MAT301

Groups and Symmetries

Gaurav Patil and Shuyang Shen

Fall 2024
These lecture notes were written and edited over iterations of group theory courses,
including MAT301 in Summer 2020 (Matt Olechnowicz and Shuyang Shen) and MAT301
in Fall 2024 (Gaurav Patil and Shuyang Shen).
We would like to thank the many students who suggested corrections to the text, especially
Callum Cassidy-Nolan, Paola Driza, Rachel Leggett, Michael Luan, Amy Mann, Bayan
Mehr, and Kelly Qiu.

i
Contents

-1 Introduction 1
-1.1 Group theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
-1.1.1 Polyhedra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
-1.1.2 Colouring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
-1.1.3 Polynomial roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
-1.1.4 Cryptography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

0 Preliminaries 6
0.1 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
0.2 Maps a.k.a. Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
0.3 Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
0.4 Integers, or: Rem(a)inders from Arithmetic . . . . . . . . . . . . . . . . . . . 13
0.4.1 Division . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
0.4.2 Congruence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
0.5 Complex numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
0.6 Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1 Groups and subgroups 20

1.1 Binary operations and groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.1.1 Visualizing binary operations . . . . . . . . . . . . . . . . . . . . . . . 21
1.1.2 Associativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.1.3 Operations table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.1.4 Definition of a group . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.1.5 Immediate consequences . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.1.6 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2.1 Basic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2.2 Groups of integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2.3 Matrix groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.2.4 Trivial group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.2.5 Cyclic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.2.6 Dihedral groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.2.7 Symmetric groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
ii
 Contents

1.3 Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.3.1 Basic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.3.2 Matrix groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.3.3 Cyclic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.3.4 Dihedral groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.3.5 Permutation groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.3.6 Generating new subgroups . . . . . . . . . . . . . . . . . . . . . . . . 34
1.4 Guises of the same group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.5 Group Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.5.1 Group actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.5.2 Concrete examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.5.3 Abstract examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
1.5.4 New actions from old . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
1.6 Orders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.6.1 Order of a group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.6.2 Order of an element . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

2 Families of groups 51
2.1 Congruence groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.1.1 Z/nZ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.1.2 U(n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.2 Cyclic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.2.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.2.2 Fundamental Theorem of Cyclic Groups . . . . . . . . . . . . . . . . 58
2.2.3 Cyclicity of U(n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.3 Dihedral groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.3.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.3.2 Order of a dihedral group . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.3.3 Subgroups of dihedral groups . . . . . . . . . . . . . . . . . . . . . . 66
2.3.4 The center of a group . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.3.5 Generators and relations . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.3.6 Infinite dihedral group . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.4 Symmetric groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.4.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.4.2 Order of the symmetric group . . . . . . . . . . . . . . . . . . . . . . 72
2.4.3 Composing and inverting permutations . . . . . . . . . . . . . . . . . 73
2.4.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.4.5 Cayley’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.4.6 Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.4.7 Inverting and composing cycles . . . . . . . . . . . . . . . . . . . . . 79
2.4.8 Permutations in terms of cycles . . . . . . . . . . . . . . . . . . . . . . 80
2.4.9 Cycle Decomposition Theorem . . . . . . . . . . . . . . . . . . . . . . 82
2.4.10 Cycle type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

iii
 Contents

2.4.11 The order of a permutation . . . . . . . . . . . . . . . . . . . . . . . . 88

2.4.12 What generates Sn ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

3 Quotients and Morphisms 92

3.1 Normality and Quotients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.1.1 Recall... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.1.2 First examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.1.3 When is (aH)(bH) a coset of H? . . . . . . . . . . . . . . . . . . . . . . 95
3.1.4 Examples of normal subgroups . . . . . . . . . . . . . . . . . . . . . . 98
3.1.5 Lingering questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.1.6 Summary of normality . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

iv
Chapter -1

Introduction

Group theory is the study of symmetry. Broadly speaking, a symmetry is an invertible

transformation of some object that preserves the object. In other words, if you can do
something to an object and leave it looking the same (or similar), you’ve found a symmetry.
Nobody’s perfect, but if your face was perfectly symmetrical, it would look the same to
you in the mirror as it looks to other people who can see you directly. The mirror shows
you a reflection of your face and leaves it looking the same—that’s a “symmetry” of your
face.
Similarly, take a regular pentagon and rotate it about its center by 72◦ . The shape of this
pentagon remains unchanged because it has rotational symmetry.
In nature, symmetries often manifest as reflectional, rotational, or translational.
Symmetries are...
• Everywhere. Symmetries show up in everything, from the shapes of galaxies all the
way down to the arrangements of fundamental particles.
• Pretty. Symmetries are visually pleasing, and we have a lot of art featuring different
forms of symmetry.
• Important. Humans use pattern recognition to understand the world, and symme-
try is one of those key “patterns”. We can study symmetries of objects to better
understand those objects and their properties.
... Which brings us back to group theory!

1
 Chapter -1. Introduction

-1.1 Group theory

“The theory of groups is a branch of mathematics in which one does something to
something and then compares the results with the result of doing the same thing to
something else, or [doing] something else to the same thing.”
—James R. Newman
Before we talk about what groups are, we want to first go over some problems in which
group theory shows up, and impress upon you the sheer applicability of group theory.

-1.1.1 Polyhedra
Consider a cube. We cannot easily “reflect” it in three-dimensional space without the aid
of a mirror, but we may rotate it in a few different ways while maintaining its shape:
• Rotating by 90◦ , 180◦ , or 270◦ about an axis through the midpoints of two opposing
edges.
• Rotating by 180◦ about an axis through the midpoints of two opposing edges.
• Rotating by 120◦ or 240◦ about a “grand diagonal”, an axis through two diagonally
opposing vertices.
You can see a visualization here. Observe that with each type of rotation, the faces of the
cube are permuted in different ways! Something that group theory studies is what these
symmetries each change and what they preserve, as well as how they interact with each
other.

-1.1.2 Colouring
Suppose we have a bunch of beads of various colors. How many necklaces can be made
out of those beads? Basic permutation results aren’t enough here: the “starting bead”
doesn’t matter but the order of the beads matter! This is a type of symmetry—“rotating”
the necklace preserves the order of the beads. Similarly, we may also “flip” the necklace
and introduce another type of symmetry.

2
 Chapter -1. Introduction

Example -1.1.1 — For a more tractable example, suppose we want to make a necklace
out of four beads that are red, yellow, green, and blue respectively. The only distinct
necklaces are the following:

Simply listing the possibilities becomes infeasible when we introduce more colors and/or
beads. Group theory to the rescue: Burnside’s lemma can be used to count items featuring
symmetries like this by counting everything equivalent up to symmetry as the same. We
will learn about this sometime in November.

-1.1.3 Polynomial roots

Remember the quadratic formula?
√ If you have a quadratic polynomial ax2 + bx + c, its
−b ± b2 − 4ac
roots are given by x = . Everyone learns this in high school.
2a
Now what if we have a cubic polynomial ax3 + bx2 + cx + d?
Well, there is also a formula for its roots: they are

b2 − 3ac
 
1 k
rk = − b+ξ C+ (k = 0, 1, 2)
3a ξk C

where s p
3 b2 − 3ac ± (b2 − 3ac)2 − 4(2b3 − 9abc + 27a2 d)3
C=
2
and √
−1 + 3i
ξ=
2
. . . you’re not expected to know this offhand.
And there is one for quartics as well:

. . . this is why they don’t teach it in high school.

3
 Chapter -1. Introduction

But mathematicians got those formulas in the 1500s, and they started looking at quintics.
They got stuck. For two hundred years. The question becomes: is the formula just really
complicated or is it straight up impossible?
Finally in 1799 Ruffini came up with a partial proof that yes, some quintic polynomials
just don’t have a solutions in nice formulas like those. Abel finished up the proof some
years after. Later Galois and Cayley developed criteria so that we know precisely which
polynomials are solvable and which ones aren’t. The tools they developed along the
way evolved into what we now call Galois theory, which, among other things, is used to
investigate the behaviour and relationship of roots of polynomials.
This will be covered in more detail in MAT401, so that’s something to look forward to.

-1.1.4 Cryptography
The art of secret messages is another inspiration to the formalization we see in group
theory.
The main idea of secret messages is to convert a message to gibberish is a reversible way.
The idea is someone who knows the secret, should be able to make sense of the gibberish.
But someone who does not know the secret should not be able to decipher the gibberish.
Ideally, you want the ability to have secret communication with many people (often people
who don’t necessarily know you) without setting up a system of secrets each time. In other
words, there have to be many possible secrets. One should be unable to narrow down
which secrets a particular person might use. A secret for us is an identifier of the exact
process of gibberishizing you re using. You want random person to be unable to try out all
secrets on your gibberish and find out your message.
Thus, we take a large ‘group’ or set of reversible transformations.

Example -1.1.2 — You may have seen the RSA cryptosystem in MAT246 (and a much
simpler proof of its mechanism can be had with group theory!). The idea is:
• Take a large modulus m which is a product of two primes p and q. Publish m
while keeping p and q secret.
• Choose an encryption key e and give it to the person with whom you wish to
communicate.
• Compute the decryption key d using e, p, q and use it to decrypt the messages
encrypted with e.
Given a fixed m, we can pick many different encryption keys e1 , . . . , en to give to dif-
ferent people—which improves the safety of communication while keeping decryption
nice and simple. These valid encryption keys are derived from the structure of the
group behind the RSA cryptosystem, and we will talk about this group in a few weeks.

4
 Chapter -1. Introduction

Exercise 1. What does the set of all valid encryption keys look like here?

Such a system allows for multiple layers of security. For example, healthcare data may
be encrypted multiple times while being sent to different agencies, so that identifiable
information is obscured and at each step, an agency only knows information essential to
their operation. When the processed data is sent all the way back to the hospital, we may
apply decryption schemes along the way and retrieve information for each patient.

5
Chapter 0

Preliminaries

0.1 Sets
A set is a collection of things under consideration, and a subset is a collection of some of
those things—including potentially all of them as well as none of them. To write down a
set, you can either write down all its elements:

{Buddy, Rex, Fido}

—or you can specify it using set-builder notation:

{x : I have a dog named x}.

The squiggles on either side are called (curly) braces.

The empty set is the set with no elements. Instead of writing it as {}, the empty set is denoted

∅
Some common sets we will be making use of are:
• N = {1, 2, 3, . . . }: set of natural numbers (sometimes including 0),
• Z = {. . . , −1, 0, 1, 2, . . . }: set of integers,
p
• Q= q
: p, q ∈ Z, q ̸= 0 : set of rational numbers,

• R: set of real numbers, and

• C: set of complex numbers.

6
 Chapter 0. Preliminaries

Given two subsets A and B of a set X, here are the most important ways to form new
subsets of X.
The union of two sets is the set of elements contained in at least one of them. That is,

A ∪ B = {x : x ∈ A or x ∈ B (or both!)}.

The intersection of two sets is the set of elements contained in both of them. That is,

A ∩ B = {x : x ∈ A and x ∈ B}.

Two sets are disjoint if their intersection is empty. Visually, disjoint sets don’t overlap.

A∪B X A∩B X
A B A B

The complement of a subset is the set of things not in it. More precisely,

A or Ac = {x ∈ X : x ̸∈ A}.

The relative complement of one set, A, in another set, B, is the set of things in B that are not
in A. That is,
B − A or B \ A = {x ∈ B : x ̸∈ A}.

A X A−B X
A B A B

Exercise 2. Write A ∩ Ac = ∅ in plain language (using the word “disjoint”), and then
prove it.

Exercise 3. Let A ⊆ X. Show that the “complement of A” is the same thing as the
“relative complement of A in X”.

Exercise 4. Show that B \ A = B ∩ Ac .

7
 Chapter 0. Preliminaries

A partition of a set is a collection of subsets that divide it up. Formally, Ai ⊆ X for all i, and
[
X= Ai ,
i

and Ai ∩ Aj = ∅ for all i ̸= j, then we say the Ai ’s partition X (also: form a partition of X).

Example 0.1.1 — The sets {1, 2} and {3, 4} partition the set {1, 2, 3, 4}, but the sets
{1, 2, 3} and {1, 2, 4} do not.

Example 0.1.2 — The intervals [k, k + 1) partition R.

... ...

−3 −2 −1 0 1 2 3

The Cartesian product of two sets A and B is the set of ordered pairs (a, b) where a is in A
and b is in B. That is,
A × B = {(a, b) : a ∈ A, b ∈ B}.

Example 0.1.3 — The “xy-plane” is the Cartesian product of R with itself.

0.2 Maps a.k.a. Functions

A function is a rule for associating a unique output to every valid input. The set of inputs
is the domain and the set of outputs (whether or not all are possible) is the codomain. If
f(x) = y we say y is the image of x under f, or the value of f at x, depending on what we
want to emphasize.
To write down a function, you can describe it in words—
“Let f be the squaring map on R.”
—or write down a formula—

f : R → R,
f(x) = x2 .

A useful “anonymous” shorthand is x 7→ x2 .

When the domain is finite, you can use two-line notation: write the elements of the domain
in a row and write their images underneath.

8
 Chapter 0. Preliminaries

Example 0.2.1 — The squaring map on the set {0, 1, 2, 3} can be written as
 
0 1 2 3
0 1 4 9

in two-line notation.

Given sets X, Y and a map f : X → Y, here are the most important objects derived from f.
The image of a subset A ⊆ X under f is the set of all the values f(a) where a ranges just
over A.
f(A) = {y : y = f(a) for some a in A}.

The image of f means the image of X under f, denoted im f = f(X).

Example 0.2.2 — Let f(x) = x2 + 1 from R to R. Then im f = [1, ∞) while f([−2, 1)) =
[1, 5].

The graph of f is the set of pairs (x, f(x)) as x ranges over X, formally

Γ (f) = {(x, y) ∈ X × Y : f(x) = y}.

Exercise 5. Isn’t the graph of f just equal to X × f(X)? Explain.

f is injective or one-to-one (or even just 1-1) if it doesn’t send different inputs to the same
output.

Exercise 6. Suppose f is injective, and f(x) = f(y). What can you deduce about x and
y?

f is surjective or onto if the image equals the codomain. Surjectivity only makes sense if you
specify a codomain!

Exercise 7. Suppose f is surjective, and y is in Y. What can you deduce about f and
X?

f is bijective if it is both injective and surjective.

9
 Chapter 0. Preliminaries

Example 0.2.3 — Injective, surjective, and bijective functions R → R.

y y y

x x x

Given another map g : Y → Z, g composed with f (or g after f) is the map obtained by
applying f and then g. That is, (g ◦ f)(x) = g(f(x)).

Exercise 8.
X Y Z
f g z1
x1 y1
x2 z2
x3 y2 z3
y3 z4
x4 z5

Express the map g ◦ f in two-line notation.

Exercise 9. Show that the composition of two injective functions is injective. Do the
same with “injective” replaced by “surjective”.

An inverse of f is a function g : Y → X such that g ◦ f = idX and f ◦ g = idY . f has an inverse

if and only if f is bijective, in which case the inverse is denoted f−1 . Be careful not to confuse
this with the reciprocal of f—f−1 (x) ̸= f(x)−1 !
Let f : X −→ Y denote a function. For S ⊆ Y, we define the preimage of S under f to be

f−1 (S) := {x ∈ X : f(x) ∈ S}.

Take care not to confuse this notation with the inverse function f−1 : The preimage is a set,
while f−1 is a function. We may talk about the preimage of sets under any function, but
only bijective ones have an inverse.

Exercise 10. What are the preimages of each element in Z under g in the previous
example? What about g ◦ f?

10
 Chapter 0. Preliminaries

Exercise 11. Let S, T be subsets of Y. Show that if S ∩ T = ∅, then

f−1 (S) ∩ f−1 (T ) = ∅.

Exercise 12. Conclude that {f−1 ({y}) : y ∈ Y} is a partition of X.

f is a self-map if Y = X. That is, f maps X to itself. A bijective self-map is called a permutation.

Example 0.2.4 —
X X
f
a a
b b
c c
d d

In two-line notation, this map is

 
a b c d
.
d b a c

Probably the most important self-map is the identity map x 7→ x, sometimes explicitly
denoted id or idX .
Self-maps are interesting because they can be iterated. Given f : X → X, the nth iterate of f
is defined as f composed with itself n times, denoted fn . For convenience, we set f0 = idX .

Exercise 13. For the function f in the previous example, write out fn for n = 0, . . . , 6.
What do you notice?

Finally, f : X → X is an involution if it’s its own inverse. That is, f2 (x) = x.

0.3 Relations
Given a set X, a relation on X is, formally, a subset R of X × X (the set of pairs (x, y) with
x and y in X). For any x, y in X, we say x is related to y (but not necessarily vice versa) if
(x, y) is in R, denoted xRy.
R is reflexive if all elements are related to themselves. That is, xRx for all x.
R is symmetric if the relation goes both ways. That is, if xRy then yRx as well (for all x and
y).
11
 Chapter 0. Preliminaries

R is antisymmetric if no two distinct elements are mutually related. That is, if xRy and yRx,
then x = y.
R is transitive if you can “remove the middleman” in a chain of relations. That is, if xRy
and yRz, then xRz.
A relation that is reflexive, symmetric, and transitive is called an equivalence relation.
Equivalence relations are denoted ∼ instead of R.
“Our human condition is such that [the relation x loves y] is, alas, neither reflexive,
symmetric, nor transitive.”
—Seth Warner, Modern Algebra

Exercise 14.
Fill out the properties of the following relations.
x, y are people R? S? T?
“x loves y”
“x is aware of y”
“x and y were married at some point”
“x is an ancestor of y”
“x looks like y (Think about the Ship of Theseus paradox!)”
“x is not younger than y”
“x has been to the same school as y”
“x is born in the same year as y”

Exercise 15. When is a relation possibly symmetric, transitive, but not reflexive?

Given an equivalence relation ∼, an equivalence class is a complete set of elements that are
all related to one another.
The equivalence class of x is denoted [x] = {y ∈ X : x ∼ y} and every equivalence class has
this form.
Exercise 16. Show that any two elements in [x] are related.

The set of all equivalence classes—a set of sets—is denoted X/∼.

Exercise 17. Show that the equivalence classes partition X.

Exercise 18. Revisit Exercise 12 by showing that the relation “x ∼ y if f(x) = f(y)” is
an equivalence relation. What are the equivalence classes?

12
 Chapter 0. Preliminaries

0.4 Integers, or: Rem(a)inders from Arithmetic

0.4.1 Division
For any integer a any nonzero integer b, there exist unique integers q and r satisfying

a = bq + r and 0 ≤ r < |b|.

q is called the quotient and r is called the remainder.

To find q and r, it’s easiest to just actually divide a by b. That gives
a r
=q+ .
b b
If b > 0, then
r |b|
0≤ < = 1.
b b
Thus jak
q= (1)
b
i.e. q is a/b rounded down. Once you know q, finding r is easy.

Example 0.4.1 — To divide a = 42 by b = 9 with remainder, start by observing that

42/9 = 4.666... so q = 4. But 9 · 4 = 36, so r = 6. In other words,

42 = 4 · 9 + 6.

When r = 0, we say b divides a and write b | a. A number is prime if it has just two
divisors. Divisibility is a transitive and reflexive relation on Z. It is neither symmetric nor
antisymmetric, but if a | b and b | a then a = b or a = −b.
A number is prime if it has just two divisors.
A common divisor of two numbers is a number dividing them both. The greatest common
divisor or gcd of two numbers is just that—the biggest of the common divisors. The gcd
has the wonderful property that if c | a and c | b then c | gcd(a, b). Two numbers are
coprime if gcd(a, b) = 1.
Dually, a common multiple of two numbers is a number they both divide. The least common
multiple or lcm of two numbers is just that—the smallest of the common multiples. Like
the gcd, the lcm has the wonderful property that if a | m and b | m then lcm(a, b) | m.
The lcm and the gcd are related by the formula

gcd(a, b) · lcm(a, b) = |ab|.

13
 Chapter 0. Preliminaries

0.4.2 Congruence
Fix an integer m, called the modulus. Say a ≡ b (mod m) if and only if m | a − b.
Congruence modulo m is an equivalence relation on Z. [Check this!] When m is clear from
context, the equivalence class of a is denoted [a], while the set of equivalence classes is
variously denoted
Z/m or Z/(m) or Z/mZ or Zm . In this course, we use Z/mZ.
The set Z/mZ inherits addition, subtraction, and multiplication from Z, meaning that you
can add, subtract, and multiply equivalence classes (of the same modulus).
In other words, one defines
[a] + [b] = [a + b] and [a] · [b] = [ab]
and then checks that these operations are well-defined.

Example 0.4.2 — Z/24Z has twenty-four elements, [0], [1], [2], . . . , [23].

[12] + [15] = [27] = [3]

[12] − [15] = [−3] = [21]

Adding and subtracting modulo 24 is like reckoning with military time.

[3] · [10] = [30] = [6]

[7]2 = [7] · [7] = [49] = [1]

Multiplication doesn’t have such a nice interpretation.

0.5 Complex numbers

Complex numbers are numbers of the form z = x + iy where x and y are real numbers and i
satisfies i2 = −1.
x and y are called the real part and imaginary part of z and denoted ℜz and ℑz respectively.
Together they are called the Cartesian coordinates of z.
Cartesian coordinates are most useful for addition and subtraction, e.g.
(a + ib) + (c + id) = (a + c) + i(b + d).
They can also be used for multiplication:
(a + ib)(c + id) = (ac − bd) + i(ad + bc)

The set of complex numbers can be visualized as the complex (or Argand) plane:
14
 Chapter 0. Preliminaries

ℑ{z}

2i

i 2+i

ℜ{z}

−2 −1 1 2

−i
3
−1 − 2
i
−2i

The (complex) conjugate of z = x + iy is the number z = x − iy.

Conjugating twice gets us back where we started:

z = x − iy = x + iy = z

which means complex conjugation is an involution.

Note that
zz = (x + iy)(x − iy) = x2 + y2
which gives us a real number.
The modulus of z is the distance between z and the origin, that is,
p
|z| = x2 + y2 .

1 z
Exercise 19. If z ̸= 0, show that z
= |z|2
.

We can also divide complex numbers by getting rid of the imaginary part in the denomina-
tor:

a + ib (a + ib)(c − id) ac + bd bc − ad
= 2 2
= 2 2
+i 2
c + id c +d c +d c + d2

The argument of z is the counterclockwise angle, in radians, from the positive real axis to
the line segment connecting z and the origin.

15
 Chapter 0. Preliminaries

ℑ{z}
−2 + 2i
2i

| − 2 + 2i| i

arg(−2 + 2i) ℜ{z}

−2 −1 1 2

−i

−2i

The polar form of z is obtained by writing z as

z = reiθ

where r = |z| is the modulus and θ = arg z is the argument.

We have Euler’s famous identity

eiθ = cos θ + i sin θ,

which can be obtained by using the Taylor series for the exponential, and recognizing the
Taylor series for sine and cosine.
Euler’s identity allows us to easily convert between the polar and Cartesian coordinates.
Polar coordinates are most useful for multiplication, division, and exponentiation owing
to identities we know about exponentiation:

(r1 eiθ1 )(r2 eiθ2 ) = r1 r2 ei(θ1 +θ2 ) ,

r1 eiθ1 r1
iθ
= ei(θ1 −θ2 ) ,
r2 e 2 r2
(re ) = rn einθ .
iθ n

The unit circle is the set of complex numbers with modulus 1. These points form a circle on
the complex plane.
ℑ{z}
i

ℜ{z}

−1 1

−i

16
 Chapter 0. Preliminaries

Let n be a natural number. If z is a complex number such that zn = 1, then z is called an

nth root of unity.

Exercise 20. Show that the nth roots of unity all have the form e2πik/n for some k in Z.

Here is an animation of the nth roots of unity where n = 3, . . . , 12.

ℑ{z}

1 ℜ{z}

0.6 Matrices
An m-by-n matrix over R is an array of real numbers with m rows and n columns.
We typically use capital letters (A, B, C, . . . ) for matrices, and lowercase letters (a, b, c, . . . )
for their entries, subscripted by row and then column.
The set of all m-by-n matrices is denoted Mm×n (R).
Note, some people write Mm×n (R) as Rm×n . This is fine, but beware—R2×2 ̸= R4 !
Given an m-by-n matrix A = (ai,j ) and an n-by-p matrix B = (bk,l ), their product AB is the
m-by-p matrix of dot products of the rows of A with the columns of B. Explicitly,
X
n
(AB)i,j = ai,k bk,j .
k=1

Example 0.6.1 — The product of the 4-by-3 matrix

 
1 1 2
0 5 3
A= 6 2 4


4 0 5

17
 Chapter 0. Preliminaries

with the 3-by-2 matrix  

2 1
B = 3 2
1 5
is the 4-by-2 matrix
 
1 1 2  
0 5 3 2 1
6 2 4 3 2
AB =   
1 5
4 0 5
   
(1, 1, 2) · (2, 3, 1) (1, 1, 2) · (1, 2, 5) 7 13
(0, 5, 3) · (2, 3, 1) (0, 5, 3) · (1, 2, 5) 18
  25
=(6, 2, 4) · (2, 3, 1) = .
(6, 2, 4) · (1, 2, 5) 22 30
(4, 0, 5) · (2, 3, 1) (4, 0, 5) · (1, 2, 5) 13 29

The transpose of an m-by-n matrix A is the n-by-m matrix whose rows are the columns of
A. That is, (AT )i,j = aj,i for all i, j. Just flip it over its diagonal:
 
 T 1 4
1 2 3
= 2 5 .
4 5 6
3 6

A square matrix is a matrix with the same number of rows as columns. If A is an n-by-n
square matrix, an inverse of A is a matrix B such that AB = BA = I. Not every matrix has
an inverse—just consider  
0 0
A=
0 0
—but when an inverse exists, it’s unique, and we denote it A−1 . A matrix whose inverse
exists is called invertible.
Finally, the determinant of a square matrix A = (aij ) is defined as follows. For a 1-by-1
matrix,  
det a = a,
and for a larger matrix,
X
n
det A = (−1)i+j ai,j det A
e i,j (2)
j=1

where i is any fixed index between 1 and n, and A

e i,j is the matrix obtained by removing
the ith row and jth column from A. This is known as row expansion.

18
 Chapter 0. Preliminaries

Example 0.6.2 — By expanding along the top row,

 
a b    
det = a det d − b det c = ad − bc.
c d

You’re also allowed to expand down any column; the formula for column expansion has the
same shape as (2) but this time the sum is over i (the rows) and it’s j (the column) that’s
fixed.
Recall that the determinant is multiplicative:

det AB = det A det B.

This fact is fundamental.

Exercise 21. Show that A is invertible if and only if det A ̸= 0.

19
Chapter 1

Groups and subgroups

1.1 Binary operations and groups

Definition 1.1.1 — Let S be a set. A binary operation on S is a function ⋆ : S × S → S
written using infix notation, like so: a ⋆ b.

In other words, a ⋆ b is the image of the element (a, b) under the function ⋆. In more other
words, a ⋆ b = ⋆(a, b).

Example 1.1.2 — Addition on the integers is a binary operation + : Z × Z → Z. The

pair (a, b) is sent to its sum a + b.

Recall that the elements of S × S are ordered pairs (a, b) with a and b both in S. A function
⋆ : S × S → S has to map each ordered pair somewhere. If a and b are distinct, then the
ordered pairs (a, b) and (b, a) are distinct, too. There’s no reason why ⋆ should send
distinct ordered pairs to the same place. Thus, we generally do not have

a ⋆ b = b ⋆ a. (1.1)

Example 1.1.3 — Subtraction on the integers is a binary operation − : Z × Z → Z.

The pair (a, b) is sent to the difference a − b. Note that a − b ̸= b − a in general.

Definition 1.1.4 — A binary operation ⋆ on a set S such that a ⋆ b = b ⋆ a for all a

and b in S is called commutative.

20
 Chapter 1. Groups and subgroups

1.1.1 Visualizing binary operations

Remember your times tables from grade school? A “times table” for a general binary
operation is called a Cayley table.

Example 1.1.5 — Here is a portion of the Cayley table for subtraction on Z:

− −2 −1 0 1 2 3
−2 0 −1 −2 −3 −4 −5
−1 1 0 −1 −2 −3 −4
0 2 1 0 −1 −2 −3
1 3 2 1 0 −1 −2
2 4 3 2 1 0 −1
3 5 4 3 2 1 0

As with matrices, Cayley tables are indexed by row and then column. The (a, b)-entry in
the Cayley table is a ⋆ b.
In any Cayley table, the elements should appear in the same order down the right as across
the top. When S is finite, Cayley tables can be used to completely describe (hence define)
binary operations.

Example 1.1.6 — Multiplication in Z/5Z looks like this:

· [0] [1] [2] [3] [4]

[0] [0] [0] [0] [0] [0]
[1] [0] [1] [2] [3] [4]
[2] [0] [2] [4] [1] [3]
[3] [0] [3] [1] [4] [2]
[4] [0] [4] [3] [2] [1]

Example 1.1.7 (Rock–Paper–Scissors) — Consider the set M = {r, p, s}, where the
elements stand for rock, paper, and scissors. Define x ⋆ y to be the winner of the match
if x ̸= y, and define x ⋆ x = x when it’s a tie. Thus, for example, r ⋆ p = p and p ⋆ s = s.

21
 Chapter 1. Groups and subgroups

Exercise 22. Fill in the Cayley table for Rock–Paper–Scissors.

⋆ r p s
r p
p s
s

1.1.2 Associativity
Question — In Rock–Paper–Scissors, what is the value of

r ⋆ p ⋆ s?

On the one hand,

(r ⋆ p) ⋆ s = p ⋆ s = s.
But on the other hand,
r ⋆ (p ⋆ s) = r ⋆ s = r.
Perhaps this is an indication that you shouldn’t play Rock–Paper–Scissors with three
people at once, but the mathematical significance of this ambiguity is due to the failure of
⋆ to be what’s called associative.
Definition 1.1.8 — A binary operation ⋆ on a set S such that (a ⋆ b) ⋆ c = a ⋆ (b ⋆ c)
for all a, b, c in S is called associative.

When an operation is associative, everything is wonderful. We’re allowed to string together

elements freely, unburdened by bothersome brackets, unoppressed by pesky parentheses,
without fear of being misapprehended.
Associativity of addition is the reason we (would) never write

1+2+3+4+5

as
1 + ((2 + (3 + 4)) + 5).
However, non-associativity of subtraction is the reason we (should) never write

1−2−3−4−5

even though, in this case, most people would argue (vehemently and to the death) that it’s
−13 because they’re reading it left to right. But what about

1 − ((2 − (3 − 4)) − 5) = 3?!

22
 Chapter 1. Groups and subgroups

In sum, associativity is about arrangements of brackets (i.e. parentheses); commutativity is

about arrangements of elements. Don’t confuse
a ⋆ (b ⋆ c) = (a ⋆ b) ⋆ c and a ⋆ (b ⋆ c) = (b ⋆ c) ⋆ a.

1.1.3 Operations table

Here is a list of common candidates for binary operations. We may check if they actually
are binary operations (that is, they are indeed functions S × S → S), and if so, decide if
they are commutative or associative.

set candidate operation? commutative? associative?

N, Z, Q, R, C + yes yes yes
N − no - -
Z, Q, R, C − yes no no
N, Z, Q, R, C × yes yes yes
N, Z ÷ no - -
Q, R, C ÷ no - -
Q× , R× , C×∗ ÷ yes no no
R max{a, b} yes yes yes
R ab no - -
R>0 ab yes no no
R3 cross product yes no no
Rn dot product no - -
vector space vector addition yes yes yes
Mn×m (R) + yes yes yes
Mn×n (R) × yes no yes
self-maps composition yes no yes
{r, p, s} rock–paper–scissors yes yes no
any set (a, b) 7→ a yes no yes

Exercise 23. Explain every “no” in the table above. (That is, find a counterexample).

1.1.4 Definition of a group

We are now prepared to define what a group is, once we introduce one additional piece of
terminology: Associativity is so natural and desirable that we’ll usually take it for granted
∗
For now, we define them as Q\{0}, R\{0}, and C\{0} respectively. See Section 1.2.1 for the motivation of this
definition.

23
 Chapter 1. Groups and subgroups

in this course.
Definition 1.1.9 — A composition law is an associative binary operation.

Definition 1.1.10 — A group is a set G with a composition law ⋆ (called its group
operation) and a distinguished element e satisfying these two axioms:
Identity: a ⋆ e = e ⋆ a = a for all a in G
Inversion: for each a in G there exists b in G such that a ⋆ b = b ⋆ a = e
We denote this (G, ⋆, e). Often we simply write (G, ⋆) or even G and let the rest be
implied.

Intuitively, the Identity axiom says “You Can Do Nothing” while the Inversion axiom says
“You Can Undo Anything”—these axioms give us the structure of the invertible symmetries
perspective on groups that we discussed in Week 1.
Any element satisfying the Identity axiom is called an identity; any element satisfying the
Inversion axiom is called an inverse.
Aside. The traditional definition of a group (which you may see in your textbooks) says that a
group is a set G with an operation ⋆ satisfying these four axioms:

Closure: a ⋆ b is in G for all a, b in G

Associativity: ⋆ is associative

Identity: there exists e in G such that a ⋆ e = e ⋆ a = a for all a in G

Inversion: for each a in G there exists b in G such that a ⋆ b = b ⋆ a = e

The traditional definition is unsatisfactory for a few technical reasons. First, the Closure axiom
becomes redundant once you ask ⋆ to be a binary operation. Second, the Associativity axiom
is moreso a property of the operation than of the elements. Third, the traditional definition
doesn’t clarify that the ‘e’ that’s asserted to exist in the Identity axiom is the same ‘e’ that
appears in the Inversion axiom.

1.1.5 Immediate consequences

Exercise 24. Show that the identity element e is unique. [That is, show that if e ′ is
another element of G that satisfies the Identity axiom, then e ′ = e.]

24
 Chapter 1. Groups and subgroups

Exercise 25. Show that for any a in G, there is a unique element b such that a ⋆ b =
b ⋆ a = e. [That is, show that if b ′ is another inverse for a, then b ′ = b.]

Remark 1.1.11. This unique element is called the inverse of a and denoted a−1 .

Exercise 26. What is the inverse of a ⋆ b?

Exercise 27. Show that we can perform right cancellation:

a⋆c=b⋆c
a=b

and left cancellation:

c⋆a=c⋆b
a=b

for all elements a, b, c in any group.

Note the importance of the sides of the expressions we are working with: We do not have in
general that c ⋆ a = b ⋆ c implies a = b [When do we have this?].
The definition of groups is actually a little stronger than we require. In fact, we can weaken
the axioms so that we only check one side of the equalities.

Exercise 28. Let G be a set with an associative binary operation ⋆ and a distinguished
element e satisfying these two axioms:
• a ⋆ e = a for all a in G (Axiom of Right Identity)
• for each a in G there exists b in G such that a ⋆ b = e (Axiom of Right Inversion)
These are like the group axioms, except they’re only required to hold “on one side”. In
this exercise you will prove that any structure (G, ⋆, e) satisfying these weaker axioms
is actually already a group.
a) Prove that G has the right-cancellation property: a ⋆ c = b ⋆ c implies a = b.
b) An idempotent is an element i such that i⋆i = i. Show that e is the only idempotent
in G. How is this related to left-cancellation?
c) Show that every right inverse is a left inverse.
d) Show that e is a left identity.
e) Explain why we are done.

25
 Chapter 1. Groups and subgroups

1.1.6 Notation
Composition laws have lots of notations, like ⋆, ∗, ◦, ·, ×, ⊗, +, ⊕, ... But when we’re
dealing with a single group, there’s only one composition law involved—so we can get
away with not writing it at all. (It’s also kind of annoying to write ⋆ all the time.) This is
called the multiplicative notation.
If the composition law is commutative, the group is called abelian. Some people write the
composition law in abelian groups using a plus sign (+), but we’ll stick to the multiplicative
notation except in very concrete cases, like Z/nZ under addition.
Indeed, we will use more notation inspired by those found in multiplication and addition:

Definition 1.1.12 — Let G be a group with identity element e and let g ∈ G. For each
integer n, define gn as follows:
if n > 0, put
gn = g · . . . · g
| {z }
n times

if n < 0, put
gn = (g−1 )−n
and if n = 0, put
g0 = e.

This notation is extremely useful for simplifying long expressions:

aabbbcdddd = a2 b3 cd4 .

For abelian groups written additively, “gn ” becomes “ng”:

a + b + c + b − a = 2b + c.

To summarize,
notation multiplicative additive
operation a · b or ab a+b
identity e or 1 0
inverses a−1 −a
powers an na
Note in particular that we will use multiplicative notation for function composition.
Multiplicative notation is convenient as it behaves largely the same way multiplication
does.

26
 Chapter 1. Groups and subgroups

Proposition 1.1.13 (Exponent Laws)

For all g in G and all n, m in Z,
1. gn gm = gn+m
2. (gn )m = gnm
3. (g−1 )n = (gn )−1 = g−n

Proof. Exercise.

Exercise 29. Let G be a group. Suppose a2 = e for every a in G. Show that G is

abelian.

1.2 Examples
1.2.1 Basic groups
Wherever addition is defined, we may have a group; the set in question must contain zero
(the additive identity) and be closed under negation (to form additive inverses).
Thus N is not a group under addition, but Z, Q, R, and C are.
Similarly, wherever multiplication is defined, if the set contains 1 (the multiplicative
identity) and is closed under reciprocation (to form multiplicative inverses), then we have
a group. Using a superscript × to denote the set of “multiplicatively invertible” elements,
we find that Q× , R× , and C× are Q\{0}, R\{0}, and C\{0} respectively, and they are all groups
under multiplication.

Exercise 30. What is Z× ?

1.2.2 Groups of integers

The set Z/nZ of residue classes modulo n forms an abelian group under addition: the
identity element is [0] and the inverse of [a] is [−a]. This is called the additive group (of
integers) modulo n.
What about multiplication? Sure, we can multiply classes, and [1] is in Z/nZ, but—
inverses?
Actually, not every class has a multiplicative inverse. For example, in Z/6Z,

[3][4] = [12] = [0]

27
 Chapter 1. Groups and subgroups

so neither [3] nor [4] are invertible. (If they were, say [3][a] = [1], then we’d have

[4] = [4][1] = [4][3][a] = [0][a] = [0]

which can’t happen modulo 6.) However,

[5][5] = [25] = [1]

so [5] is invertible.
By restricting our attention to invertible elements, we obtain:

(Z/nZ)×

known as the multiplicative group (of integers) modulo n, a.k.a. U(n).

Example 1.2.1 — (Z/6Z)× = {[1], [5]} and (Z/8Z)× = {[1], [3], [5], [7]}.

Exercise 31. What are the invertible elements of Z/nZ? What are the invertible
elements of Z/pZ if p is a prime?

1.2.3 Matrix groups

The set of invertible n × n matrices forms a group under matrix multiplication, called the
general linear group.

GLn (R) = {A ∈ Mn×n (R) : det A ̸= 0}

We can also consider matrices with entries in Z/mZ. However, det A ̸= [0] is no longer
enough—we need det A to be invertible modulo m. That is,

GLn (Z/mZ) = {A ∈ Mn×n (Z/mZ) : det A ∈ (Z/mZ)× }

Example 1.2.2 — In Z/12Z, writing a instead of [a],

" #
2 1
det = 24 − 13 = 8 − 3 = 5
3 4

which is invertible because 52 = 25 ≡ 1 (mod 12). The inverse matrix is

" #−1 " # " # " # " #
2 1 −1 4 −1 4 11 20 55 8 7
=5 =5 = = .
3 4 −3 2 9 2 45 10 9 10

28
 Chapter 1. Groups and subgroups

1.2.4 Trivial group

Let G be a set with one element, which we’ll call e. There is only one possible binary
operation on G:

G×G→G
(e, e) 7→ e

This yields the trivial group.

Exercise 32. Check that the trivial group is a group. What is its Cayley table?

1.2.5 Cyclic groups

A cyclic group is a group in which every element is an integer power∗ of a single element,
called a generator. We write
G = ⟨g⟩
to mean G is cyclic with generator g.
For example, in the group of integers under addition, every integer is an integer multiple
of 1, so
Z = ⟨1⟩.

Similarly, in the additive group modulo n, every element can be written as a sum of [1]’s.
Therefore,
Z/nZ = ⟨[1]⟩.

The group µn is the set of complex nth roots of unity under multiplication. That is,

µn = {z ∈ C : zn = 1}.

Since the nth roots of unity are of the form e2πik/n , we may write

µn = ⟨e2πi/n ⟩.

1.2.6 Dihedral groups

The dihedral group Dn is the set of reflections and rotations — SYMMETRIES — of a
regular n-gon, under composition.

∗
multiple in additive notation

29
 Chapter 1. Groups and subgroups

Example 1.2.3 — Consider the regular pentagon with its rotations

and reflections

Let r denote the one-fifth clockwise turn and let f denote the flip over the vertical axis.
Then every rotation is a power of r.
We can also express every reflection in terms of just f and r. For example, to produce a
flip across the orange axis, first flip across the red axis (f) then turn two-fifths clockwise
(r2 ), yielding fr2 .

f r2

1.2.7 Symmetric groups

Let X be a set. The symmetric group on X is the set of all permutations∗ on X under
composition. This group is denoted
SX
Special cases: the symmetric group on {1, . . . , n} is denoted Sn while the symmetric group
on N is denoted S∞ .
For example, the only two elements of S2 are the permutations
! !
1 2 1 2
id = and τ =
1 2 2 1

in two-line notation). The function τ2 sends 1 → 2 → 1 and 2 → 1 → 2, so τ2 = id. Thus

S2 = ⟨τ⟩ is cyclic with 2 elements.
∗
i.e. bijective self-maps

30
 Chapter 1. Groups and subgroups

We will explore these groups in more detail in later weeks.

1.3 Subgroups
Definition 1.3.1 — A group H is a subgroup of a group G if H is a subset of G and
the group operation on H is the same as the operation on G. The subgroup relation is
written
H≤G
meaning “H is a subgroup of G”.

To check if a subset H of a group G is a subgroup, one must show

(i) ab ∈ H for all a, b in H (the operation in G restricted to H is still a binary operation)
(ii) e ∈ H (the identity element of G is in H)
(iii) a−1 ∈ H for all a in H (the inverse in G of every element of H is in H)
(i) shows that the group operation in G restricts to a function H × H → H while (ii) and (iii)
show that H is a group.∗

Proposition 1.3.2 (Subgroup Criterion)

Let G be a group and let H ⊆ G. Then H ≤ G if and only if H is non-empty and
ab−1 ∈ H for all a, b in H.

Proof. The ‘only if’ (forward implication) is easy. For the converse, we show that the three
properties (i), (ii), and (iii) hold, albeit in a different order.
Start with (ii). Since H is non-empty, there is some element a in H. By hypothesis, aa−1 ∈ H.
But aa−1 = e, so e ∈ H.
Next, (iii). Let a ∈ H. By (ii) and the hypothesis, ea−1 ∈ H. But ea−1 = a−1 , so a−1 ∈ H.
Finally, we show (i). Let a, b ∈ H. By (iii) and the hypothesis, a(b−1 )−1 ∈ H. But
(b−1 )−1 = b, so ab ∈ H.

Exercise 33. Show that if H1 , H2 are subgroups of G, then H1 ∩ H2 is a subgroup of G.

Let us now look at some common subgroups of groups.

∗
A priori, H might have its own identity e ′ , but since the operations coincide, e ′ e ′ = e ′ in G, so e ′ = e. Same
for inverses.

31
 Chapter 1. Groups and subgroups

1.3.1 Basic groups

In additive notation, “ab−1 ” means a − b. Thus, under addition,

Z ≤ Q ≤ R ≤ C,

because the difference of two integers (resp. rationals, reals) is an integer (resp. rational,
real). (Of course, 0 ∈ Z.)
Similarly,
Q× ≤ R× ≤ C× ,
because the quotient of two nonzero rational (resp. real) numbers is rational (resp. real).
(Also, 1 ∈ Q× .)

1.3.2 Matrix groups

A matrix group is a subgroup of GLn for some n. (The definition works the same regardless
of what set the entries are in).
The most important matrix groups are the special linear groups—matrices that preserve
volume (| det A| = 1) and orientation (det A > 0, only relevant when the entries are in Q or
R),
SLn = {A ∈ GLn : det A = 1},
the orthogonal groups—matrices that preserve distance (which necessarily preserves volume
[prove it!]),
On = {A ∈ GLn : A−1 = AT },
and the special orthogonal groups—matrices that preserve distance, volume, and orientation,

SOn = {A ∈ GLn : A−1 = AT , det A = 1}.

To show that SLn ≤ GLn , just note that I ∈ SLn because det I = 1, and if A, B ∈ SLn , then

det AB−1 = det A · det B−1 = 1 · 1−1 = 1

so AB−1 ∈ SLn .
Exercise 34. Prove that On is a matrix group. Conclude that SOn is a matrix group.

There are many other examples of matrix groups.

Exercise 35. Show that the set of matrices of the form

" #
1 x
(x ∈ R)
0 1

32
 Chapter 1. Groups and subgroups

is a matrix group.

1.3.3 Cyclic groups

If G = ⟨g⟩ is cyclic and k is any fixed integer, it follows from 1.1.13 that the set of integer
powers of gk is a subgroup of G. That is,

⟨gk ⟩ ≤ G.

For example, we saw that Z under addition forms a cyclic group generated by 1. Thus,
⟨k⟩ ≤ Z for every integer k. In particular, the set of even numbers is a subgroup of Z.

Exercise 36. Is the set of odd numbers a subgroup of Z?

We’ll talk more about cyclic groups in Week 4.

1.3.4 Dihedral groups

In Dn , if r is any rotation and f is any flip, then ⟨r⟩ and ⟨f⟩ are (two of the) subgroups of
Dn .
We’ll talk more about dihedral groups in Week 4.

1.3.5 Permutation groups

A permutation group is a subgroup of a symmetric group. For example, the four permuta-
tions
! !
1 2 3 4 1 2 3 4
e= , ρ= ,
1 2 3 4 2 1 3 4
! !
1 2 3 4 1 2 3 4
σ= , τ=
1 2 4 3 2 1 4 3

constitute a subgroup of S4 called the Klein four-group∗ , denoted V.

Exercise 37. Show that V is a group by completing its Cayley table.

∗
Also see Klein Four.

33
 Chapter 1. Groups and subgroups

◦ e ρ σ τ
e
ρ
σ
τ

We’ll talk more about symmetric groups and permutation groups in weeks 5 and 6.

1.3.6 Generating new subgroups

Let G be a group and let S ⊆ G. The subgroup generated by S is the set of all possible
combinations of the elements of S (and their inverses) using the composition law in G. It is
denoted ⟨S⟩. Formally, we have

⟨S⟩ = {g±1 ±1
1 . . . gk : k ≥ 0 and gi ∈ S}.

(By allowing k = 0 we include the empty product, which we always take to be e. In

particular, the empty set generates the trivial group!)

Proposition 1.3.3
Let S ⊆ G. Then ⟨S⟩ ≤ G.

Proof. ⟨S⟩ is non-empty, because we can always form the empty product to get e. And if
a, b ∈ S then
a = gϵ1 1 . . . gϵkk and b = hδ11 . . . hδl l ,
where gi , hj ∈ S and ϵi , δj ∈ {1, −1}. Thus

ab−1 = gϵ1 1 . . . gϵkk h−δ

l
l
. . . h−δ
1 .
1

All k + l terms are in S and all the exponents are 1 or −1, so ab−1 ∈ ⟨S⟩.
Remark 1.3.4. ⟨S⟩ is in fact the smallest subgroup containing S—any subgroup containing
S must contain all elements of the form g±1 ±1
1 . . . gk , and so it must contain ⟨S⟩.

If S is finite, say S = {g1 , . . . , gn }, then we write

⟨g1 , . . . , gn ⟩ instead of ⟨{g1 , . . . , gn }⟩.

If S is a singleton (i.e. n = 1), say S = {g}, then ⟨S⟩ = ⟨g⟩ is called the cyclic subgroup
generated by g. Of course, the whole group G is cyclic iff G = ⟨g⟩ for some g in G.

34
 Chapter 1. Groups and subgroups

Example 1.3.5 — In the additive group Q,

⟨ 21 , 13 ⟩ = { n2 + m
3
: n, m ∈ Z}

because Q is abelian. Putting this expression on a common denominator yields

n m 3n + 2m
+ = .
2 3 6
Since 3(−3) + 2(5) = 1, this subgroup is actually cyclic—every element is an integer
multiple of 61 .

Example 1.3.6 — In the multiplicative group Q× ,

⟨2, 3⟩ = {2n 3m : n, m ∈ Z}

is the subgroup of fractions whose numerator and denominator (in lowest terms) are
divisible by 2 and 3 only. For example,

2 256 1
6, , , ∈ ⟨2, 3⟩
3 243 1024
but 5 is not.

Exercise 38. Show that the group in the above example cannot be cyclic. (That is,
show that there is no g ∈ Q× such that ⟨g⟩ = ⟨2, 3⟩.

1.4 Guises of the same group

We said that there’s only one group of one element (the trivial group), so µ1 , S1 , and any
trivial subgroup of another group—no matter what the identity element is—must be the
trivial group, despite the different names. You may also have noticed that many groups
under different names seem to behave in the exact same way—µ2 , Z/2Z, Z× , S2 , and D1 , for
example, all have two elements, the identity and an involution. Their Cayley tables thus
look identical other than the names of the elements. All their group-theoretic properties
that we know of are the same. So we are tempted to call them all the same group. But
what does it mean, exactly, for two groups to be “the same”?
In general, the answer to this question allows us to export the answer to a specific question
to a class of questions.

35
 Chapter 1. Groups and subgroups

Question — When are two mathematical objects the same? In other words, when can
we interchange one thing with the other in a question?

As it turns out, the answer depends on how we define sameness. For example, consider
the two sets S = {a, b, c, d, e} and T = {1, 2, 3, 4, 5}. Of course, the elements have different
names, so they’re different in that manner. However, they do have the same size, and we
may view these two sets as interchangeable in set-theoretic contexts.

Exercise 39. Consider the two questions “how many three-letter words can you make
from the letters in S (with repeats)?” and “how many functions are there from {1, 2, 3}
to T ?”. Why are these two questions in fact the same question?

The notion of “interchangeable” sets allows us to answer questions concerning finite sets
by just answering questions about sets {1, 2, 3, ..., n}.∗
When we introduce additional structure on the object in question, the notion of “sameness”
also expands to include these structure. For example, if we are to ask questions about
a vector space V, we know these questions have the same answers as with the vector
space W if there is an invertible linear transformation T : V → W—linear transformations
preserve vector addition and scalar multiplication. This allows us to reduce the study of all
finite dimensional vector spaces to just the study of Fn , where F is the corresponding field.
Returning to groups, we may ask when two groups are the same. An understanding of
this type will allow us to answer questions about groups more easily, by treating many
groups as guises of other groups that we’ve already studied.

Definition 1.4.1 — Given two groups (G, ⋆) and (H, ◦), we say G and H are isomorphic
if there is a bijection ϕ : G → H such that for all a, b ∈ G, we have

ϕ(a ⋆ b) = ϕ(a) ◦ ϕ(b).

∼ H.
Such a ϕ is called an isomorphism. We denote this G =

Remark 1.4.2. The operation on the left-hand side is done in G (before f is applied) whereas
the operation on the right-hand side is done in H (after f is applied). In the future we will
drop the operation entirely and write

ϕ(ab) = ϕ(a)ϕ(b)

if no further information is given.

Note that this immediately gives ϕ(eG ) = eH and ϕ(g−1 ) = ϕ(g)−1 .
∗
The situation for infinite sets is a bit messier, and we will not get into that within this course except to note
that two infinite sets may not actually have the same “size” (i.e. cardinality).

36
 Chapter 1. Groups and subgroups

∼ is an equivalence relation.
Exercise 40. Show that =

Exercise 41. Show that R× is isomorphic to GL1 (R). (Is the latter group abelian?)

A visual way to check isomorphisms of small groups is to draw both Cayley tables, colour
each element of one group uniquely, and colour the corresponding elements in the other
group the same colour. If the resulting patterns match, then the groups are isomorphic.
For example, writing ω = e2πi/3 , we can consider Z/3Z and µ3 = {1, ω, ω2 }:

Z/3Z [0] [1] [2] µ3 1 ω ω2

[0] [0] [1] [2] 1 1 ω ω2

[1] [1] [2] [0] ω ω ω2 1
[2] [2] [0] [1] ω2 ω2 1 ω

Why does this work? Suppose ■■ = ■ in the group on the left,

By definition of our colouring scheme, f(■) must be red, f(■) must be blue, and f(■■) =
f(■) must be green. The colour of f(■)f(■) is the colour of the cell in the f(■)-row and
the f(■)-column. This is green if and only if f(■) = f(■).

Exercise 42. Show that the groups listed at the beginning of this section are isomor-
phic.

Exercise 43. Show that Z/4Z, µ4 , (Z/5Z)× , are isomorphic.

Show that V and (Z/8Z)× are isomorphic. Show that they are not isomorphic to the
groups above.

1.5 Group Actions

1.5.1 Group actions
“GROUPS, AS MEN, WILL BE KNOWN BY THEIR ACTIONS.”
—Guillermo Moreno
Here are all 16 = 24 possible square, non-edge puzzle pieces (idealized).

37
 Chapter 1. Groups and subgroups

Say two puzzle pieces are rotation-equivalent if you can rotate one so it looks like the other.
How many puzzle pieces are there up to rotation-equivalence?

The answer is 6. Each puzzle piece has one of the following shapes:

Mathematically, we just determined the orbits of the action of the group C4 on the set of
puzzle pieces.

38
 Chapter 1. Groups and subgroups

Definition 1.5.1 — Let G be a group and X be a set. A left action of G on X is a function

G × X → X, denoted (g, x) 7→ g . x, satisfying
1) identity: e . x = x for all x in X
2) compatibility: g . (h . x) = gh . x for all x in X and g, h in G
We write G ↷ X to mean G acts on X.

Proposition 1.5.2
Let G ↷ X. Then “x ∼ y iff g . x = y for some g in G” is an equivalence relation.

Proof. By the identity axiom, ∼ is reflexive: e . x = x so x ∼ x for all x in X. By the compatibility

axiom, ∼ is transitive: if x ∼ y and y ∼ z then g . x = y and h . y = z for some g, h in G, so

hg . x = h . g . x = h . y = z.

By both axioms together, ∼ is symmetric: if x ∼ y then g . x = y so

g−1 . y = g−1 . g . x = g−1 g . x = e . x = x.

Definition 1.5.3 — The equivalence classes of the relation above are called orbits. The
orbit of x is denoted OrbG (x) or Gx, and the set of equivalence classes is denoted X/G.
That is,

OrbG (x) = Gx = {g . x : g ∈ G}
and
X/G = {Gx : x ∈ X}.

Remark 1.5.4. Note that, since orbits are equivalence classes, the set of all orbits form a
partition of X.

Exercise 44. Is gX = {g . x : x ∈ X} interesting?

Example 1.5.5 — Consider again the action of C4 on the set of puzzle pieces. There
are 6 orbits, indicated by colour. Some orbits are small (like those of and ) while
others are large (like those of , , and ). Note that the total number of puzzle pieces
is the sum
1 + 4 + 4 + 2 + 4 + 1 = 16
because the orbits partition the set.

39
 Chapter 1. Groups and subgroups

Remark 1.5.6. A right group action is just like a left group action except that the function
goes from X × G to X and is denoted (x, g) 7→ x . g. The difference stems from flipping the
compatibility axiom into (x ↷ g) ↷ h = x ↷ gh—the action by gh on x is performed g
and then h, as opposed to h and then g (as in function composition).
The other notations are flipped as well: we write X ↶ G to mean G acts on X on the right
(“right-acts”), and the orbit space of a right action is denoted G\X.
Below we will introduce many types of group actions for future reference.

1.5.2 Concrete examples

Example 1.5.7 — For any X and G, we have the trivial action g . x = x for all x in X,
for each g in G. Every orbit is trivial.

Example 1.5.8 — The group µ2 acts on any group G by inversion: for all x in G,

1.x = x
−1 . x = x−1

Since the identity and involutions are self-inverse, while other elements are not, the
orbits are either pairs {x, x−1 } where x ̸= x−1 or singletons {x} where x = x−1 (including
{e}).

Example 1.5.9 — The group Z does not act on an arbitrary group G by n . g = gn ,

because neither identity nor compatibility hold:

0 . g = g0 = e ̸= g in general, and

(n + m) . g = gn+m = gn gm whereas n . m . g = n . (gm ) = gmn .

Example 1.5.10 — Dn acts on the n-gon (more precisely, parts of the n-gon) by flips
and rotations on the right: we agreed that e.g. fr should be interpreted as first flip, then
rotate.

Example 1.5.11 — Consider the equilateral triangle with colored vertices. Now we
have its rotations

40
 Chapter 1. Groups and subgroups

and reflections

We may say that D3 acts on the red, green, and blue vertices of the triangle, such that
rotation by 120◦ clockwise sends the red vertex to the (original position of the) green
vertex. We may also say that Dn acts on the (unlabelled) edges of the triangle. We
may even say that Dn acts on the set of six possible configurations of the equilateral
triangle.
The orbit of any vertex is the whole set of vertices; the same holds for any edge, mutatis
mutandis.

Exercise 45. For n ≥ 4 an n-gon has 12 n(n − 3) diagonals. How many orbits of those
are there?

Example 1.5.12 — Given a fixed origin, O3 (R) acts on objects in the three-dimensional
space by rotations and reflections about that origin. SO3 (R) acts on them by rotations
only.

Exercise 46. Show that Dn is a subgroup of O2 (R) using their geometric definitions.

Example 1.5.13 — SX ↷ X for any set X in the obvious way: σ . x = σ(x). In particular,
Sn acts on {1, ..., n}.

Example 1.5.14 — The affine group Aff(R) of functions

x 7→ ax + b (a ̸= 0)

acts on R in the obvious way. There is only one orbit—for any two real numbers u and
v, the affine map x 7→ x + v − u sends u to v.

Exercise 47. For which pairs of distinct points (u1 , u2 ) and (v1 , v2 ) does there exist an
affine map sending ui to vi ?

41
 Chapter 1. Groups and subgroups

Example 1.5.15 — Assume for simplicity that we can identify musical notes with
their fundamental frequencies.
On a standard modern 88-key piano,∗ the A above middle C sounds the frequency 440
Hz, and the ratio between the frequencies f0 < f1 of two successive keys (white–black,
black–white, or white–white) is
f1 √
12
= 2 = 1.059...
f0
Thus the frequencies of the 88 keys (starting four octaves below 440) ranges from 27.5
Hz to just over 4186 Hz.†
For reference, the human hearing range is commonly given as 20 Hz to 20000 Hz,
which, in musical terms, ranges from a fifth below the bottommost piano-note to two
octaves and a minor third above the topmost piano key.
A unifying feature of nearly all musical traditions is the observation that two frequen-
cies sound “the same” when the ratio between them is an integer power of 2. This is
called octave equivalence.‡
So, let
X = {440 · 2k/12 : k ∈ Z}
be the “idealized” Western musical scale, whose elements we’ll refer to as pitches, and
let G = ⟨2⟩ be the cyclic subgroup of R× generated by 2. Then G ↷ X by transposition
by octaves, and the orbit space X/G has exactly twelve elements (called pitch classes):

A = {. . . , 220, 440, 880, . . . } D♯ = {. . . , 311, 622, 1244, . . . }

A♯ = {. . . , 233, 466, 932, . . . } E = {. . . , 329, 659, 1318, . . . }
B = {. . . , 246, 493, 987, . . . } F = {. . . , 349, 698, 1396, . . . }
C = {. . . , 261, 523, 1046, . . . } F♯ = {. . . , 369, 739, 1479, . . . }
C♯ = {. . . , 277, 554, 1108, . . . } G = {. . . , 391, 783, 1567, . . . }
D = {. . . , 293, 587, 1174, . . . } G♯ = {. . . , 415, 830, 1661, . . . }
∗
tuned to concert pitch
√ in equal temperament
†
27.5 · 287/12 = 3520 4 2 = 4186.009...
‡
Whether or not human perception of octave equivalence is innate or learned is still unclear, though
Jacoby et al. (2019) found evidence for the latter by studying an isolated tribe living in the Bolivian
Amazon.

Example 1.5.16 — They say you’re supposed to “rotate your mattress” every so often
to prevent it from sagging. Certain models can (and therefore must) also be “flipped”.
Together, these physical manoeuvres give an action of D2 on the ideal mattress (only

42
 Chapter 1. Groups and subgroups

theoretical).
Unfortunately, D2 is not cyclic. That means you can’t “cycle” the mattress through all
possible configurations by repeating a single, easy-to-remember action. Consequently,
mattress companies have devised complicated mattress-flipping schemes detailing
when and how to flip your mattress for maximum performance.

1.5.3 Abstract examples

Example 1.5.17 — Every group G acts on itself by multiplication: g . x = gx. There is
only one orbit because for any two group elements x, y we can take g = yx−1 and get
g . x = y.

Example 1.5.18 — Every group G acts on itself by conjugation—the act of operating

on an object by an element on one side and its inverse on the other: g . x = gxg−1 .
The orbits are called conjugacy classes and their number is called the class number of G,
denoted k(G).

Remark 1.5.19. Both these actions have right versions: G ↶ G by x . g = xg and x . g =

g−1 xg.

1.5.4 New actions from old

A variety of new actions can be constructed from a given action G ↷ X.

Example 1.5.20 (Restricted action) — A subset Y of X is called G-invariant if g . y ∈ Y

for all y in Y and all g in G. If Y ⊆ X is G-invariant, then G acts on Y the same way it
acts on X.

Exercise 48. Show that orbits are G-invariant, and that every G-invariant subset is a
union of (zero or more) orbits.

Example 1.5.21 (Subgroup action) — Let H ≤ G. Then H acts on X the same way G
does.

Example 1.5.22 (Function action) — Let Y be a set and let Y X denote the collection of
all functions X → Y. Then G acts on Y X on the right by f . g = (x 7→ f(g . x)).

43
 Chapter 1. Groups and subgroups

Exercise 49. Show that the function action is indeed an action—that it satisfies the
identity and compatibility axioms.

Exercise 50 (Opposite action). Let x . g = g−1 . x for all x in X and all g in G. This turns
the left action of G on X into an equivalent right action. In terms of the original (left)
action, what is the orbit of x under the opposite (right) action? Show that the opposite
of the opposite is the original.

1.6 Orders
1.6.1 Order of a group
With all the structures on G, we would like to count it.

Definition 1.6.1 — The order of a group G, denoted o(G) or |G|, is the number of
elements in G. If that number is infinite, we say the group has infinite order, and we
write o(G) = ∞.

Example 1.6.2 — o(Q) = ∞,

o(Z/nZ) = o(µn ) = n,
o(Sn ) = n!, and
o(Dn ) = 2n.

Theorem 1.6.3 (Lagrange)

Let G be a finite group and let H ≤ G. Then o(H) divides o(G).

Proof. Define a right group action of H on G by

g . h = gh.

We may check that this is indeed a group action—g . e = g, and (g . h) . h ′ = ghh ′ =

g . (hh ′ ).
Now consider the orbit of some g ∈ G. We claim now that OrbH (g) = {gh : h ∈ H},
denoted gH. Why? Because:
Take any g ′ ∈ gH, we know that g ′ = gh for some h, which immediately tells us g . h = g ′
and so g ′ ∈ OrbH (g). On the other hand, for any g ′ ∈ OrbH (g), we know there is some h
such that g . h = gh = g ′ . That is to say, g ′ ∈ gH.
44
 Chapter 1. Groups and subgroups

We also claim that the orbits all have the same size. Indeed, given an orbit xH, consider the
function H → xH defined by h 7→ xh. This function is
• injective: if h1 and h2 map to the same place, then xh1 = xh2 , so by left-cancellation,
h1 = h2 , and
• surjective: if y ∈ xH, then y ∈ OrbH (x), so y = xh for some h in H, so h maps to y.
Thus h 7→ xh is a bijection, so H and xH have the same size. In particular, every orbit has
size o(H).
Since orbits are equivalence classes, they partition G. All the orbits have the same size
o(H). Since G is finite, o(G) is a multiple of o(H) i.e. o(H) divides o(G).

During the course of the proof, several very important concepts came up.
• the group action—H acting on G by right multiplication—is called the right subgroup
action of H on G,
• the orbits—the sets of the form xH—are called left cosets (of H (in G)),
• the set of orbits—the set of left cosets of H in G—is denoted G/H, and
• the number of orbits—the size of the set G/H—is called the index (of H (in G)),
denoted [G : H].
Remark 1.6.4. Note that only in the end of the proof did we use finiteness of G. The rest of
the proof of Lagrange’s theorem shows that

o(G) = [G : H]o(H)

even if any these quantities is ∞. If G is finite, then

o(G)
[G : H] = .
o(H)

Example 1.6.5 — Consider G = Z/24Z and H = ⟨[4]⟩. The cosets of H, written

additively since G is abelian, are

H = [0] + H = {[0], [4], [8], [12], [16], [20]},

[1] + H = {[1], [5], [9], [13], [17], [21]},
[2] + H = {[2], [6], [10], [14], [18], [22]}, and
[3] + H = {[3], [7], [11], [15], [19], [23]}.

Clearly, the cosets partition Z/24Z. Each coset has 6 elements (the order of H) and
there are 4 cosets in total (the index of H). And indeed, 6 times 4 is 24.

45
 Chapter 1. Groups and subgroups

Exercise 51. Show that if a group G has a finite subgroup of finite index, then G is
finite.

Exercise 52. Give an example of an infinite group with an infinite subgroup of infinite
index.

Exercise 53. Show that [Z : ⟨n⟩] = n for each n > 0. What is [Z : ⟨0⟩]?

Exercise 54. Let G = Z and let H = ⟨n⟩. In the notation from the proof of Lagrange’s
theorem, show that a ∼ b if and only if a ≡ b (mod n). What does this prove about
the sets Z/⟨n⟩ and Z/nZ?

Remark 1.6.6. Note that we often denote the subgroup H = ⟨n⟩ by nZ, hence the Z/nZ
notation.
Remark 1.6.7. Note that in this case, our G/H is itself a group under gH + g ′ H = (g + g ′ )H.
This is not true in general—we will explore when G/H is a group in the future when we
talk about normal subgroups.
* !+
1 2 3
Exercise 55. Let G = S3 and H = . Can you put a group structure on
2 1 3
G/H using the group operation from G?

1.6.2 Order of an element

Definition 1.6.8 — The order of an element g, denoted o(g) or |g|, is the least positive
integer exponent n such that gn = e, or ∞ if no such exponent exists. In symbols,

o(g) = min{n ≥ 1 : gn = e}.

Example 1.6.9 — The identity element is the only element of order 1. That is, o(g) = 1
iff g = e.

Example 1.6.10 — o(g) = 2 iff g is an involution and not the identity.

Example 1.6.11 — The orders of the elements of Z/12Z are

46
 Chapter 1. Groups and subgroups

[a] [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]
o([a]) 1 12 6 4 3 12 2 12 3 4 6 12

Lemma 1.6.12 (Division)

Let G be a group and let g ∈ G be an element of finite order. Then gm = e if and only
if o(g) | m.

Proof. Let n = o(g). If n | m then

gm = (gn )m/n = em/n = e

by the exponent laws, since m/n is an integer.

Suppose gm = e. Divide m by n with remainder to get m = qn + r for some integers q
and r satisfying 0 ≤ r < |n| = n. By the exponent laws,

gm = gqn+r = (gn )q gr = eq gr = gr .

But by hypothesis, gm = e. Thus gr = e. Since n is the least positive integer annihilating g,

and r < n, it must be the case that r = 0. Thus m = qn i.e. n = o(g) | m.

Exercise 56. Suppose g has infinite order. Show that gm = e iff m = 0.

Corollary 1.6.13
Let n = o(g) be finite. Then gi = gj iff i ≡ j (mod n).

Proof. gi = gj iff gj−i = e iff n | j − i iff i ≡ j (mod n).

The following fundamental Theorem connects the two meanings of the word “order”.

Theorem 1.6.14
The order of an element is equal to the order of the cyclic subgroup it generates.

Proof. Suppose g has infinite order. Then there is no nonzero k such that gk = e, and it
follows that ⟨g⟩ is infinite.
So, suppose g has finite order n. We want to show two things: first, that

⟨g⟩ = {e, g, g2 , ..., gn−1 }

47
 Chapter 1. Groups and subgroups

and second, that the set on the right-hand side actually has n elements.
First: The elements of ⟨g⟩ are, by definition, integer powers of g. Given gk , let r be the
remainder of k when it’s divided by n. Then k ≡ r (mod n) so, by the Corollary, gk = gr .
Note that 0 ≤ r < n, so every power of g equals something in this list.
Second: To show that this list has no repeats, suppose gi = gj for some 0 ≤ i ≤ j < n. Then
i ≡ j (mod n), again by the Corollary. But i and j are both less than n, so i = j.
Putting it all together, we conclude that the elements of ⟨g⟩ are just the n powers e, g, g2 , . . . ,
gn−1 . Therefore o(⟨g⟩) = n = o(g).

Exercise 57 (Corollary-Exercise). Let G be a finite group. Show that o(g) divides

o(G).

Exercise 58. Show that {n ∈ Z : gn = e} is a subgroup of Z with index o(⟨g⟩).

Exercise 59. Let ϕ : G → H be an isomorphism of groups. Show that o(g) = o(ϕ(g)).

The relationship between o(x), o(y), and o(xy) is complicated. One might hope that

o(xy) = o(x)o(y)

—but that’s false in general. For example, o(xx−1 ) = 1 regardless of what o(x) is.

Example 1.6.15 — " #

0 −1
A=
1 0
has order 4 (because A2 = −I), and
" #
0 1
B=
−1 −1

has order 3 (because B3 = I), but

" #" # " #
0 −1 0 1 1 1
AB = =
1 0 −1 −1 0 1

has infinite order.

Here’s a very special case where we can say something.

48
 Chapter 1. Groups and subgroups

Proposition 1.6.16
Let g be an element of finite order. Then

o(g)
o(gk ) =
gcd(o(g), k)

Proof. Let n = o(g), m = o(gk ), and d = gcd(n, k).

On the one hand,
(gk )n/d = gkn/d = gnk/d = (gn )k/d = e,
n
so m | by the Division Lemma.
d
On the other hand,
gkm = (gk )m = e,
so n | km, again by the Division Lemma. But that means

n km k
| = m.
d d d
n k n
Since and are coprime, by the magic of number theory, | m.
d d d

Example 1.6.17 — [2] has order 10 in U(11) because

[2]2 = [4], [2]3 = [8], [2]4 = [5], [2]5 = [10], [2]6 = [9],
[2]7 = [7], [2]8 = [3], [2]9 = [6], and [2]10 = [1].

We can easily compute the order of any other element now. For example,

10 10
o([5]) = o([2]4 ) = = = 5,
gcd(10, 4) 2

10 10
o([7]) = o([2]7 ) = = = 10,
gcd(10, 7) 1
and
10 10
o([10]) = o([2]5 ) = = = 2.
gcd(10, 5) 5

Exercise 60. Show that o(g) = o(g−1 ) for all g in G.

Here’s another very special case where we can say something.

49
 Chapter 1. Groups and subgroups

Proposition 1.6.18
If x and y commute (that is, xy = yx) and have coprime orders, then o(xy) = o(x)o(y).

Proof. Let m = o(x) and n = o(y). Since x and y commute,

(xy)mn = xmn ymn = e.

Thus, by the Division lemma, o(xy) | mn.
To show that o(xy) = nm, let k = o(xy). Then

e = (xy)mk = xmk ymk = ymk

because (xy)k = e and xm = e. By the Division lemma, n = o(y) | mk. Since gcd(n, m) = 1,
we have n | k. By symmetry, m | k.
Now m = o(x) and n = o(y) both divide k = o(xy). Since they’re coprime, their product
divides o(xy) as well. Therefore o(xy) = o(x)o(y).

Exercise 61. Show that o(xy) = o(yx) regardless of whether x and y commute.

50
Chapter 2

Families of groups

2.1 Congruence groups

2.1.1 Z/nZ
As we’ve seen, the additive group of integers modulo n is cyclic:

Z/nZ = ⟨[1]⟩.

But [1] is not the only generator! [−1] = [n − 1] also works, and in concrete cases we can
find many others:

Example 2.1.1 — In Z/12Z we have

⟨[7]⟩ = {[7], [2], [9], . . . , [5], [0]} = Z/12Z

and
⟨[8]⟩ = {[8], [4], [0]} ⪇ Z/12Z
so [7] generates Z/12Z while [8] does not.

Exercise 62. Can you guess, in general, how to tell whether [a] generates Z/nZ?

Proposition 2.1.2
[a] generates Z/nZ iff gcd(a, n) = 1.

Remark 2.1.3. By the Lemma in the handout, gcd(a, n) is independent of the representative—
if [a] = [b], then gcd(a, n) = gcd(b, n).

51
 Chapter 2. Families of groups

Proof. First of all, [a] generates Z/nZ iff o([a]) = n. [Why?] In particular, since Z/nZ =
⟨[1]⟩, we have o([1]) = n. Expressing [a] = a[1] as a “power” of the generator, we can
appeal to the order formula:

o([1]) n
o([a]) = = .
gcd(o([1]), a) gcd(n, a)

It follows that o([a]) = n iff gcd(n, a) = 1.

Example 2.1.4 — The generators of Z/12Z are [1], [5], [7], and [11]. The other classes(’s
representatives) are divisible by 2 or 3.

The number of generators of Z/nZ is therefore φ(n), as defined in the GCD and φ handout.

2.1.2 U(n)
Aside from cyclic groups (like the additive groups Z/nZ) the most important finite abelian
groups are the multiplicative groups U(n). In some sense, studying these particular groups
tells you everything you need to know about finite abelian groups.
Recall that U(n) a.k.a. (Z/nZ)× is the set of invertible residue classes modulo n. As
promised, we’re going to explain
1. how to tell if a given class is invertible, and
2. how to find the inverse of an invertible class.
Definition 2.1.5 — Fix an integer n. A residue class [a] is invertible if there exists [b]
such that [a][b] = [1].

The inverse of [a] is unique if it exists, and we denote it [a]−1 .

Example 2.1.6 — In Z/16Z, [5] and [11] are invertible because [5][13] = [65] = [1] and
[11][3] = [33] = [1]. We have [5]−1 = [13] and [11]−1 = [3].

How do we tell apart invertible and non-invertible classes? Let’s start with a simple
criterion for the latter.

Lemma 2.1.7
If [a][b] = [0] but [a], [b] ̸= [0], then neither [a] nor [b] is invertible modulo n.

52
 Chapter 2. Families of groups

Proof. Suppose [b] had an inverse, say [c]. Then [b][c] = [1]. But [a][b] = [0] by hypothesis.
Since multiplication is associative,

[a][b][c] = [a] and [a][b][c] = [0],

contradicting the assumption that [a] was nonzero.

Example 2.1.8 — Continuing in Z/16Z, neither [4] nor [8] are invertible because
[4][8] = [32] = [0].

As you may have guessed from the last two examples, coprimality with n has something
to do with invertibility.

Lemma 2.1.9
If gcd(a, n) > 1 then [a] is not invertible modulo n.

Proof. Let d be a nontrivial common divisor of a and n. Then the integer b = n/d is not
divisible by n. However, ab = an/d is divisible by n, because the quotient ab/n is the
integer a/d. Thus,
[a][b] = [ab] = [0]
so [a] is not invertible.

Theorem 2.1.10
If gcd(a, n) = 1 then [a] is invertible modulo n.

Proof. Consider the multiplication-by-[a] map,

f : Z/nZ → Z/nZ
[x] 7→ [a][x].

If [a][x] = [a][y], then n divides ax − ay = a(x − y). But since n and a are coprime, that
means n divides x − y, and so [x] = [y]. Thus f is injective.
But! An injection from a finite set to itself must be bijective (by the pigeonhole principle).
In particular, [1] is in the image of f. In other words, there exists some [b] in Z/nZ such
that [a][b] = [1]. That means [a] is invertible!

This fulfills our first promise—[a] is invertible iff gcd(a, n) = 1.

For the second promise—we need a second proof!

53
 Chapter 2. Families of groups

Second proof of the Theorem. Since a and n are coprime, by “Bézout’s Euclidean algorithm,”
there exist integers s and t such that

as + nt = 1.

Reducing this equation modulo n, we obtain

as ≡ 1 (mod n)

or, what is equivalent,

[a][s] = [1].
That means [a] is invertible!

Example 2.1.11 — Let’s compute the inverse of [123] modulo 1024, if it exists. First,
we run the Euclidean algorithm to compute gcd(1024, 123).

1024 = 8 · 123 + 40
123 = 3 · 40 + 3
40 = 13 · 3 + 1

Thus gcd(1024, 123) = 1, so [123] is invertible modulo 1024. To find its inverse, we
work backwards:

1 = 40 − 13 · 3
= 40 − 13 · (123 − 3 · 40) = 40 · 40 − 13 · 123
= 40 · (1024 − 8 · 123) − 13 · 123 = 40 · 1024 − 333 · 123

Thus 40·1024−333·123 = 1, so [−333][123] = [1], meaning that [123]−1 = [−333] = [691].

With these theorems in hand, it’s now clear that the order of the multiplicative group of
integers modulo n is the totient of n:

o(U(n)) = φ(n).

(Yes, so few symbols mean so many words!)

Example 2.1.12 —
U(7) = {[1], [2], [3], [4], [5], [6]}
and
U(12) = {[1], [5], [7], [11]}.

54
 Chapter 2. Families of groups

More generally, for p prime,

U(p) = {[1], . . . , [p − 1]}.

Remark 2.1.13. This result tells us that Z/pZ is actually a field!

We are also in a position to derive a fundamental result in number theory, concerning the
multiplicative orders of invertible residue classes.

Theorem 2.1.14 (Euler–Fermat)

If a and n are coprime, then
aφ(n) ≡ 1 (mod n).

Proof. By Exercise 57, ao(G) = e in any group G. Thus [a]φ(n) = [1] in U(n).
Remark 2.1.15. Compare this proof with the very long proofs you may have seen in
MAT246 or MAT315!

2.2 Cyclic groups

2.2.1 Definitions
Recall that a cyclic group is a group in which every element is an integer power∗ of a single
element, called a generator. We write
G = ⟨g⟩
to mean G is cyclic with generator g.

Example 2.2.1 —
The trivial group is cyclic.
Z = ⟨1⟩ is infinite cyclic.
Z/nZ = ⟨[1]⟩ is finite cyclic of order n.
µn = ⟨e2πi/n ⟩ is also finite cyclic of order n.
Let G be any group and let h ∈ G. Then H = ⟨h⟩ is a cyclic subgroup of G of order
o(g).
⟨2⟩ ≤ Q× is the subgroup of powers of 2:

1 1 1
..., , , , 1, 2, 4, 8, . . .
8 4 2

∗
multiple in additive notation

55
 Chapter 2. Families of groups
√ √
⟨1 + 2⟩ ≤ R× is the subgroup of powers of 1 + 2:
√ √ √ √
. . . , 3 − 2 2, −1 + 2, 1, 1 + 2, 3 + 2 2, . . .

√ √
Exercise 63. Prove that every integer power of 1 + 2 has the form a + b 2 for some
integers a and b.

In certain contexts, the generator of a cyclic group has a special name.

Definition 2.2.2 — A generator for µn is called primitive nth root of unity, while a
(representative of a) generator for U(n) is called a primitive root modulo n.

Example 2.2.3 — Consider some “small” √ groups of roots of unity. In particular, let
(2πi)/8 1 (2πi)/12 1
ζ=e = 2 (1 + i) and ξ = e
√ = 2 ( 3 + i). Then

µ8 = ⟨ζ⟩ = {1, ζ, ζ2 , ζ3 , ζ4 , ζ5 , ζ6 , ζ7 }
= {1, ζ, i, ζ3 , −1, ζ5 , −i, ζ7 }

and

µ12 = ⟨ξ⟩ = {1, ξ, ξ2 , ξ3 , ξ4 , ξ5 , ξ6 , ξ7 , ξ8 , ξ9 , ξ10 , ξ11 }

= {1, ξ, ξ2 , i, ξ4 , ξ5 , −1, ξ7 , ξ8 , −i, ξ10 , ξ11 }

ℑ{z}

i ξ2 ζ
4
ξ ξ
ζ3
ξ5 ℜ{z}
1
−1
ξ11
ξ 7 ζ7
ξ10
ζ5 ξ8 −i

The only nontrivial proper subgroups of µ8 are µ4 = {1, i, −1, −i} and its subgroup
µ2 = {1, −1}. On the other hand, µ12 has several additional subgroups aside from these,
like µ3 = {1, ξ4 , ξ8 } and µ6 = {1, ξ2 , ξ4 , ξ6 , ξ8 , ξ10 }. However,

µ8 ≰ µ12 .

56
 Chapter 2. Families of groups

Observe that all these subgroups are cyclic.

Exercise 64. Show that a complex number ζ can be a primitive nth root of unity for
at most one positive integer n. What is n in terms of ζ?

Here’s the most basic criterion for determining whether a finite group is cyclic.

Proposition 2.2.4
A finite group of order n is cyclic iff it has an element of order n.

Proof. Exercise.

Exercise 65. In the Proposition, why is it necessary that the group be finite?

Example 2.2.5 (A non-cyclic group) — U(8) = {[1], [3], [5], [7]} is not cyclic, as every
element squares to the identity.

Example 2.2.6 (Another non-cyclic group) — The dihedral group Dn has order 2n,
but no element has such large order, because rotations have order at most n, and flips
are involutions. Thus Dn is not cyclic.

Example 2.2.7 (Yet another non-cyclic group) — Fix n > 1. Let G be the set of subsets
of {1, . . . , n} under the △ operation. In Homework 1, you proved G is a group. Today,
we prove G is not cyclic. We have

A2 = A△A = (A \ A) ∪ (A \ A) = ∅ ∪ ∅ = ∅

for all A in G, so no element has order 2n .

There is always at least one group of any given finite order n, namely Cn , the cyclic group
of order n. We say “the”, because...

Proposition 2.2.8
Any two cyclic groups of the same order are isomorphic.

57
 Chapter 2. Families of groups

Proof. Let G = ⟨g⟩ and H = ⟨h⟩ be cyclic groups of order n and define f : G → H by
f(gi ) = hi .
Before we can say anything about f, we have to check that it’s actually well-defined—that
the proposed value on an element is independent of the exponent. So, if gi = gj then i ≡ j
(mod n), so j = i + mn. Thus

hj = hi+mn = hi (hn )m = hi

because o(h) = n. Thus f is well-defined.

Now, f preserves group structure by the exponent laws:

f(gi gj ) = f(gi+j ) = hi+j = hi hj = f(gi )f(gj ).

Moreover, f is injective because if f(gi ) = f(gj ) then hi = hj , so n | i − j; but o(h) = o(g),

so gi−j = e. Since G and H have the same size, f is bijective, hence an isomorphism. Thus
G= ∼ H.

Exercise 66. Show that any two infinite cyclic groups are isomorphic.

2.2.2 Fundamental Theorem of Cyclic Groups

What do subgroups of cyclic groups look like? If G = ⟨g⟩ then certainly, for each integer k,
the cyclic subgroup generated by gk is a subgroup of G, i.e. ⟨gk ⟩ ≤ G.
You’d be hard-pressed to find a subgroup that isn’t cyclic, because...

Every subgroup of a cyclic group is cyclic.

Lemma 2.2.9
Let G = ⟨g⟩ be a cyclic group and let H ≤ G have finite index k. Then H = ⟨gk ⟩.

Proof. First, we prove there exists some positive integer d such that gd ∈ H.
• If H is trivial, then G must be finite, so we may take d = o(G).
• If H is nontrivial, then H contains some nonidentity element gm (as every element of
G looks like this), so we may take d = |m|.
Next, let d be the least positive integer such that gd ∈ H. We claim gm ∈ H iff d | m
(cf. Division Lemma).
(⇐) If d | m, then gm = (gd )m/d ∈ H because gd ∈ H.
(⇒) If gm ∈ H, then m = qd + r for some 0 ≤ r < d, and gr = gm (gd )−q ∈ H. Since d was
least, we get r = 0, so d | m.
58
 Chapter 2. Families of groups

It follows immediately that H = ⟨gd ⟩.

Finally, we prove that d = k. Since k is the index and G may be infinite, we use cosets. The
cosets of
H = {. . . , g−d , e, gd , g2d , g3d , . . . }
are just
gr H = {. . . , gr−d , gr , gr+d , gr+2d , gr+3d , . . . }
for 0 ≤ r < d. Since gr H ̸= gs H for r ̸= s in this range (lest gr−s be in H and d divide r − s)
there are exactly d cosets.

Exercise 67. Exactly one situation is not covered by the Lemma. What is it?

To complete this remaining case, we consider a digression back into indices.

Exercise 68. Let G be a group and K, H be subgroups of G satisfying K ≤ H ≤ G.

Prove that
[G : K] = [G : H][H : K].

Corollary 2.2.10
Let G = ⟨g⟩ be cyclic. If H1 and H2 have the same index in G, then H1 = H2 .

Proof. Let k = [G : H1 ] = [G : H2 ]. If k < ∞, we are done.

If k = ∞, suppose gn ∈ H1 for n ̸= 0. Then ⟨gn ⟩ ≤ H1 and so [G : H1 ] ≤ [G : ⟨gn ⟩] = n. We
must have instead H1 = {e}. Similarly, H2 = {e}.
Remark 2.2.11. Note that this accounts for when the index is ∞.

Example 2.2.12 — Suppose G = ⟨g⟩ has order n. For each d | n, the subgroup ⟨gd ⟩
has order n/d and index d, because
n n
o(⟨gd ⟩) = o(gd ) = =
gcd(n, d) d

and
o(G) n
[G : ⟨gd ⟩] = = = d.
o(⟨gd ⟩) n/d

Example 2.2.13 — Suppose G = ⟨g⟩ has order ∞. For each d ̸= 0, the subgroup ⟨gd ⟩
has order ∞ and index d. What about d = 0?

59
 Chapter 2. Families of groups

The Lemma, its Corollary, and the two Examples show that a cyclic group has exactly one
subgroup of every possible index.∗ What’s interesting is that the converse is true (at least in
the finite case). In fact:

Theorem 2.2.14
Let G be a finite group. If G has at most one subgroup of each index (equivalently,
order), then G is cyclic.

We will leave the proof for later. In the meantime, let’s see an application of the FToCG.

Example 2.2.15 — Recall Bézout’s theorem: if gcd(a, b) = d then there exist integers
s and t such that
as + bt = d.
We proved Bézout’s theorem by way of the Extended Euclidean algorithm. Using the
fact that every subgroup of Z is cyclic, we can give another proof of Bézout’s theorem.
Consider ⟨a, b⟩ which is the subgroup of Z generated by a and b; its elements are
integer combinations of a and b.
Since Z = ⟨1⟩ is cyclic, its subgroup ⟨a, b⟩ must be cyclic as well, so

⟨a, b⟩ = ⟨d⟩

for some integer d (a “power” of the generator 1). Since ⟨d⟩ = ⟨−d⟩, we may take d to
be positive.
Now, since a, b ∈ ⟨d⟩ we have d | a and d | b. To prove that d = gcd(a, b) we use
the other inclusion: d ∈ ⟨a, b⟩ so there exist integers n and m such that d = an + bm.
Then if c | a and c | b then c | an + bm. In (other) words, if c divides a and b then c
divides every integer combination of a and b. In particular, every common divisor of
a and b divides d. Since d divides a and b, d is the greatest common divisor of a and
b.

Exercise 69. What happens in the most degenerate case a = b = 0?

2.2.3 Cyclicity of U(n)

Theorem 2.2.16
U(n) is cyclic iff n = 2, 4, pk , or 2pk for some odd prime p and positive integer k.

∗
Gallian calls this result the Fundamental Theorem of Cyclic Groups, but he only proves it for finite cyclic
groups, and states it in terms of orders instead of indices.

60
 Chapter 2. Families of groups

Remark 2.2.17. We will prove only the “if” direction, leaving the “only if” for PS3.

Proof. U(2) is the trivial group, which is cyclic; and U(4) = {[1], [3]} = ⟨[3]⟩ is cyclic, too.

Let p be an odd prime. We will postpone the cyclicity of U(p) to a homework problem.
For now, let’s assume this fact.

Next up, let k > 1. To show U(pk ) is cyclic, we will take the product of two elements of
orders pk−1 and p − 1 to obtain an element of order pk−1 (p − 1) = o(U(pk )).
The first element is [p + 1]. Observe that, for any n ≥ 1, the binomial theorem says
     
n n n n−1 n n−2 n
(p + 1) = p + p + p ··· + p+1 (*)
1 2 n−1

where  
n n! n(n − 1)(n − 2) . . . (n − i + 1)
= = .
i i!(n − i)! i(i − 1)(i − 2) . . . 1
Choosing n
= pk−1 , we see

that each term in (*)—except the last one—is divisible by p
k
pm pm pm −1
[because i = i i−1 .] Thus
k−1
[p + 1]p = [1] in U(pk )

so o([p + 1]) | pk−1 .

On the other hand, choosing n = pk−2 in (*), every term—except the last two—is divisible
by pk . Thus
k−2
[p + 1]p = [pk−1 + 1] ̸= [1] in U(pk )
so o([p + 1]) > pk−2 . Therefore o([p + 1]) = pk−1 . [If p = 2, then the last three terms remain,
and we cannot make this conclusion; that is why we require p to be odd.]
The second element is obtained in a different manner. Let g be a primitive root modulo
p [This is where we assume the cyclicity of U(p).] and let m be the order of [g] in U(pk ).
Then
gm ≡ 1 (mod p)
m
because p | pk , so p − 1 divides m. If we let h ≡ g p−1 (mod pk ), it follows from the order
formula that [h] has order p − 1 in U(pk ).
Putting these together, we deduce that U(pk ) is generated by [p + 1][h].

Finally, for U(2pk ), let g be an odd primitive root modulo pk . [This is possible precisely
because p is odd—the class of g modulo pk therefore contains representatives of both
parities.] We claim U(2pk ) = ⟨[g]⟩. Indeed, gcd(g, 2pk ) = 1 because gcd(g, pk ) = 1 and
gcd(g, 2) = 1, so [g] is invertible. Note that the elements of ⟨[g]⟩ are distinct modulo 2pk
because the same is true modulo pk . Since φ(2pk ) = φ(pk ), we have ⟨[g]⟩ = U(2pk ).

61
 Chapter 2. Families of groups

Exercise 70. Show that U(2k ) is not cyclic for k ≥ 3 by exhibiting “too many” sub-
groups of order 2.

2.3 Dihedral groups

2.3.1 Definitions
Definition 2.3.1 — A group is called dihedral if it’s generated by a pair of nontrivial
(identity) involutions.

Compare this with the definition of cyclic: generated by a single element.

When we introduced dihedral groups way back, we said they were symmetries of n-gons
under composition. Well, those groups actually are dihedral in this new sense!
To see why, let r be the one-nth clockwise rotation, and let f be the flip over the vertical
axis. Then fr is an involution, because, geometrically, fr is a flip (over the axis through
the bottommost left-side vertex, to be precise). Moreover, f and fr generate every other
symmetry, because f(fr) = r is the minimal rotation. And as we saw previously, a flip and
a minimal rotation are all you need to express every possible flip and rotation.

f fr

f fr

f fr

62
 Chapter 2. Families of groups

Exercise 71. Is the trivial group dihedral?

Equivalent definition
While the “two-involutions” definition is very useful theoretically, the “geometric” defi-
nition is very useful in practice. The following Proposition describes how to canonically
switch back and forth between these two perspectives.

Proposition 2.3.2
Let G be a dihedral group with generating involutions a and b. Then G is generated
by the flip f = a and the rotation r = fb, which are distinct and satisfy frf = r−1 .
Conversely, if a group G is generated by two distinct elements f and r such that
o(f) = 2 and frf = r−1 , then G is dihedral with generating involutions f and fr.
Moreover, passing from one description the another and back again leaves us where
we started.

Proof. We just showed the backward direction. For the forward, we have to show that if
f = a and r = ab then G = ⟨f, r⟩ and f ̸= r and frf = r−1 .
Plainly,
⟨a, ab⟩ ≤ ⟨a, b⟩;
and since a(ab) = b, we have b ∈ ⟨a, ab⟩, so

⟨a, b⟩ ≤ ⟨a, ab⟩.

Thus ⟨a, b⟩ = ⟨a, ab⟩ i.e. G = ⟨f, r⟩.

Next, f = r iff a = ab iff b = e, but b must have order 2. So f ̸= r.
Finally, we show that frf = r−1 . By our definitions of f and r, we have

frf = a(ab)a = a2 ba = ba.

On the other hand, since a and b are self-inverse,

r−1 = (ab)−1 = b−1 a−1 = ba.

Thus frf = r−1 .

To show that these descriptions are equivalent, just note that

(a, b) −→ (a, ab) −→ (a, aab) = (a, b)

and
(f, r) −→ (f, fr) −→ (f, ffr) = (f, r).
63
 Chapter 2. Families of groups

The identity
frf = r−1
is called the fundamental relation between flip and rotation.

f r f

r−1

Since f is self-inverse, we can rewrite this identity as

rf = fr−1 .

It follows that
rk f = fr−k
for all integers k. This enables us to simplify long strings of r’s and f’s—we can move any
rotation past f, at the cost of inverting it.

Exercise 72. Which of the following are trivial for any Dn ?

a) frffrf
b) rffrfrrfr
c) rrrrrfr
d) rfrrfr
e) frfrfrfrf

2.3.2 Order of a dihedral group

Let G = ⟨r, f⟩ be a dihedral group with flip f and rotation r. Then the order of f is 2, but
the order of r could be anything!

Theorem 2.3.3
If o(r) is finite, then o(G) = 2o(r).

Proof. Let n = o(r). The fundamental relation between flip and rotation allows us to write
every element of G as either rk (a rotation) or frk (a flip) for some integer k. Thus

G = {e, r, r2 , . . . , rn−1 , f, fr, fr2 , . . . , frn−1 }

64
 Chapter 2. Families of groups

which means o(G) ≤ 2n. (We don’t know if these are all distinct—the list could contain
repeats.)
For i, j between 0 and n − 1, we have

fri = frj ⇐⇒ ri = rj ⇐⇒ i ≡ j (mod n)

which can only happen if i = j (since 0 ≤ i, j < n). Thus, in our list above, the flips are
distinct from each other, and the rotations are distinct from each other.
Yet are flips and rotations distinct from each other? We know that f ̸= r, but what about fri
and rj ? We have
fri = rj ⇐⇒ f = rj−i ⇐⇒ f ∈ ⟨r⟩.
So to prove that fri ̸= rj , we just have to show that f ̸∈ ⟨r⟩.
Suppose to the contrary that f ∈ ⟨r⟩. Then f commutes with r, so rf = fr. But, by the
fundamental relation, rf = fr−1 . Together, these imply r = r−1 , so r2 = e. Thus r has order
1 or 2. Since f ∈ ⟨r⟩ has order 2, r must have order 2. However, the 2-element cyclic group
⟨r⟩ can only have one element of order 2. Therefore f = r, a contradiction.

We use the notation Dn to denote the dihedral group with a rotation of order n. By the
theorem, o(Dn ) = 2n.

Example 2.3.4 — The symmetry group of an equilateral triangle has 6 elements.

Just like with cyclic groups, we also speak of Dn as the dihedral group of order 2n, because
there is only ever one such group:

Proposition 2.3.5
Any two dihedral groups of the same order are isomorphic.

Proof. Omitted (straightforward, but tedious).

Not-really-dihedral groups
While it’s difficult to draw regular n-gons for n = 1 and 2, that hasn’t stopped crazy
mathematicians group theorists from talking about their symmetry groups, D1 and D2 .

Exercise 73. What are D1 and D2 ? [Use the theorem.]

Proposition 2.3.6
Dn is abelian iff n = 1 or 2.

65
 Chapter 2. Families of groups

Proof. If Dn is abelian, then r and f commute; conversely, if r and f commute, then so does
every pair of elements in the group they generate.
As we saw at the end of the proof of Theorem 2.3.3, fr = rf is equivalent to r2 = e, which
is equivalent to n = 1 or 2.

These two “smallest” dihedral groups are so boring and so unlike their non-abelian bigger
siblings that for the rest of this class we will assume n ≥ 3.

2.3.3 Subgroups of dihedral groups

Last lecture, we proved that cyclicity is a hereditary property: subgroups of cyclic groups
are cyclic. The analogous assertion for dihedral groups is:

Subgroups of dihedral groups are dihedral—or cyclic.

(Doesn’t sound as nice, does it?)

Theorem 2.3.7
Let G be a dihedral group and let H ≤ G. Then H is either
c) cyclic: ⟨rk ⟩ where k = 12 [G : H], or
d) dihedral: ⟨rk , fri ⟩ where k = [G : H] and 0 ≤ i < k.

We will explore the proof of this (at least in the finite dihedral case) in Problem Set 3.

Example 2.3.8 — The subgroup of all rotations in Dn is ⟨r⟩. Its order is n, so its index
is 2. Subgroups of the rotation subgroup are cyclic subgroups of the whole group.

Example 2.3.9 — Consider D8 . By inscribing a square in an octagon∗ we can give an

example of a dihedral subgroup of D8 . If r is the one-eighth turn, then r2 is a quarter
turn. We have fr2 f = r−2 = (r2 )−1 , so the subgroup ⟨r2 , f⟩ is dihedral. Writing out its
elements,
⟨r2 , f⟩ = {e, r2 , r4 , r6 , f, fr2 , fr4 , fr6 }.
Its index is 2.
∗
or an octagon in a square!

Exercise 74. How many distinct subgroups does Dn have?

66
 Chapter 2. Families of groups

2.3.4 The center of a group

When G is abelian, all its elements commute with each other, and life is easy.
What about when G is not abelian? Do we still have any such nice elements?
The groups Dn for n ≥ 3 are the first non-abelian groups that we’re studying systematically.
One way to measure the “non-abelianness” of a group is to ask which elements commute
with everything. Such elements are called central and their union is called the center.

Definition 2.3.10 — Let G be a group. The center of G is the set

Z(G) = {g ∈ G : gh = hg for all h in G}.

Exercise 75. Show that Z(G) ≤ G, with equality if and only if G is abelian.

Example 2.3.11 (Z(Dn )) — Let ri be a rotation. Since rotations commute with rota-
tions, let frj be a flip. Then
(ri )(frj ) = frj−i
whereas
(frj )(ri ) = frj+i .
These are equal if and only if r−i = ri , i.e. ri is an involution. Geometrically, the only
involutive rotation is the 180-degree turn, possibly only when n is even. Algebraically,
r−i = ri implies i ≡ −i (mod n), and if n | 2i then n must be even and i must be n/2.
Either way, rn/2 is the only rotation that commutes with a flip. In particular, no flip
commutes with r when n > 2, so we conclude

⟨e⟩ if n is odd
Z(Dn ) =
⟨rn/2 ⟩ if n is even.

When a group’s center is trivial—like Dn ’s is when n is odd—we say the group is centerless.
" #
a b
Example 2.3.12 (Z(GL2 (R))) — Let A = ∈ Z(GL2 (R)). Since A commutes
c d
" #
0 1
with every invertible matrix, A must commute with the particular matrix B =
1 0

67
 Chapter 2. Families of groups

(which is invertible). We have

" #" # " #
a b 0 1 b a
AB = =
c d 1 0 d c

and " #" # " #

0 1 a b c d
BA = =
1 0 c d a b
so if AB = BA then a = d and b = c.
" #
1 0
Next, consider the non-invertible matrix N = . We have
0 0
" #" # " #
a b 1 0 a 0
AN = =
c d 0 0 c 0

and " #" # " #

1 0 a b a b
NA = = .
0 0 c d 0 0
As
" we# mentioned, N is not invertible so it is not an element of GL2 (R)—but I + N =
2 0
is! If A(I + N) = (I + N)A then AI + AN = IA + NA. Since AI = IA, we have
0 1
AN = NA. Therefore b = c = 0.
Putting these together, we conclude that an invertible matrix A that commutes with
every other invertible matrix must be of the form
" #
a 0
A= = aI
0 a

with a ̸= 0. After verifying that these so-called “scalar matrices” do commute with
everything else, we conclude that Z(GL2 (R)) = {aI : a ̸= 0}.

Exercise 76. Generalize this example to GLn .

2.3.5 Generators and relations

The dihedral group Dn is generated by an involution f and another element r of order n,
satisfying the fundamental relation frf = r−1 .

68
 Chapter 2. Families of groups

A convenient way to encode this information about Dn is to write down its presentation:

Dn = ⟨r, f | rn = e, f2 = e, frf = r−1 ⟩.

In general, a presentation of a group G takes the form

G = ⟨S | R⟩

where S is a set of generators for G and R is a list of relations they satisfy.∗

Implicit in every presentation is the assumption that the relations are “complete” in the
sense that the generators satisfy no other relations except for those that you can prove from
R. Thus, when we write Dn using the presentation above, the relations rn = e and f2 = e
encode that o(r) = n and o(f) = 2. Moreover, it can be shown that you cannot derive the
identity f = r using only the relations f2 = e, rn = e, frf = r−1 .

Exercise 77. Suppose G is generated by the involutions a and b. Write down a

presentation for G.

Exercise 78. Suppose G = ⟨g⟩ is cyclic of order n. Write down a presentation for G.

2.3.6 Infinite dihedral group

You might be tempted to wonder what happens if we let n = ∞ in the presentation of Dn ,

Dn = ⟨r, f | rn = e, f2 = e, frf = r−1 ⟩.

Well, it doesn’t quite make sense to say “r∞ = e”. But it does make sense to say o(r) = ∞,
which means r satisfies no nontrivial relation of the form rn = e. The other relations—
f2 = e and frf = r−1 —have no dependence on n, so they remain. It seems reasonable,
therefore, to define the infinite dihedral group as

D∞ = ⟨r, f | f2 = e, frf = r−1 ⟩.

To justify the name, we prove:

Proposition 2.3.13
The group D∞ is a dihedral group of infinite order.

Proof. D∞ is generated by the involutions f and fr.

∗
In this context, a relation is an equality between two expressions formed using products of elements of S
and their inverses.

69
 Chapter 2. Families of groups

To interpret D∞ geometrically, consider this sequence of nested n-gons, depicted here for
3 ≤ n ≤ 12:

Pointwise, as n → ∞, the sides of the n-gons converge to the black line, and the vertices
converge to the black dots. The rotational symmetry becomes a translational symmetry—
we can shift the black line, left or right, any integer number of side-lengths. The vertical
reflection symmetry remains present all the while—we can flip the “∞-gon” across the
middle, or indeed across the perpendicular to any vertex or midpoint.

70
 Chapter 2. Families of groups

Exercise 79. Show that

Aff(Z) = {x 7→ ax + b : a, b ∈ Z and the map has an inverse with coefficients in Z}

is infinite dihedral.

2.4 Symmetric groups

2.4.1 Fundamentals
Recall from Lecture 4:
Definition 2.4.1 — Let X be a set. The symmetric group on X is the set of all permuta-
tions∗ on X under composition. This group is denoted

SX .
∗
i.e. bijective self-maps

71
 Chapter 2. Families of groups

Remark 2.4.2. Other notations you might encounter include

Sym(X) and SX .
The symmetric group on {1, . . . , n} is denoted Sn while the symmetric group on N is
denoted S∞ .
The elements of a symmetric group are usually denoted by Greek letters: σ [SIGMA], τ
[TAU], α, β, γ, etc. The elements of the underlying set X are sometimes called letters due to
the fact that, historically, arbitrary sets were written like
a, b, c, . . .
i.e. using letters, literally. Consequently you may hear people refer to Sn as “the symmetric
group on n letters” and to its elements as “substitutions”.

Definition 2.4.3 — Let σ be a permutation. A letter i is said to be moved if σ(i) ̸= i

and fixed if σ(i) = i. A transposition is a permutation that moves just two points.

Example 2.4.4 — The identity map ϵ fixes everything, and is the only element in any
symmetric group to do so.

We may regard SX as a group action acting on X by function evaluation:

Example (Example 1.5.13) — SX ↷ X for any set X in the obvious way: σ . x = σ(x).
In particular, Sn acts on {1, ..., n}.
What are the orbits of this action?

2.4.2 Order of the symmetric group

Proposition 2.4.5
If X has n elements, then SX has n! elements.

Proof. Let X = {x1 , . . . , xn }. To write down a permutation of X, we just have to fill in the
bottom row of !
x1 x2 x3 x4 x5

in such a way that each xi appears exactly once (here, n = 5 for concreteness).
Clearly, there are n possible things to put in the first column (under x1 ). Write something
down, like x3 : !
x1 x2 x3 x4 x5
x3
72
 Chapter 2. Families of groups

and move on to the next column.

Since our map must be a permutation, we can’t send x2 to the same place we sent x1 (here,
x3 ). But the remaining possibilities (here, x1 , x2 , x4 , and x5 ) are all available, so there are
n − 1 possible things we can put in the second column (under x2 ). Pick one, say x5 :
!
x1 x2 x3 x4 x5
x3 x5

and move on.

Continuing in this manner, by the time we get to xn , there’ll only be one unused element—
so write it in:
! ! !
x1 x2 x3 x4 x5 x1 x2 x3 x4 x5 x1 x2 x3 x4 x5
, , .
x3 x5 x4 x3 x5 x4 x1 x3 x5 x4 x1 x2

Every permutation of X can be obtained this way. At each stage, the number of choices de-
creases by 1, from n at the start to 1 at the end, so there are n(n − 1) . . . 1 = n! permutations
altogether.

n 1 2 3 4 5 6 7 8 9 10
n! 1 2 6 24 120 720 5040 40320 362880 3628800

Exercise 80. Let X be an infinite set. Show that Sym(X) is infinite. [Hint: Let
x1 , x2 , x3 , . . . be infinitely many distinct elements of X and let x0 be another; define
σi : X → X to do nothing except swap x0 and xi .]

2.4.3 Composing and inverting permutations

Permutations, like any functions, are composed right-to-left: στ sends x to σ(τ(x)), not
τ(σ(x)). When written in two-line notation, function composition can be denoted by
juxtaposition, as in
! ! !
1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6
=
2 4 6 1 3 5 3 6 2 5 1 4 6 5 4 3 2 1

Perhaps a more intuitive way to compose permutations in two-line notation is to (i) stack
them vertically and (ii) rearrange the columns so that the top one’s bottom line matches
the bottom one’s top line; then (iii) the composite permutation is just the mapping defined

73
 Chapter 2. Families of groups

by the outermost lines.

!
1 2 3 4 5
!
1 5 4 3 2 1 2 3 4 5
!=
1 5 4 3 2 2 1 5 4 3
2 1 5 4 3

The inverse of a permutation in two-line notation is obtained by just (i) interchanging the
lines and (ii) sorting the columns so the top line matches the original.
!−1 ! !
a b c x y z c x a y z b a b c x y z
= =
c x a y z b a b c x y z c z a b x y

The order of a permutation is its order as an element of the symmetric group it lives in.

2.4.4 Examples
Example 2.4.6 — Let X = {1, 2, 3, 4}. The functions σ, τ, and ϵ, defined for all i by
σ(i) = 5 − i, τ(i) ≡ i − 1 (mod 4), and ϵ(i) = i, are permutations of X, hence elements
of S4 . In two-line notation, they are written
! !
1 2 3 4 1 2 3 4
σ= τ=
4 3 2 1 4 1 2 3
!
1 2 3 4
ϵ=
1 2 3 4

The orders of σ, τ, and ϵ are 2, 4, and 1.

Example 2.4.7 — The permutation f : N → N defined by


n + 1 n is odd
f(n) =
n − 1 n is even

can be written in two-line notation as

!
1 2 3 4 5 6 7 8 9 ...
f=
2 1 4 3 6 5 8 7 10 . . .

74
 Chapter 2. Families of groups

Similarly, the permutation g : Z → Z defined by


n + 2 n is odd
g(n) =
n − 2 n is even

can be expressed as
!
. . . −2 −1 0 1 2 3 4 5 ...
g=
. . . −4 1 −2 3 0 5 2 7 ...

f is an involution while g has infinite order.

Example 2.4.8 — Label the vertices of a pentagon like so:

1
5 2

4 3

Each symmetry of the pentagon defines a permutation of its vertices, namely by i 7→ j

iff the vertex labelled i goes where the vertex labelled j was. Thus r defines the
permutation !
1 2 3 4 5
ρ=
2 3 4 5 1
while f defines the permutation
!
1 2 3 4 5
ϕ= .
1 5 4 3 2

ρ has order 5 while ϕ has order 2.

Example 2.4.9 — Let G be a group. Then every g in G defines a permutation in

Sym(G) by x 7→ gx, which can be visualized using the Cayley table of G. We showed
in Tutorial 3 Question 2b that this is a bijection.
For example, if G = U(7), writing a instead of [a], we have

75
 Chapter 2. Families of groups

· 1 2 3 4 5 6
1 1 2 3 4 5 6
2 2 4 6 1 3 5
3 3 6 2 5 1 4
4 4 1 5 2 6 3
5 5 3 1 6 4 2
6 6 5 4 3 2 1
Each row can be interpreted as a permutation of the first row, so that e.g. 2 and 3 define
the permutations
! !
1 2 3 4 5 6 1 2 3 4 5 6
and
2 4 6 1 3 5 3 6 2 5 1 4

respectively. What are their orders?

2.4.5 Cayley’s Theorem

Before we go on and talk more about the elements of the symmetric group, the last example
already hints at something important about its subgroups.
The Fundamental Theorems of Cyclic and Dihedral Groups provide complete descriptions of
all these groups’ possible subgroups.
So it’s natural to ask next: what are all the possible subgroups of symmetric groups?

Subgroups of symmetric groups are. . . every group!

If any theorem deserves to be called the Fundamental Theorem of Symmetric Groups, it’s the
following.

Theorem 2.4.10 (Cayley)

Every group is isomorphic to some permutation group∗ .
∗
Recall that a permutation group is any subgroup of a symmetric group.

Proof. [This was previously covered in Tutorial 3 Question 2, but we include the proof for
completeness’ sake.]
Fix a group G. For every g ∈ G, we call the map in Example 2.4.9

fg : G → G, fg (x) = gx.

Tutorial 3 Question 2b gives a proof this is a bijection, but knowing G is a group, the proof
could be simplified greatly:
76
 Chapter 2. Families of groups

• fg (x) = fg (y) iff gx = gy iff x = y, so fg is injective.

• For any y ∈ G, fg (g−1 y) = gg−1 y = y, so fg is surjective.
Now consider the map
F : G → Sym(G), g 7→ fg .

This is an injection because fg = fh only if fg (e) = g = h = fh (e). Thus we may restrict the
codomain to F(G), and automatically get that F is a bijection G → F(G).
Now for any g, h, x ∈ G, we also have

F(gh)(x) = fgh (x) = ghx = fg (hx) = fg (fh (x)) = (F(g) ◦ F(h))(x).

Which tells us F(gh) = F(g) ◦ F(h), and so F is an isomorphism.

Example 2.4.11 — Consider G = D2 . Let g1 , g2 , g3 , g4 be e, r, f, fr respectively.

To quickly determine the corresponding permutations σ1 , σ2 , σ3 , and σ4 , we write the
Cayley table of G in terms of the gi ’s:

· g1 g2 g3 g4
g1 g1 g2 g3 g4
g2 g2 g1 g4 g3
g3 g3 g4 g1 g2
g4 g4 g3 g2 g1
Then our four permutations are
! !
1 2 3 4 1 2 3 4
σ1 = ϵ = , σ2 = ,
1 2 3 4 2 1 4 3
! !
1 2 3 4 1 2 3 4
σ3 = , σ4 =
3 4 1 2 4 3 2 1

which form the order-4 dihedral subgroup ⟨σ2 , σ3 ⟩ in S4 .

2.4.6 Cycles
What do the following permutations (from above) have in common?
! ! !
1 2 3 4 1 2 3 4 5 1 2 3 4 5 6
τ= , ρ= , and “3” =
4 1 2 3 2 3 4 5 1 3 6 2 5 1 4

77
 Chapter 2. Families of groups

If we write i → j to mean σ(i) = j then we can visualize the action of each of these
permutations on their respective letter-sets:
τ : 1 → 4 → 3 → 2 → 1 → ...
ρ : 1 → 2 → 3 → 4 → 5 → 1 → ...

“3” : 1 → 3 → 2 → 6 → 4 → 5 → 1 → . . .
All the elements are in one loop! Permutations which act cyclically like this have a special
name.
Definition 2.4.12 — Let k ≥ 2. A k-cycle, denoted (a1 a2 . . . ak ) where the ai ’s are
distinct, is a permutation that moves ai to ai+1 for 1 ≤ i < k, moves ak to a1 , and fixes
everything else.
An infinite cycle is a permutation denoted

(. . . a−1 a0 a1 a2 . . . )

that moves ai to ai+1 for all i ∈ Z, and fixes everything else.

A cycle is a k-cycle for some k or an infinite cycle.

The three cyclic permutations above can be expressed in cycle notation as

τ = (1 4 3 2)
ρ = (1 2 3 4 5)

“3” = (1 3 2 6 4 5)
Remark 2.4.13. Cycles can be written in multiple ways, namely by “rotating” their innards:
(a1 a2 . . . ak ) = (a2 a3 . . . ak a1 ) = . . . = (ak a1 . . . ak−1 ).
These are all equally valid, but if it makes sense to do so, we prefer to write the smallest
letter first.
Exercise 81. Show that σ is a transposition iff σ is a 2-cycle.

Theorem 2.4.14
k-cycles have order k.

Proof. Let σ = (a1 a2 . . . ak ) be a k-cycle. Then σk is the identity map, because it “moves”
each ai to ai and fixes everything else. But if 1 < i < k, then σi moves a1 to ai+1 , so σi is
not the identity map. Thus o(σ) = k.

78
 Chapter 2. Families of groups

Example 2.4.15 — All nontrivial elements in S3 are cycles:

! ! !
1 2 3 1 2 3 1 2 3
= (2 3), = (1 2), = (1 3),
1 3 2 2 1 3 3 2 1
! !
1 2 3 1 2 3
= (1 2 3), and = (1 3 2).
2 3 1 3 1 2

Exercise 82. If σ = (a1 a2 . . . ak ) is a k-cycle in SX , the cyclic group H generated by σ

has order k.
We may consider the group action of H acting on X by restricting the natural action of
SX to its subgroup H. What are the orbits of this action?

2.4.7 Inverting and composing cycles

Cycles are permutations, so anything you can do to permutations, you can do to cycles.
The catch is you’re not always guaranteed to end up with a cycle.
The inverse of a cycle is obtained by writing it backwards:
(a b c d e)−1 = (e d c b a) = (a e d c b).
The inverse of a cycle is always a cycle.
The product of two cycles is their composition—remember, right-to-left!
!
1 2 3
(1 2)(2 3) = = (1 2 3)
2 3 1
The product of two cycles need not be a cycle.

Example 2.4.16 — The composition of any cycle with its inverse is the identity map,
which is not a cycle.

Example 2.4.17 — The product of transpositions

(1 2)(3 4)

is the permutation !
1 2 3 4
2 1 4 3

79
 Chapter 2. Families of groups

which is not a cycle.

Example 2.4.18 — The product of cycles

(1 2 3)(2 3 4 5)

is the permutation !
1 2 3 4 5
2 1 4 5 3
which is not a cycle.

So, what’s the point of cycles?

2.4.8 Permutations in terms of cycles

Let’s revisit why the permutation
!
1 2 3 4 5
σ=
2 1 4 5 3

is not a cycle. We see that σ swaps 1 and 2 and permutes 3, 4, and 5 cyclically. A cycle is
supposed to move things in one loop, but σ moves things in two loops. So although σ isn’t
a cycle, it is a product of the cycles
(1 2) and (3 4 5)
in the sense that
σ = (1 2)(3 4 5) = (3 4 5)(1 2).

Let’s look at another example. The permutation

!
1 2 3 4 5 6 7 8
τ=
4 7 1 3 8 6 2 5

moves 1 to 4, 4 to 3, and 3 to 1, so there’s a 3-cycle (1 4 3); it moves 2 to 7 and 7 to 2, so

there’s a 2-cycle (2 7); similarly, it swaps 5 and 8, so there’s another 2-cycle (5 8); and 6
isn’t in any cycle, so we ignore it. Thus
τ = (1 4 3)(2 7)(5 8).

Can every permutation be written as a product of cycles? Well, yes...

(1 2 3 4 5) = (1 2)(2 3)(3 4)(4 5) = (1 5)(1 4)(1 3)(1 2) = (1 4 5)(1 2 3)
80
 Chapter 2. Families of groups

...but that’s probably not what you were expecting.

Without any further restrictions, there is no end to the number of possible “cycle decompo-
sitions” of a given permutation. What’s special about our decompositions of σ and τ above
is that the constituent cycles do not interact.

Definition 2.4.19 — Two permutations are disjoint if they do not move any of the
same elements. In other words, σ and τ are disjoint if for all x in X, σ(x) ̸= x implies
τ(x) = x.

Example 2.4.20 — The permutations (1 2) and (3 4) are disjoint, while the permuta-
tions (1 2) and (2 3) are not.

Observe that (1 2)(3 4) = (3 4)(1 2) are the same permutation. On the other hand,
!
1 2 3
(1 2)(2 3) = = (1 2 3)
2 3 1

while !
1 2 3
(2 3)(1 2) = = (1 3 2)
3 1 2
so (1 2)(2 3) ̸= (2 3)(1 2) are not the same permutation.

Lemma 2.4.21
Disjoint permutations commute.

Proof. We want to show that στ = τσ or, equivalently, that

σ(τ(x)) = τ(σ(x))
for all x in X. So, let x ∈ X. There are four cases to consider.
Case 1: σ(x) = x and τ(x) = x. This is easy:
σ(τ(x)) = σ(x) = x and τ(σ(x)) = τ(x) = x.

Case 2: σ(x) ̸= x and τ(x) = x. Here,

σ(τ(x)) = σ(x)
but what about τ(σ(x))? Since σ(x) ̸= x and σ is injective, σ(σ(x)) ̸= σ(x)! Thus σ moves
σ(x). Since σ and τ are disjoint, τ fixes σ(x). In other words,
τ(σ(x)) = σ(x).
81
 Chapter 2. Families of groups

Case 3: σ(x) = x and τ(x) ̸= x. This is just Case 2 with the roles of σ and τ reversed.
Case 4: σ(x) ̸= x and τ(x) ̸= x. Impossible, because σ and τ are disjoint!

Exercise 83. Can non-disjoint permutations commute?

With this terminology in hand, we can now state the most important theorem on permuta-
tions:

2.4.9 Cycle Decomposition Theorem

Theorem 2.4.22 (Cycle Decomposition Theorem)
Every permutation can be written as a product of disjoint cycles. Moreover, the
decomposition is unique up to reordering of factors.

Before we prove this theorem, let’s consider what permutations do to elements in X to

provide some intuition for this idea.
For any σ ∈ SX , consider the following relation:

Proposition 2.4.23
Let X be a set and let σ ∈ SX . Define x ∼ y iff σk (x) = y for some integer k. Then ∼ is
an equivalence relation on X.

Proof. We give two proofs of this fact:

Firstly, this directly follows from the fact that ⟨σ⟩ is a subgroup of SX and hence the action
of H = ⟨σ⟩ acting on X (by restricting the natural action of SX to its subgroup H) yields

σk . x = σk (x).

Which yields precisely this equivalence relation as per Proposition 1.5.2.

Secondly, we also explored a relation similar to this relation in Problem Set 1 Q2:

Problem (PS1Q2). Let f be a self-map on a set X. For x, y ∈ X, define x ∼0 y if and only

if there are some integers n, m ≥ 0, such that fn (x) = fm (y). Then ∼0 is an equivalence
relation.

Now when σ is a permutation, we claim that ∼ and ∼0 define the same relation.
If x ∼ y, we have σk (x) = y = σ0 (y), so x ∼0 y.

82
 Chapter 2. Families of groups

If x ∼0 y, we have σn (x) = σm (y). Without loss of generality, assume m < n and n − m = k.

Then σm (σk (x)) = σm (y). Since σ is injective, we conclude that σk (x) = y, so x ∼ y. We
again conclude that ∼ is an equivalence relation.

The orbit of a particular x ∈ X under this action is denoted ⟨σ⟩x. We call this the orbit of x
under σ∗ , and the set of all orbits the orbits of σ.

Lemma 2.4.24
Let σ ∈ SX and let x ∈ X. Then

⟨σ⟩x = {σk (x) : k ∈ Z}.

Proof. Exercise. [This should be very short.]

Example 2.4.25 — The identity permutation ϵ is the empty product. It generates the
trivial group, which acts on X by
ϵ . x = x.
The orbits of ϵ are the singleton subsets ⟨ϵ⟩x = {x}.

Example 2.4.26 — Let X = {1, . . . , 7}. The cycle σ = (1 2 3 4 5) in S7 generates a cyclic

subgroup of order 5. H = ⟨σ⟩ acts on X by

1 7→ 2 7→ 3 7→ 4 7→ 5 7→ 1, 6 7→ 6, 7 7→ 7.

The orbits of σ are

⟨σ⟩1 = {1, 2, 3, 4, 5}, ⟨σ⟩6 = {6}, and ⟨σ⟩7 = {7}.

Example 2.4.27 — Let X = {1, . . . , 8} and let

!
1 2 3 4 5 6 7 8
τ= = (1 2 5 7)(4 8 6)(3).
2 5 3 8 7 4 1 6

Then the orbits of τ are

⟨τ⟩1 = {1, 2, 5, 7}, ⟨τ⟩4 = {4, 6, 8}, and ⟨τ⟩3 = {3}.

[Isn’t this a lot shorter than the solution we had in PS1Q2c)?]

∗
Instead of “⟨σ⟩”.

83
 Chapter 2. Families of groups

Example 2.4.28 — Multiplication by 2 on U(7) effects the non-cyclic permutation

(1 2 4)(3 6 5)

(a pair of 3-cycles), while multiplication by 3 there is the 6-cycle

(1 3 2 6 4 5).

Example 2.4.29 — On N,

n + 1 n is odd
f(n) =
n − 1 n is even
!
1 2 3 4 5 6 7 8 9 ...
=
2 1 4 3 6 5 8 7 10 . . .
= (1 2)(3 4)(5 6)(7 8) . . .

is the product of infinitely many 2-cycles (transpositions), while on Z,


n + 2 n is odd
g(n) =
n − 2 n is even
!
. . . −2 −1 0 1 2 3 4 5 ...
=
. . . −4 1 −2 3 0 5 2 7 ...
= (. . . 4 2 0 −2 −4 . . . )(. . . −1 1 3 5 7 . . . )

is the product of two ∞-cycles (“chains”).

Example 2.4.30 — Let

[
µ2∞ = µ2n = µ1 ∪ µ2 ∪ µ4 ∪ µ8 ∪ . . .
n≥0

n
be the set of all 2n th roots of unity, i.e. ζ in C such that ζ2 = 1 for some n ≥ 0. Then
κ(z) = z3 is a permutation of µ2∞ . [Exercise.] Let ζ = e2πi/8 and ξ = e2πi/16 be primitive
8th and 16th roots of unity, respectively. Then
!
1 −1 i −i ζ ζ3 ζ5 ζ7 ξ . . .
κ=
1 −1 −i i ζ3 ζ ζ7 ζ5 ξ3 . . .

84
 Chapter 2. Families of groups

= (i −i)(ζ ζ3 )(ζ5 ζ7 )(ξ ξ3 ξ9 ξ11 )(ξ5 ξ15 ξ13 ξ7 ) . . .

is the product of infinitely many finite cycles of increasing length. [The precise cycle
structure of κ is related to the order of [3] in U(2n ).]

Exercise 84. Show that σ fixes x iff ⟨σ⟩x = {x}. Such orbits are called trivial.

Exercise 85. If ⟨σ⟩x has l elements, then

⟨σ⟩x = {x, σ(x), σ2 (x), . . . , σl−1 (x)}.

We may re-define cycles using orbit language now:

Definition 2.4.31 — A cycle is a permutation with just one nontrivial orbit. If that
orbit has size l, the cycle is called an l-cycle and l is called the length.

Remark 2.4.32. ∞-cycles are also called chains because they aren’t really “cycles” in the
literal sense.∗
Remark 2.4.33. Trivial orbits, i.e. orbits of size 1, a.k.a. fixed points, are also called “1-cycles”
(another oxymoron).

Exercise 86. Let σ ∈ Sn be an l-cycle. How many letters does σ fix?

We now have everything we need to properly state and prove the Cycle Decomposition
Theorem.

Theorem 2.4.34 (Cycle Decomposition Theorem – Final Form)

Let X be a set and let σ ∈ SX . Then there exists pairwise-disjoint cycles σi ∈ SX such
that Y
σ= σi .
i

Moreover, this decomposition into cycles is unique up to reordering. That is, if τj ∈ SX

are any other pairwise-disjoint cycles whose product is σ, then there exists a bijection
φ such that τj = σφ(j) for all j.

Remark 2.4.35. Note that the product notation makes sense because disjoint cycles com-
mute.
∗
But if we’re being picky, isn’t the term “infinite cyclic group” literal nonsense as well?

85
 Chapter 2. Families of groups

Proof. Pick one xi from each nontrivial orbit of σ and define σi : X → X by the formula

σ(x) if x ∈ ⟨σ⟩xi
σi (x) =
x otherwise.
You should check that each σi is a bijection (one-to-one and onto). In fact, each σi is a cycle,
because ⟨σ⟩xi is its only nontrivial orbit.
This identity also shows that σi and σj are disjoint when i and j are distinct, because
i ̸= j ⇐⇒ xi ̸= xj ⇐⇒ xi ̸∼ xj ⇐⇒ ⟨σ⟩xi ∩ ⟨σ⟩xj = ∅.

To prove that Y
σi = σ
i
we evaluate both sides at an arbitrary x in X. So, let x ∈ X.
• If σ fixes x, then the σ-orbit of x is trivial, so x ̸∼ xi for any i. Thus σi (x) = x for all i,
and we have Y

σi (x) = x = σ(x).
i

• If σ moves x, then the σ-orbit of x is nontrivial, so x ∼ xi0 for some i0 . Since the xi ’s
are one per orbit, x ̸∼ xi for any other i ̸= i0 . Thus σi0 (x) = σ(x), while σi (x) = x.
Since disjoint permutations commute [and since composition is associative], we have
Y
 Y

σi (x) = σi0 σi (x) = σi0 (x) = σ(x).
i i̸=i0

Q
Thus σ = i σi is the desired decomposition of σ as a product of disjoint cycles.
Q
To show uniqueness, suppose σ = j τj where τj ∈ SX are pairwise disjoint cycles. Since
Q i ) ̸= xi , there must exist a unique j = ψ(i) such that τj (xi ) ̸= xi . [Exists because
σ(x
j τj = σ; unique because τj are Q disjoint.] Since τj is a cycle, it has only one nontrivial
class, which must be ⟨τj ⟩xi . Since j τj = σ, it follows that τj (x) = σ(x) for all x ∼ xi . But
σ(x) = σi (x) for these same x! Therefore, τj = σi . The desired bijection is then φ = ψ−1 .

2.4.10 Cycle type

Now’s as good a time as any to make the following definition.

Definition 2.4.36 — Let σ ∈ SX . For each l = 1, 2, . . . , ∞ let cl be the number of orbits

of size l. The cycle type of σ is the ordered list of the cl ’s.

Remark 2.4.37. When X is finite, say of order n, cycle types are written as n-tuples
(c1 , c2 , c3 , . . . , cn ). When X is infinite, we separate c∞ from the rest of the list with a
semicolon: (c1 , c2 , c3 , . . . ; c∞ ).

86
 Chapter 2. Families of groups

Example 2.4.38 — The cycle type of ϵ in Sn is (n, 0, . . . , 0).

Example 2.4.39 — The cycle types of

!
1 2 3 4 5
σ= = (1 2)(3 4 5) ∈ S5 and
2 1 4 5 3
!
1 2 3 4 5 6 7 8
τ= = (1 4 3)(2 7)(5 8) ∈ S8
4 7 1 3 8 6 2 5

are (0, 1, 1, 0, 0) and (1, 2, 1, 0, 0, 0, 0, 0).

Example 2.4.40 — The cycle types of

“2” = (1 2 4)(3 6 5)

and
“3” = (1 3 2 6 4 5)
in Sym(U(7)) are (0, 0, 2, 0, 0, 0) and (0, 0, 0, 0, 0, 1), respectively.

Example 2.4.41 — The cycle type of


n + 1 n is odd
f(n) =
n − 1 n is even
= (1 2)(3 4)(5 6)(7 8) . . .

on N is (0, ∞, 0, . . . ; 0). On the other hand, the cycle type of


n + 2 n is odd
g(n) =
n − 2 n is even
= (. . . 4 2 0 −2 −4 . . . )(. . . −1 1 3 5 7 . . . )

on Z is (0, 0, 0, . . . ; 2).

Exercise 87. What’s the cycle type of κ(z) = z3 on µ2∞ ?

87
 Chapter 2. Families of groups

Exercise 88. What’s the cycle type of a general cycle?

Proposition 2.4.42
When X is finite, then so is every orbit, and we have
X
lcl = n.
l≥1

Proof. The orbits partition X.

2.4.11 The order of a permutation

We first give a generalization of Proposition 1.6.18:

Proposition 2.4.43
Let G be a group and let a, b ∈ G. If a and b commute and have finite order, then

o(ab) | lcm(o(a), o(b)).

Proof. Write n = o(a) and m = o(b). Since a and b commute,

(ab)lcm(n,m) = alcm(n,m) blcm(n,m) = e.

Thus, by the Division Lemma, o(ab) | lcm(n, m).

Remark 2.4.44. Note that this proof is identical to the first half of Proposition 1.6.18, except
with lcm(n, m) in place of mn.
Note that gcd(n, m) = 1 implies lcm(n, m) = nm.
When G is a symmetric group and a and b are disjoint permutations, the Proposition gets
an upgrade:

Theorem 2.4.45
If σ and τ are disjoint permutations of finite order, then

o(στ) = lcm(o(σ), o(τ))

Proof. Write n = o(σ) and m = o(τ). By the Proposition, we just have to prove that
lcm(n, m) ≤ o(στ).

88
 Chapter 2. Families of groups

Let k = o(στ). Then

σk τk = (στ)k = ϵ
because σ and τ commute. Now, since σ and τ are disjoint, it follows that every power of
σ is disjoint from every power of τ. [Why?]
In particular, σk and τk are disjoint! And if σk moves an element, then τk cannot move it
back; therefore σk and τk must both equal ϵ. By the Division Lemma, n = o(σ) | k and
m = o(τ) | k. That is, k = o(στ) is a common multiple of m and n, and so lcm(n, m) ≤
k.

Exercise 89. Show that the hypothesis of disjointness is necessary.

Corollary 2.4.46
Suppose σ has cycle type (c1 , c2 , . . . ; c∞ ). If c∞ = 0 and cl = 0 for all but finitely many
l, then
o(σ) = lcm{l : cl > 0}.

Proof. By the CDT,

YY
cl
σ= σl,i
l≥1 i=1

where the σl,i are pairwise disjoint cycles of lengths l. By the preceding Theorem, the order
of σ is

o(σ) = lcm{o(σl,i ) : l ≥ 1 and 1 ≤ i ≤ cl }

= lcm{1,
| .{z
. . , 1}, |2, .{z . . , 3}, . . . }
. . , 2}, |3, .{z
c1 c2 c3

= lcm{l : cl > 0}.

Example 2.4.47 —

o (1 2)(3 4 5) = lcm(2, 3) = 6
and

o (1 4 3)(2 7)(5 8) = lcm{3, 2, 2} = 6.

Example 2.4.48 —

o (1 2)(3 4)(5 6)(7 8) . . . ) = lcm{2, 2, 2, 2, . . . } = 2.

89
 Chapter 2. Families of groups

Exercise 90. Suppose σ has cycle type (c1 , c2 , . . . ; c∞ ). Show that o(σ) = ∞ iff c∞ > 0
or cl ̸= 0 for infinitely many l.

2.4.12 What generates Sn ?

We’ve seen the generating sets of Cn and Dn . What about Sn ? [We restrict our attention to
finite symmetric groups because the infinite group is much more complicated.]

Proposition 2.4.49
Any k-cycle (a1 . . . ak ) can be written as a product of transpositions

(a1 . . . ak ) = (a1 a2 ) . . . (ak−1 ak ).

Proof. It suffices to check the right hand side: If i ̸= aj for some j, the right hand fixes i. If
i = aj for j ̸= k, then the right hand side sends it to
aj 7→ aj+1 .
If i = ak , the right hand side sends it to
ak 7→ ak−1 7→ · · · 7→ a1 .
So the right hand side evaluates to the cycle (a1 . . . ak ).

Corollary 2.4.50
Any permutation in Sn can be written as a product of transpositions. That is, Sn is
generated by the transpositions in Sn .

Proof. Exercise.

Proposition 2.4.51
Sn is generated by the transpositions

(1 2), (2 3), . . . , (n − 1 n).

Proof. For any transposition (i j) ∈ Sn with i < j, we observe that

(i j) = (i i + 1 . . . j − 1 j)(j − 1 j − 2 . . . i + 1 i).
Proposition 2.4.49 decomposes the cycles into transpositions swapping adjacent numbers.

90
 Chapter 2. Families of groups

Theorem 2.4.52
Sn is generated by the n-cycle σ = (1 2 . . . n) and the transposition τ = (1 2).

Proof. We claim that

(k k + 1) = σk−1 τσ1−k .
To prove this, we proceed via induction. For k = 1, we are done. Suppose the claim holds
for k = K, then

σ(K+1)−1 τσ1−(K+1) = σ(σk−1 τσ1−k )σ−1

= (1 2 . . . n)(K K + 1)(1 2 . . . n)−1
= (K + 1 K + 2).

Induction tells us σ and τ generate all transpositions in Proposition 2.4.51, so they together
generate Sn .

Exercise 91. Give an example of an n-cycle and a transposition that do not together
generate Sn .

You can find a number of other generating sets of Sn , but these—the set of all transpositions,
the set of adjacent transpositions, and the set {(1 2 . . . n), (1 2)}—are the most commonly
seen.
Exercise 92. For n ≥ 3, show that Sn must be generated by at least two elements.
[That is, Sn is not cyclic.]

91
Chapter 3

Quotients and Morphisms

3.1 Normality and Quotients

We spent a lot of time in the first half of the course talking about groups in their own right,
with a sharp focus on their elements and their subgroups.
In the second part, we turn our attention to how groups are related to other groups,
how new groups are constructed from old ones, and how groups interact with other
mathematical objects.

3.1.1 Recall...
Let H be a subgroup of G.
A left coset of H in G is a subset of the A right coset of H in G is a subset of the
form form
aH = {ah : h ∈ H} Ha = {ha : h ∈ H}
where a ∈ G. where a ∈ G.
Left cosets are precisely the equivalence classes of the left coset relation† : a ∼ b iff b−1 a ∈ H
iff aH = bH. The set of equivalence classes (a.k.a. left cosets) is denoted G/H.

Exercise 93. The space of right cosets of H in G is denoted H\G. What is the analogous
equivalence relation, Ha = Hb iff ... ∈ H?

Remark 3.1.1. When we talk about “cosets” without specifying left or right, we always
mean left cosets.
The significance of these relations being equivalence relations is that two cosets (of the same
type) are either identical or disjoint—cosets cannot partially overlap.
†
Originating from the orbits of the right action of H on G by left multiplication.

92
 Chapter 3. Quotients and Morphisms

Finally, we record the trivial but highly useful observation that aH = H if (and only if)
a ∈ H. In particular, eH = H.
The notation for the coset space sometimes produces familiar objects.

Example 3.1.2 — Let G = Z, H = nZ, then G/H can be written as...

Z/nZ!

Question — Let H ≤ G. When is the coset space G/H itself a group?

3.1.2 First examples

How do we put a group structure on G/H?
The elements of G/H are cosets. Although we could, in theory, put any old composition
law on G/H, the operation on G/H ought to be somehow related to (or derived from) the
operation on G.
How should we combine aH and bH? Well, the most naive way to interpret (aH)(bH) is to
view it as the set of all pairwise products of elements of aH with elements of bH, namely
{ah1 bh2 : h1 , h2 ∈ H}.∗
Let’s see if that works.

Example 3.1.3 — Let G = Z and H = 10Z. Consider the cosets 2 + 10Z and 3 + 10Z.
We have

(2 + 10Z) + (3 + 10Z) = {. . . , −8, 2, 12, . . . } + {. . . , −7, 3, 13, . . . }

= {. . . , −15, −5, 5, 15, 25, . . . }
= 5 + 10Z

which looks just like the addition [2] + [3] = [5] in Z/10Z.

Example 3.1.4 — Let G = R and H = Z. Consider the cosets −0.7 + Z and 2.5 + Z.
We have

(−0.7 + Z) + (2.5 + Z) = {. . . , −0.7, 0.3, 1.3, . . . } + {. . . , −0.5, 0.5, 1.5, . . . }

= {. . . , −1.2, −0.2, 0.8, 1.8, 2.8, . . . }
= 1.8 + Z

∗
After all, we already interpret aH that way.

93
 Chapter 3. Quotients and Morphisms

which looks just like the addition {−0.7} + {2.5} = {1.8} in R/Z.

Notice that, in both examples, the sum of two cosets is again a coset: the “coset of the
sum”.

√3.1.5√— Let G = R and H = ⟨−1⟩. Consider the cosets πH = {π, −π} and
×
Example
√
2H = { 2, − 2}. We have
√ √ √
(πH)( 2H) = {π, −π}{ 2, − 2}
√ √
= {π 2, −π 2}
√
= π 2H.

Example 3.1.6 — Let G = R× and H = (0, ∞). Consider the cosets 1H = H and
−1H = (−∞, 0). We have

(1H)(−1H) = {xy : x > 0, y < 0} = {z : z < 0} = −1H

while
(−1H)(−1H) = {xy : x < 0, y < 0} = {z : z > 0} = 1H.

Notice that, in both examples, the product of two cosets is again a coset—the “coset of the
product”.

How about a non-abelian example?

Example 3.1.7 — Let G = D6 and H = ⟨r2 ⟩ = {e, r2 , r4 }. Consider the cosets fH =

{f, fr2 , fr4 } and rH = {r, r3 , r5 }. We have

(fH)(rH) = {f, fr2 , fr4 }{r, r3 , r5 }

= {fr, fr3 , fr5 ,
fr3 , fr5 , fr7 ,
fr5 , fr7 , fr9 }
= {fr, fr3 , fr5 }
= frH

and again we see that the product of cosets is the “coset of the product”.

Our next and final example serves to dispel the hasty conclusion that multiplying cosets
always works out.

94
 Chapter 3. Quotients and Morphisms

Example 3.1.8 — Let G = Dn for n > 2 and H = ⟨f⟩ = {e, f}. Consider the cosets
rH = {r, rf} = {r, fr−1 } and frH = {fr, frf} = {fr, r−1 }. We have

(rH)(frH) = {r, fr−1 }{fr, r−1 }

= {rfr, rr−1 ,
fr−1 fr, fr−1 r−1 }
= {f, e, r2 , fr−2 }

but this is not a coset of H = {e, f}—it’s too big!

Exercise 94. Suppose m = o(H). Show that m ≤ |(aH)(bH)| ≤ m2 .

3.1.3 When is (aH)(bH) a coset of H?

In general, (aH)(bH) = cH iff
⊆: for all h1 , h2 in H there exists h3 in H such that ah1 bh2 = ch3 ,
and
⊇: for all h3 in H there exist h1 , h2 in H such that ch3 = ah1 bh2 .
If the product of cosets is a coset, which coset is it?

Lemma 3.1.9
Suppose (aH)(bH) = cH. Then c ∼ ab, i.e., cH = abH.

Proof. Certainly a ∈ aH and b ∈ bH, so ab ∈ (aH)(bH). But if (aH)(bH) = cH, then there
exists h in H such that ab = ch, in which case cH = chH = abH.

In other words, if the product of cosets, (aH)(bH), is a coset of H, then it must be the “coset
of the product”, abH. So, when do we have the equality (aH)(bH) = abH?

Lemma 3.1.10
If b−1 hb ∈ H for all h in H, then (aH)(bH) = abH.

Proof. We always have (aH)(bH) ⊇ abH because abh = (ae)(bh) for all h ∈ H. Therefore,
the equality (aH)(bH) = abH holds if and only if

(aH)(bH) ⊆ abH. (1)

95
 Chapter 3. Quotients and Morphisms

On the level of elements, (1) is equivalent to

For all h1 , h2 in H there exists h3 in H such that ah1 bh2 = abh3 . (2)
So, suppose it’s true that b−1 hb ∈ H for all h in H. Then, for all h1 , h2 in H, we have
ah1 bh2 = abb−1 h1 bh2 = ab (b−1 h1 b)(h2 )
| {z }
h3

so (2) holds, proving the Lemma.

Remark 3.1.11. The expression b−1 hb is an instance of conjugation, the act of operating on
an object by an element on one side and its inverse on the other. If two objects are related
by conjugation, then they are called conjugates (of each other).
Remember, our goal is to put a binary operation on the set G/H such that it becomes a
group.
We’ve been attempting to use the most naive operation possible—(aH)(bH) = {ah1 bh2 :
h1 , h2 ∈ H}—to achieve our goal.
The above Lemmas show that in order for coset multiplication to work for all cosets of H
in G, we’d better have b−1 Hb ⊆ H for all b in G.

Definition 3.1.12 — A subgroup H of a group G is called normal if b−1 Hb ⊆ H for all

b in G, in which case we write H ⊴ G.∗
∗
\trianglelefteq

We summarize the above investigations in the following Theorem.

Theorem 3.1.13
Let H ⊴ G. Then G/H is a group under coset multiplication, called the quotient group of
G by H.

Proof. By the Lemmas, coset multiplication is a binary operation on G/H. It’s associative
because


(aH)(bH) (cH) = (abH)(cH) = (ab)cH


= a(bc)H = (aH)(bcH) = (aH) (bH)(cH)
so it’s a composition law. The identity is H itself: for all aH in G/H,
(aH)H = aH;
and the inverse of aH is a−1 H, because
(aH)(a−1 H) = aa−1 H = eH = H.

96
 Chapter 3. Quotients and Morphisms

Exercise 95. The order of the group G/H is ...

Since H is always clear from context∗ it’s common to write a, instead of aH, for a typical
element of the quotient group G/H.

Example 3.1.14 — Every subgroup H of an abelian group G is normal, because

g−1 hg = hg−1 g = h for all h in H and g in G. In particular,
3.1.3 nZ ⊴ Z for all n in Z
3.1.4 Z ⊴ R.
3.1.5 ⟨−1⟩ ⊴ R× .
3.1.6 (0, ∞) ⊴ R× .
The corresponding quotient groups are
∼ Cn )
3.1.3 Z/nZ, the integers modulo n (=
3.1.4 R/Z = T, the “reals modulo 1” (Problem Set 1 Extra Problem 3a)
∼ (0, ∞))
3.1.5 R× /⟨−1⟩, the “unsigned” reals under multiplication (=
∼ C2 )
3.1.6 R× /(0, ∞) = {(0, ∞), (−∞, 0)}, the “group of signs” (=

Taking a quotient is kind of like ignoring the information encoded by the subgroup. In
Z/10Z, we’re ignoring everything but the ones digits. In R/Z, we’re ignoring the integer
parts and just focusing on the fractional parts. In R× /⟨−1⟩, we’re ignoring signs: −x = x
for all x. And in R× /(0, ∞), we’re ignoring magnitude: 10100 = 10−100 = 1.

Example 3.1.15 — Subgroups of nonabelian groups are not necessarily normal:

3.1.7 ⟨r2 ⟩ ⊴ D6 because (frk )−1 r2 (frk ) = r−2 for all k ∈ Z, but
̸ Dn for n > 2 because r−1 fr = fr2 ̸∈ ⟨f⟩.
3.1.8 ⟨f⟩ ⊴
The corresponding quotient group D6 /⟨r2 ⟩ is isomorphic to D2 , because its order is 4—

o(D6 ) 12
o(D6 /⟨r2 ⟩) = [D6 : ⟨r2 ⟩] = = =4
o(⟨r ⟩)
2 3

—and it’s not cyclic—every element squares to the identity:

2
g⟨r2 ⟩ = g2 ⟨r2 ⟩ = ⟨r2 ⟩

∗
It’s right there in the notation G/H!

97
 Chapter 3. Quotients and Morphisms

because g2 ∈ ⟨r2 ⟩ whether g is a flip (g2 = e) or a rotation (g2 = r2k ).

Explicitly, we may write out the elements of D6 /⟨r2 ⟩ in full as

e = ⟨r2 ⟩ = {e, r2 , r4 } r = r⟨r2 ⟩ = {r, r3 , r5 }

f = f⟨r2 ⟩ = {f, fr2 , fr4 } fr = fr⟨r2 ⟩ = {fr, fr3 , fr5 }

Taking the quotient of a finite group by a (proper) nontrivial normal subgroup always
yields a smaller (nontrivial) group. Although quotients may be “slippery” to work with,
they are, in a certain sense, simpler.

3.1.4 Examples of normal subgroups

Example 3.1.16 — For any group G, the trivial subgroup 1 and the group G itself are
always normal subgroups of G. If these are the only two normal subgroups of G, then
∼ G and G/G =
G is called simple.∗ The quotients are G/1 = ∼ 1 respectively.
∗
Alas, the trivial group (which is the simplest group) is not considered a simple group—for the same
reason that 1 (which is the first number) is not considered a prime number.

Example 3.1.17 — Subgroups of abelian groups are normal.

Exercise 96. Are abelian subgroups of nonabelian groups normal?

Exercise 97. Are normal subgroups abelian?

Proposition 3.1.18
The normal subgroups of Dn are Dn , every subgroup of ⟨r⟩, and, just when n is even,
∼ 1, Dn /⟨rk ⟩ =
⟨r2 , f⟩ and ⟨r2 , fr⟩. The corresponding quotients are Dn /Dn = ∼ Dk for each
∼ ∼
k | n, and Dn /⟨r , f⟩ = Dn /⟨r , fr⟩ = C2 .
2 2

∼ D1 , the Proposition may be summarized as saying “quotients of dihedral

Since C2 =
groups are dihedral”. We omit the proof for now.

Exercise 98. Show that quotients of cyclic groups are cyclic.

98
 Chapter 3. Quotients and Morphisms

Example 3.1.19 —

V = ⟨(1 2)(3 4), (1 3)(2 4)⟩ = {ϵ, (1 2)(3 4), (1 3)(2 4), (1 4)(2 3)}

is a normal subgroup of S4 because conjugation preserves cycle type, and V contains

all possible double-transpositions in S4 . The quotient S4 /V has order 4!/4 = 6, so it’s
either D3 or C6 . It cannot be cyclic because S4 has no elements of order 6, so it must be
D3 .

Exercise 99 (PS1XQ3b). Show that every proper subgroup of the quaternion sub-
group Q = ⟨i, j, k⟩ is normal in Q.∗ What are the possible quotients of Q?
∗
Recall that i, j, k are the unit quaternions satisfying i2 = j2 = k2 = ijk = −1.

Exercise 100. Show that subgroups of index 2 are normal.

Example 3.1.20 — Z(G) is a normal subgroups of G.

Example 3.1.21 — Automorphisms of the form cg (x) = gxg−1 are called inner; auto-
morphisms not of this form are called outer.∗ Inner automorphisms constitute a normal
subgroup of Aut(G) [Why?]; the quotient is denoted Out(G).
∗
Some jokesters like to abbreviate “outer automorphism” to “outermorphism”.

Exercise 101. What’s an outer automorphism of Z? Of Z/nZ?

3.1.5 Lingering questions

Let H ≤ G. We tried to put a natural group structure on G/H (the set of left cosets of H in
G), and it worked—but only after making the additional assumption that H was normal in
G, i.e.,
g−1 hg ∈ H for all h in H and g in G.

This raises a few questions.

1. What about right cosets?
2. Is normality necessary?
3. What’s normality all about?

99
 Chapter 3. Quotients and Morphisms

1. Right cosets?
Let’s start with 1.
Had we worked with right cosets instead of left ones, Lemma 2 would have said (Ha)(Hb) =
Hab if aha−1 ∈ H for all h in H, because
h1 ah2 b = h1 ah2 a−1 ab = (h1 )(ah2 a−1 ) ab.
| {z }
∈H

And in order for the multiplication to have worked for all cosets, we would have required
H to satisfy
aha−1 ∈ H for all h in H, for all a in G
or, more succinctly,
aHa−1 ⊆ H for all a in G.
This is equivalent to normality as previously defined, because (a−1 )−1 = a.
So, using right cosets instead of left ones doesn’t change the required “normality” condition.
But, does it change the group?

Proposition 3.1.22 (Equality of Left and Right Cosets)

If H ⊴ G, then aH = Ha for all a in G.

Proof. Suppose H is normal. To show aH = Ha we must show

for all h1 in H there exists h2 in H such that ah1 = h2 a
and vice versa. This is accomplished using the identities
ah1 = ah1 a−1 a and h2 a = a a−1 h2 a
| {z } | {z }
∈H ∈H

where both underbraced elements are in H because H is normal.

Exercise 102. Show that the converse holds, namely, that if Ha = aH for all a in G,
then H is normal.

So, using right cosets instead of left ones doesn’t change the group elements. And it doesn’t
change the group operation, either, as the following Corollary shows.

Corollary 3.1.23
If H ⊴ G, then (aH)(bH) = (Ha)(Hb) for all a, b in G.

Proof. (aH)(bH) = abH = Hab = (Ha)(Hb).

In short, if H is normal, then H\G = G/H. \ (ツ) /

100
 Chapter 3. Quotients and Morphisms

2. Normality—needed?
Next, 2. Is normality necessary? Well, if we want the product of any two cosets to be a
coset, then H has to be normal.

Lemma 3.1.24
Let H ≤ G. If (aH)(bH) ∈ G/H for all a, b in G, then H ⊴ G.

Proof. We know from last lecture that if (aH)(bH) is a coset, then it must be the coset abH.
In particular, for all a in G,

(aH)(a−1 H) = aa−1 H = eH = H.

Thus if h ∈ H and a ∈ G, then aha−1 = (ah)(a−1 e) ∈ H.

3. Normality—what?
Finally, 3. What’s normality all about?
The key idea is conjugacy.
In general, any expression involving g on one side and g−1 on the other is an instance of
conjugation. For example, x and gxg−1 (as well as g−1 xg) are conjugate elements. Similarly,
H and gHg−1 are conjugate subgroups.
Informally, conjugacy can be seen as a “change in perspective”, and, loosely speaking,
normal subgroups are the ones which look the same from all “points of view”.

Example 3.1.25 — Let σ = (1 2 3)(4 5) ∈ S5 and let

!
1 2 3 4 5
γ= .
5 4 3 2 1

Which permutation is γσγ−1 ?

Well, γ−1 (1) = 5, σ(5) = 4, and γ(4) = 2, so γσγ−1 (1) = 2. Similarly,

γ−1 σ γ
2 7−→ 4 7−→ 5 7−→ 1

so γσγ−1 (2) = 1. Continuing in this manner, we get

γσγ−1 = (1 2)(3 5 4).

101
 Chapter 3. Quotients and Morphisms

So what? Well, with a little rewriting...

γσγ−1 = (1 2)(3 5 4) = (5 4 3)(2 1)

which looks exactly like σ = (1 2 3)(4 5) but with the letters replaced by their images
under γ.

Proposition 3.1.26
For all k-cycles and all γ’s,

γ(a1 . . . ak )γ−1 = (γ(a1 ) . . . γ(ak ))

Combined with the homomorphic property of conjugation , we see that conjugacy in Sn

amounts to relabelling.

Example 3.1.27 — In Dn ,
frf−1 = frf = r−1 .
In other words, observing a clockwise rotation from behind makes it look like a
counterclockwise rotation.

from behind after a rotation(still behind)

Similarly,
rfr−1 = fr−2 ,
which means that observing a flip with your head tilted makes it look like a flip over a
tilted axis.

tilting your head watching someone flip

102
 Chapter 3. Quotients and Morphisms

Example 3.1.28 (Change of Basis) — Let V be a vector space with two bases v1 , . . . , vn
and w1 , . . . , wn . Every vector x in V can be uniquely written as a linear combination of
basis vectors:
x = a 1 v1 + a 2 v2 + · · · + a n vn
where a1 , . . . , an are scalars. In other words, x corresponds to the vector (a1 . . . an )T
with respect to the basis v1 , . . . , vn . To express x in terms of the basis w1 , . . . , wn , we
first express each vi :
X
n
vi = bij wj
j=1

Then
X X X X X X
n n n n n
! n
x= a i vi = ai bij wj = ai bij wj =: aj′ wj
i=1 i=1 j=1 j=1 i=1 j=1

so that x corresponds to the vector (a1′ . . . an′ )T with respect to the basis w1 , . . . , wn .
Thus, “change of coordinates” is given by the formula
    
a1′ b11 . . . bn1 a1
 .   . ... ..   .. 
 ..  =  ..
 
   .  .  
′
an b1n . . . bnn an

or, briefly, (ai′ ) = B(ai ) where B is the change of basis matrix.

103
 Chapter 3. Quotients and Morphisms

For example, if we wish to go from the “default” basis v1 = (1, 0), v2 = (0, 1) to the
“custom” basis w1 = (1, −1), w2 = (1, 2), we can use the change of basis matrix:
" #
1 2 −1
B= .
3 1 1

The duck’s beak lies at the tip of the vector 4v1 + 2v2 , which is (4 2)T w.r.t. “default”
coordinates. Where’s the duck’s beak w.r.t. “custom” coordinates?
" # ! ! !
1 2 −1 4 1 8−2 2
= =
3 1 1 2 3 4+2 2

Indeed, one can check directly that 4v1 + 2v2 = 2w1 + 2w2 . So far, so good.
Now imagine we have a linear operator∗ T on V, given by some matrix Told w.r.t. the
first basis v1 , . . . , vn . To represent T completely in terms of the second basis w1 , . . . , wn ,
we need a matrix Tnew which
(1) accepts input in new coordinates,
(2) changes it back to old coordinates,
(3) performs the operation of Told , and
(4) returns an answer in new coordinates.
In other words,
Tnew = BTold B−1
is the conjugate of Told by B.

104
 Chapter 3. Quotients and Morphisms

The duck leaves.

For example, let T be a 90-degree counterclockwise rotation. In the “default” basis, the
matrix of T is " # " #
cos π2 −sin π2 0 −1
Told = = .
sin π2 cos π2 1 0
We can express T entirely in the “custom” basis by conjugating Told by B:
" #" #" # " #
1 2 −1 0 −1 1 1 1 1 −5
Tnew = BTold B−1 = = .
3 1 1 1 0 −1 2 3 2 −1
∗
“Linear operator” is to “endomorphism” as “linear transformation” is to “homomorphism”.

Thus, revolving the duck about the origin moves the tip of its beak to
" # ! ! !
1 1 −5 2 1 2 − 10 −8/3
= = .
3 2 −1 2 3 4−2 2/3

Conjugacy as automorphism
For each g in G, the function cg (x) = gxg−1 is an automorphism of G. [This is PS2Q2.]
If G is abelian, then cg (x) = gxg−1 = gg−1 x = ex = x for all x in G, so each cg is the
identity map—conjugation doesn’t do anything. So, conjugation is only interesting in the
nonabelian setting.
Like any automorphism, conjugation moves subgroups to other subgroups. [Why?] Since
every subgroup contains e, the distinct conjugates of a given subgroup always overlap.∗
∗
Do not confuse conjugates and cosets. The distinct cosets of a given subgroup never overlap.

105
 Chapter 3. Quotients and Morphisms

For example, here are the conjugates of ⟨f⟩ in D3 and of ⟨r3 , f⟩ in D6 :

r f fr3
r f r 2 r2
4
r r5

e e r3
fr fr2 fr fr5
fr4 fr2

Meanwhile, ⟨r2 , f⟩ ⊴ D4 , so all its conjugates are itself:

r f fr

e
2
r fr2
r3 fr3

The next Proposition proves that the final picture is accurate—normal subgroups are
self-conjugate.

Proposition 3.1.29
Let H ⊴ G. Then gHg−1 = H for all g in G.

Proof. Exercise. [Show that H ⊆ gHg−1 for all g in G.]

Exercise 103. Let H ≤ G and let f ∈ Aut(G) such that f(H) ≤ H. Is it necessarily true
that f(H) = H?

3.1.6 Summary of normality

We summarize our findings in this handy theorem:

106
 Chapter 3. Quotients and Morphisms

Theorem 3.1.30
Let G be a group and let H ≤ G. The following are equivalent.
1. H ⊴ G.
2. gH = Hg for all g in G.
3. G/H is a group under coset multiplication.
4. H is self-conjugate: gHg−1 = H for all g in G.

107