Curve complexes and finite index subgroups of mapping class groups (Q2498411): Difference between revisions

For any compact connected orientable surface \(S\) with boundary, its extended mapping class group \(Mod(S)\) consists of isotopy classes of homeomorphisms preserving each boundary component. This paper concerns injections of finite index subgroups of \(Mod(S)\) into \(Mod(S)\), and in particular asks when such an injection is the restriction of an inner automorphism. This may be regarded as a kind of rigidity for \(Mod(S)\). Let \(S_{g,n}\) denote a surface of genus \(g\) with \(n\) holes. The main results in this paper (when combined with earlier work of Korkmaz, Irmak, Bell and Margalit) is the following sharp result (Theorem 1.3): if \(S\) is homeomorphic to \(S_{0,n}\) (where \(n=2,3,4\)) or \(S_{1,n}\) (where \(n=0,1,2\)), then there exists an isomorphism of a finite-index subgroup of \(Mod(S)\) into \(Mod(S)\) which is not the restriction of an inner automorphism. If \(S\) is not homeomorphic to one of these surfaces, then every injection of a finite index subgroup into \(Mod(S)\) is the restriction of an inner automorphism. Irmak proved this for \(S_{2,0}\) and \(S_{g,n}\) for \(g\geq 2\), Bell and Margalit for \(g=0, n\geq 5\), and Korkmaz for \(S_{0,4}\), \(S_{1,1}\) and \(S_{1,0}\). The answer is no for \(S_{0,2}\) and \(S_{0,3}\) and yes for \(S_{0,0}\) and \(S_{0,1}\) (the cases when \(Mod(S)\) is finite). Thus the present paper concentrates on \(S = S_{1,n}\). The main tool in the proof is the complex of curves \(\mathcal{C}(S)\) introduced by Harvey. The vertices of this simplical complex correspond to isotopy classes of simple closed curves on \(S\); two vertices span an edge if they possess disjoint representatives. Ivanov showed that every isomorphism between finite index subgroups of \(Mod(S)\) extends to an inner automorphism, by showing that every simplicial automorphism of \(\mathcal{C}(S)\) is induced by an element of \(Mod(S)\). The main technical result (Theorem 1.4), then, is that, if \(n\geq 3\) a simplicial map of \(\mathcal{C}(S_{1,n})\) which preserves pairs of vertices corresponding to intersecting pairs of curves is induced by an element of \(Mod(S_{1,n})\). This result has group-theoretic applications to finite index subgroups of \(Mod(S)\). A group \(G\) is co-Hopfian \(\Longleftrightarrow\) every injective homomorphism \(G\longrightarrow G\) is an automorphism. The authors prove (Theorem 2.1), except when \(S \approx S_{0,4},S_{1,0},S_{1,1}\), that every finite index subgroup of \(Mod(S)\) is co-Hopfian. One particularly striking example, which is discussed extensively in the paper (\S 4.2), concerns the two-holed torus \(S_{1,2}\). Recall that the abstract commensurator \(Comm(G)\) of a group \(G\) is the group of equivalence classes of isomorphisms of finite index subgroups of \(G\); two such isomorphisms are equivalent \(\Longleftrightarrow\) they agree on the intersection of their domains. Because \(Mod(S_{1,2})\) and \(Mod(S_{0,5})\) are have isomorphic finite-index subgroups, they have isomorphic commensurators. However, \(Comm(Mod(S_{1,2}))\) contains \(Mod(S_{1,2})\) with index \(5\), leading to the existence of non-geometric isomorphisms of finite-index subgroups. This is closely related to the isomorphism of \(\mathcal{C}(S_{1,2})\) with \(\mathcal{C}(S_{0,5})\). This quite readable paper is self-contained and provides a useful overview of the current state of the subject.