Root numbers and ranks in positive characteristic (Q2577011): Difference between revisions

Let \(K\) be a global field and let \({\mathcal E}_\eta\) be an elliptic curve over \(K(T)\). It is known, by a theorem of Néron-Lang, that the group \({\mathcal E}_\eta (K(T))\) is finitely generated. The curve \({\mathcal E}_\eta\) uniquely extends to a minimal regular proper elliptic fibration \(\mathcal E \to {\mathbb P}^1_K\) and, for almost all (i.e., all but finitely many) \(t\in {\mathbb P}^1_K\), the specialization \({\mathcal E}_t\) of \({\mathcal E}\) at \(T=t\) is an elliptic curve over \(K\). By Silverman's specialization theorem [\textit{J. H. Silverman}, J. Reine Angew. Math. 342, 197--211 (1983; Zbl 0505.14035)], for almost all \(t\) the specialization map \({\mathcal E}_\eta (K(T)) = \mathcal E ({\mathbb P}^1_K) \to {\mathcal E}_t(K) \) is injective, and so in particular \[ \text{rank} ({\mathcal E}_\eta (K(T))) \leq \text{rank}({\mathcal E}_t(K)). \] The elliptic curve \({\mathcal E}_\eta\) is said to have elevated rank if the above inequality is strict for almost all \(t\). So far, the only known technique for constructing curves of elevated rank depends on the parity conjecture: in practice, one tries to construct \(\mathcal E \to {\mathbb P}^1_K\) so that \(W({\mathcal E}_t) = - (-1)^{\text{rank}({\mathcal E}_\eta)}\) for almost all \(t\), where \(W(E)\) denotes the global root number of the elliptic curve \(E\). Examples of this kind, for \(K={\mathbb Q}\), were found by \textit{J. W. S. Cassels} and \textit{A. Schinzel} [Bull. Lond. Math. Soc. 14, 345--348 (1982; Zbl 0474.14010)] and by \textit{D. E. Rohrlich} [Compos. Math. 87, No. 2, 119--151 (1993; Zbl 0791.11026)]. These examples are quadratic twists, and this implies that the elliptic curves so constructed are isotrivial, i.e., their \(j\)-invariant is a constant independent of \(T\). Also, standard conjectures would imply that there are no non-isotrivial elliptic curves over \({\mathbb Q}(T)\) of elevated rank. In the function field case, however, the authors are able to exhibit an example of a 2-parameter family of elliptic curves of elevated rank which are not isotrivial. Their main theorem can be restated as follows: Let \(F=\kappa(u)\) where \(\kappa\) is a finite field of characteristic \(p\neq 2\), let \(c,d\in\kappa^\times\) and define \[ {\mathcal E}_\eta : y^2= x^3+(c(T^2+u)^{2p} +du)x^2 -(c(T^2+u)^{2p} +du) x . \] If the parity conjecture over \(F\) is true, then \({\mathcal E}_\eta\) has elevated rank. It is easy to check that the curves \({\mathcal E}_\eta\) in the theorem are not isotrivial. To prove the theorem, the authors show that in fact the following more precise statements are true: the rank of \({\mathcal E}_\eta((F(T))\) is 1; \({\mathcal E}_t(F)\) has positive rank for \(t\neq\infty\); \(W({\mathcal E}_t) =1\) for all \(t\). Under the parity conjecture, this implies that \(\text{rank} ({\mathcal E}_t(F))\geq 2\) for \(t\neq\infty\).