On the variation of Tate-Shafarevich groups of elliptic curves over hyperelliptic curves (Q2581373)

Let \(F = {\mathbb F}_q(t)\) be a rational function field over the finite field \({\mathbb F}_q\) and \(E\) an elliptic curve over \(F\) with conductor \(({\mathfrak p})\cdot \infty\), where \(({\mathfrak p})\) is a prime ideal in \(A := {\mathbb F}_q[t]\). Suppose that \(E\) has split multiplicative reduction at the place at infinity. The paper studies the arithmetic of \(E\) over a quadratic extension \(K\) of \(F\) subject to certain conditions. Under some non-vanishing condition for the \(L\)-series of \(E/K\), the author obtains a remarkable formula that allows to evaluate the size of the Tate-Shafarevich group Ш\((E/K)\) of \(E/K\) in terms of other, more accessible invariants of \(E/F\) and \(K/F\). In the given situation it is known that \(E\) is a quotient \(\pi:J \to E\) (supposed to be chosen minimal) of the Drinfeld Jacobian variety \(J = J_0({\mathfrak p})\). Let \(M\) be the free \({\mathbb Z}\)-module generated by the supersingular Drinfeld invariants in characteristic \(({\mathfrak p})\). Then \({\text{ rank}}(M) = \dim J + 1\), \(M\) may be regarded as a space of automorphic forms, and is provided with a canonical \({\mathbb Z}\)-valued scalar product \(\langle~ , ~\rangle\). Furthermore, the relations between the Neron models of \(E\) and \(J\) define a canonical element \(H_E\) of \(M\). On the other hand, assume that the size \(q\) of \({\mathbb F}_q\) is odd, and that \({\mathfrak d}\) is an irreducible element of \(A\), and put \(K := F(\sqrt{\mathfrak d}\)). Assume further that \(({\mathfrak p})\) is inert in \(K\). Via the theory of supersingular Drinfeld modules and their endomorphism rings, the field \(K\) determines another element \(H_K\) of \(M\). The author's main results are as follows: Theorem 1: \(L(E_K,1) = 0\) if and only if \(\langle H_E, H_K\rangle = 0\). Theorem 2: Suppose that \(\langle H_E, H_K\rangle \neq0\). (This implies in particular, due to Tate, that the group \(E(F)\) of \(F\)-rational points is finite.) Then \(\#\text Ш(E/K)\) may be calculated through an explicitly given formula that involves only \(\langle H_E, H_K\rangle\), \(\#\text ШE(F)\), and other invariants of \(E/F\) that may be determined via Tate's algorithm at the places \(({\mathfrak p})\) and \(\infty\). As the author points out, the variation of \# Ш \((E/K)\) over different \(K\) depends only on the relative position of \(H_K\) and \(H_E\) in \(M\). While some of the assumptions (\({\mathfrak d}\) irreducible of odd degree) are made for purely technical reasons (such that the explicit formulae of [\textit{H.-G. Rück} and \textit{U. Tipp}, Doc. Math. 5, 365--444 (2000; Zbl 1012.11039)] may be applied), others (\(q\) odd, \(E\) of the given shape with conductor \(({\mathfrak p})\cdot \infty\)) seem to be of a more principal nature. Apart from proofs of the described results (and others), the paper contains an elaborated and highly instructive example (Example 5.5), where all the data given in Theorem 1.2 are calculated.