Zeros of pairs of quadratic forms (Q1637083)

Consider two quadratic forms \(Q_i(X_1,\ldots,X_n)\), \(i=1,2\), in \(n\) variables over a number field \(k\). The present paper addresses two questions. The first question asks if the Hasse principle holds for the isotropy of the pair \((Q_1,Q_2)\), i.e., do these two forms have a nontrivial common zero over \(k\) if they have a nontrivial common zero over each completion of \(k\)? The second question asks if the weak approximation property holds, i.e., if the set of rational points of this system is not empty, does this imply that this set is dense in the set of rational \(S\)-adelic points where \(S\) is any finite set of places of \(k\)? \textit{J.-L. Colliot-Thélène} et al. [J. Reine Angew. Math. 373, 37--107 (1987; Zbl 0622.14029)] have shown that if \(n\geq 9\), then the Hasse principle holds provided each local zero be nonsingular, and, still for \(n\geq 9\), that also the weak approximation principle holds under the assumption that the projective variety \(\mathcal{V}\) defined by the equation \(Q_1( {\mathbf{X}})=Q_2({\mathbf{X}})=0\) be nonsingular, by which one means that \[ \text{rk}\begin{pmatrix} \nabla Q_1(\mathbf x)\\ \nabla Q_2(\mathbf x)\end{pmatrix}=2 \] for each \({\mathbf{x}}\in\overline{k}^n\setminus\{ 0\}\) with \(Q_1( {\mathbf{x}})=Q_2({\mathbf{x}})=0\). For \(n=5\), there are examples of pairs \((Q_1,Q_2)\) where the intersection \(\mathcal{V}\) is smooth but the Hasse principle (resp. the weak approximation principle) fails. However, \textit{J.-L. Colliot-Thélène} et al. [J. Reine Angew. Math. 374, 72--168 (1987; Zbl 0622.14030)] conjectured that the Hasse principle holds for \(n\geq 6\) provided that the intersection be nonsingular. The present paper provides an important contribution toward a solution of this conjecture. It is shown that if \((Q_1,Q_2)\) is a system of forms of \(8\) variables with nonsingular intersection over a number field \(k\), then the Hasse principle as well as the weak approximation principle hold. The main new ingredient that allows the author to extend the methods of Colliot-Thélène, Sansuc and Swinnerton-Dyer is the following purely local result. Let \(k_v\) be the completion of a number field \(k\) at a finite place \(v\) and suppose that the residue field has \(\geq 32\) elements. Let \((Q_1,Q_2)\) be a pair of quadratic forms in \(8\) variables over \(k_v\) with nonsingular intersection. Then there exist \(a_1,a_2\in k_v\) such that the quadratic form \(a_1Q_1+a_2Q_2\) splits off (at least) three hyperbolic planes.