AN INTROCUTION TO MODULAR FORMS

SIDDHARTH KOTHARI

Definition 0.1. Let k ∈ Z. A modular form of weight k for the group SL2 (Z) is a function
f : H → C such that
• f is holomorphic on H,  
k f (τ ), for all a b ∈ SL (Z) and τ ∈ H,
 
• (Modularity condition)f aτ +b
cτ +d = (cτ + d) 2
c d
• |f (τ )| is bounded as Im(τ ) → ∞.
Remark 0.2. Note that the zero function is a modular form of any weight. Further, observe that
there are no non-zero modular forms of odd weight: Set a = −1 = d and b = 0 = c in the modularity
condition. Then f (τ ) = (−1)k f (τ ) =⇒ f (τ ) = −f (τ ) =⇒ f (τ = 0) for all τ ∈ H.
This definition poses many questions. Why is the function define only on H := {z ∈ C : I(z) > 0},
the upper-half plane? Why is f composed with the function (aτ + b)(cτ + d)−1 , and what does
SL2 (Z) have to do with any of this? And, most importantly, why would one care to learn more
about these kind of objects? Where do they fit in the landscape of mathematics? We aim to
answer, at least partially, these questions here.

1. The Modularity Condition and Homogeneous Functions on Lattices

We start our discussion with a lattice, which generalizes the ‘discreteness’ of the relation between
Z to R, in an attempt to ’draw out’ requiremnt number 2 from the definition of a modular form.
Definition 1.1. Let ω1 and ω2 be complex numbers. Define the lattice generated by ω1 and ω2 ,
by Λ(ω1 , ω2 ) := {a1 ω1 + a2 ω2 : ai ∈ C}. We require ω1 and ω2 to be linearly independent, that is,
ω1 /ω2 ̸∈ R, so that Λ(ω1 , ω2 ) is not a line. Any set of this form is called a lattice.
A trivial way in which we can create another lattice out of a pre-existing one, say Λ, is simply by
scaling it by some non-zero complex number λ: λΛ := {λa : a ∈ Λ}. With this in mind, consider
a function F : L → C, where L is the set of all lattices. If F is such that F (λΛ) = λk F (Λ) for all
λ ∈ C× and Λ ∈ L for some integer k, then it does seem plausible to call it a generalization of a
linear map, since the λ comes out of the ride, albeit by some factor k that may not be one. We call
such an F homogeneous (on lattices) of weight k.
For any linearly independent ω1 and ω2 , we can consider the function G(ω1 , ω2 ) := F (Λ(ω1 , ω2 )).
The homogeneity of F implies that
G(αω1 , αω2 ) = F (Λ(αω1 , αω2 )) = F (αΛ(ω1 , ω2 )) = αk G(ω1 , ω2 ).
Taking α to be ω1−1 , we have G(1, ω2 /ω2 ) = ω1k G(ω1 , ω2 ) = F (Λ(1, ω2 /ω1 )), so G is completely
determined by where F sends lattices generated by {1, τ } for some τ ∈ C. This allows us to
consider a uni-variate function f (τ ) := G(1, τ ) = F (Λ(1, τ )).
However, there are constraints on f : given two complex numbers τ1 ̸= τ2 , we may have Λ(1, τ1 ) =
Λ(1, τ2 ), meaning we have f (τ1 ) = f (τ2 ). And when does this happen?
Proposition 1.2. Two pairs (1, τ1 ) and (1, τ2 ) are such that Λ(1, τ1 ) = Λ(1, τ2 ) if and only if there
exists integers a, b, c and d such that ad − bc = 1 and
    
a b τ1 τ
= 2 .
c d 1 1
1
2 SIDDHARTH KOTHARI

Indeed, these ‘change of basis’ like matrices have a special name.

Definition 1.3. The special linear group over Z of degree 2, denoted by SL2 (Z), is defined to be
the set of 2 × 2 matrices with integer entries and with determinant 1.
Remark 1.4. SL2 (Z) forms a group under matrix multiplication.
Therefore, we have
f (τ1 ) = F (Λ(1, τ1 )) = F (Λ(cτ1 + d, aτ1 + b))
  
aτ1 + b
= F (cτ1 + d)Λ 1,
cτ1 + d
  
k aτ1 + b
= (cτ1 + c) F Λ 1,
cτ1 + d
 
k aτ1 + b
= (cτ1 + d) f ,
cτ1 + d
meaning that f satisfies the modularity condition!
 
a b
Remark 1.5. Note that the map τ 7→ aτ +b
cτ +d with ∈ SL2 (Z) defines a left group action on H.
c d
Indeed, the formula  
aτ + b (ad − bc)I(τ )
I = ,
cτ + d |cτ + d|2
applicable for all a, b, c, d ∈ R and τ ∈ C such that τ ̸= −d/c, means that γτ ∈ H for all τ ∈ H and
γ ∈ SL2 (Z). It can also be shown that γ1 (γ2 τ ) = (γ1 γ2 )τ for γ1 , γ2 ∈ SL2 (Z) simply by expanding
everything out.
Before we end this section, we introduce the first non-trivial modular form.
Definition 1.6. The Eisenstein series of weight k for k ≥ 1 is a function from H to C defined by
X 1
Ek (τ ) = .
2
(m + nτ )k
(m,n)∈{Z −(0,0)}

This definition is quite latticy. Indeed,

X 1
Ek (τ ) = ,
ωk
ω∈Λ(1,τ )∗

where Λ(1, τ )∗ denotes the non-zero elements of Λ(1, τ ). From our previous work, it should not
come as a suspire that Ek satisfies the modularity condition. Spieling out the details,
Proof. We have
(cτ + d)k
 
aτ + b X 1 X
Ek = k =
cτ + d (m(cτ + d) + n(aτ + b))k
 
aτ +b
(m,n)∈Z2 −{(0,0)} m+n cτ +d (m,n)∈Z2 −{(0,0)}
X 1
= (cτ + d)k
((md + nb) + τ (cm + an))k
(m,n)∈Z2 −{(0,0)}
X 1
= (cτ + d)k .
! (m′ + n′ τ )k
d c
(m′ ,n′ )=(m,n)
b a
 AN INTROCUTION TO MODULAR FORMS 3
 
d c
As ad − bc = 1, the map from Z2 to Z2 defined by (m, n) 7→ (m, n) is a bijection, which
b a
completes the proof. Eisenstein series satisfy curious identities involving ζ(s) and σ(n). Also, note
that E42 = E8 and E6 E4 = E10 , and in general, any E2k can be expressed as a polynomial expression
in E4 and E6 , a fact whose generalization we will explore later.
■

2. q−series for a modular form

Recall the Fouries series of a function, defined below.
Proposition 2.1. Let f : R → R be a 2π−periodic function. Then we have that
∞
X
f (x) = fˆ(n)einx ,
n=−∞
where Z π
1
f (n) = f (x)e−inx dx,
2π −π
for all n ∈ Z.
The key take-away here was to analyze a possibly complex function by breaking it up into
several manageable parts. More precisely, we expressed f as a linear combination of the exponential
functions, einx , which is itself a combination of the fundamental, the most basic, periodic functions:
sin(nx) and cos(nx). It terns out we can do something similar for modular forms (Observe that
modular forms have ’a lot’ of periodicity built into them. In particular, f (τ + 1) = f (τ ).) as well,
as illustrated in the next theorem.
Theorem 2.2. Let f : H → C be a function such that
• f is holomorphic,
• f (τ + 1) = f (τ ) for all τ ∈ H,
• and |f (τ )| is bounded as J(τ ) → ∞.
Then there exist complex numbers (ai )∞ i=0 such that
∞
X
f (τ ) = an e2πinτ .
n=1

Before going into the formal proof, note that the above series resembles a power series expansion
in e2πiτ . Thus, we define the function q(τ ) := e2πiτ from H to the punctured unit disk.
The proof is broken up into a series of lemmas.
Lemma 2.3. The function defined by q(τ ) := e2πiτ for all τ ∈ H is locally invertible and maps
surjectively to the punctured disk D′ := {z ∈ C : 0 < |z| < 1}.
Proof. Write τ = x + yi where y > 0. Then e2πiτ = e2πi(x+yi) = e2πix−2πy = e−2πy e2πix . Thus,
|q(τ )| = |e−2πy | = e−2πy ∈ (0, 1). Therefore, q(τ ) ∈ D′ . Next, being locally invertible is equivalent
to having a non-zero derivative throughout a function’s domain. We have q ′ (τ ) = 2πie2πiτ , which
is clearly non-zero as the complex exponential is never zero. Conversely, a z ∈ D′ can be written
as z ′ = au, where 1 > a > 0 is such that a = |z|, and hence can be written in the form e−2πy for
y > 0, and u is a unit complex number, which can also be written as e2πix for some real x. ■
Lemma 2.4. Let the function g : D′ → C be defined by g(q) = f (τ ) where e2πiτ = q. Then g is
well-defined, holomorphic and bounded.
Next, we compute the q−series of the Eisenstein series.
4 SIDDHARTH KOTHARI

Theorem 2.5. For k ≥ 4, we have

∞
2(2πi)k X
Ek (τ ) = 2ζ(k) + σk−1 (n)e2πin .
(k − 1)!
n=1

To prove this, we assume the Poisson summation formula.

3. The Vector space of Modular forms

Let Mk denote the vector space of all modular forms of weight k of SL2 (Z).
Lemma 3.1. If k < 0, then Mk = {0}.
Theorem 3.2. Every Mk is finite dimensional, and dim Mk is [k/12] + 1 if k ̸≡ 2 modulo 12 and
dim Mk is [k/12] if k is 2 modulo 12.