Hausdorff dimension and exact approximation order in \(\mathbb{R}^n\) (Q6912166)

For a non-increasing function \(\psi:\mathbb{N} \rightarrow\mathbb{R}_+\), a point \(x \in\mathbb{R}^n\) is said to be \(\psi\)-approximable if there are infinitely many points \(p/q = (p_1/q, \dots, p_n/q) \in\mathbb{Q}^n\) such that \N\[\N \left\Vert x -\frac{p}{q}\right\Vert < \psi(q), \N\]\Nwhere the norm \(\Vert \cdot \Vert\) is the supremum norm. The set of such points is denoted \(W_n(\psi)\).\N\NThe paper under review is concerned with the set of points which are \(\psi\)-approximable, but which are not \(c \psi\)-approximable for any \(c < 1\), i.e. the set \N\[ \NE_n(\psi) = W_n(\psi) \setminus \bigcup_{c < 1} W_n(c \psi).\N\]\NThe authors first make the additional assumption that \(\psi(s) = o(s^{-(n+1)/n})\), which removes the limiting case where Dirichlet's classical approximation theorem is relevant. Answering a question of \textit{V. Beresnevich} et al. [Math. Ann. 321, No. 2, 253--273 (2001; Zbl 1006.11039)], it is shown that the Hausdorff dimension of \(E_n\) is equal to that of \(W_n(\psi)\).\N\NThe authors apply methods and results from parametric geometry of numbers. A weaker result follows from known results, but to get the full version, they need to strengthen known technology substantially.\N\NThe paper ends with a concluding section with nice open problems, where the methods of the paper seem relevant.