Introduction and Statements of Results
In his last letter to Hardy, Ramanujan [3 pp. 354–355] introduced seventeen q-series, which he called mock theta functions. According to Ramanujan, a mock theta function is a function F whose asymptotic behavior near roots of unity resembles that of weakly holomorphic modular forms, however, there does not exist a single weakly holomorphic modular form whose difference from F remains bounded as q approaches any root of unity radially. In his seminal thesis, Zwegers [11] explained the modularity of Ramanujan’s mock theta functions by “completing” them to so-called harmonic Maass forms.1
Next recall that Dyson [5] introduced the rank2 statistic to give a combinatorial explanation of partition congruences of Ramanujan (see [10]). Note that the generating function of the number of partitions with even rank minus the number of partitions with odd rank is Ramanujan’s third order mock theta function
f(q):=n≥0∑(−q)n2qn2.
2020 Mathematics Subject Classification. 05A17, 11P81, 11P84.
Key words and phrases. mock theta functions, overpartitions, partitions, two-color partitions, universal mock theta functions.
Here, and throughout, ∣q∣<1 and for a∈C and n∈N0∪{∞}, we set (a)n=(a;q)n:=∏j=0n−1(1−aqj).
Moreover, MacMahon [9] studied partitions without sequences which are related to mixed mock modular forms.3 More precisely, using work of Fine [6], Andrews [2] showed that the generating function of partitions without sequences is related to Ramanujan’s third order mock theta function χ(q). Note that Ramanujan’s mock theta functions are specializations of the universal mock theta functions (see [8 (3.2) and above (5.1)])
g3(ζ;q):=n≥0∑(ζ,ζq;q)n+1qn(n+1),g2(ζ;q):=n≥0∑(ζ,ζq;q)n+1(−q)nq2n(n+1).(1)
Here ζ∈C∖{0} and for r∈N and a1,…,ar∈C, we let (a1,…,ar)n:=(a1,…,ar;q)n:=(a1)n⋯(ar)n.
In [4 Theorem 1.3], Lovejoy, Mahlburg, and the second author gave a combinatorial interpretation of (for d≥3 and d>2r)
(q2d;q2d)∞(−qr,−qd−r;qd)∞g2(−qr;qd).
Moreover, [4 (1.2)] yields a combinatorial statistic that is related to (for d≥3 and d>2r)
(−qr,−qd−r;qd)∞g3(−qr;qd).
Note that d=2r was excluded in [4]. In this case, the residue classes r and −r (mod d) coincide. Thus (−qr,−qd−r;qd)∞=(−qr;q2r)∞ and each part ≡r (mod 2r) comes in two copies. This naturally leads to overpartitions and two-color partitions. This is a key motivation for our combinatorial framework.
In this note, we consider restricted partitions, overpartitions, and colored partitions and relate the corresponding generating functions to the universal mock theta functions. We begin with different q-hypergeometric series representations of g3 and g2. To the authors’ knowledge these representations are not known.
Theorem 1.1. Let ζ=qm for any m∈Z.
(1) We have
g3(ζ;q)=(ζ,ζq)∞(q)∞n≥0∑(q)n(ζ,ζq)nqn.(2)
(2) We have
g2(ζ;q)=(ζ,ζq)∞(q2;q2)∞n≥0∑(q2;q2)n(ζ,ζq)nqn.(3)
With a suitable specialization of the variables ζ and q and after multiplying by appropriate infinite products, the identities in Theorem 1.1 yield the generating functions for classes of restricted partitions. The summation index records either the largest part in a certain residue class or the number of parts in that class, while the remaining factors impose restrictions on the parts in other residue classes.
We now describe these partition statistics explicitly. Here and throughout, we assume d,r∈N with d≥2r. First, recall that an overpartition of n is a partition of n in which the first occurrence of a part may be overlined. Let p(n) denote the number of overpartitions of n and set p(0):=1. For instance, p(2)=4 enumerated by 2, 2, 1+1, 1+1. We now consider overpartitions of n with the following restrictions:
parts are divisible by r,
parts that are divisible by 2r are non-overlined,
the largest part is divisible by 2r,
each part ≡r(mod2r) can appear at most twice and if it appears twice, then one of the two occurrences must be overlined.
Let opr[1](n) denote the number of such overpartitions of n and set opr[1](0):=1. For example, op2[1](12)=5 enumerated by 12, 8+4, 8+2+2, 4+4+4, 4+4+2+2. Let OPr[1](q):=∑n≥0opr[1](n)qn denote the corresponding generating function.
Theorem 1.2. We have
OPr[1](q)=(q2r;q2r)∞(−qr;q2r)∞2g3(−qr;q2r).
Next, for d>2r, we consider partitions of n with the following restrictions:
parts are ≡0,±r(modd),
each part ≡±r(modd) is greater than every part divisible by d.
Let pd,r[2](n) denote the number of such partitions of n and set pd,r[2](0):=1.
For example, p4,1[2](5)=3 enumerated by 5, 3+1+1, 1+1+1+1+1. Let Pd,r[2](q):=∑n≥0pd,r[2](n)qn denote the corresponding generating function.
Theorem 1.3. We have, for d>2r,
Pd,r[2](q)=(qd;qd)∞1Q3(qr;qd).
Next, recall that a two-color partition of n is a partition of n in which each part appears in one of two fixed colors, say blue or green. Parts of the same size but with different colors are regarded as distinct and if a part occurs both in blue and in green, then we list the green parts first. Let p2(n) denote the number of two-color partitions of n and set p2(0):=1. For example, p2(2)=5 enumerated by 2g, 2b, 1g+1g, 1g+1b, 1b+1b. We consider two-color partitions of n with the following restrictions:
parts are divisible by r,
parts divisible by 2r can appear only in blue,
each part ≡r(mod2r) is greater than parts divisible by 2r.
Let pr[2](n) denote the number of such two-color partitions of n and set pr[2](0):=1. For example, p1[2](8)=11 enumerated by 8b, 6g+2g, 6b+2g, 6g+2b, 6b+2b, 4b+4b, 2g+2g+2g+2g, 2g+2g+2g+2b, 2g+2g+2b+2b, 2g+2b+2b+2b, 2b+2b+2b+2b. Let Pr[2](q):=∑n≥0pr[2](n)qn denote the corresponding generating function.
Theorem 1.4. We have
Pr[2](q)=(q2r;q2r)∞1Q3(qr;q2r).
Now, we consider overpartitions of n with the following restrictions:
parts are divisible by r but not by 4r,
parts ≡2r(mod4r) are non-overlined,
each part ≡r(mod2r) can appear at most twice, and if it appears twice, it appears once overlined and once non-overlined,
if there are exactly m parts ≡2r(mod4r) (counted with multiplicity), then every part ≡r(mod2r) is less than 2rm. In particular, if there is no part ≡2r(mod4r), then there is no part ≡r(mod2r).
Let opr[2](n) denote the number of such overpartitions of n and set opr[2](0):=1. For example, op1[2](6)=3 enumerated by 6, 2+2+2, 2+2+1+1. Let OPr[2](q):=∑n≥0opr[2](n)qn denote the corresponding generating function.
Theorem 1.5. We have
OPr[2](q)=(q4r;q4r)∞(−qr;q2r)∞2g2(−qr;q2r).
Next, for d>2r, we consider partitions of n with the following restrictions:
parts are ≡d(mod2d) or ≡±r(modd),
if there are exactly m parts ≡d(mod2d) (counted with multiplicity), then each part ≡±r(modd) is larger than dm.
Let pd,r[4](n) denote the number of such partitions of n and set pd,r[4](0):=1. For example, p4,1[4](5)=3 enumerated by 5, 3+1+1, 1+1+1+1+1. Let Pd,r[4](q):=∑n≥0pd,r[4](n)qn denote the corresponding generating function.
Theorem 1.6. We have, for d>2r,
Pd,r[4](q)=(q2d;q2d)∞1g2(qr;qd).
Finally, for d=2r, we consider two-color partitions of n with the following restrictions:
parts are divisible by r but not by 4r,
parts ≡2r(mod4r) appear only in blue,
if there are exactly m parts ≡2r(mod4r) (counted with multiplicity), then every part ≡r(mod2r) is larger than 2rm.
Let pr[3](n) denote the number of such two-color partitions of n and set pr[3](0):=1. For example, p2[3](8)=10 enumerated by 6g+2g, 6b+2g, 6g+2b, 6b+2b, 4b+4b, 2g+2g+2g+2g, 2g+2g+2g+2b, 2g+2g+2b+2b, 2g+2b+2b+2b, 2b+2b+2b+2b. Let Pr[3](q):=∑n≥0pr[3](n)qn denote the corresponding generating function.
Theorem 1.7. We have
Pr[3](q)=(q4r;q4r)∞1g2(qr;q2r).
This note is organized as follows. In Section 2, we recall q-series transformations which are required to prove Theorem 1.1. Section 3 presents the proofs of Theorems 1.1–1.7. Section 4 is devoted to open problems for future research.
Acknowledgements
The authors have received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 101001179).
Preliminaries
In this section we recall q-series identities4 which are used in the proofs of our results. We start with Fine’s identities. To state these, following [6 equation (1.1)], for a∈C, b=q−m for every m∈N, and ∣w∣<1, we define
F(a,b;w;q):=n≥0∑(bq)n(aq)nwn.
By [6 equation (11.3)], we have the following identity.
Lemma 2.1. We have
(q)∞(bq,wq)∞F(0,b;w;q)=1−w1n≥0∑(q)n(b,w)nqn.
Next, from [6 (12.3)], we obtain the following lemma.
Lemma 2.2. We have
(1−w)F(0,b;w;q)=n≥0∑(bq,wq)n(bw)nqn2.
Next we require the following transformation from [7 (III.10)].
Lemma 2.3. We have
n≥0∑(q,baq,caq)n(bcaq,d,e)n(deaq)n=(baq,caq,deaq)∞(daq,eaq,bcaq)∞n≥0∑(q,daq,eaq)n(deaq,b,c)n(bcaq)n.
Moreover, we require the following identity due to Euler (see [1 (2.2.5)]).
Lemma 2.4. For ∣w∣,∣q∣<1, we have
n≥0∑(q)nwn=(w)∞1.
Proof of Theorems 1.1–1.7
We start with the proof of Theorem 1.1.
Proof of Theorem 1.1. Combining Lemmas 2.1 and 2.2, we obtain
n≥0∑(bq,wq)n(bw)nqn2=(bq,wq)∞(q)∞n≥0∑(q)n(b,w)nqn.
Letting (b,w)=(ζ,ζq), it follows that
n≥0∑(ζq,ζq2)nqn(n+1)=(ζq,ζq2)∞(q)∞n≥0∑(q)n(ζ,ζq)nqn.
This and the definition of g3 from (1.1) conclude the proof of Theorem 1.1 (1). Next, using the definition of g2 in (1.1) and the fact that
w→0lim(−wq)nwn=q2n(n+1),
we have
(1−ζ)(1−ζq)g2(ζ;q)=w→0limn≥0∑(ζq,ζq2)n(−q,−wq)nwn.(4)
Now, letting (a,b,c,d,e)=(q,ζq,ζ,−q,−wq) in Theorem 2.3, we get
n≥0∑(ζq,ζq2)n(−q,−wq)nwn=(ζq,ζq2,w)∞(−q,−wq,q)∞n≥0∑(q,−q,−wq)n(wq,ζq,ζ)nqn.
Taking the limit w→0 on both sides, and using (3.1), we conclude the proof of Theorem 1.1 (2). □
We next prove Theorem 1.2.
Proof of Theorem 1.2. Using the definition of OPr[1](n), it is not hard to see that
OPr[1](q)=n≥0∑(q2r;q2r)n(−qr;q2r)n2q2rn.
Here, for n∈N, the largest part is 2rn, while n=0 accounts for the empty partition. Letting q↦q2r and ζ=−qr in Theorem 1.1 (1), we thus obtain the claim. □
Now we establish Theorem 1.3.
Proof of Theorem 1.3. By definition of pd,r[2](n), we have
Pd,r[2](q)=n≥0∑(qd;qd)n(qdn+r,qdn+d−r;qd)∞qdn.
Here, for n∈N, the largest part divisible by d is dn, while n=0 accounts for partitions with no parts divisible by d. Letting q↦qd and ζ=qr in Theorem 1.1 (1), we thus obtain the claim. □
Next, we prove Theorem 1.4.
Proof of Theorem 1.4. By definition of pr[2](n), we obtain
Pr[2](q)=n≥0∑(q2r;q2r)n(q2rn+r;q2r)∞2q2rn.
Letting q↦q2r and ζ=qr in Theorem 1.1 (1), we thus obtain the claim. □
Next, we show the following elementary lemma which we require to prove Theorems 1.5–1.7.
Lemma 3.1. For k∈N and m∈N0, (q2k;q2k)mqkm is the generating function for partitions into exactly m parts which are ≡k(mod2k).
Proof. Introduce a variable w to record the number of parts. For each allowed part k(2j+1) (j∈N0), we have
1−wqk(2j+1)1=r≥0∑wrqk(2j+1)r.
The term indexed by r corresponds to taking r copies of this part. Thus the coefficient of wm in
j≥0∏1−wqk(2j+1)1=(wqk;q2k)∞1
is the generating function for partitions into exactly m parts congruent to k(mod2k). Letting q↦q2k and w↦wqk in Theorem 2.4, we obtain
n≥0∑(q2k;q2k)nqknwn=(wqk;q2k)∞1.
Comparing the coefficients of wm on both sides gives the claim. □
Now we show Theorem 1.5.
Proof of Theorem 1.5. Let m∈N0 denote the number of parts ≡2r(mod4r) (including repetitions) of an overpartition counted by opr[2](n). By (2), these parts are non-overlined. By Theorem 3.1 with k=2r, they are generated by
(q4r;q4r)mq2rm.
By (4), the parts ≡r(mod2r) lie in {r(2k−1):1≤k≤m}. Moreover, by (3), each of these parts appears at most once non-overlined and at most once overlined. Hence these parts are generated by
j=0∏m−1(1+qr(2j+1))2=(−qr;q2r)m2.
Multiplying these two generating functions and summing over m, we obtain
OPr[2](q)=m≥0∑(q4r;q4r)m(−qr;q2r)m2q2rm.
Letting q↦q2r and ζ=−qr in Theorem 1.1 (2), we thus obtain the claim. □
Next, we prove Theorem 1.6.
Proof of Theorem 1.6. Let m∈N0 be the number of parts ≡d(mod2d) (including repetitions) of a partition counted by pd,r[4](n). By Theorem 3.1 with k=d, these parts are generated by
(q2d;q2d)mqdm.
By (2), the parts ≡r(modd) are at least dm+r and the parts ≡−r(modd) are at least dm+d−r. These parts are therefore generated by
(qdm+r,qdm+d−r;qd)∞1.
Multiplying these two generating functions and summing over m, we obtain
Pd,r[4](q)=m≥0∑(q2d;q2d)m(qdm+r,qdm+d−r;qd)∞qdm.
Letting q↦qd and ζ=qr in Theorem 1.1 (2), we thus obtain the claim. □
Finally, we establish Theorem 1.7.
Proof of Theorem 1.7. Let m∈N0 denote the number of parts ≡2r(mod4r) (including repetitions) of a two-color partition counted by pr[3](n). By (2), these parts appear in blue, so by Theorem 3.1 with k=2r, they are generated by
(q4r;q4r)mq2rm.
By (2) and (3), the parts ≡r(mod2r) are at least 2rm+r. They are therefore generated by
(q2rm+r;q2r)∞21.
Multiplying these two generating functions and summing over m, we obtain
Pr[3](q)=m≥0∑(q4r;q4r)m(q2rm+r;q2r)∞2q2rm.
Letting q↦q2r and ζ=qr in Theorem 1.1 (2), we thus obtain the claim. □
Open problems
We conclude this paper with the following problems for future research:
(1) Numerically,5 we found the following congruences:
op1[1](12n+8)≡0(mod2),
op1[1](20n+j)≡0(mod2)with j∈{8,16},
op1[1](28n+j)≡0(mod2)with j∈{8,16,20}.
These lead to the following conjecture. Let (⋅⋅) denote the Legendre symbol.
Conjecture 4.1. For any odd prime p and k∈N0, we have
2∣op1[1](4pn+4k)for all n∈N0⟺(p8k+1)=−1.
(2) For op1[2](n), numerically, we observe the following congruences:
op1[2](18n+j)≡0(mod4)with j∈{8,14},
op1[2](50n+j)≡0(mod4)with j∈{16,26,36,46},
op1[2](98n+j)≡0(mod4)with j∈{26,40,54,68,82,96}.
These suggest the following general pattern.
Conjecture 4.2. Let p be an odd prime with p≡1(mod8). Then for all n∈N0 and 1≤k≤p−1, we have
op1[2](2p2n+4p2−1+2pk)≡0(mod4).
University of Cologne, Department of Mathematics and Computer Science, Weyertal 86-90, 50931 Cologne, Germany
Email address: [email protected]
Email address: [email protected]