Current Developments in Mathematics, 2008

Unearthing the Visions of a Master: Harmonic

Maass Forms and Number Theory
Ken Ono

For my teacher Basil Gordon on his 75th birthday

Abstract. Together with his collaborators, most notably Kathrin

Bringmann and Jan Bruinier, the author has been researching harmonic
Maass forms. These non-holomorphic modular forms play central roles
in many subjects: arithmetic geometry, combinatorics, modular forms,
and mathematical physics. Here we outline the general facets of the
theory, and we give several applications to number theory: partitions
and q-series, modular forms, singular moduli, Borcherds products, ex-
tensions of theorems of Kohnen-Zagier and Waldspurger on modular
L-functions, and the work of Bruinier and Yang on Gross-Zagier formu-
lae. What is surprising is that this story has an unlikely beginning: the
pursuit of the solution to a great mathematical mystery.

Modular forms are central in contemporary mathematics. Indeed, mod-

ular forms play crucial roles in algebraic number theory, algebraic topology,
arithmetic geometry, combinatorics, number theory, representation theory,
and mathematical physics. The recent history of the subject includes (to
name a few) great successes on the Birch and Swinnerton-Dyer Conjecture,
Mirror Symmetry, Monstrous Moonshine, and the proof of Fermat’s Last
Theorem. These celebrated works are dramatic examples of the evolution of
mathematics; indeed, it would have been impossible to prophesy them fifty
years ago.
Instead of travelling back in time to the 1950s, our story (also see [165])
begins in 1887, in a village in India. Our mathematics,1 which is about har-
monic Maass forms, begins with the legend (see [6, 39, 40, 110, 111, 133,
165]) of the great mathematician Srinivasa Ramanujan, and the mathemat-
ics he conjured from his death bed.
The author thanks the support of the NSF, the Manasse family, and the Hilldale
Foundation.
1
This is an exposition of the author’s lectures at the ’08 Harvard-MIT CDM
Conference.

2009
c International Press
347
348 K. ONO

1. The gift from Kumbakonam

Ramanujan was born on December 22, 1887 in Erode, a town about 250
miles southwest of Chennai (formerly known as Madras). He was a Brahmin,
a member of India’s priestly caste. He grew up in Kumbakonam, a small
town roughly 150 miles south of Chennai, where his father Srinivasa was
a cloth merchant’s clerk. Kumbakonam, which is situated on the banks of
the sacred Cauvery River, was (and remains today) a cosmopolitan center
of the rural Indian district of Tanjore in the state of Tamil Nadu.
As a boy, Ramanujan was a stellar student. Indeed, he won many
awards2 at Town High School in Kumbakonam. Thanks to his exemplary
performance, Ramanujan won a scholarship to Government College. His
life took a dramatic turn when a friend loaned him a copy of G. S. Carr’s
Synopsis of elementary results in pure mathematics, which G. H. Hardy
would later describe (see page 3 of [111]) as
“...the ‘synopsis’ it professes to be. It contains enunciations3 of 6165 theo-
rems, systematically and quite scientifically arranged, with proofs which are
often little more than cross-references...”
The amateur Ramanujan became infatuated with mathematics research,
recording his “discoveries” in notebooks, imitating Carr’s format. He typi-
cally offered no proofs of any kind. His addiction for mathematics made it
impossible for him to focus on schoolwork at Government College, and he
unceremoniously flunked out. He would later get a second chance, a schol-
arship to attend Pachaiyappa’s College in Madras. His obsession again kept
him from his school work, and he flunked out a second time (see [41]).
By 1907, the gifted Ramanujan was an academic failure. There was no
room for him in India’s system of higher education. His friends and parents
must have recognized his genius, and, to their credit, they continued to
support him by allowing him to work on mathematics full bore. There are
vivid accounts (for example, see page 67 of [133]) of Ramanujan hunched
over his slate on the porch of his house, and in the halls of Sarangapani
Temple working feverishly on mathematics.
“....Ramanujan would sit working on the pial (porch) of his house. . . , legs
pulled into his body, a large slate spread across his lap, madly scribbling,. . .
When he figured something out, he sometimes seemed to talk to himself,
smile, and shake his head with pleasure.”
For the next few years, Ramanujan continued his research in near isola-
tion.4 Thankfully, Ramanujan secured a job as a clerk at the Madras Port
Trust which provided a salary, and left time for mathematics. Although

2
Present day visitors to Town High School can view copies of these treasured awards.
3B. C. Berndt has informed me that there are actually 4417 theorems in Carr’s book.
4He did share his work with S. Narayana Aiyar, a friend who earned a Master’s in
mathematics.
 HARMONIC MAASS FORMS AND NUMBER THEORY 349

some Indian patrons acknowledged his genius, he was unable to find suit-
able mentors since Indian mathematicians did not understand his work. He
was unable to share his passion and raison d’être with others.
After years in purgatory, Ramanujan boldly wrote distinguished English
mathematicians. He wrote M. J. M. Hill, and then H. F. Baker and E. W.
Hobson, without success. Then on January 16, 1913, he wrote G. H. Hardy, a
young analyst and number theorist at Cambridge University. With his letter
he included nine pages of mathematical scrawl. C. P. Snow later elegantly
recounted (see pages 30–33 of [112]) Hardy’s reaction to the letter:
“One morning in 1913, he (Hardy) found, among the letters on his breakfast
table, a large untidy envelope decorated with Indian stamps. When he opened
it...he found line after line of symbols. He glanced at them without enthu-
siasm. He was by this time...a world famous mathematician, and ...he was
accustomed to receiving manuscripts from strangers. ....The script appeared
to consist of theorems, most of them wild or fantastic... There were no proofs
of any kind... A fraud or genius? ...is a fraud of genius more probable than
an unknown mathematician of genius? ...He decided that Ramanujan was,
in terms of ...genius, in the class of Gauss and Euler...”
In response to the letter, Hardy invited Ramanujan to Cambridge.
Although Hindu beliefs forbade such travel at the time, Ramanujan’s mother
had a vision from Goddess Namagiri granting him permission to accept the
invitation. Ramanujan left his life in south India for Cambridge University,
home of some of the world’s most distinguished mathematicians. He arrived
on April 14, 1914.
Over the next five years, Ramanujan published extensively on a wide
variety of topics: hypergeometric series, elliptic functions, modular forms,
probabilistic number theory, the theory of partitions and q-series, among
others. He would write over thirty papers, including seven with Hardy. He
was named a Fellow of Trinity College, and he was elected a Fellow of the
Royal Society (F.R.S.). News of his election spread quickly, and in India he
was hailed as a national hero.

Ramanujan at his Master’s graduation ceremony

350 K. ONO

Sadly, the story of Ramanujan’s life ends tragically. He fell ill towards
the end of his stay in England. He returned to south India in the spring of
1919 seeking a return to health and a forgiving climate. Unfortunately, his
health steadily declined over the course of the ensuing year, and he died on
April 26, 1920 in Madras. He was thirty two years old. Amazingly, despite
his illness, he spent the last year of his life, again in mathematical isolation,
conjuring a most beautiful theory, one which was nearly lost to the world.

Ramanujan’s 1919 passport photograph5

The legend of Ramanujan has continued to grow with the ever-increasing

importance of his mathematics. To explain, it is appropriate to recall
Hardy’s own words6 written shortly after Ramanujan’s death:
“Opinions may differ about the importance of Ramanujan’s work, the kind of
standard by which it should be judged, and the influence which it is likely to
have on mathematics of the future. ...He would probably have been a greater
mathematician if he could have been caught and tamed a little in his youth.
On the other hand he would have been less of a Ramanujan, and more of a
European professor, and the loss might have been greater than the gain....”
In view of the last eighty five years of progress in number theory, it is
clear that the loss would have been much greater than the gain. On one
hand, as Hardy did, we may largely base our conclusion on the contents of
Ramanujan’s notebooks, which, apart from the “lost notebook” that con-
tained the work of his last year, were known to mathematicians at the time
of his death. They are a repository of thousands of cryptic entries on eval-
uations and identities of strangely named functions. Through the tireless
efforts of B. C. Berndt, adding to the accumulated effort of earlier mathe-
maticians such as Hardy, G. N. Watson, B. M. Wilson, and R. A. Rankin,
a clear picture has emerged which reveals Ramanujan’s incredible gift for
5These images illustrate Ramanujan’s poor health in 1919. He seems to have lost
considerable weight.
6See page xxxvi of [110].
 HARMONIC MAASS FORMS AND NUMBER THEORY 351

formulas and combinatorial relations (see [32, 33, 34, 35, 36, 111, 175]).
Perhaps he was the greatest such mathematician of all time.
On the other hand, this flattering assessment is grossly inadequate, for it
does not take into account Ramanujan’s impact on Hardy’s “mathematics
of the future”. Indeed, number theory has undergone a tremendous evo-
lution since Ramanujan’s death, and today it bears no resemblance to the
number theory of his day. The subject is now dominated by the arithmetic
and analytic theory of automorphic and modular forms, the study of Dio-
phantine questions under the rubric of arithmetical algebraic geometry, and
the emergence of computational number theory and its applications. These
subjects boast many of the most celebrated achievements of 20th century
mathematics such as: Deligne’s proof of the Weil Conjectures, the solution
to Gauss’ Class Number Problem by Goldfeld, Gross, and Zagier, Wiles’
proof of Fermat’s Last Theorem, and Borcherds’s work on the infinite prod-
uct expansions of automorphic forms. A proper assessment of Ramanujan’s
greatness must then take into account the remarkable fact that his work,
the portion which was known to Hardy, makes intimate contact with all of
these notable achievements. Clearly, Ramanujan was a great anticipator.7
His work provided examples of deeper structures, and suggested important
questions which are now inescapable in the panorama of modern number
theory (see [165]).
This brings us to the mysterious mathematics Ramanujan developed
during his last year, which was nearly lost. For nearly 60 years, the only
information about this work was contained in the surviving portion of
Ramanujan’s last letter to Hardy, dated January 20, 1920. The task of
teasing his theory from this letter has been one of the greatest challen-
ges for Ramanujan historians. Then in 1976, G. E. Andrews unearthed
Ramanujan’s “lost notebook”, the sheafs of mathematical scrawl containing
the results Ramanujan described in this last letter. What are the secrets of
the mathematical scrawl Ramanujan penned during his last days? What is
its impact on Hardy’s “mathematics of the future”?

2. Ramanujan’s playground
The new mathematics described in this paper is born from the enigmatic
theory Ramanujan conjured during the last year of his life. To set the stage,
it makes sense to first ponder some of Ramanujan’s early work which we
think led him to conceive this theory.
At first glance, the stuff of partitions seems like mere child’s play:
4 = 3 + 1 = 2 + 2 = 2 + 1 + 1 = 1 + 1 + 1 + 1.
Therefore, there are 5 partitions of the number 4. As happens in num-
ber theory, seemingly simple problems, such as the business of adding and
counting, can quickly lead to difficult and beautiful problems (for example,
7I first heard Ramanujan described this way by Manjul Bhargava.
352 K. ONO

see [4, 16, 24]). A partition of the natural number n is any non-increasing
sequence of natural numbers whose sum is n. The number of partitions of
n is denoted by p(n) (by convention, we let p(0) := 1).
2.1. Ramanujan’s work on partitions. Ramanujan was perhaps the
first mathematician to seriously investigate the properties of this function
p(n). He sought a formula for p(n), one which describes the phenomenal
rate of growth suggested by the table below.

n p(n)

0 1
1 1
2 2
3 3
4 5
5 7
.. ..
. .
50 204226
.. ..
. .
200 3972999029388
.. ..
. .
1000 24061467864032622473692149727991
Together with Hardy, Ramanujan gave a remarkable asymptotic formula
[113, 114], which was perfected by H. Rademacher [168, 169] two decades
later to obtain a formula which is so accurate that it can be used to com-
pute individual values of p(n). More precisely, Rademacher defined explicit
functions Tk (n) such that for all positive n we have
∞

(2.1) p(n) = Tk (n).
k=1
The function T1 (n) alone gives the Hardy-Ramanujan asymptotic formula
1 √
(2.2) p(n) ∼ √ · eπ 2n/3 .
4n 3
The rate at which Rademacher’s series converges is astonishing; for example,
the first eight terms give the approximation
p(200) ≈ 3, 972, 999, 029, 388.004,
which nicely compares with the exact value in the table above. These works
stand out further in importance since they mark the birth of the circle
method, now one of the fundamental tools in analytic number theory.
Ramanujan also investigated the divisibility properties of p(n). At first
glance, the combinatorial definition of the partition function gives no reason
 HARMONIC MAASS FORMS AND NUMBER THEORY 353

to believe that it possesses any interesting arithmetic properties. There is

nothing, for example, which would lead us to think that p(n) should exhibit
a preference to be even rather than odd. A natural suspicion might be
that the values of p(n) are distributed evenly modulo 2. Numerics seem
to confirm this suspicion, as well as its obvious generalization when 2 is
replaced by 3. However, when we replace 3 by 5, something quite different
happens; we discover that 3,611 (many more than the expected one-fifth) of
the first 10,000 values of p(n) are divisible by 5.
The explanation must have been clear to Ramanujan when he first saw
the famous table of values of p(n) computed by P. A. MacMahon, a well-
known combinatorialist and major in the British Royal Artillery. These
values, starting with n = 0, were listed in five columns, and so Ramanujan
would have seen something like the following.

1 1 2 3 5

7 11 15 22 30

42 56 77 101 135

176 231 297 385 490

627 792 1002 1255 1575

1958 2436 3010 3718 4565.

What is striking, of course, is that every entry in the last column is a

multiple of 5. This phenomenon, which persists, explains the apparent aber-
ration above, and was the first of Ramanujan’s ground-breaking discoveries
on the arithmetic of p(n). He proved [173, 174], for every non-negative
integer n, that
p(5n + 4) ≡ 0 (mod 5),
p(7n + 5) ≡ 0 (mod 7),
p(11n + 6) ≡ 0 (mod 11).
Ramanujan conjectured (and in some cases proved) [38, 173, 174] that
there are further congruence properties in which the moduli are powers of
5, 7, or 11.
Subsequent works [28, 208] by A. O. L. Atkin and Watson resolved
the conjectured congruences of Ramanujan, and in an important paper [29]
Atkin discovered completely new congruences modulo some further small
primes. The author proved [162] that Atkin’s phenomenon generalizes to
all prime moduli exceeding 3, and in later work Ahlgren and the author
further extended [2, 3] these results to include all moduli coprime to 6. In
354 K. ONO

particular, it turns out that there are such Ramanujan-type congruences for
every modulus M coprime to 6. For example, we have that
(2.3) p(4063467631n + 30064597) ≡ 0 (mod 31).
This comprehensive theory requires deep works of Deligne, Serre, and
Shimura [83, 187, 197].

2.2. Ramanujan’s playground is a testing ground. Ramanujan’s

work on p(n), and the research it inspired, underscores the fact that the the-
ory of partitions has historically served as a delightful “testing ground” for
some of the deepest developments in the theory of modular forms. Indeed,
(2.1) and (2.2) are the first triumphs of the circle method, while congruences
such as (2.3) arise from the interplay between the Deligne-Serre theory of
-adic Galois representations, the “language” of the proof of Fermat’s Last
Theorem, and Shimura’s theory of half-integral weight modular forms.
To make this connection between partitions and modular forms, we sim-
ply require Euler’s combinatorial generating function
∞
 ∞
 1
(2.4) n
p(n)q = = 1 + q + 2q 2 + 3q 3 + 5q 4 + 7q 5 + · · · .
1 − qn
n=0 n=1

This formal power series is essentially a “weakly holomorphic modular form”.

 a bit more precise, let Γ ⊂ SL2 (Z) be a subgroup. An
To make this
element γ = ac db ∈ SL2 (Z) acts on H, the upper-half of the complex
plane, by the linear fractional transformation γz := az+b
cz+d . Loosely speaking,
a weight k modular form on Γ is a holomorphic function f on H which
satisfies
(2.5) f (γz) = (cz + d)k f (z)
for all γ ∈ Γ, which in addition is holomorphic “at the cusps”. A weakly
holomorphic modular form satisfies this definition but is permitted to have
poles at cusps.
At the heart of the matter is Dedekind’s weight 1/2 modular form, the
eta-function:
∞

η(z) := q 1/24 (1 − q n ),
n=1

where q := e2πiz . For z ∈ H, it turns out that

πi 1
η(z + 1) = e 12 η(z) and η(−1/z) = (−iz) 2 η(z).
Since SL2 (Z) is generated by
   
1 1 0 −1
T := and S := ,
0 1 1 0
 HARMONIC MAASS FORMS AND NUMBER THEORY 355

it follows that η(z) is a weight 1/2 modular form on SL2 (Z) with a “multiplier
system”. Since η(z) is nonvanishing on H, (2.4) then implies that
∞

1 1
(2.6) = p(n)q n− 24
η(z)
n=0
is a weight −1/2 weakly holomorphic modular form.
2.3. How did Ramanujan think of modular forms? Since we view
Ramanujan’s results on the partition function, and the research it inspired,
as a testing ground for the theory of modular forms, we are compelled to ask
how Ramanujan thought of modular forms. A brief inspection of most of
his works (for example, see [38, 175]) suggests that his view was consistent
with the classical treatment found in most textbooks (for example, see [71,
84, 125, 134, 143, 155, 164, 177, 185, 193, 196]) on modular forms.
Indeed, the bulk of his work on modular forms depends on the properties of
the q-series
∞

P (q) := 1 − 24 σ1 (n)q n ,
n=1
∞
Q(q) := 1 + 240 σ3 (n)q n ,
n=1
∞
R(q) := 1 − 504 σ5 (n)q n ,
n=1

(where σν (n) := d|n dν ),
which are the classical Eisenstein series E2 (z),
E4 (z) and E6 (z) from the theory of modular forms. Therefore, much of his
work followed classical lines.
This assessment, however, ignores Ramanujan’s extensive work on com-
binatorial q-series, which at first glance have nothing to do with the theory of
modular forms. Ramanujan, a master manipulator of power series, thought
deeply about combinatorial power series such as infinite products, which are
often modular like Dedekind’s eta-function η(z), and Jacobi’s identity
∞
 
(1 − q 2n )(1 + t2 q 2n−1 )(1 + t−2 q 2n−1 ) =
2
t2n q n .
n=1 n∈Z
He also thought deeply about power series he referred to as Eulerian series,
such as
∞
 2
qn
(2.7) Ω(t; q) := 1 + ,
(1 − tq)2 (1 − tq 2 )2 · · · (1 − tq n )2
n=1
which seems to have nothing to do with modular forms. However, in some
rare cases such series turn out to coincide with modular forms.
We now give one such coincidence, an identity which is particularly
relevant for Ramanujan’s work on partitions. We first recall some elementary
356 K. ONO

combinatorial notions. One may visualize a partition λ1 + λ2 + · · · + λk as a

Ferrers diagram, a left justified array of dots consisting of k rows in which
there are λi dots in the ith row:

• • ··· • λ1 nodes

• • ··· • λ2 nodes

.. ..
. .

• ··· • λk nodes.

The Durfee square of such a Ferrers diagram is the largest square of nodes in
the upper left hand corner of the diagram. The boundary of a Durfee square
naturally then divides a partition into a perfect square and two partitions
whose parts do not exceed the side length of the Durfee square.
Example 2.1. The Ferrers diagram of the partition 5+5+3+3+2+1 is:
.
• • • .. • •

..
• • • . • •

.
• • • ..
··· ··· ···
• • •

• •

•
Therefore, this partition decomposes as the Durfee square of size 9, and the
two partitions 2 + 2, and 3 + 2 + 1.
Armed with these notions, we prove the following q-series identity for
the generating function for p(n).
Theorem 2.2. The following combinatorial identity is true:
∞
 ∞
  ∞ 2
1 qm
p(n)q n = = 1 + = Ω(1; q).
1 − qn (1 − q)2 (1 − q 2 )2 · · · (1 − q m )2
n=0 n=1 m=1
In particular, we have that
∞ ∞
1   q m − 24
2 1
1 1
= q − 24 − 24
1
= q + .
η(z) 1 − qn (1 − q)2 (1 − q 2 )2 · · · (1 − q m )2
n=1 m=1
 HARMONIC MAASS FORMS AND NUMBER THEORY 357

Proof. For every positive integer m, the q-series

∞

1
= am (n)q n
(1 − q)(1 − q 2 ) · · · (1 − q m )
n=0
is the generating function for am (n), the number of partitions of n whose
summands do not exceed m. Therefore by the discussion above, the q-series
2 ∞

qm
= bm (n)q n
(1 − q)2 (1 − q 2 )2 · · · (1 − q m )2
n=0
is the generating function for bm (n), the number of partitions of n whose
Ferrers diagram has a Durfee square of size m2 . The theorem follows by
summing in m. 
Although they are quite rare, there are further examples of such q-series
identities in which an Eulerian series is essentially a modular form. Among
them, perhaps the most famous are the Rogers-Ramanujan identities [171,
182, 183, 184]:
∞ ∞ 2
1 qn
= 1 + ,
(1 − q 5n+1 )(1 − q 5n+4 ) (1 − q)(1 − q 2 ) · · · (1 − q n )
n=0 n=1
(2.8) ∞ ∞
 1  2
q n +n
= 1 + .
(1 − q 5n+2 )(1 − q 5n+3 ) (1 − q)(1 − q 2 ) · · · (1 − q n )
n=0 n=1
These infinite products are essentially the Fourier expansions of weight 0
weakly holomorphic modular forms, the type studied classically by Jacobi,
Klein, and Siegel, and again more recently by Kubert and Lang in their
work on modular units (for example, see [139, 186]).
Remark 1. The literature on such identities is extensive (for example,
see works by Andrews, Gordon, Göllnitz, and Slater [8, 102, 101, 195] to
name a few), and the pursuit of further identities and their interpretations
remains an active area of research largely due to applications in combina-
torics, Lie theory, number theory and physics (for example, see [13], [129]
and [145] to name a few). In this direction, W. Nahm [159] has very in-
teresting work related to the question of when a basic hypergeometric-type
series is automorphic.
These identities stand out since there is no reason to believe, for function
theoretic reasons, that an Eulerian series should ever be modular. Indeed,
there is no general theory of transformation laws for Eulerian series.
It is not difficult to imagine Ramanujan’s mindset. Understanding this
quandary, it seems that Ramanujan, largely motivated by his work on par-
titions and the Rogers-Ramanujan identities, spent the last year of his life
thinking deeply about the “near” modularity of Eulerian series. He under-
stood the importance of developing a “new theory”, one which overlaps in
spots with the classical theory of modular forms.
He discovered the mock theta functions.
358 K. ONO

3. Ramanujan’s last letter to Hardy

Hardy’s romantic story of Ramanujan opens with the delivery of the
landmark letter from Ramanujan in 1913. Our tale begins with Ramanujan’s
last letter to Hardy, dated January 12, 1920, just three months before his
death. We quote (see pages 220–224 of [39]):
“I am extremely sorry for not writing you a single letter up to now.. . . I dis-
covered very interesting functions recently which I call “Mock” ϑ-functions.
Unlike the “False” ϑ-functions (studied partially by Prof. Rogers in his in-
teresting paper) they enter into mathematics as beautifully as the ordinary
theta functions. I am sending you with this letter some examples.”
This letter contained 17 examples including:
∞
 2
qn
f (q) := 1 + ,
(1 + q)2 (1 + q 2 )2 · · · (1 + q n )2
n=1
∞
 2
q 2n +2n
(3.1) ω(q) := ,
(1 − q) (1 − q 3 )2 · · · (1 − q 2n+1 )2
2
n=0
∞
(−1)n (1 − q)(1 − q 3 ) · · · (1 − q 2n−1 )q n
λ(q) := .
(1 + q)(1 + q 2 ) · · · (1 + q n−1 )
n=0

At first glance, these series indeed seem to resemble the Eulerian series in
Theorem 2.2 and (2.8). For example, in terms of (2.7), we have that both
1
= q − 24 Ω(1; q).
1
f (q) = Ω(−1; q) and
η(z)
However, as Ramanujan asserts, series such as f (q), ω(q) and λ(q) are not
modular; they are what he calls mock theta functions.
Most of the surviving text of the letter, which amounts to roughly 4
typewritten pages, consists of explicit formulas for these 17 strange for-
mal power series. The theoretical content is rather obtuse, and consists of
an elementary discussion on the asymptotics of Eulerian series and their
behavior near points on the unit disk. He gives no indication of how he
derived his 17 examples. He even divides these examples into groups based
on their “order”, a term he never defines. With such flimsy clues, how could
one rederive Ramanujan’s theory? What did he mean by a mock theta
function?
Despite these formidable challenges, a few mathematicians such as
G. E. Andrews, L. Dragonette,8 A. Selberg, and Watson [7, 85, 190, 206,
207] investigated Ramanujan’s mock theta functions for what they were,
a list of enigmatic power series. For example, Andrews and Dragonette

8Leila Dragonette is better known under her married name, Leila Bram.
 HARMONIC MAASS FORMS AND NUMBER THEORY 359

[7, 85] investigated Ramanujan’s claimed asymptotic formula for the mock
theta function
(3.2)
∞
 ∞
 2
qn
f (q) = Ω(−1; q) = α(n)q n := 1 + .
(1 + q)2 (1 + q 2 )2 · · · (1 + q n )2
n=0 n=1

They proved, for positive n, that


(−1)n−1 π n
− 144
1
α(n) ∼ ·e 6 .
2 n − 24
1

Remark 2. Strictly speaking, Dragonette first proved the asymptotic

[85], while Andrews refined her work to obtain [7] the asymptotic with an
improved error term.
Despite the absence of a theory, or much less, just a simple useful def-
inition of a mock theta function, these few early works bolstered the belief
that Ramanujan had discovered something important. Watson, in his own
words,9 proclaimed [206]:
“Ramanujan’s discovery of the mock theta functions makes it obvious that
his skill and ingenuity did not desert him at the oncoming of his untimely
end. As much as any of his earlier work, the mock theta functions are an
achievement sufficient to cause his name to be held in lasting remembrance.
To his students such discoveries will be a source of delight and wonder until
the time shall come when we too shall make our journey to that Garden of
Proserpine (a.k.a. Persephone)...”

4. The “lost notebook” and Dyson’s “challenge for the future”

By the mid 1970s, little progress had been made on the mock theta func-
tions. They remained a list of enigmatic power series, without any apparent
connection to the theory of modular forms, or any other comprehensive
theory for that matter.
Then in the spring of 1976, Andrews discovered Ramanujan’s “lost note-
book” in an old box of papers hidden away in the Trinity College Library
at Cambridge University. This notebook, consisting of over 100 pages of
mathematical scrawl [176], was archived among papers from Watson’s es-
tate. Miraculously, the “lost notebook” had somehow survived a circuitous
journey from India in the early 1920s to lie forgotten in the Trinity College
Library archives. The journey was indeed miraculous, for the contents of
the box almost met a catastrophic end in 1968 when Rankin saved them
just a few days before they were scheduled to be burned. Although the

9This text is from Watson’s 1936 Presidential Address to the London Mathematical
Society.
360 K. ONO

manuscript was never truly lost, it was long forgotten, and buried among
Watson’s random papers. Andrews proclaimed [9]:
“. . . the fact that its existence was never mentioned by anyone for over 55
years leads me to call it “lost”.
The discovery of the “lost notebook” was the catalyst which made it
possible to begin chipping away at the puzzle of Ramanujan’s mock theta
functions. On top of giving 2 further mock theta functions, adding to the
17 from the last letter and 3 defined by Watson [206], the pages contained
many clues: striking identities and relations. As usual, these were given
without proof.
To illustrate the value of these clues, we consider some examples of
Ramanujan’s claimed identities, the famous “mock theta conjectures.” To
state them, we first fix notation. For non-negative integers n, let

n−1
(4.1) (x)n := (x; q)n := (1 − xq j ),
j=0
and let
∞

(4.2) (x)∞ := (x; q)∞ := (1 − xq j ),
j=0
where an empty product equals 1. Let f0 (q), f1 (q), Φ(q), and Ψ(q) be the
mock theta functions
∞
 2
qn
f0 (q) := ,
(−q)n
n=0
∞
 2
q 5n
Φ(q) := −1 +
(q; q 5 )n+1 (q 4 ; q 5 )n
n=0
(4.3) ∞
 2
q n +n
f1 (q) := ,
(−q)n
n=0
∞
 2
q 5n
Ψ(q) := −1 + .
(q 2 ; q 5 )n+1 (q 3 ; q 5 )n
n=0
These q-series resemble (2.7) and the Eulerian series in the Rogers-
Ramanujan identities (2.8). The mock theta conjectures are a list of ten
identities involving these 4 functions. Thanks to work of Andrews and F.
Garvan [26], these ten identities follow from the truth of the following pair
of identities.
Conjecture (Mock Theta Conjectures). The following identities are
true:
(q 5 ; q 5 )∞ (q 5 ; q 10 )∞
= f0 (q) + 2Φ(q 2 ),
(q; q 5 )∞ (q 4 ; q 5 )∞
(q 5 ; q 5 )∞ (q 5 ; q 10 )∞
= f1 (q) + 2q −1 Ψ(q 2 ).
(q 2 ; q 5 )∞ (q 3 ; q 5 )∞
 HARMONIC MAASS FORMS AND NUMBER THEORY 361

These surprising identities equate simple infinite products to linear

combinations of mock theta functions. These infinite products, like those
in Theorem 2.2 and (2.8), are well known to essentially coincide with the
Fourier expansions of certain weakly holomorphic modular forms. Therefore,
the truth of these identities directly related mock theta functions to modu-
lar forms. These clues from the “lost notebook” finally placed Ramanujan’s
mock theta functions in the vicinity of the theory of modular forms.
Unfortunately, these clues would not prove to be enough to rederive
Ramanujan’s theory. Indeed, mathematicians even had to wait 10 years,
until the work of D. Hickerson [118], just for a proof of these particularly
vicious q-series identities.
Nonetheless, Andrews’s discovery of the “lost notebook” in 1976 almost
immediately sparked an explosion of research on the mock theta functions,
largely spearheaded by Andrews. By the late 1990s, works by Andrews, Y.-S.
Choi, H. Cohen, F. Dyson, Garvan, B. Gordon, Hickerson, R. McIntosh,
M. Wakimoto [9, 11, 12, 14, 15, 21, 26, 27, 72, 73, 74, 75, 76, 80,
103, 104, 118, 119, 204], among numerous others, revealed many of the
deeper properties of the mock theta functions. Thanks to these works, and
too many others to list, Ramanujan’s 22 mock theta functions had been
related to a surprising collection of subjects:

• Artin L-functions in number theory

• Hypergeometric functions
• Partitions
• Lie theory
• Mordell integrals
• Modular forms
• Polymer chemistry
• ...

Despite this flurry of activity, the essence of Ramanujan’s theory re-

mained a mystery. The puzzle of his last letter to Hardy, thanks to the “lost
notebook,” had morphed into the enigmatic web of Ramanujan’s 22 mock
theta functions. The presence of this web strongly suggested the existence
of a theory, and it also demanded a solution. In his plenary address at
the Ramanujan Centenary Conference at the University of Illinois in 1987,
Freeman Dyson beautifully summed up the situation [91]:
“The mock theta-functions give us tantalizing hints of a grand synthesis still
to be discovered. Somehow it should be possible to build them into a coher-
ent group-theoretical structure, analogous to the structure of modular forms
which Hecke built around the old theta-functions of Jacobi. This remains a
challenge for the future. My dream is that I will live to see the day when our
young physicists, struggling to bring the predictions of superstring theory into
correspondence with the facts of nature, will be led to enlarge their analytic
machinery to include mock theta-functions...But before this can happen, the
362 K. ONO

purely mathematical exploration of the mock-modular forms and their mock-

symmetries must be carried a great deal further.”

5. Back to the future

By the late 1990s, the vast literature on Ramanujan’s mock theta func-
tions contained many important clues for Dyson’s “challenge for the future”.
In addition to the identities comprising the mock theta conjectures, there
were further clues such as q-series identities relating mock theta functions to
Lambert-type series and indefinite theta series. We recall two such identities
involving the mock theta functions f (q) and f0 (q) (see (3.1) and (4.3)). In
1936, Watson [206] proved that
∞
 2
 (−1)n q (3n2 +n)/2
qn 2
f (q) = 1 + = · ,
(1 + q)2 (1 + q 2 )2 · · · (1 + q n )2 (q)∞ 1 + qn
n=1 n∈Z
and in 1986 Andrews [12] proved that
⎛ ⎞
∞
 qn2 1 ⎜   ⎟
f0 (q) = n
= ·⎜
⎝ − ⎟ (−1)j q 52 n2 + 12 n−j 2 .
⎠
(−q) (q)∞
n=0 n+j≥0 n+j<0
n−j≥0 n−j<0

Such identities served as motivation for the 2002 Ph.D. thesis, written
under the direction of D. Zagier, of S. Zwegers [218, 219]. Indeed, Zwegers
researched the following two questions of Zagier (see page 2 of [219]):
(1) How do the mock ϑ-functions fit in the theory of modular forms?
(2) Is there a theory of indefinite theta functions?
His thesis brilliantly addressed these questions by combining and ex-
tending ideas from a number of sources such as works of Lerch [146, 147]
on the functions
 (−1)n eπi(n2 +n)τ +2πinν
(5.1) ,
1 − e2πinτ +2πiu
n∈Z
where τ ∈ H, ν ∈ C, u ∈ C/(Zτ + Z), work of L. Göttsche and Zagier
on indefinite ϑ-functions [105], the theory of Jacobi forms [93], Mordell
integrals [157, 158], and works of Andrews [10, 12, 14, 15].
Zwegers related Ramanujan’s mock theta functions to real analytic
vector-valued modular forms. Loosely speaking, he “completed” Ramanu-
jan’s mock theta functions by adding a non-holomorphic function, a so-called
period integral, to obtain real analytic functions which obey desired modular
transformation laws. We shall recount some of his work10 in Section 6. The
real analytic modular forms of Zwegers turned out to be examples of har-
monic Maass forms (see Section 7.1 for the definition) which were defined
10Zagier delivered a Séminaire Bourbaki lecture on these recent works on Ramanujan’s
mock theta functions in 2007 [213].
 HARMONIC MAASS FORMS AND NUMBER THEORY 363

about the same time by Bruinier and Funke [63], a coincidence which cat-
alyzed much of the research described here.
These developments sparked an immediate explosion in a wide number
of new directions. Indeed, recent works by the author, Andrews, Eguchi,
Hikami, Kac, Lawrence, Malmendier, Mellit, Okada, Wakimoto, and Zagier
[15, 18, 21, 26, 92, 120, 121, 122, 130, 131, 144, 152, 153, 211] apply
this theory to:

• Donaldson invariants
• Gauge theory
• Representation theory of Lie superalgebras
• Knot theory
• Mathematical physics
• Probability theory
• Topology.

The author and his collaborators have aimed to employ this new per-
spective to answer deep questions about many of the number theoretic topics
captured by the web of Ramanujan’s mock theta functions. Here we describe
the implications of this theory to:

• Partitions and q-series

• Modular forms
• Traces of singular moduli
• Borcherds products
• Modular L-functions à la Kohnen-Waldspurger and Kohnen-Zagier.

Remark 3. This paper is not intended to be a comprehensive treatise;

instead, it is only meant to serve as an overview of the main aspects of this
program.
The remainder of the paper is organized as follows. In Section 6, we
give one beautiful example of Zwegers work, as it pertains to Ramanujan’s
mock theta functions f (q) and ω(q), and we briefly describe his important
construction of a weight 1/2 non-holomorphic Jacobi form. In Section 7 we
give the formal definition of a harmonic Maass form, and we present many of
their basic properties. In Section 8 we give various examples of such forms.
In the remaining sections we summarize our results on the following topics:
Section 9 (Dyson-Ramanujan theory of partition congruences).
In an effort to provide a combinatorial explanation of Ramanujan’s con-
gruences
p(5n + 4) ≡ 0 (mod 5),
p(7n + 5) ≡ 0 (mod 7),
p(11n + 6) ≡ 0 (mod 11),
364 K. ONO

Dyson introduced [90] the so-called rank of a partition, a delightfully simple

statistic. The rank of a partition is defined to be its largest part minus the
number of its parts.
For example, the table below includes the ranks of the partitions of 4.

Partition Rank Rank mod 5

4 4−1=3 3
3+1 3−2=1 1
2+2 2−2=0 0
2+1+1 2 − 3 = −1 4
1+1+1+1 1 − 4 = −3 2
One observes that each residue class modulo 5 occurs exactly once in the
last column above. Based on further numerics, Dyson made the following
conjecture whose truth provides a combinatorial explanation of Ramanujan’s
congruences modulo 5 and 7.
Conjecture (1944, Dyson). The partitions of 5n + 4 (resp. 7n + 5)
form 5 (resp. 7) groups of equal size when sorted by their ranks modulo 5
(resp. 7).
In 1954, Atkin and H. P. F. Swinnerton-Dyer proved [31] Dyson’s con-
jecture.11 In view of the more general theory of partition congruences, which
includes examples such as
p(4063467631n + 30064597) ≡ 0 (mod 31),
it is natural to investigate the role that Dyson’s rank plays in the general
theory of partition congruences. Using a new class of harmonic Maass forms,
which will be described in Section 8.4, the author and K. Bringmann [47, 55]
have obtained general results in this direction. These works will be described
in Section 9.
Section 10 (Eulerian series as modular forms).
The Rogers-Ramanujan identities
∞
 ∞
 2
1 qn
= 1 + ,
(1 − q 5n+1 )(1 − q 5n+4 ) (1 − q)(1 − q 2 ) · · · (1 − q n )
n=0 n=1
∞ ∞ 2
1 q n +n
=1+ ,
(1 − q 5n+2 )(1 − q 5n+3 ) (1 − q)(1 − q 2 ) · · · (1 − q n )
n=0 n=1
and the mock theta conjectures provide examples of Eulerian series as modu-
lar forms. Thanks to the framework of the theory of harmonic Maass forms,
we have a better understanding of the modular transformation properties
of certain Eulerian series. In Section 10 we discuss work by the author,
11A short calculation reveals that the obvious generalization of the conjecture cannot
hold for 11.
 HARMONIC MAASS FORMS AND NUMBER THEORY 365

Bringmann, and R. Rhoades [57] on Eulerian series as modular forms. This

work immediately gives many new identities such as
 q n(n+1) (−q 2 ; q 2 )n  q n(n+1) (−q 2 ; q 2 )n (q 4 ; q 4 )5∞
2 + 2 = 2 .
(q; q 2 )n+1 (−q; q 2 )n+1 (q 2 ; q 2 )4∞
n≥0 n≥0

Section 11 (Exact formulas).

As described earlier, Rademacher perfected the Hardy-Ramanujan
asymptotic (2.2) to obtain an exact formula for p(n) (for example, see
[168, 169]). To state his formula, let Is (x) be the usual I-Bessel func-
tion of order s, and let e(x) := e2πix . Furthermore, if k ≥ 1 and n are
integers, then let
  x 
1 k 
(5.2) Ak (n) := χ12 (x) · e ,
2 12 12k
x (mod 24k)
x2 ≡−24n+1 (mod 24k)

where the sum runs over the residue classes modulo 24k, and where
 
12
(5.3) χ12 (x) := .
x
If n is a positive integer, then one version of Rademacher’s formula reads
∞
  √ 
− 34 Ak (n) π 24n − 1
p(n) = 2π(24n − 1) · I3 .
k 2 6k
k=1
We address the following classical partition problem.
Problem. Determine exact formulas for Ne (n) (resp. No (n)), the number
of partitions of n with even (resp. odd) Dyson rank.
Thanks to Rademacher’s formula, and the obvious fact that
p(n) = Ne (n) + No (n),
it turns out that this question is equivalent to the problem of deriving exact
formulas for the coefficients
α(n) = Ne (n) − No (n)
of the mock theta function f (q). This deduction is a simple modification of
the proof of Theorem 2.2.
As mentioned earlier, Andrews and Dragonette had already proved that

(−1)n−1 π n
− 144
1
α(n) ∼ ·e 6 .
2 n − 24
1

This result falls short of the problem of obtaining an exact formula, and as a
consequence represents the obstruction to obtaining formulas for Ne (n) and
No (n). In his plenary address “Partitions: At the interface of q-series and
modular forms”, delivered at the Millenial Number Theory Conference at
the University of Illinois in 2000, Andrews highlighted this classical problem
366 K. ONO

by promoting his conjecture12 of 1966 (see page 456 of [7], and Section 5 of
[17]) for the coefficients α(n).
Conjecture. (Andrews-Dragonette)
If n is a positive integer, then
 
∞ (−1) 2  A
k+1
n− k(1+(−1)k )  √ 
2k 4 π 24n − 1
α(n) = π(24n − 1)− 4
1
·I1 .
k 2 12k
k=1

The author and Bringmann [53] proved this conjecture using work of
Zwegers and the theory of Maass-Poincaré series. Since Ne (n) = (p(n) +
α(n))/2 and No (n) = (p(n) − α(n))/2, the proof of the conjecture, combined
with Rademacher’s exact formula, provides the desired formulas for Ne (n)
and No (n).
The proof of the Andrews-Dragonette Conjecture is a special case of a
more general problem.
Problem. Determine exact formulas for the coefficients of holomorphic
parts of harmonic Maass forms.
We shall also give general theorems which provide such exact formulas
for harmonic Maass forms with weight ≤ 1/2. These new results include the
classic results of Rademacher and Zuckerman [169, 170, 216, 217] which
give exact formulas for the Fourier coefficients of negative weight weakly
holomorphic modular forms. Indeed, these results follow since weakly holo-
morphic modular forms are harmonic Maass forms. They also recover some
results of Bruinier, Hejhal, and Niebur [62, 117, 160, 161] for harmonic
Maass forms of non-positive weight.
Section 12 (Applications to classical modular forms).
Using the properties of various differential operators, it is possible to
address old problems about classical modular forms. For example, we ad-
dress the classification of linear relations among cuspidal Poincaré series,
and we obtain a theorem which detects the vanishing of Hecke eigenvalues
for integer weight newforms. We also show that Lehmer’s Conjecture, which
asserts that none of the coefficients of the discriminant function
∞ ∞

Δ(z) = τ (n)q n := q (1 − q n )24 = q − 24q + 252q 3 − · · ·
n=1 n=1
vanish, is implied by the irrationality of the coefficients of the “holomorphic
part” of a certain Maass-Poincaré series.
The work on linear relations among cuspidal Poincaré series is contained
in Rhoades’s Ph.D. thesis [179] (also see [178]). The theorem which detects
vanishing Fourier coefficients is in a recent paper by J. H. Bruinier, Rhoades,
and the author [68]. These results are described in Section 12.

12This conjecture is suggested as a speculation by Dragonette in [85].

 HARMONIC MAASS FORMS AND NUMBER THEORY 367

Section 13 (Generating functions for singular moduli).

Let j(z) be the usual modular function for SL2 (Z)
j(z) = q −1 + 744 + 196884q + 21493760q 2 + · · · .
The values of modular functions such as j(z) at imaginary quadratic
arguments in H are known as singular moduli. Singular moduli are alge-
braic integers which play many roles in number theory. For example, they
generate class fields of imaginary quadratic fields, and they parameterize
isomorphism classes of elliptic curves with complex multiplication.
In an important paper [212], Zagier gave a new proof of Borcherds’ fa-
mous theorem on the infinite product expansions of integer weight modular
forms on SL2 (Z) with Heegner divisor. This proof, as well as all of the
results of [212], are connected to his beautiful observation that the generat-
ing functions for traces of singular moduli are essentially weight 3/2 weakly
holomorphic modular forms.
Zagier’s paper has inspired an extraordinary number of research pa-
pers with generalizations in a variety of directions in works by the author,
Bringmann, Bruinier, D. Choi, W. Duke, A. Folsom, J. Funke, O. Imamoḡlu,
P. Jenkins, D. Jeon, S.-Y. Kang, C. Kim, R. Masri, A. Miller, A. Pixton,
J. Rouse, and A. Toth [54, 58, 64, 65, 77, 78, 87, 88, 89, 95, 154]. In
Section 13 we describe one theorem which gives typical examples of such
generating functions. In addition to giving further generating functions for
singular moduli, we shall also describe a beautiful theorem of Duke [87]
related to the classical observation that
√
eπ 163
= 262537412640768743.9999999999992 . . . .
is nearly an integer.
Section 14 (Borcherds Products).
Recently, Borcherds [42, 43, 44] provided a striking description for
the exponents in the naive infinite product expansion of many modular
forms, those forms with a Heegner divisor. He proved that the exponents in
these infinite product expansions are certain coefficients of modular forms of
weight 1/2. For example, let c(n) denote the integer exponents one obtains
by expressing the classical Eisenstein series E4 (z) as an infinite product:
∞
 ∞

E4 (z) = 1 + 240 σ3 (n)q n = (1 − q)−240 (1 − q 2 )26760 · · · = (1 − q n )c(n) .
n=1 n=1

Borcherds proved that there is a weight 1/2 meromorphic modular form


G(z) = b(n)q n = q −3 + 4 − 240q + 26760q 4 + · · · − 4096240q 9 + . . .
n≥−3

with the property that c(n) = b(n2 ) for every positive integer n.
These results may be generalized where the Borcherds exponents are
given by certain Fourier coefficients of weight 1/2 harmonic Maass forms.
368 K. ONO

As a nice example, consider Ramanujan’s mock theta function (see (3.1))

∞ 2
q 2n +2n
ω(q) =
n=0
(q; q 2 )2n+1
1 q4 q 12
= + + + ··· .
(1 − q)2 (1 − q)2 (1 − q 3 )2 (1 − q)2 (1 − q 3 )2 (1 − q 5 )2
Define integers a(n) by the expression
  
a(n)q n := −2q 1/3 ω(q 1/2 ) + ω(−q 1/2 )
n∈Z+1/3

= −4 q 1/3 − 12 q 4/3 − 24 q 7/3 − 40 q 10/3 − . . . .

These new results on generalized Borcherds products imply that
∞  √ ( n )a(n2 /3)
 1 + −2q n − q 2n 3 √ √
√ = 1 − 8 −2q − (64 − 24 −2)q 2
(5.4) n=1 1 − −2q n − q 2n
√ √
+ (384 + 168 −2)q 3 + (64 − 1768 −2)q 4 + . . .
is a modular form on the congruence subgroup Γ0 (6).
A brief indication of these results is given in Section 14.
Section 15 (Derivatives and values of L-functions).
Once armed with a generalized Borcherds-type theorem, one is then able
to construct modular forms with a prescribed divisor, a twisted Heegner
divisor, with the additional property that one can determine the field of
definition of the product Fourier expansion. For√ example, the modular form
in (5.4) clearly has Fourier coefficients in Z[ −2], thanks to the fact that
Ramanujan’s mock theta function ω(q) has integer coefficients. This phe-
nomenon is extremely useful in arithmetic geometry since it allows us to
determine whether certain divisors vanish in the Jacobian of a modular
curve, a condition which plays a central role in the work of Gross and Zagier
on the Birch and Swinnerton-Dyer Conjecture. Here we explain how to
make use of these generalized Borcherds products to extend deep theorems
of Waldspurger and Kohnen.
In the 1980s, Waldspurger [205], and Kohnen and Zagier [135, 136,
137] established that certain half-integral weight modular forms serve as
generating functions of a new type. Using the Shimura correspondence
[192], they proved that certain coefficients of half-integral weight cusp forms
essentially are square-roots of central values of quadratic twists of modular
L-functions.
When the weight is 3/2, these results appear prominently in works re-
lated to the Birch and Swinnerton-Dyer Conjecture. For example, Tunnell
[202] made great use of explicit examples of these results in his work on the
ancient “congruent number problem”: the determination of those positive
integers which are areas of right triangles with rational sidelengths. More
generally, these results of Kohnen, Waldspurger and Zagier play central roles
 HARMONIC MAASS FORMS AND NUMBER THEORY 369

in the deep works of Gross, Zagier and Kohnen [108, 107] on the Birch and
Swinnerton-Dyer Conjecture.
The author and Bruinier [67] have generalized this theorem of
Waldspurger and Kohnen to prove that the Fourier coefficients of weight
1/2 harmonic Maass forms encode the vanishing and nonvanishing of both
the central values and derivatives of quadratic twists of weight 2 modular
L-functions.
Here we describe a special case of the main theorem. Suppose that
∞

(5.5) G(z) = BG (n)q n ∈ S2 (p)
n=1
is a weight 2 newform with prime level p. As usual, we let
∞
 BG (n)
(5.6) L(G, s) =
ns
n=1
denote its Hecke L-function. If Δ is a fundamental discriminant of a qua-
dratic field coprime to p, then we let L(G, χΔ , s) be the quadratic twist Hecke
L-function
∞
BG (n)χΔ (n)
(5.7) L(G, χΔ , s) = ,
ns
n=1
  √
where χΔ (•) := Δ • denotes the Kronecker character for Q( Δ). It is well
known that L(G, s) and L(G, χΔ , s) have functional equations relating their
values at s and 2 − s.
Here is a special case of the main result in [67].
Theorem 5.1. Assume the hypotheses and notation above. In addition,
suppose that the sign of the functional equation of L(G, s) is (G) = −1.
Then there is a weight 1/2 harmonic Maass form f (z) on Γ0 (4p) with
Fourier expansion
 
f (z) = c+ n
g (n)q + c− n
g (n)Γ(1/2; 4π|n|y)q ,
n
−∞ n<0

where y = Im(z) and Γ(α; t) is the usual incomplete Gamma-function (see

(7.5)), satisfying the following:
 
(1) If Δ < 0 is a fundamental discriminant for which Δ
p = 1, then

L(G, χΔ , 1) = αG · |Δ| · c− 2
g (Δ) ,
where αG is an explicit non-zero constant.  
(2) If Δ > 0 is a fundamental discriminant for which Δ
p = 1, then
L (G, χΔ , 1) = 0 if and only if c+
g (Δ) is algebraic.

Remark 4. Theorem 5.1 is a special case of the general result which

holds for all levels, and any arbitrary sign.
370 K. ONO

Example 5.2. Here we present an example which numerically illustrates

the most general form of Theorem 5.1 for the weight 2 newform G which
corresponds to the conductor 37 elliptic curve
E : y 2 = x3 + 10x2 − 20x + 8.
The table below includes some of the coefficients of a suitable f which were
numerically computed by F. Strömberg (also see [69]).

Δ c+
g (−Δ) L (E(Δ), 1) = L (G, χΔ , 1)

−3 1.0267149116 . . . 1.4792994920 . . .
−4 1.2205364009 . . . 1.8129978972 . . .
−7 1.6900297463 . . . 2.1107189801 . . .
.. .. ..
. . .
−136 −4.8392675993 . . . 5.7382407649 . . .
−139 −6 0
−151 −0.8313568817 . . . 6.6975085515 . . .
.. .. ..
. . .
−815 121.1944103120 . . . 4.7492583693 . . .
−823 312 0

Strictly speaking, the cases where Δ = −139 and −823 were not obtained
numerically. We have that L (E(−139), 1) = L (E(−823), 1) = 0 by the
Gross-Zagier formula. The evaluations c+ g (139) = −6 and cg (823) = 312
+

arise from explicit generalized Borcherds products (for example, see Example
8.3 of [67]). The rank 3 elliptic curve E(−139) is quite famous, for it was
used as input data for Goldfeld’s celebrated effective solution to Gauss’s
“Class Number Problem”. For the other Δ in the table, the derivatives are
non-vanishing and the coefficients c+ g (−Δ) are transcendental.

Theorem 5.1 relates the algebraicity of coefficients of harmonic Maass

forms to the vanishing of derivatives of modular L-functions. It is then
natural to ask whether these ideas can be used to exactly compute these
derivatives. In other words, can the theory of harmonic Maass forms be
used to obtain the deep formulas (as well as generalizations) of Gross and
Zagier which relate heights of Heegner points to such derivatives? Bruinier
and T. Yang have used these results, combined with their theory of Green’s
functions and theta lifts for harmonic Maass forms, to show that this is
indeed the case. In addition to formulating a deep conjecture about deriva-
tives of L-functions and heights (see Section 15.2), they have proved the
following striking theorem (see [70]).
Theorem 5.3. If G ∈ S2 (N ) is a weight 2 newform with the property
that the sign of the functional equation of L(G, s) is (G) = −1, then there
 HARMONIC MAASS FORMS AND NUMBER THEORY 371

is a weight 1/2 harmonic Maass form f , a weight 3/2 cusp form g, and a
Heegner divisor Z(f ) whose Neron-Tate height pairing is given by
√
2 N 

Z(f ), Z(f )N T = L G, 1).
πg2
Remark 5. To ease notation in Theorem 5.3, we did not describe the
relationship between G, g, f and Z(f ). Loosely speaking, they are related
as follows. We let g be a weight 3/2 cusp form whose image under the
Shimura correspondence is G. The harmonic Maass form f in Theorems 5.1
and 5.3 is selected so that its principal part is defined over the number field
generated by the coefficients of G, and also satisfies
ξ 1 (f ) = g−2 g.
2

Here ξ 1 is the differential operator considered in Lemma 7.4. The Heegner

2
divisor Z(f ) is then defined using the principal part of f .
These works shall be described in Section 15.

6. The mock theta functions of Ramanujan d’après Zwegers

Here we give a typical example of Zwegers’s results on Ramanujan’s
mock theta functions, and we briefly describe his construction of a weight
1/2 non-holomorphic Jacobi form.

6.1. Ramanujan’s mock theta functions f (q ) and ω(q). We first

consider Ramanujan’s third order mock theta function f (q). Important
results concerning its modularity properties were first obtained by G. N.
Watson in [206]. Although f (q) is not the Fourier expansion of a usual
meromorphic modular form, in this classic paper Watson determined its
complicated modular transformation properties. Watson’s modular trans-
formation formulas are difficult to grasp at first glance; they involve an-
other third order mock theta function, as well as terms arising from Mordell
integrals.
Zwegers [218] nicely packaged Watson’s results in terms of real analytic
vector-valued modular forms. We now describe this result. As before, let
ω(q) be the third order mock theta function defined by (3.1). Define the
vector-valued function F (z) by
(6.1)
F (z) = (F0 (z), F1 (z), F2 (z))T := (q − 24 f (q), 2q 3 ω(q 2 ), 2q 3 ω(−q 2 ))T .
1 1 1 1 1

Similarly, let G(z) be the vector-valued non-holomorphic function defined by

(6.2)
√  i∞
(g1 (τ ), g0 (τ ), −g2 (τ ))T
T
G(z) = (G0 (z), G1 (z), G2 (z)) := 2i 3  dτ,
−z −i(τ + z)
372 K. ONO

where the gi (τ ) are the cuspidal weight 3/2 theta functions

∞
  
n 1 3πi(n+ 1 )2 τ
g0 (τ ) := (−1) n + e 3 ,
n=−∞
3
∞  
1 3πi(n+ 1 )2 τ
(6.3) g1 (τ ) := − n+ e 6 ,
n=−∞
6
∞  
1 3πi(n+ 1 )2 τ
g2 (τ ) := n+ e 3 .
n=−∞
3

Using these vector-valued functions, Zwegers defined the vector-valued func-

tion H(z) by

(6.4) H(z) := F (z) − G(z).

The following description of H(z) is the main result of [218].

Theorem 6.1. (Zwegers)

The function H(z) is a vector-valued real analytic modular form of weight
1/2 satisfying
⎛ −1 ⎞
ζ24 0 0
H(z + 1) = ⎝ 0 0 ζ3 ⎠ H(z),
0 ζ3 0

⎛ ⎞
√ 0 1 0
H(−1/z) = −iz ·⎝1 0 0 ⎠ H(z),
0 0 −1

where ζn := e2πi/n . Furthermore, H(z) is an eigenfunction of the Casimir

∂2
operator Ω 1 := −4y 2 ∂z∂z ∂
+ iy ∂z + 3 with eigenvalue 16
3
, where z = x + iy,
 2   16 
∂z = 2 ∂x − i ∂y , and ∂z = 2 ∂x + i ∂y .
∂ 1 ∂ ∂ ∂ 1 ∂ ∂

Remark 6. This Casimir operator is essentially the weight 1/2 hyper-

bolic Laplacian operator (see (7.1)) which plays an important role in the
definition of harmonic Maass forms.

This beautiful theorem nicely describes the modular transformations of

both f (q) and ω(q). In particular, they are essentially the “holomorphic
parts” of components of a 3-dimensional vector-valued weight 1/2 real ana-
lytic modular form. The “non-holomorphic” parts of these components are
“period integrals” of classical weight 3/2 functions which turn out to be sin-
gle variable theta functions. The functions in Theorem 6.1 can be thought
of as prototypes for the theory of harmonic Maass forms.
 HARMONIC MAASS FORMS AND NUMBER THEORY 373

6.2. Zwegers’s weight 1/2 non-holomorphic Jacobi form. In his

thesis, Zwegers constructed weight 1/2 harmonic Maass forms by making use
of the transformation properties of Lerch sums. Here we briefly recall some
of these important results which address the difficult problem of constructing
weight 1/2 harmonic Maass forms.
For τ ∈ H, u, v ∈ C \ (Zτ + Z), Zwegers defined the function
z 1/2  (−w)n q n(n+1)/2
(6.5) μ(u, v; τ ) := · ,
ϑ(v; τ ) 1 − zq n
n∈Z

where z := e2πiu , w := e2πiv , q := e2πiτ and

 2 /2
(6.6) ϑ(v; τ ) := eπiν wν q ν .
ν∈Z+ 12

Remark 7. We stress that q := e2πiτ in this subsection, which is a brief

departure from our convention that q := e2πiz .
Zwegers (see Section 1.3 of [219]) proves that μ(u, v; τ ) satisfies the
following important properties.
Lemma 6.2. Assuming the notation above, we have that
μ(u, v; τ ) = μ(v, u; τ ),
μ(u + 1, v; τ ) = −μ(u, v; τ ),
z −1 wq − 12
μ(u + τ, v; τ ) = −μ(u, v; τ ) − iz − 2 w 2 q − 8 ,
1 1 1

μ(u, v; τ + 1) = ζ8−1 μ(u, v; τ ) (ζN := e2πi/N )

 
− 12 πi(u−v)2 /τ u v 1 1
(τ /i) e μ , ;− = −μ(u, v; τ ) + h(u − v; τ ),
τ τ τ 2i
where  ∞ πix2 τ −2πxz
e dx
h(z; τ ) := .
−∞ cosh πx
Remark 8. The integral h(z; τ ) is known as a Mordell integral.
Lemma 6.2 shows that μ(u, v; τ ) is nearly a weight 1/2 Jacobi form,
where τ is the modular variable. Zwegers then uses μ to construct weight
1/2 harmonic Maass forms. He achieves this by modifying μ to obtain a
function μ which he then uses as building blocks for such Maass forms. To
make this precise, for τ ∈ H and u ∈ C, let c := Im(u)/Im(τ ), and let
(6.7)
    
(−1)ν− 2 sgn(ν) − E (ν + c) 2Im(τ ) e−2πiνu q −ν /2 ,
1 2
R(u; τ ) :=
ν∈Z+ 12

where E(x) is the odd function

 x
e−πu du = sgn(x)(1 − β(x2 )),
2
(6.8) E(x) := 2
0
374 K. ONO

∞ 1
where for positive real x we let β(x) := x u− 2 e−πu du.
Using μ and R, Zwegers defines the real analytic function
i
(6.9) (u, v; τ ) := μ(u, v; τ ) + R(u − v; τ ).
μ
2
Zwegers construction of weight 1/2 harmonic Maass forms depends on the
following theorem (see Section 1.4 of [219]).
Theorem 6.3. Assuming the notation and hypotheses above, we have
that
(u, v; τ ) = μ
μ (v, u; τ ),
(u + 1, v; τ ) = z −1 wq − 2 μ
1
μ (u + τ, v; τ ) = − μ(u, v; τ ),
 
− 12 πi(u−v)2 /τ u v 1
(u, v; τ + 1) = −(τ /i) e
ζ8 μ 
μ , ;− =μ (u, v; τ ).
τ τ τ
 
Moreover, if A = αγ βδ ∈ SL2 (Z), then
 
u v ατ + β
= χ(A)−3 (γτ +δ) 2 e−πiγ(u−v) /(γτ +δ) · μ
1 2

μ , ; (u, v; τ ),
γτ + δ γτ + δ γτ + δ
 1

where χ(A) := η(Aτ )/ (γτ + δ) 2 η(τ ) .

Theorem 6.3 shows that μ(u, v; τ ) is essentially a weight 1/2 non-holomo-

rphic Jacobi form. In analogy with the classical theory of Jacobi forms, one
may then obtain harmonic Maass forms by making suitable specializations
for u and v by elements in Qτ +Q, and by multiplying by appropriate powers
of q. Without this result, it would be very difficult to explicitly construct
examples of weight 1/2 harmonic Maass forms.
Harmonic Maass forms of weight k are mapped to classical modular
forms (see Lemma 7.4), their so-called shadows, by the differential operator
∂
ξk := 2iy k ·
.
∂τ
The following lemma makes it clear that the shadows of the real analytic
 can be described in terms of weight 3/2 theta functions.
forms arising from μ
Lemma 6.4. [Lemma 1.8 of [219]] The function R is real analytic and
satisfies
∂R √
(u; τ ) = 2y − 2 e−2πc y ϑ(u; −τ ),
1 2

∂u
where c := Im(u)/Im(τ ). Moreover, we have that
∂ i 
R(aτ − b; τ ) = − √ e−2πa y (−1)ν− 2 (ν + a)e−πiν τ −2πiν(aτ −b) .
2 1 2

∂τ 2y 1
ν∈Z+ 2
 HARMONIC MAASS FORMS AND NUMBER THEORY 375

7. Harmonic Maass forms

For the remainder of the paper, we shall assume that the reader is
familiar with the classical theory of elliptic modular forms (for example,
see [71, 84, 125, 134, 143, 155, 164, 177, 185, 193, 196]).
D. Niebur [160, 161] and D. Hejhal [117] constructed certain non-
holomorphic Poincaré series which turn out to be examples of harmonic
Maass forms. Bruinier [61] made great use of these Poincaré series in his
early work on Borcherds lifts and Green’s functions. He then realized the
importance of developing a “theory of harmonic Maass forms” in its own
right. Later in joint work with Funke [63], he developed the fundamental
results of this theory, some of which we describe here. After making the
necessary definitions, we shall discuss Hecke operators and various differen-
tial operators. The interplay between harmonic Maass forms and classical
modular forms shall play an important role throughout this paper.
7.1. Definitions. In 1949, H. Maass introduced the notion of a Maass
form13 (see [149, 150]). He constructed these non-holomorphic automorphic
forms using Hecke characters of real quadratic fields, in analogy with Hecke’s
theory [115] of modular forms with complex multiplication (see [180] for a
modern treatment).
To define these functions, let Δ = Δ0 be the hyperbolic Laplacian
 2 
∂ ∂2
Δ := −y 2
+ ,
∂x2 ∂y 2
where z = x+iy ∈ H with x, y ∈ R. It is a second-order differential operator
which acts on functions on H, and it is invariant under the action of SL2 (R).
A Maass form for a subgroup Γ ⊂ SL2 (Z) is a smooth function f : H → C
satisfying:
 
(1) For every ac db ∈ Γ, we have
 
az + b
f = f (z).
cz + d
(2) We have that f is an eigenfunction of Δ.
(3) There is some N > 0 such that
f (x + iy) = O(y N )
as y → +∞.
Furthermore, we call f a Maass cusp form if
 1
f (z + x)dx = 0.
0
There is now a vast literature on Maass forms thanks to the works of many
authors such as Hejhal, Iwaniec, Maass, Roelcke, Selberg, Terras, Venkov,
13In analogy with the eigenvalue problem for the vibrating membrane, Maass referred
to these automorphic forms as Wellenformen, or waveforms.
376 K. ONO

among many others (for example, see [116, 117, 124, 126, 149, 150, 181,
191, 199, 200, 203]).
This paper concerns a generalization of this notion of Maass form. Fol-
lowing Bruinier and Funke [63], we define the notion of a harmonic Maass
form of weight k ∈ 12 Z as follows. As before, we let z = x + iy ∈ H with
x, y ∈ R. We define the weight k hyperbolic Laplacian Δk by
 2   
∂ ∂2 ∂ ∂
(7.1) Δk := −y 2
+ + iky +i .
∂x2 ∂y 2 ∂x ∂y
For odd integers d, define d by

1 if d ≡ 1 (mod 4),
(7.2) d :=
i if d ≡ 3 (mod 4).
Definition 7.1. If N is a positive integer (with 4 | N if k ∈ 12 Z \ Z),
then a weight k harmonic Maass form on Γ ∈ {Γ1 (N ), Γ0 (N )} is any smooth
function M : H → C satisfying the following:
 
(1) For all A = ac db ∈ Γ and all z ∈ H, we have
  
az + b (cz + d)k M (z) if k ∈ Z,
M =  c 2k −2k
cz + d d d (cz + d)k M (z) if k ∈ 12 Z \ Z.
  √
Here dc denotes the extended Legendre symbol, and z is the
principal branch of the holomorphic square root.
(2) We have that Δk M = 0. 
(3) There is a polynomial PM = n≤0 c+ (n)q n ∈ C[q −1 ] such that
M (z) − PM (z) = O(e−
y )
as y → +∞ for some  > 0. Analogous conditions are required at
all cusps.
Remark 9. Maass forms and classical modular forms are required to sat-
isfy moderate growth conditions at cusps, and it is for this reason that har-
monic Maass forms are often referred to as “harmonic weak Maass forms”.
The term “weak” refers to the relaxed condition Definition 7.1 (3) which
gives rise to a rich theory. For convenience, we use the terminology “har-
monic Maass form” instead of “harmonic weak Maass form”.
Remark 10. We refer to the polynomial PM as the principal part of
M (z) at ∞. Obviously, if PM is non-constant, then M (z) has exponential
growth at ∞. Similar remarks apply at all cusps.
Remark 11. Bruinier and Funke [63] define two types of harmonic
Maass forms based on varying the growth conditions at cusps. For a group
Γ, they refer to these spaces as Hk (Γ) and Hk+ (Γ). Definition 7.1 (3) corre-
sponds to their Hk+ (Γ) definition.
Remark 12. Since holomorphic functions on H are harmonic, it follows
that weakly holomorphic modular forms are harmonic Maass forms.
 HARMONIC MAASS FORMS AND NUMBER THEORY 377

Remark 13. Here we recall the congruence subgroups. If N is a positive

integer, then define the level N congruence subgroups Γ0 (N ), Γ1 (N ), and
Γ(N ) by
  
a b
Γ0 (N ) := ∈ SL2 (Z) : c ≡ 0 mod N ,
c d
  
a b
Γ1 (N ) := ∈ SL2 (Z) : a ≡ d ≡ 1 mod N, and c ≡ 0 mod N ,
c d
  
a b
Γ(N ) := ∈ SL2 (Z) : a ≡ d ≡ 1 mod N, and b ≡ c ≡ 0 mod N .
c d
Remark 14. For k ∈ 12 Z \ Z, the transformation law in Definition 7.1
(1) coincides with those in Shimura’s theory of half-integral weight modular
forms [192].
Remark 15. Later we shall require the classical “slash” operator. For
convenience,
a b we recall its definition here. Suppose that k ∈ 12 Z. For A =
c d ∈ SL2 (Z) (Γ0 (4) if k ∈ 2 Z \ Z), define j(A, z) by
1
√
cz + d if k ∈ Z,
(7.3) j(A, z) :=  c  −1 √
d εd cz + d if k ∈ 12 Z \ Z,
√
where εd is defined by (7.2), and where z is the principal branch of the
holomorphic square root as before. For functions f : H → C, we define the
action of the “slash” operator by
 
−2k −2k az + b
(7.4) (f |k A)(z) := j(A, z) f (Az) = j(A, z) f .
cz + d
Notice that Definition 7.1 (1) may be rephrased as
(M |k A) (z) = M (z).
Remark 16. We shall also consider level N weight k ∈ 12 Z forms with
Nebentypus χ. To define such forms, suppose that N is a positive integer
(with 4 | N if k ∈ 12 Z \ Z), and let χ be a Dirichlet character
a b modulo N . To
define these forms, one merely requires, for every c d ∈ Γ0 (N ), that
14

  
az + b χ(d)(cz + d)k M (z) if k ∈ Z,
M =  c 2k −2k
cz + d d
k
d χ(d)(cz + d) M (z) if k ∈ 12 Z \ Z.
Throughout, we shall adopt the following notation. If Γ ⊂ SL2 (Z) is a
congruence subgroup, then we let
Sk (Γ) := weight k cusp forms on Γ,
Mk (Γ) := weight k holomorphic modular forms on Γ,
Mk! (Γ) := weight k weakly holomorphic modular forms on Γ,
Hk (Γ) := weight k harmonic Maass forms on Γ.
14This replaces (1) in Definition 7.1.
378 K. ONO

Furthermore, if χ is a Dirichlet character modulo N , then we let

Sk (N, χ) := level N weight k cusp forms with Nebentypus χ,
Mk (N, χ) := level N weight k holomorphic modular forms with
Nebentypus χ,
Mk! (N, χ) := level N weight k weakly holomorphic modular forms
with Nebentypus χ,
Hk (N, χ) := level N weight k harmonic Maass forms with Nebentypus χ.
When the Nebentypus character is trivial, we shall suppress χ from the
notation.
The real analytic forms in Theorem 6.1 provide non-trivial examples
of weight 1/2 harmonic Maass forms. More generally, the work of Zwegers
[218, 219], shows how to complete all of Ramanujan’s mock theta functions
to obtain weight 1/2 harmonic Maass forms. In Section 8, we shall present
further examples of harmonic Maass forms.
7.2. Fourier expansions. In this paper we consider harmonic Maass
forms with weight 2 − k ∈ 12 Z with k > 1. Therefore, throughout we assume
that 1 < k ∈ 12 Z.
Harmonic Maass forms have distinguished Fourier expansions which are
described in terms of the incomplete Gamma-function Γ(α; x)
 ∞
(7.5) Γ(α; x) := e−t tα−1 dt,
x
and the usual parameter q := e2πiz .The following characterization is
straightforward (for example, see Section 3 of [63]).
Lemma 7.2. Assume the notation and hypotheses above, and suppose
that N is a positive integer. If f (z) ∈ H2−k (Γ1 (N )), then its Fourier expan-
sion is of the form
 
(7.6) f (z) = c+ n
f (n)q + c−
f (n)Γ(k − 1, 4π|n|y)q ,
n

n
−∞ n<0
where z = x + iy ∈ H, with x, y ∈ R.
As Lemma 7.2 reveals, f (z) naturally decomposes into two summands.
In view of this fact, we make the following definition.
Definition 7.3. Assuming the notation and hypotheses in Lemma 7.2,
we refer to 
f + (z) := c+
f (n)q
n

n
−∞
as the holomorphic part of f (z), and we refer to

f − (z) := c−
f (n)Γ(k − 1, 4π|n|y)q
n

n<0
as the non-holomorphic part of f (z).
 HARMONIC MAASS FORMS AND NUMBER THEORY 379

Remark 17. A harmonic Maass form with trivial non-holomorphic part

is a weakly holomorphic modular form. We shall make use of this fact as
follows. If f1 , f2 ∈ H2−k (Γ) are two harmonic Maass forms with equal non-
holomorphic parts, then f1 − f2 ∈ M2−k ! (Γ).
7.3. The ξ-operator and period integrals of cusp forms. Har-
monic Maass forms are related to classical modular forms thanks to the
properties of differential operators. The first nontrivial relationship depends
on the differential operator
∂
(7.7) ξw := 2iy w ·.
∂z
The following lemma,15 which is a straightforward refinement of a propo-
sition of Bruinier and Funke (see Proposition 3.2 of [63]), shall play a central
role throughout this paper.
Lemma 7.4. If f ∈ H2−k (N, χ), then
ξ2−k : H2−k (N, χ) −→ Sk (N, χ)
is a surjective map. Moreover, assuming the notation in Definition 7.3, we
have that
∞

ξ2−k (f ) = −(4π)k−1
c−
f (−n)n
k−1 n
q .
n=1
Thanks to Lemma 7.4, we are in a position to relate the non-holomorphic
parts of harmonic Maass forms, the expansions

f − (z) := c−
f (n)Γ(k − 1, 4π|n|y)q ,
n

n<0
with “period integrals” of modular forms. This observation was critical in
Zwegers’s work on Ramanujan’s mock theta functions.
To make this connection, we must relate the Fourier expansion of the
cusp form ξ2−k (f ) with f − (z). This connection is made by applying the
simple integral identity
 i∞
e2πinτ
(7.8) 2−k
dτ = i(2πn)1−k · Γ(k − 1, 4πny)q −n .
−z (−i(τ + z))
This identity follows by the direct calculation
 i∞  i∞ 2πin(τ −z)
e2πinτ e
2−k
dτ = 2−k
dτ = i(2πn)1−k · Γ(k−1, 4πny) q −n .
−z (−i(τ + z)) 2iy (−iτ )
In this way, we may think of the non-holomorphic parts of weight 2 − k
harmonic Maass forms as period integrals of weight k cusp forms, where one
applies (7.8) to
 i∞ ∞ 2πinτ
n=1 a(n)e
dτ,
−z (−i(τ + z))2−k
15The formula for ξ
2−k (f ) corrects a typographical error in [63].
380 K. ONO

where ∞ n −
n=1 a(n)q is a weight k cusp form. In short, f (z) is the period
integral of the cusp form ξ2−k (f ).
In addition to this important observation, we require the following fact
concerning the nontriviality of certain principal parts of harmonic Maass
forms.
Lemma 7.5. If f ∈ H2−k (Γ) has the property that ξ2−k (f ) = 0, then the
principal part of f is nonconstant for at least one cusp.
Sketch of the Proof. This lemma follows from the work of Bruinier
and Funke [63]. Using their pairing {•, •}, one finds that {ξ2−k f, f } = 0
thanks to its interpretation in terms of Petersson norms. On the other hand,
Proposition 3.5 of [63] expresses this quantity in terms of the principal part
of f and the coefficients of the cusp form ξ2−k (f ). An inspection of this
formula reveals that at least one principal part of f must be nonconstant. 

7.4. The D-operator. In addition to the differential operator ξ2−k ,

which defines the surjective map
ξ2−k : H2−k (N, χ) −→ Sk (N, χ),
we consider the differential operator
1 d
(7.9) D := · .
2πi dz
We have the following theorem for integer weights.
Theorem 7.6. Suppose that 2 ≤ k ∈ Z and f ∈ H2−k (N ), then
Dk−1 (f ) ∈ Mk! (N ).
Moreover, assuming the notation in (7.6), we have

Dk−1 f = Dk−1 f + = c+
f (n)n
k−1 n
q .
n
−∞

To prove this theorem, we must first recall some further differential op-
erators, the Maass raising and lowering operators (for example, see [63, 71])
Rk and Lk . They are defined by
 
∂ −1 ∂ ∂
Rk = 2i + ky = i −i + ky −1 ,
∂z ∂x ∂y
 
2 ∂ ∂ ∂
Lk = −2iy = −iy 2
+i .
∂ z̄ ∂x ∂y
With respect to the Petersson slash operator (7.4), these operators satisfy
the intertwining properties
Rk (f |k γ) = (Rk f ) |k+2 γ,
Lk (f |k γ) = (Lk f ) |k−2 γ,
 HARMONIC MAASS FORMS AND NUMBER THEORY 381

for any γ ∈ SL2 (R). The Laplacian Δk can be expressed in terms of Rk and
Lk by
(7.10) −Δk = Lk+2 Rk + k = Rk−2 Lk .
If f is an eigenfunction of Δk satisfying Δk f = λf , then
(7.11) Δk+2 Rk f = (λ + k)Rk f,
(7.12) Δk−2 Lk f = (λ − k + 2)Lk f.
For any positive integer n we put
Rkn := Rk+2(n−1) ◦ · · · ◦ Rk+2 ◦ Rk .
1 d d
We also let Rk0 be the identity. The differential operator D := 2πi dz = q dq
satisfies the following relation
Rk = −4πD + k/y.
The next well known lemma (for example, see (4.15) of [148]) is often re-
ferred to as Bol’s identity.
Lemma 7.7. Assuming the notation and hypotheses above, we have
1
Dk−1 = Rk−1 .
(−4π)k−1 2−k
By Lemma 7.7, we see that Dk−1 defines a linear map
Dk−1 : M2−k
!
(N ) −→ Mk! (N ).
Theorem 7.6 asserts that this map extends to harmonic Maass forms. More-
over, the theorem provides a simple description of the images.
Proof of Theorem 7.6. It is clear that Dk−1 f has the transforma-
tion behavior of a modular form of weight k.
We now show that Lk Dk−1 f = 0. This implies that Dk−1 f is holomor-
phic on H. By Lemma 7.7, it suffices to show that Lk R2−kk−1
f = 0. Since
Δ2−k f = 0, it follows from (7.11) by induction that
k−2
Δk−2 R2−k f = (2 − k)R2−k
k−2
f.
Using (7.10), we obtain
k−1
Lk R2−k k−2
f = (Lk Rk−2 )R2−k f = (−Δk−2 − (k − 2))R2−k
k−2
f = 0.
Finally, the growth behavior of f at the cusps implies that Dk−1 f is
meromorphic at the cusps. Therefore, Dk−1 indeed extends to H2−k (N ).
To complete the proof, we compute the Fourier expansion of Dk−1 f .
Assuming the notation in (7.6), a straightforward calculation gives

f (n)Γ(k − 1, 4πny)(−y)
k−2
R2−k f (z) = c+ 2−k 2πinz̄
e
n
−∞

+ (k − 2)!2 c−
f (n)(−y)
2−k 2πinz̄
e .
n<0
382 K. ONO

k−1
Moreover, R2−k f has the Fourier expansion

k−1
R2−k f (z) = c+
f (n)(−4πn)
k−1 n
q .
n
−∞

In particular, we have

Dk−1 f = Dk−1 f + = c+
f (n)n
k−1 n
q .
n
−∞

The first two formulas follow from the Fourier expansion of f and the differ-
k−2
ential equations Δk−2 R2−k f = (2 − k)R2−k
k−2 k−1
f and Δk R2−k f = 0. The third
formula is a consequence of the second and Lemma 7.7. 
k−2
Remark 18. Note that g := y k−2 R2−k f is a harmonic Maass form
of weight 2 − k in the (slightly more general) sense of Section 3 of [63].
Moreover, ξ2−k g = y −k L2−k g = R2−k
k−1
f . This can also be used to compute
the Fourier expansions in the proof of Theorem 7.6.
Theorem 7.6 implies that the coefficients c+ f (n), for non-zero n, are
obtained by dividing the nth coefficient of some fixed weakly holomorphic
modular form by nk−1 . Therefore we are compelled to determine the image
of the map
Dk−1 : H2−k (N ) −→ Mk! (N ).
It is not difficult to see that this map is not generally surjective. Our next
result determines the image of Dk−1 in terms of regularized inner products.
The following result is obtained by Bruinier, Rhoades and the author in [68].
Theorem 7.8. If 2 ≤ k ∈ Z, then the image of the map
Dk−1 : H2−k (N ) −→ Mk! (N )
consists of those h ∈ Mk! (N ) which are orthogonal to cusp forms with respect
to the regularized inner product, which also have constant term 0 at all cusps
of Γ0 (N ).
To make Theorem 7.8 precise, we must define what it means for a weakly
holomorphic modular form to be orthogonal to cusp forms. To this end, we
first recall the regularized inner product.
We stress again that k ≥ 2 is an integer. Obviously, Γ0 (N ) has finite
index in Γ(1) = SL2 (Z). We define a regularized inner product of g ∈ Mk (N )
and h ∈ Mk! (N ) as follows. For T > 0 we denote by FT (Γ(1)) the truncated
fundamental domain
FT (Γ(1)) = {z ∈ H : |x| ≤ 1/2, |z| ≥ 1, and y ≤ T }
for Γ(1). Moreover, we define the truncated fundamental domain for
Γ0 (N ) by

FT (N ) = γFT (Γ(1)).
γ∈Γ0 (N )\Γ(1)
 HARMONIC MAASS FORMS AND NUMBER THEORY 383

Following Borcherds [44], we define the regularized inner product (g, h)reg
as the constant term in the Laurent expansion at s = 0 of the meromorphic
continuation in s of the function

1 dx dy
lim g(z)h(z)y k−s .
[Γ(1) : Γ0 (N )] T →∞ FT (N ) y2

Arguing as in Section 6 of [44], it can be shown that (g, h)reg exists for
any g ∈ Mk (N ) and h ∈ Mk! (N ). (It also exists for g ∈ Mk (N ) and h ∈
Hk (N ). But note that it does not exist in general if g and h are both
weakly holomorphic with poles at the cusps.) For cusp forms g and h, the
regularized inner product reduces to the classical Petersson inner product.

Remark 19. If h ∈ Mk! (N ) has vanishing constant term at every cusp

of Γ0 (N ), then

reg 1 dx dy
(g, h) = lim g(z)h(z)y k .
[Γ(1) : Γ0 (N )] T →∞ FT (N ) y2

The following theorem provides a formula for the regularized inner prod-
uct in terms of Fourier coefficients.

Theorem 7.9. If g ∈ Mk (N ) and f ∈ H2−k (N ), then

(−1)k 
k−1 reg
(g, R2−k f) = wκ · cg (0, κ)c+
f (0, κ),
[Γ(1) : Γ0 (N )]
κ∈Γ0 (N )\P 1 (Q)

where cg (0, κ) (resp. c+f (0, κ)) denotes the constant term of the Fourier
expansion of g (resp. f ) at the cusp κ ∈ P 1 (Q), and wκ is the width of the
cusp κ.

Proof. For simplicity, we carry out the proof only in the special case
k−2
Γ0 (1) = SL2 (Z). We put H := y k−2 R2−k k−1
f , and let h := R2−k f = y −k L2−k H.
Since the constant terms at all cusps of h vanish, we have

k−1 reg dx dy
(g, R2−k f) = lim g(z)h(z)y k
T →∞ FT (1) y2

dx dy
= lim g(z)(L2−k H) 2
T →∞ FT (1) y

∂
= lim g(z)( H) dz dz̄
T →∞ FT (1) ∂ z̄

= − lim ¯ ∧ g(z) dz.
(∂H)
T →∞ FT (1)
384 K. ONO

Using the holomorphy of g, we obtain, by Stokes’ theorem, the expression


(g, R2−k f ) = − lim
k−1 reg
d(H(z)g(z) dz)
T →∞ FT (1)

= − lim H(z)g(z) dz
T →∞ ∂FT (1)
 1/2
= lim H(x + iT )g(x + iT ) dx.
T →∞ x=−1/2

The integral over x gives the constant term in the expansion of H(x +
iT )g(x + iT ), and it can be computed using the Fourier expansion

−2πinz
H(z) = (−1)k f (n)Γ(k − 1, 4πny)e
c+
n
−∞

+ (−1)k (k − 2)!2 c−
f (n)e
−2πinz

n<0

of H (see the proof of Theorem 7.6) and the expansion of g. To com-

plete the proof, notice that in the limit T → ∞ the only contribution is
(−1)k cg (0)c+
f (0). 

Proof of Theorem 7.8. By Theorem 7.9, it follows that if g ∈ Sk (N ),

then
k−1 reg
(g, R2−k f ) = 0.
Therefore, if f ∈ H2−k (N ), it follows from Lemma 7.7 and Theorem 7.9 that
Dk−1 f is orthogonal to cusp forms.
Conversely, assume that h ∈ Mk! (N ) is orthogonal to cusp forms and has
vanishing constant term at every cusp of Γ0 (N ). By Lemma 3.11 of [63],
we may chose f ∈ H2−k (N ) such that the principal parts of Dk−1 f and h
at the cusps agree up to the constant terms. Since the constant terms of h
and Dk−1 f vanish, they trivially agree as well. Consequently,
h − Dk−1 f ∈ Sk (N ).
In view of Theorem 7.9 and the hypothesis on h, we find that h − Dk−1 f is
orthogonal to cusp forms, and so it is zero. 

7.5. Hecke operators. It is natural to investigate the action of the

Hecke operators on harmonic Maass forms. Here we illustrate how to com-
bine facts about Hecke operators with Lemma 7.4 to obtain weakly holo-
morphic modular forms from harmonic Maass forms.
First we recall the classical Hecke operators. Suppose that p is prime,
and that k ∈ 12 Z. Suppose that

F (z) = aF (n)q n
n
−∞
 HARMONIC MAASS FORMS AND NUMBER THEORY 385

is a weight k weakly holomorphic modular form on Γ0 (N ) with Nebentypus

χ. If k ∈ Z, then the Hecke operator Tk (p) is defined by
  
(7.13) F | Tk (p) := aF (pn) + χ(p)pk−1 aF (n/p) q n .
n
−∞

If k = λ + 1
2 ∈ 2Z
\ Z, then we have that
1

  
(−1)λ n

F | Tk (p) := 2
aF (p n) + χ(p) pλ−1 aF (n)
n
−∞
p

(7.14) +χ(p2 )p2λ−1 aF (n/p2 ) q n .

Remark 20. For the sake of uniformity, we abuse notation in the case of
the half-integral weight Hecke operators. Our operator Tk (p) is traditionally
referred to as T (p2 ) or Tk (p2 ).
These operators map modular forms to modular forms, and they de-
fine linear transformations on Mk (N, χ) and Sk (N, χ). Using Lemma 7.2,
it is not difficult to see that the definitions of these Hecke operators
extend to harmonic Maass forms and their spaces in the obvious way. Using
Lemmas 7.2 and 7.4, it is then not difficult to obtain the following useful
theorem. As before, here we suppose that 2 − k ∈ 12 Z with k > 1.
Theorem 7.10. Suppose that f (z) ∈ H2−k (N, χ), and that p  N is a
prime for which ξ2−k (f ) ∈ Sk (N, χ) is an eigenform of Tk (p) with eigen-
value λ(p).
(1) If k ∈ 12 Z \ Z, then
f | T2−k (p) − p2−2k λ(p)f
!
is a weakly holomorphic modular form in M2−k (N, χ).
(2) If k ∈ Z, then
f | T2−k (p) − p1−k λ(p)f
!
is a weakly holomorphic modular form in M2−k (N, χ).

8. Examples of harmonic Maass forms

In this section we present further examples of harmonic Maass forms.
The interested reader should consult Zwegers’s works [218, 219] to see those
weight 1/2 harmonic Maass forms whose holomorphic parts are given in
terms of Lambert-type series and generating functions for indefinite binary
quadratic forms. Recent work of Bringmann, Folsom, and the author [49]
concerns further properties of these forms, and it also gives related weight
3/2 harmonic Maass forms. All of these harmonic Maass forms f have the
property that ξw (f ) is a linear combination of single variable theta functions.
Here we consider further examples of harmonic Maass forms. We recall
the classical Eisenstein series E2 (z), Zagier’s weight 3/2 Eisenstein series,
386 K. ONO

certain Maass-Poincaré series, and a combinatorial power series which plays

a central role in the Dyson-Ramanujan theory of partition congruences.

8.1. The Eisenstein series E2 (z). The differential operator D :=

1 d
2πi dz = q · dq
d
plays an important role in the theory of modular forms and
q-series. Indeed, as an operator on formal power series we have that
∞  ∞
 
n
D a(n)q := na(n)q n .
n=h n=h

Ramanujan proved [172] that

D(E4 ) = (E4 E2 − E6 )/3 and D(E6 ) = (E6 E2 − E8 )/2,
where Ek , for every even integer k ≥ 2, is the standard Eisenstein series
∞
2k 
Ek (z) := 1 − σk−1 (n)q n .
Bk
n=1

Here Bk denotes the usual kth Bernoulli number and σk−1 (n) := d|n dk−1 .
As is well known, for even k ≥ 4 we have that Ek (z) is a weight k
modular form on SL2 (Z). However, the Eisenstein series E2 (z)
∞

E2 (z) = 1 − 24 σ1 (n)q n
n=1

is not modular, and so the derivative of a modular form is not (in general)
a modular form.
It is not difficult to pin down the obstruction which prevents E2 (z) from
being a modular form. Indeed, it is simple to show (for example, see page
113 of [134]), for z ∈ H, that
6
(8.1) z −2 E2 (−1/z) = E2 (z) + .
πiz
Using this fact, it is not difficult to modify E2 (z) to obtain a weight 2
harmonic Maass form. Using (8.1), and the fact that
   
1 1 0 1
SL2 (Z) = , ,
0 1 −1 0
one finds that
3
(8.2) E2∗ (z) := − + E2 (z),
πy
where y = Im(z), is a weight 2 harmonic Maass form on SL2 (Z) in the more
general sense of Bruinier and Funke. It fails to satisfy Definition 7.1 (3).
Obviously, its holomorphic part is E2 (z), and its non-holomorphic part is
the simple function − πy
3
.
 HARMONIC MAASS FORMS AND NUMBER THEORY 387

Remark 21. Although E2∗ (z) is not a harmonic Maass form in the strict
sense of Definition 7.1, it still behaves well under the differential operator
ξ2 . One easily checks that ξ2 (E2∗ (z)) is constant, and so it is a modular form
of weight 0. This is in line with Lemma 7.4.
8.2. Zagier’s Eisenstein series. In their work on the intersection the-
ory of Hilbert modular surfaces [123, 210], Hirzebruch and Zagier required
a non-holomorphic weight 3/2 Eisenstein series G(z) on Γ0 (4). The non-
holomorphic part of this Eisenstein series is essentially the “period integral”
of the classical Jacobi theta function
Θ0 (z) := 1 + 2q + 2q 4 + 2q 9 + 2q 16 + · · · .
This is analogous with the work of Zwegers on Ramanujan’s mock theta
functions (see Theorem 6.1 and the discussion after Lemma 7.4) in that the
non-holomorphic parts of his harmonic Maass forms are also period integrals
of single variable theta functions. Unlike Zwegers’s harmonic Maass forms,
where the holomorphic parts turned out to be combinatorial q-series such as
the mock theta functions of Ramanujan, the holomorphic part of the Zagier
form is the generating function for Hurwitz class numbers. More precisely,
we have (in Zagier’s notation) the following theorem.
Theorem 8.1. If H(0) = ζ(−1) = − 12 1
, and if H(n) denotes the usual
Hurwitz class numbers for discriminants −n, then the function
∞
 ∞
1
β(4πn2 y)q −n
2
G(z) = H(n)q n + √
16π y n=−∞
n=0
is a weight 3/2 harmonic Maass form
 ∞ on Γ0 (4) in the sense of Bruinier and
Funke. Here we have that β(s) := 1 t −3/2 e−st dt.
Remark 22. Zagier’s β-function is essentially the incomplete Gamma-
function.
As in the case of E2∗ (z), we have that G(z) is a weight 3/2 harmonic
Maass form on Γ0 (4) in the more general sense of Bruinier and Funke. Again,
the culprit is Definition 7.1 (3). Nevertheless, one sees that ξ 3 (G) is a
2
multiple of Θ0 (z), which is in line with the phenomenon in Lemma 7.4. The
only difference is that Θ0 (z) is not a cusp form.
Remark 23. In recent work, Bringmann and Lovejoy [51] have found a
striking identity relating a certain rank generating function for overpartitions
to G(z).
8.3. Poincaré series. Here we describe certain Maass-Poincaré series.
Throughout, we rely on classical special functions whose properties and
definitions may be found in [1, 20]. Such Poincaré series, which were first
given by D. Hejhal, are discussed in [56, 62, 65, 117].
Suppose that k ∈ 12 Z, and that N is a positive integer (with 4 | N if
k ∈ 12 Z \ Z). Let m be an integer, and let ϕm : R+ → C be a function which
388 K. ONO

satisfies ϕm (y) = O(y α ), as y → 0, for some α ∈ R. If e(α) := e2πiα as

usual, then let
(8.3) ϕ∗m (z) := ϕm (y)e(mx).
This function is fixed by the group of translations
   
1 n
Γ∞ := ± : n∈Z .
0 1
Using the slash operator (7.4), we define the Poincaré series P(m, k, N,
ϕm ; z) by

(8.4) P(m, k, N, ϕm ; z) := (ϕ∗m |k A)(z).
A∈Γ∞ \Γ0 (N )

The Fourier expansions of these series are given in terms of the Kloost-
erman sums
  mv+nv 
v(c)× e if k ∈ Z,
(8.5) Kk (m, n, c) :=   c 2kc 2k  mv+nv 
v(c)× v εv e c if k ∈ 12 Z \ Z.
In the sums above, v runs through the primitive residue classes modulo c,
and v denotes the multiplicative inverse of v modulo c. The following lemma
gives the fundamental properties of such Poincaré series (for example, see
Proposition 3.1 of [65] where N = 4).
Lemma 8.2. If k > 2 − 2α, then the following are true.
(1) Each Poincaré series P(m, k, N, ϕm ; z) is a weight k Γ0 (N )-
invariant function.
(2) Near the cusp at ∞, the function P(m, k, N, ϕm ; z) − ϕ∗m (z) has
moderate growth. Near the other cusps, P(m, k, N, ϕm ; z) has mod-
erate growth.
(3) If P(m, k, N, ϕm ; z) is twice continuously differentiable, then it has
the locally uniformly absolutely convergent Fourier expansion

P(m, k, N, ϕm ; z) = ϕ∗m (z) + a(n, y)e(nx),
n∈Z
where
∞

a(n, y) := c−k Kk (m, n, c)
c>0
c≡0 (mod N )
 ∞    
−k y mx
z ϕm e − 2 2 − nx dx.
−∞ c |z|2
2 c |z|
We now recall the classical family of Poincaré series (for example, see
[125, 167]) which arises from Lemma 8.2. If 2 ≤ k ∈ 12 Z, N is positive, and
if m is an integer, then let
∞

(8.6) P (m, k, N ; z) = q m + a(m, k, N ; n)q n =: P(m, k, N, e(imy); z).
n=1
 HARMONIC MAASS FORMS AND NUMBER THEORY 389

These series are modular, and their Fourier expansions are given in terms
of the I-Bessel and J-Bessel functions, and the Kloosterman sums above.
Theorem 8.3. If 2 ≤ k ∈ 12 Z, and m, N ≥ 1 (with 4 | N if k ∈ 12 Z \ Z),
then the following are true.
1) We have that P (m, k, N ; z) ∈ Sk (N ), and for positive integers n
we have
k
 n  k−1  Kk (m, n, c)
2
a(m, k, N ; n) = 2π(−1) 2 ·
m c
c>0
c≡0 (mod N )
 √ 
4π mn
· Jk−1 .
c
2) We have that P (−m, k, N ; z) ∈ Mk! (N ), and for positive integers n
we have
k
 n  k−1  Kk (−m, n, c)
2
a(−m, k, N ; n) = 2π(−1) 2 ·
m c
c>0
c≡0 (mod N )
  
4π |mn|
· Ik−1 .
c

Now we recall one family of Maass-Poincaré series which have appeared

in earlier works [62, 65, 117, 160]. To define them, again suppose that
2 < k ∈ 12 Z, and that N is a positive integer (with 4 | N if k ∈ 12 Z \ Z). To
employ Lemma 8.2, we first select an appropriate function ϕ.
Let Mν, μ (z) be the usual M -Whittaker function which is a solution to
the differential equation
 
4 −μ
1 2
∂2u 1 ν
+ − + + u = 0.
∂z 2 4 z z2
For complex s, let
Ms (y) := |y|− 2 M k sgn(y), s− 1 (|y|),
k

2 2

and for positive m let ϕ−m (z) := M1− k (−4πmy). We now let
2

(8.7) Q(−m, k, N ; z) := P(−m, 2 − k, N, ϕ−m ; z).

Lemma 8.2 leads to the following theorem (see [54, 56, 62, 117]).
Theorem 8.4. If 2 < k ∈ 12 Z and m, N ≥ 1 (with 4 | N if k ∈ 12 Z \ Z),
then Q(−m, k, N ; z) is in H2−k (N ). Moreover, if Ik−1 is the usual I-Bessel
function, then

Q(−m, k, N ; z) = (1−k) (Γ(k − 1, 4πmy) − Γ(k − 1)) q −m + cm (n, y) q n .
n∈Z
390 K. ONO

1) If n < 0, then
! n ! 1−k
! ! 2
cm (n, y) = 2πik (1 − k) Γ(k − 1, 4π|n|y) ! !
m   
 K2−k (−m, n, c) 4π |mn|
× · Jk−1 .
c c
c>0
c≡0 (mod N )

2) If n > 0, then
 n  1−k
2
cm (n, y) = −2πi Γ(k)
k
m   
 K2−k (−m, n, c) 4π |mn|
× · Ik−1 .
c c
c>0
c≡0 (mod N )

3) If n = 0, then
 K2−k (−m, 0, c)
cm (0, y) = −2k π k ik mk−1 .
ck
c>0
c≡0 (mod N )

These three families of Poincaré series are closely related. The follow-
ing theorem gives the explicit relationships between the cusp forms P (m, k,
N ; z), the weakly holomorphic modular forms P (−m, k, N ; z), and the har-
monic Maass forms Q(−m, k, N ; z). They are related by the ξ2−k and Dk−1
differential operators.
Theorem 8.5. Suppose that 2 < k ∈ 12 Z, and that m, N ≥ 1 (with 4 | N
if k ∈ 12 Z \ Z). Then the following are true.
(1) Under the operator ξ2−k , we have that
ξ2−k (Q(−m, k, N ; z)) = (4π)k−1 mk−1 (k − 1) · P (m, k, N ; z) ∈ Sk (N ).
(2) Under the operator Dk−1 , if k is an integer, then we have that
Dk−1 Q(−m, k, N ; z) = −mk−1 Γ(k) · P (−m, k, N ; z) ∈ Mk! (N ).
Proof. The claims follow easily from the explicit expansions in Theo-
rems 8.3 and 8.4. 
8.4. A family of weight 1/2 harmonic Maass forms. In his work
[218, 219] (also see Section 6.2), Zwegers describes the holomorphic parts of
certain harmonic Maass forms in an explicit way. These series are presented
as reciprocals of theta functions multiplied with Lambert-type series, and
are also given as generating functions for indefinite theta series.
Here we present an infinite family of Maass forms which arise from the
Eulerian series
∞
 2
qn
(8.8) R(w; q) := 1 + ,
(wq; q)n (w−1 q; q)n
n=1
 HARMONIC MAASS FORMS AND NUMBER THEORY 391

where, as before, we let

(a; q)n := (1 − a)(1 − aq) · · · (1 − aq n−1 ).
To place this formal power series in context, we recall that we have
already seen two important specializations in w. Thanks to Theorem 2.2,
by setting w = 1 we have that
∞

R(1; q) = p(n)q n .
n=0
By letting w = −1, we obtain the series
∞
 2
qn
R(−1; q) = 1 + .
(1 + q)2 (1 + q 2 )2 · · · (1 + q n )2
n=1
This series is Ramanujan’s third order mock theta function f (q), which by
Theorem 6.1 is essentially the holomorphic part of a weight 1/2 harmonic
Maass form. In [55], Bringmann and the author generalized this fact.
To make this precise, suppose that 0 < a < c are integers,   and let
2c
ζc := e2πi/c . If fc := gcd(c,6) , then define the theta function Θ ac ; τ by
(8.9)
a    
 aπ(6m + 1) τ 
Θ ; τ := m
(−1) sin · θ 6m + 1, 6fc ; ,
c c 24
m (mod fc )

where
 2
(8.10) θ(α, β; τ ) := ne2πiτ n .
n≡α (mod β)

Throughout,
a  let c := lcm(2c2 , 24), and let "c := c /24. It is well known that
Θ c ; c τ is a cusp form of weight 3/2 (for example, see [192]).  
Using this cuspidal theta function, we define the function S1 ac ; z by
the period integral
a    1  i∞  
−i sin πa c 2 Θ ac ; c τ
(8.11) S1 ; z := √ c  dτ.
c 3 −z̄ −i(τ + z)
 
Using this notation, define D ac ; z by
a  a  c
(8.12) D ; z := −S1 ; z + q − 24 R(ζca ; q c ).
c c
 
Theorem 8.6. If 0 < a < c, where c is odd, then D ac ; z is a Maass
form of weight 1/2 on Γ1 (144fc2 "c ).
Remark 24. For even c, these Maass forms also satisfy transformation
laws with respect to Γ1 congruence subgroups. Our proof of Theorem 8.6
in [55] is long and complicated. It is based on a modification of earlier
work of Watson [206, 207]. Zagier [213] found a shorter proof of this result
392 K. ONO

-function. His proof of

using a clever application of Zwegers’s work on the μ
Theorem 8.6 relies on the fact that
q −1/24 R(e(α); q) η(3z)3 /η(z)
− q − 6 e(−α)μ(3α, −z; 3z)
1

−πiα
=
e −e πiα ϑ(3α; 3z)
+ q − 6 e(α)μ(3α, z; 3z).
1
(8.13)

9. Dyson-Ramanujan theory of partition congruences

As mentioned in the introduction, the partition function has been a
“testing ground” for the theory of modular forms. In number theory we are
interested in the divisibility properties of p(n) and the size of p(n). Here
we consider the congruence properties of the partition function p(n). It
will turn out that these harmonic Maass forms have arithmetic properties
which may be studied using the theory of -adic Galois representations as
developed by Deligne, Serre, and Swinnerton-Dyer [83, 187, 197]. These
new congruences imply new partition congruences.

9.1. Ramanujan’s congruences. Ramanujan proved the striking

congruences
p(5n + 4) ≡ 0 (mod 5),
p(7n + 5) ≡ 0 (mod 7),
p(11n + 6) ≡ 0 (mod 11),

and conjectured generalizations where the moduli are powers of 5, 7 and

11. These conjectures have been resolved in works by Atkin and Watson
[28, 208].
These congruences are just the tip of the iceberg. In the late 1960s, Atkin
[29] proved further congruences for small prime moduli. More recently, the
author proved [162] that there are infinitely many such congruences for
any prime modulus Q ≥ 5. This result was subsequently generalized by S.
Ahlgren and the author [2, 3], and these results imply the following theorem.

Theorem 9.1. If M is coprime to 6, then there are infinitely many

non-nested arithmetic progressions An + B for which

p(An + B) ≡ 0 (mod M ).

Remark 25. By non-nested, we mean that there are infinitely many

arithmetic progressions An + B, with 0 ≤ B < A, with the property that
there are no progressions which contain another progression.

Remark 26. It is believed that there are no such congruences with

moduli which are not coprime to 6 (for example, see [5, 163]).
 HARMONIC MAASS FORMS AND NUMBER THEORY 393

Example 9.2. Here are the simplest examples of these congruences for
the prime moduli 17 ≤ M ≤ 31:

p(48037937n + 1122838) ≡ 0 (mod 17),

p(1977147619n + 815655) ≡ 0 (mod 19),
p(14375n + 3474) ≡ 0 (mod 23),
p(348104768909n + 43819835) ≡ 0 (mod 29),
p(4063467631n + 30064597) ≡ 0 (mod 31).

9.2. Dyson’s rank. In an effort to provide a combinatorial explanation

of Ramanujan’s congruences modulo 5 and 7, Dyson introduced [90] the so-
called “rank” of a partition. The rank of a partition is defined to be its
largest part minus the number of its parts. Let N (m, n) denote the number
of partitions of n with rank m, and for integers 0 ≤ r < t let N (r, t; n) denote
the number of partitions of n with rank congruent to r modulo t. Dyson
noticed that ranks modulo 5 (resp. 7) apppeared to divide the partitions of
5n + 4 (resp. 7n + 5) into 5 (resp. 7) groups of equal size, thereby providing
an elegant explanation for these two congruences of Ramanujan.
Atkin and Swinnerton-Dyer [31] confirmed Dyson’s conjecture by prov-
ing the following result.

Theorem 9.3. The following are true:

(1) If 0 ≤ r < 5, then
∞ 
 
p(5n + 4)
N (r, 5; 5n + 4) − q n = 0.
5
n=0

(2) If 0 ≤ r < 7, then

∞ 
 
p(7n + 5)
N (r, 7; 7n + 5) − q n = 0.
7
n=0

Remark 27. Dyson [90] also noted that the rank fails to explain the Ra-
manujan congruence with modulus 11. A short calculation reveals that the
first failure already occurs with p(6) = 11. In view of this difficulty, Dyson
further postulated the existence of another statistic, the so-called “crank”,
that could be used to provide an explanation for all three Ramanujan con-
gruences. In 1988, Andrews and Garvan [25] found the crank, and they
confirmed Dyson’s speculation that it explains the three Ramanujan con-
gruences. Recent work of Mahlburg [151] establishes that the Andrews-
Dyson-Garvan crank plays an even more central role in the theory partition
congruences. His work establishes congruences modulo arbitrary powers of
all primes ≥ 5. Other work by Garvan, Kim and Stanton [99] gives a differ-
ent “crank” for several other Ramanujan congruences.
394 K. ONO

9.3. Maass form generating functions for Dyson’s ranks. Al-

though the rank does not explain Ramanujan’s partition congruence with
modulus 11, it turns out that it indeed plays a central role thanks to the
arithmetic properties of certain harmonic Maass forms. Indeed, it turns out
that the generating functions for N (r, t; n) are related to harmonic Maass
forms. To make this connection, one simply generalizes the proof of Theo-
rem 2.2 to obtain the following proposition concerning the power series
∞ 2
qn
R(w; q) := 1 +
(wq; q)n (w−1 q; q)n
n=1
considered earlier in (8.8).
Proposition 9.4. As a formal power series, we have that
∞
 ∞

R(w; q) := 1 + N (m, n)wm q n .
n=1 m=−∞

Thanks to Theorem 8.6, it is not difficult to relate the functions N (r, t; n)

to the coefficients of weight 1/2 harmonic Maass forms. In this direction,
Bringmann and the author proved the following theorem in [55], which also
sheds further light on Theorem 9.3.
Theorem 9.5. Assume the notation from Section 8.4. If 0 ≤ r < t are
integers, where t is odd, then
∞  
p(n) t n− t
N (r, t; n) − q
t
n=0
is the holomorphic
 part of a weight 1/2 harmonic Maass form on
Γ1 144ft "t . Moreover, the non-holomorphic part of this Maass form is
2

1
it2   
A(r, t, m) · γ(t, y; n)q − t n .
2
√
t 3 m (mod f
t) n≡6m+1 (mod 6ft )

Here we have that


t−1    
πj πj(6m + 1)
(9.1) A(r, t, m) := (−1) m
ζt−rj sin sin ,
t t
j=1
and we have that
 
i 1
γ(t, y; n) := ·Γ ; 4π "t n y .
2
2
2π "t
Proof. This result follows easily from a standard argument involving
the orthogonality of roots of unity. In particular, observe that
∞
 ∞
1 1  −rj
t−1
(9.2) n
N (r, t; n)q = p(n)q n + ζt · R(ζtj ; q).
t t
n=0 n=0 j=1
 HARMONIC MAASS FORMS AND NUMBER THEORY 395

By Theorem 8.6, it then follows that

∞  
p(n) t n− t
N (r, t; n) − q 24
t
n=0
is the holomorphic part of a Maass form of weight 1/2, one which is given
as an appropriate weighted sum of the Maass forms D at ; z .
The result then follows from the explicit description of the non-holomo-
rphic parts of these forms. For integers 0 < a < t, we have
(9.3)
a  ∞ ∞  πa  1
    i sin t2
− 24t t n− 24t
D ;z = q + N (m, n)ζtam q + √t
t 3
n=1 m=−∞
   
aπ(6m + 1) 
γ(t, y; n)q − t n .
2
× (−1)m sin
t
m (mod ft ) n≡6m+1 (mod 6ft )

This expansion follows easily from

 a  i sin  πa   2  
1

t
 aπ(6m + 1)
−S1 ;z = √t m
(−1) sin
t 3 t
m (mod ft )

  i∞ 2
ne2πin t τ
×  dτ,
n≡6m+1 (mod 6ft ) −z
−i(τ + z)

and the identity (recall (7.8))

 i∞ 2
ne2πin t τ  2
 dτ = γ(t, y; n) · q − t n .
−z −i(τ + z)

Remark 28. It is not difficult to deduce Theorem 9.3 from Theorem 9.5
using Atkin’s U -operator. In short, one must merely prove that certain
harmonic Maass forms, which turn out to be weakly holomorphic modular
forms, are identically 0.
Using quadratic twists, one can use Theorem 9.5 to obtain weakly holo-
morphic modular forms. Theorem 1.4 of [55] asserts the following result.
Theorem 9.6. If 0 ≤ r < t are integers, where t is odd, and P  6t is
prime, then
  
p(n) t n− t
N (r, t; n) − q 24
t
n≥1  
 
24t n−t −24t
P
=− P

 
is a weight 1/2 weakly holomorphic modular form on Γ1 144ft2 "t P 4 .
396 K. ONO

Proof. The non-holomorphic parts of the Maass forms in Theorem 9.5

have the property that their coefficients are supported on a fixed square
class, numbers of the form −"t n2 . This square class is easily annihilated by
taking linear combinations of quadratic twists. In particular, suppose that
P  6t is prime. For this prime P, let

P−1
v  2πiv
g := e P
P
v=1
a 
be the usual Gauss sum with respect to P. Define the function D t;z P by
a  g   v  a 
P−1  
1 −P
v
(9.4) D ;z := P D ; z | 1 ,
t P P t 2 0 1
v=1
 
where | 1 is the usual slash operator (7.4). By construction, D at ; z P is
2 a 
the P-quadratic twist
a  of D  nt ;z . In other words, the nth coefficient
 a  in the
q-expansion of D t ; z P is P times the nth coefficient of D t ; z . For the
non-holomorphic part, this follows from the fact that the factors γ(t, y; n)
appearing in (9.3) are fixed by the transformations in (9.4).  
Generalizing classical facts about twists of modular forms, D at ; z P is
a Maass form of weight 1/2 on Γ1 (144ft2 "t P 2 ). By (9.3), it follows that
 
a  −"t a 
(9.5) D ;z − D ;z
t P t P

is a Maass form of weight 1/2 on Γ1 (144fc2 "t P 2 ) with the property that
 2 2
its non-holomorphic part is supported on summands of the form ∗q − t P n .
These terms are annihilated by taking the P-quadratic twist of this Maass
form. Consequently, we obtain a weakly holomorphic modular form of
weight 1/2 on Γ1 (144ft2 "t P 4 ). Combining these observations with (9.2),
we obtain the theorem. 
Remark 29. All of the results in this section continue to hold for even
t thanks to Theorem 6.3 and (8.13).
9.4. Implications for partition congruences. Here we apply
Theorem 9.6 to shed light on the congruence properties of Dyson’s func-
tions N (r, t; n). We shall employ a method first used by the author in [162]
in his work on p(n) to show that Dyson’s rank partition functions themselves
uniformly satisfy Ramanujan-type congruences. Bringmann and the author
have proved the following theorem (see Theorem 1.5 of [55], as well as [47]).
Theorem 9.7. Let t be a positive odd integer, and let M be a positive
integer coprime to 6. Then there are infinitely many non-nested arithmetic
progressions An + B such that for every 0 ≤ r < t we have
N (r, t; An + B) ≡ 0 (mod M ).
 HARMONIC MAASS FORMS AND NUMBER THEORY 397

Remark 30. Since we have


t−1
p(An + B) = N (r, t; An + B),
r=0

Theorem 9.7 provides a combinatorial decomposition of the partition func-

tion congruence
p(An + B) ≡ 0 (mod M ).

Sketch of the Proof of Theorem 9.7. For simplicity, we suppose

that M = Qj , where 5 ≤ Q  t is prime, and j is a positive integer. The
proof depends on Theorem 9.6, the observation that certain “sieved” parti-
tion rank generating functions are weakly holomorphic modular forms. This
result then reduces the proof of Theorem 9.7 to the fact that any finite
number of half-integral weight cusp forms with integer coefficients are anni-
hilated, modulo a fixed prime power, by a positive proportion of half-integral
weight Hecke operators.
To be precise, suppose that f1 (z), f2 (z), . . . , fs (z) are half-integral weight
cusp forms where
fi (z) ∈ Sλi + 1 (Γ1 (4Ni )) ∩ OK [[q]],
2

and where OK is the ring of integers of a fixed number field K. If Q is prime

and j ≥ 1 is an integer, then the set of primes L for which
(9.6) fi | Tλi + 1 (L) ≡ 0 (mod Qj ),
2

for each 1 ≤ i ≤ s, has positive Frobenius density. As before, here Tλi + 1 (L)
2
(see Section 7.5) denotes the usual L2 index Hecke operator of weight λi + 12 .
Suppose that P  6tQ is prime. By Theorem 9.6, for every 0 ≤ r < t
∞

F (r, t, P; z) = a(r, t, P; n)q n
n=1
  
p(n) t
(9.7) := N (r, t; n) − q t n− 24
    t
24t n−t −24t
P
=− P

is a weakly holomorphic modular form of weight 1/2 on Γ1 (144ft2 "t P 4 ).

Furthermore, by the work of Ahlgren and the author [3], it follows that
∞
  t
(9.8) P (t, P; z) = p(t, P; n)q n := p(n)q t n− 24
   
n=1 24t n−t −24t
P
=− P

is a weakly holomorphic modular form of weight −1/2 on Γ1 (576"t P 4 ). In

particular, all of these forms are modular with respect to Γ1 (576ft2 "t P 4 ).
398 K. ONO

Since Q  576ft2 "t P 4 , a result of Treneer (see Theorem 3.1 of [201]),

generalizing earlier observations of Ahlgren and the author [3, 162], implies
that there is a sufficiently large integer m for which

a(r, t, P; Qm n)q n ,
Qn

for all 0 ≤ r < t, and 

p(t, P; Qm n)q n
Qn

are all congruent modulo Qj to forms in the graded ring of half-integral

weight cusp forms with algebraic integer coefficients on Γ1 (576ft2 "t P 4 Q2 ).
The system of simultaneous congruences (9.6), in the case of these forms,
guarantees that a positive proportion of primes L have the property that
these forms modulo Qj are annihilated by the index L2 half-integral weight
Hecke operators. Theorem 9.7 now follows mutatis mutandis as in the proof
of Theorem 1 of [162]. One merely interprets this annihilation, using (7.14),
in terms of the N (r, t; n). 

Two remarks.
1) The simultaneous system (9.6) of congruences follows from a straightfor-
ward generalization of a classical observation of Serre (see Section 6 of [188]).
2) Treneer states her result for weakly holomorphic modular forms on Γ0 (4N )
with Nebentypus. We are using a straightforward extension of her result
to Γ1 (4N ).
9.5. Related recent works. Recently there have been many works
on the relationship between Dyson’s rank and the Andrews-Garvan crank
[25]. Several years ago, Atkin and Garvan [30] discovered a striking partial
differential equation which related the rank and crank generating functions.
Their work then inspired the recent investigation of Andrews on Durfee
symbols and higher partition statistics [19].
Armed with the new understanding of partition ranks in terms of har-
monic Maass forms, much light has been shed recently on these results
from the perspective of modular forms and Maass forms. In this direction
there are papers by Bringmann, Garvan, Lovejoy, Mahlburg, and Osburn
[48, 50, 52], in which theorems concerning the automorphic properties of
partition generating functions, asymptotics, and congruences are obtained
for various partition statistics such as those arising in the work of Andrews
[19] on rank moments. In a recent paper, Bringmann and Zwegers [60]
investigate further the phenomenon of the the Atkin-Garvan “crank-rank”
PDE from the perspective of non-holomorphic Jacobi forms.
In another recent development, M. Monks, an undergraduate at MIT,
has discovered a deep connection between the power series R(±i; q −1 ), some
false theta functions of Rogers, and the generating function for Dyson’s
ranks restricted to partitions into distinct parts. Her work [156] allows
 HARMONIC MAASS FORMS AND NUMBER THEORY 399

one to determine the behavior of the holomorphic part of the Maass form
D(±1/4; z) in Theorem 8.6 for complex z in the lower half of the complex
plane.

10. Eulerian series as modular forms

Eulerian series are combinatorial formal power series which are con-
structed from basic hypergeometric series. As described in Section 2.3, there
are famous examples of Eulerian series which essentially are modular forms.
For example, we have the celebrated Rogers-Ramanujan identities
∞
  ∞ 2
1 qn
= 1 + ,
(1 − q 5n+1 )(1 − q 5n+4 ) (1 − q)(1 − q 2 ) · · · (1 − q n )
n=0 n=1
∞ ∞ 2
1 q n +n
=1+ .
(1 − q 5n+2 )(1 − q 5n+3 ) (1 − q)(1 − q 2 ) · · · (1 − q n )
n=0 n=1

These identities provide series expansions of infinite products which cor-

respond to weight 0 modular forms. As another example, we have Theo-
rem 2.2, which asserts that the partition number generating function satisfies
∞
  ∞ 2
1 qn
= 1 + .
1 − qn (1 − q)2 (1 − q 2 )2 · · · (1 − q n )2
n=1 n=1

Since this series is essentially the reciprocal of Dedekind’s weight 1/2 mod-
ular form, this provides another example of an Eulerian series which is a
modular form.
The literature on such identities is extensive, and the pursuit of further
identities and their interpretations remains an active area of research largely
due to applications in combinatorics, Lie theory, number theory and physics
(for example, see [13], [129], and [145] to name a few).
Here we illustrate how one may easily obtain modular forms from Euler-
ian series by forcing the cancellation of non-holomorphic parts of harmonic
Maass forms. To this end, we again consider Dyson’s generating function
R(w; q)
∞
 ∞
 ∞
 2
qn
R(w; q) := 1 + N (m, n)wm q n = 1 + ,
(wq; q)n (w−1 q; q)n
n=1 m=−∞ n=1

and the functions N (r, t; n).

By conjugating Ferrers boards, we have the trivial relation
(10.1) N (r, t; n) = N (t − r, t; n).
However, these functions also satisfy [31] some highly non-trivial sporadic
identities such as
(10.2) N (1, 7; 7n + 1) = N (2, 7; 7n + 1) = N (3, 7; 7n + 1).
400 K. ONO

Atkin and Swinnerton-Dyer [31] proved some surprising further identities

such as (see also (5.19) of [119])
(10.3)
∞
(q; q 7 )2∞ (q 6 ; q 7 )2∞ (q 7 ; q 7 )2∞ 
− = (N (0, 7; 7n + 6) − N (1, 7; 7n + 6)) q n .
(q; q)∞
n=0

This identity expresses a weight 1/2 modular form as a linear combination

of Eulerian series.
This example is a special case of a much more general phenomenon
which gives highly nontrivial relationships between Dyson’s partition rank
functions. Generalizing it, Bringmann, Rhoades, and the author [57] used
the theory of harmonic Maass forms to obtain several infinite families of
modular forms as Eulerian series. Here we present examples which arise
directly from Theorem 9.5.
Theorem 10.1. Suppose that t ≥ 5 is prime, 0 ≤ r1 , r2 < t and 0 ≤ d <
t. Then the following are true:
 
(1) If 1−24d
t = −1, then
∞

(N (r1 , t; tn + d) − N (r2 , t; tn + d)) q 24(tn+d)−1
n=0

is a weight 1/2 weakly

 1−24d  holomorphic modular form on Γ1 (576t6 ).
(2) Suppose that t = 1. If r1 , r2 ≡ 2 (±1 ± α) (mod t), where α
1

is any integer for which 0 ≤ α < 2t and 1 − 24d ≡ α2 (mod 2t),

then
∞

(N (r1 , t; tn + d) − N (r2 , t; tn + d)) q 24(tn+d)−1
n=0

is a weight 1/2 weakly holomorphic modular form on Γ1 (576t6 ).

Remark 31. Example (10.3) is the t = 7 and d = 6 case of Theorem 10.1

(2). In this case, the only choices of r1 and r2 satisfying the hypotheses are
0, 1, and 6. Since N (1, 7; n) = N (6, 7; n), (10.3) is the only nontrivial
example of Theorem 10.1 (2) in this case. The proof of the theorem will
show, for all other pairs of r1 and r2 (apart from the trivial examples such
as those arising from (10.1)), that
∞

(N (r1 , t; tn + d) − N (r2 , t; tn + d)) q 24(tn+d)−1
n=0
is not a weakly holomorphic modular form. In other words, the correspond-
ing Maass forms turn out to have non-trivial non-holomorphic parts.
Remark 32. The “Mock theta Conjectures” (see Section 4), which were
proved by Hickerson, are instances of pairs of harmonic Maass forms which
 HARMONIC MAASS FORMS AND NUMBER THEORY 401

differ by a weakly holomorphic modular form. Using arguments along the

lines of the proofs in this section, A. Folsom has provided [94] a “Maass
form” proof of these conjectures.

Remark 33. In a paper, S.-Y. Kang [132] has further developed the re-
lationship between basic hypergeometric series and non-
holomorphic Jacobi forms using the work of Zwegers. In particular, she
gives (see Theorems 1.2 and 1.3 of [132]) two explicit families of weight
1/2 modular forms, described in terms of theta functions and Dedekind’s
eta-function, which are sums of basic hypergeometric series.

Theorem 10.1 gives modular forms as differences of the generating func-

tions for the functions N (r, t; tn + d). There are similar theorems where
the rank modulus t is independent of the modulus of the arithmetic pro-
gression of the sizes of the partitions. As before, for integers 0 < a < t,
2t
let ft := gcd(t,6) , t := lcm(2t2 , 24), and let "t := t /24. We then have the
following theorem which is contained in [57].

Theorem 10.2. Suppose that t > 1 is an odd integer. If 0 ≤ r1 , r2 < t

are integers, and P  6t is prime, then
 t
(N (r1 , t; n) − N (r2 , t; n)) q t n− 24
n≥1  
 
24t n−t −24t
P
=− P

 
is a weight 1/2 weakly holomorphic modular form on Γ1 144ft2 "t P 4 .

Theorems 10.1 and 10.2 depend on Theorem 9.5. However, one may
obtain many further results of this type by making use of results which are
analogous to Theorem 9.5. This is done in [57], where we obtain several
further infinite families of Eulerian modular forms. These families give rise
to further identities such as
(q 4 ; q 4 )5∞  q n(n+1) (−q 2 ; q 2 )n  q n(n+1) (−q 2 ; q 2 )n
2 = + .
(q 2 ; q 2 )4∞ (q; q 2 )2n+1
n≥0
(−q; q 2 )2n+1
n≥0

10.1. Proofs of Theorems 10.1 and 10.2. Theorem 10.2 follows

immediately from Theorem 9.6 since the p(n)/t summands cancel when
taking the difference of the relevant generating functions.
Now we turn to the proof of Theorem 10.1. By Theorem 9.5, for any
0 ≤ r < t we have
∞ 
 
p(n) 24t2 n−t2  "
A(r, t, n) · γ(t, y; n)q −t n
2 2
N (r, t; n) − q +
t
n=0 n∈Z
402 K. ONO

" t, n) is a
is a weight 1/2 harmonic Maass form on Γ1 (576t4 ). Here A(r,
complex number given by
√ 
(10.4) " t, n) = i 8
A(r, A(r, t, m),
m (mod 2t)
6m+1≡n (mod 12t)

where γ(t, y; n) and A(r, t; m) are defined in Theorem 9.5. Applying the
Atkin U (t2 ) operator, we have, by a straightforward generalization of Propo-
sition 1.5 of [192], that
(10.5)
∞  
p(n) 24n−1  "
A(r, t, n) · γ(t, y; n)q −n
2
R(r, t; z) := N (r, t; n) − q +
t
n=0 n∈Z

is a weight 1/2 harmonic Maass form on Γ1 (576t4 ).

Now we prove Theorem 10.1 (1). By a straightforward generalization of
the classical argument on twists of modular forms (for example,
 see Proposi-
tion 22 of [134]), the quadratic twist of R(r, t; z) by t· , say R(r, t; z)t , is a
weight 1/2 harmonic Maass form on Γ1 (576t6 ). In particular, −1 t R(r, t; z)t
has an expansion of the form
  ∞   
−1 1 − 24n p(n) 24n−1
R(r, t; z)t = N (r, t; n) − q
t t t
n=0

+ " t, n) · γ(t, y; n)q −n2 .
A(r,
n∈Z
tn
 
We find that R(r, t; z) − −1 6
t R(r, t; z)t is on Γ1 (576t ), and its non-holomo-
rphic part is supported on terms −t n . By taking the quadratic
2 2
 ·  of the form q
twist of this form again by t , to annihilate these non-holomorphic terms,
one then finds that
  
p(n) 24n−1
N (r, t; n) − q
t
n≥0,
( 1−24n
t )=−1
is weight 1/2 weakly holomorphic modular form. Using the orthogonality of
Dirichlet characters modulo t, and facts about twists again, it follows that
∞ 
 
p(tn + d) 24(tn+d)−1
N (r, t; tn + d) − q
t
n=0
is a weight 1/2 weakly holomorphic modular form. Theorem 10.1 (1) follows
by taking the difference of these forms when r = r1 and r2 . Since taking
twists of twists can be viewed as a single twist by the trivial character, we
find that the resulting form is on Γ1 (576t6 ).
Now we turn to the proof of Theorem 10.1 (2). Here we argue directly
with (9.1) and (10.5). Using the theory of twists of Maass forms again, we
see that the restriction of R(r1 , t; z)−R(r2 , t; z) to forms whose holomorphic
 HARMONIC MAASS FORMS AND NUMBER THEORY 403

parts are supported on exponents of the form 24(tn + d) − 1, is a weight 1/2

harmonic Maass form on Γ1 (576t6 ).
It suffices to show that the non-holomorphic part of this form is zero
under the given hypotheses on r1 and r2 . By (10.5), one sees that the non-
holomorphic part is supported on terms of the form q −n . By construction,
2

these n satisfy n ≡ α (mod 2t), for some 0 ≤ α < 2t with 1 − 24d ≡ α2

(mod 2t). Therefore, by (9.1) and (10.5), it suffices to show that
A(r1 , t, m) − A(r2 , t, m) = 0,
when 6m + 1 ≡ α (mod 12t). By (9.1), we have
t−1 
     
πj πjα
A(r1 , t, m) − A(r2 , t, m) = (−1) m
ζt−r1 j − ζt−r2 j sin sin .
t t
j=1
 
Since sin(x) = 1
2i eix − e−ix , we have that
t−1 
 πij πij πij πij
(1−2r1 −α) (−1−2r1 −α)
e t
(1−2r1 +α)
−e t −e t
(−1−2r1 +α)
+e t

j=1
πij πij πij πij

(1−2r2 −α) (−1−2r2 −α)
−e t
(1−2r2 +α)
+e t +e t
(−1−2r2 +α)
−e t = 0.

This follows since ±1 − 2ri ± α, for i = 1 and 2, are even and coprime to t.

11. Exact formulas

Armed with the theory of harmonic Maass forms, one may obtain exact
formulas for the coefficients of generating functions which turn out to be
holomorphic parts of such forms. By employing the method of Poincaré
series as described in Section 8.3, one obtains such results. Here we first
describe the recent resolution of the Andrews-Dragonette Conjecture, and
then we give exact formulas for generic harmonic Maass forms with weight
≤ 1/2.

11.1. The Andrews-Dragonette conjecture. Rademacher famous-

ly employed the modularity of the generating function for p(n) to perfect
the Hardy-Ramanujan asymptotic formula
1 √
(11.1) p(n) ∼ √ · eπ 2n/3
4n 3
to obtain his “exact” formula. To state his formula, let Is (x) be the usual
I-Bessel function of order s. Furthermore, if k ≥ 1 and n are integers, then
let
  x 
1 k 
(11.2) Ak (n) := χ12 (x) · e ,
2 12 12k
x (mod 24k)
x2 ≡−24n+1 (mod 24k)
404 K. ONO

where the sum runs over the residue classes modulo 24k, and where
 
12
(11.3) χ12 (x) := .
x
If n is a positive integer, then Rademacher’s formula reads [168]
∞
  √ 
− 34 Ak (n) π 24n − 1
(11.4) p(n) = 2π(24n − 1) · I3 .
k 2 6k
k=1

Here we address the problem of obtaining such exact formulas for Ne (n)
(resp. No (n)), the number of partitions of n with even (resp. odd) rank. To
obtain these results, we begin with the simple observation that
∞

f (q) = R(−1; q) = 1 + (Ne (n) − No (n))q n
n=1
∞ 2
qn
(11.5) =1+ .
(1 + q)2 (1 + q 2 )2 · · · (1 + q n )2
n=1

In view of (11.4) and (11.5), since

p(n) = Ne (n) + No (n),
our task is reduced to the problem of deriving exact formulas for the coeffi-
cients α(n) of the mock theta function
∞

f (q) = R(−1; q) = 1 + α(n)q n .
n=1

The problem of estimating the coefficients α(n) has a long history, one
which even precedes Dyson’s definition of partition ranks. Indeed, Ramanu-
jan’s last letter to Hardy already includes the claim that
  ⎛  ⎞
exp π 6 − 144
n 1
⎜ exp 2 π 6 − 144 ⎟
1 n 1
α(n) = (−1)n−1 +O⎝ ⎠.
2 n − 241
n − 24
1

Typical of his writings, Ramanujan offered no proof of this claim. Dragonette

proved this claim in her 1951 Ph.D. thesis [85], and Andrews [7] subse-
quently improved upon Dragonette’s work, and he proved16 that
√  
[ n ] (−1) 2  A
k+1 k(1+(−1)k )
 2k n − 4
α(n) = π(24n − 1)− 4
1

k
k=1
 √ 
π 24n − 1
(11.6) · I1 + O(n
).
2 12k

16This is a reformulation of Theorem 5.1 of [7] using the identity I 1 (z) =  2 2 ·

1

2 πz
sinh(z).
 HARMONIC MAASS FORMS AND NUMBER THEORY 405

This result falls short of the problem of obtaining an exact formula for
α(n). Andrews and Dragonette formulated the following conjecture (see
page 456 of [7], and Section 5 of [17]) for the coefficients α(n).
Conjecture. (Andrews-Dragonette)
If n is a positive integer, then
 
∞ (−1) 2  A
k+1
n− k(1+(−1)k )  √ 
2k 4 π 24n − 1
α(n) = π(24n − 1)− 4
1
·I1 .
k 2 12k
k=1
Bringmann and the author have proved [53] the following theorem.
Theorem 11.1. The Andrews-Dragonette Conjecture is true.
Remark 34. In her Ph.D. thesis (see [97, 98]), S. Garthwaite gener-
alized the proof of Theorem 11.1, and she obtained exact formulas for the
coefficients of Ramanujan’s mock theta function ω(q). These results will be
further generalized in the next subsection.
 1 of the Proof of Theorem 11.1. By Theorem 6.1, we have
Sketch
that D 2 ; z (see (8.12)) is a harmonic   weight 1/2 Maass form on Γ0 (144)
with Nebentypus character χ12 = 12· .
The idea behind the proof is simple. We  1 shall
 construct a Maass-
Poincaré series which we shall show equals D 2 ; z . The method is anal-
ogous to material in Section 8.3. The proof of the conjecture then follows
from the fact that the formulas in the Andrews-Dragonette Conjecture give
the coefficients of this Maass-Poincaré series.
 a b k ∈ 2 + Z. We define a class of Poincaré series Pk (s; z).
1
Suppose that
For matrices c d ∈ Γ0 (2), with c ≥ 0, define the character χ(·) by
(11.7)
    b
a b e − 24 if c = 0,
χ := 1   −1
c d i−1/2 (−1) 2 (c+ad+1) e − a+d
24c − 4 + 8
a 3dc
· ω−d,c if c > 0,
where
(11.8) ωd,c := eπis(d,c) .
Here s(d, c) denotes the classical Dedekind sum.
Throughout, let z = x + iy, and for s ∈ C, k ∈ 1
2 + Z, and y ∈ R \ {0},
and let
Ms (y) := |y|− 2 M k sgn(y), s− 1 (|y|),
k
(11.9)
2 2

where Mν,μ (z) again is the M -Whittaker function. Furthermore, let

 πy   x 
ϕs,k (z) := Ms − e − .
6 24
Using this notation, define the Poincaré series Pk (s; z) by
2 
(11.10) Pk (s; z) := √ χ(M )−1 (cz + d)−k ϕs,k (M z).
π
M ∈Γ∞ \Γ0 (2)
406 K. ONO

Here Γ∞ again is the subgroup of translations in SL2 (Z). 

The defining series is absolutely convergent for Pk 1 − k2 ; z for k <
1/2, and is conditionally convergent when k = 1/2. We are interested in
P 1 34 ; z , which we define by analytically continuing the Fourier expansion.
2
This argument is not straightforward (see Theorem 3.2 and  Corollary
 4.2 of
[53]). Thanks to the properties of Mν,μ , we find that P 1 34 ; 24z is a Maass
2
form of weight 1/2 for Γ0 (144) with Nebentypus χ12 .
A long calculation gives the following Fourier expansion
    
3 − 12 1 πy
· q − 24
1
P1 ;z = 1 − π · Γ ,
2 4 2 6

0 ∞

1 1
(11.11) + γy (n)q n− 24 + β(n)q n− 24 ,
n=−∞ n=1

where for positive integers n we have

(11.12)
 
∞
 (−1)
k+1

A2k n − k(1+(−1)k )  √ 
− 14
2
4 π 24n − 1
β(n) = π(24n − 1) · I1 .
k 2 12k
k=1
 
The Poincaré series P 1 34 ; z was defined so that (11.12) coincides with the
2
conjectured expressions for the coefficients α(n).
For convenience, we let
 
3
(11.13) P (z) := P 1 ; 24z .
2 4
Canonically decompose P (z) into a non-holomorphic and a holomorphic part
(11.14) P (z) = P − (z) + P + (z).
In particular, we have that
∞

+ −1
P (z) = q + β(n)q 24n−1 .
n=1
1 
Since P (z) and D 2 ; z are Maass forms of weight 1/2 for Γ0 (144) with
Nebentypus χ12 , (11.11) and (11.12) imply that the proof of the conjecture
reduces to proving that these forms are equal. This conclusion is obtained
after a lengthy and somewhat complicated argument. 

11.2. Exact formulas for harmonic Maass forms with weight

≤ 1/2. Generalizing the results of the previous section, Bringmann and the
author have obtained exact formulas for the coefficients of the holomorphic
parts of harmonic Maass forms with weight 2−k ≤ 1/2 [59]. Suppose that f
is in H2−k (N, χ), the space of weight 2 − k harmonic Maass forms on Γ0 (N )
 HARMONIC MAASS FORMS AND NUMBER THEORY 407

with Nebentypus character χ, where we assume that 32 ≤ k ∈ 12 Z. As usual,

we denote its Fourier expansion by
 
(11.15) f (z) = c+
f (n)q n
+ c−
f (n)Γ(k − 1, 4π|n|y)q .
n

n
−∞ n<0

It is our objective to determine exact formulas for the coefficients c+f (n) of
the holomorphic part of f .
We now define the functions which are required for these exact formulas.
Throughout, we let k ∈ 12 Z, and we let χ be a Dirichlet character modulo
 a b 4 | N whenever k ∈ 2 Z \ Z. Using this character, for a matrix
1
N , where
M = c d ∈ Γ0 (N ), we let

χ(d)   if k ∈ Z,
(11.16) Ψk (M ) :=
χ(d) dc 2k
d if k ∈ 12 Z \ Z,
 
where d is defined by (7.2),andwhere dc is the usual extended Legendre
symbol. In addition, if T = ac db ∈ SL2 (Z), then we let

(11.17) μ(T ; z) := (cz + d)2−k .

Moreover, for pairs of matrices S, T ∈ SL2 (Z), we then let

μ(T ; Sz)μ(S; z)
(11.18) σ(T, S) := .
μ(T S; z)
Using this notation, we now define certain generic Kloosterman sums which
are naturally associated with cusps of Γ0 (N ).
Suppose that ρ = cρρ = L−1 ∞, (L ∈ SL2 (Z)) is a cusp of Γ0 (N ) with
a

cρ |N and gcd(aρ , N ) = 1. Let tρ and κρ be the cusp width and parameter

of ρ with respect to Γ0 (N ) (see 11.21). Suppose that c > 0 with cρ |c and
cρ  c. Then for integers n and m we have the Kloosterman sum
N

(11.19)
Kc (2 − k, ρ, χ, m, n)
   
σ(L−1 , S) 2πi (m + κρ )a
:= · exp + nd ,
Ψk (L−1 S) c tρ
0<d<c
0<a<ct
aρ a≡− cc (mod N
cρ
)
ρ
(ad,c)=1
a b
where S := c d ∈ SL2 (Z) is the unique matrix defined using the integers
a, c, and d. Using properties of σ and Ψk , one can easily show that (11.19)
is well-defined.
For convenience, we let SN be a subset of SL2 (Z) with the property that
# −1 $
S ∞ : S ∈ SN
408 K. ONO
 
represents the set of inequivalent cusps of Γ0 (N ). For M = ac db ∈ SL2 (Z),
we define
 
k−2 az + b
(11.20) fM (z) := (cz + d) f ,
cz + d
√
where z is the principal branch of the holomorphic square root. Using
this notation, we have the Fourier expansion of a form f at a cusp ρ. More
precisely, if L ∈ SN with ρ = L−1 ∞, then we have
 n+κρ
(11.21) fρ (z) = a+
ρ (n)q tρ
+ fρ− (τ ).
n∈Z
We define the principal part of f at ρ by
 m+κρ
(11.22) Pf,ρ (z) := a+
ρ (m)q tρ
.
m+κρ <0

We shall use the principal parts of a form f to determine our exact

formulas. To this end, we identify, for each cusp ρ, its contribution to the
exact formula. To make this precise, let M = L−1 and μ = L∞. For positive
n, we then define
(11.23)
! ! k−1
ik 2π ! (−m + κμ ) ! 2
A(N, 2 − k, χ, ρ, m, c; n) := − ! !
tμ ! tμ n !
 % 
 Kc (2 − k, μ, χ, −m, −n) 4π n| − m + κμ |
× · Ik−1 .
c c tμ
c>0
cμ |c, cN c
μ

Here tμ and κμ are the cusp parameters for μ as in the notation above.
Using this notation, we define the order N Kloosterman approximation
of c+
f (n) by

  N

(11.24) C(f, N ; n) := a+
ρ (m) A(N, 2 − k, χ, ρ, m, c; n).
L∈SN m+κρ <0 c=1

Moreover, we define C(f, ∞; n) in the obvious way.

Remark 35. We stress again that L and ρ are related (throughout this
section) by the formula ρ = L−1 ∞.
Theorem 11.2. If f ∈ H2−k (N, χ) with 2 ≤ k ∈ 12 Z, then for positive n
we have
f (n) = C(f, ∞; n).
c+
Remark 36. Using the asymptotic behavior of I-Bessel functions, an
inspection of the principal parts of f gives a minimal N for which
C(f, N ; n) ∼ c+
f (n).
 HARMONIC MAASS FORMS AND NUMBER THEORY 409

Remark 37. Theorem 11.2 includes the classic results of Rademacher

and Zuckerman [169, 170, 216, 217] in the very special case of those
f ∈ H2−k (N, χ) for which 2 − k < 0 and f − = 0. It also recovers some
results of Bruinier, Hejhal, and Niebur [62, 117, 160, 161] for harmonic
Maass forms of non-positive weight. Recent work by Bruinier and Strömberg
[69] addresses the problem of efficiently computing coefficients of harmonic
Maass forms.
For weight 2 − k = 12 , we have a conditional result. To make it precise,
we say that a form f ∈ H 1 (N, χ) is good if the Maass-Poincaré series cor-
2
responding to nontrivial terms in the principal parts of f are individually
convergent.
Theorem 11.3. If f ∈ H 1 (N, χ) is good, then there is a finite set SΘ (f )
2
of complex numbers such that for positive n we have
f (n) = C(f, ∞; n) + μ
c+
for some μ ∈ SΘ (f ). Moreover, if n = dm2 for some d | N and m ∈ Z+ ,
then μ = 0.
Remark 38. We believe that all f ∈ H 1 (N, χ) are good. In earlier work
2
we deduced convergence of such Maass-Poincaré series by making using of
relationships between Kloosterman sums and Salié sums (see Section 4 of
[53]), and by generalizing work of Goldfeld and Sarnak [100] on sums of
Kloosterman sums (see [96]). It seems likely that a careful application of
these ideas will prove that each such f is indeed good.
Sketch of the Proof of Theorem 11.2 and 11.3. By the general
theory of Maass-Poincaré series, which generalizes Section 8.3, we have an
explicit linear combination of Maass-Poincaré series, say f ∈ H2−k (N, χ),
whose principal parts agree with the principal parts of f up to additive
constants. The complex number C(f, ∞; n) is the nth coefficient of the
holomorphic part of this linear combination. For the weight 2−k = 1/2, one
must argue further (for example, using the equidistribution of CM points)
to establish that these formulas are convergent.
There are then three possibilites:
Case 1. We have that f − f is a holomorphic modular form. It can only
be nonzero when 2 − k = 12 , in which case the Serre-Stark Basis Theorem
[189] implies that f − f is a linear combination of theta functions, giving
Theorem 11.3.
Case 2. We have that f − f is a weakly holomorphic modular form which is
not a holomorphic modular form. Such a form must have a pole at a cusp.
However, this cannot happen since we constructed f so that the principal
parts of f − f are constant.
Case 3. We have that f −f is a harmonic Maass form with a non-trivial non-
holomorphic part. However, Lemma 7.5 shows that all such harmonic Maass
410 K. ONO

forms have at least one principal part which is nonconstant. Therefore, this
possibility never occurs. 

12. Applications to modular forms

Here we use the theory of harmonic Maass forms to address two classical
problems in the theory of modular forms:
• Linear relations among Poincaré series
• Vanishing of Hecke eigenvalues.
We are able to obtain new results by making use of Lemma 7.4, the fact
that
ξ2−k : H2−k (Γ) −→ Sk (Γ).
These classical problems, which pertain to properties of the Fourier coeffi-
cients of cusp forms in Sk (Γ), can be investigated by leveraging information
from harmonic Maass forms in H2−k (Γ).

12.1. Relations among classical Poincaré series. It is well known

that the Poincaré series (see Section 8.3)
{P (m, k, N ; z) : m ≥ 1}
span the space Sk (N ). Since the space Sk (N ) is finite dimensional, there
exist many relations among these Poincaré series. In his classic text on auto-
morphic forms, Iwaniec [125] states the following seemingly simple problem.
Problem. Find all the linear relations between P (m, k, N ; z).
Remark 39. Iwaniec states this problem for more general multiplier
systems, but the techniques here also work in that setting.
The next theorem, due to Rhoades (see [178, 179]), follows from the
theory of harmonic Maass forms.
Theorem 12.1. Suppose that 2 ≤ k ∈ 12 Z, and suppose that I is a finite
set of positive integers. Then we have that

αm P (m, k, N ; z) ≡ 0
m∈I
!
if and only if there is a form in M2−k (N ) whose principal part at ∞ is
 αm
· q −m
mk−1
m∈I

with trivial principal parts at all other cusps.

Example 12.2. In practice it is not difficult to implement Theorem 12.1.
For example, consider the two dimensional space S24 (1). Obviously, there is
 HARMONIC MAASS FORMS AND NUMBER THEORY 411

a linear relation between the Poincaré series P (1, 24, 1; z), P (2, 24, 1; z) and
P (3, 24, 1; z). Using the formulas in Theorem 8.3, we find that
P (1, 24, 1; z) ∼1.00100852 · q + 132.988977 · q 2 + 189296.261 · q 3 + · · ·
P (2, 24, 1; z) ∼0.00001585 · q + 2.45743136 · q 2 + 114.854805 · q 3 + · · ·
P (3, 24, 1; z) ∼0.00000201 · q + 0.01023411 · q 2 + 0.88465633 · q 3 + · · · .
From these numerics, we find that
−0.00000207832·P (1, 24, 1; z)+0.00427703·P (2, 24, 1; z)+P (3, 24, 1; z) ∼ 0.
Although it is not possible to make this approximation precise by using
explicit Fourier expansions, Theorem 12.1 reduces this problem to a sim-
ple calculation. We find the exact relation thanks to the existence of the
modular form E14 /Δ3
E14 (z) 1 48 195660
3
= 3+ 2 − + ··· .
Δ(z) q q q
Therefore, we then find that the two coefficients in the linear combination
above are exactly
−195660 48 · 223
≈ −0.00000207832 . . . and ≈ 0.00427703 . . . .
323 323
Theorem 12.1 is a simple consequence of the theory of harmonic Maass
forms.

Proof of Theorem 12.1. Assume that m∈I αm P (m, k, N ; z) ≡ 0.
Define a weight 2 − k harmonic Maass form f by
 αm
f= · Q(−m, k, N ; z)
mk−1
m∈I
 
= c+
f (n)q n
+ c−
f (n)Γ(k − 1, 4π|n|y)q .
n

n
−∞ n<0

By Lemma 7.4 and Theorem 8.5, we find that


ξ2−k (f ) = (4π)k−1 (k − 1) αm P (m, k, N ; z)
m∈I

= −(4π) k−1
(k − 1) c−
f (−n)n
k−1 n
q .
n≥1

By our assumption on the sum over the Poincaré series, we know that
c−
f (n) = 0 for all n < 0, and so f ∈ M2−k (N ). Namely, in terms of the
!

Maass-Poincaré series Q(−m, k, N ; z) in Section 8.3, we have that

 αm
· Q+ (−m, k, N ; z) ∈ M2−k
!
(N )
mk−1
m∈I

is the weakly holomorphic form that we desire.

412 K. ONO

Conversely, assume that f is such a weakly holomorphic modular form.

From the expansion for the coefficients of Q+ (−m, k, N ; z) in Theorem 8.4,
we find that
1  αm
f" := −f + · · Q(−m, k, N ; z) ∈ H2−k (N )
Γ(k) mk−1
m∈I
is a harmonic Maass form with trivial principal parts at every cusp. By
Lemma 7.5, it then follows that f" is in M2−k
! (N ). Theorem 8.5 then implies
" 
that 0 = ξ2−k (f ) = m∈I αm P (m, k, N ; z). 
12.2. Algebraicity and the vanishing of Hecke eigenvalues. De-
spite the fact that we have a fairly complete theory of algebraicity for forms
in Mk! (N, χ), thanks to the q-expansion principle, the theory of Eisenstein
series and newforms, the question of algebraicity for harmonic Maass forms
remains open. As we shall now see, these questions (in the case of inte-
ger weight) are intimately related to the vanishing of eigenvalues of Hecke
operators.
Remark 40. The algebraicity of the coefficients of half-integral weight
harmonic Maass forms is much more mysterious. For certain weight 1/2
forms, we shall relate algebraicity to the vanishing of derivatives of quadratic
twists of weight 2 modular L-functions (see Section 15).
To make this precise, we shall restrict our attention to those f ∈ H2−k
(N, χ) for which ξ2−k (f ) ∈ Sk (N, χ) is a Hecke eigenform. For reasons which
will become apparent, we shall concentrate on those forms for which
g
(12.1) ξ2−k (f ) = ,
g2
where g is a normalized newform and g denotes its usual Petersson norm.
To illustrate the nature of our result, we consider two examples of Maass-
Poincaré series which are not weakly holomorphic modular forms. The
Maass-Poincaré series f := 11! 1
· Q(−1, 12, 1; z) ∈ H−10 (1) satisfies (12.1)
for g = Δ(z), the unique normalized weight 12 cusp form on the full modu-
lar group. The first few coefficients17 of its holomorphic part are
1 65520
· Q+ (−1, 12, 1; z) ∼ q −1 − − 1842.89472q − 23274.07545q 2 − · · · .
11! 691
There is little reason to believe that these coefficients are rational or alge-
braic. On the other hand, we shall prove that the Maass-Poincaré series
3! · Q(−1, 4, 9; z) ∈ H−2 (9) has the property that 3! · Q (−1, 4, 9; z) has
1 1 +

rational coefficients. Its first few terms are

1 1 49 5 48 8 771 11
(12.2) · Q+ (−1, 4, 9; z) = q −1 − q 2 + q − q − q + ··· ,
3! 4 125 512 1331
and f := 3!1 ·Q(−1, 4, 9; z) satisfies (12.1) for the unique normalized newform
in S4 (9).
17This corrects a typographical error for the constant term in [68].
 HARMONIC MAASS FORMS AND NUMBER THEORY 413

Our next theorem explains the distinction between these two cases. To
make this precise, let g ∈ Sk (N, χ) be a normalized newform, and let Fg be
the number field obtained by adjoining the coefficients of g to Q.
Definition 12.3. Assuming the hypotheses above, we say that a har-
monic Maass form f ∈ H2−k (N, χ) is good for g if it satisfies the following
properties:
(i) The principal part of f at the cusp ∞ belongs to Fg [q −1 ].
(ii) The principal parts of f at the other cusps of Γ0 (N ) are constant.
(iii) We have that ξ2−k (f ) = g−2 g.
Remark 41. For every such g, there is an f which is good for g. More-
over, such an f is unique up to a weakly holomorphic form in M2−k ! (N, χ)
with coefficients in Fg . Such f can be constructed explicitly using Poincaré
series.
In joint work with Bruinier and Rhoades [68], the author has obtained
the following result which explains this phenomenon.
Theorem 12.4. Let g ∈ Sk (N, χ) be a normalized newform with com-
plex multiplication (see (12.6)). If f ∈ H2−k (N, χ) is good for g, then all
coefficients of f + are in Fg (ζM ), where ζM := e2πi/M and M = N D, where
D is the discriminant of the field of complex multiplication.
Remark 42.
i) The rationality of Q+ (−1, 4, 9; z) in (12.2) is an example of Theorem 12.4.
In this case Q(−1, 4, 9; z) is good for η(3z)8 , the unique CM newform in
S4 (9).
ii) The field Fg in Theorem 12.4 is explicit, and is determined by Hecke
characters.
iii) Let g ∈ Sk (N, χ) and f ∈ H2−k (N, χ) be as in Theorem 12.4. The proof
of the theorem implies that all of the coefficients of f + belong to Fg (c+ f (1)).
In fact, the proof gives this conclusion for any newform g, not just those
with CM.
iv) In the examples we know, it turns out that the coefficients of f + are
actually contained in Fg . It is natural to ask whether this is true in general.
The proof of Theorem 12.4 relies on the fact that some Hecke eigenvalues
of g vanish. A simple generalization of the proof of Theorem 12.4 can be
used to detect the vanishing of the Fourier coefficients of a newform.

Theorem 12.5. Suppose that g = ∞ n=1 cg (n)q ∈ Sk (N, χ) is a nor-
n

malized newform, and suppose that f ∈ H2−k (N, χ) is good for g. If p  N

is a prime for which cg (p) = 0, then c+
f (pn) is algebraic for every positive n
coprime to p.
As in Remark 42 iv), it seems possible that the algebraic coefficients of
f + are always in Fg . We address this in the next example when N = 1.
414 K. ONO

Example 12.6. Here we consider Lehmer’s Conjecture on the nonvan-

ishing of Ramanujan’s τ -function, where
∞

Δ(z) = τ (n)q n .
n=1
This example generalizes easily to all level 1 Hecke eigenforms.
Although Theorem 12.5 relates Lehmer’s Conjecture to the alleged tran-
scendence of the coefficients, say aΔ (n), of 11! 1
· Q+ (−1, 12, 1; z), it turns out
that more is true. Lehmer’s Conjecture is implied by the irrationality of
these coefficients.
We make use of explicit formulas. Using the classical Eisenstein series E4
and E6 , and the classical j-function j(z), we define polynomials Jm (x) ∈ Z[x]
by
∞
 E4 (z)2 E6 (z) 1
(12.3) Jm (x)q m := · = 1 + (x − 744)q + · · · .
Δ(z) j(z) − x
m=0
For each m we then let jm (z) = Jm (j(z)). If p is prime, then define the
modular functions
24 p
(12.4) A p (z) := (1 + p 11
) + jp (z) − 264 σ9 (m)jp−m (z),
B12
m=1
 
24
(12.5) Bp (z) := −τ (p) −264 + + j1 (z) .
B12
Here B12 = − 2730
691
is the 12th Bernoulli number. Using the principal part
of Q(−1, 12, 1; z) combined with the fact that Δ(z) is an eigenform of the
Hecke algebra, Theorem 7.10 implies (also see [166]), for primes p, that
∞
   Ap (z) + Bp (z)
p11 aΔ (pn) − τ (p)aΔ (n) + aΔ (n/p) q n = .
n=−p
E4 (z)E6 (z)
These weight −10 modular forms have integer coefficients. Now suppose
that τ (p) = 0 for a prime p. Then aΔ (np) is rational for every n coprime
to p.
Remark 43. Results similar to the example above have also been ob-
tained recently by Boylan and Guerzhoy [45, 109].
Here we prove Theorem 12.4 by combining facts about ξ2−k , with Hecke
theory and the theory of complex multiplication. We first begin with an
important proposition.

Proposition 12.7. Let g = ∞ n=1 b(n)q ∈ Sk (N, χ) be a normalized
n

newform with integer weight k ≥ 2, and let Fg be the number field obtained
by adjoining the coefficients of g to Q. Then there is a harmonic Maass
form f ∈ H2−k (N, χ) which satisfies:
(i) The principal part of f at the cusp ∞ belongs to Fg [q −1 ].
 HARMONIC MAASS FORMS AND NUMBER THEORY 415

(ii) The principal parts of f at the other cusps of Γ0 (N ) are constant.

(iii) We have that ξ2−k (f ) = g−2 g.
Proof. Let H2−k,∞ (N, χ) be the subspace of those f ∈ H2−k (N, χ)
whose principal parts at the cusps other than ∞ are constant. Note that
!
H2−k (N, χ) = H2−k,∞ (N, χ) + M2−k (N, χ).
Arguing as in Section 3 of [63], the restriction of ξ2−k to H2−k,∞ (N, χ) de-
fines a surjective map to Sk (N, χ). One now argues as in the proof of Lemma
7.3 of [67] using the pairing {g, f } = (g, ξ2−k (f )), where f ∈ H2−k,∞ (N, χ)
and g ∈ Sk (N, χ). 
Remark 44. The harmonic Maass form f satisfying (i)–(iii) is unique up
!
to the addition of a weakly holomorphic form in M2−k (N, χ) with coefficients
in Fg and a pole possibly at infinity and constant principal part at all other
cusps.
For completeness, here we briefly recall the notion of a newform with
complex multiplication (for example, see Chapter 12 of [125] or Section 1.2
of [164]). Let D < 0 √ be the fundamental discriminant of an imaginary
 field K = Q( D). Let OK be the ring of integers of K, and let
quadratic
χK := D • be the usual Kronecker character associated to K. Let k ≥ 2,
and let c be a Hecke character of K with exponent k − 1 and conductor fc ,
a non-zero ideal of OK . By definition, this means that
c : I(fc ) −→ C×
is a homomorphism, where I(fc ) denotes the group of fractional ideals of K
prime to fc , and
c(αOK ) = αk−1
for α ∈ K × for which α ≡ 1 mod× fc . To c we naturally associate a Dirichlet
character ωc defined, for every integer n coprime to fc , by
c(nOK )
ωc (n) := .
nk−1
We then let

(12.6) ΦK,c (z) := c(a)q N (a) ,
a
where a varies over the ideals of OK prime to fc , and where N (a) is the
usual ideal norm. It is well known that ΦK,c (z) ∈ Sk (|D| · N (fc ), χK · ωc )
is a normalized newform. These are newforms with complex multiplication.
By construction, if we let
∞

ΦK,c (z) = b(n)q n ,
n=1
then
(12.7) b(n) = 0 whenever χK (n) = −1.
416 K. ONO

This follows since every prime p for which χK (p) = −1 is inert.

∞Proof nof Theorem 12.4. Suppose that f is good for a CM form g =

n=1 b(n)q , and let D = Dg be the fundamental
√ discriminant of the asso-
ciated imaginary quadratic field K = Q( D). By Lemma 7.4, we then have
that
∞

ξ2−k (f ) = g−2 g = −(4π)k−1 c−
f (−n)n
k−1 n
q .
n=1
Since g has complex multiplication, (12.7) implies that c−
f (n) = 0 when
χK (−n) = −1. Because D < 0, this means that
(12.8) c−
f (n) = 0 when χK (n) = 1.

Let M = N D. We write χ0 for the trivial character modulo |D|. Since

D | N , a simple generalization of standard facts about twists of modular
forms implies that
u := f ⊗ χ0 + f ⊗ χK
is in H2−k (M, χ). The Fourier expansion of u = u+ + u− is given by

u+ (z) = 2 c+ n
f (n)q ,
n
−∞
χK (n)=1

u− (z) = 2 c−
f (n)Γ(k − 1, 4π|n|y)q .
n

n<0
χK (n)=1

The non-holomorphic part u− vanishes, and so u is a weakly holomorphic

modular form.
We now claim that for any integer b, f (z + b/D) has principal parts at
all cusps in Fg (ζM )[q −1 ]. To see this, we let γ ∈ Γ(1) and consider the cusp
γ∞. There exists a γ " ∈ Γ(1) and α, β, δ ∈ Z such that
   
D b α β
γ=γ " .
0 D 0 δ
Hence, the Fourier expansion of f (z + b/D) at the cusp γ∞ is given by
 
α β
f |γ
"| .
0 δ
By the assumption of f , it is holomorphic at the cusp ∞, unless γ
" ∈ Γ0 (N ),
in which case it is equal to
 
α β
f| .
0 δ
Since δ | D2 | M , the principal part at ∞ of this modular form is contained
in Fg (ζM )[q −1 ], proving the claim. This implies that the twists f ⊗ χ0 ,
f ⊗ χD , have principal parts at all cusps in Fg (ζM )[q −1 ]. Therefore, the
same is true for u.
 HARMONIC MAASS FORMS AND NUMBER THEORY 417

The action of Aut(C/Q(ζN )) commutes with the action of SL2 (Z) on

modular functions for Γ(N ) (for example, see Theorem 6.6 in Chapter 6.2
and the diagram before Remark 6.7 in Shimura’s book [193]). Using the
action of Aut(C/Fg (ζM )) on weakly holomorphic modular forms, we see that
uσ has the same properties for any σ ∈ Aut(C/Fg (ζM )). Moreover, uσ has
the same principal parts as u at all cusps. Hence the difference u − uσ is a
holomorphic modular form which vanishes at the cusp ∞. Since 2 − k ≤ 0,
this implies that u = uσ . Consequently, u is defined over Fg (ζM ). So for
all n ∈ Z with χK (n) = 1, we have that c+ f (n) ∈ Fg (ζM ). In particular,
cf (1) ∈ Fg (ζM ).
+

We now use the Hecke action on f and g. By Theorem 7.10, we have

that
f |2−k T (m) = m1−k b(m)f + f  ,
where f  ∈ M2−k
! (Γ0 (N ), χ) is a weakly holomorphic form with coefficients
in Fg . In view of the formula for the action of the Hecke operators on the
Fourier expansion, we obtain for any prime p that
c+
f (pn) + χ(p)p
1−k +
cf (n/p) = p1−k b(p)c+ +
f (n) + cf  (n),

where c+f  (n) ∈ Fg . Hence an inductive argument shows that all coefficients
cf (n) are contained in the extension Fg (c+
+
f (1)). This concludes the proof of
the theorem since we have already established that c+ f (1) is in Fg (ζM ). 

The proof of Theorem 12.5 is similar to the proof of Theorem 12.4, and
so we only give a brief indication of how the proof must be modified.
Sketch proof of Theorem 12.5. If p  N is a prime, then for every
positive integer m we have that
cg (p)cg (pm ) = cg (pm+1 ) + χ(p)pk−1 cg (pm−1 ).
Therefore, if p  N is a prime for which cg (p) = 0, then we have that
cg (pm+1 ) = −χ(p)pk−1 cg (pm−1 ),
which in turn implies that
⎧ m
⎨ k−1 2
−χ(p)p if m is even,
cg (pm ) =
⎩0 otherwise.
A standard argument gives a harmonic Maass form whose Fourier coefficients
are supported on terms whose exponents n have the property that p exactly
divides n. By the multiplicativity of the Fourier coefficients of newforms, it
then follows by the observation above that the non-holomorphic part of this
form is identically zero. In other words, this particular harmonic Maass form
is a weakly holomorphic modular form with suitable principal parts at cusps.
The proof now follows mutatis mutandis as in the proof of Theorem 12.4. 
418 K. ONO

13. Generating functions for singular moduli

Let j(z) be the usual modular function for SL2 (Z)

j(z) = q −1 + 744 + 196884q + 21493760q 2 + · · · .

The values of modular functions such as j(z) at imaginary quadratic ar-

guments in H are known as singular moduli. Singular moduli are algebraic
integers which play many roles in number theory. For example, they generate
class fields of imaginary quadratic fields, and they parameterize isomorphism
classes of elliptic curves with complex multiplication.
In an important paper, Zagier [212] proved that the traces of these
numbers are Fourier coefficients of certain weight 3/2 modular forms. To
illustrate his results, we begin by comparing the classical evaluations
 √ 
j −1+2 −3 − 744 j(i) − 744
= −248, = 492,
3 √  2

1 + −7
j − 744 = −4119,
2

with the Fourier coefficients of the modular form

η(z)2 · E4 (4z)
g(z) := − = −q −1 + 2 − 248q 3
η(2z)η(4z)6
(13.1) + 492q 4 − 4119q 7 + 7256q 8 − · · · .

The appearance of singular moduli as the initial coefficients of the modular

form g(z) is not a coincidence.
We now make this more precise. For integers λ, let M!λ+ 1 (4) be the
2
space of weight λ + 12 weakly holomorphic modular forms on Γ0 (4) satisfying
the “Kohnen plus-space” condition. A form satisfies this condition if its q-
expansion has the form

(13.2) a(n)q n .
(−1)λ n≡0,1 (mod 4)

Throughout let d ≡ 0, 3 (mod 4) be a positive integer, let H(d) be the

Hurwitz-Kronecker class number for the discriminant −d, and let Qd be
the set of positive definite integral binary quadratic forms (note. including
imprimitive forms)

Q(x, y) = [a, b, c] = ax2 + bxy + cy 2

with discriminant DQ = −d = b2 − 4ac. For each Q, let τQ be the unique

root in H of Q(x, 1) = 0. The singular modulus f (τQ ), for any modular
invariant f (z), depends only on the equivalence class of Q under the action
 HARMONIC MAASS FORMS AND NUMBER THEORY 419

of Γ := PSL2 (Z). If ωQ ∈ {1, 2, 3} is given by

⎧
⎪
⎨2 if Q ∼Γ [a, 0, a],
(13.3) ωQ := 3 if Q ∼Γ [a, a, a],
⎪
⎩
1 otherwise,
then, for a modular invariant f (z), define the trace Tr(f ; d) by
 f (τQ )
(13.4) Tr(f ; d) := .
ωQ
Q∈Qd /Γ

Theorems 1 and 5 of Zagier’s paper [212] imply the following theorem.

Theorem 13.1. If f (z) ∈ Z[j(z)] has a Fourier  expansion with constant
term 0, then there is a finite principal part Af (z) = n≤0 af (n)q n for which

Af (z) + Tr(f ; d)q d ∈ M!3 (4).
2
0<d≡0,3 (mod 4)

Remark 45. The earlier claim in (13.1) is the f (z) = j(z) − 744 case of
this theorem.
Remark 46. Zagier’s paper [212] includes many generalizations of this
theorem. One of his results proves that the generating functions of the
traces of jm (z) := m (j(z) − 744) |T (m) are also explicit weight 3/2 weakly
holomorphic modular forms. This result is of particular significance because
it reduces the computation of Hilbert class polynomials to a straightforward
calculation.
13.1. Further generating functions for traces of singular mod-
uli. Zagier’s work has inspired an enormous number of works [54, 58, 64,
65, 77, 78, 87, 88, 89, 95, 127, 128, 154] by the author, Bringmann,
Bruinier, D. Choi, W. Duke, A. Folsom, J. Funke, O. Imamoḡlu, P. Jenkins,
D. Jeon, S.-Y. Kang, C. Kim, R. Masri, A. Miller, A. Pixton, J. Rouse,
and A. Toth. These papers give theorems related to the phenomenon that
the coefficients of modular forms and harmonic Maass forms are often the
“traces” of some sort of singular moduli. For brevity, here we discuss one
construction which is based on Maass-Poincaré series.
For fundamental discriminants D1 , let χD1 denote the associated genus
character for positive definite binary quadratic forms whose discriminants
are multiples of D1 . If λ is an integer and D2 is a non-zero integer for which
(−1)λ D2 ≡ 0, 1 (mod 4) and (−1)λ D1 D2 < 0, then define the twisted trace
of a modular invariant f (z), say TrD1 (f ; D2 ), by
 χD1 (Q)f (τQ )
(13.5) TrD1 (f ; D2 ) := .
ωQ
Q∈ Q|D1 D2 | /Γ

Many of the works mentioned above prove that such traces are often
coefficients of Maass forms and weakly holomorphic modular forms. Here we
420 K. ONO

give one illustrative example of this phenomenon using some of the Poincaré
series constructed in Section 8.3. For convenience, we briefly recall their
construction. Let k := λ + 12 , where λ is an arbitrary integer, and let
Mν, μ (z) be the usual M -Whittaker function. For s ∈ C and y ∈ R − {0},
we define
Ms (y) := |y|− 2 M k sgn(y), s− 1 (|y|).
k

2 2

Suppose that m ≥ 1 is an integer with (−1)λ+1 m ≡ 0, 1 (mod 4). As before,

let
ϕ−m,s (z) := Ms (−4πmy)e(−mx),
where z = x + iy and e(w) := e2πiw , and let Γ∞ denote the translations in
SL2 (Z). With this notation, define the Poincaré series

(13.6) Fλ (−m, s; z) := (ϕ−m,s |k A)(z)
A∈Γ∞ \Γ0 (4)

for Re(s) > 1. Here |k denotes the usual half-integral weight k “slash oper-
ator”. If prλ is Kohnen’s projection operator (see page 250 of [136]) to the
weight λ + 12 plus-space for Γ0 (4), then for λ ∈ {0, 1} define Fλ (−m; z) by
  
2 Fλ −m, 2 ; z  | prλ if λ ≥ 2,
3 k
(13.7) Fλ (−m; z) := 
2(1−k)Γ(1−k) Fλ −m, 1 − 2 ; z | prλ if λ ≤ −1.
3 k

Remark 47. For λ = 0 or 1 we also have series Fλ (−m; z). Their

construction requires more care due to questions of convergence (see [54,
65]).

Arguing as in Section 8.3, if λ ≥ −6 with λ = −5, then Fλ (−m; z) ∈

M!λ+ 1 (4). For such λ, we denote the corresponding Fourier expansions by
2

(13.8) Fλ (−m; z) = q −m + bλ (−m; n)q n ∈ M!λ+ 1 (4).
2
n≥0
(−1)λ n≡0,1 (mod 4)

For other λ, namely λ = −5 or λ ≤ −7, it turns out that the Fλ (−m; z) are
Maass forms in Hλ+ 1 (4). We denote their expansions by
2

(13.9) Fλ (−m; z) = Bλ (−m; z) + q −m + bλ (−m; n)q n ,
n≥0
(−1)λ n≡0,1 (mod 4)

where Bλ (−m; z) is the “non-holomorphic” part of Fλ (−m; z).

Example 13.2. If λ = 1 and −m = −1, then we find the form in (13.1)

−F1 (−1; z) = g(z) = −q −1 + 2 − 248q 3 + 492q 4 − 4119q 7 + 7256q 8 − · · · .

 HARMONIC MAASS FORMS AND NUMBER THEORY 421

Generalizing Zagier’s results, we have that the coefficients bλ (−m; n) of

the Fλ (−m; z) are traces of singular moduli for functions defined by Niebur
[160]. If Is (x) denotes the usual I-Bessel function, and if λ > 1, then let
 1
(13.10) Fλ (z) := π Im(Az) 2 Iλ− 1 (2πIm(Az))e(−Re(Az)).
2
A∈Γ∞ \SL2 (Z)

For λ = 1, a similar construction holds. Indeed, Niebur [160] showed that

F1 (z) = 12 (j(z) − 744), where this function is the analytic continuation, as
s → 1+ , of
 1
−12 + π Im(Az) 2 Is− 1 (2πIm(Az))e(−Re(Az)).
2
A∈Γ∞ \SL2 (Z)

We then have the following theorem.

Theorem 13.3. (Bringmann and Ono; Theorem 1.2 of [54])
If λ, m ≥ 1 are integers for which (−1)λ+1 m is a fundamental discriminant
(note. which includes 1), then for each positive integer n with (−1)λ n ≡ 0, 1
(mod 4) we have
2(−1)[(λ+1)/2] n 2 − 2
λ 1

bλ (−m; n) = λ · Tr(−1)λ+1 m (Fλ ; n) .

m2
Remark 48. A version of Theorem 13.3 holds for integers λ ≤ 0. This
follows from a duality (see Theorem 1.1 of [54]) of Fourier expansions. Sup-
pose that λ ≥ 1, and that m is a positive integer for which (−1)λ+1 m ≡ 0, 1
(mod 4). For every positive integer n with (−1)λ n ≡ 0, 1 (mod 4), this
duality implies that
bλ (−m; n) = −b1−λ (−n; m).
Example 13.4. For λ = 1, Theorem 13.3 relates b1 (−m; n) to traces
and twisted traces of F1 (z) = 12 (j(z) − 744). These are Theorems 1 and 6 of
Zagier’s paper [212].
Theorem 13.3 is obtained by reformulating, as traces of singular moduli,
exact expressions for the coefficients bλ (−m; n). The proof follows from
the classical fact that certain half-integral weight Kloosterman sums are
essentially Salié sums. To define these sums, suppose that 0 = D1 ≡ 0, 1
(mod 4). If λ is an integer, D2 = 0 is an integer for which (−1)λ D2 ≡ 0, 1
(mod 4), and N is a positive multiple of 4, then define the generalized Salié
sum Sλ (D1 , D2 , N ) by
  
N x2 − (−1)λ D1 D2
Sλ (D1 , D2 , N ) := χD1 , x,
4 N
x (mod N )
x2 ≡(−1)λ D1 D2 (mod N )
 
2x
(13.11) ×e ,
N
422 K. ONO

where χD1 (a, b, c), for a binary quadratic form Q = [a, b, c], is given by
⎧
⎪
⎨0  if (a, b, c, D1 ) > 1,
D1
(13.12) χD1 (a, b, c) := if (a, b, c, D1 ) = 1 and Q represents
⎪
⎩
r
r with (r, D1 ) = 1.
Remark 49. If D1 = 1, then χD1 is trivial. Therefore, if (−1)λ D2 ≡ 0, 1
(mod 4), then
  
2x
Sλ (1, D2 , N ) = e .
N
x (mod N )
x2 ≡(−1)λ D2 (mod N )

Half-integral weight Kloosterman sums are essentially equal to such Salié

sums, a fact which plays a fundamental role throughout the theory of half-
integral weight modular forms. The following proposition is due to Kohnen
(see Proposition 5 of [136]).
Proposition 13.5. Suppose that N is a positive multiple of 4. If λ is an
integer, and D1 and D2 are non-zero integers for which D1 , (−1)λ D2 ≡ 0, 1
(mod 4), then
N − 2 (1 − (−1)λ i)(1 + δodd (N/4)) · Kλ ((−1)λ D1 , D2 , N ) = Sλ (D1 , D2 , N ).
1

As a consequence, we may rewrite the formulas for the bλ (−m; n) using

Salié sums. The following proposition describes these Salié sums as Poincaré-
type series over CM points.
Proposition 13.6. Suppose that λ is an integer, and that D1 is a funda-
mental discriminant. If D2 is a non-zero integer for which (−1)λ D2 ≡ 0, 1
(mod 4) and (−1)λ D1 D2 < 0, then for every positive integer a we have
 χD1 (Q) 
Sλ (D1 , D2 , 4a) = 2 e (−Re (AτQ )) .
ωQ
Q∈Q|D1 D2 | /Γ A∈Γ∞ \SL
√2 (Z)
|D1 D2 |
Im(AτQ )= 2a

Proof. For every integral binary quadratic form

Q(x, y) = ax2 + bxy + cy 2
of discriminant (−1)λ D1 D2 , let τQ ∈ H be as before. Clearly τQ is equal to

−b + i |D1 D2 |
(13.13) τQ = ,
2a
and the coefficient b of Q solves the congruence
(13.14) b2 ≡ (−1)λ D1 D2 (mod 4a).
Conversely, every solution of (13.14) corresponds to a quadratic form with an
associated CM point thereby providing a one-to-one correspondence between
the solutions of
b2 − 4ac = (−1)λ D1 D2 (a, b, c ∈ Z, a, c > 0)
 HARMONIC MAASS FORMS AND NUMBER THEORY 423

and the points of the orbits

# $
AτQ : A ∈ SL2 (Z)/ΓτQ ,
Q

where ΓτQ denotes the isotropy subgroup of τQ in SL2 (Z), and where Q
varies over the representatives of Q|D1 D2 | /Γ. The group Γ∞ preserves the
imaginary part of such a CM point τQ , and preserves (13.14). However, it
does not preserve the middle coefficient b of the corresponding quadratic
forms modulo 4a. It identifies the congruence classes b, b + 2a (mod 4a)
appearing in the definition of Sλ (D1 , D2 , 4a). Since χD1 (Q) is fixed under
the action of Γ∞ , the corresponding summands for such pairs of congruence
classes are equal. Proposition 13.6 follows since #ΓτQ = 2ωQ , and since
both ΓτQ and Γ∞ contain the negative identity matrix. 

Sketch of the Proof of Theorem 13.3. Here we prove the cases

where λ ≥ 2. The argument when λ = 1 is identical. For λ ≥ 2, the exact
formula for bλ (−m; n) is
√
bλ (−m; n) = (−1)[(λ+1)/2] π 2(n/m) 2 − 4 (1 − (−1)λ i)
λ 1

  √ 
Kλ (−m, n, c) 4π mn
× (1 + δodd (c/4)) · Iλ− 1 .
c 2 c
c>0
c≡0 (mod 4)

Using Proposition 13.5, where D1 = (−1)λ+1 m and D2 = n, for integers

N = c which are positive multiples of 4, we have

c− 2 (1 − (−1)λ i)(1 + δodd (c/4)) · Kλ (−m, n, c) = Sλ ((−1)λ+1 m, n, c).

1

These identities, combined with the change of variable c = 4a, give

∞
(−1)[(λ+1)/2] π 1  Sλ ((−1)
λ+1 m, n, 4a)
(n/m) 2 − 4
λ
bλ (−m; n) = √ √
2 a=1
a
 √ 
π mn
· Iλ− 1 .
2 a

Using Proposition 13.6, this becomes

2(−1)[(λ+1)/2] π  χ(−1)λ+1 m (Q)

(n/m) 2 − 4
λ 1
bλ (−m; n) = √
2 ωQ
Q∈Qnm /Γ
∞
  Iλ− 1 (2πIm(AτQ ))
2
√ · e(−Re(AτQ )).
a=1 A∈Γ∞ \SL2 (Z)
a
√
mn
Im(AτQ )= 2a
424 K. ONO

The definition of Fλ (z) in (13.10), combined with the obvious change of

√ 1
variable relating 1/ a to Im(AτQ ) 2 , gives

2(−1)[(λ+1)/2] n 2 − 2
λ 1
 χ(−1)λ+1 m (Q)
bλ (−m; n) = λ ·π
m 2 ωQ
Q∈Qnm /Γ
 1
Im(AτQ ) 2 · Iλ− 1 (2πIm(AτQ ))e(−Re(AτQ ))
2
A∈Γ∞ \SL2 (Z)

2(−1)[(λ+1)/2] n 2 − 2
λ 1

= λ · Tr(−1)λ+1 m (Fλ ; n).

m2


13.2. The “24-Theorem”. Exact formulas for traces of singular mod-

uli can lead to nice number theoretic consequences. Here we mention one
such application which is related to the classical observation that
√
(13.15) eπ 163
= 262537412640768743.9999999999992 . . .
is nearly an integer. To make this precise, we recall some classical facts. A
primitive positive definite binary quadratic form Q is reduced if |B| ≤ A ≤
C, and B ≥ 0 if either |B| = A or A = C. If −d < −4 is a fundamental
discriminant, then there are H(d) reduced forms with discriminant −d. The
set of such reduced forms, say Qred
d , is a complete set of representatives
 for
Qd /SL2 (Z). Moreover, each such reduced form has 1 ≤ A ≤ d/3 (see
page 29 of [82]), and has the property that τQ lies in the usual fundamental
domain for the action of SL2 (Z)
(13.16)
   
1 1 1
F = − ≤ Re(z) < and |z| > 1 ∪ − ≤ Re(z) ≤ 0 and |z| = 1 .
2 2 2
Since J1 (z) := j(z) − 744 = q −1 + 196884q + · · · , it follows that if Gred (d) is
defined by
 √
(13.17) Gred (d) = eπBi/A · eπ d/A ,
Q=(A,B,C)∈Qred
d

then Tr(d) − Gred (d) is “small”, where Tr(d) := Tr(J1 ; d). In other words,
q −1 provides a good approximation for J1 (z) for most points z. This is
illustrated by (13.15) where H(163) = 1.
It is natural to investigate the “average value”
Tr(d) − Gred (d)
,
H(d)
which for d = 163 is −0.0000000000008 . . . . Armed with the exact formulas
for Tr(d), it turns out that a uniform picture emerges for a slightly perturbed
 HARMONIC MAASS FORMS AND NUMBER THEORY 425

average, one including some non-reduced quadratic forms. For each positive
integer A, let Qold
A,d denote the set
(13.18)
A,d = {Q = (A, B, C) : non-reduced with DQ = −d and |B| ≤ A}.
Qold
Define Gold (d) by
 √
(13.19) Gold (d) = eπBi/A · eπ d/A
.
√ √
d/2≤A≤ d/3
Q∈Qold
A,d

The non-reduced forms Q contributing to Gold (d) are those primitive dis-
criminant −d forms for which τQ is in the bounded region obtained by
connecting the two endpoints of the lower boundary of F with a horizontal
1
line. Since Tr(d) is subexponential in |d| and H(d)  |d| 2 +
, the following
numerics are quite surprising:
⎧
⎪−24.672 . . . if d = 1931,
Tr(d) − Gred (d) − Gold (d) ⎨
= −24.483 . . . if d = 2028,
H(d) ⎪
⎩
−23.458 . . . if d = 2111.
Recently, Duke has proved [87] a result which implies the following theorem.
Theorem 13.7. As −d ranges over negative fundamental discriminants,
we have
Tr(d) − Gred (d) − Gold (d)
lim = −24.
−d→−∞ H(d)
Here we explain the source of −24 in the limit
Tr(d) − Gred (d) − Gold (d)
(13.20) lim = −24.
−d→−∞ H(d)
Using Theorem 13.3 and Propositions 13.5 and 13.6, it is not difficult to
reformulate the exact formulas for Tr(d) to read
 √
Tr(d) = −24H(d) + S(d, c) sinh(4π d/c),
c>0
c≡0 (4)

where S(d, c) is the Salié sum


S(d, c) = e(2x/c).
x2 ≡−d (mod c)

As a consequence, the “24 Theorem” is equivalent to the assertion that

  
4π √
S(d, c) sinh d = o (H(d)) .
√ c
c> d/3
c≡0 (4)
426 K. ONO

This follows from the fact the sum over c ≤ d/3 is essentially Gred (d) +
Gold (d). The sinh factor contributes the size of q −1 in the Fourier expansion
of a singular modulus, and the summands in the Kloosterman sum provide
the corresponding “angles”. The contribution Gold (d) arises from the fact
that the Kloosterman sum cannot distinguish between reduced and non-
reduced forms. In view of Siegel’s theorem that H(d) 
d 2 −
, (13.20)
1

follows from a bound for such sums of the form  d 2 −γ , for some γ >
1

0. Such bounds are implicit in Duke’s proof of this result [87], and more
generally in his famous work on bounding coefficients of half-integral weight
cusp forms [86].

14. Borcherds products

Here we describe work of the author and Bruinier [67] on Borcherds
products. To describe the context of this work, we first recall a famous result
of Borcherds on infinite product expansions of modular forms possessing a
Heegner divisor.

14.1. The “classical” Borcherds products. We begin by recall-

ing that Δ(z), the unique normalized weight 12 cusp form with respect
to SL2 (Z), has a Fourier expansion given by the infinite product
∞

(14.1) Δ(z) = q (1 − q n )24 .
n=1

Most modular forms do not possess simple infinite product expansions. It

was widely believed that such elegant product expansions only belong to the
domain of those modular forms whose divisors are supported at cusps, such
as the modular units (for example, see [139]).
Then in the 1990s, Borcherds [42, 43, 44] provided a striking description
for the exponents in the naive infinite product expansions of a much larger
class of modular forms, those modular forms possessing a Heegner divisor.
For example, let c(n) denote the integer exponents one obtains by expressing
E4 (z) as an infinite product:
(14.2)
∞ 
 ∞

E4 (z) = 1 + 240 d3 q n = (1 − q)−240 (1 − q 2 )26760 · · · = (1 − q n )c(n) .
n=1 d|n n=1

Borcherds’s theory implies that there is a weight 1/2 meromorphic modular

form

G(z) = b(n)q n = q −3 + 4 − 240q + 26760q 4 + · · · − 4096240q 9 + . . .
n≥−3

with the property that c(n) = b(n2 ) for every positive integer n. Notice that
the product in (14.1) also satisfies this phenomenon, where the product
 HARMONIC MAASS FORMS AND NUMBER THEORY 427

exponents are the coefficients of the weight 1/2 modular form

 2
12Θ0 (z) = 12 q n = 12 + 24q + 24q 4 + 24q 9 + · · · .
n∈Z

These two examples illustrate his general result that a modular form on
SL2 (Z) with a Heegner divisor has an infinite product expansion where the
exponents are coefficients of certain weight 1/2 weakly holomorphic modular
forms.
Remark 50. The work of Borcherds [42, 43, 44] is given in the more
general context of automorphic forms on orthogonal groups.
Remark 51. It is possible to derive explicit formulas for the exponents
in the infinite product expansions of generic modular forms, not just those
with a Heegner divisor (for example, see [66]). However, these formulas are
in general quite complicated.
We now briefly recall the most classical case of this work of Borcherds.
We shall refer to a complex number τ ∈ H of the form
√
−b + b2 − 4ac
τ=
2a
with a, b, c ∈ Z, gcd(a, b, c) = 1, and b2 − 4ac < 0 as a CM point, and
we denote its discriminant by the integer dτ := b2 − 4ac. A meromorphic
modular form f (z) on SL2 (Z) is said to have a Heegner divisor if its zeros
and poles are supported at the cusp at infinity and at CM points.
To state Borcherds’s results, we require a special sequence of weight
1/2 modular forms in M!1 (4). These forms will be distinguished by their
2
q-expansions. For each nonnegative integer d ≡ 0, 3 (mod 4) let fd (z) ∈
M!1 (4) be the unique modular form with a Fourier expansion of the form
2

fd (z) = q −d + A(D, d)q D .
D>0

That these forms are well defined follows from Lemma 14.2 of [43]. More-
over, they form a basis of M!1 (4).
2
The form f0 (z) is the classical Jacobi theta-function
(14.3) f0 (z) = Θ0 (z) = 1 + 2q + 2q 4 + 2q 9 + 2q 16 + · · · ,
and the form f3 (z) is given by the expression
D(Θ0 (z))E10 (4z) Θ0 (z) (D(E10 (z))|V (4)) 152
f3 (z) = − − Θ0 (z)
(14.4) 2Δ(4z) 10Δ(4z) 5
= q −3 − 248q + 26752q 4 − 85995q 5 + · · · ,

where D := q · d
dq .
428 K. ONO

Remark 52. It is straightforward to inductively compute the fd (z) using

f0 (z) and f3 (z). To compute fd (z) for d ≥ 4, one computes fd−4 (z)j(4z),
and then iteratively subtracts multiples of those fj (z) with 0 ≤ j ≤ d − 4.
For completeness, we include the initial terms of the first few fd (z) below.
(14.5)
f0 (z) = 1 + 2q + 2q 4 + 2q 9 + · · · ,
f3 (z) = q −3 − 248q + 26752q 4 − 85995q 5 + 1707264q 8 − 4096248q 9 + · · · ,
f4 (z) = q −4 + 492q + 143376q 4 + 565760q 5 + 18473000q 8
+ 51180012q 9 + · · · ,
f7 (z) = q −7 − 4119q + 8288256q 4 − 52756480q 5 + · · · ,
f8 (z) = q −8 + 7256q + 26124256q 4 + 190356480q 5 + · · · .
Using these forms, we may now state Borcherds’s famous theorem on the
infinite product expansion of those meromorphic forms on SL2 (Z) possessing
a Heegner divisor. Let MH denote the set of integer weight meromorphic
modular forms on SL2 (Z) with a Heegner divisor, integer coefficients, and
leading coefficient 1. Obviously, MH is closed under multiplication.
If H(−n) denotes the usual Hurwitz class number of discriminant −n,
"
then define H(z) by
1 
"
H(z) :=− + H(−n)q n
12
(14.6) 1<n≡0,3 (mod 4)
1
= − + q 3 /3 + q 4 /2 + q 7 + q 8 + q 11 + 4q 12 /3 + · · · .
12

If f (z) = n≥n0 A(n)q n ∈ M!1 (4), then define Ψ(f (z)) by
2
∞

Ψ(f (z)) := q −h
2
(14.7) (1 − q n )A(n ) ,
n=1
"
where h is the constant term of f (z)H(z). Borcherds [42] proved the fol-
lowing theorem.
Theorem 14.1. The map Ψ satisfies
Ψ : M!1 (4) −→ MH .
2

Furthermore, it is an isomorphism. The weight of Ψ(f ) is A(0), and the

multiplicity of the zero of Ψ(f ) at a CM point of discriminant D < 0 is

A(Dn2 ).
n>0

Remark 53. The isomorphism Ψ respects the additive structure of

M!1 (4) and the multiplicative structure of MH .
2
 HARMONIC MAASS FORMS AND NUMBER THEORY 429

Example 14.2. If f (z) = 12f0 (z) = 12Θ0 (z), then

∞

f (z) = A(n)q n = 12 + 24q + 24q 4 + 24q 9 + 24q 16 + · · · .
n=0
Therefore, it follows that
∞

Ψ(f (z)) = q (1 − q n )24 .
n=1
Obviously, Ψ(f (z)) is the classical Δ-function. Its weight is A(0) = 12, and
its divisor is supported at the cusp at infinity.
Example 14.3. If f (z) = 3f3 (z), then
∞

f (z) = A(n)q n = 3q −3 − 744q + 80256q 4 − 257985q 5 + 5121792q 8
n=−3
− 12288744q 9 + · · · .
By Theorem 14.1, it follows that Ψ(f (z)) is a weight A(0) = 0 meromorphic
modular form whose divisor is supported at the cusp at infinity and a triple
zero at ω = e2πi/3 . Therefore, it follows that Ψ(f (z)) must be j(z), and this
is confirmed by
Ψ(f (z)) = q −1 (1 − q)−744 (1 − q 2 )80256 (1 − q 3 )−12288744 · · ·
= q −1 + 744 + 196884q + 21493760q 3 + · · · .
Example 14.4. If f (z) = 4f0 (z) + f3 (z), then
∞

f (z) = A(n)q n = q −3 + 4 − 240q + 26760q 4
n=−3
− 85995q 5 + · · · − 4096240q 9 + · · · .
Theorem 14.1 implies that Ψ(f (z)) is a holomorphic modular form of weight
4 with leading coefficient 1. Therefore, it must be that Ψ(f (z)) = E4 (z).
This explains (14.2).
14.2. Harmonic Maass forms and generalized Borcherds prod-
ucts. Motivated by questions related to derivatives of modular L-functions,
the author and Bruinier [67] derived Borcherds-type products using har-
monic Maass forms. These results are quite technical, and a thorough treat-
ment would occupy more space than is warranted in this expository paper.
For brevity, we will be content with a brief indication of the flavor of these
results, followed by two illustrative examples.
These results are phrased in terms of vector-valued weight 1/2 harmonic
Maass forms. Let (V, Q) be a non-degenerate rational quadratic space of
signature (b+ , b− ), and let L ⊂ V be an even lattice with dual L . The dis-
criminant group L /L, together with the Q/Z-valued quadratic form induced
by Q, is called the discriminant form of the lattice L.
430 K. ONO

As usual, we let Mp2 (R) denote the metaplectic two-fold cover

 of SL2 (R).
The elements of this group are pairs (M, φ(τ )), where M = ac db ∈ SL2 (R)
and φ : H → C is a holomorphic function with φ(τ )2 = cτ + d. The group
law is defined by

(M, φ(τ ))(M  , φ (τ )) = (M M  , φ(M  τ )φ (τ )).

We denote the integral metaplectic group, the inverse image of Γ := SL2 (Z)
under the covering map, by Γ " := Mp2 (Z). It is well known that Γ " is gen-
1 1
 0 −1  √ 
erated by T := (( 0 1 ) , 1), and S := 1 0 , τ . One has the relations
  
S 2 = (ST )3 = Z, where Z := −1 0
0 −1 , i is the standard generator of the
" We let Γ
center of Γ. " ∞ :=
T  ⊂ Γ.
"
We now recall the Weil representation associated with the discriminant
form L /L (for example, see [43], [62]). It is a representation of Γ " on the
group algebra C[L /L]. We denote the standard basis elements of C[L /L]


by eh , h ∈ L /L, and write

·, · for the standard scalar product (antilinear
in the second entry) such that
eh , eh  = δh,h . The Weil representation ρL
associated with the discriminant form L /L is the unitary representation of
" on C[L /L] defined by
Γ

(14.8) ρL (T )(eh ) := e(h2 /2)eh ,

e((b− − b+ )/8) 
(14.9) ρL (S)(eh ) :=  e(−(h, h ))eh .
|L /L| h ∈L /L

If f : H → C[L /L] is a function, we write f = λ∈L /L fh eh for its
decomposition in components with respect to the standard basis of C[L /L].
Let k ∈ 12 Z, and let Mk,ρ
!
L
denote the space of C[L /L]-valued weakly holo-
morphic modular forms of weight k and type ρL for the group Γ. "
Now assume that k ≤ 1. A twice continuously differentiable function
f : H → C[L /L] is called a vector-valued harmonic Maass form of weight k
with respect to Γ " and ρL if it satisfies:
"
(i) f (M τ ) = φ(τ )2k ρL (M, φ)f (τ ) for all (M, φ) ∈ Γ;
(ii) Δk f = 0;
(iii) We have that

f (τ ) = Pf (τ ) + O(e−εv ),

as v → ∞, for some Fourier polynomial

 
Pf (τ ) = c+ (n, h)e(nτ )eh
h∈L /L n∈Z+Q(h)
−∞n≤0

and some ε > 0.

We write Hk,ρL for the vector space of such harmonic Maass forms.
 HARMONIC MAASS FORMS AND NUMBER THEORY 431

Generalizing Lemma 7.2, where the modular variable is now τ = u + iv,

we have that any f ∈ Hk,ρL decomposes as f = f + + f − , where
 
(14.10a) f + (τ ) = c+ (n, h)e(nτ )eh ,
h∈L /L n∈Q
n
−∞
 
(14.10b) f − (τ ) = c− (n, h)W (2πnv)e(nτ )eh ,
h∈L /L n∈Q
n<0
∞
and W (x) = Wk (x) := −2x e−t t−k dt = Γ(1 − k, 2|x|) for x < 0.
In [67], the theory of theta liftings is applied to such harmonic Maass
forms to obtain Borcherds products with twisted Heegner divisors on X0 (N ).
To this end, let N be a positive integer. We consider the rational quadratic
space

(14.11) V := {X ∈ Mat2 (Q) : tr(X) = 0}

with the quadratic form Q(X) := −N det(X). We let L be the lattice

  
b −a/N
(14.12) L := : a, b, c ∈ Z .
c −b

Then the dual lattice is given by

  
 b/2N −a/N
(14.13) L := : a, b, c ∈ Z .
c −b/2N

We identify L /L with Z/2N Z. Here the quadratic form on L /L is identified

with the quadratic form x → x2 on Z/2N Z.
If D ∈ Z, let LD be the set of vectors X ∈ L with Q(X) = D/4N .
Notice that LD is empty unless D is a square modulo 4N . For r ∈ L /L
with r2 ≡ D (mod 4N ) we define

LD,r := {X ∈ L : Q(X) = D/4N and X ≡ r (mod L)}.

 
b/2N −a/N
If X = c −b/2N
∈ LD,r , then the matrix
   
a b/2 0 N
(14.14) ψ(X) := =X
b/2 N c −N 0

defines an integral binary quadratic form of discriminant D = b2 − 4N ac =

4N Q(X) with b ≡ r (mod 2N ). We have that Γ0 (N ) acts on LD,r , and the
number of orbits of LD,r is finite if D = 0.
Let Δ ∈ Z be a fundamental discriminant and r ∈ Z such that Δ ≡ r2
(mod 4N ). Following [107], we define a generalized genus character for
432 K. ONO
 
b/2N −a/N
λ= c −b/2N
∈ L by putting
⎧ Δ 
⎪ n , if Δ | b − 4N ac and
2
⎪
⎪
⎨ (b2 − 4N ac)/Δis a square modulo 4N
χΔ (λ) = χΔ ([a, b, N c]) :=
⎪
⎪ and gcd(a, b, c, Δ) = 1,
⎪
⎩
0, otherwise.
Here [a, b, N c] is the integral binary quadratic form corresponding to λ,
and n is any integer prime to Δ represented by one of the quadratic forms
[N1 a, b, N2 c] with N1 N2 = N and N1 , N2 > 0.
The cusps of Γ0 (N ) correspond to Γ0 (N )-classes of primitive isotropic
vectors in L. Here we let ,  ∈ L be the isotropic vectors
   
0 1/N  0 0
= ,  = .
0 0 1 0
The 1-dimensional lattice
K = L ∩ ⊥ ∩ ⊥
is positive definite. We find that L splits into
(14.15) L = K ⊕ Z ⊕ Z,
1 0 
and K  /K ∼= L /L. Then we have K = Z  0 −1 . For λ ∈ K ⊗ R, we write
1 0
λ > 0 if λ is a positive multiple of 0 −1 .
We now define twisted Heegner divisors on the modular curve X0 (N ).
Let Δ be a fundamental discriminant and let r ∈ Z such that Δ ≡ r2
(mod 4N ). For any vector λ ∈ L of negative norm, the orthogonal comple-
ment λ⊥ ⊂ V (R) defines a point H(λ) in Gr(V ) ∼ = H. For h ∈ L /L and
a negative rational number m ∈ Z + sgn(Δ)Q(h), we consider the twisted
Heegner divisor
 χΔ (λ)
(14.16) HΔ,r (m, h) := H(λ) ∈ Div(X0 (N ))Q ,
w(λ)
λ∈LdΔ,hr /Γ0 (N )

where d := 4N m sgn(Δ) ∈ Z. Note that d is a discriminant which is con-

gruent to a square modulo 4N and which has the opposite sign as Δ. Here
w(λ) is the order of the stabilizer of λ in Γ0 (N ). We also require the degree
zero divisor
(14.17) yΔ,r (m, h) := HΔ,r (m, h) − deg(HΔ,r (m, h)) · ∞.
We have yΔ,r (f ) = HΔ,r (f ) when Δ = 1. By the theory
√ of complex multi-
plication, the divisor HΔ,r (m, h) is defined over Q( Δ).
Recall that ρ̃L = ρL for Δ > 0, and ρ̃L = ρ̄L for Δ < 0. Suppose that
f ∈ H1/2,ρ̃L is a harmonic weak Maass form of weight 1/2 with representation
ρ̃L . We denote the coefficients of f = f + + f − by c± (m, h) as in (14.10).
 HARMONIC MAASS FORMS AND NUMBER THEORY 433

Using the Fourier coefficients of the principal part of f , we finally define the
twisted Heegner divisor associated to f by
 
(14.18) HΔ,r (f ) := c+ (m, h)HΔ,r (m, h) ∈ Div(X0 (N ))R ,
h∈L /L m<0
 
(14.19) yΔ,r (f ) := c+ (m, h)yΔ,r (m, h) ∈ Div(X0 (N ))R .
h∈L /L m<0

Notice that yΔ,r (f ) = HΔ,r (f ) when Δ = 1.

Theorem 14.5. Assume the notation and hypotheses above. Let f ∈
H1/2,ρ̃L be a harmonic Maass form with real coefficients c+ (m, h) for all
m ∈ Q and h ∈ L /L. Moreover, assume that c+ (n, h) ∈ Z for all n ≤ 0.
Then infinite product
  Δ + 2
ΨΔ,r (z, f ) = e((ρf, , z)) [1 − e((λ, z) + b/Δ)]( b )c (|Δ|λ /2,rλ) ,
λ∈K  b (Δ)
λ>0

where ρf, is a certain Weyl vector (see (4.8) of [67]), converges for y suf-
ficiently large and has a meromorphic continuation to all of H with the
following properties.
(i) It is a meromorphic modular form for Γ0 (N ) with a unitary char-
acter σ which may have infinite order.
(ii) The weight of ΨΔ,r (z, f ) is c+ (0, 0) when Δ = 1, and is 0 when
Δ = 1.
(iii) The divisor of ΨΔ,r (z, f ) on X0 (N ) is given by HΔ,r (f ) + CΔ,r (f ),
where CΔ,r (f ) is supported at cusps.
(iv) The “regularized theta integral” of f defined in (5.5) of [67], satis-
fies
⎧ + 
⎨−c (0, 0)(log(4πN ) + Γ +(1))
⎪
ΦΔ,r (z, f ) = −4 log |ΨΔ,r (z, f )y c (0,0)/2 |, if Δ = 1,
⎪
⎩ √
2 Δc(0, 0)L(1, χΔ ) − 4 log |ΨΔ,r (z, f )|, if Δ = 1.
Unlike the results in [43], such generalized Borcherds products typically
transform with a multiplier system of infinite order under Γ0 (N ). The fol-
lowing criterion is obtained which determines when the multiplier system
has finite order.
Theorem 14.6. Suppose that Δ = 1. Let f ∈ H1/2,ρ̃L be a harmonic
Maass form with real coefficients c+ (m, h) for all m ∈ Q and h ∈ L /L.
Moreover, assume that c+ (n, h) ∈ Z for all n ≤ 0. The following are equiv-
alent.
(i) The character σ of the function ΨΔ,r (z, f ) defined in Theorem 14.5
is of finite order.
(ii) The coefficients c+ (|Δ|λ2 /2, rλ) are rational for all λ ∈ K  .
434 K. ONO

Remark 54. Theorem 14.6 is particularly simple to implement for

weakly holomorphic modular forms f (for example, see Lemma 6.5 of [67]).
In this case the rationality of all the coefficients is implied by the rational-
ity of the principal part, and so Theorem 14.6 implies that the σ for the
corresponding generalized Borcherds product has finite order.

Example 14.7 (Twisted modular polynomials). Here we use Theorems

14.5 and 14.6 to deduce the infinite product expansion of twisted modular
polynomials found by Zagier (see Section 7 of [212]).
The weight 1/2 harmonic Maass forms in question are the same weakly
holomorphc modular forms fd (z) which appear in (14.5). Theorems 14.5
and 14.6 then give a meromorphic modular form ΨΔ (z, fd ) := ΨΔ,r (z, fd ) of
weight 0 for the group Γ = SL2 (Z) whose divisor on X(1) is given by
 χΔ (λ)
HΔ (d) := HΔ,r (d/4, d/2) = · H(λ).
w(λ)
λ∈LΔd /Γ

Classically, the quotient LΔd /Γ corresponds to the Γ-classes of integral

binary quadratic forms of discriminant Δd. Moreover, for sufficiently large
Im(z), we have the product expansion
∞ 
 Δ 2
(14.20) ΨΔ (z, fd ) = [1 − e(nz + b/Δ)]( b )cd (Δn ) .
n=1 b (Δ)

From these properties it follows that

  χ (λ)
(14.21) ΨΔ (z, fd ) = j(z) − j(H(λ)) Δ .
λ∈LΔd /Γ

As an example, let Δ := 5 and d := −3. There are two classes of binary

quadratic forms of discriminant −15, represented
√
by [1,
√
1, 4] and [2, 1, 2], and
−1+ −15 −1+ −15
their corresponding CM points are and . It is well known
2 4 √
that the singular moduli of j(τ ) of these points are − 191025 − 85995 5, and
√ 2 2
− 2 + 2
191025 85995
5. The function f−3 has the Fourier expansion

f−3 = q −3 − 248 q + 26752 q 4 − 85995 q 5 + 1707264 q 8 − 4096248 q 9 + . . . .

Multiplying out the product over b in (14.20), we obtain the infinite product
expansion
√ ∞
 √ c−3 (5n2 )
j(z) + 191025
+ 85995  1− 5 n
+ q 2n
2 2 √5 1+ 2√ q
Ψ5 (z, f−3 ) = = .
j(z) + 191025
− 85995
5 1+ 1+ 5 n
+ q 2n
2 2 n=1 2 q

Example 14.8 (A generalized Borcherds product for Ramanujan’s ω(q)).

Here we give an example of a generalized Borcherds product arising from
 HARMONIC MAASS FORMS AND NUMBER THEORY 435

Ramanujan’s mock theta function

∞ 2
q 2n +2n 1 q4
ω(q) := = +
(q; q 2 )2n+1 (1 − q)2 (1 − q)2 (1 − q 3 )2
n=0
q 12
(14.22) + + ··· .
(1 − q)2 (1 − q 3 )2 (1 − q 5 )2
Zwegers’s work (see Theorem 6.1) completes this q-series in a way which
allows him to produce a vector valued harmonic Maass form.
Define integers a(n) by
 
− 2q 1/3 ω(q 1/2 ) + ω(−q 1/2 )

=: a(n)q n = −4 q 1/3 − 12 q 4/3 − 24 q 7/3 − 40 q 10/3 − . . . .
n∈Z+1/3

If we let √
1 + −2X − X 2
P (X) := √ ,
1 − −2X − X 2
then Theorems 14.5 and 14.6 imply that
∞
 n 2
(14.23) Ψ(z) = P (q n )( 3 )a(n /3)
n=1

is a modular function on Γ0 (6).

Here we work out an expression for Ψ(z). We use the Hauptmodul for
Γ∗0 (6), the extension of Γ0 (6) by all Atkin-Lehner involutions, which is
   4
∗ η(z)η(2z) 4 4 η(3z)η(6z)
j6 (z) = +4+3
η(3z)η(6z) η(z)η(2z)
= q −1 + 79 q + 352 q 2 + 1431 q 3 + . . . .
Let α1 and α2 be the Heegner points
√ √
−2 + −2 2 + −2
α1 := and α2 := .
6 6
We have j6∗ (α1 ) = j6∗ (α2 ) = −10. Hence j6∗ (z) + 10 is a rational function
on X0 (6) whose divisor consists of the 4 cusps with multiplicity −1 and
the points α1 , α2 with multiplicity 2. The unique normalized cusp form of
weight 4 for Γ∗0 (6) is
δ(z) := η(z)2 η(2z)2 η(3z)2 η(6z)2
= q − 2 q 2 − 3 q 3 + 4 q 4 + 6 q 5 + 6 q 6 − 16 q 7 − 8 q 8 + . . . .
Using these functions, we find that
φ(z) := Ψ(z) · (j6∗ (z) + 10)δ(z)
436 K. ONO

is a holomorphic modular form of weight 4 for Γ0 (6) with divisor 4(α1 ).

Using the classical Eisenstein series, it turns out that
√ √ √
450φ(z) = (3360 − 1920 −2)δ(z) + (1 − 7 −2)E4 (z) + (4 − 28 −2)E4 (2z)
√ √
+ (89 + 7 −2)E4 (3z) + (356 + 28 −2)E4 (6z).
Putting this all together, (14.23) becomes
∞  √ ( n )a(n2 /3)
 1 + −2q n − q 2n 3 φ(z)
√ = ∗
n=1
1 − −2q − qn 2n (j6 (z) + 10)δ(z)
√ √ √
= 1 − 8 −2q − (64 − 24 −2)q 2 + (384 + 168 −2)q 3
√
+ (64 − 1768 −2)q 4 + · · · .

15. Derivatives and values of modular L-functions

Harmonic Maass forms are very useful for investigating the central values
and derivatives of weight 2 modular L-functions. Recent works by the au-
thor, Bruinier, and Yang involve the interplay between generalized Borcherds
products, harmonic Maass forms, and earlier celebrated theorems of Gross
and Zagier, Kohnen and Zagier, and Waldspurger.
To explain these results, we first recall some standard definitions. Let
∞
f (z) = a(n)q n ∈ S2k
new
(M )
n=1
be a newform of even integral weight 2k on Γ0 (M ). For Re(s)  0, let
∞
 a(n)
(15.1) L(f, s) =
ns
n=1
be its L-function. Let
 D denote the fundamental discriminant of a quadratic
√
field, and let χD = D • denote the Kronecker character for the field Q( D).
The D-quadratic twist of f (z), denoted fD (z), is given by
∞

fD (z) = χD (n)a(n)q n ,
n=1
and for Re(s)  0 its L-function is given by
∞
 χD (n)a(n)
L(f, χD , s) = .
ns
n=1
These L-functions have analytic continuations to C and satisfy well
known functional equations. If we define Λ(f, s) by
Λ(f, s) := (2π)−s Γ(s)M s/2 L(f, s),
then there is an  ∈ {±1} for which
Λ(f, s) =  · Λ(f, 2k − s).
 HARMONIC MAASS FORMS AND NUMBER THEORY 437

Furthermore, if D is a fundamental discriminant which is coprime to M ,

then
Λ(fD , s) =  · χD (−M )Λ(fD , 2k − s).
Remark 55. The number  above is referred to as the sign of the func-
tional equation of L(f, s). If f (z) ∈ S2k
new (M ) is a newform, then

 = (−1)k λM ,
where λM is the eigenvalue of f (z) under the Fricke involution W (M ). In
other words, we have
(f |2k W (M )) (z) = λM f (z).
The values L(f, χD , k) are the central critical values of the L-functions
L(f, χD , s).
Our motivating problem is to describe the behavior of the values
L(f, χD , k), as D varies. Notice that if χD (−M ) = −1, then L(f, χD , k) =
0. Therefore at least half of these L(f, χD , k) are trivially zero. As we shall
see, the “nontrivial zeros” are much more mysterious.
Much of the interest in central critical values of modular L-functions
follows from their connection to the Birch and Swinnerton-Dyer. Here we
briefly recall important facts and results.
Suppose that E/Q is an elliptic curve of conductor N (E) (see [194] for
background on elliptic curves), and let, for Re(s)  0,
∞
aE (n)
(15.2) L(E, s) =
ns
n=1
be its Hasse-Weil L-function. In particular, if p  N (E) is prime, then we
have
NE (p) = p + 1 − aE (p),
where NE (p) denotes the number of points on the reduction of E modulo p.
By the work of Breuil, Conrad, Diamond, Taylor, and Wiles [46, 81,
198, 209], we have the following deep result which confirmed the Shimura-
Taniyama Conjecture.
Theorem 15.1. If E/Q is an elliptic curve of conductor N (E), then
there is a newform fE (z) ∈ S2new (N (E)) for which
L(E, s) = L(fE , s).
For elliptic curves E/Q, Theorem 15.1 implies that L(E, s) has an ana-
lytic continuation to C, and so the analytic behavior of L(E, s) at s = 1 is
well defined. The Birch and Swinnerton-Dyer Conjecture (see Appendix C
of [194]) gives arithmetic significance to this behavior.
Conjecture. If E/Q is an elliptic curve, and if rk(E) is the Mordell-
Weil rank of E over Q, then
ords=1 (L(E, s)) = rk(E).
438 K. ONO

Remark 56. This is the “weak” form of the Birch and Swinnerton-Dyer
Conjecture. The strong conjecture gives an arithmetic formula for
 
lim (s − 1)−rk(E) · L(E, s)
s→1
in terms of arithmetic invariants associated to E.
Although this conjecture remains open, we have the following strong
result which follows from the work of Kolyvagin, Gross, and Zagier [108,
138].18
Theorem 15.2. If E/Q is an elliptic curve for which ords=1 (L(E, s)) ∈
{0, 1}, then
ords=1 (L(E, s)) = rk(E).
Now we consider quadratic twists of elliptic curves and modular forms.
Let E/Q be an elliptic curve given by the Weierstrass equation
(15.3) E : y 2 = x3 + ax2 + bx + c,
where a, b and c are integers. If D is a square-free integer, then let E(D)
denote the D−quadratic twist of E given by
(15.4) E(D) : y 2 = x3 + aDx2 + bD2 x + cD3 .
√
The curves E and E(D) are isomorphic over Q( D).
Suppose that E/Q is an elliptic curve, and that
∞

fE (z) = aE (n)q n ∈ S2new (N (E))
n=1
is the weight two newform associated to E by Theorem 15.1. Similarly, if D
is a fundamental discriminant, then let
∞

fE(D) (z) = aE(D) (n)q n
n=1
be the newform associated to E(D). If p  DN (E) is prime, then it is simple
to check that  
D
aE(D) (p) = aE (p).
p
Consequently, the newform fE(D) (z) is the unique newform whose Hecke
eigenvalues, for primes p  DN (E), equal those of the quadratic twist fE
by χD .
In view of the Birch and Swinnerton-Dyer Conjecture, and Theorem 15.2,
we are compelled to study central values and derivatives of weight 2 modu-
lar L-functions. In this direction we have the celebrated works of Kohnen,
Zagier and Waldspurger, and also the work of Gross and Zagier [108].
18Coates and Wiles [79] earlier proved that L(E, 1) = 0 =⇒ rk(E) = 0 for elliptic
curves E/Q with complex multiplication.
 HARMONIC MAASS FORMS AND NUMBER THEORY 439

It turns out that the Fourier coefficients of half-integral weight cusp

forms often interpolate the “square-roots” of the central critical values of
the L-functions of the quadratic twists of even weight newforms.
Here we make this precise. First suppose that N is odd and square-free,
and suppose further that
∞

(15.5) g(z) = b(n)q n ∈ Sk+
new
1 (4N )
2
n=1

is a Kohnen newform.19 There is a unique newform, say f (z) ∈ S2k new (N ),

associated to g(z) under Shimura’s correspondence. The coefficients of

g(z) determine the central critical values of many of the quadratic twists
L(f, χD , s). Let ν(N ) denote the number of distinct prime divisors of N ,
and let
f, f  (resp.
g, g) denote the Petersson inner product on S2k (N )
(resp. Sk+ 1 (4N )). Generalizing a result in [137], Kohnen proved the fol-
2
lowing important theorem in [136].
Theorem 15.3. Assume the notation in the preceding discussion. If
 | N is prime, then let λ ∈ {±1} be the eigenvalue of the Atkin-Lehner
involution
(f |2k W (Q )) (z) = λ f (z).
 
If (−1)k D > 0 and D has the property that D = λ for each prime  | N ,
then

f, f  · π k
L(f, χD , k) = · |b(|D|)|2 .
k− 12
2 ν(N ) (k − 1)!|D|
g, g
For all other fundamental discriminants D with (−1)k D > 0 we have
b(|D|) = 0.
Example 15.4. If f (z) = Δ(z) ∈ S12 , then it turns out that the associ-
new (4) is
ated Kohnen newform in S13/2
∞
 E4 (4z) · D(Θ0 (z)) D(E4 (4z)) · Θ0 (z)
g(z) = b(n)q n = −
2 16
n=1
= q − 56q 4 + 120q 5 − . . . .
d
Again, here D denotes the differential operator D := q dq . For positive
fundamental discriminants D, we then have that

Δ, Δ · π 6
L(Δ, χD , 6) = 11 · b(D)2 .
120D
g, g
2

19These are the half-integral weight cuspidal Hecke eigenforms in Kohnen’s paper
[136].
440 K. ONO

Waldspurger proved a similar theorem which holds in greater generality.

For every fundamental discriminant D, define D0 by

|D| if D is odd,
(15.6) D0 :=
|D|/4 if D is even.
The following is a convenient reformulation of Waldspurger’s theorem [205].

Theorem 15.5. If f (z) = ∞ n=1 a(n)q ∈ S2k (M ) is an even weight
n new

newform and δ ∈ {±1} is the sign of the functional equation of L(f, s), then
there is a positive integer N with M | N , a Dirichlet character χ modulo
4N , a nonzero complex number Ωf , and a nonzero half-integral weight Hecke
eigenform
∞

gf (z) = bf (n)q n ∈ Sk+ 1 (4N, χ)
2
n=1
with the property that there are arithmetic progressions of fundamental dis-
criminants D coprime to 4N for which δD > 0 and
bf (D0 )2
L(f, χD , k) = D · · Ωf ,
k− 12
D0
where D is algebraic. For all other D with δD > 0, we have bf (D0 ) = 0.
Moreover, the coefficients a(n), bf (n) and the values of χ are in OK , the
ring of integers of some fixed number field K.
15.1. Extension of the Kohnen-Waldspurger theorem. The
author and Bruinier [67] have generalized this theorem of Waldspurger and
Kohnen to prove that the Fourier coefficients of weight 1/2 harmonic Maass
forms encode the vanishing and nonvanishing of both the central values and
derivatives of quadratic twists of weight 2 modular L-functions.
Here we describe a special case of the main result of [67].
Theorem 15.6. Suppose that
∞
G(z) = BG (n)q n ∈ S2 (p)
n=1
is a weight 2 newform with prime level p. In addition, suppose that the sign
of the functional equation of L(G, s) is (G) = −1. Then there is a weight
1/2 harmonic Maass form f (z) on Γ0 (4p), say
 
f (z) = c+ n
g (n)q + c− n
g (n)Γ(1/2; 4π|n|y)q ,
n
−∞ n<0
which satisfies the following:
 
Δ
(1) If Δ < 0 is a fundamental discriminant for which p = 1, then

L(G, χΔ , 1) = αG · |Δ| · c− 2
g (Δ) ,
where αG is an explicit non-zero constant.
 HARMONIC MAASS FORMS AND NUMBER THEORY 441
 
Δ
(2) If Δ > 0 is a fundamental discriminant for which p = 1, then
L (G, χΔ , 1) = 0 if and only if c+
g (Δ) is algebraic.

Remark 57. Theorem 15.6 is a special case of the general result which
holds for all levels, and any arbitrary sign. Notice that the result concerns
all weight 2 newforms, not just those which correspond to modular elliptic
curves.
Remark 58. The Maass form f (z) in Theorem 15.6 is unique up to the
addition of a weight 1/2 weakly holomorphic modular form with coefficients
in FG , the number field obtained by adjoining the coefficients of G(z) to Q.
In view of this ambiguity, it would be very interesting to pin down a choice
of f (z) which in turn gives a precise formula relating L (G, χΔ , 1) to c+
g (Δ)
in Theorem 15.6 (2).
Theorem 15.6 is obtained by combining the Gross-Zagier formula with
Borcherds products arising from harmonic Maass forms and general tran-
scendence theorems for canonical differentials of the third kind on modular
curves. Here we give a brief sketch of the proof of this theorem.
Sketch of the Proof of Theorem 15.6. By Kohnen’s theory,
there is a half-integral weight newform
∞
(15.7) g(z) = bg (n)q n ∈ S +
3 (4p),
2
n=1
unique up to a multiplicative constant, which lifts to G under the Shimura
correspondence. We can (and do) choose g so that its coefficients are in FG ,
the totally real number field obtained by adjoining the Fourier coefficients
of G to Q. We prove that there is a weight 1/2 harmonic Maass form on
Γ0 (4p) in the plus space, say
 
(15.8) fg (z) = c+ n
g (n)q + c− n
g (n)Γ(1/2, 4π|n|y)q ,
n
−∞ n<0
whose principal part Pfg has coefficients in FG , which also enjoys the prop-
erty that ξ 1 (fg ) = g−2 g, where g denotes the usual Petersson norm.
2
By Lemma 7.4, if n > 0, then
√
(15.9) bg (n) = −4 πng2 · c−g (−n).
Theorem 15.6 (1) now follows from Theorem 15.3.
The proof of Theorem 15.6 (2) is more difficult, and it involves a detailed
study of Heegner divisors. We establish that the algebraicity of the coeffi-
cients c+
g (Δ) is dictated by the vanishing of certain twisted Heegner divisors
in the Jacobian of X0 (p), which when combined with the work of Gross and
Zagier [108] and Scholl and Waldschmidt, then implies Theorem 15.6 (2).
To make this precise, let d < 0 and Δ > 0 be fundamental discriminants
which are both squares modulo p. Let Qd,p be the set of discriminant d =
b2 −4ac integral binary quadratic forms aX 2 +bXY +cY 2 with the property
442 K. ONO

that p | a. For these pairs of discriminants, we define the twisted Heegner

divisor HΔ (d) by
 τQ
(15.10) HΔ (d) := χΔ (Q) · ,
wQ
Q∈QΔd,p /Γ0 (p)

where χΔ denotes the generalized genus character corresponding to the

decomposition Δ · d as in [107], τQ is the unique root of Q(x, 1) in H as
before, and wQ denotes the order of the stabilizer
√ of Q in Γ0 (p). Then
HΔ (d) is a divisor on X0 (p) defined over Q( Δ). We use these twisted
Heegner divisors to define the degree 0 divisor

(15.11) yΔ (d) := HΔ (d) − deg(HΔ (d)) · ∞.

Finally, we associate a divisor to fg by letting


g (n)yΔ (n) ∈ Div (X0 (p)) ⊗ FG .
c+ 0
(15.12) yΔ (fg ) :=
n<0

Let J be the Jacobian of X0 (p), and let J(F ) denote the points of J
over a number field F . The Hecke algebra acts on J(F ) ⊗ C, which by the
Mordell-Weil Theorem is a finite dimensional √ vector space. We show that
the point corresponding to yΔ (fg ) in J(Q( Δ)) ⊗ C is in its G-isotypical
component. Moreover, we show that the following are equivalent:
√
(i) The Heegner divisor yΔ (fg ) vanishes in J(Q( Δ)) ⊗ C.
(ii) The coefficient c+
g (Δ) is algebraic.
(iii) The coefficient c+
g (Δ) is contained in FG .
To obtain these results, we explicitly construct modular functions with pre-
scribed Heegner divisors using results related to Theorems 14.5 and 14.6. By
work of Scholl and Waldschmidt, the vanishing of the points corresponding
to these divisors is equivalent to the algebraicity of the Fourier expansions
of these modular functions. Using the Hecke algebra and the explicit formu-
las provided by the generalized Borcherds product, we then find that this
algebraicity is dictated by single coefficients of the form c+ g (Δ).
We then obtain the following generalization of the well known Gross-
Kohnen-Zagier theorem [107]. Namely, we show that
 √
yΔG
(−n)q n = g(τ ) ⊗ yΔ (fg ) ∈ S +
3 (4p) ⊗ J(Q( Δ)),
2
n>0

where yΔG (−n) denotes the projection of y (−n) onto its G-isotypical compo-
Δ
nent. This result, when combined with the Gross-Zagier theorem [108],
√ gives
the conclusion that the Heegner divisor yΔ (fg ) vanishes in J(Q( Δ)) ⊗ C
if and only if L (G, χΔ , 1) = 0, thereby proving Theorem 15.6 (2).

 HARMONIC MAASS FORMS AND NUMBER THEORY 443

15.2. Recent work of Bruinier and Yang. Building on the results

of the last subsection, it is natural to seek exact formulas for derivatives of L-
functions using techniques arising from the theory of harmonic Maass forms.
Theorem 15.6 makes the connection between derivatives and coefficients of
harmonic Maass forms, but it does not give exact formulas. This problem
remains open. Nevertheless, the question of fundamental interest really
should be: Can one directly obtain exact arithmetic formulas (perhaps in
terms of heights) for certain derivatives of L-functions?
To this end, Bruinier and Yang [70] have recently proven some striking
theorems about Faltings heights of CM cycles and derivatives of L-functions.
In some important cases they are able to explicitly evaluate derivatives of
certain Rankin-type L-functions in terms of Faltings heights. Their work de-
pends critically on Borcherds lifts and the theory of harmonic Maass forms.
It is impossible to survey their work in any detail here. Instead, we give
the flavor of their exciting program. Loosely speaking, they use the principal
parts of suitable harmonic Maass forms f to construct Heegner divisors on
certain Shimura varieties. This depends on earlier work of Kudla, recent
work on Greens functions, and Borcherds lifts. For the cusp forms ξ(f ),
they investigate a certain “Rankin-type” L-function, and they conjecture
an exact formula relating its derivative at s = 0 to the Faltings height
pairing of a Heegner divisor with a CM cycle. The conjectured formula is
of the form

(15.13)
Ẑ(f ), Z(U )F al = κ · L (ξ(f ), U, 0),

where κ is explicit.
They have obtained several deep results in the direction of this conjec-
ture. Here we highlight one implication of their work in the classical setting
of the original formula of Gross and Zagier. Suppose that G is a normalized
newform of weight 2 for Γ0 (N ) whose Hecke L-function L(G, s) satisfies an
odd functional equation. Then there is a weight 3/2 cusp form g, which cor-
responds to G under the Shimura correspondence, and it turns out that the
“Rankin-type” L-function is proportional to L(G, s + 1). Their conjecture
therefore then implies a Faltings height pairing formula for L (G, 1).
As in the work of the author and Bruinier [67] (see Theorem 5.1), there
is a weight 1/2 harmonic Maass form f with vanishing constant term such
that ξ(f ) = g−2 g, with the additional property that the coefficients of its
principal part are in the number field generated by the eigenvalues of G. In
this case the Heegner divisor Z(f ) defines an explicit point in the Jacobian of
X0 (N ), which lies in the G isotypical component (see the sketch of the proof
of Theorem 15.6). In this setting, they prove the following comprehensive
theorem.
444 K. ONO

Theorem 15.7 (Bruinier and Yang [70]). Assuming the notation and
hypotheses above, we have that the Neron-Tate height of Z(f ) is given by
√
2 N 

Z(f ), Z(f )N T = L G, 1).
πg2
This beautiful theorem directly gives an arithmetic formula for the cen-
tral derivative of weight 2 modular L-functions with odd sign. It is natural
to ask how Theorem 15.7 relates to the classical Gross-Zagier formula. The
short answer is that it implies it.
To see this, suppose that E is an elliptic curve over Q. Assume that
its L-function L(E, s) has an odd functional equation so that the central
critical value L(E, 1) vanishes. Let N = N (E) be the conductor of E, and
let X0 (N ) be the moduli space of cyclic isogenies of degree N of generalized
elliptic curves. Let K be an imaginary quadratic field such that N is the
norm of an integral ideal of K, and write D for the discriminant of K. We
may consider the divisor Z(D) on X0 (N ) given by elliptic curves with com-
plex multiplication by the maximal order of K. By the theory of complex
multiplication, this divisor is defined over K, and its degree h is given by
the class number of K. Hence the divisor y(D) = trK/Q (Z(D) − h · (∞))
has degree zero and is defined over Q. By the modularity of elliptic curves,
we obtain a rational point y E (D) on E using a modular parametrization
X0 (N ) → E. Arguing a little further (so as to produce L(E, χD , 1)),
one can show that Theorem 15.7 then implies the Gross-Zagier theorem as
stated below.
Theorem 15.8 (Gross-Zagier [108]). Assume the notation and hypothe-
ses above. Then the canonical height of y E (D), denoted
y E (D), y E (D)N T ,
satisfies


y E (D), y E (D)N T = C |D|L (E, 1)L(E, χD , 1).
Here C is an explicit non-zero constant which is independent of K, and χD
is the Kronecker character for K/Q.
To place Theorem 15.7 in proper context, we stress that the work of
Gross and Zagier has inspired an enormous body of further research on
height pairings of algebraic cycles on Shimura varieties. For instance, Zhang
considered heights of Heegner type cycles on Kuga-Sato fiber varieties over
modular curves in [214], and the heights of Heegner points on compact
Shimura curves over totally real fields in [215]. Gross and Keating discov-
ered a connection between arithmetic intersection numbers of Hecke corre-
spondences on the product of two copies of the modular curve X(1) over Z
and the coefficients of the derivative of the Siegel-Eisenstein series of genus
three and weight two [106]. This has inspired the extensive program of
Kudla, Rapoport and Yang which relates Arakelov intersection numbers on
Shimura varieties of orthogonal type to derivatives of Siegel-Eisenstein series
and modular L-functions (for example, see [140], [141], [142]).
 HARMONIC MAASS FORMS AND NUMBER THEORY 445

In all of these works, the connection between a height pairing and the
derivative of an automorphic L-function arises in an indirect way. The idea
has been to identify the local height pairings in the Fourier coefficients of a
suitable integral kernel function (often given by an Eisenstein series), which
takes an automorphic form φ to the special value of the derivative of an
L-function associated to φ.
This recent work of Bruinier and Yang gives a new approach for obtain-
ing identities between certain height pairings on Shimura varieties of orthog-
onal type and derivatives of automorphic L-functions. As described above,
it is based on the Borcherds lift [44] and its generalization in [62, 63, 67].
Their approach directly gives formulas for the Faltings height pairing of
arithmetic Heegner divisors and CM cycles.

16. Epilogue
Although the mock theta functions are humble in origin, they have
earned a distinguished role in the legend of Ramanujan. Andrews and
Berndt confirm this in their article “Your hit parade: the top ten most
fascinating formulas in Ramanujan’s lost notebook” [22]. In their amusing
informal poll, Ramanujan’s work on Dyson’s ranks20 and the mock theta
functions rank first and second! Based on the mathematics born out of
these works, as described here, it is a safe bet that ranks and mock theta
functions will continue to hold these top spots into the foreseeable future.
They certainly have the author’s vote!
It is appropriate to end this paper with a few words about the title:
“Unearthing the visions of a master: harmonic Maass forms and number
theory”. Although Ramanujan’s last works provided the first examples of
such forms, his untimely death and the enigmatic nature of his writings
resulted in a great mystery. We will never know how he came up with the
mock theta functions. We certainly cannot pretend to know what he fully
intended to do with them. However, it is clear that he understood that the
mock theta functions would go on to play important roles in number theory,
his “visions”. The author likes to think that this paper contains some of
these visions.

Acknowledgements
The author thanks George E. Andrews, Dick Askey, Bruce B. Berndt,
Matt Boylan, Kathrin Bringmann, Jan Bruinier, Amanda Folsom, Pavel
Guerzhoy, Paul Jenkins, Jeremy Lovejoy, Maria Monks, Rob Rhoades,
Heather Swan-Rosenthal, Frank Thorne, Tonghai Yang and Luanlei Zhao
for their many helpful comments. Without this assistance the author would
have been unable to write this paper.
20Although Dyson defined the notion of a partition rank in 1944, it is clear that
Ramanujan understood the notion in 1920 because of certain identities he recorded in the
“Lost Notebook”.
446 K. ONO

References
[1] M. Abramowitz and I. Stegun, Pocketbook of mathematical functions, Verlag Harri
Deutsch (1984).
[2] S. Ahlgren, Distribution of the partition function modulo composite integers M ,
Math. Ann. 318 (2000), no. 4, pages 795–803.
[3] S. Ahlgren and K. Ono, Congruence properties for the partition function, Proc.
Natl. Acad. Sci. USA 98 (2001), no. 23, pages 12882–12884.
[4] S. Ahlgren and K. Ono, Addition and counting: the arithmetic of partitions, Notices
Amer. Math. Soc. 48 (2001), no. 9, pages 978–984.
[5] S. Ahlgren and K. Ono, Congruences and conjectures for the partition function,
q-series with applications to combinatorics, number theory, and physics (Urbana,
IL, 2000), Contemp. Math., 291, Amer. Math. Soc., Providence, RI, 2001, pages
1–10.
[6] P. V. Seshu Aiyar and R. Ramachandra Rao, Srinivasa Ramanujan (1887–1920),
Collected papers of Ramanujan (Ed. G. H. Hardy, et. al.), Cambridge Univ. Press,
Cambrdige, 1927, pages xi–xx.
[7] G. E. Andrews, On the theorems of Watson and Dragonette for Ramanujan’s mock
theta functions, Amer. J. Math. 88 No. 2 (1966), pages 454–490.
[8] G. E. Andrews, On the general Rogers-Ramanujan theorem, Amer. Math. Soc.,
Providence, 1974.
[9] G. E. Andrews, An introduction to Ramanujan’s lost notebook, Amer. Math.
Monthly 86 (1979), pages 89–108.
[10] G. E. Andrews, Hecke modular forms and the Kac-Peterson identities, Trans.
Amer. Math. Soc. 283 (1984), pages 451–458.
[11] G. E. Andrews, Combinatorics and Ramanujan’s “lost” notebook, Surveys in
combinatorics 1985 (Glasgow, 1985), London Math. Soc. Lecture Note Ser., 103,
Cambridge Univ. Press, Cambridge, 1985, pages 1–23.
[12] G. E. Andrews, The fifth and seventh order mock theta functions, Trans. Amer.
Math. Soc. 293 (1986), no. 1, pages 113–134.
[13] G. E. Andrews, q-series: Their development and application in analysis, number
theory, combinatorics physics, and computer algebra, CBMS No. 66, Amer. Math.
Soc., Providence, RI, 1986.
[14] G. E. Andrews, Ramanujan’s fifth order mock theta functions as constant terms,
Ramanujan revisited (Urbana-Champaign, Ill., 1987), Academic Press, Boston,
MA, 1988, pages 47–56.
[15] G. E. Andrews, Mock theta functions, Theta functions—Bowdoin 1987, Part 2
(Brunswick, ME, 1987), pages 283–298, Proc. Sympos. Pure Math., 49, Part 2,
Amer. Math. Soc., Providence, RI, 1989.
[16] G. E. Andrews, The theory of partitions, (Reprint of the 1976 original), Cambridge
Univ. Press, Cambridge, 1998.
[17] G. E. Andrews, Partitions: At the interface of q-series and modular forms, Rankin
Memorial Issues, Ramanujan J. 7 (2003), pages 385–400.
[18] G. E. Andrews, Partitions with short sequences and mock theta functions, Proc.
Natl. Acad. Sci. USA, 102 No. 13 (2005), pages 4666–4671.
[19] G. E. Andrews, Partitions, Durfee symbols, and the Atkin-Garvan moments of
ranks, Invent. Math. 169 (2007), pages 37–73.
[20] G. E. Andrews, R. Askey and R. Roy, Special functions, Cambridge Univ. Press,
Cambridge, 1999.
[21] G. E. Andrews and B. C. Berndt, Ramanujan’s lost notebook. Part I, Springer,
New York, 2005.
 HARMONIC MAASS FORMS AND NUMBER THEORY 447

[22] G. E. Andrews and B. C. Berndt, Your hit parade: the top ten most fascinating
formulas in Ramanujan’s Lost notebook, Not. Amer. Math. Soc. 55 (2008), pages
18–30.
[23] G. E. Andrews, F. Dyson, and D. Hickerson, Partitions and indefinite quadratic
forms, Invent. Math. 91 No. 3 (1988), pages 391–407.
[24] G. E. Andrews and K. Eriksson, Integer partitions, Cambridge Univ. Press,
Cambridge, 2004.
[25] G. E. Andrews and F. Garvan, Dyson’s crank of a partition, Bull. Amer. Math.
Soc. (N.S.) 18 No. 2 (1988), pages 167–171.
[26] G. E. Andrews and F. Garvan, Ramanujan’s “lost” notebook. VI. The mock theta
conjectures, Adv. in Math. 73 (1989), no. 2, pages 242–255.
[27] G. E. Andrews and D. Hickerson, Ramanujan’s “lost” notebook. VII. The sixth
order mock theta functions, Adv. Math. 89 (1991), no. 1, pages 60–105.
[28] A. O. L. Atkin, Proof of a conjecture of Ramanujan, Glasgow Math. J. 8, (1967)
pages 14–32.
[29] A. O. L. Atkin, Multiplicative congruence properties and density problems for p(n),
Proc. London Math. Soc. (3) 18 (1968), pages 563–576.
[30] A. O. L. Atkin and F. Garvan, Relations between the ranks and cranks of partitions,
Ramanujan J. 7 (2003), pages 343–366.
[31] A. O. L. Atkin and H. P. F. Swinnerton-Dyer, Some properties of partitions, Proc.
London Math. Soc. 66 No. 4 (1954), pages 84–106.
[32] B. C. Berndt, Ramanujan’s notebooks. Part I, Springer-Verlag, New York, 1985.
[33] B. C. Berndt, Ramanujan’s notebooks. Part II, Springer-Verlag, New York, 1989.
[34] B. C. Berndt, Ramanujan’s notebooks. Part III, Springer-Verlag, New York, 1991.
[35] B. C. Berndt, Ramanujan’s notebooks. Part IV, Springer-Verlag, New York, 1994.
[36] B. C. Berndt, Ramanujan’s notebooks. Part V, Springer-Verlag, New York, 1998.
[37] B. C. Berndt, Number theory in the spirit of Ramanujan, Student Math. Library,
34, Amer. Math. Soc., Providence, RI, 2006.
[38] B. C. Berndt and K. Ono, Ramanujan’s unpublished manuscript on the partition
and tau functions with proofs and commentary, The Andrews Festschrift (Maratea,
1998). Sem. Lothar. Combin. 42 (1999), Art. B42c, 63 pp. (electronic).
[39] B. C. Berndt and R. A. Rankin, Ramanujan: letters and commentary, History of
Mathematics, 9. Amer. Math. Soc., Providence, RI, 1995.
[40] B. C. Berndt and R. A. Rankin, Ramanujan: essays and surveys, History of Math-
ematics, 22, Amer. Math. Soc., Providence, RI, 2001.
[41] B. C. Berndt and C. A. Reddi, Two exams taken by Ramanujan in India, Amer.
Math. Monthly, 111 (2004), pages 330–339.
[42] R. E. Borcherds, Automorphic forms on Os+2,2 (R)+ and generalized Kac-Moody
algebras, Proceedings of the International Congress of Mathematicians, Vol. 1, 2
(Zürich, 1994), pages 744–752, Birkhäuser, Basel, 1995.
[43] R. E. Borcherds, Automorphic forms on Os+2,2 (R) and infinite products, Invent.
Math. 120 (1995), pages 161–213.
[44] R. E. Borcherds, Automorphic forms with singularities on Grassmannians, Invent.
Math. 132 (1998), 491–562.
[45] M. Boylan, Arithmetic properties of certain level one mock modular forms, Int. J.
Number Th., accepted for publication.
[46] C. Breuil, B. Conrad, F. Diamond, and R. Taylor, On the modularity of elliptic
curves over Q; wild 3-adic exercises, J. Amer. Math. Soc. 14 (2001), 843–939.
[47] K. Bringmann, On certain congruences for Dyson’s ranks, Int. J. Numb. Th.,
accepted for publication.
[48] K. Bringmann, On the explicit construction of higher deformations of partition
statistics, Duke J. Math., 144 (2008), pages 195–233.
448 K. ONO

[49] K. Bringmann, A. Folsom, and K. Ono, q-series and weight 3/2 Maass forms,
Compositio Math. 145 (2009), pages 541–552.
[50] K. Bringmann, F. Garvan, and K. Mahlburg, Partition statistics and quasiweak
Maass forms, Int. Math. Res. Notices, (2008), rmn. 124.
[51] K. Bringmann and J. Lovejoy, Overpartitions and class numbers of binary quadratic
forms, Proc. Natl. Acad. Sci., USA, 106 (2009), pages 5513–5516.
[52] K. Bringmann, J. Lovejoy, and R. Osburn, Automorphic properties of generat-
ing functions for generalized rank moments and Durfee symbols, Int. Math. Res.
Notices, accepted for publication.
[53] K. Bringmann and K. Ono, The f (q) mock theta function conjecture and partition
ranks, Invent. Math. 165 (2006), pages 243–266.
[54] K. Bringmann and K. Ono, Arithmetic properties of coefficients of half-integral
weight Maass-Poincaré series, Math. Ann., 337 (2007), pages 591–612.
[55] K. Bringmann and K. Ono, Dyson’s ranks and Maass forms, Ann. of Math.,
accepted for publication.
[56] K. Bringmann and K. Ono, Lifting cusp forms to Maass forms with an application
to partitions, Proc. Natl. Acad. Sci., USA 104, No. 10 (2007), pages 3725–3731.
[57] K. Bringmann, K. Ono, and R. Rhoades, Eulerian series as modular forms, J.
Amer. Math. Soc., 21 (2008), pages 1085–1104.
[58] K. Bringmann, K. Ono, and J. Rouse, Traces of singular moduli on Hilbert modular
surfaces, Int. Math. Res. Notices, no. 47 (2005), pages 2891–2912.
[59] K. Bringmann and K. Ono, Coefficients of harmonic weak Maass forms, Proc. Con-
ference on Partitions, q-series, and modular forms (Univ. Florida 2008), accepted
for publication.
[60] K. Bringmann and S. Zwegers, Rank-crank type PDE’s and non-holomorphic Jacobi
forms, Math. Res. Lettrs., accepted for publication.
[61] J. H. Bruinier, Borcherds products and Chern classes of Hirzebruch-Zagier divisors,
Invent. Math. 138 (1999), pages 51–83.
[62] J. H. Bruinier, Borcherds products on O(2, l) and Chern classes of Heegner divisors,
Springer Lect. Notes Math. 1780, Springer-Verlag (2002).
[63] J. H. Bruinier and J. Funke, On Two Geometric Theta Lifts, Duke Math. Journal
125 (2004), pages 45–90.
[64] J. H. Bruinier and J. Funke, Traces of CM-values of modular functions, J. Reine
Angew. Math. 594 (2006), pages 1–33.
[65] J. H. Bruinier, P. Jenkins, and K. Ono, Hilbert class polynomials and traces of
singular moduli, Math. Ann. 334 (2006), pages 373–393.
[66] J. H. Bruinier, W. Kohnen, and K. Ono, The arithmetic of the values of modular
functions and the divisors of modular forms, Compositio Math. 140 (2004), pages
552–566.
[67] J. H. Bruinier and K. Ono, Heegner divisors, L-functions and harmonic weak Maass
forms, Annals of Math., accepted for publication.
[68] J. H. Bruinier, K. Ono, and R. Rhoades, Differential operators and harmonic weak
Maass forms, Math. Ann. 342 (2008), pages 673–693.
[69] J. H. Bruinier and F. Strömberg, Computations of harmonic weak Maass forms,
preprint.
[70] J. H. Bruinier and T. Yang, Faltings heights of CM cycles and derivatives of
L-functions, Invent. Math., accepted for publication.
[71] D. Bump, Automorphic forms and representations, Cambridge Univ. Press,
Cambridge, 1998.
[72] Y.-S. Choi, Tenth order mock theta functions in Ramanujan’s lost notebook, Invent.
Math. 136 No. 3 (1999), pages 497–569.
[73] Y.-S. Choi, Tenth order mock theta functions in Ramanujan’s lost notebook. II,
Adv. Math. 156 (2000), no. 2, pages 180–285.
 HARMONIC MAASS FORMS AND NUMBER THEORY 449

[74] Y.-S. Choi, Tenth order mock theta functions in Ramanujan’s lost notebook. III,
Proc. Lond. Math. Soc. (3) 94 (2007), no. 1, pages 26–52.
[75] Y.-S. Choi, Tenth order mock theta functions in Ramanujan’s lost notebook. IV,
Trans. Amer. Math. Soc. 354 (2002), no. 2, pages 705–733.
[76] Y.-S. Choi, Identities for Ramanujan’s sixth-order mock theta functions, Q. J.
Math. 53 (2002), no. 2, pages 147–159.
[77] D. Choi, D. Jeon, S.-Y. Kang, and C. Kim, Traces of singular moduli of arbitrary
level modular functions, Int. Math. Res. Notices (2007), No. 22.
[78] D. Choi, D. Jeon, S.-Y. Kang, and C. Kim, Exact formulas for traces of sin-
gular moduli of higher level modular functions, J. Number Th. 128 (2008),
pages 700–707.
[79] J. Coates and A. Wiles, On the conjecture of Birch and Swinnerton-Dyer, Invent.
Math. 39 (1977), 223–251.
[80] H. Cohen, q-identities for Maass waveforms, Invent. Math. 91 No. 3 (1988), pages
409–422.
[81] B. Conrad, F. Diamond, and R. Taylor, Modularity of certain potentially Barsotti-
Tate Galois representations, J. Amer. Math. Soc. 12 (1999), 521–567.
[82] D. Cox, Primes of the form x2 + ny 2 , Wiley and Sons, New York, 1989.
[83] P. Deligne, Formes modulaires et reṕresentations -adiques, Sém. Bourbaki 355
(1969).
[84] F. Diamond and J. Shurman, A first course in modular forms, Springer-Verlag,
New York, 2005.
[85] L. Dragonette, Some asymptotic formulae for the mock theta series of Ramanujan,
Trans. Amer. Math. Soc. 72 No. 3 (1952), pages 474–500.
[86] W. Duke, Hyperbolic distribution problems and half-integral weight Maass forms,
Invent. Math. 92 (1988), no. 1, 73–90.
[87] W. Duke, Modular functions and the uniform distribution of CM points, Math.
Ann. 334 (2006), no. 2, pages 241–252.
[88] W. Duke, O. Imamoḡlu, and A. Toth, Cycle integrals of the j-function and weakly
harmonic modular forms, preprint.
[89] W. Duke and P. Jenkins, Integral traces of singular values of weak Maass forms,
Algebra and Number Theory, 2 (2008), pages 573–593.
[90] F. Dyson, Some guesses in the theory of partitions, Eureka (Cambridge) 8 (1944),
pages 10–15.
[91] F. Dyson, A walk through Ramanujan’s garden, Ramanujan revisited (Urbana-
Champaign, Ill. 1987), Academic Press, Boston, 1988, pages 7–28.
[92] T. Eguchi and K. Hikami, Superconformal algebras and mock theta functions,
preprint.
[93] M. Eichler and D. Zagier, The theory of Jacobi forms, Birkhaüser, Boston, 1985.
[94] A. Folsom, A short proof of the Mock Theta Conjectures using Maass forms, Proc.
Amer. Math. Soc. 136 (2008), pages 4143–4149.
[95] A. Folsom and R. Masri, Limiting distribution of traces of Maass-Poincaré series,
preprint.
[96] A. Folsom and K. Ono, Duality involving the mock theta function f (q), J. London
Math. Soc. 77 (2008), pages 320–334.
[97] S. Garthwaite, On questions of congruence and size for modular forms and Maass-
Poincaré series, Ph.D. Thesis (Advisor: K. Ono), University of Wisconsin, 2007.
[98] S. Garthwaite, Vector-valued Maass-Poincaré series, Proc. Amer. Math. Soc., 136
(2008) no. 2, pages 427–436.
[99] F. Garvan, D. Kim, and D. Stanton, Cranks and t-cores, Invent. Math. 101 (1990),
pages 1–17.
[100] D. Goldfeld and P. Sarnak, Sums of Kloosterman sums, Invent. Math. 71 (1983),
pages 243–250.
450 K. ONO

[101] H. Göllnitz, Partitionen mit differenzenbedingungen, J. reine angew. Math. 225

(1967), pages 154–190.
[102] B. Gordon, Some continued fractions of the Rogers-Ramanujan type, Duke Math.
J. 32 (1965), pages 741–748.
[103] B. Gordon and R. McIntosh, Some eighth order mock theta functions, J. London
Math. Soc. 62 No. 2 (2000), pages 321–335.
[104] B. Gordon and R. McIntosh, Modular transformations of Ramanujan’s fifth and
seventh order mock theta functions, Ramanujan J. 7 (2003), pages 193–222.
[105] L. Göttsche and D. Zagier, Jacobi forms and the structure of Donaldson invariants
for 4-manifolds with b+ = 1, Selecta Math. (N.S.) 4 (1998), pages 69–115.
[106] B. Gross and K. Keating, On the intersection of modular correspondences, Invent.
Math. 112 (1993), 225–245.
[107] B. Gross, W. Kohnen, and D. Zagier, Heegner points and derivatives of L-series.
II, Math. Ann. 278 (1987), pages 497–562.
[108] B. Gross and D. Zagier, Heegner points and derivatives of L-series, Invent. Math.
84 (1986), pages 225–320.
[109] P. Guerzhoy, On weak Maass-Modular grids of even integral weights, Math. Res.
Lettrs., accepted for publication.
[110] G. H. Hardy, Srinivasa Ramanujan (1887-1920), Collected papers of Ramanujan
(Ed. G. H. Hardy, et. al.), Cambridge Univ. Press, Cambrdige, 1927, pages xxi–
xxxvi.
[111] G. H. Hardy, Ramanujan: Twelve lectures on subjects suggested by his life and
work, Cambridge Univ. Press, Cambridge, 1940.
[112] G. H. Hardy (with a foreward by C. P. Snow), A mathematician’s apology,
Cambridge Univ. Press, Cambridge, 1992.
[113] G. H. Hardy and S. Ramanujan, Une formule asymptotique pour le nombre des
partitions de n [Comptes Rendus, 2 Jan. 1917] (French) [An asymptotic formula
for the number of partitions of n] Collected papers of Srinivasa Ramanujan, pages
239–241, AMS Chelsea Publ., Providence, RI, 2000.
[114] G. H. Hardy and S. Ramanujan, Asymptotic formulae in combinatory analysis
[Proc. London Math. Soc. (2) 16 (1917), Records for 1 March 1917]. Collected
papers of Srinivasa Ramanujan, 244, AMS Chelsea Publ., Providence, RI, 2000.
[115] E. Hecke, Mathematische Werke, Vandenhoeck & Ruprecht, Third edition,
Göttingen, 1983.
[116] D. A. Hejhal, The Selberg trace formula for P SL(2, R), Springer Lect. Notes in
Math. 548, Springer-Verlag, Berlin 1976.
[117] D. A. Hejhal, The Selberg trace formula for P SL(2, R), Springer Lect. Notes in
Math. 1001, Springer-Verlag, Berlin, 1983.
[118] D. Hickerson, A proof of the mock theta conjectures, Invent. Math. 94 (1988), pages
639–660.
[119] D. Hickerson, On the seventh order mock theta functions, Invent. Math. 94 No. 3
(1988), pages 661–677.
[120] K. Hikami, Mock (false) theta functions as quantum invariants, Regul. Chaotic
Dyn. 10 (2005), pages 163–174.
[121] K. Hikami, Transformation formula of the “second” order mock theta function,
Lett. Math. Phys. 75 (2006), pages 93–98.
[122] K. Hikami, Hecke type formula for the unified Witten-Reshetikhin invariants as
higher order mock theta functions, Int. Math. Res. Not. 7 (2007), 32 pp.
[123] F. Hirzebruch and D. Zagier, Intersection numbers of curves on Hilbert modular
surfaces and modular forms with Nebentypus, Invent. Math. 36 (1976), pages 57–
113.
[124] H. Iwaniec, Non-holomorphic modular forms and their applications, Modular forms
(ed. R. Rankin), Ellis Horwood Ltd., Chichester, 1984.
 HARMONIC MAASS FORMS AND NUMBER THEORY 451

[125] H. Iwaniec, Topics in classical automorphic forms, Grad. Studies in Math. 17,
Amer. Math. Soc., Providence, RI, 1997.
[126] H. Iwaniec, Spectral methods of automorphic forms, Graduate Studies in Math.,
53, Amer. Math. Soc., Providence, RI, 2002.
[127] P. Jenkins, Traces of singular moduli, modular forms, and Maass forms, Ph.D.
Thesis (Advisor: K. Ono), University of Wisconsin, 2006.
[128] P. Jenkins, Kloosterman sums and traces of singular moduli, J. Number Th. 117
(2006), pages 301–314.
[129] V. G. Kac, Infinite dimensional Lie algebras, Cambridge Univ. Press, Cambridge,
1990.
[130] V. G. Kac and M. Wakimoto, Integrable highest weight modules over affine superal-
gebras and number theory, Lie theory and geometry, Progr. Math. 123 Birkhäuser,
Boston, 1994, pages 415–456.
[131] V. G. Kac and M. Wakimoto, Integrable highest weight modules over affine super-
algebras and Appell’s function, Comm. Math. Phys. 215 (2001), pages 631–682.
[132] S.-Y. Kang, Mock Jacobi forms in basic hypergeometric series, Compositio Math.
145 (2009), pages 553–565.
[133] R. Kanigel, The man who knew infinity: A Life of the Genius Ramanujan,
Washington Square, 1992.
[134] N. Koblitz, Introduction to elliptic curves and modular forms, Springer-Verlag,
New York, 1993.
[135] W. Kohnen, Newforms of half-integral weight, J. Reine Angew. Math. 333 (1982),
32–72.
[136] W. Kohnen, Fourier coefficients of modular forms of half-integral weight, Math.
Ann. 271 (1985), 237–268.
[137] W. Kohnen and D. Zagier, Values of L-series of modular forms at the center of
the critical strip, Invent. Math. 64 (1981), no. 2, pages 175–198.
[138] V. A. Kolyvagin, Finiteness of E(Q) and the Tate-Shafarevich group of E(Q) for a
subclass of Weil curves (Russian), Izv. Akad. Nauk, USSR, ser. Matem. 52 (1988),
522–540.
[139] D. Kubert and S. Lang, Modular units, Grundlehren der Mathematischen Wis-
senschaften [Fundamental Principles of Mathematical Science], 244, Springer-
Verlag, New York-Berlin, 1981.
[140] S. Kudla, Central derivatives of Eisenstein series and height pairings, Ann. of
Math. (2) 146 (1997), 545–646.
[141] S. Kudla, Special cycles and derivatives of Eisenstein series, in Heegner points
and Rankin L-series, Math. Sci. Res. Inst. Publ. 49, Cambridge University Press,
Cambridge, 2004.
[142] S. Kudla, M. Rapoport, and T. H. Yang, Modular forms and special cycles on
Shimura curves, Annals of Math. Studies series, vol 161, Princeton Univ. Publ.,
2006.
[143] S. Lang, Introduction to modular forms, Springer-Verlag, Berlin, 1995.
[144] R. Lawrence and D. Zagier, Modular forms and quantum invariants of 3-manifolds,
Asian J. Math. 3 (1999), pages 93–107.
[145] J. Lepowsky and R. L. Wilson, A new family of algebras underlying the Rogers-
Ramanujan identities and generalizations, Proc. Natl. Acad. Sci., USA 78 (1981),
pages 7254–7258.
[146] M. Lerch, Bemerkungen zur Theorie der elliptischen Functionen, Jahrbuch über
die Fortschritte der Mathematik 24 (1892), pages 442–445.
[147] M. Lerch, Poznámky k theorii funkcí elliptických, Rozpravy České Akademie Císaře
Františka Josefa pro vědy, slovesnost a umění v praze (IIC1) I 24 (1892), pages
465–480.
452 K. ONO

[148] J. Lewis and D. Zagier, Period functions for Maass wave forms. I., Ann. Math.
153 (2001), pages 191–258.
[149] H. Maass, Über eine neue Art von nichtanalytischen automorphen Funktionen und
die Bestimmung Dirichletscher Reihen durch Funktionalgleichungen, Math. Ann.
121 (1949), pages 141–183.
[150] H. Maass, Die Differentialgleichungen in der Theorie der elliptische Modulfunktio-
nen, Math. Ann. 125 (1953), pages 235–263.
[151] K. Mahlburg, Partition congruences and the Andrews-Garvan-Dyson crank, Proc.
Natl. Acad. Sci. USA 102 (2005), no. 43, pages 15373–15376.
[152] A. Malmendier and K. Ono, SO(3)-Donaldson invariants of CP 2 and mock theta
functions, submitted for publication.
[153] A. Mellit and S. Okada, Joyce invariants for K3 surfaces and mock theta functions,
preprint.
[154] A. Miller and A. Pixton, Arithmetic traces of non-holomorphic modular invariants,
Int. J. of Number Theory, accepted for publication.
[155] T. Miyake, Modular forms, Springer-Verlag, Berlin, 2006.
[156] M. Monks, Number theoretic properties of generating functions related to Dyson’s
rank for partitions into distinct parts, preprint.
 ∞ at2 +bt
[157] L. J. Mordell, The value of the definite integral −∞ eect +d dt, Quarterly J. of
Math. 68 (1920), pages 329–342.
 ∞ at2 +bt
[158] L. J. Mordell, The definite integral −∞ eect +d dt and the analytic theory of num-
bers, Acta Math. 61 (1933), pages 323–360.
[159] W. Nahm, Conformal field theory and torsion elements of the Bloch group,
Frontiers of number theory, physics, and geometry I, Springer-Verlag, Berlin, 2007,
pages 67–132.
[160] D. Niebur, A class of nonanalytic automorphic functions, Nagoya Math. J. 52
(1973), pages 133–145.
[161] D. Niebur, Construction of automorphic forms and integrals, Trans. Amer. Math.
Soc. 191 (1974), pages 373–385.
[162] K. Ono, Distribution of the partition function modulo m, Ann. of Math. (2) 151
(2000), no. 1, pages 293–307.
[163] K. Ono, Arithmetic of the partition function, Special functions 2000: current per-
spective and future directions (Tempe, AZ), NATO Sci. Ser. II Math. Phys. Chem.,
30, Kluwer Acad. Publ., Dordrecht, 2001, pages 243–253.
[164] K. Ono, The web of modularity: arithmetic of the coefficients of modular forms and
q-series, CBMS Regional Conference Series in Mathematics, 102, Amer. Math.
Soc., Providence, RI, 2004.
[165] K. Ono, Honoring a gift from Kumbakonam, Notices Amer. Math. Soc. 53 (2006),
no. 6, pages 640–651.
[166] K. Ono, A mock theta function for the Delta-function, Proceedings of the 2007
Integers Conference, de Gruyter, Berlin, 2009, pages 141–156.
[167] H. Petersson, Über die Entwicklungskoeffizienten der automorphen Formen, Acta
Math. 58 (1932), pages 169–215.
[168] H. Rademacher, On the expansion of the partition function in a series, Ann. of
Math. (2) 44, (1943), pages 416–422.
[169] H. Rademacher, Topics in analytic number theory, Die Grundlehren der mathema-
tischen Wissenschaften, Band 169, Springer Verlag New York-Heidelberg, 1973.
[170] H. Rademacher and H. S. Zuckerman, On the Fourier coefficients of certain modu-
lar forms of positive dimension, Ann. of Math. (2) 39 (1938), no. 2, pages 433–462.
[171] S. Ramanujan, Problem 584, J. Indian Math. Soc. 6 (1914), pages 199–200.
[172] S. Ramanujan, On certain arithmetical functions, Trans. Camb. Phil. Soc. 22
(1916), pages 159–184.
 HARMONIC MAASS FORMS AND NUMBER THEORY 453

[173] S. Ramanujan, Some properties of p(n), the number of partitions of n, [Proc.

Cambridge Philos. Soc. 19 (1919), pages 207–210]. Collected papers of Srinivasa
Ramanujan, 210–213, AMS Chelsea Publ., Providence, RI, 2000.
[174] S. Ramanujan, Congruence properties of partitions, Math. Z. 9 (1921), no. 1–2,
pages 147–153.
[175] S. Ramanujan, Collected papers of Ramanujan, Cambridge Univ. Press, Cam-
bridge, 1927.
[176] S. Ramanujan, The lost notebook and other unpublished papers, Narosa, New Delhi,
1988.
[177] R. A. Rankin, Modular forms and functions, Cambridge Univ. Press, Cambridge-
New York-Melbourne, 1977.
[178] R. Rhoades, Linear relations among Poincaré series, preprint.
[179] R. Rhoades, Interplay between Weak Maass Forms and Modular Forms and Sta-
tistical Properties of Number Theoretic Objects, Ph.D. thesis (Advisor: K. Ono),
University of Wisconsin, 2008.
[180] K. Ribet, Galois representations attached to eigenforms with Nebentypus, Springer
Lect. Notes in Math. 601, (1977), pages 17–51.
[181] W. Roelcke, Automorphe formen in der hyperbolishce Ebene, I, Math. Ann. 167
(1966), pages 292–337.
[182] L. J. Rogers, Second memoir on the expansion of certain infinite products, Proc.
London Math. Soc. 25 (1894), pages 318–343.
[183] L. J. Rogers, On two theorems of combinatory analysis and some allied identities,
Proc. London Math. Soc. 16 (1917), pages 315–336.
[184] L. J. Rogers, Proof of certain identities in combinatory analysis, Proc. Cambridge
Philos. Soc. 19 (1919), pages 211–214.
[185] P. Sarnak, Some applications of modular forms, Cambridge Tracts in Math, 99,
Cambridge Univ. Press, Cambridge, 1990.
[186] B. Schoeneberg, Elliptic modular functions: an introduction, Springer-Verlag,
New York-Heidelberg, 1974.
[187] J.-P. Serre, Congruences et formes modulaires (d’apres H. P. F. Swinnerton-Dyer),
Sem. Bourbaki 416 (1971–1972), pages 74–88.
[188] J.-P. Serre, Divisibilité de certaines fonctions arithmétiques, Enseign. Math. 22
(1976), pages 227–260.
[189] J.-P. Serre and H. Stark, Modular forms of weight 1/2, Springer Lect. Notes 627,
Springer-Verlag, Berlin, (1977), pages 27–67.
[190] A. Selberg, Über die Mock-Thetafunktionen siebenter Ordnung, Arch. Math. Natur.
idenskab, 41 (1938), pages 3–15 (see also Coll. Papers, I, pages 22–37).
[191] A. Selberg, Harmonic analysis and discontinuous groups in weakly symmetric
Riemannian spaces, J. Indian Math. Soc. 20 (1956), pages 47–87.
[192] G. Shimura, On modular forms of half integral weight, Ann. of Math. 97 (1973),
pages 440–481.
[193] G. Shimura, Introduction to the arithmetic theory of automorphic functions,
Reprint of the 1971 original. Publications of the Math. Soc. of Japan, 11. Kan
Memorial Lectures, 1. Princeton Univ. Press, Princeton, NJ, 1994.
[194] J. Silverman, The arithmetic of elliptic curves, Springer-Verlag, New York, 1986.
[195] L. J. Slater, Further identities of the Rogers-Ramanujan type, Proc. London Math.
Soc. Ser. no. 2 54 (1952), pages 147–167.
[196] W. Stein, Modular forms, a computational approach, Amer. Math. Soc.,
Providence, RI, 2007.
[197] H. P. F. Swinnerton-Dyer, On -adic representations and congruences for coeffi-
cients of modular forms, Springer Lect. Notes. 350 (1973), pages 1–55.
[198] R. Taylor and A. Wiles, Ring-theoretic properties of certain Hecke algebras, Ann.
of Math. 141 (1995), 553–572.
454 K. ONO

[199] A. Terras, Harmonic analysis on symmetric spaces and applications, I, Springer-

Verlag, New York, 1985.
[200] A. Terras, Harmonic analysis and symmetric spaces and applications, II, Springer-
Verlag, Berlin, 1988.
[201] S. Treneer, Congruences for the coefficients of weakly holomorphic modular forms,
Proc. London Math. Soc., 93 (2006), pages 304–324.
[202] J. Tunnell, A classical Diophantine problem and modular forms of weight 3/2
Invent. Math. 72 (1983), no. 2, pages 323–334.
[203] A. Venkov, Spectral theory of automorphic functions, Kluwer Acad. Publ.,
Dordrecht, 1990.
[204] M. Wakimoto, Representation theory of affine superalgebras at the critical level,
Proc. Int. Congress of Math., Vol. II, (Berlin, 1998), Doc. Math. 1998, Extra Vol.,
II, pages 605–614.
[205] J.-L. Waldspurger, Sur les coefficients de Fourier des formes modulaires de poids
demi-entier, J. Math. Pures Appl. (9) 60 (1981), no. 4, pages 375–484.
[206] G. N. Watson, The final problem: An account of the mock theta functions, J.
London Math. Soc. 2 (2) (1936), pages 55–80.
[207] G. N. Watson, The mock theta functions (2), Proc. London Math. Soc. (2) 42
(1937), pages 274–304.
[208] G. N. Watson, Ramanujan’s Vermutung über Zerfällungsanzahlen, J. Reine Angew.
Math. 179 (1938), pages 97–128.
[209] A. Wiles, Modular elliptic curves and Fermat’s last theorem, Ann. of Math. 141
(1995), 443–551.
[210] D. Zagier, Nombres de classes et formes modulaires de poids 3/2, C. R. Acad. Sci.
Paris Sér. A-B 281 (1975), pages 883–886.
[211] D. Zagier, Vassiliev invariants and a strange identity related to the Dedekind eta-
function, Topology 40 (2001), pages 945–960.
[212] D. Zagier, Traces of singular moduli, Motives, Polylogarithms and Hodge Theory,
Part I. International Press Lecture Series (Eds. F. Bogomolov and L. Katzarkov),
International Press, 2002, pages 211–244.
[213] D. Zagier, Ramanujan’s mock theta functions and their applications [d’après
Zwegers and Bringmann-Ono], Séminaire Bourbaki, 60éme année, 2006–2007, no.
986.
[214] S. Zhang, Heights of Heegner cycles and derivatives of L-series, Invent. Math. 130
(1997), pages 99–152.
[215] S. Zhang, Heights of Heegner points on Shimura curves, Ann. of Math. 153 (2001),
pages 27–147.
[216] H. S. Zuckerman, On the coefficients of certain modular forms belonging to sub-
groups of the modular group, Trans. Amer. Math. Soc., 45 (1939), pages 298–321.
[217] H. S. Zuckerman, On the expansions of certain modular forms of positive dimen-
sion, Amer. J. Math. 62 (1940), pages 127–152.
[218] S. P. Zwegers, Mock ϑ-functions and real analytic modular forms, q-series with
applications to combinatorics, number theory, and physics (Ed. B. C. Berndt and
K. Ono), Contemp. Math. 291, Amer. Math. Soc., (2001), pages 269–277.
[219] S. P. Zwegers, Mock theta functions, Ph.D. Thesis (Advisor: D. Zagier), Univer-
siteit Utrecht, 2002.

Department of Mathematics, University of Wisconsin, Madison, Wisconsin

53706 USA
E-mail address: [email protected]