Mathematics 751/762

Further Group Theory/Representation

Theory

Professor J Moori
School of Mathematical Sciences
University of KwaZulu-Natal, Pietermaritzburg

Semester 2, 2010
Contents

1 Permutation Groups 2
1.1 Permutation Representations . . . . . . . . . . . . . . . . . . . . 2
1.2 Permutation Groups . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 The Sylow Theorems 12

2.1 p - Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Sylow Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3 Finite Direct Products 29

3.1 External and Internal Products . . . . . . . . . . . . . . . . . . . 29
3.2 Basis Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 The Fundamental Theorem of Finite Abelian Groups . . . . . . 35

4 Normal Series 43
4.1 Jordan-Hölder Theorem . . . . . . . . . . . . . . . . . . . . . . . 43
4.2 Soluble (Solvable) Groups . . . . . . . . . . . . . . . . . . . . . . 46

5 Representation Theory of Finite Groups 50

5.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2 Characters of Finite Groups . . . . . . . . . . . . . . . . . . . . . 61
5.3 Tensor Products and Products of Characters . . . . . . . . . . . 87
5.4 Restriction to a Subgroup . . . . . . . . . . . . . . . . . . . . . . 92
5.5 Induced Representations . . . . . . . . . . . . . . . . . . . . . . . 96

1
Chapter 1

Permutation Groups

1.1 Permutation Representations

Theorem 1.1.1 (Cayley) Every group G is isomorphic to a subgroup of SG .
In particular if |G| = n, then G is isomorphic to a subgroup of Sn .

Proof. For each x ∈ G, define Tx : G −→ G by Tx (g) = xg. Then Tx is one-to-

one and onto; so that Tx ∈ SG . Now if we define τ : G −→ SG by τ (x) = Tx ,
then τ is a monomorphism. Hence G ∼ = Image(τ ) ≤ SG .

Definition 1.1.1 The homomorphism τ defined in Theorem 1.1.1 is called the

left regular representation of G.

Note 1.1.1 Cayley’s Theorem is not that useful when the group G is large
or when G is simple. Following results (Theorem 1.1.3 and Corollary 1.1.4 )
provide substantial improvement over Cayley’s Theorem. Notice that A5 ≤ S5
and Cayley’s Theorem asserts that A5 is also a subgroup of S60 .

Corollary 1.1.2 Let GL(n, |F ) denote the general linear group over a field
|F. If G is a finite group of order n, then G can be embedded in GL(n, |F),
that is G is isomorphic to a subgroup of GL(n, |F).

Proof. Let Tx be as in Cayley’s Theorem. Assume that G = {g1 , g2 , · · · , gn }.

Let Px = (aij ) denote the n × n matrix given by aij = 1 |F if Tx (gi ) = gj
and aij = 0 |F , otherwise. Then Px is a permutation matrix, that is a
matrix obtained from the identity matrix by permuting its columns. Define
ρ : G −→ GL(n, |F ) by ρ(x) = Px , then it is not difficult to check that ρ is a
monomorphism.

2
Note 1.1.2 If Pn denotes the set of all n × n permutation matrices, then Pn
is a group under the multiplication of matrices and Pn ∼
= Sn .

Example 1.1.1 Consider the Klein four group V4 = {e, a, b, c, }. Then we

have

Ta (e) = a.e = a, Ta (a) = a2 = e, Ta (b) = ab = c, Ta (c) = ac = b;

Tb (e) = b, Tb (a) = c, Tb (b) = e, Tb (c) = a;

Tc (e) = c, Tc (a) = b, Tc (c) = e, Tc (b) = a.

Hence the permutation matrices are

     
0 1 0 0 0 0 1 0 0 0 0 1
     
 1 0 0 0   0 0 0 1   0 0 1 0 
Pe = I4 , Pa =  , Pb =  , Pc =  .
     
 0 0 0 1   1 0 0 0   0 1 0 0 
     
0 0 1 0 0 1 0 0 1 0 0 0

So that V4 ∼
= {I4 , Pa , Pb , Pc } ≤ GL(4, |F)

Exercise 1.1.1 (i) Show that for n ≥ 2, Sn is isomorphic to a subgroup of

An+2 .
(ii) Use part (i) to show that A∞ contains an isomorphic copy of every finite
group. (See below for the definition of A∞ .)

| and let F be the set of all α ∈ SX such that α

Definition 1.1.2 Let X = N
moves finitely many elements of X. Then F ≤ SX . Let A∞ denote the subgroup
of F generated by all 3-cyles of F. It can be shown that A∞ is a simple group.

Theorem 1.1.3 (Generalized Cayley Theorem) Let H be a sugroup of G

and let X be the set of all left cosets of H in G. Then there is a homomorphism
ρ : G −→ SX such that
gHg −1 .
\
Ker(ρ) =
g∈G

Proof. For any x ∈ G, define ρx : X −→ X by ρx (gH) = x(gH). Then ρx is

well-defined, one-to-one and onto. So that ρx ∈ SX . Now define ρ : G −→ SX
by ρ(x) = ρx for all x ∈ G. Then ρ is a homomorphism. We claim that
Ker(ρ) =
T −1 .
g∈G gHg
Let x ∈ Ker(ρ). Then ρx = ρ(x) is the identity permutation on X. Hence
ρx (gH) = gH for all g ∈ G. So that xgH = gH, ∀g ∈ G. So g −1 xg ∈ H, ∀g ∈ G.
This implies that x ∈ gHg −1 , ∀g ∈ G. Thus Ker(ρ) ⊆ −1 .
T
g∈G gHg Now if

3
x∈
T −1 , then x ∈ gHg −1 , ∀g ∈ G. So that xgH = gH for all g ∈ G,
g∈G gHg
that is ρx is the identity permutation. Hence x ∈ Ker(ρ), so
T −1 ⊆
g∈G gHg
Ker(ρ).

Definition 1.1.3 The homomorphism ρ defined above (Theorem 1.1.3) is called

the permutation representation of G on the left cosets of H in G. The ker-
nel of ρ, Ker(ρ) =
T −1 ,
g∈G gHg is called the core of H in G.

Exercise 1.1.2 If ρ is the permutation representation of G on the left cosets

of H in G, then show that
(i) Ker(ρ) ≤ H, (ii) G/Ker(ρ) is isomorphic to a subgroup of SX ,
where X = G/H = {gH | g ∈ G}.

Corollary 1.1.4 If G is an infinite group such that contains a proper subgroup

of finite index, then G contains a proper normal subgroup of finite index.

Proof. Let H ≤ G such that [G : H] = n. Let X = G/H be the set of all left
cosets of H in G. Then |X| = n and there is a homomorphism ρ : G −→ Sn
such that Ker(ρ) =
T −1 .
g∈G gHg Since G/Ker(ρ) is isomorphic to a subgroup
of Sn , G/Ker(ρ) is finite. Obviously Ker(ρ)  G, and since Ker(ρ) ≤ H < G,
Ker(ρ) 6= G. Note that Ker(ρ) 6= {1G }.

Corollary 1.1.5 If G is a simple group containing a proper subgroup H of

finite index n, then G is isomorphic to a subgroup of Sn .

Proof. By Theorem 1.1.3, there exists a homomorphism ρ : G −→ Sn such that

−1 and Ker(ρ) ≤ H. Since Ker(ρ)  G and G is simple,
T
Ker(ρ) = g∈G gHg
Ker(ρ) = G or Ker(ρ) = {1G }. Since H < G and Ker(ρ) ≤ H, Ker(ρ) 6= G.
∼ Image(ρ) ≤ Sn .
Thus Ker(ρ) = {1G }. Hence ρ is a monomorphism; so that G =

Exercise 1.1.3 Prove that if λ and ρ are left and right regular representations
of G, then λ(a) commutes with ρ(b) for all a, b ∈ G.

Exercise 1.1.4 (i)∗ Let G be a group of order 2m k, where k is odd. Prove that
if G contains an element of order 2m , then the set of all elements of odd order
in G is a normal subgroup. (Hint: Consider G as permutations via Cayley’s
Theorem, and show that it contains an odd permutation).
(ii) Show that a finite simple group of even order must have order divisible
by 4.

4
Exercise 1.1.5 (Poincare) If H and K are subgroups of G having finite index,
then H ∩ K has finite index. (Hint: [G : H ∩ K] ≤ [G : H][G : K].)

Exercise 1.1.6 Let G be a finite group and H ≤ G with [G : H] = p, where p

is the smallest prime divisor of |G|. Prove that H is normal in G.

Exercise 1.1.7 Prove that A6 has no subgroup of prime index.

Definition 1.1.4 (Conjugate subgroups) Let G be a group and H ≤ G we

define H g by
H g : = gHg −1 = {ghg −1 | h ∈ H}.

Then H g is called the conjugate of H by g. It is routine to check that H g ≤

G, ∀g ∈ G.

Definition 1.1.5 (Normalizer) If H ≤ G, the normalizer of H in G, de-

noted by NG (H), is defined by

NG (H) : = {g | g ∈ G, gHg −1 = H}.

If H  G, then NG (H) = G.

Exercise 1.1.8 (i) Show that G ≥ NG (H)  H. (ii) If H  K, where H and

K are subgroups of G, then NG (H) ≥ K.

Theorem 1.1.6 Let G be a group and H ≤ G. Let X = {gHg −1 | g ∈

G}. Then there exists a homomorphism φ : G −→ SX such that Ker(φ) =
T −1 .
g∈G gNG (H)g

Proof. Define φg : X −→ X by φg (g 0 Hg 0−1 ) = g(g 0 Hg 0−1 )g −1 . Then φg is

well-defined and φg ∈ SX . Now define φ : G −→ SX by φ(g) = φg . Then φ is a
homomorphism: ∀a, g ∈ G we have φ(ab) = φab and

φab (gHg −1 ) = ab(gHg −1 )b−1 a−1 = a(bgHg −1 b−1 )a−1 = a(φb (gHg −1 ))a−1
= φa (φb (gHg −1 )) = (φa ◦ φb )(gHg −1 ),

hence φab = φa ◦ φb on X and φ is a homomorphism.

If g ∈ Ker(φ), then φ(g) = φg is the identity permutation on X. So ∀g 0 ∈
G we have φg (g 0 Hg 0−1 ) = g 0 Hg 0−1 . Therefore g(g 0 Hg 0−1 )g −1 = g 0 Hg 0−1 ; so
g 0−1 gg 0 Hg 0−1 g −1 g 0 = H, that is g 0−1 gg 0 H(g 0−1 gg 0 )−1 = H. Hence g 0−1 gg 0 ∈
NG (H) and we deduce that g ∈ g 0 NG (H)g 0−1 , ∀g 0 ∈ G. This shows that
Ker(φ) ⊆
T −1 .
g∈G g.NG (H)g (1)

5
 If a ∈
T −1 , then a ∈ gNG (H)g −1 for all g ∈ G. Thus there is
g∈G gNG (H)g
g 0 ∈ NG (H) such that a = gg 0 g −1 . Now we have, for all g ∈ G,

φa (gHg −1 ) = agHg −1 a−1 = gg 0 g −1 gHg −1 gg 0−1 g −1

= gg 0 Hg 0−1 g −1 = gHg −1 ,

since g 0 ∈ NG (H). This shows that φa is the identity on X. Thus a ∈ Ker(φ)

and hence Ker(φ) ⊇
T −1 .
g∈G gNG (H)g (2) Now from (1) and (2) we obtain
that Ker(φ) =
T −1
g∈G gNG (H)g .

Note 1.1.3 The homomorphism φ given in Theorem 1.1.6, is called the per-
mutation representation of G on the conjugates of H.

Exercise 1.1.9 Prove that a subgroup H of G is normal if and only if it has

only one conjugate in G.

Exercise 1.1.10 If H and K are conjugate subgroups of G, then H ∼

= K. Give
an example to show that the converse may be false.

Exercise 1.1.11 if λ and ρ are left and right regular representations of S3 ,

show that λ(S3 ) and ρ(S3 ) are conjugate subgroups of S6 .

Exercise 1.1.12 Let G be a finite group with proper subgroup H. Prove that
G is not the set-theoretic union of all conjugates of H. Give an example in
which H is not normal and this union is a subgroup.

Exercise 1.1.13 (i) Assume H < K < G. Show that NK (H) = NG (H) ∩ K.
(ii) Prove that NG (xHX −1 ) = xNG (H)x−1 .

Exercise 1.1.14 If H and K are subgroups of G, show that NG (H ∩ K) ≥

NG (H) ∩ NG (K). Give an example in which the inclusion is proper.

Exercise 1.1.15 ∗ Let G be an infinite group containing an element x 6= 1G

having only finitely many conjugates. Prove that G is not simple.

1.2 Permutation Groups

Definition 1.2.1 Let G be a group and X be a set. We say that G acts on X
if there exists a homomorphism ρ : G −→ SX . Then ρ(g) ∈ SX for all g ∈ G.
The action of ρ(g) on X, that is ρ(g)(x), is denoted by xg for any x ∈ X. We
say that G is a permutation group on X.

6
Example 1.2.1 (i) If G ≤ SX , then obviously G acts on X naturally.
(ii) By Cayley’s theorem any group G acts on itself and the action is given
by ag = ga, ∀a ∈ G, for g ∈ G.
(iii) If H ≤ G, then G acts on G/H, the set of all left cosets of H in G. The
action is given by: for g ∈ G, (aH)g = gaH, ∀a ∈ G.
(iv) If H ≤ G, then G acts on the set of all conjugates of H in G by
(aHa−1 )g = g(aHa−1 )g −1 = gaHa−1 g −1 .

Definition 1.2.2 (Orbits) Let G act on a set X and let x ∈ X. Then the
orbit of x under the action G is defined by

xG : = {xg | g ∈ G}.

Theorem 1.2.1 Let G act on a set X. The set of all orbits of G on X form a
partition of X.

Proof. Define the relation ∼ on X by x ∼ y if and only if x = y g for some g ∈ G.

Then ∼ is an equivalence relation on X (check) and [x] = {xg | g ∈ G} = xG .
Hence the set of all orbits of G on X partitions X.

Example 1.2.2 (i) If G acts on itself by the left regular representation, then
∀g ∈ G we have g G = {g h | h ∈ G} = {hg | h ∈ G} = Gg = G. Hence under the
action of G, we have only one orbit, namely G itself.
(ii) If G acts on G/H, the set of left cosets of H in G, then ∀aH ∈ G/H we
have

(aH)G = {(aH)g | g ∈ G} = {gaH | g ∈ G} = G/H.

In this case we have only one orbit, namely G/H.

(iii) In the case when G acts on itself by conjugation, that is for g ∈ G we
have ∀x ∈ G xg : = gxg −1 , then

xG = {xg | g ∈ G} = {gxg −1 | g ∈ G} = [x]

the conjugacy class of x in G. Note that |xG | = |[x]| = [G : CG (x)]. In this case
the number of orbits is equal to the number of conjugacy classes of G.
(iv) If G acts on the set of all its subgroups by conjugation, that is H g =
gHg −1 , ∀g ∈ G, ∀H ≤ G, then for a fixed H in G we have

H G = {H g | g ∈ G} = {gHg −1 | g ∈ G}

7
the set of all conjugates of H in G. Later we will prove that the number of
conjugates of H in G is equal to [G : NG (H)]. Hence |H G | = [G : NG (H)]. In
this case the number of orbits of G is equal to the number of conjugacy classes
of subgroups of G.

Definition 1.2.3 (Stabilizer) If G acts on a set X and x ∈ X then the sta-

bilizer of x in G, denoted by Gx is the set Gx = {g | xg = x}. That is Gx is
the set of elements of G that fixes x.

Theorem 1.2.2 Let G act on a set X. Then

(i) Gx is a subgroup of G for each x ∈ X.
(ii) |xG | = [G : Gx ], that is the number of elements in the orbit of x is equal to
the index of Gx in G.

Proof. (i) Since x1G = x, 1G ∈ Gx . Hence Gx 6= Ø. Let g, h be two elements

−1 −1
of Gx . Then xg = xh = x. So (xg )h = (xh )h = x1G = x, and therefore
−1
xgh = x, ∀x ∈ X. Thus gh−1 ∈ Gx .
(ii) Since
−1
xg = xh ⇔ x = xhg ⇔ hg −1 ∈ Gx
⇔ (Gx )g = (Gx )h,

the map γ : xG −→ G/Gx given by γ(xg ) = (Gx )g is well-defined and one-to-

one. Obviously γ is onto. Hence there is a one-to-one correspondence between
xG and G/Gx . Thus |xG | = |G/Gx |.

Exercise 1.2.1 Let G act on a set X. If y = xg for some x, y ∈ X, show that

g −1 Gx g = Gxg = Gy .

Corollary 1.2.3 If G is a finite group acting on a finite set X then ∀x ∈

X, |xG | divides |G|.

Proof. By Theorem 1.2.2 we have |xG | = [G : Gx ] = |G|/|Gx |. Hence |G| =

|xG | × |Gx |. Thus |xG | divides |G|.

Theorem 1.2.4 (Applications of Theorem 1.2.2) (i) If G is a finite group,

then ∀g ∈ G the number of conjugates of g in G is equal to [G:CG (g)].
(ii) If G is a finite group and H is a subgroup of G, then the number of
conjugates of H in G is equal to [G:NG (H)].

8
Proof. (i) Since G acts on itself by conjugation, using Theorem 1.2.2 we have
|g G | = [G:Gg ]. But since

g G = {g h | h ∈ G} = {hgh−1 | h ∈ G} = [g]

and

Gg = {h ∈ G | g h = g} = {h ∈ G | hgh−1 = g} = {h ∈ G | hg = gh} = CG (g),

we have
|G|
|g G | = |[g]| = [G:Gg ] = [G:CG (g)] = .
|CG (g)|
(ii) Let G act on the set of all its subgroups by conjugation. Then by The-
orem 1.2.2 we have |H G | = [G:GH ]. Since H G = {H g | g ∈ G} = {gHg −1 | g ∈
G} = [H] and GH = {g ∈ G | H g = H} = {g ∈ G | gHg −1 = H} = NG (H) we
|G|
have |[H]| = |H G | = [G:GH ] = [G:NG (H)] = |NG (H)| .

Theorem 1.2.5 (Cauchy - Frobenius ) Let G be a finite group acting on a

finite set X. Let n denote the number of orbits of G on X. Let F (g) denote the
1
number of elements of X fixed by g ∈ G. Then n =
P
|G| g∈G F (g).

Let x ∈ X. Since there are |Gx | elements in

P
Proof. Consider S = g∈G F (g).
G that fix x, x is counted |Gx | times in S. If ∆ = xG , then ∀y ∈ ∆ we have
|∆| = |xG | = |y G | = [G:Gx ] = [G:Gy ]. Hence |Gx | = |Gy |. Thus ∆ contributes
[G:Gx ].|Gx | to the sum S. But [G:Gx ].|Gx | = |G| is independent to the choice
of ∆ and hence each orbit of G on X contributes |G| to the sum S. Since we
have n orbits, we have S = n|G|.

Definition 1.2.4 (Transitive Groups) Let G be a group acting on a set X. If

G has only one orbit on X, then we say that G is transitive on X, otherwise we
say that G is intransitive on X. If G is transitive on X, then xG = X∀ x ∈ X.
This means that ∀x, y ∈ X, ∃g ∈ G such that xg = y.

Note 1.2.1 If G is a finite transitive group acting on a finite set X, then

Theorem 1.2.2 (ii) implies that |xG | = |X| = |G|/|Gx |. Hence |G| = |X| × |Gx |.

Definition 1.2.5 (Multiply Transitive Groups) Let G act on a set X and

let |X| = n and 1 ≤ k ≤ n be a positive integer. We say that G is k - transitive
on X if for every two ordered k - tuples (x1 , x2 , · · · , xk ) and (y1 , y2 , · · · , yk ) with
xi 6= xj and yi 6= yj for i 6= j there exists g ∈ G such that xi g = yi for
i = 1, 2, · · · , k. The transitivity introduced in Definition 1.2.4 is the same as 1
- transitive.

9
Exercise 1.2.2 Let G be a group acting on a set X. Assume that |X| = n. Let
1 ≤ k ≤ n be a positive integer.
(i) Show that if G is k - transitive, then G is also (k − 1) transitive, when
k > 1.
(ii) If ∃H ≤ G such that H is k - transitive on X, then G is also k - transitive.

Exercise 1.2.3 Let G be a transitive group on a set X, |X| = k ≥ 2. Show

that G is k - transitive on X if and only if Gx is (k − 1) - transitive on X − {x},
for every x ∈ X.

Theorem 1.2.6 If G is a k - transitive group on a set X with |X| = n, then

|G| = n(n − 1)(n − 2) · · · (n − k + 1)|G[x1 ,x2 ,···,xk ] |

for every choice of k- distinct x1 , x2 , · · · , xk ∈ X, where G[x1 ,x2 ,···,xk ] denote the
set of all elements g in G such that xi g = xi , 1 ≤ i ≤ k.

Proof. Let x1 ∈ X. Then since G is k - transitive, we have |G| = n×|Gx1 | (1)

and Gx1 is (k − 1) - transitive, by Exercise 1.2.3, on X − {x1 }. Choose x2 ∈
X − {x1 }. Then since Gx1 is (k − 1) - transitive on X − {x1 } we have |Gx1 | =
|X − {x1 }| × |(Gx1 )x2 |, that is |Gx1 | = (n − 1) × |G[x1 ,x2 ] | and G[x1 ,x2 ] is (k − 2)
- transitive on X − {x1 , x2 }. (2) Notice that (1) and (2) imply that |G| =
n(n − 1) × |G[x1 ,x2 ] |. If we continue this way, we will get

|G| = n(n − 1)(n − 2) · · · (n − k + 1)|G[x1 ,x2 ,···,xk ] |.

Theorem 1.2.7 Let G act transitively on a finite set X with |X| > 1. Then
there exists g ∈ G such that g has no fixed points.

Proof. By the Cauchy - Frobenius theorem we have

1 X
1=n = F (g)
|G| g∈G
1 X
= [F (1G ) + F (g)]
|G| g∈G−{1 G}

1 X
= [|X| + F (g)].
|G| g∈G−{1 G}

If F (g) > 0 for all g ∈ G, then we have

1 X 1
1= [|X| + F (g)] ≥ [|X| + |G| − 1]
|G| |G|
g∈G−{1 G}

|X| − 1
≥ 1+ > 1,
|G|

10
which is a contradiction. Hence ∃ g ∈ G such that F (g) = 0.

Exercise 1.2.4 Let G be a group of permutations on a set X and let x, y ∈ X.

If xt = y for some t ∈ G, prove that Gx ∼
= Gy . (Hint: Use Exercise 1.2.1.)

Exercise 1.2.5 Use Cauchy - Frobenius to prove Lagrange’s Theorem. (Hint:

Consider the left - regular action of G.)

Exercise 1.2.6 If G is a finite group and c is the number of conjugacy classes

1
x∈G |CG (x)|.
P
of elements of G, show that c = |G| (Hint: Consider the conju-
gation action of G on its elements and use Cauchy - Frobenius Theorem).

Exercise 1.2.7 Let G be a finite group of order pn , where p is a prime. Assume

that G acts on a set X with p not dividing |X|. Prove that there exists x ∈ X
such that xg = x for all g ∈ G. [Hint: use Corollary 1.2.3.]

Exercise 1.2.8 Assume that V ia vector space of dimension n over ZZ p and

GL(n, p) is the corresponding general linear group acting on V . If G is a
subgroup of GL(n, p) with |G| = pm , prove that there exists a non-zero vector
v ∈ V such that gv = v for all g ∈ G. [Hint: since G ≤ GL(n, p), G acts on V .]

11
Chapter 2

The Sylow Theorems

The converse of Lagrange’s theorem is not true in general. That is if n divides

|G|, G does not necessarily have a subgroup of order n. In fact A4 is a group
of order 12 and has no subgroup of order 6, although 6 |12. (See Exercise I.3(1)
# 11, Mathematics 340)
The Norweigian mathematician L. Sylow in 1872 proved that the converse
of Langrage’s theorem is true when n = pk , for p prime. We will devote the
present chapter to the study of p - groups, Sylow theorems and their applica-
tions.

2.1 p - Groups
Definition 2.1.1 Let p be a prime. A group G is called a p - group if every
element g ∈ G has order pk for some k.

Lemma 2.1.1 If G is a finite abelian group such that p | |G|, then there exists
g ∈ G such that o(g) = p.

Proof. Since p | |G|, |G| =

6 1. So ∃ h ∈ G such that h 6= 1G . Assume that o(h) =
l. If l = kp, then o(hk ) = p. Assume that p 6 | l. Consider H =< h > . Then since
G is abelian, H  G and G/H is also abelian. We know that |G/H| = |G|/l and
since p | |G| and p 6 | l, p | |G|/l. Since the group G/H has order less than |G|
and since p divides |G/H|, by induction on |G| we can say that ∃ aH in G/H
with o(aH) = p. Now consider the natural homomorphism φ : G −→ G/H
given by φ(g) = gH. Then φ(a) = aH and 1G/H = H = (aH)p = [φ(a)]p . So
φ(ap ) = H =< h > . That is ap H = H and hence ap ∈ H. Hence o(ap ) | l. Let
0 0
o(ap ) = l0 . Then l0 | l and p does not divide l0 . Now (al )p = (ap )l = 1G implies

12
 0 0 0 0 0
that o(al ) = 1, or o(al ) = p. If al = 1G , then (aH)l = al H = H implies that
0
p | l0 which is not possible. Hence o(al ) = p.

Theorem 2.1.2 (Cauchy) If G is a finite group such that p | |G|, then ∃ g ∈

G with o(g) = p.
S
Proof. Let partition G into its conjugacy classes. Then G = Z(G) g ∈Z(G)
/ Cg (∗)
where Cg denotes the conjugacy class of g (that is Cg = [g]). If g ∈ Z(G), then
Cg = [g] = {g}. If g 6∈ Z(G), then CG (g) 6= G and |Cg | = [G:CG (g)] > 1. Also
∀g 6∈ Z(G), we have |CG (g)| < |G|. If p | |CG (g)| for some g 6∈ Z(G), then since
|CG (g)| < |G|, the proof will follow by induction. So assume that ∀g 6∈ Z(G)
we have p 6 | |CG (g)|. Since p | |G|, we must have p | [G:CG (g)] for all g 6∈ Z(G).
Using (*) we have the following equation (class equation )
X
|G| = |Z(G) + [G:CG (gi )] (∗∗)
gi 6∈Z(G)

where gi runs over a complete set of class representatives of G outside Z(G).

Since p | |G| and p | [G:CG (gi )] for all gi , we must have p | |Z(G)|. Since Z(G)
is a finite abelian group Z(G) contains an element of order p by Lemma 2.1.1

Theorem 2.1.3 (A characterization of finite p-groups) A finite group G

is a p-group if and only if |G| = pn for some n.

Proof. Assume that G is a finite p-group. If there exists a prime q such that
p 6= q and q | |G|. Then by Cauchy’s Theorem there exists g ∈ G such that
o(g) = q. Since g ∈ G and since G is a p-group, o(g) = pk for some k. That is
q = pk , which is a contradiction. Hence |G| = pn for some n.
Conversely, if |G| = pn for some n, then ∀g ∈ G we have o(g) | pn . Hence G
is a p-group.

Theorem 2.1.4 If G is a non-trivial finite p-group, then |Z(G)| > 1.

Proof. Consider the class equation |G| = |Z(G)| +

P
gi 6∈Z(G) [G:CG (gi )] (∗∗)
where gi runs over all the class representatives of G outside Z(G). Assume that
|G| = pn . Since gi 6∈ Z(G), CG (gi ) < G. Hence |CG (gi )| = pki , where ki < n.
Then we have [G:CG (gi )] = pn−ki , with n − ki > 0. Thus p | [G:CG (gi )] for all
gi and since p | |G|, by (**) we must have p | |Z(G)|. Thus |Z(G)| > 1.
Notice that the above theorem implies that if G is a non-trivial finite p-
group, then |Z(G)| = ps , 1 ≤ s ≤ n.

13
Exercise 2.1.1 Let p be a prime. (i) Prove that a subgroup of a p-group is
also a p-group.
(ii) Prove that a factor group of a p-group is also a p-group.
(iii) If H  G such that H and G/H are p-groups, show that G is also a
p-group.

Exercise 2.1.2 Let G be a finite p-group of order pn . If 0 ≤ k ≤ n, prove that

G contains a normal subgroup of order pk .

Exercise 2.1.3 Let G be a finite p-group and let {1G } 6= H  G. Prove that
H ∩ Z(G) 6= {1G }.

Exercise 2.1.4 If G is a finite p-group, show that every normal subgroup of

order p is contained in Z(G). (Hint: Use Exercise 2.1.3.)

Exercise 2.1.5 *Let G be a finite p-group and let H be a proper subgroup of

order pk . Show that H can be embedded in a subgroup of order pk+1 . Conclude
that every maximal subgroup of G has index p.

Lemma 2.1.5 If G is a non-abelian group, then G/Z(G) is not cyclic.

Proof. Assume that G/Z(G) =< x.Z(G) > for some x ∈ G. Let a, b ∈ G.
0
Then aZ(G) = (xZ(G))k = xk Z(G) and bZ(G) = xk Z(G) for some k, k 0 ∈ ZZ .
0 0
This implies that a = xk z and b = xk z 0 , z, z 0 ∈ Z(G). Now ab = xk zxk z 0 =
0 0 0 0 0
xk xk zz 0 = xk+k zz 0 and ba = xk z 0 xk z = xk xk z 0 z = xk +k zz 0 . Hence ab = ba,
a contradiction.

Corollary 2.1.6 If |G| = pn , then |Z(G)| =

6 pn−1 .

Proof. If |Z(G)| = pn−1 , then |G/Z(G)| = p. Hence G/Z(G) is cyclic and

so by Lemma 2.1.5 G must be an abelian group. But G abelian implies that
Z(G) = G, and this produces a contradiction. Thus Z(G) 6= pn−1 .

Corollary 2.1.7 If G is a finite group of order p2 , then G is abelian.

Proof. Since G is a p-group, {1G } < Z(G) ≤ G. So we have two possibilities

for |Z(G)|, either |Z(G)| = p or |Z(G)| = p2 . Now |Z(G)| = p is not possible
by Corollary 2.1.6. Hence |Z(G)| = p2 which shows that |G| = |Z(G)|; so that
G = Z(G). Therefor G is abelian.

Exercise 2.1.6 If G is a finite group of order p2 , show that G ∼ = ZZ p2 or

G∼= ZZ p × ZZ p . (Hint: Use Corollary 2.1.7 and Theorem I.4.6 of Math 340.)

14
2.2 Sylow Theorems
Definition 2.2.1 (Sylow p-subgroups) Let G be a group and p a prime.
Suppose that P is a subgroup of G. We say that P is a Sylow p-subgroup of
G if P is a maximal p-subgroup. This means that if Q is a p-subgroup of G
such that Q ⊇ P, then Q = P.

Theorem 2.2.1 If Q is a p-subgroup of a group G, there exists a Sylow p-

subgroup of G that contains Q.

Proof. Let F = {H | H is a p-subgroup of G and H ⊇ Q}. Then applying

Zorn’s Lemma to F we deduce that F contains a maximal element. Let P be
a maximal element of F. Since P ∈ F, then P is a p-group (a p-subgroup of G
) and P ⊇ Q. Assume that R is a p-subgroup of G such that R ⊇ P . Then
R ⊇ Q and hence R ∈ F. Now maximality of P implies that R = P. Hence P
is a maximal p-subgroup of G. Thus P is a Sylow p-subgroup of G.

Note 2.2.1 The set of all Sylow p-subgroups of G is denoted by Sylp (G). That
is Sylp (G) = {P | P is a Sylow p-subgroup of G}.

Theorem 2.2.2 Let G be a group and P ∈ Sylp (G). Then every conjugate of
P is also a Sylow p-subgroup of G. Moreover, if for some prime p, G has only
one Sylow p-subgroup P, then P  G.

Proof. Let g ∈ G. Assume that there is a p-subgroup Q in G such that

Q ⊇ gP g −1 . Then g −1 Qg ⊇ P. Since g −1 Qg is a p-subgroup of G (why?),
maximality of P implies that g −1 Qg = P. Hence Q = gP g −1 ; so that gP g −1 is
a Sylow p-subgroup of G.
If |Sylp (G)| = 1, then Sylp (G) = {P } and since gP g −1 ∈ Sylp (G) for all
g ∈ G, we have gP g −1 = P for all g in G. Thus P is a normal subgroup of G.

Lemma 2.2.3 Let P ∈ Sylp (G).

(i) If x ∈ NG (P )/P is a non-identity element then o(x) is not a power of p.
(ii) If g ∈ G such that o(g) = pk for some k, then gP g −1 = P implies that
g ∈ P.

Proof. (i) Assume that o(x) = pk for some k > 0. Then < x > is a p-subgroup
of NG (P )/P . By the correspondence theorem there is H ≤ NG (P ) such that
H ⊇ P and H/P =< x >. Since P and H/P are p- groups so is H. Since H
is a p-subgroup of G with H ⊇ P, we must have H = P. Thus < x > is the

15
identity subgroup of NG (P )/P . That is x is the identity of NG (P )/P which is
a contradiction.
(ii) Consider the natural homomorphism π : NG (P ) −→ NG (P )/P . Let
g ∈ G with o(g) = pk . If gP g −1 = P, then g is an element of NG (P ). Now
k
[π(g)]p k = π(g p ) = π(1G ) = 1 implies that π(g) has order a power of p. By
part (i) we must have that π(g) = 1; so g ∈ Ker(π) = P.

Exercise 2.2.1 Let P be a finite p-group. Assume that α is a homomorphism

from P into SX . Then the size of each orbit of P on X are a power of p.

Theorem 2.2.4 (Sylow’s theorem) Let G be a finite group of order pm q

where p is a prime and (p, q) = 1. Then
(i) All Sylow p-subgroups of G are conjugate.
(ii) The number np of Sylow p-subgroups of G is such that np ≡ 1(mod p)
and np | q.
(iii) If H ≤ G, then H ∈ Sylp (G) if and only if |H| = pm .

Proof. (i) Let X be the set of all conjugates of P, where P is a Sylow p-subgroup
of G. Assume that X = {P1 , P2 , · · · , Pk }, with P1 = P. Then G acts on X by
conjugation. Define ψ : G −→ SX by ψ(g) = ψg where ψg (Pi ) = gPi g −1 for
all g ∈ G, 1 ≤ i ≤ n. Thus ψ is a homomorphism and ψ|P : P −→ SX is also a
homomorphism. Now Exercise 2.2.1 implies that the size of each orbit of P are
a power of p. Obviously {P } is an orbit of size 1, we claim that this is the only
orbit of size 1 under the action of P on X. Assume that ∆ = {Pi } is an orbit
of size 1. Then for all g ∈ P we have ψg (Pi ) = gPi g −1 = Pi , and by Lemma
2.2.3 (ii) we must have g ∈ Pi . and hence Pi = P.
Thus if i 6= 1, then the orbit of P containing Pi has length more than
1 which is a power of p by Exercise 2.2.1. Thus |X| = 1 + lp = k. Hence
k ≡ 1(mod p). (1)
We will show that X is the complete set of Sylow p-subgroups of G, that
is X = Sylp (G). Assume that there exists P 0 ∈ Sylp (G) such that P 0 6∈ X.
By restricting the map ψ to P 0 , the size of each orbit of P 0 on X is a power
of p. Assume that there is an orbit of size one under the action P 0 on X,
say ∆0 = {Pi }, for some 1 ≤ i ≤ k. Then using the same argument shows
that Pi = P 0 , which is a contradiction to the assumption that P 0 6∈ X. So
every orbit of P 0 on X has length more than 1 which is a power of p. Thus
|X| = k = l0 p. (2) But relation (2) contradicts the relation (1). This shows
that there is no Sylow p-subgroup outside X. Hence X = Sylp (G). Therefore
every Sylow p-subgroup of G is a conjugate of P .

16
 (ii) By (i) np = |Sylp (G)| = |X| = k ≡ 1(mod p). Since np = [G : NG (P )] ,
we have np | pm q. Now np | pm q and (np , pm ) = 1 imply that np | q.
(iii) Suppose that |H| = pm . Then obviously H is a maximal p-subgroup of
G; so H ∈ Sylp (G). Conversely, assume that H ∈ Sylp (G). Then by Lemma
2.2.3 and Cauchy’s Theorem [Theorem 2.1.2] p does not divide [NG (H):H].
Since
[G:H] = [G:NG (H)] × [NG (H) : H] = np × [NG (H) : H]

and since p does not divide np , we have p not dividing [G:H]. So that ([G:H], p) =
1 and hence (|G|/|H|, p) = 1. Let |H| = pn for some n ∈ N
| . Then |G|/|H| =

pm−n q. Now (pm−n q, p) = 1 implies that m = n. Therefor |H| = pm .

Exercise 2.2.2 Let G be a finite group and let P ∈ Sylp (G). Show that P  G
if and only if P is the unique Sylow p-subgroup of G.

Theorem 2.2.5 Let G be a finite group and let p be a prime. If pk | |G|, then
G contains a subgroup of order pk .

Proof. If pk | |G|, then |G| = pm q, where m ≥ k and (p, q) = 1. Let P ∈

Sylp (G). Then |P | = pm . Since 0 ≤ k ≤ m by Exercise 2.1.2 there exists H ≤ P
such that |H| = pk . Now H ≤ P and P ≤ G, imply H ≤ G.

Exercise 2.2.3 Let G be a group and H ∈ Sylp (G). Let x ∈ G such that
x 6∈ H and o(x) = pn for some n ∈ N
| . Then x 6∈ NG (H).

Lemma 2.2.6 Let H ≤ G. Assume that P, Q ∈ Sylp (H) such that P 6= Q. If

P ∗ and Q∗ are Sylow p-subgroups of G such that P ∗ ⊇ P and Q∗ ⊇ Q, then
P ∗ 6= Q∗ .

Proof.
If P ∗ = Q∗ , then hP, Qi ≤ P ∗ = Q∗ . Hence hP, Qi is a p-group. Since P
and Q are subgroups of H, we have hP, Qi ≤ H. Now since hP, Qi ⊇ P and P
is a maximal p-subgroup of H, we have hP, Qi = P. Similarly we can show that
hP, Qi = Q. This produces the contradiction P = Q.

Theorem 2.2.7 If H ≤ G, then np (G) ≥ np (H).

Proof. If P ∈ Sylp (H), then by Theorem 2.2.1 there exists P ∗ ∈ Sylp (G) such
that P ∗ ⊇ P . Now apply Lemma 2.2.6.

Theorem 2.2.8 (Frattini Argument) Let G be a finite group with H  G.

If P ∈ Sylp (H), then G = HNG (P ).

17
Proof. Obviously HNG (P ) ⊆ G (in fact HNG (P ) ≤ G, why?). Let g ∈ G.
Then gP g −1 ⊆ gHg −1 = H. Hence gP g −1 is a Sylow p-subgroup of H. Since
all Sylow p- subgroups of H are conjugate in H, there exists h ∈ H such that
hP h−1 = gP g −1 . So P = h−1 gP (h−1 g)−1 and hence h−1 g ∈ NG (P ). This
shows that g ∈ HNG (P ). Thus G ⊆ HNG (P ) and therefore G = HNG (P ).

Exercise 2.2.4 Let G be a finite group and, for each prime divisor p of |G|,
choose a Sylow p-subgroup of G. Prove that G is generated by these subgroups.

Exercise 2.2.5 Let G be a finite group, and suppose that, for every prime
divisor p of |G|, every Sylow p-subgroup of G is normal in G. Prove that G is
the direct product of its Sylow subgroups.

Exercise 2.2.6 *Let G be a finite group and let P be a Sylow p-subgroup of

G. If H / G, prove that H ∩ P is a Sylow p-subgroup of H and HP/H is a
Sylow p-subgroup of G/H. (Hint: Compare orders.)

Exercise 2.2.7 Prove that a Sylow 2-subgroup of A5 has exactly five conju-
gates.

Exercise 2.2.8 Show that there is a finite group G with Sylow p-subgroups
A, B, and C, for some prime p, such that A ∩ B = {1G } and A ∩ C 6= {1G }.
(Hint: Take G = S3 × S3 .)

Exercise 2.2.9 Prove that every subgroup of order 8 in S4 contains V4 .

Example 2.2.1 Find a Sylow p-subgroup for the group GL(3, p).

Solution. In general

|GL(n, q)| = (q n − 1)(q n − q) · · · (q n − q n−1 ).

So
GL(3, p) = (p3 − 1)(p3 − p)(p3 − p2 ) = p3 (p3 − 1)(p2 − 1)(p − 1).
 
1 a b
If H ∈ Sylp (G), then |H| = p3 . Consider H = {
 
 0 1 c  | a, b, c ∈ Z
Z p }.


0 0 1
3
Then H ≤ GL(3, p) and |H| = p . Hence H ∈ Sylp (G).
We can use the results discussed so far to determine the properties of groups
of small orders. Sylow’s theorem is frequently used for the non-existence of
simple groups of various orders.

18
Example 2.2.2 There is no simple group of order 312.

Solution: Let G be a group of order 312 = 23 × 3 × 13. Then n13 ≡ 1 (mod 3)

and n13 | 23 × 3. This shows that n13 = 1. Hence if P ∈ Syl13 (G), then P  G.
Thus G is not simple.

Example 2.2.3 There is no simple group of order 30.

Solution: Let G be a group of order 30. Since n5 ≡ 1 (mod 5) and n5 | 6,

we must have n5 ∈ {1, 6}. Similarly n3 ≡ 1 (mod 3) and n3 | 10 imply that
n3 ∈ {1, 10}. If G is simple, then n3 6= 1 6= n5 . Hence n3 = 10 and n5 = 6.
Let Syl5 (G) = {Hi | 1 ≤ i ≤ 6}. Since Hi ∩ Hj ≤ Hi , for i 6= j we have
Hi ∩ Hj = {1G }. This shows that we have 6 × 4 = 24 elements of order 5 in
G. Similarly we will show that there are 10 × 2 = 20 elements of order 3 in
G. So |G| ≥ 1 + 24 + 20 = 45, which is a contradiction to the assumption that
|G| = 30.

Exercise 2.2.10 Show that there are no simple groups of order 12, 28 or 56.

Exercise 2.2.11 Use Sylow’s theorem to show that any group of order 15 is
cyclic.

Note 2.2.2 (Dihedral group D2n ) For n ≥ 2, The dihedral group D2n
is a group of order 2n generated by elements a, b with the following relations

an = 1G , b2 = 1G , bab = a−1 .

Theorem 2.2.9 Let G be a finite group of order 2p where p is an odd prime.

Then G is either Cyclic or Dihedral.

Proof. See Mathematics 340 notes.

Theorem 2.2.10 If G is a finite group of order pq, where p > q and p, q are
primes. Then either G is cyclic or G =< a, b > with relations ap = bq = 1G
and b−1 ab = ar , where r 6≡ 1 (mod p), rq ≡ 1 (mod p) and q | p − 1.

Proof. By Cauchy’s theorem [Theorem 2.1.2] there exist a, b ∈ G such that

o(a) = p and o(b) = q. Let P =< a > . Then P is a Sylow p-subgroup of G and
np ≡ 1 (mod p) with np | q. Hence there exists k ≥ 0 such that np = 1 + kp.
So q = l(1 + kp) for some l. Since q is a prime, we have two possibilities, either
l = 1 and 1 + kp = q or l = q and 1 + kp = 1. Since p > q, 1 + kp = q is not
possible. Hence we must have 1 + kp = 1. So k = 0 and hence np = 1. This

19
implies that P  G. Now consider Q =< b > . Then Q ∈ Sylq (G) and by the
same argument we obtain nq = 1 + k 0 q where k 0 = 0 or 1 + k 0 q = p.
Case 1: If k 0 = 0, then nq = 1 and hence Q  G. Since P and Q are
normal subgroups of G and P ∩ Q = {1G } and P Q = G, we have G ∼
= P ×Q ∼
=
ZZ p × ZZ q ∼
= ZZ pq .
Case 2: If 1 + k 0 q = p. Then q | p − 1 and nq = p 6= 1 Thus Q is not
normal in G. Since P  G, b−1 ab = ar for some r. If r ≡ 1 (mod p), then
ar = a; so b−1 ab = a. This shows that G is abelian, and hence Q  G which is a
q
contradiction. Hence r 6≡ 1 (mod p). Now b−1 ab = ar implies that b−q abq = ar ;
q
so that a = ar . Hence rq ≡ 1 (mod p).

Note 2.2.3 (Quaternion group Qn ) Let Q be a group of order 8 generated

by elements a, b with relations

a4 = 1G , b2 = a2 , b−1 ab = a−1 .

Then Q is called the Quaternion group.

For n ≥ 3 we can define the Generalized Quaternion group Qn to be a
group of order 2n generated by two elements a, b with relations
n−1 n−2
a2 = 1G , b2 = a2 , b−1 ab = a−1 .

Theorem 2.2.11 D8 and Q are the only non-abelian groups of order 8.

Proof. See Mathematics 340 notes (Exercise page 28).

Theorem 2.2.12 Assume that G is a finite group of order 12. If G is not

isomorphic to A4 , then there exists an element g ∈ G such that o(g) = 6.

Proof. Let H ∈ Syl3 (G). Then H is a cyclic group of order 3. Assume that
H =< a >, where o(a) = 3. Then [G:H] = 4 and G acts on the left cosets of
H. Consider the homomorphism φ : G → S4 which gives this representation.
We know that
gHg −1 .
\
Ker(φ) =
g∈G

(See Theorem 1.1.3, The generalized Cayley’s theorem). Since Ker(φ) ≤ H by

Exercise 1.1.2, we have that Ker(φ) = {1G } or Ker(φ) = H. If Ker(φ) = {1G },
then G is isomorphic to a subgroup of S4 . Since |G| = 12, we must have G ∼
= A4
[Note: if N ≤ S4 such that |N | = 12, then N = A4 (check)]. But this contradicts
our assumption that G 6=∼ A4 . Hence Ker(φ) = H. Thus H  G and n3 = 1.

20
This implies that a and a2 are the only elements of order 3 in G. So we have
the following cases:
(i) a is conjugate to a2 in G.
(ii) a is not conjugate to a2 in G.
In case (i) |[a]| = 2 and we have |CG (a)| = 12/2 = 6.
In case (ii) |[a]| = 1 and we have |CG (a)| = 12/1 = 12.
Hence in both cases |CG (a)| is even . Thus ∃ b ∈ CG (a) such that o(b) = 2.
Now since ab = ba, we have o(ab) = 6.

Exercise 2.2.12 If n 6= 4, show that An is the only proper normal subgroup

of Sn .

Theorem 2.2.13 (Non-abelian groups of order 12) Up to isomorphism we

have only three non-abelian groups of order 12.

Proof. We know that A4 is a non-abelian group of order 12. Let G be a

non-abelian group of order 12 that is not isomorphic to A4 . Then by Theorem
2.2.12 there exists an element s ∈ G of order 6. Let H =< s > . Then H is
cyclic of order 6 and H = {1G , s, s−1 , s2 , s−2 , s3 }. Since G/H ∼
= ZZ 2 , H  G.
Since |G| = 22 × 3, we must have two possibilities for a Sylow 2-subgroup of G,
that is if P is a Sylow 2-subgroup of G, then P ∼= V4 or P ∼= ZZ 4 .
(i) If P ∼ = V4 , then ∃ t ∈ G such that o(t) = 2 and t 6∈ H (i.e t 6= s3 ).
Then G =< t, s > and we have t2 = s6 = 1G . Since G is not abelian and since
tst ∈ H, we have tst = s−1 . Hence G ∼
= D12 .
(ii) If P ∼
= ZZ 4 , then P =< t >, o(t) = 4. Obviously t 6∈ H. Since G =
H ∪ tH, t2 ∈ H. Now o(t2 ) = 2 implies that t2 = s3 . Since H  G, we have
t−1 st ∈ H. So t−1 st = s or t−1 st = s−1 . If t−1 st = s, then st = ts which implies
that G =< t, s > is an abelian group. Therefore t−1 st = s−1 . That is (st)2 = t2 .
Hence G =< t, s > is a group with relations s6 = t4 = 1G , t2 = s3 = (st)2 .

Note 2.2.4 The group of order 12 defined by G =< a, b > and the relations
given in (ii) above is denoted by T . In fact we have five non-isomorphic groups
of order 12, namely two abelian groups

ZZ 12 , ZZ 6 × ZZ 2 ∼
= ZZ 2 × ZZ 2 × ZZ 3

and three non-abelian groups

A4 , D12 ∼
= ZZ 2 × S3 , T.

21
Theorem 2.2.14 If H is a p-subgroup of G which that is contained in exactly
one Sylow p-subgroup P of G, then NG (H) ⊆ NG (P ).

Proof. Let x ∈ NG (H). Then H = xHx−1 ⊆ xP x−1 . Since P ∈ Sylp (G), xP x−1 ∈
Sylp (G); so xP x−1 = P . Hence x ∈ NG (P ).

Theorem 2.2.15 Assume that np (G) is finite and P ∈ Sylp (G). If H ≤ G

such that H ⊇ NG (P ), then NG (H) = H.

Proof. Since np (G) ≥ np (H), np (H) is also finite. Let x ∈ NG (H). Then since
H ⊇ NG (P ) ⊇ P , P ∈ Sylp (H). Since x ∈ NG (H) and P ⊆ H, P x ⊆ H x = H.
Thus P x is also a Sylow p-subgroup of H. hence ∃ h ∈ H such that P = (P x )h .
This shows that xh ∈ NG (P ). Since NG (P ) ⊆ H, we must have xh ∈ H.
Therefore x ∈ H and we deduce that NG (H) = H.

Theorem 2.2.16 Assume that np (G) is finite and P ∈ Sylp (G). If H ≤ G

such that H ⊇ NG (P ), then [G:H] ≡ 1 (mod p).

Proof. Since np (G) ≥ np (H), np (H) is finite. Thus np (G) ≡ 1 (mod p) and
np (H) ≡ 1 (mod p). Since H ⊇ NG (P ) ⊇ P, P ∈ Sylp (H). Thus [G:NG (P )] ≡
1 (mod p) and [H:NH (P )] ≡ 1 (mod p). Since NH (P ) = H ∩ NG (P ) = NG (P )
and since [G:NG (P )] = [G:H] × [H:NG (P )], it follows that [G:H] ≡ 1 (mod p).

Exercise 2.2.13 (i) If G is a group of order 119, show that G is cyclic.

(ii) Let G be a group of order 595. Show that G has a unique normal
subgroup N of order 5. By considering G/N , or otherwise, deduce that G is
cyclic.

Exercise 2.2.14 Prove that (i) a Sylow 2-subgroup of S5 is isomorphic to D8 ,

(ii) a Sylow 2-subgroup of S6 is isomorphic to D8 × ZZ 2 .

Exercise 2.2.15 * (i) Find Z(D2n ). (Hint: First show that every element in
D2n is of the form ai bj .)
(ii) Deduce that Z(D2n ) is non-trivial if and only if n is even.

Exercise 2.2.16 Let G be a finite group and a, b ∈ G be two distinct elements

of order 2 in G. Let H = ha, bi. Show that H ∼
= D2n for some n.

Exercise 2.2.17 Show that D8 is not generated by its elements of largest order.

Exercise 2.2.18 Let G be a finite group with normal subgroups H and K. If

∼ G/K, is H ∼
G/H = = K?

22
Exercise 2.2.19 For every divisor d of 24, show that ∃H ≤ S4 having order d.
Moreover, if d 6= 4, show that any two subgroups of order d in S4 are isomorphic.

Exercise 2.2.20 Prove that D6 ∼

= S3 .

Exercise 2.2.21 Prove that D12 ∼

= S3 × ZZ 2 .

Exercise 2.2.22 If G is a non-abelian p-group of order p3 , prove that Z(G) =

G0 , the commutator subgroup of G.

Exercise 2.2.23 Let G be a Sylow p-subgroup of GL(3, p) (see Example 2.2.1).

If p is an odd prime, show that xp = 1G = I3 for all x ∈ G. (Hint: Expand the
polynomial (x − I3 )p in ZZ p [x]).

Exercise 2.2.24 What is the order of a Sylow p-subgroup of GL(n, p)?

Exercise 2.2.25 *Give an example of two non-isomorphic groups G and H

such that, for each integer d > 0, the number of elements in G of order d is the
same as the number of of elements in H of order d.

Exercise 2.2.26 Let G be a finite group and let P ∈ Sylp (G). If H is a normal
subgroup of G containing P , then prove that P  H implies P  G. (Hint: Use
the Frattini argument.)

Example 2.2.4 Let G be a group of order p2 q where p and q are distinct

primes. Show that G is not simple.

Solution: We will show that G has a normal Sylow subgroup (for either p
or q ). Let P ∈ Sylp (G) and Q ∈ Sylq (G).
(i) If p > q, then since np ≡ 1 (mod p) and np | q, we must have np = 1 + kp
and q = k 0 np for some k ≥ 0 and k 0 > 0. This gives q = k 0 (1 + kp) = k 0 + k 0 kp.
Hence k = 0. Thus np = 1. So P  G.
(ii) Let p < q. If nq = 1, then Q  G. Assume that nq 6= 1. Then
nq = 1 + kq, k > 0 and nq | p2 . Since nq | p2 , we have nq = p or nq = p2 . Since
p < q and nq = 1 + kq with k > 0, nq = p is not possible. Thus nq = p2 and we
have p2 = 1 + kq. So kq = (p2 − 1) = (p − 1)(p + 1). Hence q | (p − 1)(p + 1).
Since q is a prime we have q | p − 1 or q | p + 1. Since q > p, we must have
q | p + 1. Now again q > p implies that q = p + 1. Thus p = 2 and q = 3 and
|G| = 12. If G ∼
= A4 , then G has a normal Sylow 2-subgroup isomorphic to V4 .
If G 6∼
= A4 , then by Theorem 2.2.12, G contains an element of order 6. Let g ∈ G

23
such that o(g) = 6 and assume that H =< g >= {1G , g, g −1 , g 2 , g 3 , g 4 } ∼
= ZZ 6 .
Let Q = {1G , g 2 , g 4 }. Then Q ∼
= ZZ 3 and Q ∈ Syl3 (G). Obviously Q  H and
H  G. Now since aQa−1 ≤ aHa−1 = H for all a ∈ G and Q is the only Sylow
3-subgroup of H, we must have that aQa−1 = Q. Thus Q  G.

Example 2.2.5 If G is a simple group of order 60, then G ∼

= A5 .

Solution: Let G be simple and |G| = 60 = 22 × 3 Then it is not difficult to

see that n2 ∈ {3, 5, 15}. Let P ∈ Syl2 (G) and let N = NG (P ). Then we have
n2 = [G:N ].
If n2 = 3, then |N | = 20 and we must have a homomorphism ρ : G −→ S3
with Ker(ρ) = K, where K ≤ N < G. Since G is simple, we must have
K = {1G } and hence G is isomorphic to a subgroup of S3 . This is impossible
since |G| = 60 does not divide 6. Thus n2 ∈ {5, 15}.
If n2 = 5, then we have a homomorphism ρ : G −→ S5 with Ker(ρ) = K =
{1G }. Thus we can say that G is a subgroup of S5 . If G 6= A5 , then S5 = GA5 .
Since ZZ 2 ∼
= S5 /A5 = GA5 /A5 ∼= G/A5 ∩ G, A5 ∩ G is a normal subgroup of
index 2 in G, which is a contradiction. Thus G is a isomorphic to a subgroup
of A5 . Now |G| = 60 = |A5 | implies that G ∼
= A5 .
Finally, suppose that n2 = 15. Assume that ∀P, Q ∈ Syl2 (G) we have
P ∩ Q = {1G }. Then we have 15 × (4 − 1) = 45 elements in Sylow 2 - subgroups
of G. Since n5 ≡ 1 (mod 5) and n5 | 12, we must have n5 = 6. Hence we have
6 × (5 − 1) = 24 elements of order 5 in G. So far we produced 45 + 24 = 69 > 60
elements in G which is a contradiction. Hence ∃P, Q ∈ Syl2 (G) with P ∩ Q 6=
{1G }. Since |P | = |Q| = 4, P and Q are abelian. Let M = NG (P ∩ Q). Then
P, Q ≤ M and hence 4 | |M |. Since G is simple, M 6= G. If |M | = 4, then
M = P = Q, which is not possible. Hence |M | > 4, so that |M | ∈ {12, 20}.
If |M | = 20, then G has a subgroup of index 3 which is not possible (why?).
Thus M | = 12 and [G:M ] = 5. By a similar argument used in the case n2 = 5
[applied to M in place of N ] we can show that G ∼
= A5 . But this produces a
contradiction because n2 (A5 ) = 5 (see Exercise 2.2.7).

Example 2.2.6 There are no non-abelian simple groups of order less than 60.

Solution: Let G be a non-abelian group such that |G| = n with n < 60.
(i) If n = p, for some prime p, then G ∼
= ZZ p . Thus G is abelian, a contra-
diction.
(ii) If n = p2 , then G is abelian, a contradiction.

24
 (iii) If n = pk , k ≥ 3, then Z(G)  G. Since G is not abelian, Z(G) 6= G.
Also Z(G) 6= {1G } since G is a p-group. Thus Z(G) is a non-trivial normal
subgroup of G and hence G is not simple.
(iv) Assume that n = 2m × pk , where p is an odd prime and m > 0. Then
m ∈ {1, 2, 3, 4} and k ∈ N
| .

if m = 1, then np ≡ 1 (mod p) and np | 2 imply that np = 1. So G is not

simple.
If m = 2, then np ≡ 1 (mod p) and np | 4 imply that np = 1 or (np = 4
and p = 3). If G is simple, then G is isomorphic to a subgroup of S4 . Now
|G| = 22 × 3k implies that k = 1 and G ∼= A4 . This produces a contradiction
since A4 is not simple.
If m = 3, then np ≡ 1 (mod p) and np | 8 imply that np = 1 or (np = 4 and
p = 3) or (np = 8 and p = 7). Suppose that G is simple. If np = 4 and p = 3,
then G is isomorphic to a subgroup of S4 and |G| = 23 × 3k . In this case we
must have k = 1 and hence G ∼= S4 , a contradiction since S4 is not simple. If
np = 8 with p = 7, then |G| = 23 × 7 = 56. But Exercise 2.2.10 shows that G
can not be simple.
Finally, if m = 4, then we must have |G| = 24 × 3 In this case n2 | 3 and
n2 ≡ 1 (mod 2) imply that n2 ∈ {1, 3}. If G is simple, then n2 = 3, and G is
isomorphic to a subgroup of S3 . But this is impossible, since 48 > 6.
(v) If n = 2b × pm × q k , where p, q are distinct odd primes and m 6= 0 6= k.
If b = 0 and m = k = 1, then by Theorem 2.2.10, G is not simple [because for
example if if p > q, then P =< a >
G]. If b = 0, m = 2 and k = 1, then
|G| = p2 × q and G is not simple by Example 2.2.4.
Since |G| < 60, by the above results the remaining possibilities for |G| are
30 = 2 × 3 × 5 or 42 = 2 × 3 × 7. If |G| = 30, then G is not simple by Example
2.2.3. If |G| = 42, then n7 ≡ 1 (mod 7) and n7 | 6. Hence n7 = 1 and G is not
simple.

Remark 2.2.1 Simple groups play an important role in the theory of groups.
Simple groups are not easy groups, they have complicated structures. They
play a role analogous to that of prime numbers in the study of ZZ . The search
for finite simple groups started in 1870 with C. Jordan [1838 - 1922] who es-
tablished simplicity of An and linear groups over fields of prime power order.
The Classification of Finite Simple groups was completed in 1981. This
problem has a history of 150 years. Its proof is made of 15000 journal pages.
The classification theorem is :
CFS: Every finite simple group is isomorphic to a group of prime order,

25
an alternating group, one of the finite groups of Lie type or one of 26
sporadic simple groups.

Theorem 2.2.17 (i) If G is a finite group that has no subgroup of index 2 and
G ≤ Sn for some n, then G ≤ An .
(ii) If P ∈ Sylp (Sn ) for some odd prime p, then P ∈ Sylp (An ) and further-
more |NAn (P )| = 12 |NSn (P )|.

Proof. (i) If G 6≤ An , then An < GAn ≤ Sn , and hence GAn = Sn . But since
GAn /An ∼
= G/(An ∩ G), we have 2 = [Sn :An ] = |GAn /An | = |G/An ∩ G|. This
implies that An ∩ G is a subgroup of index 2 in G, a contradiction.
(ii) If P ∈ Sylp (Sn ) with p odd, then |P | is odd. Hence P has no subgroup
of index 2. Since P ≤ Sn , by (i) we have P ≤ An . Thus P ∈ Sylp (An ). Now
using the Frattini argument we have Sn = NSn (P ).An . So NSn (P ) 6≤ An and
|NSn (P )| × |An | |NSn (P )|
|Sn | = = × An
|NSn (P ) ∩ An | |NSn (P ) ∩ An |
and hence [NSn (P ):NSn (P ) ∩ An ] = 2. Now since NAn (P ) = NSn (P ) ∩ An , we
deduce that [NSn (P ):NAn (P )] = 2.

Example 2.2.7 Show that there is no simple group of order 264.

Solution : Assume that G is a simple group of order 264. Then |G| =

264 = 23 × 3 × 11 and it is not difficult to see that n11 = 12. Using the
representation on the left cosets of NG (P ) where P ∈ Syl11 (G), we would have
G is isomorphic to a subgroup of S12 . Since G is simple, it has no subgroup of
index 2. Thus G ≤ A12 , by Theorem 2.2.17.
Since n11 = 12 = [G:NG (P )], we have |NG (P )| = 264/12 = 22. Now by The-
orem 2.2.17 (ii) and Note 2.2.5 (see bellow) we have |NA12 (P )| = 21 |NS12 (P )| =
1
2 (11 × 10) = 55. Since G ≤ A12 and since P ∈ Syl11 (G), P ≤ A12 . Thus
NG (P ) ≤ NA12 (P ), so that |NG (P )| must divide |NA12 (P )| = 55. However 22
does not divide 55, a contradiction. Thus G is not simple.

Note 2.2.5 Since Sylow 11- subgroups of S12 are precisely the groups gener-
ated by a 11-cycle, and since distinct Sylow 11-subgroups of S12 intersect in the
identity, we have
The number of 11-cycles of S12
n11 (S12 ) =
The number of 11-cycles in a Sylow 11-subgroup
(12!)/11 12!
= = .
10 11 × 10
Hence if P ∈ Syl11 (S12 ), we have |NS12 (P )| = 11 × 10 = 110.

26
Note 2.2.6 If P ∈ Sylp (Sn ), where P is cyclic of order p (in this case p2 does
n(n−1)···(n−p+1)
not divide n!), then we can show that np (Sn ) = p(p−1) . In this case if
n = p or n = p + 1, we have |NSn (P )| = p(p − 1).

Exercise 2.2.27 (i) Show that there is no simple group of order 3393. (Hint:
Use Sylow 3-subgroups.)
(ii) Show that there is no simple group of order 1755. (Hint: Use Sylow 3-
subgroups to show G ≤ S13 and look at the normalizer of a Sylow 13-subgroup.)

Example 2.2.8 (A simple group of order 168) Consider the general lin-
ear group GL(2, 7), the set of all non-singular 2 × 2 matrices over GF (7) = ZZ 7 .
Consider the following subgroup of GL(2, 7), called the Special Linear Group
SL(2, 7) :

SL(2, 7) = {A | A ∈ GL(2, 7), det(A) = 1}

  
 a b 
=   | ad − bc = 1, a, b, c, d ∈ Z
Z7 .
 c d 

If a = 0, then bc = −1. Hence b 6= 0 and there are 6 choices for b (notice

that c =−b−1 ). 
Since we have 7 choices for d, we have 7×6 matrices in SL(2, 7)
0 b
of type  .
c d
If a 6= 0, then d = a−1 (1 + bc) and we have 6 choices for a. Since there are
7 choicesfor b and c respectively, a total of 6 × 72 matrices in SL(2, 7) of type
a b
  with a 6= 0 is produced.
c d
Hence we have 7 × 6 + 6 × 72 = 336 = 168 × 2 elements in SL(2, 7).
A simple calculation will show that Z(SL(2, 7)) is equal to the group {I2 , −I2 },
where I2 is the 2 × 2 identity matrix. Since Z(SL(2, 7))  SL(2, 7), we have
the factor group SL(2, 7)/Z(SL(2, 7)). This factor group is called the Special
Projective Linear group P SL(2, 7). Then |P SL(2, 7)| = (168 × 2)/2 = 168.
By calculating the conjugacy classes of SL(2, 7), we are able to calculate the
conjugacy classes of P SL(2, 7) which we list in Table 1. (See Exercise 2.2.28
bellow.)

Table 1

1 2 3 4 7 7 order of elements
I2 A B C D D−1 class representative
1 21 56 42 24 24 size of each class

27
 Now we show that P SL(2, 7) is simple: Assume that N  P SL(2, 7) such
that |N | 6= 1 and N 6= P SL(2, 7). Then N must be a union of conjugacy
classes of P SL(2, 7). Obviously |N | ≤ 84 and |N | | 168. Since N contains
I2 and at least one other conjugacy class, |N | ≥ 22. Hence 22 ≤ |N | ≤ 84.
Since |N | | 168, we must have |N | ∈ {24, 28, 42, 56, 84}. This shows that |N |
is even. Since I2 ∈ N , we deduce that N contains the conjugacy class of
A. If D ∈ N , then D−1 ∈ N . Then N has at least 1 + 21 + 24 + 24 = 70
elements. Since 70 does not divide 168, N must contain class B or class C.
Then |N | > 84 which is not possible. Hence D, D−1 6∈ N . Now assume that
|N | = 1 + 21 + 42a + 56b where {a, b} = {0, 1}. It is not difficult to see that
1 + 21 + 42a + 56b 6∈ {24, 28, 42, 56, 84}. Hence P SL(2, 7) is simple.
It can be shown that there is up to isomorphism, a unique simple group of
order 168. But it is not always the case that there is at most one simple group
of a given order. In fact Schottenfels in 1900 proved that A8 and P SL(3, 4)
are not isomorphic. These two simple groups have the same order 20160. (See
J.J. Rotman, page 176.)

Exercise 2.2.28 Let G = P SL(2, 7) Use GAP to compute the conjugacy

classes of G. For each class representative g ∈ G, determine o(g), |[g]| and
|CG (g)|. If P ∈ Syl2 (G), show that P ∼
= D8 .

28
Chapter 3

Finite Direct Products

3.1 External and Internal Products

Definition 3.1.1 Let H and K be two groups. Define the external direct
product of H and K by

H × K = {(h, k) | h ∈ H, k ∈ K}, (h, k) ◦ (h0 , k 0 ) = (hh0 , kk 0 ).

Then (H × K, ◦) is a group and 1H×K = (1H , 1K ).

Proposition 3.1.1 (i) {1H } × K  H × K, H × {1K }  H × K.

(ii) {1H } × K ∼
= K and H × {1K } ∼
= H.
(iii) {1H } × K ∩ H × {1K } = {1H×K }.
(iv) H × K ∼ = K × H.
(v) H × K is abelian if and only if H and K are abelian.

Proposition 3.1.2 If (m, n) = 1, then ZZ mn ∼

= ZZ m × ZZ n .

Proof. We know that ZZ k , ∀k ∈ ZZ , is a cyclic group. Let ZZ m =< x > and

ZZ n =< y >. Then o(x) = m and o(y) = n. We claim that o(x, y) = mn. Let
o(x, y) = k. Then since (x, y)mn = (xmn , y mn ) = (1 ZZ m , 1 ZZ n ), we have k | mn.
Also (1 ZZ m , 1 ZZ n ) = (x, y)k = (xk , y k ) implies that xk = 1 ZZ m and y k = 1 ZZ n .
Hence m | k and n | k. Since (m, n) = 1 and m | k and n | k, mn | k. Therefore
k = mn and hence o(x, y) = mn. Since ZZ m × ZZ n has mn elements and since
(x, y) ∈ ZZ m × ZZ n we have ZZ m × ZZ n =< (x, y) >∼
= ZZ mn .

Theorem 3.1.3 Let G be a group with H, K  G. Assume that G = HK and

H ∩ K = {1G }. Then G ∼
= H × K.

29
Proof. Define ψ : G −→ H × K by ψ(hk) = (h, k). Then we claim that ψ is
an isomorphism.
ψ is well defined: Suppose that g = hk = h0 k 0 . Then we have h0−1 h = k 0 k −1 .
Thus h0 −1 h = k 0 k −1 ∈ H ∩ K; so h0−1 h = k 0 k −1 = {1G } and hence h = h0 and
k = k 0 . Thus (h, k) = (h0 , k 0 ).
ψ is a homomorphism: Let g, g 0 ∈ G. Then g = hk and g 0 = h0 k 0 . We
have gg 0 = hkh0 k 0 . (*) Firstly we will show that elements of H commute with
elements of K. Consider h ∈ H and k ∈ K. Then hkh−1 k −1 = h(kh−1 k −1 ) ∈
H, since H  G. Also hkh1 k −1 = (hkh−1 )k −1 ∈ K, since K  G. Thus
hkh1 k −1 = 1G and hence hk = kh ∀h ∈ H and ∀k ∈ K.
Now using the relation (*) we get gg 0 = hh0 kk 0 , hence we have

ψ(gg 0 ) = (hh0 , kk 0 ) = (h, k)(h0 , k 0 ) = ψ(g)ψ(g 0 ).

It is routine to check that ψ is one-to-one and onto. (Check)

Note 3.1.1 All conditions in the Theorem 3.1.3 are necessary. For example
consider the group G = S3 with H = A3 and K =< (12) >. Then H  G and
H ∩ K = {1G } and G = A3 K , since |A3 K| = 3×2 = 6 = |G|. But G ∼
1 6 A3 × K,
=
because A3 × K ∼
= ZZ 3 × ZZ 2 ∼
= ZZ 6 and S3 6∼
= ZZ 6 . Notice that K is not normal
in S3 .

Exercise 3.1.1 Let G = H × K. Assume that H1  H and K1  K Show that

H1 × K1  G and
G/H1 × K1 ∼
= H/H1 × K/K1 .

Exercise 3.1.2 Let G = H × K. Show that

G/H × {1K } ∼
=K and G/{1H } × K ∼
= H.

Definition 3.1.2 (Internal Direct Product ) Let G be a group and H, K

G. We say that G is the internal direct product of H and K if G = HK
and H ∩ K = {1G }.

Note 3.1.2 If G = H × K, then G is the internal direct product of H × {1K }

and {1K } × K. [Use Proposition 3.1.1]

Note 3.1.3 If G = H × K = H × L, then K ∼

= L by Exercise 3.1.2

Example 3.1.1 Let p be a prime and n ∈ N

| . Consider the group

Epn = ZZ p × ZZ p × · · · × ZZ p (n f actors).

30
 Then Epn is an abelian group of order pn with the property that g p = 1Epn
for all g ∈ Epn . Because if g ∈ Epn , then g = (g1 , g2 , · · · , gn ) where gi ∈ ZZ p .
Since ZZ p is a cyclic group of order p, gi p = 1 ZZ p . Hence

g p = (g1 , g2 , · · · , gn )p = (g1 p , g2 p , · · · , gn p ) = (1 ZZ p , 1 ZZ p , · · · , 1 ZZ p ).

This group is called the Elementary abelian group of order pn .

3.2 Basis Theorem

In this section all group are Abelian. Additive notation will be used throughout
this section. The group operations will be denoted by (+), the identity by 0,
the inverse of an element a will be −a. We will use na for an and a + H for the
coset aH. However, the factor groups will be denoted G/H. The term direct
sum will be used for direct product and hence we use the notation H ⊕ K
for H × K.

Note 3.2.1 In this additive notation we have (a + b)n = n(a + b) = na + nb

for all a, b ∈ G.

Note 3.2.2 If X ⊆ G such that X is not empty. Then

< X > = {x1 a1 x2 a2 · · · xn an | xi ∈ X, ai ∈ ZZ n ∈ N

| }

= {a1 x1 + a2 x2 + · · · + an xn | xi ∈ X, ai ∈ ZZ n ∈ N
| },

the set of all linear combinations of elements of X with integer coeficients.

Definition 3.2.1 (p-primary groups) Let G be an abelian group. If G is a

p-group, then we say that G is a p-primary group.

Theorem 3.2.1 (Primary Decomposition) Let G be a finite abelian group.

Then G is a direct sum of p-primary groups.

Proof. Let Gp denote the set of all elements of G of order some power of p.
That is Gp = {x | x ∈ G, o(x) = pα , for some α}. Then Gp is a subgroup of G:
Because if x, y ∈ Gp , then o(x) = pα and o(y) = pβ , for some α and β. Then
α+β α+β
(xy −1 )p = xpα+β (y −1 )p , since G is abelian
α
−1 p β p
= (xpα )p β [(y ) ] = 1G

31
implies that o(xy −1 ) | pα+β . Hence xy −1 ∈ Gp and therefor Gp ≤ G.
Since G is abelian, Gp G. We will show that G = Gp . Let g ∈ G. Since
L
p | |G|
G is finite o(g) = n < ∞. Assume that n = p1 α1 p2 α2 · · · pk αk where pi are dis-
tinct primes . Let ni = n/pαi . It is not difficult to see that gcd(n1 , n2 , · · · , nk ) =
1 and hence ∃mi , 1 ≤ i ≤ k such that m1 n1 + m2 n2 + · · · + mk nk = 1. Thus
k
X
mi ni g = g (∗).
i=1

Since pi αi mi ni = nmi , we have pi αi (mi ni )g = (nmi )g = 0. This shows that

mi ni g ∈ Gpi for all 1 ≤ i ≤ k. Now
k
X k
X X
g= mi ni g ∈ Gpi ⊆ Gp
i=1 i=1 p | |G|
P
implies that G = Gp . In order to show that this sum is a direct sum, we
p | |G|
must show that for all prime q such that q | |G| we have
X
Gq ∩ Gp = {0} (∗∗).
p | |G|,p6=q

Assume that g belongs to the left hand side of (∗∗). Then q α g = 0 for some
gp , where gp ∈ Gp . Now gp ∈ Gp implies that ∃αp such that
P
α and g =
p | |G|,p6=q
pαp gp = 0. Let t = pαp . Then (t, q α ) = 1 and tg =
Q P
tgp = 0.
p | |G|,p6=q p | |G|,p6=q
Since (t, q α ) = 1 ∃a, b ∈ ZZ such that at + bq α = 1. This gives atg + bq α g = g.
Since q α g = tg = 0, we have g = 0. Therefore G =
L
Gp .
p | |G|

Lemma 3.2.2 Assume that G is an abelian group such that pG = {0} for some
prime p (that is every non-identity element of G has order p). Then G is a
vector space over F = ZZ p . Moreover if G is finite, then G is a direct sum of
cyclic groups of order p.

Proof. We know that pG = {pg | g ∈ G} = {0}. Define the scalar product

(.) : ZZ p × G −→ G by k.g = kg. Then G is a vector space over ZZ p .
If G is finite, then |G| = pn where n is the dimension of G over ZZ p .
Hence there exists a base B = {x1 , x2 , · · · , xn } for G over ZZ p . Thus G =<
x1 , x2 , · · · , xn > and hence

G =< x1 > ⊕ < x2 > ⊕ · · · ⊕ < xn >∼

= ZZ p ⊕ ZZ p ⊕ · · · ⊕ ZZ p .

32
Note 3.2.3 (Elementary abelian p-groups) If G is an abelian group such
that pG = {0}. Then we say that G is an elementary abelian p-group. Then
∀g ∈ G, o(g) = p if and only if g 6= 0. If G is finite, then |G| = pn for some
n∈N | and G ∼
= ZZ p × ZZ p × · · · ZZ p ( n factors).
For example V4 ∼
= ZZ 2 × ZZ 2 is the elementary abelian 2-group of order 4.

Theorem 3.2.3 (Basis Theorem) Every non-trivial finite abelian group G

is a direct sum of cyclic p-groups.

Proof. By Theorem 3.2.1, every finite abelian group is a direct sum of p-

groups. Hence we need only to prove the theorem for abelian p-groups. Assume
that G is a finite abelian group such that pn G = {0}. For the proof we use
induction on n. If n = 1, then G is an elementary abelian p-group and proof
will follow from Lemma 3.2.2. Now assume that G is a finite abelian p-group
such that pn+1 G = {0}. Let H = pG. Then pG ≤ G and pn H = pn+1 G = {0}.
Thus by the induction hypothesis H is a direct sum of cyclic p-groups. Let
H =< y1 > ⊕ < y2 > ⊕ · · · ⊕ < yt > where < yi > are cyclic p-groups. Since
yi ∈ H = pG, ∃zi ∈ G such that yi = pzi , 1 ≤ i ≤ t. let N be the group
generated by z1 , z2 , · · · , zt . We will show that

N =< z1 > ⊕ < z2 > ⊕ · · · ⊕ < zt > .

Pt
It is obvious that N = i=1 < zi >. Now let g be an element of < zi >
Pt Pt
∩ j6=i < zj >. Then g = ki zi = j6=i kj zj , kj ∈ ZZ . (*)
If p | ki for all 1 ≤ i ≤ t, then we have
t
X
(ki /p)yi = ki zi = (kj /p)yj
j6=i

which implies that

t
X
(ki /p)yi ∈< yi > ∩ < yj >= {0}.
j6=i

Thus (ki /p)yi = 0. So ki zi = 0 and hence g = 0.

Now assume that ∃ kl such that p does not divide kl . Since by (*) pg =
Pt
pki zi = ki yi = j6=i kj yj , 1 ≤ i ≤ t, we must have ki yi = 0, 1 ≤ i ≤ t. This
shows that kl yl = 0 and hence o(yl ) | kl . (**). Now p | o(yl ), and o(yl ) | kl
imply that p | kl , a contradiction.
Therefore g = 0 and N = ⊕ti < zi >.

33
 Next we will show that there exists a subgroup M of G such that G = N ⊕M .
For this purpose define G[p] by

G[p] = {g | g ∈ G, pg = 0}.

Then G[p] is an elementary abelian p-subgroup of G. Assume that o(yi ) = mi .

Then mi yi = p(mi zi ) = 0 implies that mi zi ∈ G[p] for all 1 ≤ i ≤ t. Hence
∃ x1 x2 , · · · , xl ∈ G[p] such that

G[p] =< m1 z1 , m2 z2 , · · · , mt zt , x1 , x2 , · · · , xl > .

Let M =< x1 , x2 , · · · , xl >. We claim that G = N ⊕ M .

Pt Pl
Firstly, if g ∈ M ∩ N, then g = i ki zi = i=1 bi xi for some ki , bi ∈ ZZ . So
t
X l
X l
X
pg = ki yi = bi pxi = bi (0) = 0,
i=1 i=1 i=1

since xi ∈ G[p]. Hence ki yi = 0 ∀i , 1 ≤ i ≤ t. So that ki = ki0 mi for some ki0 .

Pt 0 Pl Pt 0 Pl
Now g = i=1 ki mi zi = i=1 bi xi implies i=1 ki mi zi − i=1 bi xi = 0. Now
independency of {m1 z1 , m2 z2 , · · · , mt zt , x1 , x2 , · · · , xl } implies that ki0 .mi zi =
0 , 1 ≤ i ≤ t and bi xi = 0 , 1 ≤ i ≤ l. Thus g = 0 and hence M ∩ N = {0}.
Pt Pt
Secondly, if g ∈ G, then pg ∈ H. So that pg = i=1 ai yi = p( i=1 ai zi ).
Pt Pt
This shows that p(g − i=1 ai zi ) = 0 and g − i=1 ai zi ∈ G[p]. Therefore ∃ ki , bi
such that
t
X t
X l
X
g− ai zi = ki .mi zi + bi xi ,
i=1 i=1 i=1
that is
t
X l
X
g= (ai + ki mi )zi + bi xi ∈ N + M.
i=1 i=1
Thus G = N ⊕ M .

Corollary 3.2.4 Let G be a finite abelian group. Then

G = ZZ m1 ⊕ ZZ m2 ⊕ · · · ⊕ ZZ mk

where mi | mi+1 , 1 ≤ i ≤ k − 1.

Proof. By Theorem 3.2.1 we have G = Gp1 ⊕Gp2 ⊕· · ·⊕Gpt , where p1 , p2 , · · · , pt

are distinct primes. Now by the basis theorem (Theorem 3.2.3) each Gpi
is a direct sum of cyclic pi -groups. Let ZZ pi ei be the cyclic pi -group in the
decomposition of Gpi having largest order. Then

G = ZZ p1 e1 ⊕ ZZ p2 e2 ⊕ · · · ⊕ ZZ pt et ⊕ H,

34
where H is the direct sum of remaining summands. Since

ZZ p1 e1 ⊕ ZZ p2 e2 ⊕ · · · ⊕ ZZ pt et = ZZ m

where m = p1 e1 p2 e2 · · · pt et , we have G = ZZ m ⊕H. Repeating the above process

for H, we obtain H = ZZ n ⊕H1 . Obviously n divides m and G = ZZ m ⊕ ZZ n ⊕H1 .
Since G is finite, this process will end after a finite number of steps.

Note 3.2.4 The above decomposition is called Canonical Decomposition .

In this decomposition, the abelian group

G = ZZ m1 ⊕ ZZ m2 ⊕ · · · ⊕ ZZ mk

is said to be of type (m1 , m2 , · · · , mk ).

3.3 The Fundamental Theorem of Finite Abelian

Groups
Definition 3.3.1 If G is a finite elementary abelian p-group then G ∼
= ZZ p ⊕
ZZ p ⊕ · · · ⊕ ZZ p . Let d(G) denote the number of cyclic summands in the above
expression. Then d(G) is the dimension of G over ZZ p . Hence if |G| = pn , then
d(G) = n.

Lemma 3.3.1 If G is a finite abelian p-group, then pn G/pn+1 G is an elemen-

tary abelian p-group.

Proof. Let x ∈ pn G/pn+1 G. Then x = pn g + pn+1 G, for some g ∈ G and

xp = px = pn+1 g + pn+1 G = pn+1 G = 0.

This shows that if x 6= 0, then o(x) = p. Since pn G/pn+1 G is an abelian

p-group, we have shown that pn G/pn+1 G is an elementary abelian p-group.

Lemma 3.3.2 Assume that G is a finite abelian group of order pbk such that
it is a direct sum of b copies of cyclic groups of order pk . Then for all n < k
we have d(pn G/pn+1 G) = b.

Proof. Let G = G1 ⊕ G2 ⊕ · · · ⊕ Gb , where Gi ∼

= ZZ pk . Then

p n G = pn G 1 ⊕ pn G 2 ⊕ · · · ⊕ p n G b ,

and
pn+1 G = pn+1 G1 ⊕ pn+1 G2 ⊕ · · · ⊕ pn+1 Gb .

35
So

pn G/pn+1 G = pn G1 /pn+1 G1 ⊕ pn G2 /pn+1 G2 ⊕ · · · ⊕ pn Gb /pn+1 Gb .

Since Gi , 1 ≤ i ≤ b, are cyclic, pn Gi /pn+1 Gi are also cyclic (check). Now

since pn Gi /pn+1 Gi are cyclic elementary abelian p-groups, they must be cyclic
groups of order p. Thus

pn G/pn+1 G = ZZ p ⊕ ZZ p ⊕ · · · ⊕ ZZ p , b copies

and hence d(pn G/pn+1 G) = b.

Theorem 3.3.3 Let G be a finite abelian p-groups. Assume that G = C1 ⊕C2 ⊕

· · · ⊕ Cl is a decomposition for G into cyclic p-groups. Then d(pn G/pn+1 G) is
the number of cyclic summands in G having order greater than or equal to pn+1 .

Proof. Let Bk be the direct sum of all cyclic groups of order pk in the decom-
position. Suppose that bk is the number of cyclic groups in Bk . Then

G = B1 ⊕ B2 ⊕ · · · ⊕ Bt ,

pn G = pn Bn+1 ⊕ pn Bn+2 ⊕ · · · ⊕ pn Bt

and
pn+1 G = pn+1 Bn+2 ⊕ pn+1 Bn+3 ⊕ · · · ⊕ pn+1 Bt .

Hence

pn G/pn+1 G = pn Bn+1 ⊕ pn Bn+2 /pn+1 Bn+2 ⊕ · · · ⊕ pn Bt /pn+1 Bt ,

and therefore

d(pn G/pn+1 G) = d(pn Bn+1 ) ⊕ d(pn Bn+2 /pn+1 Bn+2 ) ⊕ · · · ⊕ d(pn Bt /pn+1 Bt ).

Since Bk is the direct sum of bk copies of cyclic groups of order pk , by Lemma

3.3.2, we have

d(pn Bn+1 ) = d(pn Bn+1 /pn+1 Bn+1 ) = bn+1 ,

d(pn Bn+2 /pn+1 Bn+2 ) = bn+2 ,
..
.
d(pn Bt /pn+1 Bt = bt .

Hence we have d(pn G/pn+1 G) = bn+1 + bn+2 + · · · + bt .

36
Definition 3.3.2 If G is a finite abelian p-group, we define U (n, G) by

U (n, G) = d(pn G/pn+1 G) − d(pn+1 G/pn+2 G).

Note 3.3.1 Since pn G/pn+1 G and pn+1 G/pn+2 G are elementary abelian p-
groups, U (n, G) depends only on G and not on any choice of decomposition of
G into cyclic p-groups

Theorem 3.3.4 Assume that G is a finite abelian p-group. Then any two
decompositions of G into cyclic p-groups have the same number of cyclic groups
of same order.

Proof. Using Theorem 3.3.3 we have

d(pn G/pn+1 G) = bn+1 + bn+2 + · · · + bt

and
d(pn+1 G/pn+2 G) = bn+2 + bn+3 + · · · + bt .
Hence U (n, G) = bn+1 . This shows that U (n, G) is equal to the number of cyclic
groups of order pn+1 in any decomposition of G. Since U (n, G) is independent
of the choice of decomposition, the proof follows.

Corollary 3.3.5 Let G1 and G2 be two finite abelian p-groups. Then G1 ∼

= G2
if and only if U (n, G1 ) = U (n, G2 ) for all n ≥ 0.

Proof. Assume that G1 ∼ = G2 . Then using this isomorphism we can show that
pn G1 /pn+1 G1 ∼
= pn G2 /pn+1 G2 , ∀n ≥ 0, and hence U (n, G1 ) = U (n, G2 ) for all
n ≥ 0. Conversely, if U (n, G1 ) = U (n, G2 ) for all n ≥ 0, then the proof follows
from Theorem 3.3.3

Example 3.3.1 (i) We list the abelian groups of order 8 = 23 .

U (0, G) U (1, G) U (2, G) G
3 0 0 ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2
1 1 0 ZZ 2 ⊕ ZZ 4
0 0 1 ZZ 8
(ii) We list the abelian groups of order 16 = 24 .

U (0, G) U (1, G) U (2, G) U (3, G) G

4 0 0 0 ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2
2 1 0 0 ZZ 2 ⊕ ZZ 2 ⊕ ZZ 4
1 0 1 0 ZZ 2 ⊕ ZZ 8
0 2 0 0 ZZ 4 ⊕ ZZ 4
0 0 0 1 ZZ 16

37
Note 3.3.2 It is easy to see that the number of non-isomorphic abelian groups
of order pm equals the number of partitions of m. This number is independent
of the prime p. For example the number of non-isomorphic abelian groups of
order p5 is 7 which is obtained from the list of partitions of 5:

Partitions of 5 Abelian groups of order p5

1, 1, 1, 1, 1 ZZ p ⊕ ZZ p ⊕ ZZ p ⊕ ZZ p ⊕ ZZ p
2, 1, 1, 1 ZZ p2 ⊕ ZZ p ⊕ ZZ p ⊕ ZZ p
2, 2, 1 ZZ p2 ⊕ ZZ p2 ⊕ ZZ p
3, 1, 1 ZZ p3 ⊕ ZZ p ⊕ ZZ p
3, 2 ZZ p3 ⊕ ZZ p2
4, 1 ZZ p4 ⊕ ZZ p
5 ZZ p5

Lemma 3.3.6 Suppose that G and G0 are two finite abelian groups. Let f ∈
Hom(G, G0 ). Then for all primes p we have f (Gp ) ⊆ G0p .

Proof. Since G and G0 are abelian, Gp ≤ G and G0p ≤ G0 . Let x be an element

of f (Gp ). Then x = f (y) for some y ∈ Gp . Since y ∈ Gp , ∃k such that pk y = 0.
So
00 = f (0) = f (pk y) = pk f (y) = pk x.

This shows that x ∈ G0p .

Theorem 3.3.7 Assume that G and G0 are two finite abelian groups. Then
G∼
= G0 if and only if Gp ∼
= G0p for all prime p.

Proof. If G ∼
= G0 , then ∃ f ∈ Hom(G, G0 ) such that f is an isomorphism. Then
by Lemma 3.3.6, f (Gp ) ⊆ G0p . Thus we have a homomorphism f : Gp −→ G0p .
Let x ∈ G0p . Then ∃ k such that pk x = 00 . Since f : G −→ G0 is onto, we have
an element y ∈ G such that f (y) = x. Now

f (pk y) = pk f (y) = pk x = 0.

implies pk y ∈ Ker(f ). Since f : G −→ G0 is one-to-one, pk y = 0 and hence

y ∈ Gp . Thus x ∈ f (Gp ). This shows that f : Gp −→ G0p is an isomorphism
and f (Gp ) ∼
= G0 . p
Conversely, suppose that Gp ∼
= G0p for all p. Then since G =
L
Gp and
p | |G|
G0 = G0p , we have G ∼
= G0 .
L
p | |G0 |

38
Theorem 3.3.8 (Fundamental Theorem of Finite Abelian Groups) Let
G be a finite abelian group. Any two decompositions of G into cyclic p-groups
have the same number of summands of each order.
L
Proof. Since G = Gp , the proof follows by Theorems 3.3.4 and 3.3.7
p | |G|

Note 3.3.3 (Application of the Fundamental Theorem of Finite Abelian Groups)

We are now able to determine the number of non-isomorphic classes of abelian
groups of order n. Assume that G is an abelian group of order n. Let
n= p1 α 1 p2 α 2 · · · pk αk . Then
(i) G = A1 ⊕ A2 ⊕ · · · ⊕ Ak , where Ak = Gpk and |Ak | = pk αk .
(ii) For each A ∈ {A1 , A2 , · · · , Ak } with |A| = pα we have A ∼
= ZZ pβ1 ⊕
ZZ pβ2 ⊕ · · · ⊕ ZZ pβt where β1 ≥ β2 ≥ · · · ≥ βt and β1 + β2 + · · · + βt = α. Note
that (β1 , β2 · · · Bt ) is a partition of α.
(iii) The decomposition given for G in (i) and (ii) is unique.
Therefore the number of non-isomorphic classes of abelian groups of order
n is equal to π(α1 ) × π(α2 ) × · · · × π(αk ), where π(αi ) denotes the number of
partitions of αi .

Example 3.3.2 If n = 1440 = 25 × 32 × 5, then we have π(5) = 7, π(2) = 2

and π(1) = 1. Hence there are 7 × 2 × 1 = 14 non-isomorphic abelian groups of
order 1440. First consider the following list:

pα Partitions of α Abelian Groups of order pα

5 ZZ 25
4, 1 ZZ 24 ⊕ ZZ 2
3, 2 ZZ 23 ⊕ ZZ 22
25 3, 1, 1 ZZ 23 ⊕ ZZ 2 ⊕ ZZ 2
2, 2, 1 ZZ 22 ⊕ ZZ 22 ⊕ ZZ 2
2, 1, 1, 1 ZZ 22 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2
1, 1, 1, 1, 1 ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2
32 2 ZZ 32
1, 1 ZZ 3 ⊕ ZZ 3
5 1 ZZ 5

Now we obtain the abelian groups of order 1440 by taking one abelian group
from each of the three lists (right hand column above) and taking their direct
sum:

39
 ZZ 25 ⊕ ZZ 32 ⊕ ZZ 5 ∼
= Z1440 ,
ZZ 25 ⊕ ZZ 3 ⊕ ZZ 3 ⊕ ZZ 5 ∼
= ZZ 480 ⊕ ZZ 3 ,
ZZ 24 ⊕ ZZ 2 ⊕ ZZ 32 ⊕ ZZ 5 ∼
= ZZ 720 ⊕ ZZ 2 ,
∼ ZZ 240 ⊕ ZZ 6 ,
ZZ 24 ⊕ ZZ 2 ⊕ ZZ 3 ⊕ ZZ 3 ⊕ ZZ 5 =
ZZ 23 ⊕ ZZ 22 ⊕ ZZ 32 ⊕ Z5 ∼= ZZ 360 ⊕ ZZ 4 ,
ZZ 23 ⊕ ZZ 22 ⊕ ZZ 3 ⊕ ZZ 3 ⊕ Z5 ∼
= ZZ 120 ⊕ ZZ 12 ,
∼
ZZ 23 ⊕ ZZ 2 ⊕ Z2 ⊕ ZZ 32 ⊕ Z5 = ZZ 360 ⊕ ZZ 2 ⊕ Z2 ,
∼ ZZ 120 ⊕ ZZ 6 ⊕ Z2 ,
ZZ 23 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 3 ⊕ ZZ 3 ⊕ Z5 =
ZZ 22 ⊕ ZZ 22 ⊕ ZZ 2 ⊕ ZZ 32 ⊕ Z5 ∼= ZZ 180 ⊕ ZZ 4 ⊕ ZZ 2 ,
ZZ 22 ⊕ ZZ 22 ⊕ Z2 ⊕ ZZ 3 ⊕ ZZ 3 ⊕ ZZ 5 ∼= ZZ 60 ⊕ ZZ 12 ⊕ Z2 ,
ZZ 22 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 32 ⊕ ZZ 5 ∼
= ZZ 180 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ,
ZZ 22 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 3 ⊕ ZZ 2 ⊕ ZZ 5 ∼
= ZZ 60 ⊕ ZZ 6 ⊕ ZZ 2 ⊕ ZZ 2 ,
ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 32 ⊕ ZZ 5 ∼
= ZZ 90 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ,
ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 3 ⊕ ZZ 3 ⊕ ZZ 5 ∼
= ZZ 30 ⊕ ZZ 6 ⊕ ZZ 2 ⊕ ZZ 2 ⊕ ZZ 2 .
The above is a complete list of all abelian groups of order 1440. Every
abelian of order 1440 is isomorphic to precisely one of the groups listed above.
No two of the groups in this list are isomorphic.

Exercise 3.3.1 List the abelian groups of order (i) 720 ; (ii) 1800 .

Exercise 3.3.2 If G = ZZ 60 × ZZ 45 × ZZ 12 × ZZ 36 , find

(i) the number of elements of order 2 .
(ii) the number of subgroups of index 2 in G.

Exercise 3.3.3 If H and K are elementary p-primary abelian groups, then

L
show that d(H K) = d(H) + d(K).

Exercise 3.3.4 If G and H are finite p-primary abelian groups, then show
L
that U (n, G H) = U (n, G) + U (n, H).

G∼
L L
Exercise 3.3.5 (i) Let G and H be finite abelian groups. If G =H H,
show that G ∼
= H.
B∼
L L
(ii) Suppose that A, B, and C are finite abelian groups. If A = A C,
then B ∼= C.

Definition 3.3.3 (Exponent) Let G be a group and n ∈ N

| . We say that n

is the exponent of G if n is the smallest positive integer such that Gn = {1G }.

Example 3.3.3 If G is an elementary abelian p-group of order pk , then pG =

{0}. Hence p is the exponent of G.

40
Example 3.3.4 Let G = ZZ m1 ⊕ ZZ m2 ⊕· · ·⊕ ZZ mk be a canonical decomposition
for the group G. Then |G| = m1 m2 · · · mk . For any x = (g1 , g2 , · · · , gk ) ∈ G
we have mk x = (mk g1 , mk g2 , · · · , mk gk ) and since mi | mk ∀ i, 1 ≤ i ≤ k,
we have mk x = (0, 0, · · · , 0) = 0. If ∃ m > 0 such that mG = {0}, then
∀x = (g1 , g2 , · · · , gk ) we have mx = (mg1 , mg2 , · · · , mgk ) = (0, 0, · · · , 0) = 0.
Thus we must have mgi = 0 for all gi ∈ ZZ mi , for all 1 ≤ i ≤ k. Hence mi | m
for all 1 ≤ i ≤ k . So that mk | m and therefore mk is the exponent of G.

Theorem 3.3.9 (Uniqueness of canonical decomposition) Let G be a fi-

nite abelian group. Assume that

G = ZZ m1 ⊕ ZZ m2 ⊕ · · · ⊕ ZZ mk

and
G = ZZ n1 ⊕ ZZ n2 ⊕ · · · ⊕ ZZ nl
are two canonical decompositions for G. Then mi = ni ∀i and k = l.

Proof. Since mk and nl are the exponent of G, we must have mk = nl . Now

let G1 = ZZ m1 ⊕ ZZ m2 ⊕ · · · ⊕ ZZ mk−1 and G2 = ZZ n1 ⊕ ZZ n2 ⊕ · · · ⊕ ZZ nl−1 .
Then G = G1 ⊕ ZZ mk = G2 ⊕ ZZ nl . Since mk = nl , we have ZZ mk = ZZ nl .
∼ G2 . Using induction we can say that
Now Exercise 3.3.5 implies that G1 =
mi = ni , 1 ≤ i ≤ k − 1 and k − 1 = l − 1. Hence mi = ni , 1 ≤ i ≤ k and l = k.

In the rest of this section we give some applications. For example in Theorem
3.3.10 we will show that the converse of Lagrange’s Theorem is true for finite
abelian groups.

Theorem 3.3.10 Let G be a finite abelian group. If m divides |G|, then G has
a subgroup of order m

Proof. Case 1: Assume that G is a finite abelian p-group, |G| = pα and

m = pβ , where β ≤ α. Since p | |G|,, there exists x ∈ G such that o(x) = p..
Let N =< x > . Then N  G and G/N is a finite abelian p-group of order
pα−1 . Using induction on |G|, since |G/N | < |G|, there exists H̄ ≤ G/N with
|H̄| = pβ−1 . So there exists H ≤ G such that H̄ = H/N. Now |H| = |H̄| × |N |
gives |H| = pβ−1 × p = pβ .
Lk
Case 2: Let G = i=1 Gpi , where all primes pi are distinct and each GPi is
a finite abelian pi -group of order pαi i and |G| = n = pα1 1 × pα2 2 × · · · × pαk k . Since
m | |G| , m = pβ1 1 pβ2 2 · · · pβk k , 0 ≤ βi ≤ αi . Since GPi is a finite abelian pi -group,
by Case 1, there exists Hi ≤ GPi with |Hi | = pβi i . Now let H =
Lk
i=1 Hi . Then
H ≤ G and |H| = pβ1 1 pβ2 2 · · · pβk k = m.

41
Theorem 3.3.11 Let G be a finite abelian group such that |G| is not divisible
by the square of any prime. Then G is cyclic.

Proof. Let |G| = n. Then n = p1 × p2 × · · · × pk , where pi are distinct primes.

Then

G = ZZ p1 ⊕ ZZ p2 ⊕ · · · ⊕ ZZ pk
∼
= ZZ p1 ×p2 ×···×pk , by Proposition 3.1.2
∼
= Zn .

Exercise 3.3.6 Find the order of (8, 4, 10) in the group ZZ 12 × ZZ 60 × ZZ 24 .

[Hint : if (g1 , g2 , · · · , gm ) ∈ G1 × G2 × · · · × Gm , then o(g1 , g2 , · · · , gm ) =
L.C.M (o(gi ) , 1 ≤ i ≤ m.)

Exercise 3.3.7 Find all subgroups of ZZ 2 × ZZ 4 .

Exercise 3.3.8 The set of all elements of finite order in an abelian group G
form a subgroup . We call this subgroup the torsion subgroup of G.
(i) Find the order of the torsion subgroups of ZZ 4 × ZZ × ZZ 2 and ZZ 12 × ZZ × ZZ 12 .
(ii) Find the torsion subgroup of the group (C| ∗ , .).

42
Chapter 4

Normal Series

4.1 Jordan-Hölder Theorem

Definition 4.1.1 (Normal Series) A normal series of a group G is a finite
sequence G0 , G1 , · · · , Gn of subgroups of G such that

G = G0  G1  G2  · · ·  Gn = {1G }. (∗)

The factor groups Gi /Gi+1 are called factors of (*). The number of factors of
order greater than 1 is called the length of (*).

Definition 4.1.2 Two normal series of G are equivalent if there is a one-to-

one correspondence between the factors of each series such that corresponding
factors are isomorphic. Obviously the equivalent normal series of G have the
same length.

Example 4.1.1 Let G = ZZ 30 =< x >, where o(x) = 30. Consider the follow-
ing two normal series:

G < x5 >  < x10 > {1G } (∗)

and
G < x2 >  < x6 > {1G }. (∗∗)

That is
ZZ 30  ZZ 6 >  ZZ 3  {1G } (∗)

and
ZZ 30  ZZ 15 >  ZZ 5  {1G }. (∗∗)

The factor groups of (*) are ZZ 5 , ZZ 2 and ZZ 3 while those of (**) are ZZ 2 , ZZ 3
and ZZ 5 . Hence (*) and (**) are equivalent.

43
Definition 4.1.3 (Composition Series) A composition series is a normal
series
G = G0  G1  G2  · · ·  Gn = {1G } (∗)
such that Gi+1 is a maximal normal subgroup of Gi .

Exercise 4.1.1 If H  G, prove that H is a maximal normal subgroup if and

only if G/H is simple or H = G.

Proposition 4.1.1 A normal series is a composition series if and only if its

factors are simple groups or {1}.

Proof. Proof follows from Exercise 4.1.1.

Definition 4.1.4 (Refinement) Assume that the following are two normal
series for a group G

G = G0  G1  G2  · · ·  Gn = {1G }, (∗)

G = H0  H1  H2 · · ·  Hm = {1G }. (∗∗)
Then we say that (**) is a refinement of (*), if every Gi is an Hj , but not
necessarily every Hj is a Gi . In this case obviously length(∗∗) ≥ length(∗).

Proposition 4.1.2 (i) A composition series is a normal series of maximal

length.
(ii) Every finite group has a composition series.

Proof. (i) Suppose that (*) is a composition series for G

G = G0  G1  G2  · · ·  Gn = {1G }. (∗)

Since Gi+1 is a maximal normal subgroup of Gi , we are not able to insert any
other subgroup of G between them. Hence (*) has no non-trivial refinement
and therefore it has the maximal length.
(ii) If G is simple, then G1G is the only composition series for G. So assume
6 H
G.
that G is not simple. Hence there exists a subgroup H such that {1G } =
Since G is a finite group, W.L.O.G we can assume that H is a maximal normal
subgroup of G. Now since |H| < |G|, by induction hypothesis there exists a
composition series for H, namely

H = H0  H1  H2  · · ·  Hn = {1G }. (∗).

Now consider the following normal series for G

G  H  H1  H2  · · ·  Hn = {1G }. (∗∗).

It is easy to see that (**) is a composition series for G.

44
Theorem 4.1.3 Let H  G. Suppose that (*) is a composition series for G

G = G0  G1  G2  · · ·  H  H1  H2  · · ·  Hn = {1G }. (∗)

Then
G/H  G1 /H  G2 /H  · · ·  H/H (∗∗)

is a composition series for G/H.

Proof. Proof follows from the fact (Gi /H)/(Gi+1 /H) ∼

= Gi /Gi+1 and the
assumption that (*) is a composition series for G.

Lemma 4.1.4 (Zassenhaus Lemma, 1935) Let G be a group and A  A∗ ≤

G and B  B ∗ ≤ G. Then

A(A∗ ∩ B)  A(A∗ ∩ B ∗ ), B(B ∗ ∩ A)  B(A∗ ∩ B ∗ )

and
A(A∗ ∩ B ∗ )/A(A∗ ∩ B) ∼
= B(A∗ ∩ B ∗ )/B(B ∗ ∩ A).

Proof. See Lemma 5.8, Rotman page 77.

Exercise 4.1.2 (i) Show that an abelian group has a composition series if and
only if it is finite.
(ii) Give an example of an infinite group which has a composition series.

Exercise 4.1.3 If G is a finite group having a normal series with factor groups
H0 , H1 , · · · , Hn , show that |G| = |Hi |.
Q

Exercise 4.1.4 Find all the composition series for each of the following groups:

S3 , A3 , S 4 .

Theorem 4.1.5 (Schreier, 1926) Let G be a group. Then any two normal
series of G have refinements which that equivalent.

Proof. Suppose that (*) and (**) are two normal series for G:

G = G0  G1  G2  · · ·  Gn = {1G }, (∗)

G = H0  H1  H2  · · ·  Hm = {1G }. (∗∗).

Let Gij = Gi+1 (Gi ∩Hj ), 0 ≤ i ≤ n, 0 ≤ j ≤ m. Then Gij ⊇ Gij+1 and applying
the Zassenhaus Lemma (for A = Gi+1 , A∗ = Gi , B = Hj+1 , B ∗ = Hj )

45
we get Gij+1  Gij . Note that Gi0 = Gi+1 (Gi ∩ H0 ) = Gi+1 Gi = Gi and
Gim = Gi+1 (Gi ∩ Hm ) = Gi+1 . Then Gij form a normal series (∗0 ) for G:

G = G00  G01  G02  · · ·  G0m = G1  G11  · · ·  G1m = G2  · · ·

Gi−1m = Gi  · · ·  Gn−1m = Gn = {1G }. (∗0 )

It is easy to see that (∗0 ) is a refinement of (∗) of length nm.

Similarly, if we define Hij = Hj+1 (Hj ∩ Gi ), 0 ≤ i ≤ n, 0 ≤ j ≤ m, then
Hij ⊇ Hi+1j and Hi+1j  Hij . We have a normal series (∗∗0 ) for G of length
mn which is a refinement of (∗∗). Now by Zassenhaus lemma

Hij /Hi+1j = Hj+1 (Hj ∩ Gi )/Hj+1 (Hj ∩ Gi+1 ) ∼

=

Gi+1 ((Hj ∩ Gi )/Gi+1 (Hj+1 ∩ Gi ) = Gij /Gij+1 .

Thus (∗0 ) and (∗∗0 ) are equivalent.

Theorem 4.1.6 (Jordan-Hölder Theorem, 1868 and 1889) Any two com-
position series of a group G are equivalent.

Proof. Suppose that (*) and (**) are two composition series for G. Then since
(*) and (**) are normal series of G, by Theorem 4.1.5 they have equivalent
refinements. But since a composition series is a normal series of maximal length
(Proposition 4.1.2), (*) and (**) admit no non-trivial refinements. Hence (*)
and (**) must be equivalent.

4.2 Soluble (Solvable) Groups

Definition 4.2.1 (Soluble Groups) A Group G is said to be soluble (or
Solvable, by American mathematicians) if it has a normal series whose factors
are all abelian. Such a normal series is called soluble series. Groups which
that not soluble are said to be insoluble.

The notion of solubility of groups was introduced by Galois in connection

with solving polynomial equations by radicals.

Remark 4.2.1 It is clear that non-abelian simple groups are insoluble.

Exercise 4.2.1 Show that any abelian group is soluble.

46
Exercise 4.2.2 (i) Show that Sn is soluble for n ≤ 4.
(ii) Show that the dihedral group D2n is soluble for all n.

Exercise 4.2.3 Show that any refinement of a soluble series is a soluble series.

Exercise 4.2.4 If G is a soluble group with a composition series, prove that

G is finite.

Exercise 4.2.5 Let H  G. If G has a composition series, prove that G has a

composition series one of whose term is H.

Theorem 4.2.1 If G is a soluble group and H ≤ G, then H is also soluble.

Proof. Let (*) be a soluble series for G:

G = G0  G1  G2  · · ·  Gn = {1G }. (∗)

We will show that (**) below is a soluble series for H:

H = H ∩ G ≥ H ∩ G1 ≥ · · · ≥ H ∩ Gn = {1G }. (∗∗)

Since H ∩ Gi+1 = (H ∩ Gi ) ∩ Gi+1 and Gi+1  Gi , we have H ∩ Gi+1  H ∩ Gi .

Thus (**) is a normal series. Now by the 3rd isomorphism theorem we have

(H ∩ Gi )/(H ∩ Gi+1 ) ∼
= (H ∩ Gi )Gi+1 /Gi+1 ≤ Gi /Gi+1 .

Since Gi /Gi+1 is abelian, (H ∩ Gi )/(H ∩ Gi+1 ) is also abelian and hence (**)
is a soluble series for H.

Theorem 4.2.2 If G is a soluble group and H  G, then G/H is also soluble.

Proof. Let (*) be a soluble series for G:

G = G0  G1  G2  · · ·  Gn = {1G }. (∗)

Since H  G, HGi is a subgroup G and since Gi+1  Gi , we have HGi+1  HGi .

Therefore we have the following normal series for G:

G = HG0  HG1  HG2  · · ·  HGn = H  {1G }. (∗∗)

Now Consider the following series:

G/H = HG0 /H ≥ (HG1 )/H ≥ (HG2 )/H ≥ · · · ≥ H/H = {1G/H }. (∗ ∗ ∗)

47
Since H and HGi+1 are normal subgroups of HGi , we have HGi+1 /H HGi /H
and hence (***) is a normal series for G/H.We also have

(HGi /H)/(HGi+1 /H) ∼

= (HGi )/(HGi+1 ) (by the 2nd isomorphism theorem)

= ((HGi+1 )Gi )/(HGi+1 ) ∼

= Gi /(Gi ∩(HGi+1 ) (by the 3rd isomorphism theorem)
∼
= (Gi /Gi+1 )/((HGi+1 ∩ Gi )/Gi+1 ) (by the 2nd isomorphism theorem).
So (HGi /H)/(HGi+1 /H) is a factor group of the abelian group Gi /Gi+1 and
hence it is abelian. Therefore (***) is a soluble series for G/H.

Theorem 4.2.3 Let G be a group and H  G. If H and G/H are soluble, then
G is soluble.

Proof. Since G/H is soluble, it has a soluble series (*):

G/H  K̄1  K̄2  · · ·  K̄n = {1G/H }. (∗)

Now since K̄i ≤ G/H, there exists Ki ≤ G such that K̄i = Ki /H, 1 ≤ i ≤ n.
Since K̄i+1  K̄i , we have Ki+1 /H  Ki /H and Ki+1  Ki . Now

K̄i /K̄i+1 = (Ki /H)/(Ki+1 )/H ∼

= Ki /Ki+1

imply that Ki /Ki+1 is abelian. Consider the series (**):

G  K1  K2  · · ·  Kn = H. (∗∗)

Since H is soluble, there is a soluble series (***) for H:

H  H1  H2  · · ·  Hm = {1G }. (∗ ∗ ∗)

Putting together (**) and (***), we obtain a soluble series for G.

Theorem 4.2.4 A non-trivial finite group G is soluble if and only if it has a

composition series whose factors are cyclic groups of prime order.

Proof. Assume that G is soluble and |G| = n > 1. If n = p for some prime p,
then G is a cyclic group of order p and hence G  {1G } is the only composition
series for G.
Now assume that n is not a prime. If G is abelian, then G is not simple.
If G is non-abelian, then since G is soluble, G is not simple. Thus for both
abelian and non-abelian cases G contains a non-trivial normal subgroup H.
Now since H and G/H are soluble groups of order less than n, by induction
hypothesis they have composition series with factors of prime order. As in the
proof of Theorem 4.2.3 we obtain a composition series for G whose factors are
cyclic groups of prime order. [Note that K̄i /K̄i+1 ∼
= Ki /Ki+1 implies Ki /Ki+1
is cyclic of prime order.]

48
Theorem 4.2.5 Sn is not soluble for n ≥ 5.
Proof. Consider the normal series Sn  An  1Sn (∗). Since the factors of (∗)
are ZZ 2 and An which are simple, (∗) is a composition series for Sn . Since by
Jordan-Hölder Theorem any other composition series for Sn is equivalent to
(∗), ZZ 2 and An are the only composition factors. Now the result follows from
Theorem 4.2.4.

Theorem 4.2.6 Let G = H × K such that H and K are soluble. Then G is

also soluble.

Proof. Since G/H ∼

= K, G/H is soluble. Now, by Theorem 4.2.3, solubility of
H and G/H implies that G is soluble.
Corollary 4.2.7 If G is a finite p-group, then G is soluble.
Proof. If |G| = 1, then clearly G is soluble. So assume |G| > 1. Since G is a
p-group, Z(G) 6= {1G } and hence |G/Z(G)| < |G|. Now since G/Z(G) is a finite
p-group, by induction hypothesis G/Z(G) is soluble. Since Z(G) is an abelian
group, it is soluble. Now solubility of G/Z(G) and Z(G) implies G is soluble.

Exercise 4.2.6 If N and M are soluble subgroups of G with N  G, show that

M N is a soluble subgroup of G.

Exercise 4.2.7 (i) If |G| = p2 q, where p and q are primes, show that G is
soluble.
(ii)If |G| = pq n , where p and q are primes with p < q, show that G is soluble.
[Hint: Use Sylow’ Theorem and Exercise 4.2.6.]

Exercise 4.2.8 A group G having a normal series

G = G0  G1  G2  · · ·  Gn = {1G } (∗)

with each factor group Gi /Gi+1 cyclic and each Gi+1 is normal in G is called
Super-soluble group. Show that a soluble group need not be super-soluble.
[Hint: Use S4 .]

Exercise 4.2.9 Show that the following two statements are equivalent
(i) every finite group of odd order is soluble;
(ii) every finite non-abelian simple group has even order.

Note 4.2.1 J G Thompson and W Feit in 1963 proved that every finite non-
abelian simple group has even order. [See Pacific Journal of Mathematics, 13
(1963),775–1029.]

49
Chapter 5

Representation Theory of
Finite Groups

5.1 Basic Concepts

Definition 5.1.1 Let G be a group. Let f : G −→ GL(n, |F) be a homomor-
phism. Then we say that f is a Matrix Representation of G of degree n (or
dimension n), over the field |F.

If Ker(f ) = {1G }, then we say that f is a faithful representation of G. In this

situation G ∼
= Image(f ); so that G is isomorphic to a subgroup of GL(n, |F).

Example 5.1.1 (i) The map f : G −→ GL(1, |F) given by f (g) = 1 |F for
all g ∈ G is called the trivial representation of G over |F. Notice that
GL(1, |F ) = |F ∗ .
(ii) Let G be a permutation group acting on a finite set X, where X =
{x1 , x2 , · · · , xn }. Define π : G −→ GL(n, |F) by π(g) = πg for all g ∈ G, where
πg is the permutation matrix induced by g on X. That is πg = (aij ) an
n × n matrix having 0 |F and 1 |F as entries in such way that

aij = 1 |F if g(xi ) = xj
= 0 |F otherwise.

Then π is a representation of G over |F, and π is called the permutation

representation of G on X.
(iii) Take X = G in part (ii). Define a permutation action on G by g : x −→
xg for all x ∈ G. Then the associated representation π is called the right
regular representation of G.

50
Exercise 5.1.1 Let N  G. Assume that ρ is a representation of G/N . Define
ρ : G −→ GL(n, |F ), where n is the degree of ρ̂, by ρ(g) = ρ(gN ). Then show
that ρ is a representation of G.

Exercise 5.1.2 Let N  G. Assume that ρ is a representation of degree n on

G. If N ≤ Ker(ρ), then show that the mapping ρ : G/N −→ GL(n, |F) given
by ρ(gN ) = ρ(g) is a representation of G/N.

Theorem 5.1.1 Let G be a group. Then the derived subgroup G0 lies in the
kernel of any representation of G of degree 1.

Proof. Assume that f : G −→ GL(1, |F) is a representation of degree one of

G. Let a, b ∈ G. Then

f (aba−1 b−1 ) = f (a)f (b)[f (a)]−1 [f (b)]−1 .

Since GL(1, |F ) = F ∗ is abelian we have

f (aba−1 b−1 ) = f (a)[f (a)]−1 f (b)[f (b)]−1 = 1 |F .

Thus [a, b] = aba−1 b−1 ∈ Kerf. Since G0 is generated by the set of all commu-
tators, G0 ⊆ Ker(f ).

Exercise 5.1.3 Let |F = GF (q) be the Galois Field of q elements, where

q = pk for some prime p. Show that |GL(n, |F)| = (q n −1).(q n −q) · · · (q n −q n−1 ).

Definition 5.1.2 (Special Linear Group SL(n, |F)) Let |F be any field.

SL(n, |F ) = {A | A ∈ GL(n, |F), det(A) = 1 |F }.

Then it is not difficult to show that SL(n, |F) ≤ GL(n, |F).

Theorem 5.1.2 Let F = GF (q) with q = pk for some prime p. Then

SL(n, |F)  GL(n, |F)

and
|SL(n, |F )| = |GL(n, |F)|/(q − 1).

Proof. Let ρ : GL(n, |F ) −→ F ∗ be given by ρ(A) = det(A), for all A ∈

GL(n, |F ). Then ρ is a homomorphism (check ). If a ∈ |F ∗ , then
 
a 0 0 ··· 0
 

 0 1 0 ··· 0 

 
ρ: 
 0 0 1 ··· 0 
 7−→ a,
 .. 

 . 

0 0 0 ··· 1

51
so that ρ is onto. We also have Ker(ρ) = {A | A ∈ GL(n, |F), det(A) =
1 |F } = SL(n, |F ). Since Ker(ρ)  GL(n, |F), SL(n, |F)  GL(n, |F). Now
since GL(n, |F )/Ker(ρ) ∼ = Image(ρ), we have that GL(n, |F)/SL(n, |F) ∼
=
F ∗ . Hence
|GL(n, |F )/SL(n, |F)| = |F ∗ | = q − 1.

Corollary 5.1.3 If ρ : G −→ GL(n, |F) is a representation of G, then ρ(g) ∈

SL(n, |F ) for all g ∈ G0 .
Proof. Let h = [a, b] be a commutator in G. Then we have ρ(h) = ρ(aba−1 b−1 ) =
ρ(a)ρ(b)ρ(a−1 )ρ(b−1 ). Now since

det(ρ(h)) = det(ρ(a)).det(ρ(b)).det(ρ(a−1 ))det(ρ(b−1 ))

= det(ρ(aa−1 )).det(ρ(bb−1 ))
= det(ρ(1G )).det(ρ(1G ))
= det(In ).det(In ) = 1,

we have ρ(h) ∈ SL(n, |F ).

Exercise 5.1.4 (Special triangular Group) Let ST L(N, |F) denote the set
of all invertible lower triangular n × n matrices whose diagonal entries are all
1 |F . Then ST L(n, |F ) ≤ GL(n, |F). Show that if |F = GF (q) where q = pk
for some prime p, then ST L(n, |F) is Sylow p-subgroup of GL(n, |F).

Definition 5.1.3 (Characters) Let f : G −→ GL(n, |F) be a representation

of G over the field |F. The function χ : G −→ |F defined by χ(g) = tr(f (g))
is called the character of f.

Definition 5.1.4 (Class functions) If φ : G −→ |F is a function that is

constant on conjugacy classes of G, that is φ(g) = φ(xgx−1 ), ∀x ∈ G, then we
say that φ is a class function.

Lemma 5.1.4 A character is a class function.

Proof. Let χ be a character of G. Then χ is afforded by a representation

ρ : G −→ GL(n, |F ). Let g ∈ G; then ∀x ∈ G we have

χ(xgx−1 ) = tr(ρ(xgx−1 ))
= tr(ρ(x).ρ(g).ρ(x−1 ))
= tr(ρ(x).ρ(g).[ρ(x)]−1 )
= tr(ρ(g)), see Note 5.1.1 below
= χ(g).

52
Note 5.1.1 Similar matrices have the same trace. If A = (aij ) and B = (bij )
are two matrices, then
n X
X n n X
X n
tr(AB) = ( aij bji ) = ( bji aij ) = tr(BA).
i=1 j=1 j=1 i=1

Now if B = P AP −1 , then tr(B) = tr(P AP −1 ) = tr(P −1 P A) = tr(A).

Note 5.1.2 If χ is a character afforded by a representation ρ : G −→ GL(n, |F),

then χ is not linear (in general) :

χ(gg 0 ) = tr(ρ(gg 0 )) = tr(ρ(g)ρ(g 0 )) 6= tr(ρ(g)) × tr(ρ(g 0 )) = χ(g) × χ(g 0 ).

Later we will show that χ is linear if and only if deg(ρ) = 1.

Definition 5.1.5 (Equivalent Representations) Two representations

ρ, φ : G −→ GL(n, |F ) are said to be equivalent if there exists a n × n matrix
P over |F such that P −1 ρ(g)P = φ(g), ∀g ∈ G.

Since similar matrices have the same trace, it follows that equivalent represen-
tations have the same character.

Theorem 5.1.5 Equivalent representations have the same character.

Proof. Let χ1 and χ2 be characters afforded by ρ1 and ρ2 two representations

of degree n over a field |F. Assume that ρ1 is equivalent to ρ2 . Then there is a
n × n matrix P such that P −1 ρ1 (g)P = ρ2 (g), ∀g ∈ G. Now ∀g ∈ G we have

χ2 (g) = tr(ρ2 (g)) = tr(P −1 ρ1 (g)P ) = tr(ρ1 (g)) = χ1 (g).

Hence χ1 = χ2 .

Definition 5.1.6 Let S be a set of (n × n) matrices over |F. We say that S

is reducible if ∃ m, k ∈ N
| , and there exists P ∈ GL(n, |F) such that ∀A ∈ S

we have
 
B 0
P AP −1 =  
C D

where B is an m×m matrix, D and C are k ×k and k ×m matrices respectively.

Here 0 denotes the m × k zero matrix.

53
 If there is no such P , we say that S is irreducible.
If C = 0, the zero k × m matrix, for all A ∈ S, then we say that S is fully
reducible.
We say that S is completely reducible if ∃ P ∈ GL(n, |F) such that
 
B1 0 · · 0
 
−1
 0 B2 0 · 0 
P AP =  , ∀A ∈ S,
 
 · · · · · 
 
0 0 · · Bk
where each Bi is irreducible.
Example 5.1.2 Let |F = C| ; consider
  
 a −b 
S=   | a, b ∈ C
| .
 b a 
   
1 −i 1 i
Then S is a reducible set over C| . Let P =   . Then P −1 =  
0 1 0 1
and
   
a −b a + ib 0
P −1  P =   , ∀a, b ∈ C
| .
b a b a − ib
 
1 i
In fact we can show that S is fully reducible. For this let P =  .
i 1
     
1 −i a −b a − ib 0
Then P −1 = 1
2
  and P −1  P =  .
−i 1 b a 0 a + ib
 
 1 1 
Exercise 5.1.5 Let S =   . Show that S is reducible, but it is not
 0 1 
fully reducible.

Definition 5.1.7 Let f : G −→ GL(n, |F) be a representation of G over |F.

Let S = Im(f ) = {f (g) | g ∈ G} . Then S ⊆ GL(n, |F). We say that f is
reducible, fully reducible, or completely reducible if S is reducible, fully
reducible or completely reducible.

Definition 5.1.8 (Sum of representations) let ρ : G −→ GL(n, |F) and

φ : G −→ GL(m, |F ) be two representations of G over |F. Define
ρ + φ : G −→ GL(n + m, |F ) by
 
ρ(g) 0n×m
(ρ + φ)(g) :=   = ρ(g) ⊕ φ(g),
0m×n φ(g)
for all g ∈ G. Then ρ + φ is a representation of G over |F, of degree n + m.

54
 If χ1 and χ2 are the characters of ρ and φ respectively, and if χ is the
character of ρ + φ, then ∀g ∈ G we have
 
ρ(g) 0
χ(g) = trace   = tr(ρ(g)) + tr(φ(g)) = χ1 (g) + χ2 (g) = (χ1 + χ2 )(g).
0 φ(g)

Hence χ = χ1 + χ2 .

Example 5.1.3 Let G =< a, b  > such that 2 2

 a =b =  1G and  ab = ba. Define
1 0 −1 0
f : G −→ GL(2, C| ) by f (a) =   , f (b) =   . Then f is a
0 −1 0 1
faithful representation of degree 2. It is not difficult to see that f is completely
reducible

Exercise 5.1.6 Represent the permutations of S3 as permutation matrices.

Calculate the character of this representation.

Theorem 5.1.6 (Maschke’s theorem) Let G be a finite group. Let f be a

representation of G over a filed |F whose characteristic is either 0 or is a prime
that does not divide |G|. If f is reducible, then f is fully reducible.

Proof. In fact we will show that if there is a matrix P such that ∀g ∈ G

 
A(g) 0
P −1 f (g)P =  
B(g) C(g)

then there is a matrix Q such that

 
A(g) 0
∀g ∈ G, Q−1 f (g)Q =  .
0 C(g)
 
Ir 0
Let f (g) = P −1 f (g)P and let L =   where r and s are the degrees
T Is
of A(g) and C(g) respectively, and T is an s × r matrix. We need to determine
T such that L is an invertible matrix over |F independent of g and
 
A(g) 0
L−1 f (g)L =   , ∀g ∈ G. (1)
0 C(g)

Then Q = P L. Relation (1) implies that

     
A(g) 0 Ir 0 Ir 0 A(g) 0
  =  ,
B(g) C(g) T Is T Is 0 C(g)

55
that is    
A(g) 0 A(g) 0
 = .
B(g) + C(g)T C(g) T.A(g) C(g)
Hence we find that

B(g) + C(g)T = T.A(g), ∀g ∈ G. (2).

Since f is a matrix representation of G, f (gh) = f (g).f (h) for all g, h ∈ G.

So for all g and h in G we have
    
A(gh) 0 A(g) 0 A(h) 0
 =  ,
B(gh) C(gh) B(g) C(g) B(h) C(h)

that is
   
A(gh) 0 A(g)A(h) 0
 = .
B(gh) C(gh) B(g)A(h) + C(g)B(h) C(g)C(h)

We obtain the relations:

(i) A(gh) = A(g)A(h)
(ii) C(gh) = C(g)C(h)
(iii) B(gh) = B(g)A(h) + C(g)B(h).

Relations (i) and (ii) show that A and C are matrix representations of G
over |F . By multiplying (iii) and A(h−1 ) we obtain that

B(gh)A(h−1 ) = B(g) + C(g)B(h)A(h−1 ) ∀g, h ∈ G. (3)

If we fix g and let h runs over all elements of G, then using (3) we get

B(gh)A(h−1 ) = C(g)B(h)A(h−1 )
X X X
B(g) +
h∈G h∈G h∈G

B(h)A(h−1 ).
X
= |G|B(g) + C(g) (4)
h∈G
Now let x = gh. Since h runs over all elements of G, so also does x. Hence

B(gh)A(h−1 ) = B(x)A(x−1 g) = B(x)A(x−1 )A(g)

X X X

h∈G x∈G x∈G

B(x)A(x−1 ))A(g).
X
=( (5)
x∈G

Now relations (4) and (5) give

B(x)A(x−1 ))A(g) = |G|B(g) + C(g)( B(h)A(h−1 )).

X X
( (6)
x∈G h∈G

56
Since the characteristic of |F does not divide |G|, |G| =
6 0 |F in |F. Hence we
can divide both sides of relation (6) by |G|. We get
1 X 1 X
( B(x)A(x−1 ))A(g) = B(g) + C(g)( B(h)A(h−1 )). (7)
|G| x∈G |G| h∈G

Finally by comparing relations (7) and (2), if we let

1 X
T = ( B(x)A(x−1 )),
|G| x∈G

then T satisfies the relation (2).

Theorem 5.1.7 [The general form of Maschke’s theorem] Let G be a finite

group and |F a field whose characteristic is either 0 or is a prime that does not
divide |G|. Then every representation of G over |F is completely reducible.

Proof. Let f be a representation of G over |F. If f is irreducible, then it is

completely reducible. Hence assume that f is reducible. Then by Maschke’s
theorem f is fully reducible, and therefore for all g ∈ G, f (g) is similar to
 
A(g) 0
 
0 C(g)
.
Since A and B are representations of G over F , we can apply Maschke’s
theorem to these representations. Repeating this process we obtain that f (g)
is similar to  
B1 (g) 0 · · 0
 
 0 B2 (g) 0 · 0 
,
 

 · · · · · 
 
0 0 · · Bk (g)
where Bi , 1 ≤ i ≤ k are all irreducible representations of G over |F.

Theorem 5.1.8 [Schur’s Lemma] Let ρ and φ be two irreducible representa-

tions, of degree n and m respectively, of a group G over a field |F. Assume
that there exists an m × n matrix P such that P ρ(g) = φ(g)P for all g ∈ G.
Then either P = 0m×n or P is non-singular so that ρ(g) = P −1 φ(g)P (that is
ρ and φ are equivalent representations of G).

Proof. Let r = rank(P ). Then there are non-singular matrices L and M such
that P = LEr M , where
 
0m−r×r 0m−r×n−r
Er =   ,
Ir 0r×n−r
m×n

57
and L and M are m × m and n × n matrices respectively. Since for all g ∈ G
we have P ρ(g) = φ(g)P, we obtain that

LEr M ρ(g) = φ(g)LEr M.

Hence
Er M ρ(g)M −1 = L−1 φ(g)LEr . (1)

Using the relation (1), we can partition the matrices M ρ(g)M −1 and L−1 φ(g)L
in the following way, provided that r 6= 0, and m 6= r or n 6= r:
 
A(g) B(g)
M ρ(g)M −1 =   ,
C(g) D(g)
n×n

and  
A0 (g) B 0 (g)
L−1 φ(g)L =   ,
C 0 (g) D0 (g)
m×m

where A(g) is r × r, B(g) is r × n − r, C(g) is n − r × r, D(g) is n − r × n − r,

A0 (g) is m − r × m − r, B 0 (g) is m − r × r, C 0 (g) is r × m − r, D0 (g) is r × r.
We can easily deduce that
 
0m−r×r 0m−r×n−r
Er M ρ(g)M −1 =  
A(g) B(g)

and  
B 0 (g) 0m−r×n−r
L−1 φ(g)LEr =  .
D0 (g) 0r×n−r
Now using the relation (1) we must have
   
0m−r×r 0m−r×n−r B 0 (g) 0m−r×n−r
 = .
A(g) B(g) D0 (g) 0r×n−r

Hence B 0 (g) = 0m−r×r and B(g) = 0r×m−r for all g ∈ G. This shows that ρ and
φ are reducible, which is a contradiction. Thus either r = 0 or m = n = r. If
r = 0, then P = 0m×n . If m = n = r, then P is invertible and ρ(g) = P −1 φ(g)P.

Definition 5.1.9 (Algebraically Closed Fields) . A field F is said to be

Algebraically closed if every polynomial equation p(x) = 0F with P (x) ∈ F [x]
has all its roots in F.

58
 For example the complex field C| is an algebraically closed field by the Fun-
damental Theorem of Algebra. The first proof for the fundamental theorem of
algebra was given by Gauss in his doctoral dissertation in 1799 at the age of
22 (in fact gauss gave several independent proofs, he published his last proof in
1849 at the age of 72).
For an algebraically closed field, the Schur’s Lemma (Theorem 5.1.8) has
the following noteworthy corollary.

Corollary 5.1.9 If ρ is an irreducible representation of degree n of a group G

over an algebraically closed field F , then the only matrices which commute with
all matrices ρ(g), g ∈ G, are the scalar matrices aIn , a ∈ F.

Proof. Let P be an n × n matrix such that P ρ(g) = ρ(g)P , for all g ∈ G. Then
for any a ∈ F we have

(aIn − P )ρ(g) = ρ(g)(aIn − P ), f or all g ∈ G (1)

Let m(x) = det(xIn − P ) be the characteristic polynomial of P. Since m(x) is

a polynomial over F and F is algebraically closed, there is a0 ∈ F such that
m(a0 ) = 0F . Hence det(a0 In − P ) = 0. So that a0 In − P is singular. Now using
relation (1) and Schur’s Lemma, we must have a0 In − P = 0. Thus P = a0 In .

Exercise 5.1.7 Let G be a finite group and ρ and φ be representations of

degrees n and m respectively, over a a field F. Assume that char(F ) does
not divide the order of G. Let S be an m × n matrix over F . Show that
Sρ(g) = φ(g)S for all g in G if and only if there is an m × n matrix T over F
such that S =
P −1 )T ρ(g).
g∈G φ(g

Exercise 5.1.8 Maschke’s Theorem becomes false if the hypothesis that char(F )
does not divide the order of G is omitted. Let F = GF (2) and G =< a >
be acyclic  of order 2. Define ρ : G → GL(2, F ) by ρ(1G ) = I2 and
 group
1 1
ρ(a) =   . Show that ρ is reducible but not fully reducible. (Hint use
0 1
Exercise 5.1.5.)

Exercise 5.1.9 Show that Corollary 5.1.9 is false if F = |R .

Theorem 5.1.10 Let ρ and φ be two inequivalent irreducible representations,

of degrees n and m respectively, of a group G over a field F . If T is an m × n
matrix over F , then
φ(g −1 )T ρ(g) = 0m×n .
X

g∈G

59
Proof. Let S =
P −1 )T ρ(g). Then for any x ∈ G we have
g∈G φ(g

φ(g −1 )T ρ(gx) = φ(x)φ(x−1 g −1 )T ρ(gx) =

X X
Sρ(x) =
g∈G g∈G

φ((gx)−1 )T ρ(gx)) = φ(x)( φ(z −1 )T ρ(z)) = φ(x)S.

X X
φ(x)(
g∈G z∈G

Since ρ and φ are inequivalent irreducible representations of G, by Schur’s

lemma we must have S = 0m×n .

Definition 5.1.10 Let G be a finite group and F a field such that char(F )
does not divide the order of G. If ρ and φ are two functions from G into F , we
define an inner product <, > by the following rule:
1 X
< ρ, φ >= ρ(g)φ(g −1 ),
|G| g∈G

where 1
|G| stands for |G|−1 in F.

Theorem 5.1.11 The inner product <, > defined above is bilinear and sym-
metric:

(i) < ρ1 + ρ2 , φ >=< ρ1 , φ > + < ρ2 , φ >,

(ii) < ρ, φ1 + φ2 >=< ρ, φ1 > + < ρ, φ2 >,

(iii) < aρ, φ >= a < ρ, φ >=< ρ, aφ >, for all a ∈ F,

(iv) < ρ, φ >=< φ, ρ > .

Proof. the bilinear properties (i), (ii) and (iii) are easy to verify. Let us prove
the the symmetry:
1 X 1 X 1 X
< ρ, φ >= ρ(g)φ(g −1 ) = ρ(g −1 )φ(g) = φ(g)ρ(g −1 ) =< φ, ρ > .
|G| g∈G |G| g∈G |G| g∈G

Note 5.1.3 If ρ : G −→ F ∗ is a group homomorphism, then

1 X 1 X 1 X 1
< ρ, ρ >= ρ(g)ρ(g −1 ) = ρ(1G ) = 1F = ×(|G|1F ) = 1F .
|G| g∈G |G| g∈G |G| g∈G |G|

60
5.2 Characters of Finite Groups
In this section, unless explicit exception is made, the group G will be finite and
all representations and matrices will be over the complex field C| .

Note 5.2.1 By the general form of Maschke’s theorem (Theorem 5.1.7), all
representations of G are completely reducible.

Note 5.2.2 If If ρ : G −→ GL(n, C)

| is a representation of G, then we denote
the (i, j) entry of ρ(g) by ρij (g). Hence we can regard ρij is a map from G into
C| .

Theorem 5.2.1 [Orthogonality of irreducible representations] Let G be a finite

group and ρ and φ two irreducible representations of G.

(i) If ρ and φ are inequivalent, then < ρrs , φij >= 0, for all i, j, r, s.

(ii) < ρrs , ρij >= δis δjr /deg(ρ).

Proof. (i) Using Theorem 5.1.10 we have

φ(g −1 )Ejr ρ(g) = 0m×n ,

X
(1)
g∈G

where Ejr is the m × n matrix with (j, r) entry 1 and other entries 0, with
n = deg(ρ) and m = deg(φ). Now from (1) we get
1 X
φ(g −1 )Ejr ρ(g) = 0m×n . (2)
|G| g∈G

Since the (i, s) entry of the left hand-side of the relation (2) is
1 X
φij (g −1 )ρrs (g) =< φij , ρrs >,
|G| g∈G

we have < φij , ρrs >= 0.

(ii) Let Sjr = 1 P −1 )E Then for any x ∈ G we have (see
|G| g∈G ρ(g jr ρ(g).
Exercise 5.1.7) Sjr ρ(x) = ρ(x)Sjr , and from Corollary 5.1.9 we deduce that Sjr
is an scalar matrix. Hence let Sjr = λjr In , where λjr ∈ C| . Then we have
1 X
λjr In = ρ(g −1 )Ejr ρ(g). (3)
|G| g∈G

By comparing the (i, s) entry of the left and right hand-side of (3), we get
1 X
λjr δis = ρij (g −1 )ρrs (g).
|G| g∈G

61
That is
λjr δis =< ρij , ρrs > .

Since < ρij , ρrs >=< ρrs , ρij >, we get

λjr δis =< ρij , ρrs >=< ρrs , ρij >= λsi δrj . (4)

Now if i 6= s or j 6= r, we have < ρij , ρrs >= 0 and (ii) holds. Suppose that
i = s and j = r. Then by (4) we have

< ρij , ρji >= λjj = λii . (5)

Hence we have
λ11 = λ22 = · · · = λnn = λ ∈ C| ,

so that n n
X X
nλ = λii = < ρi1 , ρ1i >, by (5),
i=1 i=1
n
1 X
ρ1i (g −1 )ρi1 (g))
X
= (
i=1
|G| g∈G
n
1 X X
= ( ρ1i (g −1 )ρi1 (g)). (6)
|G| g∈G i=1
Pn −1 )ρ
Since i=1 ρ1i (g i1 (g) is the (1, 1)−entry of ρ(g −1 )ρ(g) and since ρ(g −1 )ρ(g) =
ρ(g −1 g) = ρ(1G ) = In , we have
n
ρ1i (g −1 )ρi1 (g) = 1.
X

i=1

Now relation (6) implies that

1 X 1
nλ = 1= × |G| = 1.
|G| g∈G |G|

Hence λ = n1 . Therefore by (5) we get

1
< ρij , ρji >= = δii δjj /deg(ρ).
n

Theorem 5.2.2 [Orthogonality of irreducible characters] Let G be a finite group

and ρ and φ two irreducible representations of G. If χρ and χφ are characters
of ρ and φ respectively, then

(i) < χρ , χφ >= 1 if ρ and φ are equivalent, and

< χρ , χφ >= 0 otherwise,

62
 (ii) < χρ , χρ >= 1.

Proof. (i) Let n = deg(ρ) and m = deg(φ). Then we have

1 X
< χρ , χφ > = χρ (g)χφ (g −1 )
|G| g∈G
n n
1 X X
{[ ρii (g)][ φjj (g −1 )]}
X
=
|G| g∈G i=1 i=1
n X
m
1 X
ρii (g)φjj (g −1 )]
X
= [
i=1 j=1
|G| g∈G
Xn X m
= < ρii , φjj > . (1)
i=1 j=1

If ρ and φ are inequivalent, then < ρii , φjj >= 0 by part (ii) of Theorem
5.2.1, Hence using the relation (1) above we get < χρ , χφ >= 0.
If If ρ and φ are equivalent, then χρ = χφ by Theorem 5.1.5. Now we have

< χρ , χφ > = < χρ , χρ >

n X
X n
= < ρii , ρjj >, by (1),
i=1 j=1
n n
X X 1
= < ρii , ρii >= , by T heorem 5.2.1,
i=1 i=1
n
1
= n× = 1.
n
(ii) As above.

Corollary 5.2.3 Two irreducible representations of a finite group G are equiv-

alent if and only if they have the same characters.

Proof. Let ρ and φ be two irreducible representations of G. If ρ and φ are

equivalent then χρ = χφ by Theorem 5.1.5. Conversely assume that χρ = χφ .
Then by Theorem 5.2.2 (part (ii)) we have < χρ , χφ >= 1. Thus ρ and φ are
equivalent by Theorem 5.2.2.

Note 5.2.3 Maschke’s theorem (Theorem 5.1.7) implies that if ρ is a represen-

Pk
tation of G, then ρ is equivalent to i=1 ρi where ρi , s are irreducible represen-
Pk
tations of G. We also have χρ = i=1 χρi .

Pk
Exercise 5.2.1 If ρ is a representation of G such that ρ is equivalent to i=1 ρi
where ρi , s are irreducible representations of G, then ρi are unique up to equiv-
alence.

63
Theorem 5.2.4 (Generalisation of Corollary 5.2.3) Two representations
of a finite group G are equivalent if and only if they have the same characters.

Proof. Let ρ and φ be two representations of G. If ρ and φ are equivalent then

χρ = χφ by Theorem 5.1.5.
Conversely assume that χρ = χφ . Assume that an irreducible representation
ψi appears mi times in ρ and ni times in φ. Then adding dummy terms if
Pk Pk Pk
necessary, we have ρ ∼ i=1 mi ψi and φ ∼ i=1 ni ψi . Then χρ = i=1 mi χψi
Pk Pk Pk
and χφ = i=1 ni χψi . Since χρ = χφ , we have i=1 mi χψi = i=1 ni χψi . Hence
for any j we have
k
X
mj = < mj χψj , χψj >=< mi χψi , χψj >
i=1
k
X
= < ni χψi , χψj >= nj .
i=1
Pk Pk
Thus i=1 mi ψi = i=1 ni ψi . So ρ ∼ φ.

Definition 5.2.1 (Irreducible Characters) If χρ is a character afforded by

a representation ρ of G, then we say that χρ is an irreducible character if ρ
is an irreducible representation. Notice that if χ is an irreducible character of
G and if φ is a representation of G such that χφ = χ, then φ is also irreducible
by Theorem 5.2.4.

Theorem 5.2.5 The set of all irreducible characters of G is a linearly inde-

pendent set over C| .

Proof. Let χ1 , χ2 , · · ·, χm be a finite set of distinct irreducible characters of a

finite group G. assume that there are λ1 , λ2 , · · ·, λm in Csuch
| that

λ1 χ1 + λ2 χ2 + · · · + λm χm = 0, (1)

the zero function from G into C| . Since χi 6= χj for i 6= j, χi and χj are afforded
by inequivalent irreducible representations of G (by Theorem 5.2.4). Hence we
have < χi , χj >= 1 if i = j and 0 otherwise. Now using the relation (1) above
we get
m
X
0 = < 0, χj >=< λi χi , χj >
i=1
m
X
= < λi < χi , χj >= λj , 1 ≤ j ≤ m.
i=1

Therefore {χ1 , χ2 , · · ·, χm } is a linearly independent set over C.

|

64
Theorem 5.2.6 If G has r distinct conjugacy classes of elements, then G has
at most r irreducible characters.

Proof. Let S = {[g1 ], [g2 ], ···, [gr ]} be the set of all conjugacy classes of elements
| Define fi : S −→
of G and let V be the vector space of functions from S into C.
C| , for 1 ≤ i ≤ r, by fi ([gi ]) = 1 and fi ([gj ]) = 0 if i 6= j. Then {f1 , f2 , · · ·, fr }
is a basis for V . Thus dim(V ) = r.
Let Irr(G) denote the set of all distinct irreducible characters of G. Since
a character is a class function, we can regard Irr(G) as a subset of V. Now
since Irr(G) is a linearly independent subset of V by Theorem 5.2.5, we have
|Irr(G)| ≤ dim(V ), that is |Irr(G)| ≤ r.
Pk
Exercise 5.2.2 (i) If χ = i=1 λi χi , where χi are distinct irreducible char-
acters of G and λi are non-negative integers, show that < χ, χ >=
Pk 2
i=1 λi .

(ii) If χ is a character of G, then show that χ is irreducible if and only if

< χ, χ >= 1.

Assume that {C1 , C2 , · · ·, Cr } is the set of all conjugacy classes of elements

of G with C1 = 1G . Unless otherwise stated gi will denote an arbitrary element
of the conjugacy class Ci and we put

hi = |Ci | = |G|/|CG (gi )|.

Let Irr(G) = {χ1 , χ2 , · · ·, χk } be the set of all distinct irreducible charac-

ters of G with the assumption that χ1 is the character afforded by the trivial
representation ρ(g) = 1 for all g in G.

Theorem 5.2.7 We have the following

= |G|,
P
(i) g∈G χ1 (g)

= 0, if i 6= 1,
P
(ii) g∈G χi (g)
Pr
(iii) j=1 hj χi (gj ) = δ1i |G|.

Proof. (i) Since χ1 (g) = 1 for all g in G, we have

X X
χ1 (g) = 1 = |G|.
g∈G g∈G

65
 (ii) If i 6= 1, then < χi , χ1 >= 0. hence
1 X 1 X
0= χi (g)χ1 (g −1 ) = χi (g).
|G| g∈G |G| g∈G

= 0 × |G| = 0.
P
Thus g∈G χi (g)
= |G| and hence
P
(iii) If i = 1, then by part (i) we have g∈G χ1 (g)

X r
X
|G| = χ1 (g) = hj χ1 (gj ).
g∈G j=1

If i 6= 1, then by part (ii) we have

P
g∈G χi (g) = 0 = δ1i and hence

X r
X
0 = δ1i = χi (g) = hj χi (gj ).
g∈G j=1

Exercise 5.2.3 (i) Let ρ be an irreducible representation of G. Show that

deg(ρ) = 1 if and only if ker(ρ) ≥ G0 .

(ii) Show that all irreducible representations of an abelian group are of degree
one.

Note 5.2.4 If χρ and χφ are two characters of G, we know that

1 X
< χρ , χφ >= χρ (g)χφ (g −1 ).
|G| g∈G

Hence
r
1 X
< χρ , χφ > = hi χρ (gi )χφ (gi−1 )
|G| i=1
r
hi
χρ (gi )χφ (gi−1 )
X
=
i=1
|G|
r
1
χρ (gi )χφ (gi−1 ).
X
=
i=1
|CG (gi )|

Example 5.2.1 Consider the symmetric group S3 . Representing the elements

of S3 as permutation matrices, we obtain the following faithful representation
π : S3 −→ GL(3, C| )
     
1 0 0 0 1 0 1 0 0
     
 0 1 0  , π((12)) =  1 0 0  , π((23)) =  0 0 1  ,
π(1S3 ) =      

0 0 1 0 0 1 0 1 0

66
      
0 0 1 0 1 0 0 0 1
     
 0 1 0  , π((123)) =  0 0 1  , π((132)) =  1 0 0  .
π((13)) =      

1 0 0 1 0 0 0 1 0
Then it is easy to see that

χπ (1S3 ) = 3, χπ ((12)) = χπ ((13)) = χπ ((23)) = 1, χπ ((123)) = χπ ((132)) = 0.

Notice that χπ (g) is equal to the number of fixed points of g on {1, 2, 3}.
Now
1
< χπ , χπ > = {1[χπ (1S3 )]2 + 3[χπ ((12))]2 + 2[χπ ((123))χπ ((132))]}
6
1
= {1[9] + 3[1] + 2[0]} = 12/6 = 2.
6
this shows that χπ (and hence π) is not irreducible. Hence χπ = χi + χj , where
χi and χj are two distinct irreducible characters of S3 . (Note: If χ is a character
of a group G such that < χ, χ >= 2, then there are χi , χj ∈ Irr(G) such that
Pk
χ = χi + χj , i 6= j. Because if χ = i=1 λi χ, then < χ, χ >= 2 implies
k
X k
X k
X
2 =< χ, χ >=< λi χ, λi χ >= λ2i .
i=1 i=1 i=1

Hence there are i 6= j for which λi = λj = 1. So that χ = χi + χj .)

Since
deg(χπ ) = 3 = χπ (1S3 ) = χi (1S3 ) + χj (1S3 ),

W.L.O.G we may assume that deg(χi ) = 1 and deg(χj ) = 2. Let us now consider
the actions of χ1 (the trivial character) and χπ on the conjugacy clases of S3 :

S3 C1 C2 C3
Class Rep 1S3 (12) (123)
hi 1 3 2
χ1 1 1 1
χπ 3 1 0
Now
1
< χπ , χ1 > = {1[χπ (1S3 )χ1 (1S3 ] + 3[χπ ((12))χ1 ((12))] + 2[χπ ((123))χ1 ((132))]}
6
1
= {3 + 3 + 0} = 1.
6
thus χ1 appears only once in χπ and hence χπ = χ1 +χ2 where χ2 is a nontrivial
irreducible character of S3 . Now we have

χ2 (g) = χπ (g) − χ1 (g) = χπ (g) − 1, f or all g ∈ S3 .

67
So that we have

χ2 (1S3 ) = 3 − 1 = 2, χ2 ((12)) = χ2 ((13)) = χ2 ((23)) = 1 − 1 = 0

and
χ2 ((123)) = χ2 ((132)) = 0 − 1 = −1.

Since |Irr(S3 )| ≤ 3, We may have one more irreducible character [in fact
later we will show that for any finite group G, |Irr(S3 )| = r, the number of
conjugacy classes if G.]. Define ρ : S3 −→ C| by ρ(g) = 1 if g is even and
ρ(g) = −1 if g is odd. Then ρ is a representation of degree 1 with χρ = ρ.
Notice that χρ 6= χ1 and

χρ (1S3 ) = χρ ((132)) = χρ ((123)) = 1

and
χρ ((12)) = χρ ((13)) = χρ ((23)) = −1.

Since deg(χρ ) = 1, χρ is irreducible (note that

1 1
< χρ , χρ >= [1(1) + 3(−1)(−1) + 2(1)(1)] = [1 + 3 + 2] = 1.)
6 6
This character is the third irreducible character of S3 , namely χ3 . We are now
able to produce the following table for S3 , which is called the Character Table
of S3 over C| .

Class Rep 1S3 (12) (123)

hi 1 3 2
χ1 1 1 1
χ2 2 0 −1
χ3 1 −1 1
Notice that

< χ1 , χ2 >=< χ1 , χ3 >=< χ2 , χ3 >= 0;

X
χ1 (g) = 1(1) + 3(1) + 2(1) = 6 = |S3 |;
g∈G
X
χ2 (g) = 1(2) + 3(01) + 2(−1) = 0;
g∈G
X
χ3 (g) = 1(1) + 3(−1) + 2(1) = 0.
g∈G

68
 Let φ : S3 −→ GL(2, C| ) be given by
     
1 0 0 1 −1 −1
φ(1S3 ) =   , φ((12)) =   , φ((13)) =  ,
0 1 1 0 0 1
     
1 0 −1 −1 0 1
φ((23)) =   , φ((123)) =   , φ((132)) =  .
−1 −1 1 0 −1 −1
Then φ is a faithful representation of S3 with χφ = χ2 . Hence φ is an irreducible
representation of S3 .

Theorem 5.2.8 (Regular Representation) Let χπ be the character afforded

by the right regular representation of G. Let k = |Irr(G)|. Then we have
Pk
(i) χπ = i=1 χi (1G )χi ,
Pk 2
(ii) χπ (1G ) = i=1 [χi (1G )] = |G|,
Pk
(iii) χπ (g) = i=1 χi (1G )χi (g) = 0, for all g ∈ G − {1G }.
Pk
Proof. Assume that χπ = i=1 ni χi , where ni are non-negative integers. We
claim that ni = deg(χi ). We know that π(g) is a permutation on G, for all
g ∈ G. Since xg = x if and only if g = 1G , π(g) moves all the letters if g 6= 1G .
Hence χπ (g) =| G |, if g = 1G , and 0 otherwise.
Since
k
X k
X
< χπ , χj >=< ni χi , χj >= ni < χi , χj >,
i=1 i=1
by the orthogonality of irreducible characters we have

< χπ , χj >= nj < χj , χj >= nj .

Thus
1 X
nj = < χπ , χj >= χπ (g)χj (g −1 )
| G | g∈G
1
= (| G | χj (1G )) = χj (1G ) = deg(χj ), f or all j.
|G|
(i) By above
k
X k
X
χπ = ni χi = χi (1G )χi .
i=1 i=1
(ii) Since χπ (1G ) =| G |, by part (i) we have
k
X k
X
χπ (1G ) =| G |= χi (1G )χi (1G) = [χi (1G )]2 .
i=1 i=1

69
 (iii) Since χπ (g) = 0 for all g ∈ G − {1G }, by part (i) we have
k
X
0 = χπ (g) = χi (1G )χi (g).
i=1

Exercise 5.2.4 Let A be a square matrix over a field F. Assume that for some
| we have An = I, the identity matrix. If F contains all the nth roots of
n∈N
1, show that A is similar to a diagonal matrix.

Lemma 5.2.9 If ρ is a representation of G and g is an element of G, show

that there is a representation φ of G such that φ is equivalent to ρ and φ(g) is
a diagonal matrix.

Proof. Let | G |= n. Then g n = 1G , so that [ρ(g)]n = Im , where m = deg(ρ).

Since C| contains all the solutions for the equation xn = 1, ρ(g) is similar to
a diagonal matrix Dg . So there is a non-singular matrix P such that Dg =
P ρ(g)P −1 . Now define φ : G → GL(m, C)
| by φ(h) = P ρ(h)P −1 , for all h in G.

Then φ is a representation of G equivalent to ρ with φ(g) diagonal.

Definition 5.2.2 (Algebraic Integers) A complex number α is said to be an

Algebraic Integer if it is a root of an equation of the form

xn + a1 xn−1 + a2 xn−2 + · · · + an−1 x + an = 0, ai ∈ ZZ .

Remark 5.2.1 (Algebraic Numbers) A complex number α is said to be an

Algebraic Number if there is p(x) ∈ Q[x]
| such that P (α) = 0. It can be
|
shown that the set of all algebraic numbers is a subfield of C. If α is not an
algebraic number, then we say that it is Transcendental. For example i and
√
2 are algebraic numbers (in fact they are algebraic integers). Hermite, C
(1822–1905) and later Hilbert, D proved that e is transcendental. Lindemann,
CLF (1852–1939) in 1882 proved the transcendence of π. Hilbert’s 7th problem
is concerned with the transcendence of complex numbers of the form ab :
Hilbert’s Seventh Problem If a, b ∈ C| such that a is an algebraic number
/ {0, 1}, and b is an irrational algebraic number, then ab is transcendental.
and a ∈
A O Gelfond in 1934 proved that Hilbert’s seventh problem is true. For
√
example 2 2, 2i and ii are transcendental. But what about the case when
a and b are both transcendental? It is not known whether π π , π e or ee is
transcendental. However note that since
1 1
eπ = = = i−2i ,
e−π i2i
eπ is transcendental.

70
 Now we establish some basic results on algebraic integers. In th following
we show that the set of all algebraic integers form a ring. This ring plays a
fundamental role in Number Theory.

Lemma 5.2.10 Let α1 , α2 , · · · , αk be complex numbers, not all zero, and sup-
pose that α ∈ C| satisfy equations of the form
k
X
ααi = aij αj , i = 1, 2, · · · , k, (1)
j=1

where aij ∈ ZZ . Then α is an algebraic integer.

Proof. Equations in (1) represents a linear homogeneous system for α1 , α2 , · · · , αk .

Since, by the hypothesis, system (1) has non-zero solution, the determinant of
the coefficient matrix must be equal to zero, that is
 
α − a11 −a12 −a13 · · · −a1k
 

 −a21 α − a22 −a23 · · · −a2k 

 
det 
 ··· ··· ··· ··· ···  = 0.

 

 ··· ··· ··· ··· ··· 

−ak1 −ak2 −ak3 · · · α − akk

We can see that the above determinant is a monic polynomial of degree k in α

with integer coefficients. Hence α is an algebraic integer.

Lemma 5.2.11 If α and β are algebraic integers, so are α + β and αβ.

Proof. Suppose that α and β satisfy the following polynomial equations

αr = a1 αr−1 + a2 αr−2 + · · · + ar−1 α + ar , ai ∈ ZZ ,

β s = b1 β s−1 + b2 β s−2 + · · · + bs−1 β + bs , bi ∈ ZZ .

Then for any non-negative integer l, αl can be written as a linear combination

(with integer coefficients) of 1, α, α2 , · · · , αr−1 . Similarly for any non-negative
integer m, β m can be written as a linear combination (with integer coefficients)
of 1, β, β 2 , · · · , β s−1 .
Let α1 , α2 , · · · , αk be the products αi β j , where i.j ∈ ZZ , 0 ≤ i ≤ r − 1 and
0 ≤ j ≤ s − 1, arranged in some fixed order. Then any product of the form
αl β m can be represented in terms of αi β j , that is in terms of α1 , α2 , · · · , αk with
integer coefficients. Hence there are equations
k
X
(α + β)αi = cij αj , 0 ≤ i ≤ k, cij ∈ ZZ ,
j=1

71
 k
X
(αβ)αi = dij αj , 0 ≤ i ≤ k, dij ∈ ZZ .
j=1

Now Lemma 5.2.10 implies that α + β and αβ are algebraic integers.

Exercise 5.2.5 Let α ∈ Q.

| If α is an algebraic integer, show that α ∈ Z
Z.

Theorem 5.2.12 If χ is a character of a group G, then for any g ∈ G, χ(g) is

an algebraic integer.

Proof. Since G is finite, g n = 1G for some n ∈ N

| . Let ρ be a representation

of degree m of G that affords χ. Then [ρ(g)]n = Im , and by Lemma 5.2.9 ρ(g)

is similar to a diagonal matrix. W.L.O.G we may assume that ρ(g) itself is
diagonal (because similar matrices have the same trace). So let
 
1 0 0 ··· 0
 
 0
 2 0 · · · 0  
 
ρ(g) = diag(1 , 2 , · · · , m ) =  · · · · · · · · · · · · · · · 

,
 
 ··· ··· ··· ··· ··· 
 
0 0 · · · · · · m

with i ∈ C| . Now since [ρ(g)]n = Im , we have ni = 1, which imply that i ’s

are nth roots of unity and hence are all algebraic integers. Since χ(g) =
Pm
trace(ρ(g)) = i=1 i , by Lemma 5.2.11 we have that χ(g) is an algebraic
integer. In fact χ(g) is a sum of nth roots of unity, where n = o(g).
If bar (-) denotes the complex conjugation a + bi = a − bi in C| , then we
have the following result on the conjugation of character values:

Corollary 5.2.13 If χ is a character of a group G, then for any g ∈ G we

have χ(g −1 ) = χ(g).
Pm
Proof. By the Theorem 5.2.12, we have χ(g) = j=1 j , where i ’s are nth
roots of unity with n = o(g) and ρ(g) similar to diag(1 , 2 , · · · , m ). Since
ρ(g −1 ) = [ρ(g)]−1 , ρ(g −1 ) is similar to

[diag(1 , 2 , · · · , m )]−1 = diag(−1 −1 −1

1 , 2 , · · · , m )
Pm −1 2kj π
and hence χ(g −1 ) = j=1 j . We know that j = exp( n i) for some kj ∈ ZZ
such that 0 ≤ kj ≤ n − 1. Since j j = |j |2 = 1, we deduce that j = −1
j for
all 0 ≤ j ≤ m. Hence
m m m
χ(g −1 ) = −1
X X X
j = j = ( j ) = χ(g).
j=1 j=1 j=1

72
Exercise 5.2.6 Let ρ be a representation of a group G. Assume that χ is the
character afforded by ρ. Show that

(i) |χ(g)| ≤ χ(1G ), for all g ∈ G.

(ii) If |χ(g)| = χ(1G ), then ρ(g) is a scalar matrix.

(iii)* χ(g) = χ(1G ) if and only if g ∈ ker(ρ).

Definition 5.2.3 (F-Algebra) If F is afield and A is a vector space over F ,

then we say that A is an F -Algebra if

(i) A is a ring with identity,

(ii) for all λ ∈ F and x, y ∈ A, we have λ(xy) = λ(x)y = x(λy).

Example 5.2.2 (i) Mn×n (F ) is the algebra of all n × n matrices over a field
F.

(ii) Let V be vector space over F . Consider End(V ) = L(V, V ) = HomF (V, V ).
Then End(V ) is a ring with identity under the addition and compo-
sition of linear transformations on V , that is (f + g)(α) = f (α) + g(α)
and (f ◦ g)(α) = f (g(α)). for all α ∈ V and all f, g ∈ End(V ). De-
fine the scalar multiplication on End(V ) by (λf )(α) = λf (α), for all
f ∈ End(V ) and for all α ∈ V. Then End(V ) is a vector space over F.
Now

[λ(f ◦ g)](α) = λ[(f ◦ g)(α)] = λ[f (g(α))]

= (λf )(g(α)) = [(λf ) ◦ g](α).

Hence λ(f ◦ g) = (λf ) ◦ g. Similarly we have

[λ(f ◦ g)](α) = λ[(f ◦ g)(α)]

= (f ◦ g)(λα), since f ◦ g ∈ End(V )
= f (g(λα)) = f (λg(α))
= [f ◦ (λg)](α),

so that λ(f ◦ g) = f ◦ (λg).

Thus we have shown that End(V )is an F -Algebra.

Now we introduce another example of an algebra, the one which plays an

important part in the theory of representations, namely the Group Algebra
C| [G].

73
Definition 5.2.4 (Group Algebra) Let G be a finite group and F be any
field. Then by F [G] we mean the set of formal forms { : λg ∈ F }.
P
g∈G λg .g
We define the operations on F [G] by
P P P
(i) g∈G λg g + g∈G µg g := g∈G (λg + µg )g,

λ ∈ F,
P P
(ii) λ( g∈G λg g) := g∈G (λλg )g,
P P P P
(iii) ( g∈G λg g).( g∈G µg g) := g∈G [ h∈G λh µh−1 g ]g.

Notice that the definition of multiplication given in (iii) is the result of

assuming linearity and the multiplication in G. Under the above operations,
F [G] is an F-algebra. The element of F [G] for which λg = 1F and λh = 0F if
h 6= g is identified by g, that is 1F .g = g. Under this identification we embed G
into F [G] and in fact G becomes a basis for F [G].

Remark 5.2.2 (i) If |G| > 1, then F [G] has always zero divisors: Let g ∈ G
such that o(g) = m 6= 1. Then 1G − g and 1G + g + g 2 + · · · + g m−1 are
two non-zero elements of F [G], and we have

(1G − g).(1G + g + g 2 + · · · + g m−1 ) = 1G − g m = 1G − 1G = 0.

√
(ii) Consider the group G = V4 = {e, a, b, c}, F = C| . let e + ia + 2 b and
√
2 b − ic be two elements of C[G].
| Then
√ √ √ √ √
(e + ia + 2 b).( 2 b − ic) = 2 b − ic + 2 iab − i2 ac + 2b2 − i 2 bc
√ √ √
= 2 b − ic + 2 ic + b + 2e − i 2 a
√ √ √
= 2e − i 2 a + ( 2 + 1)b + i( 2 − 1)c.

Alternatively we can use part (iii) of the Definition 5.2.4, and we get
√ √ √ √
(e + ia + 2 b).( 2 b − ic) = (0 + 0 + 2 × 2 + 0)e
√
+ (1 × 0 + i × 0 + 2 × −i + 0)a
√ √
+ ( 2 + i × −i + 0 + 0)b + (−1 + 2 + 0 + 0)c.

(iii) Obviously G is a subgroup of UF [G] .

Exercise 5.2.7 (i) Let G = D8 =< r, s : r4 = s2 = e, rs = sr−1 > . Assume

that α = r2 + r − 2s and β = −3r2 + rs are two elements of the integral
group ring ZZ [G]. Compute βα, αβ − βα and βαβ.

74
 (ii) Consider the following elements of ZZ [S3 ]:

α = 3(1 2) − 5(2 3) + 14(1 2 3), β = 6e + 2(2 3) − 7(1 2 3),

where e = 1S3 . Compute the following elements of ZZ [S3 ] : 2α−3β, βα, αβ.

Definition 5.2.5 (Class Sums) Let C1 , C2 , · · · , Cr be the conjugacy classes

of elements of G. For 1 ≤ i ≤ r we define the Class Sums Ki by Ki =
P
g∈Ci g.
Then clearly Ki ∈ F [G] and we have the following result:

Theorem 5.2.14 The set {K1 , K2 , · · · , Kr } is a basis for the centre of the
group ring C| [G].

Proof. If g ∈ G, then g −1 Ci g = Ci . Hence we have

g −1 Ki g = g −1 ( g −1 hg = h0 = K i .
X X X
h)g =
h∈Ci h∈Ci h0 ∈Ci

Thus Ki g = gKi for all g ∈ G. Hence Ki ∈ Z(C[G]),

| |
where Z(C[G]) denotes
the centre of C| [G].
Since distinct conjugacy classes are disjoint, {K1 , K2 , · · · , Kr } is a linearly
P
independent set (why?). Now let u = |
g∈G λg g be an element of Z(C[G]). Let
x ∈ G. Then
λg (xgx−1 )x,
X X
xu = λg xg =
g∈G g∈G

λxgx−1 (xgx−1 )x,

X X
ux = λg (gx) =
g∈G g∈G

and since ux = xu, we get

P −1 P −1
g∈G λg (xgx )x = g∈G λxgx−1 (xgx )x. Hence
λg = λxgx−1 for all x ∈ G. Thus the coefficients of all conjgates of g are the
Pr Pr
Thus {K1 , K2 , · · · , Kr }
P
same in u. Hence u = i=1 (λi g∈Ci g) = i=1 λi Ki .
is a basis for Z(C| [G]).

Remark 5.2.3 Let ρ be a representation of G. Then ρ is a homomorphism

from G into GL(n, C| ) for some n ∈ N
| . We can extend ρ by linearity to an

C| -algebra homomorphism ρ : C[G]

| −→ Mn×n (C).
| |
Conversely if ρ : C[G] −→
Mn×n (C| ) is a representation of C[G]
| |
(that is ρ is an C-algebra homomorphism),
then ρ(1G ) = In . It follows that for all g ∈ G, ρ(g) is non-singular and [ρ(g)]−1
is equal to ρ(g −1 ). Hence the restriction of ρ to G (note that G ⊆ C[G])
| will be
a representation of G over C| .

|
Theorem 5.2.15 If ρ is an irreducible representation of degree m of C[G] with
the character χ, then

75
 (i) ρ(Ki ) = di Im , di ∈ C| ,

Pr
(ii) Ki Kj = k=1 λijk Kk , λijk ∈N
| ∪ {0},

Pr
(iii) di dj = k=1 λijk dk ,

(iv) di = hi χ(gi )/χ(1G ), hi = |Ci |, gi ∈ Ci .

Proof. (i) Since Ki ∈ Z(C| [G]) by Theorem 5.2.14, ρ(Ki ) commutes with all
elements of ρ(G). Now since ρ is irreducible, it follows from Corollary 5.1.9 that
ρ(Ki ) = di Im for some di ∈ C| .
(ii) Since Ki , Kj ∈ Z(C| [G]), we have Ki Kj ∈ Z(C[G]).
| So by Theorem
Pr
5.2.14, ∃λijk ∈ C| such that Ki Kj = k=1 λijk Kk . If we write this equation
in terms of elements of G, then since the coefficients on the left hand-side are
non-negative integers, λijk must be non-negative integers.
(iii) Using parts (i) and (ii), we get

r
X
di dj Im = ρ(Ki )ρ(Kj ) = ρ(Ki Kj ) = ρ( λijk Kk )
k=1
r
X r
X
= λijk ρ(Kk ) = ( λijk dk )Im .
k=1 k=1
Pr
Hence di dj = k=1 λijk dk .
(iv) By part (i) we have
X X
hi χ(gi ) = χ(g) = trace(ρ(g))
g∈Ci g∈Ci
X X
= trace( ρ(g)) = trace(ρ( g)) = trace(ρ(Ki ))
g∈Ci g∈Ci
= trace(di Im ) = mdi = di χ(1G ).

Hence di = hi χ(gi )/χ(1G ).

Corollary 5.2.16 The di ’s in Theorem 5.2.15 are algebraic integers.

Pr
Proof. By part (iii) of Theorem 5.2.15, we have di dj = k=1 λijk dk , where
λijk are non-negative integers. For a fixed j, let B be the r × r matrix (λijk )
and D be the column matrix (dk )r×1 . Then we have (dj Ir )D = BD and hence
(B − dj Ir )D = 0r×r .

76
 Since by Theorem 5.2.15 (part (iv)) we have

d1 = h1 χ(1G )/χ(1G ) = h1 = 1 6= 0,

D is a non-zero matrix. Hence B − dj Ir is a singular matrix, so that det(B −

dj Ir ) = 0. Since λijk are integers, the equation det(B − dj Ir ) = 0 produces a
polynomial equation for dj with integer coefficients and leading coefficient of
±1. Thus dj is an algebraic integer.

Note 5.2.5 If Ci is a conjugacy class of G, then Ci0 = {g ∈ G : g −1 ∈ Ci } is

also a conjugacy class of G. Obviously Ci = Ci0 if and only if g ∼ g −1 for all
g ∈ Ci .

Theorem 5.2.17 (Orthogonality relations) Let Irr(G) = {χ1 , χ2 , · · · , χk }.

Then
1 P
(i) |G| g∈G χi (g)χj (g) = δij , row-orthogonality.

Pk
(ii) s=1 χs (gi )χs (gj ) = δij 0 |CG (gj )|, column-orthogonality relation.

Proof. (i)
1 X
δij = < χi , χj >= χi (g)χj (g −1 )
|G| g∈G
1 X
= χi (g)χj (g)
|G| g∈G

by Corollary 5.2.13.
Pr
(ii) We know that Ki Kj = m=1 λijm Km . Then 1G occurs in the expansion
of Ki Kj if and only if i = j0 (that is gi is conjugate to gj−1 ). Thus λij1 = 0 if
i 6= j 0 and λij1 = hi if i = j 0 . For each 1 ≤ s ≤ k, using Theorem 5.2.15 we get

di dj = [hi χs (gi )/χs (1G )] × [hj χs (gj )/χs (1G )]

r
X
= λijm [hm χs (gm )/χs (1G )].
m=1

Thus r
X
hi hj χs (gi )χs (gj ) = λijm hm χs (1G )χs (gm ).
m=1
Therefore
k
X r
X k
X
hi h j χs (gi )χs (gj ) = [λijm hm χs (1G )χs (gm )] =
s=1 m=1 s=1

77
 k
X r
X k
X
λij1 h1 χs (1G )χs (1G ) + [λijm hm χs (1G )χs (gm )] = λij1 |G| + 0,
s=1 m=2 s=1

by Theorem 5.2.8. This show that

k
X
χs (gi )χs (gj ) = λij1 |G|/hi hj
s=1
= 0 × |G|/hi hj = 0, if i 6= j 0
= hi × |G|/hi hj = |G|/hj = |CG (gj )|, if i = j 0 .
Pk
Hence s=1 χs (gi )χs (gj ) = δij 0 |CG (gj )|.
Pk
Exercise 5.2.8 Show that s=1 χs (gi )χs (gj ) = δij |CG (gj )|.

Theorem 5.2.18 (The number of irreducible characters) The number of

irreducible characters of a group G equals the number of comjugacy classes of
G.

Proof. Let Irr(G) = {χ1 , χ2 , · · · , χk } and let r be the number of conjugacy

classes of G. Then by the Theorem 5.2.6 we have k ≤ r. Now let

S = {(χ1 (gi ), χ2 (gi ), · · · , χk (gi )) : 1 ≤ i ≤ r}.

We claim that S is a linearly independent subset of C| k . Assume that ∃λi ∈ C|

such that r
X
λi (χ1 (gi ), χ2 (gi ), · · · , χk (gi )) = 0.
i=1
Pr
Then we must have i=1 λi χs (gi ) = 0, 1 ≤ s ≤ k. So for each j we have
r
X
[ λi χs (gi )]χs (gj ) = 0, 1 ≤ s ≤ k.
i=1

Hence
k X
X r
[ λi χs (gi )]χs (gj ) = 0 f or all j.
s=1 i=1

So that
r
X k
X
λi [ χs (gi )χs (gj )] = 0 f or all j.
i=1 s=1

Now applying Theorem 5.2.17 (ii), we get

r
X
λi δij 0 |CG (gj )| = 0.
i=1

78
That is λj 0 |CG (gj )| = 0, so that λj 0 = 0 for all 1 ≤ j ≤ r. This shows that
λj = 0 for all 1 ≤ j ≤ r. Thus S is a linearly independent subset of C| k , and
hence we have
r = |S| ≤ dim(C| k ) = k.

Therefore r = k as required.

Note 5.2.6 Let ∆ be the r × r matrix (χi (gj )) = (aij ). Then ∆ is called the
character table of G. The rows are indexed by the irreducible characters of G
and the columns by the conjugacy classes of G. We take the first row and first
column to be indexed by the trivial character and 1G respectively, that is χ1 is
the trivial character and g1 = 1G . Theorem 5.2.18 shows that columns of ∆ are
linearly independent, and hence ∆ is non-singular. In particular the rows of ∆
are also linearly independent.

Exercise 5.2.9 Compute ∆−1 . [Hint: First let B = (bjl )r×r , where bjl =
1
|CG (gj )| χl (gj ). Then show that B = ∆−1 ].

Note 5.2.7 Property (ii) in Theorem 5.2.17 implies that

r
[χs (gi )]2 = 0 if gi not conjugate to gi−1
X

s=1
= |CG (gi )| otherwise.
Pr 2
In particular we have s=1 [χs (1G )] = |G|, which is the result we proved in
Theorem 5.2.8. This shows that the sum of squares of the degrees of the irre-
ducible characters of G is |G|.

Exercise 5.2.10 Let ρ be a representation of G of degree m. Define ρ∗ from

G into GL(m, C| ) by ρ ∗ (g) = [ρ(g −1 )]t , transpose of ρ(g −1 ). Then show that

(i) ρ∗ is a representation of degree m of G,

(ii) χρ∗ (g) = χρ (g), for all g ∈ G,

(iii) If ρ is irreducible, so is ρ∗,

(iv) If ρ ∼ φ, then ρ∗ ∼ φ ∗ .

Theorem 5.2.19 The degree of an irreducible representation of a finite group

G divides |G|.

79
 hk χi (gk )
Proof. By Theorem 5.2.15 (iv), χi (1G ) are algebraic integers for all k and i.
By Theorem 5.2.12, each χj (gk ) is an algebraic integer. Hence
r X
r
X χi (gk )
α= hk χj (gk )
j=1 k=1
χi (1G )

is an algebraic integer by Lemma 5.2.11. Now

r X r X
X χi (g)χj (g) X 1
α = = [χi (g)χj (g)]
j=1 g∈G
χi (1G ) χ (1 )
j=1 g∈G i G
r X
X 1
= [χi (g)χj (g −1 )]
j=1 g∈G
χi (1 G )
r X r
1 X |G| X
= [ [χi (g)χj ∗ (g)] = δij∗ = |G|/χi (1G ),
χi (1G ) j=1 g∈G χi (1G ) j=1

by Theorem 5.2.17, part (i). Hence α ∈ Q.

| Since α is an algebraic integer and

α ∈ Q,
| we must have α ∈ Z
Z . Thus χi (1G ) divides |G|.

Note 5.2.8 The integers λijk defined in the Theorem 5.2.15 (ii) are called
Class Algebra Constants. In the following corollary we will produce a for-
mula for these integers. This formula plays an important role in the application
of character theory of finite groups.

Corollary 5.2.20
r
|G| X χs (gi )χs (gj )χs (gk )
λijk = .
|CG (gi )||CG (gj )| s=1 χs (1G )
Pr
Proof. Let ρs denote the representation that affords χs . Since Ki Kj = m=1 λijm Km ,
we have r
X
ρs (Ki )ρs (Kj ) = ρs (Ki Kj ) = λijm ρs (Km ). (∗)
m=1

Now since ρs (Ki ) = di In and ρs (Kj ) = dj In , where n is the degree of ρs (see

Theorem 5.2.15), we have

χs (gi ) χs (gj )
ρs (Ki ) = hi In and ρs (Kj ) = hj In ,
χs (1G ) χs (1G )

by Theorem 5.2.15. Now by using the relation (*) above, we get

r
χs (gi ) χs (gj ) X χs (gm )
hi × hj = λijm hm .
χs (1G ) χs (1G ) m=1 χs (1G )

80
Hence r
X
λijm hm χs (gm ) = hi hj χs (gi )χs (gj )/χs (1G ), (1)
m=1

multiplying by both sides of (1) by χs (gk ) and summing from s = 1 to s = r

we obtain
r r r
X X X χs (gi )χs (gj )χs (gk )
λijm hm [ χs (gm )χs (gk )] = hi hj . (2)
m=1 s=1 s=1
χs (1G )

Since r
X
χs (gm )χs (gk ) = δkm |CG (gk )|
s=1

by Exercise 5.2.7, we have

r r
X X χs (gi )χs (gj )χs (gk )
λijm hm δkm |CG (gk )| = hi hj .
m=1 s=1
χs (1G )

So that
r
X χs (gi )χs (gj )χs (gk )
λijk hk |CG (gk )| = hi hj .
s=1
χs (1G )
Thus
r
|G||G| X χs (gi )χs (gj )χs (gk )
λijk |G| = .
|CG (gi )||CG (gj )| s=1 χs (1G )
This gives the desired formula for λijm .

Example 5.2.3 (i) If g 2 = 1G , then χi (g) ∈ ZZ for all χi ∈ Irr(G) : Because

g 2 = 1G implies that g = g −1 . If g = 1G , then χi (g) = χi (1G ) = deg(χi ) ∈ ZZ .
If g is not the identity, then o(g) = 2 and hence χi (g) is a sum of 2nd roots of
unity. Since the roots are 1 and -1, clearly χi (g) ∈ ZZ .
(ii) If g is conjugate to g −1 , then χi (g) ∈ |R for all χi ∈ Irr(G) : Because g
conjugate to g −1 implies that χi (g) = χi (g −1 ) = χi (g). Thus χi (g) ∈ |R .
(iii) Assume that g ∈ G is an element of order three and g ∼ g −1 . Then
χi (g) ∈ ZZ for all χi ∈ Irr(G) : Because g ∼ g −1 implies that χi (g) ∈ |R
for all χi ∈ Irr(G), by part (ii). Let χi (g) = ε1 + ε2 + · · · + εm , where i ∈
√ √
3 3
{1, − 12 − 1
2 i, − 2 + 2 i}. Assume that for some j, 1 ≤ j ≤ m, we have
√ √
1 3 1 3
j = − − i or j = − + i.
2 2 2 2
Then since χi (g) ∈ |R , j must also appear in χi (g). Now since j + j = −1,
we deduce that χi (g) ∈ ZZ .

81
Example 5.2.4 (Character Table of S4 ) In S4 there are 5 conjugacy classes
and hence Irr(S4 ) = 5. Consider the map ρ2 : S4 −→ C| given by ρ2 (α) = 1 if
α is even and ρ2 (α) = −1 if α is odd. Then ρ2 is a representation of degree 1
and hence ρ2 = χρ2 . let denote this character by χ2 . Then we have

χ2 (1S4 ) = χ2 ((1 2)(3 4)) = χ2 ((1 2 3)) = 1

and
χ2 ((1 2)) = χ2 ((1 2 3 4)) = −1.

So we have the following table:

Classes of S4 1S4 (1 2)(3 4) (1 2) (1 2 3 4) (1 2 3)

hi 1 3 6 6 8
χ2 1 1 −1 −1 1

Since
1
< χ2 , χ2 > = [1 + 3(1)(1) + 6(−1)(−1) + 6(−1)(−1) + 8(1)(1)]
24
1
= [1 + 3 + 6 + 6 + 8] = 24/24 = 1,
24
χ2 is irreducible.
Now let π : S4 −→ GL(4, C| ) be the natural permutation representation of
S4 . Then χπ (g) is equal to the number of fixed points of g on the set {1, 2, 3, 4}.
Then we have

Classes of S4 1S4 (1 2)(3 4) (1 2) (1 2 3 4) (1 2 3)

hi 1 3 6 6 8
χπ 4 0 2 0 1

It is not difficult to see that < χπ , χπ >= 2 and < χπ , χ1 >= 1, where χ1 is the
trivial charascter. hence χπ = χ1 + χ3 , where χ3 is an irreducible character of
degree 4 − 1 = 3. Then we have χ3 (g) = χπ (g) − χ1 (g), for all g in S4 .
Now it remains to find two more irreducible characters of S4 , namely χ4
P5 2
and χ5 . Since i=1 [χi (1S4 )] = |G| = 24, we have

[χ4 (1S4 )]2 + [χ5 (1S4 )]2 = 24 − (1 + 1 + 9) = 24 − 11 = 13 = 4 + 9.

This implies that we can assume deg(χ4 ) = 2 and deg(χ4 ) = 3. So far we have

82
the following information for the character table of S4 :

Classes of S4 1S4 (1 2)(3 4) (1 2) (1 2 3 4) (1 2 3)

hi 1 3 6 6 8
χ1 1 1 1 1 1
χ2 1 1 −1 −1 1
χ3 3 −1 1 −1 0
χ4 2 a b c d
χ5 3 e f g h

We are able to complete the character table by means of the orthogo-

nality relations. First notice that, since g ∼ g −1 for all g ∈ S4 , we have
{a, b, c, d, e, f, g, h} ⊆ |R . Using Example 5.2.3, parts (i) and (ii), we have
{a, e, b, f, d, h} ⊆ ZZ . Now using the orthogonality of the first two columns
we get
1 + 1 − 3 + 2a + 3e = 0,

so that
2a + 3e = 1. (1)
P5
Since i=1 [χi ((1 2)(3 4))]2 = |CS4 ((1 2)(3 4))|, by Note 5.2.7, we have
24
1 + 1 + 1 + a2 + e2 = = 8,
3
and hence
a2 + e2 = 5. (2)

Using relations (1) and (2) we obtain a = 2 and e = −1.

Similarly the orthogonality of the first column with columns three and five
give
1 − 1 + 3 + 2b + 3f = 0, 1 + 1 + 0 + 2d + 3h = 0.

We deduce that
2b + 3f = −3 (3)

and
2d + 3h = −2. (4)

Since
5
X 24
[χi ((1 2))]2 = |CS4 ((1 2))| = =4
i=1
6
and
5
X 24
[χi ((1 2 3))]2 = |CS4 ((1 2 3))| = = 3,
i=1
8

83
we get
b2 + f 2 = 4 − 3 = 1 (5)

and
d2 + h2 = 3 − 2 = 1. (6)

Now relations (3) and (5) imply that b = 0 and f = −1. Similarly relations (4)
and (6) imply that d = −1 and h = 0. At this stage we produce the following
information on the character table of S4 :

Classes of S4 1S4 (1 2)(3 4) (1 2) (1 2 3 4) (1 2 3)

hi 1 3 6 6 8
χ1 1 1 1 1 1
χ2 1 1 −1 −1 1
χ3 3 −1 1 −1 0
χ4 2 2 0 c −1
χ5 3 −1 −1 g 0

Using the orthogonality of columns 3 and 4 we obtain

1+1−1+0×c−g =0

and hence g = 1. Now the orthogonality of columns 4 and 5 gives

1 − 1 + (−1) × 0 + c(−1) + 1 × 0 = 0,

and hence c = 0. This completes the character table of S4 :

Classes of S4 1S4 (1 2)(3 4) (1 2) (1 2 3 4) (1 2 3)

hi 1 3 6 6 8
χ1 1 1 1 1 1
χ2 1 1 −1 −1 1
χ3 3 −1 1 −1 0
χ4 2 2 0 0 −1
χ5 3 −1 −1 1 0

Exercise 5.2.11 (Characters of cyclic groups) . Let G = hxi be a cyclic

group of order n. Let e2kπi/n be the nth roots of unity in C| , k = 0, 1, 2, ..., n − 1.
Define ρk : G → C| ∗ by ρk (xm ) = [e2kπi/n ]m . Show that ρk define the n distinct
irreducible representations of G.

Exercise 5.2.12 Use the above exercise to construct the character table of the
cyclic groups of order 2, 3, 4. 5 and 6.

84
Exercise 5.2.13 Calculate the character table of the |V 4 , the Klien 4-group.

Note 5.2.9 If G is an abelian group, then all irreducible representations of G

are of degree 1. In general, we would like to know how many of the irreducible
representations of an arbitrary group G are of degree 1. In the following theorem
we give the answer to this question.

Theorem 5.2.21 Let G be a finite group. The number of representations of G

of degree 1 is equal to [G : G0 ].

Proof. If ρ is a representation of degree one of G, then by Exercise 5.2.3(i)

we have Ker(ρ) ⊇ G0 . Now the define φ : G/G0 → C| ∗ by φ(gG0 ) = ρ(g), for
all g ∈ G. Then φ defines a representation of degree one for the group G/G0
(note that since G0 ⊆ Ker(ρ), φ is well-defined.) Since G/G0 is abelian, it
has [G:G0 ] conjugacy classes. Hence the group G/G0 has [G:G0 ] irreducible
characters. Since G/G0 is abelian, all its irreducible characters are of degree
one (see Exercise 5.2.3, part (ii).) Now consider the natural homomorphism
π : G → G/G0 . If ψ is a representation of G/G0 of degree one, then ψ ◦ π is a
representation of degree one of G. Now it is not difficult to see that we have a
one-to-one correspondence between the set of all representations of degree one
of G and the set of all the irreducible representations of the group G/G0 .

Exercise 5.2.14 Compute the character table of A4 .

Exercise 5.2.15 Calculate the character tables of Q (the quaternion group of

order 8) and D8 . Show that they have same character tables.

Exercise 5.2.16 Calculate the character table of A5 . Recall that A5 is a non-

abelain simple group of order 60 and hence (A5 )0 = A5.

Note 5.2.10 In the Exercises 5.1.1 and 5.1.2 we observed that there is a one-
to-one correspondence between representations of G/N and representations of
G with kernel containing N. Furthermore it is not difficult to show that, under
this correspondence, irreducible representations correspond to irreducible rep-
resentations. We put this result in terms of characters in the following theorem.
If χ is a character afforded by a representation ρ of G, we define ker(χ) to be
ker(ρ).

Theorem 5.2.22 Let N  G.

(i) If χ is a character of G and N  ker(χ), then χ is constant on cosets of

N in G and χ
b on G/N defined by χ(gN
b ) = χ(g) is a character of G/N.

85
 (ii) If χ
b is a character of G/N, then the function χ defined by χ(g) = χ(gN
b )
is a character of G.

(iii) In both (i) and (ii), χ ∈ Irr(G) iff χ

b ∈ Irr(G/N ).

Proof. (i) and (ii) follow from Exercise 5.1.1 and 5.1.2.
(iii) Let S be a set of coset representatives of N in G. Then we have
1 X
1 = hχ, χi = χ(g) · χ(g −1 )
|G| g∈G
1 X
= |N | · χ(g) · χ(g −1 )
|G| g∈S
1 X
= |N | · χ(gN
b b −1 N )
) · χ(g
|G| g∈S
1
)−1
X
= |N | · χ(gN
b ) · χ(gN
b
|G| gN ∈G/N
1
)−1 = hχ,
X
= χ(gN
b ) · χ(gN
b b χi
b .
|G/N | gN ∈G/N

Example 5.2.5 Consider the group G = S4 . Let N = {e, (1 2)(3 4), (1 3)(2 4), (1 4)(2 3)}
G. If Ci = [gi ] is a class of G, then C
ci = [gi N ] is a class of G/N. However, dis-

tinct classes in G may produce equal classes in G/N. Referring to the character
table of S4 (See Example 5.2.4), we see that

{χ| χ ∈ Irr(G), N ⊆ ker(χ)} = {χ1 , χ2 , χ4 }.

Hence Irr(G/N ) = {χ
b1 , χ b4 }. Using the character table of S4 we have
b2 , χ

Classes of S4 1S4 (1 2)(3 4) (1 2) (1 2 3 4) (1 2 3)

χ1 1 1 1 1 1
χ2 1 1 −1 −1 1
χ2 2 2 0 0 −1

Observe that columns 1 and 2 are identical, as are columns 3 and 4. Deleting
repeats, we obtain the character table of G/N

Classes of G/N N (1 2)N (1 2 3)N

χ
b1 1 1 1
χ
b2 1 −1 1
χ
b4 2 0 −1

86
we can see that G/N is a group of order 6 and it is not abelian. Hence G/N ∼
=
S3 . (See the character table of S3 in the Example 5.2.1)

Note 5.2.11 If N  G, then the character table of G determines whether or

not G/N is abelian. There is no way to determine from the character table of
G whether or not N is abelian.

Corollary 5.2.23 Let g ∈ G and N  G. Then |CG (g)| ≥ |CG/N (gN )|.

Proof. We know that

Irr(G/N ) = {χ|
b χ ∈ Irr(G), N ⊆ ker(χ)}.

)−1
X
|CG/N (gN )| = χ(gN
b ) · χ(gN
b
b∈Irr(G/N )
χ
X X
= χ(gN
b ) · χ(gN
b )= |χ(gN
b )|2
b∈Irr(G/N )
χ b∈Irr(G/N )
χ
X
2
= {|χ(g)| | χ ∈ Irr(G), N ⊆ ker(χ)}
X
≤ |χ(g)|2 = |CG (g)|.
χ∈Irr(G)

5.3 Tensor Products and Products of Characters

We have seen in Section 5.1 that the sum of any two characters of a group G
is again a character. Our aim in this section is to show that if χ and ψ are
characters of a group G, then the product χψ defined by

(χψ)(g) = χ(g) · ψ(g), ∀g ∈ G

is a character of G. It is clear that if χ and ψ are class functions on G, then so

is χψ.

Definition 5.3.1 (Tensor Product of Matrices) [Kronecker Product] Let

P = (pij )m×m and Q = (qij )n×n be two matrices. Define the mn × mn matrix
P ⊗ Q by
 
p11 Q p12 Q ··· p1m Q
 

 p21 Q p22 Q ··· p2m Q 

P ⊗ Q := ((pij )Q) =  .. .. .. ..
.
.
 

 . . . 

pm1 Q pm2 Q · · · pmm Q

87
 Then we have

trace(P ⊗ Q) = p11 trace(Q) + p22 trace(Q) + · · · + pmm trace(Q)

= trace(P ) · trace(Q).

Exercise 5.3.1 Show that

0 0 0 0
(P ⊗ Q) · (P ⊗ Q ) = (P P ) ⊗ (QQ ).

Definition 5.3.2 Let T and U be two representations of G. We define the

tensor product T ⊗ U by

(T ⊗ U )(g) := T (g) ⊗ U (g),

where ⊗ on the RHS is defined by Definition 5.3.1.

Theorem 5.3.1 Let T and U be two representations of G. Then

(i) T ⊗ U is a representation of G,

(ii) χT ⊗U = χT · χU .

Proof.

(i) ∀g, h ∈ G, we have

(T ⊗ U )(gh) = T (gh) ⊗ U (gh)

= (T (g) · T (h)) ⊗ (U (g) · U (h))
= (T (g) ⊗ U (g)) · (T (h) ⊗ U (h)), by the Exercise 5.3.1
= (T ⊗ U )(g) · (T ⊗ U )(h), by the Definition 5.3.2.

(ii)

χT ⊗U (g) = trace((T ⊗ U )(g))

= trace(T (g) ⊗ U (g))
= trace(T (g)) · trace(U (g))
= χT (g) · χU (g)
= (χT · χU )(g).

Hence χT ⊗U = χT · χU .

88
 Let G = H × K be the direct product of H and K. Let T : H −→ GL(m, C)
|

and U : K −→ GL(n, C| ) be representations of H and K respectively. Define

the direct product T ⊗ U as follows: Let g ∈ G. Then g can be written uniquely
in the form hk, h ∈ H, k ∈ K. Define

(T ⊗ U )(g) := T (h) ⊗ U (k),

where ⊗ on the RHS is the tensor product given in Definition 5.3.1. Then T ⊗U
is a representation of degree mn of G = H × K. Furthermore

χT ⊗U (g) = χT (h) · χU (k), where g = hk.

The following theorem assures us that all characters of G can be constructed

in this way from characters of H and K.

Theorem 5.3.2 Let G = H × K be the direct product of the groups H and K.

Then the direct product of any irreducible character of H and any irreducible
character of K is an irreducible character of G. Moreover, every irreducible
character of G can be constructed in this way.

Proof. Let χT ∈ Irr(H) and χU ∈ Irr(K). Let χ = χT ⊗U = χT · χU . then χ

is a character of G. We claim χ ∈ Irr(G). Let g ∈ G, then ∃! h ∈ H, k ∈ K
such that g = hk. So
X X X
|χ(g)|2 = |χT (h) · χU (k)|2
g∈G h∈H k∈K
X X
= |χT (h)|2 · |χU (k)|2
h∈H k∈K
X X
= |χT (h)|2 · |χU (k)|2
h∈H k∈K
= |H| · |K| = |G|.
X
Hence 1
|G| |χ(g)|2 = 1; so hχ, χi = 1. Thus χ ∈ Irr(G).
g∈G

Assume that |Irr(H)| = r and |Irr(K)| = s, then we obtain rs irreducible

0 0
characters of G = H × K in this way. If g1 = hk and g2 = h k are two elements
0 0
in G, then g1 ∼ g2 if and only if h ∼ h in H and k ∼ k in K. Then

# ofconjugacy classes of G = # of conjugacy classes of H

× # of conjugacy classes of K
= rs.

89
Hence |Irr(G)| = rs.

Exercise 5.3.2 Use Theorem 5.3.2 to construct the character tables of

(i) |V 4 ∼
= ZZ 2 × ZZ 2 ,

(ii) G ∼
= S3 × ZZ 2 .

Definition 5.3.3 (Powers of Characters) Let χ be a character of G. For

| ∪ {0} define χn by
n∈N

χn (g) := (χ(g))n , ∀g ∈ G.

Then χ0 = χ1 (the trivial character of G) and by Theorem 5.3.1 (using induction

on n) we can show that χn is a character of G. If χ is a faithful character of G
(that is ker(ψ) = {1G }), then the powers of χ can produce a lot of information
about the character table of G.

Lemma 5.3.3 (Vandermonde Matrix) If α1 , α2 , · · · , αr are distinct com-

plex numbers, then the matrix
 
1 α1 α12 · · · α1r−1
 

 1 α2 α22 · · · α2r−1 

A= .. .. .. .. ..

.
 

 . . . . 

1 αr αr2 · · · αrr−1

is invertible.

Proof. Suppose that x1 , x2 , · · · , xr are indeterminates. Consider

 
1 x1 x21 · · · xr−1
1
 

 1 x2 x22 · · · xr−1
2


∆ = det  .. .. .. .. ..
.
.
 

 . . . . 

1 xr x2r · · · xr−1
r

Y
We can show that ∆ = ± (xi − xj ), where
i<j

Y
(xi − xj ) = (x1 − x2 ) · (x1 − x3 ) · · · (x1 − xr ) ·
i<j
(x2 − x3 ) · (x2 − x4 ) · · · (x2 − xr )
···
(xr−1 − xr ).

90
Since αi , are distinct
Y
det(A) = ± (αi − αj ) 6= 0.
i<j

Hence A is invertible.

Theorem 5.3.4 Let χ be a faithful character of G. Suppose that r = |{χ(g)| g ∈

G}|. Then every irreducible character of G is a constituent of one of the powers
χ0 , χ1 , · · · , χr−1 .

Proof. Let {χ(g)| g ∈ G} = {α1 , α2 , · · · , αr }, where αi , are distinct. For

i = 1, 2, · · · , r define

Gi = {g ∈ G| χ(g) = αi }.

Take α1 = χ(1G ). Then G1 = ker(χ). Since χ is faithful, we have G1 = {1G }.

Let ψ ∈ Irr(G). We claim ψ, χj 6= 0 for some j with 0 ≤ j ≤ r − 1. Let
X
βi = ψ(g), 1 ≤ i ≤ r.
g∈Gi

Then
X
β1 = ψ(g) = ψ(g) = ψ(1G ) = deg(ψ) 6= 0.
g∈G1

Now for j ≥ 0 we have

D E 1 X j
ψ, χj = (χ )(g) · ψ(g)
|G| g∈G
1 X
= [χ(g)]j · ψ(g)
|G| g∈G
r X
1 X
= [χ(g)]j · ψ(g)
|G| i=1 g∈G
i
r
1 X
= (αi )j βi . (5.1)
|G| i=1

Let A = (aij )r×r , where aij = (αi )j−1 . Let b = (β1 , β2 , · · · , βr )1×r . Then A is
invertible by Lemma 5.3.3. Since β1 6= 0, b 6= 0. Hence bA 6= 0. Now by 5.1 we
have
r
X D E
(j + 1)th entry of bA = βi (αi )j = |G| χj , ψ .
i=1

Thus ∃ j, 0 ≤ j ≤ r − 1 such that χj , ψ 6= 0.

91
Example 5.3.1 (i) Let G be a group, |G| =
6 1. Let χπ be the character of the
right regular representation of G. Then χπ (1G ) = |G| and χπ (g) = 0 for
all g ∈ G \ {1G }. Hence χπ is faithful and {χπ (g)| g ∈ G} = {|G|, 0}. Thus
r = 2 and every irreducible character of G is a constituent of χ0π = χ1 or
χπ . [See Theorem 5.2.8].

(ii) Let G = S4 (See Example 5.2.4). Let χ = χ3 . Then

r = |{χ3 (g)| g ∈ S4 }| = |{3, 1, 0, −1}| = 4

and ker(χ3 ) = {1G }. Now χ2 = χ1 +χ3 +χ4 +χ5 (check) and χ3 , χ2 = 1.

(check). Hence every irreducible character of S4 is a constituent of χ2 or
χ3 .

Exercise 5.3.3 In the Example 5.3.1(ii), find χ3 in terms of irreducible char-

acters of S4 .

Exercise 5.3.4 Let χ, ψ and φ be characters of a group G. Show that hχψ, φi =

D E
χ, ψφ = hψ, χφi .

Exercise 5.3.5 If χ is a character of a group G, it can be shown that χ2 =

χS + χA , where
1 2 
χS = χ (g) + χ(g 2 )
2
1 2 
χA = χ (g) − χ(g 2 )
2
for all g ∈ G. [See James & Liebeck]. Let G = S4 (See Example 5.2.4 and
5.3.1). If χ = χ3 , calculate the values of χS and χA . Express χS and χA as a
linear combinations of the irreducible characters χ1 , χ2 , · · · , χ5 of S4 .

5.4 Restriction to a Subgroup

Let G be a group and H ≤ G. If ρ : G −→ GL(n, C)
| is a representation of
G, then ρ↓H : H −→ GL(n, C)
| given by (ρ↓H)(h) = ρ(h), ∀h ∈ H, is a
representation of H. We say that ρ↓H is the restriction of ρ to H. If χρ is the
character of ρ, then χρ ↓H is the character of ρ↓H. We refer to χρ ↓H as the
restriction of χρ to H.

Remark 5.4.1 It is clear that deg(ρ) = deg(ρ↓H). However, ρ irreducible

does not imply (in general) that ρ↓H is irreducible. For example, let G = D8 =
a, b| a4 = b2 = 1G , b−1 ab = a−1 . Let χ ∈ Irr(G) be given by

92
 D4 1 a a2 b ab
hi 1 2 1 2 2
χ 2 0 −2 0 0

Let H = hai = {1, a, a2 , a3 } ∼

= ZZ 4 . Then χ↓H is given by

H 1 a a2 a3
hi 1 1 1 1
χ↓H 2 0 −2 0

(Note: a ∼ a3 in G).
Since H is abelian and deg(χ↓H) = 2, χ↓H is reducible.

Theorem 5.4.1 Let H ≤ G. Let ψ be a character of H. Then there is an

irreducible character χ of G such that hχ↓H, ψiH 6= 0.

Proof. Let Irr(G) = {χ1 , χ2 , · · · , χk }. Let π be the right regular representation

of G. Then

 |G| if g = 1 ,
G
χπ (g) =
 0 if g 6= 1G ,

k
X
and χπ = χi (1G ) · χi . [See Theorem 5.2,8].
i=1
Now
1
hχπ ↓H, ψiH = (χπ ↓H)(1G ) · ψ(1G )
|H|
1
= · |G| · ψ(1G )
|H|
|G|
= · ψ(1G ),
|H|

and
k
X
hχπ ↓H, ψiH = χi (1G ) hχi ↓H, ψiH
i=1

imply that
k
X |G|
χi (1G ) hχi ↓H, ψiH = · ψ(1G ) 6= 0.
i=1
|H|

Thus hχi ↓H, ψiH 6= 0 for some 1 ≤ i ≤ k.

93
Theorem 5.4.2 Let H ≤ G. Let χ ∈ Irr(G) and let Irr(H) = {ψ1 , ψ2 , · · · , ψr }.
r
X
Then χ↓H = di ψi , where di ∈ N ˙
| ∪{0} and
i=1
r
X
d2i ≤ [G : H]. (5.2)
i=1

Moreover, we have equality in 5.2 if and only if χ(g) = 0, ∀g ∈ G \ H.

Proof. We have
r
X 1 X
d2i = hχ↓H, χ↓Hi = χ(h) · χ(h).
i=1
|H| h∈H

Since χ is irreducible,
1 X
1 = hχ, χiG = χ(g) · χ(g)
|G| g∈G
1 X 1 X
= χ(g) · χ(g) + χ(g) · χ(g)
|G| g∈H |G| g∈G\H
r
|H| X
= d2 + K,
|G| i=1 i
X X
where K = 1
|G| χ(g)χ(g). Since K = 1
|G| |χ(g)|2 , K ≥ 0. Thus
g∈G\H g∈G\H

r
|H| X
d2 = 1 − K ≤ 1,
|G| i=1 i
so
r
X |G|
d2i ≤ = [G : H].
i=1
|H|

Also K = 0 if and only if |χ(g)|2 = 0 for all g ∈ G \ H. Hence K = 0 if and

only if χ(g) = 0, ∀g ∈ G \ H.

Exercise 5.4.1 Let G = S4 , H = h(1 2 3 4), (1 3)i . Then H ≤ S4 . Show

that H ∼
= D8 . For each irreducible character χ of G express χ↓H in terms of
irreducible characters of H.

Exercise 5.4.2 Use the restriction of the irreducible characters of S4 to find

the character table of A4 .

Exercise 5.4.3 Let G be a group and let H be an abelian subgroup of G of

index n. Prove that deg(χ) ≤ n for every irreducible character χ of G.

94
Proposition 5.4.3 (An Application: Character Table of D2n ) Let D2n =<
a, b|an = b2 = 1, bab = a−1 . Then
0 0
(i) (D2n ) =< a >∼
= ZZ n , if n is odd, and (D2n ) =< a2 >∼
= ZZ n/2 , if n is
even.

(ii) If χ ∈ Irr(D2n ), then deg(χ) ∈ {1, 2}.

(iii) |{χ ∈ Irr(D2n ) : deg(χ) = 1}| = 2 or 4 if n is odd or even respectively.

(iv) |{χ ∈ Irr(D2n ) : deg(χ) = 2}| = (n − 1)/2 or (n − 2)/2 if n is odd or

even respectively.

(v) The number of conjugacy classes of D2n = (n + 3)/2 or (n + 6)/2 if n is

odd or even respectively.

Proof.

(i) Let G = D2n and A =< a >∼

= ZZ n . Then A  G and [G : A] = 2.
0
Hence G/A is abelian, so that G ≤ A. Now let H =< a2 >. Then since
0 0
aba−1 b−1 = a2 , we have a2 ∈ G . Thus H ≤ G ≤ A. Now

|H| = o(a2 ) = o(a)/gcd(2, o(a)) = n/gcd(2, n)

= n if n is odd

= n/2 if n is even.

Hence H = A ∼ = ZZ n if n is odd and H ∼ = ZZ n/2 if n is even. Thus

0 0
G =A∼ = ZZ n if n is odd and G = H ∼= ZZ n/2 if n is even.

(ii) Since A is abelian, for any χ ∈ Irr(G) we have deg(χ) ≤ [G : A], by

Exercise 5.4.3. So that for any χ ∈ Irr(G) we have deg(χ) ≤ 2.

(iii)
0
|{χ ∈ Irr(D2n ) : deg(χ) = 1}| = [G : G ] = [G : A] = 2 if n is odd

= [G : H] = 2n/(n/2) = 4 if n is even.

(iv) Let l = |{χ ∈ Irr(D2n ) : deg(χ) = 1}| and k = |{χ ∈ Irr(D2n ) : deg(χ) =
2}|. Then l ∈ {2, 4} and l + k = r = the number of conjugacy classes of
G. We have the following two cases:

95
 Case 1: l = 2, that is n is odd. In this case
r
X
|G| = 2n = [χi (1G )]2 = 1 + 1 + 4k = 2 + 4k,
i=1

which implies that k = (n − 1)/2.

Case 2: l = 4, that is n is even. In this case
r
X
|G| = 2n = [χi (1G )]2 = 1 + 1 + 1 + 1 + 4k = 4 + 4k,
i=1

which implies that n = 2 + 2k and k = (n − 2)/2.

(v) If n is odd, then

r = l + k = 2 + (n − 1)/2 = (n + 3)/2.

If n is even
r = l + k = 4 + (n − 2)/2 = (n + 6)/2.

Exercise 5.4.4 Use the above result (Proposition 5.4.3) to compute the char-
acter tables of D8 and D10 .

Exercise 5.4.5 Use similar arguments in the above proposition (Proposition

5.4.3), to show that if G is a non-abelian group of order p3 , then |Irr(G)| =
p2 + p − 1.

5.5 Induced Representations

Definition 5.5.1 (Transversal) Let H ≤ G. By a right transversal of H in
G we mean a set of representatives for the right cosets of H in G.

Theorem 5.5.1 Let H ≤ G and T be a representation of H of degree n. Extend

T to G by T 0 (g) = T (g) if g ∈ H and T 0 (g) = 0 if g 6∈ H. Let {x1 , x2 , · · · , xr }
be a right transversal of H in G. Define T ↑G by
 
T 0 (x1 gx−1
1 ) T 0 (x1 gx−1
2 ) · · · T 0 (x1 gx−1
r )
 
−1
 0
 T (x2 gx1 ) T 0 (x2 gx−1
2 ) · · · T 0 (x2 gx−1
r )


T ↑G(g) = 

.. .. .

 . ··· . .. 

 
T 0 (xr gx−1
1 ) T 0 (xr gx−1 0 −1
2 ) · · · T (xr gxr )
 r
= T 0 (xi gx−1
i ) , ∀g ∈ G.
i,j=1

Then T ↑G is a representation of G of degree nr.

96
Proof. Since each T 0 (xi gx−1
i ) is a sub matrix of degree n, (T ↑G)(g) is a nr×nr
matrix. We need to show that

(T ↑G)(gh) = (T ↑G)(g) · (T ↑G)(h), ∀g, h ∈ G.

This is equivalent to showing that [by comparing blocks on both sides] for all
fixed i, j ∈ {1, 2, · · · , r}

r
(xi ghx−1 T 0 (xi gx−1 −1
X
0 0
T j ) = k ) · T (xk hxj ). (5.3)
k=1

Case I: If xi ghx−1
j 6∈ H, then the left-hand side of (5.3) is zero. But in this
case we must have xi gx−1 −1
k 6∈ H or xk hxj 6∈ H, for each k ∈ {1, 2, · · · , r}. So
the right hand side of (5.3) is also zero. [Note: If xi gx−1 −1
k , xk hxj ∈ H, then
xi ghx−1 −1 −1
j = (xi gxk ) · (xk hxj ) ∈ H.]

Case II: Assume that v = xi ghx−1

j ∈ H. The element xi g belongs to exactly
one right coset, say xi g ∈ Hxs for some s ∈ {1, 2, · · · , r}. So u = xi gx−1
s ∈ H. If
k 6= s, then xi gx−1
k 6∈ H. thus the right hand side of (5.3) reduces to one term
with k = s. Thus (5.3) reduces to

−1
T (v) = T 0 (v) = T 0 (xi gx−1 0
s ) · T (xs hxj )

= T 0 (u) · T 0 (u−1 v)
= T (u) · T (u−1 v), since u, v ∈ H
= T (uu−1 v) = T (v), since T is a representationof H

which is true.

Definition 5.5.2 The representation T ↑G defined above is said to be induced

from the representation T of H. Let φ be the character afforded by T. Then the
character afforded by T ↑G is called induced character from φ and is denoted
by φG .

If we extend φ to G by φ0 (g) = φ(g) if g ∈ H and φ0 (g) = 0 if g 6∈ H, then

r
trace(T 0 (xi gx−1
X
φG (g) = trace((T ↑G)(g)) = i ))
i=1
r
φ0 ((xi gx−1
X
= i )).
i=1

|G|
Notice that φG (1G ) = rn = |H| φ(1).

97
Proposition 5.5.2 If φ is a character of H ≤ G, then φG is independent of
the choice of transversal.

Proof. Let S1 = {t1 , t2 , · · · , tr } and S2 = {s1 , s2 , · · · , sr } be two right transver-

sals of H in G. Suppose that
r
φ0 (ti gt−1
X
φG (g) = i ).
i=1

Without loss of generality we may assume that si ∈ Hti for i = 1, 2, · · · , r.

Then si = hi ti , where hi ∈ H for i = 1, 2, · · · , r. Now si gi s−1 −1 −1
i = hi ti gti hi ; so
si gs−1 −1
i ∈ H if and only if ti gti ∈ H. Thus

φ0 (si gs−1 0 −1
i ) = φ (ti gti ).

Hence
r r
φ0 (ti gt−1 φ0 (si gs−1
X X
φG (g) = i )= i ).
i=1 i=1

Proposition 5.5.3 The values of the induced character φG are given by

1 X 0
φG (g) = φ (xgx−1 ), ∀g ∈ G.
|H| x∈G

Proof. We know
r
φ0 (xi gx−1
X
φG (g) = i ), ∀g ∈ G,
i=1

where {x1 , x2 , · · · , xr } is a right transversal of H in G. By the Proposition 5.5.2

we also have
r
φ0 (hxi gx−1 −1
X
φG (g) = i h ), ∀h ∈ H.
i=1

Hence
r r
φ0 (hxi gx−1 −1
X XX
G
φ (g) = i h ).
h∈H h∈H i=1

φ0 (xgx−1 ).
X
So |H|φG (g) =
x∈G

Proposition 5.5.4 Let H ≤ G. Assume that φ is a character of H and g ∈ G.

Let [g] denote the conjugacy class of G containing g.

98
 (i) If H ∩ [g] = Φ, then φG (g) = 0,
m
X φ(xi )
(ii) If H ∩ [g] 6= Φ, then φG (g) = |CG (g)| ,
i=1
|CH (xi )|

where x1 , x2 , · · · , xm are representatives of conjugacy classes of H that fuse to

[g]. (That is H ∩[g] breaks up into m conjugacy classes of H with representatives
x1 , x2 , · · · , xm .)

Proof. By Proposition 5.5.3 we have

1 X 0
φG (g) = φ (xgx−1 ).
|H| x∈G

If H ∩ [g] = Φ, then xgx−1 6∈ H for all x ∈ G, so φ0 (xgx−1 ) = 0, ∀x ∈ G and

φG (g) = 0.

Now assume that H ∩ [g] 6= Φ. As x ranges over G, xgx−1 covers [g] exactly
|CG (g)| times, so
1 X
φG (g) = × |CG (g)| φ0 (y)
|H| y∈[g]
|CG (g)| X
= φ(y)
|H| y∈[g]∩H
m
|CG (g)| X
= [H : CH (xi )]φ(xi )
|H| i=1
m
X φ(xi )
= |CG (g)| .
i=1
CH (xi )

Example 5.5.1 Let G = S4 and H = A4 . The fusion map A4 −→ S4 is given

by

1S 4 −→ 1S4
[(1 2)(3 4)] −→ [(1 2)(3 4)]
[(1 2 3)] −→ [(1 2 3)]
[(1 3 2)] −→ [(1 2 3)].

H = A4 1S 4 (1 2)(3 4) (1 2 3) (1 3 2)
|CH (x)| 12 4 3 3
|CG (x)| 24 8 3 3

Let φ be a character of H. Then by Proposition 5.5.4 we have

99
 φ(1S4 )
φG (1S4 ) = 24 × 12 = 2φ(1S4 ),
φ((1 2)(3 4))
φG ((1 2)(3 4)) = 8× 4 = 2φ((1 2)(3 4)),
h i
φ((1 2 3)) φ((1 3 2))
φG ((1 2 3)) = 3× 3 + 3 ,
= φ((1 2 3)) + φ((1 3 2)),
φG ((1 2 3 4)) = φG ((1 2)) = 0, since (1 2 3 4), (1 2) 6∈ H.

Let Irr(A4 ) = {φ1 , φ2 , φ3 , φ4 } (See Exercise 5.2.14). Then we have the following
table for the values of φG
i on the classes of G = S4

G 1S 4 (1 2)(3 4) (1 2) (1 2 3 4) (1 2 3)
φG
1 2 2 0 0 0
φG
2 6 −2 0 0 0
φG
3 2 2 0 0 −1
φG
4 2 2 0 0 −1

Notice that, as sum of irreducible characters χi of S4 , we have

φG
1 = χ1 + χ2 ,
φG
2 = χ3 + χ5 ,
φG
3 = φG
4 = χ4 .

Exercise 5.5.1 Let G = S4 and H = h(1 2 3)i ∼

= ZZ 3 .

(i) If Irr(G) = {χ1 , χ2 , · · · , χ5 }, find χi ↓H as sums of irreducible characters

of ψ1 , ψ2 , ψ3 of ZZ 3 .

(ii) Calculate the induced characters ψiG , i = 1, 2, 3, as sums of the irreducible

characters χi of G.

Exercise 5.5.2 Let H = S4 with Irr(H) = {ψ1 , ψ2 , · · · , ψ5 }. Find ψiS5 , i =

1, 2, · · · , 5, by considering S4 as a subgroup of S5 . Starting with the trivial
character χ1 and alternating character χ2 of S5 , use induced characters ψiS5 , i =
1, 2, · · · , 5, to complete the character table of S5 .

100
Index

action on a set, 6

Cayley’s Theorem, 2
conjugate, 5
conjugate subgroups, 5
core of H, 4

general linear group, 2

Generalized Cayley Theorem, 3

left regular representation, 2

normalizer, 5

orbit, 7

permutation representation, 4
permutation group, 6
permutation matrix, 2

stabilizer, 8

101