Explicit classification for torsion subgroups of rational points of elliptic curves. (Q1862915): Difference between revisions

According to the theorem of Mordell-Weil, the group \(E(K)\) of rational points of an elliptic curve \(E\) over a number field \(K\) is finitely generated. Hence we have the isomorphism \[ E(K) \cong E_{tors}(K) \times {\mathbb Z}^r, \] where \(E_{tors}(K)\) is the torsion subgroup and \(r\) is the rank of \(E\) over \({\mathbb Q}\). If \(K = {\mathbb Q}\) is the rational number field, a theorem of Mazur gives us the exact structure of the torsion group \(E_{tors}({\mathbb Q})\). (A similar theorem, due to Kenku and Momose, gives the structure of the torsion group \(E_{tors}(K)\) when \(K\) is a quadratic number field. (We mention in this connection the boundedness conjecture proved by Merel, according to which the torsion group \(E_{tors}(K)\) is bounded for any number field \(K\).) In general, it is considered as easy to determine the torsion group \(E_{tors}(K)\) for a definite number field \(K\). But to determine \(E_{tors}(K)\) for some classes of number fields \(K\) involves a certain amount of work (see for instance the papers of the reviewer and his coauthors, where \(E_{tors}(K)\) is determined for elliptic curves \(E\) with integral \(j\)-invariant over quadratic, cubic and certain biquadratic number fields \(K\). For this we refer to the book of \textit{S. Schmitt} and the reviewer, Elliptic curves - a computational approach, Berlin: W. de Gruyter (2003; Zbl 1195.11078). But how to determine the torsion group \(E_{tors}(K)\) for \(E\) given in terms of an equation over \(K\)? And how to find a generator if \(E_{tors}(K)\) is cyclic? These questions are answered by the authors in the case of \(K={\mathbb Q}\). The non-cyclic case was dealt with recently by K. Ono. It remains therefore to consider the cyclic case. In this case, by Mazur's Theorem, it suffices to consider elliptic curves \(E\) of the special form \[ y^2 = x(x+M)(x+N) \] with \[ M = m+\sqrt{D} \quad \text{ and } \quad N = m-n \sqrt{D}, \] where \(D \not= 1\), \(m\) and \(n \not= 0\) are in \({\mathbb Z}\). (Of course, \(M\) and \(N\) are members of the quadratic field \({\mathbb Q}(\sqrt{D})\) and are conjugates of each other.) The methods of the paper are rather elementary. We notice that the first parametrizations going back to Tate and used by the reviewer and his coauthors (see the book cited above) were given by \textit{D. S. Kubert} [Proc. Lond. Math. Soc., III. Ser. 33, 193--237 (1976; Zbl 0331.14010)]. Kubert applied the parametrizations to prove a partial case of the boundedness conjecture. It should be not too difficult to extend the main result to quadratic fields \(K\) making use of the theorem of Kenku and Momose. By the way, the Theorem of Nagell and Lutz can be generalized to number fields \(K\) (see the cited book).