scientific article; zbMATH DE number 4023387
In 1909, Thue showed that for each algebraic number \(\alpha\) of degree \(n\), there are only finitely many rational numbers \(x/y\) \((x,y\in \mathbb{Z}\), \(y>0)\) such that \(| \alpha -x/y| <y^{-(n+1+\varepsilon)}\) for every \(\varepsilon >0\). This was the first improvement of Liouville's inequality \(| \alpha -x/y| >c(\alpha)y^{-n}\) for all \(x/y\in \mathbb{Q}\). Thue's ideas were later extended and generalized by Siegel, who proved Thue's result with the exponent \(2n^{1/2}+\varepsilon\) instead of \(n+\varepsilon,\) Dyson and Gel'fond, who reduced the exponent to \((2n)^{1/2}+\varepsilon,\) and Roth, who reduced the exponent to the best possible value \(2+\varepsilon.\) The author gives a very useful overview of the extensions of Thue's method, and sketches the mathematical ideas behind his own refinement of Dyson's method [Acta Math. 148, 255--296 (1982; Zbl 0505.10015)]. We briefly describe the contents of the paper under review. First the author describes some necessary tools in diophantine approximation theory, such as heights and Siegel's lemma. Using Siegel's lemma, he constructs the auxiliary polynomial needed in Dyson's method. With the help of this polynomial, and with Dyson's lemma guaranteeing a low order of vanishing of this polynomial at certain prescribed points, the author proves the main theorem. This roughly states that if some algebraic number \(\alpha\) has a good rational approximation \(u/v\), then no other number \(\beta\) in the field \(\mathbb{Q}(\alpha)\) can have good rational approximations. (In fact, the author proves a more general result in which algebraic numbers are approximated by elements of a given number field, and in which approximation is considered also with respect to non-Archimedean valuations.) The most striking consequence of this main theorem is that in certain cases it gives a good effective improvement of Liouville's inequality: once a good rational approximation \(u/v\) of \(\alpha\) is known, say \(| \alpha -u/v| < c(\alpha)v^{-\eta_1}\) where \(c(\alpha)\) is some explicit expression in \(\alpha\) and \(2<\eta_1<n\), then for every \(\beta\in {\mathbb{Q}}(\alpha)\) there are effectively computable numbers \(c_1\) and \(c_2\), depending on \(\alpha\), \(\beta\), \(u/v\), and \(\eta_ 2\) approximately equal to \(2n/\eta_1\), such that for all \(x/y\in \mathbb{Q}\) with \(y>c_1\), \(| \beta -x/y| > c_2y^{-\eta_ 2}.\) For instance, if \(\alpha\) is the root of \(x^n-mx+1 = 0\) closest to \(m^{-1}\) then for sufficiently large \(n\) and \(m\), \(\eta_2\) can be taken 13.21. At the end of the paper, the author discusses some generalizations of Dyson's lemma, such as the result of \textit{H. Esnault} and \textit{E. Viehweg} [Invent. Math. 78, 445--490 (1984; Zbl 0532.10020)] providing a new proof for Roth's result. [For the entire collection see Zbl 0618.00005.]
- Local positivity and effective Diophantine approximation
- scientific article; zbMATH DE number 3857222 (Why is no real title available?)
- scientific article; zbMATH DE number 165242 (Why is no real title available?)
- scientific article; zbMATH DE number 1210339 (Why is no real title available?)
- scientific article; zbMATH DE number 2150709 (Why is no real title available?)
- On Inhomogeneous Extension of Thue–Roth’s Type Inequality with Moving Targets
- Proof of the Baker-Feldman theorem via linear forms in two logarithms
- Dyson's lemma for polynomials in several variables (and the theorem of Roth)
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3765855)