K-Theory and the Quantization Commutes with Reduction Problem
2014-06-04NigelHIGSONYanliSONG
Nigel HIGSON Yanli SONG
1 Introduction
The purpose of this paper is to examine the quantization commutes with reduction phenomenon from the point of view of topologicalK-theory andK-homology.The phenomenon,which concerns the quantization of Hamiltonian actions of compact groups on compact symplectic manifolds,was discovered by Guillemin and Sternberg(see[16])and by now it has been extensively studied;see for example[24]or[14,Section 8.10]for summaries.The usual contexts are symplectic geometry or Kahler geometry or geometric invariant theory,but even the topological point of view that is the concern of this paper has received a good deal of attention,and much of what appears below can be found in one form or another in earlier works.Our new contribution is an explanation,in the later sections of the paper,of the role Bott periodicity and the Weyl character formula in the transition from the commutative to the noncommutative cases of the quantization commutes with reduction problem.In addition,perhaps the paper will help to introduce the quantization commutes with reduction to a new audience in noncommutative geometry.
Here,in brief,is an overview of the phenomenon as seen from aK-theoretical perspective.Denote byDthe Dolbeault operator on a closed complex(or almost-complex,or stable complex)manifoldM.IfEis any line bundle(or indeed any vector bundle at all)onM,then we can coupleDtoEso as to form a new elliptic operatorDE.The Fredholm index ofDEis an integer,of course,but if a compact connected Lie groupGacts onMandE,then this integeris the virtual dimension of a more refined index,which is a virtual representation

The trace of this virtual representation can be calculated using the fixed-point methods of Atiyah,Bott and Segal.For example,ifMis the Riemann sphere C∞,equipped with the natural rotation action ofU(1),then the index is given by the formula

The two terms represent contributions from the two fixed-points,and the integersw0andw∞are the weights of the representations ofU(1)on the fibers ofEover these points.
In favorable circumstances there is however an entirely dif f erent formula that gives not the character but the multiplicities of individual irreducible representations ofGin the index.The new formula is not organized around fixed-point data at all.
We will concentrate on the multiplicity of the trivial representation.Suppose that∇is aG-invariant connection onEthat is compatible with some Hermitian structure.The Kostant moment map

associated to∇is defined by

whereX∈g,LXdenotes the infinitesimal action ofXon sections ofE,and∇Xis covariant dif f erentiation in the direction of the Killing vector field associated toX(the dif f erence of these two Differential operators is multiplication by a smooth function).Assume that 0∈g∗is a regular value of the Kostant moment map,and thatGacts(locally)freely on the submanifold

In the Kahler or symplectic contexts,these two conditions are equivalent,but in general we can simply assume that both conditions hold,and then we can form the reduced manifold(or orbifold if the action is only locally free)

It is also a(stable)complex manifold,and the line bundleEdescends a line bundleE0onM0.The quantization commutes with reduction formula is the rather remarkable identity

The left side is the multiplicity of the trivial representation ofGin Index(DM,E)and the right side is the integer index of the associated Dolbeault operator on the reduced manifold.The formula is striking in that it is nearly as far from a formula of fixed-point type as it could possibly be.
For example,on the Riemann sphere,ifthen the reduced manifoldM0is a point and Index=1.So a quick check shows that the formulas(1.1)and(1.3)areconsistent with one another.If the signs are reversed,thenis a point with the opposite stable complex structure and Index,so the formulas are still consistent.
Actually,the formula(1.3)is not quite correct in the general context that we are considering,which reaches beyond symplectic manifolds.We shall begin by considering abelian group actions on stable complex manifolds(or orbifolds)and review some fairly well-known arguments that lead to a proof that(1.3)is correct for these.But in the nonabelian case the best available result involves tensor powers of the line bundleE,and asserts that

refer to[21]or[15],for example.This is our objective,and as we stated at the beginning the main goal of this paper is to explain how the nonabelian result follows rather easily from a slightK-theoretic strengthening of the abelian result(applied to the maximal torus),thanks to Bott periodicity and the Weyl character formula.
Necessary preliminaries onK-theory and stable complex manifolds are set out in Sections 2–4.In Section 8 we review(and generalize very slightly)the proof of the formula(1.3)for circle actions.To go further we need to review some facts about orbifolds,and this is done in Sections 10–11.This allows us to pass from circles to tori in Section 12,and then to general groups in Section 13.Finally,many index-theoretic arguments become a bit more conceptual when set within the context ofK-homology,and we shall do this for the quantization commutes with reduction phenomenon in the more or less independent Section 9.
2 Complex Manifolds and K-Theory
IfXis a compact Hausdorf fspace,then as usual we shall denote byK(X)the Atiyah-HirzebruchK-theory group ofX,or in other words the Grothendieck group generated from the semigroup of isomorphism classes of complex vector bundles onX.IfXis a locally compact Hausdorf fspace,then we shall denote byK(X)the Atiyah-HirzebruchK-theory group“with compact supports”.This is the reducedK-theory of the one-point compactification ofX.
Individual vector bundles do not determine elements inK-theory whenXis not compact,but bounded complexes of vector bundles

do determineK-theory elements whenever the support of the complex(the support is the smallest closed subset ofXoutside of which the complex is fiberwise exact)is compact.See for example[6]for this construction,which in fact accounts for allK-theory elements.
The most important instance of the construction ofK-theory elements from complexes is that of the Bott element of ann-dimensional complex vector spaceV,which is the element inK(V)determined by the complex

of vector bundles with the indicated constant fibers,for which the Differentials atare given by contraction withv.The support of this complex is
The Bott element can be generalized,as follows.IfVis a complex vector bundle over a locally compact spaceX,then by applying the Bott element construction to the fibers of theVwe obtain a complex of vector bundles over the locally compact spaceVwhose support isthe zero section.If the original base spaceVis compact,then we obtain an element ofK(V),called the Thom element.IfXis noncompact,then we obtain a Thom homomorphism

It maps the element ofK(X)associated to a compactly supported complexE∗overXto the element ofK(V)associated to the tensor product of the pullback ofE∗toVwith the“Thom” complex just described(the tensor product complex is compactly supported).The Thom homomorphism is an isomorphism.
For our purposes,the most important feature ofK-theory is the existence of wrong-way maps

associated to continuous mapsf:M1→M2between,for now,almost-complex smooth manifolds(we shall consider broader classes of spaces in due course).The construction,which was carried out by Atiyah and Hirzebruch in their earliest works onK-theory(see[3],but see for example[18]for a complete treatment),is functorial and homotopy-invariant.As a result,the construction for smooth embeddings determines the construction in general,since we can factor any map,up to homotopy as a composition

whereVis a finite-dimensional complex vector space,the first map is a smooth embedding and the second is the projection,which is homotopy inverse to a smooth embedding.As for embeddings,the mapf∗is defined as follows in the special case where the normal bundleNM1admits a complex vector bundle structure for which there is an isomorphism

of complex vector bundles overM1(this is the only case that shall concern us,and in any case the general case quickly reduces to this one).The wrong-way map is defined to be the composition

of the Thom homomorphism with the natural map onK-theory groups associated to the inclusion of the normal bundleNM1as an open tubular neighborhood inM2.
3 Stable Complex Structures
Recall that a stable complex structure on a real vector bundleVis a complex vector bundle structure on some direct sum(the first summand is the trivial bundle of rankk).Two stable complex structures onV,given by complex structures onandare equivalent if there exist trivial complex vector bundlesandand an isomorphism of complex vector bundles

The definition of stable complex structure may be applied to the tangent bundle of a smooth manifold:a stable complex structure on a smooth manifold is a stable complex structure onTM,and a stable complex manifold is a smooth manifold equipped with an equivalence class of stable complex structure.
The theory of wrong-way maps sketched in the last section extends to stable complex manifolds:Any smooth,or indeed continuous map between even-dimensional stable complex manifolds induces a wrong-way map

with the same homotopy invariance and functoriality properties as before.1Moreover,the restriction to even-dimensional manifolds can be dropped if one introduces the odd K-theory group as well.One way to see this is to note that a stable complex structure on an even-dimensionalMis the same as an almost-complex structure on someNow apply Bott periodicity.2This requires that we identify R2kwith Ck,which we do by mapping (x1,···,x2k)to(x1+ix2,···,x2k−1+ix2k).
4 Compact Groups
LetGbe a compact Lie group.Everything that we have just discussed generalizes to equivariantK-theory(see[23]),defined by using equivariant complex vector bundles over compact or locally compactG-spaces(and indeed it is not necessary for these things to assume that the compact groupGis a Lie group).
There is one detail worth spelling out:In this paper we shall assume that the trivial bundles appearing in the definition of stable complex structure are trivial as equivariant bundles:The action ofGon the constant fiber is the trivial action(for other purposes it would suffice to assume that equivariant Bott periodicity holds for these trivial bundles).This has the following useful consequence.
Lemma 4.1Let M be a G-equivariant smooth manifold and let

be a G-equivariant complex structure,giving M to be the structure of a G-equivariant stable complex manifold.Let g be an element of G and let F⊆M be the fixed manifold of g.
(a)The complex structure J over F restricts to a complex structure

giving F the structure of a stable complex manifold.
(b)The complex structure J induces a complex structure on the normal bundle

ProofThe subbundleis precisely the subbundle on whichGacts trivially in each fiber.By equivariance it is invariant underJ,and so(a)is proved.Item(b)is an immediate consequence of(a).
Wrong-way maps in equivariantK-theory are defined for continuous,equivariant maps between smooth,equivariant,stable complex manifolds.They are compatible with wrong-way maps in the non-equivariant theory in a number of ways.For instance ifis a mapbetween stable complex manifolds on whichGacts trivially,then there is a commuting diagram

Figure 1
with the vertical maps given by tensor product of bundles(viewed upstairs as bundles with trivialG-action)with representations(viewed as equivariant bundles pulled back from a onepoint space).Incidentally,the vertical maps are isomorphisms(see[23,Proposition 2.2]).
5 Quantization
Of special interest in the theory of wrong-way maps is the collapse of a complex or evendimensional stable complex manifold to a point,

According to the Atiyah-Singer index theorem,if we identifyKG(pt)with the representation ringR(G),and ifEis a smooth,G-equivariant complex vector bundle over a closed almostcomplexG-manifoldM,then the wrong-way map

takes the class3We will abuse the notation slightly and write E for both the bundle and its K-theory class.to the Fredholm index of the Dolbeault operator onMcoupled toE:

In recognition of the special role played by the wrong-way map,we shall introduce the following notations and terminologies.
Definition 5.1Let M be a smooth,even-dimensional,G-equivariant almost complex manifold and let α∈KG(M)be any K-theory class.Let p:M→ptbe the map that collapses M to a point.We shall call the class

the quantization of the pair(M,α),and denote it by

Let us now formulate the simplest instances of the theorems that we wish to prove in this paper.
Definition 5.2If α∈KG(M)is any K-theory class,then denote by

the multiplicity of the trivial representation within the virtual representation QG(M,α)∈R(G).
Definition 5.3Assume that0∈g∗is a regular value of μand that G acts freely on thesubmanifold

Definition 5.4be any K-theory class.Assume thatis regular valueof μand assume that G acts freely on the submanifold μ−1[0]⊆M.Denote by α0∈K(M0)the class that maps to the restriction of αto μ−1[0]under the pullback isomorphism

Now assume thatMis an even-dimensional stable complexG-manifold,thatEis aG-equivariant Hermitian line bundle onM,equipped with a connection using which we define a moment mapμ,thatis a regular value and thatGacts freely onWe will note later that the reduced manifoldM0carries a natural stable complex structure.
Theorem 5.1If M is closed and G=U(1),then

for all k.
Theorem 5.2Whether or not M is closed and whether or not G=U(1),if α∈KG(M)is any K-theory class,then

for all sufficiently large k>0.
6 A Bordism Construction
The definition of stable complex manifold applies equally well to manifolds with boundary,and every stable complex structure on a smooth manifoldWinduces one on∂Wthanks to the isomorphism

determined by an outward-pointing normal vector field on the boundary.
An important property of the wrong-way map construction in the context of stable complex manifolds,and hence of quantization,is its bordism invariance:IfMis the boundary ofW,and ifextends toW,then the composition

in which the first map is restriction to the boundary,is the zero map.In particular,we have the following proposition.
Proposition 6.1Ifis the boundary of a compact stable complex manifold W,and α1and α2are the restrictions M1and M2of a K-theory class on W,then
This may be applied in the following context.For the rest of this sectionG=U(1),letMbe smooth,closedG-equivariant stable complex manifold,and letEbe aG-equivariant smooth Hermitian line bundle overM.Let

be the Kostant moment map associated to some Hermitian connection onE,as in the introduction,where the Lie algebra generator

is chosen so that

for allIt follows from(1.2)that ifis fixed byG,then the action ofGon the fiberis given by the formula

So the valueμ(m)is an integer,namely the weight of the representation ofU(1)on the lineEm.Conversely,ifandμ(m)is non-integral,thenmis not fixed byG,so the action is locally free near theG-orbit ofm.Thus,we have the following lemma.
Lemma 6.1The G-action on M is locally free everywhere except on the integral level sets of μ.
Choose real numbers such that

and such thataandbare regular values ofμ(the exact values are not important and make no essential dif f erence in the construction that follows).By the lemma,the groupGacts locally freely onIn fact we shall assume in this section that the action is free on these level sets and the others in between(the locally free case will be considered later).
We are going to construct a bordism of stable complexG-manifolds fromMto a disjoint union of two manifoldsThe two manifolds are easy to describe topologically:

TheG-actions on these manifolds are the obvious ones inherited fromM.Moreover on the complements of the collapse regions the manifolds agree with the corresponding open subsets ofMas stable complexG-manifolds.The complementary closed sets

are fixed-set components inandrespectively.So according to Lemma 4.1,they acquire fromandthe structure of stable complex manifolds in their own right,aswell as submanifolds with complex normal bundles.The normal bundles may be described as follows.

whereU(1)acts onthrough theG-action,and onCthrough the weight−1 action,whileGacts only on the left hand factor of the product.The same holds forMb.
As for the fixed-set stable complex structures onMaandMb,they may be described as follows.Ifcis any regular value ofμ,and ifGacts freely on the level setthen we can form the“reduced”manifoldMc=μ−1[c]/Gand equip it with a stable complex structure through the identification

that maps 1∈R to anyG-invariant vector fieldYdefined onMnearwithY(μ)>0(this gives-invariant stable complex structure),and the identification

that maps 1∈R to theG-invariant vector field ondetermined by the Lie algebra generatorin(6.2).

Let us now construct the bordism.Define

Herekis any value that is less than the minimum value of the function(μ−a)(μ−b).With this choice ofk,the function

is regular onM×C at the values 0 andk,soPis indeed a smooth submanifold with boundary ofM×C.
Equip C with the weight−1 action ofU(1).4Other choices of action are possible,and give useful variations on the construction(see Section 9).The diagonal action ofU(1)onM×C is free onP(this uses our hypothesis that the action ofGonMis free betweenμ=aandμ=b),so the quotient

is a smooth manifold with boundary.The action ofGon the first factor ofM×C alone descends to an action ofGonW.The manifoldPinherits aG-andU(1)-invariant stablecomplex structure fromM×C.The quotient bundleTP/GoverWinherits aG-invariant stable complex structure fromTP.Now use the identification

that maps the trivial summand R to the subbundle ofTP/U(1)spanned by the vector fieldXgenerating the diagonalU(1)-action to equipWwith a stable complex structure.The stable complexG-manifoldWso-obtained will be our bordism.
Consider the boundary component ofPwith

By the definition of the constantkthat appears here,if(m,z)is a point in this boundary component,thenThe map

where the dot denotes the action ofG=U(1)onM,identifies the corresponding boundary component of the quotientWwith the stable complexG-manifold−M.
Consider next the boundary component ofPwith

It separates into two parts:(1)μ≤a,(2)b≤μ.Write the corresponding boundary component ofW=P/U(1)as

accordingly.
The open subset ofwithidentifies as aG-stable complex manifold with the similar part ofMvia the map(6.10),and the complement,the closed submanifold withμ=a,identifies with the smooth manifold

via the projection fromM×C toM.Nearμ−1[a]⊆M,the manifoldMhas a product form

in such a way thatμis the projection onto the R-factor.We find from this that

nearμ=avia the projection

The structure of the normal bundle can be determined from this.Similar remarks apply to
Now suppose thatWe can think ofαas a class

by makingU(1)act through its identification withG.Then we can use the projection map from the manifoldPin(6.9)toM(which is a proper map)to pull backαto a class inFinally,since the action ofU(1)onPis free,we can use the pullback isomorphism

to obtain a class infromα.Its restriction to the boundary componentMis the original classα.We will use the same notationαfor the restrictions toand we arrive at the following formula from the bordism invariance of quantization:

As was pointed out by Duistermaat et al[13],this may be used to give very simple proofs of Theorems 5.1–5.2 in the caseG=U(1).5The arguments also have a lot in common with the works of Lusztig and Kosniowski(see[20])and Atiyah and Hirzebruch(see[4]).These authors worked in the holomorphic context,and Duistermaat et al worked in the symplectic context,but the arguments for stable complex manifolds are the same.The argument uses the well-known fixed-point formulas inK-theory that we shall review in the next section.
7 Fixed-Point Formulas
We shall assume thatGis a torus.We review the basic fixed-point formulas in equivariantK-theory with a view to extend them a little in subsequent sections.
Definition 7.1Denote by R(G)gthe localization of the character ring R(G)at g∈G,or in other words,the ring of fractions obtained by inverting all virtual representations whosecharacter at g is non-vanishing.Similarly,denote bythelocalization of the R(G)-module KG(X).
Denote by F⊆M the fixed set of g(which is the same as the G-fixed set since g∈G is a topological generator).As explained in Lemma4.1,the fixed-set is a submanifold of M and it carries a canonical stable complex structure.Denote by ι:F→M the inclusion.
Theorem 7.1(see[23,Proposition 4.1]or[5,Theorem 1.1])The restriction map

in localized equivariant K-theory is an isomorphism.
Definition 7.2If H is any G-equivariant complex vector bundle over a compact G-space X,then write

The relevance of this class to the present discussion is that the composition

is given by multiplication byλ(NF),whereNFis the normal bundle ofFinM.This is evident from the definition of wrong-way maps in Section 2.
Theorem 7.2(see[5,Lemma 2.7])If NF is the normal bundle of F in M(with the complex structure given in Lemma4.1),then the element λ(NF)is invertible in the localized equivariant K-theory ring KG(F)g.
Remark 7.1IfFis noncompact,then the statement “λ(NF)is invertible” should be taken to mean “multiplication byλ(NF)is an invertible map fromKG(F)gto itself”,where“multiplication byλ(NF)” means “the alternating sum of tensor products with the bundles Λp(NF∗)”.
The definition of quantization as a wrong-way map and functoriality of wrong-way maps now immediately give the following Atiyah-Bott-Segal formula,in which we have altered slightly our usage of the symbolF.
Theorem 7.3(Compare[5,Proposition 2.8])

for every α∈KG(M).Here the sum is over the connected components of the fixed set in M.
Remark 7.2We could of course consider the fixed set as a single manifold,as we have done up to now,rather than as the union of its components.But calculations are usually done component-wise,so the above formulation of the fixed-point theorem is usually the most convenient.
8 Quantization Formula for Circle Actions
Our goal in this section is to prove Theorems 5.1–5.2 forG=U(1)following the method of Duistermaat et al(see[13]).The method has been repeated,with variations,several times in the literature(see for example[12,22]).Our purpose in repeating it one more is to record enough details to determine how the method generalizes to the contexts that we shall consider in the coming sections.
To begin with,we shall calculate in the case of a stable complexG-manifoldMand equivariant line bundleEfor which over each component of theG-fixed set inMthe groupG=U(1)acts on line bundleEwith non-negative weight.Obviously the formula(6.12)makes this assumption relevant to our ultimate goal,and we shall come back to the relevance of(6.12)after handling the special case.
Consider the ringof rational Laurent series with finite singular parts.Becauseis in fact a field,the inclusion

extends to a homomorphism

and it is in the Laurent series ring that we shall do our calculations.The advantage of doing so is that we can speak of the “coefficient ofz0” in the contribution of each component of the fixed set(which is an element ofR(G)g)to the overall quantizationQ(M,Ek·α)(which is an element ofR(G)).
Lemma 8.1If F is a component of the fixed set in M,and if G acts with positive weight on the fibers of the line bundle E over F,then the coefficient of z0in

is zero for all.When α=1the coefficient is zero for all k>0.
ProofTo keep the notation simple we shall consider the caseα=1 andk=1.The other cases are easy variations.
SinceGacts trivially onFthe natural map

is an isomorphism.So for eachs∈Z we can speak of the coefficient ofzsin any elementγ∈KG(F),and define the valuation ofγto be the leastsfor which the coefficient is nonzero.With this terminology,our aim is to show that

The key point is that the valuation ofλ(NF)is nonpositive:

and moreover the corresponding leading coefficient is a unit in the ringK(F).
According to Theorem 7.2 there is someβ∈KG(F)and an elementγ∈R(G)(whose character is non-vanishing atg∈G)such that

Since the leading coefficients ofλ(NF)andγare units,we have that

so

Now

and

where the first inequality is a result of Figure 1.Our assumption onEcompletes the proof.
Now suppose thatGacts with weight zero on the fibers ofEover some componentFof the fixed set.Lemma 8.1 does not apply,but a dif f erent hypothesis allows us to reach the same conclusion.Namely assume that the normal bundle is a line bundle of weight 1.6Actually it would be enough to have such a summand in the normal bundle.Then since

we see that val(λ(NF))=−1,which is a bit better than(8.1).The argument in the lemma shows thatonce again.

We now return to the bordism constructed in Section 6 and the formula(6.12).The above arguments show that

and also

when
We can analyzeM≤ain a similar way,but using the fieldof Laurent series with only finitely many positive order terms,and the valuation that indicates the highest nonzero coefficient instead of the lowest.We find that ifFis a component of the fixed set whereGacts onEwith a negative weight,thenFdoes not contribute toQG(M≤a,Ek·α)0.
Let us assume now that(as in the statements of Theorems 5.1–5.2)Gacts locally freely onμ−1[0].Then the only component of theG-fixed set inM≤aremaining to be analyzed isF=Ma.This does contribute toQG(M≤a,Ek·α)0.To calculate what it contributes,use the Neumann series with remainder

The previous arguments show that when we substitute this into the quantity

only the constant term 1 contributes to thez0term(for the last term,useNF−nin place ofEin the argument used in the proof of Lemma 8.1).As a result

and similarly

for
These calculations complete the proofs of Theorem 5.1 and theG=U(1)case of Theorem 5.2.If 0 is a regular value ofμ,and ifGacts freely onμ−1[0],then for sufficiently smalla,Mais bordant toM0,indeed isomorphic to it.So we can substituteM0forMain the formulas above.
Remark 8.1Denote bythe vector space of doubly infinite Laurent series.This is no longer a ring,of course,but it contains both of the ringsthat we used above.In the situation where 0 is a regular value ofμandGacts freely onμ−1[0],we may use this fact to define an alternative quantization

as follows:For each componentFof the fixed-point set,expand its contribution toQG(M,E)(via the fixed-point formula)in eitheraccording asGacts on the fibers ofEoverFby a positive or negative weight.Then add up these contributions into obtain(8.3).
The quantization is “local” in the sense that it is explicitly localized near the fixed set of theG-action.Using Lemma 8.1,one can see that no fixed set contributes to the coefficient ofz0at all!On the other hand,the calculations above relate the local and original quantizations through the following attractive formula:

HereL0is the line bundleoverM0associated to the weight 1 action ofGon C(viewed as an ordinary line bundle,with no group action).
Of course,the quantization commutes with reduction formula can be immediately deduced from(8.4),and to some extent the formula “explains” the relevance of reduced manifolds likeM0to index theory and fixed-point theory.Therefore,it is an interesting problem to derive counterparts of(8.4)for other groups.
9 Some Remarks on K-Homology
In this section we shall briefly examine the quantization commutes with reduction problem from the perspective ofK-homology.
K-homology is the dual theory to(equivariant)K-theory,and it is related toK-theory by functorial pairings

(hereKGdenotes equivariantK-theory,as in the rest of this paper,andKGis equivariantK-homology).In factK-homology is characterized by these pairings in the sense that for a fixedXevery transformation

that is natural inYand aKG(Y)-module homomorphism determines a class in theK-homology groupKG(X),and conversely.7This holds for reasonable spaces X at least,for instance,the smooth closed manifolds that we shall be considering.See[10]for more on this point of view onK-homology.
Along with the above abstract definition there are concrete constructions ofK-homology.A functional-analytic definition was proposed by Atiyah[1]and worked out in detail by Kasparov(see[19]).An element of the analytic groupis determined by an equivariant Fredholm operator on aG-Hilbert space.The Hilbert space is required to carry an action of the algebra of continuous functions on the compact spaceX,and the operator should be compatible with this action in a certain sense.See[17]for a complete treatment of Kasparov’s theory,at least in the non-equivariant case.
There are also geometric definitions.In the approach of Baum(see[8]),an element of the geometric groupis determined by a closed,even-dimensional stable complex(or,more generally spinc)G-manifoldM,together with an equivariant complex vector bundleEonMand a continuous mapffromMtoX.A natural transformation

is defined by associating to(M,E,f)a Fredholm operator obtained from the∂E-operator(the algebraXacts on the associated Hilbert space ofL2-forms onMthroughf,which determines a homomorphism fromC(X)toC(M)).
For a full treatment of the natural transformationQin the non-equivariant case(see[9]).The further issues in the equivariant case are considered in[7].
As our notation suggests,we want to view the natural transformation as a quantization map,and indeed whenXis a point,it may be identified as such,thanks to the index theorem.But what about reduction?
View the quotient mapX→X/Gas aG-equivariant map to a space with a trivialG-action.It induces a map in equivariant homology

(either analytic or geometric),and because theG-action is trivial on the right-hand side there is a standard isomorphism

which is analogous to the similar isomorphism inK-theory.Combining the two displays we obtain reduction homomorphisms in geometric and analyticK-homology

Simply by naturality of the transformationQ,these fit into a“quantization commutes with reduction diagram”

At this stage there is no“quantization commutes with reduction problem”since the diagram automatically commutes!The problem arises when one asks for explicit formulas for the reduction maps.
In fact there is a simple,concrete formula for the analytic reduction map:Take an equivariant Fredholm operator on aG-Hilbert space,and decompose it and the Hilbert space into theirG-isotypical parts:

Note that each isotypical componentHσ,although not aC(X)-module,is aC(X/G)-module.
IfXis a point,then there is an isomorphism that maps a Fredholm operatorto its Fredholm index.The composition of quantization followed by reduction in the diagram is computed from a geometric cycle,and the multiplicities with which the irreducible representations ofGoccur in the index of the associated elliptic operator.This is half of the quantization commutes with reduction formula.
Unfortunately,there seems to be no general formula for the geometric reduction map:The problem is that the easily defined isomorphism(9.1)needs to be inverted,and there is in general no known way to do this in the geometric theory.
However,the methods of the previous section may be extended very slightly to handle some cases whenG=U(1).
Assume thatMis a closed,even-dimensional stable complexG-manifold and thatEis a smooth equivariant Hermitian line bundle onM.Define,as usual,a moment map by applying Kostant’s formula to a connection onE.Assume that there are regular values±anear 0∈R so that the action ofGis free on the level sets
Construct a manifoldM0by starting from the region ofMwith−a≤μ≤aand collapsing each of the orbits withμ=±ato points.We might also write

in the notation of Section 6.
The detailed construction ofM0can be carried out in two stages.The first one is exactly described in Section 6 and results inThe second one is similar,usingas input,except that we cut at−a,of course.But in addition,during the construction described in Section 6 we equip the second factor inwith the weight+1 action ofU(1)rather than the weight−1 action.
The ef f ect of this is that the normal bundle ofcarries a weight+1 action ofG,rather than a weight−1 action as before.As for the fixed-set stable complex structure onit is the opposite of the stable complex structureM−areceived as a reduced manifold ofM.
We have the following theorem.
Theorem 9.1Given a G-equivariant geometric cycle(M,E,f),the multiplicity of the trivial representation in the class

is the K-homology class of the triple(M0,E0,f0),where M0is as above,E0is the line bundle on M0obtained from E,and f0:M0→X/G is the continuous map induced from f.
To make the theorem plausible,it should be noted that geometric cycles related by a bordism(see[8–9])determine the same geometricK-homology class.The theorem is proved by using this and essentially the same fixed-point techniques used in the previous section(so there is nothing new in the approach).
Theorem 9.1 is related to the usual statement of quantization commutes with reduction in the following way.If the action ofGis free throughout the region−a≤μ≤a,thenM0can be constructed more simply as a fiber bundle

over the reduced manifoldM0,with complex projective fibers.Now it is built into Baum’s geometric theory that if one cycle is a bundle over the other with CP1fibers(with a compact structure group preserving the complex structure on the projective space),then the two cycles determine the same cohomology class(see[8–9]).So in this instance we can replacein the above theorem withand thus recover the usual quantization commutes with reduction result.
A final remark:If we make further assumptions about the freeness of theG-action on sufficiently many level sets(or pass to orbifolds),then we can calculate all the multiplicities ingeometrically by shifting the construction above to otherWe obtain a sort of“Fourier series”

in the group
10 Orbifolds
As we move fromto other groups it will become helpful to drop the hypothesis thatGacts freely onand assume only that the action is locally free.The reason is that we shall work inductively on the rank ofG,and it is much easier to keep track of local freeness during the inductive stages than to keep track of freeness.
To accommodate the change we shall work with orbifolds.The following simplified version of the general definition is the most convenientfor our purposes(compare[25]).
Definition 10.1An orbifold structure on a locally compact Hausdorf fspace M is a triple

consisting of a compact Lie group H,a smooth manifold P that is equipped with a locally free action of H,and a homeomorphism q:P/H→M.
We shall also use the term orbifold,meaning a space with an orbifold structure.The dimension of an orbifoldis the dimension ofPminus the dimension ofH.
Definition 10.2Two orbifold structures(H1,P1,q1)and(H2,P2,q2)on the same space M are equivalent if there is a third orbifold structure of the form(H1×H2,P,q)and a commuting diagram

Figure 2
in which
(a)the individual actions of H1and H2on the manifold P are free,not just locally free,
(b)the maps p1and p2are H1-equivariant and H2-equivariant,respectively,and
(c)these maps identify P1and P2with the quotient manifolds P/H2and P/H1,respectively.
Example 10.1If=(H,P,q)is an orbifold structure onM,and the action ofHonPis free,not just locally free,then the mapq:P→MidentifiesMwith the smooth manifoldH/P,andis equivalent to the trivial orbifold structure({e},M,id)onM.
Definition 10.3The orbifold K-theory of an orbifold structure=(H,P,q)is the equivariant K-theory group

Equivalences between orbifold structures determine isomorphisms between orbifoldK-theory groups via pullbacks,as in the following diagram:

We define stable complex orbifolds as follows.
Definition 10.4A stable complex structure on an orbifold=(H,P,q)is an H-equivariant stable complex structure on the quotient bundle TP/h,wherehdenotes the subbundle of TP tangent to the action of H.
Next,it is easy to generalize all these definitions to the equivariant context,in which a compact groupGacts onM.For instance in aG-equivariant orbifold structure=(H,P,q)on aG-spaceM,the manifoldPis required to carry a smoothG-action commuting with theH-action and the mapqis required to beG-equivariant.In addition,thenG-equivariantK-theory ofis theG×H-equivariantK-theory ofP,and aG-equivariant stable complex structure is aG×H-equivariant stable complex structure onTP/h.
We shall also use concepts such as bordism(of orbifold structures involving the same groupH)and smooth maps between orbifolds(again,in the simplest situations involving the same groupH)without further explanation.
We shall need suitable quantization maps

in the orbifold context.A geometric definition(in terms of Bott periodicity,wrong-way maps,and so on)does not appear to be known.However a quantization map(10.1)is easy to define using transversally elliptic operators(see[2]):The stable complex structure on=(H,P,q)determines aG×Hequivariant andH-transversally elliptic Dolbeault operator onP.Restricting to theH-invariant sections we obtain a Fredholm operator with index inR(G).More generally,coupling with classes in the group(P)gives us the quantization map that we need.
In any case we shall assume given quantization maps(10.1)with the following properties(that are routines to verify from the analytic definition):
(a)R(G)-linearity and independence of presentation.The quantization map isR(G)-linear and compatible with the isomorphisms in Figure 2.
(b)Compatibility with the wrong-way maps.Ifι:is aG-equivariant embedding,then the following diagram is commutative:

(c)Bordism-invariance.If(,α)is bordant tothen
(d)Trivial action.IfG=G2×G1,andG1fixes the manifoldPin an orbifold structure(P,H,q)pointwise,then the diagram

is commutative(compare Figure 1).
At the very end of the paper we shall use one additional property,borrowed from the theory of the analytic index.We shall only formulate it in the context we need,though it holds more generally.
(e)Multiplicative axiom.LetFbe a stable complexG-manifold,let,and assume that

Let=(G,P,q)be any stable complex orbifold and let.If we form the locally compact space

and the orbifold structure

then

This is a mild generalization(to orbifolds)of the multiplicative property of the index used by Atiyah and Singer(compare[6,Axiom B3],and compare[9]for a more modern proof which adapts easily to the present circumstances).
11 Fixed-Point Formulas for Orbifolds
Unless otherwise indicated,we shall assume that the Lie groups(G,H,etc.)occurring in this section are compact,connected and abelian(that is,they are tori).Our aim is to formulate and prove fixed-point formula in the orbifold context,and then use it to generalize Theorems 5.1–5.2.
The orbifold fixed-point formula is not a completely straightforward generalization of the manifold case,mostly becauseis aG-equivariant orbifold structure onM,then there is a dif f erence between being fixed byGinMand being fixed byGinP:G-fixed-points inMdo not necessarily lift toG-fixed-points inP.In what follows we shall indicate how to handle this point.At the same time we shall work in a slightly more general context than that of Section 7 by studying the fixed set of a subtorusrather than the fixed set ofGitself.
Lemma 11.1Letbe a G-equivariant orbifold structure,and letbe asubtorus.If F is the fixed set of T in M,then the setis a disjoint union of a locallyfinite collection of smooth,closed submanifolds of P,each of which is the fixed-point manifold of a torus L ProofLetg∈Tbe a topological generator.For each pointthere existssuch that the elementfixesp.The connected component of the identity in the subgroup ofG×Hgenerated by(g,h)is a torus that projects ontowith finite kernel.We call this a good torus.The fixed-point set of any good torus inis a smooth closed submanifold ofPthat is contained inand the union of all such fixed-point subsets isThe fixed-point sets for di ff erent good tori inare disjoint,because if a point was fixed by two di ff erent good tori it would be fixed by the torus generated by the pair,and this would contain a circle subgroup ofH,contrary to the assumption that the action ofis locally free.Finally,there are only finitely many distinct good tori with nonempty fixed-point sets that meet any given compact set inis a locally finite disjoint union of smooth,closed submanifolds ofP,as required. Corollary 11.1The datumis a G-equivariant orbifold structure on F. Definition 11.1LetG-equivariant orbifold structure on a G-space M,and let g∈G.The localized K-theory groupis defined by Theorem 11.1Assume that g∈G generates a torus T⊆G,and we denote by F⊆M the T-fixed set in M.The restriction map in localized K-theory is an isomorphism. ProofThe standard argument,which uses the cohomological properties of theK-theory functor,can be adapted to the current situation. It suffices to show that the localized groupis zero.CoverMFbyG-invariant open setsUsuch thathas the form where Γ is the isotropy group of a point inq−1[U]andVis some Γ-space(here we are making an exception to our general rule:Γ need not be connected).There is an induction isomorphism The ringR(G)acts on the right-hand side through the projection map from Γ toGand the associated homomorphism fromR(G)toR(Γ). The image of Γ inGcan not contain the generatorg∈T,so there is a character ofGthat is trivial on the image of Γ but nontrivial ong.From this it follows that and of course it follows in turn that Sois zero by a Mayer-Vietoris argument. Assume now that=(H,P,q)is equipped with aG-equivariant,stable complex structure.Let us continue to denote byga generator of a torusand bythe fixed set ofg.Denote bythe orbifold normal bundle,that is,the normal bundle toinP.The proof of Lemma 4.1 shows that the normal bundle has aG×H-invariant complex structure,and we denote by the class from Definition 7.2. Theorem 11.2Let=(H,P,q)be a G-equivariant,stable complex orbifold.Assume that g∈G generates a torus T⊆G,and denote by F⊆M the T-fixed set in M.The class becomes invertible when viewed as an element of the localized ring We need a small algebraic computation. Lemma 11.2Let f:Γ→G be a homomorphism between compact abelian Lie groups(Γneed not be connected),and assume that f has finite kernel.Let L be the connected component of the identity inΓand let g∈G be a topological generator of the image of L.If ϕ:Γ→U(1)is any character that is nontrivial on L,then the element(1−ϕ)∈R(Γ)is invertible in the localized ring R(Γ)⊗R(G)R(G)g. ProofThere exists some0 such thatvanishes on the kernel off.There exists then a charactersuch thatIndeed,by Frobenius reciprocity,any constituent of the induced representationwill do.Now we can write as required(the element(1−ψ)can not vanish ongbecauseφ,and henceφN,is nontrivial onL). Proof of Theorem 11.2It suffices to prove that the restriction ofto anyG×Horbitis invertible in the localized equivariantK-theory of that orbit.If Γ is the isotropy group of the orbit,then and the restriction oftakes the form where the charactersare the Γ-weights of the restriction of the normal bundle toO.IfLis the connected component of the identity in Γ,thenLis a good torus in the sense of the proof of Lemma 11.1,and the charactersφjare nontrivial even when restricted toL.The invertibility of this product follows from Lemma 11.2. With these preliminaries in place,we obtain the following fixed-point formula to orbifolds,exactly as in the manifold case(and as in the manifold case,we shall adjust our notation a bit in the formulation of the theorem below). Theorem 11.3Let=(H,P,q)be a G-equivariant,even-dimensional stable complexorbifold,and let g∈G be a topological generator of a torus,then for every,where the sum is over the components of the T-fixed set in M. Now we can formulate and prove our orbifold extensions of Theorems 5.1–5.2.Let be aG-equivariant,even-dimensional,stable complex orbifold.Assume thatGdecomposes as a product withLetbe aG×Hequivariant smooth Hermitian line bundle onP,with connection,and let be the moment map associated to theT-action by Kostant’s formula.It is of course a smooth(G×H)-invariant function.We want to examine the quantization and in particular the multiplicity of the trivial representation ofT.To this end,we construct the bordism of stable complexG-orbifolds as in Section 6.8Note that even if M itselfis a manifold,unless we explicitly hypothesize that the action of T is free betweenµ=a andµ=b(rather than locally free,which is guaranteed)the bordism will only be an orbifold bordism. Lemma 11.3 This may be proved by using the method described in Section 8 ,but first we need to make a note of one further property of quantization in the orbifold case,to be added to the list at the end of Section 10: (f)Change of groups.Ifϕ:J→G×His a homomorphism of tori,then the homomorphism fromJ×HtoG×Hdetermined byφand multiplication inHinduces aK-theory map fromIn additionφand the projection fromG×HtoGinduce a map fromR(G)toR(J).The following diagram commutes: As with the other properties in our list,this is straightforward to verify if quantization is defined by using the index of transversally elliptic operators. Proof of Lemma 11.3We shall use the fixed-point formula and calculate the contribution of each componentFof theT-fixed set inM.LetL⊆G×Hbe the corresponding good torus(see the proof of Lemma 11.1)that fixes the inverse image ofFinPpointwise,and consider the multiplication map Because the associated homomorphismR(G)→R(L)is injective,it suffices to calculate the quantization map whereis any lift of the topological generatorg∈T.But ifS⊆Pis the inverse image ofF,then sinceLacts trivially onSthere exists an isomorphism and we can argue exactly as in the proof of Lemma 8.1 or the discussion following it that as required. Let us now turn to the second component in the bordism(11.1),namely the orbifoldAt this point we need to make the further assumption concerningas we did in Section 8,that it is disjoint from theT-fixed set.Then we can proceed as above and as in Section 8 to compute. To formulate the final result denote by the map induced fromThen denote byMathe quotient topological spaceMbyT.Then form theG1-equivariant stable complex orbifold structure constructed from Theorem 11.4is any regular value of μ between0and1,and T acts locally freely Similarly,we have the following theorem. Theorem 11.5If a∈Ris any regular value of μ between0and1,T acts locally freely on,and,then for all In both theorems,we can replaceif 0 is a regular value(but for our purposes the formulations above are better). LetGbe a torus.Letbe a smooth,G-equivariant,even-dimensional,stable complex orbifold structure on aG-spaceM,withHa torus.Letbe aG-equivariant,smooth Hermitian line bundle on(that is,aG×H-equivariant Hermitian line bundleEonP).Let be a Kostant moment map(associated to a-equivariant connection onE).Finally,assume thatGacts without fixed-points on It is convenient not to assume that 0∈g∗is a regular value ofμ,but in any case every neighborhood of 0∈g∗contains a regular valuea,and we may form the reduced orbifold structure on the locally compact spacewhere,as in the previous section,the map is obtained by factoring(12.1)through the quotient mapq:P→M. We may also form the reduced orbifold line bundleMore generally,ifis any class inthen restricting towe obtain from it a class We want to prove the following versions of the quantization commutes with reduction theorem,in whichandrefer to the multiplicities of the trivial representation of the full torusGin the given quantizations(up to now we have been dealing with the trivial representation of some circle subgroup). Theorem 12.1If M is closed,then for all regular values a∈g∗sufficiently close to0. Theorem 12.2If α is any class in?KG(?M),then for all regular values a∈g∗sufficiently close to0and all sufficiently large k>0. The lower bound onkis independent of the particular choice ofa.Of course,if 0∈g∗isa regular value then we can simply takea=0. Proof of Theorem 12.1IfGhas dimension one,then the result is covered by Theorem 11.4.We shall prove the theorem in general by induction on the dimension ofG,and the only real difficulty comes from having to carry along the hypothesis thatGacts without fixed-points on In order to handle this point,we use the change of groups property of quantization(see item(e)following Lemma 11.3)and argue as follows.IfGhas dimension more than one,then we can find a finite covering by a product of tori such that if a point inMis fixed by eitherG1orG2then in fact it is fixed byG(just chooseG1andG2so that they are not contained in any of the finitely many proper isotropy subgroups inG). Now the fixed-points ofGonly occur whereμassumes integral values in g∗,and it is a simple matter to further adjustG1andG2so that in the decomposition neither of the axesandcontains any of the finitely many g∗-integral points in the range ofμ(the number is finite by compactness). By Theorem 11.4 again, for all regularsufficiently close to zero,where the subscript on the left-hand side indicates the multiplicity of the trivial representation ofG1.Now apply the induction hypothesis to conclude the theorem,but for alla=(a1,a2)sufficiently close to zero for whicha1anda2are separately regular values.However ifais any regular value sufficiently close to zero,then we can find a neighborhood aroundaconsisting entirely of regular values,and within that neighborhood an element(a1,a2)to which our argument applies.By bordism invariance so the theorem is proved. The proof of Theorem 12.2 is similar.The only additional point is that for any given class,we can replaceMby an open and relatively compactG-invariant subset that contains the support of.The compactness allows us to use the same argument as just given. In this final section we shall consider arbitrary connected and compact groups.We begin by discussing reduction in the context of non-abelian groups and stable complex manifolds,since it is slightly more complicated than in the abelian case(which in ef f ect we handle by writing a torus as a product of circles). LetGbe a compact connected Lie group.LetMbe aG-equivariant stable complex manifold(or orbifold),assume that 0∈g∗is a regular value of a moment map associated to a connection on an equivariant Hermitian line bundle,as in the introduction,and assume thatGacts locally freely on the inverse image of 0.There exists an exact sequence of vector bundles where the quotient map sends a tangent vectorXto the formand another exact sequence whereM0=μ−1[0]/G.Choosing equivariant splittings of both we obtain an isomorphism Equip the equivariant bundlewith an equivariant complex structureJthat exchanges to the two summands,with0 for allX∈g(the angle brackets denote the pairing between.The above isomorphism then endowswith a stable complex structure,assumingMhas been given one. We shall prove the following theorem,which is essentially due to Meinrenken(see[21]). Theorem 13.1Assume thatis a regular value of μ and that G acts locally freelyonare the reduced stable complex orbifold and orbifold line bundle,respectively,then for all sufficiently large k>0. We shall reduce the theorem to the abelian case,and the first step is Lemma 13.1 below.See for example[11]for this approach(where it is credited to Michele Vergne). Fix a maximal torusTinG.Recall that the Weyl character formula describes the restriction map as follows.A choice of positive roots determines a complex structure on the vectorthat is invariant under the adjoint action ofT,and for which the weights of the associated complexT-representation are precisely the negative roots(this convention suits well our purposes).The characters of the irreducible representations ofG,when restricted toT,take the form Most of the detailed aspects of this formula are not relevant to us,but to fix the notation: (a)φ:t→u(1)is the highest(infinitesimal)weight of the representation. (b)ρ:t→u(1)is the half-sum of the positive roots. (c)The exponentials are characters ofTdefined by (d)The sum is over the Weyl group,and the signs are given by the sign representation of the Weyl group. (e)The denominator is the Euler class inKG(pt)=R(T)of(g/t)∗,thought of as aT-equivariant vector bundle over a point.Thus All we need to extract from the character formula(13.2)is that since the restriction map(13.1)is injective,the multiplicity of the trivial representation in anyG-representationVis equal to the multiplicity of the trivial representation ofTin the virtual representation Hence we have the following lemma. Lemma 13.1The multiplicity of the trivial representation of G in the quantization QG(M,E)∈R(G)is equal to the multiplicity QT(M,E·Δ)0of the trivial weight of T in the quantization Our goal now is to compute theT-multiplicityQT(M,E·Δ)0by using the results of the previous section.The first step is to decompose g∗as (this is possible since t∗is not only a quotient of g∗,but also embeds in g∗as the linear functionals on g that vanish on all the root spaces).Denote by the two components of the moment map under the decomposition(13.4).The first one is the moment map for theT-action.We will use the second map to pull back the Bott generatorβof the complex vector space(g/t)∗toM.Since the mapμis obviously homotopic to the zero map,and the pullback of the Bott generator toMalong this map is Δ∈KT(M)(that is,it is the trivial virtual vector bundle with fiber Δ as described in(13.3)above),we find that and as a result Next,since the Bott element is supported onthe pullback classis supported onis any(T-invariant)open ball aroundthen the pullback of the Bott element defines a class in Let us write (which is of course an open subset ofM). Lemma 13.2If k is any integer,then ProofThis is just an instance of the excision property whereιis the inclusion ofUintoMandαis any class inK(U).See property(b)of the quantization map in the list at the end of Section 10. Lemma 13.3If the open ballis chosen to be sufficiently small,then0is aregular value of the T-moment map and T acts locally freely on the inverse image of zero. ProofThis follows from the fact thatand our assumptions concerning regularity and locally free actions forG. Now form the locally compact space and the orbifold whereqis the obvious projection mapping. Lemma 13.4If k is sufficiently large,then ProofThis follows from Theorem 12.2. It now remains to identify the quantization of the orbifoldthat appears in Lemma 13.4 to the quantization of the orbifoldthat appears in Theorem 13.1. The orbifoldis a fiber bundle(with structure groupG)overwith fiberG×TB.The fiber is in turn a bundle over the complex flag varietywith fiber the open ballB. TheK-theory classis the fiber product ofwith the Thom class onG×TB.The latter class hasG-equivariant quantization(index)1∈R(G),so using the multiplicative property(f)of quantization in the list at the end of Section 10 we find that This is precisely the relation we require,and Theorem 13.1 is proved. [1]Atiyah,M.F.,Global Theory of Elliptic Operators,In Proc.Internat.Conf.on Functional Analysis and Related Topics(Tokyo,1969),Univ.of Tokyo Press,Tokyo,1970,21–30. [2]Atiyah,M.F.,Elliptic Operators and Compact Groups,Lecture Notes in Mathematics,Vol.401,Springer-Verlag,Berlin,1974. [3]Atiyah,M.F.and Hirzebruch,F.,Riemann-Roch theorems for dif f erentiable manifolds,Bull.Amer.Math.Soc.,65,1959,276–281. [4]Atiyah,M.F.and Hirzebruch,F.,Spin-manifolds and Group Actions,in Essays on Topology and Related Topics(Mémoires Dédiés`a Georges de Rham),Springer-Verlag,New York,1970,18–28. [5]Atiyah,M.F.and Segal,G.B.,The index of elliptic operators.II,Ann.of Math.(2),87,1968,531–545. [6]Atiyah,M.F.and Singer,I.M.,The index of elliptic operators.I,Ann.of Math.(2),87,1968,484–530. [7]Baum,P.,Oyono-Oyono,H.,Schick,T.,et al.,Equivariant geometricK-homology for compact Lie group actions,Abh.Math.Semin.Univ.Hambg.,80(2),2010,149–173. [8]Baum,P.F.and Douglas,R.G.,K-Homology and Index Theory,in Operator Algebras and Applications,Part I(Kingston,Ont.,1980),Vol.38,Amer.Math.Soc.,Providence,RI,1982,117–173. [9]Baum,P.F.,Higson,N.and Schick,T.,On the equivalence of geometric and analyticK-homology,Pure Appl.Math.Q.,3,2007,1–24. [10]Block,J.and Higson,N.,Weyl Character Formula inK-Theory,in Proceedings of the RIMS Thematic Year 2010 on Perspectives in Deformation Quantization and Noncommutative Geometry,Kyoto University,World Scientific,2013,299–334. [11]Cannas da Silva,A.and Guillemin,V.,Quantization of Symplectic Orbifolds and Group Actions,in Northern California Symplectic Geometry Seminar,Vol.196,Amer.Math.Soc.Transl.Ser.2,Amer.Math.Soc.,Providence,RI,1999,1–12. [12]Cannas da Silva,A.,Karshon,Y.and Tolman,S.,Quantization of presymplectic manifolds and circle actions,Trans.Amer.Math.Soc.,352(2),2000,525–552. [13]Duistermaat,H.,Guillemin,V.,Meinrenken,E.and Wu,S.,Symplectic reduction and Riemann-Roch for circle actions,Math.Res.Lett.,2(3),1995,259–266. [14]Guillemin,V.,Ginzburg,V.L.and Karshon,Y.,Moment Maps,Cobordisms,and Hamiltonian Group Actions,Vol.98,Mathematical Surveys and Monographs,Amer.Math.Soc.,Providence,RI,2002. [15]Guillemin,V.and Sternberg,S.,Homogeneous quantization and multiplicities of group representations,J.Funct.Anal.,47(3),1982,344–380. [16]Guillemin,V.and Sternberg,S.,Geometric quantization and multiplicities of group representations,Invent.Math.,67(3),1982,515–538. [17]Higson,N.and Roe,J.,AnalyticK-Homology,Oxford Mathematical Monographs,Oxford University Press,Oxford,2000. [18]Karoubi,M.,K-Theory,Springer-Verlag,Berlin,1978. [19]Kasparov,G.G.,Topological invariants of elliptic operators,I,K-homology,Izv.Akad.Nauk SSSR Ser.Mat.,39(4),1975,796–838. [20]Kosniowski,C.,Applications of the holomorphic Lefschetz formula,Bull.London Math.Soc.,2,1970,43–48. [21]Meinrenken,E.,On Riemann-Roch formulas for multiplicities,J.Amer.Math.Soc.,9(2),1996,373–389. [22]Metzler,D.S.,AK-theoretic note on geometric quantization,Manuscripta Math.,100(3),1999,277–289. [23]Segal,G.B.,EquivariantK-theory,Inst.Hautesétudes Sci.Publ.Math.,34,1968,129–151. [24]Sjamaar,R.,Symplectic reduction and Riemann-Roch formulas for multiplicities,Bull.Amer.Math.Soc.(N.S.),33(3),1996,327–338. [25]Vergne,M.,Equivariant index formulas for orbifolds,Duke Math.J.,82(3),1996,637–652.


























12 Quantization Formula for Torus Actions










13 Non-Abelian Case






















杂志排行
Chinese Annals of Mathematics,Series B的其它文章
- C∗-Algebraic Intertwiners for Degenerate Principal Series of Special Linear Groups
- Geometric Property(T)∗
- Permanence of Metric Sparsification Property under Finite Decomposition Complexity∗
- Conjugacy Classes and Characters for Extensions of Finite Groups∗
- Coarse Embedding into Uniformly Convex Banach Spaces∗
- Laplacians and Spectrum for Singular Foliations∗
