# http://www.elsevier.

com/locate/aim
Advances in Mathematics 190 (2005) 278–299
Four identities related to third order mock theta
functions in Ramanujan’s lost notebook
Hamza Yesilyurt
Department of Mathematics, University of Illinois, 1409 West Green Street, Urbana, IL 61801, USA
Received 10 February 2003; accepted 17 December 2003
Communicated by Michael Hopkins
Abstract
We prove, for the ﬁrst time, a series of four related identities from Ramanujan’s lost
notebook. The identities are connected with third order mock theta functions.
1. Introduction
In his last letter to Hardy, Ramanujan introduced mock theta functions
[9, pp. 127–131]. Included in this letter were the four third order mock theta
functions:
¯
ff ðqÞ ¼
¸
N
n¼0
q
n
2
ð1 þ qÞ
2
ð1 þ q
2
Þ
2
?ð1 þ q
n
Þ
2
; ð1:1Þ
¯
ffðqÞ ¼
¸
N
n¼0
q
n
2
ð1 þ q
2
Þð1 þ q
4
Þ?ð1 þ q
2n
Þ
; ð1:2Þ
¯
ccðqÞ ¼
¸
N
n¼1
q
n
2
ð1 À qÞð1 À q
3
Þ?ð1 À q
2nÀ1
Þ
; ð1:3Þ
ARTICLE IN PRESS
doi:10.1016/j.aim.2003.12.007
¯wwðqÞ ¼
¸
N
n¼0
q
n
2
ð1 À q þ q
2
Þð1 À q
2
þ q
4
Þ?ð1 À q
n
þ q
2n
Þ
: ð1:4Þ
They satisfy the equations
2
¯
ffðÀqÞ À
¯
ff ðqÞ ¼
¯
ff ðqÞ þ 4
¯
ccðÀqÞ ¼ ðÀq; qÞ
À1
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
n
2
; ð1:5Þ
4¯wwðqÞ À
¯
ff ðqÞ ¼ 3ðq; qÞ
À1
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
3n
2
¸
2
; ð1:6Þ
where we used the standard notation
ða; qÞ
n
¼ ðaÞ
n
:¼ ð1 À aÞð1 À aqÞyð1 À aq
nÀ1
Þ;
ða; qÞ
N
¼
¸
N
n¼0
ð1 À aq
n
Þ; jqjo1:
Watson [10] proved (1.5) and (1.6). Andrews [1] also gave certain generalizations of
(1.5) and (1.6). Third order mock theta functions are related to the rank of a
partition deﬁned by Dyson [5] as the largest part minus the number of parts. Let us
deﬁne Nðm; nÞ as the number of partitions of n with rank m: The generating function
for Nðm; nÞ is given by
¸
N
m¼ÀN
¸
N
n¼0
Nðm; nÞq
n
t
m
¼
¸
N
n¼0
q
n
2
ðtqÞ
n
ðt
À1

n
; jqjo1; jqjojtjoj1=qj: ð1:7Þ
The third order mock theta functions deﬁned by (1.1) through (1.4) can be expressed
in terms of this generating function. Third order mock theta functions and their
applications to the rank are detailed by Fine [7]. A comprehensive literature survey
on mock theta functions is given by Andrews [2].
We prove, for the ﬁrst time, a series of four related identities from Ramanujan’s
lost notebook. These identities are deﬁned and their connections to (1.5) and (1.6)
are given in Section 3. Proofs of these identities are provided in Sections 4–7. In
addition, we will show in Section 8 that one of the identities can be used to prove the
following identity:
ðqÞ
2
N
ðtÞ
N
ðt
À1

N
¼
¸
N
n¼ÀN
ðÀ1Þ
n
q
nðnþ1Þ=2
1 À tq
n
: ð1:8Þ
Identity (1.8) was proved by Evans [6, Eq. (3.1)] following a different approach.
Equality (1.8) is also given in a different form by Ramanujan on p. 59 of the lost
notebook [9]. Partition theory implications of the product
ðqÞ
N
ðtqÞ
N
ðt
À1

N
are discussed in [8].
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 279
2. Deﬁnitions and preliminary results
We ﬁrst recall Ramanujan’s deﬁnitions for a general theta function and some of its
important special cases. Set
f ða; bÞ :¼
¸
p
n¼ÀN
a
nðnþ1Þ=2
b
nðnÀ1Þ=2
; jabjo1: ð2:1Þ
Basic properties satisﬁed by f ða; bÞ include [4, p. 34, Entry 18]
f ða; bÞ ¼ f ðb; aÞ; ð2:2Þ
f ð1; aÞ ¼ 2f ða; a
3
Þ; ð2:3Þ
f ðÀ1; aÞ ¼ 0 ð2:4Þ
and if n is an integer,
f ða; bÞ ¼ a
nðnþ1Þ=2
b
nðnÀ1Þ=2
f ðaðabÞ
n
; bðabÞ
Àn
Þ: ð2:5Þ
If n ¼ 1; (2.5) becomes
f ða; bÞ ¼ af ða
À1
; a
2
bÞ: ð2:6Þ
Two other formulas satisﬁed by f ða; bÞ are [4, p. 46, Entry 30]
f ða; bÞ þ f ðÀa; ÀbÞ ¼ 2f ða
3
b; ab
3
Þ; ð2:7Þ
f ða; bÞ À f ðÀa; ÀbÞ ¼ 2af ðba
À1
; ab
À1
a
4
b
4
Þ: ð2:8Þ
The function f ða; bÞ satisﬁes the well-known Jacobi triple product identity [4, p. 35,
Entry 19]
f ða; bÞ ¼ ðÀa; abÞ
N
ðÀb; abÞ
N
ðab; abÞ
N
: ð2:9Þ
Some important special cases of (2.1) and (2.9) are
jðqÞ :¼ f ðq; qÞ ¼
¸
N
n¼ÀN
q
n
2
¼ ðÀq; q
2
Þ
2
N
ðq
2
; q
2
Þ
N
; ð2:10Þ
cðqÞ :¼ f ðq; q
3
Þ ¼
¸
N
n¼0
q
nðnþ1Þ=2
¼ ðÀq; qÞ
2
N
ðq; qÞ
N
; ð2:11Þ
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 280
f ðÀqÞ :¼ f ðÀq; Àq
2
Þ ¼
¸
N
n¼ÀN
ðÀ1Þ
n
q
nð3nÀ1Þ=2
¼ ðq; qÞ
N
: ð2:12Þ
By using (2.10) and (2.11), and elementary product manipulations, we ﬁnd that
cðÀqÞ ¼
ðq; qÞ
N
ðq
2
; q
4
Þ
N
¼
ðq
2
; q
2
Þ
N
ðÀq; q
2
Þ
N
; ð2:13Þ
jðÀqÞ ¼
ðq; qÞ
N
ðÀq; qÞ
N
: ð2:14Þ
Also, after Ramanujan deﬁne
wðqÞ :¼ ðÀq; q
2
Þ
N
: ð2:15Þ
Other basic properties of the functions j; c; f and w are [4, p. 39, Entry 24]
f ðqÞ
f ðÀqÞ
¼
cðqÞ
cðÀqÞ
¼
wðqÞ
wðÀqÞ
¼
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
jðqÞ
jðÀqÞ

; ð2:16Þ
wðqÞ ¼
f ðqÞ
f ðÀq
2
Þ
¼
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ ﬃ
jðqÞ
cðÀqÞ
3

¼
jðqÞ
f ðqÞ
¼
f ðÀq
2
Þ
cðÀqÞ
; ð2:17Þ
f
3
ðÀq
2
Þ ¼ jðÀqÞc
2
ðqÞ: ð2:18Þ
Combining (2.17) and (2.18), we obtain
f
3
ðÀqÞ ¼ cðqÞj
2
ðÀqÞ: ð2:19Þ
We will frequently use Euler’s identity
ðÀq; qÞ
N
¼ ðq; q
2
Þ
À1
N
: ð2:20Þ
For any real number a; let
f
a
ðqÞ :¼
¸
N
n¼0
q
n
2
ð1 þ aq þ q
2
Þ?ð1 þ aq
n
þ q
2n
Þ
; ð2:21Þ
where jqjo1: For jqjo1; jqjojtjojqj
À1
; let
Gðt; qÞ :¼
¸
N
n¼0
q
n
2
ðtqÞ
n
ðt
À1

n
: ð2:22Þ
We need Euler’s famous generating function for partitions,
Gð1; qÞ ¼ ðq; qÞ
À1
N
: ð2:23Þ
For a proof of (2.23) see [7, p. 13, eq. (12.311)]. We need variations of two
representations for Gðt; qÞ due to Fine [7].
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 281
Lemma 2.1. For jtjo1;
Gðt; qÞ ¼ ð1 À tÞ
¸
N
n¼0
t
n
ðt
À1

n
ð2:24Þ
¼ 1 À t
À1
þ t
À1
¸
N
n¼0
ðtqÞ
n
ðt
À1

n
: ð2:25Þ
Proof. Following Fine [7, pp. 1, Eq. (1.1)], we deﬁne
Fða; b; tÞ :¼
¸
N
n¼0
ðaqÞ
n
ðbqÞ
n
t
n
:
In this notation, Lemma 2.1 can be written as
Gðt; qÞ ¼ ð1 À tÞFð0; t
À1
; tÞ ð2:26Þ
¼ 1 À t
À1
þ t
À1
Fð0; t
À1
; tqÞ: ð2:27Þ
Eq. (2.26) is Eq. (12.3) on p. 13 of [7] with b replaced by t
À1
follows from Eq. (2.4) on p. 2 of [7].
Observe that (2.25) is valid in the region jqjojtjojqj
À1
: Also as noted by Fine [7,
p. 51, Eq. (25.6)], Gðt; qÞ satisﬁes a third order q-difference equation. We sketch a
proof here since it is stated without a proof in [7].
Lemma 2.2. For jqjo1 and jqjojtjoj1=qj; Gðt; qÞ satisﬁes the q-difference equation
1
1 À tq
Gðtq; qÞ þ
qt
3
1 À t
Gðt; qÞ ¼ 1 À qt
2
: ð2:28Þ
Proof. Let
Mðt; qÞ :¼
¸
N
n¼0
ðtqÞ
n
ðt
À1

n
; ð2:29Þ
so that by (2.25),
Gðt; qÞ ¼ 1 À t
À1
þ t
À1
Mðt; qÞ: ð2:30Þ
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 282
Using deﬁnition (2.29) and algebraic manipulation, we obtain
Mðt; qÞ ¼
¸
N
n¼0
ðtqÞ
n
ðt
À1

n
¼
¸
N
n¼0
ðtqÞ
n
ð1 À t
À1
q
nþ1
Þ
ðt
À1

nþ1
¼
¸
N
n¼0
ðtqÞ
n
ðt
À1

nþ1
À t
À1
q
¸
N
n¼0
ðtq
2
Þ
n
ðt
À1

nþ1
¼
¸
N
n¼0
ðtqÞ
n
ðt
À1

nþ1
À t
À1
qð1 À t
À1
Þ
¸
N
n¼0
ðtq
2
Þ
n
ðt
À1
Þ
nþ2
¼
1
tq
¸
N
n¼0
ðtqÞ
nþ1
ðt
À1

nþ1
À
1
ðtq
2
Þ
2
t
À1
qð1 À t
À1
Þ
¸
N
n¼0
ðtq
2
Þ
nþ2
ðt
À1
Þ
nþ2
¼
1
tq
Mðt; qÞ À 1 ð Þ À
1 À t
À1
t
3
q
3
Mðtq; qÞ À 1 À
tq
2
1 À t
À1

: ð2:31Þ
Now, Lemma 2.2 follows from (2.31) together with (2.30) after rearrangement.
For convenience, deﬁne
Vðt; qÞ :¼
1
1 À t
Gðt; qÞ: ð2:32Þ
Lemma 2.2 then takes the following form:
Vðtq; qÞ þ qt
3
Vðt; qÞ ¼ 1 À qt
2
: ð2:33Þ
Observe that
Vðt
À1
; qÞ ¼ ÀtVðt; qÞ: ð2:34Þ
The basic property (2.34) will be used many times in the sequel without comment.
The partial fraction decomposition of Vðt; qÞ is given by [8, Eq. (7.10)]
Vðt; qÞ ¼ 1 þ
t
ðqÞ
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
3nðnþ1Þ=2
1 À tq
n
: ð2:35Þ
We will need the following lemma due to Atkin and Swinnerton-Dyer [3].
Lemma 2.3. Let q; 0oqo1; be ﬁxed. Suppose that WðzÞ is an analytic function of z;
except for possibly a ﬁnite number of poles, in every region, 0oz
1
pjzjpz
2
:
If
WðzqÞ ¼ Az
k
WðzÞ
for some integer k (positive, zero, or negative) and some constant A; then either WðzÞ
has k more poles than zeros in the region jqjojzjp1; or WðzÞ vanishes identically.
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 283
3. Four identities of Ramanujan
We now offer the four identities from Ramanujan’s lost notebook that we plan to
prove.
Entry 3.1 (Ramanujan [9, p. 2, no. 3]). Suppose that a and b are real with a
2
þ b
2
¼
4: Then, if f
a
ðqÞ is deﬁned by (2.21),
b À a þ 2
4
f
a
ðÀqÞ þ
b þ a þ 2
4
f
Àa
ðÀqÞ À
b
2
f
b
ðqÞ
¼
ðq
4
; q
4
Þ
N
ðÀq; q
2
Þ
N
¸
N
n¼1
1 À bq
n
þ q
2n
1 þ ða
2
b
2
À 2Þq
4n
þ q
8n
: ð3:1Þ
If we take a ¼ 0 and b ¼ 2; then, by using (2.14) and elementary product
manipulations, we see that (3.1) reduces to (1.5) in the notation of (2.10) as follows:
2
¯
ffðÀqÞ À
¯
ff ðqÞ ¼ ðÀqÞ
À1
N
jðÀqÞ:
Entry 3.2 (Ramanujan [9, p. 2, no. 4]). If a and b are real with a
2
þ ab þ b
2
¼ 3; then
ða þ 1Þf
Àa
ðqÞ þ ðb þ 1Þf
Àb
ðqÞ À ða þ b À 1Þf
aþb
ðqÞ
¼ 3
ðq
3
; q
3
Þ
2
N
ðq; qÞ
N
¸
N
n¼1
1
1 þ abða þ bÞq
3n
þ q
6n
: ð3:2Þ
In (3.2), take a ¼ b ¼ 1 and use (2.14); then one obtains (1.6) in the notation of
(2.10) as
4¯wwðqÞ À
¯
ff ðqÞ ¼ 3ðqÞ
À1
N
j
2
ðÀq
3
Þ:
We changed the notation that Ramanujan used in the left-hand side of the next entry
to avoid confusion. Also note that the series on the right side below is f
ﬃﬃ
3
p
ðqÞ in the
notation of (2.21).
Entry 3.3 (Ramanujan [9, p. 17, no. 5]). With f
a
ðqÞ deﬁned by (2.21),
1 þ
ﬃﬃﬃ
3
p
2
f
À1
ðÀqÞ þ
3 þ
ﬃﬃﬃ
3
p
6
f
1
ðÀqÞ
¼
¸
N
n¼0
q
n
2
ð1 þ
ﬃﬃﬃ
3
p
q þ q
2
Þ?ð1 þ
ﬃﬃﬃ
3
p
q
n
þ q
2n
Þ
þ
2
ﬃﬃﬃ
3
p cðÀqÞ
ðq
4
; q
4
Þ
N
ðq
6
; q
6
Þ
N
¸
N
n¼1
1
1 þ
ﬃﬃﬃ
3
p
q
n
þ q
2n
: ð3:3Þ
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 284
Entry 3.4 (Ramanujan [9, p. 17, no. 6]). With
¯
ffðqÞ deﬁned by (1.2),
1
2
ð1 þ e
pi=4
Þ
¯
ffðiqÞ þ
1
2
ð1 þ e
Àpi=4
Þ
¯
ffðÀiqÞ
¼
¸
N
n¼0
q
n
2
ð1 þ
ﬃﬃﬃ
2
p
q þ q
2
Þ?ð1 þ
ﬃﬃﬃ
2
p
q
n
þ q
2n
Þ
þ
1
ﬃﬃﬃ
2
p cðÀqÞðÀq
2
; q
4
Þ
N
¸
N
n¼1
1
1 þ
ﬃﬃﬃ
2
p
q
n
þ q
2n
: ð3:4Þ
Note that the series on the right side above is f
ﬃﬃ
2
p
ðqÞ in the notation of (2.21).
4. Proof of Entry 3.1
Let a ¼ 2 cosðyÞ; b ¼ 2 sinðyÞ; and t ¼ e
iy
: Then, it is easy to verify that
f
a
ðqÞ ¼ GðÀt; qÞ; f
Àa
ðqÞ ¼ Gðt; qÞ; f
b
ðqÞ ¼ Gðit; qÞ ð4:1Þ
and
b À a þ 2
4
¼ À
1 À i
4t
ð1 À itÞð1 À tÞ;
b þ a þ 2
4
¼
1 þ i
4t
ð1 À itÞð1 þ tÞ;
b
2
¼
i
2t
ð1 À t
2
Þ; a
2
b
2
À 2 ¼ À2cosð4yÞ ¼ Àðt
4
þ t
À4
Þ: ð4:2Þ
Using (4.1) and (4.2), we can rewrite (3.1) as
ði À 1Þ
4t
ð1 À itÞð1 À tÞGðÀt; ÀqÞ þ
ð1 þ iÞ
4t
ð1 þ tÞð1 À itÞGðt; ÀqÞ
À
i
2t
ð1 À t
2
ÞGðit; qÞ
¼
ðq
4
; q
4
Þ
N
ðÀitqÞ
N
ðit
À1

N
ðÀq; q
2
Þ
N
ðt
4
q
4
; q
4
Þ
N
ðt
À4
q
4
; q
4
Þ
N
: ð4:3Þ
Multiplying both sides of (4.3) by 1 þ it; we obtain
ði À 1Þð1 À t
4
Þ
4t
1
1 þ t
GðÀt; ÀqÞ À
i
1 À t
Gðt; ÀqÞ þ
i À 1
1 À it
Gðit; qÞ

¼
ð1 þ itÞðq
4
; q
4
Þ
N
ðÀitqÞ
N
ðit
À1

N
ðÀq; q
2
Þ
N
ðt
4
q
4
; q
4
Þ
N
ðt
À4
q
4
; q
4
Þ
N
: ð4:4Þ
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 285
Using deﬁnition (2.32) and dividing both sides of (4.4) by ði À 1Þð1 À t
4
Þ=ð4tÞ; we see
that (3.1) is equivalent to the identity
VðÀt; ÀqÞ À iVðt; ÀqÞ þ ði À 1ÞVðit; qÞ
¼ À2ð1 þ iÞt
ðq
4
; q
4
Þ
N
ð1 þ itÞðÀitqÞ
N
ðit
À1

N
ðÀq; q
2
Þ
N
ð1 À t
4
Þðt
4
q
4
; q
4
Þ
N
ðt
À4
q
4
; q
4
Þ
N
¼ À2ð1 þ iÞt
ðq
4
; q
4
Þ
N
ðÀitÞ
N
ðit
À1

N
ðÀq; q
2
Þ
N
ðt
4
; q
4
Þ
N
ðt
À4
q
4
; q
4
Þ
N
¼ À2ð1 þ iÞt
ðq
4
; q
4
Þ
2
N
f ðit; Àit
À1

ðÀq; q
2
Þ
N
ðq; qÞ
N
f ðÀt
4
; Àt
À4
q
4
Þ
; ð4:5Þ
where in the last step we used the Jacobi triple product identity (2.9). We will verify
that (4.5) is valid for jqjojtjojq
À1
j for any ﬁxed jqjo1: Let
Lðt; qÞ :¼ VðÀt; ÀqÞ À iVðt; ÀqÞ þ ði À 1ÞVðit; qÞ;
Rðt; qÞ :¼ À2ð1 þ iÞt
ðq
4
; q
4
Þ
2
N
f ðit; Àit
À1

ðÀq; q
2
Þ
N
ðq; qÞ
N
f ðÀt
4
; Àt
À4
q
4
Þ
:
The proof of Entry 3.1 will be complete once we show that Rðt; qÞ À Lðt; qÞ 0: This
will be achieved by showing that Rðt; qÞ À Lðt; qÞ satisﬁes a q-difference equation of
the sort stated in Lemma 2.3 and has no poles, thereby, forcing it to vanish
identically.
Note that if we deﬁne kðzÞ :¼ f ðcz; c
À1
z
À1
qÞ; then by (2.6) we have
kðzqÞ
kðzÞ
¼
f ðczq; c
À1
z
À1
Þ
f ðcz; c
À1
z
À1

¼
c
À1
z
À1
f ðcz; c
À1
z
À1

f ðcz; c
À1
z
À1

¼ ðczÞ
À1
: ð4:6Þ
Following the same reasoning of (4.6), we obtain
Rðtq; qÞ
Rðt; qÞ
¼
tq
t
f ðitq; Àit
À1
Þ
f ðit; Àit
À1

f ðÀt
4
q
4
; Àt
À4
Þ
f ðÀt
4
; Àt
À4
q
4
Þ
¼ q
ðitÞ
À1
ðÀt
4
Þ
À1
¼ iqt
3
:
Let us verify now that Lðt; qÞ also satisﬁes the same q-difference equation. To that
end,
Lðtq; qÞ À iqt
3
Lðt; qÞ
¼ VðÀtq; ÀqÞ À iVðtq; ÀqÞ þ ði À 1ÞVðitq; qÞ
À iqt
3
fVðÀt; ÀqÞ À iVðt; ÀqÞ þ ði À 1ÞVðit; qÞg
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 286
¼ fVðÀtq; ÀqÞ À qt
3
Vðt; ÀqÞg À ifVðtq; ÀqÞ þ qt
3
VðÀt; ÀqÞg
þ ði À 1ÞfVðitq; qÞ À iqt
3
Vðit; qÞg
¼ 1 À ðÀqÞt
2
À ið1 À ðÀqÞðÀtÞ
2
Þ þ ði À 1Þð1 À qðitÞ
2
Þ
¼ 1 þ qt
2
À ið1 þ qt
2
Þ þ ði À 1Þð1 þ qt
2
Þ ¼ 0;
where we employed (2.33). Now Lemma 2.3 implies that Rðt; qÞ À Vðt; qÞ either has
at least 3 poles in the region jqjojzjp1; or vanishes identically. But Rðt; qÞ À Vðt; qÞ
has at most 3 poles, namely at t ¼ 1; À1; and Ài in that region, and they are all
removable as we shall demonstrate. It sufﬁces to show that t ¼ 1 is a removable
singularity. Thus,
lim
t-1
ð1 À tÞLðtÞ ¼ lim
t-1
ð1 À tÞ fVðÀt; ÀqÞ À iVðt; ÀqÞ þ ði À 1ÞVðit; qÞg
¼ lim
t-1
ð1 À tÞ
1
1 þ t
GðÀt; ÀqÞ À
i
1 À t
Gðt; ÀqÞ þ
i À 1
1 À it
Gðit; qÞ

¼ À i lim
t-1
Gðt; ÀqÞ ¼ ÀiðÀq; ÀqÞ
À1
N
; ð4:7Þ
by (2.23).
Next, by two applications of (2.9) and (2.20),
lim
t-1
ð1 À tÞRðtÞ
¼ lim
t-1
ð1 À tÞ À2ð1 þ iÞt
ðq
4
; q
4
Þ
2
N
f ðit; Àit
À1

ðÀq; q
2
Þ
N
ðq; qÞ
N
f ðÀt
4
; Àt
À4
q
4
Þ
¸
¼ À2ð1 þ iÞ lim
t-1
ð1 À tÞt
ðq
4
; q
4
Þ
2
N
f ðit; Àit
À1

ðÀq; q
2
Þ
N
ðq; qÞ
N
ðt
4
; q
4
Þ
N
ðt
À4
q
4
; q
4
Þ
N
ðq
4
; q
4
Þ
N
¼ À2ð1 þ iÞ lim
t-1
ð1 À tÞt
ðq
4
; q
4
Þ
N
f ðit; Àit
À1

ðÀq; q
2
Þ
N
ðq; qÞ
N
ð1 À t
4
Þðt
4
q
4
; q
4
Þ
N
ðt
À4
q
4
; q
4
Þ
N
¼ À
ð1 þ iÞðq
4
; q
4
Þ
N
f ði; ÀiqÞ
2ðÀq; q
2
Þ
N
ðq; qÞ
N
ðq
4
; q
4
Þ
N
ðq
4
; q
4
Þ
N
¼ À
ð1 þ iÞf ði; ÀiqÞ
2ðÀq; q
2
Þ
N
ðq; qÞ
N
ðq
4
; q
4
Þ
N
¼ À
ð1 þ iÞðÀi; qÞ
N
ðiq; qÞ
N
ðq; qÞ
N
2ðÀq; q
2
Þ
N
ðq; qÞ
N
ðq
4
; q
4
Þ
N
¼ À
ð1 þ iÞð1 þ iÞðÀiq; qÞ
N
ðiq; qÞ
N
2ðÀq; q
2
Þ
N
ðq
4
; q
4
Þ
N
¼ Ài
ðÀq
2
; q
2
Þ
N
ðÀq; q
2
Þ
N
ðq
4
; q
4
Þ
N
¼ À
i
ðÀq; q
2
Þ
N
ðq
4
; q
4
Þ
N
ðq
2
; q
4
Þ
N
¼ À
i
ðÀq; q
4
Þ
N
ðÀq
3
; q
4
Þ
N
ðq
4
; q
4
Þ
N
ðq
2
; q
4
Þ
N
¼ À
i
ðÀq; ÀqÞ
N
: ð4:8Þ
Hence, by (4.7) and (4.8), Lðt; qÞ À Rðt; qÞ has a removable singularity at t ¼ 1: By
our earlier remarks, this completes the proof of Entry 3.1.
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 287
5. Proof of Entry 3.2
Our proof of Entry 3.2 is similar to our proof of Entry 3.1. Since 3 ¼
a
2
þ ab þ b
2
¼ ða À bÞ
2
þ 3ab ¼ ða þ bÞ
2
À ab; we must have jabjo4: Assume with-
out lost of generality that jajojbj; and let a ¼ 2 cosðyÞ: Solving a
2
þ ab þ b
2
¼ 3 for
b gives b ¼ ÀcosðyÞ8
ﬃﬃﬃ
3
p
sinðyÞ: We will take b ¼ ÀcosðyÞ þ
ﬃﬃﬃ
3
p
sinðyÞ ¼ 2 sinðy À
p=6Þ; since replacing y by Ày gives the other value for b:
Let t ¼ e
iy
and r ¼ e
2pi=3
: Using this parametrization we obtain
a ¼ t þ t
À1
; b ¼ r
À1
t þ rt
À1
; and a þ b ¼ Àrt À r
À1
t
À1
;
which, in turn, implies that
f
Àa
ðqÞ ¼ GðtÞ; f
Àb
ðqÞ ¼ Gðr
À1
tÞ; and f
aþb
ðqÞ ¼ GðrtÞ:
One can easily verify that
a þ 1 ¼
1 À t
3
tð1 À tÞ
; b þ 1 ¼
rð1 À t
3
Þ
tð1 Àr
À1

; and a þ b À 1 ¼ À
r
À1
ð1 À t
3
Þ
tð1 À rtÞ
:
Now, the left side of (3.2) which we recall below, becomes
ða þ 1Þf
Àa
ðqÞ þ ðb þ 1Þf
Àb
ðqÞ À ða þ b À 1Þf
aþb
ðqÞ
¼
1 À t
3
tð1 À tÞ
GðtÞ þ
rð1 À t
3
Þ
tð1 À r
À1

Gðr
À1
tÞ þ
r
À1
ð1 À t
3
Þ
tð1 ÀrtÞ
GðrtÞ
¼
1 À t
3
t
ðVðtÞ þ rVðr
À1
tÞ þ r
À1
VðrtÞÞ:
While the right-hand side of (3.2), after observing that
abða þ bÞ ¼ À2 cosð3yÞ ¼ Àðt
3
þ t
À3
Þ;
reduces to
3ðq
3
; q
3
Þ
2
N
ðq; qÞ
N
ðt
3
q
3
; q
3
Þ
N
ðt
À3
q
3
; q
3
Þ
N
:
Thus, Entry 3.2 is equivalent, by (2.9), to the identity
VðtÞ þ rVðr
À1
tÞ þ r
À1
VðrtÞ ¼
3tðq
3
; q
3
Þ
3
N
f ðÀqÞf ðÀt
3
; Àt
À3
q
3
Þ
: ð5:1Þ
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 288
Let NðtÞ and DðtÞ denote the right and left sides of (5.1), respectively. We will verify that
NðtÞ À DðtÞ satisﬁes the q-difference equation NðtqÞ À DðtqÞ ¼ Àqt
3
ðNðtÞ À DðtÞÞ
without any poles in jqjojtjp1: Then using Lemma 2.3, we conclude that NðtÞ À
DðtÞ 0:
We employ (4.6) with c ¼ À1; and t and q replaced by t
3
and q
3
; respectively, to
deduce that
NðtqÞ
NðtÞ
¼
tq
t
1
ðÀt
3
Þ
À1
¼ Àqt
3
:
Next,
DðtqÞ þ qt
3
DðtÞ
¼ VðtqÞ þ rVðr
À1
tqÞ þ r
À1
VðrtqÞ þ qt
3
fVðtÞ þrVðr
À1
tÞ þ r
À1
VðrtÞg
¼ VðtqÞ þ qt
3
VðtÞ þ rfVðr
À1
tqÞ þ qt
3
Vðr
À1
tÞg þ r
À1
fVðrtqÞ þ qt
3
VðrtÞg
¼ 1 À qt
2
þ rð1 À qðr
À1

2
Þ þ r
À1
ð1 À qðrtÞ
2
Þ
¼ 1 þ r þ r
À1
À qt
2
ð1 þr
À1
þ rÞ ¼ 0;
where we used (2.33). Lemma 2.3 now implies that either NðtÞ À DðtÞ vanishes or has
three more poles than zeros in jqjojtjp1: But NðtÞ À DðtÞ has at most three poles,
namely at t ¼ 1; r; r
À1
; and they are all removable as we demonstrate. It sufﬁces to
show that t ¼ 1 is removable.
By (2.23),
lim
t-1
ð1 À tÞDðtÞ ¼ lim
t-1
ð1 À tÞfVðtÞ þ rVðr
À1
tÞ þ r
À1
VðrtÞg
¼ lim
t-1
ð1 À tÞ
1
1 À t
GðtÞ þ r
1
1 À r
À1
t
Gðr
À1
tÞ þ r
À1
1
1 Àrt
GðrtÞ

¼ lim
t-1
GðtÞ ¼
1
f ðÀqÞ
:
By the Jacobi triple product identity (2.9),
lim
t-1
ð1 À tÞNðtÞ ¼ lim
t-1
ð1 À tÞ
3tðq
3
; q
3
Þ
3
N
f ðÀqÞf ðÀt
3
; Àt
À3
q
3
Þ
¼ lim
t-1
ð1 À tÞ
3tðq
3
; q
3
Þ
2
N
f ðÀqÞð1 À t
3
Þðt
3
q
3
; q
3
Þ
N
ðt
À3
q
3
; q
3
Þ
N
¼
1
f ðÀqÞ
:
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 289
We have shown that NðtÞ À DðtÞ has a removable singularity at t ¼ 1: By our
earlier remarks this completes the proof of Entry 3.2.
6. Proof of Entry 3.3
If a ¼ 1; b ¼
ﬃﬃﬃ
3
p
in Entry 3.1, then
ﬃﬃﬃ
3
p
À 1 þ 2
4
f
1
ðÀqÞ þ
ﬃﬃﬃ
3
p
þ 1 þ 2
4
f
À1
ðÀqÞ À
ﬃﬃﬃ
3
p
2
f
ﬃﬃ
3
p
ðqÞ
¼
ðq
4
; q
4
Þ
N
ðÀq; q
2
Þ
N
¸
N
n¼1
1 À
ﬃﬃﬃ
3
p
q
n
þ q
2n
1 þ q
4n
þ q
8n
:
Multiplying both sides by 2=
ﬃﬃﬃ
3
p
; we ﬁnd that
3 þ
ﬃﬃﬃ
3
p
6
f
1
ðÀqÞ þ
1 þ
ﬃﬃﬃ
3
p
2
f
À1
ðÀqÞ
¼ f
ﬃﬃ
3
p
ðqÞ þ
2
ﬃﬃﬃ
3
p
ðq
4
; q
4
Þ
N
ðÀq; q
2
Þ
N
¸
N
n¼1
1 À
ﬃﬃﬃ
3
p
q
n
þ q
2n
1 þ q
4n
þ q
8n
:
We need to show then that
2
ﬃﬃﬃ
3
p cðÀqÞ
ðq
4
; q
4
Þ
N
ðq
6
; q
6
Þ
N
¸
N
n¼1
1
1 þ
ﬃﬃﬃ
3
p
q
n
þ q
2n
¼
2
ﬃﬃﬃ
3
p
ðq
4
; q
4
Þ
N
ðÀq; q
2
Þ
N
¸
N
n¼1
1 À
ﬃﬃﬃ
3
p
q
n
þ q
2n
1 þ q
4n
þ q
8n
:
ð6:1Þ
Recall that c is deﬁned by (2.11). Now,
ðq
6
; q
6
Þ
N
ðÀq; q
2
Þ
N
¸
N
n¼1
ð1 À
ﬃﬃﬃ
3
p
q
n
þ q
2n
Þð1 þ
ﬃﬃﬃ
3
p
q
n
þ q
2n
Þ
1 þ q
4n
þ q
8n
¼
ðq
6
; q
6
Þ
N
ðÀq; q
2
Þ
N
¸
N
n¼1
1 À q
2n
þ q
4n
1 þ q
4n
þ q
8n
¼
ðq
6
; q
6
Þ
N
ðÀq; q
2
Þ
N
¸
N
n¼1
1
1 þ q
2n
þ q
4n
¼
ðq
6
; q
6
Þ
N
ðq
2
; q
2
Þ
N
ðÀq; q
2
Þ
N
ðq
6
; q
6
Þ
N
¼
ðq
2
; q
2
Þ
N
ðÀq; q
2
Þ
N
¼ cðÀqÞ;
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 290
where in the last step (2.13) is used. Equality (6.1) now follows, and so the proof of
Entry 3.3 is complete.
7. Proof of Entry 3.4
Let a ¼ e
ip=4
: Clearly, using the notation of (2.22), we have
¯
ffðqÞ ¼ Gði; qÞ; and
f
ﬃﬃ
2
p
ðqÞ ¼ GðÀa; qÞ: We can then restate Entry 3.4 as
1 þ a
2
Gði; iqÞ þ
1 þa
À1
2
Gði; ÀiqÞ À GðÀa; qÞ ¼
1
ﬃﬃﬃ
2
p
cðÀqÞðÀq
2
; q
4
Þ
N
ðÀaqÞ
N
ðÀa
À1

N
:
Dividing both sides by ð1 þ aÞ=2 and employing (2.9), we arrive at
Gði; iqÞ þ a
À1
Gði; ÀiqÞ À
2
1 þa
GðÀa; qÞ ¼
ﬃﬃﬃ
2
p
cðÀqÞf ðÀqÞðÀq
2
; q
4
Þ
N
f ða; a
À1

:
ð7:1Þ
If we replace q by iq; (7.1) becomes
Gði; ÀqÞ þ a
À1
Gði; qÞ À 2VðÀa; iqÞ ¼
ﬃﬃﬃ
2
p
cðÀiqÞf ðÀiqÞðq
2
; q
4
Þ
N
f ða; aqÞ
: ð7:2Þ
The following identities will be needed for the remainder of the proof:
f ða; a
À1
qÞf ðÀa; Àa
À1
qÞ ¼ ð1 À iÞðÀq
4
; q
4
Þ
N
f
2
ðÀqÞ; ð7:3Þ
f ða; aqÞf ðÀa; ÀaqÞ ¼ ð1 À iÞðÀq
4
; q
4
Þ
N
f
2
ðÀiqÞ; ð7:4Þ
f ða; a
À1
qÞ ¼ cðiqÞ þ acðÀiqÞ; ð7:5Þ
f ðÀa; Àa
À1
qÞ ¼ cðiqÞ À acðÀiqÞ; ð7:6Þ
f ða; aqÞ ¼ cðÀqÞ þ acðqÞ; ð7:7Þ
f ðÀa; ÀaqÞ ¼ cðÀqÞ À acðqÞ: ð7:8Þ
We now offer proofs for all six identities.
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 291
To prove (7.3) we employ (2.9) to ﬁnd that
f ða; a
À1
qÞf ðÀa; Àa
À1
qÞ ¼ ðÀaÞ
N
ðÀa
À1

N
f ðÀqÞðaÞ
N
ða
À1

N
f ðÀqÞ
¼ ði; q
2
Þ
N
ðÀiq
2
; q
2
Þ
N
f
2
ðÀqÞ
¼ ð1 À iÞðÀq
4
; q
4
Þ
N
f
2
ðÀqÞ:
Clearly, (7.4) is obtained by replacing q by iq in (7.3). Recall that cðqÞ ¼ f ðq; q
3
Þ:
From (2.7) and (2.8),
f ða; a
À1
qÞ þ f ðÀa; Àa
À1

¼ 2f ða
2
q; a
À2
q
3
Þ ¼ 2f ðiq; Àiq
3
Þ ¼ 2cðiqÞ; ð7:9Þ
f ða; a
À1
qÞ À f ðÀa; Àa
À1

¼ 2af ða
À2
q; a
2
q
3
Þ ¼ 2af ðÀiq; iq
3
Þ ¼ 2acðÀiqÞ ð7:10Þ
Equalities (7.9) and (7.10) readily imply (7.5) and (7.6). And ﬁnally we obtain (7.7)
and (7.8) by replacing q by iq in (7.5) and (7.6), respectively.
We now return to (7.2) and use (7.4), (7.8), and (2.13) with q replaced by iq to
deduce that
Gði; ÀqÞ þ a
À1
Gði; qÞ À 2VðÀa; iqÞ
¼
ﬃﬃﬃ
2
p
cðÀiqÞf ðÀiqÞðq
2
; q
4
Þ
N
f ða; aqÞ
¼
ﬃﬃﬃ
2
p
cðÀiqÞf ðÀiqÞðq
2
; q
4
Þ
N
f ðÀa; ÀaqÞ
f ða; aqÞf ðÀa; ÀaqÞ
¼
ﬃﬃﬃ
2
p
cðÀiqÞf ðÀiqÞðq
2
; q
4
Þ
N
ðcðÀqÞ À acðqÞÞ
ð1 À iÞðÀq
4
; q
4
Þ
N
f
2
ðÀiqÞ
¼ a
cðÀiqÞðq
2
; q
4
Þ
N
ðcðÀqÞ À acðqÞÞ
ðÀq
4
; q
4
Þ
N
f ðÀiqÞ
¼ a
ðq
2
; q
4
Þ
N
ðcðÀqÞ À acðqÞÞ
ðÀq
4
; q
4
Þ
N
ðÀq
2
; q
4
Þ
N
¼ a
ðq
2
; q
4
Þ
N
ðcðÀqÞ À acðqÞÞ
ðÀq
2
; q
2
Þ
N
¼ aðq
2
; q
4
Þ
2
N
ðcðÀqÞ À acðqÞÞ: ð7:11Þ
It sufﬁces now to prove (7.11).
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 292
Let
Kðt; qÞ :¼aVðit; iqÞ À aVðÀit; iqÞ þ iVðt; iqÞ þ iVðÀt; iqÞ
þ ð1 À iÞVðÀat; ÀqÞ À ð1 þ iÞVðat; ÀqÞ: ð7:12Þ
The identity,
Kðt; qÞ ¼ À 4a
À1
t
f
3
ðÀq
4
Þf ða
À1
t; at
À1

f ðÀiqÞf ða; a
À1
qÞf ðÀt
4
; Àt
À4
q
4
Þ
À 2ð1 þ iÞt
cðÀq
2
Þf
2
ðÀqÞf ða
À1
t; at
À1

f ðt; t
À1
qÞf ðÀit; it
À1
qÞf ðÀit
2
; it
À2
q
2
Þ
; ð7:13Þ
together with Entry 3.1 will be used to verify (7.11). We will not prove (7.13), because
its proof is very similar to that of (5.1). The q-difference equation satisﬁed by Kðt; qÞ
is Kðtq; qÞ ¼ Àaqt
3
Kðt; qÞ: It then sufﬁces, by Lemma 2.3, to verify that the residues
at four of the six singularities match those of the two representations (7.12) and
(7.13) of Kðt; qÞ: It is easily veriﬁed that t ¼ Àa is a zero for the two representations
(7.12) and (7.13) of Kðt; qÞ: Therefore, one only needs to check the residues at any
three of the six singularities. If we knew the two other zeros whose existence is
guaranteed by Lemma 2.3, we then would be able to reduce the right-hand side of
(7.13) to a single product, but we are unable to determine these two zeros.
Let us deﬁne, by using (4.5),
Eðt; qÞ :¼ VðÀt; ÀqÞ À iVðt; ÀqÞ þ ði À 1ÞVðit; qÞ ð7:14Þ
¼ À2ð1 þ iÞt
ðq
4
; q
4
Þ
2
N
f ðit; Àit
À1

ðÀq; q
2
Þ
N
ðq; qÞ
N
f ðÀt
4
; Àt
À4
q
4
Þ
: ð7:15Þ
We will verify by using (7.12) and (7.14) that
Gði; ÀqÞ þ a
À1
Gði; qÞ À 2VðÀa; iqÞ
¼
1
að1 À iÞ
ðEða; iqÞ þ EðÀa; iqÞÞ þ
1
2
iKða; qÞ þ
1
2
aKða; ÀqÞ: ð7:16Þ
Equalities (7.13) and (7.15) will then be used to verify that (7.16) reduces to the right-
hand side of (7.11), which will complete the proof of Entry 3.4.
Using (7.14), we have
Eðt; qÞ þ EðÀt; qÞ ¼ ð1 À iÞfVðt; ÀqÞ þ VðÀt; ÀqÞ À Vðit; qÞ À VðÀit; qÞg:
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 293
Setting t ¼ a; we ﬁnd that
Eða; qÞ þ EðÀa; qÞ
¼ ð1 À iÞfVða; ÀqÞ þ VðÀa; ÀqÞ À Vðia; qÞ À VðÀia; qÞg
¼ ð1 À iÞfVða; ÀqÞ þ VðÀa; ÀqÞ À VðÀa
À1
; qÞ À Vða
À1
; qÞg
¼ ð1 À iÞfVða; ÀqÞ þ VðÀa; ÀqÞ À aVðÀa; qÞ þ aVða; qÞg:
Replacing q by iq and dividing by að1 À iÞ; we obtain
1
að1 À iÞ
ðEða; iqÞ þ EðÀa; iqÞÞ
¼ a
À1
Vða; ÀiqÞ þ a
À1
VðÀa; ÀiqÞ À VðÀa; iqÞ þ Vða; iqÞ: ð7:17Þ
By (7.12),
Kða; qÞ ¼aVðia; iqÞ ÀaVðÀia; iqÞ þ iVða; iqÞ þ iVðÀa; iqÞ
þ ð1 À iÞVðÀi; ÀqÞ À ð1 þ iÞVði; ÀqÞ
¼aVðÀa
À1
; iqÞ À aVða
À1
; iqÞ þ iVða; iqÞ þ iVðÀa; iqÞ
þ ð1 À iÞVðÀi; ÀqÞ À ð1 þ iÞVði; ÀqÞ
¼iVðÀa; iqÞ þ iVða; iqÞ þ iVða; iqÞ þ iVðÀa; iqÞ
À ið1 À iÞVði; ÀqÞ À ð1 þ iÞVði; ÀqÞ
¼2iVða; iqÞ þ 2iVðÀa; iqÞ À 2iGði; ÀqÞ: ð7:18Þ
Combining (7.17) and (7.18), we ﬁnd that
1
að1 À iÞ
ðEða; iqÞ þ EðÀa; iqÞÞ þ
1
2
iKða; qÞ þ
1
2
aKða; ÀqÞ
¼ a
À1
Vða; ÀiqÞ þ a
À1
VðÀa; ÀiqÞ À VðÀa; iqÞ þ Vða; iqÞ
À Vða; iqÞ À VðÀa; iqÞ þ Gði; ÀqÞ
À a
À1
Vða; ÀiqÞ À a
À1
VðÀa; ÀiqÞ þ a
À1
Gði; qÞ
¼ Gði; ÀqÞ þ a
À1
Gði; qÞ À 2VðÀa; iqÞ:
This proves our ﬁrst claim that (7.16) holds.
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 294
Using (7.13), (2.3), (2.6), and (2.19) with q replaced by q
4
; we ﬁnd that
Kða; qÞ ¼ À 4
f
3
ðÀq
4
Þf ð1; qÞ
f ðÀiqÞf ða; a
À1
qÞf ð1; q
4
Þ
À 2að1 þ iÞ
cðÀq
2
Þf
2
ðÀqÞf ð1; qÞ
f ða; a
À1
qÞf ðÀia; ia
À1
qÞf ð1; q
2
Þ
¼ À 4
f
3
ðÀq
4
ÞcðqÞ
f ðÀiqÞf ða; a
À1
qÞcðq
4
Þ
À 2að1 þ iÞ
cðÀq
2
Þf
2
ðÀqÞcðqÞ
f ða; a
À1
qÞf ða
À1
; aqÞcðq
2
Þ
¼ À 4
f
3
ðÀq
4
ÞcðqÞ
f ðÀiqÞf ða; a
À1
qÞcðq
4
Þ
þ 2ð1 À iÞ
cðÀq
2
Þf
2
ðÀqÞcðqÞ
f
2
ða; a
À1
qÞcðq
2
Þ
¼ À 4
j
2
ðÀq
4
ÞcðqÞ
f ðÀiqÞf ða; a
À1

þ 2ð1 À iÞ
cðÀq
2
Þf
2
ðÀqÞcðqÞf ðÀa; Àa
À1

f
2
ða; a
À1
qÞcðq
2
Þf ðÀa; Àa
À1

:
Using (7.3) and (7.6) above, we deduce that
Kða; qÞ ¼ À 4
j
2
ðÀq
4
ÞcðqÞ
f ðÀiqÞf ða; a
À1

þ 2ð1 À iÞ
cðÀq
2
Þf
2
ðÀqÞcðqÞðcðiqÞ ÀacðÀiqÞÞ
f ða; a
À1
qÞcðq
2
Þð1 À iÞðÀq
4
; q
4
Þ
N
f
2
ðÀqÞ
¼ À 4
j
2
ðÀq
4
ÞcðqÞ
f ðÀiqÞf ða; a
À1

þ 2
cðÀq
2
ÞcðqÞcðiqÞ
f ða; a
À1
qÞcðq
2
ÞðÀq
4
; q
4
Þ
N
À 2a
cðÀq
2
ÞcðqÞcðÀiqÞ
f ða; a
À1
qÞcðq
2
ÞðÀq
4
; q
4
Þ
N
¼ À 4
j
2
ðÀq
4
ÞcðqÞ
f ðÀiqÞf ða; a
À1

þ 2
cðÀq
2
ÞcðqÞf
2
ðq
2
Þ
f ðÀiqÞf ða; a
À1
qÞcðq
2
ÞðÀq
4
; q
4
Þ
N
À 2a
cðÀq
2
ÞcðqÞf
2
ðq
2
Þ
f ðiqÞf ða; a
À1
qÞcðq
2
ÞðÀq
4
; q
4
Þ
N
; ð7:19Þ
where we used (2.17) in the form f ðqÞcðÀqÞ ¼ f
2
ðÀq
2
Þ with q replaced by iq and Àiq;
respectively. But by (2.16),
cðÀq
2
Þf
2
ðq
2
Þ
cðq
2
ÞðÀq
4
; q
4
Þ
N
¼
f ðq
2
Þcðq
2
Þf ðÀq
2
Þ
cðq
2
ÞðÀq
4
; q
4
Þ
N
¼
ðÀq
2
; Àq
2
Þ
N
ðq
2
; q
2
Þ
N
ðÀq
4
; q
4
Þ
N
¼
ðÀq
2
; q
4
Þ
N
ðq
4
; q
4
Þ
N
ðq
2
; q
4
Þ
N
ðq
4
; q
4
Þ
N
ðÀq
4
; q
4
Þ
N
¼
ðq
4
; q
8
Þ
N
ðq
4
; q
4
Þ
2
N
ðÀq
4
; q
4
Þ
N
¼
ðq
4
; q
4
Þ
2
N
ðÀq
4
; q
4
Þ
2
N
¼ j
2
ðÀq
4
Þ; ð7:20Þ
where we used Euler’s identity (2.20), and (2.14). Using (7.20) in (7.19), (2.17) in the
form f ðqÞcðÀqÞ ¼ f
2
ðÀq
2
Þ with q replaced by iq and Àiq; respectively, (7.5), and
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 295
(2.14), we deduce that
Kða; qÞ ¼ À 4
j
2
ðÀq
4
ÞcðqÞ
f ðÀiqÞf ða; a
À1

þ 2
j
2
ðÀq
4
ÞcðqÞ
f ðÀiqÞf ða; a
À1

À 2a
j
2
ðÀq
4
ÞcðqÞ
f ðiqÞf ða; a
À1

¼ À 2
j
2
ðÀq
4
ÞcðqÞ
f ða; a
À1

1
f ðÀiqÞ
þ a
1
f ðiqÞ

¼ À 2
j
2
ðÀq
4
ÞcðqÞ
f ða; a
À1
qÞf
2
ðq
2
Þ
ðcðiqÞ þ acðÀiqÞÞ
¼ À 2
j
2
ðÀq
4
ÞcðqÞ
f
2
ðq
2
Þ
¼ À2
ðq
4
; q
4
Þ
2
N
cðqÞ
ðÀq
4
; q
4
Þ
2
N
ðÀq
2
; Àq
2
Þ
2
N
¼ À 2
ðq
4
; q
4
Þ
2
N
cðqÞ
ðÀq
4
; q
4
Þ
2
N
ðÀq
2
; q
4
Þ
2
N
ðq
4
; q
4
Þ
2
N
¼ À2
cðqÞ
ðÀq
2
; q
2
Þ
2
N
¼ À 2ðq
2
; q
4
Þ
2
N
cðqÞ; ð7:21Þ
where in the last step we used (2.20). Thus, by (7.21),
1
2
iKða; qÞ þ
1
2
aKða; ÀqÞ ¼ Àiðq
2
; q
4
Þ
2
N
cðqÞ À aðq
2
; q
4
Þ
2
N
cðÀqÞ: ð7:22Þ
Let us evaluate now Eða; qÞ: By (2.3), (2.6), and (7.15),
Eða; qÞ ¼ À 2ð1 þ iÞa
ðq
4
; q
4
Þ
2
N
f ðia; ia
À1

ðq; qÞ
N
ðÀq; q
2
Þ
N
f ð1; q
4
Þ
¼ À 2ð1 þ iÞa
ðq
4
; q
4
Þ
2
N
f ðÀa
À1
; ÀaqÞ
ðq; qÞ
N
ðÀq; q
2
Þ
N
f ð1; q
4
Þ
¼ð1 þ iÞ
ðq
4
; q
4
Þ
2
N
f ðÀa; Àa
À1

ðq; qÞ
N
ðÀq; q
2
Þ
N
cðq
4
Þ
: ð7:23Þ
Employ (2.11) and (2.20) to deduce that
ðq
4
; q
4
Þ
2
N
ðq; qÞ
N
ðÀq; q
2
Þ
N
cðq
4
Þ
¼
ðq
4
; q
4
Þ
2
N
ðq
2
; q
2
Þ
N
ðq; q
2
Þ
N
ðÀq; q
2
Þ
N
ðÀq
4
; q
4
Þ
2
N
ðq
4
; q
4
Þ
N
¼
ðq
4
; q
4
Þ
N
ðq
2
; q
2
Þ
N
ðq
2
; q
4
Þ
N
ðÀq
4
; q
4
Þ
2
N
¼
ðq
2
; q
2
Þ
N
ðÀq
2
; q
2
Þ
2
N
ðq
2
; q
2
Þ
N
ðÀq
4
; q
4
Þ
2
N
¼
ðÀq
2
; q
2
Þ
2
N
ðÀq
4
; q
4
Þ
2
N
¼
ðÀq
2
; q
4
Þ
2
N
ðÀq
4
; q
4
Þ
2
N
ðÀq
4
; q
4
Þ
2
N
¼ ðÀq
2
; q
4
Þ
2
N
:
ð7:24Þ
Using (7.24) in (7.23), we obtain
Eða; qÞ ¼ ð1 þ iÞðÀq
2
; q
4
Þ
2
N
f ðÀa; Àa
À1
qÞ: ð7:25Þ
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 296
Similarly, we obtain
EðÀa; qÞ ¼ ð1 þ iÞðÀq
2
; q
4
Þ
2
N
f ða; a
À1
qÞ: ð7:26Þ
Combining (7.25) and (7.26), we arrive at
1
að1 À iÞ
fEða; qÞ þ EðÀa; qÞg
¼
1 þ i
að1 À iÞ
ðÀq
2
; q
4
Þ
2
N
ð f ða; a
À1
qÞ þ f ðÀa; Àa
À1
qÞÞ ¼ 2aðÀq
2
; q
4
Þ
2
N
cðiqÞ;
ð7:27Þ
by (7.9). Finally, replacing q by iq in (7.27), we deduce that
1
að1 À iÞ
fEða; iqÞ þ EðÀa; iqÞg ¼ 2aðq
2
; q
4
Þ
2
N
cðÀqÞ: ð7:28Þ
Adding (7.22) and (7.28) together, we ﬁnd that (7.16) reduces to the right-hand side
of (7.11), i.e.,
1
2
iKða; qÞ þ
1
2
aKða; ÀqÞ þ
1
að1 À iÞ
fEða; iqÞ þ EðÀa; iqÞg
¼ Àiðq
2
; q
4
Þ
2
N
cðqÞ À aðq
2
; q
4
Þ
2
N
cðÀqÞ þ 2aðq
2
; q
4
Þ
2
N
cðÀqÞ
¼ aðq
2
; q
4
Þ
2
N
ðcðÀqÞ À acðqÞÞ:
This completes the veriﬁcation of (7.11), since we have already veriﬁed (7.16). Hence,
the proof of Entry 3.4 is complete.
8. Proof of 1.8
Let us recall Eqs. (2.35) and (5.1), which is the equivalent form of Entry 3.2. Thus,
VðtÞ þ rVðr
À1
tÞ þ r
À1
VðrtÞ ¼
3tðq
3
; q
3
Þ
3
N
ðqÞ
N
f ðÀt
3
; Àt
À3
q
3
Þ
; ð8:1Þ
VðtÞ ¼ 1 þ
t
ðqÞ
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
3nðnþ1Þ=2
1 À tq
n
; ð8:2Þ
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 297
where r ¼ e
2pi=3
: Using (8.2) in (8.1), we obtain
3tðq
3
; q
3
Þ
3
N
ðqÞ
N
f ðÀt
3
; Àt
À3
q
3
Þ
¼ 1 þ
t
ðqÞ
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
3nðnþ1Þ=2
1 À tq
n
þ r þ
rt
ðqÞ
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
3nðnþ1Þ=2
1 À r
À1
tq
n
þ r
À1
þ
r
À1
t
ðqÞ
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
3nðnþ1Þ=2
1 À rtq
n
¼
t
ðqÞ
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
3nðnþ1Þ=2
1
1 À tq
n
þ
r
1 À r
À1
tq
n
þ
r
À1
1 Àrtq
n

¼
3t
ðqÞ
N
¸
N
n¼ÀN
ðÀ1Þ
n
q
3nðnþ1Þ=2
1 À t
3
q
3n
:
Then, we have
ðq
3
; q
3
Þ
3
N
f ðÀt
3
; Àt
À3
q
3
Þ
¼
¸
N
n¼ÀN
ðÀ1Þ
n
q
3nðnþ1Þ=2
1 À t
3
q
3n
: ð8:3Þ
Now, (1.8) follows if one replaces q
3
by q and t
3
by t; respectively, and employs (2.9)
in (8.3).
Acknowledgments
I would like to thank my adviser Professor Bruce C. Berndt for his guidance and
assistance at all stages of this work.
References
[1] G.E. Andrews, On basic hypergeometric series, mock theta functions, and partitions. I, Quart.
J. Math. Oxford Ser. 17 (2) (1966) 64–80.
[2] G.E. Andrews, Mock theta functions, Theta functions—Bowdoin 1987, Part 2, Proceedings of
Symposia in Pure Mathematics Vol. 49, Brunswick, ME, 1987, pp. 283–298.
[3] A.O.L. Atkin, P. Swinnerton-Dyer, Some properties of partitions, Proc. London. Math. Soc. 4 (4)
(1954) 84–106.
[4] B.C. Berndt, Ramanujan’s Notebooks, Part III, Springer, New York, 1991.
[5] F.J. Dyson, Some guesses in the theory of partitions, Vol. 8, Eureka, Cambridge, 1944,
pp. 10–15.
[6] R.J. Evans, Generalized Lambert series, in: B.C. Berndt, H.G. Diamond, A.J. Hildebrand (Eds.),
Analytic Number Theory Allerton Park, IL, 1995), Vol. 1, Birkha¨ user, Boston, 1996, pp. 357–370.
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 298
[7] N.J. Fine, Basic Hypergeometric Series and Applications, Mathematical Surveys and Monographs,
Vol. 27, American Mathematical Society, Providence, RI, 1988.
[8] F.G. Garvan, New combinatorial interpretations of Ramanujan’s partition congruences mod 5,7 and
11, Trans. Amer. Math. Soc. 305 (1988) 47–77.
[9] S. Ramanujan, The Lost Notebook and Other Unpublished Papers, Narosa, New Delhi, 1988.
[10] G.N. Watson, The ﬁnal problem: an account of the mock theta functions, J. London Math. Soc.
11 (1936) 55–80.
ARTICLE IN PRESS
H. Yesilyurt / Advances in Mathematics 190 (2005) 278–299 299