Elements of given order in Tate-Shafarevich groups of abelian varieties in quadratic twist families (Q2027510)

Let \(A\) an abelian varieties over a number field \(K\). While the torsion of the finitely generated group \(A(K)\) is not hard to compute, the rank of \(A(K)\) remains one of the most difficult invariant attached to \(A\). For an integer \(n\), the classical \(n\)-descent methods almost gives a way to compute \(A(K)\) and the failure of this method is controlled by the \(n\)-torsion \(\Sha(A)[n]\) of the Tate-Shafarevich \(\Sha(A)\) group of \(A\). For every element \(s\in K^*/K^{*2}\), one can consider the quadratic twist \(A_s\) of \(A\) by \(k\): \(A_s\) is a \(K\)-abelian variety isomorphic to \(A\) after extension to \(K(\sqrt{s})\). One expects that, varying \(s\), the arithmetic of \(A_s\) over \(K\) might be very complicated and hence that, for many \(s\), the \(n\)-descend method applied to \(A_s\) fails. Let us order the \(A_s\) via the height of \(s\) (which, if \(K=\mathbb Q\), is the absolute value of the unique square free element in the equivalence class of \(s\in \mathbb Q/\mathbb Q^{*2}\)). Then the authors propose the following conjecture (extending a previous conjecture of Delaunay if \(A\) has dimension \(1\)). Conjecture: If \(n\geq 2\) is an integer, then \(\Sha(A_s)\) has an element of order \(n\) for a positive proportion of \(s\in K/K^{*2}\). If \(A\) is of dimension \(1\) over \(\mathbb Q\), it has a rational \(2\)-torsion point and \(n\) is a power of \(2\), Conjecture 1 is quite well-understood by the work of many people (Feng, Smith, Xiong, Zaharescu). The main contribution of the paper under consideration, is the verification of Conjecture 1 in several examples in higher dimension and when \(n\) is not a power of \(2\). To give a feeling of the results obtained, let us mention here three of them (corresponding to Theorem 1.3, Corollary 1.7 and Theorem 1.11 respectively). \begin{enumerate} \item Let \(r\geq 1\) be an integer. For any \(\varepsilon> 0\), there are infinitely many \(K\)-elliptic curve \(E\), non isomorphic over \(\overline K\), for which a proportion of at least \(1-\varepsilon\) of twists \(E_s\) have \(\vert \Sha(E_s)[3]\vert\geq 9^r\). \item Let \(E\) be the \(\mathbb Q\)-elliptic curve defined by \(y^2+xy+y=x^3+x^2-13x-219\). For at least \(50\%\) of positive square-free \(s \equiv 1\ (\bmod 8)\) that are coprime to \(5\), \(E_s(\mathbb Q)\) has rank \(0\) and \(\vert \Sha(E_s)[5]\vert\geq 25\). \item Let \(J\) the Jacobian of the curve \(y^3=x^4-x\). Let \(p \equiv 1\ (\bmod 9)\) a prime and \(F\) a number field of degree \(2d\) over \(\mathbb Q\) containing \(\mathbb Q(\zeta_p,J[p])\). Then there is an \(F\)-abelian variety \(A\) isogenous to \(J\) such that at least \(50\%\) of quadratic twists \(A_s\) have rank \(0\) and satisfy \(\vert \Sha(A_s)[p]\vert\geq p^{3d}\). \end{enumerate}