The Hilbert scheme of points for supersingular abelian surfaces (Q841216)

The relation between Hilbert schemes of finite-length subschemes and crepant resolution of singularities, via the McKay correspondence, is fairly well understood now after the work of Ito and many others in the case of surfaces in characteristic zero. For higher-dimensional varieties there is progress but still much more to be done. As far as the reviewer can tell, this is the first serious attempt to understand the situation in positive characteristic. Many of the constructions are valid for subschemes of arbitrary length on arbitrary smooth varieties, but the detailed results are limited to \({\text{Hilb}}^{2}(S)\) for \(S\) a smooth surface, which is the most accessible case but quite possibly also the most interesting one. Another, more specific use of \({\text{Hilb}}^{n}(S)\) is in Beauville's generalised Kummer construction. In this case \(S=A\) is an abelian surface: the Hilbert-Chow morphism \({\text{Hilb}}^{n}(A)\to{\text{Sym}}^{n}(A)\), composed with addition in \(A\), gives a map \({\text{Hilb}}^{n}(A)\to A\) and the fibre over \(0\) is called the generalised Kummer variety \({\text{Km}}^{n}(A)\). In characteristic zero, both \({\text{Hilb}}^{n}(A)\) and \({\text{Km}}^{n}(A)\) are smooth (and have trivial dualising sheaf), but in positive characteristic this fails for \({\text{Km}}^{n}(A)\). This phenomenon, too, is explored here in characteristic~\(2\) for \(n=2\). It turns out that \({\text{Hilb}}^{2}(A)\to A\) is a quasifibration, i.e.\ all the fibres are nonsmooth. The third strand of this paper is to study the singularities of the fibres, especially \({\text{Km}}^{2}(A)\), in this case. They depend strongly on \(A\), via the singularities of the usual Kummer variety \(A/\pm 1\), which were studied by Shioda and by Katsura in the 1970s. Thus the author is led to revisit a large part of classical surface singularity theory, but in characteristic~\(2\). A key role is played by Artin's wild involutions on surfaces. These, and the consequences of blowing up the associated surface singularities, are studied in Section~1 of the paper. The blow-up does not resolve the singularity (it is not even normal in general) but does behave cohomologically like a resolution, and this provides a counterpart in the cases under discussion to the McKay correspondence. This relation to the Hilbert scheme is discussed in Sections~2 and~3. A byproduct is examples of nonrational canonical singularities (another positive characteristic phenomenon). Section~4 deals with the special case of rational double points (quotient singularities need not be rational in positive characteristic), listing the cases that may arise on \(G\)-Hilbert schemes. The next part of the paper, Sections~5--7, is concerned with the \(S=A\) and \(G=\{\pm 1\}\), the Kummer surface case. The most difficult case by far is when \(A\) is supersingular: to understand it involves an excursion into Laufer's theory of minimal elliptic singularities. Within this, the superspecial case \(A=E\times E\), where \(E=(y^2=x^3+x)\), requires its own treatment. Section~8 is a technical one about local algebraic conditions on symmetric products of surfaces. This, and the study of symmetric products of abelian surfaces that follows in Section~9, allows us to identify the Kummer variety \(A/\pm 1\) with the closed fibre of the addition map \({\text{Sym}}^{2}(A)\to A\). Now everything fits together and the relation between \({\text{Km}}^{2}(A)\) and \(A/\pm 1\) and its singularities is fully elucidated in the final Section~10.