Small generating sets for the Torelli group (Q422816)

In this interesting paper, the author proves a conjecture of Dennis Johnson that the Torelli subgroup of the mapping class group of a surface has a finite generating set, the size of which grows cubically in the genus of the surface. The mapping class group \(Mod_{g,n}\) of a surface \(\Sigma_{g,n}\) of genus \(g\) with \(n\) boundary components is the group of isotopy classes of orientation-preserving self-homeomorphisms of \(\Sigma_{g,n}\) pointwise fixing the boundary. The \textit{Torelli subgroup} \(\mathcal I_{g,n}\) of \(Mod_{g,n}\) consists of those mapping classes that act trivially on the first homology group \(H_1(\Sigma_{g,n};\mathbb Z).\) It was first proved by \textit{D. Johnson} [Ann. Math. (2) 118, 423--442 (1983; Zbl 0549.57006)] that for \(g\geqslant 3\) and \(n\leqslant 1,\) \(\mathcal I_{g,n}\) is finitely presentable, with a generating set of size growing exponentially in \(g.\) In the same paper, Johnson conjectures that \(\mathcal I_{g,n},\) \(n\leqslant 1,\) has a generating set of size growing cubically in \(g\). The main result of this paper proves Johnson's conjecture. For \(g\geqslant 3,\) the author finds a generating set of \(\mathcal I_{g,0}\) of size at most \(57\binom{g}{3},\) and that of \(\mathcal I_{g,1}\) of size at most \(57\binom{g}{3}+2g+1\). The generators come from the mapping classes supported on certain subsurfaces of \(\Sigma_{g,n}\) of genus \(3\) with at most \(3\) boundary components. The author shows that to generate \(\mathcal I_{g,n}\), one needs \(\binom{g}{3}\) such subsurfaces, each of which provides at most \(57\) generating mapping classes. The new tool used in the proof and introduced in this paper is the \textit{handle graph} on which the Torelli subgroup acts cocompactly. Using the same technique and a recent result of \textit{R. Hain} [``Fundamental groups of branched covering and the Torelli group in genus \(3\)'', in preparation], the author gives a new and more conceptual proof of Johnson's result that \(\mathcal I_{g,n}\) is finitely generated for \(g\geqslant 3\) and \(n\leqslant1.\)