Lancaster University

Department of Mathematics and Statistics

MATH325/425: Representation Theory of Finite Groups

Contents

1. Group representations 2
1.1. Introduction 3
1.2. Group representations 5
1.3. Equivalence of representations 9
1.4. Subrepresentations 10
1.5. Direct sums of representations 11
1.6. Workshop Questions 12
1.7. Homework Questions 13
Summary 14
2. CG-modules 14
2.1. Definitions and equivalences of notions 14
2.2. Conventions 17
2.3. Faithful modules 18
2.4. New modules from old ones 19
2.5. Inner products 23
2.6. Maschke’s theorem 26
2.7. Workshop Questions 28
2.8. Homework Questions 29
Summary 30
3. CG-homomorphisms 30
3.1. Definitions 30
3.2. Schur’s lemma 39
3.3. Representations of finite abelian groups 42
3.4. Spaces of CG-homomorphisms 45
3.5. Workshop Questions 52
3.6. Homework Questions 54
Summary 55
4. Irreducible CG-modules and the Group Algebra 55
4.1. Composition factors 55
4.2. Group algebra as CG-module 58
Representation Theory of Finite Groups page 2

4.3. The decomposition of CG 60

4.4. Workshop Questions 66
4.5. Homework Questions 67
Summary 67
Summary 68
5. Characters 68
5.1. Definitions and examples 68
5.2. Inner products 75
5.3. Orthonormality of irreducible characters 77
5.4. Character table of a finite group 80
5.5. Workshop Questions 84
Summary 84
References 84
Index 86

The course aims to introduce the basic concepts and results of the classical theory
of complex representations of finite groups. The study of the representations and
characters of a finite group combines two of the major strands of algebra covered
in the second year: linear algebra and groups. A group is an abstract algebraic
structure. A representation of a group gives a way of working with it in terms of
matrices, which are rather more concrete, and have the advantage that they can be
used in computational algebra. Characters condense the information encoded in the
matrices of a group representation.
Our main objectives in this course :
• Representations of finite groups.
• The group algebra of a finite group and its modules.
• Correspondence between modules and representations.
• Module homomorphisms and their spaces.
• Maschke’s theorem, Schur’s lemma and other key results.
• Characters and main results.
For convenience, there is an appendix comprising the essential notions from linear
algebra and group theory that we need.

1. Group representations
Representation theory concerns the study of groups in terms of matrices. The moti-
vation for this approach is that groups are abstract objects, while matrices are rather
more concrete. In other words, it is simpler to work with matrices. Also, matrices
enable us to use computational algebra to find answers to specific questions.
We first set up the notation that we will use throughout, unless otherwise speci-
fied, together with results assumed to be known from MATH103, MATH220 and
MATH225.
Representation Theory of Finite Groups page 3

1.1. Introduction.
Notation 1.1 (Notation and Assumed background).

• We let n ≥ 1 be a positive integer.

• The set of all n × n matrices with coefficients in C is denoted Mn (C).
• Mn (C) is an abelian group for the addition of matrices and the additive
identity element is the n × n zero matrix.
• Matrix multiplication makes Mn (C) into a unital ring, where the identity
element is the identity matrix of size n, i.e. the matrix
 
1 0 ... 0
0
 
.. 1
0 1
 . 0 

..
In =  . . =  ∈ Mn (C)

. . .
 .. . . . . .. 
0 ... 0 1
0 1

• In particular In A = A = AIn for all A ∈ Mn (C).

• The ring Mn (C) is commutative if and only if n = 1, as then M1 (C) = C. In
general,
for A, B ∈ Mn (C), then AB 6= BA ∈ Mn (C) if n > 1.
• A matrix A ∈ Mn (C) is invertible if and only if there exists a matrix X ∈
Mn (C) with AX = In = XA. If this condition is satisfied, the matrix X is
unique and is called the inverse of A, written A−1 .
• We write GLn (C) for the set of invertible n × n matrices.
• The product of two invertible matrices is invertible since if A and B are
invertible then
(AB)−1 = B −1 A−1
In particular A is invertible if and only if A−1 is invertible.
• Matrix multiplication is associative (i.e. (AB)C = A(BC) for any matrices
A, B, C ∈ Mn (C)) and the identity matrix In is the identity element for the
multiplication of matrices.
• So GLn (C) is a multiplicative group with identity element In .
• GLn (C) is called the general linear group of degree n over C.
• In particular, GL1 (C) = C× is the multiplicative group of nonzero complex
numbers.
• A ∈ Mn (C) is invertible if and only if its determinant is nonzero:
A ∈ GLn (C) ⇐⇒ det A 6= 0
• Equivalently, A ∈ Mn (C) is invertible if and only if its reduced echelon form is
the identity matrix. Then, the product of the elementary matrices performing
the passage from A to its reduced echelon form In is the matrix A−1 .
• GLn (C) is a subset of the additive group Mn (C) but it is not a subgroup of
Mn (C). Indeed, GLn (C) is a group for the multiplication of matrices but not
for the addition, whereas Mn (C) is a group for the addition of matrices but
not for the multiplication.

Let V, W be two C-vector spaces with dim V = n and dim W = m. Let ϕ : V → W

be a linear transformation. That is,
ϕ(v + v 0 ) = ϕ(v) + ϕ(v 0 ) and ϕ(λv) = λϕ(v) , ∀ v, v 0 ∈ V and ∀ λ ∈ C
Representation Theory of Finite Groups page 4

Choose bases B = {e1 , . . . , en } of V and B 0 = {f1 , . . . , fm } of W . The matrix associ-

ated to ϕ is the matrix
X
[ϕ]BB0 ∈ Mm×n (C) = (aij ) where ϕ(ej ) = aij fi
1≤i≤m

That is, the j-th column of [ϕ]BB0 is the image of the j-th basis vector of the domain
V in the coordinates of the basis vectors of the range W . Explicitly, if
X X
v= λj ej then ϕ(v) = aij λj fi
1≤j≤n 1≤j≤n
1≤i≤m

In terms of matrix multiplication, with n = m = 2 we have in particular

    
a11 a12 λ1 a11 λ1 + a12 λ2
ϕ(λ1 e1 + λ2 e2 ) = =
a21 a22 λ2 a21 λ1 + a22 λ2
= (a11 λ1 + a12 λ2 )f1 + (a21 λ1 + a22 λ2 )f2

Therefore,

given bases B of V and B 0 of W ,

to each linear transformation ϕ : V → W
corresponds a unique matrix
[ϕ]BB0 ∈ Mm×n (C)

This is a bijective correspondence which we will use henceforth, and which is essential
to the study of group representations.
As noted above, the matrix expression of a linear transformation depends on the
choice of a basis for the vector spaces. In particular, if ϕ = IdV and B = B 0 , then
[IdV ]BB = In , where n = dim V . But if B 6= B 0 , then [IdV ]BB0 6= In .
Example 1.2. Let V = C2 and take B = {e1 , e2 } be the standard basis of V and
another basis B 0 = {f1 , f2 } of V with
   
3 −1
f1 = 3e1 + e2 = and f2 = −e1 =
1 0
Conversely,

The change of basis matrix from B to B 0 is the matrix

whose columns are the

The converse change of basis is given by the inverse matrix:

Representation Theory of Finite Groups page 5

Now let ϕ : V → V be the linear transformation defined by

ϕ(e1 ) = 2e1 , ϕ(e2 ) = e1 + e2 or equivalently

0
How do we get the matrix [ϕ]BB0 of ϕ expressed in the basis B 0 ? The answer is

That is,
ϕ(f1 ) =
=
ϕ(f2 ) = giving the matrix
0
[ϕ]BB0 = (cf. Figure 1)

Figure 1. The linear transformation ϕ in the bases B and B 0

1.2. Group representations.

Definition 1.3. Let G be a group. A representation of G is a group homomorphism
ρ : G → GLn (C) for some positive integer n. The number n is called the degree of ρ.

From now on G denotes a finite group and we write 1 for the multiplicative identity
element in G.
In other words, a function ρ : G → GLn (C) is a representation of G if and only if
the following two conditions hold:
(i) for all g ∈ G, the matrix ρ(g) is an invertible linear transformation of the
same vector space of dimension n for some positive integer n, and
(ii) ρ(g)ρ(h) = ρ(gh) for all g, h ∈ G.
From elementary properties of group homomorphisms, if ρ : G → GLn (C) is a repre-
sentation, then
ρ(1) = In and ρ(g −1 ) = ρ(g)−1 for all g ∈ G.

Suppose that G is a finite group given by generators and relations

G = hg1 , . . . , gd | R1 = 1, . . . , Rt = 1i where Ri = Ri (g1 , . . . , gd )
Representation Theory of Finite Groups page 6

are words in the gj ’s, i.e. the Ri ’s are monomials in the gj ’s. For instance a cyclic
group G of order n is presented as
G = hg | g n = 1i
while a dihedral group D2m of order 2m admits the presentations
D2m = ha, b | am = b2 = 1 , ba = a−1 bi = hs, t | s2 = t2 = 1 , (st)m = 1i
In particular a presentation of a group is not unique. However a function ϕ : G →
GLn (C) is a group homomorphism if and only if
Ri (ϕ(g1 ), . . . , ϕ(gd )) = 1
whatever the presentation.
Example 1.4.
(i) For any group G and any n ∈ N, the function ρ : G → GLn (C) defined by
ρ(g) = In for all g ∈ G is a representation of G, since for all g, h ∈ G we have

In particular, this example proves that every group has representations of

arbitrarily large degree.
(ii) Let G = ha | a2 = 1i be a cyclic group of order 2. Define ρ : G → GL2 (C) by
 n
n −5 12
ρ(a ) = for n = 1, 2.
−2 5
Since

we have that . So ρ is a group homomor-

phism G → GL2 (C), i.e. a representation of G of degree 2.
(iii) More generally, let G = ha | an = 1i be a cyclic group of order n ≥ 1. Let
2πi
ζ = e n ∈ C be a complex n-th root of 1. Define
ρ : G → C× , ρ(aj ) = ζ j
and
 j
j cos θ − sin θ
τ : G → GL2 (C) , τ (a ) = for 0 ≤ j < n.
sin θ cos θ
Then ρ and τ are both representations of G, of degrees 1 and 2 respectively.
They both represent the cyclic group Cn as a group generated by the rotation
2π
about the origin through an angle as in Figure 2.
n
(iv) The dihedral group of order 8 is the group of isometries of a square, and its
presentation is:
D8 = ha, b | a4 = b2 = 1 , ba = a−1 bi .
That is, a is an anti-clockwise rotation arount the origin and through an angle
π
, and b is the symmetry about the x-axis, both expressed in the standard
2
basis (see Figure 3).
Therefore, the natural way to define a representation of D8 is to take
Representation Theory of Finite Groups page 7

Figure 2. The representation of Cn as group of rotations

Figure 3. The dihedral group D8 as group of isometries

and define the function ρ : G → GL2 (C) with

ρ(ai bj ) = Ai B j for 0 ≤ i ≤ 3 and j = 0, 1 .
We may recall from linear algebra that A and B are the matrices of the linear
transformations of the plane corresponding to a and b respectively. We also
can check algebraically that and . Therefore
ρ is a group homomorphism, i.e. a representation of D8 of degree 2.
Definition 1.5. Let G be a group of order n and V the C-vector space of dimension
n with basis {eg | g ∈ G}. The regular representation of G is the representation
ρ : G → GLn (C) , ρ(g) = (ag,h )g,h∈G
where 
1 in the (gh, h) coefficient
ag,h = for all g ∈ G.
0 otherwise

In other words, left multiplication by g on the elements of G is a permutation of

the group elements: h 7→ gh for all h ∈ G, and the regular representation is the
corresponding permutation matrix.
Definition 1.6. A matrix with exactly one nonzero entry in each row and column,
this entry being 1 in each case, is called a permutation matrix .

In particular, ρ(1) = In , while if e1 is the first basis vector of V , then for any g ∈ G,
the first column of ρ(g) has a 1 in the g-th row and 0 elsewhere.
Representation Theory of Finite Groups page 8

Example 1.7. Let G = {1, a, a2 , a3 } ∼ = C4 be a (cyclic) group of order 4. Using the

4
ordered basis {e1 , ea , ea2 , ea3 } of C , the regular representation of G is defined by

and then ρ(aj ) = ρ(a)j for all 0 ≤ j ≤ 3.

The regular representation of a finite group G is a special instance of a permutation

representation. That is, a representation ρ in which each matrix ρ(g) is a permutation
matrix for all g ∈ G. We shall come back to such representations when discussing
permutation modules.
Another example of a permutation representation is the trivial representation, seen
in Example 1.4.
Definition 1.8. Given a finite group G, a representation ρ of G of degree n is called
a trivial representation if ρ(g) = In for all g ∈ G. Unless otherwise specified, we shall
refer to the trivial representation of G as the trivial representation of G of degree 1,
i.e.
1G : G → GL1 (C) ∼ = C× , 1G (g) = 1 for all g ∈ G.

These two extreme examples of permutation representations show the crucial facts
about representation theory of groups: on the one hand the regular representation
gives an exhaustive list of a group in terms of matrices, but this becomes cumbersome
for large “complicated” groups (e.g. a large nonabelian finite simple group). On the
other hand, the trivial representation of a group is very easy to handle for computing,
but does not allow us to gather much information about the group structure, as all
finite groups “look” alike. This leads us to the following concept.
Definition 1.9. A representation ρ : G → GLn (C) of G is faithful if ker(ρ) = {1}.

So ρ is faithful if and only if the only element g ∈ G with ρ(g) = In is g = 1. Note

that because GLn (C) is a multiplicative group, the kernel of a group homomorphism
G → GLn (C) is the subset
ker(ρ) = {g ∈ G : ρ(g) = In } of G.
Moreover we know (from MATH225) that ker(ρ) is a normal subgroup of G. That is,
gkg −1 ∈ ker(ρ) for all k ∈ ker(ρ) and all g ∈ G.

In particular, the regular representation of any group is faithful, while the trivial
representation is faithful if and only if the group is trivial.
The fundamental isomorphism theorem for groups asserts that if φ : G → H is a
group homomorphism, then G/ ker(φ) ∼= im(φ).
Proposition 1.10. A representation ρ of G is faithful if and only if im(ρ) ∼
= G.

Proof. 
Representation Theory of Finite Groups page 9

Example 1.11. Let

D8 = ha, b | a4 = b2 = 1 , ba = a−1 bi .
The representation of D8 seen in Example 1.4, ρ : G → GL2 (C), with
ρ(ai bj ) = Ai B j for 0 ≤ i ≤ 3 and j = 0, 1
where    
0 −1 1 0
A= and B = in GL2 (C) ,
1 0 0 −1
is a faithful representation of D8 . One can check that the 8 matrices Ai B j for 0 ≤ i ≤ 3
and j = 0, 1 are all distinct, or prove that im(ρ) ∼
= D8 . We leave the details in exercise.

1.3. Equivalence of representations.

As pointed out above, the matrix expression of a linear transformation depends on
the choice of bases for the domain and the codomain. Therefore, two distinct matrices
may be the same linear transformation, but written in different bases. For our pur-
poses, it suffices to consider the case when domain and range of a linear transformation
are the same. Let ϕ : V → V be a linear transformation, and for convenience, if we
take the same basis B = {e1 , . . . , en } at both ends, we simply write [ϕ]B . Referring
to Example 1.2, we note that two matrices express the same linear transformation if
and only if they are similar (or conjugate). Specialising this to group representations
defines the concept of equivalent representations.
Definition 1.12. Let ρ : G → GLn (C) and σ : G → GLm (C) be two representations
of G. We say that ρ is equivalent to σ if n = m and there exists T ∈ GLn (C) such
that σ(g) = T −1 ρ(g)T for all g ∈ G. In this case, we write σ ∼ ρ.

In other words, two representations ρ and σ of G of degree n are equivalent if and

only if there exist two bases Bρ and Bσ of Cn and invertible linear transformations
{ϕg | g ∈ G} of V such that
ρ(g) = [ϕg ]Bρ and σ(g) = [ϕg ]Bσ for all g ∈ G.
Example 1.13. Consider the regular representation of C4 as in Example 1.7,
 
0 0 0 1 e1 7−→ ea
1 0 0 0 ea 7−→ ea2
ρ : G → GL4 (C) , ρ(a) =  
0 1 0 0 ea2 7−→ ea3
0 0 1 0 ea3 7−→ e1
and then ρ(aj ) = ρ(a)j for all 0 ≤ j ≤ 3. Write B = {e1 , ea , ea2 , ea3 }, so that
ρ(a) = [ρ(a)]B .
Let B 0 = {f1 , f2 , f3 , f4 } the basis of C4 where
f1 = e1 + ea + ea2 + ea3 giving
f2 = e1 − iea − ea2 + iea3 giving
f3 = e1 − ea + ea2 − ea3 giving
f4 = e1 + iea − ea2 − iea3 giving
The change of basis matrices are
Representation Theory of Finite Groups page 10

and we calculate

0
[Id]BB0 [ρ(a)]B [Id]BB =

which defines the representation τ : G → GL4 (C) by

and then τ (aj ) = τ (a)j for 0 ≤ j ≤ 3. Since T −1 ρ(a)T = τ (a) for

T =

we conclude that ρ ∼ τ .

1.4. Subrepresentations.
Definition 1.14. Let ρ : G → GLn (C) be a representation of G and write V = Cn
for the underlying vector space. A subspace W ⊆ V is G-invariant if ρ(g)w ∈ W for
all w ∈ W and all g ∈ G.
If W is a G-invariant subspace of V , the representation ρ defines a subrepresentation
of G as follows. Let C = {e1 , . . . , em } be a basis for W , with m ≤ n. Put
ρW : G −→ GLm (C) , ρW (g) = [ρW (g)]C = (agij ) for all g ∈ G
where
X
ρW (g)ej = ρ(g)ej = agij ei for coefficients agij ∈ C, ∀ g ∈ G.
1≤i≤m

This is well defined since W is G-invariant, that is,

ρ(g)(ej ) ∈ W is a linear combination of {e1 , . . . , em }
for all g ∈ G and all 1 ≤ j ≤ m. We call ρW proper if W is a proper subspace of V .

In Example 1.13, we observe that the subspaces

Wj = hfj i are G-invariant. Indeed,

and similarly for the powers of a, which is immediate from the matrix [τ (a)]B0 .
Definition 1.15. A representation ρ : G → GLn (C) of G of degree n is irreducible
if n > 0 and ρ does not possess any proper nonzero subrepresentation. If ρ is not
irreducible, then we say that ρ is reducible.

In particular, any representation of degree 1 of any finite group is irreducible.

Representation Theory of Finite Groups page 11

Example 1.16. Let G = ha | a3 = 1i and ρ : G → GL3 (C) defined by

 
0 0 1
ρ(a) = 1 0 0
0 1 0

where the matrix is expressed in the basis {e1 , e2 , e3 } of G. Then ρ is reducible since
the subspace W = hf1 = e1 + e2 + e3 i is G-invariant. We have

that is,

ρW : G → GL1 (C) ∼
= C× , ρW (a) = 1 is the trivial representation of G.

2πi
Similarly, let ζ = e 3 ∈ C be a primitive cube root of 1, and put f2 = e1 + ζ 2 e2 + ζe3
and f3 = e1 + ζe2 + ζ 2 e3 . Then

ρ(a)(f2 ) =

ρ(a)(f3 ) =

So hf2 i and hf3 i are G-invariant subspaces too. In the basis B 0 = {f1 , f2 , f3 } of C3 ,
the above calculation shows that

[ρ(a)]B0 =

Suppose that ρ is a reducible representation of G of degree n, with invariant subspace

W of dimension m ≤ n. Then ρ is equivalent to a representation of the form
 
ρW (g) A
[ρ(g)] = for all g ∈ G
0 B

where ρW (g) ∈ GLm (C) is a subrepresentation of ρ, where A ∈ Mm×(n−m) (C) and

B ∈ GLn−m (C).

1.5. Direct sums of representations.

Definition 1.17. Let ρj : G → GLnj (C) for j = 1, 2. The direct sum of ρ1 and ρ2 is
the representation
 
ρ1 (g) 0
ρ1 ⊕ ρ2 : G → GLn1 +n2 (C) , (ρ1 ⊕ ρ2 )(g) =
0 ρ2 (g)
Representation Theory of Finite Groups page 12

of G of degree n1 + n2 . Inductively, the direct sum ρ1 ⊕ · · · ⊕ ρk of representations

ρj : G → GLnj (C) for 1 ≤ j ≤ k is the representation
 
ρ1 (g) 0 ... 0
. .. 
 0 ρ2 (g) . . . 

 . .. 
(ρ1 ⊕ · · · ⊕ ρk )(g) = 
 .
.  ∈ GLN (C)
. 
 . . .
 .. .. ..

0 
0 ... 0 ρk (g)
of G of degree N = n1 + · · · + nk .

For instance, in Example 1.16, ρ is the direct sum of three irreducible representations
of degree 1.

1.6. Workshop Questions.

Exercise 1.6.1. Let G = ha | a2 = 1i be a group of order 2. Find λ ∈ C such that
the function
ρ:G → GL2 (C)
 j
j 10 −11 (j = 0, 1) is a representation of G.
a 7→
λ −10
Exercise 1.6.2. Let G = D12 , and set
 iπ √ !
1 3
    
e3 0 0 1 2√ 2
1 0
C= π , D= , E= , F = .
0 e−i 3 1 0 − 2
3 1
2
0 −1
Show that ρ1 , ρ2 , ρ3 and ρ4 are representations of G, where
ρ1 (bi aj ) = Di C j ,
ρ2 (bi aj ) = (−D)i C 3j ,
ρ3 (bi aj ) = Di (−C)j ,
ρ4 (bi aj ) = F i E j .
Which of these four representations are faithful? Which of them are equivalent to
each other?
Exercise 1.6.3. Let G be a finite group and ρ, τ two equivalent representations of
G. Prove that ρ is irreducible if and only if τ is irreducible.
Exercise 1.6.4. Let G = GLm (C) and H = Mm (C). Decide which of the following
assertions hold and briefly prove your claims. Recall that GL1 (C) ∼
= C× as
multiplicative groups.
(i) det : G → C× is a representation of G.
(ii) det : H → C× is a representation of H.
(iii) Tr : G → C× is a representation of G.
(iv) Tr : H → C× is a representation of H.
(v) det : G → C× is a faithful representation of G.
Exercise 1.6.5. Let n ≥ 3 and G = ha, b | an = b2 = 1 , ba =a−1 bi a dihedral

2π cos θ − sin θ
group of order 2n. Let θ = n and consider the matrices A = and
sin θ cos θ
 
1 0
B= .
0 −1
Representation Theory of Finite Groups page 13

(i) Prove that the function

ρ : G → GL2 (C) , ρ(ak bl ) = Ak B l for all 0 ≤ k < n and l = 0, 1
is a representation of G.
(ii) Prove that ρ is faithful.
(iii) Find a basis B of C2 such that the matrix [ρ(a)]B is diagonal.
Exercise 1.6.6. Two of the following three representations of the group
G = ha | a12 = 1i are equivalent. Find which ones, and justify your answer.
 
0 −1
(i) ρ1 (a) =
1 0
 
0 −ω 2 2πi
(ii) ρ2 (a) = where ω = e 3 is a complex primitive cube root of 1.
ω 0
 
0 −ω 2πi
(iii) ρ3 (a) = 2 where ω = e 3 is a complex primitive cube root of 1.
1 −ω
Exercise 1.6.7.
(i) Let N be a normal subgroup of a finite group G and ρ : G/N → GLn (C) a
faithful representation of the quotient group G/N . Prove that the function
ρb : G → GLn (C) , ρb(g) = ρ(b
g ) where
gb ∈ G/N is the equivalence class in G/N containing g is a representation of
G with kernel N .
(ii) Conversely, prove that any representation ρ : G → GLn (C) defines a faithful
representation of the quotient group ρ : G/ ker(ρ) → GLn (C) with
ρ(bg ) = ρ(g) for any g ∈ gb and any gb ∈ G/ ker(ρ).

1.7. Homework Questions.

Exercise 1.7.1. Let G = D8 , and set
       
i 0 0 1 0 1 1 0
C= , D= , E= , F = .
0 −i 1 0 −1 0 0 −1
Show that ρ1 , ρ2 , and ρ3 are representations of G, where
ρ1 (bi aj ) = Di C j ,
ρ2 (bi aj ) = (−D)i C 2j ,
ρ3 (bi aj ) = F i E j .
Which of these three representations are faithful? Which of them are equivalent to
each other?
Exercise 1.7.2. Let G be a finite group and ρ, τ two equivalent representations of
G. Prove that ρ is faithful if and only if τ is faithful.
Exercise 1.7.3. Let G = ha , b | a3 = b2 = 1, b a = a−1 bi be the dihedral group
of order 6 (or equivalently G = S3 with a = (123) and b = (23)), and let ρ be the
regular representation of G.
(i) Write the matrices ρ(a) and ρ(b) in the ordered basis {e1 , ea , ea2 , eb , eab , ea2 b }
of C6 .
(ii) Prove that ρ is faithful.
(iii) Find a trivial subrepresentation of ρ.
Representation Theory of Finite Groups page 14

Summary.
The new notions studied in this section are:
• representations of a finite group G
• special representations: regular, trivial, permutation, faithful
• equivalence of representations
• subrepresentations
• direct sums of representations

2. CG-modules

In this section we introduce the concept of a CG-module, and we show that CG-
modules and representations of G are equivalent notions.

2.1. Definitions and equivalences of notions.

In the previous section, we have seen above that to each representation of a group,
say ρ : G → GLn (C) the associated vector space V = Cn is equipped with an action
of G. That is a function
f :G×V →V , f (g, v) = ρ(g)(v) ∈ V for all g ∈ G and all v ∈ V ,
subject to
(i) f (1, v) = v for all v ∈ V
(ii) f (gh, v) = f (g, ρ(h)(v)) since ρ(gh)(v) = ρ(g)(ρ(h)(v)) because ρ is a group
homomorphism
(iii) f is linear in the second variable, i.e.
(a) f (g, v + w) = f (g, v) + f (g, w) since ρ(g)(v + w) = ρ(g)(v) + ρ(g)(w),
and
(b) f (g, λv) = λf (g, v) since ρ(g)(λ)(v) = λρ(g)(v) because ρ(g) ∈ GLn (C)
is a linear transformation of V .
Definition 2.1. Given a finite group G, a CG-module is a C-vector space equipped
with an action G × V → V of G. That is, for all g ∈ G and all v ∈ V , there is an
element gv ∈ V subject to the following axioms. For all u, v ∈ V , all λ ∈ C, and all
g ∈ G,
(i) 1v = v
(ii) (gh)v = g(hv)
(iii) g(u + v) = gu + gv
(iv) g(λv) = λ(gv)

We assume throughout that modules are finite dimensional C-vector spaces, and
that G acts on the left. In some textbooks, the modules are allowed to be infinite
dimensional vector spaces over any field, or the group G acts on the right (i.e. vg
Representation Theory of Finite Groups page 15

instead of gv for g ∈ G and v ∈ V ). Often, textbooks present the theory of a more

general notion of module, namely that of a module over a ring.
By definition, if V is a CG-module, then for all g ∈ G the map v 7→ gv is a linear
transformation of V , while the axiom (gh)v = g(hv) translates in terms of modules
that the association of g to its action on V is a group homomorphism G → L(V, V ),
where L(V, V ) is the set of linear transformations V → V .
In Section 1.2 we have seen that given a CG-module V we can let each element g ∈ G
act as a linear transformation of V . Explicitly, say that B = {v1 , . . . , vn } is a basis
of V as vector space and pick g ∈ G. Let [g]B ∈ Mn (C) be the matrix whose j-th
column is the column vector gvj expressed as a linear combination of the elements
of B. Explicitly, the (i, j) coefficient of [g]B is the complex number aij where
n
X
gvj = akj vk
k=1

In particular, [1]B = In and this is independent from the choice of B. Indeed, 1v = v

for any v ∈ V .
Axioms (ii) and (i) show that
v = 1v = (g −1 g)v = g −1 (gv) for all g ∈ G and all v ∈ V ,
that is, for each g ∈ G, the matrix corresponding to the action of g has an inverse,
namely the matrix corresponding to g −1 . Therefore,
[g]B ∈ GLn (C) and [g]−1 −1
B = [g ]B

Axioms (ii) and (iii) show that the correspondence

g 7→ [g]B is a group homomorphism G −→ GLn (C)
In other words,
given a CG-module V , we obtain a representation of G of degree dim V .

The above gives the method to obtain a representation of G from a CG-module, using
matrices. Conversely, suppose that ρ : G → GLn (C) is a representation of G of
degree n. Set V = Cn for the space of column vectors and define
gv = ρ(g)v for all v ∈ V and all g ∈ G.
The left-hand side of the equality denotes the group action, whereas the right-hand
side is a matrix multiplication. It is straightforward that this action of G makes the
vector space V into a CG-module. So
given a representation of G of degree n = dim V , we get a CG-module V .
The statements in the two boxes above say that CG-modules and representations of G
are equivalent notions, and make the equivalence explicit, i.e. provide the dictionary
CG-modules ↔ Representations of G.
Definition 2.2. If a representation ρ and a CG-module V are in such a correspon-
dence, we say that ρ affords (or corresponds to) V .
Notation 2.3. For convenience of notation, if we consider a CG-module V spanned
as vector space by nonzero vectors, v1 , . . . , vn ∈ V , for some n ∈ N and some finite
group G, we simply write V = hv1 , . . . , vn i.
Example 2.4. Let G = ha | a4 = 1i be a cyclic group of order 4.
Representation Theory of Finite Groups page 16

(i) ρ : G → C× , ρ(aj ) = 1 for 0 ≤ j ≤ 3 affords the trivial 1-dimensional

CG-module V = hvi, i.e. with the trivial G-action,
(ii) ρ : G → C× , ρ(aj ) = ij for 0 ≤ j ≤ 3 affords the (nontrivial) 1-
dimensional CG-module V = hvi with the G-action,
 j
j 0 −1
(iii) ρ : G → GL2 (C) , ρ(a ) = affords the 2-dimensional CG-
1 0
module V = hv1 , v2 i with G-action

Note that ρ(a2 ) =  

0 0 0 1
 1 0 0 0
(iv) The permutation representation ρ : G → GL4 (C) , ρ(aj ) =  0 1 0

0
0 0 1 0
affords the 4-dimensional CG-module V = hv1 , v2 , v3 , v4 i with G-action
Representation Theory of Finite Groups page 17

and then inductively,

2.2. Conventions.
Let V be a vector space and G a finite group with a function G × V → V . We want
to find a minimum set of conditions to check in order to determine whether V is a
CG-module or not.
Suppose that G has a presentation
G = hg1 , . . . , gr | R1 = · · · = Rt = 1i where Ri = Ri (g1 , . . . , gr ) are the relations,
and that B = {v1 , . . . , vn } is a basis of V . Then V has a structure of CG-module,
not unique in general, if there exists a well-defined G-action on V . That is, if we can
define a product
gi vj for all 1 ≤ i ≤ r and all 1 ≤ j ≤ n
subject to the following conditions for all the indices:

(i) gi vj ∈ V
(ii) Ri (g1 , . . . , gr )vj = vj for all 1 ≤ i ≤ t. In particular, if gi has order ni , i.e.
gini = 1, then
gi · (gi · · · (gi ·vj )·) = vj = 1vj
| {z }
ni terms

Extending linearly these products to V defines a group action G × V → V which

equips V with a structure of CG-module.
Example 2.5. Let G = hg | g n = 1i be a cyclic group of order n ≥ 1 and let V = hvi
2πi
be a 1-dimensional vector space. Put ζ = e n for a primitive n-th root of 1 in C, and
define
gv = ζv
The extension of this product to all of CG and all of V gives linear combinations
g i (λv) = ζ i λv
In particular, if n = 2, then ζ = −1 and we have
g(λv) = −λv for all λ ∈ C.

Example 2.6 (A non-example). For example, if G = hg | g 3 = 1i ∼ = C3 and

∼
V = hvi = C a vector space of dimension one, we cannot define gv = 2v because
then which contradicts axiom (i) of Defi-
nition 2.1.

We now translate in terms of modules some of the concepts defined for representations.
From a historical perspective, group representations using group homomorphisms
appeared earlier than modules, but the modern approach of representation theory of
finite groups is done via CG-modules instead. Also, for computational reasons, these
latter are much easier to handle, especially when looking for specific answers about a
given finite group.
Representation Theory of Finite Groups page 18

2.2.1. Permutation modules.

Definition 2.7. Let n ∈ N and G a subgroup of the symmetric group Sn . The CG-
module V with basis {v1 , . . . , vn } on which G acts by gvi = vgi for all g ∈ G and all
1 ≤ i ≤ n is called the permutation module for G. We call {v1 , . . . , vn } the natural
basis of V . In particular, each matrix [g]B is a permutation basis.

For completeness, the above definition can be somewhat misleading, because “the”
permutation CG-module is not unique. It depends on how we identify G with a sub-
group of the symmetric group. In particular there may be several possible symmetric
groups. The following examples show this fact. There is no canonical way to define
“the” permutation CG-module in general, unless for cyclic groups or full symmetric
groups. A cyclic group of order n is identified with the subgroup generated by the
n-cycle (1 2 . . . n) ∈ Sn unless otherwise specified.
Example 2.8.
(i) Let G = hg | g 4 = 1i ∼
= C4 . The permutation CG-module is the 4-dimensional
vector space V with natural basis B = {v1 , v2 , v3 , v4 } on which the action of
g is given by the matrix

[g]B =

Here we identify the element g ∈ G with the permutation

(ii) Let G = hg, h | g 2 = h2 = (gh)2 = 1i ∼ = C2 × C2 .
(a) A possible permutation CG-module is the 4-dimensional vector space V
with natural basis B = {v1 , v2 , v3 , v4 } on which the action of G is given
by the matrices

[g]B = and [h]B =

Here we identify g, h ∈ G with the permutations

respectively.
(b) Another permutation CG-module is also the 4-dimensional vector space
V with natural basis B = {v1 , v2 , v3 , v4 }, but this time the action of G is
given by the matrices

[g]B = and [h]B =

In this case, we identify the elements g, h ∈ G with the permutations

respectively. One can check that the
matrix of gh = hg corresponds to .

2.3. Faithful modules.

Definition 2.9.
Representation Theory of Finite Groups page 19

(i) The kernel of a CG-module V is the subset

ker(V ) = {g ∈ G | gv = v} of G.
(ii) A CG-module V is faithful if ker(V ) = {1}, that is if the only element g ∈ G
for which gv = v for all v ∈ V is the identity element.
(iii) The trivial CG-module is the vector space V of dimension 1 with gv = v
for all g ∈ G, and all v ∈ V . More generally, a CG-module V is trivial if
ker(V ) = G.
Lemma 2.10. Let V be a CG-module. Then ker(V ) is a normal subgroup of G.

Proof. We need to show that for any h ∈ ker(V ) and g ∈ G we have ghg −1 ∈ ker(V ).
That is, for each v ∈ V , we need to check that

We calculate

where (1) and (3) hold by associativity of the G-action on V and (2) holds by as-
sumption on h ∈ ker(V ). 
Example 2.11.
(i) Let G = hg | g 2 = 1i ∼
= C2 and let V = hv1 , v2 i the CG-module defined by
gv1 = v2 and gv2 = v1
Then ker(V ) = and so V is
(ii) Let G = hg, h | g 2 = h2 = 1 , gh = hgi ∼
= C2 × C2 and let V = hv1 , v2 i the
CG-module defined by
gv1 = hv1 = v2 , gv2 = hv2 = v1
Then ker(V ) = and so V is

Note that the condition for a CG-module to be faithful concerns all v ∈ V . It is quite
possible for a nonidentity element of G to fix some vectors in V . We will see such
instances in the exercises.

2.4. New modules from old ones.

In this section, we construct new modules from given ones. Mainly, we translate in
terms of modules the ad hoc notions seen for group representations.

2.4.1. CG-submodules.
Definition 2.12. Let V be a CG-module. A subset U of V is a CG-submodule of V ,
or simply a submodule of V , if U is a G-invariant subspace of V , that is, U is a
subspace of V such that gu ∈ U for all g ∈ G and all u ∈ U .

In particular, a CG-submodule of a CG-module V is a subspace U of V such that U

is itself a CG-module for the same G-action.
Definition 2.13. A nonzero CG-module V is irreducible if the only CG-submodules
of V are {0} and V . If V is not irreducible, we say that V is reducible.
Representation Theory of Finite Groups page 20

In particular a reducible CG-module V must have dimension at least 2 as any 1-

dimensional vector space has no nonzero proper subspace.
Suppose that U is a submodule of V , and take a basis B = {u1 , . . . , um } of U .
Complete B into a basis B 0 = {u1 , . . . , um , vm+1 , . . . , vn } of V . For each g ∈ G, the
matrix [g]B0 of the action of g on V has a block triangular form:

where the coefficients in the columns of A ∈ Mm×(n−m) (C) and B ∈ GL(n−m) (C),
stacked on top of each other, are the coordinates of gvm+1 , . . . , gvn in the basis B 0 .
Conversely, if a given CG-module V has a basis B such that each matrix [g]B has the
above form, then the subspace W spanned by the first m vectors of the basis B is
mapped to itself by multiplication by any [g]B . That is W is a CG-submodule of V ,
saying also that V is reducible.
In other words, V is reducible if and only if there is a basis B such that each matrix
[g]B has the above form, with a zero block in the bottom left corner.

2.4.2. Direct sums.

Definition 2.14. Let G be a finite group and V1 , . . . , Vn be CG-modules. The direct
sum of V1 , . . . , Vn is the CG-module with underlying vector space the direct sum
M
V1 ⊕ · · · ⊕ Vn also written Vj for short.
1≤j≤n

Recall from linear algebra that a vector space V is a direct sum V1 ⊕ · · · ⊕ Vn if and
only if each element v ∈ V can be written in a unique way as a linear combination
v = v1 + · · · + vn with vj ∈ Vj for all 1 ≤ j ≤ n.
In other words, the following two conditions hold:

• V is generated by V1 , . . . , Vn (generating set)

• there is no nontrivial linear combination of vectors of V1 , . . . , Vn , one in each
Vj . (linear independence)

If V = V1 ⊕ · · · ⊕ Vn , then a basis of V is obtained by taking the union of bases of

each Vj .
In particular, if n = 2, then V = V1 ⊕ V2 is equivalent to saying that:

• each element of V can be written as v1 + v2 for some vj ∈ Vj for j = 1, 2

• V1 ∩ V2 = {0}

The latter condition only holds if we have a sum of two terms.

It follows, in the context of CG-modules, that the same notion of direct sum works,
provided we consider CG-modules V, V1 , . . . , Vn instead of vector spaces.
Representation Theory of Finite Groups page 21

Example 2.15. Let G = hg | g 4 = 1i and take the CG-modules


gv = v1
 1


gv2 = −v2
Vj = hvj i , 1 ≤ j ≤ 4 where

 gv3 = iv3
 gv = −iv
4 4

Then V = V1 ⊕ V2 ⊕ V3 ⊕ V4 is the CG-module of dimension 4 with basis B =

{v1 , v2 , v3 , v4 } and on which g acts by the matrix [g]B = .

Example 2.16. Let G = ha, b | a4 = b2 = 1 , ba = a−1 bi be the dihedral group of

order 8 and V = hv1 , v2 i the CG-module defined by
av1 = −v2 , av2 = v1 , bv1 = v1 and bv2 = −v2
Let W = hwi be the trivial CG-module, that is, aw = bw = w. Then dim V = 2 and
dim W = 1. The CG-module V ⊕ W has dimension 3, and consists of all the elements
λv1 + µv2 + νw with λ, µ, ν ∈ C.
We have for example

Example 2.17. Let G = hg | g 3 = 1i be a group of order 3, and V the permuta-

tion CG-module with natural basis {v1 , v2 , v3 }. The 1-dimensional subspace U of V
spanned by the vector u = v1 + v2 + v3 is a trivial CG-submodule of V since gu = u.
Put w1 = v2 − v1 , w2 = v3 − v1 , and let W be the subspace spanned by w1 , w2 . We
calculate
Representation Theory of Finite Groups page 22

Which proves that gw ∈ W for all w ∈ W and so W is a CG-submodule of V since

g generates G.
We want to prove that V = U ⊕ W . Since U and W are both submodules, we know
that U + W is a CG-submodule of V too. To prove that the sum is direct and equal
to all of V , it then suffices to show that U ∩ W = {0} because then
U +W =U ⊕W has dimension 3 = dim(V ), and so U ⊕ W = V.
Suppose
Then which holds iff There-
fore U ∩ W = {0} as we wanted to show.
In this example we see different matrix forms of the same linear transformation of
V defined by multiplication by g ∈ G. Namely, in the natural basis B we have the
permutation matrix
 
0 0 1
[g]B = 1 0 0
0 1 0
whereas in the basis B 0 we have

[g]B0 = T −1 [g]B T = where T =

is the change of basis matrix.

In particular, we note a block decomposition in which the only nonzero blocks are
diagonal:
 
1 0 0
[g]B0 =  0 −1 −1 
0 1 0
∗ 0
 
of the form for some matrices 0 and ∗ of appropriate sizes.
0 ∗

Suppose that V = ⊕ri=1 Vi and that Bi is a basis of Vi for all 1 ≤ i ≤ r. We can

amalgamate the bases B1 , B2 , . . . , Br to obtain a basis B of V . Take any g ∈ G.
Given any basis vector v1 in B1 , we have gv1 ∈ V1 , since V1 is a CG-submodule of V .
So gv1 may be written as a linear combination of the vectors in B1 and the coefficient
in gv1 of any basis vector in Bi for i 6= 1 is zero. Taking each basis Bi in turn, we see
that
[g]B1 0 ... 0
 
 0 [g]B2 . . . 0 
[g]B = 
 .. . .
.. . .. .. 
. 
0 0 ... [g]Br

Last, an example of a sum of modules which is not direct.

Example 2.18. Consider the cyclic group G = ha | a4 = 1i of order 4, and V the
permutation CG-module with natural basis B = {v1 , v2 , v3 , v4 }. For j = 2, 3, 4, let
uj = vj − v1 . Let also w1 = v3 − v1 , w2 = v4 − v2 and w3 = v1 + v2 + v3 + v4 . Hence,
consider the subspaces
V1 = hu2 , u3 , u4 i and V2 = hw1 , w2 , w3 i.
Representation Theory of Finite Groups page 23

By definition

au2 =

au3 = that is [a]B = in matrix form

au4 =
Thus, V1 is a CG-submodule of V , since a generates the entire group G. Similarly,
we have

that is in matrix form, saying that V2 is also a CG-submodule of V .

Let us form the subspace V1 + V2 of V , i.e.

V1 + V2 = {u + w | u ∈ V1 , w ∈ V2 } ⊆ V
We want to show that V1 + V2 = V and that this sum is not direct. By arguing
on the dimensions, since dim V1 + dim V2 = , we know that the sum
cannot be direct. More precisely, dim(V1 + V2 ) = dim V1 + dim V2 − dim(V1 ∩ V2 ), and
a routine computation gives , and On
the other hand, w3 ∈
/ V1 , because the equation
X
v1 +v2 +v3 +v4 = λi ui =
has no solutions for λ2 , λ3 , λ4 . Therefore
V1 ∩ V2 = hw1 , w2 i which is a 2-dimensional CG-module, and
dim(V1 + V2 ) = dim V1 + dim V2 − dim(V1 ∩ V2 ) =

implying that V1 + V2 = V .

2.5. Inner products.

We review inner products on complex vector spaces. As usual, the complex conjugate
of z = x + iy ∈ C is the complex number z = x − iy.
Definition 2.19. Let V be a vector space over C. A (complex) inner product on V
is a function
(, ) : V ×V →C
subject to the following axioms: for all u, v, w ∈ V and all λ ∈ C
(i) Conjugate symmetry: (v, w) = (w, v).
(ii) Linearity in the first argument:
(u + v, w) = (u, w) + (v, w) and (λv, w) = λ(v, w).
(iii) Positive definite: (v, v) ≥ 0 with equality if and only if v = 0.
Representation Theory of Finite Groups page 24

Two vectors v, w ∈ V are orthogonal if (v, w) = 0.

Given a subspace U of V , we call
U ⊥ = {v ∈ V | (v, u) = 0 for all u ∈ U }
the orthogonal complement of U . An orthogonal basis of V is a basis of V in which
the vectors are pairwise orthogonal. That is, {v1 , . . . , vn } is an orthogonal basis of V
if (vj , vk ) = 0 whenever k 6= j. If moreover (vj , vj ) = 1 for all j, then we say that B
is an orthonormal basis of V .

Note that the inner product of two elements of V is a complex number. Thus in
general it is meaningless to write (v, w) ≥ 0, since inequalities of complex numbers
do not make sense. However, taking w = v in axiom (i) shows that (v, v) = (v, v). So
(v, v) is a real number for any v ∈ V , and axiom (iii) makes sense.
It is immediate that if some vector space V is equipped with the standard inner
product, then V ⊥ = {0}. Moreover, if U is a subspace of V , the equality V = U ⊕ U ⊥
shows that
dim V = dim U + dim U ⊥ (cf. MATH220)
Example 2.20. Let V = C3 with basis {v1 , v2 , v3 }. The standard inner product on
V is the function
!
X X X
V ×V →C , λj vj , µk vk = λj µj
1≤j≤3 1≤k≤3 1≤j≤3

that is, (vj , vk ) = 0 if j 6= k and (vj , vj ) = 1. For instance

In particular, for any root of unity λ ∈ C, since λ−1 = λ, we have


1 if i = j
(λvi , λvj ) = λλδi,j = for i, j ∈ {1, 2}.
0 otherwise.
Example 2.21. Let V = C2 with the standard basis {e1 , e2 } and with the standard
inner product
 (ek , ej ) = δk,j = 1 if k = j and 0 otherwise (k, j = 1, 2). Let U =
1
he1 + ie2 = i. The orthogonal complement U ⊥ of U is spanned by all the vectors
i
 
a
ae1 + be2 = ∈ V such that (e1 + ie2 , ae1 + be2 ) = 0. We calculate
b

which gives the constraint

Therefore, putting a = 1, we get , giving So we conclude that U ⊥
has dimension 1 and is spanned by

Now, given an inner product ( , ) of vector space on a CG-module V , we define an

inner product on V which is G-invariant, i.e. a function h , i : V × V → C such
that the following two conditions hold:
(i) h , i is an inner product of the vector space V , and
(ii) hgv, gv 0 i = hv, v 0 i for all v, v 0 ∈ V and for all g ∈ G.
Representation Theory of Finite Groups page 25

Definition 2.22. Let G be a finite group and V a CG-module. Suppose that ( , ) is

a given inner product on V . Define,
X
hv, wi = (gv, gw) for all v, w ∈ V.
g∈G

Then h , i is an inner product on V and it has the property that

hgv, gwi = hv, wi for all g ∈ G and all v, w ∈ V .

We now prove that Definition 2.22 makes sense, provided G is finite (because of the
sum!).

Proof. We need to check the axioms of Definition 2.19. For (i), we note that for all
v, w ∈ V and all g ∈ G, we have

X
hv, wi = (gv, gw) =
g∈G

For (ii), let u, v, w ∈ V , and let λ ∈ C. We calculate

hu + v, wi =

=
and hλv, wi =

Thus (ii) holds. Finally, for axiom (iii), we note that for any v ∈ V , we have

with equality if and only if each summand (gv, gv) is zero, i.e. if and only if gv = 0
for all g ∈ G. Since 1v = v we must have v = 0. Therefore h , i is an inner product
as claimed.
For the last claim, let g ∈ G and v, w ∈ V . We have

hgv, gwi =

since h runs through G if and only if k = hg does. 

Representation Theory of Finite Groups page 26

2.6. Maschke’s theorem.

We use inner product on CG-modules to study them. The main result that we will
obtain in doing so is Maschke’s theorem. We first need a technical result.
Proposition 2.23. Let U be a CG-submodule of a CG module V and suppose that
h , i is a G-invariant inner product on V . Then U ⊥ is a CG-submodule of V and
U ⊕ U ⊥ = V as CG-modules.

Proof. By MATH220, we know that U ⊕ U ⊥ = V as vector spaces, and so all that we

need to show is that U ⊥ is a CG-submodule of V . By assumption, U ⊥ is a subspace
of V . Let g ∈ G and u ∈ U ⊥ . We want to prove that gu ∈ U ⊥ , and so we calculate
for an arbitrary element v ∈ U

where (1) holds because we suppose that h , i is G-invariant and (2) holds because
U is a CG-module and so for any g ∈ G and any v ∈ U . 
Example 2.24. Let G = S3 and V the permutation module for G with natural basis
{v1 , v2 , v3 }. Take the standard inner product ( , ) on V and use it to define the inner
product h , i of Definition 2.22. Then hvi , vj i = for all 1 ≤ i, j ≤ 3 since
|G| = 6 and (gvi , gvj ) = 1 if i = j and 0 otherwise.
Let u = v1 + v2 + v3 . Then U = hui is a trivial CG-submodule of V , since
for all g ∈ G. We have
hu, ui =

We want to find U ⊥ with respect to h , i. Let v = λ1 v1 + λ2 v2 + λ3 v3 ∈ V . Then

v ∈ U ⊥ if and only if hv, ui = 0. Now,
X
hv, ui = (gv, gu) =
g∈G

Let us write the vectors gv as column vectors:

 
λ1
we have v = λ2  and we calculate

λ3
X
gv =
g∈G

So v ∈ U ⊥ if and only if

Therefore W = U ⊥ = .
Representation Theory of Finite Groups page 27

Let B = {u, v2 , v3 } and B 0 = {u, v1 − v2 , v2 − v3 } be bases of V . For all g ∈ G the

matrices [g]B and [g]B0 have the form
   
∗ ∗ ∗ ∗ 0 0
[g]B = 0 ∗ ∗ [g]B0 = 0 ∗ ∗
0 ∗ ∗ 0 ∗ ∗
The zeroes in the first column of each matrix reflect the fact that U is a CG-submodule
of V . The “block-diagonal” shape of [g]B0 is due to the fact that
{ v1 − v2 , v2 − v3 } is a basis for the CG-submodule U ⊥ of V = U ⊕ U ⊥ .
Note indeed that U ∩ U ⊥ = {0} and since there are only two subspaces, this forces
the sum U + U ⊥ to be direct U ⊕ U ⊥ ⊆ V (cf. MATH220). Moreover dim V = 3 =
dim U + dim U ⊥ , saying that we have equality V = U ⊕ U ⊥ .

We are ready to state a major result in representation theory, namely Maschke’s

theorem. It basically allows us to reduce the study of all CG-modules to only irre-
ducible CG-modules. We shall also prove later that there are finitely many irreducible
CG-modules. In other words, they can be classified for each finite group.
Theorem 2.25 (Maschke’s theorem). Let V be a CG-module and U a CG-submodule
of V . Then there is a CG-submodule W of V such that V = U ⊕ W .

Proof. By Proposition 2.23, the CG-module U ⊥ , with respect to the inner product
h , i defined using the standard inner product ( , ) on V , is the required direct sum
complement of U in V . 

Here is the matrix interpretation of Maschke’s theorem: if we can choose a basis B

of a CG-module V such that each matrix [g]B has the form
 
∗ ∗
0 ∗
then there exists a basis B 0 of V such that each matrix [g]B0 has the form
 
∗ 0
0 ∗
So if ρ is a reducible representation of G of degree n, then we already know that ρ is
equivalent to a representation in which each matrix is of the form
 
[g]C A
[g]B 7→ for all g ∈ G
0 B
where [g]C is the matrix of g on the given submodule, and where A and B are matrices
of some appropriate size, with B square and invertible. Then Maschke’s theorem says
that ρ is equivalent to a representation of the form
 
[g]C 0 0
[g]B0 7→
0 [g]C 00
with [g]C 0 ∈ GLk (C) for some k < n and [g]C 00 ∈ GLn−k (C)
As a consequence of Maschke’s theorem we have the following.
Definition 2.26. Let V be a CG-module. Then V is completely reducible if V is a
direct sum V = ⊕ri=1 Ui , where each Ui is an irreducible CG-submodule of V .
Theorem 2.27. If V is a nonzero CG-module, then V is completely reducible.
Representation Theory of Finite Groups page 28

Proof. We proceed by induction on n = dim V . If n = 1, the result holds, as then V is

Assume that n > 1 and (induction hypothesis) each nonzero CG-module
of dimension less than n is completely reducible. Two things may happen: either
V is , or V is If V is then the theorem holds for V .
Otherwise V is and so V has a CG-submodule, say U . By
Theorem 2.25, there exists a CG-submodule W of V with V = U ⊕ W . Since both
CG-modules U and W have dimension less than n, we have by induction that
U = ⊕ri=1 Ui and W = ⊕si=1 Wi
where each Ui and each Wi is an irreducible CG-submodule of V . It follows that
r
! s
!
M M
V = Ui ⊕ Wi as required.
i=1 i=1

So every nonzero CG-module is a direct sum of irreducible CG-modules. It is in

this sense that the irreducible CG-modules are the building blocks out of which all
the CG-modules are made. So to study all the CG-modules it suffices to study the
irreducible CG-modules.

2.7. Workshop Questions.

Exercise 2.7.1. Let G = hg, h | g 2 = h2 = (gh)2 i ∼
= C2 × C2 and V = hv1 , v2 i a
vector space of dimension 2. Which of the following makes V into a CG-module?
(i) gv1 = v2 , gv2 = v1 , hv1 = −v1 , hv2 = −v2
(ii) gv1 = −v2 , gv2 = v1 , hv1 = −v1 , hv2 = −v2
(iii) gv1 = v2 , gv2 = v1 , hv1 = −v1 , hv2 = v2
(iv) gv1 = −v2 , gv2 = −v1 , hv1 = v1 , hv2 = v2
Exercise 2.7.2. Let G = hg = (1 2 3) , h = (1 2) | g 3 = h2 = 1 , hg = g 2 hi ∼
= S3
2πi
and write ζ = e 3 . Which of the following makes the given vector space V into a
CG-module? And if V is a CG-module, find its kernel.
(i) V = hvi , gv = v , hv = √−v. √
1 3 3 1
(ii) V = hv1 , v2 i , gv1 = v1 + v2 , gv2 = − v1 − v2 and
2 2 2 2
hv1 = v1 , hv2 = v2 .
(iii) V = hv1 , v2 i , gv1 = ζv1 , gv2 = ζ 2 v2 and hv1 = v2 , hv2 = v1 .
(iv) V = hv1 , v2 i , gv1 = ζv1 , gv2 = ζ 2 v2 and hv1 = −v1 , hv2 = −v2 .
(v) V = hv1 , v2 , v3 i , gvj = vj+1 for j = 1, 2 and gv3 = v1 , and hvj = vj for
1 ≤ j ≤ 3.
(vi) V = hv1 , v2 , v3 i , gvj = vj−1 for j = 2, 3 and gv1 = v3 , and hvj = v4−j
for 1 ≤ j ≤ 3.
Exercise 2.7.3. Let G be an abelian group and ρ : G → GLn (C) a representation
of G. Write V for the corresponding CG-module, and
Eλ (g) = {v ∈ V : ρ(g)v = λv} for the λ-eigenspace of ρ(g) in V , for any g ∈ G and
λ ∈ C.
(i) For any g ∈ G, prove that every eigenspace of ρ(g) is a CG-submodule of V .
(ii) V is faithful if and only if there is no non-identity element g ∈ G such that
E1 (g) = V .
Representation Theory of Finite Groups page 29

(iii) Suppose that G is not abelian and that V (i.e. ρ) is faithful. Prove the
following
(a) There exists no basis B of V for which all the matrices [ρ(g)]B are
simultaneously diagonal.
(b) Suppose that [ρ(g)]B is scalar for some g ∈ G. Prove that g ∈ Z(G),
where Z(G) = {z ∈ G | xz = zx , ∀ x ∈ G} is the centre of G.

Exercise 2.7.4. Let G = hg | g 4 = 1i be a cyclic group of order 4 and V the

CG-module
  B = {v1 , v2 , v3 } on which g acts by the matrix
with basis
0 −1 1
[g]B = 1 0 1. Find CG-submodules V1 , V2 , and V3 of V such that
0 0 1
V = V1 ⊕ V2 ⊕ V3 .

Exercise 2.7.5. Let G = hg | g 4 = 1i be a cyclic group of order 4 and V the

3-dimensional CG-module with basis B and given by
 
2 5 5
[g]B = −1 −2 −1 .
0 0 1
Find submodules W1 , W2 , W3 of V such that V = W1 ⊕ W2 ⊕ W3 .

Exercise 2.7.6. Let G = Sn and V the permutation CG-module with natural basis
{v1 , . . . , vn }. Find ker(V ).

Exercise 2.7.7. Let G = hg | g 3 = 1i be a group of order 3 and let V = hv1 , v2 , v3 i

the CG-module defined by
gv1 = v2 , gv2 = −v1 − v2 , and gv3 = v2 + v3
Find all the CG-submodules of V .

Exercise 2.7.8. Let n be an integer n ≥ 2, let G = hg | g n = 1i be a cyclic group of

order n and let V be the permutation CG-module of dimension n with natural basis
{v1 , . . . , vn }. Find all the irreducible CG-submodules of V .

Exercise 2.7.9. Let G = ha, b | a4 = b2 = 1 , ba = a−1 bi be the dihedral group of

order 8 and let V be a CG-module of dimension 4 and basis B = {v1 , v2 , v3 , v4 } on
which a and b act by
   
0 0 0 1 0 1 0 0
1 0 0 0
 and [b]B = 1 0 0 0 .
 
[a]B = 
0 1 0 0 0 0 0 1
0 0 1 0 0 0 1 0
Let U be the subspace of V spanned by the two vectors u1 = v1 + v3 and
u2 = v2 + v4 .

(i) Prove that U is a CG-submodule of V , and decompose U into a direct sum

of irreducible submodules.
(ii) Prove that U ⊥ = hv1 − v3 , v2 − v4 i with respect to the inner product h , i
used in the proof of Maschke’s theorem.

2.8. Homework Questions.

Representation Theory of Finite Groups page 30

Exercise 2.8.1. Let G = ha, b | a4 = b2 = 1 , ba = a−1 bi be the dihedral group of

order 8, and V the CG-module of dimension 2, with basis B = {v1 , v2 }, where the
action of a and b are given by the matrices:
   
i 0 0 1
[a]B = and [b]B =
0 −i 1 0
(i) Calculate (3a − 2a−1 b)(v1 + 5v2 ).
(ii) Prove that V is faithful.
(iii) Let x = a + a3 . Find W = {v ∈ V : xv = 0}. Decide whether W is a
CG-submodule of V and prove your claim.
Exercise 2.8.2. Let G = C8 = hai, and V ={ v1 , v2 } be the CG-module defined by
av1 = 7v1 + 10v2 , av2 = −5v1 − 7v2 .
Find all CG-submodules of V .
Exercise 2.8.3. Let G = C4 = {1, a, a2 , a3 }, and let V = hv1 , v2 , v3 , v4 i be the
CG-module with
av1 = v2 , v2 = v3 , av3 = v4 , av4 = v1 .
(i) Let U1 = hv1 + v2 + v3 + v4 i, so that U1 is a 1-dimensional CG-submodule of
V . Find a CG-submodule W of V such that V = U1 ⊕ W .
(ii) Find three other 1-dimensional CG-submodules U2 ,U3 and U4 of V such
that V = U1 ⊕ U2 ⊕ U3 ⊕ U4 . [Hint: consider vectors of the form
v1 + λv2 + λ2 v3 + λ3 v4 for suitable values of λ.]

Summary.
The notions studied in this section are:
• CG-modules
• CG-submodules
• quotients of CGmodules
• direct sums of CG-modules
• correspondence between modules and representations
• special and modules: regular, trivial, permutation, faithful
• irreducible and completely reducible CG-modules
• Maschke’s theorem

3. CG-homomorphisms
3.1. Definitions.
In this section we introduce and study the notion of homomorphism as it applies to
CG-modules.
Definition 3.1. Let G be a finite group and V, W two CG-modules. A function
ϕ : V → W is a CG-homomorphism if ϕ satisfies the following two conditions:
Representation Theory of Finite Groups page 31

(i) ϕ is a linear transformation, that is,

ϕ(v + v 0 ) = ϕ(v) + ϕ(v 0 ) and ϕ(λv) = λϕ(v)
for all v, v 0 ∈ V and all λ ∈ C.
(ii) ϕ(gv) = gϕ(v) for all g ∈ G and all v ∈ V .
If W = V , then we call ϕ a CG-endomorphism, if ϕ is invertible, we call it a CG-
isomorphism, and if ϕ is an invertible endomorphism, we call it a CG-automorphism.

3.1.1. Crash review of linear transformations.

Given a function ϕ : V → W between vector spaces, then ϕ is a linear transformation
if and only if it satisfies the axiom:
ϕ(av + a0 v 0 ) = aϕ(v) + a0 ϕ(v 0 ) for all a, a0 ∈ C and all v, v 0 ∈ V .

Using linear algebra techniques, it turns out that ϕ is a linear transformation if and
only if ϕ can be written as the n × m matrix [ϕ] with the j-th column being the
column vector of the P coordinates of ϕ(vj ) in the basis {w1 , . . . , wn }. Then, the image
of a vector v = j aj vj ∈ V is the vector of W whose coordinates in the basis
{w1 , . . . , wn } are the coefficients of the column vector resulting from the product
a1
 

[ϕ] · .. . For instance, the function

 .
am
V = hv1 , v2 i , W = hw1 , w2 , w3 i and ϕ : V → W with
ϕ(v1 ) = w1 + 3w2 − w3 , ϕ(v2 ) = −6w3
is a linear transformation, given by the matrix

[ϕ] =

and to calculate for instance the image of v = 2v1 − 5v2 , we multiply

[ϕ] · v =

that is, ϕ(v) = .

Instead, the function ψ : V → W with ψ(v1 ) = w12 and ψ(v2 ) = w1 + w2 + 2w3 is
not a linear transformation because ψ(v1 ) is not a homogeneous linear polynomial.
Below we come back to CG-homomorphisms as matrices in greater detail.

To check that a given linear transformation ϕ : V → W is a CG-homomorphism it

suffices to check that the equality
ϕ(gv) = gϕ(v)
holds for all g in a given set of generators of G and all v in a given basis of V .
Example 3.2. Let V = W = hv1 , v2 i = C2 and ϕ : V → V the linear transformation
defined as follows:
ϕ(v1 ) = v1 + 2v2 and ϕ(v2 ) = −2v1 − v2
Representation Theory of Finite Groups page 32

Let G = hg | g 6 = 1i a cyclic group of order 6 and define a G-action on V as follows:

gv1 = −v2 and gv2 = v1 + v2 . Then V is a CG-module since gv ∈ V for all v ∈ V
and we can check (exercise) that g 6 vj = vj for j = 1, 2. So V is a CG-module for this
G-action.
We show that ϕ is a CG-homomorphism.
Since ϕ is given as a matrix, ϕ must be a linear transformation by basic principles
of linear algebra (cf. MATH220). So all we need to check is the axiom involving the
G-action. That is, we need to prove that ϕ(gvj ) = gϕ(vj ) for j = 1, 2, as g generates
G and v1 , v2 span V . We calculate:
ϕ(gv1 ) =
gϕ(v1 ) =
ϕ(gv2 ) =
gϕ(v2 ) =
and so ϕ(gvj ) = gϕ(vj ) for j = 1, 2 as required.

It is often easier to work with matrices. For this, choose bases B = {e1 , . . . , en } of V
and B 0 = {f1 , . . . , fm } of W . Then ϕ is given by the matrix
A = [ϕ]BB0 = (aij ) ∈ Mm×n (C) (recall the notation [ϕ]domain
range )

in which the j-th column is ϕ(ej ) written as linear combination of the fi ’s, for all
1 ≤ j ≤ n. That is, the coefficients
m
X
aij are defined by ϕ(ej ) = aij fi for all 1 ≤ j ≤ n.
i=1

Then if the set {g1 , . . . , gr } generates G, the linear transformation ϕ given by the
matrix A is a CG-homomorphism if and only if
A[gi ]B = [gi ]B0 A for all 1 ≤ i ≤ r.
Example 3.3. Let us review Example 3.2 in terms of matrices. Using the basis
B = {v1 , v2 } of V as basis for domain and range, the linear transformation ϕ is given
by the matrix [ϕ]B = The matrix of the action of g on V , using the same

basis B of V , is [g]B = We calculate

[ϕ]B [g]B =

[g]B [ϕ]B =

which is indeed equivalent to the above computation in Example 3.2,

ϕ(gv1 ) = gϕ(v1 ) = and
ϕ(gv2 ) = gϕ(v2 ) =

Non-example.
Representation Theory of Finite Groups page 33

Let V = hv1 , v2 i be a 2-dimensional vector space, and G = hg | g 4 = 1i a cyclic group

of order 4. Then V can be equipped with the following G-action
gv1 = −v1 and gv2 = v2
Now let ϕ : V → V be the linear transformation
ϕ(v1 ) = v2 , ϕ(v2 ) = v1 + v2
That is in terms of matrices in the basis B = {v1 , v2 }, we have

[g] = and [ϕ] =

Then ϕ is not a CG-homomorphism because

[ϕ][g] = whereas

[g][ϕ] =

That is, gϕ(v) 6= ϕ(gv) for some v ∈ V , e.g. we can take v = v1 .

For convenience, if the bases of domain and range are “obvious” and no confusion is
possible, we simply write [g], respectively [ϕ] for the matrix of the action of g on V ,
respectively a linear transformation ϕ, as in the Non-example above.
The well-known notions of kernel and image of a linear transformation generalise to
CG-modules in the natural way as follows.
Definition 3.4. Let ϕ : V → W be a CG-homomorphism.
(i) The kernel of ϕ is the subspace
ker(ϕ) = {v ∈ V | ϕ(v) = 0} of V .
(ii) The image of ϕ is the subspace
im(ϕ) = {ϕ(v) | v ∈ V } of W .
Proposition 3.5. Let V and W be CG-modules and let ϕ : V → W be a CG-
homomorphism. The following hold.
(i) ker(ϕ) is a CG-submodule of V .
(ii) im(ϕ) is a CG-submodule of W .
(iii) ϕ is injective if and only if ker(ϕ) = {0}.
(iv) ϕ is surjective if and only if im(ϕ) = W .

Proof. We know from MATH220 that ker(ϕ) is a subspace of V and im(ϕ) a subspace
of W . So to show the first two parts we only need to prove that gv ∈ ker(ϕ) for all
g ∈ G and v ∈ ker(ϕ), and gw ∈ im(ϕ) for all g ∈ G and w ∈ im(ϕ). Pick any
v ∈ ker(ϕ) and g ∈ G. We have
(∗)
ϕ(gv) =
where (∗) holds because ϕ is a CG-homomorphism. Therefore gv ∈ ker(ϕ) as required.
Now pick any w ∈ im(ϕ) and g ∈ G. So by definition of im(ϕ), there exists v ∈ V
with ϕ(v) = w. The equality
gw =
Representation Theory of Finite Groups page 34

holds because ϕ is a CG-homomorphism, and so gw ∈ im(ϕ) as required.

The last two parts have been proved in MATH220 (for vector spaces), so that both
these statements hold.


Observe that if ϕ is a CG-endomorphism, or simply a linear transformation V → V ,

the kernel of ϕ is the eigenspace for the eigenvalue 0.
Example 3.6.
(i) For any CG-modules V, W , the zero map, 0 : V → W , i.e. with 0(v) = 0
and the identity map IdV : V → V , i.e. with IdV (v) = v, for all v ∈ V are
CG-homomorphisms. The zero map is injective if and only if V = {0}, and it
is surjective if and only if W = {0}. The identity map is a CG-automorphism
for any CG-module V .
(ii) Let V = hv1 , v2 i and G = hg | g 2 = 1i ∼
= C2 . Then V is a CG-module for the
G-action
 
0 1
[g] = given in terms of matrices.
1 0

 ϕ : V → V be the CG-homomorphism given by the matrix [ϕ] =

Let
2 −1
. One can check that [ϕ][g] = [g][ϕ] =
−1 2
The kernel of ϕ is the eigenspace for the eigenvalue , i.e. the subspace of V
spanned by the solutions to

Using MATH220 techniques, we work out the reduced-echelon form of [ϕ] to

solve this system of equation. The outcome is
   
2 −1 1 −2
−1 2 0 1
The rank of the reduced-echelon form of [ϕ] being implies that the sys-
tem has , and so we conclude that ker(ϕ) =
and ϕ is It follows that, because dim V is finite and ϕ is a linear
transformation, then ϕ is also
(iii) Let V = hv1 , v2 i and G = hg | g 2 = 1i ∼
= C2 with the same G-action as
 above.

1 1
Let ψ : V → V be the CG-homomorphism given by the matrix [ψ] = .
1 1
One can check that [ψ][g] = [g][ψ] = [ψ].
Proceeding as above, to find the kernel of ψ we calculate the reduced-
echelon form of [ψ], which is

giving ker(ψ) =

In other words, ker(ψ) is the CG-submodule of V spanned by

The image of ψ is the CG-submodule of V spanned by ,
because
 
x
ψ =
y
Representation Theory of Finite Groups page 35

In particular, ψ is neither injective nor surjective.

Example 3.7. Let G = hg = (1 2 3)| g 3 = 1i ≤ S3 be the subgroup of order 3 of
the symmetric group S3 . Let V be the permutation CG-module with natural basis
B = {v1 , v2 , v3 }. Put
u = v1 + v2 + v3 , w1 = v2 − v1 , w2 = v3 − v2 , and then let
U = hui and W = hw1 , w2 i be the CG-submodules with bases {u} and {w1 , w2 },
respectively. Observe that
V =U ⊕W with
Define πU : V → V by
3
! 3
X X
πU λi vi =( λi )u for all λ1 , λ2 , λ3 ∈ C.
i=1 i=1

We prove that πU is a CG-homomorphism, that ker(πU ) = W , and that im(πU ) = U .

Proof that πU is a CG-homomorphism: by definition πU is a linear transformation.
We need to show that πU (gvi ) = gπU (vi ) for i = 1, 2, 3, since g generates G and
v1 , v2 , v3 span V . We calculate

Alternatively, we can use matrices to show that πU is a CG-homomorphism. Hence,

the matrix of πU in the natural basis is

[πU ]B = since

and it is straightforward to check that [πU ]B commutes with [g] = that

is, A[πU ] = [πU ]A.

Proof that ker(πU ) = W and im(πU ) = U : We know that im(πU ) is the subspace of
V spanned by the columns of [πU ], because they encode the images by πU of the basis
vectors of the domain V . Thus we have

im(πU ) =

Finally, ker(πU ) is the eigenspace of [πU ] for the eigenvalue 0, that is,

ker(πU ) =

Note that w1 = ∈ ker(πU ) and w2 = ∈ ker(πU ). Since ker(πU ) is a

subspace of V , we have W ⊆ ker(πU ). Moreover, we know that
dim ker(πU ) = dim V − dim im(πU ) = = dim W
and therefore ker(πU ) = W as required.

Example 3.7 leads to a relationship between CG-homomorphisms and direct sums.

Representation Theory of Finite Groups page 36

Proposition 3.8. Let V be a CG-module with CG-submodules U1 , . . . , Ur such that

V = ⊕ri=1 Ui . For a given v ∈ V , consider the unique expression
r
X
v= ui with each ui ∈ Ui . For 1 ≤ j ≤ r define
i=1

πj : V → V by πj (v) = uj .
Then each πj is a CG-homomorphism.
Definition 3.9. Let V be a CG-module with CG-submodules U1 , . . . , Ur such that
V = ⊕ri=1 Ui . For each 1 ≤ j ≤ r, the CG-homomorphism πj : V → V mapping
πj (v) = uj is called the projection of V onto Uj .
Remark 3.10. These maps πj are called projections for obvious reasons. In linear
algebra, given a vector space V and any subspace U of V , we may choose a basis BV
of V which contains a basis BU of U . Thus, any vector v ∈ V can be written as a sum
v = u + w, where u ∈ U and w lies in a complement of U in V . The projection of V
onto U is the linear transformation πU : V → V which sends each v = u + w ∈ V to
u. (E.g. in the complex plane C2 , we have two obvious projection maps π1 , π2 , where
π1 (z1 , z2 ) = (z1 , 0) and π2 (z1 , z2 ) = (0, z2 ).) If moreover U and W are CG-modules,
Proposition 3.8 asserts that the projection maps are CG-homomorphisms.

Proof. Since each πj is a linear map, the only thing left to prove is that the equality
πj (gv)P= gπj (v) holds for all g ∈P G, all v ∈ V , and all j. Given g ∈ G and
r r
v = u
i=1 i ∈ V , we have gv = i=1 gui . Moreover gui ∈ Ui for all i because
each Ui is a CG-submodule of V . So πj (gv) = proving that πj is a
CG-homomorphism. 

The following example describes the matrices of projection CG-homomorphisms.

Example 3.11. Let V = hv1 , v2 i and G = hg | g 2 = 1i a group of order 2. Consider
the G-action on V defined by
gv1 = v2 and gv2 = v1
Thus V is a CG-module. Put u = v1 + v2 and U = hui. Then U is a CG-submodule
of V (with G-action). Let w = v1 − v2 and W = hwi. Then W is also a
CG-submodule of V on which G acts by gw = . We have V = U ⊕ W , since

v1 = and v2 =
The projections onto U and W can be expressed in terms of matrices, using the basis
B = {u, w} of V . Then

[πU ]B = and [πW ] =

Observe that [πU ]2 = [πU ]. This means that πU is the identity map on U and the zero
map on W . (Similarly for W and πW .)

A CG-homomorphism is called a CG-isomorphism if it is invertible. If there exists a

CG-isomorphism θ : V → W , we say that V and W are isomorphic CG-modules,
and we write V ∼
= W.
Recall that an equivalence relation is a relation which is reflexive, symmetric and
transitive.
Representation Theory of Finite Groups page 37

If θ : V → W and φ : W → U are CG-isomorphisms, then the inverse θ−1 : W → V

and the composition φθ : V → U are isomorphisms too. So, the relation
V and W are isomorphic
is an equivalence relation on the class of CG-modules.
In particular, if θ : V → W is a CG-isomorphism, we may use θ and θ−1 to translate
back and forth between V and W , and show that they share many properties.
Example 3.12. Let θ : V → W be a CG-isomorphism.
(i) dim V = dim W since the set {v1 , . . . , vn } is a basis of V if and only if
{θ(v1 ), . . . , θ(vn )} is a basis of W .
(ii) V is irreducible if and only if W is irreducible since U is a CG-submodule
of V if and only if θ(U ) is a CG-submodule of W .
(iii) V has a trivial CG-submodule if and only if W does too since U is a trivial
CG-submodule of V if and only if θ(U ) is a trivial CG-submodule of W .

We now bring together the notions of isomorphic CG-modules and equivalent repre-
sentations of G for a finite group G.
Theorem 3.13. Let G be a finite group and ρ, σ : G → GLn (C) two representations
of G of degree n. Let V and W be the CG-modules of dimension n afforded by ρ and
σ respectively, say in the bases B = {v1 , . . . , vn } of V and B 0 = {w1 , . . . , wn } of W .
Then V and W are isomorphic if and only if ρ and σ are equivalent, i.e.
V ∼
= W ⇐⇒ ρ ∼ σ

Proof. By definition, the G-action on V and W in the bases B and B 0 respectively is

given as follows: for all 1 ≤ i ≤ n and all g ∈ G,
X X
gvj = ρ(g)(vj ) = aij vj and gwj = σ(g)(wj ) = bij wj
1≤i≤n 1≤i≤n

and these give the coordinates of the j-th columns of the matrices of ρ(g) and σ(g),
expressed in the bases B and B 0 ,
[ρ(g)]B = (aij ) and [σ(g)]B0 = (bij ) for all g ∈ G.
By definition ρ ∼ σ if and only if there exists T ∈ GLn (C) such that, for all g ∈ G,

T −1 [ρ(g)]B T = [σ(g)]B0 ⇐⇒
(1)
⇐⇒

Because T ∈ GLn (C) is invertible, T can be regarded as the matrix of an invertible

linear transformation of the n-dimensional vector spaces V and W . So without loss,
0
we may assume that T is the matrix of ϕ : W → V , i.e. T = [ϕ]BB . The equalities (1)
for all g ∈ G translate in terms of CG-modules as

That is, ϕ is a CG-isomorphism.

Therefore, ρ ∼ σ ⇐⇒ V ∼ = W.

Representation Theory of Finite Groups page 38

Example 3.14. Let G = ha, b | a4 = b2 = 1 , ba = a−1 bi be a dihedral group of order

8, and let ρ, σ : G → GL2 (C) be the resprentations given by the matrices
   
0 1 1 0
[ρ(a)]B = , [ρ(b)]B = and
−1 0 0 −1
   
i 0 0 1
[σ(a)]B0 = , [σ(b)]B0 =
0 −i 1 0
0 2
for some bases B = {v1 , v2 } and B = {w1 , w2 } of C . The modules V and W afforded
by ρ and σ are thus equipped with the G-actions

We want to show that V ∼ = W , or equivalently ρ ∼ σ. That is, we want to find

a b
T ∈ GL2 (C) such that [ρ]B T = T [σ]B0 . Put T = and solve the system of
c d
equations resulting from the product

and

This gives the constraints . That is, there is one free

variable, say a and choosing a value for a determines the values of b, c and d. As we
want T to be invertible, we need a 6= 0. Taking a = 1 gives the solution

T = and we check that [ρ]B T = T [σ]B0

0
For completeness, let us recall that T = [Id]BB is the change of basis matrix (from B 0 ,
spanning W , to B, spanning V ).

We end this section with the version of the isomorphism theorem as it applies to
CG-modules.
Theorem 3.15. Let V and W be CG-modules and ϕ : V → W a CG-homomor-
phism. Then there exists a CG-submodule U of V such that
V = ker(ϕ) ⊕ U and U ∼ = im(ϕ)

Proof. We know that ker(ϕ) is a CG-submodule of V . So Maschke’s theorem 2.25

says that there exists a CG-submodule U of V with V = That is, each
v ∈ V can be written as v = for unique elements k ∈ ker(ϕ) and u ∈ U .
Define the function
ϕ : U → im(ϕ) , ϕ(u) = ϕ(u) for all u ∈ U .
In other words, ϕ is the CG-homomorphism ϕ restricted to the submodule U of V for
the domain. Explicitly, if ϕ
e is the function ϕ with its image restricted to the subspace
im(ϕ), so that ϕ
e is a surjective CG-homomorphism, then we have
ϕ

 inclusion /  inclusion '

U im(ϕ) 
ϕ
/ /
e
V W
7
ϕ
Representation Theory of Finite Groups page 39

Therefore, ϕ is a well-defined function, as each element of the domain (U ) has a

unique image in the range im(ϕ).
We show that ϕ is a CG-isomorphism. We first note that ϕ is a CG-homomorphism:

where (1) holds by definition of ϕ and (2) holds because ϕ is a CG-homomorphism.

(More generally, restricting the domain of a CG-homomorphism gives a CG-homomor-
phism.)
Suppose now that u ∈ ker(ϕ). That is

by definition of a direct sum. So ker(ϕ) = {0} saying that ϕ is injective.

For the surjectivity of ϕ, pick w ∈ im(ϕ). That is, w = ϕ(v) for some v ∈ V . As
recalled above, the direct sum decomposition V = ker(ϕ) ⊕ U says that we can write
v = k + u for some unique elements k ∈ ker(ϕ) and u ∈ U . So

∼
showing that w ∈ im(ϕ). Therefore ϕ : U
= / im(ϕ) is a CG-isomorphism. 

3.2. Schur’s lemma.

Schur’s lemma is a fundamental result in representation theory. Its proof uses ele-
mentary notions of linear algebra.
Definition 3.16. Let V be a vector space and ϕ : V → V a linear transformation
of V . If there exists a nonzero vector v ∈ V and some λ ∈ C such that ϕ(v) = λv,
then λ is called an eigenvalue of ϕ and v an eigenvector. The eigenvalues are the
roots of the characteristic polynomial of ϕ. The characteristic polynomial cV (λ) of ϕ
is the determinant of the matrix [ϕ]B − λIn where B is a basis of V and n = dim(V ).
Also, cV (λ) is independent of the choice of B.

For instance, the identity map and the zero map, respectively
IdV : V → V , IdV (v) = v and 0 : V → V , 0(v) = 0 for all v ∈ V
are linear tranformations of V for any vector space V . Moreover IdV has a unique
eigenvalue, namely 1, and every nonzero v ∈ V is an eigenvector for IdV . Similarly 0
is the unique eigenvalue of 0 and every nonzero vector is an eigenvector.
By the fundamental theorem of algebra, any non constant polynomial with coefficients
in C has a root in C and so any linear transformation of V has an eigenvalue, because
its characteristic polynomial splits, i.e. factorises into a product of linear factors:
f (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 = an (x − α1 ) · · · (x − αn ) ∈ C[x]
where α1 , . . . , αn ∈ C are the roots of f (x), not necessarily all distinct.
Representation Theory of Finite Groups page 40

Theorem 3.17 (Schur’s lemma).

Let V and W be irreducible CG-modules, and ϕ : V → W a CG-homomorphism.
(i) Either ϕ = 0, or ϕ is a CG-isomorphism.
(ii) If W = V , then ϕ is a scalar multiple of IdV .

Before the proof, let us observe that if V is a CG-module and 0 6= λ ∈ C a scalar,

then the function
ϕλ = λ IdV : V → V , ϕλ (v) = λv for all v ∈ V
is a CG-automorphism for any 0 6= λ ∈ C. The question one may ask is: “Suppose that
V is irreducible. Are there any other CG-automorphisms V → V than multiplications
by nonzero scalars?”. Part (ii) of Schur’s lemma answers it: “No, there aren’t”.

Proof. (i) Suppose that ϕ 6= 0, that is, there exists v ∈ V with ϕ(v) 6= 0. Then {0} =
6
im(ϕ) is a CG-submodule of W . Since W is irreducible we must have im(ϕ) = W ,
i.e. ϕ is surjective.
On the other hand, ker(ϕ) 6= V because ϕ 6= 0. Since ker(ϕ) is a CG-submodule
of V and V is irreducible, we must have ker(ϕ) = {0}, i.e. ϕ is injective. So ϕ is
invertible, and therefore ϕ is a CG-isomorphism.
(ii) As recalled above, the linear transformation ϕ has an eigenvalue, say λ ∈ C, and
so For any g ∈ G and v ∈ V , the equality

shows that ϕ−λ IdV is a CG-homomorphism, and so ker(ϕ−λ IdV ) is a CG-submodule

of V . Moreover 0 6= v ∈ ker(ϕ−λ IdV ) says that ker(ϕ−λ IdV ) is a nonzero submodule
of V . Since V is irreducible, we must have ker(ϕ−λ IdV ) = V , i.e.
for all v ∈ V , showing that (ϕ − λ IdV ) is the Equivalently, we have proved
that . 

Drawing upon Lemma 3.17 (ii), we obtain the following very useful criteria for testing
the irreducibility of CG-modules.
Proposition 3.18. If V is a non-zero CG-module such that every CG-homomorphism
from V to V is a scalar multiple of IdV , then V is irreducible.

Note that if V is a CG-module of dimension 1, then V is irreducible, because V is

irreducible as vector space, and any linear transformation V → V is a scalar multiple
of the identity (cf. MATH220). So Proposition 3.18 holds.

Proof. We prove the contrapositive, that is, we suppose that V is reducible, and we
need to show that there exists a CG-endomorphism of V which is not a scalar multiple
of IdV . Since V is reducible, we can choose a proper nonzero CG-submodule U of V .
By Maschke’s theorem 2.25, there is a CG-submodule W of V satisfying V = .
So we can write v = for some unique elements u ∈ U and w ∈ W .
By Proposition 3.8, the projection map πU : V → V onto U , defined by πU (u+w) = u
for all u ∈ U and w ∈ W , is a CG-endomorphism of V , and πU is not a scalar multiple
of IdV . Indeed, πU (u) = u = Id(u) for all u ∈ U , whereas πU (w) = 0 = 0 Id(w) for all
w ∈ W.
Thus we have proved that if V is reducible, then there are CG-endomorphisms of V
which are not a scalar multiple of IdV . 
Representation Theory of Finite Groups page 41

We now give the matrix version of Schur’s lemma 3.17 and Proposition 3.18.
Recall that a square matrix is scalar if it is a multiple of the identity matrix, i.e. of
the form λIn for some λ ∈ C and n ∈ N. For instance,
   
2 0 2 0
is scalar, but is not scalar.
0 2 0 −1
Scalar matrices have the property that they commute with any other square matrix
of same size, that is
(λIn )A = A(λIn ) for all A ∈ Mn (C), all λ ∈ C and all n ∈ N.
Conversely a square matrix which commutes with any matrix of same size must be
a scalar matrix. (Scalar matrices form the centre of the ring Mn (C).) Now, a linear
map is a scalar multiple of the identity if and only if its matrix form is scalar and
this does not depend on the choice of a basis.
Corollary 3.19. Let ρ : G → GLn (C) be a representation of a finite group G. Then
ρ is irreducible if and only if every matrix A ∈ Mn (C) which satisfies
Aρ(g) = ρ(g)A for all g ∈ G is scalar.

Proof. Consider the vector space of column vectors Cn as a CG-module by defining

gv = ρ(g)v for all g ∈ G and all v ∈ Cn . Let A ∈ Mn (C). The linear transformation
v 7→ Av of Cn is a CG-homomorphism if and only if

that is, if and only if

Suppose now that ρ is irreducible. Theorem 3.17 says that A is

Conversely, suppose that ρ is reducible. Proposition 3.18 shows that there exists a
CG-endomorphism ϕ of V which is not a scalar multiple of IdV , that is, A ∈ Mn (C)
is not scalar. 

Corollary 3.19 is an efficient way to decide whether a module is irreducible or not.

Example 3.20.
(i) Let G = hg | g 3 = 1i be a group of order 3 and ρ : G → GL2 (C) the
representation defined by
 
0 −1
ρ(g) =
1 −1

Since the matrix commutes with ρ(g), Corollary 3.19 proves that
ρ is (See also Proposition 3.21 below.)
(ii) Let G = ha, b | a4 = b2 = 1 , ba = a−1 bi be a dihedral group of order 8 and
σ : G → GL2 (C) the representation of G defined by
   
0 1 1 0
σ(a) = and σ(b) =
−1 0 0 −1
Assume that the matrix
 
α β
A= for some α, β, γ, δ ∈ C
γ δ
Representation Theory of Finite Groups page 42

commutes with both σ(a) and σ(b). From the equalities Aσ(b) = σ(b)A and
Aσ(a) = σ(a)A, we obtain and , respectively. So

A=

By Corollary 3.19, we conclude that σ is

3.3. Representations of finite abelian groups.

We start with an easy, but crucial, observation about dimensions of irreducible CG-
modules for a finite abelian group G. (Recall that G is abelian if gh = hg for all
g, h ∈ G, i.e. the multiplication is commutative.)

Proposition 3.21. Let G be an abelian group. Then every irreducible CG-module

has dimension 1.

Proof. Let V be an irreducible CG-module, and take any x ∈ G. Because G is abelian,

we have

and so the linear transformation

ϕx : V → V , ϕx (v) = xv for all v ∈ V

is a CG-homomorphism. By Schur’s lemma 3.17, since V is irreducible ϕx must be a

scalar multiple of the identity map IdV for each x ∈ G, say ϕx = λx Id, which gives
ϕx (v) = λx v ∈ hvi for each v ∈ V and each x ∈ G. Therefore, for any v ∈ V , the
subspace hvi ⊆ V is a CG-submodule of V . Since V is irreducible and v 6= 0, we
must have V = hvi and dim V = 1 as required. 

Example 3.22. Let G = hg | g 4 = 1i be a cyclic group of order 4, and consider the

representation ρ : G → GL2 (C) defined by
 
0 −1
ρ(g) = .
1 0

Call V = hv1 , v2 i the corresponding CG-module, so that and We

know from Proposition 3.21 that V is not irreducible. Indeed, ρ(g) commutes with
any , with a, b ∈ C such that

So, how do we find the irreducible submodules of V ?

The short answer is that we need to find the eigenspaces for the linear transformation
ρ(g). The characteristic polynomial of ρ(g) is

c(λ) =

We then calculate the eigenspaces for both eigenvalues i and −i. These are the
irreducible submodules of V .
Representation Theory of Finite Groups page 43

Then V = Ei ⊕ E−i as vector space, by MATH220. But we also have guj ∈ Ej for
any uj ∈ Ej for j = ±i since E±i are the eigenspaces for the linear transformation
ρ(g) which defines the G-action on V . Therefore E±i are CG-submodules of V .
Since dim E±i = 1, the above is a direct sum decomposition of V into irreducible
submodules.

3.3.1. Putting all abelian pieces together.

Each finite abelian group can be written as:
G = Cn1 × Cn2 × · · · × Cnr
for some integers n1 , . . . , nr . Let ai be a generator for the i-th term Cni . Set
gi = (1, . . . , 1, ai , 1, . . . , 1) with ai in the i-th position.
Then G = hg1 , . . . , gr i with each gi of order ni and gi gj = gj gi for all i, j.
Now let ρ : G → GLn C be an irreducible representation of G. By Proposition 3.21,
we have n = 1, so that each ρ(g) ∈ GL1 (C) ∼
= C× is a nonzero complex number. For
each generator gj of G, put
ρ(gj ) = λj for all 1 ≤ j ≤ r.
n
Since gj has order nj , we must have λj j = 1, which says that λj is a complex nj -th
root of unity. That is, we can write for some integer k with 0 ≤ k < j.
So the values λ1 , . . . , λr determine ρ, because any element g ∈ G is of the form g =
g1d1 · · · grdr , for some integers d1 , . . . , dr , and ρ is a group homomorphism. Explicitly,
ρ(g) =
For convenience, we denote this representation by ρ = ρλ1 ,...,λr .
Theorem 3.23. Let G be a finite abelian group, say G = Cn1 × · · · × Cnr . There
are exactly |G| non-equivalent irreducible representations of G, and any such is of the
form ρλ1 ,...,λr , where λj is an nj -th root of unity.

Proof. The above argument shows that any irreducible representation of G has the
form ρλ1 ,...,λr .
Conversely, if λj is an nj -th root of unity for all 1 ≤ j ≤ r, then the map ρ : G →
GL1 (C) given by

is an irreducible representation of G. Since there are choices for each λj , the

number of representations of the form ρλ1 ,...,λr is
No two are equivalent, because GL1 (C) is abelian, which means that T −1 AT = A for
any 1 × 1 invertible matrices A and T . 

In terms of modules, Theorem 3.23 asserts that for any finite abelian group G as
above, there are exactly |G| distinct isomorphism classes of irreducible CG-modules,
all of dimension 1. Accordingly to the notation in the theorem, we call each of these
Vλ1 ,...,λr . The action of any element g1d1 · · · grdr ∈ G on Vλ1 ,...,λr is as follows:
Representation Theory of Finite Groups page 44
2πi
Example 3.24. Let G = hg | g n = 1i be a cyclic group of order n, and put ζ = e n
for a primitive n-th root of 1 in C. Here primitive means that n is the smallest
positive integer for which ζ n = 1. Theorem 3.23 asserts that there are n irreducible
representations of G, and they are all the ρζ j for 0 ≤ j ≤ n − 1, where

Example 3.25. Let G = hg1 , g2 | g12 = g22 = 1 , g1 g2 = g2 g1 i ∼

= C2 ×C2 . Theorem 3.23
asserts that there are four irreducible CG-modules, say V1 , V2 , V3 and V4 , where each
Vi is a 1-dimensional vector space. We choose bases {vi } of Vi , for all 1 ≤ i ≤ 4, such
that

In other words, the CG-module:

V1 is afforded by
V2 is afforded by
V3 is afforded by
V4 is afforded by

Example 3.26. Let G = hg | g 6 = 1i ∼ = C6 be a cyclic group of order 6. By

Theorem 3.23, there are six non-equivalent irreducible CG-modules, and each has
πi
dimension 1. Explicitly, if we let λ = e 3 , then the six CG-modules

V0 = hv0 i, . . . , V5 = hv5 i

are equipped with the G-action:

G × Vj → Vj ,

Now, let H = hx, y | x2 = y 3 = 1 , xy = yxi ∼

= C2 × C3 ∼
= C6 ∼
= G. The irreducible
2πi
CH-modules are as follows. Let ζ = e and put for all j = 0, 1 and k = 0, 1, 2
3

So in particular, xw11 = and yw11 = while xw02 = and

yw02 =
Since we are taking the same group, cyclic of order 6, we should get the same irre-
ducible modules. So, how do the Vj and Wjk relate to each other?
∼
The answer is as follows. First, observe that G / H via g 7→ xy. Using this
=

2 3
identification and the equalities λ = ζ and λ = −1, we get the correspondence table
(i.e. the isomorphisms, in technical terms):
Representation Theory of Finite Groups page 45

We conclude this section by proving the converse to Proposition 3.21.

Proposition 3.27. If each irreducible CG-module has dimension 1, then G is abelian.

Proof. Let V be a faithful CG-module. By Theorem 2.27, direct consequence to

Maschke’s theorem 2.25, V decomposes as a direct sum
M r
V = Vi where each Vi is an irreducible CG-submodule of V .
i=1

By assumption, dim Vi = 1 for all i. Suppose that {vi } is a basis of Vi for all i, so
that B = {v1 , . . . , vr } is a basis of V . Since each Vi is a CG-submodule of V , we
have gvi ∈ Vi for all g ∈ G. Thus each matrix [g]B is As matrices
(i.e. AB = BA for any diagonal matrices A and B), we have

Since V is faithful, we have gh = hg for all g, h ∈ G, and hence G is abelian. 

3.4. Spaces of CG-homomorphisms.

In this section we study the set of CG-homomorphisms between CG-modules, for
any finite group G. From basic facts studied in linear algebra the set of linear
transformations L(V, W ) for any two vector spaces V and W is also a vector space.
Now, if we replace “vector spaces” by “CG-modules”, and “L(V, W )” by “set of
CG-homomorphisms”, we also obtain a vector space. The addition and scalar mul-
tiplication are defined similarly: for any CG-modules V and W , for any λ ∈ C, and
for any CG-homomorphisms ϕ, ϕ0 : V → W , we define the CG-homomorphisms
(ϕ + ϕ0 ) and (λϕ) : V → W by:
(ϕ + ϕ0 )(v) = ϕ(v) + ϕ0 (v)
(λϕ)(v) = λϕ(v) for all v ∈ V
Proposition 3.28. Let V and W be CG-modules, ϕ, ϕ0 : V → W two CG-
homomorphisms and λ ∈ C. Then the maps (ϕ + ϕ0 ) and (λϕ) : V → W are also
CG-homomorphisms. Therefore the set of CG-homomorphisms from V to W is a
vector space.

Proof. We know that (ϕ + ϕ0 ) and (λϕ) are linear transformations since L(V, W ) is a
vector space. Also, for all v ∈ V ,
(1)
(ϕ + ϕ0 )(v) = ϕ(v) + ϕ0 (v) = ϕ0 (v) + ϕ(v) = (ϕ0 + ϕ)(v)
where (1) holds because W is a vector space, hence an abelian group. So addition
is a commutative binary operation on the set of CG-homomorphisms V → W . The
additive identity element is the zero map 0 : V → W and the additive inverse of ϕ is
the CG-homomorphism
−ϕ : V → W where (−ϕ)(v) = −ϕ(v).
Now, for all g ∈ G and all v ∈ V , we have by definition

(ϕ + ϕ0 )(gv) =

showing that ϕ+ϕ0 is a CG-homomorphism. Therefore the set of CG-homomorphisms

V → W is an abelian group. In addition,
Representation Theory of Finite Groups page 46

(λϕ)(gv) =

showing that λϕ is a CG-homomorphism. So the set of CG-homomorphisms V → W

is a vector space. 

Definition 3.29. Let V and W be CG-modules. The vector space of all CG-
homomorphisms ϕ : V → W is written HomCG (V, W ).

Our focus in this section is on the vector space HomCG (V, W ), which of course depends
on V and W . Thanks to Maschke’s theorem 2.25, the strategy consists in “gluing
together” CG-homomorphisms between irreducible submodules of V and W . Then,
the key argument is given by Schur’s lemma 3.17. In particular, as we shall see below,
the dimension of the Hom-space “counts”, in some sense, the number of isomorphic
irreducible submodules of V and W .
We begin with a consequence of Schur’s lemma (Theorem 3.17), which gives the
dimension of this space if V and W are both irreducible.

Proposition 3.30. If V and W are irreducible CG-modules, then

1 if V ∼

=W
dim(HomCG (V, W )) =
0 if V  W

Proof. The proposition is immediate from Schur’s lemma 3.17. Indeed, since V and W
are irreducible, any CG-homomorphism V → W is: zero if and only if or a
CG-isomorphism if and only if So suppose that V ∼= W and let ϕ : V → W
−1
be a CG-isomorphism. For any φ ∈ HomCG (V, W ) we have ϕ φ ∈ HomCG (V, V ). By
Theorem 3.17 (ii) there exists λ ∈ C such that ϕ−1 φ = Then φ = and
so we have
HomCG (V, W ) =

which is a 1-dimensional vector space. 

From irreducible modules to arbitrary ones, the result about the dimension of the
vector space HomCG (V, W ) uses the complete reducibility of CG-modules and Propo-
sition 3.30. We first prove a useful result.

Proposition 3.31. Let V , V1 , V2 , W , W1 and W2 be CG-modules. The following

hold.

(i) The vector spaces HomCG (V, W1 ⊕ W2 ) and HomCG (V, W1 ) ⊕ HomCG (V, W2 )
are isomorphic. In particular,

dim(HomCG (V, W1 ⊕ W2 )) = dim(HomCG (V, W1 )) + dim(HomCG (V, W2 ))

(ii) The vector spaces HomCG (V1 ⊕ V2 , W ) and HomCG (V1 , W ) ⊕ HomCG (V2 , W )
are isomorphic. In particular,

dim(HomCG (V1 ⊕ V2 , W )) = dim(HomCG (V1 , W )) + dim(HomCG (V2 , W ))

Representation Theory of Finite Groups page 47

Proof. We only prove the first part, as the second is very similar, and so left as an
exercise. Consider the diagram:
π1 , π2 are the projection maps and
: W2T where
ϕ2
σ1 , σ2 the natural inclusions.
σ2 π2
ϕ 
V / W1 ⊕ W2
J
σ1 π1
ϕ1
$
W1
that is, σ1 : W1 → W1 ⊕ W2 , σ1 (w) = for all w ∈ W1 , and similarly
σ2 : W2 → W1 ⊕ W2 , σ2 (w) = for all w ∈ W2 . Note that π1 , π2 , σ1 , σ2 are
CG-homomorphisms. Moreover,
πi σ j = , i, j = 1, 2 and σ1 π1 + σ2 π2 =
where δij = 0 whenever i 6= j and δii = 1. We prove that the following function f is
an invertible linear transformation:
f : HomCG (V, W1 ⊕ W2) −→ HomCG (V, W1 ) ⊕ HomCG (V, W2 )
ϕ : V → W1 ⊕ W2 7−→

For short, put H = HomCG (V, W1 ⊕ W2 ) and Hj = HomCG (V, Wj ) for j = 1, 2. Note
that f is well defined, as composition of CG-homomorphisms is a CG-homomorphism
and domains and ranges match up. Also, it is immediate from the axioms defining
linear transformations that f is a linear transformation, that is,
f (ϕ + ϕ0 ) = f (ϕ) + f (ϕ0 ) and f (λϕ) = λf (ϕ) for all ϕ, ϕ0 ∈ H and all λ ∈ C.
To show that f is bijective, we find its inverse. Put
g : H1 ⊕ H2 −→ H
, where for each v ∈ V ,
(ϕ1 , ϕ2 ) 7−→ σ1 ϕ1 + σ2 ϕ2
(σ1 ϕ1 + σ2 ϕ2 )(v) = . (It may be useful to track the

compositions in the diagram above.) Let ϕ ∈ H. We have for all v ∈ V

gf (ϕ)(v) =
=

=
and similarly,
f g(ϕ1 , ϕ2 )(v) =
=

=
Therefore f is an invertible linear transformation (with inverse g), proving that H ∼
=
H1 ⊕ H2 as vector spaces.

Representation Theory of Finite Groups page 48

Iterating Proposition 3.31 ad libitum yields

Corollary 3.32. Let V1 , . . . , Vr , W1 , . . . Ws be CG-modules. The vector spaces
M r M s Mr Ms


HomCG Vi , Wj and HomCG (Vi , Wj )
i=1 j=1 i=1 j=1

are isomorphic. In particular,

 Mr s
M  Xr X
s


dim HomCG Vi , Wj = dim(HomCG (Vi , Wj ))
i=1 j=1 i=1 j=1

In the case when all Vi and Wj are irreducible, Proposition 3.30 and Corollary 3.32
enable us to calculate dim(HomCG (V, W )) for any CG-modules V and W .
Corollary 3.33. Let V be a CG-module which can be written as a direct sum V =
⊕ri=1 Ui of irreducible CG-submodules Ui , and let W be any irreducible CG-module.
Then
dim(HomCG (V, W )) = dim(HomCG (W, V ))
and this number is equal to the number of terms Ui with Ui ∼
= W.

Proof. By Corollary 3.32, we have

dim HomCG (V, W ) =

dim HomCG (W, V ) =

By Proposition 3.30 we have

dim HomCG (Ui , W ) = dim HomCG (W, Ui ) =

The result follows. 

In the next section, we will return to this result and its applications. We end this
section with a collection of thoroughly worked out examples.
2πi
Example 3.34. Let G = hg | g 3 = 1i be a group of order 3 and put ζ = e 3 ∈ C
for a primitive cube root of 1. (Recall that 1 + ζ + ζ 2 = 0 ans that ζ 2 = ζ −1 .) Let
V = hv1 , v2 , v3 i and W = hw1 , w2 i be the CG-modules for the G-action given by the
matrices  
1 0 0  
ζ 0
[g]V = 0 ζ 0  and [g]W =
0 1
0 0 ζ2
with respect to the given bases {v1 , v2 , v3 } of V and {w1 , w2 } of W . Let Vj = hvj i for
1 ≤ j ≤ 3 and Wk = hwk i for k = 1, 2. Then
V = V1 ⊕ V2 ⊕ V3 and W = W1 ⊕ W2
with V1 and W2 CG-modules and V2 ∼ = W1 the irreducible CG-module on
which g acts by multiplication by . In particular, V1 ∼
= W2 and V2 ∼
= W1 are the only
isomorphic pairs of irreducible submodules of V and W . So Corollary 3.33 implies
that
dim(HomCG (V, W )) = dim(HomCG (W, V )) = dim(HomCG (W, W )) =
and dim(HomCG (V, V )) = .
Representation Theory of Finite Groups page 49

More precisely, we can find a basis of the above vector spaces by “matching up”
isomorphic pairs. For example, we work out a basis of HomCG (V, W ) as follows: we
only have isomorphisms
θ1 θ2
∼ ∼
V1
= / W2 , v1 7→ w2 , and V2
= / W1 , v2 7→ w1
A CG-homomorphism V → W is given by a matrix, where the (k, l) coeffi-
cient represents a CG-homomorphism Vl → Wk . Thus Corollary 3.32 says that to
find a basis of HomCG (V, W ) we “glue” together the bases of the nontrivial spaces
HomCG (Vk , Wl ). In our case, we get the basis

where ϕ1 , ϕ2 essentially are the isomorphisms θ1 : V1 → W2 and θ2 : V2 → W1 .

For HomCG (W, V ), we obtain a basis by taking the matrices of B.
For HomCG (V, V ) since the irreducible submodules are pairwise distinct, any homo-
morphism is given by a matrix, i.e. a linear combination of

and finally,
HomCG (W, W ) =

Next example shows the two different approaches possible when looking for bases of
Hom-spaces.
Example 3.35. Let G = hg | g 2 = 1i, and V, W, X the CG-modules of dimension
2, 1 and 4 respectively, say
V = hv1 , v2 i , W = hwi and X = hx1 , x2 , x3 , x4 i
with G-action
 
  1 0 0 0
0 1 0 −1 0 0
[g]V = , [g]W = (−1) and [g]X =  
1 0 0 0 0 1
0 0 1 0
Find a bases of HomCG (V, W ), of HomCG (V, V ) and of HomCG (X, V ).
Solution. We have two methods. We find bases of HomCG (V, W ) and HomCG (V, V )
using both methods, and use the second method to find a basis of HomCG (X, V ).
method 1: direct calculation
 
a b
Consider first HomCG (V, V ). We want to find all A = such that
c d

which gives and So we have 2 free variables, say c, d. Taking (c, d) =

(1, 0) and (0, 1) gives two elements forming a basis for HomCG (V, V ), namely,
Representation Theory of Finite Groups page 50

Similarly for HomCG (V, W ), we solve the matrix equation

which shows that any such A must be of the form , i.e. a basis of HomCG (V, W )
is the set

method 2: matching up of irreducible submodules

Since G is abelian, V must be the sum of two -dimensional submodules, which are in
fact the of [g]V . The characteristic polynomial is ,
and a routine computation gives
V = E1 ⊕ E−1 where E1 = and E−1 =
Note that E1  E−1 , because they are equipped with different G-actions, i.e. there
is no nonzero 1 × 1 matrix A = (a) such that .
Therefore, Corollary 3.32 shows that

With respect to the basis {u1 , u2 } of V , a basis of HomCG (V, V ) is thus

Similarly for HomCG (V, W ), we note that W ∼ = E−1 via w 7→ u2 . So a basis for
HomCG (V, W ) with respect to the bases {u1 , u2 } of V and {w} of W is the set
Finally, we want to find a basis of HomCG (X, V ). We know that V = E1 ⊕ E−1 .
We find the eigenspaces of [g]X which gives a direct sum decomposition of X into
irreducible submodules. The characteristic polynomial of [g]X is

We calculate the eigenspaces F±1 . These are immediate from the previous computa-
tions:
F1 = and F−1 =
So, the “matching picture” of all CG-homomorphisms X → V is as follows,

indicating that dim HomCG (X, V ) = . In terms of matrices with respect to the bases
of eigenvectors

we we can take the set of matrices (dim V ×dim X = dim(range)×dim(domain))

Representation Theory of Finite Groups page 51

We see from this last example that if one of the spaces is “large”, here dim X = 4,
then it is more convenient to decompose it into a direct sum of irreducible submodules
and then match up the isomorphic irreducible modules.

We end with another example, slightly more complicated because the group is not
abelian and because the modules are not given as direct sum of irreducible submod-
ules. This last example therefore encompasses various difficulties.
Example 3.36. Let G = S3 be the symmetric group of degree 3, and let V and W
be the permutation CG-module and the trivial CG-module respectively. Find bases
of HomCG (V, W ) and of HomCG (W, V ).

Solution. We decompose V and W into a direct sum of irreducible submodules, and

then we use the “matching up method”, i.e. Proposition 3.30 and Schur’s lemma 3.17.
Consider the natural basis B = {v1 , v2 , v3 } of V and let {w} be a basis for W . Recall
that G permutes the vectors in B by acting on the indices. For instance
(1 2)v1 = v2 , (1 2)v2 = v1 and (1 2)v3 = v3
whereas gw = w for all g ∈ G. Put
u0 = v1 + v2 + v3 , u1 = v1 − v2 , u2 = v2 − v3
so that the subspaces
U0 = hu0 i and U1 = hu1 , u2 i of V are CG-submodules of V with V = U0 ⊕ U1
Then the action of G on V in the basis {u0 , u1 , u2 } is given by

[a = (1 2)] = and [b = (1 3)] =

since

Because G = h(1 2), (1 3)i these two matrices determine the G-action on any CG-
module. The block diagonal form of both matrices shows that V = U0 ⊕ U1 . As
dim U0 = 1, we know that U0 is irreducible. We prove that U1 is also irreducible using
Corollary 3.19. Consider the ordered basis {u1 , u2 } of U1 . The matrices of a and b
are respectively
A= ,B =
 
x y
We want to prove that a matrix M = ∈ M2 (C) such that AM = M A and
z t
BM = M B is necessarily scalar. The equations AM = M A and BM = M B give

from which we deduce that

is indeed Therefore U2 is by Corollary 3.19.

Let us now work out the space of CG-homomorphisms V → W . Note that U0 is
a CG-submodule of V and therefore isomorphic to But G does not act
Representation Theory of Finite Groups page 52

trivially on U1 . That is, V has exactly irreducible CG-submodule isomorphic to

W . Therefore, dim(HomCG (V, W )) = dim(HomCG (W, V )) = by Corollary 3.33. To
find bases of HomCG (V, W ) and HomCG (W, V ), we can either use the basis {u0 , u1 , u2 }
of V , or its natural basis. Independently of our choice, any CG-homomorphism is
given by a matrix. To avoid any confusion in the notation, let us separate the
entries in such a matrix by commas, i.e.
write (x, y, z) instead of (x y z) for x, y, z ∈ C.
If we consider {u0 , u1 , u2 } then any CG-homomorphism V → W must send u1 , u2 to ,
and can map u0 to any scalar multiple of This is because (x, y, z) ∈ HomCG (V, W )
if and only if

if and only if and (Note that the matrices of a and b on W are

both ) So a basis for HomCG (V, W ) is

If we take the natural basis of V instead then θ = (x, y z) ∈ HomCG (V, W ) if and
only

So we must have and that is, θ(λ1 v1 + λ2 v2 + λ3 v3 ) =

So is a basis of HomCG (V, W ).
For HomCG (W, V ), we note that the CG-homomorphisms are given by the
matrices of those in HomCG (V, W ), so that a basis is

3.5. Workshop Questions.

Exercise 3.5.1. Let V be a CG-module and a ∈ C. Define ϕ : V → V by
ϕ(v) = av for all v ∈ V .
(i) Prove that ϕ is a CG-homomorphism.
(ii) Prove that ϕ is a CG-automorphism if and only if a 6= 0.
(iii) Give the matrix form of ϕ as linear transformation V → V .
(iv) Prove that ϕφ = φϕ for any CG-homomorphism φ : V → V .
Representation Theory of Finite Groups page 53

Exercise 3.5.2. Let G = hg | g 3 = 1i be a group of order 3 and V, W two

CG-modules of dimension 2 and basis B = {v1 , v2 } and B 0 = {w1 , w2 } respectively.
Suppose that the action of g on V and W is given by the matrices
   
9 −7 −5 −3
[g]B = and [g]B0 =
13 −10 7 4
Find λ, µ ∈ C such that the map
ϕ : V →W given by ϕ(v1 ) = −6w1 + λw2 and ϕ(v2 ) = −3w1 + µw2
is a CG-homomorphism.
Exercise 3.5.3. Let G = S3 be the symmetric group of degree 3. You may assume
without proof that G is the group generated by the two transpositions a = (1 2) and
b = (2 3). Let ρ : G → GL2 (C) be the representation of G given by
   
4 3 1 0
ρ(a) = and ρ(b) =
−5 −4 −3 −1
(i) By finding all the matrices A ∈ GL2 (C) such that Aρ(g) = ρ(g)A for all
g ∈ G determine whether ρ is irreducible.
(ii) Let V be the module afforded by ρ. Find ϕ ∈ EndCG V such that
ϕ(av) = bv for all v ∈ V .
Exercise 3.5.4. Let G = hg | g n i be a cyclic group of order n ≥ 2 and
V = hv1 , . . . , vn i the permutation CG-module. Prove that V is not irreducible using
Schur’s lemma.
Exercise 3.5.5. Let G = hg | g 4 = 1i be a group of order 4 and V, W the
CG-modules with bases {v1 , v2 } and {w1 , w2 } on which G acts by
gv1 = 3v1 + 4v2 , gv2 = −2v1 − 3v2 , gw1 = −3w1 − w2 , gw2 = 8w1 + 3w2
Find a basis for HomCG (V, W ).
Exercise 3.5.6. Let G = hg | g 4 = 1i be a cyclic group of order 4 and V, W the
CG-modules of dimension 2 and 3 with bases {v1 , v2 } and {w1 , w2 , w3 } respectively,
defined by the G-actions:
gv1 = iv1 + (i − 1)v2 , gv2 = −v2 and
gw1 = −w2 , gw2 = w1 , gw3 = −w3
Find a basis for HomCG (V, W ) and for HomCG (W, V ).
Exercise 3.5.7. Let φ : V → V be a CG-endomorphism of a CG-module V and
suppose that φφ = φ, where φφ is the composition of φ with itself. Prove that there
exist CG-submodules U, W of V such that V = U ⊕ W with φ(u) = 0 for each
u ∈ U and φ(w) = w for each w ∈ W .
Exercise 3.5.8. Let G be a finite group and V a CG-module. Put
V0 = {v ∈ V : gv = v , ∀ g ∈ G}
(i) Prove that V0 is a CG-submodule of V .
(ii) Prove that for V = CG we have dim(CG)0 = 1.
(iii) Let W be a CG-module and write
W0 = {w ∈ W : gw = w , ∀ g ∈ G}
Suppose that W is isomorphic to V . Prove that W0 is isomorphic to V0 .
Representation Theory of Finite Groups page 54

Exercise 3.5.9. Let U, V, W be CG-modules and φ1 , φ2 ∈ HomCG (U, V ) and

ψ ∈ HomCG (V, W ).

(i) Prove that the composition ψφ1 is a CG-homomorphism U → W .

(ii) Prove that the function (φ1 + φ2 ) : U → V defined by
(φ1 + φ2 )(u) = φ1 (u) + φ2 (u) for all u ∈ U makes HomCG (U, V ) into an
abelian group.
(iii) Suppose that U = V . Prove that HomCG (U, U ) is a ring for the addition
defined above and the multiplication being the composition of
CG-homomorphisms. Give the multiplicative and the additive identity
elements. Is HomCG (U, U ) commutative as ring?

Exercise 3.5.10. Let G = hg | g 4 = 1i be a cyclic group of order 4 and V a

3-dimensional CG-module with basis B = {v1 , v2 , v3 } and on which G acts by
 
−1 0 1 − i
[g]B = 1 + i i −1 + 3i
0 0 −i

(i) Check that the G-action on V is well-defined, that is, the function g 7→ [g]B
is a group homomorphism G → GL3 (C).
(ii) Decompose V into a direct sum of irreducible submodules.
(iii) Find a basis for EndCG (V ).

3.6. Homework Questions.

Exercise 3.6.1. Let G = hg | g 3 = 1i be a group of order 3 and V, W two

CG-modules of dimension 2 and bases B = {v1 , v2 } and B 0 = {w1 , w2 } respectively.
Suppose that the action of g on V and W is given by the matrices
   
1 −1 4 3
[g]B = and [g]B0 =
3 −2 −7 −5
Find λ, µ ∈ C such that the map
ϕ : V →W given by ϕ(v1 ) = 3w1 + λw2 and ϕ(v2 ) = −w1 + µw2
is a CG-homomorphism.

Exercise 3.6.2. Let G = ha, b | a4 = b2 = 1 , ba = a−1 bi be a dihedral group of

order 8 and consider the representation ρ : G → GL2 (C) of G defined by
   
13 −5 11 −4
ρ(a) = and ρ(b) =
34 −13 30 −11
Is ρ irreducible? Prove your claim.

Exercise 3.6.3. Let G = ha, b | a2 = b2 = 1 , ab = bai ∼

= C2 × C2 . Give a complete
set of non-isomorphic irreducible CG-modules. Hence, give a basis for EndCG (CG).

Exercise 3.6.4. Let G = ha | a4 = 1i be a group of order 4 and V, W the

CG-modules with basis B = {v1 , v2 } and B 0 = {w1 , w2 } on which G acts by
av1 = −v2 , av2 = v1 , aw1 = −3w1 − w2 and aw2 = 8w1 + 3w2 .
Find a basis for HomCG (V, W ).
Representation Theory of Finite Groups page 55

Summary. The notions studied in this section summarise the basic properties of
CG-homomorphisms and some applications.
• CG-homomorphisms and basic properties;
• isomorphism theorem;
• Schur’s lemma and its corollaries;
• classification of the irreducible representations of finite abelian groups;
• vector spaces of CG-homomorphisms.

4. Irreducible CG-modules and the Group Algebra

In this section, we study the group algebra of a finite group G. We will see that
CG is “the most important” CG-module for any finite group G, in the sense that
CG contains all the building blocks of any representation of G in the sense we shall
describe in this section. In group algebra, the word algebra refers to an algebraic
structure which is simultaneously a ring and a vector space.

4.1. Composition factors.

Definition 4.1. Let G be a finite group. The group algebra of G is the set CG of all
elements of the form
X
ag g where ag ∈ C for all ag ∈ C.
g∈G

CG is a unital ring for the addition and multiplication defined as follows:

X X X
ag g + bg g = (ag + bg )g
g∈G g∈G g∈G
X
X
X X X

ag g bg g = (ag bh )(gh) = agh−1 bh g
g∈G g∈G g,h∈G g∈G h∈G
X X
for all ag g , bg g ∈ CG.
g∈G g∈G

In particular, X X
0CG = 0g and 1CG = δ1,g g
g∈G g∈G
where δ1,g = 1 for g = 1 and δ1,g = 0 otherwise.
Moreover CG is also a C-vector space of dimension |G|. The natural basis of CG is
the set G = {g | g ∈ G}.
Example 4.2. Let G = hg | g 3 = 1i. Then CG is the vector space with basis
G = {1, g, g 2 } and consists of all the elements of the form
a1 1 + ag g + ag 2 g 2 for a1 , ag , ag2 ∈ C.
In the ring CG, we calculate for instance for x = 1 + 2g and y = −1 + g + g 2 ,
Representation Theory of Finite Groups page 56

Given a finite group G and a CG-module V , we know by Theorem 2.27 that V is

completely reducible, that is, we can write V = ⊕ri=1 Ui for some irreducible CG-
submodules Ui of V . But such a direct sum decomposition is not unique as the
following example shows.
Example 4.3. Let G be any finite group, and let V = hv1 , v2 i be a 2-dimensional
CG-module, with gv = v for all g ∈ G and all v ∈ V . Put u = v1 + v2 . Then gu = u
for all g ∈ G, saying that hui is an irreducible CG-submodule of V . On the other
hand, it is immediate that hvi i are irreducible CG-submodules of V too. Moreover
V =
The three CG-submodules hv1 i , hv2 i , hui of V are all distinct, even though they
are all isomorphic to the CG-module.

Therefore we cannot expect a canonical decomposition of a CG-module V into a

direct sum of irreducible CG-submodules. But there is a way around this apparent
difficulty in the following sense.
r
M
Proposition 4.4. Let V be a CG-module and put V = Ui where each Ui is an
i=1
irreducible CG-submodule of V .
(i) If U is any irreducible CG-submodule of V , then U ∼
= Ui for some 1 ≤ i ≤ r.
s
(ii) If V ∼
M
= Wj where each Wj is an irreducible CG-module, then r = s and
j=1
we can permute the indices so that Ui ∼
= Wi for all 1 ≤ i ≤ r.

Proof. (i) Let u be a nonzero vector in U . By definition of the direct sum, we can
write for unique vectors ui ∈ Ui for all i. Since u 6= 0, we know
that there is at least one i for which Consider the projection πi : U → U
defined in Definition 3.9, i.e. πi (v1 + · · · + vr ) = vi for all vj ∈ Uj , for 1 ≤ j ≤ r.
In particular, im(πi ) is a submodule of the irreducible CG-module U and
isomorphic to a submodule of the irreducible CG-module The only
possibility is therefore im(πi ) =
(ii) We proceed by induction on the number r of summands U1 , . . . , Ur of V . If r = 1,
then V is so that V = and the assertion holds. Suppose
that r ≥ 2. The induction hypothesis states that if a CG-module X can be written
p q
∼
M M
as direct sums X = Xi = Yj of irreducible CG-submodules Xi and Yj with
i=1 j=1
p < r, then p = q and Xi ∼= Yi for all i up to a permutation of the indices. Let
r s
Ui and suppose that there is an isomorphism V ∼
M M
V = = Wj for some positive
i=1 j=1
integer s and irreducible CG-modules Wj . Note that s ≥ 2 because we assume that
V is By part (i), we have an isomorphism for some 1 ≤ k ≤ r.
Without loss, we re-label the indices of the Ui ’s so that Let

By induction we have
Representation Theory of Finite Groups page 57

Therefore r = s and we can permute the indices so that Ui ∼

= Wi for all 1 ≤ i ≤ r. 
r
M
In other words, a decomposition of V = Ur into a direct sum of irreducible CG-
i=1
submodules is not unique, strictly speaking, but it is unique up to isomorphism. This
property is often referred to as the unique decomposition property of CG-modules,
and it motivates the following definition.
Definition 4.5. Let V be a CG-module and U an irreducible CG-module. We say
that U is a composition factor of V if V has a CG-submodule isomorphic to U .
We say that V and W have a common composition factor if there is an irreducible
CG-module which is a composition factor of both V and W .

Proposition 4.4 shows that every CG-module is characterised by its composition fac-
M r
tors, that is, if V is a CG-module, then we can write V = Ui with each Ui an
i=1
irreducible CG-module, and each Ui is unique up to isomorphism. Moreover the mul-
tiplicity of an irreducible summand isomorphic to Ui is independent of the direct sum
decomposition of V . In other words, if
V = U1 ⊕ · · · ⊕ Ur = W1 ⊕ · · · ⊕ Wr
then for any 1 ≤ j ≤ r
{i | 1 ≤ i ≤ r , Ui ∼
= Uj } = {i | 1 ≤ i ≤ r , Wi ∼
= Uj }
Example 4.6. Let G = hg | g 3 = 1i be a group of order 3, and let V = hv1 , v2 i and
W = hw1 , w2 , w3 i be the CG-modules with G-action
V : gv1 = v2 gv2 = −v1 − v2
W : gw1 = w2 gw2 = w3 gw3 = w1
That is, as matrices expressed in the bases B = {v1 , v2 } of V and B 0 = {w1 , w2 , w3 }
of W we have

[g]B = and [g]B0 =

2πi
Let ζ = e 3 be a complex primitive cube root of unity. Recall that ζ 3 = 1 6= ζ so that
ζ satisfies the equality ζ 2 + ζ + 1 = 0. This is because ζ is a root of the polynomial
t3 − 1 ∈ C[t] , which factorises as (t − 1)(t2 + t + 1)
Since ζ 6= 1 we must have that ζ is a root of t2 + t + 1.
Let u = v1 − ζv2 ∈ V and x = w1 + ζ 2 w2 + ζw3 ∈ W . Put U = hui and X = hxi. We
calculate
gu =
gx =
saying that u and x are of [g]B and [g]B0 respectively and both for
the eigenvalue ζ. So U and X are isomorphic CG-submodules of V and W . More
precisely, the linear transformation
ϕ : U →X defined by ϕ(u) = x is a CG-isomorphism.
Since U and X have dimension they are both Therefore V and W
have a common composition factor.
Representation Theory of Finite Groups page 58

Using Schur’s lemma 3.17, we can detect common composition factors in two given
modules using homomorphisms.
Proposition 4.7. Let V and W be CG-modules. Then, V and W have a common
composition factor if and only if HomCG (V, W ) 6= {0}.
r
M s
M
Proof. Write V = Vi and W = Wj with each Vi and each Wj irreducible. By
i=1 j=1
Corollary 3.32,

Moreover, by Proposition 3.30 we have

So HomCG (V, W ) 6= {0} if and only if some Vi is isomorphic to some Wj . 

Example 4.8. Carrying on with Example 4.6, let C = {u, u0 } and C 0 = {x, x0 , x00 }
be bases of V and W formed by eigenvectors of [g]B and [g]B0 respectively. Say,
u0 = ζv1 −v2 , so that gu0 = ζ 2 u0 , and x0 = w1 +w2 +w3 , spanning a trivial submodule,
and x00 = w1 + ζw2 + ζ 2 w3 , giving gx00 = ζ 2 x00 . Then, any CG-homomorphism
ϕ : V → W is a linear combination of

[ϕ1 ]CC 0 = and of [ϕ2 ]CC 0 =

Explicitly, any ψ ∈ HomCG (V, W ) has the form ψ = that is,

ψ(λu + λ0 u0 ) = for a, b, λ, λ0 ∈ C.
In particular dim HomCG (V, W ) = which proves that V and W have common
composition factors counted with multiplicities.

4.2. Group algebra as CG-module.

Definition 4.9. Let G be a finite group. The group algebra of G is the set CG of all
elements of the form
X
ag g where ag ∈ C for all ag ∈ C.
g∈G

CG is a unital ring for the addition and multiplication defined as follows:

X X X
ag g + bg g = (ag + bg )g
g∈G g∈G g∈G
X
X
X X X

ag g bg g = (ag bh )(gh) = agh−1 bh g
g∈G g∈G g,h∈G g∈G h∈G
X X
for all ag g , bg g ∈ CG.
g∈G g∈G

In particular, X X
0CG = 0g and 1CG = δ1,g g
g∈G g∈G
where δ1,g = 1 for g = 1 and δ1,g = 0 otherwise.
Representation Theory of Finite Groups page 59

Moreover CG is also a C-vector space of dimension |G|. The natural basis of CG is

the set G = {g | g ∈ G}.
Example 4.10. Let G = hg | g 3 = 1i. Then CG is the vector space with basis
G = {1, g, g 2 } and consists of all the elements of the form
a1 1 + ag g + ag 2 g 2 for a1 , ag , ag2 ∈ C.
In the ring CG, we calculate for instance for x = 1 + 2g and y = −1 + g + g 2 ,

Given any finite group G, the regular CG-module is the vector space CG with basis
{G} and G-action induced by left multiplication in G, that is, for each g ∈ G left
multiplication by g is a linear transformation of CG, and so gives a well-defined
G-action:
X X X
µg : CG → CG , µg ( ah h) = ah gh = ag− 1h h
h∈G h∈G h∈G

The corresponding group representation is called the regular representation of G. In

particular the regular CG-module has dimension |G|. The objective of this section is
to investigate CG in-depth, and the reason is that CG is the most natural represen-
tation of G. It is faithful and a permutation CG-module.
Example 4.11. For any finite group G, the regular CG-module is a permutation
CG-module with natural basis {G}. That is, we choose an ordering of the elements
of G, say
G = {g1 = 1, g2 , . . . , gn } where |G| = n,
and then, for each 1 ≤ j ≤ n the matrix [gj ]{G} ∈ GLn (C) is the permutation matrix
whose nonzero entry in the column i is the coefficient (i, kj,i ), where kj,i is defined by
the rule
gj gi = gkj,i

Example 4.12. Let G = hg, h | g 2 = h2 = (gh)2 = 1i ∼ = C2 × C2 be a Klein

four group. Then CG is a vector space of dimension and basis In
particular, a vector x ∈ CG can be written as for uniquely
defined µj ∈ C. The G-action on CG is determined by the action of g and h on the
basis vectors, so eight multiplications G × CG → CG:

Or equivalently in terms of matrices

[g] = and [h] =

Note that if we identify G with the subgroup

G̃ = of S4

then CG is the permutation module for G̃.

Representation Theory of Finite Groups page 60

Example 4.13. Let G = hg | g 3 = 1i be a group of order 3. The elements of CG

have the form λ1 1 + λ2 g + λ3 g 2 with each λi ∈ C. We have
1(λ1 1 + λ2 g + λ3 g 2 ) =

g(λ1 1 + λ2 g + λ3 g 2 ) =

g 2 (λ1 1 + λ2 g + λ3 g 2 ) =
Taking matrices with respect to the natural basis G of CG gives the regular repre-
sentation of G. Note that all the matrices we get are permutation matrices

1 7→ g 7→ and g 2 7→

4.3. The decomposition of CG.

Let G be a finite group and CG the regular CG-module. Let V be any CG-module.
Suppose that we want to find HomCG (CG, V ). By definition of a CG-homomorphism,
ϕ ∈ HomCG (CG, V ) if and only if ϕ is a linear transformation CG → V , regarded as
vector spaces, and for all g ∈ G we have
X
ϕ(gα) = gϕ(α) for all α = ah h
h∈G

It follows that
ϕ(g) =
=
In other words,
ϕ is entirely determined by

More generally:
Definition 4.14. Let G be a finite group and V a CG-module. Then V is cyclic if
there exists v ∈ V such that
V = hvi = {αv | α ∈ CG} = CG · v
Any such v is called a generator of V .

So the above shows that CG is a cyclic CG-module, with generator 1CG . But also the
zero CG-module {0} is cyclic, “generated” by the zero vector 0.
For any cyclic CG-module, say V = hvi, then any CG-homomorphism ϕ : V → W
with domain V is determined by ϕ(v) because
V = hvi = CGhvi = {αv | α ∈ CG}
so that for any CG-homomorphism ϕ : V → W and every u ∈ V , there exists
αu ∈ CG such that u = αu v and by definition of CG-homomorphism we calculate
ϕ(u) =
that is, the image ϕ(v) of v determines the image ϕ(u) of u for all u ∈ V .
The property of a CG-homomorphism being determined by a single value is valid only
for cyclic modules hvi. In particular, as vector space a cyclic module has dimension
at most |G| and any irreducible module is cyclic (cf. exercises) but the converse is
false in general because CG = h1CG i is cyclic but not irreducible in general.
Representation Theory of Finite Groups page 61

The conclusion we draw from the above discussion is the following result.
Proposition 4.15. Let V be a CG-module. Then HomCG (CG, V ) is isomorphic to
V as vector space.

Proof. Define
f : HomCG (CG, V ) −→ V , f (ϕ) = ϕ(1) (where 1 = 1CG )
We have seen above that ϕ(1) ∈ V determines ϕ, in the sense that ϕ(α) =
for all α ∈ CG. By definition, for any ϕ, ψ ∈ HomCG (CG, V ) and any λ ∈ C, we have
f (ϕ + ψ) =
f (λϕ) =
Therefore f is a linear transformation. To prove that f is bijective, define
 
CG → V
s : V −→ HomCG (CG, V ) , s(v) =
α 7→ αv
We first need to check that s is well defined, i.e. that s(v) is indeed a CG-homomor-
phism. This is immediate from the module axioms, indicated as (∗) over the equalities
below: for any α, β ∈ CG, any λ and any g ∈ G, we have


s(v) (α + β) =


s(v) (λα) =
(saying that s(v) is a linear transformation)



g s(v) (α) =
Similarly, we check that s is a linear transformation, i.e. s(λv + µw) = λs(v) + µs(w),
for all λ, µ ∈ C and all v, w ∈ V , and this too is immediate from the module axioms
(left as exercise). To show that s = f −1 we calculate

f s(v) =

sf (ϕ) =

Therefore f and s are inverse of each other and since they are linear transformations,
we conclude that f : HomCG (CG, V ) → V is an isomorphism of vector spaces. 

The main property of the regular CG-module of any finite group is the following.
Theorem 4.16. Let G be a finite group. Any irreducible CG-module is a composition
factor of the regular CG-module.

Proof. Write the regular CG-module as a direct sum of irreducible CG-submodules,

Mr
say CG = Ui . Let W be an irreducible CG-module and choose a nonzero vector
i=1
w ∈ W . To prove that W is a composition factor of CG, we want to show that
W ∼= Ui for some i, or equivalently, that there is a nonzero CG-homomorphism
CG → W . Define (cf Proof of Proposition 4.15)
ϕ : CG → W , for all α ∈ CG.
Representation Theory of Finite Groups page 62

Then, ϕ is a linear transformation, by definition of CG as CG-module and the module

axioms. Moreover, for all g ∈ G and all α ∈ CG we have
ϕ(gα) =
By definition ϕ(1) = w 6= 0, so that ϕ is a nonzero CG-homomorphism. Therefore
im(ϕ) is a nonzero CG-submodule of W , and since W is irreducible, we must have
im(ϕ) = W . Corollary 3.32 gives the isomorphism of vector spaces
r
HomCG (CG, W ) ∼
M
= HomCG (Ui , W )
i=1

where the right hand side is a nonzero vector space. So Corollary 3.33 says that there
is an index i such that W ∼= Ui . Therefore W is a composition factor of CG. 

Because CG is a finite dimensional vector space, CG has finitely many submodules.

Corollary 4.17. Up to isomorphism, there are finitely many irreducible CG-modules.

The overarching conclusion of our findings is that to get a complete list of all the
representations of a finite group G, it suffices to find all the irreducible CG-modules.
Moreover, all the irreducible CG-modules are composition factors of the regular CG-
module. In practice this is hardly doable for an arbitrary group since we want to find
a basis of CG such that each square |G| × |G| matrix giving the action of a group
element has a block diagonal form. The case of finite abelian groups is manageable,
because we know that square matrices can be put in diagonal form simultaneously.
But difficulties arise already for the simplest nonabelian group, S3 ∼ = D6 .
Example 4.18.

(i) Let G = hg | g 3 = 1i. Then dim CG = 3, with basis {1, g, g 2 }. The matrix of
g ∈ G is

[g] = with characteristic polynomial

cg (λ) =
We calculate the eigenspaces (calculs abbreviated, gaps to fill in exercise):
 
−1 0 1
E1 :  1 −1 0 E1 =
0 0 0
 
−ζ 0 1
Eζ :  1 −ζ 0 Eζ =
0 0 0
 2 
−ζ 0 1
Eζ 2 :  1 −ζ 2 0 Eζ 2 =
0 0 0
Then CG = E1 ⊕ Eζ ⊕ Eζ 2 and each irreducible CG-module must be isomor-
phic to one of the three composition factors E∗ for ∗ ∈ {1, ζ, ζ 2 }. Then any
CG-module is isomorphic to a direct sum of composition factors.
Representation Theory of Finite Groups page 63

(ii) Let G = hg, h | g 2 = h2 = 1 , gh = hgi ∼

= C2 × C2 . Then (cf. Example 4.12
too), we know that CG is the sum of four 1-dimensional CG-submodules
because G is abelian. A routine computation to find the eigenspaces (left as
exercise) yields the submodules:
V1 =
V2 =
V3 =
V4 =
CG = V1 ⊕ V2 ⊕ V3 ⊕ V4
(iii) Let G = ha, b | a3 = b2 = 1 , ba = a−1 bi be a dihedral group of order 6. Let
2πi
ζ = e 3 and
v 0 = 1 + a + a2 u0 = b + ba + ba2
v1 = 1 + ζ 2 a + ζa2 u1 = b + ζba + ζ 2 ba2
v2 = 1 + ζa + ζ 2 a2 u2 = b + ζ 2 ba + ζba2
We calculate avi = and aui = for i = 0, 1, 2. Write A = hai
and B = hbi for the cyclic subgroups of G generated by a and b respectively.
Recall that A is a normal subgroup of G, but B is not normal in G. We
observe from the above computations that each vi and each ui spans a CA-
submodule of CG. Indeed, we get for i = 1 for instance:

au1 =
=
Therefore the 1-dimensional spaces hvi i and hui i are CA-modules. Moreover
we calculate
bv0 =
Therefore the 2-dimensional spaces
U0 = hv0 , u0 i U3 = hv1 , u2 i and U4 = hv2 , u1 i
are still CA-modules and also CB-modules. In other words U0 , U3 and U4 are
all three CG-submodules of CG because they are simultaneously CA- and
CB-modules and G is generated by a and b.
Next we prove that U3 = hv1 , u2 i is irreducible using Corollary 3.19, i.e. by
finding all the matrices which commute with those of a and b. Let
 
x y
C= ∈ GL2 (C) and suppose that
z t
       
ζ 0 ζ 0 0 1 0 1
C = C and that C = C
0 ζ2 0 ζ2 1 0 1 0
   
ζ 0 0 1
where [a]{v1 ,u2 } = and [b]{v1 ,u2 } =
0 ζ2 1 0
Solving for x, y, z, t these equations gives respectively
Representation Theory of Finite Groups page 64

so that and that is, C must be a matrix of the form

C= . Therefore U3 is irreducible. Similarly we get that U4 is
an irreducible CG-module. However U0 is reducible. Indeed, the subspaces
U1 = and U2 =
are CG-submodules of U0 . We note that U1 is a trivial CG-submodule of CG.
With U2 we calculate a(v0 − u0 ) = and b(v0 − u0 ) =
Now we prove that the set X = {v0 , v1 , v2 , u0 , u1 , u2 } is a basis of CG by
calculating the determinant of the matrix whose columns are the six vectors
of X written in the basis B = {1, a, a2 , b, ba, ba2 } of CG. We calculate
1 1 1
1 ζ2 ζ 0
1 ζ ζ2
=
1 1 1
0 1 ζ ζ2
1 ζ2 ζ
where in the first equality we added the second and third rows to the first
row of each 3 × 3 diagonal submatrix (and used that 1 + ζ + ζ 2 = 0. It
follows that the subspace of CG spanned by the columns has dimension 6,
and so {v0 , v1 , v2 , u0 , u1 , u2 } is a basis of CG. Then we have a direct sum
decomposition
CG = U1 ⊕ U2 ⊕ U3 ⊕ U4 with each Ui irreducible.
Note that U1 is a CG-module and that U1 6∼ = U2 because b(v0 − u0 ) =
∼
On the other hand, U3 = U4 because there is a CG-isomorphism
sending v1 to u1 and u2 to v2 . By Theorem 4.16 there are exactly three non-
isomorphic irreducible CG-modules, namely U1 ,U2 and U3 . Correspondingly,
every irreducible representation of G is equivalent to ρ1 ,ρ2 or ρ3 where
ρ1 (bi aj ) =
ρ2 (bi aj ) =
ρ3 (bi aj ) =

So far, we have proved that every irreducible CG-module occurs as composition factor
of the regular CG-module. The above examples show that some irreducible modules
appear more than once in CG. Our next objective is to determine how often each
irreducible CG-module occurs as composition factor of CG.
r
M
Theorem 4.19. Suppose that CG = Ui is a decomposition of CG into a direct
i=1
sum of irreducible CG-submodules. Let U be any irreducible CG-module. Then the
number of terms Ui isomorphic to U is equal to dim U .

Proof. Proposition 4.15 states that dim U = dim(HomCG (CG, U )). Corollary 3.33
shows that dim(HomCG (CG, U )) is equal to the number of Ui ’s with Ui ∼
= U. 
Example 4.20. In each of the three regular modules in Example 4.18, the theorem
holds.
Representation Theory of Finite Groups page 65

Our last main result establishes a relation between the dimensions of the irreducible
CG-modules of a finite group G and the order of G. Corollary 4.17 asserts that
any irreducible CG-module is a composition factor of the regular CG-module CG,
so that we obtain an exhaustive finite list of all the isomorphism types of irreducible
CG-modules by decomposing the regular CG-module.
Definition 4.21. Let V1 , . . . , Vk be irreducible CG-modules. Assume that no two are
isomorphic and that any irreducible CG-module is isomorphic to some Vi . Then, we
say that {V1 , . . . , Vk } form a complete set of nonisomorphic irreducible CG-modules.
Example 4.22. In example 4.18, a complete set of irreducible CG-modules for the
dihedral group G ∼
= D6 is {U1 , U2 , U3 }. Indeed, since U3 ∼
= U4 , we can write
CG ∼ = U1 ⊕ U2 ⊕ U3 ⊕ U3
Theorem 4.23. Suppose that V1 , . . . , Vk form a complete set of nonisomorphic irre-
ducible CG-modules. Then
X k
dim(Vi )2 = |G|
i=1
Moreover we may assume that V1 is the trivial CG-module of dimension 1.

Proof. We decompose the regular CG-module CG = ⊕rj=1 Uj as a direct sum of irre-

ducible CG-modules and put dim(Vi ) = di for all 1 ≤ i ≤ k. Theorem 4.19 says that
for each i the number of terms Uj isomorphic to Vi is equal to di . So the dimension
of CG is equal to the sum d1 + · · · + dr with di = dim(Vi ) and di appearing di -times
for each i. That is,

|G| =

Finally, any group has a trivial representation ρ : G → C× where ρ(g) = 1 for all
g ∈ G. That is, each group has a trivial CG-module. Since G is finite, we can let
* +
X
V1 = v1 = g ⊆ CG be the trivial CG-submodule of CG of dimension 1.
g∈G


Example 4.24.
(i) Let G be a finite abelian group. We have seen in Theorem 3.23 that G has
non-equivalent representations, all of degree That is, in terms of CG-
modules, a complete set of nonisomorphic CG-modules has irreducible
CG-modules, all of dimension
|G| =

(ii) Let G be a dihedral group of order 6 as in Example 4.18. We have seen that
CG ∼= U1 ⊕ U2 ⊕ U3 ⊕ U3
is the direct sum of four irreducible CG-modules: the trivial U1 of dimension
, a nontrivial -dimensional CG-module U2 and two isomorphic copies of a
CG-module U2 of dimension Thus
6 = |G| =
Representation Theory of Finite Groups page 66

Nonisomorphic groups can have the same list of dimensions of a complete set of their
nonisomorphic irreducible CG-modules. In particular, any two abelian groups of same
order will have the same list of dimensions of irreducible modules, namely one.
Now given a nonabelian finite group G, we know that there must be at least one
irreducible module of dimension one. But how many 1-dimensional CG-modules are
there in total? Can we find this number, or bound it, using the group structure of
G? The answer is positive and provided by the next result.
Proposition 4.25. Let G be a finite group and
[G, G] = h[g, h] = ghg −1 h−1 | g, h ∈ Gi
the commutator, or derived, subgroup of G. There are exactly |G/[G, G]| irreducible
CG-modules of dimension 1. Equivalently there are exactly |G/[G, G]| irreducible
representations G → C× of G of degree 1.

Recall that [G, G] is the smallest normal subgroup N of G such that the factor group
G/N is abelian.
Example 4.26. Let G = Sn be the symmetric group of degree n (and order n!),
for some n ≥ 3. Then [G, G] = An is the alternating group of degree n, i.e. the
subgroup of Sn formed by all the even permutations on n letters. Since Sn /An ∼ = C2 ,
any symmetric group of degree at least 3 has exactly two 1-dimensional CG-modules.
One of these is the trivial module, and the other is the signature, i.e.

× 1 if σ ∈ An is even
ε : Sn → C , ε(σ) =
−1 if σ ∈
/ An is odd
(Compare with the dihedral group in Example 4.18 and recall that D6 ∼ = S3 .)

4.4. Workshop Questions.

Exercise 4.4.1. Let G = hg | g 4 = 1i be a cyclic group of order 4 and V, W the
CG-modules with bases B = {v1 , v2 } and B 0 = {w1 , w2 } on which G acts by
gv1 = −v2 , gv2 = v1 , gw1 = −3w1 − w2 and gw2 = 8w1 + 3w2
(i) Find a basis for HomCG (V, W ).
(ii) Find all common composition factors of V and W , if any.
Exercise 4.4.2. Let n be an integer n ≥ 2, and G = hg | g n = 1i be a cyclic group
of order n. Give a complete set of nonisomorphic irreducible CG-modules.
Exercise 4.4.3. Let G be a group of order 12 and {V1 , . . . , Vk } be a complete set of
nonisomorphic irreducible CG-modules. List all the possible integers k and hence all
the sequences of dimensions
dim V1 , dim V2 , . . . , dim Vk .
Exercise 4.4.4. Let G = hg, h | g 3 = h3 = 1 , gh = hgi ∼
= C3 × C3 . Give a
complete set of nonisomorphic irreducible CG-modules. Hence, give a basis for
HomCG (CG, CG).
Exercise 4.4.5. Let G be a finite group. Find a basis of the vector space
HomCG (CG, CG).
Exercise 4.4.6. Let G be the symmetric group S3 . Let V = hu, vi where
2πi
u = 1 + ξ 2 (1 2 3) + ξ(1 3 2) and v = (1 2) + ξ 2 (2 3) + ξ(1 3) with ξ = e 3

Recall that ξ is a primitive cube root of 1 in C, so that ξ 3 = 1 andξ 2 + ξ + 1 = 0.

Representation Theory of Finite Groups page 67

(i) Prove that V is a CG-submodule of CG and that V is irreducible.

(ii) Give a complete
Pset of nonisomorphic irreducible CG-submodules of CG.
(Hint: let s = σ sgn(σ)σ ∈ CG, where sgn(σ) is the signature of σ.)
(iii) For each irreducible submodule U of CG find dim(HomCG (CG, U )).
Exercise 4.4.7. Let U1 , . . . , Uk be a complete set of nonisomorphic irreducible
CG-modules for a finite group G. Assume that V, W are CG-modules with
dim(HomCG (V, Ui )) = di and
dim(HomCG (Ui , W )) = ei for all 1 ≤ i ≤ k
X
Prove that dim(HomCG (V, W )) = di e i .
1≤i≤k

Exercise 4.4.8. Let n be an integer n ≥ 2, and G = hg | g n = 1i be a cyclic group

of order n. Decompose CG into a direct sum of irreducible submodules.
Exercise 4.4.9. Let G = ha, b | a4 = b2 = 1 , ba = a−1 bi be a dihedral group of
order 8. Decompose CG into a direct sum of irreducible submodules.

4.5. Homework Questions.

Exercise 4.5.1. Let G = hg | g 4 = 1i be a cyclic group of order 4 and V, W the
CG-modules with bases B = {v1 , v2 } and B 0 = {w1 , w2 , w3 } on which G acts by
gv1 = v2 , gv2 = −v1 ,
gw1 = 17w1 − 10w2 , gw2 = 29w1 − 17w2 , gw3 = −w3
(i) Find a basis for HomCG (V, W ).
(ii) Find all common composition factors of V and W , if any.
Exercise 4.5.2. Let G be a group of order 18 and {V1 , . . . , Vk } be a complete set of
nonisomorphic irreducible CG-modules. List all the possible integers k and hence all
the sequences of dimensions
dim V1 , dim V2 , . . . , dim Vk .
Exercise 4.5.3. Let G = hg | g 2 = 1i be a group of order 2 and V a CG-module of
dimension 3 and basis {v1 , v2 , v3 } on which G acts by
gv1 = −9v1 + 4v2 − 8v3 gv2 = −4v1 + v2 − 4v3 gv3 = 8v1 − 4v2 + 7v3
Let u1 = v1 + 2v2 + 2v3 , u2 = 2v1 − v2 + 2v3 , u3 = 2v1 − 2v2 + v3 , and hence Ui the
vector space spanned by ui , for i = 1, 2, 3.
(i) Prove that Ui is a CG-submodule of V for i = 1, 2, 3, and that
V = U1 ⊕ U2 ⊕ U3 .
(ii) Let λ ∈ C and u = v1 + v2 + λv3 ∈ V . Find the value(s) of λ for which the
subspace U spanned by u is a CG-submodule of V , and find all i ∈ {1, 2, 3}
for which U ∼
= Ui .

Summary. The notions studied in this section are concerned with the structure of
the group algebra CG of a finite group G. We record in particular the following
concepts and results:
• CG as CG-module and its decomposition;
• composition
Pk factors;
• i=1 dim(Vi )2 = |G|, for any finite group G, where {V1 , . . . , Vn } is a complete
set of nonisomorphic CG-modules.
Representation Theory of Finite Groups page 68

Summary. The notions studied in this section summarise the basic properties of the
group algebra CG of a finite group G.
• composition factors and detection using CG-homomorphisms;
• CG-module structure of CG;
• decomposition of CG and complete set of nonisomorphic CG-modules.

5. Characters
Given an abstract finite group G, we have investigated CG-modules, which roughly
allow us to associate to G a set of matrices in a consistent way reflecting the group
structure and providing us with a way to extract some properties of G. For instance, a
finite group all of whose irreducible modules have dimension 1 must be abelian. In this
section we go a step further in extracting from each matrix a single number: its trace.
This seemingly ingenuous procedure has very useful and enlightening consequences.

5.1. Definitions and examples.

Given a square matrix A = (aij ) ∈ MXn (C), the trace of A, denoted Tr(A), is the sum
of the diagonal coefficients Tr(A) = aii ∈ C of A.
1≤i≤n

Recall the following properties of the trace learned in MATH103 and MATH220: for
any A, B ∈ Mn (C) and λ ∈ C,
(i) Tr(In ) = n,
(ii) Tr(A + B) = Tr(A) + Tr(B),
(iii) Tr(AB) = Tr(BA),
(iv) if A is invertible, then Tr(ABA−1 ) = Tr(B). In particular, if λ1 , . . . , λd are
the distinct eigenvalues of B, with λi of multiplicity mi for each i, then
X
Tr(B) = mi λi
1≤i≤d

The proofs are left in exercise. Let us point out a subtlety of part (iii). Namely,
the trace is not multiplicative, i.e. Tr(AB) 6= Tr(A) Tr(B) for A, B ∈ Mn (C) in
general. Take for instance A = B = I2 . Then Tr(AB) = Tr(I2 ) = 2, whereas
Tr(A) Tr(B) = Tr(I2 )2 = 22 = 4.
Part (iv) says that the trace is an invariant of similitude. That is, if B and B 0 are
similar, then Tr(B) = Tr(B 0 ). Recall that similar precisely means that there exists
an invertible matrix A such that B 0 = ABA−1 .However  matrices having the same
2 0
trace need not be similar. For instance I2 and have the same trace but are
0 0
not similar.
Definition 5.1. Let ρ : G → GLn (C) be a representation of degree n ≥ 1 of a finite
group G and V the corresponding CG-module.
Representation Theory of Finite Groups page 69

(i) The character afforded by ρ, or by V is the function

χ : G → C , χ(g) = Tr ρ(g) (2)
We may also write χρ (or χV ) if we need to specify the representation (or the
module). Hence χ is a character of the group G.
(ii) The degree of χ is
deg(χ) = χ(1) = dim V = deg(ρ)
(iii) If deg(χ) = 1, we call χ a linear character of G. In particular, if ρ is the
trivial representation of G, then χ is the trivial character of G
(iv) If ρ : G → GL|G| (C) is the regular representation of G, then χ is called the
regular character of G.
(v) If ρ is irreducible, then we call χ an irreducible character of G.
Example 5.2. Let G be a finite group. Then G is abelian if and only if all the
irreducible characters of G have degree 1.
Example 5.3. Let G = hg | g 2 = 1i and ρ : G → GL2 (C) the regular representation.
0 1
That is, ρ(g) = . Then
1 0

χρ : G → C , χρ (1) = and χρ (g) =

Let τ : G → GL1 (C) be the representation of G defined by

τ (1) = (1) and τ (g) = (−1)
Then
χτ (1) = and χτ (g) =

Using the elementary properties of the trace, we note that characters are invariants
of groups in the following sense.
Proposition 5.4. Let χ : G → C be the character of a finite group G afforded by
some representation ρ : G → GLn (C), and V the corresponding CG-module.
(i) Suppose that g, g 0 ∈ G are conjugate. Then χ(g) = χ(g 0 ). In other words, the
character of a representation takes the same value on all the elements in a
given conjugacy class of G.
(ii) Suppose that ρ0 ∼ ρ. Then χρ0 = χ. That is, equivalent representations have
the same character.
(iii) Suppose that V 0 ∼
= V . Then χV 0 = χ. That is, isomorphic CG-modules have
the same character.

Proof. For part (i) let h ∈ G such that g 0 = hgh−1 . Choosing a basis of Cn , this
equality translates in terms of matrices as:
ρ(g 0 ) = since ρ is a group homomorphism.
Now, by property (iv) of the trace, we have

Tr(ρ(g 0 )) =

That is, by definition of χ, χ(g 0 ) = χ(g) as required.

Representation Theory of Finite Groups page 70

For part (ii), suppose that ρ0 ∼ ρ. By definition, this means that there exists T ∈
GLn (C) such that ρ0 (g) = T ρ(g)T −1 for all g ∈ G. As above we have
χρ0 (g) =
The last part is the translation of part (ii) in terms of modules. 
Remark 5.5.
(i) Proposition 5.4 (i) says that a character is a class function. The definition of
a class function is precisely a function f : G → C which takes the same value
on all the elements in a conjugacy class of G (and therefore f is determined
by the values it takes on a set of representatives of the conjugacy classes).
Recall that the conjugacy classes of G are all the subset of G of the form
g G = {hgh−1 | h ∈ G} for g ∈ G
In particular, two conjugacy classes are either equal or disjoint. Hence if
{g1 , . .[
. , gc } is a set of representatives of the conjugacy classes of G, then
G= giG is the disjoint union of its distinct conjugacy classes. Thus
1≤i≤c
a character is entirely determined by the images of g1 , . . . , gc .
A well-known theorem asserts that the irreducible characters of a finite group
form a basis of the C-vector space of class functions on G.
(ii) Proposition 5.4 (iii) implies that
there can be at most as many distinct irreducible characters as
there are nonisomorphic irreducible CG-modules.
We will prove below that there are in fact exactly as many irreducible char-
acters as there are nonisomorphic irreducible CG-modules.
Example 5.6. Let G = hg | g 3 = 1i and V the permutation CG-module with natural
basis B = {v1 , v2 , v3 }. Thus

[g]B = , [g 2 ]B =

so that χV (g) = χV (g 2 ) = and χV (1) =

Now let B 0 = {w1 , w2 , w3 } be the basis of V formed by the eigenvectors of [g]B , namely
w1 = v1 + v2 + v3 , w2 = v1 + ζ 2 v2 + ζv3 , w3 = v1 + ζv2 + ζ 2 v3
2πi
with gw1 = w1 , gw2 = ζw2 and gw3 = ζ 2 w3 , where ζ = e 3 . Then

[g]B0 = , [g 2 ]B0 =

so that χV (g) = χV (g 2 ) = and χV (1) = Recall indeed that ζ

is a primitive cube root of 1, that is, a root of the polynomial x2 + x + 1.
Note that V is in fact the regular CG-module. Now, the irreducible characters of G
are all linear, i.e. of degree 1, because G is abelian. These are the characters afforded
by the three irreducible CG-modules Wi = hwi i for 1 ≤ i ≤ 3. For convenience, write
χi instead of χWi for 1 ≤ i ≤ 3. We calculate for instance
χ2 (g) = Tr([g]{w2 } ) =
Representation Theory of Finite Groups page 71

We summarise all these calculations in a table:

χ1 (1) = χ1 (g) = χ1 (g 2 ) =
χ2 (1) = χ2 (g) = χ2 (g 2 ) =
χ3 (1) = χ3 (g) = χ3 (g 2 ) =

We will come back to such tables shortly. An immediate observation from this exam-
ple is that if G is abelian, then each conjugacy class is reduced to a single element,
and each irreducible character has degree 1. Hence let us take the example of a
non-abelian group.
Example 5.7. Let G = S3 be the symmetric group of degree 3. Put g = (1 2 3) and
h = (1 2). Recall that two elements of the symmetric group Sn are conjugate if and
only if they have the same cycle type. So there are 3 conjugacy classes, represented by
1, g, h, and it follows that any character χ of G is determined by the values χ(1), χ(g)
and χ(h). Let V be the permutation CG-module with natural basis B = {v1 , v2 , v3 }.
We calculate

χV (1) = , χ(g) = and χ(h) =

Now a complete set of nonisomorphic CG-modules is formed by the modules:

W1 = , W2 = and W3 =

where ε(σ) denotes the signature of a permutation σ ∈ G, and where w3 =

and w4 = So the action of G on W3 is defined by
gw3 = , gw4 = , hw3 =
using that gh = hg 2 and h2 = g 3 = 1. Write χi = χWi for 1 ≤ i ≤ 3. We calculate
χ1 (1) = χ1 (g) = χ1 (h) =
χ2 (1) = χ2 (g) = χ2 (h) =
χ3 (1) = χ3 (g) = χ3 (h) =
Observe that the first column, corresponding to the conjugacy class of 1 ∈ G, gives
the degree of the considered character.

Together with (matrix) representations of G and CG-modules, characters provide us

therefore with a third tool which we can use to analyse the structure of, and compute
with, abstract groups. In the “triangle” (3) below, we visualise our “representation
tools”: given a finite group G,

• the top left vertex of our “triangle” is formed by a “nice set” of matrices, one
for each element of G,
• the top right vertex is the corresponding vector space, essentially some set
Cn on which each g ∈ G acts by a linear transformation, and
• the bottom vertex is a complex function which takes constant values on con-
jugacy classes of elements of G.

These three vertices allow us to use computational algebra in order to study the
structure of an abstract group given by a set of generators and relations.
Representation Theory of Finite Groups page 72

ρ : G → GLn (C) o / V
(3)
g 7→ ρ(g) g · v = ρ(g)v
h 7

( w
χ:G → C
g 7→ Tr(ρ(g))

We now present some results on characters of finite groups, and see what “informa-
tion” about G we can extract from them.
Let ρ : G → GLn (C) be a representation of degree n ≥ 1 and affording the character
χ : G → C. Suppose that ρ is not irreducible, say ρ = ρ1 ⊕ ρ2 with ρj of degree nj ,
and let χj be the character afforded by ρj for j = 1, 2. So n = n1 + n2 and we can
find a basis of Cn such that
 
ρ1 (g) 0
ρ(g) = for all g ∈ G.
0 ρ2 (g)
In particular,
χ(g) =
Extending from two to a finite number of subrepresentations and using the theorem
of complete reducibility, we have shown the following fact.
Proposition 5.8. Let ρ : G → GLn (C) be a representation of a finite group G. Sup-
pose that ρ = ρ1 ⊕ · · · ⊕ ρt is a decomposition of ρ into irreducible subrepresentations.
Denote χj the character of ρj for each 1 ≤ j ≤ t and χ the character of ρ. Then
X
χ(g) = χj (g) for all g ∈ G.
1≤j≤t

In other words, the character of a representation is the sum of the characters of its
irreducible subrepresentations, counted with their multiplicities.

We leave the interpretation of Proposition 5.8 in terms of CG-modules and their

composition factors as an exercise. Instead, we outline an immediate consequence.
Corollary 5.9. Let ρ : G → GLn (C) be a representation of a finite group G. Let
g ∈ G be an element of order t and suppose that ρ(g) has eigenvalues λ1 , . . . , λr with
multiplicities m1 , . . . , mr respectively, then
X
χ(g) = mi λi , where λti = 1 for all 1 ≤ i ≤ r.
1≤i≤r

(That is, each eigenvalue of ρ(g) is a t-th root of 1.)

Furthermore,
χ(g −1 ) = χ(g), ∀ g ∈ G, and χ(g) ∈ R if g and g −1 are conjugate.

Proof. Let g ∈ G of order t. By assumption, ρ(g) is conjugate to the diagonal matrix

λ1
 
 ...  with each λi repeated mi times.
λr
Representation Theory of Finite Groups page 73

Now, because g t = 1, we must have ρ(g)t = In and hence

t
λ1


In =  ..  = which forces for all indices.

.
λr
For the last claim, recall that if z ∈ C is a root of 1, then

Therefore, continuing with the above notation for the eigenvalues of ρ(g), all of which
we know to have modulus 1, we calculate

χ(g) =

since ρ(g) has eigenvalues λ1 , . . . , λr if and only if the eigenvalues of ρ(g −1 ) = ρ(g)−1
are
Finally, if g and g −1 are conjugate, then
χ(g) = which implies that

Theorem 5.10. Let G be a finite group and χ : G → C the character of G
afforded by some representation ρ : G → GLn (C) of G. Let V be the corresponding
CG-module.
(i) χ(1) = dim(V ).
(ii) |χ(g)| ≤ χ(1) for all g ∈ G, with equality if and only if ρ(g) = λIn for some
λ ∈ C with |λ| = 1.
(iii) ker(ρ) = {g ∈ G | χ(g) = χ(1)}.

Proof. Part (i) is immediate from the observation that

χ(1) =
For part (ii), let g ∈ G and λ1 , . . . , λn the eigenvalues of ρ(g), not necessarily distinct.
Then
χ(g) = and we also have

By Corollary 5.9, each λj is a complex |G|-th root of 1. Therefore, each eigenvalue

must have modulus 1, i.e. |λj | = 1 for all j. By the triangle inequality, it follows that

|χ(g)| = as required.

Moreover, equality holds if and only if all the λj are

when we regard C as the 2-dimensional real vector space. Since all the λj have
modulus 1, they must be and so ρ(g) = for some complex number λ with
modulus
For part (iii), we have by definition ker(ρ) = {g ∈ G | ρ(g) = In }, and since Tr(In ) =
χ(1), we have an inclusion
ker(ρ) ⊆
Representation Theory of Finite Groups page 74

Conversely, the assumption χ(g) = χ(1) implies that χ(g) is a positive real number
(integer in fact), and by (ii) we must have

ρ(g) = for some λ ∈ C with

Now χ(g) = , with |λ| = 1, is a positive real number if and

only if λ = and therefore ρ(g) = i.e. g ∈ ker(ρ), as required. 

Definition 5.11. Let χ be a character of a finite group G. The kernel of χ is

ker(χ) = {g ∈ G | χ(g) = χ(1)} = ker(ρ)

We call χ faithful if ker(χ) = {1}.

In particular, the kernel of a character of G is a normal subgroup of G.

The next example illustrates the above result and serves as an introduction to the
next section.

Example 5.12. Let G = hg, h | g 3 = h3 = 1 , gh = hgi ∼ = C3 × C3 and ρ : G →

GL9 (C) the regular representation. We have
Mseen, by classifying the irreducible rep-
resentations of abelian groups, that ρ = ρj,k where ρj,k (g n hm ) = (ζ jn+km ) and
0≤j,k≤2
2πi
ζ = e , for all j, k, n, m ∈ {0, 1, 2}. (Recall that ζ 2 = ζ −1 = ζ.) In particular each of
3

the nine irreducible representations has degree 1 and ρ0,0 is the trivial representation.
Take for instance ρ1,2 : G → C× , defined by ρ1,2 (g) = ζ and ρ1,2 (h) = ζ 2 . Let us
calculate χ1,2 :

χ1,2 (1) = χ1,2 (g) = χ1,2 (g 2 ) =

χ1,2 (h) = χ1,2 (gh) = χ1,2 (g 2 h) =
χ1,2 (h2 ) = χ1,2 (gh2 ) = χ1,2 (g 2 h2 ) =

We observe that |χ1,2 (x)| = 1 and that χ1,2 (x) = χ1,2 (x−1 ) for all x ∈ G, as expected.
In addition, we calculate

ker(χ1,2 ) =

In particular (as we already knew), χ1,2 is not faithful. We also observe that characters
of linear representations (i.e. of degree 1) are essentially “the same”, as 1 × 1 matrices
with complex coefficients are generally identified with their unique coefficient.
Let us pick any other irreducible representation, say ρ2,0 . We calculate:

χ2,0 (1) = χ2,0 (g) = χ2,0 (g 2 ) =

χ2,0 (h) = χ2,0 (gh) = χ2,0 (g 2 h) =
χ2,0 (h2 ) = χ2,0 (gh2 ) = χ2,0 (g 2 h2 ) =

Here ker(χ2,0 ) =
Because G is abelian, each element is its own conjugacy class. Let us compare χ1,2
and χ2,0 using the inner product of Definition 5.13 below. We define

1 X 1 X
hχ1,2 , χ2,0 i = χ1,2 (g j hk )χ2,0 (g j hk ) = χ1,2 (g j hk )χ2,0 (h−k g −j )
|G| 0≤j,k≤2 |G| 0≤j,k≤2
Representation Theory of Finite Groups page 75

by Corollary 5.9. Using that each element has order 3, i.e ζ 2 = ζ −1 , g 2 = g −1 and
h2 = h−1 we compute (following left to right, top to bottom of χ1,2 ):
1
hχ1,2 , χ2,0 i = χ1,2 (1)χ2,0 (1) + χ1,2 (g)χ2,0 (g 2 ) + χ1,2 (g 2 )χ2,0 (g)+
9
+ χ1,2 (h)χ2,0 (h2 ) + χ1,2 (gh)χ2,0 (g 2 h2 ) + χ1,2 (g 2 h)χ2,0 (gh2 )+
+ χ1,2 (h2 )χ2,0 (h) + χ1,2 (gh2 )χ2,0 (g 2 h) + χ1,2 (g 2 h2 )χ2,0 (gh)+

=
Similarly
hχ1,2 , χ1,2 i =

In the following sections we will explain the equalities obtained in Example 5.3, which
are not coincidences. As a motivation for this scope, let us mention that these equa-
tions are cornerstones in group theory, and they have been key in the early stages
of the classification of finite simple groups in particular. But also characters provide
very efficient tools in demonstrating important properties of finite groups, such as
Burnside’s pa q b theorem, which asserts that any group whose order is the product of
two distinct prime powers is soluble (or solvable).

5.2. Inner products.

We define the inner product of characters by extending that of representations given
in Definition 2.22.
Definition 5.13. Let ρ1 : G → GLn1 (C) and ρ2 : G → GLn2 (C) be two representa-
tions of a finite group G, and let χ1 and χ2 be their respective characters. The inner
product of χ1 , χ2 is the complex number
1 X 1 X
hχ1 , χ2 i = χ1 (g)χ2 (g) = χ1 (g)χ2 (g −1 ) (by Corollary 5.9).
|G| g∈G |G| g∈G

It is straightforward from the proof following Definition 2.22 that h , i is indeed an

inner product, i.e. subject to the axioms
(i) Conjugate symmetry: hχ1 , χ2 i = hχ2 , χ1 i.
(ii) Linearity in the first argument:
hχ1 + χ01 , χ2 i = hχ1 , χ2 i + hχ01 , χ2 i and hλχ1 , χ2 i = λhχ1 , χ2 i.
(iii) Positive definite: hχ1 , χ1 i ≥ 0 with equality if and only if χ1 = 0 is the zero
function.
It follows that we also have a notion of orthogonality for characters.
Definition 5.14. Let χ1 , χ2 : G → C be two characters of a finite group G. We say
that χ1 and χ2 are orthogonal to each other if hχ1 , χ2 i = 0.
Example 5.15.
(i) Let G be any finite group and χ : G → C the trivial character. That is,
χ(g) = 1 for all g ∈ G. Then

hχ, χi =
Representation Theory of Finite Groups page 76

(ii) Let G = hg | g 3 = 1i ∼
= C3 and χ the regular character of G. That is,
   
0 0 1 0 1 0
χ(1) = Tr(I3 ) = 3 , χ(g) = Tr 1 0 0 = 0 , χ(g 2 ) = Tr 0 0 1 = 0
0 1 0 1 0 0
which give

hχ, χi =
Let χ1 be the trivial character of G and χ2 : G → C the linear character
2πi
afforded by ρ2 : G → GL1 (C) with ρ2 (g) = ζ = e 3 . We calculate
hχ1 , χ1 i = as noted above

hχ2 , χ2 i =

hχ1 , χ2 i =
because 1 + ζ + ζ 2 = 0

For the next result, let us recall a very important and useful equation.
Theorem 5.16 (Class equation). Let G be a finite group and g ∈ G. Let
g G = {hgh−1 | h ∈ G} denote the conjugacy class containing g ∈ G and
\
Z(G) = {g ∈ G| gh = hg , ∀ h ∈ G} = CG (g) the centre of G,
g∈G
where CG (g) = {h ∈ G | hg = gh} is the centraliser of an element g in G. Suppose
that g1 , . . . , gl form a set of representatives of the conjugacy classes of G of size greater
than 1. Note that Z(G) is the set of elements of G whose conjugacy class has size 1
and we have
G = Z(G) ∪ g1G ∪ · · · ∪ glG (disjoint union).
Then X
|G| = |Z(G)| + |giG | where |giG | = |G : CG (gi )|
1≤i≤l
|G|
and |G : CG (g)| denotes the index of CG (g) in G, i.e. the number of distinct
|CG (g)|
cosets G/CG (g), for any g ∈ G.
Example 5.17. Let G = S3 . Then G has 3 conjugacy classes represented by ,
s= and t = Their respective sizes are

, |G : CG (s)| = , |G : CG (t)| =
Now Z(G) = {1} by direct computation. We check

where for short we indicate in the indices representatives of the corresponding classes.
Example 5.18. Let G = ha, b | a6 = b2 = 1 , ba = a−1 bi be a dihedral group of
order 12. We calculate that G has 6 conjugacy classes, represented by 1, a, a2 , a3 , b
and ab, of size 1, 2, 2, 1, 3, 3 respectively, and that Z(G) = {1, a3 }. The centralisers
are as follows:
CG (a) = CG (a2 ) = CG (a4 ) = CG (a5 ) = hai ∼
= C6 of index 2,
CG (aj b) = ha3 , aj bi ∼
= C2 × C2 of index 3 for 0 ≤ j ≤ 5.
Representation Theory of Finite Groups page 77

The class equation gives the equality

Proposition 5.19. Let χ1 , χ2 be characters of a finite group G, and suppose that G

has exactly l conjugacy classes, represented by g1 , . . . , gl .
(i) hχ1 , χ2 i = hχ2 , χ1 i.
X χ1 (gi )χ2 (gi )
(ii) hχ1 , χ2 i = .
1≤i≤l
|CG (gi )|

Proof. For the first part, we have by Corollary 5.9

1 X
hχ1 , χ2 i = χ1 (g)χ2 (g −1 ) =
|G| g∈G

For the second part, let g G = {hgh−1 | h ∈ G} as in Theorem 5.16. So χ(k) = χ(g)
for any k ∈ g G , and k ∈ g G if and only if k −1 ∈ (g −1 )G . It follows that
1 X
hχ1 , χ2 i = χ1 (g)χ2 (g −1 ) =
|G| g∈G

By the class equation,

|G|
|g G | = |G : CG (g)| = for all g ∈ G,
|CG (g)|
and so
hχ1 , χ2 i =

5.3. Orthonormality of irreducible characters.

Let G be a nontrivial finite group and V1 , . . . , Vt a complete set of nonisomorphic
irreducible CG-modules with di = dim(Vi ) and t ≥ 2. By Theorem 4.19, the group
algebra of G decomposes into a direct sum
CG = V1d1 ⊕ · · · ⊕ Vtdt
In particular, we can choose proper submodules W1 , W2 of CG with no common
composition factors, i.e. HomCG (W1 , W2 ) = {0}, and such that
CG = W1 ⊕ W2 and 1 = e1 + e2 for some ei ∈ Wi for i = 1, 2.
Expressed in this way, the elements e1 , e2 are also elements of CG, and so must be of
the form X (i)
ei = a(i)
g g for some ag ∈ C.
g∈G

Proposition 5.20. With the above notation, the following hold.

(i) Let wj ∈ Wj . Then ei wj = δi,j wj for all i, j ∈ {1, 2}. In particular, e2i = ei
and e1 e2 = e2 e1 = 0.
(ii) Let χ1 be the character afforded by W1 . Then
1 X
e1 = χ1 (g −1 )g ∈ CG
|G| g∈G
Representation Theory of Finite Groups page 78

(iii) Let χ1 be the character afforded by W1 . Then

hχ1 , χ1 i = χ1 (1) = dim(W1 )
Definition 5.21. An element e ∈ CG subject to e2 = e is called idempotent of CG.
In particular, any group algebra has two idempotents: 1 and 0.

The main result about inner products shows that irreducible characters are orthonor-
mal vectors in the vector space of class functions, i.e. any two distinct irreducible
characters are orthogonal to each other and each one has norm 1. By “norm”, we
mean the norm defined by taking the square root of the inner product of a vector
with itself (as in the euclidean space).
Theorem 5.22 (Orthonormality of irreducible characters). Let χ1 , χ2 be irreducible
characters of a finite group G. Then

1 if χ1 = χ2
hχ1 , χ2 i =
0 otherwise

Proof. Without loss suppose that |G| > 1 and write

CG = V1d1 ⊕ · · · ⊕ Vtdt as direct sum of its composition factors (t ≥ 2, d ≥ 1).
Let Ui = Vidi for 1 ≤ i ≤ t, so that CG = U1 ⊕· · ·⊕Ut and Ui and Uj have no common
composition factors for any i 6= j. Let χi denote the character of Vi for all
Mi. Then
di
χUi = di χi since Ui is the direct sum Vi . For any i, writing CG = Ui ⊕ Uj , we
j6=i
M

are in the assumptions of Proposition 5.20 (iii) with Ui instead of W1 and Uj
j6=i
instead of W2 . So we get the equality

By the axioms of inner products,

which proves that for each i.

Now regarding the decomposition CG = U1 ⊕ U2 ⊕ U3 ⊕ · · · ⊕ Ut of CG, consider
the character χ = d1 χ1 + d2 χ2 of U1 ⊕ U2 . Proposition 5.20 (iii) says that
hχ, χi =
On the other hand, by the axioms of inner products and using that hχ1 , χ2 i = hχ2 , χ1 i
(by Proposition 5.19),
hχ, χi =
=
=
Comparing both values for hχ, χi, and since d1 , d2 are positive integers, we must have
hχ1 , χ2 i = as required.

Representation Theory of Finite Groups page 79

Corollary 5.23. Let χ be any character of a finite group G. Suppose that χ =

a1 χ1 +· · ·+ar χr as sum of irreducible characters of G, with the coefficients ai positive
integers. Then
X
hχ, χi = a2i
1≤i≤r

In particular, if χ is the regular character of G, then hχ, χi = |G|.

Proof. By the properties of inner products, we calculate

* +
X X
hχ, χi = ai χi , aj χ j
1≤i≤r 1≤j≤r

= as required.

In particular, if χ is the regular character of G, then we may assume that the dis-
tinct composition factors of the regular CG-module are V1 , . . . , Vr with dimensions
a1 , . . . , ar equal to the multiplicity of each Vi as composition factor of CG, by Theo-
rem 4.19. Now Theorem 4.23 asserts that
|G| =


Corollary 5.24. Let χ be a character of a finite group G. Then
χ is irreducible ⇐⇒ hχ, χi = 1

Proof. We need to show that if hχ, χi = 1, then χ is irreducible. Since P

each represen-
tation is completely reducible (Maschke’s theorem), we can write χ = i ai χi , where
χi are irreducible characters of G and ai positive integers. Then
X
hχ, χi = a2i
i

Therefore, if hχ, χi = 1, we must have χ = χ1 , i.e. there is a unique nonzero coefficient

and it must be 1. 
Definition 5.25. Let χ be a character of G, say χ = a1 χ1 + · · · + ar χr with the χi
irreducible characters of G and the ai positive integers. The irreducible characters χi
appearing in such a decomposition of χ are called the constituents of χ.

Let us end this section with the answer to the question raised above: How many
distinct irreducible characters does a finite group have?
Indeed, we have seen that if G is a finite group and V, W two irreducible CG-modules,
V ∼= W =⇒ χV = χW in Proposition 5.4.
The difficult assertion is the converse.
Theorem 5.26. Let G be a finite group. Suppose that V, W are irreducible CG-
modules. Then
V ∼= W ⇐⇒ χV = χW
Representation Theory of Finite Groups page 80

Proof. Suppose that χV = χW and write

= V1a1 ⊕ · · · ⊕ Vk k and W ∼
V ∼ a b
= V1b1 ⊕ · · · ⊕ Vk k
as direct sums of irreducible CG-modules (possibly
X some ai , bi can Xbe zero), with
corresponding characters χi = χVi . Thus χV = aj χ j , χ W = bj χj and
1≤j≤k 1≤j≤k
X
hχV , χi i = h aj χj , χi i = ai and similarly hχW , χi i = bi
1≤j≤k

By assumption, χV = χW , which shows that ai = bi for all i. Therefore, V ∼

= W as
required. 
2πi
Example 5.27. Let G = hg | g 3 = 1i and ζ = e 3 a primitive cube root of 1. Recall
the equation 1 + ζ + ζ 2 = 0. Let V = hv1 , v2 i and W = hw1 , w2 i with
   
0 −1 ζ 0
[g]V = and [g]W =
1 −1 0 ζ2
We calculate
[g 2 ]V = and [g 2 ]W = and so,

χV (1) = χW (1) =
χV (g) = and χW (g) =
χV (g 2 ) = and χW (g 2 ) =
Therefore χV = χW and V ∼ = W (which we already knew).

5.4. Character table of a finite group.

So far we have not said anything about the number of irreducible distinct characters
of G, i.e. how many nonisomorphic irreducible modules a given group has. The only
fact we know is that there are finitely many, and this number coincides with the
number of nonisomorphic CG-modules. For completeness, we give the result without
proof.
Theorem 5.28. The number of distinct irreducible characters of a finite group G is
equal to the number of conjugacy classes of G.

As in Theorem 5.16, let us denote the conjugacy classes of G

[
g G = {hgh−1 | h ∈ G} and so G = giG
1≤i≤r

where g1 , . . . , gr are representatives of all the distinct conjugacy classes of G (including

those of size 1).
We know that there are as many distinct irreducible characters as there are conjugacy
classes, and moreover, each character is a class function, i.e. it suffices to know the
value of a character on one element per conjugacy class. In particular, g, h ∈ G are
conjugate if and only if χ(g) = χ(h) for each character of G, or more briefly, for each
irreducible character of G.
Definition 5.29 (Character table).
Let G be a finite group and g1 , . . . , gr representatives of the distinct conjugacy classes
of G. Let χ1 , . . . , χr denote the irreducible characters of G. The r × r matrix with
(i, j) coefficient χi (gj ) is called the character table of G.
Representation Theory of Finite Groups page 81

Much information about the group structure can be recovered from its character table,
and a few exercises below point in that direction. We end the material of the course
with a few examples and observations.
Example 5.30. Let G = hg | g n = 1i ∼ = Cn for some positive integer n. Then G
has n irreducible characters, χ1 , . . . , χn , all of degree 1, and where χk (g j ) = ζ (k−1)j
2πi
and ζ = e n ∈ C. Also, G has n distinct conjugacy classes, each of size 1. So the
character table of G has the form:
1 g ... gj . . . g n−1
χ1 1 1 ... 1 ... 1
j n−1
χ2 1 ζ ... ζ ... ζ
.. .. .. .. ..
. . . ... . ... .
χk 1 ζ k−1 . . . ζ (k−1)j . . . ζ n−k+1
.. .. .. .. ..
. . . ... . ... .
n−1 n−j
χn 1 ζ ... ζ ... ζ

In particular, for n = 4, then ζ = i and the character table of C4 is

Example 5.31. Let G = hg, h | g 4 = h2 = 1 , gh = hgi ∼ = C4 × C2 . Then G has

eight conjugacy classes and eight irreducible characters, say χj,k for 1 ≤ j ≤ 4 and
k = 1, 2. From Theorem 3.23, we gather
1 g g2 g3 h gh g 2 h g 3 h
χ1,1 1 1 1 1 1 1 1 1
χ1,2 1 1 1 1 −1 −1 −1 −1
χ2,1 1 i −1 −i 1 i −1 −i
χ2,2 1 i −1 −i −1 −i 1 i
χ3,1 1 −1 1 −1 1 −1 1 −1
χ3,2 1 −1 1 −1 −1 1 −1 1
χ4,1 1 −i −1 i 1 −i −1 i
χ4,2 1 −i −1 i 1 i 1 −i

The following result is useful in finding the character table of a finite group G with
a normal subgroup N such that we know the characters of the quotient group G/N ,
e.g. when G/N is abelian.
Proposition 5.32. Let G be a finite group and N a normal subgroup of G. Write
ĝ = gN for the elements of G/N , i.e. the cosets of N in G. Suppose that χ : G/N →
C is a character of G/N . The function
e:G→C
χ , e(g) = χ(ĝ) for all g ∈ G
χ
is a character of G of same degree as χ. Moreover, χ is irreducible if and only if χ
e
is irreducible. Here ĝ denotes the coset gN of G/N which contains g, for any group
element g ∈ G.

Proof. Let ρ : G/N → GLn (C) be a representation of G/N affording χ. So ρ is a

group homomorphism. Define
ρe : G → GLn (C) , ρe(g) = ρ(ĝ) , ∀ g ∈ G
Representation Theory of Finite Groups page 82

Thus ρe is a group homomorphism since

ρe(gh) =
So ρe is a representation of G of same degree as ρ and by definition, the character χ
e
of ρe takes the values
χ
e(g) =
(Note that N ≤ ker(e
ρ).)
Moreover, ρ is irreducible if and only if ρe is irreducible. Indeed, if τ : G/N → GLm (C)
is a subrepresentation of G/N , for some 1 ≤ m ≤ n, then the same argument as above
yields a subrepresentation τe : G → GLm (C) of G. Conversely, any subrepresentation
θe : G → GLm (C) of ρe must also be a group homomorphism with N ≤ ker(θ), e
i.e. θe defines a representation of the quotient group G/N by (some version of) the
isomorphism theorem for groups. 

The next example shows when Proposition 5.32 can be useful.

Example 5.33. Let G = ha, b | a5 = b2 = 1 , ba = a−1 bi ∼ = D10 be a dihedral group
of order 10. We want to find the conjugacy classes of G. The elements of G are of
the form aj b or aj for 1 ≤ j ≤ 4, and since b = b−1 we calculate
baj b−1 =
Thus the conjugacy class of aj is for each j, which gives the distinct classes:

The conjugacy class of b contains

aba−1 =
a2 ba−2 =
a3 ba−3 =
a4 ba−4 =
So the conjugacy class of b is

Thus G has 4 conjugacy classes, represented by 1, a, a2 and b. Hence there are 4

irreducible characters: χ1 , χ2 , χ3 , χ4 . We know that
deg(χ1 )2 + deg(χ2 )2 + deg(χ3 )2 + deg(χ4 )2 = 10 (4)
and that we can choose χ1 to be the trivial character. So, for Equation (4) to hold,
we see that the only possibility is
deg(χ1 ) = deg(χ2 ) = and deg(χ3 ) = deg(χ4 ) =
so that
Now, N = ha | a4 i / G and G/N ∼ = hb̂ | b̂2 = 1i ∼
= C2 . In particular, G/N is abelian
and has two irreducible characters: the trivial and
λ : G/N → C , λ(b̂) = −1
So λ gives a linear character of G, by Proposition 5.32, say
χ2 : G → C , χ2 (aj bk ) = for 1 ≤ j ≤ 4 and k = 0, 1.
(Note that (∗) holds because λ is a linear character!)
Hence, we still need to find two irreducible characters χ3 , χ4 of G of degree 2.
Representation Theory of Finite Groups page 83

For this, we use the fact that G ∼

= D10 is the group of isometries of a pentagon,
2π
in which a corresponds to a rotation about the origin through an angle and b
5
corresponds to a symmetry about a line (through the origin), as in Figure 4.

Figure 4. Dihedral group of order 10 acting on a regular pentagon

By linear algebra, the matrix of a in the standard basis is

A= and that of b is B =

Taking the character of this (natural) representation of G we get χ3 : G → C with

χ3 (1) =
χ3 (a) =
χ3 (a2 ) =

χ3 (b) =

(Note that if A is the matrix of a rotation through α, then A2 is the matrix of a

rotation through 2α.)
Using the corollary to Schur’s lemma 3.19, we calculate that the only matrices which
commute with A and with B are scalar, so that χ3 is irreducible. (This avoids us to
handle trigonometric equalities instead, if we would use the orthogonality relations.)
Similarly, we get another irreducible representation and the character χ4 of G by
4π
taking for A the matrix of a rotation about the origin through an angle = 2α.
5
This gives
χ4 (1) =
χ4 (a) =
χ4 (a2 ) =
χ4 (b) =

2π
To summarise, the character table of G is, with α = ,
5
Representation Theory of Finite Groups page 84

5.5. Workshop Questions.

Exercise 5.5.1. Let G be a finite group and χ a character of G. For any g, h ∈ G,
if g −1 = h−1 gh then χ(g) ∈ R.
Exercise 5.5.2. Let G be a finite group and χ a character of G. For g ∈ G, if
g 2 = 1, then χ(g) ∈ Z and χ(1) − χ(g) is even.
Exercise 5.5.3. Let χ be the regular character of a finite group G. Prove that
hχ, χi = |G|.
Exercise 5.5.4. Let G be a subgroup of Sn and V = hv1 , . . . , vn i the permutation
module for G. Prove that the character of V is the function which maps each
permutation g ∈ G to the number of letters in {1, . . . , n} which are fixed by g.
Exercise 5.5.5. Let G be a finite group and χ, λ : G → C two characters of G.
Suppose that λ is linear. Define
λ ⊗ χ : G → C , (λ ⊗ χ)(g) = λ(g)χ(g) , ∀ g ∈ G
Then λ ⊗ χ is a character of G of same degree as χ, and λ ⊗ χ is irreducible if and
only if χ is irreducible.
Exercise 5.5.6. Calculate and compare the character tables of D8 and Q8 where
Q8 = hg, h | g 4 = 1 , g 2 = h2 , hg = g −1 hi
Exercise 5.5.7. The objective is to find the character table for the alternating
group A4 .
(i) Let V = hg = (1 2)(3 4) , h = (1 3)(2 4)i ∼
= C2 × C2 . Prove that V / A4 and
the quotient group is isomorphic to hk = (1 2 3)i ∼
= C3 .
(ii) Find the four conjugacy classes of A4 . Hence deduce the degrees of the
irreducible characters of A4 .
(iii) Using Proposition 5.32 and the irreducible characters of A4 /V , complete the
character table of A4 .
Exercise 5.5.8. Let χj be an irreducible character of a finite group G. Put
1 X
ej = χj (g −1 )g ∈ CG
|G| g∈G
Show that if χ1 , . . . , χt is the list of all the irreducible characters of G, then
1 = e1 + · · · + et

Summary. The notions studied in this section summarise the basic properties of
characters of finite groups.
• definition and elementary results;
• inner products of characters and orthogonality relations;
• character table of a finite group.

References
[1] G. James, M. Liebeck, Representations and characters of groups, Second edition, Cam-
bridge University Press, New York, 2001.
[2] S. Lang, Algebra, revised third edition, Vol. 211, Springer-Verlag, New York, 2002.
[3] J. Rotman, Advanced modern algebra Prentice Hall, 2002.
Representation Theory of Finite Groups page 85

[4] J.-P. Serre, Linear representations of finite groups, Vol. 42, Springer-Verlag, New York,
1977.
 Index
∼
=, 36 identity map, 39
identity matrix, 3
action of a group on a set, 14 image
affording representation, 15 of a CG-homomorphism, 33
algebra, 55 index of a subgroup, 76
inner product, 23
centraliser, 76
G-invariant, 19
centre, 41
invariant of similitude, 68
centre of G, 76
isomorphic CG-modules, 36
CG, 55, 58
CG-submodule, 19 ker(ρ), 8
CG-automorphism, 31 kernel
CG-endomorphism, 31 of a CG-homomorphism, 33
CG-homomorphism, 30 of a module, 19
CG-isomorphism, 31, 36
character linear transformation, 3
of a group, 69 linear extension, 17
afforded by a representation, or a module, linear transformation
69 matrix of a linear transformation, 4
degree, 69
faithful, 74 Mn (C), 3
inner product, 75 matrix
irreducible, 69 change of basis, 4
kernel, 74 inverse, 3
linear, 69 invertible, 3
orthogonality, 75 module
regular, 69 CG-module, 14
trivial, 69 cyclic, 60
character table, 80 irreducible, 19
characteristic polynomial, 39 permutation module, 18
class function, 70 reducible, 19
common composition factor, 57
commutator subgroup, 66 natural basis, 18
complete set of nonisomorphic irreducible natural basis, 55, 59
CG-modules, 65 notation of CG-module, 15
completely reducible module, 27
complex nj -th root of unity, 43 orthogonal basis, 24
complex conjugate, 23 orthogonal complement, 24
composition factor, 57 orthogonal vectors, 24
conjugacy class, 70 orthonormal basis, 24
conjugate symmetry, 23
constituent, 79 permutation matrix, 7
positive definite, 23
derived subgroup, 66 projection map, 36
direct sum of modules, 20
domain, 4 range, 4
regular representation of G, 59
eigenvalue, 39 representation, 5
eigenvector, 39 ρλ1 ,...,λr , 43
equivalence relation, 36 degree of, 5
direct sum, 11
fundamental theorem of algebra, 39 equivalent, 9
faithful, 8
general linear group, 3
irreducible, 10
group algebra, 55, 58
kernel, 8
HomCG (V, W ), 46 permutation, 8
reducible, 10
In , 3 regular, 7
idempotent of CG, 78 trivial, 8
86
Representation Theory of Finite Groups page 87

scalar matrix, 41
similar matrices, 68
subrepresentation, 10
proper, 10
subspace
G-invariant, 10

Tr, 68
trace of a square matrix, 68

unique decomposition property, 57

word, 6

zero map, 39