Random methods in 3-manifold theory (Q2631145)

A Heegaard splitting is a decomposition of a 3-manifold into a union of two handlebodies of genus \(g\), glued together by a homeomorphism of their boundaries. Every closed 3-manifold has a Heegaard splitting and the minimum such \(g\) is called the Heegaard genus of the manifold. Let \(S_g\) be the closed orientable surface of genus \(g\), and let \(MCG(S_g)\) be the mapping class group of \(S_g\). As isotropic gluing maps give homeomorphic 3-manifolds, every element \(\phi\) of \(MCG(S_g)\) gives a 3-manifold \(M(\phi)\) of Heegaard genus at most \(g\). In [\textit{N. M. Dunfield} and \textit{W. P. Thurston}, Invent. Math. 166, No. 3, 457--521 (2006; Zbl 1111.57013)], a model of random 3-manifolds known as random Heegaard splittings was introduced: an element \(\omega_n\), arising from a random walk of length \(n\) on the mapping class group \(MCG(S_g)\), gives rise to a 3-manifold \(M(\omega_n)\). In [\textit{J. Maher}, J. Topol. 3, No. 4, 997--1025 (2010; Zbl 1207.37027)], the second author showed that the probability that \(M(\omega_n)\) is hyperbolic tends to 1 as \(n\) tends to infinity, but with no information about the rate of convergence. In this paper, the authors show that the rate is exponentially fast, that is there are constants \(K\) and \(c<1\) such that the probability that \(M(\omega_n)\) is not hyperbolic is at most \(K e^n\). Denote by \(\sigma(\mu)\) the group generated by the support of the probability distribution \(\mu\) generating the random walk. A subgroup of the mapping class group is complete if its limit set in the Thurston boundary is equal to the full boundary. {Theorem 1.} Let \(G\) be a finitely generated complete subgroup of \(MCG(S_g)\) and let \(\mu\) be a finitely supported probability distribution on \(G\) whose semi-group support \(\sigma(\mu)\) is equal to \(G\). Then the probability that \(M(\omega_n)\) is hyperbolic and of Heegaard genus \(g\) tends to 1 exponentially fast. Denote by \(d_{sp}(\phi)\) the distance between the disc sets of the two handlebodies in the curve complex of the Heegaard surface which is called the Heegaard splitting distance of a Heegaard splitting corresponding to an element \(\phi\in MCG(S_g)\). Theorem 1 is an immediate consequence of the following assertion. {Theorem 2.} For any complete subgroup \(MCG(S_g)\), there is a finitely generated subgroup \(H< G\) such that for any finitely supported probability distribution \(\mu\) whose semi-group support \(\sigma(\mu)\) is a group containing \(H\), the Heegaard splitting distance \(d_{sp}(\phi)\) grows linearly with exponential decay, that is there are numbers \(K, L>0\) and \(c<1\) such that \({\mathbf P}(d_{sp}(\phi)\leq L n) \leq K\, c^n\). The Casson invariant \(\lambda(M)\) is an integer-valued invariant defined for a 3-manifold \(M\) which is a closed orientable integral homology sphere. Let \(\mathcal K\) be the subgroup of the mapping class group generated by Dehn twists in separating curves. The group \(\mathcal K\) is complete, and every element \(\phi\in {\mathcal K}\) gives an integral homology sphere \(M(\phi)\). {Theorem 3.} There is a finitely generated subgroup \(H<{\mathcal K}\) such that for any finitely supported symmetric probability distribution \(\mu\) whose semi-group support \(\sigma(\mu)\) contains \(H\), the probability \(p_n(k)\) that \(M(\omega_n)\) has Casson invariant \(\lambda(M(\omega_n)) = k\) and is hyperbolic and of Heegaard genus \(g\) is asymptotically \(p_n(k)\sim c/{\sqrt{n}}\), for some constant \(c>0\) which depends on \(\mu\). Furthermore, if \(\Lambda_n\) is the conditional distribution of Casson invariants conditioned on \(M(\omega_n)\) hyperbolic and of Heegaard genus \(g\), then \(\Lambda_n/{\sqrt{n}}\) converges in distribution to a normal distribution as \(n\) tends to infinity. The paper is organized as follows. In Section 2 the authors prove Theorem 3. In Section 3 the authors review some basic results in coarse geometry, before using these to show the first property in Section 4 and some basic results about train tracks and shadow sets in Section 5, and then use these to show the second property in Section 6. In Sections 7 and 8 the authors use some more coarse geometry to prove exponential decay for the distance from the disc set and for the Heegaard splitting distance.