Infinite family of elliptic curves of rank at least 4 (Q611897)

The article exhibits a one-dimensional parametric family of elliptic curves over \(\mathbb{Q}\) with rank at least four. While this rank is not too big in comparison with the last results obtained by \textit{N. D. Elkies} [``Three lectures on elliptic surfaces and curves of high rank'', \url{arXiv:0709.2908}], the equations of the proposed elliptic curves are quite simpler. The method developed in the paper follows the approach of \textit{E. Brown} and \textit{B. T. Myers} [Am. Math. Mon. 109, No. 7, 639--649 (2002; Zbl 1083.11037)], who were able to build families of elliptic curves of rank at least three. The core of the method is a base change from \(\mathbb{Q}(t)\) to the function field of a curve of higher genus (the best choice being a genus one curve with positive rank). The main theorem stated in the paper asserts that for infinitely many values of \(u\in\mathbb{Q}\), the elliptic curve \[ y^2+(u^2-u-3)xy=x^3+(u^2-u-3)x^2-x+1 \] has rank at least 4 over \(\mathbb{Q}\), since it contains the four linearly independent points: \[ (0,1),\qquad (1,1),\qquad (u,u+1), \qquad \left(\frac19,\frac1{54}(9+3u-3u^2+v)\right), \] where \((u,v)\) are the coordinates of a rational point lying on the curve: \[ v^2=9u^4-18u^3-9u^2-18u+2569. \] This is a rank 2 elliptic curve whose Mordell-Weil group is explicitly described in the article. Assuming the parity conjecture, some of the curves in the family should have rank at least 5. Since the finitely many values of \(u\) for which the assertions of the theorem aren't true come from rational points of certain higher genus curves, they cannot be explicitly given. The author provides a list of values of the parameter \(u\) for which the rank of the corresponding curve can be proven to be at least four (or five under the parity conjecture). In order to find such a family, the author starts from the curve \(E_t: y^2+txy=x^3+tx^2-x+1\) over \(\mathbb{Q}(t)\) and looks for a polynomial base change \(t(u)\) such that the curve \(E_{t(u)}\) admits a \(\mathbb{Q}(u)\)-rational point with \(x\)-coordinate of the form \(x=au+b\in\mathbb{Q}[u]\). For \(\deg_u t(u)>2\) there is no solution, and for \(\deg_u t(u)=1 \) the only points with that shape are integer linear combinations of \((0,1)\) and \((1,1)\). For the remaining quadratic case, only the values \(t(u)=u^2-u-3\) and \(t(u)=-u^2+u+1\) provide interesting points. Indeed, the second value would provide a second parametric family of curves with analogous properties to that of the Main Theorem. Once the base change \(t(u)=u^2-u-3\) is chosen, a search for extra points on the curve \(E_{t(u)}\) is conducted, which drives to the rational points of the second elliptic curve appearing in the Main Theorem. The possible linear dependency of the four points on \(E_{t(u)}\) presented above amounts to fractional identities between certain parameters; in all the cases these identities describe rational curves of genus greater than one, having finitely many rational points by Faltings theorem; these points give the values of the parameter \(u\) for which the assertions of the Main Theorem fail. At several points of the proofs, hard computations are necessary. They are performed by means of SAGE and Magma.