Density of rational points on a family of del Pezzo surfaces of degree one (with an appendix by Jean-Louis Colliot-Thélène) (Q2155185)

The paper under review concerns the density of rational points on certain del Pezzo surfaces of degree 1 over fields of characteristic zero, but really the motivation comes from the problem of unirationality over non-closed fields, say \(k\). Here, it is natural to restrict to \(k\)-minimal del Pezzo surfaces, and then unirationality is essentially open only for degree 1 surfaces. More precisely, the Picard number (over \(k\)) can only be 1 or 2 for these surfaces, and in the latter case unirationality follows from [\textit{J. Kollár} and \textit{M. Mella}, Am. J. Math. 139, No. 4, 915--936 (2017; Zbl 1388.14096)], but the Picard number 1 case seems completely open. It thus comes naturally to consider the problem of density of rational points of del Pezzo surfaces of degree 1 (and Picard number 1) as a test case, since this is clearly implied by unirationality. Previous (unconditional) results in this direction include [\textit{A. Várilly-Alvarado}, Algebra Number Theory 5, No. 5, 659--690 (2011; Zbl 1276.11114)] and [\textit{C. Salgado} and \textit{R. van Luijk}, Adv. Math. 261, 154--199 (2014; Zbl 1296.14018)], but there are also conditional results using root numbers by the first author in [J. Lond. Math. Soc., II. Ser. 99, No. 2, 295--331 (2019; Zbl 1459.11141)]. In detail, the authors consider the degree 6 surfaces \(S\) in weighted projective space \(\mathbb P[1,1,2,3]\) given by \[y^2 = x^3 + az^6+bz^3w^3+cw^3 \tag{1}\] where \(a,b,c\in k\) with \(ab\neq 0\) and \(4ac\neq b^2\) (the last condition is not stated explicitly in the paper, but together they ensure that the equations indeed define a del Pezzo surface of degree 1). The main result states the sufficient condition that \(S(k)\) is Zariski dense if there are \(z,w\neq 0\) such that the elliptic curve resulting upon substituting into (1) has positive rank over \(k\). Moreover, this condition is also necessary if \(k\) is of finite type over \(\mathbb Q\). To prove the density, the authors argue with the (isotrivial) elliptic surface defined by (1) (which has torsion-free Mordell-Weil group as a consequence of the above conditions). Formally, it is obtained from \(S\) by blowing up the base point \((0:0:1:1)\) of the linear system \(|-K_S|\). Starting from a non-trivial point \(R\) in a fibre of (1), the authors exhibit an auxilliary (possibly reducible) curve \(C_R\) which serves as trisection of the elliptic fibration. The curve is chosen as a special member of the linear system \(|-3K_S|\) which is singular at \(R\) and at two other points. It was pointed out by Kollàr that \(C_R\) can also be derived from a cubic surface which is \(3:1\) dominated by \(S\) in a natural way. If the fibre \(F\) in question is smooth, then the authors prove that \(R\) can be chosen in such a way that \begin{itemize} \item \(C_R\) contains an irreducible component over \(k\) which gives a section of the fibration, or \item \(C_R\) is geometrically integral of genus 0, or \item \(C_R\) is geometrically integral of genus 1 for \(R\) varying over an open subset of \(F\). \end{itemize} The proof of the potential density in the main theorem then proceeds by a careful case-by-case analysis. The necessity of the given condition, over fields of finite type over \(\mathbb Q\), builds on a theorem Colliot-Thélène which improves upon Merel's theorem for bounding the torsion in families of elliptic curves. The paper concludes with three interesting examples illustrating the techniques developed.