On algebraic numbers of small height: Linear forms in one logarithm (Q1323868): Difference between revisions

The authors establish a lower bound for \(|\alpha -1|\) where \(\alpha\) is a complex algebraic number \(\neq 1\). This lower bound is completely explicit in terms of the degree \(D\) of \(\alpha\) and its Mahler measure \(M(\alpha)\). Given any number \(\mu>0\) with \(\mu\geq \log M(\alpha)\), the authors prove \[ |\alpha -1|\geq \exp \Bigl\{- \bigl( \textstyle{3\over 2} \sqrt{D\mu \log_ + (D/\mu)}+ 2\mu+ \log_ + (D/\mu) \bigr) \Bigr\}, \] where \(\log_ + x= \max\{ \log x,0\}\). From this inequality, they deduce \[ | \alpha-1 |\geq \max \{2, M(\alpha) \}^{-3 \sqrt{D\log D}}, \] which refines a result of \textit{M. Mignotte} [Ann. Fac. Sci. Toulouse 1, 165-170 (1979; Zbl 0421.10022)] and improves on Liouville's inequality concerning the dependency in \(D\). The above mentioned result follows from a more general theorem from which the authors also deduce an application to Lehmer's problem as well as the conjecture of \textit{A. Schinzel} and \textit{H. Zassenhaus} [Mich. Math. J. 12, 81-85 (1965; Zbl 0128.034)]. A corollary of this theorem which is sufficient for these applications is the following: Let \(E\) be a number field of degree \(D\) over \(\mathbb{Q}\), let \(\alpha_ 1, \dots, \alpha_ K\) be distinct elements of \(E\) with \(K\geq 2\), and let \(G'\) be a set of embeddings of \(E\) into \(\mathbb{C}\). For each \(\sigma\in G'\), choose \(\phi_ \sigma\in i\mathbb{R}\) and determinations of \(\log (\sigma\alpha_ k)\) for \(k=1,\dots, K\). Then the average values \[ \Lambda_ \sigma= \textstyle {1 \over K} \sum_{k=1}^ K |\log (\sigma \alpha_ k)+ \phi_ \sigma| \] satisfy \[ - \sum_{\sigma\in G'} \log \Lambda_ \sigma \leq {{2D} \over K} \sum_{k=1}^ K h(\alpha_ k)+ {D\over {K-1}} \biggl( 2+\log \biggl( {{2K+1} \over 4} \biggr) \biggr)+ {K\over {K-1}} | G'|, \] where \(h(\beta)\) stands for the absolute logarithmic Weil height of \(\beta\) and \(| G'|\) denotes the cardinality of \(G'\). The method of proof uses the technique of interpolation determinants of \textit{M. Laurent} [Acta Arith. 66, 181- 199 (1994; see the review below)].