Some additive combinatorics problems in matrix rings

From MaRDI portal



Abstract: We study the distribution of singular and unimodular matrices in sumsets in matrix rings over finite fields. We apply these results to estimate the largest prime divisor of the determinants in sumsets in matrix rings over the integers.


Let \({\mathbb F}_q\) be the finite field with \(q\) elements, \({\mathcal M}_n({\mathbb F}_q)\) the set of matrices with entries in \({\mathbb F}_q\), \(\text{Gl}_n({\mathbb F}_q)\) the group of invertible matrices, \(\text{Sl}_n({\mathbb F}_q)\) the group of matices of determinant \(1\) and \({\mathcal Z}_n({\mathbb F}_q)\) the set of singular matrices, \(n\geq 2\). For two sets \({\mathcal A},{\mathcal B}\subset{\mathcal M}_n({\mathbb F}_q)\) let \[ N_{n,q}({\mathcal A},{\mathcal B})=\#\{A+B\in{\mathcal Z}_n({\mathbb F}_q):A\in{\mathcal A},B\in{\mathcal B} \}, \] \[ T_{n,q}({\mathcal A},{\mathcal B})=\#\{A+B\in\text{Sl}_n({\mathbb F}_q):A\in{\mathcal A},B\in{\mathcal B} \}. \] The authors prove that \[ \left|N_{n,q}({\mathcal A},{\mathcal B})-\frac{\#{\mathcal Z}_n({\mathbb F}_q)\#{\mathcal A}\#{\mathcal B}}{q^{n^2}}\right|=O(q^{n^2-\frac{5}{2}}\sqrt{\#{\mathcal A}\#{\mathcal B}}) \] and that \[ \left|T_{n,q}({\mathcal A},{\mathcal B})-\frac{\#\text{Sl}_n({\mathbb F}_q)\#{\mathcal A}\#{\mathcal B}}{q^{n^2}}\right|=O(q^{n^2-2}\sqrt{\#{\mathcal A}\#{\mathcal B}}). \] For \({\mathcal A},{\mathcal B},{\mathcal C},{\mathcal D} \subset{\mathcal M}_n({\mathbb F}_q)\) and \(H\in {\mathcal M}_n({\mathbb F}_q)\) let \(R({\mathcal A},{\mathcal B},{\mathcal C},{\mathcal D};H) \) be the number of solutions to the equation \[ H=(A+B)(C+D),A\in{\mathcal A},B\in{\mathcal B},C\in{\mathcal C},D\in{\mathcal D}. \] Then, uniformly over all matrices \(H\in \text{Gl}_n({\mathbb F}_q)\), we have \[ R({\mathcal A},{\mathcal B},{\mathcal C},{\mathcal D};H)=\frac{1}{q^{n^2}}\#{\mathcal A}\#{\mathcal B}\#{\mathcal C}\#{\mathcal D}+O(q^{n^2-\frac12}\sqrt{\#{\mathcal A}\#{\mathcal B}\#{\mathcal C}\#{\mathcal D}}). \] Finally, given a set \({\mathcal T} \) of integers, let \[ {\mathcal M}_n({\mathcal T}):=\{T=(t_{ij})\in {\mathcal M}_n({\mathbb Z}):t_{ij}\in{\mathcal T}, 1\leq i,j\leq n\}. \] For a non-zero integer \(m\) let \( \omega(m)\) denote the number of its distinct prime factors. Then there exists a positive constant \(c_0(n)\) such that if \({\mathcal A},{\mathcal B}\) are subsets of \(\{1,\ldots,N\}\) with \[ \min\{\#{\mathcal A},\#{\mathcal B}\}>c_0(n)(\log N)^{\frac{2n^2}{3}-1}(\log\log N)^{\frac{2n^2}{3}} \] and \(N\) is sufficiently large, then \[ \omega(W({\mathcal A},{\mathcal B}))\gg \min\{\#{\mathcal A},\#{\mathcal B}\}, \] where \[ W({\mathcal A},{\mathcal B}):=\prod_{A\in{\mathcal M}_n({\mathcal A})}\prod_{B\in{\mathcal M}_n({\mathcal B})}\det(A+B). \]











This page was built for publication: Some additive combinatorics problems in matrix rings

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1958673)