Characterizing Jacobians via trisecants of the Kummer variety (Q990181)

The paper gives a geometric answer to the classical Schottky problem which asks for a characterization of a principally polarized abelian variety to be a Jacobian. Fay's trisecant formula implies that the Kummer variety of a Jacobian admits a 4-dimensional family if trisecants. Conversely, improving a result of Gunning, \textit{G. E. Welters} showed [Ann. Math. (2) 120, 497--504 (1984; Zbl 0574.14027)] that a principally polarized abelian variety (ppav) is a Jacobian, if its Kummer variety admits a formal 1-dimensional family of flexes. Note that a flex can be considered as a degeneration of a trisecant. In the same paper Welters asked (conjectured?) whether the existence of one trisecant suffices to characterize Jacobians among ppav's. The present paper gives a positive answer to this question. There are 3 types of trisecants, (1) all 3 points coincide, (2) 2 points coincide and (3) all 3 points are distinct. A former paper of the author [in: Integrable linear equations and the Riemann-Schottky problem. Ginzburg, Victor (ed.), Algebraic geometry and number theory. In Honor of Vladimir Drinfeld's 50th birthday. Basel: Birkhäuser. Progress in Mathematics 253, 497--514 (2006; Zbl 1132.14032)] dealt with the first case. The paper under review gives the analogous proof for the last two cases. It is proved that a ppav is a Jacobian if a certain differential equation is satisfied by explicit theta functions and this is shown to be equivalent to the existence of a trisecant line of the corresponding Kummer variety.