MAS 305 Algebraic Structures II

Notes 3 Autumn 2006

Group homomorphisms
Definition Let G and H be groups, and let φ: G → H be a function. Then φ is a
homomorphism if (g1 g2 )φ = (g1 φ)(g2 φ) for all g1 , g2 in G.

Note that we are writing homomorphisms as functions on the right. Also note that
multiplication on the left-hand side of the equation is in G, while multiplication on the
right-hand side is in H.
If φ is a homomorphism then 1G φ = 1H and g−1 φ = (gφ)−1 (proof: exercise).
A homomorphism which is a bijection is called an isomorphism. If there is an
isomorphism from G to H then G is isomorphic to H, written G ∼ = H.

Definition If φ: G → H is a homomorphism then the image of φ is {gφ : g ∈ G}, which

is written Im(φ), and the kernel of φ is {g ∈ G : gφ = 1H }, which is written as ker(φ).

Theorem If φ: G → H is a homomorphism then

(a) Im(φ) is a subgroup of H;

(b) ker(φ) is a normal subgroup of G;

(c) g1 φ = g2 φ if and only if g1 and g2 are in the same coset of ker(φ), so there is a
bijection between the set of cosets of ker(φ) in G and Im(φ).

Proof (a) 1H = 1G φ ∈ Im(φ), so Im(φ) is not empty.

If h1 and h2 are in Im(φ) then there are g1 , g2 in G with h1 = g1 φ and h2 = g2 φ.
Then h−1 −1 −1 −1
1 h2 = (g1 φ) (g2 φ) = (g1 )φ(g2 φ) = (g1 g2 )φ ∈ Im(φ).
Thus Im(φ) 6 H.

1
 (b) 1G φ = 1H so 1G ∈ ker(φ), so ker(φ) is not empty.
If g1 , g2 are in ker(φ) then g1 φ = g2 φ = 1H so g−1 −1 −1
1 φ = 1H = 1H and (g1 g2 )φ =
(g−1 2 −1
1 )φ(g2 φ) = 1H = 1H , so g1 g2 ∈ ker(φ).
This shows that ker(φ) 6 G.
If g ∈ ker(φ) and x ∈ G then (x−1 gx)φ = (x−1 )φ(gφ)(xφ) = (xφ)−1 1H (xφ) = 1H
so x−1 gx ∈ ker(φ). Therefore ker(φ) E G.

(c)

g1 φ = g2 φ ⇐⇒ (g2 φ)(g1 φ)−1 = 1H

⇐⇒ (g2 φ)(g−1
1 φ) = 1H
⇐⇒ (g2 g−1
1 )φ = 1H
⇐⇒ g2 g−1
1 ∈ ker(φ)
⇐⇒ g1 and g2 are in the same right coset of ker(φ). 

Part (c) of this theorem gives the bijection between right cosets of ker(φ) in G and
elements of Im(φ) in H shown below. We know how to combine the latter using the
group operation in H. Using the bijection, this gives a way of combining right cosets
of ker(φ) in G. We shall now show that we can define a group operation on the right
cosets of a normal subgroup of G without knowing anything about another group H.

G H

ker(φ) - u1H

ker(φ)x - uh
1

ker(φ)y - uh
2

.. ..
. .

Definition If K E G then the quotient group G/K has as its elements the cosets of K
in G, with binary operation ◦ defined by Kx ◦ Ky = Kxy for x, y in G.

2
 Note that, since K is normal in G, we do not need to specify whether the cosets are
right cosets or left cosets.

Theorem The above definition of ◦ is well defined, and (G/K, ◦) is indeed a group.

Proof First we show that the definition of ◦ is not affected by which element of the
coset we use in the definition.
Suppose that Kx1 = Kx2 and Ky1 = Ky2 . Then x2 = kx1 and y2 = hy1 for some h,
k in K. Hence x2 y2 = kx1 hy1 = kx1 hx1−1 x1 y1 . But K is normal in G, so x1 hx1−1 ∈ K, so
kx1 hx1−1 ∈ K. Therefore Kx2 y2 = Kx1 y1 .
This shows that ◦ is well defined.
Next we have to check that the four group axioms are satisfied.

(a) Kxy is a coset of K, so ◦ is closed.

(b) Given x, y, z in G, we have (Kx ◦ Ky) ◦ Kz = (Kxy) ◦ Kz = K(xy)z = Kx(yz) =

Kx ◦ Kyz = Kx ◦ (Ky ◦ Kz), so ◦ is associative.

(c) K is the identity for ◦, because K = K1G and K ◦Kx = K1G ◦Kx = K(1G x) = Kx,
and, similarly, Kx ◦ K = Kx.

(d) The inverse of Kx is Kx−1 , because Kx ◦ Kx−1 = K(xx−1 ) = K1G = K and

Kx−1 ◦ Kx = K(x−1 x) = K1G = K.

Hence G/K is a group. 

Theorem Let K E G. The function θ: G → G/K defined by xθ = Kx is a homo-

morphism from G onto G/K (called the canonical homomorphism for K) and its kernel
is K.

Proof For all x, y in G, xθ◦yθ = Kx◦Ky = Kxy = (xy)θ, so θ is a homomorphism. It is

clear that θ is onto. Moreover, x ∈ ker(θ) ⇐⇒ xθ = K ⇐⇒ Kx = K ⇐⇒ x ∈ K. 

Permutation groups
Here we develop an important homomorphism and use it to find a normal subgroup of
Sn .

Definition A transposition is a permutation interchanging two objects and leaving the

rest fixed.

Lemma Every permutation is a product of transpositions.

3
Proof Every permutation is a product of cycles, and (1 2 3 . . . n) = (1 2)(1 3) . . . (1 n). 

Lemma If the permutation h contains the cycle (a1 a2 . . . an ) and the permutation g
takes ai to bi for i = 1, . . . , n then g−1 hg contains the cycle (b1 b2 . . . bn ).

Proof For 1 6 i 6 n − 1, bi (g−1 hg) = ai hg = ai+1 g = bi+1 , and bn g−1 hg = an hg =

a1 g = b1 . 

The following pictures illustrates the proof.

h = (a1 a2 . . . an ) . . .
g ↓ ↓ ... ↓ ...
−1
g hg = (b1 b2 . . . bn ) . . .

Definition The cycle structure of a permutation is the list of the lengths of its cycles,
in non-descending order.

The previous lemma shows that g−1 hg has the same cycle structure as h.

Example In S9 , if h = (1 4 5)(2 6 7 8) then the cycle structure of h is 1, 1, 3, 4. If

g = (1 6 3 4 9)(7 8 2) then g−1 hg = (6 9 5)(7 3 8 2), with the same cycle structure.

Definition In the symmetric group Sn , let c(g) denote the number of cycles of the
permutation g, including those cycles of length 1. Define sign(g) = (−1)n−c(g) . Say
that g is even if sign(g) = +1 and g is odd if sign(g) = −1.

Lemma Put m(g) equal to the number of cycles of g which have even length. Then
sign(g) = (−1)m(g) .

Proof For each permutation g, we have

n= ∑ (length of cycle).
cycles of g

So, modulo 2,
n = m(g) × 0 + (c(g) − m(g)) × 1,
so n − c(g) = −m(g) = m(g) mod 2, so (−1)n−c(g) = (−1)m(g) . 

Example In S9 , c(h) = 4, so sign(h) = (−1)9−4 = (−1)5 = −1. Also, m(h) = 1, so

sign(h) = (−1)1 = −1.

We are about to show that the function sign is a homomorphism. Untypically, this
homomorphism is written as a function on the left rather than on the right.

4
Theorem The function sign is a homomorphism from Sn to the group {1, −1} under
multiplication.
Proof We need to show that sign(gh) = sign(g) sign(h) for all permutations g and h.
First suppose that h is a transposition, say h = (1 2). Then sign(h) = −1. If 1 and 2
are in the same cycle of g then gh has two cycles in place of this, with the other cycles
the same as those of g. If 1 and 2 are in different cycles of g, then gh unites these two
cycles into one, leaving the other cycles the same as those of g. So c(gh) = c(g) ± 1,
so sign(gh) = (−1) sign(g) = sign(g) sign(h).
In general, suppose that h = h1 h2 . . . hs where the hi are transpositions. Then s − 1
applications of the previous step show that sign(h) = (−1) × (−1)s−1 = (−1)s and
s applications of the previous step show that sign(gh) = sign(g) × (−1)s . Hence
sign(gh) = sign(g) × sign(h). 
The kernel of sign is a normal subgroup of Sn , called the alternating group An . It
consists of all the even permutations. By a preceding theorem, the number of cosets
of An in Sn is equal to the size of Im(sign), which is equal to 2 if n > 2. That is,
|Sn : An | = 2, so |An | = n!/2 if n > 2.

Linear groups
Homomorphisms also give us a powerful method of finding subgroups of general lin-
ear groups. Here are two examples.  
1 n
First, consider the map φ from (Z, +) to GL(2, R) defined by nφ = . We
0 1
have     
1 n 1 m 1 n+m
(nφ)(mφ) = = = (n + m)φ
0 1 0 1 0 1
  
1 n
for all integers n and m, so φ is a homomorphism. Therefore : n ∈ Z is a
0 1
subgroup of GL(2, R).
Secondly, consider the determinant map det from GL(2, R) to (R \ {0} , ×), which
takes the 2 × 2 matrix M to its determinant det(M). It can be shown that det(AB) =
(det A)(det B) for all 2×2 real matrices A and B, so det is a homomorphism. Therefore,
the kernel of det is a normal subgroup of GL(2, R). This normal subgroup consists of
all the invertible real 2 × 2 matrices with determinant 1; it is called the special linear
group SL(2, R).
This can be generalized, by replacing R by any field F and replacing 2 by any
positive integer n. The determinant map is still a homomorphism, and so GL(n, F) has
a normal subgroup SL(n, F) consisting of the invertible n × n matrices with entries in
F and determinant 1.

5
Isomorphism theorems
The next four theorems are collectively known as ‘the isomorphism theorems’. Apart
from the first one, there is no universal agreement about which one is which, or which
results form part of which theorem.

First Isomorphism Theorem If G, H are groups and φ: G → H is a homomorphism,

then G/ ker(φ) ∼
= Im(φ).

Proof Put K = ker(φ). Define ψ: G/K → Im(φ) by (Kx)ψ = xφ. The theorem just
before the diagram shows that x and y are in the same coset of K if and only if xφ = yφ,
so ψ is well defined and one-to-one. Obviously, ψ is onto. Furthermore,

[(Kx)ψ][(Ky)ψ] = (xφ)(yφ)
= (xy)φ because φ is a homomorphism
= (Kxy)ψ
= [(Kx) ◦ (Ky)]ψ

and so ψ is a homomorphism. 

From now on we shall write Kx◦Ky just as (Kx)(Ky) when working in the quotient
group G/K.

Lemma If φ1 : G1 → G2 and φ2 : G2 → G3 are group homomorphisms then φ1 φ2 : G1 → G3

is a homomorphism.

Proof Exercise.

Correspondence Theorem Let K E G. There is a bijection between the set of sub-

groups of G containing K and the set of subgroups of G/K. The bijection preserves
inclusion. Normal subgroups of G containing K correspond to normal subgroups of
G/K.

Proof Let θ: G → G/K be the canonical homomorphism.

If K 6 H 6 G and h ∈ H then Kh ⊆ H, so H is a union of cosets of K. Let Hψ
be the set of these cosets, which is a subset of G/K. In fact, Hψ = {Kh : h ∈ H} =
{hθ : h ∈ H} = H/K. Since H is the union of the cosets in Hψ, it is clear that ψ is
one-to-one. It is also clear that if K 6 H1 6 H2 6 G then H1 ψ 6 H2 ψ, so ψ preserves
inclusion.
Define θ̃: H → G/K by hθ̃ = hθ for h in H. Then θ̃ is a homomorphism, so Im(θ̃)
is a subgroup of G/K. But Im(θ̃) = Hψ, so Hψ 6 G/K.

6
 If T 6 G/K, put T̄ = {g ∈ G : gθ ∈ T }. If g ∈ K then
gθ = K = identity in G/K ∈ T,
so K 6 T̄ and T̄ is not empty. If g1 and g2 are in T̄ then (g−1 −1
1 g2 )θ = (g1 θ) g2 θ ∈ T ,
because θ is a homomorphism, so g−1 1 g2 ∈ T̄ . Hence T̄ 6 G. If the coset Kg is in T
then g ∈ T̄ , so T = T̄ ψ. Hence ψ is onto, and so ψ is a bijection.
Finally, Hψ E G/K if and only if (Kg)−1 (Kh)(Kg) ⊆ H for all g in G and all h
in H. But (Kg)−1 (Kh)(Kg) = K(g−1 hg). Since K ⊆ H, K(g−1 hg) ⊆ H if and only
if g−1 hg ∈ H. Moreover, g−1 hg ∈ H for all g in G and all h in H if and only if
H E G. 
I like to draw a Hasse diagram to represent a group. There is one dot for each
subgroup that I am considering. If one subgroup is contained in another, there is a
line, or series of lines, from the first to the second, in a generally upwards direction,
possibly passing through other dots. The Correspondence Theorem says that the Hasse
diagram for G/K is just the part of the Hasse diagram for G that is above K.
G v - v G/K

H2 v - v H2 /K

H1 v - v H1 /K

K v - v

K/K = 1G/K = {K}

{1G } v

Second Isomorphism Theorem If K and N are normal subgroups of G with K 6 N

then (G/K)/(N/K) ∼
= G/N.
Proof Let θ be the canonical homomorphism from G onto G/K. The Correspon-
dence Theorem shows that N/K E G/K and so we can indeed form the quotient group
(G/K)/(N/K). Let φ be the canonical homomorphism from G/K onto (G/K)/(N/K).
Then θφ is a homomorphism from G onto (G/K)/(N/K), and
ker(θφ) = {g ∈ G : gθφ = identity in (G/K)/(N/K)}
= {g ∈ G : gθ ∈ ker(φ)}
= {g ∈ G : gθ ∈ N/K}
= N,

7
by the Correspondence Theorem.
Now the First Isomorphism Theorem gives

G/N = G/ ker(θφ) ∼
= Im(θφ) = (G/K)/(N/K). 

θ - v φ
G/N v  v G G/K - v (G/K)/(N/K)

v v v N/K v
 
1G/N N - - 1(G/K)/(N/K)

v v

K - 1G/K

v {1G }

This shows that it is legitimate to regard the part of the Hasse diagram for G that is
above N as the Hasse diagram for G/N whether or not we take the intermediate step
of regarding the part of the Hasse diagram for G that is above K as the Hasse diagram
for G/K.

Third Isomorphism Theorem If G is a group, K E G and H 6 G, and KH is defined

to be {kh : k ∈ K and h ∈ H} then
(a) KH is a subgroup of G containing K and H;

(b) K ∩ H is a normal subgroup of H;

(c) KH/K ∼
= H/K ∩ H.

Proof We could prove parts (i) and (ii) be elementary means, but instead we prove
them as a byproduct of proving part (iii).
Let θ be the canonical homomorphism from G onto G/K, and let θ̃ be the re-
striction of θ to H. Then Im(θ̃) is a subgroup of G/K. But Im(θ̃) = {Kh : h ∈ H},
which, under the Correspondence Theorem, corresponds to the following subgroup of
G containing K:

{g ∈ G : gθ = Kh for some h in H} = {g ∈ G : Kg = Kh for some h in H}

g ∈ G : gh−1 ∈ K for some h in H

=
= {g ∈ G : g = kh for some k in K and some h in H}
= KH.

8
So KH is a subgroup of G containing K, and Im(θ̃) = KH/K. Obviously, KH con-
tains H. This gives (i).
Moreover, ker(θ̃) = {h ∈ H : hθ = identity} = ker(θ) ∩ H = K ∩ H, so K ∩ H is a
normal subgroup of H. This gives (ii).
The First Isomorphism Theorem gives H/K ∩H = H/ ker(θ̃) ∼ = Im(θ̃) = KH/K. 

In any Hasse diagram, the convention is that if subgroups H and K are shown then
so also is K ∩ H, which is the largest subgroup contained in both of them. It is also
conventional to show the smallest subgroup containing them both, which is just KH if
either of them is normal in the whole group.
The following Hasse diagram depicts the Third Isomorphism Theorem. The parts
of the diagram bettween KH and K, and between H and H ∩ K, represent isomorphic
groups.

v G

v KH
@
@
@
H v @v K
@
@
@
@v K ∩H

v {1G }