The index of signature operators on Lipschitz manifolds (Q788364)

This paper is a part of a general program leading to index theory of elliptic operators on compact Lipschitz manifolds. The ultimate goal of this program (by now successfully realized in the joint work of the author and \textit{D. Sullivan} [see the following review] is an analytic proof of the topological invariance of the rational Pontryagin classes. The feasibility of such an approach follows from Sullivan's theorem on existence of essentially unique Lipschitz structure on every compact topological manifold of dimension 4. The Atiyah-Singer index theorem asserts that two integers (the analytic index and the topological index) associated to an elliptic differential operator on a compact smooth manifold are equal. A central part in this theory is played by the (twisted) signature operators. The paper under review shows how to do analysis on Lipschitz manifolds so that the signature operators can be defined. It is then proved that these operators are Fredholm between appropriately defined Sobolev spaces of forms with coefficients in Lipschitz vector bundles, so that the analytic index is defined. Finally this analytic index is shown to be a Lipschitz invariant. This last fact is a consequence of the excision theorem, which states that under certain conditions indices of two signature operators on two different Lipschitz manifolds having a common open part are equal. To carry this out the author has to define basic objects, e.g. Riemannian metrics, differential forms, exterior derivative etc., and redo much of ''standard elliptic machinery'' on Lipschitz manifolds. Of course, not all the pieces can be defined, and, for those that can, not all results remain true. However, in his earlier paper [Invent. Math. 61, 227-249 (1980; Zbl 0456.58027)] the author did much of the required analysis in a different non-smooth setting. As a matter of fact, the treatment of the signature operator (without twisting) in the paper under review is very close to this earlier work. Genuinely new ideas were required to define the twisted signature operators, to prove that they are Fredholm (Rellich-type result), and to compute their index. The difficulties encountered are illustrated e.g. by the fact that the Sobolev spaces (of forms with first derivatives in \(L^ 2)\) vary drastically when the Riemannian metric is changed.