GALOIS GROUPS OF APÉRY-LIKE SERIES MODULO PRIMES

XAVIER CARUSO; FLORIAN FÜRNSINN; DANIEL VARGAS-MONTOYA; WADIM ZUDILIN

doi:10.1017/S0004972725100932

GALOIS GROUPS OF APÉRY-LIKE SERIES MODULO PRIMES

Part of: Algebraic number theory: local and $p$-adic fields Differential equations in the complex domain Combinatorics Sequences and sets

Published online by Cambridge University Press: 13 February 2026

XAVIER CARUSO

FLORIAN FÜRNSINN

DANIEL VARGAS-MONTOYA

and

WADIM ZUDILIN

Show author details

XAVIER CARUSO: Affiliation:
IMB, CNRS; Université de Bordeaux , Talence, France e-mail: xavier@caruso.ovh
FLORIAN FÜRNSINN*: Affiliation:
Faculty of Mathematics, University of Vienna , Vienna, Austria
DANIEL VARGAS-MONTOYA: Affiliation:
IMT, Université Paul Sabatier , Toulouse, France e-mail: daniel.vargas-montoya@math.univ-toulouse.fr
WADIM ZUDILIN: Affiliation:
IMAPP, Radboud University Nijmegen , Nijmegen, Netherlands e-mail: w.zudilin@math.ru.nl
*: e-mail: florian.fuernsinn@univie.ac.at

Article contents

Abstract
Introduction
The numbers
Adaptation to Apéry-like sequences
Footnotes
References

Rights & Permissions

Abstract

We compute the Galois groups of the reductions modulo a prime number p of the generating series of Apéry numbers, Domb numbers and Almkvist–Zudilin numbers. We observe in particular that their behaviour is governed by congruence conditions on p.

Keywords

Apéry numbers D-finite series Galois groups

MSC classification

Primary: 11S20: Galois theory

Secondary: 05A10: Factorials, binomial coefficients, combinatorial functions 11B50: Sequences (mod $m$) 33C05: Classical hypergeometric functions, $_2F_1$ 34M15: Algebraic aspects (differential-algebraic, hypertranscendence, group-theoretical)

Information

Type: Research Article
Information: Bulletin of the Australian Mathematical Society , Volume 114 , Issue 1 , August 2026 , pp. 65 - 78

DOI: https://doi.org/10.1017/S0004972725100932 [Opens in a new window]
Creative Commons: This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright: © The Author(s), 2026. Published by Cambridge University Press on behalf of Australian Mathematical Publishing Association Inc.

1 Introduction

The Apéry numbers

$$ \begin{align*} \alpha_n = \sum_{k=0}^n{\binom nk}^2{\binom{n+k}n}^2 \quad\text{for}\; n=0,1,2,\dots, \end{align*} $$ alpha Subscript n Baseline equals sigma summation Underscript k equals 0 Overscript n Endscripts StartBinomialOrMatrix n Choose k EndBinomialOrMatrix squared StartBinomialOrMatrix n plus k Choose n EndBinomialOrMatrix squared for n equals 0 comma 1 comma 2 comma ellipsis comma

are a famous sequence of integer numbers, mostly known for playing a prominent role in Apéry’s proof [Reference Apéry3] of the irrationality of $\zeta (3)$ . Their generating series enjoys many interesting properties. It is D-finite, that is, it satisfies a linear differential equation with polynomial coefficients. Starting from the differential equation, integrality of the coefficients of its solutions is highly remarkable. The Apéry numbers grow in a controlled manner, making their generating series a G-function. Moreover, letting p be an odd prime number, it is known [Reference Gessel16, Theorem 1] that the Apéry numbers have the p-Lucas property for all prime numbers p (see also the general result for Apéry-like numbers in [Reference Malik and Straub20]). That is, $\alpha _{np+\ell }\equiv \alpha _n \alpha _\ell \pmod p$ whenever $0 \leq \ell < p$ . On the level of generating functions, this property translates to the congruence $f_\alpha \equiv A_p\cdot f_\alpha ^p \pmod p$ , where denotes the truncation of $f_\alpha $ at order p. This relation also shows that the reduction of the generating function $f_\alpha \bmod p$ is algebraic over the field of rational functions $\mathbb {F}_p(t)$ and that the extension it generates is Galois with cyclic Galois group of order dividing $p-1$ .

Algebraicity modulo (almost) all primes p is a phenomenon observed for many classes of D-finite series. More precisely, for diagonals of multivariate rational or, equivalently, algebraic functions, this is a consequence of a theorem of Furstenberg [Reference Furstenberg15]. For hypergeometric functions, it can be deduced from work by Christol [Reference Christol9] and was made explicit by Vargas-Montoya [Reference Vargas-Montoya22]. Further, if Christol’s conjecture [Reference Christol10] proves to be true, the result of Furstenberg would apply to all globally bounded D-finite series.

Having algebraic equations for the generating function modulo prime numbers, it is natural to consider their Galois groups for each prime number p. For D-finite series whose reductions modulo infinitely many prime numbers p are algebraic, it was (conjecturally) observed in [Reference Caruso, Fürnsinn and Vargas-Montoya6] that these Galois groups show some uniformity properties across different primes and seem to be related to the differential Galois group of the minimal differential equation satisfied by the series. As an example in [Reference Caruso, Fürnsinn and Vargas-Montoya6, Section 2.1.5], the Galois groups of the reductions of the Apéry series were computed for many prime numbers using a computer algebra system. The pattern that unfolded was stated without a proof and the aim of this paper is to provide one.

Fixing a prime number p and starting from the algebraic relation given by the p-Lucas property, the Galois group of $f_\alpha \bmod p$ can canonically be seen as a subgroup of $\mathbb {F}_p^\times $ via the embedding

$$ \begin{align*}\begin{array}{rcl} \operatorname{\mathrm{Gal}}(\mathbb{F}_p(t,f_\alpha)/\mathbb{F}_p(t)) & \longrightarrow & \mathbb{F}_p^\times, \\ \sigma & \mapsto & \sigma(f_\alpha)/f_\alpha. \end{array}\end{align*} $$

We aim to determine the image of this map. Throughout the article, we denote by S the unique subgroup of $\mathbb {F}_p^\times $ of index $2$ ; it is the subgroup of squares, hence the notation.

Theorem 1.1.

• If $p \equiv 1, 5, 7, 11 \pmod {24}$ , then $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(t, f_\alpha )/\mathbb {F}_p(t)) = S$ .
• If $p \equiv 13, 17, 19, 23 \pmod {24}$ , then $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(t, f_\alpha )/\mathbb {F}_p(t)) = \mathbb {F}_p^\times $ .

This result is the Galois theoretic shadow of the following factorisation property of $A_p$ .

Theorem 1.2. There exists a polynomial $B_p \in \mathbb {F}_p[t]$ such that:

• if $p \equiv 1, 5, 7, 11 \pmod {24}$ , then $A_p = B_p^2$ ;
• if $p \equiv 13, 17, 19, 23 \pmod {24}$ , then $A_p = (t^2 - 34t + 1) \cdot B_p^2$ .

The main ingredient in our proof is the fact that the Apéry series $f_\alpha $ is related to the generating function of the Franel numbers

by a simple rational substitution, namely $f_\alpha = (1+x)\cdot h^2$ , where the variables t and x are linked by the relation $t = {x(1-8x)}/{(1+x)}$ .

The sequence of Apéry numbers has two companions: the (alternating version of the) Domb numbers

$$ \begin{align*} \delta_n = (-1)^n\sum_{k=0}^n \binom{2k}k \binom{2n-2k}{n-k} {\binom nk}^2 \quad\text{for}\; n=0,1,2,\dots \end{align*} $$ delta Subscript n Baseline equals left parenthesis negative 1 right parenthesis Superscript n Baseline sigma summation Underscript k equals 0 Overscript n Endscripts StartBinomialOrMatrix 2 k Choose k EndBinomialOrMatrix StartBinomialOrMatrix 2 n minus 2 k Choose n minus k EndBinomialOrMatrix StartBinomialOrMatrix n Choose k EndBinomialOrMatrix squared for n equals 0 comma 1 comma 2 comma ellipsis

and the Almkvist–Zudilin numbers

$$ \begin{align*} \xi_n=\sum_{k=0}^n(-1)^{n-k}3^{n-3k}\frac{(3k)!}{k!^3}\binom n{3k}\binom{n+k}n \quad\text{for}\; n=0,1,2,\dots, \end{align*} $$

in the following sense. The three sequences satisfy similar difference equations of order 2 and degree 3, and their generating series

$$ \begin{align*} f_\alpha = \sum_{n=0}^\infty\alpha_n t^n, \quad f_\delta = \sum_{n=0}^\infty\delta_n t^n \quad\text{and}\quad f_\xi = \sum_{n=0}^\infty\xi_n t^n \end{align*} $$

admit modular parametrisations via the Hauptmoduln of the three subgroups of index 2 lying between $\Gamma _0(6)$ and its normaliser in $\operatorname {SL}_2(\mathbb R)$ (see [Reference Chan and Verrill7, Reference Chan and Zudilin8]).

Thanks to [Reference Chan and Zudilin8, Theorem 2.2], the argument for $f_\alpha $ extends to the series $f_\delta $ and $f_\xi $ with the new relations:

$$ \begin{align*} f_\delta = (1-8x) \cdot h^2 & \quad \text{with } t = \frac{x(1+x)}{1-8x}, \\ f_\xi = (1+x)(1-8x) \cdot h^2 & \quad \text{with } t = \frac{x}{(1+x)(1-8x)}. \end{align*} $$

StartLayout 1st Row 1st Column f Subscript delta Baseline equals left parenthesis 1 minus 8 x right parenthesis dot h squared 2nd Column with t equals StartFraction x left parenthesis 1 plus x right parenthesis Over 1 minus 8 x EndFraction comma 2nd Row 1st Column f Subscript xi Baseline equals left parenthesis 1 plus x right parenthesis left parenthesis 1 minus 8 x right parenthesis dot h squared 2nd Column with t equals StartFraction x Over left parenthesis 1 plus x right parenthesis left parenthesis 1 minus 8 x right parenthesis EndFraction period EndLayout

Theorem 1.3. Set

$$ \begin{align*} A_{\delta,p}=\sum_{n=0}^{p-1}\delta_nt^n\in\mathbb F_p[t]. \end{align*} $$ upper A Subscript delta comma p Baseline equals sigma summation Underscript n equals 0 Overscript p minus 1 Endscripts delta Subscript n Baseline t Superscript n Baseline element of double struck upper F Subscript p Baseline left bracket t right bracket period

upper A Subscript delta comma p Baseline equals sigma summation Underscript n equals 0 Overscript p minus 1 Endscripts delta Subscript n Baseline t Superscript n Baseline element of double struck upper F Subscript p Baseline left bracket t right bracket period

Then, there exists a polynomial $B_{\delta ,p} \in \mathbb {F}_p[t]$ such that:

• if $p \equiv 1 \pmod {6}$ , then $A_{\delta ,p} = B_{\delta ,p}^2$ and $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(t, f_\delta )/\mathbb {F}_p(t)) = S$ ;
• if $p \equiv 5 \pmod {6}$ , then $A_{\delta ,p} = (64t^2-20t+1) B_{\delta ,p}^2$ and $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(t, f_\delta )/\mathbb {F}_p(t)) = \mathbb {F}_p^\times $ .

Theorem 1.4. Set

$$ \begin{align*} A_{\xi,p}=\sum_{n=0}^{p-1}\xi_nt^n\in\mathbb F_p[t]. \end{align*} $$ upper A Subscript xi comma p Baseline equals sigma summation Underscript n equals 0 Overscript p minus 1 Endscripts xi Subscript n Baseline t Superscript n Baseline element of double struck upper F Subscript p Baseline left bracket t right bracket period

upper A Subscript xi comma p Baseline equals sigma summation Underscript n equals 0 Overscript p minus 1 Endscripts xi Subscript n Baseline t Superscript n Baseline element of double struck upper F Subscript p Baseline left bracket t right bracket period

Then, there exists a polynomial $B_{\xi ,p} \in \mathbb {F}_p[t]$ such that:

• if $p \equiv 1,3 \pmod {8}$ , then $A_{\xi ,p} = B_{\xi ,p}^2$ and $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(t, f_\xi )/\mathbb {F}_p(t)) = S$ ;
• if $p \equiv 5,7 \pmod {8}$ , then $A_{\xi ,p} = (81t^2+14t+1) B_{\xi ,p}^2$ and $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(t, f_\xi )/\mathbb {F}_p(t)) = \mathbb {F}_p^\times $ .

In Section 3.3, we collect more examples of series where we computationally observed similar patterns.

2 The Apéry numbers

Throughout this section, we fix an odd prime number p and set .

Lemma 2.1. The extension $\mathbb {F}_p(x)/\mathbb {F}_p(t)$ is a quadratic extension, which is generated by $(t^2-34t+1)^{1/2}$ , with nontrivial element $\sigma :x\mapsto {(1-8x)}/{(8+8x)}$ of the Galois group.

Proof. We note that x is a solution of the quadratic equation ${8x^2+(t-1)x+t=0}$ over $\mathbb {F}_p(t)$ with discriminant $\Delta =t^2-34t+1$ . An easy calculation shows that ${(1-8x)}/{(8+8x)}$ also solves this equation.

Let

h defined as summation from n=0 to infinity of summation from k=0 to n of (n over k)^3 x^n in F sub p [[x]] and H defined as summation from n=0 to p-1 of summation from k=0 to n of (n over k)^3 x^n in F sub p [x].

By Lucas’ theorem (on evaluating binomial coefficients modulo p), we easily check that the coefficients of h satisfy Lucas’ congruences and we deduce that $h=H h^p$ . Thus, $h^2 = H^{-1/e}$ with .

Lemma 2.2. The polynomial H is not an nth power in $\mathbb {F}_p[x]$ for $n\geq 2$ .

Proof. The series h satisfies the second-order differential equation

(2.1) $$ \begin{align} x(x+1)(8x-1)\frac{d^2h}{dx^2} + (24x^2+14x-1) \frac{dh}{dx} + (8x+2)h = 0. \end{align} $$ x left parenthesis x plus 1 right parenthesis left parenthesis 8 x minus 1 right parenthesis StartFraction d squared h Over d x squared EndFraction plus left parenthesis 24 x squared plus 14 x minus 1 right parenthesis StartFraction d h Over d x EndFraction plus left parenthesis 8 x plus 2 right parenthesis h equals 0 period

By [Reference Caruso, Fürnsinn and Vargas-Montoya6, Lemma 2.1.3], H is not an eth power of a polynomial for any e, as soon as H has a root different from $0, -1, 1/8$ . One checks that H has degree $p-1$ and, using $\binom {p-1}{n}\equiv (-1)^n \pmod p$ , that it has leading coefficient $1$ . Also, it is clear that its constant coefficient is $1$ , so it does not have $0$ as a root. Assuming that

$$ \begin{align*} \textstyle H=\big(x-\tfrac{1}{8}\big)^m (x+1)^{p-1-m} \end{align*} $$

for some m, we deduce by comparing the constant coefficients that $(-\tfrac 18)^m\equiv 1 \pmod p$ . Comparing the coefficient of x yields

$$ \begin{align*} m\big(-\tfrac{1}{8}\big) ^{m-1}+ (p-1-m) \equiv 2 \pmod p, \end{align*} $$

so we deduce $m=(p-1)/3$ if $p\equiv 1 \pmod 3$ and $m=(2p-1)/3$ otherwise. If $p\equiv 1 \pmod 3$ , we can then compare coefficients of $x^2$ and find the contradiction $13\equiv 10\pmod p$ . If $p\equiv 2 \pmod 3$ , the contradiction already arises at the constant coefficient. Thus, H has a root different from $-1$ and $1/8$ , and the claim follows.

The field $\mathbb {F}_p(x)$ contains all eth roots of unity, where we keep the notation ${e=(p-1)/2}$ , as they are already in $\mathbb {F}_p$ . Kummer theory then implies that the extension $\mathbb {F}_p(x, h^2)/\mathbb {F}_p(x)$ is an abelian extension of degree e whose Galois group is canonically isomorphic to S.

We now consider the tower of extensions:

Field extension diagram. Top: F sub p (x, h-squared) = F sub p (x, e-th root of H). Middle: F sub p (x) connected by Gal = S. Bottom: F sub p (t) connected by Gal = angle bracket sigma close angle bracket.

and aim at studying the Galois properties of the total extension $\mathbb {F}_p(x,h^2)/\mathbb {F}_p(t)$ . To do this, we first determine the action of $\sigma $ on H.

Lemma 2.3. We have

$$ \begin{align*}H=\begin{cases} \sigma(H)\cdot (x+1)^{p-1} & \text{if } p\equiv 1 \pmod 6,\\ - \sigma(H)\cdot (x+1)^{p-1} & \text{if } p\equiv 5 \pmod 6. \end{cases}\end{align*} $$

StartLayout 1st Row upper H equals StartLayout Enlarged left brace 1st Row 1st Column sigma left parenthesis upper H right parenthesis dot left parenthesis x plus 1 right parenthesis Superscript p minus 1 Baseline 2nd Column if p identical to 1 left parenthesis mod 6 right parenthesis comma 2nd Row 1st Column minus sigma left parenthesis upper H right parenthesis dot left parenthesis x plus 1 right parenthesis Superscript p minus 1 Baseline 2nd Column if p identical to 5 left parenthesis mod 6 right parenthesis period EndLayout EndLayout

Proof. We first notice that $\sigma $ is an involution and that $G=\sigma (H) \cdot (x+1)^{p-1}$ is a polynomial of degree $p-1$ . Additionally, one checks that H is a solution of the differential equation (2.1). Expressing G, ${dG}/{dx}$ and ${d^2G}/{dx^2}$ as $\mathbb {F}_p(x)$ -linear combinations of $\sigma (H)$ and its successive derivatives (just by applying the chain and product rules), we find that G is also a solution of the same differential equation. Therefore, we must have

(2.2) $$ \begin{align} H=aG = a\cdot \sigma(H) \cdot (x+1)^{p-1} \end{align} $$

with $a\in \mathbb {F}_p$ . Recalling that

$$ \begin{align*} \sigma(H)=\sum_{n=0}^{p-1}\sum_{k=0}^n \binom{n}{k}^3\bigg(\frac{1-8x}{8+8x}\bigg)^n, \end{align*} $$ sigma left parenthesis upper H right parenthesis equals sigma summation Underscript n equals 0 Overscript p minus 1 Endscripts sigma summation Underscript k equals 0 Overscript n Endscripts StartBinomialOrMatrix n Choose k EndBinomialOrMatrix cubed left parenthesis StartFraction 1 minus 8 x Over 8 plus 8 x EndFraction right parenthesis Superscript n Baseline comma

sigma left parenthesis upper H right parenthesis equals sigma summation Underscript n equals 0 Overscript p minus 1 Endscripts sigma summation Underscript k equals 0 Overscript n Endscripts StartBinomialOrMatrix n Choose k EndBinomialOrMatrix cubed left parenthesis StartFraction 1 minus 8 x Over 8 plus 8 x EndFraction right parenthesis Superscript n Baseline comma

we evaluate (2.2) at $x=-1$ to obtain $H_{|x=-1}=a$ . We set and introduce the hypergeometric series

where denotes the Pochhammer symbol. We let also G denote the truncation at $y^p$ of g. Then, $h={g}/{(1-2x)}$ (see [Reference Straub and Zudilin21, Example 3.4]). Moreover, g is p-Lucas as a series in y and thus $G = g^{1-p}$ . From the fact that h is also p-Lucas, we deduce the equation

$$ \begin{align*} H=h^{1-p} = \frac{(1-2x)^{p-1}}{g^{p-1}}=(1-2x)^{p-1}G. \end{align*} $$ upper H equals h Superscript 1 minus p Baseline equals StartFraction left parenthesis 1 minus 2 x right parenthesis Superscript p minus 1 Baseline Over g Superscript p minus 1 Baseline EndFraction equals left parenthesis 1 minus 2 x right parenthesis Superscript p minus 1 Baseline upper G period

Evaluating this expression at $x=-1$ gives $a = G_{|x=-1} = G_{|y=1}$ .

Thus, if $p\equiv 1 \pmod 3$ , that is, $p\equiv 1 \pmod 6$ , we obtain the congruences

$$ \begin{align*}a\equiv \sum_{k=0}^{p-1}\binom{-1/3}{k}\binom{-2/3}{k} \equiv \sum_{k=0}^{p-1}\binom{(p-1)/3}{k}\binom{(2p-2)/3}{k}\equiv\binom{p-1}{(p-1)/3}\equiv 1 \pmod p.\end{align*} $$

a identical to sigma summation Underscript k equals 0 Overscript p minus 1 Endscripts StartBinomialOrMatrix negative 1 divided by 3 Choose k EndBinomialOrMatrix StartBinomialOrMatrix negative 2 divided by 3 Choose k EndBinomialOrMatrix identical to sigma summation Underscript k equals 0 Overscript p minus 1 Endscripts StartBinomialOrMatrix left parenthesis p minus 1 right parenthesis divided by 3 Choose k EndBinomialOrMatrix StartBinomialOrMatrix left parenthesis 2 p minus 2 right parenthesis divided by 3 Choose k EndBinomialOrMatrix identical to StartBinomialOrMatrix p minus 1 Choose left parenthesis p minus 1 right parenthesis divided by 3 EndBinomialOrMatrix identical to 1 left parenthesis mod p right parenthesis period

If $p\equiv 2 \pmod 3$ , that is, $p\equiv 5 \pmod 6$ , we obtain similarly

$$ \begin{align*}a\equiv \sum_{k=0}^{p-1}\binom{-1/3}{k}\binom{-2/3}{k} \equiv \sum_{k=0}^{p-1}\binom{(2p-1)/3}{k}\binom{(p-2)/3}{k}\equiv\binom{p-1}{(p-1)/3}\equiv -1 \pmod p.\end{align*} $$

We have used the Chu–Vandermonde identity $\sum _{k=0}^r\binom {r}{k}\binom {s}{k}=\binom {r+s}{r}$ and the classical congruence $\binom {p{-}1}{k}\equiv (-1)^k \pmod p$ .

Proposition 2.4. The field extension $\mathbb {F}_p(x,h^2)/\mathbb {F}_p(t)$ is an abelian Galois extension.

Proof. We first determine all the prolongations of $\sigma : \mathbb {F}_p(x) \to \mathbb {F}_p(x)$ to an automorphism of $\mathbb {F}_p(x, h^2)$ . Such a prolongation is uniquely determined by the value of $\sigma (h^2)$ , which needs to satisfy $\sigma (h^2)^e=\sigma (H)^{-1}$ . Thus,

(2.3) $$ \begin{align} \sigma(h^2) = u \cdot h^2 \cdot (x+1)^2 \end{align} $$

for some $u \in \mathbb {F}_p$ . Additionally, according to Lemma 2.3, u has to be a square if ${p\equiv 1 \pmod 6}$ and not a square otherwise. Consequently, there are precisely e prolongations of the automorphism $\sigma $ . There are clearly also e prolongations of the identity (given by the group S of squares in $\mathbb {F}_p$ acting by multiplication on $h^2$ ). Hence, the extension $\mathbb {F}_p(x,h^2)/\mathbb {F}_p(t)$ is Galois. Moreover, any prolongation of $\sigma $ clearly commutes with any element of S, making the Galois group abelian.

Lemma 2.5. There exists a prolongation of $\sigma $ to $\mathbb {F}_p(x, h^2)$ that is an involution if and only if one can take $u=\pm \tfrac {8}{9}$ in (2.3).

Proof. We have

$$ \begin{align*}\sigma^2(h^2)= \sigma(u h^2 (x+1)^2) = u^2 h^2 (x+1)^2 (\sigma(x)+1)^2= u^2 \big(\tfrac{9}{8}\big)^2 h^2.\end{align*} $$

The claim follows.

Proposition 2.6. The Galois group of $\mathbb {F}_p(x,h^2)/\mathbb {F}_p(t)$ is

$$ \begin{align*}\operatorname{\mathrm{Gal}}(\mathbb{F}_p(x,h^2)/\mathbb{F}_p(t)) = \begin{cases} \mathbb{Z}/(p{-}1)\mathbb{Z}\simeq \mathbb{Z}/e\mathbb{Z}\times \mathbb{Z}/2\mathbb{Z} & \text{if } p\equiv 3 \pmod 4,\\ \mathbb{Z}/(p{-}1)\mathbb{Z} & \text{if } p\equiv 13, 17 \pmod {24},\\ \mathbb{Z}/e\mathbb{Z}\times \mathbb{Z}/2\mathbb{Z}& \text{if } p\equiv 1, 5 \pmod {24}. \end{cases} \end{align*} $$

StartLayout 1st Row upper G a l left parenthesis double struck upper F Subscript p Baseline left parenthesis x comma h squared right parenthesis divided by double struck upper F Subscript p Baseline left parenthesis t right parenthesis right parenthesis equals StartLayout Enlarged left brace 1st Row 1st Column double struck upper Z divided by left parenthesis p minus 1 right parenthesis double struck upper Z asymptotically equals double struck upper Z divided by e double struck upper Z times double struck upper Z divided by 2 double struck upper Z 2nd Column if p identical to 3 left parenthesis mod 4 right parenthesis comma 2nd Row 1st Column double struck upper Z divided by left parenthesis p minus 1 right parenthesis double struck upper Z 2nd Column if p identical to 13 comma 17 left parenthesis mod 24 right parenthesis comma 3rd Row 1st Column double struck upper Z divided by e double struck upper Z times double struck upper Z divided by 2 double struck upper Z 2nd Column if p identical to 1 comma 5 left parenthesis mod 24 right parenthesis period EndLayout EndLayout

This corresponds respectively to the cases where precisely one of the two numbers ${u =\pm \frac {8}{9}}$ gives an involution $\sigma : h^2\mapsto u h^2 (1+x)^2$ in $\operatorname {\mathrm {Gal}}({}\mathbb {F}_p(x, h^2))/\mathbb {F}_p(t)$ , neither of them do or both of them do.

Proof. Note that $\frac {8}{9} = 2 \cdot (\tfrac 2 3)^2.$ Moreover, $2$ is a square modulo p if and only if ${p\equiv \pm 1 \pmod 8}$ and $-1$ is a square modulo p if and only if $p\equiv 1 \pmod 4$ .

If $p\equiv 1 \pmod 6$ then, for any prolongation of $\sigma $ , as in (2.3), u must be a square; if $p\equiv 5 \pmod 6$ , the opposite must hold. As for $p\equiv 3 \pmod 4$ , precisely one of the two values of $\pm \frac {8}{9}$ is a square, so we always have precisely one adequate choice for u. If $p\equiv 1 \pmod 4$ , either both of them are a square or neither of them are.

• If $p\equiv 1 \pmod {24}$ , then u has to be a square and $\pm \frac {8}{9}$ are both squares.
• If $p\equiv 5 \pmod {24}$ , then u has to be a nonsquare and $\pm \frac {8}{9}$ are both nonsquares.
• If $p\equiv 13 \pmod {24}$ , then u has to be a square, but $\pm \frac {8}{9}$ are both nonsquares.
• If $p\equiv 17 \pmod {24}$ , then u has to be a nonsquare, but $\pm \frac {8}{9}$ are both squares.

This concludes the proof of the second statement.

For the first statement, we note that $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(x,h^2)/\mathbb {F}_p(t))$ always contains a cyclic group of order e, corresponding to $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(x,h^2)/\mathbb {F}_p(x)) \simeq S$ . This leaves us only two possible choices: either $\mathbb {Z}/(p-1)\mathbb {Z}$ or $\mathbb {Z}/e\mathbb {Z}\times \mathbb {Z}/2\mathbb {Z}$ (which actually collapse in the case where $p\equiv 3 \pmod 4$ ). The direct factor $\mathbb {Z}/2\mathbb {Z}$ occurs if and only if there exists an element of the Galois group of order $2$ , which does not belong to S, that is, a prolongation of $\sigma $ which is an involution. We conclude using Lemma 2.5.

We recall that $f=h^2\cdot (x+1)$ . This shows that $\mathbb {F}_p(t, f)\subseteq \mathbb {F}_p(x, h^2).$ Moreover, $\mathbb {F}_p(t, f)$ is a cyclic extension of $\mathbb {F}_p(t)$ , of degree $p-1$ or e, as explained in [Reference Caruso, Fürnsinn and Vargas-Montoya6, Subsection 2.1.5].

Proof of Theorem 1.1

We distinguish cases according to the congruence class of p modulo $24$ .

• If $p\equiv 1 \pmod {24}$ , the Galois group of $\mathbb {F}_p(x, h^2)/\mathbb {F}_p(t)$ is not cyclic, according to Proposition 2.6. Thus, $\mathbb {F}_p(t, f)$ is a proper subfield of $\mathbb {F}_p(x, h^2)$ , necessarily of degree e over $\mathbb {F}_p(t)$ .
• If $p\equiv 5 \pmod {24}$ , the Galois group of $\mathbb {F}_p(x, h^2)/\mathbb {F}_p(t)$ is not cyclic, according to Proposition 2.6. We conclude again that $\mathbb {F}_p(f,t)$ has degree e over $\mathbb {F}_p(t)$ .
• If $p\equiv 7 \pmod {24}$ , the Galois group of $\mathbb {F}_p(x, h^2)/\mathbb {F}_p(t)$ is cyclic, because e is odd. Then, we need to check whether f is in the unique subfield of $\mathbb {F}_p(x, h^2)$ of index 2. This is equivalent to checking whether it is invariant under the unique element of the Galois group of order $2$ . According to Proposition 2.6, this element $\sigma $ is uniquely given as the prolongation of $x\mapsto {(1-8x)}/{(8+8x)}$ with $\sigma (h^2)=u\cdot h^2\cdot (x+1)^2$ , where the prefactor u has to be $\frac {8}{9}$ . We have
$$ \begin{align*}\textstyle \sigma(f)= \sigma(h^2) (\sigma(x)+1) = \frac{8}{9} h^2 (x+1)^2 (\sigma(x)+1) = h^2\cdot (x+1) = f.\end{align*} $$

So indeed, f is fixed by $\sigma $ and the extension has degree e.
• If $p\equiv 11 \pmod {24}$ , one proceeds as for $p \equiv 7 \pmod {24}$ , with the same conclusion.
• If $p\equiv 13 \pmod {24}$ , the group $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(x, h^2)/\mathbb {F}_p(t))$ is cyclic and the unique element of order $2$ in the Galois group is an element of S, sending $h^2$ to $-h^2$ . Thus, f is not fixed and $\mathbb {F}_p(t, f)\simeq \mathbb {F}_p(x, h^2)$ , having degree $p-1$ over $\mathbb {F}_p(t)$ .
• The case $p\equiv 17 \pmod {24}$ works as $p\equiv 13 \pmod {24}$ , with the same conclusion.
• If $p\equiv 19 \pmod {24}$ , we proceed as in the case $p\equiv 7 \pmod {24}$ , with the exception that $u=-\frac {8}{9}$ . Thus, $\sigma (f)=-f$ and f is not in the unique subfield of index $2$ of $\mathbb {F}_p(x, h^2)$ . Thus, $\mathbb {F}_p(t, f)$ has degree $p-1$ .
• If $p\equiv 23 \pmod {24}$ , we proceed as for $p\equiv 19 \pmod {24}$ , as $u=-\frac {8}{9}$ again.

This concludes the proof.

Proof of Theorem 1.2

We recall that we have $f = (1+x) h^2$ in $\mathbb {F}_p(t,f)$ . Therefore, $f^e = (1+x)^e H \in \mathbb {F}_p(x)$ , from which we deduce that $A_p = (1+x)^{p-1} H^2$ . In particular, $A_p$ is a square in $\mathbb {F}_p(x)$ . Since moreover $\mathbb {F}_p(x)=\mathbb {F}_p(t, \sqrt {\Delta })$ with ${\Delta = t^2-34t+1}$ , we can write $A_p=(u+v\sqrt {\Delta })^2$ for $u, v\in \mathbb {F}_p(t)$ . As $A_p\in \mathbb {F}_p(t),$ we must necessarily have $2uv=0$ . Hence, either $u = 0$ , in which case $A_p = (t^2-34t+1) v^2$ , or $v = 0$ , in which case $A_p = u^2$ . Given that $A_p$ is a square in $\mathbb {F}_p(t)$ if and only if the extension $\mathbb {F}_p(t,f)/\mathbb {F}_p(t)$ has degree e, we conclude using Theorem 1.1.

3 Adaptation to Apéry-like sequences

The generalised Apéry numbers $a_n(r,s)=\sum _{k=0}^n \binom {n}{k}^r \binom {n+k}{n}^s$ are p-Lucas for all primes p [Reference Deutsch and Sagan14]. So $f_{r,s}=A_{r,s} f_{r,s}^p$ for the generating series $f_{r,s}$ of $a_n(r,s)$ and its truncation $A_{r,s}$ at order p. For $(r,s)\in \{(0,0), (1, 0), (0,1), (1,1), (2,0)\}$ , the generating function $f_{r,s}$ is algebraic. (For $(r,s)=(0,1)$ , it is given by $(2t)^{-1}((1-4t)^{-1/2}-1)$ ; for $(r,s)=(1,1)$ , it is given by $(1 - 6t + t^2)^{-1/2}$ ; and for $(r,s)=(2,0)$ , it is given by $(1-4t)^{-1/2}$ .) Only for the cases $(r,s)\in \{(2,2), (4,0)\}$ , we computationally observed that $A_{r,s}$ is a square for a class of prime numbers depending on some congruence conditions: the case $(r,s)=(2,2)$ is the regular Apéry numbers, and the case $(r,s)=(4,0)$ is studied in Section 3.3 (see level $10$ in Table 2). So generalising the investigation in this direction does not show many new interesting patterns.

As explained in the introduction, the Domb numbers and the Almkvist–Zudilin numbers are closely related to Apéry numbers. The proofs of very similar patterns, as stated in Theorems 1.3 and 1.4, proceed analogously to the Apéry case. We report in Sections 3.1 and 3.2 on the changes that need to be made.

In the last section, Section 3.3, more sequences of similar flavour are analysed.

3.1 The Domb numbers

We set . The relation to $h(x)$ is given by ${(1-8x)}^{-1} f_\delta (t_\delta )=h(x)^2$ . In Lemma 2.1, we replace the involution $\sigma $ by $\sigma _\delta :x\mapsto {(1+x)}/{(8x-1)}$ , and the generator of the field extension $\mathbb {F}_p(x)/\mathbb {F}_p(t_\delta )$ is $(64t_\delta ^2-20t_\delta +1)^{1/2}$ .

As in Lemma 2.3, we obtain

$$ \begin{align*}H=\begin{cases} \sigma_\delta(H)\cdot (8x-1)^{p-1} & \text{if } p\equiv 1 \pmod 6,\\ - \sigma_\delta(H)\cdot (8x-1)^{p-1} & \text{if } p\equiv 5 \pmod 6. \end{cases}\end{align*} $$

StartLayout 1st Row upper H equals StartLayout Enlarged left brace 1st Row 1st Column sigma Subscript delta Baseline left parenthesis upper H right parenthesis dot left parenthesis 8 x minus 1 right parenthesis Superscript p minus 1 Baseline 2nd Column if p identical to 1 left parenthesis mod 6 right parenthesis comma 2nd Row 1st Column minus sigma Subscript delta Baseline left parenthesis upper H right parenthesis dot left parenthesis 8 x minus 1 right parenthesis Superscript p minus 1 Baseline 2nd Column if p identical to 5 left parenthesis mod 6 right parenthesis period EndLayout EndLayout

In Lemma 2.5, the automorphism $\sigma _\delta $ can be extended as $\sigma _\delta (h^2)=u \cdot h^2 \cdot (8x-1)^2$ for u being either a square or a nonsquare, depending on the congruence class of $p\bmod 6$ . Such a prolongation is an involution on $\mathbb {F}_p(x, h^2)$ if and only if one can take $u=\pm \frac {1}{9}$ . One of these values is always a square, while the other one is a square if and only if $p\equiv 1 \pmod 4$ . Thus,

$$ \begin{align*}\operatorname{\mathrm{Gal}}(\mathbb{F}_p(x, h^2)/\mathbb{F}_p(t_\delta)) = \begin{cases} \mathbb{Z}/(p-1)\mathbb{Z}\simeq \mathbb{Z}/e\mathbb{Z}\times \mathbb{Z}/2\mathbb{Z} & \text{if } p\equiv 7, 11 \pmod{12},\\ \mathbb{Z}/(p-1)\mathbb{Z} & \text{if } p\equiv 5 \pmod{12},\\ \mathbb{Z}/e\mathbb{Z}\times \mathbb{Z}/2\mathbb{Z}& \text{if } p\equiv 1 \pmod{12}, \end{cases} \end{align*} $$

again corresponding respectively to the case of one, zero or two prolongations of $\sigma _\delta $ to an involution on $\mathbb {F}_p(x, h^2)$ .

Finally, $f_\delta \in \mathbb {F}_p(x, h^2)$ because of the relations stated above. We distinguish the following cases.

• If $p\equiv 1\pmod {12}$ , then the Galois group of $\mathbb {F}_p(x, h^2)/\mathbb {F}_p(t_\delta )$ is not cyclic and $\mathbb {F}_p(t_\delta , f_\delta )$ is a subfield of $\mathbb {F}_p(x, h^2)$ of order e.
• If $p\equiv 5\pmod {12}$ , then the Galois group $\mathbb {F}_p(x, h^2)/\mathbb {F}_p(t_\delta )$ is cyclic. The question then becomes to determine whether $f_\delta $ is fixed by $\sigma _\delta $ . However, the unique element of order $2$ in the group in this case belongs to the group S sending $h^2$ to $-h^2$ and thus $f_\delta $ is not fixed.
• If $p \equiv 7 \pmod {12}$ , then the Galois group of $\mathbb {F}_p(x, h^2)/\mathbb {F}_p(t_\delta )$ is cyclic. There is one involution extending $\sigma _\delta $ , corresponding to $u=\frac {1}{9}$ . We compute
$$ \begin{align*}\textstyle \sigma_\delta(f_\delta)= \sigma_\delta(h^2) (1-8\sigma_\delta(x)) = \frac{1}{9} h^2 (1-8x)^2 (1-8\sigma_\delta(x)) = h^2\cdot (1-8x) = f_\delta,\end{align*} $$
so $f_\delta $ is fixed and thus contained in a proper subfield of $\mathbb {F}_p(x, h^2)$ of order e.
• If $p \equiv 11 \pmod {12}$ , the prolongation of $\sigma _\delta $ corresponds to $u=-\frac {1}{9}$ and, this time, $f_\delta $ changes sign under $\sigma _\delta $ .

3.2 The Almkvist–Zudilin numbers

For the Almkvist–Zudilin numbers, we set and find the involution of $\mathbb {F}_p(x)$ fixing $\mathbb {F}_p(t_\xi )$ . This time, we obtain the simple relation $H=\sigma (H)\cdot x^{p-1}$ for all prime numbers p. The involution $\sigma _\xi $ can be extended to $\mathbb {F}_p(x, h^2)$ by $\sigma _\xi (h)^2=u h^2 x^2$ with $u\in S$ ; it is an involution if and only if we can take $u=\pm 8$ . Now, $8$ is a square if and only if $p\equiv \pm 1 \pmod 8$ and $-1$ is a square if and only if $p\equiv 1 \pmod 4$ . Thus,

$$ \begin{align*} \operatorname{\mathrm{Gal}}(\mathbb{F}_p(x, h^2)/\mathbb{F}_p(t_\xi)) = \begin{cases} \mathbb{Z}/(p-1)\mathbb{Z}\simeq \mathbb{Z}/e\mathbb{Z}\times \mathbb{Z}/2\mathbb{Z} & \text{if } p\equiv 3,7 \pmod 8,\\ \mathbb{Z}/(p-1)\mathbb{Z} & \text{if } p\equiv 5 \pmod 8,\\ \mathbb{Z}/e\mathbb{Z}\times \mathbb{Z}/2\mathbb{Z}& \text{if } p\equiv 1 \pmod 8. \end{cases} \end{align*} $$

Finally, $f_\xi \in \mathbb {F}_p(x, h^2)$ and investigations depending on the congruence class of $p\pmod 8$ yield the following statements.

• If $p\equiv 1 \pmod 8$ , the Galois group $\operatorname {\mathrm {Gal}}(\mathbb {F}_p(x, h^2)/\mathbb {F}_p(t_\xi ))$ is not cyclic and thus, $f_\xi $ is in a proper subfield of degree e.
• If $p\equiv 3 \pmod 8$ , the unique prolongation of $\sigma _\xi $ as an involution corresponds to $u=-8$ and because of
$$ \begin{align*}\textstyle \sigma_\xi (f_\xi)= \sigma(h^2) (1+\sigma_\xi(x))(1-8\sigma_\xi(x)) = h^2 (1-8x)(1+x)= f_\xi,\end{align*} $$
we conclude that $f_\xi $ lies in a proper subfield of $\mathbb {F}_p(x, h^2)$ of degree e.
• If $p\equiv 5 \pmod 8$ , the unique element of order $2$ in $\operatorname {\mathrm {Gal}}({}\mathbb {F}_p(x, h^2))/\mathbb {F}_p(t_\xi )$ belongs to the subgroup of squares of $\mathbb {F}_p^\times $ and sends h to $-h$ and thus does not fix $f_\xi $ . So $\mathbb {F}_p(t_\xi , f_\xi )=\mathbb {F}_p(x, h)$ .
• If $p\equiv 7 \pmod 8$ , the prolongation of $\sigma _\xi $ as an involution corresponds to $u=+8$ and, this time, $f_\xi $ changes sign under it and thus is not fixed by it. Again, $\mathbb {F}_p(t_\xi , f_\xi )=\mathbb {F}_p(x, h)$ .

3.3 A zoo of further examples

In this section, we present further sequences which seem to exhibit similar, sometimes a bit more intricate, patterns as the three sequences we discussed so far. Our observations are of a computational nature, although we believe that working out proofs deploying similar techniques is possible.

First, we consider the three other sequences of Zagier’s sporadic examples of integral sequences satisfying a three-term recurrence relation [Reference Zagier, Harnad and Winternitz24]. It is known that these sequences are p-Lucas for all primes p [Reference Malik and Straub20]; therefore, their generating series $f(t)$ satisfies the relation $f(t) \equiv A_p(t) f(t)^p \pmod p$ , where $A_p(t) \in \mathbb {F}_p[t]$ is the truncation at $t^p$ of $f(t) \bmod p$ . As before, we are interested in the factorisation of $A_p(t)$ of the form $A_p(t) = P(t) B_p(t)^2$ . Table 1 shows how $P(t)$ seems to vary with respect to p.

Table 1

Zagier’s sporadic examples [Reference Almkvist, van Straten and Zudilin2, Reference Zagier, Harnad and Winternitz24].

A table listing Zagier's sporadic examples across six rows and three columns, detailing mathematical sequences, Legendre symbol conditions, and polynomial coefficients. See long description.

Table 1 Long description

The table is organized into three columns: Sequence, Conditions, and Coefficient P of t.

Row 1: O E I S A 2 2 9 1 1 1. Condition: Legendre symbol negative 1 over p equals 1. Coefficient: 1.

Row 2: 2 times negative 1 to the n power times the summation from k equals 0 to n of binomial n k cubed times the quantity binomial 4 n minus 5 k minus 1 over 3 n plus binomial 4 n minus 5 k over 3 n. Condition: Legendre symbol negative 1 over p equals negative 1. Coefficient: 1 minus 2 2 t plus 1 2 5 t squared.

Row 3: O E I S A 2 9 0 5 7 5. Condition: Legendre symbol negative 2 over p equals 1. Coefficient: 1.

Row 4: Summation from k equals 0 to n of binomial n k squared times binomial 2 k over n squared. Condition: Legendre symbol negative 2 over p equals negative 1. Coefficient: 1 minus 2 4 t plus 1 6 t squared.

Row 5: O E I S A 2 9 0 5 7 6. Condition: Legendre symbol negative 1 over p equals 1. Coefficient: 1.

Row 6: Double summation from k equals 0 to n and l equals 0 to n of binomial n k squared times binomial n l times binomial k l times binomial k plus l over n. Condition: Legendre symbol negative 1 over p equals negative 1. Coefficient: 1 minus 1 8 t plus 2 7 t squared.

Table 2

Examples of sequences connected to modular forms.

A table listing mathematical sequences connected to modular forms, categorized by level, conditions, and coefficients. See long description.

Table 2 Long description

The table contains four columns: Sequence, Level, Conditions, and Coefficient P of t.

* Row 1: O E I S A 2 7 4 7 8 6. Level 5. Condition: Legendre symbol -5 over p equals 1. Coefficient: 1.

* Row 2: Same sequence and level. Condition: Legendre symbol -5 over p equals -1. Coefficient: 1 minus 4 t minus 16 t squared.

* Row 3: O E I S A 1 8 1 4 1 8. Level 6. Condition: Legendre symbols -3 over p and -6 over p both equal 1. Coefficient: 1.

* Row 4: Same sequence and level. Condition: -3 over p equals -1 and -6 over p equals 1. Coefficient: 1 plus 4 t.

* Row 5: Same sequence and level. Condition: -3 over p equals 1 and -6 over p equals -1. Coefficient: 1 minus 32 t.

* Row 6: Same sequence and level. Condition: -3 over p and -6 over p both equal -1. Coefficient: open parenthesis 1 plus 4 t close parenthesis times open parenthesis 1 minus 32 t close parenthesis.

* Row 7: O E I S A 1 8 3 2 0 4. Level 7. Condition: -7 over p equals 1. Coefficient: 1.

* Row 8: Same sequence and level. Condition: -7 over p equals -1. Coefficient: open parenthesis 1 plus t close parenthesis times open parenthesis 1 minus 27 t close parenthesis.

* Row 9: O E I S A 0 0 5 2 6 0. Level 10. Condition: -5 over p and -10 over p both equal 1. Coefficient: 1.

* Row 10: Same sequence and level. Condition: -5 over p equals -1 and -10 over p equals 1. Coefficient: 1 plus 4 t.

* Row 11: Same sequence and level. Condition: -5 over p equals 1 and -10 over p equals -1. Coefficient: 1 minus 16 t.

* Row 12: Same sequence and level. Condition: -5 over p and -10 over p both equal -1. Coefficient: open parenthesis 1 plus 4 t close parenthesis times open parenthesis 1 minus 16 t close parenthesis.

* Row 13: O E I S A 2 8 4 7 5 6. Level 11. Condition: -11 over p equals 1. Coefficient: 1.

* Row 14: Same sequence and level. Condition: -11 over p equals -1. Coefficient: 1 minus 20 t plus 56 t squared minus 44 t cubed.

* Row 15: O E I S A 3 8 6 2 8 3. Level 17. Condition: -17 over p equals 1. Coefficient: 1.

* Row 16: Same sequence and level. Condition: -17 over p equals -1. Coefficient: 1 minus 16 t minus 66 t squared minus 48 t cubed minus 127 t to the 4 power.

* Row 17: O E I S A 2 1 9 6 9 2. Level 18. Condition: -1 over p and -2 over p both equal 1. Coefficient: 1.

* Row 18: Same sequence and level. Condition: -1 over p equals -1 and -2 over p equals 1. Coefficient: 1 minus 12 t.

* Row 19: Same sequence and level. Condition: -1 over p equals 1 and -2 over p equals -1. Coefficient: 1 minus 16 t.

* Row 20: Same sequence and level. Condition: -1 over p and -2 over p both equal -1. Coefficient: open parenthesis 1 minus 12 t close parenthesis times open parenthesis 1 minus 16 t close parenthesis.

* Row 21: O E I S A 3 8 6 7 0 8. Level 20. Condition: -1 over p and -5 over p both equal 1. Coefficient: 1.

* Row 22: Same sequence and level. Condition: -1 over p equals -1 and -5 over p equals 1. Coefficient: 1 minus 4 t.

* Row 23: Same sequence and level. Condition: -1 over p equals 1 and -5 over p equals -1. Coefficient: 1 minus 12 t plus 16 t squared.

* Row 24: Same sequence and level. Condition: -1 over p and -5 over p both equal -1. Coefficient: open parenthesis 1 minus 4 t close parenthesis times open parenthesis 1 minus 12 t plus 16 t squared close parenthesis.

* Row 25: O E I S A 3 8 6 7 9 3. Level 23. Condition: -23 over p equals 1. Coefficient: 1.

* Row 26: Same sequence and level. Condition: -23 over p equals -1. Coefficient: open parenthesis 1 minus 3 t plus 2 t squared plus t cubed close parenthesis times open parenthesis 1 minus 11 t plus 22 t squared minus 19 t cubed close parenthesis.

We pursue our investigations with other sequences associated to modular functions (of a given level) in Table 2. Thanks to [Reference Adamczewski, Bell and Delaygue1, Reference Beukers, Tsai and Ye4], all the series appearing in this table are p-Lucas except for levels $17$ , $20$ and $23$ . However, in these cases, they continue to satisfy a relation of the form $f(t) \equiv A_p(t) f(t)^p \pmod p$ , where now, $A_p(t) \in \mathbb {F}_p(t)$ is a rational function; this follows from the main theorem of [Reference Vargas-Montoya23]. As a conclusion, in all cases, $A_p(t)$ is well defined and it makes sense to study its factorisation as ${A_p(t) = P(t) B_p(t)^2}$ as before.

In all the examples of Tables 1 and 2, we computationally observed that the patterns for the polynomials $P(t)$ depend on explicit quadratic residue conditions for the prime p. These patterns are very much in line with those recorded in Theorems 1.1, 1.2, 1.3 and 1.4, and we are confident that one can apply, case by case, a similar strategy to establish the claims in the tables rigorously. The principal feature of the underlying series $f(t)$ is the presence of suitable rational parametrisations $t=t(x)$ such that $f(t(x))=\rho (x)h(x)^2$ for a function $h(x)$ solving a second-order linear differential equation and, in turn, related to an arithmetic $_2F_1$ hypergeometric function $g(y)$ via another parametrisation $y=y(x)$ , namely, $h(x)=\lambda (x)g(y(x))$ . In Apéry’s case (but also for the Domb and Almkvist–Zudilin sequences), all the intermediate functions are rational in x:

$$ \begin{align*} t(x)=\frac{x(1-8x)}{1+x}, \quad \rho(x)=1+x, \quad y(x)=\frac{27x^2}{(1-2x)^3}, \quad \lambda(x)=\frac1{1-2x}. \end{align*} $$

However, in general, $\rho (x)$ , $y(x)$ and $\lambda (x)$ are algebraic, which can make the structure of the corresponding Galois group more involved. The existence of such x-parametrisations is a consequence of modular parametrisations of all such Apéry-like sequences; the details of the latter can be found in the corresponding references from which our examples originate. We leave the details to the interested reader as exercises.

Going further, the observations on the splitting pattern of $A_p(t)$ as $P(t)B_p(t)^2$ suggest that the conditions on p depend only on the quadratic residues of the negative divisors of the squarefree part of the corresponding level. We also leave it to the interested reader to formulate and prove a precise statement about the pattern that unfolds here. We conclude by remarking that these observations provide more computational evidence for the conjectures that were formulated in [Reference Caruso, Fürnsinn and Vargas-Montoya6] on uniformity properties of Galois groups of reductions of D-finite series.

Acknowledgements

We thank Alin Bostan for his interest in this project and for pointing out that the Domb numbers and the Almkvist–Zudilin numbers share the same behaviour as the Apéry numbers. F.F. and W.Z. thank the Max Planck Institute for Mathematics (Bonn, Germany) for their hospitality and financial support in May 2025, where the initial discussion on the project commenced.

Footnotes

F.F. was funded by a DOC Fellowship (27150) of the Austrian Academy of Sciences at the University of Vienna. Further he thanks the French–Austrian project EAGLES (ANR-22-CE91-0007 and FWF grant 10.55776/I6130) for financial support. X.C., F.F. and D.V.-M. thank Austria’s Agency for Education and Internationalisation (OeAD) and Campus France for providing funding for research stays via WTZ collaboration project/Amadeus project FR02/2024. The work of W.Z. was supported in part by the NWO grant OCENW.M.24.112.

References

Adamczewski, B., Bell, J. and Delaygue, É., ‘Algebraic independence of $G$

-functions and congruences “à la Lucas”’, Ann. Sci. Éc. Norm. Supér. (4) 52(3) (2019), 515–559.CrossRef Google Scholar

Almkvist, G., van Straten, D. and Zudilin, W., ‘Generalizations of Clausen’s formula and algebraic transformations of Calabi–Yau differential equations’, Proc. Edinb. Math. Soc. (2) 54(2) (2011), 273–295.CrossRef Google Scholar

Apéry, R., ‘Irrationalité de $\zeta (2)$

et $\zeta (3)$

’, Astérisque 61 (1979), 11–13.Google Scholar

Beukers, F., Tsai, W.-L. and Ye, D., ‘Lucas congruences using modular forms’, Bull. Lond. Math. Soc. 57(1) (2025), 69–78.CrossRef Google Scholar

Bostan, A., Boukraa, S., Maillard, J.-M. and Weil, J.-A., ‘Diagonals of rational functions and selected differential Galois groups’, J. Phys. A 48(50) (2015), 504001–504030.CrossRef Google Scholar

Caruso, X., Fürnsinn, F. and Vargas-Montoya, D., ‘Galois groups of reductions modulo $p$

of D-finite series’, Preprint, 2025, arXiv:2504.09429 [math.NT].Google Scholar

Chan, H. H. and Verrill, H., ‘The Apéry numbers, the Almkvist–Zudilin numbers and new series for $1/\pi$

’, Math. Res. Lett. 16(3) (2009), 405–420.CrossRef Google Scholar

Chan, H. H. and Zudilin, W., ‘New representations for Apéry-like sequences’, Mathematika 56(1) (2010), 107–117.CrossRef Google Scholar

Christol, G., ‘Fonctions et éléments algébriques’, Pacific J. Math. 125(1) (1986), 1–37.CrossRef Google Scholar

Christol, G., ‘Fonctions hypergéométriques bornées’, Groupe Travail Anal. Ultramétrique 8(14) (1986–1987), 1–16.Google Scholar

Cooper, S., ‘Sporadic sequences, modular forms and new series for $1/\pi$

’, Ramanujan J. 29(1–3) (2012), 163–183.CrossRef Google Scholar

Cooper, S., Ramanujan’s Theta Functions (Springer, Cham, 2017).CrossRef Google Scholar

Cooper, S., Ge, J. and Ye, D., ‘Hypergeometric transformation formulas of degrees 3, 7, 11 and 23’, J. Math. Anal. Appl. 421(2) (2015), 1358–1376.CrossRef Google Scholar

Deutsch, E. and Sagan, B. E., ‘Congruences for Catalan and Motzkin numbers and related sequences’, J. Number Theory 117(1) (2006), 191–215.CrossRef Google Scholar

Furstenberg, H., ‘Algebraic functions over finite fields’, J. Algebra 7 (1967), 271–277.CrossRef Google Scholar

Gessel, I., ‘Some congruences for Apéry numbers’, J. Number Theory 14(3) (1982), 362–368.CrossRef Google Scholar

Huber, T., Schultz, D. and Ye, D., ‘Series for $1/\pi$

of level 20’, J. Number Theory 188 (2018), 121–136.CrossRef Google Scholar

Huber, T., Schultz, D. and Ye, D., ‘Level 17 Ramanujan–Sato series’, Ramanujan J. 52(2) (2020), 303–322.CrossRef Google Scholar

Huber, T., Schultz, D. and Ye, D., ‘Ramanujan–Sato series for $1/\pi$

’, Acta Arithmetica 207(2) (2023), 121–160.CrossRef Google Scholar

Malik, A. and Straub, A., ‘Divisibility properties of sporadic Apéry-like numbers’, Res. Number Theory 2 (2016), 5.CrossRef Google Scholar

Straub, A. and Zudilin, W., ‘Positivity of rational functions and their diagonals’, J. Approx. Theory 195 (2015), 57–69.CrossRef Google Scholar

Vargas-Montoya, D., ‘Algébricité modulo $p$

, séries hypergéométriques et structures de Frobenius fortes’, Bull. Soc. Math. France 149(3) (2021), 439–477.Google Scholar

Vargas-Montoya, D., ‘Monodromie unipotente maximale, congruences “à la Lucas” et indépendance algébrique’, Trans. Amer. Math. Soc. 377 (2023), 167–202.Google Scholar

Zagier, D., ‘Integral solutions of Apéry-like recurrence equations’, in: Groups and Symmetries, CRM Proceedings and Lecture Notes, 47 (eds. Harnad, J. and Winternitz, P.) (American Mathematical Society, Providence, RI, 2009), 349–366.CrossRef Google Scholar

Table 1 Zagier’s sporadic examples [2, 24].Table 1 Long description.

Table 2 Examples of sequences connected to modular forms.Table 2 Long description.

Article contents

GALOIS GROUPS OF APÉRY-LIKE SERIES MODULO PRIMES

Abstract

Keywords

MSC classification

Information

1 Introduction

2 The Apéry numbers

Proof of Theorem 1.1

Proof of Theorem 1.2

3 Adaptation to Apéry-like sequences

3.1 The Domb numbers

3.2 The Almkvist–Zudilin numbers

3.3 A zoo of further examples

Acknowledgements

Footnotes

References

Save article to Kindle

Save article to Dropbox

Save article to Google Drive

Reply to: Submit a response

Your details

You have entered the maximum number of contributors

Conflicting interests