Deformations of minimal cohomology classes on abelian varieties (Q2809268)

Matsusaka's theorem as generalized by \textit{Z. Ran} [Invent. Math. 62, 459--479 (1981; Zbl 0474.14016)], is the criterion that a ppav \( ( A, \Theta )\) is a Jacobian precisely if there is an effective curve \(C \subset A\) whose cohomology class is minimal, i.e., \([C] = [\Theta]^{g-1} / {(g-1)}!\), and then \(( A, \Theta )= J(C)\). \textit{O. Debarre} showed that the translations of the Brill Noether loci \(W_d\) and \(- W_d\) are the only subvarieties in \(J(C)\) whose cohomology class is minimal, namely it is \( [\Theta]^{g-d} / {(g-d)}!\) [J. Algebr. Geom. 4, No. 2, 321--35 (1995; Zbl 0839.14018)]. Debarre's conjecture says that the existence of a minimal cohomological class of dimension \(1 \leq d \leq g-2\) forces \(( A, \Theta )\) to be a Jacobian, but for the case \(g=5\) and \(d=2\), where intermediate Jacobians of cubic \(3\)-folds also satisfy the condition, because the class of the Fano surface of lines in minimal. He proved that the Jacobian locus is an irreducible component of the locus of ppav which carry a minimal class. The present paper presents the interesting contribution that, if a Jacobian could deform together with the variety \(W_d\) out the Jacobian locus, then the curve should be hyperelliptic. The proof is based on careful and detailed investigations of the deformation properties of (i) the loci \(W_d\) (ii) of the maps \(C(d) \to W_d\) and (iii) of the inclusions \(W_d \hookrightarrow J(C)\). The demonstration requires that \(C(d) \to W_d\) is a small resolution, so the curve must not to be hyperelliptic. An intriguing open question is to understand what happens along the hyperelliptic loci.