Dyson's lemma for polynomials in several variables (and the theorem of Roth) (Q797625)

The theorem of \textit{K. F. Roth} [Mathematika 2, 1--20 (1955; Zbl 0064.28501)] asserts that for every algebraic number \(\alpha\) there are finitely many good approximations \(p/q\in\mathbb{Q}\) verifying \(| \alpha -p/q| \leq(1/q)^k\) provided \(k>2\). (Weaker results on \(k\) where obtained by Thue, Siegel, Dyson and Gelfond). E. Bombieri reconsidered and generalized recently Dyson's approach to handle auxiliary polynomials in two variables in order to obtain, for certain \(\alpha\), effective bounds on \(q\) for the exponent \(k>(2\cdot \deg \alpha)^{1/2}\) of Dyson. In our paper we generalize the Bombieri-Dyson approach to polynomials in several variables. Roughly speaking we prove that if an hypersurface in \(\mathbb{C}^n\) has very bad singularities in (\(\alpha,\ldots,\alpha)\) and all its Galois conjugates, it has ``nearly'' no singularity in an approximation point \((p_1/q_1,\ldots,p_n/q_n)\). The estimates given are strong enough to imply the theorem of Roth. The proof of the generalized ``lemma of Dyson'' is using deep results from the projective geometry: positivity for direct images of dualizing sheaves (used before to study the Kodaira dimension of fiber spaces) and Kodaira type vanishing theorems of \(\mathbb{Q}\)-divisors.