Imaginary quadratic points on toric varieties via universal torsors (Q309267)

The paper under review is concerned with Manin's conjecture for rational points on Fano varieties for the special case of toric varieties over imaginary quadratic number fields. Manin's conjecture gives a prediction for the density of rational points on a Fano variety \(X\) defined over a number field \(k\). More precisely, Manin conjectured that there exists an open subset \(U\) of \(X\) such that \[ \sharp\{x\in U: H(x)\leq B\} \sim C_{X,H,k} B(\log B)^{r-1}, \] where \(H(x)\) is an anticanonical height function, \(r\) the rank of the Picard group of \(X\), and \(C_{X,H,k}\) a constant that received an interpretation by \textit{E. Peyre} [Duke Math. J. 79, No. 1, 101--218 (1995; Zbl 0901.14025)]. Manin's conjecture has been proved for smooth projective toric varieties over any number field by \textit{V. V. Batyrev} and \textit{Y. Tschinkel} [J. Algebr. Geom. 7, No. 1, 15--53 (1998; Zbl 0946.14009)] using methods from harmonic analysis. In [in: Nombre et répartition de points de hauteur bornée. Paris: Société Mathématique de France. 91--258 (1998; Zbl 0959.14007)] \textit{P. Salberger} introduced a new idea to attack Manin's conjecture via the socalled `universal torsor method'. Using this method he gave a proof of Manin's conjecture for complete smooth split toric varieties over \(\mathbb{Q}\) with anticanonical sheaf generated by global sections. The paper under review builds on the work of Salberger and generalizes the method to imaginary quadratic number fields. Manin's conjecture is proved for smooth equivariant compactifications \(X\) of a split torus \(T\) over an imaginary quadratic number field \(k\), such that the anticanonical sheaf of \(X\) is generated by its global sections. The leading constant is shown to coincide with the prediction by Peyre. There are at least two main difficulties in generalizing the universal torsor method to number fields. On the one hand, if the class number of \(k\) is larger than one, then there may be more than one isomorphism class of torsors, whereas in the case of split toric varieties over the rational numbers there is only one isomorphism class of torsors. Secondly, the fibers of a parametrization \(\tilde{\pi}: \tilde{\mathcal{T}}\rightarrow \tilde{X}\) of a proper model \(\tilde{X}\) over the ring of integers \(\mathcal{O}_k\), are either empty or isomorphic to \((\mathcal{O}_k^\times )^r\). In the case of \(k=\mathbb{Q}\) or an imaginary quadratic number field, one can hence count points on the preimage of \(\tilde{\pi}\) and divide by \(\omega_k^r\). If \(\mathcal{O}_k^\times\) is not finite, one is often lead to lattice counting problems in unbounded regions. The first part of the paper under review still allows for general number fields, which includes finding parametrizations via universal torsors, explicit expressions for the height function, and the introduction of a generalized Möbius function (similarly as in [Peyre, loc. cit.] and [Salberger, loc. cit.]). Starting at the process of Möbius inversion, which removes coprimality conditions in the counting argument, the paper specializes to imaginary quadratic number fields. The final part of the proof then proceeds via lattice point counting arguments, again building on Salberger's work [loc. cit.].