Theta functions on the theta divisor (Q968579)

Let \(\mathbb{H}_n\) denote the Siegel upper half space of dimension \(n\). Then the classical Riemann theta function \(\theta(z,\tau)\) on \(\mathbb{C}^n\times\mathbb{H}_n\) gives rise to another function \(\eta(z,\tau)\) defined by the formula \[ \eta=^t(\theta_i)\circ (\theta_{ij})^c\circ (\theta_j), \] where \((\theta_i)\) stands for the column vector of first-order partial derivatives of \(\theta\) in the \(\mathbb{C}^n\)-direction and \((\theta_{ij})^c\) denotes the adjoint matrix of the Hessian matrix of \(\theta\). In the present paper, the author proves that the restriction of this function \(\eta\) to the vanishing locus \(\theta^{-1}(0)\) of Riemann's theta function has some remarkable properties. His main result states that is in fact a theta function of order \(n+1\) and weight \((n+5)/2\) on the theta divisor \(\Theta:= \theta^{-1}(0)\) and that it transforms under the action of the congruence subgroup \(\Gamma_{1, 2}\subset\text{Sp}(2n,\mathbb{Z})\) on \(\mathbb{H}_n\) with an automorphy factor \(\text{det}(c\tau+ d)^{(n+5)/2}\) on \(\Theta\). As the author shows in the sequel, this theorem has a precise geometric interpretation. Namely, if \((A,\Theta)\) is a complex principally polarized abelian variety with theta divisor \(\Theta\) then the zero locus of the function intersects the smooth locus \(\Theta^s\) of \(\Theta\) in the ramification locus of the Gauss map \(\Gamma: \Theta^s\to \mathbb{P}(T_0 A)^*\). Moreover, the latter result implies that \(\eta\) vanishes identically on the theta divisor if and only if \(A\) is a decomposable principally polarized abelian variety. As for the special case of Jacobians of curves, the geometric interpretation of the function \(\eta\) leads to a precise description of the locus of Weierstrass points on the underlying Riemann surface, mainly in terms of theta functions and their derivatives. Finally, the author derives a closed formula for the function in the particular case of \(n= 2\). This formula expresses \(\eta^3\) in terms of ``thetanullwerte'' with even characteristics and the Riemann theta function \(\theta\) itself. As the author points out, the description of the vanishing locus of the function \(\eta\) is also related to some recent work of \textit{S. Grushevsky} and \textit{R. Salvati Manni} on singularities of theta divisors [Int. Math. Res. Not. 2007, No. 15, Article ID rnm045 (2007; Zbl 1132.14026)].