Rational solutions of certain equations involving norms (Q1812468)

The main object of the paper under review is a Diophantine equation of the form \(P(t)=N(x_1,\dots ,x_n)\) where \(P(t)\) is a polynomial with coefficients in a number field \(k\) and \(N\) is the norm form of a degree \(n\) extension \(K/k\). The authors are interested in the Hasse principle and density of solutions in \(k\) of the above equation. They consider the case where \(P(t)\) has exactly two roots in \(k\) and no other roots in an algebraic closure \(\bar k\). In this case the equation can be brought to the form \[ t^{a_0}(1-t)^{a_1}=\alpha N(x_1,\dots ,x_n) \tag{*} \] where \(\alpha\in k^*\), and \(a_0\) and \(a_1\) are positive integers. The main result of the paper says that if \(k=\mathbb Q\) and \((a_0,a_1,n)=1\), the Brauer-Manin obstruction is the only obstruction to the Hasse principle and weak approximation on any smooth and projective model of the variety \(X\) given by (*). Moreover, if there is no Brauer-Manin obstruction to the Hasse principle on such a model, then the \(\mathbb Q\)-rational points are Zariski dense in \(X\). Note that the projection to the coordinate \(t\) defines on \(X\) a structure of fibration whose generic fibre is a principal homogeneous space under a norm torus. There is a general question posed in \S 4 of the paper by \textit{J.-L. Colliot-Thélène} and \textit{Sir P. Swinnerton-Dyer} [J. Reine Angew. Math. 453, 49-112 (1994; Zbl 0805.14010)] whether the Brauer-Manin obstruction is the only one for fibrations whose generic fibre is a homogeneous space under a connected linear algebraic group. The fibration considered in the paper under review has at most three degenerate fibres (at 0, 1 and \(\infty\)). Thus it is, in a sense, the first instance where the known methods (descent and fibration) break down being applicable to the case of at most two degenerate fibres. The proof consists of two parts. First, the authors construct a universal torsor over the smooth locus of \(X\) and bring it to the form \[ r_0N(y_1,\dots ,y_n)+r_1n(z_1,\dots ,z_n)=1. \] Second, they use the Hardy-Littlewood circle method in the form of \textit{B. J. Birch, H. Davenport} and \textit{D. J. Lewis} [Mathematika 9, 75-82 (1962; Zbl 0105.02804)] to prove that this torsor satisfies the Hasse principle and weak approximation. Thus the paper under review presents the first example of successful combination of the method of descent with the circle method. Remark. In a recent paper by \textit{J.-L. Colliot-Thélène, D. Harari} and \textit{A. Skorobogatov} [Valeurs d'un polynôme à une variable représentées par une norme, Preprint, 2003, available per Internet at \url{http://www.math.u-psud.fr/~colliot}] the hypothesis \((a_0,a_1,n)=1\) has been removed. The authors used the method of ``open descent'' applied to a partial compactification of the variety given by (*).