i i

“book” — 2005/1/23 — 20:41 — page 132 — #146

i i

132 2. BASIC THEORY OF GROUPS

Definition 2.6.16. Let a and b be elements of a group G. We say

that b is conjugate to a if there is a g ∈ G such that b = gag −1 .

You are asked to show in the Exercises that conjugacy is an

equivalence relation and to find all the conjugacy equivalence classes
in several groups of small order.

Definition 2.6.17. The equivalence classes for conjugacy are called

conjugacy classes.

Note that the center of a group is related to the notion of conju-

gacy in the following way: The center consists of all elements whose
conjugacy class is a singleton. That is, g ∈ Z(G) ⇔ the conjugacy
class of g is {g}.

Exercises 2.6

2.6.1. Consider any surjective map f from a set X onto another set
Y . We can define a relation on X by x1 ∼ x2 if f (x1 ) = f (x2 ).
Check that this is an equivalence relation. Show that the associated
partition of X is the partition into “fibers” f −1 (y) for y ∈ Y .
The next several exercises concern conjugacy classes in a group.
2.6.2. Show that conjugacy of group elements is an equivalence re-
lation.
2.6.3. What are the conjugacy classes in S3 ?
2.6.4. What are the conjugacy classes in the symmetry group of the
square D4 ?
2.6.5. What are the conjugacy classes in the dihedral group D5 ?
2.6.6. Show that a subgroup is normal if, and only if, it is a union
of conjugacy classes.

2.7. Quotient Groups and Homomorphism Theorems

Consider the permutation group Sn with its normal subgroup of even
permutations. For the moment write E for the subgroup of even per-
mutations and O for the coset O = (12)E = E(12) consisting of odd

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 133 — #147
i i

2.7. QUOTIENT GROUPS AND HOMOMORPHISM THEOREMS 133

permutations. The subgroup E is the kernel of the sign homomor-

phism ǫ : Sn −→ {1, −1}.
Since the product of two permutations is even if, and only if,
both are even or both are odd, we have the following multiplication
table for the two cosets of E:
E O
E E O
O O E
The products are taken in the sense mentioned previously; namely
the product of two even permutations or two odd permutations is
even, and the product of an even permutation with an odd permu-
tation is odd. Thus the multiplication on the cosets of E reproduces
the multiplication on the group {1, −1}.
This is a general phenomenon: If N is a normal subgroup of a
group G, then the set G/N of left cosets of a N in G has the structure
of a group.

The Quotient Group Construction

Theorem 2.7.1. Let N be a normal subgroup of a group G. The set

of cosets G/N has a unique product that makes G/N a group and
that makes the quotient map π : G −→ G/N a group homomorphism.

Proof. Let A and B be elements of G/N (i.e., A and B are left cosets
of N in G). Let a ∈ A and b ∈ B (so A = aN and B = bN ). We
would like to define the product AB to be the left coset containing
ab, that is,
(aN )(bN ) = abN.
But we have to check that this makes sense (i.e., that the result is
independent of the choice of a ∈ A and of b ∈ B). So let a′ be
another element of aN and b′ another element of bN . We need to
check that abN = a′ b′ N , or, equivalently, that (ab)−1 (a′ b′ ) ∈ N . We
have

(ab)−1 (a′ b′ ) = b−1 a−1 a′ b′

= b−1 a−1 a′ (bb−1 )b′ = (b−1 a−1 a′ b)(b−1 b′ ).
Since aN = a′ N , and bN = b′ N , we have a−1 a′ ∈ N and b−1 b′ ∈ N .
Since N is normal, b−1 (a−1 a′ )b ∈ N . Therefore, the final expression
is a product of two elements of N , so is in N . This completes the
verification that the definition of the product on G/H makes sense.

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 134 — #148
i i

134 2. BASIC THEORY OF GROUPS

The associativity of the product on G/N follows from repeated

use of the definition of the product, and the associativity of the
product on G; namely

(aN bN )cN = abN cN = (ab)cN = a(bc)N

= aN bcN = aN (bN cN ).
It is clear that N itself serves as the identity for this multiplication
and that a−1 N is the inverse of aN . Thus G/N with this multipli-
cation is a group. Furthermore, π is a homomorphism because
π(ab) = abN = aN bN = π(a)π(b).
The uniqueness of the product follows simply from the surjectiv-
ity of π: in order for π to be a homomorphism, it is necessary that
aN bN = abN . ■

The group G/N is called the quotient group of G by N . The

map π : G → G/N is called the quotient homomorphism. Another
approach to defining the product in G/N is developed in Exercise
2.7.2.

Example 2.7.2. (Finite cyclic groups as quotients of Z). The

construction of Zn in Section 1.7 is an example of the quotient group
construction. The (normal) subgroup in the construction is nZ =
{ℓn : ℓ ∈ Z}. The cosets of nZ in Z are of the form k + nZ = [k]; the
distinct cosets are [0] = nZ, [1] = 1 + nZ, . . . , [n − 1] = n − 1 + nZ.
The product (sum) of two cosets is [a]+[b] = [a+b]. So the group we
called Zn is precisely Z/nZ. The quotient homomorphism Z → Zn
is given by k 7→ [k].

Example 2.7.3. Now consider a cyclic group G of order n with

generator a. There is a homomorphism ϕ : Z → G of Z onto G
defined by ϕ(k) = ak . The kernel of this homomorphism is precisely
all multiples of n, the order of a; ker(ϕ) = nZ. I claim that ϕ
“induces” an isomorphism ϕ̃ : Zn → G, defined by ϕ̃([k]) = ak =
ϕ(k). It is necessary to check that this makes sense (i.e., that ϕ̃ is
well defined) because we have attempted to define the value of ϕ̃ on
a coset [k] in terms of a particular representative of the coset. Would
we get the same result if we took another representative, say k + 17n
instead of k? In fact, we would get the same answer: If [a] = [b],
then a − b ∈ nZ = ker(ϕ), and, therefore, ϕ(a) − ϕ(b) = ϕ(a − b) = 0.
Thus ϕ(a) = ϕ(b). This shows that the map ϕ̃ is well defined.

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 135 — #149
i i

2.7. QUOTIENT GROUPS AND HOMOMORPHISM THEOREMS 135

Next we have to check the homomorphism property of ϕ̃. This

property is valid because ϕ̃([a][b]) = ϕ̃([ab]) = ϕ(ab) = ϕ(a)ϕ(b) =
ϕ̃([a])ϕ̃([b]).
The homomorphism ϕ̃ has the same range as ϕ, so it is surjective.
It also has trivial kernel: If ϕ̃([k]) = 0, then ϕ(k) = 0, so k ∈ nZ =
[0], so [k] = [0]. Thus ϕ̃ is an isomorphism.

Example 2.7.4. Take the additive abelian group R as G and the

subgroup Z as N . Since R is abelian, all of its subgroups are normal,
and, in particular, Z is a normal subgroup.
The cosets of Z in R were considered in Exercise 2.5.11, where
you were asked to verify that the cosets are parameterized by the
set of real numbers t such that 0 ≤ t < 1. In fact, two real numbers
are in the same coset modulo Z precisely if they differ by an integer,
s ≡ t (mod Z) ⇔ s − t ∈ Z. For any real number t, let [[t]] denote
the greatest integer less than or equal to t. Then t − [[t]] ∈ [0, 1) and
t ≡ (t − [[t]]) (mod Z). On the other hand, no two real numbers
in [0, 1) are congruent modulo Z. Thus we have a bijection between
R/Z and [0, 1) which is given by [t] 7→ t − [[t]].
We get a more instructive geometric picture of the set R/Z of
cosets of R modulo Z if we take, instead of the half–open interval
[0, 1), the closed interval [0, 1] but identify the endpoints 0 and 1:
The picture is a circle of circumference 1. Actually we can take a
circle of any convenient size, and it is more convenient to take a circle
of radius 1, namely
{e2πit : t ∈ R} = {e2πit : 0 ≤ t < 1}.
So now we have bijections between set R/Z of cosets of R modulo
Z, the set [0, 1), and the unit circle T, given by
[t] 7→ t − [[t]] 7→ e2πit = e2πi(t−[[t]]) .
Let us write ϕ for the map t 7→ e2πit from R onto the unit circle,
and ϕ̃ for the map [t] 7→ ϕ(t) = e2πit . Our discussion shows that
ϕ̃ is well defined. We know that the unit circle T is itself a group,
and we recall that that the exponential map ϕ : R → T is a group
homomorphism, namely,
ϕ(s + t) = e2πi(s+t) = e2πis e2πit = ϕ(s)ϕ(t).
Furthermore, the kernel of ϕ is precisely Z.
We now have a good geometric picture of the quotient group R/Z
as a set, but we still have to discuss the group structure of R/Z. The
definition of the product (addition!) on R/Z is [t] + [s] = [t + s]. But
observe that
ϕ̃([s] + [t]) = ϕ̃([s + t]) = e2πi(s+t) = e2πis e2πit = ϕ̃(s)ϕ̃(t).

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 136 — #150
i i

136 2. BASIC THEORY OF GROUPS

Thus ϕ̃ is a group isomorphism from the quotient group R/Z to T.

Our work can be summarized in the following diagram, in which
all of the maps are group homomorphisms, and the map π is the
quotient map from R to R/Z.
ϕ qqqqq

R qqqqq
qqq
qqqq
T
qqqqqqqqqq
qqq
q

π ∼
= ϕ̃
qqqqq qqqqq
qqqqqqqq

R/Z

Example 2.7.5. Recall from Exercise 2.4.20 the “Ax + b” group or

affine group Aff(n) consisting of transformations of Rn of the form
x) = Ax
TA,bb (x x + b,
where A ∈ GL(n, R) and b ∈ Rn . Let N be the subset consisting of
the transformations TE,bb , where E is the identity transformation,
x) = x + b .
TE,bb (x
The composition rule in Aff(n) is
TA,bb TA′ ,bb′ = TAA′ ,Abb′ +bb .
The inverse of TA,bb is TA−1 ,−A−1b . N is a subgroup isomorphic to the
additive group Rn because
TE,bb TE,bb′ = TE,bb+bb′ ,
and N is normal. In fact,
−1
TA,bb TE,cc TA,b
b = TE,Acc .
Let us examine the condition for two elements TA,bb and TA′ ,bb′ to be
congruent modulo N . The condition is
TA−1 b′ TA,bb = TA′ −1 ,−A′ −1b ′ TA,bb = TA′ −1 A,A′ −1 (bb−bb′ ) ∈ N.
′ ,b

This is equivalent to A = A′ . Thus the class of TA,bb modulo N is

[TA,bb ] = {TA,bb′ : b ′ ∈ Rn }, and the cosets of N can be parameterized
by A ∈ GL(n). In fact, the map [TA,bb ] 7→ A is a bijection between
the set Aff(n)/N of cosets of Aff(n) modulo N and GL(n).
Let us write ϕ for the map ϕ : TA,bb 7→ A from Aff(n) to GL(n),
and ϕ̃ for the map ϕ̃ : [TA,bb ] 7→ A from Aff(n)/N to GL(n). The
map ϕ is a (surjective) homomorphism, because
ϕ(TA,bb TA′ ,bb′ ) = ϕ(TAA′ ,Abb′ +bb ) = AA′ = ϕ(TA,bb )ϕ(TA′ ,bb′ ),
and furthermore the kernel of ϕ is N .

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 137 — #151
i i

2.7. QUOTIENT GROUPS AND HOMOMORPHISM THEOREMS 137

The definition of the product in Aff(n)/N is

[TA,bb ][TA′ ,bb′ ] = [TA,bb TA′ ,bb′ ] = [TAA′ ,bb+Abb′ ].
It follows that
ϕ̃([TA,bb ][TA′ ,bb′ ]) = ϕ̃([TAA′ ,bb+Abb′ ]) = AA′ = ϕ̃([TA,bb ])ϕ̃([TA′ ,bb′ ]),
and, therefore, ϕ̃ is an isomorphism of groups.
We can summarize our findings in the diagram:
ϕ
Aff(n) qqqqq
qqqqq
qqqqq
qq
GL(n)
qqqqqqqqqq
qqq
q

π ∼
= ϕ̃
qqqqq qqqqq
qqqqqqqq

Aff(n)/N

Homomorphism Theorems
The features that we have noticed in the several examples are
quite general:

Theorem 2.7.6. (Homomorphism theorem). Let ϕ : G −→ G be

a surjective homomorphism with kernel N . Let π : G −→ G/N be
the quotient homomorphism. There is a group isomorphism ϕ̃ :
G/N −→ G satisfying ϕ̃ ◦ π = ϕ. (See the following diagram.)
ϕ
G
qqqqq

G
qqq
qqqq
qqqqq

qqqqqqqqqqq
qqq

π ∼
= ϕ̃
qqqqq qqqqq
qqqqqqq

G/N

Proof. There is only one possible way to define ϕ̃ so that it will

satisfy ϕ̃ ◦ π = ϕ, namely ϕ̃(aN ) = ϕ(a).
It is necessary to check that ϕ̃ is well-defined, i.e., that ϕ̃(aN )
does not depend on the choice of the representative of the coset aN .
Suppose that aN = bN ; we have to check that ϕ(a) = ϕ(b). But
aN = bN ⇔ b−1 a ∈ N = ker(ϕ)
⇔ e = ϕ(b−1 a) = ϕ(b)−1 ϕ(a)
⇔ ϕ(b) = ϕ(a).

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 138 — #152
i i

138 2. BASIC THEORY OF GROUPS

The same computation shows that ϕ̃ is injective. In fact,

ϕ̃(aN ) = ϕ̃(bN ) ⇒ ϕ(a) = ϕ(b)
⇒ aN = bN.
The surjectivity of ϕ̃ follows from that of ϕ, since ϕ = ϕ̃ ◦ π.
Finally, ϕ̃ is a homomorphism because
ϕ̃(aN bN ) = ϕ̃(abN ) = ϕ(ab) = ϕ(a)ϕ(b) = ϕ̃(aN )ϕ̃(bN ).
■

A slightly different proof is suggested in Exercise 2.7.1.

The two theorems (Theorems 2.7.1 and 2.7.6 ) say that normal
subgroups and (surjective) homomorphisms are two sides of one coin:
Given a normal subgroup N , there is a surjective homomorphism
with N as kernel, and, on the other hand, a surjective homomor-
phism is essentially determined by its kernel.
Theorem 2.7.6 also reveals the best way to understand a quotient
group G/N . The best way is to find a natural model, namely some
naturally defined group G together with a surjective homomorphism
ϕ : G → G with kernel N . Then, according to the theorem, G/N ∼ =
Ḡ. With this in mind, we take another look at the examples given
above, as well as several more examples.

Example 2.7.7. Let a be an element of order n in a group H.

There is a homomorphism ϕ : Z → H given by k 7→ ak . This
homomorphism has range hai and kernel nZ. Therefore, by the ho-
momorphism theorem, Z/nZ ∼ = hai. In particular, if ζ = e2πi/n , then
ϕ(k) = ζ k induces an isomorphism of Z/nZ onto the group Cn of
nth roots of unity in C.

Example 2.7.8. The homomorphism ϕ : R → C given by ϕ(t) =

e2πit has range T and kernel Z. Thus by the homomorphism theorem,
R/Z ∼ = T.

Example 2.7.9. The map ϕ : Aff(n) → GL(n) defined by TA,bb 7→ A

is a surjective homomorphism with kernel N = {TE,bb : b ∈ Rn }.
Therefore, by the homomorphism theorem, Aff(n)/N ∼
= GL(n).

Example 2.7.10. The set SL(n, R) of matrices of determinant 1

is a normal subgroup of GL(n, R). In fact, SL(n, R) is the kernel
of the homomorphism det : GL(n, R) → R∗ , and this implies that
SL(n, R) is a normal subgroup. It also implies that the quotient
group GL(n, R)/SL(n, R) is naturally isomorphic to R∗ .

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 139 — #153
i i

2.7. QUOTIENT GROUPS AND HOMOMORPHISM THEOREMS 139

Example 2.7.11. Consider G = GL(n, R), the group of n-by-n

invertible matrices. Set Z = G ∩ RE, the set of invertible scalar
matrices. Then Z is evidently a normal subgroup of G, and is, in
fact, the center of G. A coset of Z in G has the form [A] = AZ =
{λA : λ ∈ R∗ }, the set of all nonzero multiples of the invertible
matrix A; two matrices A and B are equivalent modulo Z precisely
if one is a scalar multiple of the other. By our general construction
of quotient groups, we can form G/Z, whose elements are cosets of
Z in G, with the product [A][B] = [AB]. G/Z is called the projective
linear group.
The rest of this example is fairly difficult, and it might be best
to skip it on the first reading. We would like to find some natural
realization or model of the quotient group. Now a natural model for
a group is generally as a group of transformations of something or the
other, so we would have to look for some objects which are naturally
transformed not by matrices but rather by matrices modulo scalar
multiples.
At least two natural models are available for G/Z. One is as
transformations of projective (n − 1)–dimensional space Pn−1 , and
the other is as transformations of G itself.
Projective (n−1)–dimensional space consists of n-vectors modulo
scalar multiplication. More precisely, we define an equivalence rela-
tion ∼ on the set Rn \ {00} of nonzero vectors in Rn by x ∼ y if there
is a nonzero scalar λ such that x = λyy . Then Pn−1 = (Rn \ {00})/ ∼,
the set of equivalence classes of vectors. There is another picture
of Pn−1 that is a little easier to visualize; every nonzero vector x is
equivalent to the unit vector x /||xx||, and furthermore two unit vec-
tors a and b are equivalent if and only if a = ±bb; therefore, Pn−1
is also realized as S n−1 /±, the unit sphere in n–dimensional space,
modulo equivalence of antipodal points. Write [x x] for the class of a
nonzero vector x .
There is a homomorphism of G into Sym(Pn−1 ), the group of
invertible maps from Pn−1 to Pn−1 , defined by ϕ(A)([x]) = [Ax]; we
have to check, as usual, that ϕ(A) is a well-defined transformation
of Pn−1 and that ϕ is a homomorphism. I leave this as an exercise.
What is the kernel of ϕ? It is precisely the invertible scalar matrices
Z. We have ϕ(G) ∼ = G/Z, by the homomorphism theorem, and thus
G/Z has been identified as a group of transformations of projective
space.
A second model for G/Z is developed in Exercise 2.7.6, as a group
of transformations of G itself.

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 140 — #154
i i

140 2. BASIC THEORY OF GROUPS

Everything in this example works in exactly the same way when

R is replaced by C. When moreover n = 2, there is a natural realiza-
tion of GL(n, C)/Z as “fractional linear transformations” of C. For
this, see Exercise 2.7.5.

Proposition 2.7.12. (Correspondence of subgroups) Let ϕ : G −→

G be a homomorphism of G onto G, and let N denote the kernel of
ϕ.
(a) The map B 7→ ϕ−1 (B) is a bijection between subgroups of
G and subgroups of G containing N .
(b) Under this bijection, normal subgroups of G correspond to
normal subgroups of G.

Proof. For each subgroup B of G, ϕ−1 (B) is a subgroup of G by

Proposition 2.4.12, and furthermore ϕ−1 (B) ⊇ ϕ−1 {e} = N .
To prove (a), we show that the map A 7→ ϕ(A) is the inverse of
the map B 7→ ϕ−1 (B). If B is a subgroup of G, then ϕ(ϕ−1 (B))
is a subgroup of G, that a priori is contained in B. But since ϕ is
surjective, B = ϕ(ϕ−1 (B)).
For a subgroup A of G containing N , ϕ−1 (ϕ(A)) is a subgroup of
G which a priori contains A. If x is in that subgroup, then there is an
a ∈ A such that ϕ(x) = ϕ(a). This is equivalent to a−1 x ∈ ker(ϕ) =
N . Hence, x ∈ aN ⊆ aA = A. This shows that ϕ−1 (ϕ(A)) = A,
which completes the proof of part (a).
Let B = ϕ−1 (B). For part (b), we have to show that B is normal
in G if, and only if, B is normal in G.
Suppose B is normal in G. Let g ∈ G and x ∈ B. Then
ϕ(gxg −1 ) = ϕ(g)ϕ(x)ϕ(g)−1 ∈ B,
because ϕ(x) ∈ B, and B is normal in G. But this means that
gxg −1 ∈ ϕ−1 (B) = B, and thus B is normal in G.
Conversely, suppose B is normal in G. For ḡ ∈ G and x̄ ∈ B,
there exist g ∈ G and x ∈ B such that ϕ(g) = ḡ and ϕ(x) = x̄.
Therefore,
ḡx̄ḡ −1 = ϕ(gxg −1 ).
But gxg −1 ∈ B, by normality of B, so ḡx̄ḡ −1 ∈ ϕ(B) = B. There-
fore, B is normal in G. ■

Proposition 2.7.13. Let ϕ : G −→ G be a surjective homomor-

phism with kernel N . Let K be a normal subgroup of G and let K =
ϕ−1 (K). Then G/K ∼ = G/K. Equivalently, G/K ∼ = (G/N )/(K/N ).

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 141 — #155
i i

2.7. QUOTIENT GROUPS AND HOMOMORPHISM THEOREMS 141

Proof. Write ψ for the quotient homomorphism ψ : G −→ G/K.

Then ψ ◦ ϕ : G −→ G/K is a surjective homomorphism, because it
is a composition of surjective homomorphisms. The kernel of ψ ◦ ϕ is
the set of x ∈ G such that ϕ(x) ∈ ker(ψ) = K; that is, ker(ψ ◦ ϕ) =
ϕ−1 (K) = K. According to the homomorphism theorem, Theorem
2.7.6,
G/K ∼ = G/ ker(ψ ◦ ϕ) = G/K.
More explicitly, the isomorphism G/K −→ G/K is
xK 7→ ψ ◦ ϕ(x) = ϕ(x)K.
Using the homomorphism theorem again, we can identity G with
G/N . This identification carries K to the image of K in G/N ,
namely K/N . Therefore,
(G/N )/(K/N ) ∼ = G/K ∼= G/K.
■

The following is a very useful generalization of the homomor-

phism theorem.

Proposition 2.7.14. Let ϕ : G → G be a surjective homomorphism

of groups with kernel K. Let N ⊆ K be a subgroup that is normal
in G, and let π : G → G/N denote the quotient map. Then there is
a surjective homomorphism ϕ̃ : G/N → G such that ϕ̃ ◦ π = ϕ. (See
the following diagram.) The kernel of ϕ̃ is K/N ⊆ G/N .
ϕ
G
qqqqq

G
qqq
qqqq
qqqqq

qqqqqqqqqq
qqq
q

π ϕ̃
qqqqq qqqqq
qqqqqqqq

G/N

Proof. Let us remark that the conclusion follows from Proposition

2.7.13 and the homomorphism theorem. The map ϕ̃ is
G/N −→ (G/N )/(K/N ) ∼ = G/K ∼ = G.
However, it is more transparent to prove the result from scratch,
following the model of the homomorphism theorem.
As in the proof of the homomorphism theorem, there is only one
way to define ϕ̃ consistent with the requirement that ϕ̃ ◦ π = ϕ,
namely ϕ̃(aN ) = ϕ(a). It is necessary to check that this is well

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 142 — #156
i i

142 2. BASIC THEORY OF GROUPS

defined and a homomorphism. But if aN = bN , then b−1 a ∈ N ⊆

K = ker(ϕ), so ϕ(b−1 a) = e, or ϕ(a) = ϕ(b). This shows that the
map ϕ̃ is well defined. The homomorphism property follows as in
the proof of the homomorphism theorem. ■

Corollary 2.7.15. Let N ⊆ K ⊆ G be subgroups with both N and

K normal in G. Then xN 7→ xK defines a homomorphism of G/N
onto G/K with kernel K/N .

Proof. The statement is the special case of the Proposition with

G = G/K and ϕ : G → G/K the quotient map. Notice that applying
the homomorphism theorem again gives us the isomorphism
(G/N )/(K/N ) = ∼ G/K.

Example 2.7.16. What are all the subgroups of Zn ? Since Zn =

Z/nZ, the subgroups of Zn correspond one to one with subgroups of
Z containing the kernel of the quotient map ϕ : Z → Z/nZ, namely
nZ. But the subgroups of Z are cyclic and of the form kZ for some
k ∈ Z. So when does kZ contain nZ? Precisely when n ∈ kZ, or
when k divides n. Thus the subgroups of Zn correspond one to one
with positive integer divisors of n. The image of kZ in Zn is cyclic
with generator [k] and with order n/k.

Example 2.7.17. When is there a surjective homomorphism from

one cyclic group Zk to another cyclic group Zℓ ?
Suppose first that ψ : Zk → Zℓ is a surjective homomorphism
such that ψ[1] = [1]. Let ϕk and ϕℓ be the natural quotient maps of
Z onto Zk and Zℓ respectively. We have maps
ϕ
k ψ
Z−→Zk −→Zℓ ,

and ψ ◦ ϕk is a surjective homomorphism of Z onto Zℓ such that

ψ ◦ ϕk (1) = [1]; therefore, ψ ◦ ϕk = ϕℓ . But then the kernel of ϕk
is contained in the kernel of ϕℓ , which is to say that every integer
multiple of k is divisible by ℓ. In particular, k is divisible by ℓ.
The assumption that ψ[1] = [1] is not essential and can be elimi-
nated as follows: Suppose that ψ : Zk → Zℓ is a surjective homomor-
phism with ψ([1]) = [a]. The cyclic subgroup group generated by [a]
is all of Zℓ , and in particular [a] has order ℓ. Thus there is a sur-
jective homomorphism Z → Zℓ defined by n 7→ [na], with kernel ℓZ.

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 143 — #157
i i

2.7. QUOTIENT GROUPS AND HOMOMORPHISM THEOREMS 143

It follows from the homomorphism theorem that there is an isomor-

phism θ : Zℓ → Zℓ such that θ([1]) = [a]. But then θ−1 ◦ ψ : Zk → Zℓ
is a surjective homomorphism such that θ−1 ◦ψ([1]) = θ−1 ([a]) = [1].
It follows that k is divisible by ℓ.
Conversely, if k is divisible by ℓ, then kZ ⊆ ℓZ ⊆ Z. Since Z is
abelian, all subgroups are normal, and by the corollary, there is a
surjective homomorphism Zk → Zℓ such that [1] 7→ [1].
We conclude that there is a surjective homomorphism from Zk
to Zl if, and only if, ℓ divides k.

Proposition 2.7.18. Let ϕ : G −→ Ḡ be a surjective homomor-

phism with kernel N . Let A be a subgroup of G. Then
(a) ϕ−1 (ϕ(A)) = AN = {an : a ∈ A and n ∈ N },
(b) AN is a subgroup of G containing N .
(c) AN/N ∼ = ϕ(A) ∼= A/(A ∩ N ).

Proof. Let x ∈ G. Then

x ∈ ϕ−1 (ϕ(A)) ⇔ there exists a ∈ A such that ϕ(x) = ϕ(a)
⇔ there exists a ∈ A such that x ∈ aN
⇔ x ∈ AN.
Thus, AN = ϕ−1 (ϕ(A)), which, by Proposition 2.7.12, is a subgroup
of G containing N . Now applying Theorem 2.7.6 to the restriction
of ϕ to AN gives the isomorphism AN/N ∼ = ϕ(AN ) = ϕ(A). On the
other hand, applying the theorem to the restriction of ϕ to A gives
A/(A ∩ N ) ∼
= ϕ(A). ■

Example 2.7.19. Let G be the symmetry group of the square, which

is generated by elements r and j satisfying r4 = e = j 2 and jrj =
r−1 . Let N be the subgroup {e, r2 }; then N is normal because
jr2 j = r−2 = r2 . What is G/N ? The group G/N has order 4
and is generated by two commuting elements rN and jN each of
order 2. (Note that rN and jN commute because rN = r−1 N , and
jr−1 = rj, so jrN = jr−1 N = rjN .) Hence, G/N is isomorphic to
the group V of symmetries of the rectangle. Let A = {e, j}. Then
AN is a four–element subgroup of G (also isomorphic to V) and
AN/N = {N, jN } ∼ = Z2 . On the other hand, A ∩ N = {e}, so
A/(A ∩ N ) ∼=A∼ = Z2 .

Example 2.7.20. Let G = GL(n, C), the group of n-by-n invertible

complex matrices. Let Z be the subgroup of invertible scalar matri-
ces. G/Z is the complex projective linear group. Let A = SL(n, C).

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 144 — #158
i i

144 2. BASIC THEORY OF GROUPS

Then AZ = G because for any invertible matrix X, we have X =

det(X)X ′ , where X ′ = det(X)−1 X ∈ A. On the other hand, A ∩ Z
is the group of invertible scalar matrices with determinant 1; such a
matrix must have the form ζE where ζ is an nth root of unity in C.
We have G/Z = AZ/Z = A/(A∩Z) = A/{ζE : ζ is a root of unity}.
The same holds with C replaced by R, and here the result is
more striking, because R contains few roots of unity. If n is odd, the
only nth root of unity in R is 1, so we see that the projective linear
group is isomorphic to SL(n, R). On the other hand, if n is even,
then −1 is also an nth root of unity and the projective linear group
is isomorphic to SL(n, R)/{±1E}.

Exercises 2.7

2.7.1. Let ϕ : G −→ G be a surjective homomorphism with kernel

N . Let π : G −→ G/N be the quotient homomorphism. Show that
for x, y ∈ G, x ∼ϕ y ⇔ x ∼π y ⇔ x ∼N y. Conclude that the
map ϕ̃ : G/N −→ G defined by ϕ̃(aN ) = ϕ(a) is well defined and
bijective.
2.7.2. Here is a different approach to the definition of the product
on G/N , where N is a normal subgroup of G.
(a) Define the product of arbitrary subsets A and B of G to be
{ab : a ∈ A and b ∈ B}.
Verify that this gives an associative product on subsets.
(b) Take A = aN and B = bN . Verify that the product AB in
the sense of part (a) is equal to abN . Your verification will
use that N is a normal subgroup of G.
(c) Observe that it follows from parts (a) and (b) that (aN )(bN ) =
abN is a well–defined, associative product on G/N .
2.7.3. Consider the affine group Aff(n) consisting of transformations
of Rn of the form TA,bb (x x + b (A ∈ GL(n, R) and b ∈ Rn ).
x) = Ax
(a) Show that the inverse of TA,bb is TA−1 ,−A−1b .
−1
(b) Show that TA,bb TE,cc TA,b
b = TE,Ac . Conclude that N = {TE,bb :
n
b ∈ R } is a normal subgroup of Aff(n).
2.7.4. Suppose G is finite. Verify that
|A| |N |
|AN | = .
|A ∩ N |

i i

i i
 i i
“book” — 2005/1/23 — 20:41 — page 145 — #159
i i

2.7. QUOTIENT GROUPS AND HOMOMORPHISM THEOREMS 145

2.7.5. Consider the set of fractional linear transformations of the

complex plane with ∞ adjoined, C ∪ {∞},
az + b
Ta,b;c,d (z) =
cz + d
a b
· ¸
where is an invertible 2-by-2 complex matrix. Show that this
c d
is a group of transformations and is isomorphic to GL(2, C)/Z(GL(2, C)).

2.7.6. Recall that an automorphism of a group G is a group isomor-

phism from G to G. Denote the set of all automorphisms of G by
Aut(G).
(a) Show that Aut(G) of G is also a group.
(b) Recall that for each g ∈ G, the map cg : G −→ G defined
by cg (x) = gxg −1 is an element of Aut(G). Show that the
map c : g 7→ cg is a homomorphism from G to Aut(G).
(c) Show that the kernel of the map c is Z(G).
(d) In general, the map c is not surjective. The image of c
is called the group of inner automorphisms and denoted
Int(G). Conclude that Int(G) ∼ = G/Z(G).
2.7.7. Let D4 denote the group of symmetries of the square, and N
the subgroup of rotations. Observe that N is normal and check that
D4 /N is isomorphic to the cyclic group of order 2.

2.7.8. Find out whether every automorphism of S3 is inner. Note

that any automorphism ϕ must permute the set of elements of or-
der 2, and an automorphism ϕ is completely determined by what it
does to order 2 elements, since all elements are products of 2–cycles.
Hence, there can be at most as many automorphisms of S3 as there
are permutations of the three–element set of 2–cycles, namely 6; that
is, |Aut(S3 )| ≤ 6. According to Exercises 2.5.13 and 2.7.6, how large
is Int(S3 )? What do you conclude?

2.7.9. Let G be a group and let C be the subgroup generated by

all elements of the form xyx−1 y −1 with x, y ∈ G. C is called the
commutator subgroup of G. Show that C is a normal subgroup and
that G/C is abelian. Show that if H is a normal subgroup of G such
that G/H is abelian, then H ⊇ C.

2.7.10. Show that any quotient of an abelian group is abelian.

2.7.11. Prove that if G/Z(G) is cyclic, then G is abelian.
2.7.12. Suppose G/Z(G) is abelian. Must G be abelian?

i i

i i