On higher eta-invariants and metrics of positive scalar curvature (Q1611335)

The paper addresses the problem of distinguishing Riemannian metrics of positive scalar curvature on closed Spin-manifolds, which admit at least one such metric, and extends some earlier results of [\textit{B. Botvinnik} and \textit{P. B. Gilkey}, Math. Ann. 302, No. 3, 507-517 (1995; Zbl 0835.58034)]. In order to put things in context let us recall main result of Botvinnik-Gilkey: Let \(M\) be a closed Spin manifold of odd dimension \(m\geq 5\) with non-trivial finite fundamental group \(G\). Asume that \(M\) admits Riemannian metric \(g\) of positive scalar curvature and that the number \(r_m(G)\) of virtual representations of \(G\) of certain class (depending on \(m\text{ Mod }4\)) greater than \(0\). Then \(M\) admits a countable family of mutually non-bordant metrics of positive scalar curvature. Moreover, the metrics can be distinguished by the eta-invariant of Dirac operator twisted by a flat bundle (coming from a representation of \(G\)). In the present paper the authors extend this result in two directions: For the first they do not confine themselves to odd-dimensional manifolds, and for the second they impose much less restrictive assumptions on the fundamental group of \(M\) (both odd-dimensionality of the manifold and finiteness of its fundamental group play crucial role in proofs of Botvinnik-Gilkey results). Their main result can be summarized as follows: If \(G= \Gamma\times\Delta\), \(\Gamma= \Gamma'\times F\) for some finitely generated group \(\Gamma'\) and finite group \(F\) with \(r_m(F)> 0\) for some odd \(m\), where \(5\leq m\leq\dim M\), \(\Delta\) is virtually nilpotent or Gromov hyperbolic and \(H^{\dim M-m-4i}(\Delta; Q)\neq 0\) for some integer \(i\), and \(M\) admits metric of positive scalar curvature, then \(M\) admits a countable family of mutually non-bordant metrics of positive scalar curvature. In order to prove this result, the authors use heavy machinery of Mishchenko-Fomenko index theory (and its further refinements) and Lott higher eta-invariants in place of ``standard'' index theory of elliptic boundary value problems and eta-invariants used in the Botvinnik-Gilkey paper.