Weil-Châtelet divisible elements in Tate-Shafarevich groups. II: On a question of Cassels (Q2261611): Difference between revisions

Let \(A\) be an abelian variety over an algebraic number field \(k\) with absolute Galois group \(\text{Gal}_k\), and let \(A^t\) be its dual abelian variety. For the Weil-Châtelet group \(H^1(k,A)\), let \(\text{div}(H^1(k,A))=\cap_{n\in\mathbb{N}}nH^1(k,a)\) be the subgroup of divisible elements. The author defines the locally divisible \(H^1\) as the kernel \(H^1_{\text{div}}(k,A)=\text{ker}\left(H^1(k,A)\rightarrow\bigoplus_vH^1(k_v,A)/\text{div}(H^1(k_v,A))\right)\) containing the global cohomology classes which locally become divisible. The main result of this paper is as follows: If we assume that (i) \(H^1(k(A_p^t)/k, A_p^t)=0\), (ii) the \(\text{Gal}_k\)-modules \(A_p^t\) and \(\text{End}(A_p^t)\) have no common irreducible subquotient, in particular if \(p\gg 0\), then \(\text{div}(H^1(k,A))_{p^{\infty}}=H_{\text{div}}^1(k,A)_{p^{\infty}}\) holds and the Tate-Shafarevich group \({\text Ш}(A/k)\) is \(p\)-divisible in \(H^1(k,A)\). For elliptic curves over \(\mathbb{Q}\), this yields the following: Let \(E\) be an elliptic curve over \(\mathbb{Q}\). Then the following holds. (1) \({\text Ш}(E/\mathbb{Q})\) is \(p\)-divisible in \(H^1(\mathbb{Q},E)\) for all primes \(p>7\). (2) There is at most one odd prime number \(p=3,5,\) or \(7\), and then at most two quadratic twists \(E^{\tau}/\mathbb{Q}\) of \(E/\mathbb{Q}\), such that \({\text Ш}(E^{\tau}/\mathbb{Q})\) is not \(p\)-divisible in \(H^1(\mathbb{Q},E^{\tau})\). (3) If \(p=5\) or \(7\) and \({\text Ш}(E/\mathbb{Q})\) is not \(p\)-divisible in \(H^1(\mathbb{Q},E)\), then \(E\) has semistable reduction outside \(p\). For Part I see Compos. Math. 149, No. 5, 729--753 (2013; Zbl 1300.11056).