Topics in Algebra

Topics in Algebra
Lecture notes, Autumn 2021

Mikko Korhonen
Topics in Algebra: Group Theory Mikko Korhonen
Contents
1 Introduction 3
1.1 Isomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 Order of an element . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Cyclic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 On the structure of (Z/nZ)× . . . . . . . . . . . . . . . . . . . 18
1.6 Homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.7 Dihedral groups . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.8 Quaternion group Q8 . . . . . . . . . . . . . . . . . . . . . . . 27
1.9 Cosets and products of subsets . . . . . . . . . . . . . . . . . . 28
1.10 Normal subgroups and quotient groups . . . . . . . . . . . . . 33
1.11 Conjugacy classes of elements . . . . . . . . . . . . . . . . . . 41
1.12 Cauchy’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.13 Number of conjugacy classes . . . . . . . . . . . . . . . . . . . 46
1.14 Conjugacy classes of subgroups . . . . . . . . . . . . . . . . . 47
1.15 p-groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2 Symmetric and alternating Groups 51

2.1 Cycle decomposition . . . . . . . . . . . . . . . . . . . . . . . 51
2.2 Orders of permutations . . . . . . . . . . . . . . . . . . . . . . 54
2.3 Conjugacy classes in Sn . . . . . . . . . . . . . . . . . . . . . . 55
2.4 Alternating groups . . . . . . . . . . . . . . . . . . . . . . . . 56
2.5 Conjugacy classes in An . . . . . . . . . . . . . . . . . . . . . 58
2.6 An is simple for n ≥ 5 . . . . . . . . . . . . . . . . . . . . . . 59
2.7 Group actions . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3 Linear groups 67
3.1 On finite fields . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2 Basic properties of GLn (q) and SLn (q) . . . . . . . . . . . . . 68
3.3 Polynomial rings . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.4 Characteristic polynomials and eigenvalues . . . . . . . . . . . 71
3.5 Conjugacy classes of GL2 (q) . . . . . . . . . . . . . . . . . . . 72
3.6 Minimal polynomials . . . . . . . . . . . . . . . . . . . . . . . 76
3.7 Orders of elements . . . . . . . . . . . . . . . . . . . . . . . . 78
3.8 Factorizations of polynomials . . . . . . . . . . . . . . . . . . 79
3.9 Generators for GL2 (F) and SL2 (F) . . . . . . . . . . . . . . . 82
3.10 Simplicity of PSL2 (q) . . . . . . . . . . . . . . . . . . . . . . . 84
1
4 Normal series 89
4.1 Characteristic subgroups . . . . . . . . . . . . . . . . . . . . . 89
4.2 Commutator subgroups and solvability . . . . . . . . . . . . . 90
4.3 Jordan–Hölder theorem . . . . . . . . . . . . . . . . . . . . . . 95
4.4 Nilpotent groups . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.5 Upper central series . . . . . . . . . . . . . . . . . . . . . . . . 103
4.6 Higher commutators . . . . . . . . . . . . . . . . . . . . . . . 104
5 Constructing groups 108

5.1 Direct products . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.2 Classification of finitely generated abelian groups . . . . . . . 111
5.3 Automorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.4 Elementary abelian p-groups . . . . . . . . . . . . . . . . . . . 119
5.5 Semidirect products . . . . . . . . . . . . . . . . . . . . . . . . 120
5.6 Application: Groups of order pq (p, q primes) . . . . . . . . . 127
5.7 Application: p-groups of order p3 . . . . . . . . . . . . . . . . 129
5.8 Application: Groups of order 2n (n odd) . . . . . . . . . . . . 134
6 Sylow theory 136

6.1 Double cosets . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
6.2 Sylow theorems . . . . . . . . . . . . . . . . . . . . . . . . . . 137
6.3 Sylow subgroups of subgroups and quotients . . . . . . . . . . 141
6.4 Number of p-subgroups of given order . . . . . . . . . . . . . . 143
6.5 Nilpotent groups and Sylow subgroups . . . . . . . . . . . . . 144
6.6 Groups of certain orders are solvable . . . . . . . . . . . . . . 146
6.7 Fitting subgroup and normal Sylow subgroups . . . . . . . . . 148
6.8 Groups of order p2 q (p and q distinct primes) . . . . . . . . . 151
6.9 Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.10 Groups of squarefree order . . . . . . . . . . . . . . . . . . . . 165
Exercises 170
2
1 Introduction
These lecture notes will cover some of the fundamentals of group theory, with
a particular focus on finite groups and their structure.
There are no formal prerequisites, but some previous exposure to basic
algebra and group theory is expected. However, we will develop the theory
from the beginning, starting with the basic definitions and examples.
Definition 1.1. A group is a pair (G, ∗), where G is a set and ∗ : G×G → G
is a binary operation (a, b) 7→ a ∗ b satisfying the following properties:
(i) Associativity: (a ∗ b) ∗ c = a ∗ (b ∗ c) for all a, b, c ∈ G
(ii) There exists an identity element e ∈ G such that e ∗ a = a ∗ e = a for

all a ∈ G.
(iii) For all a ∈ G, there exists an inverse b ∈ G such that a ∗ b = b ∗ a = e.
The cardinality |G| of the set G is called the order of the group (G, ∗).
Remark 1.2. (a) Usually we will denote a group (G, ∗) just by G.
(b) An identity element e ∈ G is unique. For if e0 is another identity element,

then e0 = e ∗ e0 = e.
(c) Inverse elements are also unique. For if b and b0 are inverses of a, then
b = e ∗ b = (b0 ∗ a) ∗ b = b0 ∗ (a ∗ b) = b0 ∗ e = b0 .
(d) In general we use multiplicative notation. That is, the multiplication

a ∗ b of two elements a, b ∈ G is usually denoted by ab, and the identity
element is usually denoted by 1 or 1G . The inverse of a ∈ G is denoted
by a−1 .
(e) For all a, b ∈ G we have (ab)−1 = b−1 a−1 .
(f) Every group G satisfies the following cancellation laws:
• If ab = ac, then b = c.
• If ac = bc, then a = b.
Definition 1.3. Let G be a group. Two elements a, b ∈ G are said to

commute if ab = ba. A group (G, ∗) is called abelian if all pairs of elements
commute, i.e., if ab = ba for all a, b ∈ G.
3
If (G, ∗) is an abelian group, we will sometimes use additive notation:

that is, we write a ∗ b as a + b for a, b ∈ G, the inverse of a ∈ G is denoted
by −a and the identity element is denoted by 0.
Example 1.4. Examples of abelian groups:
(i) (R, +): identity element is 0 ∈ R, the inverse of x ∈ R is −x ∈ R.

Similarly e.g. (Z, +) and (Q, +).
(ii) (R \ {0}, ·): identity element is 1 ∈ R, inverse of x ∈ R \ {0} is 1/x.

Similarly e.g. (C \ {0}, ·) and (R>0 , ·).
Example 1.5 (Integers modulo n). Let n > 0 be an integer. For an integer
a ∈ Z, we denote by a the set of all integers b ∈ Z such that b ≡ a mod n.
In other words,
a = a + nZ.
Note that a = b if and only if a ≡ b mod n. We define the integers modulo
n as
Z/nZ = {a : a ∈ Z} = {0, 1, . . . , n − 1}.
We define addition and multiplication in Z/nZ by
a+b=a+b for all a, b ∈ Z

a · b = ab for all a, b ∈ Z
(These operations are well-defined: if a = a0 and b = b0 , then a + b = a0 + b0

and ab = a0 b0 .) Then (Z/nZ, +) is an abelian group of order n, called the
additive group of integers modulo n.
Multiplication in Z/nZ is associative and has multiplicative identity 1.
But (Z/nZ, ·) is not a group, since not every a has an inverse (for example
a = 0). In general, for a there exists b such that a · b = 1 if and only if
gcd(a, n) = 1. The set of invertible elements
(Z/nZ)× = {a : gcd(a, n) = 1}
forms a group ((Z/nZ)× , ·), which we call the multiplicative group of integers
modulo n.
The order of (Z/nZ)× is given by the number of integers 1 ≤ a ≤ n such
that gcd(a, n) = 1. In other words |(Z/nZ)× | = φ(n), where φ is the Euler
totient function.
4
Remark 1.6. Recall that for p any prime number, we have φ(p) = p − 1,
and φ(pk ) = pk−1 (p − 1) for all k > 1. More generally, for n > 0 with prime
factorization n = pk11 · · · pkt t , we have

1 1
φ(n) = n 1 − ··· 1 − .
p1 pt
As a consequence of this formula, for n, m > 0 with gcd(m, n) = 1, we have

φ(nm) = φ(n)φ(m).
Example 1.7. Recall that a ring is a triple (R, +, ·), where R is a set and
+ and · are binary operations on R such that the following hold:
• (R, +) is an abelian group;
• Multiplication · is an associative binary operation on R with identity

1 ∈ R;
• Multiplication is distributive over +, in other words (a + b)c = ac + bc

and a(b + c) = ab + ac for all a, b, c ∈ R.
A ring R is commutative if ab = ba for all a, b ∈ R. For a ring R, we

denote the set of invertible elements in R by R× , in other words
R× = {x ∈ R : xy = yx = 1 for some y ∈ R}.
Then (R× , ·) is a group, and called the group of units of R. A field is a

commutative ring (R, +, ·) such that every non-zero element has an inverse,
in other words R× = R \ {0}.
For example, (Z/nZ, +, ·) is a commutative ring with group of units
((Z/nZ)× , ·), and Z/nZ is a field if and only if n is a prime number. Other
examples of fields are (Q, +, ·), (R, +, ·), and (C, +, ·). The integers (Z, +, ·)
form a commutative ring which is not a field.
Definition 1.8 (Symmetric groups). Let Ω be a set. The symmetric group

on Ω is the group (Sym(Ω), ◦), where
Sym(Ω) = {f : Ω → Ω | f is a bijection }
and ◦ is composition of functions. The identity element is the identity map

1 : Ω → Ω defined by 1(x) = x for all x ∈ Ω. If Ω = {1, . . . , n} for some
integer n > 0, we denote Sym(Ω) = Sn . In this case | Sym(Ω)| = n! (n
factorial).
5
If Ω = {ω1 , . . . , ωn }, then for g ∈ Sym(Ω) we denote

ω1 ω2 · · · ωn
g= ,
ω10 ω20 · · · ωn0
where g(ωi ) = ωi0 for 1 ≤ i ≤ n.

Definition 1.9. Let n > 0 be an integer, and let (i1 , . . . , ik ) be a k-tuple of
distinct integers from {1, 2, . . . , n}. The k-cycle corresponding to (i1 , . . . , ik )
is the permutation σ ∈ Sn defined by
σ(it ) = it+1 for all 1 ≤ t < k,

σ(ik ) = i1 ,
σ(d) = d for all d ∈ {1, 2, . . . , n} \ {i1 , . . . , ik }.
We denote σ = (i1 i2 · · · ik ). A 2-cycle is called a transposition.

Example 1.10. In S3 , every element is a cycle: we have
S3 = {(1), (1 2), (1 3), (2 3), (1 2 3), (1 3 2)}.
With the group operation we have defined, we have e.g. (1 2)(1 3) = (1 3 2)

and (1 2 3)(2 3) = (1 2). In Sn for n ≥ 4, you also have elements σ ∈ Sn
which are not cycles, for example σ = (1 2)(3 4) and σ = (1 2 3)(4 5).
Example 1.11. Exercise: If |Ω| ≥ 3, then Sym(Ω) is not abelian.
Definition 1.12 (General linear group). Let n > 0 be an integer, and let F
be a field (for example, F = Q or F = C). The general linear group of degree
n is the group (GLn (F), ·), where GLn (F) is the set of invertible (n × n)-
matrices with entries in F, and the group operation is matrix multiplication.
It is easy to check that GLn (F) is a group, with the identity matrix
 
1
 1 
In := 
 
 . . . 

1
as the identity element.

Example 1.13. A square matrix with entries in a field is invertible if and
only if it has nonzero determinant. Thus for example:

a b
GL2 (R) = : a, b, c, d ∈ R and ad − bc 6= 0 .
c d
6
Example 1.14. Let p be a prime. An important example of a field is

(Z/pZ, +, ·), which we will denote by Fp . We denote GLn (p) := GLn (Fp )
for all n > 0. Then for example:

1 0 1 1 1 0 0 1 0 1 1 1
GL2 (2) = , , , , , .
0 1 0 1 1 1 1 0 1 1 1 0
1.1 Isomorphisms
Definition 1.15. Let (G, ∗) and (H, ?) be groups. A map ϕ : G → H is
called an isomorphism, if ϕ is a bijection and ϕ(a ∗ b) = ϕ(a) ? ϕ(b) for all
a, b ∈ G. If there exists an isomorphism G → H, we say that G and H are
isomorphic and denote this by G ∼ = H.
If two distinct groups are isomorphic, they have the exact same multipli-
cation table, just with different names for the elements (the different names
are given by an isomorphism ϕ). Therefore as groups their internal struc-
ture is the same, and they have the same group-theoretical properties. For
example, if G ∼ = H, then G is abelian if and only if H is abelian.
Example 1.16. Let (G, ∗) be a group with G = {1, a, b} and with the
following multiplication table:
∗ 1 a b
1 1 a b
a a b 1
b b 1 a
Suppose that (H, ?) is a group such that G ∼ = H, and let ϕ : G → H

be an isomorphism. Denote 1 := ϕ(1), a := ϕ(a), and b0 = ϕ(b). Then
0 0
H = {10 , a0 , b0 }. Since ϕ(x) ? ϕ(y) = ϕ(x ∗ y) for all x, y ∈ G, we see that the
multiplication table of H is as follows:
? 10 a0 b0
10 10 a0 b0
a0 a0 b0 10
b0 b0 10 a0
So the table is exactly the same as for (G, ∗), just the names of the
elements are different. In this particular example, you can also check that
(G, ∗) ∼
= (Z/3Z, +).
7
Example 1.17. An isomorphism ϕ : G → G is called an automorphism.

For any group G, the identity map ϕ : G → G defined by ϕ(g) = g is an
automorphism. If G is an abelian group, then the inverse map g 7→ g −1
defines an automorphism of G.
Example 1.18. For any field F, we have GL1 (F) ∼
= F× .
Example 1.19. Let V be an n-dimensional vector space over a field F. Then
the general linear group on V is the group (GL(V ), ◦), where GL(V ) is the
set of all bijective linear maps V → V , and ◦ is composition of functions.
. For f ∈ GL(V ), let Af be the (n × n)-matrix
Fix a basis e1 , . . ., en of VP
n
Af = (Aij ), where f (ei ) = j=1 Aji ej for all 1 ≤ i ≤ n. Then the map
f 7→ Af defines an isomorphism GL(V ) → GLn (F).
Remark 1.20. Isomorphism forms an equivalence relation in any nonempty
collection G of groups. In other words, for all A, B, C ∈ G the following hold:
• A∼
= A. (reflexivity)
• If A ∼
= B, then B ∼
= A. (symmetry)
• If A ∼
= B and B ∼
= C, then A ∼
= C. (transitivity)
One type of problem that group theorists have studied since the early
days of group theory are various classification problems, which are roughly
speaking of the form “classify all groups G with property P up to isomor-
phism”. A solution to such a problem would be a collection G (described as
explicitly as possible) of groups such that:
• Every X ∈ G is a group with property P ;
• If X, Y ∈ G and X 6= Y , then X ∼
6= Y ;
• If G is a group with property P , then G ∼

= X for some X ∈ G.
Here are two fundamental classification problems in finite group theory.

Problem 1.21. Let n ≥ 0 be an integer. Classify all groups of order n up to
isomorphism. In other words, describe a set Gn such that all of the following
hold:
(i) X is a finite group of order n for all X ∈ Gn ;
(ii) If X, Y ∈ Gn and X 6= Y , then X ∼

6= Y ;
(iii) If G is a group of order n, then G ∼

= X for some X ∈ Gn .
8
The cardinality of a set Gn as in Problem 1.21 is called the group number

of n, and denoted by gnu(n) := |Gn |. (Exercise: show that 1 ≤ gnu(n) < ∞
for all integers n > 0.)
Problem 1.22. Let n > 0 be an integer. What is the value of gnu(n)?
There is no hope currently for a complete solution for these problems,
except for specific values of n. The history goes back to Cayley1 , who in
1854 introduced the abstract definition of groups and classified groups of
order 6. In modern notation, Cayley showed that if G is a finite group of
order 6, then G ∼
= Z/6Z or G ∼ = S3 .
Currently the value of gnu(n) is known for n < 2048. It is known that
gnu(2048) > 1774274116992170 ≈ 1.77 · 1015
but the precise number of groups of order 2048 remains unknown2,3 . The
value of gnu(n) is also known for special values of n, such as prime powers
n = pk with k ≤ 7. For example, for any prime p, we have gnu(p) = 1,
gnu(p2 ) = 2, gnu(p3 ) = 5, and
(
14, if p = 2;
gnu(p4 ) =
15, if p ≥ 3.

51,
 if p = 2;
5
gnu(p ) = 67, if p = 3;

61 + 2p + 2 gcd(p − 1, 3) + 2 gcd(p − 1, 4), if p ≥ 5.

The results and the theory that we develop during this course will allow
us to describe Gn and gnu(n) for some specific values of n. See Table 1 for a
list of small values.4
Besides classifying various classes of groups up to isomorphism, another
thing that we want to understand in group theory is the structure of a given
group G. By “structure”, we usually mean the various algebraic properties5
1
A. Cayley. On the theory of groups, as depending on the symbolic equation θn = 1.
Philos. Mag., 4(7):40–47, 1854.
2
J. H. Conway, H. Dietrich, E. A. O’Brien, Counting groups: gnus, moas, and other
exotica. Math. Intelligencer 30 (2008), no. 2, 6–18.
3
B. Eick, M. Horn, A. Hulpke, Constructing groups of ‘small’ order: recent results and
open problems. Algorithmic and experimental methods in algebra, geometry, and number
theory, 199–211, Springer, Cham, 2017.
4
See for example https://oeis.org/A000001 for more values.
5
Here is a formal definition: A property P is an algebraic property (or isomorphism-
invariant) if it is preserved under isomorphisms. In other words, if G and H are isomorphic
groups, then G has property P if and only if H has property P .
9
n gnu(n) n gnu(n) n gnu(n) n gnu(n) n gnu(n)

0 0 13 1 26 2 39 2 52 5
1 1 14 2 27 5 40 14 53 1
2 1 15 1 28 4 41 1 54 15
3 1 16 14 29 1 42 6 55 2
4 2 17 1 30 4 43 1 56 13
5 1 18 5 31 1 44 4 57 2
6 2 19 1 32 51 45 2 58 2
7 1 20 5 33 1 46 2 59 1
8 5 21 2 34 2 47 1 60 13
9 2 22 2 35 1 48 52 61 1
10 2 23 1 36 14 49 2 62 2
11 1 24 15 37 1 50 5 63 4
12 5 25 2 38 2 51 1 64 267
Table 1: Values of gnu(n) for 0 ≤ n ≤ 64.
that are associated with G and its elements. The sections that follow will
introduce some of the basic algebraic properties that are studied in group
theory.
1.2 Order of an element

By associativity, we do not need to worry about brackets when multiplying
elements in a group G. For example, for n = 4 and for any a1 , a2 , a3 , a4 ∈ G
it is easy to see that the products
a1 (a2 (a3 a4 )), a1 ((a2 a3 )a4 ), (a1 a2 )(a3 a4 ), (a1 (a2 a3 ))a4 , ((a1 a2 )a3 )a4
are all equal, and thus can be denoted by a1 a2 a3 a4 without ambiguity.

More generally, we have generalized associativity: for a1 , a2 , . . ., an ∈ G
the product a1 a2 · · · an ∈ G is uniquely determined. That is, we get the same
element regardless of the choice of brackets (Exercise: prove this by induction
on n).
Thanks to generalized associativity, we can define powers of elements in
a group.
Definition 1.23. Let G be a group. For a ∈ G, we denote a0 = 1G , a1 = a,
and more generally an = aa · · · a (n times) for an integer n > 1. We define
a−n = (a−1 )n for all n > 0. It is easy to check that for all a ∈ G and for all
m, n ∈ Z, we have am an = am+n and (am )n = amn .
Remark 1.24. If G is an abelian group and if we are using additive notation,
we shall instead write 0a = 0, 1a = a, and na = a + · · · + a (n times) for
integers n > 1. Similarly (−n)a = −(na) for all n > 0.
10
Definition 1.25. Let G be a group and let g ∈ G. The order |g| of g is

defined as follows. If the elements g, g 2 , g 3 , g 4 , . . . are all distinct, we say
that g has infinite order and define |g| = ∞. If there exist distinct integers
r, s such that g r = g s , then we say that g has finite order. When g has finite
order, there exists n > 0 such that g n = 1. In this case we call the smallest
such n the order of g and define |g| = n.
Example 1.26. (i) Let G = Sn and consider a k-cycle σ = (i1 · · · ik ).
Then |σ| = k.
(ii) Let G = (Z/nZ, +). Then 1 has order n in G.
(iii) Let G = (Z, +). Then every a ∈ Z \ {0} has infinite order in G.
Lemma 1.27. Let G be a group and let g ∈ G be an element of finite order
|g| = n.
(i) Let m ∈ Z. Then g m = 1 if and only if n divides m.
n
(ii) Let k ≥ 0. Then |g k | = gcd(n,k)
.
(iii) Let k ≥ 0 and suppose that k | n. Then |g k | = nk .
(iv) {g k : k ∈ Z} = {1, g, . . . , g n−1 }.

Proof. (i) By Euclidean division, we can write m = qn + r with q ∈ Z and
0 ≤ r < n. Now g m = (g n )q g r = g r since g n = 1, so we have g m = 1
if and only if g r = 1. Since n is the smallest positive integer such that
g n = 1, we have g r = 1 if and only if r = 0, which is equivalent to
n | m.
(ii) First note that
(g k )n/ gcd(k,n) = (g n )k/ gcd(k,n) = 1.
Next consider d ≥ 0 such that (g k )d = 1. Then by (i) we have n | kd,

which implies
n k
| d.
gcd(k, n) gcd(k, n)
n k
Since gcd( gcd(k,n) , gcd(k,n) ) = 1, it follows that
n
| d.
gcd(k, n)
n
Therefore |g k | = gcd(k,n)
.
11
(iii) Immediate from (ii).
(iv) By (i), we have g i = g j if and only if i ≡ j mod n, from which the

claim follows.
We end this section with some results about the order of the product xy
of two elements x and y in a group G.
Lemma 1.28. Let G be a group and let x, y ∈ G. Then |xy| = |yx|.
Proof. We will show that for n > 0, we have (xy)n = 1 if and only if (yx)n =
1, from which the lemma follows. For this, we can write
(xy)n = (xy)(xy) · · · (xy) = x(yx) · · · (yx)y = x(yx)n−1 y,
so (xy)n = 1 if and only if x(yx)n−1 y = 1. Multiplying both sides on the left

by y, we see that x(yx)n−1 y = 1 is equivalent to (yx)n y = y, which in turn
is equivalent to (yx)n = 1, by cancelling out y.
Lemma 1.29. Let G be a group and let x, y ∈ G be elements of finite orders

|x| = m and |y| = n. Suppose that xy = yx and that gcd(m, n) = 1. Then
|xy| = mn.
Proof. Since xy = yx, it is easy to see that (xy)d = xd y d for all integers d ∈ Z.
In particular, we have (xy)mn = 1 since xm = 1 and y n = 1. Suppose now
that (xy)d = 1. Then xd = y −d , and powering both sides by n shows xdn = 1.
Therefore m | dn, which implies m | d since gcd(m, n) = 1. Similarly we have
y d = x−d , and powering both sides by m shows that y dm = 1, which implies
n | d. Therefore m | d and n | d, which gives mn | d since gcd(m, n) = 1.
Thus |xy| = mn.
Note that both assumptions in Lemma 1.29 are necessary:
• Consider any nontrivial element x ∈ G of finite order and let y = x−1 .

Then xy = yx. But |xy| = 1, so |xy| =
6 |x||y|.
• Consider G = S3 and any x, y ∈ G with |x| = 2 and |y| = 3. Then

gcd(|x|, |y|) = 1. But |xy| = 2, so |xy| =
6 |x||y|.
In general, suppose that x, y ∈ G are nontrivial elements of finite order.

There are examples where xy has infinite order (Exercise 1.3). If xy has finite
order, what do |x| and |y| tell us about |xy|? By the following result, the
answer is “absolutely nothing”.
12
Theorem 1.30 (Miller, 1900). Let m, n, ` > 1 be integers. Then there exists
a finite group G which contains elements x and y such that |x| = m, |y| = n,
and |xy| = `.
Proof. Omitted.
As a historical note, the first proof of Theorem 1.30 was given in a paper
of G. A. Miller6 from 1900. Let k = max(m, n, `). In his paper, Miller
constructs explicitly permutations x, y ∈ Sk+2 such that |x| = m, |y| = n,
and |xy| = `. The proof by Miller is a somewhat lengthy calculation involving
many different cases. Since 1900, the result has been rediscovered7 several
times and shorter proofs have been found.
1.3 Subgroups
One basic thing that we want to understand about a group G is its subgroup
structure, that is, the groups that are contained inside of G.
Definition 1.31. Let (G, ∗) be a group. A subset H ⊆ G is a subgroup, if
H is closed under the binary operation ∗ and H equipped with ∗ is a group.
We denote this by H ≤ G.
Note that if H is a subgroup G, then H and G must have the same
identity element. Indeed, suppose that x = 1H is the identity element of
H. Then x2 = x, so multiplying both sides by x−1 shows that x = 1G .
Throughout these notes, we will use the following lemma to check whether a
subset H ⊆ G is a subgroup.
Lemma 1.32. Let G be a group. A subset H ⊆ G is a subgroup if and only
if all of the following hold:
(i) 1G ∈ H;
(ii) ab ∈ H for all a, b ∈ H;
(iii) a−1 ∈ H for all a ∈ H.
Proof. If H is a subgroup of G, then 1G ∈ H as seen above, so (i) holds. Pro-
perty (ii) holds by definition, and (iii) holds since H is a group and inverses
are unique in G. Conversely if (i) — (iii) hold then H is a subgroup; we only
need to check associativity and this follows immediately from associativity
in G.
6
G. A. Miller, On the Product of Two Substitutions, American Journal of Mathematics,
Vol. 22, No. 2 (Apr., 1900), pp. 185-190.
7
J. König, A note on the product of two permutations of prescribed orders, European
J. Combin. 57 (2016), 50–56.
13
Example 1.33. Examples of subgroups:
(i) Let G be a group. Then G itself is a subgroup of G, and moreover the

trivial subgroup {1} is always a subgroup of G.
(ii) (Z, +) ≤ (Q, +) ≤ (R, +) ≤ (C, +).
(iii) GLn (Q) ≤ GLn (R) ≤ GLn (C).
(iv) Let n > 0 be an integer and let F be a field. We define the special linear
group SLn (F) as
SLn (F) = {A ∈ GLn (F) : det(A) = 1}.
Since det(AB) = det(A) det(B) and det(A−1 ) = 1/ det(A), it follows

that SLn (F) is a subgroup of GLn (F).
Example 1.34. If G is finite and H ⊆ G is a nonempty subset closed under

products, then H is a subgroup. This follows from the fact that a g ∈ H has
some finite order n, so g n = 1 is in H, and g n−1 = g −1 is also in H. (More
generally, this works if every element of G has finite order.)
However, in general closure under the group operation is not enough.
Consider H = Z≥0 in G = (Z, +). Then H equipped with + contains the
identity element and is closed under addition. But clearly H is not closed
under taking inverses, so H is not a subgroup.
Definition 1.35. Let H be a subgroup of a group G. If H G, we say that

H is a proper subgroup. If H 6= {1}, we say that H is a nontrivial subgroup.
Lemma 1.36. Let G be a group. Then the intersection of any collection of

subgroups of G is a subgroup.
Proof. Any subgroup contains the identity, so the same is true for their in-
tersection. The intersection is also closed under products and inverses since
the subgroups are.
Definition 1.37. Let G be a group and S ⊆ G. We define the subgroup

generated by S as the intersection of all subgroups of G that contain S. We
denote the subgroup generated by S with hSi := ∩S⊆H≤G H. For x1 , . . .,
xn ∈ G, we denote hx1 , . . . , xn i := h{x1 , . . . , xn }i.
Definition 1.38. A group G is said to be finitely generated if G = hx1 , . . . , xn i

for some x1 , . . . , xn ∈ G.
Lemma 1.39. Let G be a subgroup and S ⊆ G. Then:
14
(i) hSi is the smallest subgroup of G that contains S. In other words, if

S ⊆ H ≤ G, then hSi ≤ H.
(ii) We have
hSi = {g ∈ G : g = sε11 · · · sεnn for some n ≥ 0, si ∈ S and εi ∈ {1, −1}}.
Here for n = 0, we interpret sε11 · · · sεnn = 1 (empty product).
Proof. (i) Since hSi is the intersection of some subgroups of G, it is itself

a subgroup of G. For S ⊆ H ≤ G, we have hSi ≤ H by the definition
of hSi as the intersection of all subgroups of G that contain S.
(ii) Let H = {g ∈ G : g = sε11 · · · sεnn for some si ∈ S and εi ∈ {1, −1}}.

We have S ⊆ H and clearly H is contained in every subgroup of G
that contains S, in particular H ⊆ hSi. We will show that H is a
subgroup, from which it follows that hSi = H. To this end, it is clear
that 1 ∈ H and that H is closed under multiplication. Closure under
inverses follows from (sε11 · · · sεnn )−1 = s−ε
n
n
· · · s−ε
1 , so H is a subgroup.
1
1.4 Cyclic groups

Definition 1.40. A group G is said to be cyclic if G = hgi for some g ∈ G.
Example 1.41. For all n > 0, the group Z/nZ is cyclic of order n, and Z is
infinite cyclic.
In this section, we will describe the basic properties of cyclic groups.
Lemma 1.42. Let G be a group and g ∈ G.
(i) hgi = {g k : k ∈ Z}.
(ii) If g has finite order |g| = n, then hgi is a subgroup of order |g| and
hgi = {1, g, . . . , g n−1 }.
Proof. (i) Follows from Lemma 1.39.
(ii) Follows from (i) and Lemma 1.27.
Lemma 1.43. Let G and H be cyclic groups. Then G ∼

= H if and only if
|G| = |H|. In particular:
15
(i) If G has finite order n, then G ∼

= Z/nZ.
(ii) If G has infinite order, then G ∼
= Z.
Proof. If G ∼ = H, then |G| = |H|. Conversely if G = hgi and H = hhi are

cyclic groups of equal order, check that g k 7→ hk defines an isomorphism
G → H. Then (i) follows since Z/nZ is cyclic of order n, and (ii) since Z is
infinite cyclic.
Lemma 1.43 justifies the following notation.
Definition 1.44. We denote a cyclic group of order n by Cn .
Next we will describe the subgroups of cyclic groups.
Lemma 1.45. Every subgroup of a cyclic group is cyclic.
Proof. Let G = hgi be cyclic and let H ≤ G be a subgroup. Then there exists
g d ∈ H with d ≥ 0, choose d to be minimal. We will show that H = hg d i. It
is clear that hg d i ≤ H.
To show that H ≤ hg d i, let h ∈ H and write h = g k with k ∈ Z. By
Euclidean division, we can write k = qd + r with q ∈ Z and 0 ≤ r < d. Then
g k = (g d )q g r ,
which implies g r ∈ H. By minimality of d we have r = 0, so h = g qd ∈

hg d i.
Lemma 1.46. Let G = hgi be a cyclic group of order n > 0.
(i) Let k ∈ Z. Then hg k i = hg gcd(k,n) i.
(ii) Every subgroup of G has the form hg d i for some d | n.
(iii) Let d | n. Then G has a unique subgroup H of order d, and H = hg n/d i.
(iv) Let k ∈ Z. Then hg k i = G if and only if gcd(k, n) = 1.
(v) The number of generators of G is equal to φ(n).
Proof. (i) By Bézout’s lemma, we can write λk + µn = gcd(k, n) for some

λ, µ ∈ Z. Since g n = 1, this shows that g gcd(k,n) = g λk ∈ hg k i, so
hg gcd(k,n) i ≤ hg k i. On the other hand |g k | = |g gcd(k,n) | by Lemma 1.27
(ii)–(iii), so hg gcd(k,n) i = hg k i.
(ii) Since gcd(k, n) divides n, this is immediate from (i).
16
(iii) Let H ≤ G be a subgroup of order d, where d | n. By (ii) we have

0 0
H = hg d i for some d0 | n. Since |g d | = n/d0 by Lemma 1.27 (iii), it
follows that d = |H| = n/d0 , so d0 = n/d.
(iv) We have hg k i = G if and only if |g k | = n, so this follows from Lemma

1.27 (ii).
(v) Follows from (iv).
Remark 1.47. Let G = hgi be cyclic of order n. By Lemma 1.46 (ii) – (iii),
the number of subgroups of G is equal to the number of divisors of n, with
d | n corresponding to the subgroup hg n/d i of order d. For example if G = hgi
is cyclic of order 6, then G has a total of 4 subgroups: {1}, hg 2 i, hg 3 i, and
hgi.
Lemma 1.48. Let G = hgi be a cyclic group of infinite order.
(i) Let d ∈ Z \ {0}. Then hg d i is infinite cyclic.
(ii) Every subgroup of G has the form hg d i for some d ≥ 0.

0
(iii) hg d i ≤ hg d i if and only if d0 | d.
0
(iv) hg d i = hg d i if and only if d = ±d0 .
(v) hg d i = G if and only if d = ±1.
Proof. (i) Note first that since g has infinite order, for all k, ` ∈ Z we have
g k = g ` if and only if k = `. If hg d i is finite, we have g dk = 1 = g 0 for
some k > 0, which gives dk = 0 and thus d = 0.
(ii) By Lemma 1.45, every subgroup of G is of the form hg k i for some k ∈ Z.

If k ≥ 0 we are done with k = d. If d < 0, then hg k i = hg −k i and the
claim follows with d = −k.
0 0 0
(iii) We have hg d i ≤ hg d i if and only if g d ∈ hg d i. Now g d ∈ hg d i if and
0
only if g d = g d k for some k ∈ Z, equivalently d = d0 k for some k ∈ Z
0
since g has infinite order. Therefore g d ∈ hg d i if and only if d0 | d.
0
(iv) By (iii), we have hg d i = hg d i if and only if d0 | d and d | d0 , which is
equivalent to d = ±d0 .
(v) Immediate from (iv).
17
1.5 On the structure of (Z/nZ)×

We will now apply the basic results on cyclic groups to describe the structure
of (Z/nZ)× in some cases.
Definition 1.49. Let G be a group and d > 0 an integer. We denote by
od (G) the number of elements of order d in G, in other words the cardinality
of the set {x ∈ G : |x| = d}.
In this section, we will make use of the fact that will be proven later using
Lagrange’s theorem (Corollary 1.81): if G is a finite group, the order of any
x ∈ G divides the order of G. It follows then for a finite group G of order n,
we have X
n = |G| = od (G).
d|n
We begin with an elementary result which does not involve any group
theory, but can be proven using some basic properties of cyclic groups.
P
Theorem 1.50. Let n > 0 be an integer. Then n = d|n φ(d).
Proof. Let G be a cyclic group of order n. Each element of order d generates
a subgroup of order d, and there is only such subgroup in G by Lemma 1.46
(iii). Therefore od (G) is equal to the number of generators for a cyclic group
of order d, and so od (G) = φ(d) by Lemma 1.46 (v). We conclude then that
X X
n = |G| = od (G) = φ(d).
d|n d|n
Theorem 1.51. Let G be a finite group such that for all d > 0, the number
of solutions to xd = 1 in G is at most d. Then G is cyclic.
Proof. Denote n = |G|. We have od (G) = φ(d)sd (G), where sd (G) is the
number of cyclic subgroups of order d in G. (This is because two different
cyclic subgroups
P of order d cannot have elements of order d in common.)
Since n = d|n od (G), we have
X X
φ(d)sd (G) = φ(d) (1.1)
d|n d|n
by Theorem 1.50. If we assume that xd = 1 has at most d solutions for all

d > 0, then sd (G) ≤ 1 for all d | n. It follows then from (1.1) that sd (G) = 1
for all d | n. In particular sn (G) = 1, so G is cyclic.
18
For a field F, recall that we denote the multiplicative group (F \ {0}, ·)

by F× . We will now apply Theorem 1.51 to prove the following fundamental
result about finite fields.
Theorem 1.52. Let F be a finite field. Then F× is cyclic.
Proof. Over any field F, a polynomial of degree d > 0 has at most d roots.
In particular, the polynomial xd − 1 ∈ F[x] has at most d roots, so there are
at most d solutions to xd = 1 in F× . By Theorem 1.51, the multiplicative
group F× is cyclic.
In particular, we have shown that (Z/pZ)× is cyclic for any prime p > 0.
In general (Z/nZ)× is not cyclic; for example x2 = 1 for all x ∈ (Z/8Z)× , so
(Z/8Z)× does not have an element of order 4 = φ(8). We now consider the
question of when (Z/pn Z)× is cyclic for a prime number p. We will see that
(Z/pn Z)× is cyclic if p > 2, and usually not if p = 2. First we need some
notation and a lemma.
Definition 1.53. Let n 6= 0 be an integer and let p be a prime. We denote

by νp (n) the largest integer α ≥ 0 such that pα | n. (In other words, the
unique integer α ≥ 0 such that pα | n and pα+1 - n.)
For example, for n = 60 = 22 · 3 · 5 we have ν2 (n) = 2, ν3 (n) = 1,

ν5 (n) = 1, and νp (n) = 0 for any prime p > 5.
Lemma 1.54. Let p > 2 be a prime, and let z ≡ 1 mod p. Then the
following hold:
(i) νp (z p − 1) = νp (z − 1) + 1.
k
(ii) νp (z p − 1) = νp (z − 1) + k for all k > 0.
Proof. (i) Write z = 1 + tp, where t ≥ 1. Then νp (z − 1) = α + 1, where

pα is the largest power of p dividing t.
Applying the binomial formula, we get

p p p p 2 2 p
z −1 = (1+tp) −1 = tp+ t p +· · ·+ tp−1 pp−1 +tp pp .
1 2 p−1
The first term in the sum is certainly divisible by pα+2 , and not by pα+3 .
It is easy to see that all of the other terms in the sum are divisible by
pα+3 , so we conclude that pα+2 is the largest power of p dividing z p − 1.
In other words νp (z p − 1) = νp (z − 1) + 1.
19
(ii) For k = 1 the claim is just (i). Suppose that k > 1. Then
k k−1 k−1
νp (z p − 1) = νp ((z p )p − 1) = νp (z p − 1) + 1
by (i), so the claim follows by induction on k.
Theorem 1.55. Let p > 2 be a prime. Then (Z/pn Z)× is cyclic for all
n ≥ 1.
Proof. For n = 1, the claim follows from Theorem 1.52. Suppose then that
n > 1. Our aim is to find an element of order φ(pn ) = pn−1 (p − 1) in
(Z/pn Z)× . For this it will suffice to find an element of order pn−1 and (p − 1),
as then by Lemma 1.29 their product will have order pn−1 (p − 1).
Let g ∈ Z be such that the image of g is a generator of (Z/pZ)× , in other
words g has order p−1 in (Z/pZ)× . Then the order of g in (Z/pn Z)× must be
divisible by p − 1, so a suitable power of g will have order p − 1 in (Z/pn Z)× .
For an element of order pn−1 , we claim that the image of z = 1 + p has
k
order pn−1 in (Z/pn Z)× . Indeed, by Lemma 1.54 (ii) we have νp (z p − 1) =
n−1 n−2
k + 1 for all k > 0. Therefore z p ≡ 1 mod pn and z p 6≡ 1 mod pn , so
z has order pn−1 in (Z/pn Z)× .
Lemma 1.56. Let z be an odd integer and k > 0. Then the following
statements hold:
k
(i) If z ≡ 1 mod 4, then ν2 (z 2 − 1) = ν2 (z − 1) + k.
k
(ii) If z ≡ 3 mod 4, then ν2 (z 2 − 1) = ν2 (z + 1) + k.
Proof. Exercise.
Theorem 1.57. If n ≥ 3, then (Z/2n Z)× is not cyclic.
Proof. Exercise. (Apply Lemma 1.56.)
1.6 Homomorphisms
Definition 1.58. Let G and H be groups. A map ϕ : G → H is called a
homomorphism, if ϕ(xy) = ϕ(x)ϕ(y) for all x, y ∈ G.
Rather than only looking at structure-preserving maps which are bijective
(isomorphisms), it is natural to consider structure-preserving maps which are
not necessarily bijections (homomorphisms).
For example, note that if ϕ : G → H is an injective homomorphism, then
it is easily shown that ϕ(G) is a group and moreover that ϕ provides an
20
isomorphism G ∼ = ϕ(G) (see Lemma 1.63 (vi) below). In this case we say
that G embeds into H. The question whether a given group G embeds into
another group H comes up all the time in group theory.
If ϕ is not injective, a homomorphic image ϕ(G) is still a group: the
multiplication table of ϕ(G) is like that of G, except some elements of G are
identified. For example, consider the map
ϕ : Z → Z/nZ, ϕ(x) = x for all x ∈ Z.
This map is a surjective homomorphism; the image of Z is Z/nZ which as a

group is like (Z, +), except a, b ∈ Z with a ≡ b mod n are identified.
Lemma 1.59. Let G and H be groups and let ϕ : G → H be a homomor-

phism. Then:
(i) ϕ(x1 x2 · · · xk ) = ϕ(x1 )ϕ(x2 ) · · · ϕ(xk ) for all x1 , x2 , . . . , xk ∈ G;
(ii) ϕ(1G ) = 1H ;
(iii) ϕ(x−1 ) = ϕ(x)−1 for all x ∈ G;
(iv) ϕ(xn ) = ϕ(x)n for all x ∈ G and n ∈ Z;
Proof. (i) For k = 2 this is the definition of a homomorphism, for k > 2

apply ϕ(x1 x2 · · · xk ) = ϕ(x1 )ϕ(x2 · · · xk ) and induction on k.
(ii) We have ϕ(1G )ϕ(1G ) = ϕ(1G 1G ) = ϕ(1G ); then multiplying both sides
of ϕ(1G )ϕ(1G ) = ϕ(1G ) with ϕ(1G )−1 gives ϕ(1G ) = 1H .
(iii) We have ϕ(x)ϕ(x−1 ) = ϕ(1G ) = 1H by (ii); then multiplying both sides

of ϕ(x)ϕ(x−1 ) = 1H on the left with ϕ(x)−1 gives ϕ(x−1 ) = ϕ(x)−1 .
(iv) Follows from (i) – (ii) for n ≥ 0. Then for n < 0 the claim follows from
(iii), since xn = (x−n )−1 .
Definition 1.60. Let G and H be groups and let ϕ : G → H be a homo-

morphism. The kernel of ϕ is the set
Ker ϕ = {g ∈ G : ϕ(g) = 1}.
Example 1.61. Let G be any group and let g ∈ G. Since g m g n = g m+n for
all m, n ∈ Z, we have a homomorphism ϕ : Z → G defined by ϕ(k) = g k
for all k ∈ Z. If g has infinite order, then Ker ϕ = {0}. If g has finite order
|g| = n, then Ker ϕ = nZ.
21
Example 1.62. Let n > 0 be an integer and let F be a field. We have

det(AB) = det(A) det(B) for all A, B ∈ GLn (F); in other words the deter-
minant map det : GLn (F) → F \ {0} is a group homomorphism. The kernel
consists of A ∈ GLn (F) with det(A) = 1, in other words, the kernel is SLn (F).

phism. Then:
(i) Ker ϕ is a subgroup of G.
(ii) For x, y ∈ G we have ϕ(x) = ϕ(y) if and only if xy −1 ∈ Ker ϕ.
(iii) ϕ is injective if and only if Ker ϕ = {1}.
(iv) If X ≤ G, then ϕ(X) ≤ H.
(v) If Y ≤ H, then ϕ−1 (Y ) ≤ G.
(vi) If ϕ is injective, then ϕ(X) ∼

= X for all X ≤ G, and in particular
∼
ϕ(G) = G.
(vii) If ϕ is injective, then |ϕ(g)| = |g| for all g ∈ G.
Proof. (i) We have 1 ∈ Ker ϕ by Lemma 1.59 (ii). If x, y ∈ Ker ϕ then

ϕ(xy) = ϕ(x)ϕ(y) = 1 so xy ∈ Ker ϕ. Furthermore ϕ(x−1 ) = ϕ(x)−1 =
1 by Lemma 1.59 (iii), so Ker ϕ is a subgroup.
(ii) We have ϕ(x) = ϕ(y) if and only if ϕ(x)ϕ(y)−1 = 1, which is equivalent

to ϕ(xy −1 ) = 1 by Lemma 1.59 (iii).
(iii) If ϕ is injective, then Ker ϕ = {1} as otherwise ϕ(x) = ϕ(1) for some
g 6= 1. Conversely suppose that Ker ϕ = {1}. Then ϕ(x) = ϕ(y)
implies x = y by (ii), so ϕ is injective.
(iv) Since 1G ∈ X, we have ϕ(1G ) = 1H ∈ ϕ(X) by Lemma 1.59 (ii).

Closure under products follows from the definition of a homomorphism,
and closure under inverses from Lemma 1.59 (iii).
(v) We have 1G ∈ ϕ−1 (Y ) since ϕ(1G ) = 1H ∈ Y . If x, y ∈ ϕ−1 (Y ), then

ϕ(x), ϕ(y) ∈ Y and thus ϕ(x)ϕ(y) = ϕ(xy) ∈ Y , giving xy ∈ ϕ−1 (Y ).
We also have ϕ(y)−1 = ϕ(y −1 ) ∈ Y by Lemma 1.59 (iii), so y −1 ∈
ϕ−1 (Y ). Thus ϕ−1 (Y ) ≤ G.
(vi) If ϕ is injective, then the restriction of ϕ to X provides an isomorphism

X → ϕ(X).
22
(vii) We have ϕ(g)n = ϕ(g n ) for all g ∈ G by Lemma 1.59 (iv). Thus if ϕ is
injective, then ϕ(g)n = 1 if and only if g n = 1, so |ϕ(g)| = |g|.
Definition 1.64. Let G be a group and g ∈ G. The inner automorphism

corresponding to g is the map γg : G → G defined by γg (x) = gxg −1 for all
x ∈ G.
Lemma 1.65. Let G be a group and g ∈ G. Then γg : G → G is an
automorphism for all g ∈ G. In particular:
(i) (gxg −1 )k = gxk g −1 for all x ∈ G and k ∈ Z;
(ii) |gxg −1 | = |x| for all x ∈ G.
Proof. Let g ∈ G. We have
γg (xx0 ) = gxx0 g −1 = (gxg −1 )(gx0 g −1 ) = γg (x)γg (x0 )
for all x, x0 ∈ G, so γg is a homomorphism. We have gxg −1 = gyg −1 if

and only if x = y by the cancellation laws, so γg is injective. Finally γg is
surjective, since x = g(g −1 xg)g −1 = γg (g −1 xg) for all x ∈ G. Thus γg is
an automorphism. Now claims (i) and (ii) follow from Lemma 1.59 (iv) and
Lemma 1.63 (vii).
Next we will prove Cayley’s theorem, which shows that every group is
isomorphic to a subgroup of Sym(Ω) with |Ω| = |G|. In particular, it follows
that a finite group G can be embedded into Sn with n = |G|. For most finite
groups this is not optimal, in the sense that usually one can find smaller
n such that G embeds into Sn . (Consider for example G = Sn , for which
|G| = n! is much bigger than n.)
Theorem 1.66 (Cayley’s theorem). Let G be a group. For g ∈ G, define
maps Lg , Rg : G → G by Lg (x) = gx and Rg (x) = xg for all x ∈ G. Then:
(i) Lg and Rg are bijections for all g ∈ G.
(ii) The map ϕ : G → Sym(G) defined by ϕ(g) = Lg for all g ∈ G is an
injective homomorphism.
(iii) The map ψ : G → Sym(G) defined by ϕ(g) = Rg−1 for all g ∈ G is an
injective homomorphism.
Proof. (i) For Lg , injectivity follows from the cancellation laws (if gx = gy,
then x = y); surjectivity follows from x = g(g −1 x). In the same way
we see that Rg is a bijection.
23
(ii) We have Lg Lh (x) = Lg (hx) = g(hx) = (gh)x = Lgh (x), so Lgh = Lg Lh

for all g, h ∈ G. Thus ϕ is a homomorphism. For injectivity, if Lg = Lh ,
then in particular Lg (1) = Lh (1), so g = h.
(iii) Similar to (ii).
The maps Lg in Theorem 1.66 are called left translation maps, and the
homomorphism ϕ is called the left regular representation. Similarly the Rg
are called right translations and the homomorphism ψ is called the right
regular representation.
1.7 Dihedral groups

We have found that the structure of cyclic groups — that is, groups generated
by one element — is fairly simple to describe. What about groups generated
by two elements? In this case very little can be said in general, as is already
hinted by Theorem 1.30. One special case where we can give a classification
result is when G is generated by two elements of order two.
Definition 1.67. A group G is said to be dihedral, if G = ha, bi for some

a, b ∈ G with |a| = |b| = 2.
Lemma 1.68. A group G is dihedral if and only if G = hx, yi for some

x, y ∈ G such that |x| = 2, xyx−1 = y −1 , and x 6∈ hyi.
Proof. Suppose that G is dihedral, say G = ha, bi with |a| = |b| = 2. We

claim that x = a and y = ab works. It is clear that G = hx, yi. Moreover
|x| = 2, and xyx−1 = a(ab)a = ba = y −1 .
It remains to check that x 6∈ hyi. To this end, if x ∈ hyi, then x and y
must commute, which implies xyx−1 = y. On the other hand xyx−1 = y −1 ,
so y = y −1 and thus y 2 = 1. But since |x| = 2 and x ∈ hyi, we must have
x = y. In other words a = ab, which is impossible since b 6= 1.
For the other direction, suppose that G = hx, yi for some x, y ∈ G such
that |x| = 2, xyx−1 = y −1 , and x 6∈ hyi. Let a = x and b = xy. Then
b2 = (xyx)y = y −1 y = 1 and b 6= 1 since x 6∈ hyi, so |b| = 2. Thus G = ha, bi
with |a| = |b| = 2.
Lemma 1.69. Suppose that G is a group such that G = hx, yi with |x| = 2,
xyx−1 = y −1 , and x 6∈ hyi. Then:
(i) G = {xi y j : i ∈ {0, 1} and j ∈ Z}.
24
(ii) For all i, i0 ∈ {0, 1} and j, j0 ∈ Z we have

i
(xi y j )(xi0 y j0 ) = xi+i0 y (−1) 0 j+j0 .
(iii) For all i, i0 ∈ {0, 1} and j, j0 ∈ Z we have xi y j = xi0 y j0 if and only if

i = i0 and y j = y j0 .
(iv) If |y| < ∞, then |G| = 2|y|.
Proof. The proof of (i) and (ii) is based on the following observation. We
have xyx−1 = y −1 , so by Lemma 1.65 (i) we have xy k x−1 = y −k and thus
xy k = y −k x for all k ∈ Z. (1.2)
This relation will allow us to write any product of involving x, y, and y −1 in

the form xi y j .
(i) By Lemma 1.39, every g ∈ G can be written in the form g = sε11 · · · sεnn ,
where si ∈ {x, y} and εi ∈ {1, −1} for all 1 ≤ i ≤ n. We will show that
g can be written in the form xi y j for some i, j ∈ Z. If n = 0 or n = 1
this is clear, so we suppose that n > 1 and proceed by induction on n.
Then
εn−1 εn
g = sε11 · · · sn−1 sn = xi y j sεnn
by induction. If sn = x, then g = xi (y j x) = xi+1 y −j by (1.2). If sn = y,
then g = xi y j y εn = xi y j+εn . This completes the proof of (i).
(ii) If i0 = 0, then (xi y j )(xi0 y j0 ) = xi+i0 y j+j0 as claimed. If i0 = 1, then
(xi y j )(xi0 y j0 ) = xi (y j x)y j0 = xi (xy −j )y j0 = xi+i0 y −j+j0
by (1.2).
(iii) Suppose that xi y j = xi0 y j0 for some i, i0 ∈ {0, 1} and j, j0 ∈ Z. Then

xi−i0 = y j0 −j , so from x 6∈ hyi it follows that xi−i0 = 1. This implies
i = i0 , and then from xi y j = xi0 y j0 we conclude that y j = y j0 .
(iv) Follows from (i) and (iii).
As a consequence of Lemma 1.69, the structure of a dihedral group G =

hx, yi with |x| = 2, xyx−1 = y −1 , and x 6∈ hyi is determined by the order of
y. More precisely, we have the following result.
25
Lemma 1.70. Let G = hx, yi and H = hz, wi be groups such that the follo-
wing hold:
• |x| = 2, xyx−1 = y −1 , and x 6∈ hyi;
• |z| = 2, zwz −1 = w−1 , and z 6∈ hwi.
Then G ∼
= H if and only if |y| = |w|.
Proof. Suppose first that G ∼ = H. Then in particular |G| = |H|. If G and H
have infinite order, then by Lemma 1.69 (iv) we must have |y| = |w| = ∞.
Similarly if |G| = |H| is finite, then |G| = 2|y| and |H| = 2|w| by Lemma
1.69 (iv), so |y| = |w|.
For the other direction, suppose that |y| = |w|. Apply Lemma 1.69
(i) – (iv) to conclude that we can define an isomorphism ϕ : G → H by
ϕ(xi y j ) = z i wj for all i, j ∈ Z. (The details are left as an exercise.)
At this point we have essentially classified the dihedral groups, but we
have not yet shown that they exist. One way to do this would be to use the
multiplication rule in Lemma 1.69 (ii), in which case the main thing to check
is that this defines an associative binary operation. We will later see this as
part of a more general construction (semidirect products), so we will instead
proceed with the following construction using symmetric groups.
We construct groups G = hx, yi such that |x| = 2, xyx−1 = y −1 , x 6∈ hyi;
for all possible values of |y|. In view of Lemma 1.70, this provides us with
all dihedral groups up to isomorphism.
• |y| = 1: Consider x, y ∈ S2 with x = (1 2) and y = (1). We denote

D2 = hx, yi.
• |y| = 2: Consider x, y ∈ S4 with x = (1 2) and y = (3 4). We denote

D4 = hx, yi.
• 3 ≤ |y| < ∞: write n = |y|. Consider x, y ∈ Sn , where bijections

x, y : {1, . . . , n} → {1, . . . , n} are defined by
x(i) = n + 1 − i for all 1 ≤ i ≤ n,
and y is the n-cycle

y = (1 2 · · · n).
We denote D2n = hx, yi. (Note that D6 = h(1 3), (1 2 3)i = S3 .)
26
• |y| = ∞: Consider x, y ∈ Sym(Z), where the bijections x, y : Z → Z

are defined by
x(i) = −i for all i ∈ Z,

y(i) = i + 1 for all i ∈ Z.
We denote D∞ = hx, yi.
In all of the cases above, it is readily checked that |x| = 2, xyx−1 = y −1 ,

x 6∈ hyi, and that |y| is as claimed. We call D∞ the infinite dihedral group,
and for n ≥ 1 we call D2n the dihedral group of order 2n.
Theorem 1.71. Let G be a dihedral group, say G = ha, bi with |a| = |b| = 2.
The following statements hold:
(i) If |ab| is infinite, then G ∼

= D∞ .
(ii) If |ab| = n is finite, then G ∼
= D2n .
Proof. By the proof of Lemma 1.68, for x = a and y = ab we have G = hx, yi

with |x| = 2, xyx−1 = y −1 , and x 6∈ hyi. Then the theorem follows from
Lemma 1.70 and the construction of D∞ and D2n .
1.8 Quaternion group Q8

An important example of a non-abelian group is the quaternion group of
order 8, which is denoted by Q8 . We construct Q8 as a group of 2 × 2
complex matrices as follows:
Q8 = {±1, ±i, ±j, ±k}
with

1 0 i 0
1= , i= ,
0 1 0 −i

0 1 0 i
j= , k= ,
−1 0 i 0
where i denotes the imaginary unit i2 = −1 in C. A calculation shows that

the following relations hold:
i2 = j 2 = k 2 = −1
27
ij = k = −ji
jk = i = −kj
ki = j = −ik
from which it follows that Q8 is a non-abelian8 subgroup of GL2 (C).
Remark 1.72. This construction is closely related to the quaternion algebra

H, which is isomorphic to the following R-algebra of matrices:

a + bi c + di
R1 ⊕ Ri ⊕ Rj ⊕ Rk = : a, b, c, d ∈ R
−c + di a − bi
Remark 1.73. We have previously seen that the dihedral group D8 is anot-
her example of a non-abelian group of order 8. However, we have Q8 ∼6= D8 .
One way to see this is to note that D8 and Q8 do not have the same number
of elements of order 2. (Exercise: For G = D8 and G = Q8 , determine the
order of each element.)
1.9 Cosets and products of subsets

Let G be a group and S, T ⊆ G. We define the product of the subsets S and
T by
ST = {st : s ∈ S, t ∈ T }.
Since the group operation is associative, we also have associativity for pro-
ducts of subsets: A(BC) = (AB)C for any subsets A, B, C ⊆ G. It follows
then that for any finite number of subsets A1 , A2 , . . ., Ak ⊆ G the product
A1 A2 · · · Ak is uniquely determined and equals {a1 a2 · · · ak : ai ∈ Ai }.
For g ∈ G and S ⊆ G, we define the left and right translates of S by g as
gS = {gs : s ∈ S},
Sg = {sg : s ∈ S},
respectively. Recall from Cayley’s theorem (Theorem 1.66) that the left
translation maps x 7→ gx and right translation maps x 7→ xg are bijections
for all g ∈ G. Thus |gS| = |Sg| = |S| for all S ⊆ G and g ∈ G; moreover
S = T if and only if gS = gT if and only if T g = Sg for all S, T ⊆ G and
g ∈ G.
8
Quaternions were discovered by Hamilton in 1843. Here is a quote from the paper
where he first defined quaternions: “... though it must, at first sight, seem strange and
almost unallowable, to define that the product of two imaginary factors in one order differs
(in sign) from the product of the same factors in the opposite order (ji = −ij)..”
28
Definition 1.74. Let H be a subgroup of a group G. A left coset is a subset

of the form gH, where g ∈ G. Similarly, a right coset is a subset of the form
Hg, where g ∈ G.
For any H ≤ G, it is clear that G is the union of left cosets: we have

[
G= gH.
g∈G
Moreover, it follows from the next two lemmas that the left cosets of H are
pairwise disjoint: for all a, b ∈ G we have either aH = bH or aH ∩ bH = ∅.
Lemma 1.75. Let H be a subgroup of a group G and let a ∈ G. Then

aH = H if and only if a ∈ H.
Proof. If aH = H, then a ∈ H since 1 ∈ H and a = a1. Conversely if

a ∈ H, then aH ⊂ H is clear and H ⊆ aH follows from h = a(a−1 h) for all
h ∈ H.
Lemma 1.76. Let H be a subgroup of a group G and let a, b ∈ G. The

following statements are equivalent:
(i) aH = bH,
(ii) b−1 a ∈ H,
(iii) a ∈ bH,
(iv) b ∈ aH,
Proof. We prove (i) ⇒ (ii) ⇒ (iii) ⇒ (iv) ⇒ (i):
(i) ⇒ (ii): We have a ∈ aH = bH, so a = bh for some h ∈ H; therefore

b−1 a ∈ H.
(ii) ⇒ (iii): If b−1 a = h ∈ H, then a = bh ∈ bH.
(iii) ⇒ (iv): If a = bh for some h ∈ H, then b = ah−1 ∈ aH.
(iv) ⇒ (i): If b = ah for some h ∈ H, then a−1 b ∈ H. By Lemma 1.75

we have a−1 bH = H, so multiplying both sides by a gives
bH = aH.
29
Analogously to Lemma 1.76, for right cosets we have Ha = Hb if and

only if ba−1 ∈ H if and only if a ∈ Hb if and only if b ∈ Ha. Therefore the
right cosets of H are also pairwise disjoint.
Definition 1.77. Let H be a subgroup of G. Then the cardinality of the
set of left cosets {aH : a ∈ G} is called the index of H in G and denoted by
[G : H]. (It is possible that [G : H] is infinite. For example, for H = {1} we
have [G : H] = |G|.)
Remark 1.78. By the axiom of choice9 , we can always find a subset S ⊆ G
such that S contains precisely one element from every left coset of H in G.
In other words:
• Any left coset of H in G is of the form aH for some a ∈ S;
• For all a, b ∈ S we have aH = bH if and only if a = b.
Then |S| = [G : H], and G can be written as the disjoint union
[
G= xH.
x∈S
Such a subset S is called a set of left coset representatives for H in G, or

a left transversal for H in G. A right transversal for right cosets of H is
defined analogously.
Example 1.79. Let G = S3 and let H be the subgroup H = {(1), (1 2)}.
Then [G : H] = 3, and the left cosets of H are
H = {(1), (1 2)}
(1 3)H = {(1 3), (1 2 3)}
(2 3)H = {(2 3), (1 3 2)}
The right cosets of H are
H = {(1), (1 2)}
H(1 3) = {(1 3), (1 3 2)}
H(2 3) = {(2 3), (1 2 3)}
Note that H is both a left and right coset, but neither H(1 3) nor H(2 3) is
a equal to a left coset.
9
The use of axiom of choice in this remark is unavoidable. It was shown in Theorem 1
of “K. Keremedis, Some equivalents of AC in algebra. II., Algebra Universalis 39 (1998),
no. 3–4, 163–169.” that the axiom of choice holds if and only if every subgroup of every
abelian group has a left transversal.
30
We could also define the index of a subgroup using right cosets, since the
set of left cosets and the set of right cosets have the same cardinality10 . In
any case, since |aH| = |H| for all a ∈ G, in the case of finite groups we get
Lagrange’s theorem:
Theorem 1.80 (Lagrange’s theorem). Let G be a finite group and let H ≤ G.

Then |H| · [G : H] = |G|. In particular, the order of H divides the order of
G.
Corollary 1.81. Let G be a finite group and g ∈ G. Then |g| divides |G|.
Proof. The claim follows from Lagrange’s theorem with H = hgi, since |H| =
|g| by Lemma 1.42 (ii).
Corollary 1.82. Let G be a finite group. Then g |G| = 1 for all g ∈ G.
Proof. Since g |g| = 1, the claim follows from Corollary 1.81.
Corollary 1.83. Let G be a finite group of prime order. Then G is cyclic.
Proof. Let x ∈ G be a nontrivial element. Since |x| > 1 divides |G|, we have
|x| = |G| since |G| is a prime number. Since hxi is a subgroup of order |x|
(Lemma 1.42 (ii)), we conclude that G = hxi.
From Corollary 1.83 and Lemma 1.43, it follows that gnu(p) = 1 for all
primes p.
Example 1.84. Let G = S3 . By Lagrange’s theorem, we know that any

subgroup of G has order 1, 2, 3, or 6. We see then that G has a total of 6
subgroups:
• Order 1: trivial subgroup {(1)};
• Order 2: three cyclic subgroups h(1 2)i, h(1 3)i, and h(2 3)i;
• Order 3: one cyclic subgroup h(1 2 3)i;
• Order 6: the group G itself.
Example 1.85. Let G = Q8 . As in the previous example, we can apply

Lagrange’s theorem to see that G has a total of 6 subgroups:
• Order 1: trivial subgroup {1};

10
You can check that aH 7→ Ha−1 is a well-defined bijection between the left cosets
and the right cosets of H in G.
31
• Order 2: cyclic subgroup {1, −1};
• Order 4: three cyclic subgroups hii, hji, and hki;
• Order 8: the group G itself.
Lemma 1.86. Let K ≤ H ≤ G. Then:
(i) [G : K] is finite if and only if both [G : H] and [H : K] are finite.
(ii) If [G : K] is finite, then [G : K] = [G : H][H : K].
Proof. Let X be a set of left coset representatives for H in G, and let Y be

a set of left coset representative for K in H (Remark 1.78).
We will show that Z = {xy : x ∈ X, y ∈ Y } is a set of left coset
representatives for K in G, and that |Z| = |X × Y |. Indeed, consider a left
coset gK, where g ∈ G. Then g ∈ xH for some x ∈ X, so g = xh for some
h ∈ H. Now h ∈ yK for some y ∈ Y , so g ∈ xyK. Thus every left coset of
H is of the form zK for some z ∈ Z.
Next we show that if xyK = x0 y 0 K for some x, x0 ∈ X and y, y 0 ∈ Y ,
then x = x0 and y = y 0 . To this end, if xyK = x0 y 0 K, then xy ∈ x0 y 0 K, and
thus x ∈ x0 H since y, y 0 ∈ H. Therefore x = x0 since X is a set of left coset
representatives for H in G. Thus xyK = xy 0 K, which implies yK = y 0 K
and so y = y 0 since Y is a set of left coset representatives for K in H.
We have seen that Z is a set of left coset representatives for K, and
moreover we have proven that [G : K] = |Z| = |X × Y |. Since |X| = [G : H]
and |Y | = [H : K], both claims (i) and (ii) follow.
Lemma 1.87. Let H and K be a subgroups of a group G. Suppose that

[G : H] and [G : K] are finite. Then [G : H ∩ K] is finite, and [G : H ∩ K] ≤
[G : H][G : K].
Proof. Each left coset of H ∩ K is the intersection of a left coset of H and a

left coset of K, since g(H ∩K) = gH ∩gK for all g ∈ G. Thus H ∩K has only
finitely many left cosets, and the total number is at most [G : H][G : K].
Lemma 1.88. Let G be a group and let H and K be finite subgroups of G.

Then
|H||K|
|HK| = .
|H ∩ K|
Proof. We have [
HK = hK
h∈H
32
and each left coset of K has size |K|, so |HK| = s|K|, where s is the number
of left cosets of the form hK with h ∈ H.
We show next that s = [H : H ∩ K]. To this end, we can apply Lemma
1.76. For x, y ∈ H we have xK = yK if and only if y −1 x ∈ H ∩ K. Therefore
xK = yK if and only if x(H ∩ K) = y(H ∩ K), from which it follows that
the sets {hK : h ∈ H} and {h(H ∩ K) : h ∈ H} have the same size. In other
words s = [H : H ∩ K], so by Lagrange’s theorem
|H||K|
|HK| = [H : H ∩ K]|K| = .
|H ∩ K|
as claimed.
Lemma 1.89. Let H and K be subgroups of a group G. Then HK is a
subgroup if and only if HK = KH.
Proof. Exercise.
1.10 Normal subgroups and quotient groups

Given a group G, we are interested in finding its subgroups and homomor-
phic images. Subgroups are found inside of the group G, and using nor-
mal subgroups we can construct all possible groups of the form ϕ(G), where
ϕ : G → H is a homomorphism.
Definition 1.90. Let G be a group and let N ≤ G. We say that N is a
normal subgroup of G, if g −1 N g = N for all g ∈ G. We denote this by
N E G.
Example 1.91. (a) In any group G, the trivial subgroup {1} and the group
G itself are normal subgroups of G.
(b) Let G be abelian. Then every subgroup of G is a normal subgroup.
(c) Let G = S3 . Then the cyclic subgroup h(1 2 3)i of order 3 is a normal
subgroup of G, but the cyclic subgroup H = h(1 2)i is not normal: we
have g −1 Hg = h(1 3)i or g −1 Hg = h(2 3)i for g ∈ G \ H.
(d) Exercise: Show that if H ≤ G with [G : H] = 2, then H E G.
(e) Exercise: Show that every subgroup of Q8 is normal.
(f) The property of being a subgroup is transitive: H ≤ K ≤ G implies
H ≤ G. The same is not true for normality. Let G = D8 = hx, yi with
|x| = 2, |y| = 4, and xyx−1 = y −1 . Then for H = hxi and K = hx, y 2 i
we have H E K E G but H 6E G.
33
Lemma 1.92. Let G be a group and let N ≤ G. The following statements

are equivalent:
(i) N is a normal subgroup;
(ii) gN = N g for all g ∈ G;
(iii) (xN )(yN ) = xyN for all x, y ∈ G.
(iv) g −1 N g ⊆ N for all g ∈ G;
Proof. We prove (i) ⇒ (ii) ⇒ (iii) ⇒ (iv) ⇒ (i):
(i) ⇒ (ii): Since g −1 N g = N for all g ∈ G, we have g(g −1 N g) = gN and

thus N g = gN for all g ∈ G.
(ii) ⇒ (iii): By associativity and (ii), we have (xN )(yN ) = x(N y)N =
x(yN )N = xyN N = xyN for all x, y ∈ G.
(iii) ⇒ (iv): For all g ∈ G, we have (g −1 N )(gN ) = N , so (g −1 N g)N = N ,

which implies g − N g ⊆ N .
(iv) ⇒ (i): We have g −1 N g ⊆ N for all g ∈ G. Applying this with g −1

this gives gN g −1 ⊆ N , so N = g −1 (gN g −1 )g ⊆ g −1 N g for all
g ∈ G. Thus g −1 N g = N for all g ∈ G.

phism. The following statements hold:
(i) Ker ϕ is a normal subgroup of G.
(ii) If N E G, then ϕ(N ) E ϕ(G).
(iii) If K E H, then ϕ−1 (K) E G.
Proof. (i) We know that Ker ϕ is a subgroup by Lemma 1.63 (i). It is a

normal subgroup since ϕ(x) = 1 implies ϕ(gxg −1 ) = ϕ(g)ϕ(x)ϕ(g)−1 =
ϕ(g)ϕ(g −1 ) = 1 for all g ∈ G.
(ii) We have ϕ(N ) ≤ ϕ(G) by Lemma 1.63 (iv). For all g ∈ G we have
ϕ(g)−1 ϕ(N )ϕ(g) = ϕ(g −1 N g) = ϕ(N ), since N is normal, so ϕ(N ) E
ϕ(G).
34
(iii) We have ϕ−1 (K) ≤ G by Lemma 1.63 (v). For all g ∈ G and x ∈
ϕ−1 (K), we have ϕ(g −1 xg) = ϕ(g)−1 ϕ(x)ϕ(g) ∈ K since ϕ(x) and
K E G. This shows that g −1 xg ∈ ϕ−1 (K) for all g ∈ G and x ∈ ϕ−1 (K),
so by Lemma 1.92 we have ϕ−1 (K) E G.
The main importance of normal subgroups is they can be used to con-

struct quotient groups, which turn out to be precisely the homomorphic ima-
ges of a group11 . This construction proceeds as follows. Let N E G. We
consider the set
G/N = {gN : g ∈ G}
of left cosets of N in G. For X, Y ∈ G/N , the product XY (product of
subsets of G) is also in G/N by Lemma 1.92 (iii). Thus we can define a
binary operation · on G/N by
X · Y = XY for all X, Y ∈ G/N.
Note that by Lemma 1.92 (iii), we have
(xN ) · (yN ) = xyN
for all x, y ∈ G. This makes it clear that (G/N, ·) is a group: associativity

holds since it holds in G, the identity element is N , and (xN )−1 = x−1 N for
all x ∈ G.
Definition 1.94. Let G be a group and N E G. We call the group (G/N, ·)

the quotient group of G modulo N .
Note that for N E G we have |G/N | = [G : N ], so by Lagrange’s theorem

|G/N | = |G|/|N | if G is finite.
Remark 1.95. We emphasize that the construction of the quotient group

G/N only makes sense when N is a normal subgroup.
Indeed, consider a subgroup H ≤ G which is not normal. By Lemma
1.92 (iv), we can find a g ∈ G be such that g −1 Hg 6⊆ H. We claim then
that the product of left cosets (g −1 H)(gH) is not a left coset. If this were
the case, then g −1 HgH = xH for some x ∈ G. But 1 ∈ g −1 HgH, so 1 ∈ xH
which implies x ∈ H. Therefore g −1 HgH = H, from which we conclude
that g −1 Hg ⊆ H, a contradiction. Therefore (g −1 H)(gH) is not a left coset,
which means that we cannot define the binary operation on G/N as above.
11
See Remark 1.99 below.
35
Example 1.96. (a) In Example 1.5, the construction of Z/nZ is precisely

the construction of the quotient group of Z by the normal subgroup nZ.
(b) For N = G, the quotient G/N = {G} is trivial.
(c) For N = {1}, the left cosets N in G are singletons {g}, g ∈ G. Multi-
plication in G/N is defined as {g} · {h} = {gh} for all g, h ∈ G, so it is
obvious that G/N ∼= G.
(d) Let G = S3 and consider the normal subgroup N = h(1 2 3)i. Then
G/N = {N, (1 2)N }, and G/N ∼
= C2 .
Theorem 1.97. Let G be a group and let N be a normal subgroup of G. Then

the map π : G → G/N defined by x 7→ xN is a surjective homomorphism
with Ker π = N .
Proof. It is clear that π is surjective, and π is a homomorphism since we

have (xN )(yN ) = xyN by definition. The fact that Ker π = N follows from
Lemma 1.75.
We will call the homomorphism π : G → G/N in Theorem 1.97 the
canonical homomorphism for N .
Theorem 1.98. Let G be a group and let ϕ : G → H be a homomorphism.

Then G/ Ker ϕ ∼
= ϕ(G), with an isomorphism given by x Ker ϕ 7→ ϕ(x).
Proof. Consider the map ψ : G/ Ker ϕ → ϕ(G) defined by ψ(x Ker ϕ) = ϕ(x)
for all x ∈ G. By Lemma 1.63 (ii) and Lemma 1.76, we have x Ker ϕ =
y Ker ϕ if and only if ϕ(x) = ϕ(y), so ψ is well-defined and injective. More-
over ψ is clearly surjective, and it is a homomorphism since ϕ is. Therefore
ψ is an isomorphism and G/ Ker ϕ ∼ = ϕ(G).
Remark 1.99. By Theorem 1.97 and Theorem 1.98, a group is isomorphic
to a homomorphic image of G if and only if it is isomorphic to G/N for some
N E G.
Theorem 1.98 is sometimes called the first isomorphism theorem. We will

next prove the rest of the standard isomorphism theorems for groups.
Theorem 1.100. Let G be a group, let H ≤ G and K E G. Then HK ≤ G,

H ∩ K E H and
HK/K ∼ = H/H ∩ K.
36
Proof. By Lemma 1.92 we have hK = Kh for all h ∈ H. Therefore HK =

KH, so HK is a subgroup by Lemma 1.89. Since H and K are subgroups
we have H ∩ K ≤ H, and moreover H ∩ K is a normal subgroup of H since
h−1 (H ∩ K)h = H ∩ h−1 Kh = H ∩ K for all h ∈ H.
Finally consider the map ϕ : H → HK/K defined by ϕ(x) = xK for
all x ∈ K. It is clear that ϕ is a homomorphism, and ϕ is surjective since
hkK = hK = ϕ(h) for all h ∈ H and k ∈ K by Lemma 1.75. By Lemma
1.75 we have Ker ϕ = H ∩ K, so
H/H ∩ K ∼
= HK/K
follows from Theorem 1.98.
Theorem 1.101. Let G be a group and let M and N be normal subgroups

of G such that M ≤ N . Then M/N is a normal subgroup of G/N and
G/N ∼
= G/M.
M/N
Proof. Let π : G → G/N be the canonical homomorphism. We define
G/N
ϕ:G→ , ϕ(g) = π(g)(M/N ) for all g ∈ G.
M/N
We leave it as an exercise to check that ϕ is a surjective homomorphism with

Ker ϕ = M . It follows then from Theorem 1.98 that
G/N
G/M ∼
= ,
M/N
as claimed.
Theorem 1.102 (Correspondence theorem). Let G be a group, N E G, and

let π : G → G/N be the canonical homomorphism. The following statements
hold:
(i) For all H ≤ G we have π(H) = HN/N .
(ii) For all N ≤ M ≤ G, we have π −1 (M/N ) = M .
(iii) If N ≤ M ≤ G, then M/N ≤ G/N .
(iv) Every subgroup of G/N is of the form M/N for a unique subgroup
N ≤ M ≤ G.
37
(v) Let N ≤ M ≤ G. Then M E G if and only if M/N E G/N .
Proof. (i) By Theorem 1.100 the product HN is a group, so we can form

the quotient HN/N . For all h ∈ H and n ∈ N we have hnN = hN =
π(h) by Lemma 1.75, so it follows that HN/N = π(H).
(ii) It is clear that M ⊆ π −1 (M/N ). Conversely if x ∈ π −1 (M/N ), then

xN = yN for some y ∈ M . This implies x ∈ yN , which gives x ∈ M
by N ≤ M . Therefore π −1 (M/N ) ⊆ M .
(iii) By (i) M/N is the image of M under π, so it is a subgroup by Lemma

1.63 (iv).
(iv) Let X ≤ G/N . Consider M = π −1 (X). It is easy to see that N ≤ M ≤

G, and moreover we have π(M ) = π(π −1 (X)) = X since π is surjective.
Thus X = π(M ) = M/N by (i). Uniqueness of M is immediate from
(ii).
(v) If M E G, then M/N = π(M ) E G/N by (i) and Lemma 1.93 (ii).
Conversely, if M/N E G/N , then M = π −1 (M/N ) E G by (ii) and
Lemma 1.93 (iii).
Definition 1.103. A group G is said to be simple if G 6= {1} and the only

normal subgroups of G are {1} and G.
Example 1.104. The abelian simple groups are known:
(a) If G is cyclic of prime order, then G has only two subgroups; {1} and G
itself. Therefore G is simple.
(b) Exercise: If G is abelian, then G is simple if and only if G is cyclic of

prime order.
Remark 1.105. Suppose that G is a non-trivial finite group. Let N1 / G be

a proper normal subgroup such that |N1 | is as large as possible. Then there
are no normal subgroups of G with N1 ≤ M E G other than M = G and
M = N1 , so by the correspondence theorem G/N1 is simple. If N1 = {1} then
G is simple, otherwise by the same argument we can find a proper normal
subgroup N2 / N1 such that N1 /N2 is simple. Continuing in this manner, we
eventually construct a series
G = N0 . N1 . N2 . · · · . Nt . Nt+1 = {1}
38
such that Ni /Ni+1 is simple for all 0 ≤ i ≤ t. Such a series is called a com-
position series. (In a later section we will prove the Jordan-Hölder theorem,
which states that the isomorphism types of the simple factors occurring in a
composition series are unique.)
So a finite group is made up of simple groups, and this suggests an

inductive approach towards understanding all finite groups (“Hölder pro-
gram”):
(1) Determine all finite simple groups.
(2) (Extension problem) Given two finite groups A and B, determine all
finite groups G with a normal subgroup N such that N ∼ = A and
G/N ∼ = B.
One of the biggest achievements in group theory is the solution to (1):

the classification of finite simple groups (CFSG). It has been shown that any
finite simple group is isomorphic to one of the following:
• Cyclic of prime order;
• Alternating group An with n ≥ 5; (defined in a later section)
• A group of Lie type;
• One of the 26 sporadic simple groups. (See Table 2.)
The original proof of the classification consisted of hundreds of articles

totalling tens of thousands of pages, starting from around 1955 with the
Brauer–Fowler theorem and finishing in 2004 with the Aschbacher–Smith
volume on quasithin groups. A second-generation proof led by Gorenstein–
Lyons–Solomon is in progress, and as of 2021 a total of 9 volumes have been
published (AMS Mathematical Surveys and Monographs).
What about the extension problem? This is completely infeasible in gene-
ral, one reason being that we cannot even expect to classify all finite p-groups.
However, there are many interesting things to be said about extensions of
two groups A and B — meaning groups G with a normal subgroup N such
that N ∼ = A and G/N ∼ = B. We will later find various ways of constructing
and recognizing such extensions.
39
Group Order Year
M11 24 · 32 · 5 · 11 1895
M12 26 · 33 · 5 · 11 1899
M22 27 · 32 · 5 · 7 · 11 1900
M23 27 · 32 · 5 · 7 · 11 · 23 1900
M24 210 · 33 · 5 · 7 · 11 · 23 1900
J1 23 · 3 · 5 · 7 · 11 · 19 1966
J2 27 · 33 · 52 · 7 1967
J3 27 · 35 · 5 · 17 · 19 1969
J4 221 · 33 · 5 · 7 · 113 · 23 · 29 · 31 · 37 · 43 1975
Co1 221 · 39 · 54 · 72 · 11 · 13 · 23 1969
Co2 218 · 36 · 53 · 7 · 11 · 23 1969
Co3 210 · 37 · 53 · 7 · 11 · 23 1969
Fi22 217 · 39 · 52 · 7 · 11 · 13 1969
Fi23 218 · 313 · 52 · 7 · 11 · 13 · 17 · 23 1969
Fi024 221 · 316 · 52 · 73 · 11 · 13 · 17 · 23 · 29 1969
HS 29 · 32 · 53 · 7 · 11 1968
McL 27 · 36 · 53 · 7 · 11 1969
He 210 · 33 · 52 · 73 · 17 1969
Ru 214 · 33 · 53 · 7 · 13 · 29 1972
Suz 213 · 37 · 52 · 7 · 11 · 13 1969
O’N 29 · 34 · 5 · 73 · 11 · 19 · 31 1973
HN 214 · 36 · 56 · 7 · 11 · 19 1974
Ly 28 · 37 · 56 · 7 · 11 · 31 · 37 · 67 1971
Th 215 · 310 · 53 · 72 · 13 · 19 · 31 1974
B 241 · 313 · 56 · 72 · 11 · 13 · 17 · 19 · 23 · 31 · 47 1974
M 246 · 320 · 59 · 76 · 112 · 133 · 17 · 19 · 23 · 29 · 31 · 41 · 47 · 59 · 71 1974
Table 2: The sporadic simple groups, their orders, and the year when they
were first discovered.
40
1.11 Conjugacy classes of elements

Definition 1.106. Let G be a group. We say that two elements g, h ∈ G are
conjugate in G, if there exists x ∈ G such that h = x−1 gx. We will denote
g x := x−1 gx.
Two conjugate elements of a group are “similar” and many of their pro-
perties (as elements of G) are the same. For example, they must have the
same order (Lemma 1.65 (ii)). As seen in the next example, in general linear
groups conjugacy corresponds to the similarity of matrices and to change of
basis.
Example 1.107. Let V be a finite-dimensional vector space over a field F.
Let f, g ∈ GL(V ). Let e1 , . . ., en be a basisPof V and let (Aij ) be the matrix
of f with respect to this basis, so f (ei ) = nj=1 Aji ej for all 1 ≤ i ≤ n.
Suppose that g is conjugate to f , say g = p−1 f p for some p ∈ GL(V ). Let
e01 , . . ., e0n be the basis of V defined 0 −1
Pn by 0ei := p (ei ) for all 1 ≤ i ≤ n. Then
0
it is immediate that g(ei ) = j=1 Aji ej for all 1 ≤ i ≤ n. In other words,
the matrix of g is the same as that of f , just with respect to a different basis.
Conversely, if g is obtained from f by a change of basis, then g and f are
conjugate in GL(V ).
For g, h ∈ G, denote g ∼ h if g and h are conjugate in G. It is easy to
see that ∼ is an equivalence relation:
• g ∼ g, since g = x−1 gx for x = 1G ;
• If g ∼ g 0 with g 0 = x−1 gx, then g = y −1 g 0 y with y = x−1 ; thus g 0 ∼ g;
• If g ∼ g 0 and g 0 ∼ g 00 with g 0 = x−1 gx and g 00 = y −1 g 0 y, then g 00 =
(xy)−1 gxy; thus g ∼ g 00 .
The equivalence class of g ∈ G under ∼ is called the conjugacy class of
x in G and denoted by xG . Then G is partitioned into a disjoint union of
conjugacy classes, and the conjugacy class of x in G is
xG = {g −1 xg : g ∈ G}.
An element y ∈ xG is called a representative of the conjugacy class xG .
Example 1.108. If G is an abelian group, then xG = {x} for all x ∈ G.
Example 1.109. Let G = Sn and consider a k-cycle σ = (i1 · · · ik ). As an
exercise, show that for all g ∈ Sn , we have
g(i1 · · · ik )g −1 = (g(i1 ) · · · g(ik )).
Conclude that σ G = {g ∈ G : g is a k-cycle}.
41
Example 1.110. Let G = S3 . Then G has three conjugacy classes: {(1)}

(trivial element), {(1 2), (1 3), (2 3)} (2-cycles), and {(1 2 3), (1 3 2)} (3-
cycles).
Example 1.111. Let G = Q8 = {±1, ±i, ±j, ±k}. Exercise: determine the
conjugacy classes of Q8 (there are a total of five).
Lemma 1.112. Let G be a group and H ≤ G. Then H E G if and only if
H is a union of conjugacy classes of G.
Proof. If H is a union of conjugacy classes of G, then H E G since gC g −1 =
C for any conjugacy class C and g ∈ G. Conversely, suppose that H E G.
Then for all x ∈ H we have xG ⊆ H since gxg −1 ∈ H for all g ∈ G, and
therefore H is the union of the conjugacy classes of its elements.
Definition 1.113. Let G be a group and x ∈ G. The centralizer of x in G
is CG (x) = {g ∈ G : gx = xg}, i.e., the set of all elements of G that commute
with x.
Lemma 1.114. Let G be a group. Then:
(i) CG (x) is a subgroup of G for all x ∈ G.
(ii) |xG | = [G : CG (x)] for all x ∈ G. In particular if G is finite, then |xG |

divides |G|.
Proof. (i) It is clear that 1 ∈ CG (x) for all x ∈ G. Suppose that g, h ∈
CG (x). Then (gh)x = g(hx) = g(xh) = (gx)h = (xg)h = x(gh),
so gh ∈ CG (x) and CG (x) is closed under multiplication. We have
gx = xg, and multiplying this equation from the left and the right with
g −1 gives xg −1 = g −1 x, in other words g −1 ∈ CG (x). Thus CG (x) is
closed under taking inverses and is a subgroup.
(ii) For a, b ∈ G we have axa−1 = bxb−1 if and only if (b−1 a)x = x(b−1 a)−1 ,
i.e. b−1 a ∈ CG (x). Therefore axa−1 = bxb−1 if and only if aCG (x) =
bCG (x), so the map axa−1 7→ aCG (x) is a well-defined bijection between
the sets xG and {aCG (x) : a ∈ G}. Thus |xG | = [G : CG (x)].
Definition 1.115. Let G be a group. The center of G is defined as
Z(G) = {g ∈ G : gx = xg for all x ∈ G}.
Note that Z(G) is a subgroup, since it is the intersection of all centralizers

of G.
42
Lemma 1.116. Let G be a group. Then Z(G) E G, and any subgroup of

Z(G) is a normal subgroup of G.
Proof. Follows from the fact that g −1 xg = x for all x ∈ Z(G).
Example 1.117. Some basic facts:
(i) A group G is abelian if and only if G = Z(G).
(ii) For g ∈ G, we have g ∈ Z(G) if and only if g G = {g}.
(iii) Exercise: For all n ≥ 3, we have Z(Sn ) = {1}.
(iv) Let G = S3 . Then CG (x) = hxi for all x ∈ G \ {1}. For G = S4 and
x = (1 2), we have CG (x) = {(1), (1 2), (3 4), (1 2)(3 4)}.
(v) Exercise: If G/Z(G) is cyclic, then G is abelian.
(vi) Exercise: Let N E G be such that |N | = 2. Prove that N ≤ Z(G).
Definition 1.118. For any subset S ⊂ G, we define the centralizer of S as
CG (S) = {g ∈ G : gx = xg for all x ∈ S}.
Similarly to the center, we see that CG (S) is a subgroup since it is the inter-
section of the centralizers of elements of S. In particular, we have CG (H) ≤ G
for any H ≤ G.
Proposition 1.119 (Class equation). Let G be a finite group. Let x1 , . . ., xt

be representatives for the conjugacy classes of G with more than one element.
Then
Xt
|G| = |Z(G)| + [G : CG (xi )].
i=1
Proof. Since a conjugacy class xG has size 1 if and only if x ∈ Z(G), we have
a disjoint union
[t
G = Z(G) ∪ xGi .
i=1
Pt Pt
Thus |G| = |Z(G)| + i=1 |xG
i | = |Z(G)| + i=1 [G : CG (xi )] by Lemma
1.114.
43
1.12 Cauchy’s theorem

As an application of the class equation, we will prove Cauchy’s theorem.
Theorem 1.120 (Cauchy’s theorem). Let p be a prime. If G is a finite
group such that p | |G|, then G contains an element of order p.
Proof. By induction on |G|, the theorem being vacuously true when |G| = 1.
We can suppose then that |G| > 1 and p | |G|. By the class equation, we
have
t
X
|G| = |Z(G)| + [G : CG (xi )], (1.3)
i=1
where x1 , . . ., xt are representatives for the non-central conjugacy classes,
i.e. xi 6∈ Z(G).
If p | |CG (xi )| for some 1 ≤ i ≤ t, then we are done by applying induction
on CG (xi ).
Thus we may assume that p - |CG (xi )| for all 1 ≤ i ≤ t, in which case
we must have p | [G : CG (xi )] for all 1 ≤ i ≤ t. By (1.3) it follows that
p | |Z(G)|. If Z(G) 6= G we are again done by applying induction on Z(G).
Thus we can assume that G = Z(G), in which case G is abelian. Let
x ∈ G be a non-identity element. If p divides |x|, then a suitable power of
x will be an element of order p (Lemma 1.27 (iii)). If p does not divide |x|,
then p must divide the order of G/hxi. In this case, by induction the image
of some g ∈ G in G/hxi has order p. Then |g| is divisible by p, so a suitable
power of g will be an element of order p in G.
Remark 1.121. Can we generalize Cauchy’s theorem? That is, suppose that
n is an integer that divides |G|, not necessarily a prime. Must G contain an
element of order n? Certainly this is not true in general, for example if G is
non-cyclic of order n. (Explicit example: G = S3 has order divisible by 6,
but G does not contain an element of order 6.) Problem: Suppose that n is
not a prime number. Does there exist a finite group G such that n divides
|G|, but G has no element of order n?
As an example application of Cauchy’s theorem, we will classify groups
of order 2p, where p is a prime.
Theorem 1.122. Let G be a group such that |G| = 2p, where p > 2 is a
prime. Then G ∼
= C2p or G ∼
= D2p .
Proof. By Cauchy’s theorem, there exists x, y ∈ G with |x| = 2 and |y| = p.
We have hyi/G since hyi has index 2 in G (Example 1.91 (iv)), so xyx−1 = y i
for some i ∈ Z. Since x2 = 1, we have
2
y = x2 yx−2 = xy i x−1 = (xyx−1 )i = y i ,
44
so i2 ≡ 1 mod p. Since p is prime, this implies i ≡ ±1 mod p. If i ≡ 1

mod p, then xy = yx. Thus |xy| = 2p by Lemma 1.29, so G = hxyi ∼
= C2p . If
−1 −1
i ≡ −1 mod p, then xyx = y . In this case G is dihedral (Lemma 1.68),
so by Theorem 1.71 we have G ∼= D2p .
Since D6 = S3 , in particular we have proven the following result12 .
Corollary 1.123 (Cayley, 1854). Let G be a group such that |G| = 6. Then
G∼
= C6 or G ∼= S3 .
Cauchy’s theorem tells us that if p is a prime dividing the order of a finite
group G, then G contains an element of order p. What can we say about the
number of elements of order p?
Theorem 1.124. Let G be a finite group and let p be a prime such that
p | |G|. Then the number of x ∈ G such that xp = 1 is a multiple of p.
Proof. By Cauchy’s theorem, there exists an element of order p in G, which

generates subgroup U ≤ G of order |U | = p. We will use U to partition the
set Ω = {x ∈ G : xp = 1} into pieces of size p, which implies |Ω| ≡ 0 mod p.
For x ∈ G, define
(
xU, if x ∈ CG (U ).
[x] := −1
{uxu : u ∈ U }, if x 6∈ CG (U ).
It is an exercise (Exercise 1.33) to prove that the sets [x] partition G, in

other words for all x, y ∈ G we have [x] = [y] or [x] ∩ [y] = ∅. Moreover, [x]
contains p elements for all x ∈ G. Thus the sets [x] partition G into pieces
of size p.
We now claim that the sets [x] also partition Ω, which proves the theorem.
To this end, it will suffice to prove that for x ∈ Ω, we have [x] ⊆ Ω. Indeed,
if x ∈ CG (U ), then (xu)p = xp up = 1 for all u ∈ U , so [x] = xU ⊆ Ω.
If x 6∈ CG (U ), then (uxu−1 )p = uxp u−1 = 1, so uxu−1 ∈ Ω for all u ∈ U .
Therefore [x] = {uxu−1 : u ∈ U } ⊆ Ω.
12
Proven by Cayley in 1854. Two decades later, in a paper (American Journal of
Mathematics , Vol. 1, No. 1 (1878), pp. 50–52) Cayley states (mistakenly) the following:
“... if n = 6, there are three groups, a group 1, α, α2 , α3 , α4 , α5 (α6 = 1); and two
groups 1, β, β 2 , α, αβ, αβ 2 (α2 = 1, β 3 = 1), viz: in the first of these αβ = βα; while in
the other of them we have αβ = β 2 α ...”
45
Remark 1.125. A more general result was proven by Frobenius in 1895. He

showed that if G is a finite group and n | |G|, then the number of x ∈ G such
that xn = 1 is a multiple of n. There are several proofs of this Frobenius’
theorem, but none of them are very short.
Frobenius also conjectured that if n divides |G| and if the number of so-
lutions to xn = 1 is exactly n, then the set of solutions forms a subgroup.
Almost 100 years later, Frobenius’ conjecture was proven by Iiyori and Ya-
maki (1991), using the classification of finite simple groups.
Recall the notation od (G) for the number of elements of order d in G.
Corollary 1.126. Let G be a finite group and let p be a prime such that
p | |G|. Then op (G) ≡ −1 mod p.
Proof. Follows from Theorem 1.124, since xp = 1 if and only if x = 1 or
|x| = p.
Remark 1.127. Corollary 1.126 can also be improved in some cases. For
example, if p2 | |G|, then one can show13 that op (G) ≡ p − 1 mod p2 or
op (G) ≡ −1 mod p2 . The precise value of op (G) modulo p2 depends on the
structure of p-subgroups of G.
Corollary 1.128. Let G be a finite group and let p be a prime such that
p | |G|. Then the number of subgroups of order p in G is ≡ 1 mod p.
Proof. Let r be the number of subgroups of order p in G. Note that if U
and V are subgroups of order p in G, by Lagrange’s theorem we have U = V
or U ∩ V = {1}. Moreover, any element of U \ {1} has order p. Thus the
number of elements of order p in G is equal to r(p − 1), so r ≡ 1 mod p by
Corollary 1.126.
1.13 Number of conjugacy classes

Let G be a finite group. Denote the number of conjugacy classes of G by
k(G). What can we say about k(G)? Certainly k(G) ≤ |G|, and k(G) = |G|
if and only if G is abelian. What about lower bounds for k(G)? The first
results go back to Landau14 , who showed in 1903 that |G| can be bounded
from above in terms of k(G). Using Landau’s method, one can find the
following explicit bound15 .
13
M. Herzog, Counting group elements of order p modulo p2 , Proc. Amer. Math. Soc.
66 (1977), 247–250.
14
E. Landau, Über die Klassenzahl der binären quadratischen Formen von negativer
Discriminante., Math. Ann. 56 (1903), no. 4, 671–676.
15
M. Newman, A bound for the number of conjugacy classes in a group., J. London
Math. Soc. 43 (1968), 108–110.
46
log log n
Theorem 1.129. Let G be a finite group of order n. Then k(G) > log 2
.
Proof. Omitted.
The bound in Theorem 1.129 is very weak, and later much better bounds
have been found. But it does follow from Theorem 1.129 that k(G) → ∞ as
|G| → ∞. Therefore for all c > 0, up to isomorphism there exist only finitely
many finite groups G with k(G) = c.
Example 1.130. Let G be a finite group.
(a) It is clear that k(G) = 1 only if G is trivial.
(b) Exercise: Show that k(G) = 2 if and only if G ∼

= C2 .
(c) Exercise: Show that k(G) = 3 if and only if G ∼

= C3 or G ∼
= S3 .
(d) Currently the finite groups with k(G) ≤ 14 have been classified16 .
(e) Denote by f (c) the number of finite groups (up to isomorphism) with
k(G) = c. Then we have the following values of f (c):
c 1 2 3 4 5 6 7 8 9 10 11 12
f (c) 1 1 2 4 8 8 12 21 26 38 35 32
Remark 1.131. For all c > 1, it is possible to construct an infinite group G
with k(G) = c.
1.14 Conjugacy classes of subgroups

Similarly to the conjugacy of elements, we will define conjugacy of subgroups.
Definition 1.132. Let G be a group and H, K ≤ G. A conjugate of H is a

subgroup of the form g −1 Hg for g ∈ G. We say that the subgroups H and K
are conjugate, if K = g −1 Hg for some g ∈ G. We will denote H g := g −1 Hg
for all g ∈ G.
We first observe that a conjugates of a subgroup H are subgroups and

isomorphic to H.
Lemma 1.133. Let G be a group and H ≤ G. Then g −1 Hg is a subgroup

of G and g −1 Hg ∼
= H for all g ∈ G.
16
A. Vera-López, J. Sangroniz, The finite groups with thirteen and fourteen conjugacy
classes, Mathematische Nachrichten, 280(5–6), 676–694, (2007).
47
Proof. Since the map x 7→ g −1 xg is an automorphism (Lemma 1.65), the

lemma follows from Lemma 1.63 (iv) and (vi).
Example 1.134. If H and K are conjugate subgroups of a group G, then

H ∼= K by Lemma 1.133. The converse is not true, as seen by the fol-
lowing example. Let G = S4 and consider H = {(1), (1 2)} and K =
{(1), (1 2)(3 4)}. Then H and K are both cyclic of order two, so H ∼
= K.
However, H and K are not conjugate in G; for σ ∈ G we have σHσ −1 =
{(1), (σ(1) σ(2))} =
6 K.
Two isomorphic groups have the same multiplication table, so as groups

we can think of them “as the same”. But as subgroups, they might behave
in very different ways; this is because any given group can usually be repre-
sented in many different ways. For subgroups, conjugacy is better notion of
similarity than isomorphism.
Example 1.135. Consider G = D8 , so G = hx, yi with |x| = 2, |y| = 4,

xyx−1 = y −1 and x 6∈ hyi. Consider H = hxi and K = hy 2 i. Both H and K
are cyclic of order 2, so H ∼
= K. Now H is not a normal subgroup of G, but
K E G since K = Z(G).
Definition 1.136. Let G be a group and H ≤ G. The normalizer of H in

G is defined as
NG (H) = {g ∈ G : g −1 Hg = H}.
(i) NG (H) is a subgroup of G for all H ≤ G.
(ii) The number of conjugates of H in G is equal to [G : NG (H)].
Proof. The proofs of (i) is similar to the proofs of Lemma 1.114 (i). For (ii),
show that xNG (H) = yNG (H) if and only if xHx−1 = yHy −1 and argue as
in Lemma 1.114 (ii). The details are left as an exercise.
Remark 1.138. Let G be a group and H ≤ G. It is clear that H E NG (H),

and moreover NG (H) is the largest subgroup of G that contains H as a
normal subgroup. In particular H E G if and only if NG (H) = G.
Lemma 1.139. Let G be a group and N E G. Let N ≤ H ≤ G. Then

NG/N (H/N ) = NG (H)/N .
Proof. Exercise.
48
1.15 p-groups
Definition 1.140. Let p be a prime. A group G is said to be a p-group, if
|g| is a power of p for all g ∈ G.
By Lagrange’s theorem (Corollary 1.81) and Cauchy’s theorem (Corollary

1.120), a finite group G is a p-group if and only if |G| is a power of p.
Example 1.141. Let p be a prime.
(a) The cyclic group of order pk is a p-group for all k ≥ 0.
(b) Q8 is a 2-group of order 8.
(c) If n is a power of 2, then D2n is a 2-group. The infinite dihedral group

D∞ is not a 2-group, since it contains elements of infinite order.
(d) (Prüfer p-group) The following is an example of an infinite p-group. Con-

sider the following set of complex numbers:
k
Z(p∞ ) = {z ∈ C : z p = 1 for some k ≥ 1}.
Then Z(p∞ ) is a subgroup of C× , and Z(p∞ ) is an infinite abelian p-

group.
Lemma 1.142. Let G be a finite p-group and let N E G. If N 6= {1}, then

N ∩ Z(G) 6= {1}.
Proof. Since N is a normal subgroup, it must be a union of conjugacy classes

of G (Lemma 1.112), say
N = {1} ∪ C1 ∪ · · · ∪ Ct .
The order of a conjugacy class divides the order of the group (Lemma 1.114
(ii)), so |Ci | is a power of p for all 1 ≤ i ≤ t.PThus we cannot have |Ci | > 1
for all 1 ≤ i ≤ t, as otherwise |N | = 1 + ti=1 |Ci | is not divisible by p.
Therefore Ci = {x} for some 1 ≤ i ≤ t, in which case x ∈ N ∩ Z(G) and the
lemma follows.
The following is the special case N = G of Lemma 1.142.
Corollary 1.143. Let G be a non-trivial finite p-group. Then Z(G) 6= {1}.
Proposition 1.144. Let G be a finite p-group of order p2 . Then G is abelian.
49
Proof. By Corollary 1.143 we have Z(G) 6= {1}, so G/Z(G) has order 1 or

order p. Therefore G/Z(G) is cyclic, which by an exercise implies that G is
abelian (Example 1.117 (v)).
Example 1.145. For any prime p, there exist non-abelian p-groups of order
p3 . For this we can consider the Heisenberg group
  
 1 a b 
Hp = 0 1 c  : a, b, c ∈ Z/pZ .
0 0 1
 
Exercise: Show that Hp is a non-abelian subgroup of GL3 (Z/pZ) such that

|Hp | = p3 . Is H2 ∼
= Q8 or H2 ∼
= D8 ?
Proposition 1.146. Let G be a finite p-group of order |G| = pα . Then for
all 0 ≤ β ≤ α, there exists N E G with |N | = pβ .
Proof. We proceed by induction on |G|, the case |G| = 1 being clear. Suppose
then that |G| > 1. If β = 0, we can take N = {1}, so suppose that β > 0.
By Corollary 1.143 there exists a non-identity element x ∈ Z(G). Replacing
x with a suitable power of x, we can assume that |x| = p. Then K = hxi is
subgroup of order p, and K E G since x ∈ Z(G). By induction there exists
N/K E G/K with |N/K| = pβ−1 . Then N E G with |N | = pβ . (Remember
Theorem 1.102.)
Proposition 1.147. [“Normalizers grow”] Let G be a finite p-group and H

be a proper subgroup of G. Then H NG (H).
Proof. By induction on |G|, the case |G| = 1 and |G| = p are clear. Suppose
then that |G| > p. By Corollary 1.143 we have Z(G) 6= {1}. Thus if
Z(G) ≤ H, then by induction
H/Z(G) NG/Z(G) (H/Z(G)) = NG (H)/Z(G),
and thus H NG (H). If Z(G) 6≤ H, then the claim follows since Z(G) ≤
NG (H).
Proposition 1.148. Let G be a finite p-group and H ≤ G. If [G : H] = p,

then H E G.
Proof. By Proposition 1.147, we have H NG (H). If [G : H] = p, this

implies NG (H) = G, so H E G.
50
2 Symmetric and alternating Groups

In this section, we will establish some basic facts about finite symmetric
groups Sn and alternating groups An . For example, we will describe the
conjugacy classes in both Sn and An . We will also show that An is a non-
abelian simple group if n ≥ 5.
2.1 Cycle decomposition

Definition 2.1. Let σ = (i1 · · · ik ) and τ = (j1 · · · j` ) be cycles in Sn . We
say that σ and τ are disjoint, if {i1 , . . . , ik } ∩ {j1 , . . . , j` } = ∅.
Lemma 2.2. Let σ = (i1 · · · ik ) and τ = (j1 · · · j` ) be cycles in Sn . If σ
and τ are disjoint, then στ = τ σ.
Proof. Exercise.
We will see that any permutation σ ∈ Sn can be written as a product of
pairwise disjoint cycles, and that this factorization is essentially unique.
Definition 2.3. Let σ ∈ Sn . For 1 ≤ i ≤ n, the orbit of i under σ is the set
orbσ (i) = {σ k (i) : k ∈ Z}.
Lemma 2.4. Let σ ∈ Sn . Then any two orbits of σ are either disjoint or
equal. In particular, Ω = {1, . . . , n} decomposes into a disjoint union
Ω = orbσ (i1 ) ∪ · · · ∪ orbσ (it )
for some i1 , . . . , it ∈ Ω.
Proof. Let orbσ (i) and orbσ (j) be two orbits of σ. If they are not disjoint,
there exist some k, ` ∈ Z such that σ k (i) = σ ` (j). Then i = σ `−k (j) which
implies orbσ (i) ⊆ orbσ (j), similarly j = σ k−` (i) which implies orbσ (j) ⊆
orbσ (i). Therefore orbσ (i) = orbσ (j).
Example 2.5. For example, consider σ ∈ S3 . Then:
• σ = (1) has three orbits: {1}, {2}, {3}.
• σ = (1 2) has two orbits: {1, 2}, {3}.
• σ = (1 3) has two orbits: {1, 3}, {2}.
• σ = (2 3) has two orbits: {1}, {2, 3}.
• σ = (1 2 3) and σ = (1 3 2) have a single orbit: {1, 2, 3}.
51
Lemma 2.6. Let σ ∈ Sn . Let O = orbσ (i) be an orbit of σ. Let k ≥ 1 be

minimal such that σ k (i) = i. Then:
(i) |O| = k and O = {i, σ(i), . . . , σ k−1 (i)}.
(ii) σ(O) = O, and the map σ 0 : O → O induced by σ is the k-cycle
(i σ(i) · · · σ k−1 (i)) in Sym(O).
Proof. (i) For any ` ∈ Z, we can write ` = r + qk for some 0 ≤ r < k.
Then if σ ` (i) = i, we have σ r (i) = i since σ k (i) = i. By minimality
of k this implies r = 0. Therefore σ ` (i) = i if and only if k | `. As a
consequence, for all s, t ∈ Z we have σ s (i) = σ t (i) if and only if s ≡ t
mod k. Thus O = {i, σ(i), . . . , σ k−1 (i)} and |O| = k.
(ii) Immediate from (i).
Proposition 2.7. Let σ ∈ Sn . Then σ can be written as a product of pairwise

disjoint cycles: we can write σ = π1 · · · πt such that:
(i) πi is a ki -cycle for all 1 ≤ i ≤ t;
(ii) The cycles π1 , . . ., πt are pairwise disjoint;
(iii) k1 + · · · + kt = n.
This factorization is also unique, up to the order of factors: if σ =
· · · πs0 is another factorization satisfying properties (i) – (iii), then t = s
π10
and {π1 , . . . , πt } = {π10 , . . . , πt0 }.
Proof. Let O1 = orbσ (i1 ), . . ., Ot = orbσ (it ) be the orbits of σ on Ω =
{1, . . . , n}, so by Lemma 2.4 we have a disjoint union Ω = O1 ∪ · · · ∪ Ot . For
1 ≤ r ≤ t, let kr = |Or |. It follows from Lemma 2.6 that σ is equal to the
following product of disjoint cycles:
σ = (i1 σ(i1 ) · · · σ k1 −1 (i1 )) · · · (it σ(it ) · · · σ kt −1 (it )).
Thus by defining πr = (ir σ(ir ) · · · σ kr −1 (ir )) for 1 ≤ r ≤ t, we have
σ = π1 · · · πt such that conditions (i) – (iii) hold.
Suppose that σ = π10 · · · πs0 is another factorization satisfying properties
(i) – (iii), say πr0 = (j1 j2 · · · j`r ) for 1 ≤ r ≤ s. Then it is clear that each
set Or0 = {j1 , j2 , . . . , j`r } is an orbit of σ on Ω, and O10 , . . ., Os0 are all the
orbits of σ by (iii). Since the orbits are uniquely determined by σ, it follows
that s = t and by reordering the factors (Lemma 2.2) we can assume that
O1 = O10 , . . ., Ot = Ot0 . Then by Lemma 2.6 (ii) we have πi = πi0 for all
1 ≤ i ≤ t.
52
Definition 2.8. By Proposition 2.7, for each σ ∈ Sn there exist unique

integers k1 ≥ · · · ≥ kt ≥ 1 such that k1 + · · · + kt = n and σ is a product of
pairwise disjoint cycles of lengths k1 , . . ., kt . We call (k1 , . . . , kt ) the partition
of n corresponding to σ.
Example 2.9. (a) In S3 , each element is a cycle.
(b) In S4 , the possible cycle decompositions are as follows, where {1, 2, 3, 4} =

{i, j, k, l}:
• (i)(j)(k)(l) = (1) (trivial element). Partition (1, 1, 1, 1).

• (i j)(k)(l) = (i j) (2-cycle). Partition (2, 1, 1).
• (i j)(k l) (product of disjoint 2-cycles). Partition (2, 2).
• (i j k)(l) = (i j k) (3-cycle). Partition (3, 1).
• (i j k l) (4-cycle). Partition (4).
Remark 2.10. In the factorization σ = π1 · · · πt in Proposition 2.7, the

cycles πi of length 1 correpond to the orbits of length 1, in other words, to
the fixed points of σ. Thus we can write σ = π1 · · · πs , where s ≥ 0 and
the πi are pairwise disjoint cycles of length > 1. By Proposition 2.7, this
decomposition is unique up to the ordering of the factors.
Corollary 2.11. The symmetric group Sn is generated by cycles.
Proof. Immediate from Proposition 2.7.

On the other hand, any cycle can be written as a product of 2-cycles:
(a1 a2 · · · ak ) = (a1 ak )(a1 ak−1 ) · · · (a1 a2 ).
Therefore we have the following corollary:
Corollary 2.12. The symmetric group Sn is generated by 2-cycles.
Not all 2-cycles are needed to generate Sn , as for example S3 = h(1 2), (2 3)i.
Example 2.13. Some more examples on generators of Sn :
(a) Exercise: Show that Sn is generated by (1 2), (1 3), . . ., (1 n).
(b) Exercise: Show that Sn is generated by (1 2), (2 3), . . ., (n − 1 n).
(c) Exercise: Show that Sn = h(1 2), (1 2 · · · n)i.
53
2.2 Orders of permutations

The order of a permutation can be conveniently expressed in terms of its
cycle decomposition.
Lemma 2.14. Let σ ∈ Sn be a k-cycle. Then:
(i) |σ| = k.
(ii) Let d | k. Then σ d is a product of d pairwise disjoint cycles of length

k/d.
Proof. Exercise.
Example 2.15. The 4-cycle (1 2 3 4) has order 4 and (1 2 3 4)2 = (1 3)(2 4).
Lemma 2.16. Let σ ∈ Sn , and suppose that σ = π1 · · · πt where πi are

pairwise disjoint cycles, and πi is a ki -cycle for 1 ≤ i ≤ t. Then |σ| =
lcm(k1 , . . . , kt ).
Proof. Since the cycles are pairwise disjoint, they commute pairwise, and
therefore σ d = π1d · · · πtd for all d ∈ Z. By Lemma 2.14 and the uniqueness
of the cycle decomposition, we have σ d = 1 if and only if πid = 1 for all
1 ≤ i ≤ t. Since |πi | = ki (Lemma 2.14), it follows that σ d = 1 if and only if
ki | d for all 1 ≤ i ≤ t. Since lcm(k1 , . . . , kt ) is precisely the smallest positive
integer such that ki | lcm(k1 , . . . , kt ) for all 1 ≤ i ≤ t, the lemma follows.
Example 2.17. In S5 , the element (1 2)(3 4 5) has order 6, while (1 2)(3 4)

has order 2.
Example 2.18. What is the largest order of an element of Sn ? For example

in S3 , the largest order is 3, for a 3-cycle. In S5 , the largest order is 6, for
example for (1 2 3)(4 5). Denote the maximal order of an element of Sn by
g(n). Below is a list of small values:
n 1 2 3 4 5 6 7 8 9 10 11 12 13 14
g(n) 1 2 3 4 6 6 12 15 20 30 30 60 60 84
Exercise: Using the values of g(n) given above, find an element of largest
order in Sn for 1 ≤ n ≤ 14.
54
2.3 Conjugacy classes in Sn

Another application of the cycle decomposition is the description of the con-
jugacy classes of Sn .
Proposition 2.19. Let σ, τ ∈ Sn . Then σ and τ are conjugate if and only

if they correspond to the same partitions of n.
Proof. Suppose that σ = π1 · · · πt such that πi a ki -cycle and that the cycles
πi are pairwise disjoint. For conjugation of k-cycles in Sn , we have
g(i1 · · · ik )g −1 = (g(i1 ) · · · g(ik )) (2.1)
for all g ∈ Sn . It follows from (2.1) that
gσg −1 = (gπ1 g −1 )(gπ2 g −1 ) · · · (gπt g −1 )
is a product of pairwise disjoint cycles of lengths k1 , . . ., kt . By uniqueness

of the cycle decomposition (Proposition 2.7), we conclude that conjugate
elements in Sn correspond to the same partition of n.
Conversely, suppose that σ, τ ∈ Sn correspond to the same partitions
of n. In other words, there exist integers k1 ≥ · · · ≥ kt > 0 such that
k1 + · · · + kt = n and
σ = (a11 a12 · · · a1k1 )(a21 · · · a2k2 ) · · · (at1 · · · atkt )

τ = (b11 b12 · · · b1k1 )(b21 · · · b2k2 ) · · · (bt1 · · · btkt )
are the decompositions of σ and τ into a product of pairwise disjoint cycles.

Define g ∈ Sn by

a11 a12 · · · a1k1 a21 · · · a2k2 · · · at1 · · · atkt
g := .
b11 b12 · · · b1k1 b21 · · · b2k2 · · · bt1 · · · btkt
By (2.1), we get
gσg −1 = g(a11 a12 · · · a1k1 )g −1 g(a21 · · · a2k2 )g −1 · · · g(at1 · · · atkt )g −1

= (g(a11 ) g(a12 ) · · · g(a1k1 ))(g(a21 ) · · · g(a2k2 )) · · · (g(at1 ) · · · g(atkt ))
= τ.
Thus σ and τ are conjugate in Sn .
Example 2.20. With Proposition 2.19, it is straightforward to find repre-

sentatives for the conjugacy classes for any given n. For example, in S4 we
have the following representatives for the 5 conjugacy classes:
55
representative partition size of class

(1) (1, 1, 1, 1) 1
(1 2) (2, 1, 1) 6
(1 2 3) (3, 1) 8
(1 2)(3 4) (2, 2) 3
(1 2 3 4) (4) 6
The size of each class is counted by calculating the number of cycles with
the corresponding decomposition. For example, the number of 3-cycles (i j k)
is 4·3·2
3
= 8. This is because there are 4 · 3 · 2 choices for i, j, k; then we divide
by 3 so that we do not count
(i j k) = (j k i) = (k i j)
three times. Similarly, the number of permutations (i j)(k l) is 12 ( 4·3

2
· 2·1
2
) = 3.
4·3 2·1
Indeed, there are 2 choices for (i j), then 2 choices for (k l); then divide
by 2 to avoid counting (i j)(k l) = (k l)(i j) twice. Similar arguments work
for counting the size of any given conjugacy class in Sn .
2.4 Alternating groups

In this section, we will define alternating groups, and show that it is the
unique subgroup of index 2 in Sn . There are multiple ways to define the
alternating groups, all of which boil down identifying a notion of “parity” for
permutations.
Definition 2.21. Let σ ∈ Sn . An unordered pair {i, j} with 1 ≤ i < j ≤ n

is an inversion of σ if σ(i) > σ(j). The set of all inversions of σ is denoted
by I(σ).
Lemma 2.22. Let σ, τ ∈ Sn . Then |I(στ )| ≡ |I(σ)| + |I(τ )| mod 2.
Proof. The set I(στ ) consists of {i, j} with 1 ≤ i < j ≤ n such that either
• {i, j} 6∈ I(τ ) and {τ (i), τ (j)} ∈ I(σ); or
• {i, j} ∈ I(τ ) and {τ (i), τ (j)} 6∈ I(σ).
Thus I(στ ) = (X \ Y ) ∪ (Y \ X), where
• X = I(τ )
• Y = set of pairs {i, j} with {τ (i), τ (j)} ∈ I(σ).
56
We have |X \ Y | = |X| − |X ∩ Y | and similarly |Y \ X| = |Y | − |X ∩ Y |.

Therefore |I(στ )| = |X| + |Y | − 2|X ∩ Y |. Moreover |Y | = |I(σ)| (Exercise
2.5), so we conclude that |I(στ )| ≡ |I(σ)| + |I(τ )| mod 2.
Therefore the map sgn : Sn → {1, −1} defined by
sgn(σ) = (−1)|I(σ)|
for all σ ∈ Sn is a homomorphism. (Here {1, −1} is a cyclic group of order

2 under multiplication.) We call sgn the sign homomorphism.
For σ = (1 2) we have I(σ) = {{1, 2}}, so sgn(σ) = −1 and thus sgn is a
surjective homomorphism for n ≥ 2.
Definition 2.23. The alternating group of degree n is An := Ker(sgn), where

sgn : Sn → {1, −1} is the sign homomorphism. A permutation σ ∈ Sn is said
to be even if σ ∈ An , and odd if σ 6∈ An .
For n ≥ 2 the sign homomorphism is surjective, so |An | = n!/2.
Lemma 2.24. Any transposition is an odd permutation.
Proof. Any transposition σ is conjugate to τ = (1 2), say σ = gτ g −1 for some

g ∈ Sn . Then sgn(σ) = sgn(g) sgn(τ ) sgn(g)−1 = sgn(τ ). As we have seen
above |I(τ )| = 1, so sgn(σ) = sgn(τ ) = −1.
Remark 2.25. Exercise: For σ = (i j) ∈ Sn with 1 ≤ i < j ≤ n, calculate

|I(σ)|.
We know that every σ ∈ Sn can be written as a product σ = t1 t2 · · · td of

transpositions. Since the sign map is a homomorphism, by Lemma 2.24 we
have that σ is even if d is even, and odd if d is odd. Note that there might
be multiple ways of writing a permutation as a product of transpositions, for
example (1 2 3) = (1 2)(2 3) = (1 3)(1 2) = (3 4)(3 4)(1 2)(2 3).
Lemma 2.26. Let n ≥ 3. Then An is generated by 3-cycles.
Proof. Any σ ∈ An can be written as a product of even number of transpo-

sitions, say σ = t1 t01 · · · td t0d for some transpositions ti , t0i ∈ Sn . Thus it will
suffice to show that any product of two transpositions is contained in the sub-
group generated by the 3-cycles. To this end, let t = (i j) and t0 = (k l) be
two transpositions. If t and t0 are disjoint, then tt0 = (i j)(k l) = (i j k)(j k l).
If {i, j} = {k, l}, then tt0 = 1. The remaining possibility is that {i, j} and
{k, l} have one element in common, say t = (i j) and t0 = (j l). In this case
tt0 = (i j l).
57
Proposition 2.27. Let n ≥ 2. Suppose that H < Sn is such that [Sn : H] =

2. Then H = An .
Proof. If n = 2, then S2 has order 2 and A2 is trivial, so the claim is obvious.

Suppose that n ≥ 3. Since H has index 2, it is a normal subgroup and
σ 2 ∈ H for all σ ∈ Sn . For a 3-cycle (i j k) we have (i j k) = (i k j)2 , so H
contains every 3-cycle. It follows from Lemma 2.26 that H = An .
Example 2.28. Arguing as in the proof of Proposition 2.27, we find that if

n ≥ 3, then An contains no subgroup of index 2. This provides a counterex-
ample to the converse of Lagrange’s theorem: for example |A4 | = 12, so 6
divides the order of A4 , but A4 has no subgroup of order 6.
Lemma 2.29. Let n ≥ 2 and G ≤ Sn . Then either G ≤ An , or G ∩ An is a

normal subgroup of index 2 in G.
Proof. Exercise.
2.5 Conjugacy classes in An

What are the conjugacy classes in An ? Since An is a normal subgroup of Sn ,
we at least know that An is a union of conjugacy classes Sn ; namely it is the
union of the conjugacy classes of even permutations.
Hence for σ ∈ An , we have σ Sn ⊆ An . Are all elements of σ Sn conjugate in
An ? The answer is no in general. Consider for example A3 , which contains
the conjugacy class (1 2 3)S3 = {(1 2 3), (1 3 2)}. But A3 is abelian, so
(1 2 3) and (1 3 2) are not conjugate in A3 . An exercise shows that (1 2 3)
and (1 3 2) are not conjugate in A4 either, but they are conjugate in An for
n ≥ 5.
In general it turns out that either σ Sn = σ An , or σ Sn = σ An ∪ (σ 0 )An is the
union of two distinct An -conjugacy classes of equal size. In the latter case we
say that the conjugacy class splits in An . When does a conjugacy class split?
The answer is provided by the following lemma, in terms of centralizers.
Lemma 2.30. Let G be a finite group and let N /G be such that [G : N ] = 2.

Let g ∈ N . The following statements hold:
(i) If CG (g) 6≤ N , then g N = g G .
(ii) If CG (g) ≤ N , then g G = g N ∪ (g 0 )N for some g 0 ∈ N , and moreover

|g N | = |(g 0 )N | = |g G |/2.
58
Proof. (i) Suppose that CG (g) 6≤ N . It will suffice to prove that for x ∈
G \ N , we have xgx−1 ∈ g N . Let h ∈ CG (g) \ N . Since x ∈ / N and
∼ −1
h 6∈ N , we have hx ∈ N because G/N = C2 . Since hxh = x we get
xgx−1 = (xh)g(xh)−1 ∈ g N .
(ii) Suppose that CG (g) ≤ N , so CG (g) = CN (g). Then
|g G | = [G : CN (g)] = [G : N ][N : CN (g)] = 2|g N |
by Lemma 1.114, so |g N | = |g G |/2. Let g 0 ∈ g G \ g N , say g 0 = xgx−1 .
Then CG (g 0 ) = xCG (g)x−1 ≤ N , so by the same argument |(g 0 )N | =
|g G |/2. Since distinct conjugacy classes of N are disjoint, we conclude
that g G = g N ∪ (g 0 )N .
Lemma 2.31. Suppose that n ≥ 5. Then 3-cycles are conjugate in An .

Proof. By Lemma 2.30, it will suffice to verify that for a 3-cycle σ = (i j k) ∈
An , we have CSn (σ) 6≤ An . For this, take τ = (l m) a transposition disjoint
from σ. Then τ ∈ CSn (σ) (Lemma 2.2) and τ 6∈ An .
With the same proof, one can generalize Lemma 2.31 to show that for
k odd, the k-cycles for k ≤ n − 2 are conjugate in An . More generally, one
can prove that for σ ∈ An the conjugacy class σ Sn splits in An if and only
if the cycle decomposition of σ consists of disjoint cycles with distinct odd
lengths. (For example, in A4 the 3-cycles split into two classes, since they
are a product of a 3-cycle and a 1-cycle.)
2.6 An is simple for n ≥ 5

Theorem 2.32. Let n ≥ 5. Then An is simple.
Proof. Let N E An be a non-trivial normal subgroup. It will suffice to prove
that N contains a 3-cycle, since then N will contain all 3-cycles by Lemma
2.31, and thus N = An by Lemma 2.26.
Let g ∈ N be a nontrivial element, and write its decomposition into a
product of disjoint cycles as g = π1 π2 · · · πt . We consider the different possi-
bilities for the cycle decomposition, and show that in each case we can find
a 3-cycle in N .
Case 1: Some cycle has length ≥ 4. By replacing g with a conjugate and

rearranging the πi , we can assume that π1 = (1 2 3 · · · s) with s ≥ 4. Set
σ = (1 2 3). Then
σgσ −1 = (σπ1 σ −1 )(σπ2 σ −1 ) · · · (σπt σ −1 )
59
= (σπ1 σ −1 )π2 · · · πt
= (2 3 1 4 · · · s)π2 · · · πt .
Here we have used the fact that σ is disjoint from π2 , . . ., πt and thus
commutes with them. Since σgσ −1 ∈ N by normality of N , it follows that
σgσ −1 g −1 = (2 3 1 4 · · · s)(1 2 3 · · · s)−1 = (1 2 4) ∈ N .
Case 2: Each cycle has length ≤ 3, and there are at least two 3-cycles. By
replacing g with a conjugate and rearranging the πi , we can assume that
π1 = (1 2 3) and π2 = (4 5 6). Set σ = (1 5)(2 6 3 4). Then
σgσ −1 = (σπ1 σ −1 )(σπ2 σ −1 )(σπ3 σ −1 ) · · · (σπt σ −1 )

= (σπ1 σ −1 )(σπ2 σ −1 )π3 · · · πt
= (1 3 2)(4 5 6)π3 · · · πt .
Thus σgσ −1 g −1 = (1 2 3) ∈ N .
Case 3: There is exactly one 3-cycle, and rest of the cycles have length ≤ 2.
By replacing g with a conjugate and rearranging the πi , we can assume that
g = (1 2 3)π2 · · · πt , where πi are 2-cycles. Then g 2 = (1 3 2)π22 · · · πt2 =
(1 3 2) ∈ N .
Case 4: All cycles have length ≤ 2. There must be at least 2 cycles since
g ∈ An , so by replacing G with a conjugate, we may assume that g =
(1 2)(3 4)π3 · · · πt . Let σ = (1 2 3). Calculating as in the previous cases
shows that σgσ −1 g −1 = (1 3)(2 4) ∈ N . Then for τ = (1 5 3), we have
τ (1 3)(2 4)τ −1 (1 3)(2 4) = (1 3 5) ∈ N .
The cases above exhaust all possibilities, so we conclude that N contains

a 3-cycle and thus N = An .
2.7 Group actions

Group actions are a convenient way to study the possible ways a group G
(or homomorphic images of G) can be realized as a group of permutations.
Thanks to the orbit–stabilizer theorem (Theorem 2.40), group actions are
also fundamental in various counting arguments that come up in finite group
theory.
Definition 2.33. Let G a group and let X be a nonempty set. An action

of G on X is map α : G × X → X, (g, x) 7→ g · x such that:
60
(i) 1 · x = x for all x ∈ X;

(ii) (gh) · x = g · (h · x) for all g, h ∈ G and x ∈ X.
In this case we say that G acts on X, and such an X is called a G-set.
We will often write gx instead of g · x for g ∈ G and x ∈ X. The action of an
element g ∈ G is the map αg : X → X defined by αg (x) = gx for all x ∈ X.
Example 2.34. Let G be a group.
(a) Let H ≤ G and consider the set X = {xH : x ∈ G} of left cosets of H
in G. Then G acts on X via g · (xH) = gxH.
(b) G = Sn acts naturally on X = {1, 2, . . . , n}, via σ · i = σ(i) for σ ∈ Sn
and i ∈ X.
(c) G acts on itself by conjugation. That is, we have an action of G on
X = G defined by g · x = gxg −1 for all g, x ∈ G.
The study of group actions of G on a set X is really the same thing as the
study of group homomorphisms G → Sym(X), which are called permutation
representations. This is seen from the next lemma:
Lemma 2.35. Let G be a group. Then the following statements hold.
(i) If G acts on a set X, then αg : X → X is a bijection for all g ∈ G.
(ii) If G acts on a set X, then ϕ(g)(x) := g · x defines a homomorphism
ϕ : G → Sym(X).
(iii) If ϕ : G → Sym(X) is a homomorphism, then g · x := ϕ(g)(x) defines
an action of G on X.
Proof. (i) If G acts on a set X, then it follows from (i) and (ii) in Definition
2.33 that αg αg−1 = idX = αg−1 αg for all g ∈ G. Therefore αg is a
bijection with inverse αg−1 .
(ii) In other words, we have defined ϕ(g) = αg for all g ∈ G. Thus ϕ is a
well-defined map ϕ : G → Sym(X) by (i). Moreover, we have αg αh =
αgh for all g, h ∈ G by (ii) in Definition 2.33, so ϕ is a homomorphism.
(iii) In other words, we have defined a map α : G × X → X by α(g, x) =
g · x = ϕ(g)(x) for all g ∈ G and x ∈ X, and we need to check that this
is an action of G on X. Definition 2.33 (i) is satisfied since ϕ(1) must
be the identity map (Lemma 1.59 (ii)). Definition 2.33 (ii) is satisfied
since ϕ(gh) = ϕ(g)ϕ(h) for all g, h ∈ G.
61
As is clear from Lemma 2.35, each action of G on a set X corresponds to

a unique homomorphism G → Sym(X), and conversely each homomorphism
G → Sym(X) corresponds to a unique action of G on X. An action of G on
X is faithful if the corresponding homomorphism is injective, equivalently if
the kernel of the homomorphism is trivial.
Definition 2.36. Let G be a group acting on a set X. The orbit of x ∈ X

under the action of G is the set Gx = {g · x : g ∈ G}. We will also sometimes
denote orbG (x) := Gx.
Lemma 2.37. Let G be a group acting on a set X. For any two orbits Gx
and Gy, either Gx = Gy or Gx ∩ Gy = ∅.
Proof. Note first that
Ggx = Gx for all g ∈ G. (2.2)
Suppose then that there exists z ∈ Gx ∩ Gy, say z = gx = hy for some

g, h ∈ G. Then Gx = Gz and Gy = Gz by (2.2), so Gx = Gy. Thus any two
orbits are either equal or disjoint, which proves the lemma.
By Lemma 2.37, if a group G acts on a set X, the orbits partition X. In
other words, we have a disjoint union
[
X= Gx
x∈S
where S is a set of representatives for the orbits. If there is only one orbit,
we say that the action of G on X is transitive. Note that G acts transitively
on any orbit Gx.
If X is a finite set, say
X = Gx1 ∪ · · · ∪ Gxt
with x1 , . . . , xt ∈ X representatives for the orbits, then
|X| = |Gx1 | + · · · + |Gxt |.
We next explain how the sizes of the orbits are calculated.
Definition 2.38. Let G be a group acting on a set X. The stabilizer of

x ∈ X is the set Gx = {g ∈ G : gx = x}. We will also sometimes denote
stabG (x) = Gx .
62
For a group G acting on a set X, the kernel of the action is the kernel of
the corresponding homomorphism G → Sym(X). It is clear that the kernel
is equal to \
Gx ,
x∈X
T
so the action of G on X is faithful if and only if x∈X Gx = {1}.
Lemma 2.39. Let G be a group acting on a set X. Then the following hold:
(i) Gx ≤ G for all x ∈ X.
(ii) gGx g −1 = Ggx for all g ∈ G and x ∈ X.
(iii) Let x ∈ X. Then gGx = hGx if and only if gx = hx.
Proof. (i) Let x ∈ X. It is immediate from the definitions that 1 ∈ Gx

and that g, h ∈ Gx implies gh ∈ Gx . Moreover if g ∈ Gx , then g −1 · x =
g −1 · (g · x) = 1 · x = x, so g −1 ∈ Gx . Thus Gx ≤ G.
(ii) Let g ∈ G and x ∈ X. For h ∈ Gx , we have ghg −1 (gx) = gh(g −1 gx) =

ghx = gx, so gGx g −1 ≤ Ggx . The same argument shows that g −1 Ggx g ≤
Gg−1 gx = Gx , so Ggx ≤ gGx g −1 and thus gGx g −1 = Ggx .
(iii) We have gGx = hGx if and only if h−1 g ∈ Gx , which is a equivalent to

h−1 gx = x. Clearly h−1 gx = x if and only if gx = hx.
Theorem 2.40 (Orbit–Stabilizer theorem). Let G be a group and suppose

that G acts on a set X. Let x ∈ X. Let Ω be the set of left cosets of Gx .
Then the map
ψ : Ω → Gx
defined by ψ(gGx ) = gx is a bijection. In particular | orbG (x)| = [G : Gx ].
Proof. The fact that ψ is well-defined and injective follows from Lemma 2.39
(iii). Since ψ is clearly surjective, it is a bijection. Thus |Ω| = |Gx|, so
[G : Gx ] = |Gx|.
Remark 2.41. Let G be a group acting on two sets X and Y . We say that
X and Y are equivalent (isomorphic) as G-sets, if there exists a bijection
ψ : X → Y such that ψ(gx) = gψ(x) for all x ∈ X. For equivalent G-sets X
and Y , the action of G is essentially the same, just the names of the elements
that G acts on are different (and the different names or labels are provided
63
by the bijection ψ). This is analogous to the isomorphism of groups (recall

e.g. Example 1.16).
It is clear that for the bijection ψ : Ω → Gx in Theorem 2.40, we have
ψ(gω) = gψ(ω) for all ω ∈ Ω, so Ω and Gx are equivalent as G-sets.
Thus any transitive G-set is equivalent to the action of G on the left
cosets of some subgroup of G. More precisely, if G acts transitively on X,
then the action of G on X is equivalent to the action of G on the left cosets
of Gx .
The orbit–stabilizer theorem tells us that the size of an orbit is the index
of the stabilizer of a point. We will now see how many of the results and
concepts are special cases of the orbit–stabilizer theorem. These include,
for example, the fact that the size of a conjugacy class is the index of the
centralizer.
Example 2.42. Let G be a group. Some basic examples of group actions,

along with the orbits and the stabilizers.
(a) Let σ ∈ Sn . Consider the natural action of G = hσi on X = {1, 2, . . . , n}.
• Orbits: orbσ (i) for 1 ≤ i ≤ n.

• Stabilizers: stabG (i) = hσ k i, where k > 0 is minimal such that
σ k (i) = i.
(b) Let H ≤ G and consider the set X = {xH : x ∈ G} of left cosets of H

in G. Then G acts on H via g · (xH) = gxH.
• Orbits: Only one orbit, orbG (H) = X.

• Stabilizers: Stabilizer of xH is stabG (xH) = xHx−1 .
(c) G acts on itself by left multiplication. That is, we have an action of G

on X = G defined by g · x = gx for all g, x ∈ G. The corresponding
homomorphism G → Sym(G) is the left regular representation (Theorem
1.66).
• Orbits: Only one orbit, orbG (x) = G for all x ∈ G.

• Stabilizers: stabG (x) = {1} for all x ∈ G.
(d) Let H ≤ G. We have H acting on X = G by left multiplication.
• Orbits: Cosets of H, with orbH (x) = Hx for all x ∈ G.

• Stabilizers: stabH (x) = {1} for all x ∈ H.
64
(e) G acts on itself by conjugation; that is, we have an action of G on X = G

via g · x = gxg −1 for all g, x ∈ G.
• Orbits: Conjugacy classes, i.e. orbG (x) = xG .

• Stabilizers: Centralizers, i.e. stabG (x) = CG (x).
(f) G acts on the set of its subgroup by conjugation; that is, we have an
action of G on X = {H : H ≤ G} via g · H = gHg −1 for all g ∈ G and
H ≤ G.
• Orbits: Conjugacy classes of subgroups.

• Stabilizers: Normalizers, i.e. stabG (H) = NG (H).
(g) G = Sn acts naturally on X = {1, 2, . . . , n}, via σ · i = σ(i) for σ ∈ Sn

and i ∈ X.
• Orbits: Only one orbit, orbG (i) = X for all 1 ≤ i ≤ n.

• Stabilizers: stabG (i) is isomorphic to Sym(X \ {i}) ∼
= Sn−1 .
(h) Let V be a finite-dimensional vector space over a field F, with n = dim V .
Then G = GL(V ) as on V via g · v = g(v) for all g ∈ G and v ∈ V .
• Orbits: Two orbits, {0} and V \ {0}.

• Stabilizers: stabG (0) = G, and for v ∈ V \{0} the stabilizer stabG (v)
is isomorphic to the group of matrices of the following form:
 
1 ∗ ··· ∗
0 
,
 
 ..
. A 
0
where A ∈ GLn−1 (F).
Lemma 2.43. Let G be a finite group and let H ≤ G with [G : H] = n. Then

the action of G on the left cosets of H induces a homomorphism ϕ : G → Sn ,
with \
Ker ϕ = gHg −1 .
g∈G
Proof. The action of G on the left cosets of H induces a homomorphism ϕ :

G → Sn , as we have seen in Lemma 2.35. The kernel of this homomorphism
is the intersection of all stabilizers. The stabilizer of a left coset gH is equal
to gHg −1 , so Ker ϕ = ∩g∈G gHg −1 .
65
We finish this section with one more example. Consider G = GL2 (2) and
let V = F22 with basis e1 , e2 . In this case, the action of G on V \ {0} provides
a homomorphism
ϕ : GL2 (2) → S3 ,
since there are only 3 non-zero vectors V ; indeed V \ {0} = {e1 , e2 , e1 + e2 }.
It is clear that the action of G is faithful, so ϕ must be injective. Since
| GL2 (2)| = |S3 |, the map ϕ is an isomorphism, and so GL2 (2) ∼ = S3 . Label-
ling the non-zero vectors e1 , e2 , e1 + e2 as 1, 2, 3 we have

1 0 0 1
ϕ = (1) ϕ = (1 2)
0 1 1 0

0 1 1 0
ϕ = (1 2 3) ϕ = (1 3)
1 1 1 1

1 1 1 1
ϕ = (1 3 2) ϕ = (2 3)
1 0 0 1
66
3 Linear groups
In this section, we will prove some basic results about the groups GLn (F) and
SLn (F), where F is a field. Our focus is on the case where F is a finite field,
but many of the methods described in this section will work over arbitrary
fields as well.
3.1 On finite fields

The main example of a finite field of Fp = (Z/pZ, +, ·), the field of integers
modulo p, where p is a prime.
Let F be a finite field. Let p be the characteristic of F, i.e., the smallest
positive integer such that
p1 = |1 + ·{z
· · + 1} = 0.
p summands
Since F is a field, it follows that p is a prime number, and moreover

{m1 : m ∈ Z}
is a subfield of F isomorphic to Fp , called the prime field. Thus we can
identify Fp = {m1 : m ∈ Z}.
Now F is a vector space over Fp (scalar multiplication is just multiplication
by elements of Fp ). Let α1 , . . ., αn be a basis for F over
Pn Fp . Then any element
of F is uniquely expressed as a linear combination i=1 ci αi with ci ∈ Fp , so
it follows that |F| = pn .
Thus every finite field has order equal to a prime power. Conversely, for
any prime power there exists a finite field of that order.
Theorem 3.1. Let q = pn , where p is a prime. Then there exists a field F
with |F| = q, and moreover F is unique up to isomorphism.
Proof. Omitted.
Thus for each prime power q = pn there exists a unique field of order q
up to isomorphism, and we denote such a field by Fq . Note that in Theorem
1.52, we proved that the multiplicative group of a finite field is cyclic. In
other words, we have F× ∼
q = Cq−1 for any prime power q.
Example 3.2. We know that F× × 2

4 is cyclic of order 3, say F4 = {1, α, α },
where α3 = 1. Then
0 = α3 − 1 = (α − 1)(α2 + α + 1),
so α2 + α + 1 = 0 since α 6= 1. Since F4 = {0, 1, α, α2 }, it is clear how to
multiply elements in F4 .
67
· 0 1 α α2
0 0 0 0 0
1 0 1 α α2
α 0 α α2 1
α2 0 α2 1 α
The addition operator in F4 is determined by the fact that α2 + α + 1 = 0,

for example α2 + α = 1 and α + 1 = α2 . Note that (F4 , +) is not cyclic, since
x + x = 0 for all x ∈ F4 .
+ 0 1 α α2
0 0 1 α α2
1 1 0 α2 α
α α α2 0 1
α2 α2 α2 1 0
3.2 Basic properties of GLn (q) and SLn (q)

Let q be a power of a prime. For GLn (Fq ) and SLn (Fq ), we use notation
GLn (q) := GLn (Fq ) and SLn (q) := SLn (Fq ).
Lemma 3.3. | GLn (q)| = (q n − 1)(q n − q) · · · (q n − q n−1 ).
Proof. An n × n matrix is invertible if and only if its columns are linearly

independent, i.e. if they form a basis of the vector space Fnq . Thus | GLn (q)|
is equal to the number of ordered bases of Fnq , as a vector space over Fq . Now
vectors (v1 , . . . , vn ) form a basis if and only if v1 6= 0, and vi 6∈ hv1 , . . . , vi−1 i
for all 1 < i ≤ n. Thus there are q n − 1 choices for v1 . For each such v1 there
are q n −q possible choice for v2 , since there are q vectors in hv1 i. Similarly for
v3 we have q n − q 2 choices, since there are q 2 vectors in hv1 , v2 i. Continuing
in this manner, we see that there are a total of
(q n − 1)(q n − q)(q n − q 2 ) · · · (q n − q n−1 )
bases of GLn (q).

For the order of SLn (q), we note that SLn (q) is the kernel of the homo-
morphism det : GLn (q) → F× q . It is easy to see that this homomorphism is
surjective, so by the first isomorphism theorem GLn (q)/ SLn (q) ∼ = F×q and
| SLn (q)| = | GLn (q)|/(q − 1).
Lemma 3.4. Let n ≥ 1. Then the following hold:
68
(i) Z(GLn (q)) = {λIn : λ ∈ F×

q }, the group formed by scalar matrices in
GLn (q).
(ii) Z(SLn (q)) = {λIn : λ ∈ F× n

q , λ = 1}.
Proof. Exercise.
The group PGLn (q) := GLn (q)/Z(GLn (q)) is called the projective general
linear group (over Fq ), while PSLn (q) := SLn (q)/Z(SLn (q)) is called the
projective special linear group (over Fq ). It can be proven that if n ≥ 2,
then PSLn (q) is a non-abelian simple group, except for PSL2 (2) ∼ = S3 and
PSL2 (3) ∼
= 4A .
3.3 Polynomial rings

Let F be a field. The polynomial ring F[t] in the indeterminate t (or variable)
is a the vector space over F with basis 1, t, t2 , t3 , . . . where multiplication is
defined as ! ! !
X X X X
ai ti · bj tj = ai bj tk .
i≥0 j≥0 k≥0 i+j=k
(We denote t0 = 1.) Then (F[t], +, ·) is a commutative ring with multiplica-

tive identity 1.
Each non-zero p(t) ∈ F[t] can be written uniquely in the form
p(t) = a0 + a1 t + · · · + an−1 tn−1 + an tn
for some n ≥ 0 and a0 , . . ., an ∈ F, where an 6= 0. In this case we call n

the degree of p(t) and denote deg p(t) = n. A polynomial p(t) is constant, if
p(t) = 0 or deg p(t) = 0; in other words p(t) = a0 for some a0 ∈ F.
The ring F[t] is an Euclidean domain, in other words, there is a division
algorithm. Suppose that f (t) ∈ F[t] is non-zero. Then for all p(t) ∈ F[t]
there exist unique q(t), r(t) ∈ F[t] such that
p(t) = q(t)f (t) + r(t), where r(t) = 0 or deg r(t) < deg f (t).
If r(t) = 0, we say that f (t) divides p(t) and denote f (t) | p(t).
For every p(t) ∈ F[t] and c ∈ F, the division algorithm gives
p(t) = q(t)(t − c) + d
for some d ∈ F. Thus it follows that p(c) = 0 if and only if t − c divides p(t).
A non-constant polynomial p(t) is irreducible (over F), if p(t) cannot be
written as the product of two non-constant polynomials in F[t]. In other
69
words, if p(t) = a(t)b(t), then a(t) or b(t) is constant. For example t2 + 1 is

irreducible in R[t], but has factorization
t2 + 1 = (t − i)(t + i)
in C[t].
We call a non-zero polynomial p(t) ∈ F[t] monic if it has leading coefficient
1, in other words
p(t) = tn + cn−1 tn−1 + · · · + c1 t + c0
for some c0 , . . ., cn−1 ∈ F.

Since F[t] is an Euclidean domain, greatest common divisors exist. Let
f (t), g(t) ∈ F[t]. Then d(t) = gcd(f (t), g(t)) is the unique monic polynomial
such that:
• d(t) | f (t) and d(t) | g(t);
• If p(t) is such that p(t) | f (t) and p(t) | g(t), then p(t) | d(t).
By Bézout’s lemma there exist a(t), b(t) ∈ F[t] such that
a(t)f (t) + b(t)g(t) = gcd(f (t), g(t)).
Since F[t] is Euclidean, it is a unique factorization domain, so we have

unique factorization of polynomials into irreducibles. Let p(t) ∈ F[t] be non-
constant. Then
p(t) = αp1 (t)k1 · · · pr (t)kr
for some α ∈ F, monic irreducible polynomials p1 (t), . . ., pr (t), and integers
k1 , . . ., kr > 0. Moreover, this factorization of p(t) is unique, up to the
ordering of the factors p1 (t), . . ., pr (t). For example, the factorization of
t3 − 1 into irreducibles is
t3 − 1 = (t − 1)(t2 + t + 1) in F2 [t],
t3 − 1 = (t − 1)3 in F3 [t],
t3 − 1 = (t − 1)(t2 + t + 1) in F5 [t],
t3 − 1 = (t − 1)(t − 2)(t − 3) in F7 [t].
70
3.4 Characteristic polynomials and eigenvalues

Let F be a field and let V be a finite-dimensional vector space over F. For a
linear map f : V → V , recall that an eigenvector of f with eigenvalue α ∈ F
is a non-zero vector v ∈ V such that f (v) = αv. For α ∈ F, we denote by Vα
the eigenspace
Vα = {v ∈ V : f (v) = αv}.
Let A ∈ GLn (F). Recall that the characteristic polynomial of A is the
monic polynomial pA (t) ∈ F[t] defined by
pA (t) = det(tIn − A).
(Here we are taking the determinant of a matrix with entries in F[t].)

a b
Example 3.5. Let A ∈ GL2 (F), say A = . Then
c d

t − a −b
pA (t) = det = (t − a)(t − d) − bc
−c t − d
= t2 − (a + d)t + (ad − bc)
= t2 − tr(A)t + det(A).
(Here tr(A) is the trace of the matrix A, which is defined the sum of the
diagonal entries of A.)
We recall some basic properties of pA (t). First note that the characteristic
polynomial is the same for any conjugate of A.
Lemma 3.6. Let A ∈ GLn (F). Then any conjugate of A has the same
characteristic polynomial as A.
Proof. We have
pXAX −1 (t) = det(tIn − XAX −1 )
= det(X(tIn − A)X −1 )
= det(X) det(tIn − A) det(X)−1
= pA (t)
for all X ∈ GLn (F).
.
Lemma 3.7. Let A ∈ GLn (F). Then A has an eigenvector in Fn with
eigenvalue α ∈ F if and only if α is a root of pA (t).
Proof. There is an eigenvector in Fn with eigenvalue α ∈ F if and only if
αIn − A has nontrivial kernel, which is equivalent to det(αIn − A) = 0. Since
pA (α) = det(αIn − A), the result follows.
71
3.5 Conjugacy classes of GL2 (q)

In section, we will give a description of the conjugacy classes of GL2 (q).
Lemma 3.8. Let V = (Fq )2 . Suppose that A ∈ GL2 (q) has an eigenvector
in V with eigenvalue α ∈ Fq . Then the following hold:
(i) If the eigenspace Vα is 2-dimensional, then

α 0
A=
0 α
is a scalar matrix.
(ii) If A has another eigenvalue β 6= α, then A has characteristic polynomial

(t − α)(t − β) and is conjugate to

α 0
sα,β = .
0 β
(iii) If the eigenspace Vα is 1-dimensional and α is the only eigenvalue of A,

then A has characteristic polynomial (t − α)2 and is conjugate to

α 1
uα = .
0 α
(iv) Let α, α0 , β, β 0 ∈ Fq . Then sα,β and sα0 ,β 0 are conjugate in GL2 (q) if and
only if {α, β} = {α0 , β 0 }.
(v) Let α, α0 ∈ Fq . Then uα , uα0 are conjugate in GL2 (q) if and only if
α = α0 .
Proof. (i) In this case

V = Vα , so Av = αv for all v ∈ V which makes it
α 0
clear that A = .
0 α
(ii) We are assuming that α and β are eigenvalues of A, so there exist non-
zero vectors vα ∈ Vα and vβ ∈ Vβ . Now vβ 6∈ Vα , so {vα , vβ } is linearly
independent and thus a basis of V . Then the matrix of A with respect
to this basis is
α 0
sα,β = ,
0 β
so A is conjugate to sα,β in GL2 (q).
72
(iii) Let v ∈ Vα be nonzero, and extend this to a basis {v, w} of V . We have

Av = αv and Aw = βw + γv for some β, γ ∈ Fq , so the matrix of A
with respect to this basis is

α γ
.
0 β
Thus the characteristic polynomial of A is (t − α)(t − β), and so β = α

since α is the only eigenvalue of A.
Note that γ 6= 0, as otherwise w would be an eigenvector. Thus we can
replace v by v 0 = γv to get another basis {v 0 , w}, for which Av 0 = αv 0
and Aw = αw + v 0 , so the matrix of A is

α 1
uα =
0 α
with respect to the basis {v 0 , w}. Therefore A is conjugate to uα in

GL2 (q).
(iv) Suppose that sα,β and uα0 ,β 0 are conjugate. Then they must have the
same characteristic polynomial. Since their characteristic polynomials
are (t − α)(t − β) and (t − α0 )(t − β 0 ), we conclude {α, β} = {α0 , β 0 }.
Conversely, suppose that {α, β} = {α0 , β 0 }. If α = α0 and β = β 0 , then
sα,β = sα0 ,β 0 . Otherwise α = β 0 and β = α0 , and we should show that
sα,β is conjugate to sα0 ,β 0 = sβ,α . Let e1 , e2 be the standard basis of
column vectors. Then with respect to the basis {e2 , e1 } the matrix of
sβ,α is the same as that of sα,β , so they are conjugate
in GL2 (q). (For
0 1
example via the change of basis matrix .)
1 0
(v) If uα and uα0 are conjugate, they must have the same characteristic
polynomial. The characteristic polynomials of uα and uα0 are (t − α)2
and (t − α0 )2 , respectively, so α = α0 . Conversely if α = α0 then
uα = uα0 .
Lemma 3.9. Suppose that A ∈ GL2 (q) has no eigenvectors in (Fq )2 , and let
p(t) = t2 + βt + α be the characteristic polynomial of A. Then:
(i) The polynomial p(t) is irreducible, and A is conjugate to

0 −α
Cp(t) = .
1 −β
73
(ii) For any polynomial p(t) ∈ Fq [t], the matrix Cp(t) in (i) has characteristic
polynomial equal to p(t).
(iii) Let p(t), q(t) ∈ Fq [t] be irreducible monic polynomials of degree 2. Then
Cp(t) and Cq(t) are conjugate if and only if p(t) = q(t).
Proof. (i) As a degree two polynomial p(t) is irreducible if and only if it

has no roots in Fq . This is the case since A has no eigenvectors in
V = (Fq )2 .
Now pick a non-zero vector v ∈ V . Then Av 6∈ hvi since A has no
eigenvectors, so {v, Av} is a basis of V . Thus we can write A2 v =
−α0 v − β 0 Av for some α0 , β 0 ∈ Fq . Then with respect to the basis
{v, Av} the matrix of A is
0 −α0

C= ,
1 −β 0
so A is conjugate to C in GL2 (q). The characteristic polynomial of

C is equal to t2 + β 0 t + α0 . Since conjugate matrices have the same
characteristic polynomial, we have p(t) = t2 + β 0 t + α0 and thus α0 = α
and β 0 = β. In other words C = Cp(t) .
(ii) Straightforward calculation.
(iii) If Cp(t) and Cq(t) are conjugate, they must have the same characteristic
polynomial, so p(t) = q(t) by (ii). Conversely if p(t) = q(t), then
Cp(t) = Cq(t) .
In view of Lemma 3.9, to calculate the total number of conjugacy classes,

we should count the number of irreducible polynomials t2 + βt + α ∈ Fq [t].
Lemma 3.10. Let q be a prime power. The number of irreducible polynomi-

als in Fq [t] of the form t2 + βt + α is equal to q(q − 1)/2.
Proof. Exercise.
With Lemma 3.8 and Lemma 3.9, we have found representatives for all
conjugacy classes of GL2 (q).
• q − 1 classes corresponding to scalar matrices (Lemma 3.8 (i)).
• (q − 1)(q − 2)/2 classes consisting of diagonal matrices with distinct

eigenvalues (Lemma 3.8 (ii));
74
• q − 1 classes consisting of non-diagonalizable matrices with an eigenva-

lue (Lemma 3.8 (iii));
• q(q −1)/2 classes with an irreducible characteristic polynomial (Lemma

3.9 and 3.10);
Thus there are a total of k(G) = q 2 − 1 conjugacy classes in G. We

will now calculate the size of each conjugacy class g G , which is equal to
[G : CG (g)]. A scalar matrix g ∈ G is central, so |g G | = 1. For diagonal
matrices with distinct eigenvalues, we have |CG (g)| = (q − 1)2 , so |g G | =
q(q+1). For g ∈ G non-diagonalizable with one eigenvalue, we have |CG (g)| =
q(q − 1) and so |g G | = (q − 1)(q + 1). Finally for g ∈ G with irreducible
characteristic polynomial we have |CG (g)| = (q − 1)(q + 1) and so |g G | =
q(q − 1). (The details of these calculations are left as an exercise.) We
summarize the information about the conjugacy classes in Table 2.
Example 3.11. Here is an illustration in the smallest case, G = GL2 (2).

We have following representatives for the conjugacy classes:

1 0
• Scalar: .
0 1

1 1
• Non-diagonalizable with one eigenvalue: u1 = .
0 1

0 1
• Irreducible characteristic polynomial: Cp(t) = with irreducible
1 1
characteristic polynomial p(t) = t2 + t + 1 (This is the only irreducible
degree 2 polynomial in F2 [t].)
Thus there are a total of 3 conjugacy classes. (One could also deduce this
from the fact that G ∼
= S3 .)
Example 3.12. Exercise: Determine similarly representatives for the con-

jugacy classes of GL2 (3), and the size of each conjugacy class. (There are 8
conjugacy classes.)
Example 3.13. Exercise: Describe representatives for the conjugacy classes

of elements of order 3 in GL2 (7). (There are 5 conjugacy classes of elements
of order 3 in GL2 (7).)
75
Type Number of classes Representatives Characteristic Size of class

polynomial

α 0
Scalar q−1 (t − α)2 1
0 α

α 0
Diagonalizable (q − 1)(q − 2)/2 , α 6= β (t − α)(t − β) q(q + 1)
0 β

α 1
Non-diagonal, q−1 (t − α)2 (q − 1)(q + 1)
0 α
one eigenvalue

0 −α
Irreducible q(q − 1)/2 t2 + βt + α q(q − 1)
1 −β
characteristic (irreducible)
polynomial
Table 3: Conjugacy classes in G = GL2 (q).
3.6 Minimal polynomials

For A ∈ GLn (F) and a polynomial f (t) ∈ F[t] with f (t) = ak tk + ak−1 tk−1 +
· · · + a1 t + a0 , we denote by f (A) the polynomial evaluated at the matrix A;
in other words
f (A) = ak Ak + ak−1 Ak−1 + · · · + a1 A + a0 In .
Theorem 3.14 (Cayley–Hamilton theorem). Let A ∈ GLn (F) with charac-

teristic polynomial pA (t) ∈ F[t]. Then pA (A) = 0.
Proof. Omitted. (The usual proof is an application of the Jordan normal

form.)
For our purposes, we will mostly need the Cayley–Hamilton theorem in
degree n = 2, in which case it is easy to prove. For A ∈ GL2 (F), we have
pA (t) = t2 − tr(A)t + det(A),
where tr(A) is the trace of A. It is straightforward to verify that
A2 − tr(A)A + det(A)I2 = 0.
A minimal polynomial for A ∈ GLn (F) is a monic polynomial m(t) ∈ F[t]

of minimal degree such that m(A) = 0.
Lemma 3.15. Let m(t) be a minimal polynomial of A ∈ GLn (F). Let f (t) ∈
F[t]. Then f (A) = 0 if and only if m(t) divides f (t).
76
Proof. If m(t) divides q(t) the lemma is clear. For the other direction, use the
division algorithm to write f (t) = q(t)m(t) + r(t) where deg r(t) < deg m(t);
by minimality of m(t) we must have r(t) = 0.
It follows from Lemma 3.15 that a minimal polynomial of A is unique.
We will denote it by mA (t). By the Cayley–Hamilton theorem, mA (t) divides
the characteristic polynomial pA (t).
Lemma 3.16. Let A ∈ GLn (F). Then any conjugate of A has the same
minimal polynomial as A.
Proof. Let g ∈ GLn (F). For any polynomial p(t), we have p(gAg −1 ) =
gp(A)g −1 , so p(A) = 0 if and only if p(gAg −1 ) = 0. The lemma follows.
Lemma 3.17. Let A ∈ GLn (F). Then deg mA (t) ≤ n.
Proof. By Cayley–Hamilton mA (t) divides pA (t), from which the result fol-
lows.
In degree n = 2, by the Cayley–Hamilton theorem the minimal polyno-
mial of A ∈ GL2 (F) has either degree 2 or degree 1. Moreover, it is clear
that the degree is 1 if and only if A is equal to a scalar matrix

α 0
0 α
for some α ∈ F× . Thus if A is not a scalar matrix, then by Cayley–Hamilton

the minimal polynomial of A is the characteristic polynomial of A. We record
this observation in the following lemma.
Lemma 3.18. Let A ∈ GL2 (F) be a non-scalar matrix. Then the minimal
polynomial of A is the characteristic polynomial pA (t) = t2 −tr(A)t+det(A)I2
of A.
Example 3.19. Let F be a field of characteristic p > 0. Suppose that

A ∈ GL2 (F) is such that |A| = p. Since p is prime, this is equivalent to
Ap = I2 and A 6= I2 . On the other hand, by the binomial theorem
p
p
X p i
(A − I2 ) = A (−1)p−i = Ap − I2 = 0
i=0
i
since pi ≡ 0 mod p for all 0 < i < p. Therefore the minimal polynomial

of A divides (A − I2 )p . Since the minimal polynomial has degree ≤ 2 by

Cayley–Hamilton, we conclude that mA (t) = (t − 1)2 .
77
Therefore the only eigenvalue of A is 1, and by Lemma 3.8 we conclude

that A is conjugate to the matrix

1 1
.
0 1
In particular, there is a unique conjugacy class of elements of order p in
GL2 (F).
3.7 Orders of elements

Let q be a prime power. Using the minimal polynomial, we can describe the
order of A ∈ GLn (q). Indeed, we have Ak = In if and only if Ak − In = 0,
which by Lemma 3.15 is equivalent to mA (t) dividing the polynomial tk − 1.
Thus we have the following result.
Lemma 3.20. Let A ∈ GLn (q). Then |A| is the smallest integer k > 0 such
that mA (t) divides tk − 1.
We now illustrate this in degree n = 2 for matrices of determinant 1, and
as an application give a proof of a theorem of Miller on the order of a product
of two elements (Theorem 1.30). First we need a lemma.
Lemma 3.21. Let F be a field of characteristic 6= 2. Let A ∈ SL2 (F) be such
that A has order 2. Then A = −I2 .
Proof. Exercise.
Theorem 3.22 (Miller, 1900). Let m, n, ` > 1 be integers. Then there exists
a finite group G which contains elements x and y such that |x| = m, |y| = n,
and |xy| = `.
Proof. Pick a prime p such that p does not divide 2mn`. The image of p in
(Z/2mn`Z)× has some finite order d, so q = pd is equal to 1 modulo 2mn`.
We will construct elements x, y ∈ SL2 (q) such that |x| = 2m, |y| = 2n, and
|xy| = 2`. Then since −I2 is the unique element of order 2 in SL2 (q) (Lemma
3.21), we conclude that the images of x, y, xy in PSL2 (q) = SL2 (q)/h−I2 i
have orders m, n, `, respectively.
The basic idea is that if A ∈ SL2 (q) is non-scalar, then the minimal
polynomial and the characteristic polynomial of A are equal (Lemma 3.18),
and given by
mA (t) = t2 − tr(A)t + 1.
Thus by Lemma 3.20, the order of a non-scalar matrix A ∈ SL2 (q) is com-
pletely determined by its trace.
78
Now 2m, 2n, 2` divide |F× ×

q | = q − 1, so since Fq is cyclic (Theorem 1.52),
×
we can find α, β, γ ∈ Fq with orders |α| = 2m, |β| = 2n, and |γ| = 2`. Let
δ ∈ Fq and define

α 0 β 1
x= , y= .
δ α−1 0 β −1
Now x has characteristic polynomial (t − α)(t − α−1 ) and is thus similar

to a diagonal matrix with entries α, α−1 . Therefore |x| = 2m, and similarly
|y| = 2n. We have
αβ α
xy = .
δβ δ + (αβ)−1
Now with a suitable choice of δ we can make tr(xy) equal to any element of
Fq we wish, in particular we can choose δ such that xy has order 2`.
Indeed, take δ = γ + γ −1 − αβ − (αβ)−1 , in which case tr(xy) = γ + γ −1
and xy has characteristic polynomial (t − γ)(t − γ −1 ). Then xy is similar to
a diagonal matrix with entries γ, γ −1 , so |xy| = 2`.
3.8 Factorizations of polynomials

In general, when classifying the conjugacy classes of GLn (F), one can reduce
to the case where the characteristic polynomial is equal to p(t)k for some
irreducible polynomial p(t) and integer k > 0. This is a consequence of the
following lemma.
Lemma 3.23. Let A ∈ GLn (F). Suppose that p(t), q(t) ∈ F[t] are polynomi-
als such that gcd(p(t), q(t)) = 1 and p(A)q(A) = 0. Then V = Fn decomposes
as a direct sum
V = Ker p(A) ⊕ Ker q(A).
Proof. Since gcd(p(t), q(t)) = 1, by Bézout’s identity in F[t] we have
a(t)p(t) + b(t)q(t) = 1
for some a(t), b(t) ∈ F[t]. Thus for all v ∈ V we have
v = In v = a(A)p(A)v + b(A)q(A)v. (3.1)
Since p(A)q(A) = 0, we have p(A)b(A)q(A) = b(A)p(A)q(A) = 0, so we

conclude that b(A)q(A)v ∈ Ker p(A). Similarly a(A)p(A) ∈ Ker q(A), so
we have shown that V = Ker p(A) + Ker q(A). To show that the sum is
direct, we need to verify that Ker p(A) ∩ Ker q(A) = 0. To this end, if
v ∈ Ker p(A) ∩ Ker q(A), then it is immediate from (3.1) that v = 0.
79
We illustrate Lemma 3.23 for GL3 (2). Now | GL3 (2)| = 168 = 23 · 3 · 7.
By Cauchy’s theorem there exist elements of order 3 and 7 in GL3 (2). How
to find such elements, and how to classify them up to conjugacy?
We consider elements of order 3 first. Suppose that A ∈ GL3 (2) is such
that |A| = 3. Then the minimal polynomial of A divides t3 − 1, which
factorizes as
t3 − 1 = (t − 1)(t2 + t + 1).
Since A 6= 1, the minimal polynomial must be divisible by t2 + t + 1, and
thus the characteristic polynomial is divisible by t2 +t+1. The characteristic
polynomial has degree 3, so it must be the product of t2 + t + 1 and a factor
of degree 1. This factor cannot be t since A does not have zero as eigenvalue,
so it is t − 1 and so pA (t) = (t − 1)(t2 + t + 1) = t3 − 1.
The polynomials t − 1 and t2 + t + 1 are coprime, so by Lemma 3.23 we
have
F32 = Ker(A − 1) ⊕ Ker(A2 + A + 1).
Since 1 is a root of pA (t), we have dim Ker(A − 1) ≥ 1. A straightforward
check shows that dim Ker(A − 1) = 1, so dim Ker(A2 + A + 1) = 2. Let
v ∈ Ker(A − 1) be non-zero, so A(v) = v. The linear map induced by A on
Ker(A2 +A+1) must have minimal polynomial and characteristic polynomial
A2 + A + 1, so by Lemma 3.9 it has a basis {w, w0 } such that A(w) = w0 and
A(w0 ) = w + w0 .
We conclude then that with respect to the basis {v, w, w0 }, the matrix of
A is  
1 0 0
0 0 1 ,
0 1 1
and so A is conjugate to this matrix. In particular we have shown that
there is a unique conjugacy class of elements of order 3 in GL3 (2), with a
representative given by the matrix above.
It turns out there are two conjugacy classes of elements of order 7 in
GL3 (2). Let A ∈ GL3 (2) be such that |A| = 7. Then A7 = 1, so the
minimal polynomial of A divides t7 − 1 ∈ F2 [t], which has a factorization into
irreducible polynomials as
t7 − 1 = (t − 1)(t3 + t + 1)(t3 + t2 + 1)
in F2 [t].
Since the minimal polynomial of A has degree at most 3 (Lemma 3.17),
it follows that the minimal polynomial of A is either t − 1, t3 + t + 1, or
t3 + t2 + 1. It cannot be t − 1 since A 6= 1, so mA (t) = t3 + t + 1 or t3 + t2 + 1.
80
Both possibilities can be realized, and these correspond to the two conjugacy
classes of elements of order 7. This is seen from the next lemma.
Lemma 3.24. Let A ∈ GL3 (q) with minimal polynomial p(t) = t3 + γt2 +
βt + α irreducible in Fq [t]. Then:
(i) We have pA (t) = p(t), and A is conjugate to the matrix
 
0 0 −α
Cp(t) = 1 0 −β 
0 1 −γ
(ii) For any irreducible polynomial p(t) ∈ Fq [t], the matrix Cp(t) as in (i)
has its characteristic and minimal polynomial equal to p(t).
(iii) Let p(t), q(t) ∈ Fq [t] be irreducible monic polynomials of degree 3. Then
Cp(t) and Cq(t) are conjugate in GL3 (q) if and only if p(t) = q(t).
Proof. (i) Since the minimal polynomial divides pA (t) and deg pA (t) = 3,
it follows that pA (t) = p(t).
Let V = F3q and pick a non-zero vector v ∈ V . Now A has no eigenvalues
since pA (t) = p(t) is irreducible, so Av 6∈ hvi. Thus {v, Av} is linearly
independent. We claim that A2 v 6∈ hv, Avi. For if we had A2 v ∈
hv, Avi, then the action of A on W = hv, Avi would induce a linear
map A0 : W → W . Now dim W = 2, so A0 has minimal polynomial
mA0 (t) of degree ≤ 2. On the other hand p(A0 ) = 0, so mA0 (t) | p(t), a
contradiction since p(t) is irreducible.
Thus A2 6∈ hv, Avi. Then {v, Av, A2 v} is a basis of V , and we can write
A3 v = −α0 v − β 0 Av − γ 0 A2 v for some α0 , β 0 , γ 0 ∈ Fq . Then with respect
to the basis {v, Av, A2 v}, the matrix of A is
0 0 −α0
 
C = 1 0 −β 0  ,
0 1 −γ 0
so A is conjugate to C in GL3 (q). A straightforward calculation shows
that C has characteristic polynomial equal to q(t) = t3 + γ 0 t2 + β 0 t + α0 .
Since A and C have the same characteristic polynomial we get q(t) =
p(t), and so C = Cp(t) .
(ii) A calculation shows that the characteristic polynomial is equal to p(t).
Since the minimal polynomial divides the characteristic polynomial and
since p(t) is irreducible, we conclude that p(t) is also the minimal po-
lynomial.
81
(iii) If Cp(t) and Cq(t) are conjugate, they have the same characteristic po-
lynomial, so p(t) = q(t) by (ii). Conversely if p(t) = q(t), then Cp(t) =
Cq(t) .
As we saw before, for an element A ∈ GL3 (2) of order 7, the minimal

polynomial mA (t) is irreducible and equal to t3 + t + 1 or t3 + t2 + 1. Thus
it follows from Lemma 3.24 that A is conjugate to
 
0 0 1
P = 1 0 1
0 1 0
if mA (t) = t3 + t + 1, and A is conjugate to

 
0 0 1
Q = 1 0 0
0 1 1
if mA (t) = t3 + t2 + 1. Note that by Lemma 3.24, we have mP (t) = t3 + t + 1

and mQ (t) = t3 + t2 + 1.
Example 3.25. Exercise: We have | GL3 (7)| = 26 · 34 · 73 · 19. Find represen-

tatives for conjugacy classes of elements of order 19 in GL3 (7). (There are 6
classes. Use the factorization t19 − 1 = (t − 1)(t3 + 2t − 1)(t3 + 3t2 + 3t −
1)(t3 + 4t2 + t − 1)(t3 + 4t2 + 4t − 1)(t3 + 5t2 − 1)(t3 + 6t2 + 3t − 1) in F7 [t].)
3.9 Generators for GL2 (F) and SL2 (F)

Let F be a field. In GL2 (F), a transvection is a matrix which has one of the
following forms:

1 x 1 0
0 1 y 1
where x, y ∈ F× .
We have the following result.
Lemma 3.26. Any two transvections are conjugate in GL2 (F).
Proof. Immediate from Lemma 3.8 (iii).
Lemma 3.27. The special linear group SL2 (F) is generated by transvections.
82

a b
Proof. Let A ∈ SL2 (F), say A = . We note first that multiplying
c d
A on the left with a transvection corresponds to row operations; and simi-
larly multiplying A on the right with a transvection corresponds to column
operations. That is, for all λ ∈ F:

1 λ a b a + λc b + λd
=
0 1 c d c d

1 0 a b a b
=
λ 1 c d c + λa d + λb

a b 1 λ a b + λa
=
c d 0 1 c d + λc

a b 1 0 a + λb b
=
c d λ 1 c + λd d
Suppose that c 6= 0. Then with row and column operations, we can reduce
A into the identity matrix:
1 b0 1 b0

a b 1 0
→ → →
c d c d 0 d0 0 d0
where d0 = 1 since row and column operations do not change the determinant.
Thus there exist transvections x1 , . . ., xt , y1 , . . ., ys such that
x1 · · · xt Ay1 · · · ys = I2
which implies that A is contained in the subgroup generated by transvections.

If c = 0, then we must have a 6= 0. Then adding the first row to the
second changes A into a matrix where the entry corresponding to c is non-
zero. It follows from the previous case that A is contained in the subgroup
generated by transvections.
Remark 3.28. In general for GLn (F), a transvection is defined as follows.
For 1 ≤ i, j ≤ n, let Ei,j be the (n × n) matrix with 1 as the (i, j) entry
(row i, column j) and zeroes elsewhere. Then in GLn (F), a transvection is
a matrix of the form In + λEi,j , where λ ∈ F and i 6= j. Similarly to the
proof of Lemma 3.27, with Gaussian elimination one can show that SLn (F)
is generated by transvections.
Lemma 3.29. The general linear group GL2 (F)is generated
by the set of
λ 0
transvections, together with matrices of the form , where λ ∈ F× .
0 1
83
Proof. Exercise.
Let q be a prime power. The set of generators for SL2 (q) provided by
Lemma 3.27 has size 2q, while the set of generators for GL2 (q) in Lemma
3.29 has size 3q. It turns out we can find a much smaller set of generators, and
indeed both groups can always be generated by two elements. For example,
× ∼
suppose that q 6= 2 and let ζ ∈ F× q be a generator for Fq = Cq−1 . Then for

ζ 0 0 ζ 0 −1 1
x= x = y=
0 1 0 ζ −1 −1 0
we have GL2 (q) = hx, yi and SL2 (q) = hx0 , yi. This follows from a 1962 result
of Steinberg.
3.10 Simplicity of PSL2 (q)

Let q be a prime power. We know that PSL2 (2) ∼ = S3 and PSL2 (3) ∼
= A4 are
not simple. In this section, we will prove that PSL2 (q) is simple for q > 3.
We will first do this for q = 4 and q > 5. After that we will prove that
PSL2 (5) ∼
= A5 , so then PSL2 (5) is simple since A5 is (Theorem 2.32).
Lemma 3.30. Let α ∈ F× q be such that α 6= ±1. Suppose that A ∈ SL2 (q)
has characteristic polynomial pA (t) = (t − α)(t − α−1 ). Then A is conjugate
to
α 0
S=
0 α−1
in SL2 (q).
Proof. By Lemma 3.9 we know that A is conjugate to S in GL2 (q), in other
words XAX −1 = S for some X ∈ GL2 (q). Let λ = det(X). Now S commutes
with the diagonal matrix −1
λ 0
D= ,
0 1
so DXA(DX)−1 = S. Now det(DX) = det(D) det(X) = 1, so DX ∈ SL2 (q)
conjugates A into S in SL2 (q).
Lemma 3.31. Suppose that q = 4 or q > 5 and let A ∈ SL2 (q) be a non-
scalar matrix. Then there exists B ∈ SL2 (q) such that ABA−1 B −1 is conju-
gate to a matrix of the form
λ 0
0 λ−1
for some λ ∈ F×
q with λ 6= ±1.
84
Proof. Let V = F2q . Since A is non-scalar, there exists a vector v ∈ V \ {0}

which is not an eigenvector of A (Exercise 3.18). In other words we have
Av 6∈ hvi, so {v, Av} is a basis of V . Write A2 v = αv + βAv. Then with
respect to the basis {v, Av}, the matrix of A is

0 α
.
1 β
Since A has determinant 1, we must have α = −1. Thus by replacing A with

a conjugate, we may assume that

0 −1
A= .
1 β
Define
µ 0
B= .
0 µ−1
Then B ∈ SL2 (q), and ABA−1 B −1 is equal to
−1
µ−2

0 −1 µ 0 β 1 µ 0 0
= .
1 β 0 µ−1 −1 0 0 µ β − βµ−2 µ2
Hence ABA−1 B −1 has characteristic polynomial (t − µ2 )(t − µ−2 ), so by

Lemma 3.30 it is conjugate to
2
µ 0
0 µ−2
in SL2 (q). Since we are assuming q = 4 or q > 5, there exists µ ∈ F×

q such
that µ2 6= ±1, so the lemma follows.
Theorem 3.32. Suppose that q = 4 or q > 5, and let N be a normal subgroup

of SL2 (q). Then N = {1}, N = Z(SL2 (q)) = {±I2 }, or N = SL2 (q).
Proof. Let N E SL2 (q) be such that N 6= {1} and N 6= Z(SL2 (q)). We will
show that N contains every transvection, which by Lemma 3.27 implies that
N = SL2 (q).
Since N is not contained in Z(SL2 (q)), there exists a non-scalar matrix
A ∈ N . Then by Lemma 3.31, we can find B ∈ SL2 (q) such that ABA−1 B −1
is conjugate in SL2 (q) to a diagonal matrix

λ 0
D=
0 λ−1
85
for some λ ∈ F×q with λ 6= ±1. Now ABA B

−1 −1
∈ N since N is a normal
subgroup, so we conclude that D ∈ N . By Lemma 3.30, we find that N
contains every matrix in SL2 (q) with characteristic polynomial (t−λ)(t−λ−1 ).
Therefore
λ xλ−1
−1
λ 0 1 x
= ∈N
0 λ−1 0 λ 0 1
for all x ∈ Fq . Similarly by taking transposes

1 0
∈N
x 1
for all x ∈ Fq . This completes the proof of the theorem.
Corollary 3.33. Suppose that q = 4 or q > 5. Then PSL2 (q) is simple.
Proof. By Theorem 3.32, there only normal subgroups N E SL2 (q) with
Z(SL2 (q)) ≤ N ≤ SL2 (q) are N = Z(SL2 (q)) and N = SL2 (q). Thus by the
correspondence theorem PSL2 (q) = SL2 (q)/Z(SL2 (q)) is simple.
Theorem 3.34. PGL2 (5) ∼
= S5 .
Proof. For GL2 (5), we have anatural action
on a 2-dimensional vector space
1 0
V = F25 over F5 . Let e1 = and e2 = be the standard basis of column
0 1
vectors. Since GL2 (5) acts on V , it also acts on the set of 1-dimensional
subspaces of V :
Ω = {hvi : v ∈ V \ {0}}.
(The action is the obvious one: g · hvi = hgvi for all g ∈ GL2 (5) and v ∈
V \ {0}.) The set Ω is known as the projective line P1 (F5 ) over F5 , or the
Grassmannian Gr(1, V ).
It is readily seen that there are 6 subspaces in Ω; hence there is a homo-
morphism ϕ : GL2 (5) → S6 corresponding to this action. By Exercise 3.18
the kernel Ker ϕ consists of the scalar matrices, thus ϕ(GL2 (5)) ∼
= PGL2 (5)
by the first isomorphism theorem.
We label the points in Ω by 1,2,3,4,5,6 as follows:
1 2 3 4 5 6
he1 i he2 i he1 + e2 i h2e1 + e2 i h3e1 + e2 i h4e1 + e2 i

1 1
Then for example is mapped to (2 3 4 5 6) by ϕ. For S5 we will
0 1
find a faithful action on 6 points, such that for the corresponding homomor-
phism ψ : S5 → S6 we have ψ(S5 ) = ϕ(GL2 (5)). Then S5 ∼ = ϕ(G) ∼
= PGL2 (5)
and the theorem follows.
86
To this end, let Ω0 be the set of subgroups of order 5 in S5 , and consider

the action of S5 on Ω0 by conjugation. Each subgroup in Ω0 is generated by
a 5-cycle, and there are a total of
5·4·3·2·1
= 24
5
5-cycles in S5 . In each subgroup of order 5 there are a total of four 5-cycles,
and distinct subgroups of order 5 have no elements of order 5 in common.
Therefore the total number of subgroups of order 5 in S5 is
|Ω0 | = 24/4 = 6.
Let ψ : S5 → S6 be the homomorphism corresponding to the action of S5

on Ω0 . We label the elements of Ω0 has H1 , H2 , H3 , H4 , H5 , H6 as follows:
H1 H2 H3 H4 H5 H6
h(1 2 3 4 5)i h(1 2 3 5 4)i h(1 5 2 3 4)i h(1 3 4 5 2)i h(1 3 2 4 5)i h(1 2 4 3 5)i
We have chosen this labeling so that for σ = (1 2 3 4 5) ∈ H1 we have

1 1
ψ(σ) = (2 3 4 5 6) = ϕ .
0 1
Note that this implies that Ker ψ = {(1)} since the only normal subgroups
of S5 are {(1)}, A5 , and S5 (Exercise 2.12). Thus S5 ∼
= ψ(S5 ).
For τ = (1 2) and σ = (1 2 3 4 5) we have S5 = hτ, σi (Exercise 2.2), so
∼
S5 = ψ(S5 ) = hψ(τ ), ψ(σ)i. A calculation shows that ψ(τ ) = (1 4)(2 3)(5 6)
and we know that ψ(σ) = (2 3 4 5 6), so
ψ(S5 ) = h(1 4)(2 3)(5 6), (2 3 4 5 6)i.
Next we will show that ψ(S5 ) ≤ ϕ(GL2 (5)) and thus ψ(S5 ) = ϕ(GL2 (5))
since |S5 | = 120 and |ϕ(GL2 (5))| = | PGL2 (5)| = 120. To this end, it will
thatψ(τ ), ψ(σ) ∈ ϕ(GL2 (5)). We already saw that ψ(σ) is
suffice to prove
1 1
the image of . For ψ(τ ) we would need a matrix A ∈ GL2 (5) that
0 1
maps the first basis vector e1 to a scalar multiple of 2e1 + e2 (and vice versa),
and the second basis vector e2 to a scalar multiple of e1 + e2 (and vice versa).
This determines A up to a scalar, and indeed a straightforward calculation
shows that
2 3
ϕ = (1 4)(2 3)(5 6) = ψ(τ ).
1 3
This completes the proof of the theorem.
87
Corollary 3.35. PSL2 (5) ∼

= A5 . In particular, PSL2 (5) is simple.
Proof. We note that PSL2 (5) can be embedded as a normal subgroup of
PGL2 (5). Indeed, this is true in general: let π : GLn (q) → PGLn (q) be
the natural quotient map, so Ker π is the set of scalar matrices in GLn (q)
by Lemma 3.4. Then π(SLn (q)) E PGLn (q) since SLn (q) E GLn (q), and
π(SLn (q)) ∼
= PSLn (q) by the first isomorphism theorem and Lemma 3.4.
Now | PSL2 (5)| = 60 and | PGL2 (5)| = 120, so PSL2 (5) is a normal sub-
group of index 2 in PGL2 (5). Since A5 is the unique subgroup of index 2 in
S5 (Proposition 2.27), by Theorem 3.34 we have PSL2 (5) ∼= A5 . Since A5 is
simple (Theorem 2.32), we conclude that PSL2 (5) is simple.
Here is a list of some exceptional isomorphisms involving PSL2 (q) and
PGL2 (q):
PSL2 (2) ∼
= S3
PGL2 (3) ∼
= S4
∼
PSL2 (3) = A4
PSL2 (4) ∼
= A5
∼
PGL2 (5) = S5
PSL2 (5) ∼
= A5
PSL2 (7) ∼
= PSL3 (2)
∼
PSL2 (9) = A6
PSL4 (2) ∼
= A8
88
4 Normal series
4.1 Characteristic subgroups
Definition 4.1. Let G be a group. A subgroup H ≤ G is a characteristic
subgroup, if ϕ(H) = H for every automorphism ϕ : G → G. We denote this
by H char G.
Example 4.2. Some basic examples:
(a) In any group G, the trivial subgroup {1} and the group G itself are
characteristic subgroups.
(b) For any group G, we have Z(G) char G.
(c) If G is a finite group such that G has exactly one subgroup N with |N | =
d, then N char G. This is because for any automorphism ϕ : G → G, the
image ϕ(N ) is also a subgroup of order d. Thus if G is finite cyclic, then
every subgroup of G is characteristic.
Lemma 4.3. Let G be a group. Then the following statements hold:
(i) Suppose that H char G. Then H E G.
(ii) Suppose that H char K char G. Then H char G.
(iii) Suppose that H char K E G. Then H E G.
Proof. (i) Let g ∈ G and consider the inner automorphism γg : G → G
defined by γg (x) = gxg −1 . Since H is characteristic in G, we have
H = γg (H) = gHg −1 . Therefore H E G.
(ii) Let ϕ : G → G be an automorphism of G. Since K is characteristic in G,
we have ϕ(K) = K and so the restriction of ϕ to K is an automorphism
ϕ0 : K → K of K. Since H is characteristic in K, we have ϕ0 (H) = H,
and thus ϕ(H) = H.
(iii) Let g ∈ G. Since K is a normal subgroup, the inner automorphism
γg : G → G restricts to an automorphism γg0 : K → K of K. Because H
is characteristic in K, we have γg0 (H) = H, in other words gHg −1 = H.
Thus H E G.
Example 4.4. Not all normal subgroups are characteristic. For example,
consider the subgroup G = h(1 2), (3 4)i of S4 . One checks that conjugation
by g = (1 3 2 4) gives an automorphism ϕ : G → G, where ϕ(x) = gxg −1
for all x ∈ G. Then ϕ((1 2)) = (3 4) and ϕ((3 4)) = (1 2), so the normal
subgroups h(1 2)i and h(3 4)i are not characteristic in G.
89
4.2 Commutator subgroups and solvability

Let G be a group. For x, y ∈ G we have xy = yx if and only if x−1 y −1 xy = 1.
We call x−1 y −1 xy the commutator of x and y, and denote it by
[x, y] := x−1 y −1 xy.
Definition 4.5. Let G be a group. The commutator subgroup of G is the

subgroup [G, G] := h[x, y] : x, y ∈ Gi; in other words, the subgroup generated
by the commutators in G. More generally, for subgroups H, K ≤ G, we define
the commutator subgroup of H and K as
[H, K] := h[x, y] : x ∈ H, y ∈ Ki.
Remark 4.6. In general it is not true that the set of commutators is a sub-
group. In other words, it may happen that [G, G] is not the set of commuta-
tors in G. Later in this section we will provide an example for G = SL2 (R).
(For finite groups, the smallest example has order |G| = 96.)
At this point we provide the following example which shows that for
H, K ≤ G the subgroup [H, K] is not necessarily equal to the set of commuta-
tors [h, k] with h ∈ H and k ∈ K. Let G = A4 and consider H = h(1 2)(3 4)i
and K = h(1 2 3)i. Then
{[h, k] : h ∈ H, k ∈ K} = {(1), (1 3)(2 4), (1 4)(2 3)}

[H, K] = {(1), (1 2)(3 4), (1 3)(2 4), (1 4)(2 3)}
There are also many examples where [G, G] is equal to the set of all
commutators, this holds for example for G = Sn or G = An . The Ore
conjecture states that every element of a finite non-abelian simple group is
a commutator. The Ore conjecture is now a theorem, and its proof was
completed by Liebeck, O’Brien, Shalev, and Tiep (J. Eur. Math. Soc., 2010)
using the classification of finite simple groups.
Recall that for elements g and x in a group G, we denote g x = x−1 gx.

Thus [g, x] = g −1 g x for all g, x ∈ G.
Lemma 4.7. Let G be a group and x, y, z ∈ G. Then the following identities

hold:
(i) [x, y]−1 = [y, x].
(ii) xy = yx[x, y].
(iii) [xy, z] = [x, z]y [y, z].
90
(iv) [x, yz] = [x, z][x, y]z .
(v) [x, y]z = [xz , y z ].
Proof. Exercise.
Note that by Lemma 4.7, we have [H, K] = [K, H] for any subgroups H
and K of a group G.
Lemma 4.8. Let G be a group and x, y ∈ G. Suppose that x and y commute

with [x, y]. Then
n(n−1)
(xy)n = xn y n [y, x] 2
for all n ∈ Z≥0 .
Proof. Exercise.
Lemma 4.9. Let G be a group and H ≤ G. Then H E G if and only if

[H, G] ≤ H.
Proof. Suppose that H E G. Then [h, g] = h−1 g −1 hg ∈ H for all h ∈ H

and g ∈ G, since g −1 hg ∈ H by normality. Thus [H, G] ≤ H. Conversely,
suppose that [H, G] ≤ G. Then in particular [h, g] = h−1 g −1 hg ∈ H for all
h ∈ H and g ∈ G, so h[h, g] = g −1 hg ∈ H for all g ∈ G and h ∈ H. Thus
H E G.
Lemma 4.10. Let G be group. Then the following statements hold:
(i) G is abelian if and only if [G, G] = {1}.
(ii) If [G, G] ≤ N ≤ G, then N E G. In particular [G, G] E G.
(iii) G/[G, G] is abelian. For N E G, the quotient G/N is abelian if and

only if [G, G] ≤ N .
Proof. (i) Clear.
(ii) Suppose that [G, G] ≤ N ≤ G. Let g ∈ G and n ∈ N . Then g −1 ng =

n(n−1 g −1 ng) = n[n, g] ∈ N because [n, g] ∈ G. Therefore N E G.
(iii) We have [x[G, G], y[G, G]] = [x, y][G, G] = [G, G] in G/[G, G], so G/[G, G]
is abelian. If N E G, then by (i) the quotient G/N is abelian if and
only if [xN, yN ] = N for all x, y ∈ G, which is equivalent to [x, y] ∈ N
for all x, y ∈ G.
91
Example 4.11. (i) For any abelian group G, we have [G, G] = {1}.
(ii) For G = S3 , we have [G, G] = h(1 2 3)i. More generally [Sn , Sn ] = An

for n ≥ 3.
(iii) Exercise: Let p be a prime and let G be a non-abelian p-group of order

p3 . Show that [G, G] has order p and [G, G] = Z(G).
(iv) Let q be a power of a prime and n ≥ 2. For G = GLn (q), the deter-
minant provides a homomorphism det : G → F× ×
q . Since Fq is abelian,
so is G/ Ker(det) = G/ SLn (q), so it follows from Lemma 4.10 (ii) that
[G, G] ≤ SLn (q). (It is also easy to check directly that det([x, y]) = 1
for all x, y ∈ GLn (q).)
Lemma 4.12. Let F be a field and |F| > 3. Then SL2 (F) = [SL2 (F), SL2 (F)].
Proof. (This lemma would follow from the fact that PSL2 (F) is simple when
|F| > 3, but we will give the following computational proof.) We know that
SL
2 (F)is generated
by transvections (Lemma 3.27), i.e., matrices of the form

1 x 1 0
and for x, y ∈ F. Thus it will suffice to prove that every
0 1 y 1
transvection is contained in the commutator subgroup. To this end, we have
−1 −1
1 x a 0 1 −x a 0 1 x a 0
, = −1
0 1 0 a 0 1 0 a 0 1 0 a
2

1 (a − 1)x
=
0 1
for all a ∈ F× and x ∈ F. Since |F| > 3, we can choose a ∈ F× such that
0
a 6= ±1, so (a2 − 1) 6= 0. Then choosing x = a2x−1 gives
−1
1 x0

1 x a 0
, = ∈ [SL2 (F), SL2 (F)]
0 1 0 a 0 1

0 1 0
for any x ∈ F. Taking transposes shows that ∈ [SL2 (F), SL2 (F)],
x0 1
which completes the proof of the lemma.
Lemma 4.13. Let F be a field and |F| ≥ 3. Then [GL2 (F), GL2 (F)] =
SL2 (F).
Proof. Exercise. (Argue as in Lemma 4.12.)
92
We now give an example of an element in the commutator subgroup

which is not equal to a commutator. By Lemma 4.12 we have SL2 (R) =
[SL2 (R), SL2 (R)], and we will show that −I2 ∈ SL2 (R) is not a commutator
in SL2 (R).
Lemma 4.14. The matrix −I2 ∈ SL2 (R) is not a commutator in SL2 (R).
Proof. Suppose that [A, B] = A−1 B −1 AB = −I2 for some A, B ∈ SL2 (R).
Then B −1 AB = −A, so tr(A) = tr(−A) and thus tr(A) = 0. Thus the
characteristic polynomial of A is
pA (t) = t2 − tr(A)t + det(A) = t2 + 1,
which is an irreducible polynomial in R[t]. As in the proof of Lemma 3.9, it
follows that A is conjugate to the matrix

0 −1
1 0
in GL2 (R). For every g ∈ GL2 (R) we have [Ag , B g ] = [A, B]g = −I2 and
g g
,B ∈
A SL2 (R), so without loss of generality we may assume that A =
0 −1
.
1 0
It follows from A−1 B −1 AB = −I2 that
A−1 BA = −B.

0 −1 x y
With A = , this implies that B is of the form B = for
1 0 y −x
some x, y ∈ R. But then det(B) = −(x2 + y 2 ), which cannot be equal to 1
for x, y ∈ R, a contradiction.
Remark 4.15. However, note that −I2 is a commutator in SL2 (C): for
example for

0 −1 i 0
A= B=
1 0 0 −i
we have [A, B] = −I2 . (We have seen these matrices before in Section 1.8,
where they appeared in the definition of the quaternion group Q8 .)
Suppose that G is a group generated by a subset S, say G = hSi. In
general it is not true that [G, G] is generated by the commutators [x, y] with
x, y ∈ S. For example, let n ≥ 4 and consider the symmetric group G = Sn .
Then G is generated by S = {(1 2), (1 2 · · · n)}, and it is an exercise to
check that {[x, y] : x, y ∈ S} does not generate [G, G]. The correct result in
this direction is given by the following lemma.
93
Lemma 4.16. Let G be a group and suppose that G = hSi for some S ⊆ G.
Then
[G, G] = h[x, y]g : x, y ∈ S and g ∈ Gi.
Proof. Exercise.
Definition 4.17. Let G be a group. We define the derived series of G as the

following series of subgroups: G(0) = G, G(1) = [G, G], G(2) = [G(1) , G(1) ], . . .
and in general G(k) = [G(k−1) , G(k−1) ] for k > 1.
For any group G, we have
G = G(0) D G(1) D G(2) D G(3) D · · ·
and G(i) /G(i+1) is an abelian group for all i ≥ 0. If there exists an integer
k ≥ 0 such that G(k) = {1}, then we say that G is solvable.
Example 4.18. The following groups are solvable: any abelian group, fi-
nite dihedral groups D2n , infinite dihedral group D∞ , quaternion group Q8 ,
symmetric groups S3 and S4 , any finite p-group.
Example 4.19. A group G is said to be perfect if G = [G, G]. It is clear

that a non-trivial perfect group cannot be solvable, since the G(k) = G for all
k ≥ 0. By Lemma 4.12, the special linear groups SL2 (F) are perfect whenever
F is a field with |F| > 3.
Lemma 4.20. Let ϕ : G → H be a homomorphism between two groups G

and H. Then hϕ(S)i = ϕ(hSi) for any S ⊆ G.
Proof. Exercise.
Lemma 4.21. Let ϕ : G → H be a homomorphism between two groups G

and H. Then the following statements hold:
(i) ϕ([x, y]) = [ϕ(x), ϕ(y)] for all x, y ∈ G.
(ii) ϕ([A, B]) = [ϕ(A), ϕ(B)] for all A, B ≤ G.
(iii) ϕ(G(k) ) = ϕ(G)(k) for all k ≥ 0.
Proof. Claim (i) is an exercise, while (ii) follows from Lemma 4.20 and (i).
Claim (iii) follows from (ii), by induction on k.
Lemma 4.22. Let G be a group. Then G(k) char G for all k ≥ 0, and in
particular G(k) E G.
94
Proof. Follows from Lemma 4.21 (iii) and Lemma 4.3 (i).
Lemma 4.23. Let G be a group. Then the following hold:
(i) If G is solvable, then every subgroup of G is solvable.
(ii) Let N E G. Then G is solvable if and only if both N and G/N are
solvable.
Proof. (i) For any H ≤ G we have H (k) ≤ G(k) for all k ≥ 0. Thus if G is
solvable and G(k) = {1}, then H (k) = {1} and so H is solvable as well.
(ii) Let π : G → G/N be the canonical homomorphism. Suppose first

that G is solvable, say G(k) = {1}. Then N is solvable by (i). We
have π(G)(k) = π(G(k) ) = {1} by Lemma 4.21, so π(G) = G/N is also
solvable.
Conversely, suppose that both N and G/N are solvable. Let k ≥ 0 be
such that (G/N )(k) is trivial. Since π(G)(k) = π(G(k) ), it follows that
G(k) ≤ Ker π = N . Now let ` ≥ 0 be such that N (`) = {1}. Then
G(k+`) ≤ N (`) = {1}, so G(k+`) = {1} and G is solvable.
Lemma 4.24. Let G be a group and let H, K ≤ G be solvable subgroups such

that H E G. Then HK is solvable.
Proof. Exercise.
Remark 4.25. Suppose that H, K ≤ G are solvable subgroups such that

HK is a subgroup. Then it is not in general true that HK is solvable, an
example is provided by G = A5 which is not solvable, but G = HK for
H = A4 and K cyclic of order 5.
4.3 Jordan–Hölder theorem

Definition 4.26. Let G be a group. A series is a sequence
G = G1 D G2 D · · · D Gs D Gs+1 = {1}
of subgroups such that Gi+1 E Gi for all 1 ≤ i ≤ s. We call s the length

of the series. The subgroups Gi are called the terms of the series, and the
groups Gi /Gi+1 are called quotients or factors of the series.
95
For example, if G is a solvable group with G(k) = {1}, then the derived
series is a series
G = G(0) D G(1) D · · · D G(k−1) D G(k) = {1}
where the quotients G(i) /G(i+1) are abelian groups. In fact, the existence of
a series with abelian quotients is equivalent to solvability, as we will prove
next.
Lemma 4.27. Let G be a group. Then G is solvable if and only if there
exists a series
G = G1 D G2 D · · · D Gs D Gs+1 = {1}
such that Gi /Gi+1 is abelian for all 1 ≤ i ≤ s.

Proof. If G is solvable, then as noted before the lemma, the derived series
provides a series of subgroups in G with abelian quotients. Conversely, sup-
pose that
G = G1 D G2 D · · · D Gs D Gs+1 = {1}
is a series of subgroups such that Gi /Gi+1 is abelian for all 1 ≤ i ≤ s. It
follows from Lemma 4.10 (iii) that [G, G] = G(1) ≤ G2 .
In general, we claim that
G(k) ≤ Gk+1
for all 1 ≤ k ≤ s. We prove this by induction on k, the case k = 1 having

been proven already. Suppose that 1 < k ≤ s. By induction G(k−1) ≤ Gk .
Now Gk /Gk+1 is abelian, so by Lemma 4.10 (iii) we have
Gk+1 ≥ [Gk , Gk ] ≥ [G(k−1) , G(k−1) ] = G(k) .
This completes the proof of the claim. With k = s we get G(s) ≤ Gs+1 = {1},
so G(s) = {1} and G is solvable.
Definition 4.28. Let G be a nontrivial group. A series
G = G1 D G2 D · · · D Gs D Gs+1 = {1}
is called a composition series, if Gi /Gi+1 is simple for all 1 ≤ i ≤ s.

Example 4.29. Examples of composition series:
(a) Exercise: Let G = Cpk , where p is a prime. Show that G has only one
composition series.
96
(b) Exercise: Let G = Cpq , where p and q are distinct primes. How many
composition series does G have?
(c) Let G = D8 = hx, yi with |x| = 2, |y| = 4 and xyx−1 = y −1 . Then G has
several composition series:
G . hyi . hy 2 i . {1}
G . hx, y 2 i . hxi . {1}
G . hx, y 2 i . hxy 2 i . {1}
G . hx, y 2 i . hy 2 i . {1}
G . hxy, y 2 i . hxyi . {1}
G . hxy, y 2 i . hxy 3 i . {1}
G . hxy, y 2 i . hy 2 i . {1}
(d) If G 6= {1} is simple, then G has a unique composition series G . {1}.

(e) Exercise: Show that Z does not admit a composition series.
Definition 4.30. Let G be a group. Two series
G = G1 D G2 D · · · D Gs D Gs+1 = {1}
and
G = H1 D H2 D · · · D Ht D Ht+1 = {1}
are said to be equivalent, if s = t and there exists a permutation (π(1), . . . , π(s))
of (1, . . . , s) such that Gi /Gi+1 ∼
= Hπ(i) /Hπ(i)+1 for all 1 ≤ i ≤ s.
Lemma 4.31. Let G be a group. Suppose that B E A ≤ G and C E G.
Then:
(i) B ∩ C E A ∩ C and A∩C ∼
=
B(A∩C)
B∩C
E A. B B
(ii) BC E AC and AC ∼
= A
.
BC B(A∩C)
Proof. (i) Let g ∈ A ∩ C. Then (B ∩ C)g = B g ∩ C g = B ∩ C, because

B g = B by B E A and C g = C by C E G. Moreover, we have
B E B(A ∩ C) since B E A. Thus
B(A ∩ C) ∼ A∩C A∩C
= =
B (A ∩ C) ∩ B B∩C
by Theorem 1.100. Since B and A ∩ C are normal subgroups of A, the
product B(A ∩ C) is a normal subgroup of A, and so by the correspon-
dence theorem
B(A ∩ C) A
E .
B B
97
(ii) For all a ∈ A, we have (BC)a = B a C a = BC since B E A and

C E G. We also have (BC)c = BC for all c ∈ C, since C ≤ BC.
Thus BC g = BC for all g ∈ AC, in other words BC E AC. For the
isomorphism, we have
AC A(BC) ∼ A
= =
BC BC A ∩ BC
by Theorem 1.100. Now the claim follows from Exercise 4.18, where it
is shown that A ∩ BC = B(A ∩ C).
Lemma 4.32. Let G be a group that has a composition series. If N is

a nontrivial proper normal subgroup of G, then both N and G/N have a
composition series.
Proof. Let
G = G1 . G2 . · · · . Gs . Gs+1 = {1}
be a composition series for G, so Gi /Gi+1 is simple for all 1 ≤ i ≤ s.
For N , taking intersections gives the following series:
N = G1 ∩ N D G2 ∩ N D · · · D Gs ∩ N D Gs+1 ∩ N = {1}.
Now
Gi ∩ N ∼ (Gi ∩ N )Gi+1 Gi
= E
Gi+1 ∩ N Gi+1 Gi+1
by Lemma 4.31 (i), so for all 1 ≤ i ≤ s, the quotient (Gi ∩ N )/(Gi+1 ∩ N ) is
isomorphic to a normal subgroup of Gi /Gi+1 . Since each Gi /Gi+1 is simple,
we conclude that each quotient (Gi ∩N )/(Gi+1 ∩N ) is either trivial or simple,
and thus N admits a composition series.
For G/N , taking the images of Gi in G/N we get the following series:
G/N = G1 N/N D G2 N/N D · · · D Gs N/N D Gs+1 N/N = {1}.
For the quotients, for all 1 ≤ i ≤ s we have

Gi N/N ∼ Gi N ∼ Gi ∼ Gi /Gi+1
= = =
Gi+1 N/N Gi+1 N Gi+1 (N ∩ Gi ) Gi+1 (N ∩ Gi )/Gi+1
by Lemma 4.31 (ii) and Theorem 1.101. Thus each quotient GGi+1 i N/N
N/N
is
isomorphic to a quotient of Gi /Gi+1 . Since Gi /Gi+1 is simple, each quotient
Gi N/N
Gi+1 N/N
must be trivial or simple. Therefore G/N admits a composition
series.
98
Theorem 4.33 (Jordan-Hölder theorem). Let G be a nontrivial group. Sup-

pose that G has a composition series. Then any two composition series of G
are equivalent.
Proof. Denote by `(G) the minimal length of a composition series of G. We

will prove the theorem by induction on `(G). If `(G) = 1, then G is simple
and the result is obvious. Suppose then that `(G) > 1, and let
G = G1 . G2 . · · · . Gs . Gs+1 = {1}
and
G = H1 . H2 . · · · . Ht . Ht+1 = {1}
be two composition series of G. It suffices to prove the claim in the case
where s = `(G).
If G2 = H2 , then the result follows by applying induction on G2 , since
`(G2 ) ≤ `(G1 ) − 1. Thus we can assume that G2 6= H2 . Then
G D G2 H2 . G2 ,
so it follows from simplicity of G/G2 that G = G2 H2 . Moreover, the inter-

section K2 = G2 ∩ H2 is a normal subgroup of G. By isomorphism theorems,
we have
G/G2 = G2 H2 /G2 ∼ = H2 /K2 ,
and similarly G/H2 = G2 H2 /H2 ∼ = G2 /K2 .
By Lemma 4.32 we can find some composition series of K2 , say
K2 . K3 . · · · . Ku . Ku+1 = {1}.
Now we have several composition series of G = G1 :
G1 . G2 . G3 . · · · . Gs . Gs+1 = {1} (4.1)

G1 . G2 . K2 . · · · . Ku . Ku+1 = {1} (4.2)
G1 . H2 . K2 . · · · . Ku . Ku+1 = {1} (4.3)
G1 . H2 . H3 . · · · . Ht . Kt+1 = {1} (4.4)
Applying induction on G2 , it follows that the series (4.1) and (4.2) are equiva-
lent. Since G1 /G2 ∼= H2 /K2 and G1 /H2 ∼ = G2 /K2 , the series (4.2) and (4.3)
are equivalent. Applying induction on H2 , it follows that the series (4.3)
and (4.4) are equivalent. We conclude then that series (4.1) and (4.4) are
equivalent, as claimed.
99
4.4 Nilpotent groups

An important class of solvable groups is that of nilpotent groups. (Among
finite groups, the typical example is a finite p-group.) There are multiple
ways to define nilpotent groups, we will use the following definition.
Definition 4.34. Let G be a group. We define the lower central series of G

as the following series of subgroups: γ1 (G) = G, and γk (G) = [γk−1 (G), G] for
all k > 1. (So for example γ2 (G) = G(1) = [G, G], and γ3 (G) = [[G, G], G].)
If there exists an integer k ≥ 1 such that γk (G) = {1}, then we say that G
is nilpotent.
(i) For any homomorphism ϕ : G → H, we have ϕ(γk (G)) = γk (ϕ(G)) for

all k ≥ 1.
(ii) γk (G) char G for all k ≥ 1, so in particular γk (G) E G.
Proof. (i) By induction on k. For k = 1 the claim is clear, and for k > 1
the claim follows by applying Lemma 4.21 (ii) and induction.
(ii) Follows from (i).
Lemma 4.36. Let G be a nilpotent group. Then every subgroup and quotient
group of G is nilpotent.
Proof. Suppose that G is nilpotent and let k ≥ 1 be such that γk (G) = {1}.
For any subgroup H ≤ G we have γk (H) ≤ γk (G), so γk (H) = {1} and H
is nilpotent. For quotients, let N E G be a normal subgroup. Applying
Lemma 4.35 (i) to the canonical homomorphism π : G → G/N , we see that
γk (G/N ) = γk (G)N/N = {1}. Thus G/N is also nilpotent.
Lemma 4.37. Let G be a group. Then G(k) ≤ γk+1 (G) for all k ≥ 0. In
particular if G is nilpotent, then G is solvable.
Proof. By induction on k. For k = 0 we have G(0) = G = γ1 (G). Suppose

then that k > 0. Then
G(k) = [G(k−1) , G(k−1) ] ≤ [γk (G), γk (G)] ≤ [γk (G), G] = γk+1 (G),
since G(k−1) ≤ γk (G) by induction.
100
By Lemma 4.37, we know that nilpotent groups are solvable. The converse
is not true: for example G = S3 is solvable since G(2) = {(1)}, but γk (G) =
[G, G] for all k ≥ 2, so G is not nilpotent. Another example is provided by
the following subgroup of GL2 (F), where F is a field:

λ ζ ×
B= : λ, µ ∈ F , ζ ∈ F .
0 µ
Assuming that |F| > 2, a calculation shows that B is solvable but not nilpo-
tent.
Another characterization of nilpotent groups can be given in terms of
series. We know by Lemma 4.35 that for any group, in the lower central
series
G = γ1 (G) D γ2 (G) D γ3 (G) D · · ·
the terms γi (G) are characteristic subgroups in G. Thus each quotient
γi (G)/γi+1 (G) is a normal subgroup of G/γi+1 (G) by the correspondence
theorem. Moreover, we have [γi (G), G] = γi+1 (G), so in fact γi (G)/γi+1 (G)
is contained in the center Z(G/γi+1 (G)) of G/γi+1 (G). The existence of a
series with this property characterizes nilpotent groups.
Lemma 4.38. Let G be a group. Then G is nilpotent if and only if there
exists a series
G = G1 D G2 D · · · D Gs D Gs+1 = {1}
such that Gi E G for all 1 ≤ i ≤ s + 1, and Gi /Gi+1 ≤ Z(G/Gi+1 ) for all

1 ≤ i ≤ s. (Such a series is called a central series.)
Proof. If G is nilpotent, then as noted before the lemma, the lower central
series is such a series. Conversely, suppose that there exists a series
G = G1 D G2 D · · · D Gs D Gs+1 = {1}
such that Gi E G for all 1 ≤ i ≤ s + 1, and Gi /Gi+1 ≤ Z(G/Gi+1 ) for all

1 ≤ i ≤ s.
We claim that for all 2 ≤ k ≤ s + 1, we have γk (G) ≤ Gk . First note that
G/G2 = G1 /G2 ≤ Z(G/G2 ), so G/G2 is abelian, and thus γ2 (G) = [G, G] ≤
G2 . For k > 2 we proceed by induction on k. We have Gk−1 /Gk ≤ Z(G/Gk ),
or equivalently [Gk−1 , G] ≤ Gk . By induction γk−1 (G) ≤ Gk−1 , so then
γk (G) = [γk−1 (G), G] ≤ [Gk−1 , G] ≤ Gk
as claimed. Applying this with k = s + 1, we get γs+1 (G) ≤ Gs+1 = {1}, so

γs+1 (G) = {1} and G is nilpotent.
101
Lemma 4.39. Let G be a nilpotent group and N E G such that N 6= {1}.

Then N ∩ Z(G) 6= {1}.
Proof. Let k ≥ 1 be the largest integer such that N ∩ γk (G) 6= {1}, so then
N ∩ γk+1 (G) = {1}. (Such an integer exists since N is nontrivial, and since
G is nilpotent.)
Then [N ∩ γk (G), G] ≤ [N, G] ≤ N since N is a normal subgroup, and
[N ∩γk (G), G] ≤ [γk (G), G] = γk+1 (G). Thus [N ∩γk (G), G] ≤ N ∩γk+1 (G) =
{1}, and so [G, N ∩ γk (G)] = {1}. Therefore N ∩ γk (G) is in the center of G,
and in particular N ∩ Z(G) 6= {1}.
Applying Lemma 4.39 with G = N , we get the following corollary.
Corollary 4.40. Let G be a nontrivial nilpotent group. Then Z(G) 6= {1}.
For solvable groups, we have seen that they are closed under extensions:
in other words if N E G is such that N and G/N are both solvable, then G
is solvable. This fails for nilpotent groups: for example G = S3 has a normal
subgroup N with N ∼ = C3 and G/N ∼ = C2 , but G is not nilpotent. However,
we can now prove the following result.
Lemma 4.41. Let G be a nilpotent group and N ≤ Z(G). Then G is nilpo-
tent if and only if G/N is nilpotent. In particular, G is nilpotent if and only
if G/Z(G) is nilpotent.
Proof. If G is nilpotent, then G/N is nilpotent by Lemma 4.36. Conversely,
suppose that G/N is nilpotent, say γk (G/N ) = {1}. By Lemma 4.21 applied
to the canonical homomorphism G → G/N , we have γk (G/N ) = γk (G)N/N .
Therefore γk (G) ≤ N , and then γk+1 (G) = [γk (G), G] = {1} since N is
central.
As a corollary, we prove that finite p-groups are nilpotent.
Lemma 4.42. Let G be a finite p-group. Then G is nilpotent.
Proof. By induction on |G|. The case |G| = 1 is trivial, so suppose that
|G| > 1. By Corollary 1.143 we have Z(G) 6= {1}. Thus by induction
G/Z(G) is nilpotent, and then G is nilpotent by Lemma 4.41.
As we have seen already, several of the results we have proven before
for finite p-groups hold more generally for nilpotent groups. The following
proposition is a generalization of Proposition 1.147.
Proposition 4.43. [“Normalizers grow”] Let G be a nilpotent group and H
be a proper subgroup of G. Then H NG (H).
Proof. Exercise. (Hint: Show that there exists k ≥ 1 such that γk (G) 6≤ H
and γk+1 (G) ≤ H.)
102
4.5 Upper central series

Let G be a group. By the correspondence theorem, there exists a normal
subgroup Z(G) ≤ Z 2 (G) ≤ G such that Z 2 (G)/Z(G) = Z(G/Z(G)). By
definition, we have
g ∈ Z 2 (G) if and only if [g, x] ∈ Z(G) for all x ∈ G.
Similarly, we define Z 3 (G) to be unique subgroup Z 2 (G) ≤ Z 3 (G) E G such

that Z 3 (G)/Z 2 (G) = Z(G/Z 2 (G)). Then
g ∈ Z 3 (G) if and only if [g, x] ∈ Z 2 (G) for all x ∈ G.
Continuing in this manner, we can define the upper central series, which
gives another way of defining nilpotent groups. The upper central series is
the series
1 = Z 0 (G) ≤ Z 1 (G) ≤ Z 2 (G) ≤ Z 3 (G) ≤ · · ·
of normal subgroups of G, such that Z 0 (G) = {1}, Z 1 (G) = Z(G), and for
k > 1 we have Z k (G) ≤ Z k−1 (G) E G and Z k (G)/Z k−1 (G) = Z(G/Z k−1 (G)).
Then for all k ≥ 1, we have
g ∈ Z k (G) if and only if [g, x] ∈ Z k−1 (G) for all x ∈ G.
Lemma 4.44. Let G be a group. Then Z k (G) char G for all k ≥ 0.
Proof. Exercise.
Lemma 4.45. Let G be a nilpotent group, and let c ≥ 1 be such that

γc+1 (G) = {1}. Then γc−i+1 (G) ≤ Z i (G) for all 1 ≤ i ≤ c. In particu-
lar Z c (G) = G.
Proof. Exercise.
Lemma 4.46. Let G be a group and suppose that Z c (G) = G for some c ≥ 1.
Then γc−i+1 (G) ≤ Z i (G) for all 1 ≤ i ≤ c. In particular γc+1 (G) = {1}.
Proof. Exercise.
By Lemma 4.45 and Lemma 4.46, a group G is nilpotent if and only if
Z (G) = G for some c ≥ 1. Moreover, we have Z c (G) = G if and only if
c
γc+1 (G) = {1}. We will call the smallest integer c ≥ 1 such that γc+1 (G) =
{1} the class of a nilpotent group G.
103
4.6 Higher commutators

Let G be a group. For n ≥ 3, we define the higher commutator of elements
x1 , x2 , . . ., xn ∈ G recursively as
[x1 , x2 , . . . , xn ] := [[x1 , . . . , xn−1 ], xn ].
Therefore for example

[x, y, z] = [[x, y], z]
and
[x, y, z, w] = [[[x, y], z], w]
for all x, y, z, w ∈ G. Similarly for subgroups H1 , . . ., Hn ≤ G we define the
higher commutator subgroup of H1 , . . ., Hn as
[H1 , . . . , Hn ] := [[H1 , . . . , Hn−1 ], Hn ].
Thus for example

[H, K, L] = [[H, K], L]
and for k ≥ 2 we have
γk (G) = [G, G, . . . , G] .
| {z }
k times
In our definition we have used “left-normed” commutators. Taking com-

mutators is not associative, in the sense that for x, y, z ∈ G it is possible
that
[[x, y], z] 6= [x, [y, z]]
and for H, K, L ≤ G it is possible that
[[H, K], L] 6= [H, [K, L]].
(Examples can be found already in the smallest non-abelian group, G = S3 .)

Lemma 4.47. Let G be a group and k ≥ 2. Then
γk (G) = h[x1 , x2 , . . . , xk ] : x1 , x2 , . . . , xk ∈ Gi
Proof. Exercise.
By Lemma 4.47, we have yet another definition of nilpotence: a group G
is nilpotent if and only if there exists k ≥ 1 such that
[x1 , x2 , . . . , xk ] = 1
104
for all x1 , . . ., xk ∈ G. For example:
G is nilpotent of class 1 ⇔ [x, y] = 1 for all x, y ∈ G.

G is nilpotent of class ≤ 2 ⇔ [x, y, z] = 1 for all x, y, z ∈ G.
G is nilpotent of class ≤ 3 ⇔ [x, y, z, w] = 1 for all x, y, z, w ∈ G.
···
G is nilpotent of class ≤ c ⇔ [x1 , x2 , . . . , xc+1 ] = 1 for all x1 , x2 , . . . , xc+1 ∈ G.
Lemma 4.48. Let H, K, L E G. Then [HK, L] = [H, L][K, L] and [H, KL] =
[H, K][H, L].
Proof. Exercise.
Lemma 4.49. Let H1 , . . ., Hk , A, B E G, where k ≥ 1. Then
[H1 , . . . , Hr−1 , AB, Hr+1 , . . . , Hk ]

= [H1 , . . . , Hr−1 , A, Hr+1 , . . . , Hk ][H1 , . . . , Hr−1 , B, Hr+1 , . . . , Hk ]
for all 1 ≤ r ≤ k.
Proof. Case 1: r = 1. Applying Lemma 4.48 repeatedly, we get
[AB, H2 , . . . , Hk ] = [[AB, H2 ], H3 , . . . , Hk ] (definition)

= [[A, H2 ][B, H2 ], H3 , . . . , Hk ] (by Lemma 4.48)
= [[[A, H2 ][B, H2 ], H3 ], H4 , . . . , Hk ] (definition)
= [[[A, H2 , H3 ][B, H2 , H3 ], H4 , . . . , Hk ] (by Lemma 4.48)
= ···
= [A, H2 , . . . , Hk ][B, H2 , . . . , Hk ]
as claimed.
Case 2: r = k. In this case
[H1 , . . . , Hk−1 , AB] = [H1 , . . . , Hk−1 , A][H1 , . . . , Hk−1 , B]
by Lemma 4.48.
Case 3: 1 < r < k. In this case
[H1 , . . . , Hr−1 , AB, Hr+1 , . . . , Hk ]

= [[H1 , . . . , Hr−1 , AB], Hr+1 , . . . , Hk ] (definition)
105
= [[H1 , . . . , Hr−1 , A][H1 , . . . , Hr−1 , B], Hr+1 , . . . , Hk ] (by Case 2)

= [[H1 , . . . , Hr−1 , A], Hr+1 , . . . , Hk ][[H1 , . . . , Hr−1 , B], Hr+1 , . . . , Hk ] (by Case 1)
= [H1 , . . . , Hr−1 , A, Hr+1 , . . . , Hk ][H1 , . . . , Hr−1 , B, Hr+1 , . . . , Hk ] (definition)
This completes the proof of the lemma in all cases.
Lemma 4.50. Let H1 , H2 , . . ., Hk , A be normal subgroups of G. Suppose

that among the Hi ’s there are at least d which are equal to A. Then
[H1 , H2 , . . . , Hk ] ≤ γd (A).
Proof. By induction on k. If k = 1 the claim is clear, and for k = 2 we

have [H1 , A] ≤ A and [A, H2 ] ≤ A since A is a normal subgroup. Suppose
then that k > 2. Note that γt (A) char A (Lemma 4.35), so γt (A) is a normal
subgroup of G for all t ≥ 1 by Lemma 4.3 (iii). Thus if Hk 6= A, then by
induction
[H1 , . . . , Hk ] = [[H1 , . . . , Hk−1 ], Hk ]

≤ [γd (A), Hk ]
≤ γd (A).
Similarly if Hk = A, then by induction
[H1 , . . . , Hk ] = [[H1 , . . . , Hk−1 ], A]

≤ [γd−1 (A), A]
= γd (A).
This completes the proof of the lemma.
Proposition 4.51 (Fitting’s theorem). Let G be a group and let A, B E G

be nilpotent. Then AB is nilpotent normal subgroup of G.
Proof. Since A and B are normal subgroups, it follows that AB is a normal

subgroup. For nilpotence, let a and b be such that γa+1 (A) = {1} and
γb+1 (B) = {1}. Now for any k ≥ 2, applying Lemma 4.49 repeatedly we get
γk (AB)
= [AB, AB, . . . , AB]
| {z }
(k times )
= [A, AB, . . . , AB][B, AB, . . . , AB]

= [A, A, AB, . . . , AB][A, B, AB . . . , AB][B, A, AB . . . , AB][B, B, AB, . . . , AB]
106
= ···
Y
= [X1 , X2 , . . . , Xk ]
Xi ∈{A,B}
In other words, γk (AB) is the product of the 2k higher commutator subgroups

[X1 , X2 , . . . , Xk ] with Xi equal to A or B for each 1 ≤ i ≤ k.
Consider one such commutator subgroup [X1 , . . . , Xk ]. Let k = d + e,
where d is the number of Xi ’s which are equal to A, and e is the number of
Xi ’s equal to B. By Lemma 4.50 we have
[X1 , . . . , Xk ] ≤ γd (A) ∩ γe (B).
Suppose that k > a+b. Then d > a or e > b, so γd (A) = {1} or γe (B) = {1},
and therefore [X1 , . . . , Xk ] = {1}. Thus we conclude that γa+b+1 (AB) = {1},
in other words, AB is nilpotent.
As a consequence of Proposition 4.51, a subgroup generated by finitely
many nilpotent normal subgroups is nilpotent. In particular for a finite group
G, this implies the existence of a unique maximal nilpotent normal subgroup.
Indeed, suppose that G is finite and let M be a nilpotent normal subgroup
of G such that |M | is maximal. Then if N E G is nilpotent, then M N is
nilpotent by Proposition 4.51. Thus M N = M by maximality of M , so in
fact N ≤ M and M contains every nilpotent normal subgroup of G. This
of course implies that M is unique17 . We call M the Fitting subgroup of G
and denote it by F (G). In summary, the Fitting subgroup is characterized
by the following properties:
• F (G) E G and F (G) is nilpotent;
• For all N E G nilpotent, we have N ≤ F (G).
As a straightforward consequence, the Fitting subgroup is a characteristic

subgroup of G.
Lemma 4.52. Let G be a finite group. Then F (G) char G.
Proof. Exercise.
17
For if M 0 is another nilpotent normal subgroup of maximal order, then M 0 ≤ M , and
thus M 0 = M by maximality of M 0 .
107
5 Constructing groups
In this section, we will consider various ways of constructing new groups from
old ones.
5.1 Direct products

Definition 5.1. Let G1 , . . ., Gn be groups, where n ≥ 2. The direct product
of G1 , . . ., Gn is the cartesian product G1 × · · · × Gn equipped with the
following product:
(g1 , . . . , gn )(g10 , . . . , gn0 ) = (g1 g10 , . . . , gn gn0 )
where gi , gi0 ∈ Gi for all 1 ≤ i ≤ n. Then G1 × · · · × Gn equipped with this

operation is a group. The identity element is (1G1 , . . . , 1Gn ) and the inverse
of (g1 , . . . , gn ) ∈ G1 × · · · × Gn is the element (g1−1 , . . . , gn−1 ).
Lemma 5.2. Let G1 , . . ., Gn be groups.
(i) If |Gi | < ∞ for all 1 ≤ i ≤ n, then |G1 × · · · × Gn | = |G1 | · · · |Gn |.
(ii) (A × B) × C ∼
= A × (B × C) for any groups A, B, and C.
(iii) Let (π(1), . . . , π(n)) be a permutation of {1, . . . , n}. Then
Gπ(1) × · · · × Gπ(n) ∼
= G1 × · · · × Gn .
ci := {1}×· · ·×{1}×Gi ×{1}×· · ·×{1}. Then G

(iv) Let G ci E G1 ×· · ·×Gn
and
ci ∼
(G1 × · · · × Gn )/G = G1 × · · · × Gi−1 × Gi+1 × · · · × Gn .
Proof. Exercise.
Definition 5.3. Let G be a group and let H1 , . . ., Hn be subgroups of G.
We say that G is the direct product of H1 , . . ., Hn if the map
H1 × · · · × Hn → G, (h1 , . . . , hn ) 7→ h1 · · · hn for all hi ∈ Hi
is an isomorphism.
Lemma 5.4. Let G be a group and let H, K E G. Suppose that H ∩K = {1}.
Then hk = kh for all h ∈ H and k ∈ K.
Proof. Exercise.
108
Lemma 5.5. Let G be a group and let H1 , . . ., Hn be subgroups of G. Then

G is the direct product of H1 , . . ., Hn if and only if all of the following
conditions hold:
(i) G = H1 · · · Hn ;
(ii) Hi is a normal subgroup of G for all 1 ≤ i ≤ n;
(iii) Hi ∩ (H1 · · · Hi−1 Hi+1 · · · Hn ) = {1} for all 1 ≤ i ≤ n.
Proof. Suppose first that G is the direct product of H1 , . . ., Hn ; in ot-

her words, we assume that the map ϕ : H1 × · · · × Hn → G defined by
ϕ(h1 , . . . , hn ) = h1 · · · hn is an isomorphism. Then (i) holds since ϕ is sur-
jective. For (ii), note that we have
ϕ ({1} × · · · × {1} × Hi × {1} × · · · × {1}) = Hi .
Since ϕ is a surjective homomorphism and since {1} × · · · × {1} × Hi × {1} ×

· · · × {1} is a normal subgroup of H1 × · · · × Hn (Lemma 5.2 (iv)), it follows
from Lemma 1.93 that Hi is a normal subgroup of G. For statement (iii),
let h ∈ Hi ∩ (H1 · · · Hi−1 Hi+1 · · · Hn ), say h = hi = h1 · · · hi−1 hi+1 · · · hn with
hj ∈ Hj for all 1 ≤ j ≤ n. Then
h = ϕ(1, . . . , 1, hi , 1, . . . , 1) = ϕ(h1 , . . . , hi−1 , 1, hi+1 , . . . , hn ),
so hj = 1 for all 1 ≤ j ≤ n since ϕ is injective. Thus h = 1, and we conclude

that (iii) holds.
For the other direction, suppose that (i), (ii), and (iii) hold. We will show
that ϕ is an isomorphism. First note that by (iii) we have Hi ∩ Hj = {1} for
all i 6= j, so by (ii) and Lemma 5.4 we have hi hj = hj hi for all hi ∈ Hi and
hj ∈ Hj . It follows then that we have
(h1 h2 · · · hn )(h01 h02 · · · h0n ) = (h1 h01 )(h2 h02 ) · · · (hn h0n )
for all hi , h0i ∈ Hi . In other words
ϕ(h1 , . . . , hn )ϕ(h01 , . . . , h0n ) = ϕ(h1 h01 , . . . , hn h0n )
for all hi , h0i ∈ Hi , so ϕ is a homomorphism. It follows from (i) that ϕ is

surjective. For injectivity, by Lemma 1.63 (iii) it will suffice to check that
Ker ϕ = {1}. Suppose that ϕ(h1 , . . . , hn ) = h1 · · · hn = 1. Since hi hj = hj hi
for i 6= j, it follows that h−1 i = h1 · · · hi−1 hi+1 · · · hn for all 1 ≤ i ≤ n.
Therefore hi = 1 for all 1 ≤ i ≤ n by (iii). This proves that Ker ϕ = {1} and
completes the proof of the lemma.
109
The special case of Lemma 5.5 with two factors (n = 2) is the one that
comes up most often, so we state it in the following lemma.
Lemma 5.6. Let G be a group and let H, K ≤ G. Then G is the direct

product of H and K if and only if G = HK and H, K E G, and H ∩K = {1}.
Example 5.7. (a) Consider the dihedral group G = D4 = h(12), (34)i.

Then G is the direct product of H = h(12)i and K = h(12)i, so G ∼ =
C2 × C2 .
(b) Let G = (Z/8Z)× . Then G = 1, 3, 5, 7 , and an easy calculation shows

that G is the direct product of H = 3 and K = 5 . Thus G ∼ = C2 ×C2 .

(c) Consider G = S3 and the subgroups H = h(12)i and K = h(123)i. Then
G = HK, H ∩ K = {1}, and K is a normal subgroup of G. But H is not
a normal subgroup, so G is not the direct product of H and K. (This is
also clear since H × K is abelian, and G is non-abelian.)
Example 5.8. Let n ≥ 3. Then as we have seen in an exercise (Theorem

1.57, Exercise 1.15), the group (Z/2n Z)× is not cyclic. In fact, we can now
prove that
(Z/2n Z)× ∼
= C2n−2 × C2 .
To this end, Exercise 1.15 (d) shows that H = h5i is a cyclic subgroup of
order 2n−2 in (Z/2n Z)× . We claim then that (Z/2n Z)× is the direct product
of H and K = h−1i, from which the isomorphism follows since K is cyclic
of order 2.
Since H and K are normal subgroups, it will suffice to prove that H ∩K =
1, for then |HK| = |H||K| = 2n−1 = |(Z/2n Z)× | and thus (Z/2n Z)× = HK.
In other words, we should show that −1 6∈ h5i. If we had −1 ≡ 5k mod 2n ,
then in particular −1 ≡ 5k mod 4. But 5k ≡ 1 mod 4 for all k, so this is a
contradiction. Thus H ∩ K = 1 and we get the isomorphism
(Z/2n Z)× ∼
= C2n−2 × C2 .
Previously (Theorem 1.55) we proved that for any odd prime p, the group
(Z/pn Z)× is cyclic for all n > 0. Moreover, it can be shown using the Chinese
remainder theorem that if a, b > 0 are integers with gcd(a, b) = 1, then
(Z/abZ)× ∼
= (Z/aZ)× × (Z/bZ)× .
Applying this on an integer k > 0 with prime factorization k = pn1 1 · · · pnt t ,

we can describe the structure of (Z/kZ)× , since the structure of (Z/pni i Z)×
is now known for all 1 ≤ i ≤ t.
110
Lemma 5.9. Let G = H ×K. Then for any A ≤ H and B ≤ K, the product
A × B is a subgroup of G.
Proof. Exercise.
Remark 5.10. The converse of Lemma 5.9 fails; in general it is not true
that every subgroup of H × K is of the form A × B for some A ≤ H and
B ≤ K. Example: Let G be nontrivial and consider the direct product G×G
and the diagonal subgroup ∆ = {(g, g) : g ∈ G}.
Lemma 5.11. Let H and K be finite groups such that gcd(|H|, |K|) = 1. If
N ≤ H × K, then N = A × B for some A ≤ H and B ≤ K.
Proof. Exercise.
Lemma 5.12. Let G and H be groups. Then the following hold:
(i) (G × H)(k) = G(k) × H (k) for all k ≥ 0.
(ii) γk (G × H) = γk (G) × γk (H) for all k ≥ 1.
(iii) Z k (G × H) = Z k (G) × Z k (H) for all k ≥ 0.
(iv) The direct product G×H is solvable if and only if G and H are solvable.
(v) The direct product G × H is nilpotent if and only if G and H are nil-
potent.
Proof. Exercise.
Lemma 5.13. Let G and H be finite groups. Then the Fitting subgroup
F (G × H) = F (G) × F (H).
Proof. Exercise.
5.2 Classification of finitely generated abelian groups

In this section, we will provide the classification theorem of finitely generated
abelian groups, thus in particular we classify the finite abelian groups. In
the finite case, the first proof was given in 1870 by Kronecker.
The classification theorem could be proven naturally as part of the theory
involving the Smith normal form and the classification of finitely generated
modules over principal ideal domains. To keep these notes self-contained, we
shall instead follow a short proof due to Rado18 .
18
R. Rado, A proof of the basis theorem for finitely generated Abelian groups. J. London
Math. Soc. 26 (1951), 74–75.
111
The key result is that any finitely generated abelian groups is a direct pro-
duct of cyclic groups. For this we first need a lemma. (Note that throughout
this section we will use additive notation for abelian groups.)
Lemma 5.14. Let G = hx1 , . . . , xk i be an abelian group. Let c1 , . . ., ck ≥ 0
be integers such that gcd(c1 , . . . , ck ) = 1. Then there exists y1 , . . . , yk ∈ G
such that both of the following hold:
(i) G = hy1 , . . . , yk i;
(ii) y1 = c1 x1 + · · · + ck xk .
Proof. Let s = c1 + · · · + ck . If s = 1, then ci = 1 for some 1 ≤ i ≤ k and
cj = 0 for all j 6= i; in this case we can choose y1 = xi and {y2 , . . . , yk } =
{x1 , . . . , xi−1 , xi+1 , . . . , xk }.
Suppose then s > 1 and proceed by induction on s. At least two of the
ci are non-zero, and without loss of generality we can assume c1 ≥ c2 > 0.
Clearly G = hx1 , x1 + x2 , x3 , . . . , xk i and gcd(c1 − c2 , c2 , c3 , . . . , ck ) = 1. Now
by induction there exist y1 , . . . , yk ∈ G such that G = hy1 , . . . , yk i and
y1 = (c1 − c2 )x1 + c2 (x1 + x2 ) + c3 x3 + · · · + ck xk
= c1 x 1 + c2 x 2 + · · · + ck x k ,
which proves the lemma.
Theorem 5.15. Let G be a finitely generated abelian group, and suppose
that G is generated by k elements. Then G is isomorphic to a direct product
of ≤ k cyclic groups.
Proof. Let k ≥ 1 be the smallest possible size of a generating set for G. If
k = 1, then G is cyclic and there is nothing to prove; assume k > 1 in what
follows.
Choose generators G = hx1 , x2 , . . . , xk i such that |x1 | is as small as pos-
sible. We will show that G is the direct product of hx1 i and hx2 , . . . , xk i. If
this is not the case, then by Lemma 5.6 their intersection is not trivial; thus
there exist a1 , . . ., ak ∈ Z such that a1 x1 6= 0 and
a1 x1 + a2 x2 + · · · + ak xk = 0.
We can assume that 0 < a1 < |x1 |. Moreover, by replacing some of the xi
with −xi if necessary, we can assume a2 , . . . , ak ≥ 0 without loss of generality.
Let d = gcd(a1 , a2 , . . . , ak ). Since gcd(a1 /d, a2 /d, . . . , ak /d) = 1, by
Lemma 5.14 we can find y1 , . . . , yk ∈ G such that G = hy1 , . . . , yk i and
a1 a2 ak
y1 = x1 + x2 + · · · + xk .
d d d
112
But then we have dy1 = a1 x1 + a2 x2 + · · · + ak xk = 0, and thus
|y1 | ≤ d ≤ a1 < |x1 |,
which contradicts the minimality of |x1 |.

Therefore G is the direct product of hx1 i and hx2 , . . . , xk i. The theorem
follows by applying induction on hx2 , . . . , xk i.
Remark 5.16. The assumption that G is finitely generated is essential in

Theorem 5.15. For example, an exercise shows that G = (Q, +) is not a
direct product of cyclic groups.
Definition 5.17. Let G be a group. We will use the following notation:

G0 = {1} (trivial group), G1 = G, and for n > 1 we denote Gn = G × · · · × G
(n times). Clearly Gn × Gm ∼ = Gn+m for all integers n, m ≥ 0.
We have seen that any finitely generated abelian group is isomorphic to
a direct product Cn1 × · · · × Cnk × Zn for some integers n1 , . . ., nk > 1 and
n ≥ 0. This decomposition is not unique, as is seen by the following lemma.
Lemma 5.18. Let m, n > 0 be such that gcd(m, n) = 1. Then Cmn ∼

=
Cm × Cn .
Proof. Clearly Cm × Cn contains elements x and y with |x| = m and |y| = n.

Since xy = yx, by Lemma 1.29 we have |xy| = mn, so Cm × Cn is generated
by xy.
For example, we have C15 ∼= C3 ×C5 by Lemma 5.18, so the decomposition
of C15 into a direct product of cyclic groups is not unique. However, if we
assume that the integers ni in the decomposition Cn1 × · · · × Cnk × Zn are
prime powers, we do get the following uniqueness result.
Theorem 5.19. Let G be a finitely generated abelian group. Then:
(i) G ∼
= Cn1 × · · · × Cnk × Zr for some k ≥ 0, prime powers n1 , . . ., nk > 1,
and r ≥ 0.
(ii) The integers n1 , . . ., nk , r are unique19 .

19
To be more precise, this means that if G ∼ = Cm1 × · · · × Cm` × Zs for some prime
powers m1 , . . ., m` > 1 and s ≥ 0; then s = r, k = `, and (m1 , . . . , mk ) is a permutation
of (n1 , . . . , nk ).
113
Proof. For an integer n > 1 with prime factorization n = pα1 1 · · · pαt t , we have
Cn ∼
= Cpα1 1 × · · · × Cpαt t
by Lemma 5.18. This together with Theorem 5.15 proves (i).
Thus we can assume G ∼
= Cn1 ×· · ·×Cnk ×Zr where k ≥ 0, n1 , . . . , nk > 1
are prime powers, and r ≥ 0. An exercise shows that the set
tor(G) = {g ∈ G : |g| < ∞}
of elements of finite order is a subgroup of G, and moreover
tor(G) ∼
= Cn1 × · · · × Cnk ,
∼
G/ tor(G) = Zr .
Therefore if G ∼ = Cm1 × · · · × Cm` × Zs for some other prime powers
m1 , . . . , m` > 1 and s ≥ 0, we must have
Cm1 × · · · × Cm` ∼
= Cn1 × · · · × Cnk ,
Zs ∼
= Zr .
By an exercise Zr ∼= Zs if and only if r = s, so r is uniquely determined
t
by G. Next let p be a prime. Consider G(pt ) := {g ∈ G : g p = 1}, for t ≥ 1.
Then G(pt ) is a subgroup of G, and for all t ≥ 1 we have
G(pt )/G(pt−1 ) ∼ c
= (Cp ) t ,
where ct is the number of nj such that nj = pα for some α ≥ t (Exercise).
Then for t ≥ 1, we have that ct − ct+1 is the number of nj such that nj = pt .
On the other hand, by the same argument ct − ct+1 is the number of mj such
that mj = pt . We conclude then that k = ` and (m1 , . . . , mk ) is a permutation
of (n1 , . . . , nk ), so the integers n1 , . . ., nk are uniquely determined by G.
Example 5.20. By Theorem 5.19, we have a complete classification of all
finite abelian groups. As an illustration, in Table 4 we provide a list of all
finite abelian groups of order at most 12, up to isomorphism.
Example 5.21. In general, what is the number of finite abelian groups of

order n? Write the prime factorization of n as n = pk11 · · · pkt t . Exercise:
Use Theorem 5.19 to show that up to isomorphism, the number of finite
abelian groups of order n is p(k1 ) · · · p(kt ), where p(k) denotes the number
of partitions of an integer k > 0.
For example p(4) = 5, since the partitions of 4 are 4, 1 + 3, 2 + 2, 1 + 1 + 2,
and 1 + 1 + 1 + 1. And for any prime number p, there are a total of 5 abelian
groups of order p4 up to isomorphism:
Cp4 Cp × Cp3 Cp2 × Cp2 Cp × Cp × Cp2 Cp × Cp × Cp × Cp
114
n Abelian groups of order n

1 C1
2 C2
3 C3
4 C4 , C2 × C2
5 C5
6 C2 × C3
7 C7
8 C8 , C4 × C2 , C2 × C2 × C2
9 C9 , C3 × C3
10 C2 × C5
11 C11
12 C4 × C3 , C2 × C2 × C3
Table 4: Abelian groups of order 1 ≤ n ≤ 12.
Lemma 5.22. Let G be a finite abelian group of order n > 0. Let n =

pα1 1 · · · pαt t be the prime factorization of n. Then the following hold.
(i) G contains a unique subgroup Pk of order pαk k for all 1 ≤ k ≤ t.
(ii) G is the direct product of P1 , . . ., Pt .

Proof. (i) By Theorem 5.19, we can assume that
G = (Cpm(1,1) × · · · × Cpm(1,r1 ) ) × · · · × (Cpm(t,1) × · · · × Cpm(t,rt ) )

1 1 t t
Prk
for some integers m(i, j) > 0, where j=1 m(k, j) = αk for all 1 ≤ k ≤
t. Define
Pk = Cpm(k,1) × · · · × Cpm(k,rk )
k k
for 1 ≤ k ≤ t. Then Pk is a subgroup of order pαk in G, and G =

P1 × · · · × Pt . Since Pi for i 6= k have order coprime to pk , we have
αk
Pk = {x ∈ G : xp = 1}.
In other words, Pk is the set of elements with order dividing pαk , and
thus it is the only subgroup of order pαk k in G.
(ii) Clear from the description of Pk in the proof of (i).
Example 5.23. Let G be a finite abelian p-group, say G ∼

= C p α1 × · · · × C p αt
with α1 ≥ · · · ≥ αt > 0.
115
(a) Exercise: For H ≤ G, we have H ∼

= Cpβ1 × · · · × Cpβt for some β1 ≥ · · · ≥
βt ≥ 0.
(b) Exercise: In (a), we have βi ≤ αi for all 1 ≤ i ≤ t.
(c) Exercise: Prove that results analogous to (a) and (b) hold for the quotient
G/H.
(d) Exercise: Let H ≤ G. Show that there exists a subgroup of G that is

isomorphic to G/H.
(e) Exercise: Show that Q8 does not have a subgroup isomorphic to Q8 /Z(Q8 ).
5.3 Automorphisms
Let G be a group. Recall (Example 1.17) that an automorphism of G is an
isomorphism ϕ : G → G. We will denote the set of all automorphisms of G
by Aut(G). It is easily seen that if ϕ, ψ ∈ Aut(G); then so is ϕψ ∈ Aut(G)
and ϕ−1 ∈ Aut(G). Moreover, the identity map I : G → G defined by
I(g) = g for all g ∈ G is always an automorphism of G.
Thus we conclude that Aut(G) is a subgroup of Sym(G). We call the
group Aut(G) the automorphism group of G.
Remark 5.24. Exercise: Show that if G ∼

= H, then Aut(G) ∼
= Aut(H).
Recall (Lemma 1.65) also that each element g ∈ G defines an inner
automorphism γg : G → G by γg (x) = gxg −1 for all x ∈ G. Note that
γ1 is the identity map, and moreover γgh = γg γh and γg−1 = γg−1 for all
g, h ∈ G. Therefore the set of all inner automorphisms forms a subgroup
of Aut(G). We denote Inn(G) = {γg : g ∈ G} and call Inn(G) the inner
automorphism group of G. The quotient Aut(G)/ Inn(G) is called the outer
automorphism group of G and denoted by Out(G).
Lemma 5.25. Let G be a group. Then the following hold:
(i) Inn(G) is a normal subgroup of Aut(G);
(ii) Inn(G) ∼
= G/Z(G).
Proof. (i) Let ϕ ∈ Aut(G). A calculation shows that ϕγg ϕ−1 = γϕ(g) for
all g ∈ G, thus Inn(G) E Aut(G).
116
(ii) We define a map ψ : G → Inn(G) by ψ(g) = γg . Since γgh = γg γh for

all g, h ∈ G, it is clear that ψ is a surjective homomorphism. We have
ψ(g) = 1 if and only if gxg −1 = x for all x ∈ G; equivalently g ∈ Z(G).
Thus Ker ψ = Z(G), so G/Z(G) ∼ = Inn(G).
Remark 5.26. The following observation is easy but often useful. Suppose
that G is generated by some set S. Then any automorphism ϕ is determined
by its values on S. That is, if ϕ, ϕ0 ∈ Aut(G) are such that ϕ(x) = ϕ0 (x) for
all x ∈ S, then ϕ = ϕ0 . This is easily seen for example with Lemma 1.39.
We now calculate automorphism groups for some examples of groups.
Lemma 5.27. Let G = hgi be cyclic of order n > 0.
(i) For k ∈ Z, define the map ψk : G → G by x 7→ xk . Then Aut(G) =

{ψk : gcd(k, n) = 1}.
(ii) Aut(G) ∼
= (Z/nZ)× .
Proof. (i) It is clear that ψk is a homomorphism for all k ∈ Z. If gcd(k, n) =
1, then by Lemma 1.46 (iv) we have hgi = hg k i, so ψk is surjective and
thus a bijection since G is finite. Therefore ψk ∈ Aut(G) for all k ∈ Z
with gcd(k, n) = 1.
Conversely, let ϕ ∈ Aut(G). Then G = hϕ(g)i, so by Lemma 1.46 (iv)
we have ϕ(g) = g k for some k ∈ Z such that gcd(k, n) = 1. Then it
is clear that ϕ(x) = xk for all x ∈ G, so ϕ = ψk . This completes the
proof of (i).
0
(ii) We have ψk = ψk0 if and only if g k = g k , which is equivalent to k ≡ k 0
mod n. Thus we have an injective map defined by
ψ : (Z/nZ)× → Aut(G), ψ(k) = ψk for all k ∈ Z with gcd(k, n) = 1.
It is clear ψ is also surjective, so ψ is a bijection. For all k, k 0 ∈ Z

we have ψk ψk0 = ψkk0 , so we conclude that ψ is an isomorphism and
Aut(G) ∼ = (Z/nZ)× .
Lemma 5.28. Let G = hgi be infinite cyclic. Then Aut(G) ∼

= C2 .
Proof. Similar to Lemma 5.27. Since the only generators of G are g and g −1
(Lemma 1.48), it is easy to see that the only nontrivial automorphism of G
is the inverse map x 7→ x−1 .
117
Lemma 5.29. Aut(S3 ) ∼

= S3 .
Proof. Let G = S3 . Consider ϕ ∈ Aut(G). We have S3 = h(12), (123)i, so ϕ
is determined by the images of (12) and (123). Since an automorphism must
map an element to another element with the same order (Lemma 1.63 (vii)),
we have
ϕ((12)) ∈ {(12), (13), (23)},

ϕ((123)) ∈ {(123), (132)}.
Thus | Aut(G)| ≤ 3 · 2 = 6. On the other hand, by Lemma 5.25 we have

| Inn(G)| = |G/Z(G)| = 6 since Z(G) = {1}. Therefore Aut(G) = Inn(G) ∼
=
G/Z(G) ∼ = G.
Lemma 5.30. Let G and H be finite groups such that gcd(|G|, |H|) = 1.
Then Aut(G × H) ∼
= Aut(G) × Aut(H).
Proof. Exercise.
Example 5.31. (a) The automorphism group of a cyclic group is not ne-
= (Z/8Z)× ∼
cessarily cyclic; for example Aut(C8 ) ∼ = C2 × C2 .
(b) The automorphism group of an abelian group is not necessarily abelian.
Exercise: Aut(C2 × C2 ) ∼
= S3 .
(c) Exercise: Aut(S4 ) ∼
= S4 .
Theorem 5.32. Let G be a group and let H ≤ G. For g ∈ NG (H), define
fg : H → H by fg (x) = gxg −1 for all x ∈ H.
(i) fg is an automorphism of H for all g ∈ NG (H).
(ii) The map NG (H) → Aut(H) defined by g 7→ fg is a homomorphism with

kernel CG (H). In particular CG (H) E NG (H), and NG (H)/CG (H) is
isomorphic to a subgroup of Aut(H).
Proof. (i) By Lemma 1.65, each fg is an injective homomorphism. For

g ∈ NG (H) we have gHg −1 = H, so fg is also surjective. Thus fg is an
automorphism of H for all g ∈ NG (H).
(ii) Similarly to Lemma 5.25 (ii).
Theorem 5.32 is sometimes called the “N/C-theorem” (normalizer-centralizer

theorem).
118
5.4 Elementary abelian p-groups

Let p be a prime, and G be a finite p-group of order |G| = pn . We say that
G is elementary abelian, if G ∼
= Cp × Cp × · · · × Cp (n times).
Lemma 5.33. Let G be a finite group. Then G is an elementary abelian

p-group if and only if G is abelian and xp = 1 for all x ∈ G.
Proof. Exercise.
Suppose now that (G, +) is an elementary abelian p-group, with additive
notation. The (G, +) becomes a vector space over the field Fp = Z/pZ as
follows. Let λ ∈ Fp , where λ ∈ Z. We define scalar multiplication by λ on G
by
λ · g = λg
for all g ∈ G. Since pg = 0 for all g ∈ G, this operation is well defined.
An exercise shows this makes (G, +) into a vector space over Fp . Since G
is finite, we have |G| = pn , where n is the dimension of G as an Fp -vector
space.
Let g1 , . . ., gn be a basis for G has an Fp -vector space. Then each g ∈ G
can be expressed uniquely in the form
g = α1 g1 + · · · + αn gn
with αi ∈ Fp . Thus G = hg1 , . . . , gn i and G is the direct product of hg1 i, . . .,

hgn i.
If ϕ : G → G is a homomorphism, then ϕ(x + y) = ϕ(x) + ϕ(y) for all
x, y ∈ G. Moreover, we have ϕ(kx) = kϕ(x) for all x ∈ G and k ∈ Z, so
in fact ϕ is also a Fp -linear map. In particular, this shows us that Aut(G)
is precisely the set of Fp -linear bijections G → G, in other words Aut(G) =
GL(G). We have proven the following result.
Lemma 5.34. Let G be an elementary abelian p-group of order pn . Then

Aut(G) ∼
= GLn (p).
We illustrate the identification of Aut(G) with GLn (p) in the case where
n = 2. Let G be an elementary abelian p-group of order p2 . Let g1 , g2 be a
basis of G. Consider ϕ ∈ Aut(G). Then
ϕ(g1 ) = g1a g2c

ϕ(g2 ) = g1b g2d
119
for some a, b, c, d ∈ Z. With the isomorphism Aut(G) → GL2 (p) (correspon-

ding to the basis g1 , g2 ), the automorphism ϕ corresponds to the matrix

a b
c d
in GL2 (p).
For example, we have Aut(C2 ×C2 ) ∼
= GL2 (2) and Aut(C3 ×C3 ) ∼
= GL2 (3)
by Lemma 5.34.
5.5 Semidirect products

By Lemma 5.6, a group G is the direct product of two subgroups H and K
precisely when the following two conditions hold:
(a) G = HK and H ∩ K = {1}. Equivalently, every g factors as g = hk for

unique h ∈ H and k ∈ K.
(b) Both H and K are normal in G.
If we assume that (a) holds (so G = HK with unique factorization), but

only assume that H is a normal subgroup, we say that G is a semidirect
product of H and K.
Example 5.35. (a) The direct product of any two groups is also a semidi-
rect product.
(b) For G = S3 , consider H = h(1 2 3)i and K = h(1 2)i. Then G = HK,
H ∩ K = {(1)}, and H is a normal subgroup. Thus G is the semidirect
product of H and K, and in this case K is not a normal subgroup.
(c) More generally, G = Sn is the semidirect product of H = An and K =

h(1 2)i.
(d) G = S4 is the semidirect product of
V = {(1), (1 4)(2 3), (1 3)(2 4), (1 2)(3 4)}
and S3 .
(e) Any dihedral group G is a semidirect product of cyclic subgroups H = hyi

and K = hxi, where |x| = 2, xyx−1 = y −1 , and x 6∈ hyi. (See Lemma
1.68 and Lemma 1.69.)
120
(f) The quaternion group G = Q8 is not the semidirect product of two

nontrivial subgroups H and K, since H ∩ K contains Z(Q8 ) = {1, −1}.
Given two groups H and K, up to isomorphism there is only one group
which is the direct product of H and K. But for semidirect products there can
be several possibilities. We know from Example 5.35 (b) that S3 is isomorphic
to a semidirect product of C2 and C3 . But the direct product C2 × C3 is also
a semidirect product of C2 and C3 , and certainly C2 × C3 ∼ 6= S3 .
So how can we construct all possible semidirect products of two groups?
There is a natural way to do this, which is motivated by looking at the group
operation in a semidirect product more closely.
Consider a group G with subgroups H and K such that the following
hold:
• G = HK,
• H ∩ K = {1},
• H E G.
Given two elements g = hk and g 0 = h0 k 0 with h, h0 ∈ H and k, k 0 ∈ K,
their product should also be of the form gg 0 = h00 k 00 for unique h00 ∈ H and
k 00 ∈ K. We can determine h00 and k 00 as follows. We have
gg 0 = hkh0 k 0 = (hkh0 k −1 )kk 0 (5.1)
and here kh0 k −1 ∈ H since H is a normal subgroup. Therefore h00 = hkh0 k −1
and k 00 = kk 0 . In other words, h00 = hfk (h0 ) and k 00 = kk 0 where fk is an
automorphism of H as in Theorem 5.32. Moreover by Theorem 5.32, the
map ψ : K → Aut(H) defined by ψ(k) = fk is a homomorphism.
So any semidirect product G = HK with H E G comes with a homo-
morphism ψ : K → Aut(H) which determines group operation of G: we
have
(hk)(h0 k 0 ) = (hψ(k)(h0 ))(kk 0 )
for all h, h0 ∈ H and k, k 0 ∈ K. This motivates the definition of an external
semidirect product of H and K, which we define as follows.
Let H and K be groups and let ψ : K → Aut(H) be a homomorphism.
The external semidirect product H oψ K of H and K (corresponding to the
homomorphism ψ) is the cartesian product
H oψ K := H × K
equipped with the following group operation:
(h, k) · (h0 , k 0 ) = (hψ(k)(h0 ), kk 0 )
121
for all h, h0 ∈ H and k, k 0 ∈ K.

We will next verify the the external semidirect product is indeed a group,
and that it is the semidirect product of H and K with the various properties
that we expect from it.
Theorem 5.36. Let H and K be groups, and let ψ : K → Aut(H) be a

b = {(h, 1) : h ∈ H}
homomorphism. Then Hoψ K is a group. Moreover for H
b = {(1, k) : k ∈ K} the following hold:
and K
b is a normal subgroup of H oψ K, and H ∼
(i) H =Hb with an isomorphism
given by h 7→ (h, 1).
b is a subgroup of H oψ K, and K ∼
(ii) K =Kb with an isomorphism given by
k 7→ (1, k).
(iii) H oψ K = H
bK b ∩K
b and H b = {1}.
b = {(h, 1) : h ∈ H}
In particular H oψ K is the semidirect product of H
and Kb = {(1, k) : k ∈ K}.
Proof. We begin by showing that the binary operation on H oψ K is associ-

ative. Let h, h0 , h00 ∈ H and k, k 0 , k 00 ∈ K. Then
(h, k) · ((h0 , k 0 ) · (h00 , k 00 )) = (h, k) · (h0 · ψ(k 0 )(h00 ), k 0 k 00 )

= (h · ψ(k)(h0 · ψ(k 0 )(h00 )), k(k 0 k 00 ))
= (h · ψ(k)(h0 ) · (ψ(k) ◦ ψ(k 0 ))(h00 ), k(k 0 k 00 )) (5.2)
= (h · ψ(k)(h0 ) · ψ(kk 0 )(h00 ), k(k 0 k 00 )) (5.3)
where (5.2) holds since ψ(k) is a homomorphism, and (5.3) holds since ψ is
a homomorphism. Similarly
((h, k) · (h0 , k 0 )) · (h00 , k 00 ) = (h · ψ(k)(h0 ), kk 0 ) · (h00 , k 00 )

= (h · ψ(k)(h0 ) · ψ(kk 0 )(h00 ), kk 0 (k 00 )).
Therefore (h, k)·((h0 , k 0 )·(h00 , k 00 )) = ((h, k)·(h0 , k 0 ))·(h00 , k 00 ) for all h, h0 , h00 ∈
H and k, k 0 , k 00 ∈ K.
For all h ∈ H and k ∈ K we have (h, k) · (1, 1) = (h · ψ(k)(1), k) = (h, k)
and (1, 1) · (h, k) = (1 · ψ(1)(h), k) = (h, k); therefore (1, 1) is the identity
element.
For inverse elements, the correct element is suggested by the formula (5.1).
Indeed, we have (hk)−1 = k −1 h−1 = h00 k 00 with h00 = k −1 h−1 k and k 00 = k −1 .
122
This suggests that (ψ(k −1 )(h−1 ), k −1 ) should be the inverse of (h, k). We
check that this is indeed the case. Now
(h, k) · (ψ(k −1 )(h−1 ), k −1 ) = (h · ψ(k)(ψ(k −1 )(h−1 ), kk −1 )

= (h · ψ(kk −1 )(h−1 ), 1)
= (h · h−1 , 1) = (1, 1)
and similarly
(ψ(k −1 )(h−1 ), k −1 ) · (h, k) = (ψ(k −1 )(h−1 ) · ψ(k −1 )(h), k −1 k)

= (ψ(k −1 )(h−1 h), 1)
= (1, 1).
Therefore (h, k) is invertible with inverse (ψ(k −1 )(h−1 ), k −1 ). We conclude

that H oψ K is a group.
Since (h, k) 7→ k is a surjective homomorphism H oψ K → K with kernel
equal to H,
b we conclude that H b is a normal subgroup. It is easily seen that
the map h 7→ (h, 1) is an isomorphism H → H, b so H ∼=H b and claim (i) holds.
A similar calculation shows that k 7→ (k, 1) is an injective homomorphism
with image equal to K, b so (ii) holds. For (iii), we have H oψ K = H bKb
since (h, k) = (h, 1) · (1, k) for all h ∈ H and k ∈ K. It is obvious that
Hb ∩Kb = {(1, 1)}. This completes the proof of (i) – (iii) and the theorem.
Remark 5.37. Note that if ψ : K → Aut(H) is the trivial homomorphism

defined by ψ(k)(h) = h for all k ∈ K and h ∈ H, then H oψ K = H × K.
Lemma 5.38. Let G = H oψ K be an external semidirect product, where
ψ : K → Aut(H) is a homomorphism. Let H b and K
b be as in Theorem 5.36.
Then the following statements are equivalent:
(i) G = H × K.
(ii) ψ : K → Aut(H) is trivial.
(iii) K
b is a normal subgroup of G.
Proof. Exercise.
Let H oψ K be an external semidirect product of H and K. Then it is
clear that
(h, 1) · (h0 , 1) = (hh0 , 1),

(1, k) · (1, k 0 ) = (1, kk 0 ),
123
(h, 1) · (1, k) = (h, k),

(1, k) · (h, 1) · (1, k)−1 = (ψ(k)(h), 1),
for all h, h0 ∈ H and k, k 0 ∈ K. Thus in practice when working with an

external semidirect product H oψ K, we can denote (h, k) by hk, (h, 1) by
h, and (1, k) by k for all h ∈ H and k ∈ K.
Example 5.39. Using Theorem 5.36 we can construct the dihedral groups
as follows. Let H = hyi be cyclic, and let K = hxi be cyclic of order 2. We
have a homomorphism ψ : K → Aut(H), where ψ(x) : H → H is the inverse
map h 7→ h−1 on H. Let G = H oψ K. Then in G we have the following
(writing y for (y, 1) and x for (1, x)):
|x| = 2, xyx−1 = y −1 , x 6∈ hyi.
Therefore G is dihedral (Lemma 1.68). If H is finite cyclic of order n, then

G∼= D2n , and if H is infinite cyclic, then G ∼
= D∞ .
Theorem 5.40. Suppose that G = HK, where H E G, K ≤ G, and H∩K =
{1}. Then the following hold:
(i) G ∼
= H oψ K, where ψ : K → Aut(H) is the homomorphism defined by
ψ(k)(h) = khk −1 for all h ∈ H and k ∈ K.
(ii) Suppose that H ∼ = H 0 and K ∼

= K 0 . Then any external semidirect
product H oψ K is isomorphic to H 0 oψ0 K 0 for some homomorphism
ψ 0 : K 0 → Aut(H 0 ).
Proof. (i) We define ϕ : H oψ K → G by ϕ(h, k) = hk for all h ∈ H and
k ∈ K. As in (5.1), for all h, h0 ∈ H and k, k 0 ∈ K we have
ϕ(h, k)ϕ(h0 , k 0 ) = (hk)(h0 k 0 ) = (h · ψ(k)(h))(kk 0 ) = ϕ(h · ψ(k)(h), kk 0 ),
so ϕ is a homomorphism. Moreover ϕ is a bijection since G = HK and

H ∩ K = {1}, so ϕ is an isomorphism and G ∼ = H oψ K.
(ii) Exercise.
By Theorem 5.36 and Theorem 5.40, we have at least on some level a

systematic way of studying semidirect products of two groups H and K up
to isomorphism. Theorem 5.40 tells us that all possible semidirect products
of two groups H and K can be constructed as an external semidirect product.
However, there are many examples where H oψ K ∼ = H oψ0 K for ψ 6= ψ 0 .
.
124
Example 5.41. Let H be a non-abelian group. Consider the direct product

G = H × H. Then G is the direct product of H × {1} and {1} × H, so
G = H oψ H with ψ as the trivial map.
Now consider the diagonal subgroup
K = {(h, h) : h ∈ H}.
It is clear that K ≤ G and that K ∼ = H. Furthermore, we have that G is the
semidirect product of H × {1} and K, since (x, y) = (xy −1 , 1)(y, y) for all
x, y ∈ H. By Theorem 5.40, we have G ∼ = H oψ0 H, where ψ 0 : H → Aut(H)
is defined by ψ 0 (x)(y) = xyx−1 for all x, y ∈ H. Therefore
H × H = H oψ H ∼
= H oψ0 H,
and here ψ 0 6= ψ since H is non-abelian.
Example 5.42. Any group G is a semidirect product in a trivial way, since
G∼ = G oψ {1} ∼ = {1} oψ0 G. But there are many examples of groups which
can be written non-trivially as a semidirect product in many different ways.
That is, we can have H oψ K ∼ = H 0 oψ0 K 0 with H, K, H 0 , K 0 non-trivial and
pairwise non-isomorphic.
Consider the dihedral group G = D12 of order 12, with generators G =
hx, yi such that |x| = 2, |y| = 6, xyx−1 = y −1 . Then G is the semidirect
product of H = hyi and K = hxi, so G ∼ = C6 oψ C2 . On the other hand,
an exercise shows that G is also the semidirect product of H 0 = hy 2 i and
K 0 = hx, y 3 i, so G ∼
= C3 oψ0 (C2 × C2 ).
Lemma 5.43. Let H and K be groups and let ψ, ψ 0 : K → Aut(H) be
homomorphisms. Then the following hold:
(i) Suppose that ψ 0 = ψϕ for some ϕ ∈ Aut(K). Then H oψ K ∼
= H oψ0 K.
(ii) Suppose that there exists φ ∈ Aut(H) such that ψ 0 (x) = φψ(x)φ−1 for
all x ∈ K. Then H oψ K ∼ = H oψ0 K.
Proof. (i) Exercise: check that the map H oψ K → H oψ0 K defined by
(h, k) 7→ (h, ϕ−1 (k)) is an isomorphism.
(ii) Exercise: check that the map H oψ K → H oψ0 K defined by (h, k) 7→
(φ(h), k) is an isomorphism.
Remark 5.44. Besides direct products and semidirect products, there are
various other generalizations that have been studied in group theory. Suppose
that G is a group and G = HK for some subgroups H and K. Here are some
notions that appear in the literature.
125
(a) Factorizations: Without any further assumptions, we say that G =

HK is a factorization of G. Even this level of generality has been stu-
died. Factorizations of finite non-abelian almost simple20 groups has
been studied extensively, and is still an active area of research21 . Facto-
rizations of almost simple groups come up very naturally in permutation
group theory, and have also found applications in the study of Cayley
graphs.
(b) Zappa-Szép products: If H ∩ K = {1}, we say that the factorization

G = HK is exact. In this case G is called the Zappa-Szép product of
H and K. One has the notion of an external Zappa-Szép product, for
which the analogue of Theorem 5.40 holds. A result of Douglas22 gives
a complete classification of Zappa-Szép products of finite cyclic groups
H∼ = Cm and K ∼ = Cn .
(c) Central products: If hk = kh for all h ∈ H and k ∈ K, we say that

G is the central product of H and K. In this case H and K are normal
subgroups, H ∩ K ≤ Z(G), and we are not necessarily assuming that
H ∩ K is trivial. (If H ∩ K = {1}, we get the direct product of H and
K.)
There is a notion of an external central product and an analogue of The-
orem 5.40, which allows us to construct all possible central products
of H and K. The construction requires two subgroups H0 ≤ Z(H),
K0 ≤ Z(K), and an isomorphism ψ : H0 → K0 .
One defines an external central product as
H ◦ψ K = (H × K)/N,
where N is the normal subgroup N = {(h, k) ∈ H0 × K0 : ψ(h) = k −1 }

of H × K.
Let H b be the images of H×{1} and {1}×K in H◦ψ K, respectively.
b and K
Then the following hold:
• H∼
=Hb
20
A non-abelian simple group G has trivial center, so G ∼ = Inn(G) and we can identify
G as a normal subgroup of Aut(G). A finite group X is almost simple if G ≤ X ≤ Aut(G)
for some non-abelian finite simple group G.
21
See for example “T. C. Burness, C. H. Li, On solvable factors of almost simple groups.,
Adv. Math. 377 (2021)”
22
J. Douglas, On finite groups with two independent generators. I–IV., Proc. Nat.
Acad. Sci. U.S.A. 37 (1951).
126
• K∼
=Kb
• H ◦ψ K is the central product of H
b and K,
b with H b∼
b ∩K = H0 .
Given a central product G = HK, one has G ∼ = H ◦ψ K with H0 =

H ∩ K = K0 and ψ the identity map; thus the analogue of Theorem 5.40
holds23 .
5.6 Application: Groups of order pq (p, q primes)

Let p and q be distinct primes. What are the groups of order pq?
Lemma 5.45. Let G be a group such that |G| = pq, where p > q are primes.
Then G ∼
= Cp oψ Cq for some ψ : Cq → Aut(Cp ).
Proof. By Cauchy’s theorem, there exist subgroups H ≤ G and K ≤ G
with |H| = p and |K| = q. We will show that H E G. If this is not the
case, then H 6= g −1 Hg for some g ∈ G. But then H ∩ g −1 Hg = {1}, so
|Hg −1 Hg| = p2 > |G| by Lemma 1.88, a contradiction. Now the lemma
follows from Theorem 5.40.
Proposition 5.46. Let G be a group such that |G| = pq, where p > q are
primes. If q - p − 1, then G is cyclic.
Proof. We have Aut(Cp ) ∼ = (Z/pZ)× ∼= Cp−1 by Lemma 5.27 and Theorem

1.52. Thus if q - p − 1, then every homomorphism ψ : Cq → Aut(Cp ) is
trivial, and consequently Cp oψ Cq = Cp × Cq ∼
= Cpq .
Proposition 5.47. Let G be a group such that |G| = pq, where p > q are
primes. Suppose that q | p − 1. Then the following hold:
(i) There exists a non-trivial homomorphism ψ : Cq → Aut(Cp ), so that

Xp,q = Cp oψ Cq is non-abelian.
(ii) Any group of order pq is isomorphic to Cpq or Xp,q .
Proof. (i) Let Cq = hxi. Since q | p − 1, there exists an element σ order

q in Aut(Cp ) ∼
= Cp−1 (Lemma 5.27 and Theorem 1.52). Thus there
exists a homomorphism ψ : Cq → Aut(Cp ) such that ψ(x) = σ. Then
Xp,q := Cp oψ Cq is non-abelian, since ψ is nontrivial.
23
For details, see for example M. Suzuki’s book “Group Theory I”, Proposition 4.17,
p. 138.
127
(ii) By Lemma 5.45, any group of order pq is isomorphic to Cp oψ0 Cq for

some homomorphism ψ 0 : Cq → Aut(Cp ).
If ψ 0 is trivial, then Cp oψ0 Cq = Cp × Cq ∼
= Cpq .
If ψ 0 is non-trivial, then ψ 0 (x) is an element of order q in Aut(Cp ).
Since Aut(Cp ) is cyclic, it has a unique subgroup of order q, and so
ψ 0 (x) = σ k = ψ(x)k for some k ∈ Z coprime to q. Thus ψ 0 = ψψk , where
ψk : Cq → Cq is the automorphism of Cq defined by ψk (x) = xk (Lemma
5.27 (i)). It follows then from Lemma 5.43 (i) that Cp oψ0 Cq ∼
= Cp oψ Cq .
With Proposition 5.46 and Proposition 5.47, we have completed the clas-
sification of groups of order pq, where p and q are distinct primes. We
summarize them in the following.
Theorem 5.48. Let p and q be primes and assume that p > q. Then the
following hold:
(a) If q - p − 1, there is a unique group of order pq up to isomorphism, the
cyclic group Cpq .
(b) If q | p − 1, there are two groups of order pq up to isomorphism, the
cyclic group Cpq and a non-trivial semidirect product Cp oψ Cq .
Remark 5.49. As seen in the proof of Proposition 5.47, if q | p − 1 and we
access to an automorphism of order q in Aut(Cp ), a non-trivial semidirect
product Cp oψ Cq can be described explicitly.
First note that due to the isomorphism Aut(Cp ) ∼ = (Z/pZ)× , finding an
automorphism of order q in Aut(Cp ) is equivalent to finding an integer t such
that t 6≡ 1 mod p and tq ≡ 1 mod p. Indeed, then the powering map g 7→ g t
on Cp defines an automorphism of order q.
Now writing Cp = hxi and Cq = hyi, we have xp = y q = 1, and yxy −1 =
xt . Moreover, these relations determine the group operation on Cp oψ Cq
completely.
Example 5.50. Some special cases:
(i) Every group of order 33 is cyclic.
(ii) There are two groups of order 21 up to isomorphism. The cyclic group
of order 21, and a non-trivial semidirect product G = C7 oψ C3 . We
have 2 6≡ 1 mod 7 and 23 ≡ 1 mod 7, so as in Remark 5.49, the map
x 7→ x2 defines an automorphism of order 3 for C7 . It follows that
there exist x, y ∈ G such that G = hx, yi, where |x| = 7, |y| = 3, and
yxy −1 = x2 .
128
(iii) As special case of Theorem 5.48, we recover Theorem 1.122. If p > 2 is

prime, then any group of order 2p is isomorphic to C2p or D2p .
5.7 Application: p-groups of order p3

Let p be a prime. There is only one group of order p up to isomorphism,
the cyclic group Cp of order p (Corollary 1.83). We have seen (Proposition
1.144) that every group of order p2 is abelian, and in this case by Theorem
5.19 we have the following result.
Theorem 5.51. Let p be a prime. There are only two groups of order p2 up
to isomorphism, the cyclic group Cp2 and the direct product Cp × Cp .
In this section, we will apply the results on semidirect products to classify

all groups of order p3 . By Theorem 5.19 there are exactly three abelian groups
of order p3 up to isomorphism: Cp3 , Cp × Cp2 , and Cp × Cp × Cp .
We will next see that there are two non-abelian groups of order p3 up to
isomorphism, so in total we have gnu(p3 ) = 5 groups of order p3 . Thus the
number of groups does not depend on p. However, we will have to deal with
the cases p = 2 and p > 2 separately, and we begin with the case p = 2.
Theorem 5.52. Let G be a non-abelian group of order 8. Then G ∼

= D8 or
∼
G = Q8 .
Proof. Note first that G must contain an element y ∈ G of order 4. For

otherwise x2 = 1 for all x ∈ G, which is only possible if G is abelian. We
now consider the elements in G \ hyi. Clearly any x ∈ G \ hyi must have
order 2 or order 4, and moreover G = hx, yi. Now hyi is a normal subgroup
of G since it has index 2, so xyx−1 ∈ hyi. Since |xyx−1 | = |y|, we have
xyx−1 ∈ {y, y −1 }. We cannot have xyx−1 = y since otherwise G = hx, yi
would be abelian, so xyx−1 = y −1 for all x ∈ G \ hyi.
Case 1: There exists x ∈ G \ hyi with |x| = 2.

In this case G = hx, yi with |x| = 2, |y| = 4, xyx−1 = y −1 , and x 6∈ hyi. Thus
G is dihedral by Lemma 1.68, and G ∼ = D8 by Theorem 1.71.
Case 2: For all x ∈ G \ hyi, we have |x| = 4.

In this case, any element of order 2 must be contained in hyi, so η = y 2 is the
unique element of order 2 in G. In particular η ∈ Z(G). Let x ∈ G\hyi. Then
|x| = 4, so x2 = η and xyx−1 = y −1 . Similarly for z = xy we have |z| = 4,
z 2 = η, and zyz −1 = y −1 . The elements 1, η, x±1 , y ±1 , z ±1 are distinct, so
G = {1, η, x±1 , y ±1 , z ±1 }.
129
Moreover x−1 = ηx, y −1 = ηy, and z −1 = ηz, so we conclude that
G = {1, η, x, ηx, y, ηy, z, ηy}.
One can check that the elements of G satisfy the following relations:
x2 = y 2 = z 2 = η
xy = z = η(yx)
yz = x = η(zy)
zx = y = η(xz)
ηg = gη for all g ∈ G
Which may remind you of the relations established for the quaternion
group Q8 = {±1, ±i, ±j, ±k}, with −1, i, j, k corresponding to η, x, y, z
respectively. At this point the theory of generators and relations would easily
allow us to prove that the bijection Q8 → G defined by
1 7→ 1 −1 7→ η
i 7→ x −i 7→ ηx
j 7→ y −j 7→ ηy
k 7→ z −k 7→ ηz
is an isomorphism. This could also be checked manually.

Alternatively, we can compute the left regular representation ϕ : G →
Sym(G) and calculate

1 η x ηx y ηy z ηz
ϕ(x) =
x ηx η 1 z ηz ηy y

1 η x ηx y ηy z ηz
ϕ(y) =
y ηy ηz z η 1 x ηx
Relabeling the elements of G as 1, 2, . . ., 8, we find that
G∼
= ϕ(G) = hϕ(x), ϕ(y)i ∼
= h(1 3 2 4)(5 7 6 8), (1 5 2 6)(3 8 4 7)i.
These arguments also apply for Q8 since it is a non-abelian group and all
elements of Q8 \ hii have order 4, so we conclude
Q8 ∼
= h(1 3 2 4)(5 7 6 8), (1 5 2 6)(3 8 4 7)i
and thus G ∼
= Q8 .
130
Lemma 5.53. Let p be a prime, and let G be a non-abelian group of order

p3 . Then |Z(G)| = p, G/Z(G) ∼
= Cp × Cp , and Z(G) = [G, G].
Proof. By Corollary 1.143 the center of G is nontrivial, so |Z(G)| = p or
|Z(G)| = p2 . Since G is non-abelian, G/Z(G) cannot be cyclic (Exercise
1.31), so we must have |Z(G)| = p. Thus G/Z(G) is a non-cyclic group of
order p2 , which implies G/Z(G) ∼
= Cp × Cp by Theorem 5.51. In particular
G/Z(G) is abelian, so [G, G] ≤ Z(G). Because [G, G] 6= 1 and |Z(G)| = p,
we conclude that Z(G) = [G, G].
Lemma 5.54. Let p > 2 be a prime, and let G be a non-abelian group of
order p3 . Then (xy)p = xp y p for all x, y ∈ G.
Proof. By Theorem 5.53 the commutator subgroup is central of order p, so
by Lemma 4.8 we have
p(p−1)
(xy)p = xp y p [y, x] 2
p(p−1)
for all x, y ∈ G. Since [G, G] has order p and since 2
is a multiple of p,
p(p−1)
p p p
we have [y, x] 2 = 1. Thus (xy) = x y for all x, y ∈ G.
Lemma 5.55. Let p > 2 be a prime, and let G be a non-abelian group of
order p3 . Then one of the following holds:
(i) G contains an element of order p2 , and G ∼
= Cp2 oψ Cp for some ψ :
Cp → Aut(Cp2 ).
(ii) xp = 1 for all x ∈ G, and G ∼
= (Cp × Cp ) oψ Cp for some ψ : Cp →
Aut(Cp × Cp ).
Proof. By Lemma 5.54, the map ϕ : G → G defined by ϕ(x) = xp for all
x ∈ G is a homomorphism. Since G/Z(G) ∼ = Cp × Cp by Lemma 5.53, we
have ϕ(G) ≤ Z(G). Now |Z(G)| = p, so by the first isomorphism theorem
either | Ker ϕ| = p2 or | Ker ϕ| = p3 . We consider the two possibilities.
Case 1: | Ker ϕ| = p2 .
In this case, there exists y ∈ G such that y p 6= 1. Then we must have
|y| = p2 , and H = hyi is a normal subgroup of G by Proposition 1.148. Now
H ∩ Ker ϕ = hy p i has order p, so there exists x ∈ Ker ϕ such that x 6∈ H.
Then for K = hxi we have H ∩ K = {1} and G = HK, so G is the semidirect
product of H and K. By Theorem 5.40, (i) holds.
Case 2: | Ker ϕ| = p3 .
In this case, we have xp = 1 for all x ∈ G. By Proposition 1.146 there exists a
131
normal subgroup H E G with |H| = p2 , and by Theorem 5.51 we must have

H ∼= Cp × Cp . Now choose y ∈ G \ H, and let K = hyi, so K ∼
= Cp . Then
H ∩ K = {1}, and G is the semidirect product of H and K. By Theorem
5.40, (ii) holds.
In view of Lemma 5.55, for p > 2 what remains is to classify the possible
semidirect products Cp2 oψ Cp and (Cp × Cp ) oψ Cp .
Proposition 5.56. Let p > 2 be a prime. There exists a unique non-abelian
semidirect product Cp2 oψ Cp up to isomorphism.
Proof. Let Cp = hxi and Cp2 = hyi. First we note that a semidirect product
Cp2 oψ Cp is non-abelian if and only if ψ : Cp → Aut(Cp2 ) is non-trivial,
equivalently ψ is injective.
By Theorem 1.55 and Lemma 5.27, we have Aut(Cp2 ) ∼ = Cp(p−1) . Thus
we can find a homomorphism ψ : Cp → Aut(Cp2 ) such that ψ(x) has order
p. Then Cp2 oψ Cp is non-abelian.
Consider any non-trivial map ψ 0 : Cp → Aut(Cp2 ). Then ψ 0 (x) is an
automorphism of order p. Since Aut(Cp2 ) is cyclic of order p(p − 1), it has
a unique subgroup of order p, and therefore ψ 0 (x) = ψ(x)k for some k ∈ Z
coprime to p. Thus ψ 0 = ψψk , where ψk ∈ Aut(Cp ) is the automorphism
defined by ψk (x) = xk (Lemma 5.27 (i)). By Lemma 5.43 (i), we conclude
that Cp2 oψ Cp ∼= Cp2 oψ0 Cp .
Proposition 5.57. Let p > 2 be a prime. There exists a unique non-abelian
semidirect product (Cp × Cp ) oψ Cp up to isomorphism.
Proof. Let Cp = hgi, and suppose that Cp × Cp= hx,yi. We can identify
a b
Aut(Cp × Cp ) = GL2 (p), where a matrix A = corresponds to the
c d
a c b d
automorphism defined
by x 7→ x y , y 7→ x y .
1 1
The matrix has order p in GL2 (p), so there exists a homomor-
0 1
1 1
phism ψ : Cp → Aut(Cp × Cp ) such that ψ(g) = .
0 1
1 1
By Example 3.19 any element of order p in GL2 (p) is conjugate to .
0 1
Thus if ψ 0 : Cp → Aut(Cp ×Cp ) is a non-trivial homomorphism, then there ex-
ists φ ∈ Aut(Cp ×Cp ) such that ψ 0 (g) = φψ(g)φ−1 . Then ψ 0 (g0 ) = φψ(g0 )φ−1
for all g0 ∈ Cp , so by Lemma 5.43 (ii), we have (Cp × Cp ) oψ Cp ∼ = (Cp ×
Cp ) oψ0 Cp .
Thus for any prime p, there are exactly two non-abelian groups G of order
p3 , up to isomorphism:
132
• If p = 2, we have G ∼
= D8 or G ∼
= Q8 .
• For p > 2, we have G ∼
= Cp2 oψ Cp or G ∼
= (Cp × Cp ) oψ0 Cp , where the
semidirect products are non-trivial.
For p > 2, we will realize the nontrivial semidirect products in the fol-
lowing examples. This amounts to finding automorphisms of order p in
Aut(Cp2 ) and Aut(Cp × Cp ).
Example 5.58. First note that Aut(Cp2 ) ∼ = (Z/p2 Z)× , with k ∈ (Z/p2 Z)×
corresponding to the automorphism ψk : g 7→ g k of Cp2 . We saw in the proof
of Theorem 1.55 that 1 + p has order p in (Z/p2 Z)× , so the map ψp+1 is an
automorphism of order p for Cp2 . Consider then Cp2 = hxi and Cp = hyi. We
have a homomorphism ψ : Cp → Aut(Cp2 ) defined by ψ(y) = ψ1+p . Then
Cp2 oψ Cp = hx, yi with |x| = p2 , |y| = p, and yxy −1 = xp+1 .
Moreover, the group operation in Cp2 oψ Cp is completely determined by
these relations.
Example 5.59. Let x, y be a basis of Cp × Cp , and let Cp = hzi be another
∼
cyclic group of order p.
As we have seen Section 5.4, we have Aut(Cp ×Cp ) =
a b
GL2 (p), with matrix corresponding to the automorphism of Cp × Cp
c d
a c b d 1 1
defined by x 7→ x y and y 7→ x y . The matrix ∈ GL2 (p) has order
0 1
p, so the map x 7→ x, y 7→ xy defines an automorphism ϕ : Cp ×Cp → Cp ×Cp
of order p.
Thus we have a homomorphism ψ 0 : Cp → Aut(Cp × Cp ) defined by
ψ(z) = ϕ. Then (Cp × Cp ) oψ0 Cp = hx, y, zi such that:


 |x| = |y| = |z| = p

xy = yx


 hx, yi ∼
= Cp × Cp
zxz = x and zyz −1 = xy
−1

The group operation in (Cp × Cp ) oψ Cp is completely determined by these

relations.
Alternatively, a nontrivial semidirect product (Cp × Cp ) oψ Cp can be
realized as the Heisenberg group
  
 1 a b 
Hp = 0 1 c  : a, b, c ∈ Fp .
0 0 1
 
133
As seen in Exercise 1.37, for p > 2 the group Hp is non-abelian of order p3 ,

and xp = 1 for all x ∈ Hp .
5.8 Application: Groups of order 2n (n odd)

Lemma 5.60. Let G be a finite group of order mn, where gcd(m, n) = 1.
Suppose that G contains a normal subgroup N of order n. Then N is the
unique subgroup of order n in G, and
N = {x ∈ G : xn = 1}.
Proof. Exercise.
Theorem 5.61. Let G be a finite group of order 2n, where n is odd. Then G
contains a unique subgroup N of order n. Moreover N E G and G ∼ = N oψ C2
for some homomorphism ψ : C2 → Aut(N ).
Proof. Consider the left regular representation ϕ : G → S2n of G. Since ϕ

is injective, we can identify G with ϕ(G) and assume that G ≤ S2n . For an
element g ∈ G, its orbits on the 2n points in G are precisely the right cosets
of hgi in G. Therefore any g ∈ G is a product of |G|/|g| disjoint cycles of
length |g|. Consider an element x ∈ G of order 2, which exists by Cauchy’s
theorem. Then
x = (a1 a01 )(a2 a02 ) · · · (an a0n )
is a product of n transpositions, so x is an odd permutation. Thus G is not
contained in A2n , which implies that N = G ∩ A2n is a normal subgroup of
order n in G (Lemma 2.29).
Uniqueness of N follows from Lemma 5.60, and G ∼ = N oψ C2 since G is
∼
the semidirect product of N and hxi = C2 .
Theorem 5.62. Let N be a finite group of odd order n, and let C2 = hxi.
Then N oψ C2 ∼
= N oψ0 C2 if and only if ψ(x) and ψ 0 (x) are conjugate in
Aut(N ).
Proof. If ψ(x) = ϕψ(x)ϕ−1 for some ϕ ∈ Aut(N ), it follows from Lemma

5.43 (ii) that N oψ C2 ∼
= N oψ0 C2 .
Conversely, suppose that there exists an isomorphism
π : N oψ C2 → N oψ0 C2 .
Since N is the unique subgroup of order n in both groups, we have π(N ) = N ,

and thus the restriction of π to N provides an automorphism ϕ : N → N .
134
For the generator x of C2 , we have π(x) = n0 x for some n0 ∈ N . Since π is

a homomorphism, we get
π(xnx−1 ) = π(x)π(n)π(x)−1 = n0 xϕ(n)x−1 n−1 0 −1
0 = n0 ψ (x)(ϕ(n))n0
for all n ∈ N . Since π(xnx−1 ) = ϕ(ψ(x)(n)) for all n ∈ N , we conclude that

ϕ ◦ ψ(x) = γn0 ◦ ψ 0 (x) ◦ ϕ,
where γn0 is the inner automorphism n 7→ n0 nn−1 0 .
Then ϕψ(x)ϕ−1 = γn0 ψ 0 (x), so it will suffice to prove that γn0 ψ 0 (x) and
ψ 0 (x) are conjugate in Aut(N ). To this end, we note that γn0 ψ 0 (x) and ψ 0 (x)
both have order 2, so by Theorem 1.71 they generate a subgroup which is
dihedral of order 2k, where k is the order of γn0 . Now k is odd since |N | is, so
all elements of order 2 are conjugate in D2k (Exercise 5.29), and in particular
γn0 ψ 0 (x) and ψ 0 (x) are conjugate.
Corollary 5.63. Let N be a finite group of odd order n. Let t be the number
of conjugacy classes of elements of order 2 in Aut(N ). Then up to isomor-
phism, there are t + 1 semidirect products of the form N oψ C2 .
Proof. Follows from Theorem 5.62. (We get one group from the trivial homo-
morphism ψ(x) = 1, which gives the direct product N × C2 . The remaining
t groups are given by ψ : C2 → Aut(N ) with ψ(x) of order 2.)
Example 5.64. Exercise: As an application of the results of this section,
classify the groups of order 30, up to isomorphism. (Similarly, groups of
order 66 and 70.)
Example 5.65. Suppose that p > 2 is a prime. We will classify groups of
order 2p2 . Any group N of order p2 is isomorphic to Cp2 or Cp × Cp . We
know the following:
• Aut(Cp2 ) ∼
= (Z/p2 Z)× ∼= Cp(p−1) , so there is a unique element of order
2 in Aut(Cp2 ). Thus by Corollary 5.63, there are 2 semidirect products
Cp2 oψ C2 , up to isomorphism.
• Aut(Cp × Cp ) ∼= GL2 (p), and by results of Section 3.5 there are two
conjugacy classes of elements of order 2 in GL2 (p), with representatives

−1 0 1 0
and .
0 −1 0 −1
Therefore by Corollary 5.63, there are 3 semidirect products of the form
(Cp × Cp ) oψ C2 , up to isomorphism.
By Theorem 5.61, we conclude that there are 2 + 3 = 5 groups of order
2
2p , up to isomorphism.
135
6 Sylow theory
In this section, we will prove the fundamental theorems of Sylow on finite
groups. Throughout this section, a group G will always be a finite group and
p will be a prime.
6.1 Double cosets

Let H and K be a subgroups of a group G. An (H, K)-double coset is a set
of the form HxK, where x ∈ G.
Lemma 6.1. Let H, K ≤ G be finite. Then
|H||K|
|HxK| =
|x−1 Hx ∩ K|
for all x ∈ G.
Proof. Since |HxK| = |x−1 HxK|, the claim follows from Lemma 1.88 and
the fact that |x−1 Hx| = |H|.
Lemma 6.2. Let H, K ≤ G. Then for any two (H, K)-double cosets HxK
and HyK, either HxK = HyK or HxK ∩ HyK = ∅.
Proof. If HxK ∩HyK is nonempty, there exists h, h0 ∈ H and k, k 0 ∈ K such
that hxk = h0 yk 0 . Then x = h−1 h0 yk 0 k −1 , so HxK = Hh−1 h0 yk 0 k −1 K =
HyK.
By Lemma 6.2, for H, K ≤ G the (H, K)-double cosets partition G, i.e.
G is a disjoint union [
G= HxK.
x∈G
Assuming that G is finite, there exist representatives for the double cosets,
that is, elements x1 , . . ., xt ∈ G such that
t
[
G= Hxi K
i=1
and Hxi K 6= Hxj K for i 6= j. Combining this with Lemma 6.1, we get the
following result.
Lemma 6.3. Let G be a finite group and H, K ≤ G. Let x1 , . . ., xt ∈ G be
representatives for the (H, K)-double cosets. Then
t
X |H||K|
|G| = .
i=1
|x−1
i Hxi ∩ K|
136
6.2 Sylow theorems

Theorem 6.4 (Sylow I). Let G be a finite group and let p be a prime. If pβ
divides |G|, then G contains a subgroup of order pβ .
Proof. We proceed by induction on the order of G, the case where |G| = 1
being obvious. Clearly we can assume that β > 0. The main idea is to use
the class equation (Proposition 1.119), which tells us that
t
X
|G| = |Z(G)| + [G : CG (xi )],
i=1
where x1 , . . ., xt are representatives for the non-central conjugacy classes of

G.
If p divides |Z(G)|, then by Cauchy’s theorem Z(G) contains a subgroup
N of order p. Now N is a normal subgroup of G since it is central, and
by induction G/N contains a subgroup H/N of order pβ−1 . Then H is a
subgroup of G with |H| = pβ . We may assume then that p does not divide
|Z(G)|. Since p divides |G|, it follows from the class equation that there
exists 1 ≤ i ≤ t such that [G : CG (xi )] is not divisible by p. We have
|G| = [G : CG (xi )] · |CG (xi )|, so we conclude that pβ must divide |CG (xi )|.
Now CG (xi ) is a proper subgroup since xi is not in the center, so by induction
CG (xi ) contains a subgroup of order pβ .
Let p be a prime and G a finite group. A p-subgroup of G is a subgroup
H ≤ G such that H is a p-group. A Sylow p-subgroup of G is a p-subgroup
H ≤ G with |H| = pα , where pα is the largest power of p that divides |G|.
Equivalently, a Sylow p-subgroup of G is a p-subgroup H ≤ G such that p
does not divide [G : H].
We will denote the number of Sylow p-subgroups in G by np (G). Note
that np (G) ≥ 1 by Sylow I.
Lemma 6.5. Let G be a finite group and let P be a Sylow p-subgroup of G.
The following statements hold:
(i) If P E G, then P contains every p-subgroup of G.
(ii) Let Q be a p-subgroup of G. Then Q ∩ NG (P ) = Q ∩ P .
(iii) P E G if and only if np (G) = 1.

Proof. (i) Suppose that P E G. Let x ∈ G be an element of p-power order.
The order of xP in G/P must also be a p-power. On the other hand,
by Lagrange’s theorem the order of xP divides [G : P ]. Since p does
137
not divide [G : P ], it follows that xP is the identity, and thus x ∈ P .

So P contains every element of p-power order, and therefore it contains
every p-subgroup of G.
(ii) Note that P is a Sylow p-subgroup of NG (P ) and P E NG (P ). So since
Q ∩ NG (P ) is a p-subgroup of NG (P ), we have Q ∩ NG (P ) ≤ P by
(i). Therefore Q ∩ NG (P ) ≤ Q ∩ P . The reverse inclusion follows from
P ≤ NG (P ), so Q ∩ NG (P ) = Q ∩ P .
(iii) Suppose that P E G. If Q is a Sylow p-subgroup of G, then Q ≤ P by
(i). Since |Q| = |P | it follows that Q = P , and so np (G) = 1.
For the other direction, suppose that np (G) = 1. Every conjugate of P
is a Sylow p-subgroup of G, so we must have g −1 P g = P for all g ∈ G.
In other words, the subgroup P is a normal subgroup of G.
Theorem 6.6. Let p be a prime and let P be a Sylow p-subgroup of G. Let

Q ≤ G be a p-subgroup of G. Then there exists g ∈ G such that Q ≤ g −1 P g.
Proof. Consider the decomposition of G into a disjoint union of (P, Q)-double
cosets:
[t
G= P xi Q
i=1
where x1 , . . ., xt ∈ G and the double cosets P xi Q are distinct. Then

t
X t
X
|G| = |P xi Q| = |P |[Q : x−1
i P xi ∩ Q]
i=1 i=1
by Lemma 6.1. Therefore

t
X
[G : P ] = [Q : x−1
i P xi ∩ Q].
i=1
Since [G : P ] is not divisible by p, it follows that for some 1 ≤ i ≤ t we

have [Q : x−1
i P xi ∩ Q] not divisible by p. Since Q is a p-group, this implies
Q = x−1
i P x −1
i ∩ Q, in other words Q ≤ xi P xi .
Theorem 6.7 (Sylow II). Let p be a prime. Any two Sylow p-subgroups of
G are conjugate.
Proof. Let P and Q be Sylow p-subgroups of G. By Theorem 6.6, there
exists g ∈ G such that Q ≤ g −1 P g. Since |Q| = |P | = |g −1 P g|, it follows
that Q = g −1 P g.
138
Theorem 6.8 (Sylow III). Let p be a prime, and let pα be the largest power
of p dividing |G|. Write |G| = pα m. Then the following hold:
(i) np (G) divides m,
(ii) np (G) = [G : NG (P )] for any Sylow p-subgroup P ≤ G,
(iii) np (G) ≡ 1 mod p.
Proof. Let P ≤ G be a Sylow p-subgroup of G. By Sylow II and Lemma
1.137 we have np (G) = [G : NG (P )], which divides
m = [G : P ] = [G : NG (P )][NG (P ) : P ].
Thus (i) and (ii) hold. For (iii), we consider the (P, NG (P ))-double coset
decomposition of G. Let x1 , . . ., xt ∈ G be representatives for the (P, NG (P ))-
double cosets in G with x1 = 1. Then
t
[
G= P xi NG (P )
i=1
and x2 , . . . , xt 6∈ NG (P ). Since P x1 NG (P ) = NG (P ), we have

t
X
|G| = |NG (P )| + |P xi NG (P )|
i=2
t
X |P ||NG (P )|
= |NG (P )| + (6.1)
i=2
|NG (P ) ∩ x−1
i P xi |
Xt
= |NG (P )| + |NG (P )|[P : P ∩ x−1
i P xi ] (6.2)
i=2
where (6.1) holds by Lemma 6.1 and (6.2) holds by Lemma 6.5 (ii). Therefore
t
X
[G : NG (P )] = 1 + [P : P ∩ x−1
i P xi ].
i=2
For 2 ≤ i ≤ t we cannot have P = P ∩ x−1 i P xi , since this would imply P =

x−1
i P x i and contradict x i ∈
6 NG (P ). We conclude then that [P : P ∩ x−1
i P xi ]
is divisible by p for all 2 ≤ i ≤ t, and thus [G : NG (P )] ≡ 1 mod p. By (ii),
we have np (G) ≡ 1 mod p.
Example 6.9. There are many examples where using Sylow III, we can
deduce from the order |G| of a finite group that it has a normal Sylow p-
subgroup.
139
(a) Suppose that |G| = 6 = 2 · 3. Then n3 (G) divides 2, so n3 (G) = 1 or

n3 (G) = 2. On the other hand n3 (G) ≡ 1 mod 3, so n3 (G) = 1. Thus
G has a normal subgroup of order 3 (as we have already seen previously,
for example in Lemma 5.45).
(b) Suppose that |G| = pq, where p > q are primes. By Sylow III np (G)
divides q, so np (G) = 1 or q. Since p > q we have q 6≡ 1 mod p, so by
Sylow III np (G) = 1. Therefore G has a normal subgroup of order p.
Similarly we know that nq (G) divides p, so nq (G) = 1 or nq (G) = p.
Since nq (G) ≡ 1 mod q, if q - p − 1, then nq (G) = 1. If q | p − 1, we
have nq (G) = p when G = Cp oψ Cq is a nontrivial semidirect product.
(c) Suppose that |G| = 40 = 23 · 5. Then n5 (G) divides 8 = 23 , so n5 (G) is

equal to 1, 2, 4, or 8. Since 2, 4, 8 6≡ 1 mod 5, we conclude n5 (G) = 1
and so G has a normal subgroup Q of order 5. By Sylow I, there exists
a subgroup P ≤ G with |P | = 8 = 23 . Then G is the semidirect product
of Q and P , so
G∼ = C 5 oψ P
for some homomorphism ψ : P → Aut(C5 ). We know that there are 5
groups of order |P | = 8 up to isomorphism. Thus the classification of
groups of order 40 is equivalent to classifying the following semidirect
products:
• C5 oψ C8 (3 groups up to isomorphism)
• C5 oψ (C4 × C2 ) (4 groups up to isomorphism)
• C5 oψ (C2 × C2 × C2 ) (2 groups up to isomorphism)
• C5 oψ D8 (3 groups up to isomorphism)
• C5 oψ Q8 (2 groups up to isomorphism)
For now we omit the details of how these semidirect products are classi-
fied up to isomorphism. However, the conclusion in the end is that there
are 3 + 4 + 2 + 3 + 2 = 14 groups of order 40.
(d) Suppose that |G| = 42 = 2 · 3 · 7. Then n7 (G) divides 6, so n5 (G) is equal

to 1, 2, 3, or 6. Since 2, 3, 6 6≡ 1 mod 7, we must have n7 (G) = 1 and
so G has a normal subgroup of order 7.
(e) Suppose that |G| = 56 = 23 · 7. Then n7 (G) divides 8, so n7 (G) equals

1, 2, 4, or 8. Since 2, 4 6≡ 1 mod 7, we have n7 (G) = 1 or n7 (G) = 8.
If n7 (G) = 1, then there is a normal subgroup of order 7 in G. Suppose
that n7 (G) = 8. For distinct 7-Sylows Q, Q0 we have Q ∩ Q0 = {1} since
140
they have prime order, so there are a total of 8 · (7 − 1) = 48 = |G| − 8

elements of order 7 in G. This implies that there can be only one Sylow
2-subgroup, as otherwise we would get too many elements. In that case
there is a normal subgroup of order 8.
In conclusion, G will always a have a normal Sylow subgroup (either a
2-Sylow or a 7-Sylow, or both). It follows that either G ∼ = P oψ C7 or
∼
G = C7 oψ0 P , where |P | = 8. Classifying these semidirect products up
to isomorphism, one finds that there are a total of 13 groups of order 56,
up to isomorphism.
Lemma 6.10. Let P be a Sylow p-subgroup of G. Let H be a subgroup of G
such that NG (P ) ≤ H. Then NG (H) = H.
Proof. It is clear that H ≤ NG (H). For the converse, let x ∈ NG (H). Since
P ≤ H, we have xP x−1 ≤ H. Since P and xP x−1 are Sylow p-subgroups
of H, they are conjugate in H (Sylow II), so there exists h ∈ H such that
hP h−1 = xP x−1 . Then h−1 x ∈ NG (P ), so x ∈ HNG (P ) = H. Therefore
NG (H) ≤ H, and thus NG (H) = H.
6.3 Sylow subgroups of subgroups and quotients

Let G be a finite group and let p be a prime. We will denote the set of Sylow
p-subgroups of G by Sylp (G). That is, if pα is the largest power of p dividing
G, we have
Sylp (G) = {P ≤ G : |P | = pα }.
Given a subgroup H ≤ G, or a normal subgroup N E G, what can we
say about Sylp (H), Sylp (N ), and Sylp (G/N )? The next two lemmas provide
some answers.
Lemma 6.11. Let H ≤ G. Then
Sylp (H) ⊆ {P ∩ H : P ∈ Sylp (G)}.
Proof. Exercise.
One can find examples where Sylp (H) $ {P ∩ H : P ∈ Sylp (G)}. (Exer-
cise.)
Lemma 6.12. Let G be a finite group, let p be a prime, and let N E G.
Then
Sylp (N ) = {P ∩ N : P ∈ Sylp (G)}
and
Sylp (G/N ) = {P N/N : P ∈ Sylp (G)}.
141
Proof. Exercise.
Lemma 6.13. Let G be a finite group, let p be a prime. Then:
(i) If H ≤ G, then np (H) ≤ np (G).
(ii) If N E G, then np (N ) | np (G).
(iii) If N E G, then np (G/N ) | np (G)
Proof. (i) Follows from Lemma 6.11.
(ii) By Lemma 6.12, every Sylow p-subgroup of N has the form P ∩ N for
some P ∈ Sylp (G). Now
(P ∩ N )g = P g ∩ N g = P g ∩ N
for all g ∈ G. Thus G acts by conjugation on the set of Sylow p-

subgroups of N , and this action is transitive since Sylow subgroups are
conjugate in G (Sylow II). Thus np (N ) = [G : NG (P ∩ N )] by the
orbit-stabilizer theorem. Now NG (P ) ≤ NG (P ∩ N ), so
np (G) = [G : NG (P ∩N )][NG (P ∩N ) : P ∩N ] = np (N )[NG (P ∩N ) : P ∩N ].
Thus np (N ) | np (G).
(iii) For P ∈ Sylp (G), we have P N/N ∈ Sylp (G/N ). Then NG/N (P N/N ) =
NG (P N )/N by Lemma 1.139. Thus
np (G/N ) = [G/N : NG (P N )/N ] = [G : NG (P N )].
On the other hand NG (P ) ≤ NG (P N ) (since (P N )x = P x N x = P x N ),

so
np (G) = [G : NG (P N )][NG (P N ) : NG (P )]
= np (G/N )[NG (P N ) : NG (P )].
In particular np (G/N ) | np (G).
We note that tere are examples where H ≤ G and np (H) - np (G). The
smallest example occurs for G = A5 . There exists a subgroup H < G such
that H ∼= S3 . An exercise shows that n2 (H) = 3 and n2 (G) = 5, so n2 (H)
does not divide n2 (G).
More generally for N E G, we have the following formula:
142
np (G) = np (N )np (G/N )np (NP N (P ∩ N ))

We will omit the proof, but this formula is due to Marshall Hall Jr.,
who studied properties of the numbers np (G) in a paper from 1967. Among
other things, he proved that there does not exist a finite group G such that
n3 (G) = 22 (although 22 ≡ 1 mod 3).
6.4 Number of p-subgroups of given order

Let G be a finite group and let p be a prime. Suppose that pn is the largest
power of p that divides |G|. By Sylow III, we know that the number of
subgroups of order pn is ≡ 1 mod p. It turns out the same is true for the
number of subgroups of order pk , for every 0 ≤ k ≤ n. We will prove this in
the following.
Theorem 6.14 (Frobenius, 1895). Let G be a finite group, and let p be a

prime. If pk divides |G|, then the number of subgroups of order pk in G is
≡ 1 mod p.
Proof. Denote by rk the number of subgroups of order pk in G. We will

proceed by induction on k. For k = 0 we have r0 = 1, while for k = 1
we have r1 ≡ 1 mod p by Corollary 1.128. Suppose then that k ≥ 1 and
that rk ≡ 1 mod p for all finite groups G with pk | |G|. We will show that
rk+1 ≡ 1 mod p.
To this end, let P1 , . . ., Prk be the subgroups of order pk in G. For
1 ≤ i ≤ rk , let λi be the number of subgroups of order pk+1 that contain Pi .
Note that λi is the number of subgroups of order p in NG (Pi )/Pi . Indeed, if
Pi Q and |Q| = pk+1 , then Pi E Q by Proposition 1.148. Thus Q ≤ NG (Pi ),
and Q/Pi is a subgroup of order p in NG (Pi )/Pi . Conversely, if Q/Pi is a
subgroup of order p in NG (Pi )/Pi , then Q is a subgroup of order pk+1 that
contains Pi .
Since Pi is contained in some Sylow p-subgroup (Theorem 6.6) and since
normalizers grow in p-groups (Proposition 1.147), the group NG (Pi )/Pi is
divisible by p. Thus the number of subgroups of order p in NG (Pi )/Pi is ≡ 1
mod p by Corollary 1.128. That is, we have λi ≡ 1 mod p for all 1 ≤ i ≤ rk .
Similarly, let Q1 , . . ., Qrk+1 be the subgroups of order pk+1 in G. For
1 ≤ i ≤ rk+1 , we denote by µi the number of subgroups of order pk that
are contained in Qi . By induction, we know that µi ≡ 1 mod p for all
1 ≤ i ≤ rk+1 .
Then
λ1 + · · · + λrk = µ1 + · · · + µrk+1 ,
143
because both sides of the equation count the number of pairs (Pi , Qj ) with
Pi ≤ Qj . We have λi ≡ 1 mod p and µj ≡ 1 mod p for all 1 ≤ i ≤ rk and
1 ≤ j ≤ rk+1 , so it follows that
rk ≡ rk+1 mod p.
We conclude then that rk+1 ≡ 1 mod p.
6.5 Nilpotent groups and Sylow subgroups

We have seen previously that a finite abelian group is isomorphic to the direct
product of its Sylow subgroups (Lemma 5.22). More generally, this property
characterizes finite nilpotent groups.
Proposition 6.15. Let G be a finite group. The following statements are

equivalent:
(i) G is nilpotent.
(ii) Every Sylow subgroup of G is normal.
(iii) G is the direct product of its Sylow p-subgroups.
Proof. We will prove (i) ⇒ (ii) ⇒ (iii) ⇒ (i).
(i) ⇒ (ii): Suppose that G is nilpotent, and let P ≤ G be a Sylow p-subgroup

of G. By Lemma 6.10 we have NG (NG (P )) = NG (P ). Therefore NG (P ) can-
not be a proper subgroup of G, since normalizers grow in nilpotent groups
(Proposition 4.43). Thus NG (P ) = G, which means that P E G.
(ii) ⇒ (iii): Write the prime factorization of |G| as |G| = pα1 1 · · · pαt t . For
1 ≤ i ≤ t, let Pi ≤ G be a Sylow pi -subgroup of G. By assumption Pi E G.
Then any product Pi1 · · · Pit of some Pi ’s is a subgroup.
Moreover we claim that |Pi1 · · · Pit | = |Pi1 | · · · |Pit |, and this is easily
proven by induction. Indeed, first for t = 1 the claim is trivial. For t > 1
we have Pi1 ∩ (Pi2 · · · Pit ) = {1}, since by induction Pi2 · · · Pit is a subgroup
with order |Pi2 | · · · |Pit |, which is coprime to |Pi1 |. Thus |Pi1 (Pi2 · · · Pit )| =
|Pi1 | · |Pi2 · · · Pit | = |Pi1 | · |Pi2 | · · · |Pit | by Lemma 1.88.
It follows then that P1 P2 · · · Pt is a subgroup of order pα1 1 · · · pαt t = |G|, so
G = P1 P2 · · · Pt . For 1 ≤ i ≤ t we have Pi ∩ (P1 · · · Pi−1 Pi+1 · · · Pt ) = {1},
since P1 · · · Pi−1 Pi+1 · · · Pt is a subgroup of order |P1 | · · · |Pi−1 | · |Pi+1 | · · · |Pt |
which is coprime to |Pi |. It follows then from Lemma 5.5 that G is the direct
product of the subgroups P1 , . . ., Pt .
144
(iii) ⇒ (i): In this case G is isomorphic to P1 × · · · × Pt , where Pi is a finite

pi -group for some prime pi . Since finite p-groups are nilpotent (Lemma 4.42),
and since the direct product of nilpotent groups is nilpotent (Exercise 5.6),
we conclude that G is nilpotent.
For a finite group and a prime p, we define
\
Op (G) := P,
P ∈Sylp (G)
the intersection of all Sylow p-subgroups of G. It is clear that Op (G) char G,

and in particular Op (G) E G. Note also that since Sylow p-subgroups are
conjugate, for any Sylow p-subgroup P ≤ G we have Op (G) = ∩g∈G P g .
Lemma 6.16. Let G be a finite group and let Q E G be a p-group. Then

Q ≤ Op (G).
Proof. By Theorem 6.6, there exists Sylow p-subgroup P such that Q ≤ P .

Then for g ∈ G we have Q = Qg ≤ P g , so in fact Q ≤ P g for all g ∈ G. Since
Sylow p-subgroups are conjugate (Sylow II), we conclude that Q is contained
in every Sylow p-subgroup of G. In other words, Q ≤ Op (G).
By Lemma 6.16, the subgroup Op (G) is the largest normal p-subgroup of
G. That is:
• Op (G) is a p-group and Op (G) E G;
• If Q is a p-group and Q E G, then Q ≤ Op (G).
Lemma 6.17. Let G be a finite group and let p1 , . . ., pt be the prime divisors
of |G|. Then F (G) is the direct product of Op1 (G), . . ., Opt (G).
Proof. Since F (G) is nilpotent, every Sylow subgroup of F (G) is normal

(Proposition 6.15). Let Pi E F (G) be a Sylow pi -subgroup of F (G). Since
Pi is the unique Sylow pi -subgroup of F (G), we have Pi char F (G) E G and
so Pi E G. Thus Pi ≤ Opi (G). On the other hand Opi (G) is a nilpotent
normal subgroup of G, so Opi (G) ≤ F (G). Therefore Opi (G) ≤ Pi , since Pi
is the unique Sylow pi -subgroup of G. Thus Pi = Opi (G).
We have shown that Op1 (G), . . ., Opt (G) are the Sylow subgroups of F (G),
so the lemma follows from Proposition 6.15.
145
6.6 Groups of certain orders are solvable

A classical application of Sylow theorems are proofs of solvability of groups
with certain orders. For example, we already now that any finite group of
order pn , where p is a prime, must be solvable (Lemma 4.42). Burnside’s
theorem states that any group of order pn q m (p, q primes) is solvable. The
easiest proof of this fact uses character theory, although later a purely group-
theoretical proof was found (Goldschmidt, Bender, Matsuyama). We will give
a proof in the special case where m = 1.
Proposition 6.18. Let p and q be distinct primes and n > 0 an integer.

Then any group of order pn q is solvable.
Proof. Let G be a finite group of order pn q, where p and q are distinct primes
and n > 0 is an integer. We will first prove that G cannot be simple.
We have np (G) | q, so np (G) = 1 or np (G) = q. If np (G) = 1, then G
has a normal Sylow p-subgroup. Suppose then that np (G) = q, in which case
q ≡ 1 mod p. Among the q Sylow p-subgroups, choose two P1 , P2 such that
the order of D = P1 ∩ P2 is as large as possible.
We claim that q divides |NG (D)|. If this is not the case, then NG (D) is a
p-group, so there exists a Sylow p-subgroup R such that NG (D) ≤ R. On the
other hand D P1 and normalizers grow in p-groups (Proposition 1.147), so
D NG (D) ∩ P1 . But now NG (D) ≤ R, so D R ∩ P1 , which contradicts
the assumption that |D| is as large as possible. Thus q divides |NG (D)|.
In particular, there exists a q-Sylow subgroup Q ≤ NG (D). Now Q
acts on the Sylow p-subgroups by conjugation, and this action is faithful
and transitive since by np (G) = q we have [G : NG (P1 )] = q and thus
NG (P1 ) = P1 . In other words, each Sylow p-subgroup is equal to P1y for some
y ∈ Q. On the other hand D ≤ P1 , so D = Dy ≤ P1y since Q ≤ NG (D).
Therefore D is contained in every Sylow p-subgroup of G. This implies that
\
D= P
P ∈Sylp (G)
and in particular D E G.
If D 6= {1}, then D is a nontrivial proper normal subgroup of G, so G is
not simple. Suppose then that D = {1}. Since we have assume that D is
the largest possible intersection of two Sylow p-subgroups of G, in this case
P ∩ P 0 = {1} for any distinct Sylow p-subgroups of G. Thus there are a total
of (pn − 1)q = |G| − q non-identity elements in the Sylow p-subgroups of G,
and all of these elements have order divisible by p. This leaves space for at
most q elements among the Sylow q-subgroups of G, so there can be at most
146
one Sylow q-subgroup in G. Thus nq (G) = 1, in which case G has a normal

subgroup of order q.
This completes the proof of the fact that G cannot be simple. To prove
that G is solvable, we can proceed by induction on n. For n = 1 we already
know that G has a normal Sylow subgroup N (Lemma 5.45). In this case
N and G/N are both cyclic, so G is solvable (Lemma 4.23). Suppose that
n > 1, and let N be a proper non-trivial normal subgroup of G. Then N
0
is either a p-group, a q-group, or a group of order pn q with n0 < n, so it
is solvable by induction. Similarly G/N is solvable by induction, so G is
solvable (Lemma 4.23).
Proposition 6.19. Let G be a finite group order p2 q 2 , where p and q are
primes. Then G is solvable.
Proof. We already know that groups of order p, q, p2 , q 2 , pq, p2 q, and pq 2 are
solvable (Proposition 6.18, Lemma 4.42). Thus it will suffice to prove that
G is not simple.
We can assume that np (G) > 1, as otherwise G has a normal Sylow p-
subgroup. Then np (G) = q or np (G) = q 2 . In both cases q 2 ≡ 1 mod p, so p
divides q 2 − 1 = (q − 1)(q + 1). Since p is a prime, we conclude that p | q ± 1.
We can also assume that nq (G) > 1, and then similarly q | p ± 1. It is then
straightforward to see that we must have q = p ± 1. For example if p | q − 1
and q | p + 1, then p < q ≤ p + 1 which implies q = p + 1.
Suppose then without loss of generality that q = p + 1. Then either p or
q must be even, so p = 2 and q = 3. In this case |G| = 36 = 22 · 32 . Let
P ≤ G be a 3-Sylow subgroup of G. We have [G : P ] = 4, so the action of
left cosets of P gives a homomorphism ϕ : G → S4 with Ker ϕ ≤ P (Lemma
2.43). Since S4 is not divisible by 9, we conclude that 3 | | Ker ϕ|, so Ker ϕ
is a non-trivial proper normal subgroup of G.
Proposition 6.20. Let G be a finite group of order |G| = pqr, where p <
q < r are primes. Then G is solvable.
Proof. We know that groups of order p, q, r, pq, pr, and qr are solvable, so
it will suffice to prove that G is not simple.
To this end, we will prove that G has a normal Sylow subgroup. If this
is not the case, then np (G), nq (G), nr (G) > 1. Note that np (G) divides
qr, so np (G) ≥ q. Similarly nq (G) divides pr and nq (G) ≡ 1 mod p, so
nq (G) = r or nq (G) = pr; in particular nq (G) ≥ r. Finally nr (G) divides pq
and nr (G) ≡ 1 mod r, so nr (G) = pq. We can now count elements of prime
order in G:
• np (G) ≥ q, so we have ≥ q(p − 1) elements of order p;
147
• nq (G) ≥ r, so we have ≥ r(q − 1) elements of order q;
• nr (G) = pq, so we have pq(r − 1) elements of order r;
This gives us at least
pq(r − 1) + r(q − 1) + q(p − 1) = pqr + rq − r − q > pqr
elements of prime order (since rq > r+q), contradicting |G| = pqr. Therefore
there must be a normal Sylow subgroup.
6.7 Fitting subgroup and normal Sylow subgroups

Lemma 6.21. Let G be a finite solvable group. Then CG (F (G)) ≤ F (G).
In other words, we have CG (F (G)) = Z(F (G)).
Proof. Denote F = F (G) and C = CG (F ). Suppose, for the sake of contra-

diction, that C 6≤ F . Consider CF/F , the image of C in G/F . Then CF/F
is a non-trivial group, and it is solvable since G is. Thus there exists k > 0
such that (CF/F )(k) = C (k) F/F is nontrivial, but (CF/F )(k+1) = {1}.
By F E G we have C = CG (F ) E G. Furthermore C (k) char C E G, so
C (k) E G. Next we will prove that C (k) is nilpotent. To this end, first note
that [C (k) , C (k) ] ≤ F since (CF/F )(k+1) = C (k+1) F/F is trivial. On the other
hand C (k) centralizes F , so [C (k) , [C (k) , C (k) ]] = {1} and C (k) is nilpotent of
class ≤ 2.
We have shown that C (k) is a nilpotent normal subgroup of G, which im-
plies that C (k) ≤ F . But this is a contradiction, since (CF/F )(k) = C (k) F/F
is nontrivial. Therefore we must have C ≤ F , which implies C = Z(F ).
Suppose that G is a finite solvable group, so then CG (F (G)) = Z(F (G))
by Lemma 6.21. Since F (G) E G, we have a map π : G → Aut(F (G)), with
Ker π = CG (F (G)) = Z(F (G)) (from the conjugation action, see Theorem
5.32). By the first isomorphism theorem, the quotient G/Z(F (G)) is isomor-
phic to a subgroup of Aut(F (G)). Thus G can be seen as a extension of two
groups related to F (G):
• There is a normal subgroup N E G, the center of F (G).
• The quotient G/N is isomorphic to a subgroup of Aut(F (G)).
Conceptually, the point of this is that to a large extent the structure of

a finite solvable group G is controlled by its Fitting subgroup F (G), since G
is an extension of Z(F (G)) and a subgroup of Aut(F (G)).
148
We will mostly apply this in the case where F (G) is abelian: in this
case, by Lemma 6.21 the quotient G/F (G) is isomorphic to a subgroup of
Aut(F (G)).
Lemma 6.22. Let G be a finite group of order |G| = p2 q, where p and q are
distinct primes. Then G has a normal Sylow subgroup.
Proof. We know that G is solvable (Proposition 6.18), so F (G) 6= {1}.

If q | |F (G)|, then being nilpotent, F (G) has a normal Sylow q-subgroup
Q. Then Q char F (G) E G and thus Q E G.
We can assume then that q - |F (G)|, so |F (G)| = p or |F (G)| = p2 .
If |F (G)| = p, then F (G) ∼ = Cp and by Lemma 6.21 the quotient G/F (G)
∼
embeds into Aut(Cp ) = Cp−1 , which is absurd since |G/F (G)| = pq > p − 1.
Therefore |F (G)| = p2 , in which case F (G) is a normal Sylow p-subgroup of
G.
Lemma 6.23. Let G be a finite group of order |G| = p2 q 2 , where p and q

are distinct primes. Then G has a normal Sylow subgroup.

If p2 | |F (G)|, then as a nilpotent group F (G) has a normal Sylow p-
subgroup P of order p2 . Since P char F (G) E G, we have P E G. Similarly
if q 2 | |F (G)|, we see that a normal Sylow q-subgroup of F (G) is a normal
Sylow q-subgroup of G.
Thus we may assume that p2 and q 2 do not divide |F (G)|. Since F (G) 6=
{1}, the possibilities are |F (G)| = p, |F (G)| = q, and |F (G)| = pq. In all
of these cases F (G) must be cyclic (since it is nilpotent), so by Lemma 6.21
the quotient G/F (G) embeds into Aut(F (G)).
If |F (G)| = p, then G/F (G) is a group of order pq 2 that embeds into
Aut(F (G)) ∼ = Aut(Cp ) ∼ = Cp−1 , which is impossible. For a similar reason
|F (G)| = q is impossible.
The only remaining case is |F (G)| = pq. In this case F (G) is a nilpotent
group of order pq, so F (G) ∼ = Cp × Cq . Then Aut(F (G)) ∼ = Aut(Cp ) ×
∼
Aut(Cq ) = Cp−1 × Cq−1 . Now G/F (G) is a group of order pq that embeds
into Aut(F (G)), so this is a contradiction since pq > (p − 1)(q − 1).
Lemma 6.24. Let G be a finite group of order |G| = p3 q, where p and q are
distinct primes. Then G has a normal Sylow subgroup, except when p = 2,
q = 3, and G ∼= S4 .
If q | |F (G)|, then as a nilpotent group F (G) has a normal Sylow q-
subgroup Q of order q. Since Q char F (G) E G, we have Q E G. Similarly
149
if p3 | |F (G)|, we see that a normal Sylow p-subgroup of F (G) is a normal

Sylow p-subgroup of G.
Therefore we can assume that q and p3 do not divide |F (G)|, in which
case |F (G)| = p or |F (G)| = p2 . In both cases F (G) is abelian, so by Lemma
6.21 the quotient G/F (G) embeds into Aut(F (G)).
If |F (G)| = p, then G/F (G) is a group of order p2 q that embeds into
Aut(F (G)) ∼ = Aut(Cp ) ∼= Cp−1 , which is impossible.
Therefore we must have |F (G)| = p2 , and in this case F (G) ∼ = Cp2 or
∼ ∼
F (G) = Cp × Cp . Suppose first that F (G) = Cp2 . Then G/F (G) embeds
into
Aut(F (G)) ∼
= Aut(Cp2 ) ∼= (Z/p2 Z)× ∼= Cp(p−1) .
In particular G/F (G) is abelian of order pq, so it has a normal subgroup
N/F (G) E G/F (G) of order p. Then N E G is a normal subgroup of order
p3 , which is not possible when |F (G)| = p2 .
Therefore F (G) ∼
= Cp × Cp . In this case G/F (G) embeds into Aut(Cp ×
∼
Cp ) = GL2 (p). We have
|G/F (G)| = pq
| GL2 (p)| = (p − 1)2 p(p + 1)
so pq divides (p − 1)2 p(p + 1), which implies q | p ± 1. On the other hand,

since G does not have a normal Sylow p-subgroup, we have np (G) > 1. By
Sylow III this implies np (G) = q and q ≡ 1 mod p, so p | q − 1. Then
p < q ≤ p ± 1, so it follows that q = p + 1. Then either p or q is even, so we
must have p = 2 and q = 3.
Thus we have |G| = 24 = 23 · 3, with F (G) ∼= C2 × C2 . Since n3 (G) > 1,
it follows from Sylow III that n3 (G) = 4. Then for a 3-Sylow subgroup Q, we
have [G : NG (Q)] = 4, so NG (Q) is a group of order 6. We consider the action
of G of the Sylow 3-subgroups by conjugation, which gives a homomorphism
(Lemma 2.43)
ϕ : G → S4
with \
Ker ϕ = NG (Q)g .
g∈G
Since K = Ker ϕ is a proper subgroup of NG (Q), it must have order 1, 2, or

3. The order of K cannot be 3, since there is no normal Sylow 3-subgroup
in G.
We consider the possibility that |K| = 2. In this case G/K is a group of
order 12 = 22 · 3, so by Lemma 6.22 it has a normal Sylow 2-subgroup or a
normal Sylow 3-subgroup. If there exists a normal subgroup N/K E G/K
150
of order 22 = 4, then N E G is a normal subgroup of order 23 , which is not

possible when F (G) ∼ = C2 × C2 . The other possibility is that there exists a
normal subgroup N/K E G/K of order 3, in which case N E G is a normal
subgroup of order 2 · 3 = 6. We know that any group of order 6 has a normal
Sylow 3-subgroup, so there exists a unique subgroup Q E N of order 3. Then
Q char N E G implies that Q E G, again impossible when F (G) ∼ = C2 × C2 .
Thus |K| = 2 is not possible, and we must have |K| = 1. In other words,
the homomorphism ϕ : G → S4 is injective, and since |G| = 24 = |S4 | it
must be an isomorphism. Hence G ∼ = S4 .
6.8 Groups of order p2 q (p and q distinct primes)

Let p and q be distinct primes. We consider the classification of groups of
order p2 q. As a first step, let G be a finite group of order |G| = p2 q, with a
Sylow p-subgroup P ≤ G and a Sylow q-subgroup Q ≤ G. By Lemma 6.22,
we know that G has a normal Sylow subgroup, so either P or Q is a normal
subgroup. Since P ∩ Q = {1}, we have |P Q| = |P ||Q| = |G|, so G = P Q
and G is a semidirect product of P and Q (or of Q and P ).
If both P and Q are normal, then G is the direct product of P and Q
thus isomorphic to one of the following groups:
Cp 2 × Cq
Cp × Cp × Cq .
It remains to consider the case where G is a nontrivial semidirect product.

In this case, we know (Theorem 5.40) that G is isomorphic to a nontrivial
semidirect product of one of the following forms:
Cp2 oψ Cq ,
(Cp × Cp ) oψ Cq ,
Cq oψ Cp2 ,
Cq oψ (Cp × Cp ).
We begin with the following proposition, which is similar to Theorem

5.62.
Proposition 6.25. Let K = hxi be a finite cyclic group, and let H be a
finite abelian group. Let ψ, ψ 0 : K → Aut(H) be homomorphisms. Suppose
that gcd(|H|, |K|) = 1. Then the following statements are equivalent:
(i) H oψ K ∼
= H oψ0 K.
151
(ii) There exists ϕ ∈ Aut(K) such that ψ(ϕ(x)) is conjugate to ψ 0 (x) in

Aut(H).
Proof. (i) ⇒ (ii): Suppose that there exists an isomorphism σ : H oψ K →
H oψ0 K. Since gcd(|H|, |K|) = 1, by Lemma 5.60 in both groups H oψ K
and H oψ0 K the subgroup H is the unique subgroup of order |H|. Therefore
σ(H) = H, so the restriction π = σ|H : H → H is an automorphism of H.
The image σ(x) can be written in the form hk with h ∈ H and k ∈ K,
say σ(x) = h0 xd with h0 ∈ H and d ∈ Z. We have a surjective homomor-
phism τ : H oψ0 K → K defined by hk 7→ k for all h ∈ H and k ∈ K.
Then τ σ must be also surjective, so τ σ(x) = xd must be a generator of K.
Therefore gcd(d, |K|) = 1, and the map ψd : K → K defined by x 7→ xd is
an automorphism of K (Lemma 5.27). Thus σ(x) = h0 ψd (x).
We now consider the action of x on H. Let h ∈ H. Then
σ(xhx−1 ) = σ(ψ(x)(h)) = (π ◦ ψ(x))(h).
Using the fact that σ is a homomorphism and H is abelian, we have
σ(xhx−1 ) = σ(x)σ(h)σ(x)−1 (since σ is a homomorphism)

= h0 x d
π(h)x−d h−1
0
= h0 · ψ (x )(π(h)) · h−1
0 d
0
= ψ 0 (xd )(π(h)) (since H is abelian)
= (ψ 0 ψd (x))(π(h)).
Therefore we conclude that
(π ◦ ψ(x))(h) = (ψ 0 (ψd (x)) ◦ π)(h)
for all h ∈ H. Thus π ◦ ψ(x) = ψ 0 (ψd (x)) ◦ π, so πψ(x)π −1 = ψ 0 ψd (x).

Hence ψ(x) and ψ 0 ψd (x) are conjugate in Aut(H), so (ii) holds with
ϕ = ψd−1 .
(ii) ⇒ (i): Suppose that there exists ϕ ∈ Aut(K) such that ψ(ϕ(x)) and
ψ 0 (x) are conjugate in Aut(H). Then there exists π ∈ Aut(H) such that
πψ(ϕ(x))π −1 = ψ 0 (x). For all k ∈ Z, we have
ψ 0 (xk ) = ψ 0 (x)k
= (πψ(ϕ(x))π −1 )k
= π(ψ(ϕ(x))k π −1
= π(ψ(ϕ(xk ))π −1
152
since ψ and ϕ are homomorphisms. Thus πψ(ϕ(y))π −1 = ψ 0 (y) for all y ∈ K,

so it follows from Lemma 5.43 (ii) that H oψ0 K ∼
= H oψϕ K. By Lemma 5.43
(i) we have H oψϕ K = H oψ K, so we conclude that H oψ K ∼
∼ = H oψ0 K.
For the classification of groups of order p2 q, we will begin with the cases
where the normal Sylow subgroup is cyclic.
Lemma 6.26. Let p and q be distinct primes. Let N be the number of
nontrivial semidirect products Cq oψ (Cp × Cp ), up to isomorphism. Then:
(i) N = 0, if p - q − 1.
(ii) N = 1, if p | q − 1.
Proof. If p - q − 1, then there are no elements of order p in Aut(Cq ) ∼ = Cq−1 ,
and thus every homomorphism ψ : Cp × Cp → Cq is trivial. This proves (i).
For (ii), suppose that p | q − 1. Let ψ, ψ 0 : Cp × Cp → Aut(Cq ) be non-
trivial homomorphisms. Since Aut(Cq ) ∼ = Cq−1 is cyclic, the images Im ψ
and Im ψ 0 must be cyclic of order p. In fact Im ψ = Im ψ 0 , since Cq−1 has a
unique subgroup of order p.
By the first isomorphism theorem Ker ψ = hxi and Ker ψ 0 = hx0 i are
cyclic of order p. Considering Cp × Cp as a vector space over Fp , we can
extend to a basis {x, y} of Cp × Cp by choosing any y 6∈ hxi. In this case
Im ψ = hψ(y)i.
Since Im ψ = Im ψ 0 , we can find y 0 ∈ Cp × Cp such that ψ 0 (y 0 ) = ψ(y).
Then y 0 6∈ hx0 i, so {x0 , y 0 } is another basis of Cp × Cp as a vector space over
Fp .
We can now define a linear map ϕ : Cp × Cp → Cp × Cp by ϕ(x0 ) = x
and ϕ(y 0 ) = y. Since ϕ maps basis to a basis, it is an isomorphism of vector
spaces, and thus ϕ ∈ Aut(Cp × Cp ). Then we have ψϕ = ψ 0 . Indeed,
ψϕ(x0 ) = ψ(x) = 1 = ψ 0 (x0 )
ψϕ(y 0 ) = ψ(y) = ψ 0 (y 0 )
and thus ψϕ(g) = ψ 0 (g) for all g ∈ Cp × Cp , since this holds for g in the basis
{x0 , y 0 }.
Since ψϕ = ψ 0 , it follows from Lemma 5.43 (ii) that Cq oψ (Cp × Cp ) ∼ =
Cq oψ (Cp × Cp ). We have shown that there is at most one nontrivial semi-
0
direct product Cq o (Cp × Cp ), up to isomorphism. To finish the proof of (ii),

it remains to check that there exists at least one non-trivial homomorphism
ψ : Cp × Cp → Aut(Cq ). For this, take {x, y} a basis of Cp × Cp and define
ψ(x) = I (identity) and ψ(y) = σ, where σ ∈ Aut(Cq ) is an automorphism of
order p. (Such an automorphism exists since p | q − 1 and Aut(Cq ) ∼ = Cq−1 .)
Then ψ defines a non-trivial homomorphism Cp × Cp → Aut(Cq ).
153
Remark 6.27. Alternatively, one can prove Lemma 6.26 by showing that
any semidirect product Cq oψ (Cp × Cp ) is isomorphic to (Cq oψ0 Cp ) × Cp
for some homomorphism ψ 0 : Cp → Aut(Cq ). Then the result follows from
Theorem 5.48.

nontrivial semidirect products Cq oψ Cp2 , up to isomorphism. Then:
(i) N = 0, if p - q − 1.
(ii) N = 1, if p | q − 1 and p2 - q − 1.
(iii) N = 2, if p2 | q − 1.
Proof. Let Cp2 = hxi. Let ψ, ψ 0 : Cp2 → Aut(Cq ) be homomorphisms. By

Proposition 6.25, we know that Cq oψ Cp2 ∼ = Cq oψ0 Cp2 if and only if there
exists ϕ ∈ Aut(Cp2 ) such that ψ(ϕ(x)) and ψ 0 (x) are conjugate in Aut(Cq ).
On the other hand we know that Aut(Cq ) ∼ = (Z/qZ)× ∼ = Cq−1 is cyclic,
∼
so Cq oψ Cp2 = Cq oψ0 Cp2 if and only if there exists ϕ ∈ Aut(Cp2 ) such that
ψ(ϕ(x)) = ψ 0 (x). We will show that this is equivalent to hψ(x)i = hψ 0 (x)i.
To this end, first note that automorphisms ϕ ∈ Aut(Cp2 ) are maps defined
by ϕd : x 7→ xd , where gcd(d, p) = 1. Thus if ψ(ϕd (x)) = ψ 0 (x), then ψ(x)d =
ψ 0 (x). Since gcd(d, p) = 1, we have hψ(x)i = hψ 0 (x)i. Conversely, suppose
that hψ(x)i = hψ 0 (x)i. The image ψ(x) must have order dividing p2 . Thus
since ψ 0 (x) generates hψ(x)i, by Lemma 1.46 we have ψ 0 (x) = ψ(x)d = ψ(xd )
for some d ∈ Z with gcd(d, p) = 1. Therefore ψ(ϕd (x)) = ψ 0 (x).
We have shown that Cq oψ Cp2 ∼ = Cq oψ0 Cp2 if and only if hψ(x)i = hψ 0 (x)i.
Since Aut(Cq ) is cyclic q − 1, it has a unique subgroup of each order dividing
q − 1. Thus hψ(x)i = hψ 0 (x)i if and only if |ψ(x)| = |ψ 0 (x)|. Since x has
order p2 , the possible orders for |ψ(x)| are 1, p, and p2 . We can now easily
finish the proof:
(i): If p - q −1, then there are no elements of order p or p2 in Aut(Cq ) ∼

= Cq−1 .
Thus the only possibility is ψ(x) = 1, in which case ψ is trivial.
(ii): If p | q − 1 and p2 - q − 1, then there is σ ∈ Aut(Cq ) with |σ| = p, but

no element of order p2 . We can define a homomorphism ψ : Cp2 → Aut(Cq )
by ψ(x) = σ, in which case |ψ(x)| = p. Thus there is exactly one non-trivial
semidirect product.
(iii): If p2 | q − 1, there is σ ∈ Aut(Cq ) with |σ| = p2 . We can define

homomorphisms ψ, ψ 0 : Cp2 → Aut(Cq ) with ψ(x) = σ and ψ 0 (x) = σ p .
154
Then |ψ(x)| = p2 and |ψ 0 (x)| = p. Thus there are exactly two non-trivial
semidirect products.

nontrivial semidirect products Cp2 oψ Cq , up to isomorphism. Then:
(i) N = 0, if q - p − 1.
(ii) N = 1, if q | p − 1.
Proof. Let Cq = hxi. Note that Aut(Cp2 ) ∼ = (Z/p2 Z)× ∼ = Cp(p−1) for all
primes p. Thus by applying Proposition 6.25 and arguing exactly as in the
proof of Lemma 6.28, we see that Cp2 oψ Cq ∼ = Cp2 oψ0 Cq if and only |ψ(x)| =
|ψ 0 (x)|.
Now the order of |ψ(x)| is 1 or q, so any two non-trivial semidirect pro-
ducts Cp2 oψ Cq must be isomorphic.
If q - p − 1, there is no element of order q in Aut(Cp2 ) ∼
= Cp(p−1) . Thus if
q - p − 1, any homomorphism ψ : Cq → Aut(Cp2 ) is trivial, and so (i) holds.
If q | p − 1, then there exists an element σ ∈ Aut(Cp2 ) of order q. Then
we can define a non-trivial homomorphism ψ : Cq → Aut(Cp2 ) by ψ(x) = σ.
Thus (ii) holds in this case.
With Lemma 6.26, Lemma 6.28, and Lemma 6.29, what remains is to
consider semidirect products of the form (Cp ×Cp )oψ Cq . Here Aut(Cp ×Cp ) ∼
=
GL2 (p), and for the classification we will need a few more facts about the
structure of GL2 (p).
Lemma 6.30. Let p be a prime. Then there exists an element of order p + 1

in GL2 (p).
Proof. Let K = Fp2 be a finite field of order p2 . As a vector space over Fp , we

have dim K = 2 (see Section 3.1). We know that the multiplicative group of a
finite field is cyclic, so K× ∼
= Cp2 −1 . Now p + 1 divides p2 − 1 = (p − 1)(p + 1),
so there exists an element ε ∈ K× of order p + 1.
Define a map ϕ : K → K by ϕ(x) = εx for all x ∈ K. It is clear that ϕ
is an Fp -linear map, and ϕ is a bijection since it has an inverse defined by
x 7→ ε−1 x. In other words, we have ϕ ∈ GL(K), where again K is considered
as a 2-dimensional vector space over Fp . Since ε has order p + 1 in K× , it is
easy to see that |ϕ| = p + 1.
Thus we have shown that GL(K) contains an element of order p + 1.
The lemma follows, since K has dimension 2 over Fp and hence GL(K) ∼ =
GL2 (p).
155
Lemma 6.31. Let p be a prime, and let q be an odd prime dividing p + 1.

Then Sylow q-subgroups of GL2 (p) are cyclic.
Proof. Note that | GL2 (p)| = (p − 1)2 p(p + 1). Let q k be the largest power
of q dividing | GL2 (p)|. We have gcd(p − 1, p + 1) = 2, so since q is odd and
q | p + 1, the largest power of q dividing | GL2 (p)| is the largest power of q
dividing p + 1. In other words, q k divides p + 1.
By Lemma 6.30, there exists x ∈ GL2 (p) with |x| = p + 1. Then y =
(p+1)/q k
x has order |y| = q k , so hyi is a Sylow q-subgroup of GL2 (p). Since
Sylow q-subgroups are conjugate, we conclude that every Sylow q-subgroup
of GL2 (p) is cyclic.
Theorem 6.32. Let p and q be distinct primes. Let N be the number of

nontrivial semidirect products (Cp × Cp ) oψ Cq , up to isomorphism. Then:
(i) N = 2, if q = 2.
(ii) N = 1, if q > 2 and q | p + 1.

q+3
(iii) N = 2
, if q > 2 and q | p − 1.
(iv) N = 0, if q > 2 and q - p ± 1.
Proof. First we note that if q = 2, the result follows from Example 5.65,
which shows that there are exactly 2 nontrivial semidirect products of the
form (Cp × Cp ) oψ C2 , up to isomorphism. We will assume then for the result
of the proof that q > 2.
Now Aut(Cp ×Cp ) ∼ = GL2 (p), and | GL2 (p)| = (p−1)2 p(p+1). Therefore if
q - p±1, there is no element of order q in GL2 (p), and so every homomorphism
ψ : Cq → Aut(Cp × Cp ) is trivial; as claimed by (iv). Thus we can assume
that q | p ± 1.
Let Cq = hxi. We consider the two possibilities in turn.
Case 1: q | p + 1.
By Cauchy’s theorem there exists an element of order q in Aut(Cp × Cp ), so
there is a nontrivial homomorphism Cq → Aut(Cp × Cp ). We will show next
that any two nontrivial semidirect products of the form (Cp × Cp ) oψ Cq are
isomorphic, as claimed by (ii).
To this end, let ψ, ψ 0 : Cq → Aut(Cp ×Cp ) be non-trivial homomorphisms,
so ψ(x) and ψ 0 (x) are automorphisms of order q. By Lemma 6.31, the Sylow
q-subgroups of Aut(Cp × Cp ) ∼ = GL2 (p) are cyclic. In particular each Sylow
q-subgroup of GL2 (p) has a unique subgroup of order q, so by conjugacy of
Sylow q-subgroups we conclude that any two subgroups of order q in GL2 (p)
156
are conjugate (Exercise 6.7). Therefore, there exists π ∈ Aut(Cp × Cp ) such

that πhψ(x)iπ −1 = hψ 0 (x)i.
Then πψ(x)π −1 is a generator for hψ 0 (x)i, so there exists d ∈ Z such that
gcd(d, q) = 1 and ψ 0 (x) = (πψ(x)π −1 )d = πψ(xd )π −1 . Now ψd : Cq → Cq
defined by x 7→ xd is an automorphism of Cq , and we have shown that ψψd (x)
and ψ 0 (x) are conjugate in Aut(Cp × Cp ). It follows therefore from Proposi-
tion 6.25 that (Cp × Cp ) oψ Cq ∼
= (Cp × Cp ) oψ0 Cq .
Case 2: q | p − 1.
First note that in this case we have p > 2. We consider Cp × Cp as a vector
space over Fp , and fix a basis {v, w} of Cp × Cp . Then we can identify
Aut(Cp × Cp ) = GL2 (p), with a matrix

a b
c d
denoting the automorphism
v 7→ v a wb
w 7→ v c wd
of Cp × Cp .
Since q | p−1, there exists an element µ ∈ F×
p of order q. (In other words,
we can find an integer µ such that µ 6≡ 1 mod p and µq ≡ 1 mod p.) Then
hµi is a subgroup of order q in F× p and each of its elements is a root of the
q q
polynomial t − 1 ∈ Fp [t]. Thus t − 1 splits into linear factors
tq − 1 = (t − 1)(t − µ) · · · (t − µq−1 ).
Thus if g ∈ GL2 (p) has order q, then g q −1 = 0 and so the minimal polynomial
of g divides tq − 1. In this case either g is a scalar matrix of the form µk I2 , or
it has a minimal polynomial (t − µi )(t − µj ) with µi 6= µj (note that tq − 1 has
no repeated roots). In any case, it follows that g is conjugate to a diagonal
matrix of the form
α 0
0 β
with αq = β q = 1. By conjugating g with the matrix

0 1
1 0
(which correponds to change of basis from {v, w} into {w, v}), we may assume
that α 6= 1.
157
Consider a non-trivial homomorphism ψ : Cq → GL2 (p). Then ψ(x) is an

element of order q in GL2 (p). It follows from the discussion in the previous
paragraph that there exists a π ∈ GL2 (p) such that
a
−1 µ 0
πψ(x)π =
0 µb
for some a, b ∈ Z with µa 6= 1. Now µa has order q in F× p , so it generates
hµi and there exists d ∈ Z with gcd(d, q) = 1 such that µad = µ. Then
ϕ : Cq → Cq , x 7→ xd defines an automorphism of Cq , and

−1 −1 d µ 0
πψ(ϕ(x))π = (πψ(x)π ) = .
0 µbd
It follows then from Proposition 6.25 that (Cp ×Cp )oψ Cq ∼
= (Cp ×Cp )oψ0 Cq ,
0
where ψ : Cq → GL2 (p) is the homomorphism defined by

0 µ 0
ψ (x) = .
0 µbd
Next, for 0 ≤ t ≤ q − 1, we define Gt = (Cp × Cp ) oψt Cq , where ψt : Cq →
GL2 (p) is the homomorphism defined by

0 µ 0
ψ (x) = .
0 µt
So far we have proven that any nontrivial semidirect product (Cp × Cp ) oψ Cq
is isomorphic to one of the groups
G0 , G1 , . . . , Gq−1 .
Next we will use Proposition 6.25 to determine the isomorphisms between
these groups. We will prove the following:
Claim: Gt ∼= Gs if and only if t = s, or ts ≡ 1 mod q.

By Proposition 6.25, we have Gt ∼ = Gs if and only if there exists ϕ ∈ Aut(Cq )
such that ψt (ϕ(x)) and ψs (x) are conjugate in GL2 (p). Any such ϕ is of the
form ϕ = ϕd : Cq → Cq , x 7→ xd for some d ∈ Z with gcd(d, q) = 1.
Thus for 0 ≤ t, s ≤ q − 1, we have Gt ∼ = Gs if and only if there exists
d ∈ Z with gcd(d, q) = 1 such that the matrices
d
µ 0 µ 0
and
0 µtd 0 µs
are conjugate in GL2 (p). Two diagonal matrices are conjugate if and only if
they have the same diagonal entries, so these two matrices are conjugate if
and only if one of the following holds:
158
• µd = µ and µtd = µs , in which case µs = (µd )t = µt .
• µd = µs and µtd = µ, in which case µts = (µs )t = (µd )t = µ.
In the first case µs = µt implies t ≡ s mod q, so t = s. In the second

case µts = µ implies ts ≡ 1 mod q. The claim follows.
With the claim proven, we can easily see which of the groups
G0 , G1 , . . . , Gq−1
are isomorphic to each other. Indeed, the groups G0 , G1 , Gq−1 are not
isomorphic any other Gi , while for 1 < i < q − 1 we have Gi ∼
= Gi0 , where i0
is the inverse of i modulo q. (Note that 1 and −1 are the only elements in
F×q which are equal to their own inverse.) Thus there are a total of
q−3 q+3
3+ =
2 2
groups among G0 , G1 , . . . , Gq−1 , up to isomorphism.
Let p and q be distinct primes. With Lemma 6.26, Lemma 6.29, Lemma
6.28, and Theorem 6.32, we have completed the classification of groups of
order p2 q. As a summary, the number of groups of order p2 q, up to isomor-
phism, is given in Table 5.
6.9 Transfer
Let G be a finite group. In this section we will discuss the transfer homo-
morphism, which is a certain homomorphism G → H/[H, H] defined for any
subgroup H ≤ G. For our purposes we will only consider the case where H
is abelian, in which case the transfer is a homomorphism G → H. (This will
suffice for our purposes, and the more general definition is not much more
difficult.)
Let G be a finite group, and let H ≤ G be an abelian subgroup. Let T
be a left transversal for H in G, in other words, a set of representatives for
the left cosets of H in G. Then |T | = [G : H], and
{gH : g ∈ G} = {tH : t ∈ T }.
We can then define the “dot action” of G on T as follows. For all g ∈ G

and t ∈ T , we have gtH = t0 H for a unique t0 ∈ T , which we denote by
t0 = g • t. It is straightforward to verify that this defines an action of G on
T , so the following properties hold:
159
p - q − 1, N =2
q - p ± 1. Abelian: C p2 q
Cp × Cpq
Non-abelian: None
p > 2, q = 2. N =5
Abelian: C2p2
Cp × C2p
Non-abelian: Cp2 o C2 ∼= D2p2 (1 group)
(Cp × Cp ) o C2 (2 groups)
p = 2, q = 3. N = 5.
Abelian: C12
C2 × C6
Non-abelian: (C2 × C2 ) o C3 ∼
= A4
C3 o (C2 × C2 ) ∼
= S3 × C2 ∼
= D12
C3 o C4 (1 group)
p2 | q − 1. N =5
Abelian: C p2 q
Cp × Cpq
Non-abelian: (Cq o Cp ) × Cp (1 group)
Cq o Cp2 (2 groups)
q > 3, N =4
p | q − 1, Abelian: C p2 q
p2 - q − 1 Cp × Cpq
Non-abelian: (Cq o Cp ) × Cp (1 group)
Cq o Cp2 (1 group)
p > 2, q > 2, N =3
q | p + 1. Abelian: C p2 q
Cp × Cpq
Non-abelian: (Cp × Cp ) o Cq (1 group)
q > 2, N = q+9
2
q | p − 1. Abelian: C p2 q
Cp × Cpq
Non-abelian: (Cp2 ) o Cq (1 group)
(Cp × Cp ) o Cq ( q+3
2 groups)
Table 5: For p and q distinct primes, the number N of groups of order p2 q,

up to isomorphism.
160
– 1 • t = t for all t ∈ T ;
– (g • h) • t = (gh) • (t) for all t ∈ T and g, h ∈ G;
– For all g ∈ G, the map t 7→ g • t is a bijection T → T .
(This action is equivalent to action of G on the left cosets of H.)

For all g ∈ G and t ∈ T we have gtH = (g • t)H, so (g • t)−1 gt ∈ H.
Therefore we can define a map vT : G → H by
Y
vT (g) = (g • t)−1 gt
t∈T
for all g ∈ G. We call vT the transfer map from G to H. Note that since H
is abelian, the product over T is defined uniquely and does not depend on
the ordering of T .
We will first check that the transfer map does not depend on the choice
of T , and after that we verify that it is a homomorphism.
Lemma 6.33. Let H ≤ G be an abelian subgroup of a finite group G. Then
vT = vS for any left transversals T, S of H in G.
Proof. Let T and S be left transversals of H in G. We denote the dot action
of G on T by • and on S by •0 . Then gtH = (g • t)H and gsH = (g •0 s)H
for all g ∈ G, t ∈ T , and s ∈ S.
For each t ∈ T , there exists a unique s ∈ S such that sH = tH. In this
case s = tht for a unique ht ∈ H, so
S = {tht : t ∈ T }.
For all t ∈ T , we have
(g •0 (tht ))H = gtht H = gtH = (g • t)H = (g • t)hg•t H.
Therefore
(g •0 (tht )) = (g • t)hg•t . (6.3)
Now we can calculate that
Y
vS (g) = (g •0 s)−1 gs
s∈S
Y
= (g •0 (tht ))−1 gtht
t∈T
Y
= ((g • t)hg•t )−1 gtht (by (6.3))
t∈T
161
Y
= h−1 −1
g•t (g • t) gtht
t∈T
Y Y Y
= h−1
g•t · (g • t)−1 gt · ht (since H is abelian)
t∈T t∈T t∈T
Y Y
= h−1
t · vT (g) · ht (since t 7→ g • t is a bijection)
t∈T t∈T
= vT (g)
for all g ∈ G. Thus vS = vT .

Thus for an abelian subgroup H ≤ G, we can write v : G → H for the
transfer map.
Lemma 6.34. Let H ≤ G be an abelian subgroup of a finite group G. Then
the transfer map v : G → H is a homomorphism.
Proof. Fix a left transversal T for H in G. For all x, y ∈ G, we have
Y
v(xy) = ((xy) • t)−1 xyt
t∈T
Y
= (x • (y • t))−1 xyt (since • is an action)
t∈T
Y
= (x • (y • t))−1 x(y • t)(y • t)−1 yt
t∈T
Y Y
= (x • (y • t))−1 x(y • t) · (y • t)−1 yt (since (g • t)−1 gt ∈ H)
t∈T t∈T
Y Y
= (x • t)−1 xt (y • t)−1 yt (since t 7→ y • t is a bijection)
t∈T t∈T
= v(x)v(y).
so v : G → H is a homomorphism.
Before applying the transfer homomorphism, we will need one more lemma
for calculating the value of v(g) for g ∈ G.
Lemma 6.35. Let H be an abelian subgroup of a finite group G, and let T
be a left transversal for H in G. Let g ∈ G. Let T0 ⊆ T be representatives
for the orbits of hgi on T under the dot action, and for t ∈ T0 denote by nt
the size of the hgi-orbit
hgi • t = {g k • t : k ∈ Z}.
Let v : G → H be the transfer map. Then the following statements hold:
162
P
(i) t∈T0 nt = [G : H];
(ii) nt divides |g| for all t ∈ T0 ;
(iii) t−1 g nt t ∈ H for all t ∈ T0 ;
(iv) v(g) = t∈T0 t−1 g nt t.

Q
Proof. Since T is the disjoint union of the hgi-orbits, (i) follows. Claim (ii)
follows since the size of an orbit divides the order of the group, so nt divides
|hgi| = |g|.
For (iii), let t ∈ T0 . Then the hgi-orbit of t is
{g k • t : k ∈ Z} = {g k • t : 0 ≤ k < nt }.
For each t0 = g k • t in this orbit, in the product v(g) we have a factor
(g • t0 )−1 gt0 = (g • (g k • t))g(g k • t) = (g k+1 • t)−1 g(g k • t).
Thus
Y Y
(g • t0 )−1 gt0 = (g k+1 • t)−1 g(g k • t)
t0 =g k •t 0≤k<nt
0≤k<nt
Y
= (g nt −k • t)−1 g(g nt −k−1 • t)
0≤k<nt
= (g • t)−1 g nt (g 0 • t)
nt
= t−1 g nt t,
which proves (iii). Claim (iv) also follows from this calculation, since
[
T = hgi • t.
t∈T0
This completes the proof of the lemma.

As an example application, we can prove the following.
Theorem 6.36. Let G be a finite group and [G : Z(G)] = n. Then the map
g 7→ g n is a homomorphism G → G.
Proof. We will prove that the map g 7→ g n is exactly the transfer map v :
G → Z(G), and thus a homomorphism. To this end, let T be a left transversal
for Z(G) in G.
163
Let g ∈ G, and choose T0 ⊆ T and integers nt for all t ∈ T0 with

properties as in Lemma 6.35 (i) – (iv). Then t−1 g nt t ∈ Z(G) and thus
t−1 g nt t = g nt ∈ Z(G) for all t ∈ T0 . Moreover
Y
v(g) = g nt = g n ,
t∈T0
P
since t∈T0 nt = [G : Z(G)] = n.
Theorem 6.37 (Burnside). Let G be a finite group and P ≤ G a Sylow
p-subgroup of G. Suppose that P ≤ Z(NG (P )) (in other words, NG (P ) =
CG (P )). Then P has a normal complement in G: that is, there exists a
subgroup K E G such that G = KP , K ∩ P = {1} and |K| = [G : P ].
Proof. Since P is central in NG (P ), in particular it is abelian and we have
the transfer homomorphism v : G → P . We will find that K = Ker v is a
normal complement for P . To this end, we first prove the following.
Claim: Let x, y ∈ P . If x and y are conjugate in G, then x = y.

Suppose that x, y ∈ P and y = gxg −1 for some g ∈ G. Since P is abelian,
we have P ≤ CG (x). Then gP g −1 ≤ CG (gxg −1 ) = CG (y). But we also have
P ≤ CG (y) since y ∈ P , so P and gP g −1 are Sylow p-subgroups of CG (y).
By Sylow II they are conjugate in CG (y), so there exists h ∈ CG (y) such that
hgP g −1 h−1 = P . Then hg ∈ NG (P ). But P is central in NG (P ), so hg must
commute with x. Hence
x = hgxg −1 h−1 = hyh−1 = y,
where the last equality holds since h ∈ CG (y).
We will use the claim to evaluate the transfer v : G → P on elements of

P . Let T ⊆ G be a left transversal for P P in G. Consider g ∈ P , and let
T0 ⊆ T and nt be as in Lemma 6.35. Then t∈T0 nt = [G : P ], and
Y
v(g) = t−1 g nt t,
t∈T0
where t−1 g nt t ∈ P for all t ∈ T0 . We have g nt ∈ P and t−1 g nt t ∈ P elements

of P which are conjugate in G, so by the claim above g nt = t−1 g nt t. Therefore
Y
v(g) = g nt = g [G:P ]
t∈T0
for all g ∈ P .
164
This proves that P ∩ Ker v = {1}, since gcd(|P |, [G : P ]) = 1 for a Sylow

p-subgroup P . Hence v restricted to P is injective, which shows that v must
be a surjective homomorphism. By the first isomorphism theorem K = Ker v
has order [G : P ], and the theorem follows.
6.10 Groups of squarefree order

An integer n > 1 is said to be squarefree if it is not divisible by the square of
any prime. In this case n = p1 p2 · · · pt for some distinct primes p1 , . . ., pt . In
this section, we will give a complete classification of all groups of squarefree
order.
For n squarefree, we will see that the number of groups of order n is
X Y pf (p,m) − 1
gnu(n) = ,
n
p−1
m|n p| m
n/m|φ(m) p prime
where:
• φ is the Euler totient function,
• f (p, m) is the number of prime divisors q of m such that q ≡ 1 mod m.

This result was originally proven by Hölder (1895).
Lemma 6.38. Let G be a group of order |G| = p1 p2 · · · pt , where p1 < p2 <
· · · < pt are primes. Then the following are equivalent:
(i) G is nilpotent.
(ii) G is abelian.
(iii) G is cyclic.
Proof. (i) ⇒ (ii): If G is nilpotent, it is isomorphic to the direct product of
its Sylow subgroups. In this case the Sylow subgroups of G are cyclic, so G
is abelian.
(ii) ⇒ (iii): Follows from the fact that Cm × Cn ∼ = Cmn if gcd(m, n) = 1.
(Lemma 5.18.)
(iii) ⇒ (i): This is clear.
Theorem 6.39. Let G be group of order |G| = p1 p2 · · · pt , where p1 < p2 <
· · · < pt are primes. Then for all 1 ≤ i ≤ t, there exists a normal subgroup
N E G such that |N | = pi pi+1 · · · pt . (In particular, G has a normal Sylow
pt -subgroup.)
165
Proof. If t = 1 the claim is trivial, so we assume that t > 1 and proceed by

induction on t.
Let P ≤ G be a Sylow p1 -subgroup of G. Recall that NG (P )/CG (P ) is iso-
morphic to a subgroup of Aut(P ). On the other hand p1 is the smallest prime
divisor of |G|, so the order of Aut(P ) ∼ = Cp1 −1 is coprime to |G|, and thus
NG (P )/CG (P ) is trivial. In other words NG (P ) = CG (P ), so P ≤ Z(NG (P )).
By Burnside’s normal complement theorem (Theorem 6.37), there exists a
normal subgroup K E G with |K| = [G : P ] = p2 · · · pt .
With this we can complete the proof. For i = 1 we can choose N = G.
For 2 ≤ i ≤ t, by induction there exists a normal subgroup N E K such that
|N | = pi pi+1 · · · pt . Since gcd(|N |, [K : N ]) = 1, it follows from Lemma 5.60
that N is the unique subgroup of its order in K. Thus N char K E G, which
gives N E G and completes the proof.
Theorem 6.40. Let G be group of order |G| = p1 p2 · · · pt , where p1 < p2 <

· · · < pt are primes. Then G is solvable.
Proof. For t = 1 the claim is trivial, so assume that t > 1 and proceed by
induction on t. By Theorem 6.39, there exists a normal subgroup N E G
with |N | = p2 · · · pt . By induction N is solvable, and G/N is solvable since
it is cyclic; thus G is solvable (Lemma 4.23).
Theorem 6.41. Let G be a group of order |G| = p1 p2 · · · pt , where p1 <

p2 < · · · < pt are primes. Then the Fitting subgroup F (G) has a cyclic
complement K in G. That is, there exists a cyclic subgroup K ≤ G such that
K ∩ F (G) = {1} and G = F (G)K.
Proof. From Theorem 6.40 we know that G is solvable. Thus by Lemma

6.21 we have CG (F (G)) = F (G), and so G/F (G) embeds into Aut(F (G)).
Since F (G) is nilpotent, by Lemma 6.38 we have F (G) ∼ = Cd for some d | n.
∼ ×
Thus G/F (G) embeds into Aut(Cd ) = (Z/dZ) , so in particular G/F (G) is
abelian.
Hence G/F (G) is cyclic of order n/d by Lemma 6.38. It is generated by
a coset gF (G) for some g ∈ G, in which case g must have order divisible
by n/d. Then g 0 = g k for k = |g|d/n is an element of order n/d in G. Let
K = hg 0 i. Then K is a cyclic subgroup of order n/d = [G : F (G)]. We have
K ∩ F (G) = {1} since gcd(d, n/d) = 1. Then G = F (G)K since
|F (G)| · |K|
|F (G)K| = = |F (G)| · |K| = |G|.
|F (G) ∩ K|
166
With Theorem 6.41, we have essentially described the structure of a group

G with squarefree order |G|. Indeed, by Theorem 6.41 we know that G ∼ =
F (G)oψ K, where F (G) and K are both cyclic. Moreover, since CG (F (G)) =
F (G) by the solvability of G, the homomorphism ψ : K → Aut(F (G)) must
be injective. This means that there is an element of order [G : F (G)] in
Aut(F (G)).
With this in mind, we can now construct all groups of squarefree order.
Let n > 1 be a squarefree integer. Let m | n, and write Cm = hyi and
Cn/m = hxi. Let σ ∈ Aut(Cm ) be an automorphism of order n/m (if one
exists). Then we define
Gm,σ = Cm oψ Cn/m ,
where ψ : Cn/m → Aut(Cm ) is the homomorphism defined by ψ(x) = σ.
Theorem 6.42. Let n > 1 be squarefree and let G be a group of order
n. Then there exists m | n and an automorphism σ ∈ Aut(Cm ) such that
G∼= Gm,σ .
Proof. Let m be the order of F (G). We have F (G) ∼ = Cm since F (G) is
nilpotent (Lemma 6.38). By Theorem 6.41, there exists a cyclic subgroup
K = hxi of order n/m such that G = F (G)K and F (G) ∩ K = {1}.
Since G is solvable, we have F (G) = CG (F (G)). Thus the homomorphism
ψ : K → Aut(F (G)) defined by ψ(k)(h) = khk −1 for all k ∈ K and h ∈ F (G)
is injective. In particular ψ(x) = σ is an element of order n/m in Aut(F (G)).
By Theorem 5.40 we have
G∼
= F (G) oψ K
and it is clear that F (G) oψ K ∼
= Gm,σ .
Lemma 6.43. Let n > 1 be squarefree, let m | n and let σ ∈ Aut(Cm ) be an
automorphism of order n/m. Then F (Gm,σ ) = Cm .
Proof. We have Gm,σ = H oψ K, where H ∼ = Cm and K ∼
= Cn/m . We will
prove that F (Gm,σ ) = H. Since H E Gm,σ , we have H ≤ F (Gm,σ ).
For the other inclusion, first we note that
F (Gm,σ ) = H(F (Gm,σ ) ∩ K)
since Gm,σ = HK and H ≤ F (Gm,σ ). Furthermore since F (Gm,σ ) is nilpo-
tent of squarefree order, it is cyclic by Lemma 6.38. In particular F (Gm,σ )
centralizes H. Since the homomorphism ψ : K → Aut(H) is injective, no
non-identity element of K centralizes H, and thus F (Gm,σ ) ∩ K = {1}. We
conclude then that F (Gm,σ ) = H.
167
Theorem 6.44. Let n > 1 be squarefree. Let m, m0 | n and let σ, σ 0 ∈

Aut(Cm ) be automorphisms of orders n/m and n/m0 , respectively. Then
Gm,σ ∼= Gm0 ,σ0 if and only if m = m0 and hσi = hσ 0 i.
Proof. Suppose that Gm,σ ∼ = Gm0 ,σ0 . Then F (Gm,σ ) ∼= F (Gm0 ,σ0 ), so by
0
Lemma 6.43 we have m = m . Let Cn/m = hxi. Since Aut(Cm ) is abe-
lian, by Proposition 6.25 there exists d ∈ Z with gcd(d, n/m) = 1 such that
σ d = σ 0 . Since |σ| = |σ 0 | = n/m, it follows that hσi = hσ 0 i.
Conversely, suppose that m = m0 and hσi = hσ 0 i. Then there exists d ∈ Z
with gcd(d, n/m) = 1 such that σ d = σ 0 . It follows then from Proposition
6.25 that Gm,σ ∼ = Gm,σ0 , as claimed by the theorem.
Let n > 1 be squarefree. From Theorem 6.42 we know that any group G
of order n is isomorphic to Gm,σ for some m | n and some automorphism σ ∈
Aut(Cm ) of order n/m. Furthermore, from Theorem 6.44 we know precisely
when two groups Gm,σ , Gm0 ,σ0 are isomorphic: we have Gm,σ ∼
= Gm0 ,σ0 if and
0 0
only if m = m and σ, σ generate the same cyclic subgroup of Aut(Cm ).
With this we can conclude the following.
Theorem 6.45. Let n > 1 be squarefree and m | n. Let gm be the number
of groups G of order n with |F (G)| = m, up to isomorphism. Then gm is the
number of subgroups of order n/m in Aut(Cm ), which is given by
Y pf (p,m) − 1
gm = .
n
p−1
p| m
p prime
Here f (p, m) denotes the number of prime divisors q of m such that q ≡ 1

mod p.
Proof. It follows from Theorem 6.42, Lemma 6.43, and Theorem 6.44 that
gm is the number of subgroups of order n/m in Aut(Cm ).
To prove the formula for gm , write n/m = p1 · · · pt with pi distinct primes,
and m = q1 · · · qs with qi distinct primes. For 1 ≤ i ≤ t, let di be the number
of subgroups of order pi in Aut(Cm ). Then we have
Y
gm = di .
1≤i≤t
(Indeed, since Aut(Cm ) is abelian, any subgroup of order n/m in Aut(Cm )

is a direct product of unique subgroups P1 , . . ., Pt with |Pi | = pi .)
It remains to verify that the number d = di of subgroups of order p = pi
in Aut(Cm ) is equal to
pf (p,m) − 1
.
p−1
168
For this, note that
Aut(Cm ) ∼
= Aut(Cq1 ) × Aut(Cq2 ) × · · · × Aut(Cqs )
∼
= Cq1 −1 × Cq2 −1 × · · · × Cqs −1 .
As a cyclic group, each Cqi −1 either has no elements of order p, or a unique

subgroup of order p. Therefore in Aut(Cm ) there is a subgroup X isomorphic
f (p,m)
to Cp , and furthermore X contains every subgroup of Aut(Cm ) with order
p.
The value of d is thus given by the number of subgroups of order p in
f (p,m) f (p,m)
Cp . Any non-identity element of Cp has order p, and the subgroup
generated by it contains p − 1 elements of order p. Combining this with
the fact that distinct subgroups of order p have no elements of order p in
f (p,m)
common, we conclude that the number of subgroups of order p in Cp is
equal to
pf (p,m) − 1
.
p−1
Note that Aut(Cm ) ∼
= (Z/mZ)× has order φ(m), so Aut(Cm ) contains an
element of order n/m only if n/m divides φ(m). Thus we have the following
formula for the number of groups of squarefree order n.
X Y pf (p,m) − 1
gnu(n) = ,
n
p−1
m|n p| m
n/m|φ(m) p prime
where f (p, m) is defined as above.
169
Exercises
In each exercise, G is a group. We recall notation from the lecture notes:
• |S| = cardinality of a set S.
• |x| = order of an element x ∈ G.
• od (G) = the number of elements of order d in G.
• For S ⊆ G and g ∈ G, we denote S g := g −1 Sg.
• For S, T ⊆ G, we denote ST = {st : s ∈ S, t ∈ T }.
• R>0 = {x ∈ R : x > 0}.
• Similarly Z≥0 = {0, 1, 2, 3, . . .}, etc.
• gnu(n) = number of finite groups of order n, up to isomorphism.
• νp (n) = the largest integer α ≥ 0 such that pα | n (for p prime and
n 6= 0).
• Fp is the field (Z/pZ, +, ·) of integers modulo p (for p prime).
• k(G) = number of conjugacy classes in G.
Section 1
1.1 Let n > 0 be an integer. Prove that 1 ≤ gnu(n) < ∞. Why gnu(0) = 0?
1.2 Examples of non-abelian groups:
(a) Let Ω be a set with |Ω| ≥ 3. Show that Sym(Ω) is not abelian.
(b) Let F be a field and let n ≥ 2 be an integer. Show that GLn (F) is
not abelian.

−1 1 −1 0
1.3 Let G = GL2 (Q). Define x, y ∈ G by x = and y = .
0 1 0 1
Show that |x| = 2, |y| = 2, and |xy| = ∞.
1.4 Let G be a group.
(a) Suppose that x2 = 1 for all x ∈ G. Prove that G is abelian.

(b) Suppose that (xy)2 = x2 y 2 for all x ∈ G. Prove that G is abelian.
(c) Suppose that G is abelian. Prove that (xy)n = xn y n for all n ∈ Z.
170
1.5 Suppose that Ω and Ω0 are sets such that |Ω| = |Ω0 |. (In other words, sup-
pose that there exists a bijection Ω → Ω0 .) Prove that Sym(Ω) ∼ = Sym(Ω0 ).
1.6 Let G and H be groups, and let ϕ : G → H be an isomorphism.
(a) Prove that the inverse map ϕ−1 : H → G is an isomorphism.

(b) Let ϕ0 : G → H be an isomorphism. Prove that ϕ0 = ϕψ for a
unique automorphism ψ : G → G.
(c) Prove that G is abelian if and only if H is abelian.
(d) Prove that G is cyclic if and only if H is cyclic.
(e) Prove that |x| = |ϕ(x)| for all x ∈ G.
(f) For d > 0, let od (G) be the cardinality of the set {x ∈ G : |x| = d}.
Prove that od (G) = od (H) for all d > 0.
1.7 Prove the following statements.
(a) (Q, +) ∼
6= (R, +).
(b) (Q, +) ∼
6= (Z, +).
(c) (R, +) ∼
= (R>0 , ·). (Hint: Use the fact that ex+y = ex ey .)
(d) (Q, +) ∼
6= (Q>0 , ·).
1.8 Let H G. Prove that G is generated by G \ H = {x ∈ G : x 6∈ H}.

1.9 Let G = (Z, +).
(a) Prove that G = h2, 3i. Find a generator for the subgroup h4, 6, 12i
of G.
(b) For a, b ∈ Z, what are the generators of ha, bi?
(c) Find a generator for h2i ∩ h3i and h14i ∩ h12i.
(d) For a, b ∈ Z, what are the generators of hai ∩ hbi?
1.10 Let p be a prime number and G a finite group. Let x ∈ G and k ∈ Z≥0 .
k k
Suppose that xp has order divisible by p. Prove that |x| = pk |xp |.
1.11 Let x ∈ G have order mn, where gcd(m, n) = 1. Prove that there exist
unique elements y, z ∈ G such that x = yz = zy and |y| = m, |z| = n. (Hint:
Use Bézout’s lemma.)
171
1.12 Let G = hgi be a cyclic group.
(a) Suppose that |G| = 174. Describe all subgroups of G.

(b) Suppose that |G| = p2 , where p is a prime. Describe all subgroups
of G.
(c) Suppose that |G| = pq, where p and q are distinct primes. Describe
all subgroups of G.
(d) Suppose that |G| = pk , where p is a prime and k ≥ 1. Describe all
subgroups of G.
(a) Suppose that the number of subgroups of G is finite. Prove that G

is finite.
(b) Suppose G 6= {1} and that G has only two subgroups, {1} and G.
What can you say about the structure of G?
(c) Let G be a group and suppose that G has exactly three subgroups,
{1}, H, and G. What can you say about the structure of G?
(d) Let G be a group and suppose that for any H ≤ G, either H = {1}
or H ∼
= G. What can you say about the structure of G?
172
1.14 Let G = (Q, +).
(a) The following subgroup

2 12
h , i
3 7
of (Q, +) is cyclic. Which elements generate it?
(b) Show that every finitely generated subgroup of G is cyclic.
(c) Let H and K be nontrivial subgroups of G. Show that H ∩ K is
nontrivial.
(d) Let H be the set of rational numbers of the form
a
,
2k
where a ∈ Z and k ≥ 0. Prove that H is a subgroup of (Q, +) and
that H is not cyclic.
(e) Let H be the set of all rational numbers with finite decimal ex-
pansion (for example 1/8 = 0.125 ∈ H, but 1/3 = 0.333 . . . 6∈ H).
Prove that H is a subgroup of (Q, +). Is H cyclic?
1.15 Let z be an odd integer and k > 0.

k
(a) For z ≡ 1 mod 4, prove that ν2 (z 2 − 1) = ν2 (z − 1) + k.
k
(b) For z ≡ 3 mod 4, prove that ν2 (z 2 − 1) = ν2 (z + 1) + k.
(c) For n ≥ 3, prove that (Z/2n Z)× is not cyclic.
(d) For n ≥ 2, prove that 5 has order 2n−2 in (Z/2n Z)× . What are the
orders of 3 and 7?
1.16 Let G be a group. For m ∈ Z, we define G(m) = {x ∈ G : xm = 1}.
(a) Let G be an abelian group. Show that G(m) ≤ G for all m ∈ Z.

(b) Suppose that G = hgi is cyclic of finite order n. Let m ≥ 0 be an
integer. Find a generator for G(m). What is the order of G(m)?
(c) What does G(m) look like when G is infinite cyclic?
(d) For G = S3 , calculate G(m) for m = 1, 2, 3, 4, 5, 6. Which ones are
subgroups?
1.17 Let S ⊆ G and let ϕ : G → H be a homomorphism. Prove that

hϕ(S)i = ϕ(hSi).
173
1.18 (See Lemma 1.70.) Let G = hx, yi and H = hz, wi be groups such that
the following hold:
• |x| = 2, xyx−1 = y −1 , and x 6∈ hyi;

• |z| = 2, zwz −1 = w−1 , and z 6∈ hwi.
Suppose that |y| = |w|. Prove that the map ϕ : G → H defined by ϕ(xi y j ) =
z i wj for all i, j ∈ Z is a well-defined isomorphism.
1.19 (See Section 1.7.) Construction of dihedral groups:
(a) Let n ≥ 3. Define x, y ∈ Sn as x : {1, . . . , n} → {1, . . . , n},

x(i) = n + 1 − i for all 1 ≤ i ≤ n, and y = (1 2 · · · n). Prove that
hx, yi is dihedral of order 2n.
(b) Consider x, y ∈ Sym(Z), where the bijections x, y : Z → Z are
defined by
x(i) = −i for all i ∈ Z,

y(i) = i + 1 for all i ∈ Z.
Prove that hx, yi is an infinite dihedral group.
1.20 Let G = hx, yi be dihedral, where |x| = 2, xyx−1 = y −1 , and x 6∈ hyi.

Prove that every element of G \ hyi has order 2.
1.21 Prove that every proper subgroup of Q8 is cyclic. What are the sub-
groups of Q8 ?
1.22 Let H and K be subgroups of a group G. Show that HK is a subgroup
if and only if HK = KH. Find an example where HK is not a subgroup of
G.
1.23 (Basic observations about normal subgroups.)
(a) Let H be a subgroup of G such that [G : H] = 2. Prove that

H E G.
(b) Let G = Q8 . Show that every subgroup of G is normal.
(c) Let G = D8 . Find subgroups H E K E G such that H 6E G.
174
(a) Let H ≤ G. Suppose that G = HH g for some g ∈ G. Prove that

G = H.
(b) Let H, K ≤ G and suppose that G = HK. Prove that G = H x K y
for all x, y ∈ G.
1.25 Let G = Sym(Z), the group formed by bijections Z → Z.
(a) Show that H = {σ ∈ G : σ(x) = x for all x ≤ 0} is a subgroup of

G.
(b) For g ∈ G, show that H g = {σ ∈ G : σ(x) = x for all x ∈
g −1 (Z≤0 )}.
(c) Find an element g ∈ G such that H g ⊂ H and H g 6= H.
1.26 Let G = GL2 (R).
(a) Show that

1 x
H= :x∈Z
0 1
is a subgroup of G.

α 0
(b) Consider a diagonal matrix g = with α, β ∈ R \ {0}.
0 β
Describe H g . For which values of α and β does H g ⊆ H hold?
What about H g = H? Find α and β such that H g ⊂ H, but
H g 6= H.
1.27 Suppose that G is abelian. Prove that G is simple if and only if G is

cyclic of prime order.
1.28 A proper subgroup M < G is said to be maximal, if M ≤ H ≤ G implies
H = M or H = G.
(a) Let M < G be a maximal subgroup. If M is a normal subgroup,

prove that G/M is cyclic of prime order.
(b) Show that there are no maximal subgroups in (Q, +).
175
1.29 (Conjugacy classes for some examples.)
(a) Determine the conjugacy classes of G = Q8 .

(b) Determine the conjugacy classes of G = D2 , G = D4 , G = D6 .
(c) Determine the conjugacy classes of G = D2n for n ≥ 4 even. (You
should find k(G) = (n + 6)/2 classes.)
(d) Determine the conjugacy classes of G = D2n for n ≥ 5 odd. (You
should find k(G) = (n + 3)/2 classes.)
(e) Determine the conjugacy classes of G = D∞ . (There are an infinite
number of conjugacy classes.)
1.30 Let n ≥ 3. Show that Z(Sn ) = {1}.

1.31 Prove that if G/Z(G) is cyclic, then G is abelian.
1.32 Prove that if N E G and |N | = 2, then N ≤ Z(G).
1.33 (See Theorem 1.124.) Let p be a prime number. Suppose that G contains
a subgroup U of order p. For x ∈ G, define [x] = xU if x ∈ CG (U ), and
[x] = {uxu−1 : u ∈ U } if x 6∈ CG (U ).
(a) Show that for all x, y ∈ G, either [x] = [y] or [x] ∩ [y] = ∅.
(b) Show that [x] contains exactly p elements for all x ∈ G.
(c) Suppose that x ∈ G is such that xp = 1. Show that y p = 1 for all
y ∈ [x].
(d) Prove that if G has only finitely many elements of order p, then
op (G) ≡ −1 mod p.
1.34 Let G be a finite group.
(a) Prove that k(G) = 2 if and only if G ∼

= C2 .
(b) Prove that k(G) = 3 if and only if G ∼
= C3 or G ∼
= S3 . (Hint: Use
the class equation.)
1.35 Let G be a group and N E G. Let N ≤ H ≤ G. Then NG/N (H/N ) =

NG (H)/N .
1.36 Let p be a prime. Let G be a non-abelian p-group of order p3 . Prove
that k(G) = p2 + p − 1.
176
1.37 Consider the Heisenberg group

  
 1 a b 
Hp = 0 1 c  : a, b, c ∈ Fp .
0 0 1
 
(a) Show that Hp is a non-abelian subgroup of GL3 (p) such that |Hp | =
p3 .
(b) Determine Z(Hp ).
(c) If p > 2, prove that xp = 1 for all x ∈ Hp .
(d) By part (a), the group H2 is a non-abelian group of order 8. Is
H2 ∼
= Q8 or H2 ∼= D8 ?
(a) Prove that CG (g x ) = CG (g)x for all g, x ∈ G.

(b) Suppose that G is finite. Prove that the number of pairs (x, y) ∈
G × G such that xy = yx is equal to k(G)|G|. Conclude that the
probability that two randomly chosen elements of G commute is
equal to k(G)/|G|.
Section 2
2.1 Let σ = (i1 · · · ik ) and τ = (j1 · · · j` ) be cycles in Sn . If σ and τ are
disjoint, then στ = τ σ.
2.2 (Generators for Sn .)
(a) Show that Sn is generated by (1 2), (1 3), . . ., (1 n).

(b) Show that Sn is generated by (1 2), (2 3), . . ., (n − 1 n).
(c) Show that Sn = h(1 2), (1 2 · · · n)i.
2.3 Let σ ∈ Sn be a k-cycle.
(a) Prove that |σ| = k.

(b) Let d | k. Prove that σ d is a product of d pairwise disjoint cycles
of length k/d.
177
2.4 Denote the maximal order of an element of Sn by g(n). Below is a list

of small values:
n 1 2 3 4 5 6 7 8 9 10 11 12 13 14
g(n) 1 2 3 4 6 6 12 15 20 30 30 60 60 84
Using the values of g(n) given above, find an element of largest order in Sn
for 1 ≤ n ≤ 14.
2.5 Let σ, τ ∈ Sn . Consider the set Y of pairs {i, j} such that {τ (i), τ (j)} ∈
I(σ). Prove that |Y | = |I(σ)|.
2.6 Let σ = (i j) ∈ Sn with 1 ≤ i < j ≤ n. Calculate |I(σ)|, the number of
inversions of σ.
2.7 Let n ≥ 3. Prove that An has no subgroup of index 2.
2.8 Let n ≥ 4. Prove that CSn (An ) = {1}. Conclude that Z(An ) = {1}.
2.9 Let n ≥ 2 and G ≤ Sn . Then either G ≤ An , or G ∩ An is a normal
subgroup of index 2 in G.
2.10 Find representatives for the conjugacy classes of S5 , and the size of each
conjugacy class.
2.11 Let G = S4 .
(a) Prove that N = {(1), (1 2)(3 4), (1 3)(2 4), (1 4)(2 3)} is a normal
subgroup of G.
(b) Find all the normal subgroups of G. (Hint: A normal subgroup is
the union of conjugacy classes.)
2.12 Suppose that n ≥ 5 and N E Sn . Prove that N = {1}, N = An , or

N = Sn .
2.13 Prove that (1 2 3) and (1 3 2) are not conjugate in A4 . Find all the
conjugacy classes in A4 .
2.14 Let σ ∈ Sn be an n-cycle. Prove that CSn (σ) = hσi.
2.15 Let O = orbσ (i) be an orbit of σ ∈ Sn . Prove that if g ∈ CSn (σ), then
g(O) is an orbit of σ. Prove then that CSn (σ) acts on the orbits of σ.
178
2.16 Let n ≥ 3.
(a) Let σ ∈ An . Suppose that in the cycle decomposition of σ, there

is a cycle of even length. Prove that CSn (σ) 6≤ An .
(b) Let σ ∈ An . Suppose that in the cycle decomposition of σ, there
are two cycles with equal length. Prove that CSn (σ) 6≤ An . (Hint:
If the length is even, use the previous part. If the length is odd,
use an odd permutation that swaps the two orbits corresponding
to the cycles.)
(c) Prove that for σ ∈ An , the conjugacy class σ Sn splits in An if and
only if the cycle decomposition of σ consists of disjoint cycles with
distinct odd lengths. (Hint: For one direction, use the other parts
of this exercise. For the other direction, use exercises 2.14 and
2.15.)
2.17 Let G be a group and H ≤ G. Consider the action of G on the set
X = {gH : g ∈ G}
of left cosets of H in G. Prove that G acts transitively on X, and that

stabG (xH) = xHx−1 for all x ∈ G.
2.18 Let G = GL2 (F), where F is a field. Let V = F2 , with standard basis
e1 , e2 of column vectors. Describe stabG (e1 ) and stabG (e2 ). Prove that for
all v ∈ F2 \ {0}, the stabilizer stabG (v) is conjugate in G to stabG (e1 ).
2.19 Consider the normal subgroup N = {(1), (1 2)(3 4), (1 3)(2 4), (1 4)(2 3)}
of G = S4 .
(a) Show that G acts on N \ {1} by conjugation.

(b) Prove that the homomorphism ϕ : G → S3 corresponding to this
action is surjective with Ker ϕ = N .
(c) Prove that G/N ∼ = S3 .
2.20 Let G ≤ Sn such that G ∼

= Q8 . Prove that n ≥ 8.
2.21 Let G be a finite group and H < G a proper subgroup. Let [G : H] =
r > 1.
(a) Prove that for all x ∈ G, there exists 0 < k ≤ r such that xk ∈ H.
(b) Prove that if G is simple, then |G| divides r!.
179
2.22 Let n ≥ 5. Let H be a proper subgroup of An . Prove that [An : H] ≥ n.

2.23 Let G be a finite group and let p be the smallest prime divisor of |G|.
Suppose that there is a subgroup H < G such that [G : H] = p. Prove that
H E G.
2.24 (On inclusions between symmetric and alternating groups.)
(a) Let n ≥ 1. Prove that Sn is isomorphic to a subgroup of Sn+1 .

(b) Let n ≥ 1. Prove that An is isomorphic to a subgroup of An+1 .
(c) Let n ≥ 3. Prove that Sn is isomorphic to a subgroup of An+2 .
(d) Prove that S3 is not isomorphic to a subgroup of A4 .
(e) Let n ≥ 4. Prove that Sn is not isomorphic to a subgroup of An+1 .
Section 3
3.1 Let n ≥ 1. Prove the following:
(a) Z(GLn (q)) = {λIn : λ ∈ F×

q }, the subgroup of scalar matrices in
GLn (q).
(b) Z(SLn (q)) = {λIn : λ ∈ F× n
q , λ = 1}, the subgroup of scalar matri-
ces in SLn (q).
(Hint: For 1 ≤ i, j ≤ n; let Ei,j be the (n × n) matrix with 1 as the (i, j)

entry (row i, column j) and zeroes elsewhere. For all 1 ≤ i, j ≤ n with i 6= j,
we have In + Ei,j ∈ SLn (q). Thus if a matrix commutes with every element of
SLn (q), it will commute with Ei,j for all i 6= j. Use this fact to show that if a
matrix commutes with every element of SLn (q), it must be a scalar matrix.)
3.2 Let p(t) ∈ F[t] be a polynomial of degree 2 or degree 3. Prove that p(t)
is an irreducible over F if and only if p(t) has no root in F. Find an example
of a polynomial p(t) ∈ F[t] of degree 4 such that p(t) has no roots, but p(t)
is not irreducible.
3.3 Let q be a prime power. Prove that the number of irreducible polyno-
mials in Fq [t] of the form t2 + βt + α is equal to q(q − 1)/2. (Hint: count the
number of reducible polynomials.)
3.4 Determine the irreducible monic polynomials of degree 2 in F2 [t] and
F3 [t].
180
3.5 Let q be a prime power. Prove that the number of irreducible polyno-
mials in Fq [t] of the form t2 + βt + 1 is equal to (q − 1)/2 if q is odd, and q/2
if q is even.

α 0
3.6 Let G = GL2 (q) and g = ∈ G, where α, β ∈ F× q and α 6= β.
0 β
Describe CG (g) and calculate |CG (g)|.

α 1
3.7 Let G = GL2 (q) and g = ∈ G, where α ∈ F× q . Describe CG (g)
0 α
and calculate |CG (g)|.
3.8 Suppose that t2 + βt+ α ∈ Fq [t] be an irreducible polynomial over Fq .
0 −α
Let G = GL2 (q) and g = ∈ G.
1 −β
(a) Show that
CG (g) = {λI2 + µg : λ, µ ∈ Fq and (λ, µ) 6= (0, 0)}
and conclude that |CG (g)| = (q − 1)(q + 1).

(b) Prove that C = {λI2 + µg : λ, µ ∈ Fq } equipped with matrix
multiplication and addition is a field with q 2 elements.
3.9 Let p(t) ∈ F[t] and write p(t) = tn + cn−1 tn−1 + · · · + c1 t + c0 , where
ci ∈ F. Define  
0 0 · · · 0 −c0
1 0
 · · · 0 −c1  
A = 0 1
 · · · 0 −c2  .
 .. .. . . .. 
. . . . 
0 0 · · · 1 −cn−1
(a) Prove that the characteristic polynomial of A is equal to p(t).

(b) Prove that the minimal polynomial of A is equal to p(t). (Hint:
Let e1 , . . . , en be the standard basis of column vectors. Show first
that for any polynomial q(t) of degree < n, we have q(A)e1 6= 0.)
3.10 Describe representatives for the conjugacy classes in GL2 (3), and de-
termine the size of each conjugacy class. (There are a total of 8 conjugacy
classes.)
181
3.11 Describe representatives for the conjugacy classes of elements of order 3

in GL2 (7). (There are 5 conjugacy classes of elements of order 3 in GL2 (7).)
3.12 Describe representatives for the conjugacy classes of elements of order 3
in GL2 (5).
3.13 We have | GL3 (7)| = 26 · 34 · 73 · 19, so by Cauchy’s theorem GL3 (7)
contains elements of order 19. Find representatives for conjugacy classes of
elements of order 19 in GL3 (7). (There are 6 classes. Use the factorization
t19 − 1 = (t − 1)(t3 + 2t − 1)(t3 + 3t2 + 3t − 1)(t3 + 4t2 + t − 1)(t3 + 4t2 + 4t −
1)(t3 + 5t2 − 1)(t3 + 6t2 + 3t − 1) in F7 [t], which you do not need to verify.)
3.14 Let F be a field of characteristic 6= 2. Let A ∈ SL2 (F) be such that A
has order 2. Then A = −I2 .
3.15 Let m, n > 1 be integers.
(a) Find elements x, y ∈ PSL2 (C) = SL2 (C)/{±I2 } such that |x| = m,
|y| = n, and |xy| = ∞. (Hint: Imitate the proof of Theorem 3.22,
using primitive roots of unity in C.)
(b) Find x, y ∈ PSL2 (C) such that |x| = m, |y| = ∞, |xy| = n.
3.16 Let m > 1 be an integer.
(a) Find matrices x, y ∈ GL2 (C) such that |x| = m, |y| = ∞, and
|xy| = ∞.
(b) Find matrices x, y ∈ GL2 (C) such that |x| = ∞, |y| = ∞, and
|xy| = m.
3.17 Let F be a field. Prove that GL2 (F) is generated

by the set of trans-
λ 0
vections, together with matrices of the form , where λ ∈ F× .
0 1
3.18 Let A ∈ GLn (F), and let V = Fn . Suppose that every v ∈ V \ {0} is an
eigenvector of A. Prove that A must be a scalar matrix.
3.19 Prove that PGL2 (3) ∼ = S4 and PSL2 (3) ∼
= A4 . (Hint: Consider the action
of GL2 (3) on the set of 1-dimensional subspaces of F23 .)
182
Section 4
4.1 Let G be a group and x, y, z ∈ G. Prove the following identities:
(a) [x, y]−1 = [y, x].

(b) xy = yx[x, y].
(c) [xy, z] = [x, z]y [y, z].
(d) [x, yz] = [x, z][x, y]z .
(e) ϕ([x, y]) = [ϕ(x), ϕ(y)] for any homomorphism ϕ : G → H.
4.2 Let x, y ∈ G.
(a) Suppose that x commutes with [x, y]. Prove that [xn , y] = [x, y]n
for all n ∈ Z≥0 .
(b) Suppose that both x and y commute with [x, y]. Prove that
n(n−1)
(xy)n = xn y n [y, x] 2 for all n ∈ Z≥0 .
4.3 Let G = Q8 . Determine [G, G].

4.4 Let G = D2n . Determine [G, G].
4.5 Let p be a prime and let G be a non-abelian p-group of order p3 . Show
that [G, G] has order p and [G, G] = Z(G).
4.6 Let G = Sn , where n ≥ 3. Prove that [G, G] = An .
4.7 Let S be a subset of G.
(a) Let φ : G → H be a homomorphism. Prove that hφ(S)i = φ(hSi).

(b) Let g ∈ G. Prove that hSig = hS g i.
(c) Prove that CG (hSi) = CG (S).
4.8 Let G be a group and suppose that G = hSi for some S ⊆ G.
(a) Prove the following:
[G, G] = h[x, y]g : x, y ∈ S and g ∈ Gi.
(b) Find an example where [G, G] is larger than h[x, y] : x, y ∈ Si.

(Hint: Sn = h(1 2), (1 2 · · · n)i.)
183
4.9 Let G be a group and H, K ≤ G. Prove that [H, K] E hH, Ki. (Hint:
Use Exercise 4.1 (c).)
4.10 Let G be a group and H, K E G. Prove that [H, K] ≤ H ∩ K.
4.11 (Examples of solvable groups.)
(a) Prove that all dihedral groups are solvable (including the infinite
dihedral group D∞ ).
(b) Let p be a prime. Prove that any finite p-group is solvable.
(c) Prove that Q8 is solvable.
(d) Prove that S3 is solvable.
4.12 Let G be a group and let H, K ≤ G be solvable subgroups such that

H E G. Prove that HK is solvable.
4.13 Let F be a field such that |F| ≥ 3. Prove that [GL2 (F), GL2 (F)] =
SL2 (F).
4.14 We have GL2 (2) = SL2 (2). Determine [GL2 (2), GL2 (2)].
4.15 Let G = Cpk , where p is a prime. Show that G has only one composition
series.
4.16 Let G = Cpq , where p and q are distinct primes. How many composition
series does G have?
4.17 Show that Z does not admit a composition series.
4.18 (Dedekind’s rule) Let G be a group. Let A, B, C ≤ G be such that
B ≤ A. Prove that A ∩ BC = B(A ∩ C).
4.19 Let G be a non-trivial group.
(a) Let G be a finite abelian group. Prove that G has a composition

series where each quotient is cyclic of prime order.
(b) Suppose that G is abelian. Prove that G has a composition series
if and only if G is finite.
(c) Suppose that G is solvable. Prove that G has a composition series
if and only if G is finite.
4.20 Let G be a finite group. Prove that |xG | ≤ |[G, G]| for all x ∈ G.
184
4.21 Let F be a field, and let B be the set of upper triangular matrices in
GL2 (F):
λ ζ ×
B= : λ, µ ∈ F , ζ ∈ F .
0 µ
(a) Prove that B ≤ GL2 (F).

(b) Suppose that |F| > 2. Prove that [B, B] = U , where

1 ζ
U= :ζ∈F .
0 1
What happens when |F| = 2?

(c) Suppose that |F| > 2. Prove that B is solvable but not nilpotent.
4.22 Let G be a nilpotent group and let N E G such that N 6= {1}. Prove
that [N, G] N .
4.23 Let G be a nilpotent group and H be a proper subgroup of G. Then
H NG (H). (Hint: Show that there exists k ≥ 1 such that γk (G) 6≤ H and
γk+1 (G) ≤ H.)
(a) Prove that Z k (G) char G for all k ≥ 0.

(b) Suppose that γc+1 (G) = {1} for some c ≥ 1. Prove that γc−i+1 (G) ≤
Z i (G) for all 1 ≤ i ≤ c. (In particular Z c (G) = G.)
(c) Suppose that Z c (G) = G for some c ≥ 1. Prove that γc−i+1 (G) ≤
Z i (G) for all 1 ≤ i ≤ c. (In particular γc+1 (G) = {1}.)
185
(a) (Hall–Witt identity) Let x, y, z ∈ G. Prove the following identity:
[x, y −1 , z]y [y, z −1 , x]z [z, x−1 , y]x = 1
(b) (Three subgroups lemma) Let H, K, L ≤ G. Suppose that two of

the following higher commutator subgroups are trivial:
[H, K, L], [K, L, H], [L, H, K].
Prove that [H, K, L] = [K, L, H] = [L, H, K] = {1}.

(c) Let N be a normal subgroup. Suppose that two of the following
higher commutor subgroups are contained in N :
[H, K, L], [K, L, H], [L, H, K].
Prove that all three of them are contained in N .

(d) Let i, j ≥ 1. Prove that [γi (G), γj (G)] ≤ γi+j (G). (Hint: Proceed
by induction on i, and use (c).)
(e) Prove that G(k) ≤ γ2k (G) for all k ≥ 0.
4.26 Let G be a group and k ≥ 2. Prove that
γk (G) = h[x1 , x2 , . . . , xk ] : x1 , x2 , . . . , xk ∈ Gi.
4.27 Let G = S3 .
(a) Find elements x, y, z ∈ G such that [[x, y], z] 6= [x, [y, z]].
(b) Find subgroups H, K, L ≤ G such that [[H, K], L] 6= [H, [K, L]].
4.28 Let G be a group and H, K, L E G. Prove that [HK, L] = [H, L][K, L]

and [H, KL] = [H, K][H, L].
186
4.29 Let G be a finite group.
(a) Prove that F (G) char G.

(b) Prove that if K E G, then F (K) ≤ F (G).
(c) Find an example where K ≤ G and F (K) 6≤ F (G).
(d) Suppose that G is solvable and G 6= {1}. Prove that F (G) 6= {1}.
Section 5
ci := {1}×· · ·×{1}×Gi ×{1}×· · ·×{1}.
5.1 Let G1 , . . ., Gn be groups. Let G
Then Gi E G1 × · · · × Gn and
ci ∼
(G1 × · · · × Gn )/G = G1 × · · · × Gi−1 × Gi+1 × · · · × Gn .
5.2 Let G be a group and let H, K E G. Suppose that H ∩ K = {1}. Then

hk = kh for all h ∈ H and k ∈ K.
5.3 Let H and K be groups. Then for any A ≤ H and B ≤ K, the product
A × B is a subgroup of H × K.
5.4 Let H and K be finite groups.
(a) Suppose that gcd(|H|, |K|) = 1. Prove that if N ≤ H × K, then

N = A × B for some A ≤ H and B ≤ K.
(b) Suppose that H 6= {1}. Prove that H × H has a subgroup which
is not of the form A × B.
5.5 Let G = H × K. Define the projection maps πH : G → H and πK : G →

K by πH (h, k) = h and πK (h, k) = k for all h ∈ H and k ∈ K.
(a) Prove that πH and πK are surjective homomorphisms.

(b) Prove that Ker πH ∼
= K and Ker πK ∼ = H.
(c) For any subgroup X ≤ G, prove that X ≤ πH (X) × πK (X). Find
an example where X 6= πH (X) × πK (X).
(d) Generalize (a) and (b) for G = H1 × H2 × · · · × Hk , where H1 , H2 ,
. . ., Hk are groups and k ≥ 2.
187
5.6 Let G and H be groups. Prove the following statements:
(a) (G × H)(k) = G(k) × H (k) for all k ≥ 0.

(b) γk (G × H) = γk (G) × γk (H) for all k ≥ 1.
(c) Z k (G × H) = Z k (G) × Z k (H) for all k ≥ 0.
(d) The direct product G × H is abelian if and only if G and H are
abelian.
(e) The direct product G × H is solvable if and only if G and H are
solvable.
(f) The direct product G × H is nilpotent if and only if G and H are
nilpotent.
5.7 Let G and H be finite groups. Prove that F (G × H) = F (G) × F (H).

5.8 Prove that (Q, +) is not isomorphic to a direct product of cyclic groups.
5.9 Let G = hx1 , . . . , xk i be a finitely generated abelian group. Prove that
any subgroup H ≤ G is generated by ≤ k elements. (Hint: Proceed by
induction and consider H ∩ hx2 , . . . , xk i.)
5.10 Let G be an abelian group. Let tor(G) = {g ∈ G : |g| < ∞}, the set of
elements of finite order in G.
(a) Prove that tor(G) is a subgroup of G.

(b) Prove that G/ tor(G) has no non-trivial element of finite order.
(c) Let G and H be abelian groups. If G ∼ = H, prove that tor(G) ∼ =
∼
tor(H) and G/ tor(G) = H/ tor(H).
(d) Suppose that G = Cn1 ×· · ·×Cnk ×Zr , where k ≥ 0, n1 , . . . , nk > 1,
and r ≥ 0. Prove that
tor(G) ∼
= Cn1 × · · · × Cnk ,
∼
G/ tor(G) = Zr .
(e) Find a non-abelian group X such that tor(X) is not a subgroup of

X.
188
5.11 Let (G, +) be an abelian group.
(a) Prove that for all n > 0, the set nG = {ng : g ∈ G} is a subgroup
of G.
(b) If G and H are abelian groups such that G ∼= H, prove that nG ∼ =
∼
nH and G/nG = H/nH.
(c) Let r, s ≥ 0. Prove that Zr ∼= Zs if and only if r = s. (Hint: use
part (b).)
5.12 Let (G, +) be an abelian group. Recall from Exercise 1.16 that for all
m ∈ Z, the set G(m) = {g ∈ G : mg = 0} is a subgroup of G.
(a) Let (G1 , +), . . ., (Gt , +) be abelian groups and G = G1 × · · · × Gt .

Prove that G(m) = G1 (m) × · · · × Gt (m) for all m ∈ Z.
(b) If m | n, prove that G(m) ≤ G(n).
(c) Let G be cyclic of order n. Prove that G/G(d) ∼
= Cn/d .
(d) Let G be cyclic of order n and let d | d0 | n. Prove that G(d)/G(d0 ) ∼
=
Cd/d0 .
5.13 Let G the following subgroup of GL2 (Q):

1 1 2 0
G= , .
0 1 0 1
Let U be the following subgroup of GL2 (Q):

1 x
U= :x∈Q .
0 1
Prove that H = G ∩ U is not finitely generated. (This example shows that a

subgroup of a finitely generated group is not necessarily finitely generated.)
5.14 Let n > 0 be an integer and write the prime factorization of n as n =
pk11 · · · pkt t , where pi are distinct primes and ki > 0 for all 1 ≤ i ≤ t. Prove
that up to isomorphism, the number of finite abelian groups of order n is
p(k1 ) · · · p(kt ), where p(k) denotes the number of partitions of an integer
k > 0. (For example p(4) = 5, since the partitions of 4 are 4, 1 + 3, 2 + 2,
1 + 1 + 2, and 1 + 1 + 1 + 1. Other values: p(1) = 1, p(2) = 2, p(3) = 3,
p(4) = 5, p(5) = 6, p(6) = 11.)
189
5.15 Let G be a finite abelian p-group, say G ∼

= Cpα1 × · · · × Cpαt with
α1 ≥ · · · ≥ αt > 0.
(a) Prove that for H ≤ G, we have H ∼

= Cpβ1 × · · · × Cpβt for some
β1 ≥ · · · ≥ βt ≥ 0.
(b) Prove that in (a), we have βi ≤ αi for all 1 ≤ i ≤ t.
(c) Prove that for G/H, we have G/H ∼ = Cpγ1 × · · · × Cpγt for some
γ1 ≥ · · · ≥ γt ≥ 0.
(d) Prove that in (c), we have γi ≤ αi for all 1 ≤ i ≤ t.
5.16 Let G be a finite abelian group and let H ≤ G be a subgroup. Prove

that there exists K ≤ G such that K ∼
= G/H. (Hint: Use exercise 5.15).
5.17 Let G = Q8 .
(a) Prove that G/Z(G) ∼

= C2 × C2 .
(b) Prove that G has no subgroup isomorphic to C2 × C2 .
5.18 Prove that there does not exist a group G with G/Z(G) ∼ = Q8 . (Hint:
Q8 has a subgroup of order 2, which is contained in every nontrivial subgroup
of Q8 .)
5.19 Let G and H be groups. Prove that if G ∼
= H, then Aut(G) ∼
= Aut(H).
Find an example where Aut(G) ∼
= Aut(H), but G 6∼
= H.
5.20 Prove that Aut(C2 × C2 ) ∼
= S3 .
5.21 Let G and H be finite groups.
(a) Suppose that gcd(|G|, |H|) = 1. Prove that Aut(G×H) ∼

= Aut(G)×
Aut(H).
(b) Prove that Aut(G)×Aut(H) is isomorphic to a subgroup of Aut(G×
H).
(c) Find an example where Aut(G × H) ∼6= Aut(G) × Aut(H).
5.22 Prove that Aut(S4 ) ∼

= S4 .
5.23 Let p be a prime and let G be a finite group. Prove that G is an
elementary abelian p-group if and only if G is abelian and xp = 1 for all
x ∈ G.
190
5.24 Let (G, +) be an abelian group. Suppose that there exists a prime p
such that pg = 0 for all g ∈ G.
(a) For x ∈ Z/pZ, define x · g = xg for all g ∈ G. Prove that this

makes G into a vector space over Fp = Z/pZ.
(b) Let V be a vector space over Fp . Show that for (G, +) = (V, +),
we have pg = 0 for all g ∈ G.
(c) Let ϕ : G → G be map. Show that ϕ is a homomorphism if and
only if ϕ is a Fp -linear map.
5.25 When is the order of Inn(G) a prime number?

5.26 Let H and K be groups, and let G = H oψ K be an external semidirect
product, where ψ : K → Aut(H) is a homomorphism. Let H b and K b be as
in Theorem 5.36. Prove that the following statements are equivalent:
(i) G = H × K.
(ii) ψ : K → Aut(H) is trivial.
(iii) K
b is a normal subgroup of G.
5.27 Let H and K be groups.
(a) Suppose that H ∼ = H 0 and K ∼ = K 0 . Prove that any external

semidirect product H oψ K is isomorphic to H 0 oψ0 K 0 for some
homomorphism ψ 0 : K 0 → Aut(H 0 ).
(b) Let ψ, ψ 0 : K → Aut(H) be homomorphisms. Suppose that ψ 0 =
ψϕ for some ϕ ∈ Aut(K). Prove that H oψ K ∼
= H oψ0 K.
(c) Let ψ, ψ 0 : K → Aut(H) be homomorphisms. Suppose that there
exists φ ∈ Aut(H) such that ψ 0 (x) = φψ(x)φ−1 for all x ∈ K.
Prove that H oψ K ∼= H oψ0 K.
5.28 Let G be a finite group of order mn, where gcd(m, n) = 1. Suppose that
G contains a normal subgroup N of order n. Then N is the unique subgroup
of order n in G, and
N = {x ∈ G : xn = 1}.
5.29 Let G = D2n (dihedral group), where n > 0 is odd. Prove that all
elements of order 2 are conjugate in G.
5.30 Classify groups of order 30 up to isomorphism.
191
Section 6
6.1 Let G be a finite group and let p be a prime. We denote the set of Sylow
p-subgroups of G by Sylp (G).
(a) Let H ≤ G. Prove that
Sylp (H) ⊆ {P ∩ H : P ∈ Sylp (G)}.
(b) Find an example of a group G and H ≤ G such that
Sylp (H) $ {P ∩ H : P ∈ Sylp (G)}.
6.2 Let G be a finite group and let p be a prime.
(a) Let N E G. Prove that
Sylp (N ) = {P ∩ N : P ∈ Sylp (G)}.
(b) Let N E G. Prove that
Sylp (G/N ) = {P N/N : P ∈ Sylp (G)}.
6.3 Let G be a group of order |G| = 20 = 22 · 5.
(a) Show that G is isomorphic to some semidirect product C5 oψ P ,

where |P | = 22 .
(b) Classify semidirect products C5 oψ C4 , up to isomorphism. (There
are 3 in total.)
(c) Classify semidirect products C5 oψ (C2 × C2 ), up to isomorphism.
(There are 2 in total.)
6.4 (Examples of H ≤ G such that np (H) - np (G).)
(a) Let G = S5 and H = S4 . Show that n3 (H) = 4 and n3 (G) = 10.

(b) Let G = A5 . Show that there is a subgroup H < G such that
H∼= S3 . Prove that n2 (H) = 3 and n2 (G) = 5.
6.5 Let G and H be finite groups. Prove that np (G × H) = np (G)np (H).

6.6 Let p and q be distinct primes. Classify semidirect products of the form
Cq oψ (Cp × Cp × · · · × Cp ), up to isomorphism.
192
6.7 Let G be a finite group and let p be a prime.
(a) Suppose that a Sylow p-subgroup of G is cyclic. Prove that if

pk | |G|, then any two subgroups of order pk are conjugate in G.
(b) Find an example where pk | |G|, and G contains two subgroups of
order pk which are not conjugate in G.
6.8 Let G be a finite abelian group and H ≤ G. Prove that for the transfer
map v : G → H, we have v(g) = g [G:H] for all g ∈ G.
6.9 Let G be a group of squarefree order |G|. Prove that for all d | n, there
exists a subgroup of order d in G.
193

Topics in Algebra

Uploaded by

Copyright:

Available Formats

Topics in Algebra

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Topics in Algebra

Uploaded by

Copyright:

Available Formats

Topics in Algebra

Lecture notes, Autumn 2021

2 Symmetric and alternating Groups 51

5 Constructing groups 108

6 Sylow theory 136

(i) Associativity: (a ∗ b) ∗ c = a ∗ (b ∗ c) for all a, b, c ∈ G

(ii) There exists an identity element e ∈ G such that e ∗ a = a ∗ e = a for

(iii) For all a ∈ G, there exists an inverse b ∈ G such that a ∗ b = b ∗ a = e.

Remark 1.2. (a) Usually we will denote a group (G, ∗) just by G.

(b) An identity element e ∈ G is unique. For if e0 is another identity element,

(d) In general we use multiplicative notation. That is, the multiplication

(e) For all a, b ∈ G we have (ab)−1 = b−1 a−1 .

(f) Every group G satisfies the following cancellation laws:

Definition 1.3. Let G be a group. Two elements a, b ∈ G are said to

If (G, ∗) is an abelian group, we will sometimes use additive notation:

Example 1.4. Examples of abelian groups:

(i) (R, +): identity element is 0 ∈ R, the inverse of x ∈ R is −x ∈ R.

(ii) (R \ {0}, ·): identity element is 1 ∈ R, inverse of x ∈ R \ {0} is 1/x.

a+b=a+b for all a, b ∈ Z

(These operations are well-defined: if a = a0 and b = b0 , then a + b = a0 + b0

As a consequence of this formula, for n, m > 0 with gcd(m, n) = 1, we have

• (R, +) is an abelian group;

• Multiplication · is an associative binary operation on R with identity

• Multiplication is distributive over +, in other words (a + b)c = ac + bc

A ring R is commutative if ab = ba for all a, b ∈ R. For a ring R, we

R× = {x ∈ R : xy = yx = 1 for some y ∈ R}.

Then (R× , ·) is a group, and called the group of units of R. A field is a

Definition 1.8 (Symmetric groups). Let Ω be a set. The symmetric group

and ◦ is composition of functions. The identity element is the identity map

If Ω = {ω1 , . . . , ωn }, then for g ∈ Sym(Ω) we denote

where g(ωi ) = ωi0 for 1 ≤ i ≤ n.

σ(it ) = it+1 for all 1 ≤ t < k,

We denote σ = (i1 i2 · · · ik ). A 2-cycle is called a transposition.

S3 = {(1), (1 2), (1 3), (2 3), (1 2 3), (1 3 2)}.

With the group operation we have defined, we have e.g. (1 2)(1 3) = (1 3 2)

as the identity element.

Example 1.14. Let p be a prime. An important example of a field is

Suppose that (H, ?) is a group such that G ∼ = H, and let ϕ : G → H

Example 1.17. An isomorphism ϕ : G → G is called an automorphism.

• Every X ∈ G is a group with property P ;

• If G is a group with property P , then G ∼

Here are two fundamental classification problems in finite group theory.

(ii) If X, Y ∈ Gn and X 6= Y , then X ∼

(iii) If G is a group of order n, then G ∼

The cardinality of a set Gn as in Problem 1.21 is called the group number

gnu(2048) > 1774274116992170 ≈ 1.77 · 1015

n gnu(n) n gnu(n) n gnu(n) n gnu(n) n gnu(n)

Table 1: Values of gnu(n) for 0 ≤ n ≤ 64.

1.2 Order of an element

are all equal, and thus can be denoted by a1 a2 a3 a4 without ambiguity.

Definition 1.25. Let G be a group and let g ∈ G. The order |g| of g is

(ii) Let G = (Z/nZ, +). Then 1 has order n in G.

(iii) Let k ≥ 0 and suppose that k | n. Then |g k | = nk .

(iv) {g k : k ∈ Z} = {1, g, . . . , g n−1 }.

(ii) First note that

(g k )n/ gcd(k,n) = (g n )k/ gcd(k,n) = 1.

Next consider d ≥ 0 such that (g k )d = 1. Then by (i) we have n | kd,

(iii) Immediate from (ii).

(iv) By (i), we have g i = g j if and only if i ≡ j mod n, from which the

Lemma 1.28. Let G be a group and let x, y ∈ G. Then |xy| = |yx|.

Definition 1.35. Let H be a subgroup of a group G. If H G, we say that