betaFeedback

WorksMathematics

On a Problem of Gordon and Hughes: The 11-adic Properties of Meromorphic Modular Functions on X0(22)\mathrm{X}_0(22)

Abstract

Congruence families occur throughout the theory of modular forms. These families superficially resemble one another, though they can vary substantially in their overall difficulty. One major complication resides in the topology of the modular curve associated with a given congruence family. All previously studied families are associated with curves of genus 0 or 1. The existence of a family associated with a curve of genus 2 (and the associated difficulties of a proof thereof) was first speculated upon by Gordon and Hughes. We give the first explicit congruence family associated with X0(22)\mathrm{X}_0(22), a modular curve of genus 2, and prove it. We also propose two families of conjectures associated with the same modular curve.

Introduction

In 1981 Gordon and Hughes studied Ramanujan type congruence families for the distinct partition function, denoted by q(n)q(n), modulo powers of 5 and 7. They [16 Section 1] speculated on the possibility of analogous congruences by powers of 11:

“As far as congruences for q(n)q(n) mod 11γ11^\gamma are concerned, the problem is greatly complicated by the fact that the group Γ0(22)\Gamma_0(22) has genus 2. The computational difficulties seem too formidable to justify pursuing this problem here.”

It is nearly a half-century since this paper was published, and no congruence families have been found associated with Γ0(22)\Gamma_0(22) or any other congruence subgroup corresponding to a genus 2 modular curve. In this paper, we give an explicit example of such a family and propose a broader class of such genus 2 congruence families. The congruences are exhibited by the 2-elongated plane partition diamond function d2(n)d_2(n) rather than q(n)q(n); nevertheless, the associated group is Γ0(22)\Gamma_0(22), as Gordon and Hughes had anticipated. Likewise, the difficulties, theoretical and computational, are as formidable as Gordon and Hughes foresaw. The principal theoretical novelty of the proof is the use of localizing functions that are not modular units, i.e., whose divisors are not supported on the cusps. We believe that the methods developed to handle such functions have implications beyond the present family, both for modular congruences and for the ℓ\ell-adic theory of modular forms more generally.

Background

The theory of modular congruence relationships has been an active area of research since Ramanujan’s groundbreaking work [22] on the divisibility properties of the partition function. Consider the original congruence families discovered and proved by Ramanujan, Watson [24], and Atkin [6]:

If n,α∈N and 24n≡1(modℓα), then p(n)≡0(modℓβ),(1)\text{If } n,\alpha\in\mathbb{N} \text{ and } 24n \equiv1 \pmod{\ell^\alpha}, \text{ then } p(n) \equiv0 \pmod{\ell^\beta}, \tag*{(1)}

with

β:={αif ℓ∈{5,11},⌊α2⌋+1if ℓ=7.\beta:= \begin{cases} \alpha& \text{if } \ell\in\{5,11\},\\ \left\lfloor\frac{\alpha}{2} \right\rfloor+ 1 & \text{if } \ell= 7. \end{cases}

Since the proof of these families, the arithmetic of the partition function has been an active area of research. In his seminal work, Ono [18] proved that for every prime ℓ≥5\ell\ge5 there are infinitely many congruences of the form p(An+B)≡0(modℓ)p(An+B) \equiv0 \pmod{\ell}, and Ahlgren [1] extended this to all moduli coprime to 6 (see also Ahlgren and Ono [4]). Ahlgren and Boylan [3] used tools from the theory of Galois representations and modular forms to show that Ramanujan’s partition congruences are the only ones which are of the form p(ℓn+β)≡0(modℓ)p(\ell n+\beta) \equiv0 \pmod{\ell}. For prime powers, Folsom, Kent, and Ono [13] and Boylan and Webb [10] established a strong form of ℓ\ell-adic stability for the UℓU_\ell-action. Ahlgren, Allen, and Tang [2] proved that congruences like Atkin’s hold modulo every prime ℓ≥5\ell\ge5.

2010 Mathematics Subject Classification. 11P81, 11P82, 11P83, 30F35.

Key words and phrases. Elongated plane partitions, localization method, modular curve, modular forms, partition congruences, Riemann surface.

This richness of arithmetic properties has been found to exist within the coefficients of many different modular forms. We are particularly interested in ℓ\ell-adic patterns analogous to (1.1). A striking aspect of the subject is the substantial variability in the difficulty of a given congruence family. Some admit short proofs by methods that were already available to Ramanujan. Others remain open to this day.

One major difficulty lies in the topology of the associated modular curve, e.g., the cusp count and the genus. The most difficult problems generally occur when the associated modular curve has many cusps. Naively speaking, in these cases the action of the relevant UℓU_{\ell} operator can no longer be controlled within a single, finitely generated space of modular functions, and the ℓ\ell-adic estimates required to close the induction break down. Families of this kind, such as those arising from traces of singular moduli and Hurwitz class numbers, have so far resisted even the most sophisticated techniques currently available [9, 17].

Another significant factor in the difficulty of a congruence family lies with the genus. For example, the associated modular curve for (1.1) is simply X0(ℓ)X_0(\ell). In all three cases the cusp count is 2. However, for ℓ∈{5,7}\ell\in\{5, 7\}, the curve has genus 0, while the curve has genus 1 for ℓ=11\ell= 11. Watson, who had not completed the proof of (1.1) for ℓ=11\ell= 11, speculated in 1938 that the proof would not be especially difficult, but perhaps “recht langweilig” [24]. Upon his completion of the proof for ℓ=11\ell= 11 in 1967, Atkin [6 Section 5] agreed with Watson’s speculation.

It has long been understood that families associated with curves of positive genus are difficult to treat by conventional techniques. In addition to the aforementioned remarks by Gordon and Hughes, Calegari [12 Section 4.2] has also pointed out that the standard proof methods “become essentially infeasible as soon as [the associated modular curve] has genus >0> 0”. The difficulty of a proof notwithstanding, it is interesting to note that no explicit congruence families for genus 2 appear to have been conjectured prior to this work. To the best of our knowledge, the possibility of such a family has been raised in the literature only by Gordon and Hughes [16]. To test the expectation that such a family should be accessible to current methods, we undertook a search for a suitable example. A computational search for q(n)q(n), the function originally considered by Gordon and Hughes, produced no such candidate. A natural alternative is provided by the Fourier coefficients of the following eta quotient, whose congruence families have already served as a valuable testing ground for genus 0 and 1 [7, 8], and which therefore suggested itself as a promising source of a genus 2 example:

Dk(q):=∑n≥0dk(n)qn=(q2;q2)∞k(q;q)∞3k+1.D_k(q) := \sum_{n\geq0} d_k(n)q^n = \frac{(q^2;q^2)_\infty^k}{(q;q)_\infty^{3k+1}}.

The Fourier coefficients dk(n)d_k(n) have an elegant combinatorial interpretation, discovered by Andrews and Paule [5] in their “Partition Analysis” series of papers. It is among the Fourier coefficients of D2(q)D_2(q) that we find a relevant family of interest, the main result of this paper:

Theorem 1.1. Let n,α∈Z≥1n,\alpha\in\mathbb{Z}_{\geq1} such that 8n≡1(mod11α)8n \equiv1 \pmod{11^\alpha}. Then

d2(n)≡0(mod11α+12).d_2(n) \equiv0 \pmod{11^{\frac{\alpha+1}{2}}}.

Localization and difficulty of the proof

Gordon and Hughes’s assessment of the difficulty proves accurate. The genus of X0(22)X_0(22), together with difficulties in its corresponding space of eta quotients, necessarily complicates our attempts at a proof of Theorem 1.1. As in previous approaches (see [7, 8]) via modular equation manipulations, the idea was straightforward: to study d2(n)d_2(n) in the progressions relevant to us, we construct a sequence of functions LαL_\alpha of the form

Lα:=ϕα∑n≥08n≡1(mod11α)d2(n)q⌊n11α⌋+1.L_\alpha:= \phi_\alpha\sum_{\substack{n\geq0\\8n\equiv1\pmod{11^\alpha}}} d_2(n)q^{\left\lfloor\frac{n}{11^\alpha}\right\rfloor+1}.

With the right holomorphic prefactors ϕα\phi_\alpha, we can arrange so that LαL_\alpha is a modular function over Γ0(22)\Gamma_0(22), i.e., a meromorphic function on X0(22)X_0(22) with poles only at the cusps. We can construct a pair of modified Atkin U11U_{11} operators U(i)U^{(i)}, with i∈{0,1}i \in\{0,1\}, such that

U(a)(Lα)=Lα+1,U^{(a)}(L_\alpha)=L_{\alpha+1},

with a≡α(mod2)a \equiv\alpha\pmod2. We want to place LαL_\alpha into a suitable subspace of the meromorphic functions over X0(22)X_0(22), in which the basis functions are subject to certain modular equations. We then want to study how U(i)U^{(i)} affects those basis functions. This is done by directly examining a handful of initial cases, and then using the associated modular equations to construct more general relationships. Recall that X0(22)X_0(22) is a compact Riemann surface with cusps

{[∞],[111],[12],[0]}.\left\{[\infty],\left[\frac{1}{11}\right],\left[\frac{1}{2}\right],[0]\right\}.

Since X0(22)X_0(22) is compact, every non-constant meromorphic function on it has a pole. Generally, we prefer to restrict these poles to the cusps. We consider the space of holomorphic functions over the curve except for a pole (of possibly large order) at the cusp [0][0], and denote this space as M(0)(X0(22))\mathcal{M}^{(0)}(X_0(22)). Using a straightforward computational procedure, we can use the Dedekind eta quotient to construct suitable functions with integer Fourier coefficients, and with orders −3-3, −4-4, −5-5 at [0][0]. We denote these functions x,y,zx,y,z, respectively. Their explicit form is given below (see (2.3)–(2.5)).

Combinations of these three functions can be used to construct functions of order −n-n for n≥6n \ge6. The only missing orders are −1-1, −2-2. By the Weierstrass gap theorem [20] exactly two pole orders at [0][0] are not attained, i.e., as xx, yy, zz have orders −3-3, −4-4, −5-5, the gaps are 1 and 2. Thus, we have

M(0)(X0(22))=C[x]⊕yC[x]⊕zC[x].\mathcal{M}^{(0)}(X_0(22))=\mathbb{C}[x]\oplus y\mathbb{C}[x]\oplus z\mathbb{C}[x].

Suppose LαL_\alpha has a pole at [0][0] and nowhere else. Then we could express it as a suitable polynomial combination of xx, yy, and zz. However, the poles of LαL_\alpha are supported at [0][0] and [12]\left[\frac{1}{2}\right]. In context of the localization method, we can multiply LαL_\alpha by some modular eta quotient which lives in M(0)(X0(22))\mathcal{M}^{(0)}(X_0(22)), but has positive order at [12]\left[\frac{1}{2}\right]. Indeed, a very natural candidate exists. Consider,1

w:=(q2;q2)∞11(q11;q11)∞11(q;q)∞11(q22;q22)∞11=1+11y+112z.w:=\frac{(q^2;q^2)_\infty^{11}(q^{11};q^{11})_\infty^{11}}{(q;q)_\infty^{11}(q^{22};q^{22})_\infty^{11}}=1+11y+11^2z.

Indeed, w3L1∈M(0)(X0(22))w^3L_1\in\mathcal{M}^{(0)}(X_0(22)). We can therefore express L1L_1 as a rational polynomial with summands of the form

yjzkxm(1+11y+112z)3\frac{y^jz^kx^m}{(1+11y+11^2z)^3}

with m∈N0m\in\mathbb{N}_0, j,k∈{0,1}j,k\in\{0,1\}, and j+k≤1j+k\le1.

However, such an approach raises various difficulties upon attempts to apply it to LαL_\alpha generally. In the first place, the denominator is governed by a degree 55 modular equation, while xx is governed by a degree 33 modular equation. As such, we will be forced to examine how U(i)U^{(i)} affects 1815 different combinations, with an additional factor of 2 for the value of ii, and a factor of 3 for the possible presence of yy and zz. This immediately forces well over 10410^4 large polynomial relations to be verified. Moreover, the presence of yy and zz in ww ensures that one must tediously remove monomials yazby^az^b with a+b≥2a+b\ge2 from resultant representations. Replacing zz by ww does not resolve this complication, because such a substitution would incur high powers of 11 in the denominators of coefficients in LαL_\alpha. This immediately destroys the utility of LαL_\alpha as a witness to divisibility by powers of 11. Another alternative is to work with a C[w,w−1]\mathbb{C}[w,w^{-1}] module of rank 5. In this case, one still requires the localization technique and needs to verify at least 3⋅1043\cdot10^4 associated relations. Finally, the methods of Paule and Radu [19], and of Garvan, Sellers, and the second author [15 Section 3] do not appear to adapt to the present setting.

These obstacles are consistent with Gordon and Hughes’s original assessment. We instead construct the localizing factor from xx alone. To remove the poles of LαL_\alpha at [12]\left[\frac{1}{2}\right], note that the expansion of xx near [12]\left[\frac{1}{2}\right] in terms of the local variable q^\widehat{q} is the following

x(τ2τ+1)=−211+bq^+O(q^2)with b≠0.x\left(\frac{\tau}{2\tau+1}\right)=-\frac{2}{11}+b\widehat{q}+O(\widehat{q}^{2})\quad\text{with }b\ne0.

Thus ord⁡[1/2](2+11x)=1\operatorname{ord}_{[1/2]}(2+11x)=1 and

(2+11x)ψLα∈⨁v∈{x,y,z}vC[x].(2)(2+11x)^\psi L_\alpha\in\bigoplus_{v\in\{x,y,z\}}v\mathbb{C}[x]. \tag*{(2)}

References

  1. [1]S. Ahlgren, Distribution of the partition function modulo composite integers M, Math. Ann. 318 (2000), 795–803.DOI
  2. [2]S. Ahlgren, P. B. Allen, and S. Tang, Congruences like Atkin’s for the partition function, Trans. Amer. Math. Soc. Ser. B 9 (2022), 1044–1064.arxiv.org/abs/2112.09481
  3. [3]S. Ahlgren and M. Boylan, Arithmetic properties of the partition function, Invent. Math. 153 (2003), 487–502.DOI
  4. [4]S. Ahlgren and K. Ono, Congruence properties for the partition function, Proc. Natl. Acad. Sci. USA 98 (2001), 12882–12884.pubmed.ncbi.nlm.nih.gov/11606715
  5. [5]G. Andrews and P. Paule, MacMahon’s partition analysis XIII: Schmidt type partitions and modular forms, J. Number Theory. 234 (2022), 95–119.DOI
  6. [6]A. Atkin, Proof of a conjecture of Ramanujan, Glasgow Math. J. 8 (1967), 14–32.DOI
  7. [7]K. Banerjee and N. A. Smoot, The localization method applied to k-elongated plane partitions and divisibility by 5, Math. Z. 309 (46), (2025).arxiv.org/abs/2208.07065
  8. [8]K. Banerjee and N. A. Smoot, 2-Elongated plane partitions and powers of 7: the localization method applied to a genus 1 congruence family, Canad. J. Math. (2025), 1–43.
  9. [9]M. Beazer, 3-adic properties of Hecke traces of singular moduli, Master thesis, Brigham Young University (2021), online available at http://hdl.lib.byu.edu/1877/etd11831.
  10. [10]M. Boylan and J. Webb, The partition function modulo prime powers, Trans. Amer. Math. Soc. 365 (2013), 2169–2206.DOI
  11. [11]K. Bringmann, P. Guerzhoy, B. Kane, M. Mertens, and L. Rolen, On Up-congruences for meromorphic modular forms via supersingularity, (2026), arXiv:2606.14020.
  12. [12]F. Colegari, Congruences between modular forms, Arizona winter school (2013), online available at https://www.math.uchicago.edu/~fcale/papers/AWS.pdf.
  13. [13]A. Folsom, Z. Kent, and K. Ono, ℓ-adic properties of the partition function (with an appendix by N. Ramsey), Adv. Math. 229 (2012), no. 3, 1586–1609.
  14. [14]O. Forster, Lectures on Riemann Surfaces, Springer-Verlag (1981).DOI
  15. [15]F. Garvan, J. Sellers, and N. A. Smoot, Old meets new: connecting two infinite families of congruences modulo powers of 5 for generalized Frobenius partition functions, Adv. Math. 454 (2024), p. 109866.DOI
  16. [16]B. Gordon and K. Hughes, Ramanujan Congruences for q(n), In: M. Knopp (eds) analytic number theory, lecture notes in mathematics 899, Springer, Berlin, Heidelberg (1981).DOI
  17. [17]P. Guerzhoy, Congruences for traces of singular moduli and Hurwitz–Kronecker class numbers, (2026), arXiv:2602.19544.arxiv.org/abs/2602.19544
  18. [18]K. Ono, Distribution of the partition function modulo m, Ann. of Math. (2) 151 (2000), 293–307.
  19. [19]P. Paule and S. Radu, The Andrews–Sellers family of partition congruences, Adv. Math. 230 (2012), 819–838.pubmed.ncbi.nlm.nih.gov/23471147
  20. [20]P. Paule and S. Radu, A Proof of the Weierstrass Gap Theorem Not Using the Riemann–Roch Formula, Ann. Comb. 23, 963–1007 (2019).
  21. [21]S. Radu, An algorithmic approach to Ramanujan–Kolberg identities, J. Symb. Comp. 68 (2015), 225–253.DOI
  22. [22]S. Ramanujan, Some properties of p(n), the number of partitions of n, Proc. Cambridge Phil. Soc. 19 (1919), 207–210.
  23. [23]N. A. Smoot, “A Congruence Family For 2-Elongated Plane Partitions: An Application of the Localization Method”, Journal of Number Theory 242, pp. 112–153 (2023).DOI
  24. [24]G. Watson, Ramanujans Vermutung über Zerfällungsanzahlen, J. Reine Angew. Math. 179 (1938), 97–128.DOI

Paper details

vlnc:W_pbj2s8r7
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 14

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents