Abstract
String functions are important building blocks of characters of integrable highest modules over affine Kac–Moody algebras. Kac and Peterson computed string functions for affine Lie algebras of type \(A_{1}^{(1)}\) in terms of Dedekind eta functions. We obtain new symmetries for string functions by exploiting their natural setting of Hecke-type double-sums, where special double-sums are expressed in terms of Appell–Lerch functions and theta functions, where we point out that Appell–Lerch functions are the building blocks of Ramanujan’s classical mock theta functions. We then demonstrate the utility of the new symmetries by giving new proofs of classical string function identities.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let q be a complex number where \(q:=e^{2\pi i \tau }\) and \(\tau \in \mathfrak {H}:=\{z\in \mathbb {C} | \text{ Im }{z}>0\}\). The goal of this paper is to find new symmetries for string functions for affine Kac–Moody algebras and then use the new symmetries to give new proofs of classical string function identities. We do so by putting the string functions into the setting of Hecke-type double-sums as studied by the first author and Hickerson [15]. Let us start with a brief reminder of what a string function is in this case. For details, see [17, 18].
For the affine Kac–Moody algebra \(\mathfrak {g}=A_1^{(1)}\) the simple roots are \(\alpha _0=\delta -\alpha \) and \(\alpha _1=\alpha \), where \(\alpha \) is the root of \(A_1\) Lie subalgebra and \(\delta \) is the imaginary root. Denote by \(\Lambda _0\) and \(\Lambda _1\) corresponding fundamental weights. Denote by \(P_+\) the subset of integral dominant weights in the weight lattice \(\mathbb {Z}\Lambda _0\bigoplus \mathbb {Z}\Lambda _1,\;\; P_+=\{a\Lambda _0+b\Lambda _1, a,b\in \mathbb {Z}_{\ge 0}\}\).
Let \(L(\Lambda ), \Lambda \in P_+\) be an irreducible \(\mathfrak {g}\)-module with highest weight \(\Lambda \in P_+\) of level N:
Define the character of \(L(\Lambda )\) as a function
Here a sum is taken over all weights of a module; these are weights that occur in \(L(\lambda )\), i.e., \(\lambda =\Lambda -A\alpha _0-B\alpha _1\), where \(A,B\in \mathbb {Z}_{\ge 0}\).
It is possible to express the character via the subset of weights of a module called maximal weights. A weight \(\lambda \in P(\Lambda )\) is called maximal if \(\lambda +\delta \not \in P(\Lambda )\). We will parametrize maximal weights of the module as
For a maximal weight \(\lambda \in P(\Lambda )\) the string function of \(\lambda \) is defined as
where
The character of \(L(\Lambda )\) can be expressed in terms of theta functions and string functions as (See ([17, 18, 25] for details and for how to extend it to other Kac–Moody algebras)
where \(\Theta _{n,m}(z,q)\) is a Jacobi theta function of degree m and characteristic n
The \(C_{m,\ell }^N(q)\) are level-N A\(_1^{(1)}\) string functions, see (1.1).
We recall the definition of Dedkind’s eta-function:
The character can be computed as (see [25, 26] and references therein)
which gives the expression for string functions as
This is our starting point.
2 Main results
To state our main results, we define Hecke-type double-sum as follows. Let \(x,y\in \mathbb {C}\backslash \{ 0 \}\), and set
then
We also recall notation for infinite products and theta functions. We have
where in the last line the equivalence of product and sum follows from Jacobi’s triple product identity. We note that \(j(q^n;q)=0\) for \(n\in \mathbb {Z}.\) Let a and m be integers with m positive. Define
Before we state our main results, we recall the classic symmetries for string functions. They are
We also remind the reader of our definition for \(s_{\Lambda }(\lambda )\) (1.2). Our new symmetries read
Theorem 2.1
We have
Corollary 2.2
We have
Corollary 2.3
For \(K\equiv \ell \pmod 2\), we have
Theorem 2.1 and its two corollaries provide an alternative way for interpreting and proving the string function expressions for affine Lie algebras of type \(A_{1}^{(1)}\) found in Kac and Peterson [18]. In [18, pp. 219–220], Kac and Peterson give several examples of string functions for affine Lie algebras of type \(A_{1}^{(1)}\) that have beautiful evaluations in terms of theta functions. See also [19, 20]. If we fix a positive integer m, their string functions are of the form [18, p. 260]:
where N and n are integers with \(n\equiv N \pmod 2\). The string functions \(c_{\lambda }^{\Lambda }(\tau )\) are closely related to the real quadratic fields \(\mathbb {Q}(\sqrt{m(m+2)})\) and to Hecke indefinite modular forms. We have replaced Kac and Peterson’s notation (n, N, m) with \((m,\ell ,N)\) of [25]. Here our focus will be on double-sum evaluations. A partial list of string functions from [18, pp. 219–220] reads
Level 1:
Level 2:
Level 4:
Level 6:
Level 8:
Level 10:
In this paper we rewrite the double-sum form of string functions (2.5) using an expression found in [25], where from [25] and [18, p. 260] we have that
Kac and Peterson appeal to modularity to prove the string function identities [18, p. 220]. Specifically, they use the transformation law for string functions under the full modular group, together with the calculation of the first few terms in the Fourier expansions of the string functions. They also take advantage of the fact that a modular form vanishing at cusps to sufficiently high order is zero. There are many methods to compute string functions, see [18, pp. 222–223] for a brief outline. In our present work, we will take a different approach.
Once string function identities (2.6a)–(2.11c) have been written in terms of suitable Hecke-type double-sums, see (2.15)–(2.22c), their modularity can be determined using results of [15], see in particular the formulas found in Sect. 3.3.
We recall a simple example of more general results found in [15]. For generic \(x,z\in \mathbb {C} \backslash \{ 0\}\) we define our Appell–Lerch function as
One of the simplest results from [15] then reads
Such formulas were found by using a heuristic that relates Appell–Lerch functions to divergent partial theta functions, see [15, Section 3] and [23, Section 4].
For an easy demonstration of (2.14), we consider the case level \(N=1\). It is straightforward to see from (2.5) that (2.6a) is equivalent to showing:
One sees from (2.14) that
where we used the fact that \(j(x;q)=0\) if and only if x is an integral power of q and elementary product rearrangements. On the other hand, if we make a slight variation in the inputs, we have
where \(\phi (q)\) is a sixth-order mock theta function [4].
In general, one needs to be aware of the number of theta quotients that must be dealt with in simplifying the results. Double-sums of the form \(f_{1,n+1,1}\) require computing \(n^2\) theta quotients, see in particular Theorem 3.8. For example, one sees from (2.5) that (2.11b) is equivalent to showing
which requires working with one-hundred theta quotients!
Looking for ways to reduce the number of theta quotients in Theorem 3.8 or for better ways to write Theorem 3.8 is an important question. Many of Ramanujan’s mock theta function identities involve a single quotient of theta functions. One wants to somehow understand how he was led to these single quotient identities. As an example, we can revisit the four tenth-order mock theta functions [8, 24]. Three of them read
and they can all be expressed in terms of Hecke-type double-sums [6, 7, 15]. The four functions also satisfy six single-quotient theta function identities [6,7,8, 24]. For one of them, we have [8, 24]
The six identities were originally found in Ramanjuan’s lost notebook [24]. What led Ramanujan to these identities is a mystery. In fact, in Andrews and Berndt’s final volume on Ramanujan’s lost notebook [3, p. 396], they state
“It is inconceivable that an identity such as (2.17) could be stumbled upon by a mindless search algorithm without any overarching theoretical insight.”
A proof of (2.17) using Theorem 3.8 is involved [22], so one would like to know if there is a better form of Theorem 3.8 which sheds light on what guided Ramanujan to so many single-quotient that function identities.
Double-sums of the form \(f_{K+1,K+1,1}\) that appear in Theorem 2.1 only have \(K+1\) theta quotients that need to be summed. Using our main theorem and its corollaries, string function identities (2.7a)–(2.11c) are equivalent to the respective double-sum evaluations:
Level 2:
Level 4:
Level 6:
Level 8:
Level 10:
In particular our example in (2.16) becomes (2.22b), which only requires working with six theta quotients, see Theorem 3.10.
In [15] we demonstrated that the double-sum formulas give straightforward proofs of the classical mock theta function identities, and in particular the formulas give new proofs of the mock theta conjectures [13, 14]. The focus here is to demonstrate the robustness of our approach in application to string functions. We evaluate eight string functions using double-sum formulas. Four double-sum computations are pieced together from previous results, and four double-sum computations are new.
In Sect. 3, we recall background information on theta functions, Appell–Lerch functions, and Hecke-type double-sums. In Sect. 4, we discuss classical string function relations in the environment of Hecke-type double-sums. In Sect. 5, we obtain (2.12), where the corrected proof comes from a sketch found in [15, Example 1.3]. In Sect. 6, we prove our main result: Theorem 2.1 and its two corollaries, which one could consider to be new string function relations. In the remaining sections, we use Theorem 2.1 to prove even-level string function identities.
In Sect. 7, we evaluate the level \(N=2\) string function \(c_{11}^{11}\) (2.7a). In Sect. 8, we evaluate the level \(N=4\) string function \(c_{40}^{40}-c_{04}^{40}\) (2.8b), where the proof is essentially from [23]. In Sect. 9, we evaluate the level \(N=6\) string functions \(c_{51}^{33}\) (2.9a) and \(c_{51}^{51}+c_{15}^{51}\) (2.9b). In Sect. 9, we also evaluate the level \(N=6\) string function \(c_{51}^{51}-c_{15}^{51}\) (2.9c), where the proof is essentially from [23]. In Sect. 10, we evaluate the level \(N=8\) string function \(c_{62}^{44}\) (2.10a), where the proof is essentially from [15]. In Sect. 11, we evaluate the level \(N=10\) string function \(c_{91}^{55}\) (2.11b).
For a brief historical perspective, around same time as [18], Lepowsky and Primc gave q-hypergeometric expressions for the \(sl_2\) string functions [21]. There have been many types of proofs of identities (2.18a)–(2.22c) over the years since they were first announced. For example, in [1, 2] Andrews proves many of the identities using the constant term method and Bailey pair method, respectively. There have been string functions which generalize \(C_{m,\ell }^{N}(q)\) [12, 16]. There have also been numerous further developments concerning generalized modularity of string functions such as a recent result by Dong, Kac, Ren [9] using algebraic methods. Placing string functions into the double-sum setting [15, Example 1.3] has recently led to Dousse and Osburn’s resolution [10] of a problem of Byrne et al [5].
3 Preliminaries
3.1 Theta functions
We collect some frequently encountered product rearrangements:
Following from the definitions are the following general identities:
if \(n\ge 1\), \(\zeta _n\) is a primitive n-th root of unity.
A convenient form of the Weierstrass three-term relation for theta functions is,
Proposition 3.1
For generic \(a,b,c,d\in \mathbb {C} \backslash \{0 \}\)
We collect several useful results about theta functions in terms of a proposition [13, 14]:
Proposition 3.2
For generic \(x,y,z\in \mathbb {C} \backslash \{0\}\)
We finish the subsection by giving a few examples of elementary theta function identities and product rearrangements in action.
Lemma 3.3
We have
and
Proof of Lemma 3.3
For the first summand in (3.4), we apply (3.2d) to \(J_{3,6}\) to obtain
For the second summand, we begin by applying (3.2f) to \(J_{6,60}\) to have
where we have used the product rearrangement \({\overline{J}}_{0,1}=2J_2^2/J_1\) and simplified. Applying (3.2f) to the product \(J_{5,30}{\overline{J}}_{5,30}\) yields
where for the second equality we used the product rearrangements for \({\overline{J}}_{1,3}\) and \(J_{1,6}\), and for the third equality we simplified.
For second identity (3.5), we use the product rearrangements for \({\overline{J}}_{0,1}\), \({\overline{J}}_{1,6}\), and \({\overline{J}}_{1,3}\) and simplify to have
where for the last equality we used (3.2f). Identity (3.2d) gives us
Hence we can rewrite the right-hand side as
where we have used the product rearrangement \(J_{1,5}J_{2,5}=J_{1}J_{5}\) and (3.2c). \(\square \)
3.2 Appell–Lerch functions
The Appell–Lerch function satisfies several functional equations and identities [15, 27]:
Proposition 3.4
For generic \(x,z\in \mathbb {C} \backslash \{0\}\)
Some simple evaluations of the Appell–Lerch function follow.
Corollary 3.5
We have
3.3 Hecke-type double-sums
We recall a few basic properties of Hecke-type double-sums. We have a proposition and a corollary
Proposition 3.6
[15, Proposition 6.3] For \(x,y\in \mathbb {C}\backslash \{0\}\) and \(R,S\in \mathbb {Z}\)
When \(b<a\) we adopt the summation convention that
which has the useful consequence
Usually when (3.9) is used in this paper, the two summands of theta functions in the second row are always equal to zero. In such cases, it is due to the fact that \(j(x;q)=0\) if and only if x is an integral power of q. The only exception occurs in the last subsection of Sect. 9.
Corollary 3.7
[15, Corollary 6.4] We have two simple specializations:
We also have the property [15, (6.2)]:
One can determine the theta function expressions of string functions using results of [15]. However, there are two recurring issues that must be kept in mind. The first issue is the number of theta quotients that must be computed, and the second issue is potential singularities.
To state our results, we introduce the useful
In [15, Theorem 1.3], we specialize \(n=1\), to have
Theorem 3.8
Let p be a positive integer. For generic \(x,y\in \mathbb {C} \backslash \{0\}\)
where
The specialization for \(p=1\) will be of importance. It is just (2.14):
Corollary 3.9
We have
For another useful result, we specialize [15, Theorem 1.4] to \(a=b=n\), \(c=1\).
Theorem 3.10
Let n be a positive integer. Then
where
and
Theorem 3.10 has the following specializations.
Corollary 3.11
We have
where
Corollary 3.12
We have
where
Corollary 3.13
We have
where
Corollary 3.14
We have
where
Corollary 3.15
We have
where
We finish with a few example computations that can be found in the literature.
Lemma 3.16
We have
Proof of Lemma 3.16
All three identities have been shown using the appropriate specializations of Theorem 3.10. The first can be found in [15, Section 7] and the last two in [23, Lemma 3.11]. \(\square \)
4 Classical string function relations and Hecke-type double-sums
In this section we use the environment of Hecke-type double-sums and their functional equations such as (3.9) and (3.14) to show that
satisfies the same classical symmetries for string functions. We recall the classical identities [25, (3.4), (3.5)]:
Proof for Identity (4.2):
We have
where symmetry allows us to write
\(\square \)
Proof for Identity (4.3):
Specializing (3.9) with \((R,S)=(1,-1)\) gives
As a result,
\(\square \)
Proof for Identity (4.4):
Specializing (3.9) with \((R,S)=(1,0)\) gives
where for the second equality we used (3.14). Hence
\(\square \)
5 Deriving the general integral-level N string function
We derive (2.12). We recall the notation from [25] that \(m, \ N\in \mathbb {N}\), \(\ell \in \{ 0,1,2,\dots , N\}\), \(m\equiv \ell \pmod 2\), where N is the level. In [25, p. 236], see also [11, (3.17)], one finds the Hecke-type form for the general integral-level string function:
Proposition 5.1
We have
where \(S_{m,\ell }^{N}(q)\) is given by (4.1).
We begin with an identity from [25]:
where from (1.2):
We rewrite the above double-sum:
where we have used \(k\rightarrow -k\), \(j\rightarrow j+1\). We recall identity (3.12)
and the fact that \(j(z;q)=0\) if and only if \(z=q^{n}\), \(n\in \mathbb {Z}\). This gives
for the second equality the condition \(m\equiv \ell \pmod 2\) forces that
For the third equality, the x and y in \(f_{1,N+1,1}(x,y,q)\) can be swapped because of symmetry in the definition of (2.1).
6 Proof of the main theorem and its two corollaries
The driving force of our proof is the following proposition.
Proposition 6.1
We have
Proof of Proposition 6.1
We have
For the second identity one replaces \((-1)^{(u-v)/2}\) with \((-1)^{(u+v)/2}\). This also results in the \((-1)^j\) becoming \((-1)^n\). \(\square \)
Proof of Theorem 2.1
We prove the identity for the initial plus sign. The proof for the minus sign is analogous. For (2.2), we have
where the first equality follows (3.9) with \((R,S)=(1,0)\) and the third from (6.1). \(\square \)
Proof of Corollaries 2.2 and 2.3
To prove (2.3), we let \(\ell =K\) in (2.2). To prove (2.4), we set \(m=K\) in (2.2) to have
where the second equality follows from (3.9) with \((R,S)=(-1,1)\), the third equality follows from (3.14), and the fourth follows from (3.9) with \((R,S)=(0,-1)\). \(\square \)
7 Computing-level \(N=2\) string functions
7.1 The string function \(c_{11}^{11}\) (2.7a)
It suffices to show (2.18a):
In Corollary 3.11, we find potential singularities, so we consider
We first focus on (3.18) and write
where we have used (3.6e) in the penultimate equality. Continuing with product rearrangements, we have
where in the second line we used (3.2b) and (3.2c). Now we consider
where in the second equality we used (3.2b) and (3.2c), and in the third equality we simplified. Assembling the pieces, we have
9 Computing-level \(N=6\) string functions
9.1 The string function \(c_{51}^{33}\) (2.9a)
It suffices to show (2.20a):
In Corollary 3.13 we immediately see that
Hence, it follows from Corollary 3.13 that
where we have used (3.2b), collected terms, and then used (3.3a).
9.2 The string function \(c_{51}^{51}+c_{15}^{51}\) (2.9b)
It suffices to show (2.20b):
We use Corollary 3.13. Considering (3.22), we see that two of the four summands vanish right away giving us
where we have used (3.2b) and (3.2c). Thus we have
where we have used (3.2b), (3.2c), and simplified. Regrouping terms, using (3.2c), and simplifying, we have
where we have used (3.2d). Employing (3.3a), (3.3b), and (3.2a), yields
where for the second equality we used (3.2f), and for the third equality we used (3.2a). Applying (3.2e) gives
where the second equality follows from elementary product rearrangements.
9.3 The string function \(c_{51}^{51}-c_{15}^{51}\) (2.9c)
It suffices to show (2.20c):
From Lemma 3.16, we have
where for the second equality we used (3.9) with \((R,S)=(0,1)\), and for the third we used (3.2b) and (3.14).
11 Computing-level \(N=10\) string functions
11.1 The string function \(c_{91}^{55}\) (2.11b)
It suffices to show (2.22b):
We use Corollary 3.15. In (3.26), we have
Hence
where we have used (3.2b). Using (3.2c) and regrouping terms yields
where we have used (3.3a). We focus on the last two summands:
If we specialize Proposition 3.1 with \(a=q^{15}\), \(b=q^{10}\), \(c=q^{6}\), \(d=q^{5}\), we have
Using (3.2c) we can then write
The result follows from the two identities in Lemma 3.3.
12 Concluding remarks and a question for further research
We have derived the expressions for even-level string functions for an irreducible highest weight module of Kac–Moody Lie algebra \(A_1^{(1)}\) in terms of double-sums of the form \(f_{K+1,K+1,1}(x,y,q)\). These double-sums may be expressed in terms of theta functions and Appell–Lerch functions: the building blocks of Ramanujan’s mock theta functions.
There are also questions for future research. For example, can one use a variation of the techniques in this paper in order to reduce the number of theta quotients that appear in the initial expressions for the odd-level \(A_{1}^{(1)}\) string functions?
Data availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
References
Andrews, G.E.: Hecke modular forms and Kac–Peterson identities. Trans. Am. Math. Soc. 283(2), 451–458 (1984)
Andrews, G.E.: The fifth and seventh order mock theta functions. Trans. Am. Math. Soc. 493(1), 113–134 (1986)
Andrews, G.E., Berndt, B.C.: Ramanujan’s Lost Notebook. Part V. Springer, Berlin (2018)
Andrews, G.E., Hickerson, D.R.: Ramanujan’s “lost’’ notebook. VII: the sixth order mock theta functions. Adv. Math. 89(1), 60–105 (1991)
Byrne, E., Horlemann, A.-L., Khathuria, K., Weger, V.: Density of free modules over finite chain rings. Linear Algebra Appl. 651, 1–15 (2002)
Choi, Y.-S.: Tenth order mock theta functions in Ramanujan’s lost notebook. Invent. Math. 136(3), 497–569 (1999)
Choi, Y.-S.: Tenth order mock theta functions in Ramanujan’s lost notebook II. Adv. Math. 156(2), 180–285 (2000)
Choi, Y.-S.: Tenth order mock theta functions in Ramanujan’s lost notebook III. Proc. Lond. Math. Soc. 3(94), 26–52 (2007)
Dong, C., Kac, V., Ren, L.: Trace functions of the parafermion vertex operator algebras. Adv. Math. 348, 1–17 (2019)
Dousse, J., Osburn, R.: A \(q\)-multisum identity arising from finite chain ring probabilities. Electron. J. Combin. 29(2), 7 (2022)
Distler, J., Qiu, Z.: BRS cohomology and a Feigin–Fuchs representation of Kac–Moody and parafermionic theories. Nucl. Phys. B 336, 533–546 (1990)
Hatayama, G., Kirillov, A.N., Kuniba, A., Okado, M., Takagi, T., Yamada, Y.: Character formulae of \(\widehat{sl_n}\)-modules and inhomogeneous paths. Nuclear Phys. B 536(3), 575–616 (1999)
Hickerson, D.R.: A proof of the mock theta conjectures. Invent. Math. 94(3), 639–660 (1988)
Hickerson, D.R.: On the seventh order mock theta functions. Invent. Math. 94(3), 661–677 (1988)
Hickerson, D.R., Mortenson, E.T.: Hecke-type double sums, Appell–Lerch sums, and mock theta functions, I. Proc. Lond. Math. Soc. (3) 109(2), 382–422 (2014)
Kuniba, A., Nakanishi, T., Suzuki, J.: Characters in conformal field theories from thermodynamic Bethe ansatz. Mod. Phys. Lett. A 8(18), 1649–1659 (1993)
Kac, V.: Infinite-Dimensional Lie Algebras. Cambridge University Press, Cambridge (1990)
Kac, V., Peterson, D.: Infinite-dimensional Lie algebras, theta functions and modular forms. Adv. Math. 53, 125–264 (1984)
Kac, V., Wakimoto, M.: Modular invariant representations of infinite-dimensional Lie algebras and superalgebras. Proc. Nat. Acad. Aci. USA 85(14), 4956–4960 (1988)
Kac, V., Wakimoto, M.: Classification of modular invariant representations of affine algebras. Adv. Ser. Math. Phys. 7, 138–177 (1989)
Lepowsky, J., Primc, M.: Structure of the Standard Modules for the Affine Lie Algebra \(A_{1}^{(1)}\). Contemporary Mathematics, vol. 46. AMS, Providence (1985)
Mortenson, E.T.: On the tenth-order mock theta functions. J. Aust. Math. Soc. 104(1), 44–62 (2018)
Mortenson, E.T.: A heuristic guide to evaluating triple-sums. Hardy-Ramanujan J. 43, 99–121 (2021)
Ramanujan, S.: The Lost Notebook and Other Unpublished Papers. Narosa Publishing House, New Delhi (1988)
Schilling, A., Warnaar, S.O.: Conjugate Bailey Pairs. Contemp. Math. 197, 227–255 (2002)
Warnaar, S.O.: 50 years of Bailey’s lemma. In: Betten, A., et al. (eds.) Algebraic Combinatorics and Applications, pp. 333–347. Springer, Berlin (2001)
Zwegers, S.P.: Mock theta functions. Ph.D. Thesis, Universiteit Utrecht (2002)
Acknowledgements
This research was supported by the Theoretical Physics and Mathematics Advancement Foundation BASIS, Agreement No. 20-7-1-25-1. We would also like to thank O. Warnaar and E. Feigin for helpful comments and suggestions. The authors state that there are no conflicts of interest.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Mortenson, E.T., Postnova, O. & Solovyev, D. On string functions and double-sum formulas. Res Math Sci 10, 15 (2023). https://doi.org/10.1007/s40687-023-00379-x
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s40687-023-00379-x