On a two-valued sequence and related continued fractions in power series fields

Bill Allombert, Nicolas Brisebarre, Alain Lasjaunias

Abstract

We explicitly describe a noteworthy transcendental continued fraction in the field of power series over $\mathbb{Q}$, having irrationality measure equal to 3. This continued fraction is a generating function of a particular sequence in the set {1, 2}. The origin of this sequence, whose study was initiated in a recent paper, is to be found in another continued fraction, in the field of power series over $\mathbb{F}_3$, which satisfies a simple algebraic equation of degree 4, introduced thirty years ago by D. Robbins.

Introduction

In the study of Diophantine approximation, the irrationality measure $\nu(\alpha)$ quantifies how well an irrational number $\alpha$ can be approximated by rational numbers. For real algebraic numbers of degree $d \ge 2$, Roth’s Theorem establishes that $\nu(\alpha) = 2$. However, the landscape changes significantly when we transition from real numbers to the field of formal power series. In fields of positive characteristic, such as $\mathbb{F}_q((T^{-1}))$, algebraic elements can exhibit irrationality measures much larger than 2, a phenomenon that has intrigued number theorists for decades.

This paper investigates a specific transcendental power series $\theta$ over the rational numbers $\mathbb{Q}$. By analyzing its continued fraction expansion, we demonstrate that $\theta$ possesses an irrationality measure of exactly 3. This result is particularly significant as it provides a clear, quantitative marker of transcendence in characteristic zero, where algebraic elements are constrained by the analogue of Roth's Theorem.

The Robbins Connection and Sequence Construction

The sequence at the heart of this study is not arbitrary. Its origins lie in a famous quartic equation introduced by David Robbins: $$x^4 + x^2 - Tx + 1 = 0$$ This equation, defined over the finite field $\mathbb{F}_3$, has a solution in the field of power series whose continued fraction expansion is related to a specific sequence of 1s and 2s.

The authors take this sequence, originally identified in the context of finite fields, and treat it as a sequence of coefficients for a power series over $\mathbb{Q}$. Specifically, they define an infinite word $W$ over the alphabet {1, 2} through a recursive block construction:

  • $W_0 = 1, 2$
  • $W_n = W_{n-1}, 2, W_{n-2}, 2, W_{n-1}$ The resulting power series $\theta$ is the generating function of this sequence.

Continued Fraction Expansion in Power Series Fields

In the field of power series $\mathbb{Q}((T^{-1}))$, the "integers" are polynomials in $T$. Any power series can be represented as a continued fraction: $$\theta = [a_0; a_1, a_2, a_3, \dots]$$ where each $a_n$ (the partial quotients) is a polynomial. For most transcendental numbers, the degrees and coefficients of these partial quotients are chaotic. However, for the series $\theta$ derived from the Robbins sequence, the expansion is remarkably structured.

The authors prove that the degrees $d_n = \deg(a_n)$ follow a predictable, 4-step periodic pattern. By establishing recursive relations for these degrees, they show that while most $d_n$ are equal to 1, specific indices $d_{4n+1}$ grow according to the formula: $$d_{4n+1} = \frac{3\ell_n + \ell_{n-1} + 1}{2}$$ where $\ell_n$ is a sequence related to the lengths of the word blocks $W_n$.

Key Results: Irrationality Measure and Transcendence

The primary contribution of this work is the formal calculation of the irrationality measure $\nu(\theta)$. By constructing a sequence of "best" rational approximations $R_n/S_n$ and analyzing the growth of the degrees of $S_n$ relative to the approximation error, the authors prove: $$\nu(\theta) = \lim \sup_{n \to \infty} \frac{\deg(S_{n+1})}{\deg(S_n)} = 3$$ This value of 3 is a "noteworthy" result. In the field of power series over a field of characteristic zero (like $\mathbb{Q}$), an algebraic element must have an irrationality measure of 2. Thus, the fact that $\nu(\theta) = 3$ serves as a rigorous proof that $\theta$ is transcendental.

Conjectures and Future Work

While the degrees of the partial quotients are proven, the exact coefficients of the polynomials $a_n$ remain a subject of conjecture. The authors propose that these coefficients involve a sequence of rational numbers $r_n$ defined by: $$r_n = \frac{4(2\ell_n - \ell_{n-1} + 1)}{25}$$ Numerical evidence strongly supports this conjecture, but a full algebraic proof remains elusive.

Furthermore, the authors suggest that this specific case—using the pair (1, 2)—is a window into a broader class of "automatic" continued fractions. While similar structures might exist for other pairs $(a, b)$, technical requirements like "pseudo-coprimality" may lead to more complex or "perturbed" expansions in other cases. The study highlights the "accidental beauty" of these singular mathematical objects, bridging the gap between combinatorics on words and Diophantine approximation.

Summary updated

Abstract

We explicitly describe a noteworthy transcendental continued fraction in the field of power series over Q, having irrationality measure equal to 3. This continued fraction is a generating function of a particular sequence in the set {1, 2}. The origin of this sequence, whose study was initiated in a recent paper, is to be found in another continued fraction, in the field of power series over $\mathbb{F}\_3$, which satisfies a simple algebraic equation of degree 4, introduced thirty years ago by D. Robbins.
BibTeX
@article{allombert2016two,
  author = {Bill Allombert and Nicolas Brisebarre and Alain Lasjaunias},
  title = {{On a two-valued sequence and related continued fractions in power series fields}},
  journal = {The Ramanujan Journal},
  year = {2016},
}
Figure