Generic vanishing and minimal cohomology classes on abelian varieties (Q2464032)

Let \((A, \Theta)\) be an indecomposable principally polarised abelian variety of dimension \(n\) defined over the complex numbers. A subvariety \(Y \subset A\) of dimension \(1 \leq d \leq n-2\) has minimal cohomology class if \([Y] = \frac{\Theta^{n-d}}{(n-d)!}\) (the minimality refers to the fact that such a cohomology class lives in \(H^{2(n-d)}(A, {\mathbb Z})\) and is not divisible). If \((A, \Theta)\) is general of dimension at least four, it contains no subvarieties of minimal class. In fact we have only two examples of subvarieties of minimal class: the subvarieties \(W_d(C)\) in the Jacobian of a smooth curve \(C\), and the Fano surface \(F\) of a smooth cubic threefold embedded in the intermediate Jacobian. By a conjecture of \textit{O. Debarre} [J. Algebr. Geom. 4, No.~2, 321--335 (1995; Zbl 0839.14018)], these are the only possible examples. This conjecture is known to be true in dimension at most four, cf. \textit{Z. Ran} [Invent. Math. 62, 459--479 (1980; Zbl 0474.14016)], but in higher dimension few things are known about subvarieties of minimal class. In the paper under review, the authors introduce a new approach to the characterisation of these subvarieties: they consider the probably more tractable cohomological properties of the twisted ideal sheaf of the subvariety. They conjecture that this twisted ideal sheaf has a certain generic vanishing property (the \(GV\)-property for short) if and only if the subvariety has minimal class. The main result of the paper is the proof of one of the conjectured implications: if \(Y\) is a geometrically nondegenerate subvariety of a principally polarised abelian variety whose ideal sheaf has the \(GV\)-property, the cohomology class of \(Y\) is minimal. The proof is based on the preceeding works of the authors on the Fourier-Mukai transform and their Generic Vanishing criterion [\textit{G. Pareschi} and \textit{M. Popa}, J. Am. Math. Soc. 16, No.~2, 285--302 (2003; Zbl 1022.14012); \url{arXiv:math.AG/0608127}]. They also introduce a new tool: the theta-dual \(V(Y)\) of a subvariety \(Y\) which is the locus of theta-translates containing \(Y\). They show that if the ideal sheaf of \(Y\) has the \(GV\)-property, the theta-dual has maximal dimension and \(Y-V(Y)\) is a translate of the theta-divisor. The authors also conjecture that this property of the theta-dual characterises subvarieties of minimal class. \newline It seems fair to say that the mentioned conjectures are rather heroic statements and we are far from the understanding the general picture. Nevertheless the authors' contribution to the subject is very inspiring and I encourage the reader to find out more about it in this beautiful paper.