Failure of the Hasse principle for Enriques surfaces (Q633599): Difference between revisions

Following Manin's original work describing how the Brauer group of a variety can obstruct the existence of rational points, his construction (known as the Brauer--Manin obstruction) has been generalised in various ways to give several different obstructions. Those considered in this article, in increasing order of strength, are: the algebraic Brauer--Manin obstruction; the Brauer--Manin obstruction; and the Brauer--Manin obstruction applied to finite étale torsors, known as the étale-Brauer obstruction. Various authors have constructed examples in which these obstructions give genuinely different information about the rational points. In this well-explained article, the authors fill a gap in the literature by constructing an Enriques surface \(X\) over the rational numbers on which the étale-Brauer obstruction shows that there are no rational points, yet there is no algebraic Brauer--Manin obstruction to the Hasse principle. This leaves open the possibility that the absence of rational points is explained by the (necessarily transcendental) Brauer--Manin obstruction; the authors relate this to the corresponding question for a \(K3\) surface. In the notation of the article: \textbf{Theorem 1.1.} There exists an Enriques surface \(X/\mathbb Q\) such that \[ X(\mathbb A_\mathbb Q)^{\text{ét},\mathbf{Br}} = \emptyset \qquad \text{and} \qquad X(\mathbb A_\mathbb Q)^{\mathbf{Br}_1} \neq \emptyset. \] Moreover, if \(X(\mathbb A_{\mathbb Q})^{\mathbf{Br}} = \emptyset\), then \(Y(\mathbb A_\mathbb Q)^{\mathbf{Br}\, Y \setminus \mathbf{Br}_1 Y} = \emptyset\), where \(Y\) is a \(K3\) double cover of \(X\). The variety \(X\) in question is defined as the quotient of a \(K3\) surface \(Y\) by a fixed-point free involution; \(Y\) itself is explicitly given as the intersection of three quadrics in \({\mathbb P}^5\), parametrised by three positive integers \(a,b,c\) subject to certain arithmetic conditions (described in Theorem 1.2). In particular, \((a,b,c)=(12,111,13)\) satisfies the conditions. Since \(Y\) is simply connected, the only finite étale torsors over \(X\) are twists of \(Y \to X\). To show that there is an étale-Brauer obstruction to the existence of rational points on \(X\), it is enough to show that there is a Brauer--Manin obstruction to the existence of rational points on every twist of \(Y\); the construction of \(Y\) ensures that this comes down to a known example of a Brauer--Manin obstruction, given by \textit{B. J. Birch} and \textit{H. P. F. Swinnerton-Dyer} [J. Reine Angew. Math. 274--275, 164--174 (1975; Zbl 0326.14007)]. Proving the absence of an algebraic Brauer--Manin obstruction to rational points on \(X\) is more intricate, and involves computing \(\mathbf{Pic} (\bar{X})\) and its Galois cohomology. This is accomplished using a fibration of \(Y\) into genus 1 curves; explicit equations for divisors, and the Galois action on them, are given in appendices.