Del Pezzo surfaces of degree 4 and their relation to Kummer surfaces (Q993656)

Although this article has, in its author's words, ``little pretence to originality'', it collects together and explains several pieces of projective geometry widely separated in time. The oldest is the description of the desingularised Kummer surface \(K\) of the Jacobian of a genus~\(2\) curve \(C\) as the intersection of three quadrics in \({\mathbb P}^5\), which has been known for at least a hundred and forty years. Another is the description of a degree~\(4\) del Pezzo surface by a pencil of quadrics in \({\mathbb P}^4\). Over an algebraically closed field there is not very much to add by now, but once questions of rationality arise there is newer material. Suppose that \(C\) has a rational Weierstrass point. Then the system of quadrics defining \(K\) can be used to define a projection onto a hyperplane in \({\mathbb P}^5\), and this determines a finite morphism of degree~\(2\) onto a degree~\(4\) del Pezzo. It was oberved rather recently by Flynn that every degree~\(4\) del Pezzo arises in this way, and that fact becomes immediately apparent from the construction in this paper. What is proved here is that the image \(X\) is obtained by blowing up the images in \({\mathbb P}^2\) of the other five Weierstrass points of \(C\) under the composition \(C\to {\mathbb P}^1\to {\mathop{\mathrm{Sym}}\nolimits}^2{\mathbb P}^1={\mathbb P}^2\), where the second map is the Veronese embedding of \({\mathbb P}^1\) as a plane conic. Moreover, the morphism \(K\to X\) is \(J[2]\)-equivariant and also gives a morphism from the twisted Kummer surface to the twisted del Pezzo. The proofs, which work over any field \(k\) of characteristic \(\neq 2\) and cardinality \(>5\), are the real point of the paper. A crucial intermediary is the scheme \(S_X\) of the five points in \({\mathbb P}^1_k\) that correspond to the five singular quadrics in the pencil defining \(X\). Using these five points allows a reduction to the case of quasi-split del Pezzo surfaces, i.e.\ those with a rational line. The connection with Kummer surfaces comes, in part, from reproving its description as an intersection of quadrics in terms of \(k\)-torsors. In this way the \(J[2]\)-equivariance also appears naturally.