Average rank in families of quadratic twists: a geometric point of view (Q1751044)

Let \(E\) be an elliptic curve defined by the equation:\(y^2=x^3+Ax+B\), for \(A,B\in \mathbb Q\). For a square-free integer \(d\), let \(E_d\) be the quadratic twist of \(E\) by \(d\). The purpose of this article is to study the averages of rank \(r_{d}\) and analytic rank \(r_{\text{an},d}\) of \(E_d(\mathbb Q)\) in varying \(d\) for fixed \(A\) and \(B\). Heretofore,the problem for the average of rank have been studied for the distribution of rank based on the ordering by size of \(|d|\). The author considers this problem for the distribution of rank based on geometrical ordering by canonical height. Let \(\hat{h}_{E_d}\) be the canonical height of \(E_d\). Define \(\eta_d(A,B)\) by \(\log \eta_d(A,B)=\min\{\hat{h}_{E_d}(P) | P\in E_d(\mathbb Q)\backslash E_d(\mathbb Q)_{\text{tors}}\}\) if \(r_d\ge 1\) and \(\eta_d(A,B)=\infty\), otherwise. Define a finite set \(\mathcal{H}(Y)=\{d:\text{square-free}\mid \eta_d(A,B)\le Y \}\) and the average of analytic rank to be \(\mathcal{AR}_{\text{an}}(Y)=(\sum_{d\in\mathcal{H}(Y)}~r_{\text{an},d})/\sharp \{\mathcal{H}(Y)\}\). It should be noted that \(E_d\) of rank \(0\) are excluded in this setting. By using the square-free sieve of \textit{F. Gouvêa} and \textit{B. Mazur} [J. Am. Math. Soc. 4, No. 1, 1--23 (1991; Zbl 0725.11027)] and an upper bound for the number of rational points of bounded height on smooth surfaces due to \textit{D. R. Heath-Brown} [Ann. Math. (2) 155, No. 2, 553--598 (2002; Zbl 1039.11044)], the author obtains the following lower bound: \(\liminf_{Y\to\infty}\mathcal{AR}_{\text{an}}(Y)>1\). This implies similar results for the average of rank \(r_d\) under the parity conjecture, and further for that of corank of \(p^\infty\)-Selmer group of \(E_d\) for any prime number \(p\). He conjectures that \(\sharp\{d\in\mathcal{H}(Y), r_d=i\}\sim\sharp\mathcal{H}(Y)/2\) for \(i=1,2\). This conjecture implies that \(\sharp\{d\in\mathcal{H}(Y), r_d\ge 3\}=o(\sharp\mathcal{H}(Y))\) and \(\lim_{Y\to\infty}\mathcal{AR}(Y)=3/2\).