Canonical rings of Gorenstein stable Godeaux surfaces (Q1749765)

The numerical Godeaux surfaces are the smooth minimal surfaces of general type with \(h^0(\Omega^2_S)=0\) and \(K_S^2=1\). They are the minimal surfaces of general type with minimal values of all standard invariants, attracting then the attention of several scholars. Despite that, a complete classification of them is still lacking. In a classical paper [J. Fac. Sci., Univ. Tokyo, Sect. I A 25, 75--92 (1978; Zbl 0399.14025)] \textit{M. Reid} proved that the torsion subgroup of the Picard group (equivalently, the algebraic fundamental group) of a numerical Godeaux surface is cyclic of order at most \(5\), and classified those of order at least \(3\) obtaining three nice explicit unirational family. The two missing cases are still mysterious under several points of view. Recently, there has been a lot of attention on the stable surfaces, the surfaces that appear in a modular compactification of the Gieseker moduli space of surfaces of general type. In their previous paper [Math. Nachr. 290, No. 5--6, 794--814 (2017; Zbl 1388.14104)], the authors prove that the stable numerical Godeaux surfaces that are Gorenstein have torsion group cyclic of order at most \(5\), generalizing then the first part of Reid's theorem. In this paper, they complete the generalization of the mentioned theorem to the Gorenstein stable case by classifying all stable Gorenstein numerical Godeaux surface with torsion subgroup of the Picard group of order at least \(3\) obtaining three unirational families, each containing one of the families described by Reid as dense open subset. In particular every stable Gorenstein numerical Godeaux surface with torsion subgroup of the Picard group of order \(3\) or more is smoothable. For the case of torsion \(2\) they show that every canonical curve of the natural double cover of a Gorenstein stable numerical Godeaux surface with torsion subgroup of the Picard group of order \(2\) is honestly hyperelliptic of arithmetic genus \(3\). This should in principle help in extending any description (still missing in literature) of the moduli space of smooth numerical Godeaux surface with torsion subgroup of the Picard group of order \(2\) to the stable Gorenstein case. Their classification, as Reid's original one, is obtained by classifying the possible canonical rings of the universal covers. The main difficulty in extending Reid's classification to the case of small torsion is that then the canonical ring has a lot of generators and relations. In the simply connected case this ring has at least 13 generators and 54 relations. We are not able to write down this ring for any of the smooth simply connected numerical Godeaux surfaces in literature. It is noteworthy that the authors manage to do that for a stable numerical Godeaux surface with trivial torsion subgroup of the Picard group, writing it as subring of a polynomial ring in three variables (with 13 explicit generators). It would be nice (although not easy) to write explicitely a smoothing of this surface: that would give a family of canonical rings of simply connected (conjecturally) smooth numerical Godeaux surfaces.