Abstract. In this paper we give an algorithm for computing the 2-Selmer group of an elliptic curve which has complexity 0(L D (O-5, c t )), where D is the absolute discriminant of the curve. Our algorithm is unconditional but the complexity estimate assumes the GRH and a standard conjecture on the distribution of smooth reduced ideals. This improves on the corresponding algorithm of Birch and Swinnerton-Dyer, which has complexity of V When trying to compute the Mordell-Weil group of an elliptic curve one normally first computes the 2-Selmer group. This is a group which contains a subgroup isomorphic to E(Q)/2E(Q). Whilst computing the 2-Selmer group is certainly an effective procedure there is no known effective procedure for computing the subgroup isomorphic to
E(Q)/2E(Q), and thus for computing E(Q).However all is not lost as the 2-Selmer group gives one an upper bound on the rank of the elliptic curve, and this upper bound is often attained in practice. To measure the complexity of our algorithm we set This is a function which interpolates between polynomial time, a = 0, and exponential time, a = 1. In this note we show the complexity of computing the 2-Selmer group is 0(L D (O-5, Cj)), where D denotes the absolute discriminant of the elliptic curve, under the assumption of the GRH and a standard conjecture on the distribution of reduced smooth ideals.Let E be our elliptic curve given byWe shall assume that the elliptic curve has no points of order 2 defined over Q. This is certainly the most difficult case for finding the 2-Selmer group. The modern method of computing the 2-Selmer group in this case goes back to the paper of Birch and Swinnerton-Dyer [1]. In their method a search is carried out for the quartics which represent the homogeneous spaces given their invariants. This method is certainly fast for small values of D; however it is not hard to see that its complexity is at least O(VD); see [1,11]. In the present paper we shall show that the "old-fashioned" technique, which uses the arithmetic of number fields, combined with a method derived from a paper of Brumer and Kramer [2] will determine the 2-Selmer group in our stated time. Our complexity is therefore much better than the complexity of the algorithm of Birch and SwinnertonDyer. However due to numerous improvements to the method of Birch and SwinnertonDyer, most notably the ones due to Cremona [8], we expect that in practice the method of Glasgow Math.