Smith theory, \(L^2\)-cohomology, isometries of locally symmetric manifolds, and moduli spaces of curves (Q2443214)

The main theme of the thesis is the study of finite order diffeomorphisms (homotopic to the identity) of aspherical manifolds (and their universal covering). The general idea is that (under suitable assumptions) such diffeomorphisms should not exist. Indeed, the main technical result of the paper is the following. Theorem. Let \(M\) be an aspherical manifold such that it admits a complete Riemannian metric of finite volume whose lift to the universal covering is of bounded geometry. Make the following technical assumptions: \(M\) is the interior of a compact manifold with boundary, its Euler characteristic is non-zero, and the fundamental group is residually finite. Then there is no non-trivial diffeomorphism of \(M\) of finite order which is homotopic to the identity, and the deck transformation action does not normalize any non-trivial compact Lie group of diffeomorphisms of the universal covering. The most important consequence of this result are restrictions on the isometry groups. In particular, Avramidi proves: {\parindent=6mm \begin{itemize} \item[-] If \(M\) satisfies the assumptions of the theorem above and \(\pi_1(M)\) is irreducible (i.e.~no finite index subgroup splits as a non-trivial direct product) then one has the dichotomy that any complete finite volume Riemannian (or even Finsler) metric on \(M\) either is locally symmetric (and non-positively curved) or the isometry group of the universal covering is discrete and contains \(\pi_1(M)\) as a finite index subgroup. \item[-] In particular, if one starts with an irreducible locally symmetric space of finite volume with non-zero Euler characteristic, then any complete finite volume metric on \(M\) is either a constant multiple of the locally symmetric metric, or the isometry group of \(\tilde M\) as subgroup with universally bounded index (depending only on the locally symmetric metric) \item[-] If \(M\) is the Teichmüller space of a (possibly punctured) surface (but not of a sphere with less than \(5\) or a torus with less than \(3\) punctures) with a complete Finsler metric which is invariant under the extended mapping class group and which has finite covolume, then there are no other isometries than the extended mapping class group. \end{itemize}} The two special cases follow from the general one using standard results on the topology and geometry of either Teichmüller space or locally symmetric spaces. Farb and Weinberger have obtained similar results for compact aspherical manifolds. The main point here is to generalize to non-compact manifolds. Observe that the technical theorem stated above can be restated as follows: under the assumptions made the fixed-point set of the finite order diffeomorphism to be considered is the whole manifold. The idea is to establish this using homological properties of this fixed-point set, via a generalization of Smith theory. If \(M\) is closed, one can use the fundamental class of \(M\). In the more general case, Avramidi manages to find conditions which rule out that \(M\) is \(\mathbb{Z}/p\)-homology equivalent to a bundle over a lower dimensional submanifold (the normal bundle of the hypothetical fixed-point set of a diffeomorphism of order \(p\)). This homological condition essentially says that ``\(M\) contains too many homology classes with non-trivial intersection''. The validity of this condition (for a suitable finite covering) is proved for the manifolds as described above using \(L^2\)-Betti numbers and Lück's approximation theorem for them. As an additional complication, along the proof the original manifold \(M\) has to be replaced by another quotient \(M_*\) of \(\tilde M\) which is homotopy equivalent, but without a priori control on the end structure. This problem is overcome by the use of Thurston's measure homology. The very nice paper ends with a section which constructs many examples of infinite volume complete metrics on the manifolds as considered above, where the conclusion on the isometry group of the universal covering does not hold.