Your Hit Parade:

The Top Ten Most

Fascinating Formulas in
Ramanujans Lost Notebook
George E. Andrews and Bruce C. Berndt

t 7:30 on a Saturday evening in March

1956, the first author sat down in
an easy chair in the living room of
his parents farm home ten miles east
of Salem, Oregon, and turned the TV
channel knob to NBCs Your Hit Parade to find
out the Top Seven Songs of the week, as determined by a national survey and sheet music
sales. Little did this teenager know that almost
exactly twenty years later, he would be at Trinity
College, Cambridge, to discover one of the biggest
hits in mathematical history, Ramanujans Lost
Notebook. Meanwhile, at that same hour on that
same Saturday night in Stevensville, Michigan,
but at 9:30, the second author sat down in an
overstuffed chair in front of the TV in his parents farm home anxiously waiting to learn the
identities of the Top Seven Songs, sung by Your
Hit Parade singers, Russell Arms, Dorothy Collins
(his favorite singer), Snooky Lanson, and Gisele
MacKenzie. About twenty years later, that authors
life would begin to be consumed by Ramanujans
mathematics, but more important than Ramanujan to him this evening was how long his parents
would allow him to stay up to watch Saturday
night wrestling after Your Hit Parade ended.
George E. Andrews is professor of mathematics at Pennsylvania State University, University Park. His email address is andrews@math.psu.edu.
Bruce C. Berndt is professor of mathematics at the University of Illinois at Urbana-Champaign. His email address is berndt@math.uiuc.edu.
Partially supported by National Science Foundation Grant
DMS 0457003.
Research partially supported by grant MDA H98230-071-0088 from the National Security Agency.

18

Just as the authors anxiously waited for the

identities of the Top Seven Songs of the week years
ago, readers of this article must now be brimming
with unbridled excitement to learn the identities of
the Top Ten Most Fascinating Formulas from Ramanujans Lost Notebook. The choices for the Top
Ten Formulas were made by the authors. However,
motivated by the practice of Your Hit Parade, but
now extending the survey outside the boundaries
of the U.S., we have taken an international survey to determine the proper order of fascination
and amazement of these formulas. The survey
panel of 34 renowned experts on Ramanujans
work includes Nayandeep Deka Baruah, S. Bhargava, Jonathan Borwein, Peter Borwein, Douglas
Bowman, David Bradley, Kathrin Bringmann, Song
Heng Chan, Robin Chapman, Youn-Seo Choi, Wenchang Chu, Shaun Cooper, Sylvie Corteel, Freeman
Dyson, Ronald Evans, Philippe Flajolet, Christian
Krattenthaler, Zhi-Guo Liu, Lisa Lorentzen, Jeremy Lovejoy, Jimmy McLaughlin, Steve Milne, Ken
Ono, Peter Paule, Mizan Rahman, Anne Schilling,
Michael Schlosser, Andrew Sills, Jaebum Sohn,
S. Ole Warnaar, Kenneth Williams, Ae Ja Yee,
Alexandru Zaharescu, and Doron Zeilberger. A
summary of their rankings can be found in the
last section of our paper. Just as the songs changed
weekly on Your Hit Parade, the choices for the Top
Ten Most Fascinating Formulas also change from
week to week. The reason is simple. There are
so many fascinating results in the lost notebook
that thinking about a particular formula during
the week will naturally generate increased appreciation for it, if not increased understanding, and
vault it into the Top Ten, meanwhile shoving a not
so recently contemplated formula out of the Top
Ten.

Notices of the AMS

Volume 55, Number 1

It was the practice of Your Hit Parade to present

the songs in reverse order of popularity to build
up excitement about the identity of the Number
One Song of the Week. (Of course, after the identification of the Number Two Song, you would
have had to have slept through the program if you
did not then immediately deduce the identity of
the Number One Song to be announced about five
minutes later.) However, it is the fate of popular
songs to lose their popularity and fade off the
charts. In fact, the immense popularity of Your Hit
Parade, beginning on radio in 1935 and switching
to TV in 1950, began to rapidly decline with the
advent of Rock n Roll. On April 24, 1959, Your
Hit Parade aired for the last time. It was said that
Snooky Lanson could not compete with Elvis Presley in singing the latters famous hit song, Hound
Dog. But although popular songs may fade away,
Ramanujans theorems do not fade away. They
will remain as fascinating to mathematicians of
future generations as they are to mathematicians
of our present generation.
So here they areThe Top Ten Most Fascinating
Formulas From Ramanujans Lost Notebookas
determined by the authors, ordered by our panel of experts, and presented in reverse order of
popularity among the members of the panel.

Notation
We employ the familiar notation in the theory of
q-series. For each nonnegative integer n, let
(1)

(a; q)n := (1 a)(1 aq) (1 aq n1 ),

(a; q) := lim (a; q)n ,

n

Theorem 1 (Corrected, p. 200). Define, for

any real numbers and s and for any
complex numbers , , , and , such that
Re( + + + ) > 3,
s () :=
e(n+s)i
.
( + s + n) ( s n) ( + s + n) ( s n)

January 2008

When a calculus student is asked to differentiate a quotient, she immediately turns to the
familiar quotient rule from differential calculus.
In this instance, as expected, the square of the denominator appears in the differentiated formula,
and in the numerator she obtains the difference of
two products of bilateral hypergeometric series.
But, according to (3), we see that Ramanujan had
the remarkable insight to see that this difference
of products of series could be evaluated in closed
form!
To explain the origin of (3), we review some
background about bilateral hypergeometric series.
For every integer n, define, in contrast to the notation in the previous section and elsewhere in this
paper,
(a + n)
(a)n :=
.
(a)
The bilateral hypergeometric series p Hp is defined for complex parameters a1 , a2 , . . . , ap and
b1 , b2 , . . . , bp by
#
"
a1 , a2 , . . . , ap ;
z :=
p Hp
b1 , b2 , . . . , bp ;

X
(a1 )n (a2 )n (ap )n n
z .
(b1 )n (b2 )n (bp )n
n=

The series p Hp is said to be well-poised if

On page 200 of his lost notebook [62], Ramanujan

offers two results on certain bilateral hypergeometric series. The second follows from a theorem
of J. Dougall [38], and we will not discuss it here.
The first gives a formula for the derivative of a
quotient of two certain bilateral hypergeometric
series. Ramanujans formula needs to be slightly
corrected, but what is remarkable is that such a
formula exists!

n=

t2 () ( + 1) ( + 1) ( + 1) ( + 1)

Re(b1 + b2 + + bp a1 a2 ap ) > 1.

|q| < 1.

No. 10. A Remarkable Bilateral Hypergeometric Sum

d s ()
=
d t ()

+++4
i sin{ (s t)} 2 sin( 21 )
ei( )(++2s+2t)/2

(3)

With the use of DAlemberts ratio test, it can

be checked that p Hp converges only for |z| = 1,
provided that [65, p. 181, eq. (6.1.1.6)]

with the convention that (a; q)0 := 1. Set

(2)

Then

a1 + b 1 = a2 + b 2 = = ap + b p .

In 1907, Dougall [38] showed that a well-poised

5 H5 series could be evaluated at z = 1. We do not
provide here his evaluation in terms of gamma
functions, but simply emphasize its importance
in proving Theorem 1. Dougall [38] also evaluated
in closed form a general 2 H2 at z = 1, from which
one can deduce the following bilateral form of the
binomial series theorem. If a and c are complex
numbers with Re(c a) > 1 and z is a complex
number with |z| = 1, then
#
"
(1 z)ca1 (1 a) (c)
a;
z =
(4) 1 H1
.
c;
(z)c1
(c a)
For further remarks on (4), see the paper by the
second author and W. Chu [26].
The proof of Theorem 1 now proceeds as follows. Use the familiar quotient rule to differentiate

Notices of the AMS

d s ()
.
d t ()
19

Combine the two products of series in the numerator into one double series. After some rearrangement, we find that the inner series of
the resulting double series is surprisingly a wellpoised 5 H5 , which can be summed by Dougalls
theorem. There remains a single bilateral sum,
which we can evaluate by using (4). Theorem 1
now follows. See [26] for complete details.

Theorem 2. With f (q), (q), and R(q) defined

by (6), (5), and (7), respectively, and with = ( 5+

1)/2,
Zq 2
f (t)f 2 (t 5 )

(8)
dt
53/4
t
0
Z /2
d
q
=2
5/2
1
cos ((R(q))
) 1 5 53/2 sin2
(9)

No. 9. Some Challenging Integrals for Your

Calculus Students
In his lost notebook [62], Ramanujan records approximately 15 equalities between two different
types of integrals. On the left-hand sides are integrals of Dedekind eta-functions, (z), and on the
right-hand sides are differences of two incomplete
integrals of the first kind. To describe these identities, of a type never before seen, we introduce
Ramanujans notations for theta functions.
Define, following Ramanujan, for |q| < 1,
(5)

(q) :=

n=0

q n(n+1)/2 =

(q 2 ; q 2 )
,
(q; q 2 )

and, for q = e2 iz and Im z > 0,

(6)

f (q) :=

(1)n q n(3n1)/2

n=

= (q; q) =: e2 iz/24 (z).

The product representations in (5) and (6) (and
(35)) are instances of the Jacobi triple product
identity. An incomplete elliptic integral of the first
kind is an integral of the type
Z
d
q
,
0 < /2,
0
1 k2 sin2

where k, 0 < k < 1, is called the modulus of the

integral. The complete elliptic integral of the first
kind is that above when = /2 and is denoted
by K(k), i.e.,
Z /2
d
q
K(k) :=
.
0
1 k2 sin2

To illustrate Ramanujans formulas, we record

a triple of integral formulas found on page 52
in Ramanujans lost notebook [62]. Recall first
that the RogersRamanujan continued fraction is
defined by
(7)

q
q2
q3
q 1/5
+
+
+
1
1
1
1 +
2
5
3
5
(q
;
q
)
(q
;
q
)

= q 1/5
,
|q| < 1,
5
4
5
(q; q ) (q ; q )

R(q) :=

where the product representation is a consequence

of the RogersRamanujan identities (23) and (24).

20

Z 2 tan1 (53/4 qf 3 (q5 )/f 3 (q))

0

d
q
1 5 53/2 sin2

(10)
Z 2 tan1 (51/4 q(q5 )/(q))

d
q
.
= 5
0
1 51/2 sin2

Readers do not have to be told that these are

rather uncommon integral formulas that would
present a challenge to any integral zealot not
steeped in the mathematics of Ramanujan surrounding theta functions. Theorem 2 and the
other formulas in the lost notebook of this sort
were first proved by S. Raghavan and S. S. Rangachari [57], but some of their proofs are probably
not like those of Ramanujan, because they used
the theory of modular forms. Motivated by this
fact, Berndt, H. H. Chan, and S.S. Huang [25] found
proofs for all of Ramanujans approximately 15
formulas employing only results from Ramanujans notebooks [61] and lost notebook [62]. See
also Chapter 15 of our book [10].
We briefly indicate some of the ingredients in
the proofs of (8)(10) and Ramanujans further
formulas of this sort.
First, to prove (8)(10), three different transformation formulas for incomplete elliptic integrals
are needed. One of them is the duplication formula given in the following lemma [21, p. 106, Entry
17(vi)], [61].
Lemma 3. Suppose that 0 < , 12 < 12 . If
p
1
cot tan( 2 ) = 1 x sin2 , then
Z
Z
d
d
q
q
=
.
(11)
2
2
0
0
1 x sin
1 x sin2

Ramanujan found many transformations like

(11). Some can be found in Entry 7 of Chapter 17
in [61], [21, pp. 104114], but others are scattered
throughout his notebooks.
Secondly, modular equations play key roles in
some proofs.
Thirdly, differential equations are sometimes
necessary. For (8), only the simple differential
equation
R(q) f 5 (q)
R (q) =
5q f (q 5 )
is needed. However, for other integral identities,
more difficult differential equations are required.
We give one such example.

Notices of the AMS

Volume 55, Number 1

Lemma 4. Let v be defined by

(12)

v := v(q) := q

Theorem 6. If 0 < < 1 and x > 0, then

f (q)f (q 15 )
f (q 3 )f (q 5 )

!3

(15)

 


x
1
1
sin(2 n) = x 2 4 cot( )
F
n
n=1

 p

1 X X J1 4 m(n + )x
p
+
x

2
m(n + )

Then
dv
= f (q)f (q 3 )f (q 5 )f (q 15 )
dq
p
1 10v 13v 2 + 10v 3 + v 4 .

m=1 n=0

The differential equation of Lemma 4 is not easy

to prove and is crucial in proving the following
theorem of Ramanujan from page 51 in his lost
notebook.
Theorem
5. Let v be defined by (12), and let =

( 5 + 1)/2. Then
Zq
(13)
f (t)f (t 3 )f (t 5 )f (t 15 )dt
0

1
=
5

Z 2 tan1 (1/5)

1
5

2 tan1

111vv 2
1+vv 2

q
1

d
9
25

sin

Although we are able to prove Theorems 2 and

5, as well as Ramanujans further claims about
such integrals, we would not have been able to do
so without knowing the formulas at the start. In
other words, we do not know what led Ramanujan
to believe that such formulas existed. A few further
theorems were established in [25], but we know
of no other formulas of this kind in the literature.
As indicated at the end of [25], these formulas
appear to be connected with elliptic curves. But,
in summary, further study and understanding are
necessary.

No. 8. A Double Sum of Bessel Functions

and Sums of Two Squares
On page 335 in [62], Ramanujan records two identities, each involving a double series of Bessel
functions. We discuss only one of the two identities.
The first identity involves the ordinary Bessel
function J1 (z), where
J (z) :=

n=0

(1)n
n! ( + n + 1)

 +2n
z
,
2

0 < |z| < , C.

To state Ramanujans claim, we need to also define
(
[x],
if x is not an integer,
(14) F(x) =
1
x 2,
if x is an integer,
where, as customary, [x] is the greatest integer
less than or equal to x.

January 2008

 p

J1 4 m(n + 1 )x
p

m(n + 1 )

Note that the series on the left-hand side of (15)

is finite, and it is discontinuous if x is an integer.
To examine the right-hand side, we recall that [73,
p. 199], as x ,




2 1/2
(16)
J (x)
cos x 21 14 .
x

Hence, as m, n , the terms of the double series

on the right-hand side of (15) are asymptotically
equal to

 p

cos 4 m(n + )x 43
1

(n + )3/4
2x1/4 m3/4

 p
3
cos 4 m(n + 1 )x 4
.

(n + 1 )3/4

Thus, if indeed the double series on the right side

of (15) does converge, it converges conditionally
and not absolutely.
We might ask what led Ramanujan to consider
(15). Recall that Ramanujan visited G. H. Hardy in
Cambridge during the years 19141919. Early in
this stay, Hardy focused attention on the classical
circle and divisor problems. We show that the first
double series of Bessel functions (15) is related
to the circle problem; the second series of Bessel
functions is related to the divisor problem and a
series of G. Vorono [70].
Let r2 (n) denote the number of representations
of the positive integer n as a sum of two squares.
Since each representation of an integer can be
associated with a lattice point in the plane, we can
write, with r2 (0) := 1,
X
r2 (n) = x + P (x),
(17)
0nx

where the prime on the summation sign on the

left side indicates that if x is an integer, only 12 r2 (x)
is counted. One of the most famous unsolved problems in the theory of numbers is to determine the
correct order of magnitude of the error term
P (x) as x . This is the circle problem. It was

shown by Gauss that P (x) = O( x), as x .

Notices of the AMS

21

W. Sierpinski [64] in 1906, and then Hardy [43],

[44, pp. 243263] in 1915 proved that
(18)
 1/2

X
X

x
r2 (n)
r2 (n) = x +
J1 (2 nx).
n
n=1
0nx
Thus, using (18) in (17), we can obtain a representation for P (x) as an infinite series of Bessel
functions. Observe that the series on the righthand side of (18) is similar to the inner series on
the right side of (15). In fact, one can derive the
following corollary [29] of Theorem 6.
Corollary 7. For any x > 0,
X
r2 (n)
(19)
0nx

 q


X
J1 4 m(n + 1 )x
X

4
q
= x + 2 x

1
m(n + )
n=0 m=1
4

X
X
r (n) 2 (n+b)a
r2 (n) 2 (n+a)b
2

e
=
,
e
n+a
n+b
n=0
n=0

 q


J1 4 m(n + 43 )x
q

m(n + 34 )

Ramanujan might therefore have derived (15)

in anticipation of applying it to the circle problem.
However, we have no record of any further work of
Ramanujan on the circle problem. Because r2 (n)
does not arise on the right-hand side of (19), it
may be that (19) is potentially more useful in
the circle problem than (18). On the other hand,
double series are usually more difficult to estimate
than a single infinite series. We do not provide
details, but it is not difficult to show that (19) can
be derived from (18), and conversely.
In [43], Hardy proved that


(20)
P (x) O x1/4 log1/4 x .

In the past 90 years, this result has been improved

only once. In 2003, K. Soundararajan [67] proved
that
!
3(21/3 1)/4
1/4 (log log x)
1/4
P (x) O x log x
.
(log log log x)5/8

In fact, (18) was not employed by Hardy in his

proof of (20). The identity (18) is more useful in
obtaining an upper bound for P (x), and it has been
the starting point of almost all investigations on
finding upper bounds for P (x). In particular, Sierpinski [64] used (18) to prove that P (x) = O(x1/3 ),
giving the first improvement
on Gausss upper

bound P (x) = O( x). Since 1906, the exponent

1/3 has been gradually whittled down by a succession of several mathematicians. Currently, the
best result,


(21)
P (x) = O x131/408 (log x)18,627/8320

is due to M. N. Huxley [49]. Note that

.3149 . . . . It is conjectured that


P (x) = O x1/4+ ,
22

131
408

for every > 0. Thus, the theorems of Hardy and

Soundararajan are thought to be much closer to
the correct order of magnitude of P (x) than (21).
Returning to (15), Berndt and A. Zaharescu [29]
have found a long, difficult proof of (15), but with
the order of summation reversed. However, they
[30] recently found a completely different proof of
(15) with the order of summation as prescribed by
Ramanujan. A corollary of their proofs is that the
order of summation in the double series in (15)
can be reversed without affecting the equality. Can
one directly prove that the order of summation in
(15) can be reversed? This would seem to be an
extremely difficult problem in view of the delicate
convergence of the double series.
In this same paper [43], Hardy relates a beautiful identity of Ramanujan connected with r2 (n),
namely, for a, b > 0, [43, p. 283], [44, p. 263],

which is not given elsewhere in any of Ramanujans

published or unpublished work. This is further evidence that Hardys work on sums of squares had
captured Ramanujans attention.

No. 7. Hadamard Products

On page 57 of Ramanujans lost notebook [62], we
find one of the most peculiar of all of Ramanujans
formulas
2

X
an q n
(22)
=
(q; q)n
n=0

n=1

aq 2n1
1+
n
1 q y1 q 2n y2 q 3n y3

where
1
,
(1 q)2 (q)
y2 = 0,
y1 =

y3 = y1

X
q + q3
(2n + 1)q 2n+1
y13
,
2
3
(1 q )(1 q )
1 q 2n+1
n=0

y4 = y1 y3 ,

and (q) is defined by (5). The most perplexing

aspect of this formula, when first encountering it,
is that the left-hand side is quite familiar. Indeed,
it occurs in an identity originally published by
L. J. Rogers and Ramanujan [63], namely,
2

X
1
an q n
=
(q;
q)
(aq;
q)
n
n=0

j 2j j(5j1)/2
2j (aq; q)
X
(1)
a
q
(1

aq
)
j1
.
1 +
1 qj
(q; q)j1
j=1

Notices of the AMS

Volume 55, Number 1

From this identity it is a simple exercise to deduce

the celebrated RogersRamanujan identities
(23)

n=0

1
qn
=
(q; q)n
(q; q 5 ) (q 4 ; q 5 )

No. 6. An Integral with Character

With a modest change in notation, on page 207
in his lost notebook [62], Ramanujan offers the
following formula for a character analogue of the
Dedekind eta-function.

and
(24)

X
1
q n +n
=
.
2 ; q 5 ) (q 3 ; q 5 )
(q;
q)
(q
n

n=0

However, the right-hand side of (22) is bizarre,

since it does not appear to lie in the classical
theory of q-hypergeometric series, even though it
does contain familiar objects such as the classical
theta function (q). Upon reflection, the central
idea dawns. This is the Hadamard product for the
entire function of the variable a
2

X
an q n
,
(q; q)n
n=0
where for our purposes here q is a fixed parameter
with 0 < q < 1.
The first published proof of (22) [7] by Andrews
claims a larger domain of validity than is actually
delivered by the argument of [7]. In fact, the claim
1
that (22) is valid for 0 < q < 4 has to be modified
to 0 < q < 0.00406. However, empirical studies
suggest that, in fact, (22) is valid for 0 < q < 1.
Subsequently, several further papers have been
written developing this topic. A companion identity found on page 26 of the lost notebook gives
a Hadamard product expansion for the entire
function of a

X
2
an q n .
(25)
n=0

This Hadamard product of (25) was examined in

[8], and again it was claimed that the Hadamard
expansion is valid for 0 < q < 41 . Again (following standard methods from the theory of implicit
functions [53]), this interval must be constricted
to 0 < q < 0.00792.
By completely different methods, W. Hayman
[45] asymptotically established the formulas for
the zeros of (22), as well as for those of a
generalization
(26)

n=0

Theorem
8. Let (n) denote the Legendre symbol
 
n
,
and
recall
that f (q) is defined by (6). Then,
3
for 0 < q < 1,
(27)
!
Z

Y
1 1 f 9 (t) dt
1/9
n n(n)
q
(1q )
= exp C
,
9 q f 3 (t 3 ) t
n=1
where

3 3
L(2, ) = L (1, ),
(28)
C :=
4
where L(s, ) denotes the Dirichlet L-function associated with the character .
The second equality in (28) was not given
by Ramanujan. In fact, after the first equality,
Ramanujan recorded two question marks ??, evidently indicating some doubt about the truth of his
formula. However, Ramanujan is indeed correct.
Theorem 8 was first proved by S. H. Son [66] in
an incomplete form, because the identity of C was
not addressed. Berndt and Zaharescu [28] gave a
completely different proof of Theorem 8 in which
the value of C in (28) naturally emerged from their
proof. A nagging question now loomed. Is (27) an
isolated identity, or are there further examples of
this sort?
It turns out that one should not think of Theorem 8 in terms of Ramanujans function f (q),
but instead in terms of Eisenstein series.
Suppose that is a nontrivial primitive character modulo N and that, for n 0, Bn, denotes the
nth generalized Bernoulli number defined by [69,
p. 12]

N
X
X
tn
(n)tent
=
,
B
n,
eNt 1
n!
n=0
n=1

|t| < 2 /N.

Let k 2 be an integer, and choose a character

such that (1) = (1)k . If q := e2 iz , the function

an q n
.
(q; q)n (bq; q)n

T. Huber [47], [48] generalized the methods of [7]

to obtain results similar to Andrews for the series
(26) as well as for

X
(c/(aq); q)n an q n(n+1)/2
.
(q; q)n (bq; q)n
n=0
Finally, M. Ismail and C. Zhang [51], along completely new lines, developed a method involving
elliptic integrals for proving (22) and related formulas. The power of their method allows them
to prove that the domain 0 < q < 1 is, at least,

January 2008

eventually valid for all zeros of sufficiently large

modulus.

(29)

Ek, (z) := 1

2k X X
(d)d k1 q n
Bk, n=1 d|n

is an Eisenstein series of weight k and character

on the congruence subgroup 0 (N) of the full
modular group. We note that (with the correct
choice of ) the integrand in (30) below is such an
Eisenstein series.
Theorem 9. Suppose that is real, that k 2 is
an integer, and that is a nontrivial Dirichlet character that satisfies the condition (1) = (1)k .

Notices of the AMS

23

Then, for 0 < q < 1,

q

(30)

n=1

= exp C

where

Z1
q

and
(1 q n )(n)n

k2

dt
k1 n
,

(d)d t

t
n=1 d|n
X
X

C = L (2 k, ).
One consequence of Theorem 9 is that it provides an explanation of another identity found in
Ramanujans lost notebook and
 first proved by
Andrews [3]. If we let (n) = n5 , which is an even
character, and k = 2, then it can be shown that
Theorem 9 yields the identity
(31)
!
Z

Y
1 1 f 5 (t) dt
1/5
n (n)
,
q
(1q )
= exp C5
5 q f (t 5 ) t
n=1
where
C5 = L (0, ).

Although we shall not provide details, which can

be found in [1], (31) is equivalent to an identity for
the RogersRamanujan continued fraction R(q)
given by
!

Z
51
1 1 f 5 (t) dt
(32)
R(q) =
exp
2
5 q f (t 5 ) t
and found on page 46 in Ramanujans lost notebook. To deduce the left-hand side of (32) from
the left-hand side of (31), use the identity for
R(q) found in (7). Further examples of Theorem
9 can be found in [1]. Considerable generalizations of Theorem 9 have been established by
Y. Yang [74] and R. Takloo-Bighash [68].

No. 5. Sums of Tails of Eulers Partition

Products
In the middle of page 14 of the lost notebook
[62] appear, at first glance, two of the strangest
formulas in the entire volume, namely,
(33)

R(q) = 1 +

Note that, in each of (33) and (34), Ramanujan is

summing the difference between an infinite product, S(q), and its nth partial product. The function
R(q) is the generating function for the excess of
the number of partitions into distinct parts with
even rank over those with odd rank. In [14], it is
shown that almost all the coefficients of R(q) are
0 (i.e., asymptotically 100%), and that every integer
appears as a coefficient infinitely often.
One would hope that there is some general
principle from which both (33) and (34) would
emerge. However, the first proofs of (33) and (34)
in [5] are essentially the long culmination of a
long struggle with these identities and provide no
general insight.
Subsequently, D. Zagier [75] proved a similar
result

X
((q; q) (q; q)n ) = D(q)(q; q)
n=0

 
denotes the Jacobi symbol. It was subwhere 12
n
sequently discovered that N. J. Fine [41, p. 14] had
earlier proved that
!



X
1
1
1
1
.

=
D(q) +
(q; q)
(q; q)n
(q; q)
2
n=0
All of these examples taken together strongly
suggest that a common rationale lies behind all
these discoveries. This turns out to be the case. In
[18], the following lemma was proved.
P
Lemma 10. Suppose that f (z) = n=0 (n)z n is
analytic for |z| < 1. If is a complex number for
which

X
| (n)| <
and
lim n ( (n)) = 0,

and

n=0

then
S(q)

1
(q; q 2 )n+1

= S(q)D(q 2 ) +

where
S(q) := (q; q) =
D(q) :=

24



1 X 12
2
nq n /24 ,
2 n=1 n

n=0

1
(S(q) (q; q)n ) = S(q)D(q) + R(q)
2
n=0

(34)

X
q n(n+1)/2
.
(q; q)n
n=1

1
,
(q; q 2 )

X
qn
1
+
,
2 n=1 1 q n

1
R(q),
2

lim

z1

X
d
( (n)) .
(1 z)f (z) =
dz
n=0

This unleashed numerous results similar to (33)

and (34) including seven further results in [18] and
others by G. H. Coogan and Ono [37]. For example
[18, p. 407], if
(35)

Notices of the AMS

(q) =

(1)n q n =

n=

(q; q)
,
(q; q)

Volume 55, Number 1

then Andrews, Jimenez, and Ono [18] proved that

!

X
(q; q)n
(q)
(q; q)n
n=0
=

X
(q; q)n1
qn
2
.
(q) n=1 (q; q)n1 (1 q n )2

It should be noted that both Zagier [75] and the

aforementioned authors were primarily concerned
with the applications of such results to finding
values of certain L-functions.
In [5], the first author asked for combinatorial
proofs of (33) and (34). Recently, in a beautiful
paper [36], W. Y. C. Chen and K. Q. Ji provided the
combinatorial proofs that were sought by several
researchers.
Finally, noting that Lemma 10 was a natural
next step beyond Abels Lemma, Andrews and
P. Freitas [15] established an infinite family of extensions of Abels Lemma and applied their results
to obtain further q-series identities.

No. 4. A Continued Fraction with Three

Limit Points
Let = e2 i/3 . On page 45 in his lost notebook
[62], Ramanujan writes, for |q| < 1,
(36)
lim

1
1
1
1
1 1 + q 1 + q2 1 + qn + a
2

=
where
(37)

:=

n+1
n1

(q 2 ; q 3 )
,
(q; q 3 )

1 a2 (2 q; q)
.
1 a (q; q)

Of course, because of the appearance of the limiting variable n on the right side of (36), Ramanujans
claim is meaningless as it stands, but properly interpreted, the claim (36) is correct and interesting.
Ramanujan is indicating that (36) has three limits,
depending upon the residue class modulo 3 in
which n lies. This should be compared to the
classical theorem in the theory of continued fractions, which asserts that if all the elements of
a divergent continued fraction are positive, then
the even and odd approximants approach distinct
limits [54, pp. 9697]. Before further discussing
why (36) belongs to the Top Ten in our Hit Parade
of Ramanujans fascinating formulas, we restate it
in the more standard fashion in which it was first
proved by Andrews, Berndt, J. Sohn, A. J. Yee, and
A. Zaharescu [13]. See also, [10, Entry 8.2.2].

January 2008

Theorem 11. Let N 1 = 3v + , where = 0 or

1. Then
(38)
!
1
1
1
1
lim
N
1 1 + q 1 + q 2 1 + q N1 + a
= 2

+1 (q 2 ; q 3 )
,
1 (q; q 3 )

where is defined in (37).

In his notebooks [61, Vol. 2, p. 290], for

0 < |q| < 1, Ramanujan offered the continued
fraction
(q 2 ; q 3 ) 1
q
q3
q5
=
(q; q 3 )
1 1 + q 1 + q2 1 + q3
1
1
1
1
(39)
,
=
1 q 1 + 1 q 2 + 1 q 3 + 1

which was first proved by Andrews, Berndt, Jacobsen, and Lamphere [11], and later proved more
simply by Andrews, Berndt, Sohn, Yee, and Zaharescu [12]. Thus, when a = 0, the continued
fraction on the left side of (36) or (38) is the same
as the continued fraction on the far right side of
(39), but with q replaced by 1/q.
Remarkably, (q 2 ; q 3 ) /(q; q 3 ) appears in the
three limits on the right side of (36) or (38).
In this sense, Ramanujans result (36) is analogous to his theorem on the divergence of the
RogersRamanujan continued fraction R(q) (defined in (7)), found on pages 374 and 382 in his
third notebook [61], and first proved by Andrews,
Berndt, Jacobsen, and Lamphere [11], [22, p. 30,
Entry 11]. In the latter result, Ramanujan explicitly
determines the limits of the even and odd indexed
approximants of the divergent RogersRamanujan
continued fraction for |q| > 1 and shows that these
limits can be expressed in terms of R(1/q) and
R(1/q 4 ).
If a 0, the generalized continued fraction
in (36) converges in the sense that when n is
confined to any one of the three residue classes
modulo 3, the limit of the left side exists and is
equal to that claimed on the right side of (36).
Ramanujans result is an example in the theory
of the general convergence of continued fractions
due to L. Jacobsen [52] in 1986; see also her book
with H. Waadeland [54, pp. 4144]. This is another illustration of Ramanujan having discovered a
fundamental idea long ahead of his time.
In [13], the authors construct a large class of
continued fractions with three limit points in the
sense of Theorem 11. However, their theorem does
not cover Theorem 11, which apparently lies at a
deeper level. Ramanujans identity (36) has been
generalized by D. Bowman and J. McLaughlin [31,
Theorem 3], who have established an identity having any number n 3 limit points that reduces to

Notices of the AMS

25

(36) when n = 3. Ramanujans continued fraction

(38) has limit period 1, i.e., is a limit 1-periodic
continued fraction. The work of Bowman and
McLauglin in [31, especially Theorem 3] and [32,
especially Theorem 7] significantly increases our
understanding of limit 1-periodic continued fractions with n limits. A completely different proof of
Theorem 11 arising from the theory of orthogonal
polynomials has recently been given by M. E. H.
Ismail and D. Stanton [50].

No. 3. Cranks
As usual, define the RogersRamanujan functions
G(q) and H(q) by
(40)
2
2

X
X
qn
q n +n
G(q) :=
and
H(q) :=
.
(q; q)n
(q; q)n
n=0
n=0
One of the formulas in the lost notebook [62, p.
20] that has had the most impact on subsequent
research in the theory of partitions is given by
(41)
(q; q)
= A(q 5 ) q( + 1 )2 B(q 5 )
(q; q) ( 1 q; q)
+q 2 ( 2 + 2 )C(q 5 ) q 3 ( + 1 )D(q 5 ),
where is any primitive fifth root of unity and

(q 5 ; q 5 ) G2 (q)

A(q)
=
,

H(q)

B(q) = (q 5 ; q 5 ) G(q),

(42)
5
5

C(q)
=
(q
;
q
)

H(q),

5
5
2

(q
;
q
)

H (q)

.
D(q) =
G(q)

Identity (41) was proved first by F. Garvan [42],

who used it to give a new proof of Ramanujans
famous congruence for the partition function p(n)
[58], [60, pp. 210213]
(43)

p(5n + 4) 0 (mod 5),

n 0.

Further proofs of (41) were later given by A. B.

Ekin [40] and Berndt, H. H. Chan, S. H. Chan, and
W.C. Liaw [23].
More important, however, was the use of (41)
by Garvan [42] and subsequently by Andrews and
Garvan [16] to provide the answer to a tantalizing
question posed by F. J. Dyson [39]. Namely, is there
a partition statistic that provides a combinatorial
interpretation for [59], [60, p. 230]
p(11n + 6) 0 (mod 11),

n 0,

in the same way that Dysons rank provides such

an interpretation for (43)? Dyson conjectured that
such a statistic exists, and he named it the crank.
In [16], it was shown that if
(44)

26

X
X
(q; q)
c(m, n)z m q n ,
=
(zq; q) (z 1 q; q)
n=0 m=

then, except for n = 1, c(m, n) is the number of

partitions of n with crank m, where the crank is
given as follows.
Definition 12. For any partition , let ( ) denote
the largest part of , ( ) denote the number of
ones in , and ( ) denote the number of parts of
larger than ( ). The crank c( ) is then given
by
(
( ),
if ( ) = 0,
c( ) =
( ) ( ),
if ( ) > 0.
The generating function (44) for cranks along
with Ramanujans identity (41) form the starting point for K. Mahlburgs deep and fascinating
study [55] of a variety of congruence theorems for
c(m, n). Thus, the results of K. Ono [56] on congruences for p(n) have now been refined by Mahlburg
with congruences for the related c(m, n).
There is compelling evidence that the last topic
on which Ramanujan worked before he died was
cranks [24] (although, of course, he would not
have used this terminology).

No. 2. The Mock Theta Functions

Perhaps the greatest surprise for the first author,
when he began to thoroughly examine the pages in
the lost notebook in May 1976, was the appearance
of formulas such as
(q 2 ; q 5 ) (q 3 ; q 5 ) (q 5 ; q 5 )
(45) 0 (q) =
(q 2 ; q 10 ) (q 8 ; q 10 )
+1

q 5n

(q; q 5 )n+1 (q 4 ; q 5 )n
n=0

where
(46)

0 (q) := 1 +

q n (q; q 2 )n

n=1

is a fifth order mock theta function, and where we

are utilizing the notation (2) and (1). This formula
was initially published without proof in [2], giving
an initial introduction to the lost notebook.
It turns out that there are five formulas in the
lost notebook equivalent to (45); each is related
to one of the fifth order mock theta functions
connected with the first RogersRamanujan function G(q), defined by (40). Moreover, there are
five further formulas associated with five other
fifth order mock theta functions, which are related to the second RogersRamanujan function
H(q), also defined by (40). In [17], Andrews and
Garvan proved that the five identities within each
class are equivalent. In other words, if one is true,
they all are true; and if one is false, they all are
false. Indeed, building on the work of Garvan [42],
Andrews and Garvan [17] were able to show that
(45) and the remaining four formulas from the
first class were, in fact, equivalent to the following
assertion about partitions [17, p. 243]:

Notices of the AMS

Volume 55, Number 1

and

The number of partitions of 5n

with rank congruent to 1 modulo 5
equals the number of partitions of
5n with rank congruent to 0 modulo 5 plus the number of partitions
of n with unique smallest part and
with all other parts twice the
smallest part.

f1 (q) = 1 +

The mock theta conjectures became known as a

contraction for the assertions that both sets of
five formulas are indeed true.
All these results (typified by (45)) were completely unexpected, primarily owing to the following words of G. N. Watson [72, p. 274] from the
introduction of his paper on the fifth order mock
theta functions:
but I have failed to construct
a complete and exact transformation theory of the functions, on the
lines of the transformation theory of functions of the third order,
and in view of the complexity of
all the series which are involved,
I am becoming somewhat skeptical concerning the existence of
an exact transformation theory for
functions of the fifth order.



qf0 (q) zj(q 6 z 5 , q 30 ) + z 4 j(q 24 z 5 , q 30 )



+f1 (q) z 2 j(q 12 z 5 , q 30 ) + z 3 j(q 18 z 5 , q 30 )
+2

X
(1)r q 15r +15r +3 z 5r +5
1 zq 6r +2
r =

+2

X
(1)r q 15r +15r +3 z 5r
,
1 z 1 q 6r +2
r =

This, of course, is only the beginning of the

story. S. P. Zwegers [76] has developed a fully general theory of mock theta functions, and
K. Bringmann and Ono [33] have extended that
work to a theory of Maass wave forms that has,
among many other things, provided a proof of the
AndrewsDragonette conjecture [6].
Most recently, Bringmann, Ono, and R. C. Rhodes
[35] have used their development of the mock theta
functions as the holomorphic parts of Maass wave
forms to obtain a general theorem that has as
corollaries the mock theta conjectures [46].

One of the romantic episodes in the theory of

partitions (and one of the events that most highlights Ramanujans incredible insight) is Dysons
discovery of the rank of a partition. The rank of a
partition is defined to be the largest part of
minus the number of parts of . It is not difficult
to show [39], [20] that the generating function for
N(m, n), the number of partitions of n with rank
m, is given by
(47)
2

X
X
X
qn
N(m, n)z m q n =
.
(zq; q)n (q/z; q)n
n=0
n=0 m=
Dysons objective in defining the rank was
to provide a combinatorial explanation of Ramanujans congruence (43). In Dysons words [39],
although we can prove that the partitions of
5n + 4 can be divided into five equally numerous
subclasses, it is unsatisfactory to receive from the
proofs no concrete idea of how the division is to be
made. He conjectured (among other things) that
if N(m, Q, n) denotes the number of partitions of
n with rank congruent to m modulo Q, then, for
0 m 4,
1
(48)
N(m, 5, 5n + 4) = p(5n + 4).
5
Thus, the partitions of 5n + 4 would be divided
into five equinumerous classes.
In 1954, A. O. L. Atkin and H. P. F. SwinnertonDyer [20] proved all of Dysons conjectures. Their
proof of (48) is a magnificent tour de force in the
theory of theta functions and related series.
Now comes the great surprise unearthed by
Garvan from Ramanujans lost notebook [62]. Let
be a primitive fifth root of unity, let

where
j(z, q) = (z; q) (q/z; q) (q; q) ,

January 2008

(q) := 1 +

qn
f0 (q) = 1 +
,
(q; q)n
n=1

q n +n
.
(q; q)n
n=1

No. 1. The Panels Top ChoiceRanks

As noted in [17], a proof of the mock theta conjectures would allow the same treatment for fifth
order functions that Watson himself provided for
the third order mock theta functions [71].
D. Hickerson [46] proved the mock theta conjectures. Suffice it to say that his method of proof
began with Hecke-like representations of the mock
theta functions given in [4] and built from there
magnificent theta series identities such as
z 2 j(z, q)j(z, q 3 )(q 2 ; q 2 )
=
j(z, q 2 )

(q) := 1 +

Notices of the AMS

n=0

n=0

q 5n
,
5
(q; q )n+1 (q 4 ; q 5 )n
2

q 5n
,
2
5
(q ; q )n+1 (q 3 ; q 5 )n

27

and let A(q), B(q), C(q), and D(q) be defined by

(42). On page 20 of the lost notebook [62], we find
the identity
2

X
qn
(49)
=
(q; q)n ( 1 q; q)n
n=0
A(q 5 ) + ( + 1 2)(q 5 )+
qB(q 5 ) + ( + 1 )q 2 C(q 5 )
(

( + 1 )q 3 D(q 5 ) ( 2 + 2 2)

(q 5 )
q5

Garvan showed that all of the theorems proved

by Atkin and Swinnerton-Dyer [20] for N(m, 5, n)
can be deduced from (49). Thus the conjectures
originating with Dyson in an effort to provide a
combinatorial proof of (43) turn out to be provable
via Ramanujans formula (49).
This is not the end of the story. Bringmann and
Ono [33] have undertaken a further analysis of
(47) and (48) to yield many deeper results relating
the rank to partition congruences. In addition, the
rank has been extended in [9] to a more general
class of partitions with applications to congruences for the AtkinGarvan moments of ranks
[19].

Analysis of the Voting of the Top Ten

Entries
Your Hit Parade The Top Ten Formulas
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.

Ranks
Mock Theta Functions
Cranks
Continued Fraction with 3 Limit Points
Sums of Tails
Integral with Character
Hadamard Products
Double Sum of Bessel Functions
Some Challenging Integrals
Bilateral Hypergeometric Series

Eight of the Top Ten received votes for first place.

It is interesting that the entry receiving the most
first place votes was not our first place winner, but
instead the Mock Theta Functions garnered 12
first place votes to outdistance Ranks in second
place with 6 and Cranks with 5. Although finishing
in 6th place, the Integral with Character received
4 first place votes indicating that some voters
really appreciated the beauty of this unusual looking formula. Also, despite ranking in 7th place,
three voters gave the Hadamard Products their

28

top vote. Challenging Integrals in the 9th position

received 2 first place votes, with the Continued
Fraction with 3 Limit Points and the Bilateral Hypergeometric Series, despite being shoved to last
place, garnering the remaining first place votes.
The Sums of Tails in 5th place and the Bessel
Functions in 8th place were unfortunately shut
out of the first place rankings.
Ranks received uniformly high marks from
most voters with 8 seconds to augment its half
dozen first place votes. No one ranked Ranks lower than 8th place. In contrast, Cranks in third place
received votes at all ten positions, with two voters
ranking Cranks at number 9 and one at number
10. No one ranked Mock Theta Functions last, but
every other place received at least one vote from
the panel, with only four ranking Mock Theta
Functions in the second position and two ranking
them in the 9th slot. The three top vote getters
clearly set themselves apart from the remaining
seven entries, with Mock Theta Functions losing
to Ranks by a total of only four points.
The Continued Fraction with Three Limit
Points was appreciated by most voters. Although
capturing only 1 first place vote, 2 second place
votes, and 3 third place votes, it had 21 votes in
the 47 range, which was enough, by just four
points, to beat out Sums of Tails, which was
hampered by each of 3 ninth and tenth place
rankings. The Continued Fraction had votes at
every position, while Sums of Tails had votes at
every position except the top one.
The Integral with Character in sixth place prevailed over the Hadamard Products by one measly
vote. As indicated above, several voters loved one
of these two entries, but, on the other hand, the
Integral with Character had 4 last place votes and
the Hadamard Products had even more, namely, 5
last place votes. Strangely, no one voted Hadamard
Products in the fifth position, while the Integral
with Character captured votes at all positions.
Although receiving 8 votes in either the second
or third positions, Bessel Functions received several in the lower echelons, including 5 in the tenth
position. In contrast, the Challenging Integrals
captured only 3 seconds and no thirds to finish six
points behind Bessel Functions. Although receiving only 2 tenth place votes, Challenging Integrals
received a large number of eighth and ninth place
tallies to doom the integrals to ninth place.
With 9 votes at the seventh rung and 12 votes
at the bottom rung, Bilateral Hypergeometric Series was relegated to the bottom position in the
panelists voting. However, some voters indeed
did appreciate Ramanujans remarkable Bilateral
Hypergeometric Series identity, as it received at
least one vote at each position.
The authors are grateful to several members of
the panel for their corrections and kind and useful
suggestions.

Notices of the AMS

Volume 55, Number 1

References

[24]

[1] S. Ahlgren, B. C. Berndt, A. J. Yee, and A.

Zaharescu, Integrals of Eisenstein series and
derivatives of L-functions, Internat. Math. Res. Not.
2002, No. 32, 17231738.
[2] G. E. Andrews, An introduction to Ramanujans
lost notebook, Amer. Math. Monthly 86 (1979),
89108.
, Ramanujans lost notebook III: The
[3]
RogersRamanujan continued fraction, Adv. Math.
41 (1981), 186208.
[4]
, The fifth and seventh order mock theta
functions, Trans. Amer. Math. Soc. 293 (1986),
113134.
, Ramanujans lost notebook V: Eulers
[5]
partition identity, Adv. Math. 61 (1986), 156164.
[6]
, Partitions: At the interface of q-series and
modular forms, Ramanujan J. 7 (2003), 385400.
[7]
, Ramanujans lost Notebook. VIII: The entire RogersRamanujan function, Adv. Math. 191
(2005), 393407.
, Ramanujans lost Notebook. IX: The par[8]
tial theta function as an entire function, Adv. Math.
191 (2005), 408422.
, Partitions, Durfee symbols, and the Atkin[9]
Garvan moments of ranks, Invent. Math. 169 (2007),
3773.
[10] G. E. Andrews and B. C. Berndt, Ramanujans Lost
Notebook, Part I, Springer, New York, 2005.
[11] G. E. Andrews, B. C. Berndt, L. Jacobsen, and R.
L. Lamphere, The continued fractions found in the
unorganized portions of Ramanujans notebooks,
Memoirs, Amer. Math. Soc., No. 477, 99 (1992).
[12] G. E. Andrews, B. C. Berndt, J. Sohn, A. J. Yee, and
A. Zaharescu, On Ramanujans continued fraction
for (q 2 ; q 3 ) /(q; q 3 ) , Trans. Amer. Math. Soc. 355
(2003), 23972411.
, Continued fractions with three limit points,
[13]
Adv. Math. 192 (2005), 231258.
[14] G. E. Andrews, F. J. Dyson, and D. R. Hickerson,
Partitions and indefinite quadratic forms, Invent.
Math. 91 (1988), 391407.
[15] G. E. Andrews and P. Freitas, Extension of Abels
lemma with q-series implications, Ramanujan J. 10
(2005), 137152.
[16] G. E. Andrews and F. G. Garvan, Dysons crank of a
partition, Bull. Amer. Math. Soc. 18 (1988), 167171.
, Ramanujans lost" notebook, VI: The mock
[17]
theta conjectures, Adv. Math. 73 (1989), 242255.
[18] G. E. Andrews, J. Jimnez-Urroz, and K. Ono, qseries identities and values of certain L-functions
Duke Math. J. 108 (2001), 395419.
[19] A. O. L. Atkin and F. G. Garvan, Relations between ranks and cranks of partitions, Ramanujan
J. 7 (2003), 343366.
[20] A. O. L. Atkin and H. P. F. Swinnerton-Dyer, Some
properties of partitions, Proc. London Math. Soc. 66
(1954), 84106.
[21] B. C. Berndt, Ramanujans Notebooks, Part III,
Springer-Verlag, New York, 1991.
, Ramanujans Notebooks, Part V, Springer[22]
Verlag, New York, 1998.
[23] B. C. Berndt, H. H. Chan, S. H. Chan, and W.-C.
Liaw, Cranks and dissections in Ramanujans lost
notebook, J. Comb. Thy. (A), 109 (2005), 91120.

January 2008

[25]

[26]

[27]

[28]

[29]
[30]
[31]

[32]

[33]

[34]
[35]

[36]
[37]

[38]

[39]
[40]

[41]

[42]

[43]

[44]
[45]

[46]

, CranksReally the final problem, submitted

for publication.
B. C. Berndt, H. H. Chan, and S.-S. Huang, Incomplete elliptic integrals in Ramanujans lost
notebook, in q-Series from a Contemporary Perspective, M. E. H. Ismail and D. Stanton, eds., American
Mathematical Society, Providence, RI, 2000, pp.
79126.
B. C. Berndt and W. Chu, Two entries on bilateral hypergeometric series in Ramanujans lost
notebook, Proc. Amer. Math. Soc. 135 (2007),
129134.
B. C. Berndt and R. A. Rankin, Ramanujan: Essays and Surveys, American Mathematical Society,
Providence, RI, 2001; London Mathematical Society,
London, 2001.
B. C. Berndt and A. Zaharescu, An integral
of Dedekind eta-functions in Ramanujans lost
notebook, J. Reine Angew. Math. 551 (2002), 3339.
, Weighted divisor sums and Bessel function
series, Math. Ann. 335 (2006), 249283.
, Weighted divisor sums and Bessel function
series, II, in preparation.
D. Bowman and J. McLaughlin, Continued fractions with multiple limits, Adv. Math. 210 (2007),
578606.
, Asymptotics and sequential closures of continued fractions and generalizations, submitted for
publication.
K. Bringmann and K. Ono, The f (q) mock theta
function conjecture and partition ranks, Invent.
Math. 165 (2006), 243266.
, Dysons ranks and Maass forms, Ann. Math.,
to appear.
K. Bringmann, K. Ono, and R. C. Rhoades, Eulerian series as modular forms, J. Amer. Math. Soc., to
appear.
W. Y. C. Chen and K. Q. Ji, Weighted forms of Eulers
theorem, J. Comb. Thy. (A) 114 (2007), 360372.
G. H. Coogan and K. Ono, A q-series identity
and the arithmetic of Hurwitz zeta functions, Proc.
Amer. Math. Soc. 131 (2003), 719724.
J. Dougall, On Vandermondes theorem and some
more general expansions, Proc. Edinburgh Math.
Soc. 25 (1907), 114132.
F. J. Dyson, Some guesses in the theory of
partitions, Eureka (Cambridge) 8 (1944), 1015.
A. B. Ekin, Some properties of partitions in terms
of crank, Trans. Amer. Math. Soc. 352 (2000),
21452156.
N. J. Fine, Basic Hypergeometric Series and Applications, American Mathematical Society, Providence,
RI, 1988.
F. G. Garvan, New combinatorial interpretations of
Ramanujans partition congruences mod 5, 7, and
11, Trans. Amer. Math. Soc. 305 (1988), 4777.
G. H. Hardy, On the expression of a number as
the sum of two squares, Quart. J. Math. (Oxford) 46
(1915), 263283.
, Collected Papers, Vol. II, Oxford University
Press, Oxford, 1967.
W. K. Hayman, On the zeros of a q-Bessel function, in Complex Analysis and Dynamical Systems
II, Contemp. Math. 382 (2005), 205216.
D. Hickerson, A proof of the mock theta
conjectures, Invent. Math. 94 (1988), 639660.

Notices of the AMS

29

[47] T. Huber, Zeros of Generalized RogersRamanujan

Series and Topics from Ramanujans Theory of Elliptic Functions, Ph.D. Thesis, University of Illinois
at Urbana-Champaign, 2007.
[48]
, Hadamard products for generalized Rogers
Ramanujan series, J. Approx. Thy., to appear.
[49] M. N. Huxley, Exponential sums and lattice points
III, Proc. London Math. Soc. (3) 87 (2003), 591609.
[50] M. E. H. Ismail and D. Stanton, Ramanujan continued fractions via orthogonal polynomials, Adv.
Math. 203 (2006), 170193.
[51] M. E. H. Ismail and C. Zhang, Zeros of entire functions and a problem of Ramanujan, Adv. Math. 209
(2007), 363380.
[52] L. Jacobsen, General convergence of continued
fractions, Trans. Amer. Math. Soc. 294 (1986),
477485.
[53] S. G. Krantz, The Implicit Function Theorem,
Birkhuser, Boston, 2002.
[54] L. Lorentzen and H. Waadeland, Continued Fractions with Applications, North Holland, Amsterdam,
1992.
[55] K. Mahlburg, Partition congruences and the
Andrews-Gordon-Dyson crank, Proc. Nat. Acad. Sci.
(USA) 102 (2005), 1537315376.
[56] K. Ono, Distribution of the partition function
modulo m, Ann. Math. 151 (2000), 293307.
[57] S. Raghavan and S. S. Rangachari, On Ramanujans elliptic integrals and modular identities,
in Number Theory and Related Topics, Oxford
University Press, Bombay, 1989, pp. 119149.
[58] S. Ramanujan, Some propertities of p(n), the number of partitions of n, Proc. Cambridge Philos. Soc.
19 (1919), 207210.
, Congruence properties of partitions, Proc.
[59]
London Math. Soc. 2 (1920), xviii.
[60]
, Collected Papers, Cambridge University
Press, Cambridge, 1927; reprinted by Chelsea,
New York, 1962; reprinted by the American
Mathematical Society, Providence, RI, 2000.
[61]
, Notebooks (2 volumes), Tata Institute of
Fundamental Research, Bombay, 1957.
[62]
, The Lost Notebook and Other Unpublished
Papers, Narosa, New Delhi, 1988.
[63] S. Ramanujan and L. J. Rogers, Proof of certain
identities in combinatory analysis, Proc. Cambridge
Philos. Soc. 19 (1919), 211216.
[64] W. Sierpinski, O pewnem zagudnieniu z rachunku
funkeyj asymptotycznych, Prace Mat.-Fiz. 17
(1906), 77118.
[65] L. J. Slater, Generalized Hypergeometric Functions,
Cambridge University Press, Cambridge, 1966.
[66] S. H. Son, Some integrals of theta functions, in
Number Theory, Fifth Conference of the Canadian Number Theory Association, R. Gupta and K.
S. Williams, eds., American Mathematical Society,
Providence, RI, 1999, pp. 329339.
[67] K. Soundararajan, Omega results for the divisor
and circle problems, Internat. Math. Res. Not. 2003,
No. 36, 19871998.
[68] R. Takloo-Bighash, A remark on a paper of
Ahlgren, Berndt, Yee, and Zaharescu, Internat. J.
Number Thy. 2 (2006), 111114.
[69] J. Urbanowicz and K. S. Williams, Congruences for
L-functions, Kluwer, Dordrecht, 2000.

30

[70] G. Vorono, Sur une fonction transcendante et ses

applications la sommation de quelques sries,
Ann. Sci. Ecole Nor. Sup. (3) 21 (1904), 207267;
459533.
[71] G. N. Watson, The final problem: an account of
the mock theta functions, J. London Math. Soc. 11
(1936), 5580; reprinted in [27, pp. 325347].
[72]
, The mock theta functions (2), Proc. London
Math. Soc. 42 (1937), 274304.
, A Treatise on the Theory of Bessel Functions,
[73]
2nd ed., University Press, Cambridge, 1966.
[74] Y. Yang, On integrals of Eisenstein series and
derivatives of L-series, Internat. Math. Res. Not.
2004, No. 6, 303307.
[75] D. Zagier, Vassilier invariants and a strange identity related to the Dedekind eta-function, Topology
40 (2001), 945960.
[76] S. P. Zwegers, Mock Theta Functions, Ph.D. Thesis,
Universiteit Utrecht, 2002.

Notices of the AMS

Volume 55, Number 1