Integral points on symmetric affine cubic surfaces (Q6120313)

Let \(\mathcal{U}\) be a finite type scheme over \(\mathbb{Z}\). A common approach to decide wether the set \(\mathcal{U}(\mathbb{Z})\) of \(\mathbb{Z}\)-points of \(\mathcal{U}\) is empty or not is to mimic the analogous necessary conditions over number fields. We say that \(\mathcal{U}\) satisfies the \textit{integral Hasse principle} if \[ U(\mathbb{R})\times\prod_{p\text{ prime}}\mathcal{U}(\mathbb{Z}_p) \neq \emptyset \Rightarrow \mathcal{U}(\mathbb{Z}) \neq \emptyset. \] where \(U\) is the generic fiber of \(\mathcal{U}\). There are families of varieties that satisfies the integral Hasse principle, e.g. for all but finitely many \(n \in \mathbb{Z}\) the variety defined by \(x_1^3 + x_2^3 + x_3^2 - x_1 x_2 x_3 = n\) satisfies the integral Hasse principle (see [\textit{A. Ghosh} and \textit{P. Sarnak}, Invent. Math. 229, No. 2, 689--749 (2022; Zbl 1510.11088)]), but there are counter examples (section 8 of [\textit{J.-L. Colliot-Thélène} and \textit{F. Xu} Compos. Math. 145, No. 2, 309--363 (2009; Zbl 1190.11036)]). It is also common to go one step further and define an obstruction analogous to the Brauer-Manin obstruction called, naturally, the \textit{integral Brauer-Manin obstruction}. This obstruction is defined through a pairing \[ \mathrm{BM}(-,-): \left( U(\mathbb{R}) \times\prod_{p\text{ prime}}\mathcal{U}(\mathbb{Z}_p) \right) \times \mathrm{Br} \: U \to \mathbb{Q} / \mathbb{Z}, \] explicitly it is the set of points \((x_p)_{p } \in U(\mathbb{R}) \times \prod_{p\text{ prime}} \mathcal{U}(\mathbb{Z}_p)\) such that \(\mathrm{BM}((x_p)_p,\alpha) = 0 \) for every \(\alpha \in \mathrm{Br} \: U\). We say that there is no \textit{Brauer-Manin obstruction} to the integral Hasse principle such set is non-empty. The counter example previously cited is explained by the integral Brauer-Manin obstruction. \par Let \(f \in \mathbb{Z}[u]\) be a monic cubic polynomial. The main result of the article is that for all but finitely many \(n \in \mathbb{Z}\) the affine surface \(\mathcal{U}_n\) defined by \[ f(x_1) + f(x_2) + f(x_3) = n \] has no integral Brauer-Manin obstruction. This generalises a result due to Colliot-Thélène and Wittenberg which is exactly the case \(f(u) = u^3\). The main result is achieved by explicit computations of a compactification of \(\mathcal{U}_n\) and the Brauer group of such compactification. \par Finally the article concludes with an application to the surface defined by the sum of three tetrahedral numbers and an explicit example failing both weak approximation and strong approximation.