Zeros of Stern polynomials in the complex plane

David Altizio · 2025

Read Zeros of Stern polynomials in the complex plane by David Altizio (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-03847.txt.

Introduction

David Altizio, writing alone, studies where the roots of Stern polynomials lie in the complex plane. These polynomials extend the classical Stern sequence of integers, with a doubling rule that either multiplies by the variable or adds neighbouring terms. An earlier conjecture placed every root in the half-plane where the real part stays below one.

Altizio proves partial progress: a closed disk centred at two with radius one contains no roots at all. The argument uses the Parabola Theorem for convergence of complex continued fractions. As a corollary, a conjecture of Ulas and Ulas follows: the polynomial at a prime index is irreducible over the integers.

Catalog record

CREATORDavid Altizio
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-03847

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Zeros of Stern polynomials in the complex plane David Altizio Abstract The classical Stern sequence of positive integers was extended to a polynomial sequence S n ​ ( λ ) S_{n}(\lambda) by Klavžar et. al. by defining S 0 ​ ( λ ) = 0 S_{0}(\lambda)=0 , S 1 ​ ( λ ) = 1 S_{1}(\lambda)=1 , and S 2 ​ n ​ ( λ ) = λ ​ S n ​ ( λ ) , S 2 ​ n + 1 ​ ( λ ) = S n ​ ( λ ) + S n + 1 ​ ( λ ) . S_{2n}(\lambda)=\lambda S_{n}(\lambda),\quad S_{2n+1}(\lambda)=S_{n}(\lambda)+S_{n+1}(\lambda). Dilcher et. al. conjectured that all roots of S n ​ ( λ ) S_{n}(\lambda) lie in the half-plane { Re w < 1 } \{\operatorname{Re}w<1\} . We make partial progress on this conjecture by proving that { | w − 2 | ≤ 1 } ⊆ ℂ \{|w-2|\leq 1\}\subseteq\mathbb{C} does not contain any roots of S n ​ ( λ ) S_{n}(\lambda) . Our proof uses the Parabola Theorem for convergence of complex continued fractions. As a corollary, we establish a conjecture of Ulas and Ulas by showing that S p ​ ( λ ) S_{p}(\lambda) is irreducible in ℤ ⁡ [ λ ] \mathbb{Z}[\lambda] whenever p p is a positive prime. 1 Introduction The Stern sequence (s_{n})_{n\geq 0} of positive integers, named after Moritz Abraham Stern, is given by s_{0}=0 , s_{1}=1 , and s_{2n}=s_{n},\quad s_{2n+1}=s_{n}+s_{n+1}. (1.1) There is a vast body of literature (e.g. [ 2 , 10 , 12 , 15 ] and the references therein) about the Stern sequence. Perhaps the most striking property of the Stern sequence is that every positive rational number appears exactly once in the sequence (s_{n+1}/s_{n})_{n\geq 1} , giving an explicit enumeration of \mathbb{Q}^{+} . Stern proved this fact 15 years before Cantor introduced the notion of a countable set! In [ 8 ] , Klavžar, Milutinović, and Petr define a polynomial analogue S_{n}(\lambda) of the Stern sequence via S_{0}(\lambda)=0 , S_{1}(\lambda)=1 , and S_{2n}(\lambda)=\lambda S_{n}(\lambda),\quad S_{2n+1}(\lambda)=S_{n}(\lambda)+S_{n+1}(\lambda). (1.2) The first 16 terms of the sequence S_{n}(\lambda) are given in Table 1 . This table suggests many patterns, several of which are recorded in Proposition 1 . \begin{array}[]{c|l||c|l}n&S_{n}(\lambda)&n&S_{n}(\lambda)\\ \hline\cr 1&1&9&1+2\lambda+\lambda^{2}\\ 2&\lambda&10&\lambda+2\lambda^{2}\\ 3&1+\lambda&11&1+3\lambda+\lambda^{2}\\ 4&\lambda^{2}&12&\lambda^{2}+\lambda^{3}\\ 5&1+2\lambda&13&1+2\lambda+2\lambda^{2}\\ 6&\lambda+\lambda^{2}&14&\lambda+\lambda^{2}+\lambda^{3}\\ 7&1+\lambda+\lambda^{2}&15&1+\lambda+\lambda^{2}+\lambda^{3}\\ 8&\lambda^{3}&16&\lambda^{4}\\ \end{array} Table 1: The first 16 terms of the sequence S_{n}(\lambda) . While there are several papers which analyze the sequence S_{n}(\lambda) (see [ 3 , 4 , 5 , 17 ] and the references therein), not much is known about the zeros of Stern polynomials. Because S_{2n}(\lambda)=\lambda S_{n}(\lambda) for all n\geq 1 , it suffices to examine the roots of S_{n}(\lambda) only when n is odd. Define \mathcal{S}\coloneqq\{z\in\mathbb{C}:S_{n}(z)=0\text{ for some odd }n\geq 1\}. In [ 5 , Theorem 2.1] , Gawron proves that 0 , {-}1 , {-}\tfrac{1}{2} , and -\tfrac{1}{3} are the only rational zeros of any Stern polynomial by showing, more generally, that any real number in the closed interval [-\tfrac{1}{4},\tfrac{1}{4}] cannot be in \mathcal{S} . His methods extend easily to show that any complex number z with |z|\leq\tfrac{1}{4} cannot be in \mathcal{S} . For completeness, we record the proof here. Theorem 1 . If z\in\mathcal{S} , then |z|>\tfrac{1}{4} . Proof. Let z be any complex number with |z|\leq\tfrac{1}{4} . Let b_{n}\coloneqq|S_{n}(z)| for each positive integer n . We show more strongly that b_{2n+1}>\tfrac{1}{2}\max\{b_{n},b_{n+1}\}>0 (1.3) for all n\geq 1 . The proof of ( 1.3 ) proceeds by induction on n . The base case, n=1 , follows because b_{3}=|z+1|\geq\tfrac{3}{4}>\tfrac{1}{2}=\tfrac{1}{2}\max\{1,|z|\}=\tfrac{1}{2}\max\{b_{1},b_{2}\}. There are two cases to consider. First assume n=2k is even. Then \displaystyle b_{4k+1} \displaystyle=|S_{4k+1}(z)|=|S_{2k+1}(z)+S_{2k}(z)| \displaystyle=|zS_{k}(z)+S_{2k+1}(z)|\geq b_{2k+1}-\tfrac{1}{4}b_{k} \displaystyle\geq b_{2k+1}-\tfrac{1}{2}b_{2k+1}=\tfrac{1}{2}b_{2k+1}. Moreover, b_{2k+1}\geq\frac{1}{2}b_{k} , and thus \max\{b_{2k},b_{2k+1}\}=b_{2k+1}>0 . In this case, our inequality is proved. Now assume n=2k+1 is odd. Then \displaystyle b_{4k+3} \displaystyle=|S_{4k+3}(z)|=|S_{2k+1}(z)+S_{2k+2}(z)| \displaystyle=|S_{2k+1}(z)+zS_{k+1}(z)|\geq b_{2k+1}-\frac{1}{4}b_{k+1} \displaystyle\geq b_{2k+1}-\tfrac{1}{2}b_{2k+1}=\tfrac{1}{2}b_{2k+1}. Moreover, b_{2k+1}\geq\frac{1}{2}b_{k+1} , and so in this case we also have \max\{b_{2k},b_{2k+1}\}=b_{2k+1}>0 . We have exhausted both cases, completing the proof of Theorem 1 . ∎ In [ 3 ] , Dilcher et. al. focus more specifically on the complex roots of S_{n}(\lambda) . Their paper makes the following conjecture. Conjecture 1 . All elements of \mathcal{S} lie in the half-plane \{\operatorname{Re}w<1\} . By generalizing the Enestrom-Kakeya theorem, they prove Conjecture 1 for several classes of positive integers n taking the form 2^{n}\pm k , where k is fixed and 2^{n}\geq k . These are the only two papers the author could find which discuss the complex zeros of S_{n}(\lambda) . Figure 1 shows a snapshot of \mathcal{S} . One striking feature of this figure is the contrasting behavior of these roots within the half-planes \{\operatorname{Re}w\geq 0\} and \{\operatorname{Re}w<0\} . These differences present difficulties in fully characterizing the geometry of \mathcal{S} . Figure 1: The zeros of S_{n}(\lambda) in the range \{a+bi:-4\leq a\leq 1,|b|\leq 3\} , where 1\leq n<2^{21} is odd. In this paper, we partially resolve Conjecture 1 by establishing the following result. Theorem 2 . All roots of S_{n}(\lambda) lie outside the disk \{\left|z-2\right|\leq 1\}\subseteq\mathbb{C} . To prove Theorem 2 , we use a continued fraction representation for a ratio of Stern polynomials (Theorem 3 ) independently discovered by Reznick [ 12 ] and Schinzel [ 13 ] . This allows us to use the Parabola Theorem (Theorem 4 ) to show that, for certain values of z\in\mathbb{C} , the denominators of these continued fractions can never be zero. Along the way, we establish inequalities in \mathbb{C} relating to the sums 1+z+\cdots+z^{n-1} which may be of independent interest. The most notable of these inequalities is Theorems 7 , which proves a lower bound for this geometric series whenever \operatorname{Re}z\geq 1 . As a corollary, we obtain the following surprising fact, resolving a conjecture of Ulas and Ulas ( [ 17 ] ). Corollary 1 . For each prime number p , the Stern polynomial S_{p}(\lambda) is irreducible in \mathbb{Q} . Ulas and Ulas had verified this conjecture computationally for the first million primes p . Additionally, Schinzel in [ 14 ] proved Corollary 1 for all primes p<2017 by using finite differences to bound the leading coefficient of any proper divisor of S_{n}(\lambda) . However, these previous attempts to prove the conjecture were algebraic in nature, whereas our proof depends on the analytic properties of \mathcal{S} . Acknowledgments. This paper is adapted from the author’s PhD dissertation [ 1 ] . The author thanks his advisor, Dr. Bruce Reznick, for helpful correspondence. 2 Stern Polynomial Preliminaries Notation. Write \mathbb{C} as usual for the field of complex numbers, and let \hat{\mathbb{C}}\coloneqq\mathbb{C}\cup\{\infty\} be the extended complex plane. For all positive integers n , we introduce the (non-standard) notation (\lambda)_{n}\coloneqq 1+\lambda+\cdots+\lambda^{n-1}=\frac{\lambda^{n}-1}{\lambda-1}\in\mathbb{Z}[\lambda]. For continuity purposes, we set (1)_{n}=n . For any set S\subseteq\mathbb{C} , define the auxiliary sets S^{+} and S^{-} via S^{+}\coloneqq S\cap\{\operatorname{Im}w\geq 0\}\quad\text{and}\quad S^{-}\coloneqq S\cap\{\operatorname{Im}w\leq 0\}. Furthermore, for U\subseteq\hat{\mathbb{C}} , set U^{-1}\coloneqq\{w^{-1}:w\in U\}\subseteq\hat{\mathbb{C}} . Finally, for all z\in\mathbb{C} and r\geq 0 , let B_{r}(z) be the open Euclidean ball centered at z with radius r , and let B_{r}[z] denote the closed Euclidean ball centered at z with radius r . We begin by recording some basic properties of the Stern polynomials. All of these can be proven by induction. Proposition 1 . Let n be a nonnegative integer.

Read the complete text