A height gap theorem for coefficients of Mahler functions (Q6172684)

A \(k\)-Mahler series is a power series \(f(z)=\sum_{n=0}^\infty a_n z^n\in \overline{\mathbb Q}[[z]]\) solution of a linear functional equation of the form \[ \sum_{j=0}^d p_j(z)f(z^{k^j})=0 \] for some integers \(k\ge 2\) and \(d\ge 0\), and \(p_j\in \overline{\mathbb Q}[z]\), \(p_0p_d\neq 0\). In this paper, the authors shows that the possible asymptotic growths of the coefficients \(a_n\) fall into five categories. Here, the growth of the sequence \(a_n\) is measured by the logarithmic Weil height \(h(a_n)\). Precisely, their main result is the following striking ``Height Gap Theorem'': Let \(f(z)=\sum_{n=0}^\infty a_n z^n\) be a Mahler series. Then one of the following properties holds (i) \(h(a_n)\in O \cap \Omega(n)\). (ii) \(h(a_n)\in O \cap \Omega(\ln(n)^2)\). (iii) \(h(a_n)\in O \cap \Omega(\ln(n))\). (iv) \(h(a_n)\in O \cap \Omega(\ln\ln(n))\). (v) \(h(a_n)\in O(1)\). (They write \(x_n \in O \cap \Omega(y_n)\) when a sequence \(x_n\) is both \(O(y_n)\) and \(\Omega(y_n)\) with respect to another sequence \(y_n\), and \(O\) and \(\Omega\) are Landau's classical notations.) This excludes certain asymptotic behaviors, for instance it is not possible that \(a_n\sim \ln(n)^3\) and conversely the authors provide explicit examples of Mahler series lying in each one of these five classes. Inside the class of \(k\)-Mahler series lie two classical classes of combinatorial flavor: the class of \(k\)-automatic series, itself inside the class of \(k\)-regular series. Their methods also enable the authors to prove the following characterization: Let \(f(z)=\sum_{n=0}^\infty a_n z^n\) be a \(k\)-Mahler series. Then (i) \(f\) is \(k\)-automatic if and only if \(h(a_n) \in O(1)\), that is if and only if \(a_n\) takes values in a finite set. (ii) \(f\) is \(k\)-regular if and only if \(h(a_n)\in O(\ln(n))\). The first case extends to \(k\)-automatic series with coefficients in a field of characteristic 0. As a by-product, they recover a conjecture of Becker on \(k\)-regular functions which was recently settled by Bell, Chyzak, Coons, and Dumas [\textit{J. P. Bell} et al., Trans. Am. Math. Soc. 372, No. 5, 3405--3423 (2019; Zbl 1475.30006)].