Two truncated identities of Gauss

Victor J. W. Guo1 and Jiang Zeng2

1
Department of Mathematics, East China Normal University,
Shanghai 200062, People’s Republic of China
[email protected], http://math.ecnu.edu.cn/~jwguo
2
Université de Lyon; Université Lyon 1; Institut Camille Jordan, UMR 5208 du CNRS;
43, boulevard du 11 novembre 1918, F-69622 Villeurbanne Cedex, France
[email protected], http://math.univ-lyon1.fr/~zeng

Abstract. Two new expansions for partial sums of Gauss’ triangular and square numbers
series are given. As a consequence, we derive a family of inequalities for the overpartition
function p(n) and for the partition function p1 (n) counting the partitions of n with distinct
odd parts. Some further inequalities for variations of partition function are proposed as
conjectures.
Keywords: Partition function; Overpartition function; Gauss’ identities; Shanks’ identity
AMS Subject Classifications: 11B65; 11P81; 05A17

1 Introduction
The partition function p(n) has the generating function
∞
X ∞
Y
n 1
p(n)q = n
= 1 + q + 2q 2 + 3q 3 + 5q 4 + 7q 5 + 11q 6 + · · · . (1.1)
n=0 n=1
1−q

Two classical results in the partition theory [1, p. 11] are Euler’s pentagonal number
theorem
∞
à ∞
!
Y 1 X
n
1+ (−1)j (q j(3j−1)/2 + q j(3j+1)/2 ) = 1, (1.2)
n=1
1 − q j=1

and Euler’s recursive formula for computing p(n):

∞
X
p(n) + (−1)j (p(n − j(3j − 1)/2) + p(n − j(3j + 1)/2)) = 0, (1.3)
j=1

where p(m) = 0 for all negative m.

Recently, Merca [7] has stumbled upon the following inequality:

p(n) − p(n − 1) − p(n − 2) + p(n − 5) 6 0, (1.4)

1
and then, Andrews and Merca [3] proved more generally that, for k ≥ 1,
k−1
X
(−1)k−1 (−1)j (p(n − j(3j + 1)/2) − p(n − (j + 1)(3j + 2)/2)) > 0 (1.5)
j=0

with strict inequality if n > k(3k + 1)/2.

The proof of (1.5) in [3] is based on the truncated formula of (1.2):
k · ¸
q (k+1)n+(2) n − 1
k−1 ∞
1 X j j(3j+1)/2 2j+1 k−1
X
(−1) q (1 − q ) = 1 + (−1) , (1.6)
(q; q)∞ j=0 n=k
(q; q)n k−1 q

where
∞
Y · ¸
n (a; q)∞ M (q; q)M
(a; q)∞ = (1 − aq ), (a; q)M = , and = .
n=0
(aq M ; q)∞ N q (q; q)N (q; q)M −N

Motivated by Andrews and Merca’s work [3], in this paper we shall prove new truncated
forms of two identities of Gauss [1, p. 23]:
∞
X 2 (q; q)∞
1+2 (−1)j q j = , (1.7)
j=1
(−q; q)∞
∞
X (q 2 ; q 2 )∞
(−1)j q j(2j+1) (1 − q 2j+1 ) = , (1.8)
j=0
(−q; q 2 )∞

and derive similar overpartition function and special partition function inequalities.
Theorem 1. For |q| < 1 and k > 1, there holds
à k
! ∞ · ¸
(−q; q)∞ X 2
X (−q; q)k (−1; q)n−k q (k+1)n n − 1
j j k
1+2 (−1) q = 1 + (−1) .
(q; q)∞ j=1 n=k+1
(q; q)n k q
(1.9)
The overpartition function p(n), for n > 1, denotes the number of ways of writing the
integer n as a sum of positive integers in non-increasing order in which the first occurrence
of an integer may be overlined or not, and p(0) = 1 (see Corteel and Lovejoy [5]). It is
easy to see that
∞
X Y∞
n 1 + qn
p(n)q = n
= 1 + 2q + 4q 2 + 8q 3 + 14q 4 + 24q 5 + 40q 6 + · · · .
n=0 n=1
1 − q

Corollary 2. For n, k > 1, there holds

à k
!
X
(−1)k p(n) + 2 (−1)j p(n − j 2 ) >0 (1.10)
j=1

2
with strict inequality if n > (k + 1)2 . For example,

p(n) − 2p(n − 1) 6 0, (1.11)

p(n) − 2p(n − 1) + 2p(n − 4) > 0,
p(n) − 2p(n − 1) + 2p(n − 4) − 2p(n − 9) 6 0.

Theorem 3. For |q| < 1 and k > 1, there holds

k−1
(−q; q 2 )∞ X
2 2
(−1)j q j(2j+1) (1 − q 2j+1 )
(q ; q )∞ j=0
∞
X · ¸
k−1 (−q; q 2 )k (−q; q 2 )n−k q 2(k+1)n−k n − 1
= 1 + (−1) . (1.12)
n=k
(q 2 ; q 2 )n k−1 q2

Let p1 (n) denote the number of partitions of n in which odd parts are not repeated.
It is easy to see that
∞
X ∞
Y 1 + q 2n−1
p1 (n)q n = = 1 + q + q 2 + 2q 3 + 3q 4 + 4q 5 + 5q 6 + 7q 7 + · · · .
n=0 n=1
1 − q 2n

Corollary 4. For n, k > 1, there holds

k−1
X ³ ´
(−1)k−1 (−1)j p1 (n − j(2j + 1)) − p1 (n − (j + 1)(2j + 1)) > 0 (1.13)
j=0

with strict inequality if n > (2k + 1)k. For example,

p1 (n) − p1 (n − 1) − p1 (n − 3) + p1 (n − 6) 6 0,
p1 (n) − p1 (n − 1) − p1 (n − 3) + p1 (n − 6) + p1 (n − 10) − p1 (n − 15) > 0.

A nice combinatorial proof of (1.3) was given by Bressoud and Zeilberger [4]. It would
be interesting to find a combinatorial proof of (1.5), (1.10) and (1.13). A combinatorial
proof of (1.11) will be given in Section 4.

2 Proof of Theorem 1
Generalizing Shanks’ work [8, 9], Andrews, Goulden, and Jackson [2, Theorem 1] estab-
lished the following identity
n
X n
(bq; q)n X (b/a; q)j aj q (n+1)j
2
(b; q)j (1 − bq 2j )(b/a; q)j aj q j
= . (2.1)
j=0
(1 − b)(q; q)j (aq; q)j (aq; q)n j=0 (q; q)j

When b = 1 and a = −1, the identity (2.1) reduces to

3
 n
X n
X
2 (−1; q)j (q; q)n q (n+1)j
1+2 (−1)j q j = (−1)j . (2.2)
j=1 j=0
(q; q)j (−q; q)n

By (2.2) and the q-binomial theorem (see [1, Theorem 2.1]), we have
à k
! k
(−q; q)∞ X (−q; q)∞ X (−1; q)j (q; q)k q (k+1)j
j j2
1+2 (−1) q = (−1)j
(q; q)∞ j=1
(q; q)∞ j=0 (q; q)j (−q; q)k
k
X (−1; q)j (−q k+1 ; q)∞ q (k+1)j
= (−1)j
j=0
(q; q)j (q k+1 ; q)∞
k
X ∞
X
j (−1; q)j (−1; q)i q (k+1)(i+j)
= (−1) . (2.3)
j=0 i=0
(q; q)j (q; q)i

By induction on k, it is easy to see that, for n > 1,

k
X (−1; q)j (−1; q)n−j (−q; q)k (−1; q)n−k
(−1)j = (−1)k .
j=0
(q; q)j (q; q)n−j (1 − q n )(q; q)n−k−1 (q; q)k

Hence, letting i + j = n, the right-hand side of (2.3) can be written as

∞ X
X k
(−1; q)j (−1; q)n−j q (k+1)n
(−1)j
n=0 j=0
(q; q)j (q; q)n−j
∞
X
k (−q; q)k (−1; q)n−k q (k+1)n
= 1 + (−1)
n=1
(1 − q n )(q; q)n−k−1 (q; q)k
X∞ · ¸
k(−q; q)k (−1; q)n−k q (k+1)n n − 1
= 1 + (−1) ,
n=k+1
(q; q)n k q
as desired.

3 Proof of Corollary 2
By Theorem 1, we see that the generating function for the sequence
( Ã k
!)∞
X
(−1)k p(n) + 2 (−1)j p(n − j 2 )
j=1 n=0
is given by
X∞ · ¸
k (−q; q)k (−1; q)n−k q (k+1)n n − 1
(−1) + ,
n=k+1
(q; q)n k q

which clearly has nonnegative coefficients of q m for m > 1 and has positive coefficients of
q m for m > (k + 1)2 . This completes the proof.

4
4 A combinatorial proof of (1.11)
Let Pn denote the set of all overpartitions of n. We now construct a mapping φ : Pn →
Pn−1 as follows: For any λ = (λ1 , . . . , λk ) ∈ Pn , let


 (λ1 , . . . , λk−1 ), if λk = 1,


(λ1 , . . . , λk−1 , λk − 1), if λk 6= 1, 1,
φ(λ) =

 (λ1 , . . . , λk−2 , 1̂, . . . , 1), if λk = 1,

 | {z }
λk−1 1’s

where 1̂ = 1 if λk−1 is overlined and 1̂ = 1 otherwise.

It is easy to see that |φ−1 (µ)| 6 2 for any µ ∈ Pn−1 . For example, for n = 4, the
mapping φ gives
4 7→ 3, 4 7→ 3, (3, 1) 7→ 3, (3, 1) 7→ (1, 1, 1), (3, 1) 7→ (1, 1, 1), (3, 1) 7→ 3,
(2, 2) 7→ (2, 1), (2, 2) 7→ (2, 1), (2, 1, 1) 7→ (2, 1), (2, 1, 1) 7→ (2, 1), (2, 1, 1) 7→ (2, 1),
(2, 1, 1) 7→ (2, 1), (1, 1, 1, 1) 7→ (1, 1, 1), (1, 1, 1, 1) 7→ (1, 1, 1).
This proves that p(n) 6 2p(n − 1).

5 Proof of Theorem 3 and Corollary 4

Shanks [8, 9] proved that
Xn Xn
j(2j−1) j(2j+1) (q; q 2 )j (q 2 ; q 2 )n q j(2n+1)
1+ (q +q )= 2 ; q 2 ) (q; q 2 )
,
j=1 j=0
(q j n

or equivalently,
n−1
X n−1
X
j(2j+1) 2j+1 (q; q 2 )j (q 2 ; q 2 )n q j(2n+1)
q (1 + q )= . (5.1)
j=0 j=0
(q 2 ; q 2 )j (q; q 2 )n

By (5.1) (with q replaced by −q) and the q-binomial theorem (see [1, Theorem 2.1]),
we have
k−1
(−q; q 2 )∞ X
(−1)j q j(2j+1) (1 − q 2j+1 )
(q 2 ; q 2 )∞ j=0
k−1
(−q; q 2 )∞ X (−q; q 2 )j (q 2 ; q 2 )k q (2k+1)j
= 2 2 (−1)j
(q ; q )∞ j=0 (q 2 ; q 2 )j (−q; q 2 )k
k−1
X (−q; q 2 )j (−q 2k+1 ; q)∞ q (2k+1)j
= (−1)j
j=0
(q 2 ; q 2 )j (q 2k+2 ; q)∞
k−1
X ∞
X
j (−q; q 2 )j (−q −1 ; q 2 )i q (2k+1)(i+j)+i
= (−1) . (5.2)
j=0 i=0
(q 2 ; q 2 )j (q 2 ; q 2 )i

5
By induction on k, it is easy to see that, for n > 1,
k−1
X (−q; q 2 )j (−q −1 ; q 2 )n−j q n−j (−q; q 2 )k (−q; q 2 )n−k q n−k
(−1)j = (−1)k−1
.
j=0
(q 2 ; q 2 )j (q 2 ; q 2 )n−j (1 − q 2n )(q 2 ; q 2 )n−k (q 2 ; q 2 )k−1

Hence, letting i + j = n, the right-hand side of (5.2) can be written as

∞ X
X k−1
(−q; q 2 )j (−q −1 ; q 2 )n−j q (2k+1)n+n−j
(−1)j
n=0 j=0
(q 2 ; q 2 )j (q 2 ; q 2 )n−j
∞
X (−q; q 2 )k (−q; q 2 )n−k q 2(k+1)n−k
= 1 + (−1)k−1 .
n=1
(1 − q 2n )(q 2 ; q 2 )n−k (q 2 ; q 2 )k−1
∞
X · ¸
(−q; q 2 )k (−q; q 2 )n−k q 2(k+1)n−k n − 1
= 1 + (−1)k−1 .
n=k
(q 2 ; q 2 )n k−1 q2

This proves Theorem 3. The proof of Corollary 4 is similar to that of Corollary 2 and is
omitted here.

6 Open problems
In this section, we propose a common generalization of (1.5), (1.10) and (1.13). Let
m, r be positive integers with 1 6 r 6 m/2. Consider the generalized partition function
Jm,r (n) defined by
∞
X 1
Jm,r (n)q n = . (6.1)
n=0
(q r ; q m )∞ (q m−r ; q m )∞ (q m ; q m )∞

It is easy to see that

J2,1 (n) = p(n), J3,1 (n) = p(n), J4,1 (n) = p1 (n).
Moreover, if r < m/2, then Jm,r (n) can be understood as the number of partitions of n
into parts congruent to 0, ±r modulo m. Now, Jacobi’s triple product identity implies
(see [6, p. 375]) that
∞
X
r m m−r m m m
(q ; q )∞ (q ; q )∞ (q ; q )∞ = 1 + (−1)j (q j(mj+m−2r)/2 + q j(mj+m+2r)/2 ). (6.2)
j=1

It follows from (6.1) and (6.2) that Jm,r (n) satisfies the recurrence formula:
∞
X
Jm,r (n) + (−1)j (Jm,r (n − j(mj − m + 2r)/2) + Jm,r (n − j(mj + m − 2r)/2)) = 0,
j=1
(6.3)
where Jm,r (s) = 0 for all negative s.

6
Conjecture 5. For m, n, k, r > 1 with r 6 m/2, there holds
k−1
X ³ ´
k−1
(−1) (−1)j Jm,r (n − j(mj + m − 2r)/2) − Jm,r (n − (j + 1)(mj + 2r)/2) > 0
j=0
(6.4)
with strict inequality if n > k(mk + m − 2r)/2.
In fact, when m = 2 and r = 1, the conjectural inequality (6.4) is equivalent to
à k−1
!
X
(−1)k−1 p(n) + 2 (−1)j p(n − j 2 ) > p(n − k 2 ) (6.5)
j=1

with strict inequality if n > k 2 . It is clear that (6.5) is stronger than the proved inequality
(1.10) (with k replaced by k − 1).
Finally, along the same line of thinking, we consider the sequence {t(n)}n≥0 (see
A000716 in Sloane’s database of integer sequences [10]) defined by
∞
X 1
t(n)q n = (6.6)
n=0
(q; q)3∞

= 1 + 3q + 9q 2 + 22q 3 + 51q 4 + 108q 5 + 221q 6 + 429q 7 + 810q 8 + 1479q 9 + · · · .

Clearly, the number t(n) counts partitions of n into 3 kinds of parts. Now, invoking the
identity of Jacobi [6, p. 377]:
∞
X
(q; q)3∞ = (−1)j (2j + 1)q j(j+1)/2 ,
j=0

we derive the recurrence formula:

X∞
(−1)j (2j + 1)t(n − j(j + 1)/2) = 0, (6.7)
j=0

where t(m) = 0 for all negative m.

We end the paper with the following conjecture:
Conjecture 6. For n, k > 1, there holds
k
X
k
(−1) (−1)j (2j + 1)t(n − j(j + 1)/2) > 0
j=0

with strict inequality if n > (k + 1)(k + 2)/2. For example,

t(n) − 3t(n − 1) 6 0,
t(n) − 3t(n − 1) + 5t(n − 3) > 0,
t(n) − 3t(n − 1) + 5t(n − 3) − 7t(n − 6) 6 0.

7
References
[1] G.E. Andrews, The Theory of Partitions, Cambridge University Press, Cambridge, 1998.
[2] G.E. Andrews, I.P. Goulden, and D.M. Jackson, Shanks’ convergence acceleration trans-
form, Padé approximants and partitions, J. Combin. Theory Ser. A 43 (1986), 70–84.
[3] G.E. Andrews and M. Merca, The truncated pentagonal number theorem, preprint, 2012.
[4] D.M. Bressoud and D. Zeilberger, Bijecting Euler’s partitions-recurrence, Amer. Math.
Monthly 92 (1985), 54–55.
[5] S. Corteel and J. Lovejoy, Overpartitions, Trans. Amer. Math. Soc. 356 (2004), 1623–1635.
[6] G.H. Hardy and E.M. Wright, An Introduction to the Theory of Numbers, 6th Ed., Oxford
University Press, Oxford, 2008.
[7] M. Merca, Fast algorithm for generating ascending compositions, J. Math. Modelling and
Algorithms 11 (2012), 89–104.
[8] D. Shanks, A short proof of an identity of Euler, Proc. Amer. Math. Soc. 2 (19 51), 747–749.
[9] D. Shanks, Two theorems of Gauss, Pacific J. Math. 8 (1958), 609–612.
[10] N.J.A. Sloane, On-Line Encyclopedia of Integer Sequences, http://oeis.org/