On best simultaneous approximations (Q4375241)

Let \(\alpha= (\alpha_1, \dots \alpha_s) \in\mathbb{R}^s\). An integer point \(\zeta= (p, a_1, \dots, a_s)\in \mathbb{Z}^{s+1}\) is said to be the best simultaneous approximation to \(\alpha \) if \[ D(\zeta)= \max_{1\leq j\leq s} | p\alpha_j- a_j| <\max_{1\leq j\leq s} | q \alpha_j- b_j | \] for all \(q\) with \(1\leq q\leq p\) and all \((b_1, \dots, b_s)\in\mathbb{Z}^s \setminus \{(a_1, \dots, a_s)\}\). Suppose that all the numbers \(\alpha_j\) are irrational. Let \(\dim_\mathbb{Z} \alpha\) be the maximal number of elements of sets \(\{\alpha_{i_1}, \dots, \alpha_{i_m}\} \subseteq \{1,\alpha_1, \dots, \alpha_s\}\) such that \(\alpha_{i_1}, \dots, \alpha_{i_m}\) are linearly independent over \(\mathbb{Z}\). Arrange all the best simultaneous approximations to \(\alpha\) as an infinite sequence \(\zeta^{(\nu)} =(p^{(\nu)}, a_1^{(\nu)}, \dots, a_s^{(\nu)})\), \(\nu=1,2, \dots\), such that \(p^{(1)}< \cdots <p^{(\nu)} <p^{(\nu+1)} <\dots\) and \(D(\zeta^{(1)}) >\cdots >D(\zeta^{(\nu)}) >D(\zeta^{ (\nu+1)}) >\cdots\), and then set \[ M_\nu[\alpha] =\left(\begin{matrix} p^{(\nu)} & a_1^{(\nu)} & \dots & a_s^{(\nu)}\\ \dots& \dots & \dots & \dots\\ \\ p^{(\nu+s)} & a_1^{(\nu+s)} & \dots & a_s^{(\nu+s)} \end{matrix}\right). \] The author announces with a sketch of proof the following: Let \(s\geq 3\). Then there exists a set of vectors \(\alpha= (\alpha_1, \dots, \alpha_s)\) with cardinal number of the continuum such that \(1,\alpha_1, \dots, \alpha_s\) are linearly independent over \(\mathbb{Z}\) and the rank of \(M_\nu[\alpha]\leq 3\) for all \(\nu\in \mathbb{N}\). (In particular, \(\text{det} M_\nu [\alpha]=0\) for all \(\nu\), whereas \(\dim_\mathbb{Z} \alpha= s+1\).) This result provides a counterexample to a conjecture of \textit{J. C. Lagarias} [Pac. J. Math. 102, 61-88 (1982; Zbl 0497.10025)].