Tiling the Morse sequence (Q1190472)

Let us first recall that the Prouhet-Thue-Morse sequence is defined as the fixed point of the substitution \(\sigma: 0\to 01\), \(1\to 10\) beginning by 0, hence this sequence is equal to 0110 1001 1001 0110\dots. The author is interested in tiling this sequence (i.e. in covering it by disjoint copies of given finite patterns). The Morse sequence is clearly tiled by the words 0 and 1, but also by the words \(\sigma^ n(0)\) and \(\sigma^ n(1)\) for every integer \(n\): as these two words are arbitrarily long, this implies that the associated dynamical system has rank at most 2 (a precise definition is given in the paper, roughly speaking a sequence is of rank at most \(r\) if it can be covered with \(r\) long words, allowing a small proportion of errors), and it is known that the Morse sequence is not of rank one [\textit{A. del Junco}, Can. J. Math. 29, 655-663 (1977; Zbl 0335.28010)]. The author defines two quantities \(F\) and \(F^*\) which essentially say what precise proportion of the Morse sequence can be tiled by one word, with or without errors. He sketches a proof that \(F=F^*=2/3\). The interest of this result is that it implies the simplicity of the spectrum. In a second part the author replaces words by ``words with holes'' and defines quantities \(T\) and \(T^*\) analogous to \(F\) and \(F^*\) (the rank being replaced by the so-called funny rank). He proves that \(T\geq 5/6\), and conjectures that \(T=5/6\) (the value of \(T^*\) remains an open question). The author ends his paper with interesting problems concerning the funny rank, the Fibonacci and Rudin-Shapiro sequences.