Representation Theory of Finite

Groups
arXiv:1001.0462v2 [math.RT] 21 Dec 2022

Dr. Anupam Singh

Professor (Mathematics, IISER Pune)

Indian Institute of Science Education and Research (IISER) Pune,

Dr Homi Bhabha Road, Pashan, Pune 411008 INDIA
email : [email protected]

Revised November 2022

First Draft December 2009
 Contents

Chapter 1. Introduction 7
1.1. Written in December 2022 7
1.2. Written in December 2009 7

Chapter 2. Representation of a Group 11

2.1. Commutator Subgroup and One Dimensional Representations 14

Chapter 3. Maschke’s Theorem 17

Chapter 4. Schur’s Lemma 23

Chapter 5. Representation Theory of Finite Abelian Groups over C 25

5.1. Example of Representation over Q 27

Chapter 6. The Group Algebra k[G] 29

Chapter 7. Constructing New Representations 31

7.1. Subrepresentation of a Representation 31
7.2. Sum of Representations 31
7.3. Adjoint Representation 31
7.4. Restriction of a Representation 33
7.5. Lift of a Representation 33
7.6. Tensor Product of two Representations 33
7.7. Decomposition of the Representation V ⊗ V 35
7.8. Induced Representation 37

39

Chapter 8. Matrix Elements 41

Chapter 9. Character Theory 43

3
4 CONTENTS

Chapter 10. Orthogonality Relations 45

Chapter 11. Main Theorem of Character Theory 49

11.1. Regular Representation 49
11.2. The Number of Irreducible Representations 50
11.3. Artin-Wedderburn Decomposition 51
11.4. Exercises on Character Table 52

Chapter 12. Examples 55

12.1. Groups having Large Abelian Subgroups 56
12.2. Character Table of Some Groups 57
12.3. Characters of Direct Product 61

Chapter 13. Character Table of S5 63

Chapter 14. Restriction of a Representation 65

14.1. Character Table of A5 66

Chapter 15. Central Characters 69

15.1. Algebraic Integers 71

Chapter 16. Burnside’s pq Theorem 73

16.1. Solvable Groups 75

77

Chapter 17. Induced Representation 79

17.1. Induced Representations from Subgroups 79
17.2. Construction of Induced Representation 81

Chapter 18. Character of the Induced Representation 83

18.1. Induced Representation via Tensor Product 84
18.2. Induced Representations via Invariant-functions 85

Chapter 19. Fröbenius Reciprocity 87

Chapter 20. Mackey’s Irreducibility Criteria 91

20.1. Restriction of an Induced Representation 91
20.2. Mackey-Wigner Irreducibility Criteria 93

Chapter 21. Representations of a Semidirect Product - Wigner’s Little Group Method 97

21.1. Representations of groups of order pq and p3 98
 CONTENTS 5

Chapter 22. Characters of the group GL2 (q) 101

22.1. Conjugacy classes in GL2 (q) 102
22.2. Parabolic Induction - preparation 103
22.3. The principal series representations 103
22.4. The Steinberg Representations 106
22.5. Cuspidal representations 108
22.6. Characters of GL2 (q) 109

Bibliography 111
 CHAPTER 1

Introduction

1.1. Written in December 2022

Since the first version of these notes, the author has received several feedback from
students as well as teachers (two places, IIT Kanpur and IMSc Chennai, are worth ac-
knowledging). Over the years the author got opportunities to teach a part of the material
in various workshops and a full course at IISER Pune. The author expresses his gratitude
to all students who were part of these exciting courses. The present version is revised with
some more material added. The most important part on Induced Representation which was
missing earlier is added now along with several more examples. Please let me know your
feedback as and when you refer to these notes.

1.2. Written in December 2009

This is a class note for the course on the Representation Theory of Finite Groups taught
by the author at IISER Pune to undergraduate students. We study the character theory of
finite groups and illustrate how to get more information about groups. Burnside’s theorem
is one of the very good applications. It states that every group of order pa q b , where p, q are
distinct primes, is solvable. We will always consider finite groups unless stated otherwise.
All vector spaces will be considered over general fields in the beginning but for the purpose
of character theory, we assume the field is that of complex numbers.
We assume knowledge of the basic group theory and linear algebra. The point of view
I projected to the students in the class is that we have studied linear algebra hence we
are familiar with the groups GL(V ), the general linear group, or GLn (k) in the matrix
notation. The idea of representation theory is to compare (via homomorphisms) finite
(abstract) groups with these linear groups (somewhat concrete) and hope to a gain better
understanding of them.
The students were asked to read about “linear groups” from the book by Alperin and
Bell (mentioned in the bibliography) from the chapter with the same title. We also revised,
side-by-side in the class, Sylow’s Theorem, Solvable groups and motivated ourselves for
Burnside’s pq-theorem.
7
8 1. INTRODUCTION

The aim to start with an arbitrary field was to give the feeling that the theory is
dependent on the base field and it gets considerably complicated if we move away from
characteristic 0 algebraically closed field. This we illustrate by giving an example of a
higher dimensional irreducible representation of cyclic group over Q while all its irreducible
representations are one dimensional over C. This puts things in perspective as to why we
are doing the theory over C and motivates us to develop the “Character Theory”.
 PART – I

This part contains basic definitions and some important foundational Theorems.

9
 CHAPTER 2

Representation of a Group

Let G be a finite group. Let k be a field. We will assume that characteristic of k is 0,

e.g., k = C, R or Q. Although often it would be enough to assume char(k) ∤ |G|.

Definition 2.1 (Representation). A representation of G over k is a homomorphism

ρ : G → GL(V ) where V is a vector space of finite dimension over the field k. The vector
space V is called representation space of G, and its dimension is the dimension of
representation.

Strictly speaking, the pair (ρ, V ) is called a representation of G over the field k. However,
if there is no confusion we would simply call either ρ or V , a representation of G.
Let us fix a basis {v1 , v2 , . . . , vn } of V . Then, each ρ(g) can be written in a matrix
form with respect to this basis. This would give a map ρ̃ : G → GLn (k) which is a group
homomorphism. Thus, a representation of G can also be thought of as such group homo-
morphism to GLn (k). This idea is often useful in computing things as is the case in Linear
Algebra.

Definition 2.2 (Invariant Subspace). Let ρ be a representation of G, and W ⊂ V be a

subspace. The space W is said to be a G-invariant (or G-stable) subspace if ρ(g)(w) ∈ W
∀w ∈ W and ∀g ∈ G.

Notice that once we have a G-invariant subspace W we can restrict the representation to
this subspace and define another representation ρW : G → GL(W ) where ρW (g) = ρ(g)|W .
Hence, W is also called a subrepresentation.

Example 2.3 (Trivial Representation). Let G be a group, and k a field. Let V be a

vector space over k. Then ρ(g) = 1 for all g ∈ G is a representation. This is called trivial
representation. In this case, every subspace of V is an invariant subspace.

Example 2.4. Let G = Z/mZ and k = C. Let V be a vector space of dimension n.

2πir
(1) Suppose dim(V ) = 1. Define ρr : Z/mZ → C∗ by 1 7→ e m for 1 ≤ r ≤ m − 1.
(2) Define ρ : Z/mZ → GL(V ) by 1 7→ T where T ∈ GL(V ) satisfying T m = 1. For
2πir1 2πir2
example, in the case dim(V ) = 2 we can take T = diag{e m , e m }. There is
11
12 2. REPRESENTATION OF A GROUP

a general theorem in Linear Algebra which says that any such matrix, over C, is
diagonalizable.

Example 2.5. Let G = Z/mZ and k = R. Let V = R2 with basis {e1 , e2 }. Then, we
have representations of Z/mZ:
" #
cos 2πr
m − sin 2πr
m
ρr : 1 7→
sin 2πr
m cos 2πr
m

where 0 ≤ r ≤ m − 1. Notice that we get m distinct representations.

Example 2.6. Let φ : G → H be a group homomorphism. Let ρ be a representation of

H. Then ρ ◦ φ is a representation of G.

Example 2.7. Let G = Dm = ha, b | am = 1 = b2 , ab = bam−1 i be the dihedral group

with 2m elements. We have representations ρr defined by:
" # " #
cos 2πr
m − sin 2πr
m 0 1
a 7→ 2πr 2πr , b 7→ .
sin m cos m 1 0
Verify that these are group homomorphisms. Notice that the earlier representation of Z/mZ
is a restriction of this.

Example 2.8 (Permutation Representation of Sn ). Let Sn be the symmetric group on

n symbols, and k be a field. Let V = kn with the standard basis {e1 , . . . , en }. We define a
representation ρ of Sn on V by its action on the basis vectors as follows: ρ(σ)(ei ) = eσ(i) for
σ ∈ Sn . Notice that while defining this representation we don’t need to specify any field.
That is if we write the matrix of this representation the entries are from Z.

Example 2.9 (Group Action). Let G be a group and k a field. Suppose G is acting on
a finite set X, i.e., we have a group action G × X → X. We denote by k[X] = {f | f : X →
k}, the set of all maps. Clearly, k[X] is a vector space of dimension |X|. The elements
ex : X → k defined by ex (x) = 1 and ex (y) = 0 if x 6= y, form a basis of k[X]. This action
gives rise to a representation of G on the space k[X] as follows: ρ : G → GL(k[X]) given by
(ρ(g)(f ))(x) = f (g −1 x) for x ∈ X.
If we take G = Sn and X = {1, 2, . . . , n} we get back the Example 2.8.

Example 2.10 (Regular Representation). Let G be a group of order n, and k a field.

Let V = k[G] be the n-dimensional vector space with basis as elements of the group itself.
We define L : G → GL(k[G]) by L(g)(f )(h) = f (g −1 h), called the left regular representation.
Similarly, R(g)(f )(h) = f (hg), defines right regular representation R of G. Prove that these
representations are injective. Further, these representations are obtained by the action of
 2. REPRESENTATION OF A GROUP 13

G on the set X = G by left multiplication or right multiplication. In fact, one can make
k[G] an algebra by defining the following multiplication:
X
(f ∗ f ′ )(t) = f (x)f ′ (x−1 t).
x∈G

Note that this is the convolution multiplication (not the usual point-wise multiplication).
This algebra k[G] is called the group algebra of G over k.
Example 2.11. Let G = Q8 = {±1, ±i, ±j, ±k}, and k = C. We define a 2-dimensional
representation of Q8 by: " # " #
0 1 0 i
i 7→ , j 7→ .
−1 0 i 0
Example 2.12 (Galois Theory). Let K = Q(θ) be a finite extension of Q. Let G =
Gal(K/Q). We take V = K, a finite-dimensional vector space over Q. We have a natural
representation of G as follows: ρ : G → GL(K) defined by ρ(g)(x) = g(x). Take θ = ζ, some
nth root of unity, and show that the cyclic groups Z/mZ have representations over the field
Q of possibly dimensions more than 2. This could be thought of as a reinterpretation of
the statement of the Kronecker-Weber theorem.

Definition 2.13 (Equivalence of Representations). Let (ρ, V ) and (ρ′ , V ′ ) be two rep-
resentations of G over a field k. The representations (ρ, V ) and (ρ′ , V ′ ) are said to be
G-equivalent (or equivalent) if there exists a linear isomorphism T : V → V ′ such that
ρ′ (g) = T ρ(g)T −1 for all g ∈ G. The following diagram helps in understanding this require-
ment:
T / V′
V
ρ(g) ρ′ (g)
 
T / V′
V
Let ρ be a representation. Fix a basis of V , say {e1 , . . . , en }. Then ρ gives rise to a map
G → GLn (k) which is a group homomorphism. Notice that, if we change the basis of
V then we may get a different map for the same ρ. However, these would be equivalent
as representations, i.e. these would differ by conjugation with respect to a fixed matrix
(namely the base change matrix).
We have seen that a representation can have possibly a subrepresentation. This moti-
vates us to define:
Definition 2.14 (Irreducible Representation). A representation (ρ, V ) of G is called
irreducible if it has no proper G-invariant subspace, i.e., the only G-invariant subspaces
are 0 and V .
14 2. REPRESENTATION OF A GROUP

Example 2.15. The trivial representation is irreducible if and only if it is one-dimensional.

Example 2.16. In the case of Permutation representation the subspace W =< (1, 1, . . . , 1) >
P
and W ′ = {(x1 , . . . , xn ) | xi = 0} are two irreducible Sn invariant subspaces. In fact,
this representation is a direct sum of these two, and hence completely reducible.

Example 2.17. One-dimensional representation is always irreducible. If |G| ≥ 2 then

the regular representation is not irreducible.

Exercise 2.18. Let G be a finite group. In the definition of a representation, let us

not assume that the vector space V is finite-dimensional. Prove that there exists a finite-
dimensional G-invariant subspace of V .

Hint: Fix v ∈ V non-zero and take W the subspace generated by ρ(g)(v) ∀g ∈ G.

Exercise 2.19. A representation of dimension 1 is a map ρ : G → k ∗ . There are exactly

two one-dimensional representations of Sn over C. There are exactly n one-dimensional
representations of the cyclic group Z/nZ over C.

Exercise 2.20. Prove that every finite group can be embedded inside symmetric group
Sn , for some n as well as a linear group GLm for some m.

Hint: Make use of the regular representation. This representation is also called “God-
given” representation. Later in the course, we will see why it’s so.

Exercise 2.21. Is the above exercise true if we replace Sn by An and GLm by SLm ?

Exercise 2.22. Prove that the cyclic group Z/pZ has a representation of dimension
p − 1 over Q.

Hint: Make use of the cyclotomic field extension Q(ζp ) and consider the map left mul-
tiplication by ζp . This exercise shows that representation theory is deeply connected to the
Galois Theory of field extensions.

2.1. Commutator Subgroup and One Dimensional Representations

Let G be a finite group. Consider the set of elements {xyx−1 y −1 | x, y ∈ G} and G′ the
subgroup generated by this subset. This subgroup is called the commutator subgroup
of G. We list some of the properties of this subgroup as an exercise here.

Exercise 2.23. (1) G′ is a normal subgroup.

(2) G/G′ is Abelian.
(3) G′ is smallest subgroup of G such that G/G′ is Abelian.
 2.1. COMMUTATOR SUBGROUP AND ONE DIMENSIONAL REPRESENTATIONS 15

(4) G′ = 1 if and only if G is Abelian.

(5) For G = Sn , G′ = An ; G = Dn =< r, s | r n = 1 = s2 , srs = r −1 > we have
G′ =< r >∼ = Z/nZ and Q′8 = Z(Q8 ).

Let Gb be the set of all one-dimensional representations of G over C, i.e., the set of all
group homomorphisms from G to C∗ . For χ1 , χ2 ∈ G b we define multiplication by:

(χ1 χ2 )(g) = χ1 (g)χ2 (g).

b is an Abelian group.
Exercise 2.24. Prove that G
b we have G′ ⊂ ker(χ). Hence we can prove,
We observe that for a χ ∈ G
b∼
Exercise 2.25. Show that G = G/G′ .
b for G = Z/nZ, Sn and Dn .
Exercise 2.26. Calculate directly G

Let G be a group. The group G is called simple if G has no proper normal subgroup.

Exercise 2.27. (1) Let G be an Abelian simple group. Prove that G is isomorphic
to Z/pZ where p is a prime.
(2) Let G be a simple non-Abelian group. Then G = G′ .
 CHAPTER 3

Maschke’s Theorem

From the previous chapter, we recall,

Definition 3.1 (Irreducible Representation). A representation (ρ, V ) of G is called

irreducible if it has no proper G-invariant subspace, i.e., only G-invariant subspaces are
0 and V .

Let (ρ, V ) and (ρ′ , V ′ ) be two representations of G over a field k. We can define direct
sum of these two representations (ρ ⊕ ρ′ , V ⊕ V ′ ) as follows: ρ ⊕ ρ′ : G → GL(V ⊕ V ′ ) given
by (ρ ⊕ ρ′ )(g)(v, v ′ ) = (ρ(g)(v), ρ′ (g)(v ′ )) where v ∈ V and v ′ ∈ V ′ . In the matrix notation,
if we have two representations ρ : G → GLn (k) and ρ′ : G → GLm (k), then ρ ⊕ ρ′ is given
(with respect to an appropriate basis) by
" #
ρ(g) 0
g 7→ .
0 ρ′ (g)
This motivates us to look at those “nice” representations which can be obtained by taking
the direct sum of irreducible ones.

Definition 3.2 (Completely Reducible). A representation (ρ, V ) of G is called com-

pletely reducible if it is a direct sum of irreducible ones. Equivalently, if V = W1 ⊕ . . . ⊕
Wr , where each Wi is G-invariant irreducible representation.

This brings us to the following questions:

(1) Is it true that every representation is a direct sum of irreducible ones?
(2) How many irreducible representations are there for a given G over k?
The answer to the first question is affirmative for finite groups when char(k) ∤ |G|. This
is Maschke’s theorem proved below (it is also true for Compact Groups where it is called
the Peter-Weyl Theorem). The other case when char(k) | |G| comes under the subject of
“Modular Representation Theory”, and the answer to the first question is ‘No’. We will
answer the second question over the field of complex numbers (and possibly over R) in this
course which is the subject of “Character Theory”. For the theory over Q, the subject is
called ‘rationality questions’ (refer to the book by Serre).
17
18 3. MASCHKE’S THEOREM

Theorem 3.3 (Maschke’s Theorem). Let k be a field and G be a finite group. Suppose
char(k) ∤ |G|, i.e. |G| is invertible in the field k. Let (ρ, V ) be a finite-dimensional represen-
tation of G. Let W be a G-invariant subspace of V . Then, there exists W ′ , a G-invariant
subspace, such that V = W ⊕ W ′ .
Conversely, if char(k) | |G| then there exists a representation, namely the regular rep-
resentation, and a proper G-invariant subspace which does not have a G-invariant comple-
ment.

We can re-state the above theorem as follows: Let k be a field and G be a finite group.
Then, every finite-dimensional representation of G over k has the property that whenever
it has a G-invariant subspace it has a G-invariant complement if and only if char(k) ∤ |G|.
This gives us:

Proposition 3.4 (Complete Reducibility). Let k be a field and G a finite group with
char(k) ∤ |G|. Then, every finite-dimensional representation of G is completely reducible.

Now we are going to prove the above results. We need to recall the notion of projection
from ‘Linear Algebra’. Let V be a finite-dimensional vector space over the field k.

Definition 3.5. An endomorphisms π : V → V is called a projection if π 2 = π.

Some examples of projection are given as exercises below.

Let W ⊂ V be a subspace. A subspace W ′ is called a complement of W if V = W ⊕W ′ .
Sometimes, we also say that W has a direct summand. It is a simple exercise in ‘Linear
Algebra’ to show that such a complement always exists (see exercise below), and there could
be many of them.

Lemma 3.6. Let π be an endomorphism. Then, π is a projection if and only if there

exists a decomposition V = W ⊕ W ′ such that π(W ) = 0, π(W ′ ) = W ′ , and π restricted to
W ′ is identity.

Proof. Let π : V → V be such that π(w, w′ ) = w′ . Then, clearly π 2 = π.

Now suppose π is a projection. We claim that V = ker(π) ⊕ Im(π). Let x ∈ ker(π) ∩
Im(π). Then there exists y ∈ V such that π(y) = x and x = π(y) = π 2 (y) = π(π(x)) =
π(0) = 0. Hence ker(π) ∩ Im(π) = 0. Now, let v ∈ V . Then v = (v − π(v)) + π(v), and
we see that π(v) ∈ Im(π) and v − π(v) ∈ ker(π) since π(v − π(v)) = π(v) − π 2 (v) = 0.
Now, let x ∈ Im(π), say x = π(y). Then π(x) = π(π(y)) = π(y) = x. This shows that π
restricted to Im(π) is the identity map. 

Remark 3.7. The reader familiar with ‘Jordan/Rational Canonical Form Theory’
would recognise the following. The minimal polynomial of π is X(X − 1) which is a product
 3. MASCHKE’S THEOREM 19

of distinct linear factors. Hence π is a diagonalizable linear transformation with eigenvalues

0 and 1. Hence, there exists a basis so that the matrix of π is diag{0, . . . , 0, 1, . . . , 1}. This
will give another proof of the above Lemma.

Proof of the Maschke’s Theorem. Let ρ : G → GL(V ) be a representation. Let W

be a G-invariant subspace of V . Let W0 be a (vector space) complement, i.e., V = W0 ⊕ W .
We have to produce a complement which is G-invariant. Let π be a projection corresponding
to this decomposition, i.e., π(W0 ) = 0, and π(w) = w for all w ∈ W . We define an
endomorphism π ′ : V → V by ‘averaging technique’ as follows:
1 X
π′ = ρ(t)−1 πρ(t).
|G|
t∈G

We claim that π ′ is a projection. We note that π ′ (V ) ⊂ W since πρ(t)(V ) ⊂ W and W

1 P
is G-invariant. In fact, π ′ (w) = w for all w ∈ W since π ′ (w) = |G| −1
t∈G ρ(t) πρ(t)(w) =
1 P −1 1 P −1
|G| t∈G ρ(t) π(ρ(t)(w)) = |G| t∈G ρ(t) (ρ(t)(w)) = w (note that ρ(t)(w) ∈ W and π
takes it to itself). Let v ∈ V . Then π ′ (v) ∈ W . Hence π ′2 (v) = π ′ (π ′ (v)) = π ′ (v) as we
have π ′ (v) ∈ W and π ′ takes any element of W to itself. Hence π ′2 = π ′ .
Now we write decomposition of V with respect to π ′ , say V = W ′ ⊕ W where W ′ =
ker(π ′ ) and Im(π ′ ) = W . We claim that W ′ is G-invariant which will prove the theorem.
For this we observe that π ′ is a G-invariant homomorphism, i.e., π ′ (ρ(g)(v)) = ρ(g)(π ′ (v))
for all g ∈ G and v ∈ V .
1 X
π ′ (ρ(g)(v)) = ρ(t)−1 πρ(t)(ρ(g)(v))
|G|
t∈G
1 X
= ρ(g)ρ(g)−1 ρ(t)−1 πρ(t)ρ(g)(v)
|G|
t∈G
1 X
= ρ(g) ρ(tg)−1 πρ(tg)(v)
|G|
t∈G
′
= ρ(g)(π (v)).

This helps us to verify that W ′ is G-invariant. Let w′ ∈ W ′ . To show that ρ(g)(w′ ) ∈ W ′ .

For this we note that π ′ (ρ(g)(w′ )) = ρ(g)(π ′ (w)) = ρ(g)(0) = 0. This way we have produced
G-invariant complement of W .
For the converse
nP let char(k) | |G|.o We take the regular representation V = k[G].
P
Consider W = g∈G αg g | αg = 0 . We claim that W is G-invariant but it has no
G-invariant complement. This requires a bit of effort and we leave it to an interested reader
to either work out herself or look up some textbook. 
20 3. MASCHKE’S THEOREM

Remark 3.8. In the proof of Maschke’s theorem, one can start with a symmetric
bilinear form and apply the trick of averaging to it. In that case, the complement will
be the orthogonal subspace. Conceptually, I like that proof better however it requires
familiarity with the bilinear form to be able to appreciate that proof. Later we will do that
in some other context. An enthusiastic reader can work out the exercise below.

Proof of the Proposition (Complete Reducibility). Let ρ : G → GL(V ) be a rep-

resentation. We use induction on the dimension of V to prove this result. Let dim(V ) = 1.
It is easy to verify that one-dimensional representation is always irreducible. Let V be of
dimension n ≥ 2. If V is irreducible we have nothing to prove. So we may assume V has
a G-invariant proper subspace, say W with 1 ≤ dim(W ) ≤ n − 1. By Maschke’s Theorem,
we can write V = W ⊕ W ′ where W ′ is also G-invariant. But now dim(W ) and dim(W ′ )
both are less than n. By the induction hypothesis, they can be written as a direct sum of
irreducible representations. This proves the proposition. 

Exercise 3.9. Let V be a finite-dimensional vector space. Let W ⊂ V be a subspace.

Show that there exists a subspace W ′ such that V = W ⊕ W ′ .

Hint: Start with a basis of W and extend it to a basis of V .

Exercise 3.10. Show that when V = W ⊕ W ′ the map π(w + w′ ) = w is a projection

map. Verify that ker(π) = W ′ and Im(π) = W . The map π is called a projection on W .
Notice that this map depends on the chosen complement W ′ .

Exercise 3.11. (1) Complement of a subspace is not unique. Let us consider

2
V = R . Take a line L passing through the origin. It is a one-dimensional subspace.
Prove that any other line is a complement.
(2) Let W be the one-dimensional subspace x-axis. Choose the complement space as
y-axis and write down the projection map. What if we chose the complement as
the line x = y?

The exercises below show that Maschke’s theorem may not be true if we don’t have a
finite group.

Exercise 3.12. Let V = {(. . . , a−1 , a0 , a1 , a2 , . . .) | ai ∈ R} be the space (a vector space

of infinite dimension) and G = Z. Define ρ(1) to be the shift operator and ρ(n) is obtained
by composing ρ(1) n-times. Show that this is a representation of Z. It has an invariant
subspace < (. . . , 1, 1, 1, . . .) >. Is it completely reducible?

Exercise 3.13. Consider a two dimensional representation of R as follows:

 3. MASCHKE’S THEOREM 21

!
1 a
a 7→ .
0 1
It leaves a one-dimensional subspace fixed generated by (1, 0) but it has no invariant com-
plementary subspace. Hence this representation is not completely reducible.

Exercise 3.14. Let k = Z/pZ. Consider a two-dimensional representation of the cyclic

group G = Z/pZ of order p over k of characteristic p defined as in the previous example.
Find a subspace to show that Maschke’s theorem does not hold.

Exercise 3.15. Let V = Cn be the n-dimensional complex vector space. Let G be a

finite group.
P
(1) Show that h, i : V × V → C given by h(z1 , . . . , zn ), (w1 , . . . , wn )i = ni=1 zi w̄i is a
non-degenerate inner product. (Non-degenerate amounts to showing V ⊥ = 0 or
equivalently the matrix associated with the form is invertible).
(2) Let ρ : G → GL(Cn ) be a representation. Show that H : V × V → C defined by
1 X
H(z, w) = hρ(g)z, ρ(g)wi
|G|
g∈G

is an inner product.
(3) Show that H is G-invariant, i.e., for each x ∈ G we have H(z, w) = H(ρ(x)z, ρ(x)w).
(4) Show that if W is a G-invariant subspace of V then W ⊥ is also G-invariant, and
V = W ⊥ W ⊥ . This gives another proof of Maschke’s theorem.
 CHAPTER 4

Schur’s Lemma

The broad question we would like to deal with is that can we classify all representations,
up to some kind of equivalence (defined below). Further, in the wake of Maschke’s Theorem,
considering irreducible representations will be enough.

Definition 4.1 (G-map). Let (ρ, V ) and (ρ′ , V ′ ) be two representations of G over field
k. A linear map T : V → V ′ is called a G-map (between two representations) if it
satisfies the following:
ρ′ (t)T = T ρ(t) ∀t ∈ G.
The following diagram helps in understanding this condition:
T / V′
V
ρ(t) ρ′ (t)
 
T / V′
V
The G-maps are also called intertwiners.

Exercise 4.2. Prove that two representations of G are equivalent if and only if there
exists an invertible G-map.

In the case representations are irreducible the G-maps are easy to decide as follows:

Proposition 4.3 (Schur’s Lemma). Let (ρ, V ) and (ρ′ , V ′ ) be two irreducible represen-
tations of G (of dimension ≥ 1). Let T : V → V ′ be a G-map. Then, either T = 0 or T is
an isomorphism. Moreover, if T is non-zero then T is an isomorphism if and only if the
two representations are equivalent.

Proof. Let us consider the subspace ker(T ). We claim that it is a G-invariant subspace
of V . For this let us take v ∈ ker(T ). Then T ρ(t)(v) = ρ′ (t)T (v) = 0 implies ρ(t)(v) ∈
ker(T ) for all t ∈ G. Since V is irreducible we get, either ker(T ) = 0 or ker(T ) = V . In
the case ker(T ) = V the map T = 0.
Hence we may assume ker(T ) = 0, i.e., T is injective. Now we consider the subspace
Im(T ) ⊂ V ′ . We claim that it is also G-invariant. For this let y = T (x) ∈ Im(T ). Then
23
24 4. SCHUR’S LEMMA

ρ′ (t)(y) = ρ′ (t)T (x) = T ρ(t)(x) ∈ Im(T ) for any t ∈ G. Hence Im(T ) is G-invariant. Since
V ′ is irreducible we get either, Im(T ) = 0 or Im(T ) = V ′ . Since T is injective Im(T ) 6= 0
and hence Im(T ) = V ′ . This proves that in this case T is an isomorphism. 

Exercise 4.4. Let V be a vector space over C, and T ∈ End(V ) be a linear transfor-
mation. Show that there exists a one-dimensional subspace of V left invariant by T . Show
by example that this need not be true if the field is R instead of C.

Hint: Show that T has an eigenvalue then the corresponding eigenvector will do the job.

Corollary 4.5. Let (ρ, V ) be an irreducible representation of G over C. Let T : V → V

be a G-map. Then, T = λ.Id for some λ ∈ C and Id is the identity map on V .

Proof. Let λ be an eigen-value of T corresponding to the eigen-vector v ∈ V , i.e.,

T (v) = λv. Consider the subspace W = ker(T − λ.Id). We claim that W is a G-invariant
subspace. Since T and scalar multiplications are G-maps so is T − λ. Hence the kernel is
G-invariant (as we verified in the proof of Schur’s Lemma). One can do this directly also
see the exercise below.
Since W 6= 0 and is G-invariant subspace of irreducible representation V , we get W = V .
This gives T = λ.Id. 
Thus, if (ρ, V ) is an irreducible representation. Then, the set of all G-linear maps
EndG (V ) contains scalars k ⊂ EndG (V ) ⊂ End(V ), and is a (finite-dimensional) division
ring. In the case k = C, EndG (V ) is C the set of scalars only.

Exercise 4.6. Let T and S be two G-maps. Show that ker(T + S) is a G-invariant
subspace.
 CHAPTER 5

Representation Theory of Finite Abelian Groups over C

Throughout this chapter, G denotes a finite Abelian group.

Proposition 5.1. Let k = C and G be a finite Abelian group. Let (ρ, V ) be an irre-
ducible representation of G. Then, dim(V ) = 1.

Proof. The proof is a simple application of Schur’s Lemma. We will break it down in
the step-by-step exercises below. 

Exercise 5.2. With notation as in the proposition,

(1) for g ∈ G consider ρ(g) : V → V . Prove that ρ(g) is a G-map. (Hint: ρ(g)(ρ(h)(v)) =
ρ(gh)(v) = ρ(hg)(v) = ρ(h)(ρ(g)(v)).)
(2) Prove that there exists λ (depending on g) in C such that ρ(g) = λ.Id. (Hint: Use
the corollary of Schur’s Lemma.)
(3) Prove that the map ρ : G → GL(V ) maps every element g to a scalar map, i.e., it
is given by ρ(g) = λg .Id where λg ∈ C.
(4) Prove that the dimension of V is 1. (Hint: Take any one-dimensional subspace of
V . It is G-invariant. Use Maschke’s theorem on it as V is irreducible.)

Proposition 5.3. Let k = C and G be a finite Abelian group. Let ρ : G → GL(V ) be

a representation of dimension n. Prove that we can choose a basis of V such that ρ(G) is
contained in diagonal matrices.

Proof. Since V is a representation of a finite group we can use Maschke’s theorem to

write it as a direct sum of G-invariant irreducible ones, say V = W1 ⊕ . . . ⊕ Wr . Now using
Schur’s lemma we conclude that dim(Wi ) = 1 for all i and hence in turn we get r = n. By
choosing a vector in each Wi we get the required result. 

Corollary 5.4. Let G be a finite group (possibly non-commutative). Let ρ : G → GL(V )

be a representation. Let g ∈ G. Then there exists a basis of V such that the matrix of ρ(g)
is diagonal.
25
26 5. REPRESENTATION THEORY OF FINITE ABELIAN GROUPS OVER C

Proof. Consider H =< g >⊂ G and ρ : H → GL(V ) the restriction map. Since H
is Abelian, using the above proposition, we can simultaneously diagonalise elements of H.
This proves the required result. 
Remark 5.5. In ‘Linear Algebra’ we prove the following result: A commuting set of
diagonalizable matrices over C can be simultaneously diagonalised. The proposition above
is a version of the same result. We also give a warning about the corollary above that if we
have a finite subgroup G of GLn (C) then we can take a conjugate of G in such a way that
a particular element becomes diagonal.
Now this leaves us the question to determine all irreducible representations of an Abelian
group G. For this, we need to determine all group homomorphisms ρ : G → C∗ .
Exercise 5.6. Let G be a finite group (not necessarily Abelian). Let χ : G → C∗ be a
group homomorphism. Prove that |χ(g)| = 1 and hence χ(g) is a root of unity.
b be the set of all group homomorphisms from G to the multiplicative group C∗ .
Let G
bb b to C∗ .
Let us also denote G for the group homomorphisms from G
Exercise 5.7. With the notation as above,
(1) Prove that for G = G1 × G2 we have G b∼=G b1 × G
b2 .
(2) Let G = Z/nZ. Prove that G b = {χk | 0 ≤ k ≤ n − 1} is a group generated by χ1
2πir
of order n where χ1 (r) = e n and χk = χk1 . Hence Z/nZ\∼ = Z/nZ.
(3) Use the structure theorem of finite Abelian groups to prove that G ∼ b
= G.
bb
(4) Prove that G is naturally isomorphic to G given by g 7→ eg where eg (χ) = χ(g) for
all χ ∈ G.
Exercise 5.8 (Fourier Transform). For f ∈ C[Z/nZ] = {f | f : Z/nZ → C} we define
ˆ
f ∈ C[Z/nZ] by,
n−1 n−1
1X 1X
fˆ(q) = f (k)e(−kq) = f (k)χq (−k).
n n
k=0 k=0
Pn−1 ˆ Pn−1 ˆ Pn−1 Pn−1 ˆ 2
Show that f (k) = q=0 f (q)e(kq) = q=0 f (q)χq (k) and n1 2
k=0 |f (k)| = q=0 |f (q)| .
Pn−1
Exercise 5.9. On C[Z/nZ] let us define an inner product by hf, f ′ i = n1 j=0 f (j)f¯′ (j)
where bar denotes complex conjugation. Prove that {χk | 0 ≤ k ≤ n − 1} form an orthonor-
mal basis of C[Z/nZ]. Let X
f= cχ χ.
\
χ∈Z/nZ
Calculate the coefficients using the inner product and compare this with the previous exer-
cise.
 5.1. EXAMPLE OF REPRESENTATION OVER Q 27

Now we show that the converse of the Proposition 5.1 is also true.

Theorem 5.10. Let G be a finite group. Every irreducible representation of G over C

is 1 dimensional if and only if G is an Abelian group.

Proof. Let all irreducible representations of G over C be of dimension 1. Consider

the regular representation ρ : G → GL(V ) where V = C[G]). We know that if |G| ≥ 2
this representation is reducible and is an injective map (also called faithful representation).
Using Maschke’s theorem we can write V as a direct sum of irreducible ones and they are
given to be of dimension 1. Hence there exists a basis (check why?) {v1 , . . . , vn } of V
such that the subspace generated by each basis vector are invariant. Hence ρ(G) consists
of diagonal matrices with respect to this basis which is an Abelian group. Hence G ∼ = ρ(G)
is an Abelian group. 

5.1. Example of Representation over Q

5.1.1. An Irreducible Representation of Z/pZ. Consider G = Z/pZ where p is an

odd prime. Let K = Q(ζ) where ζ is a primitive pth root of unity. Let us consider the left
multiplication map lζ : K → K given by x 7→ ζx. Consider the basis {1, ζ, ζ 2 , . . . , ζ p−2 } of
K. Then lζ (1) = ζ, lζ (ζ i ) = ζ i+1 for 1 ≤ i ≤ p − 1 and lζ (ζ p−2 ) = ζ p−1 = −(1 + ζ + ζ 2 +
. . . + ζ p−2 ) and the matrix of lζ is:
 
0 0 0 · · · 0 −1
1 0 0 · · · 0 −1
 
 
0 1 0 · · · 0 −1
 
 .. .. .. .. 
. . . . 
0 0 0 · · · 1 −1
The map 1 7→ lζ defines a representation ρ : G → GLp−1 (Q). It is an irreducible represen-
tation of G.

5.1.2. An Irreducible Representation of the Dihedral Group D2p . Notice that

the Galois group Gal(K/Q) ∼ = Z/(p−1)Z comes with a natural representation on K. Let σ ∈
Gal(K/Q). Then σ is a Q-linear map which gives representation Gal(K/Q) ∼ = Z/(p−1)Z →
GLp−1 (Q). If we consider slightly different basis of K, namely, {ζ, ζ 2 , . . . , ζ p−1 } then the
matrix of each σ is a permutation matrix. In fact this way Gal(K/Q) ֒→ Sp−1 , the symmetric
group. However, this representation is not irreducible (the element ζ + ζ 2 + · · · + ζ p−1 is
invariant and gives decomposition).
Notice that the Galois automomrphism σ : K → K given by ζ 7→ ζ −1 is an order 2
element. We claim that σ normalizes lζ , i.e., σlζ σ = lζ −1 . Since σlζ σ(ζ i ) = σlζ (ζ −i ) =
28 5. REPRESENTATION THEORY OF FINITE ABELIAN GROUPS OVER C

σ(ζ 1−i ) = ζ i−1 = lζ −1 (ζ i ). Let us denote the subgroup generated by σ as H1 and the
subgroup generated by lζ by H2 . Then H1 H2 is a group of order 2p where H2 is a normal
subgroup of order p. Hence H1 H2 ∼ = D2p . This gives representation of order p − 1 of
2 −1
D2p = hr, s | rp = 1 = s , srs = r i given by D2p → GL(K) such that r 7→ lζ and s 7→ σ.

Exercise 5.11. Prove that the representation constructed above is irreducible.

Exercise 5.12. Write down the above representation concretely for D6 and D10 .
 CHAPTER 6

The Group Algebra k[G]

Let R be a ring (possibly non-commutative) with 1.

Definition 6.1. A (left) module M over a ring R is an Abelian group (M, +) with a
map (called scalar multiplication) R × M → M satisfying the following:
(1) (r1 + r2 )m = r1 m + r2 m for all r1 , r2 ∈ R and m ∈ M .
(2) r(m1 + m2 ) = rm1 + rm2 for all r ∈ R and m1 , m2 ∈ M .
(3) r1 (r2 m) = (r1 r2 )m for all r1 , r2 ∈ R and m ∈ M .
(4) 1.m = m for all m ∈ M .

Notice that this definition is the same as the definition of a vector space over a field.
Analogous to definitions there we can define submodules and module homomorphisms.

Example 6.2. If R = k (a field) or D (a division ring) then the modules are nothing
but vector spaces over R.

Example 6.3. Let R be a PID (a commutative ring such as Z or polynomial ring k[X]
etc). Then R × R . . . × R and R/I for an ideal I are modules over R. The structure theory
of modules over PID states that any module is a direct sum of these kinds. However, over
a non-PID things could be more complicated.

Example 6.4. In the non-commutative situation, the simple/semisimple rings are stud-
ied.

A module M over a ring R is called simple if it has no proper submodules. And a

module M is called semisimple if every submodule of M has a direct complement. It is
also equivalent to saying that M is a direct sum of simple modules.
A ring R is called semisimple if every module over it is semisimple. And a ring R is
called simple if it has no proper two-sided ideal.

Exercise 6.5. (1) Is Z a semisimple ring or simple ring?

(2) When Z/nZ a semisimple or simple ring?
(3) Prove that the ring Mn (D) where D is a division ring is a simple ring and the
module D n thought as one of the columns of this ring is a module over this ring.
29
30 6. THE GROUP ALGEBRA k[G]

All of the representation theory definitions can be very neatly interpreted in module
theory language. Given a field k and a group G, we form the ring
 
X 
k[G] = αg g | αg ∈ k
 
g∈G

called the group ringof G. We can also

Pdefine k[G] =P {f | f : G → k}. We define following

P P
operations on k[G]: g∈G αg g + g∈G βg g = g∈G (αg + βg ) g, λ g∈G αg g =
P
g∈G (λαg )g and the multiplication (recall the convolution definition) by
    !  
X X X X X X
 αg g .  βg g = αt βt−1 g g =  αt βs  g.
g∈G g∈G g∈G t∈G g∈G ts=g∈G

With the above operations, k[G] is an algebra called the group algebra of G (clearly it’s
a ring). A representation (ρ, V ) for G is equivalent to taking a k[G]-module V (see the
exercises below).

Exercise 6.6. Prove that k[G] is a ring as well as a vector space of dimension |G|. In
fact, it is a k-algebra.

Exercise 6.7. Let k be a field. Let G be a group. Then,

(1) (ρ, V ) is a representation of G if and only if V is a k[G]-module.
(2) W is a G-invariant subspace of V if and only if W is a k[G]-submodule of V .
(3) The representations V and V ′ are equivalent if and only if V is isomorphic to V ′
as k[G]-module.
(4) V is irreducible if and only if V is a simple k[G]-module.
(5) V is completely reducible if and only if V is a semisimple module.

We can rewrite Maschke’s Theorem and Schur’s Lemma in modules language:

Theorem 6.8 (Maschke’s Theorem). Let G be a finite group and k a field. The ring
k[G] is semisimple if and only if char(k) ∤ |G|.

Proposition 6.9 (Schur’s Lemma). Let M, M ′ be two non-isomorphic simple R mod-

ules. Then HomR (M, M ′ ) = {0}. Moreover, HomR (M, M ) is a division ring.

Proof. The proof is left as an exercise. 

Exercise 6.10. Let D be a finite-dimensional division algebra over C then D = C.

Exercise 6.11. Let R be a finite-dimensional algebra over C and M a simple module

over R. Suppose M is a finite-dimensional over C. Then HomR (M, M ) ∼
= C.
 CHAPTER 7

Constructing New Representations

Here we will see how we can get new representations out of the known ones. All of
the representations are considered over any field k. Let (ρ, V ) be a representation of G.
The character of (corresponding to) representation ρ is a map χ : G → k is defined by
χ(t) = tr(ρ(t)) where tr is the trace of the corresponding matrix. We will deal with
“character theory” in the following chapters.

7.1. Subrepresentation of a Representation

Suppose we have a representation (ρ, V ) of G, i.e., we have ρ : G → GL(V ). Let W

be a G-invariant subspace of V . Then, we can define a subrepresentation (ρ̃, W ) of G i.e.,
ρ̃ : G → GL(W ) by ρ̃(t)(w) = ρ(t)(w). In matrix notation,! if we choose a basis of W first
A(g) B(g)
and extend it to a basis of V then ρ(g) = and ρ̃(g) = A(g).
C(g)

7.2. Sum of Representations

Let (ρ, V ) and (ρ′ , V ′ ) be two representations of the group G of dimension n and n′
respectively. We define (ρ ⊕ ρ′ , V ⊕ V ′ ) by (ρ ⊕ ρ′ )(t)(v, v ′ ) = (t(v), t(v ′ )). This is called the
sum of the two representations. In matrix notation if the ! representations are ρ : G → GLn (k)
ρ(t)
and ρ′ : G → GLn′ (k) then (ρ ⊕ ρ′ )(t) = .
ρ′ (t)

Exercise 7.1. If χ and χ′ are characters of ρ and ρ′ respectively then the character of
ρ ⊕ ρ′ is χ + χ′ .

If ρ is a representation and V = W ⊕ W ′ is a G-invariant decomposition then ρ is a sum

of its two subrepresentations. Clearly, the idea of the sum of two representations defined
here can be extended to a finite sum.

7.3. Adjoint Representation

Let V be a vector space over k of dimension n. Let us recall some basic linear algebra.
A linear map f : V → k is called a linear functional. We denote V ∗ = Homk (V, k), the
31
32 7. CONSTRUCTING NEW REPRESENTATIONS

set of all linear functionals. We define operations on V ∗ by (f1 + f2 )(v) = f1 (v) + f2 (v)
and (λf )(v) = λf (v) and it becomes a vector space. The vector space V ∗ is called the dual
space of V .

Exercise 7.2. Let V be a vector space over k with a basis {e1 , . . . , en }. With the
notation as above,
(1) Check that V ∗ is a vector space with basis e∗i which are defined by e∗i (ej ) = δij .
Hence it has the same dimension as V . After fixing a basis of V we can obtain a
basis of V ∗ this way which is called a dual basis with respect to the given one.
(2) Show that V is “naturally” isomorphic (by a map which is defined without the
requirement of a basis and is an isomorphism) to V ∗∗ .

Let T : V → V be a linear map. We define a map T ∗ : V ∗ → V ∗ by T ∗ (f )(v) = f (T (v))

represented in diagram as follows:
T
V ❅ /V
❅ ❅❅
❅❅
❅ f
T ∗ (f ) ❅ 
k.
The map T ∗ is called dual or adjoint of T .

Exercise 7.3. Fix a basis of V and consider the dual basis of V ∗ with respect to that
(as in the previous exercise). Let A = (aij ) be the matrix of T . Show that the matrix of
T ∗ with respect to the dual basis is tA = (aji ), the transpose matrix.

Now, with the knowledge of the dual/adjoint, we can define adjoint representation. Let
ρ : G → GL(V ) be a representation. We define the adjoint representation (ρ∗ , V ∗ ) as
follows: ρ∗ : G → GL(V ∗ ) where ρ∗ (g) = ρ(g−1 )∗ (the inverse in this definition is forced
to make ρ∗ a group homomorphism). In the matrix form if we have a representation
τ : G → GLn (k) then τ ∗ : G → GLn (k) is given by τ ∗ (g) = t τ (g)−1 .

Exercise 7.4. With the notation as above,

(1) Show that ρ∗ is a representation of G of the same dimension as ρ.
(2) Fix a basis and suppose τ is the matrix form of ρ then show that the matrix form
of ρ∗ is τ ∗ .
(3) Prove that if ρ is irreducible then so is ρ∗ .

The last exercise will become easier in the case of complex representations once we define
characters as we will have a simpler criterion to test when a representation is irreducible.
 7.6. TENSOR PRODUCT OF TWO REPRESENTATIONS 33

Exercise 7.5. When k = C, we get that the character of ρ∗ , χ∗ = χ̄. Further, χ

is irreducible if and only if χ∗ is irreducible (Over C, this would be easier to prove using
orthogonality relations).

7.4. Restriction of a Representation

Let (ρ, V ) be a representation of the group G. Let H be a subgroup. Then (ρ, V ) is a

representation of H also obtained by composing H ֒→ G → GL(V ) denoted as ρH or ρ|H
or Res(ρ).

7.5. Lift of a Representation

Let N be a normal subgroup of G. Thus, we have the quotient map π : G → G/N .

Then any representation of G/N gives rise to a representation of G:
π ρ
G → G/N → GL(V ).
Moreover, if the representation of G/N is irreducible then the representation of G remains
irreducible. We use this technique to obtain 1-dimensional representations by taking N =
[G, G].

7.6. Tensor Product of two Representations

There is a general method to construct the tensor product of two R modules M and
M ′, namely M ⊗R M ′ which is again an R-module. It can be done even when R is not
commutative. The tensor product is an R-module denoted as M ⊗R M ′ which is defined
by the following universal property: The module (T, φ) where φ : M × M ′ → T is a bilinear
map, is said to be the tensor product (to be denoted as M ⊗ M ′ ) if for any R-module P
and any bilinear map β : M × M ′ → P there exists a unique linear map β̃ such that the
following diagram commutes:
β
M × M▼′ /P
O
▼▼▼
▼▼▼
▼ β̃
φ ▼▼&
M ⊗ M′

i.e., β = β̃φ. At the moment we need this only for vector spaces which is easier to deal
with.
Let V and V ′ be two vector spaces over M
k. First, we define the tensor product of two
vector spaces as follows. Consider V = k(v, v ′ ) (where (v, v ′ ) serve as a basis)
(v,v′ )∈V ⊕V ′
and the subspace W spanned by elements (v1 + v2 , v ′ ) − (v1 , v ′ ) − (v2 , v ′ ), (v, v1′ + v2′ ) −
34 7. CONSTRUCTING NEW REPRESENTATIONS

(v, v1′ ) − (v, v2′ ), a(v, v ′ ) − (av, v ′ ) and a(v, v ′ ) − (v, av ′ ) for all v, v1 , v2 ∈ V , v ′ , v1′ , v2′ ∈ V ′
and a ∈ k. Then the tensor product denoted as V ⊗ V ′ is given by the quotient space V/W.
Most of the time we don’t need to worry about the construction but rather get used to the
idea of using it. Thus a more practical way is as follows.
P
Tensor product of V and V ′ is a vector space V ⊗ V ′ = { ri=1 vi ⊗ vi′ | vi ∈ V, vi′ ∈ V ′ }
with the following properties:
P P
• ( ri=1 vi ⊗ vi′ ) + ( si=1 wi ⊗ wi′ ) = v1 ⊗ v1′ + · · · + vr ⊗ vr′ + w1 ⊗ w1′ + · · · + ws ⊗ ws′ .
• (v1 + v2 ) ⊗ v ′ = v1 ⊗ v ′ + v2 ⊗ v ′ and v ⊗ (v1′ + v2′ ) = v ⊗ v1′ + v ⊗ v2′ .
P P P
• λ ( ri=1 vi ⊗ vi′ ) = ri=1 λvi ⊗ vi′ = ri=1 vi ⊗ λvi′ .
Let {e1 , . . . , en } be a basis of V and {e′1 , . . . , e′m } be that of V ′ . Then {ei ⊗ e′j | 1 ≤ i ≤
n, 1 ≤ j ≤ m} is a basis of the vector space V ⊗ V ′ hence the dimension of V ⊗ V ′ is nm.
Warning: Elements of V ⊗ V ′ are not exactly of kind v ⊗ v ′ but they are supposed to be
a finite sum of these ones!

Exercise 7.6. For any vector space W and given a bilinear map φ : V × V ′ → W there
exists a linear map φ̃ : V ⊗ V ′ → W such that φ(v, v ′ ) = φ̃(v ⊗ v ′ ). Thus, Bilk (V × V ′ , ?) =
Endk (V ⊗ V ′ , ?).

Given linear maps T ∈ End(V ) and S ∈ End(V ′ ) we can define a linear map T ⊗ S ∈
P P
End(V ⊗ V ′ ) as follows: (T ⊗ S)( v ⊗ v ′ ) = T (v) ⊗ S(v ′ ).

Exercise 7.7. Given the matrix of T and S compute the matrix of T ⊗ S. If the matrix
of T with respect to the basis {ei } is A = (aij ) and the matrix of S with respect to the
basis {e′i } is B = (bij ) then the matrix of T ⊗ S with respect to the basis {er ⊗ e′s } will turn
out to be (aij B).

Exercise 7.8. Show that T race(T ⊗ S) = T race(T )T race(S).

With the definition of tensor product in hand, we can define new representations as
follows. Let (ρ, V ) and (ρ′ , V ′ ) be two representations of the group G of dimension n and
n′ respectively. We define a representation (ρ ⊗ ρ′ , V ⊗ V ′ ) of G as follows:
X X
(ρ ⊗ ρ′ )(t)( v ⊗ v′ ) = (ρ(t)(v) ⊗ ρ′ (t)(v ′ )).

Exercise 7.9. Choose a basis of V and V ′ . Let A = (aij ) be the matrix of ρ(t) and
B = (blm ) be that of ρ′ (t). What is the matrix of (ρ ⊗ ρ′ )(t)?

Exercise 7.10. If χ and χ′ are characters of ρ and ρ′ respectively then the character
of ρ ⊗ ρ′ is χχ′ .
 7.7. DECOMPOSITION OF THE REPRESENTATION V ⊗ V 35

If (ρ, V ) is a representation of G then V ⊗d = V ⊗ · · · ⊗ V, Symd V, ∧d (V ) are also

representations of G. This way starting from one representation we can get many represen-
tations. Though, even if you start from an irreducible representation the above-constructed
representations need not be irreducible (for example V ⊗ V given above) but often they
contain other irreducible representations. Writing down the direct sum decomposition of
tensor representations is an important topic of study. Often it happens (for semisimple
Lie algebras) that we need a much smaller number of representations (called fundamental
representations) of which tensor products contain all irreducible representations (called
highest weight theory in the case of semisimple Lie algebras).

Exercise 7.11. Let (ρ, V ) be a representation of G and χ be its character. Show that
the character of ρ⊗m is χm .

Exercise 7.12. Can we decompose V ⊗n in terms of Symi (V ) and ∧j (V )?

Exercise 7.13. Consider the group SL2 (R). This group has a natural representation
on 2-dimensional real vector!space V2 . We write this, say with basis {x, y}, as follows: for
a b
g ∈ SL2 (R) write g = then,
c d
g.x = ax + cy and g.y = bx + dy.

Now consider a real vector space Vd+1 for d ≥ 1 with basis

{xd , xd−1 y, xd−2 y 2 , . . . , xy d−1 , y d }.

We define a representation ρd+1 : SL2 (R) → GL(Vd+1 ) as follows:

ρd+1 (g)(xi y j ) := (g.x)i (g.y)j

where the right-hand side is expanded as polynomials. Show that ρd+1 is an irreducible
representation. Further, ρd+1 is nothing by symmetric power representation symd (V2 ).
Thus, V2 is an example of fundamental representation in this case.

7.7. Decomposition of the Representation V ⊗ V

Let G be a finite group. Suppose that (ρ, V ) and (ρ′ , V ′ ) are representations of G. Then
we can get a new representation of G from these representations defined as follows (recall
from Section 7.6):

ρ ⊗ ρ′ : G → GL(V ⊗ V ′ )
(ρ ⊗ ρ′ )(g)(v ⊗ v ′ ) = ρ(g)(v) ⊗ ρ′ (g)(v ′ )
36 7. CONSTRUCTING NEW REPRESENTATIONS

Now suppose ρ and ρ′ are representations over C and χ and χ′ are the corresponding
characters, then the character of ρ ⊗ ρ′ is χχ′ given by (χχ′ )(g) = χ(g)χ′ (g). However, even
if ρ and ρ′ are irreducible ρ ⊗ ρ′ need not be irreducible.
Let (ρ, V ) be a representation of G. We consider (ρ ⊗ ρ, V ⊗ V ). As defined above it
is a representation of G with character χ2 where χ2 (g) = χ(g)2 . Recall from section 7.6 it
can be decomposed (as a vector space) as

V ⊗ V = Sym2 (V ) ⊕ ∧2 (V ).

We prove below that both Sym2 (V ) and ∧2 (V ) are G-spaces.

Theorem 7.14. Let (ρ, V ) be a C-representation of G. Then, V ⊗ V = Sym2 (V ) ⊕

∧2 (V ) where each of the subspaces Sym2 (V ) and ∧2 (V ) are G-invariant.

Proof. Let {v1 , . . . , vn } be a basis of V . Then {vi ⊗ vj | 1 ≤ i, j ≤ n} is a basis of

V ⊗ V with dimension n2 . Consider the linear map θ defined on the basis of V ⊗ V by
θ(vi ⊗ vj ) = vj ⊗ vi and extended linearly. We observe that θ can also be defined without the
P P
help of any basis by θ(v ⊗ w) = w ⊗ v since if v = ai vi and w = bj vj , then θ(v ⊗ w) =
P P P P P
θ ( ai vi ⊗ bj vj ) = ai bj θ(vi ⊗ vj ) = ai bj (vj ⊗ vi ) = (bj vj ⊗ ai vi ) = (w ⊗ v).
2
Observe that θ = 1 and we take the subspaces of V ⊗ V corresponding to eigenvalues 1
and −1:

Sym2 (V ) = {x ∈ V ⊗ V | θ(x) = x}
∧2 (V ) = {x ∈ V ⊗ V | θ(x) = −x}

Note that {(vi ⊗vj +vj ⊗vi ) | 1 ≤ i ≤ j ≤ n} is a basis for Sym2 (V ) and hence its dimension
is n(n+1)
2 and {(vi ⊗ vj − vj ⊗ vi ) | 1 ≤ i < j ≤ n} is a basis for ∧2 (V ) and hence dimension
is n(n−1)
2 .
We claim that Sym2 (V ) and ∧2 (V ) both are G-invariant. Suppose that v ⊗ w ∈
Sym2 (V ) and g ∈ G then we have
 X 
θ(ρ(g)(v ⊗ w)) = θ ρ(g) λij (vi ⊗ vj + vj ⊗ vi )
X 
= θ λij (ρ(g)vi ⊗ ρ(g)vj + ρ(g)vj ⊗ ρ(g)vi )
X
= λij [θ(ρ(g)vi ⊗ ρ(g)vj ) + θ(ρ(g)vj ⊗ ρ(g)vi )]
X
= λij [ρ(g)vj ⊗ ρ(g)vi + ρ(g)vi ⊗ ρ(g)vj )]
X 
= ρ(g) λij (vj ⊗ vi + vi ⊗ vj )
= ρ(g)(v ⊗ w).
 7.8. INDUCED REPRESENTATION 37

We also see that Sym2 (V ) ∩ ∧2 (V ) = {0}. Also, their individual dimensions add up to
n(n+1)
2 + n(n−1)
2 = n2 = dim(V ⊗ V ), hence we have V ⊗ V = Sym2 (V ) ⊕ ∧2 (V ) as G-
spaces. 

Theorem 7.15. The characters of Sym2 (V ) and ∧2 (V ) are χS and χA respectively

given by
1 2

χS (g) = χ (g) + χ(g 2 )
2
1 2

χA (g) = χ (g) − χ(g 2 ) .
2

Proof. Suppose that |G| = d. Then for any g ∈ G, (ρ(g))d = I. Thus, m(X), the
minimal polynomial of ρ(g), divides the polynomial p(X) = X d − 1. Since p(X) has distinct
roots so will m(X) and hence ρ(g) is diagonalisable.
Let {e1 , · · · en } be an eigenbasis for V and let {λ1 , · · · , λn } be the corresponding eigenval-
ues. Then, from the proof of the previous theorem, it follows that {(ei ⊗ ej − ej ⊗ ei ) | i < j}
is an eigenbasis for ∧2 (V ) with corresponding eigenvalues {λi λj | i < j}. We now have
 
X 1 X 2 X 2
1 2

χA (g) = T r(ρ ⊗ ρ)(g) = λi λj = λi − λi = χ (g) − χ(g 2 ) .
2 2
i<j

Now, we can calculate χS as χ2 = χS + χA . 

7.8. Induced Representation

See the chapter 17.

 PART – II

This part deals with the Character Theory where we mostly work with representations over
C.

39
 CHAPTER 8

Matrix Elements

Now onwards we assume that the field k = C. We also denote S1 = {α ∈ C | |α| = 1}.
Let G be a finite group. Then, C[G] = {f | f : G → C} is a vector space of dimension |G|.
Let f1 , f2 be two functions from G to C, i.e., f1 , f2 ∈ C[G]. We define a map ( , ) : C[G] ×
C[G] → C as follows,
1 X
(f1 , f2 ) = f1 (t)f2 (t−1 ).
|G|
t∈G
Note that ( , ) is a symmetric bilinear form.
Let ρ : G → GL(V ) be a representation. We can choose a basis and get a map in matrix
form ρ : G → GLn (C) where n is the dimension of the representation. This means we have,
 
a11 (g) a12 (g) · · · a1n (g)
 
 a21 (g) a22 (g) · · · a2n (g) 

g 7→  . .. .. 
. .. 
 . . . . 
an1 (g) an2 (g) · · · ann (g)
where the matrix entries are aij : G → C, i.e, aij ∈ C[G]. The maps aij ’s are called matrix
elements of ρ. Thus, to a representation ρ we can associate a subspace W of C[G] spanned
by aij . It can be shown that this subspace does not depend on the chosen basis. In
what follows, we will explore relations between these subspaces W associated to irreducible
representations of a finite group G. Let ρ1 , . . . , ρr , · · · be irreducible representations of G of
dimension n1 , · · · , nr , · · · respectively. We don’t know yet whether there are finitely many
irreducible representations which we will prove later. Let W1 , · · · , Wr , · · · be associated
subspaces of C[G] to the irreducible representations.

Theorem 8.1. Let (ρ, V ) and (ρ′ , V ′ ) be two non-equivalent irreducible representations
of G of dimension n and n′ respectively. Let aij and bij be the corresponding matrix elements
with respect to some fixed basis of V and V ′ respectively. Then,
(1) (ail , bmj ) = 0 for all i, 
j, l, m.
 1 if i = j and l = m
(2) (ail , amj ) = n δij δlm = n
1
0 otherwise.

41
42 8. MATRIX ELEMENTS

Proof. Let T : V → V ′ be a linear map. Define

1 X
T0 = ρ(t)T (ρ′ (t))−1 .
|G|
t∈G

Then, T0is a G-linear map. Using Schur’s Lemma 4.3, we get T 0 = 0. Let us denote the
matrix of T by xlm . Then, ijth entry of T 0 is zero for all i and j, i.e.,
1 XX
ail (t)xlm bmj (t−1 ) = 0.
|G|
t∈G l,m

Since T is arbitrary linear transformation the entries xlm are arbitrary complex num-
ber hence can be treated as indeterminate. Hence, coefficients of xlm are 0. This gives
1 P −1
|G| t∈G ail (t)bmj (t ) = 0 hence (ail , bmj ) = 0 for all i, j, l, m.
1 P
Now, let us consider T : V → V , a linear map. Again define T 0 = |G| t∈G ρ(t)T (ρ(t))
−1

which is a G-map. From Corollary 4.5, we get that T 0 = λ.Id where λ = n1 tr(T ) =
1 P 1 P 0 1 P
n l xll = n l,m xlm δlm since n.λ = tr(T ) = |G| t∈G tr(T ) = tr(T ). Now using matrix
elements we can write ijth term of T : 0

1 XX 1X
ail (t)xlm amj (t−1 ) = λδij = xlm δlm δij .
|G| n
t∈G l,m l,m

Again T is an arbitrary linear map so its matrix elements xlm can be treated as indetermi-
nate. Comparing coefficients of xlm we get:
1 X 1
ail (t)amj (t−1 ) = δlm δij .
|G| n
t∈G

Which gives (ail , amj ) = n1 δlm δij . 

Corollary 8.2. If f ∈ Wi and f ′ ∈ Wj with i 6= j then (f, f ′ ) = 0, i.e., (Wi , Wj ) = 0.

Exercise 8.3. Prove that, in both cases, T 0 is a G-map.

Exercise 8.4. Let aij and a′ij be the set of two different matrix elements of the repre-
sentation ρ with respect to a different basis. Then, there exists a liner transformation P ,
namely the base change, such that [aij (g)] = P [a′ij (g)]P −1 . Using this we can show that
the subspace < aij >=< a′ij >⊂ C[G].
 CHAPTER 9

Character Theory

We have C[G], space of all complex-valued functions on G which is a vector space of

dimension |G|. We define an inner product h , i : C[G] × C[G] → C by
1 X
hf1 , f2 i = f1 (t)f2 (t).
|G|
t∈G

Exercise 9.1. With the notation above,

(1) Prove that h , i is an inner product (non-degenerate) on C[G].
(2) If f1 and f2 take value in S1 ⊂ C then hf1 , f2 i = (f1 , f2 ◦ ι) where ι : g 7→ g −1 .

Definition 9.2 (Character of a Representation). Let (ρ, V ) be a representation of G.

The character of (corresponding to) a representation ρ is a map χ : G → C defined by
χ(t) = tr(ρ(t)) where tr is the trace of corresponding matrix. The character of an irreducible
representation is said to be irreducible character.

Strictly speaking, it is χρ but for the simplicity of notation we write χ only when it is clear
which representation it corresponds to.

Exercise 9.3. Let A, B ∈ GLn (k). Prove the following:

(1) tr(AB) = tr(BA).
(2) tr(A) = tr(BAB −1 ).

Exercise 9.4. Usually, we define the trace of a matrix. Show that in the above definition
of character it is a well-defined function. That is, prove that χ(t) doesn’t change if we choose
a different basis and calculate the trace of ρ(t). This is another way to say that trace is an
invariant of the conjugacy classes of GLn (k).

Exercise 9.5. If ρ and ρ′ are two isomorphic representations, i.e., they are G-equivalent,
then the corresponding characters are the same. The converse of this statement is also true
which we will prove later.

Proposition 9.6. If ρ is a representation of dimension n, and χ is the corresponding

character then,
43
44 9. CHARACTER THEORY

(1) χ(1) = n, the dimension of the representation.

(2) χ(t−1 ) = χ(t) for all t ∈ G where ¯ denotes the complex conjugation.
(3) χ(tst−1 ) = χ(s) for all t, s ∈ G, i.e., character is constant on the conjugacy classes
of G.
(4) For f ∈ C[G] we have (f, χ) = hf, χi.

Proof. For the proof of part two, we use Corollary 5.4 to calculate χ(t). From that
corollary every ρ(t) can be diagonalised (at a time not simultaneously which is enough for
our purposes), say diag{ω1 , · · · , ωn }. Since t is of finite order, say d, we have ρ(t)d = 1. That
is each ωjd = 1 means the diagonal elements are dth root of unity. We know that roots of
unity satisfy ωj−1 = ωj . Hence χ(t−1 ) = tr(ρ(t)−1 ) = ω1−1 + · · · + ωn−1 = ω1 + · · · + ωn = χ(t).
We can also prove this result by the upper triangulation theorem, i.e., every matrix ρ(t)
can be conjugated to an upper triangular matrix. Combining this with the fact that the
trace is the same for the conjugate matrices, gives the result. 

Definition 9.7 (Class Function). A function f : G → C is called a class function if f

is constant on the conjugacy classes of G. We denote the set of class functions on G by H.

Exercise 9.8. With the notation as above,

(1) Prove that H is a subspace of C[G].
(2) The dimension of H is the number of conjugacy classes of G.
P
(3) Let cg = x∈G xgx−1 ∈ C[G]. The centre of the group algebra C[G] is spanned
by cg .
(4) Let χ be a character corresponding to some representation of G. Then, χ ∈ H.

Proposition 9.9. Let (ρ, V ) and (ρ′ , V ′ ) be two representations of the group G and
χ, χ′ be the corresponding characters. Then,
(1) The character of the sum of two representations is equal to the sum of characters,
i.e., χρ⊕ρ′ = χ + χ′ .
(2) The character of the tensor product of two representations is the product of two
characters, i.e., χρ⊗ρ′ = χχ′ .

Proof. The proof is a simple exercise involving matrices. 

This way we can define the sum and product of characters which is again a character.
 CHAPTER 10

Orthogonality Relations

Let G be a finite group. Let W1 , W2 , . . . , Wh , . . . be irreducible representations of G

of dimension n1 , n2 , . . . , nh , . . . over C. Let χ1 , χ2 , . . . , χh , . . . are corresponding characters,
called irreducible characters of G. We will fix this notation from now onwards. We
will prove that the number of irreducible characters and hence the number of irreducible
representations are finite and equal to the number of conjugacy classes.
In the last chapter, we introduced an inner product h, i on C[G]. We also observed that
character of any representation belongs to H, the space of class functions.

Theorem 10.1. The set of irreducible characters {χ1 , χ2 , . . .} form an orthonormal set
of (C[G], h, i). That is,
(1) If χ is a character of an irreducible representation then hχ, χi = 1.
(2) If χ and χ′ are two irreducible characters of non-isomorphic representations then
hχ, χ′ i = 0.

Proof. Let ρ and ρ′ be non-isomorphic irreducible representations and (aij ) and (bij )
P P
be the corresponding matrix elements. Then χ(g) = i aii (g) and χ′ (g) = j bjj (g). Then
P P P
hχ, χ′ i = (χ, χ′ ) = ( i aii , j bjj ) = i,j (aii , bjj ) = 0 from Theorem 8.1 and Proposi-
P P
tion 9.6 part 4. Using the similar argument we get hχ, χi = (χ, χ) = ( i aii , j bjj ) =
P Pn 1
i (aii , bii ) = i=1 n = 1 where n is the dimension of the representation ρ. 

Corollary 10.2. (1) The set of irreducible characters form a linearly independent
subset of C[G].
(2) The number of irreducible characters is finite.

Proof. Since the irreducible characters form an orthonormal set they are linearly in-
dependent. Hence their number has to be less than the dimension of C[G] which is |G|,
hence finite. 

Once we prove that two representations are isomorphic if and only if their characters are
the same this corollary will also give that there are finitely many non-isomorphic irreducible
representations. We prove this for the irreducible representations first.
45
46 10. ORTHOGONALITY RELATIONS

Proposition 10.3. Let (ρ, V ) and (ρ′ , V ′ ) be two irreducible representations with char-
acters χ and χ′ respectively. Then, (ρ, V ) and (ρ′ , V ′ ) are equivalent if and only if χ = χ′ .

Proof. Let χ = χ′ and we need to show V and V ′ are G-equivalent. On the contrary,
suppose V and V ′ are not equivalent. Let aij and bij be the corresponding matrix elements
P P
of ρ and ρ′ respectively. Then, χ = ′
i aii and χ = j bjj . Now, from Theorem 8.1
′
P P P ′
hχ, χ i = h aii , bjj i = i,j haii , bjj i = 0. However, hχ, χ i = hχ, χi = 1, a contradiction.
Thus, V and V ′ must be G-equivalent. 
We are going to use the above results to analyse the general representation of G and
identify its irreducible components.

Theorem 10.4. Let (ρ, V ) be a representation of G with character χ. Let V decompose

into a direct sum of irreducible representations:
V = V1 ⊕ · · · ⊕ Vm .
Then the number of Vi isomorphic to Wj (a fixed irreducible representation) is equal to the
scalar product hχ, χj i.

Proof. Let φ1 , . . . , φm be the characters of V1 , . . . , Vm . Then χ = φ1 + · · · + φm . Also

P P
hχ, χj i = i hφi , χj i = φi =χj hφi , χj i = the number of Vi isomorphic to Wj . 

Corollary 10.5. With the notation as above,

(1) The number of Vi isomorphic to a fixed Wj does not depend on the chosen decom-
position.
(2) Let (ρ, V ) and (ρ′ , V ′ ) be two representations with characters χ and χ′ respectively.
Then V ∼ = V ′ if and only if χ = χ′ .

Proof. The proof of part 1 is clear from the theorem above. For the proof of part 2, it
is clear that if V ∼ = V ′ we get χ = χ′ . Now suppose χ = χ′ . Let V ∼
n
= W1n1 ⊕ · · · ⊕ Wh h and
V′ ∼ m
= W1m1 ⊕ · · · ⊕ Wh h be the decomposition as a direct sum of irreducible representations
(we can do this using Maschke’s Theorem) where ni , mj ≥ 0. Suppose χ1 , χ2 , . . . , χh be
the irreducible characters of W1 , . . . , Wh . Then χ = n1 χ1 + · · · + nh χh and χ′ = m1 χ1 +
· · · + mh χh . However as χi ’s form an orthonormal set they are linearly independent. Hence
χ = χ′ implies ni = mi for all i. Hence V ∼ = V ′. 
From this corollary, it follows that the number of irreducible representations is the same
as the number of irreducible characters which is less than or equal to |G|. In fact, later we
will prove that this number is equal to the number of conjugacy classes. The above analysis
also helps to identify whether a representation is irreducible by use of the following:
 10. ORTHOGONALITY RELATIONS 47

Theorem 10.6 (Irreducibility Criteria). Let χ be the character of a representation

(ρ, V ). Then hχ, χi is a positive integer and hχ, χi = 1 if and only if V is irreducible.

Proof. Let V ∼ n
= W1n1 ⊕ · · · ⊕ Wh h . Then χ = n1 χ1 + · · · + nh χh and
X
hχ, χi = hn1 χ1 + · · · + nh χh , n1 χ1 + · · · + nh χh i = n2i .
i
Hence hχ, χi = 1 if and only if one of the ni = 1, i.e, χ = χi for some i. Hence the result. 

Exercise 10.7. Let χ be an irreducible character. Show that χ is so. Hence a repre-
sentation ρ is irreducible if and only if ρ∗ is so.

Exercise 10.8. Use the above criteria to show that the “Permutations representation”,
and “Regular Representation” (defined in the second chapter) are not irreducible if |G| > 1.

Exercise 10.9. Let ρ be an irreducible representation and τ be a 1-dimensional repre-

sentation. Show that ρ ⊗ τ is irreducible.
 CHAPTER 11

Main Theorem of Character Theory

Now we deal with the main question of how many irreducible representations are there
for a given group G.

11.1. Regular Representation

Let G be a finite group and χ1 , . . . , χh be the irreducible characters of dimension

n1 , . . . , nh respectively. Let L be the left regular representation of G with the corresponding
character l.
Exercise 11.1. The character l of the regular representation is given by l(1) = |G| and
l(t) = 0 for all 1 6= t ∈ G.

Theorem 11.2. Every irreducible representation Wi of G is contained in the regular

representation with multiplicity equal to the dimension of Wi (which is denoted by ni ).
Hence, l = n1 χ1 + · · · + nh χh .

Proof. In the view of Theorem 10.4 the number of times χi is contained in the regular
1 1
representation is given by hl, χi i = |G| l(1)χi (1) = |G| |G|ni = ni . This proves the required
result. 
If we are asked to construct irreducible representations of a finite group we don’t know
where to look for them. This theorem ensures a natural place, namely the regular represen-
tation, where we can find all of them. For this reason, one may also call it a “God Given
Representation”. Some further properties are listed here as an exercise.
Exercise 11.3. With the notation as above,
P
(1) The degree ni satisfy hi=1 n2i = |G|.
P
(2) If 1 6= s ∈ G we have hi=1 ni χi (s) = 0.
Hints: This follows from the formula for l as in the theorem by evaluating s = 1 and
s 6= 1.
These relations among characters will be useful to determine the character table of the
group G.
49
50 11. MAIN THEOREM OF CHARACTER THEORY

11.2. The Number of Irreducible Representations

Now we will prove the main theorem of the character theory.

Theorem 11.4 (Main Theorem). The number of irreducible representations of G (up
to isomorphism) is equal to the number of conjugacy classes of G.
The proof of this theorem will follow from the following proposition. We know that the
irreducible characters χ1 , . . . , χh ∈ H and form an orthonormal set (see Theorem 10.1). We
prove that, in fact, they form an orthonormal basis of H and generate as an algebra whole
of C[G].
Proposition 11.5. The irreducible characters of G form an orthonormal basis of H,
the space of class functions.
To prove this we will make use of the following:
Lemma 11.6. Let f ∈ H be a class function on G. Let (ρ, V ) be an irreducible repre-
P
sentation of G of degree n with character χ. Let us define ρf = t∈G f (t)ρ(t) ∈ End(V ).
Then, ρf = λ.Id where λ = |G|
n hf, χi.

Proof. We claim that ρf is a G-map and use Schur’s Lemma to prove the result. For
any g ∈ G we have,
X X X
ρ(g)ρf ρ(g −1 ) = f (t)ρ(g)ρ(t)ρ(g −1 ) = f (t)ρ(gtg −1 ) = f (g−1 sg)ρ(s) = ρf .
t∈G t∈G s∈G

Hence ρf is a G-map. From Schur’s Lemma (see 4.5) we get that ρf = λ.Id for some λ ∈ C.
Now we calculate the trace of both sides:
X X 1 X
λ.n = tr(ρf ) = f (t)tr(ρ(t)) = f (t)χ(t) = |G| f (t)χ(t) = |G|hf, χi.
|G|
t∈G t∈G t∈G
|G|
Hence we get λ = n hf, χi. 
Proof of the Proposition 11.5. We need to prove that irreducible characters span
H as being orthonormal they are already linearly independent. Let f ∈ H. Suppose f is
orthogonal to each irreducible χi , i.e., hf, χi i = 0 for all i. Then we will prove f = 0.
Since hf, χi i = 0 it implies hf, χi i = 0 for all i. Let 
ρi be the corresponding
 irreducible
|G|
representation. Then from the previous lemma ρi f = n hf, χi i .Id = 0 for all i. Now
let ρ be any representation of G. From Maschke’s Theorem, it is a direct sum of ρi ’s, the
irreducible ones. Hence ρf = 0 for any ρ.
In particular we can take the regular representation L : G → GL(C[G]) for ρ and we
P P
get Lf = 0. Hence Lf (e1 ) = 0 implies t∈G f (t)L(t)(e1 ) = 0, i.e., t∈G f (t)et = 0 hence
f (t) = 0 for all t ∈ G. Hence f = 0. 
 11.3. ARTIN-WEDDERBURN DECOMPOSITION 51

Proposition 11.7. Let s ∈ G and let rs be the number of elements in the conjugacy
class of s. Let {χ1 , . . . , χh } be irreducible representations of G. Then,
P
(1) We have hi=1 χi (s)χi (s) = |G| rs .
P
(2) For t ∈ G not conjugate to s, we have hi=1 χi (s)χi (t) = 0.

Proof. Let us define a class function ft by ft (t) = 1 and ft (g) = 0 if g is not conjugate
P
to t. Since irreducible characters span H (see 11.5) we can write ft = hi=1 λi χi where
1 r t Ph
λi = hft , χi i = |G| rt χi (t). Hence ft (s) = |G| i=1 χi (t)χi (s) for any s ∈ G. This gives the
required result by taking s conjugate to t and not conjugate to t. 

11.3. Artin-Wedderburn Decomposition

There is a more conceptual proof of the main theorem mentioned in the last sec-
tion. It follows from a general decomposition theorem on semisimple algebras called Artin-
Wedderburn decomposition. We prove this in the case k = C.

Theorem 11.8 (Artin-Wedderburn). Let G be a finite group and k a field such that
char(k) ∤ |G|. Then
k[G] ∼
= Mn1 (D1 ) × . . . Mnr (Dr )
where Di are division algebras over k. In the case k = C, C[G] ∼ = Mn (C) × . . . × Mn (C).
1 r

Proof. We prove for the case k = C. We have C[G] ∼ n

= W1n1 ⊕ · · · ⊕ Wh h as C[G]-
module. We compute HomC[G] (C[G], C[G]) in two different ways. If we think of C[G] as a
ring (possibly non-commutative) then HomC[G] (C[G], C[G]) ∼
= C[G]opp . On the other hand
HomC[G] (C[G], C[G]) ∼ n n
= HomC[G] (W1n1 ⊕ · · · ⊕ Wh h , W1n1 ⊕ · · · ⊕ Wh h )
h
M
∼
= HomC[G] (Wini , Wini )
i=1
M
∼
= Mni (EndC[G] (Wi ))
M
∼
= Mni (C).

This gives us C[G]opp ∼

= ⊕hi=1 Mni (C). 

Exercise 11.9. Let R be a ring (possibly non-commutative) then EndR (R) ∼

= Ropp ,
the opposite ring.

Exercise 11.10. Prove the main theorem using the Artin-Wedderburn decomposition
by computing the dimension of the centre on both sides.

Exercise 11.11. There are no finite dimensional division algebras over C.

52 11. MAIN THEOREM OF CHARACTER THEORY

Exercise 11.12. Verify the steps of the proof in the above theorem using Schur’s
Lemma.

Exercise 11.13. Find the Artin Wedderburn decomposition of a cyclic group Z/nZ
over fields C, R and Q.

Exercise 11.14. Show that R[Q8 ] ∼

= R × R × R × R × H where H is the real quaternion
algebra.

Exercise 11.15. Show that R[D4 ] ∼

= R × R × R × R × M2 (R).

Exercise 11.16. Show that C[Q8 ] ∼

= C[D4 ] ∼
= C4 × M2 (C).

Exercise 11.17. What is the decomposition of Q[S3 ]?

11.4. Exercises on Character Table

Character table can be thought of as a matrix of size h, the number of conjugacy classes
in G, with complex entries. We denote here as X.

Exercise 11.18. Show that the character table X of a finite group G is an invertible
matrix.

Exercise 11.19. Is it possible to have a zero column in a character table?

Exercise 11.20. Let G be a finite group. The elements s, t ∈ G are conjugate if and
only if χ(s) = χ(t) for all (irreducible) characters χ of G.

Hint: Consider the class functions fs and ft which take vale 1 on the conjugacy classes
corresponding to s and t respectively and 0 elsewhere. Now use that H is spanned by
irreducible characters.

Exercise 11.21. An element g ∈ G is called real if g is conjugate to g −1 . A character

χ is called real if it takes all real values. Prove that the number of real conjugacy classes is
the same as the number of real irreducible characters.

Hint: When χ is irreducible, so is χ̄. Thus, χ → χ permutes the rows of X. Thus,

X = P X where P is a permutation matrix. Now, if we interchange the columns under g →
g −1 once again we get a permutation of X, which is in fact Xr = X (because χ(t−1 ) = χ(t)).
Thus, Xr = X = XQ for some permutation matrix Q. Now, P X = XQ gives P = XQX −1
hence trace(P ) = trace(Q). Note that the trace of a permutation matrix is the number of
fixed points and we are done.

Exercise 11.22. What is the determinant of the character table?

 11.4. EXERCISES ON CHARACTER TABLE 53

Hint: Let us begin with computing

!  
X |G| |G| |G|
t
XX = χm (gi )χ̄m (gj ) = diag ,..., ,...,
m
r1 rl rh
Q Qh
using column orthogonality relations. Thus, det(X)det(X) = l |G| r1 = i=1 |ZG (gi )| where
gi are representative of conjugacy classes. Now, make use of X = P X where P is a
permutation matrix to get the answer.

Exercise 11.23. Prove that the sum of any row in the character table is a non-negative
integer.

Hint: Look at the action of G on itself by conjugation. Take the corresponding

representation ρ : G → GL(k[G]). Calculate the character τ of ρ, i.e, show that τ (g) =
P P
|ZG (g)|. Now, write τ = hi=1 mi χi and prove that mi = hτ, χi i = hj=1 χi (gj ).
P
Exercise 11.24. Let χ be an irreducible character. Show that g∈G χ(g) = 0.

Hint: Consider the inner product of χ with the trivial character.

Exercise 11.25. Let ρ : G → GL(V ) be an irreducible representation and χ be its

character. Then, ker(ρ) is {g ∈ G | χ(g) = χ(1)}. Thus, ker(ρ) can be determined from
the character table as being a normal subgroup it is a union of certain conjugacy classes.

Hint: Suppose dimension of V is n. We know χ(g) = λ1 + · · · + λn where ρ(g)

can be thought of diagonal matrix diag(λ1 , . . . , λn ) and each λi are roots of unity. Then,
|χ(g)| ≤ n. When can equality occur gives the answer.

Definition 11.26. Let ρ : G → GL(V ) be a representation and χ be its character. The

kernel of character χ is defined to be ker(χ) = {g ∈ G | χ(g) = χ(1)}.

Exercise 11.27. A normal subgroup of G is a disjoint union of conjugacy classes.

Exercise 11.28. Any normal subgroup can be obtained by looking at the intersection
of the kernel of some characters.

Hint: Let N be a normal subgroup. Then, G acts on G/N giving rise to a representa-
tion, say ρ. Let ψ be its character. The kernel of character ψ is precisely N . However, we
P
don’t know if this character is irreducible. But, we can write ψ = mi χi . Then, ker(ρ) is
the intersection of kernels of χi appearing in this sum.

Exercise 11.29. Let χ be a character which takes all values 0 except at e. Show that
χ is a multiple of the regular character χreg .
54 11. MAIN THEOREM OF CHARACTER THEORY

1
Hint: Let us compute hχ, 1i = |G| χ(e) 6= 0 integer (here 1 is the trivial character).
Thus, χ(e) = hχ, 1i.|G| = hχ, 1i.χreg (e). Hence, χ = hχ, 1i.χreg .

Exercise 11.30. Let (ρ, V ) be a faithful representation of G. Then, every irreducible

representation of G is a subrepresentation of V ⊗m , for some m.

Hint: Let χ be the character of ρ and dimension be n. Let {χ(g) | g 6= e} =

Q
{α1 , . . . , αm }. Consider φ = χ mi=1 (χ − αi 1) in C[G]. Clearly, φ is a class function and
Q Qm
φ(g) = 0 if g 6= e whereas φ(e) = χ(e) m i=1 (χ(e) − αi ) = n i=1 (n − αi ) 6= 0. Hence, φ
is a multiple of the regular character χreg , i.e., φ = rχreg . Now, let χi be an irreducible
character. Let us assume contrary that χi doesn’t appear in χm for all m, i.e., hχi , χm i = 0
for all m. Then, hχi , φi = 0 and hence hχi , rχreg i = 0 which is a contradiction that χreg
contains all irreducible characters.

Exercise 11.31. Let χ be a non-linear (i.e., dimension is greater than 1) irreducible

character of G. Prove that there exists a conjugacy class where χ takes value 0.

Exercise 11.32. (1) Let G be a finite group of size n with the number of conju-
|G|
gacy classes h. Suppose |[G,G]| = r. Then, n + 3r ≥ 4h.
Hint: We know that G has h irreducible characters of dimension say, ni . We
also know that there are r irreducible characters of dimension 1. So, n = |G| =
Ph 2
P 2
P 2
P 2
i=1 ni = ni =1 ni + ni ≥2 ni = r + ni ≥2 ni ≥ r + 4(h − r).
(2) Commuting Probability Given a finite group G, the commuting probability
2
cp(G) = |{(g1 ,g2)∈G|G||g21 g2 =g2 g1 }| . Show that if G is non-Abelian, cp(G) ≤ 58 . This is
due to Erdös and Turan.
P P |G|
Hint: We write |{(g1 , g2 ) ∈ G2 | g1 g2 = g2 g1 }| = g∈G |ZG (g)| = g∈G |Cl(g)| =
h
|G|h (as each Cl(g) appears as many times in the sum). Hence, cp(G) = |G| . Using
h 1 3r
the previous exercise cp(G) = |G| ≤ 4 + 4n . Now, since G is non-Abelian we have
|G|
|[G, G]| ≥ 2, i.e, r = |[G,G]| ≤ |G|
This gives us cp(G) ≤ 14 + 4n
2 .
3r
≤ 14 + 83 = 58 .
(3) Show, by computing the cp(G) for groups of order 8, that equality holds in the
above.
 CHAPTER 12

Examples

Let G be a finite group. Let ρ1 , . . . , ρh be irreducible representations of G over C with

corresponding characters χ1 , . . . , χh . We know that the number h is equal to the number
of conjugacy classes in G. We can make use of this information about G and make a
character table of G. The character table is a matrix of size h × h of which rows are
labelled as characters and columns as conjugacy classes.

r1 = 1 r2 · · · rh
g1 = e g2 · · · gh
χ1 n1 = 1 1 · · · 1
χ2 n2
.. ..
. .
χh nh

where g1 , g2 , . . . denote representative of the conjugacy class and ri denotes the number
of elements in the conjugacy class of gi . The following proposition summarizes the results
proved about the characters. We also recall the inner product on C[G] defined by hf1 , f2 i =
1 P
|G| t∈G f1 (t)f2 (t).

Proposition 12.1. With the notation as above, we have,

(1) The number of conjugacy classes is the same as the number of irreducible characters
which is the same as the number of non-isomorphic irreducible representations.
(2) Two representations are isomorphic if and only if their characters are equal.
(3) A representation ρ with character χ is irreducible if and only if hχ, χi = 1.
(4) |G| = n21 + n22 + · · · + n2h where n1 = 1 corresponds to the trivial character.
(5) The characters form an orthonormal basis of H, i.e.,
X X
χi (t)χj (t) = rl χi (gl )χj (gl ) = δij |G|.
t∈G gl

That is, the rows of the character table are orthonormal.

55
56 12. EXAMPLES

(6) The columns of the character table also form an orthogonal set, i.e.,
h
X |G|
χi (gl )χi (gl ) =
rl
i=1

and
h
X
χi (gl )χi (gm ) = 0
i=1
where gl and gm are representative of different conjugacy classes.
(7) The character table matrix is an invertible matrix.
(8) The degree of irreducible representations divide the order of the group, i.e., ni | |G|.

The proof of the last statement will be done later.

Warning : If the character table of two groups are the same that does not imply that
the groups are isomorphic. Look at the character tables of Q8 and D4 . In fact, all non-
Abelian groups of order p3 (there are two non-isomorphic ones) have the same character
table (find a reference for this).

12.1. Groups having Large Abelian Subgroups

First, we will give another proof of the Theorem 5.10 using Character Theory.

Theorem 12.2. Let G be a finite group. Then G is Abelian if and only if all irreducible
representations are of dimension 1, i.e., ni = 1 for all i.

Proof. With the notation as above let G be Abelian. We have

|G| = n21 + · · · + n2h

where h = |G|. Hence the only solution to the equation is ni = 1 for all i. Now suppose
ni = 1 for all i. Then the above equation implies h = |G|. Hence each conjugacy class has
a size of 1 and the group is Abelian. 
|G|
Proposition 12.3. Let G be a group and A be an Abelian subgroup. Then ni ≤ |A| for
all i.

Proof. Let ρ : G → GL(V ) be an irreducible representation. We can restrict ρ to A

and denote it by ρA : A → GL(V ) which may not be irreducible. Let W an irreducible
ρA - invariant subspace of V . The above theorem implies dim(W ) = 1. Say W =< v >
where v 6= 0. Consider V ′ =< {ρ(g)v | g ∈ G} >⊂ V . Clearly V ′ is G-invariant and as
V is irreducible V ′ = V . Notice that ρ(ga)v = λρ(g)v, i.e., ρ(ga)v and ρ(g)v are linearly
 12.2. CHARACTER TABLE OF SOME GROUPS 57

dependent. Hence V ′ =< {ρ(g1 )v, . . . , ρ(gm )v} > where gi are representatives of the coset
gi A in G/A. This implies dim(V ) ≤ m = |G| |A| . 

Corollary 12.4. Let G = Dn be the dihedral group with 2n elements. Any irreducible
representation of Dn has dimension 1 or 2.

12.2. Character Table of Some Groups

Example 12.5 (Cyclic Group). Let G = Z/nZ. All representations are one-dimensional
2πi
hence giving character. The characters are χ1 , . . . , χn given by χr (s) = e n rs for 0 ≤ s ≤ n.

Example 12.6 (S3 ).

S3 = {1, (12), (23), (13), (123), (132)}
We already know the one-dimensional representations of S3 which are trivial representations
and the sign representation. The characters for these representations are themselves. Now
we use the formula 6 = |G| = n21 + n22 + n23 = 1 + 1 + n23 gives n3 = 2. Now we use
hχ3 , χ1 i = 16 {1.2.1 + 3.a.1 + 2.b.1} = 0 and hχ3 , χ2 i = 61 {1.2.1 + 3.a.(−1) + 2.b.1 = 0}. Hence
solving the above equations 3a + 2b = −2 and −3a + 2b = −2 gives a = 0 and b = −1.

r1 = 1 r2 = 3 r3 = 2
g1 = 1 g2 = (12) g3 = (123)
χ1 n1 = 1 1 1
χ2 n2 = 1 −1 1
χ3 n3 = 2 a=0 b = −1

Example 12.7 (Q8 ).

Q8 = {1, −1, i, −i, j, −j, k, −k}
The commutator subgroup of Q8 = {1, −1} and Q8 /{±1} ∼ = Z/2Z × Z/2Z. Hence it has
4 one-dimensional representations which are lifted from Z/2Z × Z/2Z. Notice that the
first Z/2Z component is the image of i and the second one of j. Now we use 8 = |G| =
n21 + n22 + n23 + n24 + n25 = 1 + 1 + 1 + 1 + n25 gives n5 = 2. Again using orthogonality of χ5
with other known χi ’s we get the following equations:
1.2.1 + 1.a.1 + 2.b.1 + 2.c.1 + 2.d.1 = 0
1.2.1 + 1.a.1 + 2.b.1 + 2.c.(−1) + 2.d.(−1) = 0
1.2.1 + 1.a.1 + 2.b.(−1) + 2.c.1 + 2.d.(−1) = 0
1.2.1 + 1.a.1 + 2.b.(−1) + 2.c.(−1) + 2.d.1 = 0
58 12. EXAMPLES

This gives the solution a = −2, b = 0, c = 0 and d = 0.

r1 = 1 r2 = 1 r3 = 2 r4 = 2 r5 = 2
g1 = 1 g2 = −1 g3 = i g4 = j g5 = k
χ1 n1 = 1 1 1 1 1
χ2 n2 = 1 1 1 −1 −1
χ3 n3 = 1 1 −1 1 −1
χ4 n4 = 1 1 −1 −1 1
χ5 n5 = 2 a = −2 b = 0 c = 0 d = 0

Example 12.8 (D4 ).

D4 = {r, s | r 4 = 1 = s2 , rs = sr −1 }

The commutator subgroup of D4 is {1, r 2 } = Z(D4 ). And D4 /Z(D4 ) ∼ = Z/2Z × Z/2Z hence
there are 4 one dimensional representations as in the case of Q8 . We can also compute the
rest of it as we did in Q8 . We also observe that the character table is the same as Q8 .

r1 = 1 r2 = 1 r3 = 2 r4 = 2 r5 = 2
g1 = 1 g2 = r 2 g3 = r g4 = s g5 = sr
χ1 n1 = 1 1 1 1 1
χ2 n2 = 1 1 1 −1 −1
χ3 n3 = 1 1 −1 1 −1
χ4 n4 = 1 1 −1 −1 1
χ5 n5 = 2 a = −2 b = 0 c = 0 d=0

Example 12.9 (A4 ).

A4 = {1, (12)(34), (13)(24), (14)(23), (123), (132), (124), (142), (134), (143), (234), (243)}

We note that the 3 cycles are not conjugate to their inverses. Here we have

H = {1, (12)(34), (13)(24), (14)(23)} ∼

= Z/2Z × Z/2Z

a normal subgroup of A4 and A4 /H ∼ = Z/3Z. This way the 3 one dimensional irreducible
representations of Z/3Z lift to A4 as we have maps A4 → A4 /H ∼= Z/3Z → GL1 (C) given
3 2 2 2 2
by ω where ω = 1. We use 1 + 1 + 1 + n4 = 12 = |A4 | to get n4 = 3. Now we take the
 12.2. CHARACTER TABLE OF SOME GROUPS 59

inner product of χ4 with others and get the equations:

1.3.1 + 3.a.1 + 4.b.1 + 4.c.1 = 0

1.3.1 + 3.a.1 + 4.b.ω + 4.c.ω 2 = 0
1.3.1 + 3.a.1 + 4.b.ω 2 + 4.c.ω = 0

This gives us the character table.

r1 = 1 r2 = 3 r3 = 4 r4 = 4
g1 = 1 g2 = (12)(34) g3 = (123) g4 = (132)
χ1 n1 = 1 1 1 1
χ2 n2 = 1 1 ω ω2
χ3 n3 = 1 1 ω2 ω
χ4 n4 = 3 a = −1 b=0 c=0

Example 12.10 (S4 ). The group S4 has 2 one-dimensional representations given by

trivial and sign. We recall that the permutation representation (Example 2.16) gives rise
to the subspace V = {(x1 , x2 , . . . , xn ) ∈ Cn | x1 + x2 + · · · + xn = 0} which is an n − 1
dimensional irreducible representation of Sn . We will make use of this to get a 3 dimensional
representation of S4 and corresponding character χ3 . A basis of the space V is {e1 − e2 , e2 −
e3 , e3 − e4 } and the action is given by:

(12) : (e1 − e2 ) 7→ e2 − e1 = −(e1 − e2 )

(e2 − e3 ) 7→ e1 − e3 = (e1 − e2 ) + (e2 − e3 )
(e3 − e4 ) 7→ (e3 − e4 )
 
−1 1 0
 
So the matrix is  0 1 0 and χ3 ((12)) = 1. The action of (12)(34) is e1 − e2 7→
0 0 1
e2 − e1 = −(e1 − e2 ), e2
− e3 7→ e1 − e4 =  (e1 − e2 ) + (e
 2 − e3 ) + (e3 − e4 ) and e3 −
−1 1 0
 
e4 7→ e4 − e3 = −(e3 − e4 ) and the matrix is  0 1 0 . So χ3 ((12)(34)) = −1. The
0 1 −1
action of (123) is e1 − e2 7→ e2 − e3 , e2 − e3 7→ e3 − e1 = −(e1 − e2 )− (e2 − e3) and
0 −1 1
 
e3 − e4 7→ e1 − e4 = (e1 − e2 ) + (e2 − e3 ) + (e3 − e4 ). So the matrix is 1 −1 1 and
0 0 1
χ3 ((123)) = 0. And the action of (1234) is e1 − e2 7→ e2 − e3 , e2 − e3 7→ e3 − e4 and
60 12. EXAMPLES
 
0 0 −1
 
e3 − e4 7→ e4 − e1 = −(e1 − e2 ) − (e2 − e3 ) − (e3 − e4 ). So the matrix is 1 0 −1 and
0 1 −1
χ3 ((1234)) = −1. This gives χ3 . We can get another character χ4 = χ3 .χ2 corresponding
to the representation P erm ⊗ sgn. We can check that this is different from others so
corresponds to a new representation and is also irreducible as hχ4 , χ4 i = 1.
To find χ5 we can use orthogonality relations and get equations.

r1 = 1 r2 = 6 r3 = 3 r4 = 8 r5 = 6
g1 = 1 g2 = (12) g3 = (12)(34) g4 = (123) g5 = (1234)
χ1 n1 = 1 1 1 1 1
χ2 n2 = 1 −1 1 1 −1
χ3 n3 = 3 1 −1 0 −1
χ4 n4 = 3 −1 −1 0 1
χ5 n5 = 2 a=0 b=2 c = −1 d=0

The representation to which χ5 corresponds is the 2-dimensional irreducible represen-

tation of S4 /V4 ∼
= S3 . Let us understand this isomorphism. Consider the action of S4 on
the set X = {x1 = {1, 2} ⊔ {3, 4}, x2 = {1, 3} ⊔ {2, 4}, x3 = {1, 4} ⊔ {2, 3}}. It’s a transitive
action with kernel V4 .

Example 12.11 (Dn , n even).

Dn = {a, b | an = 1 = b2 , ab = ba−1 }
n
and the conjugacy classes are {1}, {a 2 }, {aj , a−j } for all 1 ≤ j ≤ n2 − 1, {aj b | j even}
n
and {aj b | j odd}. The center is {1, a 2 }. The subgroup generated by a2 is a commutator
and is normal hence there are 4 one-dimensional representations. The rest of them are
two-dimensional (refer to Corollary 12.4) representations defined in the Example 2.7. Using
them we can make the character table.

Example 12.12 (Dn , n odd).

Dn = {a, b | an = 1 = b2 , ab = ba−1 }

and the conjugacy classes are {1}, {aj , a−j } for all 1 ≤ j ≤ n−1 j
2 and {a b | 0 ≤ j ≤ n − 1}.
The commutator subgroup is generated by a, and is normal and hence there are 2 one-
dimensional representations. The rest of them are two-dimensional (refer to Corollary 12.4)
representations defined in the Example 2.7. Using them we can make the character table.
 12.3. CHARACTERS OF DIRECT PRODUCT 61

Exercise 12.13. Write down a criterion such that a representation is conjugate to its
own adjoint. Find examples of representations which are not equivalent to their adjoint
(analyse one-dimensional representations).
Exercise 12.14. Let G be a group acting on a finite set X. Let us denote the corre-
sponding representation on k[X] by λX .
(1) A function f ∈ k[X] is fixed by λX if and only if f is constant on orbits of G in
X. In particular, dimk k[X]G is the number of orbits of G in X.
(2) Show that the character of this representation χλX (g) is the number of fixed points
of g in X.
(3) Show that k[X] = W ⊕ W0 , where W is space of constant functions and W0 =
P
{f ∈ k[X] | x∈X f (x) = 0}, is a G decomposition.
Exercise 12.15. In continuation to Exercise 12.14, now let us assume that G is doubly
transitive on X.
1 P
(1) Consider the diagonal action of G on X × X. Show that |G| g∈G χλX×X (g) = 2.
∼
(2) Show that k[X] ⊗ k[X] = k[X × X] as G-modules and hence χλX×X = χ2λX .
(3) Show that 2 = hχλX , χλX iG .
(4) Show that W0 is irreducible by showing that the character has norm 1.

12.3. Characters of Direct Product

Let G and G′ be two finite groups. Let (ρ, V ) be a representation of G and (ρ′ , V ′ ) be a
representation of G′ . Then, we can define (ρ⊗ρ′ , V ⊗V ′ ) a representation of G×G′ as follows:
P P
ρ × ρ′ : G × G′ → GL(V ⊗ V ′ ) such that (ρ ⊗ ρ′ )(g, g ′ )( v ⊗ v ′ ) = ρ(g)(v) ⊗ ρ(g ′ )(v ′ ).
Exercise 12.16. Show that ρ ⊗ ρ′ is a representation of G × G′ .
Exercise 12.17. Let χ be the character of ρ and χ′ be that of ρ′ . Show that the
character of ρ ⊗ ρ′ is χχ′ where (χχ′ )(g, g ′ ) = χ(g)χ′ (g′ ).
Proposition 12.18. Consider C-representations. If ρ and ρ′ are irreducible represen-
tations of G and G′ respectively then ρ ⊗ ρ′ is an irreducible representation of G × G′ .
Proof.
1 X
hχχ′ , χχ′ i = χ(t)χ′ (t′ )χ̄(t)χ̄′ (t′ )
|G × G′ |
(t,t′ )∈G×G′
! !
1 X X
= χ(t)χ̄(t) χ′ (t′ )χ̄′ (t′ )
|G||G′ |
t∈G t′ ∈G′
′ ′
= hχ, χihχ , χ i = 1.
62 12. EXAMPLES


Thus, if ρ1 . . . ρr are irreducible representations with irreducible characters χ1 , . . . , χr
of G, and ρ′1 . . . ρ′s are irreducible representations with irreducible characters χ′1 , . . . , χ′s of
G′ then ρi ⊗ ρ′j are irreducible representations of G ⊗ G′ with characters χχ′ .
From the Fundamental Theorem of Abelian groups, any Abelian group is a product of
cyclic groups. For the cyclic groups, all irreducible representations can be easily written
by finding out an appropriate root of unity in C. This section gives a complete method to
writing down all representations for an Abelian group explicitly.

Exercise 12.19. Compute the character table of all Abelian groups up to order 25.

Exercise 12.20. Show that the above Proposition need not be true for a semi-direct
product (try the Dihedral groups).

Once we understand the induced representations we will be able to find a method to

get characters of a semi-direct product of groups.
 CHAPTER 13

Character Table of S5

With the theory developed so far, we are now ready to construct the character table for
the symmetric group S5 . The size of this group is |S5 | = 5! = 120. Two elements σ and σ ′
of Sn are conjugate if and only if their cycle structure is the same when they are written as
products of disjoint cycles. Thus, the conjugacy classes in S5 and the size of classes are as
follows:
|(gi )| 1 10 20 15 30 20 24
gi 1 (12) (123) (12)(34) (1234) (12)(345) (12345)

We already know 2 one-dimensional (and hence irreducible) representations. One is the

trivial representation and the other the sign representation which sends each transposition
to −1.
For any Sn , we know an irreducible representation of dimension n − 1 (as in the
Example 2.16 obtained by the action of Sn on the subspace V of Cn given by V =
{(x1 , x2 , . . . , xn ) ∈ Cn | x1 + x2 + · · · + xn = 0}). We fill this information into the character
table. We need to find 4 more irreducible characters.
Let us take the tensor product of the trivial representation with any other representation
(ρ, V ) gives a representation isomorphic to (ρ, V ) (since χ1 χV = χV ). Hence we only need to
consider the tensor product of the second and the third representation namely sgn⊗V whose
1
character is χ2 χ3 . Check hχ2 χ3 , χ2 χ3 i = 120 ((4)2 + 10(−2)2 ) + 20(1)2 + 20(1)2 + 24(−1)2 ) =
1 = hχ3 , χ3 i. Thus this representation turns out to be irreducible. Let χ4 = χ2 χ3 6= χ3 .
We include this character χ4 into the character table:

|(gi )| 1 10 20 15 30 20 24
gi 1 (12) (123) (12)(34) (1234) (12)(345) (12345)
χ1 1 1 1 1 1 1 1
χ2 1 −1 1 1 −1 −1 1
χ3 4 2 1 0 0 −1 −1
χ4 4 −2 1 0 0 1 −1

We now consider χ = χ3 and examine the representations χS and χA :

63
64 13. CHARACTER TABLE OF S5

|(gi )| 1 10 20 15 30 20 24
gi 1 (12) (123) (12)(34) (1234) (12)(345) (12345)
χA 6 0 0 −2 0 0 1
χS 10 4 1 2 0 1 0
1
We check that χA is irreducible as hχA , χA i = 120 ((6)2 + 15(−2)2 + 24(1)2 ) = 1. Thus
χ5 = χA is the fifth irreducible character of S5 .
Suppose that χ6 and χ7 are the other 2 irreducible characters. Since every representation
of a finite group can be written as a direct sum of irreducible ones we have, χS = m1 χ1 +
m2 χ2 + · · · + m7 χ7 where mi = hχS , χi i. Calculations show that hχS , χS i = 3, hχS , χ1 i = 1
P 2
and hχS , χ3 i = 1. We also have mi = hχS , χS i = 3. Thus χS = χ1 + χ3 + ψ, where ψ is
an irreducible character. We rewrite ψ = χS − χ1 − χ3 explicitly and get
|(gi )| 1 10 20 15 30 20 24
gi 1 (12) (123) (12)(34) (1234) (12)(345) (12345)
ψ 5 1 −1 1 −1 1 0
1
Since ψ is irreducible, hψ, ψi = 120 (52 + 10 + 20 + 15 + 30 + 20) = 1 as expected. Let
χ6 = ψ then χ7 = χ2 χ6 will be another new irreducible character. We have thus found the
character table of S5 :
Character Table of S5
|(gi )| 1 10 20 15 30 20 24
gi 1 (12) (123) (12)(34) (1234) (12)(345) (12345)
χ1 1 1 1 1 1 1 1
χ2 1 −1 1 1 −1 −1 1
χ3 4 2 1 0 0 −1 −1
χ4 4 −2 1 0 0 1 −1
χ5 6 0 0 −2 0 0 1
χ6 5 1 −1 1 −1 1 0
χ7 5 −1 −1 1 1 −1 0
Using this and restriction we can find the character table of A5 which we do later.
 CHAPTER 14

Restriction of a Representation

Let (ρ, V ) be a representation of the group G, i.e., ρ : G → GL(V ) is a group homomor-

phism. Suppose H is a subgroup of G. Then we can restrict the map ρ to H denoted as
ResGH ρ or ρ|H to obtain a representation for H. Note that even if (ρ, V ) is an irreducible
representation of G, (ResG H ρ, V ) need not be irreducible for H.

Example 14.1. Consider a 2-dimensional irreducible representation of the Dihedral

Group and restrict it to its cyclic subgroup.

In this chapter we explore the connection between the characters of a group G and any of
its subgroups H. Recall from Chapter 9 we define an inner product on C[G] denoted as h, i.
We denote this inner product on C[H] by h, iH thinking of H as a group in itself.

Proposition 14.2. Let G be a group and H be a subgroup. Let ψ be a non-zero character

of H. Then there exists an irreducible character χ of G such that hResG
H χ, ψiH 6= 0.

Proof. Let χ1 , χ2 , . . . , χh be the irreducible characters of G. We have the regular

P
representation of G: χreg = hi=1 χi (1)χi , with values χreg (1) = |G|, χreg (g) = 0 for all
g 6= 1. Thus,
1 |G|ψ(1)
hResG H χreg , ψiH = χreg (1)ψ(1) = 6= 0.
|H| |H|
Ph
Therefore, hResGH χreg , ψiH =
G
i=1 χi (1)hResH χi , ψiH 6= 0. Since all quantities involved in
this sum are non-negative at least one of the hχi , ψiH 6= 0. 

Theorem 14.3. Let G be a group and H a subgroup. Let χ be an irreducible character

of G. Suppose ψ1 , ψ2 , . . . , ψk are all irreducible characters of H and ResG
H χ = d1 ψ1 +d2 ψ2 +
· · · + dk ψk for some d1 , d2 , . . . , dk integers. Then,
k
X |G|
d2i ≤
|H|
i=1

and equality occurs if and only if χ(g) = 0 for all g ∈

/ H.

65
66 14. RESTRICTION OF A REPRESENTATION

Pk
Proof. Thinking of H as a group we have hResG G
H χ, ResH χiH =
2
i=1 di . Since χ is
an irreducible character of G we have,
 
1 X 1  X X
1 = hχ, χi = χ(g)χ(g) = χ(h)χ(h) + χ(g)χ(g) 
|G| |G|
g∈G h∈H g ∈H
/

|H| 1 X
= hResG
H χ, Res G
H χiH + χ(g)χ(g).
|G| |G|
g ∈H
/

|H| G G 1 P
Rewriting the above we get, |G| hResH χ, ResH χiH =1− |G| g ∈H
/ χ(g)χ(g) which implies
X |G| 1 X |G|
d2i = hResG G
H χ, ResH χiH = − χ(g)χ(g) ≤ .
|H| |H| |H|
g ∈H
/

1 P P 2
Moreover equality occurs if and only if |H| / χ(g)χ(g) = 0, i.e.,
g ∈H / |χ(g)| = 0 which
g ∈H
happens if and only if |χ(g)| = 0 and hence χ(g) = 0 for all g ∈
/ H. 

We will apply the above theorem for index two subgroups and get the following,

Corollary 14.4. Let G be a group and H be a subgroup of index 2. Let χ be an

irreducible character of G. Then one of the following happens :
(1) ResGH χ = ψ is an irreducible character of H. This happens if and only if there
exists g ∈ G and g ∈/ H such that χ(g) 6= 0.
G
(2) ResH χ = ψ1 + ψ2 where ψ1 and ψ2 are irreducible characters of H. This happens
if and only if χ(g) = 0 for all g ∈
/ H.

Proof. Let us write ResG χ = d1 ψ1 + d2 ψ2 + · · · + dk ψk then from Theorem 14.3 we

P 2 P 2H
get di ≤ 2. Further, di = 2 if and only if χ(g) = 0 for all g ∈ / H, i.e., in this case
P 2
d1 = d2 = 1 is the solution and we get (2). Otherwise, di = 1 and the only solution is
d1 = 1 which gives the first case. 

14.1. Character Table of A5

Now we apply the results obtained so far on G = S5 and get the characters of its
subgroup H = A5 . We aim to write down the character table of A5 . For this, we will use
the Corollary 14.4 and the character table of S5 derived earlier in the previous chapter. The
conjugacy classes of A5 and their corresponding sizes are as follows:
|(gi )| 1 20 15 12 12
gi 1 (123) (12)(34) (12345) (13452)
 14.1. CHARACTER TABLE OF A5 67

From the character table of S5 it follows that Resχ1 = Resχ2 , Resχ3 = Resχ4 and Resχ6 =
Resχ7 are irreducible characters of A5 . Hence we get the partial character table of A5 as
follows.
|(gi )| 1 20 15 12 12
gi 1 (123) (12)(34) (12345) (13452)
ψ1 1 1 1 1 1
ψ2 4 1 0 −1 −1
ψ3 5 −1 1 0 0
ψ4 n4 = 3 a1 a2 a3 a4
ψ5 n5 = 3 b1 b2 b3 b4

We know that 12 + 42 + 52 + n4 2 + n5 2 = 60 and hence n4 2 + n5 2 = 18. The only possible

integral solutions of this equation are n4 = n5 = 3.
Now we see that the only irreducible character of S5 whose restriction to A5 is not
irreducible is χ5 . Since ψ1 , ψ2 , ψ3 are all obtained by restriction of the characters other than
χ5 it follows from Theorem 14.3 that only possibly hψ4 , Resχ5 iA5 6= 0 and hψ5 , Resχ5 iA5 6=
0. Now from Corollary 14.4 it follows that Resχ5 is a sum of two characters of A5 and
hence Resχ5 = ψ4 + ψ5 . Thus we have a1 = −b1 , a2 = −2 − b2 , a3 = 1 − b3 and a4 = 1 − b4 .
Now we use the orthogonality relations for characters of A5 and get hψ4 , ψ1 i = 3 +
20a1 + 15a2 + 12a3 + 12a4 = 0, hψ4 , ψ2 i = 12 + 20a1 − 12a3 − 12a4 = 0, and hψ4 , ψ3 i =
15 − 20a1 + 15a2 = 0. Solving these equations we get a1 = 0, a2 = −1 and a3 + a4 = 1.
Hence we have the following:

|(gi )| 1 20 15 12 12
gi 1 (123) (12)(34) (12345) (13452)
ψ1 1 1 1 1 1
ψ2 4 1 0 −1 −1
ψ3 5 −1 1 0 0
ψ4 3 0 −1 a3 a4 = 1 − a3
ψ5 3 0 −1 b3 = 1 − a3 = a4 b4 = 1 − a4 = a3

Proposition 14.5. Every element of A5 is conjugate to its own inverse. Hence the
entries of the character table are real numbers.

Proof. Clearly, it is enough to prove that the representatives of the conjugacy classes
are conjugate to their own inverse. It is clear for the element 1, (123) and (12)(34). For
68 14. RESTRICTION OF A REPRESENTATION

others, we check that

(12345)−1 = (54321) = (15)(24)(12345)(15)(24)
(13452)−1 = (25431) = (12)(35)(13452)(12)(35)
Now we know that χ(g−1 ) = χ(g) and g being conjugate to g−1 this is also equal to χ(g).
Hence χ(g) = χ(g) gives χ(g) ∈ R for all g ∈ A5 . 
The above proposition implies that a3 , a4 , b3 and b4 are real numbers. From hχ4 , χ4 i = 1
we get a23 + a24 = 3. Substituting a4 = 1 − a3 we get a23 − a3 − 1 = 0. And the solutions are
√ √
1+ 5 1− 5
a3 = = b4 , a 4 = = b3
2 2
or √ √
1− 5 1+ 5
a3 = = b4 , a4 = = b3 .
2 2
Since the values of ψ4 and ψ5 on other conjugacy classes are the same, both the above
solutions would give the same set of irreducible characters. Hence without loss of generality,
we may take the first set of solutions. This gives the complete character table of A5 as
follows:
Character Table of A5
|(gi )| 1 20 15 12 12
gi 1 (123) (12)(34) (12345) (13452)
ψ1 1 1 1 1 1
ψ2 4 1 0 −1 −1
ψ3 5 −1 1 0√ 0√
1+ 5 1− 5
ψ4 3 0 −1 2√ 2√
1− 5 1+ 5
ψ5 3 0 −1 2 2

Exercise 14.6. Prove that every element of Sn is conjugate to its own inverse and
hence character table consists of real numbers.

Remark : In fact more is true that every element of Sn is conjugate to all those
powers of itself which generates the same subgroup (called rational conjugacy). Hence it is
true that characters of Sn always take value in integers. However, this is not true for An ,
for example, check A4 and A5 . In general, for An , they may not be even real-valued.
 CHAPTER 15

Central Characters

Let G be a finite group. Let ρ1 , . . . , ρh be all irreducible C-representations of G of

dimension n1 , · · · , nh respectively with corresponding characters χ1 , · · · , χh . We also fix
g1 , g2 , . . . , gh to be a representative of conjugacy classes of G with the respective conjugacy
classes size r1 , . . . , rh .
For any give representation ρ : G → X X
GL(V ) we can define an algebra homomorphism ρ̃ : C[G] → End(V ) by αg g 7→ αg ρ(g).
g g
We know that Z(C[G]), the center of C[G], is spanned by the elements cg1 , . . . , cgh where

X
cgi = t
t∈G
t=sgi s−1

i.e., the sum of all conjugates of gi . Here gi are representatives of the conjugacy classes.

Exercise 15.1. (1) Show that ρ̃ is an algebra homomorphism, that is, it is a vector
space homomorphism as well as ring homomorphism.
P
(2) Let g αg g ∈ Z(C[G]) then αg = αsgs−1 for any s ∈ G.
(3) Show that the center of Mn (k) (and End(V )) is the set of all scalar matrices (works
over any field k).

Exercise 15.2. Take ρ to be a 1-dimensional irreducible representation of a cyclic

group. Show that ρ̃ is not injective even if ρ is so.

We wish to understand how ρ̃ behaves on the center.

Proposition 15.3. Let ρ be an irreducible representation and z ∈ Z(C[G]). Then

ρ̃(z) = λ.I for some λ ∈ C (λ depends on z) and I is the identity transformation. Further,
λ = Tdim(ρ)
r(ρ̃(z))
.
69
70 15. CENTRAL CHARACTERS

Proof. We claim that ρ̃(z) ∈ End(V ) is a G-map and then use Schur’s Lemma. Let
P
z = g αg g then αg = αsgs−1 for any s ∈ G (as z is in the center). Then,
!
X X X
ρ̃(z)(ρ(t)v) = ρ̃ αg g (ρ(t)v) = αg ρ(g)(ρ(t)v) = ρ(t) αg ρ(t−1 gt)(v)
g g g
X
= ρ(t) αtut−1 ρ(u)v = ρ(t)ρ̃(z)v.
u

In the last step, we use αu = αtut−1 . Since ρ is irreducible, Corollary 4.5 (Schur’s Lemma)
implies that ρ̃(z) = λ.I for some λ ∈ C. 

With the notation as above let us consider algebra homomorphisms ρ̃i corresponding to
the irreducible representations ρi . In the light of above proposition let us denote ρ̃i (cgj ) =
λij Ini where λij ∈ C and Ini denotes the identity transformation. Further, we can take
traces on both sides and get,
X
ni λij = tr(ρ̃i (cgj )) = tr(ρi (t)) = rj χi (gj )
t∈G
t=sgj s−1

where rj is the number of conjugates of gj . This gives,

rj χi (gj ) χi (gj )
λij = = rj .
ni χi (1)
Thus, each irreducible representation ρi determines λi a function called central character
given the map λi : G → C defined on the conjugacy classes by λi (gj ) = λij .

Definition 15.4. To an irreducible representation ρ, the map λ : G → C defined on the

χ(g )
conjugacy classes by λ(gj ) = rj χ(1)j is said to be the central character associated to the
irreducible representation ρ. Clearly, it is a class function.

The next proposition says that the values of central characters are algebraic integers.

Proposition 15.5. Each λij is an algebraic integer.

Proof. Let us consider M = cg1 Z ⊕ · · · ⊕ cgh Z ⊂ Z(C[G]) = cg1 C ⊕ · · · ⊕ cgh C. Clearly,

M is a Z-submodule. It is a subring also (multiply cgj , cgk by writing them as elements in
C[G]). Take cgj , cgk ∈ Z(C[G]). Then cgj cgk ∈ Z(C[G]), in fact cgj cgk ∈ M . Hence we can
h
X
write cgj cgk = ajkl cgl where ajkl are integers. By applying ρ̃i we get,
l=1
h
X
(λij Ini )(λik Ini ) = ρ̃i (cgj )ρ̃i (cgk ) = ρ̃i (cgj cgk ) = ajkl λil Ini .
l=1
 15.1. ALGEBRAIC INTEGERS 71

P
This gives λij λik = hl=1 ajkl λil where each ajkl ∈ Z.
Now we take N = λi1 Z ⊕ · · · ⊕ λih Z ⊂ C which is a finitely generated Z-module and
λij N ⊂ N for all j. This implies λij is an algebraic integer. 

Lemma 15.6. For any s ∈ G and χ character of a representation, χ(s) is an algebraic

integer.

Proof. Let s ∈ G be of order d in G. Then ρ(s) is of the order less than or equal to
d. Since the field is C we can choose a basis such that the matrix of ρ(s), say A, becomes
diagonal (see 5.4 and also proof in 9.6). Clearly, Ad = 1 implies diagonal elements are the
root of the polynomial X d − 1 and hence are algebraic integer (being roots of unity). As a
sum of algebraic integers is again an algebraic integer we get a sum of diagonals of A which
is χ(s) is an algebraic integer. 
Now we can prove the main theorem of this chapter,

Theorem 15.7. The order of an irreducible representation divides the order of the
group, i.e., ni divides |G| for all i.

Proof. Let ρi be an irreducible representation of degree ni with character χi . From

1 P −1
the orthogonality relations we have, 1 = hχi , χi i = |G| t∈G χi (t)χi (t ). Thus,
h
X h
X
|G| = rj χi (gj )χi (gj−1 ) = ni λij χi (gj−1 ).
j=1 j=1
h
X |G|
After re-writing we get, λij χi (gj−1 ) = . The left side of this equation is an algebraic
ni
j=1
integer (using Proposition 15.5 and Lemma 15.6) and the right side is a rational number.
Hence |G|
ni is an algebraic integer as well as an algebraic number, which implies must be an
integer. This gives the required result, ni divides |G|. 

15.1. Algebraic Integers

Read about this topic.

 CHAPTER 16

Burnside’s pq Theorem

As an application of the character theory, we prove Burnside’s pq Theorem (see the

original article [Bu]). Let p 6= q be primes. We know that any group of order p is cyclic,
of order p2 is Abelian and pa is solvable. In this chapter, we prove that any group of order
pa q b is also solvable. In general groups of order divisible by three different primes need not
be solvable (think of A5 ).

Exercise 16.1. A group of order pr , for some r, is called a p-group. Let G be a p group
of order pr . Prove the following:
(1) The center of G is non-trivial.
(2) G has a subgroup of order ps for all 0 ≤ s ≤ r.
(3) Prove that for r = 2, G is always Abelian.
(4) Let G be acting on a finite set X. Show that

|X| ≡ |X G | (mod p)

where X G = {x ∈ X | gx = x∀g ∈ G}.

(5) Use the above to show that |Z(G)| ≡ 0 (mod p).
(6) Show that G is solvable.

We continue with the notation in the previous chapter and recall,

χ (g )
• the values of central character λij = rj ini j are algebraic integers.
• the values of characters χi (t) are algebraic integers.
Let us begin with a basic result.

Lemma 16.2. With the notation as above suppose rj and ni are relatively prime. Then,
either ρi (gj ) is in the center of ρi (G) or tr(ρi (gj )) = χi (gj ) = 0.

Proof. As in the proof of Proposition 9.6 and Lemma 15.6 we can choose a basis such
that the matrix of ρi (gj ) = diag{ω1 , · · · , ωni } and hence χi (gj ) = ω1 + · · · + ωni where ωk ’s
are d-th root of unity (here d is the order of gj ). Now, |ω1 + · · · + ωni | ≤ 1 + · · · + 1 = ni ,
χ (g )
hence | ini j | ≤ 1.
73
74 16. BURNSIDE’S pq THEOREM

χ (g )
In the case | ini j | = 1 we must have ω1 = ω2 = . . . = ωni . This implies that the matrix
of ρi (gj ) = diag{ω1 , · · · , ω1 } is central, and hence in this case ρi (gj ) belongs in the center
of ρi (G).
χ (g ) χ (g )
Now, suppose that | ini j | < 1. Let us denote α := ini j . We will show that α = 0.
Since rj and ni are relatively prime we can find integers l, m ∈ Z such that rj l + ni m = 1.
Then, λij = rj α gives lλij = (1 − ni m)α = α − mχi (gj ). Since λij and χi (gj ) both are
algebraic integers (see Proposition 15.5 and Lemma 15.6) we get α is an algebraic integer.
Now we use a little bit of Galois Theory (see Section 14.5 [DF]) of Cyclotomic extension.
Let ζ be a primitive d-th root of unity and let us consider the Galois extension K = Q(ζ)
of Q. Let ωk = ζ ak then we can write α = n1i (ω1 + · · · + ωni ) = n1i (ζ a1 + · · · + ζ ani ). For
σ ∈ Gal(K/Q) the element σ(α) is also of the same Y kind, and hence |σ(α)| ≤ 1. This
implies that the norm of α defined by N (α) = σ(α) has |N (α)| < 1. Since α is
σ∈Gal(K/Q)
an algebraic integer so are σ(α), and hence the product N (α) is an algebraic integer. Since
N (α) is also invariant under all σ ∈ Gal(K/Q) it is a rational number hence it must be an
integer. However as |N (α)| < 1 this gives N (α) = 0. This eventually gives the required
result χi (gj ) = ni α = 0. 

In what follows we may assume G is non-trivial and of order involving at least two
different primes.

Proposition 16.3. Let G be a finite group and C be a conjugacy class of g ∈ G.

If |C| = pr , where p is a prime and r ≥ 1, then there exists a non-trivial irreducible
representation ρ of G such that ρ(C) is contained in the centre of ρ(G). In particular, G is
not a simple group.

Proof. On contrary let us assume that ρi (C) is not contained in the center of ρi (G)
for all irreducible representations ρi of G. Then, from previous Lemma if (pr , ni ) = 1, i.e,
p ∤ ni , then we must have χi (g) = 0 for g ∈ C.
P
Consider the character of the regular representation χreg = hi=1 ni χi . Then for any
1 6= s ∈ G we have
h
X h
X
0 = χreg (s) = ni χi (s) = 1 + ni χi (s).
i=1 i=2
P
Let us take g ∈ C (note that g 6= 1 since |C| > 1). Then hi=2 ni χi (g) = −1 which we may
P
re-write after dividing by p as hi=2 npi χi (g) = − 1p . On the left hand side the term is either
0 (when p ∤ ni then χi (g) = 0) or an algebraic integer (when p | ni , npi is an integer). Thus,
the overall left side is an algebraic integer while the right-hand side is a rational number
 16.1. SOLVABLE GROUPS 75

hence it should be an integer which is a contradiction. Hence, there exists a non-trivial

irreducible representation ρi such that ρi (C) is contained in the center of ρi (G).
Now, let us take the above ρ : G → GL(V ) and we have ρ(C) ⊂ Z(ρ(G)) where C is
a conjugacy class of non-identity element. We want to show that G can not be simple. If
ker(ρ) 6= 1 it will be a normal subgroup of G and thus G is not simple. In case ker(ρ) = 1,
ρ is an injective map and Z(ρ(G)) 6= 1. But the centre is always a normal subgroup which
again implies G is not simple. 
With these results in hand we are ready to prove our main Theorem.

Theorem 16.4 (Burnside’s Theorem). Every group of order pa q b , where p, q are distinct
primes, is solvable.

Proof. We use induction on a + b. If a + b = 1 then G is a p-group and hence G is

solvable. Now assume a + b ≥ 2, and any group of order pr q s with r + s < a + b is solvable.
Let Q be a Sylow q-subgroup of G. If Q = {e} then b = 0 and G is a p-group and hence
solvable (see the Exercise 16.1). So let us assume Q is non-trivial. Since Q is a q-group
(prime power order) it has a non-trivial centre. Let 1 6= t ∈ Z(Q). Then
t ∈ Q ⊂ CG (t) ⊂ G
and hence |CG (t)| = pl q b for some 0 ≤ l ≤ a which gives [G : CG (t)] = pa−l .
First, we claim that G is not simple. If G = CG (t) then t ∈ Z(CG (t)) = Z(G),
i.e., Z(G) is a non-trivial normal subgroup and G is not simple. Hence we may assume
|CG (t)| = pl q b with l < a. Then using the formula we get the size of the conjugacy class
of t: |C(t)| = |C|G|
G (t)|
= pa−l . Here C(t) denotes the conjugacy class of t. Now, using the
previous proposition we get G is not simple.
Since we know that G is not simple, it has a proper normal subgroup, say N . The order
of N and G/N both satisfy the hypothesis hence they are solvable. Thus G is solvable (see
Proposition 10, Section 6.1 [DF]). 
The most amazing and celebrated result in this direction is Feit-Thompson Theorem
(see [FT]): Every finite group of odd order is solvable. Thus, non-Abelian finite simple
groups must be of even order.

16.1. Solvable Groups

Read about this topic.

 PART – III

One of the important techniques to produce representations is the method of induction.

77
 CHAPTER 17

Induced Representation

Let G be a finite group and H its subgroup. Let k be a field. Note that the various
definitions and constructions make sense over any field k. However, whenever we talk about
characters we consider only C-representations. Given a representation ρ : G → GL(V ) we
can restrict this representation to H and get ρ|H : H → GL(V ) where ρH (h) = ρ(h). This
is also written as ResG
H (ρ) or ρ ↓. The idea is to do the other way round, i.e., to begin with,
a representation of a subgroup H and get a “method” which gives a representation of G.
There are several ways to understand this. We will see some of them here.

17.1. Induced Representations from Subgroups

Let H be a subgroup of G, and let V be a G-representation. We can restrict the actin

of G to H and consider V as a H-representation. Let W be a H-invariant subspace of V .
Note that even though V is irreducible as a G representation it need not be irreducible as
a H representation. We consider gW ⊂ V for any g ∈ G. Notice that for any h ∈ H the
subspaces gW and ghW are the same. Hence, if at all, there are possible as Xmany distinct
subspaces gW as coset representatives g ∈ G/H. If we consider W = ′ gW then
g∈G/H
W′ ⊂ V is a G-subspace. Now, we are ready to define induced representation.

Definition
M17.1. The G-representation V is called induced from the H-subrepresentation
W if V = gW where sum varies over representatives of the left cosets of H. The
g∈G/H
induced module V is denoted by IndG
H W or Ind(W ) or W ↑.

Let us try to understand what V looks like if it is an induced module. Let s1 = e, s2 , . . . , sr

be a set of representatives of distinct cosets of G/H, i.e, G = H ∪ s2 H ∪ . . . ∪ sr H. Then,

Proposition 17.2. Let G be a group with H a subgroup. Let V be a G-representation

and W be an H-subrepresentation of V . Suppose V = IndG
H W . Then,

(1) si W = si hW for any h ∈ H.

T
(2) si W sj W = {0} when i 6= j.
79
80 17. INDUCED REPRESENTATION

(3) the G action on the set {si W | si ∈ G/H} permutes si W ’s (that is acts transitively
with stabiliser H).
(4) dim V = |G/H| dim W .
Here s1 = e, s2 , . . . , sr is a set of representatives of distinct cosets of G/H.

Proof. Since W is a H-representation,

M it is clear that si W = si hW for any h ∈ H.
Since V = IndG
H W = gW it follows that {x ∈ G | xW = W } = H. We can also
g∈G/H
define G action on {si W | 1 ≤ i ≤ r} by g.si W = (gsi )W . This is a transitive action and
the stabiliser of W is H.
Fix a basis of W , say, w1 , . . . wl . Then, a basis of V would be

w1 , . . . , wl , s2 w1 , . . . , s2 wl , . . . , sr w1 , . . . sr wl

where si w1 , . . . , si wl forms a basis of si W . Thus, dim V = |G/H| dim W . 

Exercise 17.3. With notation as in the proof above, show that

{w1 , . . . , wl , s2 w1 , . . . , s2 wl , . . . , sr w1 , . . . sr wl }

is a basis of V . In fact, si w1 , . . . , si wl forms a basis of si W .

Let us try to understand how G-representation action would be on V = Ind(W ) in

terms of the given H-representation W . What really happens is that the induced module
action is fully determined by two things:
• G action on G/H, and
• H action on W .
Since G × G/H → G/H given by g.xH = (gx)H is a permutation action, let us write this
as follows: g.si H = sσ(i) H where σ ∈ Sr depends on g. That is, gsi = sσ(i) hi for some
hi ∈ H. Now we can write the G-module action on the basis vectors as follows:

g.(si wj ) = sσ(i) (hi wj )

where si wj are basis vectors of V (as in the proof of the Proposition above). Let us work
out some examples.
P
Example 17.4. Consider the regular representation of G on C[G] = { αg eg | αg ∈ C}
given by x.eg = exg . Now consider the trivial subgroup H = {1} and W = {αe1 | α ∈ C}.
We claim that C[G] = IndG H W . This can be seen following the explanation above. The coset
representatives of G/H would be the elements of G with group multiplication as an action.
So, the basis of Ind(W ) would be {se1 = es | s ∈ G} and the representation would be
 17.2. CONSTRUCTION OF INDUCED REPRESENTATION 81

g.se1 = egs which is the regular representation. Thus, regular representation is an induced
representation, induced from a 1-dimensional representation of the trivial subgroup.

Example 17.5. Let G =!Z/4Z and V = Ce1 + Ce2 be the representation Z/4Z →
i
GL2 (C) given by 1̄ 7→ . Consider the subgroup H = h2̄i ∼
= Z/2Z and W =< e1 >.
−i
P
Since W is invariant under whole G, we see that g∈G/H gW = W . Thus, V is not induced
by W in this case.

Example 17.6. Once again let us consider the regular representation of G on C[G] =
P
{ αg eg | αg ∈ C} given by x.eg = exg . Now consider any subgroup H and W =
P
{ h∈H αh eh | αh ∈ C}. Clearly, W is a H subrepresentation. We claim that C[G] =
IndG H W . This can be seen as follows. The coset representatives of G/H, say are s1 , . . . , sr
and the basis vectors of W are {eh | h ∈ H}. Thus, basis vectors of IndG H W would
be {si eh | h ∈ H, i = 1, . . . , r} = {eg | g ∈ G}. Further, the representation would be
g.si eh = gesi h = egsi h which is the regular representation. Thus, regular representation is
P
induced representation, induced from H-representation W = { h∈H αh eh | αh ∈ C}.

Example 17.7. Let us consider the representation of the dihedral group G = Dm =

ha, b | am = 1 = b2 , ab = bam−1 i on V = Ce1 ⊕ Ce2 defined by:
" 2πi # " #
em 0 1
a 7→ 2πi , b 7→ .
e− m 1 0
Consider the representation of H = hai ∼ = Z/mZ on the subspace W =< e1 >. We claim
G
that IndH W = V . Let us verify this. We can take coset representatives of G/H to be
{1, b}. Then, the basis vectors for IndG H W would be {e1 , be1 = e2 }. Now let us determine
2πi 2πi
the action of G. Check that a.e1 = e e1 (this is given) and a.be1 = ba−1 e1 = e− m be1 =
m
2πi
e− m e2 . Similarly, be1 = e2 and b.be1 = 1.e1 = e1 . Note that in this case an irreducible
representation, after induction, gives an irreducible representation.

Exercise 17.8. Given H and W there is a unique G-representation IndG

HW .

17.2. Construction of Induced Representation

This formula suggests that we can always construct an induced G-module from a given
H module W . We don’t need to have a G-representation V a priori. Let us do this. Let
s1 , . . . , sr be representatives of G/H. Consider, V = W ⊕ W ⊕ · · · ⊕ W , r-copies. We have
G × G/H → G/H given by g.xH = (gx)H. Now, define the G-representation as follows:

g.wei = hi (w)ej
82 17. INDUCED REPRESENTATION

where gsi H = sj H gives gsi = sj hi for some hi ∈ H. That is, (0, . . . , 0, w, 0, . . . , 0) with w
at i-th place gets mapped to (0, . . . , 0, hi (w), 0, . . . , 0) with hi (w) at j-th place.

Exercise 17.9. Check that we have a G-representation with its character given by the
formula of induced character.

Example 17.10. Consider the group Q8 and H = hii ∼ = Z/4Z. Take the 1-dimensional
∗
representation θ : H → C given by i 7→ ι (here the first i is an element of Q8 , and the
complex one is denoted by ι to avoid confusion). What’s the induced representation IndQ 8
H θ?
Let the representation of H be W =< w >, i.e, i.w = ιw. Let us fix a representatives of
Q8 /H to be {1, j} and a basis of Ind(W ) to be {1w, jw}.
! Now, for i ∈ Q8 , i.1w = i.w = ιw
ι
and i.jw = −jiw = −jιw = −ιjw, hence i 7→ . Similarly, for j ∈ Q8 , j.1w = jw
−ι
!
−1
and j.jw = −w, hence j 7→ . Note that here we get an irreducible representation
1
after induction.

Example 17.11. Let us look at Example 17.5 again. Fix the coset representatives
of Z/4Z/h2i to be {0̄, 1̄}. Take the representation of h2̄i ∼
= Z/2Z/ to be h2̄i → C∗ given
by 2̄ 7→ −1. Suppose the representation space is W =< w >. Take a basis of Ind(W )
to be {w, 1̄w} which would be a 2-dimensional representation given by 1̄.w ! = 1̄w and
−1
1̄.1̄w = 2̄w = −w. Thus, the induced representation is given by 1̄ 7→ .
1
 CHAPTER 18

Character of the Induced Representation

As we have seen earlier that the induced character is determined by the G action on G/H
together with the H representation W . So, if we know the character of H-representation
W , say θ, we should be able to determine the character of the induced representation
V = Ind(W ), namely Ind(θ) (or IndG H θ).

Theorem 18.1. Let θ be the character of the H-representation W . Then, the character
χ = IndG
H θ is given as follows:
X 1 X
IndG
H θ(g) = χ(g) = θ(s−1
i gsi ) = θ(xgx−1 )
si |H| x∈G
s−1 gsi ∈H xgx−1 ∈H
i

where s1 = e, s2 , . . . , sr is a set of representatives of distinct cosets of G/H.

Proof. For g ∈ G, we need to compute χ(g) = tr(g) on V . For this, we make use
of V = s1 W ⊕ s2 W ⊕ · · · ⊕ sr W on which G acts by permutation. Let us consider the
basis {w1 , . . . , wl , s2 w1 , . . . , s2 wl , . . . , sr w1 , . . . sr wl }. More precisely, the action is given by
g.(si wj ) = sσ(i) (hi wj ) where gsi = sσ(i) hi for some hi ∈ H. Imagine that we have to write
the matrix of g with respect to this basis. When g.si W 6= si W it contributes 0 to the trace.
So, the only contribution for the trace of g will come when g.si W = si W . We do this more
precisely below.
When g.si W = sj W with i 6= j it won’t contribute to the trace of g. Thus, we need to
look at the components where g.si W = si W . In this case, gsi = si hi for some hi ∈ H, that
is, s−1
i gsi ∈ H. Thus,
X
χ(g) = tr(g) = tr(g|si W )
si
g.si W =si W
X X
= tr(s−1
i gsi ) = θ(s−1
i gsi ).
si si
s−1 gsi ∈H s−1 gsi ∈H
i i

The second equality follows as the size of each coset is |H|. 

83
84 18. CHARACTER OF THE INDUCED REPRESENTATION

Example 18.2. Consider H = {1} the trivial subgroup of G. The, character of the
trivial representation on H is θ : H → C given by θ(1) = 1. We can use the above formula to
1 X
compute the character IndG G
H θ as follows. First of all, IndH θ(1) = θ(x1x−1 ) =
|H| x∈G
x1x−1 ∈H
X 1 X
−1
θ(1) = |G|, and for g 6= 1, IndG
H θ(g) = θ(xgx ) = 0 as xgx−1 ∈ H if and
|H|
x∈G x∈G
xgx−1 ∈H

only if g = 1. Thus, IndG

H θ is the regular character on G.

Example 18.3. Let us consider the subgroup H = hai ∼ = Z/mZ of the dihedral group
m 2
G = Dm = ha, b | a = 1 = b , ab = ba m−1 i. Take the character θ : H → C given by
2πi
θ(a) = e . We can take coset representatives of G/H to be {1, b}. Then, IndG
m
H θ(1) =
1 X 1 X 2m X
−1 G −1
θ(x1x ) = θ(1) = = 2. Similarly, IndH θ(a) = θ(si asi ) =
|H| x∈G m m si
x∈G
x1x−1 ∈H s−1 asi ∈H
2πi X
i
−1 −1 − 2πi
θ(a) + θ(b ab) = θ(a) + θ(a ) = e m +e m , and IndG
H θ(b) = θ(s−1
i bsi ) = 0
si
s−1 bsi ∈H
i
/ H as well as b−1 bb ∈
as 1b1 ∈ / H. Note that in this case an irreducible character, after
induction, gives an irreducible character.

Given f ∈ C[H] we can extend f to f˙ ∈ C[G] as follows: f˙(g) = f (g) if g ∈ H and

f˙(g) = 0 otherwise. With this notation, we can introduce the following map
IndG
H C[H] → C[G]
:
X 1 X ˙
IndG
H f (g) = f˙(s−1
i gsi ) = f (xgx−1 ).
|H|
si ∈G/H x∈G

Thus, in the view of the above theorem, a character gets mapped to a character.

18.1. Induced Representation via Tensor Product

Let G be a group and H be a subgroup. Let k be a field. Suppose we are given W a

k[H]-module which amounts to having a representation of H. We would like to construct
a k[G]-module. Since H is a subgroup of G, we already have k[G] a k[H]-module (think of
ring/algebra-extension).

Definition 18.4. Now we define induced representation using the tensor product as
follows:
IndGH W := k[G] ⊗k[H] W
which is naturally a k[H]-module. We make it a k[G]-module as follows: for φ ∈ k[G] define
P P
φ.( i fi ⊗ wi ) := ( i φfi ) ⊗ wi .
 18.2. INDUCED REPRESENTATIONS VIA INVARIANT-FUNCTIONS 85

Exercise 18.5. Show that with the above definition IndG

H W is a k[G]-module.

Exercise 18.6. Show that k[G] is a free k[H]-module of rank |G/H|, and use this to
show that the two definitions of induced modules are the same. We can do this by choosing
coset representatives {s1 , . . . , sr } of G/H which ensures that every element of k[G] is of the
form β1 s1 + · · · βr sr where βj ∈ k[H]. Then,
X X X
(β1 s1 + · · · + βr sr ) ⊗ w = β1 s1 ⊗ w + · · · + βr sr ⊗ w = s1 ⊗ β1 w + · · · + sr ⊗ βr w.

Again, by fixing a basis of W , say, w1 , . . . , wl , we get a basis

{s1 ⊗ w1 , . . . , sr ⊗ w1 , . . . , s1 ⊗ wl , . . . , sr ⊗ wl }

of IndG
HW .

Thus, we see that this definition is the same as the Definition 17.1. This definition is
quite conceptual and makes it clear that the construction of an induced module can be
naturally done. Thus, IndG H is a functor which sends any k[H]-module W to the k[G]-
module k[G] ⊗k[H] W . This is quite a in sync with the IndG
H sending any class function on
H to a class function on G.

18.2. Induced Representations via Invariant-functions

There is yet another way to define induced representations via invariant maps. Each of
these definitions has the advantage of being generalised to different setups. The following
definitions are more often used when we are dealing with infinite groups, such as compact
groups, Lie groups etc. Once again we begin with a subgroup H of G and a representation
W of H. We take
V = {f : G → W | f (xh−1 ) = hf (x)}
the set of H invariant maps from G to W . We define G action on V as follows: (gf )(x) =
f (g −1 x). We first check that gf ∈ V . Notice that if we fix {s1 , . . . sr } as left coset repre-
sentatives for G/H then the maps in V are defined once they are specified on si ’s.

Exercise 18.7. (1) Prove that gf ∈ V . Note that (gf )(xh−1 ) = f (g−1 xh−1 ) =
−1
hf (g x) = h(gf )(x).
(2) f ∈ V is defined by its values on {s1 , . . . , sr }. Because, f (g) = f (si h) = h−1 f (si )
where g ∈ si H.
(3) Show that V is a vector space of dimension r. dim(W ).
(4) Prove that V ∼= IndGH W . For this, consider the functions fi ∈ V which map si H
to W and other elements of G to 0.
86 18. CHARACTER OF THE INDUCED REPRESENTATION

Warning: If we take V as left-H-invariant functions then we need to make G act on

the right. Be aware of which definition you are using.
Now we are going to see that the induced module satisfies a certain kind of universal
property:

Proposition 18.8. Let H be a subgroup of G and W be a H-module. Then, for any G

module M and any H-homomorphism φ : W → M there exists a unique G-homomorphism
φ̃ : IndG
H W → M such that the following diagram commutes:

W❅ / IndG W
❅❅ H
❅❅ ✇
❅ ✇
✇
φ ❅❅ ✇ φ̃
{✇
M
where W embeds in IndG G ∼ G
H W as 1.W component. Thus, HomH (W, ResH M ) = HomG (IndH , M )
as a k-vector space.
P P
Proof. Given φ we define φ̃ : k[G] ⊗ W → M as follows: φ̃( α ⊗ w) = αφ(w).
Check that φ̃ is a k[G]-module homomorphism. Now we know that any map on the induced
module can be determined by its value on W . 

Exercise 18.9. Prove that IndG

H W is a relatively free kH-module, i.e., prove the above
proposition.

Exercise 18.10. Show that induction is transitive. That is, if H is a subgroup of G

and K is a subgroup of H, i.e, K ⊂ H ⊂ G, then for any K-module W we have
IndG H G
H IndK W = IndK W.

Exercise 18.11. We have the permutation representation ρ of Sn on the set X =

{1, 2, . . . , n}. Consider Sn as a subgroup of Sn+1 by fixing the last symbol n + 1. Determine
S
the representation and character of IndSn+1 n
ρ.
 CHAPTER 19

Fröbenius Reciprocity

Let G be a group and H be a subgroup. We can define the restriction map ResGH : C[G] →
G
C[H] by simply restricting the function f : G → C to f |H =: ResH (f ).
We can also define the induction map IndGH : C[H] → C[G] as follows (see Theorem 18.1).
For any function f on H we define induced function IndG H f : G → C by
r
X 1 X ˙
IndG
H f (g) = f˙(si gs−1
i )= f (xgx−1 )
|H|
i=1 x∈G

where f˙ is an extension of f on G defined to be 0 outside H.

Exercise 19.1. (1) Prove that if f is a class function on H then IndGH f is a class
function on G. This is clear if we look at the second formula.
(2) Prove that if f is a character on H then IndG H f is so. We have proved this in the
G
last chapter that IndH f is a character of the induced representation.

Recall that for a group G we define the following inner product on the class functions:
1 X
hφ1 , φ2 iG = φ1 (t)φ2 (t).
|G|
t∈G

The Fröbenius Reciprocity theorem tells us that ResG G

H and IndH are adjoints of each other.
More precisely we have the following,

Theorem 19.2 (Fröbenius Reciprocity). Let G be a group and H be a subgroup. Let f

be a class function on H and ψ be a class function on G. Then,

hIndG G
H f, ψiG = hf, ResH ψiH .

Proof. The proof is computational using the formula.

1 X 1 XX ˙
hIndG
H f, ψiG = IndG
H f (t)ψ(t) = f (xtx−1 )ψ(t)
|G| |G||H|
t∈G t∈G x∈G
1 XX ˙
= f (xtx−1 )ψ(t).
|G||H|
x∈G t∈G
87
88 19. FRÖBENIUS RECIPROCITY

Now, note that f˙ is zero outside H, so, the non-zero terms in the inside sum would appear
only when xtx−1 ∈ H. That is, when xtx−1 = h or t = x−1 hx for some h ∈ H. Thus,

1 XX ˙ 1 XX ˙
hIndG
H f, ψiG = f (xtx−1 )ψ(t) = f (h)ψ(x−1 hx)
|G||H| |G||H|
x∈G t∈G x∈G h∈H
1 XX 1 X
= f (h)ψ(h) = hf, ResG G
H ψiH = hf, ResH ψiH .
|G||H| |G|
x∈G h∈H x∈G

This completes the proof. 

Let us also define hV1 , V2 iG := dimC HomG (V1 , V2 ).

Exercise 19.3. (1) Let χ1 and χ2 be characters of V1 and V2 respectively. Then

hχ1 , χ2 i = hV1 , V2 i.
(2) Prove that hW, ResG G
H V iH = hIndH W, V iG where W is a H-representation and V
is a G-representation.

An interesting question to study is if we can get all possible characters of a given group
G by induction of representations of proper subgroups. Let us define R(G) to be the free
Abelian subgroup generated by all irreducible characters of G, i.e., R(G) = Zχ1 + · · · + Zχh .
Notice that R(G) ⊗ C = H, the space of class functions. Thus, ResG H : R(G) → R(H) and
G
IndH : R(H) → R(G).

IndG
H
"
R(H) R(G)
b
ResG
H

Y
Now, we can take R(H) → R(G) and ask the question if it is a surjective map.
H<G
There are a couple of theorems due to Brauer and Artin (see [Se] Part 2) which answer
these questions. We illustrate this with an example.

Example 19.4. Let us consider the Kleins four group G = V4 = {e, a, b, c}. Consider
the subgroups H1 =< a >, H2 =< b >, H3 =< c >. Let χ1 , χ2 , χ3 be the non-trivial
characters of H1 , H2 and H3 respectively. Let 1 = φ1 , φ2 , φ3 and φ4 are the characters of
G.
 19. FRÖBENIUS RECIPROCITY 89

r1 = 1 r2 = 1 r3 = 1 r4 = 1
e a b c
1 = φ1 n1 = 1 1 1 1
φ2 n2 = 1 1 −1 −1
φ3 n3 = 1 −1 1 −1
φ4 n4 = 1 −1 −1 1

Then IndG G
H1 1 = 1 + φ2 and IndH1 χ1 = φ3 + φ4 and similarly others. Let us check this:
IndG −1 G −1
H1 1(g) = 1̇(g) + 1̇(b gb) = 21̇(g) = (1 + φ2 )(g) and IndH1 χ1 (g) = χ̇1 (g) + χ̇1 (b gb) =
2χ̇1 (g) = (φ3 + φ4 )(g). This defines a maps

3
Y
R(Hi ) → R(G)
i=1

of which image contains 1 + φ2 , 1 + φ3 , 1 + φ4 , φ3 + φ4 , φ2 + φ4 , φ2 + φ3 . We claim that the

image is of index 2. We check that 2φj belongs to the image and 1 doesn’t. However, the
image is the whole of it if we work over Q.

Example 19.5. In the case of the dihedral group Dm consider H = hai. Let θr , for
2πir
0 ≤ r ≤ m − 1, be the character of H given by a 7→ e− m . Then, for r 6= 0, IndG H θr is
an irreducible character of dimension 2 when θr (a) 6= ±1. Note that r and m − r give the
same character. Determine the image R(H) → R(G).

Exercise 19.6. Determine the induced representations from the Abelian subgroups for
G = S3 , S4 , Q8 and determine the subgroup in R(G).

Exercise 19.7. Consider the following character φ of S3 .

r1 = 1 r2 = 3 r3 = 2
g1 = 1 g2 = (12) g3 = (123)
φ 2 0 −1
90 19. FRÖBENIUS RECIPROCITY

Consider S3 as a subgroup of S4 by fixing the last symbol 4. Recall the character table of
S4 as follows:
r1 = 1 r2 = 6 r3 = 3 r4 = 8 r5 = 6
g1 = 1 g2 = (12) g3 = (12)(34) g4 = (123) g5 = (1234)
χ1 1 1 1 1 1
χ2 1 −1 1 1 −1
χ3 3 1 −1 0 −1
χ4 3 −1 −1 0 1
χ5 2 0 2 −1 0

(1) Use the Fröbenius reciprocity to show that hIndSS43 φ, χi i = hφ, ResSS43 χi i is 0 for
i = 1, 2 and is 1 for i = 3, 4, 5. Thus, conclude that IndSS43 φ = χ3 + χ4 + χ5 .
(2) Compute the character IndSS43 φ using the formula and show that

r1 = 1 r2 = 6 r3 = 3 r4 = 8 r5 = 6
g1 = 1 g2 = (12) g3 = (12)(34) g4 = (123) g5 = (1234)
S4
IndS3 φ 8 0 0 −1 0

Now, we can also directly write IndSS43 φ = χ3 + χ4 + χ5 decomposed as a sum of

irreducible characters of S4 .
 CHAPTER 20

Mackey’s Irreducibility Criteria

One of the purposes of defining induced representation is to be able to construct irre-

ducible representations of the given group G. One wonders if knowing irreducible repre-
sentations of various proper subgroups is of some help. Let H be a subgroup of G with a
representation (φ, W ) and character θ. Let IndG
H W be the induced representation of G with
the corresponding character IndH θ. To check if IndG
G
H θ is irreducible we need to compute

hIndG G G G
H θ, IndH θiG = hθ, ResH IndH θiH

and see if this is 1. Thus, in this process we need to understand ResG G

H IndH θ, more generally,
the decomposition of ResG G
H IndH W . For this, we take up a slightly more general situation
as follows.

20.1. Restriction of an Induced Representation

Let H and K be two subgroups of G. We are given a representation (φ, W ) of H. We

would like to understand the restriction of the G-representation IndG H W to the subgroup
G G
K, that is, ResK IndH W . To understand this we would require the concept of double
coset decomposition of G with respect to (K, H). That is, we define a relation ∼ on G
as follows: for g, g′ ∈ G we say g ∼ g′ if there exists k ∈ K and h ∈ H such that g ′ = kgh.
The equivalence classes are called (K, H)-double cosets of G. Let S = {s1 , . . . , sl } be a set
of representatives of (K, H)-double cosets of G.

Exercise 20.1. Show that ∼ is an equivalence relation on G. Further, G is a disjoint

union of KsH where s ∈ S.

Equivalently, we can write s ∈ K\G/H. Now, for any s ∈ S we denote the subgroups
Hs := sHs−1 ∩ K ≤ K.

Exercise 20.2. Consider the group( GL2!(q) and its Borel subgroup
) B consisting of all
a b
upper triangular matrices, i.e., B = | a, c ∈ F∗q , b ∈ Fq . Compute B\GL2 (q)/B
c
( !)
1
by showing that a set of representatives can be chosen to be I, .
1
91
92 20. MACKEY’S IRREDUCIBILITY CRITERIA

Exercise 20.3. Consider the group( SL2 (q)!and its Borel subgroup
) B consisting of all
a b
upper triangular matrices, i.e., B = | a ∈ F∗q , b ∈ Fq . Compute B\SL2 (q)/B
a−1
( !)
−1
by showing that a set of representatives can be chosen to be I, .
1

Exercise 20.4. Consider the Dihedral group Dn = ha, b | an = 1 = b2 , ab = ba−1 i and

the subgroup N = hai. Compute N \Dn /N .

Exercise 20.5. Consider the group GLn (q) and its Borel subgroup Bn consisting of all
upper triangular matrices. Compute Bn \GLn (q)/Bn ∼ = Sn by showing that the representa-
tives can be chosen to be the permutation matrices.

Now, since we are given a representation (φ, W ) of H we can consider representations

sφs−1 on sHs−1 . Let us set the following notation for this: (φs , Ws ) a representation of
Hs := sHs−1 ∩ K. Here we have Ws = W and φs (x) = φ(s−1 xs) for all x ∈ Hs . Just to
distinguish the action of sHs−1 from that of H we denote the representation space by Ws
here (though it’s the same space W ).

Theorem 20.6. The representation ResG G

K IndH W is decomposed as follows:
M
ResG G ∼ IndK
K IndH W = Hs W s
s∈K\G/H

where s varies over a set of representatives of the double cosets.

M
Proof. Denote V = IndG
HW = tW and recall G acts on {tW | t ∈ G/H} with
t∈G/H
stabiliser H. X X
Step 1: For s ∈ S, let us denote by V (s) = xW = ksW . We claim that
x∈KsH k∈K
(1) each M
V (s) is a K-representation (this is clear by second formula) and,
(2) V = V (s) (because G is a disjoint union of the (K, H) double cosets).
s∈S
M
Step 2: We claim that V (s) ∼
= IndK
Hs Ws . First, we note that V (s) = k(sW ).
k∈K/Hs
The sum part follows from the definition and for the direct sum, we note that {x ∈ K |
x(sW ) = sW } = Hs . More precisely, if k1 sW = k2 sW then s−1 k2−1 k1 sWM = W , that
−1 −1
is, s k2 k1 s ∈ H, thus, k1 ≡ k2 (mod Hs ). This proves that V (s) = k(sW ) =
k∈K/Hs
IndK
Hs (sW ).
 20.2. MACKEY-WIGNER IRREDUCIBILITY CRITERIA 93

Step 3: Now for the final step we show that Ws ∼ = sW as Hs -modules. This follows
from writing out the definition in both cases explicitly. For g ∈ Hs write g = shs−1 for
some h ∈ H, then g.Ws = s−1 gs(W ) = h(W ) and g.sW = shs−1 sW = sh(W ).
This completes the proof. 
Summary (steps to write the decomposition):
Given:
• H and K subgroups of G, and
• (φ, W ) a representation of H.
Want: to understand ResG G
K (IndH W ) in terms of given data?
Steps:
• Choose S a set of representatives of the double cosets K\G/H.
• Construct the subgroups Hs of K and its representation φs on W (denoted as Ws
to indicate a different action).
• Now consider the induced representations IndK Hs Ws which gives the final answer.

20.2. Mackey-Wigner Irreducibility Criteria

Now we can get back to our main problem where we can apply the Theorem 20.6 by
taking K = H. Let us recall the main problem. Let G be a group and H be a subgroup with
a representation (φ, W ) with associated character θ. We would like to know when IndG HW
is an irreducible representation of G, equivalently when the corresponding character IndG
Hθ
is an irreducible character.

Definition 20.7. Two representations (ρ1 , V1 ) and (ρ2 , V2 ) of G are said to be disjoint
if V1 and V2 do not have a common irreducible representation when they are written as a
sum of irreducible ones.

Proposition 20.8. The representations (ρ1 , V1 ) and (ρ2 , V2 ) of G are disjoint if their
corresponding characters are orthogonal.

Proof. The characters of ρ1 and ρ2 will not have any common irreducible character
when they are written as a sum of irreducible ones. Now the result follows using the
orthogonality relations. 

Theorem 20.9 (Mackey). Let H be a subgroup of G and (φ, W ) be a representation of

H. For s ∈ G, define the subgroups Hs = sHs−1 ∩ H ≤ H, and a representation (φs , Ws )
of Hs where Ws = W with the action φs : Hs → GL(W ) given by φs (x) = φ(s−1 xs). Then,
the representation IndG
H W is irreducible if and only if
94 20. MACKEY’S IRREDUCIBILITY CRITERIA

(1) W is an irreducible representation of H, and

(2) for all s ∈ G − H the representations (φs , Ws ) and ResH
Hs (φ, W ) of Hs are disjoint.

Proof. To prove this consider the corresponding characters. Let θ be the character of
φ. Then, IndG if hIndG
H W is irreducible if and only X
G G G
H θ, IndH θiG = hθ, ResH IndH θiH = 1.
Now using Theorem 20.6 ResG G
H IndH θ = IndHHs θs where θs is the character of φs .
s∈H\G/H
Thus, IndG
H W is irreducible if and only if
X
1 = hIndG G
H θ, IndH θiG = hθ, IndH
Hs θs iH
s∈H\G/H
X X
= hθ, IndH
Hs θs iH = hResH
Hs θ, θs iHs
s∈H\G/H s∈H\G/H

(we use reciprocity for H and Hs here). Now each term in the last sum is a non-negative
integer and for s = e we have the corresponding term hResH Hs θ, θs iHs = hθ, θiH ≥ 1. Thus,
IndGH W is irreducible if and only if hIndG θ, IndG θi = 1 if and only if hθ, θi
H H G H = 1 and
H
hResHs θ, θs iHs = 0. These are precisely the required conditions. 

A particular case of the above situation arises when we have a normal subgroup of G. In
this case, we can rephrase the above condition and eventually get what is called Wigner’s
little group method. This is applied to get irreducible representations of a semidirect
product.

Corollary 20.10. Suppose N is a normal subgroup of G and (φ, W ) is a representation

of N . For any s 6∈ N define representations of N by φs (x) = φ(s−1 xs). Then, IndG
N W is
irreducible if and only if
(1) φ is an irreducible representation of N , and,
(2) φ and φs are not isomorphic for any s ∈ G − N .

Now we summarise the above so that we can apply it in a practical situation.

20.2.1. Summary (steps to check irreducibility): What’s Given:

• H a subgroup of G, and
• (φ, W ) a representation of H.
Want: to understand when IndG
H W is an irreducible representation of G.
Steps:
• Check if φ is an irreducible representation of H. If yes proceed further.
• Choose S a set of representatives of the double cosets H\G/H.
 20.2. MACKEY-WIGNER IRREDUCIBILITY CRITERIA 95

• For all s ∈ S, construct the subgroups Hs of H and its representation φs on W

(denoted as Ws to indicate a different action). Note if H is normal all Hs = H but
φs could still be different. Also, for s = e the subgroup Hs = H and φs = φ.
• Check if (φs , Ws ) and ResH Hs (φ, W ) of Hs are disjoint for all s 6= e. If yes, the
induced representation is irreducible.

20.2.2. Summary (steps to check irreducibility when we have a normal sub-

group): Given:

• N a normal subgroup of G, and

• (φ, W ) a representation of N .

Want: to understand when IndG

N W is an irreducible representation of G.
Steps:

• Check if φ is an irreducible representation of N . If yes, proceed further.

• Choose S a set of representatives of the double cosets N \G/N .
• For all s ∈ S, consider the conjugate representations φs on W of N (denoted as
Ws to indicate a different action). That is, φs : N → GL(W ) is given by φs (x) =
φ(s−1 xs).
• Check if the representations φs and φ of N are not isomorphic for all s 6= e.
This can be done by computing their character. If yes, the induced representation
IndG
N W is irreducible.

Exercise 20.11. Consider the Dihedral group Dn = ha, b | an = 1 = b2 , ab = ba−1 i

with n > 2 and the subgroup N = hai.

(1) N is a normal subgroup of G.

(2) There are n 1-dimensional representations of N as N ∼ = Z/nZ.
(3) We can choose N \Dn /N = {e, b}.
(4) Given φ a 1-dimensional non-trivial representation (that is φ(a) 6= 1) of N we define
the representation φb of N by φb (x) = φ(b−1 xb). Check φb (a) = φ(bab) = φ(a−1 ).
Hence, φ and φb are not isomorphic if and only if φ 6= φ̄ (which is if and only if
φ(a) 6= ±1).
(5) Similarly, check the same for any φbai and note that φ and φbai are not isomorphic.
Although this is not really needed at the end.
(6) Show, using Mackey’s criteria, that the induced representation IndG N φ is an irre-
ducible representation where φ(a) 6= ±1 is a primitive n-th root of unity.
96 20. MACKEY’S IRREDUCIBILITY CRITERIA

Exercise 20.12. Consider the group SL2 (q) and the Borel subgroup
( ! )
a b
B= | a ∈ F∗q , b ∈ Fq .
a−1
( !)
−1
(1) We can choose B\SL2 (q)/B = I, s = .
1
(2) Fix a group homomorphism ω : F∗q → C∗ . Consider the 1-dimensional representa-
tion φ : B → C given by
!
a b
φ = ω(a).
a−1
Verify that it’s a representation. ( ! )
a
(3) Compute Bs = B ∩ sBs−1 = T = | a ∈ F∗q .
a−1
!
a
(4) Then, ResB B
Bs φ is given by ResBs φ = ω(a).
a−1
! !
a a−1
(5) Now, φs : Bs → C is given by φs (x) = φ(s−1 xs). Check, φs =φ =
a−1 a
ω(a−1 ).
(6) Thus, φ is irreducible if and only if ResB 2
Bs φ 6= φs if and only if ω 6= 1.
 CHAPTER 21

Representations of a Semidirect Product - Wigner’s Little

Group Method

Let G be a group with A a normal Abelian subgroup. Suppose G = A ⋊ H for some

subgroup H. Hence we have A ∩ H = 1 and G = AH. Now, all irreducible representations
of A are 1-dimensional. Let  be the set of all characters (equivalently 1-dim irreducible
representations) of A. Define an action of G on  as follows: Let χ ∈ Â. Now, for a given
g ∈ G we define χg : A → C∗ by χg (x) = χ(g −1 xg) which is again a character.

Exercise 21.1. Verify that G × Â → Â given by (g, χ) 7→ χg is an action.

This action can be restricted to H and thus H acts on the set of characters of A as
follows: H × Â → Â given by h.χ 7→ χh . Let us denote the stabiliser of χ by Hχ = {h ∈
H | χh = χ}. This allows us to construct a subgroup of G,
Gχ = A.Hχ ∼
= A ⋊ Hχ
called the little groups. Now, the character χ ∈ Â can be trivially extended to Gχ as
follows: χ : Gχ → C∗ given by χ(a.h) = χ(a) for all a ∈ A and h ∈ Hχ . Thus, χ is a
character of Gχ of dimension 1.

Exercise 21.2. Understand all of the above for the dihedral group Z/nZ ⋊ Z/2Z.

Now, let ρ be an irreducible representation of Hχ . Again, using the projection map

Gχ → Hχ given by a.h 7→ h we can get an irreducible representation ρ of Gχ . Let us denote
the representation θχ,ρ := IndG
Gχ χ ⊗ ρ of G.

Theorem 21.3. Let G ∼ = A⋊ H where A is a normal Abelian subgroup of G. For χ ∈ Â

and ρ an irreducible representation of H, extend them to the “little subgroups” Gχ . Now,
define θχ,ρ = IndG
Gχ χ ⊗ ρ representations of G. Then,
(1) θχ,ρ is irreducible.
(2) If θχ,ρ and θχ′ ,ρ′ are isomorphic then χ, χ′ are in the same orbit of Â/H and ρ ∼
= ρ′ .
(3) Every irreducible representation of G is isomorphic to one of the θχ,ρ .

We leave proof of this to the reader.

97
98 21. REPRESENTATIONS OF A SEMIDIRECT PRODUCT - WIGNER’S LITTLE GROUP METHOD
( !)
a b
Exercise 21.4. Consider the affine group A(q) = ⊂ GL2 (q). Write down
0 1
all irreducible representations of A(q).
( !)
a b
Exercise 21.5. Consider the Borel subgroup B = ⊂ GL2 (q). Write down
0 d
all irreducible representations of B.
( !)
a b
Exercise 21.6. Consider the group ZU = ⊂ GL2 (q). Write down all
0 a
irreducible representations of ZU .

21.1. Representations of groups of order pq and p3

In this section, p, q are primes and p 6= q. We know any group of order p is cyclic and
any group of order p2 is Abelian. Thus, all irreducible representations of such groups are
1-dimensional. The representation theory of such groups can be done easily following the
theory in Chapter 5. Thus, we will deal with non-Abelian groups here.

21.1.1. Groups of order pq. Let G be a group of order pq. We may assume q < p.
When q ∤ (p − 1) it can be shown that the group is Abelian, in fact, cyclic (See Example on
page 135, 143 and 181 [DF]).
Thus, we deal with the case when q | (p−1) and G are non-Abelian. Denote by r = p−1 q .
In this case, there is a unique non-Abelian group, up to isomorphism, of order pq. Such
groups can be also realised as matrix subgroup in the following way.

Example 21.7. Consider the finite field Fp and the group

( ! )
1 x ∗
G= | x ∈ Fp , y ∈ Fp .
y

(1) Show that G is a group of order p(p − 1). ! !

1 1 1
(2) Fix ζ to be a generator of the cyclic group F∗p . Take X = and Y = .
ζ 1
What are the orders of X and Y ? Show that X and Y generate G?
(3) Take A = X r and B = Y . Then show that G =< A, B > is a nonabelian subgroup
of G of order pq.
(4) Show that the number of conjugacy classes of G is q + r.
(5) Show that the commutator subgroup [G, G] =< B > is of order p. Hence, the
number of 1-dimensional representations of G is pq
p = q.
 21.1. REPRESENTATIONS OF GROUPS OF ORDER pq AND p3 99

(6) Use the det : G → F∗p and show that image size is q. Thus, we can lift 1-dimensional
representations to get q 1-dimensional representations of G.

Example 21.8. Consider the group of order 21, G = Z/3Z ⋉ Z/7Z.

We note that G = Z/qZ ⋉ Z/pZ which can be explicitly written as follows:

G = ha, b | aq = 1 = bp , ab = bal i
where l is an element of order q in F∗p . In this case, we will see that the remaining r irreducible
representations are each of dimension q. These are obtained as induced representations of
1-dimensional representations of the subgroup < B >.

Exercise 21.9. Complete the last step.

Theorem 21.10. Let G be a non-Abelian group of order pq where p, q are prime and
q | (p − 1). Then, G has q irreducible representations of dimension 1 and p−1
q irreducible
representations of dimension q.

21.1.2. Groups of order p3 . We deal with the non-Abelian groups of order p3 . When
p = 2 its either D4 or Q8 and we have already seen the character table of these groups.
So, we may assume p is odd. Once again, up to isomorphism, there are 2 such groups. We
recommend well-written online notes by Conrad [Co] or [DF] for a concrete description of
these groups.
We only require the following structural description:

Proposition 21.11. Let G be a non-Abelian group of order p3 with center Z. Then,

|Z| = p, [G, G] = Z, and G/Z ∼
= Z/pZ × Z/pZ.

Thus, such groups have p2 representations of dimension 1 obtained as a lift of G/Z.

The remaining ones are p − 1 of them and each of degree p.

Exercise 21.12. Complete the last step. These representations are obtained as induced
representations of an Abelian normal subgroup N =< a, Z > of order p2 where a ∈ G − Z.

Note that the character table of both non-isomorphic groups would be the same as we
have noted similar phenomena for groups of order 8.

Exercise 21.13. Collect the examples of groups where only two distinct character
degrees are possible. In fact, the classification of such groups is known which you may
explore further.
 CHAPTER 22

Characters of the group GL2 (q)

All representations considered here are over C. We consider the finite field Fq , with
q ≥ 5 odd (this is simply for the convenience of some calculations involved). The group
GL2 (q) = {X ∈ M2 (Fq ) | det(X) 6= 0}
and
SL2 (q) = {X ∈ M2 (Fq ) | det(X) = 1} .

Exercise 22.1. Let us warm up with some basic calculations.

(1) Show that |GL2 (q)| = (q 2 − 1)(q 2 − q) and |SL2 (q)| = (q 2 − 1)(q − 1).
(2) Show that det : GL2 (q) → F∗q is a surjective group homomorphism with kernel the
group SL2 (q).
(3) Show that the center Z(GL2 (q)) = {λ.I | λ ∈ F∗q }, the scalar matrices.
! !
1 t 1 0
Now define x12 (t) := and x21 (s) := for t, s ∈ k. The matrices x12 (t)
0 1 s 1
and x21 (s) are called the elementary matrices.

Exercise 22.2. Let us show that the elementary matrices generate the group SL2 (q).
(1) Compute w(t) = x12 (t)x21 (−t−1 )x12 (t). What is w(−1)? Compute h(t) = w(t)w(−1).
(2) For c 6= 0, verify the following for an element in SL2 (q),
! ! ! !
a b 1 (a − 1)c−1 1 1 (d − 1)c−1
= .
c d 1 c 1 1
(3) Show that the group SL2 (q) is generated by the set of all elementary matrices
{x12 (t), x21 (s) | t, s ∈ Fq }.

Exercise 22.3. Let us compute the 1-dimensional representations.

(1) Show that [SL2 (q), SL2 (q)] = 1 and hence the only 1 dimensional representation of
SL2 (q) is the trivial one.
(2) Show that [GL2 (q), GL2 (q)] = SL2 (q) and hence GL2 (q) has q − 1 representations
of dimension 1.
101
102 22. CHARACTERS OF THE GROUP GL2 (q)

(3) Since F∗q ∼

= Z/(q − 1)Z (why?) we have q − 1 representations of dimension 1. This
would give q − 1 representations of GL2 (q) of dimension 1.

22.1. Conjugacy classes in GL2 (q)

Let us try to determine the number of conjugacy classes in this group. The conjugacy
classes in GL2 (q) are as follows. This can be determined by rational canonical form theory.
Let M(x) and C(x) be the minimal and characteristic polynomial respectively. Clearly,
C(x) could be any degree 2 polynomial with non-zero constant term and M(x) could be any
polynomial of degree ≤ 2.
Central type: When C(x) = (x − λ)2 and M(x) = (x − λ), this gives the conjugacy
classes of central matrices {λI} for λ ∈ F∗q . There are total q − 1 such conjugacy classes
each of size 1.
Split semisimple type: When C(x) = (x −!a)(x − b) = M(x). This corresponds to
a
the conjugacy classes of diagonal matrices where a, b ∈ F∗q and a 6= b (so as they
b
are excluded from the central type). There are total (q−1)(q−2)
2 such classes each of size
q(q + 1). Note that the centralizer of such elements is all of the diagonal matrices hence of
size (q − 1)2 .
Anisotropic type: When C(x) = x2 +ax+b = M(x) is irreducible. These elements are
those of which minimal polynomial and characteristic polynomials are ! equal and irreducible
0 −b
of degree 2 and a representative is the companion matrix . These correspond to
1 −a
picking an element in Fq2 which is not in Fq and two of these will give the same class. Thus,
2
there are q 2−q such classes. The centralizer of such an element is polynomials in the same
matrix and hence ∼ = F∗q2 of size q 2 − 1. Thus, the conjugacy class size is q 2 − q.
Unipotent type: When C(x) = (x − λ)2 and 2
! M(x) = (x − λ) . This corresponds to
λ 1
the conjugacy classes of Jordan matrices where λ ∈ F∗q and the centraliser is of
λ
size q(q − 1) and hence the conjugacy class size is q 2 − 1. There are q − 1 such types.
Thus we get,

Theorem 22.4. The total number of conjugacy classes in GL2 (q) is

(q − 1)(q − 2) q 2 − q q−2 q
(q − 1) + + + q − 1 = (q − 1)(2 + + ) = (q − 1)(q + 1).
2 2 2 2
So, far we have found q − 1 characters each of dimension 1. We need to find more
irreducible characters. Let us set some notation here. We have det : GL2 (q) → F∗q given by
 22.3. THE PRINCIPAL SERIES REPRESENTATIONS 103

A 7→ det(A). Now, if ψ : F∗q → C∗ is a representation then lifting this to GL2 (q) we get,
ψ ◦ det.

Proposition 22.5. The group GL2 (q) has q − 1 representations of dimension 1. These
are given by ψ ◦ det where ψ : F∗q → C∗ is a group homomorphism.

22.2. Parabolic Induction - preparation

Now we consider certain representations of a parabolic subgroup and induce these rep-
resentations in the whole group. In our case, the parabolic subgroup we consider is simply
the Borel subgroup of upper triangular matrices. In this process, we get several irreducible
representations. ( !)
a b
Let us consider the subgroup B = ⊂ GL2 (q), namely the Borel subgroup.
0 d
Fix group homomorphisms ψ, φ : F∗q → C∗ . These are 1-dimensional representations of the
cyclic group F∗q . This can be done by fixing a primitive (q − 1)th root of unity, say ζ. Then
all characters of F∗q (being a cyclic group of order q −1) will be given by mapping a generator
of F∗q to ζ, ζ 2 , . . . , ζ q−1 = 1. Now, using ψ, φ we define a representation ρψ,φ : B → C∗ given
by !
a b
ρψ,φ = ψ(a)φ(d).
0 d

Exercise 22.6. Show that ρψ,φ is a representations of B. That is, show that these are
group homomorphisms.

Exercise 22.7. Further, show that these are all possible 1-dimensional representations
of B.

Now we deal with the two cases ψ 6= φ and ψ = φ separately. We induce the representation
ρψ,φ of B to GL2 (q) and analyse them further. In the first case when ψ 6= φ the induced
representations turn out to be irreducible and are called the principal series representations.
In the second case when ψ = φ, the induced representations are not irreducible and these
turn out to be a direct sum of a 1-dimensional representation what is called the Steinberg
representations.

22.3. The principal series representations

We continue the notation from the previous section. Now define the induced repre-
GL (q)
sentations ρ̂ψ,φ = IndB 2 ρψ,φ . We want to determine when these representations are
irreducible.
104 22. CHARACTERS OF THE GROUP GL2 (q)

Exercise 22.8. The dimension of the representation ρ̂ψ,φ is q + 1. This follows by a

straightforward computation (see Proposition 17.2).

Now we use the algorithm 20.2.1 to check the irreducibility of these induced representations.
Before that, we need to determine B\GL2 (q)/B.
( !)
1
Exercise 22.9. Show that we can choose I, s = as a set of representatives
1
for B\GL2 (q)/B. This is equivalent to applying row-column operations on a matrix where
we are not allowed row-column flipping. Thus, in the end, we will be left with one of these
kinds of matrices.

Now compute Bs as follows:

( ! )
a
Exercise 22.10. Bs = B ∩ sBs−1 = T = | a, d ∈ F∗q . This subgroup T is
d
called a split maximal torus.

Exercise 22.11. Since T = F∗q × F∗q the irreducible representations are obtained by
taking a product of the individual components which themselves are cyclic. We see that
these representations are given by {ψ.φ} where ψ, φ are characters of F∗q and the map
(ψ.φ)(a, b) = ψ(a)φ(b).

Exercise 22.12. Then, ResB

Bs ρψ,φ is a map on Bs = T given by
!
B a
(ResBs ρψ,φ ) = ψ(a)φ(d).
d

Let us compute the conjugate maps and note the subtle difference.

Exercise 22.13. Now, (ρψ,φ )s : Bs → C is a map on Bs = T given by (ρψ,φ )s (x) =

ρψ,φ (s−1 xs). Check,
! !
a d
(ρψ,φ )s = ρψ,φ = ψ(d)φ(a) = ResB
Bs ρφ,ψ .
d a

Now we are ready to prove the following:

GL2 (q)
Proposition 22.14. When ψ 6= φ the representation ρ̂ψ,φ = IndB ρψ,φ is irreducible
of dimension q + 1.
GL2 (q)
Proof. From Mackey’s criteria Theorem 20.9, the representation ρ̂ψ,φ = IndB ρψ,φ
is irreducible if and only if
 22.3. THE PRINCIPAL SERIES REPRESENTATIONS 105

(1) ρψ,φ is irreducible (which it is by being 1-dimensional) representation of B.

(2) ResBBs ρψ,φ and (ρψ,φ )s are disjoint (for this we need to check if these are equal or
not as they are 1-dimensional) representations of Bs = T.
GL (q)
Hence, we get ρ̂ψ,φ = IndB 2 ρψ,φ is irreducible if and only if ResB
Bs ρψ,φ 6= (ρψ,φ )s as
maps. Since we are given φ 6= ψ we get the result. 

However, we need to also understand if we get distinct irreducible representations. These

are certainly different from the earlier ones obtained so far which were 1-dimensional as these
are (q + 1)-dimensional.

Proposition 22.15. When ψ1 6= φ1 and ψ2 6= φ2 , the representations ρ̂ψ1 ,φ1 ∼ = ρ̂ψ2 ,φ2
(q−1)(q−2)
if and only if (ψ1 , φ1 ) = (ψ2 , φ2 ) or (ψ1 , φ1 ) = (φ2 , ψ2 ). Thus, there are total 2
distinct such irreducible representations.

Proof. In view of Proposition 18.8 we need to determine,

GL2 (q) GL2 (q)

HomGL2 (ρ̂ψ1 ,φ1 , ρ̂ψ2 ,φ2 ) = HomGL2 (IndB ρψ1 ,φ1 , IndB ρψ2 ,φ2 )
GL2 (q) GL2 (q)
= HomB (ρψ1 ,φ1 , ResB IndB ρψ2 ,φ2 ) Using reciprocity
= HomB (ρψ1 ,φ1 , ρψ2 ,φ2 ⊕ IndB
Bs (ρψ2 ,φ2 )s )

= HomB (ρψ1 ,φ1 , ρψ2 ,φ2 ) ⊕ HomB (ρψ1 ,φ1 , IndB

Bs (ρψ2 ,φ2 )s )

= HomB (ρψ1 ,φ1 , ρψ2 ,φ2 ) ⊕ HomBs (ResB

Bs ρψ1 ,φ1 , (ρψ2 ,φ2 )s )

where we use reciprocity again at the last step. The first term on the right here is either
1 or 0 dimension depending on ρψ1 ,φ1 is isomorphic to ρψ2 ,φ2 or not as both are irreducible
representations. Thus, dim HomB (ρψ1 ,φ1 , ρψ2 ,φ2 ) = 1 if ψ1 = ψ2 and φ1 = φ2 and 0 other-
wise.
The second term has representations of Bs = T given by diag(a, d) 7→ ψ1 (a)φ1 (d) and
diag(a, d) 7→ ψ2 (d)φ2 (a). These two representations are equal if and only if ψ1 = φ2 and
ψ2 = φ1 . Thus, dim HomBs (ResB Bs ρψ1 ,φ1 , (ρψ2 ,φ2 )s = 1 if ψ1 = φ2 and φ1 = ψ2 , and 0
otherwise. 

(q−1)(q−2)
Thus so far we get (q − 1) irreducible representations of dimension 1 and 2
irreducible representations, namely the principal series, of dimension q + 1.
106 22. CHARACTERS OF THE GROUP GL2 (q)

22.4. The Steinberg Representations

We continue with the setup in the previous section and discuss what happens when
ψ = φ. We have, ρψ := ρψ,ψ : B → C∗ given by

!
a b
ρψ = ψ(ad).
0 d

GL (q)
We induce this representation and get ρ̂ψ = IndB 2 ρψ which is a representations of
dimension (1 + q). We want to determine if these representations give rise to irreducible
ones. It turns out that these are not irreducible but can be easily broken to get some
new irreducible representations, called the Steinberg representations. We can use Mackey’s
GL (q)
criteria as in the proof or Proposition 22.14 and show that ρ̂ψ = IndB 2 ρψ are not
irreducible as ResB
Bs ρψ , (ρψ )s both are same and given by

!
a
7→ ψ(ad).
d

Exercise 22.16. Show that, (ρψ )s : Bs = T → C is given by (ρψ )s (x) = ρψ (s−1 xs).
Check,
! !
a d
(ρψ )s = ρψ = ψ(ad) = ResB
B s ρψ .
d a

Exercise 22.17. We may also note that if ψ is a trivial character then so is ρψ , i.e,
ρψ (A) = 1 for all A ∈ B. Does the induced representation ρ̂ψ sound familiar in this case?

GL (q)
Now, let us try to understand the decomposition of ρ̂ψ = IndB 2 ρψ . First, we determine
how many irreducible representations are involved in ρ̂ψ with the help of Proposition 18.8.

Proposition 22.18. ρ̂ψ is a sum of exactly two irreducible representations.

GL (q)
Proof. Suppose, ρ̂ψ = IndB 2 ρψ = m1 ρ1 + · · · + mh ρh written as a sum of irre-
ducible representations. We need to determine mi s. From the Maschke’s Theorem if a
L mi
module V is written as a direct sum of irreducibles V = Wi then HomC[G] (W, W ) =
L L P 2
Mmi (EndC[G] (Wi )) = Mmi (C). Thus, dim HomC[G] (W, W ) = mi can help us de-
termine mi s.
 22.4. THE STEINBERG REPRESENTATIONS 107

In view of this, we need to compute the dim HomGL2 (ρ̂ψ , ρ̂ψ ).

GL2 (q) GL2 (q)
HomGL2 (ρ̂ψ , ρ̂ψ ) = HomGL2 (IndB ρψ , IndB ρψ )
GL (q) GL (q)
= HomB (ρψ , ResB 2 IndB 2 ρψ ) Using reciprocity
= HomB (ρψ , ρψ ⊕ IndB
Bs (ρψ )s )

= HomB (ρψ , ρψ ) ⊕ HomB (ρψ , IndB

Bs (ρψ )s )

= HomB (ρψ , ρψ ) ⊕ HomBs (ResB

Bs ρψ , (ρψ )s ) Using reciprocity

= HomB (ρψ , ρψ ) ⊕ HomBs (ResB B

Bs ρψ , ResBs ρψ ) Use the exercise above

= C ⊕ C.

Note that at the last step we have all 1-dimensional representations (hence are irreducible)
of B and Bs = T respectively.
Thus, dim HomGL2 (ρ̂ψ , ρ̂ψ ) = 2 and using the equation m21 + · · · + m2h = 2, combined
with the fact it is not irreducible, we conclude that 12 + 12 = 2 is the only solution. 

GL (q)
Now, since ρ̂ψ = IndB 2 ρψ is a sum of two irreducible representations we need to de-
termine them. Fortunately, in the present case, we can easily show that it has 1-dimensional
representations contained in it and thus the complement would be the required irreducible
one.
GL2 (q)
Proposition 22.19. The representation ρ̂ψ = IndB ρψ contains a copy of the 1-
dimensional representation ψ ◦ det.

Proof. For this, we need to analyse the representation space. Recall, we begin with
GL2 (q)
the one dimensional B-space W and ! get V = IndB W . If W =< e > then the action of
a b
B is given by ρψ . That is, .e = ψ(ad)e, in another words A.e = ψ(det(A))e for an
0 d
A ∈ B. Now, the induced module V = {f : GL2 (q) → W | f (xA−1 ) = A.f (x), A ∈ B} and
GL2 (q)-action is given by (gf )(x) = f (g−1 x). We also remark (which will be immediately
useful) that while writing an element g ∈ GL2 (q) as g = xA−1 with A−1 ∈ B we can take
x to be of det(x) = 1.
We claim that the function f0 : GL2 (q) → W defined by f0 (g) = ψ(det(g −1 ))e has the
following property:
(1) f0 is in V .
(2) < f0 >⊂ V is a GL2 (q)-invariant subspace of V .
(3) The action of GL2 (q) on f0 is multiplication by ψ ◦ det.
108 22. CHARACTERS OF THE GROUP GL2 (q)

To check the first one,

f0 (xA−1 ) = ψ(det(Ax−1 ))e = ψ(det(A) det(x−1 ))e = ψ(det(A))ψ(det(x−1 ))e = A.f0 (x).
Now for the remaining ones we note,
(g′ f0 )(x) = f0 (g′−1 x) = ψ(det(x−1 g′ ))e = ψ(det(g ′ ))ψ(det(x−1 ))e = ψ(det(g′ ))f0 (x).

GL (q) L
Thus, ρ̂ψ = IndB 2 ρψ ∼ = (ψ ◦ det) Stψ where Stψ is an irreducible representation
of GL2 (q) of dimension (q + 1) − 1 = q, called the Steinberg representations.

Proposition 22.20. For every ψ, a character of F∗q , we have an irreducible represen-

tation Stψ of dimension q. There are (q − 1) such irreducible characters of GL2 (q).

22.5. Cuspidal representations

So far, we have found (q − 1) irreducible representations of dimension 1, (q−1)(q−2)

2
irreducible representations of dimension (q + 1) and (q − 1) irreducible representations of
dimension q. Thus, so far we have found
(q − 1)(q − 2)
(q − 1) + + (q − 1)
2
irreducible characters of GL2 (q). Since the number of irreducible characters is the same as
2
the number of conjugacy classes, it remains to find q 2−q irreducible characters. We can also
try to determine their degrees:
q 2 −q
X  
2
(q − 1)(q − 2) q(q − 1)3
n2i 2
= |G| − (q − 1).1 + (q + 1)2 + (q − 1).q 2 = .
2 2
i=1

A possible solution of this equation with ni | |GL2 (q)| is that each ni = q − 1. This is indeed
the case. The remaining representations are said to be cuspidal representations. We
will discuss that here briefly without going into too much in detail. Once again we look at
certain subgroups of which representations
( !we can induce. )
a b
Consider the subgroup ZU = | a ∈ F∗q , b ∈ Fq .
a
!
1 1
Exercise 22.21. Compute the centralizer of u = and show that it is ZU . Note
1
that this (
group is!actually the
) product (not a direct product) of the centre Z = Z(GL2 (q))
1 b
and U = | b ∈ Fq .
1
 22.6. CHARACTERS OF GL2 (q) 109

Now given a character ψ of F∗q (there are q − 1 such) and φ of Fq (there are q such) we
define a representation ✵ψ,φ : ZU → C as follows:
!
a b
✵ψ,φ = ψ(a)φ(b).
a

Exercise 22.22. These are all possible 1-dimensional representations of ZU .

Now, we define ✵ ˜ ψ,φ = IndGL2 (q) ✵ψ,φ which are of dimension q 2 − 1. As it turns out
ZU
that these are not irreducible. We work with certain ψ and show that it can be decomposed
as a direct sum of two irreducible representations
GL (q) GL2 (q)
IndZU2 ✵ψ,φ = IndT ψ ⊕ πψ
Let us begin with understanding the anisotropic maximal torus T . The field Fq2 can be
embedded in M2 (q) giving rise to T ⊂ GL2 (q).

Exercise 22.23. Consider Fq2 over Fq as 2-dimensional vector space. For any, θ ∈ Fq2
consider the left multiplication map on Fq2 giving rise to a linear map. This embeds Fq2 in
M2 (q).

Now F∗q2 is cyclic group of order q 2 − 1 thus has as many 1-dimensional representations.
Consider the non-trivial Galois automorphism σ : Fq2 → Fq2 given by x 7→ xq . This fixed
field is Fq . Thus, the maps ψ : F∗q2 → C∗ also include the ones for F∗q . We say ψ is a regular
character of T = F∗q2 if ψσ 6= ψ. Equivalently, ψ q 6= ψ or ψ q−1 6= 1. Thus, these are ψ
which are not of order q − 1, hence total such number would be (q 2 − 1) − (q − 1) = q 2 − q.
Also, not that such ψ would come in pairs {ψ, ψ ◦ σ}.
GL (q) GL (q)
Proposition 22.24. For a regular character ψ of T , we have IndZU2 ✵ψ,φ = IndT 2 ψ⊕
πψ where πpsi is an irreducible representation of dimension q − 1. The πψ ∼
= πψ′ if and only
′ (q 2 −q)
if ψ = ψ ◦ σ. Thus, these are the remaining 2 irreducible cuspidal representations.

This section is not complete and a bit of effort is required to complete all the steps.
On other hand knowing the irreducible representations, as we do by now, one can simply
compute the characters to verify that we get all of them.

22.6. Characters of GL2 (q)

Given the description above write the character table.

 Bibliography

[AB] Alperin, J. L.; Bell R. B., “Groups and Representations”, Graduate Texts in Mathe-
matics, 162, Springer-Verlag, New York, 1995.
[Bu] Burnside, W., “On Groups of Order pα q β ”, Proc. London Math. Soc. (2) 1 (1904),
388-392.
[Co] Conrad, Keith, “Groups of order p3 ”, online notes available at
https://kconrad.math.uconn.edu/blurbs/grouptheory/groupsp3.pdf
[CR1] Curtis C. W.; Reiner I., “Methods of representation theory vol I, With applications to
finite groups and orders”, Pure and Applied Mathematics, A Wiley-Interscience Publi-
cation. John Wiley & Sons, Inc., New York, 1981.
[CR2] Curtis C. W.; Reiner I., “Methods of representation theory vol II, With applications
to finite groups and orders”, Pure and Applied Mathematics, A Wiley-Interscience Pub-
lication. John Wiley & Sons, Inc., New York, 1987.
[CR3] Curtis C. W.; Reiner I., “Representation theory of finite groups and associative al-
gebras”, Pure and Applied Mathematics, Vol. XI Interscience Publishers, a division of
John Wiley & Sons, New York-London 1962.
[D1] Dornhoff L., “Group representation theory. Part A: Ordinary representation theory”,
Pure and Applied Mathematics, 7. Marcel Dekker, Inc., New York, 1971.
[D2] Dornhoff, L., “Group representation theory. Part B: Modular representation theory”.
Pure and Applied Mathematics, 7. Marcel Dekker, Inc., New York, 1972.
[DF] Dummit D. S.; Foote R. M. “Abstract algebra”, Third edition, John Wiley & Sons,
Inc., Hoboken, NJ, 2004.
[FH] Fulton; Harris, “Representation theory: A first course” Graduate Texts in Mathemat-
ics, 129, Readings in Mathematics, Springer-Verlag, New York, 1991.
[Gr] Grove, Larry C., “Groups and characters”, Pure and Applied Mathematics (New York).
A Wiley-Interscience Publication. John Wiley & Sons, Inc., New York, 1997. x+212.
[M] Musili C. S., “Representations of finite groups” Texts and Readings in Mathematics,
Hindustan Book Agency, Delhi, 1993.
[Sc] P. Schneider, “Modular representation theory of finite groups”. Springer, Dordrecht,
2013.
111
112 BIBLIOGRAPHY

[FT] Feit, Walter; Thompson, John G., “Solvability of groups of odd order” Pacific J. Math.
13 (1963), 775-1029.
[L] Lang, “Algebra”, Second edition, Addison-Wesley Publishing Company, Advanced Book
Program, Reading, MA, 1984.
[JL] James, Gordon; Liebeck, Martin, “Representations and characters of groups”, Second
edition, Cambridge University Press, New York, 2001.
[Se] Serre J.P., “Linear representations of finite groups” Translated from the second French
edition by Leonard L. Scott, Graduate Texts in Mathematics, Vol. 42, Springer-Verlag,
New York-Heidelberg, 1977..
[Si] Simon, B., “Representations of finite and compact groups” Graduate Studies in Math-
ematics, 10, American Mathematical Society, Providence, RI, 1996.
[St] B. Steinberg, “Representation theory of finite groups. An introductory approach”. Uni-
versitext. Springer, New York, 2012.