Effectivity of Brauer-Manin obstructions on surfaces (Q631800)

The Brauer-Manin obstruction, first described by Manin in 1970, is an explanation for why some algebraic varieties, defined over a number field, fail to have any global rational points despite having points over every completion of the base field. The obstruction is based on the Brauer group of the variety, and any attempt to compute the obstruction depends on being able to compute the Brauer group. Manin's original interest was in rational surfaces; in this case, the Brauer group consists entirely of \textit{algebraic} elements, which are (in principle) possible to calculate once a finite set of generators for the geometric Picard group is known. An effective procedure, using explicit Galois cohomology, for computing the algebraic part of the Brauer group in this case, together with the associated obstruction, was described in the authors' previous article [Adv. Math. 218, No. 1, 1--27 (2008; Zbl 1142.14013)]. Elements of the Brauer group which are not algebraic are called \textit{transcendental}. The present article addresses the problem of computing the obstruction associated to the whole Brauer group, including the transcendental part, in the case of an algebraic surface where generators for the geometric Picard group are known. The specific result is as follows. Theorem 1. Let \(X\) be a smooth projective geometrically irreducible surface over a number field \(k\), given by a system of homogeneous polynomial equations. Assume that the geometric Picard group \(\mathrm{Pic}(X_{\bar{k}})\) is torsion-free and generated by finitely many divisors, each with a given set of defining equations. Then for each positive integer \(n\) there exists an effective description of a space \(X_n \subseteq X({\mathbb A}_k)\) which satisfies \[ X({\mathbb A}_k)^{\mathrm{Br}(X)} \subseteq X_n \subseteq X({\mathbb A}_k)^{\mathrm{Br}(X)[n]}, \] where \(\mathrm{Br}(X)[n] \subseteq \mathrm{Br}(X)\) denotes the \(n\)-torsion subgroup. In particular, \(X({\mathbb A}_k)^{\mathrm{Br}(X)}\) is effectively computable provided that \(|\mathrm{Br}(X)/\mathrm{Br}(k)|\) can be bounded effectively. A particular case of interest where there are transcendental elements in the Brauer group is when \(X\) is a K3 surface. Obtaining generators for the geometric Picard group and bounds on the Brauer group are both difficult problems, but known in some cases. The effective procedure described is based on explicit calculations in étale Čech cohomology. The first step is to fibre the surface \(X\) over the projective line and use known techniques on curves, applied to the generic fibre, to obtain elements in \(\mathrm{Br} \;\bar{k}(X)\) which may be ramified, but generate the whole of \(\mathrm{Br}(X_{\bar{k}})[n]\). Repeating this with a different fibration allows the calculation of a system of generators which are geometrically unramified. There is then an intricate descent procedure to obtain elements of \(\mathrm{Br}(X)\) generating the \(n\)-torsion. Finally, the obstruction is computed in the same way as in the previous article [loc.\,cit.].