Expander graphs, gonality, and variation of Galois representations (Q428158)

The authors address questions relating the generic behavior of abelian schemes over curves over number fields to the behavior on the individual fibers. The theorems proved include: {1}) If \(k\) is a number field, and \(E_1\) and \(E_2\) are elliptic curves over \(k(T)\) that are not geometrically isogenous, then for each \(d \geq 1\) and all but finitely many \(\ell\) (depending on \(d\)), there are only finitely many points \(t\) of \(\mathbb{P}^1_k\) defined over fields whose degree over \(k\) is less than \(d\), such that the restrictions of \(E_1\) and \(E_2\) to \(t\) have isomorphic \(\ell\)-torsion (as \(G_{k_1}\)-modules, where \(k_1\) is the field of definition of \(t\)). {2}) Let \(\mathcal{A} \to U\) be a principally polarized abelian scheme of relative dimension \(g\), where \(U\) is a smooth, geometrically connected curve over a number field \(k\). If the image of the geometric monodromy representation of \(\mathcal{A} \to U\) is Zariski dense in \(\text{Sp}_{2g}(\mathbb{Z})\), then for each \(d \geq 1\) and all but finitely many \(\ell\) (depending on \(d\)), there are only finitely many points \(t\) of \(U\) defined over fields whose degree over \(k\) is less than \(d\), such that the arithmetic monodromy action on the fiber of \(\mathcal{A}\) above \(t\) does not contain \(\text{Sp}_{2g}(\mathbb{F}_{\ell})\). {3}) Letting \(\mathcal{A}\), \(U\), \(k\), and \(d\) be as above, then there exists \(\ell(d)\) such that for all primes \(\ell > \ell(d)\), there are only finitely many points \(t\) of \(U\) defined over fields whose degree over \(k\) is less than \(d\), such that the \(\ell\)-torsion on the fiber of \(\mathcal{A}\) above \(t\) is nonzero. The main underlying result, from which everything follows, is the following. Let \(U\) a smooth, geometrically connected curve over a number field \(k\), and let \(U_i \to U\) be an infinite family of étale covers. Then for all \(d \geq 1\) and all but finitely many \(i\), the curve \(U_i\) has only finitely many points whose degrees are bounded by \(d\), so long as a certain family of graphs associated to the \(U_i\) is \textit{esperantist}. A family of graphs being esperantist is a generalization of it being \textit{expander}. The link to rational points is that one can show that, when the family of aforementioned graphs is esperantist, one can show that the genus and the gonality of the \(U_i\) tend to \(\infty\). A result of Faltings and Frey then allows one to conclude the existence of finitely many low degree points on the \(U_i\). The first two theorems above follow relatively quickly from the underlying result, whereas for the third theorem, it requires somewhat more effort to show that the relevant family of graphs is esperantist. The paper is a very interesting synthesis of many different areas of mathematics.