On the orthogonality of generalized pattern sequences (Q6934102)

Let \(b\) be an integer and \(w\) be a word over a \(b\)-letter alphabet. For every positive integer \(n\) the sequence \(e_{b,w}(n)\) counts the number of occurrences of \(w\) in the \(w\)-representation of \(n\) (possibly with leading zeros; the number is well-defined if \(w\) contains non-zero digits). Now the sequence \((a_{b,m,w}(n))\) is defined as \(a_{b,m,w}(n)=exp\left(\frac{2\pi i e_{b,w}}{m}\right)\), and is called a pattern sequence. For example, \(a_{2,2,1}\) is the Thue-Morse word and \(a_{2,2,11}\) is the Rudin-Shapiro word, both over the alphabet \(\{-1,1\}\).\N\NNow a generalized pattern sequence is defined as a letter-to-letter product of several pattern sequences with the same \(b\) and \(m\) but with different \(w\); in particular, the same pattern sequence can appear in this product several times. A generalized pattern sequence is denoted by \((a_{b,m,S}(n))\), where \(S\) is a finite set with multiplicities of words \(w\).\N\NThe main result stated (but not proven) in this very dense abstract is a necessary and sufficient condition for the equality\N\[\N\lim_{N\to\infty} \frac{1}{N}\sum_{n=0}^{N}a_{b,m,S}(n)=0.\N\]\NThe condition is stated in terms of an associated matrix \(M_{b,m,S}\) and a column vector \(A_{b,m,S}(0)\). Several examples are given.\N\NFor the entire collection see [Zbl 1569.05001].