 Research
 Open Access
 Published:
Note on the problem of Ramanujan’s radial limits
Advances in Difference Equations volume 2014, Article number: 191 (2014)
Abstract
Ramanujan in his deathbed letter to GH Hardy concerned the asymptotic properties of modular forms and mock theta functions. For the mock theta function $f(q)$, he claimed that as q approaches an even order 2k root of unity ζ,
where $(1q)(1{q}^{3})(1{q}^{5})\cdots (12q+2{q}^{4}\cdots )={\prod}_{n=1}^{\mathrm{\infty}}\frac{1{q}^{n}}{{(1+{q}^{n})}^{2}}$. Recently, Folsom, Ono and Rhoades have proved two closed formulas for the implied constant and formulated an open problem which is related to their two theorems. In this note, we give a new proof on the problem of the two theorems by using some results about the generating functions of convex compositions given by GE Andrews and AppellLerch sums.
MSC:11F37, 11F03, 11F99.
1 Introduction
In his deathbed letter to Hardy, Ramanujan gave no definition of mock theta functions but just listed 17 examples and a qualitative description of the key properties that he had noticed. Since that time, many papers studying the 17 specific examples have been written by many famous mathematicians such as Watson, Selberg and Andrews [1]. Due to the work of Zweger [2, 3], Bringmann and Ono [4, 5], Zagier [6] and others, we are able to recognize Ramanujan’s mock theta functions as holomorphic parts of certain harmonic weak Maass forms of weight $1/2$, originally defined by Bruinier and Funke [7]. This realization has resulted in many applications in combinatorics, number theory, physics and representation theory.
While the theory of the weak Maass forms has led to a flood of applications in many disparate areas of mathematics, it is still not the case that we fully understand the deeper framework surrounding the contents of Ramanujan’s last letter to Hardy. Here we revisit Ramanujan’s original claims from his deathbed letter [8], which begins by summarizing the asymptotic properties, near roots of unity, of the Eulerian series which were modular forms. He then asked whether others with similar asymptotic were necessary for the summation of a modular form and a function which is $O(1)$ at all roots of unity. In fact, the recent work by Griffin et al. [9] has confirmed that there are no weakly holomorphic modular forms which exactly cut out the singularities of Ramanujan’s mock theta functions.
Claim (Ramanujan [8])
As q approaches a primitive even order 2k root of unity ζ radially within the unit disk, then
where the mock theta function $f(q)$ is defined by
and
Throughout this paper, let $q={e}^{2\pi i\tau}$. The function $f(q)$ is convergent for $q<1$ and those roots of unity q with odd order. For the even order roots of unity, $f(q)$ has exponential singularities. For example, $f(0.994)\sim 1.08\cdot {10}^{31}$, $f(0.996)\sim 1.02\cdot {10}^{46}$, $f(0.998)\sim 6.41\cdot {10}^{90}$.
In order to cut out the exponential singularity at $q=1$, Ramanujan found the function $b(q)$ which is modular form up to multiplication by ${q}^{\frac{1}{24}}$, defined in his notation as
Ramanujan’s last letter also inspired the problem of determining the asymptotic of the coefficients of mock theta functions such as $f(q)$. Andrews [10] and Dragonette [11] obtained asymptotic for coefficients of $f(q)$, then Bringmann and Ono [5] proved an exact formula for these coefficients. In the recent work, Folsom et al. [12, 13] provided two closed formulas for the implied constant $O(1)$.
Theorem 1.1 (FolsomOnoRhoades Theorem 1.1 of [12, 13])
If ζ is a primitive even order 2k root of unity, as q approaches ζ radially within the unit disk, then
Remark

(1)
Theorem 1.1 makes Ramanujan’s claim a special case of a more general result.

(2)
Since empty products equal 1, then Theorem 1.1 shows that
$$\underset{q\to 1}{lim}(f(q)+b(q))=4.$$(5) 
(3)
Zudilin [14] has given an elementary proof of Theorem 1.1 by using Dyson’s rank function and the AndrewsGarvan crank function in his recent work.
In the meantime, Folsom et al. [12] proved a different form formula for the $O(1)$ constant as follows.
Theorem 1.2 (FolsomOnoRhoades Theorem 1.3 of [12])
If ζ is a primitive even order 2k root of unity, as q approaches ζ radially within the unit disk, then
Remark The authors left an open problem as a challenge for someone to find an elementary proof to show that the constants appearing in Theorem 1.1 match those appearing in Theorem 1.2. Interesting enough, Theorem 1.2 possesses a proof in [12] that makes use of qseries transformations only, while Theorem 1.1 is a particular instance of a much more general result whose proof uses a machinery of mock theta functions [13]. The principal goal of this note is to give a new proof of the problem without using the relation of these two theorems.
2 Statement of results
As pointed out by many authors, Theorem 1.1 is a special case of a more general one, which surprisingly relates two wellknown qseries: Dyson’s rank function $R(\omega ;q)$ and the AndrewsGarvan crank function $C(\omega ;q)$. They play a prominent role in studying integer partition congruences.
In order to define these series, we let
and for all $n\in \mathbf{Z}$, we denote
It follows that
Following the paper [14] by Zudilin and the notation above, we can rewrite $f(q)$ and $b(q)$ as follows:
For the results of Theorems 1.1 and 1.2, we have
where
and
where
and
Here $\psi (q)$ and $\varphi (q)$ are two thirdorder mock theta functions in Ramanujan’s Lost Notebook.
Recall that Dyson’s rank function is given by
where $N(m,n)$ is the number of partitions of n with rank m.
The rank of a partition is defined to be its largest part minus the number of its parts. If $\omega \ne 1$ is a root of unity, it is known that $R(\omega ;q)$ is (up to a power of q) a mock theta function which is the holomorphic part of a harmonic Maass form of weight $1/2$.
The AndrewsGarvan crank function is defined by
where $M(m,n)$ is the number of partitions of n with crank m.
For any roots of unity ω, $C(\omega ;q)$ is (up to a power of q) a modular form. Otherwise, we need the qhypergeometric series $U(\omega ;q)$ which arises in the study of strongly unimodal sequences; it is defined by
where $\mu (m,n)$ is the number of strongly unimodal sequences of size n with rank m.
Theorem 1.1 is a special case of the following theorem which relates these three qseries, here we define ${\zeta}_{n}:={e}^{\frac{2\pi i}{n}}$.
Theorem 2.1 (FolsomOnoRhoades Theorem 1.2 of [13])
Let $1\le a<b$ and $1\le h<k$ be integers with $gcd(a,b)=gcd(h,k)=1$ and $b\mid k$. If ${h}^{\prime}$ is an integer satisfying $h{h}^{\prime}\equiv 1\phantom{\rule{0.25em}{0ex}}(mod\phantom{\rule{0.25em}{0ex}}k)$, as q approaches ${\zeta}_{k}^{h}$ radially within unity disk, then
By taking $a=1$, $b=2$, and $m=2k$, so that ${\zeta}_{b}^{a}=1$, $\zeta ={\zeta}_{m}^{h}$ is a primitive even order 2k root of unity, Theorem 1.1 follows directly because of the fact that $f(q)=R(1;q)$, $b(q)=C(1;q)$ and $\mu (q)=U(1;q)$.
Based on the results above, in this note, we try to explain the relation between the two theorems, our results are as follows.
Theorem 2.2 Let ζ be a primitive even order 2k root of unity, as q approaches ζ radially within the unit disk, we have

(1)
if $k\equiv 0\phantom{\rule{0.25em}{0ex}}(mod\phantom{\rule{0.25em}{0ex}}2)$, then
$$\underset{q\to \zeta}{lim}(f(q){(1)}^{k}b(q))=4\mu (\zeta )=4\psi (\zeta ),$$(21) 
(2)
if $k\equiv 1\phantom{\rule{0.25em}{0ex}}(mod\phantom{\rule{0.25em}{0ex}}2)$, then
$$\underset{q\to \zeta}{lim}(f(q){(1)}^{k}b(q))=4\mu (\zeta )=2\varphi (\zeta ).$$(22)
Obviously, we can get the following corollary from Theorem 2.2 directly.
Corollary 2.3 Let ζ be a primitive even order 2k root of unity,

(1)
if $k\equiv 0\phantom{\rule{0.25em}{0ex}}(mod\phantom{\rule{0.25em}{0ex}}2)$, then
$$\mu (\zeta )=\psi (\zeta ),$$(23) 
(2)
if $k\equiv 1\phantom{\rule{0.25em}{0ex}}(mod\phantom{\rule{0.25em}{0ex}}2)$, then
$$2\mu (\zeta )=\varphi (\zeta ).$$(24)
Remark

(1)
One can use the uniqueness property of limits to obtain the result of Theorem 2.2 by combining with the relationship between these three functions
$$2\varphi (q)f(q)=f(q)+4\psi (q)=b(q).$$(25) 
(2)
In this note, we prove Theorem 2.2 by using the result related with the generating functions of convex compositions given by Andrews [15] and the online Encyclopedia of Integer Sequences [16] as well as AppellLerch sums, we get the desired results.
3 Proof of the theorem
Before we give the proof of the theorem, we would like to first introduce the recent work given by Andrews [15], which is about the theory of concave and convex compositions that linked the related generating functions to combinations of classical, false, or mock theta functions and other AppellLerch sums.
Following Andrews, the strictly convex composition of n is defined by
where ${a}_{1}<{a}_{2}<\cdots <{a}_{R}<c>{b}_{1}>{b}_{2}>\cdots >{b}_{S}$, and c is called the central part and is ≥0. We denote the number of strictly convex compositions of n by ${X}_{d}(n)$. For example, ${X}_{d}(5)=6$.
The related generating functions of ${X}_{d}(n)$ are defined as
In this work, Andrews related the above generating functions to the classical and mock theta functions as follows (Theorem 1 of [15]):
where $\psi (q)$ is defined in (15), and $\alpha (q)$ is defined by
It has been termed a secondorder mock theta function by McIntosh [17].
In his deathbed letter, Ramanujan defined four thirdorder mock theta functions, three of them are as follows:
Folsom, Ono and Rhoades used the relation between these three functions in (25) and the basic hypergeometric series
to prove Theorem 1.2 and to show the new forms for $\psi (q)$ and $\varphi (q)$ in (15), (16), respectively.
Following Andrews, we find that
where $\mu (q)$ is defined in (13) of Theorem 1.1.
By combining with formula (28), we can deduce that
For the even k, let ${\zeta}_{k}={e}^{\frac{2\pi i}{k}}$ be any root of unity, we have the fact that prefactor ${(q;q)}_{\mathrm{\infty}}$ of $\alpha (q)$ vanishes at any even root of unity, while the $\alpha (q)$ has finite limits. As q approaches an even order root of unity ${\zeta}_{k}$, then
namely,
For the above fixed even k, we denote the primitive even order 2k root of unity by ${\zeta}_{2k}$. As q approaches ${\zeta}_{2k}$ radially within the unit disk, we have
The first claim in Theorem 2.2 is proved.
Folsom et al. in [12] found that a search in the Online Encyclopedia of Integer Sequences [16] shows that the values of the function $G({e}^{t})$ have some relations with the coefficients of the expansion of $f({e}^{t})+b({e}^{t})$ while $t\to {0}^{+}$, where $G(q)$ is defined by
This function exists only for q, a root of unity. By studying the asymptotic relationship between $U(1;{e}^{t})$ and $G({e}^{t})$, Folsom, Ono and Rhoades guessed that there might be a relation at the other roots of unity, and then they computed the first fifth odd roots of unity ${\zeta}_{k}$, here ${\zeta}_{k}={e}^{\frac{2\pi i}{k}}$, k is odd positive integer. Surprisingly, the results of the computation clearly showed that for odd roots of unity ${\zeta}_{k}$,
Combining with the Online Encyclopedia of Integer Sequences research, we can reach the conclusion that for any odd order k root of unity ${\zeta}_{k}$, we still have the formula of (38). By the way, one can find that the function $G(q)$ converges in the domain $q<1$, because it is obvious that we can rewrite it as follows:
Applying formula (37), we then get
The search in [16] shows that this qseries matches the mock theta function $\varphi (q)$. In fact, in the proof of Theorem 1.2, Folsom, Ono and Rhoades obtained the same form for the mock theta function $\varphi (q)$ in formula (16). In this case, if q is a root of unity, we have
On the other hand, for any odd order k root of unity ${\zeta}_{k}$, from the property of $U(\omega ;q)$, we have
and
Finally, for odd integer k, if ${\zeta}_{k}$ is any primitive odd order k root of unity, as q approaches ${\zeta}_{k}$, we have
Namely,
For the above fixed odd k, we denote the primitive even order 2k root of unity by ${\zeta}_{2k}$. As q approaches ${\zeta}_{2k}$ radially within the unit disk, we have
Since ${\zeta}_{2k}^{2}={\zeta}_{k}$, ${\zeta}_{2k}$ is the root of ${\zeta}_{k}$. Then we get
We have proved the second claim in Theorem 2.2.
On the other hand, we can prove the second claim in Theorem 2.2 in a new way which is related to the properties of AppellLerch sums.
The definition of an AppellLerch sum is given by
where
For the integers a and positive integers m, we define
and
Then $m(x,q,z)$ can be expressed as a bilateral sum [18]:
then we get
Let $\omega =1$, we have
Recall Proposition 2.6 in [19] that
For the odd integer k and $x=1$, we can eliminate $m(1,q,1)$ by using the result above as follows:
Then we have
For odd integer k, if ζ is a root of unity of order 2k, as $q\to \zeta $, we see that the factor $\frac{j(1;q)}{{J}_{1}}=2{(q;q)}_{\mathrm{\infty}}^{2}$ vanishes while the summation is finite at unity, then we have
where $\frac{{J}_{1,2}^{2}}{{J}_{1}}=b(q)$.
Finally, we get
The second claim in Theorem 2.2 is also completed.
References
 1.
Andrews GE, Berndt BC: Ramanujan’s Lost Notebook. Springer, New York; 2009.
 2.
Zweger, S: Mock theta functions. Ph.D. Thesis (Advisor: D. Zaier), Universiteit Utrecht (2002)
 3.
Zweger S: Mock ϑ functions and real analytic modular forms. Contemp. Math. In qSeries with Applications to Combinatorics, Number Theory, and Physics. Am. Math. Soc., Providence; 2001:269277.
 4.
Bringmann K, Ono K: Dyson’s ranks and Maass forms. Ann. Math. 2010, 171: 419449. 10.4007/annals.2010.171.419
 5.
Bringmann K, Ono K:The $f(q)$ mock theta function conjecture and partition ranks. Invent. Math. 2006, 165: 243266. 10.1007/s0022200504935
 6.
Zagier, D: Ramanujan’s mock theta functions and their applications (after Zwegers and OnoBringmann). In: Séminaire Bourbaki, vol. 2007/2008, Astérisque. Exp. 326 pp. 143164 (2009)
 7.
Bruinier JH, Funke J: On two geometric theta lifts. Duke Math. J. 2004, 125: 4590. 10.1215/S0012709404125138
 8.
Berndt BC, Rankin RA: Ramanujan: Letters and Commentary. Am. Math. Soc., Providence; 1995.
 9.
Griffin M, Ono K, Rolen L: Ramanujan’s mock theta functions. Proc. Natl. Acad. Sci. USA 2013, 110: 57655768. 10.1073/pnas.1300345110
 10.
Andrews GE: On the theorems of Watson and Dragonette for Ramanujan’s mock theta functions. Am. J. Math. 1966, 88: 454490. 10.2307/2373202
 11.
Dragonette L: Some asymptotic formulae for the mock theta series of Ramanujan. Trans. Am. Math. Soc. 1952, 72: 474500. 10.1090/S00029947195200499278
 12.
Folsom A, Ono K, Rhoades RC: Ramanujan’s radial limits. Contemp. Math. 2012, 334: 112.
 13.
Folsom A, Ono K, Rhoades RC: Mock theta functions and quantum modular forms. Forum Math., Pi 2013., 1: Article ID e2
 14.
arXiv: 1309.2053v2
 15.
Andrews GE: Concave and convex compositions. Ramanujan J. 2013, 31: 6782. 10.1007/s1113901293946
 16.
Sloane, NJA: The online Encyclopedia of Integer Sequences. http://oeis.org
 17.
McIntosh RJ: Second order mock theta function. Can. Math. Bull. 2007, 50: 284290. 10.4153/CMB20070289
 18.
Ramanujan S: The Lost Notebook and Other Unpublished Papers. Narosa Publishing House, New Delhi; 1988.
 19.
arXiv: 1208.6316
Acknowledgements
We thank the editor and the referees for their valuable suggestions to improve the quality of this article. In addition, this article is supported by the Natural Science Foundation of China under Grant (11271283), the Natural Science Foundation of Shaanxi Province of China under Grant (2012JM1021), the Foundation of CMIRI of Shaanxi Province under Grant (2013JMR11) and the Key Subject Foundation under Grant (14SXZD002).
Author information
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
The authors declare that the study was realized in collaboration with the same responsibility. All authors read and approved the final manuscript.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0), which permits use, duplication, adaptation, distribution, and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Received
Accepted
Published
DOI
Keywords
 modular forms
 mock theta functions
 generating functions