On the Glasner property for matrices with polynomial entries (Q2079494)

Let $\mathbb{T}=\mathbb{R}/\mathbb{Z}$ be the unit torus identified with $[0,1)$. \textit{S. Glasner} [Isr. J. Math. 32, 161--172 (1979; Zbl 0406.54023)] showed that for any infinite set $X\subseteq\mathbb{T}$ and any $\varepsilon>0$, there is an $n\in\mathbb{N}$ such that $nX$ is $\varepsilon$-dense in $\mathbb{T}$. Recently, \textit{K. Bulinski} and \textit{A. Fish} [``Glasner property for unipotently generated group actions on tori'', Isr. J. Math. (to appear)] generalized Glasner's result by considering dilations $A(n)X$, where $X\subseteq\mathbb{T}^d=(\mathbb{R}/\mathbb{Z})^d=[0,1)^d$ and $A(x)$ is a $d\times d$ matrix of integer polynomials. In particular, they showed that for any $\varepsilon>0$, there is some $k_{d,A}(\varepsilon)$ such that if $X\subseteq\mathbb{T}^d$ is of cardinality at least $k_{d,a}(\varepsilon)$, then for some $n\in\mathbb{N}$ the dilation $A(n)X$ is $\varepsilon$-dense in $\mathbb{T}^d$, that is $$\sup_{\zeta\in\mathbb{T}^d}\inf_{x\in X}\|\zeta-A(n)x\|\leqslant\varepsilon,$$ where $\|\cdot\|$ is the distance on $\mathbb{T}^d$ induced by the Euclidean norm. Further, they have the bound \[ k_{d,A}(\varepsilon)\leqslant c(d,e)H^{d(d+1)}\varepsilon^{-d(d+1)(2e+1)+o(1)}\text{ as }\varepsilon\to 0, \] where $e$ and $H$ are the largest degree and absolute value of the non-constant coefficients of the polynomial entries of $A(x)$, respectively, and $c(d,e)$ is a positive constant depending only on $d$ and $e$. In this paper, the author improves the bound on $k_{d,A}(\varepsilon)$ to \[ k_{d,A}(\varepsilon)\leqslant c(d,e)H^{(3d+1)/2+o(1)}\varepsilon^{-d(2d+1)e-(7d+1)/2+o(1)}\text{ as }\varepsilon\to 0. \] Here, it is not so much the better bound, but the author's method that will probably be of use to most readers. While Bulinski's and Fish's bound is based on the classical Hua bound of complete rational exponential sums, the current paper exploits the fact that the set of moduli for which Hua's bound is optimal is rather sparse. In particular, it uses more refined information about complete rational exponential sums and some well-known results about the arithmetic structure of the integers.