On the number of Mordell-Weil generators for cubic surfaces (Q452394): Difference between revisions

The Mordell-Weil Theorem tells us that for a smooth plane cubic curve over \(\mathbb{Q}\) there is a finite ``basis'' \(B\) for the rational points. That is to say there is a finite set of rational points from which all such points may be generated by using chords and tangents. There are two well known open questions in this connection. The first asks whether for any \(b\) there is a curve for which the smallest such basis \(B\) has \(\# B\geq b\) --- or equivalently, whether the ranks of elliptic curves over \(\mathbb{Q}\) are unbounded. The second asks whether smooth cubic surfaces have a finite basis, in a similar sense. This interesting paper relates to both these questions. One considers an arbitrary field \(K\), and a smooth cubic surface \(S\) in \(\mathbb{P}^3\), defined over \(K\). Given a finite set \(B\subseteq S(K)\) one then produces a succession of sets \[ B=B_0\subseteq B_1\subseteq\dots\subseteq S(K) \] by taking \(B_{n+1}\) to consist of points \(R\in S(K)\) such that either \(R\in B_n\), or there exist \(P,Q\in B_n\) and a line \(\ell\not\in S\) such that \(\ell\cdot S=P+Q+R\). If we now set \(\text{Span}(B)=\cup B_n\) the question is whether or not \(\text{Span}(B)=S(K)\). At first sight this seems to be a natural generalization of the chord and tangent process for cubic curves, but there is an important difference. It is permissible to take \(P=Q\in B_n\). This can correspond to several different points \(R\), depending on which line \(\ell\), in the tangent plane at \(P\), one chooses. Thus when \(P=Q\) there is no sense in which \(R\) is the unique ``sum'' of \(P\) and \(Q\). With the above definition of \(\text{Span}(B)\) the first result is that if \(\# K\geq 13\), and if \(S\) contains two skew lines both defined over \(K\), then there is a basis with just one element. Indeed one can take \(B=\{P\}\) for any \(P\) on one of the two lines, provided that it is not an Eckardt point. Previously we had no examples in which \(S\) was known to be finitely generated. The second theorem states that for any \(b\) there is a smooth cubic surface \(S_b\) defined over \(K=\mathbb{Q}\) for which any generating set, if it existed at all, would need to have \(\# B\geq b\). This is the analogue of elliptic curves having unbounded ranks. The surface \(S_b\) takes the shape \(x^3+y^3+z(z^2+Mw^2)=0\) in which the integer \(M\) has many suitably chosen prime factors.