The uniformization of the moduli space of principally polarized abelian 6-folds (Q1985541)

The authors present a uniformisation of the moduli space \(\mathcal{A}_6\) of principally polarised abelian 6-folds in the following sense: they construct a generically finite dominating morphism from a variety \({\operatorname {Hur}}\) that parametrises certain degree 27 covers of \(\mathbb{P}^1\). This is especially interesting because \(\mathcal{A}_6\) is mysterious: at the time the paper was written, nothing was known about its Kodaira dimension, although it has since been shown in [\textit{M. Dittmann} et al., Algebra Number Theory 15, No. 1, 271--285 (2021; Zbl 1469.11113)] that its second plurigenus is positive so it has non-negative Kodaira dimension. ``Wake an algebraic geometer in the dead of night, whispering `27'\dots'', one of the authors famously wrote elsewhere. Of course these are the twenty-seven lines on a cubic surface, and the covers are Galois with Galois group the Weyl group of \(E_6\). Take a Lefschetz pencil of cubic surfaces, restrict to the lines, normalise and take the Stein factorisation and one has Kanev's construction, a degree 27 cover \(C\) of \(\mathbb{P}^1\) with branching at each of the 24 branch points given by a reflection in the Weyl group. The space \({\operatorname {Hur}}\) is the parameter space of such covers. The curve \(C\) has genus 46 and a symmetric correspondence \(\tilde D\) of degree 10, given by the incidence relation among the lines. This induces an endomorphism \(D\) of \(JC\), satisfying \((D-1)(D+5)=0\), and the image of \(D-1\) is the Prym-Tyurin-Kanev variety which turns out to be a principally polarised abelian 6-fold. All this is due to Kanev: the basic result of this paper is that the map \(PT\colon H\to\mathcal{A}_6\) that this induces is generically finite and therefore, since both spaces are irreducible of dimension 21, dominating. Here, in other words, is a construction of the general principally polarised abelian 6-fold. The proof of this result is by degeneration to the boundary of \(\mathcal{M}_{0,24}\), where one has a local toroidal description of \(PT\). The theory of degeneration of Torelli and Prym maps is known and the same techniques are applicable here, though not directly. Indeed some computer calculations involving \(E_6\) are used to show that the differential of \(PT\) has maximal rank. Hurwitz spaces can be complicated, however. An approach to studying \({\operatorname {Hur}}\) is via its \(S_{24}\) cover by the space \(\mathcal{H}\) of \(E_6\)-covers of \(\mathbb{ P}^1\) with 24 labelled branch points. This can be compactified to \(\overline{\mathcal{H}}\) in a way compatible with the action of \(S_{24}\) and the forgetful map to \(\mathcal{M}_{0,24}\) , giving a compactification \(\overline{\operatorname {Hur}}\) that extends the Prym-Tyurin-Kanev map \(PT\) to the Satake compactification of \(\mathcal{A}_6\). It is clearly of interest to understand the birational geometry of these spaces. One would hardly have expected \(\overline{\mathcal{H}}\) to be unirational (even before the result of Dittmann, Salvati Manni and Scheithauer, which precludes that) and indeed its canonical class is big. The canonical class of \(\overline{\operatorname{Hur}}\) is harder to describe but a second main result of this paper is that there is a boundary divisor \(E\), contracted by the extension of \(PT\) to a rational map to the perfect cone compactification \(\overline{\mathcal{A}}_6\), such that \(K_{\overline{\operatorname {Hur}}}+E\) is big. In the course of proving this fact, the authors discover a lot about the geometry of the spaces concerned. In particular they link the splitting of the Hodge bundle on \(\overline{\operatorname {Hur}}\) into Hodge eigenbundles with the Brill-Noether theory of \(C\): in particular, the \(+1\)-eigenbundle corresponds generically to the image of the Petri map. This analysis leads to a description of the ramification divisor of \(PT\), in a way strongly reminiscent of the corresponding result for the usual Prym map. There is also a sense in which Pryms appear as limits of Prym-Tyurin-Kanev varieties.