Effective bounds on the size of the Tate-Shafarevich group (Q1815210)

The first author and Szpiro conjectured that, for every \(\varepsilon>0\), there exists an effectively computable constant \(c_\varepsilon>0\) depending only on \(\varepsilon\) such that the order of the Tate-Shafarevich group \(\text{ ш}_E\) of an elliptic curve \(E\) over \(\mathbb{Q}\) satisfies the inequality \[ |\text{ ш}_E |<c_\varepsilon N_E^{{1 \over 2}+ \varepsilon}, \] where \(N_E\) denotes the conductor of \(E\) over \(\mathbb{Q}\). The authors of the paper under review prove this conjecture for any family \({\mathcal E}\) of elliptic curves \(E\) over \(\mathbb{Q}\) whose \(j\)-invariant \(j(E)\) belongs to a finite set and, moreover, is different from 0 and \(12^3\). In general, the curves \(E\in {\mathcal E}\) are assumed to be modular (for which it suffices that they are semistable) and to satisfy the Birch and Swinnerton-Dyer conjecture. In many cases, however, the bound obtained is unconditional, e.g. for CM-elliptic curves of rank 0. However, bounds are also obtained in the exceptional cases in which \(j(E)\) equals 0 or \(12^3\), but they are slightly weaker. Actually, only curves \(E\) of ranks 0 or 1 are treated in the paper, and the exponent \({1\over 2}\) of \(N_E\) in the Goldfeld-Szpiro conjecture is replaced by 95/192 or, for \(CM\)-curves, by 59/120, if \(j(E)\neq 0,12^3\), by 79/120 if \(j(E)=0\), and by 37/60 if \(j(E)=12^3\). But the authors point out that their results extend to curves of higher rank if one assumes S. Lang's conjecture on a lower bound for the regulator \(R_E\) of \(E\) over \(\mathbb{Q}\) or, with weaker bounds, if one uses estimates of Hindry and Silverman instead of Lang's conjecture. The main ingredients of the proofs are an analysis of the constants in the Birch and Swinnerton-Dyer conjecture (especially with respect to varying twists) and an upper bound on the size of the \(n\)-th derivative of the \(L\)-series of a cusp form (twisted by an arbitrary character) at the center of the critical strip-following Friedlander and Iwaniec. The Goldfeld-Szpiro conjecture plays an important role in connection with some other conjectures. In particular, this conjecture for rank 0 semistable elliptic curves implies a version of the ABC conjecture.