On positional representation of integer vectors (Q2666919)

\textit{A. Vince} [Congr. Numerantium 98, 199--212 (1993; Zbl 0801.05023); SIAM J. Discrete Math. 6, No. 3, 501--521 (1993; Zbl 0826.52019)] showed that if \(M\) is an expansive integer \(m\times m\)-matrix, i.e., all its eigenvalues have absolute value \(>1\), then there exists a finite digit set \(\mathcal{D}\subset\mathbb{Z}^m\) such that every \(x\in\mathbb{Z}^m\) can be written in the form \(x=\sum_{k=0}^n M^kd_k\) with \(n\geq 0\) and \(d_k\in\mathcal{D}\) for \(k=0,\ldots ,n\). On the other hand, Vince showed that such a finite digit set \(\mathcal{D}\) cannot exist if \(M\) is not expansive. The authors consider the problem of representing integer vectors with sums as above where the matrix \(M\) is not necessarily expansive, but where negative powers of \(M\) are allowed. Given a finite subset \(\mathcal{D}\) of \(\mathbb{Z}^m\) with \(0\in\mathcal{D}\), and an \(m\times m\)-matrix \(M\) with integer entries and of determinant \(\Delta\not= 0\), define the set \[ \text{Fin}_{\mathcal{D}}(M):=\left\{ \sum_{k\in I} M^kd_k:\, \begin{array}{l}I\subset\mathbb{Z}\ \text{finite, } I\not=\emptyset,\\ d_k\in\mathcal{D}\ \text{for } k\in I\end{array}\right\}, \] i.e., the set of all vectors \(x\in\mathbb{Q}^m\) that can be expressed as a finite sum of terms \(M^kd_k\) with \(k\in\mathbb{Z}\) and \(d_k\in\mathcal{D}\). One easily verifies that \(\text{Fin}_{\mathcal{D}}(M)\subseteq\bigcup_{k\in\mathbb{Z}_{\geq 0}} \Delta^{-k}\mathbb{Z}^m\). The main result of the authors is that given \(M\) as above, there always exists a finite subset \(\mathcal{D}\subset\mathbb{Z}^m\) such that \(\mathbb{Z}^m\subset \text{Fin}_{\mathcal{D}}(M)\). The proof is constructive; it is based on a careful study of the Jordan Normal Form of \(M\). The authors also give a necessary and sufficient condition for \(\text{Fin}_{\mathcal{D}}(M)\) to be equal to \(\bigcup_{k\in\mathbb{Z}_{\geq 0}} \Delta^{-k}\mathbb{Z}^m\), in fact they show that if \(\mathcal{D}\) is such that \(\mathbb{Z}^m\subset \text{Fin}_{\mathcal{D}}(M)\), then \(\text{Fin}_{\mathcal{D}}(M)=\bigcup_{k\in\mathbb{Z}_{\geq 0}} \Delta^{-k}\mathbb{Z}^m\) if and only if there is a non-negative integer \(\ell\) such that all entries of \(M^{\ell}\) are divisible by \(\Delta\). In particular, if \(\det M=\pm 1\) and \(\mathcal{D}\) is such that \(\mathbb{Z}^m\subset \text{Fin}_{\mathcal{D}}(M)\), then \(\mathbb{Z}^m=\text{Fin}_{\mathcal{D}}(M)\).