Visibility of the Shafarevich-Tate group at higher level (Q2469680)

Let \(A/K\) be an abelian variety of a number field \(K\). Let \(\text Ш(A/K)\) be the Tate-Shafarevich group of \(A/K\). Let \(\iota: A \hookrightarrow C\) be an embedding of \(A\) into another abelian variety \(C/K\). Let \(Q/K\) be the quotient abelian variety. A class \(c\in \text Ш(A/K)\subset H^1(K,A)\) is called visible if \(\iota_*(c)=0\) or, equivalently, if \(c=\delta(P)\) for some \(P\in Q(K)\), where \(\delta: Q(K)\to H^1(K,A)\) is the connecting homomorphism in Galois cohomology. Let \(B\subset C\) be another abelian subvariety, such that \(A+B=C\) and \(A\cap B\) is finite. \textit{A. Agashe} and \textit{W. Stein} [J. Number Theory 97, No. 1, 171--185 (2002; Zbl 1101.14055)] proved that for almost all primes \(\ell\) there is a homomorphism \(B(K)/\ell B(K) \to \text Ш(A/K)[\ell]\) whose kernel has \(\mathbb{F}_\ell\)-dimension at most the rank of \(A(K)\). In this paper a refinement of the above Theorem is proven: let \(R\subset \text{End}_K(C)\) be a commutative subring that leaves \(A\) and \(B\) stable, \(\mathfrak{m}\) a maximal ideal of \(R\) and let \(\ell\) be the residue characteristic. Suppose \(B(K)\) is finite. Then \((B(K)/\ell B(K))[\mathfrak{m}] \to \text Ш(A/K)[\mathfrak{m}]\) is injective. Agashe and Stein used their result to prove the existence of visible elements in \(\text Ш(A_f,K)\) where \(A_f\) is a newform abelian subvariety of \(J_0(N)= \text{Jac}(X_0(N))\). In this paper the authors use their to prove the existence of visible subgroups at a higher level: Let \(\delta_1,\delta_p: X_0(pN) \to X_0(N)\) be the degeneracy maps let \(\varphi: J_0(N)\to J_0(pN)\) be \(\delta_1^*+\delta_p^*\). The author prove the existence of elements in \(\text Ш(A_f,K)\) that are not visible by \(J_0(N)\), but visible by \(J_0(pN)\), i.e., visible at a higher level. At the end of the paper it is conjectured that for each modular abelian variety \(A/\mathbb{Q}\) and each \(c\in \text Ш(E/K)\) there exists a modular abelian variety \(C\) such that \(c\) is visible by \(A\). Some computational evidence for this conjecture is provided.