A finiteness condition for complex continued fraction algorithms

DOI10.1090/PROC/17380MaRDI QIDQ6896436zbMATH OpenFDO

Authors C. Kalle, Fanni M. Sélley, Jörg M. Thuswaldner

Publication date 21 November 2025

Published in Proceedings of the American Mathematical Society (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1090/PROC/17380

zbMATH Keywords

complex continued fractions finite range condition

Mathematics Subject Classification ID

Continued fractions and generalizations (11J70) Ergodic theory (37A99)

Review text

The paper is concerned with \textit{continued fraction algorithms} that allow a representation of real or complex numbers by digits (the \textit{partial quotients}) \(a_1,a_2,a_3,\ldots\) taken from an infinite set \N\[\Nz=\cfrac{1}{a_1+\cfrac{1}{a_2+\cdots}}. \N\]\NThe authors give a historical overview from Euclid's algorithm (iteration of the Gauss map) to other types, pointing out connections with Lagrange's best approximation,\textit{nearest integer continued fractions}, several types Hurwitz algorithms etc.\N\NThe layout of the paper is as follows:\N\N\S1 Introduction (\(3\frac{1}{2}\) pages).\N\N\S2 The Hurwitz algorithm (\(1\frac{1}{2}\) pages).\N\NHere papers by \textit{A. Hurwitz} [Acta Math. 11, 187--200 (1888; JFM 20.0201.01)] and \textit{J. Hurwitz} [Acta Math. 25, 231--290 (1902; JFM 33.0221.04)] are referred to; it is a generalization of the nearest integer algorithm for numbers in the complex plane.\N\NAs usual the set of \textit{Gaussian integers} is given by\N\[\N\mathbb{Z}[i]=\{a+bi\,:\,a,b\in\mathbb{Z}\},\ i=\sqrt{-1};\ U=\{z\in\mathbb{C}\,:\,-\frac{1}{2}\leq\text{Re}\,z,\,\text{Im}\,z<\frac{1}{2}\}.\N\]\N(\(\text{Re}z\) and \(\text{Im}z\) are the real and imaginary parts of \(z\) respectively)\N\NThe Hurwitz algorithm is defined through the iterates of the Hurwitz map\N\[\NT\,:\, U\setminus \{0\}\rightarrow U;\ z\mapsto \frac{1}{z}-\lfloor \frac{1}{z}\rfloor_U\tag{*}\N\]\Nwhere the floor function \(\lfloor . \rfloor_{1/2}\) is replaced by \(\lfloor\cdot\rfloor_U : \mathbb{C}\rightarrow U\) with \(z\rightarrow z-w\), \(w\) the unique Gaussian integer satrisfying \(z-w\in U\).\N\NFor a given \(n\)-tuple \(b_1,b_2,\ldots,b_n\in\mathbb{Z}[i],\ n\geq 2\), the \textit{cylinder}-sets \(\Delta(b_1,\ldots,b_n)=T^{-1}_{b_1}(\Delta(b_2,\ldots,b_n))\) play an important role (here \(b\in\mathbb{Z}[i],\ \Delta(b)=\{z\in U\, :\, \lfloor\frac{1}{L}\rfloor_U=b\}\) and \(T_b\) is the restriction of \(T\) to \(\Delta(b)\)).\N\NThus \(\Delta(b_1,\ldots,b_n)\) is the set of points \(z\in\mathbb{Z}[i]\) that have the first digits \(a_i=b_i\ (1\leq i\leq n)\) in their Hurwitz expansion.\N\N\S3 A generalization: the \(\alpha\)-Hurwitz algorithm (\(2\frac{1}{2}\) pages).\N\NHere the main object of the paper appears: a perturbation of the Hurwitz map. For \(\mathbf{\alpha}=(\alpha_1,\alpha_2) \in \mathbb{R}^2\), \(T\) in formula \((\ast)\) is made to depend on the \textit{shifted} domain \N\[\NU_{\mathbf{\alpha}}=\{z\in\mathbb{C}\, :\, \alpha_1-1\leq \text{Re} z <\alpha_1,\ \alpha_2-1\leq \text{Im} z <\alpha_2\},\]\Nand \(\lfloor\cdot\rfloor_{\mathbf{\alpha}}\) is the function \N\[\N\lfloor\cdot\rfloor_{\mathbf{\alpha}} : \mathbf{C}\rightarrow U_{\mathbf{\alpha}},\ z\mapsto z-w,\N\]\Nwhere \(w\in\mathbb{Z}[i]\) is the unique Gaussian integer with \(z-w\in U_{\mathbf{\alpha}}\).\N\NThe \(\mathbf{\alpha}\)-Hurwitz map is given by \N\[\NT_{\mathbf{\alpha}} : U_{\mathbf{\alpha}}\setminus \{0\} \rightarrow U_{\mathbf{\alpha}},\ z\mapsto \frac{1}{z}-\lfloor \frac{1}{z}\rfloor_{\mathbf{\alpha}}.\N\]\NThe main result is now\N\NTheorem 3.1. Let \(p,q,r,s\in\mathbb{N}\) be such that \((\frac{p}{q},\frac{r}{s})\in\mathcal{D}\). For \(\mathbf{\alpha}= (\frac{p}{q},\frac{r}{s})\) the \(\mathbf{\alpha}\)-Hurwitz algorithm satisfies the finite range condition. Equivalently, the map \(T_{\mathbf{\alpha}}\) admits a partition with finitely many atoms of \(U_{\mathbf{\alpha}}\).\N\N\S4 Proof of the main result (\(3\frac{1}{2}\) pages).\NProof of Theorem 3.1.\N\N\S5 Perspectives (\(1\) page).\NTouches upon different choices of the \(\alpha_1,\alpha_2\), the optimal constant \(L\) for which the convergents of the \(\mathbf{\alpha}\)-Hurwitz algorithm satisfy \(|z-p_n/q_n|<L|q_n|^2\) (compare [\textit{R. B. Lakein}, Monatsh. Math. 77, 396--403 (1973; Zbl 0307.10033)]), investigation of shifted versions of \textit{J. Shallit}'s algorithm [Springer Proc. Math. Stat. 43, 321--339 (2013; Zbl 1315.11009)].\N\NReferences (\(26\) items).

Cites work

This page was built for publication: A finiteness condition for complex continued fraction algorithms

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6896436)