Joachim Wehler

Modular Forms
and Elliptic Curves
DRAFT, Release 1.50

April 16, 2021

2

c Joachim Wehler, 2020, 2021

I prepared these notes for the participants of the lectures. The lecture took place at
the mathematical department of LMU (Ludwig-Maximilians-Universität) at
Munich. The first time during the winter term 2017/18 and then during the winter
term 2020/21. Some parts of the notes rely on notes of an unpublished lecture
of O. Forster.

Compared to the oral lecture in class these written notes contain some additional
material.

I thank all participants - in particular W. Hensgen - for pointing out some errors
and for their proposals for improvement. Please report to

[email protected]

any further errors or typos, adding the version of the lecture notes.

Release notes:
• Release 1.50: Update acknowledgment.
• Release 1.49: Chapter 5. Remark 5.15 moved from Chapter 7. Chapter 7.
Revision.
• Release 1.48: Chapter 6. Revision, Lemma 6.21 added.
• Release 1.47: Chapter 5. Proposition 5.9 added.
• Release 1.46: Chapter 3 and Chapter 2. Revision.
• Release 1.45: Chapter 4. Revision.
• Release 1.44: Chapter 5. Revision.
• Release 1.43: Chapter 5. Theorem 5.25 proof corrected.
• Release 1.42: Chapter 5. Theorem 5.25 corrected. Chapter 6 minor revisions.
• Release 1.41: Chapter 7. Section 7.2 added. Chapter 5. Theorem 5.23 expanded.
Theorem 5.25 expanded and corollary integrated.
• Release 1.40: Chapter 7. Section 7.1 added.
• Release 1.39: Chapter 4. Proposition 4.25 expanded.
• Release 1.38: Chapter 6. Sections 6.1 and 6.2 added.
• Release 1.37: Chapter 5. Section 5.2, Lemma 5.24 replaced by new version,
subsequent Corollary added, Theorem 5.25 adapted, renumbering, Section 5.3
added.
• Release 1.36: Chapter 5. Section 5.2, Proposition 5.28 added, minor revision.
• Release 1.35: Chapter 5. Section 5.2 added.
• Release 1.34: Chapter 5. Section 5.1, minor revision.
• Release 1.33: Chapter 4. Example 4.39 expanded.
• Release 1.32: Chapter 5. Section 5.1 added.
• Release 1.31: Chapter 4. Section 4.2, Definition 4.19 expanded, Lemma 4.30
and Theorem 4.31 clarified, Remark 4.27 added, renumbering, minor revision.
Section 4.3 added.
 3

• Release 1.30: Chapter 4. Section 4.2, Example 4.36 added, minor revision.
• Release 1.29: Chapter 3, Lemma 3.4, exponent of determinant changed.
Chapter 4, Example 4.4 added. Renumbering, Section 4.2 added.
• Release 1.28: Chapter 4. Proposition 4.8, proof corrected. Section 4.1, minor
revisions.
• Release 1.27: Chapter 2, Remark 2.30, formulas corrected by extended
Legendre symbol and Kronecker symbol. Chapter 4, Proposition 4.6 added,
minor revisions, renumbering. List of results updated.
• Release 1.26: Chapter 3, proof of Lemma 3.4: formula corrected. Chapter 4,
minor revisions
• Release 1.25: Chapter 4, Section 4.1 added.
• Release 1.24: Minor revisions.
• Release 1.23: Minor revisions.
• Release 1.22: Chapter 3 completed.
• Release 1.21: Section 3.2, some typos corrected.
• Release 1.20: Minor revisions.
• Release 1.19: Minor revisions.
• Release 1.18: Chapter 2, Remark 3.5 added.
• Release 1.17: Minor revisions.
• Release 1.16: Chapter 2, Remark 2.8, Example 2.11, Lemma 2.26 corrected.
Chapter 3, Section 3.1 reordered.
• Release 1.15: Chapter 3, Remark 3.7 added, Remark 3.7 expanded, minor
revisions, renumbering.
• Release 1.14: Minor revisions.
• Release 1.13: Chapter 3, Section 3.2 added. List of results added.
• Release 1.12: Chapter 1, Proposition 1.19 added, renumbering. Chapter 2,
Remark 2.17 added, renumbering. Chapter 3, Section 3.1 added.
• Release 1.11: Chapter 2 completed.
• Release 1.10: Chapter 2, Section 2.2 correction of some typos.
• Release 1.9: Chapter 1, revision of Corollary 1.9 and 1.10, Lemma 1.13,
Corollary 1.15 and Theorem 1.21. Further minor revisions.
• Release 1.8: Chapter 2 minor revision.
• Release 1.7: Chapter 2 minor revision, some additions.
• Release 1.6: Chapter 2 minor revision.
• Release 1.5: Chapter 2, Sections 2.1 and 2.2 added.
• Release 1.4: Chapter 1, minor revision.
• Release 1.3: Introduction added.
• Release 1.2: Chapter 1, minor revision.
• Release 1.1: Complete revision, starting with Chapter 1.
Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

PARI files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Part I General Theory

1 Elliptic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.1 The field of elliptic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 The Weierstrass ℘-function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3 Abel’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2 The modular group Γ and its Hecke congruence subgroups Γ0 (N) . . . 31

2.1 The moduli space of complex tori and group actions . . . . . . . . . . . . . 31
2.2 Topology of the orbit space of the Γ -action . . . . . . . . . . . . . . . . . . . . . 46
2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces . . . . . . . . . . 63

3 The algebra of modular forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

3.1 Modular forms and cusp forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.2 Eisenstein series and the algebra of modular forms of Γ . . . . . . . . . . 90
3.3 Generalization to Hecke congruence subgroups Γ0 (N) . . . . . . . . . . . . 117

4 Elliptic curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

4.1 Embedding tori as plane cubic hypersurfaces . . . . . . . . . . . . . . . . . . . 125
4.2 Elliptic curves over subfields of C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.3 Elliptic curves over finite fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

5 Introduction to Hecke theory and applications . . . . . . . . . . . . . . . . . . . . 181

5.1 Hecke operators of the modular group and their eigenforms . . . . . . . 181
5.2 The Petersson scalar product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
5.3 Numerology: Lagrange, Jacobi, Ramanujan, Mordell . . . . . . . . . . . . . 217

Part II Advanced Theory

v
vi Contents

6 Application to imaginary quadratic fields . . . . . . . . . . . . . . . . . . . . . . . . . 229

6.1 Imaginary quadratic fields and tori with complex multiplication . . . . 229
6.2 Modular polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
6.3 Fractional ideals and the class number formula . . . . . . . . . . . . . . . . . . 251

7 Outlook: Modular elliptic curves, monstrous moonshine . . . . . . . . . . . . 261

7.1 Modular elliptic curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
7.2 Monstrous moonshine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266

List of results and some outlooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Introduction

Modular Forms and elliptic curves are a classical domain from mathematics. At
least, since the proof of Fermat’s last conjecture the domain attracts widespread at-
tention. The domain of Modular Forms integrates the three mathematical disciplines
Complex Analysis, Algebraic Geometry, and Algebraic Number Theory.

Elliptic curves can be investigated by different mathematical methods.

• Algebraic geometry: Elliptic curves are the zero-sets of polynomials.

• Complex analytic geometry: When considered as Riemann surfaces then elliptic

curves are complex tori.

• Arithmetic geometry: Elliptic curves can be defined over different fields, e.g.
over Q or F p and over the ring Z.
The algebraic point of view identifies complex elliptic curves with smooth cubic
hypersurfaces of P2 . Hence elliptic curves are the first ones in the series of cubic
hypersurfaces in complex projective space Pn , n ≥ 2. The higher dimensional
cubics are also challenging examples, see [29, Chap.V, §4], [31].

Chapter 1 introduces elliptic functions as doubly periodic, meromorphic func-

tions in the complex plane C. The period group is a lattice Λ ⊂ C. The main tool to
study elliptic function is the residue theorem.

Chapter 2 characterizes elliptic functions as meromorphic functions on the

torus C/Λ . The classes of biholomorphically equivalent complex tori are the orbits
of a group action of the modular group

Γ ×H →
− H

The quotients

1
2 Introduction

Γ0 (N) \H
of the restricted action of the Hecke congruence subgroups

Γ0 (N) ⊂ Γ

are Riemann surfaces with compactifications X0 (N).

Chapter 3 introduces modular forms and cusp forms as holomorphic functions

on H ∪ {∞} with a certain transformation behaviour with respect to the Γ0 (N)-action.
These functions unreveal themselves as meromorphic differential forms on the mod-
ular curve X0 (N). The Riemann-Roch theorem computes the dimensions of the vec-
tor spaces of modular forms and cusp forms.

Chapter 4 makes a new start from the viewpoint of algebraic geometry: Complex
tori embed as elliptic curves into the complex projective plane. They can be repre-
sented as non-singular cubic curves, defined as zero set of a Weierstrass polynomial.
The main tool to represent a complex torus as an elliptic curve is the Weierstrass ℘-
function and its differential equation. Conversely, each complex elliptic curve arises
as embedding of a complex torus.
Choosing different fields for the coefficents of the Weierstrass polynomials al-
lows to consider elliptic curves defined over C, Q, Z or even over the finite fields F p .
The focus of the chapter are some relations between complex analytic geometry, al-
gebraic geometry and arithmetic geometry.

Chapter 5 considers families of Hecke operators, which are linear endomor-

phisms on the finite dimensional vector spaces of modular forms. Each family of
Hecke operators acts on a given vector space of modular forms of fixed weight. On
the corresponding subspace of cusp forms the family can be diagonalized simulta-
neously, a result which relies on the Petersson scalar product. As an application the
chapter proves some theorems from algebraic number theory related to Lagrange,
Jacobi, Ramanujan, Mordell and other mathematicians.

In the second, more advanced part of the lecture notes Chapter 6 deals with
deeper applications of the theory of modular forms to algebraic number theory. The
chapter investigates the relation between tori with complex multiplication and imag-
inary quadratic number fields. The main role is played by the modular invariant j.
As an application the chapter proves a lower bound of the class number.

The final Chapter 7 gives an outlook to the modularity theorem for elliptic curves,
which plays the dominant role in Wiles’ proof of the Fermat conjecture. A second
section gives an outlook to the moonshine relation between the Fourier coefficients
of the modular j-invariant and the dimensions of the irreducible representations of
the monster group, which at day culminates in the work of Borcherds.

It is also the aim of these lecture notes to illustrate their results and the outlook
by a series of numerical calculations using a computer algebra system.
PARI files

Chapter 1:
• Weierstrass_p_function_08: Laurent expansion of the
Weierstrass ℘-function and its derivative, cf. Remark 1.14.

Chapter 2
• modular_curve_as_covering: Genus, degree and further parameters of
modular curve X0 (N), cf. Remark 2.30.

Chapter 3
• Congruence_subgroup_19: Dimension of Mk (Γ0 (pn )) via Riemann-Roch,
cf. Remark 3.27.

Chapter 4
• Torus_to_Weierstrass_equation_01: The plane cubic of an
embedded torus, cf. Example 4.4

• Elliptic_curve_plot_06: Plot of some plane cubics, cf. Example 4.29.

• Elliptic_curve_02: Invariants of some plane cubics, cf. Example 4.29.

• Elliptic_curve_weierstrass_equation_12: Global minimal

Weierstrass polynomial and its discriminant, cf. Example 4.39.

• Elliptic_curve_Hasse_estimate_05: Hasse estimate for number

of F p -rational points, cf. Remark 4.42.

• Elliptic_curve_04_02: Analytic rank of elliptic curves, cf. Remark 4.47.

3
4 PARI files

Chapter 5
• Hecke_matrix_02: Hecke operator T2 ∈ End(S28 (Γ )), cf. Example 5.14.

• Congruence_subgroup_20: Oldforms and newforms for different Γ0 (N),

cf. Example 5.30.

• Modular_forms_theta_01: 4-squares theorem and similar

decompositions, cf. Example 5.39.

Chapter 6
• modular_polynomial_01: Modular polynomials of different levels, cf.
Example 6.9.

• modular_polynomial_02: The polynomials Φm and their factorization, cf.

Example 6.11.

Chapter 7
• Elliptic_curve_taniyama_10: Illustration of Wiles’ theorem by some
numerical examples, cf. Example 7.7.

The PARI-files of these lecture notes are contained in a separate folder. It is

recommended to download all of them together, because some of them will call
other files. After downloading the PARI files into the working directory of PARI a
given file can be called by the command

\r filename

For more information see the PARI manual from the PARI homepage.
 Part I
General Theory
6

The context of Chapter 1 is complex analysis on domains in the complex plane.

The base field is the field C of complex numbers. We develop the theory of elliptic
functions, i.e. doubly periodic meromorphic functions. Classical tools are Laurent
series and the residue theorem for meromorphic functions. The period group of
doubly periodic functions are lattices in C.
The view point of Chapter 2 are compact Riemann surfaces. We continue work-
ing with the base field C, but now the domain is a complex manifold. The most
simple example of a compact Riemann surface is the Riemann sphere P1 (C). Mero-
morphic functions are holomorphic maps into the Riemann sphere. A second class
of compact Riemann surfaces are complex tori. The doubly periodic meromorphic
functions from Chapter 1 are meromorphic functions on a complex torus. The main
tools from the theory of Riemann surfaces are line bundles, divisors and the Theo-
rem of Riemann-Roch. The set of isomorphy classes of complex tori turns out to be
the orbit space of the modular group SL(2, Z) acting on the upper half plane H. We
show that the compactified orbit space is biholomorphic equivalent to P1 (C).
Chapter 4 switches from the analytic theory to an algebraic context. We con-
sider the Riemann sphere from an algebraic point of view and study non-singular
curves C in the projective space P1 (C), now equipped with the Zariski topology.
These curves are the zero set of one or more homogeneous polynomials. The small-
est subfield k ⊂ C which comprises their coefficients is the field of definition of C.
As a consequence we extend the range of base fields under consideration to the
field Q and to the residue fields F pn of Z. The Weierstrass ℘-function and its deriva-
tive ℘ 0 from Chapter 1 embedd a given torus into P2 (C). The image is as an elliptic
curve E. It is the zero set of a Weierstrass polynomial F, a homogeneous polyno-
mial of degree 3. If all coefficients of F are rational and if in addition a Q-valued
point of E exists, then E is a rational elliptic curve. The existence of a rational
point O ∈ E(Q) allows to equip E(Q) with an additive group structure having O
as neutral element. Rational elliptic curves have a Weierstrass polynomial F with
integer coefficients. We study the reduction E p of E modulo the primes p ∈ Z. The
sequence (card E p ) p prime encodes important arithmetic properties of a rational el-
liptic curve.
Chapter 3 takes up the study of the modular group from Chapter 2. We ex-
tend the operation of Γ := SL(2, Z) on H to an operation on the vector space of
holomorphic functions on H. Those functions which are symmetric with respect
to this operation are named resp. modular forms and cusp forms. All of them de-
velop into a Fourier series around the point ∞. Due to their symmetry modular
forms correspond bijectively to meromorphic sections of the line bundle Ω 1 and
its tensor products defined on the compactified orbit space P1 (C). The complex
vector spaces of sections are finite dimensional, their dimension can be computed
by the theorem of Riemann-Roch. Taking the tensor product as multiplication pro-
vides the direct sum of these vector spaces with the structure of a finitely gener-
ated graded algebra forms M∗ (Γ ) = ⊕k Mk (Γ ). The grading is given by the weight
of the modular forms. The algebra of modular forms M∗ (Γ ) contains the ideal of
cusp forms S∗ (Γ ) = ⊕k Sk (Γ ). On M∗ (Γ ) acts the family of Hecke operators as an
Abelian algebra of endomorphisms. The ideal S∗ (Γ ) is stable with respect to this
action. Each component Sk (Γ ) has a basis of eigenforms with respect to the Hecke
algebra. For each eigenform f the family of eigenvalues reveals remarkable rela-
tions from the arithmetic of the Fourier coefficients of f . These relations make up
the magic of modular forms.
Chapter 1
Elliptic functions

The chapter starts with complex analysis in the plane. Subsequently the results are
translated into the language of Riemann surface. Here they are combined with some
results from algebraic topology. A final section applies the Riemann-Roch theorem
to prove Abel’s theorem about divisors on complex tori.

1.1 The field of elliptic functions

The present section deals with complex analysis in the plane, i.e. we study holomor-
phic and meromorphic functions on domains in the complex plane C.

The trigonometric functions sin and cos are examples of periodic holomorphic
functions, while tan and cot are periodic meromorphic. All periods of these func-
tions are integer multiples of respectively 2π and π. Elliptic functions are meromor-
phic with period group isomorphic to Z ⊕ Z.

Recall that a meromorphic function on a domain G ⊂ C is a holomorphic function

f : G\S → C

with a discrete closed subset S ⊂ G of poles of f - but no essential singularities.

Definition 1.1 (Period of a meromorphic function). Consider a meromorphic

function f on a domain G. A number ω ∈ C is a period of f if
• for all z ∈ G also z ± ω ∈ G

• and for all z ∈ G \ P

f (z + ω) = f (z).

9
10 1 Elliptic functions

Only seemingly Definition 1.1 excludes the poles from the definition of period-
icity. A pole z0 of f is an isolated singularity of f . Therefore the function f expands
into a convergent Laurent series around z0
∞
f (z) = ∑ an · (z − z0 )n
n≥n0

with coefficients
1 Z
f (z)
an = dz for suitable ε > 0.
2πi (z − z0 )n+1
|z−z0 |=ε

If f has the period ω then the Laurent expansions of f around z0 and z0 + ω are the
same because the integrands attain the same value at z and z + ω. Hence periodicity
of f includes also the poles.

Example 1.2 (Periodic meromorphic functions).

1. The functions sin, cos : C → C are holomorphic with period ω = 2π.

2. The tangent-function is meromorphic on C with singularities (π/2) + kπ, k ∈ Z,

and period ω = π.

3. The exponential function C → C∗ , z 7→ e2πiz , is holomorphic with period ω = 1.

If a meromorphic function f has a period ω 6= 0 one can ask for the set of all
periods. Apparently elements of the form kω, k ∈ Z are also periods of f . How
many independent periods of f exist?

Definition 1.3 (Discrete subgroup). A subgroup Γ ⊂ (C, +) is a discrete sub-

group, if the subspace topology of Γ is discrete, i.e. any g ∈ Γ has an open neigh-
bourhood U ⊂ C with
U ∩ Γ = {g}.

Proposition 1.4 (Period group). Consider a non-constant meromorphic function f

on a domain G ⊂ C. The set Γf of all periods of f is a discrete subgroup of C. It is
named the period group of f .

Proof. i) Apparently Γf is a subgroup.

ii) If Γf were not discrete then a convergent sequence (ωn )n∈N of pairwise distinct
periods ωn ∈ Γf would exist.
1.1 The field of elliptic functions 11

Hence for an arbitrary but fixed z0 ∈ G, which is not a pole of f , and for all n ∈ N

f (z0 ) = f (z0 + ωn ).

The identity theorem implies that f is constant, a contradiction, q.e.d.

Proposition 1.5 (Discrete subgroups of (C, +)). Each discrete subgroup

Γ ⊂ (C, +)

belongs to one of the following types:

1. Rank = 0:
Γ = {0}
2. Rank = 1:
Γ = Zω = {m · ω : m ∈ Z}
with a suitable element ω ∈ C∗ .

3. Rank = 2:

Γ = Zω1 + Zω2 = {m1 · ω1 + m2 · ω2 : m1 , m2 ∈ Z}

with ω1 , ω2 ∈ C linearly independent over the base field R. The subgroup Γ is

named a lattice.

For a proof see [1, Chap. 7, Sect. 2.1].

Note. If we specify a lattice in the form

Λ = Zω1 + Zω2
we will assume that the basis (ω1 , ω2 ) is positively oriented, i.e.
 
Re ω1 Re ω2
det > 0.
Im ω1 Im ω2

Definition 1.6 (Elliptic function). Consider a non-constant meromorphic function f

on a domain G ⊂ C with period group Γf .
• If rank Γf ≥ 1 then f is named periodic.
12 1 Elliptic functions

• If rank Γf = 2 then f is named doubly periodic. Its period group

Γf = Zω1 + Zω2

is also named the period lattice of f , see Figure 1.1. The subset

P = {λ1 ω1 + λ2 ω2 : 0 ≤ λ1 , λ2 < 1}

is the fundamental period parallelogram of f with respect to the basis (ω1 , ω2 ).

A doubly periodic meromorphic function on C is an elliptic function with

respect to a lattice Λ if
Λ ⊂ Γf
i.e. if all lattice points ω ∈ Λ are periods of f .

Fig. 1.1 Period lattice Λ

Which elliptic functions do exist? Proposition 1.7 shows that it is not interesting
to study the subclass of holomorphic elliptic functions. Therefore we will focus on
1.1 The field of elliptic functions 13

meromorphic functions with poles. See Proposition 1.8, and its corollaries about the
value attainment of elliptic functions.

Proposition 1.7 (Constant elliptic functions). Any holomorphic elliptic function

f :C→C

is constant.

Proof. An entire function f is holomorphic and attends the maximum of its modulus
on its closed period parallelogram which is a compact set. Being a bounded entire
function, f is constant according to Liouville’s theorem, q.e.d.

Proposition 1.8 (Residue theorem for elliptic functions). An elliptic function f

with period parallelogram P satisfies

∑ resζ ( f ) = 0.
ζ ∈P

Proof. The boundary ∂ P comprises only finitely many poles of f . Therefore we can
choose a number a ∈ C such the boundary of the translated period parallelogram

Pa := a + P

contains no poles of f . The residue theorem implies

1 Z
f (z) dz = ∑ resζ ( f ) = ∑ resζ ( f ).
2πi ∂ Pa ζ ∈P ζ ∈P
a

The integrand on the left-hand side is doubly periodic. Therefore the integration
along opposite sides of the period parallelogram cancels and the whole integral van-
ishes, q.e.d.

Corollary 1.9 (Poles and zeros of elliptic functions).

1. There are no elliptic functions f with one pole of order = 1 modulo the period
lattice Γf and no other pole mod Γf .

2. There are no elliptic functions f with only one zero of order = 1 modulo the
period lattice Γf and no other zero mod Γf .

Proof. 1. A pole of order 1 at a ∈ C has resa ( f ) 6= 0, which contradicts Proposition 1.8.

2. Apply part 1) to the elliptic function 1/ f , q.e.d.

14 1 Elliptic functions

Corollary 1.10 (Counting poles and zeros of elliptic functions). Any non-constant
elliptic function f attains modulo each lattice

Λ ⊂ Γf

all values a ∈ C ∪ {∞} with the same multiplicity. In particular, f has mod Λ the
same number of poles and zeros taken with multiplicity.
Proof. The case a 6= ∞ reduces to the second claim by considering the function f − a.
It has the same poles as f . Therefore it suffices to show that f has the same number
of zeros and poles, taken with multiplicity. Moreover it suffices to prove the theorem
for Λ = Γf .

We apply Proposition 1.8 to the meromorphic function f 0 / f and obtain

!
f0
∑ resζ f = 0.
ζ ∈P

Each residue evaluates to

! (
f0 k f has a zero of order k at ζ
resζ =
f −k f has a pole of order k at ζ

q.e.d.

1.2 The Weierstrass ℘-function

In the present section we consider an arbitrary but fixed lattice

Λ := Z ω1 + Z ω2 .

We use the notation Λ 0 := Λ \ {0}. Moreover P denotes the fundamental period

parallelogram of Λ .

The ℘-function of Λ is a distinguished elliptic function: All elliptic functions of

the given lattice Λ derive from ℘ and its derivative ℘ 0 .

Lemma 1.11 (Lattice constants). For k > 2 the infinite series

1
GΛ ,k := ∑ 0 ωk
ω∈Λ

is absolutely convergent. Its value is named the lattice constant of Λ .

1.2 The Weierstrass ℘-function 15

Proof. We choose the exhaustion of Λ by the sequence (Λn )n∈N of disjoint sets of
indices of increasing modulus

Λn := {µ · ω1 + ν · ω2 ∈ Λ : µ, ν ∈ Z; |µ|, |ν| ≤ n and (|µ| = n or |ν| = n)},

i.e.
˙
[
Λ= Λ .
n∈N n
Then
card Λn = 8n.
A suitable constant c exists with |ω| ≥ c · n for all n ∈ N and for all ω ∈ Λn . There-
fore !
1 8·n 8 1
∑ |ω|k ≤ ck · nk ≤ ck · nk−1 .
ω∈Λn

For k > 2
∞ 1
∑ nk−1 < ∞.
n=1

Therefore the claim follows from

!
1 ∞ 1
∑ = ∑ ∑ , q.e.d.
ω∈Λ 0
ω k n=1 ω∈Λn ωk

We will study elliptic functions with period lattice Λ . According to Corollary 1.9
a candidate f for the most simple example would have only one pole mod Λ , the
pole having order = 2 and residue = 0. The Laurent expansion of f around z0 = 0
would start
1
f (z) = 2 + a1 · z + O(2).
z
Being doubly periodic, f would have poles exactly at the points ω ∈ Λ . The follow-
ing Theorem 1.12 shows that such a function actually exists.

Theorem 1.12 (Weierstrass ℘-function). For each lattice Λ the series

!
1 1 1
℘(z) := 2 + ∑ −
z ω∈Λ 0 (z − ω)2 ω 2

is absolutely and compactly convergent for all z ∈ C \ Λ . It defines an even elliptic

function ℘ with respect to Λ , named the Weierstrass ℘-function of Λ . The pole set
of ℘ is Λ , each pole has order = 2.
If we consider only the first summand with fixed z then

1 1
∼ .
(z − ω)2 ω 2
16 1 Elliptic functions

Therefore the series with these summands alone does not converge. We will show
that the difference
1 1
2
− 2
(z − ω) ω
is proportional to
1
.
ω3
Therefore the series ℘(z) converges.

Note: TeX reserves the separate symbol “backslash wp” to denote the Weierstrass
function ℘.

Proof. We recall the definition of compact convergence of a sequence ( fν )ν∈N of

meromorphic functions on a domain G ⊂ C, cf. [62]: For each compact subset K ⊂ G
exists an index ν0 ∈ N such that
• for all ν > ν0 the function fν has no pole in K

• the sequence ( fν )ν>ν0 is uniformly convergent on K.

i) Meromorphic with pole set Λ : For each compact set K ⊂ C only finitely many
summands of ℘ have a pole in K. We show that the remaining series converges
absolutely and uniform on K: Choose an arbitrary but fixed R > 0. For

|z| ≤ R and 2R ≤ |ω|

holds
|ω|
− |z| ≥ 0.
2
The triangle inequality

|ω| |ω| |ω|

|z − ω| = |ω − z| ≥ ||ω| − |z|| = − |z| + ≥
2 2 2

implies the estimate

1 1 ω 2 − (z − ω)2 | − z2 + 2zω| |z|2 + 2 · |z| · |ω|

− = ≤ 2
≤ 4 · ≤
(z − ω)2 ω 2 ω 2 · (z − ω)2 ) |ω|4
!
|ω|
|ω|2 ·
2

|z|2 |z| 4 · R2 R R
≤ 4· + 8 · ≤ +8· = 10 ·
|ω|4 |ω|3 2 · R · |ω|3 |ω|3 |ω|3
independent from |z| < R.
1.2 The Weierstrass ℘-function 17

According to Lemma 1.11 the series

1
∑ 0 ω3
ω∈Λ

converges absolutely. For |z| ≤ R we decompose the series with respect to the sum-
mation over ω ∈ Λ
" !# " !#
1 1 1 1 1
+ − + −
z2 ω∈Λ 0∑
,|ω|<2R
(z − ω)2 ω 2 ∑
ω∈Λ ,|ω|≥2R
(z − ω)2 ω 2

The first summand is meromorphic with poles exactly at the points z ∈ Λ ∩ D2R (0).
The second summand converges absolutely and compactly on D2R (0) due to Lemma 1.11,
and defines a holomorphic function on D2R (0).

Because R can be choosen arbitrary, the series ℘ is compact convergent and de-
fines a meromorphic function on C with pole set Λ .

ii) Even function with periods from Λ :

• ℘ is even: It is admissible to rearrange the summation by replacing ω by −ω,
because the series is absolute convergent
! !
1 1 1 1 1 1
℘(−z) = 2 + ∑ − = 2+ ∑ − =
z ω∈Λ 0 (−z − ω)2 ω 2 z ω∈Λ 0 (−z − (−ω))2 (−ω)2
!
1 1 1
= 2+ ∑ − = ℘(z).
z ω∈Λ 0 (z − ω)2 ω 2
• ℘ has periods from Λ : The derivative is

−2 −2 1
℘ 0 (z) = 3
+ ∑ 3
= −2 ∑ 3
z ω∈Λ 0
(z − ω) ω∈Λ (z − ω)

Apparently for each fixed ω0 ∈ Λ

℘ 0 (z + ω0 ) = ℘ 0 (z)

after rearranging the summation by replacing ω by ω − ω0 . As a consequence

℘(z + ω j ) = ℘(z) + c j

with two suitable constants c j ∈ C, j = 1, 2.

For j = 1, 2 choosing the specific arguments

z := −ω j /2,
18 1 Elliptic functions

and using that ℘ is even implies

℘(ω j /2) = ℘((−ω j /2) + ω j ) = ℘(−ω j /2) + c j = ℘(ω j /2) + c j

Hence
c j = 0, q.e.d.

Lemma 1.13 (Half-periods of a lattice). Consider a lattice

Λ = Zω1 + Zω2

and denote by ( )
ω1 ω2 ω1 + ω2
Z(ω1 ,ω2 ) := , ,
2 2 2
the set of half-periods of its fundamental period parallelogram with respect to
(ω1 , ω2 ). The derivative ℘ 0 of the Weierstrass function ℘ of Λ has mod Λ
• a pole of order = 3 at 0 ∈ C and no other poles,

• and exactly three zeros, represented by the points of Z(ω1 ,ω2 ) .

Proof. Consider u ∈ Z(ω1 ,ω2 ) . According to Theorem 1.12 the function ℘ has no
pole at u. The derivative ℘ 0 has the same pole set as ℘. Because ℘ is even, the
function ℘ 0 is an odd elliptic function

℘ 0 (u) = −℘ 0 (−u) = −℘ 0 (−u + 2u) = −℘ 0 (u).

Therefore
℘ 0 (u) = 0,
i.e. each point of Z(ω1 ,ω2 ) is a zero of ℘ 0 within the fundamental period parallelo-
gram. Hence ℘ 0 has at least three zeros mod Λ .
Moreover ℘ 0 like ℘ has a single pole at 0 ∈ C mod Λ . The pole of ℘ 0 has
order = 3. Corollary 1.10 implies that ℘ 0 has exactly three zeros mod Λ . As a
consequence, the points of Z(ω1 ,ω2 ) represent exactly the zeros of ℘ 0 , and each zero
has order = 1, q.e.d.

Note that the determination of the two zeros mod Λ of℘is much more delicate [19].

Remark 1.14 (Weierstrass function ℘). Figure 1.2 shows the Laurent expansion
around 0 ∈ C of the Weierstrass function ℘ and its derivative ℘ 0 for the two
lattices Λ
1.2 The Weierstrass ℘-function 19

• with basis (1, ρ = e2πi/3 )

• and basis (1, i).

See PARI file “Weierstrass p function 08”. The numerical calculation confirms that
℘ is an even function and that ℘ 0 vanishes at the half-lattice points of Λ .

Fig. 1.2 Laurent expansion of ℘ and ℘ 0 and vanishing at half-lattice points

Theorem 1.12 shows that the ℘-function of the lattice Λ has all points from Λ as
periods. Corollary 1.15 shows the other direction: There are no additional periods,
i.e. Γ℘ = Λ .
Corollary 1.15 (Period lattice of ℘ and ℘ 0 ). The Weierstrass function ℘ of the
lattice Λ and its derivative ℘ 0 have the period lattice Λ , i.e.

Λ = Γ℘ = Γ℘ 0

i.e. the only periods of ℘ and of ℘ 0 are the points from Λ .

Proof. We know
Λ ⊂ Γ℘ ⊂ Γ℘ 0
20 1 Elliptic functions

Here the first inclusion has been shown in Theorem 1.12, while the second inclusion
is obvious. We show
Γ℘ 0 ⊂ Λ :
Let ω ∈ Γ℘ 0 be a period of ℘ 0 . Then

℘ 0 (ω/2) = ℘ 0 ((ω/2) − ω) = ℘ 0 (−ω/2) = −℘ 0 (ω/2).

Either ω/2 is not a pole of℘ 0 . Then ω/2 is a zero of℘ 0 and 1.13 implies ω/2 ∈ Λ /2
or ω ∈ Λ .
Or ω/2 is a pole of ℘ 0 . Then Theorem 1.12 implies ω/2 ∈ Λ , and in
particular ω ∈ Λ , q.e.d.

Proposition 1.16 (Coefficients of the Laurent expansion of ℘). The Weierstrass

function ℘ of Λ has the Laurent expansion around zero

1 ∞
℘(z) = + ∑ a2k · z2k
z2 k=1

with coefficients derived from the lattice constants of Λ

a2k = (2k + 1) · GΛ ,2k+2 .

Proof. We determine for ω 6= 0 the Taylor series of

1 1
−
(z − ω)2 ω 2

around 0 ∈ C: Taking the derivative

!
1 d −1
2
=
(z − ω) dz z−ω

and expanding into a geometric series

−1 1 1/ω ∞
= = = (1/ω) · ∑ (1/ω)ν · zν
z − ω ω − z 1 − (z/ω) ν=0

shows
1 ∞ ∞ 1
ν ν−1
= (1/ω) · ∑ ν · (1/ω) · z = ∑ (ν + 1) · ν+2 · zν
(z − ω)2 ν=1 ν=0 ω

and
1 1 ∞ 1
2
− 2 = ∑ (ν + 1) ν+2 · zν .
(z − ω) ω ν=1 ω
1.2 The Weierstrass ℘-function 21

Because ℘ is even and the series converges absolutely, see Theorem 1.12, we obtain
after rearrangement and using that ℘ is even
! !
1 1 1 1 ∞ 1
℘(z) = 2 + ∑ − = 2 + ∑ (ν + 1) ∑ ν+2 zν =
z ω∈Λ 0 (z − ω)2 ω 2 z ν=1 ω∈Λ 0
ω
!
1 ∞ 1
= 2 + ∑ (2k + 1) ∑ z2k , q.e.d.
z k=1 ω∈Λ 0
ω 2k+2

Theorem 1.17 (Differential equation of ℘). The Weierstrass function ℘ of the

lattice Λ satisfies the differential equation
2
℘ 0 = 4 ·℘3 − g2 ·℘− g3

with the constants

g2 := gΛ ,2 := 60 · GΛ ,4 , g3 := gΛ ,3 := 140 · GΛ ,6

derived from the lattice constants.

Proof. To simplify the notation we omit the index Λ from the lattice constants GΛ ,k .
Due to Proposition 1.16 the Laurent expansions start

1 1
℘(z) = + ∑ (2k + 1) · G2k+2 · z2k = 2 + 3 · G4 · z2 + 5 · G6 · z4 + O(6)
z2 k≥1 z

2
℘ 0 (z) = − + 6 · G4 · z + 20 · G6 · z3 + O(5)
z3
4 24 · G4
℘ 0 (z)2 = − − 80 · G6 + O(2)
z6 z2
1
℘(z)2 = + 6 · G4 + 10 · G6 · z2 + O(4)
z4
1 6 · G4 3 · G4 1 9 · G4
℘(z)3 = 6
+ 2 + 10 · G6 + 2 + 5 · G6 + O(2) = 6 + 2 + 15 · G6 + O(2)
z z z z z
Therefore all summands of order < 1 of the sum
2
℘ 0 − 4 ·℘3 + g2 ·℘+ g3

cancel and
2
℘ 0 − 4 ·℘3 + g2 ·℘+ g3 = O(2).
The left-hand side is an elliptic function. The right-hand side shows that the function
is even holomorphic and vanishes at 0 ∈ C. According to Proposition 1.7 the left-
hand side is constant, i.e. zero, q.e.d.
22 1 Elliptic functions

Theorem 1.18 (Field of elliptic functions). The field M (Λ ) of elliptic functions

with respect to the lattice Λ is a field extension of C of transcendence degree = 1,
more precisely
M (Λ ) = C(℘)[℘ 0 ]
with ℘ 0 algebraic over the field C(℘) with quadratic minimal polynomial

F(T ) = T 2 − 4 ·℘3 + g2 ·℘+ g3 ∈ C(℘)[T ]

with the lattice constants

g2 := 60 · GΛ ,4 , g3 := 140 · GΛ ,6 .

Note the equality C(℘)[℘ 0 ] = C(℘)(℘ 0 ) because℘ 0 is algebraic over C(℘).

Proof. i) Due to Theorem 1.17 we have

F(℘ 0 ) = 0 ∈ C(℘).

The derivative ℘ 0 is odd while ℘ is even. Therefore ℘ 0 ∈

/ C(℘), and the minimal
polynomial F must have degree ≥ 2.

ii) Consider an elliptic function f ∈ M (Λ ), f 6= 0. We show f ∈ C(℘, ℘ 0 ).

• We first consider the case that f has pole set Λ .

The unique pole of f mod Λ has order k ≥ 2 due to Corollary 1.9. We prove

f ∈ C(℘,℘ 0 )

by induction on the pole order k.

If k = 2 then for suitable c ∈ C the function

f − c ·℘
has at most a single pole mod Λ of order < 2. Therefore the function is holomor-
phic, and even constant due to Corollar 1.7.

For the induction step assume k > 2. On one hand, for even k = 2m a suitable
constant c ∈ C exists such that f − c ·℘ m has only poles of order < k. Therefore

f − c ·℘ m ∈ C(℘,℘ 0 )

by induction assumption. On the other hand, for odd k = 2m + 1 a suitable con-

stant c ∈ C exists such that

f − c ·℘ m−1 ·℘ 0
1.2 The Weierstrass ℘-function 23

has only poles of order < k. By induction assumption

f − c ·℘ m−1℘ 0 ∈ C(℘,℘ 0 ).

In both cases
f ∈ C(℘,℘ 0 ).
• Eventually, we consider the case of an arbitrary pole set of f .

Then f has mod Λ only finitely many poles zν ∈ P\Λ , ν = 1, ..., n. The function ℘
is holomorphic in P \ Λ . Hence for suitable constants k1 , ..., kn ∈ N the function
n
f (z) · ∏ (℘(z) −℘(zν ))kν
ν=1

has poles at most in Λ . According to the first part

n
f (z) · ∏ (℘(z) −℘(zν ))kν ∈ C(℘,℘ 0 ).
ν=1

Therefore
f ∈ C(℘,℘ 0 ), q.e.d.

Proposition 1.19 (Discriminant). For a given lattice

Λ = Zω1 + Zω2 ⊂ C

the cubic polynomial

GΛ (T ) := 4T 3 − g2 · T − g3 ∈ C[T ], g2 := 60 · GΛ ,4 , g3 := 140 · GΛ ,6

has three pairwise distinct zeros: The values ℘(ω) ∈ C at the half-period points
of Λ
ω ∈ ZΛ := {ω1 /2, ω2 /2, (ω1 + ω2 )/2}
As a consequence, the discriminant
1
discr(GΛ ) = · (g32 − 27 · g23 ) ∈ C
24
of the polynomial GΛ (T ) is non-zero, in particular

g32 − 27 · g23 6= 0.

Proof. 1. Pairwise distinct zeros: The differential equation of℘ 0 from Theorem 1.17
implies
℘ 0 (ω)2 = 4 ·℘(ω)3 − g2 ·℘(ω) − g3 = GΛ (℘(ω))
24 1 Elliptic functions

Due to Lemma 1.13

ω ∈ ZΛ =⇒ GΛ (℘(ω)) = 0
and the function ℘ assumes at the half-period ω ∈ ZΛ the value ℘(ω) with mul-
tiplicity at least = 2 because ℘ 0 (ω) = 0. In case

℘(ω) = ℘(ω 0 ) for ω 6= ω 0 ∈ ZΛ

the elliptic function℘would assume the value℘(ω) with multiplicity at least = 4.

We would obtain a contradiction to Corollary 1.10 that the elliptic function ℘ as-
sumes mod Λ each value with multiplicity = 2. Hence

℘(ω) 6= ℘(ω 0 ).

2. Discriminant: One checks the discriminant formula

1
discr(GΛ ) = · (g32 − 27 · g23 ) ∈ C,
24
e.g. see [33, Cor. 3.4]. By definition the discriminant of a polynomial is non-zero
iff the zeros of the polynomial are pairwise distinct, q.e.d.

1.3 Abel’s theorem

The zeros and poles of an elliptic functions cannot be prescribed in an arbitrary way:
Abel’s Theorem states a sufficient and necessary condition.

We continue fixing an arbitrary lattice

Λ := Z ω1 + Z ω2 ,
0
keeping the notations Λ := Λ \ {0} and P denoting the fundamental period paral-
lelogram of Λ .

Proposition 1.20 (Zero and pole divisor). Consider an elliptic function and denote
by A = (ai )i=1,...,n its family of zeros in P and by B = (bi )i=1,...,n its family of poles
in P. Then n ≥ 2 and
∑ ai − ∑ bi ∈ Λ .
i=1,...,n i=1,...,n

Note that each point p of A and B appears as often as its multiplicity as zero or pole
of f indicates. By Corollary 1.9 holds n ≥ 2, and by Corollary 1.10 both families A
and B have the same cardinality.
1.3 Abel’s theorem 25

Proof. We consider a translation

Pa := P + a

of the fundamental parallelogram such that f has neither zeros nor poles on the
boundary ∂ Pa . With a counter-clockwise orientation of ∂ Pa the residue theorem
gives !
1 Z f 0 (z) f 0 (z)
z· dz = ∑ res p z · .
2πi ∂ Pa f (z) p∈Pa f (z)

• Right-hand side: Assume ord( f ; p) = k ∈ Z∗ . Then

(
p is a zero of f of order k if k > 0
p is a pole of f of order k if k < 0

In a neighbourhood of p
f 0 (z) k
= + f1 (z)
f (z) z − p
with f1 holomorph. As a consequence

f 0 (z) f 0 (z) p·k

z· = (p + (z − p)) · = + f2 (z)
f (z) f (z) z − p

with f2 holomorph. Therefore

!
f 0 (z)
res p z· = p·k
f (z)

and !
f 0 (z) n n
∑ res p z· = ∑ ai − ∑ bi .
p∈Pa f (z) i=1 i=1

• Left-hand side: Denote by A, B,C, D the vertices of Pa , counter-clockwise ori-

ented.
Z
f 0 (z) Z
f 0 (z) Z 0
f (z)
z· dz = (z−(z+ω2 ))· dz = −ω2 dz = −ω2 ·[log f (z)]BA
f (z) f (z) f (z)
−
→ −→ −
→ −
→
AB+CD AB AB

Here log f denotes a branch of the logarithm in a simply-connected neighbour-

hood of the line AB.
Z
f 0 (z)
z· dz = ω2 · (log f (A) − log f (B)).
f (z)
−
→ −→
AB+CD

Because
26 1 Elliptic functions

f (A) = f (B)
an integer m2 ∈ Z exists with

log f (A) − log f (B) = 2πi · m2 .

−
→ −→
A similar argument for the lines BC and DA provides a second constant m1 ∈ Z,
such that finally

1 Z f 0 (z)
· z· dz = m1 · ω1 + m2 · ω2 ∈ Λ , q.e.d.
2πi f (z)
∂ Pa

Theorem 1.21 (Abel’s theorem). Consider n ≥ 2 and two finite families of complex
points
A := (ai )i=1,...,n , B := (bi )i=1,...,n ⊂ C
with A and B pairwise disjoint mod Λ . Then are equivalent:
• There exists an elliptic function f ∈ M (Λ ) which has mod Λ the zero set A and
the pole set B

• The points of A and B satisfy

n n
∑ ai − ∑ bi ∈ Λ .
i=1 i=1

Proof. Proposition 1.20 proves that the zero set and the pole set of f satisfy the con-
dition stated above. In order to prove the opposite direction of the theorem assume
two sets A and B with the property stated above. Without restriction we may assume
n
∑ (a j − b j ) = 0
j=1

as an equation, not just a congruence; otherwise replace a1 by a suitable point which

is congruent mod Λ . The proof for the existence of a suitable elliptic function f is
by induction.
Start of induction n = 2: If
a1 − b1 = b2 − a2
then consider
a1 + a2 + b1 + b2
z0 := .
4
Geometrically the point z0 is the center of the parallelogram with vertices the
points a1 , a2 , b1 , b2 . We may assume z0 = 0, otherwise we translate all coordinates
by −z0 . As a consequence, the two conditions state
1.3 Abel’s theorem 27

a1 = −a2 and b1 = −b2 .

Depending on the position relative Λ of a1 and b1 we now distinguish the following

cases for the choice of f :
• a1 , b1 ∈
/ Λ : The even function

g(z) := ℘(z) −℘(a1 )

has
– a simple zero at a1 and at a2 if a1 6≡ a2 mod Λ , and a zero of order = 2 at
a1 ≡ a2 otherwise

– and a pole of order = 2 at z = 0 .

An analogous result holds for the function

h(z) := ℘(z) −℘(b1 ).

Hence the quotient

g ℘−℘(a1 )
f := =
h ℘−℘(b1 )
has zeros exactly the zeros of g and poles exactly the zeros of h, while the poles
of g and h at the origin cancel.

• a1 ∈ Λ : Then b1 ∈
/ Λ . Set
1
f :=
℘−℘(b1 )
• b1 ∈ Λ : Then a1 ∈
/ Λ . Set
f := ℘−℘(a1 )
Hence f is elliptic with prescribed zeros and poles.

Induction step n ≥ 2, n 7→ n + 1: If mod Λ

(a1 − b1 ) + ... + (an − bn ) + (an+1 − bn+1 ) ≡ 0
then choose ãn ∈ C with

ãn ≡ (a1 − b1 ) + ... + (an−1 − bn−1 ) + an .

Then mod Λ
(a1 − b1 ) + ... + (an−1 − bn−1 ) + (an − a˜n ) ≡ 0
and
(ãn + an+1 ) − (bn + bn+1 ) ≡ 0.
Depending on the position of ãn relative Λ we distinguish the following cases:
28 1 Elliptic functions

• ãn 6≡ a j for all j = 1, ..., n: First we exclude the two exceptional cases ãn ≡ bn
and ãn ≡ bn+1 : They lead to respectively

an+1 ≡ bn+1 and an+1 ≡ bn

which has been excluded by assumption.

For the remaining general cases, by induction assumption two elliptic functions
exist
f1 ∈ M (Λ )
with zeros (a1 , ..., an ) and poles (b1 , ..., bn−1 , ãn ) and

f2 ∈ M (Λ )

with zeros (ãn , an+1 ) and poles (bn , bn+1 ).

• ãn ≡ an : Then
(a1 − b1 ) + ... + (an−1 − bn−1 ) ≡ 0
By induction assumption two elliptic functions exist

f1 ∈ M (Λ )

with zeros (a1 , ..., an−1 ) and poles (b1 , ..., bn−1 ) and

f2 ∈ M (Λ )

with zeros (an , an+1 ) and poles (bn , bn+1 ).

• ãn 6≡ an , ãn ≡ a j for at least one j = 1, ..., n − 1, w.l.o.g. ãn ≡ a1 : Then

(an − b1 )(a2 − b2 ) + ... + (an−1 − bn−1 ) ≡ 0

and
(a1 + an+1 ) − (bn + bn+1 ) ≡ 0.
Similarly to the previous case by induction assumption two elliptic functions
exist
f1 ∈ M (Λ )
with zero set (a2 , ..., an ) and pole set (b1 , ..., bn−1 ) and

f2 ∈ M (Λ )

with zero (a1 , an+1 ) and poles (bn , bn+1 ).

In all cases the elliptic function

f := f1 · f2 ∈ M (Λ )

has the prescribed zeros (a1 , ..., an+1 ) and poles (b1 , ..., bn+1 ), q.e.d.
1.3 Abel’s theorem 29

Corollary 1.22 (Elliptic functions with pole of order = 3). An elliptic function f ∈ M (Λ )
with a pole mod Λ at 0 ∈ Λ of order = 3 and no other poles mod Λ , has exactly
three zeros a1 , a2 , a3 mod Λ . They satisfy

a1 + a2 + a3 = 0 mod Λ .

Corollary 1.22 applies to f = ℘ 0 , see Lemma 1.13.

The lenghty case-by-case analysis in the proof of Theorem 1.21 indicates: To

define elliptic functions as functions in the complex plane with a certain priodicity
is not the best context. Therefore we will consider in Section 2.1 elliptic functions
simply as meromorphic functions on the torus. The torus is a compact Riemann
surface. Abel’s theorem generalizes to all compact Riemann surfaces. Phrasing and
proving the theorem uses advanced results from the theory of Riemann surfaces,
see [21, §20].
Chapter 2
The modular group Γ and its Hecke congruence
subgroups Γ0 (N)

2.1 The moduli space of complex tori and group actions

An elliptic function f with lattice Λ , as studied in Chapter 1, has as natural domain

of definition not the complex plane C. Due to its periodicity the natural domain of
definition for f is the quotient C/Λ , a complex torus: Elliptic functions are mero-
morphic functions on the torus.
Therefore the present section investigates complex analysis on complex tori. We
will prove that the equivalence classes of complex tori up to biholomorphic isomor-
phisms depend on one complex parameter. The parameter space can be choosen as
quotient of the upper half-plane

H := {τ ∈ C : Im τ > 0}

with respect to the action of SL(2, Z) as group of fractional-linear transformations.

The basis of a lattice

Λ = Zω1 + Zω2
is not uniquely determined. Two positively oriented bases generate the same lattice
if and only if they are equivalent modulo SL(2, Z), i.e. iff they belong to the same
orbit of the following SL(2, Z)-action:

Lemma 2.1 (Different bases of a given lattice). Denote by

   
ω1 ω2
M := (ω1 , ω2 ) basis o f R2 with det >0
| |

the set of positively oriented bases of R2 . The set R of all lattices Λ ⊂ C is the set
of (left) equivalence classes of M

SL(2, Z)\M

31
32 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

with respect to the equivalence relation induced by the map

    0   
ω1 ω1 ω1
SL(2, Z) × M → M, γ, 7→ := γ · .
ω2 ω20 ω2

Proof. The invertible matrices γ ∈ M(2 × 2, Z) have determinant ±1. Because both
bases are positively oriented we have det γ = 1, q.e.d.

Two tori belonging to different lattices may be biholomorphically equivalent.

The following Definition 2.2 introduces a name for the relation between lattices
with isomorphic tori.
Definition 2.2 (Similar lattices). Two lattices Λ , Λ 0 ⊂ C are similar if a complex
number a ∈ C∗ exists with
Λ0 = a·Λ.

Proposition 2.3 (Holomorphic maps between tori). Consider a holomorphic map

F : C/Λ1 → C/Λ2

between two tori with F(0) = 0. Then a uniquely determined number a ∈ C exists
such that the following diagram commutes

µa
C C
p1 p2
F
C/Λ1 C/Λ2

Here µa denotes the multiplication by a and pi , i = 1, 2, denote the canonical pro-

jections.
In particular, any holomorphic map between tori which fixes the neutral element is
a group homomorphism.
Note: Due to the commutativity of the diagram in Proposition 2.3 the multiplier a ∈ C
satisfies
a · Λ1 ⊂ Λ2 .
Proof. The composition
F ◦ p1 : C →
− C/Λ2
starts from the simply-connected domain C. Therefore it lifts to the total space C of
the covering projection p2 . It provides a holomorphic map

F̃ : C →
− C,
2.1 The moduli space of complex tori and group actions 33

uniquely determined by the condition F̃(0) = 0. Comutativity of the lifting diagram

shows: For all z ∈ C, ω ∈ Λ1 ,

F̃(z + ω) − F̃(z) ∈ Λ2 .

The lattice Λ2 is discrete and F̃ is continous. Therefore

F̃(z + ω) − F̃(z)

does not depend on z. As a consequence for all z ∈ C

F̃ 0 (z + ω) − F̃ 0 (z) = 0.

Hence F̃ 0 is a holomorphic and elliptic. Therefore F̃ 0 is a constant a ∈ C, and for

all z ∈ C
F̃(z) = a · z
The uniqueness of a follows from the fact that µa induces the zero map F iff for
all z ∈ C holds a · z ∈ Λ2 . Because Λ2 ⊂ C is discrete and a · 0 = 0 we have a · z = 0
for all z ∈ C, which implies a = 0, q.e.d.

Corollary 2.4 (Lattices of biholomorphically equivalent tori). Two tori

T1 = C/Λ1 and T2 = C/Λ2

are biholomorphically equivalent iff their lattices Λ1 and Λ2 are similar.

Proof. i) Assume the existence of a biholomorphic map

F : T1 → T2

satisfying without restriction F(0) = 0. Proposition 2.3 provides a commutative di-

agram
µa
C C
p1 p2
F
C/Λ1 C/Λ2

with µa the multiplication by a complex number a ∈ C, satisfying

µa · Λ1 ⊂ Λ2 .

Analogously, there exists a complex number b ∈ C with µb the unique lift of F −1

fixing the origin. Then the map
µb ◦ µa = µb·a
34 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

is the unique lift of

idC/Λ1 = F −1 ◦ F
fixing the origin. Hence
b · a = 1.
In particular
a, b ∈ C∗
and
a · Λ1 = Λ2 .
ii) Assume
Λ2 = a · Λ1
with a ∈ C∗ . Then the multiplication
µa : C → C,
induces a biholomorphic map F : T1 → T2 , q.e.d.

Theorem 2.5 (Biholomorphic equivalence classes of complex tori).

1. Any complex torus is biholomorphic equivalent to a torus T = C/Λ with a nor-
malized lattice
Λ = Z·1+Z·τ ⊂ C
with τ ∈ H in the upper half-plane.

2. The period τ of a normalized lattice is determined up to a fractional linear trans-

formation with integer coefficients. Hence two tori

C/Λτ and C/Λτ 0

with normalized lattices

Λτ = Z · 1 + Z · τ and Λτ 0 = Z · 1 + Z · τ 0

are biholomorphic equivalent iff

a τ +b
τ0 =
c τ +d
for a matrix  
ab
∈ SL(2, Z).
cd

Proof. 1. Consider an arbitrary torus T = C/Λ with a lattice satisfying

Λ = Zω1 + Zω2
2.1 The moduli space of complex tori and group actions 35

and det (ω1 ω2 ) > 0. Then

ω2
τ :=
ω1
satisfies Im τ > 0 and the lattice Λ is similar to the normalized lattice

Λτ := Z · 1 + Z · τ

due to
1
· Λ = Λτ .
ω1
According to Corollary 2.4 the two tori T and T 0 := C/Λτ are biholomorphic
equivalent.

2. i) Assume  
a τ +b ab
τ0 = γ · τ = with γ := ∈ SL(2, Z).
c τ +d cd
Set    
ω2 τ
:= γ · , i.e. ω2 = aτ + b, ω1 = cτ + d.
ω1 1
The base change implies

Z · ω1 + Z · ω2 = Z · 1 + Z · τ = Λτ .

As a consequence
1
· Λτ = Λτ 0
ω1
because
ω2
τ0 = .
ω1
The lattices Λτ and Λτ 0 are similar. Hence the tori C/Λτ and C/Λτ 0 are biholo-
morphic equivalent according to Corollary 2.4.

ii) For the opposite direction assume that two tori with normalized lattices

C/Λτ and C/Λτ 0 , τ, τ 0 ∈ H,

are biholomorphic equivalent. According to Corollary 2.4 a number α ∈ C∗ exists

with
Λτ 0 = α · Λτ .
As a consequence
Λτ 0 = Z · α + Z · α · τ.
Therefore a matrix  
ab
γ= ∈ SL(2, Z)
cd
36 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

exists with  0  
τ ατ
=γ· ,
1 α
i. e.
a · ατ + b · α a · τ + b
τ0 = = , q.e.d.
c · ατ + d · α c · τ + d

Theorem 2.5 shows the consequences when two points of H are related by an
element of the group SL(2, Z). We formalize this relation by the concept of a group
acting on a topological space.

Definition 2.6 (Group action).

1. Consider a topological group G with neutral element e ∈ G and a topological
space X. A left action of G on X is a continuous map

φ : G×X →
− X, (g, x) 7→ gx,

also written g(x) := gx, satisfying the following properties: For all g1 , g2 ∈ G and
for all x ∈ X
• ex = x

• g1 (g2 x) = (g1 g2 )x

2. Consider a group action G on X.

• The action is faithful if for all g ∈ G, g 6= e, exists x ∈ X with gx 6= x, i.e. no
group element different from e acts in a trivial way.

• A fixed point of the group action is a point x ∈ X with gx = x for all g ∈ G.

The action is fixed-point-free if it has no fixed points.

• The isotropy group of a point x ∈ X is the subgroup

Gx := {g ∈ G : gx = x} ⊂ G.

• The orbit of x ∈ X is the subspace

{gx ∈ X : g ∈ G} ⊂ X

Two points of X are equivalent with respect to the group action if they belong
to the same orbit. The set of equivalence classes is the orbit space G\X, the
set of orbits of the action. The canonical map to the orbit set is denoted

π :X →
− G\X, x 7→ [x].
2.1 The moduli space of complex tori and group actions 37

• A fundamental domain F of the group action is a subset F ⊂ X such that the

restriction
π|F : F →− G\X
is bijective. Depending on eventual additional properties of the group action
one looks for fundamental domains with specific additional properties.

• The action is transitive if the whole set X is a single orbit, i.e. if for any two
points x1 , x2 ∈ X exists g ∈ G with x2 = gx1 .

Apparently all points of the same orbit have conjugated isotropy groups because
Ggx = g · Gx · g−1 .

Theorem 2.5 deals with the specific group action from Definition 2.7.

Definition 2.7 (Action of SL(2, Z) on H). The SL(2, Z)-left action on H is the map
 
aτ + b ab
Φ : SL(2, Z) × H → H, (γ, τ) 7→ γ(τ) := , γ := .
cτ + d cd

Remark 2.8 (Action of SL(2, Z) and extensions).

1. Imaginary part: If τ ∈ H and
 
ab
γ= ∈ GL(2, R)+ := {A ∈ GL(2, C) : det A > 0}
cd

then
Im τ
Im γ(τ) = det(γ) ·
|cτ + d|2
As a consequence
τ ∈ H =⇒ γ(τ) ∈ H.
Hence the map from Definition 2.7 extends to an action

aτ + b
GL(2, R)+ × H →
− H, γ(τ) :=
cτ + d
which apparently restricts to an action of all subgroups of GL(2, R)+ , in particu-
lar to Γ ⊂ GL(2, R)+ .
38 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Due to Theorem 2.5 the biholomorphic equivalence classes of complex tori cor-
respond bijectively to the points of the orbit space SL(2, Z)\H. Hence the orbit
space is named the moduli space of 1-dimensional complex tori.

Note: The orbit space is read “H modulo SL(2, Z)”.

2. Action of Γ on H∗ : We have

z ∈ R =⇒ γ(z) ∈ R ∪ {∞} and z ∈ Q =⇒ γ(z) ∈ Q ∪ {∞}

Moreover, due to
az + b a + (b/z)
=
cz + d c + (d/z)
we can extend γ to the argument ∞ by setting


∞ if d 6= 0 and c = 0

γ(∞) :=
a


otherwise
c
Set
H∗ := H ∪
˙ Q∪
˙ {∞}
Then the extension

Φ : SL(2, Z) × H∗ →
− H∗ , (γ, z) 7→ γ(z),

is well-defined.

Note. The proof of Theorem 2.28 will provide H∗ with a specific topology which
induces on Q ⊂ H a topology different from the subspace topology of Q ⊂ C and
also different neighbourhoods of ∞ ∈ H∗ .

Definition 2.9 (Modular group, Hecke conguence subgroups and cusps).

1. The modular group is the group

Γ := SL(2, Z)

2. For a positive integer N ∈ N∗ the Hecke congruence subgroup of level N is the

subgroup   
a b
Γ0 (N) := ∈ Γ : c ≡ 0 mod N ⊂ Γ
c d
The canonical left action of Γ from Remark 2.8 on H∗ restricts to a left action

ΦN : Γ0 (N) × H∗ → H∗ .
2.1 The moduli space of complex tori and group actions 39

3. Points τ ∈ H with non-trivial isotropy group, i.e. Γτ ) {±id}, are named elliptic
points of Γ0 (N). Half the order of the isotropy group

ord Γ0 (N)τ
hτ :=
2
is named the period of τ. An orbit is named an elliptic orbit if at least one point
of the orbit - and hence all points - are elliptic.

4. The orbits of the points τ ∈ Q ∪ {∞} are named the cusps of Γ0 (N). The set
of cusps is denote by cusp(Γ0 (N)). The width (Deutsch: “Breite”) of the cusp
passing through
τ ∈ Q ∪ {∞}
is the index of the isotropy groups

hτ := [Γτ : Γ0 (N)τ ]

Remark 2.10 (Modular group and congruence subgroups).

1. Apparently,
Γ = Γ0 (1)
2. For all N ≥ 1 the element
−id ∈ Γ0 (N)
acts trivially on H, i.e. for all τ ∈ H

−id ∈ Γ0 (N)τ

The elements
±id ∈ Γ0 (N)
are the only elements which act trivially on H. Therefore some authors apply the
name modular group to the quotient

PSL(2, Z) := SL(2, Z)/{±id}.

We will not follow this convention. Because one can study also the action of
certain subgroups of SL(2, Z) which do not contain the element −id.

3. Corollary 2.18 will show that the isotropy groups of all elliptic points from Def-
inition 2.9 are finite groups. But the isotropy group Γ∞ ⊂ Γ is an infinite group,
more specific
 
11
Γ∞ =< ±T > with translation T := ∈Γ
01
40 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

4. Theorem 2.28 will show the importance to consider the group action not only on
the open upper half-plane H but also on its “closure” H∗ .

Example 2.11 (Cusps of Γ , Γ0 (2), Γ0 (4) and Γ0 (11)). See [17, Sect. 3.8], and the
PARI commands mfcusps, mfnumcusps, mfcuspwidth.
1. For N ∈ N the number ε∞ (Γ0 (N)) of cusps is

∑ φ (gcd(d, N/d))
d|N

with the sum extending over all positive divisors d of N, and φ denoting the Euler
function defined as
Φ(n) := card (Z/nZ)∗
2. The group Γ has a single cusp
• Orbit of the point 0:
Q ∪ {∞}
with width = 1. All rationals p/q ∈ Q belong to the Γ -orbit of ∞, see
Remark 2.8.

3. The group Γ0 (2) has two cusps

• Orbit of the point 0:
( )
p
∈ Q : (p, q) = 1, q odd
q

with width = 2.

• Orbit of the point 1/2:

( )
p
∈ Q : (p, q) = 1, p 6= 0, q even ∪ {∞}
q

with width = 1.

4. The group Γ0 (4) has three cusps

• Orbit of the point 0:
( )
p
∈ Q : (p, q) = 1, q odd
q
2.1 The moduli space of complex tori and group actions 41

with width = 4.

• Orbit of the point 1/2:

( )
p
∈ Q : (p, q) = 1, p 6= 0, q even, not divisible by 4
q

with width = 1.

• Orbit of the point 1/4:

( )
p
∈ Q : (p, q) = 1, p 6= 0, q divisible by 4 ∪ {∞}
q

with width = 1.

5. The group Γ0 (11) has two cusps

• Orbit of the point 0:
( )
p
∈ Q : (p, q) = 1, q not divisible by 11
q

with width = 11.

• Orbit of the point 1/11:

( )
p
∈ Q : (p, q) = 1, p 6= 0, q divisible by 11 ∪ {∞}
q

with with width = 1.

Lemma 2.12 (Divisor lemma). Consider three integers a, b, c ∈ Z with

gcd(a, b, c) = 1.

Then exists an integer r ∈ Z such that

gcd(a + r · b, c) = 1.

Proof. E.g. [36, Kap. II, §3 Lemma]: We claim that

r := ∏ p
p|c,p-a
42 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

satisfies the claim. Otherwise assume the existence of a prime q satisfying

q | gcd(a + r · b, c).

• Case q | a: Then q | rb. Because q - r then q | b. Therefore

gcd(a, b, c, ) 6= 1,

a contradiction.

• Case q - a: Then q | r by definition of r. Because q | a + rb also q | a, a contradic-

tion, q.e.d.

Lemma 2.13 (Exact sequence of congruence subgroups). Consider N ∈ N∗ . There

exists a canonical exact sequence of Abelian groups
π
1→
− Γ (N) →
− Γ−
→ SL(2, Z/NZ) →
− 1

with the principal congruence subroup of level N

   
10
Γ (N) := A ∈ Γ : A ≡ mod N
01

and
SL(2, Z/NZ) := {B ∈ M(2 × 2, Z/NZ) : det B = 1 ∈ (Z/NZ)∗ }
and the residue morphism

π :Γ →
− SL(2, Z/NZ), A 7→ A.

Proof. i) kernel: Apparently ker π = Γ (N).

ii) Surjectivity: E.g.see [36, Kap. II, §3 Satz]. Assume a given
 
α β
A=  ∈ SL(2, Z/NZ)
γ δ
We may assume  
αβ
A= ∈ M(2 × 2, Z),
γ δ
and w.l.o.g γ 6= 0, otherwise set γ = N. The condition about the determinant

det A = 1 ∈ Z/NZ

implies
αδ − β γ ≡ 1 mod N,
2.1 The moduli space of complex tori and group actions 43

and therefore
gcd(γ, δ , N) = 1.
Lemma 2.12 provides an integer r ∈ Z satisfying

gcd(γ, d) = 1

for
d := δ + rN.
On one hand, from the determinant

αd − β γ = αδ − β γ + αrN = 1 + sN

with a suitable integer s ∈ Z. On the other hand

gcd(γ, d) = 1

provides an equation
γy − dx = s
with suitable integers x, y ∈ Z.

Consider the matrix

 
α + xN β + yN
à := ∈ M(2 × 2, Z)
γ d
Then
π(Ã) = π(A) = A
and
det à = (α + xN) · d − (β + yN) · γ = 1 + sN + N(xd − yγ) = 1,
which implies à ∈ Γ and finishes the proof of the Lemma, q.e.d.

Proposition 2.14 (Index of Hecke congruence subgroups). For any integer N ≥ 1

holds the index formula
!
1
[Γ : Γ0 (N)] = N · ∏ 1 +
p|N
p

Here the product ranges over all prime factors p of N.

In particular, all Hecke congruence subgroups have finite index in the modular
group.
Proof. See also [53, Chap. 1.6].
44 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

i) Transitivity of the index formula: We extend the exact sequence from

Lemma 2.13 to the commutative diagram
π
1 Γ (N) Γ SL(2, Z/NZ) 1

id ψ

1 Γ (N) Γ0 (N) Γ0 (N)/Γ (N) 1

The induced map

ψ : Γ0 (N)/Γ (N) →
− SL(2, Z/NZ)
is injective with image
 
α β
im ψ = ⊂ SL(2, Z/NZ)
0 α −1

Apparently
[Γ0 (N) : Γ (N)] = card(im ψ) = φ (N) · N
with the Euler function φ . The exact sequence of the first line of the diagram implies

[Γ : Γ (N)] = card SL(2, Z/NZ).

The transititvity of the index function implies

[Γ : Γ0 (N)] · [Γ0 (N) : Γ (N)] = [Γ : Γ (N)]

ii) Factorisation: The prime factorisation of integers

k
e
N = ∏ pjj
j=1

implies the factorisation of the Euler function

!
k 1
φ (N) = ∏ pe j · 1 −
j=1 p j

with the Euler factors !

e e 1
φ (p ) = p · 1 − ,
p

and the factorisation of the SL(2, −)-groups due to the Chinese remainder theorem
k  
e
SL(2, Z/NZ) ' ∏ SL 2, Z/p j j Z .
j=1

One has
2.1 The moduli space of complex tori and group actions 45
!

e

3e 1
card SL 2, Z/p j j Z = p j j 1− 2 ,
pj
see [53, Chap. 1.6] and [17, Exerc. 1.2.3]. For a proof see also [28, Cor. 2.8] and use
for the reduction from GL to SL the exact sequence of the determinant
det
1→ − GL (2, Z/pe Z) −→ (Z/pe Z)∗ →
− SL (2, Z/pe Z) → − 1.

And also
[Γ : Γ (N)] = card SL(2, Z/NZ),
see part i). It follows
k  
e
[Γ : Γ (N)] = ∏ card SL 2, Z/p j j Z =
j=1
! !
k 1 k 1
3e 2 ej
=∏ pj j
1− 2 = N · ∏ pj · 1− 2 =
j=1 pj j=1 pj
! ! !
k 1 1 k 1
e
= N2 · ∏ p j · 1 −
j
1+ = N 2 · φ (N) · ∏ 1 +
j=1 pj pj j=1 pj
Hence
[Γ : Γ (N)]
[Γ : Γ0 (N)] = =
[Γ0 (N) : Γ (N)]
!
2 k 1
N · φ (N) · ∏ j=1 1 + !
pj 1
= = N ·∏ 1+ , q.e.d.
N · φ (N) p|N
p

Example 2.15 (Right cosets of congruence subgroups). See the PARI-command

mfcosets.
1. The subgroup Γ0 (2) has index

[Γ : Γ0 (2)] = 3

with the set of right coset representatives

    
0 −1 2 0 −1
id, J := ,U =
1 0 1 −1

for  
−1 1
U :=
−1 0
46 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

2. The subgroup Γ0 (4) has index

[Γ : Γ0 (4)] = 6

with the set of right coset representatives

           
10 0 −1 10 0 −1 0 −1 10
, , , , , .
01 1 0 11 1 2 1 3 21

2.2 Topology of the orbit space of the Γ -action

The section continues the study of the moduli space of complex tori. We denote by
 
aτ + b ab
Φ : Γ × H → H, (γ, τ) 7→ γ(τ) := , γ :=
cτ + d cd
the continuous Γ -left action introduced in Remark 2.8. Our final aim is to provide
the orbit space
Y := Γ \H
with an analytic structure and to compactify the resulting Riemann surface to obtain
a compact Riemann surface X. The latter will be named the modular curve of the
action Φ. We proceed along the following steps:
• Constructing a Hausdorf topology on Y , see Theorem 2.21.

• Constructing an analytic structure on Y , see Section 2.3 Proposition 2.27.

• Compactifying Y to a compact Riemann surface X, see Section 2.3 Theorem

2.28.

To provide the orbit space of a group action with a suitable topological or differ-
entiable structure is always a subtle task. In the present context the problem is inten-
sified by the fact that the modular group does not operate freely, see Theorem 2.16:
There are elliptic points. The difficulty consist in dealing with the elliptic points
with respect to the Hausdorff property and with respect to complex charts on the
quotient.

The modular group Γ contains the two distinguished elements

• Reflection at the y-axis and dividing by squared absolute value:

−τ
 
0 −1
S := ∈ Γ , S(τ) = −1/τ = 2 .
1 0 |τ|
2.2 Topology of the orbit space of the Γ -action 47

• Translation by one:
 
1 1
T := ∈ Γ , T (τ) = τ + 1.
0 1

Theorem 2.16 (Fundamental domain and elliptic points of the Γ -action).

1. The modular group Γ is generated by the two elements S, T ∈ Γ which satisfy

S4 = id, (ST )6 = id.

2. The orbit space Y = Γ \H of the action

Φ : Γ ×H →
− H, (γ, τ) 7→ γ(τ),

maps bijectively to the set

D := {τ ∈ H : |τ| > 1 and −1/2 ≤ Re τ < 1/2}∪{τ ∈ H : |τ| = 1 and −1/2 ≤ Re τ ≤ 0},

the standard fundamental domain of Φ.

3. The action Φ has exactly two elliptic points τ0 ∈ D. Their isotropy groups are
cyclic:
• τ0 = i with
Γi =< S >⊂ Γ , ord Γi = 4,
• τ0 = ρ := e2πi/3 with

Γρ =< ST >⊂ Γ , ord Γρ = 6.

48 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Fig. 2.1 Fundamental domain D of the modular group Γ shaded, from [29]

Figure 2.1 shows the fundamental domain D of the modular group. In order to
exclude pairs of congruent points within D one has to exclude part of the boundary
of D. This convention is followed by [29] but not by all references. Most references
name fundamental domain the closure D.

Fig. 2.2 Some translates of the fundamental domain of Γ , from [52, Chap. VII, Fig. 1]
2.2 Topology of the orbit space of the Γ -action 49

Figure 2.2 shows some translates of the fundamental domain of Γ under the
elements T, S ∈ Γ and their products.

Proof. The three parts of the subsequent proof do not correspond one to one to the
three parts of the theorem. Note that the subgroup

{±id} ⊂ Γ

operates on H in a trivial way and that S2 = −id.

i) The domain D: Denote by

G :=< S, T >⊂ Γ
the subgroup generated by the two distinguished elements. We show: Any given
point τ ∈ H is equivalent to a point τ 0 ∈ D mod G.

We choose an element γ0 ∈ G with Im γ0 (τ) maximal. Such choice is possible

because  
Im τ ab
Im γ(τ) = for γ = ∈ G.
|c · τ + d|2 cd
The translation T ∈ Γ does not change the imaginary part of its argument. Therefore
we may assume
−1/2 ≤ Re γ0 (τ) < 1/2.
In addition, |γ0 (τ)| ≥ 1 because otherwise
!
− γ0 (τ)
Im (S ◦ γ0 )(τ) = Im > Im γ0 (τ).
|γ0 (τ)|2

• If |γ0 (τ)| > 1 then τ 0 := γ0 (τ).

• If |γ0 (τ)| = 1 but 0 < Re γ0 (τ) ≤ 1/2 then

τ 0 := (S ◦ γ0 )(τ).

In both cases τ 0 ∈ D.

ii) Isotropy groups and fundamental domain: Consider two points τ, τ 0 ∈ D with
 
ab
τ 0 = γ(τ), γ = ∈Γ.
cd
We show τ = τ 0 , and the latter equation implies γ ∈ Γτ .

We may assume without restriction Im τ 0 ≥ Im τ. Then

50 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Im τ
Im τ 0 = .
|c · τ + d|2
implies
|c · τ + d| ≤ 1.
√
Because Im τ ≥ 3/2 only the cases c = 0, ±1 are possible:
• Case 1, c = 0: Then d = ±1 and a = d. Without restriction a = d = 1, hence

τ 0 = τ + b, b ∈ Z.

Because Re τ 0 = (Re τ) + b we obtain b = 0; i.e. τ = τ 0 and

 
10
γ =± = ± id ∈ Γτ .
01

• Case 2, c = 1: Then |τ + d| ≤ 1, which implies d ∈ {0, ±1}.

Case 2a, d = 0: If d = 0 then b = −1 and

a·τ −1
τ0 = = a − (1/τ), a ∈ Z.
τ
Due to τ ∈ D the equation

|c · τ + d| = |τ| ≤ 1

implies |τ| = 1. Hence

1/τ = τ and τ 0 + τ = a ∈ Z

which implies
Re τ 0 + Re τ = a ∈ Z and Im τ 0 = Im τ.
Therefore: Either a = 0 and τ = i which implies

τ 0 = −(1/τ) = i = τ

and  
0 −1
γ =± = ±S ∈ Γi .
1 0
Or a = −1 and τ = ρ which implies

τ 0 = −1 − (1/τ) = ρ = τ

and  
−1 −1
γ =± = ±(ST )2 ∈ Γρ .
1 0
Case 2b, d = ±1: If d = ±1 then
2.2 Topology of the orbit space of the Γ -action 51

|c · τ + d| = |τ ± 1| ≤ 1.

Because of τ, τ 0 ∈ D the case d = −1 with |τ −1| ≤ 1 is excluded. Therefore d = 1

leaving
|τ + 1| = |τ − (−1)| ≤ 1.
As a consequence, τ = ρ and

a·ρ +b
ad − bc = a − b = 1 and τ 0 = = a + ρ,
ρ +1

hence a = 0, τ 0 = ρ = τ and
 
0 −1
γ= = ST ∈ Γρ .
1 1

• Case 3, c = −1: This case reduces to case 2, c = 1, by considering −γ0 .

iii) G = Γ : Consider an arbitrary element γ ∈ Γ and the point

τ0 := 2 · i ∈ D.
Due to part i) an element γ0 ∈ G exists with

(γ0 ◦ γ)(τ0 ) ∈ D.

Due to part ii) the two points of D coincide

τ0 = (γ0 ◦ γ)(τ0 ) ∈ D,

i.e.
γ0 ◦ γ ∈ Γτ0 = {±id}
which implies γ ∈ G, q.e.d.

Remark 2.17 (Presentation of the modular group and fundamental domain of its
subgroups).
1. Consider the element
 
−1 1
U := −T S = ∈ SL(2, Z)
−1 0

which satisfies U 3 = id. Apparently also

G =< S, U > .
52 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Therefore Theorem 2.16 implies that the two elements S and U generate the
group SL(2, Z). One can show that all relations of the generators are generated
by
S4 = U 3 = S2U · (US2 )−1 = id,
i.e. that
SL(2, Z) =< S, U; S4 ,U 3 , S2U · (US2 )−1 >
is a presentation of SL(2, Z), see [36, Kap. II, §2 Bemerk.].

2. Denote by
PSL(2, Z) := SL(2, Z)/{±id}
the corresponding projective-linear group, which acts faithful on H, and by

π : SL(2, Z) →
− PSL(2, Z), A 7→ A,

the canonical quotient map. The projective-linear group has the two cyclic sub-
groups
H1 :=< S >' Z/2Z and H2 :=< U >' Z/3Z.
One can show
2 3
PSL(2, Z) = H1 ∗ H2 =< S,U; S ,U >
as the free product, see the short argument from [3] and also [52, Chap. VII, § 1, Rem.].

3. For a subgroup K ⊂ Γ with finite index J := [G : K] < ∞ each decomposition

˙
[
Γ= γ · K, γj ∈ Γ ,
j∈J j

into left cosets provides a fundamental domain

˙
D(K) := γ −1 (D)
[
j∈J i

of the induced K-action, which derives from the standard fundamental domain D
of Γ . In the following we will always choose fundamental domains D(K) ⊂ H
which originate as translates of D. These fundamental domains are measurable
subsets of H.

Corollary 2.18 (Elliptic points). For N ≥ 1 the action

Φ : Γ0 (N) × H →
− H

has finitely many elliptic orbits. Each elliptic point τ ∈ H has the finite isotropy
group
Γ0 (N)τ = Γτ ∩ Γ0 (N).
2.2 Topology of the orbit space of the Γ -action 53

Proof. Theorem 2.16 shows: The action of Γ has exactly two elliptic points i, ρ ∈ D,
and both have finite isotropy groups. The index formula from Proposition 2.14 im-
plies the finiteness of
[Γ : Γ0 (N)] =: k.
According to Remark 2.17 each finite coset decomposition
k
[
Γ= γ j · Γ0 (N), γ j ∈ Γ ,
j=1

provides the fundamental domain of the Γ0 (N)-action

k
D(Γ0 (N)) := γ −1
[
j (D)
j=1

e.g. see Figure 2.3. For each j = 1, ..., k and each point τ ∈ γ −1 j (D) the group mor-
phism
− Γγ j (τ) , γ 7→ γ j · γ · γ −1
Γ0 (N)τ → j ,

injects Γ0 (N)τ into Γγ j (τ) . Hence the elliptic points of Γ0 (N) contained in D(Γ0 (N))
are contained in the finite set
k k
γ −1 γ −1
[ [
j (i) ∪ j (ρ), q.e.d.
j=1 j=1

Figure 2.3 shows the fundamental domain of Γ0 (13). The fundamental domain
D(Γ0 (13)) splits into 14 disjoint translates of D according to the index formula from
Proposition 2.14
[Γ : Γ0 (13)] = 14
For further explanations of Figure 2.3 see [17, Fig. 3.1 and Exerc. 3.1.4].
54 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Fig. 2.3 Fundamental domain of Γ0 (13), adapted from [17, Fig. 3.1]

The following Lemma 2.19 about the local behaviour of the action Φ covers the
main step for proving the Hausdorff property of the quotient topoplogy with respect
to the quotient map
H→ − Γ \H.

Lemma 2.19 (Properly discontinuous group action). Any two points τ1 , τ2 ∈ H

have neighbourhoods
Ui ⊂ H, i = 1, 2,
such that for all γ ∈ Γ :

γ(U1 ) ∩U2 6= 0/ =⇒ γ(τ1 ) = τ2 .

In particular: Any point τ ∈ H has a neighbourhood U ⊂ H such that for all γ ∈ Γ :

γ(U) ∩U 6= 0/ =⇒ γ ∈ Γτ .

The set U has no elliptic points except possibly τ.

2.2 Topology of the orbit space of the Γ -action 55

Fig. 2.4 Local behaviour of the action of the modular group, see Lemma 2.19

Figure 2.4 visualizes Lemma 2.19: If an element γ ∈ Γ maps a point of the dis-
tinguished neighbourhood U1 of τ1 to the distinguished neighbourhhood U2 of τ2 ,
then γ maps τ1 to τ2 . The add-on states: If γ ∈ Γ does not fix τ, then γ moves all
points from U to the complement of U.

Proof. 1. Introducing a finiteness condition: We choose two relatively compact

neighbourhoods
Vi ⊂⊂ H of τi , i = 1, 2,
and show: Only finitely many matrices γ ∈ Γ exists with
56 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

γ(V1 ) ∩V2 6= 0.
/

The idea is to consider the extension of the group action of Γ to SL(2, R), see
Remark 2.8,
 
aτ + b ab
Φ : SL(2, R) × H → H, (γ, τ) 7→ γ(τ) := , γ := ,
cτ + d cd

which provides a compact isotropy group of the point i

SL(2, R)i = SO(2, R) :

One checks
γ(i) = i ⇐⇒ a = d and b = −c,
hence  
ab
∈ SO(2, R).
cd
The action Φ of SL(2, R) is transitive because any point τ ∈ H belongs to the
orbit of i: If
τ = x + iy
then γτ (i) = τ for the matrix

x
 
√
 y √ 
y  
 = √1 · y x ∈ SL(2, R).

γτ := 
  y 01
 1
0 √
y

As a consequence for any two points e1 , e2 ∈ H and any γ ∈ SL(2, R):

γ(e1 ) = e2 ⇐⇒ γ ∈ γe2 · SO(2, R) · γe−1

1

according to the following commutatice diagram

γ
e1 e2

γe−1
1
γe2

i i

If a matrix γ ∈ Γ satisfies γ(V1 ) ∩V2 6= 0/ then also for the closure

γ(V 1 ) ∩V 2 6= 0/

and
2.2 Topology of the orbit space of the Γ -action 57

γe2 · SO(2, R) · γe−1

[
G := {γ ∈ SL(2, R) : γ(V 1 ) ∩V 2 6= 0}
/ = 1
.
e1 ∈V 1 ,e2 ∈V 2

Because γe and its inverse γe−1 depend continuously on e ∈ H, the latter set on
the right-hand side is the continuous image of the compact set

V 2 × SO(2, R) ×V 1

under the continuous map

− SL(2, R), (e2 , γ, e1 ) 7→ γe2 · γ · γe−1

V 2 × SO(2, R) ×V 1 → 1
,

and therefore compact itself. Its intersection

G∩Γ

with the discrete group Γ is finite.

2. Shrinking V1 and V2 : According to the first part only finitely many γ ∈ Γ exist
with
γ(V1 ) ∩V2 6= 0.
/
In particular, the set

GΓ := {γ ∈ Γ : γ(V1 ) ∩V2 6= 0/ and γ(τ1 ) 6= τ2 }

is finite. Therefore, after finitely many steps we can shrink the neighbourhoods V1
and V2 to neighbourhoods U1 and U2 with

γ(U1 ) ∩U2 = 0/ iff γ(τ1 ) 6= τ2 :

For each γ ∈ GΓ we have

γ(τ1 ) 6= τ2 .
Because H is a Hausdorff space we choose for each γ ∈ GΓ two disjoint neigh-
bourhoods of respectively γ(τ1 ) and τ2 in H

U1,γ ⊂ H, U2,γ ⊂ H and U1,γ ∩U2,γ = 0,

/

and define two neighbourhoods of respectively τ1 and τ2 in H

! !
−1
\ \
U1 := V1 ∩ γ (U1,γ ) , U2 := V2 ∩ U2,γ .
γ∈GΓ γ∈GΓ

Then for all γ ∈ GΓ

γ(U1 ) ∩U2 = 0.
/
58 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

3. Isotropy group Γτ : The add-on is the specific case

τ = τ1 = τ2 .

Apply part 2) and set

U := U1 ∩U2
4. Elliptic points form a discrete subset: Each
 
ab
γ= ∈ Γ , γ 6= ±id,
cd

fixes at most one point z ∈ H: If z ∈ H and γ(z) = z then z satisfies the quadratic
equation
cz2 + (d − a)z − b = 0
If
c 6= 0 or d 6= a or b 6= 0
the equation has at most two solutions, counted with multiplicity. The same
quadratic equation is also satisfied by z. Because z 6= z, the equation has no fur-
ther solution. Hence there is no

z0 ∈ H, z0 6= z,

satisfying γ(z0 ) = z0 . For any elliptic point τ 0 ∈ U exists

γ ∈ Γτ 0 , γ 6= ±id.

Because
τ 0 ∈ γ(U) ∩U
the previous part of the proof implies

γ ∈ Γτ

and therefore τ = τ 0 , q.e.d.

Lemma 2.20 (Openness of the canonical projection onto the orbit space). De-
note by
p:H→ − Y := Γ \H
the canonical projection onto the orbit space of the action

Φ : Γ ×H →
− H

and provide Y with the corresponding quotient topology. Then p is an open map.

Proof. By definition of the quotient topology a subset U ⊂ Y is open iff

2.2 Topology of the orbit space of the Γ -action 59

p−1 (U) ⊂ H

is open.

For each arbitrary but fixed γ ∈ Γ its orbit map, the restriction

Φ(γ, −) : H →
− H, τ 7→ γ(τ),

is continuous with inverse the orbit map of γ −1

Φ(γ −1 , −) : H →
− H, τ 7→ γ −1 (τ).

Hence the orbit map Φ(γ, −) is a homeomorphism.

In order to prove the openness of p we consider an open set U ⊂ H and prove the
openness of p(U) ⊂ Y : The inverse image

p−1 (p(U)) =
[
γ(U)
γ∈Γ

is open as union of open sets. Hence

p(U) ⊂ Y

is open by definition of the quotient topology, q.e.d.

Theorem 2.21 (Hausdorff topology of the orbit space). The quotient topology on
the orbit space
Y := Γ \H
is Hausdorff.

Proof. Consider two distinct points y1 6= y2 ∈ Y . For j = 1, 2 chooose two points τ j ∈ H

with p(τ j ) = y j . The assumption

p(τ1 ) 6= p(τ2 )

implies: For all γ ∈ Γ

γ(τ1 ) 6= τ2 .
As a consequence, Lemma 2.19 provides two open neighbourhoods U j ⊂ H of re-
spectively τ1 and of τ2 with
γ(U1 ) ∩U2 = 0/
for all γ ∈ Γ . Hence
p(U1 ) ∩ p(U2 ) = 0.
/
60 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Lemma 2.20 implies that the sets p(U j ) ⊂ Y, j = 1, 2 are open. Hence they are
disjoint open neighbourhoods of respectively

y1 = p(τ1 ) and y2 = p(τ2 ), q.e.d.

With Theorem 2.21 the present section has obtained its first goal. Its second goal
is to embed the result into the general context: Which properties ensure that a set
of equivalence classes becomes a Hausdorff space when provided with the quotient
topology with respect to the canonical projection? We introduce the general concept
of a proper group action in Definition 2.22 and prove the Hausdorff criterion from
Lemma 2.24.

Definition 2.22 (Proper group action). Consider a continuous action

G × X → X, (g, x) 7→ g(x),

of a locally compact group G on a locally compact space X. The action is a proper

group action if
σ : G × X → X × X, (g, x) 7→ (x, g(x)),
is a proper map, i.e. if the inverse image of a compact set is again compact.

Note. The image of the map σ from Definition 2.22 is the equivalence relation
R ⊂ X ×X
which the group action induces on X.

Corollary 2.23 (Proper action of the modular group). The action of the modular
group
Φ : Γ ×H → H
is a proper group action.
Proof. Consider a compact subset

K ⊂ H × H.

In order to show the compactness of

σ −1 (K) ⊂ Γ × H

it suffices to show that σ −1 (K) is sequentially compact. We claim: Any sequence

(γν , τν )ν∈N
2.2 Topology of the orbit space of the Γ -action 61

in σ −1 (K) has a convergent subsequence. Due to compactness of K we may assume

that the image sequence

(σ (γν , τν ))ν∈N = (τν , γν (τν ))ν∈N

is convergent. Set

x1 := lim τν , x2 := lim γν (τν ) (pointwise convergence).

ν→∞ ν→∞

We have to derive from the convergent sequences of points in H

(τν )ν∈N and (γν (τν ))ν∈N

the existence of a convergent subsequence of group elements (γν )ν∈N ∈ Γ :

Lemma 2.19 implies the existence of open neighbourhoods Ui ⊂ H of xi , i = 1, 2,

such that for all γ ∈ Γ
γ(U1 ) ∩U2 6= 0/ =⇒ γ(x1 ) = x2 .
Due to the pointwise convergence for all ν ∈ N but finitely many indices

τν ∈ U1 and γν (τν ) ∈ U2 ,

hence
γν (U1 ) ∩U2 6= 0.
/
The lemma implies
γν (x1 ) = x2
The set
{γ ∈ Γ : γ(x1 ) = x2 }
has the same cardinality as the isotropy group Γx1 . The latter is a finite group due to
Theorem 2.16. A subsequence (γν )ν∈N becomes stationary, hence convergent, q.e.d.

Lemma 2.24 (Hausdorff criterion for an equivalence relation). Let X be a topo-

logical space and
R ⊂ X ×X
an equivalence relation on X. Then the set of equivalence classes X/R provided with
the quotient topology with respect to the canonical map

p:X →
− X/R

is a Hausdorff spaces if both of the following conditions are satisfied:

• The map p is open
62 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

• The equivalence relation R ⊂ X × X is closed.

Proof. See [18, Chap. VII, 1.6]: Consider x, y ∈ X with

p(x) 6= p(y)

as classes in X/R. Then (x, y) ∈

/ R. Because R ⊂ X × X is closed, there is a neigh-
bourhood
U ×V ⊂ (X × X) \ R
of (x, y) ∈ X × X. Openness of p implies the existence of neigbourhoods

p(U), p(V ) ⊂ X/R

of respectively p(x) and p(y) with

p(U) ∩ p(V ) = 0,
/ q.e.d.

With the new concepts and results the proof of the Hausdorff property from
Theorem 2.21 reads as follows:

Corollary 2.25 (Hausdorff topology of the orbit space). The quotient topology on
the orbit space
Y := Γ \H
is Hausdorff.

Proof. We apply Lemma 2.24: First, Lemma 2.20 states that p is an open map.
Secondly, we have
R = σ (Γ × H) ⊂ H × H
According to Corollary 2.23 the map

σ : Γ ×H → H×H

is proper, and therefore a closed map. Therefore Lemma 2.24 implies: The orbit
space Y , provided with the quotient topology, is a Hausdorff space, q.e.d.

As a consequence of the Hausdorff property of the orbit space Y each orbit

p−1 (y) ⊂ H, y ∈ Y,

of the Γ -action is closed.

2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 63

2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces

The elliptic points of the group action are those points which complicate the intro-
duction of a complex structure on the orbit space Y . To overcome this difficulty we
make a detailed study of the group action in the neighbourhood of an elliptic point.
Fortunately, here the action restricts to the action of the isotropy group and the lat-
ter turns out as a group of rotations. The statement of the following Lemma 2.26 is
trivial for non-elliptic points.

For elliptic points it relies on the Lemma of Schwarz. Lemma 2.26 states: Locally
the isotropy group of an elliptic point τ ∈ H with period hτ acts as the group of
rotations of a disk around the origin with angles a multiple of 2π/hτ .

Lemma 2.26 (Isotropy group acting as group of rotations). For any point τ ∈ H
exist
• a neighbourhood Uτ ⊂ H of τ such that for all γ ∈ Γ

γ(Uτ ) ∩Uτ 6= 0/ =⇒ γ ∈ Γτ ,

• and a fractional linear transformation λτ ∈ GL(2, C) satisfying

'
λτ : Uτ −
→ ∆τ

with a disc ∆τ = {z ∈ ∆ : |z| < rτ } for a suitable radius rτ > 0

such that: For all γ ∈ Γτ the conjugate map

γ˜τ := λτ ◦ γ ◦ λτ−1 : ∆τ → ∆τ

from the commutative diagram

γ
Uτ γ(Uτ )

λτ λτ
γ˜τ
∆τ ∆τ

is a rotation around 0 ∈ ∆τ . The angle of rotation is an integer multiple of 2π/hτ .

Proof. Due to Theorem 2.16 w.l.o.g.

τ ∈ {i, ρ}.

Denote by
64 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

z−τ
λτ : P1 → P1 , λτ (z) := ,
z−τ
the fractional linear automorphism of P1 with λ (H) = ∆ and

λτ (τ) = 0, λτ (τ) = ∞.

The following diagram commutes for all γ ∈ Γτ

γ
H H

λτ λτ
γ˜τ
∆ ∆

The conjugate
γ̃τ := λτ ◦ γ ◦ λτ−1 : ∆ → ∆
has the fixed point 0 ∈ ∆ . The non-trivial isotropy groups are

Γi =< S > and Γρ =< ST > .

For τ ∈ {i, ρ} one checks

|γ̃τ 0 (0)| = |γ 0 (τ)| = 1

because in particular
|S 0 (i)| = |(ST )0 (ρ)| = 1.
The Lemma of Schwarz implies that

γ̃τ : ∆ →
− ∆

is a rotation
γ̃τ (z) = ek·(2πi/hτ ) · z, k ∈ {0, 1, ..., hτ − 1}
We choose a disk ∆τ ⊂ ∆ such that the neighbourhood of τ ∈ H

Uτ := λτ−1 (∆τ )

satisfies the properties from Lemma 2.19. Then restricting the above digram proves
the claim, q.e.d.

In order to obtain a complex structure on the orbit space, the Hausdorff space
Y = Γ \H,
2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 65

we introduce charts on Y with holomorphic transition functions using the canonical

projection
p:H→ − Y.
Around the class p(τ) ∈ Y of an elliptic point τ ∈ H we may consider the isotropy
groups Γτ as cyclic groups of rotations of a disk ∆τ due to Lemma 2.26. Assume

(1/2) · ord Γτ = hτ =: k.

Then in a neighbourhood ∆τ of λτ (τ) the function

z 7→ zk

foliates the k different segments of the disk ∆τ , which are eqivalent under the Γτ -action,
to an open set Wτ ⊂ C, see Figure 2.5, and induces a chart
φτ
p(Uτ ) −→ Wτ .

Fig. 2.5 The three equivalent segments of the disk ∆τ of an elliptic point τ ∈ H with hτ = 3

Proposition 2.27 (Analytic structure of the orbit space). On the orbit space

Y = Γ \H

exists an analytic structure such that Y becomes a Riemann surface and the canon-
ical projection
p:H→Y
becomes a holomorphic map.
Considered from a general point of view the complex manifold Y is the coarse
moduli space of complex tori.
66 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Proof. i) Homeomorphisms: For each τ ∈ H choose a neighbourhood Uτ ⊂ H of τ

with the properties from lemma 2.26. Using the notation from the lemma define

powτ : ∆τ → C, z 7→ zhτ ,

and set
Wτ := powτ (∆τ )
Eventually, define a chart of Y around y = p(τ) ∈ Y

φτ : p(Uτ ) → Wτ

as the commutative completion of the following diagram

p|Uτ
Uτ p(Uτ )

λτ φτ
powτ
∆τ Wτ

The map φτ is
• well-defined and bijective according to Lemma 2.26 because for all γ ∈ Γ \ Γτ

γ(Uτ ) ∩Uτ = 0.
/

The different inverse images under p|Uτ of a given point from p(Uτ ) are identi-
fied by powτ ◦ λτ ,

• continuous, because p is open,

• open, because λτ and powτ are open.

As a consequence, φτ is a homeomorphism.

ii) Explicit form of the transition functions: We show that the family

A := (φτ : p(Uτ ) → Wτ )τ∈H

is a complex atlas of Y .

Assume two charts

φτi : p(Uτi ) → Wτi
around points yi = p(τi ) ∈ Y, i = 1, 2, and assume

p(Uτ1 ) ∩ p(Uτ2 ) 6= 0.
/
2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 67

To simplify the subscripts we set

Ui := Uτi , φi := φτi , λi := λτi , hi := hτi ,Wτi := Wi , i = 1, 2.

Consider a point
x ∈ p(U1 ) ∩ p(U2 ),
see Figure 2.6.
68 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Fig. 2.6 Transition functions of Y

2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 69

We show that the transition function

φ2,1 := φ2 ◦ φ1−1

is holomorphic in a neighbourhood of φ1 (x) ∈ W1 ⊂ C: By definition of φi two

points τ̃i ∈ Ui , i = 1, 2, exist with

λi (τ̃i )hi = φi (x), i = 1, 2.

They satisfy τ̃2 = γ(τ̃1 ) for a suitable γ ∈ Γ . Hence for q ∈ C in a neighbourhood

of φ1 (x) the transition functions attains the value
 h2
φ2,1 (q) = (λ2 ◦ γ ◦ λ1−1 )(q1/h1 )

We have to prove: Taking the h1 -root does not prevent the holomorphy of the map.

iii) Holomorphy of the transition functions: We prove the non-trivial case h1 > 1,
i.e. we assume that the orbit x ∈ X is elliptic.
• Case φ1 (x) = 0 ∈ W1 , i.e. τ̃1 = τ1 :

The point
τ1 = τ̃1
is elliptic. Therefore also τ̃2 = γ(τ˜1 ) ∈ U2 is elliptic. Because τ2 is the only
elliptic point in U2 we obtain τ˜2 = τ2 , and

τ2 = γ(τ1 ) and h := h2 = h1 .

By construction of λi , i = 1, 2, the composition of fractional linear

transformations
λ2 ◦ γ ◦ λ1−1
is a fractional linear transformation of P1 mapping 0 to 0 and ∞ to ∞. Therefore
a matrix  
ab
A= ∈ GL(2, C)
cd
exists with
λ2 ◦ γ ◦ λ1−1 = A
and
A · (0 1)> = (0 1)> , A · (1 0)> = (1 0)> .
With respect to the injection

→ P1 , z 7→ (z : 1),
C ,−

– the vector (0 1)> ∈ C2 represents the point (0 : 1) ∈ P1 and the point 0 ∈ C,

70 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

– while the vector (1 0)> ∈ C2 represents the point (1 : 0) ∈ P1 and the point
∞ ∈ C ∪ {∞}.
As a consequence  
a0
A=
0d
and
 h  h
φ2,1 (q) = (λ2 ◦ γ ◦ λ1−1 )(q1/h ) = A(q1/h ) = ((a · q1/h )/d)h = (a/d)h · q.

Therefore φ2,1 is linear, and hence holomorphic in a neighbourhood

of 0 = φ1 (x).

• Case φ2 (x) = 0 ∈ W2 , i.e. τ̃2 = τ2 :

The previous case shows the holomorphy of φ1,2 in a suitable neighbourhood.

Because the map is bijective, its inverse φ2,1 is holomorphic.

• Case φ1 (x) ∈ W1 arbitrary, i.e. τ̃1 ∈ U1 arbitrary:

We reduce this case to the other two cases. Choose τ3 ∈ D with p(τ3 ) = x and
consider the chart
φτ3 : Uτ3 → Wτ3
with the shorthand
U3 := Uτ3 , φ3 := φτ3 ,W3 := Wτ3 .
In a suitable neighbourhood W of x with

W ⊂ (p(U1 ) ∩ p(U2 ) ∩ p(U3 ))

we have
φ1,2 = φ1,3 ◦ φ3,2 .
The right-hand side is holomorphic due to part i) and ii). Therefore φ1,2 is
holomorphic, q.e.d.

The important steps in the proof of Proposition 2.27 and its prerequisites:

• The modular group acts properly.

• Isotropy groups act as rotations (Lemma of Schwarz).

'
• Suitable fractional-linear maps λ : H −
→ ∆.
2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 71

The upper half-plane has a canonical boundary point, the point ∞. The extension of
the group action from Remark 2.8 shows that one has to consider also the rational
line Q, but not R \ Q, as part of the boundary.
When extending the Euclidean topology of H to a topology on H∗ , one has to
define the neighbourhoods in H∗ of the additional points from

Q ∪ {∞}.

First, surprisingly as a neighbourhood basis of ∞ ∈ H∗ one does not take the

complements of compact sets in C as one does for the point

∞ ∈ P1

from the Riemann sphere, alias projective space. Instead one takes the sets

{z ∈ C : Im z > R} ∪ {∞}, R > 0,

i.e. the complements of sets with bounded imaginary part. Secondly, one translates
neighbourhoods of ∞ to neighbourhoods of a rational q ∈ Q by fractional-linear
transformations γ ∈ Γ with
γ(∞) = q.
The resulting neighbourhood basis of a point q ∈ Q are disks in H with q a point on
the Euclidean boundary of the disk, see Figure 2.7. The second step is determined
by the requirement, that the modular group acts as group of homeomorphisms
on H∗ . The new topology of H∗ induces on Q the discrete topology. Apparently the
latter is finer than the topology induced on Q by the Euclidean topology of R.

Fig. 2.7 Neighbourhoods of infinity and of some rational points from H∗ , from [17, Fig. 2.5]

Theorem 2.28 (The modular curve of the modular group). Consider the action
of the modular group Γ on the extended upper half-plane
72 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

H∗ := H ∪ Q ∪ {∞}.

according to Remark 2.8.

1. On the orbit space
X := Γ \H∗
exists the structure of a compact Riemann surface such that the following dia-
gram commutes with canonical projections p and p∗ and canonical injections in
the horizontal direction

H H∗

p p∗

Y X

2. The Riemann surface X is biholomorphic equivalent to P1 .

Proof. i) Topology on H∗ : Take as neighbourhood basis of ∞ for the topology on H∗

the sets
UR := {τ ∈ H : Im τ > R} ∪ {∞}, R > 0,
and as neighbourhood basis of

q := a/c ∈ Q, gcd(a, c) = 1,

the sets  
a∗
γ(UR ), R > 0, γ = ∈Γ,
c∗
because
γ(∞) = q
due to Remark 2.8. These neighbourhood bases together with the open sets of H
generate a topology on H∗ with H ⊂ H∗ an open subspace.
Note. Because fractional linear transformations from Γ take lines to circles or lines
the neighbourhoods of γ(UN ) are circles tangent to the real axis at the distinguished
point a/c ∈ Q. Hence neighbourhoods of ∞ and of rational points are completely
different from their counterpart with respect to the Euclidan topology, see
Figure 2.7.

ii) Hausdorff quotient topology on Γ \H∗ : We provide the set

X := Γ \H∗

with the quotient topology with respect to the canonical projection

2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 73

p∗ : H∗ → X.
Then p∗ is an open map: The proof is the same as the corresponding proof for
Lemma 2.20. The proof uses only the fact that the quotient arises from a group
action.

In order to show that X is Hausdorff we consider two points

xi = p∗ (τi ) ∈ X, τi ∈ H∗ , i = 1, 2, and assume x1 6= x2 .
We distinguish the following cases:
• τ1 ∈ H, τ2 ∈ Q ∪ {∞}: W.l.o.g. τ2 = ∞. For each
 
ab
γ= ∈Γ
cd

the formula
Im τ Im τ
Im γ(τ) = 2
= ≤
|c · τ + d| (c · Re τ + d)2 + c2 · (Im τ)2

1 
c=0


 2 c=0 Im τ
≤ Im τ · d ≤ ≤ max{Im τ, 1/Im τ}
1  1 c 6= 0
c 6= 0


 2
c · (Im τ)2 Im τ
shows
Im γ(τ) ≤ max{Im τ, 1/(Im τ)},
independently from γ. Therefore a relatively compact neighbourhood

U1 ⊂⊂ H

of τ1 and a neighbourhood
U2 ⊂ H∗
of τ2 = ∞ exist such that for all γ ∈ Γ

γ(U1 ) ∩U2 = 0.
/

As a consequence
p∗ (U1 ) ∩ p∗ (U2 ) = 0.
/
• τ1 , τ2 ∈ H: This case has been considered already by Corollary 2.25.

H
iii) Compactness of X: Denote by D ⊂ H the closure of D with respect to H.
Then
74 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

D := D ∪ {∞} ⊂ H∗
H

is compact with respect to H∗ : Consider an open covering (Ui )i∈I of D. If ∞ ∈ Ui0

then for suitable R > 0:
H
{τ ∈ D : Im τ > R} ∪ {∞} ⊂ Ui0 .

The remaining set

H
{τ ∈ D : Im τ ≤ R}
is compact in H, and therefore covered by a finite subcovering of (Ui )i∈I . As a
consequence X is compact as continuous image via p∗ of the compact set

D ⊂ H∗ .

One has
Y = X \ {p∗ (∞)}.
Therefore Y ⊂ X is an open subset.

iv) Charts of the modular curve: In order to define a chart

φ∞ : p∗ (U∞ ) → W∞
around the cusp
x = p∗ (∞) ∈ X
we consider the neighbourhood of ∞ ∈ H∗

U∞ := U2 = {τ ∈ H : Im τ > 2} ∪ {∞}.

We define (
e2πi·τ 6 ∞
if τ =
pow∞ : U∞ → ∆ , τ →
7
0 if τ = ∞
and set
W∞ := pow∞ (U∞ ) ⊂ ∆ .
For two points τi ∈ U∞ , i = 1, 2,

p∗ (τ2 ) = p∗ (τ1 ) ⇐⇒ τ2 = γ(τ1 ) for a suitable γ ∈ Γ

⇐⇒ τ2 = τ1 + m for a suitable m ∈ Z, due to Theorem 2.16,

⇐⇒ pow∞ (τ2 ) = pow∞ (τ1 ).
We define the chart φ∞ as the commutative completion of the diagram
2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 75

p∗ |U∞
U∞ p∗ (U∞ )

pow∞ φ∞

W∞

One shows analogously to the proof of Proposition 2.27 that the map φ∞ is a home-
omorphism.

Eventually we have to show that the charts of Y and the chart φ∞ are compatible, It
suffices to prove the compatibility of the chart

φ∞ : p∗ (U∞ ) → W∞

around p∗ (∞) with the chart

φτ : p(Uτ ) = p∗ (Uτ ) → Wτ

around an arbitrary point p(τ), τ ∈ H. We may assume that Uτ does not contain an
elliptic point, and therefore that

φτ ◦ (p∗ |Uτ )

is injective and biholomorphic. Hence the transition function of the two charts is
holomorphic.

v) Modular curve X ' P1 : The image of the fundamental domain

Y = p(D)
is homeomorphic to the plane C: Identify points on the left and right boundary of
D and equivalent points of D on the left part of the unit circle and on the right part.
Therefore the compactification of Y is homeomorphic to the 1-point compactifica-
tion of C, i.e. X is homeomorphic to S2 . The only complex structure on the sphere S2
is the complex projective space P1 (C), cf. [63], q.e.d.

For the biholomorphy

X ' P1
a formal proof will be given in Corollary 3.16. This second proof uses the modular
invariant j from the theory of modular forms.
76 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Definition 2.29 (Modular curve of Γ ). The compact Riemann surface from Theo-
rem 2.28
X(1) := X(Γ ) := Γ \H∗
is named the modular curve of Γ or simply the modular curve.

An analogous construction of a modular curve can be made for each congruence

subgroup Γ0 (N) ⊂ Γ .

Remark 2.30 (Modular curves of the congruence subgroups). Generalizing the con-
struction of the modular curve

X(1) = Γ \H∗

from Theorem 2.28 one can introduce for each Hecke congruence subgroup

Γ0 (N) ⊂ Γ , N ∈ N,

a complex structure on the orbit space

X0 (N) := Γ0 (N)\H∗ = (Γ0 (N)\H) ∪

˙ Cusp(Γ0 (N)),

see [17, Prop. 2.4.2ff.] The resulting compact Riemann surface X0 (N) is named the
modular curve of Γ0 (N).

1. Branched covering of modular curves: The inclusion of groups

→ Γ (1)
Γ0 (N) ,−

induces a holomorphic branched covering

f : X0 (N) →
− X(1)

of compact Riemann surfaces. The covering has degree

!
1
deg f = [Γ : Γ0 (N)] = N · ∏ 1+ ,
p|N
p

see [17, Exerc. 3.1.1] and Proposition 2.14.

2. Genus of modular curves: The Riemann-Hurwitz formula, e.g., [63], applied to

the covering
− X(1) ' P1
f : X0 (N) →
computes the genus of the modular curves of the Hecke congruence subgroups
as
2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 77

b( f )
g(X0 (N)) = 1 + − deg f
2
with
b( f ) = ∑ b( f ; x)
x∈X0 (N)

the total branching order of f .

A detailed investigation of the elliptic points and of the cusps of the action
of Γ0 (N) shows the genus formula

deg f ε2 ε3 ε∞
g(X0 (N)) = 1 + − − −
12 4 3 2
with ε2 , ε3 and ε∞ denoting the number of elliptic points of respectively period 2
and 3 and the number of cusps of the action of Γ0 (N), see [17, Theor. 3.1.1].

The ε-constants can be computed as

ε∞ = ∑ Φ(gcd(d, N/d)),
d|N

see Example 2.11, and ε2 and ε3 are the numbers of solutions in Z/NZ of the
equations respectively

x2 + 1 = 0 and x2 − x + 1 = 0.

Using the quadratic residue symbol for n ∈ {−1, −3} and extended to all primes p

   0 if p = 2
−1
:= 1 if p ≡ 1 mod 4
p 
−1 if p ≡ 3 mod 4


and 
  0
 if p = 3
−3
:= 1 if p ≡ 1 mod 3
p 
−1 if p ≡ 2 mod 3


one obtains, see [17, Cor. 3.7.2, Exerc. 3.7.6] and [53, Theor. 1.43]:
   
−1
∏ p|N 1 + p if 4 - N


ε2 =


0 otherwise

and
78 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)
   
−3
∏ p|N 1 + p if 9 - N


ε3 =


0 otherwise
3. Estimate of the genus g(X0 (N)): As a corollary of the genus formula the constants
from part 2) satisfy the estimate

ε2 2ε3
2 · g(X0 (N)) − 2 + + + ε∞ > 0.
2 3
4. Modular curves of prime level: For a prime p ∈ N and k := p + 1 the modular
curve X0 (p) has the genus
$ %
 k

 − 1 if k ≡ 2 mod 12
 12



g(X0 (p)) =
$ %
k





 12 otherwise

In particular,

g(X0 (2)) = 0, g(X0 (11)) = 1, g(X0 (13)) = 0 and g(X0 (17)) = 1

5. Numerical example: See the PARI-file “modular curve as covering”. For the
computation of the extended quadratic residue symbol PARI provides the com-
mand
kronecker(a,p).
The Kronecker symbol is defined for arbitrary integer a ∈ Z and arbitrary primes
including p = 2 by extending the Legendre symbol:
• Odd prime p:  
a
kronecker(a, p) := for odd p
p
• Even prime p = 2:
(
0 if a even
kronecker(a, 2) := 2
(−1)(a −1)/8 otherwise

The Kronecker symbol satisfies

• Odd prime p:
   2    
−4 2 −1 −1
kronecker(−4, p) = = · =
p p p p
2.3 Modular curves X(Γ0 (N)) as compact Riemann surfaces 79

• Even prime p = 2:
   
−1 −3
kronecker(−4, 2) = 0 = and kronecker(−3, 2) = −1 =
2 2

Hence 
∏ p|N (1 + kronecker(−4, p))
 if 4 - N
ε2 =

0 otherwise


and 
∏ p|N (1 + kronecker(−3, p))
 if 9 - N
ε3 =

0 otherwise


Figure 2.8 shows the invariants of the covering

X0 (p) →
− X(1)

for all primes p < 100.

80 2 The modular group Γ and its Hecke congruence subgroups Γ0 (N)

Fig. 2.8 Modular curves X0 (N) →

− X(1)
Chapter 3
The algebra of modular forms

Modular forms are holomorphic functions on H distinguished by two properties:

1. On one hand, they have the modular group as group of symmetries.

2. On the other hand, they expand into a Fourier series around the point ∞ with
rational Fourier coefficients after normalization.
The interplay of these two properties is coined “The magic of modular forms” [65].
Therefrom arises a series of remarkable identities between the Fourier coefficients
of modular forms. In many cases these identities imply interesting statements about
arithmetic functions. We will see examples in Section ?? and Chapter 6.

In general, one introduces modular forms as functions on H with a certain trans-

formation behaviour with respect to the action of the modular group or its con-
gruence subgroups. The basic examples are the Eisenstein series, see Section 3.2.
A meromorphic quotient of certain modular forms is the modular invariant j, see
Definition 3.15, a key object of the whole theory. From a more abstract point of
view, modular forms are differential forms on the modular curves, which are com-
pact Riemann surfaces, see Theorem 3.17. The Riemann-Roch theorem computes
the dimensions of the different vector spaces of modular forms.

3.1 Modular forms and cusp forms

Modular forms are holomorphic functions on H which transform in a simple man-

ner under the operation of the congruence subgroups Γ0 (N) and extend to H∗ in
a certain holomorphic way. These functions have a discrete non-Abelian group of
symmetries.

81
82 3 The algebra of modular forms

We introduce modular forms as analytical objects which are invariant under

transformation of the modular group Γ or its Hecke subgroups Γ0 (N). In a sec-
ond step, such invariant objects can be considered analytical objects on the modular
curves X0 (N). Because the modular curves are compact Riemann surfaces one can-
not expect interesting objects when restricting to invariant holomorphic functions.
Instead one has to consider more general holomorphic and meromorphic differential
forms.

We start from the general context of representation theory of the group Γ , i.e. we
introduce a Γ -action on certain vector spaces of holomorphic differential forms.

Proposition 3.1 (Γ -invariant differential forms). Denote by

W 1 := H 0 (H, ΩH1 )

the vector space of holomorphic differential forms on the upper half-plane H.

1. The map
Φ 1 : W 1 × Γ → W 1 , (ω, γ) 7→ Φ 1 (ω, γ) := γ ∗ ω,
is a Γ -right action. For
 
ab
γ= ∈ Γ , ω = f dτ ∈ W 1 and τ ∈ H
cd

holds
dτ
Φ 1 (ω, γ)(τ) := (γ ∗ ω)(τ) = ( f ◦ γ)(τ)
(cτ + d)2
2. For even k ∈ N set
⊗k/2
Ω k/2 := ΩH and W k/2 := H 0 (H, Ω k/2 )

and
Φ k/2 : W k/2 × Γ → W k/2 , (ω, γ) 7→ Φ k/2 (ω, γ) := γ ∗ ω.
For  
ab
γ= ∈ Γ , ω = f (dτ)k/2 ∈ W k/2 and τ ∈ H
cd
holds
(dτ)k/2
Φ k/2 (ω, γ)(τ) := ((γ ∗ )k/2 ω)(τ) = ( f ◦ γ)(τ)
(cτ + d)k
3. For even k ∈ N,
 
ab
γ= ∈ Γ and ω = f (dτ)k/2 ∈ W k/2
cd
3.1 Modular forms and cusp forms 83

holds the equivalence which characterizes invariant differential forms:

1
γ ∗ ω = ω ⇐⇒ f (τ) = ( f ◦ γ)(τ) for all τ ∈ H.
(cτ + d)k

Note. In the context of modular forms Ω k/2 denotes the tensor product of Ω 1 not
the exterior product.

Proof. 1. We have

Φ 1 (ω, γ1 · γ2 ) = (γ1 · γ2 ) ∗ ω = γ2∗ (γ1∗ ω) = Φ 1 (Φ 1 (ω, γ1 ), γ2 )

Moreover
!
aτ + b a(cτ + d) − (aτ + b)c 1
(dγ)(τ) = d = 2
dτ = dτ
cτ + d (cτ + d) (cτ + d)2

Therefore
dτ
(Φ 1 ( f dτ, γ))(τ) := (γ ∗ ( f dτ))(τ) = f (γ(τ)) (dγ)(τ) = ( f ◦ γ)(τ)
(cτ + d)2

2. For even k > 2 the proof is analogous by applying the derivation separately for
each tensor factor dτ, q.e.d.

Proposition 3.1 shows: For even k ∈ N the invariant differential forms from the
fixed space of the Γ -representation
f dτ k/2 ∈ (W k/2 )Γ
are characterized by holomorphic coefficient functions f ∈ O(H) satisfying

1
f (τ) = ( f ◦ γ)(τ) ·
(cτ + d)k

for all  
ab
γ= ∈ Γ and τ ∈ H.
cd
If these coefficent functions extend holomorphically to ∞ ∈ H∗ they will be named
modular functions of weight = k. The factors

1
(cτ + d)k

from their transformation law will be isolated and termed factors of automorphy,
see Definition 3.3.
84 3 The algebra of modular forms

For the translation

   
ab 11
T := := ∈ Γ0 (N)
cd 01
the factors of automorphy satisfy for all τ ∈ H

1
= 1.
cτ + d
Hence

T ∗ ( f (dτ)k/2 ) = f (dτ)k/2 ⇐⇒ f (τ + 1) = f (τ) for all τ ∈ H,

and the latter equality means that the function f has the period = 1. Hence f expands
into a convergent Fourier series
∞
f (τ) = fˆ(q) = ∑ an · qn , q = e2πi·τ ,
n=−∞

similar to the trigonometric function

1 2πi·z
+ e−2πi·z .


cos(2πz) = e
2
The holomorphic map
H → ∆ ∗ , τ 7→ e2πiτ ,
surjects to the punctured unit disk

∆ ∗ := {z ∈ C : 0 < |z| < 1}.

Here neighbourhoods of ∞ ∈ H∗ with respect to the topology of H∗ from Theorem 2.28

are mapped to neighbourhoods of zero. A meromorphic function on H with period = 1
induces a corresponding meromorphic function fˆ on ∆ ∗ according to the following
diagram

f
H C

fˆ

∆∗

Set
∞
fˆ(q) := f (τ) = ∑ an · qn , q = e2πi·τ .
n=−∞
3.1 Modular forms and cusp forms 85

Definition 3.2 (q-expansion and Fourier coefficients). Consider a meromorphic

function f on H invariant with respect to the operation of T and the induced mero-
morphic function fˆ on ∆ ∗ with

fˆ(q) := fˆ(e2πiτ ) := f (τ).

1. The function f is named meromorphic at ∞

• if fˆ has an isolated singularity at 0 and

• if this singularity is a pole or a removable singularity.

2. If f is meromorphic at ∞ then the induced function fˆ has a Laurent expansion
around 0
∞
fˆ(q) = ∑ an · qn , q = e2πiτ ,
n=n0

with a suitable n0 ∈ Z. The Laurent expansion is named the q-expansion of f ,

and the elements (an )n≥n0 are the Fourier coefficients of f .

3. A function f , meromorphic at ∞, is named holomorphic at ∞ if n0 ∈ N, i.e.

if 0 ∈ ∆ is a removable singularity of fˆ. In this case one defines

f (∞) := fˆ(0).

Note: For a meromorphic function f on H, invariant with respect to T :

f holomorphic at ∞ ⇐⇒ lim | f (τ)| < ∞ ⇐⇒
Im τ→∞

⇐⇒ ∃ R > 0 : | f (τ)| bounded for Im τ ≥ R.

Modular forms will not be necessarily invariant under the operation of Γ0 (N), but
they multiply with a factor of automorphy which is a holomorphic function without
zeros. We define the concept in a slightly more general context.
Definition 3.3 (Factor of automorphy). Consider the left group action
 
aτ + b ab
Φ : GL(2, R)+ × H → H, (γ, τ) 7→ γ(τ) := ,γ= ,
cτ + d cd

which extends the group action of the modular group Γ to GL(2, R)+ . The map

h : GL(2, R)+ × H → H, h(γ, τ) := cτ + d,

is named the factor of automorphy of Φ.

The translation T ∈ SL(2, Z) has the factor of automorphy h(T, −) = 1.

86 3 The algebra of modular forms

Lemma 3.4 (Transformation of the factors of automorphy).

1. The factors of automorphy satisfy the composition rule

h(B · A, τ) = h(B, A(τ)) · h(A, τ); A, B ∈ GL(2, R)+ , τ ∈ H.

2. Consider the vector space

V := { f : H →
− C}

of complex-valued functions on H. For any k ∈ Z the map

V × GL(2, R)+ →
− V, ( f , γ) 7→ f [γ]k ,

with
f [γ]k (τ) := f (γ(τ)) · h(γ, τ)−k · (det γ)k−1 , τ ∈ H,
is a right group action of GL(2, R)+ on the vector space V .

In the definition of f [γ]k from Lemma 3.4 the exponent of det γ is not uni-
form in the literature. We use the exponent (k − 1) from [17, Exerc. 1.2.11],
while [35, Chap. III, §3] uses the exponent k/2. The choice of the exponent (k − 1)
will later simplify Definition 5.10.
Note that in Lemma 3.4 the group GL(2, R)+ is written on the right-hand side of
the vector space V .

Proof. Consider A, B ∈ GL(2, R)+ and τ ∈ H.

1.    
τ A(τ)
A· = · h(A, τ).
1 1
Here we distinguish between the matrix product of A with a vector on the left-
hand side and the fractional linear transformation A(τ) on the right-hand side.

Applying this identity to the product B · A

   
τ (B · A)(τ)
(B · A) · = · h(B · A, τ).
1 1

Left-hand side:
       
τ A(τ) B(A(τ))
B· A· = B· · h(A, τ) = · h(B, A(τ)) · h(A, τ).
1 1 1

Equating with the right-hand side gives for the lower row

h(B · A, τ) = h(B, A(τ)) · h(A, τ),

3.1 Modular forms and cusp forms 87

hence for the upper row

(B · A)(τ) = B(A(τ)).
2. Starting with the right-hand side we compute

(( f [B]k )[A]k )(τ) = ( f [B]k )(A(τ)) · h(A, τ)−k (det A)k−1 =

= f (B(A(τ))) · h(B, A(τ))−k · h(A, τ)−k · (det B)k−1 · (det A)k−1 =

= f ((B · A)(τ)) · h(B · A, τ)−k · (det (B · A))k−1 = f [B · A]k (τ).
Here the penultimate equality holds due the result from part 1) and due to the
equality
(B · A)(τ) = B(A(τ))
from the proof of part i). Hence

f [B · A]k = ( f [B]k )[A]k , q.e.d.

Remark 3.5 (Meromorphy and holomorphy at the cusps). The Fourier expansion
from Definition 3.2 is well-defined for holomorphic functions f : H →
− C wich are
invariant with respect to the translation T . The modular group Γ has only a sin-
gle cusp, the orbit comprising ∞ and all rationals q ∈ Q. But Hecke congruence
subgroups Γ0 (N), N ≥ 2, have several cusps, each comprising a different subset of
rationals q ∈ N.

a
Fix a Hecke congruence subgroup Γ0 (N), a weight k ∈ Z, and a rational q = ∈ Q.
c
How to extend a holomorphic function
f :H→
− C
with
f [γ]k = f
for all γ ∈ Γ0 (N) to the point q ∈ Q in a meromorphic or holomorphic way?

We consider the canonical surjective group homomorphism

− PSL(2, Z) := Γ /{±1}, α 7→ α.
Γ→

The image Γ ∞ ⊂ PSL(2, Z) of the isotropy group Γ∞ ⊂ Γ is generated by the

translation:
Γ ∞ =< T >= {T k : k ∈ Z} ' Z
If α ∈ Γ with
q = α(∞)
88 3 The algebra of modular forms

then
Γ0 (N)q ' α −1 · Γ0 (N)q · α ⊂ Γ∞
or
−1
Γ0 (N)q ' α · Γ0 (N)q · α ⊂ Γ ∞ ' Z
For    
ab N −1 ∗ ∗
α= define γ := α · T · α = ∈ Γ0 (N)q
cd −Nc2 ∗

Then γ ∈ Γ0 (N)q . Because < T N > ⊂ Γ∞ is a subgroup of finite index, the same
holds for
< γ > ⊂ Γq and for < γ > ⊂ Γ q
We obtain
1 ≤ hq := [Γq : Γ0 (N)q ] = [Γ q : Γ0 (N)q ] < ∞

Lemma 3.4 shows

( f [α]k )[T hq ]k = f [α · T hq ]k = f [γ · α]k = ( f [γ]k )[α]k = f [α]k

Hence f [α]k is invariant under T hq and has a Fourier series
∞
( f [α]k )(τ) = ∑ an · q̂n , q̂ := e2πi·τ/hq
n=−∞

The function f [α]k is named meromorphic respectively holomorphic at ∞ iff f [α]k

has at q̂ = 0 at most a pole respectively a removable singularity. Note that the order
of f [α]k at q̂ = 0 does not depend on the choice of α.

The fundamental object of the whole theory is the modular form. We intro-
duce modular forms of the modular group Γ respectively its Hecke congruence
subgroups Γ0 (N) as the coefficient functions of the invariant differential forms from
Proposition 3.1 which extend to ∞ ∈ H∗ . It turns out that modular forms are mero-
morphic differential forms on the modular curve X0 (N), see Theorem 3.17.

Definition 3.6 (Modular forms and automorphic forms).

1. Modular forms and automorphic forms of the modular group: Consider an
integer k ∈ Z and a meromorphic function f on H.
• The function f is weakly modular of weight k with respect to Γ if for all γ ∈ Γ
and for all τ ∈ H
 
1 ab
f (γ(τ)) · k
= f (τ), γ = .
(cτ + d) cd
3.1 Modular forms and cusp forms 89

• The function f is an automorphic form of weight k with respect to Γ if f is

weakly modular of weight k and meromorphic at ∞.

• The function f is a modular form of weight k with respect to Γ if f is weakly

modular of weight k, holomorphic on H and holomorphic at ∞.

• A modular form f of weight k vanishing at all points of the cusp of Γ , i.e.

f (∞) = 0

is a cusp form of weight k with respect to Γ .

• An automorphic function with respect to Γ is an automorphic form of weight = 0,

i.e. satisfying
f ◦γ = f.
Similarly, a modular function is a modular form of weight = 0.

• On denotes by
Ak (Γ ) ⊃ Mk (Γ ) ⊃ Sk (Γ )
the vector spaces of automorphic forms with respect to Γ of weight k, the
subspace of modular forms, and the subspace of cusp forms.

2. Modular forms of congruence subgroups: Consider a congruence subgroup Γ0 (N)

with a positive integer N ∈ N∗ .
• A modular form of weight k ∈ Z, k ∈ Z, with respect to Γ0 (N) is a holomor-
phic function
f :H→C
with the following properties: For all γ ∈ Γ0 (N) holds

f [γ]k = f ,

and for all α ∈ Γ the function f [α]k is holomorphic at ∞ in the sense of

Remark 3.5.

• A modular form f of weight k ∈ Z, k ∈ Z, with respect to Γ0 (N) is named

a cusp form of weight k ∈ Z with respect to Γ0 (N) if for all α ∈ Γ the
function f [α]k is holomorphic at ∞, and its Fourier series in the sense of Re-
mark 3.5 vanishes at q̂ = 0.
3. Automorphic forms of congruence subgroups: An automorphic form of weight k, k ∈ Z,
with respect to Γ0 (N) is a meromorphic function on H with the following prop-
erties: For all γ ∈ Γ0 (N) holds
f [γ]k = f ,
90 3 The algebra of modular forms

and for all α ∈ Γ the function f [α]k is meromorphic at ∞ in the sense of

Remark 3.5.

We denote by
Ak (Γ0 (N)) ⊃ Mk (Γ0 (N)) ⊃ Sk (Γ0 (N))
the complex vector space of respectively automorphic forms, modular forms, and
cusp forms of weight k with respect to Γ0 (N).

Definition 3.6, part 2) and 3) generalize part 1): For γ ∈ Γ holds det γ = 1, hence
1
f [γ]k (τ) = ( f ◦ γ)(τ) ·
(cτ + d)k
Note that Definition 3.6 uses the weight convention from [17, Chap. 1.1]. Some
sources use a different weight convention. In the literature also the notation to dis-
criminate between modular functions and modular forms is not uniform.

As a consequence of Theorem 2.16 and Lemma 3.4 it is sufficient to require in

Definition 3.6, part 1) the transformation law only for the two generators S, T of Γ .

Remark 3.7 (Modular forms of odd weight). A modular form of a congruence sub-
group Γ0 (N), N ∈ N, of odd weight k ∈ Z satisfies for all τ ∈ H and for the particular
element  
−1 0
γ= ∈ Γ0 (N)
0 −1
the equation

f (τ) = f [γ]k (τ) = f (γ(τ)) · (−1)k = f (τ) · (−1)k .

Therefore f = 0 is the only modular function of Γ0 (N) of odd weight.

3.2 Eisenstein series and the algebra of modular forms of Γ

We will now construct for the modular group Γ = SL(2, Z) explicit modular
forms Gk of even weight k ≥ 4 and a distinguished holomorphic function G2 , which
behaves in a different way under the Γ -action. To achieve the construction we
need the convergence of certain infinite series. Lemma 3.8 provides the necessary
prerequisites.
3.2 Eisenstein series and the algebra of modular forms of Γ 91

The whole section will use the unbranched covering projection

− ∆ ∗ , τ 7→ q(τ) := e2πi·τ ,
q:H→
to study the number theoretic properties of modular forms f by considering the
Fourier coefficients of their Fourier expansion
∞
fˆ(q) = ∑ an · qn
n=0

Lemma 3.8 (Fourier series and partial fraction). Assume k ∈ N. Consider τ ∈ C

and set
q := e2πi·τ .
1. For k ≥ 2 and Im τ 6= 0
 ∞

 ∑ rk−1 · qr Im τ > 0
(−2πi)k  r=1

1
∑ (τ + n)k = (k − 1)! ·  ∞
n∈Z
(−1)k · ∑ rk−1 · q−r

 Im τ < 0
r=1

2. The double series

∞
∑ d · qmd
m,d=1

satisfies the prerequisites of the rearrangement theorem, i.e. there exists a con-
stant K such that for all finite subsets A, B ⊂ N∗ holds

∑ d · |q|md ≤ K
m∈A,d∈B

3. For even k ≥ 2, Im τ > 0,

!
1 2 (2πi)k ∞
∑ ∑ (mτ + n)k = · ∑ σk−1 (n) qn
m∈Z∗ n∈Z (k − 1)! n=1

is absolutely convergent.

Here for each k ≥ 2 the arithmetic function

σk−1 : N∗ → N∗ , σk−1 (n) := ∑ d k−1 ,

d|n

adds the (k − 1)-powers of the positive integer divisors of n.

Proof. 1. i) We first consider the case Im τ > 0. The proof compares the partial
fraction and the Fourier expansion of the cotangent function.
92 3 The algebra of modular forms

• Partial fraction, see [62]:

!
1 ∞ 1 1
π · cot(πτ) = + ∑ +
τ m=1 τ +m τ −m

• Fourier series
cos(πτ) 1+q ∞
π · cot(πτ) = π = (−iπ) · = −πi − 2πi ∑ qm .
sin(πτ) 1−q m=1

Here the second equality results from the Euler formulas

1 1
cos πτ = (eiπτ + e−iπτ ) and sin πτ = (eiπτ − e−iπτ )
2 2i
which imply

eiπτ + e−iπτ q+1 1+q

cot(πτ) = i · = i· = (−i) ·
eiπτ − e−iπτ q−1 1−q
The third equality uses the geometric series

1+q ∞ ∞
= (1 + q) ∑ qm = 1 + 2 ∑ qm
1−q m=0 m=1

which is absolutely convergent because |q| < 1 due to τ ∈ H.

Equating both expansions gives

!
1 ∞ 1 1 ∞
+∑
τ m=1
+
τ +m τ −m
= −πi − 2πi ∑ qm ,
m=1

and successive (k − 1)-times differentiation with respect to τ shows

1 (−2πi)k ∞ k−1 m
∑ k
= ·∑m q .
m∈Z (τ + m) (k − 1)! m=1

ii) For Im τ < 0 we use

(τ + n)k = (−1)k · (−τ − n)k

Hence
1 1 1
∑ (τ + n)k = (−1)k · ∑ (−τ − n)k = (−1)k · ∑ (−τ + n)k
n∈Z n∈Z n∈Z

Part i) applied to −τ implies

3.2 Eisenstein series and the algebra of modular forms of Γ 93

1 (−2πi)k
(−1)k · ∑ (−τ + n)k = · (−1)k · ∑ rk−1 · q−r
n∈Z (k − 1)! r=1

which proves the claim.

2. Consider τ ∈ H with q := e2πiτ real and 0 < q < 1.

For any real x ∈ [0, 1[ the geometric series satisfies
∞ 1
∑ xd = 1 − x
d=0

Taking the derivative and multiplying by x shows

∞ x
∑ d · xd = (1 − x)2
d=1

For arbitrary but fixed m ≥ 1 set x := qm < 1

! !
∞ ∞
md
∞ ∞
m d
∞ qm
∑ ∑ d ·q = ∑ ∑ d · (q ) = ∑ (1 − qm )2
m=1 d=1 m=1 d=1 m=1

The last series is convergent because for fixed q

lim |1 − qm |2 = 1.
m→∞

Set
∞ qm
K := ∑
m=1 (1 − qm )2
As a consequence
∑ d · qmd ≤ K.
m∈A,d∈B

For arbitrary τ ∈ H we have

q := e2πiτ with 0 ≤ |q| < 1

Hence w.l.o.g. we may assume q real with 0 ≤ q < 1, which proves the claim.

3. We split the exterior summation ∑ into

m∈Z∗

∑ where Im(mτ) < 0 and ∑ where Im(mτ) > 0.

m<0 m>0

Due to part 1) for each m ∈ Z∗ exists the interior series

94 3 The algebra of modular forms
∞
∑ rk−1 · q−rm if m < 0




(2πi)k r=1

1
∑ (mτ + n)k = (k − 1)! ·  ∞
n∈Z 
k−1 rm
∑ r ·q


 if m > 0
r=1

Hence for even k ≥ 2

! !
1 (2πi)k ∞ ∞
k−1 rm
∑ ∑ (mτ + n)k = 2· ·∑ ∑r ·q .
m∈Z∗ n∈Z (k − 1)! m=1 r=1

The latter series is convergent: For k ≥ 4 any rearrangement of the original double
series is permitted. For k = 2 the result of part 2) allows to apply the rearrange-
ment theorem to the double series
∞
∑ r · qrm , |q| < 1.
m,r=1

We rearrange the series by changing the indices of summation from (m, r)

to (n = rm, d|n):
! !
∞ ∞ ∞ ∞
∑ ∑ rk−1 · qrm = ∑ ∑ d k−1 · qn = ∑ σk−1 (n) · qn .
m=1 r=1 n=1 d|n n=1

Eventually for all even k ≥ 2

!
1 (2πi)k k
∑∗ ∑ (mτ + n)k = 2· · ∑ σk−1 (n) · qn , q.e.d.
m∈Z n∈Z (k − 1)! n=1

Eisenstein series are the lattice constants of normalized lattices

Λ1,τ = Z · 1 + Z · τ
when considered as functions of the argument τ ∈ H. Eisenstein series are explicit
representations of modular forms.

Theorem 3.9 (Eisenstein series, k ≥ 4). Consider an even integer k ≥ 4.

1. The Eisenstein series
0 1
Gk (τ) := ∑
(c,d)∈Z2
(cτ + d)k

is absolutely and compactly convergent for τ ∈ H.

The series defines a modular form Gk of Γ of weight k with

3.2 Eisenstein series and the algebra of modular forms of Γ 95

Gk (∞) = 2ζ (k).

Here
∞
ζ (s) := ∑ (1/ns ), Re s > 1,
n=1

denotes the Riemann ζ -function.

2. The Eisenstein series Gk ∈ Mk (Γ ) has the Fourier expansion

(2πi)k ∞
Gk (τ) = 2ζ (k) + 2
(k − 1)! ∑ σk−1 (n) · qn , q = e2πiτ .
n=1

Proof. i) Convergence and transformation: For τ ∈ H consider the normalized lat-

tice
Λτ := Z + Zτ
and for k ≥ 4 its lattice constants GΛτ ,k . Apparently

Gk (τ) := GΛτ ,k

The series GΛ ,k is absolute convergent due to Lemma 1.11, hence allows arbitrary
rearrangement. With respect to the translation

T : H → H, τ 7→ T (τ) := τ + 1,

we obtain
Gk (τ) = Gk (τ + 1).
With respect to the reflection

−1
S : H → H, τ 7→ S(τ) :=
τ
we have
1 1 0 1
Gk (S(τ)) · h(S, τ)−k = · Gk (−(1/τ)) = k · ∑ = Gk (τ).
τk τ (c,d)∈Z2 (−c(1/τ) + d)k

Because S and T generate the modular group, Lemma 3.4 implies for all γ ∈ Γ and
for all τ ∈ H
Gk (τ) = Gk (γ(τ)) · h(γ, τ)−k , i.e. Gk = Gk [γ]k

ii) Fourier expansion and holomorphy: We split the summation over

(c, d) ∈ Z2 , (c, d) 6= (0, 0),

into the simple sum for (c, d) with c = 0 and d ∈ Z∗ and the double sum with c ∈ Z∗
and d ∈ Z. Then Lemma 3.8, part 3) implies
96 3 The algebra of modular forms

0 1 1 1
Gk (τ) = ∑ = ∑ k+ ∑ =
(c,d)∈Z2
(cτ + d)k d∈Z∗ d c∈Z∗ ,d∈Z (cτ + d)k

2 (2πi)k ∞
= 2ζ (k) + · ∑ σk−1 (n) qn , q = e2πi·τ .
(k − 1)! n=1
The Fourier expansion, a convergent power series, proves the holomorphy of Gk in
H and at the point ∞, q.e.d.

In order to get rid of the transcendent values in the Fourier expansion of Gk one
divides by the value of the Riemann ζ -function.
Definition 3.10 (Normalized Eisenstein series). For even k ≥ 4 one defines the
normalized Eisenstein series

Gk
Ek := ∈ Mk (Γ )
2ζ (k)

Corollary 3.11 (Normalized Eisenstein series). For each k ≥ 4 all Fourier coeffi-
cients of the normalized Eisenstein series are rational, more precisely

2k ∞
Ek (τ) = 1 − · ∑ σk−1 (n) · qn ∈ Mk (Γ ), q = e2πi·τ
Bk n=1

Proof. 1. Euler relation of Bernoulli numbers: See [20, Kap. VII, Satz 4.1]. We
relate the cotangent function to the Bernoulli numbers (Bn )n∈N . The Bernoulli
numbers are the generators of the power series

z ∞ Bn n
=: ∑ ·z .
ez − 1 n=0 n!

We compute

cos z eiz + e−iz e2iz + 1 2i

cot z = = i · iz −iz
= i · 2iz
= i + 2iz
sin z e −e e −1 e −1
and use the vanishing of Bernoulli numbers with odd index ≥ 3 and B0 = 1, B1 = −1/2

2iz 2iz ∞ B2µ ∞ 2µ

2µ µ 2
z·cotz = iz+ = iz+1− + ∑ ·(2iz) = 1+ ∑ (−1) ·B2µ ·z2µ
e2iz − 1 2 µ=1 (2µ)! µ=1 (2µ)!

• On one hand from the definition of the Bernoulli numbers - replacing z by πz -

∞ 22µ
πz · cot(πz) = 1 + ∑ (−1)µ · (2µ)! · π 2µ · B2µ · z2µ
µ=1
3.2 Eisenstein series and the algebra of modular forms of Γ 97

• On the other hand the partial fraction of the cotangent function

∞ 2z
πz · cot(πz) = 1 + z ∑
ν=1 z2 − ν 2

and the geometric series

!µ
1 1 ∞ z2
2 2
=− 2· ∑
z −ν ν µ=0 ν2

imply !µ !
2
∞1 ∞ z2
πz · cot(πz) = 1 + 2z · ∑ − 2 · ∑ =
ν=1 ν µ=0 ν 2
!
∞ ∞ 1
= 1 − 2 ∑ ∑ 2µ z2µ
µ=1 ν=1 ν

Comparing the coefficients of both series representing

πz · cot(πz)

implies the Euler relation

∞ 1 2µ
µ+1 2
2 · ζ (2µ) := 2 · ∑ = (−1) · π 2µ · B2µ
ν=1 ν 2µ (2µ)!

2. Rationality of the Fourier coefficients: The Fourier expansion of the Eisenstein

series Gk from Theorem 3.9 and Part 1) show

Gk (τ) 2k ∞
Ek (τ) := = 1 − · ∑ σk−1 (n) · qn ∈ Mk (Γ ).
2ζ (k) Bk n=1

All Bernoulli numbers are rational due to the recursion formula for N ∈ N∗
N  
N +1
∑ n · Bn = 0 with B0 = 1.
n=0

As a consequence, all Fourier coefficients of Ek are rational, q.e.d.

Concerning the first values of Bernoulli numbers

!
2k
k, ζ (k), Bk , −
Bk
98 3 The algebra of modular forms

one has
(2, π 2 /6, 1/6, −24)
(4, π 4 /(2 · 32 · 5), −1/30, 240)
(6, π 6 /(33 · 5 · 7), 1/42, −504)
(8, π 8 /(2 · 33 · 52 · 7), −1/30, 480)
(10, π 10 /(35 · 5 · 7 · 11), 5/66, −264)
(12, 691 · π 12 /(36 · 53 · 72 · 11 · 13), −691/2730, 65520/691)

For k ∈ {4, 6, 8} the numerical value of Bk implies that all Fourier coefficients
of Ek are not only rational but are integers.

Remark 3.12 (Eisenstein series).

1. Rationality of the Fourier coefficients: The result from Corollary 3.11 is remark-
able: The invariance of the holomorphic function Ek under the right action of
the modular group implies the rationality of the Fourier coefficients of Ek . The
Fourier coefficients of these holomorphic functions are not arbitrary numbers
from C but rationals from Q. This rationality is the reason for the “magic of
modular forms” revealed by later applications. See also Proposition 3.14.

2. Weight k ≥ 4: For even k > 2 the Eisenstein series Gk relate to the lattice
constants GΛ ,k of an arbitrary lattice with positively oriented basis

Λ = Zω1 + Zω2

as follows: Set τ := ω2 /ω1 ∈ H. Then

1
GΛ ,k = Gk (τ).
ω1k

3. Weight k = 2: Similar to the series

∞ (−1)n−1
∑ n
n=1

in one dimension, in two dimensions the series

0 1
∑ (mτ + n)2
(m,n)∈Z2
3.2 Eisenstein series and the algebra of modular forms of Γ 99

is not absolutely convergent. Anyhow, in both cases some specific orders of sum-
mation provide certain finite values. In the 2-dimensional case we define the
Eisenstein series
!
∞ 1 1
G2 (τ) := 2 · ∑ 2 + ∑ ∑ 2
, τ ∈ H.
n=1 n m∈Z∗ n∈Z (mτ + n)

Lemma 3.8, part 3) implies that the sum on the right-hand side is convergent with
Fourier series
∞
G2 (τ) = 2 · ζ (2) − 8π 2 · ∑ σ1 (n) · qn , q = e2πi·τ .
n=1

The equality
ζ (2) = π 2 /6
implies for the normalized Eisenstein series of weight k = 2

G2 (τ) ∞
E2 (τ) := = 1 − 24 · ∑ σ1 (n) · qn , q = e2πi·τ .
2ζ (2) n=1

The convergent Fourier representation shows: G2 is holomorphic on H ∪ {∞} due

to Lemma 3.8, part 3), and invariant with respect to translation.

4. Transformation of G2 : With respect to the reflection S ∈ Γ the Eisenstein series G2

transforms as
1 2πi
(G2 [S]2 )(τ) := G2 (Sτ) · 2 = G2 (τ) −
τ τ
or
(G2 [γ]2 )(τ) = G2 (τ) − 2πic · h(γ, τ)−1
for general  
ab
γ= ∈Γ,
cd
see [36, Kap. III, § 6, Abschn. 1] or [17, Exerc. 1.2.8]. In particular, G2 is not
weakly modular. The lack of modularity is due to the conditional convergence
of G2 .

Proposition 1.19 dealt with the discriminant of the cubic polynomial from the dif-
ferential equation of the Weierstrass function ℘ of a lattice. Definition 3.13 shows:
These discriminants are the values of a modular form.
Definition 3.13 (Discriminant form). The discriminant form is the modular form
of Γ of weight k = 12 defined as

∆ := g32 − 27 · g23 : H → C
100 3 The algebra of modular forms

derived from the Eisenstein series

g2 := 60 · G4 , g3 := 140 · G6 .

In Definition 3.13 the letters g2 and g3 designate modular functions, not the con-
stants from Theorem 1.17, which denote the value at the period τ of a given normal-
ized lattice
Λ = Z · 1 + Z · τ.

Proposition 3.14 (Fourier coefficients of the dicriminant form). All Fourier co-
efficients of the normalized discriminant form

∆
∈ M12 (Γ )
(2π)12

are integers:
∆ (q) ∞
12
= ∑ τn · qn ∈ Z{q}, q = e2πi·τ ,
(2π) n=0

with τ0 = 0 and τ1 = 1.
In particular ∆ ∈ S12 (Γ ) is a cusp form of weight 12, and the zero at ∞ ∈ H∗ has
order = 1.

The Fourier coefficients τn , n ≥ 0, are the subject of certain of Ramanujan’s well-

known conjectures, see Definition 5.31.
Proof. i) Fourier series: We refer to Definition 3.10 of the normalized Eisenstein
series and their Fourier series expansion from Corollary 3.11

G4 ∞
E4 = 32 · 5 · 4
= 1 + 240 · ∑ σ3 (n) · qn
π n=1

33 · 5 · 7 G6 ∞
E6 = · 6 = 1 − 504 · ∑ σ5 (n) · qn .
2 π n=1

We have
22 4 23
g2 = 60 · G4 = π · E4 and g3 = 140 · G6 = 3 π 6 · E6 .
3 3
Hence
26 3 3
12 E4 − E6
2 3 2
12 E4 − E6
∆ = g32 − 33 g23 = π 12 · (E − E6
2
) = (2π) · = (2π) ·
33 4 26 · 33 1728
3.2 Eisenstein series and the algebra of modular forms of Γ 101

and eventually
∆ E43 − E62 E43 − E62
= = .
(2π)12 26 · 33 1728
We set as shorthand
∞
Xk (q) := ∑ σk (n) · qn
n=1

and obtain
∆ (1 + 240 · X3 )3 − (1 − 504 · X5 )2
=
(2π)12 26 · 33
When denoting by
Q = Q(q)
the numerator of the latter fraction, then

Q = (3 · 240 · X3 + 3 · 2402 · X32 + 2403 · X33 ) + (1008 · X5 − 5042 · X52 )

ii) Fourier coefficients τ0 = 1, τ1 = 1: In lowest order in q with

X3 (q) = q + O(2) and X5 (q) = q + O(2)

we get
Q(q) = 3 · 240 · q + 1008 · q + O(2) = 1728 · q + O(2)
which proves τ0 = 0 and τ1 = 1.

iii) Fourier coefficients τn ∈ Z, n ≥ 2: The remaining part of the proof reduces to

the claim that also all other coefficients of the series Q are divisible by
1728 = 26 · 33 = (22 · 3)3 = 123
We compute

Q = (3 · 240 · X3 + 3 · 2402 · X32 + 2403 · X33 ) + (1008 · X5 − 5042 · X52 )

Q ≡ 3 · 240 · X3 + 1008 · X5 ≡ 122 (5 · X3 + 7 · X5 ) mod 123

To complete the proof we have to show that all coefficients of the power series

5 · X3 + 7 · X5

are divisible by 12. We show even more: For all d ∈ Z

5 · d3 + 7 · d5 ≡ 0 mod 12

According to the Chinese remainder theorem the claim reduces to the two congru-
ences
5 · d 3 + 7 · d 5 ≡ 0 mod 3 and 5 · d 3 + 7 · d 5 ≡ 0 mod 4
102 3 The algebra of modular forms

or
d 5 ≡ d 3 mod 3, mod 4.
The latter two congruences reduce to
• If d 3 6≡ 0 mod 3 then d 2 ≡ 1 mod 3,

• and: if d 3 6≡ 0 mod 4 then d 2 ≡ 1 mod 4

The congruences have to be checked for

d ∈ {1, 2} ⊂ Z/3Z = {0, 1, 2} and d ∈ {1, 3} ⊂ Z/4Z = {0, 1, 2, 3},

which can be easily done, q.e.d.

Definition 3.15 (Modular invariant j). The modular invariant j is the automor-
phic function, defined as the quotient of two modular forms of weight = 12,

g32 g3
j := 1728 · = 1728 · 3 2 2 .
∆ g2 − 27g3

Corollary 3.16 (Modular curve and modular invariant j). The modular invari-
ant is a biholomorphic map
'
→ P1
j : X(1) −
which maps the cusp of X(1) to ∞ ∈ P1 .

Proof. Proposition 1.19 implies that j is a holomorphic function on H because the

denominator ∆ has no zeros in H. The Fourier expansion of

∆
(2π)12

from Proposition 3.14 and the equality

G4 (∞) = 2ζ (4) 6= 0

from Theorem 3.9 show that j has a pole of order = 1 at the point ∞. Because j is
even an automorphic function, it is a meromorphic function on the modular curve,
i.e. a holomorphic map
'
→ P1 .
j : X(1) −
Because all fibres of j have the same multiplicity = 1, the map j is biholomorphic,
q.e.d.
3.2 Eisenstein series and the algebra of modular forms of Γ 103

Continuing with the investigation started in Proposition 3.1 we now set out to
show: Modular forms of weight k are those meromorphic differential forms on the
modular curves X0 (N) which are multiples of a certain divisor which depends on
the level N and the weight k. Hence modular forms are the holomorphic sections of
certain line bundles on the modular curves. Applying the Riemann-Roch theorem
we draw several conclusions. In detail:

• We recall the Riemann-Roch theorem on a compact Riemann surface X with

genus g(X) for a divisor D ∈ Div(X), and also Serre duality, see [63]:

dim H 0 (X, OD ) − dim H 1 (X, OD ) = 1 − g(X) + deg D.

We will use also the notation

deg OD := deg D.

The canonical divisor K ∈ Div(X) has degree

deg K = 2g(X) − 2.

Serre-duality allows to replace the cohomology group H 1 by the dual of a coho-

mology group H 0
H 1 (X, OD ) = H 0 (X, OK−D )∨ .
Moreover, for each divisor D ∈ Div(X) with deg D < 0

dim H 0 (X, OD ) = 0.

• We derive a formula for the dimension of the vector spaces Mk (Γ ). Section 3.3
will generalize the computation to Mk (Γ0 (N)).

• Chapter 4 will show: Any complex elliptic curve arises from a complex torus.
More precisely, the elliptic curve is parametrized by the two meromorphic func-
tions ℘Λ and ℘Λ0 defined on a suitable torus C/Λ , see Theorem 4.28.

We recall the canonical projection to the modular curve

p : H∗ → X := Γ \H∗ ,
skipping the ∗-superscript used in the notation from Theorem 2.28. We introduce
the class notation
[τ] = p(τ) ∈ X.
In addition we use the floor function: For x ∈ R

bxc := max{n ∈ Z : n ≤ x}.

104 3 The algebra of modular forms

Theorem 3.17 (Modular forms are meromorphic differential forms on the modular
curve). Consider an even weight k ≥ 0.
1. Differential forms: The map of vector spaces is injective
 
α : Mk (Γ ) → H 0 X, M ⊗O Ω ⊗k/2 , f 7→ ω f ,

with ω f the well-defined meromorphic differential form on X induced by the Γ -invariant

modular form on H
f (τ) (dτ)k/2
2. The distinguished divisor: Consider the divisor
$ % $ %
k k k
Dk := · [i] + · [ρ] + · [∞] ∈ Div(X).
4 3 2

The attachment
U 7→ ΩΓ ,k (U), U ⊂ X open,
with
n o
ΩΓ ,k (U) := ω ∈ H 0 (U, M ⊗O Ω ⊗k/2 ) : ω 6= 0 and div ω ≥ −Dk |U ∪ {0}

is a sheaf ΩΓ ,k on X. The sheaf ΩΓ ,k

• is a subsheaf of the sheaf of meromorphic differential forms, i.e.

ΩΓ ,k ⊂ M ⊗ O Ω ⊗k/2

• and is isomorphic to the line bundle

LΓ ,k := O(k/2)·K+Dk ∈ Pic(X)

of multiples of the divisor −((k/2) · K + Dk ) ∈ Div(X)) with K ∈ Div(X) the

canonical divisor of X.

3. Meromorphic differential forms on the modular curve: The image of the map α
from part 1) is
α(Mk (Γ )) = H 0 (X, ΩΓ ,k ).

In particular, sections of the sheaf ΩΓ ,k , i.e. certain meromorphic differential

forms, correspond bijectively to holomorphic sections of the line bundle LΓ ,k .

Proof. Cf. [50, §4]. The numbers of the different parts of the proof do not corre-
spond one-to-one to the sections of the theorem.
3.2 Eisenstein series and the algebra of modular forms of Γ 105

i) Definition of α: If f ∈ Mk (Γ ) then Proposition 3.1 implies that the modular

group leaves invariant the differential form
f (τ) (dτ)k/2
on H. Hence the latter defines a meromorphic form ω f on X with pullback

f (τ) (dτ)k/2 := p∗ ω f :

One considers first the form ω f outside the three points which are the orbits of
the two elliptic points and the cusp. Secondly one extends ω f holomorphically into
these exceptional points by using Riemann’s extension theorem for bounded holo-
morphic functions.

ii) Distinguished divisor: The sheaf ΩΓ ,k is well-defined. The definition of the

divisor Dk results from the following relation between the order of f at t ∈ H∗ and
the order of ω f at [t] ∈ X:

ord[∞] ω f + k/2
 t =∞
ordt f =

ht · ord[t] ω f + (k/2) · (ht − 1) t ∈ H


Here ht denotes the period of t. The order of a meromorphic differential form

 
ω ∈ H 0 X, M ⊗O Ω ⊗k/2

at a point x ∈ X is defined as follows: Choose a chart of X around x

q:U →
− V ⊂C

and represent
ω|U = f · (dq)k/2
Then
ordx ω := ordx f ,
independently from the choice of the chart. To prove the order relation for t ∈ {i, ρ, ∞}
we use the local form of
p : H∗ → X
and the charts around t from Theorem 2.28 and Proposition 2.27. We denote by τ
points from Ut ⊂ H and by q points from Wt ⊂ C:
• For t = ∞ ∈ H∗ we recall the chart pow∞ around t and the chart φ∞ around the
cusp [∞] from the diagram
106 3 The algebra of modular forms

p|U∞
U∞ p(U∞ )

pow∞ φ∞

W∞

For τ ∈ U∞ (
e2πiτ τ=6 ∞
q(τ) := pow∞ (τ) :=
0 τ =∞
We denote by τ the points on U∞ ⊂ H and by q the points on W∞ ⊂ ∆ ⊂ C. Then

dq = 2πiq dτ and (dq)k/2 = (2πi)k/2 qk/2 (dτ)k/2

which implies
1 (dq)k/2
(dτ)k/2 = ·
(2πi)k/2 qk/2
The order relation relates

ord∞ f = ord( fˆ; q = 0) = ord(p∗ ω f ; τ = ∞)

to
ord[∞] ω f = ord((φ∞−1 )∗ ω f ; q = 0) = ord(h; q = 0)
with
h(q) (dq)k/2 := (φ∞−1 )∗ ω f
The commutative diagram

p = φ −1 ◦ pow∞ implies p∗ = pow∗∞ ◦ (φ∞−1 )∗

On one hand
p∗ ω f = f (τ) (dτ)k/2
On the other hand,
 
pow∗∞ (φ∞−1 )∗ ω f = pow∗∞ h(q) (dq)k/2 = h(q(τ)) · (2πi)k/2 · q(τ)k/2 (dτ)k/2

Hence
f (τ) = h(q(τ)) · (2πi)k/2 · q(k(τ))k/2
and
fˆ(q) = h(q) · (2πi)k/2 · qk/2
which implies
ord( fˆ; q = 0) = ord(h; q = 0) + k/2
3.2 Eisenstein series and the algebra of modular forms of Γ 107

or
ord∞ f = ord[∞] ω f + k/2

• For t ∈ {i, ρ} we recall the chart φt from the following diagram

p|Ut
Ut p(Ut )

λt φt
powt
∆t Wt

The map
'
λt : Ut −
→ ∆t
is biholomorphic. For the order calculation we assume w.l.o.g.

λt = id and Ut = ∆t

The isotropy group Γt acts on Ut by multiplication by an e-th root of unity with

e := ht

For τ ∈ Ut
q(τ) := powt (τ) := τ e
Therefore

dq = e · τ e−1 dτ and (dq)k/2 = ek/2 · τ (k/2)·(e−1) (dτ)k/2

If
ω f (q) = g(q) (dq)k/2
then

(p∗ ω f )(τ) = ω f (q(τ)) = g(q(τ)) · ek/2 · τ (k/2)·(e−1) (dτ)k/2 = f (τ) (dτ)k/2

As a consequence
f (τ) = g(q(τ)) · τ (k/2)·(e−1) · ek/2
By definition
ord[t] ω f = ord0 g
Hence

ordt f = e · ord0 g + (k/2) · (e − 1) = e · ord[t] ω f + (k/2) · (e − 1) =

108 3 The algebra of modular forms

= ht · ord[t] ω f + (k/2) · (ht − 1)

Summing up

ordt f = ord[t] ω f if t ∈ D \ {i, ρ, ∞}

ordi f = 2 · ord[i] ω f + k/2
ordρ f = 3 · ord[ρ] ω f + k
ord∞ f = ord[∞] ω f + k/2
or, because for all t ∈ H∗ holds ordt f ≥ 0:

ord[t] ω f = ordt f ≥ 0 if t ∈ D \ {i, ρ, ∞}

ord[i] ω f = 1/2 · ordi f − k/4 ≥ −k/4, hence ord[i] ω f ≥ −bk/4c

ord[ρ] ω f = 1/3 · ordρ f − k/3, hence ord[ρ] ω f ≥ −bk/3c
ord[∞] ω f = ord∞ f − k/2, hence ord[∞] ω f ≥ −k/2
which finishes the proof of the order relations. As a consequence
( $ % $ % )

⊗k/2
 k k k
α(Mk (Γ )) ⊂ ω ∈ H X, M ⊗O Ω
0
: div ω ≥ − · [i] − · [ρ] − · [∞]
4 3 2

Conversely, each ω ∈ H 0 (X, ΩΓ ,k ) defines the modular form f ∈ Mk (Γ ) with

f (τ) (dτ)k/2 = p∗ ω

iii) Line bundle: It is well known that the sheaves of multiples of a divisor are line
bundles. The sheaf of multiples of the sum of two divisors is the tensor product of
the sheaves of multiples of each summand, see [63], q.e.d.

The formulas in the proof of Theorem 3.17, part 2) show: While a modular
form f on H∗ is holomorphic when considered as a function, the representing dif-
ferential form ω f on the modular curve X may have poles. Actually, it is the divisor
of ω f which encodes the information about f .

In the language of line bundles, modular forms of weight = k are the sections of
the line bundle LΓ ,k on the modular curve:

Mk (Γ ) ' H 0 (X, LΓ ,k )
3.2 Eisenstein series and the algebra of modular forms of Γ 109

Example 3.18 (Cusp forms S2 (Γ )).

1. There are no cusp forms of weight = 2, i.e.

S2 (Γ ) = 0.

Proof. Each cusp form

f ∈ Sk (Γ )
is holomorphic on H∗ and vanishes at ∞ ∈ H∗ . Hence

ordτ f ≥ 0, τ ∈ H, and ord∞ f ≥ 1.

The order relations from Theorem 3.17 imply

ω f := α( f ) ∈ H 0 (X(1), ΩΓ ,k )

satisfies for τ ∈ D \ {[i], [ρ], [∞]}

ord[τ] ω f = ord[τ] f ≥ 0,

in particular for k = 2:

ord[i] ω f = (1/2)(ordi f − 1) ≥ −(1/2), hence ord[i] ω f ≥ 0

ord[ρ] ω f = (1/3)(ordρ f − 2) ≥ −(2/3), hence ord[ρ] ω f ≥ 0

ord[∞] ω f = ord∞ f − 1 ≥ 0
Therefore
ΩΓ ,2 ' Ω
and ω f is holomorphic, i.e.
ω f ∈ H 0 (X, Ω )
The group H 0 (X, Ω ) vanishes because

dim H 0 (X, Ω ) = g(X) = g(P1 ), q.e.d.

2. The formula generalizes to Hecke congruence subgroups as the isomorphy of

vector spaces:
S2 (Γ0 (N)) ' H 0 (X0 (N), Ω )
which implies
dim S2 (Γ0 (N)) = g(X0 (N)),
see Corollary 3.28.

The calculation from Example 3.18 carries over to arbitrary even weights k ∈ Z.
110 3 The algebra of modular forms

Corollary 3.19 (Dimension of Mk (Γ )). The vector spaces Mk (Γ ) of modular forms

of even weight k ≥ 0 have the following dimensions:
(
bk/12c k ≡ 2 mod 12
dim Mk (Γ ) =
bk/12c + 1 otherwise

For the dimension of the vector space Sk (Γ ) of cusp forms see Lemma 3.23.
Proof. We have
Mk (Γ ) ' H 0 (X, LΓ ,k ).
For k = 0 we have on the compact Riemann surface X ' P1

dim M0 (Γ ) = dim H 0 (X, O) = 1.

Therefore we are left with even k > 0:

i) Serre duality and Riemann-Roch theorem: By Serre-duality for each k > 0
dim H 1 (X, LΓ ,k ) = dim H 0 (X, LΓ∨,k ⊗ O OK ) = dim H 0 (X, O−(k/2)·K−Dk +K )
Now
deg (−(k/2) · K − Dk + K) = ((k/2) − 2) − bk/4c − bk/3c ≤
≤ ((k/2)−2)−((k/4)−1)−((k/3)−1) = k/2−k/4−k/3 = k ·(1/2−1/4−1/3) < 0
As a consequence

0 = dim H 0 (X, LΓ∨,k ⊗ O OK ) = dim H 1 (X, LΓ ,k )

The Riemann-Roch theorem implies for LΓ ,k = O(k/2)K+Dk

dim Mk (Γ ) = dim H 0 (X, LΓ ,k ) = 1 − g(X) + (k/2)(−2) + bk/4c + bk/3c + k/2 =

= 1 − k + bk/4c + bk/3c + k/2 = 1 − (k/2) + bk/3c + bk/4c

ii) Evaluating the formula for dim Mk (Γ ): We consider the different values of
even k mod 12 separately. We make the ansatz

k = α · 12 + β , β ∈ {0, 2, ..., 10}

and compute for all values of β

1 − (k/2) + bk/3c + bk/4c

• Residue β = 0:

1 − 6α + 4α + 3α = 1 + α = 1 + bk/12c

• Residue β = 2:

1−(6α +1)+b4α +2/3c+b3α +2/4c = 1−(6α +1)+4α +3α = α = bk/12c

3.2 Eisenstein series and the algebra of modular forms of Γ 111

• Residue β ∈ {4, 6, 8, 10}:

1 − (6α + β /2) + b4α + β /3c + b3α + β /4c =

= 1−6α −β /2+4α +bβ /3c+3α +bβ /4c = 1+α +(−β /2+bβ /3c+bβ /4c) =
= 1 + α = 1 + bk/12c
q.e.d.

In particular
dim M0 = 1, dim M2 = 0, dim M4 = dim M6 = dim M8 = dim M10 = 1, dim M12 = 2.
The non-zero vector spaces in the last line are spanned by the Eisenstein series G4 , G6
and their products. Only M12 contains a non-zero cusp form, namely the discrimi-
nant modular form ∆ .

Apparently the numerical value dim H 0 (X, LΓ ,k ) can be obtained without applying
the theorem of Riemann-Roch and Serre duality in the proof of Corollary 3.19: On
the modular curve X ' P1 each line bundle L is determined by its degree m, and
for m ≥ 0
dimH 0 (P1 , L ) = m + 1.
We gave the proof in the above form, because Theorem 3.26 will generalize the
proof to the modular curves X0 (N) of the Hecke congruence subgroups.

Corollary 3.20 (Weight formula for modular forms). Consider an even integer k ≥ 4
and a modular form f ∈ Mk (Γ ). Then the sum of the orders of f satisfies the follow-
ing equation:

1 1 0 k
ord∞ f + ordi f + ordρ f + ∑ ordτ f =
2 3 τ 12

Here the symbol ∑0τ denotes the summation over all points τ ∈ D \ {i, ρ}.

Proof. We recall the relation between the order of f ∈ Mk (Γ ) at a point τ and the
order of the differential form ω f ∈ H 0 (X, M ⊗O Ω ⊗k/2 ) at the point [τ] ∈ X estab-
lished during the proof of Theorem 3.17:

ord∞ f = ord[∞] ω f + k/2

ordi f = 2 ord[i] ω f + k/2

ordρ f = 3 ord[ρ] ω f + k
ordτ f = ord[τ] ω f , τ ∈ D \ {i, ρ, ∞}.
112 3 The algebra of modular forms

The modular curve X has genus g = 0 according to Corollary 3.16. Therefore the
line bundle ΩX has degree = −2, and the tensor product ΩX ⊗k/2 has degree

= (−2) · k/2 = −k

which implies
deg div(ω f ) = −k.
As a consequence

1 1 0
ord∞ f + ordi f + ordρ f + ∑ ordτ f =
2 3 τ
! ! !
k k k 0
= ord[∞] ω f + + ord[i] ω f + + ord[ρ] ω f + + ∑ ord[τ] ω f =
2 4 3 τ

k k k k k k k
= deg div(ω f ) + + + = −k + + + = , q.e.d.
2 4 3 2 4 3 12

Corollary 3.21 (Specific values and Fourier coefficents of the modular invariant j).

1. The modular invariant

− P1
j : X(1) →
attains the distinguished values

j([i]) = 1728 and j([ρ]) = 0.

2. All Fourier coefficents (bn )n≥−1 of the modular invariant

∞
j(q) = ∑ bn · qn
n=−1

are integers, and b−1 = 1.

Proof. 1. Corollary 3.20 implies G4 (ρ) = 0 and G6 (i) = 0. The definition

g2 3
j = 1728 ·
g2 − 27g3 2
3

implies the values of j at the distinguished orbits.

2. To compute the Fourier expansion of

g32
j = 26 · 33 ·
∆
3.2 Eisenstein series and the algebra of modular forms of Γ 113

we use the definition

22 4 26
g2 = · π · E4 , hence g32 = 3 · π 12 · E43 ,
3 3
and secondly the formula from Proposition 3.14

∆ (τ) ∞
12
= ∑ τn · qn ∈ Z{q}, q = e2πi·τ .
(2π) n=1

We obtain
!
  g3   26 3 12 1
j = 26 · 33 · 2 = 26 · 33 · ·E ·π · =
∆ 33 4 ∆

E43 E43 1 E43

= = = · =
∆ /(2π)12 ∞
τ · qn q 1+ ∞
τ · qn−1
∑ n ∑ n
n=1 n=2
!m
1 ∞ ∞ 1
= · E43 · ∑ (−1)m · ∑ τn · qn−1 = · (1 + O(q))
q m=0 n=2 q
by applying the formula of the geometric series and using

E4 (q) = 1 + O(q)

from Corollary 3.11 and the fact, that E4 has integer Fourier coefficents, q.e.d.

Corollary 3.22 (The field of automorphic functions).

1. If an automorphic function f ∈ A0 (Γ ) is holomorphic on H with Fourier series

f (τ) = ∑ cn · qn , c−k 6= 0,
n≥−k

for a suitable k ∈ N, then a poylnomial P(X) ∈ C[X] of degree k exists with

• f = P( j),

• and all coefficients of P(X) are Z-linear combinations of the Fourier coeffi-
cients c−k , ..., c−1 , c0 .
2. The field of automorphic functions, which equals the field of meromorphic func-
tions on the modular curve X, is

A0 (Γ ) = C( j).
114 3 The algebra of modular forms

Thanks to Corollary 3.16 the second part of Corollary 3.22 is a disguised version of
the well-known fact that the field of meromorphic functions on P1 is C(z), the field
of rational functions in one complex variable.
Proof. 1. The proof goes by induction on k: If k = 0 than f is a modular func-
tion, hence equals the constant c0 . For the induction step k − 1 7→ k consider the
function
g := f − c−k · jk .
The function g has at most a pole of order = k − 1 at ∞. Therefore the induction
assumption applies to g.

2. Each automorphic function f ∈ A0 (Γ ) can be considered a meromorphic function f ∈ M (P1 )

on the projective space. Denote by

A := {x1 , ..., xn } ⊂ H ∪ {∞} and B = {y1 , ..., yn } ⊂ H ∪ {∞}

the respective zero set and pole set of f , counted with multiplicity. Both sets have
the same cardinality because a non-constant holomorphic map between compact
Riemann surfaces attains each value with the same multiplicity, see [63]. Then
n j − j(xν )
g := ∏ ∈ A0 (Γ )
ν=1 j − j(yν )

has the same zero set and pole set as f . The quotient f /g has no poles and zeros
on H ∪ {∞}. According to part 1) a constant a ∈ C exists with

f
a=
g
or
f = a · g ∈ C( j), q.e.d.

Lemma 3.23 (Exact sequence for cusp forms). Denote by

ε : Mk (Γ ) → C, f 7→ f (∞),

the evaluation with kernel Sk (Γ ) and by

µ∆ : Mk−12 (Γ ) → Mk (Γ ), f 7→ f · ∆ ,

the multiplication with the discriminant form. Then for even k ≥ 4: The following
sequence of vector space morphisms is exact
∆ µ ε
0 → Mk−12 (Γ ) −→ Mk (Γ ) →
− C → 0.

In particular:
3.2 Eisenstein series and the algebra of modular forms of Γ 115

• For even k ≥ 4:
Mk (Γ ) = Sk (Γ ) ⊕ C · Gk ,
and the multiplication by the discriminant modular form ∆ defines a vector space
isomorphism
'
µ∆ : Mk−12 (Γ ) −
→ Sk (Γ ).
• For k < 0:
dim Mk = 0
As a consequence, the formulas from Corollary 3.19 allow to compute the di-
mension of Sk (Γ ). A partial generalization of Lemma 3.23 will be proved in
Corollary 3.28.
Proof. i) Exactness on the right-hand side: For even k ≥ 4 the Eisenstein series Gk ∈ Mk (Γ )
satisfies
Gk (∞) = 2 · ζ (k) 6= 0.
Therefore ε is surjective.
ii) Exactness in the middle: If f ∈ Mk (Γ ) with ε( f ) = 0, i.e.
f ∈ Sk (Γ )
then the quotient
f
g :=
∆
is holomorphic on H, because ∆ has no zero on H due to Proposition 1.19. In
addition g is holomorphic at the point ∞ because ∆ has a zero of order 1 at ∞ ac-
cording to Proposition 3.14. Hence g ∈ Mk−12 (Γ ). In addition, the composition

ε ◦ µ∆ = 0

is obvious.
iii) Exactness on the left-hand side: Apparently the map µ∆ is injective.

iv) As a consequence

dim Sk (Γ ) = 0 for k ∈ {0, 2, 4, 6, 8, 10}.

Then iteratively for all even k < 0

dim Mk (Γ ) = 0

because for all even k ∈ Z

∆µ
Mk−12 (Γ ) −→ Sk (Γ )
is injective, q.e.d.
116 3 The algebra of modular forms

The result dim Mk (Γ ) = 0 if k < 0 generalizes to Hecke congruence subgroups:

For k < 0
dim Mk (Γ0 (N)) = 0,
see [66, Prop. 3].

Theorem 3.24 (Algebra of modular forms). The associative algebra

!
M
M∗ (Γ ) := M2k (Γ ), +, ·
k≥0

of modular forms of the module group Γ is commutative and freely generated by the
two Eisenstein series
G4 ∈ M4 (Γ ) and G6 ∈ M6 (Γ ).
The subalgebra of cusp forms
M
S∗ (Γ ) := S2k (Γ ) ⊂ M∗ (Γ )
k≥0

is a principal ideal generated by ∆ ∈ S12 (Γ ).

Proof. One checks for k 6= k0

M2k (Γ ) ∩ M2k0 (Γ ) = {0}.

i) Generated by G4 and G6 : We prove the claim about Mk (Γ ) by reducing succes-

sively the weight. The even integer k ≥ 4 decomposes as

k = 4·α +6·β

with non-negative integers α, β ∈ Z. Hence

 
β β
Gα4 · G6 ∈ Mk and ε Gα4 · G6 6= 0

Consider an arbitrary element f ∈ Mk . For suitable λ ∈ C

   
β β
ε f − λ · Gα4 · G6 = f − λ · Gα4 · G6 (∞) = 0.

According to Lemma 3.23 an element g ∈ Mk−12 exists with

β
f − λ · Gα4 · G6 = ∆ · g.

The claim follows by repeating the argument with g ∈ Mk−12 (Γ ), and using

M4 (Γ ) = C · G4 and M6 (Γ ) = C · G6
3.3 Generalization to Hecke congruence subgroups Γ0 (N) 117

due to Corollary 3.19.

ii) Freely generated: The generators G4 and G6 are even algebraically independent.
Assume the existence of a polynomial P ∈ C[T1 , T2 ] with

P(G4 , G6 ) = 0.

We may assume that all monomials of P(G4 , G6 ) have the same weight. The case
∗
P(G4 , G6 ) = a · Gm
4 + G6 · P̂(G4 , G6 ), a ∈ C ,

is excluded because G4 (i) 6= 0, but G6 (i) = 0. Analogously the case

∗
P(G4 , G6 ) = b · Gm
6 + G4 · P̂(G4 , G6 ), b ∈ C ,

is excluded because G6 (ρ) 6= 0, but G4 (ρ) = 0. Hence the product G4 G6

divides P(G4 , G6 ), i.e.

P(G4 , G6 ) = G4 G6 · P̂(G4 , G6 ) = 0.

Iteration of the argument for

P̂(G4 , G6 ) = 0 with deg P̂ < deg P

proves the claim P = 0 by descending induction.

iii) Ideal of cusp forms: According to Lemma 3.23 any cusp form is a multiple of ∆ ,
q.e.d.

3.3 Generalization to Hecke congruence subgroups Γ0 (N)

In an analogous way to Theorem 3.17, also modular forms of a Hecke congruence

subgroup Γ0 (N), N ∈ N, are meromorphic differential forms on the corresponding
modular curve X0 (N). Essentially, one has to take into account that a congruence
subgroup has more than one cusp and that cusps may have distinct width. Therefore
the width of the cusps enters as an additional parameter when determining local
coordinates around the cusps of the modular curve. In addition, one cannot use a
standard atlas like on X = P1 but one has to deal with atlases in a generic way.

Recall from Remark 2.30 the branched covering

− X(Γ ) ' P1
f : X0 (N) →
of degree
118 3 The algebra of modular forms
!
1
deg f = N · ∏ 1+
p|N
p

and the genus formula

deg f ε2 ε3 ε∞
g(X0 (N)) = 1 + − − −
12 4 3 2
Moreover, we recall the canonical projection from the extended half-plane

p : H∗ →
− X0 (N) := Γ0 (N)\H∗

The aid to represent modular forms of Mk (Γ0 (N)), which are defined as functions
on H∗ , as meromorphic differential forms on the modular curve X0 (N) are the divi-
sors
$ % $ %
k k k
Dk := · ∑ x+ · ∑ x + · ∑ x ∈ Div(X0 (N)), even k ≥ 0.
4 x∈X (N) 3 x∈X (N) 2 x∈X (N)
0 0 0
x elliptic x elliptic x cusp
hx =2 hx =3

They have degree $ % $ %

k k k
deg Dk = · ε2 + · ε3 + · ε∞
4 3 2
For each fixed even k ≥ 0 the divisor Dk singles out those meromorphic differential
forms from  
H 0 X0 (N), M ⊗ O Ω ⊗k/2

which arise from modular forms under the canonical injection

 
− H 0 X0 (N), M ⊗ O Ω ⊗k/2 , g 7→ ωg ,
α : Mk (Γ0 (N)) →

with ωg the well-defined differential form satisfying

p∗ ωg (τ) = g(τ) (dτ)k/2 .

Lemma 3.25 (Order relation). Consider a Hecke congruence subgroup Γ0 (N) and
an even weight k ≥ 0.
1. Definition of α: The map
 
− H 0 X0 (N), M ⊗ O Ω ⊗k/2 , g 7→ ωg ,
α : Mk (Γ0 (N)) →

with ωg the well-defined differential form satisfying

p∗ ωg (τ) = g(τ) (dτ)k/2

3.3 Generalization to Hecke congruence subgroups Γ0 (N) 119

is well-defined.

2. Order relation: The order of a modular form g and the order of the meromorphic
differential form
ωg = α(g)
relate as 
ord[τ] ωg + k/2
 [τ] cusp
ordτ g =

hτ · ord[τ] ωg + k/2 · (hτ − 1) τ ∈ H


Proof. Each modular form g ∈ Mk (Γ0 (N)) is a holomorphic function on H. The

function g defines on H the Γ0 (N)-invariant holomorphic differential form

g(τ) (dτ)k/2

Therefore g(τ) (dτ)k/2 defines on the open Riemann surface

Y0 (N) := p(H)

a meromorphic differential form

 
ωg ∈ H 0 Y0 (N), M ⊗ Ω ⊗k/2

with singularities at most the finitely many orbits of elliptic points τ ∈ H.

Moreover Γ0 (N) has only finitely many cusps, see Example 2.11. For
arbitrary q ∈ Q and
β ∈ Γ with q = β (∞)
each modular form g ∈ Mk (Γ0 (N)) has a well-defined order

ordq g := ord∞ g[β ]k

independent from the choice of β . Hence ωg extends as meromorphic differential

form to the compactified modular curve
 
⊗k/2
ωg ∈ H 0 X0 (N), M ⊗ ΩX (N)
0

A cumbersome calculation by using charts, see [17, Sect. 3.3.], shows that the order
relations from the proof of Theorem 3.17 carry over: These formula show that the
map α is well-defined. The map α is injective, and via pullback of differential forms
also surjective, q.e.d.
120 3 The algebra of modular forms

On a compact Riemann surface line bundles and divisors correspond to each

other bijectively, see [63]. Theorem 3.26 introduces the line bundles of meromorphic
differential forms induced by the distinguished divisors Dk on the modular curves.
Theorem 3.26 (The line bundle of modular forms). Consider a Hecke congru-
ence subgroup
Γ0 (N), N ∈ N,
and an even weight k ≥ 0.
On the modular curve X0 (N) with canonical divisor K ∈ Div(X0 (N)) the line
bundle
LN,k := O(k/2)·K+Dk ∈ Pic(X0 (N))
with the divisor
$ % $ %
k k k
Dk := · ∑ x+ · ∑ x + · ∑ x ∈ Div(X0 (N))
4 x∈X (N) 3 x∈X (N) 2 x∈X (N)
0 0 0
x elliptic x elliptic x cusp
hx =2 hx =3

of degree $ % $ %
k k k
deg Dk = · ε2 + · ε3 + · ε∞ ≥ 0
4 3 2
satisfies:
• Meromorphic differential forms:

LN,k ' ΩN,k

with the subsheaf of meromorphic differential forms on X0 (N)

ΩN,k ⊂ M ⊗ O Ω ⊗k/2

defined by

ΩN,k (U) := {ω ∈ H 0 (U, M ⊗ O Ω ⊗k/2 ) : ω 6= 0 and div ω ≥ −Dk |U} ∪ {0}

for each open U ⊂ X0 (N)

• Global sections: Consider the canonical map

− H 0 (X0 (N), M ⊗ O Ω ⊗k/2 ), g 7→ ωg ,

α : Mk (Γ0 (N)) →

with ωg the unique differential form with

p∗ ωg = g(τ) (dτ)k/2 .

Here p : H∗ →
− X0 (N) denotes the canonical projection. Then
3.3 Generalization to Hecke congruence subgroups Γ0 (N) 121

H 0 (X0 (N), LN,k ) ' H 0 (X0 (N), ΩN,k ) =

h  i
− H 0 X0 (N), M ⊗ O Ω ⊗k/2
= im α : Mk (Γ0 (N)) →

• Cohomology: For k ≥ 2
H 1 (X0 (N), LN,k ) = 0
• Dimension of vector space of sections: For k ≥ 2

dim H 0 (X0 (N), LN,k ) = (g(X0 (N)) − 1) · (k − 1) + deg Dk ,

and for k = 0

LN,0 ' O, hence dim H 0 (X0 (N), LN,0 ) = 1.

Proof. We set
g := g(X0 (N))
i) Estimating the degree of (k/2) · K + Dk : By definition

LN,k = O(k/2)·K+Dk .

Using the estimates

bk/4c ≥ (k/4) − 1/2 and bk/3c ≥ (k/3) − 2/3

and the genus formula for X0 (N) we obtain for the degree of the corresponding
divisor for k ≥ 2
! !
k k−2 k−2 k
deg ((k/2) · K + Dk ) ≥ · (2g − 2) + · ε2 + · ε3 + · ε∞ =
2 4 3 2
!
k−2 ε2 2
= (2g − 2) + · 2g − 2 + + · ε3 + ε∞ + ε∞ =
2 2 3
!
k − 2 deg f
= (2g − 2) + · + ε∞ > 2g − 2
2 6

ii) Vanishing of dim H 1 (X0 (N), LN,k ): We apply Serre’s duality theorem in the form
∨
H 1 (X0 (N), LN,k ) = H 0 (X0 (N), LN,k ⊗ O Ω )∨

and estimate for even k ≥ 2 due to part i)

deg (−((k/2) · K + Dk ) + K) < (2 − 2g) + (2g − 2) = 0

Hence
∨
dim H 1 (X0 (N), LN,k ) = dim H 0 (X0 (N), LN,k ⊗OΩ) = 0
122 3 The algebra of modular forms

iii) Riemann-Roch theorem for LN,k : The Riemann-Roch theorem proves the di-
mension formula

dim H 0 (X0 (N), LN,k ) = 1 − g + deg ((k/2) · K + Dk ) =

= 1 − g + deg ((k/2) · K) + deg Dk = (k − 1)(g − 1) + deg Dk , q.e.d.

Remark 3.27 (Modular forms of Hecke congruence subgroups). Figure 3.1 shows
different modular curves as coverings of P1 and the dimension of their vector spaces
of modular forms, see PARI file ”Congruence subgroup 19”. The last column shows
the numerical value
dim H 0 (X(Γ0 (N)), LN,k )
computed by applying the Riemann-Roch theorem to the line bundle LN,k .

Fig. 3.1 Dimension formulas for modular forms of weight k = 4 of Γ0 (pn )

Apparently
M0 (Γ0 (N)) ' C, N ∈ N,
which implies
3.3 Generalization to Hecke congruence subgroups Γ0 (N) 123

dim M0 (Γ0 (N)) = 1 and dim S0 (Γ0 (N)) = 0.

Corollary 3.28 (Codimension of cusp forms of Hecke congruence subgroups).

Consider a congruence subgroup

Γ0 (N), N ∈ N.

1. For even weight k ≥ 4 the subspace of cusp forms

Sk (Γ0 (N)) ⊂ Mk (Γ0 (N))

has codimension = ε∞ , the number of cusps of X0 (N). In particular, the codimen-

sion does not depend on the weight k.

2. For weight k = 2 cusp forms are exactly the holomorphic differential forms on
the modular curve. Hence

dim S2 (Γ0 (N)) = g(X0 (N)).

For the computation of ε∞ in the general case see Example 2.11.

Proof. Set
g := g(X0 (N))
and consider the divisor Dk ∈ Div(X0 (N)) from Theorem 3.26, and define the divisor

D∞
k := ∑ (−1) · x + Dk ∈ Div(X0 (N))
x∈X0 (N)
x cusp

and the correponding line bundle

LN,k
∞
= O(k/2)·K+D∞k ∈ Pic(X0 (N)

with
deg LN,k
∞
= −ε∞ + deg LN,k .
The order relations imply for even k ≥ 2

H 0 X0 (N), LN,k
∞


=

= ωh ∈ H 0 (X0 (N), ΩN,k ) : ordx ωh ≥ (k/2) − 1 for all cusps x ∈ X0 (N) =



= {h ∈ Mk (Γ0 (N)) : ordt h ≥ 1 for all t ∈ H∗ with [t] cusp} = Sk (Γ0 (N))

1. k ≥ 4: The formula from the proof of Theorem 3.26 shows for even k ≥ 4
124 3 The algebra of modular forms

k − 2 deg f
deg LN,k
∞
≥ 2g − 2 + · > 2g − 2
2 6
 
because deg f ≥ 1. Hence H 1 X0 (N), LN,k
∞ = 0 because

∞ ∨
deg (LN,k ) ⊗ O Ω 1 < −(2g − 2) + (2g − 2) = 0,

and the Riemann-Roch theorem implies

dim Sk (Γ0 (N)) = dim H 0 X0 (N), LN,k

∞


=

= dim H 0 X0 (N), LN,k − ε∞ = dim Mk (Γ0 (N)) − ε∞ .

2. k = 2: The explicit form of D∞

2 shows

D∞
2 = 0 ∈ Div(X0 (N)),

which implies
LN,2
∞
= OK = Ω 1
and
S2 (X0 (N)) = H 0 (X0 (N), Ω ), q.e.d.
Chapter 4
Elliptic curves

The present chapter ties quite closely with Section 2.1 on tori. We now bridge com-
plex analysis and algebraic geometry.
Section 4.1 shows that each complex torus is an algebraic variety in the complex
projective space P2 . The variety is the zero set of a cubic polynomial.
Section 4.2 introduces elliptic curves as projective-algebraic varieties of genus = 1
defined over arbitrary subfields of C. We show that elliptic curves correspond bijec-
tively to the smooth zero set of cubic polynomials, the Weierstrass polynomials. The
passage to algebraic geometry allows to investigate the points of elliptic curves with
coordinates in distinguished subfields of C.

4.1 Embedding tori as plane cubic hypersurfaces

Generalizing the 1-dimensional projective space P1 we introduce the higher-dimensional

complex projective spaces Pn .

Definition 4.1 (n-dimensional projective space). For n ∈ N consider the quotient

Pn := (Cn+1 \ {0})/ ∼

with respect to the equivalence relation

z = (z0 , ..., zn ) ∼ w = (w0 , ..., wn ) :<=> ∃λ ∈ C∗ : w = λ · z

and with the canonical projection onto equivalence classes

π : Cn+1 \ {0} →
− Pn , z 7→ [z].

For z = (z0 , ..., zn ) ∈ Cn+1 \ {0} the expression

125
126 4 Elliptic curves

(z0 : ... : zn ) := π(z) ∈ Pn

is named the homogeneous coordinate of π(z). The quotient topology is named

the Euclidean topology of Pn . The space Pn is named the n-dimensional complex
projective space.
One considers Cn ⊂ Pn via the injection
Cn ,−
→ Pn , (z1 , ..., zn ) 7→ (1 : z1 : ... : zn ).

Definition 4.2 (Standard atlas of Pn ). The standard atlas of Pn is the family

U := (Ui )i=0,...,n

with the open sets

Ui := {(z0 : ... : zn ) ∈ Pn : zi 6= 0}, i = 0, ..., n.

and for each i = 0, ..., n as i-th standard chart the homeomorphism

zi−1 zˆi zi+1

!
n z0 zn
φi : Ui →
− C , (z0 : ... : zn ) 7→ , ..., , , , ..., .
zi zi zi zi zi

Here the “hat” indicates omission of the term. The transformation between two
charts φ j and φi with i < j is the map

ψi j := φi ◦ φ j−1 : φ j (Ui ∩U j ) →
− φi (Ui ∩U j ),
!

w0 wi−1 wi+1 w j−1 1 w j+1 wn
w0 , ..., w j−1 , 1̂, w j+1 , ..., wn 7→ , ..., , 1̂, , ..., , , , ...,
wi wi wi wi wi wi wi

Theorem 4.3 shows once more the importance of the Weierstrass function ℘ and of
its differential equation for ℘ 0 in the theory of complex tori.
Theorem 4.3 (The embedding theorem for tori). Consider a complex torus

T := C/Λ

with lattice
Λ = Zω1 + Zω2 .
Set
g2 := 60 · GΛ ,4 , g3 := 140 · GΛ ,6 ,
and denote by ℘ = ℘Λ the Weierstrass function of Λ . The map
4.1 Embedding tori as plane cubic hypersurfaces 127
(
2 (1 : ℘(p) : ℘ 0 (p)) if p 6= 0
φ : T → P , p 7→
(0 : 0 : 1) if p = 0

maps the torus T biholomorphically onto the cubic hypersurface E ⊂ P2 defined by

the cubic polynomial

Fhom (X0 , X1 , X2 ) = X22 X0 − (4X13 − g2 · X1 X02 − g3 · X03 ) ∈ C[X0 , X1 , X2 ],

which is the homogenization of

F(X,Y ) = Y 2 − (4X 3 − g2 X − g3 ) ∈ C[X,Y ].

Proof. The subsequent proof of Theorem 4.3 relies heavily on the properties of ℘
and ℘ 0 from Section 1.2.

i) Equation of φ (T ): The differential equation of the Weierstrass function ℘

(℘ 0 )2 = ℘3 − g2 ·℘− g3
implies

φ (T \ {0}) ⊂ Ea f f := {(x, y) ∈ C2 : y2 = 4 · x3 − g2 · x − g3 }

The closure of the affine algebraic set

Ea f f ⊂ C2 ⊂ P2

is the projective algebraic variety E ⊂ P2 , see Figure 4.1, the zero set in P2 of

Fhom (X0 , X1 , X2 ),

cf. [29, Chap. I, Exerc. 2.9].

128 4 Elliptic curves

Fig. 4.1 Real points of Ea f f

ii) Holomorphy of φ : To check holomorphy in a neighbourhood of 0 ∈ T we choose

an open neighbourhood U ⊂ T of the origin such that ℘ 0 has no zeros in U. On
U ∗ := U \ 0
one obtains by diving the homogeneous coordinate by ℘ 0

φ |U ∗ = (1 : ℘ : ℘ 0 ) = (1/℘ 0 : ℘/℘ 0 : 1),

which converges with limit

lim φ (z) = (0 : 0 : 1) := O ∈ P2
z→0

because ℘ and ℘ 0 have poles at 0 of order respectively = 2 and 3. Hence 0 ∈ T is a

removable singularity of φ .

iii) Injectivity of φ : The holomorphic map

− P1
℘: T →

is a branched covering of degree = 2, while the holomorphic map

℘0 : T →
− P1

is a branched covering of degree = 3. To give an indirect proof of the injectivity

of φ assume two distinct points
4.1 Embedding tori as plane cubic hypersurfaces 129

z 6= w ∈ T \ {0}

with
(℘(z),℘ 0 (z)) = (℘(w),℘ 0 (w)) ∈ C2 .
Then
℘(z) = ℘(w) implies w = −z
due to Corollary 1.10 and because ℘ is an even function due to Theorem 1.12.
Moreover

℘ 0 (z) = ℘ 0 (w) = ℘ 0 (−z) = −℘ 0 (z) implies ℘ 0 (z) = 0.

Lemma 1.13 implies that

z = [ω/2], ω ∈ Λ and w = −z ∈ T

are two distinct classes of half-period points of Λ . But ℘ attains at z the value ℘(z)
with multiplicity = 2, hence ℘ assumes the value ℘(z) = ℘(w) with
multiplicity ≥ 4, a contradiction.

iv) The image of φ : The only non-affine point is

E \ Ea f f = O = φ (0).

Consider an affine point

(1 : x : y) ∈ E.
The ℘-function is a non-constant holomorphic map

− P1 ,
℘: T →

in particular it is surjective: There exists z ∈ T with

℘(z) = x.

The function ℘ is even, therefore also

℘(−z) = x.

From
y2 = 4x3 − g2 · x − g3 = 4℘(z)3 − g2 ·℘(z) − g3 = ℘ 0 (z)2
follows:
• Either
y = ℘ 0 (z) and φ (z) = (1 : x : y).
• Or
y = −℘ 0 (z) = ℘ 0 (−z) and φ (−z) = (1 : x : y)
130 4 Elliptic curves

v) Smoothness of E:
• A point (x0 , y0 ) ∈ C2 is a singular point of Ea f f iff it satisfies the three equations

0 = F(x0 , y0 ) = y20 − (4x03 − Ax0 − B), A := g2 , B := g3 ,

∂F ∂F
(x0 , y0 ) = 2y0 = 0 and (x0 , y0 ) = −12x02 + A = 0
∂y ∂x
Introducing the cubic poynomial

f (x) := 4x3 − Ax − B ∈ C[x]

the condition is equivalent to

f (x0 ) = 0 and f 0 (x0 ) = 0.

The latter condition is equivalent to x0 being a multiple zero of f , i.e. to the

vanishing of the discriminant of f

∆ f = A3 − 27B2 .

Proposition 1.19 implies ∆ f 6= 0.

• To prove the non-singularity of E at the point O = (0 : 0 : 1) ∈ P2 we consider

the standard coordinate of P2 around the point O
!
z0 z1
− C2 , (z0 : z1 : z2 ) 7→ (u, v) :=
φ2 : U2 → , .
z2 z2

We have
φ2 (E ∩U2 ) = {(u, v) ∈ C2 : f (u, v) = 0}
with
f (u, v) := u − (4 · v3 − A · u2 · v − B · u3 ).
Then the partial derivatives are

∂f ∂f
(u, v) = 1 − (2 · A · u · v − 3 · B · u2 ) and (u, v) = −12 · v2 + A · u2
∂u ∂v
hence
∇ f (0, 0) = (1, 0) 6= 0.
The non-singular projective algebraic curve E = φ (T ) ⊂ P2 is connected, hence
a compact Riemann surface with the induced analytic structure as a
hypersurface of P2 .

vi) Biholomorphy: The map

4.1 Embedding tori as plane cubic hypersurfaces 131

φ :T →
− E
between Riemann surfaces is a bijective holomorphic map, hence it is
biholomorphic, q.e.d.

Example 4.4 (Embedding tori as plane cubics). The PARI script

Torus_to_Weierstrass_equation_01 computes a cubic equation in P2
for a given torus T = C/Λ , see Figure 4.2. Real values are approximated by ratio-
nals.

Fig. 4.2 Embedding tori as plane cubics

The embedding result of Theorem 4.3 is a particular case of the general question:
How to obtain holomorphic maps

− Pn
X→

with X a compact a complex manifold X?

The underlying construction is even more general. It is not restricted to complex

analysis. Section 4.2 deals with an example from algebraic geometry. The present
section recalls the general construction in the context of compact Riemann
surfaces X, see also [63]. The pivotal objects are the twisted line bundle on Pn and
very ample line bundles on X. In the following we shall not distinguish between a
line bundle and the invertible sheaf of its holomorphic sections. We shall employ in
both cases the same notation
L ∈ Pic(X).
132 4 Elliptic curves

Definition 4.5 (The twisted line bundle on Pn ). The twisted line bundle O(1)
on Pn is defined with respect to the standard atlas of Pn by the cocycle

g = (gi j ) ∈ Z 1 (U , O ∗ )

with
zj
gi j (z0 : ... : zn ) := , 0 ≤ i 6= j ≤ n.
zi

Proposition 4.6 (Sections of O(1)). The sections of O(1) on Pn correspond bijec-

tively to the linear polynomials HPol(1) ⊂ C[z0 , ...., zn ], i.e. the map

− H 0 (Pn , O(1)), zk 7→ sk , k = 0, ..., n.

HPol(1) →

Here sk ∈ H 0 (Pn , O(1)) is defined with respect to the i-th standard chart by the
holomorphic functions

zk
sk,i : Ui →
− C, sk,i (z0 : ... : zn ) :=
zi
Proof. The map defined above on the standard charts provide a section

sk := (sk,i )i=0,...,n ∈ H 0 (Pn , O(1))

because
z j zk zk
gi j · sk, j = · = = sk,i , q.e.d.
zi z j zi

Proposition 4.6 generalizes to sections of the tensor powers

O(k) := O(1)⊗k , k ∈ N,

cf. also [63].

In order that a line bundle L on a Riemann surface X defines a holomorphic map

− Pn
φL : X →
to a projective space, the line bundle L has to be globally generated.

Definition 4.7 (Base-point, globally generated line bundle). Consider a Riemann

surface X and a line bundle L on X.
4.1 Embedding tori as plane cubic hypersurfaces 133

1. A point x ∈ X is a base-point of L if each section σ ∈ H 0 (X, L ) satisfies σ (x) = 0.

Note
σ (x) = 0 ⇐⇒ σx ∈ mx Lx
with Lx the stalk of L at x ∈ X and

mx ⊂ Ox

the maximal ideal of the stalk Ox of the structure sheaf at x ∈ X.

2. The line bundle is globally generated or base-point-free if it has no base-points,

i.e. if for each x ∈ X exists a section σ ∈ H 0 (X, L ) such that the germ σx ∈ Lx
generates the stalk Lx as Ox -module. The latter condition is equivalent to σ (x) 6= 0.

Proposition 4.8 (Holomorphic maps to projective spaces). Consider a compact

Riemann surface X and a globally generated line bundle L on X. After choosing a
basis
(σi )i=0,...,n ∈ H 0 (X, L )
the map
− Pn , x 7→ (σ0 (x) : ... : σn (x)),
φL : X →
is well-defined, independent from the chosen trivialization of L in a neighbourhood
of x ∈ X, and holomorphic. It induces an isomorphism
∗
φL (O(1)) ' L
∗ (O(1)) the pull-back of the twisted line bundle.
with φL

Proof. We set
φ := φL .
i) Definition: We consider the definition of φ in a suitable open neighbourhood U of x.
On U we choose a trivialization to identify the line bundle L with the struc-
ture sheaf O. Hence sections from H 0 (U, L ) become holomorphic functions.
Because L is globally generated there is an index j ∈ {0, ..., n} with

σ j (x) 6= 0.

Hence the point

(σ0 (x) : ... : σn (x)) ∈ Pn
is well-defined. It is independent of the choice of the chart and of the isomorphism

L |U ' O|U.

Apparently the map φ is holomorphic on U.

134 4 Elliptic curves

ii) Pullback of the twisted line bundle: For each i = 0, ..., n the sets

Xi := φ −1 (Ui ) = {x ∈ X : (σi )x ∈
/ mx Lx }

form an open covering (Xi )i=0,...,n of X because L is globally generated.

For i = 0, ..., n the canonical morphisms between the rings of local sections

O(1)(Ui ) →
− φ∗ (L )(Ui ) = L (Xi ) induced by si |Ui 7→ σi |Xi

define a canonical sheaf morphism on Pn

O(1) →
− φ∗ L

The latter corresponds to a sheaf morphism on X

φ ∗ (O(1)) →
− L,

because the two functors φ ∗ and φ∗ are adjoint, see [29, Chap. II, § 5], i.e. there is a
canonical group isomorphim

HomOX (φ ∗ (O(1)), L ) ' HomOPn (O(1), φ∗ L )

The induced morphism

φ ∗ (O(1)) →
− L
between the two line bundles is an isomorphism because it is an isomorphism on
stalks: If for a given i ∈ {0, ..., n}

x ∈ Xi ⊂ X and y := φ (x) ∈ Ui ⊂ Pn

then according to the definition of the analytic inverse image sheaf

(φ ∗ O(1))x = O(1)y ⊗OPn ,y OX,x →

− Lx , (si )y ⊗ 1 7→ (σi )x , q.e.d.

Theorem 4.9 provides a geometric criterion for the map provided by a globally
generated line bundle L to be a closed embedding.

Theorem 4.9 (Projective embedding induced by a line bundle). Consider a com-

pact Riemann surface X and a globally generated line bundle L on X. Then the
induced map
− Pn
φL : X →
is a holomorphic closed embedding iff L satisfies both of the following properties:
1. Separating points: For any two distinct points p 6= q ∈ X exists a section

σ ∈ H 0 (X, L ) with σ (p) 6= 0 but σ (q) = 0

or vice versa.
4.1 Embedding tori as plane cubic hypersurfaces 135

2. Separating tangent vectors: For all x ∈ X the map

dL ,x : {σ ∈ H 0 (X, L ) : σx ∈ mx Lx } →
− mx Lx /m2x Lx , σ 7→ [σx ],

is surjective.

For the proof of Theorem 4.9 see [63].

Remark 4.10 (Separating points and tangent vectors). We explain the meaning of
the two statements from Theorem 4.9.
1. Separating points: The homogeneous coordinates of a given point in Pn do not
fix the value of the components. But it is determined whether a component is
zero or not zero. Therefore for each pair of points p 6= q ∈ X

φL (p) 6= φL (q)

if at least one component of the left-hand side is different from zero while the
same component on the right-hand side is zero.

2. Separating tangent vectors: A holomorphic mapping is an embedding in the

neighbourhood of a point x ∈ X if the Jacobi map at x does not vanish. The
latter condition is equivalent to the fact that at x the push-forward of tangent vec-
tors is injective, equivalently the pull-back of cotangent vectors is surjective: We
consider the isomorphy of Ox -modules - respectively of complex vector spaces -

mx Lx /m2x Lx ' (mx /m2x ) ⊗Ox Lx = Tx1 ⊗Ox Lx

with the cotangent space

Tx1 := mx /m2x
of X at x. Then the map

dL ,x : {σ ∈ H 0 (X, L ) : σx ∈ mx Lx } →
− Tx1 ⊗Ox Lx

is induced by the total differential dx of holomorphic functions: A section

σ ∈ H 0 (X, L ) with σx ∈ mx Lx

factorizes in a suitable neighbourhood U of x as

σ = f · σ1

with a holomorphic function f ∈ O(U) satisfying fx ∈ mx and a holomorphic

section σ1 ∈ L (U). Then

dL ,x (σ ) = dx f ⊗ σ1 ∈ Tx1 ⊗Ox Lx
136 4 Elliptic curves

dO(1),y
{s ∈ H 0 (Pn , O(1)) : s(y) = 0} TP1n ,y ⊗OPn ,y O(1)y

φ∗ φT∗1

dL ,x
{σ ∈ H 0 (X, L ) : σ (x) = 0} 1 ⊗
TX,x OX,x Lx

Fig. 4.3 Separating tangent vectors

The commutative diagram from Figure 4.3 with

y := φ (x)

and surjective maps

dO(1),y and φ ∗
shows: The pull-back φT∗1 is surjective if and only if dL ,x is surjective.

Definition 4.11 (Very ample line bundle). A globally generated line bundle L on
a compact Riemann surface X is very ample if the induced map

− Pn
φL : X →

is an embedding.

A globally generated line bundle L on X has enough sections to define a holo-

morphic map
− Pn .
φL : X →
If L is very ample then there are enough sections that φL is even an embedding.
Due to the compactness of X its image under an embedding is always closed.

Notation 4.12. For a line bundle L on a Riemann surface X and a divisor D ∈ Div(X)
we denote by
LD := L ⊗O OD
the sheaf of meromorphic sections of L which are multiples of the divisor −D. The
sheaf LD is isomorphic to a line bundle.

Consider a line bundle L on a compact Riemann surface X. Theorem 4.13 states

a numerical criterion for the dimension of the vector spaces
4.1 Embedding tori as plane cubic hypersurfaces 137

H 0 (X, LD ), D ∈ Div(X),

which ensures that the map

− Pn
φL : X →
is well-defined and a closed embedding. This criterion is very helpful in the applica-
tions because the dimension on the vector spaces in question can often be computed
by using the theorem of Riemann-Roch in combination with Serre duality.

Theorem 4.13 (Very-ampleness criterion). Consider a compact Riemann surface X.

For a line bundle L on X are equivalent:
• The line bundle L is very ample.

• For any two not necessarily distinct points p, q ∈ X the corresponding point di-
visors
P := 1 · p, Q := 1 · q ∈ Div(X)
satisfy
dim H 0 (X, L−(P+Q) ) = dim H 0 (X, L ) − 2.

Proof. i) Assume the validity of the dimension formula: The formula implies for any
two point divisors P, Q ∈ Div(X)

H 0 (X, L−(P+Q) ) ( H 0 (X, L−P ) ( H 0 (X, L )

and each proper inclusion has codimension = 1 because it is defined by a single

linear equation.
• The equation
h0 (X, L−P ) = h0 (X, L ) − 1.
states that the kernel of the evaluation

H 0 (X, L ) →
− L p /m p L p ' C, σ 7→ [σ p ],

has codimension = 1. Hence p is not a base-point of L . As a consequence, the

line bundle L is globally generated and

− Pn
φL : X →

is well-defined.

• The equation
H 0 (X, L−(P+Q) ) ( H 0 (X, L−P )
implies that for any pair of distinct points p, q ∈ X there exists a section

σ ∈ H 0 (X, L−P ) \ H 0 (X, L−(P+Q) )

138 4 Elliptic curves

i.e. satisfying
σ (p) = 0 but σ (q) 6= 0.
Hence the sheaf L separates points.

• The dimension formula shows for any point p ∈ X with point divisor P ∈ Div(X)

H 0 (X, L−2P ) ( H 0 (X, L−P )

has codimension = 1. Hence there exists a section

σ ∈ H 0 (X, L−P ) \ H 0 (X, L−2P )

which implies the surjectivity of the composition of the canonical maps

H 0 (X, L−P ) →
− mX,p L p →
− mX,p L p /m2X,p L p

due to
dimC (mX,p /m2X,p ) = 1.
Therefore L separates tangent vectors.

ii) Assume L very ample: Theorem 4.9 implies that L separates points and
tangent vectors. Separating points implies for all point divisors P 6= Q ∈ Div(X)

dim H 0 (X, L−(P+Q) ) = dim H 0 (X, L ) − 2.

Separating tangent vectors implies for each point divisor P ∈ Div(X)

dim H 0 (X, L−2P ) = dim H 0 (X, L ) − 2, q.e.d.

Remark 4.14 (Projective embedding of compact Riemann surfaces). As a conse-

quence of the very-ampleness criterion from Theorem 4.13 one can prove: Each
compact Riemann surface X embeds holomorphically into a projective space.
1. If g(X) ≥ 2 then the tri-canonical line bundle Ω ⊗3 ∈ Pic(X) is very ample,
see [63].

2. For a complex torus X = C/Λ holds Ω ' O, hence

g(X) := dim H 0 (X, Ω ) = 1

Consider the divisor

D := 3 · 0 ∈ Div(X)
with the point 0 ∈ X being the origin. Then the line bundle
4.2 Elliptic curves over subfields of C 139

L := OD ∈ Pic(X)

is very ample according to Theorem 4.13: As a consequence of Ω ' O, for two

point divisors
P = 1 · p, Q = 1 · q ∈ Div(X)
holds
dim H 1 (X, L−(P+Q) ) = dim H 0 (X, O−D+(P+Q) ) = 0
because
deg (−D + (P + Q)) = −3 + 2 = −1 < 0
Hence

dim H 0 X, L−(P+Q) = 1 − g(X) + deg L − 2 = dim H 0 (X, L ) − 2

The Riemann-Roch theorem implies

dim H 0 (X, L ) = dim H 0 (X, OD ) = deg D = 3

Apparently, a basis of H 0 (X, OD ) is the family of linearly independent elements

(1,℘,℘ 0 ),

which gives a second proof for the embedding from Theorem 4.3.

4.2 Elliptic curves over subfields of C

Complex Analysis of several variables or Complex Analytic Geometry on one hand,

and on the other hand Algebraic Geometry are mathematical theories with many
concepts in common and also several concepts in parallel. We compile the following
GAGA-dictionary (Géométrie Algébrique et Géométrie Analytique).

Proposition 4.15 (Dictionary: Complex Analytic Geometry - Algebraic Geom-

etry).
140 4 Elliptic curves

Complex Analytic Geometry Algebraic Geometry

Complex Manifold Smooth variety

Euclidean topology Zariski topology

Compact complex manifold Smooth projective algebraic variety

Compact Riemann surface Projective algebraic curve

Genus

Stein manifold Smooth affine variety

Structure sheaf OX

Holomorphic function Regular function

Meromorphic function Rational function

Local ring OX,x

Line bundle

Very ampleness

Divisor

Sheaf ΩX of differential forms

Cohomology

Riemann-Roch Theorem

Serre Duality

The table relates concepts from complex analytix geometry on the left-hand side
with concepts from algebraic geometry on the right-hand side. But not any compact
manifold, and not even any compact Kaehler manifold has a counter part in Alge-
braic Geometry. Hence not any instance of a concept on the left-hand side of the
table induces a correponding instance on the right-hand side. On both sides there
are objects which cannot be studied by methods from the other side.
4.2 Elliptic curves over subfields of C 141

In Algebraic Geometry a variety is assumed to be irreducible. Irreducibility cor-

responds to connectedness in the case of manifolds. If one skips the requirement of
smoothness one obtains the irreducible reduced complex space as the analogue of a
variety.

Remark 4.16 recalls varieties as the basic objects of classical algebraic geometry.
Varieties are irreducible by definition, hence they are defined by prime ideals. But
different from manifolds varieties may have singularities. Hence varieties are the
analogue of reduced and irreducible complex spaces. We distinguish between their
field k of definition and the algebraic closure k where the points of the variety have
their coordinates. As a topological space varieties will be equipped with the Zariski
topology.

Recall that a field k is perfect if each irreducible polynomial with coefficients

from k is separable, i.e has pairwise distinct roots in an algebraic closure k. All
fields in the present section will be assumed perfect. The relevant examples of per-
fect fields are the fields k of characteristic = 0 or the finite fields k = F p of prime
characteristic = p.

Remark 4.16 (Algebraic variety with structure sheaf). Consider a perfect field k and
denote by k the algebraic closure of k.
1. Variety: An affine algebraic variety X is the zero set
n
X = V (a) ⊂ An (k) = k

of a prime ideal a ⊂ k[X1 , ..., Xn ], i.e.

X := {x ∈ An (k) : f (x) = 0 for all f ∈ a}.

The ring
R := k[X1 , ..., Xn ]/a
is the affine coordinate ring of X.

If a can be generated by polynomials from k[X1 , ..., Xn ], then X is defined over k,

denoted X/k. The smallest subfield of k with this property is named the field of
definition of X.

2. A projective algebraic variety X is the zero set

X = V (a) ⊂ Pn (k)

of a prime ideal
a ⊂ k[X0 , ..., Xn ],
142 4 Elliptic curves

generated by homogeneous polynomials, i.e.

X := {(x = (x0 : ... : xn ) ∈ Pn (k) : f (x) = 0 for all homogeneous f ∈ a}.

3. Zariski topology: Closed sets of An (k) are the affine algebraic subvarieties and
their finite unions. The Zariski topology on an affine algebraic variety X ⊂ An (k)
is the subspace topology induced from the Zariski topology of An (k). A basis of
the Zariski topology of X are the sets

Dh := {x ∈ X : h(x) 6= 0}, h ∈ R.

Closed sets of Pn (k) are the projective algebraic subvarieties. The Zariski topol-
ogy on a projective algebraic variety X ⊂ Pn (k) is the subspace topology induced
from the Zariski topology of Pn (k).

Non-empty open subsets of an irreducible space are irreducible itself, and are
also dense.

4. Structure sheaf and sheaf of rational functions: Consider an affine algebraic va-
riety X and an open set U ⊂ X. A function

f :U →k

is regular if it has locally the form

g
f= with g, h ∈ R, h locally without zeros in X.
h
Note that the quotient representation of f is only required to hold locally. The
sheaf OX of regular functions on an affine algebraic variety X is the contravariant
functor
U 7→ OX (U) := { f : U → k| f regular }, U ⊂ X open,
with the restriction of functions. The sheaf OX is named the structure sheaf of X.
Sections over a basis open set Dh ⊂ X form the ring

O(Dh ) = Rh := {g/hn : g ∈ R, n ∈ N}

Because the ideal

a ⊂ k[X1 , ..., Xn ]
is prime, the quotient R and all rings

OX (U), U ⊂ X open,

are integral domains. The sheaf MX of rational functions on X is the sheafifica-

tion of the presheaf
U 7→ Q(OX (U)), U ⊂ X open,
4.2 Elliptic curves over subfields of C 143

with Q(OX (U)) the quotient field of the integral domain OX (U). Sections of MX (U)
are locally quotients
g/h, g, h ∈ R, h 6= 0.
For a non-empty open set U ⊂ X holds

M (U) = Q(R) =: k(X),

independent from U. The field Q(R) is named the function field k(X) of X. The
sheaf
M∗ ⊂ M
is the multiplicative subsheaf of non-zero rational functions.

The definitions carry over to projective algebraic varieties, beaucse the latter
have an atlas of affine algebraic varieties.

5. Non-singularity: Consider an algebraic variety X and a point x ∈ X. The local

ring OX,x at x ∈ X is the ring of germs of regular functions defined in a neigh-
bourhood of x ∈ X. Its dimension defines the dimension of X at x ∈ X.
The variety X is non-singular or smooth at x ∈ X if OX,x is a regular local ring,
i.e. if
dim OX,x = dimκ(x) (mx /mx 2 ) (Dimension equals cotangent dimension).
Here
κ(x) := OX,x /mx
denotes the residue field of X at x.

6. Curve over k: A projective algebraic curve C/k is a 1-dimensional, non-singular

projective algebraic variety defined over the field k. If

C ⊂ P2 (k)

then C/k is a plane projective algebraic curve. Its genus is defined as

g(C) = dimk H 0 (C, Ω ),

the dimension of the k-vector space of regular differential forms.

7. Affine schemes: From a modern viewpoint the primary property of a variety is not
the point set
X ⊂ An (k),
the primary object is the ideal

a ⊂ k[X1 , ..., Xn ]
144 4 Elliptic curves

with minimal field of definition k, or its quotient, the ring

R := k[X1 , ..., Xn ]/a

Due to Grothendieck’s reformulation of algebraic geometry, the basic concept is

for arbitrary rings R the affine scheme

(X := Spec R, OX )

Here Spec R is a topological space with points the prime ideals of R. The structure
sheaf OX has as stalks the localizations

Rp , p ⊂ R prime ideal.

We have included the requirement of non-singularity into our definition of a

projective algebraic curve because all curves under consideration will be assumed
smooth if not stated otherwise. Note that any plane projective algebraic curve is
defined by a single homogeneous polynomial. Its degree is the degree of C/k.

Remark 4.16 for X/k involves two fields:

• First, the field of definition k is not necessarily algebraically closed. It is the
minimal field where the coefficients of a set of generators of I(X) may be taken
from.

• Secondly, the algebraically closed field k, used to visualize K/k as a point set X
from a geometrical view point. Also the condition on smoothness refers to k.

But the set X ⊂ An (k) respectively X ⊂ Pn (k) is not the only point set attached
to K/k. Definition 4.17 introduces point sets X(K) attached to all intermediate
fields k ⊂ K ⊂ k.

Definition 4.17 (K-valued point pf X/k). Consider a projective algebraic variety X/k
with
X ⊂ Pn (k).
Then for any field
k⊂K
the set of K-valued points of X is defined as

X(K) := X ∩ Pn (K) =
4.2 Elliptic curves over subfields of C 145

= {x = (x0 : ... : xn ) ∈ Pn (k) : x ∈ X and xi ∈ K f or all i = 0, ..., n}.

Remark 4.18 (GAGA).

1. Chow’s theorem: Historically, the first GAGA-result is Chow’s theorem: Any
projective complex analytic manifold X ⊂ Pn is the variety of homogeneous poly-
nomials, i.e. if X ⊂ Pn is locally defined as the zero set of holomorphic functions,
then X can be globally defined by homogenous polynomials.

2. Analytification of schemes: To formalize the analogy between complex analytic

geometry and complex algebraic geometry Grothendieck uses the language of
category theory. On one hand, let AnC denote the category of complex spaces and
holomorphic maps. On the other hand, let Al C denote the category of schemes
of finite type defined over C and regular maps. The first step attaches to a
schema X ∈ Al C a distinguished complex analytic space X an ∈ AnC . The latter
is obtained via the functor:
Φ : AnC →
− Ens
attaching to a complex analytic space Y the set

Φ(Y ) := HomC (Y, X)

of morphisms of ringed spaces. The functor Φ can be represented by a universal

object (X an , φ ), named the analytification of X. The pair comprises a complex
analytic space X an and a morphism

φ : X an →
− X,

see [46]. As a consequence: For a given complex analytic space Y each morphism

f :Y →
− X

of ringed spaces induces a unique holomorphic map

f an : Y →
− X an ,

such that the following diagram commutes

f an f
φ
X an X

The analytification φ : X an →
− X from the analytification defines for each x ∈ X an
a pull-back
146 4 Elliptic curves

φx : OX,φ (x) →
− OX an ,x
which allows to compare the local ring of regular functions on X and the local
ring of holomorphic functions on X an . As a consequence one obtains compari-
son theorems between the algebraic properties of the local rings of the structure
sheaves of X and X an , see [46, Chap. 2].

In addition, any morphism f : X → Y in AnC induces a unique holomorphic map

f an : X an →
− Y an
such that the following diagram commutes

φX
X an X

f an f
φY
Y an Y

Again, one obtains comparison theorems between the algebraic properties of the
morphisms
f : X → Y and f an : X an →
− Y an ,
see [46, Chap. 3].

3. Coherent sheaves and proper maps: The morphism

− X an
φ :X →

from the analytification of X defines the functor analytical inverse image as the
pull-back of module sheaves over the structure sheaf:

φ ∗ : Shea f O →
− Shea f O , F 7→ φ ∗ F
X X an

and
− G , 7→ φ ∗ f : φ ∗ F →
f :F → − φ ∗G .
The stalk at a point x ∈ X an of the inverse image sheaf is easy to calculate

(φ ∗ F )x = Fφ (x) ⊗OX,φ (x) OX an ,x

Serre investigated in his GAGA-paper [49] for a projective schema X ∈ Al C and

a coherent OX -module the relation between the cohomology in the algebraic and
the analytic category. Subsequently his results have been generalized to the rela-
tive case: For a proper morphism

f :X →
−Y
4.2 Elliptic curves over subfields of C 147

in Al C and a coherent OX -module sheaf F the canonical morphisms in AnC be-

tween the corresponding direct image sheaves
'
(R j f∗ F )an −
→ R j f∗an (F an ), j ∈ N,

are isomorphisms, see [46, Chap. 4].

The literal analogue of divisors on compact Riemann surfaces are divisors on

a projective algebraic curve C/k. A priori, these divisors refer to the points of the
curve C ⊂ Pn (k) with coordinates in k. In a second step we will consider those
divisors which are defined over k.

Definition 4.19 (Divisors on projective algebraic curves). Consider a projective

algebraic curve C/k with C ⊂ Pn (k).
1. A divisor on C/k is formal sum

D= ∑ nx · x
x∈C

with integers nx ∈ Z satisfying nx = 0 for all but finitely many indices. For a
point p ∈ C the corresponding point divisor is denoted

P := 1 · p

Each divisor D ∈ Div(C) has a well-defined degree

deg D := ∑ nx ∈ Z
x∈C

The additive group of divisors on C/k is denoted Div(C), its subgroup Div0 (C) ⊂ Div(C)
contains the classes of divisors of degree = 0.

2. A divisor D ∈ Div(C) is a principal divisor if D = 0 or

D = div f

for a rational function f ∈ k(C). Two divisors D1 , D2 ∈ Div(C) are equivalent,

annotated
D1 ∼ D2 ,
∗
if their difference is a principal divisor. The divisor of a rational function f ∈ k
has degree
deg (div f ) = 0.
Denote by
Prin(C) ⊂ Div0 (C)
148 4 Elliptic curves

the subgroup of principal divisors. The quotient

Cl(C) := Div(C)/Prin(C)

is named the divisor class group of C. Its subgroup

Cl0 (C) := Div0 (C)/Prin(C) ⊂ Cl(C)

is the class group of degree zero divisors.

3. A divisor
D= ∑ nx · x ∈ Div(C)
x∈C

is defined over k if it is invariant under the action of the Galois group Gk/k , i.e. if
for all σ ∈ Gk/k
D = Dσ := ∑ nx · xσ ∈ Div(C)
x∈C

4. For a divisor D ∈ Div(C) we denote by OD the sheaf of multiples of −D, i.e.

OD (U) := { f ∈ M ∗ (U) : div f ≥ −D|U} ∪ {0}, U ⊂ X open.

For projective-algebraic curves C/k the theorems of Riemann-Roch, Serre dual-

ity, the very-ampleness criterion from Theorem 4.13 and the dimension formula

dim H 0 (C, O) = 1

hold literally in the same form as on compact Riemann surfaces. Also the canonical
divisor K ∈ Div(C) has degree

deg K = 2g(C) − 2.

Hence we will apply these results also in the context of algebraic geoemtery without
further mentioning.

Definition 4.20 (Elliptic curve over k). An elliptic curve defined over k is a
pair (E, O) with
• a projective algebraic curve E/k of genus g = 1

• and a k-valued point O ∈ E

Note. In Definition 4.20 “O” denotes the upper-case letter O, not the digit 0. We
denote by
O∞ := 1 · O ∈ Div(E)
4.2 Elliptic curves over subfields of C 149

the distinguished point divisor. It is defined over k.

If the base field k and the k-valued point O are not relevant for the context, we will
often denote an elliptic curve just by E.

Note that there exist smooth projective algebraic curves C/Q of genus g = 1
without rational points, i.e. with C(Q) = 0.
/ An example is the Selmer
curve C ∈ P2 (C) defined by the homogeneous cubic equation
3X 3 + 4Y 3 + 5Z 3 = 0.

The fact g = 1 for an elliptic curve E allows a simple application of the Riemann-
Roch theorem to compute the vector spaces of multiples of a divisor on E.
Lemma 4.21 (Divisors on an elliptic curve). Consider an elliptic curve E and a
divisor D ∈ Div(E). Then


0 if deg D < 0

0 if deg D = 0, D not principal
dim H 0 (E, OD ) =


1 if deg D = 0, D principal
if deg D > 0

deg D

Proof. The Riemann-Roch theorem together with Serre duality imply for any
divisor D ∈ Div(E)

dim H 0 (E, OD ) − dim H 1 (E, OD ) = dim H 0 (E, OD ) − dim H 0 (E, O−D+K ) =

= 1 − g + deg D = deg D
Because an elliptic curve E has genus g = 1, its canonical divisor has degree

deg K = 2g − 2 = 0.

As a consequence

deg (−D + K) = deg (−D) = −deg D

i) Case deg D 6= 0: If deg D > 0 then deg (−D) < 0. As a consequence

(
deg D if deg D > 0
dim H (E, OD ) =
0
0 if deg D < 0

ii) Case deg D = 0: We consider the sheaf OD as a line bundle. Then

D principal ⇐⇒ OD ' O
150 4 Elliptic curves

On one hand, D principal implies

H 0 (E, OD ) = dim H 0 (E, O) = 1

On the other hand, for non-principal D the line bundle OD is not trivial. Hence each
regular section of s ∈ H 0 (E, OD ) has a zero. If s 6= 0, then deg D = 0 implies that s
has also at least a pole, contradicting the regularity of s. Therefore s = 0, q.e.d.

The following Lemma 4.22 will be used in the proof of Theorem 4.23. It indi-
cates why the existence of a k-valued point is requested for an elliptic curve defined
over k.
Lemma 4.22 (Divisors defined over a subfield). Consider a projective algebraic
curve C/k and a divisor
D = ∑ nx · x ∈ Div(C)
x∈X

which is defined over k. Then the finite dimensional k-vector space

∗
H 0 (C, OD ) = { f ∈ k (C) : div f ≥ −D} ∪ {0}

has a basis of rational functions from k(C), more precisely

H 0 (C, OD ) = k ⊗k { f ∈ k∗ (C) : div f ≥ −D} ∪ {0}

For the proof see [56, II, Prop. 5.8].

Elliptic curves are by definition subvarieties of a projective space Pn (k). Theorem 4.23
proves that they are even plane curves, i.e. that they embed as hypersurfaces
into P2 (k).

Theorem 4.23 (Elliptic curves are plane cubics). For an elliptic curve (E/k, O)
exist two rational functions x, y ∈ k(E) with poles only at O such that the map
(
2 (1 : x(p) : y(p)) p 6= O
φ : E → P (k), p 7→
(0 : 0 : 1) p=O

induces an isomorphism onto the plane cubic C ⊂ P2 (k), which is defined by a

homogeneous polynomial Fhom ∈ k[X0 , X1 , X2 ] of the form

Fhom (X0 , X1 , X2 ) = X0 X22 + a1 X0 X1 X2 + a3 X02 X2 − (X13 + a2 X0 X12 + a4 X02 X1 + a6 X03 )

with five suitable constants ai ∈ k, i = 1, 2, 3, 4, 6. In particular, C is defined over k

and non-singular.

Proof. We construct a very ample line bundle L = OD on E and define φ := φL .

Therefore consider the divisor
4.2 Elliptic curves over subfields of C 151

D := 3 · O∞ ∈ Div(E).

The line bundle

L := OD ∈ Pic(E)
satisfies the very ampleness criterion from Theorem 4.13: The proof follows from
the dimension formula in Lemma 4.21 because for all points p, q ∈ E

dim H 0 (E, OD−P ) = dim H 0 (E, OD ) − 1

and
dim H 0 (E, OD−(P+Q) ) = dim H 0 (E, OD ) − 2.
For n ≥ 0
∗
H 0 (E, On·O∞ ) ) = { f ∈ k (E) : f has at most a pole at O of order = n} ∪ {0}

and the dimension formula implies the proper inclusion

H 0 (E, O3·O∞ ) ) H 0 (E, O2·O∞ ) ) H 0 (E, OO∞ ) ' H 0 (E, O) = C.

Hence a basis (1, x, y) of

H 0 (E, L ) = H 0 (E, O3·O∞ )

exists with
• y ∈ H 0 (E, O3·O∞ ) ⊂ k(E) has a pole of order = 3 at O and no other pole,

• x ∈ H 0 (E, O2·O∞ ) ⊂ k(E) has a pole of order = 2 at O and no other pole,

• 1 ∈ H 0 (E, O) = H 0 (E, O1·O∞ ) ⊂ k(E) has no poles.

By assumption O ∈ E(k). Therefore the divisor D is defined over k. According to
Lemma 4.22 we may choose even x, y ∈ k(E). With respect to the basis (1, x, y) the
line bundle L defines the closed embedding

φ := φL : E → P2 (k), p 7→ (1 : x(p) : y(p)),

onto a curve C/k in P2 (k). The family

(1, x, y, x2 , xy, x3 , y2 )

of seven rational functions from the 6-dimensional vector space H 0 (E, O6·O∞ ) is
linearly dependent. Therefore the functions satisfy a linear relation with coefficients
from k. Among them the two functions x3 and y2 are the only ones having a pole
at O ∈ E of order = 6. Hence their coefficients in the linear relation are not zero.
After division we may assume that their coefficients are = ±1 and

y2 + a1 xy + a3 y = x3 + a2 x2 + a4 x + a6
152 4 Elliptic curves

with coefficients ai ∈ k. This relation defines the inhomogeneous polynomial

F(X,Y ) := Y 2 + a1 XY + a3Y − (X 3 + a2 X 2 + a4 X + a6 ) ∈ k[X,Y ]

and its homogenization, the polynomial Fhom ∈ k[X0 , X1 , X2 ] with

Fhom (X0 , X1 , X2 ) := X0 X12 + a1 X0 X1 X2 + a3 X2 X02 − (X13 + a2 X12 X0 + a4 X1 X02 + a6 X03 ).

The inhomogeneous variables (X,Y ) relate to the homogeneous variables (X0 , X1 , X2 )

according to
X1 X2
X= and Y = .
X0 X0
The curve φ (E) is a subvariety of the 1-dimensional cubic hypersurface

C := V (Fhom ) ⊂ {(x0 : x1 : x2 ) ∈ P2 (k) : Fhom (x0 , x1 , x2 ) = 0}.

The cubic C intersects the line at infinity

P1 (k) ' {(x0 : x1 : x2 ) ∈ P2 (k) : x0 = 0}

in the single point (0 : 0 : 1) with multiplicity = 3. Its affine part is the affine variety
of the polynomial F(X,Y ) ∈ k[X,Y ]

Caff = {(x0 : x1 : x2 ) ∈ P2 (k) : Fhom ((x0 : x1 : x2 )) = 0 and x0 6= 0} =

= {(x, y) ∈ A2 (k) : F(x, y) = 0} = V (F).

The polynomial F ∈ k[X][Y ], considered as polynomial in Y with coefficients from
the ring k[X], is irreducible: It does not split as

F(X,Y ) = (Y + b1 )(Y + b2 )

with polynomials b1 , b2 ∈ k[X]. Therefore Caff and also C are irreducible. Hence the
non-constant regular map
φ :E →C
is surjective, cf. [29, Chap. II, Prop. 6.8], an analogue of the theorem about non-
constant proper holomorphic maps between Riemann surfaces. As a consequence

φ (E) = C and E ' C over k

and C is non-singular, q.e.d.

In the following we will identify an abstract elliptic curve (E/k, O) with the plane
cubic from Theorem 4.23. We say that (E/k, O), more precisely its affine part, is
defined over k by the Weierstrass equation F(X,Y ) = 0.
4.2 Elliptic curves over subfields of C 153

Definition 4.24 (Weierstrass polynomial and Weierstrass equation, its discrim-

inant and j-invariant). A Weierstrass polynomial over a field k is a cubic poyno-
mial

F(X,Y ) = Y 2 + a1 XY + a3Y − (X 3 + a2 X 2 + a4 X + a6 ) ∈ k[X,Y ]

with coefficients
a1 , a2 , a3 , a4 , a6 ∈ k.
1. The equation
F(X,Y ) = 0,
i.e.
Y 2 + a1 XY + a3Y = X 3 + a2 X 2 + a4 X + a6
is named the Weierstrass equation over k defined by F.

2. The discriminant of F is

∆F := −b22 b8 − 8b34 − 27b26 + 9b2 b4 b6

with
b2 := a21 + 4a2 , b4 := 2a4 + a1 a3 , b6 := a23 + 4a6
and
b8 := a21 a6 + 4a2 a6 − a1 a3 a4 + a2 a23 − a24 .
3. The j-invariant of F is

c34
jF := , c4 := b22 − 24bb .
∆F

Starting from an elliptic curve over a field k the resulting Weierstrass polynomial
is not unique. There exist coordinate transformations defined over k which transform
a Weierstrass polynomial over k to a different form, in particular to a more simple
form.
Proposition 4.25 (Weierstrass equation in short form, discriminant and j-invariant).
Consider a Weierstrass polynomial

F(X,Y ) = Y 2 + a1 XY + a3Y − (X 3 + a2 X 2 + a4 X + a6 ) ∈ k[X,Y ]

1. Short form: If char k 6= 2, 3 then there exists a linear coordinate change

   0 
u2 · X + r

X X
7→ = 3
Y Y0 u ·Y + s · u2 · X + t

with constants
r, s,t ∈ k, u ∈ k∗
154 4 Elliptic curves

which transforms F(X,Y ) to the Weierstrass polynomial in short form

2 3
F 0 (X 0 ,Y 0 ) = Y 0 − (X 0 + A · X 0 + B), A, B ∈ k

with discriminant
∆F 0 = −24 · (4A3 + 27B2 ) ∈ k
and the same j-invariant
jF 0 = jF
2. Transforming Weierstrass polynomials in short form: Each linear coordinate
transformation over k
 0    2   0
X X u 0 X
7→ := · .
Y0 Y 0 u3 Y0

with u ∈ k∗ transforms a Weierstrass polynomial in short form

2 3
F 0 (X 0 ,Y 0 ) = Y 0 − (X 0 + A0 · X 0 + B0 ) ∈ k[X 0 ,Y 0 ]

to the Weierstrass polynomial in short form

F(X,Y ) = Y 2 − (X 3 + A · X + B) ∈ k[X,Y ].

such that the numerical constants relate as

A = u4 · A0 , B = u6 · B0 , ∆F = u12 · ∆F 0 .

retaining the j-invariant

(4A)3 (4A0 )3
jF := 1728 · = 1728 · =: jF 0
∆F ∆F 0
3. The j-invariant: Two Weierstrass polynomials in short form have the same j-invariant
if and only if they are related by an isomorphism as in part 3) defined over k -
not necessarily over k.

4. For each algebraic number j0 ∈ k exists an elliptic curve E/k( j0 ) with j-invariant
equal to j0 .

For the proof of Proposition 4.25 see [56]: For part 1) cf. Chap. III, §1; for part 2)
cf. Chap. III, Rem. 1.3; for part 3) cf. Chap. III, Prop. 1.4(b); for part 4) cf.
Chap. III, Prop. 1.4(c).

The discriminant ∆F ∈ k of a Weierstrass polynomial F(X,Y ) ∈ k[X,Y ] indicates

whether the projective algebraic variety
4.2 Elliptic curves over subfields of C 155

C = V (Fhom ) ⊂ P2 (k)

is non-singular or not. If ∆F 6= 0 then (C/k, (0 : 0 : 1)) is non-singular, i.e. an elliptic

curve.

Proposition 4.26 (Singularities of cubics). Assume char k 6= 2, 3 and consider a

plane cubic curve
C := V (Fhom ) =⊂ P2 (k)
defined by the homogenization of a Weierstrass polynomial in short form

F(X,Y ) = Y 2 − (X 3 + AX + B) ∈ k[X,Y ]

with A, B ∈ k and discriminant

∆F = −16 · (4A3 + 27B2 ).

The curve C has at most one singular point. The curve

• is non-singular iff ∆F 6= 0

• has a node iff ∆F = 0 and A 6= 0

• has a cusp iff ∆F = A = 0.

For a proof of Proposition 4.26 see [56, Chap. III, Prop. 1.4].

Remark 4.27 (Cubic projective-algebraic curves and elliptic curves). The adjunc-
tion formula for non-singular hypersurfaces implies: Any projective algebraic cu-
bic curve C ⊂ P2 has genus = 1. Hence C/k is an elliptic curve if C has at least
one k-valued point, see [29, Chap. V, Example. 1.5.1].

Theorem 4.28 is the converse of Theorem 4.3. it shows that the complex points of
an elliptic curve, defined over a subfield of C, can be parametrized by the
Weierstrass function ℘ and its derivative ℘ 0 of a suitable torus. The proof relies on
the result of Theorem 4.23 that elliptic curves are plane cubics which satisfy a
Weierstrass equation.
A very coarse analogue of the uniformization is the parametrization of the points of
the unit circle by the values of the exponential function e2πi·t on the parameter
interval [0, 1].
Theorem 4.28 (Uniformization of elliptic curves E/C by tori). For any elliptic
curve E defined over C exists a torus T = C/Λ and a biholomorphic map
'
Φ :T −
→ E(C).

Here E(C) ⊂ Pn (C) is provided with the induced Euclidean topology and analytic
structure from the complex manifold Pn (C) = Pn .
156 4 Elliptic curves

Proof. Due to Theorem 4.23 we may assume that E is defined as a plane curve,
defined by the homogenization of a Weierstrass polynomial in short form
2
F(X,Y 0 ) := Y 0 − (X 3 + a4 · X + a6 ) ∈ C[X,Y 0 ].

The substitution
Y 0 = (1/2) ·Y
shows that E can also be defined by the Weierstrass polynomial

F(X,Y ) = Y 2 − (4 · X 3 − A · X − B)

with suitable constants A, B ∈ C and discriminant

∆F = A3 − 27 · B3

due to Proposition 4.26. We search for a lattice

Λ = Zω1 + Zω2 ⊂ C

with lattice constants GΛ ,4 and GΛ ,6 satisfiying

A = 60 · GΛ , 4 and B = 140 · GΛ , 6 .

The restriction of the modular invariant

g32 (τ)
j|H : H → C, j(τ) := 1728 · 3
,
g2 (τ) − 27g23 (τ)

is surjective according to Corollary 3.16. Hence a number τ ∈ H exists with

A3
j(τ) = 1728 · .
∆F
If g2 (τ) = 0 then τ = ρ according to Corollary 3.20. We choose

Λ := Λρ .

Similarly, if g3 (τ) = 0 then τ = i, and we choose

Λ := Λi .

Otherwise, the two equations

g32 (τ) A3
j(τ) = 1728 · = 1728 ·
g32 (τ) − 27g23 (τ) ∆F

and
∆F = A3 − 27B2
4.2 Elliptic curves over subfields of C 157

imply
g32 (τ) A3
= .
g32 (τ) − 27g23 (τ) A3 − 27 · B2
Clearing the denominators by crosswise multiplication implies
!3 !2
A B
=
g2 (τ) g3 (τ)

We choose ω1 ∈ C with
!4 !6
A 1 B 1
= and = .
g2 (τ) ω1 g3 (τ) ω1

This choice is possible because we may replace ω1 by iω1 . In addition, we set

ω2 := τ · ω1 , i.e. τ = ω2 /ω1 .

Then (ω1 , ω2 ) is a positively oriented basis of the lattice

Λ := Zω1 + Zω2 .

Its lattice constants are

 4  4
1 1
GΛ ,4 = · G4 (τ) = · (1/60) · g2 (τ) = (1/60) · A
ω1 ω1

and similarly
GΛ ,6 = (1/140) · B,
i.e.
A = gΛ ,2 and B = gΛ ,3 .
The map
− P2
φ : C/Λ →
from Theorem 4.3 maps the torus C/Λ biholomorphically to the cubic hypersurface
with Weierstrass polynomial

F(X,Y ) = Y 2 − (4 · X 3 − gΛ ,2 · X − gΛ ,3 ) = Y 2 − (4 · X 3 − A · X − B), q.e.d.

Example 4.29 (Elliptic curves and cubics with a singularity). The following Figures 4.5 - 4.8
show some elliptic curves.
While Figure 4.9 and Figure 4.10 show the singular cubics node and the Neil
parabola. They have a single singularity, namely at the origin respectively a node
and a cusp. The corresponding table from Figure 4.4 shows the invariants. The role
of the conductor will be explained in Definition 4.40 later.
158 4 Elliptic curves

For the computation and the plots see the PARI files Elliptic_curve_02
and Elliptic_curve_plot_06.

Figure Weierstrass equation j-invariant discriminant conductor

Elliptic curve

4.5 Y 2 = X 3 − 3X + 3 −28 · 33 /5 −24 · 33 · 5 23 · 33 · 5

4.6 Y 2 = X3 + X 1728 −26 26

4.7 Y 2 = X3 − X 1728 26 25

4.8 Y 2 = X3 + 3 0 −24 · 35 24 · 35

Y 2 = X 3 − X − 14 −25 · 33 /661 −27 · 661 27 · 661

Y 2 + 5XY +Y = X 3 53 · 1013 /(2 · 72 ) 2 · 72 2·7

Singular cubic

4.9 Y 2 = X3 + X2 − 0 −

4.10 Y 2 = X3 − 0 −

Fig. 4.4 Numerical values of some cubic curves

4.2 Elliptic curves over subfields of C 159

Fig. 4.5 Y 2 = X 3 − 3X + 3

Fig. 4.6 Y 2 = X 3 + X

Fig. 4.7 Y 2 = X 3 − X
160 4 Elliptic curves

Fig. 4.8 Y 2 = X 3 + 3

Fig. 4.9 Y 2 = X 3 + X 2

Fig. 4.10 Y 2 = X 3
4.2 Elliptic curves over subfields of C 161

Lemma 4.30 prepares the proof of Theorem 4.31 and the Definition 4.32 of the
group structure of an elliptic curve (E/k, O). The lemma shows that for a pair (p, q)
of distinct points from E the divisor

D := P − Q ∈ Div(E)

cannot be a principal divisor. Note that the lemma gives also a second proof of
Corollary 1.9.

Lemma 4.30 (No single pole of order one). Consider an elliptic curve (E/k, O)
defined over a field k. There is no ratiomal function f ∈ k(E)∗ with

div f = P − Q

with P, Q ∈ Div(E) the elementary divisors of two distinct points p 6= q ∈ E.

Proof. Consider the point divisor

Q ∈ Div(E) of degree deg Q = 1.

Lemma 4.21 implies

dim H 0 (E, OQ ) = 1.
Hence the inclusion
k ⊂ H 0 (E, OQ )
implies
k = H 0 (E, OQ ).
Hence each f ∈ H 0 (E, OQ ) is constant, and therefore f = 0 or

div f = 0 ∈ Div(E)

which implies p = q, a contradiction, q.e.d.

Lemma 4.30 shows: The obstructions for a divisor D ∈ Div0 (E) to be a principal
divisor are divisors of type P − Q or even of type P − O∞ . A divisor D ∈ Div0 (E) is
the divisor of a rational function up to a divisor of type

P − O∞ .

Because the point P ∈ E is uniquely determined by D, the obstruction is expressed

by P alone.

We use the notation from Definition 4.19).

Theorem 4.31 (The class group of degree zero divisors on an elliptic curve).
Consider an elliptic curve (E/k, O).
162 4 Elliptic curves

1. For any divisor D ∈ Div0 (E) exists a unique point p ∈ E with

D ∼ P − O∞

Denote the corresponding map by

σ : Div0 (E) → E, D 7→ p.

2. The map σ is surjective. It induces a bijective map from the group of divisor
classes of degree = 0
'
σ : Cl0 (E) −
→E
with inverse
'
E−
→ Cl0 (E), p 7→ [P − O∞ ].

Proof. 1. Definition of σ : The divisor D + O∞ has degree = 1. Lemma 4.21 implies

dim H 0 (E, OD+O∞ ) = 1.

We choose a non-zero rational function f ∈ k(E)∗ with divisor satisfying

div f ≥ −(D + O∞ ).

Because
deg (div f ) = 0 but deg(−(D + O∞ )) = −1,
there exists a point p ∈ E with

div f = −(D + O∞ ) + P

Then
0 ∼ −(D + O∞ ) + P i.e. D ∼ P − O∞ .
If for a different point
q ∈ E, q 6= p,
also
D ∼ Q − O∞
then
0 = D − D ∼ P − Q.
∗
Hence for a suitable rational function g ∈ k (E)

P − Q = div g,

contradicting Lemma 4.30. Therefore

σ (D) := p ∈ E

is well-defined.
4.2 Elliptic curves over subfields of C 163

2. i) Surjectivity of σ : Consider an arbitrary point p ∈ E. Set

D := P − O∞ ∈ Div0 (E).

Then σ (D) = P by definition, which proves the surjectivity of σ .

ii) Induced map σ :

[D1 ] = [D2 ] =⇒ P1 − O∞ ∼ P2 − O∞ =⇒ P1 ∼ P2 =⇒ σ (D1 ) = σ (D2 )

iii) Injectivity of σ :

σ ([D1 ]) = σ ([D2 ]) =⇒ σ (D1 ) = σ (D2 ) =⇒

=⇒ ∃p ∈ E : D1 ∼ P−O∞ and D2 ∼ P−O∞ =⇒ D1 ∼ D2 =⇒ [D1 ] = [D2 ], q.e.d.

Definition 4.32 (Group structure of an elliptic curve). Consider an elliptic curve (E/k, O)
and denote by (Cl0 (E), +) the group of divisor classes of degree zero. The bijective
map from Theorem 4.31

σ : Cl0 (E) →
− E, [P − O∞ ] 7→ p,

carries over the group structure from (Cl0 (E), +) to E: The map

⊕ : E ×E →
− E, p ⊕ q := σ ([P − O∞ ] + [Q − O∞ ]),

defines an Abelian additive group (E, ⊕) with neutral element the k-valued point O ∈ E(k).

Corollary 4.33 (Abel’s Theorem for elliptic curves). Consider an elliptic curve (E/k, O).
A divisor
D = ∑ n p · P ∈ Div0 (E)
p∈E

is a principal divisor iff M

n p ∗ p = O ∈ E.
p∈E

Here the last equation refers to the group structure ⊕ on E. The product n p ∗ p
means:
• Taking n p -times the sum of a point p ∈ E if n p ≥ 0, and

• taking (−n p )-times the sum of p ∈ E if n p ≤ 0. Here the symbol p∈E

denotes the negative of p with respect to the group structure.
164 4 Elliptic curves

Proof. Because
0 = deg D = ∑ np
p∈E

we have
D = D− ∑ nP · O∞ = ∑ n p · (P − O∞ ) ∈ Div0 (E)
p∈E p∈E

Due to Theorem 4.31, part ii)

D principal ⇐⇒ [D] = 0 ⇐⇒ σ ([D]) = O ∈ E ⇐⇒

!
M
⇐⇒ σ ∑ n p · [P − O∞ ] = O ∈ E ⇐⇒ n p ∗ σ ([P − O∞ ]) = O ∈ E ⇐⇒
p∈E p∈E
M
⇐⇒ n p ∗ p = 0 ∈ E, q.e.d.
p∈E

Proposition 4.34 (Group structure of tori and elliptic curves). For a complex
torus T = C/Λ the biholomorphic map from Theorem 4.3

φ :T →
− E(C)

onto the complex points of a plane elliptic curve E is an isomorphism of groups.

Proof. 1. i) Denote by 0T ∈ T the neutral element of the group (T, ⊕). The canoni-
cal morphism

ρ : (T, ⊕) →
− (Div0 (T ), +), p 7→ 1 · p − 1 · 0T ,

is a group morphism: For j = 1, 2 and two points p j ∈ T and divisors from

Div0 (T )
D j := 1 · p j − 1 · 0T and D := 1 · (p1 ⊕ p2 ) − 1 · 0T
the difference

D1 + D2 − D = 1 · p1 + 1 · p2 − 1 · (p1 ⊕ p2 ) − 1 · 0T ∈ Div0 (T )

is a principal divisor according to Abel’s Theorem 1.21 because

p1 ⊕ p2 (p1 ⊕ p2 ) 0T = 0T ∈ Λ

ii) For a point p ∈ T the image

ρ(p) = 1 · p − 1 · 0T ∈ Div0 (T )

is a principal divisor iff p = 0T , again due to Abel’s Theorem. If we denote by

ρ :T →
− Cl0 (T )
4.2 Elliptic curves over subfields of C 165

the composition of ρ with the canonical map Div0 (T ) →

− Cl0 (T ) then

ρ :T →
− Cl0 (T )

is injective.

iii) Each divisor

D= ∑ na · a − ∑ nb · b ∈ Div0 (T ), A, B ⊂ T, A ∩ B = 0,/ na , nb > 0,
a∈A b∈B

defines a point ! !
M M
p := na · a nb · b ∈T
a∈A b∈B

Then
ρ(p) = 1 · p − 1 · 0T ∈ Div0 (T )
Due to Abel’s Theorem the difference
! !! !
M M
ρ(p)−D = 1 na · a nb · b −1·0T − ∑ na · a − ∑ nb · b ∈ Div0 (T )
a∈A b∈B a∈A b∈B

is a principal divisor because 0T = 0T ∈ Λ . Hence ρ is surjective.

iv) Part ii) and iii) imply that ρ is an isomorphism of groups.

2. Due to part 1) and Definition 4.32 the group structures respectively on domain
and codomain of
φ :T →− E(C)
are determined by the divisor class groups. Because φ is an analytical isomor-
phism it is a group isomorphism, q.e.d.

Remark 4.35 (Group structure of an elliptic curve).

1. Geometric version: The group structure from Definition 4.32 of an elliptic
curve (E/k, O) can also be obtained by a geometric construction. It relies on
the fact that a cubic and a line in P2 intersect in exactly three points, counted
with multiplicity.

See Figure 4.11: Two add two points p, q ∈ E \ {O} consider the line L(p, q)
passing through p and q. The line intersects E in a third point r ∈ E. The second
line L(r, O) intersects E in a third point, which is p ⊕ q. For more details see [63].
166 4 Elliptic curves

Fig. 4.11 Group law on an elliptic curve (Geometric construction)

2. Mordells theorem: The group structure on E(Q) provides also the set E(K)
of K-valued points with a group structure for any intermediate field

Q ⊂ K ⊂ Q.

Mordell proved: The Abelian group E(Q) of Q-valued points of E is finitely

generated, [56, Chap. VIII, Theor. 4.1]. As a consequence, E(Q), today names
the Mordell-Weil group of E/Q, decomposes as the product of an Abelian torsion
group and a free Abelian group of finite rank r < ∞

E(Q) = E(Q)tors × Zr .

Many open questions are related to the rank r.

3. Falting’s theorem: Elliptic curves have genus g = 1. Curves of genus g > 1

do no longer carry a group structure. Mordell conjectured: Projective algebraic
curves C/Q with genus g > 1 have only finitely many points with rational coor-
dinates, i.e.
4.2 Elliptic curves over subfields of C 167

card C(Q) < ∞.

The Mordell-conjecture was proved by Faltings, see [58].

Example 4.36 (Group structure of an elliptic curve). Consider an elliptic curve E/Q.
The PARI file Elliptic_curve_torsion_group investigates the torsion
group of E(Q) for the elliptic curve E/Q with Weierstrass polynomial

y2 = x3 − 43x + 166

see Figure 4.12. It shows that the torsion group E(Q)tors has a factor isomorphic
to Z/7Z. For more information, in particular for the equality

E(Q)tors = Z/7Z

see [54, Chap. VII, Theor. 7.5]. It is conjectured that the rank r of the free part of
the group E(Q) equals the analytic rank of E/Q.
168 4 Elliptic curves

Fig. 4.12 The group E(Q)tors

4.3 Elliptic curves over finite fields 169

4.3 Elliptic curves over finite fields

The present section starts the investigation of elliptic curves defined over Q. Each
elliptiv curve E/Q can also be defined by a distinguished Weierstrass polynomial
with integer coefficients, called a global minimal polynomial, see Definition 4.37
and Remark 4.38.

Entering into the field of arithmetic geometry one may reduce the integer equa-
tion modulo any prime p ∈ Z, each time obtaining an elliptic curve defined over a
finite field F p . Considered from a general point of view: An elliptic curve defined
over Q is a family of not necessarily smooth curves defined over all prime fields of
prime characteristic. Or even more general, it is a map to Spec Z, and all but but
finitely many fibres are elliptic curves over prime fields. Reduction modulo p is a
powerful tool to investigate the arithmetic information encoded in an elliptic curve
defined over Q.

Before investigating the family of elliptic curves one has to eliminate redundant
prime powers of the discriminant which otherwise would distort the result. The dis-
criminants of the different Weierstrass polynomials of an elliptic curve differ by
a non-zero rational. After multiplying with a principal nominator they differ by a
non-zero integer. Hence the reduction of the discriminants modulo a prime p ∈ Z
may differ. To generalize the discriminant criterion from Proposition 4.26 one has
to find a unique minimal discriminant. The corresponding Weierstrass polynomial
is named global minimal.

Definition 4.37 (Global minimal Weierstrass polynomial). For an elliptic curve E/Q
a Weierstrass polynomial F ∈ Z[X,Y ] with integer coefficients is global minimal if
its discriminant
∆F = ± ∏ pv(p) , v(p) ∈ N∗ ,
p prime

is minimal for each given prime factor p of ∆F in the following sense: The elliptic
curve E/Q cannot be defined by a Weierstrass polynomial Fp (X,Y ) ∈ Z[X,Y ] with
discriminant ∆Fp satisfying
ord p (∆Fp ) < v(p).

“Global minimality” is a global property because it refers to all primes of ∆F

Remark 4.38 (Global minimal Weierstrass polynomial).

1. Non-uniqeness of the discriminant: The Weierstrass polynomial of an elliptic
curve, even if is in short form, is not uniquely determined. In general, an el-
liptic curve does not have a unique discriminant. It is not the discriminant, but
the j-invariant which characterizes the elliptic curve.
170 4 Elliptic curves

2. Global minimal Weierstrass polynomial: The discriminant of a global minimal

Weierstrass polynomial
F(X,Y ) ∈ Z[X,Y ]
is uniquely determined. The last condition from Definition 4.37 implies for any
prime factor p 6= 2, 3 of its discriminant

ord p (∆F ) < 12.

Each elliptic curve over k = Q has a global minimal Weierstrass polynomial. The
result generalizes to elliptic curves over arbitrary number fields with class num-
ber = 1, see [56, Chap. VIII, Cor. 8.3] and more specific [33, Theor. 10.3]. A
global minimal Weierstrass polynomial is not necessarily in short form.

3. PARI command: The Pari command ellminimalmodel provides the global mini-
mal model for elliptic curves over number fields k with class number = 1.

Example 4.39 (Global minimal Weierstrass polynomial).

1. See PARI-file Elliptic_curve_weierstrass_equation_12. Figure 4.13,
upper example, considers the Weierstrass polynomial

F(X,Y ) = Y 2 − (X 3 + 15.625)

Its discrimant
∆F = −24 · 33 · 512
is not minimal because
512 |∆F
The global minimal Weierstrass polynomial is

Fmin (X,Y ) = Y 2 − (X 3 + 1)

with discriminant
∆Fmin = −24 · 33 .
4.3 Elliptic curves over finite fields 171

Fig. 4.13 Global minimal Weierstrass polynomial

2. The Weierstrass polynomial

F(X,Y ) = Y 2 − (X 3 − 595 · X + 5.586) ∈ Z[X,Y ]

with discriminant
∆F = 212 · 73
is global minimal, despite of the exponent 12 of the particular prime factor p = 2.

3. Consider the elliptic curve E/Q with Weierstrass polynomial in short form

1
F(X,Y ) = Y 2 − (X 3 + · X + 1) ∈ Q[X,Y ]
5
which is not global minimal because it has a non-integer cefficient. To obtain a
global minimal Weierstrass polynomial, we make the transformation
   −2 0 
X u ·X
= −3 0 with u := 5
Y u ·Y

and multiply by u6 . We obtain the Weierstrass equation

2 3 1 2 3
Y 0 = X 0 + · u−2 · u6 · X 0 + u6 i.e. Y 0 = X 0 + 53 · X 0 + 56
u
Hence E/Q is also defined by the Weierstrass polynomial

Fmin (X,Y ) = Y 2 − (X 3 + 53 · X + 56 ) ∈ Z[X,Y ]

which is global minimal because

∆Fmin = −24 · 59 · 31 · 109 ∈ Z,

172 4 Elliptic curves

see Figure 4.13, lower example.

Definition 4.40 (Reduction mod p). Consider an elliptic curve

E = (E/Q, O)

with global minimal Weierstrass polynomial F(X,Y ) ∈ Z[X,Y ].

1. Denote by
Fhom ∈ Z[X0 , X1 , X2 ]
the homogeniziation of F. For each prime p ∈ N reducing mod p the coefficients
of Fhom provides a Weierstrass polynomial

Fhom,p ∈ F p [X0 , X1 , X2 ]

with coefficients from the finite field F p . The Weierstrass polynomial defines the
projective variety

E p := Var(Fhom,p ) := {(x0 : x1 : x2 ) ∈ P2 (F p ) : Fhom,p (x0 , x1 , x2 ) = 0} ⊂ P2 (F p ),

which is named the reduction mod p of E .

2. At a prime p ∈ Z the elliptic curve E is

• stable if (E p , O) is non-singular, i.e. an elliptic curve over F p .

• semistable if (E p , O) is singular with a node.

• unstable if (E p , O) is singular with a cusp.

3. The conductor cond(E) ∈ N∗ of E/Q distinguishes the primes p with semistable
reduction from those with unstable reduction: One has

cond(E) = ∏ pe p
p

with the same prime factors p as the discriminant ∆F . The exponents of the prime
factors satisfy
e p := 1 ⇐⇒ E is semistable at p
and for p > 3
e p := 2 ⇐⇒ E is unstable at p,
For p = 2, 3 see also [54, App. C, §16].
4.3 Elliptic curves over finite fields 173

According to Proposition 4.26 the distinction between stable and not-stable at a

prime p depends on the reduction of the minimal discriminant. The following
example 4.41 shows why it is necessary to study elliptic curves by considering
global minimal Weierstrass polynomials, not just arbitrary Weierstrass
polynomials.

Example 4.41 (Reduction mod p of a global minimal Weierstrass polynomial). Con-

sider the elliptic curve (E/Q, O) from Example 4.39 defined by the inhomogeneous
Weierstrass polynomial

F(X,Y ) = Y 2 − (X 3 + 15.625) ∈ Q[X,Y ]

with discriminant
∆F = −24 · 33 · 512 ∈ Z.
Note 15.625 = 56 .

i) The Weierstrass polynomial F ∈ Z[X,Y ] is in short form but it is not global

minimal. Reducing its coefficients mod p = 5 gives the Weierstrass polynomial in
short form
F5 (X,Y ) = Y 2 − X 3 ∈ F5 [X,Y ]
with discriminant
∆F5 = 0 ∈ F5 .
Hence the Weierstrass polynomial F5 defines a singular cubic over F5 with a cusp.
Note that the discriminant ∆F contains the prime factor 512 .

ii) Therefore we have to apply first the coordinate transformation with the matrix
 2 
1/u 0
, u := 5,
0 1/u3

which transforms F(X,Y ) to the global minimal Weierstrass polynomial

2 3
F 0 (X 0 ,Y 0 ) = Y 0 − (X 0 + 1)

with discriminant
∆F 0 = −24 · 33 ∈ Z.
Reducing the coefficients of F 0 mod p = 5 creates the Weierstrass polynomial
2 3
F50 (X 0 ,Y 0 ) = Y 0 − (X 0 + 1) ∈ F5 [X 0 ,Y 0 ]

with discriminant
∆F 0 = 3 ∈ F5 .
5

The reduction of E mod p = 5 defines an elliptic curve E5 over F5 .

174 4 Elliptic curves

Remark 4.42 (Hasse estimate).

1. Important properties of an elliptic curve E/Q are encoded in the function, which
attaches to each prime p with stable E p the family

E p (F pn ), n ∈ N∗ ,

of orders of the groups of F pn -valued points of the elliptic curve E p . These groups
collect those points from E p ⊂ P2 (F p ) with all coordinates in the finite fields F pn .
We recall the following result of Hasse for the case n = 1:
√
1 + p − card E p (F p ) ≤ 2 · p.
For a proof see [33, Theor. 10.5].

Hasse’s result was conjectured by Artin. The value

1+ p

for comparison is motivated as follows: The field F p has p elements. A

Weierstrass equation
Y 2 = X 3 + AX + B
is quadratic in Y . Hence for each x ∈ F p there exist at most two
solutions (x, y) ∈ A2 (F p ). Taking into account the point at infinity O ∈ E p (F p )
gives the estimate
card E p (F p ) ≤ (1 + 2p).
In the average, i.e. by choosing the Weierstrass polynomial at random, one
could expect a 50% chance to find a solution (x, y) at all. Therefore 1 + p is a
plausible average for card E p (F p ).

One defines for any prime p the difference

a p (E) := 1 + p − card E p (F p )

between the average number and the actual number of F p -valued points of E.

2. See the PARI-file Elliptic_curve_Hasse_estimate_05 for the numer-

ical example of the Hasse inequality displayed in Figure 4.14.
4.3 Elliptic curves over finite fields 175

Fig. 4.14 The Hasse estimate

3. The sequence
(a p (E)) p prime

will turn out to be a fundamental characteristic of the elliptic curve E, see

Remark 4.45. These numbers determine also the cardinalities

card E p (F pn )

for n ≥ 2. The latter will serve as the generators of the ζ -function of E, see
Definition 4.44.
176 4 Elliptic curves

Historically, the first example of a ζ -function is due to by L. Euler. The function

has later be named Riemann ζ -function
Proposition 4.43 (ζ -function and Euler product). The Riemann ζ -function
∞ 1
ζ : {s ∈ C : Re s > 1} →
− C, ζ (s) := ∑ ns
n=1

is holomorphic. The ζ -function

1. extends to a meromorphic function on C with a single pole at s = 1. The pole has
order = 1 and residue = 1.

2. It satisfies the functional equation with respect to reflection at the point s = 1

π −s/2 · Γ (s/2) · ζ (s) = π −(1−s)/2 · Γ ((1 − s)/2) · ζ (1 − s)

with the Γ -function,

3. and it has the product representation

1
ζ (s) = ∏
p 1 − p−s

with the product taken over all primes p.

Proof. 1. See [64, Teil I, § 4].

2. See reference from part 1).

3. For each N ∈ N expand the finite product by using the geometric series and the
unique prime factor decomposition of integers
!
1 1 1 1
∏ 1 − p−s = ∏ 1 + ps + p2s + ... = ∑ 0 ns
p<N p<N

with the sum on the right-hand side taken over all positive integers with all prime
factors less than N. The limit N → ∞ proves the claim, q.e.d.

From the viewpoint of the ring Z the Riemann ζ -function is a global object.
Because it considers all primes p ∈ Z and collects as a product the local information
encoded in the geometric series of p

1 ∞
n
= ∑ 1/p−s
1 − p−s n=0

Apparently, all information about a prime p is just its value.

4.3 Elliptic curves over finite fields 177

We now switch from the primes p of Z to the elliptic curves E p /F p , which are the
fibres of a given elliptic curve E/Q over its stable primes p. First we study the
local information, i.e. the information encoded in E p /F p . The curve E p contributes
as local information at the prime p the sequence of cardinalities of points with
values in the finite fields of characteristic p
card E p (F pn ), n ∈ N∗ .

Definition 4.44 (Zeta-function of an elliptic curve over F p ). Consider an elliptic

curve E p /F p . The ζ -function of E p is the formal power series with generator derived
from the sequence (card E p (F pn ))n∈N∗
!
∞ Tn
Z(E p , T ) := exp ∑ card E p (F pn ) · ∈ Q[[T ]].
n=1 n

One checks by formal derivation

d Z0 ∞ ∞
log Z(E p , T ) = T · = T · ∑ card E p (Fnp ) · T n−1 = ∑ card E p (Fnp ) · T n



T·
dT Z n=1 n=1

Remark 4.45 (Properties of the ζ -function of an elliptic curve over F p ). The ζ -function
of an elliptic curve E p /F p
Z(E p , T ) ∈ Q[[T ]]
has the following properties:
1. Rationality: The ζ -function is a rational function over the ring Z, more precisely:

L(E p , T )
Z(E p , T ) = ∈ Z(T )
(1 − T ) · (1 − p · T )

with the quadratic polynomial

L(E p , T ) = 1 − a p (E) · T + p · T 2 , a p (E) := 1 + p − card E p (F p ).

2. Functional equation: The ζ -function satisfies the functional equation

Z(E p , 1/pT ) = Z(E p , T ).

3. Analogue of the Riemann hypothesis: Both complex zeros α, β ∈ C of L(E p , T )

have modulus
√
|α| = |β | = p.
178 4 Elliptic curves

The proof of these properties is due to Weil, [56, Chap. V, Theor. 2.2].

Weil also stated and proved a generalization for projective algebraic curves
curves C p /F p of arbitrary genus. If the curve C p is the reduction mod p of a curve

C/Q

then Weil’s theorem shows the neat relation between the curve C p and the complex
curve X(C) ⊂ Pn : The dimension of the polynomial in the numerator of
the ζ -function of C p is the first Betti number of the compact Riemann surface X/C.

Moreover, Weil made a conjecture for the general case of higher-dimensional

smooth projective algebraic varieties. Deligne proved the Weil conjecture. For an
excellent introduction to the whole subject see [29, Appendix C].

The L-series L(E) of an elliptic curve E/Q collects for each prime p ∈ Z the
local information about the reduction E p . Hence the L-series is a global object.
Definition 4.46 (L-series of an elliptic curve over Q). The L-series of an elliptic
curve E/Q is the function

1
L(E, −) : {s ∈ C : Re s > 3/2} → C, L(E, s) := ∏
p Z(E p , p−s )

Remark 4.47 (L-series of an elliptic curve E/Q).

1. The convergence of the L-series and its representation as an absolut convergent
Dirichlet series with respect to the variable s follows from Hasse’s estimate in
Remark 4.42 and the theory of Euler products, see [33, Chap. X, Cor. 10.6].
Hence L(E) is a holomorphic function.
Due to the proof of the Shimura-Taniyama-Weil conjecture the L-series of
an elliptic curve extends holomorphically to the whole complex plane. There
are several unsolved conjectures about the arithmetic information encoded by
the L-series of an elliptic curve. E.g. Birch and Swinnerton-Dyer conjectured

rank E(Q) = r

with r equal to the analytic rank defined as the order of the zero of the L-series
at s = 1
r := ord(L(E); 1).
The PARI script Elliptic_curve_04_2 computes the analytic rank for
some elliptic curves E/Q of higher analytic rank, see Figure 4.15.

For more results and conjectures see [56, App. C, § 16].

4.3 Elliptic curves over finite fields 179

2. The two polynomials in the denominator of the ζ -function Z(E p , T ) evaluate

at T = p−s as values of the Riemann ζ -function

1 1
−s
= ζ (s) and = ζ (s − 1)
1− p 1 − p1−s
They do not provide new information about the elliptic curve.

3. If one compares the L-series L(E, s) of an elliptic curve E/Q with the Riemann ζ -function ζ (s)
on the level of the two product representations

1 1
L(E, s) = ∏ −s
and ζ (s) = ∏ −s
p Z(E p , p ) p 1− p

then for each given prime p the local factor

1
Z(E p , p−s ) corresponds to
1 − p−s
The geometric series on the right hand side depends only on the prime p, while
the corresponding ζ -function on the left-hand side depends on p and on the re-
duction of the given elliptic curve E/Q.
180 4 Elliptic curves

Fig. 4.15 Analytic rank of several elliptic curves E/Q

Chapter 5
Introduction to Hecke theory and applications

5.1 Hecke operators of the modular group and their eigenforms

Definition 5.1 (Multiplicative function). A function

f : N∗ →
− N∗

is muliplicative if for each pair of coprime integers m, n ∈ N∗ holds

f (n · m) = f (n) · f (m)

Proposition 5.2 (Multiplicativeness of the power sum function). For each k ∈ N

the function
σ k : N∗ →
− N∗ , σk (n) := ∑ d k ,
d|n

is multiplicative. It has the product representation

(e j +1)k
N pj −1
σk (n) = ∏
j=1 pkj − 1

Proof. The prime decomposition

N
e
n = ∏ p j j , n ∈ N∗ ,
j=1

and the formula for the finite geometric series imply the product formula
   
k·e1 k·eN
σk (n) = pk1 + pk·2
1 + ... + p1 · ... · pkN + pk·2
N + ... + pN =

181
182 5 Introduction to Hecke theory and applications

ej
! (e j +1)k
N  ν N pj −1
=∏ ∑ pkj =∏
j=1 ν=1 j=1 pkj − 1
And the product formula implies for coprime n, m ∈ N∗

σk (n) · σk (m) = σk (n · m), q.e.d.

Corollary 5.3 (Eisenstein series are multiplicative). For each even k ≥ 4 the
Fourier coefficents of the Eisenstein series

− Bk ∞
· Ek (τ) = ∑ cν · qν , q = e2πi·τ ,
2k ν=0

are multiplicative for ν ≥ 1.

Proof. We reduce the statement of the corollary to Proposition 5.2: Corollary 3.11
shows
2k ∞
Ek (τ) = 1 − · ∑ σk−1 (ν) · qν , q = e2πi·τ
Bk ν=1
and
− Bk − Bk ∞
· Ek (τ) = + ∑ σk−1 (ν) · qν ,
2k 2k ν=1

with Fourier coefficents

cν = σk−1 (ν), ν ≥ 1, q.e.d.

Corollary 5.3 starts with the multiplicativeness of a well-known arithmetic func-

tion and derives the multiplicativeness of the Fourier coefficients of certain modular
forms. Hecke theory goes the way in the opposite direction: Hecke operators single
out certain modular forms and derive the multiplicativeness of their Fourier coeffi-
cients. Historically the first example is the normalized discriminant form

∆
(2π)12

which gives the multiplicativeness of Ramanujan’s τ-function.

Hecke operators are linear maps on modular forms, more precisely: For each
even weight k ∈ N the Hecke algebra

Hk (Γ ) ⊂ End(Mk (Γ ))

is a commutative subalgebra of endomorphisms, which restrict to endomorphisms

of the vector spaces of cusp forms Sk (Γ ) ⊂ Mk (Γ ). For fixed weight the Hecke
algebra Hk (Γ ) is generated by a family of Hecke operators (Tm )m∈N . They average
5.1 Hecke operators of the modular group and their eigenforms 183

the transformation behaviour of modular forms with respect to certain equivalence

classes of matrices
Γm ⊂ GL(2, R)+
The equivalence classes are the orbits of a left action of the modular group on these
matrices, which we have to study first.

Besides the two sequences of levels N ∈ N∗ from Definition 2.9 and weights k ∈ Z
from Definition 3.6 we now introduce a third sequence m ∈ N∗ . It is the sequence
of the sets of integral matrices with determinant = m, see Definition 5.4. These ma-
trices generalize the modular group Γ with respect to the determinant. They serve
to define the Hecke operators which average the values of modular forms of a given
weight and level. In addition, these integral matrices will serve in Section 6.2 to
average as class invariants distinguished values of the j-invariant.

Definition 5.4 (Integral matrices with positive determinant). For any m ∈ N∗

consider the set of matrices

Γm := {M ∈ M(2 × 2, Z) : det M = m} ⊂ GL(2, Q)+

and its subset   

ab
Γm,prim := ∈ Γm : (a, b, c, d) = 1
cd
of primitive matrices. We denote by

Φm : Γ × Γm → Γm , (A, M) 7→ A · M

the canonical Γ -left action, and by

Φm,prim := Φm |Γm,prim : Γ × Γm,prim →

− Γm,prim

its restriction. The respective orbits sets are denoted as the left quotients

Γ \Γm and Γ \Γm,prim .

Remark 5.5 (Restriction of the Γ -action to the primitive matrices). The subset

Γm,prim ⊂ Γm

is stable under the Γ -left action Φm .

Proof. For the proof consider any

   
αβ ab
A= ∈ Γ and M = ∈ Γm,prim .
γ δ cd
184 5 Introduction to Hecke theory and applications

The ideal generated by the coefficients of the product matrix A · M

I := (αa + β c, αb + β d, γa + δ c, γb + δ d) ⊂ Z

contains the element

δ · (αa + β c) + (−β ) · (γa + δ c) = (αδ − β γ)a + (δ β − β δ )c = a

and similar for b, c, d with the other generators of I. Therefore

(1) = (a, b, c, d) ⊂ I

which implies
(1) = I
Hence the coefficients of the product matrix A · M are coprime, i.e.

Φm (A, M) = A · M ∈ Γm,prim , q.e.d.

Lemma 5.6 shows that each orbit of the group action can be represented by a
triangular matrix which facilitates many calculations.
Lemma 5.6 (Orbit sets of the left action). Consider a positive integer m ∈ N∗ .
1. A bijective map exists
  
' ab
Γ \Γm −
→ V (m) := ∈ Γm : 0 ≤ b < d
0d

to a subset of upper triangular matrices. The orbit set of Φm has cardinality σ1 (m).

2. A bijective map exists

  
' ab
Γ \Γm,prim −
→ V (m, prim) := ∈ V (m) : (a, b, d) = 1 .
0d

The orbit set of Φm,prim has cardinality

ψ(m) := m · ∏ (1 + (1/p))
p|m
p prime

Proof. 1. We define a map

p : Γm → V (m),
which attaches to each M ∈ Γm a distinguished triangular matrix p(M) belonging
to the same orbit. And we show that the matrix p(M) depends only on the orbit
of M:
5.1 Hecke operators of the modular group and their eigenforms 185

i) Consider a fixed but arbitrary element

 
ab
M= ∈ Γm .
cd
We construct a matrix  
αβ
A= ∈Γ
γ δ
such that  
αa + β c αb + β d
A·M = ∈ V (m) :
γa + δ c γb + δ d
First, in order to remove the entry c of M we choose coprime integers γ, δ ∈ Z
with
γ · a + δ · c = 0.
In order to obtain a matrix with det = 1 we choose integers α, β ∈ Z with

α · δ − β · γ = 1.

Such a choice is possible because γ and δ are coprime. Set

 
αβ
A1 := ∈Γ.
γ δ

Then  0 0
a b
A1 · M =
0 d0
and
det (A1 · M) = det M = a0 · d 0 = m.
In particular d 0 6= 0 and possibly after replacing A1 by −A1 even a0 , d 0 > 0.

Secondly, because T ∈ Γ , for any k ∈ Z also

A2 (k) := T k · A1 ∈ Γ .

Then    0 0  0 0
a b + k · d0

1k a b
A2 (k) · M = · = .
01 0 d0 0 d0
We choose k0 ∈ Z such that

0 ≤ b0 + k0 · d 0 < d 0

and set
A := A2 (k0 ).
ii) The construction from part i) determines uniquely the element

A · M ∈ V (m) :
186 5 Introduction to Hecke theory and applications

Assume two matrices Ai ∈ Γ such that

Ai · M =: Mi ∈ V (m), i = 1, 2.

Then

M = A−1 −1 −1
1 · M1 = A2 · M2 or A21 · M1 = M2 , A21 := A2 · A1 .

Set    
αβ ai bi
A21 =: and Mi := , i = 1, 2.
γ δ 0 di
From      
αβ a b a b
· 1 1 = 2 2
γ δ 0 d1 0 d2
follows:
• γ · a1 = 0, hence γ = 0.

• From d2 = δ · d1 follows δ > 0. Together with α · δ = 1 it implies α = δ = 1.

• As a consequence a1 = a2 , d1 = d2 , and b1 + β · d1 = b2 .

• Because 0 ≤ b1 , b2 < d1 = d2 we get b1 = b2 and β = 0.

As a consequence  
10
A21 = or A1 = A2 .
01
iii) Part i) and ii) show that

p : Γm → V (m), M 7→ A · M,

is well-defined, and p(M) belongs to the same orbit as M.

iv) The map p is surjectice because V (m) ⊂ Γm . In addition,

p(M1 ) = p(M2 ) ⇐⇒ ∃ A ∈ Γ with A · M1 = M2

⇐⇒ M1 and M2 on the same orbit of Φm .

As a consequence p induces the bijective map
'
→ V (m), M 7→ p(M).
Γ \Γm −

v) The equality
card V (m) = σm (1)
follows at once from the explicit representation of the elements of V (m): Each
coefficient d ≥ 0 satisfies d|m and allows d different coefficients b.
5.1 Hecke operators of the modular group and their eigenforms 187

2. The representation of the orbit set Γ \Γm,prim follows from part 1) by restriction.
For the cardinality of V (m, prim) see [38, Chap. 5, §1].

Corollary 5.7 (Complete representatives of the orbit set). For each γ ∈ Γ :

1. The set
V (m) · γ ⊂ Γm
is a complete set of representatives of the orbit set Γ \Γm , i.e. with respect to the
canonical projection
Γm → Γ \Γm , M 7→ [M],
holds the equality of sets

Γ \Γm = {[M] : M ∈ V (m)} = {[M · γ] : M ∈ V (m)}.

2. The set
V (m, prim) · γ ⊂ Γm,prim
is a complete set of representatives of the orbit set Γ \Γm,prim .
Proof. 1. Consider a given γ ∈ Γ . We show: For arbitrary but fixed M ∈ V (m) exist
an element
M1 ∈ V (m)
and a unique matrix A ∈ Γ such that

A · M = M1 · γ.

According to Lemma 5.6, applied to M · γ −1 ∈ Γm , a unique element A ∈ Γ exists

with
M1 := A · (M · γ −1 ) ∈ V (m).
Then
M1 · γ = A · M.
Hence M1 · γ lies on the same orbit as M. As a consequence, the map

V (m) →
− V (m), [M] 7→ [M · γ],

is well-defined and injective.

2. Analogously, q.e.d.

Lemma 5.8 (Complete sets of common representatives for left/right-action).

Consider an arbitrary but fixed prime p, and the left action

Φle f t : Γ × Γp →
− Γp , (γ, M) 7→ γ · M
188 5 Introduction to Hecke theory and applications

1. Besides the set V (p) from Lemma 5.6 the set of symmetric matrices
    
p0 1 b
Vsym := ∪ : 0≤b< p
01 b p + b2

is a complete set of representatives of Φle f t too.

2. Consider a complete set V of representatives of Φle f t formed by symmetric ma-

trices. Then also the system of adjoints

V ] := {M ] : M ∈ V }, M ] := (det M) · M −1 ,

is a complete systems of representatives of Φle f t .

Proof. 1. For each 0 ≤ b < p

     
1 b 1 0 1 b
= ·
b p + b2 b 1 0 p

Hence the matrices    

1 b 1 b
and
b p + b2 0 p
belong to the same orbit of Φle f t .

2. i) Let
Φright : Γp × Γ →
− Γp , (M, γ) 7→ M · γ,
be the corresponding right action of Γ . For each complete system V of represen-
tatives of Φle f t the set of transposed elements

V > := {v> : v ∈ V }

is a complete system of representatives of Φright : Assume v ∈ Γp . Then also v> ∈ Γp

and
v> = A · M
for suitable
A ∈ Γ and M ∈ V .
Hence
v = M > · A>
with A> ∈ Γ > = Γ .

As a consequence, if V is formed by symmetric matrices, then V is also a com-

plete system of representatives for Φright .

ii) For given w ∈ Γp set

5.1 Hecke operators of the modular group and their eigenforms 189

v := (det w) · w−1 ∈ Γp .
Then

v = w] and det v = (det w)2 · det w−1 = det w and w = (det v) · v−1

Due to part i) exist

A ∈ Γ and M ∈ V
with
v = M·A
Then
det v = det M
and
v−1 = A−1 · M −1
and
w = v−1 · det v = A−1 · ((det M) · M −1 )
with
A−1 ∈ Γ −1 = Γ , q.e.d.

We recall a simple version of the elementary divisor theorem.

Proposition 5.9 (Elementary divisor theorem).
1. For each matrix
A ∈ Γm , m ∈ N∗ ,
exist two matrices γ1 , γ2 ∈ Γ with
 
e 0
γ1 · A · γ2 = 1 satisfying the divisor relation e1 |e2 .
0 e2

2. For each pair of primitive matrices

A1 , A2 ∈ Γm,prim , m ∈ N∗ ,

exist two matrices γ1 , γ2 ∈ Γ with

A2 = γ1 · A1 · γ2 .

Proposition 5.9 is a special case of the Smith normal form, a result which holds for
matrices over principal ideal domains.

Definition 5.10 (Hecke operators of the modular group). For any positive integer m ≥ 1
the m-th Hecke operator is the C-linear endomorphism of the algebra of modular
forms
190 5 Introduction to Hecke theory and applications

Tm : M∗ (Γ ) → M∗ (Γ )
which is defined on Mk (Γ ), k ∈ N, as

Mk (Γ ) → Mk (Γ ), f 7→ Tm f := ∑ f [M]k .
Γ \Γm

Here the summation extends over an arbitrary complete set of representatives M ∈ Γm

of the orbit set Γ \Γm of the left action

Γ × Γm →
− Γm .

The value f [M]k does not depend on the choice of a representative from the
orbit Γ · M because due to Lemma 3.4: For all α ∈ Γ
f [α · M]k = ( f [α]k )[M]k = f [M]k
E.g., the summation in Definition 5.10 of the Hecke operator Tm may extend over
the elements of V (m) or over the elements of V (m) · γ, γ ∈ Γ . Note T1 = id.

Theorem 5.11 (Fourier coefficients of Hecke transforms).

1. The Hecke operators from Definition 5.10 are well-defined, i.e. for each even
weight k and for all m ≥ 1

f ∈ Mk (Γ ) =⇒ Tm f ∈ Mk (Γ ).

2. For a modular form f ∈ Mk (Γ ) with Fourier expansion

∞
f (τ) = ∑ an · qn , q = e2πi·τ ,
n=0

its Hecke transform Tm f ∈ Mk (Γ ) has the Fourier expansion

∞
(Tm f )(τ) = ∑ bn · qn
n=0

with Fourier coefficients

bn = ∑ rk−1 · amn/r2 , n ∈ N,
r>0
r∈(m,n)

in particular
b0 = a0 · σk−1 (m) and b1 = am
3. The ideal of cusp forms is stable under all Hecke operators, i.e. for each m ∈ N∗

Tm (S∗ (Γ )) ⊂ S∗ (Γ ).
5.1 Hecke operators of the modular group and their eigenforms 191

Due to Theorem 5.11, part 1) we will use the same notation Tm for a Hecke operator
and its restriction to a single modular space of fixed weight.
Proof. 1. To prove that Tm f is a modular form of the same weight as f we have to
verify the properties from Definition 3.6:
• Appararently Tm f is holomorphic on H.

• Weakly modularity: Consider a fixed but arbitrary matrix γ ∈ Γ . We replace

the summation over the orbit set V (m) by the summation over the complete
set of representatives V (m) · γ, see Corollary 5.7:

(Tm f )[γ]k = ∑ ( f [M]k )[γ]k = ∑ ( f [M · γ]k ) =

M∈V (m) M∈V (m)

= ∑ f [N]k = ∑ f [M]k = Tm f .
N∈V (m)·γ M∈V (m)

• Holomorphy of Tm f at the point ∞: To compute the Fourier series of Tm f we

use the summation with
 
a b
M= ∈ V (m)
0 d

For the computation of

( f [M]k )(τ) = f (M(τ)) · h(M, τ)−k · (det M)k−1

we employ the following intermediate results:

i)
!
aτ + b
exp(2πin · M(τ)) = exp 2πin · = e2πin·(aτ/d) · e2πin·(b/d) , n ∈ N.
d

ii) (
2πib·(n/d) d if d|n
∑ e =
0≤b<d 0 otherwise
The first case follows from

e2πib·(n/d) = 1

for each of the d summands with 0 ≤ b < d. The second case follows from
the formula of the finite geometric series with argument e2πi·(n/d) :
d−1 e2πi·d·(n/d) − 1
∑ e2πib·(n/d) = e2πi·(n/d) − 1
=0
b=0
192 5 Introduction to Hecke theory and applications

iii) We have
! !
aτ + b ∞ aτ + b
f (M(τ)) = f = ∑ an · exp 2πin · =
d n=0 d
! !
∞ aτ b
= ∑ an · exp 2πin · · exp 2πin ·
n=0 d d
Hence
d−1 ∞
∑ f (M(τ)) = ∑ d · and · qn(m/d) , q = exp(2πinτ),
b=0 n=0

because
!
d−1 d−1 ∞
2πin(aτ)/d 2πin(b/d)
∑ f (M(τ)) = ∑ ∑ an · e ·e =
b=0 b=0 n=0
!
∞ d−1 ∞
2πin(aτ)/d 2πin(b/d)
= ∑ an · e · ∑e = ∑ and · e2πinaτ · d =
n=0 b=0 n=0
∞ ∞
= d · ∑ and · qna = d · ∑ and · qn(m/d)
n=0 n=0

Due to part ii) the third last sum retains only summands with index = nd.

We now use the representation of the orbit set Γ \Γm from Lemma 5.6 and
insert the Fourier series of f . We obtain due to the formula from part iii)
(Tm f )(τ) = ∑ f [M]k (τ) = mk−1 · ∑ h(M, τ)−k · f (M(τ)) =
M∈V (m) M∈V (m)

!k−1 !
k−1 1 ∞ m ∞
=m ∑ d k · ∑ and · qn(m/d) · d = ∑ d ∑ and · q n(m/d)
=
d>0 n=0 d>0 n=0
d|m d|m

!k−1 !
m ∞
= ∑ d ∑ an · q(n/d)·(m/d)
d>0 n=0
d∈(m,n)

The substitution

(d, n) 7→ (t = (n/d) · (m/d), r = m/d)

provides
!2
d2 · t d tm
n= = tm = .
m m r2
5.1 Hecke operators of the modular group and their eigenforms 193

Hence  
∞
(Tm f )(τ) = rk−1 · a(tm)/r2  q t
 
∑ ∑
t=0 r>0
r|(m,t)

with the inner sum being finite. Apparently, the Fourier series has no coeffi-
cients with negative index. Hence Tm f is holomorphic at ∞ ∈ H∗ .

2. For the proof see the Fourier series from the previous part.

3. A cusp form satisfies a0 = 0. The formula for the Fourier coefficients of Tm f

from part 2) shows that Tm f is a cusp form again, q.e.d.

The following corollary specializes the general formula from Theorem 5.11. It
gives the formula for the Fourier coefficients of the Hecke transform for prime in-
dices.
Corollary 5.12 (Fourier coefficients with prime index of Hecke transforms).
Consider a modular form f ∈ Mk (Γ ) with Fourier coefficients (an )n∈N . Then the
Fourier coefficients (bn )n∈N of its Hecke transform Tm f satisfy for each prime
index p

(
amp if p - m
bp =
amp + pk−1 · am/p if p|m

Proof. The claim follows from the third formula for the general case in Theorem 5.11:
• If p - m then r|(p, m) implies r = 1.

• If p|m then r|(p, m) implies r = 1 or r = p; note n = p, q.e.d.

Corollary 5.13 (Simultaneous eigenvector of all Hecke operators). The discrim-

inant modular form
∆ ∈ S12 (Γ )
is a simultaneous eigenform of all Hecke operators, i.e. ∆ is an eigenvector of all
Hecke operators
Tm ∈ End(S12 (Γ )), m ≥ 1.
The Fourier coefficients of the normalized discriminant modular form
194 5 Introduction to Hecke theory and applications

∆ ∞
12
= ∑ τm · qm
(2π) m=1

are the corresponding eigenvalues, i.e. for all m ∈ N∗

Tm (∆ ) = τm · ∆ .

Proof. According to Corollary 3.19 and Lemma 3.23 the vector space S12 (Γ ) of
cusp forms of weight k = 12 is 1-dimensional. It has the discriminant modular
form ∆ from Definition 3.13 as a basis. Hence ∆ is an eigenvector of each Hecke
operator Tm , m ≥ 1. To calculate the corresponding eigenvalue we recall the Fourier
expansion of the normalized discriminant modular form from Proposition 3.14:

∆ (τ) ∞
= ∑ τn · qn , τ1 = 1.
(2π)12 n=1

Theorem 5.11 provides the Fourier expansion

!
1
Tm · ∆ (τ) = τm · q + O(2).
(2π)12

Hence by comparing the linear term of the Fourier series on both sides

Tm (∆ ) = τm · ∆ , q.e.d.

Example 5.14 (Hecke operators). Consider the vector space of cusp forms S28 (Γ ).
Due to Theorem 3.19 and Lemma 3.23

dim S28 (Γ ) = 2.

The two cuspforms of weight 28

f1 := ∆ · E44 and f2 := ∆ 2 · E4

are linearly independent because

ord( f1 ; ∞) = 1 and ord( f2 ; ∞) = 2.

Hence ( f1 , f2 ) is a basis of S28 (Γ ). We consider the Hecke operator

T2 : S28 (Γ ) →
− S28 (Γ ).

The endomorphism has two distinct eigenvalues, hence it can be diagonalized. Fig-
ure 5.1 shows
• the Fourier series of f1 and f2
5.1 Hecke operators of the modular group and their eigenforms 195

• the Fourier series of T2 f1 and T2 f2

• the Hecke matrix of T2 with respect to the basis ( f1 , f2 )

• the eigenvalues of T2

• the Fourier coefficients of two corresponding eigenvectors (v1 , v2 ) of T2 , normal-

ized with linear Fourier coefficient = 1,

• the representation of the eigenvectors with respect to ( f1 , f2 )

Fig. 5.1 Hecke operator T2 ∈ End(S28 (Γ ))

The result from Figure 5.1 has been computed by the PARI script
Hecke_matrix_02.

We give an analogue to the Hecke operators from the theory of quadrics. The
preface of the textbook [17] chooses the analogue as an introduction to the theory
of modular forms. Corollary 5.13 represents the discriminant modular form ∆ as si-
multaneous eigenform of all Hecke operators (Tm )m≥1 acting on the complex vector
space
196 5 Introduction to Hecke theory and applications

V := S12 (Γ ).
The corresponding eigenvalues are the Fourier coefficients of ∆ . Remark 5.15 trans-
lates the law of quadratic reciprocity from algebraic number theory into a similar
shape.

Remark 5.15 (Reduction of quadrics mod p).

1. The Legendre symbol considers an odd prime p ∈ Z, and determines for each
integer d ∈ Z whether d is a square mod p. The Legendre symbol is defined as

  
d +1 d 6≡ 0 mod p and d quadratic rest mod p
:= −1 d 6≡ 0 mod p and d not quadratic rest mod p
p 
0 d ≡ 0 mod p


The law of quadratic reciprocity relates the Legendre symbol of two odd primes
p 6= q according to the formula
  
q
    p
 p ≡ 1 mod 4 or q ≡ 1 mod 4
p
=
q   
− q

p ≡ 3 mod 4 and q ≡ 3 mod 4
p

2. Consider an arbitrary, but fixed integer d 6= 0 and the quadratic polynomial

F(X) := X 2 − d ∈ Z[X]

which defines the affine quadric Q/Q

{x ∈ C : F(x) = 0}.

For each prime p we consider the reduction mod p

Fp (X) := X 2 − d p ∈ F p [X]

and the set Q(F p ) of points of Q with coordinates from F p .

Analogously to Remark 4.42 for elliptic curves we define

a p (Q) := card Q(F p ) − 1.

The number 1 in the definition is the average number of solutions mod p. The
values a p (Q) extend from prime indices to an (Q), n ∈ N, and satisfy the product
rule
anm (Q) = an (Q) · am (Q), n, m ∈ N∗ .
For odd prime p the number a p (Q) equals the Legendre symbol
5.1 Hecke operators of the modular group and their eigenforms 197
 
d
a p (Q) = .
p

Due to the Law of Quadratic Reciprocity and its two supplements the Legendre
symbol depends only on the residue class p mod 4d.

3. We now represent the sequence (a p (Q)) p prime as a sequence of eigenvalues of a

simultaneous eigenvector for a suitable sequence of linear endomorphisms (Tp ) p prime .
Our aim is to formulate an analogue of the modular group, of modular functions
and of Hecke operators:

Set
N = NQ := 4|d|,
and consider the multiplicative group

G := (Z/NZ)∗ ,

and the complex vector space of functions defined on G

V := { f : G → C}.

Eventually, as an analogue to the Hecke operators we introduce the family of

linear endomorphisms (Tp ) p prime of V defined as
(
f (p · n) p - N
(Tp f )(n) :=
0 p|N

for n ∈ G. We claim: All endomorphism (Tp ) p prime have a simultaneous eigen-

vector, namely the distinguished function

fQ : G → C, f (n) := an (Q).

Note that fQ is well-defined on the residue classes of G because the Legendre

symbol depends only on the residue class p mod 4d. Computing the eigenvalues
we get for all n ∈ G
(
fQ (p · n) = a pn (Q) = a p (Q) · an (Q) p - N
Tp ( fQ )(n) =
0 p|N

As a consequence, for all primes p ∈ Z

Tp ( fQ ) = a p (Q) · fQ .

It is not by chance that in Example 5.14 the Hecke endomorphism T2 ∈ End(S28 (Γ ))

can be diagonalized. Theorem 5.18 will show: The Hecke algebra of a given weight
198 5 Introduction to Hecke theory and applications

is commutative. And Theorem 5.26 will show that the Hecke operators are selfad-
joint with respect to a certain Hermitian scalar product on the cusp forms. Hence all
Hecke operators on cusp forms can be simultaneously diagonalized.

Definition 5.16 (Hecke algebra of weight k). The Hecke algebra of even weight k ∈ N

Heckek (Γ ) ⊂ End(Mk (Γ ))

is the subalgebra spanned by the family (Tm |Mk (Γ ))m∈N of Hecke operators.

To prepare the proof of Theorem 5.18 about the Hecke algebra we consider for two
matrices
 
aj bj
Mj =   ∈ V (n j ), j = 1, 2
0 dj
the components of the product
 
a1 a2 a1 b2 + b1 d2
M1 · M2 =   ∈ Γn ·n
1 2
0 d1 d2

We want to show that the products form a complete set of representatives of

the Γ -left action on Γn1 ·n2

Lemma 5.17 (Restsystem for V (n1 · n2 )). For j = 1, 2 consider two coprime num-
bers
n j ∈ N∗ , j = 1, 2.
1. For given positive numbers a j , d j with n j = a j · d j the map of restsystems

φ : {0, ..., d1 − 1} × {0, ..., d2 − 1} →

− {0, ..., d1 · d2 }, (b1 , b2 ) 7→ a1 · b2 + b1 · d2 ,

is bijective.

2. The map on the product of complete systems of representatives

V (n1 ) ×V (n2 ) →
− Γn1 ·n2 , (M1 , M2 ) 7→ M1 · M2 ,

maps injectively onto a complete system of representatives of Γ \Γn1 ·n2 .

Proof. 1. It suffices to show that φ is injective because its domain and codomain
have the same finite cardinality. Therefore assume

(b̃1 , b̃2 ) ∈ (Z/d1 Z) × (Z/d2 Z) with φ ((b1 , b2 )) = φ ((b̃1 , b̃2 ))

5.1 Hecke operators of the modular group and their eigenforms 199

i.e.
a1 · b2 + b1 · d2 ≡ a1 · b̃2 + b̃1 · d2 mod d1 d2
Then
a1 (b2 − b̃2 ) ≡ (b̃1 − b1 )d2 mod d1 d2
which implies
a1 (b2 − b̃2 ) ≡ 0 mod d2
Because
gcd(n1 , n2 ) = 1 =⇒ gcd(a1 , d2 ) = 1
therefore
b1 − b̃1 ≡ 0 mod d1 , i.e. b1 ≡ 1̃2 mod d1
which implies
b2 = b̃2 .
Analogously b1 = b̃1 .

2. In addition to part 1) one uses the bijectivity of the maps of divisor sets

{a1 ∈ N∗ : a1 |n1 }×{a2 ∈ N∗ : a2 |n2 } →

− {a ∈ N∗ : a|n1 n2 }, (a1 , a2 ) 7→ a1 a2 , q.e.d.

Theorem 5.18 (The Hecke algebras are commutative). For each weight k ∈ N
the Hecke algebra Heckek (Γ ) is commutative. It has the following properties:
1. For coprime n, m ∈ N∗ Hecke operators are multiplicative:

Tm ◦ Tn = Tm·n

2. For a prime power pr , r ≥ 1, holds

Tpr ◦ Tp = Tpr+1 + pk−1 · Tpr−1

In particular
Tpr ∈ Z[Tp ]
3. The algebra Heckek (Γ ) is generated by the family Tp , p prime.

Proof. 1. For each f ∈ Mk (Γ ) Lemma 5.17 and Lemma 3.4 imply

!
(Tm ◦Tn )( f ) = Tm (Tn f ) = ∑ (Tn f )[M2 ]k = ∑ ∑ ( f [M1 ]k )[M2 ]k =
M2 ∈V (m) M2 ∈V (m) M1 ∈V (n)

= ∑ f [M1 · M2 ]k = ∑ f [M]k = Tnm f

M1 ∈V (n) M∈V (n·m)
M2 ∈V (m)
200 5 Introduction to Hecke theory and applications

2. The proof is by induction. First one replaces Lemma 5.17 by two analogous re-
sults about restsystems. Secondly, one evaluates the formula for the Fourier coef-
ficients of Hecke transforms from Theorem 5.11 and Corollary 5.12, cf. [36, Kap. IV, §2.2].

3. Due to part 1) the family (Tm )m∈N of all Hecke operators generates the same
algebra as the family of Hecke operators with all prime power indices

Tpr , p prime , r ∈ N.

Due to part 2) for each fixed prime p the family Tpr , r ∈ N, generates the same
algebra as the single Hecke operator Tp .

The commutativity of the Hecke algebras follows from the parts 1) and 3) , q.e.d.

5.2 The Petersson scalar product

Next we introduce a Hermitian scalar product on the vector space of cusp forms of
a given congruence subgroup Γ0 (N).

Definition 5.19 (Hyperbolic volume form). The hyperbolic volume form on H is

the differential form
dx ∧ dy
dµ(τ) := , τ = x + iy ∈ H
y2

Lemma 5.20 (Invariance of the hyperbolic volume form). The hyperbolic vol-
ume form is GL(2, R)+ -invariant, i.e. for each γ ∈ GL(2, R)+ holds

γ ∗ dµ = dµ

Proof. For each  

ab
γ= ∈ GL(2, R)+
cd
we have to show: For all τ ∈ H

(γ ∗ dµ)(τ) = dµ(τ) :

For the proof one uses the usual coordinate transformation

i
dx ∧ dy = dτ ∧ dτ
2
5.2 The Petersson scalar product 201

which implies
i dτ ∧ dτ
dµ(τ) = ·
2 (Im τ)2
and
i dγ(τ) ∧ dγ(τ)
(γ ∗ dµ)(τ) = ·
2 (Im γ(τ))2
The proof of Proposition 3.1 and the formulas from Remark 2.8

Im τ dτ
Im γ(τ) = det(γ) · and dγ(τ) = det(γ) ·
|cτ + d|2 (cτ + d)2

imply
(γ ∗ dµ)(τ) = dµ(τ), q.e.d.

Proposition 5.21 (Integration over fundamental domains).

1. The standard fundamental domain D of Γ has the volume
Z
π
vol D := dµ =
D 3
2. The fundamental domain of a congruence subgroup Γ0 (N) has the volume
Z
π
vol D(Γ0 (N)) := dµ = [Γ : Γ0 (N)] · vol D = · N · ∏ (1 + (1/p)) ,
D(Γ0 (N)) 3 p|N

independent from the chosen left coset decomposition of Γ with respect to Γ0 (N).

3. Consider a continuous function

Ω :H→
− C

with fixed group

Γ (Ω ) := {A ∈ Γ : Ω ◦ A = Ω }.
Then for each subgroup K ⊂ Γ (Ω ) with ±1 ∈ K and for each pair(F1 , F2 ) of
fundamental domains of K:
Z Z
Ω dµ = Ω dµ
F1 F2

if at least one of the two integrals is finite.

Apparently, part 2) of Proposition 5.21 is a special case of part 3).

202 5 Introduction to Hecke theory and applications

Proof. 1. From the geometrical description of D, see Figure 2.1:

Z 1/2 Z ∞
! Z 1/2  
Z
dy ∞
dµ = √ 2
dx = (−1/y) √ dx =
D −1/2 1−x2 y −1/2 1−x2

Z 1/2
dx 1/2
= √ = arcsin x = π/6 − (−π/6) = π/3
−1/2 1 − x2 −1/2

2. The general formula for

vol (D(Γ0 (N))) = [Γ : Γ0 (N)] · vol D

follows from
• the index formula Proposition 2.14,

• the representation of the fundamental domain as disjoint union

[Γ :Γ0 (N)]
D(Γ0 (N)) = γ −1
[
j (D)
j=1

see Remark 2.17,

• and the invariance of the volume form dµ according to Lemma 5.20.

3. For A ∈ Γ (φ ) the usual transformation formula for integrals and the invariance
of the hyperbolic volume form dµ imply for each measurable set B ⊂ H
Z Z
Ω dµ = Ω dµ
B A(B)

If one splits H as the disjoint union of the sets A(F2 ), A ∈ K, and observes
that ±1 act as identity on H, which introduces the factor = 1/2, then
Z Z Z
Ω dµ = (1/2) · ∑ Ω dµ = (1/2) · ∑ Ω dµ =
F1 A∈K A(F2 )∩F1 A∈K F2 ∩A (F1 )
−1

Z Z
= (1/2) · ∑ Ω dµ = Ω dµ, q.e.d.
A∈K A(F1 )∩F2 F2

Definition 5.22 (Petersson scalar product). For each level N ∈ N and each even
weight k ∈ N, k ≥ 0, the Petersson scalar product on the cusp space Sk (Γ0 (N)) is
the C-sesquilinear map

< −, − >Γ0 (N) : Sk (Γ0 (N)) × Sk (Γ0 (N)) →

− C,
5.2 The Petersson scalar product 203

1 Z
< f , g >Γ0 (N) := · f (τ) · g(τ) · (Im τ)k dµ(τ)
vol D(Γ0 (N)) D(Γ0 (N))

Note
vol D(Γ0 (N)) < ∞ because [Γ : Γ0 (N)] < ∞.
Definition 5.22, see [17, Chap. 5, Def. 5.4.1], implies: If
Γ1 := Γ0 (N1 ) ⊂ Γ2 := Γ0 (N2 ) and f , g ∈ Sk (Γ2 )
then
< f , g >Γ1 =< f , g >Γ2
While [36, Kap. IV, 3.2], which considers only the modular group Γ , employs in
Definition 5.22 instead before the integral the factor = 1.

Theorem 5.23 (Petersson scalar product). The map from Definition 5.22

< −, − >Γ0 (N)

is well-defined. It is a Hermitian scalar product on the vector space Sk (Γ0 (N)) of

cusp forms.

Proof. For each continuous and bounded function

φ :H→
− C

Proposition 5.21 implies the estimate

Z
|φ | dµ ≤ vol(D(Γ0 (N))) · kφ kH < ∞
D(Γ0 (N))

To prove that the Petersson scalar product of two modular forms f , g ∈ Sk (Γ0 (N)) is
well-defined, consider the continuous function

− C, φ (τ) := f (τ) · g(τ) · (Im τ)k .

φ :H→

i) To prove the boundedness of φ it suffices to prove for any γ ∈ Γ that the restriction

(φ ◦ γ)|D

is bounded: Being continuous φ ◦ γ is bounded on each compact subset of D. Sec-

ondly, the Fourier expansions
∞ ∞
f [γ]k (τ) = ∑ αn · qnh and g[γ]k (τ) = ∑ βn · qnh , qh = e2πiτ/h ,
n=1 n=1
204 5 Introduction to Hecke theory and applications

for a suitable h ∈ N, show that both cusp forms are of type O(qh ) in the limit Im τ → ∞.
Then
|qh | = |e2πiτ/h | = e−2π·(Im τ)/h =⇒ O(q2h ) · (Im τ)k = o(qh )
ii) The function φ is Γ -invariant because for each γ ∈ Γ and τ ∈ H according to
Definition 3.6
φ (γ(τ)) = f (γ(τ)) · g(γ(τ)) · (Im γ(τ))k =
k
= f [γ]k (τ) · h(γ, τ)k · g[γ]k (τ) · h(γ, τ) · (Im τ)k · |h(γ, τ)|−2k =
= f [γ]k (τ) · g[γ]k (τ) · (Im τ)k = f (τ) · g(τ) · (Im τ)k = φ (τ)
Hence part i) and Proposition 5.21, part 3) imply that the Petersson scalar product
is independent from the choice of a fundamental domain, which finishes the proof,
q.e.d.

Note. The proof of Theorem 5.23 shows that the Hermitian scalar
product < f , g >Γ0 (N) is also defined for two modular forms f , g ∈ Mk (Γ0 (N)) if at
least one of them is a cusp form.

Recall from Remark 2.8 the operation of GL(2, Q)+ and its subgroups on H

Lemma 5.24 (Petersson scalar product). Consider

• a weight k ∈ N and two cusp forms f , g ∈ Sk (Γ )

• together with a prime p and a matrix M ∈ Γp .

Then:
1. Transformation formula for subgroups of finite index: Each subgroup K ⊂ Γ (Ω )
with finite index [Γ (Ω ) : K] < ∞ satisfies

1 Z Z
· Ω dµ = Ω dµ
[Γ (Ω ) : K] D(K) D(Γ (Ω ))

if at least one of the two integrals is finite.

2. Finiteness of the index under conjugation: The extended principal congruence

subgroup of Γ of level p

Γ̃ (p) := {L ∈ Γ : L ≡ ±1 mod p}

has finite index. In addition, define for each pair of continuous functions

h1 , h2 : H →
− C
5.2 The Petersson scalar product 205

the function

− C, Ω (h1 , h2 )(τ) := (h1 · h2 )(τ) · (Im τ)k

Ω (h1 , h2 ) : H →

Then with respect to the fixed groups

Γ (Ω ( f [M]k , g)) and Γ (Ω ( f , g[M ] ]k ))

holds the inclusion of subgroups:

Γ̃ (p) ⊂ Γ (Ω ( f [M]k , g)) and M −1Γ̃ (p)M ⊂ Γ (Ω ( f [M]k , g)),

and both subgroups have finite index, and analogously for

Γ̃ (p) ⊂ Γ (Ω ( f , g[M ] ]k )) and MΓ̃ (p)M −1 ⊂ Γ (Ω ( f , g[M ] ]k )).

3. Invariance of the index under conjugation:

[Γ : Γ̃ (p)] = [Γ : M −1Γ̃ (p)M]

Proof. Cf. [36, Kap. IV, §3.3]. Because k is fixed during the whole computation we
simplify the notation [−] := [−]k .
1. Consider a fundamental domain F of Γ (Ω ). Then each decomposition of Γ (Ω )
into the finite number [Γ (Ω ) : K] of left classes with respect to K
˙
Mν−1 · K
[
Γ (Ω ) =
ν

provides a fundamental domain of K

˙
G=
[
Mν (F )
ν

Proposition 5.21, part 3), applied to each pair (F , Mν (F )) of fundamental do-

mains of Γ (Ω ), implies
Z Z
Ω dµ = Ω dµ,
Mν F F

hence Z Z Z
Ω dµ = Ω dµ = ∑ Ω dµ =
D(K) G ν Mν (F )
Z Z
=∑ Ω dµ = [Γ (φ ) : K] · Ω dµ
ν F D(Γ (Ω ))

2. i) According to Proposition 2.14

[Γ : Γ (p)] < ∞ hence also [Γ : Γ̃ (p)] < ∞.

206 5 Introduction to Hecke theory and applications

Each element L ∈ Γ̃ (p) ⊂ Γ with decomposition

L = ±1 + p · A, A ∈ M(2 × 2, Z).

has the conjugate

L0 := MLM −1 = M(±1 + p · A)M −1 = ±1 + M(pA)M −1 = ±1 + MAM ]

Hence
L0 M = ML and L0 ∈ Γ
which implies
( f [M])[L] = ( f [L0 ])[M] = f [M].
Because Γ̃ (p) ⊂ Γ also
g[L] = g.
The two L-invariant elements
f [M] and g
satisfy for τ ∈ H the transformation formula

( f [M] · g)(L(τ)) = ( f [M] · g)(τ) · |h(L, τ)|2k :

We compute

( f [M] · g)(L(τ)) = f [M](L(τ)) · g(L(τ)) = ( f [M][L])(τ) · h(L, τ)k · g(L(τ)) =

= f [M](τ) · h(L, τ)k · g[L](τ) · h(L, τ)k = f [M](τ) · g(τ) · |h(L, τ)|2k
Multiplying by (Im L(τ))k both sides of the transformation formula for

f [M] and g

and using the transformation formula for the imaginary part of the argument
Im L(τ) = (Im τ) · h(L, τ)−2
shows
( f [M] · g)(L(τ)) · (Im L(τ))k = ( f [M] · g)(τ) · (Im τ)k
or
Ω ( f [M], g) ◦ L = Ω ( f [M], g)
Because L ∈ Γ̃ (p) is arbitrary we have shown

Γ̃ (p) ⊂ Γ (Ω ( f [M], g))

The finiteness of the index of Γ̃ (p) in the fixgroups follow from the finiteness

[Γ : Γ̃ (p)] < ∞
5.2 The Petersson scalar product 207

ii) For arbitrary k ∈ N holds

MΓ̃ (kp)M −1 ⊂ Γ̃ (k) :

If L ∈ Γ̃ (kp) then

M · L · M ] ≡ ± M · 1 · M ] mod kp = ± p · 1 mod kp

1
M·L· · M ] ≡ ± 1 mod k i.e. M · L · M −1 ∈ Γ̃ (k).
det M
Now setting k = 1 shows
MΓ̃ (p)M −1 ⊂ Γ̃ ,
while setting k = p shows

MΓ̃ (p2 )M −1 ⊂ Γ̃ (p) or Γ̃ (p2 ) ⊂ M −1Γ̃ (p)M

As a consequence

[Γ : M −1Γ̃ (p)M] ≤ [Γ : Γ̃ (p2 )] < ∞

Finally one checks

M −1Γ̃ (p)M ⊂ Γ (Ω ( f [M], g)).
iii) The claim about the subgroups related to M ] follows from part i) and ii) by
interchanging the role of f and g and the role of M and M ] .

3. We consider the two subgroups of Γ from part 2)

Γ̃ (p) ⊂ Γ and K := M −1Γ̃ (p)M.

Consider a fundamental domain F ⊂ H of Γ̃ (p). Then M −1 F is a fundamental

domain of K: For each x ∈ H also M(x) ∈ H. Hence by assumption exists a
unique γ ∈ Γ̃ (p) with
γ(M(x)) ∈ F .
Then
M −1 (γ(M(x)) = (M −1 γM)(x) ∈ M −1 F
The uniqueness of γ implies the uniqueness of M −1 γM.

We compute the volume of both fundamental domains, hereby using

the GL(2, Q)+ -invariance of the hyperbolic volume form dµ according to
Lemma 5.20
Z
[Γ : Γ̃ (p)] · vol D = vol F = dµ =
F
Z
= dµ = vol(M −1 F ) = [Γ : K] · vol D
M −1 F
208 5 Introduction to Hecke theory and applications

Hence
[Γ : Γ̃ (p)] = [Γ : K] = [Γ : M −1Γ̃ (p)M],
in particular independent from M ∈ Γp , q.e.d.

Theorem 5.25 (Selfadjointness of the Hecke operators). For each even weight k ≥ 0
the Hecke operators
Tm ∈ End(Sk (Γ )), m ∈ N,
are self-adjoint with respect to the Petersson scalar product, i.e. for all f , g ∈ Sk (Γ )

< Tm f , g >=< f , Tm g >

Proof. Due to Theorem 5.18 it suffices to prove the claim for m = p prime. Due to
Lemma 5.8 there exists a complete set V (p) of representatives of the Γ -action

Φle f t : Γ × Γp →
− Γp

such that also V (p)] , the set of adjoints of the matrices from V (p), is a complete
set of representatives.

i) Transformation formula for the integrand: We consider the function Ω (h1 , h2 )

defined as
Ω (h1 , h2 )(τ) := (h1 · h2 )(τ) · (Im τ)k , τ ∈ H,
which has been introduced in Lemma 5.24, part 3). To simplify the notations we
set for M ∈ V (p)

ΩM,le f t := Ω ( f [M]k , g) and ΩM0 ,right := Ω ( f , g[M 0 ]k )

Explicit computation with the definition of the [−]k -operation on Mk (Γ ) proves

ΩM,le f t = ΩM0 ,right ◦ M

ii) Existence of the integrals: Denote by

˙
F=
[
γ(D)
γ

the fundamental domain of Γ̃ (p) resulting from a finite coset representation of Γ

with respect to the subgroup Γ̃ (p) ⊂ Γ . We show for M ∈ V (p)
Z
ΩM,le f t dµ exists
F

by proving for γ ∈ Γ Z
ΩM,le f t dµ exists :
γ(D)
5.2 The Petersson scalar product 209

Similarly to the proof of Theorem 5.23, part ii) one checks on D an analogue of the
transformation formula. First,

ΩM,le f t ◦ γ = Ω ( f [M]k , g) ◦ γ = Ω (( f [M]k )[γ]k , g[γ]k ) = Ω ( f [M · γ]k , g[γ]k )

Due to Lemma 5.6 there exists a matrix A ∈ Γ with

A · (M · γ) = L with an upper triangular matrix L ∈ V (p)

Set
γle f t := A−1 ∈ Γ ,
then
M · γ = γle f t · L.
Secondly, using that f and g are modular forms with respect to Γ

Ω ( f [M · γ]k , g[γ]k ) = Ω ( f [γle f t · L]k , g[γ]k ) = Ω (( f [γle f t ])[L]k , g[γ]k ) =

= Ω ( f [L]k , g) = ΩL,le f t
As a consequence
Z Z Z
ΩM,le f t dµ = ΩM,le f t ◦ γ dµ = ΩL,le f t dµ
γ(D) D D

Because L is an upper triangular matrix the Fourier series of the cusp form f can
easily be evaluated at L(τ), τ ∈ H, while the automophy factor h(L, −) is a
constant and det L = p. An analogous estimate as in the proof of Theorem 5.23,
part ii) shows the boundedness of the integrand on D. Hence
Z Z
ΩM,le f t dµ = ΩL,le f t dµ < ∞
γ(D) D

iii) Left-right switching of [−]k : For given M ∈ V (p)

Z   Z
ΩM,le f t dµ = Γ (ΩM,le f t ) : Γ̃ (p) · ΩM,le f t dµ
F D(Γ (ΩM,le f t ))

because
Γ̃ (p) ⊂ Γ (ΩM,le f t )
has finite index according to Lemma 5.24, part 2). Concerning the indices
Lemma 5.24, part 3) implies

[Γ : Γ̃ (p)] = [Γ : M −1Γ̃ (p)M]

hence also
Γ (ΩM,le f t ) : Γ̃ (p) = Γ (ΩM,le f t ) : M −1Γ̃ (p)M
   
210 5 Introduction to Hecke theory and applications

Using the transformation formula from part i) and taking M −1 F as a fundamental

domain of M −1Γ̃ (p)M the calculation continues
  Z
Γ (ΩM,le f t ) : Γ̃ (p) · ΩM,le f t dµ =
D(Γ (ΩM,le f t ))

 Z
= Γ (ΩM,le f t ) : M −1Γ̃ (p)M ·

ΩM,le f t dµ =
D(Γ (ΩM,le f t ))
Z Z Z
ΩM,le f t dµ = ΩM0 ,right ◦ M dµ = ΩM0 ,right dµ
M −1 F M −1 F F
iv) Left-right switching of the Hecke operator:
Z Z
< Tp f , g >= (Tp f · g)(τ) · (Im τ)k dµ = Ω (Tp f , g) dµ
D D

For any pair of cusp forms Tp f and g the integrand is invariant under the action
of Γ as checked during the proof of Theorem 5.23, part i). Hence continuing
Z
1 Z
Ω (Tp f , g) dµ = · Ω (Tp f , g)dµ =
D [Γ : Γ̃ (p)] F

1 Z
1 Z
= · ∑ ΩM,le f t dµ = · ∑ ΩM,le f t dµ
[Γ : Γ̃ (p)] F M∈V (p) [Γ : Γ̃ (p)] M∈V (p) F

Using the result from part iii) and the property of V (p)] the calculation continues

1 Z
1 Z
· ∑ ΩM,le f t dµ = · ∑ ΩM0 ,right dµ =
[Γ : Γ̃ (p)] M∈V (p) F [Γ : Γ̃ (p)] M∈V (p) F

1 Z
1 Z
· ∑ ΩM0 ,right dµ = · Ω ( f , Tp g) dµ =
[Γ : Γ̃ (p)] F M∈V (p) [Γ : Γ̃ (p)] F
Z Z
= Ω ( f , Tp g) dµ = ( f · Tp g)(τ) · (Im τ)k dµ =< f , Tp g >, q.e.d.
D D

Theorem 5.26 (Eigenforms of the Hecke algebra of the modular group). For
each even weight k ≥ 0 the cusp space Sk (Γ ) has a basis of eigenforms of the Hecke
algebra Heckek (Γ ).
Proof. First, Theorem 5.18 implies that Heckek (Γ ) is commutative. Secondly, The-
orem 5.25 implies that all Hecke operators Tm , m ∈ N, are selfadjoint endomor-
phisms of the unitary vector space (Sk (Γ ), < −, − >). Hence the claim follows due
to a result from linear algebra about simultaneous diagonalizability of a family of
pairwise commuting, self-adjoint endomorphisms, q.e.d.
5.2 The Petersson scalar product 211

Remark 5.27 (Hecke theory of congruence subgroups). For each level N ∈ N and
even weight k ≥ 0 one can define Hecke operators on modular forms of Mk (Γ0 (N)).
The corresponding Hecke algebras are commutative, and each vector space of cusp
forms of fixed weight has an eigenbasis with respect to the family

Tm , m ∈ N and gcd(N, m) = 1.

For a proof see [34, Chap. 4, Theor. 4.22], [17], [33, Chap. IX, §6].

One should not consider a Hecke congruence subgroup Γ0 (N) as an isolated object
of a fixed level N. Instead one should focus on the whole family of Hecke
congruence subgroups
Γ0 (N), N ∈ N
For each weight k ∈ N and positive integer m ∈ N the modular spaces of different
levels
Mk (Γ0 (M)) and Mk (Γ0 (m · M))
are related.
Proposition 5.28 (Changing levels (N/m) 7→ N). Consider a weight k ∈ N, a
level M ∈ N and a positive integer m ∈ N. Set
 
m0
µm := ∈ Γm
0 1

The map
j : Mk (Γ0 (M)) →
− Mk (Γ0 (m · M)), f 7→ f [µm ]k ,
is a well-defined injection and restricts to an inclusion of cusp forms.
Under the assumptions of Proposition 5.28 one has apparently also the injection

→ Mk (Γ0 (m · N)).
Mk (Γ0 (N)) ,−

Proof. 1. Conjugation of Γ0 (M): Denote by

Γ 0 := µm−1 · Γ0 (M) · µm := {µm−1 · γ · µm : γ ∈ Γ0 (M)} ⊂ GL(2, Q)+

the µm -conjugate of Γ0 (M). If

 
ab
γ= ∈ Γ0 (M)
cd

then
     
1 1 0 m0 a b/m
µm−1 · γ · µm = · ·γ · = ∈ GL(2, Q)+
m 0m 0 1 cm d
212 5 Introduction to Hecke theory and applications

Claim: For N := m · M
Γ 0 ∩ Γ0 (M) = Γ0 (N).
The inclusion
Γ 0 ∩ Γ0 (M) ⊂ Γ0 (N)
is obvious because
M|c =⇒ m · M|c · m, i.e. N|c · m
To prove the opposite inclusion

Γ0 (N) ⊂ Γ 0 ∩ Γ0 (M) :

Each matrix    
a b a b
β= = ∈ Γ0 (N)
cN d cmM d
has the form  
a mb
β= µm−1 · γ · µm with γ := .
cM d
Here
det γ = det β = 1, hence γ ∈ Γ0 (M).
As a consequence

Γ0 (N) ⊂ Γ 0 and even Γ0 (N) ⊂ (Γ 0 ∩ Γ0 (M))

2. Changing right-multiplication to left-multiplication: Claim: For each α ∈ Γ

exists αle f t ∈ Γ such that
 
∗∗
µm · α = αle f t · ∈ M(2 × 2, Z) ∩ GL(2, Q)+ .
0∗

For the proof set

M := µm · α ∈ Γm
By Lemma 5.6 exists a matrix A ∈ Γ such that

A · M ∈ Γm

is triangular. Set
αle f t := A−1 ∈ Γ
3. Weak modularity: Assume f ∈ Mk (Γ0 (M)) and set

g := f [µm ]k .

Due to part 1) a given β ∈ Γ0 (N) has the form

β = µm−1 · γ · µm
5.2 The Petersson scalar product 213

with a suitable γ ∈ Γ0 (M). Then

g[β ]k = ( f [µm ]k )[β ]k = f [µm · β ]k = f [µm · µm−1 · γ · µm ]k =

= f [γ · µm ]k = ( f [γ]k )[µm ]k = f [µm ]k = g

Hence f [µm ]k is weakly modular with respect to Γ0 (N).

4. Holomorphy at the cusps: According to the definition one has to show: For each

g := f [µm ]k , f ∈ Mk (Γ0 (M)),

and each α ∈ Γ the function g[α]k is holomorphic at ∞ ∈ H∗ according to the

definition of Mk (Γ0 (N)). Due part 2) there exists a matrix

αle f t ∈ Γ

such that
µm · α = αle f t · θ
with a triangular matrix
 
ab
θ= ∈ M(2 × 2, Z) ∩ GL(2, Q)+
0d

Hence
g[α]k = f [µm · α]k = f [αle f t · θ ] = ( f [αle f t ]k )[θ ]k
From the Fourier expansion
∞
f [αle f t ]k = ∑ an · q̂n , q̂ = e2πi·τ/h
n=0

for a suitable h ∈ N one obtains as a consequence

∞
g[α]k (τ) = f [αle f t ]k [θ ]k (τ) = mk−1 ·d −k · ∑ an ·e2πin·(b/(dh)) · q̃n , q̃ = e2πi(τ/h)·(a/d) .
n=0

Hence g[α]k is holomorphic at ∞. And the last Fourier series shows that j maps
cusp forms to cusp forms.

If we fix the level N then each prime factor p|N defines a level changing
 
p0
(N/p) 7→ N, using µ p = ∈ Γp .
01

The subspace of Mk (Γ0 (N)) generated by the images of the maps

214 5 Introduction to Hecke theory and applications

j p,(N/p),k : Sk (Γ0 (N/p)) × Sk (Γ0 (N/p)) →

− Sk (Γ0 (N)), ( f , g) 7→ f + g[µ p ]k ,
for all prime factors p|N is named the oldspace, see Definition 5.29.

Definition 5.29 (Oldform and newform of cusp forms). Consider a level N ∈ N∗

and a weight k.
1. The image !
∑ im j p,(N/p),k ⊂ Sk (Γ0 (N))
p|N

is named the subspace

Skold (Γ0 (N)) ⊂ Sk (Γ0 (N))
of oldforms. The summation is indexed by the primes p dividing N.

2. The orthogonal complement of Skold (Γ0 (N) with respect to the Petersson scalar
product is named the subspace

Sknew (Γ0 (N)) := Skold (Γ0 (N)⊥ ⊂ Sk (Γ0 (N))

of newforms.

Apparently, by the same mappings one can also define the oldspace of modular
forms as a subspace fo Mk (Γ0 (N)). But unfortunately the Petersson scalar product
does not extend from the cusp forms to all modular forms. Hence there is no
orthogonal complement of subspace of old forms.

Example 5.30 (Oldforms and newforms). See PARI-file Congruence_subgroup_20.

• Congruence subgroup Γ0 (2): Figure 5.2 displays for the Hecke congruence sub-
group Γ0 (2) of level N = 2 the splitting of the cusp space for the first even
weights k = 0, ..., 12 into oldforms and newforms.

According to Definition 5.29 one has to consider for the splitting the single prime
factor p = 2 and the cusp forms Sk (Γ0 (1)), i.e. the cusp forms of the full modular
group Γ = Γ0 (1).

For weight k = 12 one has the 1-dimensional subspace

S12 (Γ ) = C · ∆ ⊂ S12 (Γ0 (2))

In addition also the modular form

∆2 : H →
− C, τ 7→ ∆ (2 · τ)

belongs to S12 (Γ0 (2)). The computation shows

5.2 The Petersson scalar product 215

S12 (Γ0 (2)) = spanC < ∆ , ∆2 >,

all cusp forms in S12 (Γ0 (2)) are old forms induced by S12 (Γ ).

• Congruence subgroup Γ0 (4): Similarly Figure 5.3 considers the modular group Γ0 (4)
of level N = 4. For the splitting one has to consider Sk (Γ0 (2)) and the level
change 2 7→ 4. As expected the oldforms of Γ0 (4) contain all cusp forms induced
from the cusp forms of Γ0 (2).

• Congruence subgroup Γ0 (11): The final Figure 5.4 considers the modular group Γ0 (11)
of level N = 11. For the splitting one has to consider the level change 1 7→ 11.
The old forms of Γ0 (11) derive from the modular forms ∆ and j(∆ ).

Fig. 5.2 Γ0 (2) oldforms and newforms

216 5 Introduction to Hecke theory and applications

Fig. 5.3 Γ0 (4) oldforms and newforms

Fig. 5.4 Γ0 (11) oldforms and newforms

5.3 Numerology: Lagrange, Jacobi, Ramanujan, Mordell 217

5.3 Numerology: Lagrange, Jacobi, Ramanujan, Mordell

The present section draws some conclusion from the existence of simultaneous
eigenforms of the Hecke operators. First we prove a conjecture of Ramanujan con-
cerning a certain numerical function. Secondly, we derive the four squares theorem.

Information about the Ramanujan τ-function is encoded by the modular discrim-

inant, the cusp form ∆ ∈ S12 (Γ ).

Definition 5.31 (Ramanujan τ-function). The Fourier coefficients (τn )n∈N of the
normalized discriminant form
∆ (τ) ∞
= ∑ τn · qn , q = e2πi·τ ,
(2π)12 n=1

define the Ramanujan τ-function

τ : N → Z, n 7→ τ(n) := τn .

According to Proposition 3.14 all Fourier coefficients are integers τ(n) ∈ Z.

Computing the first coefficients τ(n) up to n = 6 gives

∆ (τ)
= q − 24q2 + 252q3 − 1.472q4 + 4.830q5 − 6.048q6 + O(7)
(2π)12

From these values one verifies at once the product formula

τ(2) · τ(3) = −24 · 252 = −6.048 = τ(6),

and as a particular case of the recursion formula

τ(p2 ) = τ(p)2 − p11 · τ(p0 )

the formula for p = 2

τ(22 ) = −1.472 = τ(2)2 − 211 = (−24)2 − 2.048.

From explicit computation of the first 30 values of τ(n) Ramanujan [45]

conjectured in 1916 several generalizations. Figure 5.5 shows equations (103)
and (104) from the original paper with two of Ramanujan’s conjectures.

Note. There is a typo in equation (104) referring to the factor 2. It should read

|τ(p)| ≤ 2 · p11/2
218 5 Introduction to Hecke theory and applications

Fig. 5.5 Ramanujan’s conjecture on τ(n) from [45]

The conjecture was proved inter alia one year later by Mordell [42]. Figure 5.6
shows Mordell’s introduction from his paper.
5.3 Numerology: Lagrange, Jacobi, Ramanujan, Mordell 219

Fig. 5.6 Mordell’s introduction to [42]

A film about Ramanujan is to be recommended, named “The man who knew

infinity” (Deutsch: Die Poesie des Unendlichen).

Today Ramanujan’s product formula and recursion formula can be obtained as a

first application of Hecke theory:

Proposition 5.32 (Product and recursion formula of the Ramanujan τ-function).

The Ramanujan τ-function satisfies:
220 5 Introduction to Hecke theory and applications

1. Product formula: For all integers m, n ≥ 1

τ(m) · τ(n) = ∑ r11 · τ(mn/r2 ).

r|(m,n)

In particular, for coprime m and n:

τ(m) · τ(n) = τ(m · n).

2. Recursion formula: For all primes p and integers k ≥ 1

τ(p) · τ(pk ) = τ(pk+1 ) + p11 · τ(pk−1 ).

Proof. 1. Consider an arbitrary, but fixed m ∈ N. Corollary 5.13 from Hecke theory
derives the eigenvalue equation
!
∆ ∆
Tm 12
= τ(m) ·
(2π) (2π)12

When equating for each fixed n ∈ N the terms of order n on both sides of the
equation the coefficent formula from Theorem 5.11 implies:

∑ r11 · τ(mn/r2 ) = τ(m) · τ(n).

r|(m,n)

For coprime integers m, n we have (m, n) = 1 and the summation reduces to the
single term for r = 1:
τ(m) · τ(n) = τ(m · n)
2. We set m = pk and n = p in the product formula from part 1). Due to

(pk , p) = p

we now have two summands, referring to the indices r = 1 and r = p. We obtain

τ(pk ) · τ(p) = τ(pk+1 ) + p11 · τ(pk−1 ), q.e.d.

Remark 5.33 (Growth estimation of the Ramanujan τ-function). The final conjec-
ture of Ramanujan about the τ-function is the growth estimation

|τ(p)| ≤ 2 · p11/2 , p prime,

see Figure 5.5 equation (104) (after correction). This conjecture lies much deeper
than Ramanujan’s other conjectures. Deligne proved the growth condition of the τ-function
as a consequence of his proof of the Weil conjectures in 1974.
5.3 Numerology: Lagrange, Jacobi, Ramanujan, Mordell 221

The next application uses information encoded by modular forms from M2 (Γ0 (4)).
The “Four squares theorem” answers the question: How many possibilities exist to
represent a positive integer as the sum of four integer squares?

Fermat recalls in a letter to a friend from 1659 those results from arithmetic
which he considers his most important contribution to this field. One of these results
is the claim that any natural number can be written as the sum of four squares. The
first proof of this theorem has been given by Lagrange in 1770. The more refined
version of Theorem 5.38 is due to Jacobi in 1834. For the history of these issues
see [47].

Definition 5.34 (Representing integers as sum of squares). The number of possi-

bilities to represent a non-negative integer n as the sum of k integer squares defines
the arithmetic function
r : N∗ × N∗ → N
with n o
r(n, k) := card v = (v1 , ..., vk ) ∈ Zk : n = v21 + ... + v2k .

Note that r(n, k) counts k-tuples with positive and negative components as dif-
ferent, and distinguishes also k-tuples with the same components, but in different
order.

We will see that the representation of integers as sum of squares is related to the
congruence subgroup Γ0 (4), the relevant space of modular forms is M2 (Γ0 (4)).

Definition 5.35 (Generating function ϑ ). The series

∞
ϑ (τ, k) := ∑ r(n, k) · qn , q := e2πiτ ,
n=0

which is generated by the sequence r(n, k)n∈N , is named a ϑ -series.

Proposition 5.36 (Generating function as modular form). The generating function θ (−, 4)
is a modular form of weight 2 of the congruence subgroup Γ0 (4), i.e.

θ (−, 4) ∈ M2 (Γ0 (4)).

Its Fourier series starts

θ (τ, 4) = 1 + 8q + O(2), q = e2πiτ .

222 5 Introduction to Hecke theory and applications

Proof. Scetch, see cf. [17, Chap.1, §2; Chap. 4, §9], [36].

i) The congruence subgroup Γ0 (4) is generated as

Γ0 (4) =< T, γ >
with  
10
γ= ∈ Γ0 (4)
41
ii) The translation T apparently satisfies

ϑ (−, 4)[T ]2 = ϑ (−, 4)

The product formula

ϑ (τ, 4) = ϑ (τ)4
with
∞
ϑ (τ) = ∑ r(n, 1) · qn
n=0

reduces the weak modularity

ϑ (−, 4)[γ]2 = ϑ (−, 4)

to the transformation formula for ϑ : The difficult part is to prove

√
ϑ (−1/(4τ)) = −2iτ · ϑ (τ).

Then follows
! ! v ! !
u
τ −1 u 1 1
ϑ =ϑ = t2i +1 ·ϑ − −1 =
4τ + 1 4(−1/(4τ) − 1) 4τ 4τ
v ! ! v !
√
u u
u 1 1 u 1
= t2i +1 ·ϑ − = t2i + 1 · (−2iτ) · ϑ (τ) = 4τ + 1 · ϑ (τ),
4τ 4τ 4τ

and exponention
ϑ (−, 4) = ϑ 4
shows the transformation formula
!
τ
ϑ , 4 = (4τ + 1)2 · ϑ (τ, 4) i.e. ϑ (−, 4)[γ]2 = ϑ (−, 4)
4τ + 1

iv) The Fourier series

ϑ (τ) = 1 + r(1, 1) · q + O(q2 ) = 1 + 2 · q + O(q2 )

5.3 Numerology: Lagrange, Jacobi, Ramanujan, Mordell 223

shows
ϑ (τ, 4) = (1 + 2 · q)4 + O(2) = 1 + 8 · q + O(2), q.e.d.

Theorem 5.37 (The modular form G2,N ∈ M2 (Γ0 (N))). Denote by G2 the Eisen-
stein series from Remark 3.12. For each level N ≥ 1 set
 
N0
µN := ∈ M(2 × 2, Z) ∩ GL(2, Q)+
0 1

The function
G2,N := G2 − G2 [µN ]2 : H →
− C
is a modular function
G2,N ∈ M2 (Γ0 (N))

Proof. By definition

(G2 [µN ]2 )(τ) := G2 (µ2 (τ)) · det µN · h(µN , τ)−2 = N · G2 (Nτ)

i) Holomorphy: Due to Remark 3.12 the functions G2 and hence also G2,N are holo-
morphic in H ∪ {∞}.

ii) Weak modularity: We have to show for all α ∈ Γ0 (N):

G2,N [α]2 = G2,N .
We recall the formula from Remark 3.12
 
ab
G2 [γ]2 (τ) = G2 (τ) − 2πi · c · h(γ, τ)−2 for γ := α = ∈Γ
cd

We have to compute the transforms of the two summands of G2,N .

• G2 [α]2 : Set  
ab
γ := α = ∈ Γ0 (N) ⊂ Γ .
cd
Then
G2 [α]2 (τ) = G2 (τ) − 2πi · c · h(α, τ)−2
• (G2 [µN ]2 )[α]2 : Then
 
a Nb
µN · α = β · µN with β :=
c/N d

Here β ∈ Γ because
c ≡ 0 mod N =⇒ c/N ∈ Z
Then
224 5 Introduction to Hecke theory and applications

(G2 [β ]2 )(z) = G2 (z) − 2πi · (c/N) · h(β , z)−2

Hence with the argument z = µN (τ) = Nτ

(G2 [β ]2 )(µN (τ)) = G2 (µN (τ)) − 2πi · (c/N) · h(β , µN (τ))−2

Then

((G2 [µN ]2 )[α]2 )(τ) = (G2 [µN · α]2 )(τ) = (G2 [β · µN ]2 )(τ) =

= ((G2 [β ]2 )[µN ]2 )(τ) = (G2 [β ]2 )(µN (τ)) · h(µN , τ)−2 · N =

= G2 (µN (τ)) − 2πi · (c/N) · h(β , µN (τ))−2 · h(µN , τ)−2 · N =

= G2 (µN (τ)) · h(µN , τ)−1 · N − 2πi((c/N) · N) · (h(β , µN (τ)) · h(µN , τ))−2 =

= (G2 [µN ]2 )(τ) − 2πi((c/N) · N) · (h(β , µN (τ)) · h(µN , τ))−2
Now

h(β , µN (τ)) = h(β , Nτ) = (c/N) · Nτ + d = cτ + d and h(µN , τ) = 1

and
h(α, τ) = cτ + d
which implies
h(β , µN (τ)) = h(α, τ)
Hence
((G2 [µN ]2 )[α]2 )(τ) = (G2 [µN ]2 )(τ) − 2πi · c · h(α, τ)−2

As a consequence, the transformation formulas for both summands imply:

G2,N [α]2 = G2 [α]2 − ((G2 [µN ]2 )[α]2 ) = G2 − (G2 [µN ]2 ) = G2,N

iv) Finally one checks the Fourier expansion at the cusps using the Fourier expansion
of G2 from Remark 3.12, see [17, Chap. I, Exerc. 1.2.8.e] and [33, Chap. IX, §3, Example 4],
q.e.d.

Theorem 5.38 (Four squares theorem). For any n ∈ N, n ≥ 1, there are

r(n, 4) = 8 · ∑ d
d|n, 4-d

possibilities to represent the non-negative integer n as the sum of 4 integer squares.

In particular for 4 - n
r(n, 4) = 8 · σ1 (n) = 8 · ∑ d.
d|n

Proof. Computation within M2 (Γ0 (4)):

5.3 Numerology: Lagrange, Jacobi, Ramanujan, Mordell 225

i) Dimension: dim M2 (Γ0 (4)) = 2 according to Theorem 3.26, see also Figure 5.3.

ii) Basis: According to Theorem 5.37 for each N ≥ 2 the Eisenstein series

G2,N (τ) := G2 (τ) − N · G2 (Nτ)

is a modular form of the congruence subgroup Γ0 (N) of weight 2, i.e.

G2,N (τ) ∈ M2 (Γ0 (N)).

The Fourier expansion of G2 from Remark 3.12 provides for the two Eisenstein
series G2,2 and G2,4 the Fourier expansions
   
π2  ∞ π 2
G2,2 (τ) = − 1 + 24 · ∑  d  qn  = − (1 + 24q + O(2))
  
3 ∑ 3
n=1 d|n
d odd

and
! !
∞
2
G2,4 (τ) = −π 1+8· ∑ ∑ d q n
= −π 2 (1 + 8q + O(2)).
n=1 d|n, 4-n

Comparing terms of order 0 and 1 in the Fourier expansions shows that both
Eisenstein series are linearly independent. Hence the family (G2,2 , G2,4 ) is a basis
of M2 (Γ0 (4)). In particular, it generates θ (−, 4).

Proposition 5.36 implies by comparing Fourier coefficients of order 0 and 1

1
θ (−, 4) = − G2,4 .
π2
Equating on both sides the Fourier coefficients of order n ∈ N∗ shows

r(n, 4) = 8 · ∑ d, q.e.d.
d|n, 4-d

Example 5.39 (Modular forms via ϑ -functions). Figure 5.7 shows for some pairs (k, e)
the numbers
r(n, k, e).
Here r(n, k, e) ∈ N is the number of different representations of n as a sum of k
summands, each an e-power. These numbers are the Fourier coefficients of suitable
modular forms. The case
r(n, 4, 2) =: r(n, 4)
has been investigated in Theorem 5.38. The figure shows the output of the PARI-
script Modular_forms_theta_01.
226 5 Introduction to Hecke theory and applications

Fig. 5.7 Modular functions to compute r(−, k, e)

 Part II
Advanced Theory
Chapter 6
Application to imaginary quadratic fields

The modular j-invariant parametrizes the set of compex tori by the values of the
affine part of the modular curve X(Γ ) ' P. All complex analytic properties of a
given torus with normalized period lattice (1, τ) are encoded in the value j(τ) ∈ C
of the modular j-invariant. But j(τ) is too coarse to discriminate between the arith-
metic properties of the members of a class of biholomorphic equivalent tori.

Section 6.2 will investigate for tori with complex multiplication: Which arith-
metic information can be obtained when considering besides j(τ) also the values

j(α(τ)), α ∈ Γm,prim , m ∈ N∗ ?

References for the present chapter are [66, Chap. 6.1] and [50].

6.1 Imaginary quadratic fields and tori with complex

multiplication

As a first arithmetic property of complex tori and elliptic curves we observe: Com-
plex tori with normalized lattice generated by (1, τ) relate for certain numbers τ ∈ H
to an imaginary quadratic field. How to characterize tori with those arithmetic prop-
erties? The answer is given by Theorem 6.6.

As a general reference for number fields we recommend [22], [40], [32].

Definition 6.1 (Number field).

229
230 6 Application to imaginary quadratic fields

1. A field K is a number field if

Q ⊂ K ⊂ C with degree deg K := [K : Q] < ∞.

If deg K = 2 then K is a quadratic field.

2. The Galois group Gal(K/Q) of a number field K is the group of all field auto-
morphism of K which pointwise fix the elements of Q.

Remark 6.2 (Number field). Consider a number field K of degree

n = [K : Q]

1. Denote by
(α j ) j=1,...,n
a Q-basis of K. For each α ∈ K the multiplication

µα : K →
− K, x 7→ α · x,

is determined by the values

µα (α j ) = ∑ α jk · αk , j = 1, ..., n.
k=1,...,n

For α ∈ K norm and trace are respectively defined as

NK (α) := det(α jk ) ∈ Q and trK (α) := ∑ αjj ∈ Q

j=1,...,n

2. The discriminant of K with respect to the basis (α j ) j=1,...,n is defined as

∆K := det(trK (α j · αk )) ∈ Q

The discriminant depends on the choice of the basis: It transforms with the square
of the determinant of the base change matrix.

3. The ring of algebraic integers in K, the number ring of K, is the ring

OK := {x ∈ K : x integral over Z}

It is a lattice in K, and K is the quotient field of OK of integers. The ring OK is

the prototype of a Dedekind ring, i.e. OK is a Noetherian, 1-dimensional integral
domain, integrally closed in its quotient field.

4. An order of K is a subring Λ ⊂ OK which is a free Z-module of rank n.

6.1 Imaginary quadratic fields and tori with complex multiplication 231

5. The number field K has a Q-basis (α j ) j=1,...,n by algebraic integers α j ∈ OK .

Each element α ∈ K has a unique representation

Zα1 + ... + Zαn

α=
∆K
√
6. For a quadratic field K = Q( d) with a squarefree integer d ∈ Z one has

4 · d if d ≡ 2, 3 mod 4

∆K =

d if d ≡ 1 mod 4


and " √ #
∆K + ∆K
OK = Z
2
7. Each number field K has a primitive element, i.e.

K = Q(α)

with a suitable α ∈ K. Each primitive element satisfies a polynomial equation

P(α) = 0 with P(X) ∈ Q[X] and deg P = deg K.

8. Each quadratic number field has the form

√
K = Q( d) with d ∈ Z squarefree.

If d < 0 then K is named imaginary quadratic, if d > 0 then K is named real

quadratic.
√
9. The Galois group Gal(K/Q) of a quadratic number field K = Q( d) equals the
group Z2 . Its generator
σ1 : K →
− K, z 7→ σ1 (z),
satisfies (
−z d>0
σ1 (z) =
z d<0
Set
σ0 := idGal(K/Q) .
Then norm and trace of an element z ∈ K can be computed by using the elemen-
tary symmetric functions:

N(z) = σ0 (z) · σ1 (z) and Tr(z) = σ0 (z) + σ1 (z)

232 6 Application to imaginary quadratic fields

Note. Due to historical reasons the symbol O has two different meanings. In the
context of complex analysis O also denotes the complex structure sheaf.

Definition 6.3 (Endomorphisms of a torus). An endomorphism f of a torus

(T, 0) = C/Λ

is a holomorphic map
f : (T, 0) →
− (T, 0)
with f (0) = 0.

Recall from Proposition 2.3 that each endomorphism f of a torus fits into a
commutative diagram

µα
C C
p p
f
C/Λ C/Λ

with a unique multiplication map µα determined by an element α = α f ∈ C: For

all z ∈ C
µα (z) = α · z

Proposition 6.4 (Endomorphism ring). The set of endomorphisms of a torus

T = C/Λ

constitutes with respect to addition and composition a ring End(T ), the endomor-
phism ring of T . The map

End(T ) →
− {α ∈ C : α · Λ ⊂ Λ }, f 7→ α f ,

is an isomorphism of rings.

If not stated otherwise by the above ring isomorphism we identify the endomor-
phism ring End(T ) with a subring of C. Because each torus T is an Abelian group
its endomorphism ring End(T ) comprises at least the group Z:

Z ⊂ End(T ) ⊂ C.

Which intermediate rings are possible? The case

6.1 Imaginary quadratic fields and tori with complex multiplication 233

End(T ) = Z

is the general case, justifying a specific name for tori with additional endomor-
phisms.
Definition 6.5 (Complex multiplication (CM)).
1. A complex number τ ∈ H is a CM-point and its class [τ] ∈ X(Γ ) in the modular
curve is a CM-class, if the corresponding torus

T = C/Λ , Λ = Z ⊕ Z · τ,

satisfies
Z ( End(T )
2. The values j(τ) ⊂ C at CM-points τ ∈ H are named singular moduli or singular
values of the modular invariant j.

Theorem 6.6 states a remarkable relation between tori and number fields. It re-
veals the arithmetic structure encoded in a complex torus with complex multiplica-
tion.
Theorem 6.6 (Complex multiplication and imaginary quadratic fields). Con-
sider a point τ ∈ H.
1. Then: τ is a CM-point ⇐⇒ Q(τ) is an imaginary quadratic number field.

2. If τ is a CM-point with K := Q(τ), then the torus

T := C/Λτ , Λτ := Z + Z · τ,

has as ring of endomorphisms

End(T ) ⊂ C

an order of K.
Note that any order of an imaginary quadratic number field is a lattice.
Proof. 1. i) Consider a CM-point τ ∈ H. By definition exists α ∈ C \ Z with

α · Λτ ⊂ Λτ .

We obtain a matrix  
ab
γ= ∈ M(2 × 2, Z)
cd
with    
1 1
α· =γ· .
τ τ
234 6 Application to imaginary quadratic fields

We obtain
α = a + b · τ, α · τ = c + d · τ,
which implies b 6= 0, because α ∈
/ Z, and

b · τ 2 + (a − d) · τ − c = 0.

Because τ ∈ H is not a real number, the number field Q(τ) is imaginary quadratic.

ii) Assume that τ ∈ H belongs to an imaginary quadratic number field. Then for
suitable integers a, b, c ∈ Z with a 6= 0 because τ ∈ H
a · τ 2 + b · τ + c = 0.
From      
1 0 a 1
(aτ) · = ·
τ −c −b τ
follows aτ ∈ End(T ), ατ ∈
/ Z, and T has complex multiplication,

2. Assume that T has complex multiplication and consider an endomorphism f ∈ End(T )

operating as multiplication µα with a complex number α. Alike to the proof of
part 1, i) we obtain a matrix
 
ab
γ= ∈ M(2 × 2, Z)
cd

with    
1 1
α· =γ·
τ τ
Therefore α is an eigenvalue of γ and satisfies the eigenvalue equation, which is
an integral equation:

α 2 − (d + a) · α + ab − cd = 0.

As a consequence α ∈ OK , which implies

Z ( End(T ) ⊂ OK

Because OK is free of rank = 2 the same holds true for End(T ), and the
subring End(T ) is a sublattice of OK , i.e. an order, q.e.d.

We consider in detail the two elliptic points of the modular group Γ , see
Theorem 2.16:
Example 6.7 (CM-points).
1. Elliptic point i: The point i ∈ H determines the normalized lattice of Gaussian
integers
6.1 Imaginary quadratic fields and tori with complex multiplication 235

Λi = Z + Z · i
The torus T = C/Λi has the endomorphism ring

End(T ) = Λi

The endomorphism ring

End(T ) = Λi
equals the ring √
OK ⊂ K := Q(i) = Q( −1)
of algebraic integers from K. Recall from Corollary 3.21

j(i) = 1728 = 26 · 33 ∈ Z.

2. Elliptic point ρ = e2πi/3 : The point ρ ∈ H determines the normalized lattice

Λρ = Z + Z · ρ

The torus T = C/Λρ has the endomorphism ring

End(T ) = Λρ

The result is similar to the first example: The endomorphism ring

End(T ) = Λρ

equals the ring √

OK ⊂ K := Q(ρ) = Q( −3)
of algebraic integers from K. Recall from Corollary 3.21

j(ρ) = 0 ∈ Z.

Also the element 2ρ ∈ H belongs to K. The torus

T = C/Λ2ρ

with the normalized lattice

Λ2ρ = Z + Z · 2ρ

has as endomorphisms the multiplication with arbitrary elements

α ∈ Λ2ρ .

For the proof one uses the equation

ρ 2 + ρ + 1 = 0.
236 6 Application to imaginary quadratic fields

The endomorphism ring is

End(T ) = O
with
O = Λ2ρ ( OK
is a proper order of K. Note j(ρ) = 0 ∈ Z. A PARI-calculation shows

j(2ρ) = 54000 ∈ Z.

6.2 Modular polynomials

The present section takes up the study of imaginary quadratic fields K from the view-
point of modular forms. We relate complex multiplication to the value attainment of
the modular j-function: Theorem 6.14 proves that the value j(τ) ∈ C is an algebraic
integer for any CM-class [τ] ∈ X(Γ ). The main tool to derive the necessary proper-
ties of the modular j-invariant are the modular polynomials from Definition 6.8.

The modular polynomials are a family of polynomials (Fm (X))m∈N∗ which extract
the properties of the j-invariant j ∈ M (H) from its transformation behaviour with
respect to the operation of the modular group Γ

Γ × Γm,prim →
− Γm,prim

on the set of primitive matrices Γm,prim .

There is the following analogy to the family of Hecke operators (Tm )m∈N∗ : The
latter extract the Fourier coefficients of a given modular f ∈ Mk (Γ ) from the
transformation behaviour of f with respect to the operation of the modular group Γ

Γ × Γm →
− Γm

on the set of matrices Γm . The Hecke operators are the family of linear maps with

Tm ( f ) := ∑ f [M]k
M∈V (m)

defined as the weighted sum of over a complete system V (m) of representatives.

Similarly the modular polynomials are the family

Fm (X) = ∏ (X − j ◦ M) ∈ M (H)[X]
M∈V (m,prim)
6.2 Modular polynomials 237

defined as the product over a complete set V (m, prim) of representatives. Note the
similarity between the definition f [M]k for modular forms from Definition 3.4 and
the expression j ◦ M for the automorphic function j.

We take up the investigation of the integral matrices Γm from Section 5.1, in

particular Definition 5.4 of the set Γm,prim of primitive matrices and its complete
set V (m, prim) of representatives with respect to the canonical left Γ -operation.
The set V (m, prim) has cardinality

ψ(m) := m · ∏ (1 + (1/p))
p|m
p prime

Definition 6.8 (Modular polynomials). Consider m ∈ N∗ and the complete set V (m, prim)
of representatives of Γ \Γm,prim .
1. Each matrix M ∈ V (m, prim) induces the automorphism of the upper half-plane

M : H → H, τ 7→ Mµ (τ).

The ψ(m) functions

jM := j ◦ M ∈ V (m, prim)
are the class invariants of Γ \Γm .

2. The m-th modular polynomial is the polynomial

Fm (X) := ∏ (X − jM ) ∈ M (H)[X].
M∈V (m,prim)

Note. The modular invariant is constant along the orbits of the action of the
modular group Γ . Therefore the value of jM in Definition 6.8 does not depend
on representing a coset from Γ \Γm,prim by the matrix M ∈ V (m, prim) or by the
matrix γ · M, γ ∈ Γ .

Example 6.9 (Modular polynomials). The PARI-script modular_polynomial_01

computes for several m the modular polynomials Fm , see Figure 6.1.
238 6 Application to imaginary quadratic fields

Fig. 6.1 Modular polynomials

1. m = 2:
ψ(2) = 3
and
F2 (X, j) = X 3 +
+(− j2 + 1488 ∗ j − 162000) ∗ X 2 +
+(1488 ∗ j2 + 40773375 ∗ j + 8748000000) ∗ X+
+( j3 − 162000 ∗ j2 + 8748000000 ∗ j − 157464000000000)
2. m = 3:
ψ(3) = 4
and
F3 (X, j) = X 4 +
+(− j3 + 2232 ∗ j2 − 1069956 ∗ j + 36864000) ∗ X 3 +
+(2232∗ j3 +2587918086∗ j2 +8900222976000∗ j +452984832000000)∗X 2 +
+(−1069956∗ j3 +8900222976000∗ j2 −770845966336000000∗ j +1855425871872000000000)∗X+
+( j4 +36864000∗ j3 +452984832000000∗ j2 +1855425871872000000000∗ j)

The roots of the modular polynomial Fm (X) ∈ M (H)[X] are exactly the class
invariants jM ∈ V (m, prim) of the operation
Γ × Γm,prim →
− Γm,prim
We denote by sµ , j = 1, ..., ψ(m), the elementary symmetric functions in ψ(m)
variables. Then up to sign the functions

sµ ( j1 , ..., jψ(m) ) ∈ M (H), j = 1, ..., ψ(m),

are the coefficients of the modular polynomial Fm (X). E.g.,

6.2 Modular polynomials 239

(−1)ψ(m) sψ(m) ( j1 , ..., jψ(m) ) = (−1)ψ(m) · j1 · ... · jψ(m)

is the constant term and

(−1) · σ1 ( j1 , ..., jψ(m) ) = −( j1 + ... + jψ(m) )

is the coefficient of the term X ψ(m)−1 . Theorem 6.10 now takes a closer look on
these coefficients and proves: They are polynomials in the modular j-invariant with
integer coeffcients, i.e. the coefficients belong to the subring

Z[ j] ⊂ M (H).

See also Example 6.9.

Theorem 6.10 (Coefficients of the modular polynomials). For each m ∈ N∗

Fm (X) = ∏ (X − jM ) ∈ Z[ j][X] = Z[ j, X].

M∈V (m,prim)

Proof. For arbitrary but fixed index µ = 1, ..., ψ(m) we consider the coefficient

c := cµ := σµ ( j1 , ..., jψ(m) ) ∈ M (H)

of the modular polynomial

Fm (X) ∈ M (H)[X].

We prove step by step the following properties of the coefficient c ∈ M (H) until
arriving at c ∈ Z:
i) Weakly modular: According to Corollary 5.7 right multiplication with an
element γ ∈ Γ permutes the orbits of the Γ -left operation Φm
{[M1 ], ..., [Mψ(m) ]} = {[M1 · γ], ..., [Mψ(m) · γ]}.
Hence right multiplication does not change the value of the elementary symmetric
functions: For all j = 1, ..., ψ(m)

σµ ( j1 , ..., jψ(m) ) = σµ ( j1 ◦ γ, ..., jψ(m) ◦ γ).

Therefore c is weakly modular.

ii) Holomorphic on H: The j-invariant is holomorphic on H according to

Corollary 3.16. Therefore also the coefficient c is holomorphic.

iii) Meromorphic at ∞: The coefficient c is weakly modular according to part i).

Hence c has a Fourier expansion
240 6 Application to imaginary quadratic fields

c(τ) = ∑ aν · qν , q = e2πiτ .
ν∈Z

We have to show that the singularity q = 0 is a pole of c. For the proof we estimate
the growth of c: According to Corollary 3.21 the modular invariant has the Fourier
expansion
1
j(τ) = + P(q)
q
with P ∈ Z{q} a convergent power series with integer coefficients. For an upper
triangular matrix  
ab
α= ∈ V (m, prim)
0d
the Fourier series of the class invariant j ◦ α derives from the Fourier series of j by
substituting the variable q = e2πiτ by

e2πiα(τ) = e2πi(b/d) · q(a/d)

The equality
ad = m
implies
2
e2πiα(τ) = ζmab · (q1/m )a with ζm := e2πi/m
a primitive m-th root of unity. In particular

1
|( j ◦ α)(τ)| ≤ + O(1)
|q|a/d

for a single class invariant. Hence for suitable constants k, M

1
|c(τ)| ≤ M ·
|q|k

for small q ∈ ∆ \ {0}.

iv) Integrality c ∈ OK [ j]: Due to the part already proved, Corollary 3.22 implies
that c is a polynomial with complex coefficients in the modular invariant, i.e.
c ∈ C[ j].
We show that even
c ∈ OK [ j]
with
K := Q(ζm )
the m-th cyclotomic number field. For the proof we recall from part iii) the substi-
tution of q by
6.2 Modular polynomials 241
2
ζm ab · (q1/m )a .
As a consequence the class invariant j ◦ α has a Fourier expansion with respect
to q1/m , namely

1 
1/m a2 ab

j(α(τ)) = 2 ab
+ P (q ) · ζm ,
(q1/m )a · ζm

which shows that the Fourier coefficents of c belong to K. According to Corollary 3.22
the coefficients of the polynomial c ∈ C[ j] are Z-linear combinations of the Fourier
coefficients of c. Hence the coefficients are Z-linear combinations of the m-th roots
of unity. Therefore the coefficients of c ∈ C[ j] are algebraic integers from OK ,
i.e. c ∈ OK [ j].

v) Integrality c ∈ Z[ j]: The operation of the Galois group

Gal(K/Q) ' (Z/mZ)∗ .

extends to an operation on the set { j1 , ..., jψ (m)} of class invariants

Gal(K/Q) × { j1 , ..., jψ(m) } → { j1 , ..., jψ(m) }, (χ, jµ ) 7→ jµ χ ,

which is defined as follows: Assume χ ∈ Gal(K/Q) maps χ(ζm ) = ζm e with (e, m) = 1.

If jµ is the class invariant of the matrix
 
ab
M=
0d

then we define jµ χ as the class invariant of the matrix

 χ
ab
Mχ =
0 d

with
0 ≤ bχ < d and bχ ≡ e · b mod d.
Because the function c is fixed under the operation of Gal(K/Q), its Fourier co-
efficients belong to the fixed field Q. Due to part iv) the Fourier coefficients even
belong to
Q ∩ OK = Z, q.e.d.

We now evaluate the modular polynomials

Fm (X) ∈ Z[ j][X]

on the diagonal {X = j}, i.e. we consider the polynomials of the single variable j
242 6 Application to imaginary quadratic fields

Φm ( j) := Fm ( j, j) = ∏ ( j − jM ) ∈ Z[ j].
M∈V (m,prim)

Example 6.11 (The polynomials Φm ( j) ∈ Z[ j]). We take up Example 6.9 with Figure 6.2,
which shows some polynomials Φm . The PARI script Modular_polynomial_02
computes their factorization:
1. m = 2:

Φ2 ( j) = F2 ( j, j) = − j4 +2978∗ j3 +40449375∗ j2 +17496000000∗ j −157464000000000

splits into the factors

j − 8000, j − 1728, ( j + 3375)2

2. m = 3:

Φ3 ( j) = F3 ( j, j) = − j6 + 4464 ∗ j5 + 2585778176 ∗ j4 + 17800519680000 ∗ j3

−769939996672000000 ∗ j2 + 3710851743744000000000 ∗ j
splits into the factors

j − 54000, ( j − 8000)2 , j, ( j + 32768)2

3. The example Φ5 ( j) shows: In general Φm ( j), m prime, does not split into linear
factors over Z.

Fig. 6.2 Some polynomials Φm and their factorization

6.2 Modular polynomials 243

Lemma 6.12 (Leading coefficient of the modular polynomials). For squarefree m ≥ 2

the polynomial
Φm ( j) ∈ Z[ j]
has leading coefficient ±1.

Example 6.9 illustrates Lemma 6.12 by

Φ2 ( j) = − j4 + O(3), Φ3 ( j) = − j6 + O(5) and Φ7 ( j) = − j10 + O(9).

Proof. The function

Φm ( j) : H →
− C
is an automorphic function. Consider its Fourier expansion

ck ck−1
Φm ( j)(τ) = + + higher terms
qk qk−1
Because j has a pole at τ = ∞ of order = 1, the coefficient ck of the Fourier series is
also the leading coefficient of Φm as polynomial in j. The coefficient is the product
of the highest negative coefficients of the factors

j − jµ , µ = 1, ..., ψ(m).

For arbitrary but fixed µ the representation

1
j(τ) = + P(q)
q
implies !
aτ + b e−2πi(b/d) 
a/d 2πi(b/d)

jµ (τ) = j = + P q · e
d qa/d
with jµ corresponding to the matrix
 
ab
∈ V (m, prim).
0d

Hence the highest negative coefficient in the Fourier expansion of j − jµ is

(
1 a<d
e−2πi(b/d) a > d

Note that the case a = d is excluded because m = a · d is squarefree. In any case the
highest negative coefficent of
j − jµ
is a root of unity, and the same holds for the coefficient ck of the product
244 6 Application to imaginary quadratic fields

Φ= ∏ ( j − jM ).
M∈V (m,prim)

Because ck ∈ Z we have ck = ±1, q.e.d.

According to Theorem 6.10 the modular polynomial

ψ(m)
Fm ( j, X) := Fm (X) = ∏ (X − jµ ) ∈ Z[ j, X]
µ=1

is a polynomial in the two variables j and X with integer coefficients. It expands as

!
N
Fm ( j, X) = ∑ ∑ crs · jr · X s , crs ∈ Z.
n=0 r+s=n

Proposition 6.13 will show that Fm ( j, X) is symmetric in both variables, i.e. the
coefficients satisfy
crs = csr .
As a consequence, the highest exponent of j in Fm ( j, X) is ψ(m).

Proposition 6.13 (Symmetry of the modular polynomials). The modular polyno-

mials
Fm ( j, X) ∈ Z[ j, X], m ∈ N∗ ,
are symmetric with respect to j and X, i.e.

Fm ( j, X) = Fm (X, j).

Proof. We set R = Z[ j] with quotient field

K := Q(R) = Q( j).

i) Irreducibility of Fm (X) ∈ R[X] over K: The modular polynomial

Fm (X) = ∏ (X − jM ) ∈ R[X]
M∈V (m,prim)

has as roots the class invariants

jM ∈ V (m, prim) = { j1 , ..., jψ(m) }.

They are pairwise distinct which can be shown by comparing their Fourier coeffi-
cients. To prove that Fm (X) is an irreducible polynomial we show that the Galois
group
Gal(L/K)
6.2 Modular polynomials 245

of the splitting field

L = K( j1 , ..., jψ(m) )
of the polynomial Fm (X) ∈ R[X] acts transitively on the roots. Then no root belongs
to the fixed field K.

Elements of the Galois group can be obtained as follows: Any matrix γ ∈ Γ defines
the automorphism
L = K( j1 , ..., jψ(m) ) →
− L, f 7→ f ◦ γ.
Consider two class invariants

jµ , jν , 1 ≤ µ, ν ≤ ψ(m)

corresponding to the matrices

Mµ , Mν ∈ V (m, prim)

One has to find γ ∈ Γ such that

jν = j µ ◦ γ

i.e. one has to find γ, γ 0 ∈ Γ such that

γ 0 · Mν = Mµ · γ,

which is achieved by Proposition 5.9, part 2). Therefore Gal(L/K) operates transi-
tively on the roots of Fm (X), and the latter polynomial is irreducible over K.

ii) Common root of Fm ( j, X) and Fm (X, j): We consider the two polynomials

f (X) := Fm ( j, X) ∈ R[X] and g(X) := Fm (X, j) ∈ R[X]

For any matrix M ∈ V (m, prim) the class invariant

jM := j ◦ M

is a root of the modular polynomial f (X) ∈ R[X]. In particular for

 
1 0
α := ∈ Vm,prim
0m

holds for all τ ∈ H

!!
τ
0 = f ( jα (τ)) = Fm ( j(τ), jα (τ)) = Fm j(τ), j
m
246 6 Application to imaginary quadratic fields

Replacing the argument τ by τ · m shows for all τ ∈ H

0 = Fm ( j(τ · m), j(τ)) = Fm ( jβ (τ), j(τ)) = g( jβ (τ)) i.e. g( jβ ) = 0

with  
m0
β := ∈ Vm,prim
0 1
Also f ( jβ ) = 0 because β ∈ Vm,prim . Hence both polynomials f (X), g(X) ∈ R[X]
have the common root jβ .

iii) Symmetry: Due to Lemma 6.12 and part i) the polynomial f (X) ∈ K[X] is
irreducible with leading coefficient ±1. Hence f (X) is the minimal polynomial of
the field extension K ⊂ K[ jβ ], and g(X) is a multiple of f (X), i.e.

g(X) = h(X) · f (X)

with a polynomial h(X) ∈ K[X]. According to Theorem 6.10 both polynomials

f (X) and g(X) have their coefficients already in the ring R = Z[ j]. The ring R is
factorial as a ring of polynomials over the factorial ring Z. The Lemma of Gauss
implies that also the polynomial h[X] has all coefficients from R, and the equation

g(X) = h(X) · f (X)

holds in R[X]. Set

H( j, X) := h(X) ∈ R[X] = Z[ j][X].
Then
Fm (X, j) = H( j, X) · Fm ( j, X) ∈ Z[ j, X].
Considered as an equation in the ring Z[ j, X] of polynomials in two variables the
last equation is invariant when interchanging the role of j and X. Hence also
Fm ( j, X) = H(X, j) · Fm (X, j) ∈ Z[ j, X].
As a consequence

Fm ( j, X) = H(X, j) · H( j, X) · Fm ( j, X),

H(X, j) · H( j, X) = 1,
and
H(X, j) = H( j, X) = ±1.
In case H(X, j) = −1 the previous equation

Fm ( j, X) = H(X, j) · Fm (X, j)

would imply

Fm (X, j) = −Fm ( j, X) and Fm ( j, j) = −Fm ( j, j).

6.2 Modular polynomials 247

Then
Fm ( j, j) = 0
which contradicts the fact that Φm ( j) = Fm ( j, j) has leading coefficient ±1 accord-
ing to Lemma 6.12. Therefore
H(X, j) = 1
and
Fm ( j, X) = Fm (X, j), q.e.d.

We now derive from the theory of modular polynomials that all singular values
j(τ) are algebraic integers.

Theorem 6.14 (Singular values of the modular j-invariant). For each CM-point τ ∈ H
the j-value j(τ) ∈ C is an algebraic integer. More precisely: There exists m ∈ N∗
such that j(τ) is a root of the polynomial Φm (X) ∈ Z[X].

Proof. Consider a CM-point τ ∈ H. Theorem 6.6 implies that τ belongs to an imag-

inary quadratic field √
K = Q( d)
with d ∈ Z, d < 0 squarefree. The subring

OK ⊂ K

of algebraic integers of K is a free Z-module of rank = 2

OK = Z ⊕ Z · z

with a Z-basis (z, 1), z ∈ H. For admissible values of z see Remark 6.2.

i) A matrix α ∈ Γm,prim which fixes z: If a matrix α ∈ Γm,prim fixes a point τ ∈ H, i.e.

if
α(τ) = τ,
then
j(τ) = j(α(τ))
and vice versa. Hence the fixed points of the matrices α ∈ Γm,prim correspond to the
roots of the polynomial Φm (X) ∈ Z[X]. We determine a matrix α ∈ Γm,prim which
fixes z: The element
√
 d d < −1

ξ :=

1 + i d = −1


is an algebraic integer. Its norm

248 6 Application to imaginary quadratic fields

N(ξ ) = ξ ξ ∈ Z

is squarefree. Multiplication with ξ ∈ OK defines a Z-linear endomorphism

OK →
− OK , x 7→ ξ · x.

Denote by  
ab
α= ∈ M(2 × 2, Z) ∩ GL(2, R)+
cd
its matrix with respect to the basis (z, 1) of OK . Then
   
z z
ξ· =α·
1 1

or in components    
ξ ·z az + b
=
ξ ·1 cz + d
• The first form shows: The matrix

α ∈ M(2 × 2, Z)

which represents the Z-linear multiplication by ξ has the eigenvalue ξ and by

conjugation α = α also the eigenvalue ξ . Hence

m := det α = ξ ξ = N(ξ ) ∈ Z

is squarefree which implies

(a, b, c, d) = 1
and therefore
α ∈ Γm,prim
due to Definition 5.4.

• The component form shows: Considered as a fractional linear transformation

a·w+b
H→
− H, w 7→ α(w) :=
c·w+d
the matrix matrix α ∈ GL(2, R)+ satisfies

a·z+b ξ ·z
α(z) = = =z
c·z+d ξ ·1

ii) j(z) is an algebraic integer: Lemma 5.6 implies the existence of a matrix

αµ ∈ V (m, prim), µ ∈ {1, ..., ψ(m)},

6.2 Modular polynomials 249

with the same Γ -orbit as α with respect to the left action

Φm,prim : Γ × Γm,prim →
− Γm,prim , (A, M) 7→ A · M,

w.l.o.g. µ = 1. Then
j ◦ α = j ◦ α1
because the value of j is constant along each orbit of the Γ -operation. Due to part i)

j(z) = j(α(z)) = j(α1 (z)) = j1 (z),

which implies: The function

Φm ( j) = ∏ ( j − jM ) ∈ Z[ j]
M∈V (m,prim)

vanishes at z ∈ H:
Φm ( j)(z) = Φm ( j(z)) = 0.
The polynomial
Φm ( j) ∈ Z[ j]
has integer coefficients, and leading coefficient ±1 due to Lemma 6.12. Hence

0 = Φm ( j(z))

proves that j(z) is an algebraic integer. Hence the subring

Z[ j(z)] ⊂ C

is integral over the ring Z.

iii) j(τ) is integral over Z[ j(z)]: The given element

τ ∈ K = Q+Q·z
has a representation
az + b
τ = β (z) :=
d
with a primitive matrix
 
ab
β= ∈ Γm,prim , m := a · d.
0d

Alike to the argument from part ii) the Γ -orbit of β passes through a matrix αµ ∈ V (m, prim)
which proves
j(τ) = j(β (z)) = j(αµ (z)) = jµ (z).
Theorem 6.10 implies that jµ is integral over the ring Z[ j]. Hence
250 6 Application to imaginary quadratic fields

j(τ) = jµ (z)

is integral over Z[ j(z)].

iv) j(τ) is an algebraic integer: When combining part iii) and part ii) the
transitivity of integral dependence proves the integrality of j(τ) over Z, i.e. τ is an
algebraic integer, q.e.d.

Example 6.15 (Singular values of the modular j-function). Continuing Example 6.11
we consider a CM-class [τ] ∈ X(Γ ) and choose the level

m = 2.

1. According to Example 6.11 the polynomial

Φ2 ( j) ∈ Z[ j]

has degree = 4 and factorizes as

j − 8000, j − 1728, ( j + 3375)2

Which property is expressed by the three factors, why are the values j ∈ {8000, 1728, −3375}
related?

According to Theorem 6.14 we determine a complete set of representatives

α j ∈ Γ2,prim , j = 1, 2, 3,

with fixed points α j (z j ) = z j , z j ∈ H:

•   √
1 −2 1+i 7
α1 = has fixed point z1 =
1 0 2
which defines the lattice
√
1 + −7 √
Λ1 :=< 1, >= OK1 ⊂ K1 := Q( −7)
2
of the torus T1 := C/Λ1 .

•  
1 1
α2 = has fixed point z2 = i
−1 1
which defines the lattice
6.3 Fractional ideals and the class number formula 251
√
Λ2 :=< 1, i >= OK2 ⊂ K2 := Q( −1)

of the torus T2 := C/Λ2 .

•
√
 
0 −1
α3 = has fixed point z3 = i 2
2 0
which defines the lattice
√ √
Λ3 :=< 1, i 2 >= OK3 ⊂ K3 := Q( −2)

of the torus T3 := C/Λ3 .

Hence the three resulting fields K j , j = 1, 2, 3 are imaginary quadratic. Due to
Theorem 6.6 each of the three fields relates to a torus with complex multiplica-
tion. In addition Theorem 6.14 shows that the tori have their j-values from

{8000, 1728, −3375}

We already know j(z2 ) = j(i) = 1728. A PARI-computation completes at once

the remaining values

j(z1 ) = −3375 and j(z3 ) = 8000.

One knows that all three fields have class number hK = 1, which will confirm
Corollary 6.23.

6.3 Fractional ideals and the class number formula

A further application of the modular j-invariant to number theory is the class num-
ber formula for imaginary quadratic fields.

We first recall the divisor sequence of a Riemann surface, e.g., cf. [63].
Remark 6.16 (Class group of a compact Riemann surface). The divisor sequence
on a compact Riemann surface X

− O∗ →
1→ − M∗ →
− D→
− 0

provides the exact sequence of Abelian groups

1 → C∗ → M ∗ (X) → Div(X) → Cl(X) → 0.

because
252 6 Application to imaginary quadratic fields

Div(X)
H 0 (X, O ∗ ) = C∗ , H 0 (X, D) = Div(X), Cl(X) := and H 1 (X, M ∗ ) = 0
im[M ∗ (X) → Div(X)]

The vanishing H 1 (X, M ∗ ) = 0 is a result from the theory of compact Riemann

surfaces. It implies the isomorphy
'
→ H 1 (X, O ∗ ) = Pic(X)
Cl(X) −

between the group of divisor classes and the Picard group of isomorphism classes
of holomorphic line bundles on X.

For number fields an analogue of divisors are fractional ideals, see [4, Chap. 9].

Definition 6.17 (Fractional ideal). Consider a number field K of degree.

1. A fractional ideal of K is defined as an OK -submodule a ⊂ K for which there
exists an element d ∈ OK , d 6= 0, with

d · a ⊂ OK .

Hence a fractional ideal is a subset of the form

ã
a= ⊂K
d
with an ideal ã ⊂ OK . The element d is named a common denominator of a.

2. Fractional ideals generated by a single element α ∈ K

a = (α) := OK · α

are named principal fractional ideals.

3. For two fractionals ideals a, b ⊂ K the product

a·b

is the fractional ideal generated by all products a · b, a ∈ a, b ∈ b.

4. For a fractional ideal a 6= {0} the set

a−1 := {x ∈ K : x · a ⊂ OK }

is named the inverse of a.

Remark 6.18 (Fractional ideals). Consider a number field K with number ring OK .
6.3 Fractional ideals and the class number formula 253

1. Each non-zero ideal of OK and each non-zero fractional ideal of K is a lat-

tice of rank = deg K. The fractional ideals of K are the finitely generated OK -
submodules of K.

2. Because OK is a Dedekind ring for a fractional ideal

a 6= {0} ⊂ K

the inverse a−1 is also a fractional ideal, see [4, Theor. 9.8]. It satisfies

a · a−1 = (1)

3. With respect to multiplication the set of fractional ideals is an Abelian group J(K).
Its quotient by the subgroup of principal fractional ideals is named the ideal class
group or simply class group of K

Cl(K) := J(K)/im[K ∗ →
− J(K)]

The canonical map is

K∗ →
− J(K), x 7→ (x).
Hence two fractional ideals
a1 , a2 ∈ J(K)
define the same class in Cl(K) iff

a2 = a · a1 for a suitable a ∈ K ∗ .

As a consequence there is an exact sequence of multiplicative Abelian groups

− OK∗ →
1→ − K∗ →
− J(K) →
− Cl(K) →
− 1

4. A number ring OK is a principal ideal domain iff

hK = 1

A theorem from algebraic number theory states the finiteness of the class number

hK := ord Cl(K)

see [40, Chap. 5, Cor. 2], [32, Chap. 12, §2]. The result is an analogue to the fact
that the Neron-Severi group

NS(X) := Pic(X)/Pic0 (X)

of a compact Riemann surface is finitely generated. This similarity strongly val-

idates the analogy between the function fields of compact Riemann surfaces and
number fields.
254 6 Application to imaginary quadratic fields

5. Comparing the exact sequence of a number field with the exact divisor sequence
of a Riemann surface from Remark 6.16 one easily observes the analogy between
corresponding objects:

Compact Riemann surface X Number field K

OX∗ OK∗
MX ∗ K∗
Div(X) J(K)
Cl(X) Cl(K)

Example 6.19 (Class number).

1. There are exactly 9 imaginary quadratic fields
√
K = Q( −D), D > 0 squarefree integer,

with class number hK = 1. They belong to the values

D ∈ {1, 2, 3, 7, 11, 19, 43, 67, 163}.

2. The imaginary quadratic field

√
K = Q( −5)

has class number hK = 2. The class of the ideal

√
(2, 1 + −5) ⊂ OK

is not principal. It generates Cl(K), see [40, Chap. 5].

We now consider isomorphism classes of tori with complex multiplication and a

given number ring OK . How many isomorphism classes have OK as ring of endo-
morphisms? Proposition 6.20 shows that their number equals the class number hK .
In particular, there are only finitely many isomorphism classes of tori with endo-
morphism ring OK .
Proposition 6.20 (Class group and tori with given endomorphism ring). Con-
sider an arbitrary, but fixed imaginary quadratic field K with ring of integers OK .
There exists a bijection
'
→ {[T ] : T torus with End(T ) = OK }
Cl(K) −

between the ideal class group Cl(K) and the set of classes of biholomorphically
equivalent complex tori with ring of endomorphisms OK . The bijection is induced
by the attachement
6.3 Fractional ideals and the class number formula 255

normalized lattice Λ ∈ J(K) 7→ T = C/Λ with End(T ) = OK .

Its inverse is induced by the attachement

normalized torus T = C/Λ with End(T ) = OK 7→ Λ ∈ J(K).

Proof. i) Well-definedness of the map: Starting with a fractional ideal

Λ ⊂K⊂C

we may assume
Λ = Z+Z·ω
a normalized lattice because the ideal class is determined up to an element of K.
Due to Proposition 6.4 the torus

T := C/Λ

has the endomorphism ring

End(T ) = {α ∈ C : α · Λ ⊂ Λ }

Because Λ is an OK -module one obtains

OK ⊂ End(T ).

For the opposite inclusion consider a number α ∈ End(T ), i.e. α · Λ ⊂ Λ . The

normalization Z + Z · τ = Λ implies

α ·1 = α ∈ Λ.

ii) Well-definedness of the inverse map: Starting with a normalized torus

T = C/Λτ , Λτ = Z + Z · τ, with End(T ) = OK

we have to show that Λτ ⊂ K is a fractional ideal. After choosing a Z-basis (1, ω)

of OK the assumption

OK = End(T ) = {α ∈ C : α · Λτ ⊂ Λτ }

provides a matrix  
ab
∈ M(2 × 2, Z)
cd
with      
1 ab 1
ω· = · .
τ cd τ
Hence
ω = a + b · τ and ω · τ = c + d · τ
256 6 Application to imaginary quadratic fields

• The second equation implies

c
τ= ,
ω −α
hence Λτ ⊂ Q(OK ) = K.

• The lattice Λτ ⊂ K is an OK -module because both equations taken together imply

1, ω, τ, τ · ω ∈ Λτ

• Finally the first equation implies

b · Λτ ⊂ Z + Z · b · τ ⊂ Z + Z · ω = OK

As a consequence Λτ ⊂ K is a fractional ideal of K.

iii) The two constructions from part i) and ii) are inverse to each other, q.e.d.

Lemma 6.21 (Rigidness of orders). Consider an imaginary quadratic number

field K, an order Λ ⊂ OK and a subring R ⊂ K. If Λ ∼
= R as ring isomorphism
then Λ = R.

Proof. For the proof assume Λ = Z + Z · τ and consider a ring isomorphism

f :Λ →
− R

For suitable a, b ∈ Z holds

τ 2 = a · τ + b.
Then
f (τ)2 = f (a · τ + b) = a · f (τ) + b =⇒
=⇒ ( f (τ) − τ) · ( f (τ) + τ) = f (τ)2 − τ 2 = a · f (τ) + b − τ 2 =
= a · f (τ) + b − a · τ − b = a · ( f (τ) − τ)
Hence
f (τ) = τ or f (τ) = a − τ
In both cases

R = f (Z + Z · τ) = Z + Z · f (τ) = Z + Z · τ, q.e.d.

The class number hK of an imaginary quadratic number field K with number

ring OK relates to the singular value j(τ) of the j-invariant at an element τ ∈ H
with
6.3 Fractional ideals and the class number formula 257

OK = Z + Z · τ
The complex number j(τ) ∈ C is an algebraic integer due to Theorem 6.14.
Hence Q( j(τ)) is a number field. We show the estimate

[Q( j(τ)) : Q] ≤ hK .

Theorem 6.22 holds even in the strong form

[Q( j(τ)) : Q] = hK

See also [55, Chap. II, Theor. 4.3]. The book uses the notation

ELL(OK ) := {[T ] : T torus with End(T ) ∼

= OK }

with [T ] the biholomorphy class of the torus T .

Theorem 6.22 (Estimate of the class number). Consider a given imaginary

quadratic field K and denote by

OK = Z + Z · τ, τ ∈ H,

its number ring. Then the class number hK satisfies the estimate

[Q( j(τ)) : Q] ≤ hK .

Proof. i) Construction of a distinguished torus: Consider the given number field K

and its number ring
OK = Z + Z · τ, τ ∈ H.
Proposition 6.20 provides a normalized ideal Λ ⊂ J(K) such that the torus

T := C/Λ

has endomorphism ring

End(T ) = OK
Due to Theorem 4.3 the torus T embeds into P2 as a smooth cubic hypersurface,
i.e. as the set E = E(C) of complex points of a plane elliptic curve defined over C.

ii) The number field Q( j(τ)): Theorem 6.6 implies that the complex number τ ∈ H
from part i) is a CM-point, and Theorem 6.14 concludes that j(τ) ∈ C is an
algebraic integer. Define the number field

L := Q( j(τ))

and denote by
m := [L : Q]
258 6 Application to imaginary quadratic fields

its degree. There exist m field morphisms over Q

φi : L →
− C, i = 1, ..., m,

with pairwise distinct values φi ( j(τ)). Each extends to an element of the Galois
group Gal(C/Q). That’s made possible by extending elements from the Galois
group over Q along algebraic field extensions and along a pure transcendental field
extension: A given element

φ ∈ {φi : i = 1, ..., m}

can be considered an element

φ 1 ∈ Gal(Lsp /Q)

with Lsp ⊂ C the splitting field of L over Q. The element φ 1 extends to an element

φ 2 ∈ Gal(Q/Q)

along the algebraic extension [Q : Lsp ]. Choose a trancendence basis S of C over Q

and consider the field
K := Q(S),
and the pure transcendental extension [K : Q]. Then φ 2 extends to an element

φ 3 ∈ Gal(K/Q).

Finally φ 3 extends to an element

σ ∈ Gal(C/Q)

along the algebraic exension [C : K]. The resulting elements

σi ∈ Gal(C/Q), i = 1, ..., m,

have pairwise distinct values σi ( j(τ)).

iii) Functorial action of the Galois group: For the elliptic curve

E = Var( f ), f (X0 , X1 , X2 ) = ∑ aI · X I ∈ C[X0 , X1 , X2 ]

I

and an element σ ∈ Gal(C/Q) the curve

E σ := Var( f σ ), f σ (X0 , X1 , X2 ) := ∑ σ (aI ) · X I ,

I

defines again a plane elliptic curve. The attachment

E 7→ E σ
6.3 Fractional ideals and the class number formula 259

− E 0 defines a
from part ii) extends to a functor, i.e. each morphism Φ : E →
canonical morphism
− E0
σ
Φσ : Eσ →
with
id σ = id and (Φ ◦Ψ )σ = Φ σ ◦Ψ σ
For the proof represent a given holomorphic map

− E 0 , Φ(O) = O,
Φ :E →

by its induced multiplication map

F := µa : C →
− C, z 7→ a · z,

via the corresponding uniformization of the elliptic curves by complex tori C/Γ
and C/Γ 0 from Theorem 4.28.

F
C C z 6= 0 a · z, z 6= 0

Φ
E E0 (1 : ℘Γ (z) : ℘Γ0 (z)) (1 : ℘Γ 0 (a · z) : ℘Γ0 0 (a · z))

The diagrams induce a unique commutative completion

− E0
Φ :E →

because
a·Γ ⊂ Γ 0
and the functions ℘Γ 0 and its derivative have the period Γ 0 . The morphism Φ is
locally represented by a pair of rational maps
!
p1 p2
,
q1 q2

defined as quotient of polynomials. Eventually apply σ ∈ Gal(C/Q) to the

coefficients of these polynomials to obtain a representative of Φ σ .

As a consequence, for each σ ∈ Gal(C/Q) the map

− End(E σ ), Φ 7→ Φ σ ,
End(E) →

is a ring isomorphisms, i.e. End(E σ ) ∼

= End(E). Lemma 6.21 implies
260 6 Application to imaginary quadratic fields

End(E σ ) = End(E).

iv) The classes [E σ ]: The elliptic curve E with affine polynomial

F(X,Y ) = Y 2 − (4 · X 3 − g2 · X − g3 )

has the j-invariant

g32
j(E) = 1728 · = j(τ).
g32 − 27 · g23
Hence the elliptic curve E σ has the affine polynomial

F σ (X,Y ) = Y 2 − (4 · X 3 − σ (g2 ) · X − σ (g3 ))

and the j-invariant

σ (g2 )3
j(E σ ) = 1728 · = σ ( j(τ))
σ (g2 )3 − 27 · σ (g3 )2

The biholomorphy classes of elliptic curves are determined by the j-invariants of

the curves. The elements of the family

( j(E σi ) = σi ( j(τ))i=1,...,m

are pairwise distinct. Hence the elliptic curves

(E σi )i=1,...,m

with these j-invariants are pairwise not biholomorphic equivalent. For i = 1, ..., m
holds
End(E σi ) = OK
due to part iii). Proposition 6.20 implies that the corresponding ideal classes
in Cl(K) are pairwise distinct. Their number m is bounded by the class number

hK = card Cl(K).

Hence
m = [L : Q] ≤ hK , q.e.d.

Corollary 6.23 (Integer values of the modular invariant). Consider an imagi-

nary quadratic number field K with class number hK = 1 and number ring

OK = Z + Z · τ

Then j(τ) ∈ Z.
Proof. The proof follows from Theorem 6.14 and Theorem 6.22, q.e.d.
Chapter 7
Outlook: Modular elliptic curves, monstrous
moonshine

7.1 Modular elliptic curves

The relation between elliptic curves E/Q and modular forms is given by Wiles’
theorem, the proof of the Shimura-Taniyama-Weil conjecture for semistable elliptic
curves. Wiles’s modularity theorem has been subsequently extended to all elliptic
curves E/Q by Breuil, Conrad, Diamond and Taylor. The modularity theorem re-
lates for each elliptic curve E/Q the numbers

a p (E) := 1 + p − card E p (F p ), p prime,

to the eigenvalues of Hecke operators acting on a suitable cusp form as common

eigenform. The numbers a p (E) measure the deviation of the numbers of F p -valued
points of the reduction E(F p ) from the average value, see Remark 4.42.

A first version of the relation between rational elliptic curves and the theory of
modular forms is Conjecture 7.4.

Conjecture 7.1 (Shimura-Taniyama-Weil). For each elliptic curve E/Q exists a

level N ∈ N and a non-constant holomorphic map from a modular curve

X0 (N) →
− E(C)

Theorem 7.2 (Wiles’ modularity theorem). The Shimura-Taniyama-Weil conjec-

ture is true.

For the proof of the modularity theorem, Theorem 7.2, by Wiles in the stable and
semistable case see the symposion [15]. Concerning the remaining cases subse-
quently proved by several authors see the announcement [16] and the proof in [10].

261
262 7 Outlook: Modular elliptic curves, monstrous moonshine

When comparing rational elliptic curves E/Q and cusp forms f ∈ Sk (Γ0 (N)) one
compares the L-series of both mathematical objects. For the elliptic curve E/Q
recall the L-series L(E, −) arising from the numbers (a p (E)) p prime , see
Definitions 4.44 and 4.46, and also Remark 4.45.

For the cusp form f see Definition 7.3.

Definition 7.3 (L-series of a cusp form). Consider a level N ∈ N∗ and a

weight k ∈ N. For a cusp form f ∈ Sk (Γ0 (N)) with Fourier series
∞
f (τ) = ∑ cn · qn , q := e2πi·τ ,
n=1

one defines its L-series

∞ cn
L( f , −) : {s ∈ C : Re s > 1 + (k/2)} →
− C, L( f , s) := ∑ ns
n=1

Remark 7.4 (L-series of a cusp form).

1. The L-series L( f , −) from Definition 7.3 is well-defined. One obtains L( f , −)
from the Mellin transform of f , see [33, Chap. VIII, §5].

2. The L-series L( f , −) extends to a holomorphic function on C and satisfies the

functional equation with respect to the reflection at k

1 1
s
· Γ (s) · L( f , s) = (−1)k/2 · · Γ (k − s) · L( f , k − s)
(2π) (2π)k−s

see [33, Chap. VIII, Theor. 8.1]

The textbook [17] does not prove Theorem 7.2. But it gives an excellent
introduction to the problem and it highlights different views onto the theorem. In
joint work with his coworkers, one of the authors of [17] proved in [10] the final
step of the modularity theorem.

Theorem 7.5 (Versions of the modularity theorem).

1. Complex elliptic curve: Each elliptic curve E/C with rational modular invariant

1728 · g32
jE = ∈Q
g32 − 27 · g23
7.1 Modular elliptic curves 263

is uniformized by a modular curve, i.e. there exists an integer N ∈ N and a sur-

jective holomorphic map
X0 (N) →− E/C,
see [17, Theor. 2.5.1].

2. Rational elliptic curve: Each elliptic curve E/Q is uniformized by a modular

curve, i.e. there exists an integer N ∈ N and a surjective regular map over Q

X0 (N)Q →
− E/Q,

see [17, Theor. 7.7.2].

3. Jacobi torus: For each elliptic curve E/C with rational modular invariant

1728 · g32
jE = ∈Q
g32 − 27 · g23

exists an integer N ∈ N and a surjective holomorphic map

Jac(X0 (N)) →
− E,

from the Jacobi torus of a modular curve X0 (N), N ∈ N, which is a group homo-
morphism, see [17, Theor. 6.1.3].

4. Cusp form, version with coefficients: For each elliptic curve E/Q with conductor NE ∈ N
exists a new form
∞
f (q) = ∑ an ( f ) · qn ∈ S2 (Γ0 (NE ))
n=1

satisfying for all primes p ∈ N

a p ( f ) = a p (E),

see [17, Theor. 8.8.1].

Recall from Definition 4.40 for an elliptic curve E/Q with global minimal Weier-
strass polynomial F ∈ Z[X,Y ] the conductor

cond E = ∏ pe p ∈ Z,
p

having the same prime factors p as the discriminant ∆F and exponents e p ∈ N∗

deriving from the type of singularity of the reduction E(F p ).

5. Cusp form, version with L-series: For each elliptic curve E/Q with conductor NE ∈ N
exists a new form
∞
f (q) = ∑ an ( f ) · qn ∈ S2 (Γ0 (NE ))
n=1
264 7 Outlook: Modular elliptic curves, monstrous moonshine

with
L( f , −) = L(E, −),
see [17, Theor. 8.8.3]

Remark 7.6 (Forerunner of the Shimura-Taniyama-Weil Conjecture).

1. It is known that a non-constant holomorphic map

X0 (N) →
− E(C)

as required in Conjecture 7.1 would be defined by two automorphic functions,

see Definition 3.6,
u, v ∈ A0 (Γ0 (N))
which satisfy a Weierstrass equation of E/Q.

2. If the level N is minimal such that a non-constant holomorphic map

X0 (N) →
− E(C)

exists, then the map is induced by a new form f ∈ S2new (Γ0 (N)).

3. If the new form f ∈ S2new (Γ0 (N)) from the previous step has integer-valued
Fourier coefficients then
L( f , −) = L(E, −)
and N is the conductor of E.

4. The Fermat conjecture once stated that the Fermat equation

X n +Y n = Z n , n ∈ N, n ≥ 3,

has no non-trivial integer solutions.

For the particular exponents n = 3, 4 the Fermat conjecture was proved by Euler.
In addition it was known: The general case follows from the case of prime expo-
nents n = l ≥ 5. Before 1995 the conjecture was open for general primes l ≥ 5.

Assuming for an indirect proof of the Fermat conjecture the existence of a

non-trivial solution
(a, b, c) ∈ Z, gcd(a, b, c) = 1,
of the Fermat equation with prime exponent l ≥ 5 Frey proposed to study the
elliptic curve EFH /Q with Weierstrass equation
7.1 Modular elliptic curves 265

Y 2 = X · (X − al ) · (X − cl ),

today named the Frey-Hellegouarch curve. The relation between non-trivial so-
lutions of the Fermat equation and points on elliptic curves had already been
considered by Hellegouarch, see [30]. The curve EFH has discriminant

∆ = 24 · (abc)2l

and modular j-invariant

c34
j= with c4 = 24 · (a2l − (ac)l + c2l )
∆
The global minimal Weierstrass polynomial of EFH has conductor

N= ∏ p
p|abc

Serre and Ribet proved that the existence of EFH would contradict the validity of
the Shimura-Taniyama-Weil conjecture. Hence Wiles’ modularity theorem, The-
orem 7.2, implies that the Frey-Hellegouarche curve EFH does not exist. Hence
the Fermat conjecture is true.

A reference for all issues of this remark is [33, Chap. XII, § 1 and 4].

Example 7.7 (Modularity theorem). Figure 7.1 shows different elliptic curves E/Q
defined by a minimal Weierstrass polynomial, their conductor N, and a sample of
coefficients of their L-series and of the Fourier coefficients of the new-forms in

S2 (Γ0 (N)) = H 0 (X0 (N), Ω ),

see Corollary 3.28. The numerical result from the PARI script Elliptic_curve_taniyama_10
confirms the Modularity Theorem that a rational elliptic curves arises from a mod-
ular form.
266 7 Outlook: Modular elliptic curves, monstrous moonshine

Fig. 7.1 Relation between rational elliptic curves and new forms

7.2 Monstrous moonshine

What is “monstrous moonshine”? Monstrous moonshine is a synonym for the re-

lation between modular forms on one side and finite simple groups on the other
side.

That type of moonshine was detected in 1979, then formalized as a conjecture,

proved by Borcherds in 1992, and finally honored by a Fields Medal awarded to him
in 1998.

For an introduction see [8] and [59], on a more advanced level see [24] and [25].

Definition 7.8 (Simple group). A group G is simple if it has no normal, proper

subgroup H ⊂ G, i.e. satisfying

H 6= {e} and H 6= G

If a finite group G has a proper normal subgroup H ( G then one will try to in-
vestigate G by studying the smaller groups H and G/H. Proposition 7.9 generalizes
this construction. It formalizes in which sense simple groups are the building blocks
of all finite groups.
7.2 Monstrous moonshine 267

Proposition 7.9 (Jordan-Hölder theorem). Each finite group G has a finite, strictly
increasing sequence of subgroups, a composition series,

G0 = {e} ⊂ G1 ⊂ ... ⊂ Gn = G

such that for each i = 1, ..., n

Gi−1 ⊂ Gi
is a normal subgroup with simple quotient group

Qi := Gi /Gi−1

All composition series of G have the same lenght. Each two composition series of G
have - up to permutation - the same simple quotients counted with multiplicity.
For a proof see [38, Chap. IV, §4].

Example 7.10 (Composition series). The cyclic groups C12 has the following dif-
ferent composition series
{0} ⊂ C3 ⊂ C6 ⊂ C12
{0} ⊂ C2 ⊂ C6 ⊂ C12
{0} ⊂ C2 ⊂ C4 ⊂ C12
Each composition series of C12 has the same family of simple quotients (C2 ,C2 ,C3 )
up to permutation.

Definition 7.11 (Group representation). Consider a group G.

1. A group representation is a pair (V, ρ) with a vector space V and group morphism

ρ :G→
− GL(V )

to the group of linear automorphisms of V . As a shorthand one calls the vector

space a G-module using the notation

G ×V →
− V, (g, v) 7→ g · v := ρ(g)(v).

2. For a group representation

ρ :G→
−V
with finite dimensional V the trace of linear endomorphisms defines a map

trace ρ : G →
− C, χρ (g) := trace ρ(g),

which is named the trace of ρ

268 7 Outlook: Modular elliptic curves, monstrous moonshine

The representation V \ for monstrous moonshine is infinite-dimensional. But it

splits as the direct sum of finite-dimensional representations. For a finite-dimensional
representation the trace of an element g ∈ G only depends on the conjugacy class
of g0
[g0 ] := {g · g0 · g−1 : g ∈ G}
because invariance of the trace under cyclic permuation shows for g0 , g ∈ G

trace ρ(g · g0 · g−1 ) = trace(ρ(g) · ρ(g0 ) · ρ(g)−1 ) =

= (trace(ρ(g0 ) · ρ(g)−1 · ρ(g)) = trace ρ(g0 )

Apparently the trace of the neutral element equals the dimension of the representa-
tion
trace ρ(e) = dim V ∈ N.

Theorem 7.12 (Complete reducibility). Consider a finite group G.

1. Any G-module is completely reducible, i.e.
M
V= Vi
i∈I

with irreducible G-modules Vi , i ∈ I.

2. The irreducible G-modules correspond bijectively to the conjugacy classes of G.

Proof. See [51, §2, Theor. 7].

Remark 7.13 (Classification of finite simple groups).

1. Each finite simple group falls into exactly one of the following classes
• a set of 18 infinite sequences of groups

• a set of 26 groups, named the sporadics.

2. To the first class belongs the sequence (An )n≥5 of the group of alternating permu-
tations of n elements, the sequence C(p) p prime of cyclic groups with p elements,
and a series of Lie type groups like (PSL(n, F pk ))n .

3. The sporadics can be distinguished by their order. The next-to-biggest sporadic

is Fischer’s baby monster B with

ord B = 241 · 313 · 56 · 72 · 11 · 13 · 17 · 19 · 23 · 31 · 47 ∼ 4 · 1033

7.2 Monstrous moonshine 269

4. The largest sporadic is the monster M with

ord M = 246 · 320 · 59 · 76 · 112 · 133 · 17 · 19 · 23 · 29 · 31 · 41 · 47 · 59 · 71 ∼ 8 · 1053

The monster has 194 conjugacy classes and therefore exactly 194 irreducible
representations.

Example 7.14 (Monstrous Moonshine). Figure 7.2 shows

• on the right-hand side for the monster M the dimensions a(n) of the first irre-
ducible, complex M-modules Vn , n ≤ 18, ordered by dimension,

• and on the left-hand side the Fourier coefficients c(n) of the q-expansion of the
modular j-invariant.

Fig. 7.2 Monstrous Moonshine

What later was named “monstrous moonshine” started with the observations from
Figure 7.2
1 = 1 and 196.884 = 1 + 196.883
by McKay in 1978. The formula states
270 7 Outlook: Modular elliptic curves, monstrous moonshine

c(−1) = a(1) and c(1) = a(1) + a(2)

relating two coefficients of the q-expansion of the modular j-invariant to the dimen-
sions of the two lowest irreducible M-modules. As Borcherds remarks “moonshine
is not a poetic term referring to light from the moon. It means foolish or crazy idea
(Quatsch in German).”

Definition 7.15 (Some generalizations of the modular theory).

1. Moonshine group: A discrete subgroup G ⊂ SL(2, R) is a group of moonshine
type if for a suitable level N ∈ N∗

Γ0 (N) ⊂ G

and for a translation  

1b
γ= ∈ G ⇐⇒ b ∈ Z,
01
i.e. G contains all integer-valued translations and no other translations.

2. Hauptmodul: The orbit space of the canonical action of a group G of moonshine

type on the extended upper half plane H∗

φ : G × H∗ →
− H∗

is a compact Riemann surface X(G), the modular curve of G. For G with

genus X(G) = 0

the unique meromorphic function JG ∈ M (X(G)) with Fourier expansion nor-

malized as
1 ∞
JG (τ) = + ∑ an · qn , q = e2πi·τ ,
q n=1
is the Hauptmodul of G.

Moonshine groups G ⊂ SL(2, R) are congruent to the modular group Γ in the

following sense: The intersection

H := G ∩ Γ

satisfies the equality of indices

[G : H] = [Γ : H].
In particular, the index is finite because

[Γ : H] ≤ [Γ : Γ0 (N)] < ∞.
7.2 Monstrous moonshine 271

For a moonshine group G of genus X(G) = 0 the modular curve X(G) is P1 , and JG
is a distinguished generator of the field of meromorphic functions

M (X(G)) = C(JG ).

Example 7.16 (Hauptmodul).

• The Hauptmodul of the modular group Γ is the normalized modular j-invariant

1
JΓ (q) = j(q) − 744 = + 1960 884 · q + 210 4930 760 · q2 + 8640 2990 970 · q3 + O(4)
q

• The Hecke congruence subgroup Γ0 (2) has the Hauptmodul

1
JΓ0 (2) = + 276 · q − 20 048 · q2 + 110 202 · q3 + O(4)
q

• The normalizer of Γ0 (2) in SL(2, R), the moonshine group

Γ0 (2)+ ⊂ SL(2, R)

has the Hauptmodul

1
JΓ0 (2)+ = + 40 372 · q + 960 256 · q2 + O(3)
q

Remark 7.17 (Forerunner of the Monstrous Moonshine Theorem).

1. McKay-Thompson conjecture: McKay and Thompson conjectured the existence
of a graded, infinite-dimensional M-module

V\ = Vn\
M

n∈N

satisfying: The graded components Vn\ split as the direct sum of irreducible M-modules
with a certain multiplicity. The M-module V \ is characterized by the family of
series, later named McKay-Thompson series

T[g] (τ) = ∑ trace(g|Vn\ ) · qn , g ∈ M,

n∈N

see [60].

2. Conway-Norton conjecture: Each McKay-Thompson series T[g] is the Haupt-

modul of a moonshine group Γ[g] with modular curve of genus = 0.
272 7 Outlook: Modular elliptic curves, monstrous moonshine

Thompson decoded the data from Figure 7.2 as follows:

c(2) = 210 4930 760 = a(1) + a(2) + a(3)
and subsequently
c(3) = 2 · a(1) + 2 · a(2) + a(3) + a(4)
c(4) = 3 · a(1) + 3 · a(2) + a(3) + 2 · a(4) + a(5)
c(5) = 4 · a(1) + 5 · a(2) + 3 · a(3) + 2 · a(4) + a(5) + a(6) + a(7)
If the irreducible representations ρi , i ∈ N, of M are ordered acccording to increas-
ing dimension then the first summands of the representation from Remark 7.17 are

(V0 , ρ1 ), (V1 = {0}), (V2 , ρ1 ⊕ ρ1960 883 ), (V3 , ρ1 ⊕ ρ1960 883 ⊕ ρ210 2960 876 )

The series generated by the dimensions (dim Vn )n∈N has the Fourier expansion
∞
∑ (dim Vn+1 ) · qn =
n=−1

1
= J(τ) = + 1960 884 · q + 210 4930 760 · q2 + 210 4930 760 · q3 + O(4), q = e2πi·τ .
q

The open problem was to find explicitly a representation with the properties from
the Conway-Norton conjecture, see Remark 7.17. Borcherds proved in 1992
Theorem 7.18, see [6].

Theorem 7.18 (Monstrous Moonshine). The monster group M has a representa-

tion as automorphism group of the monster vertex algebra V \ . The representation
satisfies the properties of the Conway-Norton conjecture.
List of results and some outlooks

Part I. General Theory

Chapter 1. Elliptic functions
Focus: Analysis

Holomorphic elliptic functions are constant (Prop. 1.7)

Sum of residues vanishes for elliptic functions (Prop. 1.8)

Elliptic functions attain each value with equal multiplicity (Cor. 1.10)

Weierstrass function ℘ of a lattice (Theor. 1.12)

Differential equation of ℘ (Theor. 1.17)

Field of elliptic functions equals C(℘)[℘ 0 ] (Theor. 1.18)

Elliptic functions with prescribed zeros and poles (Theor. 1.21)

Chapter 2. The modular group Γ and its Hecke congruence

subgroups Γ0 (N)
Focus: Analysis, Topology

Holomorphic maps between complex tori (Prop. 2.3)

Moduli space of complex tori (Theor. 2.5)

273
274 List of results and some outlooks

Group action (Def. 2.6)

Modular group and Hecke congruence subgroups (Def. 2.9)

Fundamental domain of the Γ -action (Theor. 2.16)

Proper discontinuity of the Γ -action (Lem. 2.19)

Hausdorff topologyy of the orbit space of the Γ -action (Theor. 2.21, Cor. 2.25)

Analytic structure of the orbit space of the Γ -action (Prop. 2.27)

The modular curve as compact Riemann surface (Theor. 2.28)

The modular curves of the Hecke congruence subgroups (Rem. 2.30)

Chapter 3. The algebra of modular forms

Focus: Analysis

Γ -invariant differential forms (Prop. 3.1)

Modular forms and cusp forms (Def. 3.6)

Eisenstein series (Theor. 3.9)

Discriminant form ∆ (Def. 3.13)

Modular invariant j (Def. 3.15)

Modular curve and modular invariant j (Cor. 3.16)

Modular forms as meromorphic diff. forms on the modular curve (Theor. 3.17)

The field C( j) of automorphic functions (Cor. 3.22)

The algebra of modular forms and the ideal of cusp forms (Theor. 3.24)

The line bundle of modular forms (Theor. 3.26)

The codimension of cusp forms (Cor. 3.28)

Chapter 4. Elliptic curves

List of results and some outlooks 275

Focus: Analysis, algebraic geometry, arithmetic geometry

Embedding tori (Theor. 4.3, Rem. 4.14)

Globally generated line bundle (Def. 4.7)

Very-ampleness criterion (Theor. 4.13)

Elliptic curve over k (Def. 4.20)

Elliptic curves are plane cubics (Theor. 4.23)

Weierstrass polynomial (Def. 4.24)

Elliptic curves over C are tori (Theor. 4.28)

The group of classes of degree zero divisors of an elliptic curve (Theor. 4.31)

Abel’s theorem for elliptic curves (Theor. 4.33)

Reduction of an elliptic curve E/Q mod p (Def. 4.40)

The ζ -function of an elliptic curve E/F p (Def. 4.44)

The L-series of an elliptic curve E/Q (Def. 4.46)

Chapter 5. Introduction to Hecke theory and applications

Focus: Analysis, number theory

Hecke operators of the modular group (Def. 5.10)

Fourier coefficients of Hecke transforms (Theor. 5.11)

The discriminant ∆ as simultaneous eigenform (Cor. 5.13)

Commutativity of the Hecke algebra (Theor. 5.18)

Petersson scalar product (Theor. 5.23)

Self-adjointness of the Hecke operators of the modular group (Theor. 5.25)

Eigenforms of the Hecke algebra of the modular group (Theor. 5.26)

Oldforms and newforms (Def. 5.29)

276 List of results and some outlooks

Properties of the Ramanujan τ-function (Prop. 5.32)

The modular forms G2 N ∈ M2 (Γ0 (N)) (Theor. 5.37)

The 4-squares theorem (Theor. 5.38)

Chapter 6. Application to imaginary quadratic fields

Focus: Analysis, algebraic number theory

Complex multiplication (Def. 6.5)

Complex multiplication and imaginary quadratic fields (Theor. 6.6)

Modular polynomials of a given level (Def. 6.8)

Coefficients of the modular polynomials from Z[ j] (Theor. 6.10)

For CM-points τ ∈ H the singular values j(τ) ∈ C are algebraic integers (Theor.
6.14)

The ideal class group and tori with complex multiplication (Prop. 6.20)

The class number formula and the modular j-invariant (Theor. 6.22)

Chapter 7. Modular elliptic curves, monstrous moonshine

Focus: Analysis, algebraic geometry, group theory

Rational elliptic curves are modular (Theor. 7.2)

The Monstrous Moonshine theorem (Theor. 7.18)

References

References for these notes are grouped as

• Modular Forms: [9], [11] [15] [16] [19] [34] [36] [39] [42] [43] [44] [45] [50]
[52] [53] [65]

• Complex Analysis: [1] [5] [21] [27]

• Algebraic Geometry: [23] [29] [49] [54] [56]

• Number Theory: [13] [22] [26] [32] [40] [47]

• Online plattform: [57] (textbook level), [41] (research level)

The basic textbook is [17], for an elementary survey see [2].

1. Ahlfors, Lars: Complex Analysis. 2nd edition, McGraw-Hill, Tokyo (1966)

2. Alfes-Neumann, Claudia: Modulformen. Fundamentale Werkzeuge der Mathematik.
Springer, Wiesbaden (2020)
3. Alperin, Roger C.: PSL2 (Z) = Z2 ∗ Z3 . The American Mathematical Monthly. Vol. 100
(1993), 385-386
4. Atiyah, M. F.; Macdonald, I. G.: Introduction to Commutative Algebra. Addison Wesley
(1969)
5. Ben-Zvi, David, D.: Moduli Spaces. https://www.ma.utexas.edu/users/benzvi/math/pcm0178.pdf
6. Borcherds, Richard E.: Monstrous Moonshine and Monstrous Lie Superalgebras. Invent.
Math. 109, (1992), p. 405-444
7. Borcherds, Richard E.: Proceedings of the I.C.M., Vol. I (Berlin, 1998). Doc. Math. 1998,
Extra Vol. I, 607-615,
http://math.berkeley.edu/˜reb/papers/icm98/icm98.pdf Call 8.7.2020
8. Borcherds, Richard E.: What is ... The Monster? Notices of the AMS, 49,7 (2002), p. 1076-
1077
9. Borel, A.; Chowla, S.; Herz, C.; Iwasawa, K.; Serre, J-P.: Seminar on Complex Multiplication.
Lecture Notes in Mathematics 21. Springer, Berlin (1966)
10. Breuil, Christophe; Conrad, Brian; Diamond, Fred; Taylor, Richard: On the Modularity of
Elliptic Curves over Q: Wild 3-adic exercises. Journal of the American Mathematical Society.
14(4), p. 843-939 (2001)

277
278 References

11. Chenevier, Gaëtan: Introduction aux Formes Modulaires. Internet (2015)

12. Cheng, Miranda: Moonshine-I.
https://www.youtube.com/watch?v=HB5dCXQIWqM Call 8.7.2020
13. Cohen, Henri: A Course in Computational Algebraic Number Theory. Springer, Berlin (1993)
14. Conway, J.H.; Norton, S.P.: Monstrous Moonshine. Bull. London. Math. Scoc 11, (1979),
308-339
15. Cornell, Gary; Silvermann, Joseph H.; Stevens, Glenn (Eds.): Modular Forms and Fermat’s
Last Theorem. Springer, New York (1997)
16. Darmon, Henri: A proof of the full Shimura-Taniyama-Weil conjecture is announced. Notices
Americ. Math. Soc. 46,11 (1999), p. 1397-1401
17. Diamond, Fred; Shurman, Jerry: A First Course in Modular Forms. Springer, New York
(2005)
18. Dugundji, James: Topology. Allyn and Bacon, Boston (1966)
19. Eichler, Martin; Zagier, Don: On the Zeros of the Weierstrass ℘-Function. Mathematische
Annalen 258, (1982), p. 399-407
20. Fischer, Wolfgang; Lieb, Ingo: Funktionentheorie. Vieweg, Braunschweig (1980)
21. Forster, Otto: Riemannsche Flächen. Springer, Berlin (1977)
22. Forster, Otto: Algorithmische Zahlentheorie. 2. Auflage Springer Spektrum (2015)
23. Fulton, William: Algebraic Curves. An Introduction to Algebraic Geometry. The Ben-
jamin/Cummings Publishing Company (1969)
24. Gannon, Terry: Monstrous Moonshine: The first twenty-five years. Bull. London Math. Soc.
38, p. 1-33 (2006)
25. Gannon, Terry: Moonshine beyond the Monster. The bridge connecting Algebra, Modular
Forms and Physics, Cambridge University Press, Cambridge (2006)
26. Gross, Benedict: The arithmetic of elliptic curves - An update. Arabian Journal of Science
and Engineering 1 (2009), p. 95-103
http://www.math.harvard.edu/˜gross/preprints/ell2.pdf
27. Gunning, Robert C.: Lectures on Riemann Surfaces. Princeton University Press (1966)
28. Han, Juncheol: The general linear group over a ring. Bull. Korean Math. Soc. 43 (2006), No.
3, pp. 619-626
29. Hartshorne, Robin: Algebraic Geometry. Springer, New York (1977)
30. Y. Hellegouarch: Points d’order 2ph sur les courbes elliptiques. Acta Arithmetica 26 (1972),
253-263
31. Huybrechts, Daniel: The geometry of cubic hypersurfaces. (2020)
http://www.math.uni-bonn.de/people/huybrech/Notes.pdf Call 15.9.2020
32. Ireland, Kenneth; Rosen, Michael: A Classical Introduction to Modern Number Theory.
Springer, New York (1982)
33. Knapp, Anthony: Elliptic Curves. Princeton University Press. Princeton (1992)
34. Kilford, Lloyd James: Modular forms: a classical and computational introduction. Imperial
College Press, 2nd ed. (2015)
35. Koblitz, Neal: Introduction to Elliptic Curves and Modular Forms. Springer, New York, 2nd
ed. (1993)
36. Koecher, Max; Krieg, Aloys: Elliptische Funktionen und Modulformen. 2. Aufl. Springer,
Berlin (2007)
37. Lang, Serge: Algebra. Addison-Wesley, Reading, 7th ed. (1977)
38. Lang, Serge: Elliptic Functions. 2nd edition Springer, Berlin (1987)
39. Lang, Serge: Introduction to Modular Forms. Springer, Berlin (1995)
40. Marcus, Daniel A.: Number Fields. Springer, New York (1977)
41. mathoverflow. https://mathoverflow.net/
42. Mordell, Louis Joel: On Mr. Ramanujan’s empirical expansion of modular functions. Proc.
Cambridge Philosophical Society, 19 (1917), p. 117-124
43. Ono, Ken: The last words of a genius. Notices Americ. Math. Soc. 57, 11 (2010), p.1410-1419
44. Pantchichkine, Alexeı̈: Formes Modulaire et Courbes Elliptiques. https://www-fourier.ujf-
grenoble.fr/ panchish/06ensl.pdf
References 279

45. Ramanujan, Srinivasa: On certain arithmetical functions. Trans. Cambridge Philosophical So-
ciety. 22(9) (1916), p. 159-184
46. Raynaud, Michèle: Géometrie Algébrique et Géometrie Analytique. Séminaire de Géometrie
Algébrique du Bois Marie 1960-61, Exposé XII. https://arxiv.org/pdf/math/
0206203.pdf Call June 2020
47. Scharlau, Winfried; Opolka Hans: Von Fermat bis Minkowski. Eine Vorlesung über Zahlen-
theorie. Springer, Berlin (1980)
48. Schultz, Dan: Notes on Modular Forms.
https://faculty.math.illinois.edu/\%7eschult25/ModFormNotes.
pdf Call 27.10.2020
49. Serre, Jean-Pierre: Géométrie Algébrique et Géométrie Analytique. Annales de l’Institut
Fourier. IV (1955-56), p. 1-42
50. Serre, Jean-Pierre: II Modular Forms. (1957/58). In: [9].
51. Serre, Jean-Pierre: Représentations linéaires des groupes finis. Herman, Paris (1967)
52. Serre, Jean-Pierre: A Course in Arithmetic. Springer, New York (1973)
53. Shimura, Goto: Introduction to the Arithmetic Theory of Automorphic Functions. Iwanami
Shoten and Princeton University Press, o.O. (1971)
54. Silverman, Joseph: The Arithmetic of Elliptic Curves. Springer, New York (1986)
55. Silverman, Joseph: Advanced Topics in the Arithmetic of Elliptic Curves. Springer, New York
(1994)
56. Silverman, Joseph; Tate, John: Rational points on Elliptic Curves. Springer, New York (1992)
57. stackexchange. https://math.stackexchange.com/
58. Szpiro, Lucien: Séminaire sur les pinceaux arithmétiques: La conjecture de Mordell.
Astérisque 127. Société Mathématiques de France, Paris (1985)
59. Tatitscheff, Valdo: A short introduction to Monstrous Moonshine. Preprint (2019)
https://www.researchgate.net/publication/331008545_A_short_
introduction_to_Monstrous_Moonshine
60. Thompson, J. G.: Some Numerology between the Fischer-Griess monster and the Elliptic
Modular Function. Bull. Lond. Math. Soc. 11 (1979), 352-353
61. van der Geer, Gerard: Siegel Modular Forms. arXiv:math/0605346 [math.AG]
62. Wehler, Joachim: Complex Analysis. Lecture Notes (2019) http://www.mathematik.
uni-muenchen.de/\%7ewehler/20181218_Funktionentheorie_Script.
pdf
63. Wehler, Joachim: Riemann Surfaces. Lecture Notes (2020) http://www.mathematik.
uni-muenchen.de/\%7ewehler/20190530_RiemannSurfacesScript.pdf
64. Zagier, Don: Zetafunktionen und quadratische Körper. Eine Einführung in die höhere Zahlen-
theorie. Springer, Berlin (1981)
65. Zagier, Don: https://www.youtube.com/watch?v=zKt5L0ggZ3o (2015)
66. Zagier, Don: Elliptic Modular Forms and Their Applications.
https://people.mpim-bonn.mpg.de/zagier/files/doi/10.1007/
978-3-540-74119-0_1/fulltext.pdf Call 16.7.2020
Index

K-valued point, 144, 145 multiplicativeness, 182

L-series, 178 normalized, 96
ϑ -function, 221, 225 relation to lattice constants, 98
℘-function, 15 elementary divisor theorem, 189
ζ -function of an elliptic curve, 177 elliptic curve
ζ -function, 177
Abel’s theorem, 26, 163 L-series, 178
automorphic form, 88 class group, 161
definition, 148
class group divisor, 149
compact Riemann surface, 251 group structure, 161, 163, 164
elliptic curve, 161 plane curve, 150
class number, 254, 257 uniformization, 155
complete reducibility, 268 elliptic function
complex analysis - algebraic geometry, definition, 11
dictionary, 139 residue theorem, 13
composition series, 267 elliptic function field, 22
Conway-Norton conjecture, 271 elliptic point, 38
cusp, 38 embedding theorem for compact Riemann
cusp form surfaces, 138
codimension, 123
definition, 88 factor of automorphy, 85, 86
dimension formula, 110 Falting’s theorem, 165
Fermat conjecture, 264
degree four squares theorem, 224
plane curve, 144 Fourier coefficients
discrete subgroup, 10, 11 definition, 85
discriminant form of Hecke transforms, 190
definition, 99 Fourier expansion, 85
eigenform of Hecke operators, 193 fractional ideal, 252
Fourier coefficients, 100 Frey-Helleguarch curve, 264
divisor, 147 fundamental domain, 36, 47
divisor lemma, 41
doubly periodic, 12 GAGA, 145
Galois group, 229
Eisenstein series group action, 36
definition, 94 group of moonshine type, 270

281
282 Index

group representation, 267 base-point, 132

line bundle
Hasse estimate, 174 base-point free, 132
Hauptmodul, 271 globally generated, 132
Hecke algebra very ample, 136
commutativity, 199 very ampleness criterion, 137
definition, 198 local ring, 141
Hecke congruence subgroup
changing levels, 211 McKay-Thompson conjecture, 271
definition, 38 meromorphic at ∞, 85
index, 43 modualr form
Hecke operators newform, 262
definition, 189 modular curve, 71, 76
eigenforms, 193, 210 modular form
Fourier coefficients, 190, 193 G2 (N), 223
self-adjointness, 208 algebra of modular forms, 116
holomorphic at ∞, 85 as differential form on the modular curve,
holomorphic map 104
defined by a line bundle, 133 definition, 88
embedding into projective space, 134 dimension formula, 110
map to Pn , 133 newform, 214
hyperbolic volume form oldform, 214
definition, 200 section of a line bundle, 120
invariance, 200 weight formula, 111
modular function, 89
imaginary quadratic field modular group
definition, 230 action of isotropy groups, 63
estimate of class number, 257 definition, 38
integral matrices fundamental domain, 47
orbit set of Γ -action, 184 orbit space, 65
complete set of representatives, 187 presentation, 51
definition Γm , 183 modular invariant
definition Γm,prim , 183 and modular curve, 102
invariant differential forms, 82 definition, 102
isotropy group, 36 generates field of automorphic functions,
113
Jacobi torus, 262 singular value, 233, 247, 250, 260
Jordan-Hölder theorem, 267 modular polynomial
coefficients, 239
Kronecker symbol, 78 definition, 237
symmetry, 244
L-series modularity theorem, 261, 262, 265
holomorphic, 262 moduli space, 65
of a cusp form, 262 monstrous moonshine, 269
lattice monstrous moonshine theorem, 272
basis, 31 Mordell’s theorem, 165
definition, 11 multiplicative function, 181
lattice constant, 14
normalized, 34 number field
period lattice, 11 algebraic integer, 230
positively oriented basis, 31 class group, 254
similar, 32 class number, 253
Legendre symbol, 78 definition, 229
line bundel discriminant, 230
Index 283

fractional ideal, 252 definition, 217

ideal class group, 252 growth estimate, 220
imaginary quadratic, 230 product and recursion formulas, 219
norm, 230 regular function, 141
order, 230 regular local ring, 141
trace, 230 representing integers as sum of squares, 221
number ring Riemann ζ -function, 94, 95
definition, 230
Shimura-Taniyama-Weil conjecture, 261
orbit space, 36 simple finite groups
classification, 268
period simple group, 266
definition, 9 smooth point, 141
period group, 10
period parallelogram, 11
torus
Petersson scalar product
biholomorphic equivalent, 33, 34
definition, 202
complex multiplication, 233
properties, 203, 204
embedding as elliptic curve, 126
plane projective algebraic curve, 141
endomorphism, 232
positively oriented basis, 31
endomorphism ring, 232
projective algebraic curve, 141
holomorphic map, 32
projective algebraic variety, 141
ring of endomorphisms, 232
projective space
higher dimensional, 125
sections of O(1), 132 Weierstrass ℘-function
standard atlas, 126 definition, 15
twisted line bundle, 132 differential equation, 21
proper group action, 60 Laurent expansion, 20
properly discontinuous, 54 Weierstrass equation, 153
Weierstrass polynomial
q-expansion, 85 definition, 153
quadric, 196 minimal, 169

Ramanujan τ-function Zariski topology, 141