Galois coverings of moduli spaces of curves and loci of curves with symmetry (Q2439339)

Let \(g,n\) be non-negative integers with \(2g-2+n>0\) and let \(S_{g,n}\) be a closed orientable surface of genus \(g\) with \(n\) deleted points. The condition \(2g-2+n>0\) asserts that each (analytically finite) Riemann surface structure on \(S_{g,n}\) comes from a Fuchsian group. The space that parametrizes isomorphic marked (analytically finite) Riemann surface structures on \(S_{g,n}\) is the Teichmüller space \(T_{g,n}\), this being a simply-connected complex manifold of dimension \(3g-3+n\). The modular group \(\Gamma_{g,[n]}\) of \(S_{g,n}\) is the quotient of the group \({\roman Hom}^{+}(S_{g,n})\) of orientation-preserving self-homeomorphisms of \(S_{g,n}\) by its normal subgroup \({\roman Hom}^{0}(S_{g,n})\) of homeomorphisms homotopic to the identity. The group \(\Gamma_{g,[n]}\) acts as a discrete group of holomorphic automorphisms on \(T_{g,n}\), and \({\mathcal M}_{g,[n]}=T_{g,n}/\Gamma_{g,[n]}\) is the space that parametrizes isomorphic unmarked (analytically finite) Riemann surface structures on \(S_{g,n}\), called the moduli space \({\mathcal M}_{g,[n]}\). It follows that \({\mathcal M}_{g,[n]}\) is a complex orbifold of dimension \(3g-3+n\) and that \(\Gamma_{g,[n]}\) is its orbifold fundamental group. Each finite degree regular cover of \({\mathcal M}_{g,[n]}\) can be described by a finite index normal subgroup \(G\) of \(\Gamma_{g,[n]}\). Finite order elements of \(\Gamma_{g,[n]}\) act with fixed points on \(T_{g,n}\); so in order to obtain smooth finite degree regular covers, one need to ensure that \(G\) has no torsion. A general construction of finite degree regular covers of \({\mathcal M}_{g,[n]}\) is by considering finite index normal subgroups \(K\) of \(\pi_{1}(S_{g,n})\) which are invariant under its automorphisms (for instance, characteristic subgroups). The normal subgroup \(K\) provides a finite degree regular cover \(P_{K}:S_{K} \to S_{g,n}\) with deck group \(G_{K}=\pi_{1}(S_{g,n})/K\). The invariance property of \(K\) by the automorphisms of \(\pi_{1}(S_{g,n})\) asserts that every self-homeomorphism \(\phi\) of \(S_{g,n}\) lifts to a self-homeomorphism \(\psi\) of \(S_{K}\) so that \(P_{K} \circ \psi = \phi \circ P_{K}\), in particular, that \(\psi\) belongs to the normalizer of \(H\) (seen in \(\Gamma(S_{K})\), the modular group of \(S_{K}\)). The subgroup \(G\) of \(\Gamma_{g,[n]}\), consisting of those \(\phi\) with the property that \(\psi\) induces the identity automorphism of \(G_{K}\), is a finite index normal subgroup of \(\Gamma_{g,[n]}\). Unfortunately, the group \(G\) may have torsion. \textit{E. Looijenga} [J. Algebr. Geom. 3, No. 2, 283--293 (1994; Zbl 0814.14030)] provided a construction, for the same \(K\) as above, of another finite index normal subgroup of \(\Gamma_{g,[n]}\) which is torsion free. For this, he considers the natural homomorphism \(\rho_{m}:\Gamma(S_{K}) \to \mathrm{Sp}(H_{1}(\overline{S_{K}},{\mathbb Z}/m{\mathbb Z})\), where \(m \geq 2\) is an integer and \(\overline{S_{K}}\) is the compactification of \(S_{K}\) (by adding all the punctures). If \(m \geq 3\), the restriction of \(\rho_{m}\) to \(G_{K}\) is injective (if \(m=2\), one needs to take care that \(G_{K}\) does not contains an hyperelliptic involution). Consider the homomorphism \[ \rho_{K,(m)}:\Gamma_{g,[n]} \to N_{{\roman Sp}(H_{1}(\overline{S_{K}},{\mathbb Z}/m{\mathbb Z}))}(\rho_{m}(G_{K}))/\rho_{m}(G_{K}), \] which is defined by sending each \([\phi] \in T_{g,[n]}\) first to \([\psi] \in \Gamma(S_{K})\) and then applying \(\rho_{m}\) and later the corresponding projection to the quotient by \(\rho_{m}(G_{K})\). Then the kernel \(G\) of \(\rho_{K,(m)}\) will provide a finite index normal subgroup of \(T_{g,[n]}\) without torsion. In the reviewed paper, the author generalizes Looijenga's construction to provide finite degree regular covers of \(\overline{\mathcal M}_{g,[n]}\), the Deligne-Mumford compactification of \({\mathcal M}_{g,[n]}\). This last space is constructed by adding stable Riemann surfaces (obtained from \(S_{g,n}\) by pinching a finite set of suitable pairwise disjoint simple loops). The main idea to see why the extension is possible is in essence the following. Let \(K\) be a normal subgroup of \(\pi_{1}(S_{g,n})\) of finite index which is invariant under the automorphisms of \(\pi_{1}(S_{g,n})\). In order to see a point in the boundary, we consider a set \(F\) of pairwise disjoint simple loops in \(S_{g,n}\); let \(S_{g,n}^{F}\) be the (topological) stable surface obtained by pinching the loops in \(F\). Use \(P_{K}\) to lift the loops in \(F\) to obtain a set \(F_{K}\) of pairwise disjoint loops in \(S_{K}\); let \(S_{K}^{F_{K}}\) be the stable surface obtained by pinching the loops in \(F_{K}\). The map \(P_{K}\) induces a natural map \(P_{K}^{F}:S_{K}^{F} \to S_{g,n}^{F}\). Then each structure of stable Riemann surface given to \(S_{g,n}^{F}\) lifts under \(P_{K}^{F}\) to provide a structure of a stable Riemann surface to \(S_{K}^{F}\). The group \(\pi_{1}(S_{g,n})/K\) can be seen as a group of automorphisms of \(S_{K}^{F}\) whose quotient is \(S_{g,n}^{F}\), and \(P_{K}^{F}\) its natural cover map.