Beta-expansion and continued fraction expansion (Q2465888)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Beta-expansion and continued fraction expansion
scientific article

    Statements

    Beta-expansion and continued fraction expansion (English)
    0 references
    0 references
    0 references
    10 January 2008
    0 references
    For a real number \(\beta > 1\), every real number \(x \in [0,1]\) has a \(\beta\)-expansion \(x = \sum_{n=1}^\infty \varepsilon_n(x) \beta^{-n}\). Such \(\beta\)-expansions in general are non-unique, but letting \(T_\beta x = \{\beta x\}\), where \(\{y\}\) denotes the fractional part of \(y\), we obtain such an expansion where \(\varepsilon_n(x) = [\beta T_\beta^{n-1} x]\), where \([y]\) denotes the integer part of \(y\). We classify real numbers \(\beta > 1\) according to the growth rate of the length of blocks of zeros in the \(\beta\)-expansion of \(1\). Defining \[ l_n = \sup\{k \geq 0 : \varepsilon_{n+j}(1) = 0\text{ for all } 1 \leq j \leq k\}, \] we say that \(\beta \in A_0\) if \(l_n\) is bounded, \(\beta \in A_1\) if \(\limsup_{n \rightarrow \infty} l_n/n = 0\) and \(\beta \in A_2\) otherwise. The set \(A_0\) contains all Pisot numbers, i.e., real algebraic integers greater than \(1\) whose conjugates are all of modulus \(<1\). The paper under review relates the \(\beta\)-expansion to the simple continued fraction expansion of a number \(x \in [0,1)\). Let \(I(a_1, \dots, a_m)\) denote the set of all numbers in \([0,1)\) whose first \(m\) partial quotients are \(a_1, \dots, a_m\). Similarly, let \(J(\varepsilon_1, \dots, \varepsilon_n)\) denote the set of numbers \(x \in [0,1)\) whose \(\beta\)-expansion begins with the digits \(\varepsilon_1, \dots, \varepsilon_n\). The paper is concerned with the quantity \(k_n(x) = \sup \{m \geq 0 : J(\varepsilon_1(x), \dots,\varepsilon_n(x)) \subset I(a_1(x), \dots, a_m(x))\}\), i.e., the number of partial quotients determined by the first \(n\) digits in the \(\beta\)-expansion. It is shown that if either (i) \(\beta \in A_0\) and \(x\) is irrational or (ii) \(\beta \in A_1\) and \(x \notin \{x \in [0,1) : \beta^*(x) = + \infty, \beta_*(x) < +\infty\}\), then \[ \liminf_{n \rightarrow \infty} {k_n(x) \over n} = {\log \beta \over {2 \beta^*(x)}}, \quad \limsup_{n \rightarrow \infty} {k_n(x) \over n} = {\log \beta \over {2 \beta_*(x)}}, \] where \(\beta_*(x) = \liminf_{n \rightarrow \infty} \log q_n(x)/n\) and \(\beta^*(x) = \limsup_{n \rightarrow \infty} \log q_n(x)/n\) are the lower and upper Lévy constants. Here, \(q_n(x)\) denotes the denominator of the \(n\)th convergent of \(x\). This extends previous work by \textit{J. Wu} [Adv. Math. 206, No. 2, 684--694 (2006; Zbl 1156.11033)], who showed that the same conclusion holds for \(\beta = 10\) and any irrational \(x \in [0,1)\). Note that the important case of \(\beta\) being a Pisot number is contained in the result.
    0 references
    \(\beta\)-expansion
    0 references
    continued fraction expansion
    0 references
    Lévy constant.
    0 references

    Identifiers