MTL105: Preliminaries & Group Theory

Prof. Ekata Saha

Contents
1 Preliminaries 1
1.1 Relations and partitions on a set . . . . . . . . . . . . . . . . . . . 1
1.1.1 Equivalence relation . . . . . . . . . . . . . . . . . . . . . 2
1.2 Binary operations . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Group Theory 7
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Cyclic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Permutation groups . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.5 Cosets and Lagrange’s theorem . . . . . . . . . . . . . . . . . . . 25
2.5.1 An application of Lagrange’s theorem . . . . . . . . . . . . 29
2.6 Normal subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.7 Quotient subgroups . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.8 Homomorphism of groups . . . . . . . . . . . . . . . . . . . . . . 35
2.9 Direct products of groups . . . . . . . . . . . . . . . . . . . . . . 46
2.10 Conjugacy relation and Class equation . . . . . . . . . . . . . . . 51
2.10.1 Applications of class equation . . . . . . . . . . . . . . . . 52
2.11 Partial converses of the Lagrange’s theorem . . . . . . . . . . . . . 53
2.11.1 Applications of Cauchy’s theorem . . . . . . . . . . . . . . 54
2.12 Group actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.12.1 Orbit and stabilizer . . . . . . . . . . . . . . . . . . . . . 58
2.12.2 Kernel of a group action . . . . . . . . . . . . . . . . . . . 59
2.13 Automorphism group . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.14 Characteristic subgroups . . . . . . . . . . . . . . . . . . . . . . . 61
2.15 Sylow theorems and applications . . . . . . . . . . . . . . . . . . . 61
2.15.1 Applications of Sylow’s theorems . . . . . . . . . . . . . . 66
2.16 Structure theorem of finite abelian groups . . . . . . . . . . . . . . 69
2.16.1 Applications of the structure theorem of finite abelian groups 70
2.16.2 Proof of the structure theorem of finite abelian groups . . . 70
CHAPTER 1
Preliminaries

1.1 Relations and partitions on a set

We know that a Set is a well defined collection of distinct objects. We call a set A is
finite if the number of elements in the set A is finite, otherwise we call it an infinite set.
A very first example of an infinite set which one encounters is the set of natural numbers
which is denoted by N. Below is the list of notations which we will use throughout.

Notations

• For a set A, by |A| we mean the cardinality of A.

• By A ✓ B we mean that the set A is a subset of the set B and the set B is a
superset of the set A i.e. all the elements of A belong to B.

• By A ⇢ B we mean that the set A is a proper subset of B i.e. all the elements of
A belong to B and there exists at least one element in B which does not belong to
A.

• By A [ B we mean the set {c : c 2 A or c 2 B}. We call this set as the union of

the sets A and B.

• By A\B we mean the set {c : c 2 A and c 2 B}. We call this set as the intersection
of the sets A and B.

• We denote the empty set by ?.

• We say that the sets A and B are disjoint if A \ B = ?.

1
 2 §1.1. Relations and partitions on a set

• By the Cartesian product of the sets A and B we mean the set {(a, b) : a 2 A, b 2
B}. We denote this set by A ⇥ B.

Definition 1.1.1 Let A 6= ?. A relation R on A is a subset of A ⇥ A. Let a, a0 2 A.

We say that a is related to a0 with respect to the relation R if (a, a0 ) 2 R. If a is related
to a0 with respect to R, we denote it by aRa0 . If the set A = ?, then on A we have only
one relation in particular the empty relation.

Remark 1.1.2 In general for any relation R on a non-empty set A, aRa0 does not imply
that a0 Ra for all a, a0 2 A.

Definition 1.1.3 Let A be a set and R be a relation on A. We say that

• R is reflexive on A if for all a 2 A, we have (a, a) 2 R.

• R is symmetric on A if for any a, a0 2 A such that (a, a0 ) 2 R, we have (a0 , a) 2 R.

• R is anti-symmetric on A if for any a, a0 2 A if (a, a0 ), (a0 , a) 2 R, then we have

a and a0 are same in A. We denote a = a0 .

• R is transitive on A if for any a, a0 , a00 2 A such that (a, a0 ), (a0 , a00 ) 2 R, we have
(a, a00 ) 2 R.

Example 1.1.4 Let A = {1, 2, 3, 4} and R = {(1, 2), (2, 1), (1, 1)}. We see that R is
not reflexive as (2, 2), (3, 3) and (4, 4) 2
/ R, R is symmetric, and R is not transitive as
(2, 1), (1, 2) 2 R but (2, 2) 2/ R. Also R is not anti-symmetric as (2, 1), (1, 2) 2 R, but
1 6= 2.

1.1.1 Equivalence relation

Definition 1.1.5 Let A be a set and R be a relation on A. We say that R is an
equivalence relation on A if R is reflexive, symmetric and transitive.

Remark 1.1.6 If the set A is empty then the empty relation is vacuously reflexive,
symmetric and transitive, and hence it is an equivalence relation. The empty relation is
not refliexive on any non-empty set A, hence on any non-empty set A, the empty relation
is not an equivalence relation.

Exercise 1.1.7 Let R denote the set of all real numbers. Check which of the following
relations on R are equivalence relations:
§1.1. Relations and partitions on a set 3

1) {(x, y) 2 R ⇥ R : y = 2x}.

2) {(x, y) 2 R ⇥ R : x < y}.

3) {(x, y) 2 R ⇥ R : xy > 0}.

4) {(x, y) 2 R ⇥ R : y 6= x}.

5) {(x, y) 2 R ⇥ R : y 6= 2 + x}.

6) {(x, y) 2 R ⇥ R : x  y}.

7) {(x, y) 2 R ⇥ R : xy 0}.

8) {(x, y) 2 R ⇥ R : y = x}.

Definition 1.1.8 Let R be an equivalence relation on a non-empty set A. Now for

a 2 A, let us define the set
[a] := {a0 2 A : a0 Ra}.

Then the set [a] is called the equivalence class of the element a 2 A with respect to
the equivalence relation R.

Exercise 1.1.9 Let R be an equivalence relation on a non-empty set A. Then show the
following:

1) For all a 2 A, [a] 6= ?.

2) If a0 2 [a], then [a0 ] = [a].

3) For any a, a0 2 A, either [a] = [a0 ] or [a] \ [a0 ] = ?.

4) Let {[a↵ ] : ↵ 2 I} be the collection of all disjoint equivalence classes of A with

respect to R, where I is an indexing set. Then
[
A= [a↵ ].
↵2I

Now we define what do we mean by a partition on a non-empty set.

Definition 1.1.10 Let A 6= ? and P be a collection of some non-empty subsets of A.

Let us write P = {Ai }i2I where Ai 6= ? and Ai ✓ A. We say that P is a partition on
S
A if and only if A = Ai and Ai \ Aj = ? for all i 6= j.
i2I
 4 §1.2. Binary operations

Remark 1.1.11 Let A 6= ?. We can see immediately from the definition of a partition
that every equivalence relation on A gives rise to a partition on A.

Question 1.1.12 We can now ask the converse of the above remark i.e. for given any
partition on A 6= ?, do we get an equivalence relation on A 6= ??
The answer is yes. Let P = {Ai }i2I be a partition on A 6= ?. We can define a
relation R on A as follows:
For any two elements a, a0 2 A, we say that aRa0 if and only if both a, a0 2 Ai for some
i 2 I. One can check that R is an equivalence relation on A.

Hence we now know that the set of all partitions on a non-empty set A is in one to
one correspondence with the set of all equivalence relations on A. We end this section
by giving an interesting example of an equivalence relation on the set of integers Z.

Example 1.1.13 (Congruence relation on Z) Let n 2 N. Now we define a relation

‘⌘ (mod n)’ on Z as follows:
For any a, b 2 Z, we say that a ⌘ b (mod n) if and only if n|(a b). Check that ‘⌘
(mod n)’ is an equivalence relation on Z. This equivalence relation is called the congru-
ence relation modulo n on Z and the equivalence classes are called the congruence
classes modulo n. Check that there are n disjoint congruence classes modulo n.

1.2 Binary operations

In this section we will discuss about the binary operations on a non-empty set.

Definition 1.2.1 Let A 6= ?. By a binary operation ⇤ on A we mean a map from

A ⇥ A to A.

Example 1.2.2 The usual addition ‘+’ is a binary operation on the set of natural
numbers N.

Definition 1.2.3 Let A 6= ? and ⇤ be a binary operation on A.

• We say that ⇤ is associative if for all a, b, c 2 A, we have (a ⇤ b) ⇤ c = a ⇤ (b ⇤ c).

• We say that ⇤ is commutative if for all a, b 2 A, we have a ⇤ b = b ⇤ a.

Examples 1.2.4 The following are few examples of various binary operations.
§1.2. Binary operations 5

1) Define ⇤ on Z by a ⇤ b = a + b for all a, b 2 Z. This binary relation is commutative

as well as associative.

2) Define ⇤ on Z by a ⇤ b = a b for all a, b 2 Z. This binary relation is neither

commutative nor associative.

3) Define ⇤ on Z by a ⇤ b = |a + b| for all a, b 2 Z. This binary relation is commutative

but not associative.

4) Define ⇤ on Z by a ⇤ b = a for all a, b 2 Z. This binary relation is not commutative

but associative.

Question 1.2.5 Let A be a set of n elements, n 2 N. How many binary operations one
can get on A?

A binary operation ⇤ is a map from A ⇥ A to A. So ⇤ is a map from a set of n2 elements

to a set of n elements. For each element in A ⇥ A, we have exactly n choices in A to
2
map to. Since there are n2 elements in A ⇥ A, one can have nn many binary operations
on A.

Question 1.2.6 Let A be a set of n elements, n 2 N. How many commutative binary

operations one can get on A?
2
We already know that there are total nn many binary operations on A. A binary
operation ⇤ is commutative means a ⇤ b = b ⇤ a for all a, b 2 A. So for two elements
a, b 2 A, it is enough to to know where does (a, b) map to. Now there are exactly n many
elements of the form (a, a) in A ⇥ A. So there are n2 n many elements of the form
(a, b) with a 6= b. Hence we need to know where do these (n2 n)/2 + n = (n2 + n)/2
many elements map to. Again each element has n choices to get mapped to. So one can
2
get total n(n +n)/2 many commutative binary operations on A.

Definition 1.2.7 Let A 6= ? and ⇤ be a binary operation on A.

• We say, ⇤ satisfies left cancellation law if for any a, b, c 2 A, a ⇤ b = a ⇤ c implies

that b = c.

• We say, ⇤ satisfies right cancellation law if for any a, b, c 2 A, a ⇤ c = b ⇤ c implies

that a = b.
 6 §1.2. Binary operations

Example 1.2.8 Define ⇤ on Z \ {0} by a ⇤ b := |a|b for all non-zero integers a, b. This
binary operation ⇤ on Z \ {0} satisfies left cancellation law but does not satisfy right
cancellation law.

Definition 1.2.9 Let A 6= ? and ⇤ be a binary operation on A.

• An element e0 2 A is called a left identity of (A, ⇤) if for all a 2 A, e0 ⇤ a = a.

• An element e00 2 A is called a right identity of (A, ⇤) if for all a 2 A, a ⇤ e00 = a.

• An element e 2 A is called an identity of (A, ⇤) if for all a 2 A, a ⇤ e = a = e ⇤ a.

Examples 1.2.10 The following are few examples of various binary operations with
different types of identity elements.

1) Define ⇤ on Z by a ⇤ b = a + b for all a, b 2 Z. This binary relation has a left

identity, a right identity as well as an identity.

2) Define ⇤ on Z by a ⇤ b = a b for all a, b 2 Z. This binary relation has a right

identity but no left identity and no identity.

3) Define ⇤ on Z by a ⇤ b = |a + b| for all a, b 2 Z. This binary relation has no left

identity, no right identity and no identity.

4) Define ⇤ on Z by a ⇤ b = b for all a, b 2 Z. This binary relation has a left identity,

but no right identity and no identity.

Theorem 1.2.11 Let A 6= ? and ⇤ be a binary operation on A. If there exists an

identity of A, then it is unique.

Proof. Let us assume that e1 , e2 2 A be two identities of A. So we have for all a 2 A,

e1 ⇤ a = a and a ⇤ e2 = a.

Hence we get,
e1 = e1 ⇤ e2 = e2 .

This completes the proof.

CHAPTER 2
Group Theory

2.1 Introduction
Definition 2.1.1 A group is an ordered pair (G, ⇤) where G is a non-empty set and ⇤
is a binary operation on G such that the following hold:

• For all a, b, c 2 G, a ⇤ (b ⇤ c) = (a ⇤ b) ⇤ c.

• There exists an element e 2 G such that a ⇤ e = a = e ⇤ a for all a 2 G.

• For each a 2 G, there exists b 2 G such that a ⇤ b = e = b ⇤ a.

So (G, ⇤) is a group if ⇤ is associative, the identity exists in G, and for each element in
G, we have an inverse of that element in G.

Theorem 2.1.2 Let (G, ⇤) be a group. Then for each a 2 G, there exists unique b 2 G
such that a ⇤ b = e = b ⇤ a where e is the identity of (G, ⇤).

Proof. Let a 2 G. Suppose there are two elements b, c 2 G such that

a ⇤ b = e = b ⇤ a and a ⇤ c = e = c ⇤ a.

We shall show that b = c. Note that

b = b ⇤ e = b ⇤ (a ⇤ c) = (b ⇤ a) ⇤ c = e ⇤ c = c.

Hence the proof.

7
 8 §2.1. Introduction

Remark 2.1.3 For each a 2 G, the unique element b 2 G such that a ⇤ b = e = b ⇤ a, is

called the inverse of a. Generally b is denoted by a 1 .

Definition 2.1.4 Let (G, ⇤) be a group.

• (G, ⇤) is called abelian or commutative if a ⇤ b = b ⇤ a for all a, b 2 G. Otherwise

(G, ⇤) is called non-abelian or non-commutative.

• (G, ⇤) is called finite group if the number of elements of G is finite. Otherwise it

is called an infinite group.

Definition 2.1.5 By the order of a group (G, ⇤) we mean the number of elements of
G. If the number of elements of G is finite, then G is called a group of finite order or a
finite group, otherwise G is called a group of infinite order or infinite group.

Examples 2.1.6 The following are few examples of groups.

1) (Z, +) is an infinite abelian group.

2) For n 2 N, we denote the set of all the congruence classes modulo n by Z/nZ. We
define the binary operation ‘+’ on Z/nZ as follows:
For [a], [b] 2 Z/nZ, [a] + [b] := [a + b]. First we show that the operation ‘+’ is
well-defined. Let [a] = [a0 ] and [b] = [b0 ]. We need to show that [a + b] = [a0 + b0 ].
Since [a] = [a0 ], we have n | a a0 . Similarly since [b] = [b0 ], we have n | b b0 .
Therefore n | (a + b) (a0 + b0 ). This implies that [a + b] = [a0 + b0 ].
Note that [0], [1], . . . , [n 1] are distinct elements of Z/nZ. Also for any integer
a 2 Z, by division algorithm we can check that a ⌘ r (mod n) for some 0  r < n.
Therefore the cardinality of Z/nZ is exactly n. Under the operation ‘+’, Z/nZ
forms a group. So (Z/nZ, +) is a finite abelian group.

3) Consider GL2 (R), the set of all 2 ⇥ 2 invertible matrices (with respect to the usual
matrix multiplication) with the entries from R. This set under the usual matrix
multiplication is a group. This is an infinite non-abelian group.

4) Consider GL2 (Z/2Z), the set of all 2 ⇥ 2 invertible matrices (with respect to usual
matrix multiplication) with the entries from Z/2Z. Under the usual matrix multi-
plication, GL2 (Z/2Z) is a group. This is a finite non-abelian group.

Exercise 2.1.7 Let (G, ⇤) be a group. Show that

§2.1. Introduction 9

1) For all a 2 G, (a 1 ) 1
= a.

2) For all a, b 2 G, (a ⇤ b) 1
=b 1
⇤ a 1.

From now on, for simplicity we will write a · · ⇤ a} = an .

| ⇤ ·{z
n times

Exercise 2.1.8 Let (G, ⇤) be an abelian group. Show that for all a, b 2 G and n 2 Z,

(a ⇤ b)n = an ⇤ bn .

Exercise 2.1.9 Let (G, ⇤) be a group such that for all a, b 2 G,

1 1
(a ⇤ b) =a ⇤ b 1.

Show that (G, ⇤) is an abelian group.

Definition 2.1.10 Let (G, ⇤) be a group and a 2 G. Suppose there exists a positive
integer m such that am = e, where e is the identity of (G, ⇤). Then the least positive
integer n such that an = e, is called the order of a. If no such integer exists, then we
call that a is of infinite order. By the notation o(a), we refer to the order of a.

Remark 2.1.11 Note that in a group (G, ⇤), for an element a 2 G if o(a) = n, then
o(a 1 ) = n.

Theorem 2.1.12 Let (G, ⇤) be a group and a 2 G be such that o(a) = n. Let m 2 N
be such that am = e, then n|m.

Proof. By division algorithm we know there exist integers q, r such that

m = nq + r, where 0  r < n.

Therefore,

ar = am qn
= am ⇤ a qn
= e ⇤ (a 1 )qn = e, where e is the identity of G.

Since r < n and o(a) = n, we get that r = 0. This proves that m|n.

Theorem 2.1.13 Let (G, ⇤) be a finite group. Then every element of G is of finite
order.
 10 §2.1. Introduction

Proof. Let us assume that |G| = n. We have the order of the identity e is 1. Let a 2 G be
such that a 6= e. Consider the collection {e, a, a2 , . . . , an } ✓ G. By convention, a0 := e.
Since the order of G is n, there exist 0  m < r  n such that am = ar . By convention,
a0 := e. Therefore, ar m = e, which implies that o(a)  r m  n. This proves that
order of any element in (G, ⇤) is finite.

Question 2.1.14 Does there exist an infinite group whose every element is of finite
order?

Let us first consider the finite group (Z/2Z, +). Every non-zero element of this group is
of order 2. Now consider the infinite Cartesian product of Z/2Z. Since infinite Cartesian
Q
product of sets having more than one element is infinite, G := Z/2Z is infinite. We
1
consider point-wise addition as the binary operation on G. Every non-zero element of
(G, +) is of order 2.

Theorem 2.1.15 Let (G, ⇤) be a finite group with |G| = n and a 2 G. Then an = e.

Proof. If a = e, then clearly an = e. So let a 6= e. We know that o(a)  n. Let

o(a) = m. Consider {e, a, . . . , am 1 } ✓ G. If {e, a, . . . , am 1 } = G, then m = n.
Therefore an = e. So let {e, a, . . . , am 1 } 6= G, we consider x 2 G \ {e, a, . . . , am 1 } and
the set {x, xa, . . . , xam 1 }. We claim that {e, a, . . . , am 1 } \ {x, xa, . . . , xam 1 } = ?. If
xak = al , for some 0  k, l < m, then x = al k . If l k, then 0  l k < m and so
x 2 {e, a, . . . , a m 1
}. If l < k, then 0 < k l < m, and so x = (a 1 )k l 2 {e, a, . . . , am 1 }
as a 1 = am 1 . In both the cases we get x 2 {e, a, . . . , am 1 }, a contradiction. So the
sets {e, a, . . . , am 1 } and {x, xa, . . . , xam 1 } are disjoint.
Now if {e, a, . . . , am 1 } [ {x, xa, . . . , xam 1 } = G, then n = 2m and hence an =
e. If {e, a, . . . , am 1 } [ {x, xa, . . . , xam 1 } = 6 G, then we choose y 2 G so that y 2 /
{e, a, . . . , am 1
} and y 2
/ {x, xa, . . . , xa m 1
}. Following the above argument, we similarly
get

{e, a, . . . , am 1 }\{y, ya, . . . , yam 1 } = ? and {y, ya, . . . , yam 1 }\{x, xa, . . . , xam 1 } = ?.

If {e, a, . . . , am 1 } [ {x, xa, . . . , xam 1 } [ {y, ya, . . . , yam 1 } = G, we get n = 3m and

hence an = e, otherwise we continue the above process which will stop after finitely
many steps as G is finite. Hence G can be written as a disjoint union of finitely many
sets each having m elements i.e. n = rm for some positive integer r and hence an = e.
§2.2. Subgroups 11

Theorem 2.1.16 Let (G, ⇤) be a group and a 2 G be such that o(a) = n. Then for
every m 2 N,
n
o(am ) = .
gcd(m, n)
Proof. Let us assume that o(am ) = k. So amk = e, where e is the identity of G. Now
since o(a) = n, we get that n | mk. Therefore, mk = nr for some r 2 N.
Also let gcd(m, n) = d. So there exist u, v 2 N such that m = du and n = dv with
gcd(u, v) = 1. So substituting the values of m, n in mk = nr we get that, duk = dvr i.e.
uk = vr. This implies that v | uk. Since gcd(u, v) = 1, we get that v|k i.e. (n/d) | k.
Again,
n mn dun
(am ) d = a d = a d = aun = e.

As o(am ) = k, we also get that k | (n/d). This proves that k = n/d i.e.

n
o(am ) = .
gcd(m, n)

2.2 Subgroups
Definition 2.2.1 Let (G, ⇤) be a group and H be a non-empty subset of G. We say
that H is a subgroup of (G, ⇤) if (H, ⇤) is also a group.

Remark 2.2.2 In a group (G, ⇤), we always have two subgroups H = {e}, where e is
the identity of (G, ⇤) and H = G. They are called trivial subgroups of (G, ⇤). A
subgroup H of G such that H 6= {e} and H 6= G, is called a non-trivial subgroup of
(G, ⇤).

Example 2.2.3 The set of all even numbers i.e. 2Z is a subgroup of (Z, +).

Theorem 2.2.4 All subgroups of (G, ⇤) have same identity.

Proof. Let H be a subgroup of (G, ⇤). We denote the identity of (G, ⇤) by e. We show
that e is also the identity of (H, ⇤). Let us denote the identity of H by eH . So we have,

eH = eH ⇤ eH .

Also we have,
eH = eH ⇤ e, as e is the identity of (G, ⇤).
 12 §2.2. Subgroups

Therefore, eH ⇤ eH = eH ⇤ e and hence using left cancellation law we get eH = e.

Remark 2.2.5 For a 2 H, where H is a subgroup of (G, ⇤), the inverse of a in G and
H are same.

Theorem 2.2.6 Let (G, ⇤) be a group and H be a non-empty subset of G. Then H is

a subgroup of (G, ⇤) if and only if for all a, b 2 H, a ⇤ b 1 2 H.

Proof. First we assume that H is a subgroup of (G, ⇤). We shall show that for all
a, b 2 H, a ⇤ b 1 2 H. Let a, b 2 H. Since H is a subgroup of G, we have a, b 1 2 H
and therefore a ⇤ b 1 2 H.
Next assume that for all a, b 2 H, a ⇤ b 1 2 H. We show that H is a subgroup of
(G, ⇤). Since H 6= ?, there exists an element a 2 H. Now from the hypothesis we get
e = a ⇤ a 1 2 H, where e is the identity of (G, ⇤).
Now let b 2 H, so for e, b 2 H, we get from the hypothesis that b 1
= e⇤b 1
2 H. So
we have proved that every element in H has an inverse in H.
Let a, b 2 H. So a, b 1 2 H as we have already shown that each element of H has an
inverse in H. Now from the hypothesis we obtain that a ⇤ b = a ⇤ (b 1 ) 1 2 H. This
proves that ⇤ is a binary operation on H.
The associativity of the elements of H under ⇤ follows from the fact that H is a subset
of G. This completes the proof of the fact that H is a subgroup of (G, ⇤).

Theorem 2.2.7 Let (G, ⇤) be a group and {H↵ : ↵ 2 I} be a collection of subgroups of

T
G. Then H := H↵ is a subgroup of (G, ⇤).
↵2I

Proof. Since H↵ ’s are subgroups of (G, ⇤), e 2 H↵ for all ↵ 2 I where e is the identity
of (G, ⇤). Therefore H 6= ?. Now let a, b 2 H. Then a, b 2 H↵ for all ↵ 2 I. Since each
H↵ is a subgroup of (G, ⇤), we have a, b 1 2 H↵ and therefore a ⇤ b 1 2 H↵ for all ↵ 2 I.
This proves that a ⇤ b 1 2 H and hence H is a subgroup of (G, ⇤).

Definition 2.2.8 Let (G, ⇤) be a group. By center of the group (G, ⇤) we mean the
set Z(G) := {x 2 G : x ⇤ g = g ⇤ x for all g 2 G}.

Remark 2.2.9 Note that Z(G) 6= ? as the identity of G is always in Z(G). Also note
that a group (G, ⇤) is abelian if and only if G = Z(G).

Theorem 2.2.10 The center Z(G) is a subgroup of (G, ⇤).

§2.2. Subgroups 13

Proof. We have already observed that Z(G) 6= ?. So let a, b 2 Z(G). We show that
a ⇤ b 1 2 Z(G). Since b 2 Z(G), we have b ⇤ g = g ⇤ b for all g 2 G. Therefore
g ⇤ b 1 = b 1 ⇤ g for all g 2 G. Hence we have for all g 2 G,

(a ⇤ b 1 ) ⇤ g = a ⇤ (b 1
⇤ g) = a ⇤ (g ⇤ b 1 ) = (a ⇤ g) ⇤ b 1
= (g ⇤ a) ⇤ b 1
= g ⇤ (a ⇤ b 1 ).

This completes the proof.

Definition 2.2.11 Let A 6= ? be a subset of a group (G, ⇤). By the centralizer of A

in (G, ⇤) we mean the following set,

CG (A) := {g 2 G : g ⇤ a = a ⇤ g for all a 2 G}.

Exercise 2.2.12 Prove that CG (A) is a subgroup of (G, ⇤).

Remark 2.2.13 Note that Z(G) ✓ CG (A) and CG (G) = Z(G).

Let H, K be two subgroups of a group (G, ⇤). Then we define,

HK := {h ⇤ k : h 2 H, k 2 K}.

In general HK need not be a subgroup of (G, ⇤). The following theorem gives us a
necessary and sufficient condition on HK so that HK becomes a subgroup of (G, ⇤).

Theorem 2.2.14 The following are equivalent:

1) HK is a subgroup of (G, ⇤).

2) HK = KH.

3) KH is a subgroup of (G, ⇤).

Proof. First we assume that HK is a subgroup of (G, ⇤). We will show that HK = KH.
Let h ⇤ k 2 HK where h 2 H and k 2 K. So (h ⇤ k) 1 = k 1
⇤h 1
2 HK. We write
k 1 ⇤ h 1 = h1 ⇤ k1 where h1 2 H and k1 2 K. Therefore,

1
h ⇤ k = (k ⇤ h 1) 1
= (h1 ⇤ k1 ) 1
= k1 1 ⇤ h1 1 2 KH.

This proves that HK ✓ KH.

 14 §2.2. Subgroups

Now let k ⇤ h 2 KH where k 2 K and h 2 H. Note that,

k = e ⇤ k 2 HK and h = h ⇤ e 2 HK,

where e denotes the identity of (G, ⇤). Since HK is a subgroup of (G, ⇤), we obtain that
k ⇤ h 2 HK. Hence KH ✓ HK. This completes the proof of the fact that if HK is a
subgroup of (G, ⇤) then HK = KH.
Next we assume that HK = KH. We shall show that HK is a subgroup of (G, ⇤).
Let h1 ⇤ k1 , h2 ⇤ k2 2 HK where h1 , h2 2 H and k1 , k2 2 K. Now,

1
(h1 ⇤ k1 ) ⇤ (h2 ⇤ k2 ) = h1 ⇤ (k1 ⇤ k2 1 ) ⇤ h2 1 .

We know (k1 ⇤ k2 1 ) ⇤ h2 1 2 KH. We have KH = HK. So we can write (k1 ⇤ k2 1 ) ⇤ h2 1 =

h ⇤ k for some h 2 H and k 2 K. Hence (h1 ⇤ k1 ) ⇤ (h2 ⇤ k2 ) 1 = (h1 ⇤ h) ⇤ k 2 HK. This
proves that HK is a subgroup of (G, ⇤).

Remarks 2.2.15

• If (G, ⇤) is an abelian group, then for any two subgroups H, K of (G, ⇤), we get
HK is a subgroup of (G, ⇤).

• In general for non-abelian groups (G, ⇤), for any two subgroups H, K of (G, ⇤),
HK need not be a subgroup of (G, ⇤). For example, consider G = GL2 (Z/2Z)
with usual matrix multiplication. We take the subgroups
⇢✓ ◆ ✓ ◆
10 11
H= , ,
01 01
⇢✓ ◆ ✓ ◆
10 01
K= , .
01 10
Now ⇢✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
10 11 01 11
HK = , , , .
01 01 10 10
1 1
Note that order of 1 0
is 3 and as 3 - 4, hence we conclude that HK is not a
subgroup of G.
§2.3. Cyclic groups 15

2.3 Cyclic groups

Definition 2.3.1 A group (G, ⇤) is called cyclic if there exists an element a 2 G such
that G = {an : n 2 Z}. The set {an : n 2 Z} is denoted by hai.

Definition 2.3.2 Let (G, ⇤) be a cyclic group such that G = hai for some a 2 G. The
element a is called a generator of G.

Remark 2.3.3 For a cyclic group (G, ⇤), if a is a generator of G, then a 1

is also a
generator of G.

Examples 2.3.4 The following are few examples of cyclic groups.

1) The group (Z, +) is an infinite cyclic group and Z = h1i.

2) For any n 2 N, the group (Z/nZ, +) is a finite cyclic group and Z/nZ = h[1]i.

Theorem 2.3.5 Every cyclic group is abelian.

Proof. Let G = hai be a cyclic group. Let b, c 2 G. Then b = am and c = an for some
m, n 2 Z. Therefore,

b ⇤ c = am ⇤ an = am+n = an+m = an ⇤ am = c ⇤ b.

This proves that (G, ⇤) is an abelian group.

The converse of the above theorem is not true i.e. there are abelian groups which
are not cyclic. Here we give one example of an abelian group which is not cyclic. Let
G = {e, a, b, c} and define ⇤ on G by the following:

a ⇤ a = b ⇤ b = c ⇤ c = e,
a ⇤ b = b ⇤ a = c, b ⇤ c = c ⇤ b = a, a ⇤ c = c ⇤ a = b,
and e is the identity element.

This group (G, ⇤) is known as Klein’s 4 group. Clearly this group is abelian but not
cyclic.

Theorem 2.3.6 Let (G, ⇤) be a finite group of order n. Then (G, ⇤) is cyclic if and only
if there exists an element a 2 G such that o(a) = n.
 16 §2.3. Cyclic groups

Proof. Throughout the proof we denote e as the identity of (G, ⇤).

First we assume that we have an element a 2 G such that o(a) = n. We show that
G = hai, and hence G is cyclic. Since o(a) = n, we have that the number of elements
of the set {e, a, . . . , an 1 } is n. As {e, a, . . . , an 1 } ✓ G and both the sets have same
number of elements, we obtain that G = {e, a, . . . , an 1 } = hai.
Next we assume that (G, ⇤) is a cyclic group of order n. Let a be a generator of (G, ⇤).
We shall show that o(a) = n. We know that in a finite group, order of each element is
finite. So let o(a) = m. Then the number of elements in the set S := {e, a, . . . , am 1 } ✓ G
is m. Now let ak 2 G. Using division algorithm we get two integers m, r such that
k = mq + r where 0  r < m. Therefore,

ak = amq+r = amq ⇤ ar = e ⇤ ar = ar 2 S, as r < m.

This proves that S = G. Therefore m = |S| = |G| = n. Hence we have found an element
in G whose order is n.

Corollary 2.3.7 Every group of prime order is cyclic.

Proof. Let (G, ⇤) be a group of prime order p. It is enough for us to find an element of
order p in (G, ⇤). As |G| = p > 1, there exists a 6= e in G. We know that in a finite
group order of every element divides order of the group. So o(a) = 1 or p. Since a 6= e,
we get o(a) = p.

Exercise 2.3.8 Let G = hai and o(a) = n. Let k 2 N be such that 1  k  n. Then
show that ak is also a generator of G if and only if gcd(k, n) = 1.

Theorem 2.3.9 All subgroups of a cyclic group are cyclic.

Proof. Let G = hai and H be a subgroup of G. If H = {e}, where e is the identity of

(G, ⇤), then clearly H is cyclic.
So let H 6= {e}, then there exists an element b 2 H such that b 6= e. Since b 2 G
also, there exists m 2 Z such that b = am . Also b 1 = a m 2 H. Among m, m, one
is positive. So we choose n to be the least positive integer so that an 2 H. We show
that H = han i. Clearly han i ✓ H as an 2 H. It only remains to show that H ✓ han i.
Let h 2 H. Now h = as , for some s 2 Z. By division algorithm we can find integers q, r
such that s = qn + r and 0  r < n. Therefore,

ar = as qn
= as ⇤ (a 1 )qn 2 H.
§2.3. Cyclic groups 17

By the choice of n we get r = 0 as r < n. Hence s = qn. This proves that h 2 han i.
This completes the proof that H is cyclic.

Theorem 2.3.10 Let G be a finite cyclic group and H is a subgroup of G. Then |H|
divides |G|.

Proof. If H = {e}, then |H| = 1, so |H| divides |G|. So let H 6= {e}. Since H is
a subgroup of a cyclic group G, H is also cyclic. So there exists b 2 H such that
o(b) = |H|. Now b 2 G also. Let G = hai. Then b = am for some integer m. Also
b 1 = a m 2 H and o(b) = o(b 1 ). Among m, m, one is positive. Without loss of
generality let m > 0. Then
o(a)
o(am ) = .
gcd(o(a), m)
This proves that o(am ) divides o(a) i.e. |H| divides |G|.

Theorem 2.3.11 Let G be a finite cyclic group of order n. Let m 2 N be such that
m | n. Then there exists a unique subgroup of G of order m.

Proof. Let G = hai. Then o(a) = |G| = n. First we show the existence of a subgroup
of G of order m. Given that m | n. Therefore, n = mk for some k 2 N. Consider
H := hak i. Now we have,

o(a) n
o(ak ) = = = m.
gcd(o(a), k) k

Now |H| = o(ak ) = m. This proves the existence of a subgroup of G of order m.

Now we show the uniqueness of such a subgroup. Let K be a subgroup of G of order
m. We shall show that K = H. We know that K is also cyclic as K is a subgroup of a
cyclic group. Let K := hat i for some t 2 Z. Now m = |K| = o(at ). Therefore atm = e,
where e is the identity of (G, ⇤). Since o(a) = n, we have n | tm. Hence tm = nr for
some r 2 Z. This implies that t = kr. So at = akr 2 H. We have |H| = |K| and K ✓ H.
Therefore, K = H.

Remark 2.3.12 As a consequence of the above theorem we can say that in a cyclic
group of order n 2 N, there are exactly d(n) many subgroups where d(n) denotes the
number of positive divisors of n.
 18 §2.4. Permutation groups

2.4 Permutation groups

Definition 2.4.1 Let A be a non-empty set. By a permutation of A, we mean a
bijective map from A to itself.

Definition 2.4.2 A group (G, ⇤) is called a permutation group, if G consists of ‘some’

permutations of a non-empty set A and the operation ⇤ is the composition of maps.

As an example of permutation group, we look at the symmetric group with n symbols

in more detail. Let In = {1, 2, . . . , n}. Denote, Sn as the collection of all permutations
of In . Under the operation composition of maps, Sn forms a group which we refer to as
the symmetric group with n symbols. Note that |Sn | = n!. Let 2 Sn be such that
for any i 2 In , (i) := ai , ai 2 In and ai 6= aj for all i 6= j. Then one can denote by
the following two row notation:
✓ ◆
1 2 ··· n
.
a1 a2 · · · an

Example 2.4.3 We consider the group (S3 , ). We know that |S3 | = 6. Below we list
down all the elements of S3 in two row notation.
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
123 123 123 123 123 123
, , , , , .
123 132 213 231 321 312

Note that, ✓ ◆ ✓ ◆ ✓ ◆
123 123 123
= ,
132 213 312
and ✓ ◆ ✓ ◆ ✓ ◆
123 123 123
= .
213 132 231
So we see that ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
123 123 123 123
6= .
132 213 213 132
This shows that (S3 , ) is non-abelian.

Exercise 2.4.4 Show that (Sn , ) is non-abelian for all n 3.

Now we give examples of subgroups H, K of (S3 , ) such that HK is not a subgroup

of (S3 , ).

Definition 2.4.5 A permutation of In is called a k-cycle or a cycle of length k if

§2.4. Permutation groups 19

there exist distinct a1 , a2 , . . . , ak 2 In such that (ai ) = ai+1 for all 1  i  (k 1),
(ak ) = a1 , and (a) = a for all a 2 In \ {a1 , a2 , . . . , ak }. We denote as (a1 a2 · · · ak ).

By convention we consider the identity permutation as 1-cycle.

Remarks 2.4.6 Note that

1) In (S3 , ), all the non-identity permutations of can be written as cycles of length

less than or equal to 3. Below are the non-identity elements of (S3 , ):

(2 3), (1 2), (1 2 3), (1 3), (1 3 2).

2) As we see in (S3 , ), all non-identity elements are cycles, we may wonder whether
for any n, all non-identity elements of (Sn , ) are cycles?
Let us consider, ✓ ◆
1234
2 S4 .
2143
Note that ✓ ◆
1234
= (1 2)(3 4) = (3 4)(1 2).
2143
So we see by this example that not all non-identity permutations are cycles for
n 4.

Theorem 2.4.7 Let 2 Sn be a cycle. Then is a k-cycle if and only if o( ) = k.

Proof. First assume that is a k-cycle. Let us write,

= (a1 · · · ak ).

Note that for any 1  m < k,

m
(a1 ) = am+1 6= a1 .

Also, k (ai ) = ai for all 1  i  k. This proves that o( ) = k.

Next assume that is a cycle and o( ) = k. We show that is a k-cycle. Let length
of be m. So, o( ) = m as order of an m-cycle is m. Given that o( ) = k. Therefore
m = k. This proves that is a k-cycle.

Remark 2.4.8 So it justifies the convention of considering the identity permutation as

1-cycle.
 20 §2.4. Permutation groups

Definition 2.4.9 Two cycles (a1 a2 · · · ak ) and (b1 b2 · · · ar ) in (Sn , ) are called
T
disjoint cycles if {a1 , a2 , . . . , ak } {b1 , b2 , . . . , br } = ?.

Exercise 2.4.10 Let ↵, be two disjoint cycles in (Sn , ). Show that ↵ = ↵.

Theorem 2.4.11 Let 2 Sn , n 2 be such that = 1 · · · r , where i ’s are disjoint

cycles for all 1  i  r. Let o( i ) = ni for all 1  i  r. Then o( ) = lcm(n1 , . . . , nr ).

Proof. Let o( ) = t and lcm(n1 , . . . , nr ) = m. We want to show that t = m. Note that,

m
=( 1 ··· r)
m
= m
1 ··· m
r , as i ’s are disjoint cycles.
= e.

Therefore, t | m. Now to show m | t, it is enough to show that ni | t for all 1  i  r.

For that we show, it = e for all 1  i  r. Let k 2 In .

• If i (k) = k, then t
i (k) = k.

• If i (k) =
6 k, then j (k) = k for all j 6= i as i ’s are disjoint cycles. Therefore,
(k) = i (k). So it (k) = t (k) = k.

This proves that ni | t, and hence m | t. This completes the proof.

We saw earlier that ✓ ◆
1234
2 S4 ,
2143
is not a cycle but a product of two disjoint 2-cycles. So we may wonder that for any n,
can we say that any non-identity permutation in (Sn , ) is a product of disjoint cycles?
To answer this, we have the following theorem.

Theorem 2.4.12 Let n 2 and 2 Sn be such that 6= e. Then can be written as

a product of disjoint cycles of length 2.

Proof. We have that 6= e and n 2, so there exists m 2 In such that (m) 6= m. We

know that (Sn , ) is a finite group. So the order of is finite. Let o( ) = t i.e. t = e.
This implies that t (m) = m. We choose 1 < k  t to be the least positive integer so
that k (m) = m. Now we consider,

k 1
↵1 := (m (m) · · · (m)).

Case: 1 All the elements of In , which are not fixed by appear in ↵1 .

§2.4. Permutation groups 21

Let a 2 In be such that (a) 6= a and a 2 ↵1 , then the map ↵1 takes a to (a). Now
let b 2 In be such that (b) = b, then b 2/ ↵1 as l (b) = b for any positive integer l, so
↵1 (b) = b. This proves that the two maps ↵1 and and same. So itself is a cycle.
Case: 2 There are elements of In , which are not fixed by and do not appear in ↵1 .
Let m0 2 In be such that (m0 ) 6= m0 and m0 does not appear in In . Again we have,
0
t
(m0 ) = m0 . So we choose the least positive integer 1 < k 0  t such that k (m0 ) = m0 .
We write,
0
↵2 := (m0 (m0 ) · · · k 1 (m0 )).

We show that ↵1 , ↵2 are disjoint cycles. We know that m 6= m0 . Now if ↵1 , ↵2 are not
disjoint, then there exist integers p, q, 1  p  k, 1  q  k 0 such that p (m) = q (m0 ).
Now,
p+1
(m) = ( p (m)) = ( q (m0 )) = q+1 (m0 ).

Therefore,
p+i
(m) = q+i
(m0 ), for all i 2 N [ {0}.

As q  k 0 , we can choose an integer j 2 N [ {0} such that q + j = k 0 . So,

k0
m0 = (m0 ) = q+j
(m0 ) = p+j
(m).

This shows that m0 2 ↵1 , a contradiction. So ↵1 , ↵2 are disjoint cycles and hence

↵1 ↵2 = ↵2 ↵1 .

Case: 2A All the elements which are not fixed by appear in either ↵1 or ↵2 .
Let (a) 6= a. If a 2 ↵1 , then a 2
/ ↵2 , as ↵1 , ↵2 are disjoint cycles. Now a 2
/ ↵2
implies that ↵2 (a) = a. So,

↵1 ↵2 (a) = ↵1 (a) = (a).

Similarly if a 2 ↵2 , then
↵1 ↵2 (a) = ↵2 (a) = (a).

If (b) = b, then following the justification given in Case: 1 we get that b is fixed by
both ↵1 , ↵2 . This proves that = ↵1 ↵2 .
Case: 2B There are elements of In , which are not fixed by and do not appear in
↵ 1 , ↵2 .
Then again with such an element we construct a cycle ↵3 . We repeat this process,
 22 §2.4. Permutation groups

it terminates at a finite stage as In is finite. In the process, the cycles we construct are
disjoint. If we finally get r many disjoint cycles ↵1 , . . . , ↵r , then following the justification
given in Case 2A we get that
= ↵1 · · · ↵r .

This completes the proof.

Definition 2.4.13 A 2-cycle in Sn is called a transposition.

Theorem 2.4.14 Every cycle of length greater than or equal to 2 can be written as a
product of transpositions.

Proof. Let = (a1 a2 · · · ak ) be a k-cycle and k 2. We claim that

= (a1 ak ) · · · (a1 a3 ) (a1 a2 ).

We denote ↵ = (a1 ak ) · · · (a1 a3 ) (a1 a2 ). Note that (a1 a2 ) maps a1 to a2 , and a2

does not belong to the rest of the transpositions, and hence a2 is fixed by the rest (k 2)
many transpositions. So ↵ maps a1 to a2 . Now (a1 a2 ) maps a2 to a1 , (a1 a3 ) maps a1
to a3 , and a3 does not belong to the rest of the transpositions, and hence a3 is fixed by
the rest (k 3) many transpositions. So ↵ maps a2 to a3 . Now (a1 a2 ) maps a3 to a3 ,
(a1 a3 ) maps a3 to a1 , (a1 a4 ) maps a1 to a4 and a4 does not belong to the rest of the
transpositions, and hence a4 is fixed by the rest (k 4) many transpositions. So ↵ maps
a3 to a4 . Proceeding this way we get that ↵ maps ai to ai+1 for all 1  i  (k 1) and
↵ maps ak to a1 as ak belongs to only (a1 ak ). So we get that ↵ = and this completes
the proof of the theorem.
So from the above two theorems we can conclude the following:

Theorem 2.4.15 Every non-identity permutation in Sn can be written as a product of

transpositions.

At this point, we may wonder what about the identity permutation? Note that we
can write e = (1 k) (1 k) for any 2  k  n. So e can be written as a product of two
transpositions. In fact, we can show that e can only be a product of even number of
transpositions.

Theorem 2.4.16 The identity permutation e in Sn can not be written as a product of

odd many permutations.
§2.4. Permutation groups 23

Proof. We consider a polynomial in n variables as follows:

Y
P (X1 , . . . , Xn ) = (Xk Xl ).
1k<ln

For any permutation 2 Sn , we define

Y
(P (X1 , . . . , Xn )) := P (X (1), . . . , X (n)) = (X (k) X (l) ).
1k<ln

So according to our definition,

Y
e(P (X1 , . . . , Xn )) = (Xk Xl ) = P (X1 , . . . , Xn ).
1k<ln

Suppose, e = 1 · · · r where r is an odd number and 1, . . . , r are transpositions.

Let = (i j) and i < j be a transposition. We show that

(P (X1 , . . . , Xn )) = P (X1 , . . . , Xn ).

Below we write down the factors of P (X1 , . . . , Xn ):

(X1 X2 ) · · · (X1 Xn )
(X2 X3 ) · · · (X2 Xn )
..
.
(Xi 1 Xi ) · · · (Xi 1 Xj ) · · · (Xi 1 Xn )
(Xi Xi+1 ) · · · (Xi Xj 1 )(Xi Xj )(Xi Xj+1 ) · · · (Xi Xn )
(Xi+1 Xi+2 ) · · · (Xi+1 Xj 1 )(Xi+1 Xj )(Xi+1 Xj+1 ) · · · (Xi+1 Xn )
..
.
(Xj 1 Xj )(Xj 1 Xj+1 ) · · · (Xj 1 Xn )
(Xj Xj+1 ) · · · (Xj Xn )
(Xj+1 Xj+2 ) · · · (Xj+1 Xn )
..
.
(Xn 1 Xn )
 24 §2.4. Permutation groups

We see that under 1,

(Xi Xi+1 ) · · · (Xi Xj 1 ) 7!(Xj Xi+1 ) · · · (Xj Xj 1 )

= ( 1)j i 1
(Xi+1 Xj ) · · · (Xj 1 Xj ).

Also under , each factor of the form (Xi+k Xj ) 7! (Xi+k Xi ) = (Xi Xi+k )
for every 1  k  j i 1. Hence

(Xi+1 Xj ) · · · (Xj 1 Xj ) 7!(Xi+1 Xi ) · · · (Xj 1 Xi )

= ( 1)j i 1
(Xi Xi+1 ) · · · (Xi Xj 1 ).

Hence under , the factor

(Xi Xi+1 ) · · · (Xi Xj 1 )(Xi+1 Xj ) · · · (Xj 1 Xj )

remains fixed. The factor (Xi Xj ) 7! (Xj Xi ) = (Xi Xj ).

Now for the remaining factors, each of them either gets fixed under or gets mapped
to another factor among them. Note that is a bijection of In , therefore two different
factors can never get mapped to a same factor. Hence the product of all the remain-
ing factors is fixed under . This proves that (P (X1 , . . . , Xn )) = P (X1 , . . . , Xn ).
Therefore,

e(P (X1 , . . . , Xn )) = ( 1 ··· r )(P (X1 , . . . , Xn ))

=( 1 ··· r 1 )(P (X r (1) ,...,X r (n) ))

= ( 1 ··· r 1 )(P (X1 , . . . , Xn ))

= ( 1)r P (X1 , . . . , Xn )
= P (X1 , . . . , Xn ), as r is odd.

As e(P (X1 , . . . , Xn )) = P (X1 , . . . , Xn ), this gives a contradiction and hence completes

the proof.

Theorem 2.4.17 A permutation 2 Sn is a product of either even or odd number of

transpositions, but never both.

Proof. Suppose can be written as both. We write,

= 1 ··· m, where m is even and i ’s are transpositions for all 1  i  m,

§2.5. Cosets and Lagrange’s theorem 25

and

= ˜1 · · · ˜r , where r is odd and ˜i ’s are transpositions for all 1  i  r.

So, 1 = ˜r 1 · · · ˜1 1 . Note that the inverse of a transposition is always a transposition,

as ↵ ↵ = e for any transposition ↵. Now we write,

1
e= = 1 ··· m ˜r 1 · · · ˜1 1 .

This shows that e is a product of r + m, an odd many transpositions. A contradiction

to the fact that e can never be a product of odd many transpositions. Therefore any
permutation 2 Sn can be product of either even or odd many transpositions, but never
both.

Definition 2.4.18 If a permutation in Sn is a product of even number of transpositions,

then it is called an even permutation, otherwise it is called an odd permutation.

Define An to be the set of all even permutations of Sn . Clearly An 6= ? as e 2 An .

Exercise 2.4.19 Show that An is a subgroup of (Sn , ).

Definition 2.4.20 The set An of all even permutations of Sn form a group under the
composition of maps, which is called as the alternating group of n symbols.

Theorem 2.4.21 For n 2, |An | = n!/2.

Proof. Let ⌧ = (1 2). Clearly, if 2 An , then ⌧ 2 Sn \ An . So define : An ! Sn \ An

by ( ) := ⌧ for all 2 An . This map is injective as 1 ⌧ = 2 ⌧ gives 1 = 2 due to
cancellation law. Now is surjective, as for ↵ 2 Sn \An , we see that (↵ ⌧ ) = ↵ ⌧ ⌧ = ↵.
Hence |An | = |Sn \ An |. As Sn is disjoint union of An and Sn \ An , |An | = |S2n | = n!2 .

2.5 Cosets and Lagrange’s theorem

Let (G, ⇤) be a group and H be a subgroup of (G, ⇤). Let a 2 G. Then the set

aH := {a ⇤ h : h 2 H},

is called a left coset of H in G. The set

Ha := {h ⇤ a : h 2 H},
 26 §2.5. Cosets and Lagrange’s theorem

is called a right coset of H in G.

Example 2.5.1 Let us consider the Klein’s 4 group i.e. the set G = {e, a, b, c} with the
operation ⇤ as follows: a ⇤ a = b ⇤ b = c ⇤ c = e, a ⇤ b = b ⇤ a = c, a ⇤ c = c ⇤ a = b, b ⇤ c =
c ⇤ b = a. Let us consider the subgroup H := {e, a} of (G, ⇤). We look at all the left and
right cosets of H in (G, ⇤).
The below are the left cosets of H in (G, ⇤):

eH = {e, a} = H,
aH = {a, a ⇤ a} = {e, a} = H,
bH = {b, ba} = {b, c},
cH = {c, ca} = {c, b}.

The below are the right cosets of H in (G, ⇤):

He = {e, a} = H,
Ha = {a, a ⇤ a} = {e, a} = H,
Hb = {b, a ⇤ b} = {b, c},
Hc = {c, a ⇤ c} = {c, b}.

We also note that here the number of distinct left and right cosets of H in (G, ⇤) are
same.

Exercise 2.5.2 Consider the subgroup H := {e, (1 2)} of the symmetric group with 3
symbols S3 . Find out all the distinct left and right cosets of H in S3 .

Exercise 2.5.3 Consider the subgroup H := 3Z of the group (Z, +). Find out all the
distinct left and right cosets of H in (Z, +).

Remark 2.5.4 Note that a subgroup H is both left as well as right coset of H in a
group (G, ⇤) as eH = H = He. So the collection of all left cosets of H in (G, ⇤) and the
collection of all right cosets of H in (G, ⇤) are non-empty.

Theorem 2.5.5 Let (G, ⇤) be a group and H be a subgroup of (G, ⇤). Then the follow-
ing statements hold:

1) aH = H if and only if a 2 H.
§2.5. Cosets and Lagrange’s theorem 27

2) Ha = H if and only if a 2 H.

3) aH = bH if and only if a 1
⇤ b 2 H.

4) Ha = Hb if and only if a ⇤ b 1
2 H.

5) Either aH = bH or aH \ bH = ?.

6) Either Ha = Hb or Ha \ Hb = ?.
Proof. This can be proved as follows:
1) First let aH = H. So a = a ⇤ e 2 aH, as e 2 H. This implies that a 2 H. Next
let a 2 H. So aH ✓ H. Now for any h 2 H, we can write h = a ⇤ (a 1 ⇤ h). This
proves that h 2 aH as a 1 ⇤ h 2 H, and hence H ✓ aH. Therefore finally we get
aH = H.

2) First let Ha = H. So a = e ⇤ a 2 Ha, as e 2 H. This implies that a 2 H. Next

let a 2 H. So Ha ✓ H. Now for any h 2 H, we can write h = (h ⇤ a 1 ) ⇤ a. This
proves that h 2 Ha as h ⇤ a 1 2 H, and hence H ✓ Ha. Therefore finally we get
Ha = H.

3) First let aH = bH. So H = a 1 bH, and now using 1), we get, a 1 b 2 H. Next let
a 1 b 2 H. So a 1 bH = H. This implies, aH = bH.

4) First let Ha = Hb. So H = H(a ⇤ b 1 ), and now using 2), we get, a ⇤ b 1

2 H.
Next let a ⇤ b 1 2 H. So H(a ⇤ b 1 ) = H. This implies, Ha = Hb.

5) Let aH \ bH 6= ?. We show that aH = bH. Let x 2 aH \ bH, so x = a ⇤ h1

and also x = b ⇤ h2 for some h1 , h2 2 H. Therefore a ⇤ h1 = b ⇤ h2 and hence
a 1 ⇤ b = h1 ⇤ h2 1 2 H. This proves that aH = bH.

6) Let Ha \ Hb 6= ?. We show that Ha = Hb. Let x 2 Ha \ Hb, so x = h1 ⇤ a

and also x = h2 ⇤ b for some h1 , h2 2 H. Therefore h1 ⇤ a = h2 ⇤ b and hence
a ⇤ b 1 = h1 1 ⇤ h2 2 H. This proves that Ha = Hb.

Remark 2.5.6 Let H be a subgroup of (G, ⇤). From the above theorem we can conclude
that collections of all distinct left (resp. right) cosets of H forms a partition of G.

Exercise 2.5.7 Let H be a subgroup of (G, ⇤). For any two elements a, b 2 G, we define
aRb if and only if a 1 ⇤ b 2 H. Show that R is an equivalence relation on G and the
equivalence classes of G with respect to R is the collections of all distinct left cosets of
H in G
 28 §2.5. Cosets and Lagrange’s theorem

Theorem 2.5.8 Let H be a subgroup of (G, ⇤) and a 2 G. Then

|aH| = |H| = |Ha|.

Proof. To show that |aH| = |H|, we need to show that there is a bijection from H to
aH. We define, f : H ! aH, by f (h) = a ⇤ h for all h 2 H. First we show that f is
injective. Let f (h1 ) = f (h2 ) i.e. a ⇤ h1 = a ⇤ h2 . This implies that h1 = a 1 ⇤ (a ⇤ h1 ) =
a 1 ⇤ (a ⇤ h2 ) = h2 . Now we show that f is surjective. Let b 2 aH. Then we consider
h 2 H such that b = a ⇤ h. Clearly f (h) = a ⇤ h. This proves that |aH| = |H|.
To show |H| = |Ha|, define g : H ! Ha, by g(h) = h ⇤ a for all h 2 H and then
similarly we get g is a bijection. Thus |H| = |Ha|.
Therefore |aH| = |H| = |Ha|.

Theorem 2.5.9 Let H be a subgroup of (G, ⇤) and L (resp. R) be the set of all left
(resp. right) cosets of the subgroup H in G. Then |L| = |R|.

Proof. We need to show that there is a bijection from L to R. We define, : L ! R as

follows: for aH 2 L set (aH) := Ha 1 . First we need to show that is well-defined.
Let aH = bH for some a, b 2 G. So a 1 ⇤ b 2 H. This implies that H(a 1 ⇤ b) = H
i.e. Ha 1 = Hb 1 and hence is well-defined. Now we show that is injective. Let
(aH) = (bH). So Ha 1 = Hb 1 and therefore a 1 ⇤ b 2 H. This implies that
aH = bH. Thus is injective. Now we show that is surjective. Let Ha 2 R. Consider
a 1 H 2 L. Note that (a 1 H) = Ha. Thus is surjective. This completes the proof.

Definition 2.5.10 Let H be a subgroup of a group (G, ⇤). Then the index of H in
G is defined as the number of distinct left or right cosets of H in (G, ⇤). We denote the
index of H in G by [G : H].

Theorem 2.5.11 (Lagrange’s theorem) Let H be a subgroup of a finite group (G, ⇤).
Then |H| divides |G|. In fact,
|G|
[G : H] = .
|H|

Proof. Since (G, ⇤) is a finite group we have [G : H] is finite. Let x1 H, . . . , xr H be all

the distinct left cosets of H in (G, ⇤). Now as they form a partition of (G, ⇤), we have
[
G= xi H.
1ir
§2.5. Cosets and Lagrange’s theorem 29

Therefore,
X
|G| = |xi H|.
1ir

Also we know that |xi H| = |H| for all 1  i  r. This implies that

|G| = r|H|.

This proves that |H| divides |G|. Also note that

|G|
[G : H] = r = .
|H|

2.5.1 An application of Lagrange’s theorem

Theorem 2.5.12 Let H, K be two finite subgroups of a group (G, ⇤). Then

|H||K|
|HK| = T .
|H K|
T
Proof. Write A = H K. Note that A is a subgroup of (H, ⇤). We have, H is a finite
group. Let x1 A, . . . , xr A be all the distinct left cosets of A in H. Therefore, |H| = r|A|.
Now AK ✓ K and K ✓ AK as e 2 A. So K = AK. Then,
!
[ [
HK = xi A K = xi K.
1ir 1ir

T T
We show that xi K xj K = ? for all i 6= j. If xi K xj K 6= ?, then xi K = xj K i.e.
xi 1 ⇤ xj 2 K. Also xi 1 ⇤ xj 2 H as xi , xj 2 H. This implies that xi 1 ⇤ xj 2 A and hence
xj A = xj A, a contradiction. Therefore,
X
|HK| = |xi K| = r|K|.
1ir

So finally we have
|H||K| |H||K|
|HK| = = T .
|A| |H K|

Remark 2.5.13 From the above theorem we get that for any finite subgroups H, K of
 30 §2.6. Normal subgroups

(G, ⇤), |HK| = |KH|.

Question 2.5.14 Does the converse of Lagrange’s theorem hold i.e. if (G, ⇤) is a finite
group and if m | |G|, then does there exists a subgroup of (G, ⇤) of order m?

We know that if (G, ⇤) is cyclic then for every divisor m of |G|, there exists an unique
subgroup of (G, ⇤) of order m. But it is not true for all finite groups. For example
consider the alternating group of 4 symbols A4 . We have |S4 | = 24. So |A4 | = 12.
Consider all the 3-cycles of S4 :

(1 2 3), (1 3 2), (1 2 4), (1 4 2), (2 3 4), (2 4 3), (1 4 3), (1 3 4).

They are all even, so all the 3-cycles belong to A4 . We see that the number of 3-cycles in
S4 is 8. Now 6 | |A4 |. We show that A4 has no subgroup of order 6. If there is a subgroup
H of A4 of order 6, then there exists a 3-cycle = (a b c) 2 S4 such that 2 / H. We
know that o( ) = 3, so K := {e, , } is a subgroup of A4 . Note that
2 2
= 12 / H.
Thus we get,
|H||K|
|HK| = T = 18.
|H K|
But HK ✓ A4 and |A4 | = 12. This leads us to a contradiction. So there is no subgroup
of A4 of order 6.

2.6 Normal subgroups

Definition 2.6.1 Let H be a subgroup of a group (G, ⇤). We say that H is a normal
subgroup of G if gH = Hg for all g 2 G i.e. if every left coset is a right coset and
vice-versa.

Remarks 2.6.2 Note that

1) The trivial subgroups i.e. {e} and G are always normal in (G, ⇤).

2) In general, gH = Hg for some g 2 G does not imply that gh = hg for all h 2 H.

Let G = S3 and H = {e, (1 2 3), (1 3 2)}. We can check that for all 2 S3 ,
H = H . But
(2 3)(1 2 3) 6= (1 2 3)(2 3).

Theorem 2.6.3 Let H be a subgroup of a group (G, ⇤). Then the following are equiv-
alent:
§2.6. Normal subgroups 31

1) H is a normal subgroup of (G, ⇤).

2) gHg 1
✓ H for all g 2 G.

3) gHg 1
= H for all g 2 G.

Proof. First we assume that H is a normal subgroup of (G, ⇤). Since H is a normal
subgroup of (G, ⇤), by definition we have gH = Hg for all g 2 G. Therefore gHg 1 ✓ H
for all g 2 G.
Next we assume that gHg 1 ✓ H for all g 2 G. We show that H ✓ gHg 1 for all
g 2 G. Let h 2 H. Now h = g ⇤ (g 1 ⇤ h ⇤ g) ⇤ g 1 2 gHg 1 as g 1 ⇤ h ⇤ g 2 H. This
shows that H ✓ gHg 1 for all g 2 G and hence gHg 1 = H for all g 2 G.
Now assume that gHg 1 = H for all g 2 G. Then gH = Hg for all. g 2 G. So H is
a normal subgroup of (G, ⇤).

Theorem 2.6.4 The center Z(G) of a group (G, ⇤) is a normal subgroup.

Proof. We have seen that Z(G) is a subgroup of (G, ⇤). Now we show that gZ(G)g 1 ✓
Z(G) for all g 2 G. Let g 2 G and a 2 Z(G). Now a 2 Z(G) implies that a ⇤ g = g ⇤ a,
so g ⇤ a ⇤ g 1 = a ⇤ g ⇤ g 1 = a 2 Z(G). This is true for all a 2 Z(G), hence gZ(G)g 1 ✓
Z(G).

Theorem 2.6.5 Let H be a subgroup of (G, ⇤) such that [G : H] = 2. Then H is

normal in (G, ⇤).

Proof. We are given that H has two distinct left cosets and two distinct right cosets
in (G, ⇤). We want to show that gH = Hg for all g 2 G. Let g 2 H, then we have
gH = H = Hg. So let g 2/ H. Then gH \ H = ? and Hg \ H = ?. We can then write
for any g 2
/ H,
[ [
G=H gH and G = H Hg.

Therefore,
gH = G \ H = Hg for all g 2
/ G.

This prove that H is a normal subgroup of (G, ⇤).

Example 2.6.6 The alternating group with n symbols An is a normal subgroup of the
symmetric group with n symbols Sn . We know |An | = n!2 = |S2n | . So [Sn : An ] = 2 and
hence An is a normal subgroup of Sn .

Theorem 2.6.7 Let H, K be two subgroups of a group (G, ⇤). Then,

 32 §2.6. Normal subgroups

1) If H is normal then HK = KH, and hence HK is a subgroup of (G, ⇤).

2) If H, K both are normal then HK(= KH) is a normal subgroup of (G, ⇤).
T
3) If H, K both are normal then H K is a normal subgroup of (G, ⇤).

Proof. This can be proved as follows:

1) Since H is normal for any k 2 K, we have kH = Hk, so

[ [
HK = Hk = kH = KH.
k2K k2K

Therefore HK is a subgroup of (G, ⇤).

2) As H, K both are normal we have, gHg 1

= H and gKg 1
= K for all g 2 G. For
any g 2 G,

1 1 1
gHKg = gH(g ⇤ g)Kg = (gHg 1 )(gKg 1 ) = HK.

This proves that HK(= KH) is a normal subgroup of (G, ⇤).

3) Let a 2 H \ K and g 2 G. Now g ⇤ a ⇤ g 1 2 H and also g ⇤ a ⇤ g 1 2 K as H, K

both are normal subgroups of (G, ⇤). Therefore g ⇤ a ⇤ g 1 2 H \ K, and hence
g(H \ K)g 1 ✓ HK. This is true for all g 2 G, therefore the proof.

Remarks 2.6.8 Note that

1) If HK = KH is a subgroup of (G, ⇤), that does not imply that at least one of
H, K is a normal subgroup. Consider the following:

G := S4 , H := {e, (1 2)}, and K = {e, (3 4)}.

Now we note that H, K are not normal in S4 , as

/ H, and (1 3)(3 4)(1 3) = (1 4) 2

(1 3)(1 2)(1 3) = (2 3) 2 / K.

But HK = {e, (1 2), (3 4), (1 2)(3 4)} = KH is a subgroup of S4 .

2) If HK = KH is a normal subgroup of (G, ⇤), that does not imply that both of
H, K are normal subgroups. Consider the following:

G := S5 , H := h(1 2 3 4 5)i, and K = A4 .

§2.7. Quotient groups 33

So we have two subgroups H, K of S5 . We show that A5 = h(1 2 3 4 5)iA4 . Note

that H \ K = {e}. Therefore,

|H||K| 5 · 12
|HK| = = = 60 = |A5 |.
|H \ K| 1

As HK ✓ A5 and |HK| = |A5 | < 1, we conclude that A5 = h(1 2 3 4 5)iA4 . Now

A5 is normal in S5 , so h(1 2 3 4 5)iA4 is normal in S5 . We show that H is not
normal in S5 . We have,

H = {e, (1 2 3 4 5), (1 3 5 2 4), (1 4 2 5 3), (1 5 4 3 2)},

and
(1 2)(1 2 3 4 5)(1 2) = (1 3 4 5 2) 2
/ H.

This shows that H is not normal in S5 .

3) If H \ K is a normal subgroup of (G, ⇤), that does not imply that both of H, K
are normal subgroups. Consider the following:

G := S3 , H := {e, (1 2)}, and K = {e, (1 3)}.

We have already seen that HK 6= KH, and hence H, K both are not normal
subgroups of S3 . But H \ K = {e} is a normal subgroup of S3 .

2.7 Quotient groups

Let H be a normal subgroup of a group (G, ⇤). So each left coset of H is also a right
coset of H in (G, ⇤) and vice-versa. Therefore in this case we refer to them as just cosets.
Let G/H denote the set of all distinct cosets of H in (G, ⇤). We can induce an operation
on G/H from the operation on G. Define ⇤ on G/H as follows: for any aH, bH 2 G/H,

aH ⇤ bH = (a ⇤ b)H.

First we show that this operation is well-defined. Let a1 H = a2 H and b1 H = b2 H i.e.

a1 1 ⇤ a2 2 H and b1 1 ⇤ b2 2 H. We shall show that (a1 ⇤ b1 )H = (a2 ⇤ b2 )H. Now,

1
(a1 ⇤ b1 ) ⇤ (a2 ⇤ b2 ) = (b1 1 ⇤ a1 1 ) ⇤ (a2 ⇤ b2 )
= b1 1 ⇤ (a1 1 ⇤ a2 ) ⇤ b2
 34 §2.7. Quotient groups

= (b1 1 ⇤ (a1 1 ⇤ a2 ) ⇤ b1 ) ⇤ (b1 1 ⇤ b2 ).

This shows that (a1 ⇤ b1 ) 1

⇤ (a2 ⇤ b2 ) 2 H as H is normal in (G, ⇤). Hence,

(a1 ⇤ b1 )H = (a2 ⇤ b2 )H.

We note that the operation ⇤ is associative in G/H. Let aH, bH, cH 2 G/H. We have,

(aH ⇤ bH) ⇤ cH = (a ⇤ b)H ⇤ cH = (a ⇤ b ⇤ c)H = aH ⇤ (b ⇤ c)H = aH ⇤ (bH ⇤ cH),

as ⇤ is associative in G. Also note that H is the identity of G/H as aH⇤H = aH = H⇤Ha

for any a 2 G. We now show that for aH 2 G/H, we have an inverse element in G/H.
As a 2 G, we have a 1 2 G. Consider a 1 H 2 G/H. Now,

aH ⇤ a 1 H = (a ⇤ a 1 )H = H = (a 1
⇤ a)H = a 1 H ⇤ aH.

So we see that (G/H, ⇤) forms a group. This group (G/H, ⇤) is called the quotient
group of G by H.

Remarks 2.7.1 Note that

1) If (G, ⇤) is abelian, then (G/H, ⇤) is abelian as for any aH, bH 2 G/H, aH ⇤ bH =

(a ⇤ b)H = (b ⇤ a)H = bH ⇤ aH.

2) If (G, ⇤) is cyclic, then (G/H, ⇤) is cyclic as for any aH 2 G/H,

aH = (g m )H = (gH)m , where G := hgi and a = g m for some m 2 Z.

So G/H = hgHi.

3) If (G/H, ⇤) is abelian, then (G, ⇤) need not be abelian. Note that Sn /An is a group
of order 2, and hence abelian but Sn is non-abelian.

4) If (G/H, ⇤) is cyclic, then (G, ⇤) need not be cyclic. Note that Sn /An is a group of
order 2, and hence cyclic but Sn is not cyclic.

Example 2.7.2 Let (G, ⇤) = (Z, +). Consider the subgroup 3Z in (Z, +). As (Z, +)
is abelian, 3Z is a normal subgroup. Note that Z/3Z = {3Z, 1 + 3Z, 2 + 3Z}. It is
interesting to note that this is the group of distinct congruence classes modulo 3.
§2.8. Homomorphism of groups 35

2.8 Homomorphism of groups

Let (G1 , ⇤1 ) and (G2 , ⇤2 ) be two groups. We would like to see the interplay between
them. A map f : G1 ! G2 is called a homomorphism if f (a ⇤1 b) = f (a) ⇤2 f (b) for all
a, b 2 G1 .

Remark 2.8.1 Between (G1 , ⇤1 ) and (G2 , ⇤2 ), there is always at least one homomor-
phism as the map f : G1 ! G2 , defined by f (a) = e2 for all a 2 G1 , is a homomorphism
(G1 , ⇤1 ) into (G2 , ⇤2 ). Here e2 denotes the identity of (G2 , ⇤2 ). This homomorphism is
called the trivial homomorphism of (G1 , ⇤1 ) into (G2 , ⇤2 ).

Examples 2.8.2 The following are examples of group homomorphisms.

1) Consider (G1 , ⇤1 ) as the group of real numbers R with usual addition and (G2 , ⇤2 )
as the group of non-zero real numbers R \ {0} with usual multiplication. Define
f : R ! R \ {0} by f (a) := exp(a). The notation exp denotes the exponential
map.

2) Consider (G1 , ⇤1 ) as the group of invertible 2 ⇥ 2 matrices with real entries, with
the operation usual matrix multiplication. This group is denoted by GL2 (R), and
is called the general linear group of order 2 with real entries. The name
linear comes from the fact that GL2 (R) is essentially the set of all bijective linear
transformations from the vector space R2 (over R) to itself. Consider (G2 , ⇤2 )
as the group of non-zero real numbers R \ {0} with usual multiplication. Define
f : GL2 (R) ! R \ {0} by f (A) := det(A). The notation det(A) denotes the
determinant of the matrix A.

3) Consider (G1 , ⇤1 ) as Sn , the symmetric group with n symbols. Consider (G2 , ⇤2 )

as the group {1, 1} with usual multiplication. Define f : Sn ! {1, 1} by
8
<1 if is even,
f ( ) :=
: 1 otherwise.

Theorem 2.8.3 Let (G1 , ⇤1 ) and (G2 , ⇤2 ) be two groups and f : G1 ! G2 be a group
homomorphism. Then,

1) f (e1 ) = e2 , where e1 , e2 denote the identities of (G1 , ⇤1 ) and (G2 , ⇤2 ) respectively.

2) f (a 1 ) = f (a) 1
for all a 2 G1 .
 36 §2.8. Homomorphism of groups

3) f (an ) = f (a)n for all n 2 Z.

Proof. This can be proved as follows:

1) Note that f (e1 ) = f (e1 ⇤1 e1 ) = f (e1 ) ⇤2 f (e1 ). Therefore by cancellation we obtain,

f (e1 ) = e2 .

2) Note that e2 = f (e1 ) = f (a ⇤1 a 1 ) = f (a) ⇤2 f (a 1 ) for all a 2 G1 . Therefore,

f (a 1 ) = f (a) 1 for all a 2 G1 .

3) Let a 2 G1 . For n = 0, we have already seen that f (e1 ) = e2 . Let n 2 N. Assume

the induction hypothesis for n = k, i.e. f (ak ) = f (a)k . Now,

f (ak+1 ) = f (ak ⇤1 a) = f (ak ) ⇤2 f (a) = f (a)k+1 .

Now let n < 0. Let n = m where m > 0. Therefore,

f (an ) = f (a m
) = f (a 1 )m = f (a) m
= f (a)n .

This completes the proof.

As an application we see that f : Z ! Z defined by f (n) = n + 1 is not a group

homomorphism as f (0) 6= 0.

Exercise 2.8.4 Show that all the group homomorphisms from (Z, +) to itself are of the
form fm (n) := mn for all n 2 Z, for some m 2 Z.

Remarks 2.8.5 Let (G1 , ⇤1 ) and (G2 , ⇤2 ) be two groups and f : G1 ! G2 be a group
homomorphism. Then,

1) For a subgroup H of (G1 , ⇤1 ), f (H) is a subgroup of (G2 , ⇤2 ). Note that f (H) 6= ?

as e2 = f (e1 ) 2 f (H). Let a, b 2 f (H). So a = f (h1 ) and b = f (h2 ) for some
h1 , h2 2 H. Therefore,

1 1
a ⇤2 b = f (h1 ) ⇤2 f (h2 ) = f (h1 ⇤1 h2 1 ) 2 f (H).

This proves that f (H) is a subgroup of (G2 , ⇤2 ).

2) For a subgroup H of (G2 , ⇤2 ), the inverse image of H under f i.e. the set f 1 (H) :=
{g 2 G1 : f (g) 2 H} is a subgroup of (G1 , ⇤1 ). Note that f 1 (H) 6= ? as
e1 2 f 1 (H). Let a, b 2 f 1 (H). So f (a), f (b) 2 H. As H is a subgroup of (G2 , ⇤2 ),
§2.8. Homomorphism of groups 37

we have f (a) ⇤2 f (b) 1 2 H. So f (a ⇤1 b 1 ) 2 H. Therefore a ⇤1 b 2 f 1

(H). This
proves that f 1 (H) is a subgroup of (G1 , ⇤1 ).

3) For a normal subgroup H of (G1 , ⇤1 ), f (H) need not be normal in (G2 , ⇤2 ). Con-
sider the homomorphism f : S3 ! S3 defined by
8
<(1 2) if is a transposition,
f ( ) :=
:e otherwise.

We know S3 is normal subgroup of S3 , but {e, (1 2)} is not normal in S3 .

4) If f is surjective, for a normal subgroup H of (G1 , ⇤1 ), f (H) is normal in (G2 , ⇤2 ).

Let a 2 f (H) and g2 2 G2 . We have a = f (h) for some h 2 H. Since f is surjective
there exists g1 2 G1 such that f (g1 ) = g2 . Therefore,

g2 ⇤2 a ⇤2 g2 1 = f (g1 ⇤1 h ⇤1 g1 1 ) 2 f (H) as g1 ⇤1 h ⇤1 g1 1 2 H.

This proves that f (H) is normal in (G2 , ⇤2 ).

5) For a normal subgroup H of (G2 , ⇤2 ), f 1 (H) := {g 2 G1 : f (g) 2 H} is a normal

subgroup of (G1 , ⇤1 ). We have already seen that f 1 (H) is a subgroup of (G1 , ⇤1 ).
Now let g 2 G1 and a 2 f 1 (H). To show g ⇤1 a ⇤1 g 1 2 f 1 (H), we need to show
that f (g ⇤1 a ⇤1 g 1 ) 2 H. Note that

f (g ⇤1 a ⇤1 g 1 ) = f (g) ⇤2 f (a) ⇤2 f (g) 1

2 H as f (g) 2 G2 and f (a) 2 H.

This proves that f 1

(H) is a normal subgroup of (G1 , ⇤1 ).

6) Let a 2 G1 be such that o(a) = n. Then o(f (a)) | n. Note that f (a)n = f (an ) =
f (e1 ) = e2 . So o(f (a)) | n.

In general o(f (a)) need not be equal to n. For example, consider the group homo-
morphism f : Sn ! {1, 1} by
8
<1 if is even,
f ( ) :=
: 1 otherwise.

If is a 3-cycle then o( ) = 3 and o(f ( )) = 1.

 38 §2.8. Homomorphism of groups

7) Let a 2 G1 be such that o(a) = n and f be injective. Then o(f (a)) = n. Note that
if o(f (a)) = m then f (am ) = f (a)m = e2 = f (e1 ). So am = e1 , so n | m and hence
m = n.

Definition 2.8.6 A group (G2 , ⇤2 ) is called a homomorphic image of (G1 , ⇤1 ) if there

exists a surjective homomorphism f : G1 ! G2 .

Theorem 2.8.7 Let f be a surjective homomorphism from (G1 , ⇤1 ) onto (G2 , ⇤2 ). Then

1) If (G1 , ⇤1 ) is abelian then so is (G2 , ⇤2 ).

2) If (G1 , ⇤1 ) is cyclic then so is (G2 , ⇤2 ).

Proof. This can be proved as follows:

1) Let a, b 2 G2 . So there exist a0 , b0 2 G1 such that f (a0 ) = a and f (b0 ) = b. Now,

a ⇤2 b = f (a0 ) ⇤2 f (b0 ) = f (a0 ⇤1 b0 ) = f (b0 ⇤1 a0 ) = f (b0 ) ⇤2 f (a0 ) = b ⇤2 a.

This proves that (G2 , ⇤2 ) is abelian.

2) Let G1 := hgi and b 2 G2 . So there exists a 2 G1 be such that f (a) = b. Now

a = g m for some m 2 Z. Therefore,

b = f (g m ) = f (g)m 2 hf (g)i.

Thus G2 ✓ hf (g)i. Clearly hf (g)i ✓ G2 as f (g) 2 G2 . This proves that G2 =

hf (g)i, and hence cyclic.

Remark 2.8.8 If (G2 , ⇤2 ) is abelian, (G1 , ⇤1 ) need not be abelian. Also if (G2 , ⇤2 ) is
cyclic, (G1 , ⇤1 ) need not be cyclic. For example, consider the homomorphism f : Sn !
{1, 1} defined as 8
<1 if is even,
f ( ) :=
: 1 otherwise.

Here {1, 1} is abelian and cyclic with usual multiplication, but Sn is neither abelian
nor cyclic.

Definition 2.8.9 Let (G1 , ⇤1 ) and (G2 , ⇤2 ) be two groups and f : G1 ! G2 be a group
homomorphism. Then the set

{g 2 G1 : f (g) = e2 },
§2.8. Homomorphism of groups 39

where e2 is the identity of G2 , is called the kernel of f and is denoted by Kerf .

Definition 2.8.10 Let (G1 , ⇤1 ) and (G2 , ⇤2 ) be two groups and f : G1 ! G2 be a group
homomorphism. Then the set
{f (g) : g 2 G1 }

is called the image of f and is denoted by Imf .

Remark 2.8.11 Both the sets Kerf and Imf are non-empty as e1 2 Kerf and e2 2 Imf .

Exercise 2.8.12 Show that Kerf is a normal subgroup of G1 .

Exercise 2.8.13 Show that Imf is a subgroup of G2 .

Remark 2.8.14 The subgroup Imf of (G2 , ⇤2 ) need not be normal. Consider the ho-
momorphism f : S3 ! S3 defined by
8
<(1 2) if is a transposition,
f ( ) :=
:e otherwise.

Here Imf = {e, (1 2)} is not normal in S3 .

Theorem 2.8.15 Let (G1 , ⇤1 ) and (G2 , ⇤2 ) be two groups and f : G1 ! G2 be a group
homomorphism. Then f is injective if and only if Kerf = {e1 }.

Proof. First assume that Kerf = {e1 }. We show that f is injective. Let f (a) = f (b) for
some a, b 2 G1 . Then we have f (a) ⇤2 f (b) 1 = e2 i.e. f (a ⇤1 b 1 ) = e2 as f is a group
homomorphism. This shows that a ⇤1 b 1 2 Kerf , and hence a ⇤1 b 1 = e1 i.e. a = b.
Next we assume that f is injective. We show that Kerf = {e1 }. Let a 2 Kerf . So
f (a) = e2 = f (e1 ). As f is injective we get a = e1 . This proves that Kerf = {e1 }.

Definition 2.8.16 Let (G1 , ⇤1 ) and (G2 , ⇤2 ) be two groups and f : G1 ! G2 be a group
homomorphism. Then f is called an isomorphism of (G1 , ⇤1 ) and (G2 , ⇤2 ) if the map
f is bijective.

Theorem 2.8.17 Let f : G1 ! G2 be an isomorphism of (G1 , ⇤1 ) and (G2 , ⇤2 ). Then

we have, f 1 : G2 ! G1 is an isomorphism of (G2 , ⇤2 ) and (G1 , ⇤1 ).
 40 §2.8. Homomorphism of groups

Proof. Since f is bijective, the inverse map of f , denoted by f 1

exists. We write the
map f 1 : G2 ! G1 below:

f 1
(a) := a0 for all a 2 G2 , where f (a0 ) = a.

Let a, b 2 G2 . So there exist unique a0 , b0 2 G1 such that f (a0 ) = a and f (b0 ) = b. Now
a ⇤2 b = f (a0 ) ⇤2 f (b0 ) = f (a0 ⇤1 b0 ) as f is a homomorphism. So we have,

f 1
(a ⇤2 b) = a0 ⇤1 b0 = f 1
(a) ⇤1 f 1
(b).

This shows that f 1 is a homomorphism from (G2 , ⇤2 ) to (G1 , ⇤1 ). Now f 1

is bijective
as f is bijective. This completes the proof.

Remarks 2.8.18 If there is an isomorphism between (G1 , ⇤1 ) and (G2 , ⇤2 ), then we say
that the groups (G1 , ⇤1 ) and (G2 , ⇤2 ) are isomorphic. In this case we denote G1 ' G2 .

1) If the groups (G1 , ⇤1 ) and (G2 , ⇤2 ) are isomorphic, then (G1 , ⇤1 ) is abelian if and
only if (G2 , ⇤2 ) is abelian. We have already seen that if the group (G2 , ⇤2 ) is a
homomorphic image of the group (G1 , ⇤1 ) with (G1 , ⇤1 ) abelian, then (G2 , ⇤2 ) is
abelian. In this case the groups (G1 , ⇤1 ) and (G2 , ⇤2 ) are homomorphic images of
each other, and hence the remark follows.

2) If the groups (G1 , ⇤1 ) and (G2 , ⇤2 ) are isomorphic, then (G1 , ⇤1 ) is cyclic if and
only if (G2 , ⇤2 ) is cyclic. We have already seen that if the group (G2 , ⇤2 ) is a
homomorphic image of the group (G1 , ⇤1 ) with (G1 , ⇤1 ) cyclic, then (G2 , ⇤2 ) is
cyclic. In this case the groups (G1 , ⇤1 ) and (G2 , ⇤2 ) are homomorphic images of
each other, and hence the remark follows.

3) If f is an isomorphism of (G1 , ⇤1 ) and (G2 , ⇤2 ), then we have o(a) = o(f (a)) for all
a 2 G1 . We have already seen that o(a) = o(f (a)) for all a 2 G1 if f is injective.
Hence it is a particular case.

Exercise 2.8.19 Let (G1 , ⇤1 ), (G2 , ⇤2 ) and (G3 , ⇤3 ) be three groups. Then show that,

1) Gi ' Gi for all 1  i  3.

2) G1 ' G2 implies that G2 ' G1 .

3) If G1 ' G2 and G2 ' G3 then G1 ' G3 .

§2.8. Homomorphism of groups 41

Remark 2.8.20 The above exercise shows us that ' is an equivalence relation on the
collection of all groups. For a group (G, ⇤), by the isomorphic class of (G, ⇤) we mean
the collection of all groups which are isomorphic to (G, ⇤).

Example 2.8.21 The groups (R, +) and (R \ {0}, ·) are not in the same isomorphic
class. There is no element in (R, +) of finite order but 1 in (R \ {0}, ·) is of order 2.

Remarks 2.8.22 We observe the following:

• A finite group can never be isomorphic to an infinite group.

• A cyclic group can never be isomorphic to a non-cyclic group.

• An abelian group can never be isomorphic to a non-abelian group.

• Two groups can be of same order, but they may not be isomorphic. For example,
we know Z/6Z is abelian group of order 6 and S3 is non-abelian group of order 6.

Theorem 2.8.23 Any infinite cyclic group is isomorphic to the group (Z, +).

Proof. Let (G, ⇤) be an infinite cyclic group. Let a be a generator of (G, ⇤). Clearly a 6= e.
Let us define, f : Z ! G by f (n) := an . Now f (n + m) = an+m = an ⇤ am = f (n) ⇤ f (m).
So f is a homomorphism from (Z, +) to (G, ⇤). Now as G is infinite, a can not be of
finite order. So by convention, an = e if and only if n = 0. Therefore Kerf = {0}
and hence f is injective. Now let b 2 G. So b = ar for some r 2 Z. We can see that
f (r) = ar = b. Thus f is surjective. This proves that G ' Z.

Theorem 2.8.24 Any finite cyclic group of order n 2 N is isomorphic to the group
(Z/nZ, +).

Proof. Let (G, ⇤) be a finite cyclic group of order n 2 N. Let a be a generator of (G, ⇤).
So an = e. Let us define, f : G ! Z/nZ by f (am ) := [m]. First we show that f
is well-defined. Let as = at . So as t = e. So n | (s t). Therefore [s] = [t] i.e.
f (as ) = f (at ). Now f (am ⇤ ar ) = f (am+r ) = [m + r] = [m] + [r] = f (am ) + f (ar ). So f
is a homomorphism. Note that Kerf = {g 2 G : f (g) = [0]} = {am 2 G : [m] = 0} =
{am 2 G : n | m} = {e}, as am = e for all m such that n | m. So f is injective. For
[r] 2 Z/nZ, we can consider ar 2 G. We see that f (ar ) = [r] and hence f is surjective.
This proves that G ' Z/nZ.

Remark 2.8.25 We conclude that up to isomorphism there is only one cyclic group of
order n 2 N.
 42 §2.8. Homomorphism of groups

Theorem 2.8.26 (First isomorphism theorem) Let f : G1 ! G2 be a homomor-

phism of the groups (G1 , ⇤1 ) and (G2 , ⇤2 ). Then,

G1 /Kerf ' Imf.

Proof. We know that Kerf is a normal subgroup of the group (G1 , ⇤1 ). So we can
consider the quotient group (G1 /Kerf, ⇤1 ) where the operation ⇤1 on (G1 /Kerf, ⇤1 ) is
induced from the operation ⇤1 of (G1 , ⇤1 ).
We define f˜ : G1 /Kerf ! Imf by f˜(aKerf ) := f (a) for all aKerf 2 G1 /Kerf . We
need to show that the map f˜ is well-defined. Note that,

aKerf = bKerf
() a 1 b 2 Kerf
() f (a 1 b) = e2
() f (a) = f (b)
() f˜(aKerf ) = f˜(bKerf ).

This proves that the map f˜ is well-defined and also injective. Let g 2 Imf . So there
exists a 2 G1 such that f (a) = g. Consider aKerf 2 G1 /Kerf . We see that f˜(aKerf ) =
f (a) = g. This proves that f˜ is surjective.
Now we show that f˜ is a group homomorphism of (G1 /Kerf, ⇤1 ) onto (Imf, ⇤2 ). Let
aKerf, bKerf 2 G1 /Kerf . Now,

f˜(aKerf ⇤1 bKerf ) = f˜((a ⇤1 b)Kerf )

= f (a ⇤1 b) = f (a) ⇤2 f (b), as f is a homomorphism.
= f˜(aKerf ) ⇤2 f˜(bKerf ).

Therefore we conclude that f˜ is an isomorphism of the groups (G1 /Kerf, ⇤1 ) and (Imf, ⇤2 ).
Thus G1 /Kerf ' Imf .

Theorem 2.8.27 (Cayley) Every group (G, ⇤) is isomorphic to some subgroup of the
permutation group of the set G.

Proof. Let us denote the permutation group of the set G by A(G). We shall define a
homomorphism from the group (G, ⇤) to the group (A(G), ).
For that let a 2 G, define a : G ! G by a (g) := a ⇤ g for all g 2 G. First we show
that a 2 A(G). Let b 2 G. Consider a 1 ⇤ b 2 G, note that a (a 1 ⇤ b) = a ⇤ (a 1 ⇤ b) =
§2.8. Homomorphism of groups 43

b. So the map a is onto. Now let a (g1 ) = a (g2 ), so a ⇤ g1 = a ⇤ g2 . Therefore

g1 = a 1 ⇤ (a ⇤ g1 ) = a 1 ⇤ (a ⇤ g2 ) = g2 . This proves that a is injective and hence
a 2 A(G). Now we define : G ! A(G) by (g) := g for all g 2 G. Let g1 , g2 2 G.
We show that (g1 ⇤ g2 ) = (g1 ) (g2 ). Note that for a 2 G,

g1 ⇤g2 (a) = (g1 ⇤ g2 ) ⇤ a = g1 ⇤ (g2 ⇤ a) = g1 ⇤ g2 (a) = g1 ( g2 (a)) = g1 g2 (a).

This proves that the functions g1 ⇤g2 and g1 g2 are same, and hence is a group
homomorphism. Now,

Ker := {g 2 G : (g) = eA(G) }

= {g 2 G : g = eA(G) }
= {g 2 G : g ⇤ a = a for all a 2 G}
= {eG }.

Here eG , eA(G) denote the identities of the group (G, ⇤) and (A(G), ), respectively. There-
fore by first isomorphism theorem,

G ' G/Ker ' Im .

We know that Im is a subgroup of (A(G), ), hence the proof.

Corollary 2.8.28 If (G, ⇤) is a finite group of order n 2 N, then the group (G, ⇤) is
isomorphic to some subgroup of Sn .

Theorem 2.8.29 (Second isomorphism theorem) Let H, K be two subgroups of a

group (G, ⇤) with K normal in (G, ⇤). Then,

H/ (H \ K) ' HK/K.

Proof. We define, : H ! HK/K by (h) := hK for all h 2 H. First we show that

is a homomorphism. Let h1 , h2 2 H. Note that,

(h1 ⇤ h2 ) = (h1 ⇤ h2 )K = h1 K ⇤ h2 K = (h1 ) ⇤ (h2 ).

So is a homomorphism. Let xK 2 HK/K. Now there exist h 2 H, k 2 K such that

x = hk. So xK = hkK = hK. Consider h 2 H, so (h) = hK = xK. So is onto i.e.
 44 §2.8. Homomorphism of groups

Im = HK/K. Now,

Ker = {h 2 H : (h) = K} = {h 2 H : hK = K} = {h 2 H : h 2 K} = H \ K.

Therefore by first isomorphism theorem,

H/(H \ K) ' HK/K.

Remark 2.8.30 We know that even if one of H, K is not normal in (G, ⇤), then H \ K
need not be normal in (G, ⇤). For example, take K = G and H a subgroup which is not
normal in (G, ⇤). Then H \ K = H is not normal in (G, ⇤). But here we see that H \ K
is normal in H whenever K is normal in (G, ⇤).

Theorem 2.8.31 (Third isomorphism theorem) Let H, K be two normal subgroups

of a group (G, ⇤) with H ✓ K. Then,
.
(G/H) (K/H) ' G/K.

Proof. We define, : G/H ! G/K by (gH) := gK for all g 2 G. First we show that
is well-defined. Let g1 H = g2 H. Then g1 1 g2 2 H ✓ K. So g1 K = g2 K, this proves
that is well-defined. Now we show that is a homomorphism. Let g1 H, g2 H 2 G/H.
Now,

(g1 H ⇤ g2 H) = (g1 ⇤ g2 H) = g1 ⇤ g2 K = g1 K ⇤ g2 K = (g1 H) ⇤ (g2 H).

Therefore is a homomorphism. For gK 2 G/K, consider gH 2 G/H, so (gH) = gK,

and hence is onto. Note that,

Ker = {gH 2 G/H : (gH) = K} = {gH 2 G/H : gK = K} = {gH 2 G/H : g 2 K} = K/H.

Thus by first isomorphism theorem,

.
(G/H) (K/H) ' G/K.

Now we try to understand the connection between the subgroups of (G, ⇤) and the
subgroups of (G/H, ⇤) where H is normal in (G, ⇤).

Theorem 2.8.32 Let H be a normal subgroup of (G, ⇤). If K is a subgroup of (G, ⇤)

§2.8. Homomorphism of groups 45

with H ✓ K, then K/H := {kH : k 2 K} is a subgroup of (G/H, ⇤).

Proof. Define : G 7! G/H by (g) := gH for all g 2 G. Note that for g1 , g2 2 G,

(g1 ⇤ g2 ) = (g1 ⇤ g2 )H = g1 H ⇤ g2 H = (g1 ) ⇤ (g2 ), so is a homomorphism. We know
that under a homomorphism, image of a subgroup of the domain is again a subgroup of
the range. So for a subgroup K of (G, ⇤), the image of K under , i.e. (K) = {kH : k 2
K} is a subgroup of (G/H, ⇤). As H ✓ K, we can write (K) = {kH : k 2 K} = K/H,
and hence the proof.

Theorem 2.8.33 Let H be a normal subgroup of (G, ⇤). Suppose T is a subgroup of

(G/H, ⇤), then there exists a subgroup K of (G, ⇤) with H ✓ K such that T = K/H.

Proof. Define the set K := {g 2 G : gH 2 T } ✓ G. We show that K is a subgroup of

(G, ⇤). Note that e 2 K as H 2 T . So K 6= ?. Let a, b 2 K. Therefore aH, bH 2 T .
As T is a subgroup of (G/H, ⇤), we have (bH) 1 = b 1 H 2 T . Thus a ⇤ b 1 H 2 T .
Therefore a ⇤ b 1 2 K. This proves that K is a subgroup of (G, ⇤). Clearly H ✓ K as
for all h 2 H, hH = H 2 T . Now from the definition of K we can note that T = K/H.
This completes the proof.

Theorem 2.8.34 Let H be a normal subgroup of (G, ⇤). The subgroups of (G, ⇤)
containing H and the subgroups of (G/H, ⇤) are in bijective correspondence.

Proof. Let A denote the set of all subgroups of (G, ⇤) that contain H and B denote the
set of all subgroups of (G/H, ⇤). We show that there is a bijection from the set A to the
set B. We have seen that for a subgroup K of (G, ⇤) containing H, K/H is a subgroup
of (G/H, ⇤). So we can define f : A ! B by f (K) = K/H for all K 2 A. Also we have
seen that for any subgroup T of (G/H, ⇤), there exists a subgroup K of (G, ⇤) such that
H ✓ K and T = K/H. So the map f is onto. It only remains to show that f is injective.
Let f (K1 ) = f (K2 ) i.e. K1 /H = K2 /H. We show that K1 = K2 . Let k1 2 K1 . So
k1 H 2 K1 /H = K2 /H. This implies that k1 H = k2 H for some k2 2 K2 . Thus we have,
k1 1 ⇤ k2 2 H ✓ K2 . Hence k1 1 = (k1 1 ⇤ k2 ) ⇤ k2 1 2 K2 . Therefore k1 2 K2 as K2 is
a subgroup of (G, ⇤). This proves that K1 ✓ K2 . Similarly we can show that K2 ✓ K1
and hence K1 = K2 . This proves that the sets A and B are in bijective correspondence.

Example 2.8.35 As an application we can find out all the subgroups of (Z/12Z, +).
We know that all the subgroups of (Z/12Z, +) are of the form H/12Z where 12Z ✓ H
and H is a subgroup of (Z, +). We know that (Z, +) is cyclic, so all the subgroups of
(Z, +) are of the form (mZ, +) for m 2 Z. Note that mZ = mZ. So we can restrict
 46 §2.9. Direct product of groups

ourselves to m 2 N [ {0}. Now 12Z ✓ mZ if and only if m | 12. So the possibilities

for m are 1, 2, 3, 4, 6, 12. So there are total 6 subgroups of (Z/12Z, +) and they are
(Z/12Z, +), (2Z/12Z, +), (3Z/12Z, +), (4Z/12Z, +), (6Z/12Z, +) and (12Z/12Z, +).
Note that (Z/12Z, +) and (12Z/12Z, +) are the trivial subgroups of (Z/12Z, +).

Theorem 2.8.36 Let H be a normal subgroup of (G, ⇤). Let K1 , K2 be two subgroups
of (G, ⇤) containing H. Then K1 ✓ K2 if and only if K1 /H ✓ K2 /H.

Proof. Let H ✓ K1 ✓ K2 . Then clearly K1 /H ✓ K2 /H. So we assume that K1 /H ✓

K2 /H. We show that K1 ✓ K2 . Let k1 2 K1 . Therefore k1 H 2 K1 /H. As K1 /H ✓
K2 /H, we have k1 H = k2 H for some k2 2 K2 . This implies that k1 1 ⇤ k2 2 H ✓ K2 . So,

k1 1 = (k1 1 ⇤ k2 ) ⇤ k2 1 2 K2 .

As K2 is a subgroup of (G, ⇤), we have k1 2 K2 . Thus K1 ✓ K2 .

Theorem 2.8.37 Let H be a normal subgroup of (G, ⇤) and K be a subgroup of (G, ⇤)

containing H. Then K is normal in (G, ⇤) if and only if K/H is normal in (G/H, ⇤).

Proof. First suppose that K is normal in (G, ⇤). We have already seen that : G ! G/H
defined by (g) := gH for all g 2 G is a surjective homomorphism. We know that under
a surjective homomorphism, image of a normal subgroup is normal. We have K is normal
in (G, ⇤), so the image (K) = K/H is normal in (G/H, ⇤).
Next suppose that K/H is normal in (G/H, ⇤). We show that K is normal in (G, ⇤).
Let g 2 G and k 2 K. So gH 2 G/H and kH 2 K/H. Now as K/H is normal in
(G/H, ⇤),

g ⇤ k ⇤ g 1 H = gH ⇤ kH ⇤ g 1 H = gH ⇤ kH ⇤ (gH) 1
2 K/H.

Thus g ⇤ k ⇤ g 1 = k 0 for some k 0 2 K. This proves that g ⇤ k ⇤ g 1

2 K and hence K is
normal in (G, ⇤).

2.9 Direct product of groups

Let (G1 , ⇤1 ) and (G2 , ⇤2 ) be two groups. We consider the Cartesian product of the sets
G1 and G2 , i.e.
G1 ⇥ G2 := {(g1 , g2 ) : g1 2 G1 and g2 2 G2 }.
§2.9. Direct product of groups 47

We define an operation ⇤ on G1 ⇥ G2 as follows: for (a, b), (c, d) 2 G1 ⇥ G2 ,

(a, b) ⇤ (c, d) = (a ⇤1 c, b ⇤2 d).

As a ⇤1 c 2 G1 and b ⇤2 d 2 G2 , we see that ⇤ is a binary operation on G1 ⇥ G2 . Now

G1 ⇥ G2 6= ? as both G1 , G2 are non-empty. The set G1 ⇥ G2 with the operation ⇤ forms
a group, which we call as the direct product of the groups (G1 , ⇤1 ) and (G2 , ⇤2 ). The
element (e1 , e2 ) is the identity of (G1 ⇥G2 , ⇤) where e1 , e2 denote the identities of (G1 , ⇤1 )
and (G2 , ⇤2 ) respectively. For each (a, b) 2 G1 ⇥ G2 , consider a 1 2 G1 and b 1 2 G2 ,
then the element (a 1 , b 1 ) is the inverse of (a, b) in (G1 ⇥ G2 , ⇤). Now one can check
that the associativity of ⇤ on G1 ⇥ G2 follows from the associativity properties of ⇤1 and
⇤2 . With these we are done showing that (G1 ⇥ G2 , ⇤) is a group. We can take n many
groups (Gi , ⇤i ) for 1  i  n and consider their direct product G1 ⇥ · · · ⇥ Gn defining ⇤
on G1 ⇥ · · · ⇥ Gn , in a similar manner.

Exercise 2.9.1 The group (G1 ⇥G2 , ⇤) is abelian if and only if both (G1 , ⇤1 ) and (G2 , ⇤2 )
are abelian.

Example 2.9.2 The direct product S3 ⇥ Z is an infinite non-abelian group.

Exercise 2.9.3 If the group (G1 ⇥ G2 , ⇤) is cyclic then both (G1 , ⇤1 ) and (G2 , ⇤2 ) are
cyclic.

Remark 2.9.4 The converse of the above exercise need not be true.

• The group (Z ⇥ Z, +) is not cyclic though (Z, +) is cyclic. Suppose (Z ⇥ Z, +) is

cyclic and let (a, b) be a generator of (Z ⇥ Z, +). Now (1, 0), (0, 1) 2 Z ⇥ Z. So
there exist m, n 2 Z such that

(1, 0) = m(a, b) and (0, 1) = n(a, b).

Now ma = 1 and mb = 0 gives that b = 0. Also we have nb = 1. This is a

contradiction. Thus (Z ⇥ Z, +) is not cyclic.

• The group (Z/4Z ⇥ Z/6Z, +) is not cyclic though (Z/4Z, +), (Z/6Z, +) both are
cyclic. If (Z/4Z ⇥ Z/6Z, +) is cyclic, then we must have an element of order
24 in it. But we can see that order of no element in (Z/4Z ⇥ Z/6Z, +) exceeds
lcm(4, 6) = 12.
 48 §2.9. Direct product of groups

At this point we have the following theorem which is equivalent to the well-known
Chinese remainder theorem.

Theorem 2.9.5 The group Z/mZ ⇥ Z/nZ is cyclic if and only if gcd(m, n) = 1. In
other words, Z/mZ ⇥ Z/nZ ' Z/mnZ if and only if gcd(m, n) = 1.

Proof. First suppose that Z/mZ ⇥ Z/nZ is a cyclic group. We show that gcd(m, n) = 1.
Let gcd(m, n) = d > 1. Since Z/mZ, Z/nZ are finite groups of order m, n respectively,
we have ma = [0] and nb = [0] for all a 2 Z/mZ and b 2 Z/nZ. Note that,

mn ⇣n m ⌘
(a, b) = ma, nb = ([0], [0]).
d d d

Now mn/d 2 N and mn/d < mn as d > 1. Therefore order of every element in
Z/mZ ⇥ Z/nZ is strictly less than mn, a contradiction to the fact that Z/mZ ⇥ Z/nZ
is a cyclic group. Therefore d = 1.
Next suppose that gcd(m, n) = 1. We show that Z/mZ ⇥ Z/nZ is a cyclic group.
Since Z/mZ, Z/nZ are cyclic groups of order m, n respectively, there exist a 2 Z/mZ
and b 2 Z/nZ such that o(a) = m and o(b) = n. Now,

mn(a, b) = (n(ma), m(nb)) = ([0], [0]).

So o(a, b)  mn. Now if d(a, b) = ([0], [0]), then da = [0] and db = [0]. Therefore m | d
and n | d. We have gcd(m, n) = 1, so mn | d. So mn is the least positive integer so that
mn(a, b) = ([0], [0]) and hence o(a, b) = mn. This proves that Z/mZ ⇥ Z/nZ is cyclic.
Note that Z/mZ⇥Z/nZ is cyclic of order mn if and only if Z/mZ⇥Z/nZ ' Z/mnZ.

Corollary 2.9.6 If (G1 , ⇤1 ) and (G2 , ⇤2 ) are two finite cyclic groups of co-prime orders,
then (G1 ⇥ G2 , ⇤) is a cyclic group.

Proof. Suppose order of (G1 , ⇤1 ) and (G2 , ⇤2 ) are m, n respectively. Then,

G1 ' Z/mZ and G2 ' Z/nZ.

So, G1 ⇥ G2 ' Z/mZ ⇥ Z/nZ and as gcd(m, n) = 1, we have Z/mZ ⇥ Z/nZ is cyclic.
Thus G1 ⇥ G2 is cyclic.

Definition 2.9.7 Let H, K be two subgroups of a group (G, ⇤). Then (G, ⇤) is said to
be an internal direct product of H and K if
§2.9. Direct product of groups 49

1) G = HK,

2) H \ K = {e},

3) h ⇤ k = k ⇤ h for all h 2 H and k 2 K.

Examples 2.9.8

• Consider the Klein’s 4 group. Consider the subgroups H = {e, a} and K = {1, b}
of it. Note that Klein’s 4 group is an internal direct product of H and K.

• Consider A4 , the alternating group of 4 symbols. Consider the subgroups

H = {e, (1 2)(3 4), (1 3)(2 4), (1 4)(2 3)} and K = h(1 2 3)i.

Note that |HK| = 12 = |A4 | and HK ✓ A4 . Thus A4 = HK. But A4 is not an

internal direct product of H and K.

Lemma 2.9.9 Let H, K be two normal subgroups of the group (G, ⇤) and H \K = {e}.
Then h ⇤ k = k ⇤ h for all h 2 H and k 2 K.

Proof. Let h 2 H and k 2 K. Now (h ⇤ k) ⇤ (h ⇤ k) 1 = h ⇤ k ⇤ h 1 ⇤ k 1 . We see that

as H, K are normal, h ⇤ k ⇤ h 1 ⇤ k 1 2 H \ K. Therefore, (h ⇤ k) ⇤ (h ⇤ k) 1 = e. Hence
h ⇤ k = k ⇤ h.
The next theorem gives us an equivalent definition of the internal direct product of
groups.

Theorem 2.9.10 Let H, K be two subgroups of the group (G, ⇤). Then (G, ⇤) is an
internal direct product of H and K if and only if

1) G = HK,

2) H \ K = {e},

3) H, K are normal subgroups.

Proof. We see that the if part of the theorem follows from the previous lemma. So we
assume that (G, ⇤) is an internal direct product of the subgroups H and K. We show
that H, K are normal subgroups. Let g 2 G and h 2 H. As G = HK, we can have
h1 2 H and k1 2 K such that g = h1 ⇤ k1 . Therefore,

1
g⇤h⇤g = h 1 ⇤ k1 ⇤ h ⇤ k1 1 ⇤ h 1 1
 50 §2.9. Direct product of groups

= h1 ⇤ h ⇤ k1 ⇤ k1 1 ⇤ h1 1 , as the elements of H, K commute with each other.

= h1 ⇤ h ⇤ h1 1 2 H.

This proves that H is a normal subgroup. Similarly one can show that K is a normal
subgroup.

Remark 2.9.11 Let n 2 N and H1 , · · · , Hk be normal subgroups of (G, ⇤) such that G =

H1 · · · Hk and Hi \ H1 · · · Hi 1 Hi+1 · · · Hk = {e} for all 1  i  k. Then (G, ⇤) is called
the internal direct product of H1 , . . . , Hk . Note that Hi \H1 · · · Hi 1 Hi+1 · · · Hk = {e}
implies that Hj \ Hj = {e} for all i 6= j. But the other way is not true. For example
consider the symmetric group of three symbols S3 . Let us consider,

H1 := {e, (1 2)}, H2 := {e, (1 3)} and H3 := {e, (1 2 3), (1 3 2)}.

Here Hi \ Hj = {e} for all i 6= j but H1 \ H2 H3 = H1 \ S3 = H1 .

Theorem 2.9.12 Let (G, ⇤) be an internal direct product of the subgroups H and K.
Then,

1) G ' H ⇥ K.

2) G/H ' K and G/K ' H.

Proof.

1) Define f : H ⇥ K ! G by f (h, k) = h ⇤ k for all h 2 H, k 2 K. It is easy to note

that f is a group homomorphism. We show that f is bijective.
Let g 2 G. Since (G, ⇤) is an internal direct product of H, K, there exist h 2 H and
k 2 K such that g = h ⇤ k. We consider (h, k) 2 H ⇥ K so that f (h, k) = h ⇤ k = g.
Let f (h1 , k1 ) = f (h2 , k2 ) i.e. h1 ⇤ k1 = h2 ⇤ k2 for h1 , h2 2 H and k1 , k2 2 K.
Therefore, h1 1 ⇤ h2 = k1 ⇤ k2 1 . Since H \ K = {e} we get that h1 = h2 and k1 = k2 .

2) Define : K ! G/H by (k) = kH for all k 2 K. Note that is a group

homomorphism. Clearly is onto. Now,

Ker = {k 2 K : (k) = H} = {k 2 K : kH = H} = H \ K = {e}.

Therefore by first isomorphism theorem we obtain, G/H ' K. Similarly by defin-

ing : H ! G/K by (h) = hK for all h 2 H, we can show that G/K ' H.
§2.10. Conjugacy relation and Class equation 51

Remark 2.9.13 Similarly one can show that if (G, ⇤) is an internal direct product of
H1 , . . . , Hk , then G ' H1 ⇥· · ·⇥Hk and G/Hi ' H1 ⇥Hi 1 ⇥Hi+1 ⇥Hk for all 1  i  k.

Remark 2.9.14 It is noteworthy that if (G, ⇤) is direct product of two of its subgroups
then we can view (G, ⇤) as an internal direct product of its subgroups too. So we see
that the notions of direct product and internal direct product are not different.

2.10 Conjugacy relation and Class equation

Let (G, ⇤) be a group. We define a relation ⇠ on G as follows: for a, b 2 G, we say that
a ⇠ b if and only if there exist a g 2 G such that b = g ⇤ a ⇤ g 1 .

Exercise 2.10.1 Show that ⇠ is an equivalence relation on (G, ⇤).

This equivalence relation is called the conjugacy relation and if a, b 2 G are related
by conjugacy relation, then they are called the conjugates of each other. We know that
every equivalence relation gives rise to a partition on the set. Here the equivalence class
of a 2 G is called the conjugacy class of a 2 G and is denoted by cl(a). Note that
cl(a) 6= ? as a 2 cl(a). So if we have (G, ⇤) is a group of finite order then we can write,
[
G= cl(ai ),
1in

where ai ’s for all 1  i  n are all the distinct representatives of all the conjugacy classes
of G.

Exercise 2.10.2 Show that a 2 Z(G) if and only if cl(a) = {a}.

Therefore for distinct representatives xi 2

/ Z(G), 1  i  t, we can write,
[ [
G= cl(xi ) Z(G).
1it

Thus,
X
|G| = |Z(G)| + |cl(xi )|.
1it

The above equation is known as the class equation of the group (G, ⇤).
Recall that in a group (G, ⇤), for a 2 G, by the centraliser of a 2 G, we mean the
set CG (a) := {g 2 G : g ⇤ a = a ⇤ g}. We know that Z(G) ✓ CG (a) for every a 2 Z(G).
 52 §2.10. Conjugacy relation and Class equation

Theorem 2.10.3 Let (G, ⇤) be a group. For any a 2 G, we have [G : CG (a)] = |cl(a)|.

Proof. Let L denote the set of all left cosets of CG (a) in (G, ⇤). We shall show that there
is a bijection from L to cl(a). We define, f : L ! cl(a) by f (gCG (a)) := g ⇤ a ⇤ g 1 . We
need to show first that f is well-defined. Note that,

xCG (a) = yCG (a)

1
() x ⇤ y 2 CG (a)
1 1
() (x ⇤ y) ⇤ a = a ⇤ (x ⇤ y)
1
() x ⇤ a ⇤ x = y ⇤ a ⇤ y 1.

From the above we get that f is well-defined and injective. It only remains to show that
f is onto. Let b 2 cl(a). Then there exists g 2 G such that b = g ⇤ a ⇤ g 1 . We consider
gCG (a) 2 L. Then f (gCG (a)) = g ⇤ a ⇤ g 1 = b. Thus f is a bijective map from L to
cl(a). Hence
|L| = |cl(a)|.

So we can rewrite the class equation as

X |G|
|G| = |Z(G)| + .
1it
|CG (xi )|

2.10.1 Applications of class equation

1) Let (G, ⇤) be a group of order pn where p is a prime and n 2 N. Then Z(G) 6= {e}.
If Z(G) = G, then the statement is true. So let Z(G) 6= G. Consider a 2 / Z(G).
Therefore, the centraliser of a, CG (a) 6= G. Also CG (a) 6= {e} as a 2 CG (a). So
CG (a) is a proper subgroup of (G, ⇤). Thus we have,

|G|
= pn r
for some 0 < r < n.
|CG (a)|

We have,
X |G|
|Z(G)| = |G| .
1it
|CG (xi )|

Note that !
X |G|
p| |G| ,
1it
|CG (xi )|
§2.11. Partial converses of the Lagrange’s theorem 53

so we get |Z(G)| is divisible by p. Also |Z(G)| 1 as e 2 Z(G). Therefore

|Z(G)| p and hence Z(G) 6= {e}.

2) Every group of order p2 is abelian where p is prime. We have already seen that
Z(G) 6= {e}. So |Z(G)| is either p or p2 . Suppose |Z(G)| = p. Let a 2
/ Z(G) and
consider CG (a). We have,
Z(G) ✓ CG (a) ✓ G.

So |CG (a)| is either p or p2 . If |CG (a)| = p, then Z(G) = CG (a), a contradiction

as a 2
/ Z(G). If |CG (a)| = p2 , then CG (a) = G. This implies that a 2 Z(G), a
contradiction. Therefore |Z(G)| = p2 and hence Z(G) = G. This completes the
proof.

2.11 Partial converses of the Lagrange’s theorem

We know that in general for any group of order n, if m | n, there need not be a subgroup
of order m. For example, we have seen that A4 does not have a subgroup of order
6 though 6 | |A4 |. In this section we discuss the partial converses of the Lagrange’s
theorem. Recall that we have seen that for a finite cyclic group of order n if m 2 N be
such that m | n, then there exists a unique subgroup of order m. So for finite cyclic
groups the converse of Lagrange’s theorem holds. The theorem below by Cauchy tells us
that in general for any finite group, the converse of Lagrange’s theorem holds for prime
divisors of the order of the group.

Theorem 2.11.1 (Cauchy) Let (G, ⇤) be a finite group and p be a prime such that
p | |G|. Then (G, ⇤) has an element of order p.

Proof. First we prove the statement for all abelian groups of finite order. Let (G, ⇤) be
a finite abelian group and p be a prime such that p | |G|. We use induction on |G| to
prove the statement. Suppose |G| = 1, then there is no prime p such that p | |G|, so
the statement is true vacuously. Suppose the statement is true for all abelian groups of
order strictly less than |G|. Now we consider the abelian group (G, ⇤). Since |G| > 1,
there exists an element a 2 G such that a 6= e.
Case 1: Suppose, p | o(a). Then o(a) = pm for some m 2 N. We consider am 2 G.
Note that,
pm
o(am ) = = p.
gcd(pm, m)
 54 §2.11. Partial converses of the Lagrange’s theorem

Case 2: Suppose, p - o(a). Let o(a) = t. Then gcd(t, p) = 1. We consider the subgroup
H := hai. Since (G, ⇤) is abelian, we have H is normal in (G, ⇤). Therefore we can
consider the quotient group (G/H, ⇤). Now,

|G|
|G/H| = < |G|, as t > 1.
|H|

Now (G/H, ⇤) is also abelian, so by induction hypothesis (G/H, ⇤) has an element say
y of order p. We know that f : G ! G/H defined by f (g) = gH for all g 2 G is an
onto homomorphism. So for y 2 G/H, there exists an x 2 G such that f (x) = y. We
also know that o( (g)) | o(g) for any group homomorphism . So here we get p | o(x)
as o(y) | o(x). This gives us an element x 2 G such that p | o(x). Hence following the
argument in Case 1, we have an element of order p.
Therefore we have proved the statement for all finite abelian groups. Now we shall
prove the statement for all finite groups. Let (G, ⇤) be a finite group and p be a prime
such that p | |G|. We use induction on |G| to prove it. Suppose |G| = 1, then there is
no prime p such that p | |G|, so the statement is true vacuously. Suppose the statement
is true for all groups of order strictly less than |G|. Now we consider the group (G, ⇤).
Case 1: There exist a proper subgroup H of (G, ⇤) such that p | |H|. Now as |H| < |G|,
there exists an element of order p in (H, ⇤) and hence in (G, ⇤).
Case 2: There is no proper subgroup of (G, ⇤), whose order is divisible by the prime p.
We write down the class equation of (G, ⇤):

X |G|
|G| = |Z(G)| + ,
1ik
|C G (ai )|

where for all 1  i  k, ai 2/ Z(G) and they are representatives of all distinct conjugacy
classes of the group (G, ⇤). For all 1  i  k, we have ai 2 / Z(G), so CG (ai ) ⇢ G.
Therefore in this case we have p - CG (ai ) for all 1  i  k. So for all 1  i  k, we have
p | (|G|/|CG (ai )|). This implies that p | |Z(G)|. So Z(G) can not be a proper subgroup
of (G, ⇤). Hence G = Z(G), i.e. (G, ⇤) is abelian and for an abelian group we have
already proved the statement.

2.11.1 Applications of Cauchy’s theorem

1) Up to isomorphism there are only two groups of order 6.
Let (G, ⇤) be a group of order 6. By Cauchy’s theorem there exist elements of
order 2 and 3 in (G, ⇤). Ley a 2 G be such that o(a) = 2 and b 2 G such that
§2.11. Partial converses of the Lagrange’s theorem 55

o(b) = 3.
Case I : Suppose ab = ba. Therefore (a ⇤ b)n = an ⇤ bn for any n 2 N. Thus
o(ab) = lcm(o(a), o(b)) = o(a)o(b) = 6. Therefore (G, ⇤) is cyclic group of order 6
and hence G ' Z/6Z.
Case II : Suppose ab 6= ba. Consider the following subgroups:

H := {e, a} and K := {e, b, b2 }.

Note that K is a normal subgroup of (G, ⇤) as [G : K] = 2. Therefore HK is a

subgroup of (G, ⇤). Now orders of H, K are co-prime, so |H \ K| = 1. Therefore,

|H||K|
|HK| = = 6 = |G|.
|H \ K|

This proves that G = HK := {e, a, b, b2 , a ⇤ b, a ⇤ b2 }. It is a good point to note

that (G, ⇤) is not an internal direct product of H, K as ab 6= ba. We define a map
f : G ! S3 as follows:

f (e) := e, f (a) := (1 2), f (b) := (1 2 3),

f (b2 ) := f (b)2 , f (a ⇤ b) := f (a) f (b), f (a ⇤ b2 ) := f (a) f (b)2 .

So from the definition of f we see that f is a group homomorphism. We show that

f is onto. As {e, (1 2), (1 2 3), (1 3 2)} ✓ Imf , we have |Imf | 4. Also we know
that Imf is a subgroup of S3 , therefore by Lagrange’s theorem |Imf | | 6. This
proves that |Imf | = 6 and hence Imf = S3 . Since f is an onto map from a set of
6 elements to a set of 6 elements, we get that f is injective. By first isomorphism
theorem we obtain G ' S3 .
Thus we get (G, ⇤) is either isomorphic to Z/6Z or S3 and hence the proof.

2) Let (G, ⇤) be a group of order pn where p is a prime and n 2 N. Then (G, ⇤)

contains a normal subgroup of order p.
Since |G| = pn , we have |Z(G)| 1. So |Z(G)| = pr for some 0 < r  n. Now since
p | |Z(G)|, by Cauchy’s theorem there exists an element say x of order p. Consider
H := hxi ✓ Z(G). So H is normal and |H| = p. This proves the existence of a
normal subgroup of order p in (G, ⇤).

3) Let (G, ⇤) be a group of order pn where p is a prime and n 2 N with p > n. Then
(G, ⇤) contains a normal subgroup of order p.
 56 §2.12. Group actions

Since p | |G|, by Cauchy’s theorem there exists an element say x of order p. We

consider H := hxi. This proves the existence of one subgroup of order p. To show
that there exists one normal subgroup of order p, we show that (G, ⇤) has only one
subgroup of order p. Because for every subgroup H of order p, gHg 1 is also a
subgroup of order p for each g 2 G. As there is only one subgroup of order p, it
follows that H = gHg 1 for all g 2 G i.e. H is normal in (G, ⇤). So we show now
that H is the only subgroup of (G, ⇤) of order p. Suppose K is another subgroup
of order p, then H \ K = {e} as H 6= K and H \ K is subgroup of both H and
K. Therefore,
|H||K| p2
|HK| = = > pn.
|H \ K| 1
A contradiction as |G| = pn. So there is only one subgroup of order p.

Exercise 2.11.2 Let (G, ⇤) be a finite abelian group of order n. Let m 2 N be such
that m | n. Then (G, ⇤) has a subgroup of order m.

The above exercise also tells us that converse of Lagrange’s theorem holds for all
finite abelian groups.

2.12 Group actions

Let (G, ⇤) be a group and A be a non-empty set. By an action of the group (G, ⇤)
on A we mean a map from G ⇥ A ! A such that

1) (g1 ⇤ g2 ) · a = g1 · (g2 · a) for all g1 , g2 2 G and a 2 A,

2) eG · a = a for all a 2 A,

where g · a denotes the image of (g, a) and eG denotes the identity of (G, ⇤).

Examples 2.12.1

1. Let V be a R vector space. Note that the scalar multiplication is an action of the
group of non-zero real numbers with multiplication denoted by (R \ {0}, ·) on V .

2. Let (G, ⇤) be a group and A 6= ? be a set. Define G ⇥ A ! A as g · a = a for all

a 2 A, g 2 G. This action is called the trivial action of (G, ⇤) on A.

3. Let (G, ⇤) be a group. Then g1 · g2 = g1 ⇤ g2 for all g1 , g2 2 G is an action of the

group (G, ⇤) on the set G.
§2.12. Group actions 57

4. Let H be a normal subgroup of the group (G, ⇤). Then g · h = g ⇤ h ⇤ g 1

for all
g 2 G, h 2 H is an action of (G, ⇤) on H.

5. Let n 2 N and A = {1, . . . , n}. Note that the symmetric group with n symbols Sn
acts on A as follows: · m = (m).

Remark 2.12.2 We can note that if (G, ⇤) is a non-abelian group then the map g1 ·g2 =
g2 ⇤ g1 for all g1 , g2 2 G is not an action of (G, ⇤) on G.

Proposition 2.12.3 Let (G, ⇤) be a group and A 6= ? a set. Then every action of
(G, ⇤) on A gives rise to a group homomorphism from (G, ⇤) to the permutation group
on A, denoted by SA .

Proof. Suppose (G, ⇤) acts on the set A. Now for g 2 G, we have the map g :A!A
defined by g (a) = g · a for all a 2 A. Note that,

1 1
g · a1 = g · a2 =) g · (g · a1 ) = g · (g · a2 )
1 1
=) (g ⇤ g) · a1 = (g ⇤ g) · a2
=) e · a1 = e · a2
=) a1 = a2 .

For a 2 A, consider g 1 · a 2 A. Then g (g 1 · a) = g · (g 1 · a) = a. Therefore g 2 SA .

We define a map f : G ! SA defined by f (g) = g for all g 2 G. We show that f is a
group homomorphism. Let g1 , g2 2 G. We need to show that

g1 ⇤g2 = g1 g2 .

Note that for a 2 A,

g1 ⇤g2 (a) = (g1 ⇤ g2 ) · a = g1 · (g2 · a) = g1 · ( g2 (a)) = g1 ( g2 (a)) =( g1 g2 )(a).

Remark 2.12.4 On the other hand, for a group (G, ⇤) and A 6= ? set, every group
homomorphism f : G ! SA gives rise to an action of (G, ⇤) on A. Define G ⇥ A ! A by
g · a = f (g)(a). It is easy to check that this is a group action. Also the associated group
homomorphism G ! SA that we get from this action is nothing but f .
 58 §2.12. Group actions

2.12.1 Orbit and stabilizer

Suppose a group (G, ⇤) is acting on a non-empty set A. We can define ⇠ on A as follows:

a ⇠ b if and only if there exists g 2 G such that b = g · a.

Exercise 2.12.5 Prove that ⇠ is an equivalence relation on A.

Definition 2.12.6 For a 2 A, by the orbit of a, we mean the equivalence class of a.

We usually denote the orbit of a by Orb(a). So,

Orb(a) = {g · a : g 2 G}.

As we know that every equivalence relation gives a partition on the set, we can write
A as disjoint union of orbits. Below we revisit the important theorem by Lagrange in
group theory.

Theorem 2.12.7 (Lagrange) Let (G, ⇤) be a finite group and H be a subgroup of it.
Then |H| divides |G|.

Proof. We consider the action of H on G as follows: h · g := h ⇤ g for all h 2 H, g 2 G.

Since G is finite we can write, G as finite disjoint union of orbits of its elements as G.
For g 2 G, we can note that Orb(g) = {h · g : h 2 H} is in bijective correspondence with
H. So |Orb(g)| = |H| for each g 2 G. Hence clearly |H| divides |G|.

Definition 2.12.8 For a 2 A, by the stabilizer of a in (G, ⇤), we refer to the set
{g 2 G : g · a = a}, denoted by Ga .

Exercise 2.12.9 Prove that Ga is a subgroup of (G, ⇤).

Examples 2.12.10

1) Consider the action of the symmetric group with n-symbols on the set {1, . . . , n} as
follows: · m := (m) for all 2 Sn and m 2 {1, . . . , n}. Note that the stabilizer
of m in Sn is isomorphic to Sn 1 .

2) Let (G, ⇤) be a group and P be the power set of G. Consider the action of (G, ⇤)
on P as follows: g · A := gAg 1 for all g 2 G and A 2 P. Note that the stabilizer
of A in G is NG (A).
§2.12. Group actions 59

Proposition 2.12.11 Suppose a group (G, ⇤) acts on a set A 6= ?. Let a 2 A. Then

Orb(a) contains exactly [G : Ga ] many elements.

Proof. Let L denote the set of left cosets of Ga in (G, ⇤). We show that L and Orb(a)
are in one-to-one correspondence. Define f : Orb(a) ! L by f (g · a) = gGa . Note that,

g1 · a = g2 · a
() g1 1 · (g2 · a) = (g1 1 ⇤ g2 ) · a = e · a = a
() g1 1 ⇤ g2 2 Ga
() g1 Ga = g2 Ga .

This shows f is well-defined and injective. Clearly f is surjective and hence the proof.
As an application of this we prove the following theorem:

Theorem 2.12.12 Every 2 Sn can be written as a composition of disjoint cycles.

Proof. Consider the cyclic subgroup of Sn generated by . We call it G = h i. We know

that (G, ) acts on {1, . . . , n} by i · m = i (m). So we can write {1, . . . , n} as disjoint
union of orbits of elements from the set {1, . . . , n}.
Let O be an orbit and m 2 O. We know that |O| = [G : Gm ] and i ·m ! i Gm gives
us a bijective correspondence between O and the set of all left cosets of the stabilizer
Gm in (G, ). Since (G, ) is cyclic, G/Gm is cyclic. Let |G/Gm | = d. So the distinct left
cosets of Gm in (G, ) are Gm , Gm , . . . , d 1 Gm . Hence O = {m, · m, . . . , d 1 · m}.
This completes the proof.

2.12.2 Kernel of a group action

Suppose a group (G, ⇤) acts on a set A 6= ?. Then by the kernel of the group action
we refer to the following set:

Kernel := {g 2 G : g · a = a for all a 2 A}.

Clearly Kernel ✓ Ga for all a 2 A.

Exercise 2.12.13 Prove that Kernel is a subgroup of (G, ⇤).

Example 2.12.14 Let A be a subset of a group (G, ⇤). We consider the normalizer of
A in G denoted by NG (A) and the action of NG (A) on A as follows: g · a = g ⇤ a ⇤ g 1 for
 60 §2.13. Automorphism group

all g 2 NG (A) and a 2 A. Note that the Kernel of this action is CG (A), the centralizer
of A in (G, ⇤). In particular taking A = G we get that the Kernel is Z(G).

2.13 Automorphism group

An isomorphism from the group (G, ⇤) to (G, ⇤) is called an automorphism of (G, ⇤).
We denote the set of all automorphisms of (G, ⇤) by Aut(G).

Exercise 2.13.1 Show that with the composition of functions Aut(G) is a group.

Example 2.13.2 A primary set of examples of automorphisms of (G, ⇤) come from

conjugation. Fix g 2 G. Define fg : G ! G by fg (a) = g ⇤ a ⇤ g 1 for all a 2 G. It can be
checked that fg is an automorphism of (G, ⇤). This particular type of automorphisms are
called inner automorphisms of (G, ⇤). We denote the set of all inner automorphisms
of (G, ⇤) by Inn(G).

Exercise 2.13.3 Show that Inn(G) is a subgroup of Aut(G) under composition of func-
tions.

A priori, we may wonder about a possible bijective correspondence between (G, ⇤)

and (Inn(G), ). We shall note that it happens only if Z(G) = {e}. Because in general

Inn(G) ' G/Z(G).

And therefore G ' Inn(G) if Z(G) = {e}.

Proposition 2.13.4 Let H be a normal subgroup of (G, ⇤). Then the action of (G, ⇤)
on H by g · h = g ⇤ h ⇤ g 1 induces a homomorphism from (G, ⇤) to (Aut(H), ) with
kernel CG (H).

Proof. Define F : G ! Aut(H) by F (g) := fg , where fg : H ! H is defined by

fg (h) = g ⇤ h ⇤ g 1 . One can check that F is a group homomorphism. Note that,

Ker(F ) = {g 2 G : fg = I}, where I(h) = h for all h 2 H

= {g 2 G : g ⇤ h ⇤ g 1
= h for all h 2 H}
= CG (H).
§2.14. Characteristic subgroups 61

Remark 2.13.5 By the first isomorphism theorem we get from the above proposition
that
G/CG (H) ' Im(F ).

Now if we take H = G, we get CG (H) = Z(G) and Im(F ) = Inn(G). Hence

Inn(G) ' G/Z(G).

Remark 2.13.6 If we take (G, ⇤) an abelian group then from the above remark we
can observe that Inn(G) contains only the identity. Let us consider the abelian group
G = Z/3Z. We can note that Aut(Z/3Z) contains exactly two maps;

0 ! 0, 1 ! 1, 2 ! 2 and 0 ! 0, 1 ! 2, 2 ! 1.

So Aut(Z/3Z) ' Z/2Z. In this case we see that Aut(Z/3Z) 6= Inn(Z/3Z).

Exercise 2.13.7 Prove that the automorphism group of the Klein’s 4 - group is isomor-
phic to S3 .

2.14 Characteristic subgroups

Let (G, ⇤) be a group. A subgroup H of (G, ⇤) is called a characteristic subgroup
of (G, ⇤) if every automorphism of (G, ⇤) maps H onto itself i.e. if (H) = H for all
2 Aut(G).
Note that every characteristic subgroup of (G, ⇤) is normal as Inn(G) ✓ Aut(G). But
the converse is not true i.e. every normal subgroup need not be a characteristic subgroup.
For example in Klein’s 4 - group the subgroup {e, a} is normal but not characteristic.

Exercise 2.14.1 Prove that if H is the unique subgroup of (G, ⇤) of a given order, then
H is characteristic subgroup of (G, ⇤).

2.15 Sylow theorems and applications

In this section we present three theorems which are named after the Norwegian mathe-
matician Peter Ludwig Sylow. The first theorem of Sylow is again a partial converse of
Lagrange’s theorem.
 62 §2.15. Sylow theorems and applications

Theorem 2.15.1 (Sylow’s first theorem) Let (G, ⇤) be a group and p be a prime
number such that pn | |G|, where n 2 N. Then there exist subgroups of order pn .

Proof. We shall prove it by induction on |G|.

Let |G| = 1. Then there is no prime p and n 2 N such that pn | |G|. So the statement
is vacuously true.
Now we assume the induction hypothesis i.e. for all groups of order strictly less than
|G|, if any prime power pi where i 2 N divides their order then they have subgroups of
order pi .
Now we consider the group (G, ⇤) and let pn | |G|. We need to show that (G, ⇤) has
a subgroup of order pn . We consider two cases.
Case 1: Suppose (G, ⇤) has a proper subgroup H such that pn | |H|. Then by induction
hypothesis (H, ⇤) has a subgroup of order pn and hence (G, ⇤) has a subgroup of order
pn .
Case 2: Suppose (G, ⇤) has no proper subgroup H such that pn | |H|. We write down
the class equation of (G, ⇤):

X |G|
|G| = |Z(G)| + ,
1jk
|C G (aj )|

where for all 1  j  k, aj 2/ Z(G) and they are representatives of all distinct conjugacy
classes of the group (G, ⇤). Consider the subgroup CG (aj ) of (G, ⇤). As aj 2 / Z(G), we
have CG (aj ) is a proper subgroup of (G, ⇤). According to our assumption in this case
pn - |CG (aj ). Therefore we have for each 1  j  k,

|G|
p| .
|CG (aj )|

This implies that p | |Z(G)|. So by Cauchy’s theorem Z(G) has an element of order p.
We call it a. Consider H := hai. Clearly H is normal as H ✓ Z(G). Therefore we can
consider the quotient group (G/H, ⇤). Now note that the order of G/H is strictly less
than |G|. Also pn 1 divides |G/H| as pn divides |G| and H| = p. Thus by induction
hypothesis, G/H has a subgroup say T of order pn 1 . Now T = K/H for some subgroup
K of (G, ⇤) containing H. Note that |K| = |T ||H| = pn . This proves that (G, ⇤) has a
subgroup of order pi .

Definition 2.15.2 (Sylow subgroup) Let (G, ⇤) be a group such that |G| = pn m,
where p is a prime and m, n 2 N be such that gcd(p, m) = 1. Then the subgroups of
§2.15. Sylow theorems and applications 63

(G, ⇤) of order pn are called the p - Sylow subgroups.

Before we state Sylow’s second theorem, we define the notion of double coset. Let
H, K be two subgroups of a group (G, ⇤). Let x, y 2 G. We define a relation ⇠ on
(G, ⇤) as follows: define x ⇠ y if and only if there exist an h 2 H and k 2 K such that
y = h ⇤ x ⇤ k. One can check that ⇠ is an equivalence relation. Note that the equivalence
class of x 2 G is given by the set HxK := {h ⇤ x ⇤ k : h 2 H, k 2 K}. We call the
set HxK, a double coset of H, K in (G, ⇤). Now G is the union of the distinct double
cosets of H, K in (G, ⇤) as we know that distinct equivalence classes form a partition on
G.

Lemma 2.15.3 Let H, K be two finite subgroups of a group (G, ⇤). Then for any x 2 G,

|H||K|
|HxK| = .
|H \ xKx 1 |

Proof. Let x 2 G. We define a map f : HxK ! HxKx 1

by f (h ⇤ x ⇤ k) := h ⇤ x ⇤ k ⇤ x 1

for all h 2 H, k 2 K.
First we note that f is well-defined. Let h ⇤ x ⇤ k = h0 ⇤ x ⇤ k 0 for h, h0 2 H and
k, k 0 2 K. Therefore, h ⇤ x ⇤ k ⇤ x 1 = h0 ⇤ x ⇤ k 0 ⇤ x 1 i.e. f (h ⇤ x ⇤ k) = f (h0 ⇤ x ⇤ k 0 ).
Also f (h ⇤ x ⇤ k) = f (h0 ⇤ x ⇤ k 0 ), implies that h ⇤ x ⇤ k = h0 ⇤ x ⇤ k 0 . Thus f is injective.
For y 2 HxKx 1 , write y = h ⇤ x ⇤ k ⇤ x 1 for some h 2 H, k 2 K. Then consider
h ⇤ x ⇤ k 2 HxK, so we get a pre-image of y. Thus f is bijective. This proves that,

|HxK| = |HxKx 1 |
|H||xKx 1 |
=
|H \ xKx 1 |
|H||K|
= .
|H \ xKx 1 |

Remark 2.15.4 From the first theorem of Sylow we know that p - Sylow subgroups
exist in a group whose order is divisible by p. Note that if H is a p - Sylow subgroup
then gHg 1 is also a p - Sylow subgroup for each g 2 G. So the collection of subgroups
{gHg 1 : g 2 G} consists of p - Sylow subgroups of (G, ⇤). The second theorem of Sylow
tells us that they are all the p - Sylow subgroups.

Theorem 2.15.5 (Sylow’s second theorem) Let H, K be two p - Sylow subgroups

 64 §2.15. Sylow theorems and applications

of a group (G, ⇤) with |G| = pn m and gcd(p, m) = 1. Then there exists g 2 G such that
K = gHg 1 .

Proof. Consider the distinct double cosets of H, K in (G, ⇤). Since |G| is finite, the
number of distinct double cosets are also finite. Therefore we can write,
X
|G| = |Hxi K|, where Hxi K’s are distinct double cosets.
1ir

We want to show that there exists g 2 G such that K = gHg 1 . Suppose no such g
exists. Then H 6= xi Kxi 1 for all 1  i  r. Therefore H \ xi Kxi 1 ⇢ H and hence
|H \ xi Kxi 1 | = pmi where mi < n. So we have,

X X |H||K|
|G| = |Hxi K| =
1ir 1ir
|H \ xi Kxi 1 |
X p2n
=
1ir
pm i
X
= pn+(n mi )
.
1ir

This implies that pn+1 divides |G|, a contradiction. So there exists g 2 G such that
K = gHg 1 .

Remark 2.15.6 For two subgroups H, K of (G, ⇤) if there exists g 2 G such that
K = gHg 1 , then H, K are called conjugates of each other.

Remark 2.15.7 The third theorem of Sylow determines the number of distinct p -
Sylow subgroups. Before we state and prove the theorem, we define the normaliser of
a subgroup H of (G, ⇤). It is denoted by N (H) and defined as the set N (H) := {g 2
G : gHg 1 = H}. One can check that N (H) is a subgroup of (G, ⇤) and H is a normal
subgroup (N (H), ⇤).

Theorem 2.15.8 (Sylow’s third theorem) Let (G, ⇤) be a group of order pn m, where
p is a prime and n, m 2 N with gcd(p, m) = 1. Then the number of p - Sylow subgroups
of (G, ⇤), np is of the form 1 + kp where k 2 N [ {0} and np | |G|.

Proof. Let H be a p - Sylow subgroup of (G, ⇤). From the second theorem of Sylow we
know that all the p - Sylow subgroups are of the form gHg 1 for g 2 G. First we claim
that,
np = [G : N (H)].
§2.15. Sylow theorems and applications 65

Let S be the set of all distinct p - Sylow subgroups of (G, ⇤) and N be the set of all
distinct left cosets of N (H) in (G, ⇤). Define, f : S ! N by f (gHg 1 ) := gN (H). Note
that,

1 1
aHa = bHb
1 1
() (a ⇤ b)H(b ⇤ a) = H
1 1 1
() (a ⇤ b)H(a ⇤ b) =H
1
() a ⇤ b 2 N (H)
() aN (H) = bN (H).

Therefore the map f is well-defined and injective. Also f is onto as for any gN (H) 2 N ,
we consider gHg 1 2 S so that f (gHg 1 ) = gN (H). This proves that f is bijective and
hence np = [G : N (H)]. From here we get that np | |G| as np |N (H)| = |G|.

Next we consider the double cosets of H, H in (G, ⇤). Let I be the set of all g 2 G
such that HgH are the distinct double cosets of H, H. Therefore,
[
G= HgH.
g2I

We can write,
X X
|G| = |HgH| + |HgH|.
g2I, g2I,
g2N (H) g 2N
/ (H)

Now for g 2
/ N (H), we have H \ gHg 1
⇢ H, so |H \ gHg 1 | = pr where r < n. Thus,

|H||H|
|HgH| = = p2n r
= pn+(n r)
.
|H \ gHg 1 |

Therefore we have,
X
|HgH| = pn+1 u, where u 2 N [ {0}.
g2I,
g 2N
/ (H)

If the second sum is empty sum we get u = 0. Note that for g 2 N (H), gH = Hg, so
HgH = gH. Therefore,
X
|G| = |gH| + pn+1 u.
g2I
g2N (H)

Denote L := {gH : g 2 I, g 2 N (H)}. Let g1 , g2 2 N (H). Note that, g1 H = g2 H if and

 66 §2.15. Sylow theorems and applications

only if Hg1 H = Hg2 H. This proves that |L| = [N (H) : H]. So we have,

|G| = [N (H) : H]|H| + pn+1 u, as |gH| = |H|.

Therefore,
|G| pn+1 u
=1+ 2 N, as |N (H)| divides |G|.
|N (H)| |N (H)|
So we have,
pn+1 u
2 N [ {0}.
|N (H)|
Now pn+1 - |N (H)| as N (H) ✓ G and pn+1 - |G|. This proves that,

pn+1 u
p| .
|N (H)|

Therefore finally we have,

|G|
np = = 1 + kp for some k 2 N [ {0}.
|N (H)|

2.15.1 Applications of Sylow’s theorems

1) Let H be a p - Sylow subgroup of the group (G, ⇤). Then np = 1 if and only if H
is a normal subgroup.
First suppose that H is a normal p - Sylow subgroup of (G, ⇤). Then gHg 1 = H
for all g 2 G. By the second theorem of Sylow we know that all the other Sylow
subgroups are conjugates of H, therefore we have only one p - Sylow subgroup H,
i.e. np = 1. Next suppose that np = 1. We are given H is a p - Sylow subgroup.
As np = 1, we have gHg 1 = H for all g 2 G by the aid of the second theorem of
Sylow. This proves that H is a normal subgroup of (G, ⇤).

2) Let np = 1 for all primes p | |G|. Then (G, ⇤) is the direct product of its Sylow
subgroups.
Let |G| = n. We write n = pa11 · · · pakk where pi ’s are distinct primes and ai 2 N.
Given that npi = 1 for all 1  i  k. We denote each pi - Sylow subgroup by Hi . As
npi = 1 for all 1  i  k, we have Hi ’s are normal. We show that G ' H1 ⇥· · ·⇥Hk .
Note that H1 · · · Hk is a subgroup of (G, ⇤). Also Hi \ H1 · · · Hi 1 Hi+1 · · · Hk = {e}
§2.15. Sylow theorems and applications 67

as orders of Hi ’s are mutually co-prime. Therefore |H1 · · · Hk | = |G| and hence

H1 · · · Hk = G. This prove us that (G, ⇤) is an internal direct product of H1 , . . . , Hk .
Thus G ' H1 ⇥ · · · ⇥ Hk .

3) Let (G, ⇤) be a group of order pq where p, q are primes with p > q. Also let
q - (p 1). Then (G, ⇤) is a cyclic group.

Proof. We know that np can be 1, p, q or pq as np divides|G|. Also we know that

np = 1 + kp for some integer k 0. Therefore np is either 1 or q. Now q 6= (1 + kp)
for k 0 as q < p. Therefore np = 1 and this implies that the p - Sylow subgroup
of (G, ⇤) say H is normal.
Similarly nq can be 1, p, q or pq as nq divides |G|. Also nq = 1+k 0 q for some integer
k 0 0. Therefore nq is either 1 or p. Now p 6= (1 + k 0 q) for k 0 1 as q - (p 1).
Therefore nq = 1 and this implies that the q - Sylow subgroup of (G, ⇤) say K is
normal.
Since H, K are of co-prime orders we also get that H \ K = {e}. Thus (G, ⇤) is an
internal direct product of H, K and therefore G ' H ⇥ K. Note that H ' Z/pZ
and K ' Z/qZ as every group of prime order is cyclic and up to isomorphism there
is only one group of a fixed finite order. As p, q are co-prime, we have,

G ' H ⇥ K ' Z/pZ ⇥ Z/qZ ' Z/pqZ.

This completes the proof.

Remark 2.15.9 The last application plays a very important in the classification of finite
abelian groups. For example, it tells us that every group of order 15 is cyclic.

Remark 2.15.10 If in the last application we take q | (p 1), then nq can be 1 or p.

If nq = 1 then again we get the group to be cyclic. But if nq = p, then there exists
a unique non-abelian group of order pq. To find such non-abelian example we need to
study semi-direct product.

Theorem 2.15.11 Every group of order 12 has either a normal 3 - Sylow subgroup or
it is isomorphic to the alternating group of 4 symbols A4 .

Proof. Let (G, ⇤) be a group of order 12. We have |G| = 22 · 3. So the possible choices
for n3 are 1, 2, 3, 4, 6, 12. If n3 = 1, then (G, ⇤) has a normal 3 - Sylow subgroup. So
let n3 > 1. As n3 = 1 + 3k for some k 2 N, then n3 can only be 4. Let X be the
 68 §2.15. Sylow theorems and applications

set of all distinct 3 - Sylow subgroups of (G, ⇤). We denote the distinct 3 - Sylow
subgroups by P1 , P2 , P3 , P4 . First we note that the normaliser of Pi , N (Pi ) = Pi for all
1  i  4. We have seen that [G : N (Pi )] = n3 = 4. So |N (Pi )| = 3. As N (Pi ) ✓ Pi
and |N (Pi )| = 3 = |Pi |, we obtain that N (Pi ) = Pi . We define g : X ! X by
g (Pi ) := gPi g
1
for all 1  i  4. Clearly g is a bijection as gPi g 1 = gPj g 1 implies
that Pi = Pj , which is not true. Now we define a map : G ! S4 by (g) = g . We can
check that (g ⇤ g 0 ) = g⇤g0 = g g . Thus
0
is a group homomorphism. Now,

Ker = {g 2 G : gPi g 1
= Pi for all 1  i  4}
= {g 2 G : g 2 N (Pi ) for all 1  i  4}
= {g 2 G : g 2 Pi for all 1  i  4}, as N (Pi ) = Pi for all 1  i  4.
= {e}, as Pi ’s are all distinct.

Because |Pi \ Pj | can be 1 or 3. If |Pi \ Pj | = 3, then it turns out that Pi = Pj . This

proves that is injective. Now we show that is onto. Each Pi contains 2 elements of
order 3, therefore (G, ⇤) has exactly 8 elements of order 3. Since is injective, there are
exactly 8 elements of order 3. In S4 , an element has order 3 if and only if it is a 3 - cycle.
Now in S4 , there are total 8 many 3 - cycles and 3 - cycles are all even permutation. So
|Im \ A4 | 8. Also |A4 | = 12 and |Im \ A4 | | |A4 |. Therefore Im \ A4 = A4 . This
implies that A4 ✓ Im . Again,

|Im |  |G| = 12 = |A4 |.

Hence Im = A4 . This proves that G ' A4 .

Theorem 2.15.12 Let (G, ⇤) be a group of order p2 q where p, q are distinct primes.
Then (G, ⇤) has a normal Sylow subgroup.

Proof. The possible choices for np are 1, p, p2 , q, pq, p2 q. Also as np = 1+kp for some inte-
ger k 0, np can be either 1 or q. Similarly the possible choices for nq are 1, p, p2 , q, pq, p2 q
and as nq = 1 + k 0 q for some integer k 0 0, nq can be 1, p or p2 .
Now since p, q are distinct primes, either p < q or q < p.
Case I: Let p > q. Then np = 1 + kp 6= q. So np = 1 and hence (G, ⇤) has a normal p -
Sylow subgroup.
Case II: Let p < q. In this case if nq = 1, then (G, ⇤) has a normal q - Sylow subgroup.
Let nq > 1. Now nq = 1 + k 0 q 6= p as p < q. So let nq = p2 . Therefore, k 0 q = p2 1 =
(p 1)(p + 1). As q is a prime number, we have q | (p 1) or q | (p + 1). Note that
§2.16. Structure theorem of finite abelian groups 69

q - (p 1) as q > p. So q | (p + 1). Therefore we have,

p < q  (p + 1).

This is possible only if p = 2, q = 3. This implies that (G, ⇤) is a group of order 12.
Now as (G, ⇤) is a group of order 12 either (G, ⇤) has a normal 3 - Sylow subgroup or
G ' A4 . If G ' A4 , then note that {e, (1 2)(3 4), (1 3)(2 4), (1 4)(2 3)} is a normal 2-
Sylow subgroup of A4 .

2.16 Structure theorem of finite abelian groups

For a given n 2 N, we know there are only finitely many groups of order n up to
isomorphism as every group of order n is isomorphic to some subgroup of Sn and there
are only finitely many subgroups of Sn of order n. The structure theorem of finite abelian
groups helps us to determine the number of abelian groups of order n up to isomorphism.

Theorem 2.16.1 (Structure theorem of finite abelian groups) Any finite abelian
group can be written as a direct product of finite cyclic groups.

For a given n 2 N, by writing down the prime factorisation of n, we can list down all
the finite abelian groups of order n. We illustrate a few examples in the following table:

n prime factorisation All possible finite abelian groups (up to isomorphism) of order n
1 1 {0}
2 2 Z/2Z
3 3 Z/3Z
4 2⇥2 Z/4Z, Z/2Z ⇥ Z/2Z
10 2⇥5 Z/2Z ⇥ Z/5Z

In this context, let us define what do we mean by a finitely generated group.

Definition 2.16.2 A group (G, ⇤) is called finitely generated if there exists a finite
set A ✓ G such that every element of G can be written as a composition of the elements
of the set A.

We know that all cyclic groups are generated by a single element, hence they are
finitely generated. Other than cyclic groups we can consider the group (Z ⇥ Z, +). This
 70 §2.16. Structure theorem of finite abelian groups

group is finitely generated as this group is generated by the set {(1, 0), (0, 1)}. We also
have structure theorem for finitely generated abelian groups, which states that a finitely
generated abelian group is isomorphic to Zr ⇥ K where r 0 and (K, ⇤) is a
finite abelian group. So if (G, ⇤) is a finitely generated abelian group, then there
exists an r 2 N [ {0} and ni 2 N for all 1  i  t such that

G ' Zr ⇥ Z/n1 Z ⇥ · · · Z/nt Z.

2.16.1 Applications of the structure theorem of finite abelian

groups
1) Up to isomorphism there are only two abelian groups of order 12. By structure
theorem of finite abelian groups we have up to isomorphism there are only two
groups of order 12 namely

Z/3Z ⇥ Z/4Z ' Z/12Z and Z/2Z ⇥ Z/2Z ⇥ Z/3Z ' Z/2Z ⇥ Z/6Z.

2) Up to isomorphism there is only one abelian group of order 6 namely

Z/2Z ⇥ Z/3Z ' Z/6Z as gcd(2, 3) = 1.

3) We now know that every group of order 22 = 4 is abelian. From the structure
theorem of finite abelian groups we know that up to isomorphism there are only
two abelian groups of order 4. Therefore up to isomorphism there are only two
groups of order 4.

2.16.2 Proof of the structure theorem of finite abelian groups

In an abelian group, every subgroup is normal. So all the Sylow subgroups are normal.
Let (G, ⇤) be an abelian group of order n = pa11 · · · pakk where pi ’s are distinct primes and
ai 2 N. Let Hi denote the pi - Sylow subgroup of (G, ⇤) for all 1  i  k. Therefore we
can write
G ' H1 ⇥ · · · ⇥ Hk .

So now if we can show that each Hi can be written as a product of finite cyclic groups,
then we are done. Therefore suffices to prove the structure theorem for a finite abelian
group (G, ⇤) of order pn for some prime p and n 2 N. We choose a1 2 G such that o(a1 ) is
§2.16. Structure theorem of finite abelian groups 71

maximal. Say o(a1 ) = pn1 where 1  n1  n. Consider A1 := ha1 i. As (G, ⇤) is abelian,

we have A1 is a normal subgroup. So we can consider the quotient group (G/A1 , ⇤). We
choose b2 2 G such that o(b2 A1 ) is maximal in (G/A1 , ⇤). Say o(b2 A1 ) = pn2 . We know
that f : G ! G/A1 defined by f (g) := gA1 for all g 2 G is a group homomorphism.
Therefore, o(b2 A1 ) | o(b2 ). So o(b2 A1 )  o(b2 ). Since pn1 is the maximal order in (G, ⇤),
we have pn2  pn1 i.e. n2  n1 . Note that since o(b2 A1 ) = pn2 , so pn2 is the first power
of b2 that falls into A1 . Now we can write,

n2
bp2 = ai1 for some i 2 Z.

Thus,
n1 p n1 n2
bp2 = ai1 .
n1
Also bp2 = e as pn1 is the maximum of orders of all elements of the group (G, ⇤).
p n1 n2
Therefore we get that ai1 = e. Hence pn1 | ipn1 n2 i.e. pn2 | i. Write i = jpn2 for
n1 n2
some j 2 Z. So bp2 = ai1 = ajp i.e. (b2 a1 j )p 2 = e. We set a2 := b2 a1 j 2 G. Note that
n
1
o(a2 ) = pn2 . Then consider A2 := ha2 i. We show that A1 \ A2 = {e}. Let at2 2 A1 . Then
we have bt2 a1 jt 2 A1 . This implies that bt2 2 A1 . So bt2 A1 = A1 . Since o(b2 A1 ) = pn2 , we
n2
get that pn2 | t. Therefore, at2 = arp 2 = e where r 2 Z. This proves that A1 \ A2 = {e}.

Now note that A1 A2 is a normal subgroup of (G, ⇤). We then choose b3 2 G so that
the order of b3 A1 A2 is maximal in G/A1 A2 . Say o(b3 A1 A2 ) = pn3 . We first claim that
n3  n2  n1 . Since we already have n2  n1 , it is enough to show that n3  n2 . Since
n
b2 2 G is such that o(b2 A1 ) = pn2 is maximal in (G/A1 , ⇤), we get (b3 A1 )p 2 = A1 . So,

n2
bp3 2 A1 ⇢ A1 A2 .

This gives us that,

o(b3 A1 A2 )  pn2 , i.e. pn3  pn2 .
n3
Hence n3  n2 . Now bp3 2 A1 A2 , so we can write,

n3
bp3 = ai11 ai22 for some integers i1 , i2 .

n2
As (b3 )p 2 A1 , we get

n2 n3 n n2 n3 n2
(ai11 ai22 )p = ((b3 )p 2 )p = (b3 )p 2 A1 .

n2 n3
Thus (ai22 )p 2 A1 . Since o(a2 A) = pn2 , we have pn2 | i2 pn2 n3
, i.e. pn3 | i2 . Also
 72 §2.16. Structure theorem of finite abelian groups

n1
bp3 = e. Hence,
n1 n3 n1
(ai11 ai22 )p = bp3 = e.
n1 n3
Hence ai11 p 2 A1 \ A2 and we have A1 \ A2 = {e}. Therefore,

pn1 | i1 pn1 n3
i.e. pn3 | i1 .

n3
Write i1 = j1 pn3 , i2 = j2 pn3 . Set a3 := b3 a1 j1 a2 j2 , so that ap3 = e, moreover o(a3 ) = pn3 .
Set A3 := ha3 i. We now show that A3 \ A1 A2 = {e}. Let at3 2 A1 A2 , then

(b3 a1 j1 a2 j2 )t 2 A1 A2 .

n3
Hence bt3 2 A1 A2 . Therefore pn3 |t. Since ap3 = e, we get at3 = e. This proves that
A3 \ A1 A2 = {e}.
We repeat the process with the normal subgroup A1 A2 A3 . Since (G, ⇤) is a finite
group, we finally end up getting cyclic subgroups Ai = hai i of order pni for 1  i 
s, with n1 n2 ··· ns such that G = A1 A2 · · · As and for every 2  i  s,
Ai \ A1 · · · Ai 1 = {e}. This proves that (G, ⇤) is an internal direct product of the cyclic
subgroups A1 , . . . , As . Therefore G ' A1 ⇥ · · · ⇥ As .