Jacobians among abelian threefolds: a formula of Klein and a question of Serre (Q606558)

Let \(k\) be a field and let \((A,a)\) be a principally polarized abelian threefold over \(k\). Suppose \((A,a)\) is isomorphic over \(\overline{k}\) to the Jacobian of a non-hyperelliptic curve \(X_0\) of genus three. According to a refined version of Torelli's theorem, there is a curve \(X/k\) isomorphic to \(X_0\) over \(\overline{k}\), and a quadratic character \(\epsilon: \text{Gal}(\overline{k}/k)\to \{\pm1\}\), such that the twisted abelian threefold \((A,a)_\epsilon\) is \(k\)-isomorphic to \((\text{Jac} X,j)\), where \(j\) is the canonical polarization. J.-P. Serre suggested that \(\epsilon\) should be trivial if and only if certain Siegel modular form \(\chi_{18}\) evaluated at \((A,a)\) is a square in \(k^*\), up to an adequate normalization. He raised then the question of how to compute the character \(\epsilon\) in a precise way. In this paper the authors answer this question when \(k\subset \mathbb{C}\). If \(\omega_1,\omega_2,\omega_3\) is any basis of \(H^0(A,\Omega_A^1)\), they prove that \((A,a)\) is \(k\)-isomorphic to a Jacobian if and only if \(-\chi_{18}(A,\omega_1\wedge \omega_2\wedge \omega_3)\) is a square in \(k^*\). The proof is based on an explicit expression of \(-\chi_{18}(A,\omega_1\wedge \omega_2\wedge \omega_3)\) as a product of even thetanullwerte, after a proper normalization. In particular, an explicit computation of the character \(\epsilon\) is derived.