Groups and Group Actions

Richard Earl

Hilary and Trinity Terms 2014

 Syllabus
Hilary Term (8 lectures)

• Axioms for a group and for an Abelian group. Examples including geometric symmetry
groups, matrix groups (GLn , SLn , On , SOn , Un ), cyclic groups. Products of groups. [2]

• Permutations of a finite set under composition. Cycles and cycle notation. Order. Trans-
positions; every permutation may be expressed as a product of transpositions. The parity
of a permutation is well-defined via determinants. Conjugacy in permutation groups. [2]

• Subgroups; examples. Intersections. The subgroup generated by a subset of a group. A

subgroup of a cyclic group is cyclic. Connection with hcf and lcm. Bezout’s Lemma. [1.5]

• Recap on equivalence relations including congruence mod n and conjugacy in a group.

Proof that equivalence classes partition a set. Cosets and Lagrange’s Theorem; examples.
The order of an element. Fermat’s Little Theorem. [2.5]

Trinity Term (8 lectures)

• Isomorphisms, examples. Groups up to isomorphism of order 8 (stated without proof).

Homomorphisms of groups with motivating examples. Kernels. Images. Normal sub-
groups. Quotient groups; examples. First Isomorphism Theorem. Simple examples de-
termining all homomorphisms between groups. [3]

• Group actions; examples. Definition of orbits and stabilizers. Transitivity. Orbits parti-
tion the set. Stabilizers are subgroups. [2]

• Orbit-stabilizer Theorem. Examples and applications including Cauchy’s Theorem and

to conjugacy classes. [1]

• Orbit-counting formula. Examples. [1]

• The representation G → Sym(S) associated with an action of G on S. Cayley’s Theorem.

Symmetry groups of the tetrahedron and cube. [1]

Recommended Texts
• M. A. Armstrong Groups and Symmetry (Springer, 1997)

2
 STANDARD GROUP NOTATION

• Z, Q, R, C — the integers/rationals/reals/complex numbers under +.

• Cn — the cyclic group of order n.

• Zn — the integers, modulo n, under +. (Zn is isomorphic to Cn )

• Q∗ , R∗ , C∗ — non-zero rationals/reals/complex numbers under ×.

• Z∗p — the non-zero elements of Zp , where p is a prime, under ×.

• Z∗n — more generally for composite n, the units of Zn , those elements coprime with n
under ×.

• (0, ∞) — the positive real numbers under ×.

• Sym(S) — the permutations (i.e. bijections S → S) of a set S under composition.

• Sn — permutations of {1, 2, . . . n} under composition.

• An — even permutations of {1, 2, . . . , n} under composition.

• D2n — the symmetries of a regular n-sided polygon under composition.

• V or V4 — the Klein four-group Z2 × Z2 ∼

= {e, (12)(34), (13)(24), (14)(23)}.

• Q8 — the quaternion group {±1, ±i, ±j, ±k}.

• S 1 — the complex numbers with unit modulus under multiplication.

• GL(n, F ) — invertible n×n matrices with entries in the field F under matrix multiplication.

• SL(n, F ) — the subgroup of GL(n, F ) whose elements have determinant 1.

• AGL(n, F ) — the affine maps of F n .

• O(n) — orthogonal n × n real matrices (A−1 = AT ) under matrix multiplication. Also

SO(n).

• U (n) — unitary n × n complex matrices (A−1 = ĀT ) under matrix multiplication. Also
SU (n).

• Aut(G) — the automorphisms (i.e. isomorphisms G → G) of a group G under composition.

• G1 × G2 — the (direct) product group of two groups G1 and G2 .

• G/H — the quotient (or factor) group of a group G by a normal subgroup H of G.

• g — the cyclic subgroup of G generated by g ∈ G.

• S — the subgroup of a group G generated by a subset S of G.

3
4
1. THE GROUP AXIOMS

1.1 Abstraction

Abstract algebra began as a late nineteenth century construct bringing together many results
and observations from the previous century. During the twentieth century, throughout much
of pure mathematics, abstraction was to play an important role in shaping the subject. Prior
to the nineteenth century, most mathematics had begun with some interest in real-world prob-
lems. Perhaps due to the paradigm-smashing discovery of non-euclidean geometries, nineteenth
century mathematicians were more willing to cut that cord. In any case, thinking abstractly,
mathematicians began looking at the underlying rules, logic and structures that drove seem-
ingly disparate results. The power of abstraction, then, is its generality: beginning with the
rules of an abstract structure, one can begin to demonstrate results that apply to all examples
of that structure. Whilst the nature of a specific structure is likely grounded in some important
concrete examples, the proof itself emerges independent of any particular examples. Still more,
with luck, the proof will be that much more apparent as it focuses on the structure’s rules only
and there is no distraction from superficial clutter.
In this first course in abstract algebra we concentrate on groups. Groups, particularly in the
form of "substitution groups", had been apparent in the work of Galois, Gauss, Cauchy, Abel et
al. in the early nineteenth century. The general axioms for a group were first written down by
Cayley in 1849, but their importance wasn’t acknowledged at the time. Two other important
algebraic structures are rings and fields — you will likely have met the field axioms in Linear
Algebra I and Analysis I. Both these concepts are due to Richard Dedekind (1831-1916) who
was arguably the father of abstract algebra; certainly he was one of the first to fully appreciate
the power of abstract structures. Fields and rings naturally lead to examples of groups also.
It wasn’t until the 1870s that an appreciation of the merit of abstract structures was showing,
especially in the nascent algebraic number theory which grew out of efforts to prove Fermat’s
Last Theorem.

1.2 Binary Operations

Definition 1 A binary operation ∗ on a set S is a map ∗ : S × S → S. We write a ∗ b for

the image of (a, b) under ∗.

So a binary operation takes two inputs a, b from S in a given order and returns a single output
a ∗ b which importantly has to be in S. Standard examples include addition, multiplication and
composition but there are many other examples as well.

THE GROUP AXIOMS 5

Example 2 The following are examples of binary operations.
(i) +, −, × on R; ÷ is not a binary operation on R as, for example 1 ÷ 0 is undefined;
(ii) ∧, the cross product, on R3 ;
(iii) min and max on N;
(iv) ◦, composition, on the set Sym(S) of bijections of a set S to itself;
(v) matrix multiplication on the set of n × n complex invertible matrices (for a given n).

Definition 3 A binary operation ∗ on a set S is said to be associative if, for any a, b, c ∈ S,

(a ∗ b) ∗ c = a ∗ (b ∗ c) .

In particular, this means an expression such as a1 ∗ a2 ∗ · · · ∗ an always yields the same result,
irrespective of how the individual parts of the calculation are performed.

Definition 4 A binary operation ∗ on a set S is said to be commutative if, for any a, b ∈ S,

a ∗ b = b ∗ a.

Definition 5 An element e ∈ S is said to be an identity element (or simply an identity)

for an operation ∗ on S if, for any a ∈ S,

e ∗ a = a = a ∗ e.

Proposition 6 Let ∗ be a binary operation on a set S and let a ∈ S. If an identity e exists

then it is unique.

Proof. Suppose that e1 and e2 are two identities for ∗. Then

e1 ∗ e2 = e1 as e2 is an identity;
e1 ∗ e2 = e2 as e1 is an identity.

Hence e1 = e2 .

Definition 7 If an operation ∗ on a set S has an identity e and a ∈ S, then we say that b ∈ S

is an inverse of a if
a ∗ b = e = b ∗ a.

Proposition 8 Let ∗ be an associative binary operation on a set S with an identity e and let
a ∈ S. Then an inverse of a, if it exists, is unique.

Proof. Suppose that b1 and b2 are inverses of a. Then

b1 ∗ (a ∗ b2 ) = b1 ∗ e = b1 ;
(b1 ∗ a) ∗ b2 = e ∗ b2 = b2 .

By associativity b1 = b2 .

6 THE GROUP AXIOMS

Notation 9 If ∗ is an associative binary operation on a set S with identity e, then the inverse
of a (if it exists) is written a−1 .

Example 10 If we look at the binary operations in Example 2 then we note:

(i) + on R is associative, commutative, has identity 0 and x−1 : = −x for any x;
− on R is not associative or commutative and has no identity;
× on R is associative, commutative, has identity 1 and x−1 : = 1/x for any nonzero x.
(ii) ∧ on R3 is not associative or commutative and has no identity;
(iii) min on N is both associative and commutative but has no identity;
max on N is both associative and commutative and has identity 0 (being the least element
of N) though no positive integer has an inverse;
(iv) ◦ is associative, but not commutative, with the identity map x → x being the identity
element and as permutations are bijections they each have inverses;
(v) matrix multiplication on n × n invertible complex matrices is associative, but not commu-
tative, with identity element In and A−1 : = A−1 (unsurprisingly).

Remark 11 Given a binary operation ∗ on S and a subset T ⊆ S, then we have a restriction

∗ : T × T → S which need not generally be a binary operation on T . We will, quite naturally,
be interested in those subsets T for which ∗ restricts to a binary operation ∗ : T × T → T on T.
In this case, T is said to be closed under ∗.

Example 12 + is a binary operation on Z.

(i) Let m ∈ Z. The set mZ = {mn : n ∈ Z} is closed with a identity 0 and inverses.
(ii) The set {n ∈ Z : n 1} is closed under + but has no identity nor inverses.

1.3 The Group Axioms

Definition 13 A group (G, ∗) consists of a set G and a binary operation ∗ on G

∗ : G × G → G, (a, b) → a ∗ b

such that
(i) ∗ is associative — that is a ∗ (b ∗ c) = (a ∗ b) ∗ c for any a, b, c ∈ G;
(ii) there is an identity e which satisfies e ∗ a = a = a ∗ e for all a ∈ G;
(iii) for each a ∈ G there exists an inverse a−1 such that a ∗ a−1 = e = a−1 ∗ a.

Remark 14 If the operation ∗ is clear and unambiguous then we will often simply say "G is
a group" as a shorthand for "(G, ∗) is a group".

Remark 15 When verifying that (G, ∗) is a group we have to check (i), (ii), (iii) above and
also that ∗ is a binary operation — that is, a ∗ b ∈ G for all a, b, ∈ G; this is sometimes referred
to as closure.

THE GROUP AXIOMS 7

Notation 16 It is common notation to suppress the binary operation ∗ when discussing groups
generally and instead to write ab for a ∗ b. We shall do this from now on.

Notation 17 Also, when n is an integer, we will write



 aaa · · · a n > 0;

 n times
an = e n = 0;

 −1 −1 −1

 a a · · · a n < 0.
−n times

Basic rules of algebra following from the group axioms are:

Proposition 18 Let G be a group, x, y, z ∈ G and m, n ∈ Z. Then

(a) (xy)−1 = y −1 x−1 .
(b) (xn )−1 = x−n .
(c) xm xn = xm+n.
(d) (xm )n = xmn .
(e) (Cancelling on the left) If xy = xz then y = z.
(f) (Cancelling on the right) If xy = zy then x = z.

Proof. Left as exercises.

Definition 19 We say that a group G is abelian, after the Norwegian mathematician Niels
Abel (1802-1829), if the group operation is commutative — that is,

g1 g2 = g2 g1 for all g1 , g2 ∈ G.

Example 20 The sets Z, Q, R, C form abelian groups under + with e = 0 and x−1 : = −x in
each case.

Example 21 The sets Q\{0}, R\{0} C\{0} form abelian groups under × with e = 1 and
x−1 : = 1/x in each case. These groups are respectively denoted as Q∗ , R∗ , C∗ .

Remark 22 More generally if (F, +, ×) is a field then (F, +) and (F \{0}, ×) are both abelian
groups.

Example 23 The set of positive real numbers (0, ∞) form an abelian group under × with e = 1
and x−1 : = 1/x.

Example 24 Any vector space forms an abelian group under +.

Example 25 Show that the n × n invertible real matrices form a group under matrix multipli-
cation. This group is denoted GL(n, R) and is called the nth general linear group.
Show that GL (n, R) is non-abelian when n > 1.

8 THE GROUP AXIOMS

Solution. Our operation, matrix multiplication, is a binary operation because if A and B are
invertible n × n real matrices then AB is invertible with (AB)−1 = B −1 A−1 .
Matrix multiplication is associative.
The group has an identity In as InA = AIn = A for all A and In is invertible.
Every element A ∈ GL(n, R) has an inverse A−1 (in the usual sense) which is invertible as
−1 −1
(A ) = A and which satisfies AA−1 = A−1 A = In .
If n 2 then we see that
   
1 1 0 1 0 0
A =  0 1 0 , B= 1 1 0 
0 0 In−2 0 0 In−2

don’t commute as (AB)11 = 2 = 1 = (BA)11 and hence GL (n, R) is non-abelian.

There are other important matrix groups. Let n be a positive integer.

Example 26 The real invertible n × n matrices with determinant 1 form a group SL(n, R)
(the special linear group) under matrix multiplication which is non-abelian for n 2.

Example 27 The orthogonal n × n matrices form a group O(n) under matrix multiplication
which is non-abelian for n 2. (Recall a real matrix A is orthogonal if A−1 = AT .)

Example 28 The orthogonal n × n matrices with determinant 1 form a group SO(n) under
matrix multiplication which is non-abelian for n 3.
T
Definition 29 A complex square matrix A is called unitary if A−1 = A , that is the transpose
of the conjugate of A.

Example 30 Show that the unitary n × n matrices form a group U (n) under matrix multipli-
cation which is non-abelian for n 2.

Solution. This is left to Exercise Sheet 1, Question 3.

Example 31 The set S 1 = {z ∈ C : |z| = 1} forms an abelian group under multiplication.

Note that the three groups S 1 , U(1), SO(2) can all be naturally identified by

cos θ − sin θ
eiθ ↔ eiθ ↔ .
sin θ cos θ

An important (if rather elementary) family of groups is the cyclic groups.

Definition 32 A group G is called cyclic if there exists g ∈ G such that

G = gk : k ∈ Z .

Such a g is called a generator. As g i g j = g i+j = g j g i then cyclic groups are abelian.

Example 33 Z is cyclic and has generators 1 and −1.

THE GROUP AXIOMS 9

Example 34 Let n 1. The nth cyclic group Cn is the group with elements

e, g, g 2 , . . . , g n−1

which satisfy g n = e. So given two elements in Cn we define

g i+j if 0 i + j < n,
gi ∗ g j = i+j−n
g if n i + j 2n − 2.

Another important family of groups is the dihedral groups.

Example 35 (The Dihedral Groups) Let n 3 be an integer and consider a regular n-sided
polygon P in the plane. We then write D2n for the set of isometries of the plane which map the
polygon back to itself. It is clear that D2n forms a group under composition as (i) the identity
map is in D2n , (ii) the product of two isometries taking P to P is another such isometry, (iii)
the inverse of such an isometry is another such isometry, (iv) composition is associative. Here
"D" stands for "dihedral", meaning two-sided.
We consider the n = 3 or D6 case first, that is where the polygon is an equilateral triangle
as in the first diagram below.

3 4 3

r
r

1 21 2

s s

n=3 n=4
We will denote rotation anticlockwise by 2π/3 as r and denote reflection in the vertical as s.
We will also label the vertices as 1, 2, 3. It is easy to see that the following symmetries are all
different
e, r, r2, s, rs, r2 s.
One way of seeing this is by noting how these symmetries permute the vertices; note that these
six symmetries respectively permute the vertices as

1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
, , , , , .
1 2 3 2 3 1 3 1 2 2 1 3 3 2 1 1 3 2

In fact, these are the only symmetries of the triangle as there are 6 = 3! permutations of the
vertices.
When n = 4 (i.e. when the group is D8 ) we will similarly denote rotation anticlockwise
by π/2 as r and again denote reflection in the vertical as s. If we will label the vertices as

10 THE GROUP AXIOMS

1, 2, 3, 4 then we note that the following eight symmetries are distinct because of their effect on
the vertices:

1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
e = , r= , r2 = , r3 = ,
1 2 3 4 2 3 4 1 3 4 1 2 4 1 2 3
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
s = , rs = , r2 s = , r3 s = .
2 1 4 3 3 2 1 4 4 3 2 1 1 4 3 2

What is slightly unclear in this case is whether there are any further symmetries. Clearly some
permutations of the vertices cannot be performed as symmetries — for example, swapping just
3 and 4 is not possible as, afterwards, 1 and 3 would no longer be opposite one another on a
diagonal. We shall see in Proposition 36 that the eight symmetries above are indeed the only
possible symmetries of the square.

Proposition 36 Let P be a regular n-sided polygon P in the plane with r denoting anticlockwise
rotation through 2π/n about P ’s centre and s denoting reflection in an axis of P . Then the
symmetries of P are
e, r, r2 , . . . rn−1 , s, rs, r2 s, . . . rn−1 s. (1.1)

Proof. Label the vertices of P as 1, 2, . . . , n in an anticlockwise order. Let t be a symmetry

of P. Once t has been effected then the vertices will either read 1, 2, . . . n anticlockwise or
clockwise. Suppose the former and say that 1 has moved to position k where 1 k n. Then
r1−k t returns 1 to position 1 and — more generally, as the vertices are in an anticlockwise order —
all vertices to their original starting positions. Hence r1−k t = e and we see t = rk−1 . If instead
t changes the vertices to a clockwise order then ts — as s flips the polygon — keeps them in an
anticlockwise order. As before we see that ts = rk−1 for some 1 k n and hence t = rk−1 s
as s−1 = s.
Hence the symmetries in (1.1) are the only possible. To see that these 2n symmetries are
distinct, well firstly each of e, r, r2 , . . . rn−1 keep the vertices in the same anticlockwise order
but each moves vertex 1 to a different place, and each of s, rs, r2 s, . . . rn−1 s reverses the order
of the vertices and each moves vertex 1 to a different position.

Given two groups G and H, there is a natural way to make their Cartesian product G × H
into a group. Recall that as a set

G × H = {(g, h) : g ∈ G, h ∈ H}.

We then define the product group G × H as follows.

Theorem 37 Let (G, ∗G ) and (H, ∗H ) be groups. Then the operation ∗ defined on G × H by

(g1 , h1 ) ∗ (g2 , h2 ) = (g1 ∗G g2 , h1 ∗H h2 )

is a group operation. (G × H, ∗) is called the product group or the product of G and H.

THE GROUP AXIOMS 11

Proof. As ∗G and ∗H are both associative binary operations then it follows easily from the
definition to see that ∗ is also an associative binary operation on G × H. We also note
eG×H = (eG , eH ) and (g, h)−1 = (g −1 , h−1 )
as for any g ∈ G, h ∈ H
(eG , eH ) ∗ (g, h) = (g, h) = (g, h) ∗ (eG , eH );
(g −1 , h−1 ) ∗ (g, h) = (eG , eH ) = (g, h) ∗ (g −1 , h−1 ).

Definition 38 The cardinality |G| of a group G is called the order of G. We say that a group
G is finite if |G| is finite.

One way to represent a finite group is by means of the group table or Cayley table (after
the English mathematician Arthur Cayley, 1821-1895).

Definition 39 Let G = {e, g2 , g3 , . . . , gn } be a finite group. The Cayley table (or group
table) of G is a square grid which contains all the possible products of two elements from G.
The product gi gj appears in the ith row and jth column of the Cayley table

Remark 40 Note that a group is abelian if and only if its Cayley table is symmetric about the
main (top-left to bottom-right) diagonal.

Example 41 The Cayley table of D6 is given below

∗ e r r2 s rs r2 s
2
e e r r s rs r2 s
r r r2 e rs r2 s s
r2 r2 e r r2s s rs
2 2
s s r s rs e r r
rs rs s r2s r e r2
r2 s r2 s rs s r2 r e
All these product can be worked out from just the three rules
r3 = e, s2 = e, sr = r2 s.
So, for example, we see
r2 s (rs) = r2 (sr) s = r2 r2 s s = r4 s2 = r.

Proposition 42 A Cayley group table is a latin square. That is, every group element appears
precisely once in each row and in each column.

Proof. Given a particular group element gk we see that the map G → G given by g → gk g
is a bijection with inverse g → gk−1 g. This means that the kth row contains each element of G
precisely once. Likewise the map G → G given by g → ggk is a bijection with inverse g → ggk−1 ,
and this means that the kth column contains each element of G precisely once.

12 THE GROUP AXIOMS

Example 43 Note that, in both D6 and D8 , the rotations by themselves — so {e, r, r2 } when
n = 3 and {e, r, r2 , r3 } when n = 4 — make groups in their own rights. Their group tables are
given below.
∗ e r r2 r3
∗ e r r2
e e r r2 r3
e e r r2
, r r r 2 r3 e
r r r2 e
r2 r 2 r 3 e r
r 2 rr e r
r3 r 3 e r r2
We can see that these groups are the cyclic groups C3 and C4 .
Put another way this means that {e, r, r2 } is a subgroup of D6 and {e, r, r2 , r3 } is a subgroup
of D8 .
Definition 44 Let G be a group. We say that a subset H ⊆ G is a subgroup of G if the group
operation ∗ restricts to make a group of H. That is H is a subgroup of G if:
(i) e ∈ H;
(ii) whenever g1 , g2 ∈ H then g1 g2 ∈ H.
(iii) whenever g ∈ H then g −1 ∈ H.
Note that there is no need to require that associativity holds for products of elements in H as
this follows from the associativity of products in G.
Example 45 If we consider the symmetry group of a (non-square) rectangle, then two sym-
metries are reflection in the vertical, which we will denote a, and reflection in the horizontal,
which we will denote b. A further symmetry is ab which is rotation by π; we will write c = ab.

4 3

1 2
a
If we label the vertices of the rectangle as 1, 2, 3, 4 then we identify e, a, b, c as
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
e= , a= , b= , c= .
1 2 3 4 2 1 4 3 4 3 2 1 3 4 1 2
These are in fact the only symmetries of the rectangle. To see this we only have to consider
the position of vertex 1 after a symmetry has been effected; once we know where 1 has moved to
then the positions of the remaining vertices are uniquely determined by the rectangle’s geometry.
The Cayley table for this group is then
∗ e a b c
e e a b c
a a e c b
b b c e a
c c b a e

THE GROUP AXIOMS 13

Note that the group is abelian.
This group — whether thought of in terms of the symmetries of a rectangle or considered as
permutations — is often called the Klein four-group, after the German mathematician Felix
Klein (1849-1925) and denoted V or V4 . (The V stands for "vier", the German for "four".)

Example 46 Groups of order four or less. What ways are there to fill a Cayley table
for a group of order n 4? We will always have an identity element e and we will label the
remaining element a, b, c, . . . depending on the order of the group.
In the case n = 1, there is clearly only one such table.
When n = 2 then the products e2 , ea, ae are all clear from the group axioms; further for the
table to be a latin square we also need that a2 = e.
When n = 3 then the products e2 , ea, ae, be, eb are all clear from the group axioms. Further
we see that ab can be neither a nor b if the table is to form a latin square. The only remaining
possibility is that ab = e and likewise that ba = e. From there, the remainder of the table is
uniquely determined on the basis of it being a latin square.
∗ e a b
∗ e a
∗ e e e a b
, e e a , .
e e a a b e
a a e
b b e a

The n = 4 situation is slightly more complicated. If we consider ba then it can (potentially) be

either e or c.
Case (i): if ba = e then b = a−1 and so ab = e also. (See first table below). But then,
focusing on the third row and column, we must also have bc = a = cb and b2 = c. (See the
second table below.) Finally, looking at the second row, we see that a2 = c and ca = b = ac and
c2 = e is the only remaining possibility. See final table.
∗ e a b c ∗ e a b c ∗ e a b c
e e a b c e e a b c e e a b c
a a ? e ? , a a ? e ? , a a c e b
b b e ? ? b b e c a b b e c a
c c ? ? ? c c ? a ? c c b a e

But is this a genuine group table? Note, in this case, that c = a2 and b = a3 . If we swap the b
and c rows and columns, we rewrite this table as
∗ e a a2 a3
e e a a2 a3
a a a2 a3 e
a2 a2 a3 e a
a3 a3 e a a2
which we have already seen as a concrete example of a group, namely the four rotations of a
square.
Case (ii) Instead now we consider the case ba = c. As ab = e then ab = c is the only
remaining possibility. (See table below.) At this point, though, the way forward is still not

14 THE GROUP AXIOMS

clear, in the sense that we can continue to create a latin square in different ways. We could
(potentially) have a2 = b or a2 = e.

∗ e a b c
e e a b c
a a ? c ? .
b b c ? ?
c c ? ? ?

Case (ii)(a) Assuming ba = c and a2 = b then we can complete the table (See first table below).
But if we note b = a2 and c = a3 then we see we in fact have reproduced the same table as in
case (i).
∗ e a b c ∗ e a a2 a3
e e a b c e e a a2 a3
a a b c e , a a a2 a3 e
b b c e a a2 a2 a3 e a
c c e a b a3 a3 e a a2
Cases (ii)(b’) and (ii)(b”) If we assume ba = c and a2 = e then we can only complete the table
as far as the first table below. We could still have either (b’) b2 = a (see second table) or (b”)
b2 = e (see third table).

∗ e a b c ∗ e a b c ∗ e a b c
e e a b c e e a b c e e a b c
a a e c b , a a e c b , a a e c b .
b b c ? ? b b c a e b b c e a
c c b ? ? c c b e a c c b a e

The second table again is nothing new as a = b2 , c = b3 , so that this is in fact the same group
as we have just met twice already. But the final table is fundamentally different, for example
as every element is self-inverse. We recognise it as the symmetry group of the rectangle or the
Klein group V4 that we met previously.

Remark 47 In all, then, there are four latin squares of size 4 × 4 with rows and columns
(e, a, b, c). In each case, the latin square represented a group though this is not always the case
(see Exercise Sheet 1, Question 2).

To make the following discussion a little easier we introduce now the idea of the order of a
group element.

Definition 48 Let G be a group and g ∈ G. The order of g, written o(g), is the least positive
integer r such that g r = e. If no such integer exists then we say that g has infinite order.

Remark 49 Note, now, that there are unfortunately two different uses of the word order: the
order of a group is the number of elements it contains; the order of a group element is the least
positive power of that element which is the identity.

THE GROUP AXIOMS 15

 In the above we have seen that in essence there is one group of order one, one of order two,
one of order three and two of order four. Though the symbols for the elements may vary, an
order two group consists of an identity element and an element of order two — so that it is in
essence C2 ; an order three group consists of an identity, an element of order three and its square
— so that it is in essence C3 ; an order four group either consists of an identity, an element of
order four, its square and cube — so that is in essence C4 — or it consists of an identity and
three distinct elements of order two which commute — so that it is essentially V4 .
The idea which rigorously captures the idea of "in essence" is that of the isomorphism. For
two groups to be the same "in essence" means that, suitably relabelled, their Cayley tables
would be the same. That relabelling would be an isomorphism.
Definition 50 An isomorphism φ : G → H between two groups (G, ∗G ) and (H, ∗H ) is a
bijection such that for any g1 , g2 ∈ G we have
φ(g1 ∗G g2 ) = φ(g1 ) ∗H φ(g2 ).
Two groups are said to be isomorphic if there is an isomorphism between them. Being iso-
morphic is fairly easily seen to be an equivalence relation. There may well be more than one
isomorphism between isomorphic groups.
The Cayley tables of two isomorphic groups would in essence be the same as follows. If
φ : G → H is an isomorphism and G = {g1 , . . . gn } so that H = {φ(g1 ), . . . , φ(gn )} then φ would
set up the following correspondence between the Cayley tables of G and H
∗G g1 ··· gj ··· gn ∗H φ(g1 ) ··· φ(gj ) ··· φ(gn)
g1 φ(g1 )
.. ..
. φ .
−→
gi gi ∗G gj φ(gi ) φ(gi ) ∗H φ(gj )
.. ..
. .
gn φ(gn )
so that if the two tables are to properly correspond under φ we need to have in each case
φ(gi ∗G gj ) = φ(gi ) ∗H φ(gj ).
Remark 51 We have seen that there are, up to isomorphism, 1, 1, 1, 2 groups of order 1, 2, 3, 4
respectively. In general it is a very difficult problem to work out how many groups there are of
a given order. The number of different groups g for each order n up to 80 are given in the table
below.
n 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
g 1 1 1 2 1 2 1 5 2 2 1 5 1 2 1 14 1 5 1 5
n 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
g 2 2 1 15 2 2 5 4 1 4 1 51 1 2 1 14 1 2 2 14
n 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
g 1 6 1 4 2 2 1 52 2 5 1 5 1 15 2 13 2 2 1 13
n 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
g 1 2 4 267 1 4 1 5 1 4 1 50 1 2 3 4 1 6 1 52

16 THE GROUP AXIOMS

2. PERMUTATION GROUPS

Definition 52 Let S be a set. A bijection S → S is called a permutation of S and the set of

permutations of S is denoted Sym(S).
If n is a positive integer, then we write Sn for Sym ({1, 2, . . . n}) .
Notation 53 Important: though this convention is not universal, in Oxford it is usual to
write permutations on the right. So for k ∈ {1, 2, . . . , n} and σ, τ ∈ Sn then we would write
kσ for σ(k) and kστ for τ (σ(k)). So the permutation στ is the composition of σ first with τ
second.
Theorem 54 Let S be a set.
(a) Then Sym(S) forms a group under composition. It is called the symmetry group of S.
(b) If |S| 3 then Sym(S) is non-abelian.
(c) The cardinality of Sn is n!
Proof. (a) We know that the composition of two bijections is a bijection. So ◦ is indeed a
binary operation on Sym(S). Further for any f, g, h ∈ Sym(S) and x ∈ S we have
x ((f g) h) = (x (f g)) h = ((xf ) g) h = (xf ) (gh) = x (f (gh))
So ◦ is an associative binary operation. The identity of Sym(S) is easily seen to be the identity
map
idS (x) = x for all x ∈ S
and the inverse of f ∈ Sym(S) is, unsurprisingly, its inverse map f −1 .
(b) If x1 , x2 , x3 are three distinct elements of S then we can define two permutations of S
by
f : x1 → x1 , x2 → x3 , x3 → x2 , x → x for other x;
g : x1 → x2 , x2 → x1 , x3 → x3 , x → x for other x;
which do not commute as
fg : x1 → x2 , x2 → x3 , x3 → x1 , x → x for other x.
gf : x1 → x3 , x2 → x1 , x3 → x2 , x → x for other x;
(c) For f ∈ Sn there are n possibilities for 1f , but as f is 1-1, and so 1f = 2f , there are n − 1
possibilities for 2f once 1f is known and likewise n − 2 possibilities for 3f etc. In all then there
are
n × (n − 1) × (n − 2) × · · · × 1 = n!
permutations of {1, 2, . . . , n}.
One (slightly cumbersome) way of writing down a permutation σ ∈ Sn is as an array
1 2 3 ··· n
1σ 2σ 3σ · · · nσ
though we shall improve on this notation with the introduction of cycle notation.

PERMUTATION GROUPS 17
Example 55 (i) So S2 is a group of order two which contains the elements

1 2 1 2
, .
1 2 2 1

The first element is the identity and the second is self-inverse.

(ii) And S3 is a group of order six with contains the elements

1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
, , , , , .
1 2 3 2 3 1 3 1 2 1 3 2 3 2 1 2 1 3

The first element is e. If we write

1 2 3 1 2 3
σ= , τ=
2 3 1 1 3 2

then we see (amongst other things) that

1 2 3 1 2 3 1 2 3
σ2 = = ;
2 3 1 2 3 1 3 1 2
1 2 3 1 2 3 1 2 3
σ3 = = = e;
3 1 2 2 3 1 1 2 3
1 2 3 1 2 3
τ2 = = e;
1 3 2 1 3 2
1 2 3 1 2 3 1 2 3
στ = = ;
2 3 1 1 3 2 3 2 1
1 2 3 1 2 3 1 2 3
τ σ2 = = .
1 3 2 3 1 2 3 2 1

The six elements of S3 are in fact

e, σ, σ2 , τ, στ , σ2 τ .

The Cayley table for S3 is

∗ e σ σ2 τ στ σ2 τ
e e σ σ2 τ στ σ2 τ
σ σ σ2 e στ σ2τ τ
σ2 σ2 e σ σ2τ τ στ
τ τ σ2τ στ e σ2 σ
στ στ τ σ2 τ σ e σ2
σ2τ σ2 τ στ τ σ2 σ e

Remark 56 Note that the six permutations listed above as the elements of S3 are the same as
those listed in Example 35 and the above Cayley table for S3 is identical to that in Example 41
once σ, τ are replaced with r, s. This shows that D6 and S3 are in fact isomorphic.

18 PERMUTATION GROUPS
Example 57 Set

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
α= , β= , γ= ,
2 4 3 1 5 3 4 5 1 2 2 5 4 3 1

in S5 . Determine the product αβγ, the inverse of β and the order of γ.

Solution. We have
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
αβγ = = ;
2 4 3 1 5 3 4 5 1 2 2 5 4 3 1 3 2 1 4 5
−1
1 2 3 4 5 3 4 5 1 2 1 2 3 4 5
β −1 = = = ;
3 4 5 1 2 1 2 3 4 5 4 5 1 2 3
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
γ2 = , γ3 = , γ4 = ,
5 1 3 4 2 1 2 4 3 5 2 5 3 4 1
1 2 3 4 5 1 2 3 4 5
γ5 = , γ6 = = e,
5 1 4 3 2 1 2 3 4 5

so that the order of γ is 6.

There is a special type of permutation, a cycle, which shall prove useful as we shall see that
any permutation can be (essentially) uniquely decomposed as a product of cycles.

Definition 58 A permutation σ ∈ Sn is a cycle if there are distinct elements a1 , a2 , . . . , ak in

{1, 2, . . . , n} such that

ai σ = ai+1 for 1 i < k; ak σ = a1 ;

and
xσ = x for x ∈
/ {a1 , a2 , . . . , ak } .
The length of such a cycle is k and we would refer to σ as a k-cycle. Note that the order of
a k-cycle is k.

Notation 59 Cycle notation: We denote the above cycle as

(a1 a2 a3 · · · ak ) .

Note that this notation isn’t unique (in fact there are k such expressions in all) and that we
have
(a1 a2 a3 · · · ak ) = (a2 a3 a4 · · · ak a1 ) = · · · = (ak a1 a2 · · · ak−1 ) .

Example 60 Note that α, β, γ, from Example 57 can be written as

α = (124) , β = (13524) , γ = (125) (34) .

So α is a 3-cycle, β is a 5-cycle and γ is not a cycle.

PERMUTATION GROUPS 19
Definition 61 Two cycles (a1 . . . ak ) and (b1 . . . bl ) are said to be disjoint if ai = bj for all
i, j.
Proposition 62 Disjoint cycles commute.
Proof. Let α = (a1 . . . ak ) and β = (b1 . . . bl ). Then
ai αβ = ai+1 β = ai+1 , ai βα = ai α = ai+1 , for i < k;
ak αβ = a1 β = a1 , ak βα = ak α = a1 ;
bi αβ = bi β = bi+1 , bi βα = bi+1 α = bi+1 , for i < l;
bl αβ = bl β = b1 , bl βα = b1 α = b1 ;
xαβ = xβ = x, xβα = xα = x, for x ∈
/ {a1 , . . . , ak , b1 , . . . , bl } .

Theorem 63 Every permutation can be written as a product of disjoint cycles. This expression
is unique up to the cycling of elements within cycles and permuting the order of the cycles.
Proof. Let σ ∈ Sn and let a1 ∈ {1, 2, . . . , n}. Consider the sequence
a1 , a1 σ, a1 σ 2 , a1 σ 3 , . . .
As the elements of the sequence are in the set {1, 2, . . . n} then the sequence must have repe-
titions so that a1 σ i = a1 σ j for some i < j. But then a1 σ j−i = a1 is an earlier repetition of a1
and we see that a1 is in fact the first element of the sequence to repeat. Say a1 σ k1 = a1 is the
first repetition of a1 . We see then that
σ acts on the set a1 , a1 σ, a1 σ 2 , . . . , a1 σ k1 −1 as the cycle a1 a1 σ a1 σ 2 . . . a1 σ k1 −1 .
The set a1 , a1 σ, a1 σ 2 , . . . , a1 σ k1 −1 is called the orbit of a1 .
If k1 = n then σ is a cycle and we are done. If not then we take a second element a2 not in
the orbit of a1 and we can similarly see that σ acts as a second cycle on the orbit of a2 . These
orbits are disjoint for if a1 σ i = a2 σ j for some i, j then a2 = a1 σ i−j and we see that a2 was in
the orbit of a1 , a contradiction. As the set {1, 2, . . . , n} is finite then these orbits eventually
exhaust the set and we see that
σ = a1 a1 σ a1 σ 2 . . . a1 σ k1 −1 a2 a2 σ a2 σ 2 . . . a2 σ k2 −1 · · · ar ar σ ar σ 2 . . . a1 σ kr −1
where r was the number of different orbits.
Suppose now that
σ = ρ1 ρ2 · · · ρk = τ 1 τ 2 · · · τ l
are expressions for σ as products of disjoint cycles. Then 1 appears in precisely one cycle ρi and
in precisely one cycle τ j . By reordering the appearances of the cycles if necessary (as they do
commute, being disjoint) we may assume that 1 appears in ρ1 and τ 1 . By cycling the elements
of the cycles ρ1 and τ 1 , if necessary, we may assume that 1 appears at the start of each cycle.
Hence we see
ρ1 = 1 1σ 1σ 2 · · · 1σ k−1 = τ 1
where k is the size of the orbit of 1. By continuing similarly with an element not in the orbit
of 1 we can show (with permitted permuting of cycles and cycling within cycles) that ρ2 = τ 2
etc. to complete the proof.

20 PERMUTATION GROUPS
Definition 64 As a consequence of the above theorem, the lengths of the various cycles of a
permutation and the number of cycles of each such length, is well-defined. The is known as the
cycle decomposition type (or just cycle type) of the permutation.

Example 65 Write the following permutations in S9 as products of disjoint cycles.

1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
α= , β= .
5 6 9 7 3 2 4 8 1 4 9 7 5 1 3 6 2 8

Write α−1 and β 272 as products of disjoint cycles.

Solution. We see that

α α α α α α α α α
1 → 5 → 3 → 9 → 1, 2 → 6 → 2, 4 → 7 → 4, 8→8

and so
α = (1539) (26) (47) (8) .
Hence we also see
α−1 = (1935) (26) (47) (8) .
We also see
β β β β β β β β β
1 → 4 → 5 → 1, 2 → 9 → 8 → 2, 3 → 7 → 6 → 3,
so that
β = (145) (298) (376) .
We then see that

 e k is a multiple of 3;
β k = (145)k (298)k (376)k = β k − 1 is a multiple of 3;
 2
β k − 2 is a multiple of 3.
as disjoint cycles commute. Hence

β 272 = β 2 = (154) (289) (367) .

Notation 66 Suppressing 1-cycles. It is typical to not bother writing 1-cycles (or fixed
points) of permutations. So — for α, α−1 as in the previous example — we will write

α = (1539) (26) (47) and α−1 = (1935) (26) (47)

with it being understood that 8 is not moved (or more generally any unmentioned elements).

Proposition 67 Let σ = ρ1 · · · ρk be an expression for σ as disjoint cycles of lengths l1 , . . . , lk .

Then the order of σ equals
lcm(l1 , . . . lk )
where lcm denotes the lowest common multiple. (Given finitely many positive integers, their
least common multiple is the smallest positive integer which is a multiple of each of them.)

PERMUTATION GROUPS 21
Example 68 (i) How many 5-cycles are there in S11 ?
(ii) How many permutations in S8 have a cycle decomposition type of two 3-cycles and one
2-cycle?

Solution. (i) 5-cycles in S11 are of the form (a b c d e). There are 11 choices for what a might
be, 10 for b, 9 for c, 8 for d and 7 for e. However we need to remember that

(a b c d e) = (b c d e a) = (c d e a b) = (d e a b c) = (e a b c d)

and so there answer is

11 × 10 × 9 × 8 × 7
= 11088.
5
(ii) Permutations in S8 that decompose to two 3-cycles and a 2-cycles are of the form

(a b c) (d e f ) (g h) .

There are 8! ways of filling in these brackets as a . . . h but we need to remember that the above
permutation also equals

(b c a) (d e f ) (g h) = (a b c) (e f d) (g h) = (a b c) (d e f ) (h g) = (d e f ) (a b c) (g h) .

The first three are equivalent rewritings that come from cycling elements within cycles whilst the
last one comes from permuting two equal length cycles. Hence the number of such permutations
is
8! 40320
= = 1120.
3 × 3 × 2! × 2 36
(i’) Note that we could have tackled the 5-cycle question in this manner as well. Thinking of
5-cycles in S11 as having cycle type

(a b c d e) (f) (g) (h) (i) (j) (k)

we see that there are 11! ways of filling these brackets as a . . . k but 5 ways of cycling a . . . e
and 6! ways of permuting the equal length cycles (f ) . . . (k) Hence the answer is

11!
= 11088.
5 × 6!

Proposition 69 In Sn there are

n!
l1k1 × l2k2 × · · · × lrkr (k1 ! × k2 ! × · · · × kr !)

permutations with a cycle type of k1 cycles of length l1 , k2 cycles of length l2 , . . . , kr cycles of

length lr . This decomposition includes 1-cycles so that

k1 l1 + k2 l2 + · · · + kr lr = n.

22 PERMUTATION GROUPS
Proof. Put in a fixed order the ki brackets. As already argued in specific cases, there are
n! ways of filling the brackets with the numbers 1, . . . , n. However the same permutation can
be written as a product of disjoint cycles in many ways as specified in Theorem 63 (though
essentially these all being the same). From that theorem we know that there are li ways of
cycling the elements of each cycle of length li and we have ki ! of permuting the cycles of length
li . Hence n! is an overcount by a factor of

l1k1 × l2k2 × · · · × lrkr (k1 ! × k2 ! × · · · × kr !) .

Example 70 How many permutations of each cycle type are there in S7 ?

Solution. The table below contains the various numbers. We need to consider the various
ways in which 7 can be composed as other integers.

type working # type working # type working #

7×6×5
7 7!/7 720 4+2 7!/ (4 × 2) 630 3 3
70
7×6×5×4
6 7!/6 840 4 4
210 3 × 2 7×6×5×4×3×2
2×2×2×3!
105
7×6×5×4×3×2
5+2 7!/ (5 × 2) 504 2×3 3×3×2!
280 2 × 2 7×6×5×4
2×2×2!
105
7×6×5×4×3 7! 7×6
5 5
504 3+2×2 3×2×2×2!
210 2 2
21
7×6×5×4×3
4+3 7!/ (4 × 3) 420 3+2 3×2
420 e 1

Without the labels 1, 2, . . . , n two permutations of the same cycle type would be indistin-
guishable. For example, a permutation in S8 which consists of two 3-cycles and one 2-cycle
would simply look like

if we were ignorant of which of the eight objects were 1, 2, . . . , 8. Here each arrow represents
the effect of applying the permutation once. This idea can be more formally captured by the
idea of conjugates.

Definition 71 Two permutations σ, τ ∈ Sn are said to be conjugate in Sn if there exists

ρ ∈ Sn such that
σ = ρ−1 τ ρ.

Theorem 72 Two permutations σ, τ ∈ Sn are conjugate if and only if they have the same cycle
type. (i.e. for any given length, the two permutations have the same number of cycles of that
length.)

We first note the following:

PERMUTATION GROUPS 23
Lemma 73 For any cycle (a1 a2 . . . ak ) and any ρ ∈ Sn we have
ρ−1 (a1 a2 . . . ak )ρ = (a1 ρ a2 ρ . . . ak ρ).
Proof. (Of Lemma) This is left to Exercise Sheet 2, Question 3.
Proof. (Of Theorem.) Suppose that τ = ρ−1 σρ and that σ = ψ 1 ψ 2 . . . ψ r where the ψ i are
disjoint cycles. Then
τ = ρ−1 (ψ 1 ψ 2 . . . ψ r ) ρ = ρ−1 ψ 1 ρ ρ−1 ψ 2 ρ · · · ρ−1 ψ r ρ
and we see by the previous lemma that the ρ−1 ψ i ρ are disjoint cycles of the same lengths as
the ψ i . Conversely, suppose that σ and τ have the same cycle decomposition type. Then we
may line up the cycles in σ and τ of corresponding lengths as
σ = (a1 a2 . . . ak ) (b1 b2 . . . bl ) (c1 c2 . . . cm ) · · ·
↓ρ ↓ρ ↓ρ ↓ρ
τ = (α1 α2 . . . αk ) (β 1 β 2 . . . β l ) (γ 1 γ 2 . . . γ m ) · · ·
and we define ρ by ai ρ = αi , bi ρ = β i , ci ρ = γ i , etc. We then have
αi ρ−1 σρ = ai σρ = ai+1 ρ = αi+1 = αi τ for 1 i<k
αk ρ−1 σρ = ak σρ = a1 ρ = α1 = αk τ
and similarly for the other cycles.
Example 74 Let
σ = (12) (34) (567) , τ = (28) (17) (345)
be permutations in S8 .
(i) How many ρ are there in S8 such that σ = ρ−1 τ ρ?
(ii) How many ρ ∈ S8 are there which commutes with σ?
Solution. (i) We need ρ such that
ρ−1 τ ρ = (2ρ 8ρ) (1ρ 7ρ) (3ρ 4ρ 5ρ) = (12) (34) (567) .
Thinking about the different ways of rewriting (12) (34) (567) (as the same permutation) we
see that we need
(2ρ 8ρ) (1ρ 7ρ) = (12) (34) and (3ρ 4ρ 5ρ) = (567) .
and so
3ρ = 5 or 6 or 7, 2ρ = 1 or 2 or 3 or 4, 6ρ = 8.
Once we know 3ρ then 4ρ and 5ρ are known (e.g. 3ρ = 6 implies 4ρ = 7 and 5ρ = 5). Once we
know 2ρ then we know 8ρ but we still have two choices for 1ρ. In all then we see that there are
3 × 4 × 2 = 24
choosing 3ρ choosing 2ρ choosing 1ρ

such ρ.
(ii) If we replace τ with σ we can still make the same argument to realize that there are 24
such ρ that σ = ρ−1 σρ. However this is an equivalent equation to ρσ = σρ so these same 24
permutations commute with σ.

24 PERMUTATION GROUPS
Remark 75 If further asked which these 24 permutations are recall that we need

(1ρ 2ρ) (3ρ 4ρ) (5ρ 6ρ 7ρ) = (12) (34) (567) .

So the 24 permutations in fact comprise the group D8 × C3 where D8 is the symmetry group of
the square with labels 1, 2 on one diagonal and 3, 4 on another and C3 = {e, (567) , (576)} .

Recall now the definition of a permutation matrix from Linear Algebra II:

Definition 76 (i) An n × n matrix is a permutation matrix if each row and each column
contain a single entry 1 and all other entries are 0.
(ii) We can associate with σ ∈ Sn a permutation matrix Pσ such that the 1 entry in row i
of Pσ is in column iσ.
Note that the (i, j)th entry of Pσ is δ iσ j .

Example 77 With n = 3 :
     
0 1 0 0 1 0 0 0 1
P(12) =  1 0 0  , P(123) =  0 0 1 , P(132) =  1 0 0 .
0 0 1 1 0 0 0 1 0

Note that P(12) is self-inverse and that P(123) is the inverse of P(132) .

Proposition 78 (a) For σ ∈ Sn then Pσ is indeed a permutation matrix.

(b) For σ, τ ∈ Sn then Pστ = Pσ Pτ .

Proof. (a) By definition Pσ has precisely a single entry 1 in each row. And the only row with
an entry of 1 in the ith column is row iσ −1 .
(b) By definition of matrix multiplication
n n
(Pσ Pτ )ij = (Pσ )ik (Pτ )kj = δ iσ k δ kτ j = δ iστ j = (Pστ )ij .
k=1 k=1

We will make use of permutation matrices in showing that the parity of a permutation is
well-defined.

Definition 79 (i) A transposition is another term for a 2-cycle.

(ii) A permutation is said to be odd (resp. even) if it can be written as a product of an odd
(resp. even) number of transpositions.

Lemma 80 If σ is a transposition then det Pσ = −1.

Proof. This is equivalent to knowing that swapping two rows of a matrix multiplies its deter-
minant by −1. For if σ = (ij) then

det Pσ = det (In with rows i and j swapped) = − det In = 1.

PERMUTATION GROUPS 25
Theorem 81 (a) Every permutation can be written as a product of transpositions (and conse-
quently is either even and/or odd).
(b) No permutation is both even and odd.
Proof. (a) Any permutation my be written as a product of disjoint cycles by Theorem 63.
And any cycle may be written as a product of transpositions as
(a1 a2 a3 · · · ak ) = (a1 a2 ) (a1 a3 ) · · · (a1 ak )
as the product of transpositions has the effect
a1 → a2 → a2 , ai → ai → a1 → ai+1 → ai+1 for i 2.
(a1 a2 ) remainder first i−2 (a1 ai ) (a1 ai+1 ) remainder

(b) If σ is expressible as the product of k transpositions, then by the above lemma det Pσ =
(−1)k . Hence no permutation can be both even and odd.
Remark 82 Note that cycles of even (resp. odd) length are (somewhat annoyingly) odd (resp.
even). So a permutation is even if and only if its cycle type has an even number of even length
cycles.
Example 83 If we return to Example 70 then we see that the following permutations were the
even ones.
type working # type working # type working
#
7×6×5×4×3×2 7×6×5×4
7 7!/7 720 2×3 3×3×2!
280 2×2 105
2×2×2!
7×6×5×4×3 7!
5 5
504 3 + 2 × 2 3×2×2×2! 210 e 1
7×6×5
4 + 2 7!/ (4 × 2) 630 3 3
70 TOTAL 2520
Note that precisely half the permutations are even.
Proposition 84 (a) The even permutations in Sn form a subgroup An .
(b) For n 2, the order of An is 12 n!.
(c) An is non-abelian for n 4.
An is called the alternating group.
Proof. (a) If
σ = ρ1 ρ2 · · · ρ2k and τ = ψ 1 ψ 2 · · · ψ 2l
are expressions for σ and τ as products of even numbers of transpositions then
στ = ρ1 ρ2 · · · ρ2k ψ 1 ψ 2 · · · ψ 2l and σ −1 = ρ2k ρ2k−1 · · · ρ1
are clearly even. The identity is also even as e is the product of zero transpositions. Hence An
is a subgroup of Sn .

(b) The permutation (12) is odd; so the maps

An → Acn given by σ → (12)σ; Acn → An given by σ → (12)σ;
are inverses of one another and so |An | = |Acn | = 12 n!.

(c) If n 4 then note (123) and (124) are even permutations which do not commute.

26 PERMUTATION GROUPS
Example 85 (123) and (132) are not conjugate in A4 .

Solution. Suppose that σ −1 (123) σ = (132). Then (1σ 2σ 3σ) = (132). As

(132) = (213) = (321)

are the only ways to write the permutation (132) then there are three possibilities

1σ = 1, 2σ = 3, 3σ = 2, 4σ = 4;
1σ = 2, 2σ = 1, 3σ = 3, 4σ = 4;
1σ = 3, 2σ = 2, 3σ = 1, 4σ = 4.

That is σ equals (23) or (12) or (13) . As none of these is even, then (123) and (132) are not
conjugate in A4 .

Example 86 The conjugacy classes in A4 are {e} and

{(12) (34) , (13) (24) , (14) (23)} , {(123) , (134) , (214) , (324)} , {(132) , (143) , (124) , (234)} .

Solution. Note that

(123)−1 (12) (34) (123) = (23) (14) , (123)−1 (23) (14) (123) = (31) (24) .

Also

(123)−1 (134) (123) = (214) , (123)−1 (214) (123) = (324) , (134)−1 (214) (134) = (123) .

As conjugacy in A4 implies conjugacy in S4 (though not conversely) and a 3-cycle is not con-
jugate in A4 with its inverse (by the previous example), then the conjugacy classes in A4 are
as given.

PERMUTATION GROUPS 27
28 PERMUTATION GROUPS
3. MORE ON SUBGROUPS & CYCLIC GROUPS

Recall that we say a subset H of a group G is a subgroup of G if H is a group in its own

right under the restriction of G’s group operation. We write this

H G.

There is a simple check — the subgroup test — for determining whether a subset is a subgroup.

Proposition 87 (Subgroup Test) Let G be a group. Then H ⊆ G is a subgroup of G if and

only if H is non-empty and whenever x, y ∈ H then x−1 y ∈ H.

Proof. =⇒ Suppose that H G. Then e ∈ H and so H is non-empty. Further for x, y ∈ H

we have x−1 y ∈ H as H is closed under products and inverses.
⇐= Suppose that the subgroup test applies. As H = ∅ then there is some h ∈ H and
so by the test e = h−1 h ∈ H. Further if x, y ∈ H then by the test x−1 = x−1 e ∈ H and
xy = (x−1 )−1 y ∈ H. Thus H is closed under products and inverses. Finally the associativity
of the group operation on H is inherited from its associativity on G.

Example 88 The subgroups of S3 are

{e}, {e, (12)}, {e, (13)}, {e, (23)}, A3 , S3 .

Solution. The listed subgroups are certainly subgroups of S3 . To see that these are the only
subgroups suppose that H S3 . Certainly e ∈ H. If |H| = 2 then H must consist of e and a
non-trivial self-inverse element. If |H| = 3 then it must be of the form {e, g, g 2 } where g 3 = e
and A3 is the only such subgroup. If |H| 4 then H must either (i) contain all three 2-cycles
or (ii) a 2-cycle and a 3-cycle. As the product of two 2-cycles in S3 is a 3-cycle, we see case (ii)
in fact subsumes case (i). Also if H contains a 3-cycle then it contains its inverse. So, without
any loss of generality we may assume this 2-cycle and 3-cycle to be (12) and (123). But then

(13) = (12)(123), (23) = (123)(12), (132) = (123)2

and we see that H = S3 .

Example 89 The subgroups of D8 are

{e}, {e, r2 }, {e, s}, {e, rs}, {e, r2 s}, {e, r3 s},
{e, r, r2 , r3 }, {e, r2 , s, r2 s}, {e, r2 , rs, r3 s}, D8 .

Solution. The listed subgroups are certainly subgroups of D8 . To see that these are the only
subgroups suppose that H D8 . Certainly e ∈ H. If |H| = 2 then H must consist of e and
a non-trivial self-inverse element and the possibilities are listed above. If |H| = 3 then it must
be of the form {e, g, g 2 } where g 3 = e but there is no such g ∈ D8 . A subgroup of order 4 must

MORE ON SUBGROUPS & CYCLIC GROUPS 29

be of the form {e, g, g 2 , g 3 } where g 4 = e or {e, a, b, ab} where a2 = b2 = e and ab = ba. For the
former, only g = r or g = r3 will do which both lead to the same subgroup {e, r, r2 , r3 }. For
the latter we must have two reflections and a rotation; further the rotation must be r2 if it is
to commute with the reflections. So the only possibilities are {e, r2 , s, r2 s} and {e, r2 , rs, r3 s}.
If |H| 5 then H must contain a rotation and a reflection; if the rotation is r or r3 then it
and the reflection will lead to all of D8 but if there are three or more reflections then at least
one must be in a diagonal and one in the vertical or horizontal and so their product is r or r3 .
Hence H = D8 is the only subgroup of order greater than 4.

Example 90 The subgroups of C6 = {e, g, g 2 , g 3 , g 4 , g 5 } are

{e}, {e, g 3 }, {e, g 2 , g 4 }, C6 .

The only subgroups of C5 are {e} and C5 .

Solution. The only non-trivial self-inverse element in C6 is g 3 and the non-trivial solutions of
x3 = e are g 2 , g 4 . If H C6 and |H| 4 then either g ∈ H or g 5 = g −1 ∈ H (both of which
−1
lead to H = C6 ) or {e, g , g , g } ⊆ H in which case g 3 (g 2 ) = g ∈ H in which case H = C6
2 3 4

is the same conclusion.

3
If H C5 and g ∈ H then H = C5 but if g 2 ∈ H then (g 2 ) = g ∈ H, and if g 3 ∈ H then
−1 −1
(g 3 ) = g 2 ∈ H and if g 4 ∈ H then (g 4 ) = g ∈ H. So H = {e} or H = C5 .

Remark 91 You may have noticed that in each of the previous examples, |H| divides |G| and
this is indeed the case. This result is known as Lagrange’s Theorem which we will prove in
the next chapter.

Proposition 92 Let G be a group and H, K be subgroups of G. Then H ∩ K is a subgroup.

Proof. This is left as Exercise Sheet 3, Question 2.

In fact, it is very easy to generalize Proposition 92 to show that if Hi (where i ∈ I) form a

collection of subgroups of G then
Hi G.
i∈I

Thus we may make the following definition.

Definition 93 Let G be a group and S a subset of G.

(i) The subgroup generated by S, written S , is the smallest subgroup of G which contains S.
(This is well-defined as G is a subgroup of G which contains S and S is then the intersection
of all such subgroups.)
(ii) If g ∈ G, then we write g rather than the more accurate but cumbersome {g} .
(iii) If S = G then the elements of S are said to be generators of G.

Example 94 Determine S in each of the following cases:

(i) G = Z, S = {12, 42}.
(ii) G = S4 , S = {(123) , (12) (34)} .
(iii) G = Q∗ , S = {3, 32 }.

30 MORE ON SUBGROUPS & CYCLIC GROUPS

Solution. (i) Note that 6 = 42 − 3 × 12; hence 6 ∈ S and 6Z ⊆ S . But as 12 = 2 × 6 and
42 = 7 × 6 then S ⊆ 6Z. This S = 6Z.
(ii) As (123) ∈ A4 and (12) (34) ∈ A4 then certainly S ⊆ A4 . If we write σ = (123) and
τ = (12)(34) then we see that the following are also in A4 .
e, σ = (123), τ στ = (214), (τ στ )2 = (124) ,
σ 2 τ σ = (14)(23), στ σ 2 = (13)(24), τ = (12)(34), σ 2 = (132),
στ = (243), (στ )2 = (234), τ σ = (134), (τ σ)2 = (143).
Hence S = A4 .
(iii) We have 3 ∈ S and so 3m ∈ S for all m ∈ Z. Likewise 2 = 32 × 3 ∈ S so that 2n ∈ S
for all n ∈ Z. So 2n 3m ∈ S for m, n ∈ Z. But as these form a subgroup of Q∗ (see the next
example) we have
S = {2n 3m : m, n ∈ Z}.

Example 95 Show that if G is abelian and g, h ∈ G then

g, h = {g r hs : r, s ∈ Z} .
Solution. Certainly {g r hs : r, s ∈ Z} ⊆ g, h . However, when G is abelian (or indeed if just
gh = hg), then {g r hs : r, s ∈ Z} is a subgroup as follows:
(i) e = g 0 h0 ∈ {g r hs : r, s ∈ Z} ;
(ii) g k hl g K hL = g k+K hl+L ∈ {g r hs : r, s ∈ Z} ;
−1
(iii) g k hl = h−l g −k = g −k h−l ∈ {g r hs : r, s ∈ Z} .
Remark 96 In several of the results that follow, notably Proposition 97 and Theorem 102
we make use of the following fact, known as the division algorithm, which we will take as
self-evident.
• Let a, b be integers with b > 0. Then there exist unique integers q, r such that a = qb + r
and 0 r < b.
Proposition 97 Let G be a group and g ∈ G. Then
(a) g = g k : k ∈ Z .
(b) If o(g) is finite then g = e, g, g 2 , . . . , g o(g)−1 .
Proof. (a) As g k ∈ g for any integer k, so it only remains to show that H = g k : k ∈ Z is
indeed a subgroup. Using the subgroup test we note g 0 = e ∈ H and that if g k , g l ∈ H then
(g k )−1 g l = g −k g l = g l−k ∈ H.
Hence g = H.
(b) It is again clear that e, g, g 2 , . . . , g o(g)−1 ⊆ g . Also for any k ∈ Z there exist q, r ∈ Z
such that k = qo(g) + r where 0 r < o(g). Then
q
g k = g qo(g)+r = g o(g) g r = eq g r = g r ∈ e, g, g 2 , . . . , g o(g)−1 .

MORE ON SUBGROUPS & CYCLIC GROUPS 31

Remark 98 Recall that we say that a group G is cyclic if there exists g ∈ G such that G = g .
Note also that a cyclic group is necessarily abelian.

Remark 99 Note that in a finite group G, then g is a generator if and only if o(g) = |G|.

Example 100 (i) C6 is cyclic with generators g and g 5 .

(ii) C5 is cyclic with generators g, g 2 , g 3 , g 4 .
(iii) C2 × C2 is not cyclic as the elements have orders 1, 2, 2, 2.
(iv) C2 × C3 is cyclic. If C2 = {e, g} and C3 = {e, h, h2 } then (g, h) and (g, h2 ) are both
generators of C2 × C3 as they have order 6 (check!).
(v) Q is not cyclic: clearly 0 = Q and if q = 0 then 12 q ∈
/ q = qZ. By the same reasoning
we see that Q cannot be generated by finitely many elements.

Theorem 101 Let G be a cyclic group.

(a) If |G| = n is finite, then G is isomorphic to Cn .
(b) If |G| is infinite, then G is isomorphic to Z.

Proof. (a) Let g be a generator of G. Then

G = g = e, g, g 2 , . . . , g n−1

by Proposition 97 (b) and multiplication in G is as given in Example 34 as g n = e.

(b) If g is a generator of G with infinite order, then we can define a map φ : G → Z by
r
φ(g ) = r which is an isomorphism.

Theorem 102 Let G be a cyclic group and H G. Then H is cyclic.

Proof. Let G = g . If H = {e} then H = e and we are done. Otherwise, we define

n = min k > 0 : g k ∈ H .

To show that n is well-defined, note that g k ∈ H = {e} for some k = 0. As H is a subgroup

−1
then g −k = g k ∈ H also. As one of ±k is positive, n is well-defined. We will show that

H = gn .

As g n ∈ H then g n ⊆ H. Conversely say that g a ∈ H. Then, by the division algorithm, there

exist q, r ∈ Z such that a = qn + r where 0 r < n. But then

g r = g a−qn = g a (g n )−q ∈ H

as g a ∈ H and g n ∈ H. By the minimality of n then r = 0 and g a = (g n )q ∈ g n .

Corollary 103 The subgroups of Z are each of the form mZ where m ∈ Z.

32 MORE ON SUBGROUPS & CYCLIC GROUPS

Proposition 104 Let m, n be non-zero integers. By Theorem 102 we have

m, n = h m ∩ n = l

for some h, l > 0. Then h has the following properties:

(a) h|m and h|n;
(b) if x|m and x|n then x|h;
(c) there exist u, v ∈ Z such that um + vn = h. (Bézout’s Lemma)
and l has the following properties:
(d) m|l and n|l;
(e) if m|x and n|x then l|x.

Proof. Properties of h:
(a) As m = 1m + 0n ∈ m, n = h then h|m. Similarly h|n.
(c) This follows from Example 95.
(b) Say x|m and x|n. Then by Bézout’s Lemma x|um + vn and so x|h.
Properties of l:
(d) As l ∈ m then m|l. Similarly n|l.
(e) If m|x then x ∈ m . Likewise x ∈ n . So x ∈ m ∩ n = l and l|x.

Definition 105 We define h, as defined in the previous Proposition, to be the highest com-
mon factor or hcf of m and n.
We define l as defined in the previous Proposition, to be the least common multiple or
lcm of m and n.

Theorem 106 (Chinese Remainder Theorem) Let m and n be coprime natural numbers.
Then Cmn is isomorphic to Cm × Cn .
Specifically if g is a generator of Cm and h is a generator of Cn then (g, h)generates Cm ×Cn .

Proof. Certainly
(g, h)mn = ((g m )n , (hn )m ) = (en , em ) = (e, e)
so that the order of (g, h) divides mn. But on the other hand g k = e if and only if m|k and
hk = e if and only if n|k. So
(g, h)k = g k , hk = (e, e)
if and only if m|k and n|k. As m, n are coprime then, by Bezout’s Lemma, there exist u, v such
that um + vn = 1. As n|k then mn|mk and as m|k then mn|nk. So

mn | (umk + vnk) = k.

Hence the order of (g, h) is mn which equals |Cm × Cn | .

MORE ON SUBGROUPS & CYCLIC GROUPS 33

34 MORE ON SUBGROUPS & CYCLIC GROUPS
4. REPRISE ON EQUIVALENCE RELATIONS

Firstly we recall:

Definition 107 A (binary) relation ∼ on a set S is a subset of S × S.

Then, for a, b, ∈ S, we write a ∼ b if and only if (a, b) ∈ ∼ .

We might just as easily view the relation as a function ∼ : S × S → {T, F }, that is a function
with two inputs from S and output True (T ) or False (F ). The "set" ∼ would then be ∼−1 (T ).

Example 108 (i) With S = Z, we would have " (3, 4) = T " or "(3, 4) ∈ " as rather
unnatural ways of simply saying "3 4".
(ii) If S = {1, 2, 3} then < is the set {(1, 2), (1, 3), (2, 3)}.

Definition 109 We say that a relation ∼ on a set S is an equivalence relation if it is

(i) reflexive — that is a ∼ a for all a ∈ S;
(ii) symmetric — that is, whenever a ∼ b then b ∼ a;
(iii) transitive — that is, whenever a ∼ b and b ∼ c then a ∼ c.

Example 110 The following are all examples of equivalence relations:

(i) S = C with z ∼ w iff |z| = |w|;
(ii) S = GL(n, R) with A ∼ B iff there exists P ∈ GL(n, R) such that A = P −1 AP ;
(iii) S = {polygons in R2 } and ∼ is congruence;
(iv) S = P(X) and A ∼ B if |A| = |B|;
(v) S is a group and x ∼ y if x = y or x = y −1 ;
(vi) S = C 1 (R) with f (x) ∼ g(x) if f ′ (x) = g ′ (x).

Example 111 The following relations aren’t equivalence relations:

(i) S = Z with m ∼ n iff m < n as ∼ isn’t reflexive or symmetric;
(ii) S = P (X) with A ∼ B iff A ⊆ B as ∼ isn’t symmetric;
(iii) S = R [x] with p (x) ∼ q (x) iff p (a) = q (a) for some a ∈ R as ∼ isn’t transitive.

Proposition 112 Let S = Z and n 2 is an integer. If we set a ∼ b if a − b is a multiple of

n then ∼ is an equivalence relation.

Proof. (a) For any a ∈ Z we have a ∼ a as 0 is a multiple of n.

(b) If a ∼ b then a − b = kn for some integer k. Then b − a = −kn and hence b ∼ a.
(c) If a ∼ b and b ∼ c then a − b = kn and b − c = ln for integers k, l. But then

a − c = (a − b) + (b − c) = (k + l)n

and hence a ∼ c.

REPRISE ON EQUIVALENCE RELATIONS 35

Definition 113 Let G be a group and g, h ∈ G. Then g and h are said to be conjugate in G
if there exists k ∈ G such that g = k −1 hk. (Compare with Definition 71.)

Proposition 114 Conjugacy is an equivalence relation.

Proof. Let G be a group and write g ∼ h if there exists k such that g = k −1 hk. Then:
(a) ∼ is reflexive as g = e−1 ge for all g.
−1
(b) If g = k −1 hk then h = kgk −1 = (k −1 ) gk −1 and hence ∼ is symmetric.
(c) If g1 ∼ g2 and g2 ∼ g3 then there exist k1 and k2 such that

g1 = k1−1 g2 k1 and g2 = k2−1 g3 k2 .

Hence g1 = k1−1 k2−1 g3 k2 k1 = (k2 k1 )−1 g3 (k2 k1 ).

Definition 115 Given an equivalence relation ∼ on a set S with a ∈ S, then the equivalence
class of a, written ā or [a] , is the set

ā = {x ∈ S : x ∼ a} .

Example 116 Given the equivalence relation in Proposition 112 there are n equivalence classes
namely 0̄, 1̄, 2̄, . . . , n − 1. This follows from the division algorithm in Z. We see that

0̄ = nZ; 1̄ = 1 + nZ; ... n − 1 = (n − 1) + nZ = −1 + nZ.

Example 117 The conjugacy class of σ in Sn are those permutations of the same cycle type.

Example 118 The conjugacy classes of D8 are

{e} , r, r3 , r2 , s, r2 s , rs, r3 s .

Diagrammatically it is a little clearer as to what is going on

r2
r3 s

r
r2 s

r3 r2 s rs

Depending on their viewpoints, two observers might confuse reflection in the horizontal with it
in the vertical, but will be certain that the square wasn’t reflected in a diagonal; likewise they
might conflate rotation by a right angle anticlockwise with the same in a clockwise fashion.
For D10 the conjugacy classes are

{e} , r, r4 , r2 , r 3 , s, rs, r2 s, r3 s, r4 s .

36 REPRISE ON EQUIVALENCE RELATIONS

Again, diagrammatically, it is a little clearer as to what is going on
axis of r4 s axis of rs

3 2
r r

axis of r2 s axis of r3 s
r4 r

axis of s

The general cases are investigated in Exercise Sheet 5, Question 3.

Definition 119 Let S be a set and Λ be an indexing set. We say that a collection of subsets
Aλ of S (where λ ∈ Λ) is a partition of S if
(i) Aλ = ∅ for each λ ∈ Λ;
(ii) Aλ = S;
λ∈Λ
(iii) if λ = µ then Aλ ∩ Aµ = ∅, or equivalently: if Aλ ∩ Aµ = ∅ then λ = µ.

Notation 120 Given a partition P of S and a ∈ S, we will write Pa for the unique set in P
such that a ∈ Pa .

Theorem 121 Let ∼ be an equivalence relation on a set S. Then the ∼-equivalence classes
partition S.

Proof. Firstly, a ∈ ā for any a ∈ S by reflexivity; this shows that equivalence classes are
non-empty and also that their union is S. Secondly, we need to show that distinct equivalence
classes are disjoint. So suppose that c ∈ ā ∩ b̄ for a, b, c ∈ S. We need to show that ā = b̄. As
c ∈ ā then c ∼ a and likewise c ∼ b. By symmetry and transitivity it follows that a ∼ b. So if
x ∈ ā we have x ∼ a ∼ b and hence, by transitivity, x ∼ b. We have shown that ā ⊆ b̄. If we
swap the roles of a and b in the above argument then b̄ ⊆ ā and the result follows.

Theorem 122 Let S be a set.

(a) Given an equivalence relation ∼ on S then the equivalence classes of ∼ form a partition
P (∼) of S (where P (∼)a = ā for each a ∈ S).
(b) Given a partition P of S then the relation ∼P on S defined by

a ∼P b if and only if b ∈ Pa

is an equivalence relation on S.
(c) As given above, (a) and (b) are inverses of one another; that is

P (∼P ) = P and ∼P (∼) = ∼ .

In particular, there are as many equivalence relations on a set S as there are partitions of the
set S.

Proof. (a) was proven in the previous theorem. To prove (b), suppose that P is a partition of
S.

REPRISE ON EQUIVALENCE RELATIONS 37

 • Let a ∈ S. Then a ∈ Pa by definition and so a ∼P a.
• If a ∼P b then b ∈ Pa and b ∈ Pb by definition. So b ∈ Pa ∩ Pb = ∅ and hence Pa = Pb .
Thus a ∈ Pb and b ∼P a.
• If a ∼P b and b ∼P c then b ∈ Pa and c ∈ Pb . As b ∈ Pa ∩ Pb = ∅ then Pa = Pb and so
c ∈ Pa and a ∼P c.

(c) Let P be a partition of S.

A ∈ P (∼P ) ⇐⇒ there is a ∈ A such that A is the ∼P -equivalence class of a
⇐⇒ there is a ∈ A such that A = Pa
⇐⇒ A ∈ P.
Likewise
a ∼P (∼) b ⇐⇒ b ∈ (P (∼))a ⇐⇒ a ∈ b̄ ⇐⇒ a ∼ b.

Example 123 There are 52 equivalence relations on a set with 5 elements.

Solution. Let X = {1, 2, 3, 4, 5}. As the only ways to partition the integer 5 is
5= 4+1= 3+2= 3+1+1= 2+2+1 =2+1+1+1 =1+1+1+1+1
and for each such possibility there correspond the following partitions of X
Partition of 5 Partitions of X
5 1
5
4+1 1
=5
5
3+2 2
= 10
5
3+1+1 3
= 10
1 5 3
2+2+1 2! 2 2
= 15
5
2+1+1+1 2
= 10
1+1+1+1+1 1

Example 124 How many partitions are there of a set with 22 elements into 4 subsets of size
3 and 2 subsets of size 5?

Solution. The answer is 1254751898400 counted either of the following ways:

ways of filling the six sets

22 19 16 13 10 5 ways of placing the 22 elements

3 3 3 3 5 5 22!
= 4 2 .
4!2! (3!) (5!) × 4!2!
shuffling same-size subsets shuffling same-size subsets
shuffling within subsets

38 REPRISE ON EQUIVALENCE RELATIONS

4.1 Modular Arithmetic

Consider the odd and even integers. The product of two odd integers is an odd integer, no
matter what odd integers we have in mind. Likewise we can see, for example, that

Even × Odd = Even, Odd + Odd = Even,

again irrespective of the even and odd numbers we have in mind. If we fill out the addition
and multiplication tables for {Even, Odd} then we obtain

+ Even Odd × Even Odd

Even Even Odd , Even Even Even .
Odd Odd Even Odd Even Odd

You may notice that {Even, Odd} under + makes an abelian group with Even being the additive
identity.
More properly the above tables describe the arithmetic of the integers "modulo 2" or more
simply "mod 2". Modular arithmetic is the study of remainders. If we divide an integer by
2 then there are two possible remainders 0 (when the integer is even) and 1 (when the integer
is odd). We could instead rewrite the above addition and multiplication with 0 replacing Even
and 1 replacing Odd. The tables would then look like:

+ 0 1 × 0 1
0 0 1 , 0 0 0 .
1 1 0 1 0 1

Most of those calculations look fairly natural with the exception of 1 + 1 = 0, but recall the
equation is really conveying that an odd number added to an odd number makes an even
number. From the point of view of remainders, adding the two remainders of 1 makes a whole
new factor of 2; these two 1s add to clock back to 0.
In fact, modular arithmetic is sometimes also referred to as clockwork arithmetic and
another everyday example of modular arithmetic is the 12-hour clock. It would not be at all
surprising for me to say that 5 hours after 9 o’clock comes 2 o’clock or that 7 hours before 1
o’clock was 6 o’clock or that 7 three-hour shifts that started at 2 o’clock will end at 11 o’clock.
In mod 12 arithmetic we would write these calculations as

5 + 9 = 2, 1 − 7 = 6, 2 + 7 × 3 = 11.

These facts are true irrespective of what day of the week we are discussing or whether 5
represents 5am or 5pm.(The only significant difference between mod 12 arithmetic and the
12-hour clock is that we write 0, instead of 12, for noon and midnight.)
More generally, we can use the division algorithm to describe the possible remainders when
we divide by any integer n 2.

MODULAR ARITHMETIC 39
Definition 125 If we are doing arithmetic mod n, (where n 2) then, by the division algo-
rithm, there are n possible remainders, namely

0, 1, 2, 3, . . . , n − 1.

We define here rules for how to add, subtract and multiply these n remainders in mod n arith-
metic. Take a, b ∈ {0, 1, 2, . . . , n − 1}. It may well be the case that a + b, a − b or ab aren’t on
this list, but the remainders of this sum, difference and product will be. We may define mod n
addition, subtraction and multiplication by:

a+b = remainder when a + b is divided by n;

a−b = remainder when a − b is divided by n;
ab = remainder when ab is divided by n.

Notation 126 We write Zn for the set of remainders {0, 1, 2, . . . , n − 1} under the operations
of mod n arithmetic. Also we will write mod n besides a sum, difference or product to denote
that we are doing these operations in the context of mod n arithmetic.

Example 127 In mod 7 arithmetic we have

3+6 = 2 mod 7 as 3 + 6 = 9 and 9 = 1 × 7 + 2;

3−5 = 5 mod 7 as 3 − 5 = −2 and − 2 = (−1) × 7 + 5;
3×5 = 1 mod 7 as 3 × 5 = 15 and 15 = 2 × 7 + 1.

We can more concisely write down all the rules of mod 7 arithmetic with addition and multi-
plication tables:

+ 0 1 2 3 4 5 6 × 0 1 2 3 4 5 6
0 0 1 2 3 4 5 6 0 0 0 0 0 0 0 0
1 1 2 3 4 5 6 0 1 0 1 2 3 4 5 6
2 2 3 4 5 6 0 1 2 0 2 4 6 1 3 5
3 3 4 5 6 0 1 2 3 0 3 6 2 5 1 4
4 4 5 6 0 1 2 3 4 0 4 1 5 2 6 3
5 5 6 0 1 2 3 4 5 0 5 3 1 6 4 2
6 6 0 1 2 3 4 5 6 0 6 5 4 3 2 1

Definition 125 has the advantage of being unambiguous (i.e. the operations +, −, × clearly
deliver well-defined answers) but it also looks a little unnatural. For example, is it clear that
the distributive law still applies? Alternatively, we can take a more formal view of what the
arithmetic of Zn is. In Proposition 112, we met the equivalence relation on Z given by a ∼ b if
a − b is a multiple of n. We can see now that this is the same as saying

a ∼ b if and only if a = b mod n. (4.1)

We saw in Example 116 that there are then n equivalence classes 0̄, 1̄, 2̄, . . . , n − 1. An alter-
native, more formal but also more natural, definition of the arithmetic of Zn is then:

40 REPRISE ON EQUIVALENCE RELATIONS

Definition 128 Let Zn denote the equivalence classes 0̄, 1̄, 2̄, . . . , n − 1 of Z under the equiva-
lence relation (4.1). We define the operations + and × on Zn by

ā + b̄ = a + b, ā × b̄ = a × b.

Proposition 129 The operations + and × are well-defined on Zn .

Proof. How might + and × not be well-defined? Well, because the same equivalence class has
many different representatives (e.g. 1̄ = 7̄ in Z6 ) it’s feasible that we might have ā = ᾱ and
b̄ = β̄ yet a + b = α + β. Adding the same two elements shouldn’t be able to yield two different
sums. So suppose that ā = ᾱ and b̄ = β̄, then

a − α = kn and b − β = ln

for k, l ∈ Z. But then

(a + b) − (α + β) = (a − α) + (b − β) = (k + l)n

and
ab − αβ = (α + kn)(β + ln) − αβ = (kβ + lα + kln)n
and hence a + b = α + β and ab = αβ are both true so that + and × are well-defined.

Proposition 130 (a) (Zn , +) is an abelian group isomorphic to Cn .

(b) Further × is associative, commutative and distributes over +.

Proof. That (Zn , +) is an abelian group and the properties of × mentioned in (b) are all
inherited from the same properties in Z. For example, to see that the distributive law still
holds, we simply have to note for ā, b̄, c̄ ∈ Zn that

ā b̄ + c̄ = ā(b + c) [as + is well-defined in Zn ]

= a(b + c) [as × is well-defined in Zn ]
= ab + ac [by the distributive law in Z]
= ab + ac [as + is well-defined in Zn ]
= ā b̄ + ā c̄ [as × is well-defined in Zn ].

To see that (Zn , +) is indeed cyclic we need only note that 1̄ has (additive) order n.

We now note, for certain values of n, that modular arithmetic can have some unfortunate
algebraic aspects such as

3 × 5 = 0 mod 15, 4 × 3 = 0 mod 6.

It follows that one cannot divide by 3 or 5 in Z15 nor divide by 3 or 4 in Z6 . More generally
we note:

MODULAR ARITHMETIC 41
Proposition 131 Let x̄ ∈ Zn with x = 0.
(a) x̄ has a multiplicative inverse if and only if hcf(x, n) = 1. Hence if n is prime, then Zn
is in fact a field.
(b) Those x̄ with a multiplicative inverse (the so-called units) form a group Z∗n under
multiplication.

Proof. This is left as Exercise Sheet 4, Question 2.

Example 132 List the units in Z12 . Identify the group Z∗12 .

Solution. As 12 = 22 × 3 then the units of Z12 are 1, 5, 7, 11. Note that the group table is

∗ 1 5 7 11
1 1 5 7 11
5 5 1 11 7
7 7 11 1 5
11 11 7 5 1

showing that Z∗12 is isomorphic to C2 × C2 .

42 REPRISE ON EQUIVALENCE RELATIONS

5. ORDER. LAGRANGE’S THEOREM

Recall that we also defined the order o(g) of a group element:

Definition 133 Let G be a group and g ∈ G. If there is a positive integer k such that g k = e,
then the order o(g) of g ∈ G is defined as

o(g) = min {m > 0 : g m = e}.

Otherwise we say that the order of g is infinite.

Proposition 134 If G is finite, then o(g) is finite for each g ∈ G.

Proof. Consider the list

g, g 2 , g 3 , g 4 , . . .
As G is finite, then this list must have repeats. Hence there are integers i > j such that g i = g j .
So g i−j = e showing that {m > 0 : g m = e} is non-empty and so has a minimal element.

Proposition 135 If g ∈ G and o(g) is finite, then g n = e if and only if o(g)|n.

Proof. If n = ko(g) then

k
g n = g o(g) = ek = e.
Conversely, if g n = e then, by the division algorithm, there are integers q, r such that n =
qo(g) + r where 0 r < o(g). Then

g r = g n−qo(g) = g n (g o(g) )−q = e.

By the minimality of o(g) then r = 0 and so n = qo(g).

Proposition 136 If φ : G → H is an isomorphism and g ∈ G then o(φ(g)) = o(g).

Proof. We have
(φ(g))k = eH ⇐⇒ φ(g k ) = eH ⇐⇒ g k = eG
as φ is injective.

Example 137 In D8 we have

o(e) = 1, o(r2 ) = o(s) = o(rs) = o(r2 s) = o(r3 s) = 2, o(r) = o(r3 ) = 4.

Proposition 138 Let x, n be integers with n 2. Then the order o(x̄) of x̄ ∈ Zn is

n
o(x̄) = .
hcf(x, n)

ORDER. LAGRANGE’S THEOREM 43

Proof. Left to Exercise Sheet 4, Question 1.

Corollary 139 x̄ ∈ Zn is a generator if and only if hcf(x, n) = 1.

Definition 140 Let H be a subgroup of G.

Then the left cosets of H (or left H-cosets) are the sets

gH = {gh : h ∈ H}.

The right cosets of H (or right H-cosets) are the sets

Hg = {hg : h ∈ H}.

Notation 141 We write G/H for the set of (left) cosets of H in G. The cardinality of G/H
is called the index of H in G.

Remark 142 (i) Note that different elements g1 ,g2 ∈ G can represent the same (left) coset —
i.e. we can have g1 H = g2 H yet g1 = g2 .
(ii) In general, we will have gH = Hg. Obviously we will have gH = Hg if G is abelian,
and in other cases as well.

Example 143 Let G = S3 and H = {e, (12)} . Then

eH = (12) H = {e, (12)} ; He = H (12) = {e, (12)} ;

(13) H = (132) H = {(13) , (132)} ; H (13) = H (123) = {(13) , (123)} ;
(23) H = (123) H = {(23) , (123)} ; H (23) = H (132) = {(23) , (132)} .

Note here that Hg = gH in general.

Example 144 Let G = Sn and H = An . Then

σAn = An σ = An when σ is even; σAn = An σ = Sn \An when σ is odd.

Note that σAn = An σ for all σ ∈ Sn , even though Sn is not (in general) abelian.

Example 145 Let G = C∗ and H = S 1 . Then, for w ∈ C∗ , we have

wS 1 = {z ∈ C∗ : |z| = |w|} .

Example 146 Let G = Z and H = nZ. Then the (left and right) coset of r ∈ Z is r + nZ. So
there are n cosets

0 + nZ, 1 + nZ, 2 + nZ, ... (n − 1) + nZ = −1 + nZ.

So we can naturally identity Zn with Z/nZ (if only as sets for the moment).

Lemma 147 (Coset Equality Lemma) Let H G and g, k ∈ Ġ. Then

gH = kH ⇐⇒ k −1 g ∈ H.

For right cosets, Hg = Hk ⇐⇒ kg −1 ∈ H.

44 ORDER. LAGRANGE’S THEOREM

Proof. Suppose that gH = kH. Then g = ge ∈ kH and so there exists h ∈ H such that
g = kh. Hence k −1 g = h ∈ H.

Conversely suppose that k −1 g = h ∈ H. Then gH = khH ⊆ kH and kH = g (g −1 k) H =

−1
gh H ⊆ gH.

Remark 148 The relation on G given by g ∼ k ⇐⇒ k −1 g ∈ H is an equivalence relation on

G with the equivalence classes being the left cosets of H. This essentially comprises part of the
following proof of Lagrange’s Theorem when we prove that the (left) cosets partition G.

Theorem 149 (Lagrange’s Theorem) (First instances of theorem due to Lagrange in 1771.)
Let G be a finite group and H a subgroup of G. Then |H| divides |G| .

Remark 150 There are two steps to this proof. We shall prove:
(a) The (left or right) cosets of H partition G.
(b) Each (left or right) coset of H is equinumerous with H.
Both (a) and (b) in fact hold for infinite groups.

Proof. Let G be a (not necessarily finite) group G and H a subgroup of G.

(a) For any g ∈ G, note g = ge ∈ gH, so that the union of the (left) cosets is G. Now suppose
that two cosets aren’t disjoint; we’ll show that they must be equal. Say k ∈ g1 H ∩ g2 H. Then
there are h1 , h2 ∈ H such that k = g1 h1 = g2 h2 . Then g2−1 g1 = h2 h−1
1 ∈ H and g1 H = g2 H by
the Coset Equality Lemma.

(b) For any g ∈ G then h → gh is a bijection between H and gH. This map is clearly onto
and also 1-1, for if gh1 = gh2 then we see h1 = h2 by applying g −1 . Hence |gH| = |H| .

Finally, if G is finite, then we have

|G| = |G/H| × |H|

and hence |H| divides |G| .

Remark 151 Lagrange’s Theorem states that the order of a subgroup is a factor of the order
of the group. The converse does not hold — that is, if G is a finite group and k is a factor of
|G| then there need not be a subgroup H of G such that |H| = k. For example, |A4 | = 12 yet A4
has no subgroup of order 6. (See Examples 86 and 209.) The converse of Lagrange’s Theorem
is true for cyclic groups though: for if k divides n then n/k has order k in Zn .

Example 152 Find all the subgroups of (i) Z31 ; (ii) D10 ; (iii) Z5 × Z5 .

Solution. (i) As 31 is prime then a subgroup must have order 1 or 31. Hence the only
subgroups are {0̄} and Z31 itself.
(ii) The subgroups of D10 can have order 1, 2, 5 or 10. So aside from {e} and D10 we can
have order 2 subgroups of the form {e, reflection} and the only order 5 subgroup consists of
the five rotations.

ORDER. LAGRANGE’S THEOREM 45

 (iii) |Z5 × Z5 | = 25. So the subgroups can have order 1, 5 or 25. Every element in Z5 × Z5
apart from (0̄, 0̄) has order 5. The subgroups of order 5 consist of the identity (0̄, 0̄) and four
elements of order 5 each of which generate that subgroup. So there are (25 − 1)/(5 − 1) = 6
such subgroups. Specifically these are

{(0̄, 0̄), (1̄, 0̄) , (2̄, 0̄), (3̄, 0̄), (4̄, 0̄)} ; {(0̄, 0̄), (0̄, 1̄), (0̄, 2̄), (0̄, 3̄), (0̄, 4̄)} ;
{(0̄, 0̄), (1̄, 1̄), (2̄, 2̄), (3̄, 3̄), (4̄, 4̄)} ; {(0̄, 0̄), (1̄, 2̄) , (2̄, 4̄), (3̄, 1̄), (4̄, 3̄)} ;
{(0̄, 0̄), (2̄, 1̄), (4̄, 2̄), (1̄, 3̄), (3̄, 4̄)} ; {(0̄, 0̄), (1̄, 4̄), (2̄, 3̄), (3̄, 2̄), (4̄, 1̄)} .

The only other subgroups are then {(0̄, 0̄)} and Z5 × Z5 .

Corollary 153 Let G be a finite group and g ∈ G. Then o(g) divides |G|.

Proof. g = {e, g, g 2 , . . . , g o(g)−1 } is a subgroup of G with order o(g).

Remark 154 This Corollary has no converse: for example, S3 has no element of order 6.
However we shall later prove Cauchy’s Theorem which states that if p is a prime factor of |G|
then there is a group element with order p. We shall prove this for p = 2 (see Corollary 162
below).

Corollary 155 Let G be a finite group with |G| = p, a prime. Then G is cyclic.

Proof. Let g ∈ G with g = e. Then o(g) = 1 and yet o(g) divides p, so o(g) = p. Hence
| g | = p. That is g = G and G is cyclic.

Corollary 156 Let G be a finite group and g ∈ G. Then g |G| = e.

Proof. |G| is a multiple of o(g) and g o(g) = e.

Theorem 157 (Fermat’s Little Theorem, 1640) Let p be a prime and a ∈ Z such that p
does not divide a. Then ap−1 = 1 mod p.

Proof. This is just Corollary 156 with G = Z∗p as Z∗p = p − 1.

Theorem 158 (Euler’s Theorem, 1736) Let n 2 and let a ∈ Z be coprime with n. Then

aφ(n) = 1 mod n

where φ(n) = |{k : 0 < k < n, hcf(k, n) = 1}| .

Proof. This is just Corollary 156 with G = Z∗n as φ(n) = |Z∗n | .

Remark 159 (Off-syllabus) The "phi function" or "totient function" φ (n) was introduced by
Euler in 1760. It is an important number-theoretic function with the following properties.
(i) φ(p) = p − 1 for a prime p.
(ii) φ(pk ) = pk − pk−1 = pk−1 (p − 1).
(iii) φ(mn) = φ(m)φ(n) if m and n are coprime.

46 ORDER. LAGRANGE’S THEOREM

Lemma 160 Let G be a group. Then the relation ∼ on G defined by

x ∼ y ⇐⇒ x = y or x = y −1

is an equivalence relation. The equivalence classes are generally of the form x̄ = {x, x−1 } unless
x is self-inverse in which case x̄ = {x} .

Proof. Left as an exercise.

Corollary 161 (Wilson’s Theorem) If p is a prime then

(p − 1)! = −1 mod p.

Proof. If p = 2 then this just says 1 = −1 mod 2 which is true. So assume p 3. Consider
the self-inverse elements in Z∗p . We see

x̄ = x̄−1 ⇐⇒ x̄2 = 1 ⇐⇒ (x̄ − 1̄) (x̄ + 1̄) = 0̄ ⇐⇒ x̄ = 1̄ or x̄ = −1̄

as Zp is a field. So the only singleton equivalence classes of ∼ (the equivalence relation defined
in Lemma 160) are {1̄} and {−1̄} with all others being of the form {x̄, x̄−1 }. As the equivalence
classes partition Z∗p then

(p − 1)! = k̄ = k̄ = 1̄ × (−1̄) × k̄ = −1̄

k̄∈Z∗p equivalence each doubleton
classes equivalence equivalence
class classes

as the contribution to the product from each doubleton equivalence class is x̄ × x̄−1 = 1̄.

Corollary 162 Let G be a group with even order. Then G has an element of order 2.

Proof. Consider the equivalence relation on G defined in Lemma 160. If there are m doubleton
equivalence classes and n singleton equivalence classes, then we have

2m + n = |G|

as the equivalence classes partition |G|. As |G| is even then n is even but we also know n 1 as
e is self-inverse. So, in fact, n 2 and there is a non-identity element x which satisfies x = x−1
or equivalently x2 = e so that o (x) = 2.

Theorem 163 Let G be a finite group with |G| = 2p where p 3 is prime. Then G is
isomorphic to C2p or D2p .

Proof. Assume that G is not cyclic. The possible orders of elements in G are 1 (the identity
e) or 2 or p. As |G| = 2p is even then there is an element x ∈ G of order 2. (Corollary 162).
Further if g 2 = e for all g ∈ G then G ∼ n
= (Z2 ) for some n (Exercise Sheet 4, Question 5),
which is not possible here and hence there is an element y ∈ G of order p. As x has order 2
and y, y 2 , . . . , y p−1 have order p then x ∈
/ y . Hence G = y ∪ x y or more expansively

G = e, y, y 2 , . . . , y p−1 , x, xy, xy 2 , . . . , xy p−1 .

ORDER. LAGRANGE’S THEOREM 47

Now the product yx is somewhere amongst G. If yx = y i we arrive at a similar contradiction
to before. So yx = xy j for some 1 j < p. Then

(yx)2 = yxyx = (yx) xy j = y j+1 ; (yx)3 = xy j (yx)2 = xy 2j+1 ;

until more generally we find that (yx)2k = y k(j+1) and that (yx)2k+1 = xy kj+k+j . So yx has an
even order and o (yx) = 2. In particular it follows that j = p − 1. Hence

G = x, y : x2 = y p = e, yx = xy p−1

which is a presentation for D2p . We can think of x as reflection in a given axis and y as clockwise
rotation through 2π/p.

Remark 164 (Off Syllabus) Presentations. Recall that the dihedral group D2n can be defined
as
D2n = r, s : rn = e = s2 , sr = r−1 s . (5.1)
Equation (5.1) is an example of a presentation for D2n . We can think of r as a rotation
and s as a reflection if we want to make real the elements r and s, but there’s no great need
as the presentation contains everything necessary to describe the algebra of D2n or any group
isomorphic to D2n . A presentation of a group describes some generators of the group (here
r and s) and the (non-trivial) rules or relations that govern the algebra in the group. Con-
tained in the relations is enough information to show that the group contains 2n elements
e, r, , . . . , rn−1 , s, rs, . . . , rn−1 s and determine products between them. Any other string or word
in the generators can be shown to be one of these 2n elements by means of the relations. For
example, we can see that

sr3 sr2 s = sr3 sr2 s = r−3 (ss) r2 s = r−1 s = rn−1 s.

Other group presentations include

Z∼
= g , Cn ∼
= g : gn = e , Z2 ∼
= g, h : gh = hg .

There are, of course, many different ways to present the same group. Note a = s and b = rs
generate D6 . We can write the other elements as

r2 = ab, r = ba, r2 s = aba

and see that

D6 = a, b : a2 = e = b2 , bab = aba .
We need to check we have enough relations. Using the relations a2 = e = b2 we see that we need
only consider those strings (or words) which alternately go a then b. And using the relation
bab = aba we can contract substrings of bab from longer words via

a (bab) = a (aba) = ba, (bab) a = (aba) a = ab.

The only strings that can’t be contracted further in this way are e, a, b, ab, ba, aba.

48 ORDER. LAGRANGE’S THEOREM

6. HOMOMORPHISMS AND ISOMORPHISMS

Let G and H be groups. Recall that an isomorphism φ : G → H is a bijection such that

φ(g1 ∗G g2 ) = φ(g1 ) ∗H φ(g2 ) for all g1 , g2 ∈ G.
We say that two groups, G and H, are isomorphic, and write G ∼ = H, if there is an isomorphism
between the groups. Two isomorphic groups should usually be considered as manifestations of
the same group. We proved in the first half of the course that:
Theorem 165 Let G be a group with |G| = p, a prime. Then G ∼
= Cp .
Theorem 166 Let G be a group with |G| = 2p where p 3 is prime. Then G ∼
= C2p or
G∼
= D2p .
These results are sufficient to demonstrate the following:
Theorem 167 Up to isomorphism, the groups of order 7 are:
• Order 2: C2
• Order 3: C3
• Order 4: C4 or C2 × C2 ∼
= V4
• Order 5: C5
• Order 6: C6 or S3 ∼
= D6
• Order 7: C7
The general situation (re how many groups there are of a give order n) is very complicated
and depends largely on the factors of n. Even the n = 8 case (stated below) is more complicated
and beyond the scope of this first course.
Theorem 168 Up to isomorphism, the groups of order 8 are
C8 , C2 × C4 , C2 × C2 × C2 , D8 , Q8 .
Remark 169 We have yet to meet the fifth group Q8 which is usually introduced via the quater-
nions. A quaternion might be thought of as a four-dimensional version of a complex number.
A quaternion is a number of the form
q = a + bi + cj + dk (a, b, c, d ∈ R)
with quaternions adding as one might expect and multiplying distributively and associatively
according to the rules
i2 = j2 = k2 = ijk = −1.
An example of the group Q8 is then
Q8 = {±1, ± i, ±j, ±k} .

HOMOMORPHISMS AND ISOMORPHISMS 49

 The equivalent of a linear map in group theory is the homomorphism.

Definition 170 Let G and H be groups. A homomorphism φ : G → H is a map such that

φ(g1 ∗G g2 ) = φ(g1 ) ∗H φ(g2 ) for all g1 , g2 ∈ G.

Hence an isomorphism between G and H is simply a bijective homomorphism.

We also have:

Definition 171 An automorphism of a group G is an isomorphism from G to G. The

automorphisms of G form a group Aut(G) under composition. (Exercise Sheet 5, Question 5).
An endomorphism of G is a homomorphism from G to G.
(Rarely used) A monomorphism is an injective homomorphism and an epimorphism is
a surjective homomorphism.

Homomorphisms in group theory have many properties akin to linear maps in linear algebra.

Proposition 172 Let φ : G → H be a homomorphism between groups and let g ∈ G, n ∈ Z.

Then
(i) φ(eG ) = eH . (ii) φ(g −1 ) = (φ(g))−1 . (iii) φ(g n ) = (φ(g))n .

Proof. (i) We have φ(eG ) = φ(eG ∗ eG ) = φ(eG )φ(eG ) and applying φ(eG )−1 to both sides (i)
follows. For (ii) note
φ(g) φ(g −1 ) = φ(gg −1 ) = φ(eG ) = eH
demonstrating (ii). For (iii) note more generally that we can show φ(g n ) = (φ(g))n for n > 0
by induction and then for n = −k < 0 we have

φ(g n ) = φ((g −1 )k ) = (φ(g −1 ))k = (φ(g)−1 )k = φ(g)n .

Corollary 173 Let φ : G → H be a homomorphism between groups and let g ∈ G. Then

o(φ(g)) divides o(g).

Proof. Note
φ(g)o(g) = φ(g o(g) ) = φ(eG ) = eH .
In a group, k n = e if and only if n is a multiple of o(k) (by Proposition 135).

Example 174 The map φ : Z → Zn given by φ(n) = n̄ is a homomorphism as

m + n = m + n.

Example 175 If H is a subgroup of G then inclusion ι : H → G given by ι(h) = h is a

homomorphism.

50 HOMOMORPHISMS AND ISOMORPHISMS

Example 176 For any groups G, H the map φ: G → H given by φ(g) = eH is a group
homomorphism.

Example 177 Let G1 and G2 be groups. Then the maps

π 1 : G1 × G2 → G1 , (g1 , g2 ) → g1 and π 2 : G1 × G2 → G2 , (g1 , g2 ) → g2

are homomorphisms.

Example 178 The map sgn : Sn → {±1} given by

1 if σ is even
sgn(σ) =
−1 if σ is odd

is a homomorphism as, for σ, τ ∈ Sn , we have sgn(στ ) = sgn(σ) sgn(τ ).

Example 179 The map det : GL(n, R) → R∗ is a homomorphism as, for n × n matrices A, B
we have det(AB) = det A det B.

Example 180 The map trace : Mn (R) → R is a homomorphism as, for n × n matrices A, B
we have trace (A + B) = trace A + trace B.

Example 181 The map log : (0, ∞) → R is a homomorphism as for x, y > 0 we have
log(xy) = log x + log y. In fact, being a bijection, this is an isomorphism.

Example 182 The map φ : R → S 1 given by φ(x) = eix is a surjective homomorphism as

ei(x+y) = eix eiy .

Example 183 The map φ : C∗ → R∗ given by φ(z) = |z| is a group homomorphism as |zw| =
|z| |w| for z, w = 0.

Example 184 The map φ : R∗ → R∗ given by φ(x) = x2 is a group homomorphism as (xy)2 =

x2 y 2 .

Remark 185 More generally φ : G → G given by φ(g) = g 2 is a homomorphism if and only if

G is abelian. (This is left as an exercise).

Proposition 186 Let a ∈ G, a group. Conjugation by a, i.e. the map θa : G → G given by

θ a (g) = a−1 ga is an isomorphism.

Proof. Firstly we note

θa (gh) = a−1 gha = (a−1 ga)(a−1 ha) = θa (g)θa (h).

Secondly an easy check shows that θa−1 = (θa )−1 and so θa is a bijection.

Corollary 187 Let G be a group and g, h ∈ G.

(i) If g and h are conjugate then o(g) = o(h).
(ii) If g and h are conjugate then g −1 and h−1 are conjugate.

HOMOMORPHISMS AND ISOMORPHISMS 51

Proposition 188 Homomorphisms φ: Z → Z are all of the form φ(x) = nx for some n ∈ Z.

Proof. Certainly, for any n ∈ Z, we have that φ(x) = nx is a homomorphism as

φ(x + y) = n(x + y) = nx + ny = φ(x) + φ(y).

Conversely, for any homomorphism φ : Z → Z, if we set n = φ(1) then by Proposition 172 (iii)
we have for x > 0

φ(x) = φ 1 + 1 + · · · + 1 = φ(1) + φ(1) + · · · + φ(1) = nx

x times x times

and then φ(−x) = −φ(x) = −nx = n(−x).

Remark 189 As a crucial aspect of the above proof we have just noted that, if G is a cyclic
group with generator g, then any homomorphism from G is entirely determined by the value of
φ(g) as
φ(g r ) = (φ(g))r for any r ∈ Z.
More generally if g1 , . . . , gk are generators of a group G then any homomorphism from G is
entirely determined by the values φ(g1 ), . . . , φ(gk ).
This result corresponds to the similar result in linear algebra: any linear map T : V → W
is determined by the values T takes on a basis of V (or more generally on a spanning set).

In a comparable way to linear maps, we can also define the kernel and image of a homomor-
phism. As one might expect from the study of linear maps, kernels and images in group theory
are subgroups — in fact, more than this, kernels turn out to be a special type of subgroup.

Definition 190 Let φ : G → H be a homomorphism between groups. Then:

(i) the kernel of φ, written ker φ, equals

ker φ = {g ∈ G : φ(g) = eH } ⊆ G.

(ii) the image of φ, written Im φ, equals

Im φ = {φ(g) : g ∈ G} ⊆ H.

Definition 191 Let G be a group and H a subgroup of G. Then H is said to be a normal

subgroup of G if
gH = Hg for all g ∈ G
or equivalently if
g −1 hg ∈ H for all g ∈ G, h ∈ H.
If H is a normal subgroup of G then we write H ⊳ G.

Proposition 192 Let φ : G → H be a homomorphism between two groups. Then ker φ ⊳ G.

52 HOMOMORPHISMS AND ISOMORPHISMS

Proof. Say k1 , k2 ∈ ker φ and g ∈ G. Then

φ(eG ) = eH ; φ(k1 k2 ) = φ(k1 )φ(k2 ) = eH eH = eH ; φ(k1−1 ) = φ(k1 )−1 = e−1

H = eH

showing that ker φ is a subgroup of G; further as

φ(g −1 k1 g) = φ(g −1 )φ(k1 )φ(g) = φ(g)−1 eH φ(g) = eH

then ker φ is a normal subgroup of G.

Proposition 193 Let φ : G → H be a homomorphism between two groups. Then Im φ H;

Proof. Note eH = φ(eG ) ∈ Im φ. Say h1 , h2 ∈ Im φ. Then there are gi such that φ(gi ) = hi .
Note
h1 h2 = φ(g1 )φ(g2 ) = φ(g1 g2 ) ∈ Im φ; h−1
1 = φ(g1 )
−1
= φ(g1−1 ) ∈ Im φ.

Example 194 The map φ : Z → Zn given by φ(n) = n̄ has kernel nZ and has image Zn .

Example 195 The map sgn : Sn → {±1} has kernel An and image {±1} .

Example 196 The map det : GL(n, R) → R∗ has kernel SL(n, R) and image R∗ .

Example 197 The map φ : R → S 1 given by φ(x) = eix has kernel 2πZ and image S 1 .φ

Example 198 The map φ : C∗ → R∗ given by φ(z) = |z| has kernel S 1 and image (0, ∞) .

We close this chapter with the following result — it is our first step to understanding homo-
morphisms via the First Isomorphism Theorem.

Proposition 199 A homomorphism is constant on a coset of ker φ and takes different values
on different cosets.

Proof.

φ(g1 ) = φ(g2 ) ⇐⇒ φ(g2−1 g1 ) = eH ⇐⇒ g2−1 g1 ∈ ker φ ⇐⇒ g1 ker φ = g2 ker φ.

Corollary 200 Let φ : G → H be a homomorphism between two groups. Then φ is 1-1 if and
only if ker φ = {eG } .

HOMOMORPHISMS AND ISOMORPHISMS 53

54 HOMOMORPHISMS AND ISOMORPHISMS
7. NORMAL SUBGROUPS. QUOTIENT GROUPS

Notation 201 If H G, then we write G/H for the set of (left) cosets of H in G.

Let H be a subgroup of G. Recall that we say that H is a normal subgroup of G if

gH = Hg for all g ∈ G

or equivalently if
g −1 hg ∈ H for all g ∈ G, h ∈ H.
If H is a normal subgroup of G we write this H ⊳ G.

Remark 202 This does NOT mean that gh = hg for all g ∈ G and h ∈ H or that G is
abelian. Although we can easily see that all subgroups of abelian groups are normal.

Remark 203 H being a normal subgroup of G is a property of how H is contained in G and

not solely a property of H. For example, if we consider H = (12) S3 then H is normal in
H but H is not normal in S3 .

Remark 204 In any group G, it is the case that {e} and G are normal subgroups of G. If
these are the only ones then G is said to be simple.

Remark 205 If asked to show that H is a normal subgroup of G then this means showing that
both H is a subgroup of G and checking that H is normal in G.

Remark 206 Note that a subgroup H G is normal in G if and only if H is a union of

conjugacy classes.

Proposition 207 Let H G. If |G/H| = 2 then H ⊳ G.

Proof. Note that eH = H = He. So one left coset of H is H and one right coset of H is H.
As there are only two (left or right) cosets, and as (left or right) cosets partition G then the
other left coset is H c (the complement of H in G) and the other right coset is also H c . Recall
that gH = H if and only if g ∈ H and likewise Hg = H if and only if g ∈ H. Hence

if g ∈ H then gH = H = Gg;
/ H then gH = H c = Hg.
if g ∈

Example 208 For n 2, this shows An is normal in Sn . Also SO(n) is normal in O(n).

Example 209 A4 has no subgroup of order 6. In particular, Lagrange’s Theorem has no

converse.

NORMAL SUBGROUPS. QUOTIENT GROUPS 55

Solution. From Example 86, we know that the conjugcay classes of A4 have sizes 1, 3, 4, 4. If
A4 had a subgroup of order 6 = 12 |A4 | then it would be normal and hence a union of conjugacy
classes. However no selection of 1, 3, 4, 4 adds up to 6 so there is no subgroup of that order.

Example 210 (a) The normal subgroups of S3 are {e} , A3 and S3 .

(b) The normal subgroups of S4 are {e} , V4 , A4 and S4 .

Solution. (a) Recall that a normal subgroup is a union of conjugacy classes. The identity and
the 2-cycles do not form a subgroup. The identity and the 3-cycles form A3 .
(b) The conjugacy classes e, (ab) , (abc) , (abcd) , (ab) (cd) have respective sizes 1, 6, 8, 6, 3.
As the order of any subgroup is a factor of 24, this can only be achieved with these numbers as

1, 1 + 3, 1 + 8 + 3, 1 + 6 + 8 + 6 + 3,

which correspond to the subgroups {e} , V4 , A4 and S4 .

Definition 211 Let G be a group. The centre of G, denoted Z(G), is the set

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

Proposition 212 Let G be a group. Then Z(G) ⊳ G.

Proof. (i) Certainly eh = h = he for all h ∈ G and so e ∈ Z(G).

(ii) If g1 , g2 ∈ Z(G) and h ∈ G then (g1 g2 )h = g1 hg2 = h(g1 g2 ) and so g1 g2 ∈ Z(G).
(iii) If g ∈ Z(G) and h ∈ G then gh = hg =⇒ hg −1 = g −1 h and so g −1 ∈ Z(G).
(iv) Finally if g ∈ Z(G) and h ∈ G, then h−1 gh = h−1 hg = g ∈ Z(G).

Remark 213 Note that g ∈ Z(G) if and only if the conjugacy class of g is {g}.

Example 214 (a) If n 3 then Z(Sn ) = {e} .

(b) The centre of GL(n, F ) is {λIn : λ = 0} .
(c) The centre of
D2n = r, s : rn = s2 = e, r−1 s = sr
is {e} when n is odd and e, rn/2 when n is even.

Solution. (a) and (b) will not be proved here. They are tractable but non-trivial exercises.
(c) follows from Exercise Sheet 5, Question 3 and Remark 213.

Proposition 215 Let H G.

(a) The binary operation ∗ on G/H given by

(g1 H) ∗ (g2 H) = (g1 g2 )H

is well-defined if and only if H ⊳ G.

(b) If H ⊳ G then (G/H, ∗) is a group.

Definition 216 If H ⊳ G then (G/H, ∗) is called the quotient group.

56 NORMAL SUBGROUPS. QUOTIENT GROUPS

Proof. (a) Suppose that H ⊳ G. Say g1 H = k1 H and g2 H = k2 H, so that k1−1 g1 ∈ H and
k2−1 g2 ∈ H. We wish to show (g1 g2 ) H = (k1 k2 ) H. Note

(g1 g2 ) H = (k1 k2 ) H ⇐⇒ (k1 k2 )−1 (g1 g2 ) ∈ H

⇐⇒ k2−1 k1−1 g1 g2 ∈ H
⇐⇒ k2−1 k1−1 g1 k2 k2−1 g2 ∈ H.

We have already noted that k2−1 g2 ∈ H and k1−1 g1 ∈ H; as H is normal then k2−1 k1−1 g1 k2 ∈ H
also.
Conversely, suppose that ∗ is well-defined. Let h ∈ H and g ∈ G. Then

g −1 hg H = g −1 H ∗ (hH) ∗ (gH)
= g −1 H ∗ (eH) ∗ (gH)
= g −1 eg H
= eH = H

and in particular g −1 hg ∈ H. That is, H is a normal subgroup of G.

(b) Suppose now that H ⊳ G. Then part (i) has shown that ∗ is a well-defined binary
operation on G/H. Further ∗ is clearly associative as G’s group operation is associative.
Finally, we note that for any g ∈ G,

(eH) ∗ (gH) = gH = (gH) ∗ (eH)

and hence eH = H is the identity element of G/H and for any g ∈ G

g −1 H ∗ (gH) = eH = g −1 H ∗ (gH)

so that
(gH)−1 = g −1 H in G/H.

Remark 217 Quotient Structures appear throughout pure mathematics. For a set S, then we
can take an equivalence relation ∼ on S and consider the set of equivalence classes S/ ∼ . This
is an example of a quotient structure with sets. Why might we ever want to do this? We do it
all the time and have been doing so for some time. If asked the question, "what is the area of
a triangle?" we respond "half base times height". We do not ask for the triangle’s coordinates
in R2 , which would be "too much" information and just clutter the issue, but instead we are
already thinking about the triangle’s equivalence class under congruence because the answer is
most naturally given in this format. So quotient structures can be a means of throwing away
extraneous information and focussing on the most relevant information. As an example: can
1, 000, 003 be written as the sum of two squares? At first glance this looks like it might be hard
and something we need to resolve with the help of a computer. Solving the equation

x2 + y 2 = 1, 000, 003

NORMAL SUBGROUPS. QUOTIENT GROUPS 57

looks fairly impenetrable but taking this equation in mod 4 and showing that x̄2 + ȳ 2 = 3̄ has
no solutions (rather easy as x̄2 = 0̄ or 1̄ for x̄ ∈ Z4 ) shows that the original equation had no
solutions. By moving to mod 4 we have thrown away a lot of information but have reduced the
problem in size and have crucially kept enough of the right information that conclusions can be
still made about the original problem.
Of course, with sets, we start off with rather structureless objects, so there aren’t really any
expectations of a structure on S/ ∼. This is no longer the case when we look to make quotient
groups and hope that a binary operation on the quotient might naturally arise from the binary
operation on the original group.
Remark 218 So if we have a group G and plan to make some elements equivalent, but still
hope to have some sensible algebra afterwards, how should we go about this? We have met some
of these ideas already with modular arithmetic. If we wish to introduce new rules, for example
as in Z3 when we wish to say that 3 ∼ 0, then surely there will need to be further consequences
of this rule if the resulting algebra is to be sensible. So we’d likewise expect:
3 + 3 ∼ 0 + 0 = 0 and − 3 ∼ −0 = 0.
So, reasoning more generally, those elements which are equivalent to the identity will need to
form a subgroup. More than that in fact (a point which is less obvious with the abelian example
of Z3 above) we will need that
if h ∼ e and g ∈ G then g −1 hg ∼ g −1 eg = e.
The equivalence class of e will need to be a normal subgroup. From what we’ve seen earlier
(Proposition 215) this is then sufficient for G/ ∼ to be a well-defined group.
In light of these comments it may be easier to think of G/H as G mod H. From this
viewpoint H is the equivalence class of e and we should think of gH as ḡ, the equivalence class
of g instead. The somewhat cumbersome equation
g1 H ∗ g2 H = (g1 g2 ) H becomes g1 ∗ g2 = g1 g2
which looks more natural and we see elements (or their equivalence classes) just multiply the
way they did before and do so in a well-defined fashion.
Proposition 219 Let G be a group and H a subset of G. Then H is a normal subgroup of G
if and only if it is the kernel of some homomorphism from G.
Proof. Suppose that H is the kernel of some homorphism φ : G → K. Then H = ker φ is a
normal subgroup by Proposition 192. Conversely if H is normal in G then
π : G → G/H given by π(g) = gH
is a homomorphism with kernel H. This follows as
π(g1 g2 ) = (g1 g2 ) H = (g1 H) (g2 H) = π(g1 )π(g2 )
and
π(g) = H ⇐⇒ gH = H ⇐⇒ g ∈ H.

58 NORMAL SUBGROUPS. QUOTIENT GROUPS

Example 220 Let G = Z and H = nZ then
Z/nZ = {0 + nZ, 1 + nZ, . . . , (n − 1) + nZ}
and can naturally be identified with Zn = 0̄, 1̄, . . . , n − 1 , the integers mod n.

Example 221 Let G = Sn and H = An . Then

Sn /An = {An , Sn \An } = {evens, odds} ∼
= C2 .
Example 222 Let G = D2n and H = r . Then D2n / r ∼ = C2 . We have "modded out" by
so much that all that remains is a memory of whether the group element kept the polygon the
same side up or flipped the polygon over.

Example 223 Let G = C∗ and H = S 1 . Then

C∗ /S 1 ∼
= (0, ∞)
and this time we essentially "modded out" any details about argument with all elements on the
circle |z| = r being made equivalent to r.

Example 224 As a further example

S4 /V4 ∼
= S3 .
See Exercise Sheet 6, Question 5.

Example 225 Let G = AGL(n, R) denote the group of affine maps

f (x) = Ax + b, A ∈ GL(n, R) and b ∈ Rn
and T denote the subgroup of translations t(x) = x + c. Then T ⊳ G and G/T ∼
= GL(n, R).
Solution. This follows as
f −1 tf (x) = f −1 t (Ax + b) = f −1 (Ax + b + c) = A−1 ((Ax + b + c) − b) = x + A−1 c
and we can naturally identify f T with A to show the isomorphism G/T ∼
= GL(n, R).
Example 226 Let n = 2k be even and let G = D2n and H = Z (D2n ) = e, rk noting that rk
is a half turn. Then
D2n / rk = ē, r̄, r̄2 , . . . r̄k−1 , s̄, r̄s̄, . . . , r̄k−1 s̄
where
r̄k = s̄2 = ē and s̄r̄ = r̄−1 s̄
which is a presentation of Dn when k 3 and is a presentation of V4 when k = 2. So
D2n / rn/2 ∼
= Dn (for n 6) and D8 / r2 ∼
= V4 .
The real importance of quotient groups will become more apparent when we meet the First
Isomorphism Theorem.

NORMAL SUBGROUPS. QUOTIENT GROUPS 59

7.1 Congruences (Off-syllabus)

There is another way of introducing quotient groups and that is via congruences.
Definition 227 Let G be a group and ∼ an equivalence relation on G. Then we say that ∼ is
a congruence if
(i) whenever g1 ∼ h1 and g2 ∼ h2 then g1 g2 ∼ h1 h2 .
(ii) whenever g ∼ h then g −1 ∼ h−1 .
So a congruence is an equivalence relation which respects the group operations. We have
met examples of congruences already such as = mod n. It is more apparent from this definition
that G/ ∼ will be a well-defined group in a natural way.
Proposition 228 Let ∼ be a congruence on a group G. Then the congruence classes G/ ∼
naturally form a group under the group operation ḡ ∗ h̄ = gh where ḡ denotes the congruence
class of g.
Proof. ∗ is a well-defined binary operation from (i) in the definition above. Associativity in
G/ ∼ is inherited from associativity in G and we also easily see eG/∼ = ē and (ḡ)−1 = g −1 .
The following propositions show that introducing quotient groups via congruences is entirely
equivalent to introducing them via normal subgroups.
Proposition 229 Let H G. Then the equivalence relation g ∼ k ⇐⇒ gH = kH is a
congruence if and only if H is normal.
Proof. Suppose that ∼ is a congruence and that h ∈ H, g ∈ G. Then h ∼ e and so, as ∼ is
a congruence, g −1 hg ∼ g −1 eg = e showing that g −1 hg ∈ H. Conversely suppose that H ⊳ G.
Then ∼ is a congruence by Proposition 215.
Proposition 230 Let ∼ be a congruence on group G. Then
(a) ē is a normal subgroup of G.
(b) G/ ∼ is isomorphic to G/ē.
Proof. (a) Clearly e ∈ ē. If h, k ∈ ē then, as ∼ is a congruence,
h−1 k ∼ e−1 e = e
and so h−1 k ∈ ē. Hence ē G. Finally if h ∈ ē and g ∈ G then
g −1 hg ∼ g −1 g = e
and so g −1 hg ∈ ē and ē ⊳ G.
(b) G/ ∼ is naturally isomorphic to G/ē via the isomorphism
φ : G/ ∼ → G/ē, φ (ḡ) = gē.
Firstly this is well-defined: for if ḡ = h̄ then h−1 ∼ g −1 and so h−1 g ∼ g −1 g = e. Then h−1 g ∈ ē
and gē = hē. Finally
φ(ḡ h̄) = φ(gh) = (gh)ē = (gē) (hē) = φ(ḡ)φ(h̄).

60 NORMAL SUBGROUPS. QUOTIENT GROUPS

8. FIRST ISOMORPHISM THEOREM

The usefulness of quotient groups will become more apparent once we understand the impli-
cations of the First Isomorphism Theorem. This theorem will have particular importance in
answering the question: find all the homomorphisms between two groups G and H. This is
relatively straightforward when G is cyclic.
Proposition 231 The map φ : Zm → Zn given by
φ(r mod m) = kr mod n
is a well-defined homomorphism if and only if n divides km. As Zm is cyclic then every homo-
morphism Zm → Zn is of this form.
Proof. If φ is a well-defined function then as 0̄ = m̄ in Zm we need 0̄ = km to be true in Zn .
That is we need n to divide km. Conversely, suppose that n divides km so that km = nc for
some c. For φ to be well-defined, we need to ensure that
r1 mod m = r2 mod m =⇒ kr1 mod n = kr2 mod n,
or equivalently that if m divides x = r1 − r2 then n divides kx. If m divides x then x = md for
some d and hence
kx = k (md) = (km) d = (nc) d
is a multiple of n as required. It is then an easy check to see that φ(r̄) = kr̄ is a homomorphism
provided it is well-defined.
Example 232 How many homomorphisms are there (i) from Z6 to Z12 (ii) from Z12 to Z7 ?
(iii) from Z10 to D8 ?
Solution. (i) By the previous proposition we know that any homomorphism Z6 → Z12 is of
the form r̄ → kr̄ where 12|6k or equivalently 2|k. Hence k is even. But as k and k + 12 would
lead to the same homomorphism then there are in fact only six homomorphisms
n̄ → 0, n̄ → 2n̄, n̄ → 4n̄, n̄ → 6n̄, n̄ → 8n̄, n̄ → 10n̄.
(ii) Homomorphisms Z10 → D8 . As in Proposition 188, it again follows that φ is entirely
determined by φ(1) as Z10 is cyclic. Further o(1) = 10 and so o(φ(1)) divides 10. Also o(φ(1))
divides |D8 | = 8 as a consequence of Lagrange’s Theorem. Combining these facts we see o(φ(1))
divides 2. The orders of the elements of D8 are given in the table below:
g e r r2 r3 s rs r2 s r3 s
o(g) 1 4 2 4 2 2 2 2
The possible values of φ(1) are e, r2 , s, rs, r2 s, r3 s and again each of these leads to a well-defined
homomorphism.
n n
n̄ → e, n̄ → r2n , n̄ → sn , n̄ → (rs)n , n̄ → r2 s , n̄ → r3 s .

FIRST ISOMORPHISM THEOREM 61

Example 233 Find the kernels and images of the homomorphisms in Example 232.

Solution. (i).Homomorphisms from Z6 to Z12 .

φ n̄ → 0 n̄ → 2n̄ n̄ → 4n̄ n̄ → 6n̄ n̄ → 8n̄ n̄ → 10n̄

ker φ Z6 {0̄} {0̄, 3} {0̄, 2̄, 4̄} {0̄, 3̄} {0̄}
Im φ {0̄} 0̄, 2̄, 4̄, 6̄, 8̄, 10 {0̄, 4̄, 8̄} {0̄, 6̄} {0̄, 4̄, 8̄} 0̄, 2̄, 4̄, 6̄, 8̄, 10

Note in each case that |ker φ| × |Im φ| = 6 = |Z6 | .

(ii) Homomorphisms from Z10 to D8 .

n n
φ n̄ → e n̄ → r2n n̄ → sn n̄ → (rs)n n̄ → (r2 s) n̄ → (r3 s)
ker φ Z6 2̄ = {0̄, 2̄, 4̄, 6̄, 8̄} 2̄ 2̄ 2̄ 2̄
Im φ {e} {e, r} {e, s} {e, rs} {e, r2 s} {e, r3 s}

Notice in each case that |ker φ| × |Im φ| = 10 = |Z10 | .

Let G be a group and consider finding all the homomorphisms from Z to G. A particular
homomorphism φ is entirely determined by φ(1). If g = φ(1) and n = o(g) < ∞ then we
see that Im φ = g ∼ = Zn and that ker φ = nZ. We have wrapped Z into G going around
and around g and repeating with period n. The map φ is not 1-1 and we have collapsed its
domain, Z, in such a way for its image to match a subgroup of G. More technically we have
that G/ ker φ = Z/nZ is isomorphic to Im φ ∼ = Zn with the pre-image of any element of the
image being a coset of the kernel.
If we are considering homomorphisms from Zm into G these are likewise determined by φ(1̄)
as Zm is cyclic but we can only have φ(1̄) = g if o(g) = n divides m as we must have that
φ(m̄) = eG for well-definedness. In this case ker φ = n̄ and again we are collapsing Zm onto
Zm / n̄ ∼= Zn ∼= g .
More generally, when considering homomorphisms φ : G → H where G need not be cyclic,
we will still be addressing the same problem of how are we to collapse the group G by ker φ
in order to fit G onto some subgroup of H as its image. This is where the First Isomorphism
Theorem helps.

Theorem 234 (First Isomorphism Theorem, Jordan 1870) Let φ : G → H be a homomor-

phism between two groups. Then
(a) ker φ ⊳ G;
(b) Im φ H;
(c) the map g ker φ → φ(g) gives an isomorphism

G ∼
= Im φ.
ker φ

Corollary 235 Let φ : G → H be a homomorphism between two groups and assume G is finite.
Then
|G| = |ker φ| × |Im φ|

62 FIRST ISOMORPHISM THEOREM

Proof. Parts (a) and (b) have been proven already in Propositions 192 and 193.
(c) Consider the map
G
φ̄ : → Im φ given by g ker φ → φ (g) .
ker φ
Let g1 , g2 ∈ G. From Proposition 199

φ(g1 ) = φ(g2 ) ⇐⇒ g1 ker φ = g2 ker φ.

This both shows that φ̄ is well-defined and that φ̄ is 1-1. It is also clear that φ̄ is onto. Finally
φ̄ is a homomorphism as

φ̄((g1 ker φ) ∗ (g2 ker φ)) = φ̄((g1 g2 ) ker φ) = φ(g1 g2 ) = φ(g1 )φ(g2 ) = φ̄(g1 ker φ) ∗ φ̄(g2 ker φ).

The Corollary follows because the order of G/ ker φ is |G| / |ker φ| by Lagrange’s Theorem

Example 236 For sgn : Sn → {±1} , the Isomorphism Theorem reads Sn /An ∼
= {±1} .

Example 237 For det : GL(n, R) → R∗ , the Isomorphism Theorem reads

GL(n, R)/SL(n, R) ∼
= R∗ .

Example 238 For projection onto the first coordinate π 1 : G1 × G2 → G1 , the Isomorphism
Theorem reads (G1 × G2 ) / ({e} × G2 ) ∼
= G1 .

Example 239 For φ : Z → Z given by φ(x) = nx, then the Isomorphism Theorem reads
Z = Z/ {0} ∼
= nZ.

Example 240 For φ : Z → Zn given by φ (x) = x̄, then the Isomorphism Theorem reads
Z/nZ ∼
= Zn .

Example 241 For φ : Z8 → Z12 given by φ (x̄) = 3x̄ then the Isomorphism Theorem reads
Z8 / 4̄ ∼
= 3̄ .

Example 242 And for the following abelian groups, applying the Isomorphism Theorem to the
homomorphism φ(x) = x2 gives

S 1 / {±1} ∼
= S 1, Z5 / {0̄} ∼
= Z5 , Z6 / 3̄ ∼
= 2̄ .

Remark 243 The question of finding how many homomorphisms φ : G → H there are between
two groups G and H can be a difficult one if the groups are large and internally complicated,
but the Isomorphism Theorem gives us a means in principle to determine that number. The
process is as follows:
(i) Determine the normal subgroups N of G. (These are the potential kernels.)
(ii) Determine the number of subgroups in H which are isomorphic to G/N; let’s call this
number n(N). (These are the possible images when the kernel is N .)
(iii) For those normal subgroups where n(N) > 0, determine the order of Aut(G/N ).

FIRST ISOMORPHISM THEOREM 63

Then the number of homomorphisms from G to H is

n(N) × |Aut(G/N)| .
n(N)>0

To explain this formula, the Isomorphism Theorem tells us that any homomorphism will map
G onto an image which is isomorphic to G/N for some normal subgroup N. Conversely, given
a subgroup I H which is isomorphic to G/N for some normal N, there are |Aut(G/N )|
isomorphisms ι from G/N to I and for each such isomorphism the map
π ι
G → G/N → I

(where π : G → G/N is the map g → gN ) will be a homomorphism from G to H with image I.

This is a very general approach to the problem at hand. In the case of G being cyclic or small,
it may well be much easier to focus on some generators of G and consider their possible
images, recalling that o(φ(g)) divides o(g) as we did in Example 233. Approaching it that way
though, you will of course need to make sure that potential φ(g) do indeed lead to actual
homomorphisms.

Example 244 How many homomorphisms are there from S3 to C4 × C2 ?

Solution. Take g and h to be generators, respectively, of C4 and C2 .

Recall that the normal subgroups of S3 are {e} , A3 , S3 (see Example 210) and we have that

S3 / {e} ∼
= S3 , S3 /A3 ∼
= C2 , S3 /S3 ∼
= {e} .

There are no homomorphisms with kernel {e} as C4 × C2 has no subgroup isomorphic to S3

(which is not abelian).
There are 3 elements in C4 × C2 with order 2, namely (e, h) , (g 2 , h) , (g 2 , e) and to each of
these corresponds a homomorphism with kernel A3 .
The final homomorphism is the map φ(σ) = (e, e) for all σ ∈ S3 . In all, then, there are four
such homomorphisms.

(e, e) σ ∈ A3 (e, e) σ ∈ A3 (e, e) σ ∈ A3

φ1 (σ) = , φ2 (σ) = , φ3 (σ) = ,
(0̄, h) σ ∈
/ A3 (g 2 , h) σ ∈
/ A3 (g 2 , 3) σ ∈
/ A3

and φ4 (σ) = (e, e) for all σ.

Example 245 How many homomorphisms are there from A4 to S3 ?

Solution. We have seen (Example 86) that the conjugacy classes of A4 are {e} and

{(12) (34) , (13) (24) , (14) (23)} , {(123) , (134) , (214) , (324)} , {(132) , (143) , (124) , (234)} .

So the normal subgroups of A4 are {e} , V4 , A4 and we have that

A4 / {e} ∼
= A4 , A4 /V4 ∼
= C3 , A4 /A4 = {e} .

64 FIRST ISOMORPHISM THEOREM

There is no subgroup of S3 which is isomorphic to A4 and one, namely A3 , which is isomorphic
to C3 .
There is obviously one homomorphism with kernel A4 . And having ker φ = V4 means that
Im φ is isomorphic to C3 and so must be A3 ⊆ S3 . However there are two isomorphisms
from C3 to C3 and so two different ways of wrapping A4 /V4 onto A3 and hence two different
homomorphisms A4 → S3 with image A3 . Thus in all we have three homomorphisms A4 → S3
given by
 
 e σ ∈ V4  e σ ∈ V4
φ1 (σ) = (123) σ ∈ (123) V4 , φ2 (σ) = (132) σ ∈ (123) V4 , φ3 (σ) = e.
 
(132) σ ∈ (132) V4 (123) σ ∈ (132) V4

Example 246 Show that Aut(V4 ) is isomorphic to S3 .

Solution. If we write a = (12) (34) , b = (13) (24) , c = (14) (23) then any automorphism of V4
must send e to e and {a, b, c} to {a, b, c}. Hence there are at most 6 automorphisms. However
if we note V4 may be presented as
a, b : a2 = e = b2 , ab = ba = b, c : c2 = e = b2 , bc = cb = a, c : a2 = e = c2 , ac = ca
then we see that there is a symmetry in the roles of a, b, c and so any element of Sym {a, b, c}
is indeed an automorphism of V4 .
Example 247 How many homomorphisms are there from D8 to A4 ?
Solution. Recall that the conjugacy classes of D8 are
{e} , r, r3 , r2 , s, r2 s , rs, r3 s .
Hence the normal subgroups of D8 are
N1 = {e} , N2 = e, r2 , N3 = e, r, r2 , r3 ,
N4 = e, s, r2 , r2 s , N5 = e, s, r2 , r2 s , N6 = D8
and we have that (see Example 226)
D8 /N1 ∼
= D8 , D8 /N2 ∼
= V4 , D8 /N3 ∼
= D8 /N4 ∼
= D8 /N5 ∼
= C2 , D8 /N6 ∼
= {e} .
By Lagrange’s Theorem, there is no subgroup of A4 which is isomorphic to D8 . There is one
subgroup (namely V4 ) which is isomorphic to V4 and three subgroups of A4 which are isomorphic
to C2 (namely a , b and c ).
From the previous example we know there are 6 homomorphisms with kernel N2 and image V4 ;
such an example is
φ(e) = φ(r2 ) = e, φ(r) = φ(r3 ) = a, , φ(s) = φ(r2 s) = b, φ(rs) = φ(r3 s) = c.
There are 3 × 3 = 9 homomorphisms with kernel N3 or N4 or N5 and image a or b or c ;
such an example is
φ(e) = φ(s) = φ(r2 s) = φ(r2 ) = e, φ(r) = φ(r3 ) = φ(rs) = φ(r3 s) = b.
Finally there is the homomorphism with image {e} .
In all then there are 6 + 9 + 1 = 16 homomorphisms from D8 to A4 .

FIRST ISOMORPHISM THEOREM 65

66 FIRST ISOMORPHISM THEOREM
9. GROUP ACTIONS

We move now, from thinking of groups in their own right, to thinking of how groups can move
sets around — for example, how Sn permutes {1, 2, . . . , n} and matrix groups move vectors.

Definition 248 A left action of a group G on a set S is a map

ρ: G × S → S

such that:
(i) ρ(e, s) = s for all s ∈ S;
(ii)ρ(g, ρ(h, s)) = ρ(gh, s) for all s ∈ S and g, h ∈ G.

Notation 249 We will normally write g · s for ρ(g, s) and so (i) and (ii) above would now read
as:
(i’) e · s = s for all s ∈ S;
(ii’) g · (h · s) = (gh) · s for all s ∈ S and g, h ∈ G.

Remark 250 We will think of g · s ∈ S as the point that s is moved to by g.

Example 251 The group GL(n, R) acts on Rn by

A · v = Av for A ∈ GL(n, R) and v ∈ Rn

as
In v = v and (AB) v = A (Bv) for v ∈ Rn , A, B ∈ GL(n, R).

Example 252 The group GL(n, R) acts on the set Mnn (R) of real n×n matrices by conjugation

A · M = AM A−1 .

We can verify that this is a left action by noting

In ·M = In M In−1 = M and (AB) ·M = (AB) M (AB)−1 = A BMB −1 A−1 = A· (B · M ) .

Example 253 Another action of GL(n, R) on the set Mnn (R) is given by

A · M = AM.

Example 254 Let S be a polygon in R2 or a polyhedron in R3 . Then the symmetry group of

S — those isometries g that satisfy g(S) = S — acts naturally on S. The symmetry group can
separately be considered as acting on the set of vertices of the polyhedron, or edges, or faces.

GROUP ACTIONS 67
Example 255 Let S denote the set of triangles in R2 and let

G1 = v → Av + b : A ∈ GL(2, R), b ∈ R2 .
G2 = v → Av + b : A ∈ O(2), b ∈ R2 .

That is, G1 is the group of affine transformations of the plane and G2 is the group of isometries
of the plane. Then G1 and G2 act naturally on S by g · ∆ = g(∆).

And then there are various examples where a group acts on itself or subsets of itself in a
natural way.

Example 256 Let G be a group. Then we have a left action of G on itself by g · h = gh for
g, h ∈ G.

Example 257 Let G be a group. Then we have a left action of G on itself by conjugation —
that is
g · h = ghg −1 .
We clearly have e · g = g and

g · (k · h) = g · khk −1 = gkhk −1 g −1 = (gk) h (gk)−1 = (gk) · h.

Example 258 Let H be a (not necessarily normal) subgroup of a group G and let G/H denote
the set of left cosets of H. Then there is a left action of G on G/H by

g1 · (g2 H) = (g1 g2 ) H.

We can similarly consider right actions of groups:

Definition 259 A right action of a group G on a set S is a map

ρ: S × G → S

such that:
(i) ρ(s, e) = s for all s ∈ S;
(ii) ρ(ρ(s, h), g) = ρ(s, hg) for all s ∈ S and g, h ∈ G.

There is no particular benefit to considering left actions over right actions or vice versa.
Examples of right actions that we have met are:

Example 260 (i) Sn acts on {1, 2, . . . , n} by ρ(k, σ) = kσ.

(ii) Sn acts on the power set of {1, 2, . . . , n} (the set of subsets) by ρ(S, σ) = Sσ.
(ii) GL(n, R) acts on the set of 1 × n row vectors by ρ(v, A) = vA.
(iii) A group G acts on the set of right cosets of a subgroup H by ρ(Hk, g) = Hkg.
(iv) A group G acts on itself by translation by ρ(h, g) = hg.
(v) A group G acts on itself by conjugation by ρ(h, g) = g −1 hg.

68 GROUP ACTIONS
Definition 261 If a group G acts on a set S and s ∈ S then:
(i) the orbit of s, written Orb(s), is defined as

Orb(s) = {g · s : g ∈ G} ⊆ S.

If there is only one orbit then we say that the action is transitive.
(ii) the stabilizer of s, written Stab(s), is defined as

Stab(s) = {g ∈ G : g · s = s} ⊆ G.

Example 262 When Sn (right) acts on {1, 2, . . . , n} by ρ(k, σ) = kσ then there is just one
orbit. Note that

Stab(n) = {σ ∈ Sn : nσ = n} = Sym {1, 2, . . . , n − 1} ∼

= Sn−1 .

Example 263 When Sn (right) acts on the subsets of {1, 2, . . . , n} by ρ(S, σ) = Sσ then there
are n + 1 orbits, one for each possible size of |S| and if |S| = k then

Stab(S) ∼
= Sk × Sn−k .

(See Exercise Sheet 6, Question 3.)

Example 264 When GL(n, R) acts on Rn by A · v = Av, there are just two orbits {0} and
Rn \{0}. If A ∈ GL(n, R) then A0 = 0 whilst if v = 0 then v can be extended to a basis which
can be used as the columns of an invertible matrix A. Then Ae1 = v where e1 = (1, 0, . . . , 0)T .

Example 265 When GL(2, C) acts on the set M22 (C) of complex n×n matrices by conjugation

A · M = AM A−1 ,

then the an orbit either has a representative of the form

λ 0
λ, µ ∈ C
0 µ

when the matrix is diagonalizable or of the form

λ 1
λ∈C
0 λ

when the matrix is not diagonalizable.

Example 266 Let S be the set of black-or-white colourings of a square’s edges. As there are
four edges then |S| = 24 = 16. The square’s symmetry group D8 acts naturally on S and there
are six orbits with a representative of each orbit listed below

W W W W, BW W W, BBW W, BW BW, BBBW, BBBB,

where the edges’ colours are listed in clockwise order.

GROUP ACTIONS 69
Example 267 Let S be the set of black-or-white colourings of a cube’s faces. As there are six
faces then |S| = 26 = 64. The cube’s rotational symmetry group acts naturally on S and there
are 10 orbits with a representative of each orbit listed below:
all white, 1 black, 2 opposite black faces, 2 adjacent black faces,
3 black faces in a C, 3 black faces around a corner,
2 opposite white faces, 2 adjacent white faces, 1 white, all black.
Example 268 (From Example 255.) When the affine group group of R2 acts on the set of
triangles, then there is just one orbit as any triangle can be taken to any other triangle via an
affine map. The orbits when the isometry group acts are the congruence classes, as two triangle
are related by an isometry if and only if they are congruent.
Example 269 When a group G acts on itself by g · h = gh then the action is transitive and
each stabilizer is just {e} .
Example 270 When a group G acts on itself by g·h = ghg −1 then the orbit of g is its conjugacy
class and its stabilizer is the centralizer CG (g).
Example 271 When H G and G acts on the set of cosets G/H by g1 · (g2 H) = (g1 g2 ) H
then the action is transitive and the stabilizer of gH is gHg −1 as
k · gH = gH ⇐⇒ kgH = gH ⇐⇒ g −1 kg ∈ H ⇐⇒ k ∈ gHg −1 .
We conclude with three important results relating to orbits and stabilizers.
Proposition 272 The orbits of an action partition the set.
Proof. Let G be a group acting on a set S. We introduce a binary relation ∼ on S by setting,
for s, t ∈ S,
s ∼ t ⇐⇒ there exists g ∈ G such that g · s = t.
We shall show that ∼ is an equivalence relation and that the equivalence classes are the orbits.
(a) ∼ is reflexive as s = e · s for all s ∈ S;
(b) ∼ is symmetric as
s ∼ t =⇒ g · s = t for some g ∈ G
=⇒ g −1 · t = s
=⇒ t ∼ s.
(c) ∼ is transitive as
s ∼ t, t ∼ u =⇒ g · s = t and h · t = u for some g, h ∈ G
=⇒ (hg) · s = h · (g · s) = h · t = u
=⇒ s ∼ u.
Hence ∼ is an equivalence relation and, in particular, the equivalence classes partition S. For
s ∈ S, note that the equivalence class s̄ equals
s̄ = {g · s : g ∈ G} = Orb(s).

70 GROUP ACTIONS
Proposition 273 The stabilizers of an action are subgroups.

Proof. Let G be a group acting on a set S and s ∈ S. Then

e ∈ Stab(s) as e · s = s.

If g, h ∈ Stab(s) then
(gh) · s = g · (h · s) = g · s = s
showing that gh ∈ Stab(s) and

g −1 · s = g −1 · (g · s) = g −1 g · s = e · s = s

showing that g −1 ∈ Stab(s).

Proposition 274 If two elements lie in the same orbit then their stabilizers are conjugate.

Proof. If s, t lie in the same orbit of an action then there exists g ∈ G such that g · s = t. Then

h ∈ Stab(s) ⇐⇒ h·s=s
⇐⇒ h · g −1 · t = g −1 · t
⇐⇒ g · h · g −1 · t = g −1 · t = t
⇐⇒ ghg −1 · t = t
⇐⇒ ghg −1 ∈ Stab(t)
⇐⇒ h ∈ g −1 Stab(t)g.

Hence
Stab(s) = g −1 Stab(t)g.

GROUP ACTIONS 71
72 GROUP ACTIONS
10. ORBIT-STABILIZER THEOREM

Theorem 275 (Orbit-Stabilizer Theorem) Let G be a finite group acting on a set S and
let s ∈ S. Then
|G| = |Stab(s)| × |Orb(s)| .

Proof. We shall show that there is a well-defined bijection between the cosets of Stab(s) in G
and Orb(s). This will then mean by Lagrange’s Theorem that
|G|
= # cosets of Stab(s) = |Orb(s)| .
|Stab(s)|
We define the map
φ : G/Stab(s) → Orb(s) by φ(gStab(s)) = g · s.
We first need to show that φ is well-defined (i.e. that the image of the coset gStab(s) is not
dependent on the choice of representative g). Note that
gStab(s) = hStab(s) ⇐⇒ h−1 g ∈ Stab(s)
⇐⇒ h−1 g · s = s
⇐⇒ g · s = h · s.
This shows that φ is indeed well-defined. The reverse implications show that φ is 1—1. Finally
it is immediately apparent that φ is onto as every element of Orb(s) can be written as g · s for
some g.

Remark 276 In fact, φ is more than just a bijection, it is an isomorphism of actions of G. If

G acts on two sets S and T, an isomorphism of these actions is a bijection φ : S → T such that
φ(g · s) = g · (φ(s)).

Remark 277 As an immediate consequence, note that the size of an orbit must divide the
order of the group. (We already knew this to be true of stabilizers as they are subgroups.)

Corollary 278 (Lagrange’s Theorem) Let G be a group and H G. Then G acts on G/H by
g · (kH) = (gk) H.

Proof. With this action

Stab(H) = H and Orb(H) = G/H.
By the Orbit-Stabilizer Theorem
|G/H| × |H| = |G| .

ORBIT-STABILIZER THEOREM 73
Corollary 279 Let G be a group, g ∈ G and
CG (g) = {h ∈ G : gh = hg} = centralizer of g,
C(g) = h−1 gh : h ∈ G = conjugacy class of g.
Then
|CG (g)| × |C(g)| = |G| .
Proof. G acts on itself by conjugation:
g · h = ghg −1 .
For g ∈ G
Stab(g) = h ∈ G : hgh−1 = g = CG (g), Orb(g) = C(g).

Example 280 Determine the number of conjugates of (1 2 3) in A5 .

Solution. First we will find CA5 ((123)). We note
(123) ρ = ρ (123) ⇐⇒ (123) = ρ−1 (123) ρ = (1ρ 2ρ 3ρ) .
As there are three ways of writing (123) , the others being (231) and (312), then
1ρ = 1, 2ρ = 2, 3ρ = 3 or 1ρ = 2, 2ρ = 3, 3ρ = 1 or 1ρ = 3, 2ρ = 1, 3ρ = 2.
So 1, 2, 3 must cycle in one of three ways, all of them even permutations. For ρ to be even, it
must be the case that 4ρ = 4 and 5ρ = 5. Hence |CA5 ((123))| = 3 and
|A5 | 60
|C ((123))| = = = 20.
3 3
As there are, in all, 5×4×3
3
= 20 3-cycles in A5 then the conjugacy class of (123) in A5 is the
entire set of 3-cycles. (Compare this with Example 86.)
Example 281 Show that there are 12 rotational symmetries of a regular tetrahedron and 24
of a cube.
Solution. Let GT denote the tetrahedral group. A tetrahedron has 4 vertices and GT acts
transitively on this set. The stabilizer of a particular vertex v consists of the identity and the
two ±120◦ rotations about an axis through v and the midpoint of the opposite face. Hence
|GT | = |Stab(v)| × |Orb(v)| = 3 × 4 = 12.
Let GC denote the cube’s group. A cube has 8 vertices and GC acts transitively on this set.
The stabilizer of a particular vertex v consists of the identity and the two ±120◦ rotations
about the diagonal through v and its opposite vertex. Hence
|GC | = |Stab(v)| × |Orb(v)| = 3 × 8 = 24.
Either of these calculations could as easily have been performed by looking at the actions on
edges and faces. Using the faces of a cube, for example, we would have concluded
|GC | = |Stab(f)| × |Orb(f )| = 4 × 6 = 24.

74 ORBIT-STABILIZER THEOREM
Proposition 282 A group G of order pr , where p is prime, has a non-trivial centre.

Proof. Let G act on itself by conjugation. Recall that the centre of G is Z(G) = {g ∈ G : hg = gh ∀h ∈ H},
so being in Z(G) is equivalent to being a singleton orbit. There is at least one such orbit, namely
{e}. By the Orbit-Stabilizer Theorem, all the orbits have size pk for some 0 k < r. If Nk is
the number of such orbits then, as the orbits partition G we have

N0 + N1 p + N2 p2 + · · · + Nr−1 pr−1 = pr ,

and in particular we see that N0 is a multiple of p. As N0 1 then there are other singleton
orbits, i.e. other elements of Z(G).

Proposition 283 A group G of order p2 is isomorphic to Cp2 or Cp × Cp .

Proof. If G has an element of order p2 then G ∼= Cp2 . If there is no such element, then the non-
trivial elements have order p. Take a non-trivial element x from the centre Z(G) and another
element y not in x . Then the elements xi y j (0 i, j < p) are all distinct as x ∩ y = {e}.
Further as x is Z(G) then these elements multiply by

xi y j ∗ xI y J = xi+I y j+J

and we see that G ∼

= Cp × Cp .

Example 284 Let G be a group with three conjugacy classes. Show that G ∼
= C3 or G ∼
= S3 .

Solution. The conjugacy class of e is just {e}. Denote the size of the remaining two classes as
c1 and c2 . Both c1 and c2 are factors of |G| and so |G| = k1 c1 = k2 c2 for some k1 , k2 . Without
any loss of generality we may assume that k1 k2 . As the conjugacy classes partition G then

1 + c1 + c2 = k1 c1 = k2 c2

so that
1 + c2 = (k1 − 1)c1 , 1 + c1 = (k2 − 1)c2 .
Eliminating c2 and rearranging somewhat, we arrive at
k2
c1 = .
(k1 − 1) (k2 − 1) − 1
If k1 > 3 then we have the contradiction
k2 k2 k2
< = < 1.
(k1 − 1) (k2 − 1) − 1 2 (k2 − 1) − 1 2k2 − 3
If k1 = 1 then c1 = −k2 < 0, a further contradiction. So the possibilities are

k1 = 2, k2 = 3, c1 = 3, c2 = 2, |G| = 6;
k1 = 3, k2 = 3, c1 = 1, c2 = 1, |G| = 3.

As C6 has six conjugacy classes, the first possibility leads to S3 only and the second to C3 .

ORBIT-STABILIZER THEOREM 75
Theorem 285 (Cauchy’s Theorem) Let G be a finite group and let p be a prime dividing
|G| . Then G has an element of order p.

Proof. Let S denote the set

S = {(g1 , g2 , . . . , gp ) ∈ Gp : g1 g2 · · · gp = e} .

Note that |S| = |G|p−1 as the first p − 1 elements g1 , . . . , gp−1 may be chosen freely from G and
then gp = (g1 g2 · · · gp−1 )−1 is determined. Let σ = (123 · · · p) and note that there is an action
of σ ∼= Cp on S by
σ · (g1 , g2 , . . . , gp ) = (g2 , g3 , . . . , gp , g1 )
as

(g1 , g2 , . . . , gp ) ∈ S ⇐⇒ g1 g2 · · · gp = e
⇐⇒ g2 · · · gp = g1−1
⇐⇒ g2 · · · gp g1 = e
⇐⇒ (g2 , g3 , . . . , gp , g1 ) ∈ S.

We consider the orbits of this action. As | σ | = p then, by the Orbit-Stabilizer Theorem, the
orbits may have size 1 or p. If (g1 , g2 , . . . , gp ) is in an orbit of size 1 then

σ · (g1 , g2 , . . . , gp ) = (g2 , g3 , . . . , gp , g1 ) = (g1 , g2 , . . . , gp )

and hence we see that

g1 = g2 = · · · = gp and (g1 )p = e.
We also see that there is at least one singleton orbit, namely {(e, e, . . . , e)}. As the orbits
partition S then
|S| = k + lp
where k is the number of singleton orbits and l is the number of orbits of size p. As p divides
|G| then p divides |S| = |G|p−1 and hence p divides k. From our earlier comment k 1 and
hence there is at least one other singleton orbits besides {(e, e, . . . , e)}. Again from our earlier
comments, this other singleton orbit is of the form {(g, g, . . . , g)} where g p = e.

76 ORBIT-STABILIZER THEOREM
11. COUNTING ORBITS

Theorem 286 (Orbit Counting Formula) Let G be a finite group acting on a finite set S.
Then
1
# orbits = |fix(g)|
|G| g∈G
where, for g ∈ G, we define
fix(g) = |{s ∈ S : g · s = s}| .
Proof. We will consider the set
A = {(g, s) : g · s = s} ⊆ G × S
and count up |A| in two different ways. Then

|A| = |{s ∈ S : g · s = s}| = |{g ∈ G : g · s = s}| .

g∈G s∈S

The first and second sums respectively equal

|fix(g)| and |Stab(s)| .
g∈G s∈S

If the orbits are O1 , O2 , . . . , ON then

N
|Stab(s)| = |Stab(s)| [as the orbits partition S].
s∈S i=1 s∈Oi

Using the Orbit-Stabilizer Theorem, this is turn can be rewritten as

N N N
|G|
|Stab(s)| = = |G| = N |G| .
i=1 s∈Oi i=1 s∈Oi
|Oi | i=1

Hence
N |G| = |fix(g)|
g∈G

and the result follows.

Remark 287 Note that if g and h are conjugate then fix(g) = fix(h). To see this, say g =
k −1 hk, and note
g · s = s ⇐⇒ k −1 hk · s = s ⇐⇒ h · (k · s) = (k · s) .
So we can rewrite the orbit counting formula as
1
N= |fix(g)| |C(g)|
|G| g∈C
i

where we take a representative g from each of the conjugacy classes Ci .

COUNTING ORBITS 77
Remark 288 The orbit counting formula is often (incorrectly) called Burnside’s Lemma as
it was stated and proved in an 1897 text of William Burnside, but the result had been much
earlier known to Frobenius and Cauchy.

Example 289 In how many essentially different ways can a triangle’s edges be painted with n
colours. Compare your answers for n = 2 and n = 3 with those found in Sheet 6, Exercise 4.

Solution. With labels still present there are n3 colourings. So we can form the table

g # of conjugates s fixed by g fix(g) contribution

e 1 all n3 n3
r 2 C1 C1 C1 n 2n
2
s 3 C1 C1 C2 n 3n2

By the orbit counting formula there are

n3 + 3n2 + 2n n (n + 1) (n + 2)
=
6 6
essentially different colourings. Note this fomula gives 2 × 3 × 4/6 = 4 and 3 × 4 × 5/6 = 10
colourings when n = 2 and n = 3 as previously calculated.

Remark 290 How might this have been calculated directly? The triangles sides might be
coloured with 1 or 2 or 3 colourings. There are respectively n and 2 n2 and n3 such colourings
and so we again arrive at the answer
2n (n − 1) n (n − 1) (n − 2) 6n + (6n2 − 6n) + (n3 − 3n2 + 2n) n3 + 3n2 + 2n
n+ + = = .
2 6 6 6
However, once we move on to geometric objects with more symmetries, a direct approach quickly
becomes intractable.

Example 291 A cuboid has distinct dimensions. In how many (essentially different) ways can
the cuboid’s faces be painted black or white? Determine this number when two of the cuboid’s
dimensions are equal.

Solution. As the cuboid’s dimensions are distinct then its symmetry group is C2 × C2 (like
that of a rectangle). The three non-trivial elements are rotations through a half-turn about
each of the x- y- and z-axes.
There are 26 = 64 ways of colouring these faces whilst labelled. As the symmetry group is
abelian, the conjugacy classes are singleton sets; however we can see that each of the non-trivial
elements will fix the same number of colourings. Applying the orbit counting formula we arrive
at the table
g like elements s fixed by g fix(g) contribution
e 1 all 64 64
=e 3 C1 C1 C2 C2 C3 C4 16 48
Our answer is then
64 + 48
= 16 + 12 = 28.
4
78 COUNTING ORBITS
When two (but not three) of the dimensions are the same then the symmetry group is now D8 .
Arguing similarly we arrive at the table

g conjugates s fixed by g fix(g) contribution

e 1 all 64 64
r 2 C1 C2 C2 C2 C2 C3 8 16
2
r 1 C1 C2 C3 C2 C3 C4 16 16
s 2 C1 C2 C3 C2 C4 C1 16 32
rs 2 C1 C2 C2 C3 C3 C1 8 16

Our answer is then

64 + 16 + 16 + 32 + 16
= 8 + 2 + 2 + 4 + 2 = 18.
8

Example 292 How many different triples (x1 , x2 , x3 ) of positive integers are there such that
x1 + x2 + x3 = 100 and x1 x2 x3 .

Proof. Let
S = {(x1 , x2 , x3 ) : xi 1, x1 + x2 + x3 = 100} .
Then S3 acts naturally on S and in each orbit of this action there is a unique (x1 , x2 , x3 ) such
that x1 x2 x3 . So the question is equivalent to finding the number of orbits of this action.
Note that x1 can be any number from 1 to 98, and x2 any number from 1 to 99 − x1 , with x3
then determined by the choices of x1 and x2 . So
98
1
|S| = (99 − x1 ) = 99 × 98 − × 98 × 99 = 49 × 99 = 4851.
x1 =1
2

We apply the orbit counting formula as below:

g conjugates s fixed by g fix(g) contribution

e 1 (x1 , x2 , x3 ) 4851 4851
(12) 3 (x1 , x1 , x3 ) 49 147
(123) 2 (x1 , x1 , x1 ) 0 0

Hence the number of orbits (and our answer) equals

4851 + 147
= 833.
6

Example 293 How many essentially different ways are there to make a bracelet which has
three red beads, two blue beads and two white beads?

COUNTING ORBITS 79
Proof. These seven beads can be considered to occupy the vertices of a regular heptagon
and then two different colourings would be considered indistinguishable if some element of D14
connects the two. So we are again being asked to determine the number of orbits of this action
of D14 on the set of (labelled) colourings.
The total number of different colourings (whilst the positions are labelled) is

7! 5040
= = 210.
3!2!2! 6×2×2

Let r denote a rotation by 2π/7 and s denote a (fixed) reflection. Note that a rotation (in this
case) would only fix those colourings that are monochromatic (not possible here). All reflections
are in an axis that goes through a vertex and the opposite edge’s midpoint. A colouring would
be fixed if vertices and their mirror images were of the same colour; with the given beads this
is only possible if we colour the vertex on the axis red and the other six in pairs opposite one
another (one pair red, one blue, one white). There are 6 = 3! ways of doing this. Hence our
answer is
e rotations reflections

210 + 6 × 0 + 7 × 3! 210 + 42
= = 15 + 3 = 18.
14 14

Remark 294 We can already see (though the answer is still relatively small) that it would be
rather difficult to list these 18 arrangements by inspection and be confident we had not listed a
colouring twice nor missed any colouring. The eighteen colourings are in fact

RRRW W BB RRRBW W B RRW RBBW RRBRW W B RRBW RW B RBRBRW W

RRRW BBW RRW RW BB RRBRBW W RRW W RBB RRBW RBW RBRW RBW
RRRW BW B RRW RBW B RRBRW BW RRW BRBW RW RW RBB RBRW RW B

Example 295 Anticipating 10 students for his option, a tutor assigns 5 weekly slots in his
diary. In the end only 6 students choose to take the option. The tutor allows them to choose
from the available slots, stipulating only a maximum of 2 students per slot. In how many
different ways can the students choose to arrange themselves?

Solution. If there had been 10 students taking the option then the slots could have been
distributed in
10!
= 113400
(2!)5
ways. With, instead, 6 students on the option then there will be 4 unused places. Different
assignings of these places to the 4 "missing" students correspond to the same assignings of places
to the 6 students — which is precisely the number of assignings we are seeking to calculate. So,
alternatively, we can let S4 act on the unoccupied places from the 113400 original assignings and
determine the number of orbits of this action. For example, if we list the six students’ slots first
then the unoccupied ones, we see that 123441(2355) leads to the same teaching arrangements

80 COUNTING ORBITS
as 123441(5325).

g conjugates fixed by g fix(g) contribution

e 1 all 113400 113400
(12) 6 S1 S1 S2 S3 12600 75600
(123) 8 S1 S1 S1 S2 0 0
(1234) 6 S1 S1 S1 S1 0 0
(12) (34) 3 S1 S1 S2 S2 1800 5400

Note that, with up to two students allowed per slot, there are no arrangements of the form
S1 S1 S1 S1 or S1 S1 S1 S2 . Of the form S1 S1 S2 S2 there are
6!
× 5×4 = 1800
0!0!2!2!2!
choice of S1 and S2

arrangements fixed and of the form S1 S1 S2 S3 (where S2 and S3 may be the same) there are

6!
1800 + × 5×4×3 = 1800 + 10800 = 12600.
0!1!1!2!2!
choice of S1 ,S2 ,S3

Our final answer, then, is

113400 + 75600 + 5400
= 8100.
24

Remark 296 Had we sought to calculate this directly we could have counted as follows: the
unused four places could have been in four different (e.g. 123455), three different (e.g. 112234)
or two (e.g. 112233) different slots. These respectively correspond to

6! 5 3 6! 5 6!
5× , × , ×
2!1!1!1! 2 2 2!2!1!1! 3 2!2!2!

arrangements to give the answer

1800 + 5400 + 900 = 8100.

COUNTING ORBITS 81
82 COUNTING ORBITS
12. REPRESENTATIONS

Theorem 297 Given a left action of a group G on a set S there is an associated homomorphism

ρ : G → Sym(S).

To each homomorphism ρ : G → Sym(S) there is an associated left action of G on S. These

correspondences are inverses of one another.

Proof. Let g ∈ G. Then the map

ρg : S → S given by ρg (s) = g · s

is a bijection of S as it has inverse ρg−1 . Further the map

ρ : G → Sym(S) given by g → ρg

is a homomorphism as

ρgh (s) = (gh) · s = g · (h · s) = ρg (ρh (s)) = ρg ρh (s).

Conversely, given a homomorphism

ρ : G → Sym(S)

then there is an action of G on S given by

g · s = (ρ(g)) (s).

This is a left group action as for all s ∈ S and g, h ∈ G we have

e · s = (ρ(e)) (s) = idS (s) = s

and
(gh) · s = ρ(gh)(s) = ρ(g) (ρ(h)(s)) = g · (h · s) .

Corollary 298 (Cayley’s Theorem) Every finite group is isomorphic to a subgroup of some
permutation group Sn . (More generally, whether or note G is finite, our proof shows that G is
isomorphic to a subgroup of Sym(G).)

Proof. As G acts on itself, by g · h = gh, then we can consider the associated representation
ρ : G → Sym(G). Let G = {g1 , g2 , . . . , gn } be a finite group Then ρ(gi ), left multiplication by
gi , is a permutation of G. Further ρ is 1-1 as

ρ(gi ) = ρ(gj ) =⇒ ρ(gi )(e) = ρ(gj )(e) =⇒ gi = gj .

Hence G is isomorphic with the image of ρ(G) Sym(G) ∼

= Sn .

REPRESENTATIONS 83
Example 299 List the elements of S3 and describe the corresponding subgroup of S6 deter-
mined by Cayley’s Theorem.
Solution. We can list S3 as
g1 = e, g2 = (12) , g3 = (13) , g4 = (23) , g5 = (123) , g6 = (132) .
As S3 is generated by (12) and (13) then ρ(S3 ) is generated by ρ(12) and ρ(13). Note left-
multiplication by (12) has the following effect
g1 → g2 , g2 → g1 , g3 → g5 , g4 → g6 , g5 → g3 , g6 → g4 ,
and left-multiplication by (13) has the following effect
g1 → g3 , g2 → g6 , g3 → g1 , g4 → g5 , g5 → g4 , g6 → g2 .
Hence ρ(S3 ) is the subgroup of S6 generated by (12) (35) (46) and (13) (26) (45).
Example 300 Rotational Symmetry Groups of the Tetrahedron and Cube.
Let T be a regular tetrahedron and C be a cube. We will denote as GT the rotational
symmetry group of T and as GC the rotational symmetry group of C.

(a) If we label the four vertices of T as 1, 2, 3, 4 then we have a homomorphism

ρ : GT → S4
associating with each rotation in GT the induced permutation on the labelled vertices 1, 2, 3, 4; it
is a homomorphism as composition is the binary operation in each of the two groups. Further,
ρ is injective as no two distinct rotations lead to the same movement of the vertices. Hence we
have an isomorphism ρ : GT → Im ρ.
What is Im ρ? We already know from the Orbit-Stabilizer theorem that |GT | = 12 and hence
it follows that
GT ∼= Im ρ = A4 .
If we want to fully understand what these 12 rotations are, we see that they come in three types.
identity : 1 of these.
rotation of ± 2π/3 about a vertex and opposite face’s midpoint : 8 of these.
rotation of π about opposite edges’ midpoints : 3 of these.

84 REPRESENTATIONS
The breakdown of these 12 rotations as 1 + 8 + 3 might seem at odds with the conjugacy classes
of A4 which we know to have sizes 1, 4, 4, 3. But note that if we are looking at a rotation of
±2π/3 about a vertex and opposite face’s midpoint, and make sure to be looking down on the
vertex, then we can discern clockwise rotations from anticlockwise rotations, which is why these
8 rotations split into two conjugacy classes and why such a rotation is not conjugate to its
inverse.

(b) If we label the vertices of C as 1-8 then we would likewise have an injective homomor-
phism GC → S8 . However as |S8 | = 40320 it would be rather messy appreciating the structure
of the image. If, instead, we consider the four diagonals of the cube

D1 = {1, 7} , D2 = {2, 8} , D3 = {3, 5} , D4 = {4, 6}

then we have a homomorphism

ρ : GC → Sym {D1 , D2 , D3 , D4 } = S4 .

This homomorphism is again injective. To see this we will show that its kernel is trivial.
Suppose that ρ(r) = e for some rotation r ∈ GC . Then r maps each Di to Di and hence maps
1 to either 1 or 7 and 2 maps to 2 or 8 etc.. If we assume that 1 → 7 then 2 (being an adjacent
vertex of 1) must map to 8 and similarly 3 → 5, 4 → 6. However the map

1 ↔ 7, 2 ↔ 8, 3 ↔ 5, 4↔6

is not a rotation (it is −I3 ). So any rotation fixing the Di must be the identity. As before we
now have an isomorphism ρ : GC → Im ρ S4 . From the Orbit-Stabilizer Theorem we know
that |GC | = 24 and hence
GC ∼ = Im ρ = S4 .
If we list the elements of GC we see that the rotations are:

identity : 1 of these.
rotation of ± π/2 about opposite faces’ midpoints : 6 of these.
rotation of π about opposite faces’ midpoints : 3 of these.
rotation of ± π/3 about a diagonal’ midpoints : 8 of these.
rotation of π about opposite edges’ midpoints : 6 of these.

and these descriptions respectively correspond to the conjugacy classes

{e} , 4-cycles, double transpositions, 3-cycles, 2-cycles.

Example 301 How many ways are there to colour the faces of a cube using n colours?

Solution. There are 6 faces and hence n6 colourings. For the vertex-to-vertex rotations, the
six faces split as two triples of faces (those adjacent to the vertices) which must be monochrome
— hence there are n2 such colourings fixed. For the mid-edge to mid-edge rotations, the six faces
split as three pairs and so there are n3 colourings fixed. For the mid-face to mid-face quarter

REPRESENTATIONS 85
turns they split as 1 + 1 + 4, so again there are n3 fixed colourings. Finally, for the mid-face to
mid-face half turns they split as 1 + 1 + 2 + 2, so there are n4 colourings fixed.

g fix(g) conjugates contribution

e n6 1 n6
vertex to vertex through ±2π/3 n2 8 8n2
mid-edge to mid-edge n3 6 6n3
3
mid-face to mid-face through ±π/2 n 6 6n3
mid-face to mid-face through π n4 3 3n4

So the number of essentially different colourings of a cube’s faces with n colours is

n6 + 3n4 + 12n3 + 8n2

.
24
When n = 2 compare with Example 267.

Example 302 How many ways are there to colour the edges of a tetrahedron black or white,
using equal numbers of each?

Solution. The tetrahedron has six edges and so there are 6 C3 = 20 different colourings. The
identity fixes all 20 of these. Given a rotation about an axis through a vertex and the opposite
face’s midpoint, then the six edges split as two triples than need to be of the same colour. Given
a rotation about an axis through a opposite mid-points of edges, the edges split as 1 + 1 + 2 + 2
so one of the 1s and one of the 2s need to be black.
g fix(g) conjugates contribution
6
e C3 = 20 1 20
vertex to face through ±2π/3 2 8 16
mid-edge to mid-edge 4 3 12

Thus the number of essentially different colourings is

20 + 16 + 12
= 4.
12

86 REPRESENTATIONS