On Diophantine exponents and Khintchine's transference principle (Q1946750)

Consider a system of linear equations \(\Theta\mathbf x=\mathbf y\) with \(\mathbf x\in \mathbb R^m\), \(\mathbf y\in \mathbb R^n\) and \(\Theta=(\theta_{ij})\) with \(\theta_{ij}\in\mathbb R\). Let \(\Theta^T\) the transposed matrix and the ``transposed'' system \(\Theta^T\mathbf y=\mathbf x\). The \textit{transference theorems} deal with the relations between the solutions of these two systems. The supremum of real numbers \(\gamma\), such that there are infinitely many \(x\in \mathbb Z^m, y \in \mathbb Z^n\) satisfying the inequality \(|\Theta \mathbf x-\mathbf y|_\infty\leq |\mathbf x|_\infty^{-\gamma}\) where \(|\cdot|_\infty\) denotes the sup-norm in the corresponding space, is called the \textit{individual diophantine exponent} of \(\Theta\) and is denoted by \(\beta(\Theta)\). The supremum of real numbers \(\gamma\), such that for each \(t\) large enough there are \(\mathbf x \in \mathbb Z^m\), \(\mathbf y\in \mathbb Z^n\) satisfying the inequalities \(0 < |x|_\infty \leq t, |\Theta\mathbf x -\mathbf y|_\infty\leq t^{-\gamma},\) is called the \textit{uniform diophantine exponent} of \(\Theta\) and is denoted by \(\alpha(\Theta)\). Results: The aim of this paper (also available at \url{arXiv:1004.4933}) is to prove some new inequalities connecting the quantities \(\alpha(\Theta), \alpha(\Theta^T), \beta(\Theta), \beta(\Theta^T)\). The author generalizes some results of \textit{V. Jarník} [Acta Arith. 2, 1--22 (1936; Zbl 0015.29405; JFM 62.1158.04)] and \textit{A. Apfelbeck} [Czech. Math. J. 1(76), 119--147 (1952; Zbl 0048.27802)] and he finds: For all positive integers \(n, m\), not equal simultaneously to \(1\), we have \(\alpha(\Theta^T)\geq \frac{n-1}{m-\alpha(\Theta)}\) if \(\alpha(\Theta)\leq 1\) and \(\alpha(\Theta^T)\geq \frac{n-\alpha(\Theta)^{-1}}{m-1}\) if \(\alpha(\Theta)\geq 1\). The author finds the following refinement of some results of \textit{Y. Bugeaud} and \textit{M. Laurent} [``On transfer inequalities in Diophantine approximation. II'', Preprint, \url{arXiv:0811.2102}] and \textit{A. Khintchine} [Rend. Circ. Mat. Palermo 50, 170--195 (1926; JFM 52.0183.01)]. For all positive integers \(n, m\), not equal simultaneously to \(1\), there are three inequalities: \[ \begin{aligned} & \beta(\Theta^T)\geq\frac{n\beta(\Theta)+n-1}{(m-1)\beta(\Theta)+m}\\ & \beta(\Theta^T)\geq\frac{(n-1)(1+\beta(\Theta))-(1-\alpha(\Theta))}{(m-1)(1+\beta(\Theta))+(1-\alpha(\Theta))}\\ & \beta(\Theta^T)\geq\frac{(n-1)(1+\beta(\Theta)^{-1})-(\alpha(\Theta)^{-1}-1)}{(m-1)(1+\beta(\Theta)^{-1})+(\alpha(\Theta)^{-1}-1)}\end{aligned} \] The author improves also a strong version of a \textit{Kintchine's transference theorem} due to \textit{K. Mahler} [Mat. Sb., Nov. Ser. 26(68), 457--462 (1950; Zbl 0038.19001)]. In the last section, he uses his method to work not only with the diophantine exponents, but with arbitrary functions satisfying some natural growth conditions.