Ia Groups

Julia Goedecke (jg352)

Michaelmas 2016, last updated 16th January 2017

 Contents

Preamble 1
The Notes 1
How to use the notes 1
How to treat references in the notes 1
Exercises 1
Blog 1

Chapter 1. Groups and Homomorphisms 3

Addition modulo 3 3
Symmetries of a square 3
Group axioms 4
Some simple properties 5
Subgroups 6
Small detour about functions 7
Group homomorphisms 8
Images and Kernels 9
Cyclic groups 10
Cartesian products of groups 11
Dihedral groups 12

Chapter 2. Symmetric Groups 14

Definitions 14
Cycle notation 15
The sign of a permutation 17

Chapter 3. Lagrange Theorem 19

Cosets 19
The Theorem 19
Lagrange Corollaries 20
Small detour on equivalence relations 20
Applications of Lagrange 21
Left or right cosets 23

Chapter 4. Infinite Transformation Groups 24

Frieze groups 24
Matrix groups 26
Möbius group 29

Chapter 5. Quotient groups 32

Normal subgroups 32
Quotients 33
The Isomorphism Theorem 34

Chapter 6. Group actions 37

iii
CHAPTER 0. CONTENTS

Groups acting on sets 37

Orbits and Stabilisers 38
Left multiplication actions 40
Chapter 7. Conjugacy 42
Conjugation action 42
Conjugacy classes in Sn 43
Conjugacy in An 45
Conjugation used in the Möbius group 46
Conjugation action on subgroups 47
Chapter 8. Actions in action 48
Using actions 48
Polyhedron symmetry groups 50
Permutation properties of Möbius maps 51
Chapter 9. Quaternions 54
Groups of order 8 54
Quaternions 54

Chapter 10. Geometric properties of orthogonal matrices 56

Reminder of orthogonal matrices 56
Two-dimensional rotations and reflections 56
Three-dimensional rotations and reflections 57
Chapter 11. Möbius maps and cross ratio 58
Summary of results about Möbius maps 58
Cross ratio 58
Summary of conjugacy classes 59

IA Groups Page iv c Julia Goedecke, Michaelmas 2016

 Preamble

The Notes
These notes are not verbatim what I will write in the lectures, but the content is exactly the
same. The main difference is that they have more complete sentences, and they may have some
comments that I only said in lectures. I might try to make such comments green, but it may not
be consistent.
If you find any errors and typos in the notes, please do let me know (jg352), even if they look
trivial.

How to use the notes

You can read ahead of lectures, you can use them for revision, you can use them to look up
a little detail which you can’t figure out from your copy of the lectures, and probably in many
more ways. Last year’s students told me they liked using these notes when doing example sheets,
because they are (automatically) searchable. It is up to you to find out how they are most useful.
If you don’t like taking notes at all in lectures, use these. If you (like me) find that taking down
notes in lectures is actually the best way to learn something, set these notes aside for a while and
use them just to fill in gaps later. If you try to read these notes while I’m lecturing the same
material, you may get confused and probably won’t hear what I say. So my advice is to follow one
of these three possibilities: take notes in lectures; read ahead and then listen in lectures knowing
you’ve already seen it in the notes; listen in lectures hoping that it will all be written in the notes.

How to treat references in the notes

There are two types of references: referring back to something we’ve already done, or referring
forward to something that we will do later in this course, or you will do later in a different course.
The first kind are obvious: if we need to use a result or example, we like to be able to look back
to remember all the details. The second kind you can ignore on first reading if you like. They are
meant to put things in context and whet your appetite for more maths :-). You might be interested
in them later, for example when you revise the course or learn more in other courses. Think of
them as cross-references.
All references to places inside the text are hyperlinked for your convenience.

Exercises
Sometimes in the notes I will say “exercise”. The main reason to have these is for you to be
able to check your understanding by doing a fairly straight-forward exercise yourself. Example
sheet questions usually go a bit further and require thinking, whereas such exercises in the text
should be easy if you’re comfortable with the material, and a good starting point to get comfortable
with the material if you are not quite yet.

Blog
You will find summaries, some suggestions for understanding, and some links to interesting
material outside the course on my course blog. [Link] However, it is last
1
CHAPTER 0. PREAMBLE

year’s blog, so the order of material will be different in some places. For this year, I am writing up
the tips about how to learn maths at university in the blog.

IA Groups Page 2 c Julia Goedecke, Michaelmas 2016

 CHAPTER 1

Groups and Homomorphisms

Addition modulo 3
(or any whole number n).

An example of modular arithmetic is reading the clock. For whole numbers x, y and n we write

x”y pmod nq ô px ´ yq “ kn for some whole number k.

So, we treat two numbers as “the same” (or congruent) modulo n if we can subtract or add a
multiple of n from the first number to get the second number. We could imagine it more easily
on a clock: when we have reached n, we start again at 0. So the only important numbers are
really the numbers from 0 to n ´ 1 (because we can always reach one of these through addition or
subtraction of n). We can write into a table how we add these numbers (see below).

Addition modulo 3
`3 0 1 2
0 0 1 2
1 1 2 0
2 2 0 1

We only really use the numbers 0, . . . , n ´ 1.

If we compare usual addition with this addition mod n, we might come up with the following
properties:
˛ adding two numbers gives another number between 0, . . . , n ´ 1.
˛ 0`something “ the same something (0 ` a “ a).
˛ We can always find a number to add that will give 0 as result.
˛ It does not matter how we set brackets.
˛ The order of adding numbers doesn’t matter.

Symmetries of a square
As a second example we will look at symmetries of a square.

First we have to be clear about the meaning of the word symmetry. What we mean by this
is a (continuous) mapping of (the plane or space containing) a geometrical object which sends
the object onto itself. That is, the object looks the same afterwards (though the numbers of the
corners may have changed). We don’t know formally what “continuous” means yet. For now it’s
enough to know that it means we cannot “tear apart” the geometric object and stick it together
in a different way. Points which are close to each other/next to each other in the original shape
still have to be close to each other/next to each other after the mapping.
Let’s just draw all symmetries of a square that we can think of.
3
CHAPTER 1. GROUPS AND HOMOMORPHISMS

4 3 3 2 2 1 1 4

90˝ 180˝ 270˝

1 2 4 1 3 4 2 3

3 4 1 2 2 3 4 1

2 1 4 3 1 4 3 2

So we see there are two different kinds of symmetries: rotations and reflections. We will count
the rotation angle anti-clockwise (that is just a convention). But how do we know we have written
down all symmetries? Perhaps we missed some!
We can argue this way: if we take the corner labelled 1, we can map it to one of four corners.
Once we have decided that, the corner labelled 2 has to go to a corner next to where we mapped
1, so there are two possibilities. Once we have decided that as well, everything else is fixed and we
have no further choices. So we see that we should get 4 ¨ 2 “ 8 different symmetries, and we have
indeed listed them all.
Remember this and compare when we do group actions in Chapter 6, Section Orbits and
Stabilisers.
Properties:
˛ doing two symmetries after another gives another.
˛ “do nothing” before or after a symmetry doesn’t change it.
˛ For each symmetry, there is one that “undoes” it.
˛ It does not matter how we set brackets. (Harder here: try examples yourself.)
˛ Does the order matter? It does!
4 3 3 4 1 3

1 2 2 1 2 4

4 3 2 3 3 2

1 2 1 4 4 1

As you can see, these give different answers: the top combination gives rotation by 2700 , and
the bottom combination gives 900 rotation.

We will use these common properties to define an abstract structure which we can work with
instead of these specific examples. For that we need to know:
An operation is a way of combining two elements to get a new element.
For example, n ` m in Z, or a `3 b pmod 3q, or v ˝r (composition).
[More formally: it is a function X ˆ X ÝÑ X. Come back to this when you’ve done functions.]

Group axioms
Definition: A group is a set G with an operation ˚ satisfying the following axioms:
0. for all a, b P G, we have a ˚ b P G; (closure)
1. there is e P G such that for all a P G, a ˚ e “ a “ e ˚ a; (identity)
2. for each a P G there is a´1 P G such that a ˚ a´1 “ e “ a´1 ˚ a; (inverses)
3. for all a, b, c P G, we have a ˚ pb ˚ cq “ pa ˚ bq ˚ c. (associativity)
IA Groups Page 4 c Julia Goedecke, Michaelmas 2016
CHAPTER 1. GROUPS AND HOMOMORPHISMS

If we also have
4. for all a, b P G, a ˚ b “ b ˚ a, (commutativity)
then the group is called abelian.

[Strictly speaking we have to say “there exists e such that all of these axioms hold”.]

Examples: a) Integers Z with ` form a group.

b) Rationals Q with ` form a group.
c) Zn , intergers mod n, with `n form a group.
d) Q˚ “ Qzt0u with ¨ form a group. [Check all the axioms: Practice Sheet A]
e) t1, ´1u with ¨ forms a group:
¨ 1 ´1
1 1 ´1
´1 ´1 1
f) Symmetries of a regular triangle with composition form a group. We can use any regular
n-gon: c.f. dihedral group D2n later, e.g. D8 is group of symmetries of square.
g) 2 ˆ 2 invertible matrices with matrix multiplication form a group. [Check all the axioms.]
This group is called GL2 pRq (see Chapter 4).
h) Symmetry groups of 3-dimensional objects such as cube, terahedron, toblerone box, ...
under composition.

Which of these are abelian groups?

Counterexamples: a) Z with ¨: what is the inverse of 3? Axiom 2 fails.

b) Q with ¨: what is the inverse of 0? Axiom 2 fails.
c) Z4 zt0u with ¨4 : what is 2 ¨4 2? Axiom 0 fails.
d) Q˚ “ Qzt0u with division: Can you find an identity Axiom 1 fails.
e) Z with ´: is p3 ´ 4q ´ 2 the same as 3 ´ p4 ´ 2q? Axiom 3 fails.
2
f) Z with “to the power of”: p23 q2 “ 82 “ 64 but 2p3 q “ 29 “ 512. Axiom 3 fails.

[Look at examples and counterexamples on the Practice Sheets.]

Some simple properties

1 Proposition: Let pG, ˚q be a group. Then
(i) The identity is unique.
(ii) Inverses are unique.

Proof. (i) Suppose both e and e1 are identities in G. Then e ˚ e1 “ e1 as e is an identity,

and e ˚ e “ e as e1 is an identity. So e “ e1 .
1

(ii) Suppose both a´1 and b satisfy Axiom 2 for a P G. Then b “ b ˚ e “ b ˚ pa ˚ a´1 q “
pb ˚ aq ˚ a´1 “ e ˚ a´1 “ a´1 . Here we are using in order: Axiom 1, Axiom 2, Axiom 3,
Axiom 2, Axiom 1.


2 Proposition: Let pG, ˚q be a group, and let a, b P G. Then

(i) pa´1 q´1 “ a “a is the inverse of its inverse.”
(ii) pa ˚ bq´1 “ b´1 ˚ a´1 “socks and shoes”.

Proof. (i) Given a´1 , both a and pa´1 q´1 satisfy x˚a´1 “ e “ a´1 ˚x, so by uniqueness
of inverses, a “ pa´1 q´1 .
IA Groups Page 5 c Julia Goedecke, Michaelmas 2016
CHAPTER 1. GROUPS AND HOMOMORPHISMS

(ii) pa ˚ bq ˚ pb´1 ˚ a´1 q “ a ˚ pb ˚ pb´1 ˚ a´1 qq“ a ˚ ppb ˚ b´1 q ˚ a´1 q“ a ˚ pe ˚ a´1 q “ a ˚ a´1 “ e.
Similarly pb´1 ˚ a´1 q ˚ pa ˚ bq “ e, so by uniqueness of inverses, pa ˚ bq´1 “ b´1 ˚ a´1 .


Remarks: ˛ From now on we will use that “Associativity means we can leave out brackets”.
c.f. Practice Sheet B Q3.
˛ If clear from context, we leave out ˚: e.g. pabq´1 “ b´1 a´1 .
˛ We often just write G instead of pG, ˚q if the operation is clear.

Definition: A group pG, ˚q is a finite group if the set G has finitely many elements. Then the
order of G is |G|, the number of elements of G.

Exercise: Which of our examples are finite groups?

Subgroups
Let pG, ˚q be a group throughout.

Definition: A subgroup pH, ˚q ď pG, ˚q (or H ď G) is a subset H Ď G such that H with the
restricted operation ˚ from G is also a group. If H ď G and H ‰ G, we call H a proper subgroup.

Examples: a) pZ, `q ď pQ, `q ď pR, `q ď pC, `q

b) We always have pteu, ˚q ď pG, ˚q (trivial subgroup) and pG, ˚q ď pG, ˚q.
c) pt1, ´1u, ¨q ď pQ˚ , ¨q.

3 Lemma: (“Usual subgroup criterion”)

Let pG, ˚q be a group and let H Ď G be a subset. Then H ď G if and only if
(i) e P H, and “identity is in H”
(ii) for a, b P H, also a ˚ b P H, and “closed under ˚”
(iii) for a P H, also a´1 P H. “closed under inverses”

Proof. We have to use the group axioms 0.–3., applied to H. First note that associativity in
H is inherited from G, meaning Axiom 3 holds in H as soon as it holds in G. (So we don’t have
to check it.)
Clearly (i) ñ Axiom 1, (ii) ñ Axiom 0 and (iii) ñ Axiom 2.
Conversely, suppose pH, ˚q is also a group. Does eH have to be the same as the identity e P G?
Yes: eH ˚ eH “ eH in H, so also in G. In G, eH has some inverse, so eH ˚ eH ˚ e´1 H “ eH ˚ eH ,
´1

which gives eH “ e. So Axiom 1 ñ (i). Also Axiom 2 ñ (iii) by uniqueness of inverses in G, and
Axiom 0 ñ (ii) easily. 

Examples: d) The rotations of a square form a subgroup of D8 (all symmetries of the

square). Check: (i) clearly true, (ii) two rotations give another rotation, (iii) inverse is
the rotation in the opposite direction (or 360˝ ´ first rotation).
e) Even numbers form a subgroup of the integers: 2Z ď Z. (i) 0 is even, (ii) 2a`2b “ 2pa`bq,
(iii) ´2a “ 2p´aq.

4 Lemma: (“Super-efficient subgroup criterion”)

Let pG, ˚q be a group and H Ď G. Then pH, ˚q ď pG, ˚q if and only if
I. H is not empty and
II. given a, b P H, also a ˚ b´1 P H.

IA Groups Page 6 c Julia Goedecke, Michaelmas 2016

CHAPTER 1. GROUPS AND HOMOMORPHISMS

Proof. “ñ” If H is a subgroup, then e P H, so H is not empty. Combining closure under

inverses and multiplication gives II.
“ð” We show conditions (i), (ii), (iii) of the usual subgroup criterion (Lemma 3). By I, there
is some h P H, so using II on h, h we get e “ h ˚ h´1 P H. Now for all a P H, use II on e, a to get
e ˚ a´1 “ a´1 P H. Finally, for a, b P H, we have just shown that b´1 P H, so use II on a, b´1 to
get a ˚ b “ a ˚ pb´1 q´1 P H. 
Note that it is hardly ever efficient to use this one in an actual example. The efficiency comes
in when we are dealing with very general situations; if you’re working in an example, to work out
a ˚ b´1 you usually have to work out b´1 and can already see if it is in the subgroup given. So in
examples, it is mostly more sensible to use the usual subgroup criterion.

5 Proposition: (“subgroups of Z”)

The subgroups of pZ, `q are exactly nZ for n P Z.

Proof. For any n P Z, nZ is a subgroup: 0 P nZ, and for a, b P nZ, we have a “ na1 , b “ nb1
for a1 , b1 P Z, so a ´ b “ npa1 ´ b1 q P nZ, so by the super-efficient subgroup criterion (Lemma 4),
nZ is a subgroup.
Conversely, let H ď Z. We know 0 P H. If H “ t0u, it is 0Z. Otherwise pick n to be
the smallest positive integer in H. We show H “ nZ. Suppose a P H is not divisible by n, so
a “ nk ` a1 , with a1 P t1, . . . , n ´ 1u. But as H is a subgroup, nk “ n ` n ` n ` . . . ` n P H,
so a ´ nk “ a1 P H, contradicting the fact that n is the smallest positive integer in H. So every
a P H is divisible by n, i.e. H “ nZ. 

Small detour about functions

You’ll do this also in N+S.
Given two sets X and Y , a function f : X ÝÑ Y sends each x P X to a particular f pxq P Y .
X is the domain or source, Y is the codomain or target of f .

Examples: ˛ 1X : X ÝÑ X with 1X pxq “ x identity function

(sometimes written id)
˛ ι : Z ÝÑ Q with ιpnq “ n inclusion map
˛ f1 : Z ÝÑ Z with f1 pnq “ n ` 1
˛ f2 : Z ÝÑ Z with f2 pnq “ 2n
˛ f3 : Z ÝÑ Z with f3 pnq “ n2
˛ g1 : t0, 1, 2, 3, 4u ÝÑ t0, 1, 2, 3, 4u with g1 pxq “ x ` 1 if x ă 4, g1 p4q “ 4.
˛ g2 : t0, 1, 2, 3, 4u ÝÑ t0, 1, 2, 3, 4u with g2 pxq “ x ` 1 if x ă 4, g2 p4q “ 0.
˛ h : C ÝÑ C with hpzq “ az`b d
cz`d for given a, b, c, d P C. (Note: not really defined on ´ c , so
we should perhaps exclude that point from the domain. But we see a different way round
that later.) ´ ¯
α “ ac db : R2 ÝÑ R2 with αp xy q “ ax`by
` ˘
˛ cx`dy .

‚ ‚ ‚ ‚
‚ ‚ ‚ ‚
‚ ‚ ‚ ‚
‚ ‚ ‚ ‚
‚ g1
‚ ‚ g2
‚

Functions can be composed (applied one after another): if f : X ÝÑ Y and g : Y ÝÑ Z, then

g ˝f : X ÝÑ Z is defined by g ˝f pxq “ gpf pxqq. For example, from above: f2 ˝f1 pnq “ 2n ` 2,
f1 ˝f2 pnq “ 2n ` 1.
IA Groups Page 7 c Julia Goedecke, Michaelmas 2016
CHAPTER 1. GROUPS AND HOMOMORPHISMS

Two functions f, g are the same (f “ g) if they have the same source X and same target Y ,
and for all x P X we have f pxq “ gpxq.
A function f is injective if it “hits everything at most once”: if f pxq “ f pyq then x “ y. A
function f is surjective (or onto) if it “hits everything at least once”: for all y P Y there is x P X
with f pxq “ y. A function f : X ÝÑ Y is bijective if it is injective and surjective (i.e. it “hits
everything exactly once”).

Examples: ι and f2 are injective but not surjective. f3 and g1 are neither, 1X , f1 and g2 are
bijective. h is constant when ad ´ bc “ 0 [check it!], but when ad ´ bc ‰ 0, then h is injective. Is
it surjective?

Bijective functions f : X ÝÑ Y have inverses: f ´1 : Y ÝÑ X with f ˝f ´1 “ 1Y and f ´1 ˝f “

1X .

Exercise: Write down the inverses to 1X , f1 and g2 . If X is a finite set, prove f : X ÝÑ X is

surjective if and only if it is injective.

6 Lemma: The composite of bijective functions is bijective.

Proof. Exercise or see N+S. 

Group homomorphisms
We are interested in functions/maps that “preserve” or “respect” the group operation.

Definition: Let pG, ˚q and pH, ‹q be groups. Then f : G ÝÑ H is a group homomorphism (or
just homomorphism) if for all a, b P G we have f pa ˚ bq “ f paq ‹ f pbq.

In words: It does not matter if we first multiply in G and then send the answer to H, or first
send each element to H and then multiply them there.

Definition: A homomorphism that is also a bijective function is an isomorphism.

Examples: a) 1G : G ÝÑ G and ι : Z ÝÑ Q are group homomorphisms. So is f2 : Z ÝÑ Z.

The other previous examples are not (find out why). Which of the above are isomor-
phisms?
b) exp : pR, `q ÝÑ pR` , ¨q with exppxq “ ex is a group homomorphism:
exppx ` yq “ ex`y “ ex ey “ exppxq exppyq
Is it injective? Surjective? What about exp : pR, `q ÝÑ pR˚ , ¨q?
π 3π
c) Take pZ4 , `4 q and H “ t1, ei 2 , eiπ , ei 2 u with complex multiplication (4th roots of unity).
aπ
Define f : Z4 ÝÑ H by f paq “ ei 2 . Exercise: Show this is an isomorphism (i.e. a bijective
group homomorphism).
d) f : GL2 pRq ÝÑ R˚ with f pAq “ det A is a homomorphism: detpABq “ det A det B.

Definition: Two groups are called isomorphic if there is an isomorphism between them. We
write G – H and think of them as “the same” group.

7 Proposition: (“Properties of group homs”)

(i) Group homomorphisms send the identity to the identity: f peq “ e.
(ii) Group homomorphisms send inverses to inverses: f pa´1 q “ f paq´1 .
(iii) The composite of two group homomorphisms is a group homomorphism.
IA Groups Page 8 c Julia Goedecke, Michaelmas 2016
CHAPTER 1. GROUPS AND HOMOMORPHISMS

(iv) The composite of two isomorphisms is an isomorphism.

(v) The inverse of an isomorphism is an isomorphism.

Proof. (i) Let f : pG, ˚q ÝÑ pH, ‹q be a group homomorphism. Then we have f peG q “
f peG ˚ eG q “ f peG q ‹ f peG q. Now f peG q P H has an inverse, so multiplying by this we
get:
f peG q´1 ‹ f peG q “ f peG q´1 ‹ f peG q ‹ f peG q
ô eH “ f peG q.
(ii) Exercise.
(iii) Let f : G1 ÝÑ G2 and g : G2 ÝÑ G3 be group homomorphisms. Then for a, b P G1 , we
have
g ˝f pa ˚1 bq “ gpf pa ˚1 bqq “ gpf paq ˚2 f pbqq as f is group homomorphism
“ gpf paqq ˚3 gpf pbqq as g is group homomorphism
“ g ˝f paq ˚3 g ˝f pbq.
(iv) Combine (iii) and Lemma 6: the composite of bijective group homomorphisms is still a
group homomorphism and bijective.
(v) Let f : G ÝÑ H be an isomorphism. So it is a bijection and has an inverse function
f ´1 : H ÝÑ G. We show that f ´1 is also a group homomorphism (and so an isomorphism,
as it is bijective).
Take x, y P H. Then there are a, b P G with f paq “ x and f pbq “ y because f is
bijective. Then f ´1 px ˚ yq “ f ´1 pf paq ˚ f pbqq “ f ´1 pf pa ˚ bqq “ a ˚ b “ f ´1 pxq ˚ f ´1 pyq.


Images and Kernels

Definition: If f : G ÝÑ H is a group homomorphism, then the image of f
Imf “ tb P H | Da P G s.t. f paq “ bu “ “f pGq” “ tf paq | a P Gu
is the set of elements in H “hit by f ”;
and the kernel of f
Ker f “ ta P G | f paq “ eH u “ “f ´1 pteH uq”
is the set of elements in G which are mapped to the identity (the preimage of teH u).

8 Proposition: (“kernels and images are subgroups”)

Let f : G ÝÑ H be a group homomorphism. Then the image is a subgroup of H and the kernel is
a subgroup of G. In symbols: Imf ď H, Ker f ď G.

Proof. We use the “super-efficient subgroup criterion” (Lemma 4) and “properties of group
homs” (Proposition 7).
I. Imf is not empty, as f peq “ e, so eH P Imf . Similarly eG P Ker f .
II. If b1 , b2 P Imf , then there exist a1 , a2 P G with f pai q “ bi . So
f pa1 ˚ a´1
2 q “ f pa1 q ˚ f pa2 q
´1
“ b1 ˚ b´1
2 P Imf.

II for kernel: Exercise.


In fact, kernels are special subgroups:
Given f : G ÝÑ H, a P G, k P Ker f , we have f pa ˚ k ˚ a´1 q “ f paq ˚ f pkq ˚ f paq´1 “
f paq ˚ e ˚ f paq´1 “ e, so a ˚ k ˚ a´1 P Ker f also. Subgroups with this property are called normal
subgroups: see later (beginning of Chapter 5).
IA Groups Page 9 c Julia Goedecke, Michaelmas 2016
CHAPTER 1. GROUPS AND HOMOMORPHISMS

Examples: a) The identity 1G : G ÝÑ G has image Im1G “ G and kernel Ker 1G “ teu.
The inclusion map ι : Z ÝÑ Q has Imι “ Z and Ker ι “ teu. The map f2 : Z ÝÑ Z with
f2 pnq “ 2n has Imf2 “ 2Z and Ker f2 “ teu.
b) The exponential map exp : pR, `q ÝÑ pR` , ¨q has Impexpq “ R` and Kerpexpq “ t0u.
c) The homomorphism f : Z4 ÝÑ H as previously defined has Imf “ H and Ker f “ t1u.
d) The determinant homomorphism det : GL2 pRq ÝÑ pR˚ , ¨q has Impdetq “ R˚ and kernel
Kerpdetq “ tall matrices with determinant 1u“ SL2 pRq. (see later, Chapter 4 Section
Matrix groups)

This suggests some relationship between images and kernels on the one hand and injectivity
and surjectivity on the other hand.

9 Proposition: (“injectivity via kernels”)

A group homomorphism f : G ÝÑ H is
(i) surjective if and only if Imf “ H
(ii) injective if and only if Ker f “ teu.

Proof. (i) Clear by definition.

(ii) “ñ” f peq “ e, so if k P Ker f , then f pkq “ e, and so k “ e.
“ð” Given a, b s.t. f paq “ f pbq, then f pab´1 q “ e, so ab´1 “ e, so a “ b.


Cyclic groups
Notation: Write a2 “ a ˚ a, an “ a ˚ ¨ ¨ ¨ ˚ a (n factors of a), a0 “ e, a´n “ pa´1 qn “ pan q´1 .

Definition: A group pG, ˚q is cyclic if there is an element a P G such that all elements of G are
powers of a. Such an a is called a generator.

Examples: a) pZ, `q is cyclic, with generator 1 or ´1. Generators are not unique.
This is “the infinite cyclic group” ÝÑ cf “essential uniqueness of cyclic groups”
(Lemma 48 in Chapter 5).
b) t`1, ´1u with ¨ is cyclic with generator ´1.
c) pZn , `n q (integers mod n) is cyclic. 1 is a generator, but there are several other generators,
depending on n.
n “ 3: 1, 2 are generators (“2 ” ´1”)
n “ 4: 1, 3 are generators (“3 ” ´1”)
n “ 5: 1, 2, 3, 4 are generators.
d) Rotations of triangle: cyclic generated by rotation of 1200 or 2400 .
e) Similarly rotations of any regular n-gon give a cyclic group of n elements. This is the
“symmetries viewpoint” of cyclic groups. You get the same symmetries viewpoint for
pZn , `n q if you view modular arithmetic with the clock-image.

Remark: How could you view Z as a group of symmetries of something? Take an infinite line
with little diagonal lines in equal distances (e.g. every centimetre). Then the group of translation
symmetries of this line is exactly Z.

Notation: Given a group G and a P G, we write xay for the cyclic subgroup generated by a,
which is defined to be the smallest subgroup containing a.

Remark: The subgroup generated by a is in fact the set of all powers of a. To see this, we first
check that the set of all powers of a is a subgroup:
I. e “ a0 is a power of a, so the set is non-empty.
IA Groups Page 10 c Julia Goedecke, Michaelmas 2016
CHAPTER 1. GROUPS AND HOMOMORPHISMS

II. Given am , an , then am´n is also a power of a,

so by the super-efficient subgroup criterion, the set of power of a is a subgroup of G.
Now if a is in any subgroup, then by closure a ˚ a “ a2 is in the subgroup as well, and so by
induction all positive powers of a. But the inverse a´1 is also in the subgroup, so also all negative
powers of a. So the set of powers of a is the smallest subgroup containing a.

So xay is a subgroup by definition, we don’t have to prove that. We did have to prove that it
is the same as the set of powers of a, but now we can use that as well since we’ve proved it.

Definition: The order of an element a P G is the smallest positive k P N s.t. ak “ e. If no such

k exists, then a as infinite order. Write ordpaq for the order of a.

10 Lemma: (“The order is the size of the cyclic subgroup.”)

For a P G, ordpaq “ |xay|.

Proof. If ordpaq “ 8, then an ‰ e for any n P Z, so an ‰ am for all n ‰ m. Therefore xay

has infinite order.
If ordpaq “ k ă 8, then ak “ e, so xay “ ta0 “ e, a, a2 , . . . , ak´1 u because ak “ e “ a0 ,
k`1
a “ a, etc, and a´1 “ ak´1 etc. So |xay| “ k. 

We write Cn for “the” cyclic group of order n (in multiplicative notation). So e.g. Cn – pZn , `q.
c.f. Lemma 48 Section The Isomorphism Theorem, Chapter 5.

Examples: ˛ Every element in Z (other than 0) has infinite order.

˛ 1 P Zn has order n, for any n. But say 2 P Z4 has order 2, and 2 P Z6 has order 3.
˛ e P G always has order 1.

11 Proposition: Cyclic groups are abelian.

Proof. Exercise. 

Remark: Let G be a group, a1 , . . . , ak P G. Then xa1 , . . . , ak y is the subgroup generated by

a1 , . . . , ak , the smallest subgroup of G containing all ai .

Exercise: Any subgroup of a cyclic group is cyclic.

Cartesian products of groups

Given two groups pG1 , ˚1 q, pG2 , ˚2 q, we can define a group operation on the set G1 ˆ G2 “
tpa1 , a2 q | ai P Gi u (the set of ordered pairs).

Definition: The product G1 ˆ G2 of two groups G1 , G2 is the group with componentwise multi-
plication pa1 , a2 q ˚ pb1 , b2 q “ pa1 ˚1 b1 , a2 ˚2 b2 q.

Exercise: Check that it really is a group!

Examples: Combining some groups we have met:

˛ Zˆt1, ´1u has elements pn, 1q and pn, ´1q for any n P Z. “Componentwise multiplication”
here means that we add in the first component (that is the group operation of Z) and
multiply in the second component (that is the group operation of t1, ´1u). The only
elements with finite orders are p0, 1q and p0, ´1q. [Exercise: what are their orders?
IA Groups Page 11 c Julia Goedecke, Michaelmas 2016
CHAPTER 1. GROUPS AND HOMOMORPHISMS

˛
C2 ˆ C2 “ tpe, eq, pa, eq, pe, bq, pa, bqu
– t e, x, y, xyu
Clearly pa, eq ˚ pe, bq “ pe, bq ˚ pa, eq, i.e. yx “ xy. All non-identity elements have order 2.
So this is an abelian group with two generators of order 2.
˛ Here something happens that didn’t in the example above:
C2 ˆ C3 “ tpe, eq, pa, eq, pe, bq, pa, bq, pe, b2 q, pa, b2 qu
– t e, x3 , x4 , x, x2 , x5 u
– C6
Note that we always have pa1 , eq ˚ pe, a2 q “ pe, a2 q ˚ pa1 , eq. “Everything in G1 commutes with
everything in G2 .”

Comparing the two examples of products of cyclic groups, we may wonder if there is a general
pattern to this. And indeed there is.

12 Proposition: (“products of cyclic groups”)

Cn ˆ Cm – Cnm if and only if n, m are coprime.

Proof. Let Cn be generated by a (so an “ e) and Cm be generated by b. We have pa, bqk “

pa , bk q “ e if and only if ak “ e and bk “ e. If n, m are coprime, this happens for the first time
k

with k “ nm. So the element pa, bq has order nm, and therefore Cn ˆ Cm – Cnm .
If n, m have a common factor l, say n “ n1 l and m “ m1 l, then for any element par , bs q P
1 1
Cn ˆ Cm , we have par , bs qn m l “ pe, eq, and so there is no element of order nm. So Cn ˆ Cm cannot
be (isomorphic to) Cnm . 

We now look at some conditions we could check to see whether a group might be the direct
product of two groups. This result is a little bit abstract, but can come in very useful in certain
examples, for example on the example sheet.

13 Proposition: (“Direct Product Theorem”)

Let H1 , H2 ď G such that
(i) H1 X H2 “ teu (“they intersect only in e”)
(ii) for all a1 P H1 , a2 P H2 , have a1 a2 “ a2 a1 (“H1 and H2 commute in G”)
(iii) for all a P G, there are ai P Hi with a “ a1 a2 (“G “ H1 H2 ” internal product)
Then G – H1 ˆ H2 .

Proof. Define f : H1 ˆ H2 ÝÑ G by f ppa1 , a2 qq “ a1 a2 . This is a group homomorphism:

f ppa1 , a2 q ˚ pb1 , b2 qq “ f ppa1 b1 , a2 b2 qq “ a1 b1 a2 b2 “ a1 a2 b1 b2 “ f ppa1 , a2 qqf ppb1 , b2 qq using (ii). By
(iii) f is surjective. We use (i) to show that f is injective: if f ppa1 , a2 qq “ e, then a1 “ a´1 2 P
H1 X H2 . So a1 “ e “ a2 , so Ker f “ teH1 ˆH2 u. So by “injectivity via kernels” (Proposition 9), f
is injective. Therefore f is a bijective group homomorphism, so an isomorphism. 

Dihedral groups
Recall the symmetries of a regular n-gon. There are n rotations and n reflections. The rotations
0
are generated by the rotation of 360
n , let us call it r. This has order n. Any reflection has order 2.
Choose your favourite one, and call it s. In fact, r and s generate the whole group of symmetries.

The dihedral group D2n is the group of symmetries of a regular n-gon; it is generated by an
element r of order n and an element s of order 2 which satisfy srs´1 “ r´1 (or sr “ r´1 s).
IA Groups Page 12 c Julia Goedecke, Michaelmas 2016
CHAPTER 1. GROUPS AND HOMOMORPHISMS

As a set, D2n “ te, r, r2 , . . . , rn´1 , s, rs, r2 s, . . . , rn´1 su. Here we use for example sr “ r´1 s “
n´1
r s and srk “ r´k s “ rn´k s. We can swap all powers of r to the front, using the relationship
between s and r.
It is useful to have both the geometric viewpoint as symmetries or a regular n-gon and the
algebraic viewpoint as elements made out of these generators and using the given relations for any
calculations.

Exercise: Show that each rk s has order 2. (These are the reflections).

CAREFUL: D6 fl C3 ˆ C2 and D2n fl Cn ˆ C2 !!!!! (for n ą 2)

Because: Cn ˆ C2 is always abelian, D2n is not abelian for n ą 2.

Remark: There are funny border-cases as well: if you think what you would get non-geometrically
if you use the description with generating elements which satisfy some relationship, you can say:
˛ D2 – C2 : here n “ 1 so the “r” doesn’t really exist properly.
˛ D4 – C2 ˆ C2 : if you look at the relationship sr “ r´1 s and both r and s have order 2,
then r´1 “ r, so we get exactly an abelian group with two generators of order 2, as we’ve
seen before.
˛ D8 also exists: here the r has infinite order. See Section 4 about how to think of that.
How can we think of these geometrically? We have to make some slight changes to our pictures
if we want to fit these into the pattern of symmetries of n-gons. Essentially there are two options
(that I can think of):
˛ Fiddle with the “n”: you could think of the D2 as a symmetry group of a line segment
(with end points): identity, and “flipped”. And you could think of the D4 as the symmetry
of a rectangle. I find this slightly unsatisfactory, because it does not fit into the regular
n-gon pattern.
˛ Add some extra “decorations” which are invisible in the usual n-gons: think of each
vertex having an anti-clockwise direction around it. In the usual n-gon (for n ě 3),
that is automatic by the anti-clockwise numbering of vertices which we start off with.
Then any rotation leaves these “orientations” in their anti-clockwise direction, while any
reflection flips them to be clock-wise. Notice that the orientations of all vertices always
have to be the same. Then you can think of D2 as the symmetry group of a 1-gon: an
oriented point. Identity and “flip the orientation”. And D4 becomes the symmetry group
of a 2-gon, a line with oriented end-points as vertices. Identity and 180˝ rotation leave
the orientations the same, but one has the vertices the other way round. Reflection in the
vertical to the line, or reflection in the actual line, both flip the orientation, but one leaves
the vertices where they are (reflection in the line itself) and one also flips the vertices
(reflection in the vertical). Pictures in Lectures

IA Groups Page 13 c Julia Goedecke, Michaelmas 2016

 CHAPTER 2

Symmetric Groups

Definitions
Definition: A bijection from a set X to itself is also called a permutation (of X). The set of all
permutations on X is SymX.

14 Theorem: SymX with composition forms a group.

Proof. 0. If σ : X ÝÑ X and τ : X ÝÑ X, then σ ˝τ : X ÝÑ X. If σ, τ are both bijec-

tions, the composite is also bijective (cf. Lemma 6). So if σ, τ P SymX, also σ ˝τ P SymX.
1. The identity 1X : X ÝÑ X is clearly a permutation, and gives the identity element.
2. Every bijection has an inverse function which is also bijective, so if σ P SymX then
σ ´1 : X ÝÑ X is also in SymX and satisfies the axiom for the group inverse.
3. Composition of functions is always associative. (See Sheet A)


Remark: X can be an infinite set!

Definition: If X is finite, say |X| “ n, we usually use X “ t1, 2, . . . , nu and write SymX “ Sn .
This is the symmetric group of degree n.

Notation: (“two row notation”) « ff « ff

1 2 3 1 2 3 4 5
Write 1, . . . , n on the top line and their images below. E.g. P S3 or P
2 3 1 2 1 3 4 5
« ff
1 2 ¨¨¨ n
S5 . Generally, if σ : X ÝÑ X, write σ “ .
σp1q σp2q ¨ ¨ ¨ σpnq

Remark: Don’t get too used to this notation, we’ll get a better one soon.

Tricks: When composing, reorder the second element. E.g.:

» fi
« ff « ff 1 2 3 « ff
1 2 3 1 2 3 1 2 3
“ –2 1 3fl “
˝
— ffi
2 3 1 2 1 3 3 2 1
3 2 1
‚ ‚ ‚ ‚ ‚ ‚

or think of strings ‚ ‚ ‚ “

‚ ‚ ‚ ‚ ‚ ‚

Exercise: Prove that |Sn | “ n!. (i.e. degree ‰ order)

14
CHAPTER 2. SYMMETRIC GROUPS

Examples: (Examples
#« ff+ for small n) #« ff « ff+
1 1 2 1 2
n“1 S1 “ “ teup– C1 q n“2 S2 “ , – C2
1 1 2 2 1
#« ff « ff « ff « ff « ff « ff+
1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
n“3 S3 “ , , , , ,
1 2 3 2 3 1 3 1 2 2 1 3 3 2 1 1 3 2
Compare S3 to the symmetries of a regular triangle, i.e. D6 . What do you find?

Exercise: Write out all subgroups of S3 . (You should get 1 of order 3, 3 of order 2, plus the trivial
ones.)

‚ ‚ ‚ ‚ ‚ ‚

Note that S3 is not abelian: ‚ ‚ ‚ ‰ ‚ ‚ ‚

‚ ‚ ‚ ‚ ‚ ‚
We can view S3 as a subgroup of any Sn for n ě 3: fix 4, 5, . . . , n.
We can also view D2n as a subgroup of Sn . Take a regular n-gon and number the vertices:
every symmetry of the regular n-gon is also a bijection t1, 2, . . . , nu ÝÑ t1, 2, . . . , nu.
For example D8 ď S4 :
« ff « ff « ff « ff
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
D8 – te, “ r, “ r2 , “ r3 , “ s,
2 3 4 1 3 4 1 2 4 1 2 3 3 2 1 4
« ff « ff « ff
1 2 3 4 1 2 3 4 2 1 2 3 4
“ rs, “ r s, “ r3 su
4 3 2 1 1 4 3 2 2 1 4 3
Here r is rotation by 900 anti-clockwise, and s is reflection in a diagonal.

Cycle notation
Example: S3
˛ e “nothing moves”
˛ p123q “ p231q “ p312q “1 goes to 2, 2 goes to 3, 3 goes to 1”
˛ p132q
˛ p12q leave out numbers that don’t move
˛ p13q
˛ p23q

Advantages: It is easy to find the order of such a cycle. (Find all orders in S3 .).
Inverses: p123q´1 “ p321q “ p132q. Write backwards (then cycle round to get smallest number at
the front).
Composition (from right to left): p123qp12q “ p13qp2q. Look at 1 in right-most cycle. It goes to
2. Then look at 2 in the next cycle moving to the left: there 2 goes to 3. So alltogether 1 goes
to 3 and we write the 3 down. Then we look at 3. There is no 3 in the right-most cycle. In the
next cycle, 3 goes to 1. We already have 1 at the beginning, so instead of writing it again we just
close the bracket. Then we take the next number that we haven’t written down yet, 2. In the
right-most cycle, 2 goes to 1. In the next cycle, 1 goes to 2. So alltogether 2 goes to 2, so we close
the bracket. This will make most sense when you see it demonstrated in lectures.

Convention: We tend to write the smallest number at the front of the cycle.
IA Groups Page 15 c Julia Goedecke, Michaelmas 2016
CHAPTER 2. SYMMETRIC GROUPS

In S4 : p12qp34q is one element, but p12q and p34q are also separate elements. p1234qp14q “
p1qp234q.

Definition: We call pa1 a2 ¨ ¨ ¨ ak q a k-cycle. 2-cycles pa1 a2 q are also called transpositions. Two
cycles are disjoint if no number appears in both.

Example: p12q and p34q are disjoint, but p123q, p12q are not, and p123q, p34q are not.

15 Lemma: Disjoint cycles commute.

Proof. Let σ, τ P Sn be disjoint cycles. We must prove: for all a P t1, . . . , nu, we have
σpτ paqq “ τ pσpaqq. There are three cases:
˛ a in neither cycle is easy: σpτ paqq “ σpaq “ a “ τ paq “ τ pσpaqq.
˛ a in σ, not in τ : this means σpaq “ b ‰ a, τ paq “ a. Note that as σ and τ are disjoint, b is
also not in the cycle τ , so also τ pbq “ b. Then σpτ paqq “ σpaq “ b and τ pσpaqq “ τ pbq “ b.
˛ a in τ , not in σ: analogous.


Note that non-disjoint cycles may not commute: p13qp23q “ p132q but p23qp13q “ p123q.

16 Theorem: (“disjoint cycle notation works”)

For n ě 2, every permutation in Sn can be written (essentially uniquely) as a product of disjoint
cycles.

Proof. Essentially uniquely means: the order of disjoint cycles doesn’t matter, the “rotation”
of individual cycles doesn’t matter.
Let σ P Sn . We start with p1 σp1q σ 2 p1q σ 3 p1q ¨ ¨ ¨ q. As t1, . . . , nu is finite, for some k we
must have σ k p1q already in the list. In fact, if σ k p1q “ σ l p1q with l ă k, then (as σ is a bijection)
σ k´l p1q “ σ l´l p1q “ 1. So actually all σ i p1q are distinct until for some k we get σ k p1q “ 1 again.
So the first cycle is p1 σp1q σ 2 p1q ¨ ¨ ¨ σ k´1 p1qq. Then pick the smallest (or any) number which
does not appear in this cycle, say j P t1, . . . , nuzt1, σp1q, . . . , σ k´1 p1qu. Repeat to get the second
cycle pj σpjq ¨ ¨ ¨ σ l´1 pjqq. As σ is a bijection, these two cycles are disjoint. We repeat until we
have exhausted all t1, . . . , nu.
Why is this essentially unique? The order of disjoint cycles doesn’t matter because they
commute. Any j completely determines pjσpjq ¨ ¨ ¨ q because σ is a function. 

We can substitute “unique” for “essentially unique” if we cycle the smallest element to the
front in each cycle and then order the cycles by the size of the first number.
Look back at this proof when we’ve done orbits in Chapter 6, especially Lemma 52.

Definition: Writing a permutation σ P Sn in disjoint cycle notation, the list of cycle lengths is
called the cycle type of σ.

Note: Of course only up to reordering. We often (but not always) leave out singleton cycles.

Examples: p12q has cycle type 2 (a transposition).

p12qp34q has cycle type 2, 2 (double transposition).
p123qp45q has cycle type 3, 2, etc.

It is easy to see that a k-cycle has order k.

IA Groups Page 16 c Julia Goedecke, Michaelmas 2016
CHAPTER 2. SYMMETRIC GROUPS

17 Lemma: (“order by cycle type”)

For σ P Sn , the order of σ is the least common multiple (lcm) of the different cycle lengths in
disjoint cycle notation.

Proof. As disjoint cycles commute (Lemma 15), we can group each cycle together when
taking powers: if σ “ τ1 τ2 ¨ ¨ ¨ τl , with the τi all disjoint cycles, then σ m “ τ1m τ2m ¨ ¨ ¨ τlm .
If cycle τi has length ki , then τiki “ e, and τim “ e if and only if ki |m. So to get an m such
that σ m “ e, we need all ki to divide m, i.e. we need lcmpk1 , . . . , kl q|m. So the order, which is the
smallest such m, is lcmpk1 , . . . , kl q. 

Examples: Any transposition has order 2. p12qp34q also has order 2. p123qp45q has order 6.

The sign of a permutation

Permutations come in two different types: even and odd permutations. These have something
in common with even and odd integers, and also something in common with positive and negative
numbers.

18 Proposition: (“Sn is generated by transpositions”)

Every permutation is a product of transpositions.

Proof. By disjoint cycle notation, it is enough to do this for one cycle.

pa1 a2 ¨ ¨ ¨ ak q “ pa1 a2 qpa2 a3 q ¨ ¨ ¨ pak´1 ak q, which is a product of k ´ 1 transpositions. [Check
the product.] 
We want to use this for the sign of a permutation.
BUT it is not unique, e.g.
p12345q “ p12qp23qp34qp45q “ p12qp23qp12qp34qp12qp45q “ p15qp14qp13qp12q “ ¨ ¨ ¨
We want to find some property of such decompositions into transpositions that is invariant, i.e.
only depends on the permutation we started off with.

19 Theorem: (“sign is well-defined”)

Writing σ P Sn as a product of transpositions in different ways, σ is either always the product of
an even number of transpositions or always the product of an odd number of transpositions.

Idea of proof:
˛ Start off with something which is determined entirely by a given permutation: number
of disjoint cycles.
˛ Show this number changes parity when the permutation is multiplied by a transposition.
˛ When making a given permutation out of transpositions: the parity of “number of dis-
joint cycles” changes for each transposition. But end result is fixed, so parity of no of
transpositions needed must be fixed.

Proof. Write #pσq for the number of cycles in disjoint cycle notation, including singleton
cycles. So #peq “ n, #pp12qq “ n ´ 1 etc. What happens if we multiply σ by a transposition
τ “ pcdq? (wlog c ă d). Clearly composing with τ does not affect any cycles not containing c or d.
˛ If c, d are in the same σ-cycle: Say
pc a2 a3 ¨ ¨ ¨ ak´1 d ak`1 ¨ ¨ ¨ ak`l qpcdq “ pc ak`1 ak`2 ¨ ¨ ¨ ak`l qpd a2 a3 ¨ ¨ ¨ ak´1 q
So #pστ q “ #pσq ` 1.
˛ If c, d are in different cycles (this could be cycles of length 1), then
pc a2 a3 ¨ ¨ ¨ ak´1 qpd b2 b3 ¨ ¨ ¨ bl´1 qpcdq “ pc b2 b3 ¨ ¨ ¨ bl´1 d a2 ¨ ¨ ¨ ak´1 q
IA Groups Page 17 c Julia Goedecke, Michaelmas 2016
CHAPTER 2. SYMMETRIC GROUPS

So #pστ q “ #pσq ´ 1.
So for any transposition τ , #pστ q ” #pσq ` 1 pmod 2q.
Now, suppose σ “ τ1 ¨ ¨ ¨ τl “ τ11 ¨ ¨ ¨ τl11 as products of transpositions. As disjoint cycle notation
works (Theorem 16), #pσq is completely determined by σ (i.e. it has nothing to do with these
products of transpositions). But σ “ eτ1 ¨ ¨ ¨ τl , so by applying the previous result several times,
we get
#pσq ” #peq ` l ” n ` l pmod 2q
1 1
and #pσq ” #peq ` l ” n ` l pmod 2q
1
so l ” l pmod 2q. 

Definition: Writing σ P Sn as a product of transpositions σ “ τ1 ¨ ¨ ¨ τl , we call sgnpσq “ p´1ql

the sign of σ. If sgnpσq “ 1 we call σ an even permutation. If sgnpσq “ ´1 we call σ an odd
permutation.

Note: We have proved that this is well-defined, i.e. we get the same answer no matter how we
write σ as a product of transpositions.

20 Theorem: For n ě 2, sgn : Sn ÝÑ t˘1u is a surjective group homomophism.

Proof. It is a group homomorphism: sgnpσ1 σ2 q “ p´1ql1 `l2 “ p´1ql1 p´1ql2 “ sgnpσ1 qsgnpσ2 q,
where σ1 “ τ1 ¨ ¨ ¨ τl1 and σ2 “ τ11 ¨ ¨ ¨ τl12 as products of transpositions.
It is surjective: sgnpeq “ 1, sgnpp12qq “ ´1. 
Note: The hard bit is showing that it is well-defined!!! This means that if someone asks you
to define the signature homomorphism, you have to include the proof that the sign is well defined!

21 Lemma: σ P Sn is an even permutation iff the number of cycles of even length is even.

Proof. As sgn is a group homomorphism, writing σ in disjoint cycle notation σ “ σ1 ¨ ¨ ¨ σl , we

get sgnpσq “ sgnpσ1 q ¨ ¨ ¨ sgnpσl q. We have seen that a k-cycle is the product of k ´ 1 transpositions
(Proposition 18), so
sgnpσq “ p´1qk1 ´1 ¨ ¨ ¨ p´1qkl ´1
“1 iff an even number of the ki ´ 1 are odd
iff an even number of the ki are even
p “ ´1 iff an odd number of the ki are even)

Slogan: “odd length cycles are even, even length cycles are odd.”

Definition: The kernel of sgn : Sn ÝÑ t˘1u is called the alternating group: kerpsgnq “ An .

Remark: As a kernel, An ď Sn . [Indeed, as a kernel An is a normal subgroup of Sn , c.f. Chapter 5.]

You will meet/have met sgn ř in the definition of the determinant of a matrix: if A is an n ˆ n
matrix, A “ paij q, then det A “ σPSn sgnpσqa1 σp1q ¨ ¨ ¨ an σpnq .

22 Proposition: Any subgroup of Sn contains either no odd permutations, or exactly half.

Proof. See Q1 Sheet 2. Revisit this after we’ve done the Isomorphism Theorem (Thm 47):
that gives another way to prove this. 

IA Groups Page 18 c Julia Goedecke, Michaelmas 2016

 CHAPTER 3

Lagrange Theorem

Cosets
Definition: Let H ď G and a P G. The set aH – tah | h P Hu is a left coset of H. Similarly
Ha – tha | h P Hu is a right coset of H.

Examples: a) Take 2Z ď Z. Then 6 ` 2Z “ t all even numbers u “ 0 ` 2Z and 1 ` 2Z “

t all odd numbers u “ 17 ` 2Z.
b) Take G “ S3 , H “ xp12qy “ te, p12qu. The left cosets are:
eH “ p12qH “ te, p12qu; p13qH “ tp13q, p123qu “ p123qH; p23qH “ tp23qp132qu “ p132qH
c) Take G “ D6 . Recall D6 “ xr, s | r3 “ e “ s2 , rs “ sr´1 y. Take H “ xsy “ te, su. Then
the left coset rH “ tr, rs “ sr´1 u is not the same as the right coset Hr “ tr, sru. Try
the same in S3 , and find other cosets for D6 . [Notice something?]

In the next section we are going to prove some important properties of cosets. For this we
need a definition first.

Definition: Let X be a set, and X1 , X2 , . . . , Xn subsets of X. Then the Xi are called a partition
of X if Ť
n
(i) i`1 Xi “ X (“union gives all of X”) or equivalently
@x P XDi s.t. x P Xi (“every element of X is in at least one Xi ”)
(ii) Xi X Xj “ ∅ for all i ‰ j (“pairwise disjoint”) (“each element is in at most one Xi ”)

We will prove that the left cosets of H partition G, and that they all have the same size.
Note: We can have aH “ bH with a ‰ b, see example 2Z ď Z.

The Theorem
23 Theorem: (Lagrange)
If G is a finite group and H ď G, then |H| divides |G|.

Proof. First step: we prove that the left cosets partition G.

(i) For each a P G, we have a P aH, so every element is in at least one coset.
(ii) We need to prove: for a, b P G, the cosets aH and bH are either the same or disjoint.
Suppose aH, bH are not disjoint. So there exists some h1 P H such that ah1 P bH
(i.e. ah1 P aH X bH). So ah1 “ bh2 for some h2 P H. Then a “ bh2 h´1 1 , so for any h P H,
we have ah “ bh2 h´1
1 h P bH. So aH Ď bH. Similarly b “ ah 1 h´1
2 and bH Ď aH, so the
two cosets are the same.
Second step: We prove that all left cosets have the same size (as H). Let a P G. Consider
H ÐÑ aH
h ÞÝÑ ah
´1
a b ÐÝß b

19
CHAPTER 3. LAGRANGE THEOREM

Check: These are well-defined functions, and mutually inverse. So they give a bijection between
the sets H and aH.
So if |H| is finite, then |H| “ |aH| for all a P G. So putting both parts together: G is the
disjoint union of distinct cosets, all of which have size |H|, so |G| “ number of cosets ¨ |H|. 
We only need |G| finite in the very last step of the proof!
We could do the same proof with right cosets.

Definition: We write |G : H| for the number of cosets of H in G, and call it the index of H in G.

So
|G| “ |G : H||H|

24 Fact: (“same coset check”)

aH “ bH ô b´1 a P H. (see in proof: a “ bh2 h´1
1 , so b
´1
a “ h2 h´1
1 P H)

Exercise: Do the other direction.

Lagrange Corollaries
25 Corollary: (“element order divides group order”)
Let G be a finite group and a P G. Then ordpaq  |G|.

Proof. Consider the subgroup H “ xay generated by a. We know that ordpaq “ |xay|, and
by Lagrange |H|  |G|. 

26 Corollary: (“exponent divides group order”)

Let G be a finite group. Then for each a P G, a|G| “ e.

Proof. We know |G| “ ordpaq ¨ k for some k P Z. So a|G| “ paordpaq qk “ ek “ e. 

m
Recall: the exponent of G is the smallest number m s.t. a “ e for all a P G.
We have proved the statement of the corollary, but not quite the name: it should not be too
hard for you to prove the statement suggested by the name as well.

27 Corollary: (“prime order groups”)

Groups of prime order are cyclic. Moreover, such groups are generated by any of their non-identity
elements.

Proof. Say |G| “ p. Let a P G. If a “ e, then xay “ teu. Otherwise H “ xay ‰ teu, and |H|
divides |G| by Lagrange. As |G| is prime, |H| “ |G|, so ordpaq “ p and G “ xay is cylic. As the
element a was arbitrary, any non-identity element generates G. 

Small detour on equivalence relations

Definition: Let X be a set. An equivalence relation „ on X is a relation which is
(i) reflexive: x „ x for all x P X
(ii) symmetric: x „ y ñ y „ x for all x, y P X
(iii) transitive: If x „ y and y „ z, then x „ z, for all x, y, z P X.

The symbol „ is officially called “tilde”. It is also sometimes called “twiddle” (slightly less
formally). I like the fact that you can use “twiddle” also as a verb, as in “x twiddles y”.

Examples: a) X “ Z, the relation ”n defined as a ”n b pmod nq ô n  a ´ b.

IA Groups Page 20 c Julia Goedecke, Michaelmas 2016
CHAPTER 3. LAGRANGE THEOREM

b) X any set of groups, – “isomorphic to” is an equivalence relation.

This is a hidden exercise.

Definition: Given an equivalence relation „ on X, the equivalence classes are

rxs„ p“ rxsq “ ty P X | x „ yu

28 Proposition: Equivalence classes form a partition.

Proof. Let „ be an equivalence relation on the set X.

(i) Each x P X is in the class rxs.
(ii) If rxs X rys ‰ ∅, then there is some z P rxs X rys, i.e. x „ z and y „ z. Then (using
symmetry and transitivity) we can show that x „ y, and any w with x „ w also satisfies
y „ w. So rxs “ rys (because the argument is symmetric in x and y).

Are cosets equivalence classes for some equivalence relation? Easy answer: yes, just set a „ b
iff a, b in same coset. But we will find a more interesting answer.

29 Lemma: (“coset equivalence relation”)

Let H ď G. Defining a „ b if b´1 a P H gives an equivalence relation on G, whose equivalence
classes are the left cosets of H.

Proof. We check the three conditions of an equivalence relation.

Reflexivity: a´1 a P H so a „ a.
Symmetry: If b´1 a P H then also pb´1 aq´1 “ a´1 b P H. So a „ b ñ b „ a.
Transitivity: If a „ b and b „ c, we have b´1 a, c´1 b P H. So c´1 bb´1 a “ c´1 a P H, so a „ c.
By “same coset check” (Fact 24), it follows that a „ b iff aH “ bH iff a P bH and b P aH, so
equivalence classes are exactly the cosets. 
Notice: the fact that the identity is inside the subgroup gives reflexivity, closure under inverses
gives symmetry, and closure under group multiplication gives transitivity.
G is not necessarily finite!

Applications of Lagrange
In this section we will learn how to calculate with equivalence classes. Then we can use
Lagrange to prove a nice Number Theory result (Fermat-Euler Theorem) with Group Theory
methods instead.

Consider pZ, `q and, for fixed n, take the subgroup H “ nZ. The cosets are: 0 ` H “ r0s,
1 ` H “ r1s, . . . , n ´ 1 ` H “ rn ´ 1s. We can call the numbers 0, . . . , n ´ 1 representatives of
the cosets.
Calculating “mod n”: Define ras ` rbs “ ra ` bs and ras ¨ rbs “ rabs.
We need to check that these are well-defined! If ra1 s “ ra2 s and rb1 s “ rb2 s, then a1 “ a2 ` kn,
b1 “ b2 ` ln for some k, l P Z. So a1 ` b1 “ a2 ` b2 ` npk ` lq and a1 ¨ b1 “ pa2 ` knqpb2 ` lnq “
a2 b2 ` npkb1 ` la2 ` klnq. So ra1 ` b1 s “ ra2 ` b2 s and ra1 b1 s “ ra2 b2 s.
We have seen that pZn , `n q is a group. What happens with multiplication? We can only take
elements which have inverses. (called units Ñ c.f. GRM in second year)
Let Un “ tras | a coprime to nu. (We will see that these are the units.) Define the Euler
totient function ϕpnq “ |Un |. E.g. ϕppq “ p ´ 1 for p prime, ϕp4q “ 2.

30 Proposition: (“mult mod n as group”)

Un is a group under multiplication mod n.
IA Groups Page 21 c Julia Goedecke, Michaelmas 2016
CHAPTER 3. LAGRANGE THEOREM

Proof. The operation is well-defined, see above.

Closure: If a and b are coprime to n, then ab is also coprime to n. So ras, rbs P Un ñ rasrbs “
rabs P Un .
Identity: r1s (clearly)
Inverses: Let ras P Un , and consider the map “multiplication by a”
Un ÝÑ Un
rcs ÞÝÑ racs
This is injective: if rac1 s “ rac2 s, then n divides apc1 ´ c2 q, so as a and n are coprime, n divides
c1 ´ c2 , so rc1 s “ rc2 s.
As Un is finite, any injection Un ÝÑ Un is also a surjection (recall exercise from “small detour
about functions” Chapter 1, or see N+S). So there is a c such that racs “ rasrcs “ r1s. So ras has
an inverse in Un .
Associativity and commutativity follow from the corresponding properties of Z (because the
operation rasrbs “ rabs is well-defined). 

31 Theorem: (Fermat-Euler)
Let n P N, and a P Z be coprime to n. Then aϕpnq ” 1 pmod nq.
In particular (Fermat’s Little Theorem): If n “ p is prime, then ap´1 ” 1 pmod pq for any
a not a multiple of p.

Proof. As a is coprime to n, we have ras P Un . As “exponent divides group order” (Corol-

lary 26), we have ras|Un | “ r1s, which means aϕpnq ” 1 pmod nq. We have seen that for a prime p,
ϕppq “ p ´ 1. 
Comment: If you want to use this as a proof of Fermat-Euler in a “non-groups” context, you
must (at least) add the proof that Un is a group.

We can use Lagrange to help us find subgroups:

Examples: ˛ D10
possible subgroup sizes must be
1 teu
2 elements must have order 2 Ñ 5 such
5 must be cyclic, e plus 4 elements of order 5. So only one.
10 D10
˛ D8
possible subgroup sizes searching gives
1 teu
2 5 such (4 reflections, one rotation 1800 )
4 3 such (one of which is cyclic)
8 D8

Exercise: Use Lagrange to help you find all subgroups of the cyclic group Cn .

Remark: The converse of Lagrange is false! That is, if k  |G|, there is not necessarily a subgroup
of order k. E.g. |A4 | “ 12 , and A4 has no subgroup of order 6 (see Sheet 2 Q3). (C.f. Cauchy
Theorem (Thm 71): this is different for primes.)

We can also use Lagrange to determine what small groups must look like.
IA Groups Page 22 c Julia Goedecke, Michaelmas 2016
CHAPTER 3. LAGRANGE THEOREM

32 Proposition: (“groups of order 4”)

A group of order 4 is isomorphic to either C4 or C2 ˆ C2 .

Proof. Let |G| “ 4. By Lagrange, possible element orders are: 1 (only e), 2, 4. If there is an
element a P G of order 4, then G “ xay – C4 .
If not, then all non-identity elements have order 2. So by Sheet 1 Q8, G is abelian. Take two
elements of order 2, say b, c P G. Then xby “ te, bu, xcy “ te, cu, so xby X xcy “ teu. As G is abelian,
xby and xcy commute. The element bc “ cb also has order 2 and is the only element of G left, so
G “ xby ¨ xcy, so by the “Direct Product Theorem” (Proposition 13), G – xby ˆ xcy – C2 ˆ C2 . 
To determine groups of size 6 we also use Lagrange, but we need normal subgroups, so we will
only do it in the next chapter.

Left or right cosets

We could have proved Lagrange with right cosets.
As |aH| “ |H| and similarly |H| “ |Ha|, left and right cosets have the same size. Are they
the same?

Example: a) G “ pZ, `q, H “ 2Z. We have

0 ` 2Z “ t all even numbers u “ 2Z ` 0
1 ` 2Z “ t all odd numbers u “ 2Z ` 1
so left cosets equal right cosets.
Of course, if G is abelian, we have aH “ Ha for all a P G, and any subgroup H ď G.
b) G “ D6 p– S3 q generated by r, s satisfying r3 “ e “ s2 , rs “ sr´1 . |G| “ 6. Let
K “ xry – xp123qy, an order 3 subgroup.
The cosets partition G, so they must be
K “ te, r, r2 u – te, p123q, p132qu
and sK “ ts, sr “ r2 s, sr2 “ rsu “ Ks
por p12qK “ tp12q, p23q, p13qu “ Kp12qq
So for all a P G, aK “ Ka.
c) Take G “ D6 – S3 again, but H “ xsy – xp12qy, order 2.
left cosets right cosets
H “ te, su – te, p12qu H “ te, su – te, p12qu
rH “ tr, rs “ sr´1 u Hr “ tr, sr “ r2 su
– tp123q, p13qu “ p123qH – tp123q, p23qu “ Hp123q
r2 H “ tr2 , r2 s “ sru Hr2 “ tr2 , sr2 “ rsu
– tp132q, p23qu “ p132qH – tp132q, p13qu “ Hp132q
They don’t coincide.
K and H are “different kinds of subgroups”, see Chapter 5.

IA Groups Page 23 c Julia Goedecke, Michaelmas 2016

 CHAPTER 4

Infinite Transformation Groups

Frieze groups
NOTE: We will not cover all of the frieze group material in lectures. What is not covered in
lectures will not be examinable.
We saw earlier on in the course that we can think of Z as the symmetry group of a certain
infinite strip; and we saw some funny group D8 which we don’t quite know how to think about
yet. Let’s put these two ideas together and see what comes out of it.
As for the symmetries of regular polygons earlier in the course, we consider symmetries to be
(continuous) maps from the plane to the plane which preserve the geometric shape we are interested
in. In particular, this means we can’t “tear” the geometric object and stick it back together in a
different way.
So inspired by the strip for Z, we want to study:

Definition: A frieze is an infinite horizontal strip with discrete translational symmetry. (This
means, the translational symmetry group is Z, not Q or R.) A frieze group is the symmetry
group of some frieze.

To explain one important aspect of frieze groups, we look at a particular frieze:

or
¨ ¨ ¨ LΓLΓLΓLΓLΓLΓLΓLΓLΓLΓ¨ ¨ ¨

This pattern has, apart from its translational symmetry, only glide reflections.

Definition: A glide reflection is the composite of a translation in one direction and a reflection
in the direction vertical to the translational direction.

Note that if the glide reflection g translates by distance 1{2d (as well as reflecting horizontally),
then the “pure” translation of distance 1{2d may or may not also be a symmetry of the frieze. (In
the above example, it is not.) However, the translation g 2 of distance d definitely is.
So possible symmetries of such a frieze are:

˛ translations;
˛ vertical reflections (necessarily in infinitely many axes which are fixed distances apart);
˛ horizontal reflection;
˛ 180˝ rotations (necessarily in infinitely many centre points which are fixed distances
apart);
˛ glide reflections.

Exercise: ˛ Check how these different options combine. E.g. two glide reflections combine
to a translation (with double translational distance). Or a glide reflection followed by a
vertical reflection gives a 180˝ rotation. Etc.
24
CHAPTER 4. INFINITE TRANSFORMATION GROUPS

˛ Check in what relationship these different options stand to each other. E.g. if x is a
translation, and y is a 180˝ rotation, convince yourself that yx “ x´1 y. The horizontal
reflection commutes with everything.
Hint for both: it helps to colour some bits of the appropriate frieze to see where they end up.

Examples: (Possible Frieze Groups) Note that names are generally my own, not standard.
They come from the letter representation of an example frieze with those symmetries.

˛ H-type, model frieze

This has all possible symmetries.

To determine what “abstract” group the symmetry group of this is isomorphic to,
note that a glide reflection followed by a vertical reflection gives a 180˝ rotation, and
two glide reflections give a translation, as mentioned above. So we can call the “smallest
distance” glide reflection g (say with distance 1{2d) and some fixed vertical reflection v,
and the horizontal reflection h. We find that vg “ g ´1 v, but that the horizontal reflection
commutes with everything. So the symmetry group here is

FH “ D8 ˆ Z2 – D8 ˆ C2 .

(The vg 2k are vertical reflections, as are the hvg 2k`1 . The vg 2k`1 are 180˝ rotations, as
are the hvg 2k . So the horizontal reflection turns vertical reflections into rotations and vice
versa. The g 2k are the translations by distance multiples of d, the hg 2k`1 are translations
by distance odd multiples of 1{2d, and the g 2k`1 are glide reflections with distances odd
multiples of 1{2d, while the hg 2k are glide reflections with distances multiples of d. I
will draw in the vertical symmetry lines and rotation centres for the different types in
lectures.)
˛ If we “subtract” the horizontal reflection from the above, we get

VΛ-type, model frieze

That means we have vertical reflections (only the vg 2k ), 180˝ rotations (only the
2k`1
vg ), glide reflection (the g 2k`1 , so only with distances odd multiples of 1{2d), and
translations g 2k (only with distances multiples of d). So as a subgroup of FH above, this
group is
FVΛ “ D8 ˆ teu – D8 .
˛ Now we take out the vertical reflections, and so we must also take out the glide reflection,
leaving only translation and 180˝ rotation.
Z-type, model frieze

As a subgroup of FH (or FVΛ ), this is generated by a translation g 2 and a 180˝

rotation vg, so we only get elements of the kind vg 2k`1 and g 2k , i.e. only translations
and rotations. However, if we rename g 2 “ t and vg “ s, then they still satisfy st “ t´1 s
[check it!]. So we still get the abstract group

FZ – D8 ,

but it is a different subgroup of FH . (To make it clearer, think of the example 2Z ă Z:

they are both isomorphic to Z, but not the same subgroup of Z.)
˛ Instead we could have taken out the rotations (and the glide reflection), leaving transla-
tions and vertical reflections.
M-type, model frieze
IA Groups Page 25 c Julia Goedecke, Michaelmas 2016
CHAPTER 4. INFINITE TRANSFORMATION GROUPS

This is the subgroup of FH (or FLΓ ) generated by g 2 and v, so we get elements of the
type g 2k (translations) and vg 2k (vertical reflections). The group is still
FM – D8 ,
but a third different subgroup of FH . (Think of two possible subgroups of D2n inside
D4n . When the n becomes 8, then of course the index doesn’t actually get smaller!)
˛ If we just have glide reflection and horizontal reflection, i.e. taking out the vertical
reflections (and rotations) from the H-type, we get
D-type, model frieze

As we remember that the horizontal reflection h commutes with everything, we get

the group
FD “ Z ˆ Z2 – Z ˆ C2 ď D8 ˆ C2 “ FH ,
thinking of the Z ď D8 being the subgroup generated by g. (If we instead take the
subgroup generated by g 2 and h, we get the same thing, just with larger distances, which
you could think of as 2Z ˆ C2 ď FH .)
˛ Here we have taken away everything except the glide reflections (two of which give the
all-present translations).
LΓ-type, model frieze

Clearly
FLΓ – Z.
As a subgroup of our first group, it is generated by g.
˛ Finally, we have one with only translations.
P-type, model frieze
Here also
FP – Z,
generated by g 2 , and perhaps thought of as 2Z ď FH . (You could also use the subgroup
generated by gh, in which case you get half the distances between the lines.)

Remarks: ˛ One might also call frieze groups “discrete symmetry groups fixing a line”.
You can think of e.g. finite cyclic and finite dihedral groups as discrete symmetry groups
fixing a point. And we have seen the relationship via “finite vs infinite” between these
above. A non-discrete symmetry group fixing a line might be R or Q. A non-discrete
symmetry group fixing a point might be S 1 , the unit circle, thought of as complex numbers
of modulus 1, with complex multiplication; or the group you meet on Example Sheet 1
with e2πiq for q P Q. Non-discrete groups tend to have interesting topological structures,
which means you can meaningfully talk about distances and/or analysis.
˛ Frieze groups are interesting as symmetry groups of ornamental designs, hence the name.
But in higher dimensions, it gets less “only ornamental”: in higher dimensions, these kind
of symmetry groups are called crystallographic groups and are very important to people
studying structures of materials (for example crystallographers, materials scientists, some
chemists, ...). They can also be useful to biologists and medics in studying structures of
DNA or proteins.

Matrix groups
Back to standard material, will be covered in lectures.
This chapter has “transformation groups” in the title, where transformation means basically
as much as “bijective function”, possibly applied to something “geometric”. So we are now going
to study a particular type of such functions, namely matrices, which we can apply to vectors in R2
IA Groups Page 26 c Julia Goedecke, Michaelmas 2016
CHAPTER 4. INFINITE TRANSFORMATION GROUPS

or R3 (very geometric spaces), or Rn (the obvious mathematical extension of the geometric ones
we can actually visualise).

Consider Mnˆn pF q, the set of nˆn matrices over F “ R, C (some field). Matrix multiplication
is associative (e.g. as they represent functions), but not in general commutative. If we want I to
be our identity, what matrices have inverses?

Definition:

GLn pF q “ tA P Mnˆn pF q | A is invertibleu

is the general linear group.

We could also have said “take all matrices with non-zero determinant”, because these are
exactly the ones which are invertible (as maps). You will cover this (or have covered this) in V+M.

33 Proposition: GLn pF q is a group.

Proof. Check GLn pRq or GLn pCq.

˜1 ¸
1 0
The identity is .. “ I, inverses are the inverse matrices, which exist by definition of
0 .
1
GLn pF q, and the composite of invertible matrices is invertible (c.f. Lemma 6: the composite of
bijective functions is bijective, or use det AB “ det A det B). Multiplication is associative. 

34 Proposition: (“det is a group hom”)

det : GLn pF q ÝÑ F zt0u is a surjective group homomorphism (F “ R or C).

Proof. We know (from V+M, or will soon know) that det AB “ det˜
A det B, so
¸ it is a group
x
1 0
homomorphism. If A is invertible, then det A ‰ 0. Given x P F zt0u, det .. “ x, so it is
0 .
1
surjective. 

Definition: The kernel of det is the special linear group

SLn pF q “ tA P GLn pF q | det A “ 1u.

So SLn pF q ď GLn pF q as it is a kernel (and so in fact it is this special kind of subgroup called
normal, see later).

We now look at a subgroup of GLn pRq with special geometric properties.

Interlude: reminder from V+M:
AT has entries ATij “ Aji “reflect in diagonal”.
Facts:
˛ pABqT “ B T AT
˛ pA´1 qT “ pAT q´1
˛ AT A “ I ô AAT “ I ô A´1 “ AT
˛ det AT “ det A.
We are now most definitely in R.

Definition: On “ On pRq “ tA P GLn pRq | AT A “ Iu is the orthogonal group.

IA Groups Page 27 c Julia Goedecke, Michaelmas 2016
CHAPTER 4. INFINITE TRANSFORMATION GROUPS

We check it is a group: if A, B P On , then pABqT AB “ B T AT AB “ B T IB “ I. Clearly

I I “ I, so I P On . If A P On , then AT P On , because pAT qT AT “ AAT “ I. So A´1 “ AT and
T

we have inverses.
How to think of it: a matrix in On has orthonormal columns: for each column v, v T v “ 1
(normal); for different columns, wT v “ 0 (orthogonal). Such a matrix is called an orthogonal
matrix, c.f. V+M.

35 Proposition: det : On ÝÑ t˘1u is a surjective group homomorphism.

T
Proof. For A P On , we have I, so det AT A “ pdet Aq2 “ 1, so det A “ ˘1. We know
¨ A A “˛
´1
1 0
det AB “ det A det B, and det ˝ .. ‚ “ ´1, so it is a surjective group homomorphism. 
0 .
1

Definition: The kernel of det : On ÝÑ t˘1u is the special orthogonal group

SOn “ SOn pRq “ tA P On | det A “ 1u.

Their special geometric property is the following:

36 Lemma: (“orthogonal matrices are isometries”)

For A P On and x, y P Rn , we have
(1) pAxqT pAyq “ xT y (A preserves dot product)
(2) |Ax| “ |x| (A preserves length)

Proof. pAxqT pAyq “ xT AT Ay “ xT Iy “ xT y as AT A “ I. Then |Ax|2 “ pAxqT pAxq “

x x “ |x|2 , so as both sides are positive, |Ax| “ |x|.
T


Note: orthogonal matrices are also linear maps, i.e. 0 Ý

Þ Ñ 0; not all isometries are linear (for
example translations). Isometries also form a group.
If we now look at complex matrices, the equivalent to “transpose” is “complex conjugate
transpose”. So we use dagger: pA: qij “ Aji
We still have:
˛ pABq: “ B : A:
˛ pA´1 q: “ pA: q´1
˛ A: A “ I ô AA: “ I ô A´1 “ A:
˛ det A: “ det A

Definition: Un “ tA P GLn pCq | A: A “ Iu is the unitary group (the group of all unitary
matrices, also written U pnq).

Exercise: Check it is a group.

In the next result, we will want to refer to the set of complex numbers with modulus 1:
S 1 “ tz P C | |z| “ 1u is called the unit circle. It forms a group under complex multiplication.

37 Lemma: det : Un ÝÑ S 1 is a surjective group homomorphism.

Proof. 1 “ det A: A “ | det A|2 , so | det A| “ 1, so det ¨

A P S 1 . det
˛ AB “ det A det B as
λ
1 0
1
before, so it is a group homomorphism. Given λ P S , we have ˝ .. ‚ P Un with determinant
0 .
1
λ, so it is surjective. 
IA Groups Page 28 c Julia Goedecke, Michaelmas 2016
CHAPTER 4. INFINITE TRANSFORMATION GROUPS

Definition: The kernel of det : Un ÝÑ S 1 is SUn , the special unitary group.

Equivalently to the orthogonal matrices being isometries, unitary matrices preserve the com-
plex dot product:
x: y “ pAxq: pAyq
where x: y “ x1 y1 ` x2 y2 ` ¨ ¨ ¨ ` xn yn .

Möbius group
We now look at different kind of transformations, which take place on complex numbers.
Specifically, we want to study maps f : C ÝÑ C with f pzq “ az`b
cz`d , a, b, c, d P C, ad ´ bc ‰ 0.
Why ad ´ bc ‰ 0?
paz ` bqpcw ` dq ´ paw ` bqpcz ` dq
f pzq ´ f pwq “
pcw ` dqpcz ` dq
pad ´ bcqpz ´ wq
“ for all z, w P C
pcw ` dqpcz ` dq
So if ad ´ bc “ 0, then f is constant.
If c ‰ 0, what about f p´ dc q? Is it not defined?
We add a new point, 8, to C, to form the extended complex plane C Y t8u “ C8 .

Construction: (Stereographic Projection)

Take the unit sphere, cut it in the equator with the complex plane. Take a rod attached to the north
pole. Then the intersection points of the rod with the complex plane correspond to intersection
points of the rod with the sphere. The north pole corresponds to 8.
[C8 is called the one-point compactification of C (because it is compact, c.f. Met and Top
in Easter).]
We call this sphere the Riemann sphere.

az`b
Definition: A Möbius map is a map f : C8 ÝÑ C8 of the form f pzq “ cz`d , with a, b, c, d P C,
ad ´ bc ‰ 0, with f p´ dc q “ 8 and f p8q “ ac (if c ‰ 0).
[For c “ 0, f p8q “ 8.]

38 Lemma: Möbius maps are bijections C8 ÝÑ C8 .

Proof. Note that f pzq ´ f pwq shows that any Möbius map is injective on C. The inverse of
f pzq “ az`b dz´b
cz`d is gpzq “ ´cz`a . Check the composition both ways. For example, for z ‰ 8, z ‰ c
a

we have ´ ¯
dz´b
a ´cz`a `b adz ´ ab ´ bcz ` ab pad ´ bcqz
f pgpzqq “ ´ ¯ “ “ “z
dz´b
c ´cz`a ` d cdz ´ cb ´ cdz ` ad ad ´ bc

Special cases: f pgp8qq “ f p´ dc q “ 8 and f pgp ac qq “ f p8q “ a

c. (For c “ 0: f pgp8qq “ f p8q “
8.)
Similarly g ˝f “ 1C8 . 

39 Proposition: (Möbius group)

The Möbius maps form a group M under composition.
a1 z`b1 a2 z`b2
Proof. ˛ If f1 pzq “ c1 z`d1 , f2 pzq “ c2 z`d2 with ai di ´ bi ci ‰ 0, then
a2 ac11z`d
z`b1
1
` b2 pa1 a2 ` b2 c1 qz ` pa2 b1 ` b2 d1 q a3 z ` b3
f2 ˝f1 pzq “ “ “
c2 ac11z`d
z`b1
1
` d2 pc2 a1 ` d2 c1 qz ` pc2 b1 ` d1 d2 q c3 z ` d3
IA Groups Page 29 c Julia Goedecke, Michaelmas 2016
CHAPTER 4. INFINITE TRANSFORMATION GROUPS

with a3 d3 ´ b3 c3 “ pa1 a2 ` b2 c1 qpc2 b1 ` d1 d2 q ´ pa2 b1 ` b2 d1 qpc2 a1 ` d2 c1 q “ pa1 d1 ´

b1 c1 qpa2 d2 ´ b2 c2 q ‰ 0. This works for z ‰ 8, z ‰ ´ dc11 and z ‰ f1´1 p´ dc22 q “ ´ dc33 .
Again we need to check some special cases:
d
pa1 a2 `b2 c1 qp´ c 1 q`a2 b1 `b2 d1
– f1 p´ dc11 q “ 8, f2 p8q “ a2
c2 . We check f2 ˝f1 p´ dc11 q “ d
1
“
pc2 a1 `d2 c1 qp´ c 1 q`c2 b1 `d1 d2
1
a2 pc1 b1 ´a1 d1 q a2
c2 pb1 c1 ´a1 d1 q “ c2 .
– f1 p8q “ c1 , f2 p ac11 q “ ac21aa12`d
a1 `b2 c1
2 c1
“ ac33 “ f2 ˝f1 p8q.
– f1 pf1´1 p´ dc22 qq “ ´ dc22 and f2 p´ dc22 q “ 8, while f2 ˝f1 p´ dc33 q “ 8 as well.

A little explanation of what we are actually checking in the above: say for the first
bullet point, we are checking that ´ dc11 ÞÝÑ 8 ÞÝÑ ac22 when doing first f1 and then f2 in
separate steps is the same as when we do the new map with the a3 , b3 , c3 , d3 , i.e. in one
go ´ dc11 ÞÝÑ ac22 .
(When c1 “ 0 or c2 “ 0, it is also easy to check.)
So indeed f2 ˝f1 is a Möbius map.
˛ 1C8 : C8 ÝÑ C8 is 1pzq “ 1z`00z`1 with 1 ´ 0 ‰ 0, so 1 P M .
dz´b
˛ Inverses are f pzq “ ´cz`a with da ´ bc ‰ 0, so f ´1 P M .
´1

˛ Composition of functions is alway associative.



Note: M is not abelian. For f1 pzq “ 2z, f2 pzq “ z `1 have f1 ˝f2 pzq “ 2z `2, f2 ˝f1 pzq “ 2z `1.

Remark: (“point at 8 is not special”)

“Morally”, 8 is no different to any other point on the Riemann Sphere. You can get quite far with
1
the conventions “ 8 “ 0”, “ 01 “ 8” and “ a8 a
c8 “ c ”
Don’t use these in other contexts, only on the Riemann Sphere!!!!

Clearly az`b λaz`λb

cz`d “ λcz`λd for any λ ‰ 0 P C. So we don’t have a “unique representation with
a, b, c, d”. (c.f. fractions 12 “ 24 ). But

40 Proposition: (“Möbius maps via matrices”)

` ˘ az`b
The map θ : GL2 pCq ÝÑ M sending A “ ac db P GL2 pCq to fA with fA pzq “ cz`d is a surjective
group homomorphism.

Proof. θ lands in M , as A P GL2 pCq gives ad ´ bc ‰ 0. This also shows surjectivity. From
previous calculations:
pa1 a2 ` b2 c1 qz ` a2 b1 ` b2 d1
θpA2 qθpA1 qpzq “ “ θpA2 A1 qpzq.
pc2 a1 ` d2 c1 qz ` c2 b1 ` d1 d2


The kernel of this θ is

az ` b
Ker θ “ tA P GL2 pCq | z “ @zu
cz ` d
“ tλI | λ P C, λ ‰ 0u “ Z.
Here setting z “ 8 gives c “ 0, then z “ 0 ñ b “ 0, and then z “ 1 ñ a “ d. We call the matrices
λI scalar matrices. Z is the centre of GL2 pCq: those matrices which commute with everything.
C.f. Chapter 7.

Remark: Using fA “ fB iff B “ λA for some λ ‰ 0 P C, we get θ|SL2 pCq : SL2 pCq ÝÑ M also
surjective, Kerpθ|SL2 pCq q “ t˘Iu (scalar matrices of determinant 1).
IA Groups Page 30 c Julia Goedecke, Michaelmas 2016
CHAPTER 4. INFINITE TRANSFORMATION GROUPS

We can now look at what these Möbius transformations do geometrically.

41 Proposition: (“geometry of Möbius maps”)

Every Möbius map can be written as a composite of maps of the following form:
(i) f pzq “ az, a ‰ 0 dilation/rotation
(ii) f pzq “ z ` b translation
(iii) f pzq “ z1 combined inversion and reflection.

az`b
Proof. Let gpzq “ cz`d P M.
 ,2 a z  ,2 a z `
If c “ 0, (i.e. gp8q “ 8): gpzq “ ad z ` db , i.e. z d d
b
d .
piq piiq

If c ‰ 0, i.e. gp8q “ z0 P C (note z0 “ ac ): Let hpzq “ z´z1

0
, then hgp8q “ 8, so hg is of the
1
above form. And h pwq “ w ` z0 , which is a composite of type (iii) and (ii), so g “ h´1 hg is also
´1

a composite of maps of type (i), (ii) and (iii).

[Or direct calculation:
z ,2 z ` d  ,2 1  ,2 c2 
´ad`bc ´ad`bc
,2 a ` c2 “ az`b ] 
piiq c piiiq z` dc piq z` dc piiq c z` dc cz`d

Exercise: The “non-calculation way” in the above proof can be transferred into another (different)
composition (which of course has the same end result).

IA Groups Page 31 c Julia Goedecke, Michaelmas 2016

 CHAPTER 5

Quotient groups

Remember our discussion of left and right cosets, where we discovered that there were “two
different kinds” of subgroups.

Normal subgroups
Definition: A subgroup K of G is a normal subgroup if for all a P G and for all k P K we have
aka´1 P K. Write K E G.

Exercise: This is equivalent to

˛ @a P G, we have aK “ Ka (left coset=right coset)
˛ @a P G, we have aKa´1 “ K (see later: conjugacy classes in Chapter 6, in particular
Lemma 69)

From the earlier example at the end of Chapter 3, H ď D6 is not normal, but K E D6 is.
Every non-trivial group has at least two normal subgroups: which ones?

42 Lemma: (i) Any subgroup of index 2 is normal.

(ii) Any subgroup of an abelian group is normal.

Proof. (i) If K ď G of index 2, then the only two possible cosets are K and GzK
(because cosets partition G, proved in Lagrange, Thm 23). Since eK “ Ke, the other
left as well as the other right coset must both be GzK. So left and right cosets are the
same.
(ii) In an abelian group we have (the stronger condition) aka´1 “ k for all a P G, k P K.


43 Proposition: Every kernel is a normal subgroup.

Proof. Let f : G ÝÑ H be a group homomorphism with Ker f “ K. As seen in “Images and

Kernels”, for k P K, a P G, we have f paka´1 q “ f paqf pkqf paq´1 “ f paqf paq´1 “ e. So aka´1 P K,
so K E G. 
In fact, will see in the next sections that the kernels are exactly the normal subgroups.

Examples: From the result about kernels, we get the first few examples:
˛ An E Sn because it is a kernel.
˛ Similarly SLn E GLn , SOn E On and SUn E Un .
˛ The subgroup of scalar matrices, Z “ tλI | λ ‰ 0u, is a normal subgroup of GL2 pCq,
because it is the kernel of θ : GL2 pCq ÝÑ M , the map from matrices to the Möbius group.
A more direct example:
˛ In G “ D8 , the subgroup K “ xr2 y is normal.
Proof: Any element of G is either srl or rl (for some l). Check both types: e.g.
srl r2 psrl q´1 “ srl r2 r´l s “ ssr´2 “ r2 P K. The second case is an exercise.
32
CHAPTER 5. QUOTIENT GROUPS

Here is an example of how normality can be used.

44 Proposition: (“groups of order 6”)

Every group of order 6 is either cyclic or dihedral.

Proof. Let |G| “ 6. By Lagrange, possible element orders are 1, 2, 3 and 6. If there is a P G
of order 6, then G – C6 .
If there is no element of order 6, we can only have orders 2, 3 (other than e). By Sheet 1 Q7,
there must be an element r P G of order 3 (as 6 is not a power of 2). So xry E G as it has index 2.
There must also be an element s P G of order 2, by Sheet 1 Q8 (as 6 is even).
What is srs´1 ? As xry is normal, srs´1 P xry, so
˛ srs´1 “ e not possible (as r ‰ e), or
˛ srs´1 “ r ñ sr “ rs and sr has order 6 ñ G – C6 or
˛ srs´1 “ r2 “ r´1 ñ G is dihedral, i.e. G is generated r, s which satisfy
r3 “ e “ s2 , srs´1 “ r´1 , so G – D6 .


Quotients
We will now investigate when we can form a group out of cosets.
This is a part that many students find difficult to get their head round first time, because sets
(the cosets) suddenly become elements of some other set. So to get a bit of a feeling for it, I will
describe a little demonstration that can help. Take the numbers 0 to 11 with addition mod 12, and
randomly pair them up to make equivalence classes (of some not-further-determined equivalence
relation). Say we have paired up t1, 4u, t3, 8u, t7, 2u, ... If we now want to add them, we could try
saying t1, 4u ` t7, 2u “ t3, 8u because 1 ` 7 “ 8, but we could equally well have used 4 and 7, then
we’d need to get the pair of numbers containing 11. So this is not well-defined. We are trying to
define the addition of two sets by adding some representatives inside the sets, but we get different
answers for different choices.
Instead, now pair up the numbers in the cosets of the subgroup t0, 6u of Z12 . So now we
have t0, 6u, t1, 7u, t2, 8u, t3, 9u and so on up to t5, 11u. Now if you try adding two of these sets,
regardless of which representatives you choose to add you will get the same coset as your answer!
For example t1, 7u ` t3, 9u “ t4, 10u, because 1 ` 3 “ 4 or because 1 ` 9 “ 10 or because 7 ` 3 “ 10
or because 7 ` 9 “ 16 “ 4 pmod 12q. So any of the choices we make give the same answer. Try it
also with the cosets of the subgroup t0, 3, 6, 9u. Let’s do it properly mathematically now.

45 Proposition: Let K E G. The set pG : Kq of (left) cosets of K in G is a group under the

operation aK ˚ bK “ abK.

Proof. We will show that the operation is in fact well-defined, and satisfies the group axioms.
Well-defined: If aK “ a1 K and bK “ b1 K, then a1 “ ak1 and b1 “ bk2 for some k1 , k2 P K. So
1 1
a b “ ak1 bk2 “ abk3 k2 P abK for some k3 P K, as Kb “ bK.
Closure: If aK, bK are cosets, then so is abK.
Identity: eK “ K is the identity (clear from the definition of the group operation).
Inverses: a´1 K is the inverse to aK (clear from def. of the operation once it’s well-defined ).
Associativity: follows from associativity in G. 

Alternative way of using the normality in the proof of well-definedness: We start with a1 b1 “
ak1 bk2 . We want to get to a1 b1 “ abk for some k P K. So let’s try to make it like that: a1 b1 “
ak1 bk2 “ ab b´1 k1 b k2 . Here we added the b after the a because we wanted it there, but to keep
the equation the same, we have to also include b´1 . But now b´1 k1 b P K because K is normal, so
we get a1 b1 “ abk3 k2 as we did in the proof above.
IA Groups Page 33 c Julia Goedecke, Michaelmas 2016
CHAPTER 5. QUOTIENT GROUPS

Definition: This group is called the quotient group (or factor group) of G by K, and written
G{K.

Examples: a) Let G “ Z and K “ nZ (which is normal as G is abelian). The cosets are

nZ, 1 ` nZ, 2 ` nZ, . . . , pn ´ 1q ` nZ. G{K gives “addition mod n”. We write
Z{nZ “ Zn .
In fact these are the only quotient groups of Z, see “subgroups of Z” (Proposition 5).
b) Recall the example K E D6 from Section Left and right cosets, Chapter 3.
We have D6 generated by r, s with r3 “ e “ s2 , sr “ r´1 s and K “ xry E D6 (as it
has index 2). There are two cosets K and sK, so D6 {K has order 2, so D6 {K – C2 .
c) Now we use K “ xr2 y E D8 , generated by r, s with r4 “ e “ s2 , sr “ r´1 s (from
Section Normal subgroups). G{K should have |D8 | : |K| “ 4 elements. [Or you might be
more used to the notation |D8 |{|K| “ 4.]
G{K “ tK, rK “ r3 K, sK “ sr2 K, srK “ sr3 K “ rsKu
– te, x, y, xyu – C2 ˆ C2
by Propsition 32 “groups of order 4”, as all elements have order 2.

Remarks: ˛ If G is abelian, then G{K is also abelian.

˛ The set pG : Hq of left cosets also exists for non-normal subgroups H ď G, but the group
operation given above is not well-defined !!

Non-Example: Try D6 with H “ xsy. We saw the cosets in Section Left and right cosets. E.g.
rH ˚ r2 H “ r3 H “ H, but rH “ rsH and r2 H “ srH, and rssrH “ r2 H ‰ H.

46 Lemma: Given K E G, the map q : G ÝÑ G{K sending a to aK is a surjective group homo-

morphism (called the quotient map).

Proof. qpabq “ abK “ aK ˚ bK “ qpaq ˚ qpbq by definition. It is clearly surjective. 

Example: d) Let G “ Cn and H ď Cn (we’ve seen that H is also cyclic).

Say Cn “ xc | cn “ ey and H “ xck y – Cl where kl “ n.

q : Cn ÝÑ Cn {H
cm ÞÝÑ cm H
Ker q “ tcm | cm H “ Hu “ H.
What does the quotient group look like?
Cn {H “ tH, cH, c2 H, . . . , ck´1 Hu “ xcHy – Ck .

Slogan: “Subgroups and quotients of cyclic groups are cyclic.”

Remark: Quotient groups are not subgroups of G! They contain different kinds of elements (they
are not even subset). They may not even be isomorphic to any subgroups.
e.g. Z: quotient groups are Z{nZ (all finite)
subgroups are mZ (all infinite)
(In some sense “quotients are the opposite of (normal) subgroups” cf Part III Category Theory.)

The Isomorphism Theorem

We saw that all kernels are normal subgroups.
IA Groups Page 34 c Julia Goedecke, Michaelmas 2016
CHAPTER 5. QUOTIENT GROUPS

Given K EG, we have the quotient map q : G ÝÑ G{K with Ker q “ K. So “normal subgroups
are exactly the kernels of group homomorphisms”.
Now we refine this even more:

47 Theorem: (Isomorphism Theorem)

Let f : G ÝÑ H be a group homomorphism with Ker f “ K. Then K E G and G{K – Imf .

Proof. We have proved K E G already (Proposition 43).

G
f
,2 H
8B
q
 f
G{K
Define f : G{K ÝÑ H by f paKq “ f paq (so that we get f q “ f ). We check that this is well-defined:
if a1 K “ a2 K, then a´1 ´1
2 a1 P K by “same coset check” (Fact 24), so e “ f pa2 a1 q “ f pa2 q
´1
f pa1 q,
so f pa2 q “ f pa1 q, and f pa1 Kq “ f pa2 Kq as required.
f is a homomorphism: f paK ˚ bKq “ f pabKq “ f pabq “ f paqf pbq “ f paKqf pbKq.
f is injective: If f paKq “ f pbKq (i.e. f paq “ f pbq), then f pb´1 aq “ f pbq´1 f paq “ e, so
´1
b a P K, so aK “ bK by “same coset check”.
f is clearly surjective onto Imf .
So f gives an isomorphism G{K – Imf pď Hq. 

Remarks: ˛ If f : G ÝÑ H is injective, then Ker f “ teu, so G{K – G and G is isomorphic

to a subgroup of H (write G À H; e.g. inclusion).
˛ If f : G ÝÑ H is surjective, then Imf “ H, so then G{K – H.

Slogan: “Homomorphic images are quotients.”

Examples: a) Consider the determinant homomorphism det : GLn pRq ÝÑ R˚ . The kernel
is Kerpdetq “ SLn pRq “ tA P GLn pRq | det A “ 1u and the image is Imf “ R˚ , since for
x
1 0
x P R˚ , we have . “ x.
0 ..
1
So by the Isomorphism Theorem GLn pRq{SLn pRq – R˚ .
b) Recall the homomorphism θ : GL2 pCq ÝÑ M representing Möbius maps as matrices. We
worked out in Chapter 4, Section Möbius group, that the kernel is
Ker θ “ tλI | λ P C, λ ‰ 0u “ Z,
the scalar matrices. So by the Isomorphism Theorem: M – GL2 pCq{Z. We call this
GL2 pCq{Z “ PGL2 pCq, the projective general linear group.
Notice: using fA “ fB iff B “ λA for some λ ‰ 0 P C, we get θ|SL2 pCq : SL2 pCq ÝÑ M
also surjective, Kerpθ|SL2 pCq q “ t˘Iu (scalar matrices of determinant 1).
So also M – SL2 pCq{t˘Iu “ PSL2 pCq.
“The Möbius group is isomorphic to the projective special linear group.”
c) Define θ : pR, `q ÝÑ pC˚ , ¨q by θprq “ e2πir .
It is a group homomorphism: θpr ` sq “ e2πipr`sq “ e2πir e2πis “ θprqθpsq. Its kernel
is Z E R.
What is the image? We use the Isomorphism Theorem to get R{Z – Imθ “: pS1 , ¨q,
which we can take as (an alternative) definition of the unit circle.
IA Groups Page 35 c Julia Goedecke, Michaelmas 2016
CHAPTER 5. QUOTIENT GROUPS

d) Let G “ Z˚p “ Z{pZzt0u, for p prime (p ‰ 2).

The map f : G ÝÑ G defined by f paq “ a2 is a group homomorphism: pabq2 “ a2 b2
as pZ{pZq˚ is abelian. The kernel is Ker f “ t˘1u “ t1, p ´ 1u. So we see that Imf –
G{ Ker f must have order p´1
2 (these are the “quadratic residues”).

48 Lemma: (“essential uniqueness of cyclic groups”)

Any cyclic group is isomorphic to either pZ, `q or pZ{nZ, `n q for some n P N.

Proof. Let G “ xcy be cyclic. Define

f : Z ÝÑ G
m ÞÝÑ cm
f is a group homomorphism: cm1 `m2 “ cm1 cm2 .
f is clearly surjective. Ker f E Z, so either
˛ Ker f “ teu, so f is an isomorphism, so G – Z.
˛ Ker f “ Z, then G – Z{Z “ teu “ C1 .
˛ Ker f “ nZ for some n P N (these are the only subgroups of Z, see Prop. 5). Then
G – Z{nZ – Cn by the Isomorphism Theorem.


Definition: A group G is simple if it has no non-trivial proper normal subgroups. (i.e. only teu
and G).

Example: ˛ Cp with p prime is simple (but boring).

˛ A5 is simple (proof later, Theorem 65).

The finite simple groups are the building blocks of all finite groups. All finite simple groups
have been classified.
Note that for K E G with teu ‰ K ‰ G, we have |K| ă |G| and |G{K| ă |G|.

IA Groups Page 36 c Julia Goedecke, Michaelmas 2016

 CHAPTER 6

Group actions

Groups acting on sets

Recall the symmetries of an n-gons or SymX, or the Möbius maps on the Riemann sphere: the
group elements “do something” to the set of vertices or the set X or the points on the Riemann
sphere.

Definition: Let X be a set, G a group. An action of G on X is a function θ : G ˆ X ÝÑ X

satisfying
0. θpg, xq P X for all g P G, x P X;
1. θpe, xq “ x for all x P X;
2. θpg, θph, xqq “ θpgh, xq for all g, h P G, x P X.

Alternative notations:

θpg, xq “ g ¨ x “ gpxq P X θpg, ´q “ θg : X ÝÑ X

θpe, xq “ e ¨ x “ epxq “ x θe pxq “ x or θe “ 1X
θpg, θph, xqq “ θpgh, xq or gphxq “ pghqx or gphpxqq “ pghqpxq θg ˝θh “ θgh

Examples: a) Trivial action: take any group G and any set X, with θpg, xq “ x for all g, x.
“G does nothing”.
b) Sn acts on t1, . . . , nu by permutation.
c) D2n acts on (the vertices of) a regular n-gon and/or acts on t1, . . . , nu.
d) Rotations of a cube act on: faces of the cube, or diagonals, or axes (= pairs of opposite
faces), or...
e) Möbius maps act on C8 , the Riemann sphere.

Remark: Compare b), c): different groups can act on the same set. d): One group can act on
different sets.

49 Lemma: For each g P G, θpg, ´q “ θg : X ÝÑ X is a bijection.

Proof. As θpg, θpg ´1 , xqq “ θpgg ´1 , xq “ θpe, xq “ x for all x P X, we have θg ˝θg´1 “ 1X ,
the identity on X. Similarly θg´1 ˝θg “ 1X , so θg is a bijection (with inverse θg´1 ). 

50 Proposition: (“Alternative Action Definition”)

Let G be a group, X a set. Then θ : G ˆ X ÝÑ X with θpg, xq “ θg pxq is an action if and only if
ϕ : G ÝÑ SymX with ϕpgq “ θg is a group homomorphism.

Proof. “ñ” By Lemma 49, θg : X ÝÑ X is a bijection, so indeed θg P SymX. We have

ϕpghq “ θgh “ θg ˝θh “ ϕpgq˝ϕphq, so ϕ is a group homomorphism.
“ð” Given ϕ : G ÝÑ SymX a group homomorphism, and defining θ : G ˆ X ÝÑ X by θpg, xq “
ϕpgqpxq, the resulting θ is an action:
37
CHAPTER 6. GROUP ACTIONS

0. As ϕpgq “ θg P SymX, θpg, xq “ θg pxq P X.

1. ϕpeq “ 1X “ θe , so θe pxq “ x for all x.
2. ϕpghq “ ϕpgq˝ϕphq, so θg ˝θh “ θgh .


Definition: Given an action of G on X, the kernel of the action is Kerpϕ : G ÝÑ SymXq.

Note: These are all the elements that “act as the identity”. We have just shown that they
form a (normal) subgroup of G.

Ker ϕ “ K E G and G{K À SymX.

Examples: a) D2n acting on n vertices t1, . . . , nu gives ϕ : D2n ÝÑ Sn with Ker ϕ “ teu.
(This formalises our view of D2n as a subgroup of Sn .)
b) Let G be the rotations of a cube, and let it act on the axes (=pairs of opposite faces)
x, y, z.

y
z

This gives ϕ : G ÝÑ S3 . Then rotation around any of these axes by 1800 acts as the
identity on tx, y, zu. So here the kernel of the action has (at least) 4 elements: e, and
those three 1800 rotations. (In fact we will see later: these 4 are exactly the kernel, there
are no more; c.f. Section Using actions.)

Definition: An action is called faithful if Ker ϕ “ teu.

Orbits and Stabilisers

Which elements can we “reach” from some x P X? Which group elements fix x?

Definition: Given an action of G on X and x P X, the orbit of x is

orbpxq “ Gpxq “ ty P X | Dg P G s.t. gpxq “ yu Ď X

and the Stabiliser of x is

Stabpxq “ Gx “ tg P G | gpxq “ xu Ď G

The orbit of x is the set of elements we can reach from x, and the stabiliser of x is the set of
group elements which fix x.

51 Lemma: Stabpxq is a subgroup of G.

Proof. ˛ epxq “ x so Stabpxq is non-empty.

˛ If g, h P Stabpxq, then gh´1 pxq “ gpxq “ x so gh´1 P Stabpxq.
So Stabpxq is a subgroup by the super-efficient subgroup criterion (Lemma 4). 

Examples: a) Let D8 act on the corners of a square, so X “ t1, 2, 3, 4u. Then orbp1q “ X
and Stabp1q “ te, reflection in diagonal through corner 1u.
IA Groups Page 38 c Julia Goedecke, Michaelmas 2016
CHAPTER 6. GROUP ACTIONS

4 3

1 2

b) Let D8 act on the symmetry lines of a square, say called v, h, u, d as in the picture below.
Here orbphq “ th, vu, orbpuq “ tu, du.

u d
v

c) Let xp12qy act on t1, 2, 3u.

orbp1q “ t1, 2u and orbp3q “ t3u

Stabp1q “ teu and Stabp3q “ te, p12qu

d) Let the rotations of a cube act on the three axes (=pair of opposite faces) x, y, z (as
before). We can convince ourselves that orbpxq “ tx, y, zu. The stabiliser is harder. We
can find Stabpxq Ě te, any rotation in axis x p3 suchq, 1800 rotation in axis y or axis zu.
But it is hard to know if we have found all elements of the stabiliser. See later to
learn how to do that (using the Orbit-Stabiliser Theorem (Theorem 53), this is worked
out in Chapter 6, Section Applications).
e) (Look here again later to see another added example.)

Exercise: There are two more elements in the stabiliser from d) above. Can you find them using
the Taylor-Cube? You can find a net for the “Taylor-Cube” on the Moodle course page, my blog
or on Gareth Taylor’s website.

Definition: An action of G on X is called transitive if orbpxq “ X for all x P X.

“You can reach any element from any element.”

52 Lemma: The orbits of an action partition X.

Proof. Let G act on X.

(i) @x P X, x P orbpxq as epxq “ x.
(ii) If z P orbpxq X orbpyq, then Dg1 P G s.t. g1 pxq “ z and Dg2 P G s.t. g2 pyq “ z. So
x “ g1´1 pg2 pyqq. Then for any w P orbpxq with gpxq “ w, we have gpg1´1 pg2 pyqqq “ w so
w P orbpyq. So orbpxq Ď orbpyq, and similarly orbpyq Ď orbpxq, so orbits are disjoint or
equal.
Therefore the orbits partition X. 

Remember the proof that disjoint cycle notation works (Theorem 16): we were really finding
the orbits, which are disjoint and that is why it all worked.

53 Theorem: (Orbit-Stabiliser)
Let the finite group G act on the set X. Then for any x P X, we have

|G| “ |orbpxq||Stabpxq|.

Proof. Idea: We show that each point in the orbit of x corresponds to a particular coset of
Stabpxq. That lets us use Lagrange.
IA Groups Page 39 c Julia Goedecke, Michaelmas 2016
CHAPTER 6. GROUP ACTIONS

Stabpxq is a subgroup of G (by Lemma 51), so we can look at the set of cosets. Consider the
following mappings:
tcosets of Stabpxqu ÐÑ orbpxq
gStabpxq ÞÝÑ gpxq
hStabpxq ÐÝß y “ hpxq

These are well-defined: if gStabpxq “ g 1 Stabpxq, then by the “same coset check”, Fact 24,
we have g ´1 g 1 P Stabpxq, so g ´1 g 1 pxq “ x. So indeed g 1 pxq “ gpxq is just one well-defined point
in the orbit. For the other direction, if a point in the orbit satisfies y “ hpxq “ h1 pxq, then
h´1 h1 P Stabpxq, and so the two cosets coincide: hStabpxq “ h1 Stabpxq. It is now clear from
the construction that these two mappings are inverse to each other, giving us a bijection between
the set of coset of Stabpxq and the orbit of x. (Notice that we have not yet used any finiteness
conditions.)
So for a finite group G, putting this bijection together with Lagrange (Theorem 23) immediately
gives us
|G| “ |orbpxq||Stabpxq|.


Remark: To get a deeper understanding of what is going on here, notice that the coset gStabpxq
is exactly the set of all h which take x to y:
th P G | y “ hpxqu “ gStabpxq

Examples: a) In Chapter 1, Section Symmetries of a square we worked out “how big is

D2n ”. Now we can do it using Orbit-Stabiliser: D2n acts on t1, . . . , nu transitively, so
|orbp1q| “ n. Stabp1q “ te, reflection in line through 1u, so |D2n | “ n ¨ 2. The reflection
in the line through 1 represents the part of the argument where we said “once 1 is fixed,
2 can go in two different places”.
Note: if the action is transitive, then all stabilisers have same size.
b) Let xp12qy act on t1, 2, 3u as before. We saw
orbp1q “ t1, 2u and orbp3q “ t3u
Stabp1q “ teu and Stabp3q “ te, p12qu
so the sizes are different, but in each case multiply to 2.
Exercise: see what you get with e.g. xp123qp45qp6789qy acting on t1, . . . , 9u.
c) Let S4 act on X “ t1, 2, 3, 4u. We have orbp1q “ X and |S4 | “ 24, so we know that
|Stabp1q| “ |S4 |{|orbp1q| “ 24{4 “ 6. That makes it easier to find: clearly St2,3,4u – S3
fixes 1, so St2,3,4u Ď Stabp1q, but by sizes this gives St2,3,4u “ Stabp1q.

Left multiplication actions

We will now learn about some actions which come up again and again, and some of their
consequences.

54 Lemma: (left regular action)

Every group G acts on itself by left multiplication. This action is faithful and transitive.

Proof. We check the three conditions for an action:

0. @g P G, x P G, gx P G.
1. e ¨ x “ x for all x P G by identity axiom.
2. gphxq “ pghqx by associativity.
IA Groups Page 40 c Julia Goedecke, Michaelmas 2016
CHAPTER 6. GROUP ACTIONS

So this is indeed an action.

The action is faithful: if gx “ x for all x P X, then g “ e by uniqueness of identity (Prop. 1).
The action is transitive: given x, y P G, then yx´1 “ g P G and gx “ yx´1 x “ y. 
This action gives rise to the important theorem:

55 Theorem: (Cayley)
Every group is isomorphic to a subgroup of some symmetric group.

Proof. Take the left regular action of G on itself. This gives a group homomorphism
ϕ : G ÝÑ SymG, with kernel Ker ϕ “ teu as the action is faithful. So by the Isomorphism Theorem
(Theorem 47), G – Imϕ ď SymG. 

56 Lemma: (left coset action)

Let H ď G. Then G acts on the left cosets of H by left multiplication, transitively.

Proof. We check the three conditions for an action:

0. gpaHq “ gaH is a coset of H.
1. epaHq “ aH for all cosets aH.
2. g1 pg2 paHqq “ g1 ppg2 a2 qHq “ pg1 g2 qaH by associativity in G.
So this is indeed an action. Given aH, bH, we have ba´1 P G and ba´1 paHq “ bH, so the action
is transitive. 
Note: if H “ teu, this is just the left regular action.

57 Proposition: GLn pCq acts faithfully on Cn by left multiplication, with two orbits.

Proof. “applying the function to a vector”.

˛ For A P GLn pCq, v P Cn , we have Av P Cn .
˛ Iv “ v for all v P Cn .
˛ ApBvq “ pABqv for all v P Cn .
Intuitively, this action is clearly faithful as “only I acts as the identity”.
Proper proof: a linear map is determined by what it does on a basis. Take the standard basis
¨ ˛ ¨ ˛
1 0
˚.‹
˚0‹
˚ ‹
˚ .. ‹
e1 “ ˚
˚ .. ‹ , . . . , en “ ˚ ‹ .
‹ ˚ ‹
˝.‚ ˝0‚
0 1
Any matrix which maps each ek to itself must be I.
Orbits: clearly A0 “ 0 for all A P GLn pCq. Also, as A is invertible, Av “ 0 ô v “ 0. So 0
forms a singleton orbit. Given any two vectors v ‰ w P Cn zt0u, there is a matrix A P GLn pCq s.t.
Av “ w. (Prove it with methods from V+M.) 
Similarly GLn pRq acts on Rn .

Exercise: Letting SLn pCq act on Cn , what orbits do you get?

IA Groups Page 41 c Julia Goedecke, Michaelmas 2016

 CHAPTER 7

Conjugacy

The concept of conjugacy, or conjugate elements, comes up again and again in mathematics.
We study it first in terms of actions, but we will see it in many different contexts as well.

Conjugation action
Definition: Given a, b P G, the element bab´1 P G is the conjugate of a by b.

58 Lemma: (conjugation action)

Any group G acts on itself by conjugation.

Proof. We check the three conditions for an action:

0. gpxq “ gxg ´1 P G for g, x P G.
1. epxq “ exe´1 “ x for all x P G.
2. gphpxqq “ gphxh´1 q “ ghxh´1 g ´1 “ pghqxpghq´1
So this is indeed an action. 
This action is so important that the kernel, orbits and stabilisers of it have special names.

Definition: The kernel of this action is the centre of G:

ZpGq “ tg P G | gag ´1 “ a @a P Gu “ tg P G | ga “ ag @a P Gu
“elements that commute with everything”.
(This is also sometimes written CpGq but that can be confusing with centraliser notation
coming next.)
The orbits of this action are called conjugacy classes
cclpaq “ tb P G | Dg P G s.t. gag ´1 “ bu.
The stabilisers of this action are called centralisers
CG paq “ tg P G | gag ´1 “ au “ tg P G | ga “ agu
“elements that commute with a.”
Ş
Remark: ZpGq “ gPG CG pgq (Exercise)

Conjugacy and normal subgroups are very closely related. First we see:

59 Lemma: (“conjugation action restricts to normal subgroups”)

Let K E G. Then G acts by conjugation on K.

Proof. For all g P G, k P K, we have gkg ´1 P K, so the map G ˆ K ÝÑ K sending pg, kq

to gkg ´1 does land in K. We still have eke´1 “ k for all k P K, and still have pghqkpghq´1 “
gphkh´1 qg ´1 . 
And now we look the other way, using conjugacy to determine which subgroups are normal.
42
CHAPTER 7. CONJUGACY

Ť
60 Proposition: (“Normal = ccls”)
Normal subgroups are exactly those subgroups which are unions of conjugacy classes.

Proof. Let K E G. Then if k P K, so is gkg ´1 for all g P G, so cclpkq Ď K. Conversely, if

K is a union of ccls and a subgroup of G, then for all k P K, g P G, we have gkg ´1 P K, so K is
normal. 

There are many particular cases in which conjugation action has a special meaning. For
example:

61 Proposition: GLn pCq acts on Mnˆn pCq by conjugation.

Proof. Let A P Mnˆn pCq, P P GLn pCq. Then P AP ´1 P Mnˆn pCq, IAI ´1 “ A, and
pP QqApP Qq´1 “ P pQAQ´1 qP ´1 . 

How to think about this: matrices in Mnˆn pCq represent maps Cn ÝÑ Cn . Two matrices are
conjugate if they represent the same map with respect to different bases. The P is then a base
change matrix.
Special case GL2 pCq acting on M2ˆ2 pCq:
We know from V+M that we have three different types of orbits: A is conjugate to a matrix
of one of ˜
these forms:
¸
λ 0
˛ with λ ‰ µ
0 µ
˜ ¸
λ 0
˛
0 λ
˜ ¸
λ 1
˛
0 λ
C.f. V+M and eigenvectors. Also more generally Linear Algebra second year “Jordan Normal
Form”.
So, as we see above, conjugate matrices really represent the same map. There are many more
such similarities of conjugate elements, which we will see in the following sections.
Slogan: “Conjugate elements have many of the same properties.”

Exercise: Conjugate elements have the same order, same size centralisers, ... cf ccls in Sn .

You will also see another property conserved under conjugation on the example sheet, look
out for it!
Conjugacy is used a lot in all sorts of different areas of maths, like geometry, algebra, also
physics, to make situations easier to handle. For example, when considering a rotation in C about
any point, we can first translate that point to 0, then do the rotation around 0, which is very easy,
and then translate 0 back to the original point. In similar ways conjugacy is used to consider an
“easier” but entirely equivalent situation. We will see how useful it is when working with Möbius
maps in the second-to-last section of this chapter.

Conjugacy classes in Sn
In the context of Sn , we see many more properties that conjugate elements have in common.
The elements σ, τ P Sn are conjugate if Dρ P Sn s.t. ρσρ´1 “ τ . Ccls in Sn are very special.

62 Proposition: (“ccls in Sn are determined by cycle type”)

Two elements of Sn are conjugate if and only if they have the same cycle type.
IA Groups Page 43 c Julia Goedecke, Michaelmas 2016
CHAPTER 7. CONJUGACY

Proof. If pa1 ¨ ¨ ¨ ak q is a k-cycle and ρ P Sn , then the conjugate ρpa1 ¨ ¨ ¨ ak qρ´1 is the
k-cycle pρpa1 q ρpa2 q ¨ ¨ ¨ ρpak qq because

ρpa1 q ÞÝÑ a1 ÞÝÑ a2 ÞÝÑ ρpa2 q

ρpa2 q ÞÝÑ a2 ÞÝÑ a3 ÞÝÑ ρpa3 q etc.

If σ “ σ1 σ2 ¨ ¨ ¨ σl is a product of disjoint cycles, then ρσρ´1 “ ρσ1 ρ´1 ρσ2 ρ´1 ¨ ¨ ¨ ρσl ρ´1 , so it has
the same cycle type as σ.
Conversely, if σ, τ have the same cycle type, say

σ “ pa1 a2 ¨ ¨ ¨ ak qpak`1 ¨ ¨ ¨ ak`l q ¨ ¨ ¨

τ “ pb1 b2 ¨ ¨ ¨ bk qpbk`1 ¨ ¨ ¨ bk`l q ¨ ¨ ¨

˜ ¸
a1 ¨¨¨ ak`l ¨¨¨ an
then we use ρ “ . Then ρσρ´1 “ τ . So σ and τ are conjugate. 
b1 ¨¨¨ bk`l ¨¨¨ bn

We should do a sanity check: ρ really is a permutation because σ and τ are in disjoint cycle
notation.

Example: Ccls of S4

example element cycle type size of ccl size of centraliser sign

e 1111 1 24 +
p12q 211 6“ 4¨32 4 -
p12qp34q 22 3“ 4¨3
2¨2 8 +
p123q 31 8“ 4¨3¨2
3 3 +
p1234q 4 6“ 4¨3¨2¨1
4 4 -

The sizes of the centralisers are calculated using the Orbit-Stabiliser Theorem.

What can this tell us?

Recall: A normal subgroup is a union of ccls.

Example: Normal subgroups of S4 . They must contain e, the order must divide 24, they must be
a union of ccls.
order 1 teu
order 4 te, p12qp34q, p13qp24q, p14qp23qu “ V4 – C2 ˆ C2 .
Check it really is a subgroup (Klein four group).
order 12 A4 (kernel of signature)
order 24 S4

So the quotients of S4 (or homomorphic images) are:

S4 {teu “ S4 S4 {V4 “ tV4 , p12qV4 , p13qV4 , p23qV4 , p123qV4 , p132qV4 u – S3 p– D6 q

c.f. “groups of order 6” Proposition 44
S4 {A4 – C2 S4 {S4 – teu

Exercise: Repeat this for S5 .

IA Groups Page 44 c Julia Goedecke, Michaelmas 2016
CHAPTER 7. CONJUGACY

Conjugacy in An
We have seen that |Sn | “ 2 ¨ |An | and that ccls in Sn are nice. What about ccls in An ?
First thought:
cclSn pσq “ tτ P Sn | Dρ P Sn s.t. τ “ ρσρ´1 u
cclAn pσq “ tτ P An | Dρ P An s.t. τ “ ρσρ´1 u
Obviously cclAn pσq Ď cclSn pσq, as An Ď Sn . But it could be smaller.
E.g. p123q and p132q are conjugate in S3 by p23q, i.e. p23qp123qp23q “ p132q. But p23q R A3 .
[BUT in S5 , also p23qp45qp123qp23qp45q “ p132q!]
We use Orbit-Stabiliser:
|Sn | “ |cclSn pσq||CSn pσq| and
|An | “ |cclAn pσq||CAn pσq|
As |An | “ 21 |Sn |, there are two options:
˛ either cclSn pσq “ cclAn pσq and |CAn pσq| “ 12 |CSn pσq|
˛ or 21 |cclSn pσq| “ |cclAn pσq| and CAn pσq “ CSn pσq.

Definition: When |cclAn pσq| “ 21 |cclSn pσq|, we say that the conjugacy class of σ splits in An .

When does this happen?

Note that ccls cannot split into more than two in An !
Recall: Any subgroup of Sn has either none or half of its elements odd (Proposition 22).

63 Proposition: (“splitting ccls”)

For σ P An , the ccl of σ splits in An if and only if no odd permutation commutes with σ.

Proof. We have |cclAn pσq| “ 21 |cclSn pσq| if and only if CAn pσq “ CSn pσq. Clearly CAn pσq “
CSn pσq X An , so this happens iff CSn pσq Ď An , i.e. σ commutes with no odd permutation. 

Example: Ccls in A4
example element cycle type |cclS4 | odd element in CS4 ? |cclA4 |
e 1111 1 yes, e.g. p12q 1
p12qp34q 22 3 yes, e.g. p12q 3
p123q 31 8 no: see below 4
4

CS4 pp123qq “ xp123qy. See “ccls of S4 ”: the size is 3, and we know that definitely all powers of
p123q commute with it, so that is everything.

Example: Ccls in A5
example element cycle type |cclS5 | odd element in CS5 ? |cclA5 |
e 11111 1 yes, e.g. p12q 1
p12qp34q 221 15 yes, e.g. p12q 15
p123q 311 20 yes, e.g. p45q 20
p12345q 5 24 no: see below 12
12

64 Lemma: σ “ p12345q P S5 has CS5 pσq “ xσy.

IA Groups Page 45 c Julia Goedecke, Michaelmas 2016
CHAPTER 7. CONJUGACY

Proof. We first find the size using Orbit-Stabiliser: |cclS5 pσq| “ 24, |S5 | “ 120, so |CS5 pσq| “
5. Clearly all powers of σ commute with σ, so xσy ď CS5 pσq. But by sizes, CS5 pσq “ xσy. 

65 Theorem: A5 is simple.

Proof. We know that normal subgroups must be unions of ccls, must contain e and their
order must divide |A5 | “ 60. The sizes of the ccls in A5 are 1, 15, 20, 12, 12. But the only
options of adding a subset of those including 1 to give a number which divides 60 is 1 “ 1 or
1 ` 15 ` 20 ` 12 ` 12 “ 60. So only teu E A5 and A5 E A5 , so A5 is simple. 

In fact, all An for n ě 5 are simple (see GRM second year).

Conjugation used in the Möbius group

Continuing with the theme of “conjugate elements have similar properties”, we will now see in
particular what conjugate Möbius maps have in common. First we will find the conjugacy classes
of the Möbius maps.

66 Proposition: (“Conjugacy classes of Möbius maps”)

Any Möbius map is conjugate to f pzq “ νz for some ν ‰ 0, or to f pzq “ z ` 1.

Proof. We have a surjective group homomorphism θ : GL2 pCq ÝÑ M (see Prop. 40 “Möbius
maps via matrices”).
The ccls in GL2 pCq are of type
˜ ¸
λ 0 λz ` 0 λ
ÞÝÑ gpzq “ “ z λ, µ ‰ 0
0 µ 0z ` µ µ
˜ ¸
λ 0
or ÞÝÑ gpzq “ z “ 1 ¨ z identity
0 λ
˜ ¸
λ 1 λz ` 1 1
or ÞÝÑ gpzq “ “z` λ‰0
0 λ λ λ
˜ ¸ ˜ ¸
1 λ1 1 1
But in fact is conjugate to :
0 1 0 1
˜ ¸˜ ¸˜ ¸ ˜ ¸
λ 0 1 λ1 1
λ 0 1 1
“ ÞÝÑ gpzq “ z ` 1.
0 1 0 1 0 1 0 1


As an example of properties a Möbius map can have, we will look at fixed points. Any Möbius
map with c “ 0 fixes 8, and z ÞÝÑ z ` 1 only fixes 8, whereas z ÞÝÑ 2z also fixes 0. What can we
say in general?

67 Proposition: Any Möbius map with at least three fixed points is the identity.

Proof. Consider f pzq “ az`b az`b

cz`d . This has fixed points at those z which satisfy cz`d “ z ô
2 2
az ` b “ cz ` dz ô cz ` pd ´ aqz ´ b “ 0. This quadratic has at most two roots, unless c “ b “ 0
and d “ a, in which case f “identity. 

Now we see easily that (for ν ‰ 0, 1) νz has 0, 8 as fixed points, and z ` 1 only 8. Does this
transfer to their conjugates?
IA Groups Page 46 c Julia Goedecke, Michaelmas 2016
CHAPTER 7. CONJUGACY

68 Proposition: (“fixed points of Möbius maps”)

Every non-identity Möbius map has exactly one or two fixed points.

Proof. Given f P M , f ‰ id, there is h P M s.t. hf h´1 pzq “ νz or z ` 1 (for ν ‰ 0, 1).

Now f pwq “ w ô hf h´1 phpwqq “ hpwq, so f and hf h´1 have the same number of fixed points
(c.f. Sheet 4 Q1).
So f has either exactly 2 (for hf h´1 pzq “ νz, ν ‰ 0, 1) or exactly 1 (for hf h´1 pzq “ z ` 1)
fixed points. 

Conjugation action on subgroups

Going back to the general conjugation action, we can not only act on elements of a group, but
also on a different set:

69 Lemma: (conjugation action on subgroups)

Let X be the set of subgroups of G. Then G acts by conjugation on X.

Proof. We check the three conditions of an action:

0. If H ď G, then we have to show that gHg ´1 is also a subgroup:
˛ e P H ñ geg ´1 “ e P gHg ´1 so it is non-empty.
˛ Any two elements in gHg ´1 are of the form gag ´1 , gbg ´1 for some a, b P H. Then
gag ´1 pgbg ´1 q´1 “ gag ´1 gb´1 g ´1 “ gab´1 g ´1 P gHg ´1 as ab´1 P H.
So by the super-efficient subgroup criterion (Lemma 4), gHg ´1 ď G.
1. eHe´1 “ H.
2. g1 pg2 Hg2´1 qg1´1 “ pg1 g2 qHpg1 g2 q´1 .
So this is indeed an action. 

Remark: If K EG, then the ccl of K under this action is just tKu. “Normal subgroups are exactly
the ones with singleton ccls.”

Exercise: If H1 , H2 are conjugate subgroups, then they are isomorphic. In particular, they have
the same size.

The stabilisers of this action are often called normalisers (but we won’t need this word in
our course).
NG pHq “ tg P G | gHg ´1 “ Hu
“NG pHq is the largest subgroups of G such that H E NG pHq.” (Exercise)
Clearly H ď NG pHq. These normalisers will come up a lot in the second year course GRM, in
the context of Sylow subgroups for example.

We saw above that conjugate elements, i.e. elements in the same ccl, have similar properties.
Now ccls are a particular example of orbits, and in a general action, elements in the same orbit do
also have some important properties in common:

70 Lemma: (“Stabilisers of elements in same orbit are conjugate”)

Let G act on X, and let g P G, x P X. Then
Stabpgpxqq “ gStabpxqg ´1
por Ggpxq “ gGx g ´1 q

Proof. Example Sheet 4 Q2. 

So they are in particular isomorphic.

IA Groups Page 47 c Julia Goedecke, Michaelmas 2016

 CHAPTER 8

Actions in action

Using actions
Using actions, we can find out lots of things about groups. We will give some examples here.
We can use the Orbit-Stabiliser Theorem to find sizes of groups.

Example: (Cube) Let G` be the group of all rotations of a cube, acting on the vertices. So
X “ tverticesu, |X| “ 8. The action is transitive (convince yourself).
What is the stabiliser of vertex 1? It is exactly all rotations in the axis through 1 and the
diagonally opposite vertex. There are three of those (including e). So |G| “ |orbp1q||Stabp1q| “
8 ¨ 3 “ 24.

Having found the size of a group, we can also use the Orbit-Stabiliser Theorem to find sizes
of stabilisers.

Example: (Cube) Let G` be rotations of a cube as above, but now acting on the three axes
(=pairs of opposite faces), let us call them x, y, z.

y
z

We see that this action is transitive, so |orbpxq| “ 3. So |Stabpxq| “ 24{3 “ 8. This helps to find
the elements of the stabiliser: we certainly have any rotation in the axis x, which gives 4 elements
(this includes e). Then we have rotations of 1800 in the axis y and in the axis z. We know we are
missing two so we look a bit harder and find that rotation by 1800 in the axis which goes through
the midpoints of edges parallel to the axis x also fixes x. There are two of these, so now we know
all elements in the stabiliser. If we hadn’t known the size, we may not have thought of looking for
these last two.

Notice that the first four elements we mentioned keep the endpoints of the axis x (i.e. both
faces of the “pair of opposite faces”) fixed, whereas the other four switch them.
Looking at the homomorphism ϕ : G` ÝÑ S3 induced by this action, we now see that the
kernel of ϕ is the intersection of the stabilisers of x, y and z, so contains exactly four elements: e
and rotation by 1800 around any of the three axes.
All this is very nicely visualised on the “Taylor-Cube”, which you can find on the Moodle
course page, my blog or on Gareth Taylor’s website.
48
CHAPTER 8. ACTIONS IN ACTION

We have used the alternative action definition (Prop. 50) via a homomorphism to Sn above.
We can use this homomorphism also to say something about the existence of normal subgroups or
existence of particular size subgroups.

Example: G finite, H ď G of index n, G acts on left cosets of H by left multiplication. This gives
a group homomorphism (which is non-trivial for H ‰ G)
ϕ : G ÝÑ Sn (as n cosets of H)
Now Ker ϕ E G, and we know Ker ϕ ‰ G. So
either we have found a normal subgroup if we know such an H exists (it could be teu).
or if G is simple, we know Ker ϕ “ teu, so that tells us something about the possibilities of n:
n! ě |G|.
I would like you to use this as a technique rather than quote it as a result, as it is such a flexible
argument.
Further refinement: Consider
G
ϕ
,2 Sn sgn
,2 t˘1u

The kernel of this composite is normal in G. If G is simple, Kerpsgn˝ϕq “ teu or G. Look at sizes:
can it be teu? If it is G, we know that Imϕ ď An , so n!
2 ě |G|.

We’ve seen on Sheet 1 that if |G| is even, then G has an element of order 2. In fact, we can
use our knowledge of actions to prove the very useful:

71 Theorem: (Cauchy)
Let G be a finite group, p a prime dividing |G|. Then G has an element of order p.

Idea of proof: An element of order p satisfies a ¨ a ¨ ¨ ¨ a “ e, where we have p copies of a. So we

look at p-tuples pa1 , a2 , . . . , ap q which multiply to e. A tuple of form pa, a, . . . , aq is special amongst
these: it doesn’t change under “rotation” to pa2 , a3 , . . . , ap , a1 q. So we act on such p-tuples by this
“rotation”, and use the fact that orbits partition the set and some knowledge about sizes to imply
the existence of a size 1 orbit.
Proof. G, p fixed. Consider Gp “ G ˆ ¨ ¨ ¨ ˆ G (p copies of G), the set of p-tuples of G. Let
X Ď Gp be X “ tpa1 , . . . , ap q P Gp | a1 ¨ ¨ ¨ ap “ eu “tuples that multiply to the identity”. In
particular, if an element b has order p, then pb, b, . . . , bq P X. (In fact, if pb, b . . . , bq P X and b ‰ e,
then b has order p as p is prime.)
Now let H “ xh | hp “ ey – Cp be a cyclic group of order p with generator h. Let H act on
X by “rotation”: hpa1 , . . . , ap q “ pa2 , . . . , ap , a1 q. This is an action:
0. If a1 ¨ ¨ ¨ ap “ e, then a´1 1 “ pa2 ¨ ¨ ¨ ap q, so also pa2 ¨ ¨ ¨ ap qa1 “ e, so hpa1 ¨ ¨ ¨ ap q P X.
1. epa1 , . . . , ap q “ pa1 , . . . , ap q
2. hl pa1 , . . . ap q “ pal`1 , . . . , a1 , . . . , al q “ h ¨ h ¨ ¨ ¨ hpa1 , . . . , ap q
As orbits partition X, the sum of all orbit sizes must be |X|. We know |X| “ |G|p´1 (choose first
p ´ 1 entries, last is inverse of their product).
Now, as p  |G|, also p  |X|. We have
|orbpa1 , . . . , ap q| ¨ |StabH pa1 , . . . , ap q| “ |H| “ p.
So all orbits must have size 1 or p, and they sum to |X| “ p¨(something). Clearly pe, e, . . . , eq has
orbit size 1. So there must be some (in fact at least p ´ 1) other orbits of size 1 (to get p  |X|).
These look like tpa, a, . . . , aqu for some a P G, which has order p. 
Aside notice: given one element of order p, there must always be at least p ´ 2 more as well.
You can see this by considering the powers of this element, or the cyclic subgroup generated by it.
IA Groups Page 49 c Julia Goedecke, Michaelmas 2016
CHAPTER 8. ACTIONS IN ACTION

It has size p, so is Cp , which is generated by any non-identity element, i.e. has p ´ 1 elements of
order p.

Polyhedron symmetry groups

Having found the size of the group of rotations of a cube above, we might want to know what
group it is isomorphic to.

72 Proposition: G` – S4 .

Proof. Let G` act on the four diagonal of the cube. This gives a group homomorphism
ϕ : G` ÝÑ S4 . We have p1234q P Imϕ, by rotation around axis through top and bottom face. We
also have p12q P Imϕ, by rotation around axis through midpoint of edge containing 1 and 2. So by
Sheet 2 Q2(c), Imϕ “ S4 , i.e. ϕ is surjective. But |G` | “ |S4 | “ 24, so ϕ is in fact an isomorphism.
Look at the “Taylor-Cube” to visualise all this. You can find it on the Moodle course page,
my blog, or on Gareth Taylor’s website. 
Now we will find out about all symmetries of the cube.

73 Proposition: The group G of all symmetries of the cube is isomorphic to S4 ˆ C2 .

Proof. To find the size of G, let G act on the vertices of the cube. Then as G acts transitively,
|orbp1q| “ 8. The stabiliser of 1 is
Stabp1q “ te, 2 rotations in axis through 1,
3 reflections in planes through 1 and an edge coming out of 1u
which has 6 elements. So |G| “ 8 ¨ 6 “ 48.
Consider the “reflection in midpoint”, i.e. sending each point to its opposite on the cube.
Viewing this as “´I” (in R3 ), we can easily see that this symmetry, call it τ , commutes with all
other symmetries of the cube.
We now use a technique which is almost the same as in the Direct Product Theorem (Propo-
sition 13) to show that G – G` ˆ xτ y – S4 ˆ C2 .
From the comment above we know that τ commutes with all rotations. We also know that
G` X xτ y “ teu. So (by the same arguments as in the Direct Product Theorem) we have an
injective group homomorphism
G` ˆ xτ y ÝÑ G
pg, hq ÞÝÑ gh
which must also be surjective, since |G| “ |G` ˆ xτ y|. Therefore it is an isomorphism. 
You might know that the cube is an example of certain bodies called platonic solids.

Platonic Solids: There are exactly five platonic solids.

˛ Tetrahedron: faces are regular triangles; 4 faces, 4 vertices, 6 edges. A vertex is opposite
to the mid-point of a face.
˛ Cube: faces are squares; 6 faces, 8 vertices, 12 edges. A vertex is opposite to another
vertex.
˛ Octahedron: faces are regular trianges; 8 faces, 6 vertices, 12 edges. A vertex is opposite
to another vertex.
˛ Dodecahedron: faces are regular pentagons; 12 faces, 20 vertices, 30 edges. A vertex is
opposite to another vertex.
˛ Icosahedron: faces are regular triangles; 20 faces, 12 vertices, 30 edges. A vertex is
opposite to another vertex.
IA Groups Page 50 c Julia Goedecke, Michaelmas 2016
CHAPTER 8. ACTIONS IN ACTION

Having looked at the cube, we will also study the symmetry groups of octahedra and tetrahedra
in more detail. Dodecahedron and icosahedron are studied in IB GRM.
In fact, in the result about cubes, we have already also proved that the group of symmetries of
an octahedron is S4 ˆ C2 : the octahedron is dual to the cube. Put each vertex of the octahedron
on the centre of a face of the cube, and vice versa.

(Dodecahedron and Icosahedron are also dual, but they are not done here.)

Example: (Tetrahedron)
This is self-dual in the above sense. Let’s number the vertices with 1, 2, 3, 4.
Let G` be the group of rotations acting on vertices. We see that orbp1q “ t1, 2, 3, 4u and

Stabp1q “ trotations in axis through 1 and centre of opposite sideu

2π 4π
“ te, rotation by , rotation by u.
3 3
So |G` | “ 4 ¨ 3 “ 12 by Orbit-Stabiliser.
The action gives a group homomorphism ϕ : G` ÝÑ S4 . Clearly Ker ϕ “ teu (if all four
vertices are fixed, the whole tetrahedron is fixed). So G` is a subgroup of S4 of order 12.
On this information we guess A4 . Are we right?
We get p234q, p243q P Stabp1q as seen above. Similarly we have p123q, p132q P Stabp4q,
p124q, p142q P Stabp3q, p134q, p143q P Stabp2q. So we have all 3-cycles.
Rotation in axes through opposite edge-midpoints give p12qp34q, p14qp23q, p13qp24q. So we have
indeed G` – A4 .
“There is no rotation that fixes two vertices and swaps the other two.”
Now we look at all symmetries G. We get φ : G ÝÑ S4 with Ker φ “ teu. The reflection in the
plane through 1 and 2 swaps 3, 4, etc., so now Stabp1q “ xp234q, p34qy – D6 . Think of it as the
symmetries of the triangle face opposite of 1. This gives |G| “ 4 ¨ 6 “ 24, so as φ is an injection,
we get G – S4 .

Permutation properties of Möbius maps

We have seen that any Möbius map with three fixed points is the identity.

74 Lemma: (“3-point determintation of Möbius maps”)

Given f, g P M , if there are z1 , z2 , z3 P C8 distinct such that f pzi q “ gpzi q, then f “ g.
IA Groups Page 51 c Julia Goedecke, Michaelmas 2016
CHAPTER 8. ACTIONS IN ACTION

Proof. As Möbius maps are invertible, we have g ´1 f pzi q “ zi , so g ´1 f has three fixed points
and is the identity (Prop. 67). So g ´1 f “ id ñ f “ g. 
So any Möbius map is determined by 3 points.

Definition: An action of G on X is called three-transitive if the induced action on the set

tpx1 , x2 , x3 q P X 3 | xi pairwise distinctu given by gpx1 , x2 , x3 q “ pgpx1 q, gpx2 q, gpx3 qq is transitive.
This means: for any two triples x1 , x2 , x3 and y1 , y2 , y3 of distinct elements of X, there is a
g P G s.t. gpxi q “ yi .
If this g is always unique, the action is called sharply three-transitive.

75 Proposition: The Möbius group M acts sharply three-transitively on C8 .

Proof. Given z1 , z2 , z3 distinct, then

z ´ z2 z3 ´ z1
f pzq “
z ´ z1 z3 ´ z2
sends
z1 ÞÝÑ 8
z2 ÞÝÑ 0
z3 ÞÝÑ 1.
Usual special cases:
z ´ z2
If z1 “ 8 then f pzq “
z3 ´ z2
z3 ´ z1
If z2 “ 8 then f pzq “
z ´ z1
z ´ z2
If z3 “ 8 then f pzq “ .
z ´ z1
Given also w1 , w2 , w3 distinct in C8 and g P M sending
w1 ÞÝÑ 8
w2 ÞÝÑ 0
w3 ÞÝÑ 1,
´1
then we have g f pzi q “ wi .
By “3-point determination of Möbius maps” (Lemma 74) this g ´1 f is the unique Möbius map
with that property. 
Three points also determine lines/circles.
We view a line in C – R2 as a circle on the Riemann Sphere through 8.

76 Lemma: (“equation of complex circle/straight line”)

The general equation of a circle or straight line in C is Azz ` Bz ` Bz ` C “ 0, where A, C P R,
|B|2 ą AC.

Proof. This comes from |z ´ B| “ r ą 0 for a circle and |z ´ a| “ |z ´ b| with a ‰ b for a

straight line. C.f. V+M. 
Note: A “ 0 gives a straight line. A ‰ 0, B “ 0 gives a circle centred at the origin. C “ 0
gives a circle passing through the origin (radius=|centre|).

77 Proposition: Möbius maps send circles/lines to circles/lines.

Or: Möbius maps send circles on the Riemann Sphere to circles on the Riemann Sphere.
IA Groups Page 52 c Julia Goedecke, Michaelmas 2016
CHAPTER 8. ACTIONS IN ACTION

Proof. Either: Calculate directly using w “ az`b dw´b

cz`d ô z “ ´cw`a . Then Azz ` Bz ` Bz ` C “
0 ô A1 ww ` B 1 w ` B 1 w ` C 1 “ 0 with A1 , C 1 P R.
Or: Use “geometry of Möbius maps” (Prop. 41) and check for each of the three types:
Dilation/rotation and translation are straight-forward.
1 1
z : Using w “ z we have Azz ` Bz ` Bz ` C “ 0 ô Cww ` Bw ` Bw ` A “ 0 

Example: Consider f pzq “ z´i z`i . Where does the real line go?
Real line = circle containing 8, 0, 1. f maps this to the circle containing f p8q “ 1, f p0q “ ´1,
f p1q “ ´i. So to the unit circle.
The upper half plane goes to the inside (as f piq “ 0).
“Complementary components are mapped to complementary components.”
In lectures I will add a diagram.

IA Groups Page 53 c Julia Goedecke, Michaelmas 2016

 CHAPTER 9

Quaternions

Groups of order 8
78 Lemma: (“groups of order 8”)
If G has order 8, then either:
G is abelian and isomorphic to one of C8 , C4 ˆ C2 , pC2 ˆ C2 q ˆ C2 , or
G is not abelian and isomorphic to one of D8 or Q8 (dihedral or quaternion).

Proof. If G contains an element of order 8, then G – C8 .

If all non-identity elements have order 2, then G is abelian (by Sheet 1, Q7). Let a ‰ b P Gzteu.
Then by the direct product theorem (Proposition 13, Chapter 1), xa, by – xay ˆ xby. Now take
c P Gzxa, by (such a c must exist by sizes). Then again by the direct product theorem, G –
pxay ˆ xbyq ˆ xcy – C2 ˆ C2 ˆ C2 .
So now assume G has no element of order 8, but there is an element a P G of order 4. Then
H “ xay has index 2, so H E G, and G{H – C2 (by sizes). This means, for any b P GzH, bH
generates G{H, with pbHq2 “ b2 H “ H. So b2 P xay, and in particular b2 commutes with a.
If b2 “ a or a3 , then b has order 8, contradicting our assumption. So either b2 “ e or b2 “ a2 .
(This does not imply b “ a.) Also bab´1 “ al for some l, since xay E G.
2
Now a “ b2 ab´2 “ bpbab´1 qb´1 “ bal b´1 “ pbab´1 ql “ al , where we use that b2 commutes
with a for the first step. So l2 ” 1 pmod 4q, so l ” ˘1 pmod 4q.
Cases:
˛ l ” 1 pmod 4q, i.e. bab´1 “ a, so ba “ ab, so G is abelian.
– If b2 “ e, we have G “ xa, by – xay ˆ xby – C4 ˆ C2 by the direct product theorem.
– If b2 “ a2 , then pba´1 q2 “ e (as G is abelian), and G “ xa, ba´1 y – C4 ˆ C2 .
˛ l ” ´1 pmod 4q, i.e. bab´1 “ a´1 .
– If b2 “ e, then G “ xa, b | a4 “ e “ b2 , bab´1 “ a´1 y – D8 , i.e. G is dihedral. (sizes
help to make this entirely rigorous)
– If b2 “ a2 , we potentially have a new group, called Q8 , the quaternions (see next
section).

Why potentially? The relations could lead to a contradiction.
e.g. if a5 “ e, b2 “ e, then bab “ a2 ñ a “ b2 ab´2 “ ba2 b´1 “ pbabq2 “ a4 , which is not true.

Quaternions
Definition: The set of matrices
#˜ ¸ ˜ ¸ ˜ ¸ ˜ ¸ ˜ ¸˜ ¸ ˜ ¸ ˜ ¸+
1 0 ´1 0 i 0 0 1 0 i ´i 0 0 ´1 0 ´i
, , , , , ,
0 1 0 ´1 0 ´i ´1 0 i 0 0 i 1 0 ´i 0
forms a group under matrix multiplication, called the quaternions Q8 .

Exercise: Check it is a group. It is a subgroup of GL2 pCq. In fact, Q8 ď SU2 ď SL2 pCq ď GL2 pCq.

Remark: We know that this is not any of the other groups of order 8, because:
54
CHAPTER 9. QUATERNIONS

˛ it is not abelian:
˜ ¸ ˜ ¸˜ ¸
i
0 1 0 0 i
“
0
´1 0 ´i i 0
˜ ¸˜ ¸ ˜ ¸
0 1 i 0 0 ´i
“
´1 0 0 ´i ´i 0
` 0
˘
˛ it is not dihedral: all elements except p 10 01 q and ´10 ´1 have order 4.
`i 0 ˘ ` 0 1˘
We can check that taking a “ 0 ´i and b “ ´1 0 , this group does indeed satisfy a4 “ e,
b2 “ a2 and bab´1 “ a´1 .

Now that it exists, we can think of Q8 as generated by a and b satisfying a4 “ e, b2 “

2
a , bab´1 “ a´1 .
A nicer way to think of it:
Q8 “ t1, ´1, i, ´i, j, ´j, k, ´ku

with p´1q2 “ 1, i2 “ j 2 “ k 2 “ ´1, p´1qi “ ´i etc.

ij “ k, jk “ i, ki “ j, ji “ ´k, kj “ ´i, ik “ ´j
We call it anticommutative.
Translation table:
1 ´1 i j k “ ij ´i
e a “ b2
2
a b ab a “ ab2 “ b2 a
3
˜ ¸ ˜ ¸ ˜ ¸ ˜ ¸ ˜ ¸
1 0 ´1 0 i 0 0 1 0 i
0 1 0 ´1 0 ´i ´1 0 i 0

Complete it yourself as practice.

IA Groups Page 55 c Julia Goedecke, Michaelmas 2016

 CHAPTER 10

Geometric properties of orthogonal matrices

Reminder of orthogonal matrices

Recall: orthogonal matrices are real square matrices which satisfy AT A “ I. Any orthogonal
matrix has determinant ˘1. The special orthogonal group SOn E On is the subgroup of orthogonal
matrices with determinant 1. By the Isomorphism Theorem, On {SOn – C2 .
¨ ˛
1
..
79 Lemma: On “ SOn Y ˝ . 0
‚SOn .
0 1
´1

Proof. Cosets partition the group (c.f. Lagrange, Theorem 23). 

We saw that orthogonal matrices are isometries, that is, they preserve length and “angles”,
meaning the usual inner product. Equivalently, unitary matrices preserve the complex dot product.

Two-dimensional rotations and reflections

Let’s investigate O2 pRq
` more
˘ closely.
Recall: O2 “ SO2 Y 10 ´1 0
SO2 .

80 Lemma: SO2 consists of all rotations of R2 around 0.

˜ ¸ ˜ ¸
T a b ´1 d ´b
Proof. Let A P SO2 , so A A “ I and det A “ 1. Let A “ , then A “ .
c d ´c a
So AT “ A´1 implies ad˜´ bc “ 1, c “ ´b 2 2
¸ and d “ a. This gives a ` c “ 1, so set a “ cos θ “ d,
cos θ ´ sin θ
c “ sin θ “ ´b. So A “ . What does this map do to the standard basis?
sin θ cos θ
˜ ¸ ˜ ¸ ˜ ¸ ˜ ¸
1 cos θ 0 ´ sin θ
A “ and A “ , so A is rotation by θ. (See diagram in lectures.)
0 sin θ 1 cos θ
Any rotation in R2 has this form, and so is in SO2 . 

81 Corollary: Any matrix in O2 is a rotation around 0 or a reflection in a line through 0.

Proof. If A P SO2 , then A is a rotation as above. Otherwise,

˜ ¸˜ ¸
1 0 cos θ ´ sin θ
B“ for some θ (by Lemma 79)
0 ´1 sin θ cos θ
˜ ¸
cos θ ´ sin θ
“
´ sin θ ´ cos θ
This has eigenvalues 1, ´1, so it is a reflection (in the line which is the 1-eigenspace). The line
goes through 0 because it is a linear map. 
56
CHAPTER 10. GEOMETRIC PROPERTIES OF ORTHOGONAL MATRICES

Three-dimensional rotations and reflections

82 Lemma: Every matrix in SO3 is a rotation around some axis.

Proof. Let A P SO3 . Then det A “ 1, and A is an isometry (by Lemma 36), so the eigenvalues
λ satisfy |λ| “ 1 and multiply to 1. So
either D complex eigenvalues λ, λ with λλ “ 1, so the third eigenvalue is real and is `1,
or all eigenvalues are real and multiply to 1, so they are 1, 1, 1 or ´1, ´1, 1.
So we can pick an eigenvector for eigenvalue 1 as the third basis vector, and then in some
orthonormal basis we have ˜ ¸
A1 00
A“ .
00 1
Now A1 is in R2 with det A1 “ 1, and A1 is still orthogonal, so A1 P SO2 , so by Lemma 80, A1 is a
rotation and ¨ ˛
cos θ ´ sin θ 0
A “ ˝ sin θ cos θ 0‚
˚ ‹

0 0 1
in some basis. 

83 Lemma: Every matrix in O3 is the product of at most three reflections (in planes through 0).
´1 0 0
¯
Proof. Recall O3 “ SO3 Y 01 0 SO3 .
´0cos
0 ´1
θ ´ sin θ 0
¯
So if A P SO3 , we know A “ in some basis, which is
sin θ cos θ 0
0 0 1
¨ ˛¨ ˛
1 0 0 cos θ ´ sin θ 0
A “ ˝0 ´1 0‚˝´ sin θ ´ cos θ 0‚,
˚ ‹˚ ‹

0 0 1 0 0 1
the product
´ 1of0 two¯reflections.
0
If A P 0 1 0 SO3 , then it is the product of three reflections. 
0 0 ´1
´ ´1 0 0 ¯
Note: 0 ´1 0 P O3 needs the three reflections, as it is itself not a reflection in a plane.
0 0 ´1

Exercise: Think about these statements geometrically. C.f. Beardon 11.2 Orthogonal maps and
11.3.3.

IA Groups Page 57 c Julia Goedecke, Michaelmas 2016

 CHAPTER 11

Möbius maps and cross ratio

Summary of results about Möbius maps

Will be filled in later. You can do it yourself, in fact!

Cross ratio
Recall: Given distinct z1 , z2 , z3 P C8 , there exists a unique g P M such that gpz1 q “ 8,
gpz2 q “ 0, gpz3 q “ 1.

Definition: Given four distinct points z1 , z2 , z3 , z4 P C8 , their cross ratio is rz1 , z2 , z3 , z4 s “ gpz4 q
for g as above.

So r8, 0, 1, λs “ λ (for any λ ‰ 0, 1, 8).

We know this exists and is uniquely defined because M acts sharply three-transitively on C8
(Prop. 75).
Note that different authors use different permutations of 1, 2, 3, 4. (This doesn’t matter as long
as you are consistent.) Beardon is different to the one I’ve given.
Formula:
z4 ´ z2 z3 ´ z1
rz1 , z2 , z3 , z4 s “ ¨
z4 ´ z1 z3 ´ z2
with special cases as in Prop. 75.

84 Lemma: (“double transpositions fix cross-ratio”)

For z1 , z2 , z3 , z4 distinct points in C8 , we have rz1 , z2 , z3 , z4 s “ rz2 , z1 , z4 , z3 s “ rz3 , z4 , z1 , z2 s “
rz4 , z3 , z2 , z1 s.

Proof. By inspection of formula. 

So ours is the same as the one on Wikipedia.

85 Proposition: (“Möbius maps preserve cross-ratio”)

If f P M , then rz1 , z2 , z3 , z4 s “ rf pz1 q, f pz2 q, f pz3 q, f pz4 qs.

Proof. Let g be the unique Möbius map s.t. rz1 , z2 , z3 , z4 s “ gpz4 q “ λ.

f

: '
z1  ,2 8 lr  f pz q
g
1
D!

z2  ,2 0 lr  f pz q
g
2
D!
 g
,2 1 lr 
z3 f pz3 q
D!

z4  ,2 λ
g
f pz4 q
58
CHAPTER 11. MÖBIUS MAPS AND CROSS RATIO

We know that there is a unique Möbius map with equivalent properties for the f pzi q. But gf ´1 has
this property, so is the unique map. So rf pz1 q, f pz2 q, f pz3 q, f pz4 qs “ gf ´1 pf pz4 qq “ gpz4 q “ λ. 
In fact, we see from this proof:
Given z1 , z2 , z3 , z4 distinct and w1 , w2 , w3 , w4 distinct in C8 , there is f P M with f pzi q “ wi
iff rz1 , z2 , z3 , z4 s “ rw1 , w2 , w3 , w4 s.

86 Corollary: The points z1 , z2 , z3 , z4 lie on some circle or line iff rz1 , z2 , z3 , z4 s P R.

Proof. Let C be the circle/straight line through z1 , z2 , z3 . Let g be the unique Möbius
map with gpz1 q “ 8, gpz2 q “ 0, gpz3 q “ 1 (c.f. Proposition 75). Then gpz4 q “ r8, 0, 1, gpz4 qs “
rg ´1 p8q, g ´1 p0q, g ´1 p1q, g ´1 pgpz4 qqs (as Möbius maps preserve cross-ratio, Prop. 85), which in turn
is equal to rz1 , z2 , z3 , z4 s. So z4 P C ô gpz4 q “ rz1 , z2 , z3 , z4 s P R, because Möbius maps send
circle/lines to circles/lines (Prop. 77). 

Summary of conjugacy classes

Any non-identity Mbius map is conjugate to one of these four types:

Action on Riemann Sphere

Elliptic: f pzq “ νz with |ν| “ 1 2 fixed points, other points move along circles
or f pzq “ eiθ z round the fixed points
Hyperbolic: f pzq “ νz with ν P R` , ν ‰ 1 2 fixed points, other points move on circular arcs
or f pzq “ rz, r P R` , r ‰ 1 from one fixed point to the other
Loxodromic: f pzq “ νz with ν R R` , |ν| ‰ 1 2 fixed points, other points move on spirals away
or f pzq “ reiθ z from one fixed point towards the other
Parabolic: f pzq “ z ` 1 1 fixed point, other points move on circles
through the fixed point, away on one side,
towards on the other
You can see the action on the Riemann sphere in the pictures below. These images are taken
from T. K. Carne’s lectures notes for Geometry and Groups, which can be found at
[Link]/~tkc/GeometryandGroups/[Link]

IA Groups Page 59 c Julia Goedecke, Michaelmas 2016

CHAPTER 11. MÖBIUS MAPS AND CROSS RATIO

IA Groups Page 60 c Julia Goedecke, Michaelmas 2016