Finding rational points on elliptic curves using 6-descent and 12-descent (Q947500): Difference between revisions

Let \(E\) be an elliptic curve over \(\mathbb{Q}\) and let \(S^{(n)}(E/\mathbb{Q})\) be its \(n\)-th Selmer group. An \(n\)-descent procedure provides \(n\)-covering maps \(\pi_{\alpha,n}: C_{\alpha,n}\rightarrow E\) (where, usually, the index \(\alpha\) varies in \(S^{(n)}(E/\mathbb{Q})\,\)) and, as is well known, can be useful to find elements in the Mordell-Weil and Tate-Shafarevich groups of \(E\). The paper aims to combine \(n\) and \(m\)-descent to get an \(nm\)-descent data and is successful in particular when \(m=n+1\) (but there are also more general results for \(n\equiv \pm a^2\pmod m\) and \(m\equiv \pm b^2\pmod n\)). For the proof of the main theorem the author associates to an \(n\)-descent data \(C\rightarrow \mathbb{P}^{n-1}\) (a Brauer-Severi diagram, [see \textit{J. E. Cremona, T. A. Fisher, C. O'Neil, D. Simon, M. Stoll}, J. Reine Angew. Math. 615, 121--155 (2008; Zbl 1242.11039)]) a morphism of group schemes \(\chi_n: E[n]\rightarrow PGL_n\,\), then defines the Heisenberg group \(H_n\) as the counterimage of \(\chi_n(E[n])\) in \(SL_n\,\). Such groups behave well for coprime integers \(n\) and \(m\) (i.e. \(H_n\times H_m\simeq H_{mn}\,\)) and are the fundamental ingredient for the definition of \(mn\)-descent data. Explicit \(n\)-descent computations are available only for small values of \(n\). Some results on invariant theory of binary quartics and ternary cubics give a more computational approach to the problem and are used for the couples \((n,m)= (2,3)\) and \((3,4)\). These provide explicit formulas for 6 and 12-descent. In the final section the author gives a few numerical examples for some curves included in the database avaliable in \textit{W. A. Stein, M. Watkins} [Lect. Notes Comput. Sci. 2369, 267--275 (2002; Zbl 1058.11036)]. The descent described here succedes in finding a second generator for the Mordell-Weil group of each of the 35 curves of this database having rank at least 2.