Motivated by conjectures of Schlosser and Zhou on sign patterns of Fourier coefficients of infinite products, we study two families of theta quotients and their positive integral powers. We determine asymptotic formulas for their Fourier coefficients and use them to obtain explicit eventual periodic sign patterns. Our results provide new instances of the sign phenomena observed by Schlosser and Zhou and illustrate how dominant cusps determine the resulting sign patterns. These results were formalized and verified in Lean by AxiomProver.
Introduction and statement of results
In this paper, for m∈N≥3 and ℓ∈Z∖{0}, we study the sign pattern of the Fourier coefficients for the quotient
(q2,qm−2;qm)∞ℓ(q,qm−1;qm)∞ℓ.(1)
Here and throughout, ∣q∣<1 and for a∈C and n∈N0∪{∞}, we set (a)n=(a;q)n:=∏j=0n−1(1−aqj). For r∈N and a1,…,ar∈C, we let (a1,…,ar)n:=(a1,…,ar;q)n:=(a1)n⋯(ar)n. The primary motivation for considering this problem goes back to the sign pattern of Fourier coefficients of the Rogers–Ramanujan continued fraction
In this case, using the Circle method, the sign pattern of the n-th Fourier coefficient for n sufficiently large was proved by Richmond and Szekeres [16 (3.9)]. Andrews [2 Corollary 1] then gave a proof for the sign pattern of all coefficients via q-series dissections. Later, using dissections of Ramanujan’s theta functions, Ramanathan [15 Theorem 5 and Theorem 6] and Chan–Yesilyurt [7 Theorem 1.2 and Theorem 1.3] studied periodicity of the sign of Fourier coefficients for a family of infinite products resembling (1). Recently, using similar techniques, Baruah and Sarma [3 Theorem 2 and Theorem 3] investigated the sign pattern of the coefficients of (1) with m=5 and ℓ=±5. Using the Circle method, Schlosser and Zhou [17] studied the sign pattern of Fourier coefficients of theta quotients similar to (1) and made several conjectures. Three of the authors [4 Theorem 1.2] settled one of the conjectures of Schlosser and Zhou [17 Conjecture 22]. Other results in this direction can be found in [8 Section 2] and [11].
The paper is organized as follows. In § 3, we state an exact formula of Zuckerman, preliminary results on the Jacobi theta function, and Ingham’s Tauberian theorem. Proofs of Theorem 1.1 and Theorem 1.2 are given in § 5. We show Theorem 1.3 and Theorem 1.4 in § 6. § 7 is devoted to the proof of Theorem 1.5 and Theorem 1.6. Finally, we conclude this paper with § 8 by proposing questions for future research.
Acknowledgements
The first and the second author have received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 101001179). The research of the fourth author was supported by grants from the Research Grants Council of the Hong Kong SAR, China (project numbers HKU 17314122 and HKU 17305923). The authors are grateful to Ken Ono for using AxiomProver to formalize the results of this paper and to verify the proofs in Lean. These formal verifications were carried out by AxiomProver. The relevant files are available in the following repository: https://github.com/bkanemath/Sign-Patterns-of-Fourier-Coefficients-of-Theta-Quotients.
Preliminaries
Exact formulas of Rademacher and Zuckerman
Exact formulas of Rademacher and Zuckerman. We require an exact formula for Fourier coefficients of weakly holomorphic modular functions. Rademacher [13] perfected the Circle Method of Hardy and Ramanujan [10] and obtained an exact formula for the partition function. This was later extended to weakly holomorphic forms of nonpositive weight in results of Rademacher–Zuckerman [14] and Zuckerman [18] and by Ono and the second author [6]. To state the result, let Iα denote the usual I-Bessel function. We allow weakly holomorphic modular forms with a multiplier system. Throughout, for a∈N, b∈Z, with gcd(a,b)=1, we denote by [b]a the inverse of b(moda).
Theorem 2.1. Let κ≤0. Suppose that f is a weakly holomorphic modular form of weight κ on a subgroup Γ of finite index in SL2(Z). Assume that, for some δ≥0, f(τ)=∑n≫−∞a(n)qn+δ. For h,k∈N0 with 0≤h<k and gcd(h,k)=1, define fh,k by
(5) Forλ∈Q, q2λ2ϑ(λτ;τ)is a modular form of weight21(on some congruence subgroup).
Moreover, using [4 Lemma 2.1 (3)], Theorem 2.2 (3), (4), we have the following lemma.
Lemma 2.3. Suppose that τ∈H and a,b∈R.
(1) If0<a<1, then we have, asτ→i∞,
ϑ(aτ+b;τ)=−ie−πibq81−2a2(1+O(qmin{a,1−a})).
(2) Forb∈/Z, we have, asτ→i∞,
ϑ(b;τ)∼−2sin(πb)e4πiτ.
(3) If0<b<1, then we have, asz→0,
ϑ(aiz+b;iz)∼−ze−2πiab+πiae(−b2−41+b)zπ.
(4) Fora∈/Z, we have, asz→0,
ϑ(aiz;iz)∼−z2isin(πa)e−4zπ.
We also require the following asymptotics of the I-Bessel function.
Lemma 2.4.We have, asx→∞,
I1(x)∼2πxex.
A Tauberian theorem
We require a modified version of Ingham’s Tauberian Theorem [12], stated in [5 Theorem 1.1]. For Δ≥0, define
RΔ:={x+iy:x,y∈R,x>0,∣y∣≤Δx}.
Proposition 2.5. Let B(q)=∑n≥0b(n)qn be a power series whose radius of convergence is at least one and assume that the b(n) are non-negative and weakly increasing. Also suppose that ν,β,γ∈R with γ>0 exist such that
B(e−t)∼νtβetγas t→0+,B(e−z)≪∣z∣βe∣z∣γas z→0,(5)
with the latter condition holding in each region RΔ for Δ≥0. Then we have
b(n)∼2πn2β+43νγ2β+41e2γnas n→∞.
Moreover, we require the asymptotic behavior of (q)∞ as q→1.
Lemma 2.6.We have, withq=e−z,
(q)∞∼z2πe−6zπ2as z→0.
Proof of Theorem 1.1 and Theorem 1.2
In this section, we consider the case that m is odd. For m∈N>3 and ℓ∈N, we study the behavior of Gmℓ(τ), where
Gm(τ):=ϑ(2τ;mτ)ϑ(τ;mτ)=q21Fm(q),
as τ→kh∈Q. For a∈R, we write a=⌊a⌋+{a}. We have the following lemma (for general m). For m∈N>3, ℓ,k∈N, d:=gcd(m,k) with μ:=dm, and κ:=dk, define
where, here and throughout, we use ≡ to denote that the identity holds modulo 2πiZ. Now by Theorem 2.2 (1) with w=mkj+1(h+iz) (j∈{0,1}), we have, as z→0,
Let d∣m. If d=1, then μ=m, κ=k, and {mμh}={h}=0. So, by Theorem 3.1, Gm(k1(h+iz))=Oh,k(1), as z→0. We next assume d>1. Plugging in μ=dm and κ=dk, by (11), we need to maximize (note that d is odd since m is odd)
Observe that for d=3 these expressions vanish. So for the remainder of the proof, we assume d>3. For m and d fixed the expressions in (3.7) are maximized for j=2d−1. The corresponding contribution is 4mk(d−1)(d−3). Moreover, one sees that this is maximal for k minimal with gcd(m,k)=d, which occurs for k=d. The expression we need to maximize in this case is 4md(d−1)(d−3). As d∣m, this is maximized for d=m. The corresponding values of h are 2m−1 and 2m+1 which are clearly coprime to k=m. □
Lemma 3.3. For m∈N>3 odd, ℓ∈N, and h∈{2m−1,2m+1}, we have, as z→0,
Moreover, by definition, c_m^{[1]}(n)\in\mathbb{R}. So, Fm(q)=Fm(q), and thus we only need to prove the claim for 2m−1. Theorem 2.3 (3) finishes the proof. □
Proof of Theorem 1.1. By Theorem 2.2 (5), q−2m3ℓ+2ℓFmℓ(q)=q−2m3ℓGmℓ(τ) is a modular function. By Theorem 3.2 its maximal growth occurs at the cusps 2mm−1 and 2mm+1. Using Theorem 2.4, Theorem 3.3, and (1.3), Theorem 2.1 now gives, as n→∞,
In this section, we consider the case that m is even. We first determine the asymptotics of Fmℓ(−e−z) as z→0. Note that, by the fact m even and (1.2), we have
Lemma 4.1. For n∈N0, dm[ℓ](n)≥0 and {dm[ℓ](n)}n≥0 is weakly increasing.
Proof. From (4.1), it directly follows that dm[ℓ](n)≥0 for n∈N0. To prove that {dm[ℓ](n)}n≥0 is weakly increasing, it suffices to show that (1−q)Fmℓ(−q) has non-negative coefficients. By (4.1), we have
Using this with τ=2πiz, Theorem 2.6 and Theorem 2.3 (4) imply the claim.
Proof of Theorem 1.3. By Theorem 4.1, dm[ℓ](n) satisfies the monotonicity and non-negativity required to apply Theorem 2.5. Moreover, Theorem 4.2 shows that Fmℓ(−e−z) satisfies (2.1) with ν=2ℓsinℓ(m2π)1, β=0, and γ=2mπ2ℓ. By Theorem 2.5, we have, as n→∞,
dm[ℓ](n)∼2ℓ+45m41sinℓ(m2π)ℓ41n43eπm2ℓn.
By (4.2), we conclude the proof.
Proof of Theorem 1.4. The claim follows from Theorem 4.1 and (4.2).
Proof of Theorem 1.5 and Theorem 1.6
For m∈N>3 and λ∈N, we study the main growth of Gmλ(τ), where
Gm(τ):=ϑ(τ;mτ)ϑ(2τ;mτ)=q−21Fm(q),
as τ→kh∈Q. Noting that Gm=Gm−1, Theorem 3.1 implies the following:
Lemma 5.1. Let m∈N>3 and λ,k∈N. For 0≤h<k with gcd(h,k)=1, d:=gcd(m,k) with μ:=dm and κ:=dk, we have,
Gmλ(kh+kiz)=e−zAm,λ(h,k)+Oh,k(1)(as z→0).
Next, we maximize Gmλ(k1(h+iz)).
Lemma 5.2. For m∈N>3 with m≡±1(mod6) and λ∈N, Gmλ(k1(h+iz)) is maximal, as z→0, for kh∈{6mm∓1,6m5m±1}.
Proof. By taking the negative of (3.6) we see that, as z→0, Gmλ(k1(h+iz)) is maximal for the maximal value of
Let d∣m. If d=1, then μ=m, κ=k, and {mμh}={h}=0. So, by Theorem 5.1, Gmλ(k1(h+iz))=Oh,k(1) as z→0. Next, we assume d>1. Plugging in μ=dm and κ=dk, by (5.1), making the change of variables j↦d−j in the second case, thus we need to maximize
We note that for m and d fixed these expressions are maximized if ∣6j−d∣ is minimal. Now ∣6j−d∣ is minimal for j=6d∓1. In this case, (5.2) is maximal for k=d. The expression we need to maximize in this case is 12md−1. This is maximized for d=m. The maximum value is 121(1−m21), which is positive for m∈N>3. Since j=6m∓1, we have m=6j±1 and hence
gcd(m,j)=gcd(6j±1,j)=1.
Thus, for h∈mN0+j (for the plus sign) and k=m (resp. h∈mN−j and k=m for the minus sign), we have gcd(h,k)=1. Taking the representative 0≤h<m gives h=j=6m∓1 and h=m−j=65m±1. □
Next, we determine the growth of Fmλ(em2πi(h+iz)) as z→0.
Lemma 5.3. For m∈N>3 with m≡±1(mod6), λ∈N, and h∈{6m∓1,65m±1}, we have,
Proof. Again it is enough to consider the case λ=1. Moreover, by definition, γm[1](n)∈R. So, Fm(q)=Fm(q) and thus we only need to prove the claim for h=6m∓1. Using Theorem 2.2 (3), for h=6m∓1, we have, as z→0,
Proof ofTheorem 1.5. By Theorem 2.2 (5), q2m3λ−2λFmλ(q) is a modular function. By Theorem 5.2 its maximal growth occurs at the cusps 6mm∓1 and 6m5m±1. Using Theorem 5.3 and Theorem 2.4, Theorem 2.1 now gives, as n→∞,
This is the complement of [3m+4+6λ−5,34m+1+6λ−5] in [0,2m−1] and the claim follows in this case. □
Open questions
The results of this paper suggest several directions for future research.
(1) It would be interesting to extend Theorems 1.1, 1.3, and 1.5, by considering theta quotients of the form (for a,b,m,ℓ∈N with a=b)
ϑ(bτ;mτ)ℓϑ(aτ;mτ)ℓ.
For which choices of a, b, m, and ℓ do the Fourier coefficients have an eventual periodic sign pattern? Can this sign pattern be determined explicitly? More generally, can one characterize the dominant cusps in terms of a, b, and m?
(2) In Theorem 1.5, we assume gcd(m,6)=1. Can this assumption be removed? More precisely, how do the dominant cusps and the resulting sign pattern change if 2∣m or 3∣m?
(3) Our results raise a more general question: under what conditions do the Fourier coefficients of a modular theta quotient have an eventually periodic sign pattern?
Declarations
Data Availability. Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
Conflicts of interest. The authors declare that there are no conflicts of interest.
Funding. Funding declarations can be found in the Acknowledgements Section.
Numerics
In the following tables, we list the sign patterns of cm[ℓ](n) and γm[ℓ](n) for m∈{5,7}, 1≤ℓ≤6, and 70≤n≤m+69.
(m,ℓ)
(5,1)
(5,2)
(5,3)
(5,4)
(5,5)
sgn(cm[ℓ](n))
+−+−−
+−+−−
+−+−+
+−+−+
−−+−+
sgn(γm[ℓ](n))
++−−−
+++−−
+++−−
−++−−
−+++−
Table 1.
(m,ℓ)
(5,6)
(7,1)
(7,2)
(7,3)
sgn(cm[ℓ](n))
−++−+
+−+−++
+−+−+−+
−−+−++−
sgn(γm[ℓ](n))
−−++−
+++−−−+
+++−−−−
++++−−−
Table 1.
(m,ℓ)
(7,4)
(7,5)
(7,6)
sgn(cm[ℓ](n))
−+−−+−+
−+−++−+
−+−+−++
sgn(γm[ℓ](n))
−++++−−
−−+++−−
−−++++−
Table 1. Sign patterns for cm[ℓ](n) and γm[ℓ](n).
University of Cologne, Department of Mathematics and Computer Science, Weyertal 86-90, 50931 Cologne, Germany
[1]G. Andrews, The theory of partitions, Cambridge University Press, 1980.
[2]G. Andrews, Ramanujan’s “Lost” Notebook III. The Rogers–Ramanujan continued fraction, Adv. Math. 41 (1981), 186–208.DOI
[3]N. Baruah and A. Sarma, Sign patterns and congruences of certain infinite products involving the Rogers–Ramanujan continued fraction, Ramanujan J. 67 (2025), 1–25.arxiv.org/abs/2503.08517
[4]K. Bringmann, B. Heim, and B. Kane, On a sign-change conjecture of Schlosser and Zhou, J. Math. Anal. Appl. 548 (2025), p. 129339.arxiv.org/abs/2406.03453
[5]K. Bringmann, C. Jennings-Shaffer, and K. Mahlburg, On a Tauberian theorem of Ingham and Euler–Maclaurin summation, Ramanujan J. 61 (2021), 55–86.
[6]K. Bringmann and K. Ono, Coefficients of harmonic Maass forms, Proceedings of the 2008 University of Florida Conference on partitions, q-series, and modular forms, 2008.
[7]S. Chan and H. Yesilyurt, The periodicity of the signs of the coefficients of certain infinite product, Pacific J. Math. 225 (2006), 13–32.DOI
[8]S. Chern, Asymptotics for the Taylor coefficients of certain infinite products, Ramanujan J. 55 (2021), 987–1014.arxiv.org/abs/1902.10839
[9]M. Eichler and D. Zagier, The theory of Jacobi forms, Progress in Math. 55, Birkhäuser, 1985.DOI
[10]G. Hardy and S. Ramanujan, Asymptotic formulae in combinatory analysis, Proc. London Math. Soc. (2) 17 (1918), 75–115.
[11]B. He and L. Li, Some conjectures of Schlosser and Zhou on sign patterns of the coefficients of infinite products, arXiv:2509.10023.
[12]A. Ingham, A Tauberian theorem for partitions, Ann. Math. 42 (1941), 1075–1090.
[13]H. Rademacher, On the partition function p(n), Proc. Lond. Math. Soc. (3) 43 (1937), 241–254.DOI
[14]H. Rademacher and H. Zuckerman, On the Fourier coefficients of certain modular forms of positive dimension Ann. Math. 39 (1938), 433–462.DOI
[15]K. Ramanathan, Generalisations of some theorems of Ramanujan, J. Number Theory 29 (1988), 118–137.DOI
[16]B. Richmond and G. Szekeres, The Taylor coefficients of certain infinite products, Acta Sci. Math. (Szeged) 40 (1978), 347–369.
[17]M. Schlosser and N. Zhou, On the infinite Borwein product raised to a positive real power, Ramanujan J. 61 (2023), 515–543.
[18]H. Zuckerman, On the coefficients of certain modular forms belonging to subgroups of the modular group, Trans. Amer. Math. Soc. 45 (1939), 298–321.DOI