Transcendence of numbers with a low complexity expansion (Q1377590): Difference between revisions

The (block-) complexity of a sequence on a finite alphabet is the function \(n\to p(n)\), where \(p(n)\) counts the number of (different) blocks of length \(n\) occurring in the sequence. A small step towards the conjecture that any algebraic irrational number is normal in base \(k\) (hence contains all possible blocks in its base \(k\) expansion) would be to prove that, if too many blocks are missing (i.e., \(p(n)\) is small), then the number is either rational or transcendental. Let us call \(C\) the set of sequences on a given alphabet \(\{0,1,\dots, k-1\}\) having low complexity, and such that the corresponding real numbers in base \(k\) are either rational or transcendental. For automatic sequences, we have \(p(n)= O(n)\) and results of Loxton and van der Poorten seem to indicate that these sequences are in \(C\). In the paper under review the authors translate in combinatorial terms Ridout's theorem; they obtain a combinatorial criterion of transcendence that implies: Sturmian sequences (i.e., sequences of complexity \(n+k-1\) on the alphabet \(\{0,1,\dots, k-1\}\)) are in \(C\) (this was known only for subclasses); Arnoux-Rauzy sequences are in \(C\) (these sequences have complexity \(2n+1)\); fixed points with overlaps of primitive morphisms are in \(C\). This last result has been extended in the binary case by Luca Q. Zamboni (misspelled in the paper under review) and the reviewer to any fixed point of a binary morphism primitive or of constant length and non-trivial [\textit{J.-P. Allouche} and \textit{L. Q. Zamboni}, J. Number Theory 69, No. 1, 119--124 (1998; Zbl 0918.11016)].