Density of rational points on a quadric bundle in \(\mathbb{P}^3 \times \mathbb{P}^3 \) (Q2217894)

In the study of the rational points of smooth projective varieties over number fields, Fano varieties (i.e. varieties with ample anti-canonical bundle) sits on the opposite side of the hyperbolic ones: while the latter are expected to have very few rationals points, the former have lot of them. Hence, given a smooth projective Fano variety \(X\) over \(\mathbb Q\), one can try to estimates how big the set \(X(\mathbb Q)\) is. This is what Manin's conjecture is about. To make this estimate precise, one embeds \(X\subseteq \mathbb P^n\) via some multiple of the anticanonical bundle \(\omega_X\) and looks at the induced height function \(h_{X}:X(\mathbb Q)\rightarrow \mathbb R_{\geq 0}.\) Since \(h\) satisfies the Northcott's property, for every subset \(\Omega\) of \(X(\mathbb Q)\) and every \(B\in \mathbb N\) the set \[\{x\in \Omega \text{ such that } h(x)\leq B\}\] is finite, so that its cardinality \(N(\Omega,B)\) is a positive integer. Manin's conjecture predicts that there exists a thin subset \(\Omega\subseteq X(\mathbb Q)\) and a \(c\in \mathbb R_{\geq 0}\) such that \[N(X-\Omega,B)\sim cB(\log(B))^{\rho-1} \quad \text{as } B\to +\infty,\] where \(\rho\) is the Picard rank of \(X\). A refinement of the conjecture has been proposed by Peyre, who gave a conjectural interpretation constant \(c\) in terms of the geometry of \(X\). In the paper under review, the authors prove the refined conjecture for the biprojective hyper-surface \(X\subseteq \mathbb P^3\times \mathbb P^3\) cut out by the equation \[F(x,y):=x_1y^2_1+x_2y^2_2+x_3y^2_3+x_4y^2_4=0\] and the thin subset \[\Omega:=\{(x,y)\in X(\mathbb Q) \text{ such that } x_1x_2x_3x_4 \text{ is a square in } \mathbb Q \}.\] Observe that for every \(x=[x_1:x_2:x_3:x_4]\not \in \Omega\), the quadric surface \(Q_x:=x_1y^2_1+x_2y^2_2+x_3y^2_3+x_4y^2_4=0\subseteq \mathbb P^3\) is not isomorphic to \(\mathbb P^1\times \mathbb P^1\) and hence is has ``less'' \(\mathbb Q\)-rational points than it could.