On the expansions of a real number to several integer bases (Q1759752): Difference between revisions

Let \(\xi\) be a real number and \(b\geq 2\) be an integer. Then we can write \[ \xi=\left\lfloor\xi\right\rfloor+\sum_{k\geq1}\frac{a_k}{b^k} =\left\lfloor\xi\right\rfloor+0.a_1a_2a_3\dots, \] where \(\left\lfloor\xi\right\rfloor\) denotes the greatest integer less than \(\xi\), each digit \(a_k\) is an integer from \(\{0,1,\dots,b-1\}\) and infinitely many \(a_k\) are not equal to \(b-1\). The present paper concerns the complexity of the \(b\)-ary expansion. For a positive number \(n\), let \(p(n,\xi,b)\) denote the total number of distinct blocks of \(n\) digits in the \(b\)-ary expansions of \(\xi\). We call a number rich if for every \(n\) every possible block of \(n\) digits occurs in the expansion, \textit{i.e.} \(p(n,\xi,b)=b^n\). Furthermore let \(E(\xi,b)\) denote the entropy of \(\xi\) to base \(b\), \textit{i.e.} \[ E(\xi,b)=\lim_{n\to+\infty}\frac{\log p(n,\xi,b)}{n}. \] Then, the first result states that given \(\varepsilon>0\) and \(b_0\geq 2\) an integer, there exists a positive real number \(c\) depending only on \(\varepsilon\) and \(b_0\), and uncountably many real numbers \(\xi\) such that \[ E(\xi,b_0)<\varepsilon \] and \[ E(\xi,b)\leq\log b-\frac{1}{b^{cb(\log b)}},\quad\text{for every }b\geq2. \] The second result is a construction of the \(b\)-ary expansion (and therefore of a number \(\xi\)) which is rich to every integer base. Finally in the third result the author considers digit changes and the number of zeros in the \(b\)-ary expansion. In particular, he is able to show that a real number cannot have very few digit changes simultaneously with respect to two coprime bases. A similar result holds for non-zero digits.