A variant of a theorem by Ailon-Rudnick for elliptic curves (Q1708447)

Let \(\pi:\mathcal{E}\rightarrow C\) be an elliptic surface for a projective smooth curve \(C\) defined over \(\overline{\mathbb{Q}}\), with generic fiber \(E\) defined over the function field \(\overline{\mathbb{Q}}(C)\). There is a correspondence between \(\overline{\mathbb{Q}}(C)\)-rational points \(P\) of \(E\) and sections \(\sigma_P:C\rightarrow \mathcal{E}\) (i.e., maps such that \(\pi\circ \sigma_P=\mathrm{id}_C\)) and one defines \(P_t:=\sigma_P(t)\) a point in the elliptic curve \(\mathcal{E}_t:=\pi^{-1}(t)\) (for all the \(t\in C(\overline{\mathbb{Q}})\) whose fiber is an elliptic curve, i.e., for almost all of them). A recent theorem of \textit{L. DeMarco} and \textit{N. M. Mavraki} [Am. J. Math. 142, No. 2, 443--473 (2020; Zbl 1461.11102)] proves that if \(\mathcal{E}_1\) and \(\mathcal{E}_2\) are two elliptic fibrations of the same curve \(C\) with two sections \(\sigma_{P_1}\) and \(\sigma_{P_2}\) such that there is a sequence of points \(t_n\in C(\overline{\mathbb{Q}})\) verifying \[ \lim_{n\rightarrow +\infty} \widehat{h}_{(\mathcal{E}_i)_{t_n}}(P_i)_{t_n}=0 \] (for \(i=1,2\) and \(\widehat{h}\) the canonical Néron-Tate height), then there exist group morphisms \(\phi:\mathcal{E}_1\rightarrow \mathcal{E}_2\) and \(\psi:\mathcal{E}_2\rightarrow \mathcal{E}_2\) (not both trivial) such that \(\phi(P_1)=\psi(P_2)\). In the same setting, the paper under review considers couples of sections \(\sigma_{P_i},\sigma_{Q_i}\) (\(i=1,2\)) such that \((P_i)_t\) and \((Q_i)_t\) are linearly dependent for infinitely many \(t\in C(\overline{\mathbb{Q}})\) and proves that they yield isogenies \(\phi : E_1\rightarrow E_2\) and \(\psi:E_2\rightarrow E_2\) with \(\phi(P_1)=\psi(P_2)\) or a linear dependence between \(P_i\) and \(Q_i\) (\(i=1,2\)). The proof of the nontrivial case (i.e., when \(Q_i\) is not a multiple of \(P_i\)) uses estimates on heights for the (infinitely many) relations \([m_{i,t}](P_i)_t=(Q_i)_t\) to directly apply the DeMarco-Mavraki theorem (when \(\widehat{h}_{E_i}(P_i)>0\)) or to reduce to the isotrivial case (when both \(\widehat{h}_{E_i}(P_i)\) are 0, i.e., \(\sigma_{P_1}\) and \(\sigma_{P_2}\) are constant nontorsion sections) with \(\mathcal{E}_i=E_i^0\times C\), and use dominant morphisms to build the required isogenies. As an application, the authors prove a (slightly more general version of a) conjecture of \textit{J. Silverman} [J. Number Theory 48, No. 3, 330--352 (1994; Zbl 0807.14020)]: for \(\overline{\mathbb{Q}}(C)\)-independent points \(P_i,Q_i\) of elliptic curves \(E_i\) defined over \(\overline{\mathbb{Q}}(C)\) there is an effective divisor \(D \in\mathrm{Div}(C)\) such that \(\mathrm{gcd}([n_1]P_1-Q_1,[n_2]P_2-Q_2)\leqslant D\) for almost all \(n_i\) (where the divisor associated to a point \(P\) is the pull-back via \(\sigma_P\) of the \(O\)-section of an elliptic fiber \(\mathcal{E}\rightarrow C\) with generic fiber \(E\), i.e., \(\sigma^*_P(\sigma_O(C))\)) and, in the case \(Q_i=O_i\), \(\mathrm{gcd}([n]P_1,[n]P_2)=\mathrm{gcd}(P_1,P_2)\) for a set of natural numbers \(n\) with positive density in \(\mathbb{N}\).