Positive scalar curvature, higher rho invariants and localization algebras (Q2251902)

The Roe algebra \(C^*(X)\) for a proper metric space \(X\) is a coarse invariant playing a key role in the formulation of the coarse assembly map in the coarse Novikov conjecture. To encode Roe algebra with local geometric information, Yu introduced the localization algebra \(C_L^*(X)\) for \(X\) generated by bounded and uniformly continuous functions \(f: [0, \infty)\rightarrow C^*(X)\) where the propagation of \(f(t)\) is tending to \(0\) as \(t\to\infty.\) Localization algebras serves as a major tool in proving several important cases of the coarse Novikov conjecture (see [\textit{G. Yu}, \(K\)-Theory 11, No. 4, 307--318 (1997; Zbl 0888.46047); Ann. Math. (2) 147, No. 2, 325--355 (1998; Zbl 0911.19001)]). A convenient feature of localization algebras is that the course assembly map for \(X\) can be identified as \[ ev_*: K_*(C_L^*(X))\rightarrow K_*(C^*(X)) \] induced by the natural evaluation map of the localization algebra \(C_L^*(X)\) at \(t=0.\) This paper uses localization algebras and \(K\)-theory to investigate an interesting geometric situation where a manifold with boundary admits a group action. Rho invariants are variants of eta invariants giving rise to the boundary term in the Atiyah-Patodi-Singer index theorem for manifolds with boundaries. The main contribution of the paper is the formulation of a higher version of rho invariant in the language of localization algebras. Consider a complete spin manifold \(N\) without boundary, equipped with a discrete group \(\Gamma\) of isometries preserving the spin structure. It is further assumed that the action of \(\Gamma\) on \(N\) is proper and cocompact and \(N\) has a complete Riemannian metric \(h\) whose scalar curvature is uniformly positive. The higher rho invariant is a quantity lying in the \(K\)-theory of some subalgebra of localization algebra: \[ \rho(D_N, h)\in K_*(C^*_{L, 0}(N)^{\Gamma}). \] Here, \(C^*_{L, 0}(N)^{\Gamma}\) is the kernel of the evaluation map from the equivariant localization algebras \(C^*_L(N)^{\Gamma}\) to the equivariant Roe algebra \(C^*(N)^{\Gamma}\), i.e., \[ ev: C^*_L(N)^{\Gamma}\rightarrow C^*(N)^{\Gamma}. \] The authors show that this version of higher rho invariant coincides with the definition of rho invariant which was first formulated by \textit{N. Higson} and \textit{J. Roe} [Pure Appl. Math. Q. 6, No. 2, 555--601 (2010; Zbl 1227.19006)]). The main result of the paper concerns a fairly general situation of a manifold with boundary. The authors relate the higher index of the Dirac operator on the manifold to the rho invariant of the Dirac operator restricted to the boundary. Let \(M\) be an \(m\)-dimensional complete spin manifold with boundary \(\partial M\) where the metric on \(M\) has product structure near \(\partial M\) and the restriction \(h\) to \(\partial M\) has positive scalar curvature. Moreover, let \(\Gamma\) be a discrete group acting on \(M\) properly, cocompactly and isometrically and preserving the spin structure. Denote by \(D_M, D_{\partial M}\) the Dirac operators on \(M\) and \(\partial M\) respectively. The higher index \(\mathrm{Ind} D_M\) of \(D_M\) gives rise to the a class in \(K_m(C^*(M)^{\Gamma}).\) which can be mapped to \(K_{m-1}(C^*_{L, 0}(M)^{\Gamma})\) through the boundary map \[ \partial_i: K_i(C^*(M)^{\Gamma})\rightarrow K_{i-1}(C^*_{L, 0}(M)^{\Gamma}) \] of the long exact sequence of \(K\)-theory associated to the short exact sequence \[ 0\rightarrow C_{L, 0}^*(M)^{\Gamma}\rightarrow C_L^*(M)^{\Gamma}\rightarrow C^*(M)^{\Gamma}\rightarrow 0. \] Meanwhile, there is a higher rho invariant \[ \rho(D_{\partial M}, h)\in K_{m-1}(C_{L, 0}^*(\partial M)^{\Gamma}) \] The main result states that the \(K\)-boundary map of the higher index of \(D_M\) is related to the rho invariant \(\rho(D_{\partial M}, h)\) as follows: \[ \partial_m(\mathrm{Ind} D_M)=i_*(\rho(D_{\partial M}, h)). \] Here, \(i_*\) is the natural homomorphism from \(K_{m-1}(C_{L, 0}^*(\partial M)^{\Gamma})\) to \(K_{m-1}(C_{L, 0}^*(M)^{\Gamma})\) induced by the inclusion \(i: \partial M\rightarrow M.\) The main result has some interesting implications. An immediate consequence is that nonvanishing of rho invariants is an obstruction to extension of the positive scalar curvature on the boundary to the whole manifold. A further application is that the authors use the main theorem and the higher rho invariant to construct a map from Stolz' positive scalar curvature exact sequence to the \(K\)-theory long exact sequence \[ \rightarrow K_{n+1}(C^*_L(X)^{\Gamma})\rightarrow K_{n+1}(C^*(X)^{\Gamma})\rightarrow K_{n}(C^*_{L, 0}(X)^{\Gamma})\rightarrow K_{n}(C^*_{L}(X)^{\Gamma})\rightarrow \] so that all diagrams involved commute. This is based on the works of \textit{S. Stolz} [in: Proceedings of the international congress of mathematicians, ICM '94, August 3-11, 1994, Zürich, Switzerland. Vol. I. Basel: Birkhäuser. 625--636 (1995; Zbl 0848.57021)], \textit{N. Higson} and \textit{J. Roe} [\(K\)-Theory 33, No. 4, 277--299 (2004; Zbl 1083.19002); 301--324 (2004; Zbl 1083.19003); 325--346 (2004; Zbl 1085.19002)] and \textit{P. Piazza} and \textit{T. Schick} [``Rho-classes, index theory and Stolz' positive scalar curvature sequence'', \url{arXiv:1210.6892}]. But the authors' proof works for all dimensions. This result implies that the higher rho invariant could provide nontrivial secondary invariants which does not belong to the image of the course assembly map.