A finite dimensional approach to the strong Novikov conjecture
DANIEL RAMRAS RUFUS WILLETT GUOLIANG YUThe aim of this paper is to describe an approach to the (strong) Novikov conjecture based on continuous families of finite dimensional representations: this is partly inspired by ideas of Lusztig related to the Atiyah-Singer families index theorem, and partly by Carlsson's deformation K -theory. Using this approach, we give new proofs of the strong Novikov conjecture in several interesting cases, including crystallographic groups and surface groups. The method presented here is relatively accessible compared with other proofs of the Novikov conjecture, and also yields some information about the K -theory and cohomology of representation spaces.19K56, 55N15, 57R20; 20C99, 46L85, 46L80
IntroductionThe aim of this paper is to study the strong Novikov conjecture [19] for a finitely presented group Γ. If we assume that Γ has a finite classifying space BΓ, one version of this conjecture states that the analytic assembly mapis rationally injective; here the left hand side is the K -homology of BΓ and the right hand side is the K -theory of the maximal group C * -algebra of Γ. We give a definition of the analytic assembly map in Section 2 below. The strong Novikov conjecture implies the classical Novikov conjecture on homotopy invariance of higher signatures, as well as being closely related to several other famous conjectures. A naive version of our approach might proceed as follows. A finite dimensional unitary representationof Γ defines a vector bundle E ρ over BΓ via a well-known balanced product construction. E ρ defines an element [E ρ ] of the K -theory group K * (BΓ) and thus a 'detecting homomorphism'As is well-known 1 , ρ * factors through the analytic assembly map µ; hence µ(x) = 0 for any x ∈ K * (BΓ) such that there exists ρ : Γ → U(n) with ρ * (x) = 0. Thus if one can find 'enough' representations to detect all of K * (BΓ), one would have proved the strong Novikov conjecture.Unfortunately, this approach will not work: the bundles E ρ are flat, so Chern-Weil theory tells us that any 'detecting homomorphism' as in line (2) above is rationally trivial on reduced K -homology. One possible way to salvage the idea in the paragraph above is to use infinite dimensional representations. This led to the Fredholm representations of Miscenko [23], and subsequently to Kasparov's KK-theory [19]; both of these, and the closely related approach to the Novikov conjecture through the Baum-Connes conjecture [7] have proved enormously fruitful.In this paper we suggest a different approach. The central idea is not to use a single representation as in line (1) above, but instead a continuous family of representationsparametrized by a topological space X . Such a family defines a bundle E ρ over X × BΓ and thus a detecting homomorphismfrom the K -homology of BΓ to the K -theory of X via slant product with [E ρ ] ∈ K * (BΓ × X). This ρ * still factors through the analytic assembly map. The central result of this paper is that for...