Probabilistic effectivity in the subspace theorem

From MaRDI portal





The celebrated theorem of \textit{K. F. Roth} [Mathematika 2, 1--20 (1955; Zbl 0064.28501)] and its far reaching generalisation by \textit{W. M. Schmidt} [Diophantine approximation. Berlin-Heidelberg-New York: Springer-Verlag (1980; Zbl 0421.10019)] in Diophantine approximation are concerned with rational approximation of algebraic numbers or of linear forms with algebraic coefficients. For instance, Schmidt's theorem states that if \(\alpha_1, \dots, \alpha_n \in \mathbb{R}^n\) are linearly independent vectors with algebraic coordinates and \(\varepsilon >0\), then the solutions \(q \in \mathbb{Z}^n\setminus \{0\}\) to the inequality \N\[\N\prod_{i=1}^n \vert q \cdot \alpha_i \vert \le \Vert q \vert^{-\varepsilon} \N\]\Nlie in a finite union of proper subspaces of \(\mathbb{Q}^n\). Likewise, Roth's result states that for an algebraic irrational, there are finitely many solutions to \N\[\N\left\vert \alpha - \frac{p}{q} \right\vert \le \frac{1}{q^{2+\varepsilon}}.\N\]\NThe proofs of both theorems is ineffective. While it is possible to get a bound on the number of subspaces in the first case, or on the number of good rational approximations in the second case, the proofs and results does not tell us when to stop looking for solutions. The paper under review seeks to resolve this -- at least in a probabilistic way. Concretely, they obtain bounds on the density of the (vectors of) algebraic numbers for which the inequalities have solutions very large in height.\N\NThe statement of the main results of the paper is too technical to repeat here, and the proofs are very involved. It is worth mentioning that all involved constants are explicit, and as the authors remark can most certainly be optimised further.











This page was built for publication: Probabilistic effectivity in the subspace theorem

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6866754)