LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE

LANGLANDS PROGRAM

JOHANNES ANSCHÜTZ

Contents
1. Remarks on the lecture 2
2. A general introduction to the Langlands program 3
3. From modular forms to automorphic representations, part I 11
4. From modular forms to automorphic representations, part II 24
5. From modular forms to automorphic representations, part III 33
6. Langlands reciprocity for newforms 46
7. Modular curves as moduli of elliptic curves (by Ben Heuer) 55
8. The Eichler–Shimura relation (by Ben Heuer) 62
9. Galois representations associated to newforms 70
10. Galois representations for weight 1 forms (by Ben Heuer) 79
11. The Langlands program for general groups, part I 88
12. The Langlands program for general groups, part II 94
References 102

1
2 JOHANNES ANSCHÜTZ

1. Remarks on the lecture

These are (slightly revised) notes for an introductory course on the Langlands
program.1
The eventual aim of the course was to present (mostly without any indication of
proof) statements/objects/conjectures... occuring in the Langlands program for a
general reductive group over Q, and explain how they specialize to the case of GL2 .
Another aim was to present the construction of Galois representations associated
to modular forms (with an emphasis on the case of weight 1 and 2).
Originally it was planned to include more material on the local Langlands pro-
gram. But this was postponed as for the winter term 2020/2021 a lecture on this
at the university Bonn was announced by Peter Scholze.
To make the notes more readable I included some statements, which in the lecture
I said orally.
Disclaimer: I am by far not a person with serious knowledge/understanding of
the Langlands program, thus in the notes I may oversimplify/overcomplicate things,
be inaccurate, or even wrong, and miss subtelties. Thus use these notes at your
own risk and consult the mentioned references for definite/correct statements. Any
comments/hints on the notes are welcome!
I want to thank the participants of the course for the interested questions dur-
ing the lectures. Moreover, I want to thank Ben Heuer heartily for following my
suggestion to give three lectures (which ones is indicated in the respective title).

1As the lecture was planned to run parallel to the eventually cancelled trimester program of
the same name, the original title was “The arithmetic of the Langlands program”. But this title
was not accurate in describing what was presented in the lecture. Therefore I changed the title of
these revised notes.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM 3

2. A general introduction to the Langlands program

In this lecture we want to give a rough introduction to the Langlands program,
and present a first example of its arithmetic significance.
Let us recall that a connected, smooth, affine, algebraic group G over a field k is
reductive if the base change Gk to an algebraic closure of k contains no closed normal
subgroup isomorphic to the additive group Ga , or equivalently, the unipotent radical
of Gk is trivial.
For each prime number p we let
Zp := lim Z/pn , Qp := Zp ⊗Z Q
←−
be the p-adic integers resp. the field of p-adic numbers.
Let
Y
Af := {(xp )p ∈ Qp | xp ∈ Zp for all but finitely many primes p }
p

be the ring of finite adéles of Q, and let

A := Af × R
be the full ring of adéles of Q.
Let us recall that A is a locally compact topological ring. Namely, on Af there
exists a unique ring topology such that the subring
Y
Zp ⊆ Af
p

is open. On A one can take then the product topology with the usual topology on
R. An important property is that the subspace topology on the subring Q ⊆ A,
i.e., Q is embedded diagonally, is discrete.
The starting point for the Langlands program is a a reductive group G over Q,
the most important example being G = GLn for some n ≥ 1.
The above facts imply that G(A) ∼= G(Af ) × G(R) is naturally a locally compact
topological group, with G(Q) ⊆ G(A) a discrete subgroup. Namely, choose an
embedding G ⊆ Spec(Q[X1 , . . . , Xm ]) for some m ≥ 1, and take the subspace
topology of induced embedding G(A) ⊆ Am . This topology is independent of the
choice of the embedding. For more details see [GH19, Theorem 2.2.1.].
We will use the following facts:
• On any locally compact topological group H there exists a right-invariant
Haar measure, cf. [GH19, Section 3.2.], which is unique to a scalar in R>0 .
• On G(A) the right-invariant Haar measure is also left-invariant, and it
descends to a G(A)-invariant measure on G(Q)\G(A), cf. [GH19, Lemma
3.5.4.], [GH19, Lemma 3.5.3].
We now introduce one object of ultimate interest in the Langlands program for
G.
Definition 2.1. We set
L2 (G(Q)\G(A))
as the space of measurable functions f : G(Q)\G(A) → C such that
Z
|f |2 < ∞
G(Q)\G(A)
4 JOHANNES ANSCHÜTZ

where the integration is w.r.t. the G(A)-invariant measure on G(Q)\G(A).

As usual, two measurable functions in this L2 -space have to be identified if they
agree outside a set of measure zero.
The space L2 (G(Q)\G(A)) is naturally a Hilbert space via the inner product
Z
(f1 , f2 ) 7→ f1 (g)f 2 (g)dg.
G(Q)\G(A)

The action of G(A) on G(Q)\G(A) via right translation induces a (left) action of
G(A) on L2 (G(Q)\G(A)). By right-invariance of the chosen measure on G(Q)\G(A)
this action is unitary, i.e., preserves the inner product.
We can now state a principal aim of the Langlands program (in a very crude
form):

Decompose the Hilbert space L2 (G(Q)\G(A)) as a representation of G(A),

according to arithmetic data.

We will clarify a bit what we mean by “decomposing”. Set

[
G(A)
as the isomorphism classes of irreducible, unitary G(A)-representations (on Hilbert
spaces). Because of analytic issues it is too naive to hope that there exists a direct
sum decomposition
L2 (G(Q)\G(A)) “ ∼
M
=” π ⊕mπ
[
[π]∈G(A)

with mπ ∈ N ∪ {∞} multiplicity of some irreducible, unitary G(A)-representation

π.
Let us discuss the example G = Gm = GL1 . In this case,
G(Q)\G(A) ∼= Q× \A×
is a locally compact abelian group, and thus we can use (abstract) Fourier theory
to analyze the GL1 (A) = A× -representaiton
L2 (Q× \A× ).
Let us recally some result on abstract Fourier theory, following [DE14, Chapter
3].
• Let A be any locally compact abelian group (like R, S 1 = ∼ R/Z, A× , Z/n, Qp , ...).
• Let Ab be the set of isomorphism classes of irreducible, unitary representa-
tions of A.
• Then A b is in bijection to the set Homcts (A, S 1 ) of unitary, continuous char-
acters A → S 1 , i.e., each π ∈ A b is one-dimensional, and in particular
again a locally compact abelian group (via pointwise multiplication and
the compact-open topology).
ˆ
• Moreover, A ∼ = Â as topological groups via a 7→ (χ 7→ χ(a)) and A is
discrete if and only if Ab is compact.2

b embeds into Q S 1 . “⇐” Use A ∼

2“⇒” Choose L Z  A, then A ˆ
= Â and the definition of the
I I
compact-open topology.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM 5

The Plancherel theorem is a major result of abstract Fourier theory. For us its
relevance lies in the fact that it offers a different description of the space L2 (A) as
an A-representation.
Theorem 2.2 (Plancherel theorem, cf. [DE14, Chapter 3.4]). The Fourier trans-
form
F : L1 (A) ∩ L2 (A) → L2 (A)
b
defined by Z
f 7→ (χ 7→ f (x)χ(x)dµ(x))
A
extends to a unitary isomorphism
F : L2 (A) ∼
= L2 (A)
b
of Hilbert spaces.
The Fourier transform F is an A-equivariant isomorphism, when we let A act
on L2 (A)
b via
a · g(χ) := χ(a)g(χ)
2 b
for a ∈ A, g ∈ L (A), χ ∈ A. Thus,
b
L2 (A) ∼
= L2 (A)
b
as A-representations.
Let us discuss the two examples A = S 1 , and A = R..
• If A = S 1 , then
Z∼ b n 7→ χn with χn (z) = z n , z ∈ S 1 .
= A,
Thus
L2 (S 1 ) ∼
M
d
= Cχn
n∈Z
is the Hilbert space direct sum of the S 1 -equivariant subspaces
Cχn ⊆ L2 (S 1 ).
Concretely: each L2 -function f : S 1 → C can uniquely be written as
X X
f= an χn , where an ∈ C, |an |2 < ∞.
n∈Z n∈Z

Thus, we found a nice decomposition of the unitary S 1 -representation

L2 (S 1 ) into irreducible subspaces. The analytic issues are not immense.
Instead of a(n algebraic) direct sum of representations, we had to consider
L
the Hilbert space direct sum dCχn .
n∈Z
b∼
• If A = R, then R = R via
R → R, x 7→ χx , where χx : R → S 1 , y 7→ e2πixy .
b
The isomorphism
F : L2 (R) ∼
= L2 (R)
implies therefore that each f ∈ L2 (R) can be written as
Z
f (x) = g(y)χy (x)dy
R
6 JOHANNES ANSCHÜTZ

for a unique function g ∈ L2 (R). Thus,

Z
2 ∼
L (R) = χy dy
R

is a Hilbert integral of representation (cf. [GH19, Section 3.7]), a notion

which is not as immediate as the Hilbert space direct sum. Note that
we cannot do better: χx ∈ / L2 (R) for any x ∈ R, and L2 (R) contains no
irreducible subrepresentation of R, cf. [GH19, Section 3.7].3
Let us go back and consider G = Gm , i.e., A = A× .
• Factoring each n ∈ Q× into n = ±p1 . . . pk with pi prime implies
Q× \A× ∼
Y
= Z×
p × R>0 .
p

• Let us ignore the factor R>0 for the moment and look at
∼ d Cχ.
Y M
L2 ( Z× p)=
p χ

Z× 1
Z×
Q Q
with χ : p → S all continuous characters of p . Note that each
p p
character χ :
Q
Z× ∼ (Z/m)× → S 1 factors over some quotient (Z/m)× .
p = lim
←−
p m
• The characters χ are “arithmetic” data, namely
Gal(Q(µ∞ )/Q) ∼
Y
= Z×
p,
p
S 2πi
where Q(µ∞ ) = Q(e n ) is the cyclotomic extension of Q. Define
n∈N

α : Gal(Q/Q)  Gal(Q(µ∞ )/Q) ∼

Y
= Z×
p
p prime

as the natural projection.

Q ×
• Thus for each χ : Zp → S 1 the composition χ ◦ α is a (continuous)
p
1-dimensional Galois representations of Gal(Q/Q).
• By the theorem of Kronecker-Weber each 1-dimensional Galois representa-
tion
σ : Gal(Q/Q) → C×
is of the form σ = χ ◦ α for some χ.
Theorem 2.3 (Kronecker-Weber). The field Q(µ∞ ) is the maximal abelian ex-
tension of Q, i.e., each finite Galois extension F/Q with abelian Galois group
Gal(F/Q) is contained in Q(µ∞ ).
Thus, we obtain the following decomposition for G = Gm :
\
L2 (R>0 Q× \A× ) ∼
M
= Cχσ
σ : Gal(Q/Q)→C×

3Also no irreducible quotient, the statement about quotients in [GH19, Lemma 3.7.1] is wrong,
I think.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM 7

with χσ characterised by χσ ◦ α = σ. This (easy) example is prototypical for what

one aims for in the Langlands program. Note that modding out R>0 is convenient
and rather harmless. The full space
L2 (Q× \A×
b ∼
is a Hilbert integral of the representations L2 (R>0 Q× \A× ) over the space R =R
of unitary characters of R>0 .
Let us pass to general G and discuss the geometry of the space
G(Q)\G(A).
This will be much more complicated than in the case G = GL1 .
There exists a central subgroup AG ⊆ G(R), with AG ∼= Rr>0 , such that
[G] := AG G(Q)\G(A)
has finite volume, cf. [GH19, Theorem 2.6.2.]. Namely, AG can be taken as the
connected component of the maximal split subtorus of the center Z(G) of G, e.g.,
if G = GLn , then AGLn ⊆ GLn (R) is the group of scalar matrix with entries in R>0
while ASLn is trivial.
Again the full space L2 (G(Q)\G(A)) is a Hilbert space integral of the represen-
tations L2 ([G]) over AcG , and thus considering L2 ([G]) instead of L2 (G(Q)\G(A))
reduces some analytic issues.4
Although we are only interested in some arbitrary right-invariant measure on
G(A)5 let us mention that there exists a canonical measure on G(A), the Tamagawa
measure, and that for this measure the number
τ (G) := vol([G])
is called the Tamagawa number of G. The knowledge of τ (G) is arithmetically
interesting. Cf. [Col11, Appendix B], [Lur, Lecture 1]. For example, τ (SLn ) = 1,
which implies vol(SLn (R)/SLn (Z)) = ζ(2)ζ(3) . . . ζ(n), where ζ(s) = is the Rie-
mann ζ-function.
The topological spaces G(Q)\G(A) and [G] are profinite coverings of some real
manifold, e.g., if G = GLn , then
Y
GLn (Z)\( GLn (Zp ) × GLn (R)),
p

where the action of GLn (Z) is diagonally, and thus GLn (Q)\GLn (A) covers the real
manifold
GLn (Z)\GLn (R).
Similarly, [G] covers GLn (Z)R>0 \GLn (R). Cf. [GH19, Section 2.6], [Del73, (0.1.4.1)].
More precisely, for each compact-open subgroup K ⊆ G(Af ) (also called a “level
subgroup”) the quotient
[G]/K
is a real manifold.
Similarly, let K∞ ⊆ G(R) be a maximal compact (connected or not) subgroup
of G(R), e.g., if G = GLn , then one can take K∞ ∼
= SOn (R). As K∞ intersects

4For precision, let us mention that each right-invariant Haar measue on G(A) descents to [G],
cf. [GH19, Section 2.6].
5Because the resulting L2 ([G])-spaces are isomorphic.
8 JOHANNES ANSCHÜTZ

AG ⊆ G(R) trivially, the space [G] is a principal K∞ -bundle (or K∞ -torsor) over
[G]/K∞ .
In the next lecture we will discuss in more detail the case G = GL2 . For now let
us just mention that if K∞ ∼ = SO2 (R), then by Möbius transformations
 
∼ ± a b a·i+b
AG \GL2 (R)/K∞ − →H , g= 7→ g(i) :=
c d c·i+d
with
H± := {z ∈ C | Im(z) 6= 0}
the upper/lower halfplane. In particular,
[GL2 ]/K∞ ∼
Y
= GL2 (Z)\(GL2 ( Zp ) × H± )
p

Let K ⊆ G(Af ) be a compact-open subgroup, then the set

m
a
G(Q)\G(Af )/K = G(Q)gi K
i=1

is finite (cf. [GH19, Theorem 2.6.1]) and

m
[G]/K∞ K ∼
a
= Γi \X,
i=1
where
X := AG \G(R)/K∞
is a real manifold, and
Γi := G(Q) ∩ G(R)gi Kgi−1 ⊆ G(R)
is a discrete subgroup (an example of a congruence subgroup, cf. [GH19, Section
2.6]). For K is sufficiently small, the group Γi acts freely and properly discontinu-
ously on X, cf. [GH19, Definition 15.2]. In this case, the morphism
X → Γi \X
is a covering (in the sense of topology) with covering group Γi . The case we are
mostly interested in is the case G = GL2 . Then everything becomes more explicit.
Namely,
Γi \X ∼
= Γ\H±
with Γ ⊆ GL2 (Z) a discrete subgroup containing
Γ(m) := ker(GL2 (Z) → GL2 (Z/m))
for some m ≥ 0. For m ≥ 3, the action of Γ(m) is free on H± . This example will
provide the link with modular forms in the next lecture.
Note that for arbitrary G we in general have dim(Γi \X) > 0. Thus the geometry
[G], and the space L2 ([G]), is much more complicated than in case G = Gm .
The real manifolds Γi \X are interesting real manifolds, because of their arith-
metic significance. We will see that if G = GL2 , these Riemann surfaces are quasi-
projective and canonically defined over number fields.
If the space [G] is compact (which happens if and only if any Gm ⊆ G lies
in center, cf. [GH19, Theorem 2.6.2]), the analysis of L2 ([G]) is more simple. A
concrete example, when this happens is in the case that G is the units in a non-split
quaternion algebra over Q, i.e., the group of units in a Q-algebra with presentation
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM 9

hx, y|x2 = a, y 2 = b, xy = −yxi, for suitable a, b ∈ Q× . Note that G(R) ∼

= GL2 (R)
if ab < 0, and thus the Γi \X are still quotients of the upper half-plane. For more
details on quaternion algebras see [GS17, Chapter 1].
Reductive groups over Q exist in abundance. A particular, easy way to produce
many examples is via Weil restrictions. To introduce this let F/Q be a finite
extension, H reductive over F . Then the functor R 7→ H(R ⊗Q F ) on the category
of Q-algebras is representable by a reductive group
G := ResF/Q (H)
over Q, cf. [CGP15, Appendix A] or [GH19, Section 1.4]. For this reductive group
G we get
G(Q)\G(A) = H(F )\H(AF )
with AF ring of adéles for F . This implies that for the Langlands program consid-
ering all reductive groups over Q is equivalent to considering all reductive groups
over all number fields.
When H = Gm,F the desired decomposition of L2 ([G]) is (roughly) equivalent
to class field theory for F . Let us mention that the finiteness of the volume of [G]
incorporates the finiteness of the class number for F and Dirichlet’s theorem on
units for F .
For another source of examples consider the case when F is imaginary quadratic
and H = GL2,F . Then G(R) ∼ = GL2 (C), we can take K∞ ∼ = U (n) as the unitary
group and
X = AG \G(R)/K∞
is the 3-dimensional hyperbolic space H3 . The quotients Γi \X in this case are called
arithmetic hyperbolic 3-manifolds, cf. [Thu82, Section 4 & 5]. More precisely, the
hyperbolic 3-space H3 can be modelled on the quaternions
{q = x + y · i + z · j | x, y ∈ R, z > 0}
 
a b
and ∈ GL2 (C) acts on H3 via q 7→ (aq + b)(cq + d)−1 , where C = {x + y ·
c d
i | x, y ∈ R}.
After having discussed some examples, we can now give the definition of an
automorphic representation (in the L2 -sense), which is of central importance in the
Langlands program.
Definition 2.4 ([GH19, Definition 3.3]). Let G be a reductive group over Q.
An automorphic representation (in the L2 -sense) is an irreducibe unitary G(A)-
representation π which is isomorphic to a subquotient of L2 ([G]).
Later, we will also discuss a different, more algebraic notion of an automor-
phic representation. In the above definition, π is necessarily trivial on AG . This
assumption is usually harmless. By Schur’s lemma for irreducible, unitary represen-
tations we can find for any irreducible, unitary G(A)-representation π a character
χ : G(R) → C× with χ ⊗ π trivial on AG , cf. [GH19, Section 6.5]. E.g., for G = GLn
one can take χ of the form g 7→ |det(g)|s ) for some s ∈ R>0 .
The space L2 ([G]) decomposes into a “discrete” and a “continuous” part, cf.
[GH19, Section 10.4]. From an analytic perspective the case that [G] is compact is
easier as there the continuous part vanishes.
We now explain more precisely what is meant by the “discrete” and “continu-
[ denotes the set of isomorphism classes of irreducible
ous” part. Recall that G(A)
10 JOHANNES ANSCHÜTZ

unitary G(A)-representations. Generalizing the compact-open topology on A b for A

a locally compact abelian group, the set G b has a natural topology, the Fell topol-
ogy, cf. [GH19, Section 3.8]. Roughly, the Fell topology can be described as follows.
[ x ∈ π, and ϕ : π → C a continuous C-linear homomorphism.
Consider π ∈ G(A),
Then
fx,ϕ : G → C, g 7→ ϕ(gx)
is called a matrix coefficient of π. Roughly, two π, π 0 ∈ G(A)
[ are close in the Fell
topology if their matrix coefficients are close for the compact-open topology.
The following theorem holds more generally for a unitary representation of G(A).
We don’t need the exact meaning of the integral in Theorem 2.5, let us just mention
that its occurence is due to the fact that some representations of G(A) “contribute”
to L2 ([G]) without being subrepresentations (similar to the case of Fourier theory
for R).
Theorem 2.5 ([GH19, Theorem 3.9.4]). There exists a measurable multiplicity
b → {1, 2, . . . , ∞} and a measure µ on G
function m : G b such that
Z Mm(π)
L2 ([G]) ∼ ( π)dµ(π).
d
=
G
b
The m and µ are unique up changes on sets of measure 0.
With this theorem we can now explain roughly what is by the “discrete” and
“continuous” part of L2 ([G]). Points π ∈ G(A)
[ with µ(π) > 0 are “discrete” for
[ and they appear as subrepresentations in L2 ([G]) with
the measure m on G(A),
multiplicty m(π). This motivates the decomposition
L2 ([G]) ∼
= L2 ([G]) ⊕ L2 ([G]),
disc cont
2
where Ldisc ([G]) is the Hilbert sum of all irreducible, unitary subrepresentations of
G(A), and L2cont ([G]) its orthogonal complement, cf. [GH19, Section 9.1].
In a pioneering work Langlands described the continuous part via (proper) par-
abolic subgroups of G. Recall that by definition, a closed, connected subgroup
P ⊆ G (over Q) is parabolic if G/P is a proper scheme over Spec(Q), and that the
quotient of P by its unipotent radical is its Levi quotient M , cf. [GH19, Section
1.9]. For example, for GLn take a decomposition n = n1 + n2 + . . . + nk , and
P as the subgroup of block upper triangular matrices in GLn with blocks of size
n1 , . . . , nk . The Levi quotient is M = block diagonal matrices. Up to conjugation
all parabolics in GLn are of this form.
Langlands then proved that the continuous part L2cont ([G]) can be described via
“inducing” the discrete parts L2disc ([M ]) for all Levi quotients of parabolics P ⊆ G,
P 6= G, cf. [GH19, Section 10.4]). This is his theory of “Eisenstein series”, and it
is a starting point of the Langlands program.
For this course, L2disc ([G]) (actually the proper “cuspidal” subspace) will be more
important than the whole space L2 ([G]). Let us mention that [G] is compact if and
only G has no proper parabolics (defined over Q). This is in accordance with the
previous claim that L2 ([G]) = L2disc ([G]) in this case.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM11

3. From modular forms to automorphic representations, part I

Starting with this lecture we want to construct automorphic representations for
GL2 associated to modular forms.
Let us recall that we discussed last time:
• Consider a G reductive group over Q, and A = Af × R the ring of adéles.
• Then we introduce the space [G] = AG G(Q)\G(A), called the adélic quo-
tient, which has finite G(A)-invariant volume.
• We stated that the Langlands program aims to decompose the unitary
G(A)-representation
L2 ([G])
according to arithmetic data.
• In the example, G = Gm we showed that
\
L2 (R>0 Q× \A× ) ∼
M
= C ◦ χσ ,
σ : Gal(Q/Q)→C×

as an example that the desired arithmetic data arise from Galois represen-
tations of Gal(Q/Q).
• Then we discussed the geometry of [G]:
– let K∞ ⊆ G(R) be some maximal compact, connected subgroup,
– let K ⊆ G(Af ) be some compact-open subgroup.
– Then
[G]
K-torsor (in particular: a profinite covering)

[G]/K is a real manifold, at least for sufficiently small K

K∞ -torsor

m̀
[G]/KK∞ = Γi \X is a disjoint union of arithmetic manifolds
i=1

with Γi ⊆ G(R) some (discrete) congruence subgroups determined by

K, and
X := AG \G(R)/K∞ .
• Moreover, we defined that an irreducible unitary representation π of G(A)
is automorphic (in the L2 -sense) if it is isomorphic to a subquotient of
L2 ([G]).
• First examples (only mentioned implicitly) are the following:
– if G = Gm , then the automorphic representations (in the L2 -sense) of
G(A) are the unitary characters of A× , which are trivial on Q× · R>0 .
– if G is general the trivial representations (=constant functions on [G])
is automorphic because [G] has finite volume. Moreover, it lies in the
discrete part L2disc ([G]) of L2 ([G]).
• For general G there exists the orthogonal decomposition
L2 ([G]) ∼= L2 ([G]) ⊕ L2 ([G])
disc cont

with L2disc ([G]) the closure of the sum of all irreducible subrepresentations
of G(A).
12 JOHANNES ANSCHÜTZ

• Finally, we mentioned that Langlands has described L2cont ([G]) via L2disc ([M ])
for M running through Levi quotients of parabolic subgroups P ⊆ G, P 6= G
(which are defined over Q).
Today, we will discuss in more detail the case G = GL2 . First of all,
AGL ∼ = R>0 2

embedded as scalar matrices into GL2 (R). Last time we claimed that we have
GL2 (Q)\GL2 (A) ∼ b × GL2 (R)),
= GL2 (Z)\(GL2 (Z)
where
b := lim Z/m ∼
Y
Z = Zp .
←−
m p
Let us explain why this is true, cf. [Tho, Theorem 1]. The group GL2 (Af ) is
the restricted product of the groups GL2 (Qp ) with respect to the compact-open
subgroups GL2 (Zp ), i.e.,
Y
GL2 (Af ) = {(Ap )p ∈ GL2 (Qp ) | Ap ∈ GL2 (Zp ) for all but finitely many p}
p

(cf. [GH19, Proposition 2.3.1]). The Chinese remainder theorem implies that Q ⊆
Af (embedded diagonally) is dense. Moreover,
A× × b×
f =Q ·Z
using prime factorization.
The groups GL2 (Zp ), GL2 (Qp ) are generated by elementary and diagonal matri-
ces which implies that
GL2 (Af ) = GL2 (Q)GL2 (Z),
b
and thus
GL2 (Q)\GL2 (Af ) ∼ = GL2 (Q) ∩ GL2 (Z)\GL
b b ∼
2 (Z) = GL2 (Z)\GL2 (Z).
b
From here we can deduce that as desired
GL2 (Q)\GL2 (A) ∼ b × GL2 (R)).
= GL2 (Z)\(GL2 (Z)
The proof works in fact for all n ≥ 1 and shows
GLn (Q)\GLn (A) ∼ b × GLn (R)).
= GLn (Z)\(GLn (Z)
Let us define
b → GL2 (Z/m))
Km = ker(GL2 (Z)
for m ≥ 1. Then these groups are cofinal within all compact-open subgroups in
GL2 (Af ). We have
GL2 (Q)\GL2 (Af )/Km ∼ = GL2 (Z)\GL2 (Z)/K
b m

and GL2 (Z)/K

b ∼
m = GL2 (Z/m). The group SL2 (Z/m) is even generated by elemen-
tary matrices.6 This implies that the morphism SL2 (Z) → SL2 (Z/m) is surjective.
Thus, we obtain
GL2 (Z)\GL2 (Z)/K
b ∼
m = {±1}\(Z/m)
×

6 use the euclidean algorithm and the magic identity

      
a 0 1 a 1 0 1 a 0 −1
=
0 a−1 0 1 −a−1 1 0 1 1 0
for any invertible element a in some ring R.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM13

via the determinant. Now, define

K∞ := SO2 (R) ⊆ GL2 (R)
and
X := AGL2 \GL2 (R)/K∞ .
Then X ∼
= H ±
(the upper/lower halfplane) via the morphsim
 
∼ a b a·i+b
AGL2 \GL2 (R)/K∞ − → H± , g = 7→ g(i) := .
c d c·i+d
Thus, we obtain in the end
GL2 (Q)\(GL2 (Af )/Km × X) ∼
a
= Γ(m)\H±
{±1}\(Z/m)×

with
Γ(m) := Km ∩ GL2 (Z) = ker(GL2 (Z) → GL2 (Z/m)).
The fact that GL2 (Q)\(GL2 (Af )/Km ×X) is non-connected may seem annoying,
but it will turn out to be an advantage. Note that if we replace K ⊆ GL2 (Af ) by
some compact-open subgroup K such that det : K → Z b × is surjective, e.g., K
equals  
a b
K1 (m) := { ∈ GL2 (Af ) | c ≡ 0, d ≡ 1 mod m, },
c d
then GL2 (Q)\(GL2 (Af )/Km × X) is connected.
The finiteness of the volume of [G] for G = GL2 can now be checked via a direct
argument.
The compactness of K∞ Km implies that it suffices to show that the space
[GL2 ]/K∞ Km has finite volume. Writing this space as a disjoint union we can
even reduce to the statement that Γ(m)\H± has finite volume for m ≥ 1. Up to
scalar in R>0 , the GL2 (R)-invariant measure on H± is given by the volume form
1
dx ∧ dy.
y2
Indeed, we have to show

1 1
g∗ (
2
dx ∧ dy) = 2 dx ∧ dy
y y
for g ∈ GL2 (R). Now, write dx ∧ dy = 2i dz ∧ dz with z = x + iy. The statement
can now easily shown to be true for
 
1 a
g= , a ∈ R,
0 1
for AGL2 , for  
a 0
g= , a ∈ R× ,
0 1
and finally  
0 1
g= ,
1 0
i.e., for z 7→ z1 = |z|z 2 . From the magic identity mentioned before we can conclude
invariance for lower triangular matrices, and thus for all of GL2 (R).
14 JOHANNES ANSCHÜTZ

Let us now show that Γ(m)\H± has finite volume. A known fundamental domain
for SL2 (Z) on H is given by the set
D := {z ∈ H | |z| > 1, |Re(z)| < 1/2}
and finiteness of the volume of GL2 (Z)\H± follows from
Z
1
dx ∧ dy < ∞.
y2
D

Finally, as Γ(m) ⊆ GL2 (Z) is of finite index Γ(m)\H± is of finite volume, as desired.
The Langlands program incorporates in particular a close connection between
modular forms and 2-dimensional (`-adic) Galois representation of Gal(Q/Q). Dis-
cussing this connection will be a major theme of the course. First let us recall the
classical definition of modular forms, cf. [DS05]. For this, let Γ ⊆ SL2 (Z) be a
congruence subgroup, later assumed to be sufficiently small. Fix k ∈ Z.
Definition 3.1. A function f : H± → C is a modular form of weight k for Γ if
 
a b
• f (γ · z) = (cz + d)k f (z) for γ = ∈Γ
c d
• f is holomorphic
• f is holomorphic at the cusps of Γ (we will explain this in a second).
Let Mk (Γ) be the C-vector space of modular forms of weight k for Γ. Let us
also call a function weakly modular if it satisfies only the first two conditions, but
is not necessarily holomorphic at the cusps.
Let us recall that the upper halfplane H naturally emdeds SL2 (Q)-equivariantly
into the set
H∗ := H ∪ P1 (Q)
equipped with the Satake topology. The orbits of a congruence subgroup Γ ⊆
SL2 (Z) on P1 (Q) are called the cusps for Γ, and the quotient
Γ\H∗
is a natural compactification of Γ\H. Adding another copy of P1 (Q) to the lower
halfplane H− , one obtains similar objects/notions for Γ ⊆ GL2 (Z).
The condition of being holomorphic at the cusps, means now the following. For
each cusp σ a weakly modular form f for Γ has a Fourier expansion, i.e., up to
translating the cusp to ∞ ∈ P1 (Q) there exists an equality
X
f (z) = an q n/h
n∈Z
2πiz
for q = e , h the so-called smallest periodicity of f , and an ∈ C, n ∈ Z. The
weakly modular form f is then called holomorphic at this cusp if an = 0 for n < 0,
and it is called holomorphic at the cusps if it is holomorphic at every cusp. If
furthermore a0 = 0 for each cusp, then f is called a cusp form. We denote by
Sk (Γ) ⊆ Mk (Γ)
the subspace of cusp forms of weight k for Γ.
Our next aim is to describe (weakly) modular forms of weight k ∈ Z for Γ ⊆
GL2 (Z) canonically as sections of some holomorphic line bundle
ωk
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM15

over Γ\H± (at least for sufficiently small Γ).

Before doing this we discuss some useful constructions of “descent for fiber bun-
dles”, which will be used frequently.
Consider a set Z, a group H acting freely on Z (from the right). Then the
projection
Z → Z/H
deserves to be called an H-torsor. Let F be a set with a (left) action of H. The
contracted product
Z ×H F
of Z over H with F is defined as the quotient of
Z ×F
for the action h · (z, f ) 7→ (zh−1 , hf ) of H. Via the natural projection (z, f ) 7→ z
we obtain the “fiber bundle”
Z ×H F → Z/H
over Z/H whose fibers are (non-canonically) isomorphic with F . Indeed, for a class
c ∈ Z/H and the choice of z ∈ Z mapping to c, we obtain the isomorphism
F ∼
= (Z ×H F ) ×Z/H c, f 7→ (z, f ).
Be aware that − ×H − denotes the contracted product, while − ×Z/H − the fiber
product (and these are two unrelated constructions). In particular, there exists the
canonical isomorphism
Z ×F ∼
= (Z ×H F ) ×Z/H Z, (z, f ) 7→ ((z, f ), z),
i.e., the fiber bundle Z ×H F is canonically trivial after pullback to Z/H.
We can conveniently describe sections of the fiber bundle Z ×H F → Z/H.
Lemma 3.2. The set of sections s : Z/H → Z ×H F , i.e., morphism over Z/H,
identifies canonically with the set of functions ϕ : Z → F such that ϕ(zh) = h−1 ϕ(z)
for all z ∈ Z, h ∈ H.
Proof. Given s : Z/H → Z ×H F and z ∈ Z, there exists (by freeness of the action
of H on Z) a unique element f ∈ F such that
s(z) = (z, f ).
Set ϕ(z) := f . Then
(z, ϕ(z)) = s(z) = s(zh) = (zh, ϕ(zh)) = (z, hϕ(zh),
which implies
ϕ(z) = hϕ(zh)
as desired. Conversely, given as ϕ : Z → F satisfying ϕ(zh) = h−1 ϕ(z) for all
z ∈ Z, h ∈ H, the function
Z → Z × F, z 7→ (z, ϕ(z))
is, by definition, invariant for the action of H, and thus descents to a section
s : Z/H → Z ×H F.
Both constructions are inverse to each other. 
16 JOHANNES ANSCHÜTZ

Conversely, assume we are given any morphism Y → Z/H with an isomorphism

over Z
g : Y ×Z/H Z ∼
=Z ×F
for some set F . Pick z ∈ Z with image c ∈ Z/H. Then for any h ∈ H we obtain
an isomorphism
−1
g|z×F h g|g−1 (zh×F )
σh : F ∼
= Yc ×c z ∼
= Yc ×c zh ∼
= F,
which define an action of H on F . Moreover, one can check that via this action
Y ∼
= Z ×H F.
In more fancy terms, one obtains that “fiber bundles” Y → Z/H with a “trivial-
ization” over Z are equivalent to sets with an action of H.
Let us mention also the following variant, when Z = G is a group and H ⊆ G
a subgroup. A G-equivariant set Y over G/H is a space π : Y → G/H over G/H
together with a left action of G such that π is G-equivariant. In this case, one has
the canonical trivialization
G × Y1·H ∼
= Y ×G/H G, (g, y) 7→ (gy, g)
and one can conclude that the category of G-equivariant sets over G/H is just
equivalent to the category of sets equipped with a (left) action of H, because for
an H-set F the contracted product
G ×H F
is naturally G-equivariant (by letting G via left multiplication on the left factor).
Note that we required that H acts on Z (resp. G) on the right, but of course we
can similarly consider “fiber bundles” over H\Z.
We placed the constructions in the category of sets, but it is clear that if Z, H, F
are topological spaces (real manifolds, complex manifolds,...) and the action of H is
continuous (smooth, holomorphic, ...) we can perform the analogous constructions
in the category of topological spaces (real manifolds, complex manifolds,...) etc..
Let us now start to describe modular forms as sections of line bundles on modular
curves.
First of all, let us note that the line bundles
OP1C (k)
for k ∈ Z are naturally GL2 (C)-equivariant. Indeed, by taking tensor powers it is
enough to consider the case k = 1. But then OP1C (1) is associated to the natural
Gm -torsor
A2C \ {0} → P1C , (z1 , z2 ) 7→ [z1 : z2 ],
which is naturally GL2 (C)-equivariant. Next, we observe that the embedding
H± ⊆ P1C , z 7→ [z : 1]
is GL2 (R)-equivariant, and thus we obtain for n ∈ Z the GL2 (R)-equivariant line
bundle

OH± (k) := OP1C (k) ×P1C H±


H±
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM17

over H± . The stabilizer of i ∈ H± in GL2 (R) identifies with

C× ∼
= R>0 SO2 (R) ⊆ GL2 (R).
Using
H± ∼
= GL2 (R)/C×
the OH± (k) correspond, under the constructions for fiber bundles mentioned before,
to the (holomorphic) representation
χn : C× → GL1 (C) = C× , z 7→ z k .
Note that Lemma 3.2 implies that smooth sections of OH± (k) on H± identify with
smooth functions
ϕ : GL2 (R) → C
satisfying
ϕ(gz) = z −k ϕ(g), g ∈ GL2 (R), z ∈ C× .
This will be used later. From now on we assume that Γ is sufficiently small, i.e., it
acts freely on H± . Modding out the action of Γ on OH± (k) yields the holomorphic
line bundle
ω ⊗k := Γ\OH± (k)


Γ\H± ,
over Γ\H± . To get the link of sections of ω k with (weakly) modular forms of weight
k for Γ, we have to make the GL2 (R)-action on
OH± (k)
more explicit. For k ∈ Z we define the holomorphic GL2 (R)-equivariant line bundle
Lk := H± × C
over H± with GL2 (R)-acting on the left by
az + b
g · (z, λ) := ( , (cz + d)k λ)
cz + d
 
a b
for g = ∈ GL2 (R). For notational convenience let us introduce the function
c d
(a “factor of automorphy”)
 
± × a b
j : GL2 (R) × H → C , (g = , z) 7→ j(g, z) := cz + d.
c d
We leave it as an exercise to check that he map
L∗1 → OH± (1)∗ = H± ×P1C A2C \ {0},
(z, λ) 7→ (z, (λz, λ))
is a holomorphic isomorphism of GL2 (R)-equivariant line bundles over H± . Here
(and in the following) (−)∗ denotes the complement of the zero section in a line
bundle. By taking tensor powers we obtain
Lk ∼
= OH± (k)
18 JOHANNES ANSCHÜTZ

for all k ∈ Z, as GL2 (R)-equivariant holomorphic line bundles.7 Similarly to

Lemma 3.28 we get that a holomorphic section
H± → Lk , z 7→ (z, f (z))
is Γ-equivariant if and only if f (z) satisfies
f (γ · z) = j(γ, z)k f (z)
for γ ∈ Γ, i.e.,
az + b
f( ) = (cz + d)k f (z)
  cz + d
a b
if γ = ∈ Γ. In other words, we proved that
c d
H 0 (Γ\H± , ω ⊗k )
identifies with the space of weakly modular forms for Γ. In particular, we have a
natural embedding
Mk (Γ) ⊆ H 0 (Γ\H± , ω ⊗k )
with image defined by condition of being “holomorphic at the cusps”. Let us analyze
what the condition of being holomorphic at the cusps means in terms of sections of
ωk .
We claim that the ω k extend canonically to the (Satake) compactification
Γ\H±
of Γ\H± . The stabilizer Γ∞ of the cusp ∞ ∈ P1 (Q) in SL2 (Z) is the subgroup of
elements  
1 a
0 1
with a ∈ Z. Thus, it suffices to see that ω extends (canonically) along the embed-
ding
Γ∞ \(H ∪ {∞}).
Write
L+k := Lk|H .
Then the Γ∞ -action on L+
k = H × C is only via the first factor (as j(γ, z) = z for
all z ∈ H and any γ ∈ Γ∞ ). Note that via expontial map
Γ∞ \H ∼= D× := {q ∈ C× | 0 < |q| < 1}
(this was implicitly used when discussing Fourier expansions). Thus canonically,
Γ∞ \L+ ∼ = D× × C,
k
×
which clearly extends to D := {q ∈ C | |q| < 1}, namely to D × C. This proves
as desired that ω k extends canonically to the mentioned (Satake) compactification
Γ\H± of Γ\H± . We called this extension again ω k . The condition of holomorphicity
7That L and O
k H± (k) are abstractly isomorphic, as GL2 (R)-equivariant line bundles can also
be deduced from the observation that both correspond to the C× -representation z 7→ z k . This
was noted for OH± , and follows because under the embedding C× an element z ∈ C× is mapped
 
Re(z) −Im(z)
to the matrix g = and j(g, i) = z.
Im(z) Re(z)

8
The sitation is slightly different than in Lemma 3.2, namely, these Γ-equivariant vector bundles
are not associated to a representation of Γ.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM19

at the cusps now translates into the statement that the section f ∈ H 0 (Γ\H± , ω k )
extends to a section of ω k on Γ\H± .
Using a bit more theory, the Riemann surface
Γ\H±
is an algebraic curve over C, and the analytification of its canonical compactification
is Γ\H± . As Mk (Γ) is the image of H 0 (Γ\H± , ω k ) in H 0 (Γ\H± , ω k ), one can deduce
(from GAGA) that Mk (Γ) is finite-dimensional over C. Moreover, it can be shown
that ω is ample on Γ\H± which implies that Mk (Γ) = 0 for k < 0 and Mk (Γ) gets
“big” for k  0. In fact, using the theorem of Riemann Roch one can calculate the
dimension of Mk (Γ). Let us finally note that
Sk (Γ) ∼
= H 0 (Γ\H± , ω ⊗k (−D))
where D = Γ\H± \ Γ\H± is the (reduced) divisor at infinity.
For more details, we refer to [DS05, Chapter 3, Chapter 7].
We now explain how to pass from sections of ω ⊗k to functions on GL2 (A) (or
[GL2 ]).
For this we will interpret modular forms for Γ with Γ ⊆ GL2 (Z) varying among
all congruence subgroup via “modular forms” for K where K ⊆ GL2 (Af ) is varying
among all compact-open subgroups.
Let K ⊆ GL2 (Af ) be a compact-open subgroup and write
m
GL2 (Q)\(GL2 (Af )/K) × H± ) ∼
a
= Γi \H±
i=1

with Γi ⊆ GL2 (Z) some congruence subgroups. Then OH± (k) defines by pullback
from H± a (complex) line bundle ω ⊗k on
GL2 (Q)\(GL2 (Af )/K) × H± )
(we implicitly assumed that K is sufficiently small and used GL2 (Q)-equivariance
of OH± (k)).
The spaces Mk (Γi ) of modular forms for Γi , i = 1, . . . , m, (even the weakly
modular forms) embed into the space of holomorphic sections
H 0 (GL2 (Q)\(GL2 (Af )/K × H± ), ω ⊗k ).
Set
m
M
Mk (K) := Mk (Γi ).
i=1
Then Mk (K) ⊆ H 0 (GL2 (Q)\(GL2 (Af )/K × H± ), ω ⊗k ) is defined analogously by
the condition of being holomorphic at all cusps. We can even get rid of K and use
the pullback of ω ⊗k to define a G(Af )-equivariant line bundle, again written ω ⊗k ,
on
GL2 (Q)\(GL2 (Af ) × H± ).
Note that this space is no longer a complex manifold, only an inverse limit of
complex manifolds along finite covering maps. But still we can define an analog of
“holomorphic sections” of ω ⊗k on it via
H 0 (GL2 (Q)\(GL2 (Af ) × H± ), ω ⊗k )
:= lim H 0 (GL2 (Q)\(GL2 (Af )/K × H± ), ω ⊗k ).
−→
K
20 JOHANNES ANSCHÜTZ

Moreover, we get the (big space)

Mk := lim Mk (K)
−→
K

of modular forms of level k. Later we will explain that for K ⊆ GL2 (Af ) (sufficiently
small) we obtain Mk (K) as the invariants of K under some canonical action of K
on Mk .
From our discussion of fiber bundles around Lemma 3.2 we know that the pull-
back of
OH± (k)
along
GL2 (R) → H± , g 7→ g · i
is canonically trivial (as a GL2 (R)-equivariant bundle), and that smooth sections
of OH± (k) on H± identify canonically with smooth functions
ϕ : GL2 (R) → C
satisfying
ϕ(gz) = z −k ϕ(g), g ∈ GL2 (R), z ∈ C× .
Concretely we associate to a smooth section
H± → Lk = H± × C, z 7→ (z, f (z))
the smooth function
ϕf : GL2 (R) → C, g 7→ j(g, i)−k f (gi).
Note that smooth sections of Lk on H± are simply smooth functions on H± (but
this identification is not GL2 (R)-equivariant!).
We can do something similar with H± replaced by GL2 (Af ) × H± , if we slightly
extend the notion smoothness.
Definition 3.3. Let Yi , i ∈ I, a cofiltered inverse system of real manifolds such
that each transition map Yi → Yj is a finite covering. Set Y := lim Yi and let
←−
i∈I
πi : Y → Yi be the canonical projection. Then we call a function
ϕ: Y → C
smooth if ϕ = ϕi ◦ πi for some i ∈ I and some smooth function ϕi : Yi → C.
Note that by the same pattern we can also define holomorphic functions (or
sections of line bundles) on Y if the Yi is additionally assumed to be a complex
manifold. Of course, we want to apply this terminology to GL2 (Af ) × H± (or
GL2 (A) = GL2 (Af ) × GL2 (R)), which is the inverse limit of the complex manifolds
GL2 (Af )/K × H± (resp. real manifolds GL2 (Af )/K × GL2 (R)) where K is running
through the cofiltered system of compact-open subgroups of GL2 (Af ).
We obtain that smooth sections
GL2 (Af ) × H± → GL2 (Af ) × Lk , (g, z) 7→ (g, z, f (g, z))
identify with smooth functions
ϕ : GL2 (Af ) × GL2 (R) → C
satisying
ϕ(g, g∞ z) = z −k ϕ(g, g∞ )
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM21

for (g, g∞ ) ∈ GL2 (A) and z ∈ C× ⊆ GL2 (R). Concretely, for f we set
ϕf (g, g∞ ) := j(g∞ , i)−k f (g, g∞ i).
Note that the section
GL2 (Af ) × H± → GL2 (Af ) × Lk , (g, z) 7→ (g, z, f (g, z))
is invariant under the action of GL2 (Q) from the left if and only if f satisfies the
modularity condition
f (γg, γz) = j(γ, z)k f (g, z).
Using the equation
j(γg∞ , i) = j(γ, g∞ i)j(g∞ , i)
we obtain that f defines a GL2 (Q)-equivariant section of Lk if and only if the
function ϕf is GL2 (Q)-invariant, i.e.,
ϕf (γg, γg∞ ) = ϕf (g, g∞ )
for (g, g∞ ) ∈ GL2 (A).
Restricting to holomorphic sections we have thus constructed the natural mor-
phism
H 0 (GL2 (Q)\(GL2 (Af ) × H± ), ω ⊗k ) → C ∞ (GL2 (Q)\GL2 (A))
such that f (g, z) 7→ ϕf (g, g∞ ) := j(g∞ , i)−k f (g, g∞ i), where we viewed a section
of ω ⊗k on GL2 (Q)\(GL2 (Af ) × H± ) as a function
GL2 (Af ) × H± → C, (g, z) → f (g, z)
satisfying the above modularity condition.
Let us recall that
Mk = lim Mk (K)
−→
K
with Mk (K) a sum of spaces of modular forms
m
M
Mk (Γi )
i=1

for various congruence subgroups Γi ⊆ GL2 (Z), and that the image of Mk in
H 0 (GL2 (Q)\(GL2 (Af ) × H± ), ω ⊗k ) is defined by the condition of “holomorphic-
ity at the cusps”.
Thus, we have related modular forms to smooth functions on GL2 (Q)\GL2 (A).
Note that
[GL2 ] = GL2 (Q)AGL2 \GL2 (A),
but our functions ϕf are not invariant under AGL2 , namely
ϕ(g, g∞ r) = r−k ϕ(g, g∞ )
for r ∈ AGL2 ∼
= R>0 . However, it is not difficult to rectify this. Let
Y
| − |adélic : A× → R>0 , ((xp )p , x∞ ) 7→ |xp |p |x∞ |
p
×
be the adélic norm (which is trivial on Q ). Then for f ∈ Mk the function
k/2
ef (g, g∞ ) := |det(g, g∞ )|adélic ϕf (g, g∞ )
ϕ
is invariant under GL2 (Q)AGL2 , i.e., defines a smooth function [GL2 ] → C.
22 JOHANNES ANSCHÜTZ

We have to clarify when modular forms give rise to functions in L2 ([GL2 ]). For
this, pick a section f ∈ H 0 (GL2 (Q)\(GL2 (Af ) × H± , ω ⊗k ). The question is: When
is
ef ∈ C ∞ ([GL2 ])
ϕ
in L2 ([GL2 ])? Pick a compact-open subgroup K ⊆ GL2 (Af ) such that
f ∈ H 0 (GL2 (Q)\(GL2 (Af )/K × H± ), ω ωk ).
The function ϕef is L2 if and only if the smooth function ϕef ∈ C ∞ ([GL2 ]/K) ⊆
∞ 2
C ([GL2 ]) is L (as K is compact). This reduces us to the question: For Γ ⊆
GL2 (Z) some congruence subgroup, f ∈ H 0 (Γ\H± , ω ⊗k ). When is the function
ef (g) = |det(g)|k/2 (ci + d)−k f (g(i))
ΓAGL2 \GL2 (R) → C, g 7→ ϕ
square-integrable, i.e., in L2 ? We claim:
|ϕf (g)|2 = |f (g(i))|2 |Im(g(i))|k
for any g ∈ GL2 (R), or equivalently,
?
|det(g)|−1/2 |ci + d| = |Im(g(i))|1/2
 
a b
for g = ∈ GL2 (R). But this is true as both sides are invariant under
c d
• K∞ = SO2 (R)-invariant
  from the right,
1 ∗
• the subgroup ⊆ GL2 (R) acting from the left,
0 1
• the diagonal matrices in GL2 (R) acting from the left.
Hence Z Z
1
|ϕf (g)|2 dg = |f (z)|2 |y|k 2 dx ∧ dy
y
ΓAGL2 \GL2 (R) Γ\H±

(Recall that we integrate over “the” invariant measure of H± .) As Γ\H± has finitely
many cusps this leads us to a local statement at cusps. By conjugation we may
reduce to consider ∞ ∈ P1 (Q). Then write f in Fourier expansion at ∞, i.e.,
X
f (z) = an q n/h .
n∈Z

for some h ≥ 0 (depending on Γ). Then

X
|y|k f (z) = 2π|log(|q|)|k an q n/h
n∈Z

as y = log(|q|). Now for 0 < r ≤ 1

Z X i
|log(|q|)|k−2 an q n/h dq ∧ dq < ∞
2
n∈Z
0<|q|≤r

if an = 0 for n ≤ 0, i.e., if f is cuspidal.

Thus we get, as a summary of the previous discussion, that the map
Mk ⊆ H 0 (GL2 (Q)\(GL2 (Af ) × H± ), ω k ) → C ∞ ([GL2 ]), f 7→ ϕ
ef
induces a canonical inclusion
e : Sk → L2 ([GL2 ]),
Φ
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM23

where Sk := lim Sk (K).

−→
K
We can make the following observations:
• GL2 (Af ) acts on H 0 (GL2 (Q)\(GL2 (Af )×H± ), ω k ), C ∞ (GL2 (Q)\GL2 (A)),
C ∞ ([GL2 ]), L2 ([GL2 ]), Mk , Sk
• The morphism
Φ : Mk → C ∞ (GL2 (A)), f 7→ ϕf
is GL2 (A)-equivariant, but not
k/2
Mk → C ∞ ([GL2 ]), f 7→ ϕ
ef = | det |adélic ϕf
k/2
(because of the factor |det|adélic ).
• More precisely, for g ∈ GL2 (Af ) and varying K ⊆ GL2 (Af ) the isomor-
phism
g
Mk (K) −→ Mk (g −1 Kg) ⊆ Mk
g
defines the action Mk −
→ Mk as Mk = lim Mk (K).
−→
K
• GL2 (Af ) does not act on Mk (K), Sk (K), . . . for any fixed K ⊆ GL2 (Af ).
The existence of the GL2 (Af )-action is the main motivation why we switched
from Γ-modular forms to K-modular forms.
We thus see our next tasks:
• Describe Sk as a GL2 (Af )-representation.
• Characterize the image of Sk in L2 ([GL2 ]) (or of Mk in C ∞ (GL2 (Q)\GL2 (A))).
Hopefully, the link of modular forms to automorphic representations (in the
L2 -sense) becomes more clear now. For f ∈ Sk we can take the closure of the
subspace generated by the GL2 (A)-translates of f and hope that it defines an
irreducible, unitary representation. Note that the representations obtained in this
way lie automatically in the discrete part L2disc ([GL2 ]) of L2 ([GL2 ]).
24 JOHANNES ANSCHÜTZ

4. From modular forms to automorphic representations, part II

Today:
• Continue construction of automorphic representations associated to cusp
forms.
Last time:
• We constructed a GL2 (Af )-equivariant embedding
Φ : Mk → C ∞ (GL2 (Q)\GL2 (A)),
which induces an embedding
e : Sk → L2 ([GL2 ]),
Φ
k/2
which is GL2 (Af )-equivariant up to multiplying with |det|adélic , i.e., the
morphism
Φe : Sk ⊗ |det|k/2 → L2 ([GL2 ])
adélic
is GL2 (Af )-equivariant.
• Here
Mk = lim Mk (K)
−→
K⊆GL2 (Af )

with
m
M
Mk (K) = Mk (Γi )
i=1
a sum of spaces of modular forms for congruence subgroups Γi ⊆ GL2 (Z).
• Similarly:
Sk = lim Sk (K)
−→
K
with Sk (K) a sum of spaces of cusp forms for congruence subgroups.
• Today: How can one characterize the images
Φ(Mk ) ⊆ C ∞ (GL2 (Q)\GL2 (A))
resp.
e k ) ⊆ L2 ([GL2 ])?
Φ(S
.
• Next time: How does Sk decompose as a GL2 (Af )-representation?
The construction of Φ from the last lecture:
• We considered the GL2 (R)-equivariant line bundles
Lk → H± ∼
= GL2 (R)/C×
associated to the representation z 7→ z k of C× = AGL2 SO2 (R).
• By pullback we obtained line bundles ω ⊗k on GL2 (Q)\(GL2 (Af ) × H± ).
• Smooth, holomorphic,... sections of ω ⊗k identify with smooth, holomor-
phic,... functions
f : GL2 (Af ) × H± → C, (g, z) 7→ f (g, z)
satisfying
f (γg, γz) = j(γ, z)k f (g, z)
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM25

for  
a b
γ= ∈ GL2 (Q),
c d
and
j(γ, z) := cz + d.
• The function
 
± a × b
j : GL2 (R) × H → C , ( , z) 7→ cz + d.
c d
satisfies the “cocycle equation”
j(gh, z) = j(g, hz)j(h, z)
for all g, h ∈ GL2 (R), z ∈ H± .
• For f : GL2 (Af ) × H± → C as above the function
ϕf : GL2 (Af ) × GL2 (R) → C, (g, g∞ ) 7→ j(g∞ , i)−k f (g, g∞ i)
is GL2 (Q)-equivariant, i.e., lies in C ∞ (GL2 (Q)\GL2 (A)).
• We obtain by Lemma 3.2 that the map f 7→ ϕf identifies smooth sections
GL2 (Q)\GL2 (Af ) × H± → ω ⊗k
with smooth functions
ϕ : GL2 (Q)\GL2 (A) → C
satisfying
ϕ(g, g∞ z) = z −k ϕ(g, g∞ )
for all (g, g∞ ) ∈ GL2 (Af ) × GL2 (R), z ∈ C× .
Upshot and next aims:
• Describe the image of Mk in C ∞ (GL2 (Q)\GL2 (A)) under f 7→ ϕf by con-
ditions on ϕ ∈ C ∞ (GL2 (Q)\GL2 (A)).
• Imposing the condition of cuspidality will then determine the image of Sk
under
e : Sk → L2 ([GL2 ])
Φ
−k/2
as those smooth cuspidal functions such that |det|adélic lies in the image of
Φ.
• We already know that an element ϕ in the image of Mk must satisfy
ϕ(g, g∞ z) = z −k ϕ(g, g∞ )
for z ∈ C× and (g, g∞ ) ∈ GL2 (A) and that such ϕ descent to a smooth
function
fϕ : GL2 (Af ) × H± → C, (g, z) 7→ f (g, z)
satisfying modularity for GL2 (Q).
• Need to check:
– When is fϕ holomorphic?
– When is fϕ holomorphic at the cusps?
26 JOHANNES ANSCHÜTZ

When is fϕ holomorphic? This reduces to the following question: Given a

smooth function
f : H± → C.
Which condition on
ϕf : GL2 (R) → C, g 7→ j(g, i)−k f (gi)
guarantees that f is holomorphic? We outline the strategy we will follow:
• The Lie algebra
g := gl2 (R) ∼
= Mat2,2 (R)
of GL2 (R) acts on C ∞ (GL2 (R)) by deriving the action of GL2 (R) by right
translations on GL2 (R).
• Make this action explicit on ϕf .
• Then construct an element Y ∈ gC such that
Y ∗ ϕf = 0
if and only if f is holomorphic. Here gC := g ⊗R C is the complexified Lie
algebra of g.
Infinitesimal actions (cf. [GH19, Section 4.2]): We start to explain what we
meant by ”deriving the GL2 (R)” action.
• Let H any Lie group with Lie algebra h := T1 H. Here T1 H denotes the tan-
gent space of the real manifold H at the identity element, i.e., the elements
of h = T1 H are equivalences of paths
γ : (−ε, ε) → H
with ε > 0, γ(0) = 1, and two paths are equivalent if they have the same
derivative at 1 (in some chart).
• Let V be any representation of H, where V a Hausdorff topological C-vector
space.
• For X ∈ h, defined by some path γ : (−ε, ε) → H with γ(0) = 1, and v ∈ V
consider the function
γ(t)v − v
(−ε, ε) \ {0} → V, t 7→ .
t
If the limit of this function for t → 0 exists, we write
γ(t)v − v
X ∗ v := lim .
−→ t
t→0

• An element v ∈ V is called smooth if

X1 ∗ (. . . ∗ (Xn ∗ v)) . . .)
exists for all X1 , . . . , Xn ∈ h.
• We call Vsm ⊆ V the subspace of smooth vectors in V .
• Then Vsm is stable under H (but maybe different from V ).
• Moreover, Vsm is a representation of h, or equivalently a module under the
enveloping algebra for h:
M
U (h) := h⊗n /hX ⊗ Y − Y ⊗ X − [X, Y ] | X, Y ∈ hi.
n≥0

• Here: [−, −] the Lie bracket of h.

 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM27

The explicit action of g = gl2 (R) on ϕf : We now make the action of gl2 (R) on
some function
ϕf (g) := j(g, i)−k f (gi)
associated to some section
f : H± → OH± (k)
of OH± (k) more explicit. For this let us recall the equation
j(gh, z) = j(g, hz)j(h, z)
for g, h ∈ GL2 (R), z ∈ H± .
• We already know
ϕf (gz) = z −k ϕf (g)
for z ∈ C× , g ∈ GL2 (R) (by Lemma 3.2). Note that
   
1 0 0 −1
C∼= Lie(C× ) = h , iR ⊆ g.
0 1 1 0
By deriving we can conclude that C acts on ϕf via the linear form
C 7→ C, z 7→ −kz.
Thus, more precisely
 
1 0
∗ ϕf = −kϕf
0 1
and  
0 −1
∗ ϕf = −kiϕf ,
1 0
 
0 −1
because corresponds to i ∈ C.
1 0
• Consider now the subgroup
 
∗ ∗
U := ⊆ GL2 (R)
0 1
with associated Lie algebra
   
1 0 0 1
Lie(U ) = h , i ⊆ g.
0 0 0 0 R
• Define the auxiliary function
ϕ̃f : GL2 (R) × H± → C, (g, z) 7→ j(g, z)−k f (gz).
(do not confuse it with ϕ
ef which was introduced in the last lecture).
• For a fixed g ∈ GL2 (R) we obtain
∂ ϕ̃f ∂f
(g, z) = det(g)j(g, z)−k−2 (gz)
∂z ∂z
∂ det(g)
(using that the function z 7→ gz has ∂z -derivative j(g,z)2 .)
• Thus, f is holomorphic if and only if
∂ ϕ̃f
(g, i) = 0
∂z
for all g ∈ GL2 (R).
28 JOHANNES ANSCHÜTZ

• For any path γ : (−ε, ε) → U with γ(0) = 1 we can calculate

ϕf (gγ(t))
= ϕ̃f (gγ(t), i)
= j(gγ(t), i)−k f (gγ(t)i)
= j(g, γ(t)i)−k j(γ(t), i)−k f (gγ(t)i)
= ϕ̃f (g, γ(t)i)
because j(γ(t), i) = 1.
• Set
∂γ
Xγ := (0) ∈ Lie(U ).
∂t
• Then
Xγ ∗ ϕf
∂
= ∂t (ϕf (gγ(t))|t=0
∂ ϕ̃f ∂ ∂ ϕ̃f ∂
= ∂z (g, i) ∂t (γ(t) · i)|t=0 + ∂z (g, i) ∂t (γ(t) · i)|t=0 .
using the chain rule for Wirtinger derivatives.
• More concretely, consider as a first path the function
 
1+t 0
γ1 : (−ε, ε) → U, t 7→ .
0 1
Then γ1 (t) · i = (1 + t)i and thus
 
1 0 ∂ ϕ̃f ∂ ϕ̃f
∗ ϕf (g) = i (g, i) − i (g, i).
0 0 ∂z ∂z
• As a second path we take
 
1 t
γ2 : (−ε, ε) → U, t 7→ ,
0 1
then γ1 (t) · i = i + t and thus
 
0 1 ∂ ϕ̃f ∂ ϕ̃f
∗ ϕf (g) = (g, i) + (g, i).
0 0 ∂z ∂z
• In particular,
   
1 0 0 1 ∂ ϕ̃f
i ∗ ϕf (g) + ∗ ϕf (g) = 2 (g, i)
0 0 0 0 ∂z
• Thus  
i 1
∈ gC .
0 0
 
i 1
would be a candidate for Y , i.e., f is holomorphic if and only if ∗
0 0
ϕf = 0.
• But we can do better. Namely, define
 
1 −1 i
Y := 2
i 1
     .
1 1 0 1 0 −1 i 1
=2 + 2i +i
0 1 1 0 0 0
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM29

Then
k ki ∂ ϕ̃f ∂ ϕ̃f
Y ∗ ϕf (g) = − ϕf (g) + i(− )ϕf (g) +2i (g, i) = 2i (g, i),
| 2 {z 2 } ∂z ∂z
=0
and thus f is holomorphic if and only if Y ∗ ϕf = 0.
• Why is this choice better?
– Set    
0 −i 1 −1 −i
H := , X := .
i 0 2 −i 1
– Then H, X, Y is an sl2 -triple in gC .
– Namely, in the basis
   
−i i
, ,
1 1
it is given by
     
1 0 0 1 0 0
, , .
0 −1 0 0 1 0
– Note that H spans the Lie algebra iLie(SO2 (R)) ⊆ gC .
We can pass to the next question.
When is fϕ holomorphic at the cusps?
• Consider a function in Fourier expansion
X
f (z) = an e2πizn ,
n∈Z
defined on
{z ∈ C | |Re(z)| < 1, Im(z) > 1}.
• Recall that
|e2πiz | = e−2πy
with y = Im(z).
• Then an = 0 for all n < 0 if and only if
|f (x + iy)| ≤ Cy N
for some C, N ∈ R>0 and y → ∞, because for n < 0 the exponential
function e−2πny grows faster than any polynomial function.
• By definition this means f is of moderate growth, or slowly increasing, a
notion which can be generalized to all reductive group over Spec(Q).
• Let us note that moreover, an = 0 for all n ≤ 0, i.e., including n = 0, if
and only if for all N ∈ R>0
|f (x + iy)| ≤ Cy −N
for some C ∈ R>0 and y → ∞.
• In other words, f is rapidly decreasing.
Moderate growth for general reductive groups: The moderate growth con-
dition can be phrased for arbitrary reductive groups G over Q as follows
• For each embedding ρ : G ,→ SLn,Q (not into GLn,Q ) we obtain a norm
|g| := sup max {|ρ(g)i,j v |v }
v i,j=1,...,n

on G(A), where v runs through all primes and ∞.

30 JOHANNES ANSCHÜTZ

• A function ϕ : G(A) → C is of moderate growth or slowly increasing if

|ϕ(g)| ≤ C|g|N
for suitable C, N ∈ R>0 , cf. [GH19, Definition 6.4.].
• It is not difficult to see that this definition does not depend on the embed-
dding ρ (as this changes the norm only up to some polynomial functions).
We leave it as an exercise (cf. [GH19, Lemma 6.3.1.] and our above remark on
Fourier coefficients) to check that a (holomorphic) section
f ∈ H 0 (GL2 (Q)\(GL2 (Af ) × H± ), ω ⊗k ).
is holomorphic at the cusps if and only of the function ϕf ∈ C ∞ (GL2 (Q)\GL2 (A))
is of moderate growth.
We give a short summary of what we have done.
Upshot:
• We can describe the image of
Mk → C ∞ (GL2 (Q)\GL2 (A)), f 7→ ϕf .
• Namely, ϕ ∈ C ∞ (GL2 (Q)\GL2 (A)) lies in the image if and only if
– ϕ(g, g∞ z) = z −k ϕ(g, g∞ z) for (g, g∞ ) ∈ GL2 (A), z ∈ C× .
– ϕ is of moderate growth.
– For each g ∈ GL2 (Af )
Y ∗ ϕ(g, −) = 0,
where Y ∈ gC = gl2 (R)C is the element constructed before.
• Note that by our discussion these conditions match the conditions ”mod-
ularity, holomorphic at cusps, holomorphic” in the definition of a modular
form.
• For k ≥ 1 it is not difficult to see that there exists a unique, up to isomor-
0
phism, irreducible (gC , O2 (R))-module Dk−1 containing a non-zero element
v, such that Y ∗ v = 0, and C acts on v via the character z 7→ z −k , cf.
×
0
[Del73, Section 2.1] (where the dual of Dk−1 is described).
0
• Using Dk−1 one can rewrite our result as saying that

Mk ∼ 0
= Hom(gC ,O2 (R)) (Dk−1 ∞
, Cmg (GL2 (Q)\GL2 (A))),
∞
where Cmg (GL2 (Q)\GL2 (A)) ⊆ C ∞ (GL2 (Q)\GL2 (A)) denotes the subset
of functions with moderate growth.
• More on this will appear later when we discuss automorphic forms and
(gC , K∞ )-modules, cf. Section 11.
We finish the lecture by sketching how to incorporate cuspidality.
A condition for cuspidality:
• Let ϕ = ϕf ∈ C ∞ (GL2 (Q)\GL2 (A)) for some f ∈ Mk , seen as a function
on GL2 (Af ) × H± → C satisfying modularity.
• Then f ∈ Sk if and only if the integral
Z
ϕB (g) := ϕ(ng)dn
N (Q)\N (A)
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM31

vanishes for all g ∈ GL2 (A), where

 
1 ∗
N=
0 1
is the unipotent radical in the standard Borel
 
∗ ∗
B=
0 ∗
and dn an N (A)-invariant measure on N (Q)\N (A).
• Namely:
– We may assume that f , and thus ϕ = ϕf , is invariant under some
principal congruence subgroup K(m) ⊆ GL2 (Af ).
– Consider g = 1. Then using N (Q)\N (A) ∼ = (mZ b × R)/mZ:
R
ϕf (n)dn
N (Q)\N
R (A)
= j(n∞ , i)−k f (n, n∞ i)d(n, n∞ )
N (Q)\N
R R (A)
= f (a, i + x)dadx
R/mZ mZ b
R
= µ(mZ) b f (1, i + x)dx
R/mZ
= µ(mZ)a
b 0

with µ(mZ)
b the volume of mZ
b and
∞
X
f (1, z) = an q n/m
n=0

the Fourier expansion of the modular form f (1, −) for Γ(m) at the
cusp ∞.
– Indeed: The crucial point is that
Z
n
e2πi m x dx = 0
R/mZ

if n > 0 (which is part ofR Fourier theory, and follows more generally
from the statement that χ(g)dg = 0 for any compact, abelian group
A
A and any non-trivial character χ : A → C× ).
– The condition ϕB (g) = 0 for all g ∈ GL2 (A) is then equivalent to the
vanishing of the constant Fourier coefficients at all cusps.
– We leave the details as an exercise, cf. [Del73, Rappel 1.3.2.], and
[Gel75, Proposition 3.1.(vii)].
• Note: Because ϕ is GL2 (Q)-invariant for left translations, we can replace
B by any conjugate gBg −1 with g ∈ GL2 (Q).
Let us note that the way we associate functions on GL2 (A) is different in for-
mulation than in the standard references, e.g., [Gel75], which usually restrict to
modular forms on
 
a b
Γ1 (m) := { ∈ SL2 (Z) | c ≡ 0, d ≡ 1 mod m}
c d
32 JOHANNES ANSCHÜTZ

or to modular forms with nebentype χ : (Z/m)× → C× for

 
a b
Γ0 (m) := { ∈ SL2 (Z) | c ≡ 0 mod m}.
c d
Although these capture all automorphic representations generated by modular
forms, we think it is a bit unmotivated to restrict to these congruence subgroups
(or their counterparts K1 (m), K0 (m) ⊆ GL2 (Af ) in this moment). We will clarify
the relation of our approach with the classical ones in the next lecture.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM33

5. From modular forms to automorphic representations, part III

Last time:
• Let ϕ ∈ C ∞ (GL2 (Q)\GL2 (A)).
• Then ϕ = ϕf for some f ∈ Mk if and only if
– ϕ(g, g∞ z) = z −k ϕ(g, g∞ ) for (g, g∞ ) ∈ GL2 (A), z ∈ C× .
– ϕ is of moderate growth.
– For each g ∈ GL2 (Af )
Y ∗ ϕ(g, −) = 0,
where Y ∈ gC = gl2 (R)C is a suitably constructed, natural element.
• f ∈ Mk is a cusp form if and only if for all B ⊆ GL2,Q proper parabolic,
i.e., Borel, with unipotent radical N ⊆ GL2,Q :
Z
ϕf (ng)dn = 0
N (Q)\N (A)

for all g ∈ GL2 (A).

Today:
• Describe Sk as a GL2 (Af )-representation.
• This will yield our main source of examples for automorphic representa-
tions.9
Considering the whole spaces Mk , Sk (and not merely Mk (K), Sk (K) for some
compact-open subgroup K ⊆ GL2 (Af ), or even Mk (Γ), Sk (Γ) for Γ ∈ SL2 (Z) a
congruence subgroup) has the advantage that one now can use representation theory
for the locally profinite group GL2 (Af ). In this vein the following observation is
important.
Observation:
• Mk and Sk are smooth and admissible GL2 (Af )-representations in the fol-
lowing sense.
Definition 5.1. Let G be a locally profinite group (like GL2 (Af )). A representation
G on a C-vector space V is called smooth if
[
V = V K,
K⊆G

where K runs through the compact-open subgroups of G and

V K := {v ∈ V | kv = v for all k ∈ K}.
Equivalently, V is smooth if the action morphism
G×V →V
is continuous with V carrying the discrete topology. Denote by
Rep∞
C G

the (abelian) category of smooth representations of G on C-vector spaces.

9More precisely, the irreducible subrepresentations of the closure of Φ(S

e k ) ⊆ L2 ([GL2 ]) are
the automorphic representations (in the L2 -sense) associated with cusp forms.
34 JOHANNES ANSCHÜTZ

Definition 5.2. Let G be a locally profinite group. A smooth representation V of

G is called admissible if
dimC V K < ∞
for all K ⊆ G compact-open.
Let us consider some examples
MkK with MkK = Mk (K) is smooth and admissible because
S
• Mk =
K⊆GL2 (Af )
the Mk (K) are finite dimensional.
• Similarly: Sk is an admissible representation of GL2 (Af ).
• For G reductive over Q ⇒ C ∞ (G(Q)\G(A)) is a smooth G(Af )-representation,
but not admissible.
• L2 ([G]) is not a smooth G(Af )-representation.
In Section 11 we will introduce the space of automorphic forms for a reductive
group G over Q which serves as a replacement (actually a huge generalization)
for the space Mk , which is again a smooth G(Af )-representation. As we will see
smooth representations are more algebraic in nature than unitary representations,
like L2 ([G]). This makes smooth representations of locally profinite groups more
convenient to work with.
In the classical theory of modular forms Hecke operators (which might seem
to be some ad hoc definition) are important. From the representation-theoretic
perspective they arise from the action of an Hecke algebra.
The Hecke algebra (of a locally profinite group):
• G a locally profinite group.
– e.g., GL2 (Af ), GLn (Qp ), Gal(Q/Q), . . .
• For K ⊆ G compact-open, there exists a unique left-invariant resp. a unique
right-invariant Haar measure µ on G with µ(K) = 1.
• Concretely:
– By translation invariance it is sufficient to define µ for K 0 ⊆ G a
compact-open subgroup.
– Define K 00 := K 0 ∩ K compact-open. Then K 00 is of finite index in K
resp. K 0 .
– Set
[K 0 : K 00 ]
µ(K 0 ) := .
[K : K 00 ]
– This defines the desired measure and is also the unique possible choice.
– Note that µ takes actually values in Q. This is a first hint that the
theory of smooth representations of locally profinite groups is ”more
algebraic”.
• From now on, fix a left invariant Haar measure µ on G.
• Define the associated “Hecke algebra”
H(G) := Cc∞ (G)
of G as the set of locally constant, compactly supported functions G → C.
• H(G) is an algebra via convolution:
Z
f ∗ g(x) := f (xy)g(y −1 )dy
G
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM35

for f, g ∈ H(G). Note that H(G) depends implicitly on the chosen mea-
sure. However, it is possible to define H(G) intrinsically as the space of
compactly supported distributions with multiplication given by convolu-
tion of distributions, cf. [BRa, Definition 7].
• The multiplication in H(G) is a purely algebraic operation:
– Each f ∈ H(G) is a finite sum
Xm
f= ai χgi Ki
i=1
for some gi ∈ G, Ki ⊆ G compact-open subgroups.
– Wlog: K1 = . . . = Km (by shrinking).
– Let K ⊆ G be compact-open and g1 , g2 ∈ G. Then
m
a
g1 Kg2 K = hj K
j=1

for some hj ∈ G and

m
X
χg1 K ∗ χg2 K = µ(K)χhj K .
j=1

• The algebra H(G) has a lot of idempotents, namely,

1
eK := χK
µ(K)
with K ⊆ G compact-open subgroups, χK the characteristic function of K.
But in general H(G) has no identity (unless G is compact).
Modules for the Hecke algebra (cf. [BRa]):
• H(G) is a replacement for the group algebra C[G], which is better suited
to study smooth representations of G.
• If V ∈ Rep∞
C G, then
Z
f ∗ v := f (h)hvdh
G
for f ∈ H(G), v ∈ V , defines an action of H(G) on V .
• Concretely: If K ⊆ G is a compact-open subgroup and sufficiently small,
m
P
such that v ∈ V is fixed by K and f = ai χgi K with ai ∈ C, gi ∈ G, then
i=1
m
X
f ∗v = ai gi v.
i=1
• Assume K ⊆ G is a compact-open subgroup. We define the Hecke algebra
of G relative to K
H(G, K) := eK H(G)eK .
Then H(G, K) is an algebra with unit eK .
• Let V be any smooth representation of G. Then it is not difficult to see
that
V K = eK ∗ V
for any K ⊆ G compact-open.
36 JOHANNES ANSCHÜTZ

• Applying this to the smooth G-representation H(G) (via left and right
multiplication) we obtain that
H(G, K) ∼= C ∞ (K\G/K),
c
i.e., H(G) identifies with the set of K-biinvariant, compactly supported
functions on G.
The following small statement is important in our upcoming discussion of the
GL2 (Af )-representation Sk .
• If V ∈ Rep∞ C G is irreducible and V
K
6= 0, then V K is irreducible as an
H(G, K)-module.
– Indeed: If M ⊆ V K is a non-trivial H(G, K)-submodule, then
V = H(G) ∗ M
by irreducibility V and thus
V K = eK (H(G) ∗ M )) = eK H(G)eK M = M,
because eK ∗ M = M .
• Conversely one can show that if V K is irreducible or zero for all compact-
open subgroups K ⊆ G, then V is irreducible (if V 6= 0).
Namely, this property allows as to associate Hecke eigensystems to irreducible,
admissible GL2 (Af )-representations.
From irreducible, admissible GLn (Af )-representations to Hecke eigensys-
tems:
• Let V be an irreducible, admissible GLn (Af )-representation.
• Recall that the compact-open subgroups
K(m) = ker(GLn (Z) b → GLn (Z/m)), m ≥ 0,
form basis of compact-open neighborhoods of the identity in GL2 (Af ).
• From the important property and admissibility (which easily implies a ver-
sion of Schur’s lemma) we know that
– V K(m) is an irreducible H(GLn (Af ), K(m))-module for m  0.
– EndH(GLn (Af ),K(m)) (V K(m) ) ∼
= C.
• In the following we will use that for each prime p a smooth GL2 (Af )-
representation yields by restricting along the inclusion GL2 (Qp ) ⊆ GL2 (Af )
a smooth GL2 (Qp )-representation.
• Set \
S := {p prime dividing m},
m≥0, V K(m) 6=0

• Note that p ∈ S if and only if V GL2 (Zp ) = 0.

• We call a prime p unramified for V if p ∈ / S, i.e., if V GL2 (Zp ) 6= 0.
• The following observation is very important:
For each prime p the algebra H(GL2 (Qp ), GL2 (Zp )) is commutative!
• Even more is true: Take m ≥ 1 and p - m. Then H(GL2 (Qp ), GL2 (Zp ))
is central in H(GL2 (Af ), K(m)). More precisely, to a double coset of
some A ∈ GL2 (Qp ) one can associate the double coset of the element
(1, . . . , 1, A, 1, . . . , 1) ∈ GL2 (Af ) whose entry at p is A, and the identity
elsewhere, and the resulting embedding
H(GL2 (Qp ), GL2 (Zp )) → H(GL2 (Af ), K(m))
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM37

has image in the center of H(GL2 (Af ), K(m)). Namely, it is easy to see
that the double coset given by
(1, . . . , 1, A, 1, . . . , 1)
with A ∈ GL2 (Qp ) commutes with double cosets of B ∈ GL2 (Af ) if the
component of B at p is the identity matrix. Thus, centrality of H(GL2 (Qp ), GL2 (Zp ))
follows actuality from commutativity of H(GL2 (Qp ), GL2 (Zp )).
• To prove the commutativity we use Gelfand’s trick. Namely, consider the
antiinvolution
σ : GLn (Qp ) → GLn (Qp ), g 7→ g tr ,
which has the following properties:
– σ preserves GLn (Z b S ).
– σ(f1 ∗ f2 ) = σ(f2 ) ∗ σ(f1 ) for f1 , f2 ∈ H(GL2 (Qp ), GL2 (Zp )).
– The cosets GLn (Zp )gGLn (Zp ) with g ∈ GLn (Qp ) a diagonal matrix
span H(GL2 (Qp ), GL2 (Zp )) (by the elementary divisor theorem).
– Now we can prove commutativity. Namely, σ(f ) = f for all f ∈
H(GL2 (Qp ), GL2 (Zp )) and thus
f1 ∗ f2 = σ(f1 ∗ f2 ) = σ(f2 ) ∗ σ(f1 ) = f2 ∗ f1
as desired.
• The centrality of H(GL2 (Qp ), GL2 (Zp )) is important, as it allows to define
the system of Hecke eigenvalues of V (which we recall is an irreducible,
admissible GL2 (Af )-representation).
• Recall that S denotes the set of ramified primes of V , i.e., those primes p
such that V GL2 (Zp ) = 0.
• For p ∈
/ S we will now construct a morphism
σV,p : H(GL2 (Qp ), GL2 (Zp )) → C
of C-algebras, which is naturally assocatiated with V .
• For this pick p ∈
/ S, and choose v ∈ V fixed by GL2 (Zp ). Then v is fixed by
K(m) for some m ≥ 0 with p - m (by smoothness of the GL2 (Af )-action).
• As we discussed before irreducibility of V implies that V K(m) is irreducible,
and thus
EndH(GL2 (Af ),K(m)) (V K(m) ) ∼
=C
by Schur’s lemma.
• As H(GLn (Qp ), GLn (Zp )) is central in H(GL2 (Qp ), GL2 (Zp )) its acts via a
homomorphisms of H(GL2 (Af ), K(m))-modules on V K(m) .
• Thus we obtain our morphism
σV,p : H(GLn (Qp ), GLn (Zp )) → EndH(GLn (Af ),K(m) (V K(m) ) ∼
= C,
which a priori dependents on v and m. But it is not difficult to see σV,p is
invariant under enlarging m. This implies that σV,p is actually independent
of v and m, and hence canonically associated with V and p.
Definition 5.3. The system
{σV,p : H(GL2 (Qp ), GL2 (Zp ) → C}p∈S
/

is called the system of Hecke eigenvalues for V .

38 JOHANNES ANSCHÜTZ

We present a slightly different perspective to systems of Hecke eigenvalues.

Namely, let S be some finite set of primes, and define the ”Hecke algebra away
from S” as
TS := H(GL2 (ASf ), GL2 (Z
b S )),
where Y
b S :=
Z Zp
p∈S
/

and
ASf := Q ⊗Z Z
bS .
Using Gelfand’s trick one can see that TS is commutative. Moreover, if one devel-
opes the appropriate definition (cf. [Fla79]) the ”restricted tensor product”
0
TS ∼
O
= H(GL2 (Qp ), GL2 (Zp )).
p∈S
/

Now, we can also make the following definition.

Definition 5.4. A Hecke eigensystem, or system of Hecke eigenvalues, is a maxi-
mal ideal in TS where S is a finite set of primes and TS := H(GLn (ASf ), GLn (Z
b S )).

We can collect the following observations.

• TS is of countable dimension and thus each Hecke eigensystem has residue
field C.
• A Hecke eigensystem for S is equivalently a collection of C-algebra homo-
morphisms
σp : H(GLn (Qp ), GLn (Zp )) → C
for p ∈
/ S as before.
• If S ⊆ S 0 are two finite sets of primes, each Hecke eigensystem for S induces
one for S 0 . We call two Hecke eigensystems
{σp }p∈S
/ and {σp0 }p∈S
/ 0

equivalent if they agree outside some finite set of primes which contains S
and S 0 .
• As we saw before each irreducible, admissible GLn (Af )-representation yields
by the above construction a system of Hecke eigenvalues.
We will see that in the Langlands program the system of Hecke eigenvalues
associated to modular forms is expected to being arithmetically interesting.
To gain a more useful perspective on Hecke eigensystems, we need a full descrip-
tion of H(GL2 (Qp ), GL2 (Zp )).
A full description of H(GL2 (Qp ), GL2 (Zp )):
• For moment set G := GL2 (Qp ) and K = GL2 (Zp ).
• We normalize the Haar measure on G such that µ(K) = 1.
• By the elementary divisor theorem the C-vector space H(G, K) is free on
the elements  i 
p 0
K K
0 pj
with (i, j) ∈ Z2,+ := {(i, j) ∈ Z2 | i ≥ j}.
LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM39

• We claim that
'
f : C[X1 , X2±1 ] → H(G, K)
via the morphism
 
p 0
X1 7→ K K,
0 1
p 0
X2 7 → K K.
0 p
Here (and in the following) we identify a double coset with its characteristic
function.
• Let us verify the claim (without invoking Gelfand’s trick):
– The element  
p 0
K K
0 p
is central in H(G, K) as
KzK · KgK = KgK · KzK
for all g ∈ G, and z ∈ G in the (group-theoretic) center.
– This implies that f is well-defined (note that C[X1 , X2±1 ] is commuta-
tive, but H(GL2 (Qp ), GL2 (Zp )) a priori not.
– Moreover, let us check that f is surjective. By multiplying with f (X2 )
it suffices to see that each
 i 
p 0
K K
0 1
with i ≥ 1 lies in the image of f .
– Using induction one sees that for n ≥ 1 the n-fold product
   
p 0 p 0
K K · ... · K K
0 1 0 1
agrees with the set of matrices Mat2,2 (Zp )val=n in Mat2,2 (Zp ) of de-
terminant of valuation n.
– From the definition of the product in H(G, K) one concludes that the
product    
p 0 p 0
K K · ... · K K
0 1 0 1
is the characteristic function χ of Mat2,2 (Zp )val=n .
– As this characterstic function χ has the characteristic function of
 n 
p 0
K K
0 1
as a summand with coefficient 1 we can conclude (using a small induc-
tion) that it lies in the image of f , because each other double coset
occuring in Mat2,2 (Zp )val=0 has a representative a matrix of the form
 j 
p 0
0 pi
with i, j ≥ 1 and i + j = n.
– In particular, we see that H(G, K) must be commutative without using
Gelfand’s trick.
40 JOHANNES ANSCHÜTZ

– Now it is not difficult to conclude that f must be an isomorphism.

Indeed,
Spec(H(G, K)) ⊆ Spec(C[X1 , X2±1 ])
defines a subscheme whose projection to Spec(C[X2±1 ]) is free of infinite
rank, but this is by only possible for Spec(C[X1 , X2±1 ]) itself (using
integrality of the latter).
We summarize our discussion on systems of Hecke eigenvalues.
Upshot:
• Each irreducible, admissible GL2 (Af )-representation V yields canonically
a system of Hecke eigenvalues
ãp ∈ C, b̃p ∈ C×
for all primes p outside the finite set S of primes, which are ramified for V .
• Concretely: If v ∈ GL2 (Af ) is fixed by K(m) and p - m, then
 
p 0
T̃p (v) := GL2 (Zp ) GL2 (Zp ) ∗ v = ãp v,
0 1

where Tep is “the Hecke operator at p”, and

 
p 0
S̃p (v) := GL2 (Zp ) GL2 (Zp ) ∗ v = b̃p v.
0 p
• Note that the system of Hecke eigenvalues is completely determined by the
collection of complex numbers
{ãp , b̃p }p∈S
/ .

• Let Z ⊆ GL2 be the center, i.e., the subgroup of scalar matrices.

• Let ψ : A× ∼ ×
f = Z(Af ) → C be the central character of V (which exists by
Schur’s lemma). Then
b̃p = ψ((1, . . . , 1, p, 1, . . . , 1)
×
for p prime such that Z× p ⊆ Af acts trivially on V .
• The central character ψ yields another invariant of the irreducible, admis-
sible GL2 (Af )-representation.
• Note that ψ is in general not determined by the systems of Hecke eigen-
values, because finitely many p are missing and no triviality on Q× , i.e.,
“automorphicity”, is required.
We now use our discussion to obtain a decomposition of the smooth, admissible
GL2 (Af )-representation Sk . First, we collect some general observations before we
obtain the precise decomposition.
General results on the GL2 (Af )-representation Sk :
• The GL2 (Af )-representation Sk decomposes into a direct sum

Sk ∼
M
= Vi
i∈I

of irreducible, admissible representations Vi .

• Indeed:
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM41

k/2
– Up to twisting by the character |det|adélic|GL2 (Af ) we have a GL2 (Af )-
equivariant embedding Sk ⊆ L2 ([GL2 ]) and thus the C-vector space
k/2
Sk ⊗|det|adélic is equipped with the non-degenerate GL2 (Af )-equivariant
hermitian pairing
Z
(·, ·) : Sk × Sk → C, (f, g) 7→ ϕ
ff ϕeg .
[GL2 ]

– This implies by passing to orthogonal complements that each irre-

ducible GL2 (Af )-subquotient of Sk is already a direct summand.
– Using Zorn’s lemma each smooth non-zero GL2 (Af )-representation V
has an irreducible subquotient. Indeed, without loosing generality one
may assume that V is generated by some v ∈ V . Then one applies
Zorn’s lemma to the set of submodules not containing v.
We need the following result by Shalika and Piatetski-Shapiro.
Theorem 5.5 (Shalika/Piatetski-Shapiro). For each finite set S of primes and
each system of Hecke eigenvalues {ãp , b̃p }p∈S
/ there exists at most one irreducible
subrepresentation V ⊆ Sk with this (equivalence class of ) system of Hecke eigen-
values.
• This is a special case of the ”strong multiplicty one theorem” for cuspidal
automorphic representations for GLn , cf. [GH19, Theorem 11.7.2].10
• We can record the following consequence: Assume that 0 6= f ∈ Sk is fixed
by K(m), and an eigenvector for the T̃p , S̃p −operators for p - m. Then the
GL2 (Af )-representation
V := hf iGL2 (Af )
generated by f is irreducible.
– Namely, each non-zero v ∈ V yields a system of Hecke eigenvalues,
which is equivalent to the one for f . By strong multiplicity one, they
have to lie in the same subspace.
Now we will relate the Hecke operators T̃p , S̃p to their classical counterparts
acting on modular forms, and describe the decomposition of Sk via the newforms
of Atkin/Lehner.
The decomposition of Sk indexed by newforms:
• By a result of Atkin-Lehner/Casselman (cf. [Cas73]) for each irreducible
GL2 (Af )-subrepresentation V ⊆ Sk , there exists some m ∈ N and some
0 6= v ∈ V such that v is fixed by the subgroup
 
a b
K1 (m) := { | c ≡ 0, d ≡ 1 mod m} ⊆ GL2 (Z)
b
c d
• Set
Γ1 (m) := GL2 (Z) ∩ K1 (m).
• Remarks:

10More precisely, we deduce it from the mentioned stronger result for L2

cusp ([GL2 ]) via the
k/2
density of cuspidal automorphic forms, [GH19, Theorem 6.5.1.], and our embedding Sk ⊗|det|adélic .
42 JOHANNES ANSCHÜTZ

– Our notation is different than in many sources, where Γ1 (m) ⊆ SL2 (Z).
This clash of notation is harmless as
Γ1 (m)\H± = ∼ (Γ1 (m) ∩ SL2 (Z))\H.
– The minimal m for which there exists such a non-zero v ∈ V is called
the conductor of V . In this case, v is unique up to a scalar (and called
the newform of the representation).
– More details can be found in [Del73].
• We noted already that the morphism
'
Γ1 (m)\H± −→ GL2 (Q)\(GL2 (Af )/K1 (m) × H± )
[z] 7→ [(1, z)]
is an isomorphism in Section 3. Of course, the isomorphism is compatible
with ω ⊗k .
• In particular,
Mk (Γ1 (m)) ∼= Mk (K1 (m))
(and both agree with the space of classical modular forms for Γ1 (m) ∩
SL2 (Z)).
• Let us make this isomorphism more explicit and pick f˜ ∈ Mk (K1 (m)).
• Let f ∈ Mk (Γ1 (m)) be the section corresponding to f˜.
• Then
f˜(g, z) = f˜(γ, z) = j(γ −1 , z)−k f (γ −1 z),
where g = γh with γ ∈ GL2 (Q), h ∈ K1 (m).
• Note that we are using that GL2 (Q)K1 (m) = GL2 (Af ).
• Fix some prime p with p - m.
• We have the two operators T̃p , S̃p acting on Mk (K1 (m)). Thus, by transport
of structure they act on Mk (Γ1 (m)).
• Let us make this action of S̃p on Mk (Γ1 (m)) explicit.
• For a matrix γ ∈ GL2 (Q) we denote by
γp ∈ GL2 (Af )
the element whose component at p is γ ∈ GL2 (Q) ⊆ GL2 (Qp ), and the
identity matrix otherwise. This element should not be confused with
γdiag ∈ GL2 (Af ),
by which we mean the element obtained by diagonally embedding GL2 (Q) ⊆
GL2 (Af ).
• The action of S̃p on Mk (K1 (m)) is easy to write down, but to describe it
on Mk (Γ1 (m) we have to find an expression
 
p 0
= γdiag h, with γ ∈ GL2 (Q), h ∈ K1 (m).
0 p p
• Because p, m are prime there exists a, b ∈ Z, such that
 
a b
A := ∈ GL2 (Z).
m p
• Then  −1   
p 0 p 0
Adiag · · ∈ K1 (m),
0 p−1 diag
0 p p
LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM43

where we used the notation (−)diag , (−)p introduced before.

• In other words, we can take
 
p 0
γ := A−1
0 p
• With this choice of γ the operator S̃p acts on Mk (Γ1 (m)) via
f 7→ (z 7→ j(γ −1 , z)−k f (γ −1 z)).
• The group
 
a b
Γ0 (m) := { | c ≡ 1 mod m}
c d
normalizes Γ1 (m) and thus there is an action of Γ0 (m)/Γ1 (m) ∼
= (Z/m)×
via
(B, f ) 7→ j(B, z)−k f (Bz).
for B ∈ Γ0 (m), f ∈ Mk (Γ1 (m)).
• Thus, we obtain a decomposition
M
Mk (Γ1 (m)) = Mk (Γ0 (m), χ)
χ : (Z/m)× →C×

with Mk (Γ0 (m), χ), by definition, the space of modular forms for Γ0 (m)
with nebentypus χ, i.e., modular forms for Γ1 (m) satisfying
 
az + b k a b
f( ) = χ(d)(cz + d) f (z) for ∈ Γ0 (m).
cz + d c d
• For each Dirichlet character χ : (Z/m)× → C× there exists a unique adélic
character
ψχ : Q× \A× /(1 + mZ)b ∩Zb × → C× ,
such that ψχ (r) = r−k for r ∈ R>0 .
• We obtain that
Mk (Γ0 (m), χ) ∼
= Mk (K1 (m), ψχ ),
where the RHS denotes the ψχ -eigenspace for the action of A× f (the center
of GL2 (Af )) on Mk (K1 (m)).
• With this terminology we can finish the description of the operator S̃p on
Mk (Γ1 (M )).
• Namely, the action of S̃p is given on the subspace Mk (Γ0 (m), χ) ⊆ Mk (Γ1 (m))
by the action of
 −1 
−1 p 0
γ = A
0 p−1
 −1 
p 0
and acts by pk and A ∈ Γ0 (m) by χ(p).
0 p−1
• Let us describe now the action (by transport of structure) of T̃p on Mk (Γ0 (m), χ).
• We have
  p−1
a p j  a 1 0
p 0
GL2 (Zp ) GL2 (Zp ) = GL2 (Zp ) GL2 (Zp ).
0 1 0 1 0 p
j=0
44 JOHANNES ANSCHÜTZ

• We again have to find factorizations of (the shown representatives of ) the

double cosets into
γ · h, γ ∈ GL2 (Q), h ∈ K1 (m).
• We can take the factorization
   
p j p j
= ·h
0 1 p 0 1 diag
for j = 0, . . . , p − 1, with h ∈ K1 (m), i.e.,
 −1
−p−1 j

−1 p
γ = .
0 1
• For  
1 0
0 p
we find, similar to the argument above,
a p−1 b
       
−1 a b 1 0 1 0
γ = · = ·
m p 0 p−1 0 p−1 pm p
as desired.
• Putting these together we obtain
p−1
X z−j
T̃p (f ) = pk χ(p)f (pz) + f( ).
j=0
p

• If the modular form with nebentypus f ∈ Mk (Γ0 (m), χ) is written in Fourier

expansion
∞
X
f (z) = an q n , q = e2πiz ,
n=0
then more explicitly
∞
X ∞
X
T̃p (f )(z) = panp q n + pk χ(p)an q pn .
n=0 n=0
• This follows from the fact
p−1
(
X
2πi −jn p, p|n
e p =
j=0
0, otherwise.

• In particular, if f ∈ Mk (Γ0 (m), χ) is an eigenvector for T̃p and a1 = 1, then

(by considering the coefficient in front of q 1 )
Tep (f ) = pap f.
This motivates to define
1
Tp = T̃p ,
p
which is the classical Hecke operator acting on modular forms (like in [DS05,
Section 5.2.]). In other words, the Hecke eigenvalues for the Tp -operators
are the Fourier coefficients of the normalized eigenform.
We can now prove the desired theorem on decomposing the GL2 (Af )-representation
Sk .
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM45

Theorem 5.6. The GL2 (Af )-representation Sk decomposes into irreducibles as

Sk ∼
M
= hf iGL2 (Af )
f

with f running through the set of (normalized) newforms for Γ0 (N ) with nebentypus
χ for varying N and Dirichlet characters χ : (Z/N )× → C× .
We don’t give the precise definition of a newform, but refer to [DS05, Section
5.8.]. Roughly, a newform is an eigenform realising the minimal N among all
eigenforms with an equivalent system of Hecke eigenvalues.
Proof. After the above preparation we can now quote [DS05, Proposition 5.8.4],
[DS05, Theorem 5.8.2], which imply that for each system of Hecke eigenvalues
appearing in Sk there exists a unique normalized (i.e., in Fourier expansion a1 = 1)
newform with equivalent system of Hecke eigenvalues. 
In other textbooks, e.g., [Gel75, Proposition 3.1], automorphic representations
are associated directly to some f ∈ Mk (Γ0 (m), χ), i.e., without introducing the
space Mk , and the GL2 (Af )-equivariant embeddings
Mk → C ∞ (GL2 (Q)\GL2 (A)), f (g, z) 7→ ϕf (g, g∞ ) = j(g∞ , i)−k f (g, g∞ i)
resp.
k/2 k/2
Sk ⊗ |det|adélic → L2 ([GL2 ]), f 7→ ϕ
ef (g, g∞ ) = |det(g, g∞ )|adélic ϕf .
We note that under the isomorphism
Mk (Γ1 (m)) ∼
= Mk (K1 (m))
mentioned above, our formulas specialize to the one in [Gel75, Proposition 3.1.].
Moreover, the maybe unexpected factor pk/2−1 appearing in [Gel75, Lemma 3.7.]
k/2
has an easy explanation by the appearance of the twist by |det|add́elic and the
passage from (our) Tep to the classical Hecke operator Tp = 1 Tep .p
46 JOHANNES ANSCHÜTZ

6. Langlands reciprocity for newforms

Last time:
• We proved the decomposition
Sk ∼
M
= hf iGL2 (Af )
f

into irreducible GL2 (Af )-representations.

• Here f is running through the set of (normalized) newforms of weight k for
Γ0 (N ) with nebentypus χ for varying N and varying Dirichlet characters
χ : (Z/N )× → C× .
• In course, of proving the decomposition we associated to each irreducible,
admissible GL2 (Af )-representation π a system of Hecke eigenvalues
{ãp (π), b̃p (π)}p∈S
/

for S := {p prime with π GLn (Zp ) = 0}. In fact, ãp (π) resp. b̃p (π) was de-
fined to be the eigenvalue of the double coset
 
p 0
GL2 (Zp ) GL2 (Zp )
0 1
resp.  
p 0
GL2 (Zp ) GL2 (Zp )
0 p
when acting on a non-zero vector fixed by GL2 (Zp ).
• If π = hf iGL2 (Af ) for a normalized newform f ∈ Sk (Γ0 (N ), χ), then S =
{prime divisors of N } and
ãp (π) = pap , b̃p (π) = pk χ(p)
∞
an q n is the Fourier expansion of f .
P
if p ∈
/ S, where f (q) =
n=1
In this lecture, we will introduce traces of Frobenii for `-adic Galois representa-
tions (which are the ”Galois-theoretic” counter part of a system of Hecke eigenval-
ues). Then we want to state what the Langlands program predicts for Sk .
Traces of Frobenii for `-adic representations:
• Let ` be some prime.
• Q` = lim E is a topological field via colimit topology.
−→
E/Q`
• Let W be a finite dimensional Q` -vector space.
• Let σ : GQ := Gal(Q/Q) → GL(W ) be a continuous representation, also
called an “`-adic Galois representation”.
• We recall that for each prime p there exists an embedding
GQp := Gal(Qp /Qp ) → GQ ,
well-defined up to conjugacy in GQ .
• The local absolute Galois group GQp sits in the exact sequence
1 → Ip → GQp → Gal(Fp /Fp ) = (Frobgeom )Z → 1
b
p
with Ip ⊆ GQp the inertia subgroup, and
Frobarith
p : Fp → Fp , x 7→ xp
LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM47

the arithmetic Frobenius at p.

−1
• The inverse Frobgeom
p := Frobarith
p is called the geometric Frobenius.
• Let S be a finite set of primes. We say σ is unramified outside S if σ(Ip ) = 1
for all p ∈
/ S.
• If σ unramified outside S and n := dimQ` W , we can associate to σ a system
{c1,p , . . . , cn,p }p∈S
/

of elements c1,p , . . . , cn,p ∈ Q` . Namely,

ci,p := Tr(σ(Frobarith
p )|Λi W ).
• Note: The ci,p are well-defined because σ(Ip ) = 1.
• Note: The ci,p = Tr(σ(Frobgeom
p )|Λi W ) depend only on p (and not on the
embedding GQp ⊆ GQ as the trace is invariant under conjugation.
• If W is semisimple, then σ is determined by the
{c1,p , . . . , cn,p }p∈S
/

for each finite set of primes S such that W is unramified outside S, cf.
[Ser97, I-10].
• In fact, the collection {c1,p }p∈S
/ is sufficient (plus the assumption that W
semisimple and unramified outside S).
• Indeed:
– Consider the group algebra Λ := Q` [GQ,S ], where GQ,S is the quotient
of GQ by the closure of the subgroup generated by the Ip , p ∈/ S.
– Assume that W, W 0 are two semisimple, continuous GQ,S -representations
with
Tr(Frobarith
p |W ) = Tr(Frobarith
p |W 0 )
for all p ∈
/ S.
– By Chebotarev density and continuity this implies
Tr(λ|W ) = Tr(λ|W 0 ),
for all λ ∈ Λ.
– For irreducible, pairwise non-isomorphic Λ-modules W1 , . . . , Wm and
each i = 1, . . . , m, there exists µi ∈ Λ, such that
µi = 1
on Wi , but
µi = 0
on Wj , j 6= i (this is a version of the Chinese remainder theorem).
– Write
m m
⊕n0
M M
W = Wi⊕ni , W 0 = Wi i
i i=1
in isotypic components with W1 , . . . , Wm irreducible, pairwise non-
isomorphic, and ni , n0i ∈ N, possibly zero.
– Then
ni dimQ` Wi = Tr(µi |W ) = Tr(µi |W 0 ) = n0i dimQ` Wi
for all i = 1, . . . , m, i.e., ni = n0i , i = 1, . . . , m and W ∼
= W 0 as desired.
• This statement can be seen as an analog of “strong multiplicity one” for
cuspidal automorphic representations for GLn mentioned last time.
48 JOHANNES ANSCHÜTZ

• It was important that we considered representations with coefficients in

characteristic 0, i.e., it is wrong that the traces ci,p determine the (semisim-
ple) representation. By the Brauer–Nesbitt theorem it is still true that the
full system {c1,p , . . . , cn,p }p∈S
/ determines a finite dimensional semisimple
representation uniquely.
The traces of Frobenius elements contain informations of high arithmetic signifi-
cance. The following discussion is very important for understanding the arithmetic
implications of the Langlands program.
Arithmetic significance of the traces of Frobenii:
• Let σ : GQ → GL(W ) be an `-adic representation, unramified outside S.
• The traces
c1,p := Tr(σ(Frobarith
p )|W )
encode significant arithmetic information.
• For example, assume that F/Q is finite Galois extension with Galois group
Gal(F/Q) ∼= S3 , and that
σ : GQ  Gal(F/Q) → GL2 (Q` )
is inflated from the unique irreducible, 2-dimensional representation of S3 .
Set
S := {p ramified in F }
and for p ∈
/S
Frobarith
F,p

as the image of Frobarith

p ∈ GQ,S in Gal(F/Q) ∼ = S3 .
• Then:
∗
Tr(Frobarith
p |W ) = 2
⇔ Frobarith
F,p ∈ {1} ,
⇔ p splits completely in OF
∗
Tr(Frobarith
p |W ) = 0
arith
⇔ FrobF,p ∈ {(1, 2), (1, 3), (2, 3)} ,
⇔ p splits into three distinct primes in OF
∗
Tr(Frobarith
p |W ) = −1
⇔ Frobarith
F,p ∈ {(1, 2, 3), (1, 3, 2)} .
⇔ p splits into two distinct primes in OF
• Thus: Knowledge of traces of Frobenii for all `-adic Galois representations
(with finite image) implies knowledge of the decomposition of unramified
primes in all finite extensions of Q.
• Other examples of `-adic representations are the étale cohomology groups
i
Hét (XQ , Q` )
for proper, smooth schemes X over Q, and i ≥ 0.
i
• The GQ -action is induced by functoriality of Hét (−, Q` ) from the (right)
action of GQ on
XQ = X ×Spec(Q) Spec(Q).
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM49

• Proper, smooth base change together with the Grothendieck-Lefschetz trace

i
formula imply that if X has good reduction at p and p 6= `, then Hét (XQ , Q` )
is unramified at p for i ≥ 0 and
2dim(X)
X
(−1)i Tr(Frobgeom
p
i
|Hét (XQ , Q` )) = ]X (Fp ),
i=0

where X → Spec(Z(p) ) is a proper, smooth model of X at p.

From the arithmetic point of view we can now state one major point of the
Langlands program.
The whole point of Langlands reciprocity:
• Traces of Frobenii are supposed to match Hecke eigenvalues!
• Thus, arithmetic information is expected to be encoded in automorphic
information, which might be more accessible.
• The following example was taken from a post of Matthew Emerton. More
insightful posts by him can be found via his webpage: http://www.math.
uchicago.edu/~emerton/ .
– Let F be the splitting field of x3 − x −√1.
– Then F is the Hilbert class field of Q( −23), and Galois over Q with
Galois group S3 .
– Consider as before the associated 2-dimensional, irreducible Galois rep-
resentation σ : GQ → Gal(F/Q) ,→ GL2 (Q` ).
∞
an q n ,
P
– Hecke proved that there exists a normalized newform f (q) =
i=1
such that
ap = Tr(σ(Frobarith
p ))
for all primes p with p - 23.
– One can make f more explicit: Considering ramification, f must lie in
S1 (Γ0 (23), χ) with χ : (Z/23)× → C× the quadratic character deter-
mining the quadratic subfield
√
Q( −23) ⊆ Q(ζ23 ).
People knowing modular forms (or a database like LMFDB) will tell
us that we must have
∞
Y
f (q) = q (1 − q i )(1 − q 23i ).
i=1

– We can now calculate the developement of f (or look it up in LMFDB):

f (q) = q − q 2 − q 3 + . . . + q 58 + 2q 59 + . . .
and deduce how primes decompose in F , e.g., 59 splits completely!
• Perhaps the most famous example how modularity of traces of Frobenii was
used was the proof of Fermat’s last theorem.
• Namely, Fermat’s theorem was proved by Wiles/Taylor using the following
strategy (initiated by Frey, and complemented by Serre, Ribet):
– Assume up + v p + wp = 0 with u, v, w ∈ Q, uvw 6= 0 and p ≥ 3.
– Consider the elliptic curve E over Q with (affine) Weierstraß equation
y 2 = x(x + up )(x − v p )
 50 JOHANNES ANSCHÜTZ

(after possibly manipulating u, v, w a bit).

– Consider the dual σ of the Galois representation
GQ → GL(H 1 (E , Q` )) ∼
ét Q = GL2 (Q` ).
– Show that σ is modular, i.e., there exists a normalized newform f1 (q) =
∞
an q n of weight 2 such that
P
n=1

Tr(σ(Frobarith )) = ap
for almost all primes p.
– Using a theorem of Ribet, one concludes that f1 must be congruent to
a normalized newform f2 in
S2 (Γ0 (2)),
i.e., the Fourier coefficients of f1 , f2 are algebraic integers and congru-
ent modulo some prime.
– But S2 (Γ0 (2)) = 0, and thus f2 cannot exist.
– This yields the desired contradiction.
– For more details, cf. [Wil95], [Rib90].
Now, we state what is expected on the relation between Galois representations
and cusp forms.
Langlands reciprocity for newforms:
• Fix a prime ` and an isomorphism
ι : Q` ∼
= C.
• For newforms the Langlands program combined with the Fontaine-Mazur
conjecture predicts a bijection
'
LL : Amod −→ Gmod
from the set
M
Amod := {irreducible GL2 (Af )-subrepresentations π ⊆ Sk },
k≥1

to a certain set
Gmod
consisting of irreducible, 2-dimensional Galois representations
σ : GQ → GL2 (Q` )
• LL should satisfy that for all π ∈ Amod with σ := LL(π) we have
ι(Tr(σ(Frobarith
p ))) = ap (π), ι(det(σ(Frobarith
p ))) = bp (π)
In a previous ver- for p outside some specified finite set S of primes.
sion I considered • Here, the eigenvalues ap (π) resp. bp (π) are used and not the ãp (π), b̃p (π).
the geometric • Note: Such a bijection LL is uniquely determined (by the respective multi-
Frobenius, which plicity one theorems), if it exists.
was wrong. • Conjecturally, the set is S specified explicitly. Recall that a prime p un-
ramified for σ resp. π if
σ(Ip ) = 1
resp.
π GL2 (Zp ) 6= 0.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM51

• Then a prime p 6= ` is conjecturally unramified for π if and only if it is

unramified for σ := LL(π).
• The matching of Hecke eigenvalues states then more precisely that σ(Frobarith
p )
should have characteristic polynomial
X 2 − ι−1 (ap (π))X + ι−1 (bp (π))
for each unramified prime p 6= ` for π.
• It is the next aim of the course to indicate how the map LL can be con-
structed, i.e., how to associate Galois representations to normalized new-
forms. For this we follow results of Deligne, cf. [Del71b], and Deligne/Serre,
cf. [DS74].
Let us make more precise which set of Galois representations is expected to be
associated to newforms.
The set Gmod :
• The set Gmod was defined by Fontaine/Mazur in relation with their remark-
able conjecture on geometric Galois representations, cf. [FM95].
• Let σ : GQ → GL2 (Q` ) be an irreducible, 2-dimensional `-adic representa-
tion.
• Then by definition σ ∈ Gmod if and only if
– σ unramified at almost all p, i.e., σ(Ip ) = 1 for p outside some finite
set of primes S.
– σ is odd, i.e., for each complex conjugation c ∈ Gal(C/R) ⊆ GQ we
have
det(σ(c)) = −1.
– The restriction
σ` := σ|GQ` : GQ` → GL2 (Q` )
is de Rham with Hodge-Tate weights 0, ω for some ω ∈ N (note that
ω = 0 is allowed).
• If σ = LL(π) with π ⊆ Sk , then conjecturally ω = k − 1.
• Using the work of many people (Wiles, Kisin, Emerton, Skinner-Wiles,
Pan,. . . ) any σ ∈ Gmod with distinct Hodge-Tate weights is modular if
` ≥ 5, cf. [Pan19, Theorem 1.0.4.].
• If σ ∈ Gmod has finite image, then σ is modular, cf. [PS16], [GH19, Section
13.4.] (and the references therein).
• Conjecturally: HT-weights 0 ⇔ image of σ finite.
We have to make a short digression on the important notion of being de Rham.
de Rham representations:
• Let p be a prime (the previous `).
• Let K/Qp a discretely valued, non-archimedean extension with perfect
residue field (e.g., K/Qp finite).
• Let σ : GK := Gal(K/K) → GLm (Qp ) be a continuous representation.
• Then σ has coefficients in some finite extension E/Qp (by compactness of
GK as Qp carries the colimit topology here).
• We obtain a Qp -linear representation, a “local p-adic Galois representa-
tion”,
ρ : GK → GLm (E) ⊆ GLn (Qp ),
where n := m · dimQp E.
52 JOHANNES ANSCHÜTZ

• Thus, we can change our setup and let V be a finite dimensional Qp -vector
space, and ρ : GK → GL(V ) a continuous representation.
• Then V is called de Rham if
dimK (BdR ⊗Qp V )GK = dimQp V
for a certain field extension BdR of K with GK -action, cf. [BC09, Section
6].
• Namely: BdR is Fontaine’s field of p-adic periods, cf. [Fon94], [BC09, Defi-
nition 4.4.7]. Abstractly,
BdR ∼
= CK ((t)),
where CK is the completion of K for its p-adic valuation (this topology is
coarser than the colimit topology on K).
• BdR is a discretely valued field, with residue field CK , the deduced decreas-
ing filtration Fil• BdR is GK -stable and as CK -semilinear GK -representations
gr• BdR ∼
M
= CK (j),
j∈Z

where CK (j) := CK ⊗Zp Zp (j), with

Zp (1) := Tp µp∞ (CK )
the p-adic Galois representation associated to the cyclotomic character
χcyc : GK → Z× ∼
p = Aut(µp∞ (CK )).

• An important theorem of Tate, cf. [BC09, Theorem 2.2.7], states the fol-
lowing: Let χ : GK → Q×p be a character. Then
(
0 K, if χ|IK has finite image
H (GK , CK (χ)) =
0, otherwise ,
here IK ⊆ GK is the inertia subgroup.
• In particular:
H 0 (GK , CK (j)) = 0, if j 6= 0,
as χjcyc |I has infinite image if j 6= 0.
K
• Conclusion: If dimC V = 1, then V is de Rham if and only if V is isomorphic
to χ · χjcyc for some character χ : GK → Q× p with IK of finite order, and
j ∈ Z. In particular, there exists many representations, which are not de
p−1 a
Rham, e.g., the characters (χcyc ) with a ∈ Zp \ Z.
• Let us mention some general facts on de Rham representations.
– ρ : GK → GL(V ) is de Rham if ρ|GK 0 is de Rham for some finite
extension K 0 of K, cf. [BC09, Proposition 6.3.8].
– de Rham representations are stable under subquotients, duals and ten-
sor products, cf. [BC09, Section 6.1].
• Let V be a de Rham representation of dimension n. Then
V ⊗Qp CK ∼
= CK (j1 ) ⊕ . . . ⊕ CK (jn )
as CK -semilinear GK -representations, where j1 , . . . , jn ∈ Z. The unordered
collection (j1 , . . . , jn ) is called the collection of Hodge-Tate weights for V .
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM53

Thus, the condition of being de Rham puts some sort of ”integrality” condition
on a p-adic representation.
We close the lecture by indicating how the map LL will be constructed (if k = 2).
Strategy for constructing LL (roughly):
• Let π ∈ Amod , with associated system of Hecke eigenvalues
{ap , bp }p∈S
/ .

• Assume for simplicity that π ⊆ S2 , i.e., π is generated by some newform of

weight 2.
• Then we will construct (something close to) a GL2 (Af )-equivariant embed-
ding
S2 ,→ H 1 (GL2 (Q)\(GL2 (Af ) × H± ), C) := lim H 1 (GL2 (Q)\(GL2 (Af )/K × H± ), C)
−→
K

of S2 into cohomology.
• The given isomorphism ι : Q` ∼
= C yields an isomorphism
H (GL2 (Q)\(GL2 (Af ) × H ), C) ∼
1 ±
= H 1 (GL2 (Q)\(GL2 (Af ) × H± ), Q` )
of cohomology groups.
• As already mentioned for K ⊆ GL2 (Af ) a (sufficiently small) compact-open
subgroup, the complex manifold
GL2 (Q)\(GL2 (Af )/K × H± )
is actually algebraic, i.e., given by X̃K (C) for some quasi-projective scheme
X̃K → Spec(C).
• The étale comparison theorem implies
H 1 (GL2 (Q)\(GL2 (Af )/K × H± ), Q` ) ∼ = H 1 (X̃K , Q` ).
ét

• Now the miracle happens: For K ⊆ GL2 (Af ) a (sufficiently small) compact-
open subgroup
the quasi-projective scheme X̃K → Spec(C) is canonically defined over Spec(Q)!
In other words, X̃K ∼
= XK ×Spec(Q) Spec(C) for (compatible) quasi-projective
schemes XK → Spec(Q).
• In particular,
1
Hét (X̃K , Q` ) ∼ 1
= Hét (XK,Q , Q` )
by invariance of étale cohomology under change of algebraically closed base
fields. But now the RHS carries an action of GQ !
• Look now at the GQ × GL2 (Af )-module
1 1
Hét (XQ , Q` ) := lim Hét (XK,Q , Q` ).
−→
K

• Then we will check that the GQ -module

1
LL(π) := HomGL2 (Af ) (π, Hét (XQ , Q` ))
is 2-dimensional and that we can express the traces of Frobenii by Hecke
eigenvalues. The dual of this 2-dimensional representation will then be the
desired representation.
• Replacing Q` by some local system a similar strategy works for weight
k ≥ 2.
54 JOHANNES ANSCHÜTZ

• However, this does not help for k = 1. Here one has to use that a weight
1 modular form can be congruent to some modular form of weight k ≥ 2,
and use the previous case.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM55

7. Modular curves as moduli of elliptic curves (by Ben Heuer)

Last time:
• First rough sketch of construction of
'
LL : Amod −→ Gmod
M
Amod :={irreducible GL2 (Af )-subrepresentations π ⊆ Sk },
k≥1

Gmod :={σ : GQ → GL2 (Q` ), unramified outside finite set S, odd, de Rham above `}
• Crucial point for construction: For sufficiently small compact-open sub-
groups K ⊆ GL2 (Af ) the complex manifold
GL2 (Q)\(GL2 (Af )/K × H± )
(1) is algebraic over Spec(C),
(2) admits a canonical model XK over Spec(Q).
1
This gives rise to the Galois action on l-adic étale cohomology Hét (XK,Q̄ , Q` ).
Today:
• Explain why the algebraic model XK exists.
• Reinterpret notions defined before in algebro-geometric terms: modular
forms, compactification, q-expansions, adelic Hecke action, ...
Reminder on elliptic curves Reference: [Silverman: The Arithmetic...]
• Let K be any field. We have the following equivalent definitions:
Definition 7.1. An elliptic curve over K is equivalently a
– connected smooth projective curve E|K of genus 1 with a chosen point
O ∈ E(K).
– connected smooth projective algebraic group E|K of dimension 1.
– non-singular plane cubic curve, defined by a Weierstraß equation
E : y 2 = x3 + ax + b, a, b ∈ K
(if charK 6∈ {2, 3}, otherwise more terms are required, cf. [Del75]).
• Fact: The group scheme is automatically commutative (!).
• Fact: The non-singularity can be expressed as a condition on the discrimi-
nant:
E non-singular ⇔ ∆(a, b) = −16(4a3 + 27b2 ) 6= 0
(if charK ∈
/ {2, 3}).
• Fact: Given E, the Weierstraß equation is not unique. But the j-invariant
4a3
j(E) := j(a, b) = 1728
4a3 + 27b2
is independent of choice of Weierstraß equation.
• Fact: If K is algebraically closed, we have
E∼ = E 0 ⇔ j(E) = j(E 0 ).
• More generally, we can replace Spec(K) by any base scheme S.
Definition 7.2. An elliptic curve over S is a proper smooth curve E → S
with geometrically connected fibres of genus 1, together with a point 0 : S →
E.
56 JOHANNES ANSCHÜTZ

• This can be thought of as family of elliptic curves parametrised by S.

• Again, this automatically has a group structure:
Theorem (Abel). There is a unique way to endow an elliptic curve E →
S with the structure of a commutative S-group scheme s.t. 0 = identity
section.
• Zariski-locally on S, one can describe E in terms of a Weierstraß equation.
Definition 7.3. A morphism of elliptic curves E → E 0 over S is a ho-
momorphism of S-group schemes. An isogeny is an fppf-surjective homo-
morphism with finite flat kernel.
Example 7.4. For any N ∈ Z, the group scheme structure gives the mor-
phism “multiplication by N ”, which is denoted by [N ] : E → E.
• Fact: [N ] is finite flat of rank N 2 . It is finite étale if N is invertible on S.
• In particular, [N ] is an isogeny.
Definition 7.5. E[N ] := ker[N ] ⊆ E, the N -torsion subgroup. This is
a finite flat closed subgroup scheme. It is finite étale if N is invertible on
S.
• More generally, for any finite flat subgroup scheme D ⊆ E[N ] of exponent
N , ∃ unique elliptic curve E/D with isogeny ϕ : E → E/D with kernel D.
• [N ] induces a dual isogeny ϕ∨ : E/D → E with kernel E[N ]/D.
Elliptic curves over K = C Reference: [Diamond–Shurman: A first course...]
• Let E|C be an elliptic curve. Then E(C) has the structure of a compact
complex Lie group. As such, it can be canonically uniformised:
• Let Lie E be the tangent space at 0 ∈ E(C). Then Lie E ∼ = C and there is
an isomorphism of complex Lie groups
E(C) = Lie E/H1 (E, Z) = ∼ C/Λ, where Λ = ∼ Z2 .
• For any N ∈ Z, we have
E(C)[N ] = ( N1 Λ)/Λ ⊆ C/Λ.
• Note: E(C)[N ] = ( N1 Z/Z)2 = (Z/N )2 is of rank N 2 , as claimed.
• Upshot: E(C) is a 1-dimensional complex torus: A quotient of C by a
lattice Z2 ∼
= Λ ⊆ C such that Λ ⊗Z R = C.
• morphisms of complex tori := morphisms of complex Lie groups
Theorem. There is an equivalence of categories
{elliptic curves over C} → {1-dim complex tori}
• Essential surjectivity: Let Λ ⊆ C be a lattice as above. Define the Weier-
straß ℘-function
1 X1 1

℘(z, Λ) = 2 + −
z z2 (z − w)2
w∈Λ
w6=0

• Then ℘(z) = ℘(z, Λ) is holomorphic on C\Λ with complex derivative

X 1
℘0 (z) = −2 .
(z − w)3
w∈Λ
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM57

• This is clearly Λ-periodic ℘ and ℘0 are meromorphic functions on C/Λ.

Theorem. The functions ℘ and ℘0 are related by a Weierstraß equation
EΛ : ℘02 =4(℘3 + g2 (Λ)℘ + g3 (Λ))
for explicit g2 (Λ), g3 (Λ) ∈ C. We thus get an isomorphism of Lie groups
∼
(℘, ℘0 ) : C/Λ −
→ EΛ (C).
• Sending C/Λ 7→ EΛ defines a quasi-inverse
{1-dim complex tori} → {elliptic curves over C}.
Complex moduli spaces of elliptic curves
• We can now parametrise complex elliptic curves up to isomorphism.
• For this, we just have to determine when two lattices Λ1 , Λ2 satisfy
∼ C/Λ2 .
C/Λ1 =
• Fact: Any homomorphism of complex tori is multiplication by a ∈ C.
Indeed, it lifts to a holomorphic homomorphism on the Lie algebras.
⇒ this happens iff Λ1 = aΛ2 for some a ∈ C× .
• Can now uniformise as follows: Write Λ = w1 Z ⊕ w2 Z.
• The condition Λ ⊗Z R = C implies w1 /w2 ∈ C \ R = H± .
⇒ For every complex torus E, there is τ ∈ H± such that
E∼= Eτ := C/Λτ , Λτ = Z ⊕ τ Z
• Finally, for τ, τ 0 ∈ H± , we have
aτ + b
Z + τ Z = Z + τ 0 Z ⇔ ∃γ = = τ0
a b


c d ∈ GL2 (Z) : γτ =
cτ + d
• This shows: τ →
7 Eτ defines
GL2 (Z)\H = {complex tori}/ ∼∼
± ∼
= {elliptic curves over C}/ ∼
This gives our space from earlier,
SL2 (Z)\H = GL2 (Z)\H± = GL2 (Q)\(GL2 (Af )/K × H± ) for K = GL2 (Ẑ)
an interpretation as a “moduli space” (so far, in a weak sense):
SL2 (Z)\H = {elliptic curves over C}/ ∼
• C algebraically closed ⇒ j-invariant defines bijection (in fact biholomor-
phism)
∼
j : SL2 (Z)\H −
→ C.
• Via the covering map H → GL2 (Z)\H± , we get from the above:
±

∼
H± ={complex tori with ordered basis α : Z2 −
→ Λ}/ ∼
∼
={elliptic curves E over C with α : Z2 −
→ H1 (E, Z)}/ ∼
• What about moduli interpretations of other levels? Recall for N ∈ N, we
had the level K = K(N ),
a
XK := GL2 (Q)\(GL2 (Af )/K × H± ) = Γ(N )\H± ,
where
Γ(N ) = {γ ∈ GL2 (Z)|γ ≡ ( 10 01 ) mod N }.
58 JOHANNES ANSCHÜTZ

• These are precisely the automorphisms of Λ that preserve the subgroup

1
EΛ [N ] = N Λ/Λ ⊆ C/Λ.
Consequently, we have the more general statement
∼
Γ(N )\H± ={complex tori with ordered basis α : (Z/N )2 −
→ 1
N Λ/Λ}/ ∼
2 ∼
={elliptic curves E over C with α : (Z/N ) −
→ E(C)[N ]}/ ∼ .
• In between GL2 (Z) and Γ(N ), we had the groups
Γ1 (N ) = {γ ∈ GL2 (Z)|γ ≡ ( ∗0 ∗1 ) mod N },
Γ0 (N ) = {γ ∈ GL2 (Z)|γ ≡ ( ∗0 ∗∗ ) mod N }.
• The modular curves associated to these have moduli interpretations:
Γ1 (N )\H± ={elliptic curves E|C with point Q ∈ E(C)[N ] of exact order N }/ ∼
Γ0 (N )\H± ={elliptic curves E|C with cyclic subgroup C ⊆ E(C)[N ] of rank N }/ ∼ .
Motivating Example: The Legendre family
• Consider the complex manifold U = P1 \{0, 1, ∞} in the variable λ.
• Over U , we have a complex family of elliptic curves:
• The Legendre family is cut out of P2U → U by the Weierstraß equation
Eλ : Y 2 = X(X − 1)(X − λ).
Proposition 7.6. (1) We have Eλ [2] = {O, (0, 0), (1, 0), (λ, 0)}. In par-
ticular, there is a natural isomorphism Eλ [2] = (Z/2Z)2 .
(2) For any elliptic curve E over C together with a trivialisation
∼
α : (Z/2Z)2 −
→ E[2],
there is a unique point x ∈ U such that E = (Eλ )x .
• Upshot: Eλ → U is universal elliptic curve with level Γ(2), in some
sense.
• This is all algebraic! Can make sense of Eλ → U as morphism of schemes
C.
• Even better: These schemes are already defined over Q.
• Can elaborate on this argument to show that
Γ(2)\H± =: XΓ(2) (C) = P1 \{0, 1, ∞}.
∗
• NB: In particular, the compactification is XΓ(2) = P1 .
• This shows that S2 (Γ0 (2)) = 0, which we used for Fermat’s Last Theorem.11
Idea:

Get algebraic model of Γ(N )\H± over Q by passing from C to moduli of

elliptic curves over general Q-schemes S.

• Question: Is there a scheme representing the functor

PSL2 (Z) : S 7→ {elliptic curves over S}/ ∼
on schemes over Q? This would be a scheme X → Q with C-points
SL2 (Z)\H.
11We used implicitly that ω 2 ∼ Ω1 .
=
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM59

• Answer: No. An easy way to see this is the phenomenon of twists:

• Can have two non-isomorphic E ∼ 6 E 0 over Q that become isomorphic over
=
a quadratic extension K. In particular, j(E) = j 0 (E).
• Clearly, two Q-points of X agree iff they do over K. So X cannot exist!
• Same problem for the Legendre family above: This does not represent
∼
S 7→ {elliptic curves E over S with α : (Z/2Z)2 −
→ E[2]}/ ∼
because this fails to account for quadratic twists.
• Fact: Twists of elliptic curves E over a field K correspond to 1-cocycles
GK|K → Aut(E).
Quadratic twists come from the involution ([−1] : E → E) ∈ Aut(E).
• Slogan: Nontrivial automorphisms are bad for representability.12
• Upshot: In order to get a representable functor, need additional data.
Moduli problems of elliptic curves [Katz–Mazur: Arithmetic moduli...]
• Consider the moduli functors on schemes over Z[ N1 ]:
PΓ0 (N ) : S 7→ {(E|S, C ⊆ E[N ] cyclic subgroup scheme of rank N )}/ ∼,
PΓ1 (N ) : S 7→ {(E|S, Q ∈ E[N ](S) point of exact order N )/ ∼,
∼
PΓ(N ) : S 7→ {(E|S, α : (Z/N Z)2 −
→ E[N ] isom of group schemes)/ ∼ .
• We can also form “mixed moduli problems” like
PΓ1 (N )∩Γ0 (p) := PΓ1 (N ) ×PSL2 (Z) PΓ0 (p) : S 7→ {(E, Q, C)}/ ∼
• Crucial point: Let (E, C) ∈ PΓ0 (N ) (S). The automorphism [−1] : E → E
sends C 7→ C. In particular, [−1] ∈ Aut(E, C).
• In contrast, let (E, Q) ∈ PΓ1 (N ) (S), then [−1] : E → E sends Q ∈
E[N ] to −Q ∈ E[N ], which is different if N ≥ 3. In particular, [−1] 6∈
Aut(E, Q).13
• Similarly for PΓ(N ) (S). This explains the problem with the Legendre fam-
ily: N = 2 is too small to deal with the automorphism [−1]! We need
N ≥ 3.
• Upshot: The moduli problems PΓ(N ) and PΓ1 (N ) are “rigid” (:= have no
non-trivial automorphisms) for N ≥ 3, in contrast to PΓ0 (N ) and PΓ(2) .

Theorem. Let N ≥ 3 and p any prime.

(1) The moduli problems PΓ(N ) and PΓ1 (N ) are each representable by
smooth affine curves XΓ(N ) and XΓ1 (N ) over Z[ N1 ].
(2) For any n ∈ N, the moduli problem PΓ(N )∩Γ0 (pn ) is representable by a
flat affine curve XΓ1 (N )∩Γ0 (pn ) over Z[ N1 ] that is smooth over Z[ pN
1
].

12More precisely, it is the non-flatness (due to exceptional isomorphisms of elliptic curves) of the
automorphism group that causes problems. E.g., in many cases the Picard functor parametrizing
isomorphism classes of line bundles is representable, but under the assumptions, say, of properness
and geometrically integral fibers the automorphism groups of the line bundles are all Gm , and
hence flat over the base.
13And there are no exceptional isomorphisms.
60 JOHANNES ANSCHÜTZ

• Remark: In order to represent PSL2 (Z) , one can pass from schemes to the
bigger category of stacks. Get “moduli stack of elliptic curves” M1,1 .
• Between varying level structures Γ ⊆ Γ0 ⊆ Γ(N ), have forgetful morphisms
XΓ0 → XΓ .
These are all finite étale over Z[ N1 ].
• GL2 (Z/N Z) acts from the right on PΓ(N ) by precomposition
α 7→ α ◦ γ.
Alternatively, this is a left action via precomposition with γ ∨ = γ −1 det γ.
• This induces a GL2 (Z/N Z)-action on XΓ(N ) .
• Similarly, there is a natural (Z/N Z)× -action on XΓ1 (N ) .
• We can then define a modular curve
XΓ0 (N ) := XΓ1 (N ) /(Z/N Z)× .
• This does not represent P(Γ0 (N )), but it is “as close as possible”.
• In particular, XΓ0 (N ) (K) = P(Γ0 (N ))(K) for algebraically closed K.
Compactification
• The j-invariant associated to Weierstraß equations defines a finite flat func-
tion
j : X → A1 .
• By normalisation in A1 → P1 , can define
j : X ∗ → P1 ,
still a finite flat morphism. Think of X ∗ as X plus a finite divisor of points.
• Fact: X ∗ is smooth and proper.
• Reason: This can be seen using the Tate curve over Z[ N1 ][[q]].
• X ∗ (C) is the smooth compactification XC∗ of Γ\H± mentioned earlier.
• Remark: X ∗ is a moduli space of “generalised elliptic curves”, cf. [DR73].
Geometric Modular forms References: [Katz], [Loeffler: lectures notes]
• Fix N ≥ 3. Let X = XΓ1 (N ) .
• Goal: Geometric reinterpretation of modular forms as sections of sheaves
on X:
Definition 7.7. Let ω := e∗ Ω1E|X , where e : X → E is the identity section.
• Since E → X is smooth of dimension 1, this is a line bundle.
• Fact: This extends uniquely to a line bundle ω on X ∗ . Reason: The
universal elliptic curve extends to a group scheme E ∗ → X ∗ , take e∗ Ω1E ∗ |X ∗ .

Proposition 7.8. The analytification of ω ⊗k is naturally isomorphic to

the sheaf ω k of modular forms of weight k on X ∗ (C).
∼
Sketch of proof. Recall: H± is moduli space of E|C with α : Z2 −
→ H1 (E, Z).
∼
Integration α(1,0) defines an isomorphism e∗ Ω1E|C −
R
→ C.
Get canonical trivialisation of ω over H± . Check GL2 (Z)-action coincides.


• In particular, the sheaf ω already has an algebraic model over Z[ N1 ].

LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM61

Definition 7.9. For any Z[ N1 ]-algebra R, can define

Mk (Γ1 (N ), R) = Γ(XΓ∗1 (N ) ×Spec(Z[1/N ]) Spec(R), ω ⊗k ),
the R-module of modular forms of weight k with coefficients in R.
• Fact: This is a finite free Z[ N1 ]-module.
• Fact: Using Tate curves, get q-expansions in R[[q]]. Can define cusp
forms.
• Fact: If R, S are flat Z[ N1 ]-algebras (e.g. any Q-algebras), then
Mk (Γ1 (N ), R) ⊗R S = Mk (Γ1 (N ), S).
• Upshot: Modular forms are manifestly objects of algebraic geometry!
62 JOHANNES ANSCHÜTZ

8. The Eichler–Shimura relation (by Ben Heuer)

Last time:
• There is a GL2 (Z)-equivariant bijection
∼
=
H± → {elliptic curves E over C with α : Z2 −
→ H1 (E, Z)}/ ∼
τ 7→ (Eτ := C/(Z + Zτ ), α : Z2 → Z + Zτ, (1, 0) 7→ 1, (0, 1) 7→ τ )
• Corollary:
b × H± ) ∼ ∼
b2 −
GL2 (Z)\(GL2 (Z) = {elliptic curves E over C with α : Z → T E}/ ∼,
where T E := lim E[N ] is the full adélic Tate module of E.
←−
N ∈N
• This can be defined over Q!
• Set X
b → Spec(Q) as the moduli scheme representing the functor
∼
b2 −
Q−Sch → Sets, S 7→ {elliptic curves E over S with α : Z → T E}/ ∼ .
Here, T E = lim E[N ] is an inverse limit of finite, étale group schemes over
←−
N
S.
• −1 ∈ GL2 (Z)
b acts trivially on X.b
• For K ⊆ GL2 (Z)
b compact-open with K ⊆ K1 (3) (note in particular, −1 ∈ /
K) have smooth quasi-projective curve
XK = X/K.
b

• E.g.: If K = K(N ), N ≥ 3, then

XK(N )
∼
parametrizes elliptic curves E together with trivialization (Z/N )2 −
→ E[N ].
• Other examples, K = K1 (N ) Z/N ,→ E[N ] , K = K0 (N ) subgroups.
Next goal:
• Use modular curves to reinterpret Hecke operators geometrically.
• From geometry of modular curves modulo p, deduce fundamental relation
between Galois action and Hecke action: The Eichler–Shimura relation.
The adélic action:
• Clearly, GL2 (Z)
b acts on X, b2 ∼
b by precomposition on α : Z = T E.
• But: Want GL2 (Af )-action to get Hecke operators
⇒ work with elliptic curves up to quasi-isogeny.
• Recall: Af = Zb ⊗Z Q and

b × H± ) ∼
GL2 (Z)\(GL2 (Z) = GL2 (Q)\(GL2 (Af ) × H± ).
• On elliptic curves can interpret this as follows.
• Recall that the LHS geometrizes to X, b which represents the functor
b2 −∼
P : Q−Sch →Sets, S 7→{(E|S, α : Z → T E)}/isomorphism

• Let V E = T E ⊗Z Q be the rational adélic Tate module, and define

∼
P 0 : Q−Sch →Sets, S 7→{(E|S, α : A2f −
→ V E)}/quasi-isogeny14
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM63

Definition 8.1. Recall: An isogeny is a surjective homomorphism E → E 0

with finite flat kernel. A quasi-isogeny is an element of
{E → E 0 isogeny} ⊗Z Q.
A p-quasi-isogeny is an element of
{E → E 0 isogeny of degree a power of p} ⊗Z Z[ p1 ].
• Then P ∼
= P 0 . Indeed:
– Clearly have natural transformation P → P 0 . Sketch for inverse:
∼
– Take (E, α : A2f − → V E) ∈ P 0 (S). Multiply by isogeny N : E → E
−1
until α restricts to
TE → Z b2.
Reduce both sides mod M ∈ N. Let D be the kernel of the induced
map
E[M ] → (Z/M )2 ,
this stabilises for M  0. The dual isogeny E/D → E induces
∼
b2 −
α:Z → T (E/D).
– This defines unique representative in P (S). Thus P (S)  P 0 (S).
• This shows that the natural GL2 (Ẑ)-action extends to a GL2 (Af )-action.
• This does not fix

the projection to any XK because it changes E.
• Example: p0 10 acts by sending E 7→ E/D where D ⊆ E[p] is the subgroup
generated by the image of α(1, 0) under Tp E  E[p].
• Fact: All of this still works after compactification.
Geometric interpretation of Hecke operators
• Let V = C ∞ (GL2 (Af ) × H± ) be the smooth GL2 (Af )-representation of
smooth functions
f : GL2 (Af ) × H± → C, (g, z) 7→ f (g, z),
i.e., f is fixed under some compact-open subgroup in GL2 (Af ) and z 7→
f (g, z) is smooth for any g ∈ GL2 (Af ).
• Fix a prime p.
• Let ϕ ∈ H(GL2 (Qp )) be any element of the Hecke algebra at p, i.e., a locally
constant function ϕ : GL2 (Qp ) → C with compact support.
• Recall that for f ∈ V we have
Z
ϕ ∗ f (g, z) = ϕ(h)f (gh, z)dh,
GL2 (Qp )

where we chose the Haar measure on GL2 (Qp ) with volume 1 on Kp =

GL2 (Zp ).
• The integral above can be rewritten as follows:
• Consider the diagram
GL2 (Af ) × H± × GL2 (Qp )
q2 q1

GL2 (Af ) × H± GL2 (Af ) × H± ,

64 JOHANNES ANSCHÜTZ

q1 (g, z, h) =(g, z),

q2 (g, z, h) =(gh, z).
• Such a diagram is called a correspondence.
• We also have the projection to the third factor:
π : GL2 (Af ) × H± × GL2 (Qp ) → GL2 (Qp )
Lemma 8.2. For f : GL2 (Af ) × H± → C in V we have
ϕ ∗ f = q1,! (q2∗ (f ) · π ∗ (ϕ)),
where (−)∗ means the pullback of a function and (−)! integrating along the
fiber (which makes sense because ϕ has compact support).
• Proof. The fiber of q1 over (g, z) ∈ GL2 (Af ) × H± is
q1−1 (g, z) = {(g, z, h)| h ∈ GL2 (Qp )} ∼
= GL2 (Qp )
and q2∗ (f ) · π ∗ (ϕ) is the function
(g, z, h) 7→ f (gh, z)ϕ(h).
Applying q1,! , the integral over this fibre, gives
Z
ϕ(h)f (gh, z)dh = ϕ ∗ f (g, z).
GL2 (Qp )

Note that this is well-defined because ϕ has compact support. 

Passage to finite level
• Let N be coprime to p and let K := K1 (N ) ⊆ GL2 (Af ). Then Kp =
GL2 (Zp ).
• Write K p for the prime-to-p-part of K, i.e. K = Kp K p and K p ∩ Kp = 1.
• Assume that f is fixed under K, i.e., arises by pullback from a function
f˜: GL2 (Af )/K × H± → C.
• Assume ϕ ∈ H(GL2 (Qp ), Kp ), i.e., that ϕ is Kp -biinvariant.
• ϕ is the pullback of a function with finite support
ϕ̃ : Kp \GL2 (Qp )/Kp → C.
• In this case, ϕ ∗ f is again K-invariant, i.e., the pullback of a function
]
ϕ ∗ f : GL2 (Af )/K × H± → C.
• We can change the above diagram (while retaining notation) to:

(GL2 (Af )/K p × H± ) ×Kp GL2 (Qp )/K

q2 q1

GL2 (Af )/K × H± GL2 (Af )/K × H± .

Here − ×Kp − is the contracted product from Section 3 (i.e., the quotient
for the action k · (g, z, h) := (gk −1 , z, kh)) and
q2 ([g, z, h]) = [gh, z], q1 ([g, z, h]) = [g, z],
where square brackets indicate equivalence classes.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM65

• We again have a projection

π : (GL2 (Af )/K p × H± ) ×Kp GL2 (Qp )/Kp →Kp \GL2 (Qp )/Kp ,
[g, z, h] 7→[h].
• Next, mod out by GL2 (Q) on the left: Recall this gives a complex manifold
XΓ1 (N ) (C) = GL2 (Q)\(GL2 (Af )/K × H± )
• We then need to replace functions f by sections in ω k , k ∈ Z, on this.
• More precisely, we obtain
GL2 (Q)\(GL2 (Af )/K p × H± ) ×Kp GL2 (Qp )/Kp
q2 q1
(1)

XΓ1 (N ) (C) XΓ1 (N ) (C).

and for f ∈ H 0 (XΓ1 (N ) (C), ω k ), i.e., a weakly modular form of level K1 (N ),

we get
ϕ ∗ f = q1,! (q2∗ (f ) · π ∗ (ϕ)).
• To make this precise, use: We have a canonical isomorphism
q ∗ (ω k ) ∼
2 = q ∗ (ω k )
1
since ω is defined via pullback of a GL2 (Q)-equivariant line bundle on H± .
k

• Thus can regard q2∗ (f ) · π ∗ (ϕ) ∈ q2∗ (ω k ) as a section of q1∗ (ω k ).

• Thus we can sum up along the fibers to obtain the section
q1,! (q2∗ (f ) · π ∗ (ϕ)) ∈ Γ(XΓ1 (N ) (C), ω k ).
• This is a more geometric interpretation of the Hecke action on modular
forms.
Interpretation in terms of moduli
• On elliptic curves, (Equation (1)) corresponds to a Hecke correspon-
dence:
{(ι : E1 99K E2 , α)}
q2 q1

{(E2 , α0 )} {(E1 , α)}

where
– (E1 , α) is an elliptic curve over C with a Γ1 (N )-level structure
α : Z/N Z ,→ E1 [N ]
– Equivalently, this is the datum of a point Q := α(1) ∈ E1 [N ](C) of
exact order N .
– (ι : E1 99K E2 , α) is the data of elliptic curves E1 , E2 over C with α a
Γ1 (N )-level structure on E1 , and ι a p-quasi-isogeny:
– q1 : (ι : E1 99K E2 , α) 7→ (E1 , α).
– q2 : (ι : E1 99K E2 , α) 7→ (E2 , α0 ) where α0 is the composition
α ι
Z/N Z E1 [N ] ∼ E2 [N ]
(use: ι is an isomorphism on N -torsion as (p, N ) = 1). Equivalently,
this sends Q to the image Q0 in E2 [N ].
66 JOHANNES ANSCHÜTZ

– The map π sends (ι : E1 99K E2 , α) to the class of the isomorphism

∼
Tp (ι) : Tp E1 ⊗Zp Qp −
→ Tp E2 ⊗Zp Qp .
More concretely, choose any bases of Tp E1 resp. Tp E2 . Then Tp (ι) is
represented by a matrix A ∈ GL2 (Qp ). The class of A ∈ Kp \GL2 (Qp )/Kp
is independent of the chosen basis.
• Problem: Functor of p-quasi-isogenies not representable (only by ind-scheme).
• However: Mainly interested in ϕ given by the characteristic functions for
 
p 0
Kp Kp ↔ S̃p ,
0 p
 
p 0
Kp Kp ↔ T̃p .
0 1
• Then term ϕ(h) in the defining integral is supported on this double coset.
Hecke correspondences are supported on subspaces represented by schemes!
• For S̃p , consider tuples (ι : E1 99K E2 , α) of the form
([p] : E → E, α)
i.e. E1 = E2 and ι : E1 → E2 is multiplication [p].
• For T̃p , instead need to consider
(ι : E1 → E2 , α)
where ι : E1 → E2 is an isogeny of degree p.
• Equivalently, this is the datum of a cyclic subgroup ker ι ⊆ E1 of rank p.
• Last time: These are now representable by schemes, even over Spec(Q) or
suitable Spec(Z[1/N ])!
Hecke action on modular curves – perspective of algebraic geometry
• We now pass to the finite level K = K1 (N ) for some p - N .
• Corresponding modular curve XΓ1 (N ) over Z[ N1 ] represents pairs (E, Q)
over Z[ N1 ]-schemes S of an elliptic curve E over S with a point
Q ∈ E[N ](S).
• We start with hpi := S̃p . By the above, just need to multiply Q by p.
• For this, use action of (Z/N Z)× on XΓ1 (N ) .
• By the above, the correspondence for S̃p now becomes
XΓ1 (N )
p· id
∼

XΓ1 (N ) XΓ1 (N ) .

• Fact: For any a ∈ (Z/N Z)× , have canonical isomorphism a∗ ω = ω.

• Since the right map is an isomorphism, the integration over fibers is trivial.
• Everything extends to compactifications.
• Putting everything together, we have proved:
Proposition 8.3. Consider the operator defined as the composition
p∗
Mk (Γ1 (N ), Z[ N1 ]) = H 0 (XΓ∗1 (N ) , ω k ) −→ H 0 (XΓ∗1 (N ) , p∗ ω k ) = Mk (Γ1 (N ), Z[ N1 ]).
Then its base-change to C is the Hecke operator S̃p .
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM67

• For T̃p , instead of isogenies (ι : E → E 0 , α), parametrise triples (E, D, Q),

where D = ker ι finite cyclic subgroup of rank p. Get maps of moduli
functors
π1 :(E, D, Q)7→(E, Q),
π2 :(E, D, Q)7→(E/D, Q0 ), Q0 := Q + D = image of Q under E[N ] → (E/D)[N ]
• These induce a Hecke correspondence
XΓ0 (p)∩Γ1 (N )
π2 π1

XΓ1 (N ) XΓ1 (N ) .
• Fact: π1 and π2 extend to compactifications XΓ∗0 (p)∩Γ1 (N ) → XΓ∗1 (N ) .
• Fact: π1 and π2 are finite flat of degree p + 1, even on compactifications.
• Fact: There are canonical isomorphisms π1∗ ω = π2∗ ω (over Spec(Q))
• Since π1 has finite fibers, the integral occurring in the definition of the
Hecke operator becomes a discrete sum, which we can interpret as a trace
map
Trπ : π1,! π1∗ ω → ω
• Combining all this, we have proved:
Proposition 8.4. The base-change to C of the operator
π∗ Trπ
H 0 (XΓ∗1 (N ) , ω ⊗k ) −→
2
H 0 (XΓ∗1 (N )∩Γ0 (p) , ω ⊗k ) −−−→
1
H 0 (XΓ∗1 (N ) , ω ⊗k ).
is the Hecke operator T̃p .
• Note: This is all defined already over Q (even over Z[ N1 ])! Consequence:
Corollary 8.5. The eigenvalues of S̃p , T̃p on Mk (Γ1 (N ), C) are algebraic.
• To give a more explicit description of T̃p , we reinterpret this in terms of
divisors:
• Let Div(XΓ1 (N ) ) = set of divisors on X (≈ formal sums of points).
• We can then reinterpret hpi := S̃p and T̃p as operators on Div(XΓ1 (N ) ) :
S̃p : Div(XΓ1 (N ) ) → Div(XΓ1 (N ) ), [E, Q] 7→ [E, p · Q],
X
T̃p : Div(XΓ1 (N ) ) → Div(XΓ1 (N ) ), [E, Q] 7→ [E/D, Q + D],
D⊆E[p]

where E is an elliptic curve, and Q ∈ E[N ] a point of order N .

Modular curves in characteristic p
• Let E be an elliptic curve over a scheme S of characteristic p.
• Then E[N ] is étale for all p - N .
• The morphism [p] : E → E factors into
F V
→ E (p) −→ E,
E−
the Frobenius and Verschiebung isogeny. Here E (p) = E ×S,F S.
⇒ always have ker F ⊆ E[p]. This is a connected, i.e. ker F (S) = 0.
Definition 8.6. Over S = Fp , there are two cases:
68 JOHANNES ANSCHÜTZ

(1) E(Fp ) = Z/pZ. Then E is called ordinary, and E[p] = µp × Z/pZ.

(2) E(Fp ) = 0. Then E is called supersingular, and E[p] is connected.
• “supersingular” is not about smoothness. It just means “really special”:
• Fact: There are only finitely many isomorphism classes of supersingular
elliptic curves over Fp . All others are ordinary.
• Recall: XΓ1 (N ) is defined over Z[ N1 ].
• Can therefore reduce to Fp to get modular curve XΓ1 (N ),Fp → Spec(Fp ).
• Similarly for XΓ1 (N )∩Γ0 (p) representing (E, D, Q) with D ⊆ E[p] of rank p.
• Deligne–Rapoport: The fibre XΓ1 (N )∩Γ0 (p),Fp → Spec(Fp ) is of the form

D = µp
XΓ1 (N )∩Γ0 (p),Fp
D = Z/pZ

id F

XΓ1 (N ),Fp

• Two copies of XΓ1 (N ),Fp , corresponding to D = µp or D = Z/pZ as sub-

group of E[p], with transversal intersections at supersingular points (black).
• Projection is identity (deg=1) on one copy, and Frobenius (deg=p) on other.
Eichler–Shimura relation [Diamond–Shurman], [Conrad: Appendix to
Serre’s...]
• Eichler–Shimura relation expresses reduction T̃p mod p in terms of Frobe-
nius:
• Base change to algebraic closure Fp .
• Consider morphism Div0 (XΓ∗ (N ),F ) → Pic0 (XΓ∗ (N ),F )
1 p 1 p

• Fact: Pic0 (XΓ∗ (N ),F ) is points of a group scheme over Fp . In particular,

1 p
multiplication by p factors through Frobenius F : there is V = F ∨ s.t.
p=V ◦F =F ◦V
Theorem. (Eichler–Shimura relation) In Pic0 (XΓ∗ ), we have
1 (N ),Fp

T̃p = F + hpiV.
• Note: There is a version for XΓ∗ , which famously reads
0 (N ),Fp

T̃p = F + V

We get a second important variant by multiplying by F and using F F ∨ =

p.
F 2 − T̃p F + hpip = 0.
Proof. (Sketch) Let E be an ordinary elliptic curve over Z̆p = W (Fp ). Let C ⊆ E[p]
be the “canonical subgroup”:= generated by kernel of
E[p](Z̆p ) → E[p](Fp ).
LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM69

• Key fact: For D ⊆ E[p] cyclic subgroup scheme of rank p,

(p)
(1) the isogeny E → E/D reduces to F : E → E ⇔ D = C.
(p−1 )
(2) the isogeny E → E/D reduces to V : E → E ⇔ D 6= C.
(p−1 )
• Here E := E ×Fp ,F −1 Fp = base-change along inverse of Frobenius.
• F sends Q ∈ E[N ] to base-change Q(p) ∈ E (p) [N ] = E[N ](p) .
−1 −1
• Since V ◦ F = p on E[N ], this implies: V sends Q to pQ(p ) ∈ E (p ) [N ].
• Recall: There are p + 1 different subgroups D of E[p] over W (Fp ).
P
• Thus T̃p ([E, Q]) = D⊆E[p] [E/D, Q + D] in Div(XΓ1 (N ) ) reduces to
−1 −1
[E (p) , Q(p) ] + p[E (p )
, pQ(p )
] in Div(XΓ1 (N ),Fp ).
• The first summand is F [E, Q].
−1 −1 −1 −1
• By definition, hpi[E (p ) , Q(p ) ] = [E (p ) , pQ(p ) ].
• Now pass to Pic0 (XΓ∗ (N ),F ). Here: p = V F . We therefore have
1 p

(p−1 ) (p−1 )
p[E ,Q ] = pF −1 [E, Q] = V [E, Q].
(technically, we need to work with [E, Q] − [E 0 , Q0 ] to be in Div0 ). 
70 JOHANNES ANSCHÜTZ

9. Galois representations associated to newforms

Next aim:
• Realize newforms in cohomology.
• Finish construction of Galois representations associated to newforms of
weight k = 2.
• Give hints on how to proceed in the case that k ≥ 2.
We will associate classes in singular cohomology associated to modular forms (of
weight 2) by using differential forms. For this we need to recall some facts on de
Rham cohomology.
De Rham cohomology (cf. [BT13]):
• Let X be a real manifold.
• Let C ∞ (X) be the space of smooth functions f : X → C.
• Let Ai (X) be the space of smooth i-forms on X, i ≥ 0.
• The de Rham complex of X is the complex

d d
A• (X) : A0 (X) −
→ A1 (X) −
→ A2 (X) → . . .
=C ∞ (X)

with d the exterior derivative.

• The de Rham cohomology of X is then defined as the cohomology of the
de Rham complex
∗
HdR (X) := H ∗ (A• (X))
• Fascinatingly, the de Rham comparison isomorphism implies
∗
HdR (X) ∼
= H ∗ (X, C)
with RHS the sheaf cohomology of the constant sheaf C on X. This has the
interesting consequence that the de Rham cohomology is a purely topolog-
ical invariant of X.
– Sketch of proof: Sending U ⊆ X open to A• (U ) defines a complex of
sheaves
A•X
on X which is a resolution of the constant sheaf C (by the Poincaré
lemma). Using a partition of unity argument (needs X paracompact)
the sheaves AkX , k ≥ 0, are flasque. ⇒ RΓ(X, C) ∼
= Γ(X, A•X ).
• Moreover,
H ∗ (X, C) ∼
= H ∗ (X, C).
sing

with RHS = singular cohomology of X with values in C.

• If Γ is a discrete group acting properly discontinuously and freely on X,
then
d d
A• (Γ\X) ∼
= A• (X)Γ = ( A0 (X)Γ −
→ A1 (X)Γ −
→ A2 (X)Γ → . . .),
=C ∞ (X)Γ

i.e., the invariants in the de Rham complex calculate the cohomology of the
quotient Γ\X. Note that due to the potential presence of group cohomology
the statement does not hold on cohomology groups.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM71

• Finally, we set
∗
HdR (GL2 (Q)\(GL2 (Af ) × H± )) := lim HdR
∗
(GL2 (Q)\(GL2 (Af )/K × H± ))
−→
K

and
H ∗ (GL2 (Q)\(GL2 (Af ) × H± ), C) := lim H ∗ (GL2 (Q)\(GL2 (Af )/K × H± ), C),
−→
K

where the colimit is over all compact-open subgroups K ⊆ GL2 (Af ), and
the transition maps are induced by pullback.
• Then, for i ≥ 0, the (GL2 (Af )-equivariant) de Rham comparison
i
HdR (GL2 (Q)\(GL2 (Af ) × H± )) ∼
= H i (GL2 (Q)\(GL2 (Af ) × H± ), C)
holds by passing to the colimit.
Upshot:
• Can construct classes in H ∗ (X, C) using (closed) differential forms.
Cohomology classes associated to modular forms of weight 2:
• Note that
GL2 (Q)\(GL2 (Af ) × H± ) ∼
= GL2 (Q)+ \(GL2 (Af ) × H),
where GL2 (Q)+ ⊆ GL2 (Q) is the subgroup of elements of positive determi-
nant.
• Pick f ∈ H 0 (GL2 (Q)+ \(GL2 (Af ) × H), ω ⊗k ), k ∈ Z.
• We view f as a function
f : GL2 (Af ) × H → C, (g, z) 7→ f (g, z)
satisfying
f (γg, γz) = (cz + d)k f (g, z)
 
a b
for γ ∈ ∈ GL2 (Q)+ .
c d
• Assume that f is of weight k = 2.
• Then the differential form f (g, z)dz on GL2 (Af ) × H is closed and satisfies
γ ∗ (f (g, z)dz) = f (γg, γz)γ ∗ dz = det(γ)f (g, z)dz
for γ ∈ GL2 (Q)+ .
• Indeed:
– Closedness follows from holomorphicity as
∂ ∂
d(f (g, z)dz) = f (g, z)dz ∧ dz − f (z)dz ∧ dz = 0.
∂z ∂z
 
a b
– If γ = ∈ GL2 (Q), then
c d
det(γ)
γ ∗ (dz) = dz.
(cz + d)2
– Using that f is of weight 2, we can conclude.
72 JOHANNES ANSCHÜTZ

• Let
| − | = | − |adélic : Q× \A× → R>0
be the adélic norm, i.e.,
Y
|(x2 , x3 , . . . , x∞ )| := |xp |p · |x∞ |∞
p

with | − |p : Qp → R≥0 , x 7→ p−vp (x) the p-adic norm, and | − |∞ : R → R≥0

the real norm.
• | − | defines the character
χno := | − | ◦ det = |det|adélic
of GL2 (A), which is trivial on GL2 (Q).
• Thus, for
f˜(g, z) := χno ((g, 1))f (g, z)
∼ GL2 (Af ) × GL2 (R)), the form
(with (g, 1) ∈ GL2 (A) =
ηf := f˜(g, z)dz
on GL2 (Af ) × H is GL2 (Q)+ -equivariant.
• We get a map, which is not GL2 (Af )-equivariant,
1
α : M2 → HdR (GL2 (Q)+ \(GL2 (Af ) × H)) ∼
= H 1 (GL2 (Q)+ \(GL2 (Af ) × H), C)
by sending f to [ηf ].
• For a smooth GL2 (Af )-representation V and a smooth character
χ : GL2 (Af ) → C×
define V (χ) := V ⊗C χ.
• Then
α : M2 (χno ) → H 1 (GL2 (Q)+ \(GL2 (Af ) × H), C)
is GL2 (Af )-equivariant.
• Let K ⊆ GL2 (Af ) be compact-open, and set
XK := GL2 (Q)+ \(GL2 (Af )/K × H).
with canonical compactification
∗
XK := GL2 (Q)+ \(GL2 (Af ) × H∗ )
where H∗ := H ∪ P1 (Q) (equipped with Satake topology).
∗
• If f ∈ S2 (K), then ηf extends to a holomorphic differential form on XK .
2πiz
– Indeed: If q = e , then
dq = 2πiqdz,
i.e.,
1
dz =dq.
2πiq
Now express f (g, z) at each cusp in Fourier expansion, i.e., as a func-
tion of q.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM73

• The diagram
1
HdR ∗
(XK ) / H 1 (XK )
dR

' '
 
H 1 (XK ∗
, C) / H 1 (XK , C)
commutes.
• Moreover, the morphism
∗ ∗
Hc1 (XK , C) → H 1 (XK , C) = Hc1 (XK , C)
is surjective as its cokernel embeds into H 1 ({cusps}, C) = 0.
• Thus, α(S2 (K)) lies in the interior cohomology H e 1 (XK , C) ∼ ∗
= H 1 (XK , C)
of XK , which by definition is the image of the compactly supported coho-
mology Hc1 (XK , C) → H 1 (XK , C).
• Note that the above discussion applies similarly to anti holomorphic mod-
ular forms, i.e., complex conjugates of modular forms.
• This yields the map
α : S2 (K) → He 1 (XK , C), f 7→ [χno f dz],
where S2 (K) denotes the C-linear space of antiholomorphic modular forms.
• We can pass to infinite level and obtain the morphism
α ⊕ α : S2 ⊕ S2 → H e 1 (GL2 (Q)\(GL2 (Af ) × H± ), C)
with (hopefully) self-explaining notation.
Theorem 9.1 (Eichler–Shimura). The map
e 1 (GL2 (Q)\(GL2 (Af ) × H± ), C)
α ⊕ α : S2 (χno ) ⊕ S2 (χno ) → H
is a GL2 (Af )-equivariant isomorphism.
• The statement is equivalent to the analogous statement for all (sufficiently
small) compact-open subgroups K ⊆ GL2 (Af ).
• Thus, fix some K ⊆ GL2 (Af ) compact-open.
• The proof will exploit the ∪-product pairing
integrate
Hc1 (XK , C) × H 1 (XK , C) → Hc2 (XK , C) −−−−−→ C,
which under the de Rham comparison is induced from the pairing
Z
(η1 , η2 ) 7→ η1 ∧ η2
XK

of differential forms, where η1 has compact support (cf. [BT13]).

• We will use the following observation:
– We can calculate the compactly supported cohomology via Hc∗ (XK , C) ∼ =
H ∗ (A•c (XK )), where A•c denotes differential forms with compact sup-
ports.
– The canonical map
(2) A•c (XK ) → A•rd (XK )
is a quasi-isomorphism, where the RHS denotes differentials forms of
rapid decay, cf. [Bor80, Theorem 2].
• For f ∈ S2 the form ηf is rapidly decreasing.
74 JOHANNES ANSCHÜTZ

– Use that |e2πinz | = e−2πnIm(z) decreases rapidly if Im(z) → ∞ and

n ≥ 1.
• Thus α lifts naturally to a map
e : S2 (K) → Hc1 (XK , C).
α
• Consider f ∈ S2 (K), g ∈ S2 (K). Then:
∗ α
e(f ) ∪ α(f ) = 0 as dz ∧ dz = 0.
∗ Similarly for g.
∗ Using Stokes’ theorem, one proves
Z
α(f ) ∪ α(g)) =
tr(e ηf ∧ ηg .
XK
∗ This implies that the Poincaré pairing induces (up to a scalar) the
Petterson scalar product aka L2 -pairing h−, −i on S2 .
• We can now prove that β := α ⊕ α is injective.
• Indeed:
– If β(f + g) = 0, then
β(f + g) ∪ α(f ) = hf, f i = 0,
i.e., f = 0. Similarly, g = 0.
• Now we check H e 1 (XK , C) = 2dim(S2 (K)). If we succeed, the proof of the
Eichler-Shimura isomorphism is finished.
• Namely (we use lower case latters to denote dimensions):
∗ ∗ ∗
∗ eh1 (XK , C)) = h1 (XK , C) = 2 · h0 (XK , C) − χtop (XK )
∗ Here: χtop is the topological Euler characteristic.
∗ On the other hand 15:
dim(S2 (K))
∗
= h0 (XK , Ω1X ∗ )
K
RR ∗ ∗
= deg(Ω1X ∗ ) + χhol (XK ) + h1 (XK , Ω1X ∗ )
K K
∗ 0 ∗
= −χhol (XK ) + h (XK , C)
∗ Here: χhol is the holomorphic Euler characteristic.
∗ ∗
∗ χtop (XK ) = 2χhol (XK )
We can now finish our outlined strategy for associating Galois representations fo
newforms (at least in weight 2).
Galois representations associated to newforms:
• Recall that in the last two lectures we introduced a scheme
Xb → Spec(Q)
with GL2 (Af )-action such that naturally
X(C)
b ∼
= GL2 (Q)\(GL2 (Af ) × H± ).
• For K ⊆ GL2 (Af ) compact-open (plus sufficiently small), we get a quasi-
projective, smooth curve
XK → Spec(Q)
with
XK (C) ∼
= GL2 (Q)\(GL2 (Af )/K × H± )
15We use implicitly that ω 2 (−cusps) ∼ Ω1 .
=
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM75

(note the clash in notation with previous section, where XK was equal to
RHS).
• Fix a prime `.
• For K ⊆ GL2 (Af ) can consider the interior étale cohomology
e 1 (X 1 1 ∼ 1 ∗
Hét K,Q , Q` ) := Im(Hc,ét (XK,Q , Q` ) → Hét (XK,Q , Q` )) = Hét (XK,Q , Q` ).

• In the limit, we get the big cohomology space

1 b e 1 (X
H
e ét (XQ , Q` ) := lim H , Q ).
−→ ét K,Q `
K
1 1
• By naturality of Hc,ét , Hét this space has an action of
Gal(Q/Q) × GL2 (Af ),
i.e., an action of Gal(Q/Q) and an action of GL2 (Af ), and these two actions
commute.
• Fix an isomorphism ι : Q` ∼= C.
• Using étale comparison theorems (and ι) we get the isomorphisms
ι
Hét
b , Q` ) ∼
e 1 (X 1 b
= Hét (XC , Q` ) ∼
=He 1 (X(C),
b Q` ) ∼
=He 1 (X(C),
b C).
Q

• These isomorphisms are GL2 (Af )-equivariant.

• Let π ⊆ S2 be an irreducible GL2 (Af )-representation.
• Using ι, we will view each smooth GL2 (Af )-representation over C as a
smooth representation over Q` .
• The space
ρeπ := HomGL (A ) (π(χno ), He 1 (X
b , Q` ))
2 f ét Q
is two-dimensional by the Eichler-Shimura isomorphism
e 1 (X(C),
H b C) ∼
= S2 (χno ) ⊕ S2 (χno )
and because S2 is multiplicity free as a GL2 (Af )-representation.
• In summary, as a Gal(Q/Q) × GL2 (Af )-representation
(XQ , Q` ) ∼
M
1 b
(3) He ét = ρeπ ⊗Q` π(χno ),
π
where the sum is running over all irreducible GL2 (Af )-representations π ⊆
S2 (note that there is a natural morphism from the RHS to the LHS).
As we tried to explain the importance of the Langlands reciprocity for newforms
lies in the fact that Hecke eigenvalues match with traces of Frobenii. We will discuss
this now.
The Eichler–Shimura relation in étale cohomology:
B We need to show that Hecke eigenvalues match with traces of Frobenii. B
• Recall: To π ⊆ Sk with system of Hecke eigenvalues
{ap (π), bp (π) = χ(p)pk−1 }p∈S
/ ,

we want to attach a 2-dimensional `-adic representation

ρπ : Gal(Q/Q) → GL2 (Q` ),
/ S each arithmetic Frobenius Frobarith
such that for p ∈ p at p has character-
istic polynomial
X 2 − ap (π)X + χ(p)pk−1
76 JOHANNES ANSCHÜTZ

(we omit ι, ρπ in the following), i.e.,

(Frobarith
p )2 − ap (π)Frobarith
p + χ(p)pk−1 = 0.
• By passing to K1 (N )-invariants in (Equation (3)), we have (for N ≥ 3)
(XΓ∗1 (N ),Q , Q` ) ∼
M
1
Hét = ρeπ ⊗Q` π(χno )K1 (N )
π

(note XK1 (N ) ∼
= XΓ1 (N ) ).
• Let p be a prime, p - `N .
• Fix a place of Q over p. This determines an algebraic closure Fp of Fp .
• Recall: For an abelian variety A over a field L, and ` a prime, we write
T` A = lim A[`n ](L), V` A = T` A ⊗Z` Q`
←−
for the `-adic Tate module resp. the rationalized `-adic Tate module.
Lemma 9.2. Let p be prime, p - `N . Then we have
1
Hét (XΓ∗1 (N ),Q , Q` ) = V` Pic0 (XΓ∗1 (N ),Fp )∨ ,
where (−)∨ denotes the Q` -dual.
`n
Proof. The Kummer sequence 1 → µ`n → Gm −→ Gm → 1 implies
1
Hét (XΓ∗1 (N ),Q , Q` (1)) ∼
= V` Pic0 (XΓ∗1 (N ),Q ).
The Weil pairing yields a canonical isomorphism
V` Pic0 (X ∗ )∨ =∼ V` Pic0 (X ∗ )(−1).
Γ1 (N ),Q Γ1 (N ),Q

Finally, because ` 6= p
V` Pic0 (XΓ∗1 (N ),Q ) ∼
= V` Pic0 (XΓ∗1 (N ),Q ) ∼
= V` Pic0 (XΓ∗1 (N ),Fp )
p

by lifting torsion points. 

• In particular, the Galois representation
1
Hét (XΓ∗1 (N ),Q , Q` )
is unramified at p.
• On V` Pic0 (XΓ∗ (N ),F ), we have the Eichler–Shimura relation, which we re-
1 p
call reads
F 2 − Tep F + pSep = 0
with F the arithmetic Frobenius.
Proposition 9.3. Let p be a prime with p - `N . Then on Hét1
(XΓ∗ (N ),Q , Q` ), we
1
have
(Frobgeom
p )2 − Tep Frobgeom
p + pSep = 0.
• For a normalized newform f ∈ S2 (Γ0 (N ), χ) we define finally
ρπ )∨
ρf := (e
with π the irreducible GL2 (Af )-representation generated by f , and
1 b
ρeπ := HomGL2 (Af ) (π(χno ), H
e ét (XQ , Q` )).
Some analysis of ρf , cf. [Rib77]:
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM77

• Write the (normalized) newform f ∈ S2 (Γ0 (N ), χ) in Fourier expansion

X
f (q) = an q n .
n≥1

Then for p - `N , Frobarith

p has characteristic polynomial
2
X − ap X + χ(p)p
on ρf .
• Indeed:
– By the (dual of the) Eichler–Shimura relation for Frobgeom
p we get
(Frobarith
p )2 − Tep Frobarith
p + pSep
on ρf .
– Recall that Tep has eigenvalue pap on π K1 (N ) , while Sep has eigenvalue
χ(p)p2 .
– We had to twist the GL2 (Af )-action on S2 by χno = |det(−)|. Thus
Tep has eigenvalue ap on
ρeπ ⊗Q` π(χno )K1 (N ) ⊆ Hét
1
(XΓ∗1 (N ),Q , Q` ).
1 2
while Sep has eigenvalue p2 χ(p)p = χ(p).
• The determinant of ρf is
ψ · χcyc ,
where
×
χcyc : GQ → Q`
denotes the cyclotomic character, and
× ι
ψ : Q× R>0 \A× → Q` ∼
= C×
the (finite order) adélic character determined by χ.
• Indeed:
– One uses Chebotarev and checks that both sides agree on arithmetic
Frobenius elements for p - `N .
• In particular, ρf is odd, i.e., the determinant of each complex conjugation
is −1.
– This uses that for all k ≥ 1 non-triviality of Sk (Γ0 (N ), χ) implies
χ(−1) = (−1)k .
• ρf is irreducible, cf. [Rib77, Theorem (2.3)].
• Because, ρf is realized in the étale cohomology of a proper, smooth scheme
over Q, the de Rham comparison theorem in `-adic Hodge theory implies
that ρf |GQ is de Rham.
`

Outline of the construction in weights ≥ 2:

• Instead of
1 b
H
e ét (XQ , Q` )
use
1 b
He ét (XQ , Lk ),
with the local system
Lk := Symk−1 R1 f∗ (Q` )
78 JOHANNES ANSCHÜTZ

for f : E → X b the universal elliptic curve.

k
• Problem: L does not extend to a local system on the compactification.
Thus one needs an additional argument to justify base change to Fp .
• Use similar strategy to obtain analogs of the Eichler–Shimura relation/isomorphism.
• Also replace the de Rham comparison/étale comparsion theorems by their
versions with coefficients.
• More details can be found in [Del71b].
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM79

10. Galois representations for weight 1 forms (by Ben Heuer)

Last time: Constructed Galois representations attached to newforms in weight
≥ 2:
Theorem 10.1 (Deligne ’71). Let N ≥ 3 and k ≥ 2. Let ε : (Z/N Z)× → C× be
a Dirichlet character. Let f ∈ Mk (Γ0 (N ), ε) be a newform of weight k. Let K be
the number field generated by the ap (f ) and ε. Then for any place λ of K over a
prime ` - N , there is a continuous, odd, irreducible Galois representation
ρ : GQ → GL2 (Kλ ),
unramified outside N , such that the characteristic polynomial of Frobp is
X 2 − ap X + ε(p)pk−1 for all p - N `.
Galois reps associated to weight 1 modular forms
• Goal of today: prove the following Theorem

Theorem 10.2 (Deligne–Serre ’73). Let N ≥ 3. Let ε : (Z/N Z)× → C× be

a Dirichlet character. Let f ∈ M1 (Γ0 (N ), ε) be a newform of weight 1. Then
there is a continuous, odd, irreducible Galois representation
ρ : GQ → GL2 (C),
unramified outside N , such that the characteristic polynomial of Frobp is
X 2 − ap X + ε(p) for all p - N.

• This finishes the construction of newforms Galois reps.

• Note that ρ has image in GL2 (C), i.e. is an Artin representation, in contrast
to the ones for weight ≥ 2, which were `-adic. Why is that?
– Easy fact: Any Artin representation has finite image.
– In fact, ρ is defined already over a number field, so could embed into
p-adic field Kλ as before. But using C is a way of signaling “finite
image”.
– On the Galois side, see that ρ can only be finite for k = 1 since the
determinant is required to be det ρ = εχk−1 cycl , and χcycl has infinite
image.
• If ε is even, then M1 (Γ0 (N ), ε) = 0. Hence ε is odd, which implies ρ odd.
First sketch of construction
Compared to last time, the proof will be less geometry, more Galois representa-
tion theoretic:
• Reduce f mod p. This gives a “mod p modular form”
• Use congruences of modular forms to show there is a mod p Hecke eigenform
f 0 with same eigenvalues but higher weight ≥ 2.
• Lift f 0 back to characteristic 0 using “Deligne–Serre lifting”
• Attach Galois representation ρf 0 using weight ≥ 2 construction
• Reduce Galois representation mod `
• Lift Galois representation from F` to C.
80 JOHANNES ANSCHÜTZ

Introduction to mod p modular forms

• Let N ≥ 3. Let S be any Z[ N1 ]-scheme S.
• Recall: We had defined modular forms over S of weight k by
Mk (Γ1 (N ); S) := Γ(XΓ∗1 (N ),S , ω k ).
• ω is the ample line bundle of relative differentials on universal elliptic curve
• Can define q-expansions: Let S = Spec(A), then have the Tate curve T (q)
over Z((q)) ⊗Z A. Over Spec(Z((q)) ⊗Z A), have canonical isomorphism
ωGm = ωT (q)
• bundle ωGm is trivial: the canonical differential dxx on Gm is invertible
section
• Via this trivialisation, get for every cusp of XΓ∗1 (N ),A an associated q-
expansion
Mk (Γ1 (N ); S) → A[[q]].
Theorem 10.3 (Katz’ q-expansion principle). This is injective for all k at each
cusp.
• Now apply this all to S = Spec(F) where F is a finite field of characteristic
p: The resulting modular forms are called mod p modular forms.
• The following fact perhaps helps you get a feeling for mod p modular forms:
Lemma 10.4. For weight k ≥ 2, any mod p modular form is the reduction
of a modular form over a number field modulo p. In particular, in this case
Mk (Γ1 (N ); F) = Mk (Γ1 (N ); Z[ N1 ]) ⊗Z F.
Proof. Using Riemann–Roch, one can show that H 1 (XΓ∗1 (N ) , ω ⊗k ) = 0. 
• However, it is often better to have a more intrinsic definition. For example:
Example of mod p modular form: The Hasse invariant
• Recall: For any elliptic curve in characteristic p, have Verschiebung isogeny
V : E (p) → E.
• In particular, have this for the universal elliptic curve.
• On relative differentials, this defines a pullback map in the other direction
V ∗ : ω → ω ⊗p .
• Tensoring with ω −1 yields
V ∗ : O → ω ⊗(p−1) V ∗ ∈ Γ(ω ⊗(p−1) ) = Mp−1 (Γ1 (N ); F).
• This is the Hasse invariant, denoted by Ha.
• It is a mod p modular form of weight p − 1. In fact: an eigenform of level
1.
Proposition 10.5. The Hasse invariant has constant q-expansion = 1 ∈ Fp [[q]].
Proof. Sketch: Use ωT (q) = ωGm : On Gm have F = [p] and thus V = 1. 
• Note: 1 ∈ M0 (Γ1 (N ); Fp ) is also a mod p modular form with q-expansion
1.
• This does not contradict the q-expansion principle as the weights are dif-
ferent.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM81

Upshot: In characteristic p, we have a new phenomenon:

• Modular forms of different weight can have same q-expansion!
• E.g. for any modular form f of weight k, any n ∈ N, have modular form
Han · f of weight k + n(p − 1) with same q-expansion.
• In particular, this preserves eigenforms a system of eigenvalues for Hecke
operators can correspond to eigenforms of different weights.
• The multiplication map
·Ha
Mk (Γ1 (N ); S) −−→ Mk+p−1 (Γ1 (N ); S)
is injective (q-expansion principle) but not surjective!
• Reason: Ha has zeros, precisely on supersingular points of XΓ∗1 (N ) (Fp )
• This is one motivation for removing those points, which yields the theory
of p-adic modular forms.
Eisenstein series
• There is a more classical perspective on this that we want to at least men-
tion:
Definition 10.6.
For any even k ≥ 4, let
X 1
Gk := .
(m + nτ )k
(n,m)∈Z2
(n,m)6=(0,0)

This is a modular eigenform of weight k and level 1 with q-expansion

∞
X 4
2ζ(k)(1 − σk−1 (n)q n )
n=0
B k

where σk−1 (n) = d|n dk−1 and Bk is the k-th Bernoulli number.
P

Define the normalised Eisenstein series as

∞
1 4 X
Ek := Gk = 1 − σk−1 (n)q n ∈ Z(p) [[q]]
2ζ(k) Bk n=0

Lemma 10.7. Assume p ≥ 5, then in terms of q-expansions, we have

Ep−1 ≡ 1 mod p.
Proof. This follows from Kummer’s congruences for Bernoulli numbers. 

Upshot: For p ≥ 5, the reduction mod p of the Eisenstein series Ep−1 is = Ha:
Ep−1 ≡ Ha mod p “Deligne’s congruence”
Strategy (Deligne–Serre):
• Given eigenform of weight 1, reduce mod p.
• Multiply with Ha so it becomes eigenform of weight ≥ 2.
• Use construction from last lecture in this case.
Galois reps attached to mod p modular forms
82 JOHANNES ANSCHÜTZ

Theorem 10.8. Let k ≥ 1 and let f ∈ Mk (Γ1 (N ), ε; Fp ) be a mod p cuspidal

eigenform. Let Ff be the (finite) field generated over Fp by the ap (f ) and ε. Then
there is a semi-simple Galois representation
ρf : GQ → GL2 (Ff ),
unramified outside N ` such that for all ` - N p,
charpoly(Frobp |ρf ) = X 2 − ap X + ε(p)pk−1 .
Step 1: Reduce to case of k ≥ 2: If k = 1, multiply with Ha. This does not
change the q-expansion, hence is still an eigenform. But its weight is ≥ 2.
Step 2: Lift the Hecke eigensystem to characteristic 0. This is always possible
in weight ≥ 2, as we shall discuss next.
Deligne–Serre lifting lemma
• Let R be a Z[ N1 ]-algebra without zero-divisors. Let
TR ⊆ End(Mk (Γ1 (N ); R))
be the Hecke algebra: R-subalgebra generated by the Hecke operators Tl , Sl .
• A Hecke eigensystem is a ring homomorphism
ψ : TR → R.
The archetypal example of a Hecke eigensystem is:
• Let g be an eigenform in Mk (Γ1 (N ); R), this gives rise to a Hecke eigensys-
tem
ψg : TR → R, T 7→ aT s.t. T (g) = aT g.
Lemma 10.9. Assume that R = K is a field. Then any Hecke eigensystem
ψ : TK → K
comes from a unique Hecke eigenform.
Proof. • TK is an Artinian K-algebra.
• After passing to extension, it is the product of Artinian local rings Ti .
• Since TK acts faithfully on M := Mk (Γ1 (N ); K), the submodule Mi :=
Ti · M is non-zero (commutative algebra fact).
• Take any eigenvector in Mi . Then this has eigensystem ψ.
• Uniqueness: The Hecke eigenvalues determine the q-expansion. 
• Let K be a number field, p a place over p - N . Let Fp be the residue field.
• Write OK,(p) for the valuation subring of K of p-integral elements.
Lemma 10.10 (Deligne–Serre Lifting Lemma). Let k ≥ 2 and let g ∈ Mk (Γ1 (N ); Fp )
be an eigenform. Then there is a finite extension K 0 |K, an extension p0 |p and an
eigenform g̃ ∈ Mk (Γ1 (N ); OK 0 ,(p0 ) ) such that g̃ reduces to g mod p0 .
Proof. • Consider the ring homomorphism
ψg
TOK → TFp −−→ Fp .
• This corresponds to a maximal ideal m. Since
Spec(TOK ) → Spec(OK )
is locally free, can find prime ideal q ⊆ m such that q ∩ O = 0.
LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM83

• This defines a non-zero prime ideal of TK , corresponding to Hecke eigen-

system
TK → K 0 .
• By Lemma Lemma 10.9, this corresponds to an eigenform.
• Use uniqueness in Lemma Lemma 10.9 to see that this reduces to g. 
84 JOHANNES ANSCHÜTZ

Proof of Theorem Theorem 10.8

• Let f be mod p eigenform of weight ≥ 2, defined over extension Ff of Fp .
• Lift to eigenform f˜ over some number field K. Let p be a place over p.
• Associated to this, have Galois representation
ρf˜ : GQ → GL2 (Kp ).
• Since image is compact, can always find isomorphic representation of the
form
ρf˜ : GQ → GL2 (OK,p ).
• Reduce this mod p to get representation
ρf˜ : GQ → GL2 (Fp ).
• This might not yet be semi-simple Define ρf as semi-simplification
ρf := ρss
f˜

(:=direct sum of Jordan–Hölder factors). Same trace and determinant as

ρf˜.
• Since ker ρf ⊆ ker ρf˜, this is unramified outside of pN .
• Remains to show: ρf already defined over Ff : For this, STP:
ρf = σ ◦ ρf for all σ ∈ Gal(Fp |Ff ).
• Certainly, all characteristic polynomials of Frob` for ` - pN ,
X 2 − a` X + ε(`),
are preserved by σ, since a` , ε(`) ∈ Ff by definition. Now use:
Theorem 10.11 (Brauer–Nesbitt). Let F be a perfect field. Let G be a
finite group. Let V, W be two semi-simple finite dimensional representations
of G over F . Then V ∼ = W if and only if
charpoly(g|V ) = charpoly(g|W ) for all g ∈ G.
• This replaces the similar statement on traces from Section 6 if the coeffi-
cients are of positive characteristic.
• This finishes the construction of f ρf . 
Serre’s conjecture
• A quick digression before we continue with weight 1 forms:
• In 73’, Serre gave a conjecture that described exactly which mod p Galois
representations arise in this way. It became known as
Conjecture 10.12 (Serre’s Modularity Conjecture, or Serre’s Conjecture). Let
ρ : GQ → GL2 (Fp ) be a continuous Galois representation that is odd and irreducible.
Then there is a mod p eigenform g such that
ρ = ρg .
In 1987, Serre published an article refining this conjecture:
• He predicted the optimal level of g (in terms of the Artin conductor of ρ)
• He predicted the optimal weight (recipe in terms of ramification of ρ at p)
• He showed that his conjecture would imply Fermat’s Last Theorem.
• Serre’s Conjecture is now a theorem by Khare–Wintenberger (2008).
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM85

Lifting ρf to C
• Back to our earlier setup: f ∈ M1 (Γ0 (N ), ε) eigenform
• K number field generated by eigenvalues and ε.
• Let L := set of primes of Q which decompose completely in K. This is an
infinite set by Chebotarev.
• Let ` ∈ L and λ a place over ` in K. Let OK,(λ) ⊆ K be λ-integral elements.
• Now apply previous section with role of p played by varying ` ∈ L.
• For any such λ, consider reduction f of f mod λ. This is a mod ` eigenform.
• Theorem Theorem 10.8 associates a semi-simple Galois representation
ρf ,` : GQ → GL2 (F` )
(over F` as λ completely split) with charpoly of Frobp for any p - `N given
by
charpoly(Frobp |ρf ,` ) = X 2 − ap X + ε(p).
• Zeros of this are n-th units roots, which we may assume are all in OK (by
the next lemma) and thus in Fl . Thus
= (X − a)(X − b) for some a, b ∈ F×
` .

Definition 10.13. Let G` := im(ρf ,` ) ⊆ GL2 (F` ). This is a finite group.

Lemma 10.14. If f is of weight 1, there is A ∈ N such that |G` | ≤ A for all ` ∈ L.
Proof. After using a crucial estimate on the coefficients of weight 1 modular forms
this becomes purely group theoretical: use classification of subgroups of GL2 (F` ).
We shall not discuss it, but if you want to see it: [DS74, Lemma 8.4]. 
• Enlarge K so that it contains all n-th unit roots for n ≤ A. This shrinks
L.
• Also exclude the primes ≤ A from L. Then L is still infinite.
Lemma 10.15. Let G be a finite group of order prime to `. Then any representa-
tion ρ : G → GL2 (F` ) lifts to a representation
ρ̃ : G → GL2 (OK,(λ) ).
Proof. Step 1: Lift to GL2 (Z` ):
• STP: any G → GL2 (Z/`n Z) lifts to G → GL2 (Z/`n+1 Z).
• For this, consider the short exact sequence
1 → 1 + `n M2 (F` ) → GL2 (Z/`n+1 Z) → GL2 (Z/`n Z) → 1
• The first group is isomorphic to M2 (F` ) via 1 + `n x 7→ x.
• Now use non-abelian group cohomology: Consider the left-action
G acts on Map(G, GL2 (Z/`n Z) via g · ϕ := ϕ(g)ϕ(g)−1 .
• Then
Hom(G, GL2 (Z/`n Z)) = Map(G, GL2 (Z/`n Z))G .
• Apply Map to the above short exact sequence, then G gives a long exact
sequence of non-abelian group cohomology
· · · → Hom(G, GL2 (Z/`n+1 )) → Hom(G, GL2 (Z/`n )) → H 1 (G, Map(G, M2 (F` )))
• Fact in group cohomology: H 1 (G, A) = 0 if |G|, |A| finite and coprime.
86 JOHANNES ANSCHÜTZ

Step 2: Already defined over number field

• Since G is finite, ρ is semi-simple.
• Traces must be sums of roots of unity of order ≤ |G| ≤ A contained in
K.
• Brauer–Nesbitt ensures that ρ is already defined over K.
• Finally, K ∩ Z` = OK,(λ) 

Back to prove of Theorem Theorem 10.2 (Galois representation for weight 1 forms)
• Apply this to ρg : G` ,→ GL2 (F` ), get G` → GL2 (OK,(λ) )
• Compose with GQ → G` to get Galois representation
ρ̃g,` : GQ → GL2 (OK,(λ) ).
Proposition 10.16. The lift ρ̃g,` : GQ → GL2 (OK,(λ) ) is such that for all p - `N ,

charpoly(Frobp |ρ̃g,` ) = X 2 − ap X + ε(p).

The key is to vary `: Recall that we get ρ̃g,` for all ` ∈ L.
Definition 10.17. Let Y be the set of polynomials of the form (X − α)(X − β) for
α, β ∈ OK,(λ) roots of unity of order ≤ |A|. This is a finite set.

• Note: Given n ≤ A, ` ∈ L, and n-th root of unity x ∈ F× ` , there is exactly

one lift to root of unity in K (existence: n ≤ A, uniqueness: ` ≥ A).
• Consequence: If F, G ∈ Y , then F ≡ G mod λ implies F = G.
• Idea: Apply this to F = X 2 − ap X + ε(p) and G = charpoly(Frobp |ρ̃g,` )

Lemma 10.18. X 2 − ap X + ε(p) ∈ Y .

Proof. For all λ over a prime in L, there is F ∈ Y for which
charpoly(Frobp |ρg,` ) ≡ F mod λ
since ρg,` has finite image. Thus also
X 2 − ap X + ε(p) ≡ F mod λ.
Since Y is finite, there is F for which this holds for infinitely many λ. Thus
X 2 − ap X + ε(p) = F ∈ Y 

Proof of Proposition. • As G` finite of order ≤ A, have

charpoly(Frobp |ρ̃g,` ) ∈ Y
• charpoly(Frobp |ρ̃g,` ) ≡ charpoly(Frobp |ρg,` ) ≡ X 2 − ap X + ε(p) mod λ
• Thus both sides are in Y , which implies
charpoly(Frobp |ρg,` ) = X 2 − ap X + ε(p). 

Remains to prove:
• ρ is odd: This is simply because ε is odd and c has order 2.
• ρ is irreducible: Use complex analytic estimate due to Rankin 
This finishes the construction of eigenforms Galois representations!
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM87

Summary of the weight 1 construction

       
eigenforms congruence eigenforms of Deligne’s `-adic Galois complex Galois
of weight 1 weight ≥ 2 construction representations representations
reduce lift
reduce DS-lifting
     
mod ` eigenforms ·Ha mod ` eigenforms mod ` construction mod ` Galois
of weight 1 of weight ≥ 2 Theorem Theorem 10.8 representations
88 JOHANNES ANSCHÜTZ

11. The Langlands program for general groups, part I

Next aims:
• More or less precise statements (for parts of) Langlands program for GLn ,
or even general G reductive over Q.
Langlands(-Clozel-Fontaine-Mazur) reciprocity for GLn,F :
• Let F be an arbitrary number field.
• Fix some prime ` and an isomorphism C ∼ = Q` .
• The Langland(-Clozel-Fontaine-Mazur) reciprocity conjecture states that
for any n ≥ 1, there exists a (unique) bijection between
i) the set of L-algebraic cuspidal automorphic representations of GLn (AF ),
ii) the set of (isomorphism classes) of irreducible continuous representa-
tions Gal(F /F ) → GLn (Q` ) which are almost everywhere unramified,
and de Rham at places dividing `,
such that the bijection matches Satake parameters with eigenvalues of
Frobenius elements.
Comments:
• This is only a part of the Langlands program for G = ResF/Q GLn,F .
• This does not describe all automorphic representations for GLn,F , nor all
Galois representations.
• Not known for F = Q, n ≥ 2, e.g., there exists L-algebraic cuspidal au-
tomorphic representations of GL2 (A) (associated to Maassforms), which
should correspond to even irreducible, 2-dimensional `-adic representations
of Gal(Q/Q) (with finite image).
• Langlands-Tunnell: If the image is solvable, automorphicity of the Galois
representation is known, cf. [GH19, Theorem 13.4.5.]. This does not cover
almost everywhere unramified 2-dimensional, irreducible Galois represen-
tations with image SL2 (F5 ).
• Unicity follows from the (strong) multiplicity one theorems on both sides.
• Today: Explain what “L-algebraic” means.
Automorphic forms for general groups:
• G an arbitrary reductive group over Q.
• Will introduce a convenient (=more algebraic) replacement for L2 ([G]).
• Recall for G = GL2 = GL2,Q , k ∈ Z, the embedding
Φ : Mk → C ∞ (GL2 (A)), f 7→ ϕf
from Section 3.
• Recall the description of the image of Φ(Mk ) resp. Φ(Sk ), namely:
– For k ∈ Z an element ϕ ∈ C ∞ (GL2 (A)) lies in the image of
Φ : Mk → C ∞ (GL2 (A))
if and only if
∗ ϕ(g, g∞ z) = z −k ϕ(g, g∞ ) for (g, g∞ ) ∈ GL2 (A), z ∈ C× ⊆
GL2 (R).
∗ ϕ is of moderate growth (implies the vanishing of negative Fourier
coefficients).
∗ For each g ∈ GL2 (Af )
Y ∗ ϕ(g, −) = 0,
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM89

where Y ∈ gC = gl2 (R)C is a suitably constructed, natural ele-

ment (implies holomorphicity).
∗ ϕ(γg) = ϕ(g) for all γ ∈ GL2 (Q).
– f is moreover cuspidal if and only if
Z
ϕf (ng)dn = 0
N (Q)\N (A)

for all unipotent radicals N in proper parabolics P ⊆ G (defined over

Q).
• Fix a maximal compact (usually not connected) subgroup  K∞ of G(R),
−1 0
e.g., O2 (R) = SO2 (R) ∪ sSO2 (R) ⊆ GL2 (R) with s = .
0 1
Definition 11.1 (cf. [GH19, Definition 6.5]). Let G be a reductive group over Q.
An (adélic) automorphic form for G is a smooth function ϕ : G(A) → C such that
1) ϕ is of moderate growth (was defined in Section 4, not important for this
lecture).
2) ϕ is right K∞ -finite, i.e., the functions g 7→ ϕ(gk) for k ∈ K∞ ⊆ G(A)
span a finite dimensional vector space.
3) ϕ is killed by an ideal in Z(gC ) of finite codimension, where gC denotes the
(complexified) Lie algebra of G(R), and Z(gC ) the center of the enveloping
algebra U (gC ) of gC .
4) ϕ is left G(Q)-invariant, i.e., ϕ(γg) = ϕ(g) for all γ ∈ G(Q).
The Casimir element for GL2 :
• Let us check that 3) is indeed verified for ϕf : C ∞ (GL2 (A)) → C with
f ∈ Mk .
• Namely, the element Y was not in Z(gl2,C ).
• Set
       
0 −i 1 −1 −i 1 −1 i 1 0
H := , X := , Y := , Z := .
i 0 2 −i 1 2 i 1 0 1
• Then
1)
H ∗ ϕf = −kϕf
−k
as z · ϕf = z ϕf for z ∈ C× .
2) Y ∗ ϕf = 0.
• [X, Y ] = XY − Y X = H (because H, X, Y form an sl2 -triple)
• Define the Casimir operator
1
∆ := (H 2 + 2XY + 2Y X).
4
Then ∆ ∈ Z(gl2,C ).
• In fact, Z(gl2,C ) is generated, as a C-algebra, by ∆ and Z ∈ gl2,C (this
follows from the Harish-Chandra isomorphism, cf. [GH19, Theorem 4.6.1]).
• One calculates (using Y ∗ ϕf = 0)
1 1 1
∆ ∗ ϕf = (H 2 − 2XY + 2Y X) ∗ ϕf = (H 2 − 2H) ∗ ϕf = (k 2 − 2k)ϕf ,
4 4 4
thus ϕf is indeed killed by the ideal h∆ − 41 (k 2 − 2k), Z + ki ⊆ Z(gl2,C ) of
finite codimension.
90 JOHANNES ANSCHÜTZ

More on automorphic forms:

• G arbitrary reductive over Q.
• Set A(G) as the space of automorphic forms, and A([G]) as the subspace of
automorphic forms, which are invariant under AG (acting via left or right
translations on G(A)).
• Unfortunately, the spaces A([G]) and A(G) are not stable under G(R),
because K∞ -finiteness need not be preserved.
Example that K∞ -finiteness is not preserved (following Andreas Mi-
hatsch):
• Consider G = GL2 (R), K := K∞ ◦
= SO2 (R) ∼ = S 1 , and V any representa-
∞
tion on functions on G containing Cc (G). We give an example showing
that K-finiteness is not preserved under the action by right translation.
• More precisely, there exists ϕ ∈ Cc∞ (G), and h ∈ G such that ϕ is K-finite,
but not Kh := h−1 Kh-finite.
• For this: Pick h ∈ G, such that Kh ∩ K = {±1}.
• Then choose a non-zero ϕ ∈ Cc∞ (G) invariant under K, such that ϕ ≡ 0 on
an open subset of Kh (e.g., construct ϕ by pullback a function from G/K,
note that the image of Kh in G/K is of dimension > 0).
• No non-zero Kh -finite function on Kh can vanish on an open subset. Indeed,
the Kh -finite functions on Kh ∼
= S 1 are precisely the restrictions of Laurent-
polynomials.
More on automorphic forms (continued):
• Clearly, G(Af ) acts on A(G) via ϕ 7→ (g 7→ ϕ(gh)) for h ∈ G(Af ).
• Similarly, K∞ , even AG K∞ , acts on A(G)
• Less clear, but true (cf. [GH19, Proposition 4.4.2.]): gC acts on A(G).
⇒ U (gC ), in particular Z(gC ) ⊆ U (gC ), acts on A(G) by differential operators.
• The gC and K∞ -action make A(G) into a (gC , K∞ )-module, in the following
sense.
Definition 11.2 (cf. [GH19, Definition 4.5.]). A (gC , K∞ )-module is a C-vector
space V , which is simultaneously a gC - and K∞ -module, such that
1) V is a union of finite dimensional K∞ -stable subspaces.
2) For X ∈ k := Lie(K∞ )C and ϕ ∈ V we have
∂
X ∗ϕ= (exp(tX)ϕ)t=0 .
∂t
3) For h ∈ K∞ , X ∈ gC and ϕ ∈ V we have
h(X ∗ (h−1 ϕ) = (Ad(h)(X)) ∗ ϕ.
Comments:
• We required no topology on V , thus (gC , K∞ )-modules are much more
algebraic than G(R)-representations, and thus easier to handle.
• In 2) the limit is taken for the natural topology of a finite dimensional
K∞ -stable subspace of V containing ϕ.
• For each representation of G(R) on some Hilbert space C-vector space, the
space of smooth and K∞ -finite vectors is naturally a (gC , K∞ )-module, cf.
[GH19, Proposition 4.4.2.], and on unitary representations one does not
loose any informations, cf. [GH19, Theorem 4.4.4.], namely: isomorphism
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM91

classes of irreducible unitary representations of G(R) inject into isomor-

phism classes of irreducible (gC , K∞ )-modules.
• Moreover, the (gC , K∞ )-modules obtained in this way are admissible, i.e.,
for each (finite dimensional) irreducible representation σ of K∞ , the σ-
isotypic component is finite dimensional, cf. [GH19, Theorem 4.4.1.].
• Version of Schur’s lemma ⇒ On each irreducible, admissible (gC , K∞ )-
module π the center Z(gC ) of U (gC ) acts via some morphism
ωπ : Z(gC ) → C
of C-algebras, called the “infinitesimal character of π”.
• The Harish-Chandra isomorphism, cf. [GH19, Theorem 4.6.1.], furnishes
Z(gC ) ∼
= U (tC )W
for each Cartan subalgebra t ⊆ g and W the corresponding Weyl group.
Note U (tC ) ∼
= Sym• tC because tC is an abelian Lie algebra.
• As W is finite, C-algebra morphisms
Z(gC ) ∼
= U (tC )W → C
correpond bijectively to W -orbits of C-algebra homomorphisms
U (tC ) ∼
= Sym• tC → C.
Moreover,
HomC−alg (U (tC ), C) ∼
= HomC−lin (tC , C) =: t∨
C.

U (tC )W → U (tC ) is finite.

• Thus, the infinitesimal character defines a canonical W -orbit of elements in
the “weight space” t∨ C.
• Now assume that t = Lie(T ) for some maximal torus T ⊆ GR . Then
W -equivariantly
X ∗ (TC ) ⊆ t∨
C.
• An irreducible, admissible (gC , K∞ )-module π is called L-algebraic if the
W -orbit in t∨C corresponding to the infinitesimal character ωπ lies in the
finite free Z-module X ∗ (TC ) ⊆ t∨
C , cf. [BG10, Definition 2.3.1.], [Tho, Defi-
nition 93].
(gl2,C , O2 (R))-modules generated by modular forms:
• Let f ∈ Mk . We will analyze the (gl2,C , O2 (R))-module V ⊆ A(GL2 )
generated by ϕ := ϕf .
• Recall
Y ∗ ϕ = 0, z · ϕ = z −k ϕ for z ∈ C× .
• It is not difficult to see that there exists a unique, up to isomorphism,
0
(gl2,C , O2 (R))-module Dk−1 (the dual of Dk−1 in [Del73, Section 2.1]) gen-
erated by such an element ϕ, and that Dk−1 0 ∼
= V is irreducible.
• In particular, it only depends on k, not f .
• We obtain, as was mentioned some time ago,
∼ Hom(gl ,O (R)) (D0 , A(GL2 )),
Mk = 2,C 2 k−1

(note: the space H(GA ) in [Del73, Scholie 2.1.3.] agrees with our A(GL2 )
only up to inversion on GL2 (A)).
92 JOHANNES ANSCHÜTZ

• We calulated
1 2
∆∗ϕ= (k − 2k)ϕ, Z ∗ ϕ = −kϕ,
4
and this determines the infinitesimal character ωDk−1
0 .
• The W -orbit of C-algebra homomorphisms U (tC ) ∼ = C[Z, H] → C corre-
sponding to ωDk−1
0 under the Harish-Chandra isomorphism is given by the
homomorphisms
Z 7→ −k, H 7→ ±(k − 1).
• The lattice X (T ) ⊆ hZ, Hi∨
∗
C is given by C-linear maps

Z 7→ a, H 7→ b
with a, b ∈ Z and a ≡ b mod 2.
0
• Thus, Dk−1 is not L-algebraic (it is C-algebraic in the sense of [BG10]).
• However,
0
Dk−1 ⊗C |det|1/2+a
is L-algebraic for any a ∈ Z because (the (gl2,C , O2 (R))-module associated
with) |det|1/2 has infinitesimal character corresponding to
Z 7→ 1, H 7→ 0.
Upshot:
• For any G there exists the space A(G) of (adélic) automorphic forms, which
for GL2 naturally contains the ϕf with f ∈ Mk .
• The space A(G) carries a (gC , K∞ ) × G(Af )-action, and is a more algebraic
replacement for L2 ([G]) with its G(A)-action.
A new notion of automorphic representations:
• We redefine the notion of an automorphic representation using A(G) instead
of L2 ([G]).
• Namely, an automorphic representation for G is an irreducible G(Af ) ×
(gC , K∞ )-subquotient of A(G), cf. [GH19, Definition 6.8.].
• The former automorphic representations will from now on be called “L2 -
automorphic representations”.
• For informations how both notions relate, cf. [GH19, Section 6.5.].
Flath’s theorem (cf. [GH19, Theorem 5.7.1.]):
• For almost all primes p the reductive group GQp := G ×Spec(Q) Spec(Qp ) is
“unramified”, i.e., extends to a reductive group scheme
Gp → Spec(Zp ).
• Equivalently, GQp is quasi-split (=contains a Borel subgroup defined over
Qp ) and split(=contains a maximal and split torus) over an unramified
extension.
• Then
Y0
G(A) = (G(Qp ), Gp (Zp )) × G(R),
p

where Gp (Zp ) ⊆ G(Qp ) is a compact-open subgroup (cf. [GH19, Proposition

2.3.1.]), called “hyperspecial”.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM93

• Flath’s theorem (cf. [GH19, Section 5.7.] and [Fla79]) states that irre-
ducibe, admissible G(Af ) × (gC , K∞ )-modules π decompose (uniquely) into
a “restricted tensor product”
0
π∼
O
= πp ⊗ π∞
p prime

of irreducible, smooth (even admissible) GQp -representations πp and an

irreducible, admissible (gC , K∞ )-module π∞ .
• Moreover, for almost all primes p the representation πp is unramified, i.e.,
G (Z )
the group GQp is unramified and πp p p 6= 0, where Gp is a reductive model
of GQp .
• Important, cf. [GH19, Theorem 7.5.1.]: H(G(Qp ), Gp (Zp )) is again commu-
G (Z )
tative! (⇒ If πp p p 6= 0, then its dimension is one.)
Upshot:
• Thus irreducible, admissible G(Af ) × (gC , K∞ )-modules are a collection of
“local data”.
• Being automorphic puts strong relations among these local data.
• For almost all primes p get homomorphism
χp : H(G(Qp ), Gp (Zp )) → C,
similarly to case for GL2 ⇒ this will yield analog of a “system of Hecke
eigenvalues”.
• For π set
O0
πf := πp ,
p
thus
π∼
= πf ⊗ π∞ .
Definition 11.3 (L-algebraic automorphic representation). An automorphic rep-
resentation
π∼= πf ⊗ π∞
is L-algebraic if the irreducible, admissible (gC , K∞ )-module π∞ is L-algebraic, cf.
[BG10, Definition 3.1.1.].
Examples:
• An automorphic representation π ⊆ A(GL2 ) which is generated by a mod-
ular form is not L-algebraic, but the twist
1/2+a
π ⊗C |det|adélic
is for any a ∈ Z.
• A continuous character ψ : Q× \A× → C is L-algebraic if and only if
ψ = χ| − |kadélic
for some k ∈ Z, | − |adélic the adélic norm and χ : Q× \A× → C a character
of finite order.
94 JOHANNES ANSCHÜTZ

12. The Langlands program for general groups, part II

Last time:
• F arbitrary number field
• The Langlands(-Clozel-Fontaine-Mazur) reciprocity conjecture for GLn,F
says:
– ` some prime. Fix an isomorphism C ∼ = Q` .
– Then for any n ≥ 1, there exists a (unique) bijection between
i) the set of L-algebraic cuspidal automorphic representations of
GLn (AF ),
ii) the set of (isomorphism classes) of irreducible continuous repre-
sentations Gal(F /F ) → GLn (Q` ) which are almost everywhere
unramified, and de Rham at places dividing `,
such that the bijection matches Satake parameters with eigenvalues of
Frobenius elements.
• Introduced space of (adélic) automorphic forms A(G) for any G. N
0
• Flath’s theorem ⇒ automorphic representations π factorize: π ∼ = p πp ⊗
π∞ .
• Explained “L-algebraic” (=infinitesimal character of π∞ is “integral”).
Today:
• Introduce “cuspidal” automorphic representations.
• Explain Satake parameters.
• Discuss some expectations of the Langlands program for general G.
Definition 12.1 ([GH19, Definition 9.2]). Let G/Q be reductive. We call ϕ ∈
L2 ([G]) resp. ϕ ∈ A(G) cuspidal if
Z
ϕ(ng)dn = 0
N (Q)\N (A)

for all unipotent radicals N of proper parabolic subgroups P ⊆ G (defined over Q),
and almost all g ∈ G(A).
Cuspidal automorphic representations:
• Set L2cusp ([G]) resp. Acusp (G) resp. Acusp ([G]) as the subspaces of cuspidal
elements.
• There is an embedding with dense image
Acusp ([G]) ⊆ L2cusp ([G]),
i.e., cuspidal automorphic forms satisfy a growth condition strong enough
to make them L2 (they are “rapidly decreasing”), cf. [GH19, Section 6.5.].
• An L2 -automorphic representation of G(A) is cuspidal if it is isomorphic to
a subquotient (actually subrepresentation) of L2cusp ([G]).
• Similarly, an automorphic representation is cuspidal if it occurs as a sub-
quotient of Acusp (G).
• Gelfand, Piatetski-Shapiro: As a unitary G(A)-representation
\
L2cusp ([G]) ∼
M
= mπ π
[
π∈G(A)

with each mπ finite, i.e., mπ ∈ N ∪ {0}, cf. [GH19, Corollary 9.1.2].

 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM95

• In particular, L2cusp ([G]) ⊂ L2disc ([G]) (note: the trivial representation oc-
curs in L2disc ([G]) \ L2cusp ([G])).
• From
\
L2cusp ([G]) ∼
M
= mπ π
[
π∈G(A)

on deduces the decomposition (into irreducible, admissible G(Af )×(gC , K∞ )-

modules)
Acusp ([G]) ∼
M
= mπ π K∞ −finite ,
[
π∈G(A)

where π K∞ −finite ⊆ π denotes the (dense) subspace of K∞ -finite (in partic-

ular, smooth) vectors, cf. [GH19, Theorem 4.4.4.].
• Thus, the decomposition of L2cusp ([G]) is “the same as” the decomposition
of Acusp ([G]).
• Piatetski-Shapiro, Shalika, cf. [GH19, Theorem 11.3.4]: F number field,
G = ResF/Q GLn,F . Then the G(A)-representation on L2cusp ([G]) is multi-
plicity free (“multiplicity one”).
• “Strong multiplicity one”: Assume π, π 0 are cuspidal automorphic repre-
sentations for GLn,F . If πp ∼= πp0 for almost all primes p, then π ∼
= π 0 , cf.
[GH19, 11.7.2].
• Not true without cuspidality, cf. [BG10, page 38].
• Moeglin, Waldspurger, cf. [GH19, Theorem 10.7.1.], [MW89]: G = ResF/Q GLn,F .
Then L2disc ([G]) is known, up to parametrizing cuspidal automorphic rep-
resentations of GLd,F for d|n, and turns out to be multiplicity free.
We can now close a gap. Namely, we never associated an automorphic represen-
tation to a newform. But with the material presented now we can give (at least
two) formal definition. Let f ∈ Sk be a newform. Then we can either associate to f
the (irreducible) GL2 (Af ) × (gl2,C , O2 (R))-submodule in A(GL2 ) generated by the
automorphic form ϕf (cf. Section 3, or the irreducible, unitary subrepresentation
of L2 ([GL2 ]) generated by ϕ˜f , cf. Section 3. Which construction is used has to be
checked in the respective situation.
Satake parameters:
• Describe πp if π is unramified at p, i.e.,

πpGp (Zp ) 6= 0

with Gp → Spec(Zp ) a reductive model of GQp (which exists for almost all
primes p).
• Thus, describe C-algebra homomorphisms

H(G(Qp ), K) → C

for K := Gp (Zp ).
• Langlands/Satake, cf. [GH19, Theorem 7.5.1., Corollary 7.5.2.], [BG10, Sec-
tion 2.1.]: There exists
– a “natural” algebraic group Ĝ over Q,
– an automorphism Frp : Ĝ(Q) ∼= Ĝ(Q),
96 JOHANNES ANSCHÜTZ

– and a bijection
HomC−alg (H(G(Qp ), K), C)
1:1
↔
{Frobenius semisimple Ĝ(C) − conjugacy classes in Ĝ(C) o FrZp }.
• The group ĜC has the dual root datum as G. Thus for example, cf. [Bor79]:
– if G = GLn , then Ĝ = GLn,C ,
– if G = SLn , then Ĝ = PGLn,C ,
– if G = SO2n , then Ĝ = SO2n,C ,
– if G = SO2n+1 , then Ĝ = Sp2n,C ,
– if G = GSp2n , then Ĝ = GSpin2n+1,C .
• If GQp is split, then the automorphism Frp is trivial, and Frobenius semisim-
ple conjugacy classes are just conjugacy classes of semisimple elements in
Ĝ(C).
• If G = GLn , then semisimple conjugacy classes are uniquely determined by
their characteristic
N0 polynomials.
• Let π = π
p p ⊗ π ∞ be an automorphic representation of G. Then we
obtain the following analog of a system of Hecke eigenvalues:
– a finite set S of primes, such that πp is unramified for p ∈/ S,
– for each p ∈ / S a Frobenius semisimple conjugacy class cp (π) in Ĝ(C) o
FrZp .
• For GLn the eigenvalues of (each element in) cp (π) are the Satake param-
eters of π.
• Concretely, if the coset
GLn (Zp )Diag(p, . . . , p, 1, . . . , 1 )GLn (Zp ),
| {z } | {z }
i−times (n−i)−times
GL (Z )
has eigenvalue e
ap,i on πp n p , then the elements in cp (π) have character-
istic polynomial
(1−n) i(i−n)
Xn − p 2 ap,1 X n−1 + . . . + (−1)i p
e 2 ap,i X i + . . . + (−1)n e
e ap,n ,
cf. [GH19, Section 7.2.].
∞
an q n ∈ Sk (Γ0 (N ), χ)
P
• Even more concrete, for π generated by a newform f =
i=1
the cp (π), p - N, have characteristic polynomial
X 2 − p−1/2 pap X + χ(p)pk .
1/2 1/2
Recall that π ⊗C |det|adélic is L-algebraic. For π ⊗C |det|adélic we obtain the
polynomial
X 2 − ap X + χ(p)pk−1 ,
which (after choosing an isomorphism Q` ∼ = C) was the characteristic poly-
nomial of an arithmetic Frobenius at p.
The L-group:
• G reductive over Q
• ` a prime.
• With a little work (cf. [GH19, Section 7.3.]) the group Gal(Q/Q) acts on
the reductive group Ĝ over Q (the action is trivial if G is split).
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM97

• Set
L
G := Ĝ(Q` ) o Gal(Q/Q).
• An L-parameter is by definition a Ĝ(Q` )-conjugacy class of continuous ho-
momorphisms
Gal(Q/Q) → L G,
whose projection to Gal(Q/Q) is the identity.
• If F/Q is finite and G = ResF/Q GLn,F , then an L-parameter identifies with
an isomorphism class of an n-dimensional `-adic Galois representation of
Gal(F /F ).
The Buzzard–Gee conjecture for L-algebraic automorphic representa-
tions:
• Let ` be a prime.
• Fix an isomorphism ι : C ∼
= Q` .
• Let π be an L-algebraic automorphic representation of G.
• Then Buzzard–Gee (cf. [BG10, Conjecture 3.2.2.]) conjecture that there
exists an L-parameter
L
ρπ : Gal(Q/Q) → G
such that (in particular)
– if p 6= ` is unramified for π, then each arithmetic Frobenius Frobp ∈
Gal(Qp /Qp ) maps to the conjugacy class of

ι(cp (π)) ∈ Ĝ(Q` ) o FrobZp

under
Gal(Qp /Qp ) → Ĝ(Q` ) o Gal(Qp /Qp )  Ĝ(Q` ) o FrobZp ,
b

– for each continuous representation L G → GLN (Q` ), which is algebraic

on Ĝ(Q` ), the composition
ρ L
Gal(Q` /Q` ) −
→ G → GLN (Q` )
is de Rham.
• This generalizes the association of Galois representations to (a twist of) the
automorphic representation associated to some newform.
• Contrary to the case of GLn the L-parameter ρπ is not conjectured to be
unique, cf. [BG10, Remark 3.2.4.].
• Moreover, we may chose the same L-parameter ρ for different L-algebraic
automorphic representations π.
• For GLn the Buzzard–Gee conjecture predicts in addition to the previous
Langlands-Clozel-Fontaine-Mazur conjecture the existence of Galois rep-
resentations attached to possibly non-cuspidal (L-algebraic) automorphic
representations. These Galois representations are then no longer conjec-
tured to be irreducible.
The Langlands group L:
• Conjecturally, there should exist a (very big) locally compact group L,
the (global) Langlands group, with (at least) the following properties, cf.
[GH19, Section 12.6.], [Art02], [LR87]:
98 JOHANNES ANSCHÜTZ

– There exists a canonical surjection

L  Gal(Q/Q)
(in fact, L  WQ with WQ the global Weil group).
– For each n ≥ 0 the isomorphism classes Irrn (L) of continuous irre-
ducible n-dimensional C-representations of L are naturally in bijection
with the set Cspn of cuspidal automorphic representations of GLn (A).
– Fix an isomorphism Q` ∼ = C. Then there is an injection of the set
Ggeom of isomorphism classes of irreducible, almost everywhere un-
ramified `-adic Galois representations Gal(Q/Q) which are de Rham
at ` to the set Irr of isomorphism classes of irreducible representations
of L, with image corresponding to L-algebraic cuspidal automorphic
representations.
– For each place v of Q there is an injection
Lv → L
of the local Langlands groups, which is well-defined up to conjugation
in L. Here:
∗ Lv ∼= WR , the non-split extension of C× by Gal(C/R), if v = ∞.
∗ Lv ∼= WQp × SU2 (R) if v = p prime and
= IQp o Z ⊆ Gal(Qp /Qp ) ∼
WQp ∼ = IQp o Ẑ
is the subgroup of elements mapping to integral powers of Frobe-
nius.
N0
– If π ∼ = v πv ∈ Cspn corresponds to a continuous representation
ρπ : L → GLn (C), then πv should correspond to ρπ |Lv under the local
Langlands correspondence for GLn , cf. [GH19, Section 12.5.].
• If this conditions hold, then necessarily Lab ∼
= Q× \A× . Note that by global
class field theory there exists a surjection Q× \A× → Gal(Q/Q)ab .
• The existence of L is currently completely out of reach, even more conjec-
turally L should surject onto the “motivic Galois group” Gmot over C, cf.
[Art02], [LR87].
The global Langlands correspondence for general G (very, very rough):
• No precision is claimed, and all statements are close to being empty!
• For precise statements one should consult [GH19, Chapter 12] or [Art94].
• In order to parametrize non L-algebraic automorphic representations (at
least those relevant for the decomposition of L2disc ([G])) one should replace
the previous Galois version of L-parameters by (certain) L-parameters
L
L→ G
(actually in the Weil form WQ n Ĝ(C) of the L-group).
• Then one hopes to construct a surjective map (cf. [GH19, Conjecture 12.6.2.])
LL
{(certain) automorphic representations} −−→ {(certain) L − parameters}.
• The fibers of LL are called L-packets. They are possibly infinite.
• A precise construction of a local Langlands correspondence (cf. [GH19,
Conjecture 12.5.1.]) should yield a parametrization of the L-packets, cf.
[GH19, (12.23)].
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM99

• For each π ⊆ L2disc ([G]) its multiplicity should be computable via the L-
parameter LL(π) and the precise parametrizations of the L-packets, cf.
[GH19, Conjecture 12.6.3.].
• However, for non-tempered representations the above should not be rea-
sonable and one should consider Arthur parameters
L
L × SL2 (C) → G
instead of L-parameters, cf. [Art94].
Important stuff, which was not mentioned in the lecture:
• L-functions, [Bor79], [GH19, Chapter 11, Section 12.7.],
• Functoriality, [GH19, Section 12.6.],
• The trace formula, [Art05], [GH19, Section 18],
• The Langlands program for function fields,
• ...
A glimpse on Shimura varieties:
• Let K∞ ⊆ G(R) be a maximal (connected) compact subgroup.
• Let K ⊆ G(Af ) be a (sufficiently small) compact-open subgroup.
• Recall that
–
[G] → [G]/K = G(Q)AG \G(A)/K
is a profinite covering of a real manifold
–
[G]/K → XK := [G]/KK∞ = G(Q)AG \G(A)/KK∞
is a K∞ -bundle over XK , which is a disjoint union of arithmetic man-
ifolds (=quotients of a symmetric spaces by arithmetic subgroups)
• Note that implicitly XK depends on K∞ .
• Let us set (just to simplify some notations later)
X
b := lim XK .
←−
K

Stuff we encountered for G = GL2 :

(1) The upper/lower halfplane H± ∼
= G(R)/K∞ , which is naturally a complex
manifold
(2) The holomorphic embedding
H± → P1C
(3) The G(R)-equivariant vector bundles
ωk
on H± , or by pullback on XK for K ⊆ G(Af ) compact-open.
(4) For k ∈ Z the space of modular forms
Mk ⊆ H 0 (X,
b ω k ),
where the RHS denotes holomorphic sections.
∗
(5) The canonical compactification XK of XK , with the extension of ω k on it.
(6) A scheme Xb → Spec(Q), whose C-valued points are naturally isomorphic
to X.
b
100 JOHANNES ANSCHÜTZ

(7) The Gal(Q/Q) × G(Af )-representation

1 e
H
e ét (XQ , L)

associated to certain G(Af )-equivariant Q` -local systems L on Xe, e.g., L ∼

=
Q` .
(8) The Eichler–Shimura isomorphism relating Sk for k ≥ 2, to certain
1 e
H
e ét (XQ , L).

(9) The Eichler–Shimura relation expressing a relation of the Gal(Q/Q)-action

and the G(Af )-action on
e 1 (Xe , L),
Hét Q

which was a consequence of the mod p geometry of some modular curve.

(10) A map
LL : Amod → Gmod
with (conjecturally) describable image.
For general G there are unfortunately no analogs - but wait: there exist
G giving rise to Shimura varieties!
• Already (Item 1) fails for general G: The space
G(R)/K∞
need not be complex manifold! (E.g., if G = GL3 , then the real dimension
dim(GL3 (R)/AGL3 (R) SO3 (R)) = 3(3+1)
2 − 1 = 5 is odd)
• In particular, everything related to holomorphicity has no analog for such
G, e.g., Item 2, Item 3, Item 4, Item 6, Item 7,...
• That is bad news!
• Good news: For groups G giving rise to Shimura data (like ResF/Q GL2,F
for F/Q totally real, or GSp2n ,... , cf. [Lan17], [Mil05], [Del71a], [Del79]),
the real manifold
G(R)/K∞ AG
is a complex manifold.
• A Shimura datum is a pair (G, X) with G a reductive group G over Q, and
X a G(R)-conjugacy classes of morphisms h : S := ResC/R Gm → GR such
that
1) For each h ∈ X, the characters of S(R) = C× acting on Lie(GR )C (via
Ad ◦ h) are either z 7→ zz , z 7→ 1, or z 7→ zz .
2) For each h ∈ X, the group {g ∈ G(C) | h(i)gh(i)−1 = g} is compact
modulo its center.
3) The adjoint group Gad has no factor, defined over Q, for which the
projection of h ∈ X is trivial.
• E.g., take G = GL2 , and X as the conjugacy class of the usual embedding
C× → GL2 (R). Note that X ∼ = H± .
• In general, for h ∈ X the stabilizer Kh of h in G(R) is compact modulo
the center of G(R) (this follows from condition 2)) and X ∼= G(R)/Kh (∼=
G(R)/AG K∞ ).
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM
101

• For K ⊆ G(Af ) recall

XK := G(Q)\(G(Af )/K × X).
Other notation: ShK (G, X) the Shimura variety of level K attached to
(G, X).
• We get:
(1) X is a disjoint union of hermitian symmetric domains, cf. [Mil05],
[Del79].
(2) To each h ∈ X is naturally attached a parabolic subgroup Ph ⊆ G(C)
(cf. [CS17, Section 2.1]), and sending h → Ph defines an open embed-
ding
π: X ∼ = G(R)/Kh → Fl ∼ = G(C)/Ph
of X into a flag variety (this generalizes the embedding H± → P1 (C)).16
In particular, X carries a natural complex structure. More is true
(Baily-Borel): XK is naturally a quasi-projective variety, cf. [Lan17,
Theorem 2.4.1.].
(3) The pullback of G(C)-equivariant vector bundles on Fl gives rise to
G(Q)-equivariant vector bundles on G(Af ) × X. By descent one one
obtains automorphic vector bundles E on XK for any K ⊆ G(Af )
compact-open, i.e., analogs of the ω k , k ∈ Z, [Har88]. Note that G(C)-
equivariant vector bundles on Fl are equivalent to (algebraic) repre-
sentations of Ph .
(4), (5) If dimC (XK ) > 1, there do exist compactifications XK , but these are
no longer canonical. However, it is possible to define subspaces in the
cohomology of H ∗ (XK , E)), which generalize Mk for k ∈ Z, cf. [Har88].
(6) The space X b (=inverse limit of complex manifolds) arises again as the
C-points of a scheme X b → Spec(E), the canonical model, defined over
some number field (the “reflex field of the Shimura data”), cf. [Lan17,
Theorem 2.4.3.].
(8),(9),(10) Having the canonical model, one can attempt (sometimes sucessfully)
to obtain analogs of points (18),(19),(20) and finally relate (certain)
automorphic representations to (variants of) Galois representations,
cf. [Har88], [BRb]. Needless to say that everything (compactifications,
Eichler–Shimura isomorphism/relation, integral models, mod p geome-
try, interior vs intersection cohomology,...) is much more complicated!

16Viewing X as a “moduli space of (polarized) Hodge structures” clarifies this embedding, cf.
[Del79].
102 JOHANNES ANSCHÜTZ

References
[Art94] James Arthur. Unipotent automorphic representations: conjectures. Department of
Mathematics, University of Toronto, 1994.
[Art02] James Arthur. A note on the automorphic langlands group. Canadian Mathematical
Bulletin, 45(4):466–482, 2002.
[Art05] James Arthur. An introduction to the trace formula. Harmonic analysis, the trace for-
mula, and Shimura varieties, 4:1–263, 2005.
[BC09] Olivier Brinon and Brian Conrad. Cmi summer school notes on p-adic hodge theory.
2009.
[BG10] Kevin Buzzard and Toby Gee. The conjectural connections between automorphic rep-
resentations and galois representations. Automorphic forms and Galois representations,
1:135–187, 2010.
[Bor79] Armand Borel. Automorphic l-functions. In Automorphic forms, representations and
L-functions (Proc. Sympos. Pure Math., Oregon State Univ., Corvallis, Ore., 1977),
Part, volume 2, pages 27–61, 1979.
[Bor80] A. Borel. Stable and L2 -cohomology of arithmetic groups. Bull. Amer. Math. Soc.
(N.S.), 3(3):1025–1027, 1980.
[BRa] JN Bernstein and K Rumelhart. Lectures on p-adic groups. Unpublished, available at
http://www. math. tau. ac. il/˜ bernstei/Publication list.
[BRb] Don Blasius and J Rogawski. Zeta functions of shimura varieties. motives (seattle, wa,
1991), 525–571. In Proc. Sympos. Pure Math, volume 55.
[BT13] Raoul Bott and Loring W Tu. Differential forms in algebraic topology, volume 82.
Springer Science & Business Media, 2013.
[Cas73] William Casselman. On some results of atkin and lehner. Mathematische Annalen,
201(4):301–314, 1973.
[CGP15] Brian Conrad, Ofer Gabber, and Gopal Prasad. Pseudo-reductive groups, volume 26 of
New Mathematical Monographs. Cambridge University Press, Cambridge, second edi-
tion, 2015.
[Col11] P Colmez. éléments danalyse et dalgebre (et de théorie des nombres), second éd. Éditions
de lÉcole Polytechnique, Palaiseau, 2011.
[CS17] Ana Caraiani and Peter Scholze. On the generic part of the cohomology of compact
unitary shimura varieties. Annals of Mathematics, pages 649–766, 2017.
[DE14] Anton Deitmar and Siegfried Echterhoff. Principles of harmonic analysis. Springer,
2014.
[Del71a] P Deligne. Travaux de shimura, sem. bourbaki 389. Lecture Notes in Math, 244, 1971.
[Del71b] Pierre Deligne. Formes modulaires et representations e-adiques. In Séminaire Bourbaki
vol. 1968/69 Exposés 347-363, pages 139–172. Springer, 1971.
[Del73] Pierre Deligne. Formes modulaires et représentations de gl (2). In Modular functions of
one variable II, pages 55–105. Springer, 1973.
[Del75] Pierre Deligne. Courbes elliptiques: formulaire dapres j. tate. In Modular functions of
one variable, IV (Proc. Internat. Summer School, Univ. Antwerp, Antwerp, 1972),
volume 476, pages 53–73, 1975.
[Del79] Pierre Deligne. Variétés de shimura: interprétation modulaire, et techniques de construc-
tion de modeles canoniques. In Automorphic forms, representations and L-functions
(Proc. Sympos. Pure Math., Oregon State Univ., Corvallis, Ore., 1977), Part, volume 2,
pages 247–289, 1979.
[DR73] Pierre Deligne and Michael Rapoport. Les schémas de modules de courbes elliptiques.
In Modular functions of one variable II, pages 143–316. Springer, 1973.
[DS74] Pierre Deligne and Jean-Pierre Serre. Formes modulaires de poids 1. In Annales scien-
tifiques de l’École Normale Supérieure, volume 7, pages 507–530, 1974.
[DS05] Fred Diamond and Jerry Michael Shurman. A first course in modular forms, volume
228. Springer, 2005.
[Fla79] Dan Flath. Decomposition of representations into tensor products. In Automorphic
forms, representations and L-functions (Proc. Sympos. Pure Math., Oregon State Univ.,
Corvallis, Ore., 1977), Part, volume 1, pages 179–183, 1979.
 LECTURE NOTES FOR AN INTRODUCTORY COURSE ON THE LANGLANDS PROGRAM
103

[FM95] Jean-Marc Fontaine and Barry Mazur. Geometric galois representations. Elliptic curves,
modular forms, & Fermats last theorem (Hong Kong, 1993), Ser. Number Theory, I,
pages 41–78, 1995.
[Fon94] Jean-Marc Fontaine. Le corps des périodes p-adiques. Astérisque, 223:59–111, 1994.
[Gel75] Stephen S Gelbart. Automorphic forms on adele groups. Number 83. Princeton Univer-
sity Press, 1975.
[GH19] Jayce R Getz and Heekyoung Hahn. An introduction to automorphic representations.
Available at website: https: // www. math. duke. edu/ ~ hahn/ GTM. pdf , 2019.
[GS17] Philippe Gille and Tamás Szamuely. Central simple algebras and Galois cohomology,
volume 165. Cambridge University Press, 2017.
[Har88] Michael Harris. Automorphic forms and the cohomology of vector bundles on shimura
varieties. In Proceedings of the Conference on Automorphic Forms, Shimura Varieties,
and L-functions, Ann Arbor, pages 41–91, 1988.
[Lan17] Kai-Wen Lan. An example-based introduction to shimura varieties. Preprint, 2017.
[LR87] R. P. Langlands and M. Rapoport. Shimuravarietäten und Gerben. J. Reine Angew.
Math., 378:113–220, 1987.
[Lur] Jacob Lurie. Tamagawa numbers via nonabelian poincaré duality. Lectures available at
http://www.math.harvard.edu/~lurie/282y.html.
[Mil05] James S Milne. Introduction to shimura varieties. Harmonic analysis, the trace formula,
and Shimura varieties, 4:265–378, 2005.
[MW89] Colette Mœglin and J-L Waldspurger. Le spectre résiduel de {gl}(n). In Annales scien-
tifiques de l’École normale supérieure, volume 22, pages 605–674, 1989.
[Pan19] Lue Pan. The fontaine-mazur conjecture in the residually reducible case. arXiv preprint
arXiv:1901.07166, 2019.
[PS16] Vincent Pilloni and Benoı̂t Stroh. Arithmétique p-adique des formes de hilbert surcon-
vergence, ramification et modularité. Astérisque, 382:195–266, 2016.
[Rib77] Kenneth A Ribet. Galois representations attached to eigenforms with nebentypus. In
Modular functions of one variable V, pages 18–52. Springer, 1977.
[Rib90] Kenneth A. Ribet. From the taniyama-shimura conjecture to fermat’s last theorem.
Annales de la Faculté des sciences de Toulouse : Mathématiques, Ser. 5, 11(1):116–139,
1990.
[Ser97] Jean-Pierre Serre. Abelian l-adic representations and elliptic curves. CRC Press, 1997.
[Tho] Jack Thorne. Topics in automorphic forms (notes taken by chao li. course
notes available at http://www.math.columbia.edu/~chaoli/docs/AutomorphicForm.
html#thm:AFHCiso.
[Thu82] William P Thurston. Three dimensional manifolds, kleinian groups and hyperbolic ge-
ometry. Bulletin of the American Mathematical Society, 6(3):357–381, 1982.
[Wil95] Andrew Wiles. Modular elliptic curves and fermat’s last theorem. Annals of mathemat-
ics, 141(3):443–551, 1995.

Mathematisches Institut, Universität Bonn, Endenicher Allee 60, 53115 Bonn, Deutsch-
land
E-mail address: ja@math.uni-bonn.de