On normal numbers (Q772256): Difference between revisions

Let \(r,s\) be integers and suppose that \(\log r/\log s\) is irrational. Let \(t\) be an integer, where \(1<t<s\). The author shows that almost all numbers in whose ``decimal'' expansion to base \(s\) only digits \(<t\) occur are normal to base \(r\), where almost all is taken with respect to an appropriate measure. More precisely, he shows that almost always the discrepancy of the sequence \(r^n\xi\) \((1\leq n\leq N)\) modulo 1 is less than \(N^{-\alpha}\), where only the digits \(<t\) occur in the expansion of \(\xi\) to base \(s\) and \(\alpha\) depending only on \(r,s,t\). A similar but less precise result was proved simultaneously and independently by the reviewer [Colloq. Math. 7, 95--101 (1959; Zbl 0090.26004)].