The relative Mishchenko-Fomenko higher index and almost flat bundles. I: The relative Mishchenko-Fomenko index (Q2032782)

Let \(\Gamma\) be a discrete group and let \(X\) be a space with fundamental group \(\Gamma\). The higher index map for \(X\) is a canonical map from its \(K\)-homology to the \(K\)-theory of the maximal group \(C^\ast\)-algebra of \(\Gamma\). This map is equivalent to the Baum-Connes assembly map for \(\Gamma\) if \(\Gamma\) is torsionfree. If \(X\) is a manifold with boundary or, more generally, a pair of spaces \(Y\subseteq X\), then the fundamental group \(\Lambda\) of \(Y\) also becomes relevant. Chang-Weinberger-Yu [11] constructed a relative higher index map from the \(K\)-homology of the pair \((X,Y)\) to the \(K\)-theory of a suitable relative group \(C^\ast\)-algebra \(C^*(\Gamma,\Lambda)\), defined as the suspension of the cone of the canonical map \(C^*(\Lambda) \to C^*(\Gamma)\) induced by the inclusion map from \(Y\) to \(X\). Deeley and Goffeng [15,16] constructed this map differently, using the geometric \(K\)-homology of Baum and Douglas (see [14]). This article deals with a third definition of the relative higher index map and some of its applications. The new construction of the relative higher index is based on a relative version of the Mishchenko-Fomenko line bundle. One of the main theorems in the article shows that the new definition agrees with the two previous definitions. As a preparation for this, both definitions are recalled, together with some relevant background. The new definition also gives a dual relative higher index map from the \(K\)-homology of the relative group \(C^\ast\)-algebra to the \(K\)-theory of the pair. The new theory is used in two ways. First, it allows to prove sufficient conditions for the relative higher index map to be rationally injective, split injective, or invertible (see Theorem~6.5 in the article). Secondly, it is applied to relate the higher indices of a manifold \(M\) that is partitioned by a hypersurface \(N\subseteq M\) into two components, \(M = M_1 \sqcup_{N} M_2\), to the higher indices of the constituents. It is assumed that the canonical maps from the fundamental group of \(N\) to those of \(M_1\) and \(M_2\) are injective. The van Kampen Theorem implies that the fundamental group of \(M\) is the amalgamated free product of the fundamental groups of \(M_1\) and \(M_2\) over \(N\). The \(K\)-theory of the group \(C^\ast\)-algebra of this free product is related to the \(K\)-theory groups of the three constituents by a Mayer-Vietoris sequence. An important theorem in this article shows that the relative higher index of \(N\subseteq M_1\) is equal to the image of the (absolute) higher index of \(M\) under a canonical map that appears in the Mayer-Vietoris sequence. In particular, it follows that the higher index of \(N\) is zero if the higher index of \(M\) is zero. Another corollary concerns the non-vanishing of the higher index for a manifold built by cutting and pasting (see Corollary 5.4 in the article). Another consequence is the following: if the double of a compact spin manifold along its boundary is enlargeable and its higher index relative to the boundary does not vanish, then the manifold with boundary admits no metric with positive scalar curvature. The arguments in the article also work in the setting of KO-theory and the \(K\)-theory of real group \(C^\ast\)-algebras. This causes little extra work. It is useful because KO-theory enters both in the Gromov-Lawson conjecture and in the more recent work on the classification of topological insulators.