The digit exchanges in the rotational beta expansions of algebraic numbers (Q2167506)

From MaRDI portal
scientific article
Language Label Description Also known as
English
The digit exchanges in the rotational beta expansions of algebraic numbers
scientific article

    Statements

    The digit exchanges in the rotational beta expansions of algebraic numbers (English)
    0 references
    0 references
    0 references
    25 August 2022
    0 references
    \textit{A. Rényi} [Acta Math. Acad. Sci. Hung. 8, 477--493 (1957; Zbl 0079.08901)] introduced the \(\beta\)-expansion of a real number \(\xi\) with respect to a real \(\beta >1\), which is defined as follows. Define \(T_{\beta}:\, [0,1]\to [0,1)\) by \(T_{\beta}(x)=\beta x-[\beta x]\). If \(0\leq \xi\leq 1\), define the digits of \(\xi\) in its \(\beta\)-expansion by \(t_n(\beta ,\xi ):=[\beta T_{\beta}^{n-1}(\xi )]\) for \(n=0,1,2,\ldots\). Then \(\xi=\sum_{n=0}^{\infty} t_n(\beta ,\xi )\beta^{-n}\). If \(\xi>1\), one chooses an integer \(R\) such that \(\beta^{-R}\xi\in [0,1)\). Then the \(\beta\)-expansion of \(\xi\) is \(\xi =\sum_{n=1-R}^{\infty} t_n(\beta ,\xi )\beta^{-n}=\beta^{R}\sum_{n=1}^{\infty} t_n(\beta,\beta^{-R}\xi )\beta^{-n}\). \textit{Y. Bugeaud} [Integers 9, No. 3, 215--226, A20 (2009; Zbl 1194.11022)] considered the number of digit changes of the \(\beta\)-expansion of \(\xi\) in case that \(\xi\) is algebraic, and \(\beta\) is a Pisot number or Salem number. Recall that \(\beta\) is a Pisot number if \(\beta >1\), \(\beta\) is an algebraic integer and its conjugates over \(\mathbb{Q}\) other than \(\beta\) itself have absolute value smaller than \(1\). If instead, all conjugates of \(\beta\) other than \(\beta\) itself have absolute value \(\leq 1\) and at least one of them has absolute value exactly \(1\), \(\beta\) is called a Salem number. Let \(\gamma (\beta ,\xi ,N)\) denote the number of \(n\leq N\) with \(t_{n+1}(\beta ,\xi )\not= t_n(\beta ,\xi )\) for \(n\leq N\). Using a deep tool from Diophantine approximation, namely Ridout's p-adic generalization of Roth's theorem, Bugeaud showed that if \(\xi\in (0,1)\) is algebraic and \(\beta\) is Pisot or Salem, then \(\gamma (\beta,\xi, N)\gg_{\beta ,\xi} (\log N)^{3/2}/(\log\log N)^{1/2}\) as \(N\to\infty\). \textit{H. Kaneko} [Rend. Semin. Mat. Univ. Padova 136, 205--223 (2016; Zbl 1362.11073)] considered the problem of determining a lower bound for \(\nu (\beta ,\xi ,N)\), which is the number of \(n\leq N\) with \(t_n(\beta ,\xi )\not= 0\). By means of an elementary approach, very different from Bugeaud's, Kaneko showed that if again \(\beta\) is a Pisot or Salem number and \(\xi\in (0,1)\) is algebraic, with \([\mathbb{Q} (\beta ,\xi ) :\mathbb{Q} (\beta )]=D\), then \(\nu (\beta ,\xi ,N) \gg_{\beta ,\xi } (N/\log N)^{1/D}\) as \(N\to\infty\). In the paper under review, the authors prove a general theorem, which implies an improvement upon Bugeaud's lower bound for \(\gamma (\beta ,\xi ,N)\), but for a restricted class of algebraic numbers \(\xi\). On the other hand, they extend the class of \(\beta\) under consideration to so-called quasi-Pisot and quasi-Salem numbers and consider also expansions of complex algebraic numbers. If \(\beta\) is a not necessarily real algebraic integer with \(|\beta |>1\), we call \(\beta\) quasi-Pisot if each conjugate of \(\beta\) outside \(\{ \beta ,\overline{\beta}\}\) has absolute value \(<1\), and quasi-Salem, if each conjugate of \(\beta\) not in \(\{ \beta ,\overline{\beta}\}\) has absolute value \(\leq 1\), with at least one of these conjugates having absolute value exactly \(1\). We state their result. \textbf{Theorem.} Let \(\beta\) be a quasi-Pisot or quasi-Salem number and \(\xi\) an algebraic number with \(D=[\mathbb{Q} (\beta ,\xi ):\mathbb{Q} (\beta )]=D\). Let \(S\) be a finite subset of \(\mathbb{Z} [\beta ]\) with \(0\in S\). Assume that either \(S\subset\mathbb{Z}\), or \(\mathbb{Q} (\beta )\) has an imaginary quadratic subfield \(K\) and \(S\) is contained in the ring of integers of \(K\). Assume that \(\xi\) satisfies the following conditions: (i) \(\xi=\sum_{n=1-R}^{\infty} t_n\beta^{-n}\) with \(R\geq 0\) and \(t_n\in S\) for all \(n\geq 1-R\); (ii) there exist \(\pi ,A_0,A_1,\ldots ,A_D\in \mathbb{Z}[\beta ]\) with \(\pi\not= 0\), \(A_D\not= 0\) such that \(A_D\xi^D+A_{D-1}\xi^{D-1}+\cdots +A_0=0\); \((\beta -1)^{D-k}A_k/\pi \in \mathbb{Z} [\beta ]\) for \(k=1,\ldots ,D\); \\ \((\beta -1)^D\beta^nA_0/\pi\not\in \mathbb{Z}[\beta ]\) for all \(n\geq 0\). Then \( \#\{ n\in\mathbb{Z} :\, 1-R\leq n\leq N,\, t_n\not= t_{n+1}\}\gg_{\beta ,\xi} (N/\log N)^{1/D}. \) The proof is elementary. The authors apply their result to obtain an improvement of Bugeaud's result for certain algebraic numbers \(\xi\) in the case that \(\beta\) is Pisot or Salem. Further they derive lower bounds for the number of digit changes up to index \(N\) for so-called negative or rotational expansions for certain algebraic numbers \(\xi\).
    0 references
    \(\beta\)-expansions
    0 references
    nonzero digits
    0 references
    Pisot numbers
    0 references
    Salem numbers
    0 references

    Identifiers