An elementary proof of Bridy's theorem
Eric Rowland, Manon Stipulanti, Reem Yassawi
TL;DR
The paper provides an elementary proof of Bridy's bound on the size of the minimal $q$-automaton for $q$-automatic sequences arising from algebraic power series over $\mathbb{F}_q$. It achieves this by embedding algebraic sequences as diagonals of rational functions and analyzing the orbit of a Cartier-operator-based linear map $\lambda_{0,0}$ on a fixed finite-dimensional border-structured polynomial space, reducing the problem to univariate border emulations and period arguments. The main result bounds the kernel size by $|\ker_q(a(n))| \le q^{h d} + q^{(h-1)(d-1)} \mathcal{L}(h,d,d) + \lfloor\log_q h\rfloor + \lfloor\log_q \max(h,d)\rfloor + 3$, yielding Bridy’s asymptotic $(1+o(1)) q^{h d}$ as parameters grow. The article also provides numerical evidence supporting sharpness, develops a robust framework for diagonals of rational functions, and notes connections to recent multilinear generalizations, while offering conjectural structure results for univariate orbit fixed points that may guide future refinements.
Abstract
Christol's theorem states that a power series with coefficients in a finite field is algebraic if and only if its coefficient sequence is automatic. A natural question is how the size of a polynomial describing such a sequence relates to the size of an automaton describing the same sequence. Bridy used tools from algebraic geometry to bound the size of the minimal automaton for a sequence, given its minimal polynomial. We produce a new proof of Bridy's bound by embedding algebraic sequences as diagonals of rational functions.
