Realization of the mapping class group by homeomorphisms (Q884979)

Let \(M\) be a closed oriented surface, \(\mathcal{H}omeo(M)\) the group of orientation-preserving homeomorphisms of \(M\), \(\mathcal{MC}(M)\) the quotient of \(\mathcal{H}omeo(M)\) by the subgroup of homeomorphisms of \(M\) that are isotopic to the identity, and let \[ Pr:\mathcal{H}omeo(M)\to \mathcal{MC}(M) \] be the associated projection. In this paper, the author proves the following Theorem: If the genus of \(M\) is at least 6, then there is no homomorphism \(\mathcal{E}:\mathcal{MC}(M)\to \mathcal{H}omeo(M)\) such that the composition \(Pr\circ\mathcal{E}\) is the identity. The result settles an open question which has been asked by Thurston. The question for genus between \(2\) and \(5\) is still open. If one takes diffeomorphisms instead of homeomorphisms, then the result was known, and is due to Morita (in that case, for genus at least 5). Morita's proof is based on cohomological obstructions, and it is known by work of Mather and Thurston that these obstructions do not exist in the homeomorphism case. The proof in the paper under review uses geometry and topology on the surface (and no algebraic topology). A key ingredient is an analysis of an \(f\)-invariant upper-semicontinuous decomposition of \(M\) associated to each homeomorphism \(f\) of that surface.