Khintchine-type theorems for values of subhomogeneous functions at integer points (Q2659842): Difference between revisions

Given two functions \(f=\mathbb{R}^n\to\mathbb{R}^\ell\), \(\psi:\mathbb{R}_{\ge 0}\to (\mathbb{R}_{>0})^\ell\) and a norm \(\nu\) on \(\mathbb{R}^n\), define \(A_{f,\psi,\nu}=\{\mathbf{x}\in{\mathbb{R}}^n\mid |f(\mathbf{x})|\le\psi(\nu(\mathbf{x}))\}\). For \(T\in\mathbb{R}_{>0}\) and \(\boldsymbol{\varepsilon}\in(\mathbb{R}_{>0})^\ell\), let \(B_{f,\boldsymbol{\varepsilon},\nu,T}=\{\mathbf{x}\in\mathbb{R}^n\mid |f(\mathbf{x})|\le{\boldsymbol{\varepsilon}},\nu({\mathbf{x}})\le T\}\). The Lebesgue measure on a Euclidean space is denoted by \(m\). The authors assume \(f\) to be subhomogeneous, meaning that \(f\) is Borel measurable and there exists a constant \(d=d_f\in\mathbb{R}_{\ge 0}\) such that for each \(t\in(0,1)\) and each \({\mathbf{x}}\in{\mathbb{R}}^n\) one has \(|f(t{\mathbf{x}})|\le t^d|f(\mathbf{x})|\). Further, they assume that \(\psi\) is regular, meaning that \(\psi\) is Borel measurable and there exist real numbers \(a=a_\psi\in\mathbb{R}_{>1}\) and \(b=b_\psi\in{\mathbb{R}}_{>1}\) such that for each \(z\in\mathbb{R}_{>0}\) one has \(b_\psi \psi(z)\le \psi(a_\psi z)\). Furthermore, they assume that each component of \(\psi\) is non increasing. The first theorem of the paper under review is the following special case of their main results. (i) If \(m(A_{f,\psi,\nu})\) is finite (resp. infinite), then for Haar almost no (resp. almost every) \(g\in\mathrm {SL}_n(\mathbb{R})\), \(f\circ g\) is \((\psi,\nu)\) approximable, which means that \({\mathrm{Card}}\big({\mathbb Z}^n\cap A_{f,\psi,\nu}\bigr)=\infty\). (ii) Suppose \[ \sum_{k=1}^\infty \frac 1 {m(B_{f,\psi(2^k),\nu,2^k})}<\infty. \] Then for Haar almost every \(g\in\mathrm{SL}_n(\mathbb{R})\), \(f\circ g\) is uniformly \((\psi,\nu)\) approximable, which means \(B_{f,\psi(T),\nu,T}\cap({\mathbb Z}^n\setminus\{0\})\not=\emptyset\) for each sufficiently large \(T\in\mathbb R_{>0}\). This yields, in a very general framework, a necessary and sufficient condition for approximation and a sufficient condition for uniform approximation. The authors give a number of consequences of their results. For instance, in the special case where \(n=p+q\) with \(p\) and \(q\) in \(\mathbb{Z}_{\ge 1}\) and with \[ f(\mathbf{x})=\sum_{j=1}^p |x_j|^d-\sum_{k=p+1}^n |x_k|^d, \] they generalize earlier work by [\textit{J. S. Athreya} and \textit{G. A. Margulis}, J. Mod. Dyn. 12, 9--16 (2018; Zbl 1454.11126)] and [\textit{D. Kelmer} and \textit{S. Yu}, Trans. Am. Math. Soc. 373, No. 12, 8677--8695 (2020; Zbl 1477.11123)], where \(d\) was assumed to be even. The method of the two last mentioned papers as well as of the paper under review has its origin in [\textit{C. A. Rogers}, Acta Math. 94, 249--287 (1955; Zbl 0065.28201); Proc. Lond. Math. Soc. (3) 6, 305--320 (1956; Zbl 0071.27403); \textit{W. M. Schmidt}, Trans. Am. Math. Soc. 123, 178--199 (1966; Zbl 0232.10029)]. The last section \textit{Concluding Remarks} of the paper under review suggests further developments: inhomogeneous approximation, counting the number of solutions, metric number theory.