Avoidance of split overlaps (Q2214038)

A \(t\)-overlap is a finite word of the form \(xxx'\), where \(|x'|=t\) and the word \(x\) starts with \(x'\). Here \(1\)-overlaps are just usual overlaps studied since Thue's papers from the beginning of the 20th century. A split occurrence of a repetition is a word of the form \(xyz\), where \(xz\) forms the repetition; so, for example, the word \texttt{contentment} contains the factor \texttt{ntentment} which is a split occurrence of the \(2\)-overlap \texttt{ntentent}. As Ochem remarked (the proof of this statement is given in the paper for the sake of completeness), there are no infinite words over a finite alphabet which would avoid split overlaps. So, the main results of this paper are upper and lower bounds for the length of the longest word over a \(k\)-letter alphabet which avoids split \(t\)-overlaps.