On the Conjectures of Bost and of Baum-Connes and the generalized Trace Conjecture
Wolfgang Lück
Münster
http://www.math.uni-muenster.de/u/lueck/
May 2007
Outline
Equivariant homology theory.
ClassifyingG-space for proper actions.
Conjecture due toBostand toBaum-Connes.
Inheritance properties underdirected colimits.
Equivariant Chern characters.
(Generalized) Trace Conjectured.
Convention: group will always meandiscrete group.
Equivariant homology theories
Definition (G-homology theory)
AG-homology theoryH∗ is a covariant functor from the category of G-CW-pairs to the category ofZ-graded abelian groups together with natural transformations
∂n(X,A) :Hn(X,A)→ Hn−1(A) forn∈Zsatisfying the following axioms:
G-homotopy invariance;
Long exact sequence of a pair;
Excision;
Disjoint union axiom.
Definition (Equivariant homology theory)
Anequivariant homology theoryH?∗ assigns to every groupGa G-homology theoryHG∗. These are linked together with the following so calledinduction structure: given a group homomorphismα:H→G and aH-CW-pair(X,A)there are for alln∈Znatural homomorphisms
indα:HHn(X,A) → HGn(indα(X,A)) satisfying:
Bijectivity
If ker(α)acts freely onX, then indα is a bijection;
Compatibility with the boundary homomorphisms;
Functoriality inα;
Compatibility with conjugation.
Example (Equivariant homology theories)
Given aK∗ non-equivariant homology theory, put HG∗(X) := K∗(X/G);
HG∗(X) := K∗(EG×GX) Borel homology.
Equivariant bordismΩ?∗(X);
Equivariant topologicalK-homologyK∗?(X)in the sense of Kasparov.
Recall forH ⊆Gfinite
KnG(G/H)∼=KnH({•})∼=
(RC(H) neven;
{0} nodd.
Classifying spaces for proper actions
Definition (ClassifyingG-space for properG-actions,tom Dieck(1974))
A model for theclassifying G-space for proper G-actionsis a proper G-CW-complexE Gsuch that for any properG-CW-complexY there is up toG-homotopy precisely oneG-mapY →E G.
Theorem (Homotopy characterization ofEF(G)) There exists a model for E G;
Two models for E G are G-homotopy equivalent;
A proper G-CW -complex X is a model for E G if and only if for each H ∈ F the H-fixed point set XH is contractible.
We haveEG=E Gif and only ifGis torsionfree.
We haveE G={•}if and only ifGis finite.
A model forE D∞is the real line with the obvious D∞=Z o Z/2=Z/2∗Z/2-action.
Every model forED∞is infinite dimensional, e.g., the universal covering ofRP∞∨RP∞.
The spacesE Gare interesting in their own right and have often very nice geometric modelswhich are rather small.
On the other hand anyCW-complex is homotopy equivalent to G\E Gfor some groupG(seeLeary-Nucinkis (2001)).
Conjectures due to Bost and Baum-Connes
Conjecture (Baum-ConnesConjecture)
The Baum-Connes Conjecturepredicts that the assembly map KnG(E G)→Kn(Cr∗(G))
is bijective for all n∈Z.
Conjecture (BostConjecture)
The Bost Conjecturepredicts that the assembly map KnG(E G)→Kn(l1(G)) is bijective for all n∈Z.
These conjecture have versions, where one allowscoefficients in aG-C∗algebraA
KnG(E G;A) → Kn(AoCr∗G);
KnG(E G;A) → Kn(Aol1G).
There is a natural map
ι:Kn(Aol1G)→Kn(AoC∗r G) map.
The composite of the assembly map appearing in the Bost Conjecture withιis the assembly map appearing in the Baum-Connes Conjecture.
We will see that the Bost Conjecture has a better chance to be true than the Baum-Connes Conjecture.
On the other hand the Baum-Connes Conjecture has a higher potential for applications since it is related to index theory and thus has interesting consequences for instance to the Conjectures due toBass,Gromov-Lawson-Rosenberg,Novikov,Kadison, Kaplansky.
These conjecture have been proved for interesting classes of groups. Prominent papers have been written for instance by Connes,Gromov,Higson,Kasparov,Lafforgue,Mineyev, Skandalis,Yu,Weinbergerand others.
Inheritance properties under colimits
Letψ:H →Gbe a (not necessarily injective) group homomorphism.
GivenG-CW-complexY, letψ∗Y be theH-CW-complex obtained fromY byrestrictingtheG-action to aH-action viaψ.
GivenH-CW-complexX, letψ∗X be theG-CW-complex obtained fromY byinductionwithψ, i.e.,ψ∗X =G×ψX.
Consider a directed system of groups{Gi |i ∈I}with (not necessarily injective) structure mapsψi:Gi →Gfori∈I. Put G=colimi∈IGi.
LetX be aG-CW-complex.
We have the canonicalG-map
ad: (ψi)∗ψ∗iX =G×Gi X →X, (g,x)7→gx.
Define a homomorphism tnG(X): colim
i∈I HGni(ψi∗X)→ HGn(X) by the colimit of the system of maps indexed byi∈I
HGni(ψ∗iX)−−−→ Hindψi Gn ((ψi)∗ψi∗X) H
Gn(ad)
−−−−−→ HGn(X).
Definition (Strongly continuous equivariant homology theory) An equivariant homology theoryH?∗ is calledstrongly continuousif for every groupGand every directed system of groups{Gi |i ∈I}with G=colimi∈IGi the map
tnG({•}) : colim
i∈I HGni({•})→ HGn({•}) is an isomorphism for everyn∈Z.
Theorem (Bartels-Echterhoff-Lück (2007))
Consider a directed system of groups{Gi |i∈I}with G=colimi∈IGi. Let X be a G-CW -complex. Suppose thatH∗?is strongly continuous.
Then the homomorphism tnG(X) : colim
i∈I HGni(ψ∗iX) −→ H∼= nG(X) is bijective for every n∈Z.
Idea of proof.
Show thatt∗G is a transformation ofG-homology theories.
Prove that the strong continuity implies thattnG(G/H)is bijective for alln∈ZandH⊆G.
Then a general comparison theorem gives the result.
Theorem (Bartels-Echterhoff-Lück (2007))
Let{Gi |i ∈I}be a directed system of groups with G=colimi∈IGi and (not necessarily injective) structure mapsψi:Gi →G. Suppose that H?∗ is strongly continuous and for every i∈I and subgroup H ⊆Gi the assembly map
HnH(E H)→HnH({•}) is bijective.
Then for every subgroup K ⊆G (and in particular for K =G) also the assembly map
HnK(E K)→HnK({•}) is bijective.
Lemma (Davis-Lück(1998))
There are equivariant homology theoriesH?∗(−;Cr∗)andH?∗(−;l1) defined for all equivariant CW -complexes with the following properties:
If H ⊆G is a (not necessarily finite) subgroup, then
HGn(G/H;Cr∗) ∼= HnH({•};Cr∗) ∼= Kn(Cr∗(H));
HGn(G/H;l1) ∼= HHn({•};l1) ∼= Kn(l1(H));
H?∗(−,l1)is strongly continuous;
BothH?∗(−;Cr∗)andH?∗(−;l1)agree for proper equivariant CW -complexes with equivariant topological K -theory K∗?in the sense of Kasparov.
One ingredient in the proof of the strong continuity ofH?∗(−;l1)is to show
colim
i∈I Kn(l1(Gi))∼=Kn(l1(G)).
This statement does not make sense for the reduced group C∗-algebra since it is not functorial under arbitrary group homomorphisms.
For instance,Cr∗(Z∗Z)is a simpleC∗-algebra and hence no epimorphismCr∗(Z∗Z)→Cr∗({1})exists.
HenceH?∗(−;Cr∗)isnotstrongly continuous.
Theorem (Inheritance under colimits for the Bost Conjecture, Bartels-Echterhoff-Lück (2007))
Let{Gi |i ∈I}be a directed system of groups with G=colimi∈IGi and (not necessarily injective) structure mapsψi:Gi →G. Suppose that the Bost Conjecture with C∗-coefficients holds for all groups Gi. Then the Bost Conjecture with C∗-coefficients holds for G.
Theorem (Lafforgue (2002))
The Bost Conjecture holds with C∗-coefficients holds for all hyperbolic groups.
Corollary
Let{Gi |i∈I}be a directed system of hyperbolic groups with (not necessarily injective structure maps).
Then the Bost Conjecture holds with C∗-coefficients holds for colimi∈IGi.
Many recent constructions of groups with exotic properties are given by colimits of directed systems of hyperbolic groups.
Examples are:
groups with expanders;
Lacunary hyperbolic groupsin the sense ofOlshanskii-Osin-Sapir;
Tarski monsters, i.e., groups which are not virtually cyclic and whose proper subgroups are all cyclic;
Certain infinite torsion groups.
Certaingroups with expandersyield counterexamples to the surjectivity of the assembly map appearing Baum-Connes Conjecture with coefficients by a construction due
toHigson-Lafforgue-Skandalis (2002).
These implies that the mapKn(Aol1G)→Kn(Aor G)is not surjective in general.
The main critical point concerning the Baum-Connes Conjecture is that the reduced groupC∗-algebra of a group lacks certain functorial properties which are present on the left side of the assembly map. This is not true if one deals withl1(G)or groups ringsRG.
The counterexamples above raised the hope that one may find counterexamples to the conjectures due toBaum-Connes,Borel, Bost,Farrell-Jones,Novikov.
The results above due toBartels-Echterhoff-Lück (2007)and unpublished work byBartels-Lück (2007)prove all these conjectures (with coefficients) except the Baum-Connes Conjecture for colimits of hyperbolic groups.
There is no counterexample to the Baum-Connes Conjecture (without coefficients) in the literature.
Equivariant Chern character
Theorem (Artin’s Theorem) Let G be finite. Then the map
M
C⊂G
indGC: M
C⊂G
RC(C)→RC(G)
is surjective after inverting|G|, where C⊂G runs through the cyclic subgroups of G.
LetC be a finite cyclic group.
TheArtin defectis the cokernel of the map M
D⊂C,D6=C
indCD: M
D⊂C,D6=C
RC(D)→RC(C).
For an appropriate idempotentθC ∈RQ(C)⊗ZZ h 1
|C|
i
the Artin defect is after inverting the order of|C|canonically isomorphic to
θC·RC(C)⊗ZZ 1
|C|
.
Theorem (Lück(2002))
Let X be a proper G-CW -complex. LetZ⊆ΛG ⊂Qbe the subring of Qobtained by inverting the orders of all the finite subgroups of G.
Then there is a natural isomorphism chG: M
(C)
Kn(CGC\XC)⊗Z[WGC]θC·RC(C)⊗ZΛG
∼=
−→KnG(X)⊗ZΛG, where(C)runs through the conjugacy classes of finite cyclic
subgroups and WGC=NGC/C·CGC.
Example (Improvement of Artin’s Theorem)
Consider the special case whereGis finite andX ={•}Then we get an improvement of Artin’s theorem, namely,
M
(C)
Z⊗Z[WGC]θC·RC(C)⊗ZZ 1
|G|
∼=
−→RC(G)⊗ZZ 1
|G|
Example (X =E G)
In the special caseX =E Gwe get an isomorphism M
(C)
Kn(BCGC)⊗Z[W
GC]θC·RC(C)⊗ZΛG −→∼= KnG(E G)⊗ZΛG,
Conjecture (Trace Conjecture forG) The image of the trace map
K0(Cr∗(G))−→tr R
is the additive subgroup ofRgenerated by{|H|1 |H ⊂G,|H|<∞}.
Lemma
Let G be torsionfree. Then the Baum-Connes Conjecture for G implies the Trace Conjecture for G.
Proof.
The following diagram commutes because of theL2-index theoremdue toAtiyah(1974).
K0G(EG) //
∼=
K0(Cr∗(G))tr //R
K0(BG) //K0({•}) ∼= //Z
OO
Theorem (Roy(1999))
The Trace Conjecture is false in general.
Conjecture (Modified Trace Conjecture)
LetΛG⊂Qbe the subring ofQobtained fromZby inverting the orders of finite subgroups of G. Then the image of the trace map
K0(Cr∗(G))−−−→trN(G) R is contained inΛG.
Theorem (Image of the traceLueck(2002)) The image of the composite
K0G(E G)−−−→asmb K0(Cr∗(G))−−−→trN(G) R is contained inΛG.
In particular the Baum-Connes Conjecture implies the Modified Trace Conjecture.
Problem: What is the image of the trace map in terms ofG?
TakeX =E G. Elements inK∗(E G)are given by elliptic G-operatorsP over cocompact properG-manifolds with Riemannian metrics.
Problem: What is the concrete preimage of its class under chG∗? One term could be the index ofPC onMCgiving an element in K0(CGC\EC)which isK0(BCGC)after tensoring withΛG.
Another term could come from the normal data ofMC inMwhich yields an element inθC·RC(C).
The failure of the Trace Conjecture shows that this is more complicated than one anticipates. The answer to the question above would lead to a kind oforbifoldL2-index theoremwhose possible denominators, however, are not of the expected shape
n forH ⊆Gfinite.