level .
thus for our sample , sed sfrs derived with @xmath34 are more in line with the @xmath3sfrs .
this is in general agreement with the work of @xcite .
they compare uv+ir and sed sfrs and find good agreement between the sfr indicators for @xmath34 , but when a shorter time of @xmath91 is adopted , they find that the sed sfrs underestimate the true sfrs .
@xcite also compare uv+ir sfrs with sed sfrs .
however , instead of using a longer @xmath93for a decreasing sfh , they find the sed sfrs agree with uv+ir sfrs when increasing sfhs are adopted .
the sfr indicators may also differ because they probe different stellar mass ranges , and thus different star formation timescales .
the @xmath3sfrs depend on ob stars , so are averaged over @xmath94 , while sed sfrs are limited by the discrete nature of sed fitting and depend on the uv flux from o , b and a stars , which live longer .
these timescale differences could cause discrepancies for galaxies with episodic or rapidly increasing or decreasing sfhs .
however , as we stack multiple objects and thus average over many sfhs , we expect that the timescale differences should not significantly influence the measured sfrs . as discussed in section [ sec : sample_selection ] , we use a number of methods to reject agns from our sample .
however , as almost all individual objects do not have sufficient line snrs , and more importantly we do not have separate [ nii ] and @xmath3measurements , we can not use a bpt @xcite diagram to distinguish whether the line emission originates from star formation or agn . after stacking ,
there is sufficient line snr to place the binned values in the bpt and alternative diagrams , which we show in figure [ fig : bpt ] .
these diagrams allow us to assess if there is agn contamination in the stacks .
we use the data from the stacks in stellar mass in these diagrams . in the left panel ,
we show the traditional bpt diagram : [ nii]/@xmath3vs .
[ oiii]/@xmath16 .
included are the theoretical limit for star - forming ( sf ) galaxies from @xcite ( dashed line ) and the more conservative empirical division between sf galaxies and agn by @xcite ( dashed - dot line ) , both of which are derived for galaxies at @xmath95 .
the sdss dr7 galaxies @xcite are divided into three categories based on these dividing lines : sf galaxies ( blue contours ) , composite ( green contours ) , and agn ( red contours ) .
these same group definitions are used in the alternate bpt diagrams ( middle and right of figure [ fig : bpt ] ) . because we do not measure [ nii ] directly ( instead inferring a value as described in section [ sec : nii_correction ] ) , we can not use the traditional bpt diagram as a post - analysis check .
instead , we show that our inferred [ nii]/@xmath3ratios and the observed [ oiii]/@xmath16ratios , shown as open grey circles , are consistent with values for sf galaxies .
however , in agreement with other observational ( e.g. , * ? ? ?
* ; * ? ? ?
* ) and theoretical studies @xcite of distant galaxies , the data points are moved slightly up ( and possibly to the right ) in the bpt diagram compared to the average values found in the local star - forming sequence .
an alternate bpt diagram is [ sii]/@xmath3vs .
[ oiii]/@xmath16 .
however , @xmath3is not a directly measured quantity , so using the deblended @xmath3 values makes it impossible to detangle the assumptions of the [ nii ] correction with the possible presence of agn .
instead , we make an alternate diagram with only directly measured quantities : [ sii]/(h@xmath15+[nii ] ) vs. [ oiii]/@xmath16 , which we show in the middle panel of figure [ fig : bpt ] .
there is significant overlap between the different categories of sdss galaxies , so we are unable to discriminate between sf galaxies and agn using combinations of blended emission lines . a final alternate to the bpt diagram , the mass - excitation ( mex ) diagram @xcite , is useful as we directly measure the necessary lines ( [ oiii ] and @xmath16 ) and derive stellar mass with sed fitting , which we show in the right panel of figure [ fig : bpt ] .
the solid black lines are their empirical divisions between sf , composite , and agn , which are valid up to @xmath96 .
we include the sdss categories , using stellar masses from the mpa / jhu value - added catalogs @xcite , to demonstrate that while the separation is not as clean as the traditional bpt diagram , it does separate sf galaxies from agn in the local universe .
our data are consistent with the sf region as defined by @xcite , though the data do lie on the division boundary .
however , work by @xcite at @xmath55 has shown that the low redshift empirical divisions in the mex diagram incorrectly classify @xmath55 sf galaxies as agn .
they use the shift in the mass - metallicity relation between @xmath95 and 2 to derive a shift to lower mass to bring the high-@xmath17 values into agreement with the @xmath95 mex diagram ( see fig .
14 of * ? ? ?
this empirical correction ( shown as open grey circles in the right panel of figure [ fig : bpt ] ) shifts our data securely into the sf region of the diagram .
these diagnostics demonstrate that there is likely not much agn contamination to the line emission in our sample .
one of the largest sources of uncertainty in this analysis is our correction for [ nii ] in our measurement of @xmath3from the blended h@xmath15+[nii ] line .
figure [ fig : avplot ] demonstrates how much our results change when we do not correct for [ nii ] .
we use the relation between [ nii]/@xmath3and stellar mass by @xcite as a way of detangling the blended lines .
we infer the [ nii ] contribution from the weighted average mass of each stack .
because the two samples have similar masses and sfrs , and are close in redshift , the average sample properties should be fairly similar .
still , the use of this relation may introduce bias into our results .
however , our gas phase metallicities may be different from those of @xcite , due to slight differences in redshift , mass or ( s)sfr , which could affect both the slope and scaling of the [ nii]/@xmath3relation .
such trends have been proposed by @xcite , in the form of the fundamental metallicity relation ( fmr ) .
the fmr suggests that metallicity depends on both stellar mass and sfr .
we repeat our analysis using the fmr in conjunction with the @xcite [ nii]/@xmath3metallicity calibration as an alternative way of calculating the [ nii ] correction .
this allows us to test whether the trend of @xmath73versus ssfr may be due to a variation in [ nii]/@xmath3with sfr .
if the fmr is adopted , the @xmath0values are slightly lower than the @xmath0values derived using the @xcite values , which reduces the strength of the trend of @xmath73 with ssfr . in this case , the relation we observe between @xmath73and ssfr is weaker than the trend presented in section [ sec : ssfrsfrmass ] , and is consistent with no trend , as the difference is only @xmath97 .
however , the data are still suggestive of a possible decreasing trend of @xmath73with ssfr , which would imply the signal is not entirely caused by trends with metallicity .
it is also possible that there is some agn contamination at the highest masses @xcite , which would also result in an underestimate of the [ nii ] fraction . in most cases ,
our corrected @xmath3flux would be higher than the true value , leading to an overestimate of @xmath0 .
we use the mass-[nii]/@xmath3relation presented in @xcite for our analysis over the fmr due to uncertainty about metallicity relations at high redshifts .
first , the fmr is still not well tested at these redshifts @xcite .
second , recent work has also questioned whether [ nii]/@xmath3correlates with metallicity at these redshifts @xcite .
@xcite present direct measurements of [ nii]/@xmath3 , which allows us to avoid systematic problems with metallicity calibrations .
one of the strengths of our analysis is that we draw a sample from a non - targeted grism survey , with sample cuts designed to avoid bias as much as possible .
however , bias and incompleteness most likely still affect our sample .
the dustiest star - forming galaxies have very attenuated @xmath3fluxes .
our @xmath3snr selection cut , designed to avoid adding noise to our analysis , introduces bias against galaxies with large @xmath0 .
this bias affects the high mass , high sfr end of the sample , as these objects have the largest @xmath35 . as lower mass ,
lower sfr galaxies tend to have lower @xmath3luminosities , the @xmath3snr cut will also exclude objects with the largest @xmath35values in that mass range .
our continuum normalization scheme also introduces bias into our analysis .
we adopt a normalization scheme to improve the signal of our stack , but the cost is that some objects have much higher scaled @xmath3fluxes , and thus they dominate our stacks .
this biases our results towards those objects with higher @xmath3equivalent widths in a given bin .
because the line measurements are biased based on the @xmath3flux , we take the weighted average of the continuum values within a bin to ensure a fair comparison .
however , if we instead used a non - weighted average for @xmath98 , the @xmath1vs .
@xmath0relationship does not change much .
in this paper , we investigate dust attenuation in @xmath23 star - forming galaxies using data from the 3d - hst survey .
we measure both the dust towards hii regions , using balmer decrements , and the integrated dust properties , using sed modeling .
we find that there is extra attenuation towards star - forming regions . on average the total attenuation of these regions , @xmath0 ,
is 1.86 times the integrated dust attenuation , @xmath1 .
however , the amount of extra attenuation is not the same for all galaxies .
we find that the amount of extra attenuation decreases with increasing ssfr , in agreement with the results by @xcite for low - redshift galaxies .
our findings are consistent with the two - component dust model , which assumes there is a diffuse dust component in the ism and a dust component associated with the short - lived stellar birth clouds . for galaxies with high ssfr , the stellar light will be dominated by continuum emission from the younger stellar population in the birth clouds , resulting in similar attenuation toward the line and continuum emission . for more evolved galaxies , much of the stellar light
will only be attenuated by the diffuse ism , leading to larger discrepancies between the two dust measures .
the observed trend of @xmath73with ssfr may be affected by uncertainties in the [ nii ] correction and possible dust attenuation law variations .
future work is necessary to determine what role these effects have on the relation between @xmath73and ssfr .
similar to previous studies @xcite , we find less extra attenuation in distant galaxies than is found in the local universe @xcite .
this effect can also be explained by the two - component model , as lower redshift objects tend to have lower ssfrs than higher redshift galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
we find that both @xmath0and @xmath1increase with increasing sfr and stellar mass , and decreasing specific sfrs .
however , our data are biased against the dustiest objects , which may affect these trends .
we also observe there to be little redshift evolution in the @xmath0-@xmath99 relation , although uncertainties and incompleteness makes it impossible to make a definite claim .
using the balmer decrements , we calculate dust - corrected @xmath3sfrs to test the accuracy of sfrs derived from sed fitting .
we find better agreement between the sfr indicators if short sfh decay times are not allowed and the constraint @xmath100 is used .
this generally agrees with the results of past studies comparing uv+ir and sed sfrs @xcite or @xmath3and sed sfrs @xcite .
however , even with this constraint the sed sfrs slightly underestimate the @xmath3sfrs .
we note that our results are slightly impacted by incompleteness and systematic uncertainties .
first , we employ snr cuts on @xmath3to ensure quality data , which likely results in incompleteness of the dustiest galaxies .
second , to obtain significant @xmath16detections , we stack spectra , and thus our normalization scheme or incorrectly measured integrated properties could impact our measurements .
most of these issues can be overcome with future observations by a number of new multi - object near infrared spectrographs on 8 - 10 m class telescopes , among which is mosfire on keck @xcite .
these instruments have higher spectral resolutions , which will avoid blended lines .
they will also allow for deeper measurements , which will yield more accurate balmer decrements of individual objects as well as allowing for investigation of dust to higher @xmath35limits . additionally , including rest - frame mid- and far - ir data in future work
will ensure more accurate values of @xmath1 .
better measurements of @xmath0and @xmath1are necessary to better constrain the geometric distribution of dust in high redshift galaxies and the effects of the dust - to - star geometry on dust attenuation .
we thank edward taylor and nick hand for useful discussions , and david sobral and daichi kashino for sharing data for comparison .
we thank the anonymous referee for constructive comments , which have improved this paper .
this work is based on observations taken by the 3d - hst treasury program ( go 12177 and 12328 ) with the nasa / esa hst , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
we acknowledge support from stsci grant 12117.21-a .
sp is funded by the national science foundation graduate research fellowship under grant no .
dge 1106400 .
in addition to the stacked spectra ( presented in fig . [
fig : spec ] ) , we also present the average photometry for each bin ( see fig .
[ fig : phot ] ) .
the photometry covers a much greater wavelength range , so the average photometry allows us to examine the average sed shape outside the limited range covered by the grism spectra .
the photometry for each object is first normalized to match the grism normalization ( see section [ sec : stacking ] ) , and the central wavelength of each filter is de - redshifted using the object s grism redshift .
the data are averaged within each bin , using the normalized fluxes and rest - frame wavelengths .
if not all objects in the bin have coverage in a given filter , nearby filters are combined so each averaged photometric point contains at least the same number of measurements as number of objects in the bin , provided the photometry are not too widely separated in wavelength . for comparison with the average photometry
, we also show the average best - fit continuum model and the errors on the best - fit model .
plotting the continuum model and error allows us to examine the error in the amount of balmer absorption .
the average best - fit stellar population model and simulated best - fit models are calculated as described in section [ sec : stacking ] .
the @xmath101 continuum errors are estimated at every wavelength using the simulated continuum models .
the perturbations of the photometry of the individual objects do not lead to large variation in the stacked best - fit models , especially near the balmer lines .
the broad wavelength coverage of the photometry , specifically across the balmer break , provides reasonably tight constraints on the average amount of balmer absorption for the stacks . | the nature of dust in distant galaxies is not well understood , and until recently few direct dust measurements have been possible .
we investigate dust in distant star - forming galaxies using near - infrared grism spectra of the 3d - hst survey combined with archival multi - wavelength photometry .
these data allow us to make a direct comparison between dust around star - forming regions ( @xmath0 ) and the integrated dust content ( @xmath1 ) .
we select a sample of 163 galaxies between @xmath2 with @xmath3signal - to - noise ratio @xmath4 and measure balmer decrements from stacked spectra to calculate @xmath0 .
first , we stack spectra in bins of @xmath1 , and find that @xmath5 , with a significance of @xmath6 .
our result is consistent with the two - component dust model , in which galaxies contain both diffuse and stellar birth cloud dust .
next , we stack spectra in bins of specific star formation rate ( @xmath7 ) , star formation rate ( @xmath8 ) , and stellar mass ( @xmath9 ) .
we find that on average @xmath0increases with sfr and mass , but decreases with increasing ssfr .
interestingly , the data hint that the amount of extra attenuation decreases with increasing ssfr .
this trend is expected from the two - component model , as the extra attenuation will increase once older stars outside the star - forming regions become more dominant in the galaxy spectrum .
finally , using balmer decrements we derive dust - corrected @xmath3sfrs , and find that stellar population modeling produces incorrect sfrs if rapidly declining star formation histories are included in the explored parameter space . |
the prediction of a relativistically invariant quantum field theory in 2 + 1 dimensions as low energy limit of a solid state system @xcite was confirmed by investigation of graphene @xcite .
modification of the graphene structure may allow to obtain more general quantum field theories @xcite .
since solid state systems typically have edges , the study of quantum field theories with boundaries acquired additional interest .
an important tool in this context is the bulk - boundary ope . in particular , some of its singularities yield divergent observables and constitute obstructions to any experimental realization .
the coefficient function of the identity in such an ope is just the corresponding vacuum expectation value in the presence of the boundary .
integrated expectation values of current operators can be interpreted as measurable charge or spin currents and should remain finite in realistic systems . here
we study this obstruction in the simplest case of free fermions , where vacuum expectation values can be interpreted as integrated contributions of the particles in the dirac sea .
it is of course well known that such effects can prevent the lifting of a quantum mechanical system to a quantum field theory , even without the presence of a boundary . in our case the dirac sea of a single fermion yields a half - integral value of the hall conductivity @xcite , @xcite , @xcite .
consequently one needs an even number of fermions , though some of them may be shifted to infinite mass , which results in a chern - simons term instead . for boundary states
we shall find more intricate obstructions , which depend on the numerical parameters of the boundary state .
we consider spinor fields @xmath0 acted upon by the dirac operator @xmath1 on the open half plane @xmath2 , as in @xcite .
this operator is a closable symmetric operator on the domain @xmath3 of smooth functions with compact support vanishing in a neighbourhood of @xmath4 .
the possible boundary conditions are classified by its self - adjoint extensions .
the latter can be obtained by the standard von neumann theory . here
we give a modification , which is less general but better adapted to the present problem .
since @xmath5 is dense in the hilbert space @xmath6 , the image of an arbitrary vector @xmath7 in the domain of a self - adjoint extension is uniquely determined by the unextended operator , so the problem is reduced to the determination of such domains .
let @xmath8 be the adjoint of the unextended @xmath9 .
by von neumann theory , every self - adjoint extension of @xmath9 corresponds 1 - 1 to an isometry @xmath10 between the defect spaces @xmath11 where @xmath12 is a fixed but arbitrary positive number .
indeed , @xmath13 gives rise to the self - adjoint extension @xmath14 with domain @xmath15 the choice @xmath16 is standard , but we shall see that the limit of large @xmath12 allows a more direct physical interpretation . applied to ( [ dirac - op ] ) , @xmath17 is given by @xmath18 integrals over elements of the one - dimensional spaces @xmath19 where @xmath20 for @xmath21 the defect spaces consist of wave functions which become supported on an arbitrarily small neighborhood of the boundary and factorize in a universal term @xmath22 and a spinor of the form @xmath23 thus isometries @xmath24 become tensor products of a unitary map @xmath25 with the map between one dimensional subspaces of @xmath26 which takes @xmath27 to @xmath28 . from ( [ extended domain ] ) we read off that any @xmath29 satisfies @xmath30 if @xmath31 is not an eigenvalue of @xmath13 , then @xmath32 where @xmath33 is the self - adjoint operator @xmath34 the physical necessity of such a boundary condition can be understood as follows .
consider the conserved current densities @xmath35 one needs a boundary condition which implies @xmath36 , and this indeed follows from ( [ b.c.g ] ) .
we only study local boundary conditions , for which @xmath37 vanishes identically and also impose translational invariance .
thus @xmath33 has to commute with @xmath38 and with multiplication by functions of @xmath39 .
this implies that it is a real constant @xmath40 , and we obtain the boundary condition @xmath41 which we shall use in the sequel . to include the case @xmath42
, we consider @xmath40 as an element of the projective real line , including the value @xmath43 . for any @xmath40
the hamiltonian commutes with the pt transformation @xmath44 .
the space reflection @xmath45 lifts by @xmath46 to a map between systems with parameters @xmath47 and @xmath48 .
because of this duality we only need to consider positive @xmath49 .
arbitrary values of @xmath49 will only be considered at the very end of the calculation .
it turns out that the results allow a continuous extension to @xmath50 , which is the case relevant for graphene .
the cpt map @xmath51 relates the positive energy states for @xmath47 to the negative energy states for @xmath52 .
a different type of duality is given by @xmath53 , @xmath54 , which relates boundary states of complementary half planes .
the dirac operator commutes with @xmath55 and consequently with @xmath56 we denote the real eigenvalues of these hermitean operators by @xmath18 and @xmath57 .
we take @xmath58 for positive @xmath57 and @xmath59 with @xmath60 for negative @xmath57 .
the corresponding eigenvalues @xmath61 of @xmath9 satisfy @xmath62 . according to the sign of @xmath57 the spectrum of @xmath9 has two parts , which will be refered to as bulk and edge .
bulk eigenfunctions @xmath63 with given @xmath64 have the form @xmath65 with constant spinors @xmath66 . the equation @xmath67 and
the boundary condition reduce to @xmath68 and @xmath69 . for each @xmath70
there is a unique solution up to normalization , such that for the bulk part we just recover the spectrum of the dirac operator on the whole plane , with a gap @xmath71 .
we normalize these eigenfunctions with respect to the measure @xmath72 and choose phases such that @xmath73 where @xmath74 with @xmath75 edge eigenfunctions @xcite have the form @xmath76 .
we normalize them with respect to the measure @xmath77 and choose phases such that @xmath78 the equation @xmath79 yields @xmath80 and @xmath81 thus @xmath82 are one - valued functions of @xmath18 , and for any @xmath18 one has a unique edge state whenever @xmath83 and no edge state otherwise .
the latter inequality restricts the possible values of @xmath84 .
the limiting values for @xmath85 lie on the mass shell .
when we put @xmath86 then positive energy half of the mass shell is touch for @xmath87 and the negative energy half for @xmath88 .
one finds energies within the spectral gap @xmath89 of the bulk iff @xmath90 .
the linear dispersion relation for the boundary equation implies that edge states travel at the fixed velocity @xmath91 note that @xmath92 determines @xmath40 up to the replacement of @xmath40 by @xmath93 , which is the one induced by cpt .
the latter transformation changes the signature @xmath94 .
altogether , the boundary condition can be characterized by @xmath92 and @xmath94 , except for the fact that @xmath94 is undefined in the limiting cases @xmath95 .
one observes that 1 . for vanishing edge velocity , i.e. for @xmath96 and @xmath43 , edge states have @xmath97 and @xmath98 , respectively , independently of @xmath18 .
this case is probably unphysical , since all edge states have the same energy and the thermodynamic partition function diverges for a system of infinite width . since the width of the edge excitation increases linearly with @xmath18 , the thermodynamic properties of a system of a finite system would depend on the sample size .
the edge velocity is equal to the maximal bulk velocity for the cpt invariant boundary conditions @xmath95 . these yield @xmath99 and @xmath100 independently of @xmath18 . for @xmath101
there are no edge states , for @xmath102 all edge states have the same dependence on @xmath103 .
this yields particular divergences , so we will tacitly exclude @xmath95 in most of the subsequent calculations .
in @xmath104-space the half - line of edge states is tangent to the half - hyperboloid given by @xmath105 and @xmath106 at @xmath107 . for @xmath108
this happens at positive energy and for @xmath109 at negative energy .
the tangency has consequences for the behaviour at large distance from the boundary , as we shall see later .
the portion of edge energies within @xmath89 gives rise to an edge current parallel to the boundary , with edge - conductivity equal to ( in units of @xmath110 ) @xcite @xmath111 this applies whenever some subinterval of the gap is actually occupied , independently of the size of this subinterval .
when no boundary exists , fermi energies within the gap @xmath89 obviously yield the same theory .
this means that any pair @xmath112 with @xmath113 yields a boundary state of the standard lorentz invariant free fermion theory .
we want to calculate the corresponding expectation values of the current densities .
the bulk contribution does not depend on the choice of @xmath114 and edge contributions for energies within @xmath89 are finite and were calculated in @xcite .
thus it suffices to consider @xmath115 .
the required vacuum expectation values are @xmath116 where @xmath117 and @xmath118 here the variable @xmath61 in @xmath63 is the negative roots of @xmath119 .
note that @xmath120 and @xmath121 vanish identically . on the other hand eqs .
( [ phi ] , [ rho ] ) yield @xmath122 where @xmath123 the first term on the r.h.s . of eq .
( [ j2-bulk ] ) is an odd function of @xmath18 , so that its contribution to the vacuum expectation value of @xmath124 vanishes for any symmetric regularization . for the second term
we also will find a partial cancelation .
we substitute @xmath125 by @xmath126 , which yields for negative @xmath61 @xmath127 moreover @xmath128 here @xmath129 where @xmath130 and @xmath131 the two denominators @xmath132 and @xmath133 are formally related by a reflection across the boundary , with @xmath54 .
one obtains a partial fraction expansion @xmath134 with @xmath135 to regularize the momentum integrals we have to introduce a cutoff . in a solid state context
the underlying physics selects a stationary lorentz frame , in which a symmetric cutoff in @xmath18 is natural . for this cutoff
the @xmath136 integrals of these two terms cancel against each other due to the symmetry @xmath137 , @xmath138 .
the remaining terms are @xmath139 for a symmetric cutoff @xmath140 of the @xmath141-integration one finds @xmath142 to calculate the @xmath143 integral we need @xmath144 where the principal value of the logarithm has to be taken , so that @xmath145 to calculate the @xmath146 integral over the logarithmic part we use @xmath147 the argument of @xmath148 has branching points at @xmath149 .
we may move the integration path towards the cut with branching point @xmath150 .
since the discontinuity of the logarithm is @xmath151 , this yields an elementary integral .
altogether we obtain @xmath152 for the contribution from the edge functions , we have according to ( [ lambda ] ) @xmath153 we have to integrate over those @xmath154 for which @xmath83 and @xmath155 .
this yields @xmath156.\ ] ] note that @xmath157 vanishes for @xmath158 and is non - singular for @xmath159 . for @xmath109
both the bulk and the edge current densities decay algebraically at large distance from the boundary . in their sum
these terms cancel , an interesting effect which reflects the tangency of the edge state and bulk state manifolds in @xmath104 space , which was discussed in the previous section . in all cases
the singular part of the vacuum expectation value has the form @xmath160 remarkably , this equation is invariant under @xmath39 reflection together with @xmath161 , such that it remains true for @xmath162 and is in fact independent of @xmath49 .
the regular part of the expectation value is continuous in @xmath49 and vanishes for @xmath50 . for infinite @xmath49
the contribution becomes localized at the boundary and is given by the @xmath163 terms in ( [ j2bulk ] ) .
it may be interesting to rederive this result from a chern - simons lagrangian .
a single free fermion yields a divergent edge current on a half plane . for @xmath164 fermions with boundary conditions
@xmath165 the cut - off dependent edge current vanishes for @xmath166 one still has a dipolar edge current @xmath167 which in general does not vanish .
in addition there is a @xmath168 singularity , which yields an unphysical divergent particle transport in the neighborhood of the edge , unless @xmath169 since the bulk system and the boundary at @xmath4 are lorentz invariant under boosts in the @xmath39 direction , one should be able to act with lorentz transformations on the boundary conditions . in the context of relativistc quantum mechanics
this is straightforward .
the action on the parameters @xmath165 can be determined from eq .
( 3 ) or more easily from eqs .
( 8 , 9 ) , according to which boundary states are characterized by a lorentz invariant signature @xmath170 and their fixed travel velocity @xmath171 this translates into a rapidity @xmath172 given by @xmath173 , or equivalently by @xmath174 to which the rapidity of the lorentz boost is added .
when one moves from one - particle states to field theory , control of divergencies requires a cutoff which breaks lorentz invariance .
it may or may not be restored , when this cutoff is taken to infinity .
one would expect that a restoration of loretz invariance implies cpt invariance . due to the unbroken pt
invariance this should yield invariance under charge conjugation and consequently an absence of net boundary currents .
in our case we will find an unexpectedly subtlety , however . in terms of edge velocities and signatures
the two divergency cancellation conditions can be rewritten as @xmath175 where @xmath176 for @xmath177 there is no solution , in agreement with the restriction to an even @xmath164 coming from the bulk physics . for @xmath178
one needs @xmath179 , thus invariance under charge conjugation .
more generally , pairs of edge states related by charge conjegation do not contribute to the breaking of lorentz invariance or to net boundary currents , as expected .
there is an unexpected remainder of lorentz invariance in more general situation , however . when all @xmath180 of the remaining states have the same sign , the cancellation conditions are invariant under sufficiently small boosts which do not change the sign of edge velocities .
when @xmath180 of different signs occurs , no aspect of lorentz invariance is recovered when the cutoff is removed . in any case , breaking of lorentz invariance by edge states does not seem to be an argument against the realization of such systems in solid state physics , where this invariance is only a low energy phenomenon
. it will be interesting to see to what extent the @xmath40 values can be controled experimentally .
systems without time reversal invariance would be particularly interesting , but are certainly very difficult to realize . | we calculate the ground state current densities for 2 + 1 dimensional free fermion theories with local , translationally invariant boundary states .
deformations of the bulk wave functions close to the edge and boundary states both may cause edge current divergencies , which have to cancel in realistic systems .
this yields restrictions on the parameters of quantum field theories which can arise as low energy limits of solid state systems .
some degree of lorentz invariance for boosts parallel to the boundary can be recovered , when the cutoff is removed . |
low lying scalar mesons are an old problem in hadron physics , see the review by spanier and trnqvist on scalar mesons in pdg00 and the literature quoted there . in a recent analysis @xcite of @xmath0- and @xmath1-meson
nonleptonic decays distinct signals for strong enhancements in @xmath2-wave @xmath3 and @xmath4 channels have been observed , reviving the interest for these states . in the following we shall refer to the signal in the @xmath3 channel as @xmath5 and in the @xmath4 chanel as @xmath6 .
the enhancements can be well described by a breit - wigner - type resonance form , with the correct threshold behaviour _ x , bw(s ) = , where the subscript @xmath7 stands for @xmath5 or @xmath6 .
the correct threshold behaviour is guaranteed through the @xmath8-dependent width _
x(s ) = _ 0x . here
@xmath9 and @xmath10 are the masses of the mesons in the decay channel and @xmath11 . in @xcite
the following parameters were found : + 1 ) for the @xmath2-wave @xmath12 channel , @xmath13 , m_= 0.478 0.024 0.017 ; _ 0 = 0.3240.0420.021 .
2 ) for the @xmath2-wave @xmath14 channel , @xmath15 , m_= 0.797 0.019 0.042 ; _ 0 = 0.410 0.043 0.085 .
a theoretical analysis gnpt00 of the non - leptonic decay @xmath16 is in agrement with the experimental data .
the semileptonic @xmath0-decays offer , in principle , much cleaner samples than the nonleptonic decays since there occur no problems connected with the presence of a third strongly interacting particle . in the non - strange channel important experimental information comes indeed from the analysis of the final state interaction of the pions from leptonic @xmath17-decays . for the case of strange mesons , the analysis of @xmath0 decays can play a similarly important role . in a recent paper by the focus collaboration lin02 clear evidence
was found that in the semileptonic decay @xmath18 the @xmath0-meson does not decay exclusively into the hadronic vector channel , but that there is interference with a scalar contribution . in this paper
we estimate the decay rates of semileptonic @xmath0-decays into the scalar @xmath4 and @xmath3 channels .
we use the method of qcd sum rules @xcite which has been successfully applied to several semileptonic decay processes . for a recent review see @xcite and the literature quoted there . even if ( broad ) resonances in the hadronic decay channels exist , it is not clear whether they are due to interactions on the quark level , or if they are rather an effect of interactions in the purely hadronic channel , see for instance jop00 . here
a theoretical analysis can be very helpful .
the semileptonic decay is supposed to occur on the quark level and therefore the decay rate should depend crucially on the direct coupling of the resonances to the quark currents .
we consider two limiting cases ( 8,5 ) ( 2,0)(1,1)2 ( 4,2)(1,-1)2 ( 2,0)(1,0)4 ( 8,0)(2,1)2 ( 8,0)(2,-1)2 ( 0,0)(1,0)2 ( 4,2)(0,1)2 ( 6,0)(1,0)2 ( 1,0.2)@xmath0 ( 7,0.2)@xmath7 ( 4.2,3)@xmath19 ( 4,0.2)@xmath20 ( 9,0.8)@xmath21 ( 9,-0.4)@xmath22 ( 2.5,1)@xmath23 ( 5.2,1)@xmath24 * the observed two - meson final state couples to the corresponding quark - antiquark state only through the @xmath25 or @xmath26 resonance .
* the quark antiquark state couples to uncorrelated final mesons in an @xmath2-wave .
since the signals we investigate in this paper are very broad the finite width has to be taken into account .
otherwise our analysis is based on the same principles and assumptions as the used in the sum - rule analysis of other semileptonic @xmath0-decays @xcite .
we therefore refer to this paper for details and clarifications .
we investigate the semileptonic decays dx | _ , where @xmath7 might be the @xmath27 or an uncorrelated @xmath3 or @xmath14 pair in the @xmath2-state , as depicted in fig . graph .
the semileptonic decay of a @xmath0-meson with momentum @xmath28 into a scalar state with total momentum @xmath29 and invariant mass @xmath30 is described by the two form factors of the matrix element of the weak current x |j_| d=(p_d+p_x)_f_+x(t ) + ( p_d - p_x)_f_-x(t ) , where @xmath31 . in the decay rate
the form factor @xmath32 is multiplied by the difference of the lepton masses and hence is negligible for both @xmath33 and @xmath34 decays .
the differential semileptonic decay rate is given by = ^3/2(m_d^2,s , t)f_+x^2(t ) , where @xmath35 is the fermi coupling constant and @xmath36 the ckm transition element from the charmed quark to the quark @xmath24 .
the total width is = _ ( m_a+m_b)^2^m_d^2ds_0^(m_d-)^2 dt . the spectral distribution in the invariant mass @xmath37 of the hadronic final state is given by = 2 _ 0^(m_d-)^2 dt .
the @xmath0 meson in the initial state is interpolated by the pseudoscalar current j_d(x ) = |c(x)i_5 q_1(x ) , where @xmath23 is the field of the charmed quark and @xmath20 that of an up or down quark , summation over spinor and colour indices being understood but not indicated explicitly .
the final hadronic state @xmath7 is interpolated by the scalar current j_x(x ) = |q_1(x ) q_2(x ) , where @xmath24 represents a light quark field for @xmath38 , and a strange quark field for @xmath39 .
the semileptonic decay rate is obtained from the time ordered product of the two interpolating fields in eqs .
( intd ) and ( intscal ) and the weak current @xmath40 t_x(p_d^2,p_x^2,t ) = i^2d^4x d^4y 0|t [ j_d(x ) j^w_(0 ) j_x(y)]|0e^i(p_d.x - p_x.y ) . in order to select the semileptonic decay rates into the lowest lying hadrons we insert intermediate states and obtain the following double dispersion relation in the phenomenological side t_x^phen ( p_d^2,p_x^2,t)&= & ds_dds_x 0 | j_d(0)|dd|j^w_(0)|xx|j_x(0)|0 + & & . introducing 0 | j_d(0)|d&= & a_d _ d(s_d ) ; p_d^2=s_d , + x|j_x(0)|0&=&a_x _
x(s_x ) ; p_x^2=s_x , and using eq .
( ff ) we obtain t_x^phen(p_d^2,p_x^2,t)&= & ds_dds_x a_d a_x + & & ( f_+x(t , s_d , s_x)(p_d+p_x)_+f_-x(t , s_d , s_x ) ( p_d - p_x ) _ ) + & & + . in the following we concentrate on the relevant form factor @xmath41 and introduce the factor @xmath42 which multiplies the vector @xmath43 t_+x^phen(p_d^2,p_x^2,t)&= & ds_dds_x a_d a_x f_+x(t , s_d , s_x ) + & & + . for the @xmath0-meson
the density @xmath44 introduced in eq .
( constd ) is given by _
d(s_d)= ( s_d - m_d^2 ) , and we obtain a_d = f_d m_d^2/m_c , @xmath45 being the @xmath0-meson decay constant in the conventional notation . for the density @xmath46 we use the two extreme anstze mentioned above : * the quark - antiquark current @xmath47 in the @xmath48 state couples to the @xmath25 or @xmath26-resonance described by the breit - wigner distribution ( bw ) _
x(s)=_x , bw(s ) . * the quark - antiquark current @xmath47 couples to an uncorrelated meson pair , the density being described by the density of two - particle phase space _
x(s ) = , where @xmath9 , @xmath10 are the masses of the mesons in the final state
. the three - point function can be evaluated by perturbative qcd if the external momenta are in the deep euclidean region p_d^2(m_c+m_1)^2 , p_x^2 ( m_1+m_2)^2 , t ( m_c+m_2)^2 . in order to approach the not - so - deep - euclidean region and to get more information on the nearest physical singularities ,
nonperturbative power corrections are added to the perturbative contribution t^theor_+x(p_d^2,p_x^2,t)= 1 ^ 2_s_d , thds_d _ s_x , th ds_x + _ ijo_ij . the perturbative contribution is contained in the double spectral function @xmath49 .
the wilson coefficients @xmath50 multiplying the power corrections can be evaluated in perturbative qcd .
the operators @xmath51 occur in the operator expansion of the time ordered product eq .
( 3point ) ; their vacuum expectation values , the condensates @xmath52 , are introduced as phenomenological parameters . in order to suppress the condensates of higher dimension and at the same time reduce the influence of higher resonances , the series in eq .
( power ) is borel improved , leading to the mapping f(p^2 )
f(m^2 ) , . furthermore
, we make the usual assumption that the contributions of higher resonances are well approximated by the perturbative expression _
s_0d^ds_d_s_0x^ds_x , with appropriate continuum thresholds @xmath53 and @xmath54 . by equating the borel transforms of the phenomenological expression in eq.(3point3 ) and that of the `` theoretical expression '' , eq .
( power ) , we obtain the sum rule + & & = _ s_d , th^s_0d ds_d_s_x , th^s_0x ds_x _ + x(s_d , s_x , t)e^-s_d^2/m_d^2e^-s_x / m_x^2 + & & + k_+x(m_d^2,m_x^2,t ) + k_+i(m_d^2,m_x^2,t ) , where @xmath55 is some value below @xmath54 . in the zero width approximation we have of course @xmath56 .
@xmath57 are the borel transforms of the nonperturbative expressions due to the condensates and @xmath58 is an approximation to the instanton contributions , which might be important in the scalar channel @xcite . the decay constant @xmath45 and the coupling @xmath59 defined in eqs .
( constd ) , ( const ) , and(af ) can also be determined by sum rules obtained from the appropriate two - point functions . using the same procedure as described above we arrive at ( f^theor_d(m^2))^2 e^-m_d^2/m^2= _ m_c^2^s_0d ds_d(s ) e^-s / m^2 + k_d(m^2 ) , and + & & _
m_q_2 ^ 2^s_0x ds _
x(s)e^-s / m^2 + k_x(m^2)+ k_i(m^2 ) . the analysis of the two - point function for the scalar mesons , and the explicit expressions for the functions occuring in eqs .
( power2 ) , ( twopointd ) and ( twopoint ) are given in appendices a and b respectively .
the final sum rule for the form factor is obtained from eq .
( power2 ) by inserting for @xmath45 and @xmath59 the expressions @xmath60 and @xmath61 of eqs .
( twopointd ) and ( twopoint ) + & & e^m_d^2/m_d^2 ( m_d^2m_cf_d(m_d^2 ) a_x(m_x^2)1 _ ( m_a+m_b)^2^s_0xds_x_x(s_x ) e^-s_x
/ m_x^2)^-1 + & & ( _ s_d , th^s_0d ds_d_s_x , th^s_0x ds_x _ + x(s_d , s_x , t)e^-s_d^2/m_d^2e^-s_x / m_x^2 + & + & k_+x(m_d^2,m_x^2,t ) + k_+i(m_d^2,m_x^2,t ) ) .
the radiative corrections for the scalar and pseudoscalar channels are known to be large bro .
they are expected to be large in the three - point function too . by inserting the sum rule expressions for the two - point functions , eqs .
( twopointd ) and ( twopoint ) , in the denominator of the sum rule for the three - point function , eq .
( sr3 ) , we expect , at least , a partial cancellation of these corrections @xcite .
the sum rule eq .
( sr3 ) is evaluated in the same way as described in bbd91 , and we only sketch the main steps of this evaluation . in the complete theory , the right hand side of eq .
( sr3 ) should not depend on the borel variables @xmath62 .
however , in a truncated treatment there will always be some dependence left
. therefore , one has to work in a region where the approximations made are supposedly acceptable and where the result depends only moderately on the borel variables . to decrease the dependence of the results on the borel variables @xmath62 ,
we take them in the two - point functions at half the value of the corresponding variables in the three - point sum rules , _
i.e. _ , in eq.(sr3 ) we put ^2= m_d^2/2 and m_x^2= m_x^2/2 .
we furthermore choose = .
we have checked that the results do not depend crucially on this particular choice .
if the momentum transfer @xmath63 to the lepton pair is larger than a critical value @xmath64 , non - landau singularities have to be taken into account bbd91 . since anyhow we have to stay away from the physical region , _
i.e. _ we must have @xmath65 , we limit our calculation to the region @xmath66 . in this range
the t - dependence can be obtained from the sum rule ( sr3 ) directly .
it can be fitted by a monopole , and extrapolated to the full kinematical region .
since we do not take into account radiative corrections we choose the qcd parameters at a fixed renormalisaton scale of about @xmath67 : the strange and charm mass @xmath68 , @xmath69 , the up and down quark masses are put to zero .
we take for the non - strange quark condensate @xmath70 , for the strange quark condensate @xmath71 , and for the mixed quark - gluon condensate @xmath72 with @xmath73 . for the continuum threshold in the @xmath0-channel
we take from @xcite @xmath74 .
the standard value in the @xmath7 channel would be @xmath75 , yielding @xmath76 and @xmath77 . as an additional condition we use the mass constraint from the two - point function ,
as described in appendix a. we start with the decay @xmath78 and first consider the case where the scalar quark current ( intscal ) couples directly to the @xmath26-signal through a breit - wigner distribution ( [ rho1 ] ) . in fig .
2 we show the different contributions to the form factor @xmath79 in eq .
( sr3 ) , as a function of the borel variable @xmath80 , using the continuum threshold @xmath81 . as the lower limit for the fiducial region in @xmath80
we take that value of @xmath80 where the perturbative contribution is one half of the total contribution . as upper limit
we take the value @xmath82 , which is motivated in appendix a. for such a high value of the borel parameter the result is very stable but it is largely determined by the choice of the continuum model .
the instanton contribution in the fiducial region is completely negligible ; the five dimensional mixed condensate is strongly suppressed compared to the three dimensional quark condensate . in the borel variable of the @xmath7-channel
the fiducial region corresponds approximately to the range @xmath83 .
[ fig2 ] = 8.0 cm in the range @xmath84 no non - landau singularities occur for our choices of the continuum thresholds
. the momentum dependence of @xmath85 can , in this @xmath63-range , be very well approximated by a monopole expression f_+(t)=f_+(0)1-tm_p^2 , and extrapolated to the full physical region . for @xmath86 and for values of @xmath80 discussed above , we find for the form factor at @xmath87 0.48f_+(0 ) 0.55 , and for the pole mass , 1.9 m_p 2.2 .
the pole mass is considerably smaller than the mass of the charmed pseudovector - meson @xmath88 which would fit into the @xmath63-channel . in the @xmath89 decay
the pole mass also came out to be smaller bbd91 than the mass of the strange vector meson @xmath90 . in the limits of the borel variables and the continuum thresholds discussed above we obtain for the total semileptonic decay width ( d)= ( 5.5 1.0 ) 10 ^ -15 gev , where we have used @xmath91 . the same calculation done in an
zero width approximation would yield a total semileptonic decay width which is about 20 % larger .
[ fig3 ] the spectral distribution @xmath92 , eq .
( [ gaspec ] ) , where @xmath37 is the invariant mass of the @xmath93 state , is given in fig .
3 , solid line .
next we investigate the same decay under the assumption that the scalar current in eq .
( intscal ) does not couple to a resonance , but to an uncorrelated @xmath93 pair in an @xmath2-state , _ i.e. _ , we use the ansatz ( rho2 ) for @xmath94 entering the sum rules . in fig .
4 we show the dependence on the borel variable @xmath80 of the decay form factor @xmath95 for @xmath81 . note that now the density ( rho2 ) describes a two - particle phase space and therefore the dimension of @xmath85 is different from the previous case .
[ fig4 ] = 8.0 cm the masses of the pole fit to the @xmath63-dependence are practically the same as for the resonance case ; for the form factor at @xmath87 we obtain 5.0 ^-1 f_+(0 ) 7.1 ^-1 .
the total width comes out for the case of an uncorrelated @xmath93 pair in an @xmath2 state as ( d(k ) _
s|_)=(3.7 1.1)10 ^ -15 gev.[deps ] the spectral distribution for this case is also shown in fig . 3 , with a dotted line .
although there is no resonance formation the resulting distribution shows a maximum at approximately the mass of the @xmath26 , which is an effect of the decrease of the total phase space near the kinematical limits .
the evaluation of the decay @xmath96 follows exactly the same lines . here the fiducial range in @xmath80 is chosen according to the same criteria as before and goes approximately from @xmath97 to @xmath98 corresponding approximately to a range @xmath99 .
the continuum limit @xmath100 was chosen between 1 and 1.6 @xmath101 .
the resulting form factors and total decay width are , in the case of a resonance formation with a breit - wigner width 0.42f_+(0 ) 0.57 , ( d| _ ) = ( 8.0 2.5 ) 10 ^ -16 gev , [ gasibw ] and for the case of two uncorrelated @xmath102-mesons in an @xmath2-state 5 ^-1 f_+(0 ) 6 ^-1 and ( d()_s | _ ) = ( 4.5 1.0 ) 10 ^ -16 gev .
[ gasicont ] the spectral distributions for both ansatze are shown in fig .
the semileptonic decays of @xmath0-mesons into scalar hadrons are very similar to those into pseudoscalars . the same role played by the vector part in the pseudoscalar case
is played by the weak pseudovector for the decay into scalars .
therefore , the theoretical expressions are very similar , only terms proportional to the mass of the strange quark change sign .
this leads to a small reduction of the form factors compared to the decay into a pseudoscalar state .
the main difference of the semileptonic decay rates into @xmath17 and @xmath26 is due to the different phase space .
the decay into @xmath25 mesons is suppressed by the small value of the weak @xmath103 matrix element @xmath104 .
the spectral distributions of the invariant masses of the @xmath14 and @xmath12 sytems are given in figs .
their maxima are well below the masses of the @xmath26 and the @xmath25 , and also the forms are quite different from breit - wigner distributions .
the increase of the spectral ditributions is steeper and the fall - off substantially faster than for the corresponding breit - wigner forms .
this is an effect of the total phase space in the semileptonic decay .
if the scalar current does not couple directly to the resonance but only to an uncorrelated meson pair , the decay rate is reduced by a factor two compared with the decay into resonances , but nevertheless an enhancement near 0.8 and 0.5 gev is visible due to the total final state phase space ( see figs . 3 and 5 ) . with good statistics a discrimination between the two extreme cases should be possible
this would add valuable information on the nature of the intriguing low lying scalar resonances .
in reality things might be complicated by a coupling of the interpolating current intscal on the quark level to a resonance as well as to a uncorrelated meson pair . in this case on has to construct a new density @xmath105 and use it instead of the densities in eqs .
( rho1 ) or ( rho2 ) , in order to calculate the form factors @xmath106.the procedure that follows is the same .
: we would like to thank i. bediaga for fruitful discussions .
this work has been supported by cnpq , faperj , fapesp ( brazil ) , and daad ( fed .
germany ) .
in fig . 6 we display the different contributions to @xmath107 , as a function of the borel variable @xmath108 using the ansatz in eq .
( [ rho1 ] ) for @xmath94 and @xmath109 .
as expected the instanton contribution is particularly important for small values of the borel variable . in spite of the weak dependence of the sum rule results on @xmath108
, it is well known that radiative corrections could be large in this case .
therefore , we do not attach great significance to the high stability of our results .
[ fig6 ] = 8.0 cm in the zero width approximation one can obtain a sum rule for the mass by performing the logarithmic derivative of the right and left - hand sides of eq .
( twopointd ) with respect to @xmath110 , assuming that @xmath45 is independent of @xmath110 . in fig .
7 we display the logarithmic derivative of the left and right - hand sides of the sum rule ( twopoint ) , again assuming that @xmath107 is independent of the borel variable . [ fig7 ] = 8.0 cm for @xmath111 there is a reasonable overlap between the two sides of the sum rule in the borel window @xmath112 , and this defines the range for the borel variable @xmath113 in eq .
( [ sr3 ] ) . since all borel variables are related by eqs .
( m23 ) and ( mm ) , we obtain from the above range approximately @xmath114 .
[ fig8 ] = 8.0 cm in the case of @xmath5 , similar results are obtained and in fig .
8 we show ( for @xmath115 and the ansatz in eq .
( [ rho1 ] ) for @xmath116 ) that the instanton contribution ( dotted line ) is even more important in this case , giving a very stable result as a function of the borel mass . the quark condensate and mixed condensate contributions ( dashed line ) are zero since they are now proportional to the light quark mass , which we take equal to zero . in fig .
9 we show the lhs and the rhs of the sum rule generated by the logarithmic derivative of eq .
( twopoint ) with respect to @xmath110 , as a function of the borel mass , using the ansatz in eq .
( [ rho1 ] ) for @xmath116 and @xmath115 .
we see that there is again a good overlap between the two sides of the sum rule in the borel window @xmath117 .
since the mass of the particle is related with the square root of the lhs of this sum rule , from this figure we see that our result is compatible with @xmath118 found in @xcite .
[ fig9 ] = 8.0 cm
in all this work we take into account the mass of the strange quark at most squared and neglect the mass of the light quarks . for the scalar meson @xmath7 ,
we consider the particular case of @xmath26 , since the @xmath25 can be easily obtained from it by neglecting the strange quark mass .
the perturbative contributions for the two - point functions defined in sec .
iii are : _
d(s)=38 ^ 2(s - m_c^2)^2s and _ ( s)=38 ^ 2(s-2m_s^2 ) . the nonperturbative contributions including the quark and mixed condensates are k_d(m^2)=-m_c|qqe^-m_c^2/m^2 and k_(m^2)=m_s e^-m_s^2/m^2(|qq+ |ss2+m_0 ^ 22m^2|qq ) , where we have defined @xmath119 . the instanton induced contribution is given by k_i(m^2)=|n2|m_s|m_qm^2z^2_z^2/4^ dxx^2(x - z^2/4)^2e^-x^2x - z^2/4,[inst ] where @xmath120 , and we have introduced the average instanton size , @xmath121 , and the instanton number density , @xmath122 , given by @xcite @xmath123 in eq .
( [ inst ] ) @xmath124 and @xmath125 are the effective quark masses which we take to be @xmath126 and @xmath127 . the perturbative double spectral function is obtained by using the cutkosky rules and , in the case of the meson @xmath26 , is given by _ + ( s_d , s_x , t)&=&-38 ^ 2 ^ 3/2 + & & ( s_d - m_c^2)(s_x - s_xmin)(s_xmax - s_x ) , where s_xmin = m_c^2(-m_c^2+m_s^2+s_d+t)+m_s^2s_d - s_dt+(m_c^2-s_d ) 2m_c^2 and s_xmax = m_c^2(-m_c^2+m_s^2+s_d+t)+m_s^2s_d - s_dt-(m_c^2-s_d ) 2m_c^2 .
the nonperturbative contributions , including the quark and mixed condensates , which survive the double borel transformation in @xmath128 and @xmath129 are k_+(m_d^2,m_x^2,t)&= & |qqe^-m_c^2/m_d^2 , and the instanton induced contribution is given by k_+i(m_d^2,m_x^2,t)&= & 8|^4t^2|n |m_s|m_q _
0^dr ds du d ( m_d^2 - 1)m_x^4(1 + 4s)^2(s m_d^2 - 4m_x^2)^2s^2 + & & e^-|^2(r+u ) e^-m_c^2 e^-_0|^2 e^-ta_0sm_d^2 , where a_0=s(m_d^2 - 1)1 + 4s and _ 0=s+(1 + 4s)t)m_x^2s+(1 + 4s)t)-4m_x^2 .
groom et al .
( particle data group),_eur .
_ c15:437 , 2000 .
aitala et al .
_ 86:770 , 2001 c. gbel , ( for the e791 collaboration ) , hep - ex/0110052 .
r. gatto , g. nardulli , a.d .
polosa and n.a .
tornqvist , _ phys .
_ , b494:168 , 2000 j.m .
link et al ( focus collaboration ) , hep - ex/0203031 m.a .
shifman , a.i . and vainshtein and v.i .
zakharov , _ nucl .
_ , b147:385 , 1979 | semileptonic decays of @xmath0-mesons into scalar hadronic states are investigated .
two extreme cases are considered : a ) the meson decays directly into an uncorrelated scalar state of two two mesons and b ) the decay proceeds via resonance formation .
qcd sum rules including instanton contributions are used to calculate total and differential decay rates under the two assumptions .
ps .
pacs numbers : 12.38.lg , 13.30.ce , 14.65.dw |
the topic at hand is perverse sheaves and nearby cycles ; for greater accessibility of these notes , we give a summary of the definitions and necessary properties here .
occasionally , we indicate diagram chases in a proof . for ease of reading we have tried not to make this an essential point , but in case the reader should find such a chase to be a convincing informal argument
, we indicate here why it is also a convincing formal one .
every object in an abelian category @xmath3 can be considered , via yoneda s lemma , to be a sheaf , namely its functor of points , on the _ canonical topology _ of @xmath3 .
this is , by definition , the largest grothendieck topology in which all representable functors @xmath4 are sheaves , and its open covers are precisely the _ universal strict epimorphisms_. such a map is , in a more general category , a map @xmath5 such that the fibered product @xmath6 exists , the coequalizer @xmath7 exists , the natural map @xmath8 is an isomorphism , and that all of this is _ also _ true when we make any base change along a map @xmath9 , for the induced map @xmath10 . in an abelian category , however , this is all equivalent merely to the statement that @xmath11 is a surjection .
recall the definitions of the various constructions on sheaves : 1 .
kernels of maps are taken sectionwise ; i.e. for a map @xmath12 , @xmath13 .
likewise , products and limits are taken sectionwise .
cokernels are _ locally _ taken sectionwise : any section @xmath14 is , on some open cover @xmath15 of @xmath16 , of the form @xmath17 for @xmath18 .
likewise , images , coproducts , and colimits are taken locally . in an abelian category , where all of these constructions exist by assumption , these descriptions
are even prescriptive : if one forms the sheaves thus described , they are representable by the objects claimed .
therefore , the following common arguments in diagram chasing are valid : 1 .
a map @xmath19 is surjective if and only if for every @xmath20 , there is some @xmath21 such that @xmath22 .
this is code for : for every `` open set '' @xmath16 and every @xmath23 , there is a surjection @xmath24 and a section @xmath25 such that @xmath26 .
2 . if @xmath20 , then @xmath27 if and only if @xmath28 .
this is code for : if @xmath23 and @xmath29 , then there is some surjection @xmath24 and @xmath30 with @xmath26 .
3 . for @xmath31 , @xmath32 if and only if @xmath33 . here , the sum of maps @xmath34 and @xmath35 is obtained by forming the fibered product @xmath36 which covers both @xmath16 and @xmath15 , and then taking the sum of the maps @xmath37 ; the condition for equality is just the statement that a section of a sheaf vanishes if only it vanishes on an open cover .
any other arguments involving elements and some concept related to exactness can also be phrased in this way .
thus , a nave diagram - chasing argument can be converted into a rigorous one simply by replacing statements like @xmath38 with correct ones @xmath39 for some open set @xmath16 , and passing to surjective covers of @xmath16 when necessary .
all the action takes place in the derived category ; specifically , let @xmath40 be an algebraic variety and denote by @xmath41 its derived category of bounded complexes of sheaves of vector spaces with constructible cohomology . by definition , a map of complexes @xmath42 defines an isomorphism in @xmath41 if and only if its associated map on cohomology sheaves @xmath43 is an isomorphism for all @xmath44 .
we have a notation for the index - shift : @xmath45)^i$ ] ( technically , the differential maps also change sign , but we will never need to think about this ) .
the derived category @xmath41 is a `` triangulated category '' , which means merely that in it are a class of triples , called `` distinguished triangles '' , of complexes and maps @xmath46\ ] ] in which two consecutive arrows compose to zero , satisfying the axioms given in , for example , @xcite ( but see also [ sec : comments ] ) , and with the property that the associated long sequence of cohomology sheaves @xmath47 is exact ( note that @xmath48 ) = h^1(a^\bullet)$ ] ) ; we say that the @xmath49 are `` cohomological '' . if @xmath42 is given , there always exists a triangle whose third term @xmath50 is the `` cone '' of @xmath11 ; this cone is unique up to nonunique isomorphism and any commutative diagram of maps @xmath11 induces a map on cones , but this is not functorial .
it follows that the induced triangle itself is unique up to a nonunique isomorphism whose component morphisms on @xmath51 and @xmath52 are the identity maps .
a functor between two triangulated categories is `` triangulated '' if it sends triangles in one to triangles in the other .
in @xmath41 we also have some standard constructions of sheaf theory .
for any two complexes there is the `` total tensor product '' @xmath53 obtained by taking in degree @xmath54 the direct sum of all products @xmath55 with @xmath56 ( and some differentials that are irrelevant ) and its derived bifunctor @xmath57 , with @xmath58 , which is a triangulated functor in each variable .
we also have the bifunctor ( contravariant in the first argument ) @xmath59 , whose terms are @xmath60|_u)$ ] , and its derived bifunctor @xmath61 , with @xmath62 , which is triangulated in each variable .
of course , these two have an adjunction : @xmath63 for any zariski - open subset @xmath64 with inclusion map @xmath65 , there are triangulated functors @xmath66 and
@xmath67 ; if @xmath44 is the inclusion of its complement @xmath68 , then there are likewise maps @xmath69 and @xmath70 .
( technically the operation @xmath71 is only left exact on sheaves and we should write @xmath72 for its derived functor , but we will never have occasion to use the plain version so we elide this extra notation . )
they satisfy a number of important relations , of which we will only use one here : there is a functorial triangle in the complex @xmath73 : @xmath74 we will generally forget about writing @xmath75 and consider @xmath76 . there is also a triangulated duality functor @xmath77 which interchanges @xmath78 and @xmath79 , in that @xmath80 , etc . , and is an involution . in fact , if we set @xmath81 , then @xmath82 .
for any map @xmath83 of varieties , we have @xmath84 as well ( also @xmath85 , and none of them are equal ) , with the same relationships to @xmath2 , and the useful identity @xmath86 note that by these properties , we have @xmath87 . here
we give a detail - free overview of the formalism of perverse sheaves created in @xcite .
within @xmath41 there is an abelian category @xmath88 of `` perverse sheaves '' which has nicer properties than the category of actual sheaves .
it is specified by means of a `` t - structure '' , namely , a pair of full subcategories @xmath89 and @xmath90 , also satisfying some conditions we wo nt use , and such that @xmath91 there are truncation functors @xmath92 and likewise for @xmath93 , fitting into a distinguished triangle for any complex @xmath73 : @xmath94 ( where @xmath95 \circ \tau^{\geqslant 0 } \circ [ 1]$ ] ) .
this triangle is _ unique _ with respect to the property that the first term is in @xmath96 and the third is in @xmath97 .
they have the obvious properties implied by the notation : @xmath98 if @xmath99 , and likewise for @xmath100 .
furthermore , there are `` perverse cohomology '' functors @xmath101 , where of course @xmath102)$ ] and @xmath103 ; these are cohomological just like the ordinary cohomology functors .
the abelian category structure of @xmath88 is more or less determined by the fact that if we have a map @xmath12 of perverse sheaves ( this is the notation we will be using ; we will not think of perverse sheaves as complexes ) , then @xmath104 for notational convenience , we will write @xmath105 for a perverse sheaf on @xmath16 , @xmath106 for one on @xmath40 , and as usual , abandon @xmath75 and just consider @xmath107 ( for the reason expressed immediately below , this is reasonable ) . the category @xmath88 is closed under the duality functor @xmath2 , but not necessarily under the six functors defined for an open / closed pair of subvarieties
. however , it is true that @xmath108 and @xmath109 , while @xmath110 ( @xmath111 a perverse sheaf on @xmath68 ) ; we say these functors are right , left , or just `` t - exact '' .
furthermore , when @xmath65 is an affine morphism ( the primary example being when @xmath68 is a cartier divisor ) , both @xmath112 and @xmath71 are t - exact , and thus their restriction to @xmath113 is exact with values in @xmath88 .
there is also a `` minimal extension '' functor @xmath114 , defined so that @xmath115 is the image of @xmath116 in @xmath117 along the natural map @xmath118 ; it is the unique perverse sheaf such that @xmath119 and @xmath120 , but for us the most useful property is that when @xmath65 is an affine , open immersion , then we have a sequence of _ perverse sheaves _
@xmath121 \incl j _ !
\sh{m } \surj j _ { ! * } \sh{m } \incl j _ * \sh{m } \surj i^ ! j _ { ! * } \sh{m}[1];\ ] ] i.e. @xmath122 = \on{ker}(j _ ! \sh{m } \to j _ * \sh{m})$ ] and @xmath123 = \on{coker}(j _ ! \sh{m } \to j _ * \sh{m})$ ] are both perverse sheaves .
perverse sheaves have good category - theoretic properties : @xmath88 is both artinian and noetherian , so every perverse sheaf has finite length .
finally , we will use the sheaf - theoretic fact that if @xmath124 is a locally constant sheaf on @xmath40 , then @xmath125 is perverse whenever @xmath106 is .
note that since @xmath124 is locally free , it is flat , and therefore @xmath126 .
if we have a map @xmath127 such that @xmath128 ( so @xmath129 ) , the `` nearby cycles '' functor @xmath130 is defined .
namely , let @xmath131 be the universal cover of @xmath132 , let @xmath133 be its pullback , forming a diagram @xmath134_{v } \ar[dd ]
\\ z \ar[r]^i \ar[dd ] & x \ar[dd]^f & u \ar[l]_{j } \ar[dd ] & \\ & & & { \tilde{\gm } } \ar[dl]^{u } \\
\{0\ } \ar[r ] & \aff^1 & \gm \ar[l ] & } \ ] ] and set ( in this one instance , explicitly writing @xmath71 and @xmath135 as non - derived functors ) @xmath136 since @xmath137 and @xmath138 are exact , indeed @xmath139 is a left - exact functor from sheaves on @xmath16 to sheaves on @xmath68 .
many sources ( e.g. @xcite*1.1.1 ) give the definition @xmath140 ; in fact , they are the same : since @xmath141 is a covering map , if @xmath106 is a flasque sheaf on @xmath16 , then @xmath142 is flasque on @xmath143 and so acyclic for @xmath135 ( and @xmath144 acyclic for @xmath71 ) .
therefore we may form the derived functor before or after composition .
note that @xmath141 is not an algebraic map , and therefore it is not _ a priori _ clear whether @xmath145 preserves constructibility ; that it does is a theorem of deligne ( @xcite , expos xiii , thorme 2.3 for tale sheaves and expos xiv , thorme 2.8 for the comparison with classical nearby cycles ) .
the fundamental group @xmath146 acts on any @xmath147 via deck transformations of @xmath148 and therefore acts on @xmath139 and @xmath145 .
there is a natural map @xmath149 , obtained from @xmath150-adjunction , on whose image @xmath146 acts trivially .
we set , by definition , @xmath151 where @xmath152 is the `` vanishing cycles '' sheaf .
using some homological algebra tricks the above sequence induces a natural distinguished triangle @xmath153 where @xmath154 is ( morally ) the right derived functor of @xmath155 . like @xmath145
, @xmath154 has a monodromy action of @xmath146 ; this action is one of the maps on the cone of the above triangle induced by the monodromy action on @xmath145 , but as this is not functorial , one should consult the real definition in @xcite ( given for the algebraic nearby cycles , but see also the second expos ) .
lemunipotent nearby cycles there exists a unique decomposition of @xmath145 as @xmath156 , where for any choice of generator @xmath157 of @xmath146 , @xmath158 acts nilpotently on @xmath159 for any complex @xmath160 and is an automorphism of @xmath161 .
the part @xmath162 is called the functor of _ unipotent nearby cycles_. to start , we observe that for any sheaf @xmath106 on @xmath16 , we have @xmath163 , and so if @xmath106 is constructible , by the constructibility of nearby cycles so is @xmath164 ; thus , for any open @xmath165 , @xmath166 is finite - dimensional .
let @xmath167 be the subfunctor such that for any sheaf @xmath106 on @xmath16 , @xmath168 is the subsheaf of @xmath164 in which @xmath158 is nilpotent , so for each @xmath15 , @xmath169 is the generalized eigenspace of @xmath157 with eigenvalue @xmath170 .
therefore it is actually a direct summand ; we recall the general argument which works over any field @xmath171 .
if @xmath172 is an endomorphism of a finite - dimensional vector space @xmath173 , we view @xmath173 as a @xmath174$]-module with @xmath175 acting as @xmath172 . by the classification of modules over a principal ideal domain , we have @xmath176/p(x)$ ] for certain polynomials @xmath177 .
the generalized eigenspace with eigenvalue @xmath170 is then the sum of those pieces for which @xmath177 is a power of @xmath178 , and the remaining summands are a @xmath172-invariant complement in which @xmath179 acts invertibly . as the image of @xmath180 ( @xmath181 ) ,
this complement is functorial in the category of finite - dimensional @xmath174$]-modules and so we have the same decomposition in sheaves of finite - dimensional @xmath174$]-modules .
specializing again to the present situation , we get a decomposition @xmath182 ( this is the definition of @xmath183 ) .
both summands are _ a fortiori _ left - exact functors and taking derived functors , we obtain a decomposition : @xmath184 since @xmath158 is nilpotent on any @xmath168 , we may apply @xmath185 to any constructible complex of injectives , thus computing @xmath159 for any complex @xmath160 , and conclude that @xmath158 is nilpotent on each such ; by general principles it is invertible on @xmath161 .
this is what we want .
uniqueness of the decomposition is clear ; indeed , if in any category with a zero object we have objects @xmath175 and @xmath186 together with endomorphisms @xmath187 and @xmath188 respectively such that @xmath187 is nilpotent and @xmath188 invertible , then any map @xmath189 intertwining @xmath187 and @xmath188 is zero : we have @xmath190 , so @xmath191 if @xmath192 . for morphisms @xmath193 we work in the opposite category .
in particular , if we have @xmath194 as a sum of two functors as in the statement of the lemma , then the identity map on @xmath145 has no @xmath195-component on @xmath162 and no @xmath196-component on @xmath161 , and so induces isomorphisms @xmath197 and @xmath198 .
we note that this lemma is a special case of [ nearby cycles decomposition ] when the field of coefficients is algebraically closed . however , this decomposition is defined over any field .
there is a triangle , functorial in @xmath199 , @xmath200 ( see ) which , taking @xmath201 and inserting @xmath162 because the monodromy acts trivially on the first term , gives the extremely important ( for us ) triangle @xmath202 we also have a unipotent part of the vanishing cycles functor @xmath154 , and , again since the monodromy acts trivially on @xmath203 , a corresponding triangle @xmath204 if @xmath124 is any locally constant sheaf on @xmath205 with underlying vector space @xmath206 and unipotent monodromy , then @xmath207 , where @xmath146 acts on the tensor product by acting on each factor ( since @xmath124 is trivialized on @xmath148 ) .
we note the following fact , crucial to all computations in this paper : _ @xmath65 is an affine morphism .
_ indeed , @xmath68 is cut out by a single algebraic equation .
although when @xmath68 is any cartier divisor it is still locally defined by equations @xmath11 and the inclusion @xmath65 of its complement is again an affine morphism , it is not necessarily possible to glue nearby cycles which are locally defined as above ; c.f .
@xcite*remark 5.5.4 ; an explicit example will appear in @xcite .
however , it follows from [ nearby cycles computations ] that when @xmath105 is a perverse sheaf and @xmath208 has trivial monodromy , it is in fact independent of @xmath11 and gluing is indeed possible .
triangle [ eq : psiun ] already implies that nearby cycles preserve perverse sheaves .
lemnearby cycles are perverse the functor @xmath209 $ ] sends @xmath210 to @xmath211 and takes @xmath113 to @xmath212 . since @xmath65 is affine and an open immersion , @xmath71 and @xmath112 are t - exact , so for any @xmath213 , @xmath214 is in @xmath215 .
if we apply the long exact sequence of perverse cohomology to triangle [ eq : psiun ] , we therefore get in nonnegative degrees : @xmath216 for @xmath217 , the map @xmath218 is both given by a _
nilpotent _ operator and is surjective , so zero .
it follows that @xmath219 , as promised .
now let @xmath220 be a perverse sheaf .
then @xmath221}$ ] since its perverse cohomology sheaves are the kernel and cokernel of the map @xmath222 . in degrees
@xmath223 , then , we have @xmath224 this means that for @xmath225 , all the maps @xmath158 are injective and nilpotent , hence zero
. thus @xmath226 , as desired .
since @xmath209 $ ] acts on perverse sheaves , we will give it the abbreviated notation @xmath227 .
let @xmath228 be the vector space of dimension @xmath229 together with the action of a matrix @xmath230 $ ] , a unipotent ( variant of a ) jordan block of dimension @xmath231 .
let @xmath232 be the locally constant sheaf on @xmath205 whose underlying space is @xmath228 and in whose monodromy action a ( hereafter fixed choice of ) generator @xmath157 of @xmath146 acts by @xmath233 . since it is locally free , it is flat , so we will write @xmath234 rather than @xmath235 in tensor products with it ( actually , with @xmath236 ) .
it has the following self - duality properties , where @xmath237 is the dual local system and @xmath238 is the local system in whose monodromy @xmath157 acts by @xmath239 : lemjordan block dual we have @xmath240 , and @xmath241 for @xmath242 .
since @xmath243 is the local system with vector space the dual @xmath244 and monodromy @xmath245 , its monodromy is again unipotent with a single jordan block of length @xmath231 .
we fix in @xmath228 the given basis @xmath246 associated to @xmath247 , and in @xmath244 we choose a generalized eigenbasis @xmath248 in which @xmath245 has the matrix @xmath233 , so that sending @xmath249 identifies @xmath233 with @xmath245 and thus induces the desired map of local systems .
the same proof shows that @xmath250 . in general ,
then , we construct an isomorphism : @xmath251 where @xmath252 by constructing a map @xmath253 such a map can be obtained by applying @xmath254-adjunction to a map : @xmath255 since @xmath2 exchanges @xmath78 and @xmath79 , we have @xmath256 . by the property [ eq : exceptional hom localization ] of @xmath257
, we have @xmath258 note also that @xmath259 by definition and therefore the map [ eq : dual iso adjoint ] can be constructed by applying @xmath260 to a certain map on @xmath205 : @xmath261 this map , in turn , is obtained by first replacing the @xmath235 with @xmath234 ( since @xmath232 is locally free ) and applying @xmath2 to the pairing @xmath262 given by the isomorphism @xmath263 described in the first paragraph .
thus , locally [ eq : dual iso ] is the tautological isomorphism @xmath264 .
since it is a local isomorphism , it is an isomorphism . in the rest of this section
, @xmath105 is any object of @xmath113 .
the following construction is beilinson s definition of the unipotent nearby cycles : propnearby cycles construction let @xmath265 be the natural map .
then there is an inclusion @xmath266 , identifying the actions of @xmath146 , which is an isomorphism for all sufficiently large @xmath231 .
( in fact , it suffices to take @xmath231 large enough that @xmath267 annihilates @xmath268 . ) we know by [ nearby cycles are perverse ] that @xmath269 is a perverse sheaf , so taking together the triangle [ eq : psiun ] with @xmath270 and exact sequence [ eq : extension triangle ] with @xmath271 , we see that @xmath272 , where @xmath158 is the map appearing in the former triangle shifted by @xmath273 .
we also have @xmath274 where the @xmath44th coordinate of the action of @xmath157 is @xmath275 , with @xmath276 the copy of @xmath277 acting on @xmath268 considered as the @xmath44th summand .
that is , using elements , @xmath278 is sent by @xmath157 to @xmath279 .
thus , for an element of @xmath280 , we have @xmath281 , or : @xmath282 if we define a map @xmath283 by sending the element @xmath284 to the coordinates @xmath285 defined by the first formula above , then @xmath286 is injective and its image contains @xmath280 ( namely , that subspace satisfying the second equation ) . since @xmath158 ( hence @xmath287 ) is nilpotent on @xmath268 , for @xmath231 sufficiently large , @xmath288 .
we claim that @xmath286 intertwines the actions of @xmath289 and @xmath233 : @xmath290 finally , we employ the isomorphism @xmath250 of [ jordan block dual ] to give an automorphism of @xmath291 intertwining @xmath233 and @xmath239 .
coruniformity there exists an integer @xmath187 such that @xmath292 annihilates both @xmath293 and @xmath294 for all @xmath231 . by [ nearby cycles construction ] , the kernel is contained in @xmath268 and thus annihilated by that power of @xmath158 which annihilates the nearby cycles .
temporarily let @xmath295 ; then @xmath296 , so @xmath297 is again annihilated by some @xmath292 . in preparation for the next section ,
we give a generalization of this construction . for each @xmath298
there is a natural short exact sequence @xmath299 that is , for any @xmath300 , @xmath301 sends @xmath232 to the first @xmath231 coordinates of @xmath302 if @xmath303 , and to the quotient @xmath304 given by collapsing the first @xmath305 coordinates if @xmath306 ( that is , @xmath307 ) and @xmath308 .
this sequence respects the action of @xmath146 on the terms and , via [ jordan block dual ] , the @xmath309 sequence is dual to the @xmath310 sequence .
let @xmath220 ; then we have induced maps on the tensor products : @xmath311 ( we will often omit the subscript @xmath105 when no confusion is possible ) . by [ jordan block dual ] ,
these satisfy @xmath312 note that since the @xmath232 are locally free , the @xmath313 are all injective when @xmath314 and surjective when @xmath307 .
let @xmath300 and set @xmath315 we will use the following self - evident properties of the @xmath316 : lem(a , r ) factorization the @xmath316 satisfy : 1 .
[ en : alternating ] when @xmath308 , we have @xmath317 .
[ en : transitivity ] when @xmath318 and @xmath319 have the same sign and @xmath320 , we have @xmath321 .
[ en : images ] let @xmath303 , @xmath322 ; then we have : @xmath323 finally , by [ uniformity ] and ( ) , for @xmath303 , @xmath324 annihilates @xmath325 and @xmath326 . from now on
, we will assume @xmath303 .
propstable kernel and cokernel for @xmath327 , the natural maps @xmath328 and @xmath329 respectively induce isomorphisms @xmath330 and @xmath331 and @xmath332 induce isomorphisms @xmath333 using the maps @xmath334 and @xmath335 we get a square which , using [ ( a , r ) factorization ] ( , ) , we verify is commutative : @xmath336 showing that @xmath334 induces a map on kernels .
since it is injective , we get a long sequence of inclusions of kernels : @xmath337 by [ ( a , r ) factorization ] , each kernel is annihilated by @xmath324 , whose kernel is ( for @xmath338 ) the perverse sheaf @xmath339 ; thus , this sequence is contained in this sheaf .
since perverse sheaves are noetherian , this chain must have a maximum , so the kernels stabilize . for the cokernels , we apply [ eq : g duality ] to the argument of [ uniformity ] .
( one can also argue directly using the artinian property of perverse sheaves . ) for the second statement concerning kernels , since @xmath292 annihilates @xmath340 , for @xmath341 it is contained in @xmath342 , and therefore by definition in @xmath343 .
the statement on cokernels is again obtained by dualization and [ eq : g duality ] .
( a direct argument employing a diagram chase is also possible , using the fact that @xmath344 . ) departing slightly from beilinson s notation , we denote these stable kernels and cokernels @xmath345 and @xmath346 for @xmath303 ; when @xmath347 we drop it .
propkernel equals cokernel there is a natural isomorphism @xmath348 .
consider the map of short exact sequences for any @xmath231 and any @xmath349 ( to eliminate clutter we have not written the superscripts on the maps @xmath350 ) :
@xmath351 by the snake lemma , we have an exact sequence of kernels and cokernels : @xmath352 if @xmath353 , then the first and last maps are , by the second part of [ stable kernel and cokernel ] , isomorphisms .
therefore @xmath354 is an isomorphism .
since the long exact sequence of cohomology [ eq : snake sequence ] is natural , we see that @xmath354 is independent of @xmath231 and @xmath355 in the sense of the proposition : @xmath356 where the requisite commutative diagrams are produced using [ ( a , r ) factorization ] ( , ) . for the same reason ,
@xmath354 is a natural transformation between the two functors @xmath357 because they are equal , we will give a single name @xmath358 to the stable kernel and cokernel .
these are thus exact functors , and by definition of @xmath359 and [ eq : g duality ] , they commute with duality : @xmath360 . from [ nearby cycles construction ] we conclude : cornearby cycles computations for @xmath327 we have @xmath361 , and thus an isomorphism @xmath362 which is natural in the perverse sheaf @xmath105 .
a more effective , equivalent construction is obtained as follows : suppose @xmath292 annihilates @xmath268 .
then we have by [ eq : kernel and cokernel ] : @xmath363 = i^ ! j _ { ! * } ( \sh{m } \tensor f^ * \sh{l}^n)[1].\ ] ] conversely , if these equations hold , then of course @xmath292 annihilates @xmath268 .
we will refer to @xmath364 as @xmath365 , which beilinson calls the `` maximal extension functor '' and denotes without the superscript .
although there is no independent , non - unipotent analogue , we have chosen to use this notation to match that for the nearby and ( upcoming ) vanishing cycles functors , which do have such analogues .
propxi sequences there are two natural exact sequences exchanged by duality and @xmath366 : @xmath367 where @xmath368 and @xmath369 . these sequences are , respectively , the last and first halves of [ eq : snake sequence ] . for the first one , take @xmath370 , so @xmath371 and therefore @xmath372 ; we get an exact sequence @xmath373 for @xmath327 , by the second part of [ stable kernel and cokernel ] , the first map is an isomorphism , and for @xmath374 we obtain the first short exact sequence from the remaining three terms above . for the second short exact sequence
, we apply the same reasoning to [ eq : snake sequence ] with @xmath375 and then @xmath374 , with @xmath376 : @xmath377 it is obvious from these constructions and [ eq : g duality ] that the two short exact sequences are exchanged by duality . to show that @xmath378 and @xmath379 , we identify these maps in the above sequences and rewrite the claims as : @xmath380 for both , we use the fact that since @xmath381 , we have @xmath382 , as constructed in the familiar proof of the snake lemma , with the inverse interpreted as a multi - valued pullback
. then the claims are equivalent to @xmath383 which follow from [ ( a , r ) factorization ] ( , ) .
the remainder of the paper is simply what beilinson calls `` linear algebra '' ( one might argue that this has already been the case for most of the preceding ) .
take @xmath384 for a perverse sheaf @xmath385 in the above exact sequences . from the maps in these two sequences
we can form a complex : @xmath386 where @xmath387 and @xmath388 are defined by the left- and right - adjunctions @xmath389 and @xmath390 and the property that @xmath391 .
propvanishing cycles the complex [ eq : phi complex ] is in fact a complex ; let @xmath392 be its cohomology sheaf . then @xmath393 is an exact functor @xmath394 , and there are maps @xmath395 such that @xmath396 as in the following diagram : @xmath397 that [ eq : phi complex ] is a complex amounts to showing that @xmath398 , which is true by definition of the @xmath399 and adjunction . to show that @xmath393 is exact ,
suppose we have @xmath400 , so that we get a short exact sequence of complexes @xmath401 where by @xmath402 we have denoted the complex [ eq : phi complex ] padded with zeroes on both sides .
note that since @xmath403 is injective and @xmath404 surjective , @xmath402 fails to be exact only at the middle term .
therefore we have a long exact sequence of cohomology sheaves : @xmath405 which shows that @xmath393 is functorial and an exact functor .
if we apply @xmath406 to [ eq : phi complex ] , it becomes simply ( with @xmath407 ) @xmath408 which is actually exact , so @xmath409 ; i.e. @xmath392 is supported on @xmath68 . finally , to define @xmath286 and @xmath141 , let @xmath410 , and set @xmath411 in coordinates , and @xmath412 . since @xmath413 , @xmath141 factors through @xmath392 , and we have @xmath414 by [ xi sequences ] .
define a _ vanishing cycles gluing data _ for @xmath11 to be a quadruple @xmath415 as in [ vanishing cycles ] ; for any @xmath385 , the quadruple @xmath416 is such data .
let @xmath417 be the category of gluing data ; then @xmath418 is a functor .
conversely , given a vanishing cycles data @xmath419 we can form the complex @xmath420 since @xmath421 , and let @xmath422 be its cohomology sheaf .
beilinson gives an elegant framework for proving the equivalence of [ eq : phi complex ] and [ eq : inverse gluing diagram ] in @xcite*appendix . rather than proving [ gluing ] directly
, we present his technique ( with slightly modified terminology ) .
defndiads let a _
be a complex of the form @xmath423 in which @xmath424 is injective and @xmath425 is surjective ( so it is exact on the ends ) .
let the category of diads be denoted @xmath426 . let a _
be a short exact sequence of the form @xmath427 in which both @xmath428 are injections and both @xmath429 are surjections .
let the category of triads be denoted @xmath430 ; it has a _ reflection functor _
@xmath431 which invokes the natural symmetry @xmath432 , and is an involution . we can define a map @xmath433 by setting @xmath434 where the natural inclusion / projection ( resp .
projection / inclusion ) are called : @xmath435 ( note @xmath436 ) .
we define the inverse @xmath437 by the formula @xmath438 lemdiad - triad equivalence the functors @xmath439 are mutually inverse equivalences of @xmath426 with @xmath430 . before beginning the verification of the many necessary facts , we observe that the property of a sequence @xmath440 as above being in @xmath430 is equivalent to the following diagram being cartesian @xmath441 and the following smaller sequence being exact @xmath442 for @xmath443 . indeed , for [ eq : cartesian equivalent ] ,
the diagram is cartesian if and only if @xmath440 is exact in the middle , and for [ eq : small ses ] , the arrows are respectively injective and surjective by hypothesis if @xmath444 , while exactness in the middle follows from that of @xmath440 . for readability , we continue the proof as several sub - lemmas .
@xmath445 : it is easily verified that @xmath446 is an exact sequence . to see that @xmath447 is surjective and @xmath448 injective
, we consider the cartesian diagrams @xmath449 and use that @xmath425 is surjective and @xmath424 is injective .
likewise , @xmath450 is surjective and @xmath451 is injective .
@xmath452 : clearly , @xmath453 is a complex , since the sequence @xmath454 is exact ( hence a complex ) .
since @xmath455 is injective by hypothesis , @xmath456 is injective .
we must show that @xmath457 is surjective , where by [ eq : small ses ] with @xmath458 we have @xmath459 ; consider the diagram @xmath460 which is cartesian by exactness of [ eq : small ses ] . since the bottom arrow is a surjection , for @xmath461 to be a surjection it suffices to show that the left arrow is . by [ eq : cartesian equivalent ] with @xmath458
, @xmath462 is a surjection since the bottom arrow there is a surjection by hypothesis .
@xmath463 : its @xmath464 is @xmath465 ; its @xmath466 and @xmath467 are indeed @xmath466 and @xmath467 , and its @xmath468 is @xmath469 ; one checks quickly that the maps are right as well .
@xmath470 : its @xmath471 is @xmath472 since @xmath473 is an injection ; its @xmath466 and @xmath474 are obviously the original @xmath466 and @xmath474 . the small sequence [ eq : small ses ] with @xmath458 shows that @xmath475 is correct as well . finally , for @xmath476
, we must show that @xmath477 , or in other words , that @xmath478 is surjective , which follows from [ eq : cartesian equivalent ] with @xmath479 .
clearly , both of the complexes [ eq : phi complex ] and [ eq : inverse gluing diagram ] are diads . comparing them , we find that the construction of the latter is given by : cordiad reflection the reflection functor on a diad is the complex @xmath480 where @xmath481 is the natural inclusion and @xmath482 the natural projection , @xmath483 , and @xmath484 factors @xmath485 through @xmath486
. that is , @xmath487 as defined above .
we need to show that @xmath488 and @xmath489 , and prove the identities of the morphisms .
the first is easily verified directly , considering both as subobjects of @xmath490 , while for the second , we assert that the map @xmath491 induces the desired isomorphism from the latter to the former . to show that it identifies @xmath492 with @xmath493 , it suffices to check that the following diagram is cartesian : @xmath494 which follows from the definition of @xmath495 .
the identities of @xmath481 , @xmath496 , and @xmath482 are clear from these constructions , while for @xmath484 it is fastest to chase the above diagram .
thmgluing the gluing category @xmath417 is abelian ; @xmath418 and @xmath497 are mutually inverse exact functors , and so @xmath417 is equivalent to @xmath88 . that @xmath417 is abelian amounts to proving that taking coordinatewise kernels and cokernels works
that is , if we have @xmath498 and @xmath499 with maps @xmath500 , @xmath501 and such that the following diagram commutes : @xmath502 ^ -{u } \ar[d]_{{\psi^\text{un}}_f(a_u ) } & \sh{f}_z \ar[r]^-{v } \ar[d]^{a_z } & { \psi^\text{un}}_f ( \sh{m } ) \ar[d]^{{\psi^\text{un}}_f(a_u ) } \\ { \psi^\text{un}}_f ( \sh{m } ' ) \ar[r]^-{u ' } & \sh{f}_z ' \ar[r]^-{v ' } & { \psi^\text{un}}_f ( \sh{m } ' ) } \ ] ] then @xmath503 is a kernel for @xmath504 , where @xmath505 and @xmath506 are induced maps ; likewise for the cokernel ; and we must show that @xmath504 is an isomorphism if and only if the kernel and cokernel vanish .
the maps @xmath505 and @xmath506 are constructed from the natural sequence of kernels ( or cokernels ) in the above diagram , and the exactness of @xmath227 , and once they exist it is obvious from the definition of morphisms in @xmath417 that the desired gluing data is a kernel ( resp .
cokernel ) .
since @xmath113 and @xmath212 are abelian and kernels and cokernels are taken coordinatewise , the last claim follows . to show that @xmath507 and @xmath508 are mutually inverse , we interpret @xmath88 and @xmath417 as diad categories in the form given , respectively , by diagrams [ eq : phi complex ] and [ eq : inverse gluing diagram ]
. the reflection functor is given by [ diad reflection ] ; by [ xi sequences ] and the definition of @xmath509 , its value on [ eq : phi complex ] is that of the functor @xmath507 .
for the same reason , its value on [ eq : inverse gluing diagram ] is that of @xmath508 interpreted as a complex of type [ eq : phi complex ] ( the @xmath106 term is what we have previously called the value of @xmath508 ) .
since the reflection functor is an involution , @xmath508 and @xmath507 are mutually inverse .
we conclude with some musings on the theory exposited here . in the previous version http://arxiv.org/abs/1002.1686v2[arxiv:1002.1686v2 ] of these notes
, we gave a substantially different proof of [ kernel equals cokernel ] , adhering closely to that given in @xcite*key lemma .
as that proof may better illuminate the two - sided limit formalism which we also omit , the curious reader is encouraged to consult it .
the functor @xmath393 , like @xmath227 , has a familiar identity .
thmvanishing cycles and phi there is an isomorphism of functors @xmath510 $ ] and a natural distinguished triangle @xmath511 isomorphic to that in [ eq : phiun ] .
according to the definition of @xmath393 in [ vanishing cycles ] , we have a short exact sequence and , thus , a corresponding distinguished triangle of the same form : @xmath512 since @xmath513 , there is a projection map @xmath514 commuting with the inclusion of @xmath515 .
now we apply the octahedral axiom of triangulated categories as given in @xcite*(1.1.7.1 ) : @xmath516 where all the straight lines are distinguished triangles , both the ( geometric ) triangles are commutative , and the square commutes .
it is easy to see that @xmath517 must be surjective because @xmath404 is surjective ; thus , since both @xmath518 and @xmath106 are perverse , @xmath519 $ ] is also perverse , and so we have an exact sequence @xmath520 \to k \xrightarrow{\on{pr } } \sh{f } \to 0.\ ] ] but by definition , @xmath521 , and therefore @xmath519
\cong { \psi^\text{un}}_f(j^ * \sh{f})$ ] .
note that the inclusion then becomes the map @xmath286 , as defined in the proof of [ vanishing cycles ] .
rotating the other triangle in the above octahedral diagram , we have @xmath522 comparing with [ eq : phiun ] , we find that @xmath523 \cong { \phi^\text{un}}_f(\sh{f})$ ] is perverse .
conversely , starting from [ eq : phiun ] in place of the above triangle , we conclude by the octahedral axiom that @xmath523 $ ] is the cohomology of [ eq : phi complex ] , which admits a unique extension to a functor of @xmath106 compatible with the octahedral diagram .
therefore , we conclude an isomorphism of functors @xmath510 $ ] . as beilinson observes , the full nearby cycles functor @xmath524 , for @xmath525 , can be recovered from @xmath162 as applied to variations of @xmath105 . here
we must assume that the field of coefficients is algebraically closed .
lemnearby cycles decomposition there exists a unique isomorphism of functors @xmath526 @xmath527 where for any constructible complex @xmath160 on @xmath16 , @xmath528 is nilpotent on @xmath529 .
simply pursue the line of reasoning in [ unipotent nearby cycles ] but , since the field of coefficients is algebraically closed , produce the full jordan decomposition rather than just the unipotent and non - unipotent parts .
the lemma can also be deduced from @xcite*lemme 3.2.5 , which applies to the jordan decomposition of an endomorphism of any complex in the derived category .
let @xmath530 be the local system of rank @xmath170 on @xmath205 with monodromy @xmath531 ; then clearly , we have @xmath532 , where @xmath157 acts as @xmath531 on the one - dimensional vector space @xmath533 . substituting into [ eq : nearby cycles decomposition ] , we obtain : @xmath534 thus , [ nearby cycles computations ] gives a procedure for computing the full nearby cycles functor of perverse sheaves , and @xmath535 $ ] sends perverse sheaves on @xmath16 to perverse sheaves on @xmath40 .
using some general reasoning , we can extend the properties of @xmath536 $ ] from the subcategory of perverse sheaves to the entire derived category . to this end , let @xmath537 be a triangulated functor between triangulated categories with t - structures , and let the respective cores be the abelian categories @xmath3 , @xmath538 .
we will assume that the objects of @xmath539 are _ bounded above _ , meaning that @xmath540 .
lemtruncation suppose @xmath172 is right t - exact and that @xmath541 ; then @xmath172 is t - exact .
we will show that @xmath172 commutes with all truncations .
suppose we have an object @xmath542 , so that there is a distinguished triangle @xmath543 where by definition , @xmath544 \in \cat{a}[-b]$ ] . by hypothesis on @xmath172
, we have @xmath545 , @xmath546 , and @xmath547 ) \in \cat{b}[-b ] \subset \cat{d}^{\geqslant b}$ ] .
since @xmath172 is triangulated , there is a triangle @xmath548 ) \to\ ] ] and therefore , by uniqueness of the truncation triangle , it must be that @xmath549 .
this is under the hypothesis that @xmath550 ; since then @xmath551 and since @xmath552 , we can apply truncations - by - one repeatedly and conclude that for all @xmath54 , we have @xmath553 .
now suppose we have any @xmath175 , and for any @xmath54 form the distinguished triangle @xmath554 to which we apply @xmath172 . since @xmath555 , the cone of the resulting triangle @xmath556 must be isomorphic to @xmath557 , by uniqueness of cones and the truncation triangle for @xmath558 .
thus , @xmath559 .
since then @xmath172 commutes with all trunctions , it is _ a fortiori _ t - exact .
take @xmath560 $ ] ; by [ nearby cycles are perverse ] , it satisfies the hypothesis of [ truncation ] , and therefore we conclude : thmnearby cycles are exact the functor @xmath535 $ ] on the bounded derived category @xmath561 is t - exact for the perverse t - structure . likewise , @xmath562 $ ]
is t - exact . for the second statement
, we must show that @xmath562 $ ] is right t - exact and preserves perverse sheaves ; the latter claim already follows from [ vanishing cycles and phi ] . for the former
, we apply the long exact sequence to the triangle @xmath563 we have @xmath564}$ ] because of triangle [ eq : extension triangle ] , and we already know that @xmath535 $ ] is right t - exact , so the long exact sequence of perverse cohomology shows that @xmath565 when @xmath217 , as desired .
we will not prove here that @xmath535 $ ] commutes with verdier duality .
this is significantly more difficult since it necessitates enlarging the domain of a certain natural transformation ( the map @xmath354 constructed in [ kernel equals cokernel ] ) from the core of the perverse t - structure to the entire derived category .
this involves the interaction with both objects and morphisms : * the natural maps must be defined for all objects , not just those in @xmath113 ; * the maps thus obtained must commute with all morphisms , not just those between objects of @xmath113 . to see why this is difficult , consider showing merely that the @xmath354 ( and their translates ) are natural with respect to maps of the form @xmath566}$ ] , with @xmath567 and @xmath568 .
note that the argument given for the naturality of @xmath354 is not valid in this context , since kernel and cokernel constructions in the abelian category of perverse sheaves are not functorial in the entire derived category .
if @xmath458 , this is easy ; we necessarily have @xmath569 $ ] , so rotating the distinguished triangle gives a short exact sequence @xmath570 \to \sh{m } \to 0.\ ] ] conversely , this sequence constructs the distinguished triangle @xmath571 \to \on{cone}(g)$ ] by the reverse procedure . then , applying @xmath572 and @xmath573 to the sequence , we find by naturality of @xmath354 that there is a commutative diagram of short exact sequences , which implies that @xmath354 is natural with respect to @xmath350 . the analogue of this argument for @xmath574 would involve finding a sequence of the form @xmath575 representing @xmath350 .
the manner in which such a sequence does represent such a map is clear ; we get a collection of short exact sequences representing maps : @xmath576}\end{aligned}\ ] ] ( where @xmath577 and @xmath578 ) , and thus , by composition , a map @xmath566}$ ] , as desired .
this is yoneda s realization of @xmath579 ; it holds in the derived category of @xmath113 .
it is , however , a nontrivial theorem , proved in @xcite , that this is the same as @xmath580 , and in fact it is describing the morphisms that occupies the entirety of the work in that paper .
of course , once we choose to cite this result , it is a trivial consequence of [ nearby cycles computations ] that @xmath535 $ ] commutes with @xmath2 , since it is then the derived functor of a self - dual _ exact _ functor on @xmath113 .
thus , we do not expect that there will be as elementary an argument as for the perversity of nearby cycles .
in the recent preprint @xcite , autoduality of the nearby cycles functor is proven in complete generality in the complex analytic setting , and references are given there for prior results and those in the algebraic setting .
we have used the term `` maximal extension functor '' without explanation ( as did beilinson ) , but [ xi sequences ] provides sufficient rationale : applying @xmath137 to the first one and @xmath581 to the second one , the long exact sequence of perverse cohomology shows that @xmath582 are both perverse sheaves , which is as far out ( cohomologically ) as they can be given that @xmath137 is right t - exact and @xmath581 is left t - exact .
this should be compared with the defining property of the `` minimal extension '' @xmath115 , that @xmath583 $ ] and @xmath584 $ ] are perverse , so that it has a minimal presence on @xmath40 given that it extends @xmath105 .
the condition that @xmath585 and @xmath586 are perverse does not uniquely characterize @xmath587 , as one could add any perverse sheaf supported on @xmath68 without changing it , but imposing [ xi sequences ] forbids such a modification . as we will see below ,
these sequences uniquely determine @xmath587 . to do so
, consider the pair of upper and lower `` caps '' of an octahedron : @xmath588 } \ar"1,1";"1,3"^{\alpha } \ar"1,1";"2,2"_{\alpha_- } \ar"2,2";"1,3"_{\alpha_+ } \ar"2,2";"3,1"_{\beta_- } \ar"3,1";"1,1"^{[1 ] } \ar"1,3";"3,3 " \ar"3,3";"2,2"_{\beta_+[1 ] } \ar"3,3";"3,1"^{(1 - t)[1 ] } \ar@{}"2,2";"1,2"|(0.67){c } \ar@{}"2,2";"2,3"|(0.67){d } \ar@{}"2,2";"3,2"|(0.67){c } \ar@{}"2,2";"2,1"|(0.67){d } } } & & \underset{\text{\large(lower cap)}}{\xymatrix@c-1ex@r+2ex { { j _ ! \sh{m } } & & { j _ * \sh{m } } \\ & { i^ * j _ * \sh{m}\vphantom{{\xi^\text{un}}_f(\sh{m } ) } } & \\ { { \psi^\text{un}}_f(\sh{m } ) } & & { { \psi^\text{un}}_f(\sh{m})[1 ] } \ar"1,1";"1,3"^{\alpha } \ar"1,3";"2,2 " \ar"2,2";"1,1"^{[1 ] } \ar"3,1";"1,1"^{[1 ] } \ar"3,1";"2,2 " \ar"1,3";"3,3 " \ar"2,2";"3,3 " \ar"3,3";"3,1"^{(1 - t)[1 ] } \ar@{}"2,2";"1,2"|(0.67){d } \ar@{}"2,2";"2,3"|(0.67){c } \ar@{}"2,2";"3,2"|(0.67){d } \ar@{}"2,2";"2,1"|(0.67){c } } } \end{aligned}\ ] ] the triangles marked `` c '' are commutative and those marked `` d '' are distinguished ; the arrows marked @xmath589 $ ] have their targets ( but not their sources ) shifted by @xmath170 .
the octahedral axiom states that given any diagram of commutative and distinguished triangles as in ( lower cap ) we can construct a diagram as in ( upper cap ) and vice versa ( @xcite*1.1.6 ) . using these diagrams , we can derive [ eq : psiun ] and [ xi sequences ] from each other .
this idea is also present in @xcite*5.7.2 .
propxi implies psiun suppose we have functors @xmath365 and @xmath227 from @xmath113 to @xmath88 , where @xmath227 has a unipotent action of @xmath146 , and satisfying [ xi sequences ] .
then [ eq : psiun ] holds with @xmath590 $ ] .
given [ xi sequences ] , each exact sequence there corresponds to a unique distinguished triangle in @xmath41 with the same entries ; these triangles appear in ( upper cap ) , where the top and bottom maps are @xmath591 and @xmath592 $ ] since the triangles containing them are commutative .
the octahedral axiom gives us ( lower cap ) , and since the upper triangle is distinguished its cone ( the middle term ) must necessarily be @xmath593 by [ eq : extension triangle ] . therefore the bottom triangle is [ eq : psiun ] , as desired .
note that all the interior maps in ( lower cap ) are uniquely determined , since they correspond to the kernels and cokernels of the maps @xmath591 and @xmath158 of perverse sheaves in the long exact sequence of cohomology .
proppsiun implies xi given only the triangle [ eq : psiun ] , both the functor @xmath365 and its extension classes in @xmath594 and @xmath595 can be constructed with [ xi sequences ] satisfied ( except for the duality statement ) . in particular , by [ xi implies psiun ]
, @xmath365 is uniquely determined by [ xi sequences ] . given [ eq : psiun ] ,
since we have [ eq : extension triangle ] canonically we can form all the vertices of ( lower cap ) and both distinguished triangles ; the left and right maps are determined by the requirement that the triangles containing them be commutative .
the octahedral axiom gives us ( upper cap ) and @xmath587 , identified at first only as an element of @xmath41 .
from [ nearby cycles are perverse ] we know that @xmath268 is perverse ; then the long exact sequence of perverse cohomology associated to either distinguished triangle in ( upper cap ) shows that , in fact , @xmath587 is perverse , and thus those triangles correspond to exact sequences as in [ xi sequences ] .
the equations @xmath596 and @xmath597 can then be read off from the commutativity of the upper and lower triangles . since the vertical arrows come from ( lower cap ) ,
these distinguished triangles are uniquely determined up to isomorphism fixing @xmath598 and @xmath268 , as desired .
the identity of @xmath365 is somewhat mysterious , but can be made precise using the gluing category .
these computations are also given in @xcite*example 5.7.8 .
propglued functors for any perverse sheaf @xmath220 , we have the following correspondences via the gluing construction : @xmath599 where @xmath600 is the map @xmath601 in the gluing category ; in @xmath115 , we mean @xmath602 ; in @xmath587 , taking @xmath603 , we have @xmath604 and @xmath605 . using the triangle of [
vanishing cycles and phi ] , we have @xmath606 and since @xmath607 , we get an isomorphism @xmath608 ; dualizing , we have @xmath609 also . since @xmath286 is the first map in this triangle , under this identification we have @xmath610 , and therefore @xmath611 since @xmath612 .
this gives the quadruple for @xmath613 ; for @xmath614 , we dualize , since @xmath286 and @xmath141 are dual by their definition in propositions and . that the natural map is given by @xmath601 follows from the fact that this does define a map @xmath615 in the gluing category , and that its restriction to @xmath16 is the identity . for @xmath115 , we use the fact that it is the image of the natural map @xmath600 ; having already identified all the parties , this is clear from the quadruples just obtained . for the identification of @xmath587 ,
obviously , @xmath396 ; more importantly , @xmath286 is injective and @xmath141 surjective
. then the pair of exact sequences in [ xi sequences ] can be described on quadruples as being trivial over @xmath16 , and over @xmath68 the maps @xmath403 and @xmath404 are described by the following maps of quadruples : @xmath616 we take @xmath617 and @xmath618 to be the maps whose @xmath68-parts ( the @xmath16-parts are zero ) are : @xmath619 then it is clear from the definitions of @xmath286 and @xmath141 that we obtain the sequences of [ xi sequences ] ; by the uniqueness part of [ psiun implies xi ] , this uniquely determines @xmath587 , completing the proof . since the entirety of [ sec : gluing ] follows only from [ xi sequences ] , propositions and
show that the constructions of [ sec : unipotent cycles ] are irrelevant for constructing the gluing functor .
their purpose , as is evident from the order we have chosen for the theorems , is to exhibit the autoduality of @xmath227 and @xmath365 ( and , thus , @xmath393 ) .
however , beilinson s development has an aesthetic virtue ( over just using the above short proof of [ xi sequences ] ) : once [ nearby cycles are perverse ] is proven , the entire theory takes place within the abelian category of perverse sheaves .
in addition , [ nearby cycles computations ] is an ingeniously elementary , insightful , and more useful definition of a functor whose actual definition is quite obscure . | the titular , foundational work of beilinson not only gives a technique for gluing perverse sheaves but also implicitly contains constructions of the nearby and vanishing cycles functors of perverse sheaves .
these constructions are completely elementary and show that these functors preserve perversity and respect verdier duality on perverse sheaves .
the work also defines a new , `` maximal extension '' functor , which is left mysterious aside from its role in the gluing theorem . in these notes
, we present the complete details of all of these constructions and theorems . in this paper we discuss alexander beilinson s `` how to glue perverse sheaves '' @xcite with three goals .
the first arose from a suggestion of dennis gaitsgory that the author study the construction of the unipotent nearby cycles functor @xmath0 which , as beilinson observes in his concluding remarks , is implicit in the proof of his key lemma 2.1 . here
, we make this construction explicit , since it is invaluable in many contexts not necessarily involving gluing .
the second goal is to restructure the presentation around this new perspective ; in particular , we have chosen to eliminate the two - sided limit formalism in favor of the straightforward setup indicated briefly in @xcite*4.2 for d - modules .
we also emphasize this construction as a simple demonstration that @xmath1 $ ] and verdier duality @xmath2 commute , and de - emphasize its role in the gluing theorem .
finally , we provide complete proofs ; with the exception of the key lemma , @xcite provides a complete program of proof which is not carried out in detail , making a technical understanding of its contents more difficult given the density of ideas .
this paper originated as a learning exercise for the author , so we hope that in its final form it will be helpful as a learning aid for others .
we do not intend it to supplant , but merely to supplement , the original , and we are grateful to beilinson for his generosity in permitting this .
the author would like to offer three additional thanks : to gaitsgory , who explained how this beautiful construction can be understood concretely , thus providing the basis for the perspective taken here ; to sophie morel , for confirming the author s understanding of nearby and vanishing cycles as presented below ; and to mark de cataldo , for his generous contribution of time and effort to the improvement of these notes . in order to maintain readability
, we will work with sheaves of vector spaces in the classical topology on complex algebraic varieties , except in the second part of [ sec : comments ] , where we will require the field of coefficients to be algebraically closed .
for the necessary modifications to tale sheaves , one should consult beilinson s paper : aside from the shift in definitions the only change is some tate twists . for the d - modules case , one should read sam lichtenstein s undergraduate thesis , @xcite , in which the two - sided limit construction is also given in detail . |
the aim of this article is to study the geometry of mildly singular fano varieties on which there is a prime divisor of picard number one . recall that a fano variety is a normal complex algebraic variety whose anticanonical divisor has some positive multiple which is an ample cartier divisor . a first related result is given by l. bonavero , f. campana and j. a. winiewski in the sequel of articles @xcite and @xcite , where the authors classified ( toric ) fano varieties of dimension @xmath0 on which there is a divisor isomorphic to @xmath1 and later used these results to study ( toric ) complex varieties whose blow - up at a point is fano .
for instance , in the toric case we have the following result .
let @xmath2 be a smooth toric fano variety of dimension @xmath0 .
then , there exists a toric divisor @xmath3 of @xmath2 isomorphic to @xmath1 if and only if one of the following situations occurs : 1 .
@xmath4 and @xmath3 is a linear codimension 1 subspace of @xmath2 .
2 . @xmath5 and @xmath3 is a fiber of the projection on @xmath6 .
@xmath7 , where @xmath8 , and @xmath3 is either the divisor @xmath9 or the divisor @xmath10 .
@xmath2 is isomorphic to the blow - up of @xmath11 along a linear @xmath12 contained in the divisor @xmath9 , where @xmath13 , and @xmath3 is either the strict transform of the divisor @xmath9 or the strict transform of the divisor @xmath14 .
in particular , this classification leads @xmath15 .
some years later , t. tsukioka in @xcite used some arguments from @xcite and @xcite to generalize these results and proved , more generally , that a smooth fano variety @xmath2 of dimension @xmath0 containing an effective prime divisor of picard number one must satisfy @xmath15 .
the bound @xmath15 was recently proved by g. della noce in ( * ? ? ?
* remark 5.5 ) , when @xmath2 is supposed to be a @xmath16-factorial gorenstein fano variety of dimension @xmath0 with canonical singularities , with at most finitely many non - terminal points , and under the more general assumption of the existence of an effective prime divisor @xmath17 such that the real vector space @xmath18 of numerical classes of 1-cycles on @xmath2 that are equivalent to 1-cycles on @xmath3 , is one - dimensional . in the smooth case , c. casagrande and s. druel provide in @xcite a classification ( and examples ) of all cases with maximal picard number @xmath19 .
let @xmath2 be a fano manifold of dimension @xmath0 and @xmath19 .
let @xmath20 be a prime divisor with @xmath21 .
then @xmath2 is isomorphic to the blow - up of a fano manifold @xmath22 along an irreducible subvariety of dimension @xmath23 contained in a section of the @xmath6-bundle @xmath24 , where @xmath25 is a fano manifold of dimension @xmath26 and @xmath27 .
firstly , we recall in [ section : extremalcontractions ] that a mildly singular fano variety @xmath2 always has an extremal ray @xmath28 whose intersection with a given effective divisor is positive .
the rest of [ section : extremalcontractions ] is devoted to the study of these extremal contractions in the case that the given divisor has picard number one .
this allows us to prove the following result in [ section : rho=3 ] .
[ thm : rho=3 ] let @xmath2 be a @xmath16-factorial gorenstein fano variety of dimension @xmath29 with canonical singularities and with at most finitely many non - terminal points .
assume that there exists an effective prime divisor @xmath20 such that @xmath30 and that @xmath19 .
then , there is a commutative diagram @xmath31_{\widehat{\sigma } } \ar[rd]^{\sigma } \ar[dd]^\varphi & \\ \widehat{y } \ar[rd]_{\widehat{\pi } } & & y \ar[ld]^{\pi } \\ & z & } \ ] ] where @xmath32 ( resp .
@xmath33 ) corresponds to a divisorial contraction of an extremal ray @xmath34 ( resp .
@xmath35 ) which is the blow - up in codimension two of an irreducible subvariety of dimension @xmath23 , and @xmath36 is a contraction of fiber type , finite over @xmath3 , corresponding to the face @xmath37 .
moreover , @xmath38 , @xmath39 and @xmath40 are @xmath16-factorial varieties with canonical singularities and with at most finitely many non - terminal points , @xmath39 is fano and @xmath25 is a @xmath16-factorial fano variety with rational singularities .
the results of s. cutkosky on the contractions of terminal gorenstein threefolds @xcite , together with the previous result imply the following corollary .
[ thm : threefold ] let @xmath2 be a @xmath16-factorial gorenstein fano threefold with terminal singularities .
assume that there exists an effective prime divisor @xmath20 such that @xmath21 and that @xmath19 .
then , @xmath2 is factorial and it can be realized as the blow - up of a smooth fano threefold @xmath39 along a locally complete intersection curve @xmath41 .
moreover , @xmath39 is isomorphic to @xmath42 , where @xmath43 . in the case
@xmath44 , we obtain in [ section : rho=2 ] the following extension of ( * ? ? ?
* remark 3.2 and proposition 3.3 ) to mildly singular fano varieties @xmath2 with @xmath44 , on which there is an effective prime divisor of picard number one .
[ thm : rho=2,2nd contr ] let @xmath2 be a @xmath16-factorial gorenstein fano variety of dimension @xmath29 with canonical singularities and with at most finitely many non - terminal points .
assume that there exists an effective prime divisor @xmath20 such that @xmath30 and that @xmath44 .
there are two possibilities : 1 .
if @xmath3 is not nef , then there is an extremal contraction sending @xmath3 to a point .
if @xmath3 is nef , then @xmath45 is an extremal ray .
one of the following assertions must hold : * @xmath46 is of fiber type onto @xmath6 , and @xmath3 is a fiber .
* @xmath46 is a divisorial contraction sending its exceptional divisor @xmath47 to a point , and such that @xmath48 .
* @xmath46 is a small contraction and there is a flip @xmath49 and a contraction of fiber type @xmath50 such that the general fiber is isomorphic to @xmath6 , with anticanonical degree 2 .
moreover , @xmath51 is finite over the strict transform of @xmath3 in @xmath52 . in order to extend the classification results to higher dimensions
, we will restrict ourselves to the case of toric varieties . in that case , the combinatorial description of the mmp for toric varieties treated in [ section : toricmmp ] , as well as some particular properties of them , will allow us to prove the following result in [ section : rho=3toric ] .
[ thm : rho=3toric ] let @xmath2 be a @xmath16-factorial gorenstein toric fano variety of dimension @xmath29 with canonical singularities and with at most finitely many non - terminal points .
assume that there exists an effective prime divisor @xmath20 such that @xmath30 and that @xmath19 .
then , there exist @xmath16-factorial gorenstein toric fano varieties @xmath39 and @xmath25 , with terminal singularities , such that 1 .
@xmath53 , the normalized blow - up of an invariant toric subvariety @xmath54 of dimension @xmath23 ; and 2 .
@xmath22 , where @xmath43 and @xmath55 is the ample generator of @xmath56 . in the toric setting
, we obtain in [ section : rho=2toric ] results that extend l. bonavero s description of the extremal contractions in the case @xmath44 to mildly singular toric fano varieties .
if @xmath2 is supposed to have isolated canonical singularities then we obtain the following classification .
[ thm : rho=2
toric div isolated ] let @xmath2 be a @xmath16-factorial gorenstein toric fano variety of dimension @xmath0 with isolated canonical singularities .
assume that there exists an effective prime divisor @xmath20 such that @xmath21 and that @xmath44 .
then , either 1 . @xmath57 for some toric variety @xmath39 .
moreover , @xmath39 is a @xmath16-factorial gorenstein toric fano variety of dimension @xmath26 with terminal singularities and fano index @xmath58 , and @xmath59 . in particular
, @xmath2 has only terminal singularities .
@xmath2 is isomorphic to the blow - up of a toric variety @xmath39 along an invariant subvariety @xmath54 of dimension @xmath23 , contained in the smooth locus of @xmath39 .
moreover , @xmath39 is isomorphic to either * @xmath60 , * @xmath61 if @xmath62 is even , or * @xmath63 , where @xmath64 are two relatively prime integers such that @xmath65 and @xmath66 .
+ in particular , @xmath39 is a @xmath16-factorial gorenstein fano variety with @xmath67 and it has at most two singular points .
conversely , the blow - up of any of the listed varieties @xmath39 along an invariant irreducible subvariety @xmath54 of dimension @xmath23 and contained in the smooth locus of @xmath39 , leads to a toric variety @xmath2 satisfying the hypothesis .
moreover , in the case of contractions of fiber type we obtain the following result without the assumption of isolated singularities .
[ thm : rho=2 toric fibration ] let @xmath2 be a @xmath16-factorial gorenstein toric fano variety of dimension @xmath0 with canonical singularities and with at most finitely many non - terminal points .
assume that there exists an effective prime divisor @xmath20 such that @xmath21 and that @xmath44 .
let @xmath28 be an extremal ray such that @xmath68 and assume that the corresponding extremal contraction @xmath69 is of fiber type .
then , @xmath57 . moreover , @xmath39 is a @xmath16-factorial gorenstein fano variety of dimension @xmath26 with terminal singularities and fano index @xmath58 , and @xmath59 .
in particular , @xmath2 has only terminal singularities . finally
, [ section : toric1coverings ] is devoted to show that the extremal contractions studied in [ section : rho=2toric ] lift to universal coverings in codimension 1 , introduced by w. buczyska in @xcite .
see definition [ pws ] for the notion of poly weighted space ( pws ) , introduced by m. rossi and l. terracini in @xcite and proved to be universal covering spaces in codimension 1 for @xmath16-factorial toric varieties .
in particular , we obtain the following description of divisorial contractions of toric mildly fano varieties with picard number two .
it should be noticed that even if the combinatorial description of these divisorial contractions is very simple ( see lemma [ toricdivisorial ] ) and it coincides with the one of the blow - up of a subvariety of dimension @xmath23 in the smooth case , it may happen that the morphisms are not globally a blow - up of the coherent sheaf of ideals of a ( irreducible and reduced ) subvariety but only a blow - up in codimension two if the singularities are not isolated ( see example [ contreexample ] ) .
[ prop:1-covering , divisorial ] let @xmath2 be a @xmath16-factorial gorenstein toric fano variety of dimension @xmath0 with canonical singularities and with at most finitely many non - terminal points .
assume that there exists an effective prime divisor @xmath20 such that @xmath21 and that @xmath44 .
let @xmath28 be an extremal ray such that @xmath68 and let us denote by @xmath69 the corresponding extremal contraction .
assume that @xmath70 is birational .
then there exist weights @xmath71 and a cartesian diagram of toric varieties @xmath72^{\widehat{\pi}\hspace{1 cm } } \ar[d]_{\pi_x } & \mathbb{p}(\lambda_0,\ldots,\lambda_{n } ) \ar[d]^{\pi_y } \\ x \ar[r]^\pi & y } \ ] ] where vertical arrows denote the corresponding canonical universal coverings in codimension 1 , and @xmath73 is a gorenstein fano pws with canonical singularities and with at most finitely many non - terminal points such that @xmath74 .
moreover , @xmath75 is a divisorial contraction sending its exceptional divisor @xmath76 onto an invariant subvariety @xmath77 of dimension @xmath23 .
i would like to thank my supervisors , stphane druel and catriona maclean , for their unstinting help and their valuable guidance .
through out this article all varieties will be assumed to be reduced and irreducible schemes of finite type over the field @xmath78 of complex numbers .
its smooth locus will be denoted by @xmath79 , while @xmath80 denotes its singular locus .
we will use the notation and results of the minimal model program ( mmp for short ) in @xcite .
let @xmath2 be a normal projective variety , we consider @xmath81 where @xmath82 denotes the numerical equivalence .
let @xmath83 be the convex cone generated by the classes of effective 1-cycles , i.e. , @xmath84cycles with non - negative coefficients , and @xmath85 be its topological closure .
we denote by @xmath86 $ ] the class of @xmath87 in @xmath88 .
let @xmath89 be a closed subset and @xmath90 be the natural inclusion , we define @xmath91 for us , a divisor will always be a weil divisor .
let us denote by @xmath92 the class of a canonical divisor in @xmath93 . a complete normal variety @xmath2 is said to be a fano variety if there exists a positive multiple of @xmath94 which is cartier and ample .
a fano variety is therefore always projective .
the variety @xmath2 is said to be gorenstein if @xmath92 is a cartier divisor and its singularities are cohen - macaulay .
the property of being gorenstein is local and open , so the gorenstein locus of @xmath2 is the open subset containing all the gorenstein points of @xmath2 ( it contains @xmath95 , in particular ) .
we follow the usual convention , and we say that @xmath2 is a @xmath16-gorenstein variety if some positive multiple of @xmath92 is a cartier divisor ; we do not require cohen - macaulay singularities . in this case , the gorenstein index of @xmath2 is the smallest positive integer @xmath96 such that @xmath97 is a cartier divisor . in the same way as for @xmath84cycles , we can define @xmath98 as the vector space of @xmath16-cartier divisors with real coefficients , modulo numerical equivalence @xmath82 .
we denote by @xmath99 $ ] the class of @xmath3 in @xmath98 .
we have a non - degenerated bilinear form @xmath100 @xmath101,[c])\longmapsto d\cdot c\ ] ] given by the intersection product between curves and divisors in @xmath2 .
then , @xmath102 is called the picard number of @xmath2 .
it is a classical fact that @xmath103 is finite .
an extremal ray @xmath104 of @xmath85 is a 1-dimensional subcone such that if @xmath105 , @xmath106 imply that @xmath107 .
we denote @xmath108\in r}c \subseteq x.\ ] ] if @xmath3 is a @xmath16-cartier divisor and @xmath28 is a extremal ray , then the sign @xmath38 ( resp .
@xmath109 ) is well defined . from now on , all the varieties are assumed to be @xmath16-factorial unless explicitly mentioned . a contraction of @xmath2 is a projective surjective morphism @xmath110 with connected fibers , where @xmath39 is a normal projective variety .
in particular , @xmath111 .
let us recall the notion of singularities of pairs for @xmath16-factorial varieties ( see @xcite for details ) .
let @xmath2 be a @xmath16-factorial normal projective variety , and @xmath112 an effective @xmath16-divisor on @xmath2 .
let @xmath113 be a log - resolution of the pair @xmath114 , i.e. , a birational projective morphism @xmath115 whose exceptional locus @xmath116 is the union of the effective prime divisors @xmath117 s and such that @xmath118 is a simple normal crossing divisor , where @xmath119 is the strict transform of @xmath120 in @xmath39 by @xmath115 .
using numerical equivalence , we have @xmath121 the numbers @xmath122 are independent of the log - resolution and depend only on the discrete valuation that corresponds to @xmath117 .
+ suppose that all @xmath123 and that @xmath115 is a log - resolution of @xmath114 , we say that the pair @xmath114 is @xmath124 here klt means kawamata log terminal .
if the conditions above hold for one log - resolution of @xmath114 , then they hold for every log - resolution of @xmath114 .
+ we say that @xmath2 is terminal ( canonical , ... ) or that it has terminal ( canonical , ... ) singularities if @xmath125 is a terminal ( canonical , ... ) pair . from the cone theorem ( see ( * ? ?
* chapter 3 ) ) , if @xmath2 is klt then for each extremal face @xmath126 there exists a unique morphism @xmath127 with connected fibers , called the extremal contraction of @xmath128 , from @xmath2 onto a normal projective variety @xmath129 such that the irreducible curves contracted by @xmath130 to points are exactly the curves whose classes in @xmath88 belongs to @xmath128 .
we denote by @xmath131 the exceptional locus of @xmath130 , i.e. , the subset of @xmath2 where @xmath130 is not an isomorphism .
there are three possibilities : * @xmath132 ; we say that @xmath130 is of fiber type .
* @xmath130 is birational and its exceptional locus is an effective divisor @xmath133 ( prime if @xmath128 is an extremal ray ) such that @xmath134 ; we say that @xmath130 is a divisorial contraction .
* @xmath130 is birational and its exceptional locus has codimension @xmath135 in @xmath2 ; we say that @xmath130 is a small contraction .
if @xmath2 is klt and @xmath28 is a @xmath136negative extremal ray , then there is a short exact sequence @xmath137 & { \operatorname{pic}}x_r \ar[rr]^{l\mapsto { \operatorname{cont}}_r^ * l } & & { \operatorname{pic}}x \ar[rr]^{m\mapsto m\cdot [ c ] } & & \mathbb{z } , } \ ] ] where @xmath138 is a curve whose numerical class generates the extremal ray @xmath104 . in the case
when @xmath2 is a klt fano variety we have a more precise description : the mori cone of @xmath2 is finite rational polyhedral , generated by classes of rational curves ; in particular @xmath139 .
moreover , kawamata - viehweg theorem implies that @xmath140 for @xmath141 .
after the works of birkar , cascini , hacon and mckernan , if @xmath2 is a klt fano variety , then @xmath2 is a mori dream space ( see ( * ? ? ?
* corollary 1.3.2 ) and @xcite ) . for us , a @xmath142bundle ( or fibration in @xmath6 ) is a smooth morphism all of whose fibers are isomorphic to @xmath6 .
let @xmath143 be a vector bundle over a variety @xmath2 , we denote by @xmath144 the scheme @xmath145 .
finally , we recall that if @xmath146 are positive integers with @xmath147 then we define the associated weighted projective space ( wps ) to be @xmath148 where @xmath149 is the equivalence relation @xmath150
moreover , @xmath151 is a toric variety with picard number one and torsion - free class group . in general
, we say that a @xmath16-factorial complete toric variety is a poly weighted space ( pws ) if its class group is torsion - free .
in this section we study extremal contractions of mildly singular fano varieties @xmath2 that admits an effective prime divisor @xmath20 such that @xmath21 .
firstly , notice that klt fano varieties always have an extremal ray whose intersection with a given effective prime divisor is positive ( c.f .
* lemme 2 ) ) . [ rexists ]
let @xmath2 be a @xmath16-factorial klt fano variety and let @xmath20 be an effective prime divisor .
then , there exists an extremal ray @xmath28 such that @xmath68 . the cone theorem ( * ? ?
* theorem 3.7 ) implies that @xmath139 is a rational polyhedral cone generated by a finite number of extremal rays @xmath152 .
let @xmath153 be any curve such that @xmath154>0 $ ] . since @xmath87 is numerically equivalent to a positive sum of extremal curves , @xmath86=\sum_{i=1}^s a_i [ c_i]$ ] with @xmath155 and @xmath156\in r_i$ ] , we can pick one such that @xmath157 .
secondly , we have that the contraction of an extremal ray whose intersection with an effective prime divisor of picard number one is positive has at most one - dimensional fibers .
[ fibersdim1 ] let @xmath2 be a @xmath16-factorial klt variety and @xmath20 be an effective prime divisor such that @xmath21 .
let us suppose that @xmath158 and that there exists @xmath159 extremal ray such that @xmath38 .
then , @xmath160 .
in particular , the extremal contraction @xmath161 is finite on @xmath3 and all the fibers of @xmath161 are at most of dimension 1 .
we follow the proof of ( * ? ? ?
* lemma 3.1 ) .
assume , to the contrary , that @xmath162 .
the numerical class of every irreducible curve @xmath163 must belong to @xmath104 , since @xmath21 .
thus , @xmath161 sends @xmath3 to a point and @xmath164 .
if @xmath161 is birational , then we would have that @xmath165 is an effective divisor by the negativity lemma , a contradiction .
the contraction @xmath161 is then of fiber type and @xmath166 .
as @xmath167 and @xmath168 is a point , there exists a non - trivial fiber of @xmath161 disjoint of @xmath3 , and then @xmath169 , a contradiction . as a consequence we have that all the fibers of @xmath161 are at most of dimension 1 .
in fact , @xmath170 is a finite morphism : if there exists a curve @xmath163 contained in a fiber , we would have that @xmath86\in r$ ] and therefore @xmath162
. we will need the following result .
[ extcont ] let @xmath2 be a @xmath16-factorial klt fano variety of dimension @xmath0 .
let us suppose that @xmath167 and that there exists an effective prime divisor such that @xmath21 .
let @xmath28 be an extremal ray such that @xmath38 and let us denote by @xmath171 the corresponding extremal contraction .
then , 1 .
if @xmath172 is of fiber type , then @xmath173 is a @xmath16-factorial fano variety with rational singularities of dimension @xmath26 such that @xmath174 .
2 . if @xmath172 is birational and we suppose that @xmath2 has gorenstein canonical singularities with at most finitely many non - terminal points , then @xmath172 is a divisorial contraction and there exists a closed subset @xmath175 with @xmath176 such that @xmath177 , @xmath178 , @xmath179 and @xmath180 is the blow - up of a @xmath23-dimensional smooth subvariety in @xmath181 .
moreover , @xmath173 is a @xmath16-factorial fano variety with canonical singularities with at most finitely many non - terminal points .
in particular , @xmath182=e\cdot [ f]=-1,\ ] ] for every irreducible curve @xmath128 such that @xmath183\in r$ ] , where @xmath184 is the exceptional divisor of @xmath172 .
let us suppose that @xmath171 is of fiber type .
then , @xmath185 is a finite morphism , by lemma [ fibersdim1 ] , and thus @xmath186 . since @xmath173 is a @xmath16-factorial projective variety , the projection formula and the fact that @xmath21 imply that @xmath174 . in particular , every big divisor on @xmath173 is ample . let us prove that @xmath173 is a fano variety in this case . since @xmath2 is fano klt , there exists an effective @xmath16-divisor @xmath187 on @xmath173 such that @xmath188 is klt and such that @xmath189 is ample , by ( * ? ? ?
* lemma 2.8 ) .
in particular , @xmath173 has rational singularities and @xmath190 is a big divisor , and therefore ample . let us suppose now that @xmath171 is a birational contraction .
then , it follows from ( * ? ? ?
* theorem 2.2 ) that @xmath172 is a divisorial contraction given by the blow - up in codimension two of an irreducible subvariety of dimension @xmath23 on @xmath173 , and that @xmath173 is a @xmath16-factorial fano variety with canonical singularities with at most finitely many non - terminal points .
finally , the ampleness of the anticanonical divisor @xmath190 follows verbatim the proof in the smooth case given in ( * ? ? ?
* lemma 3.1 ) .
[ rholeq3 ] let @xmath2 be a gorenstein fano variety of dimension @xmath0 with canonical singularities with at most finitely many non - terminal points , and let @xmath20 be an effective prime divisor such that @xmath21 .
then , @xmath15 , by ( * ? ? ?
* remark 5.5 ) .
[ fibersp1 ] let @xmath2 be a @xmath16-factorial klt variety of dimension @xmath0 and let @xmath171 be a contraction of fiber type , associated to an extremal ray @xmath28 , such that all the fibers are one - dimensional . then ( * ? ?
* corollary 1.9 ) implies that all fibers of @xmath172 are connected and their irreducible components are smooth rational curves .
let us denote by @xmath191 $ ] the numerical class of the funtamental 1-cycle associated to the scheme theoretic fiber @xmath192 over the point @xmath193 , and by @xmath194}(x)]$ ] the numerical class of the cycle theoretic fiber over @xmath193 in the sense of j. kollr ( * ? ? ?
* i.3 ) .
since @xmath2 is @xmath16-gorenstein , the intersection number @xmath195}(x)]$ ] is independent of the closed point @xmath193 , by ( * ? ? ?
* proposition i.3.12 ) . in our context ( terminal gorenstein threefolds or toric varieties )
, @xmath171 will be additionally flat and hence the cycle theoretic fiber will coincide with the fundamental class of the scheme theoretic fiber ( see ( * ? ? ?
* definition i.3.9 and corollary i.3.15 ) ) . under the flatness hypothesis , if we suppose moreover that @xmath92 is a cartier divisor then we will have that @xmath196=2 $ ] for all @xmath193 , by generic smoothness .
we have therefore three possibilities in that case : 1 .
@xmath197 is an irreducible and generically reduced rational curve such that @xmath198 and that @xmath196=2 $ ] .
2 . @xmath191=2[c]$ ] as 1-cycles , where @xmath87 is an irreducible and generically reduced rational curve such that @xmath199 and that @xmath200 = 1 $ ] .
@xmath201 , with @xmath202 irreducible and generically reduced rational curves such that @xmath203 and that @xmath200 = -k_x\cdot [ c ' ] = 1 $ ] .
let us finish the section with two results concerning birational extremal contractions .
[ specialmoriprogram ] let @xmath2 be a @xmath16-factorial fano variety with canonical singularities .
then , for any prime divisor @xmath20 , there exists a finite sequence ( called a special mori program for the divisor @xmath165 ) @xmath204 such that , if @xmath205 is the transform of @xmath3 for @xmath206 and @xmath207 , the following hold : 1 . @xmath208 and @xmath39
are @xmath16-factorial projective varieties and @xmath209 have canonical singularities .
2 . for every @xmath210
, there exists an extremal ray @xmath211 of @xmath212 with @xmath213 and @xmath214 such that : * for @xmath215 , @xmath216 , and @xmath217 is either the contraction of @xmath211 ( if @xmath211 is divisorial ) , or its flip ( if @xmath211 is small ) ; * the morphism @xmath218 is the contraction of @xmath219 and @xmath51 is a fiber type contraction .
the following result is a particular case of ( * ? ? ?
* lemma 3.3 ) .
we include the statement with our notation for completeness .
[ gorlocus ] let @xmath220 be a @xmath16-factorial gorenstein fano variety of dimension @xmath0 with canonical singularities and with at most finitely many non - terminal points .
let @xmath221 be an effective prime divisor and let us suppose that there is a diagram @xmath222 where @xmath217 is the birational map associated to the contraction of an extremal ray @xmath223 such that @xmath213 , for @xmath224 ; as in theorem [ specialmoriprogram ] .
if @xmath225 for @xmath226 , then both @xmath227 and @xmath228 are divisorial contractions , @xmath229 is contained in the gorenstein locus of @xmath212 and @xmath230 is disjoint from the transform of @xmath231 in @xmath220 .
in this section we study the extremal contractions of mildly singular fano varieties @xmath2 on which there is an effective prime divisor of picard number one and such that @xmath19 . as we pointed out in remark [ rholeq3 ] , this is the largest possible picard number for such varieties .
compare with the smooth case ( * ? ? ?
* lemma 3.1 and theorem 3.8 ) .
since @xmath2 is a @xmath16-factorial klt fano variety , there is an extremal ray @xmath28 such that @xmath38 , by lemma [ rexists ] .
we denote by @xmath232 the associated extremal contraction .
we note that proposition [ extcont ] implies that @xmath32 is a divisorial contraction sending the effective prime divisor @xmath233 onto a subvariety @xmath54 of dimension @xmath23 .
moreover , @xmath39 is a fano variety with canonical singularities and with at most finitely many non - terminal points , such that if we denote by @xmath234 the image of @xmath3 by @xmath32 , then we have that @xmath235 and that @xmath236 .
since @xmath39 is @xmath16-factorial klt fano variety , there is an extremal ray @xmath237 such that @xmath238 , by lemma [ rexists ] .
we denote by @xmath239 the associated extremal contraction .
let us prove that @xmath70 is of fiber type .
assume , to the contrary , that @xmath70 is a birational contraction .
hence , lemma [ gorlocus ] implies that both @xmath70 and @xmath240 are divisorial contractions , and that the exceptional locus @xmath241 consists of two disjoint effective prime divisors .
since @xmath242 we have that @xmath243>0 $ ] for every irreducible curve @xmath244 .
in particular , @xmath245 , as @xmath246 , contradicting the fact that the exceptional divisors are disjoint .
let @xmath35 be the extremal ray such that @xmath247 .
then , @xmath36 can be factorized as @xmath248 , where @xmath249 and @xmath250 .
since @xmath36 has fibers of dimension 1 , both contractions must have fibers of dimension at most 1 .
notice that the general fiber of @xmath36 is not contracted by @xmath33 , hence @xmath33 must be divisorial and @xmath251 a contraction of fiber type , by the same arguments as above .
the results of s. cutkosky on the contractions of terminal gorenstein threefolds and theorem [ thm : rho=3 ] above lead to corollary [ thm : threefold ] . by theorem [ thm :
rho=3 ] , there is a diagram @xmath252^{\sigma } \ar@/_1pc/[rr]_{\varphi } & y \ar[r]^{\pi } & z},\ ] ] where @xmath253 is a divisorial contraction sending a prime divisor @xmath254 onto a curve @xmath255 , and @xmath70 and @xmath36 are both extremal contractions of fiber type whose fibers are of dimension 1 .
all these varieties are @xmath16-factorial fano varieties , and @xmath39 has terminal singularities .
moreover , @xmath2 is factorial by ( * ? ? ?
* lemma 2 ) . by (
* lemma 3 and theorem 4 ) , @xmath41 is an irreducible reduced curve which is a locally complete intersection , @xmath39 is a factorial threefold which is smooth near the curve @xmath87 , and @xmath253 is the blow - up of the ideal sheaf @xmath256 . in particular
, @xmath39 is a gorenstein fano threefold with terminal singularities and therefore ( * ? ? ?
* theorem 7 ) implies that @xmath25 is a smooth del pezzo surface and @xmath24 is a ( possibly singular ) conic bundle over @xmath25 .
we note that @xmath27 and hence @xmath257 .
let @xmath258 . since @xmath70 is finite on @xmath259 and @xmath244 , we have that @xmath260 is an effective prime divisor on @xmath261 , which is therefore ample .
let us denote by @xmath262 the locus of points of @xmath261 over which @xmath70 is not a smooth morphism . by ( * ?
* proposition ii.1.1 ) , is a closed subset of @xmath261 .
then , @xmath262 has pure codimension 1 on @xmath261 or @xmath263 , by ( * ? ? ?
* theorem 3 ) .
let us suppose that @xmath262 is not empty .
if we take @xmath264 and we denote by @xmath265 its fiber by @xmath70 , then we have that @xmath266 ( as @xmath267 ) and that the 1-cycle on @xmath39 associated to @xmath268 is of the form @xmath269=[c]+[c']$ ] , where @xmath87 and @xmath270 are ( possibly equal ) irreducible and generically reduced rational curves such that @xmath271 ( as @xmath272 ) , by remark [ fibersp1 ] .
thus , if we denote by @xmath273 the strict transform of @xmath268 on @xmath2 by @xmath32 , we will have @xmath274\geq 3 $ ] , contradicting the fact that the anticanonical degree of every fiber of @xmath240 is 2 ( see remark [ fibersp1 ] ) .
we conclude in this way that @xmath275 is a @xmath6-bundle , and then @xmath39 is an smooth threefold by ( * ? ? ?
* theorem 5 ) .
finally , let us notice that if @xmath234 is not nef then there is a birational contraction @xmath276 sending @xmath234 to a point , by ( * ? ? ?
* remark 3.2 ) , and hence ( * ? ? ?
* lemma 3.9 ) implies that @xmath277 , with @xmath43 . on the other hand ,
if @xmath234 is nef then we apply ( * ? ? ?
* proposition 3.3 ) to conclude that either there is a divisorial contraction @xmath276 sending an effective prime divisor @xmath278 to a point , or there is a contraction of fiber type @xmath279 ; small contractions are excluded since @xmath2 is a fano variety , by the same argument ( 3.3.8 ) in @xcite . in the first case ( * ? ? ?
* lemma 3.9 ) allows us to conclude , while in the second case we have that @xmath280 , by ( * ? ? ?
* lemma 4.9 ) .
in the case @xmath44 we can describe the extremal contraction associated to the other extremal ray in the mori cone of @xmath2 ( compare with ( * ? ? ? * proposition 3.3 ) ) .
we will need the following result of g. v. ravindra and v. srinivas ( see @xcite ) .
[ lefs ] let @xmath2 be a complex projective variety , smooth in codimension 1 and let @xmath281 be a ample line bundle over @xmath2 . if @xmath282 is a linear subspace which gives a base point free linear system @xmath283 on @xmath2 , then there is a dense zariski open set of divisors @xmath284 such that the restriction map @xmath285 is an isomorphism , if @xmath286 , and is injective , with finitely generated cokernel , if @xmath287 .
[ coro1 ] let @xmath2 be a @xmath16-factorial fano variety of dimension @xmath0 with klt singularities .
if @xmath167 and @xmath3 is an effective prime divisor such that @xmath21 , then @xmath3 is not an ample @xmath16-divisor .
assume , to the contrary , that @xmath3 is an ample @xmath16-divisor .
let @xmath288 such that @xmath289 is a very ample cartier divisor on @xmath2 and use the complete linear system @xmath290 to embed @xmath291 .
let us define the projective incidence variety @xmath292)\in x\times \mathbb{p}(v)\;|\;x\in e\ } \subseteq x \times \mathbb{p}(v),\ ] ] and let @xmath293 be the second projection .
let @xmath294 be the relative hilbert scheme of curves associated to the morphism @xmath293 , which is a projective scheme with countably many irreducible components .
let us denote by @xmath295 the image of the components of @xmath294 that do not dominate @xmath296 .
they are closed subsets of @xmath296 .
thus , if we take @xmath297\in \mathbb{p}(v)-\cup z_i$ ] ( a very general point on @xmath296 ) , then for every curve @xmath298 on @xmath133 there is a dominant component of @xmath294 such that @xmath298 is one the curves parametrized by this component . since the image of this dominant component is in fact the whole projective space @xmath296
, then we have that there is a curve @xmath299 on @xmath3 which is also parametrized for this component .
in particular , @xmath298 and @xmath299 are numerically equivalent .
but , since there is only one curve on @xmath3 up to numerical equivalence , we will have @xmath300 . on the other hand ,
theorem [ lefs ] implies that we can also suppose that this very general divisor @xmath301 is chosen in such a way the restriction @xmath285 is injective .
since @xmath167 and @xmath300 , we have that the inclusion @xmath302 is not surjective . therefore , the induced map on the dual spaces obtained by restriction @xmath303 is not injective .
since @xmath2 is a @xmath16-factorial klt fano variety , we have that numerical and linear equivalence coincide , by ( * ? ? ?
* lemma 2.5 ) .
hence , we have the following diagram @xmath304 & \operatorname{pic}(x)\otimes_{\mathbb{z}}\mathbb{r } \ar@{_{(}->}[l]_{\cong } \ar@{->>}[r]^{\hspace{3mm}\cong } \ar@{^{(}->}[d ] & { \operatorname{n}}^1(x ) \ar@{->>}[d ] \\
\operatorname{cl}(e)\otimes_{\mathbb{z}}\mathbb{r } & \operatorname{pic}(e)\otimes_{\mathbb{z}}\mathbb{r } \ar@{_{(}->}[l ] \ar@{->>}[r ] & { \operatorname{n}}^1(e ) } \ ] ] therefore , @xmath303 is not injective if and only if @xmath305 is not injective .
let us consider a line bundle @xmath281 on @xmath133 such that @xmath306 .
notice that @xmath133 has klt singularities since it is a general member of the ample linear system @xmath290 , by ( * ? ? ?
* lemma 5.17 ) .
in particular , @xmath133 has rational singularities and thus if we consider a resolution of singularities @xmath307 , then the leray spectral sequence leads to @xmath308 and @xmath309 for all @xmath310 . on the other hand , the short exact sequence of sheaves @xmath311 and the vanishing @xmath312 for @xmath313 , give us @xmath314 .
since @xmath2 is a normal variety with cohen - macaulay singularities , serre s duality implies @xmath315 so , by taking @xmath316 large enough at the beginning if necessary , we can suppose that @xmath317 by the kawamata - viehweg vanishing theorem .
we get that @xmath318 and hence an inclusion @xmath319 .
clearly @xmath320 and thus @xmath321 . by the projection formula , @xmath322 , a contradiction .
we end this section by proving theorem [ thm : rho=2,2nd contr ] .
if @xmath3 is not nef , then there exists an extremal ray @xmath323 such that @xmath324 , and therefore @xmath325 . since @xmath21
we must have that @xmath326 and that @xmath168 is a point . if @xmath3 is nef , then it is not ample by corollary [ coro1 ] , and thus @xmath45 is an extremal ray , as @xmath44 .
if we denote @xmath327 , the proof follows almost verbatim the proof in the smooth case given in ( * ? ? ?
* proposition 3.3 ) , up to modify the argument ( 3.3.6 ) in @xcite by using lemma [ gorlocus ] to get a contradiction if we suppose that @xmath51 is a birational contraction .
we will analyze now the toric case .
we may refer the reader to @xcite for the general theory of toric varieties and to @xcite for details of the toric mmp .
we will keep the same notation as @xcite .
let @xmath328 be a lattice , @xmath329 its dual lattice and let @xmath330 ( resp .
@xmath331 ) and @xmath332 ( resp .
@xmath333 ) be their real scalar ( resp .
rational scalar ) extensions .
let us denote by @xmath334 the natural @xmath335bilinear pairing .
let @xmath336 be a fan .
as we will see , most of the properties of our interest in the context of the mmp of the associated toric variety @xmath337 , can be translated into combinatorial properties of the fan @xmath338 .
sometimes we will write @xmath339 instead of @xmath340 in order to emphasize the dependence of @xmath341 on the lattice where primitive generators of @xmath338 belong .
let @xmath342 be the set of @xmath343 dimensional cones in @xmath338 .
in the same way , if @xmath344 is a cone , we will denote by @xmath345 the set of its @xmath346dimensional faces .
usually , we will not distinguish between @xmath84dimensional cones @xmath347 ( or @xmath84dimensional faces @xmath348 ) and the primitive vector @xmath349 generating them . if @xmath350 we will denote by @xmath351 the associated affine toric variety , and by @xmath352 the closed invariant subvariety of codimension @xmath343 .
in particular , each @xmath347 corresponds to an invariant weil divisor @xmath353 on @xmath2 ; such a cone is called a _
ray_. similarly , each cone of codimension 1 @xmath354 corresponds to an invariant rational curve on @xmath2 ; such a cone is called a _
wall_. it is a classical fact that every weil divisor on a toric variety @xmath2 is linearly equivalent to an invariant toric divisor ( see ( * ? ? ?
* theorem 4.1.3 ) ) .
the same holds for effective curves on complete toric varieties .
let @xmath337 be a complete toric variety of dimension @xmath62
. then @xmath355.\ ] ] in particular , @xmath356 is a closed rational polyhedral cone and it is strictly convex if and only if @xmath2 is a projective variety .
all affine toric varieties associated to strongly convex rational polyhedral cones are normal ( see ( * ? ? ?
* theorem 1.3.5 ) ) .
thus , a toric variety @xmath2 associated to a fan @xmath338 is also normal .
moreover , we can check if a toric variety associated to a fan @xmath338 is smooth or not by looking at all the cones @xmath344 . in fact , if we say that a cone @xmath357 is smooth if and only if the associated affine toric variety @xmath351 is smooth , then we have the following result .
[ toricsing0 ] let @xmath2 be the toric variety associated to the fan @xmath338 .
then , @xmath358 and @xmath359 moreover , given a @xmath360dimensional simplicial cone @xmath361 with generators @xmath362 , let @xmath363 and define the multiplicity of @xmath32 by @xmath364.\ ] ] then , 1 .
@xmath32 is smooth if and only if @xmath365 .
2 . let @xmath366 be a basis of @xmath367 and write @xmath368 .
then , @xmath369 3 .
if @xmath370 is a face of @xmath32 , then @xmath371.\ ] ] in particular , @xmath372 whenever @xmath370 .
[ smoothincodimk ] let @xmath2 be a @xmath373dimensional toric variety associated to a simplicial fan @xmath338 , i.e. , a fan whose cones are all simplicial , on which there is a cone @xmath32 of full dimension @xmath62 . if @xmath2 is smooth in codimension @xmath343 , namely the closed invariant subset @xmath374 is such that @xmath375 , then we can choose a basis of @xmath376 in such a way the first @xmath343 generators of the cone @xmath32 corresponds to the first @xmath343 elements of the canonical basis of @xmath377 . in general ,
most of the interesting kind of singularities can also be characterized by looking at the ( maximal ) cones belonging to the fan .
[ toricsing1 ] let @xmath32 be a strongly convex rational polyhedral cone and let @xmath351 be the corresponding affine toric variety , then the following hold . 1 .
@xmath351 is cohen - macaulay .
@xmath351 is @xmath16-factorial if and only if @xmath32 is simplicial .
3 . @xmath351 is @xmath16-gorenstein if and only if there exists @xmath378 such that @xmath379 , for every ray @xmath348 . in this case
, the gorenstein index of @xmath351 is the smallest positive integer @xmath96 such that @xmath380 .
4 . if @xmath351 is @xmath16-gorenstein then @xmath351 has klt singularities .
if @xmath351 is @xmath16-gorenstein then @xmath351 has terminal singularities of gorenstein index @xmath381 if and only if exists @xmath382 such that @xmath383 @xmath384 the element @xmath385 is uniquely determined whenever @xmath32 is of maximal dimension in the fan .
if @xmath351 is @xmath16-gorenstein then @xmath351 has canonical singularities of gorenstein index @xmath381 if and only if exists @xmath382 such that @xmath383 @xmath386 the element @xmath385 is uniquely determined whenever @xmath32 is of maximal dimension in the fan .
if @xmath351 is gorenstein then @xmath351 has canonical singularities . here
, @xmath387 denotes the set of primitive vectors @xmath388 generating all the rays @xmath389 we may refer the reader to the survey @xcite for proofs or references to proofs . by theorem [ toricsing1 ] above , if @xmath2 is a @xmath16-factorial toric variety with canonical singularities then we have the decomposition @xmath390 therefore , if @xmath2 is a @xmath16-factorial toric variety with canonical singularities and with at most finitely many non - terminal points , then the ( finite ) set of canonical points is made up by some invariant points @xmath391 , where @xmath32 are of maximal dimension in the fan .
let us introduce now the necessary elements to run the toric mmp .
let @xmath392 be a rational vector subspace ; a collection of cones @xmath393 is a degenerate fan with vertex @xmath394 if it satisfies the usual conditions of a fan with strict convexity of cones replaced by @xmath395 this coincides with the usual notion of a fan @xmath396 in the quotient space @xmath397 . in our setting , we will always deal with @xmath16-factorial complete toric varieties , i.e. , toric varieties having a simplicial fan whose support is the whole space @xmath330 , by theorem [ toricsing1 ] above and ( * ? ? ?
* theorem 3.4.1 ) . in the toric case ,
every extremal ray @xmath28 of such a variety will correspond to a invariant curve @xmath398 such that @xmath399 $ ] or , equivalently , to a wall @xmath354 .
let us suppose that @xmath400 , where @xmath401 are primitive vectors .
since @xmath338 is a simplicial fan , @xmath402 separates two maximal cones @xmath403 and @xmath404 , where @xmath405 and @xmath406 are primitive on rays on opposite sides of @xmath402 .
the @xmath407 vectors @xmath408 are linearly dependent .
hence , they satisfy a so called _ wall relation _ : @xmath409 where @xmath410 and @xmath411 for @xmath412 . by reordering if necessary
, we can assume that @xmath413 let us introduce the notation @xmath414 and @xmath415 this wall relation and the signs of the coefficients involved allow us to describe the nature of the associated contraction .
[ contraction ] let @xmath2 be a @xmath16-factorial complete toric variety of dimension @xmath62 associated to the fan @xmath416 , and suppose that @xmath28 is an extremal ray of @xmath2 .
let us remove from @xmath417 all the walls @xmath402 associated to curves from @xmath104 and for each such @xmath402 replace the two adjacent maximal cones @xmath32 and @xmath418 from @xmath419 by the cone @xmath420 .
then , @xmath421 and @xmath422 are independent of @xmath402 and , by taking respectively their faces in @xmath423 , where @xmath424 , we get a complete fan @xmath425 , degenerate with vertex @xmath426 if @xmath427 , non - degenerate if @xmath428 .
moreover , if @xmath427 then @xmath429 is a complete simplicial fan . if @xmath430 , then @xmath431 is simplicial .
furthermore , in this case the induced morphism of toric varieties @xmath432 is the contraction of the extremal ray @xmath104 in the sense of mori theory .
moreover , the exceptional locus of @xmath172 corresponds to the irreducible closed invariant subvariety @xmath433 associated to the cone @xmath434 , which is contracted onto the irreducible closed invariant subvariety @xmath435 corresponding to the cone @xmath436 ( if @xmath427 then both are equal to the whole variety @xmath2 and @xmath173 , respectively ) .
therefore , @xmath437 , @xmath438 , and @xmath439 is a flat morphism , all whose fibers are ( possibly non - reduced ) projective toric varieties of picard number 1 and dimension @xmath440 .
in general , if @xmath354 is any wall of @xmath338 ( not necessarily corresponding to an extremal ray ) , we will also have a wall relation allowing us to compute the intersection number of the curve @xmath398 with every invariant divisor @xmath441 , @xmath347 .
[ intersections ] let @xmath338 be a simplicial fan in @xmath442 and @xmath443 be a wall , separating the two maximal cones @xmath403 and @xmath404 , satisfying the wall relation @xmath409 where @xmath410 and @xmath411 for @xmath412 . then , if we denote by @xmath444 the invariant weil divisor associated to the primitive vector @xmath349 corresponding to @xmath347 and by @xmath87 the invariant curve associated to the wall @xmath402 , we have that 1 . @xmath445 = 0 $ ] for all @xmath446 .
@xmath447 = \dfrac{{\operatorname{mult}}(\omega)}{{\operatorname{mult}}(\sigma)}$ ] and @xmath448 = \dfrac{{\operatorname{mult}}(\omega)}{{\operatorname{mult}}(\sigma')}$ ] .
3 . @xmath449 = \dfrac{b_i{\operatorname{mult}}(\omega)}{b_n{\operatorname{mult}}(\sigma)}= \dfrac{b_i{\operatorname{mult}}(\omega)}{b_{n+1}{\operatorname{mult}}(\sigma')}$ ] for @xmath450 . in the setting of toric varieties
, we will be interested in analyzing the extremal contractions appearing in [ section : extremalcontractions ] in terms of the description given by theorem [ contraction ] .
let us begin by the birational case .
[ toricdivisorial ] let @xmath337 be a @xmath16-factorial gorenstein toric variety of dimension @xmath0 .
let @xmath171 be a divisorial contraction such that 1 .
@xmath451 is an invariant prime divisor on @xmath2 .
@xmath452 is an invariant subvariety of codimension two .
3 . @xmath453=k_x\cdot [ f]=-1 $ ] for every non - trivial fiber @xmath128 of @xmath172 .
let us suppose that @xmath171 is defined by the contraction of the wall @xmath400 that separates the maximal cones @xmath403 and @xmath404 .
then , up to reordering if necessary , the wall relation satisfied by these cones is of the form @xmath454 by theorem [ contraction ] , up to reordering if necessary , we can suppose that @xmath171 is defined by the relation @xmath455 where @xmath456 , @xmath457 and @xmath458 .
let us denote by @xmath459 the invariant curve associated to the wall @xmath402 . by proposition [ intersections ] ,
@xmath445 = 0 $ ] for @xmath460 .
moreover , the wall relation gives us @xmath449 = 0 $ ] for @xmath461 and @xmath462 = \dfrac{\lambda}{\alpha}v(u_n)\cdot [ c ] = \dfrac{\lambda}{\beta}v(u_{n+1})\cdot [ c].\ ] ] by hypothesis , @xmath463 = -1 $ ] and thus @xmath447 = -\frac{\alpha}{\lambda}$ ] and @xmath448 = -\frac{\beta}{\lambda}$ ] .
it is well know that for a toric variety @xmath337 , @xmath464 is a invariant canonical divisor on @xmath2 ( see ( * ? ? ?
* theorem 8.2.3 ) ) , and thus the condition @xmath200 = 1 $ ] can be translated into @xmath465 on the other hand , we should notice that we can suppose that @xmath466 .
in fact , if @xmath467 , then the equation ( 1 ) implies @xmath468 , as @xmath469 is a primitive vector .
the same argument applies to the other two pairs . by assumption
, @xmath470 is a cartier divisor , i.e. , for each maximal cone @xmath471 , there is @xmath472 with @xmath473 . in our setting
, this condition applied to the two maximal cones @xmath32 and @xmath418 tells us that there exists two elements @xmath474 such that @xmath475 @xmath476 from the equation ( 1 ) we obtain @xmath477 @xmath478 by using the equation ( 2 ) and ( 3 ) we obtain that @xmath479 and thus @xmath480 , as @xmath481 and @xmath482 . in the same way , by using the equation ( 2 ) and ( 4 ) , we deduce that @xmath430 and hence @xmath483 .
finally , we get the relation @xmath484 and @xmath485 .
let us consider now the case when @xmath171 is a contraction of fiber type .
[ araujo ] let @xmath337 be a @xmath16-factorial toric variety of dimension @xmath0 and let @xmath171 be an extremal contraction such that @xmath486 .
then , @xmath172 is a flat morphism whose generic fiber is isomorphic to @xmath6 .
let us denote by @xmath487 the locus of points of @xmath173 over which @xmath172 is not smooth .
then , ( * ? ? ?
* theorem 3 ) implies that either @xmath487 is empty or of pure codimension 1 in @xmath173 , as @xmath16-factorial toric varieties have locally quotient singularities at every point ( see ( * ? ? ?
* theorem 11.4.8 ) ) . as a consequence
we have the following description for contractions of fiber type .
[ toricfibration ] let @xmath2 be a @xmath16-factorial projective toric variety of dimension @xmath0 . let us suppose that @xmath2 is smooth in codimension two and that there is an extremal contraction @xmath171 of fiber type onto a @xmath16-factorial projective toric variety of dimension @xmath26 . then @xmath172 is a @xmath142bundle .
moreover , there is a split vector bundle @xmath143 of rank 2 on @xmath173 and an isomorphism @xmath488 over @xmath173 . since @xmath489 and @xmath490
, we can take @xmath491 general hyperplane sections in @xmath173 in order to obtain a smooth surface @xmath492 such that @xmath493 is a smooth curve .
hence @xmath494 will be a morphism from a smooth projective surface onto a smooth projective curve such that its generic fiber is isomorphic to @xmath6 .
the tsen s theorem implies that @xmath495 admits a section and therefore the fibers of @xmath495 are generically reduced .
thus , outside a codimension two closed subset of @xmath173 , the fibers of @xmath172 are generically reduced .
then , @xmath172 is a smooth morphism outside a codimension two closed subset of @xmath173 .
we conclude by the remark [ araujo ] that @xmath172 is a smooth morphism .
finally , since @xmath171 is a toric morphism between toric varieties of relative dimension 1 , it admits two disjoint invariant sections @xmath496 and @xmath497 passing through the two invariant points of all the fibers of @xmath36 . hence , ( * ? ? ?
* remark 8) implies that there exists a rank 2 split vector bundle @xmath143 such that @xmath488 .
we are now able to prove the structure theorem for toric varieties with picard number 3 . by theorem [ thm : rho=3 ]
, we obtain a diagram @xmath252^{\sigma } \ar@/_1pc/[rr]_{\varphi } & y \ar[r]^{\pi } & z},\ ] ] where @xmath253 is a divisorial contraction sending a toric prime divisor @xmath133 onto an invariant subvariety @xmath498 of codimension two on @xmath39 , and @xmath70 and @xmath36 are both extremal contractions of fiber type whose fibers are of dimension 1 .
all these varieties are @xmath16-factorial toric fano varieties . by lemma [ toricfibration ] , @xmath499 is a @xmath142bundle isomorphic to the projectivization of the rank 2 split vector bundle @xmath500 . as @xmath501 for any line bundle @xmath502
, we can suppose that @xmath503 .
on the other hand , since @xmath56 is isomorphic to @xmath504 , we can consider an ample generator @xmath55 of @xmath56 and an integer @xmath505 such that @xmath506 .
up to tensor by @xmath507 , we can always suppose that @xmath43 .
in particular , both @xmath39 and @xmath25 must have at most terminal singularities , since @xmath39 has a most a finite number of canonical singularities and @xmath70 is locally trivial .
it should be noticed that in this case @xmath54 is contained in one of the two disjoint invariant sections associated to the @xmath142bundle @xmath24 . in fact , by theorem [ contraction ] ( taking @xmath427 and @xmath508 ) we have that @xmath499 is defined by a wall relation of the form @xmath509 where @xmath400 is a wall generating the extremal ray associated to this contraction , which separates the two maximal cones @xmath403 and @xmath404 . thus @xmath510 and @xmath499 is induced by the quotient @xmath511 . since @xmath512 and @xmath513 is a finite morphism , @xmath498 is sent onto an invariant subvariety of dimension @xmath491 , a divisor on @xmath25 .
thus , @xmath514 or @xmath515 , otherwise it will be sent onto a subvariety of dimension @xmath516 . if we suppose that @xmath517 , where @xmath518 is one of these disjoint invariant sections , then for any invariant affine open subset @xmath519 such that @xmath520 we will have @xmath521 the open set @xmath522 is therefore isomorphic to an open set on @xmath39 contained in the locus where @xmath523 is an isomorphism . in particular , @xmath524 is an affine gorenstein toric variety and thus @xmath394 is also gorenstein .
we conclude in this way that both @xmath39 and @xmath25 are gorenstein varieties . as a consequence of the formula @xmath525
we have that @xmath233 is a cartier divisor .
let us prove that @xmath526 verifies the universal property of the blow - up . the short exact sequence of sheaves @xmath527 gives @xmath528 where @xmath529 since @xmath253 is a contraction .
notice that @xmath530 is a @xmath142bundle .
in fact , since @xmath92 is a cartier divisor and @xmath531=1 $ ] for any non - trivial fiber @xmath128 of @xmath32 , it follows that the scheme theoretic fiber @xmath128 is an irreducible and generically reduced rational curve on @xmath2 .
then , by ( * ? ? ?
* theorem ii.2.8 ) , @xmath530 is a @xmath142bundle and thus @xmath532 . on the other hand , the cartier divisor @xmath533 is @xmath32-ample and therefore @xmath534 for @xmath141 , by ( * ?
? * vanishing theorem 1.1 ) .
hence , the above long exact sequence becomes @xmath535 and thus @xmath536 .
let us follow @xcite and notice that @xmath253 is a local contraction supported by the cartier divisor @xmath537 .
let @xmath128 be any non - trivial fiber of @xmath32 .
then , by ( * ? ? ? * theorem 5.1 ) , the evaluation morphism @xmath538 is surjective at every point of @xmath128 . on the other hand , @xmath539 and @xmath540
is defined to be the image of @xmath541 .
thus , @xmath542 is an invertible sheaf . then , by the universal property of the normalized blow - up , @xmath32 factorizes as @xmath252^{\tau \quad } \ar@/_1pc/[rr]_{\sigma } & \overline{\operatorname{bl}_a(y ) } \ar[r]^{\quad \nu\circ\varepsilon } & y},\ ] ] where @xmath543 is the blow - up of the coherent sheaf of ideals @xmath544 and @xmath545 its normalization . since @xmath32 contracts only the irreducible divisor @xmath133 , @xmath546 contracts no divisor .
if @xmath546 is not finite , it is a small contraction sending a curve @xmath547 to a point .
the rigidity lemma ( * ? ? ?
* lemma 1.6 ) applied to the @xmath6-bundle @xmath548 and the morphism @xmath549 implies that @xmath546 contracts the divisor @xmath133 , a contradiction .
hence @xmath546 is a finite and birational morphism onto a normal variety , and therefore @xmath546 is an isomorphism by zariski s main theorem . by ( * ?
* proposition 11.4.22 ) , if @xmath337 is a @xmath16-factorial toric variety with gorenstein terminal singularities , then @xmath550 .
therefore , if we start with a toric variety @xmath2 of dimension 3 or 4 satisfying the hypothesis of theorem [ thm : rho=3toric ] , we will obtain that @xmath25 is a smooth toric variety , implying that @xmath39 and @xmath2 must be both smooth toric varieties too .
in this section we study the extremal contractions described in [ section : extremalcontractions ] for toric toric varieties with picard number 2 .
let us begin with the proof of proposition [ thm : rho=2 toric fibration ] .
the isomorphism @xmath551 , with @xmath43 , follows from lemma [ toricfibration ] and the fact that @xmath67 . on the other hand ,
the condition @xmath552 is in fact equivalent to the condition of @xmath2 being fano . to see this ,
let us recall that the adjunction formula for projectivized vector bundles @xcite gives @xmath553 where @xmath554 is the tautological divisor on @xmath555 and @xmath58 is defined in such a way @xmath556 .
now , we notice that if @xmath128 is any fiber @xmath70 , then @xmath557 = 2\xi \cdot [ f ] = 2 > 0.\ ] ] on the other hand , if @xmath41 is an irreducible reduced curve and @xmath558 is the image of @xmath87 by the section associated to the quotient @xmath559 , then @xmath560 = \deg(c)\cdot ( i_y - a).\ ] ] hence , @xmath2 is fano if and only if @xmath561 , since the numerical classes of these curves above generates the mori cone of @xmath2 .
let us prove now theorem [ thm : rho=2 toric div isolated ] .
let @xmath28 be an extremal ray such that @xmath68 and let @xmath69 be the corresponding extremal contraction .
if @xmath70 is of fiber type then proposition [ thm : rho=2 toric fibration ] leads us to the first case .
let us suppose that @xmath69 is a birational contraction . since @xmath39 is a complete and simplicial toric variety of dimension @xmath62 and picard number one , the fan of @xmath39 contains exactly @xmath407 rays .
let us denote by @xmath562 the primitive lattice vectors generating these rays . by proposition [ extcont ]
, @xmath69 is a divisorial contraction sending an irreducible invariant divisor @xmath563 onto a codimension two subvariety @xmath54 , and @xmath39 is a @xmath16-factorial toric fano variety with isolated canonical singularities .
moreover , @xmath453=k_x \cdot [ f]=-1 $ ] for every non - trivial fiber @xmath128 of @xmath70 . notice that the fan of @xmath2 contains exactly @xmath564 rays , generated by the primitive lattice vectors @xmath565 .
now , lemma [ toricdivisorial ] implies that , up to reordering if necessary , @xmath70 is induced by the wall relation @xmath566 there are exactly @xmath567 walls satisfying this relation ( corresponding to the fibers over the @xmath567 invariants points of @xmath498 ) .
namely , the walls @xmath568 separating the two maximal cones @xmath569 and @xmath570 . from this
relation and proposition [ intersections ] , we can compute : @xmath571 = -\dfrac{{\operatorname{mult}}(\omega_i)}{{\operatorname{mult}}(\sigma_i)}=-\dfrac{{\operatorname{mult}}(\omega_i)}{{\operatorname{mult}}(\sigma_i')}=-1,\ ] ] and @xmath572=\dfrac{{\operatorname{mult}}(\omega_i)}{{\operatorname{mult}}(\sigma_i)}=\dfrac{{\operatorname{mult}}(\omega_i)}{{\operatorname{mult}}(\sigma_i')}=1.\ ] ] therefore , @xmath573 .
on the other hand , remark [ smoothincodimk ] implies that @xmath574 since @xmath2 has isolated singularities .
hence , both @xmath217 and @xmath575 are smooth cones .
we get that @xmath576 is a smooth cone for @xmath577 .
thus , @xmath578 is contained in the smooth locus of @xmath39 .
the isomorphism @xmath579 follows from the description of the blow - up of a smooth toric variety along an irreducible invariant ( smooth ) subvariety ( see ( * ? ? ?
* definition 3.3.17 ) ) .
since @xmath39 has @xmath567 smooth maximal cones and a complete simplicial toric variety of dimension @xmath62 and picard number one has exactly @xmath407 maximal cones ( corresponding to @xmath407 invariant points ) , @xmath39 has at most 2 singular points and they are outside @xmath54 . in particular , @xmath39 is a @xmath16-factorial gorenstein toric fano variety with @xmath67 .
let us prove that @xmath39 is isomorphic to one of the listed varieties . since the fan of @xmath39 contains a smooth maximal cone , we can suppose that the vectors @xmath580 correspond to the first @xmath62 elements of the canonical basis of @xmath377 , by remark 6.3 .
let us write @xmath581 , with @xmath582 for @xmath583 .
for each @xmath584 we have that @xmath585 .
therefore , proposition 6.2 leads @xmath586 , for @xmath587 .
hence , we can write @xmath588 , with @xmath589 . it should be noticed that @xmath39 has isolated singularities if and only if @xmath590 . thus , @xmath591 with @xmath589 relatively prime integers .
now , @xmath39 is a gorenstein weighted projective space if and only if @xmath592 and @xmath593 , by ( * ? ? ?
* lemma 3.5.6 ) .
equivalently , @xmath65 and @xmath66 .
if @xmath594 the only possibility is @xmath595 , leading to ( a ) : @xmath596 .
let us suppose that @xmath597 and notice that @xmath598 since @xmath590 .
on the other hand , @xmath599 since @xmath600 , this condition is fulfilled when @xmath601 .
therefore , @xmath602 implies that @xmath603 or , equivalently , @xmath604 .
this leads to @xmath605 , @xmath606 and hence to ( b ) : @xmath607 and @xmath62 must be even .
finally , if @xmath608 we get the last case ( c ) : @xmath591 .
conversely , given one of these listed varieties @xmath39 with their fans as above and considering @xmath2 to be the blow - up of @xmath39 along @xmath609 , we obtain a projective toric variety satisfying the hypothesis .
in fact , since @xmath498 is contained in the smooth locus of @xmath39 we obtain that @xmath2 is a @xmath16-factorial gorenstein toric variety with isolated canonical singularities . in order to prove that @xmath2 is fano we need to analyze the second extremal contraction that corresponds to the wall relation on @xmath2 given by @xmath610 in any of the three listed cases we will obtain @xmath611 = \dfrac{n-1+a}{b}\in \mathbb{z}_{>0},\ ] ] proving that @xmath2 is a fano variety , by the cone theorem and kleiman s criterion of ampleness . as a consequence we obtain the following list of possible admissible weights @xmath612 that corresponds to varieties @xmath591 as in theorem [ thm : rho=2 toric div isolated].2 .
compare with @xcite and @xcite .
.admissible @xmath591 as in theorem [ thm : rho=2 toric div isolated].2 [ cols="^,^",options="header " , ] from the wall relation @xmath613 we can deduce the nature of the second extremal contraction @xmath614 . in the smooth case it is a contraction of fiber type onto @xmath615 . in the singular case , it is a divisorial contraction sending its exceptional locus onto a point .
moreover , @xmath616 since @xmath617 we conclude with an example showing that is the hypothesis of isolated singularities in theorem [ thm : rho=2 toric div isolated].2 can not be dropped .
we exhibit an example of a @xmath16-factorial gorenstein toric fano fivefold @xmath2 with terminal singularities , whose singular locus is one - dimensional and that admits a birational extremal contraction @xmath69 which is not a blow - up , but only a blow - up in codimension two , and where @xmath39 is a non - gorenstein @xmath16-factorial toric fano fivefold .
[ contreexample ] let us consider the fan @xmath618 generated by the vectors @xmath619 where @xmath620 is the canonical basis of @xmath621 , @xmath622 and @xmath623 .
it can be checked by using @xcite that @xmath2 is a @xmath16-factorial gorenstein fano fivefold .
moreover , it can be checked by hand that its singular locus is one - dimensional , consisting only of terminal points , and given by @xmath624 the wall relation @xmath625 determines an extremal contraction @xmath69 sending the weil divisor @xmath626 ( which is not cartier ) onto @xmath627 .
finally , the relation @xmath628 implies that @xmath629 , which is not gorenstein , by ( * ? ? ?
* lemma 3.5.6 ) .
the aim of this section is to describe the contraction of extremal rays appearing in the previous sections by using universal coverings in codimension 1 .
let @xmath2 be a complex normal algebraic variety .
a _ covering in codimension 1 _ is a finite surjective morphism @xmath630 which is unramified in codimension 1 .
namely , there exists a subvariety @xmath631 such that moreover , a _ universal covering in codimension 1 _ is a covering in codimension 1 which is universal in the sense that for any covering in codimension 1 @xmath634 there exists a ( not necessarily unique ) covering in codimension 1 , @xmath635 , such that @xmath636 . 1 .
if @xmath92 is a cartier divisor , then @xmath638 is a cartier divisor .
if @xmath2 is a fano variety , then @xmath39 is a fano variety .
3 . if @xmath2 has terminal ( resp .
canonical ) singularities , then @xmath39 also has terminal ( resp .
canonical ) singularities .
as @xmath630 is unramified in codimension 1 , there is no ramification divisor and hence @xmath639 , implying ( 1 ) . the point
( 2 ) follows from ( * ? ? ?
* proposition 5.1.12 ) , while ( 3 ) follows from ( * ? ? ?
* proposition 3.16 ) .
[ ramification ] let @xmath630 be a finite morphism of toric varieties corresponding to the map of fans @xmath640 given by the inclusion of lattices @xmath641 of finite index , so that @xmath642 and @xmath643 . then , 1 . @xmath36 is equivariant with respect to the homomorphism of tori @xmath644 .
@xmath36 is an abelian cover with galois group @xmath645 .
the ramification divisor @xmath646 is supported on the torus invariant divisors @xmath353 , with multiplicities @xmath647 defined by the condition that the integral generator of @xmath648 is @xmath649 , for every ray @xmath650 .
[ toric pi_1 ^ 1 ] let @xmath2 be a @xmath16-factorial toric variety defined by the fan @xmath651 and let @xmath652 be the sublattice of @xmath340 generated by the primitive lattice generators @xmath653 of all the rays @xmath654 . then , @xmath655 this name comes from the following observation : the fan @xmath338 has cone generators @xmath657 , and the maximal cones of @xmath338 are generated by the @xmath62-element subsets of @xmath658 . as they are linearly dependent , for some @xmath660 .
therefore , @xmath661 if and only if the primitive lattice vectors @xmath662 generate the lattice @xmath340 .
if it is the case we will have that @xmath663 , by ( * ? ?
* example 5.1.14 ) .
[ fake wps ] let @xmath2 be a fake weighted projective space of dimension @xmath62 .
there exists a unique universal covering in codimension 1 @xmath664 , canonically identifying @xmath2 as a finite geometric quotient of @xmath151 by the torus - equivariant action of @xmath665 .
[ pws ] let @xmath337 be a @xmath16-factorial complete toric variety of dimension @xmath62 .
we define the _ canonical universal covering in codimension 1 of @xmath2 _ to be the covering in codimension 1 @xmath666 corresponding to the map of fans @xmath667 moreover , we say that @xmath2 is a poly weighted space ( pws ) if @xmath668 after the recent works of m. rossi and l. terracini , there is an explicit combinatorial construction ( via gale duality ) of the canonical universal covering in codimension 1 of any @xmath16-factorial complete toric variety .
this extends theorem [ fake wps ] to higher class group rank varieties ( see ( * ? ? ?
* theorem 2.2 ) for details ) . the remaining of the section will be devoted to study contractions @xmath526 of extremal rays as in the previous section via universal coverings in codimension 1 , and without the assumption of isolated singularities in the divisorial case .
let @xmath2 be a @xmath16-factorial projective toric variety and let @xmath399\subseteq \overline{{\operatorname{ne}}}(x)$ ] be an extremal ray defining a contraction of fiber type @xmath171 , where @xmath669 is an invariant curve contracted by @xmath172 .
then , the wall @xmath400 separates two maximal cones @xmath403 and @xmath404 . following the same notation as in theorem [ contraction ]
, we have that ( up to reordering , if necessary ) the contraction of @xmath104 is defined by the projection @xmath670^{\phi \hspace{2.4 cm } } & n_{x_r}:=n_x\slash \left(\operatorname{span}(u_{\beta+1},\ldots , u_{n+1})\cap n_x \right ) . } \ ] ] write @xmath671 for primitive vectors @xmath672 in @xmath673 and positive integers @xmath674 , where @xmath675 .
then , this @xmath672 define a @xmath676-dimensional cone @xmath677 which is of maximal dimension by theorem [ contraction ] and hence it corresponds to an invariant open affine subset @xmath678 of @xmath173 . in this setting
, we have that the contraction of fiber type @xmath171 becomes ( locally ) trivial over the invariant open affine subset @xmath679 after a finite morphism of toric varieties ( possibly ramified in codimension 1 ) .
[ toriclocallytrivial ] let @xmath679 be an invariant open affine subset and let @xmath680 .
then , there is a commutative diagram of toric morphisms @xmath681^{\widehat{\varphi_r } } \ar[dd]_{\pi_{x_0 } } & \widetilde{x_{r,0 } } \ar[rd]^{\mu } & \\ & & \widehat{x_{r,0 } } \ar[ld]^{\pi_{x_{r,0 } } } \\
x_0 \ar[r]^{\varphi_r } & x_{r,0 } & } \ ] ] that satisfies the following conditions : * @xmath682 and @xmath683 are the corresponding canonical universal coverings in codimension 1 .
* @xmath684 is a finite surjective morphism such that @xmath685 has ramification order @xmath674 over @xmath686 .
* @xmath687 is a trivial fibration , whose fiber is a wps .
if @xmath171 is a toric extremal contraction of fiber type , then @xmath172 is not necessarily locally trivial , as there may be some multiple fibers .
however , we have the following combinatorial criterion to decide whether is locally trivial over an invariant open affine subset of @xmath173 ( see ( * ? ? ?
* remark 3.3 and remark 3.8 ) ) .
let @xmath688 be a wall defining the contraction and @xmath689 as in theorem [ contraction ] .
then , @xmath690 can be seen as a fan in the real vector space @xmath426 , defining a toric variety @xmath128 .
moreover , we have that @xmath691 for all @xmath193 . in this
setting , invariant fibers of @xmath172 are given by @xmath692 with @xmath693 such that @xmath694 and @xmath695 .
let @xmath696 be a fixed point of the torus action and @xmath692 the fiber of @xmath172 over @xmath697 .
then , the following are equivalent : [ prop:1-covering , fiber type ] let @xmath2 be a @xmath16-factorial gorenstein toric fano variety of dimension @xmath0 with canonical singularities and with at most finitely many non - terminal points .
assume that there exists an effective prime divisor @xmath20 such that @xmath21 and that @xmath44 .
let @xmath28 be an extremal ray such that @xmath68 and let us denote by @xmath69 the corresponding extremal contraction .
assume that @xmath70 is of fiber type .
then there exist weights @xmath700 and a cartesian diagram of toric varieties @xmath72^{\widehat{\pi}\hspace{1 cm } } \ar[d]_{\pi_x } & \mathbb{p}(\lambda_0,\ldots,\lambda_{n-1 } ) \ar[d]^{\pi_y } \\ x \ar[r]^\pi & y } \ ] ] where vertical arrows denote the corresponding universal covering in codimension 1 , and @xmath73 is a gorenstein fano pws with terminal singularities such that @xmath74 .
moreover , @xmath701 leads to an isomorphism @xmath702 by proposition [ thm : rho=2 toric fibration ] , @xmath69 is a @xmath142bundle and @xmath2 is isomorphic to @xmath555 , where @xmath39 is a @xmath16-factorial gorenstein fano variety with terminal singularities . in particular ,
if @xmath703 is an invariant open affine subset of @xmath39 with fixed point @xmath696 and @xmath692 is the fiber of @xmath70 over @xmath697 , then @xmath70 is a locally trivial fibration over @xmath703 and hence @xmath704 by remark [ toricfibers ] above .
thus , the contraction is defined by the projection @xmath705^{\operatorname{pr}_2\hspace{1.4 cm } } & \operatorname{span}(\tau)\cap n_x \cong n_y . } \ ] ] this implies that all the ramification orders @xmath674 appearing in lemma [ toriclocallytrivial ] are equal to 1 .
this holds for every invariant open affine open subset of @xmath39 and hence it follows that the induced morphism between the universal coverings in codimension 1 , @xmath706 , is a locally trivial fibration all whose fibers are isomorphic to @xmath6 .
moreover , this commutative diagram is cartesian in the category of schemes by ( * ? ? ?
* lemma 2.2.7 ) .
* @xmath2 is gorenstein @xmath712 @xmath713 is gorenstein @xmath712 @xmath714 for every @xmath715 , by ( * ? ? ?
* lemma 3.5.6 ) .
* @xmath2 is terminal @xmath712 @xmath713 is terminal @xmath712 @xmath716 for each @xmath717 , by ( * ? ? ?
* proposition 2.3 ) .
* @xmath2 is fano @xmath712 @xmath718 by proposition [ thm : rho=2 toric fibration ] and the formula ( * ? ? ?
* proposition 2.3 ) for the canonical divisor of @xmath713 .
let @xmath2 be a @xmath16-factorial projective toric variety and @xmath28 an extremal ray defining a divisorial contraction @xmath171 .
then , there is a commutative diagram of toric morphisms @xmath72^{\widehat{\varphi_r } } \ar[dd]_{\pi_x } & \widetilde{x_r } \ar[rd]^{\mu } & \\ & & \widehat{x_r } \ar[ld]^{\pi_{x_r}}\\ x \ar[r]^{\varphi_r } & x_r & } \ ] ] that satisfies the following conditions : * @xmath722 and @xmath723 are the corresponding canonical universal coverings in codimension 1 .
* @xmath684 is a finite surjective morphism unramified in codimension 1 and given by the inclusion of lattices @xmath724 of index @xmath725 , defined by the condition that the integral generator of @xmath726 is @xmath727 , where @xmath728 is the exceptional divisor of @xmath172 .
* @xmath729 is a divisorial contraction with @xmath730 and @xmath731 . *
if @xmath2 ( resp .
@xmath173 ) is a fano variety then @xmath73 ( resp . @xmath732 and
@xmath733 ) is . *
if @xmath2 ( resp .
@xmath173 ) has gorenstein singularities then @xmath73 ( resp . @xmath732 and @xmath733 ) does . * if @xmath2 has terminal ( resp .
canonical ) singularities , then all varieties in the diagram have terminal ( resp .
canonical ) singularities .
let us suppose that @xmath171 is given by the contraction of the wall @xmath400 separating the maximal cones @xmath403 and @xmath404 .
then , the wall relation satisfied by these cones ( defining the contraction ) is given by @xmath734 where @xmath410 and @xmath411 . since @xmath172 is a divisorial contraction we can suppose that ( up to reordering , if necessary ) @xmath735 and @xmath736 , by theorem [ contraction ] .
thus , @xmath737 and the contraction corresponds to the stellar subdivision of the cone @xmath738 with respect to the primitive lattice vector @xmath469 satisfying the wall relation above .
clearly we have the inclusion of lattices @xmath724 , which is of finite index since @xmath741 , by the wall relation above .
hence , we obtain an induced finite surjective morphism of toric varieties @xmath742 that is unramified in codimension 1 , by lemma [ ramification ] . now , the fan of @xmath73 is obtained by the stellar subdivision of the fan of @xmath733 with respect to the primitive vector @xmath469 satisfying the wall relation above , obtaining the desired commutative diagram that satisfies ( a ) , ( b ) and ( c ) by construction . by lemma [ divisorialcovering ]
there is a commutative diagram of toric morphisms @xmath72^{\widehat{\pi } } \ar[dd]_{\pi_x } & \widetilde{y } \ar[rd]^{\mu } & \\ & & \widehat{y } \ar[ld]^{\pi_{y}}\\ x \ar[r]^{\pi } & y & } \ ] ] where @xmath722 and @xmath743 are the corresponding canonical universal coverings in codimension 1 and @xmath744 is a finite surjective morphism defined by the inclusion of lattices @xmath745 .
it should be noticed that under the hypothesis of the statement , lemma [ toricdivisorial ] implies that @xmath746 and hence @xmath747 .
moreover , this commutative diagram is cartesian in the category of schemes by ( * ? ? ?
* lemma 2.2.7 ) .
let @xmath2 as in proposition [ prop:1-covering , divisorial ] and suppose that @xmath749 , i.e. , that @xmath709 .
the extremal divisorial contraction @xmath750 determines the shape of the fan of @xmath2 in terms of the fan of @xmath151 ( which is well known ) : the fan of @xmath151 is given by @xmath407 lattice primitive vectors @xmath751 that generates the lattice @xmath340 and that satisfy the relation @xmath752 let us suppose that @xmath70 contracts @xmath753 onto the invariant subvariety @xmath754 , of codimension two .
then , @xmath755 , by lemma [ toricdivisorial ] .
in particular , the same computation used to prove ( * ? ? ?
* lemma 3.5.6 ) shows that @xmath2 is a fano gorenstein variety if and only if @xmath714 , @xmath756 , @xmath757 and @xmath758 for every @xmath759 , where @xmath760 .
however , to the best of the author s knowledge , the characterization such @xmath2 having terminal singularities is more subtle . indeed
, if @xmath2 is a @xmath16-factorial fano gorenstein toric variety then it corresponds to a simplicial reflexive lattice polytope @xmath761 ( see @xcite for details ) . in (
* corollary 3.7 ) , b. nill characterizes all polytopes among these ones that correspond to varieties with only terminal singularities , but it does not seem easy to translate this characterization into a function of the weights @xmath146 . finally , it should be noticed that if one of the weights is equal to 1 , say @xmath762 , then we have a coordinate - wise description of the primitive vectors defining the fan of @xmath763 . namely , the canonical basis of @xmath377 together with the vector @xmath764 . in this case , we can explicitly compute the cartier data @xmath765 of @xmath92 , allowing us to decide whether the singularities of @xmath2 are terminal or not .
in particular , we compute that @xmath2 has only gorenstein terminal singularities if all the integers @xmath766 considered before , are equal or greater than 3 . the variety defined in example [ contreexample ]
satisfy this condition .
m. andreatta and j. a. winiewski , _ a view on contractions of higher - dimensional varieties _ , algebraic geometry - santa cruz 1995 , proc .
pure math .
, 62 , part 1 , amer . math . soc . , providence , ri , 1997 , pages 153 - 183 . | in this paper we study the geometry of mildly singular fano varieties on which there is an effective prime divisor of picard number one .
afterwards , we address the case of toric varieties . finally , we treat the lifting of extremal contractions to universal covering spaces in codimension 1 . |
the conditional second moment of the @xmath23 quadrature depends on the width @xmath41 and the position of the conditioning interval .
a narrow interval generally yields the lowest possible moment values , but they come at the cost of low number of data points and , consequently , lesser certainty about the validity of the result .
the best certifiable violation is obtained when the upper boundary of the confidence band reaches its minimum value .
[ fig_delta ] shows the second moment with its upper confidence band estimated from the measured data .
for all considered values of @xmath41 we have optimized over the values @xmath21 marking the center of the interval to obtain the minimum of the upper confidence band boundary .
the plotted moment can be therefore expressed as @xmath50 = \min_{x_2 } \left\ { e\bigl[x_1 ^ 2\big| x_2 - \frac{\delta}{2 } < |x_2| < x_2 + \frac{\delta}{2}\bigr ] \right\}.\ ] ] we have obtained the lowest certifiable variance for @xmath47 . $ ] ( red ) and its upper confidence band ( green ) relative to the width of the postselection interval @xmath41 .
the position of the conditioning interval given by the measured value of @xmath38 was optimized over in order to obtain minimal certifiable variance given by the upper confidence interval .
, width=250 ]
this work was partly supported by crest of jst , jsps kakenhi , apsa .
p. m. , p.z . , and
r. f. acknowledge project gb14 - 36681 g of the czech science foundation .
y. h. acknowledges support from alps .
k. m. acknowledges support from jsps .
cc c. k. hong , z. y. ou , and l. mandel , measurement of subpicosecond time intervals between two photons by interference , phys .
lett . * 59 * , 2044 ( 1987 ) .
t. kobayashi , r. ikuta , s. yasui , s. miki , t. yamashita , h. terai , t. yamamoto , m. koashi , and n. imoto , frequency - domain hong - ou - mandel interference , nature photonics * 10 * , 441 ( 2016 ) ; r. lopes , a. imanaliev , a. aspect , m. cheneau , d. boiron , and c. i. westbrook , atomic hong - ou - mandel experiment , nature * 520 * , 66 ( 2015 ) ; k. toyoda , r. hiji , a. noguchi , and s. urabe , hong - ou - mandel interference of two phonons in trapped ions , nature * 527 * , 74 ( 2015 ) ; a. m. kaufman , b. j. lester , c. m. reynolds , m. l. wall , m. foss - feig , k. r. a. hazzard , a. m. rey , and c. a. regal , two - particle quantum interference in tunnel - coupled optical tweezers science * 345 * , 306 ( 2014 ) ; a. sipahigil , k. d. jahnke , l. j. rogers , t. teraji , j. isoya , a. s. zibrov , f. jelezko , and m. d. lukin , indistinguishable photons from separated silicon - vacancy centers in diamond , phys .
lett . * 113 * , 113602 ( 2014 ) .
u. l. andersen , j. s. neergaard - nielsen , p. van loock , and a. furusawa , hybrid discrete- and continuous - variable quantum information , nature physics * 11 * , 713 ( 2015 ) .
a. ourjoumtsev , h. jeong , r. tualle - brouri , and p. grangier , generation of optical ` schrdinger cats ' from photon number states , nature * 448 * , 784 ( 2007 ) .
j. etesse , m. bouillard , b. kanseri , and r. tualle - brouri , experimental generation of squeezed cat states with an operation allowing iterative growth , phys .
lett . * 114 * , 193602 ( 2015 ) .
a. e. ulanov , i. a. fedorov , d. sychev , p. grangier , and a. i. lvovsky , loss - tolerant state engineering for quantum - enhanced metrology via the reverse hong - ou - mandel effect , nature communications * 7 * , 11925 ( 2016 ) ; h. jeong , a. zavatta , m. kang , s .- w .
lee , l. s. costanzo , s. grandi , t. c. ralph , and m. bellini , generation of hybrid entanglement of light , nature photonics * 8 * , 564 ( 2014 ) .
r. filip and p. zapletal , squeezed - state generation from single - photon sources , phys .
a * 90 * , 043854 ( 2014 ) .
k. makino , y. hashimoto , j. yoshikawa , h. ohdan , t. toyama , p. van loock , and a. furusawa , synchronization of optical photons for quantum information processing , science advances * 2 * , e1501772 ( 2016 ) .
m. cooper , e. slade , m. karpiski , and b. j. smith , characterization of conditional state - engineering quantum processes by coherent state quantum process tomography , new j. phys . * 17 * , 033041 ( 2015 ) .
y. s. teo , h. zhu , b .-
englert , j. ehek , and z. hradil , quantum - state reconstruction by maximizing likelihood and entropy , phys .
* 107 * , 020404 ( 2011 ) .
r. horodecki , p. horodecki , m. horodecki , and k. horodecki , quantum entanglement , rev .
* 81 * , 865 ( 2009 ) .
k. modi , a. brodutch , h. cable , t. paterek , and v. vedral , the classical - quantum boundary for correlations : discord and related measures , rev .
phys . * 84 * , 1655 ( 2012 ) .
single photon divided on a beam splitter and conditioned by homodyne detection can be always expressed as @xmath51 . in this state ,
@xmath52 and the moment is never squeezed .
w. vogel , nonclassical states : an observable criterion , phys .
lett . * 84 * , 0031 ( 2000 ) .
j. yoshikawa , k. makino , s. kurata , p. van loock , and a. furusawa , creation , storage , and on - demand release of optical quantum states with a negative wigner function , phys .
x * 3 * , 041028 ( 2013 ) .
j. yoshikawa , k. makino , a. furusawa , in _ engineering the atom - photon interaction _ , a. predojevi , m. w. mitchell , eds .
( springer cham , 2015 ) , chap . 8 . m. ebert , a. gill , m. gibbons , x. zhang , m. saffman , and t. g. walker , atomic fock state preparation using rydberg blockade , phys .
lett . * 112 * , 043602 ( 2014 ) .
a. d. oconnell , m. hofheinz , m. ansmann , r. c. bialczak , m. lenander , e. lucero , m. neeley , d. sank , h. wang , m. weides , j. wenner , j. m. martinis , and a. n. cleland , quantum ground state and single - phonon control of a mechanical resonator , nature * 464 * , 697 ( 2010 ) . c. galland , n. sangouard , n. piro , n. gisin , and t. j. kippenberg , heralded single - phonon preparation , storage , and readout in cavity optomechanics , phys .
* 112 * , 143602 ( 2014 ) .
m. hofheinz , e. m. weig , m. ansmann , r. c. bialczak , e. lucero , m. neeley , a. d. oconnell , h. wang , j. m. martinis , and a. n. cleland , generation of fock states in a superconducting quantum circuit , nature * 454 * , 310 ( 2008 ) .
s. p. premaratne , f.c .
wellstood , and b.s .
palmer , microwave photon fock state generation by stimulated raman adiabatic passage , nature communications * 8 * , 14148 ( 2017 ) . | quality of individual photons and their ability to interfere is traditionally tested by measuring the hong - ou - mandel photon bunching effect . however , this phase insensitive measurement only tests the particle aspect of the quantum interference .
motivated by these limitations , we formulate a witness capable of recognizing both the quality and the coherence of two photons from a single set of measurements .
we exploit the conditional nonclassical squeezing which is sensitive to both the quality of the single photons and their indistinguishability and we show that it can be used to directly reveal both the particle and the wave aspects of the quantum interference .
finally , we experimentally test the witness by applying it to a pair of independent single photons generated on demand
. quantum interference of individual photons is one of the fundamental keystones of quantum technology .
the ability to recognize whether photons from different sources can interfere is therefore an important one .
the basic tests use photon correlation measurements to ascertain presence of individual photons and two - photon bunching to verify their interference .
the experimental test of the bunching effect traditionally employs the hong - ou - mandel interference @xcite , in which the two individual photons are mixed on a balanced beam splitter and the resulting two mode field is measured by a pair of photon counters .
the bunching manifests as the lack of coincidence counts - if the coincidence probability is below one half , the interference effect is nonclassical , i. e. incompatible with description using a mixture of classical waves or classical particles .
recent experiments for a wide range of experimental platforms safely observed low values of coincidence counts and thus demonstrated this fundamental nonclassical aspect @xcite . for discrete variable quantum technology ( dv )
these features are sufficient . however , from the point of view of continuous variables ( cv ) and the more general framework of hybrid quantum technology @xcite , these experiments quantify only the particle nature of the interference and ignore the vacuum and the related phase sensitive aspects @xcite of the photon bunched state .
the full picture of two - photon interference can be obtained by replacing the photon counters with homodyne detection and performing a full tomography @xcite .
the reconstructed density matrix can then be analyzed in order to visualize and evaluate the interference caused by the the indistinguishability .
the drawback to this tomographic approach is that it can not be used as a direct witness .
the tomographic reconstruction is a process combining information from many incompatible measurements in order to obtain not directly measurable quantities , such as off - diagonal elements of the density matrix .
the elements of the reconstructed matrix contain , in principle , full information about the state .
however , the information is practically limited due to unavailability of a complete set of measurement and the need to form prior assumptions about the suitable hilbert space of the system @xcite . for this reason some elements , in particular those related to higher fock numbers , are difficult to estimate with sufficient precision .
this may lead to erroneous or indeterminate conclusions , such as attributing nonclassical properties to interference of classical signals .
fortunately , there are some phase sensitive photon correlation properties which can be used for construction of a direct witness not reliant on the tomographic approach .
two - photon interference leads to a state which , by suitably conditioning homodyne measurement , can be transformed into a state with quadrature squeezing @xcite .
this squeezing , although not strong enough to be of much use on its own , can be created neither from classical light nor from non - interfering photons and is therefore an excellent witness of nonclassical and coherence properties .
it can be also directly measured .
interestingly , the measurement itself relies only on classical correlations within the state , without exploiting its entanglement @xcite or discord @xcite .
the general witness therefore requires one further condition - the phase of the initial states needs to be randomized .
this step , while not affecting single photon states , cancels the possible false positives .
the full _ phase sensitive hong - ou - mandel effect _ therefore serves as a directly observable witness simultaneously for both particle and wave - like aspects of the two photon interference .
it also offers a new , much more challenging , benchmark against which new single photon sources can be tested . in this paper
we introduce the phase sensitive hong - ou - mandel effect as the witness for both high quality and high overlap of remotely prepared single photon states .
the witness relies on observation of nonclassical quadrature correlations in a state created by interference of two single photons .
we apply the witness to a pair of photons individually prepared in timing synchronization cavities @xcite , and we demonstrate both their high quality and coherence . in the ideal version of the hong - ou - mandel experiment ,
the two single photons indistinguishable in all but a single degree of freedom are in the initial state @xmath0 and get mixed on a balanced beam splitter .
this produces an entangled state @xmath1 with @xmath2 being the relative phase between the two output modes introduced by the beam splitter .
the two resulting modes are then subjected to intensity measurements by avalanche photo diodes ( apds ) whose action is described by their povm elements @xmath3 and @xmath4 , which provide us with probabilities @xmath5 , @xmath6 . for indistinguishable photons there is a distinct lack of coincidences represented by @xmath7 on the other hand , for completely distinguishable photons the probability of coincident detection is firmly set to @xmath8 .
the full hong - ou - mandel measurement is realized by deliberately delaying one of the photons , thus increasing the distinguishability , and measuring the coincidence rates for these different settings .
the effect is then quantified in terms of visibility , which is defined as @xmath9 where the maximization and minimization is performed over the set of different delays , and it is equal to one for perfectly indistinguishable single photons and zero for perfectly distinguishable ones .
the visibility can be also reduced due to imperfection of the single photon states , caused by presence of vacuum or higher fock terms . however , since the vacuum terms do not contribute towards coincidence rates , only the higher fock terms lead to reduction in visibility .
this means the hong - ou - mandel effect is a good indicator of distinguishability of single photons , but it provides a limited insight into their quality .
another shortcoming of the hong - ou - mandel effect is related to the technological aspect of the employed single photon detectors .
these particle detectors have a very broad detection spectrum incompatible with the much narrower spectrum of continuous homodyne detectors , which play crucial role in cv quantum information processing with light @xcite .
this discrepancy can be removed by using optical filtering , but this comes with the cost of detection efficiency and alteration of the photon statistics .
another issue is the fundamental identity of the single photon detectors as _ intensity _ detectors . as such
, the measurement is insensitive to the vacuum and can not reveal the wave portion of the interference - the relative phase between the modes - related to the phase sensitive correlations between the field quadratures . as a consequence , with the particle detectors the state ( [ state_id ] ) can not be differentiated from a completely mixed state @xmath10 .
needless to say , in cv and hybrid quantum information processing this distinction is quite crucial .
we are therefore looking for a strong unifying witness capable of verifying the quality of single photon states as well as their indistinguishability . at the same time
, we want a witness which employs high quality homodyne detectors in order to make it compatible with cv quantum information processing @xcite . and
finally , we want to observe the interference effect directly , without the need to consider different measurement bases and settings , and subsequent reconstruction of the state s density matrix .
all of these requirements are reconciled in a three step procedure , in which the two single photon states are first phase randomized , then they interfere on a balanced beam splitter , and finally they are subjected to the joint measurement of phase - correlated quadrature operators @xmath11 and @xmath12 , where @xmath13 with @xmath14 .
@xmath15 and @xmath16 obey canonical commutation relation @xmath17 = i$ ] . because of the phase insensitivity of initial single photon states , a common component of phases has no significance .
hence , quadrature operators can be expressed only with the differential component @xmath18 as @xmath19 and @xmath20 .
let us look at the ideal scenario with the state ( [ state_id ] ) and consider a homodyne measurement of mode 2 yielding a value @xmath21
. this measurement projects the state of the remaining mode 1 into a state proportional to @xmath22 .
this state has zero mean values of the quadrature operators @xmath23 and @xmath24 .
the remarkable thing is that by choosing some particular values of @xmath21 , the state in mode 1 is projected into a squeezed state , whose second moment at a certain measurement angle shows a value less than @xmath25 .
the witness , confirming that state @xmath26 was obtained by interference of two phase insensitive single photon states on a balanced beam splitter , can be therefore formulated as @xmath27 = \frac{\mathrm{tr } [ x_1 ^ 2(\delta \theta)\otimes|x_2\rangle\langle x_2| \rho]}{\mathrm{tr } [ 1\otimes|x_2\rangle\langle x_2| \rho ] } < \frac{1}{2},\end{aligned}\ ] ] where @xmath28 $ ] represents conditional mean value of an operator when condition @xmath29 is satisfied .
the phase @xmath30 that gives minimum second moment is tied with the phase of superposition @xmath2 as @xmath31 .
it is important that @xmath2 is arbitrarily controllable by classical phase locking , so @xmath30 is known _ a priori _ and can be set to an arbitrary value .
the procedure therefore requires only a single set of measurements , avoids any need for reconstruction , and ( [ directwitness ] ) is a direct witness for both the indistinguishability and quality of single photons .
the conditioning on @xmath21 is related to the generalized theory of squeezing extraction presented in ref .
@xcite , but there are some interesting distinctions . due to the ideally vanishing nature of the first moments , the condition can be specified with help of the second _ noncentral _ moment which makes it unsatisfiable by a lone single photon @xcite .
this is in contrast to observing squeezing of the conditional central moment , the variance , which can be achieved even for a single photon interacting with the vacuum and may therefore lead to false positives .
however , the conditional squeezing can be considered as a witness for quality of _ single _ photons @xcite .
for example , an imperfect single photon state represented by density operator @xmath32 can be confirmed as nonclassical for an arbitrary positive value of @xmath33 .
the method demonstrates phase - sensitive continuous variable nature of the correlations in photon anti - correlation experiments .
it is also interesting to point out that when the squeezing is replaced by another nonclassicality witness employing only single quadrature measurement , such as one proposed in ref .
@xcite , no conditional nonclassicality can be detected .
the experimental setup for the test of the phase - sensitive hong - ou - mandel effect for two single photon states is based on our former experiment reported in ref .
@xcite and it is shown in fig .
[ fig_expsetup ] .
each of the two photon sources is based on the heralded scheme with parametric down conversion contained in an optical cavity .
photon pairs are generated from the continuous - wave pump light at 430 nm at random times .
one photon of the pair immediately leaves the cavity and is detected by an avalanche photodiode , thus heralding presence of another 860 nm photon stored in the cavity .
a technique involving concatenated shutter cavity @xcite can be then used to release the second photon at a desired time .
this allows timing synchronization of the two generated photons and enables their interference on the balanced beam splitter . with this technique , the simultaneous emission occurs 15 times in a second with 300 ns synchronization window and we gathered 12,000 events within an hour . since the single photon states are generated independently , their phase is random and needs no further randomization .
the state after interference is measured by phase - sensitive homodyne detection in both output modes of the balanced beam splitter : one mode is used for conditioning to generate squeezing and the other mode is used for the verification of squeezing in the conditioned state .
the relative phase between a local oscillator and the output mode at both homodyne detections , and the relative phase of two local oscillators which is equivalent to @xmath34 are actively stabilized .
the quantum efficiency of homodyne detectors and the visibility between a signal beam and the local oscillator are both more than 99% .
any correction of losses is not applied to the results below . .
( a ) joint probability distribution @xmath35 .
( b ) normalized quadrature distribution @xmath36
. an occurrence rate of @xmath23 with the window @xmath37 is shown in histogram .
lines are gaussian distributions with the variance of conditioned @xmath36 ( solid red ) and that of vacuum fluctuation for reference ( dashed blue ) , respectively.,width=240 ] fig .
[ fig_hist ] ( a ) shows the joint measured distribution of quadratures @xmath23 and @xmath38 of the two modes emerging from the beam splitter with @xmath39 .
the quadratures were locked against each other , but their relation to the outside reference varied over time .
this effectively resulted in a statistics which is equivalent to measurement without an outside phase reference .
we will therefore omit the phase dependence from now on .
we can see that the quadrature data are distinctively correlated in a manner consistent with the target state . in accordance with the proposed witness
, we are looking for a quantitative indicator of the hong - ou - mandel coherence in the form of the second conditional moment @xmath40 $ ] .
the theory asks for conditioning on a single measurement result .
since this is impossible , we have relaxed the requirement for condition and accepted situations in which the value of quadrature @xmath38 fell into a bin of width @xmath41 , @xmath42 . due to the symmetrical nature of the distribution , we also used both positive and negative values in the post - selection .
an example of the conditional quadrature distribution of @xmath23 is shown in fig .
[ fig_hist ] ( b ) .
the conditioning window , denoted by set @xmath43 , was set to @xmath37 and it picked 1,028 of the initial 12,000 data points .
the conditional variance was found to be @xmath44 = 0.45 $ ] , which is smaller than the vacuum fluctuations variance of 0.5 .
$ ] where @xmath45 is the set of values satisfying @xmath46 .
the width of the post - selection interval was chosen to be @xmath47 .
dashed blue line marks theoretical prediction for the interference of two ideal single photon states .
solid red line shows prediction for interference of two imperfect photons produced by the cavities .
red dots represent experimental data with the light red zone marking the @xmath48 confidence band .
the inset shows detailed behavior in the area with the highest certifiable squeezing .
, width=240 ] let us now discuss the results in more detail , giving more attention to possible statistical errors , and make sure the observed squeezing is significant . for that we need to construct the confidence band for the estimated variance . under the assumption that the conditional distributions are approximately gaussian we used monte carlo method to simulate 1,000 runs of the experiment , estimated their moments and evaluated their statistics
the standard deviation of the estimated variances depends on the number of data points and therefore on the position and the width of the post - selection interval .
we have optimized over these parameters and found minimal certified variance for conditional interval width @xmath47 ( see appendix a for details ) .
the conditional variances for this scenario are shown in fig .
[ fig_condvx2 ] .
we can see that the measured data closely follow the curve predicted from knowledge of the two independent single photon states , which was obtained by their individual tomographic reconstructions @xcite .
this shows that the amount of squeezing is limited mainly by the quality of the single photons and not their coherence .
furthermore , the squeezing is observable even if we take the three standard deviations wide confidence band into account .
this confirms that the squeezing is not a statical artifact and demonstrates both the nonclassical nature of the two initial single photon states and their mutual coherence .
this first application of the proposed witness also shows its sensitivity to small imperfection of these components .
note , visibility of the traditional hong - ou - mandel interference estimated from the tomography data would be @xmath49 in our experiment . in summary
, we have suggested a phase sensitive hong - ou - mandel effect as a witness for recognizing high quality of individual photons and their ability to quantum mechanically interfere .
such interference is the basic building block of future hybrid optical technologies which combine both continuous and discreet aspects of light @xcite .
the suggested witness tests both of these properties and it does that directly , using only a single set of measurements without needing a sequence of individually incompatible measurements and without directly exploiting entanglement @xcite or discord @xcite .
we have applied the witness to the pair of on demand generated photons and demonstrated both their nonclassical nature and their mutual coherence .
the witness can be directly applied to other experimental platforms where generation of interfering fock states plays an important role , like atomic ensembles @xcite , optomechanics @xcite , and superconducting circuits @xcite . |
invariant manifolds provide geometric structures for understanding dynamical behavior of nonlinear systems under uncertainty .
some systems evolve on fast and slow time scales , and may be modeled by coupled singularly perturbed stochastic ordinary differential equations ( sdes ) .
a slow - fast stochastic system may have a special invariant manifold called a random slow manifold that capture the slow dynamics .
we consider a stochastic slow - fast system @xmath0 where @xmath1 and @xmath2 are matrices , @xmath3 is a small positive parameter measuring slow and fast scale separation , @xmath4 and @xmath5 are nonlinear lipschitz continuous functions with lipschitz constant @xmath6 and @xmath7 respectively , @xmath8 is a noise intensity constant , and @xmath9 is a two - sided @xmath10-valued wiener process ( i.e. , brownian motion ) on a probability space @xmath11 . under a gap condition and for @xmath3
sufficiently small , there exists a random slow manifold @xmath12 , @xmath13 , as in @xcite , for slow - fast stochastic system - .
when the nonlinearities @xmath14 are only locally lipschitz continuous but the system has a random absorbing set ( e.g. , in mean - square norm ) , we conduct a cut - off of the original system .
the new system will have a random slow manifold which captures the original system s slow dynamics .
the random slow manifold is the graph of a random nonlinear mapping @xmath15 , with @xmath16 determined by a lyapunov - perron integral equation @xcite , @xmath17 with @xmath18 and @xmath19 .
the random slow manifold exponentially attracts other solution orbits .
we will find an analytically approximated random slow manifold for sufficiently small @xmath3 , in terms of an asymptotic expansion in @xmath3 , as in @xcite .
this slow manifold may also be numerically computed as in @xcite . by restricting to the slow manifold ,
we obtain a lower dimensional reduced system of the original slow - fast system - , for @xmath3 sufficiently small @xmath20 where @xmath21 and @xmath22 are defined in the next section .
if the original slow - fast system - contains unknown system parameters , but only the slow component @xmath23 is observable , we conduct parameter estimation using the slow system . since the slow system is lower dimensional than the original system , this method offers an advantage in computational cost , in addition to the benefit of using only observations on slow variables .
this paper is arranged as follows . in the next section ,
we obtain an approximated random slow manifold and thus the random slow system .
then in section 3 , we present a method for parameter estimation on the slow manifold .
finally , a simple example is presented in section 4 to illustrate our method .
by a random transformation @xmath24 we convert the sde system - to the following system with random coefficients @xmath25 where @xmath26 is the stationary solution of linear system @xmath27 . here
@xmath28 is the wiener shift implicitly defined by @xmath29 .
note that @xmath18 . define a mapping ( between random samples ) @xmath30 implicitly by @xmath31 .
then @xmath32 is also a wiener process with the same distribution as @xmath33 .
moreover , @xmath34 and @xmath35 are identically distributed with @xmath36 and @xmath37 , respectively . by a time change @xmath38 and
using the fact that @xmath39 and @xmath40 are identically distributed , the system - is reformulated as @xmath41,\\ \label{fast - equation - random000fast } & & y'=by+g ( x , y + \sigma \eta(\theta_\tau\psi_\varepsilon\omega)),\end{aligned}\ ] ] where @xmath42 .
we make the following two hypotheses .
@xmath43 : there are positive constants @xmath44 , @xmath45 and @xmath46 , such that for every @xmath47 and @xmath48 , the following exponential estimates hold : @xmath49 @xmath50 : @xmath51 .
+ then there exists a random slow manifold @xmath52 for the random system - , with @xmath53 being expressed as follows @xcite , @xmath54 we can get a small @xmath3 approximation for @xmath53 .
start with the expansion @xmath55 and the integral expression @xmath56\,dr $ ] in the system - . using the taylor expansions of @xmath57 and @xmath58 at point @xmath59 and denoting @xmath60
, we obtain @xmath61 for @xmath3 sufficiently small , with @xmath62 and @xmath63\\{}&&\quad\quad\quad\quad\quad\quad + g_y ( \xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))y_1(s)\big\}\,ds,\end{aligned}\ ] ] where @xmath64 and @xmath65 satisfy the following random differential equations , respectively @xmath66 and @xmath67y_1(\tau)\\ \quad\quad\quad\quad+g_x(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\big\{a\tau\xi+\int_0^\tau f(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))\,ds\big\ } , \\
y_1(0)= h_1^{\varepsilon}(0 , \omega )
. \end{cases}\end{aligned}\ ] ] noticing that @xmath68 and @xmath69 have identical distribution , together with the fact that @xmath70 we see that @xmath71 and @xmath69 are identically distributed .
recall the transformation introduced in the beginning of this section @xmath72 and @xmath73 .
we thus obtain a lower dimensional , slow stochastic system on the slow manifold , @xmath74 moreover , @xmath75 is an approximation or a first order truncation of @xmath76 , and hence we have an approximate slow system @xmath77 this is the _ slow system _ we will work on for parameter estimation in the next section .
if the original slow - fast system - contains an unknown system parameter @xmath78 , but only the slow component @xmath23 is observable , we can conduct parameter estimation using the slow system .
in fact , this unknown parameter @xmath78 is carried over to the slow system which is now rewritten as @xmath79 for @xmath3 sufficiently small .
since this slow system is lower dimensional than the original system , this method offers an advantage in computational cost , in addition to a benefit of using only observations on slow variables .
it is often more feasible to observe slow variables than fast variables @xcite .
assume that we have observation , @xmath80 , on the slow component @xmath23 only , and let us estimate the system parameter @xmath78 .
an estimator @xmath81 for @xmath78 may be obtained with existing techniques as reviewed in our earlier work @xcite or @xcite and references therein . here
the subscript @xmath82 indicates that the parameter estimation is conducted on the slow system .
for example , @xmath81 may be obtained by minimizing the objective function @xmath83 .
we then compare this estimator @xmath81 with the estimator , @xmath84 ( without superscript @xmath82 ) , based on the original slow - fast system - , when observations @xmath85 are available for both components @xmath86 .
for example , @xmath84 may be obtained by minimizing the objective function @xmath87 $ ] .
a stochastic nelder - mead method is used to minimize the objective functions . in the next section ,
we demonstrate this method with an example .
in this section , we demonstrate our parameter estimation method based on random slow manifolds by a simple example .
consider a slow - fast stochastic system @xmath88 where @xmath89 is a real unknown positive parameter , @xmath3 is a small positive scale separation constant , @xmath8 is a constant noise intensity , and @xmath90 is a scalar wiener process . in this system ,
the nonlinear terms are not global lipschitz .
but if additionally it has an absorbing set , we can cut - off the nonlinearities without affecting the long time , slow dynamics ( almost surely ) . indeed , for arbitrary constants @xmath92 and @xmath46 , we have @xmath93\\{}&=&\big(-\frac{2m^2}{{\varepsilon}}y^2+\frac{m^2}{300{\varepsilon}}x^2y+ \frac{2mk}{{\varepsilon}}y-\frac{mk}{300{\varepsilon}}x^2+\frac{m^2\sigma^2}{{\varepsilon}}\big)dt+2m(my - k)\frac{\sigma}{\sqrt{{\varepsilon}}}\,dw_t.\end{aligned}\ ] ] therefore , @xmath94 taking @xmath95 , @xmath96 , then we see that @xmath97 and @xmath98 thus @xmath99 by the gronwall inequality , we conclude that @xmath100 this means , for fixed @xmath3 and @xmath101 , the dynamics of the system - will eventually stay in an ellipse ( almost surely ) , i.e. there is a random absorbing set .
we can then cut - off the nonlinearities outside this absorbing set to obtain a modified system which has , almost surely , the same long time , slow dynamics as the original system @xcite . in the following calculations ,
we actually have omitted this cut - off procedure for simplicity . by the random transformation , sdes system - are converted into the following system @xmath102 therefore , there exists an @xmath53 satisfying @xmath103 whose graph is a random slow manifold for the random system - .
in fact , @xmath104 has an approximation @xmath105 ( with error @xmath106 ) , @xmath107 so the approximated slow system is @xmath108with @xmath109 .
we will illustrate that the parameter estimator for @xmath78 based on this low dimensional , reduced system is a good approximation of the parameter estimator based on the original system - .
the random slow manifold is the graph of @xmath110 , where @xmath111 is as in .
it is a curve ( depending on random samples ) .
the random orbits of the system - approaches to this curve exponentially fast . figures [ figure sdes - sms001e001][figure sdes - sms005e001 ] show some orbits of the slow - fast system - and the approximate random slow manifold @xmath112 , where @xmath113 is as in , with different @xmath8 and @xmath3 values .
( red curve ) , where @xmath113 is as in , with @xmath114 and @xmath115 : @xmath116 ( left ) and @xmath117 ( right).,title="fig:",height=257 ] ( red curve ) , where @xmath113 is as in , with @xmath114 and @xmath115 : @xmath116 ( left ) and @xmath117 ( right).,title="fig:",height=257 ] ( red curve ) , where @xmath113 is as in , with @xmath118 and @xmath115 : @xmath116 ( left ) and @xmath117 ( right).,title="fig:",height=257 ] ( red curve ) , where @xmath113 is as in , with @xmath118 and @xmath115 : @xmath116 ( left ) and @xmath117 ( right).,title="fig:",height=257 ] , where @xmath113 is as in with @xmath115 : @xmath114 ( left ) and @xmath118 ( right).,title="fig:",height=257 ] , where @xmath113 is as in with @xmath115 : @xmath114 ( left ) and @xmath118 ( right).,title="fig:",height=257 ] the deterministic nelder - mead method ( nm ) is a geometric search method to find a minimizer of an objective function @xmath119 . starting from an initial guess point , it generates a new point ( reflection point , expand point , inside / outside contraction point or shrink point ) by comparing function values , and thus get a better and better estimator until the smallest objective function value in this iteration reaches the termination value ( prescribed error tolerance ) .
the algorithm of the nm method and its improvements have been widely studied and utilized @xcite .
the method has its advantage that the objective function need not to be differentiable , and it can thus be used in various applications .
as noted in @xcite , the nelder - mead method is a widely used heuristic algorithm .
only very limited convergence results exist for a class of low dimensional ( one or two dimensions ) problems , such as in the case when the objective function @xmath120 is strictly convex @xcite . for the numerical simulation , with analytical objective functions , matlab function _
fminsearch _ can be used to find a minimizer . however , when dealing with problems with noise , nm method has the disadvantage @xcite that it lacks an effective sample size scheme for controlling noise , as shrinking steps are sensitive to the noise in the objective function values and then may lead to the search in a wrong direction .
in fact an analytical and empirical evidence is known for the false convergence @xcite on stochastic function .
so we use the stochastic nelder - mead simplex method ( snm ) @xcite to mitigate the possible mistakes in the stochastic setting .
the newly developed adaptive random search in @xcite consists of a local search and a global search .
it generates a new point and new objective function @xmath121 in the @xmath122th iteration with increasing number of the sample size scheme @xmath123 .
a proper choice for @xmath123 is @xmath124 $ ] , with @xmath125 $ ] the largest integer not bigger than @xmath126 . here
@xmath127 is the sum of @xmath123 objective function values for @xmath120 .
snm leads to the convergence of @xmath120 at search points to @xmath128 ( and thus we obtain a minimizer @xmath129 ) , with probability one .
we want to estimate the parameter @xmath78 by using both the original slow - fast system and the slow system , in order to demonstrate that the slow system is appropriate for parameter estimation , when @xmath3 is sufficiently small .
take the true value @xmath116 ( say ) and numerically solve - with an initial condition @xmath130 to get @xmath131 samples of observational data @xmath132 , @xmath133 at time instants @xmath134 , @xmath135 ( save these data ) .
take two initial guesses for the unknown system parameter @xmath137 and @xmath138 randomly and solve the original slow - fast system - , with the same initial condition @xmath130 and time points @xmath134 , @xmath139 .
thus we obtain @xmath140 and @xmath141 values which depend on @xmath78 .
take two initial guesses @xmath144 and @xmath145 randomly and solve the slow system with the same initial condition @xmath146 at the same time instants @xmath134 , @xmath139 .
we thus obtain @xmath147 which depends on @xmath144 or @xmath145 . for slow - fast system - and slow manifold reduced system , with @xmath114 and @xmath115 : true values @xmath116 ( left ) and @xmath117 ( right).,title="fig:",height=264 ] for slow - fast system - and slow manifold reduced system , with @xmath114 and @xmath115 : true values @xmath116 ( left ) and @xmath117 ( right).,title="fig:",height=264 ] and estimator @xmath136 ( top ) for slow - fast system - ; and objective function @xmath149 and estimator @xmath143 ( bottom ) for slow manifold reduced system : @xmath114 , @xmath115 and true value @xmath116.,title="fig:",height=264 ] and estimator @xmath136 ( top ) for slow - fast system - ; and objective function @xmath149 and estimator @xmath143 ( bottom ) for slow manifold reduced system : @xmath114 , @xmath115 and true value @xmath116.,title="fig:",height=264 ] + and estimator @xmath136 ( top ) for slow - fast system - ; and objective function @xmath149 and estimator @xmath143 ( bottom ) for slow manifold reduced system : @xmath114 , @xmath115 and true value @xmath116.,title="fig:",height=264 ] and estimator @xmath136 ( top ) for slow - fast system - ; and objective function @xmath149 and estimator @xmath143 ( bottom ) for slow manifold reduced system : @xmath114 , @xmath115 and true value @xmath116.,title="fig:",height=264 ] and estimator @xmath136 ( top ) for slow - fast system - , and objective function @xmath149 and estimator @xmath143 ( bottom ) for slow manifold reduced system : @xmath114 , @xmath115 and true value @xmath117.,title="fig:",height=264 ] and estimator @xmath136 ( top ) for slow - fast system - , and objective function @xmath149 and estimator @xmath143 ( bottom ) for slow manifold reduced system : @xmath114 , @xmath115 and true value @xmath117.,title="fig:",height=264 ] + and estimator @xmath136 ( top ) for slow - fast system - , and objective function @xmath149 and estimator @xmath143 ( bottom ) for slow manifold reduced system : @xmath114 , @xmath115 and true value @xmath117.,title="fig:",height=264 ] and estimator @xmath136 ( top ) for slow - fast system - , and objective function @xmath149 and estimator @xmath143 ( bottom ) for slow manifold reduced system : @xmath114 , @xmath115 and true value @xmath117.,title="fig:",height=264 ] figure [ figure sdes - sm - fs001e001 ] shows the objective functions @xmath119 and @xmath149 with true @xmath116 ( left ) and @xmath117 ( right ) .
here we used @xmath150 paths to get the expectation in the objective function . for @xmath115 and @xmath114 , figures
[ s001e001a01 ] and [ s001e001a1 ] are objective function values and estimators of each iteration for slow - fast system - ( top ) and reduced system ( bottom ) with @xmath116 and @xmath117 , respectively .
we observe that , as proved in @xcite , the objective function value @xmath119 tends to @xmath128(=0 ) , while the minimizer or the parameter estimator @xmath136 provides an accurate estimation of the system parameter @xmath78 . for sufficiently small @xmath151 ,
the objective function value @xmath149 for the slow system will also get closer and closer to @xmath152 , and the minimizer or the parameter estimator @xmath143 is a good approximation of @xmath136 .
x. kan , j. duan , i. g. kevrekidis and a. j. roberts , simulating stochastic inertial manifolds by a backward - forwand approach .
_ siam j. on applied dynamical systems _ , in press , vol . 12 , 2013 .
arxiv:1206.4954 [ math.ds ] . | a parameter estimation method is devised for a slow - fast stochastic dynamical system , where often only the slow component is observable . by using the observations only on the slow component ,
the system parameters are estimated by working on the slow system on the random slow manifold .
this offers a benefit of dimension reduction in quantifying parameters in stochastic dynamical systems .
an example is presented to illustrate this method , and verify that the parameter estimator based on the lower dimensional , reduced slow system is a good approximation of the parameter estimator for original slow - fast stochastic dynamical system . [ [ mathematics - subject - classifications-2010 ] ] mathematics subject classifications ( 2010 ) + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + primary 60h30 ; secondary 60h10 ; 37d10 . [ [ keywords ] ] keywords + + + + + + + + parameter estimation ; slow - fast system ; random slow manifold ; quantifying uncertainty ; numerical optimization |
study of the nature of light scalars @xmath3 and @xmath10 , well - established part of the proposed light scalar meson nonet @xcite , is one of the central problems of nonperturbative qcd , it is important for understanding the way chiral symmetry is realized in the low energy region and , consequently , for the understanding of confinement .
many papers have already been devoted to this subject , see , for example , refs .
@xcite .
naively one could think that the scalar @xmath3 and @xmath10 mesons are the @xmath11 p - wave states with the same quark structure as @xmath4 and @xmath12 , respectively .
but now there are many indications that the above scalars are four - quark states , see , for example , refs . @xcite and references therein .
one of these indications is the suppression of the @xmath3 and @xmath10 resonances in the @xmath13 and @xmath14 reactions , respectively , predicted in 1982 @xcite and confirmed by experiment @xcite , see @xcite .
the elucidation of the mechanisms of the @xmath15 , @xmath10 , and @xmath3 resonance production in the @xmath16 collisions confirmed their four - quark structure @xcite .
light scalar mesons are produced in @xmath16 collisions mainly via rescatterings , that is , via the four - quark transitions .
as for the @xmath4 and @xmath12 , the well - known @xmath11 states , they are produced mainly via the two quark transitions ( direct couplings with @xmath16 ) . another argument in favor of the four - quark nature of the @xmath3 and @xmath10 is the fact that the @xmath17 and @xmath18 decays proceed through the kaon loop : @xmath19 , @xmath20 , i.e. via the four - quark transition @xcite .
the kaon loop model was suggested in ref .
@xcite and confirmed by the experiment ten years later @xcite .
recently the comparative study of the scalar and tensor mesons was proposed in the @xmath21 and @xmath22 reactions ( i.e. in the timelike region of @xmath23 ) @xcite . in the present study
we consider the reaction @xmath24 in the spacelike region of @xmath23 . in 2009 belle collaboration published high - statistical data on the @xmath1 reaction @xcite .
these data revealed the specific feature of the @xmath25 cross section : it turned out sizable in the region between the @xmath3 and @xmath26 resonances , that certainly indicates the presence of additional contributions .
the experimenters took into account the putative heavy isovector scalar @xmath27 with mass about @xmath28 gev ( they called it @xmath29 ) along with @xmath3 and @xmath4 together with the polynomial coherent background @xcite . in the theoretical works refs .
@xcite @xmath3 is produced mainly by loops , and the born contribution to @xmath13 plays role of coherent background , see figs .
[ amplitudes ] , [ bornvakk ] .
the kaon formfactor @xmath30 in loop diagrams @xmath31 was introduced in those papers .
the given paper is the development of the refs .
@xcite . basing on their theoretical results , we perform new fitting of the belle data simultaneously with the kloe data on the @xmath2 decay .
we show that in the kaon loop model of the @xmath32 transition the kaon formfactor @xmath30 is not required for good data description , as well as in the @xmath33 decay and @xmath34 @xcite .
the obtained @xmath3 coupling constants are in good agreement with the four - quark model prediction @xcite .
using results on @xmath1 we predict the @xmath35 , where @xmath36 is the @xmath37 invariant mass and @xmath8 is the @xmath23 virtuality .
the measurement of the cross section @xmath35 would allow additional check of the models of the @xmath3 structure and our understanding of the mechanism of the reactions @xmath38 as well as @xmath1 .
the theoretical description of the @xmath39 reaction is in sec .
[ formulas ] .
the results on the @xmath1 data description are presented in sec .
[ results ] .
the prediction of the @xmath35 is in sec .
[ vectorexcitations ] .
it is found that the role of vector excitations ( @xmath40 , @xmath41 , etc . )
is crucial for the @xmath4 and @xmath12 production : the qcd based asymptotics of the @xmath5 ( or @xmath42 ) cross section can be reached only by taking into account cancellation of the contributions of @xmath43 and @xmath44 in @xmath8 channel with the contributions of their radial excitations .
the @xmath4 contribution is expected to be dominant at large @xmath8 , while at @xmath9 it is similar to the @xmath45 contribution .
the conclusion is in sec .
[ conclusion ] .
some details are provided in appendices i - iii . noticed misprints of refs .
@xcite are mentioned in ref .
@xmath46@xmath47.,width=566 ] [ cols="^ " , ] , @xmath48 , for the same parameters as in fig .
[ ggcrossq2 ] .
solid line @xmath49 gev@xmath50 , dashed line @xmath51 gev@xmath50 , long - dashed line @xmath52 gev@xmath50.,width=377 ] here @xmath53 and @xmath54 .
the requirement @xmath55 leads to elimination of the contribution proportional to @xmath56 , so @xmath57 is @xmath58 satisfying the qcd based asymptotics .
this asymptotics is a manifestation of the radial vector excitations .
the requirement eq .
( [ zeroing ] ) do not fix both @xmath59 and @xmath60 .
the ratio @xmath61 is the question for the experiment . in the @xmath42 reaction
the situation with asymptotics and cancellation due to radial vector excitations is similar .
let us take in eq .
( [ f2 ] ) @xmath62 mev and @xmath63 mev in agreement with ref .
@xcite .
if one takes @xmath64 together with the parameters of fit 1 , the @xmath65 and @xmath66 peaks fall synchronously with @xmath67 increase , see figs .
[ ggcrossq3 ] and [ ggcrossq3smeared ] , where the sums @xmath68 and @xmath69 are shown for different values of @xmath8 .
the @xmath66 peak starts dominating only at @xmath70 gev .
the @xmath71 is shown for different values of @xmath8 in fig .
[ ggcross1q3 ] . in this case contribution of vector excitations is relatively small at @xmath72 as one could expect from general considerations .
( [ f2 ] ) , ( [ zeroing ] ) also contain cases when contribution of vector excitations exceeds the @xmath73 and @xmath74 one even at @xmath72 .
let us consider two of them . in case @xmath75 and
@xmath76 the @xmath65 and @xmath66 peaks in @xmath77 fall synchronously for @xmath78 gev , then the @xmath66 peak starts dominating , see fig .
[ ggcrossq ] .
the @xmath71 for this case is shown in fig .
[ ggcross1q ] .
the choice @xmath79 and @xmath80 leads to fig .
[ ggcrossq2 ] : the @xmath66 peak grows at low @xmath67 and falls under the value at @xmath72 only at @xmath81 gev .
starting from @xmath82 gev the @xmath66 peak dominates over the @xmath65 peak , see fig .
[ ggcrossq2 ] .
the @xmath71 is shown in fig .
[ ggcross1q2 ] .
we do nt consider scenarios which seem doubtful .
for example , for vast region of @xmath59 and @xmath60 the @xmath83 at some @xmath84 gev .
note that variation of the parameters in table i do not provide such dramatic changes at @xmath85 as possible variation of @xmath59 and @xmath60 .
new belle experiment could clarify what scenario is realized .
how vector excitations influence on contributions to the @xmath13 amplitude not involving the @xmath66 meson is the question for separate investigation .
the kaon loop contribution probably does not change dramatically because the couplings of the radial vector excitations with kaon channel , obtained in @xcite , are small .
other contributions are not so large .
the direct transition @xmath86 is small in agreement with the four - quark model scenario .
the experimental data on the @xmath13 reaction evidence in favor of the four - quark model of the @xmath3 . the data is well described with the scenario based on four - quark model : relations eqs .
( [ fourqrelations ] ) and small @xmath87 .
the obtained values of @xmath87 mean that the @xmath88 point - like coupling gives negligible contribution to @xmath2 process .
the production ( and decay ) of the @xmath3 via rescatterings , i.e. via the four - quark transitions , is the main qualitative argument in favour of the four - quark nature of the @xmath3 . the @xmath8 dependence of the cross section obtained with specific values of @xmath59 and @xmath60 ( for example , shown in figs .
[ ggcrossq3 ] , [ ggcrossq3smeared ] and [ ggcross1q3 ] ) may be quite reliable in the region @xmath78@xmath89 gev @xcite . at higher @xmath67 the obtained dependence may be treated only as a guide .
note that at very high @xmath67 one should use qcd .
the strong influence of radial vector excitations at @xmath90 gev is considered also , see figs .
[ ggcrossq]-[ggcross1q2 ] .
we do nt use the kaon formfactor @xmath30 , introduced in refs .
@xcite , since the data can be explained without it , and the processes @xmath2 as well as @xmath91 are described without this factor too . as for comparative production of @xmath65 and @xmath66 in @xmath77 , at high @xmath92 gev
the @xmath66 contribution dominates in all variants . in the intermediate region
@xmath90 gev the @xmath65 and @xmath66 peaks fall synchronously if vector excitations contribution is relatively small at @xmath72 as one could expect from general considerations .
but in principle it is not excluded that the @xmath66 peak dominates in the intermediate region too , see figs .
[ ggcrossq ] , [ ggcrossq2 ] .
the data on @xmath93 have recently appeared @xcite .
the perspectives of study of the @xmath10 and @xmath12 comparative production in this reaction are poor because the @xmath10 peak is considerably less than the @xmath12 peak in @xmath94 already @xcite .
as it was mentioned above , the mechanism of @xmath12 production in the reaction @xmath95 is similar to the mechanism of @xmath4 production in the process @xmath96 . in ref .
@xcite , we considered in detail the changeover of the dominant helicity amplitude in the processes @xmath95 and @xmath96 with increasing @xmath8 and showed that data from ref .
@xcite could be satisfactorily described even with only one radial excitation ( @xmath40 ) .
emphasize that the best process to study the @xmath66 production is the @xmath97 reaction because @xmath98 @xcite and the background is expected to be small
. the forthcoming superkekb factory , which will have the luminosity at @xmath99 times more than the kekb one @xcite , could be the best place for investigation of the scalar and tensor mesons in @xmath37 collisions .
this work was supported in part by the russian foundation for basic research ( grants nos .
13- 02 - 00039 and 16 - 02 - 00065 ) and interdisciplinary project no .
102 of the siberian branch , russian academy of sciences .
for pseudoscalar @xmath100 mesons and @xmath101 one has : @xmath102 .\end{aligned}\ ] ] for @xmath103 @xmath104 @xmath105.\end{aligned}\ ] ] note that we take into account intermediate states @xmath106 in the @xmath3 and @xmath27 propagators : @xmath107 and the same for @xmath27 .
note that @xmath108 and @xmath109 .
the amplitude of the signal process @xmath111 is @xmath112 where @xmath113 , @xmath114 and @xmath115 are the polarization vectors of @xmath116 meson and photon , the function @xmath117 is given below .
the @xmath118 .
the matrix element of the background process @xmath119 is @xmath120 the mass spectrum is @xmath121 where the mass spectrum for the signal is @xmath122 the mass spectrum for the background process @xmath123 is @xcite : @xmath124 where @xmath125 and @xmath126 the term of the interference between the signal and the background processes is written in the following way : @xmath127 where @xmath128 the @xmath129 is additional relative phase between @xmath130 and @xmath131 , does not mentioned in eqs .
( [ a0signal ] ) and ( [ mback ] ) .
it is assumed to be constant and takes into account , for example , @xmath132 rescattering effects , see @xcite . in the @xmath19 loop model @xmath117 has the following forms : for @xmath133 @xmath134\biggr\},\end{aligned}\ ] ] where @xmath135 for @xmath136 @xmath137\biggr\}.\end{aligned}\ ] ] note that @xmath138 , see eq .
( [ ikk ] ) , for @xmath139 .
the inverse propagator of the @xmath73 meson has the following expression @xmath140 we use coupling constants @xmath141 , @xmath142 gev@xmath143 and @xmath144 gev@xmath143 , obtained with the help of ref .
@xcite data .
the function to minimize , @xmath145 function , in the present paper consists of three terms : the @xmath150 represents additional terms to reach some restrictions : lower @xmath151 to reduce influence of the analytical continuation of the @xmath152 under the @xmath153 threshold on the @xmath154 , see eq .
( [ a0signal ] ) ; coupling constants close to the relations eq . ( [ fourqrelations ] ) ; not large @xmath155 and not very large @xmath156 . the @xmath150 was used to obtain fits 1 , 3 , and 4 .
fit 2 without @xmath150 shows that the @xmath65 parameters not go far from the fit 1 results .
m. n. achasov _
( snd collab . ) , phys .
b * 440 * , 442 ( 1998 ) ; m. n. achasov _
et al . _ ,
b * 485 * , 349 ( 2000 ) ; r. r. akhmetshin _ et al . _
( cmd-2 collab . ) , phys .
b * 462 * , 380 ( 1999 ) ; a. aloisio _
( kloe collab . ) , phys .
b * 537 * , 21 ( 2002 ) .
the prediction @xmath157 kev was done in 1982 in the frame of the four - quark mit bag model .
the particle data group survey of 2014 @xcite gives the following data : @xmath158 kev and @xmath159 kev , which is an order of magnitude smaller than the @xmath16 width of the tensor @xmath11 resonance @xmath160 kev .
the prediction of the ideal @xmath11 model @xmath161 is excluded experimentally .
one can see from table i that the @xmath27 parameters are not determined well on the present stage . the nature of this particle currently can not be fixed . for example
, it is possible that @xmath27 is a complex of several scalars , effectively taken into account .
the question of the @xmath27 parameters and nature is not important for the aims of the present work . | the prediction of the cross section @xmath0 based on the simultaneous description of the belle data on the @xmath1 reaction and the kloe data on the @xmath2 decay is presented .
the production of the scalar @xmath3 and tensor @xmath4 is studied in detail .
it is shown that the qcd based asymptotics of the @xmath5 cross section can be reached by the compensation of the contributions of @xmath6 and @xmath7 with the contributions of their radial excitations in @xmath8 channel . at large @xmath8 the @xmath4 contribution
is expected to be dominant , while at @xmath9 it is similar to the scalars contribution . |
the standard model which , on some large scale , is defined as a homogeneous and isotropic solution of einstein s equations for gravitationally interacting matter , has proved to be remarkably robust against various observational challenges especially of the recent past .
it is this robustness together with a list of theoretical and observational arguments which makes it hard to see any need for an alternative to the standard model .
nevertheless , there exist some simple arguments which let the standard model appear dogmatic and a replacement overdue , while most scientific activity in the field is directed towards a consolidation of the standard model .
it is fair to say that most of the work , which is directed towards consolidation , is already based implicitly on the assumption that the standard model gives the correct picture .
i here scetch a possible dialogue which we can watch without risk of being biassed by some prejudice : we have two people who try to defend their points of view and both of them might be biassed , but both advance arguments which can be proved or disproved .
this dialogue is mirrored in an ongoing debate in the field of astrostatistics about the existence of evidence for a scale of homogeneity ( e.g. , davis 1996 and pietronero 1996 ) , a subject which is also dealt with in several contributions to this volume ( see : kerscher et al .
, martnez , sylos labini et al . ) and was the subject of a panel discussion held during the meeting .
let us start with the advocate of the standard model red : `` the ( large scale ) standard model is a solution of friedmann s equations : @xmath0 where @xmath1 is the cosmological constant , @xmath2 is related to the constant curvature of the model at time @xmath3 , @xmath4 the value of the homogeneous density at time @xmath5 , and @xmath6 is the scale function of the isotropically expanding ( or contracting ) universe . ''
also : `` this model is unstable to perturbations in the density field and/or the velocity field , respectively '' , which is the well known content of gravitational instability ; we may call this property _ local gravitational instability_. in spite of this instability , red supports the following conjecture about the global properties of the universe ( a statement which will come again in various refined versions later on ) : * conjecture * ( version 1 ) : `` the universe can be _ approximated _ by the standard model , if _ averaged _ on some large scale @xmath7 , e.g. for the inhomogeneous density field we would have : @xmath8 for all times . '' here , we may think for simplicity that the brackets @xmath9 denote euclidean spatial averages of some tensor field as a function of eulerian coordinates and time , @xmath10 .
the newtonian case serves as a good illustration .
below , we shall explain that the arguments carry over to riemannian spaces and general relativity .
green replies : `` i do nt expect that spatial averaging and time evolution commute as a result of the nonlinearity of the basic system of equations . ''
blue explains : `` if we average the cosmological fields of density and velocity at some initial time @xmath3 ( at recombination ) and use these average values ( which are remarkably isotropic according to the microwave background measurements ) as initial data of a homogeneous isotropic solution , then , e.g. , the value of @xmath4 at time @xmath5 is expected to differ from the average field @xmath11 of the inhomogeneous initial data evolved to the time @xmath5 . this has been particularly emphasized by ellis ( 1984 ) . '' indeed , green is right : the explanation of non commutativity is given by blue in terms of the _ commutation rule _ for the expansion scalar defined as the divergence of the velocity field @xmath12 ( buchert , ehlers 1997 ) : @xmath13 `` equation ( 2 ) shows that , on any spatial domain @xmath14 , the evolution of the average quantity differs from the averaged evolved one , the difference being given by a fluctuation term .
for details and discussions of what follows see ( buchert 1996 and buchert & ehlers 1997 ) ; for the relation to dynamical models see ( ehlers & buchert 1997 ) . '' also : `` equation ( 2 ) only assumes mass conservation , i.e. , we follow a tube of trajectories so that the mass in the spatial averaging domain @xmath14 is conserved in time .
this is a sensible assumption , since we want to extend the spatial domain to the whole universe later on . ''
blue goes further by specifying the local dynamical law for the expansion scalar .
this is furnished by raychaudhuri s equation : `` introducing a scale factor via the volume @xmath15 , @xmath16 ( so , we do not care about the shape of the spatial domain @xmath14 ; it may expand anisotropically ) , raychaudhuri s equation for @xmath17 may be inserted into the commutation rule ( 2 ) resulting in the _ generalized friedmann equations _ : @xmath18 with the _ effective _ source term which involves averages over fluctuation terms of the expansion , the shear scalar @xmath19 and the vorticity scalar @xmath20 : @xmath21 @xmath22 thus , as soon as inhomogeneities are present , they are sources of the equation governing the average expansion
. they may be negative or positive giving rise to an additional effective ( dynamical ) density which we may measure by the dimensionless ratio @xmath23 for @xmath24 , the source due to ` backreaction ' is , on the averaging domain , equal to that of the averaged matter density .
the effective density does , in general , not obey a continuity equation like the matter density ; an effective mass is either produced or destroyed in the course of structure formation . ''
red : `` in principle you are right , but i doubt that the effect is quantitatively significant . '' green : `` irrespective of the global relevance of this term , it will play an interesting role on scales where its value is non
negligible ; for dominating shear fluctuations the ` backreaction ' could fake a ` dark matter ' component , since the mass in the standard friedmann equation will be overestimated in this case . ''
red is going to advance a strong argument in favour of the standard model : `` in newtonian theory the ` backreaction ' term ( 4 ) vanishes by averaging over the whole universe , if the latter is topologically closed , i.e. , compact and without boundary . '' indeed , he succeeds in writing the local term @xmath26 as a divergence of some vector field @xmath27 , @xmath28 , if he assumes space to be euclidean .
`` hence , using gau s theorem , we may transform the volume integral in the average @xmath29 into a surface integral over the boundary @xmath30 of the averaging domain . for compact universes without boundary ( e.g. , a 3torus @xmath31 ) this surface integral
is zero ; we obtain @xmath32 on the torus . ''
blue : `` as a side result red proved that the currently employed models for large scale structure , analytical or n body simulations , are constructed correctly : the assumption _ of periodic boundary conditions _ is equivalent to using as the 3space a hypertorus , not @xmath33 .
these models so far have been assumed to be friedmannian on average _ by construction _ rather than derivation . ''
green adds a disclaimer : `` the above argument depends on the flatness of space . ''
red : `` but , as was shown in ( buchert & ehlers 1997 ) the generalized friedmann equations ( 3 ) also hold for irrotational flows in general relativity . ''
green : `` yes , but not the fact that the local term @xmath34 can be written as a divergence of some vector field , which makes , besides spatial curvature , a crucial difference . ''
blue illustrates this last statement by green by some technical explanations : `` if one introduces normal ( gaussian ) coordinates @xmath35 and thus foliates spacetime into flow
orthogonal hypersurfaces of constant time ( this is only possible for irrotational flows ) , then eqs .
( 3 ) also hold . since eqs .
( 3 ) are equations for scalar quantities , they are manifestly covariant and will hold in any coordinate system .
the crucial difference to the newtonian treatment , however , is the non integrability of inhomogeneous deformations ( as defined below ) : write the metric of the spatial hypersurfaces as a quadratic form @xmath36 involving the one forms @xmath37 , then it is necessary and sufficient for the metric to be flat that the one forms are exact , i.e. , @xmath38 , and the coefficients @xmath39 reduce to a deformation gradient with respect to lagrangian coordinates , @xmath40 ; in other words , the coefficient matrix @xmath41 which measures the deformations is integrable .
the non integrability of @xmath41 implies the non existence of a vector field @xmath27 .
therefore , we can not shift a volume averaged quantity to a contribution on the boundary of the averaging domain and the conclusion on the vanishing of ` backreaction ' for models with non
euclidean space sections can not be drawn using red s argument . ''
green : `` this remark by blue is bad news in so far as we expect the valid theory to be general relativity on the large scales under consideration . ''
blue : `` again , we can formulate a side result concerning current models of large scale structure : if we model structure in , e.g. , an undercritical density universe ( the total density parameter being @xmath42 ) , then the model _ has to be _ interpreted as a newtonian model .
it makes sense to speak about hypersurfaces of constant negative curvature for the average model , but in that case there currently exists no proof that the average model is friedmannian for closed , curved spaceforms .
moreover , it is then not even possible to introduce a simple hypertorus topology , since this is incompatible with a hyperbolic geometry ( compare , e.g. , lachize rey & luminet 1995 ) . ''
red summarizes the preceeding findings and reformulates his statement : * conjecture * ( version 2 ) : on some large scale @xmath7 we still may approximate the metric of spatial hypersurfaces by the flat euclidean metric .
then , we have two options ( which both are connected with the requirement of periodic boundary conditions ) : either , @xmath43 we live in a `` small universe '' ( ellis 1971 , ellis & schreiber 1986 ) , i.e. , space is genuinely compact without boundary and has a finite size @xmath44 .
this we may call _ topological closure condition _ ( option a ) . or , @xmath43 the value of @xmath45 is numerically negligible on the scale @xmath7 .
this would support the generally held view and we may call this _ technical closure condition _ ( option b ) , since then option a is a justified approximation .
green : `` i agree that a compact universe is appealing , since only then we have some hope to explore a substantial fraction of its volume ; however , it may then not have flat space sections . ''
green accepts the flatness of space as a working hypothesis in order not to complicate the discussion : `` if we do nt have option a ( in which case red s argument is exact ) , we have to examine the ` backreaction ' term @xmath46 in more detail quantitatively '' .
green develops his argument : `` the terms in @xmath46 have positive contributions ( vorticity and expansion fluctuations ) , @xmath47 , and negative ones ( shear fluctuations ) , @xmath48 '' , and `` each individual fluctuation term has fixed sign and , thus , does not vanish on _ any _ scale '' .
blue : `` an immediate consequence of the positivity of these terms is that their values may decay with scale until we reach a representative volume of the universe , but as soon as we have reached this , these terms approach a finite positive value , even on a scale on which we may assume periodic boundary conditions . ''
green concludes that `` the requirement of vanishing or smallness of the sum @xmath49 implies a _ conspiracy _ between vorticity , shear and expansion fluctuations , which is not to be expected a priori . ''
the final refinement of red s statement therefore assumes the form : * conjecture * ( version 3 ) : on some large scale @xmath7 we still may approximate the metric of spatial hypersurfaces by the flat euclidean metric . in general we may not expect that the universe is genuinely periodic on the scale @xmath7
. however , on that scale , the term @xmath50 has a negligible value : either , because : @xmath43 each of the terms @xmath51 and @xmath52 is numerically small , so that the conspiracy assumption does not matter . or , because ( if the terms are not numerically small ) : @xmath43 the inhomogeneities evolve such that @xmath53 for scales approaching @xmath7 _ and for all times_.
blue : `` both options imply assumptions on the initial fluctuation spectrum , and both formulate properties of gravitational _ dynamics _ which is , in principle , testable . ''
red : `` i agree that we want , under all circumstances , avoid `` fine tuning '' ; the standard model should be generic for a wide range of dynamical models for the evolution of inhomogeneities . ''
green now argues on the grounds of gravitational dynamics : `` up until now there is no dynamical model which includes the full ` backreaction ' ( apart from perturbative studies which may capture some of the effect see futamase 1989 , 1996 , bildhauer 1990 and bildhauer & futamase 1991a , b , as well as russ et al .
1997 ) ; the main problem to construct such a model is the following : not only the inhomogeneities affect the global expansion , but also any model for the evolution of inhomogeneities will depend on how the average evolves in time . ''
blue details : the latter is principally known from the linear theory of gravitational instability : if the universe expands faster , then the inhomogeneities have a harder time to form . here , we are faced with a nonlinear self interaction problem : we may start with a flat friedmann model as background ( the average of the ricci scalar is zero ) , and some model for the inhomogeneities relative to this background . from the inhomogeneous model
we calculate the ` backreaction ' ( this was recently attempted by russ et al .
however , if there is a nonvanishing ` backreaction ' on the global scale @xmath7 , then this procedure gives us just the first step in the sense of an iteration ; in the second step we would have a curved background ( the average of the ricci scalar is nonzero ) , and we would have to construct an inhomogeneous model for a curved background including the ` backreaction ' from the first iteration . in turn
the second iteration would yield the ` backreaction ' for this model , and the full ` backreaction ' could be calculated after @xmath54 steps of this procedure provided there is convergence to a solution .
green : `` it is clear that we are far from being able to investigate such a model . for example , in a curved background we can neither use simple periodic boundary conditions , nor can we work with the standard fourier transformation ; we would have to work with eigenfunctions on curved spaces and would have to respect the compatibility with some , in general , nontrivial topologies . ''
one `` way out '' is to cheat : green bases his further argumentation on the standard newtonian model : `` we may use the standard model which is mathematically well defined as the average over a general inhomogeneous but periodic newtonian model , and let the box of the simulation extend to very large scales . then , the ` backreaction ' can be calculated for subensembles of the simulation box on scales which we consider representative for the universe . '' this possible study will at least give us some quantitative clues of the effect ; it is the subject of an ongoing work ( buchert et al .
1997 ) which blue is going to scetch in the next section .
let us assume , in agreement with conjecture 3 by red , that the space sections are euclidean and that the inhomogeneities can be subjected to periodic boundary conditions on some large scale .
usually , this scale is set to be around @xmath55mpc / h , mainly because of limits on cpu power using n body simulations .
however , a recent analysis of the iras 1.2 jy catalogue ( see kerscher et al . ,
this volume ) has demonstrated that fluctuations in the matter distribution do not vanish on that scale .
a mock catalogue of the iras sample produced by a simulation with a box size of @xmath56mpc / h enforces the fluctuations to vanish on the periodicity scale , and the corresponding analysis of the mock sample shows disagreement in all moments ( except the first ) of the matter distribution with the observed data ( see kerscher et al .
, this volume and kerscher et al . 1997 ) .
this example shows that not only fluctuations in the average density are an indication for inhomogeneities , but averages over higher
order moments of the density field ( e.g. reflected by averaged shear fluctuations ) create huge ( phase correlated ) , possibly low amplitude structures .
thus , even if we do not argue globally about the ` backreaction problem ' ( the ` backreaction ' is zero by construction due to the assumed newtonian description and the periodicity ) , this effect has to be seriously considered on scales of current all sky surveys .
this study entails , from a dynamical point of view , a quantification of _ cosmic variance _ within the standard model .
we therefore have to run simulations of a considerably larger spatial extent ; we may use for simplicity `` truncated lagrangian schemes '' .
these schemes have been shown to agree with n body results down to scales around the correlation length ( melott et al .
1994 , wei et al . 1996 ) and , thus , are suitable tools to realize boxes of gigaparsec extent . for this purpose
it may be considered sufficient to use the `` truncated first order scheme '' ( known as tza ; `` truncated zeldovich approximation '' , coles et al . 1993 ) . in a work in progress
buchert et al . (
1997 ) consider two cobe normalized cosmogonies , standard cdm and a cdm model with cosmological constant .
both cosmogonies are realized for a box of 1.8 gpc / h with an effective resolution of @xmath57 lagrangian fluid elements .
the simulation box is then subdivided into smaller boxes and the ensemble average is taken over values of the dimensionless relative ` backreaction ' ( 4 ) .
other quantities like the expected hubble constant or the expected density parameter including ` backreaction ' are also studied both numerically and analytically .
-6.2 true cm a plot of the absolute value of @xmath58 , normalized by the global mean density , for the initial conditions scdm against scale is shown in figure 1 , which already gives a clear representation of the scale dependence of the effect under study .
the absolute values are shown here , because @xmath58 might be negative or positive in subsamples , and the overall sum is , by assumption , zero .
the absolute value is then an estimate of the expected ` backreaction ' on some scale .
however , in some subsamples the effect may be smaller or larger .
three preliminary conclusions may be drawn from the first results of this study : @xmath43 the magnitude of the ` backreaction ' source term is of the same order as the mean density and higher on scales @xmath59mpc / h for scdm .
it quickly drops to a @xmath60 effect on scales of @xmath61mpc / h . @xmath43 the magnitude of the ` backreaction ' source term is proportional to the r.m.s .
density fluctuations almost independent of scale . compared to the density fluctuations it amounts to a factor of about @xmath62 for scdm .
@xmath43 using the `` zeldovich approximation '' we can calculate analytically the ` backreaction ' @xmath63 .
this calculation shows that @xmath63 is a _ growing _ function of time in an expanding universe .
this debate brought up one fundamental result which supports red : any inhomogeneous newtonian cosmology , whose flat space sections are confined to a length
scale @xmath7 on which the matter variables are periodic , averages out to the standard model . on this global scale
there is no ` backreaction ' and the cosmological parameters of the homogeneous
isotropic solutions of friedmann s equations are well defined also for the average cosmology on that scale .
setting up simulations of large scale structure in this way is correct , and a global comoving coordinate system can be introduced to scale the whole cosmology .
this validates the common way of constructing inhomogeneous models of the universe . on the other hand ,
green s arguments initiate two well justified ways of saying that this architecture is `` forced '' due to the settings of a ) excluding curvature of the space sections , and b ) requiring periodic boundary conditions : one way is to analyze the effect of ` backreaction ' on scales smaller than the periodicity scale , and base this analysis upon the standard model , however , by extending the spatial size of periodicity to very large scales .
the results obtained in the framework of a well tested approximation scheme are three fold : first , they show the importance of the influence of the inhomogeneities on average properties of a chosen spatial domain , _ although _ the `` forcing conditions '' bring the effect down to zero on the scale @xmath7 .
second , the ` backreaction ' value is numerically small on large scales , but it is _ always _ larger than the r.m.s .
density fluctuations ; in other words : taking the amplitudes of density fluctuations serious ( e.g. by normalizing the cosmogony on some large scale ) always implies the presence of , e.g. , shear fluctuations which are neglected on that scale in the standard model .
third , they show that the effect is a growing function of time . from the latter result
we may establish the notion of _ global gravitational instability _ of the standard model as opposed to the well known local instability : it states that , as soon as the ` backreaction ' has a non zero value at some time
, this value will be increasing ; the average model drifts away from the standard model .
the second `` forcing '' is due to the newtonian treatment .
being justified on smaller scales , a newtonian model is expected to fail just when we approach the _ large _ scales of periodicity which we have to consider to justify neglection of the ` backreaction ' effect .
setting up a general relativistic model unavoidably implies the presence of local curvature and will , in general , yield a nonvanishing average curvature for inhomogeneous models .
neither simple periodic boundary conditions can be employed , nor can be proved that the ` backreaction ' effect should vanish globally , at least for compact space sections without boundary .
both , red and green , are right , but if red s assumptions are weakened , the resulting cosmology has much richer properties and can not be confined to a simple box : it will take its additional degrees of freedom to evolve away from the standard model .
green s more general view suffers from the fact that an alternative model is yet not formulated , but it is definitely within reach .
i would like to thank martin kerscher , herbert wagner and arno wei for discussions , arno wei for his allowance to present fig.1 prior to publication of a common work .
some aspects of this contribution were subject of a letter correspondence with jrgen ehlers , to whom i am also thankful for discussing the manuscript .
bildhauer s. 1990 , prog .
84 , 444 bildhauer s. & futamase t. 1991a , m.n.r.a.s .
249 , 126 bildhauer s. & futamase t. 1991b , g.r.g . 23 , 1251 buchert t. 1996 , in `` mapping , measuring and modelling the universe '' , valncia 1995 , p. coles , v.j .
martnez & m.j .
bordera ( eds . ) , asp conference series , pp .
349356 buchert t. , & ehlers j. 1997 , astron .
320 , 1 buchert t. & kerscher m. , wei a.g .
1997 , work in progress coles p. , melott a.l . & shandarin s.f .
1993 , m.n.r.a.s .
260 , 765 davis m. , 1996 , in `` critical dialogues in cosmology '' , princeton 1996 , n. turok ( ed . )
, in press ehlers j. & buchert t. 1997 , g.r.g . , in press ellis g.f.r .
1971 , g.r.g . 2 , 7 ellis g.f.r .
1984 , in proc .
10th international conference on `` general relativity and gravitation '' , b. bertotti et al .
, dordrecht : reidel , 215288 ellis g.f.r . , & schreiber g. 1986 , phys .
lett . a 115 , 97 ellis g.f.r . , ehlers j. , brner g. , buchert t. , hogan c.j .
, kirshner r.p . , press w.h . , raffelt g. , thielemann f k . &
van den bergh s. 1997 , in dahlem workshop report es19 `` the evolution of the universe '' , berlin 1995 , s. gottlber & g. brner ( eds . ) , chichester : wiley , in press futamase t. 1989 , m.n.r.a.s .
237 , 187 futamase t. 1996 , phys .
d. 53 , 681 kerscher m. , schmalzing j. , buchert t. & wagner h. 1997 , m.n.r.a.s . , submitted lachize rey m. & luminet j.p .
1995 , phys .
254 , 135 melott a.l .
, buchert t. & wei a.g .
1994 , astron .
293 , 641 mukhanov v.f .
, abramo r.w . &
brandenberger r.h . , 1997 ,
78 , 1624 pietronero l. 1996 , in `` critical dialogues in cosmology '' , princeton 1996 , n. turok ( ed . ) , in press russ h. , soffel m.h .
, kasai m. & brner g. 1997 , phys .
d , in press wei a.g . , gottlber s. & buchert t. 1996 , m.n.r.a.s . | the averaging problem for inhomogeneous cosmologies is discussed in the form of a disputation between two cosmologists , one of them ( red ) advocating the standard model , the other ( green ) advancing some arguments against it .
technical explanations of these arguments as well as the conclusions of this debate are given by blue .
-0.8 true cm |
the 2d electron systems ( 2des ) present in gaas or si / sio@xmath14 structures access a wide range of electron densities , providing a wealth of experimental observations@xcite .
the 2d electrons reside in the x - y plane and also have a transverse ( @xmath5-dependent ) density @xmath15 which is confined to the lowest subband of the hetero - structure@xcite .
the 2d character arises since the higher subbands are sufficiently above the fermi energy @xmath16 .
then the physics depends only on the `` coupling parameter '' @xmath17 = ( potential energy)/(kinetic energy ) , the @xmath5-distribution @xmath15 , the spin - polarization @xmath2 , and the temperature @xmath3 which has to be significantly smaller than the fermi energy @xmath16 to preserve the 2-d character .
the @xmath17 for the 2des at the density @xmath18 is equal to the mean - disk radius @xmath19 per electron , expressed in effective atomic units which depend on the bandstructure mass @xmath20 and the `` background '' dielectric constant @xmath21 .
the @xmath5-motion in the lowest subband may have widths of @xmath22 , in gaas when @xmath1 is @xmath23 , in heterojunction insulated - gate field - effect transistors ( higfet ) which have been an object of recent studies@xcite .
similarly , other nanostructures ( e.g. , quantum dots ) contain electrons confined to a micro - region in the @xmath6-@xmath7 plane , and have a sizable @xmath5 extension@xcite .
hence layer - thickness effects are important in many areas of nanostructure physics .
appropriate correlation functionals@xcite for such systems are still unavailable , even though the exchange functional for fang - howard distributions is known @xcite .
layer thickness effects are a long standing probe of exchange and correlation theories in 3d electron slabs@xcite .
the relevance of the finite size of the 2d layers had also been considered within the quantum hall effect@xcite , and more recently in the context of the @xmath13-factor and the effective mass @xmath12 of 2-d layer@xcite . in the early days of the application of diagrammatic perturbation theory ( pt ) to condensed - matter problems
, it was normal to attempt to calculate various many - body properties like the effective mass @xmath12 , the effective @xmath13-factor @xmath24 , and corrections to the total energy using perturbation methods .
the need to go beyond the random - phase approximation ( rpa ) was rapidly appreciated and lead to the work of hubbard , rice , vosko , geldart and others@xcite .
the common experience with the generalized rpa method , when applied to 3d electrons and to ideally thin 2d layer is that it predicts a spin - polarized phase at unrealistic high densities ( low @xmath1 ) , while quantum monte carlo ( qmc ) simulations suggest a density near @xmath25 25 - 26 in 2d .
rpa methods lead to negative pair - distribution functions ( pdfs ) , incorrect local - field corrections in the response functions , and disagreement with the compressibility sum rule and other formal conditions .
recent attempts to calculate the effective mass @xmath12 for ideally thin layers@xcite show strong disagreement with the @xmath12 obtained from qmc simulations@xcite .
in fact , the main thrust of the programs of singwi , tosi et al .
( stls)@xcite , and ichimaru et al.@xcite was to overcome the shortcomings of the rpa - like approach via non - perturbative methods .
many - body calculations where different parts of the calculation ( e.g , vertex corrections , local - field corrections , etc . ) are obtained from different sources , ( e.g , vertex corrections from a model , local - field corrections or correlation energies from qmc , and some other quantities fitted to sum rules etc . ) and combined together , have also appeared . unless the same `` mixture '' is used to calculate a multitude of properties and shown to lead to consistent results , such theories have to be treated with caution .
we have introduced an approximate method for strongly correlated quantum systems where the objective is to work with the pdf of the quantum fluid , generated from an equivalent classical coulomb fluid whose temperature @xmath26 is chosen to reproduce the correlation energy of the original quantum fluid at @xmath27 .
the classical pdfs are calculated via the hyper - netted - chain ( hnc ) equation , and the method is called the chnc . as the method has been described in previous publications@xcite , and successfully applied to a variety of problems@xcite , only a brief account is given here , mainly to help the reader .
the pdfs are obtained from an integral equation which can be recast into a classical kohn - sham form where the correlation effects are captured as a sum of hnc diagrams and bridge diagrams .
@xmath28 the temperature of the classical fluid @xmath29 is chosen such that at @xmath27 the calculated classical @xmath30 recovers the known correlation energy of the fully spin - polarized 2d electron fluid at the given density .
this fitting has been done in ref .
@xcite , and @xmath26 is known as a function of @xmath1 . at finite-@xmath3 , the classical fluid temperature @xmath31 is taken to be @xmath32 this @xmath31 is used for all spin polarizations . in eq .
[ chnceq ] the indices @xmath33 label the spin states .
the @xmath34 is chosen to ensure that @xmath35 reduces to the explicitly known non - interacting pdf , @xmath36 , when the coulomb interaction @xmath37 is switched off .
thus @xmath34 takes care of pauli - exclusion effects and ensures that the `` fermi - hole '' is exactly recovered@xcite . also , @xmath38 is a sum of terms that appear in the hyper - netted - chain equation , while @xmath39 contains 3-body and higher - order diagrams not contained in the nodal term @xmath40 .
the latter depends implicitly on @xmath35 , and is evaluated via the ornstein - zernike equation .
the bridge term is very difficult to evaluate , but the hard - sphere fluid provides a good approximation .
that is , in 2d , we specify @xmath39 by specifying the hard - disk radius @xmath41 , or equivalently the packing fraction @xmath42 , and use the percus - yevik approach@xcite .
as the parallel - spin three - body clusters are suppressed by the pauli exclusion , we use only a single bridge function , viz .
, @xmath43 .
this makes the bridge interaction effectively independent of @xmath2 .
bulutay and tanatar , and also khanh and totsuji@xcite , have examined variants of chnc without a bridge function ( this is some what equivalent to neglecting back - flow terms in qmc simulations of 2d systems ) .
the hard - sphere radius @xmath44 , and the packing fraction @xmath42 are given by@xcite : @xmath45 a plot of the bridge function for a few typical cases is given in fig .
[ bridge ] . for @xmath46 and @xmath47 .
the effect of changing the temperature is also shown .
, width=340,height=377 ] the chnc method was applied to the 3d and 2d electron fluids@xcite , to dense hydrogen fluid@xcite , and also to the two - valley system in si / sio@xmath14 2des@xcite .
in each case we showed that the pdfs , energies and other properties obtained from chnc were in excellent agreement with comparable qmc results .
the advantage of the chnc method is that it affords a simple , semi - analytic theory for strongly correlated systems where qmc becomes prohibitive or technically impossible to carry out .
the classical - fluid model allows for physically motivated treatments of complex issues like three - body clustering etc .
, which are difficult within quantum methods .
the disadvantage , typical of such many - body approaches , is that at present it remains an `` extrapolation '' taking off from the results of a model fluid .
the fully spin - polarized infinitely - thin uniform 2des is the model fluid@xcite , as the @xmath8 of this one - component system is accurately known .
in this study we use a method of replacing the inhomogeneous electron distribution @xmath48 by a homogeneous density system@xcite i.e. , a constant - density approximation ( cda ) where the transverse density @xmath15 is a constant within a slab of width @xmath4 , and zero outside .
the cda avoids difficulties associated with gradient expansions noted in ref.@xcite .
also , the cda presents a unified approach to quasi-2d distributions like the fang - howard model@xcite , the quantum - well model etc . @xmath0 for such distributions is calculated as a function of the spin polarization @xmath2 , 2-d density parameter @xmath1 , the cda width @xmath4 and the temperature @xmath3 .
this enables us to determine physical quantities related to the landau fermi liquid parameters .
thus the spin susceptibility enhancement , the effective mass @xmath12 , and the land @xmath13 factor for the quasi 2d electrons are presented .
the transverse distribution @xmath15 of the 2d electrons is given by the square of the lowest subband wavefunction @xmath49 of the heterostructure , calculated within the envelope approximation .
the nature of the materials used ( e.g , si / sio@xmath14 or gaas ) and the doping profiles determine the confining potential and the electron density @xmath15 in the z - direction .
typically , @xmath15 may be modeled by one of the following forms . @xmath50
the second and third are frequently used approximate ( but adequate ) models , while we present the fourth model , the cdm .
this is an excellent constant - density approximation ( to be called the constant - density approximation , cda ) to generate the effective 2-d potential arising from most @xmath15 distributions , on suitably choosing @xmath4 .
the fang - howard ( fh ) form@xcite , contains the parameter @xmath51 , and is normalized in the range @xmath52 to @xmath53 .
the fh parameter @xmath51 is such that @xmath54 , where @xmath55 is the 2d electron density @xmath18 , and @xmath56 is the depletion charge density@xcite .
the material parameters are contained in the effective bohr radius @xmath57 defined in terms of the usual constants , @xmath21 and @xmath20 being the dielectric constant and the band mass . for the devices used in ref .
@xcite , the depletion density has been reported to be negligible@xcite
. then @xmath58 , in atomic units .
we denote the coulomb potential in an infinitely thin layer by @xmath59 , while @xmath60 is used for the effective 2-d potential of a thick layer .
the effective 2d - coulomb potential @xmath60 between two electrons having coordinates ( @xmath61 ) and ( @xmath62 ) , with @xmath63 is given by , @xmath64^{1/2}}\\\ ] ] here @xmath65 is @xmath53 for fh , while @xmath66 for the others .
the potential @xmath67 and the form factor @xmath68 reflects the effect of the @xmath5-extension of the density .
there is no analytic form for @xmath68 in the fang - howard case , while the @xmath69-space form , @xmath70 is known@xcite .
if the dielectric constants of the barrier and well material were assumed equal , then the fang - howard form factor is : @xmath71[1+\frac{q}{b}]^{-2}\ ] ] here we derive a potential @xmath72 for the constant - density model(cdm ) which is , to an excellent approximation _ electrostatically equivalent _ to the the 2d potential for any reasonable @xmath15 .
these fh - type distributions are themselves convenient fits to the self - consistent schrodinger solutions and have uncertainties of a few percent .
the cda holds well within such limits .
the potentials are explicitly shown in fig .
[ pots ] for the fh form where we have taken an extreme example with @xmath73 .
the method of replacing an inhomogeneous distribution by a uniform distribution is suggested by the observation that the non - interacting total pair - correlation function @xmath74 has the form @xmath75 , where @xmath76 is the density -profile around the fermi hole . in our case
we wish to replace the inhomogeneous @xmath15 by a constant - density distribution @xmath77 which has the same electrostatic potential in the 2-d plane as @xmath15 .
@xmath78 since the subband distribution is normalized to unity , the width @xmath4 of the constant - density slab is simply @xmath79 .
starting from different objectives , gori - giorgi et al.@xcite have already proposed this formula for determining an average density for an inhomogeneous density , in the context of pair - distributions in 2-electron atoms .
we have also shown the utility of the cda in estimating the correlation energy of the 2des in the wigner - crystal phase@xcite .
the @xmath4 of the cda is somewhat different from the `` thickness '' @xmath80 often assigned to the fh distribution .
in fact , the constant - density slab width @xmath4 for the fang - howard @xmath51 is given by @xmath81 .
form with the quasi-2d potential from the fang - howard distribution for the fh parameter @xmath51 = 0.1 , and 0.3 and the equivalent slabs at @xmath4 = 53.3 and 17.8 a.u .
respectively.,width=340,height=377 ] the quasi-2d potential for a cdm of width @xmath4 is given by @xmath82 w(r)&=&v(r)f(s),\;\ ; s = r / w , \;\ ; t=\surd{(1+s^2 ) } \\ f(s)&=&2s\left[\log\frac{1-t}{s } \,\,+1-t\right ] \end{aligned}\ ] ] this potential tends to @xmath83 for large @xmath84 , and behaves as @xmath85 for @xmath86 .
thus the short - range behaviour is logarithmic and weaker than the coulomb potential .
the @xmath87-space form of the cdm potential is : @xmath88 the form factors @xmath89 and @xmath90 tend to unity as @xmath91 .
these @xmath84-space and @xmath87-space analytic forms of the cdm lead to analytic formulae for the fh form . in our work
we assume that a given distribution has been replaced by an equivalent uniform - slab distribution , and only the final @xmath60 potential enters into the exchange - correlation calculations ( the numerical work has been checked via direct calculations as well ) . in the case of gaas - higfets ,
if @xmath56 could be neglected , the @xmath1 parameter specifies the @xmath51 parameter and hence the width @xmath4 of the cdm
. then @xmath92 and @xmath93 .
the exact exchange free energy @xmath94 for 2d layers of finite thickness can be readily evaluated using the quasi-2d potential @xmath60 and the noninteracting pair - distribution functions @xmath95 of the 2-d fluid .
only the parallel - spin case @xmath96 is relevant . also , @xmath95 for a slab of finite thickness is identical to that for an ideally thin 2d layer , both at @xmath27 and at finite-@xmath3 .
in fact , we find that the @xmath3 dependence of the @xmath94 of layers of finite thickness is very close to that of the ideally thin case . the first - order unscreened exchange free energy @xmath94 consists of @xmath97 , where @xmath98 denotes the two spin species .
at @xmath27 these reduce to the exchange energies : @xmath99 here @xmath100 , and @xmath101 .
then the exchange energy per particle at @xmath27 , i.e. , the total @xmath102 becomes @xmath103\ ] ] where @xmath104 and @xmath105 are the fractional compositions @xmath106 of the two spin species .
we define a reduced temperature @xmath107 , @xmath108 , and the species - dependent reduced chemical potentials @xmath109 by @xmath110 , reduced temperatures @xmath111 and @xmath112 , based on the two fermi energies @xmath113 and @xmath114 which are @xmath115 .
then we have : @xmath116 the @xmath117 is the fermi integral defined as usual : @xmath118 the @xmath110 are given by @xmath119 in the paramagnetic case eq .
[ exf ] reduces to the result given by isihara et al .
( see their eq.3.4 ; they use a different definition of the fermi integral ) . for small values of @xmath120 ,
the exchange energy is of the form , @xmath121\ ] ] the total exchange free energy is @xmath122 .
the accurate numerical evaluation of eq .
[ exf ] requires the removal of the square - root singularity by adding and subtracting , e.g , @xmath123 for the case where @xmath42 is negative , and @xmath124 , and so on .
a real - space formulation of @xmath94 = @xmath125 using the zeroth - order pdfs fits naturally with the approach of our study .
thus @xmath126 here @xmath127 .
in the non - interacting system at temperature @xmath3 , the antiparallel @xmath128 , viz .
, @xmath129 , is zero while @xmath130 ^ 2\ ] ] here * k * , * r * are 2-d vectors and @xmath131 is the fermi occupation number at the temperature @xmath3 . at @xmath27 @xmath132 where @xmath133 is a bessel function . as a numerical check ,
we have evaluated the exchange free energy by _ both _ methods , i.e. , via k - space and r - space formulations .
we present a convenient analytic fit to the exchange free energy which is a universal function @xmath134 , for arbitrary @xmath2 .
that is , the same function applies to any component , on using the reduced fermi temperature of the spin species .
the total @xmath94 is obtained by adding both spin contributions .
the analytic fit is : @xmath135 the fit coefficients @xmath136 are 3.27603 , 4.81484 , 3.33100 , 6.51436 .
the temperature @xmath137 is @xmath138 , appropriate to the spin polarization .
the exchange effects in the 2des decay more slowly with temperature than in the 3d case where a tanh@xmath139 factor appears in eq .
3.2 of ref . .
the above form does not explicitly contain the low - temperature logarithmic term , but it reproduces the value of 0.99382 at @xmath140 , while the numerical integration gives 0.9939497 .
similarly , at @xmath141 , 10 and 30 the fit ( integral ) returns 0.63839 ( 0.63839 ) , 0.22999 ( 0.22990 ) , and 0.13421 ( 0.13410 ) respectively .
the @xmath27 exchange energy is modified by the layer thickness @xmath4 .
the expression for @xmath95 depends only on @xmath142 . similarly , the quasi - potential @xmath60 depends on the reduced variable @xmath143 .
hence the exchange energy of the quasi-2d layer depends only on the ratio @xmath144 .
the exchange energy per electron at a density @xmath18 , given by @xmath1 , polarization @xmath2 , in a layer of width @xmath4 is given by : @xmath1452\pi rdr ei_x(r_s,\zeta , w)=\frac{1}{2}n r_s\int_0^\infty 2\pi dxh^0(x,\zeta)f(x / w_s)\ ] ] even though we have analytic forms for @xmath60 and @xmath74 as well as their fourier transforms , we have not found a convenient analytic result for the exchange energy at @xmath27 .
however , the results can be parametrized by simple analytic - fit formulae : @xmath146 here @xmath147 is the exchange energy of the ideally thin system given by eq .
[ ex_t=0 ] .
the ratio @xmath148 is a measure of the reduction in the exchange energy due to the thickness effect .
since the effect depends on @xmath144 , for a given thickness , the effect is greater for _ high density _ samples . if the depletion density @xmath56 in higfets could be neglected , and if the exchange - correlation energy @xmath0 is _ not _ included in the energy minimization which determines the fang - howard parameter @xmath51 , then @xmath149 . the inclusion of @xmath0 in self - consistently determining @xmath51 changes @xmath51 by @xmath150 for low @xmath1 , but the effect becomes less important at higher @xmath1 . at @xmath1=1 , and 10 for @xmath151 , the ratio @xmath152 is 0.652 and 0.794 respectively .
the reduction from the ideally - thin 2d form is clearly substantial .
the exchange free energy @xmath153 at finite-@xmath3 , for layers with thickness @xmath4 is found to be adequately approximated by the temperature factor of the ideally - thin system .
however , in calculating the effective mass from the finite-@xmath3 energies , we make independent calculations near @xmath27 and do not use the fit given here . .
parameters fitting the exchange energy ratio as a function of the layer width ratio @xmath144 . [ cols="^,^,^,^,^,^",options="header " , ] [ fc_table ] .
the higfet @xmath4 varies as @xmath154 and is @xmath155 18 near the spin - phase transition .
, width=340,height=377 ] qmc simulations as well as chnc calculations show that there is a spin - polarization transition ( spt ) in the ideally - thin 2d electron fluid near @xmath156 . on the other hand ,
the correlation contributions dominate over the exchange energy in the 2-valley 2d system in si mosfets , and the spt is suppressed@xcite . the rapid increase in @xmath12 , with @xmath24 remaining unchanged while @xmath1 is increased , observed by shashkin et al.@xcite was found to be consistent with this picture@xcite . in finite - thickness 2d layers ,
as the cdm width @xmath4 increases , the location of the spt is pushed to higher values , as seen in fig .
[ sptfig ] . in the case of higfets used by , e.g. ,
tan et al.@xcite , the width @xmath4 increases with @xmath1 , but only as @xmath154 .
thus at @xmath1=26 , the @xmath4 is only 18.4 , and the spt remains intact and occurs at a somewhat higher @xmath1 , as shown by de palo et al@xcite , and also in fig .
[ sptfig ] . a natural consequence of delaying the spt is to decrease the spin - susceptibility enhancement .
the effective thickness of the quasi-2d layer can be increased by suitably designing the shape of the potential well , or including an additional subband , and in this case the spt can be circumvented . however , a discussion of higher subband effects is outside the scope of this study .
= @xmath157 as a function of @xmath1 for the higfet . the experimental data of zhu et al.@xcite are also shown .
( b)variation of @xmath157 as a function of the spin - polarization for a higfet .
, width=340,height=377 ]
the results for the exchange - correlation free energy @xmath158 for the ideal 2d system and the thick - layer system contain all the information needed to calculate the spin - susceptibility enhancement , the effective mass @xmath12 and the effective land factor @xmath24 , for the ideal system and the thick layer .
in fact , the following quantities are calculated from the respective second derivatives of the energy .
@xmath159 } { \left[\partial^2 f_0(t)/\partial t^2\right ] } \\ \chi_p/\chi_s&= & ( m^*g^*)^{-1 } = 1 + \frac { \left[\partial^2 f_{xc}(\zeta)/\partial \zeta^2\right ] } { \left[\partial^2 f_0(\zeta)/\partial \zeta^2\right ] } \end{aligned}\ ] ] we use the available qmc results for the ideal 2d exchange - correlation energy at @xmath27 , and where convenient , the qmc pair - distribution functions at @xmath27 as parametrized by giri - giorgi et al@xcite .
the chnc is used to obtain the pair - functions for situations ( e.g , at finite-@xmath3 and finite thickness ) where the qmc data are simply not available or difficult to use . in most cases , replacing the qmc - pdfs with the chnc ones , or using the `` unperturbed-@xmath13 '' approximation leads to relatively small changes .
the exception is in the calculation of @xmath12 , where the `` unperturbed-@xmath13 '' approximation , eq .
[ unpertg ] , is inadequate . in fig .
[ m2d - id ] we present our results for the ideal-2des .
no experimental results are available for this case , but limited qmc simulations@xcite as well as results from diagrammatic perturbation theory ( pt ) @xcite are available
. of an ideal 2d layer ( @xmath160 ) obtained from chnc are compared with the the quantum monte carlo data of ref .
@xcite and the perturbation theory calculations of zhang et al.@xcite , and asgari et al.@xcite , i.e. , their @xmath161 calculation .
, width=340,height=377 ] the chnc based @xmath12 values have an error of at most @xmath162 2% .
the ideal 2d - layer @xmath12 shows a rapid rise around @xmath1=2 to 5 , in good agreement with the four values from qmc , and then slows down in strong contrast to the @xmath12 proposed by asgari et al .
, and zhang et al . we have displayed the model denoted @xmath161 by asgari et al . , as being their optimal choice from among the many models given in ref .
@xcite , where they also contest the analysis of zhang et al .
the pt approaches are partly semi - phenomenological in that qmc data are input into local - field factors and other ingredients of the calculation ; the choice of the vertex functions , treatment of on - shell or off - shell effects , whether to use self - consistent propagators or not , etc . , are components of the prescription used by different workers .
the strong disagreement of the pt - based @xmath12 with the qmc - based @xmath12 is notable .
the chnc method has some similar uncertainties , especially in the use of a percus - yevik hard - sphere bridge function @xmath163 to capture the back - flow like three - body contributions to the pdfs and the total energy . as seen in fig .
[ bridge ] , the @xmath3 dependence of @xmath163 seems quite small , and our initial calculations of @xmath12 , reported in ref .
@xcite were based on the zero-@xmath3 form of @xmath163 .
this leads to an @xmath12 which drops slightly below unity and remains there . in this calculation
we have used the proper @xmath3-dependent bridge function and the calculated @xmath12 is in good agreement with the qmc - based @xmath12 .
this might be somewhat coincidental , as the qmc results are also based on sensitive approximations .
however , it means that we do have a @xmath163 which is consistent with current qmc results , and hence may be used with greater confidence in studying thick-2d systems . another point in favour of our model of @xmath163
is seen in the local - field factor ( lff ) of the ideal 2des response function .
a study of the lff of the 2d response@xcite shows that the formation of singlet - pair correlations is essentially complete by @xmath164 , and after that the structure of the fluid remains more or less unchanged , until the spt is reached .
the hard - sphere model of @xmath163 provided a satisfactory description of the short - ranged features of the 2des - lff . the rapid rise in @xmath12 up to @xmath164 and
the subsequent slow - down is probably related to the formation and persistence of the singlet structure in the 2d fluid revealed by the form of the lff . in a higfet .
the pt results of zhang et al . , and asgari et al . , for the higfet are close to the ideal-2d results from qmc .
the `` unperturbed-@xmath13 '' approach to the energy using chnc - based pdfs gives a slight reduction from the ideal 2d .
for the full chnc calculation , see fig .
[ scfmstr],width=340,height=377 ] in fig .
[ m2d - qua ] we present a comparison of various theoretical calculations of the effective mass @xmath165 of the electrons in the higfet .
the pt calculations of zhang et al . , and asgari et al . ,
show a strong decrease of @xmath12 from the pt values in the ideal 2des . our calculations , using the `` unperturbed-@xmath13 '' approximation , eq .
[ unpertg ] , lead to a @xmath165 which is only slightly reduced by the thickness effect .
this @xmath165 curve is in close to that of asgari et al .
this is clearly a numerical accident . according to asgari et .
al . , the difference between the ideal @xmath12 and @xmath165 increase as @xmath1 increases . in our calculation
using the `` unperturbed-@xmath13 '' approximation , the difference , already quite small , seems to diminish as @xmath1 increases .
in fact , as @xmath1 increases , the ratio @xmath10 of the higfet layer decreases and the thickness effect may be expected to decrease , unless the difference between @xmath12 and @xmath165 is driven by some other effect .
this `` other effect '' is revealed by giving up the `` unperturbed-@xmath13 '' approximation , and using the full thick - layer 2des pair - distribution function at finite @xmath3 , calculated using the chnc , to evaluate the total free energy @xmath166 of the quasi-2d system , and hence the @xmath165 . in fig .
[ scfmstr ] we display the pdf of the quasi-2des of a higfet at @xmath47 , and compare it with the pdf of the ideal 2des .
the difference between the ideal and quasi systems is embodied in the form factor @xmath68 .
the reduced coulomb repulsion at small-@xmath84 leads to _ a large pile - up of electrons around the electron at the origin .
_ this means , the electron has to drag this charge pileup and this contributes an enhanced @xmath12 to the thermodynamic and transport properties of the quasi 2des .
the experimental results of tan et .
for @xmath12 show an increase of @xmath155 150% between @xmath1=3 and @xmath167 .
our results from the full chnc calculation , as well as the perturbation results of asgari et al . , are shown in the lower panel of fig .
[ scfmstr ] . given the failure of the pt calculations to reproduce the qmc - based @xmath12 for the ideal 2d , it is difficult to evaluate the reliability of the pt - based @xmath165 .
the pt - overestimate of @xmath12 of the ideal 2d is typical of the rpa - like character of these theories which are likely to predict spin transitions at relatively high densities . also , we believe that if the same pt prescriptions were used to evaluate the one - top value @xmath168 of the pdfs of the 2des and the quasi-2des , another measure of the short - comings of the pt methods would be revealed .
the product @xmath157 is given by the ratio of the static spin susceptibility @xmath169 to the ideal ( pauli ) spin susceptibility @xmath170 .
the long wavelength limit of the static response functions are connected with the compressibility or the spin - stiffness via the second derivative of the total energy with respect to the density or the spin polarization@xcite .
de palo et al.@xcite have calculated @xmath157 from the qmc pair distribution functions and shown that they obtain quantitative agreement with the data for very narrow 2-d systems@xcite as well as for the thicker systems found in higfets@xcite .
the chnc pdfs are close approximations to qmc results , and when used in eq .
[ unpertg ] , yield correction energies which are in good agreement with the energies obtained by de palo et al@xcite . in ref .
@xcite we showed that the rapid enhancement of @xmath157 in si / sio@xmath14 2des is a consequence of the increase in @xmath12 with @xmath1 , and that the @xmath24 does not increase with @xmath1 because there is _ no spin - phase transition _ in the 2-valley case . in the higfet system
there is a slightly delayed spt , as seen in fig .
[ susfig ] .
hence @xmath157 increases with @xmath1 , while @xmath12 also increases quite rapidly , due to the enhanced `` on - top '' correlations shown ( fig .
[ scfmstr ] ) in the pdf of the quasi-2des .
the resulting @xmath24 of the higfet is shown in fig .
[ gstar ] . at @xmath47 and @xmath171=0.1 , for an ideal 2des , and for a higfet calculated using chnc .
( b)the effective mass @xmath165 in a higfet from fully self - consistent chnc calculations finite-@xmath3 .
the experimental data are from tan et al@xcite .
, width=340,height=415 ] -factor is obtained from the qmc @xmath157 of ref .
@xcite , divided by the chnc ideal 2d @xmath12 .
the experimental higfet @xmath172 ( boxes ) is from the @xmath157 of zhu et al . , divided by the @xmath12 of tan et al .
the inset shows the spin - susceptibility enhancement @xmath157 from the @xmath173 calculated from eq .
[ unpertg ] , where the ideal 2d @xmath174 is used , and from the full chnc calculation using the @xmath175 consistent with the quasi-2d potential .
, width=302,height=340 ]
we have presented a detailed study of the effect of many - body interactions in quasi-2d electron layers using a single theoretical framework which involves the calculation of the pair - distribution functions of the system via a classical representation of the quantum fluid . a procedure for replacing the inhomogeneous transverse distributions via a constant - density model , i.e. , an equivalent _ homogeneous _ distribution , has also been demonstrated .
easy to use parametrized fit formulae for the exchange energy at zero and finite-@xmath3 have been presented . a simple numerical scheme for calculating the correlation energy of a thick 2d layer , , _ via _ a 3-d slab model combined with a 1-d rod model , has also been demonstrated .
we find that the thickness effect on the spin - phase transition etc .
, provides a clear picture of the changes in the spin - susceptibility enhancement leading to a strong increase in the @xmath13-factor , while @xmath12 is increased due to the enhancement of the `` on - top '' correlations arising from the reduction of the coulomb potential in thick layers .
however , unlike in the case of the effective mass data for si / sio@xmath14 systems@xcite , these results do not provide good quantitative agreement with the effective - mass data for higfets recently reported by tan et al
. this may be due to our use of the ideal 2des bridge function even for the higfets .
99 s. v. kravchenko and m. p. sarachik , rep . prog.phys . * 67 * , 1 ( 2004 ) t. ando , b. fowler , and f. stern , rev .
* 54 * , 437 ( 1982 ) j. zhu , h. l. stormer , l. n. pfeiffer , k. w. baldwin , and k. w. west , phys .
lett . * 90 * , 56805 ( 2003 ) p. j. poole , j. mccaffrey , r.l .
williams , j. lefebvre , j. vac .
sci . & tech .
b * 19 * , 1467 ( 2001 ) y - h kim , i - h lee , s. nagaraja , j - p leburton , r. q. hood , r. m. martin , phys .
b * 61 * , 5202 ( 2000 ) f. stern , jpn . j. appl .
* 2 * , 323 ( 1974 ) j. f. dobson and j. wang , phys .
b * 69 * , 235104 ( 2004 ) j. p.
perdew and y. wang , phys .
rev b * 45 * , 13244 ( 1992 ) a. h. macdonald and g. c. aers , phys .
b * 29 * , 5976 ( 1984 ) f. c. zhang and s. das sarma , phys .
b * 33 * , 2903 ( 1986 ) e. tutuc , s. melinte , e. p.
de poortere , m. sheyegan , and r. winkler .
cond - mat/0301027 y .- w tan , j. zhu , h. l. stormer , l. n. pfeiffer , k. n. baldwin , and k. w. west , cond - matt/0412260 s. de palo , m. botti , s. moroni , and g. senatore , cond - mat/0410145 r. asgari , b. davoudi , m. polini , m. p. tosi , g. f. giuliani , and g. vignale , phys .
b * 71 * , 45323 ( 2005 ) ; cond - mat/0412665 y. zhang and s. das sarma , phys . rev .
* 71*,45322 ( 2005 ) ; cond - mat/0312565 m. w. c. dharma - wardana , cond - mat/503246 d. j. w. geldart and s. vosko , van . j. phys .
y. kwon , d. m. ceperley , and r. m. martin , phys .
b * 50 * , 1684 ( 1994 ) k. s. singwi , m. p. tosi , r. h. land , pnd a. sjolander , phys .
rev b * 176 * , 589 ( 1968 ) s. ichimaru , rev .
* 54 * , 1017 ( 1982 ) m. w. c. dharma - wardana and f. perrot , phys .
lett . * 84 * , 959 ( 2000 ) franois perrot and m. w. c. dharma - wardana , phys . rev .
b , * 62 * , 14766 ( 2000 ) franois perrot and m. w. c. dharma - wardana , phys
lett . * 87 * , 206404 ( 2001 ) m. w. c. dharma - wardana and f. perrot .
lett . * 90 * , 136601 ( 2003 ) m. w. c. dharma - wardana and f. perrot , phys .
b , * 66 * , 14110 ( 2002 ) m. w. c. dharma - wardana and f. perrot , phys . rev .
b * 70 * , 035308 ( 2004 ) m. w. c. dharma - wardana and f. perrot , europhys
. lett . * 63 * , 660 - 666 ( 2003 ) m. w. c. dharma - wardana , europhys .
lett . * 67 * , 552 ( 2004 ) f. lado , j. chem . phys . * 47 * , 5369 ( 1967 ) y. rosenfeld , phys.rev .
a bf42 , 5978 ( 1990 ) , m. baus , j .-
p hansen j. phys .
c : solid state phys . * 19 * , l463 ( 1986 ) ; our fortran codes are available from the author . c. bulutay and b. tanatar , phys .
b * 65 * , 195116 ( 2002 ) n. q. khanh and h. totsuji , solid state com . , * 129*,37 ( 2004 ) f. f. fang and w. e. howard , phys
* 16 * , 797 ( 1966 ) f. perrot and m. w. c. dharma - wardana , phys .
a * 30 * , 2619 ( 1984 ) we thank stephania de palo and gaetano senatore for informative and helpful correspondence , and for providing us with their correction energy @xmath177 and the full diffusion - qmc correction at @xmath47 .
d. m. ceperley and b. j. alder , phsy .
* 45 * , 566 ( 1980 ) p.
gori - giorgi and j. p. perdew , prb * 69 * , 041103 ( 2004 ) a. a. shashkin , m. rashmi , s. anissimova , and s. v. kravchenko , v. t. dolgopolov , t. m. klapwijk , phys .
* 91 * , 46403 ( 2003 ) k. vakili , y. p. shkolnikov , e. tutuc , e. p. de poortere , and m. shayegan , phys .
lett . * 92*,226401 ( 2004 ) m. w. c. dharma - wardana , solid state com .
* 127 * , 783 - 788 ( 2003 ) c. attaccalite et al .
lett . * 88 * , 256601 ( 2002 ) | the exchange and correlation @xmath0 of strongly correlated electrons in 2d layers of finite width are studied as a function of the density parameter @xmath1 , spin - polarization @xmath2 and the temperature @xmath3 .
we explicitly treat strong - correlation effects via pair - distribution functions , and introduce an equivalent constant - density approximation ( cda ) applicable to all the inhomogeneous densities encountered here .
the width @xmath4 defined via the cda provides a length scale defining the @xmath5-extension of the quasi-2d layer resident in the @xmath6-@xmath7 plane .
the correlation energy @xmath8 of the quasi-2d system is presented as an interpolation between a 1d gas of electron - rods ( for @xmath9 ) coupled via a log(r ) interaction , and a 3d coulomb fluid closely approximated from the known _ three - dimensional _ correlation energy when @xmath10 is small .
results for the @xmath11 , the transition to a spin - polarized phase , the effective mass @xmath12 , the land @xmath13-factor etc . , are reported here . |
this work was supported by the brazilian agencies cnpq and fapesp .
we would like to thank a. greco , p.g .
pagliuso and r. l. doretto for fruitful discussions .
v. barzykin and d. pines , phys .
rev * b 52 * , 13 585 ( 1995 ) ; w. anderson , science * 235 * , 1196 ( 1987 ) ; m. u. ubbens and p. a. lee , phys .
b 46 * , 8434 ( 1992 ) ; p. a. lee and x. g. wen , phys .
lett . * 78*. 4111 ( 1997 ) ; j. m. tranquada _
et al _ , phys .
lett . * 78 * , 338 ( 1997 ) ; c. castellani , c. di castro , and m. grilli , z. phys . * b 103 * , 137 ( 1997 ) . t. p. devereaux and d. einzel , phys .
b * 51 * , 16336 ( 1995 ) ; t. p. devereaux et al .
b * 54 * , 12523 ( 1996 ) ; t. strohm and m. cardona , solid state commun .
* 9 * , 233 ( 1997 ) ; m. krantz and m. cardona , j. low temp . phys . * 99 * , 205 ( 1995 ) .
j. geerk , x. x. xi and g. linker , z. phys .
* b 73 * , 329 ( 1988 ) ; m. gurvitch , j. m. valles , a. m. cucolo , r. c. dynes , j. p. garno , l , f. schneemeyer , and j. v. wasczak , phys .
lett . * 63 * , 1008 ( 1989 ) ; a. fournel , i. oujia , i. sorbier , europhys . lett . * 6 * , 653 ( 1988 ) . | the electronic raman scattering has been investigated in optimally oxygen doped yba@xmath0cu@xmath1o@xmath2 single crystals as well as in crystals with non - magnetic , zn , and magnetic , ni , impurities .
we found that the intensity of the a@xmath3 peak is impurity independent and their energy to @xmath4 ratio is almost constant ( @xmath5 ) .
moreover , the signal at the b@xmath3 channel is completely smeared out when non - magnetic zn impurities are present .
these results are qualitatively interpreted in terms of the zeyher and greco s theory that relates the electronic raman scattering in the a@xmath3 and b@xmath3 channels to _ d_-cdw and superconducting order parameters fluctuations , respectively .
far a long time , the phenomenon of high-@xmath4 superconductivity in the cuprates has been considered as related to an unconventional pairing state.@xcite it is widely accepted that the unconventional mechanism in these materials is closely related to their normal state properties , e.g. , the non - fermi - liquid behavior,@xcite in spite of the lack of consensus about the correct description of this state .
the presence of several kinds of fluctuations , such as spin , flux phases , stripes or charge density waves ( cdw ) is an additional complication@xcite and many models considering one or more of these fluctuations have been proposed .
however , there is a large amount of experimental data that these models can not explain ( see , e.g. , sections 5.4 and 6.4 of refs.@xcite and @xcite , respectively ) . among these data
we mention the absence of a convincent explanation for the electronic raman scattering ( ers ) of many cuprates in the superconducting state .
the redistribution of the ers in the superconducting state has been used to study the gap order parameter in many superconductors , both conventional or unconventional , such as nb@xmath1sn,@xcite cuprates,@xcite borocarbides@xcite and recently , mgb@xmath0.@xcite in general , the ers of the cuprates presents two characteristic peaks in the b@xmath3 and a@xmath6b@xmath7 channels in the superconducting phase that disappear above @xmath4 . in the yba@xmath0cu@xmath1o@xmath2 ( y123 )
system the ers of optimally doped crystals has shown the a@xmath3 peak stronger than the b@xmath3 counterpart .
also , the maximum intensity of the a@xmath3 peak lies at a lower energy than that of the b@xmath3 one.@xcite in the overdoped state , the b@xmath3 peak shifts down converging approximately to the same position of the a@xmath3 peak@xcite while in the strongly underdoped regime neither a@xmath3 nor b@xmath3 peaks are observed .
the appearance of these peaks only below @xmath4 , leads one to believe that they are related to the superconductivity , probably being pair - breaking peaks .
some publications@xcite have already shown that , for a gap with nodes and @xmath8 order parameter at low energies , the ers efficiency obeys the @xmath9 and @xmath10 power - laws in the b@xmath3 and a@xmath11b@xmath7 channels , respectively .
besides , these works have predicted that the a@xmath3 peak should be weaker than the b@xmath3 component with their maxima appearing at the same energy @xmath12 . in spite of the good agreement between the theory and the experimentally observed power laws , the relative position and intensity of the ers peaks are in disagreement with the theoretically expected results for the pair - breaking ers response .
hence , the origin of the a@xmath3 and b@xmath3 peaks remains unclear .
+ b@xmath7 channel showing the redistribution of the ers below @xmath4 for ( a ) y123 with t@xmath13 k , ( b ) y123:ni with t@xmath14 k , and ( c ) y123:zn with t@xmath15 k single crystals.,height=325 ] in this letter we make an attempt to clarify this question by means of ers measurements in y123 single crystals doped with a small amount , 5% , of either , magnetic ni@xmath16 ( y123:ni ) or non - magnetic zn@xmath16 ( y123:zn ) impurities in the copper planes .
it is known that substituting cu by zn or ni in y123 preserves the oxygen doping level and modify only slightly the crystal structure .
also , the non - magnetic zn@xmath16 impurities , contrary to the magnetic ni@xmath16 ones , restore significant spin fluctuations in the normal state.@xcite since the presence of impurities in the cuo@xmath0 planes are known to induce a pair - breaking effect,@xcite the investigation of the change in the electronic properties by introducing magnetic and non - magnetic impurities can give an important insight into the open questions related to the physics of the cuprates .
single crystals of y123 , y123:ni and y123:zn were prepared as previously described.@xcite after a thermal treatment , the transition temperatures were measured by _ dc_-magnetization and were found to be @xmath17 , @xmath18 and @xmath19 k for the y123 , y123:ni and y123:zn samples , respectively . in terms of the oxygen level ,
all samples are in the optimally - doped state .
the raman measurements were carried out using a triple spectrometer equipped with a ln@xmath0 ccd detector .
the @xmath20 nm line of an ar@xmath21 ion laser was used as an excitation source .
the laser power at the sample was kept below @xmath22 mw on a spot diameter of about @xmath23 @xmath24 m .
the samples were cooled in an exchange he gas variable temperature cryostat , and measured in a near - backscattering configuration on the @xmath25plane . for the tetragonal symmetry @xmath26 , the choice of the @xmath27 geometry probes a combination of the a@xmath3 and @xmath28 symmetries , while choosing the @xmath29 geometry couples to excitations of b@xmath3 symmetry .
@xmath30 ( @xmath31 ) denote axes rotated by @xmath32 from the crystallographic @xmath33 axes .
channel for ( a ) y123 , ( b ) y123:ni and ( c ) y123:zn.,height=325 ] in fig.1 we present the raman spectra in the a@xmath3+b@xmath7 channel at different temperatures for the y123 , y123:ni and y123:zn samples , corrected by the thermal bose - einstein factor . in fig .
1(a ) the spectrum for y123 at @xmath34 k displays a flat background and just below t@xmath35 k starts the rearrangement of the electronic background , resulting in a broad peak in the spectral range between @xmath36 and @xmath37 @xmath38 .
the same behavior is also found in fig.1 ( b ) for y123:ni .
however , in this case the rearrangement of the ers starts only below @xmath23 k , producing a broad peak located in the same spectral range as far the pure sample . for y123:zn , fig.1(c ) ,
the gain of spectral weight in the superconducting state is also present in the a@xmath39+b@xmath7 channel , although it starts to appears at @xmath40 k , below t@xmath41 k , but is located in the same spectral range as the other two samples . in the b@xmath3 channel , fig.2 , the rearrangement of the ers is also displayed for y123 and y123:ni . for y123 , fig.2(a )
, it starts below @xmath42 k and it appears in the @xmath43 to @xmath44 @xmath38 spectral range . for y123:ni , fig.2(b ) , the broad peak first appears below @xmath45 k and it is also located in the @xmath43 to @xmath44 @xmath38 spectral range .
surprisingly , for y123:zn , fig .
2(c ) , the rearrangement of the ers is absent below @xmath4 .
the only observed effect of lowering the temperature is the broadening of the b@xmath3 phonon at @xmath46 @xmath38 .
k ( black lines ) and @xmath34 k ( gray lines ) for ( a)y123 , ( b ) y123:ni , and ( c)y123:zn single crystals in a@xmath6b@xmath7 channel.,height=325 ] k ( black lines ) and @xmath34 k ( gray lines ) for ( a ) y123 , ( b ) y123:ni , and ( c ) y123:zn in the b@xmath3 channel.,height=325 ] in order to determine the energy of the broad peaks which appear in the superconducting phase in the a@xmath6b@xmath7 and b@xmath3 channels , the pure ers response function has been obtained by subtracting the contribution of the phonons fitted to lorentzian or fano - profiles to the bose - einstein corrected raw data . in figure 3 we present the pure ers response for the a@xmath6b@xmath7 channel at @xmath22 and @xmath34 k. at @xmath22 k we found the a@xmath3 response peaks around @xmath47 , @xmath48 and @xmath49 @xmath38 for y123 ( 3a ) , y123:ni ( 3b ) and y123:zn ( 3c ) , respectively .
we notice that the a@xmath3 energy to @xmath4 ratio @xmath50 is @xmath51 for y123 , @xmath52 for y123:ni , and @xmath53 for y123:zn .
as commented above , at @xmath34 k , the superconducting rearrangement of the ers is absent and the raman spectra present no pronounced peak .
figure 4 shows the pure ers for the b@xmath3 channel at @xmath22 and @xmath34 k. the dramatic difference between the response of y123:zn compared with that of the other crystals is evident .
the peak is absent in y123:zn but it appears centered around @xmath54 @xmath38 in y123 and y123:ni .
the ratio @xmath55 is @xmath56 for y123 and @xmath57 for y123:ni .
moreover , the ers spectrum in y123:zn is almost the same at @xmath22 and @xmath34 k being also very similar to those in y123 and y123:ni at @xmath34 k , except for a small difference in the absolute value of the intensities .
observation of fig .
3 and 4 indicates that the zn substitution affects the a@xmath6b@xmath7 and b@xmath3 channels in a quite different way . while in the a@xmath6b@xmath7 channel the intensities of the peaks are almost unaffected by the impurities , in the b@xmath3 channel the peak is smeared out in the zn - doped sample .
the comparison between the b@xmath3 channel signals at @xmath22 and @xmath34 k in y123:zn indicates that the ers in the superconducting and normal states are nearly the same .
this unusual effect of zn substitution on the b@xmath3 raman response is surprising and has not been predicted by any theoretical model nor been observed by other systematic experimental work .
another relevant result is the almost constant energy to @xmath4 ratio @xmath58 for the a@xmath3 peak for all samples .
this value is in agreement with previous values obtained for the a@xmath3 peak@xcite and for the superconducting gap measured by electron tunneling spectroscopy in y123.@xcite recently , venturini et al@xcite and zeyher and greco @xcite have discussed theoretically the ers in the superconducting phase of high-@xmath4 cuprates .
the work of venturini et al.@xcite presents a model suggesting that the observed peak in the a@xmath3 + b@xmath7 channel is produced by collective spin fluctuations , being a two - magnon raman peak . in the same model , the b@xmath3 peak
is related to pair breaking .
they were able to fit the experimental raman spectrum of bi2212 to their model and obtained the correct relative position of the a@xmath3 and b@xmath3 peaks.@xcite moreover , gallais et al.@xcite have shown that the a@xmath3 peak tracks the magnetic resonance peak observed by inelastic neutron scattering@xcite at 40 @xmath59 in ni - substituted yba@xmath0cu@xmath1o@xmath60 .
these authors have interpreted this fact as an evidence for the magnetic origin of the a@xmath3 peak .
however , our results can not be interpreted in these terms .
the main reason is the strong zn - impurity dependency observed in the b@xmath3 spectra . in the venturini s framework
the b@xmath3 peak is related to pair - breaking process with their maximum at @xmath12 energy .
thus , the complete smearing out of the b@xmath3 peak observed in our experiments ( see fig
. 4 c ) would imply that @xmath61 for the zn - substituted crystal .
however , it is known that @xmath4 does not go to zero in this case and , discharging any anomalous behavior of the @xmath62 ratio , @xmath63 .
futhermore , the @xmath64 ratio for the a@xmath3 peak presents better agreement to the gap energy measured by others groups ( see , e.g. , ref .
@xcite ) than the b@xmath3 one , indicating that the a@xmath3 ers peak is related to pair - breaking .
another possible theoretical comparison could be made with the zeyher and grecos@xcite theory .
these authors used the superconducting model by cappelluti and zeyher@xcite in order to understand the ers of cuprates .
the cappelluti and zeyher model proposed that the superconductivity in cuprates is originated by the competition between the superconducting and the _ d_-cdw ( also called orbital antiferromagnetism ) order parameters .
zeyher and greco@xcite suggested that the a@xmath3 and b@xmath3 peaks are caused by amplitude fluctuations of the superconducting and _ d_-cdw order parameters , respectively .
the _ d_-cdw phase corresponds to a flux phase where current flows around each cuo@xmath0 square alternatively clockwise and counterclockwise@xcite giving rise to orbital antiferromagnetism where the only interaction present in the order parameter is the heisenberg exchange coupling between the cu@xmath16 ions.@xcite as mentioned above , our experimental data indicates that the a@xmath3 peak is related to pair - breaking . in this sense ,
our experimental ers results give support to the zeyher and greco s theory .
nevertheless , the impurity effect on the _ d_-cdw b@xmath3 peak is not well theoretically understood at the moment .
cappelluti and zeyher@xcite have shown that the zn impurities substitution does not have appreciable influence on the oxygen doping phase diagram of the cuprates . however , the effect on the b@xmath3 ers signal could be more subtle and a more detailed theoretical calculation could be used .
nonetheless , qualitatively we can elaborate about the impurity substitution effect on b@xmath3 signal as follows .
it is known that the main interaction originating the _ d_-cdw is the heisenberg coupling between the cu@xmath16 spins .
thus , it is expected that the cu@xmath16 ( @xmath65 ) substitution by zn@xmath16 ( @xmath66 ) impurity , would break down the long range coherence of the orbital antiferromagnetism , explaining the disappearence of the _ d_-cdw excitation in the b@xmath3 channel .
in fact , this is consistent with the larger depletion of @xmath4 just produced by the presence of zn@xmath16 impurities.@xcite besides , the ni@xmath16 ( @xmath67 ) impurities could be less effective in breaking down the orbital antiferromagnetism notwithstanding having also effect in decreasing @xmath4 .
moreover , as shown by gupta and gupta@xcite the charge density redistribution due to the ni@xmath16 ions is localized while the zn@xmath16 perform an extended perturbation . on this basis
, one would expect that the zn@xmath16 ions would be more effective on reducing the _ d_-cdw excitations . in conclusion ,
our results show that the a@xmath3 ers peak present a constant gap to @xmath4 ratio @xmath58 independent of the presence of magnetic or non - magnetic impurities .
also , the b@xmath3 ers peak is smeared out in the y123 when the cu@xmath16 is substituted by small amount of non - magnetic zn impurities whereas the a@xmath6b@xmath7 spectra remain insensitive to the kind of impurity .
these results could be qualitatively interpreted in terms of the of zeyher and grecos@xcite theory that relates the ers in the a@xmath3 and b@xmath3 channels to _ d_-cdw and superconducting order parameters fluctuations , respectively . |
the main result of the present paper is a `` relative version '' of the _ another presentation _ for symplectic steinberg groups obtained in @xcite .
first , we give a definition of a relative symplectic group @xmath0 for any form ideal @xmath1 in @xmath2 . for maximal form parameter @xmath3
it has already appeared in @xcite an is shown to have good homological properties .
we also consider case of a maximal form parameter and establish another `` coordinate - free '' presentation for @xmath4 . the last result is a relativisation of the @xcite , but it is not a generalisation , since we need more relations for the presentation .
moreover , in our proofs we use the results of @xcite .
namely , the main theorem of @xcite states the following .
let @xmath5 , then the symplectic steinberg group @xmath6 can be defined by the set of generators @xmath7 and relations @xmath8 where @xmath9 is an esd - transformation @xmath10 for usual generators of the symplectic steinberg group the following identities hold @xmath11 we prove the following theorem in the case of maximal form parameter .
let @xmath5 , then the relative symplectic steinberg group @xmath12 can be defined by the set of generators @xmath13 and relations @xmath14 and for usual relative generators the following identities hold @xmath15 the author plans to generalise the above result for the case of arbitrary form parameter .
in the sequel @xmath2 denotes an arbitrary associative commutative unital ring , @xmath16 denotes a free right @xmath2-module with basis numbered @xmath17 , @xmath18 , @xmath19 , @xmath20 , @xmath18 , @xmath21 , @xmath5 . for the vector
@xmath22 its @xmath23-th coordinate will be denoted by @xmath24 , i.e. @xmath25 . by @xmath26 we denote the standard symplectic form on @xmath27 , i.e @xmath28 .
we will usually write @xmath29 instead of @xmath30 .
observe that @xmath31 for any @xmath32 .
define the _ symplectic group _
@xmath33 as the group of automorphisms of @xmath27 preserving the symplectic form @xmath34 , @xmath35 for @xmath36 and @xmath37 , @xmath22 , @xmath38 , denote by @xmath9 the automorphism of @xmath27 s.t . for @xmath39 one
has @xmath40 we refer to the elements @xmath9 as the ( _ symplectic _ ) _ esd - transformations_. [ esd - properties ] let @xmath37 , @xmath41 , @xmath39 be three vectors such that @xmath38 , @xmath42 , and let @xmath43 , @xmath44
. then 1 .
@xmath45 , 2 .
@xmath46 , 3 .
@xmath47 , 4 . @xmath48 .
observe that @xmath49 and @xmath50 . in the present paper all commutators
are left - normed , @xmath51=xyx{^{-1}}y{^{-1}}$ ] , we denote @xmath52 by @xmath53 .
[ z - decomposition ] for @xmath37 , @xmath22 such that @xmath54 , @xmath38 , and @xmath36 one has @xmath55=t(u,\,v{\varepsilon_{i}},\,a)t(e_{-i},\,-ua{\varepsilon_{-i}},\,0).\ ] ] _ a form ideal _
@xmath1 is a pair of an ideal @xmath56 and a _ relative form parameter _
@xmath57 _ of level _
@xmath58 , i.e. , an additive subgroup of @xmath58 such that 1 .
@xmath59 holds @xmath60 , 2 .
@xmath61 holds @xmath62 , 3 .
@xmath63 holds @xmath64 . if @xmath65 the only possible choice for a relative form parameter is @xmath3 .
define @xmath66 and @xmath67 , where @xmath36 , @xmath23 , @xmath68 @xmath69 @xmath70 @xmath71 @xmath69 @xmath72 , @xmath73 .
we refer to these elements as the _ elementary symplectic transvections_. a normal subgroup of @xmath74 @xmath75 is called _ the relative elementary symplectic group _ corresponding to the form ideal @xmath1 . the _ symplectic steinberg group _ @xmath76 is the group generated by the formal symbols @xmath77 , @xmath78 , @xmath79 subject to the steinberg relations @xmath80=1,\text { for } h\not\in\{j ,-
i\},\ k\not\in\{i ,- j\},\\ & [ x_{ij}(r),\,x_{jk}(s)]=x_{ik}(rs),\text { for } i\not\in\{-j ,- k\},\ j\neq - k,\\ & [ x_{i ,- i}(r),\,x_{-i , j}(s)]=x_{ij}(rs{\varepsilon_{i}})x_{-j , j}(-rs^2),\\ & [ x_{ij}(r),\,x_{j ,- i}(s)]=x_{i ,- i}(2rs{\varepsilon_{i}}).\end{aligned}\ ] ] the next lemma is a straightforward consequence of lemmas [ esd - properties ] and [ z - decomposition ] .
there is a natural epimorphism @xmath81 sending the generators @xmath82 to the corresponding elementary transvections @xmath83 . in other words ,
the steinberg relations hold for the elementary transvections .
if group @xmath84 acts on group @xmath85 from the left , we will denote the image of @xmath86 under the homomorphism corresponding to the element @xmath87 by @xmath88 , the element @xmath89 by @xmath90 $ ] and the element @xmath91 by @xmath92 .
define the _ relative symplectic steinberg group _ @xmath0 corresponding to the form ideal @xmath1 as a formal group with the action of the absolute steinberg group @xmath76 defined by the set of ( relative ) generators @xmath93 subject to the following relations @xmath94=1,\text { for } h\not\in\{j ,-
i\},\ k\not\in\{i ,- j\},\\ & \llbracket x_{ij}(r),\,y_{jk}(a)]=y_{ik}(ra),\text { for } i\not\in\{-j ,-
k\},\ j\neq - k,\\ & \llbracket x_{i ,- i}(r),\,y_{-i , j}(a)]=y_{ij}(ra{\varepsilon_{i}})y_{-j , j}(-ra^2),\\ & [ y_{i ,- i}(\alpha),\,x_{-i , j}(r)\rrbracket = y_{ij}(\alpha r{\varepsilon_{i}})y_{-j , j}(-\alpha r^2),\\ & \llbracket x_{ij}(r),\,y_{j ,- i}(a)]=x_{i ,- i}(2ra{\varepsilon_{i}}),\\ & \!\,^{x_{ij}(a)}y_{hk}(b)=\!\,^{y_{ij}(a)}y_{hk}(b).\end{aligned}\ ] ] in other words we consider a free group generated by symbols @xmath95 where @xmath96 is from the absolute steinberg group and @xmath97 is from the set of relative generators , @xmath76 naturally acts on this free group via @xmath98 and then we define a relative symplectic steinberg group as a factor of the described free group modulo normal subgroup generated by kl0kl7 .
obviously , there is a natural map @xmath99 then its kernel is denoted by @xmath100 .
define the _ relative steinberg unipotent radical _ @xmath101 and the ( _ absolute _ ) _ steinberg parabolic subgroup _
@xmath102 for @xmath103 ,
@xmath104 one has @xmath105 [ heis ] one has @xmath106\leq{\langle}y_{i ,- i}(\alpha){\rangle},\qquad [ { \!\,^{(i)}u_1},\,{\langle}y_{i ,- i}(\alpha){\rangle}]=1.\ ] ] every element of @xmath107 can be expressed in the form @xmath108 [ unirad ] the restriction of the natural projection @xmath109 to @xmath107 is injective @xmath110 take an element @xmath111 . using the above corollary ,
decompose @xmath97 as @xmath112 then @xmath113 implies that @xmath114 for all @xmath23 .
[ unipotent - decomposition ] take @xmath115 such that @xmath116 and @xmath117 . denote @xmath118 then @xmath119 assume that @xmath120 , for @xmath121 the proof looks exactly the same .
since @xmath122 for @xmath123 and @xmath124 one has @xmath125 and @xmath126 so that the right hand side of the desired equality is in fact equal to @xmath127 for @xmath115 with @xmath116 and @xmath36 such that @xmath128 , define @xmath129 by lemma [ unipotent - decomposition ] , @xmath130 indeed lies in @xmath131 . moreover , the same lemma provides the following decomposition .
[ y - decomposition ] for @xmath115 such that @xmath116 , @xmath36 such that @xmath128 one has @xmath132 in particular , @xmath133 for @xmath73 and @xmath134 .
[ y - add ] for @xmath41 , @xmath135 such that @xmath136 and @xmath43 , @xmath44 such that @xmath137 , @xmath138 , one has @xmath139 obviously , @xmath140 .
moreover , @xmath141 so that the right hand side of this equality is well - defined .
now , it remains to observe that the images of the elements on both sides under @xmath142 coincide .
one has @xmath143 and @xmath144 .
[ wow ] for @xmath145 , @xmath115 one has @xmath146 we may assume that @xmath147 is an elementary transvection . for short root transvection one
has @xmath148 for long root transvection one has @xmath149 [ y - conjugated - by - ps ] for @xmath103 , @xmath115 such that @xmath116 , @xmath36 such that @xmath128 , one has @xmath150
first , observe that since @xmath151 for @xmath152 one has @xmath153 .
thus , @xmath154 i.e. , @xmath155 . by lemma
[ wow ] , @xmath156 , so the right hand side of the desired equation is well - defined .
finally , observe that the images of both sides under @xmath142 coincide .
[ switch ] for @xmath123 , @xmath157 , one has @xmath158 . for @xmath159
the claim is obvious .
let @xmath78 , then @xmath160 in the absolute situation @xmath161 , we will write @xmath162 instead of @xmath163 .
these are exactly the elements , which appear in the _ another presentation_. [ z - decomposition - for - y ] for @xmath22 such that @xmath164 , @xmath165 , @xmath36 , @xmath166 one has @xmath167 the proof of lemma 12 from the another presentation paper can be repeated verbatim . for @xmath22 such that @xmath164 , @xmath165 , @xmath36 , @xmath166 one has the following decomposition @xmath168 [ ppc ] for @xmath169 , @xmath115 such that @xmath164 , @xmath79 , @xmath43 such that @xmath128 , one has @xmath170=y(e_j,\,vr{\varepsilon_{k}},\,ar^2)y(e_{-k},\,e_jar{\varepsilon_{k}},\,0).\ ] ] denote @xmath171 , @xmath172 , @xmath173 , @xmath174
. then @xmath170=\llbracket x,\,yzw]=\llbracket x,\,y]\cdot\!\,^y\llbracket x,\,z]\cdot\!\,^{yz}\llbracket x,\,w].\ ] ] assume that @xmath175 , for @xmath176 the proof looks exactly the same . for @xmath177 one
has @xmath178=\llbracket x_{j ,- k}(r{\varepsilon_{k}}),\,y_{-k ,-
h}(v_h{\varepsilon_{h}})]=\\ = y_{j ,- h}(rv_h{\varepsilon_{k}}{\varepsilon_{h}})=y_{h ,- j}(rv_h{\varepsilon_{k}}{\varepsilon_{j}}).\end{gathered}\ ] ] for @xmath179 one has @xmath180=\llbracket x_{j ,- k}(r{\varepsilon_{k}}),\,y_{-k ,-
j}(v_j{\varepsilon_{j}})]=y_{j ,- j}(2rv_j{\varepsilon_{k}}).\ ] ] since any @xmath181 commutes with any @xmath182 for @xmath183 , @xmath184 , @xmath183 , @xmath185 , we have @xmath186=\prod_{h<0}\llbracket x_{j ,- k}(r{\varepsilon_{k}}),\,y_{h , k}(v_h{\varepsilon_{-k}})]=\\=y_{j ,- j}(2rv_j{\varepsilon_{k}})\prod_{\substack{h\neq j\\h<0}}y_{h ,- j}(rv_h{\varepsilon_{k}}{\varepsilon_{j}})=y(e_j,\,v_-r{\varepsilon_{k}},\,0).\end{gathered}\ ] ] similarly , @xmath187=\prod_{\substack{h>0}}y_{h ,- j}(rv_h{\varepsilon_{k}}{\varepsilon_{j}})=y(e_j,\,v_+r{\varepsilon_{k}},\,0).\ ] ] then , @xmath188=\!\,^{y(e_{-k},\,v_-,\,0)}y(e_j,\,v_+r{\varepsilon_{k}},\,0)=\\ = y(e_j,\,v_+r{\varepsilon_{k}},\,0)[y(e_j,\,-v_+r{\varepsilon_{k}},\,0),\,y(e_{-k},\,v_-,\,0)]=\\ = y(e_j,\,v_+r{\varepsilon_{k}},\,0)\llbracket x(e_j,\,-v_+r{\varepsilon_{k}},\,0),\,y(e_{-k},\,v_-,\,0)]=\\ = y(e_j,\,v_+r{\varepsilon_{k}},\,0)y(e_{-k},\,e_j{\langle}v_-,\,v_+{\rangle}r{\varepsilon_{k}},\,0).\end{gathered}\ ] ] next , @xmath189=\llbracket x_{j ,- k}(r{\varepsilon_{k}}),\,y_{-k , k}(a-{\langle}v_-,\,v_+{\rangle})]=\\ = \big([y_{-k , k}(a-{\langle}v_-,\,v_+{\rangle}),\,x_{k ,-
j}(r{\varepsilon_{j}})\rrbracket\big){^{-1}}=\\ = \big(y_{-k ,- j}\big((a-{\langle}v_-,\,v_+{\rangle})r{\varepsilon_{j}}{\varepsilon_{-k}}\big)y_{j ,- j}\big(-(a-{\langle}v_-,\,v_+{\rangle})r^2\big)\big){^{-1}}=\\ = y_{j ,- j}(ar^2-{\langle}v_-r{\varepsilon_{k}},\,v_+r{\varepsilon_{k}}{\rangle})y(e_j,\,e_{-k}(a-{\langle}v_-,\,v_+{\rangle})r{\varepsilon_{k}},\,0).\end{gathered}\ ] ] obviously , @xmath190 $ ] commutes with @xmath191 and @xmath192 , thus , finally , @xmath193=y_{j ,- j}(ar^2-{\langle}v_-r{\varepsilon_{k}},\,v_+r{\varepsilon_{k}}{\rangle})y_{j ,-
j}(2rv_j{\varepsilon_{k}})\cdot\\ \cdot\prod_{\substack{h\neq j\\h<0}}y_{h ,- j}(rv_h{\varepsilon_{k}}{\varepsilon_{j}})\prod_{\substack{h>0}}y_{h ,- j}(rv_h{\varepsilon_{k}}{\varepsilon_{j}})y(e_{-k},\,e_j{\langle}v_-,\,v_+{\rangle}r{\varepsilon_{k}},\,0)\cdot\\ \cdot y(e_{-k},\,e_{j}(a-{\langle}v_-,\,v_+{\rangle})r{\varepsilon_{k}},\,0)=y(e_j,\,vr{\varepsilon_{k}},\,ar^2)y(e_{-k},\,e_jar{\varepsilon_{k}},\,0).\end{gathered}\ ] ] for @xmath169 , @xmath115 such that @xmath164 , @xmath79 , @xmath43 such that @xmath128 , one has @xmath194y(e_{-k},\,e_jar{\varepsilon_{-k}},\,0).\ ] ] for @xmath32 , @xmath115 such that @xmath38 , @xmath54 , @xmath36 such that @xmath128 denote @xmath195y(e_{-i},\,ua{\varepsilon_{-i}},\,0).\ ] ] due to lemma [ z - decomposition ] one has @xmath196 . for @xmath115
such that @xmath197 , @xmath36 such that @xmath128 , one has @xmath198 by lemma [ ppc ] .
one can also obtain the following result . for @xmath22 with @xmath197
, @xmath166 one has @xmath199 the proof of lemma 15 from the another presentation paper can be repeated verbatim .
[ z - correctness ] consider @xmath32 , @xmath115 such that @xmath38 , @xmath200 , @xmath201 , @xmath79 and @xmath36 such that @xmath128 .
then one has @xmath202 denote @xmath203 , @xmath204 , @xmath205 , then by the very definition @xmath206c.\ ] ] denote @xmath207 , @xmath208 , @xmath209 , then by lemma [ ppc ] @xmath210w\ ] ] and @xmath211=\llbracket x,\,\llbracket y{^{-1}},\,z]w]=\llbracket x,\,\llbracket y{^{-1}},\,z]]\cdot\!\,^{\llbracket y{^{-1}},\,z]}\llbracket x,\,w].\ ] ] using hall
witt identity , @xmath212=\!\,^{y{^{-1}}x}\llbracket[x{^{-1}},\,y],\,z]\cdot\!\,^{y{^{-1}}z}[[z{^{-1}},\,x\rrbracket,\,y\rrbracket.\ ] ] one can check using lemma [ y - conjugated - by - ps ] that @xmath97 acts trivially on @xmath192 , i.e. , @xmath213 , thus , @xmath212=\!\,^{y{^{-1}}x}\llbracket[x{^{-1}},\,y],\,z].\ ] ] next , @xmath214=[x(e_i,\,-u,\,0),\,x(e_j,\,-e_{-i}r{\varepsilon_{i}},\,0)]=x(-u,\,-e_jr,\,0)=x(e_j,\,ur,\,0).\ ] ] denote @xmath215=x(e_j,\,ur,\,0)$ ] , one can show that @xmath216=1 $ ] , thus , @xmath212=\!\,^{y{^{-1}}}\llbracket d,\,z].\ ] ] observe that @xmath191 acts trivially on @xmath217 , thus , @xmath218=\llbracket yx,\,q]$ ] .
since @xmath219 $ ] and @xmath220=[x(e_{-i},\,e_jr{\varepsilon_{i}},\,0),\,x(e_i,\,-u,\,0)]=x(e_j,\,ur,\,0)\ ] ] also acts trivially on @xmath217 , one has @xmath218=\llbracket x,\,q]$ ] , thus , @xmath211=\!\,^y\llbracket x,\,q]=\!\,^y\llbracket x,\,\llbracket y{^{-1}},\,z]]\cdot\!\,^{y\llbracket y{^{-1}},\,z]}\llbracket x,\,w]=\llbracket d,\,z]\cdot\!\,^{y\llbracket y{^{-1}},\,z]}\llbracket x,\,w].\ ] ] denote @xmath221=\llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,e_{-j}ar{\varepsilon_{i}}{\varepsilon_{-j}},\,0)]=y(e_{-j},\,uar{\varepsilon_{-j}},\,0),\ ] ] then @xmath222=1 $ ] , thus , @xmath223}h=\!\,^{\llbracket y{^{-1}},\,z]w}h=\!\,^{q}h = h$ ] , so , @xmath211=\llbracket d,\,z]\cdot\!\,^{y}h=\llbracket d,\,z]\cdot h\cdot[h{^{-1}},\,y\rrbracket.\ ] ] since @xmath224=\llbracket x(e_j,\,-e_{-i}r{\varepsilon_{i}},\,0),\,y(e_{-j},\,uar{\varepsilon_{j}},\,0)]=y(e_{-i},\,uar^2{\varepsilon_{-i}},\,0)=c,\ ] ] one has @xmath211c=\llbracket d,\,z]h\ ] ] or , @xmath202 define the ( _ absolute _ ) _ levi subgroup _ @xmath225 .
observe that for @xmath226 one has @xmath153 and @xmath227 .
indeed , the first equality holds for @xmath228 with @xmath229 and the second one for @xmath228 with @xmath230 . for @xmath32 , @xmath115 such that @xmath38 , @xmath54 , @xmath36 such that @xmath128 , @xmath226
, one has @xmath231 since @xmath232 , one can see that @xmath233 and similarly @xmath234 .
using lemma [ wow ] one obtains , that @xmath156 .
thus , @xmath235 is well - defined .
the proof of lemma 17 from the another presentation paper works for this situation as well .
lemma [ unipotent - decomposition ] implies that for @xmath41 such that @xmath236 , @xmath237 one has @xmath238 . for @xmath239 orthogonal to both @xmath37 and @xmath41 , @xmath240 ,
one has @xmath241 .
below , in the computations this fact is frequently used without any special reference .
[ z - additivity ] for @xmath237 and @xmath32 such that @xmath242 , @xmath115 such that @xmath243 and @xmath38 , and for @xmath43 , @xmath166 such that @xmath128 , one has @xmath244 start with the right - hand side ( one can check that it is well - defined ) @xmath245y(e_{-i},\,u(a+v_{-j}b{\varepsilon_{-j}}){\varepsilon_{-i}},\,0).\end{gathered}\ ] ] decompose @xmath246 inside the commutator and use the familiar identity @xmath247=\llbracket a,\,b]\cdot\,^b\llbracket a,\,c]$ ] to obtain @xmath248=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,a)y(e_{-i},\,e_jb{\varepsilon_{i}},\,0)]=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,a)]\cdot\,^{y(e_{-i},\,v{\varepsilon_{i}},\,a)}\llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,e_jb{\varepsilon_{i}},\,0)]=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,a)]\cdot\,^{y(e_{-i},\,v{\varepsilon_{i}},\,a)}y_{(i)}(u,\,e_jb,\,0)=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,a)]\cdot\,^{y(e_{-i},\,v{\varepsilon_{i}},\,a)}y(e_j,\,ub,\,0).\end{gathered}\ ] ] observe that in general @xmath249 does not lie in @xmath250 , but @xmath251 always lies in @xmath252 .
so that we can compute the conjugate as follows @xmath253=\\ = y(e_j,\,ub,\,0)y(e_{-i},\,t(e_j,\,-ub,\,0)v{\varepsilon_{i}},\,a)\cdot y(e_{-i},\,-v{\varepsilon_{i}},\,-a)=\\ = y(e_j,\,ub,\,0)y(e_{-i},\,v{\varepsilon_{i}}-ub{\langle}e_j,\,v{\varepsilon_{i}}{\rangle},\,a)y(e_{-i},\,-v{\varepsilon_{i}},\,-a)=\\ = y(e_j,\,ub,\,0)y(e_{-i},\,-ubv_{-j}{\varepsilon_{i}}{\varepsilon_{j}},\,0).\end{gathered}\ ] ] thus , @xmath254y(e_j,\,ub,\,0)\cdot\\ \cdot y(e_{-i},\,-ubv_{-j}{\varepsilon_{i}}{\varepsilon_{j}},\,0)y(e_{-i},\,u(a+v_{-j}b{\varepsilon_{-j}}){\varepsilon_{-i}},\,0)=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,a)]y(e_j,\,ub,\,0)y(e_{-i},\,ua{\varepsilon_{-i}},\,0).\end{gathered}\ ] ] finally , it remains to observe that @xmath255 and thus one has @xmath256=1 $ ] and @xmath254y(e_{-i},\,ua{\varepsilon_{-i}},\,0)\cdot\\ \cdot y(e_j,\,ub,\,0)=y_{(i)}(u,\,v,\,a)y_{(i)}(u,\,e_jb,\,0).\end{gathered}\ ] ] for @xmath32 such that @xmath257 , @xmath115 such that @xmath38 , and @xmath36 such that @xmath128 define @xmath258 observe that the above definition coincides with the old one for @xmath41 with @xmath243 , so that we can use the same notation for the generator .
observe also that the right - hand side is well - defined .
[ w - conjugation ] for @xmath226 , @xmath32 such that @xmath257 , @xmath115 such that @xmath38 , and @xmath36 such that @xmath128 one has @xmath231
since @xmath226 one gets @xmath259 and similarly @xmath260 .
then @xmath261 obviously , in the absolute case @xmath161 for @xmath37 , @xmath22 such that @xmath242 and @xmath262 , @xmath237 , @xmath36 , one has that such an element @xmath263 lies in @xmath264 .
indeed , it is a product of elements from @xmath264 by definition .
[ w - correctness ] for @xmath237 , @xmath32 such that @xmath242 , @xmath115 such that @xmath38 , and @xmath36 such that @xmath128 one has @xmath265 denote @xmath266 , @xmath267
. then by lemma [ z - additivity ] one gets @xmath268 further , denote @xmath269 and @xmath270 .
then one has @xmath271 changing roles of @xmath23 and @xmath272 one gets @xmath273 but @xmath274 by lemma [ z - correctness ] . finally , it remains to observe that @xmath275 and @xmath276 commute with both @xmath277 and @xmath278 .
this is obvious from the fact that @xmath279 and @xmath280 lie in @xmath264 .
for @xmath37 equal to the base vector @xmath281 using lemma [ z - decomposition - for - y ] one gets @xmath282 for @xmath37 having at least two pairs of zeros the element @xmath283 does not depend on the choice of @xmath23 by lemma [ w - correctness ] .
in this situation we will often omit the index in the notation , @xmath284 for @xmath32 , @xmath285 , such that @xmath42 , @xmath286 and @xmath287 define @xmath288 our next objective is to show that @xmath289 does not depend on the choice of @xmath23 .
[ long - add ] for @xmath32 such that @xmath242 , @xmath115 such that @xmath290 , @xmath38 , @xmath36 such that @xmath128 , @xmath291 one has @xmath292 decompose @xmath293 as follows @xmath294y(e_{-i},\,u(a+b){\varepsilon_{-i}},\,0).\end{gathered}\ ] ] using @xmath247=\llbracket a,\,b]\cdot\,^b\llbracket a,\,c]$ ] we obtain @xmath295=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,a)y(e_{-i},\,0,\,b)]=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,a)]\cdot\,^{y(e_{-i},\,v{\varepsilon_{i}},\,a)}\llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,0,\,b)].\end{gathered}\ ] ] since @xmath38 one has @xmath296 and thus @xmath297y(e_{-i},\,ub{\varepsilon_{-i}},\,0)y(e_{-i},\,ua{\varepsilon_{-i}},\,0)=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,a)]\cdot\,^{y(e_{-i},\,v{\varepsilon_{i}},\,a)}\llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,0,\,b)]\cdot\\ \cdot\,^{x(e_{-i},\,v{\varepsilon_{i}},\,a)}y(e_{-i},\,ub{\varepsilon_{-i}},\,0)\cdot y(e_{-i},\,ua{\varepsilon_{-i}},\,0)=\\ = \llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,v{\varepsilon_{i}},\,0)]\cdot\,^{y(e_{-i},\,v{\varepsilon_{i}},\,a)}y_{(i)}(u,\,0,\,b)\cdot y(e_{-i},\,ua{\varepsilon_{-i}},\,0).\end{gathered}\ ] ] recall that @xmath298 acts trivially on both @xmath299 and @xmath300 , so that @xmath301y(e_{-i},\,ua{\varepsilon_{-i}},\,0)y(u,\,0,\,b)=\\ = y_{(i)}(u,\,v,\,a)y(u,\,0,\,b).\end{gathered}\ ] ] for @xmath32 having at least two pairs of zeros one has @xmath302 and @xmath303 .
[ w = zz ] for @xmath32 such that @xmath242 , @xmath115 such that @xmath38 and @xmath304 , @xmath36 such that @xmath128 one has @xmath305 by definition @xmath306 denote @xmath266 , then by the previous lemma @xmath307 and thus @xmath308 consider @xmath32 , @xmath285 , such that @xmath42 , @xmath286 , @xmath287 , @xmath237 and denote @xmath309 , @xmath310 , @xmath311 , @xmath312 . then one has @xmath313 since @xmath5 the vector @xmath41 has at least two pairs of zeros , so that one can apply the previous lemma .
[ short - is - three - long ] for @xmath272 , @xmath314 , @xmath37 , @xmath22 such that @xmath242 and @xmath315 , @xmath38 , @xmath135 such that @xmath286 , @xmath240 and @xmath287 holds @xmath316 decomposing @xmath317 inside the commutator and using @xmath318=\,^a\llbracket b,\,c]\cdot\llbracket a,\,c]$ ] we get @xmath319y(e_{-i},\,(u+v)a{\varepsilon_{-i}},\,0)=\\ = \llbracket x(e_i,\,v,\,0)x(e_i,\,u,\,0),\,y(e_{-i},\,w,\,a)]y(e_{-i},\,va{\varepsilon_{-i}},\,0)y(e_{-i},\,ua{\varepsilon_{-i}},\,0)=\\ = \!\,^{x(e_i,\,v,\,0)}\llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,w,\,a)]\cdot\llbracket x(e_i,\,v,\,0),\,y(e_{-i},\,w,\,a)]\cdot\\ \cdot y(e_{-i},\,va{\varepsilon_{-i}},\,0)y(e_{-i},\,ua{\varepsilon_{-i}},\,0)=\!\,^{x(e_i,\,v,\,0)}\llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,w,\,a)]\cdot\\ \cdot y_{(i)}(v,\,w,\,a)y(e_{-i},\,ua{\varepsilon_{-i}},\,0).\end{gathered}\ ] ] observe that @xmath320 commutes with @xmath321 , moreover , @xmath322 and acts trivially on @xmath323 , thus @xmath323 commutes with @xmath324 .
thus , we get @xmath325\cdot y(v,\,w,\,a)\cdot y(e_{-i},\,ua{\varepsilon_{-i}},\,0)=\\ = \!\,^{x(e_i,\,v,\,0)}\llbracket x(e_i,\,u,\,0),\,y(e_{-i},\,w,\,a)]\cdot\,^{x(e_i,\,v,\,0)}y(e_{-i},\,ua{\varepsilon_{-i}},\,0)\cdot\\ \cdot\,^{x(e_i,\,v,\,0)}y(e_{-i},\,-ua{\varepsilon_{-i}},\,0)\cdot y(v,\,w,\,a)\cdot y(e_{-i},\,ua{\varepsilon_{-i}},\,0)=\\ = \!\,^{x(e_i,\,v,\,0)}y_{(i)}(u,\,w,\,a)\cdot y(v,\,w,\,a)\cdot\llbracket x(e_i,\,v,\,0),\,y(e_{-i},\,-ua{\varepsilon_{-i}},\,0)]=\\ = \!\,^{x(e_i,\,v,\,0)}y(u,\,w,\,a)\cdot y(v,\,w,\,a)\cdot y(v,\,ua,\,0).\end{gathered}\ ] ] it only remains to show that @xmath326 decompose @xmath327y(e_i,\,ua{\varepsilon_{i}},\,0),\ ] ] then , @xmath328y(e_i,\,ua{\varepsilon_{i}},\,0)=\\ = \llbracket x(e_{-i}+v{\varepsilon_{i}},\,u,\,0),\,y(e_i,\,w{\varepsilon_{-i}},\,a)]y(e_i,\,ua{\varepsilon_{i}},\,0)=\\ = \llbracket x(e_{-i},\,u,\,0)x(v{\varepsilon_{i}},\,u,\,0),\,y(e_i,\,w{\varepsilon_{-i}},\,a)]y(e_i,\,ua{\varepsilon_{i}},\,0)=\\ = \!\,^{x(e_{-i},\,u,\,0)}\llbracket x(v{\varepsilon_{i}},\,u,\,0),\,y(e_i,\,w{\varepsilon_{-i}},\,a)]\cdot\\ \cdot\llbracket x(e_{-i},\,u,\,0),\,y(e_i,\,w{\varepsilon_{-i}},\,a)]y(e_i,\,ua{\varepsilon_{i}},\,0)=\\ = \!\,^{x(e_{-i},\,u,\,0)}\llbracket x(v{\varepsilon_{i}},\,u,\,0),\,y(e_i,\,w{\varepsilon_{-i}},\,a)]y(u,\,w,\,a).\end{gathered}\ ] ] finally , notice that @xmath329 commutes with @xmath330 .
in this section we assume that @xmath3 .
set @xmath339 lemmas [ w = zz ] and [ short - is - three - long ] imply that @xmath340 interchanging roles of @xmath23 and @xmath272 one has @xmath341 by lemma [ short - symmetry ] one has @xmath342 .
now we have to show that @xmath343 commutes with @xmath344 . with this end fix @xmath345 .
observe that @xmath346 commutes with @xmath347 by lemma [ w - conjugation ] .
next , we will show that @xmath348=1.\ ] ] by lemma [ short - symmetry ] one has @xmath349 and @xmath350 now , lemma [ w = zz ] implies that @xmath351 finally , it remains to obseve that @xmath352 commutes with both @xmath353 and @xmath354 by lemma [ w - conjugation ] .
[ w - symmetry ] let @xmath362 and let @xmath115 , @xmath363 be vectors having only @xmath364-th and @xmath365-th non - zero coordinates respectively ; consider also @xmath366 such that @xmath367 .
set @xmath368 .
then @xmath369 the proof of lemma 29 from the another presentation paper can be repeated almost verbatim .
the only difference is that one should use another approach to show that @xmath371 acts trivially on @xmath372 and @xmath373 .
namely , one should observe first that @xmath374 acts trivially on @xmath375 afterwards , decompose @xmath376y(e_j,\,va{\varepsilon_{j}},\,0),\ ] ] thus , @xmath377y(e_j,\,va{\varepsilon_{j}},\,0).\ ] ] since @xmath378 we obtain that @xmath379y(e_j,\,va{\varepsilon_{j}},\,0)=\\ = \!\,^{x(e_{-j},\,v,\,0)}\llbracket x(e_{-k},\,-va{\varepsilon_{k}}{\varepsilon_{j}},\,0),\,y(e_j,\,0,\,a)]\cdot y(v,\,0,\,a).\end{gathered}\ ] ] finally , it remains to observe that @xmath380 acts trivially on @xmath381 .
firstly , choosing some @xmath23 and @xmath272 such that @xmath384 one has @xmath385 now , choosing @xmath345 , taking @xmath386 , @xmath387 and any @xmath157 , and using lemma [ short - is - three - long ] we obtain @xmath388 [ p - relations ] assume that @xmath37 , @xmath417 are columns of symplectic elementary matrices , and let @xmath41 , @xmath418 , @xmath39 be arbitrary coloumns such that @xmath405 , and @xmath419 . then for any @xmath43 , @xmath420 , @xmath421 , @xmath422 , @xmath79 one has @xmath423 to check kl0 use t4 , for kl1 use t3 and t2 .
kl2 follows from the definition of the action .
below we verify the rest .
@xmath442=\\ \phantom{\rm ( kl3)\ } = ( e_{-k},\,t(e_i,\,e_{-j}r{\varepsilon_{-j}},\,0)e_j,\,b{\varepsilon_{-k}},\,0)(e_{-k},\,e_j,\,b{\varepsilon_{-k}},\,0){^{-1}}=\\ \phantom{\rm ( kl3)\ } = ( e_{-k},\,e_j+e_ir,\,b{\varepsilon_{-k}},\,0)(e_{-k},\,-e_j,\,b{\varepsilon_{-k}},\,0)=\\ \phantom{\rm ( kl3)\ } = ( e_{-k},\,e_ir,\,b{\varepsilon_{-k}},\,0)=(e_{-k},\,e_i,\,rb{\varepsilon_{-k}},\,0)=\\=(e_i,\,e_{-k},\,rb{\varepsilon_{-k}},\,0)=\!\,_{ik}(rb);\end{gathered}\ ] ] @xmath443=\\ \phantom{\rm ( kl3)\ } = ( e_{-j},\,t(e_i,\,0,\,r)e_{-i},\,b{\varepsilon_{-j}},\,0)(e_{-j},\,e_{-i},\,b{\varepsilon_{-j}},\,0){^{-1}}=\\ \phantom{\rm ( kl3)\ } = ( e_{-j},\,e_{-i}+e_ir{\varepsilon_{i}},\,b{\varepsilon_{-j}},\,0)(e_{-j},\,-e_{-i},\,b{\varepsilon_{-j}},\,0)=\\ \phantom{\rm ( kl3)\ } = ( e_{-j},\,e_ir{\varepsilon_{i}},\,b{\varepsilon_{-j}},\,-rb^2)=\\ \phantom{\rm ( kl3)\ } = ( e_{-j},\,e_ir{\varepsilon_{i}},\,b{\varepsilon_{-j}},\,0)(e_{-j},\,0,\,b{\varepsilon_{-j}},\,-rb^2)=\\ \phantom{\rm ( kl3)\ } = ( e_{-j},\,e_i,\,rb{\varepsilon_{i}}{\varepsilon_{-j}},\,0)(e_{-j},\,0,\,0,\,-rb^2)=\!\,_{ij}(rb{\varepsilon_{i}})\cdot\!\,_{-j , j}(-rb^2);\end{gathered}\ ] ] @xmath444 @xmath445=\\ \phantom{\rm ( kl3)\ } = ( e_i,\,t(e_i,\,e_{-j}r{\varepsilon_{-j}},\,0)e_j,\,b{\varepsilon_{i}},\,0)(e_i,\,-e_j,\,b{\varepsilon_{i}},\,0)=\\ \phantom{\rm ( kl3)\ } = ( e_i,\,e_j+e_ir,\,b{\varepsilon_{i}},\,0)(e_i,\,-e_j,\,b{\varepsilon_{i}},\,0)=(e_i,\,e_ir,\,b{\varepsilon_{i}},\,0)=\\ = ( e_i,\,e_i,\,rb{\varepsilon_{i}},\,0)=(e_i,0,\,0,\,2rb{\varepsilon_{i}})=\!\,_{i ,- i}(2rb{\varepsilon_{i}}).\end{gathered}\ ] ] the definition of the action together with t5 imply kl7 . obviously , @xmath454 , then @xmath455 by the previous lemma . for arbitrary column of symplectic elementary matrix @xmath37
take @xmath408 such that @xmath456 and use that @xmath449 and @xmath438 preserve action . | we give two definitions of relative symplectic steinberg group and show that they coincide .
= 2 |
similarity between the magnetic and superconducting ordering has been noted long time ago ( see ref .
and the review articles ) . in the simplest case of the bcs superconductivity the order parameter is described by a complex quantity @xmath10 with the amplitude @xmath11 that determines the cooper pair density and the phase @xmath12 related to a voltage @xmath13 via the gauge josephson equation @xmath14with @xmath15 , @xmath16 being the elementary charge .
this relation shows that , in equilibrium , the gauge - invariant quantity @xmath17 equals zero : @xmath18 .
the electrical current @xmath19 of the condensate is expressed in terms of another gauge - invariant quantity the condensate velocity @xmath20 as @xmath21 .
one of the most remarkable phenomena arising in superconducting systems is the josephson effect .
josephson@xcite showed that a condensate current @xmath22can flow in an s / i / s junction being a periodic function of the gauge - invariant phase difference between superconductors @xmath23 , where the limits of the integration @xmath24 and @xmath25 mean respectively the right and left superconductors s separated by an insulating layer i , and @xmath26 is the josephson critical current . in the case of a magnetic ordering ferromagnetic ( f ) or antiferromagnetic ( af)the order parameter is the magnetization @xmath27 ( in the antiferromagnets the vector @xmath28 is the magnetization in one of the sublattices ) .
the absolute value of @xmath28 corresponds to the amplitude @xmath11 in superconductors , and the angle @xmath29 between the @xmath30-axis and the vector @xmath28 is analogous , to some extent , to the phase @xmath31 . the temporal variation of @xmath29 is described by the equation similar@xcite to eq .
( [ i1 ] ) , @xmath32where @xmath33 is the gyromagnetic ratio , @xmath34 is the bohr magneton , @xmath35 and @xmath36 is the magnetic induction in the respectively right or left f- or af - layer composing the josephson - like structure . in eq .
( [ i3 ] ) , the @xmath30-axis is chosen to be directed along the vector @xmath37 . ) in detail since we will study the dc josepshon - like effect only .
more detailed discussion of the nonstationary relation , eq .
( [ i3 ] ) , is given in refs .
flavio , tremblay . ] the josephson - like effect in f / i / f and af / i / af junctions was studied in refs . .
it was found that in these junctions , the spin current has the form @xmath38where @xmath39 and @xmath40 is a constant related to the barrier transmittance @xmath41 . in the full analogy to the josephson
charge current @xmath42 , the spin current is nondissipative .
the absence of the dissipation for the spin current and the possibility to use this dissipationless current in spintronics has been discussed in many papers@xcite ( see also reviews ) .
unfortunately , using the spin currents in spintronics does not seem to be straightforward because the analogy between the spin and charge currents is not complete , namely , the spin current is not a conserved quantity.@xcite we will demonstrate the validity of this statement considering the proximity effect in af / f structures . in particular , we will calculate the spin current through a josephson - like af / f / af junction .
it is well known that the josephson effect in s / f / s junctions reveals new remarkable properties@xcite that are absent in the case of more conventional s / i / s or s / n / s junctions .
first , the condensate functions @xmath43 penetrating into a diffusive f - layer decay on a very short range of the order @xmath44 , where @xmath45 is the classical diffusion coefficient of the ferromagnet and @xmath7 is the exchange field ; the subindex @xmath46 denotes the diffusive case .
this expression for @xmath47 is valid if the condition @xmath48 is fulfilled , where @xmath49 is the momentum relaxation time .
usually , the exchange field @xmath7 is much larger than the critical temperature of the superconducting transition @xmath1 .
in addition , the cooper pairs wave function oscillates in space .
it consists of the singlet component @xmath50 , and the triplet component @xmath51 with zero projection of the total spin on the direction of the magnetization @xmath28 .
such oscillations lead to a change of the sign of the josephson critical current @xmath26 , and the so called @xmath52-state is realized in the junction . another interesting effect predicted in ref .
( see also the reviews ) and observed in several experiments klapwijk , sosnin , birge , chan , blamirescience , westerholt , aarts , deutscher in s / f structures , is related to a new type of superconducting correlations .
this type of correlations called `` odd triplet superconductivity '' can arise due to the proximity effect in the f - layer with a nonhomogeneous magnetization . in the case of a uniform magnetization in the f - layer , the pair wave function there has both the singlet component @xmath53 and the triplet component @xmath54 with zero projection of the total spin of the cooper pair on the direction of the magnetization .
both components oscillate in space and decay on a short length of the order of @xmath55 . in the ballistic case
, these components oscillate with the period @xmath56 .
the situation changes drastically if the magnetization in f is not uniform in the nearest vicinity of the s / f interface.@xcite in this case , not only the components @xmath57 and @xmath51 arise in the f - layer , but also the triplet component @xmath58 with nonzero projection of the total spin on @xmath28 is generated .
this component is insensitive to the exchange field and penetrates the f - layer over a long ( compared to @xmath59 ) distance .
the penetration length does not depend on the exchange field @xmath60 and is much larger than @xmath61 . in absence of the spin - orbit interaction and scattering by magnetic impurities
this length may become comparable with the penetration length of the superconducting condensate into the diffusive normal ( nonmagnetic ) metal @xmath62 .
the component @xmath63 can be called the long - range triplet component ( lrtc ) .
it is symmetric in the momentum space and is therefore also insensitive to the scattering by ordinary ( nonmagnetic ) impurities . in this respect
the lrtc differs drastically from the usual triplet component @xmath64 that describes superconductivity in @xmath65.@xcite the latter is an antisymmetric function of the momentum @xmath66 and is therefore destroyed by ordinary impurities .
the correlator @xmath63 changes sign by permutation of the operators @xmath67 and @xmath68 , as it should be according to the pauli principle , which follows from the fact that the fermi operators @xmath67 and @xmath68 anticommute at equal times .
this means that @xmath69 , i.e. , the function @xmath70 is an odd function of time @xmath71 or , in other words , it is an odd function of the frequency @xmath72 in the fourier representation .
that is why this component is called `` odd triplet component '' .
recent experiments unambiguously confirmed the existence of the lrtc in the s / f structures with a nonhomogeneous magnetization.klapwijk,sosnin,birge,chan,blamirescience,westerholt,aarts,deutscher in this paper we show that the analogy between the superconducting and magnetic order parameters in josephson - like junctions is deeper than it was indicated previously .
it turns out that in the sdw / f / sdw junction with a ballistic f - layer the `` short''- and long - range af order parameters are induced in the f - layer in an af / f structure ( we use the quotation marks for `` short''-range , to indicate that in the considered ballistic case this component does not decay on a short range , but rather oscillates with a short period ) .
the component of the af order parameter with the @xmath28 vector ( in each sublattice ) being parallel to the magnetization vector in the f - layer @xmath73 , which is supposed to be oriented along the @xmath30-axis , penetrates into the f - layer and oscillates with a short period of the order of @xmath74 . on the other hand ,
the sdw with the @xmath28 vector perpendicular to @xmath73 penetrates the f - layer over a long distance of the order of @xmath75 and decays monotonously .
this part of the sdw order parameter may be called the long - range component induced in the ferromagnetic layer . at the same time , in spite of the long - range penetration of the af correlations into the ferromagnet f , the spin current that arises at the af / f interface is not constant in space .
it oscillates in the f region with a short period .
the study of the spin current in various heterostructures is important not only from the point of view of fundamental physics but also from the point of view of possible applications of these structures in spintronics.@xcite we do not analyze the case of the bose
einstein condensation in the magnon system at a high magnon density which can be realized via the external pumping . in this case , a josephson - like effect may also arise , but it has different properties ( see the review and references therein ) .
in order to describe the af order we adopt the model developed for superconducting fe - based pnictides.chubukov09,eremin10,chubukov10,tesanovic,schmalian10 in this model the band structure is assumed to consist of two bands ( electron and hole ) with perfect or almost perfect nesting , which is taken into account by a certain parameter @xmath76 . due to this ,
a spin density wave ( sdw ) corresponding to an antiferromagnetic order parameter arises in these materials . on the basis of this model , which describes well many properties of this new class of high-@xmath1 superconductors ( see review articles mazin , norman , wilson , johnston , stewart , chubukov11rev and references therein ) , the generalized eilenberger equations for the quasiclassical green s functions @xmath77 have been derived in the previous papermoorve11 by the authors ( see also a recent paper @xcite where the definition of the green s functions differs from ours ) .
the quasiclassical formalism is especially suitable for tackling nonhomogeneous problems and the matrix functions @xmath77 allow one to find all the necessary observable quantities , such as the charge and spin currents , the density - of - states , etc .
the technique of quasiclassical green s functions is presented , e.g. , in refs .
rammer , belzigrev , kopnin .
first , we consider a josephson - like af / f / af junction composed of two leads and a thin ferromagnetic layer between the leads . in the leads
a spin density wave exists , which is a particular example of the af order and the layer between these materials is a ferromagnet in the ballistic regime ( see fig .
[ fig : setup ] ) . for simplicity , we assume that both the sdw and the f parts of the structure consist of similar two - band materials and differ from each other by different interaction constants and doping levels .
the af / f interface penetrability is supposed to be small and thus one can neglect the effect of the f - layer on the sdw leads .
the green s functions in the right ( left ) leads are given by the formula that can be obtained from eq .
( 6.4 ) of ref . by setting the superconducting order parameter to zero ,
@xmath78 ) @xmath79 \big\rbrace \ , .
\notag\end{aligned}\]]here , @xmath80 is the matsubara frequency , @xmath81 is the amplitude of the sdw , @xmath82 and @xmath83 is the tensor product of pauli matrices in the band , gorkov nambu and spin space , respectively .
the magnetization of the sdw lies in the @xmath84-@xmath30-plane at the angle @xmath85 relative to the @xmath30-axis , with the projection on the @xmath84-axis @xmath86 in the left ( right ) leads . for simplicity ,
we suppose the ideal nesting and set @xmath87 . in the f - layer ,
the quasiclassical green s functions obey the equation @xmath88 = 0 \ , , \label{m2}\]]where @xmath89 and @xmath6 is the modulus of the fermi velocity .
the matrix @xmath90 is defined as @xmath91 with the exchange field in the ferromagnet @xmath7 .
we assume that there is no impurity scattering in the f - layer ( ballistic case ) .
the impurity scattering leads to a suppression of the sdw amplitude.@xcite equation ( [ m2 ] ) is supplemented by the boundary condition which we write in the form @xmath92 \ , .
\label{m3}\]]the transmission coefficient @xmath93 is a matrix , which is assumed to be small .
this boundary condition generalizes the widely used boundary condition derived by zaitsev@xcite to the case of a two - band metal with an sdw .
we consider a simplified model for the af / f interface and assume that there are no interband transitions . in this case , the matrix @xmath93 has the form @xmath94 , where @xmath95 and @xmath96 are the matrix tunneling elements for transitions from the band @xmath97 resp .
@xmath98 of the f - layer to the same bands @xmath97 resp .
@xmath98 of the sdw leads ( see the appendix [ boundary_conditions ] ) . if these elements are equal , i.e. , @xmath99 , then @xmath100 and @xmath101 .
the boundary conditions that describe real materials are more complicated .
they must include the transitions between different bands and this process may be accompanied by spin flips .
however , we are not interested in exact calculations of the amplitude of the sdw penetrating into the ferromagnetic layer , but rather in a qualitative effect the appearance of `` short''- and long - range components
. the existence of these components does not depend on the exact form of the boundary conditions . in the lowest order of approximation in the transmission coefficient @xmath102 ( no proximity effect )
, the matrix @xmath77 in the ferromagnet has the form @xmath103 .
therefore , the boundary condition can be written in the form @xmath104 \ , , \label{m4}\]]where the matrix @xmath105 equals @xmath106and @xmath107 is an effective transmission coefficient for the sdw .
the boundary condition ( [ m4 ] ) defines the spin current through the interface .
one can see that if one of the tunneling elements ( @xmath108 or @xmath109 ) is zero , then the spin current turns to zero .
we need to find a solution of eq .
( [ m2 ] ) supplemented with the boundary condition ( [ m4 ] ) . in the considered case of a low penetrability of the sdw / f interface , the matrix @xmath77 in the f - film can be represented in the form @xmath110 , where @xmath111 is a matrix with a small `` amplitude '' which describes the sdw induced in the f - film due to the proximity effect . as usual , we represent @xmath77 as a sum of a symmetric in momentum part and the antisymmetric one , i.e. , @xmath112 .
we substitute this expression into eq .
( [ m2 ] ) and split it into two equations for the symmetric and antisymmetric parts @xmath113 & = 0 \ , , \label{m2a } \\
n_{x}^{2 } v \partial_x \check{a } + \big [ \check{\lambda } \
, , \check{s } \big ] & = 0 \ , .
\label{m2b}\end{aligned}\]]excluding the matrix @xmath114 , we obtain one equation for the matrix @xmath115 @xmath116 \big ] = 0 \ , .
\label{m6}\]]for the matrix @xmath114 we obtain a similar equation .
the commutator @xmath117 \big]}$ ] is easily calculated resulting in @xmath118 \big ] = 2 \big\lbrace ( \omega^{2 } - { \mathcal{h}}^{2 } ) \check{s } & + 2 \mathrm{i } \mathcal{h } \omega ( \hat{\sigma}_{3 } \cdot \check{s } + \check{s } \cdot \hat{\sigma}_{3 } ) \\ & + \omega^{2 } \check{s } - \mathcal{h}^{2 } \hat{\sigma}_{3 } \cdot \check{s } \cdot \hat{\sigma}_{3 } \big\rbrace \ , .
\notag\end{aligned}\ ] ] here , we used the property @xmath119 .
it is seen from the boundary conditions ( [ m4][m5 ] ) , that this matrix anticommutes with @xmath120 .
it is convenient to represent the matrix @xmath115 as a sum of two matrices @xmath121 , where @xmath122 and @xmath123 .
we will see that the matrix @xmath124 describes a `` short''-range component of the sdw penetrating into the f - layer , while @xmath125 determines the long - range component , which is perpendicular to the orientation of the magnetization in the ferromagnet . from eq .
( [ m6 ] ) we obtain the equations for @xmath124 and @xmath125 @xmath126 in the first order in the transmission coefficient @xmath127 the solutions of these equations obeying the boundary conditions are @xmath128 \cdot \cos ( \alpha/2 ) \ , , \label{m9a } \\
\check{s}_{\perp } & = 2 m s_{\omega } \mathcal{t}_{\text{eff}}^{2 } \hat{\rho}_{2 } \cdot \hat{\tau}_{1 } \cdot \hat{\sigma}_{2 } \cdot \mathcal{c}(x ) \cdot \sin ( \alpha/2 ) \
, , \label{m9b}\end{aligned}\ ] ] where @xmath129 and @xmath130 is the dimensionless amplitude of the sdw induced in the f - layer and @xmath131 with @xmath132 and @xmath133 .
the here introduced wave vectors @xmath134 and @xmath135 determine the characteristic lengths which describe the penetration of the `` short''- and long - range components of the sdw into the f - region , respectively .
the same wave vectors characterize the penetration of the short - range component , i.e. , the singlet and triplet component with zero projection of the total spin onto the @xmath30-axis , and the long - range triplet component of the superconducting wave function into the f - region of an s / f bilayer in the ballistic case.@xcite the component of the sdw @xmath124 , eq .
( [ m8a ] ) , which can be called the `` short''-range component , oscillates at a given angle @xmath136 rather fast in space with a period of the order of @xmath74 ( it is assumed that @xmath137 ) . being averaged over the angle , it decays from the interface in a power - law fashion .
the long - range component @xmath125 , eq . ( [ m8b ] ) , penetrates into the ferromagnet over a long length @xmath138 , which does not depend on the exchange field @xmath7 .
thus , there is an analogy with an s / f / s josephson junction where the ferromagnet has an inhomogeneous magnetization.@xcite in the ballistic case , the `` short''-range component of the condensate wave function oscillates in space with a period @xmath139,@xcite whereas the long - range component penetrates the ferromagnet over the length @xmath140 .
we use quotation marks for the `` short''-range component because , in the considered ballistic case , this component does not decay exponentially in the ferromagnetic layer , but oscillates and decays in a power - law fashion after averaging over the angles . in the diffusive limit in the s / f
/ s junctions , this component decays exponentially on a short distance of the order @xmath55 .
note that the `` short''-range component @xmath124 consists of the odd and even in @xmath141 parts ( correspondingly , the first and the second terms in eq .
( [ m9a ] ) ) , whereas the long - range component @xmath125 is an odd function of the frequency @xmath141 .
the same dependence has an anomalous ( gorkov s ) green s function in an s / f / s system with a nonhomogeneous magnetization @xmath28.@xcite in the last case , the short - range part consists of the singlet component and triplet component with zero projection of the total spin onto the @xmath28 vector , and the long - range part consists of the triplet component with a non - zero projection of the total spin onto the @xmath28 vector . in fig .
[ fig : re_im_c ] a ) we plot the coordinate dependence of the `` short''- ( the functions @xmath142 and @xmath143 ) and the long - range ( the function @xmath144 ) components . the former stems from the part of the matrix @xmath77 corresponding to the magnetization component of the sdw @xmath28 parallel to the magnetization in the ferromagnet @xmath145 , i.e. , the second term on the right in eq .
( [ m1 ] ) .
the long - range component stems from the sdw component perpendicular to the @xmath73 vector and turns to zero at @xmath146 .
the components of @xmath114 that determine the spin current @xmath147 can be presented in the same form as the @xmath148 matrix , i.e. , @xmath149 .
the `` longitudinal''@xmath150 , and the `` transverse''@xmath151 , parts are related with @xmath148 through the simple formulas ( see eq . ( [ m2b ] ) ) @xmath152[m11b ] ) we find the matrix @xmath114 : @xmath153 \cdot \cos ( \alpha/2 ) \ , , \label{m11c } \\
\hat{\tau}_{3 } \cdot \check{a}_{\perp } & = -2 m \frac{\mathcal{t}_{\text{eff}}^{2}}{|n_{x}| } \hat{\rho}_{2 } \cdot \hat{\tau}_{1 } \cdot \hat{\sigma}_{2 } \cdot \bar{\mathcal{c}}(x ) \cdot \sin ( \alpha/2 ) \ , , \label{m11d}\end{aligned}\]]where @xmath154 these functions are plotted in fig .
[ fig : re_im_c ] b ) and contribute to the `` short''- and the long - range components of the green s function in the f - layer , more exactly to the odd in @xmath155 part of it .
as it follows from the expression for the spin current ( see eq .
( 3.13 ) in ref . ) ) , the obtained formulas for @xmath156 lead to zero spin current inside the f - film .
this result is quite natural because the spin josephson - like current in sdw / f / sdw junction should be proportional to the penetration probabilities of both interfaces , i.e. , @xmath157 . in the symmetric case under consideration @xmath158 . in order to obtain a finite spin current
, we have to calculate the matrix @xmath114 in the next order in the transmission coefficient @xmath159 .
we can easily find the corrections @xmath160 from the normalization condition @xmath161 as it follows from this equation @xmath162 where @xmath163 and @xmath164 .
therefore , for @xmath165 we obtain the equation @xmath166where @xmath167 .
we are interested in the component @xmath168 of the matrix @xmath165 , which is proportional to the matrix @xmath169 , because only this component contributes to the spin current .
this component commutes with the matrix @xmath170 . substituting expressions ( [ m9a][m10d ] ) into eq .
( [ m14a ] ) , we obtain for @xmath171 @xmath172}{\sinh(\theta_{+ } ) \cosh(\theta_{\omega } ) } \bigg\rbrace \sin ( \alpha ) \hat{\rho}_{3 } \cdot \hat{\tau}_{3 } \cdot \hat{\sigma}_{1 } \ , .
\label{m15}\ ] ] one can write eq .
( [ m15 ] ) in the form @xmath173 where @xmath174 .
now , the spin current can be easily calculated .
we see that the sdw / f / sdw junctions are very similar to the s / f / s josephson junctions . in analogy to the `` short''- and long - range components of the superconducting condensate penetrating the ferromagnet , there are similar `` short''- and long - range components of the sdw penetrating the ferromagnet . nevertheless , there is an essential difference between the two systems .
the josephson charge current @xmath42 in an s / f / s junction is constant in space in the ferromagnet .
it is determined by the long - range component , provided the magnetization @xmath175 in the f - layer is not uniform and the thickness @xmath176 of this layer is not too small ( @xmath177 ) .
in contrast , the spin current in an sdw / f / sdw junction is not constant in the f - layer but oscillates with a period of the order of @xmath59 ( see fig .
[ fig : jspx_on_x ] ) .
it is constant in space only in the absence of the exchange field , i.e. , in sdw / n / sdw junction .
the spin current through the sdw / f / sdw junction can easily be calculated using the formula for @xmath3 ( see eq .
( 3.13 ) of ref . )
@xmath178 the notation @xmath147 means the spin current density with the spin projection @xmath179 ( resp .
@xmath84 ) flowing perpendicular to the s / f interface and the angle brackets denote the angle averaging . substituting eq .
( [ m15 ] ) into eq .
( [ spincurrgeneral ] ) we obtain @xmath180where @xmath181one can see that the spin current oscillates in space in the f - layer .
it is also seen that in the case of a normal ( nonmagnetic ) metal , i.e. , at @xmath182 , the spin current is constant in space . as follows from eq .
( [ j_c(x ) ] ) , the spin current at the interfaces it is equal to @xmath183 using eq . ( [ j_c(x ) ] )
we plot the coordinate dependence of the critical spin current with the projection of the spin onto the @xmath179-axis in fig .
[ fig : jspx_on_x ] for different values of @xmath7 .
the critical spin current oscillates with a period of the order of @xmath74 . on the left panel
we choose @xmath184 so that there is no dependence of the oscillation amplitude on @xmath179 ; on the right panel we choose @xmath185 , so that the amplitude depends on @xmath179 in a power - law fashion , which is a consequence of the integration over the angles .
thus , we see that the sdw order parameter penetrating into the ferromagnet contains two components a long - range component with @xmath28 vector perpendicular to the magnetization in the ferromagnet @xmath73 , and a `` short''-range component with @xmath28 vector parallel to the @xmath30-axis .
a similar effect arises in an s / f / s junction with the nonhomogeneous magnetization , where the long - range component is the odd triplet ( @xmath186 ) wave function of the cooper pairs , and the short - range component corresponds to the singlet and @xmath187 triplet wave function.@xcite however , the spin current @xmath3 is not analogous to the charge josephson current @xmath42it is not constant in space , but oscillates with a period @xmath139 .
now , we consider a junction formed by two materials with the sdw . these can be iron - based pnictides separated by an insulating , normal metal or ferromagnetic layer . in all these cases
the current through the junction is given by @xmath188where the critical spin current depends on the type of the barrier .
for example , for sdw / n / sdw junction it is equal to @xmath189 here , the superscript @xmath179 denotes the component of the spin current vector having the spin projection on the @xmath179-axis only . for the case of sdw / i / sdw junction
we easily obtain @xmath190 introducing @xmath191 we can write this equation in the case of ideal nesting as @xmath192 which corresponds to the ambegaokar baratoff formula for the critical josephson current with @xmath193 replaced by @xmath81 and the `` interface resistance per unit area '' @xmath194 is related to the effective transmission coefficient@xmath195 .
equation ( [ af / af ] ) for the spin `` josephson '' current in an af / i / af junction has been obtained in ref . by another method .
now , we derive an equation similar to the ferrell prange@xcite equation .
we consider a junction with sdw leads in the form of thin films .
the thickness @xmath46 of the films must be less than a characteristic length , over which the sdw vector @xmath28 restores its preferable direction .
this direction is determined by anisotropy effects .
the idea of the derivation is the same as in ref . .
considering the region near the boundary ( cf .
[ fig : ferrell_prange ] ) we use the continuity equation for the spin @xmath196 , @xmath197which can be derived easily from eq .
( 3.5 ) of ref . .
using the green s functions @xmath77 unperturbed by the proximity effect ( see eq .
( [ m1 ] ) ) , one can show that @xmath198 . from the continuity equation we get @xmath199 the formula for the spin current @xmath200 in an sdw lead with the magnetization orientation varying in space
can be easily obtained from eq .
( [ m2 ] ) in the same way as the meissner current is found in a clean superconductor ( see , e.g. , ref . ) .
it has the form ( see appendix [ ferrell_prange_appendix ] ) @xmath201 this equation formally coincides with the corresponding formula for the supercurrent in a clean superconductor if one replaces @xmath202 , @xmath203 and @xmath204 , where @xmath31 is the phase in the superconductor . substituting eq .
( [ af / af ] ) into eq .
( [ fpcont ] ) we arrive at the final result @xmath205where @xmath206 is a characteristic length inversely proportional to the critical current @xmath207 .
it equals @xmath208 for sdw / i / sdw junctions .
equation ( [ fpfinal ] ) is the analog of the ferrell prange equation for a junction composed of two materials with the sdw and separated by some barrier ( a thin insulating or normal metal layer ) .
one can easily find a solution similar to the one describing a fluxon in the josephson tunnel junctions .
it has the form @xmath209 \ , , \label{fluxon}\]]where the size of the `` fluxon '' is given by @xmath210 .
the mutual orientation changes along the @xmath30-axis from @xmath211 to @xmath212 and has the value @xmath52 in the core of the `` fluxon '' , cf . fig .
[ fig : spin_fluxon ] . in the case of the josephson fluxon
one has a circulating charge current in our case , the spin current circulates in the `` fluxon '' . up to now , we assumed the ideal nesting , @xmath87 . in the case
@xmath213 we find the quasiclassical green s function in the sdw system by inverting the matrix @xmath214 in eq .
( 3.3 ) of ref . .
it turns out that @xmath77 is a linear combination of six `` basis matrices '' @xmath215where the subscripts l and r denote the left and right sdw systems , respectively , corresponding to the setup depicted in fig .
[ fig : setup ] with the f - layer replaced by a very thin insulating barrier .
the six functions @xmath216 are @xmath217where @xmath218 .
the detuning parameter @xmath219 can be written in the form @xmath220 with @xmath221 describing the relative size mismatch of the `` hole '' and `` electron '' pockets and @xmath222 describes the ellipticity of the `` electron '' pocket , while the `` hole '' pocket is assumed to be a circle.@xcite from the quasiclassical green s function , eq . ( [ eqn : qgf_in_af ] ) , we can calculate the dependence of various quantities of interest on @xmath223 .
we begin with the density of states ( dos ) .
the dos is given by the formula @xmath224 where the angle brackets mean the averaging over the momentum directions @xmath225 inserting the expression for the quasiclassical green s function into eq .
( [ eqn : dos ] ) we obtain @xmath226this result was obtained earlier in refs . .
we plot the dos for different values of the ratio @xmath227 in fig .
[ fig : dos_af ] choosing a constant @xmath228 , where @xmath229 is the sdw order transition temperature .
note that there is a range of the values of @xmath230 where an energy gap appears in the dos , indicating an insulating phase of the system .
we can trace the transition to the metal considering the value of the dos at zero energy , @xmath231 , as a function of the parameter @xmath222 . in fig .
[ fig : wm_on_mu_phi ] we plot the dependence of the order parameter @xmath81 on @xmath222 and , in the same plot , show the dependence @xmath231 on @xmath222 ( dashed line ) .
one can observe a rather steep increase of @xmath231 in the region around @xmath232 .
the dependence of the sdw order parameter @xmath81 on @xmath222 is calculated from the self - consistency equation , eq .
( 3.19 ) of ref . , which in addition has to be averaged over the momentum directions .
it reduces to the form ( see also ref . )
@xmath233 it is of special interest@xcite to calculate the spin current through the sdw / i / sdw system . using the expression eq .
( [ eqn : qgf_in_af ] ) for the quasiclassical green s functions in the left and right leads , and inserting it into the expression for the spin current , eq .
( 3.13 ) of ref .
, we obtain the familiar form of the josephson - like current @xmath234for the spin current through the sdw / sdw interface which has only the spin projection on the @xmath179-axis .
the critical coefficient @xmath40 is dependent on @xmath76 @xmath235 in fig . [ fig : spin_current_on_mu ] we show the dependence of the critical spin current on the two parameters @xmath236 and @xmath222 , choosing in each case the other parameter constant when plotted as a function of @xmath236 , we set @xmath237 ; plotting the spin current as a function of @xmath222 we set @xmath238 .
one can see from fig .
[ fig : spin_current_on_mu ] and fig .
fig : wm_on_mu_phi that in the insulating phase the spin current is not zero . in contrary , it is at its maximum .
this addresses the question about the possibility of the appearance of the spin current in an insulator ( see discussion on p. 12 of ref . ) . analyzing the expression ( [ eqn : spin_current_on_mu ] ) for the critical spin current
we see that there is a possibility of a sign change . at small temperatures , the spin current changes sign
if , e.g. , the condition @xmath239 holds with @xmath240 at the same time .
we note the similarity of the self - consistency equation ( eqn : self_consistency ) with the self - consistency equations presented in refs . .
setting @xmath241 we can map the problem of finding the dependence of @xmath81 on @xmath221 onto the problem of finding the dependence of the superconducting order parameter on a strong exchange field resulting in the so - called larkin
fulde ferrell ( loff ) state , which is characterized by the spatial dependence of @xmath242 .
note that the possibility of the loff - like state in pnictides was noted earlier by gorkov and teitelbaum.@xcite
we have considered josephson - like junctions of the sdw / f / sdw and sdw / i / sdw types and calculated the spin current @xmath3 in these systems . in both the cases
the dependence of @xmath3 on the angle @xmath243 between the mutual orientations of the magnetization vectors @xmath28 in the right and left leads is given by eq .
( [ af / af ] ) .
the critical current density @xmath244 is determined by the interface transparencies .
we emphasize the analogy between the sdw / f / sdw junctions and the josephson s / f / s junctions with a nonhomogeneous magnetization . in both the systems
there are `` short''- and long - range components of the order parameters that penetrate the ferromagnet . in the ballistic systems under consideration
the `` short''-range sdw component penetrating the f - layer oscillates with a period of the order of @xmath245 .
it stems from the sdw component parallel to the magnetization vector @xmath73 in the ferromagnet .
the sdw component normal to the @xmath73 vector penetrates the ferromagnet on a long distance of the order of @xmath246 .
the penetration depth of this component does not depend on the exchange field @xmath7 , and , in this sense , this component is analogous to the odd triplet long - range component in s / f / s junctions with a nonhomogeneous magnetization .
however , there is an essential difference between the spin current in sdw / f / sdw junctions @xmath247 and the josephson charge current @xmath42 in s / f / s junctions . whereas the josephson current @xmath42 , as it should be , does not depend on @xmath179 ( @xmath248 ) , the spin current @xmath247 oscillates in space with a period of the order @xmath249
this behavior is a consequence of the fact that the spin current arises due to the interference between the `` short''- and long - range component of the sdw in the ferromagnet .
the spin current in sdw / i / sdw junctions depends on the angle @xmath250 in the same way as in the sdw / f / sdw junctions , i.e. , @xmath251 .
we have calculated the dependence of @xmath3 in an sdw / i / sdw junction on the parameter @xmath219 characterizing the deviation from the ideal nesting . at @xmath252 , an energy gap @xmath81 opens in the excitation spectrum in the leads with the sdw ( see fig .
[ fig : wm_on_mu_phi ] ) . in a certain region of the parameter
@xmath76 , the energy gap disappears and the leads of the junction become metals ( or semimetals ) . the spin current exists in both the cases ; it turns to zero at the neel temperature when @xmath253 .
furthermore , we have derived the ferrell prange equation for the josephson - like junctions with the sdw and found a solution describing a localized spin current distribution analogously to the fluxon in a tunnel s / i / s josephson junction .
the characteristic length of the `` fluxon '' , @xmath254 , is determined by the barrier transmittance and the thickness @xmath46 of the leads with the sdw . in the case of bulk
leads the thickness @xmath46 should be replaced by a characteristic length over which the vector @xmath28 of the sdw restores its favorable direction in the bulk , which is determined by the anisotropy effects .
the authors are grateful to a. chubukov and i. eremin for useful remarks and discussions .
we are also indebted to g. e. volovik for useful comments .
we thank sfb 491 for financial support .
here , we generalize the boundary conditions for quasiclassical green s functions to the case of a two - band material with the sdw . it is not in the scope of the paper to present a rigorous derivation of the boundary conditions with account for all the possible processes accompanying the passage of electrons through the interface .
we ignore the interband transitions and also the spin - flip processes .
in addition , we restrict ourselves to the case of a low transmittance coefficient .
thus , the derived boundary conditions , to some extent , are phenomenological .
however , in a single - band case they coincide with the boundary conditions derived by zaitsev@xcite and in case of two - band material they are compatible with the boundary conditions used in ref . .
note also that the exact form of the boundary conditions is not important for our qualitative conclusions such as the appearance of the `` short''- and long - range sdw components in the ferromagnet . in the case of weak penetration and single - band materials , the boundary conditions for the green s function @xmath77 have the form of eq .
( [ m3 ] ) , in which the matrix @xmath93 should be replaced by a scalar @xmath102 ( see ref . ) . the green s function on the left - hand side of this equation determines the charge current through the interface .
it equals the commutator multiplied by @xmath255 and consisting of the symmetric in momentum space green s functions .
this part of eq .
( [ m3 ] ) , which of course also determines the charge current , can be obtained by using the tunneling hamiltonian method ( see , e.g , ref .
volkov75 ) . in terms of the operators @xmath256 and
@xmath257 introduced in ref .
the tunneling hamiltonian can be written as follows @xmath258where @xmath259 and the coefficients @xmath260 are defined in eq .
( [ m3 ] ) in terms of the matrix elements @xmath96 describing the electron transition through the interface in each band .
the green s function @xmath261 satisfies the equation ( see , e.g , ref . )
@xmath262where @xmath263 .
this means that the self - energy part has formally the same form as in the case of the impurity scattering , but the green s function in @xmath264 describes another ( right ) lead in the equation for the green s function @xmath265 of the left lead .
the charge current through the interface is determined by the trace of a commutator @xmath266 \big\ } = \mathrm{tr } \big\ { \hat{\rho}_3 \cdot \hat{\tau}_3 \cdot \hat{\sigma}_0 \cdot \big [ \check{\mathcal{t } } \hat{g}_{\text{r } } \mathcal{\check{\mathcal{t } } } \ , , \hat{g}_{\text{l } } \big ] \big\ } \ , .
\label{ap3}\end{aligned}\ ] ] the expression in the curly brackets expressed in terms of the quasiclassical green s functions appears in the right - hand side of eq .
( [ m3 ] ) .
in order to obtain the expression for the spin current in the left ( right ) lead flowing parallel to the interface , we assume that the angle @xmath243 varies in space slowly so that the change of @xmath243 on a short coherence length @xmath267 is small .
the characteristic length @xmath268 of the @xmath243 variation is inversely proportional to the critical current @xmath269 ( see below ) and is much larger than @xmath270 .
if the angle @xmath243 is spatially constant , then the green s function @xmath271 obeys the equation @xmath272 = 0 \ , .
\label{a1}\ ] ] the solution of this equation is given by eq .
( [ m1 ] ) and may be represented in the form @xmath273where the matrix @xmath274 does not depend on @xmath243 .
the unitary matrix @xmath275 is defined as @xmath276 .
if the angle @xmath277 slowly depends on the coordinate along the interface , we have to add a correction @xmath278 to the matrix @xmath279 which depends on @xmath280@xmath281 .
the matrix @xmath114 satisfies the equation @xmath282where the matrices @xmath283 and @xmath284 are related in the same way as matrices @xmath285 and @xmath279 , i.e. , @xmath286 . here
, we used @xmath287 which follows from the normalization condition eq .
( [ m12 ] ) .
one easily finds @xmath288 | we study josephson - like junctions formed by materials with antiferromagnetic ( af ) order parameters . as an antiferromagnet
, we consider a two - band material in which a spin density wave ( sdw ) arises . this could be fe - based pnictides in the temperature interval @xmath0 , where @xmath1 and @xmath2 are the critical temperatures for the superconducting and antiferromagnetic transitions , respectively . the spin current @xmath3 in af / f / af junctions with a ballistic ferromagnetic layer and in tunnel af / i / af junctions is calculated .
it depends on the angle between the magnetization vectors in the af leads in the same way as the josephson current depends on the phase difference of the superconducting order parameters in s / i / s tunnel junctions .
it turns out that in af / f / af junctions , two components of the sdw order parameter are induced in the f - layer .
one of them oscillates in space with a short period @xmath4 while the other decays monotonously from the interfaces over a long distance of the order @xmath5 ( where @xmath6 , @xmath7 and @xmath8 are the fermi velocity , the exchange energy and the temperature , respectively ; the subindex @xmath9 denotes the ballistic case ) .
this is a clear analogy with the case of josephson s / f / s junctions with a nonhomogeneous magnetization where short- and long - range condensate components are induced in the f - layer .
however , in contrast to the charge josephson current in s / f / s junctions , the spin current in af / f / af junctions is not constant in space , but oscillates in the ballistic f - layer .
we also calculate the dependence of @xmath3 on the deviation from the ideal nesting in the af / i / af junctions .
the spin current is maximal in the insulating phase of the af and decreases in the metallic phase .
it turns to zero at the neel point when the amplitude of the sdw is zero and changes sign for certain values of the detuning parameter . |
dust is one of the most important factors for the understanding of galaxy evolution .
since hydrogen molecules are efficiently formed on the surface of dust grains , the molecular formation rate is much larger than the case without dust .
such an enrichment of molecular abundance by dust realizes a favorable condition for star formation ( e.g. , * ? ? ?
dust grains also absorb stellar light mainly at ultraviolet and optical wavelengths and re - emit in the infrared .
consequently , dust affects the spectral energy distribution ( sed ) of galaxies ( e.g. , * ? ? ?
furthermore , the formation rate of hydrogen molecules on the grain surface and the mass absorption coefficient of radiation depend strongly on the grain size distribution ( e.g. , * ? ? ?
* ; * ? ? ?
dust grains form by condensation of elements heavier than helium ( i.e. , metals ) .
metals are mainly supplied from asymptotic giant branch stars ( agb stars ) and supernovae ( sne ) , and part of them condense into dust grains ( e.g , .
dust grains are not only supplied by stars but are also destroyed by sn shocks in the interstellar medium ( ism ) ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
furthermore , it is thought that metal accretion onto the surface of grains in the ism ( referred to as `` grain growth '' in this paper ) is an important process for explaining the amount of dust in the milky way ( e.g. , * ? ? ?
* ; * ? ? ?
to the present day , there have been a lot of studies that investigate the evolution of the total dust mass in galaxies by taking into account these processes ( e.g. , * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
they assumed a representative grain size , but the efficiencies of dust destruction and grain growth depend on the grain size distribution .
thus , we should consider the evolution of the grain size distribution to understand the total dust mass precisely .
the grain size distribution is derived from observed extinction curves ( which mainly depend on the grain size distribution and the grain species ) . according to @xcite ,
if spherical grains are assumed , the extinction curve in the milky way is reproduced by @xmath12 ( @xmath13 m ; this grain size distribution is referred to as the mrn distribution ) , where @xmath6 is the grain radius and @xmath14d@xmath6 is the number density of grains in size interval @xmath15 $ ] ( see * ? ? ?
* ; * ? ? ?
* for more detailed fitting to the milky way extinction curve ) .
the situation seems to be very different for distant galaxies .
recently , @xcite discussed the extinction curves of seven quasars at high redshift ( @xmath16 ) .
they showed that these extinction curves tend to be flat at wavelengths @xmath17 m in the quasar s rest frame .
the difference between extinction curves in distant and nearby objects may indicate that different processes dominate the dust evolution at different epochs . in young galaxies , type ii sne
( sne ii ) are thought to be the dominant sources of dust because of short lifetime of their progenitors .
however , @xcite suggested that agb stars are also important sources of dust production even at galactic age less than @xmath18 gyr .
in addition , grain growth is expected to be the dominant process to increase dust mass in galaxies if the metallicity becomes larger than a certain value @xcite .
furthermore , if the metallicity reaches a sub - solar value , grain grain collisions in the ism ( shattering and coagulation ) become efficient enough to change the grain size distribution significantly ( e.g. , * ? ? ?
we call all processes affecting the grain size distribution ` dust processes ' .
these dust processes affect the different sizes of grains in galaxies .
@xcite showed that sne ii supply relatively large grains ( @xmath19 @xmath20 m ) into the ism because small grains are destroyed by reverse shocks before they are ejected into the ism ( see also * ? ? ?
* ; * ? ? ?
the size distribution of grains produced by agb stars is thought to be biased to large ( @xmath21 m ) sizes ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
furthermore , the smaller grains in the ism are more easily destroyed by interstellar shocks driven by sne ( e.g. , * ? ? ?
if grain growth occurs , since the timescale of this process is proportional to the volume - to - surface ratio of a grain , smaller grains grow more efficiently ( e.g. , * ? ? ?
after the dust grains are released into the diffuse ism , shattering can also occur .
@xcite showed that large grains ( @xmath22 m ) acquire larger velocity dispersions than the shattering threshold velocity if the grains are dynamically coupled with magnetized interstellar turbulence .
shattering is indeed a promising mechanism of small grain production ( e.g. , * ? ? ?
shattering also occurs in sn shocks @xcite . in dense regions ,
coagulation can occur , so that the grain size distribution shifts to larger sizes ( e.g. , * ? ? ?
* ; * ? ? ?
the various dust processes above in galaxies occur on timescales dependent on the metallicity , the total dust amount , the grain size distribution , and so on .
hence , it is crucial to consider all dust processes in a unified framework to understand the evolution of both the total dust amount and the grain size distribution .
there have been a number of studies on the evolution of the grain size distribution in galaxies .
@xcite discussed the evolution of grain size distribution considering dust destruction by sn shocks and grain growth .
however , they did not consider shattering and coagulation by grain
grain collisions .
@xcite suggested a dust evolution model in a multi - phase ism [ warm neutral medium ( wnm ) and cold neutral medium ( cnm ) ] , and also considered the collisional processes of dust grains .
however , they did not consider the size distribution of grains released by stars in order to simplify their model .
@xcite discussed the grain size distribution in young starburst galaxies .
they assumed that sne ii are the source of dust in these galaxies and focused on the production of small grains by shattering .
@xcite constructed a dust evolution model taking into account dust formation and destruction by sne ii along with the formation and evolution of galaxies . however , since they focused on galaxies in the high-@xmath23 universe , they did not consider dust formation in agb stars , grain growth , shattering and coagulation . in this work , we construct a dust evolution model taking into account the dust formation by sne ii and agb stars , dust destruction by sn shocks , grain growth , and shattering and coagulation , to investigate
what kind of dust processes determine the grain size distribution at each stage of galaxy evolution . in our model
, we do not consider mass exchange among various ism phases in detail ( e.g. , * ? ? ?
* ) , but our results contain the contributions of dust processes in the two ism phases , wnm ( @xmath24 k , @xmath25 ) and cnm ( @xmath26 k , @xmath27 ) by assuming these mass fractions in the ism to be constant .
this paper is organized as follows : in section [ sec : model ] we introduce the dust evolution model based on chemical evolution of galaxies . in section [ sec : result ] we examine the contribution of each dust process to the grain size distribution . section [ sec : discussion ] is devoted to the discussion on what kind of dust processes regulate the grain size distribution in galaxies .
we conclude this work in section [ sec : conclusion ] . throughout this paper the solar metallicity
is set to be @xmath28 @xcite .
in this section , we introduce our dust evolution model in a galaxy .
first , we show the basic equations of the chemical evolution model .
we then describe the dust evolutions based on the chemical evolution model , involving dust production by sne ii and agb stars , dust destruction by sn shocks , grain growth , and shattering and coagulation by grain grain collisions .
some grain processing mechanisms work in a different way in a different ism phase @xcite . in this work , while we use a one - zone model to examine the representative properties of a galaxy , we consider the effects of the dust processes in wnm and cnm by introducing the mass fractions of wnm and cnm , @xmath29 and @xmath30 . considering temperatures less than @xmath31 k , we find that an equilibrium state of two thermally stable phases ( wnm and cnm ) is established in the ism ( e.g. , * ? ? ?
thus , we calculate dust evolution taking into account a two - phase neutral ism .
we also assume that the galaxy is a closed - box ; that is , the total baryon mass @xmath32 ( the sum of the stellar mass and the ism mass in the galaxy ) is constant . since @xmath32 is just a scale factor in our work , the total dust mass just scales with @xmath32 . throughout this paper @xmath32
is set to be @xmath33 .
inflow and outflow are not considered in our model for simplicity .
since inflowing gas is considered to be not only metal poor but also dust poor , the abundance of both metals and dust are diluted with the same ( or similar ) fraction by inflow .
this effect is degenerate with a slower chemical enrichment , under a longer @xmath34 , where @xmath35 is the star formation timescale . as for outflow , since ism components ( namely gas , metals , and dust ) are blown out of a galaxy , the total gas mass in a galaxy decreases . in this case ,
star formation rate decreases at earlier phase of galaxy evolution ; that is , the effect of outflow is degenerate with a shorter @xmath34 .
thus , we just absorb the effects of inflow and outflow into @xmath34 . in this subsection
, we describe our model of chemical evolution in a galaxy . from the above assumptions ,
the equations of time evolution of the total stellar mass , @xmath36 , the ism mass , @xmath37 , and the mass of a metal species x , @xmath38 , in the galaxy are expressed as @xmath39 where @xmath40 is the galaxy age , @xmath41 is the star formation rate , @xmath42 , and @xmath43 and @xmath44 are the masses of the total baryons and total metal species x released by stars in a unit time , respectively . in this paper , we consider two dust species , carbonaceous dust and silicate dust , and we adopt two key elements of dust species ( x @xmath45 c for carbonaceous dust and x @xmath45 si for silicate dust ) in calculating dust evolution ( see section [ subsec : dustmodel ] for details ) .
we adopt @xmath46 , @xmath47 , and @xmath48 as initial conditions . in our work
, we adopt the schmidt law for the sfr : @xmath49 @xcite , and the index @xmath50 is thought to be @xmath18@xmath51 observationally ( e.g. , * ? ? ?
we here adopt @xmath52 , @xmath53 where the star formation timescale @xmath34 is a constant . for comparison ,
the case with @xmath54 is also shown in appendix [ app : schmidt ] . as long as we adopt the same star formation timescale at @xmath55 ,
there is little difference between the two cases with @xmath52 and @xmath56 .
@xmath43 and @xmath44 are written as @xmath57 \phi(m)\mbox{sfr}(t){\rm d}m , \label{eq : returnmass}\\ y_{\rm x}(t ) & = & \int^{100\;{\rm m}_{\odot}}_{m_{\rm cut}(t ) } m_{\rm x}(m , z(t - \tau_m ) ) \phi(m)\mbox{sfr}(t){\rm d}m , \label{eq : returnmetal}\end{aligned}\ ] ] where @xmath58 is the stellar initial mass function , @xmath59 is the lifetime of a star with mass @xmath60 at the zero - age main sequence , @xmath61 is the metallicity ( @xmath62 ) , and @xmath63 and @xmath64 represent the mass of remnant stars ( white dwarfs , neutron stars or black holes ) and the mass of metal species x ejected by a star of mass @xmath60 and metallicity @xmath61 , respectively . for the lifetime of stars , we adopt the formula derived by @xcite , and the formula is obtained by the fitting to the stellar models of the padova group @xcite . since its metallicity dependence is weak , we always adopt the stellar lifetime for solar metallicity as a representative value .
the lower bound of the integration , @xmath65 is the mass of a star with @xmath66 .
we adopt the salpeter imf for stellar mass range @xmath67@xmath68 @xcite : @xmath69 where @xmath70 is set to be @xmath71 , and the normalization is determined by @xmath72 to check the variation of the results with @xmath70 , we examine the case with @xmath73 ( a top heavy imf ) in appendix [ app : imf ] . for @xmath73 ,
the processes in the ism occur at earlier phases of galaxy evolution than for @xmath74 , because a larger amount of dust is supplied by stars .
however , we find that the sequence of the dominant dust processes along the age does not change so the following discussions are not altered significantly by the change of @xmath70 .
thus , we only consider @xmath74 in the following discussion . to calculate eqs .
( [ eq : returnmass ] ) and ( [ eq : returnmetal ] ) , we quote the remnant and metal mass data of stars with mass @xmath60 and metallicity @xmath61 from some previous works .
we assume that the mass ranges of agb stars and sne ii are @xmath18@xmath75 and @xmath76@xmath77 , respectively , and that all stars with initial masses more than @xmath77 evolve into black holes without ejecting any gas , metals or dust @xcite .
the data for agb stars with mass @xmath18@xmath78 and metallicity @xmath79 is taken from @xcite and the data for sne ii with mass @xmath80@xmath81 and metallicity @xmath82 is from @xcite .
we interpolate and extrapolate the data for all values of mass and metallicity ( also for the dust data in sections [ subsubsec : agb ] and [ subsubsec : snii ] ) . for dust evolution
, we consider dust production by sne ii and agb stars , dust destruction by sn shocks in the ism , grain growth in the cnm , and shattering and coagulation by grain
grain collisions in the wnm and cnm . in this work ,
as mentioned in section [ sec : intro ] , we assume a two - phase ism ( wnm and cnm ) to calculate the dust evolution ( see also section [ subsec : shattering ] ) .
we neglect the contribution of type ia sne ( sne ia ) to the production of metals and dust , and the destruction of dust .
@xcite showed that sne ia release little dust into the ism . furthermore ,
dust destruction by sne ia is expected to be insignificant to the total dust budget in galaxies ( less than 1/10 of the contribution of sne ii ; @xcite ) . as for metals , although @xcite showed that the contribution of sne ia to the silicon and carbon enrichment in the ism can be comparable to that of sne ii , the ratio between sn ia rate and sn ii rate is unknown [ @xcite suggested that it is about @xmath67 taking into account a chemical evolution model ] .
thus , to simplify the discussion , we neglect the contribution from sne ia , keeping in mind a possible underproduction of metallicity .
the dust production data we adopt contain a lot of dust species ( c , si , @xmath83 , sic , fe , fes , @xmath84 , mgo , @xmath85 , @xmath86 , @xmath87 , and @xmath88 ) @xcite .
however , the physical properties of grain species other than carbonaceous and silicate grains are not fully known .
hence , we categorize all grain species other than carbonaceous dust as silicate and calculate their growth , shattering , and coagulation by adopting the physical parameters of silicate grains .
in particular , after grain growth and coagulation occur , the dust species categorized as silicate dust do not evolve separately and our simplification can avoid the complexity arising from the compound species .
in fact , the mass of dust grains ejected by sne is dominated by si grains , which would grow into silicate grains in the oxygen - rich environments such as molecular clouds . for carbonaceous dust ,
we adopt material properties of graphite .
the adopted parameters of these two grain species are shown in table [ tab : para ] and are the same as in @xcite and @xcite .
although we calculate silicate and carbonaceous dust separately , we are interested in how the overall grain size distribution is affected by each dust process . therefore , we focus on the total grain size distribution .
[ cols="<,^,^,^,^,^,^,^,^",options="header " , ] + note .
x is the key element of dust species , @xmath89 is the mass fraction of the key element x in the grains , @xmath90 is the atomic mass of x , @xmath91 is the bulk density of dust grains , @xmath92 is the shattering threshold velocity , @xmath93 is the surface energy per unit area of grains , @xmath94 is young s modulus , and @xmath95 is poisson s ratio .
+ @xmath96we assume @xmath97 for the composition of silicate @xcite .
. + @xmath99@xcite .
+ in this work , we assume that grains are spherical .
thus , the mass of a grain with radius @xmath6 is @xmath100 where @xmath91 is the bulk density of dust grains . in our model ,
we consider that the minimum and maximum radii of grains , @xmath101 and @xmath102 , are @xmath103 m and @xmath104 m , respectively .
although the minimum size of grains is poorly known , even if @xmath105 m , the evolution of both the total dust mass and the grain size distribution does not change significantly @xcite .
the size distribution of grains produced by agb stars is not well known .
@xcite suggested that the size distribution is log - normal with a peak at @xmath21 m based on the fitting to observed seds .
@xcite have recently calculated the size distribution of sic produced by c - rich agb stars by performing dust formation calculation coupled with a hydrodynamical model .
they showed that the mass distribution , @xmath106 , is close to log - normal with a peak at @xmath107@xmath108 m , where the grain size distribution @xmath14 is defined so that @xmath109 is the number density of dust grains with radii in the range @xmath110 $ ] ( the size distribution multiplied by @xmath111 means the mass distribution per logarithmic grain radius ) .
hence , both theory and observations suggest that agb stars preferentially produce large grains ( @xmath22 m ) . in this paper
, we simply assume that the mass distribution , @xmath106 , of each species produced by agb stars is log - normal with a peak at @xmath112 m with standard deviation @xmath113 , so that the shape of the mass distribution in fig . 7 in @xcite
is reproduced .
we normalize @xmath14 by @xmath114 where @xmath115 is the dust mass released by a star with mass @xmath60 .
the size distributions of all species are assumed to be the same for simplicity .
dust mass data for agb stars with mass @xmath18@xmath116 and metallicity @xmath117 is taken from @xcite .
the size distributions of dust species other than carbonaceous dust are summed to compose the grain size distribution of silicate ( the same procedure is also applied in section [ subsubsec : snii ] ) .
we define @xmath118 as the size distribution of dust species , where @xmath119 represents the key element of dust species ( x = c for carbonaceous dust and x = si for silicate dust ) .
some fraction of dust grains in galaxies are produced in the ejecta of sne ii ( e.g. , * ? ? ?
after a sn explosion , reverse shock occurs because of interactions between the ism surrounding the sn and its ejecta , and dust grains are destroyed by sputtering in the shock ( e.g. , * ? ? ?
* ; * ? ? ?
@xcite calculated the total mass and size distribution of dust grains ejected by sne ii considering the dust destruction in the radiative and non - radiative phases of sn remnants .
we adopt the data for dust mass and size distribution derived by @xcite for sne ii with mass @xmath80@xmath120 . and @xmath121 is less than @xmath122 ( see * ? ? ?
] they also considered two cases for mixing in the helium core : unmixed and mixed models .
@xcite showed that the data from the unmixed model is in better agreement with observations than that of the mixed model .
hence , we adopt the unmixed model .
@xcite showed that the size distribution of grains supplied by sne ii is biased to large ( @xmath21 m ) grains due to the destruction of small grains by the reverse shock .
the amount and size distribution of grains injected by sne ii depend on the density of the surrounding ism because the dust destruction efficiency of the reverse shock is higher in the denser ism .
however , the trend that smaller grains are more easily destroyed does not change . in this paper , the hydrogen number density of the ism surrounding the sne ii , @xmath123 , is set to be @xmath124 as a fiducial value , but the cases with @xmath125 and @xmath126 are also examined .
dust grains in the ism are destroyed or become smaller by sputtering due to the passage of interstellar shocks driven by sne .
since the destruction changes and depends on the grain size , it is important to consider dust destruction taking into account the grain size distribution . to calculate this destruction process
, we adopt the formulae in @xcite , which we show here briefly .
the number density of dust grains with radii in the range @xmath127 $ ] after the passage of a sn shock , @xmath128 , is given by @xmath129 where @xmath130 is the number fraction of grains that are eroded from the initial radii @xmath131 $ ] to radii @xmath127 $ ] by sputtering in the sn shock and has been obtained using the models by @xcite .
note that if @xmath132 , @xmath133 .
thus , the change in the number density of grains with radii @xmath110 $ ] , @xmath134 , after the passage of a single sn shock is expressed as @xmath135 accordingly , the change of mass density , @xmath136 , is @xmath137 where @xmath138 is the total dust mass with radii @xmath127 $ ] before the dust destruction .
the dust destruction efficiency @xmath139 depends on the hydrogen number density of the ism , @xmath140 , such that dust grains are destroyed more efficiently in denser regions . as mentioned in section [ subsubsec : snii ] , @xmath141 as a fiducial value in this paper .
the equation for the time evolution of @xmath142 for dust destruction by sn shocks in the ism is expressed as @xmath143 , \label{eq : sndest}\ ] ] where @xmath144 is the sn rate and @xmath145 is the ism mass swept up by a sn shock . to express the dependence on the galaxy age
, we write @xmath146 and @xmath147 as @xmath142 and @xmath148 , respectively .
the sn rate , @xmath144 , is expressed as @xmath149 where we assume that the range of integration in eq .
( [ eq : snrate ] ) is the mass range where sne occur @xcite ( if @xmath150 , @xmath151 ) .
the ism mass swept up by a sn shock , @xmath145 , depends on the density and metallicity of the ism . in our model , we adopt the following formula used in @xcite ; @xmath152^{-0.289}.\ ] ] here , we formulate the growth process of grains taking into account the grain size distribution . in the ism , particularly in dense and cold regions , metals accrete onto the surface of grains
efficiently ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
recently , various studies have shown the importance of grain growth for dust enrichment in galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
@xcite showed quantitatively that the grain size distribution has a very important consequence for grain growth . here , we follow the formulation by @xcite and consider only grain growth of refractory dust ( silicate and carbonaceous dust in this paper ) .
although volatile grains such as water ice also exists in clouds in reality , they evaporate quickly when the clouds disappear or the gas temperature rises . for grain growth , the following equation holds : @xmath153 = 0 , \label{eq : growth}\ ] ] where @xmath154 is the growth rate of the grain radius . ) is valid for the case where only grain growth is considered , i.e. , without sputtering , shattering and coagulation . ] from eq .
( [ eq : particle ] ) , @xmath155 also , from @xcite , the rate of mass increase of a grain with radius @xmath6 is expressed as @xmath156 where @xmath89 is the mass fraction of the key species x in the grains , @xmath90 is the atomic mass of x , @xmath157 is the sticking coefficient of the key species , and @xmath158 is the collision rate of x to a grain with radius @xmath6 , defined as follows @xcite @xmath159 where @xmath160 is the number density of x in the gas phase in the cnm , @xmath161 is the boltzmann constant , and @xmath162 is the gas temperature in the cnm .
we apply @xmath163 and @xmath164 for carbonaceous and silicate dust , respectively ( table @xmath18 ) and @xmath165 k. next , we consider @xmath160 , which is evaluated by @xmath166 , \label{eq : numchange}\ ] ] where @xmath167 is the average mass density of the ism in which grain growth occurs . as
grain growth occurs , the number of gaseous metals decreases .
thus , @xmath168 is a decreasing function of time if only grain growth is concerned .
the mass density is estimated as @xmath169 , where @xmath20 is the mean atomic weight , assumed to be @xmath170 ( the mass ratio of hydrogen to helium is @xmath171 ) .
in addition , @xmath172 and @xmath173 are the mass of a hydrogen atom and the hydrogen number density in the cnm , respectively , and we apply @xmath174 .
hence , from the above four equations [ eq .
( [ eq : sizediff])([eq : numchange ] ) ] , we obtain @xmath175 .
\label{eq : radiusgrowth}\ ] ] we assume @xmath176 for simplicity , which means that when the key species collide with a dust grain , it definitely sticks . in our study , we calculate the grain growth using eqs .
( [ eq : growth ] ) and ( [ eq : radiusgrowth ] ) .
turbulence occurs in the ism ubiquitously , and it is confirmed that turbulence is maintained by thermal conduction from simulations ( e.g. , * ? ? ?
in a turbulent medium , dust grains are accelerated by turbulence ( e.g. , * ? ? ?
* ) , and they collide with each other and shattering can occur ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
@xcite suggested that the grain size distribution in the ism changes significantly by shattering due to collisions between dust grains accelerated by magnetohydrodynamic turbulence @xcite . in our model , to calculate shattering process , we adopt the grain velocity calculated by @xcite , and the shattering equation and parameters used by @xcite , whose formulation is based on @xcite .
we outline the treatment of shattering .
we define @xmath177 as the mass of grains with radii [ @xmath178 in a unit volume ( refer to as `` mass density '' in this paper ) . considering shattering in the collision between two grains with radii @xmath179 and @xmath180 ( called grains @xmath18 and @xmath51 , respectively ) , the time evolution of @xmath181 for shattering is expressed as @xmath182_{\rm shat}\hspace{-6 mm } = & -&m(a)\rho_{\rm x}(a , t){\rm d}a\!\!\int^{a_{\rm max}}_{a_{\rm
min}}\hspace{-3mm}\alpha[m(a),m(a_1)]\rho_{\rm x}(a_{1},t){\rm d}a_{1 } { \rm d}a_{1}\!\\ & + & \!\int^{a_{\rm max}}_{a_{\rm min}}\!\!\!\int^{a_{\rm max}}_{a_{\rm min}}\hspace{-3mm}\alpha[m(a_{1}),m(a_{2})]\rho_{\rm x}(a_{1},t){\rm d}a_{1}\rho_{\rm x}(a_{2},t){\rm d}a_{2 } m^{1,2}_{\rm shat}(a ) { \rm d}a_{1}{\rm d}a_{2 } , \label{eq : shattering}\end{aligned}\ ] ] and @xmath183 = \left\ { \begin{array}{ll } { \displaystyle 0 } & ( v_{1,2 } \le v_{\rm shat})\\ \frac{\displaystyle \sigma_{1,2}v_{1,2}}{\displaystyle m(a_{1})m(a_{2 } ) } & ( v_{1,2 } > v_{\rm shat } ) , \end{array } \right .
\label{eq : shatteredrate}\end{aligned}\ ] ] where @xmath184 is the total mass of shattered fragments of grain @xmath18 within size bin @xmath110 $ ] by a collision between grains @xmath18 and @xmath51 , and depends on the relative velocity of the grains .
the size distribution of shattered fragments is proportional to @xmath185 ( e.g. , * ? ? ?
the cross section of the collision between grains @xmath18 and @xmath51 is assumed to be @xmath186 $ ] . the shattering threshold , @xmath187 , is assumed to be @xmath188 , and @xmath189 for carbonaceous dust and silicate dust , respectively @xcite .
we adopt the same treatment for the relative velocity as @xcite and @xcite : each time step is divided into four small time steps , and we consider shattering under the following four relative velocities in each small time step ( i ) front collision @xmath190 , ( ii ) back - end collision @xmath191 , ( iii ) side collision @xmath192 , and ( iv ) @xmath193 . here ,
@xmath194 and @xmath195 are the velocities of the grain with radius @xmath179 and @xmath196 , respectively .
shattering can occur not only in turbulence but also in sn shocks ( e.g. , * ? ? ?
however , both of these shattering mechanisms have similar consequences on the grain size distribution , so it is difficult to separate them . to compare our work with previous studies @xcite
, we only consider shattering in turbulence . in low temperature and high density regions of the ism
, it is expected that coagulation by grain
grain collisions occurs .
indeed , @xcite observed dense filaments and showed that the ratio of the intensity in the filaments , @xmath197 , is smaller than that in the diffuse ism .
they concluded that this trend resulted from the decrease of small grains due to coagulation .
for coagulation , we adopt the formulation , the velocity of grains , and the parameters used by @xcite .
the time evolution of @xmath181 for coagulation is expressed as follows @xmath198_{\rm coag}\hspace{-6mm}=&-&m(a)\rho_{\rm x}(a , t){\rm
d}a\!\!\int^{a_{\rm max}}_{a_{\rm min}}\hspace{-3mm}\alpha[m(a_{1}),m(a)]\rho_{\rm x}(a_{1},t){\rm d}a_{1}{\rm d}a_{1}\!\\ & + & \!\int^{a_{\rm max}}_{a_{\rm min}}\!\!\!\!\int^{a_{\rm max}}_{a_{\rm min}}\hspace{-3mm}\alpha[m(a_{1}),m(a_{2})]\rho_{\rm x}(a_{1},t){\rm d}a_{1}\rho_{\rm x}(a_{2},t){\rm d}a_{2}m^{1,2}_{\rm coag}(a){\rm d}a_{1}{\rm d}a_{2},\end{aligned}\ ] ] and @xmath183 = \left\ { \begin{array}{ll } { \displaystyle 0 } & ( v_{1,2 } \ge v^{1,2}_{\rm coag})\\ \frac{\displaystyle \beta\sigma_{1,2}v_{1,2}}{\displaystyle m(a_{1})m(a_{2 } ) } & ( v_{1,2 } < v^{1,2}_{\rm coag } ) , \end{array } \right .
\label{eq : coagthreshold}\end{aligned}\ ] ] where @xmath199 is the sticking coefficient of dust grains , and @xmath200 if the mass range of @xmath201 is within @xmath202 $ ] ; otherwise @xmath203 .
we assume that coagulation occurs if the relative velocity is less than the coagulation threshold @xmath204 . in our model
, it is calculated in the same way as @xcite , @xmath205^{1/2 } \frac{\gamma^{5/6}}{e^{1/3 } r^{5/6}_{1,2 } s^{1/2 } } , \label{eq : coagthrevel}\ ] ] where @xmath206 , @xmath93 is the surface energy per unit area , and @xmath207 $ ] , where @xmath208 and @xmath209 are poisson s ratio and young s modulus of grain @xmath18 .
the parameters we used are shown in table [ tab : para ] . here , we assume @xmath210 for simplicity .
the treatment of the relative velocity is the same as for shattering . here
, using the dust processes introduced above , we show the equation for the dust mass evolution in a galaxy at each grain radius bin , so that we can finally obtain the evolution of grain size distribution . defining @xmath211 as the mass density of grains with radii @xmath212 $ ] '' to emphasize that it stands not for infinitesimal but a certain small amount . ] and a galactic age @xmath40 , it is formulated as @xmath213\\ \nonumber & & \hspace{-9pt } + \eta_{\rm cnm } \left[m(a ) \delta a\frac{\partial [ f(a , t)]}{\partial t}\right]\\ \nonumber & & \hspace{-9pt } + \eta_{\rm wnm}\left[\frac{{\rm d}\delta m_{\rm d}(a , t)}{{\rm d}t}\right]_{\rm shat , wnm } + \eta_{\rm cnm}\left[\frac{{\rm d}\delta m_{\rm d}(a , t)}{{\rm
d}t}\right]_{\rm shat , cnm}\\ & & \hspace{-9pt } + \eta_{\rm wnm}\left[\frac{{\rm d}\delta m_{\rm d}(a , t)}{{\rm d}t}\right]_{\rm coag , wnm } + \eta_{\rm cnm}\left[\frac{{\rm d}\delta m_{\rm d}(a , t)}{{\rm d}t}\right]_{\rm coag , cnm } , \label{eq : dustevo}\end{aligned}\ ] ] where @xmath29 and @xmath30 are the mass fraction of wnm and cnm in the ism , respectively . from top to bottom , the terms on the right hand side describe reduction of dust due to astration , ejection of dust from stellar sources , dust destruction by sn shocks , grain growth in the cnm , shattering in the wnm and cnm , and coagulation in the wnm and cnm . to calculate the dust processes which occur in each ism phase , as mentioned in section [ sec : intro ] ,
we assume ( 1 ) that @xmath29 and @xmath30 are constants and ( 2 ) that there are two stable phases , wnm and cnm , in the ism ( namely , the sum of @xmath29 and @xmath30 is unity ) .
the total mass of grains with radii @xmath214 $ ] ejected by stars per unit time , @xmath215 , is expressed as @xmath216 where @xmath217 is the total mass of grains with radii @xmath214 $ ] released by stars with mass @xmath60 and metallicity @xmath61 .
in this paper , as mentioned above , we consider the effects of dust formation by sne ii and agb stars , dust destruction by sn shocks in the ism , grain growth , shattering , and coagulation on the evolution of grain size distribution in galaxies . among these processes ,
dust formation by sne ii and agb stars , dust destruction and grain growth directly increase or decrease the total dust mass , while shattering and coagulation modify only the grain size distribution .
the evolution of the total dust mass in galaxies is often modeled by taking into account the former four contributions ( dust formation by sne ii and agb stars , dust destruction , and grain growth ) ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
they calculated the dust evolution by assuming a representative grain size , but the dust destruction and grain growth depend on the grain size distribution .
thus , it is unknown whether these four contributions can reproduce the grain size distribution in nearby galaxies even though they can explain the evolution of the total dust mass . in section [ subsec : w / o collision ]
, we first investigate the contributions of the processes that directly affect the total dust mass , and then in section [ subsec : w/ collision ] , we examine the effects of shattering and coagulation .
first , we consider the stellar processes including dust ejection from stars ( sne ii and agb stars ) and dust reduction via astration .
figure [ fig : staronly ] shows the result .
the size distribution is expressed by multiplying @xmath111 to show the mass distribution in logarithmic grain radius bin .
we adopt @xmath218 gyr and @xmath141 .
we also show the cases with sne ii only . as mentioned in section [ sec : model ] , since @xmath32 is just a scale factor , the shape of the size distribution does not change with @xmath32 . , and @xmath219 gyr , respectively , with @xmath218 gyr and @xmath220 .
dashed lines are cases with dust production by sne ii only and dust reduction through astration , the same color corresponding to the same age .
note that the red dashed line overlaps with the red solid line.,scaledwidth=45.0% ] from fig .
[ fig : staronly ] , throughout any galactic age , we can observe that the grain size distribution has a peak at around @xmath221 m , and that only a small amount of grains with @xmath222 m can be formed by stars . in fig .
[ fig : star_chansn ] , we show the grain size distribution for other values of @xmath140 : @xmath223 in the left panel and @xmath224 in the right panel , respectively . from figure
[ fig : star_chansn ] , we find that a larger amount of dust grains with radii less than @xmath225 m are destroyed by reverse shocks in the case of higher @xmath140 , and a smaller amount of dust is supplied to the ism .
however , even if @xmath140 changes , the trend that a small amount of dust grains with radii less than @xmath226 m are supplied to the ism does not change .
thus , stars are the sources of dust grains with large radii ( @xmath227 m ) . from figs .
[ fig : staronly ] and [ fig : star_chansn ] , we observe that dust from sne ii always dominates the grain size distribution , while the contribution of dust from agb stars is seen only around @xmath112 m at a galactic age @xmath228 gyr . from our calculation , for the case with @xmath141 , the dust mass ratios produced by agb stars and sne ii are @xmath229 , @xmath230 , and @xmath231 at @xmath232 and @xmath219 gyr , respectively . on the other hand , @xcite suggested that the contribution of agb stars to the total dust mass in galaxies approaches or exceeds that of sne ii at @xmath233 gyr .
this difference mainly results from the dust mass data adopted .
we adopt the data of @xcite , whereas @xcite adopted those of @xcite .
the dust mass of @xcite is larger than that of @xcite because of the difference in the treatment of the dust condensation and the destruction by reverse shocks .
however , even if the contribution of agb stars is larger , the result that only a small amount of grains with @xmath234 m are produced by stars does not change . from the right panel in fig .
[ fig : star_chansn ] , we find that the contribution of dust from agb stars is relatively large for @xmath235 at @xmath228 gyr . at @xmath228 gyr ,
the dust mass ratios produced by agb stars and sne ii are @xmath230 , @xmath231 , and @xmath236 for the cases with @xmath237 and @xmath126 , respectively .
this is because a larger amount of dust grains condensed in the ejecta of sne ii are destroyed by reverse shocks for higher @xmath140 .
( dotted lines ) .
the values of @xmath34 and @xmath140 are the same as in fig .
[ fig : staronly ] .
solid lines are the same as in fig .
[ fig : staronly ] , the same color corresponding to the same age .
note that the red dotted line overlaps with the red solid line.,scaledwidth=45.0% ] in fig .
[ fig : sndest ] , we show the evolution of the grain size distribution taking into account dust destruction by sn shocks in the ism in addition to the dust production by sne ii and agb stars . we also present the cases without the dust destruction ( i.e. , the same as the solid lines of fig .
[ fig : staronly ] ) .
the values of @xmath34 and @xmath140 are set to the same values as in fig .
[ fig : staronly ] . at @xmath238 gyr ,
the grain size distributions with and without the dust destruction by sn shocks in the ism are very similar to each other .
now we estimate the dust destruction timescale .
first , we introduce the sweeping timescale , @xmath239 , at which sn shocks sweep the whole ism , as @xmath240 from eq .
( [ eq : snrate ] ) , if @xmath241 is defined as @xmath242 eq .
( [ eq : snrate ] ) , with eq .
( [ eq : sfr ] ) , can be approximated as @xmath243 where @xmath241 is about @xmath244 from our calculation .
thus , if @xmath141 , @xmath245@xmath246 .
next , we approximate the dust destruction rate by introducing the dust destruction timescale , @xmath247 , as @xmath248 the right hand side of eq .
( [ eq : sndest ] ) can be approximated as @xmath249 where @xmath250 is a typical value of @xmath251 ; then , eq . ( [ eq : sndest ] ) reduces to @xmath252 since the overall efficiency of dust destruction , @xmath253 , is @xmath254 for @xmath255 @xcite , we obtain @xmath256 .
thus , the difference between the cases with and without dust destruction can not be seen at @xmath257 gyr in fig .
[ fig : sndest ] , where @xmath258 gyr .
we find that dust grains with @xmath259 m are destroyed effectively at @xmath219 gyr @xmath260 ( compare the solid and dotted lines in fig . [
fig : sndest ] ) .
since the decreasing rate of grain radius by sputtering does not depends on the grain radius , smaller grains are effectively destroyed in sn shocks , and the amount of smaller grains decreases @xcite . in fig .
[ fig : channsn ] , we show the cases with @xmath125 and @xmath261 . comparing figs .
[ fig : sndest ] and [ fig : channsn ] , we find that a larger amount of dust grains are destroyed for higher @xmath140
. the destruction effect is more pronounced at small sizes .
indeed , we observe that grains with @xmath262 m are effectively destroyed in the case with @xmath263 .
nevertheless , the result that smaller grains are effectively destroyed does not change , and we find that dust grains with radii larger than @xmath264 m mainly dominate the total dust amount in galaxies .
consequently , if the dust destruction by sputtering in sn shocks is dominant , only large ( @xmath22 m ) grains can survive in the ism .
figure [ fig : eta0.5 ] shows the evolution of the grain size distribution taking into account the dust production from stellar sources , dust destruction , and grain growth .
we adopt @xmath218 gyr , @xmath265 , and the mass fraction for the cnm , @xmath266 .
gyr and @xmath141 ( long - dashed lines ) .
we adopt @xmath266 .
dotted lines represent cases without grain growth [ the same as in fig .
( [ fig : sndest ] ) ] , the same color corresponding to the same age .
note that the red dotted line overlaps with the red long - dashed line .
green solid line represents the slope of the power - law grain size distribution with index @xmath267 [ @xmath268 @xcite ] which is thought to be the grain size distribution in the milky way . , scaledwidth=45.0% ] from fig .
[ fig : eta0.5 ] , we observe that while the grain size distributions with and without grain growth are almost the same at ages @xmath67 gyr and @xmath18 gyr , the difference is clear at @xmath219 gyr .
the effect of grain growth is prominent around @xmath269 m at @xmath219 gyr , since the total surface area of grains is dominated by grains with @xmath270 m .
the timescale of grain growth is discussed in detail in section [ sec : discussion ] . in fig . [
fig : eta0.5 ] , we also plot the slope of the mrn size distribution , @xmath271 @xcite , which is thought to be the grain size distribution in the milky way . from fig .
[ fig : eta0.5 ] , it is clear that the small grains with @xmath272 m are too few to reproduce the mrn size distribution .
however , the existence of the @xmath273 m excess is considered to be a proof of the existence of small grains @xcite .
furthermore , @xcite argued by using their infrared sed model that small grains are necessary to reproduce the near mid infrared seds of star forming galaxies .
consequently , when we consider the case in which dust production by sne ii and agb stars , dust destruction , and grain growth take place , the grain size distribution is always dominated by large grains , and we need to consider other processes to produce small grains efficiently . in the above we have investigated the dust processes which directly affect the evolution of the total dust mass in galaxies : dust production by agb stars and sne ii , dust destruction by sn shocks , and grain growth . as shown above , these processes can not produce small grains @xmath274 efficiently .
therefore , we now consider the contributions of the grain grain collisions , shattering and coagulation in turbulence , to the grain size distribution .
if these processes occur , although the total dust mass in galaxies does not change , the grain size distribution does . in the left panel of fig .
[ fig : shattering ] , we show the evolution of the grain size distribution in the galaxy with and without shattering ( all other dust processes in section [ subsec : w / o collision ] are included ) .
the right panel of fig .
[ fig : shattering ] shows the time evolution of dust - to - gas mass ratio ( @xmath275 ) for the cases with and without shattering , respectively .
we also plot the case without grain growth and shattering and the evolution of metallicity in the same panel . the parameters @xmath34 and @xmath140 are set to be @xmath3 gyr and @xmath124 , respectively .
we adopt @xmath276 . from the left panel of fig .
[ fig : shattering ] , at the early stage of galaxy evolution ( @xmath67 gyr ) the size distributions with and without shattering are similar with only a little difference at small sizes . at @xmath18 gyr , we observe that the size distribution has a bump at @xmath277 m in the case with shattering . as time passes , the amount of large grains decreases , and as a result the size distribution is dominated by small grains .
this behavior is substantially different from that of the case without shattering .
we now discuss this behavior in more detail . as shown in eq .
( [ eq : shattering ] ) , the efficiency of shattering is larger for larger amount of grains @xcite . at @xmath67 gyr ,
the efficiency of shattering is low because of the small dust abundance . as a result
, there is only a small difference between the cases with and without shattering . at @xmath278 gyr , since shattering occurs efficiently due to the increased amount of large grains , the amount of small grains increases . at the same time , we observe that the grain size distribution has little difference between the cases with and without shattering at @xmath279 m in the left panel of fig . [
fig : shattering ] .
this is because shattering of a tiny fraction of large grains can produce a large amount of small grains @xcite .
furthermore , since the number of small grains increases , the small grains dominate the total grain surface area .
consequently , grain growth occurs at the smallest grain sizes ( @xmath280 m ) , forming a bump at @xmath281@xmath282 m . at @xmath228 gyr , since the number of small grains increases , large grains are shattered more efficiently by the frequent collisions with the small grains .
consequently , comparing the grain size distribution at @xmath219 gyr with that at @xmath18 gyr , the amount of large grains decreases significantly .
furthermore , because of grain growth , the bump is shifted to a larger size at @xmath219 gyr than at @xmath18 gyr , and finally the size distribution has a large bump at @xmath283 m at @xmath219 gyr .
focusing on the grain size distribution at @xmath219 gyr , we find that if shattering occurs , the amount of grains with @xmath284 m is more than two orders of magnitude smaller than that of grains with @xmath285 m .
thus , the maximum size of grains in diffuse ism is determined not by stardust but by the process of shattering . in the right panel of fig .
[ fig : shattering ] , we find that grain growth starts to increase the total dust mass at around @xmath278 gyr as seen in the rapid increase of dust - to - gas mass ratio , and grain growth becomes more rapid in the case with shattering than in the case without shattering because of the increased number of small grains . as discussed in @xcite
, shattering contributes not only to the evolution of the grain size distribution but also to the total dust mass in galaxies indirectly through the enhanced grain growth .
thus , shattering is a very important process in understanding the evolution of the size distribution and the amount of dust grains in the ism . in fig .
[ fig : coagulation ] , we show the evolution of the grain size distribution with and without coagulation ( all the other dust processes are included ) in the left panel , and the time evolutions of dust - to - gas mass ratio with and without coagulation and of metallicity in the right panel .
the parameters adopted are the same as in fig .
[ fig : shattering ] . from the left panel of fig .
[ fig : coagulation ] , we find that there is little difference between the cases with and without coagulation at @xmath67 and @xmath286 gyr . since larger grains are coupled with the larger - scale turbulence , they can obtain larger velocity dispersions . thus , coagulation mainly occurs by collisions between small grains whose velocity dispersions are smaller than the coagulation threshold [ eq . ( [ eq : coagthrevel ] ) ] .
however , since the abundance of small grains is low , the contribution of coagulation is not seen at @xmath67 and @xmath18 gyr before shattering becomes effective .
after that , a large abundance of small grains are produced by shattering so coagulation becomes effective .
consequently , the bump at @xmath283 m shifts to a larger size by coagulation . from the right panel of fig . [ fig : coagulation ] , we find that the evolution of the total dust mass does not change significantly by coagulation , confirming the result obtained by @xcite .
if coagulation occurs , the number of small grains decreases ; as a result , the surface - to - volume ratio of grains decreases .
this effect may suppress the increase in dust mass due to grain growth .
however , since grain growth becomes inefficient to the dust evolution before coagulation becomes efficient ( the details are shown in section [ sec : discussion ] ) , the contribution of coagulation can not be observed for the total dust mass evolution .
consequently , the effect of coagulation on the evolution of the total dust mass in galaxies is negligible . from the left panel of fig .
[ fig : coagulation ] , we find that the amount of grains with @xmath287 m does not change significantly by coagulation because coagulation can not occur by collision between large grains which have larger velocity dispersions than the coagulation threshold .
thus , although the bump of the grain size distribution is shifted to a larger size by coagulation , coagulation does not affect the maximum size determined by shattering ( section [ subsec : resultshat ] ) .
shattering and coagulation occur differently in both ism phases ( wnm and cnm ) . here , by adopting @xmath288 and @xmath289 , we show the effect of ism phases . in fig .
[ fig : paradep ] , we show the evolution of the grain size distribution with @xmath288 ( left panel ) and @xmath289 ( right panel ) .
the case with @xmath290 is shown for comparison . at @xmath257 gyr ,
the grain size distributions are almost the same in all the cases , since the dust process is dominated by the production by stellar sources . at @xmath278 gyr ,
the amount of dust grains with @xmath291 m is larger for the case with larger @xmath30 because grain growth is more efficient . at @xmath219 gyr ,
the difference is clear at each grain size .
for larger @xmath29 , the decrements of the amount of dust grains with @xmath292 m is larger because shattering in the wnm is more efficient .
furthermore , for larger @xmath30 , the bump produced by grain growth around @xmath293 m shifts to a larger size .
thus , we understand that the amount of dust grains with @xmath284 m and the shift of the bump around @xmath293 m are dominated by shattering in wnm and coagulation in cnm , respectively . in addition , comparing the two panels in fig . 8
, we find that the dust amount at @xmath294@xmath295 m tends to be smaller for a larger @xmath30.@xmath295 m for the cases between ( @xmath296 ) = @xmath297 and @xmath298 is about @xmath299 at @xmath228 gyr .
] it means that the amount of dust grains with @xmath300 m is dominated not by shattering in wnm but shattering in cnm .
hence , the grain size distribution in galaxies is finally dominated by processes in wnm for large grains ( @xmath301 m ) and by processes in cnm for small grains ( @xmath21 m ) .
in section [ sec : result ] , we showed the evolution of the grain size distribution in galaxies for a variety of mixture of dust processes .
we found that the grain size distribution is dominated by large grains produced by stars ( sne ii and agb stars ) in the early stage of galaxy evolution , but as the time passes the number of small grains increases due to shattering , and the small grains grow to larger grains by grain growth .
after that , the size distribution shifts to larger sizes due to coagulation .
thus , we conclude that , while the grain size distribution in galaxies is controlled by stellar processes in the early stage of galaxy evolution , the main driver to change the size distribution is replaced with the processes in the ism ( shattering , coagulation , and grain growth ) at the later stage of galaxy evolution .
these processes ( shattering , coagulation , and grain growth ) have timescales dependent on the grain size distribution . in this section , by adopting representative grain radii , @xmath302 m , @xmath293 m , @xmath112 m , and @xmath303 m , we discuss the evolution of the grain size distribution more quantitatively . , and @xmath304 , the latter being obtained by considering only stellar processes [ the first and second terms in right hand size of eq .
( [ eq : dustevo ] ) ] .
panel ( a ) , ( b ) , and ( c ) are the cases with @xmath305 , and @xmath289 , respectively with @xmath218 gyr and @xmath141 . panel ( d ) and ( e ) are the case with @xmath306 gyr and @xmath1 gyr , respectively with @xmath307 .
solid , dotted , dashed , and dot - dashed lines represent the ratio of @xmath308 m , @xmath293 m , @xmath112 m , and @xmath303 m , respectively . ,
title="fig:",scaledwidth=45.0% ] , and @xmath304 , the latter being obtained by considering only stellar processes [ the first and second terms in right hand size of eq .
( [ eq : dustevo ] ) ] .
panel ( a ) , ( b ) , and ( c ) are the cases with @xmath305 , and @xmath289 , respectively with @xmath218 gyr and @xmath141 .
panel ( d ) and ( e ) are the case with @xmath306 gyr and @xmath1 gyr , respectively with @xmath307 .
solid , dotted , dashed , and dot - dashed lines represent the ratio of @xmath308 m , @xmath293 m , @xmath112 m , and @xmath303 m , respectively . ,
title="fig:",scaledwidth=45.0% ] , and @xmath304 , the latter being obtained by considering only stellar processes [ the first and second terms in right hand size of eq .
( [ eq : dustevo ] ) ] .
panel ( a ) , ( b ) , and ( c ) are the cases with @xmath305 , and @xmath289 , respectively with @xmath218 gyr and @xmath141 . panel ( d ) and ( e ) are the case with @xmath306 gyr and @xmath1 gyr , respectively with @xmath307 .
solid , dotted , dashed , and dot - dashed lines represent the ratio of @xmath308 m , @xmath293 m , @xmath112 m , and @xmath303 m , respectively .
, title="fig:",scaledwidth=45.0% ] + , and @xmath304 , the latter being obtained by considering only stellar processes [ the first and second terms in right hand size of eq .
( [ eq : dustevo ] ) ] .
panel ( a ) , ( b ) , and ( c ) are the cases with @xmath305 , and @xmath289 , respectively with @xmath218 gyr and @xmath141 . panel ( d ) and ( e ) are the case with @xmath306 gyr and @xmath1 gyr , respectively with @xmath307 .
solid , dotted , dashed , and dot - dashed lines represent the ratio of @xmath308 m , @xmath293 m , @xmath112 m , and @xmath303 m , respectively . ,
title="fig:",scaledwidth=45.0% ] , and @xmath304 , the latter being obtained by considering only stellar processes [ the first and second terms in right hand size of eq .
( [ eq : dustevo ] ) ] . panel ( a ) , ( b ) , and ( c ) are the cases with @xmath305 , and @xmath289 , respectively with @xmath218 gyr and @xmath141 .
panel ( d ) and ( e ) are the case with @xmath306 gyr and @xmath1 gyr , respectively with @xmath307 .
solid , dotted , dashed , and dot - dashed lines represent the ratio of @xmath308 m , @xmath293 m , @xmath112 m , and @xmath303 m , respectively .
, title="fig:",scaledwidth=45.0% ] figure [ fig : eachsizeevo ] shows the time evolution of the ratio between the size distribution functions , @xmath14 , and @xmath304 , the latter being obtained by considering only the stellar processes [ the first and second terms in the right hand size of eq .
( [ eq : dustevo ] ) ] .
panels ( a ) , ( b ) , and ( c ) are the cases with @xmath309 , and @xmath289 , respectively , for @xmath310 gyr and @xmath141 . from fig .
[ fig : eachsizeevo ] , we find that behavior of @xmath311 depends strongly on the grain radius . as mentioned above , this is because each dust process works at different grain radii on different timescales .
first , we discuss the evolution of the grain size in panel ( a ) . we find that @xmath311 at @xmath312 m starts to deviate from unity at the earliest galactic age among all four grain sizes . the process that causes this increase is shattering .
these small grains are produced by shattering between large grains produced by stars .
furthermore , as mentioned in section [ subsec : resultshat ] , since shattering of a small number of large grains can produce a large number of small grains , we can not see the change of @xmath311 for @xmath313 and @xmath303 m . at @xmath314 a few hundreds of myr , @xmath311 at @xmath315 m increases .
this increase is also due to shattering .
the reason why the effect of shattering appears at @xmath316 m later than at @xmath317 m is that the size distribution of shattered fragments is proportional to @xmath185 ( see section [ subsec : shattering ] ) . in other words ,
the shattered fragments become dominant at smaller sizes on shorter timescales than at larger sizes . at @xmath233 gyr , we find that the increase of @xmath311 at @xmath316 m is accelerated .
this indicates that another process becomes efficient , and it is grain growth . as seen from fig .
[ fig : shattering ] , as grain growth becomes efficient around @xmath18 gyr , the amount of grains with less than @xmath318 m increases significantly . at @xmath319 gyr
, we find that @xmath311 decrease at all sizes .
these decreases are due to coagulation for small grains ( @xmath320 and @xmath293 m ) and shattering for large grains ( @xmath313 and @xmath303 m ) . as we showed in section [ subsubsec : resultcoag ] , coagulation mainly occurs between small grains .
thus , the coagulation effect can not be seen at early phase of galaxy evolution when the abundance of small ( @xmath7 m ) grains is small .
shattering can also occur effectively if there is a large amount of small grains because of a high grain
grain collision rate with small grains [ cf .
( [ eq : shatteredrate ] ) ] .
in addition , the main reason why the decrements of grains with @xmath313 and @xmath303 m are different is shattering in different ism phases .
as shown in section [ subsec : paradep ] , grains with @xmath284 m are mainly dominated by shattering in wnm , while grains with @xmath300 m are dominated by shattering in cnm . in summary , at early phase of galaxy evolution ( @xmath321 myr ) , the size distribution is dominated by dust grains produced by stars , after @xmath322 myr , the dust processes in the ism begin to affect the size distribution at small size , and at @xmath319 gyr ( for @xmath310 gyr ) , various dust processes in the ism affect all sizes of grains .
panels ( b ) and ( c ) in fig .
[ fig : eachsizeevo ] show the cases with ( @xmath323 ) @xmath324 and ( 0.1 , 0.9 ) , respectively . compared with panel ( a ) , we find that @xmath311 at @xmath316 m does not decrease at @xmath219 gyr in panel ( b ) .
this is because the timescale of coagulation becomes longer for smaller @xmath30 . from panel ( c ) ( @xmath325 ) , we find that the decrement at @xmath326 m is smaller than those in the cases of panel ( a ) and ( b ) .
this is because the efficiency of shattering in wnm is smaller for smaller @xmath29 . however , from all the three panels , we can observe that the timing at which @xmath311 at all sizes changes due to the dust processes in the ism ( in this case , it is about @xmath51 gyr ) does not vary significantly by the change of ( @xmath327 ) for the same star formation timescale . in order to discuss the effect of @xmath34 on the size distribution , the results are shown for the same values of the parameters as in the panel ( a ) of fig .
[ fig : eachsizeevo ] , but for @xmath328 gyr in panel ( d ) and @xmath329 gyr in panel ( e ) . compared with panel ( a )
, we find that @xmath311 change at earlier stages for shorter @xmath34 at all sizes .
this is explained as follows . if @xmath34 is short , the amounts of dust and metals released by stars are large at early phases of galaxy evolution .
the timescales of shattering and coagulation are inversely proportional to the dust - to - gas mass ratio ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , and the timescale of grain growth is inversely proportional to metallicity ( e.g. , * ? ? ?
thus , for shorter @xmath34 , dust processes in the ism ( grain growth , shattering , and coagulation ) begin to affect the size distribution at earlier stages of galaxy evolution ( @xmath330 , @xmath51 and @xmath3 gyr for @xmath331 , and @xmath1 gyr , respectively ) .
the timescale of the change of @xmath311 is roughly estimated to be @xmath332 @xmath333 gyr ( appendix [ app : timescale ] ) .
we conclude that the grain size distribution in galaxies changes drastically through the galaxy evolution because different dust processes operate on the grain size distribution at different ages .
we constructed a dust evolution model taking into account the grain size distribution in a galaxy , and investigated what kind of dust processes dominate the grain size distribution at each stage of galaxy evolution . in this paper , we considered dust formation by sne ii and agb stars , dust destruction by sn shocks in the ism , grain growth in the cnm , and grain grain collisions ( shattering and coagulation ) in the wnm and cnm .
we found that the grain size distribution in galaxies is dominated by large grains produced by stars in the early stage of galaxy evolution , but as time passes the size distribution is controlled by processes in the ism ( grain growth , shattering , and coagulation ) and the age at which these ism processes enter depends on the star formation timescale , as @xmath334 gyr .
while dust production by sne ii and agb stars , dust destruction by sn shocks , and grain growth in the cnm directly affect the total dust mass evolution , we found that the grains are predominantly large ( @xmath335@xmath221 m ) and only a small amount of small grains ( @xmath222 m ) are produced by these processes .
if we take shattering and coagulation into account , the grain size distribution is modified significantly by these two processes .
in particular , shattering indirectly contribute to the large increase of the total dust mass : after small grains ( @xmath7 m ) are produced by shattering , grain growth becomes more effective because of the enhanced surface - to - volume ratio .
furthermore , grain growth produces a large bump in the grain size distribution around @xmath315 m .
the effects of shattering in wnm and cnm on the size distribution appear at different grain radii : while grains with @xmath292 m are mainly shattered in wnm , shattering in cnm affects grains with @xmath300 m .
furthermore , the effect of shattering , in particular shattering in wnm , is large enough to determine the maximum size of grains in the ism .
coagulation occurs effectively after the abundance of small grains is enhanced by shattering , and the grain size distribution is deformed to have a bump at a larger size ( @xmath9@xmath10 m at @xmath336 gyr ) by coagulation .
we conclude that the evolution of both the total dust mass and the grain size distribution in galaxies are related strongly to each other and the grain size distribution changes drastically through the galaxy evolution .
we thank the anonymous referee for many suggestions , which were useful to improve the quality and the clarity of this paper .
we are grateful to akio k. inoue , takashi kozasa , daisuke yamasawa , and asao habe for fruitful discussions , to satoshi okuzumi and hiroshi kobayashi for helpful discussions on the process of grain grain collision , to lars mattsson for his comments which improve the presentation of this paper and to jennifer m. stone for checking the english .
rsa acknowledges the hospitality of the members in institute of astronomy and astrophysics , academia sinica during his stay .
rsa has been supported from the grant - in - aid for jsps research under grant no .
23 - 5514 .
rsa and ttt have been also partially supported from the grand - in - aid for the global coe program `` quest for fundamental principles in the universe : from particles to the solar system and the cosmos '' from the ministry of education , culture , sports , science and technology ( mext ) of japan .
ttt have been supported by the grant - in - aid for the scientific research fund ( ttt : 23340046 , 24111707 ) commissioned by the mext and by the strategic young researcher overseas visits program for accelerating brain circulation commissioned by the jsps ( r2405 ) .
hh is supported by nsc grant 99 - 2112-m-001 - 006-my3 .
t.n . has been supported by world premier international research center initiative ( wpi initiative ) , mext , japan , and by the grant - in - aid for scientific research of the jsps ( 22684004 , 23224004 ) .
99 anders , e. , & grevesse , n. 1989 , gecoa , 53 , 197 asano , r. s. , takeuchi , t. t. , hirashita , h. , & inoue , a. k. 2013 , earth , planets , & space , 65 , 213 bernard , j. -p .
2008 , aj , 136 , 919 bertelli , g. , bressan , a. , chiosi , c. , fagotto , f. , & nasi , e. 1994 , a&as , 106 , 275 bianchi , s. , schneider , r. 2007 , mnras , 378 , 973 calura , f. , pipino , a. , & matteucci , f. 2008 , a&a , 479 , 669 chokshi , a. , tielens , a. g. g. m. , & hollenbach , d. 1993 , apj , 407 , 806 draine , b. t. 2009 , in henning th .
, grn e. , steinacker j. , eds , cosmic dust near and far , asp conference series , asp , san francisco , p. 453 draine , b. t. , & lee , h. m. 1984 , apj , 285 , 89 dwek , e. & scalo , j. m. 1980 , apj , 239 , 193 dwek , e. 1998 , apj , 501 , 643 evans , a. 1994 , the dusty universe , wiley , chichester fitzpatrick , e. l. , & massa , d. 2007 , apj , 663 , 320 gallerani , s. , maiolino , r. , juarez , y. , nagao , t. , marconi , a. , bianchi , s. , schneider , r. , mannucci , f. , oliva , t. , willott , c. j. , jiang , l. , & fan , x. 2010 , a&a , 523 , 85 gordon , k. d. , clayton , g. c. , misselt , k. a. , landolt , a. u. , & wolff , m. j. 2003 , apj , 594 , 279 groenewegen , m. a. t. 1997 , a&a , 317 , 503 harris , j. , & zaritsky , d. 2009 , apj , 138 , 1243 heger , a. , fryer , c. l. , woosley , s. e. , langer , n. , & hartmann , d. h. 2003 , apj , 591 , 288 hirashita , h. 1999a , apj , 510 , l99 hirashita , h. 1999b , apj , 522 , 220 hirashita , h. , & ferrara , a. 2002 , mnras , 337 , 921 hirashita , h. , nozawa , t. , kozasa , t. , ishii , t. t. , & takeuchi , t. t. 2005 , mnras , 357 , 1077 hirashita , h. , & omukai , k. 2009 .
mnras , 399 , 1795 hirashita , h. , & yan , h. 2009 , mnras , 394 , 1061 hirashita , h. 2010 , mnras , 407 , l49 hirashita , h. , nozawa , t. , yan , h. , & kozasa , t. 2010 , mnras , 404 , 1448 hirashita , h. & kuo , t - m .
2011 , mnras , 416 , 1340 hirashita , h. 2012 , mnras , 422 , 1263 ikeuchi , s. , & tomita , h. 1983 , pasj , 35 , 77 inoue , a. k. 2003 , pasj , 55 , 901 inoue , a. k. 2011 , eps , 63 , 1 jones , a. p. , tielens , a. g. g. m. , hollenbach , d. j. & mckee , c. f. 1994 , apj , 433 , 797 jones , a. p. , tielens , a. g. g. m. , hollenbach , d. j. 1996 , apj , 469 , 740 jones , a. p. , & nuth , j. a. iii . 2011 , a&a , 530 , 44 karakas , a. i. 2010 , mnras , 403 , 1413 kennicutt , r. c. 1998 , apj , 498 , 541 kim , s .-
m . , martin , p. g. , & hendry , p. d. 1994 , apj , 422 , 164 kobayashi , c. , umeda , h. , nomoto , k. , tominaga , n. , & ohkubo , t. 2006 , apj , 653 , 1145 koyama , h. , & inutsuka , s. 2002 , apj , 564 , 97 kozasa , t. , nozawa , t. , tominaga , n. , umeda , h. , maeda , k. , nomoto , k. 2009 , in henning th .
, grn e. , steinacker j. , eds , asp conf .
414 , cosmic dust near and far .
, san francisco , p. 43
, & hirashita , h. 2012 , mnras , 424 , l34 liffman , k. , & clayton , d. d. 1989 , apj , 340 , 853 luck , r. e. , moffett , t. j. , barnes , t. g. , iii , & gieren , w. p. 1998
, aj , 115 , 605 maiolino , r. , schneider , r. , oliva , e. , bianchi , s. , ferrara , a. , mannucci , f. , pedani , m. , & roca sogorb , m. 2004 , nat , 431 , 533 mathis , j. s. 1990 , ara&a , 28 , 37 mathis , j. s. , rumpl , w. , & nordsieck , k. h. 1977 , apj , 217 , 425 matsuura , m. , dwek , e. , meixner , m. , et al .
2011 , science , 333 , 1258 mckee , c. f. , & ostriker , e. c. 2007 , ara&a , 45 , 565 michaowski , m. j. , murphy , e. j. , hjorth , j. , watson , d. , gall , c. , & dunlop , j. s. 2010 , a&a , 522 , a15 nomoto , k. , iwamoto , k. , nakasato , n. , thielemann , f .- k . , brachwitz , f. , tsujimoto , t. , kubo , y. , & kishimoto , n. 1997 , nupha , 621 , 467 nozawa , t. , kozasa , t. , umeda , h. , maeda , k. , & nomoto , k. 2003 , apj , 598 , 785 nozawa , t. , kozasa , t. , & habe , a. 2006 , apj , 648 , 435 nozawa , t. , kozasa , t. , habe , a. , dwek , e. , umeda , h. , tominaga , n. , maeda , k. , & nomoto , k. 2007 , apj , 666 , 955 nozawa , t. , kozasa , t. , tominaga , n. , maeda , k. , umeda , h. , nomoto , k. , & krause , o. 2010 , apj , 713 , 356 nozawa , t. , maeda , k. , kozasa , t. , tanaka , m. , nomoto , k. , & umeda , h. 2011 , apj , 736 , 45 odonnell , j. e. , & mathis , j. s. 1997 , apj , 479 , 806 ormel , c. w. , paszun , d. , dominik , c. , & tielens , a. g. g. m. 2009 , a&a , 502 , 845 pei , y. c. 1992 , apj , 395 , 130 pipino , a. , fan , x. l. , matteucci , f. , calura , f. , silva , l. , granato , g. , & maiolino , r. 2011 , a&a , 525 , 61 raiteri , c. m. , villata , m. , & navarro , j. f. 1996 , a&a , 315 , 105 russell , s. c. , & dopita , m. a. 1992 , apj , 384 , 508 salpeter , e. e. 1955 , apj , 121 , 161 schmidt , m. 1959 , apj , 129 , 243 silvia , d. w. , smith , b. d. , & shull , j. m. 2012 , apj , 748 , 12 spitzer , l. jr .
1978 , physical processes in the interstellar medium ( wiley , new york ) p.207 stepnik , b. , et al .
2003 , a&a , 398 , 551 takagi , t. , vanseviius , v. , & arimoto , n. 2003 , pasj , 55 , 385 takeuchi , t. t. , hirashita , h. , ishii , t. t. , hunt , l. k. , & ferrara , a. 2003 , mnras , 343 , 839 takeuchi , t. t. , ishii , t. t. , nozawa , t. , kozasa , t. , & hirashita , h. 2005 , mnras , 362 , 592 todini , p. , & ferrara , a. 2001 , mnras , 325 , 726 valiante , r. , schneider , r. , bianchi , s. , andersen , a. c. 2009 , mnras , 397 , 1661 valiante , r. , schneider , r. , salvadori , s. , & bianchi , s. 2011 , mnras , 416 , 1916 weingartner , j. c. , & draine , b. t. 2001 , apj , 548 , 296 winters , j. m. , fleischer , a. j. , le bertre , t. , & sedlmayr , e. 1997 , a&a , 326 , 305 wolfire , m. g. , mckee , c. f. , hollenbach , d. , & tielens , a. g. g. m. 2003 , apj , 587 , 278 yamasawa , d. , habe , a. , kozasa , t. , nozawa , t. , hirashita , h. , umeda , h. , & nomoto , k. 2011 , apj , 735 , 44 yan , h. , lazarian , a. , & draine , b. t. 2004 , apj , 616 , 895 yasuda , y. , & kozasa , t. 2012 , apj , 745 , 159 zhukovska , s. , gail , h. p. , trieloff , m. 2008 , a&a , 479 , 453
in this appendix , we show dust evolution models with parameters different from the values adopted in the main text . in fig .
[ fig : index ] , we show star formation history ( sfh ) and the evolution of the grain size distribution with the schmidt law index @xmath337 and @xmath56 . to compute the sfh and grain size distribution by using star formation rate with the schmidt index @xmath54 , the sfr with the schmidt law index @xmath54 ( @xmath338 )
is expressed as @xmath339 where @xmath340 is a constant .
we define the value of @xmath340 so that it satisfies the following equation at @xmath55 : @xmath341 this is set to compare it with @xmath34 for @xmath52 easily .
thus , we obtain @xmath342 from fig .
[ fig : index ] , we find that the results are not significantly different between the cases with the @xmath337 and @xmath56 .
figure [ fig : imfindex ] shows the sfh and the evolution of the grain size distribution with the power - law index of the salpeter imf @xmath73 and @xmath71 ( fiducial value in this paper ) .
we observe that sfrs are almost the same , but the grain size distributions are different . if @xmath70 is small , that is , a large number of sne ii are produced , the abundance of dust and metals increase earlier than the case with large @xmath70 . as a result , the dust amount of each size of grains [ the values of @xmath343 with @xmath73 is larger than the case with @xmath74 .
the dust processes in the ism also become effective earlier because of the larger dust abundance .
however , we find that the trend of the evolution of the grain size distribution ( at early phases , stars are dominant sources of dust , as time passes , the processes in the ism become important ) does not change .
in section [ sec : discussion ] , we found that the timescale of the change of @xmath311 of all sizes of grains depends on star formation timescale , and the change are due to coagulation for small grains and shattering for large grains . since both of shattering and coagulation are collisional processes , the timescales scale with the grain abundance in the same way . in order to evaluate the dependence on the star formation timescale
, we compare the contributions of stars and shattering .
then , we consider the timescale of shattering , @xmath347 . since shattering is a collisional process , @xmath347 can be represented as @xmath348 where @xmath349 is the @xmath51nd moment of a grain size @xmath6 , @xmath350 is the relative velocity of grains , and @xmath351 is the number density of grains , which is given by @xmath352 where @xmath353 is the @xmath354rd moment of a grain size , @xmath91 is the bulk density of dust grains , @xmath355 is the hydrogen number density in the region where shattering occurs , and @xmath172 is the mass of the hydrogen atom . we assume the contribution of shattering to the amount of dust grains as @xmath356 , and comparing this equation with eq .
( [ appeq : star ] ) , we obtain the relation between shattering timescale and star formation timescale , @xmath357 in addition , by substituting eqs .
( [ appeq : shattime ] ) and ( [ appeq : numgra ] ) into eq .
( [ appeq : shatstar ] ) , we obtain @xmath358 to evaluate this value , we adopt @xmath359 , @xmath360 and @xmath361 ( wnm ) as a representative value .
also , from our calculation , @xmath362 , and @xmath363 for dust grains produced by stars .
then , we finally obtain @xmath364.\ ] ] thus , we conclude that the timescale of shattering , that is , the timescale of the change of @xmath311 , is proportional to @xmath365 . | dust in galaxies forms and evolves by various processes , and these dust processes change the grain size distribution and amount of dust in the interstellar medium ( ism ) .
we construct a dust evolution model taking into account the grain size distribution , and investigate what kind of dust processes determine the grain size distribution at each stage of galaxy evolution .
in addition to the dust production by type ii supernovae ( sne ii ) and asymptotic giant branch ( agb ) stars , we consider three processes in the ism : ( i ) dust destruction by sn shocks , ( ii ) metal accretion onto the surface of preexisting grains in the cold neutral medium ( cnm ) ( called grain growth ) , and ( iii ) grain grain collisions ( shattering and coagulation ) in the warm neutral medium ( wnm ) and cnm .
we found that the grain size distribution in galaxies is controlled by stellar sources in the early stage of galaxy evolution , and that afterwards the main processes that govern the size distribution changes to those in the ism , and this change occurs at earlier stage of galaxy evolution for a shorter star formation timescale ( for star formation time - scales @xmath0 and @xmath1 gyr , the change occurs about galactic age @xmath2 and @xmath3 gyr , respectively ) .
if we only take into account the processes which directly affect the total dust mass ( dust production by sne ii and agb stars , dust destruction by sn shocks , and grain growth ) , the grain size distribution is biased to large grains ( @xmath4@xmath5 m , where @xmath6 is the grain radius ) . therefore , shattering is crucial to produce small ( @xmath7 m ) grains .
since shattering produces a large abundance of small grains ( consequently , the surface - to - volume ratio of grains increases ) , it enhances the efficiency of grain growth , contributing to the significant increase of the total dust mass .
grain growth creates a large bump in the grain size distribution around @xmath8 m .
coagulation occurs effectively after the number of small grains is enhanced by shattering , and the grain size distribution is deformed to have a bump at @xmath9@xmath10 m at @xmath11 gyr .
we conclude that the evolutions of the total dust mass and the grain size distribution in galaxies are closely related to each other , and the grain size distribution changes considerably through the galaxy evolution because the dominant dust processes which regulate the grain size distribution change .
[ firstpage ] dust , extinction galaxies : evolution galaxies : ism ism : clouds galaxies : general stars : formation |
studying solar - like oscillations is a powerful way to probe the interiors of stars ( e.g. , * ? ? ?
* ; * ? ? ?
oscillations in main - sequence stars and subgiants have been measured using ground - based spectroscopy ( for a recent review see * ? ? ?
* ) and more recently by space missions such as _ corot _ ( e.g. , * ? ? ?
red giants oscillate at lower frequencies and so demand long and preferably uninterrupted time series to resolve their oscillations .
the first indications of solar - like oscillations in g and k giants were based on ground - based observations in radial velocity ( see @xcite and references therein ; @xcite ) and photometry @xcite , and on space - based photometry from the _ hubble space telescope _ ( _ hst _ ; @xcite ) , _ wire _
@xcite , _ microvariability and oscillations of stars _
( _ most _ ; @xcite ) , and _ solar mass ejection imager _ ( _ smei _ ; @xcite ) .
a major breakthrough came from observations over about 150 days with the _ corot _
space telescope , which produced clear detections in numerous stars of both radial and non - radial oscillations in the frequency range 10100 @xcite .
this increased the number of known pulsating g and k giants to nearly 800 , and enabled the first systematic study of oscillations in a large population of red clump stars @xcite .
note that the red clump comprises core he - burning stars and is the metal - rich counterpart to the horizontal branch . detecting oscillations in lower - luminosity red giants , which are h - shell - burning stars at the base of the giant branch , is very desirable for testing models of stellar evolution and also for constraining the star - formation rate in the local disk @xcite .
however , such detections are even more challenging than for clump stars because the amplitudes are substantially lower .
in addition , the oscillation frequencies are higher and are comparable to the orbital frequency of a satellite in low - earth orbit . these factors made it difficult for _ corot_to achieve such detections . in this letter , we present data from the asteroseismology program of the _ nasa _ _ kepler _ mission @xcite that reveal clear oscillations in a large sample of red giants , extending in luminosity from the red clump down to the bottom of the giant branch .
the observations were obtained over the first 34 days of science operations of the _ kepler _ mission ( q1 data ) .
we analyzed light curves having 29.4-minute sampling ( long - cadence mode ) for about 1500 stars listed as red giants in the _ kepler input catalog _ ( kic ; @xcite ) .
about @xmath6 of the q1 light curves showed large variations ( @xmath71% ) on long timescales , mostly due to m - giant pulsations , and are not analyzed here .
for the remaining sample , we used the light curves to measure the global oscillation parameters , most notably the frequency of maximum power ( ) and the mean large frequency separation ( ) .
we were able to find a power excess in @xmath81000 stars ( see also figure 5 in * ? ? ? * ) and measure the large separation for @xmath8700 stars . in this letter
we focus on the low - luminosity red giants ( corresponding to @xmath0 ) for which , as discussed in the introduction , _ corot _ results have not yet been reported . as discussed by @xcite ,
such stars generally have luminosities below about @xmath9 .
figure [ fig:01 ] shows the measured values of and for these stars .
the symbols show results obtained using the pipeline described by @xcite , who also described the method used to fit and subtract the background .
similar results were found using pipelines developed by four other groups @xcite , and the dotted lines in figure [ fig:01 ] show the region inside which @xmath880% of those measurements lay .
a detailed comparison of the results from the different pipelines is ongoing .
the tight correlation in figure [ fig:01 ] between and was discussed by @xcite and @xcite , and is seen here to extend to red giants with much higher .
this and other correlations between global oscillation parameters , and their connection to fundamental stellar parameters , will be discussed in future papers .
we now consider a subset of 50 of the low - luminosity red giants for more detailed analysis , chosen as having the best signal - to - noise ratios .
these stars are indicated in figure [ fig:01 ] with filled symbols .
figure [ fig02 ] ( left panel ) shows power spectra for six of them , spanning the full range of being considered .
each power spectrum shows a regular series of peaks , which is the clear signature of solar - like oscillations .
solar - like oscillations are p modes with high order and low degree , and the observed frequencies in main - sequence stars are reasonably well - described by the following relation : @xmath10 here , @xmath11 ( the radial order ) and @xmath12 ( the angular degree ) are integers .
the form of equation [ eq : asymptotic ] is motivated by theoretical calculations : for relatively unevolved stars , an asymptotic expansion @xcite shows that @xmath13 is approximately the inverse of the sound travel time across the star , while @xmath14 is sensitive to the sound speed gradient near the core and @xmath15 is sensitive to the surface layers .
we conventionally define three small frequency separations : @xmath16 is the spacing between adjacent modes with @xmath3 and @xmath4 , @xmath17 is the spacing between adjacent modes with @xmath5 and @xmath2 , and @xmath18 is the amount by which @xmath5 modes are offset from the midpoint between the @xmath3 modes on either side .
if equation [ eq : asymptotic ] holds then it follows that @xmath19 , @xmath20 and @xmath21 .
we shall use equation [ eq : asymptotic ] as a guide to the analysis of the observed frequencies , although its physical interpretation may be open to question in the case of red giants . to determine oscillation frequencies for each of the 50 stars in our subset
, we needed to locate the highest peaks in the power spectra .
we did this using conventional iterative sine - wave fitting , sometimes called pre - whitening or clean .
this involves finding the highest peak in the power spectrum , subtracting the corresponding sinusoidal variation from the time series and then re - calculating the power spectrum in an iterative process .
we retained only those peaks that lay in the frequency range @xmath22 5 and had heights greater than 4.5 times the fitted background level ( in amplitude ) .
damping of solar - like oscillations causes the observed power from each mode to be spread into multiple peaks under a lorentzian envelope .
if the length of the observations is significantly greater than the typical mode lifetime , one therefore expects several peaks to be extracted for each mode .
in fact , the modes are only barely resolved ( see section [ sec : collapsedechelle ] ) and so we expect iterative sine - wave fitting to give good results .
a powerful method to study the mode frequencies of solar - like oscillations is to plot them modulo the large separation , in a so - called chelle diagram @xcite .
we have displayed the frequencies of all 50 stars in a single chelle diagram by using the scaling technique described by @xcite , in an effort to carry out `` ensemble asteroseismology '' .
the result is shown in figure [ fig03]_a_. to make this diagram , we first divided the frequencies of each star by ( that is , scaling them to have a large separation of unity ) .
we used the chelle diagram to identify the radial modes and then fine - tuned the scaling factor ( ) by up to a few percent to align them on a single line in the chelle diagram , shown as a solid line in figure [ fig03]_a_. as discussed by ( * ? ? ? * their method 2 ) , this equates to assigning all stars the same value of @xmath15 .
the fact that all the stars can be aligned indicates this is a valid approximation .
this alignment is also clearly seen in the right - hand panel of figure [ fig02 ] , which shows power spectra plotted against scaled frequencies ( i.e. , divided by ) .
the chelle diagram in figure [ fig03]_a _ differs from those shown by @xcite in one important respect : only the horizontal axis has been scaled . in the vertical direction , we have plotted the original frequencies without scaling , which spreads out the points and allows us to look for variations with .
such a plot might be called a semi - scaled chelle diagram .
we can collapse the chelle diagram in the vertical direction in order to define the ridges more clearly .
figure [ fig03]_b _ shows this using a simple histogram of the peaks in figure [ fig03]_a _ , and the ridges are clearly visible .
rather than counting the number of peaks above a threshold , another approach is to sum the total power . to do this , we restricted each background - corrected power spectrum to the same frequency range as before ( @xmath23 ) , then divided the frequency scale by the large separation ( using the same fine - tuned value as above ) and folded the spectrum modulo unit spacing .
these folded spectra were summed for all 50 stars and then smoothed slightly ( with a boxcar of width 0.01 ) to give the result shown in figure [ fig03]_c_. the clarity of the ridges in figure [ fig03 ] allows us to make some statements about mode lifetimes . as mentioned in section [ sec : freqs ] , damping causes each mode to be spread into multiple peaks under a lorentzian envelope with a full - width at half maximum ( fwhm ) of @xmath24 , where @xmath25 is the mode lifetime .
there has been considerable discussion of the lifetimes of red - giant oscillations , particularly regarding whether the modes are sufficiently long - lived to allow their frequencies to be measured with sufficient accuracy to be useful for asteroseismology @xcite .
the _ corot _
results have clearly dismissed this worry for red clump stars @xcite and our _ kepler _ data do the same for the low - luminosity stars .
this is excellent news for asteroseismology .
figure [ fig03 ] was constructed by aligning the radial modes on the vertical solid line .
however , the resulting width of the @xmath3 ridge is not zero , which arises from at least three factors : ( 1 ) the resolution limit set by the duration of the observations ( 34d ) , which leads to peaks in the power spectrum having fwhm 0.3 ; ( 2 ) variations in with frequency in individual stars over the range being considered ( curvature in the chelle diagram ) ; and ( 3 ) broadening of power due to finite mode lifetimes ( see above ) .
the observed narrowness of the @xmath3 ridge sets upper limits on the second and third of these effects . here , as discussed above , we are particularly interested in the mode lifetimes .
the fwhm of the @xmath3 peak in figure [ fig03]_c _ , found by subtracting the background and fitting a lorentzian , is 0.033 .
this sets an upper limit on the intrinsic linewidths of these modes .
however , so far our measurement is in terms of , which varies over the sample .
the median is 11.5 , in which case the measured width equates to a lower limit on the mode lifetimes of about 10days . to investigate further , we divided the sample into four bins according to and formed the collapsed chelle diagram for each subset , as before .
we found the widths of the @xmath3 ridges in all four cases to be about the same in absolute terms ( @xmath80.4 ) , which is only slightly greater than the fwhm set by the resolution of the data ( see above ) .
we conclude that the mode lifetimes in these red giants is at least 10days , possibly much greater .
the @xmath4 ridge is closely parallel to the @xmath3 ridge , implying that is a nearly constant fraction of for the stars in our sample , with @xmath26 ( figure [ fig03]_c _ ) .
this frequency scaling indicates that , to a good approximation , the stars are homologous . we were able to measure the mean small separation for 38 stars in our sample .
figure [ fig04]_a _ shows the results in a so - called c - d diagram @xcite , which plots versus .
the dashed line in figure [ fig04]_a _ shows a linear fit , which has parameters @xmath27 the spread of points about this relation presumably reflects a spread in stellar masses , and is better seen by plotting the ratio of the two separations , as shown in figure [ fig04]_b _ ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the dotted line in figure [ fig03 ] marks the midpoint between @xmath3 modes .
interestingly , the @xmath5 ridge is centered slightly to the right of this line .
this indicates that the average small separation in these red giants is negative , whereas it is positive in the sun and the handful of other main - sequence stars for which it has been measured : @xmath28 cen a @xcite , @xmath28 cen b @xcite , @xmath29 pav @xcite , and @xmath25 cet @xcite .
a negative value for also indicates that equation [ eq : asymptotic ] is not satisfied in detail for these red giants , a point already noted for the g8 giant hr 7349 by @xcite , based on _ corot
_ observations .
the @xmath5 ridge in figure [ fig03 ] is notably broader than the others .
we do not attribute this to shorter mode lifetimes , but rather to the presence of mixed modes whose frequencies are shifted by avoided crossings .
mixed modes occur in evolved stars and have the characteristics of p modes in the envelope of the star and of g modes in the interior ( e.g. , * ? ? ?
a typical example is the star kic 5356201 ( see figure [ fig02 ] and filled blue symbols in figure [ fig03]_a _ ) .
the @xmath3 and 2 modes are quite close to their respective ridges , whereas there is much more scatter about @xmath5 .
a more extreme example is the star kic 4350501 ( see figure [ fig02 ] and filled red symbols in figure [ fig03]_a _ ) , which shows a particularly high peak at 87 , well below .
the proximity of this peak to the @xmath5 ridge indicates that it is likely to be a mixed mode .
a similar situation is seen in the f5 subgiant procyon , for which ground - based velocity observations show a narrow peak well below that lies close to the @xmath5 ridge in the chelle diagram @xcite .
the large peak height can be explained by the fact that mixed modes are expected to have longer lifetimes ( smaller linewidths ) than pure p modes because they have larger mode inertias ( e.g. , * ? ? ?
another good example is kic 5006817 , shown in figure [ fig : nice ] , where we clearly see multiple @xmath5 peaks in each order ( see also kic 9904059 = koi 145 in figure 1 of * ? ? ?
each of these clusters of peaks is reminiscent of a single mode that is heavily damped ( see section [ sec : collapsedechelle ] ) .
however , as mentioned above , mixed modes are expected to have longer lifetimes than pure p modes and hence to produce narrower not broader peaks .
indeed , this phenomenon of multiple @xmath5 peaks is probably what led previous authors to report short mode lifetimes in red giants ( see section [ sec : collapsedechelle ] ) .
guided by the theoretical work of @xcite , we instead interpret the observed properties of the @xmath5 ridge in terms of mixed modes .
@xcite discussed theoretical amplitudes of solar - like oscillations for a range of models on the red giant branch .
their models show mostly regular frequency patterns for evolved giants ( @xmath30 80 ) , but predict more complex spectra for low - luminosity giants .
this is mainly explained by the less efficient radiative damping in the cores of these stars , leading to a stronger interaction of p - mode and g - mode cavities , and by the less efficient trapping due to the smaller density contrast between the core and the envelope . as a consequence , more non - radial modes
could be observed than only those trapped in the envelope .
this is particularly expected to affect @xmath31 modes because the evanescent region just above the h - burning shell is smaller for these modes ( see also * ? ? ?
* ; * ? ? ?
our observation of a broadened @xmath5 ridge with a small cluster of modes per order supports the theoretical predictions by @xcite .
observations of frequency spectra in a sample of higher luminosity red giants would be a desirable follow - up study .
in addition to ridges corresponding to @xmath3 , 1 and 2 , we see clear evidence in figure [ fig03 ] for modes with @xmath2 .
the detection of these modes with photometry is difficult because of the very low amplitudes that result from geometric cancellation .
the only previous report of solar - like oscillations with @xmath2 from photometry ( except for the sun ) is a probable detection using _ corot _ in the g0 main - sequence star hd 49385 @xcite .
the ability to detect @xmath2 modes from _ kepler_observations of red giants is a significant bonus . in figures [ fig03]_a _ and _ b _ there are 12 peaks that fall along the ridge that we identify with @xmath2 , and these arise from 10 stars in the sample of 50
. these are presumably peaks that , due to the stochastic nature of the excitation , happen to rise above our detection threshold in these observations .
a more objective approach is to sum the power from all stars , as shown in figure [ fig03]_c _ , and we again see clear evidence for the @xmath2 ridge . how does the position of the @xmath2 ridge compare with expectations ? as mentioned in section [ sec : freqs ] , it is conventional to measure @xmath2 relative to @xmath5 , via the small separation
. however , given that the position of @xmath5 for these red giants is anomalous , it seems sensible to instead measure @xmath2 relative to @xmath3 .
we therefore suggest using a new small separation , , defined ( by analogy to ) as the amount by which the @xmath2 modes are offset from the midpoint between the @xmath3 modes on either side . if follows that @xmath32 . from figure [ fig03]_c
_ we see that @xmath33 .
if the asymptotic relation held ( equation [ eq : asymptotic ] ) , we would expect @xmath34 .
in that case the ratio between and would be 2.0 , whereas we observe a ratio of 2.2 .
these results should provide valuable tests for theoretical models .
these results , based on the analysis of only 34 days of data , show the tremendous potential of _ kepler _ long - cadence data for the study of solar - like oscillations in red giants .
as more data become available , the longer time series will allow detailed studies of stars covering the whole range of evolutionary states along the red giant branch and provide stringent observational tests for theories of stellar structure and evolution in this part of the h - r diagram .
we gratefully acknowledge the entire _ kepler _ team , whose outstanding efforts have made these results possible .
funding for this discovery mission is provided by nasa s science mission .
grants were received from the european research council ( erc / fp7/20072013 n@xmath35227224 , prosperity ) , and k.u.leuven ( goa/2008/04 ) . | we have measured solar - like oscillations in red giants using time - series photometry from the first 34 days of science operations of the _ kepler mission_. the light curves , obtained with 30-minute sampling , reveal clear oscillations in a large sample of g and k giants , extending in luminosity from the red clump down to the bottom of the giant branch .
we confirm a strong correlation between the large separation of the oscillations ( ) and the frequency of maximum power ( ) .
we focus on a sample of 50 low - luminosity stars ( @xmath0 , @xmath1 ) having high signal - to - noise ratios and showing the unambiguous signature of solar - like oscillations . these are h - shell - burning stars , whose oscillations should be valuable for testing models of stellar evolution and for constraining the star - formation rate in the local disk .
we use a new technique to compare stars on a single chelle diagram by scaling their frequencies and find well - defined ridges corresponding to radial and non - radial oscillations , including clear evidence for modes with angular degree @xmath2 . measuring the small separation between @xmath3 and @xmath4 allows us to plot the so - called c - d diagram of versus .
the small separation of @xmath5 from the midpoint of adjacent @xmath3 modes is negative , contrary to the sun and solar - type stars .
the ridge for @xmath5 is notably broadened , which we attribute to mixed modes , confirming theoretical predictions for low - luminosity giants .
overall , the results demonstrate the tremendous potential of _ kepler _ data for asteroseismology of red giants . |
the main idea of the renormalization group ( hereafter rg ) approach to field theory is to `` integrate out '' all fluctuations with momenta larger than some cutoff @xmath3 , and to take them into account by means of a modified dynamics for the remaining fluctuation modes with momenta smaller than @xmath3 .
this modified dynamics is governed by a scale - dependent effective action @xmath4 , the @xmath3-dependence of which is described by a functional differential equation , the exact rg equation .
flow equations can be used for a complete quantization of fundamental theories . if the latter have the classical action @xmath5 one imposes the initial condition @xmath6 at the ultraviolet ( uv ) cut - off scale @xmath7 , uses the exact rg equation to compute @xmath4 for all @xmath8 , and then sends @xmath9 and @xmath10 .
the defining property of a fundamental theory is that the `` continuum limit '' @xmath10 actually exists after the renormalization of finitely many parameters in the action .
interestingly , there are also perturbatively non - renormalizable theories which admit a @xmath10 limit .
the `` continuum limit '' of these theories is taken at a non - gaussian fixed point of the rg flow .
it replaces the gaussian fixed point which , at least implicitly , underlies the construction of perturbatively renormalizable theories .
recently , lauscher and reuter@xmath11 have constructed a new exact rg equation for the effective average action of euclidean quantum gravity .
they have found both a gaussian and a non - gaussian fixed point .
a strong evidence has been therefore obtained in favour of 4-dimensional einstein gravity being asymptotically safe.@xmath12 here , the key question we study is _ how to achieve a modification of the standard adm lagrangian where @xmath0 and @xmath1 are dynamical variables according to a prescribed renormalized trajectory_. moreover , we are interested in the cosmological applications of such a framework.@xmath13
we here use the adm hamiltonian treatment of space - time geometry .
thus , the identity @xmath14 where @xmath15 , suggests using the leibniz rule to express @xmath16 so that division by @xmath17 in the integrand of the einstein - hilbert action yields the lagrangian ( the @xmath18 factor in the numerator is set to @xmath19 with our choice of units ) @xmath20d^{3}x , \label{(3)}\ ] ] after adding to the action functional the boundary term @xmath21 , and assuming that @xmath22 is a closed manifold .
the lagrangian ( 3 ) , however , suffers from a serious drawback , because the resulting momentum conjugate to the three - metric reads as @xmath23 this would yield , in turn , a hamiltonian containing a term quadratic in @xmath24 ( since @xmath25 depends also on @xmath26 ) , despite the fact that ( 3 ) is only linear in @xmath24 when expressed in terms of first and second fundamental forms .
there is therefore a worrying lack of equivalence between @xmath25 and @xmath27 .
we thus decide to include a `` bulk '' contribution in order to cancel the effect of @xmath24 and @xmath28 in eq .
( 3 ) by writing @xmath29 as a starting action defining a gravitational theory with variable @xmath0 and @xmath1 .
more precisely , upon division by @xmath17 in the integrand of the einstein
hilbert term ( first integral in eq .
( 5 ) ) , the second and third integral in eq . ( 5 ) cancel the effect of @xmath30 and @xmath31 in eq .
( 1 ) , respectively .
the resulting lagrangian density belongs to the general family depending only on fields and their first derivatives , which is the standard assumption in local field theory .
it should also be noticed that if @xmath0 were constant , the second integral on the right - hand side of ( 5 ) would reduce to the york
gibbons hawking boundary term @xmath32 .
in other words , we are trying to generalize the standard adm lagrangian in order to consider @xmath0 as a _ dynamical field _ and investigate which dynamics is consistent with the rg approach . in this spirit
we consider hereafter the following general adm lagrangian : @xmath33d^{3}x + l_{\rm int}+l_{\rm matter } , \label{(6)}\ ] ] where the first term has the same functional form as the lagrangian of adm general relativity ( but with @xmath0 and @xmath1 promoted to dynamical variables ) , @xmath34 is an interaction term of a kinetic type which , for dimensional reasons , must be of the form ( the coefficient @xmath35 is introduced for later convenience ) @xmath36 @xmath37 being the interaction parameter .
@xmath38 is the `` matter '' lagrangian that we shall consider as given by a self - interacting scalar field .
on using the adm formalism , we take a scalar - field lagrangian @xmath39d^{3}x , \label{(8)}\ ] ] where the potential @xmath40 is , for the time being , un - determined , and @xmath41 is negative with our convention for the space - time metric .
we focus , hereafter , on cosmological models with rw symmetry . strictly speaking ,
such a name can be criticized , since we are no longer studying general relativity , nor are we simply rg - improving the einstein equations .
nevertheless , we will find that spatially homogeneous and isotropic cosmological models of the rw class can still be achieved .
in such models with lapse function @xmath42 , the full lagrangian , including scalar field , reads as ( here @xmath43 for a closed , spatially flat or open universe , respectively ) @xmath44 where hereafter we revert to dots , for simplicity , to denote derivatives with respect to @xmath45 .
thus , the resulting second - order evolution equations for @xmath46 and @xmath47 are @xmath48 & & \mu{\ddot g } -{3\over 2}\mu { { \dot g}^2\over g } + 3\mu{\dot a \over a } { \dot g } + { g \over 2 } \left(-6{{\dot a}^{2}\over a^{2 } } + 6{{\cal k } \over a^{2}}-2\lambda + 2g{d \lambda \over d g}\right)=0 , \label{eqg}\\[2 mm ] & & { \ddot \phi}+3{{\dot a}\over a}{\dot \phi } + { dv \over d\phi}=0 . \end{aligned}\ ] ] moreover , since the lagrangian ( 6 ) is independent of time derivatives of the lapse , one has the primary constraint of vanishing conjugate momentum to @xmath49 . the preservation in time of such a primary constraint yields , for our lagrangian generated from the assumption of rw symmetry , the constraint equation @xmath50 in cosmology
, the postulate of homogeneity and isotropy implies that the rg scale dependence is turned into a time dependence : @xmath51 .
the explicit law is @xmath52 , @xmath53 being a positive constant , because , when the age of the universe is @xmath45 , no fluctuation with a frequency smaller than @xmath54 can have played any role as yet .
hence the integrating - out of modes which underlies the wilson renormalization group should be stopped at @xmath55 . in the neighbourhood of a fixed point @xmath56 the evolution of the dimensionful @xmath0 and @xmath1
is approximately given by @xmath57 .
we therefore obtain the time - dependent newton parameter and cosmological term : @xmath58 unlike models where only the einstein equations are rg - improved , our framework allows for a non - trivial dynamics of the scale factor even in the absence of coupling to an external field.@xmath59 to appreciate this point , consider first the case when no scalar field exists , so that eq .
( 12 ) should not be considered .
the relation ( 14 ) suggests looking for power - law solutions of the type @xmath60 and separately consider the @xmath61 and @xmath62 case .
for @xmath61 we obtain that @xmath63 is an un - determined factor , while ( from eq .
( 13 ) ) @xmath64 which implies a power - law inflation for the `` @xmath65 '' solution , @xmath66 being larger than @xmath19 if @xmath67 , and a possible solution of the horizon problem .
note that the first equality is a relation between coupling constants which has to be satisfied , while the second simply relates the value of @xmath68 with the product @xmath69 .
since @xmath53 is not determined , @xmath66 can be made arbitrarily large . moreover ,
both @xmath70 and @xmath71 are constant , since @xmath72 if @xmath62 we find instead @xmath73 and @xmath74 where , as before , the former equation relates the values of the coupling constants , while the latter is a consistency relation .
in particular we see that , if @xmath75 , then @xmath76 must be negative .
in both cases ( i.e. @xmath62 ) , @xmath71 and @xmath70 are constant with @xmath77 next we consider the contribution of a scalar field with a self - interacting potential of the type @xmath78 , with the ansatz ( 15 ) for @xmath79 , and @xmath80 for the scalar field . the klein
gordon equation of motion ( 12 ) then yields @xmath81 and @xmath82 , which implies @xmath83 so as to have a real scalar field .
we then find , if @xmath61 , @xmath84 & & \zeta_\pm = \frac { 24 g_\star\pi[3\xi^2\lambda_\star-3\mu^{2}+5\mu \pm ( 1-\mu)\sqrt{3\mu ( 3\mu-4)-6\xi^2\lambda_\star } ] } { \xi^2(3\xi^2\lambda_\star + 2(3 - 2\mu)\mu ) } , \label{(21)}\end{aligned}\ ] ] where ( 21 ) is a consistency relation , ( 20 ) determines the value of @xmath68 , and a reality condition for @xmath85 is given by @xmath86 .
it is not difficult to see that there are physically interesting solutions with power - law inflation if @xmath37 is large enough and positive , and @xmath53 is positive .
if instead @xmath62 the only solution is for @xmath87 and we get the consistency conditions@xmath59
@xmath88 a challenging open problem is now the development of cosmological perturbation theory starting from lagrangians as in eq .
this could tell us whether formation of structure in the early universe can also be accommodated within the framework of a wilson - type@xmath89 formulation of quantum gravity .
the authors are grateful to the infn for financial support . the work of g. esposito has been partially supported by prin 2002 `` sintesi ''
. 0 o. lauscher and m. reuter , _ phys .
rev . _ * d65 * , 025013 ( 2002 ) .
s. weinberg , in _ general relativity , an einstein centenary survey _
, eds . s. w. hawking and w. israel ( cambridge university press , cambridge , 1979 ) .
a. bonanno and m. reuter , _ phys .
rev . _ * d65 * , 043508 ( 2002 ) .
a. bonanno , g. esposito and c. rubano , gr - qc 0403115 .
k. g. wilson and j. b. kogut , _ phys .
* 12 * , 75 ( 1974 ) . | models of gravity with variable @xmath0 and @xmath1 have acquired greater relevance after the recent evidence in favour of the einstein theory being non - perturbatively renormalizable in the weinberg sense .
the present paper builds a modified arnowitt
deser misner ( adm ) action functional for such models which leads to a power - law growth of the scale factor for pure gravity and for a massless @xmath2 theory in a universe with robertson walker symmetry , in agreement with the recently developed fixed - point cosmology .
interestingly , the renormalization - group flow at the fixed point is found to be compatible with a lagrangian description of the running quantities @xmath0 and @xmath1 . |
m.t . would like to thank professor daniel debra and dr .
sasha buchman for many stimulating conversations about the glisa mission concept , dr .
john w. armstrong for reading the manuscript and his valuable comments , and drs .
anthony freeman and daniel mccleese for for their constant encouragement . m.t . also acknowledges financial support through the topic research and technology development program of the jet propulsion laboratory . j.c.n.a .
acknowledges partial support from fapesp ( 2013/26258 - 4 ) and cnpq ( 308983/2013 - 0 ) .
this research was performed at the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
http://www.ligo.caltech.edu/ .
b.p . abbott _
_ , _ phys .
, _ * 116 * , 061102 ( 2016 ) .
k.s . thorne . in : _
three hundred years of gravitation _
, 330 - 458 , eds .
s.w . hawking , and w. israel ( cambridge university press : new york ) , ( 1987 ) .
http://www.virgo.infn.it/ .
schutz and m. tinto , _
_ , * 224 * , 131 ( 1987 ) .
y. grsel and m. tinto , _ phys .
d , _ * 40 * , 3884 , ( 1989 ) .
p. bender , k. danzmann , & the lisa study team , laser interferometer space antenna for the detection of gravitational waves , pre - phase a report , mpq233 ( max - planck- institt fr quantenoptik , garching ) , july 1998 .
m. tinto , j.w .
armstrong , and f.b .
estabrook , _ phys .
d , _ * 63 * , 021101(r ) ( 2000 )
. m. tinto and s.l . larson , _ phys .
d , _ * 70 * , 062002 , ( 2004 ) .
searle , p.j .
sutton , and m. tinto , _ class .
, _ * 26 * , 15 ( 2009 ) .
m. tinto and s.v .
dhurandhar , _ living rev .
relativity _ , * 17 * , 6 , ( 2014 ) .
m. tinto , j.c.n .
de araujo , o.d .
aguiar , and m.e.s .
alves , _ astroparticle physics _ , * 48 * , 50 , ( 2013 ) .
a. sesana , _ phys .
_ * 116 * , 231102 ( 2016 ) . c. cutler , and e.e .
flanagan , _ phys .
d , _ * 49 * , 2658 , ( 1994 ) . m. tinto , d. debra , s. buchman , and s. tilley , _ review of scientific instruments _ , * 86 * , 014501 , ( 2015 ) . m. tinto , j.c.n .
de araujo , h.k .
kuga , m.e.s .
alves , and o.d .
aguiar , _ class .
quantum grav .
_ , * 32 * , 185017 , ( 2015 ) .
sean t. mcwilliams proposed , independently of us , a mission concept very similar to glisa called gadfli .
both concepts were first ( independently and simultaneously ) proposed to nasa as part of its 2011 request for information ( rfi ) exercise for a lisa - like gw mission @xcite .
the derivation of the glisa sensitivity and the related discussion on the magnitude of the noises that define it have been presented in the main body and appendix of ref . @xcite .
although it was stated there that additional technology developments were needed to reduce the optical bench noise level below that of the photon - shot noise , the lisa pathfinder experiment @xcite has demonstrated this noise to be a factor of @xmath64 smaller than its anticipated value .
nasa solicitation # nnh11zda019l titled : `` concepts for the nasa gravitational - wave mission '' . http://nspires.nasaprs.com/external/ m. armano _
_ , _ phys .
lett . , _ * 116 * , 231101 , ( 2016 ) .
prince , m. tinto , s.l .
larson , and j.w .
armstrong , _ phys .
d , _ * 66 * , 122002 , ( 2002 ) . e.e .
flanagan and s.a .
hughes , _ phys . rev .
d , _ * 57 * , 4535 ( 1998 ) .
m. tinto and j.w .
armstrong , _ phys .
d , _ * 59 * , 102003 ( 1999 ) .
abbott , _ et al .
, _ * 116 * , 241103 , ( 2016 ) .
m. maggiore , `` gravitational waves .
volume 1 : theory and experiments.''(oxford university press , oxford , 2008 ) first edition , pag .
174 . to derive the redshift , @xmath32 ,
we have assumed a fiducial @xmath65cdm flat cosmology with the matter density parameter @xmath66 , the cosmological constant density parameter @xmath67 , and the value of the hubble parameter at the present time @xmath68 . to obtain the luminosity distance , @xmath69 , in terms of the chirp mass , @xmath70 ,
we have rewritten the snr as a function of the redshift by expressing @xmath33 in terms of @xmath32 ( see eq .
( 13 ) in @xcite ) .
then , for a given snr , we have solved a transcendental equation to obtain z , and from it derived the luminosity distance .
mcwilliams , http://arxiv.org/abs/1111.3708 | the geosynchronous laser interferometer space antenna ( glisa ) is a space - based gravitational wave ( gw ) mission that , for the past five years , has been under joint study at the jet propulsion laboratory , stanford university , the national institute for space research ( i.n.p.e . ,
brazil ) , and space systems loral .
if flown at the same time as the lisa mission , the two arrays will deliver a joint sensitivity that accounts for the best performance of both missions in their respective parts of the millihertz band .
this simultaneous operation will result in an optimally combined sensitivity curve that is `` white '' from about @xmath0 to @xmath1 , making the two antennas capable of detecting , with high signal - to - noise ratios ( snrs ) , coalescing black - hole binaries ( bhbs ) with masses in the range @xmath2 .
their ability of jointly tracking , with enhanced snr , signals similar to that observed by the advanced laser interferometer gravitational wave observatory ( aligo ) on september 14 , 2015 ( the gw150914 event ) will result in a larger number of observable small - mass binary black - holes and an improved precision of the parameters characterizing these sources .
together , lisa , glisa and aligo will cover , with good sensitivity , the @xmath3 frequency band .
the first direct observation of a gw signal announced by the ligo project @xcite on february 11 , 2016 @xcite , represents one of the most important achievements in experimental physics today , and marks the beginning of gw astronomy @xcite . by simultaneously measuring and recording
strain data with two interferometers at hanford ( washington ) and livingston ( louisiana ) , scientists were able to reach an extremely high level of detection confidence and infer unequivocally the gw source of the observed signal to be a coalescing binary system containing two black - holes of masses @xmath4 and @xmath5 out to a luminosity distance of @xmath6 corresponding to a red - shift @xmath7 ( the above uncertainties being at the @xmath8 percent confidence level ) . the network of the two ligo interferometers could constrain the direction to the binary system only to a broad region of the sky because the french - italian virgo detector @xcite was not operational at the time of the detection and no electromagnetic counterparts could be uniquely associated with the observed signal .
ground - based observations inherently require use of multiple detectors widely separated on earth and operating in coincidence .
this is because a network of gw interferometers operating at the same time can ( i ) very effectively discriminate a gw signal from random noise and ( ii ) provide enough information for reconstructing the parameters characterizing the wave s astrophysical source ( such as its sky - location , luminosity distance , mass(es ) , dynamical time scale , etc . )
@xcite .
space - based interferometers instead , with their six links along their three - arms , have enough data redundancy to validate their measurements and uniquely reconstruct an observed signal @xcite .
a mission such as lisa , with its three operating arms , will be able to assess its measurements noise level and statistical properties over its entire observational frequency band . by relying on a time - delay interferometric ( tdi ) measurement that is insensitive to gravitational waves @xcite
, lisa will assess its in - flight noise characteristics in the lower part of the band , i.e. at frequencies smaller than the inverse of the round - trip - light time . at higher frequencies instead , where it can synthesize three independent interferometric measurements
, it will be able to perform a data consistency test by relying on the null - stream technique @xcite , i.e. a parametric non - linear combination of the tdi measurements that achieves a pronounced minimum at a unique point in the search parameter space when a signal is present .
in addition , by taking advantage of the doppler and amplitude modulations introduced by the motion of the array around the sun on long - lived gw signals , lisa will measure the values of the parameters associated with the gw source of the observed signal @xcite .
although a single space - based array such as lisa can synthesize the equivalent of four interferometric ( tdi ) combinations ( such as the sagnac tdi combinations @xmath9 ) @xcite , its best sensitivity level is achieved only over a relatively narrow region of the mhz frequency band . at frequencies lower than the inverse of the round - trip light time
, the sensitivity of a space - based gw interferometer is determined by the level of residual acceleration noise associated with the nearly free - floating proof - masses of the onboard gravitational reference sensor and the size of the arm - length .
this is because in this region of the accessible frequency band the magnitude of a gravitational wave signal in the interferometer s data scales linearly with the arm - length . at frequencies higher than the inverse of the round - trip light time instead
, the sensitivity is primarily determined by the photon - counting statistics @xcite .
note that this noise grows linearly with the arm - length through its inverse proportionality to the square - root of the received optical power and because , in this region of the accessible frequency band , the gravitational wave signal no - longer scales with the arm - length . in summary , for a given performance of the onboard science instrument and optical configuration , the best sensitivity level and the corresponding wideness of the bandwidth over which it is reached are uniquely determined by the size of the array .
the best sensitivity level and the corresponding bandwidth over which it is reached by a space - based interferometer become particularly important when considering signals sweeping up - wards in frequency such as those emitted by coalescing binary systems containing black - holes .
as recently pointed out by sesana @xcite , it is theoretically expected that a very large ensemble of coalescing binary systems , with masses comparable to those of gw150914 , will have characteristic wave s amplitudes that could be observable by lisa over an accessible frequency region from about @xmath10 hz to about @xmath11 hz .
these frequency limits correspond to ( i ) the assumption of observing a gw150914-like signal for a period of five years ( approximately equal to its coalescing time ) , and ( ii ) the value at which the signal s amplitude is equal to the lisa sensitivity .
although one could in principle increase the size of the lisa s optical telescopes and rely on more powerful lasers so as to increase the upper frequency cut - off to enlarge the observational bandwidth , in practice pointing accuracy and stability requirements together with the finiteness of the onboard available power would result in a negligible gain . a natural way to broaden the mhz band , so as to fill the gap between the region accessible by lisa and that by the ground interferometers , is to simultaneously fly an additional interferometer with a smaller arm - length . the geosynchronous laser interferometer space antenna ( glisa )
@xcite , which has been analyzed during the past five years by a collaboration of scientists and engineers at the jet propulsion laboratory , stanford university , the national institute for space research ( i.n.p.e . ,
brazil ) , and space systems loral , would allow this @xcite . with an arm - length of about @xmath12 km , it could achieve a shot - noise limited sensitivity in the higher region of its accessible frequency band that is about a factor of @xmath13 better than that of lisa @xcite ; glisa will display a minimum of its sensitivity in a frequency region that perfectly complements those of lisa and aligo , resulting into an overall accessible gw frequency band equal to @xmath3 . to derive the expression of the joint lisa - glisa sensitivity , we first note that the noises in the tdi measurements made by the two arrays will be independent .
this means that the joint signal - to - noise ratio squared , averaged over an ensemble of signals randomly distributed over the celestial sphere and random polarization states , @xmath14 , can be written in the following form @xmath15 where the functions @xmath16 correspond to the squared optimal sensitivities ( as defined through the @xmath17 , @xmath18 , and @xmath19 modes @xcite ) of the lisa and glisa arrays respectively .
note that the factor @xmath20 in front of the integral in eq.([snrnet ] ) reflects the adopted convention throughout this article of using one - sided power spectral densities of the noises @xcite . from eq .
( [ snrnet ] ) it is straightforward to derive the following expression of the joint sensitivity squared , @xmath21 , in terms of the individual ones , @xmath22 and @xmath23 @xmath24 in fig . [ fig1 ] we plot the optimal sensitivities of lisa ( blue ) , glisa ( red ) averaged over sources randomly distributed over the sky and polarization states , together with the joint lisa - glisa ( black ) given by eq .
( [ network ] ) . for completeness , and
to visually exemplify the scientific advantages of flying simultaneously two space - based missions of different arm - lengths , we have included the anticipated sensitivity of the third - generation aligo detector ( green ) together with the amplitude of the gravitational wave signal gw150914 ( magenta ) as functions of the fourier frequency @xmath25 .
we have also included a black horizontal line showing the frequency evolution of gw150914 over a 5 years period before coalescence . shortly before the announcement made by aligo of the detection of a second signal emitted by another black - hole binary system , gw151226 , with masses roughly half of those of gw150914 @xcite
, sesana estimated that a large number of such systems @xcite could be observed by lisa while they are still spiraling around each other over periods as long as the entire five years duration of the mission . because of these compelling astrophysical reasons , to quantify the scientific advantages of flying glisa jointly with lisa we will focus our attention on signals emitted by coalescing black - hole binaries with chirp masses in the range @xmath26 .
in our analysis we will use the following expressions ( valid for circular orbits ) of the fourier transform of the amplitude of the gravitational wave signal emitted by such systems , @xmath27 , and the time , @xmath28 , it takes them to coalesce @xcite @xmath29^{5/6 } \
\frac{c}{d_l } \
f^{-7/6 } \ , \label{h } \\
t_c & = & \frac{5}{256 } \ \left[\frac{(1 + z ) g m_c}{c^3}\right]^{-5/3 } \
( \pi f_{gw})^{-8/3 } \ . \label{tc}\end{aligned}\ ] ] in eqs.([h ] , [ tc ] ) @xmath30 is the gravitational constant , @xmath31 is the speed of light , @xmath32 is the cosmological red - shift , @xmath33 is the corresponding luminosity distance , @xmath34 is the chirp mass associated with the binary system whose components have masses @xmath35 and @xmath36 , @xmath25 is the fourier frequency and @xmath37 the instantaneous frequency of the emitted gw . by using the above signal amplitude and time to coalescence , and by further assuming an integration time of @xmath38 years ( which , in the case of the gw150914 signal , defines the lower - limit of integration in the integral of the snr to be equal to @xmath39 hz ) , we have estimated the snrs achievable by the two interferometers when operating either as stand - alone or jointly .
we find lisa can observe a gw150914-like signal with a snr equal to @xmath40 , while glisa with an snr of @xmath41 because of its better sensitivity over a larger part of its observable band ( see fig .
[ fig1 ] ) .
as expected , the joint lisa - glisa network further improves upon the snr of glisa - alone by reaching a value of about @xmath42 . from these results
we can further infer that glisa will achieve a sufficiently high snr to warrant the detection of a gw150914-like signal by integrating for a shorter time .
we find that , by integrating for a period of @xmath43 days prior to the moment of coalesce of a gw150914-like system , glisa can achieve a snr of @xmath40 .
the level of snr achievable by the lisa - glisa network over a five - year integration is about @xmath44 percent higher than that of lisa alone .
this implies , from the estimated parameter precisions derived by sesana @xcite and their dependence on the value of the snr @xcite , that the lisa - glisa network will estimate the parameters associated with the gw source of the observed signal with a precision that is @xmath45 better than that obtainable by lisa alone .
it should be said , however , that this is a lower bound on the improved precision by which the parameters can be estimated since it is based only on snr considerations .
as pointed out by mcwilliams in an unpublished document @xcite , the doppler frequency shift together with the larger diurnal amplitude modulations experienced by the gw signal in the glisa tdi measurements will further improve the precision of the reconstructed parameters beyond that due to only the enhanced snr .
we will analyze and quantify this point in a follow - up article as this is beyond the scope of this letter . from the expressions of the snr and of the fourier amplitude of the gw signal , @xmath46 ( eqs .
[ snrnet ] , [ h ] ) , and by fixing the snr to a specific value for each operational configuration ( stand - alone vs. network ) , it is possible to infer the corresponding average luminosity distance to a bhb in terms of its chirp mass parameter @xcite . in fig .
[ fig2 ] we plot the results of this analysis by assuming a snr of @xmath47 .
as expected , the three derived luminosity distances are monotonically increasing functions of the chirp - mass .
for chirp - masses in the range @xmath48 glisa can see signals further away than lisa because of its better sensitivity at higher frequencies .
systems with a chirp - mass larger that @xmath49 instead can be seen by lisa at a larger luminosity distance than that achievable by glisa alone .
note that the lisa - glisa network out - performs the stand - alone configurations by as much as @xmath50 percent for bhbs with chirp - masses in the interval @xmath51 .
this results into a number of observable events that is about @xmath52 times larger than that detectable by each interferometer alone .
finally , a gw150914-like signal characterized by a chirp mass of @xmath53 can be see by lisa at an average luminosity distance of about @xmath54 , by glisa out to @xmath55 , and by the lisa - glisa system out to about @xmath56 .
in addition to derive the luminosity distance as a function of the chirp - mass ( with a chosen value of the snr ) , we have also calculated the snr as a function of the chirp - mass by assuming the snr of the lisa mission to be constant and equal to @xmath47 ( see fig .
[ fig3 ] ) .
this of course implies that the luminosity distance of the system is uniquely determined by the chirp - mass through the assumed snr value of @xmath47 for lisa .
since we are focusing on chirp - masses defined in the interval @xmath57 \ m_\odot$ ] , from fig .
[ fig2 ] we infer that the values assumed by the luminosity distance will be greater than or equal to about @xmath58 mpc .
this minimum value of the luminosity distance defines an astrophysical interesting region as it is more than four times larger than the luminosity distance to the virgo cluster of galaxies .
fig .
[ fig3 ] shows the two snrs of glisa and lisa - glisa to be decreasing functions of the chirp - mass , with their maximum values of @xmath59 and @xmath60 respectively at @xmath61 . at @xmath62 instead the glisa snr also becomes equal to @xmath47 while , as expected , the lisa - glisa snr is @xmath63 larger .
this letter has shown the scientific advantages of simultaneously flying with the lisa mission an additional , smaller size , space - based interferometer such as glisa . because of its smaller arm - length
, glisa will be able to survey the region of the gw frequency band that is in between those accessible by lisa and aligo . by covering the entire mhz and khz gw frequency band , lisa , glisa and aligo
will detect all known sources emitting in this broad frequency region , and observe signals requiring multi - band detection for understanding the physical nature of their sources . |
confinement - induced resonances ( cirs ) , originally predicted in the seminal work of olshanii @xcite and more recently observed experimentally for both bosons @xcite and fermions @xcite , have attracted a great deal of attention during the past few years .
the immediate reason is that they represent a valuable tool for the control of the atomic interactions thereby allowing to enter and probe the regime of strongly correlated bosonic or fermionic many - body systems . beyond this ,
changes of the transversal confinement potential possess direct consequences for the scattering behavior of the atoms in the waveguide such that a multitude of binary resonance profiles are accessible .
examples for the variety of scattering properties in waveguides are multi - channel confinement - induced resonances @xcite , resonant molecule - formation processes and center of mass coupling effects @xcite as well as a dual cir @xcite which leads to complete quantum suppression of scattering .
a recent experiment @xcite on cirs in transversally anisotropic waveguides has shown the necessity of an adequate theoretical approach for describing collisional processes in confined geometries of one- and two - dimensional character . a main result of this experimental work @xcite was the observation of a splitting of the cir - related loss signals in the presence of a transversal anisotropy .
two recent attempts to explain this splitting have not been successful @xcite .
the corresponding studies represent investigations of the pseudo - potential scattering with a single open transverse channel @xcite in the case of tightly confining waveguides . both the width of the cir and the full multichannel character of the problem were not taken into account . on the other hand
it has been shown that the non - separability of the center of mass and internal motion for atomic scattering in anharmonic ( transversally isotropic ) waveguide potentials changes the resonance picture qualitatively and new resonances ( anharmonicity - induced resonances ( airs ) ) are occur where molecular excited center of mass states cross the threshold @xcite .
very recently the coupling of the center of mass excitations in anharmonic isotropic and anisotropic confining potentials to the ground state has been analyzed in great detail in ab initio calculations @xcite .
there a very good agreement of the air splitting with the distance between the maxima of the atomic loss in the experiment @xcite was found .
however , it has not been analyzed what happens with the `` harmonic '' part of the investigated two - body spectrum in the trap and whether the `` harmonic '' cir agrees with previous results @xcite . in previous works
@xcite the specific ratio @xmath0 ( where @xmath1,@xmath2,@xmath3 and @xmath4 are the harmonic oscillator length , scattering length in free space , reduced atomic mass and harmonic oscillator frequency , respectively ) yielding the position of the cir was defined by the zero @xmath5 of the scattering amplitude @xmath6 in the transversal ground state in the zero - energy limit .
this point also corresponds to the absolute minimum of the transmission coefficient @xmath7 which represents an important scattering observable near the cir @xcite . in the present work
, we explore a situation where not only the ground transversal channel but also excited channels contribute to the scattering process in the waveguide and demonstrate the importance of multichannel scattering effects in the resonant region . within our description of few - channel ultracold scattering in confined geometries
we observe a mechanism which leads to the splitting of the cir under the action of a transversal anisotropy of the waveguide .
it is based on the fact that the total transmission coefficient @xmath8 , which we define as a sum of partial coefficients @xmath9 emerging from the different transverse ( ground and excited ) states labeled by @xmath10 , @xmath11 averaged over the initial populations @xmath12 , possesses its main contribution near the cir from the first excited transversal state and not from the ground state .
this follows from the observation that @xmath13 possesses a much deeper as well as broader transmission well around its minimum compared to the wells ( minima ) of @xmath14 for the excited states near the cir @xcite .
thus , even for low - energy pair collisions the scattering properties near the cir location are not determined by the partial coefficient @xmath13 , where @xmath15 , but by the behavior of the coefficients @xmath16 in excited states @xmath17 . employing this mechanism we find a splitting of the minimum of the total transmission coefficient ( 1 ) in an anisotropic trap .
the splitting is a consequence of the different dependencies of the partial coefficients @xmath18 on @xmath19 for different states in the resonant region .
a necessary prerequisite of the appearance of this splitting effect is therefore the occupation of excited transversal states in the waveguide .
we emphasize , that the present investigation is performed for harmonic traps , i.e. we investigate the influence of the anisotropy of a harmonic transversal confinement on the cirs @xcite opposite to the above - mentioned works that explore the effects of anharmonicity @xcite and anisotropy @xcite . in the following section we describe the computational approach to our multichannel scattering problem in the confined quasi-1d geometry .
the corresponding results are discussed in the third section .
the short fourth section is devoted to a discussion of the mechanism leading in the harmonic waveguide with transverse anisotropy to the splitting of the cirs .
finally we provide a brief conclusion .
to calculate the partial transmission coefficients ( the single index @xmath10 is here replaced by the double index @xmath20 indicating the quantum numbers belonging to the different transversal degrees of freedom of the waveguide ) @xmath21 describing the transmission probability from the initial transverse state @xmath22 to all possible final states @xmath23 in the course of the collision of identical bosons in a harmonic waveguide with the transverse trapping potential @xmath24 , we solve the multichannel scattering problem for the 3d hamiltonian @xmath25 depending on the relative variables @xmath26 , with the asymptotic scattering wave function for @xmath27 @xmath28 @xmath29 here the matrix elements @xmath30 of the scattering amplitude describe the transition from the initial channel with the transverse energy @xmath31 $ ] and the relative longitudinal momentum @xmath32 ( @xmath33 being the relative longitudinal collision energy and @xmath34 the total energy ) to the final open channel @xmath23 with @xmath35 .
@xmath36 are the eigenfunctions of the 2d harmonic oscillator corresponding to the eigenvalues @xmath37 .
the latter are degenerate with respect to the quantum number @xmath38 in an isotropic trap @xmath39 and @xmath40 .
the asymptotic wave function ( 4 ) is explicitly symmetric with respect to the exchange of the atoms . in order to solve the multichannel scattering problem ( 3,4 )
we extend the approach developed in @xcite for scattering in three dimensions : the expansion over the spherical harmonics on a grid ( two - dimensional discrete - variable representation ) is replaced by the expansion over the product states @xmath41 where @xmath42 are the eigenfunctions of the 1d harmonic oscillator .
the calculations are performed with the finite - range gaussian approximation @xmath43 for the interparticle interaction with @xmath44 chosen such that we have one weakly - bound state in the potential well .
we do expect that all effects shown below are to a large extent independent of the chosen interaction potential @xmath45 . as an independent check
, we have verified that the results obtained for the four open channel scattering in an isotropic waveguide @xmath39 using the screened coulomb potential @xcite are reproduced by the present approach .
note that for convergence typically 200 basis functions are used .
on @xmath19 and @xmath46 in an isotropic waveguide @xmath47.,width=321 ] from the initial ground @xmath48 into the transversally excited states @xmath49 for a few values of @xmath19.,width=321 ] let us first analyze atomic collisions in a transversely isotropic waveguide @xmath39 .
the partial transmission coefficients describing the transmission in the ground @xmath50 and in the first excited @xmath51 and @xmath52 scattering channels are shown in fig.[fig1 ] for different longitudinal collision energies @xmath46 .
obviously the position of the cir is stable with respect to variations of the energy for @xmath53 .
indeed , this position is very close to the value @xmath54 predicted in the zero - energy limit with the pseudo - potential approach @xcite and has been confirmed in subsequent numerical computations with different interatomic potentials @xcite .
hereafter we define the position of the cir in the ground state scattering channel as the minimum of the transmission coefficient @xmath55 which coincides with the zero of the imaginary part of the scattering amplitude @xmath56 in the region @xmath57 @xcite .
the minimum of the transmission coefficient @xmath58 in the first excited state ( the position of the cir in the excited scattering state ) is at a nonzero value and the corresponding transmission valley is therefore much less pronounced than in the case of @xmath59 for the ground state .
moreover , the width of the cir differs considerably for the two cases and unlike the ground state the position of the cir in the excited state is strongly dependent on the energy @xmath46 .
these facts , as we will see below , are of crucial importance for the analysis of the region near the cir .
importantly , according to ( 1 ) the total transmission coefficient is determined in a broad neighbourhood of the resonant region @xmath60 by the partial coefficients @xmath51 and @xmath52 ( if @xmath61 and @xmath62 are large enough ) @xmath63 which is due to the near - zero values of @xmath59 around its minimum ( see fig.[fig1 ] ) .
a note is in order here .
the significant occupation of transversally excited states does not necessarily arise due to a finite temperature in thermal equilibrium where the occupation probability is determined by the corresponding boltzmann distribution .
instead , the preparation of the ultracold atomic ensemble in the waveguide itself can lead to an occupation of excited states . this could e.g. be a nonadiabatic loading process into the waveguide .
the resulting nonequilibrium state would show a redistribution of its energy between the longitudinal and transversal degrees of freedom and might eventually thermalize , depending on the number of atoms and integrability aspects of the underlying system @xcite .
atomic collisions near the cir could possibly also lead to an increase of the excited state population .
indeed , the increase of the atomic loss and heating near a cir has been attributed @xcite to inelastic three - body collisions , which lead to the formation of molecules while transferring the molecular binding energy and the following energy release due to possible deexcitation of rovibrationally excited molecules to the kinetic energy of the center of mass motion of the molecule and in particular to the escaping third atom .
subsequent collisions of this third atom can lead to transverse excitations if its kinetic energy exceeds the threshold value @xmath64 .
the binding energy of the most weakly bound molecular state in the waveguide ( see fig.2 in @xcite ) exceeds already this threshold energy . using the calculated scattering amplitudes @xmath65
we have evaluated the transition probabilities @xmath66 @xcite @xmath67 where @xmath38 and @xmath68 .
the calculated probabilities @xmath69 are presented in fig.[fig3e ] for a few values of @xmath19 in the region @xmath70 i.e. between the first and second excited channel thresholds . since the probability @xmath69 approaches to the values @xmath71 one might consider collisions with the atoms accelerated due to molecules formation near cir as a possible mechanism for the emergence of considerable populations of the first excited states @xmath72 .
note also that hereafter we do not address the transmission coefficients @xmath73 with odd quantum numbers @xmath74 because these states are not coupled with the ground and excited states possessing even @xmath75 and @xmath76 . in an anisotropic waveguide @xmath77 as functions of @xmath19 calculated for a near - threshold collision energy @xmath78.,width=321 ] in the region @xmath79 of the cir for the isotropic case @xmath47 as well as for anisotropic waveguides with @xmath77 and @xmath80 calculated for @xmath81 . for @xmath82 the splitting of the minimum of the transmission coefficient can be observed . , width=283 ] in the region @xmath79 of the cir for the isotropic case @xmath47 as well as for anisotropic waveguides with @xmath77 and @xmath80 calculated for @xmath81 . for @xmath82
the splitting of the minimum of the transmission coefficient can be observed . , width=283 ] in the region @xmath79 of the cir for the isotropic case @xmath47 as well as for anisotropic waveguides with @xmath77 and @xmath80 calculated for @xmath81 .
for @xmath82 the splitting of the minimum of the transmission coefficient can be observed .
, width=283 ] in the region @xmath79 of the cir for anisotropic case @xmath83 as a function of relative population @xmath84 of the first excited manifold @xmath72 .
the splitting of the minimum of the transmission coefficient is still present even for only one percent population of the excited states.,width=321 ] we now analyze the multichannel scattering amplitude and the corresponding transmission coefficients for anisotropic waveguides for different values of the ratio @xmath82 . in fig.[fig2 ] we present the calculated partial transmission coefficients @xmath85 as a function of @xmath19 for fixed values of @xmath86 and @xmath46 . comparing this result with the isotropic case given in fig.[fig1 ]
we see that the anisotropy does not change the overall behavior of the coefficients @xmath50 and @xmath51 in the region near the cir .
however , the anisotropy splits the excited states with @xmath49 into two components and changes the @xmath52 coefficient dramatically by splitting the well of the transmission curve . as it is demonstrated in fig.[fig3 ] , the effect of the splitting of the minimum of the partial transmission coefficient @xmath87 can also be observed in the total transmission coefficient @xmath88 ( 1 ) at 5% of the relative population @xmath84 of the first exited states and persists with varying @xmath86 and @xmath46 .
fig.[fig3d ] demonstrates the dependence of the splitting on the relative population @xmath84 calculated in the region @xmath89 compatible with the experiment @xcite .
the effect of the splitting of the transmission coefficient is enhanced with increasing excited state population . and
anisotropic @xmath90 waveguides as a function of @xmath19 .
below the continuum threshold @xmath91 the calculated binding energy of the weakly - bound state is shown . between the thresholds @xmath91 and @xmath92 the calculated resonant energies @xmath93 were determined by the minimum of the transmission coefficient @xmath55 . between the @xmath91 and @xmath94
thresholds the resonant energy @xmath93 coincides with the location of the zero of @xmath95 .
bold solid curves with the full dots correspond to the isotropic case @xmath47 and the thin curves with open circles to the anisotropic case @xmath90 .
indices @xmath22 label the splitted sub - levels of the threshold energies @xmath96 of the excited states with @xmath38 . for the second excited threshold with @xmath97
only the lowest sub - level @xmath98 is shown.,width=321 ] in the computations , the colliding energy @xmath46 was chosen in agreement with the conditions of the experiment @xcite , where @xmath46 , even at maximal heating , has remained below @xmath99 , i.e. @xmath100 .
it should be noted that the pronounced shift of the minimum of the @xmath8 coefficients with varying @xmath46 is due to the considerable dependence of the partial coefficients @xmath52 and @xmath51 on @xmath33 ( see fig.[fig1 ] ) .
this effect might be responsible for the shift of the maximum of the atom loss in the experiment @xcite in the direction of increasing values for @xmath2 . using this assumption
we can fix @xmath101 as being closest to the experimental conditions @xcite by choosing from the calculated @xmath102 curves for @xmath47 the one whose position of the minimum coincides more close with the point of maximal atomic loss in the experiment .
to clarify the origin of the splitting of the partial coefficient @xmath52 we have calculated the spectrum of the near - threshold bound state and resonant states of the atomic dimer in the confining trap as a function of @xmath19 ( see fig.[fig4 ] ) .
the resonant energies @xmath103 are determined by the position of the minimum of the partial coefficient @xmath55 , where the two - body total energy @xmath34 was varied between the thresholds corresponding to the ground state @xmath104 and the second excited states @xmath105 $ ] with @xmath106 .
these resonant states were defined in @xcite as cirs with non - zero energies .
the binding energy @xmath107 of the atomic dimer in the harmonic trap with respect to the ground state threshold @xmath91 has been calculated by solving the corresponding eigenvalue problem . in analyzing the results it is important to note the different dependence of the position of the cir obtained for the different transmission curves @xmath108 in fig.[fig1 ] on the collisional energy @xmath46 .
since the resonant energy @xmath93 belonging to the cir in the ground state strongly changes with varying @xmath19 near the value @xmath0 ( see figure [ fig4 ] where a steep descent can be observed with increasing @xmath109 ) the dependence of @xmath55 on the energy @xmath46 near the minimum position is very weak ( see fig.[fig1 ] ) .
opposite to this , the dependence of the resonant energy @xmath93 for the first excited states on @xmath19 is much smoother ( see uppermost black curve with full dots in fig.[fig4 ] ) .
this leads to a considerably stronger dependence of the minimum of @xmath51 and @xmath52 on @xmath19 on @xmath33 ( see fig.[fig1 ] ) . in the anisotropic waveguide ,
the resonant curve @xmath110 of the cir of the ground state splits into two components which are shown in fig.[fig4 ] .
the energetically lower ( 0,2)-component qualitatively repeats the behavior of the resonant energy curve ( cir ) of the isotropic case .
this is why the anisotropy causes only a limited quantitative change for the @xmath50 coefficient .
however , the behavior of the energetically upper ( 2,0)-component is different ( see fig.[fig4 ] ) .
the resonance curve @xmath93 of this component is flat for the complete parameter region of @xmath19 which leads to a strong change with respect to the energy dependence of the @xmath52 coefficient , including in particular a strong change of the corresponding positions of the minima ( maxima ) with respect to @xmath109 .
the behavior of the @xmath51 coefficient is determined by the energetically lowest resonant curve @xmath93 emerging from the @xmath97 threshold in the isotropic case which is deformed only slightly by the anisotropy .
this is why the difference between @xmath111 and @xmath112 is significantly smaller when compared to the case of the @xmath52 coefficient .
the above model could potentially also explain the appearance of additional cirs with further increase of the anisotropy @xmath82 as seen in @xcite . increasing @xmath82 the energetical distance between the sub - levels characterized by @xmath113 decreases .
this leads to a considerable increase of the populations @xmath114 of higher excited states and in particular to an increase of the contribution of these states to the total transmission coefficient @xmath8 ( see eq.(1 ) ) .
the key for the understanding of the mechanism of the splitting of the cir under the action of an anisotropic harmonic trap is the diatomic spectrum of the weakly - bound and resonant states in the harmonic waveguide given in fig.6 .
it also explains why the previous considerations in the zero - energy limit near the ground state threshold did not provide any splitting of the cir defined as the singularity of @xmath115 @xcite .
actually , so far it was implicitly supposed that the resonant energy curve @xmath93 behaves linearly as it crosses the ground state threshold @xmath116 for @xmath117 and becomes subsequently a weakly - bound state for @xmath118 ( see , for example , fig.2 in @xcite ) .
therefore , it was natural to expect that in an anisotropic waveguide @xmath119 leading to the splitting of the first excited threshold @xmath120 ( see fig.6 in the present paper and fig.1(b ) in @xcite ) the resonant curve @xmath93 will also split into two components @xmath121 and @xmath122 crossing the ground state threshold at the points @xmath123 and @xmath124 with the separation @xmath125 proportional to the threshold splitting @xmath126 .
these expectations were confirmed in the experiment by measuring the distance between the maxima of the atomic loss in the anisotropic waveguide which was interpreted as @xmath125 ( see fig.3(c ) in @xcite ) .
however , our extension of the calculation of the resonant energy @xmath93 to the region @xmath118 has shown a strongly nonlinear behavior of the curve @xmath93 ( see fig.6 ) while shifting the point where the resonant curve @xmath93 crosses the ground state threshold to the value @xmath127 .
this means that at the point @xmath128 we observe a complete rearrangement of the spectrum .
the first resonant state @xmath93 becomes a new weakly - bound state once we cross this point and equivalently for the higher excited resonant states which convert into each other .
this behavior , i.e. the rearrangement of the spectrum of the two - body system at the point @xmath128 , remains in the anisotropic waveguide ( see fig.6 ) .
this is why the splitting of the singularity of the function @xmath115 was not observed near the point @xmath117 in the anisotropic harmonic waveguide @xcite but one can observe the splitting of the minimum in the effective transmission coefficient @xmath8 ( eq.(1 ) ) due to the quasi - crossing of the resonant curves @xmath122 and @xmath121 leading to the splitting of the minimum in the @xmath52 coefficient ( see fig.3 ) qualitatively equal to the width of the quasi - crossing near @xmath117 .
this width is proportional to the first excited threshold splitting @xmath126 in the presence of the anisotropy and in good agreement with the experimental value for the splitting of the maxima of the atom loss @xcite .
our investigation and following analysis demonstrates that multichannel scattering in anisotropic harmonic waveguides can lead to a splitting of the confinement - induced resonance .
a necessary ingredient is a population of at least a few percent of the transversally excited states which can certainly occur via e.g. a nonadiabatic loading process of the atoms into the waveguide .
there are several ways to improve our suggested model which would help further clarifying the behaviour of the splitting effect with varying parameters and depending on the initial preparation of the atomic ensemble .
first , one would have to take into account the velocity distribution ( distribution over the collision energy @xmath33 ) in the longitudinal direction @xcite .
shifts and splittings of the cir due to anharmonicities of the trap @xcite and the influence of the inelastic channel of molecule formation have also to be determined . it is known that the anharmonicity of the trap couples the relative and center - of - mass motion of the colliding atoms and can lead to additional anharmonicity - induced resonances ( airs ) due to the removal of the degeneracy of the center - of - mass and relative motions @xcite . however , the airs are supposed to be much narrower than the cirs and , as a consequence , more difficult to be detected experimentally because of the relatively weak anharmonic coupling with respect to the interatomic interaction @xcite . as a conclusion ,
the observation of these airs presumably represents a challenging experimental problem .
explicitly suppressing the anharmonicity of the waveguide could lead to a discrimination between the two mechanisms causing a splitting of the cirs .
we thank dr . e.haller for fruitful discussions .
the authors acknowledge financial support by the deutsche forschungsgemeinschaft and the heisenberg - landau program .
v.s.m . thanks the zentrum fr optische quantentechnologien of the university of hamburg for the warm hospitality .
m. olshanii , phys .
lett . * 81 * , 938 ( 1998 ) .
t. kinoshita , t. wenger , and d. s. weiss , science * 305 * , 1125 ( 2004 ) .
b. paredes , a. widera , v. murg , o. mandel , s. flling , i. cirac , g.v .
shlyapnikov , t.w .
hnsch , and e. bloch , nature * 429 * , 277 ( 2004 ) .
e. haller , m. gustavsson , m.j .
mark , j.g .
danzl , r. hart , g. pupillo , and h.c .
ngerl , science * 325 * , 1224 ( 2009 ) .
k. gnter , t. stferle , h. moritz , m. khl , and t. esslinger , phys .
rev . lett . * 95 * , 230401 ( 2005 ) .
s. saeidian , v.s .
melezhik , and p. schmelcher , phys . rev .
a * 77 * , 042721 ( 2008 ) .
v.s . melezhik and p. schmelcher , new j. phys . * 11 * , 073031 ( 2009 ) v. peano , m. thorwart , c. mora , and r. egger , new j. phys .
* 7 * , 192 ( 2005 ) j.i .
kim , v.s .
melezhik and p. schmelcher , phys .
lett . * 97 * , 193203 ( 2006 ) e. haller , m.j .
mark , r. hart , j.g .
danzl , l. reichsllner , v. melezhik , p. schmelcher , and h.c .
ngerl , phys .
rev . lett . * 104 * , 153203 ( 2010 ) .
s.g . peng , s.s .
bohloul , x.j .
liu , h. hu , and p.d .
drummond , phys .
a * 82 * , 063633 ( 2010 ) .
w. zhang and p. zhang , phys.rev .
a * 83 * , 053615 ( 2011 ) j.p .
kestner and l.m .
duan , new j. phys .
* 12 * , 053016 ( 2010 ) s. sala , p
schneider , and a. saenz , arxiv:1104.1561 .
t. bergeman , m. g. moore , and m. olshanii , phys .
lett . * 91 * , 163201 ( 2003 ) .
e. tiesinga , c.j .
williams , h.h .
mies , and p.s .
julienne , phys .
a * 61 * , 063416 ( 2000 ) ; v.a .
yurovsky , phys .
a * 71 * , 012709 ( 2005 ) ; v.s .
melezhik , j.i .
kim , and p. schmelcher , phys .
a * 76 * , 053611 ( 2007 ) .
melezhik , j. comp . phys . * 92 * , 67 ( 1991 ) v.s .
melezhik and c .- y .
hu , phys .
lett . * 90 * , 083202 ( 2003 ) m. rigol , phys .
lett . * 103 * , 100403 ( 2009 ) a. griesmaier , a. aghajani - talesh , m. falkenau , j. sebastian , a. greiner , and t. pfau , j. phys .
b * 42 * , 145306 ( 2009 ) . | we analyze the impact of multichannel scattering in harmonic waveguides on the positions and widths of confinement - induced resonances for both isotropic and anisotropic transversal confinement .
multichannel scattering amplitudes and transmission coefficients are calculated and used to characterize the resonant behavior of atomic collisions with varying anisotropy .
a mechanism is established which leads to a splitting of the confinement - induced resonance in the presence of anisotropy . |
globular cluster systems were extensively studied using photometry in the last two decades producing many interesting findings ( e.g. ashman & zepf 1997 for a review ) . however
, many new questions were raised that could not be answered using broad
band colors alone because population synthesis models predict ( sometimes strong ) age metallicity degeneracies ( e.g. worthey 1994 ) .
several attempts were made to obtain spectra of globular clusters in systems outside the local group ( mould et al .
1987 , mould et al .
1990 , huchra & brodie 1987 , brodie & huchra 1991 , grillmair et al .
1994 , perelmuter , brodie & huchra 1995 , hui et al .
1995 , bridges et al .
1997 , minniti et al .
however , all of these studies had to focus on the very brightest globular clusters in the host galaxies and only low signal to noise spectra were obtained .
kinematic information could be obtained for the globular clusters but , for the vast majority , line indices were either affected by high photon noise , or not measurable at all . using the new generation of 10m class telescopes , the expectation was that reliable line indices would be measurable for many globular clusters in galaxies at typical distances of the virgo or fornax galaxy clusters . among the open questions
were : do globular clusters in other galaxies show the same abundances and abundance ratios as in the local group , i.e. can they have formed by similar processes ?
are globular clusters in ellipticals more metal rich than in spirals as thought from their broad
band colors ( e.g first pointed out by cohen 1988 ) , and what is the upper metallicity limit ? can the abundances help to identify a population of globular clusters in ellipticals that formed during a merger event ? how do the abundances of globular clusters relate to those of the stars of the host galaxies ?
preliminary answers to some of these questions were discussed by brodie & huchra ( 1991 ) , but conclusive statements require better data than was available to them . here
we report the first spectroscopically derived abundances for globular clusters in ngc 1399 , the central giant elliptical galaxy in fornax .
ngc 1399 , like other cd galaxies , has an over abundant globular cluster system with @xmath16 globular clusters ( kissler - patig et al .
1997 , forbes et al .
photometric studies suggest the presence of multiple globular cluster sub populations ( e.g. ostrov et al .
1993 , kissler - patig et al .
1997 , forbes et al .
1997 ) , the origins of which are uncertain . in the next section
we describe the observations . in section 3
we derive the kinematics of our globular cluster sample and use this information to give an estimate of the galaxy mass and mass to light ratio at several effective radii . the abundances and metallicities of our globular clusters
are derived in sect .
4 . in section 5
we discuss our broad
band colors together with the line indices .
we draw conclusions from our sample in sect .
we selected 21 globular cluster candidates from the original list used by grillmair et al .
( 1994 ) and kindly provided by w. couch .
these cluster candidates were identified on deep photographic plates in the anglo - australian telescope archive and were selected on the basis of their b@xmath17r colors .
the observations were carried out with the low resolution imaging spectrograph ( lris , oke et al .
1995 ) in multi - slit mode at the keck 1 telescope on the nights of 1995 , december 22nd and 23rd .
exposures totaling 160 min were taken over the two nights for one mask with 21 slitlets + 3 centering objects .
comparison lamps and flat fields were taken after the 30 min + 20 min exposures on the night of the 22nd , and before and after the 20 min + @xmath18 min exposures in the night of the 23rd .
the ccd used was a tektronix 2048@xmath192048 with 24 @xmath20 m pixels .
the observations were made with a 600g / mm grating ( blazed at 5000 , with a dispersion 1.24 / pixel ) and 1 arcsec slitlets , resulting in an effective resolution of 5.6 and a usable wavelength range of 4000 to 6100 common to all spectra .
the reduction was done in a standard fashion under iraf , with the help of kelson s et al .
( 1997 ) expector program .
every exposure was bias corrected using an average of bias exposures taken at the beginning and end of the night , adjusted to the overscan region of each image .
the exposures were then divided by internal lamp flatfields , that were averaged and normalized to a value of unity for each slitlet individually , in order to correct for wavelength response .
the flatfielded 2d images were then corrected for the instrument x and y distortions by constructing a distortion map from the flatfield and comparison lamp exposures that was then applied to the science exposures .
the wavelength calibration and sky subtraction was carried out by the expector program .
the wavelength solution was obtained from 2530 hg , ar , ne , and kr lines in the comparison lamp spectra taken before and after the series of science exposures and shifted versus the night sky lines on the sciences exposures . the wavelength solution fitted by a third order legendre polynomial , had a typical 1@xmath21 rms below 0.2 ( @xmath22 km@xmath2 s@xmath3 ) in the region redward of 4800 where it could be well anchored to night sky lines . below 4500
the wavelength solution was less well defined and could deviate at the blue end by up to 1 .
the sky subtraction blueward of 5500 , where only a few weak sky lines are present , worked very well and between 5500 and 6100 only the expected photon noise from the brightest oxygen and sodium night sky lines remained after the subtraction . the individual wavelength
calibrated spectra were then extracted under iraf and averaged to produce high signal to noise spectra .
the flux calibration was done using long slit data of the flux standard feige 25 , observed before the target exposures on the same nights as the targets .
the response curve seems to vary slightly with the position of the spectra on the chip , so that velocities and line indices were measured both on the flux calibrated and flux uncalibrated ( hereafter fluxed and unfluxed ) spectra .
for all objects the results agreed within small fractions of the errors .
values quoted below for velocities and line indices are the average of the two methods .
the final spectra were cross - correlated with two velocity template spectra obtained during the same run ( m31 globular clusters 225 - 280 , @xmath23 km@xmath2s@xmath3 and 158 - 213 , @xmath24 km@xmath2s@xmath3 ) using the method of tonry & davis ( 1979 ) implemented in the fxcor package within iraf . only the spectral region between 4800 and 6000 was used as this is where the wavelength solution is best defined . in most cases
the cross correlation peaks were very well defined and the formal errors returned by the cross correlation were on the order of 25 km s@xmath3 . combined with the errors in the wavelength calibration , we estimate the random error to be on the order of 30 km s@xmath3 .
we checked for any systematic errors in the reduction of the individual objects by using individual exposures instead of the combined spectra , and using different comparison lamp spectra for the wavelength calibration .
the dispersion of these individual measurements was added in quadrature to the formal error .
this total error is given in table 1 , together with the heliocentric velocity , v magnitude and v - i color taken from the nw field of kissler - patig et al .
( 1997 ) , the photographic b@xmath25 magnitude and b@xmath17r color from the original list used by grillmair et al .
( 1994 ) and , for the five objects in common , the velocity quoted by grillmair et al .
note that the objects # 4 , # 8 , and # 16 turned out to be a star and two low redshift early type galaxies , and will not be discussed in the following .
figure 1 shows our velocities versus those of grillmair et al .
( 1994 ) for the five objects in common .
the velocities are in good agreement , within the errors , and no systematic shift can be identified in this sample .
we plotted the location of our globular clusters with respect to the galaxy in fig . 2 , where the rings indicate 1 and 5 galactic effective radii ( taken from goudfrooij et al .
1994 ) and the sizes of the symbols reflect the globular cluster velocities with respect to the mean sample velocity ( open being approaching , solid being receding ) .
our globular clusters are located between 2 and 7 effective radii of the galaxy and extend from the north to the west of the galaxy .
a histogram of the globular cluster velocities is shown in fig .
the mean velocity of our sample is 1293 @xmath26 km@xmath2s@xmath3 , that is offset by 150 km@xmath2s@xmath3 from the systemic velocity of ngc 1399 ( 1447 km@xmath2s@xmath3 , de vaucouleurs et al .
the mean velocities of previous samples are 1517 @xmath27 km@xmath2s@xmath3 ( grillmair et al .
1994 ) and 1353 @xmath28 km@xmath2s@xmath3 ( or 1309 @xmath26 km@xmath2s@xmath3 without an outlier , minniti et al .
the latter sample is mainly situated on the opposite side of the galaxy from ours ( south east of the galaxy ) , making it unlikely that rotation is the cause of our low mean velocity .
this is further supported by the results of grillmair et al .
( 1994 ) , who found no evidence for rotation in their more uniformly distributed sample .
the velocity dispersion derived from our sample is 302 @xmath29 km@xmath2s@xmath3 .
it is slightly lower than , but consistent within the errors with , the ones derived by grillmair et al .
( 1994 ) : @xmath30 km@xmath2s@xmath3 , and minniti et al .
( 1998 ) : @xmath31 km@xmath2s@xmath3 ( or @xmath32 km@xmath2s@xmath3 without an outlier ) . as noted in the previous studies ,
the velocity dispersion is higher than measured for the stellar component within 1.5 arcmin , suggesting that either the mass to light ratio and/or radial anisotropy changes dramatically between 1.5 and 5 arcmin .
figure 4 shows magnitude and color versus velocity .
figure 5 shows velocity versus radius .
given the small size of our sample only very clear correlations would be discernible .
apparently velocity does not strongly correlate with radius , color or magnitude of the globular clusters .
such correlations will be explored more fully in a forthcoming paper which will including all three ( grillmair et al . 1994 , minniti et al . 1998 , and this work ) globular clusters samples ( kissler - patig et al .
1998 ) .
the velocity distribution of the globular clusters can be used to give an estimate of the mass of the parent galaxy .
given the small number of velocities and the poor spatial coverage , we follow a simple method also applied by huchra and brodie ( 1987 ) .
we estimate the mass from the virial theorem on the one hand , and from the projected mass method on the other hand ( bahcall & tremaine 1981 , heisler , tremaine & bahcall 1985 ) .
the virial theorem mass for a system of @xmath33 globular clusters with measured velocities is @xmath34 where @xmath35 is the separation between the @xmath36th and @xmath37th cluster , and @xmath38 is the velocity difference between the @xmath36th cluster and the mean system velocity .
the projected mass is @xmath39 where @xmath40 is the separation of the @xmath36th globular cluster from the centroid .
the quantity @xmath41 was chosen to be 1.5 following heisler et al .
( 1985 ) , as a slight correction to @xmath42 , used in a discrete model but determined analytically from a continuous model . the quantity @xmath42 depends on the distribution of the orbital eccentricities for the globular clusters , it ranges from to @xmath43 for radial orbits , over @xmath44 for isotropic orbits , to @xmath45 for circular orbits , in an extended mass .
we adopt a projection factor of @xmath44 , which assumes purely isotropic cluster orbits in an extended mass , for comparison with the simulations of hernquist & bolte ( 1993 ) .
the results from both methods are : @xmath46 and @xmath47 within a radius of 5 arcmin or 28 kpc , adopting a distance to ngc 1399 of 19.3 mpc ( madore et al .
1997 ) .
the errors were calculated by a standard jackknife procedure ( e.g. efron & tibshirani 1993 ) .
further , we have to consider the incomplete spatial coverage of our sample .
we derived the effective radius of the globular cluster system by fitting a de vaucouleurs profile to the globular cluster surface density data of kissler - patig et al .
we obtained an effective radius r@xmath48 .
the mean radius of the spectroscopic sample is @xmath49 r@xmath50 .
richestone & tremaine ( 1984 ) showed that a single velocity dispersion observed at 1.6 r@xmath10 ( approximating our observations ) can vary by a factor of @xmath51 .
thus , mass to light ratios derived from observations like ours can vary by a factor @xmath52 .
an additional error might be present in the projected mass method : hernquist and bolte ( 1993 ) simulated a globular cluster distribution around a galaxy and noted that a sample like ours ( roughly 20 objects extending out to 30 kpc , assuming @xmath53 ) , systematically overestimates the total galaxy mass by up to 50% .
in addition , the value we adopted for the projection factor was assumed without knowledge of the real globular cluster orbits .
the total luminosity within 5 arcmin is obtained from the light profile of bicknell et al .
( 1989 ) , correcting for our assumed distance of 19.3 mpc .
we obtain a total integrated luminosity within 5 arcmin of @xmath54 and estimate a mass to light ratio within 28 kpc of @xmath55 using @xmath56 or @xmath57 ( likely to be an overestimate , see above ) using @xmath58 .
grillmair et al . (
1994 ) derived @xmath59 , but assumed a distance to ngc 1399 of 13.2 mpc , i.e. their value would fall around @xmath60 for our assumed distance .
minniti et al . (
1998 ) derived @xmath6 values between 50 and 130 , also assuming a slightly shorter distance .
all samples cover comparable galactocentric distances .
in summary the values agree when corrected for the different assumed distances .
the main result is that this mass to light ratio is about a factor 10 above that expected for an old stellar population , leading to the conclusion that , at this distance from the center ( @xmath61 r@xmath10 ) , dark matter dominates the potential .
the power of globular clusters for such studies was nicely illustrated by cohen & ryzhov ( 1997 ) , who computed @xmath62 ratios at various radii from a large sample of globular cluster velocities in m87 .
finally , we note that our results are in good agreement with the ones derived from x ray data .
jones et al . ( 1997 ) report a total mass for ngc 1399 of 4.3 to @xmath63 within 18 arcmin , and a mass to light ratio increasing from @xmath64 at 2.6 arcmin to @xmath65 at 15.8 arcmin ( for an assumed distance of 24 mpc ) .
absorption line indices were measured following the procedure described in brodie & huchra ( 1990 ) who used the lick / ids system bandpasses defined in burstein et al .
mean heights were defined in each of the pseudocontinuum regions and a straight line was drawn through their midpoints .
the difference in flux between this line and the observed spectrum in the index bandpass determines the index . in the following paragraph
we discuss some details of the line measurements that can slightly affect the index values .
we note that the lick / ids bandpasses as defined in burstein et al .
( 1984 ) were recently fine tuned and the system was extended ( see gonzales 1993 and trager 1997 ) . unless otherwise noted , we used the burstein et al .
( 1984 ) definitions in order to be compatible with the metallicity calibrations of brodie & huchra ( 1990 ) . further , unless otherwise noted , we did not artificially degrade our resolution from 5.6 to the 8.5 - 9 of the lick / ids system , again to stay consistent with brodie & huchra ( 1990 ) whose most relevant data for our comparison were uncorrected for resolution effects and had 5 resolution ( the rest of the brodie & huchra data spanned the range from 5 to 12 resolution ) .
we tested the effects of resolution by measuring indices from our original spectra and from our spectra degraded , with a gaussian filter , to 9 resolution .
as already noticed by gonzales ( 1993 ) , the broader indices ( e.g. mg2 ) are barely affected , while narrower indices ( fe , h@xmath13 , ... ) are systematically lower ( by about @xmath66 mag in our case ) when measured from the lower dispersion spectra .
this amounts only to @xmath67 of our photon noise error , but it is systematic .
we further tested our index measurement routine versus that used by the lick / ids group in order to check the determination of the continuum and the treatment of the bandpass end
points . for this purpose worthey
provides a set of high signal to noise templates and a list of index bandpasses measured by the lick / ids software ( http://www.astro.lsa.umich.edu/users/worthey/ ) on which one can test one s own program .
we found a perfect agreement between our software and worthey s ( the original used to measure the indices in the lick / ids system ) .
finally , we checked for differences using the new and old lick / ids bandpasses ( typically shifted by 0.5 - 2.6 from the original definitions ) .
no systematic effect could be found for most indices , the variations depending on the photon noise in individual resolution elements of the spectra .
however we found the index values for fe5335 ( whose definition was shifted by 2.6 on the blue side ) to be systematically lower ( by @xmath68 mag ) when using the new definition . we conclude that care should be taken when inter comparing measurements of various groups .
an exact comparison of the values quoted in table 2 with the latest lick / ids system is only possible after correcting for the systematic offsets .
finally we point out that all indices in table 2 are given in magnitudes , while the lick / ids system usually quotes the atomic indices in .
further , in the same table , the values for ngc 1399 taken from huchra et al .
( 1996 ) were measured on the spectrum without deconvolving it with the velocity dispersion of the galaxy ( @xmath69 km s@xmath3 ) .
the effect is very similar to degrading the resolution of the spectrum by almost a factor two , and the measured line indices are therefore systematically lower than trager s ( 1997 ) . in the following
we will use the data for ngc 1399 from trager ( 1997 ) when comparing the galaxy with the globular clusters .
representative fluxed spectra of a blue , a red and an extremely red globular cluster are shown in fig .
the spectra displayed were smoothed with a 3 pixel average filter .
we measured all our indices on the fluxed and unfluxed spectra .
the duplication test was carried out because of the uncertainties of flux calibrating multi slit data with flux standards taken in long slit mode .
flux calibrating is a multiplicative procedure and should not , therefore , affect the line indices .
only broad indices could be affected if the slope of the continuum dramatically changes from the unfluxed to the fluxed spectrum .
we compared the indices measured on the flux and unfluxed spectra and found the differences to be negligible .
all values given in table 2 are averaged from the results from the fluxed and unfluxed spectra .
the errors were estimated from the photon noise in the bandpasses ( see brodie & huchra 1990 ) and from the error in the wavelength calibration for indices blueward of 4500 ( only g band ) .
all spectra had comparable signal to noise and the errors were found to be similar for all objects at a given index . in fig .
7 we show the relations between mg2 and various other indices , together with the range covered by the milky way and m31 globular clusters ( shown as shaded areas ) .
the first point to note is that the globular clusters in ngc 1399 ( except for two clusters that we will discuss further below ) cover the _ entire _ range spanned by milky way and m31 globular clusters .
they do not show anomalies in the metal tracing indices such as mg and fe , or in the more age sensitive index , h@xmath13 .
apparently ngc 1399 mostly hosts globular clusters with ages and metallicities similar to those found in the spirals of the local group . in fig .
8 and 9 we plot mg2 versus the equivalent width of fe5270 , fe5335 and h@xmath13 in order to compare them to population synthesis models , to the value of the host galaxy ngc 1399 , and to similar plots for elliptical galaxies ( e.g worthey et al .
the indices of table 2 are shown as solid dots . for clusters # 2 and # 14
we also included the indices measured in the new lick / ids system ( new bandpass definitions , resolution degraded to 9 ) as open circles .
the values for ngc 1399 itself are taken from trager ( 1997 ) and shown as triangles .
for completeness we show in table 2 also the line indices derived by huchra et al .
( 1996 ) for ngc 1399 .
note that these latter did not deconvolve the galaxy spectrum with the velocity dispersion of the galaxy before measuring the line indices .
this leads to systematically lower values of huchra et al .
( 1996 ) and prevents a direct comparison with trager s or the globular cluster measurements . in fig .
8 the population synthesis models of fritze - v .
alvensleben & burkert ( 1995 ) are shown as long dashed lines for 8 and 16 gyr year old populations , with metallicity varying between z=0.001 and z=0.04 .
worthey s ( 1994 ) models are plotted as short dashed lines for 8 and 17 gyr and metallicity varying between @xmath11fe / h@xmath12=-2.0 $ ] and @xmath70 dex .
the range covered by the mg rich elliptical galaxies of worthey et al .
( 1992 ) is shown as a hatched area .
the same symbols are used in fig . 9 , except that we show the tracks for 16 , 8 , and 3 gyr for models of fritze - v .
alvensleben & burkert ( 1995 ) , and for 17 , 8 , 3 , and 1.5 gyr in the case of worthey s ( 1994 ) models .
the observed relations between fe and mg2 are well reproduced by the population synthesis models of fritze - v .
alvensleben & burkert ( 1995 ) and worthey ( 1994 ) .
given the small age dependence of these indices and our errors , we can not discriminate age differences from this plot ( fig .
worthey et al . (
1992 ) report a mg versus fe enhancement for large ellipticals .
unfortunately we have only two globular clusters with mg high enough to compare them with the affected ellipticals .
given our errors on fe , these two globular clusters are compatible with the models and also fall in the range covered by the mg rich ellipticals .
the other element observed to be unusually high in bright ellipticals is nad .
indeed the two metal rich clusters , especially # 14 , have high nad abundances that could hint to abnormal high values . in summary , both globular cluster # 2 and # 14 have values close to the ones measured for the stars in ngc 1399 .
more data for mg rich globular clusters are clearly needed to investigate this interesting point : since globular clusters can be as rich in mg as large ellipticals , their study might constrain the scenarios responsible for the mg versus fe enhancement in large ellipticals .
worthey et al . (
1992 ) summarize the three mechanisms that can be responsible for an enhanced @xmath11mg / fe@xmath12 $ ] ratio in large ellipticals versus normal ratios in smaller ellipticals : _
i ) _ different star formation timescale i.e. no time for sne ia to dilute the abundance ratios set by the earlier sne ii ; _ ii ) _ variable / flatter imf , e.g. in mergers , to produce more sne ii versus sne ia ; _ iii ) _ selective loss mechanisms retaining mg and expelling fe in giant ellipticals ( this last scenario is rather unlikely as they note ) .
while selecting between these scenarios is difficult on the basis of the properties of the starlight alone , globular clusters might help to settle the issue .
if , for example , the mg enrichment is a result of a flatter imf for star formation in mergers , globular clusters formed in mergers will show the mg over enrichment too .
if , however , the high @xmath11mg / fe@xmath12 $ ] values are the result of different star formation timescales among ellipticals , and the mg rich globular clusters are young ( a possibility discussed below ) , e.g. formed in a later merger , then they will probably not show the higher @xmath11mg / fe@xmath12 $ ] ratios .
the relation between mg2 and h@xmath13 is interesting in many respects .
first it shows the significant differences in the predictions from various models for the value of h@xmath13 at a given age and metallicity .
a discussion of the causes of these differences is out of the scope of this paper .
we concentrate on the points in common to all models .
these are , on the one hand that our globular clusters ( except for the two outliers ) could span a large range of age according to the models .
our error in h@xmath13 does not allow an age determination better than within a factor two , but our measurements are also consistent with a single age , within the errors . and we note that our values lie in the range spanned by the milky way globular clusters ( see fig . 7 ) , that span a small ( few gyr ) range in age .
the ngc 1399 globular clusters could , therefore , all have similar ages and be as old as the globular clusters in the milky way .
on the other hand , we note that the h@xmath13 values for our two metal rich clusters can not be reproduced by any model , no matter what age or metallicity is assumed .
large values for h@xmath13 were already reported ( at lower mg ) in some m31 globular clusters ( burstein et al .
1984 , brodie & huchra 1991 ) . to check alternative explanations to age for strong h@xmath13 values , burstein et al .
( 1984 ) computed a semi empirical model to estimate the effects on h@xmath13 and mg when taking a cluster like 47 tuc or m71 as starting point ( both having stubby , red horizontal branches ) and changing the horizontal branch to be blue like that of m5 ( so as to yield maximum h@xmath13 strength ) .
their result shows that h@xmath13 can be raised roughly from 1.5 to 2.5 , without significantly affecting the mg abundance .
quantitatively , this alone could explain the position of globular cluster # 2 and # 14 in fig .
this is very interesting in the light of the extended blue horizontal branches found recently by rich et al .
( 1997 ) in metal rich milky way globular clusters .
blue horizontal branches are unexpected ( and were , until recently , not observed ) at typical bulge metallicities ( e.g. fusi - pecci et al .
1992 , 1993 ) .
this is probably the reason why the contribution of a blue horizontal branch is not taken into account in any of the models at high metallicities even though it can contribute to stronger h@xmath13 values . to date , in the models of metal rich stellar populations , most of the contribution to h@xmath13 and the blue light came from stars from the turn off region , so that variations in color or h@xmath13 had to be explained by the position of the turn off point , i.e. by variations in age .
this necessarily created a strong age
metallicity degeneracy .
new models which include the possible contribution of blue horizontal branches are clearly needed to cover the whole of the possible abundance range at given age and metallicity for globular clusters .
more data for globular clusters at high mg and h@xmath13 might help in the future to fine tune the populations synthesis models .
our two clusters # 2 and # 14 are outliers in many respects and bring insight into some of the properties of very metal rich globular clusters .
unfortunately we do not have a v@xmath15i color for our cluster # 2 and the b@xmath17r color is too uncertain to rely on ( e.g. compare colors for # 10 and # 20 ) .
however , # 14 is the reddest cluster in the sample and falls in the very red tail of the color distribution ( kissler - patig et al .
1997 ) , hinting that such objects are rare .
further these two objects show that globular clusters can have solar abundances ( see table 3 ) but they are not much in excess of solar ( except perhaps for their mg abundances ) .
this is contrary to results from broad band photometric studies ( johnson & washington photometry ) in several galaxies ( e.g. secker et al .
1995 for the most extreme case ) where globular clusters were reported as having metallicities up to four times solar . while the abundances of the rest of the clusters in ngc 1399 are similar to abundances found in the milky way and m31 , the abundances of clusters # 2 and # 14 are clearly anomalous .
we note that the measurement of the h@xmath13 index is controlled in a negative sense by mg ( tripicco & bell 1995 ) .
thus , in our particular case , h@xmath13 could even be slightly underestimated . in order to further check the strength of the hydrogen lines
, we measured h@xmath14 for clusters # 2 and # 14 , using the bandpasses defined by worthey & ottaviani ( 1997 ) extending the lick / ids system ( h@xmath71 for young stellar populations , h@xmath72 for old stellar populations ) . for this purpose
we degraded the resolution of our spectra to 9 and re determined the wavelength calibration around the h@xmath14 line , before measuring the indices . in both cases we were close to the blue limit of our spectral range in a region showing a factor 10 less signal than in the region between 4800 and 6000 .
we obtained values of h@xmath73 and @xmath74 and h@xmath75 and @xmath76 for clusters # 2 and # 14 respectively . despite the large errors , all the values are unexpectedly high at the high mg values and are incompatible with high ages ( @xmath77 gyr ) according to worthey & ottaviani ( 1997 )
further , the blue continuum bandpass for h@xmath14 exactly covers the g band , which is strong in both clusters , so that our h@xmath14 indices are probably underestimates when compared to young stellar systems .
it remains unclear how to interpret these strong balmer lines , and in particular the high h@xmath13 values . as discussed above ,
they could be explained by the presence of blue horizontal branches ( which would also account for the location of the clusters in fig .
9 ) and/or the strong balmer lines could be , at least partly , associated with a younger age but new models would have to be computed for a reliable comparison .
a young age can not , by itself , explain the location of the clusters in fig .
9 . for ages
as young as 1 to 5 gyr , one would expect diluted metal lines and blue broad band colors , which are definitively not seen in these clusters .
however , a younger age would be consistent with the very high metallicities of these clusters ( see next section ) in a picture where they formed later than the other clusters from highly enriched gas .
their absolute luminosities provide no further clues .
the difference between an 15 gyr and 2 gyr old cluster ( same mass and same metallicity ) would be on the order of 2 magnitudes .
these clusters are not exceptionally luminous within our sample .
they have absolute v luminosities around @xmath78 given our assumed distance .
they could be old and have masses as high as omega centauri ( v@xmath79 , the most massive globular cluster in the milky way ) or be younger and have lower , more `` normal '' , masses .
finally , we note that three of our other globular clusters show mg values at the upper limit of the range spanned by the milky way and m31 .
two of these clusters also exhibit slightly stronger h@xmath13 values , than expected for old clusters ( although not significantly when considering the errors )
. this might be yet another sub population . clearly , larger samples of globular clusters with well defined abundances will be valuable in exploring in more detail globular cluster systems with complex formation histories .
we derived @xmath11fe / h@xmath12 $ ] values for the globular clusters using our indices and following the method of brodie & huchra ( 1990 ) .
we applied their index@xmath11fe / h@xmath12 $ ] relations to various elements and listed the results in table 3 .
we include values for the g band , cn , and the ca h@xmath80k lines for the few spectra extending far enough to the blue , as well as nad where the night sky lines could be well enough subtracted .
brodie & huchra ( 1990 ) derived the relations from milky way and m31 clusters , i.e. the relations do not extend to the high metallicities seen for some of the globular clusters in ngc 1399 . in particular , we found the relations between mg2 and @xmath11fe / h@xmath12 $ ] , as well as between mgh and @xmath11fe / h@xmath12 $ ] to be badly approximated by a linear extrapolation at higher metallicities ( mg2@xmath81 ) when compared to population synthesis models ( see fig .
for these two relations we applied a correction to the five and two most metal rich globular clusters respectively by allocating them the metallicity predicted by the 17 gyr models at their mg abundance .
in addition to the @xmath11fe / h@xmath12 $ ] values from individual indices we computed a mean @xmath11fe / h@xmath12 $ ] value from fe5270 and mg2 ( shown in column 10 of table 3 ) and a weighted mean @xmath11fe / h@xmath12 $ ] from all available indices ( column 11 ) . the weights were chosen to reflect the quality of the calibrating relation on the one hand , and the quality of the measured index on the other hand .
we assigned weights of 1 to mg2 and fe5270 , and 0.2 to all other available indices .
we estimated the errors from the errors on the indices .
the error on the final weighted mean metallicity is estimated to be 0.20 dex for most clusters , slightly higher ( @xmath82 dex ) for the clusters with high mg2 .
brodie & huchra ( 1990 ) calibrated their index metallicity relations using the metallicities from zinn & west ( 1984 ) , which are mainly based on the ca ii k line . the zinn & west scale
was recently claimed to be slightly non linear compared to the total metal abundance scale ( carreta & gratton 1997 ) .
we emphasize that our results are necessarily tied , via the brodie & huchra ( 1990 ) calibration , to the zinn & west ( 1984 ) results . in fig .
11 we plotted @xmath11fe / h@xmath12 $ ] values derived from fe5270 and mg2 versus each other and versus the weighted mean @xmath11fe / h@xmath12 $ ] values .
we also show a histogram over the weighted mean metallicity values for our clusters .
the @xmath11fe / h@xmath12 $ ] values derived from mg for high metallicities lie somewhat above the values derived from fe5270 .
as already mentioned , this is mostly due to the fact that , for these high mg2 values , the relation defined by brodie & huchra ( 1991 ) is no longer valid , and our correction crude .
it could also be due to a slight over enrichment of mg versus fe , but our data do not allow us to make any strong statement in this regard ( see last section ) . in fig .
11 ( lower left panel ) , three additional clusters seem to have deviating mg / fe ratios , however this trend is not confirmed by their fe5335 values in fig . 8 ( lower panel ) where these clusters show normal fe5335 versus mg2 abundances . for metallicities below -0.3 dex
the agreement between the @xmath11fe / h@xmath12 $ ] value derived from mg2 and fe5270 is good .
the histogram over the weighted mean @xmath11fe / h@xmath12 $ ] values shows that our sample includes globular clusters with metallicities ranging from typical values for milky way halo globular clusters to slightly above solar but there are no objects with significantly super solar abundances .
very metal poor clusters ( @xmath11fe / h@xmath12<-1.6 $ ] ) appear to be missing , but recall that the blue part of the globular color distribution in ngc 1399 is not well sampled . the mean metallicity for our sample is @xmath11fe / h@xmath12=-0.83\pm0.13 $ ] dex ( the error is the standard error of the mean ) .
the prediction from the galaxy luminosity
mean globular cluster metallicity relation of brodie & huchra ( 1991 ) is @xmath11fe / h@xmath12=-1.14\pm0.12 $ ] , or @xmath11fe / h@xmath12=-0.91\pm0.16 $ ] ( when using the relation defined from ellipticals only ) , adopting , for ngc 1399 , b@xmath83 mag ( de vaucouleurs et al .
1991 ) and a distance modulus of 31.43 ( madore et al .
1997 ) .
because of the size of our spectroscopic sample , we have to return to broad band colors to comment further on the metallicity distribution in the globular cluster system of ngc 1399 .
in this section we compare the line indices with the broad band colors of our globular clusters .
we present the indices versus our v@xmath15i colors from kissler - patig et al .
( 1997 ) , for which the typical error on the color is 0.035 mag at these magnitudes .
being of photographic origin , the uncertainties in existing b@xmath17r colors ( from grillmair 1992 ) listed in table 1 are too large ( of the order of 0.3 mag ) for our purposes . in fig .
12 we show our v@xmath15i colors versus mg2 , mgb , @xmath84fe@xmath85 ( defined as the mean of fe5270 and fe5335 , as introduced by burstein et al . 1984 ) , and h@xmath13 .
we note that the broad band color correlates well with the metal indices .
this might have been expected since neither mg or fe are very age sensitive ( e.g. worthey 1994 ) , so that age
metallicity degeneracy is not a factor here .
however , we note that v@xmath15i also correlates well ( inversely ) with h@xmath13 . as in fig . 9 ,
the v@xmath15i colors scatter by less than their typical errors , and are therefore consistent with a single age for these clusters .
but , as already noted above , the limiting factor in deriving ages are our errors in h@xmath13 which span a factor of two in age according to population synthesis models ( e.g. fritze - v . alvensleben & burkert 1995 , worthey 1994 ) . formal linear fits return the following relations : + mg2 @xmath86v@xmath15i + @xmath84fe@xmath85 @xmath87v@xmath15i + h@xmath88v@xmath15i + in fig
. 13 we plot our v@xmath15i colors versus the weighted mean [ fe / h ] ( see above section and table 3 ) as filled triangles , together with the values for all milky way globular clusters that have a reddening of less than e(b@xmath15v)=0.2 ( open circles ) .
the milky way values were taken from the mcmaster database ( harris 1996 ) and de reddened according to rieke & lebofsky ( 1985 ) . for our cluster # 14
we plotted the metallicity derived from fe5270 , as an open triangle .
for the first time , insight can be gained into the relationship between globular cluster metallicities and colors for the reddest clusters ( v@xmath15i @xmath89 mag ) .
we note that the relation may be slightly non linear in the sense that the slope gets flatter to redder colors , as predicted by the population synthesis models .
while the v@xmath15i color reflects well metallicity , as shown above , it could still be affected by age within our errors .
we stress again that the ages of our globular clusters are not well defined , but they are compatible with the ages of milky way globular clusters , as inferred from all the measured element abundances .
the most metal rich globular clusters may have formed later than the rest .
if they are , for example , half as old as the metal poor ones , their v@xmath15i color is shifted to the blue by 0.1 magnitude .
that is , if we corrected the colors of our most metal rich globular clusters for age , we would get an even flatter slope .
however since the ages are uncertain , we will not apply any corrections for age to the colors in what follows .
given the restricted number of data points for colors above v@xmath15i=1.2 , we attempted only a linear fit to the sample ( including the selected milky way globular clusters ) .
the returned relation between metallicity and v@xmath15i is : + @xmath11fe / h@xmath12 $ ] @xmath90v@xmath15i + the slope of this relation is almost twice as flat as the ones previously derived by couture et al .
( 1990 ) and kissler - patig et al .
( 1997 ) from the milky way globular clusters alone .
a non linear fit will lead to an even more dramatic result . for red colors , metallicities derived from v@xmath15i , and most probably from other broad band colors too , have been overestimated .
since most globular cluster systems in early type galaxies have red median colors ( v@xmath15i @xmath89 ) , most mean metallicities have probably been overestimated in the past .
kissler - patig et al .
( 1997 ) have shown that the color distribution of ngc 1399 has two peaks " at v@xmath15i @xmath91 mag and v@xmath15i @xmath92 mag ( confirming the multi
modal distribution found with washington photometry by ostrov et al .
1993 , also seen by forbes et al .
thus , if these two peaks are associated with two different sub populations , these two populations would have mean metallicities around [ fe / h]@xmath93 dex and [ fe / h]@xmath94 dex , very similar to the means of the halo and disk / bulge populations of the milky way .
the same applies to globular clusters in m87 which peak at v@xmath15i@xmath95 and v@xmath15i@xmath96 ( elson & santiago 1996 ) .
this corresponds to [ fe / h]@xmath97 dex and [ fe / h]@xmath98 dex , similar to ngc 1399 and again similar to the milky way .
further , the small ellipticals ngc 1374 , ngc 1379 , ngc 1387 , ngc 1427 which appear uni modal ( kissler - patig et al .
1997 ) would have mean metallicities around [ fe / h]@xmath99 , @xmath100 , @xmath101 , @xmath102 dex respectively .
that is , their globular clusters do not have solar or super solar mean abundances as previously thought .
moreover , the spread in metallicity between the different galaxies is reduced to 0.5 dex instead of 1 dex , leading to a more homogeneous picture and smaller discrepancies between metal rich globular clusters in spirals and ellipticals .
finally we note that very metal rich globular clusters such as our objects # 2 and # 14 , with v@xmath15i colors @xmath103 make up only a small fraction ( 25% ) of the total globular cluster system , as derived from the color distribution of kissler - patig et al .
if these objects are significantly younger , i.e. brighter , they might even be over represented in the magnitude limited sample of kissler - patig et al .
( 1997 ) and represent an even smaller fraction of the total globular cluster system .
we conclude from the above that v@xmath15i , and thus probably the other broad band colors too , trace metallicity relatively well in globular cluster systems . drawing a tentative conclusion from this sample , the v@xmath15i relation can be used if most objects are old and do not show a significant age spread , as it seems the case for our sample that scatters around an isochrone within the measurement errors ( see section 4 ) .
generally , this will be the case for any globular cluster system that formed at high redshift with no significant fraction of clusters that formed since @xmath104 .
further , given the spread in [ fe / h ] at a given color , metallicities from v@xmath15i will be accurate only to 0.5 dex for individual objects .
however , mean metallicities for entire globular cluster systems can probably be derived with an accuracy of 0.3 dex from v@xmath15i .
this will , for example , allow corrections for metallicity effects to distances derived from the globular cluster luminosity function , when the mean color of the globular cluster system is known , as proposed by ashman , conti & zepf ( 1995 ) .
most globular clusters in ngc 1399 have very similar mg , fe and h@xmath13 line indices to the milky way and m31 globular clusters and span the full range observed in these galaxies . the metal
poor clusters have metallicities corresponding to the mean metallicity of milky way halo clusters , the metal
rich clusters have similar metallicities to milky way bulge / disk clusters .
their mg and fe and especially h@xmath13 indices indicate that these clusters are probably as old as the milky way globular clusters , although deriving accurate ages is difficult given the observational errors and model uncertainties .
two clusters clearly stand out in their abundances .
they have mg and fe ( and na ) indices significantly higher than observed in any globular cluster in the milky way or m31 , and comparable to the integrated starlight of giant ellipticals .
their metallicities are estimated to be around , or even slightly above , solar .
whether or not these clusters exhibit a mg versus fe over abundance , as seen in the integrated starlight of giant ellipticals , remains an interesting open question .
moreover , these two clusters show abnormally high h@xmath13 ( and h@xmath14 ) values that can not be matched by population synthesis models for any age and metallicity , making any age determination very uncertain and calling for an explanation
. blue horizontal branches in these metal rich clusters would qualitatively be a solution .
the formation of these globular clusters several gigayears after the more metal poor globular clusters could also be part of the explanation but needs revisited models including blue horizontal branches at high metallicity for comparison .
this would confirm the formation of at least some globular clusters in a later phase as suggested for globular clusters forming in mergers ( schweizer 1987 , ashman & zepf 1992 ) . from the globular cluster color distribution
, we estimate that these peculiar clusters do not make up more than 5% of the total population , hinting to mergers involving little dissipation .
the overall picture of the system seems to hint at the presence of various sub
populations similar to the ones seen in the milky way , as well as a small fraction of globular clusters that formed later from highly enriched gas .
that is , processes like the ones that produced the milky way halo and disk / bulge population can be responsible for the formation of the vast majority of the globular clusters in ngc 1399 , only a small percentage of the total number of globular clusters need to have formed later from solar metallicity gas .
fritze - v .
alvensleben & gerhard ( 1994 ) showed that , at the end of a starburst induced by a merger event , the metallicity can reach solar , independently of the age of the event ( under the assumption that no gas is lost from , or accreted by , the merging galaxy pair ) .
this would allow the formation of the extremely metal rich clusters in a merger as soon as 3 gyr after the formation of the progenitor galaxies if globular clusters formed at the end of the starburst .
these few new globular clusters would be added to the globular clusters already present in the progenitors ( the majority ) .
translated into implications for the formation of ngc 1399 and its over abundant globular cluster system : the galaxy and its globular clusters are likely to have formed at an early time .
the blue population formed from material as metal poor as the galactic halo , despite the fact that it existed in a high density environment like the center of a cluster of galaxies , and it presumably formed at a similar early epoch .
the relatively low mean metallicity ( compared to previous estimates , based on broad band colors ) of the `` red '' population ( peaking at v@xmath15i=1.18 ) of @xmath11fe / h@xmath12\simeq -0.6 $ ] dex , implies formation before the gas could be enriched to solar abundances .
whether these globular clusters formed in early merger events or during the collapse of the `` bulge '' ( as might have been the case in the milky way , e.g. minniti 1995 ) can not be distinguished from our data .
however , judging by the globular cluster color distribution and the metallicity of the very red objects , only a very small fraction of the globular clusters clearly formed later from solar metallicity gas , i.e. later mergers are unlikely to be the cause for the over abundant globular cluster system of ngc 1399 . to explain the abnormally high number of globular clusters and specific frequency of ngc 1399 , alternative models are needed ( see e.g. forbes , brodie & grillmair 1997 , blakeslee et al .
1997 ) , and the role of mechanisms like stripping from neighboring galaxies ( muzzio 1987 ) , or accretion of dwarf galaxies ( hilker et al .
1997 ) must be better understood .
we obtained moderate resolution spectra for 18 globular clusters in the giant elliptical cd galaxy ngc 1399 in fornax . from the derived velocities we calculated a galaxy mass of 1 to @xmath105 , leading to a mass to light ratio of @xmath106 or @xmath107 ( depending on the estimator ) within 5 arcmin ( 6.7 r@xmath10 ) , for an assumed distance of 19.3 mpc
this would imply a dark matter dominated potential at this radius .
no correlation between magnitude , color or position with velocity could be found in our small sample .
the element abundances of most globular clusters in ngc 1399 do not differ from the ones observed in the milky way and m31 .
no different processes and formation time than the ones assumed for the formation of the milky way halo and disk / bulge globular clusters are needed to explain the majority of globular clusters in ngc 1399 .
however , two of the globular clusters in our sample clearly stand out in the strength of their line indices and show abundances similar to the starlight of giant ellipticals .
they must have formed from solar metallicity gas in a different formation process , e.g. merger event , although not necessarily much ( @xmath108 gyr ) later than the others .
further , these very metal rich clusters show unexplained high h@xmath13 ( and h@xmath14 ) abundances , incompatible with any age
metallicity combination of existing models .
these abundances can , however , be explained by blue horizontal branches . judged from the color distribution , these clusters represent less than 5% of the total number of globular clusters .
this hint at late mergers not being the cause of the over abundant globular cluster system around ngc 1399 .
metallicity degeneracy of broad band colors , as predicted by the models , is presumably artificially strengthen by not taking into account a possible contribution to the blue light from the horizontal branch at high metallicities .
band colors turned out to be very good metallicity tracers in the globular cluster system .
while for individual cases the metallicity of a globular cluster can only be derived to an accuracy of 0.5 dex from its v@xmath15i color , the mean metallicity of a globular cluster system can be determined to an accuracy of about 0.3 dex .
however , we stress that the relation derived here has a slope about twice as shallow as the ones previously extrapolated from the milky way system , making globular cluster metallicity differences between galaxies less significant and bringing estimates for red globular clusters back from super solar to solar metallicities . in a recently submitted paper cohen , blakeslee & ryzhov ( 1998 ) present a similar study in m87 .
they obtained keck / lris spectra for a larger sample of globular clusters and find a similar metallicity range as for the clusters in ngc 1399 , ranging from metal poor to solar metallicities , and an old median age for their globular cluster sample .
their spectroscopy strengthen the point that globular clusters in large ellipticals have very similar line indices to the globular clusters in the milky way and m31 .
we would like to thank the staff at the keck observatories , as well as the entire team of people , led by j.b .
oke and j.g .
cohen , responsible for the low resolution imaging spectrograph , for making the observations possible .
we would also like to thank a. phillips for his help to run his slitmask preparation and alignment software .
further , we are thankful for interesting discussions and comments from mike bolte and sandy faber .
many thanks also to uta fritze - v .
alvenlesben for the electronic version of her new models , as well as useful remarks .
thanks to dan kelson for providing his program expector prior to public release , and to luc simard for his help in using it .
thanks to scott trager and sandy faber for making their data on ngc 1399 available prior to publication .
we thank the referee j. secker for some useful comments .
c cc rc rc r r 1 & 3 36 13.8 & -35 39 24.8 & & & 21.8 & & 732 @xmath109 & 2 & 3 36 14.2 & -35 38 51.2 & & & 22.4 & 1.19 & 1094 @xmath110 & 3 & 3 36 09.4 & -35 37 32.4 & & & 22.3 & 1.02 & 1571 @xmath111&4 & 3 36 12.0 & -35 37 44.3 & & & 21.9 & 1.20 & 279 @xmath112 & 5 & 3 36 13.2 & -35 37 37.8 & & & 21.8 & 1.17 & 1775 @xmath113 & 6 & 3 36 17.8 & -35 37 50.2 & & & 22.3 & & 1386 @xmath111&7 & 3 36 16.7 & -35 37 01.7 & 21.01 & 1.23 & 21.7 & 1.31 & 1376 @xmath114 & 1677 @xmath1158 & 3 36 15.4 & -35 36 17.1 & & & 22.3 & 1.34 & @xmath116 & 9 & 3 36 19.2 & -35 36 28.7 & 21.04 & 1.25 & 21.8 & 1.33 & 1150 @xmath111&1280 @xmath11510 & 3 36 21.5 & -35 36 04.4&20.55 & 1.05 & 21.4 & 1.50 & 815 @xmath117&980 @xmath11511 & 3 36 25.0 & -35 36 28.3&21.34 & 1.17 & 22.2 & 1.50 & 1338 @xmath118 & 12 & 3 36 20.3 & -35 35 15.3&21.97 & 0.94 & 22.4 & 1.06 & 1736 @xmath111&1701 @xmath11513 & 3 36 23.4 & -35 35 37.2&21.51 & 0.91 & 22.2 & & 1247 @xmath117&14 & 3 36 24.5 & -35 35 36.8&21.17 & 1.37 & 22.1 & 1.65 & 1260 @xmath113 & 15 & 3 36 23.2 & -35 34 39.3&21.26 & 1.04 & 22.0 & 1.26 & 1523 @xmath117&16 & 3 36 25.0 & -35 34 31.5 & & & 22.1 & 1.56 & @xmath119 & 17 & 3 36 26.3 & -35 34 20.7&21.55 & 1.14 & 22.4 & 1.33 & 866 @xmath120&18 & 3 36 31.4 & -35 35 05.9&21.32 & 1.08 & 22.3 & 1.50 & 1688 @xmath121&19 & 3 36 34.5 & -35 34 52.2&21.41 & 1.29 & 22.3 & & 1150 @xmath122&20 & 3 36 28.5 & -35 33 17.0&21.63 & 1.13 & 22.3 & 0.88 & 1374 @xmath123&21 & 3 36 35.7 & -35 34 24.6&21.15 & 1.09 & 22.0 & 1.37 & 1194 @xmath124 & 1062 @xmath115 & & & & & & & & ngc 1399 ( d ) & 3 36 34.4 & -35 36 45.0 & 9.59 & 1.25 & & & 1447 @xmath125 & c r r r r r r r r r 1 & 0.072 & 0.005 & 0.067 & 0.052 & 0.040 & 0.079 & & & + 2 & 0.319 & 0.170 & 0.214 & 0.098 & 0.091 & 0.115 & & 0.154 & 0.030 + 3 & 0.111 & 0.003 & 0.089 & 0.035 & 0.022 & 0.058 & 0.060 & & + 5 & 0.094 & 0.020 & 0.071 & 0.066 & 0.059 & 0.082 & 0.096 & 0.042 & 0.033 + 6 & 0.129 & 0.038 & 0.127 & 0.021 & 0.041 & 0.088 & & & 0.026 + 7 & 0.228 & 0.078 & 0.182 & 0.053 & 0.090 & 0.060 & 0.114 & 0.174 & 0.028 + 9 & 0.174 & 0.059 & 0.123 & 0.068 & 0.072 & 0.080 & 0.136 & 0.063 & 0.027 + 10 & 0.066 & 0.022 & 0.034 & 0.028 & 0.037 & 0.098 & 0.081 & & 0.017 + 11 & 0.179 & 0.051 & 0.129 & 0.061 & 0.056 & 0.100 & & & 0.029 + 12 & 0.032 & 0.040 & 0.076 & 0.045 & 0.023 & 0.114 & 0.079 & 0.065 & + 13 & 0.066 & 0.016 & 0.067 & 0.044 & 0.008 & 0.108 & 0.096 & & 0.017 + 14 & 0.339 & 0.153 & 0.227 & 0.099 & 0.083 & 0.080 & & 0.283 & 0.034 + 15 & 0.114 & 0.026 & 0.097 & 0.064 & 0.051 & 0.088 & 0.103 & & 0.023 + 17 & 0.210 & 0.058 & 0.160 & 0.044 & 0.048 & 0.102 & 0.169 & 0.071 & 0.036 + 18 & 0.181 & 0.050 & 0.156 & 0.079 & 0.067 & 0.087 & & & 0.039 + 19 & 0.244 & 0.078 & 0.156 & 0.068 & 0.085 & 0.084 & & 0.168 & 0.062 + 20 & 0.122 & 0.032 & 0.101 & 0.048 & 0.055 & 0.082 & & & 0.048 + 21 & 0.168 & 0.051 & 0.129 & 0.077 & 0.040 & 0.078 & 0.182 & & 0.055 + & & & & & & & & & + ngc 1399 ( a ) & 0.371 & 0.201 & 0.202 & 0.081 & 0.082 & 0.052 & 0.197 & 0.261 & 0.059 + & @xmath126 & @xmath127 & @xmath128 & @xmath127 & @xmath129 & @xmath130 & @xmath131 & @xmath130 & @xmath127 + ngc 1399 ( b ) & 0.288 & 0.186 & 0.113 & 0.038 & & 0.045 & 0.223 & 0.122 & + & @xmath132 & @xmath133 & @xmath134 & @xmath135 & & @xmath136 & @xmath137 & @xmath138 & + c rrrrrrrr cc 1 & -1.51 & -1.04 & -1.74 & -1.27 & & & -1.19 & & -1.27 & -1.30 + 2 & 0.18 & -0.10 & 0.41 & 0.80 & & -0.14 & & & 0.04 & 0.12 + 3 & -1.11 & -1.37 & -1.78 & -0.96 & -1.76 & & -1.17 & & -1.24 & -1.29 + 5 & -1.28 & -0.75 & -1.42 & -1.21 & -1.36&-1.75 & -0.95 & -1.13 & -1.01 & -1.12 + 6 & -0.94 & -1.65 & -1.05 & -0.42 & & & & & -1.29 & -1.20 + 7 & -0.27 & -1.01 & -0.23 & 0.35 & -1.16 & 0.15 & -1.03 & & -0.64 & -0.55 + 9 & -0.48 & -0.71 & -0.63 & -0.49 & -0.90 & & -0.45 & & -0.59 & -0.60 + 10 & -1.56 & -1.52 & -1.40 & -1.74 & -1.53 & & -1.35 & & -1.54 & -1.53 + 11 & -0.53 & -0.85 & -0.79 & -0.4 & & & & & -0.69 & -0.67 + 12 & -1.90 & -1.18 & -1.03 & -1.13 & -1.55&-1.43 & & & -1.54 & -1.47 + 13 & -1.56 & -1.19 & -1.51 & -1.26 & -1.35 & & -1.42 & -1.24 & -1.37 & -1.37 + 14 & 0.26 & -0.08 & 0.26 & 0.99 & & 1.72 & & & 0.09 & 0.30 + 15 & -1.09 & -0.78 & -1.30 & -0.85 & -1.28 & & & & -0.93 & -0.98 + 17 & -0.38 & -1.21 & -0.65 & 0.04 & -0.52 & -1.33 & -0.73 & -0.89 & -0.79 & -0.81 + 18 & -0.42 & -0.50 & -0.81 & -0.01 & & & & & -0.46 & -0.45 + 19 & -0.17 & -0.72 & -0.25 & -0.02 & & 0.07 & & & -0.40 & -0.32 + 20 & -1.00 & -1.12 & -1.19 & -0.79 & & & -0.99 & & -1.06 & -1.04 + 21 & -0.54 & -0.53 & -0.79 & -0.38 & -0.39 & & & & -0.54 & -0.54 + & & & & & & & & & & + ngc 1399 ( b ) & 0.52 & -0.44 & 0.85 & 1.19 & -0.21 & 1.41 & & & 0.04 & 0.26 + | we report moderate resolution , high signal to noise spectroscopy of globular clusters around ngc 1399 , the central cd galaxy in the fornax cluster .
we address issues as diverse as element abundances of globular clusters versus stellar populations in ellipticals , blue horizontal branches in metal
rich globular clusters , broad band colors as metallicity tracers , possible overestimation of the age metallicity degeneracy in globular clusters , and dark matter in the halo of ngc 1399 .
we obtained spectra for 21 globular cluster candidates with multi - slit spectroscopy using the lris on the keck 1 telescope .
our sample turned out to include 18 globular clusters , 1 star , and 2 low redshift late - type galaxies ( @xmath0 ) .
the mean velocity of our globular cluster sample is @xmath1 km@xmath2s@xmath3 and its velocity dispersion @xmath4 km@xmath2s@xmath3 .
both are slightly lower than , but in agreement with , previously derived values .
we derive a mass of 1 to @xmath5 within 28 kpc for the galaxy , and a @xmath6 ratio of @xmath7 or @xmath8 @xmath9 , depending on the mass estimator .
both estimates indicate that dark matter dominates the potential at 6 r@xmath10 .
the derived element abundances for the globular clusters span the entire range observed in the milky way and m31 , with a mean metallicity of our sample around @xmath11fe / h@xmath12 \simeq -0.8 $ ] dex .
this implies that the two major sub populations reported from photometry could have formed by the same processes as the ones that formed halo and disk / bulge globular clusters in the local group spirals .
two globular clusters ( that we associate with a group of very red globular clusters , representing about 5% of the total system ) clearly stand out and exhibit metal abundances as high as observed for stellar populations in giant ellipticals .
in addition , they display surprisingly high h@xmath13 and h@xmath14 indices that are not explained by any age / metallicity combination of existing models .
the high mg and h@xmath13 values in these clusters could , however , be explained by the presence of blue horizontal branches . finally , we find that v@xmath15i and metallicity are well correlated in the globular cluster system , but that the slope of the relation is twice as flat at high metallicities as an extrapolation from the relation for milky way globular clusters .
this implies that the mean metallicities of globular cluster systems in ellipticals are lower , and cover a smaller range , than previously derived from broad band v@xmath15i colors . |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.