Abelian varieties with many endomorphisms and their absolutely simple factors (Q427935): Difference between revisions

Let \(\mathbb{Q}\) denote the field of rational numbers, and \(\overline{\mathbb{Q}}\), an algebraic closure of \(\mathbb{Q}\). An abelian variety \(B/\overline{\mathbb{Q}}\) is called an \textit{abelian \(\mathbb{Q}\)-variety} if for each Galois conjugate \(B^\sigma\) there is an isogeny \(\mu_\sigma : B^\sigma \to B\) that commutes with all endomorphisms \(\phi \in\mathrm{End}_{\overline{\mathbb{Q}}}(B)\), i.e., \(\phi \circ \mu_\sigma = \mu_\sigma \circ \phi^\sigma\) for all \(\phi \in \mathrm{End}_{\overline{\mathbb{Q}}}(B)\). A \textit{building block} is an abelian \(\mathbb{Q}\)-variety whose endomorphism algebra is a central division algebra over a totally real field \(F\), with Schur index \(t\leq 2\) and reduced degree \(t\, [F:\mathbb{Q}] = \dim B\). \textit{K. A. Ribet} [Prog. Math. 224, 241--261 (2004; Zbl 1092.11029)] and \textit{E. E. Pyle} [Prog. Math. 224, 189--239 (2004; Zbl 1116.11040)] proved in 2004 the following correspondence: Let \(A/\mathbb{Q}\) be an abelian variety of \(\mathrm{GL}_2\)-type such that \(A_{\overline{\mathbb{Q}}}\) does not have complex multiplication. Then \(A_{\overline{\mathbb{Q}}}\) decomposes up to \(\overline{\mathbb{Q}}\)-isogeny as \(A_{\overline{\mathbb{Q}}} \sim B^n\) for some building block \(B/\overline{\mathbb{Q}}\). Conversely, if \(B/\overline{\mathbb{Q}}\) is a building block then there exists a \(\mathrm{GL}_2\)-type variety \(A/\mathbb{Q}\) such that \(A_{\overline{\mathbb{Q}}} \sim B^n\) for some \(n\). The author of the paper under review generalizes Ribet-Pyle's result to a number field \(k\), and proves that the Ribet-Pyle correspondence takes place in a slightly more general context: If \(A/k\) is an abelian variety such that \(\mathrm{End}_k(A)\otimes \mathbb{Q}\) is a maximal subfield of \(\mathrm{End}_{\overline k}(A)\otimes \mathbb{Q}\), then \(A_{\overline k}\) decomposes up to \(\overline k\)-isogeny as \(A_{\overline k} \sim B^n\) for some abelian \(k\)-variety \(B\). Conversely, if \(B/\overline k\) is a \(k\)-variety, then there exists an abelian variety \(A/k\) such that \(\mathrm{End}_k(A)\otimes \mathbb{Q}\) is a maximal subfield of \(\mathrm{End}_{\overline k}(A)\otimes \mathbb{Q}\), and it satisfies that \(A_{\overline k} \sim B^n\) for some \(n\). \textrm The result for \(\mathrm{GL}_2\)-type abelian varieties over \(k\) follows from this result.