Cloitre's Self-Generating Sequence
Jeffrey Shallit
TL;DR
Let $(a_n)_{n\ge1}$ be Cloitre's self-generating sequence over \{1,2\} with the property that the sum of the terms in the $n$-th run equals $2 a_n$. The authors connect $(a_n)$ to the regular paperfolding sequence and express it as the run-length sequence $b_n$ derived from the paperfolding data, ultimately showing $a_n=b_n$. They prove Cloitre's conjecture by constructing a $16$-state synchronized automaton for $g_n$, the number of $1$'s in $a_1\cdots a_n$, and verifying the recurrence relations that imply $0\le 3 g_n-2 n\le 4$, giving $\lim_{n\to\infty} g_n/n=2/3$. The automata-based approach is extended to all paperfolding sequences via a universal automaton, and a related sequence $w_n$ (OEIS $A091960$) is analyzed with automated proofs. This work bridges paperfolding dynamics with Cloitre's sequence and demonstrates how finite automata and automated reasoning can resolve long-standing conjectures and yield generalizations.
Abstract
In 2009 Benoit Cloitre introduced a certain self-generating sequence $$(a_n)_{n\geq 1} = 1, 1, 2, 1, 1, 1, 1, 2, 1, 1, 2, 1, 1, 2, 2, \ldots,$$ with the property that the sum of the terms appearing in the $n$'th run equals twice the $n$'th term of the sequence. We give a connection between this sequence and the paperfolding sequence, and then prove Cloitre's conjecture about the density of $1$'s appearing in $(a_n)_{n \geq 1}$.
