Geometry of the best approximations (Q1593980)

Let \(\alpha_1,\dots, \alpha_s\in \mathbb{R}\) be such that \(\alpha_1,\dots, \alpha_s,1\) are linearly independent over \(\mathbb{Z}\), where \(s\geq 1\). For any \(\vec n= (n_0, n_1,\dots, n_s)\in \mathbb{Z}^{s+1}\), \(\vec n\neq \vec 0\), define \(\zeta(\vec n)= n_0+ n_1\alpha_1+\cdots n_s\alpha_s\). A non-zero integer vector \(\vec m= (m_0, m_1,\dots, m_s)\) with \(M= \max_{0\leq j\leq s}|m_j|\) is called the best approximation if \[ \zeta(\vec m)= \min\Bigl\{ \zeta(\vec n)\mid \vec n\in \mathbb{Z}^{s+1},\;\vec n\neq \vec 0,\;\max_{0\leq j\leq s}|n_j|\leq M\Bigr\}. \] The linear forms \(\zeta(\vec m)\) and the quantities \(M\) corresponding to the best approximation form monotonic sequences \(\zeta_1> \zeta_2>\dots> \zeta_\nu> \zeta_{\nu+1}>\dots\), and \(M_1< M_2<\dots< M_\nu< M_{\nu+1}<\dots\), where \(\vec m_\nu= (m_{0,\nu},\dots, m_{s,\nu})\) is the \(\nu\)th vector of the best approximation \(\zeta_\nu= \zeta (\vec m_\nu)\) and \(M_\nu= \max_{0\leq j\leq s}|m_{j,\nu}|\). We define \(\Delta_\nu^s= \det(m_{i,\nu+j})_{0\leq i,j\leq s}\) for \(\nu= 1,2,\dots\;\). Thus we can prove that \(\Delta_\nu^1= \pm 1\) for every \(\nu\in \mathbb{N}\), and that there are infinitely many \(\nu\) such that \(\Delta_\nu^2\neq 0\). Basing on a previous work of the author [Mat. Zametki 58, 394-410 (1995; Zbl 0857.11037)], the following theorem is proved: For \(s\geq 3\), there are continuum many tuples \((\alpha_1,\dots, \alpha_s)\in \mathbb{R}^s\) having the following properties: (i) \(\alpha_1,\dots, \alpha_s,1\) are linearly independent over \(\mathbb{Z}\); (ii) For sufficiently large \(\nu\) the sequence of the best approximation vectors \(\vec m_\nu\) constructed for \(\vec\alpha= (\alpha_1,\dots, \alpha_s)\) is contained in a 3-dimensional sublattice of the integer lattice \(\mathbb{Z}^{s+1}\). In particular, \(\Delta_\nu^s= 0\) for all \(\nu\geq \nu_0(\vec\alpha)\).