European Journal of Combinatorics

Manuscript Draft

Manuscript Number: YEUJC-D-14-00471

Title: Bilateral truncated Jacobi's identity

Article Type: Full Length Article

Abstract: Recently, Andrews and Merca considered the truncated version of Euler's pentagonal number
theorem and obtained a non-negative result on the coefficients of this truncated series. Guo and Zeng
showed the coefficients of two truncated Gauss' identities are non-negative and they conjectured that
the truncated Jacobi's identity also has non-negative coefficients. Mao provided a proof of this
conjecture by using an algebraic method. In this paper, we consider the bilateral truncated Jacobi's
identity and show that when the upper and lower bounds of the summation satisfy some certain
restrictions, then this bilateral truncated identity has non-negative coefficients. As a corollary, we
show the conjecture of Guo and Zeng holds. Our proof is purely combinatorial and mainly based on a
bijection for Jacobi's identity.
Manuscript

Bilateral truncated Jacobi’s identity

Thomas Y. He1 , Kathy Q. Ji2 and Wenston J.T. Zang3
1,2,3
Center for Combinatorics, LPMC-TJKLC
Nankai University, Tianjin 300071, P.R. China
1
heyao@mail.nankai.edu.cn, 2 ji@nankai.edu.cn, 3 wenston@mail.nankai.edu.cn

Abstract. Recently, Andrews and Merca considered the truncated version of Euler’s
pentagonal number theorem and obtained a non-negative result on the coefficients of
this truncated series. Guo and Zeng showed the coefficients of two truncated Gauss’
identities are non-negative and they conjectured that the truncated Jacobi’s identity also
has non-negative coefficients. Mao provided a proof of this conjecture by using an algebraic
method. In this paper, we consider the bilateral truncated Jacobi’s identity and show that
when the upper and lower bounds of the summation satisfy some certain restrictions, then
this bilateral truncated identity has non-negative coefficients. As a corollary, we show the
conjecture of Guo and Zeng holds. Our proof is purely combinatorial and mainly based
on a bijection for Jacobi’s identity.
Keywords: Partitions, Jacobi’s identity.
AMS Classifications: 05A17, 11P82, 11P83.

1 Introduction
This paper is concerned with bilateral truncated Jacobi’s identity. Recall that Jacobi’s
identity [9, p.257, Eq.(5)] (see also [19, Theorem 357]) is given by

∑
+∞
j+1
(q; q)3∞ = (−1)j (2j + 1)q ( 2 ), (1.1)
j=0

which is equivalent to
∑
+∞
j+1
(q; q)3∞ = (−1)j jq (2 ). (1.2)
j=−∞

It plays an important role in the study of partition congruences and representations of

integers as sums of squares, see, for example, Andrews [1, 2], Ewell [8], Hirschhorn [12],
Hirschhorn and Sellers [13] and Ramanujan [15, 16].
The truncated theta series were recently studied by several authors, see, for example,
Andrews and Merca [4], Chapman [7], Guo and Zeng [11], Mao [14], Shanks [17, 18], and

1
Yee [20]. In [4], Andrews and Merca considered the truncated Euler’s pentagonal number
theorem. Euler’s pentagonal number theorem is one of the most well-known theorem in
partition theory.
∑
+∞
(q; q)∞ = (−1)j q j(3j+1)/2 . (1.3)
j=−∞

This gives
1 ∑
+∞
(−1)j q j(3j+1)/2 (1 − q 2j+1 ) = 1. (1.4)
(q; q)∞ j=0

Here and throughout this paper, we have adopted the standard notation on partitions
and q-series [3]

∏
n−1 ∏
∞
(a; q)n = (1 − aq j ) and (a; q)∞ = (1 − aq j ).
j=0 j=0

Let p(n) be the number of partitions of n. From (1.4), it is easy to obtain the following
recursive formula for p(n):

∑
∞ ( ( ) ( ))
3j 2 + j 3j 2 + 5j
(−1) p n − j
−p n− −1 = 0.
j=0
2 2

Andrews and Merca [4] showed that the following truncated Euler’s pentagonal number
theorem
1 ∑
k−1
(−1) k−1
(−1)j q j(3j+1)/2 (1 − q 2j+1 )
(q; q)∞ j=0

has non-negative coefficients. This leads to the following inequality on p(n),

∑
k−1 ( ( ) ( ))
3j 2 + j 3j 2 + 5j
(−1) k−1
(−1) p n −
j
−p n− −1 ≥ 0,
j=0
2 2

where n ≥ 1 and k ≥ 1. They also conjectured that for positive integers k, R, S with
k ≥ 1 and 1 ≤ S < R/2, the following truncated Jacobi’s triple product identity

(−1)k−1 ∑
k−1
(−1)j q Rj(j+1)/2−Sj (1 − q (2j+1)S ) (1.5)
(q S ; q R )∞ (q R−S ; q R )∞ (q R ; q R )∞ j=0

has non-negative coefficients. When R = 3 and S = 1, this truncated identity is the

truncated Euler’s pentagonal number theorem and has been proved by Andrews and
Merca [4]. Guo and Zeng [11] showed this conjecture holds when R = 4 and S = 1.
Mao [14] showed this conjecture is valid by exploiting an algebraic method . Yee [20]

2
gave a combinatorial proof of this conjecture with the aid of Wright’s bijection [19] for
the Jacobi’s triple product identity.
In [11], Guo and Zeng considered the truncated Jacobi’s identity. It is easy to see that
Jacobi’s identity (1.1) can also be written as

1 ∑
∞
j+1

3
(−1)j (2j + 1)q ( 2 ) = 1. (1.6)
(q; q)∞ j=0

Define t(n) to be the number of partitions of n into three kinds of parts, then
∑ 1
t(n)q n = .
n≥1
(q; q)3∞

Thus, from (1.6), it is easy to see that for n ≥ 1,

∑
∞
(−1)j (2j + 1)t(n − j(j + 1)/2) = 0.
j=0

Guo and Zeng [11] conjectured that for n, k ≥ 1,

∑
k
(−1) k
(−1)j (2j + 1)t(n − j(j + 1)/2) ≥ 0. (1.7)
j=0

In other words, the following truncated Jacobi’s identity

(−1)k ∑
k
j+1
j ( 2 )
(−1) q (2j + 1)
(q; q)3∞ j=0

has non-negative coefficients. Mao [14] gave an algebraic proof of this conjecture.
In this paper, we consider the following bilateral truncated Jacobi’s identity

1 ∑
b
j+1

3
(−1)j jq ( 2 )
(q; q)∞ j=a

and give a sufficient condition so that this bilateral truncated series has non-negative
coefficients. More specifically, we have
Theorem 1.1. Given any two integers a ≤ b, define sg(a) := 1 if a ≥ 0 and sg(a) := −1
if a < 0. If (−1)a+b sg(a − 1)sg(b) = 1, then the coefficient of q n in

(−1)a ∑b
sg(a − 1) (−1) j
jq (j+1
2 ) (1.8)
(q; q)3∞ j=a

is non-negative
{ {for n ≥ 1 and is positive for a ̸= 0 or b ̸= 0 and
(a+ sg(a)+1 ) (b+ sg(b)+3 )}}
n ≥ max 1, min 2
2 , 2
2 .

3
 Observe that the coefficient of q n in (1.8) is equal to

∑
b
(−1) sg(a − 1)
a
(−1)j jt(n − j(j + 1)/2).
j=a

Hence, Theorem 1.1 is equivalent to the following corollary.

Corollary 1.2. When (−1)a+b sg(a − 1)sg(b) = 1 and for n ≥ 1, there holds

∑
b
(−1) sg(a − 1)
a
(−1)j jt(n − j(j + 1)/2) ≥ 0
j=a

{ {( sg(a)+1 ) ( sg(b)+3 )}}

with strict inequality if a ̸= 0 or b ̸= 0 and n ≥ max 1, min a+ 2 2 , b+ 2 2 .

Note that Guo and Zeng’s conjecture (1.7) is the special case of Corollary 1.2 with
a = −k − 1 and b = k.
As an example of Corollary 1.2, let a = −2 and b = 3, we obtain

t(n) − 3t(n − 1) + 2t(n − 3) − 3t(n − 6) ≤ 0.

Another example, let a = −3 and b = 4, we get

t(n) − 3t(n − 1) + 5t(n − 3) − 3t(n − 6) + 4t(n − 10) ≥ 0.

This paper is organized as follows. In Section 2, we first recall the definitions of syn-
chronized partitions and rooted synchronized partitions and give the generating functions
for rooted synchronized partitions. Then we state a bijection between the set of degen-
erate rooted synchronized partitions and the set of synchronized partitions without the
zero part and an involution on the set of non-degenerate rooted synchronized partitions.
Finally, we restrict these two bijections to the set of rooted synchronized partitions whose
discrepancy lies in the interval [a, b]. Thus, we obtain a partition identity, which plays an
important role in the proof of Theorem 1.1. In Section 3, we give a combinatorial proof
of Theorem 1.1 based on the partition identity in Section 2.

2 Synchronized partitions
The notions of synchronized partitions and rooted synchronized partitions were first intro-
duced in [5] which are used to give a combinatorial proof of finite form of Jacobi’s identity.
Assume that α = (α1 , α2 , . . . , αr ) is a partition with distinct parts and β = (β1 , β2 , . . . , βs )
is also a partition with distinct parts under the assumption that the last part βs may be
zero. Then a synchronized partition is a representation of (α, β) as a two-row array such

4
that some ∗ symbols may be added at the end of α or β so that they are of the same length
depending on which is of smaller length. We may denote a synchronized partition with
underlying partitions α and β by S(α, β). The difference r − s is called the discrepancy
of the synchronized partition. A synchronized partition with a positive discrepancy k can
be represented as follows:
( )
α1 α2 · · · αs αs+1 · · · αs+k
S(α, β) =
β1 β2 · · · βs ∗ ∗ ∗

and a synchronized partitions with a negative discrepancy −k (k > 0) can be represented

as follows: ( )
α1 α2 · · · αr ∗ ∗ ∗
S(α, β) = .
β1 β2 · · · βr βr+1 · · · βr+k
A synchronized partition with zero discrepancy can be simply represented as a two-row
array without any star added, which is referred to as a Frobenius partition, see Andrews
[1], Corteel and Lovejoy [6]. A rooted synchronized partition is defined as a synchronized
partition with a distinguished star symbol, which we denote by ∗¯. Clearly, a rooted
synchronized partition has an underlying synchronized partition with nonzero discrepancy.
To distinguish a synchronized partition, we may denote a rooted synchronized partition
with underlying partitions α and β by S̄(α, β).
For example, there are five rooted synchronized partitions of 2 :
( )( )( )( )( )
2 1 ∗¯ ∗¯ ∗¯ ∗ ∗ ∗¯
.
∗¯ 1 0 2 2 0 2 0

It is easy to see that the generating function of synchronized partitions equals

(−q; q)∞ (−1; q)∞ , (2.1)

and the generating function of synchronized partitions without the zero part equals

(−q; q)∞ (−q; q)∞ . (2.2)

On the other hand, the generating function of synchronized partitions with a non-negative
discrepancy k equals
k+1
q( 2 )
, (2.3)
(q; q)∞
and the generating function of synchronized partitions with a negative discrepancy −k
equals
−k+1 k
q( 2 ) q (2)
= . (2.4)
(q; q)∞ (q; q)∞

The generating functions (2.3) and (2.4) can be deduced from the Wright’s bijection
[19] for the Jacobi’s triple product identity. Let us recall that a synchronized partition

5
S(α, β) with a discrepancy k can be represented as follows: put r solid circles on the
diagonal, where r is equal to the number of parts in α. Then for j = 1, 2, . . . , r, put
αj − 1 circles in row j to the right of the diagonal and βj circles in column k + j below the
diagonal.
( For example,
) Figure
( 2.1 gives the representations
) of the synchronized partitions
7 5 4 3 2 1 6 5 4 ∗ ∗ ∗
and .
6 5 2 0 ∗ ∗ 7 5 4 3 2 1

Figure 2.1: The Wright’s Representations.

( )
Let us consider the case discrepancy k > 0. Note that k+1 2
is the generating function
k+1
( )
of partitions (k, k − 1, · · · , 1). Hence q 2 /(q; q)∞ is the generating function for a pair
of partitions (Tk , λ), where Tk = (k, k − 1, · · · , 1) and λ is a partition with no restriction.
Then Wright’s bijection [19] as illustrated by Figure 2.2 yields the generating function
(2.3).

 -

Figure 2.2: The Wright’s Bijection for k ≥ 0.

−k+1 k
The case −k < 0 is similar to the previous situation. Notice that q ( 2 ) = q (2) is
−k+1
the generating function of partition (k − 1, . . . , 1). Thus q ( 2 ) /(q; q)∞ is the generating
function for a pair of partitions (Tk−1 , λ), where Tk−1 = (k −1, · · · , 1) and λ is a partition
with no restriction. Then Wright’s bijection yields the generating function (2.4).

6
  -

Figure 2.3: The Wright’s Bijection for k < 0.

Let us define the sign of a synchronized partition as (−1)δ(S) , where δ(S) is the number
of stars in S(α, β) and define a sign of a rooted synchronized partition S̄(α, β) as (−1)δ(S̄) ,
where δ(S̄) is the number of stars in S̄(α, β) under the assumption that a star with the
bar in the top row is not counted. In other words, the sign of a rooted synchronized
partition equals (−1)k sg(k), where k is the discrepancy.
From (2.3) and (2.4), it is easy to show that the generating function of synchronized
partitions S(α, β) with sign (−1)δ(S) equals

1 ∑
+∞
k+1
(−1)k q ( 2 ) , (2.5)
(q; q)∞ k=−∞

and the generating function of rooted synchronized partitions S̄(α, β) with sign (−1)δ(S̄)
equals
1 ∑
+∞
k+1
(−1)k kq ( 2 ) . (2.6)
(q; q)∞ k=−∞

For any two integers a ≤ b, let Sa,b denote the set of synchronized partitions S(α, β)
whose discrepancy lies in the interval [a, b], and let Ra,b be the set of rooted synchronized
partitions S̄(α, β) whose discrepancy lies in the interval [a, b]. Again, from (2.3) and (2.4),
we have

Lemma 2.1. The generating function of synchronized partitions S(α, β)in Sa,b with sign
(−1)δ(S) equals
∑ 1 ∑
b
k+1
δ(S) |α|+|β|
(−1) q = (−1)k q ( 2 ) , (2.7)
(q; q)∞ k=a
S(α,β)∈Sa,b

and the generating function of rooted synchronized partitions S̄(α, β) in Ra,b with sign
(−1)δ(S̄) equals

7
 ∑ 1 ∑
b
k+1
(−1) δ(S̄) |α|+|β|
q = (−1)k kq ( 2 ) . (2.8)
(q; q)∞ k=a
S̄(α,β)∈Ra,b

A rooted synchronized partition S̄(α, β) is called degenerate if

( )
α1 · · · αs αs+1 · · · αr
S̄(α, β) =
β1 · · · βs ∗¯ ··· ∗
or ( )
α1 · · · αr ∗ ··· ∗ ∗¯
S̄(α, β) = ,
β1 · · · βr βr+1 · · · βs 0
where α1 > α2 > . . . > αr ≥ 1 and β1 > β2 > . . . > βs ≥ 1; otherwise S̄(α, β) is called
non-degenerate.
Next, we give a bijection between the set of degenerate rooted synchronized partitions
and the set of synchronized partitions without the zero part, as stated in [5].

Theorem 2.2. There is a sign preserving bijection ϕ between the set of degenerate rooted
synchronized partitions of n and the set of synchronized partitions of n that do not contain
the zero part.

Proof. For a degenerated rooted synchronized partition S̄(α, β), we can construct a syn-
chronized partition S(α′ , β ′ ) that do not contain the zero part. Let ℓ(λ) denote the number
of parts in an ordinary partition λ. We consider the following two cases.

Case 1: If ℓ(α) > ℓ(β), then delete the bar to the first ‘∗’ on the bottom row.
( ) ( )
α1 · · · αs as+1 · · · αr α1 · · · αs αs+1 · · · αr
←→ .
β1 · · · βs ∗¯ · · · ∗ β1 · · · βs ∗ ··· ∗

Case 2: If ℓ(α) < ℓ(β), then delete a zero part on the bottom row along with a barred
star on the top row.
( ) ( )
α1 · · · αr ∗ · · · ∗ ∗¯ α1 · · · αr ∗ ··· ∗
←→ .
β1 · · · βr βr+1 · · · βs 0 β1 · · · βr βr+1 · · · βs

Clearly, the procedure is reversible and it preserves the signs.

We now state a sign reversing involution on the set of non-degenerate rooted synchro-
nized partitions (see [5, Theorem 3.1]).

Theorem 2.3. There is a sign reversing involution τ on the set of non-degenerate rooted
synchronized partitions of n.

8
Proof. For a non-degenerate rooted synchronized partition S̄(α, β), we proceed to con-
struct a non-degenerate rooted synchronized partition S̄(α′ , β ′ ). We consider the following
two cases.

Case 1: The partition β has a zero part.

– If ℓ(α) > ℓ(β), then replace the zero part by a star ∗.

– If ℓ(α) < ℓ(β), then delete the whole column of the zero part.
Case 2: The partition β has no zero part.

– If ℓ(α) > ℓ(β), then replace the first ‘∗’ on the bottom row by a zero part.
– If ℓ(α) < ℓ(β), then add a zero part along with a star on the top as a column.

The above bijection can be illustrated as follows:

( ) ( )
α1 · · · αs as+1 · · · αr ℓ(α)>ℓ(β) α1 · · · αs αs+1 · · · αr
←
−−−−
−−
−−−→
− ,
β1 · · · 0 ∗ ∗
¯ ∗ β1 · · · ∗ ∗ ∗¯ ∗

( ) ( )
α1 · · · αr ∗ ∗¯ ∗ ∗ ℓ(α)<ℓ(β) α1 · · · αr ∗ ∗¯ ∗
←
−−
−−
−−
−−
−→
− .
β1 · · · βr βr+1 · · · βs−1 0 β1 · · · βr βr+1 · · · βs−1

It is easy to check that the above construction is a sign reversing involution.

Using (2.2) and (2.6) and combining Theorems 2.2 and 2.3, we are led to a combina-
torial proof of Jacobi’s identity

1 ∑∞
k+1
(−1)k kq ( 2 ) = (q; q)2∞ .
(q; q)∞ k=−∞

In [10], Joichi and Stanton also provided two combinatorial proofs of Jacobi’s identity.
In the remainder part of this section, we will restrict the bijection ϕ in Theorem 2.2
and the involution τ in Theorem 2.3 to the set Ra,b . Let Da,b denote the set of degenerate
rooted synchronized partitions in Ra,b and let Na,b denote the set of non-degenerate rooted
synchronized partitions in Ra,b . Obviously,

Ra,b = Da,b ∪ Na,b . (2.9)

We first restrict the bijection ϕ in Theorem 2.2 to the set Da,b . Let S̄(α, β) be a degen-
erate rooted synchronized partition with the discrepancy k and let S(α′ , β ′ ) = ϕ(S̄(α, β))
and its discrepancy is equal to k ′ . From the definition of ϕ, we see that S(α′ , β ′ ) is a
synchronized partition without the zero part. Furthermore, when k > 0, k ′ = k; when
k < 0, k ′ = k +1. Hence, if S̄(α, β) is in Da,b , that is, a ≤ k ≤ b, then a+(1+sg(−a))/2 ≤
k ′ ≤ b + (1 − sg(b))/2 which implies that S(α′ , β ′ ) is in Sa+(1+sg(−a))/2,b+(1−sg(b))/2 . Thus,
we have the following conclusion.

9
Corollary 2.4. The bijection ϕ in Theorem 2.2 is also a sign preserving bijection between
the set of degenerate rooted synchronized partitions of n in Ra,b and the set of synchronized
partitions of n in Sa+(1+sg(−a))/2,b+(1−sg(b))/2 that do not contain the zero part.

Let Za,b denote the set of synchronized partitions of n in Sa,b that do not contain the
zero part, from Corollary 2.4, we see that
∑ ∑
(−1)δ(S̄) q |α|+|β| = (−1)δ(S) q |α|+|β| . (2.10)
S̄(α,β)∈Da,b S(α,β)∈Za+(1+sg(−a))/2,b+(1−sg(b))/2

We next restrict the bijection τ in Theorem 2.3 to the set Na,b . It should be noted
that when we apply the involution τ into non-degenerate rooted synchronized partitions
S̄(α, β) with the zero part, whose discrepancy is b, then we delete the zero part of β
to obtain S̄(α′ , β ′ ) which has b + 1 discrepancy. This new rooted synchronized partition
S̄(α′ , β ′ ) is not in Na,b . Besides, if we apply the involution τ into non-degenerate rooted
synchronized partitions S̄(α, β) without the zero part, whose discrepancy is a, we then
add a part 0 to β to get S̄(α′ , β ′ ) which has a − 1 discrepancy. This rooted synchronized
partition S̄(α′ , β ′ ) is also not in Na,b .
Define Ta to be the set of non-degenerate rooted synchronized partitions S̄(α, β) with-
out the zero part, whose discrepancy is a and Tb to be the set of non-degenerate rooted
synchronized partitions S̄(α, β) with the zero part, whose discrepancy is b. Let Ma,b de-
note the set of non-degenerate rooted synchronized partitions in Na,b and not in Ta and
Tb . Obviously, we have
Na,b = Ma,b ∪ Ta ∪ Tb . (2.11)
It is not difficult to verify the following consequence.

Corollary 2.5. The involution τ in Theorem 2.3 is also a sign reversing involution τ on
the set Ma,b .

As a consequence of Corollary 2.5, we obtain the following identity.

∑
(−1)δ(S̄) q |α|+|β| = 0. (2.12)
S̄(α,β)∈Ma,b

10
Hence, from the relation (2.9) and (2.11), we have
∑
(−1)δ(S̄) q |α|+|β|
S̄(α,β)∈Ra,b
∑ ∑
= (−1)δ(S̄) q |α|+|β| + (−1)δ(S̄) q |α|+|β|
S̄(α,β)∈Da,b S̄(α,β)∈Na,b
∑ ∑
= (−1)δ(S̄) q |α|+|β| + (−1)δ(S̄) q |α|+|β|
S̄(α,β)∈Da,b S̄(α,β)∈Ma,b
∑ ∑
+ (−1)δ(S̄) q |α|+|β| + (−1)δ(S̄) q |α|+|β| . (2.13)
S̄(α,β)∈Ta S̄(α,β)∈Tb

Submitting (2.10) and (2.12) into (2.13), we obtain the following identity, which is useful
in the proof of our main theorem.
∑
(−1)δ(S̄) q |α|+|β|
S̄(α,β)∈Ra,b
∑
= (−1)δ(S) q |α|+|β|
S(α,β)∈Za+(1+sg(−a))/2,b+(1−sg(b))/2
∑ ∑
+ (−1)δ(S̄) q |α|+|β| + (−1)δ(S̄) q |α|+|β| . (2.14)
S̄(α,β)∈Ta S̄(α,β)∈Tb

3 Proof of Theorem 1.1

In this section, we give a combinatorial proof of Theorem 1.1. To this end, we first
interpret (1.8) in terms of rooted synchronized partitions, and then show our theorem
holds based on (2.14).
From (2.8) in Lemma 2.1, it is known that (1.8) can be interpreted as follows.

(−1)a ∑b
(−1)a ∑
sg(a − 1) (−1)j
jq (j+1
2 ) = sg(a − 1) (−1)δ(S̄) q |α|+|β| . (3.1)
(q; q)3∞ j=a
(q; q)2
∞
S̄(α,β)∈Ra,b

11
Using (2.14) in the above identity, we find that

(−1)a ∑b
sg(a − 1) (−1) j
jq (j+1
2 )
(q; q)3∞ j=a

(−1)a ∑
= sg(a − 1) (−1)δ(S̄) q |α|+|β|
(q; q)2∞
S̄(α,β)∈Ta

(−1)a ∑
+ sg(a − 1) (−1)δ(S̄) q |α|+|β|
(q; q)2∞
S̄(α,β)∈Tb

(−1)a ∑
+ sg(a − 1) (−1)δ(S) q |α|+|β| . (3.2)
(q; q)2∞
S(α,β)∈Za+(1+sg(−a))/2,b+(1−sg(b))/2

Obviously, (3.2) can be seen as the sum of three terms. We proceed to show that each
sum has non-negative coefficients under the assumption (−1)a+b sg(a − 1)sg(b) = 1. At
first, we have the following conclusion.
Lemma 3.1. Given any integer a, the coefficient of q n in
(−1)a ∑
sg(a − 1) (−1)δ(S̄) q |α|+|β| (3.3)
(q; q)2∞
S̄(α,β)∈Ta

is non-negative for n ≥ 1. Moreover, when a ̸= 0 and a ̸= 1, the coefficient of q n in (3.3)

( sg(a)+1 )
is positive for n ≥ a+ 2 2 .

Proof. Let S̄(α, β) be a non-degenerate rooted synchronized partition in Ta . From the

definition of Ta , we see that (−1)δ(S̄) = (−1)a sg(a). Furthermore, when a ̸= 0, it is clear
to see that sg(a − 1) = sg(a). So

(−1)a ∑ 1 ∑
sg(a − 1) (−1)δ(S̄) q |α|+|β| = q |α|+|β| , (3.4)
(q; q)2∞ (q; q)2∞
S̄(α,β)∈Ta S̄(α,β)∈Ta

which clearly has non-negative coefficients. When a = 0, by the definition of Ta , we see

that there is no element in Ta , so (3.3) is equal to zero.
We next show that when a ̸= 0 and a ̸= 1, the coefficient of q n in (3.4) is positive for
( sg(a)+1 )
n ≥ a+ 2 2 . It is easy to see that the coefficient of q n in (3.4) is equal to the number
of triplets (γ, δ, S̄(α, β)) of n, where γ and δ are ordinary partitions, S̄(α, β) ∈ Ta and

|γ| + |δ| + |α| + |β| = n.

( sg(a)+1 )
So it suffices to show that when a ̸= 0 and a ̸= 1 and for any n ≥ a+ 2 2 , there exists
at least one such triplet (γ, δ, S̄(α, β)). We consider the following two cases.

12
When a > 1, we set γ = δ = ∅ and
( () )
n − a2 a − 1 a − 2 · · · 1
S̄(α, β) = .
∗ ∗ ∗ ··· ∗¯

When a < 0, we set γ = δ = ∅ and

( )
∗¯( ) ∗ ∗ ··· ∗
S̄(α, β) = .
n − −a2
−a − 1 −a − 2 ··· 1
( sg(a)+1 ) (|a|+1)
In either case, it is clear that |γ|+|δ|+|α|+|β| = n and note that n ≥ a+ 2 2 = 2
(|a|)
which implies n − 2 ≥ |a|, so S̄(α, β) ∈ Ta . Thus we completes the proof.

It should be noted that when a = 0 or a = 1, there is no element in Ta which implies

the number of such triplet is equal to zero. So in these two cases, the coefficient of q n in
(3.3) is equal to zero.
The next lemma determines when the second sum of (3.2) has the non-negative coef-
ficients.
Lemma 3.2. For any integers a ≤ b, when (−1)a+b sg(a − 1)sg(b) = 1, the coefficient of
q n in
(−1)a ∑
sg(a − 1) (−1)δ(S) q |α|+|β| (3.5)
(q; q)2∞
S(α,β)∈Tb

is non-negative for n ≥ 1. Moreover, when (−1)a+b sg(a − 1)sg(b) = 1, b ̸= −1 and b ̸= 0,

( sg(b)+3 )
the coefficient of q n in (3.5) is positive for n ≥ b+ 2 2 .

Proof. Let S̄(α, β) be a non-degenerate rooted synchronized partition in Tb . From the

definition of Tb , it is known that (−1)δ(S̄) = (−1)b sg(b). Since (−1)a+b sg(a − 1)sg(b) = 1,
we have
(−1)a ∑ 1 ∑
δ(S̄) |α|+|β|
sg(a − 1) (−1) q = q |α|+|β| . (3.6)
(q; q)2∞ (q; q)2∞
S̄(α,β)∈Tb S̄(α,β)∈Tb

Clearly, the above identity has non-negative coefficients.

We proceed to show that when b ̸= −1 and b ̸= 0, the coefficient of q n in (3.6) is
( sg(b)+3 )
positive for n ≥ b+ 2 2 . It is clear to see that the coefficient of q n in (3.6) can be
combinatorially interpreted as the number of triplets (γ, δ, S̄(α, β)) of n, where γ and δ
are ordinary partitions, S̄(α, β) ∈ Tb and

|γ| + |δ| + |α| + |β| = n.

( sg(b)+3 )
So it suffices to show that when b ̸= −1, b ̸= 0 and for any positive integer n ≥ b+ 2 2 ,
there exists at least one such triplet (γ, δ, S̄(α, β)). There are two following cases. When

13
 (b+2)
b > 0 and n ≥ 2
, we set γ = δ = ∅ and
( ( ) )
n − b+1
2
b b − 1 ··· 1
S̄(α, β) = .
0 ∗ ∗ ··· ∗¯
(b+1)
When b < −1 and n ≥ , we set γ = δ = ∅ and
2
( )
∗¯
( ) ∗ ∗ ··· ∗ ∗
S̄(α, β) = .
n − −b−12
−b − 2 −b − 1 · · · 1 0

In either case, we have |γ| + |δ| + |α| + |β| = n and S̄(α, β) ∈ Tb .

It should be remarked that when b = −1 or b = 0, there is no element in Tb which

implies the number of such triplet is equal to zero. So the coefficient of q n in (3.6) is equal
to zero.
Finally, we aim to show that the third sum of (3.2) has non-negative coefficients when
(−1)a+b sg(a − 1)sg(b) = 1.
Lemma 3.3. For any integers a ≤ b, when (−1)a+b sg(a − 1)sg(b) = 1, the coefficient of
q n in
(−1)a ∑
sg(a − 1) (−1)δ(S) q |α|+|β| (3.7)
(q; q)2∞
S(α,β)∈Z 1+sg(−a) 1−sg(b)
a+ 2 ,b+ 2

is non-negative for n ≥ 1. In particular, when a = 1 or b = −1 or ab = 0 but a2 + b2 > 0,

the coefficient of q n in (3.7) is positive for all n ≥ 1.

Proof. Let On denote the set of triplets (γ, δ, S(α, β)) of n, where

• γ is an ordinary partition;

• δ is an ordinary partition;

• S(α, β) is a synchronized partition without zero parts such that a + 1+sg(−a)

2
≤
ℓ(α) − ℓ(β) ≤ b + 1−sg(b)
2
;

• |γ| + |δ| + |α| + |β| = n.

Let π = (γ, δ, S(α, β)) ∈ On , we associate π with a sign

ω(π) = (−1)a sg(a − 1)(−1)δ(S) . (3.8)

It is easy to see that the coefficients of q n in (3.7) is equal to

∑
ω(π).
π∈On

14
Hence, it suffices to show that the number of triplets in On with positive sign is not less
than the number of triplets in On with negative sign. To this end, we construct a sign-
reversing involution ψ defined on the set On . Let π = (γ, δ, S(α, β)) ∈ On , we proceed to
construct another triplet π ′ = (γ ′ , δ ′ , S(α′ , β ′ )). We consider the following four cases.
Case 1. γ = α = ∅ and β1 ≥ δ1 . We remove β1 from β to get β ′ and add β1 to δ to
generate δ ′ .
Case 2. γ = α = ∅ and β1 < δ1 . We remove δ1 from δ to get δ ′ and add δ1 to β to generate
β ′.
Case 3: γ ̸= ∅ or α ̸= ∅ and α1 ≥ γ1 . Remove α1 from α to get α′ and add α1 to γ to
generate γ ′ .
Case 4: γ ̸= ∅ or α ̸= ∅ and α1 < γ1 . Remove γ1 from γ to generate γ ′ and add γ1 to α
to get α′ .
From the definition of ψ, it is easy to see that the map ψ changes the parity of
ℓ(α) − ℓ(β). Note that (−1)δ(S) = (−1)ℓ(α)−ℓ(β) , so ψ changes the parity of the sign ω(π).
Furthermore, it is not difficult to check that if we apply the map ψ into the following
three kinds of triplets (γ, δ, S(α, β)) in On , then their images are not in the set On .
( )
(1). α = γ = ∅, ℓ(β) = − a + 1+sg(−a)
2
and δ1 > β1 .

(2). γ ̸= ∅ or α ̸= ∅, ℓ(α) − ℓ(β) = a + 1+sg(−a)

2
and γ1 ≤ α1 .
(3). γ ̸= ∅ or α ̸= ∅, ℓ(α) − ℓ(β) = b + 1−sg(b)
2
and γ1 > α1 .
We denote the set of these triplets by En . For π = (γ, δ, S(α, β)) ∈ En and note that
(−1)a+b sg(a − 1)sg(b) = 1, an easy calculation deduces that the sign of π is positive. Let
On denote the set of triplets in On and not in En . Obviously, the map ψ is a sign-reversing
involution on the set On , which implies that
∑
ω(π) = 0.
π∈On

Hence, we have ∑ ∑
ω(π) = ω(π), (3.9)
π∈On π∈En

which is non-negative since w(π) = 1 for any π ∈ En .

To show that (3.7) has positive coefficients when a = 1 or b = −1, or ab = 0 but
a + b2 > 0, we will show that in these three cases, for any positive integer n, there exists
2

at least one triple π = (γ, δ, S(α, β)) in the set En .

When a = 1 or a = 0, but b ̸= 0, let
π = (∅, ∅, S((n), ∅)).

Clearly, π satisfies that γ1 ≤ α1 and ℓ(α) − ℓ(β) = 1 = a + 1+sg(−a)

2
. So π ∈ En .

15
 When b = −1 or b = 0, but a ̸= 0, let

π = ((n), ∅, S(∅, ∅)).

Obviously π satisfies that γ1 > α1 and ℓ(α) − ℓ(β) = 0 = b + 1−sg(b)

2
. So π ∈ En . Thus,
we completes the proof of this lemma.

It should be noted that when a = b = 0, we have a + 1+sg(−a)

2
= 1, but b + 1−sg(b)
2
= 0,
which gives that a + 1+sg(−a)
2
> b + 2 . So there is no element in Za+ 1+sg(−a) ,b+ 1−sg(b) .
1−sg(b)
2 2
n
Thus, the coefficient of q in (3.7) is equal to zero.

Proof of Theorem 1.1. Recall that the bilateral truncated Jacobi’s identity has the fol-
lowing combinatorial interpretation

(−1)a ∑b
sg(a − 1) (−1)j
jq (j+1
2 ) (3.10)
(q; q)3∞ j=a

(−1)a ∑
= sg(a − 1) (−1)δ(S̄) q |α|+|β| (3.11)
(q; q)2∞
S̄(α,β)∈Ta

(−1)a ∑
+ sg(a − 1) (−1)δ(S̄) q |α|+|β| (3.12)
(q; q)2∞
S̄(α,β)∈Tb

(−1)a ∑
+ sg(a − 1) (−1)ℓ(α)−ℓ(β) q |α|+|β| . (3.13)
(q; q)2∞
S(α,β)∈Z 1+sg(−a) 1−sg(b)
a+ 2 ,b+ 2

From Lemmas 3.1, 3.2, and 3.3, we see that when (−1)a+b sg(a − 1)sg(b) = 1, the
coefficient of q n in (3.10) is non-negative for n ≥ 1.
To study the positivity of the coefficient of q n in (3.10), we first let
{ {( ) ( }}
sg(b)+3 )
a + sg(a)+1
2
b + 2
L(a, b) = max 1, min , .
2 2

We consider the following three cases:

Case 1: a ̸= 0, a ̸= 1, b ̸= 0 and b ̸= −1. From Lemma 3.1, we see that the coefficient
( sg(a)+1 )
of q n in (3.11) is positive for all n ≥ a+ 2 2 . By Lemma 3.2, it is known that the
( sg(b)+3 )
coefficient of q n in (3.12) is positive for all n ≥ b+
2
2 . Note that the coefficient of q n
in (3.13) is non-negative, so the coefficient of q in (3.10) is positive for all n ≥ L(a, b).
n

Case 2: a ̸= 0, a ̸= 1, and b = 0 or b = −1. In this case, by Lemma 3.3, we know

that the coefficient of q n in (3.13) is positive for all n ≥ 1. Since the coefficients of q n in

16
(3.11) and (3.12) are non-negative, then for all n ≥ L(a, b), the coefficient of q n in (3.10)
is positive.
Case 3: a = 1 or a = 0 but b ̸= 0. In this case, it follows from Lemma 3.3 that the
coefficient of q n in (3.13) is positive for all n ≥ 1. Hence, for all n ≥ L(a, b), the coefficient
of q n in (3.10) is positive since the coefficients of q n in (3.11) and (3.12) are non-negative.
Therefore, when a ̸= 0 or b ̸= 0, the coefficient of q n in (3.10) is positive for n ≥ L(a, b).
This completes the proof.

Acknowledgments. This work was supported by the 973 Project, the PCSIRT Project
and the National Science Foundation of China.

References
[1] G. E. Andrews, Generalized Frobenius partitions, Mem. Amer. Math. Soc. 49 (1984),
No. 301, iv+, 44 pp.
[2] G. E. Andrews and R. Roy, Ramanujan’s method in q-series congruences, Electron.
J. Combin. 4(2) (1997) #R2.
[3] G.E. Andrews, The Theory of Partitions, Cambridge Mathematical Library, Cam-
bridge University Press, Cambridge, 1998.
[4] G. E. Andrews and M. Merca, The truncated pentagonal number theorem, J. Combin.
Theory Ser. A 119 (2012) 1639–1643.
[5] W. Y.C. Chen and K. Q. Ji, Jacobi’s identity and Synchronized Partitions, arX-
iv:math/0601309.
[6] S. Corteel and J. K. Lovejoy, Frobenius partitions and the combinatorics of Ramanu-
jan’s 1 ψ1 summation, J. Combin. Theory Ser. A 97 (2002) 177–183.
[7] R. Chapman, Partition identities arising from involutions, Australas J. Combin.
27(2003) 285–291.
[8] J. A. Ewell, Completion of a Gaussian derivation, Proc. Amer. Math. Soc. 84(2)
(1982) 311–314.
[9] C. Jacobi, Fundamenta nova theoriae functionum ellipticarum, Mathematische W-
erke, Vol. 1, (1829) 49–239.
[10] J. T. Joichi and D. Stanton, An involution for Jacobi’s identity, Discrete Math. 73
(1989) 261–271.
[11] V. J. W. Guo and J. Zeng, Two truncated identities of Gauss, J. Combin. Theory
Ser. A 120 (2013) 700–707.
[12] M. D. Hirschhorn, Jacobi’s two-square theorem and related identities, Ramanujan J.
3 (1999) 153–158.

17
[13] M. D. Hirschhorn and J. A. Sellers, Two congruences involving 4-cores, The Foata
Festschrift. Electron. J. Combin. 3 (1996) #R10.
[14] R. Mao, Proofs of two conjectures on truncated series, J. Combin. Theory Ser. A 130
(2015) 15–25.
[15] S. Ramanujan, Some propertities of p(n), the number of partitions of n, Proc. Cam-
bridge Philos. Soc. 19 (1919) 207–210.
[16] S. Ramanujan, Congruence properties of partitions, Math. Zeitschr. 9 (1921) 147–
153.
[17] D. Shanks, A short proof of an identity of Euler, Proc. Amer. Math. Soc. 2 (1951)
747–749.
[18] D. Shanks, Two theorems of Gauss, Pacific J. Math. 8 (1958) 609–612.
[19] E. M. Wright, An enumerative proof of an identity of Jacobi, J. London Math. Soc.
40 (1965) 55–57.
[20] A. J. Yee, The truncated Jacobi triple product theorem, J. Combin. Theory Ser. A
130 (2015) 1–14.

18