Automaticity of the sequence of the last nonzero digits of \(n!\) in a fixed base (Q2199500)

Let \(b \ge 2\) be an integer. A sequence \((a_n)_{n \ge 0}\) taking finitely many values is said to be \(b\)-\textit{automatic} if there is a finite machine which permits to know the value of \(a_n\) by reading one after the other the digits of \(n\) in its expansion in base \(b\). Moreover, the sequence \((a_n)_n\) is said to be \textit{automatic} if it is \(k\)-automatic for some \(k\). This classical notion is recalled in Section 2 as well as its main properties -- a useful point. The author addresses the question of the automaticity of the sequence \((\ell_b(n!))_n\), where \(\ell_b(n!)\) - an integer in \([1, b-1]\) - denotes the \textit{last nonzero digit} of \(n!\) in base \(b\), defined in the following way: if one writes \(n! = b^r m\), where \(\gcd(m, b) \neq 0\), then \(\ell_b(n!) \) is congruent to \(\gcd(m,b)\) modulo \(b\). It is folklore that \((\ell_b(n!))_n\) is automatic if \(b\) is a power of a prime, or a small number like \(6\) or \(10\). It was proved in the paper quoted as [4] that the sequence \((\ell_{12}(n!))_n\) is \textit{not} automatic. The paper completely characterizes the values of \(b\) for which the sequence \((\ell_b(n!))_n\) is automatic. More precisely, Theorem 3, proved in Section 3, states the following Let \(b = p_1^{a_1} p_2^{a_2}\cdots\) with \(a_1(p_1-1) \ge a_2(p_2-1)\ge \cdots\). The sequence \((\ell_b(n!))_n\) is \(p_1\)-automatic if \(a_1(p_1-1) > a_2(p_2-1)\) or \(b = p_1^{{a_1}}\) and is not automatic otherwise. Notice that \(12=2^2 \times 3^1\) with \(2\times (2-1) = 1 \times (3-1)\).