Generalized radix representations and dynamical systems. III (Q935570)

For \(\mathbf r = (r_{1},\dots , r_d)\in \mathbb R^d\) the mapping \(\tau _{\mathbf r}:\mathbb Z^d \rightarrow \mathbb Z^d\) given by \[ \tau _{\mathbf r}(a_{1},\dots ,a_d) = (a_{2}, \dots , a_d, - \lfloor r_{1}a_{1}+\dots + r_da_d\rfloor), \] where \(\lfloor\cdot \rfloor\) denotes the floor function, is called a shift radix system if for each \(\mathbf a\in \mathbb Z^d\) there exists an integer \(k > 0\) with \(\tau _{\mathbf r}^k(\mathbf a) = 0\). The authors have studied shift radix systems and their connections to \(\beta\)-expansions and canonical number systems in previous parts of this series of papers [\textit{S. Akiyama, T. Borbély, H. Brunotte, A. Pethő} and \textit{J. M. Thuswaldner}, ``Generalized radix representations and dynamical systems. I'', Acta Math. Hung. 108, No. 3, 207--238 (2005; Zbl 1110.11003); \textit{S. Akiyama, H. Brunotte, A. Pethő} and \textit{J. M. Thuswaldner}, ``Generalized radix representations and dynamical systems. II'', Acta Arith. 121, No. 1, 21--61 (2006; Zbl 1142.11055)]. Let \({\mathcal D_{d}}\) and \(\mathcal D_{d}^0\) be the sets of vectors \(\mathbf r\) in \({\mathbb R}^{d}\) such that for all \({\mathbf a}\in{\mathbb Z}^{d}\), the sequence \((\tau^k_{{\mathbf r}}({\mathbf a}))_{k\geq 0}\) is ultimately periodic resp.\ reaches \(0\). Set \({\mathcal G}=\{A\in{\mathbb Z}[X]: \deg A<d\}\). For a polynomial \(P=p_dX^d+\cdots+p_0\in{\mathbb Z}[X]\) with \(p_0\geq 2\) and \(p_d=1\), consider the map \(T_P:{\mathcal G}\to{\mathcal G}\) mapping a polynomial \(A=\sum_{j=0}^{d-1}A_jX^j\) to \(T_p(A)=\sum_{i=0}^{d-1}(A_{i+1}-qp_{i+1})X^i\) with \(A_d=0\) and \(q=\lfloor A_0/p_0\rfloor\), resulting in \(A=(A_0-qp_0)+X T_P(A)\) with \(A_0-qp_0\in{\mathcal N}=\{0,\dots,p_0-1\}\). The polynomial \(P\) is called a canonical number system polynomial if for each \(A\in{\mathcal G}\), there exists an integer \(k>0\) with \(T_P^k(A)=0\). The set \({\mathcal C}_d^0\) consists of all \((p_0,\dots,p_{d-1})\in{\mathbb Z}^d\) with \(|p_0|\geq 2\) such that the corresponding polynomial \(P=X^d+\sum_{i=0}^{d-1}p_jX^j\) is a canonical number system polynomial; on the other hand, the set \({\mathcal C}_d^0\) consists of all \((p_0,\dots,p_{d-1})\in{\mathbb Z}^d\) with \(|p_0|\geq 2\) such that for the corresponding polynomial \(P=X^d+\sum_{i=0}^{d-1}p_jX^j\), the map \(T_P\) has only finite orbits. The authors consider the sets \[ \begin{aligned} {\mathcal C}_d(M)&=\{ (p_{d-1}/M,\dots,p_1/M)\in{\mathbb R}^{d-1}: (M,p_1,\dots,p_{d-1})\in{\mathcal C}_d\},\\ {\mathcal D}^0_d(x)&=\{(r_2,\dots,r_d)\in{\mathbb R}^{d-1}: (x,r_2,\dots,r_d)\in{\mathcal D}^0_d\}. \end{aligned} \] The authors prove that \(\text{Lim}_{M\to\infty} {\mathcal C}_d(M)=\overline{{\mathcal D}_{d-1}}\) holds for all \(d\geq 2\), where \(\text{Lim}\) denotes the topological limit. Furthermore, for \(d\geq 2\) and \(M\) a positive integer, they prove that \[ \lim_{M\to\infty} {|\{(p_1,\dots,p_{d-1})\in{\mathbb Z}^{d-1}: (M,p_1,\dots,p_{d-1})\in{\mathcal C}_d\}|\over M^{d-1}}=\lambda_{d-1}({\mathcal D}_{d-1}), \] where \(\lambda_{d-1}\) denotes the \((d-1)\)-dimensional Lebesgue measure. Furthermore, they show that \(\lim_{x\to0} \lambda_{d-1}({\mathcal D}_d^0(x) \Delta {\mathcal D}_{d-1}^0)=0\) and that \[ \lim_{M\to\infty} {|\{(p_1,\dots,p_{d-1})\in{\mathbb Z}^{d-1}: (M,p_1,\dots,p_{d-1})\in{\mathcal C}^0_d\}|\over M^{d-1}}=\lambda_{d-1}({\mathcal D}^0_{d-1}). \]