An Introduction to

the
Theory of Groups
An Introduction to
the
Theory of Groups
GEORGE W. POLITES
Department of Mathematics
Illinois Wesleyan University
INTERNATIONAL TEXTBOOK COMPANY
Scranton, Pennsylvania
Copyright, 1968, by International Textbook Company
All rights reserved. No part of the material protected by this copyright notice may be repro-
duced or utilized in any form or by any means, electronic or mechanical, including photo-
copying, recording, or by any informational storage and retrieval system, without written
permission from the copyright owner. Printed in the United States of America. Library of
Congress Catalog Card Number: 68-31222.
Preface
It is to be emphasized that this book is intended as an introduction to the
theory of groups. It is hoped that a reading of this book will create a great deal
of interest in the subject of groups, and that the reader will prolong this interest
by continued reading in many of the fme books that go into the theory in much
more depth. It is to be emphasized, too, that familiarity with and a knowledge
of most of the material here is essential to a further study and understanding of
the more advanced topics and specialized material on group theory.
The primary motive for the writing of the book was to provide a text for
the study of groups by students engaged in independent study or honors work.
It should also do very nicely for use in a special-topics course in mathematics, or
in a mathematics seminar. It is directed mainly to the advanced undergraduate
student and should provide enough material for a three-credit-hour course. The
book also may have appeal for many graduate students as a reference, or as a
supplement to a text in abstract algebra.
Finally, it should be stressed that most of the exercises in this book are of
extreme importance and should be studied closely. Many results found later on
in the text depend strongly on exercises that may occur much earlier. The
examples also should be carefully noted, and the reader should try to construct
others in addition to solving the ones given.
Bloomington, Illinois
June, 1968
v
GEORGE W. POLITES
Contents
1. Introduction ... 1
1-1. Definition of a Group . . . . . . . . . . . . . . . . . . . . . . 1
1-2. Subgroups and Cosets . . . . . . . . . . . . . . . . . . . . . . 6
1-3. Cyclic Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1-4. Isomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1-5. Normal Subgroups . . . . . . . . . . . . . . . . . . . . . . . . 13
1-6. Homomorphisms and Quotient Groups . . . . . . . . . . 16
1-7. Automorphisms.. . . . . . . . . . . . . . . . . . . . . . . . . 21
2. Solvable Groups ... 25
2-1. Permutations. Alternating Groups. . . . . . . . . . . . . . 25
2-2. Simple Groups . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2-3. Composition Series. Solvable Groups. . . . . . . . . . . . 29
2-4. Direct Products 33
3. p-Groups ... 37
3-1. Converse of Lagrange's Theorem. Double Cosets . . . . 37
3-2. p-Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3-3. Sylow Theorems . . . . . . . . . . . . . . . . . . . . . . . . . 42
3-4. Nilpotent Groups . . . . . . . . . . . . . . . . . . . . . . . . 44
3-5. The Frattini Subgroup. . . . . . . . . . . . . . . . . . . . . . 47
4. Groups of Special Order . .. 49
4-1. Groups of Order 4 and 6 . . . . . . . . . . . . . . . . . . . . 49
4-2. Groups of Order pq and p
2
q. . . . . . . . . . . . . . . . . . 51
VII
Vlll
4-3. Groups of Order p
3
4-4. Groups of Order 16
Contents
53
55
5. Galois Theory ... 63
5-1. Defmition of a Field. Extension Fields . . . . . . . . . . 63
5-2. Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5-3. Splitting Fields . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5-4. The Galois Group of a Polynomial. Fundamental
Theorem of Galois Theory . . . . . . . . . . . . . . . . . . 70
5-5. Solvability by Radicals . . . . . . . . . . . . . . . . . . . . . 75
Index ... 79
1
Introduction
1-1. DEFINITION OF A GROUP
We begin by stating several definitions of a group and showing their equiva-
lence. By a mapping of a set A into a set B we shall mean a rule which associates
with each element a in A a unique element b in B. Iff is a mapping from A
into B, we shall indicate this by f:A-+B. If a is in A, f(a) shall denote the ele-
ment (image) in B that corresponds to a under the map f By a binary operation
on a setS we shall mean a mapping from S xS into S where S xS =\(a, b);
a and b are in S. (When we say that there is a binary operation defined on S,
it is to be understood that S is a nonempty set.)
Definition 1 Let G be a set of elements and * a binary operation defined on
the elements of G. We call G a group with respect to* if the following proper-
ties hold:
1. (a *b) *C =a *(b *C) for all a, b, and c belonging to G. ( * is associative.)
2. There exists an element e belonging to G such that a *e = e *a= a for
every a in G. e is called an identity.
3. For each a belonging to G there exists an element y in G such that
a*y = y *a= e. y is called an inverse of a.
Remark 1 It must be true in any group G that a* b belongs to G whenever a
and b belong to G. (G is closed with respect to*.)
Although a group is often defined as above, properties (2) and (3) need not
say quite as much. Consider the following definition.
Definition 2 Let G be a set of elements and * a binary operation defined on
the elements of G. We call G a group with respect to *if the following proper-
ties hold:
1. (a *b) *C =a *(b *c) for all a, b, and c belonging to G.
2. There exists an element e belonging to G such that e * a = a for every a
Introduction
in G. e is called a left identity.
3. For each a belonging to G there exists an element y in G such that
y *a= e. y is called a left inverse of a.
Remark 2 Parts (2) and (3) of Definition 2 may be referred to as "left" axioms.
We could just as well have stated "right" axioms.
Note: We shall usually write ab for a*b and refer to* as "multiplication."
Theorem 1 Definition 1 is equivalent to Definition 2.
Proof" Clearly, Definition 1 implies Definition 2. Suppose G is defined by
Definition 2. We must show that ea = a implies ae = a, and ya = e implies ay = e.
Suppose ya =e. Now y belongs to G so there exists x in G such that xy =e.
Thus ay=e(ay)=(xy)(ay)=x(ya)y=x(ey)=xy=e. Therefore ya=e im-
plies ay =e. Next, let a belong to G. Then a= ea = (ay)a = a(ya) = ae, and
ea =a implies ae =a. Therefore Definition 2 implies Definition 1.
Definition 3 Let G be a set of elements and * a binary operation defined on
the elements of G. We call G a group with respect to * if the following proper-
ties hold:
1. (a*b)*c =a *(b*c) for all a, b, and c belonging to G.
2. The equations ax= b and ya = b have solutions in G, where a and bare
arbitrary elements of G.
Theorem 2 Definition 1 is equivalent to Definition 3.
Proof" Definition 1 implies Definition 3 since, if ax = b and ya = b, we
have x = a-
1
b belongs to G andy= ba-
1
belongs to G. Suppose G is defined by
Definition 3. Let c belong to G and e denote a solution of the equation xc =c.
Then ec = c. Consider the equation ex =a, a an arbitrary element in G. Then
ea = e(cx) = (ec)x =ex= a. Therefore e is a left identity for G. Since the equa-
tion xa = e is solvable in G for every a in G, each a has a left inverse in G. Thus
Definition 2 is satisfied, and hence so is Definition 1 (Theorem 1).
Before discussing some examples of groups we note the following proper-
ties and definitions.
Theorem 3 The identity of a group is unique; the inverse of a group element is
unique.
Proof" Suppose e and e' are identities for a group G. Then e' = ee' (since e
is an identity)= e (since e' is an identity). Suppose y andy' are inverses of the
group element a. Theny = ye = y(ay') = (ya)y' = ey' = y'.
Note: We shall usually write a-
1
for the inverse of a (or -a if the opera-
tion is "addition").
Introduction 3
Theorem 4 Let G be a group and a and b belong to G. Then (abr
1
= b-
1

1
and (a-
1
rl = a.
Proof" (b-
1
a-
1
) (ab)=b-
1
(a-
1
a)b=b-
1
eb=b--
1
b=e, which implies b-
1

1
= (ab rl. a(a-
1
) = e implies a= (a-
1
rl.
Definition 4 Let G be a group and a belong to G. We define a
1
=a, a
2
= aa,
and, in general, if k is a positive integer such that ak has been defined, we define
ak+l =aka. We define a
0
= e and a-k = (a-
1
)k.
Note: Since e = ek = (aa-
1
)k = ak(a-
1
)k, (a and a-
1
commute), we see
that (a -l )k is the inverse of a". Thus (akrl = (a -I )k =a -k. The reader should
verify that am an = am +n and (am t =am n hold in a group.
Definition 5 If ab = ba for all ii and b belonging to G, a group, we call G an
Abelian (commutative) group.
We now give several examples of groups.
Example 1 The integers with respect to addition, the rationals (or reals) with
respect to addition, the nonzero rationals (or reals) with respect to multiplica-
tion, and the complex numbers with respect to addition, are all examples of
Abelian groups.
Example 2 The set of all nonsingular n by n matrices and the set of all orthog-
onal n by n matrices are groups with respect to matrix multiplication.
Example 3 Consider the symmetries of the square (Fig. 1 ). These are R, a 90
D
v
D'
4
3 2
Fig. 1
rotation clockwise around its center 0; R' and R", similar rotations through
180 and 270; I, a rotation about 0 of 360; H a reflection in the horizontal
Introduction
axis through 0; V a reflection in the vertical axis through 0; D and D', reflec-
tions in the diagonals.
We define a binary operation * on these symmetries as follows: X* Y shall
be the symmetry which achieves the same rearrangement of the square as is ob-
tained by first performing Y and then X Table 1 is an example of a * table (usu-
TABLE 1
*
R
R' R"
H v D n'
I I R
R' R"
H v D n'
R R
R' R"
I D n'
v H
R' R' R"
I R v H n'
D
R" R"
I R
R' n'
D H v
H H
n'
v D I
R' R"
R
v v D H
n' R'
I R
R"
D D H n' v R
R"
I
R'
n' n' v D H
R"
R
R'
I
ally called a multiplication table, or a Cayley table). H *R, for example, is the
entry in the row designated by Hand the column headed by R. Thus HR = D'.
Inspection of the table indeed shows that * is binary, that I is the identity,
and that each element has an inverse. The associative property follows from the
fact that (XY)Z and X(YZ) is in each case the symmetry obtained by first per-
forming Z and then Y and then X. Thus the symmetries of the square form a
group with respect to * It is clear, however, that it is not Abelian.
Remark 3 We may of course consider the symmetries of any regular polygon
of n ;;:;;, 3 sides and see that in each case we obtain a non-Abelian group with re-
spect to a binary operation * defined as in Example 3. (How many symmetries
are there corresponding to a fixed n?)
Example 4 SupposeS is a set containing n elements and denote by Sn the set
of all 1-1 mappings of S onto S. Such a mapping is called a permutation of S.
[Suppose f is a mapping from A to B and x andy belong to A. If f(x) = f(y) im-
plies that x = y, then f is said to be a one to one (1-1) mapping. If b is an arbi-
trary element of B and there is an a in A such that f(a) = b, we say that f is a
mapping from A onto B.] For f and gin Sn we define f*g as follows: if*g)
(x) = f(g(x )), for all x inS. It is left to the reader to verify that Sn, together with
*,is a group. (Sn is called the symmetric group on n symbols.)
Remark 4 One could discuss the set of 1-1 mappings of an infinite setS onto
itself, with * defined as above, and still obtain a group.
Example 5 As a special case of Example 4 we consider the following. Suppose
S is a set containing four elements, which we denote by 1, 2, 3, 4. Then S
4
will
Introduction 5
consist of 24 permutations as shown.
e 2 3 4)
pl = 1 2 3 4
(1 2 3 4)
p? = 1 3 4 2
c 2 3 4)
p13 = 4 2 3 1
c 2 3 4)
pl9 = 4 1 2 3
e 2 3 4)
p2 = 2 1 4 3
e 2 3 4)
Ps = 1 4 2 3
(1234)
pl4 = 3 1 4 2
c 2 3 4)
P2o = 3 4 2 1
e 2 3 4)
p3 = 3 4 1 2
c 2 3 4)
P9= 3241
(1 2 3 4)
P1s = 2 4 1 3
(1 2 3 4)
p2l = 1 2 4 3
(1234)
p4 = 4 3 2 1
(1 2 3 4)
p 10 = 4 2 1 3
(1 2 3 4)
p 16 = 1 3 2 4
(1234)
p22 = 3 2 1 4
(1 2 3 4)
Ps = 2 3 1 4
c 2 3 4)
Pu = 2 4 3 1
(1234)
pl? = 4 3 1 2
(1234)
p23 = 1 4 3 2
(1 2 3 4)
P6 = 3124
(1 2 3 4)
pl2 = 4 1 3 2
c 2 3 4)
P
18
= 2 3 4 1
(1 2 3 4)
p24 = 2 1 3 4
Here we are using a convenient notation to represent the 1-1 mappings of S
onto S. P
9
, for example, is the map which sends 1 onto 3, 2 onto 2, 3 onto 4,
and 4 onto 1. It is to be remembered that in performing the operation * on the
elements of S
4
, we first perform the map on the right, then the one on the left.
As an example,P
3
*P
9
=P
8

Table 2 is a partial (for obvious reasons) multiplication table for S
4
.
Clearly, S
4
is non-Abelian.
TABLE 2
pl p2 p3 p4 Ps
p6 p?
Ps
p9
Pw Pu
pl2 ... p24
pl pl p2 p3 p4 Ps
p6 p?
Ps
p9
Pw Pu
pl2
p2 p2 pl p4 p3
Ps Pw Pu Ps
pl2 p6 p? p9
p3 p3 p4 pl p2 pl2 p? p6 p9
Ps Pu Pw Ps
p4 p4 p3 p2
PI
p9
Pu Pw
pl2
Ps
p? p6
Ps
Ps Ps
p9
Ps
pl2 p6
PI
p2
Pu
p? p4 p3
Pw
p6 p6 p?
Pu Pw
pl
Ps
p9 p3 p2 p!2
Ps
p4
p? p? p6
Pw Pu
p3 p12 Ps PI
p4
Ps
p9 p2
Ps Ps PI2 Ps
p9
Pw
p2 pl p?
Pu
p3 p4 p6
p9 p9
Ps PI2 Ps Pu
p4 p3 p6
Pw PI
p2 p?
Pw Pw Pu
p? p6 p2
Ps PI2
p4 PI
p9
Ps
p3
Pu Pu Pw
p6 p? p4 p9
Ps
p2 p3
Ps
p!2 pl
p12 pl2
Ps
p9
Ps
p? p3 p4
Pw
p6 p2 pl
Pu
6 Introduction
Remark 5 Sn for n ;;;;, 3 is non-Abelian.
Proof" Let x
1
, x
2
, and x
3
be distinct elements in S and define the map-
pings f and g as follows: ftxt) = x
2
,Jtx2) = x
3
,ftx
3
) = x
1
,f(y) = y for any yin
S other than x
1
, x
2
, or x
3
g(xt) = x
1
,g(x2) =x
3
,g(x
3
) = x
2
,g(y) = y for any y
inS other than x
2
or x
3
. Clearly,[ andg belong to Sn. Now (fg)(xd =f(g(x
1
))
= ftx
1
) = X2, while (gf) (x
1
) = g(f(x
1
)) = g(x
2
) = x
3
.
Exercise 1 Prove that if x
2
= e for every x in G, a group, then G is Abelian.
1-2. SUBGROUPS
AND
COSETS
It may well be the case that a group contains a subset that meets all the re-
quirements to itself be a group. Clearly, je l is one such subset. This leads to
the following definition.
Definition 6 A subset H of a group G is a subgroup of G, if His itself a group
relative to the binary operation defined in G.
Actually, one need not explicitly show that every property (as stated, say,
in Definition 1) is satisfied by a subset H of a group G in order to show it is a
subgroup. This is the content of our next theorem (and remark).
Theorem 5 A subset H of a group G is a subgroup of G if, and only if, the fol-
lowing hold:
(i) if a and b belong to H, then ab belongs to H.
(ii) if h belongs to H, then h -J belongs to H.
Proof" Clearly, if H is a subgroup of G, then (i) and (ii) hold. Conversely,
suppose H is a subset of G such that a and b in H imply ab in H, and h in Him-
plies h -! in H. Now h and h -J in H imply hh -
1
= e in H (by (i)). Since * is asso-
ciative on G, it is associative on H. Therefore His a group.
Remark 6 (i) and (ii) above are equivalent to: (iii) if a and b are in H then
ab-
1
isinH.
Proof: Clearly, (i) and (ii) imply (iii). Suppose (iii) holds and that a and b
belong to H. a in H implies a
1
= e in H (by (iii)), and so eb-
1
= b-
1
belongs to
H. Thus (iii) implies (ii). Furthermore, a(b-
1
r
1
= ab belongs to H. Thus (iii)
implies (i).
Exercise 2 Suppose H and K are subgroups of a group G. Prove that H n K is
also a subgroup of G. (H n K denotes the set of elements common to both H
and K.)
Introduction 7
Definition 7 Let G be a group and H a subgroup of G. The set Ha = l ha; h in
H, a in G, a fixed l is called a right coset of H. (A left coset aH may be defined
similarly.)
Note: If the operation in G is "addition," then the set H +a= \h +a;
h in H, a in G, a fixed l would be a right coset of H.
Remark 7 Since He= H = eH, we see that His itself a right (left) coset of H.
Also, since e belongs to H, we have that a belongs to Ha (aH). Thus every ele-
ment of G is in some right (left) coset of H. We may regard a as the representa-
tive of Ha (aH).
Theorem 6 If HailHb =f- ,then Ha = Hb. (denotes the empty set.)
Proof" Suppose x belongs to HailHb. Then we may write x = h1 a and
x=h
2
b where h
1
and h
2
belong to H. Thus h
1
a=h
2
b and so a=h
1
-
1
h2b.
Therefore ha=h(h
1
-
1
h
2
b)=h
3
b where h
3
=hh
1
-
1
h
2
belongs to H. Thus
Hac;;.Hb. Similarly one shows Hbc;;.Ha. (We shall write Ac;;.B or B"d.A to indicate
that A is a subset of B. If A is a subset of B, but =f- B, we shall write ACE or
B ~ and call A a proper subset of B.)
Remark 8 The number of elements in any right (left) coset of His the same as
the number in H. We also note that (Har
1
= a-
1
Hand (bH)-
1
= Hb-
1
((Har
1
is
the set consisting of the inverses of all elements in the coset Ha.) Thus there is a
1-1 correspondence between right and left cosets of H.
Exercise 3 Prove that (Har
1
= a-
1
H.
To illustrate the concepts of subgroup and coset we consider the following
examples.
Example 6 The subsets K
1
= \I,Hl and K
2
= \J,R,R',R"l of the group of sym-
metries of the square (Example 3) are also subgroups. The right cosets of K
1
are
the sets l!,Hl, \R,D'l, \R',Vl, and \R",Dl, while the left cosets of K
1
are
ll,Hl, lR,DL \R',Vl, and lR",D'l. The right cosets of K
2
are the sets
\l,R,R',R"l and I H,D, V,D'l. These are also the left cosets of K
2
It is to be
noted that it is not always the case that the collection of right cosets of a sub-
group is the same as the collection of left cosets.
Example 7 Suppose G is the group of integers with respect to addition and H
the subgroup of G consisting of all multiples of 4. (The reader should show that
H is a subgroup.) Then the right cosets of H are the sets H,H + 1, H + 2, and
H + 3 (see note following Definition 7). Clearly, these are also the left co sets of
H.
Note: In the above example, we may of course consider the subset con-
sisting of all multiples of any fixed integer n.
8 Introduction
Definition 8 The order of a group G is the number of elements in G. A finite
group is a group of finite order. A group has infinite order if it does not have
finite order.
Remark 9 Since each elEment of G is in some right coset of H, we write G = H
+ Hx
2
+ Hx
3
+ to indicate that the co sets H, Hx
2
, Hx
3
, ... are disjoint and
exhaust G. (We take x
1
=e.) If the number of distinct right cosets of H in G is
r, then we have G =H + Hx2 + + Hxr.
Definition 9 The number r of distinct right (left) cosets of a subgroup H in a
group G is called the index of H in G and is denoted by [G:H].
Note: Since we may take H = je l; we see that the order of G may be de-
fined as the index of the identity subgroup.
Exercise 4 In a group of order 2n, n a positive integer, prove there exists an
element besides e which is its own inverse.
We now prove one of the most important theorems of group theory.
Theorem 7 (Lagrange) If G is a group of order n, the order of every subgroup
H of G is a divisor of n.
Proof" Suppose H has order t. We decompose G according to right cosets
of H obtaining G = H + Hx
2
+ Hx
3
+ + Hxr. Now each coset Hxi contains t
elements. Since these cosets are disjoint we have n = rt.
Note: The order of a group G is the product of the order of a subgroup
H and the index of H in G.
Theorem 8 Suppose G is a group, H a subgroup of G, and K a subgroup of H.
Then [G:K] = [G:H] [H:K].
Proof" This is immediate from Lagrange's Theorem.
1-3. CYCLIC GROUPS
Of special interest are the so-called cyclic groups. We introduce them here.
Definition 10 If a group G contains an element a such that every element of G
is of the form ak for some integer k (or of the form ka if the operation is+), we
say that G is a cyclic group. a is called a generator of G.
We distinguish two cases:
1. All powers ak are distinct. Then the cyclic group G = \ ,
2
, a-
1
,
a
0
, a, a
2
, I is infinite.
2. ah = ak, h > k. Here ah-k = e (h- k > 0). Let m be the smallest posi-
Introduction
9
tive exponent for which am =e. Then a
0
, a, a
2
, , am -
1
are all distinct. For
suppose at= a
8
(0..;;; s < t..;;; m- 1). Then at-s = e (0 < t- s < m), which contra-
dicts the assumption made on m. Now every integer n can be written in the
form n = qm + r, 0..;;; r < m (division algorithm). Thus an = aqm + r = aqm ff =
ear = ar. Therefore we see that all powers of a are contained in the sequence a
0
,
a, a
2
, , am-
1
and so G =I a
0
, a, a
2
, , am-I I is a cyclic group of order m.
Remark 10 If G is a group and a belongs toG, the setH= \ak;k an integer! is
a subgroup of G. Thus every element of a group generates a cyclic subgroup.
Defmition 11 The order of the cyclic group generated by a is called the order
of the element a. a is an element of infinite order if all powers of a are distinct.
The following are immediate consequences of Lagrange's Theorem.
1. If G is a finite group, the order of each element of G is a divisor of the
order of G.
2. A group of order p, p a prime, is a cyclic group. Every element =I= e is a
generator.
3. If G has order n, then an = e for every a in G.
Proof of (3): Suppose a belongs to G and the order of a ism. Then mk =
n. Now am =e. Thus an = amk =(am )k =e.
Note: Every cyclic group is Abelian since aid= i+i = J+i = aiai, for arbi-
trary integers i andj.
The following are examples of cyclic groups.
Example 8 The integers under addition are generated by 1.
Example 9 Let G = 11, -1, i, - i I and * be the usual multiplication for complex
numbers. i and -i both generate G. (i =A, so i
2
= -1.)
Example 10 The group K
2
(see Example 6) is generated by R. (Is this the
only generator?)
Example 11 Consider the fixed positive integer 11, and let a and b be any in-
tegers. We say that a is congruent to b modulo 11 (for this we will write a= b
(mod 11)) if a- b is divisible by 11. For example, 27 = 5 (mod 11) since 27- 5
is divisible by 11. One easily shows that every integer is congruent (mod 11) to
exactly one of the integers 0, 1, 2, ... , 10. Now let G = II, 2, ... , 10 I and de-
fine an operation * on G as follows: for a and b in G, a* b is the ordinary prod-
uct of a and b "reduced" modulo 11. For example, 3 * 9 = 5 (since 39 = 27 = 5
(mod 11)). The complete* table for G is given in Table 3. Inspection of theta-
ble shows G to be an Abelian group with respect to * Now G is also cyclic,
since 2 is a generator.
10
Introduction
TABLE 3
1 2 3 4 5 6 7 8 9 10
1 1 2 3 4 5 6 7 8 9 10
2 2 4 6 8 10 1 3 5 7 9
3 3 6 9 1 4 7 10 2 5 8
4 4 8 1 5 9 2 6 10 3 7
5 5 10 4 9 3 8 2 7 1 6
6 6 1 7 2 8 3 9 4 10 5
7 7 3 10 6 2 9 5 1 8 4
8 8 5 2 10 7 4 1 9 6 3
9 9 7 5 3 1 10
8 6 4 2
10 10 9 8 7 6 5 4 3 2 1
Exercise 5 Generalize Example 11 by replacing the fixed positive integer 1I by
any positive integer m > I. Under what conditions will the set G = !I, 2, ... ,
m - II be a cyclic group?
Remark 11 It is to be noted that for any fixed positive integer m, the set G =
l 0, I, 2, ... , m - II is always a group (in fact, cyclic) if the binary operation *
defined on G is as follows: a* b is the ordinary sum of a and b "reduced" modu-
lo m. Henceforth, a group of this type will be referred to as the additive group
of integers modulo m and shall be denoted by 1/(m).
One can observe that all subgroups of the cyclic groups described in the
above examples are also cyclic. That this is the case in any cyclic group is our
next theorem.
Theorem 9 A subgroup H of a cyclic group G is itself cyclic.
Proof" Suppose G is generated by a and let m be the smallest positive
integer such that am is in H. (We assume H * le\ .) Now H C G; thus any ele-
ment of His of the form ak for some integer k. Using the division algorithm, we
may write k=qm+r Thus ak=aqm+r=(am)q ar, which im-
plies ar =(am )-q ak. Now am belongs to Hand ak belongs to H. Therefore ar is
in H. Hence r = 0 and k = qm. Therefore every element of H has the form aqm
= (am )q. Thus H is cyclic generated by am .
Remark 12 In Theorem 9, suppose G has order n. Then an = e belongs to Hand
n = qm. Thus every subgroup H of G is generated by an element of the form am
where m divides n. Conversely, if the order of G is n = mq, H = l a
0
, aq, a
2
q,
... , atm-I Jq I is a cyclic subgroup of G having order m. Thus we see the exist-
ence of a subgroup of G having order m for any divisor m of the order of G.
Theorem 10 Let G be a group, G * l e l. Then G has no proper subgroups if,
and only if, G is a finite cyclic group of prime order.
Proof" The "if' part is immediate from Lagrange's Theorem. To prove
Introduction 11
"only if," let x =I= e belong to G. Then the cyclic group generated by xis all of
G. If x has infinite order then x
2
generates a proper subgroup. Thus x must
have finite order n. If n = rs, r > 1, s > 1, then xr generates a proper subgroup of
order s. Therefore n must be a prime.
Exercise 6 Suppose G is an Abelian group. Let a and b belong to G such that
the order of a is r and the order of b is s, where r and s are relatively prime.
Show that ab has order rs. (Two integers n and m are relatively prime if their
only common factors are 1 and -1. We write (n,m) = 1 to indicate that nand m
are relatively prime.)
Exercise 7 Suppose G is cyclic of order n with generator a. Show that ak is al-
so a generator of G if, and only if, k and n are relatively prime.
Remark 13 Suppose n is a positive integer and denote by cp (n) the number of
positive integers less than n that are relatively prime ton. (cp is known as the
Euler cp function.) Then Exercise 7 says that the number of generators for a cy-
clic group of order n is cp(n ). In particular, the number of generators for the
group of Example 11 is cjJ(lO) = 4. (Since 2 is a generator, so are 32, 72, and
92, by Exercise 7.)
Exercise 8 Suppose G is an Abelian group of order 6 containing an element of
order 3. Prove G is cyclic.
Exercise 9 Show that the elements of finite order in any commutative group G
form a subgroup.
Exercise 10 Prove that a group of order pm, p a prime, must contain a sub-
group of order p. (Groups of this type are of great interest and will be discussed
in Chapter 3.)
Exercise 11 Suppose G has only one element (say a) of order 2. Show xa =
ax for all x in G.
Exercise 12 Prove that every non-Abelian group contains proper subgroups.
1-4. ISOMORPHISMS
Of prime importance in mathematics is the concept of isomorphism. We de-
fine it here for groups.
Definition 12 Let G and G' be groups W1th respect to* and*' respectively. We
say that G is isomorphic to G' if there exists a 1-1 mapping rr of G onto G' such
that if rr(a) =a' and rr(b) = b', then rr(a*b) = a'*'b'. The mapping rr is called an
isomorphism of G onto G'.
12
Introduction
Example 12 Consider the set G' = ! 1,3 ,5 ,7 l and the operation *' defined on
G' to be ordinary multiplication "reduced" modulo 8. One easily verifies that
TABLE 4
rt:
3 5 7
3 5 7
3 1 7 5
5 7 1 3
7 5 3 1
G' together with *' is a group (see Table 4). Table 5 is the * table for the subset
K
3
= \I,H, V,R'I of the group of symmetries of the square (Example 3).
TABLES
I H v
R'
I I H v
R'
H H I
R'
v
v v
R'
I H
R' R'
v H I
That K3 is a group (subgroup of the group of symmetries of the square) fol-
lows from Theorem 5.
Now consider the mapping rr:K
3
-+ G' defined by rr(J) = 1, rr(H) = 3,
rr(V) = 5, and rr(R') = 7. Clearly, rr is 1-1 and onto. Thatrr(a*b) = rr(a) *'rr(b)
for arbitrary elements a and b in K
3
, is easily verified. Thus the groups K
3
and
G' are isomorphic and rr is an isomorphism. (Is rr the only isomorphism of K
3
onto G'?)
Example 13 The mapping x -+log x is an isomorphism of the group G of posi-
tive real numbers under multiplication onto the group G' of all real numbers
under addition.
Exercise 13 Prove that if rr is an isomorphism of a group G onto a group G',
then: (i) rr (e)= e' (e is the identity of G, e' the identity of G'); (ii) rr (a)= a' im-
plies rr(a -I) = (a'r
1
.
Exercise 14 Show that the group of Example 3 is isomorphic to a subgroup of
the group S
4
. (See Example 5.) Generalize this to the group of symmetries of
any regular polygon and the groups Sn.
Remark 14 With some thought, one should realize that isomorphic groups have
the same structure. Esser.tially, the only difference is the notation used to de-
Introduction
13
scribe the elements and the nature of the operations defined on these elements.
Our next theorem completely characterizes the cyclic groups.
Theorem 11 (i) Every cyclic group G of infinite order is isomorphic to the ad-
ditive group I of integers; (ii) Every cyclic group G of order n is isomorphic to
the additive group I /(n) of integers modulo n. (See Remark 11.)
Proof: For (i), define rr:G--+ I by rr(ak) = k, where a is a generator of
G. For (ii), define rr:G--+ I/(n) by rr(ak) = k, where a is a generator of G and k is
reduced modulo n.
We have seen that many groups may be regarded as permutation groups;
that is, as groups whose elements are permutations (Exercise 14). It is somewhat
remarkable that this is true for any group. We prove this now.
Theorem 12 (Cayley) Every group G is isomorphic to a permutation group.
Proof: Let a be a fixed element of G and consider the mapping rra :G--+ G
defined by rra(x)=ax. 11a is onto, since for yin G we have rra(a-
1
y)=y.
Clearly, rra is 1-1. Thus 11a is a permutation of G for each a. Let P be the set
consisting of the mappings rra. Since rrb-J = rrb-J (why?) we have, for rra and rrb
belonging toP, (rra rrb -
1
)(x) = 11a(rrb -
1
(x )) =rraW
1
x) = a(b-
1
x) = (ab-
1
)x = 11ab -!
(x), which implies that P is a group (Remark 6). Now consider the mapping
tp:G--+ P defined by tp(a) = rra If rra = rrb (that is, rra(x) = rrb(x) for all x in G),
then ax= bx, which implies that a= b and tp is 1-1. Since tp(ab) =nab= rrarrb =
tp(a) tp(b ), tp is an isomorphism. This concludes the proof.
1-5. NORMAL SUBGROUPS
In this section we shall introduce what is perhaps the most important type
of subgroup of a group. We begin with the following theorem.
Theorem 13 Let H be a subgroup of G and x belong to G. Then K =
1
Hx
=I x-
1
hx; h in ~ is a subgroup of G.
Proof: Let x-
1
h
1
x and x-
1
h
2
x belong to K. We will show that (x-
1
h
1
x) X
(x-
1
h
2
xr
1
belongs to K (see Remark 6). Now (x-
1
h
1
x)(x-
1
h
2
xr
1
= (x-
1
h
1
x) X
(x-
1
h
2
-
1
x) = x-
1
h
1
(xx-
1
) h
2
-
1
x =
1
h
1
h
2
-
1
x =
1
h
3
x, which is an element of
K (where h
3
= h
1
h
2
-
1
is in H).
Definition 13 If G =! x
1
,x
2
, .. I is a group and His a subgroup of G, the sub-
groups x
1
-
1
Hx
1
, x
2
-
1
Hx
2
, are called conjugates of H. (We shall also refer
to the setsx
1
-
1
Hx
1
,x
2
-
1
Hx
2
, . as conjugates of HwhenH is any subset of G.
Of particular interest are the conjugates of a single element, which we discuss in
a later section.)
Note: Since e belongs to G,e-
1
He= His a conjugate of H.
14
Introduction
Definition 14 If in Definition 13 we have H = x
1
-
1
Hx
1
= x
2
-
1
Hx
2
= , we
call H a normal (self-conjugate, invariant) subgroup of G. Thus His a normal
subgroup of G if x -
1
Hx = H for all x belonging to G. We write H 6. G to denote
His a normal subgroup of G.
Remark 15 Suppose
1
Hx c:;, H for all x in G. Then x-
1
hx belongs to H for all
x in G and h in H. If k belongs to H, then xk
1
= h also belongs to H. Thus
k = x-
1
hx is in x-
1
Hx and we have H c:;,x-
1
Hx. Hence H= x-
1
Hx for allx in G,
and soH 6. G. Therefore, to show His a normal subgroup of G, we need only
show x -
1
Hx c:;, H for all x in G.
You will recall that if H is a subgroup of a group G, the collection of right
co sets of H in G is not necessarily the same as the collection of left co sets of H
(Example 6). That this cannot happen when His normal is proved next.
Theorem 14 His a normal subgroup of G if, and only if, every right coset Hx
is also a left coset xH.
Proof: Suppose H 6. G. Then x-
1
Hx = H for all x in G. Thus Hx = xH.
Conversely, if Hx = yH, then x belongs to yH, which implies that xH = yH. Thus
Hx = xH (and hence x-
1
Hx =H) for all x in G. Therefore His normal.
Exercise 15 Suppose Hand K are normal subgroups of a group G. Prove that
H n K is also normal in G.
Example 14 Every subgroup of an Abelian group is normal. In particular,
every subgroup of a cyclic group is normal.
Example 15 In the group of symmetries of the square, we have already noted
that the subsets K
1
=II, Hl and K
3
=II, H, V, R'\ are subgroups. NowK
3
is
also a normal subgroup while K
1
is not (show this). However, K
1
is normal in
K3.
Remark 16 It is clear that if H, K, and G are groups such that H C K C G and
H 6. G, then H will also be normal in K. But suppose H C K C G and H 6. K.
When isH 6. G? Certainly not always, as is seen in Example 15. Also, when is
K 6. G? We shall have more to say on this later.
Example 16 The subset H = l P
1
, P
2
, , P
12
\is a normal subgroup of the
group S
4
(That H is a subgroup follows by inspecting the upper left corner of
Table 2.)
Definition IS Let G be a group and A a nonempty subset of G. We call A a
complex. If A and B are two complexes, by the product AB we mean the set of
all products ab, a in A and b in B.
Note: To say that AB = BA where A and B are complexes of G, does not
mean that ab = ba for all a in A and bin B; but rather that, given a
1
b
1
inAB,
Introduction
15
there exists b'a' (not necessarily b
1
a
1
) in BA such that a
1
b
1
= b'a', and con-
versely. In particular, if B consists of a single element x, to say that Ax = xA
does not mean ax = xa for all a in A.
Exercise 16 Suppose Hand K are subgroups of a group G. Show thatHK is a
subgroup of G if, and only if, HK = KH.
Exercise 17 Suppose G is a group, Hand K subgroups of G. Prove that HK is a
subgroup of G if either H or K is normal in G.
Exercise 18 In Exercise 17, suppose H and K are both normal in G. Prove that
HK is normal in G.
The preceding exercises give us conditions in order that HK be a subgroup
of a group G when Hand K are subgroups of G. But what about the order of
HK? In future discussions we shall need this information. We obtain it now.
Note: We shall henceforth denote the number of elements in a finite setS
by lSI.
Theorem 15 Suppose H and K are finite subgroups of a group G. Then IHKI =
IHIIKI
IH n Kl. (It should be noted here that we are not assuming that HK is a sub-
group of G.)
Proof: Suppose x belongs to H n K and hk belongs to HK. We may write
hk = hxx-
1
k. Now hx belongs to H (x does) and x-
1
k belongs to K (x-
1
does).
Thus the element hk appears in HK at least IH n Kl times. If hk = h
1
k
1
, then
kk
1
-
1
=h-
1
h
1
, which is an element of HnK. Therefore h
1
=hkk
1
-
1
andk
1
=
h
1
-
1
hk, which implies hk appears in HK at most IH n Kl times. Hence the ele-
ment hk appears in HK exactly IH n Kl times. It follows that the number of dis-
tinct elements in HK is the product of the number of elements in H and the
number of elements in K, divided by the number of times a given element
appears in HK.
Exercise 19 Let G be a group and Z(G) =\a in G; ax =xa for all x in G l.
Prove that Z(G) is a normal subgroup of G. (Z(G) is called the center of G.)
Exercise 20 Suppose G is a group and x andy belong to G. The element
1
y -I xy is called the commutator of x andy. Prove that the set G' of all finite
products of such commutators froms a normal subgroup of G. (G' is called the
commutator subgroup, or derived group of G.)
We have seen that if His a subgroup of a group G and xH = Hx for all x in G,
then H is a normal subgroup. Thus, in the case H is not normal in G, xH = Hx
does not hold for some element x in G. If we confine our attention to just those
16
Introduction
elements x of G for which xH and Hx are indeed equal, we are led to some very
interesting and very useful results. We do this here in a more general setting, in
that we merely assume H to be a complex of G.
Definition 16 Let G be a group and H a complex of G. The set of elements in
G that commute with H is called the normalizer of H in G and is denoted by
Nc(H).
Note: A subgroup H of a group G is normal in G if, and only if, Nc(H) =
G.
Theorem 16 Nc(H) is a subgroup of G. Furthermore, [G:Nc(H)] =the num-
ber of conjugates of H.
Proof" Let x andy belong toN c(H). Then Hx = xH and Hy = yH, or H =
xHx -I and H = yHy -
1
. Thus xHx -
1
= yHy -
1
, which implies that Hx -
1
y = x -
1
yH.
Therefore x -
1
y belongs toN c(H) and hence Nc(H) is a subgroup. Next, let D =
Nc(H), S = !Dx; x in G I, and T= I x-
1
Hx; x in G!. Definef:S-+ Tby f(Dx) =
x -
1
Hx. (The reader should show that f is well-defined.) f is clearly onto. We
show f is 1-1 as follows: Suppose f(Dx) = f(Dy); that is, suppose x -
1
Hx = y -
1
Hy.
Then y
1
H = Hyx -
1
. Thus yx -
1
belongs to D, implying y belongs to Dx. There-
fore Dx = Dy. Hence the number of conjugates of H is equal to the number of
right cosets of Nc(H), which is [G:Nc(H)].
Theorem 16 is one of extreme importance in obtaining future results and we
emphasize it with our next example.
Example 17 Consider again the group G of symmetries of the square (Example
3) and let K be the complex I R, H, D'l. We first determine N c(K). By inspect-
ing Table 1 we find that the only elements x in G for which xK = Kx are I and
R'. Therefore Nc(K) =II, R'l, which is clearly a subgroup of G. Next we de-
termine the left co sets of N c(K). These are the sets !I, R'l , I R, R "I, IH, V l
and I D, D'l . Finally, we see that the conjugates of K are { R, H, D'l, { R, V,
D i, l R ", H, D l, and I R ", V, D'l. Thus the number of conjugates of K is equal
to the number ofleft cosets of Nc(K), which of course is [G:Nc(K)].
Exercise 21 Let G be a group and a belong to G. Show that the normalizer of
a is a group containing the cyclic group generated by a as a normal subgroup.
1-6. HOMOMORPHISMS
AND
QUOTIENT GROUPS
In Definition 12 we introduced the concept of isomorphism. A much more
general notion is that of homomorphism. We define it here and develop several
important results in connection with normal subgroups.
Introduction
17
Definition 17 Let G and G be groups with respect to *and *'respectively. Let
<fJ be a mapping from G into G. If <fJ(a*b) = <P(a) *' <P(b) for all a and binG,
then we call <fJ a homomorphism. The setH= l x in G; <P(x) = x for some x in G(
is called the homomorphic image of Gunder <{J.
No'!: If <Pis a homomorphism of G onto G then H = G and we say that
G and G are homomorphic.
Example 18 For G and G any groups, the mapping <{J:G--->- G defined by <P(x) =
e, e the identity of G, is clearly a homomorphism.
Example 19 Any isomorphism of G onto G is a homomorphism of G onto G.
Example 20 Consider the following groups: S
4
, and G = \1,-11 (under multi-
plication). Define the mapping <[J:S4 --->- G by <P(Pa = 1, 1 .;;; i.;;; 12, and <P(Pj) =
-1, 13 .;;;j.;;; 24. S
4
and G are homomorphic under <fJ.
Theorem 17 Let <P be a homomorphism of G into G (G and G groups). Then
H, the homomorphic image of Gin G, is a subgroup of G.
Proof: Let a and b belong to Hand let a and b be elements of G such that
<P(a) = (i, <P(b) =b. Now <P(ab) belongs to H and <P(ab) = <P(a) <P(b) = a b ~ There-
fore ab is in H. Next, let e be the identity of G and x belong to H where <P(x) =
X, x in G. Then x = <P(x) = <fJ(ex) = <fJ(e)<fJ(x) =<fJ(e)x, which implies that <P(e) is
the identity of H. (We shall denote <P(e) by i?.) Finally, let jlbelong to H where
<P(Y) = Y, yin G. Then e = <P (e)= <P (yy -I)= <P (y )<P (y -
1
) = y <P(Y -
1
J, implying that
<PCY-I) = v r ~
Note: One could speak of a homomorphism <fJ from a group G into a setS
having defined on it a binary operation *
1
Here, as in Theorem 17, it is easily
shown that the homomorphic image of Gunder <fJ is a group. However, it is nec-
essary to verify the associative property in this case.
Exercise 22 Suppose <fJ is a homomorphism of G into G and lj; a homomor-
phism ofG into G. Prove that the mapping I/I<{J:G->-G defined by (I/I<P)(x)=
I/I(<P(x)) is a homomorphism of G into G. Also prove that if <fJ and lj; are both
onto, then so is the homomorphism lj; <fJ.
Definition 18 Suppose G is a group homomorphic to G under a homomor-
e_hism <{J. Let K be the set of elements kinG such that <P(k) = e, the identity of
G. K is called the kernel of the homomorphism <{J.
Example 21 The kernel of the homomorphism of Example 18 is all of G; of
Example 20, it is ! P
1
, P
2
, . , P
12
l.
Theorem 18 The kernel K of a homomorphism <fJ from a group G onto a group
G is a normal subgroup of G. Furthermore, two elements of G have the same
image in G if, and only if, they belong to the same right (left) coset of K.
18 Introduction
Proof: Suppose k
1
and k
2
belong to K. Then 10(k
1
) = e and IO(k
2
) =e.
Now IO(k
1
k
2
-
1
)=10(kt)IO(k
2
-
1
)=e[IO(k2)]-
1
= ~ e = i which implies k1k2-
1
belongs to K. Thus K is a subgroup of G. Next, let x belong toG and k belong
to K. Suppose 10(x)=a. Theni,O(x-
1
kx)=\O(X-
1
)10(k)IO(x)=a-rea=e. This
implies thatx-
1
kx is an element of K and so K is normal in G.
To complete the proof of the theorem, let x andy belong to G where 10(x) =
1,0(y) =a. Then 10(xy-
1
) = e, implying that xf
1
belongs to K. Thus x belongs to
Ky and x andy are in the same right coset of K. Conversely, if xis inKy, then
X= ky' k in K. Suppose IOCY) =a. Then \O(X) = IO(ky) = IO(k) IOCY) = ea =a, and
we see that x andy have the same image in G.
Exercise 23 Let K be the kernel of a homomorphism 1T of a group G onto a
group G'. If K = I e l , prove that 1T is an isomorphism.
The following example serves to introduce a whole new class of groups.
Example 22 Consider the subgroupsK
1
=II, H! andK
4
=II, R'l of the group
of symmetries of the square. In Example 6 we found the right cosets of K
1
to
TABLE 6
{I,H! {R,D'! {R',V! {R",Dl
/I,H! {I,H! {R,D'} {R',V/ {R",D/
[R,D'} IR, D, D', R"l {R',H, V,I l {R",D',D,R/ {I,V,H,R'/
{R',V! !R',V! {R",D} {I,Hl
{R,D'!
{R",D} {R",D',D,R} II, V,H,R'J {R,D,D',R" J {R',H, V,II
be {I, Hl, {R, D' l, ! R', Vl, and {R", Dl. The right cosets of K
4
are l /, R' I,
{R, R"!, {H, VI, and l D, D' l'. Table 6 gives the product of any two right co-
sets of K
1
, and Table 7 the product of any two right co sets of K4
TABLE 7
{I,R'/ {R,R"/ {H,V/ {D,D'l
{I,R'/ {I,R'! {R,R"!
!H,V/ {D,D'\
{R,R"} {R,R"J {R',IJ
{D,D'! {V,H/
{H,V} . {H,V}
{D',DJ
{I,R'}
{R",R!
{D,D'!
{D,D'}
{H,V}
{R,R"! {I,R'!
One notices immediately that the product of any two right cosets of K
4
is
again a right coset of K
4
But not so forK
1
As is seen in our next theorem, the
reason for this tremendous difference is the fact that K
4
is normal. The reader
may wish to verify that K
4
is the center for the group of symmetries of the
square. (See Exercise 19 .)
Introduction
19
Theorem 19 Let G be a group and N a subgroup of G. N is normal in G if, and
only if, the product of any two right cosets of N is a right coset of N.
Proof: Suppose N b. G and let Nx and Ny be right cosets of N. Then
NxNy = N(?cN)y = N(Nx )y = NNxy = Nxy. Conversely, suppose the product of
any two right cosets of N is again a right coset of N. Then for x in G, NxNx-
1
is a right coset, and it contains the element exex-
1
=e. Thus NxN
1
=N,
which implies that xNx -
1
c;;,.N, which implies that N is normal in G (Remark 15).
Remark 17 The equationNxNy = Nxy defines a binary operation on the collec-
tion of right co sets of a normal subgroup N. Remarkably, with this concept of
"product," these cosets form a group, as we show next.
Theorem 20 The cosets of any normal subgroup N of a group G form a group
with respect to multiplication of cosets as defined above.
Proof' Since (NxNy)Nz = NxyNz = N(xy)z = Nx(yz) = NxNyz =Nx(NyNz),
the product is associative. N = Ne is the left identity, since (Ne)(Nx) = Nex = Nx,
for all co sets Nx. Finally, we have N
1
Nx = N
1
x = Ne = N, which implies that
Nx -
1
is the left inverse of Nx.
Definition 19 The group of cosets of N is called the quotient group (factor
group) of G by Nand is denoted by G/N.
Thus the quotient group of the group of symmetries of the square by its
center is the collection of cosets of K
4
given in Example 22, and Table 7 is the
multiplication table for this quotient group.
Example 23 In S
4
, the subset N = ! P
1
,P
2
,P
3
,P 4 l is a normal subgroup of the
group G = l P
1
, P
2
, , P
12
1. The right cosets of N in G are N,A = l P
5
,P
8
,P
9
,
P
12
l, and B= !P
6
,P
7
,P
10
,P
1
d. Thus G/N= lN,A,BI. (The reader should
verify all of this.)
Note: IG/N I= [ G:N].
A problem of interest is the following: when is the quotient group of a
group G by a normal subgroup N, cyclic? In the above example we have a cyclic
quotient group (why?) with both G and N noncyclic. The reader may wish to
pursue this problem further.
Theorem 21 Let G be a group and N a normal subgroup of G. Then the map-
ping '{J:G-* G/N defined by 'f!(X) = Nx is a homomorphism of G onto G/N and
N is the kernel of 'f!.
Proof' 'f! is clearly onto. Suppose 'f!(x) =Nx, '{!(y) =Ny. Then 'f!(xy) =
Nxy = NxNy = 'f!(X)'f!(y). Therefore 'f! is a homomorphism. Next, supposex be-
longs to G and <P0c) = N (the identity of G/N). Since 'f!(x) also equals Nx, we
have x belongs toN. Conversely, if x belongs toN, then 'f!(X) = Nx = N. There-
fore N is the kernel of 'f!.
Introduction
Note: The homomorphism v; as defined in Theorem 21 is usually referred
to as the natural homomorphism of G onto G/N.
Theorem 22 Suppose v; is a homomorphism of a group G onto a group G and
K is the kernel of v;. Then G is isomorphic to G/K.
Proof Define rr :G/K--"" G by rr(Kx) = x, where xis the image of x under
the homomorphism v;. We show rr is an isomorphism. (The reader should show
that rr is well-defined.) Clearly, rr is onto. Suppose rr(Kx) = x and rr(Ky) = Y. If
x = y, then v;(x) = v;(y ), which implies that x andy belong to the same coset of
K (Theorem 18). Thus Kx = Ky and rr is 1-1. Finally, since rr(KxKy) =
rr(Kxy) = xy = v;(xy) = v;(x) v;(y) = xy = rr(Kx) rr(Ky ), rr is an isomorphism.
The results of Theorems 18, 21, and 22 are of such importance that were-
state them here:
(i) The kernel of any homomorphism is a normal subgroup.
(ii) Any normal subgroup is the kernel of a homomorphism.
(iii) To within isomorphism, the only homomorphic images of a group G
are the quotient groups G/N, N any normal subgroup in G.
Example 24 Let us find all homomorphic images of the group G of symmetries
of the square. These are (to within isomorphism) the quotient groups G/N,
N 6.G. Thus we must find all normal subgroups of G. It has already been shown
that K
2
= \ /,R,R',R"l, K
3
= l!,H,V,R' 'r, and K
4
= ll,R' l are normal. Of
course, so are G and! fl. That K
5
= !I,R',D,D'j is also normal may be verified
easily. Therefore the quotient groups of G are the groups
G/K
2
= i l!,R,R',R"\, !H,V,D,D'Jl
G/K
3
= i !I,H,V,R'l, \R,R",D,D'\1
G/K
4
= t\I,R'j, !R,R"i,!H,V!, !D,D'lJ
G/K
5
= t!I,R',D,D'/,tR,R",H,V il, G/G= !G\
and
G/\1! = i \I!, \R!, ! R'j, !R"i, !Hl. ! V!, IDL ! D'! J
Now K
3
and K
5
are isomorphic (thus so are G/K
3
and G/K
5
). Hence (to within
isomorphism) the homomorphic images of G are G/K
2
, G/K
3
, G/K
4
, G/G, and
G/ !I}. These last two correspond to the homomorphisms obtained by mapping
all elements onto the identity and every element onto itself, respectively.
Note: We emphasize the following: If N
1
and N
2
are normal subgroups
of a group G and N
1
and N
2
are isomorphic, then G/N
1
and G/N
2
are isomor-
phic. (Is the converse also true?)
Exercise 24 Suppose that G is a cyclic group of order n. How many homomor-
phic images of G are there?
Introduction 21
Exercise 25 Suppose G is a cyclic group and H a subgroup of G such that
[G:H) = m. Prove that G/H is cyclic of order m (see comment preceding Theo-
rem 21).
In Exercise 20 you were introduced to the derived group G' of a group G.
The following are important properties of this group.
Exercise 26 Prove that G/G' is Abelian.
Exercise 27 Suppose N is normal in G and G/N is Abelian. Prove that G '<;;; N./
We end this section with the following useful theorem.
Theorem 23 Suppose G is a group and N a normal subgroup of G. There is a
1-1 correspondence between subgroups H of G containing Nand subgroups H' of
G/N. Furthermore, H is normal in G if, and only if, H' is normal in G/N. Also,
[G:H] = [GjN:H'].
Proof: Consider the natural homomorphism <P from G onto G/N. Clearly,
if H is any subgroup of G, then ..p(H) = H' is a subgroup of G/N (Theorem 17).
Suppose H' is a subgroup of G/N and His the subset of G such that ..p(H) = H'.
Since ..p(N) = N belongs to H', H contains N. We show H is a subgroup of G.
Let x andy belong to H. Then ..p(x) and ..p(y) belong to H'. Since H' is a group,
..p(x) ..p(y) is in H'. But ..p(x) ..p(y) = ..p(xy ). Therefore xy belongs to H. Finally,
x in H implies ..p(x) in H', which implies ..p(x) -
1
= ..p(x -
1
) is in H'. Thus x -
1
is in
Hand soH is a subgroup of G. Now if we suppose K is a subgroup of G contain-
ing N such that ..p(K) is also H ', then we must have K = H. For let h belong H.
Then ..p(h) belongs to H', which is equal to ..p(K). But then h must also be inK,
implying that H C K. Similarly, we have K C H. Next, suppose that His normal
in G and x is an element of G/N. There exists g in G such that ..p(g) = x. Thus
x-
1
H'x = ..p(g-
1
) ..p(H) ..p(g) = ..p(g-
1
Hg) = ..p(H) = H', which implies thatH' is normal
in G/N. If H' 6G/N and g belongs toG, then ..p(g-
1
Hg)=..p(gr
1
H'..p(g)=H'.
This implies that g-
1
Hg = H, and soH 6 G. It is clear that [G:H] = [G/N:H'].
1-7. AUTOMORPHISMS
In this section we discuss a special class of isomorphisms of a group G.
Definition 20 An automorphism a of a group G is an isomorphism of G onto
itself.
Theorem 24 Let A( G) be the set of automorphisms of a group G. Then A( G)
is a group with respect to product of mappings.
Proof: Since A( G) is a subset of the set of all 1-1 mappings of G onto it-
self, we need only show that A(G) is closed and that every element of A( G) has
22
Introduction
an inverse in A(G). (See Example 4 and Remark 4.) Let a:
1
and a:
2
belong to
A(G). Clearly, a:
1
a:
2
is 1-1 and onto. Let x andy be elements of G. Then
(a:1a:2)(xy) = a:1(a:2(xy)) = a:I(a:2(x)a:2(y)) = a:I(a:2(x))a:I(a:2(y)) = (a:1a:2)(x)
(a:
1
a:
2
)(y). Finally, suppose a: belongs toA(G). Then a:-
1
(xy) = a:-
1
[(a:a:-
1
)(x)
(a:a:-I)(y)] = a:-1 [a:(a:-I(x)a:-I(y))] = (a:-1 a:)(a:-I(x)a:-I(y)) =a:-I(x)a:-I(y),
implying that a: -I belongs to A (G).
Definition 21 In any group G, the element a-
1
xa is called the conjugate of x
by a. (See Definition 13.)
Theorem 25 Let G be a group and a belong to G. The mapping 1fa: G-+ G de-
fined by na(x) = a-
1
xa is an automorphism.
Proof" 1fa is onto, since, given y in G, na(aya-
1
) = y. Suppose na(x) = a-
1
xa and na(y) = a-
1
ya. If a-
1
xa = a-
1
ya, then x = y. Therefore 1fa is 1-1. Fi-
nally, we have na(xy) = a-
1
(xy)a = a-
1
xaa-
1
ya = na(x)na(y). Thus 1fa is an auto-
morphism of G.
Definition 22 The automorphisms 1fa are called inner automorphisms; all other
automorphisms are called outer.
Theorem 26 The inner automorphisms of a group G form a normal subgroup
J(G) of the group A (G).
Proof: Let 1fa and nb belong to J(G). Then (nanb)(x) = na(b-
1
xb) = a-
1
b-
1
xba = (baf
1
x(ba) = nba(x), so that 1fa1fb belongs to J(G). Also,(rra-I1fa)(x) =
1fa-t (a-t xa) = (a-
1
rl (a-
1
xa) a-t =X, implying that 1fa-1 = 1fa -t. Therefore I( G) is
a subgroup of A(G). To show I( G) 1'1 A(G), let a: belong to A(G), 1fa belong to
I( G), and x belong to G. Then (a:-
1
rraa:)(x) = a:-
1
(a-
1
a:(x)a) = a:-
1
(a-
1
)xa:-
1
(a)=
rra:-r (a)(x). Thus a:-
1
rraa: = rra:-r (a) belongs to I( G) and we have I( G) is normal
inA( G).
Theorem 27 Let G be a group and H a subgroup of G. His normal in G if, and
only if, it is invariant under all inner automorphisms of G (that is, H contains
the conjugates of each of its elements by all elements of G).
Proof: See Definition 14.
Note: One perhaps sees from Theorem 27 why normal subgroups are also
referred to as invariant, or self-conjugate subgroups.
The following example illustrates the previous notions concerning automor-
phisms.
Example 25 Table 8 lists the inner automorphisms of the group G of symme-
tries of the square. Here we have n
1
= nR; 1fR = 1fR", 1fH = nv and rrD = 1fd.
Introduction 23
All other automorphisms of G are outer. These are given in Table 9. Observe
that [A(G):/(G)] = 2. This shows that /(G) is normal in A(G). (The reader
should prove in general that, if G is a group and His a subgroup of G of index 2,
then H '::,.G.) Observe too what the inner automorphisms do to the normal sub-
groups of G.
TABLE 8
X
1TI(X) 1TR(X) 1TH(X) 1TD(X)
I. I I I I
R R R
R" R"
R' R' R' R' R'
R" R" R"
R R
H H v H v
v v H v H
D D
D' D'
D
D' D'
D D
D'
TABLE9
X CX1 (X) a
2
(X) tx3(X) a
4
(X)
I I I I I
R R R
R" R"
R' R' R' R' R'
R" R" R"
R R
H D
D'
D
D'
v
D'
D
D'
D
D v H H v
D'
H v v H
We have seen that every group is isomorphic to a permutation group (Theo-
rem 12). In particular, one can show that G = I/, R, R ', R ", H, V, D, D' l and
A(G)= \7Tio 7TR, 7TH, 7TD, a
1
, a
2
, a
3
, tx4l are isomorphic. (The reader should
show this.)
Exercise 28 Show that the elements of a group G may be divided into classes
of conjugate elements, where the number of elements of a class is a divisor of
the order of G (when G is finite). Show also that every element in Z(G) forms a
class by itself. (This is a most important result and will be applied several times
in later sections.)
Exercise 29 Suppose G is any group. Prove that the mapping r.p:G _,./(G) de-
fined by r.p(a) = 7Ta-1 is a homomorphism of G onto /(G) whose kernel is Z(G).
(In particular, verify this for Example 25.)
Exercise 30 Prove that G/Z(G) is isomorphic to !(G).
:24 Introduction
Exercise 31 Show that the only inner automorphism of an Abelian group is the
identity automorphism.
Exercise 32 Determine all automorphisms of a finite cyclic group.
Solvable Groups
2-1. PERMUTATIONS
ALTERNATING GROUPS
We now return to the symmetric groups Sn, first discussed in Example 4.
(
1 2 3 4)
In Example 5, in the case n = 4, symbols such as
4 3 1 2
were used to
represent the permutations on the set S containing 4 elements, arbitrarily la-
(
1234)
beled 1, 2, 3, 4. Here,
4 3 1 2
is the permutation which maps 1 onto 4, 2 onto
3, 3 onto 1, and 4 onto 2. Of course, this notion may be extended to S =
l1, 2, ... , n l for any positive integer n. We wish now to improve on this
method of representing permutations.
Definition 23 A permutation P on a set S = l x
1
, x
2
, , xn l will be called
a cycle of length m, if S has a subset \y
1
,y
2
, ,ym l such that P(y;) =
Yi+t, 1 <,i<m andP(ym)=y
1
, whileP(yj)=yi, m<j<,n. Such a cycle will
be denoted by(yt,Yz, ... ,ym).
If an element of S is mapped onto itself by a permutation P, we shall omit
it in our notation for P (except for the identity permutation, which will be de-
(
12345)
noted by (1 )). Thus, for P =
3 2 1 4 5
we write (13). In S
4
, (124) is the per-
mutation that maps 1 onto 2, 2 onto 4, 4 onto 1, and 3 onto 3.
(
1 2 3 4) ( 1 2 3 4 5)
NowconsiderthepermutationsP
1
=
4 3 2 1
andPz =
3 54 1 2
. Clearly,
(
1 2 3 4) (1 2 3 4) (1 2 3 4)
P
1
andP
2
cannot be written as cycles. But
4 3 2 1
=
4 2 3 1 1 3 2 4
25
26
Solvable Groups
(
1 2 3 4 5) (1 2 3 4 5) (1 2 3 4 5)
(14)(23) and = = (134)(25). Thus it ap-
, 35412 32415 15342
pears that if a permutation cannot itself be written as a cycle, then it can be
written as a product of cycles. (This indeed is the case and the reader should
show it. That is, show that if Pis a permutation on the setS= l X
1
,x
2
, .. , Xn l ,
then P may be written uniquely as a product of cycles on disjoint subsets of S.)
We note that since (x 1 x 2 xn) = (x
1
xn)(x
1
xn_
1
) (x
1
x
2
), every per-
mutation on S may be written as a product of cycles, each having length 2.
(Such a cycle is called a transposition.) Of course, these are not cycles on dis-
joint subsets of S. One also loses uniqueness in this case. For example, (124) =
(14)(12)and (I 2)(24); (134) =(I 4)(13) and ( 13)(34).
Next, consider the polynomial Ll in the variables x
1
, x
2
, . , Xn given by
Ll =(xi- Xz)(xJ- X3)(x1- X4) (xi- xn)
X (x
2
- x3)(xz- x4) (xz - xn)
X (x3 - x4) (x3 - xn)
X (xn _
1
- xJ
If the permutation (x
1
x
2
) is applied to Ll we notice that Ll is changed to
the product
Cxz- xJ)(xz- x3)(xz- x4) (xz- xn)
X (x
1
- X3)(x1- X4) (xi- xn)
X (x3 - x 4) (x3 - xn)
X (xn-1 - xn)
That is, Ll is changed to-Ll. In general, for any permutationP, we haveP(Ll) =
Llor-Ll.
Definition 24 A permutation Pis said to be even if P(Ll) = Ll;odd if P(Ll) = -Ll.
Remark 18 Every transposition is an odd permutation. Also, the product of
two even or two odd permutations is even; the product of an even and odd per-
mutation is odd. Thus an even permutation is one than can be written as a prod-
uct of an even number of transpositions, and an odd permutation one that can
be written as a product of an odd number of transpositions.
Exercise 33 Write the elements of S
4
as products of transpositions, thus show-
ing that P
1
, P
2
, ... , P
12
are the even permutations.
Note: A cycle of length m is even when m is odd; odd when m is even.
Theorem 28 Denote by An the set of all even permutations in Sn Then An is
n'
a normal subgroup of Sn and iAn I = 2.
Solvable Groups '27
Proof: Define If: Sn --7 ll, -1l by ~ P P ) = 1 if P is even and P P ) = -1 if P is
odd. One shows easily that If is a homomorphism of Sn onto l1, -1l . Clearly,
An is the kernel of If. Thus An is normal in Sn by Theorem 18. Now Sn/An is
' isomorphic to l1, -1l. Therefore [Sn :An] = 2, which implies that iAn I = n2 .
An is called the alternating group on n symbols. Later we shall show that
the alternating groups afford us with examples of a certain class of groups.
Exercise 34 Find the derived group of S
4
(See Exercise 20.) Generalize your
result to Sn.
Exercise 35 Divide A
4
into classes of conjugate elements. (See Exercise 28.)
What is Z (A
4
)?
2-2. SIMPLE GROUPS
In this section we introduce the groups that are referred to as simple groups.
Definition 25 A subgroup H of a group G is a maximal normal subgroup if H
is a normal subgroup and there exists no proper normal subgroup K of G which
properly contains H.
Theorem 29 is a corollary to Theorem 23, but we emphasize it now.
Theorem 29 Suppose H is normal in G and K a subgroup of G such that
H C K C G. Then K is normal in G if, and only if, K/H is normal in G/H.
Proof: Suppose K .6. G .H .6. G and H C K C G imply that His normal inK.
Thus K/H is a group (clearly, a subgroup of G/H). We wish to show that K/H is
a normal subgroup of G/H. Let Hg belong toG/Hand Hk belong to K/H. Then
(Hgr
1
(Hk)(Hg) = (Hg-
1
)(Hk)(Hg) = Hg-
1
kg belongs to K/H (since g-
1
kg is in
K). Therefore K/H .6. G/H. Conversely, suppose K/H .6. G/H. Then if Hg be-
longs to G/H and Hk belongs to K/H, (Hgr
1
(Hk)(Hg) = Hg-
1
kg belongs to K/H.
Now the elements of K/H are of the form Hx, where xis inK. Thus we have
g-
1
kg belongs to K, which implies that K is normal in G.
Remark 19 It follows from Theorem 29 that H is a maximal normal subgroup
of G if, and only if, the quotient group G/H contains no proper normal sub-
groups. Groups with this property are of special interest.
Definition 26 A simple group is a group containing no proper normal sub-
groups.
Example 26 Groups of prime order are simple groups.
Example 27 The alternating group A
4
is not simple since it contains the nor-
28 Solvable Groups
mal subgroup N= IP
1
,P
2
,P
3
,P
4
l. (See Example 23.) Clearly, S
4
/A
4
is a sim-
ple group.
Exercise 36 Suppose G is a finite Abelian simple group. Prove that G is a
cyclic group of prime order.
We have called a group G cyclic if it contains an element x such that every
element of G can be written as a power of x. More generally, suppose G is a
group and x
1
, x
2
, .. , are elements of G. If every element of G can be ex-
pressed as a finite product of terms of the form x {
1
x
2
q , then we say that
x
1
, x
2
, .. , generate G. Clearly, every finite group is generated by a finite set
jx
1
,x
2
, ... ,Xm l.
Exercise 37 Show that the cycles (12), (13), ... , (In) generate Sn.
Theorem 30 The cycles (123), (124), ... , (12n) generate the alternating group
An,n>2.
Proof: The elements of An may be written as a product of an even num-
ber of transpositions, each of which is a product of cycles of the form (In)
(Exercise 37). Thus we need only show that a product (lj)(lk) where 2 j
k ~ n, may be expressed as a product of elements of the form (12n) or (12nr
1
=
(ln2). For j=k we have (lj)(lk)=(l)=(l23)(132). Forj=l=k andj=2 we
have (lj)(lk) = (12)(lk) = (lk2). For j =I= k and k = 2, (lj)(lk) = (lj)(12) =
(12j). And finally, for j t= k, j and k;;;, 3, we have (lj)(lk) = (lj)(12)(12)(1k) =
(12j)(lk2).
We now show that the alternating groups An, n t= 4, are simple groups.
Clearly, A
2
and A
3
are simple, since A
2
= l(l) l and A
3
has prime order. That
A
4
is not a simple group has already been discussed (Example 27). Thus it re-
mains to prove that An is simple for n > 4. For this we need the following the-
orem.
Theorem 31 Let H be a normal subgroup of An, n > 3. If H contains a cycle
of length 3, then H =An.
Proof: We may denote the cycle by (123). Now-H normal implies that
Y
1
(123)P belongs to H for all permutations Pin An. In particular, (23kr
1
X
(123)(23k) = (lk2) belongs to H for k > 3. Now (lk2) in H implies (lk2) X
(lk2) = (12k) is in H. But the cycles (12k) generate An, and so we have H =
An.
Theorem 32 An is a simple group for n > 4.
Proof: Let H=F l(l)l be a normal subgroup of An. We must show that
H = A
17
Let P be an element of H. Now P may be written as a product of dis-
Solvable Groups 29
joint cycles, say P=P
1
P
2
P
8
Suppose at least one of these cycles, say P
1
,
has length greater than 3. Then we may write P = (123 ... m)P
2
P
3
P
8
. Con-
sider the permutation Q = (132). H normal in An implies that H contains
(123)P(132). Since (132) commutes with P
2
P
3
P
8
(why?), (123)P(l32) =
(123)(123 ... m)(132)P
2
P
3
Ps = (14 ... m23)P
2
P
3
P
8
It follows that
H contains (14 ... m23)P
2
P
3
P
8
r
1
= (14 ... m23)(m ... 321) = (124).
Thus H =An by Theorem 31.
It remains to reach the conclusion when every element of His a product of
cycles of length 2 or 3. If we suppose that two of the cycles of P, say P
1
and P
2
,
have length 3, then we may writeP.= (123)(456)P
3
P
8
Consider Q = (354).
Now (345)P(354) = (345)(123)(456)(354)P
3
Ps = (124)(365)P
3
P
8
be-
longs to H, as does (124)(365)P
3
P
8
r
1
= (16345); a contradiction.
When P has only one cycle of length 3, say P
1
= (123), then (132) belongs
toHwhere (132)=(123)P
2
P
8
(123)P
2
P
8
Again,H=An by Theorem
31.
Finally, suppose P is a product of transpositions. Then we may write P =
(12)(34)P
3
P
8
Let Q = (142). Then (124)P(142) = (13)(24)P
3
P
8
be-
longs to H, which implies that (13)(24)P
3
PsP-
1
=(14)(23)belongs to H.
Now consider R = (235). Since (253)(14)(23)(235) = (14)(25), and (14)(25) X
(23)(14) =(235), (235) is an element of H. Therefore H =An and the proof is
completed.
Remark 20 The only subgroup H of Sn, n > 4, such that [Sn :H) = 2, is An.
For suppose [Sn :H) = 2. Then His normal in Sn, implying that H nAn is nor-
mal in An. Now An is simple for n > 4. Thus H nAn = l (1) l or An. Clearly,
H nAn =An. This implies that An t:;;,_ H. Since [Sn :An] = 2, we have An =H.
(Is this result also true for n.;;:; 4?)
2-3. COMPOSITION SERIES
SOLVABLE GROUPS
At this time we will prove the well-known Jordan-Holder theorem for com-
position series and develop some useful results on solvable groups.
Definition 27 A finite sequence l e l = H
0
_f; H 1 _f; H 2 _f; _f; Hr = G of sub-
groups of a group G is called a normal series if Hi_
1
is normal in Hi, i = 1 ,2, ... ,r.
The factor groups Hi/Hi_
1
are called the factors of the normal series.
Note: In a normal series, a term may be repeated any number of times. If
this is not the case, we speak of a series without repetitions.
Definition 28 A second normal series lei = K
0
t:;;,_ K
1
t:;;,_ K
2
t:;;,_ i: Ks = G is
called a refinement of the first one if all the Hi also occur in this second series.
30
Solvable Groups
Defmition 29 Two normal series are called isomorphic if all factors of one
series are isomorphic with the factors of the second series in some order.
Defmition 30 A normal series without repetitions, which cannot be refined
without repetition, is called a composition series. The factors of a composition
series are called composition factors.
Remark 21 A normal series \ e l = H
0
C H
1
C C Hr = G is a composition
series if, and only if, H;_
1
is a maximal normal subgroup of H; for each i. That
is, if, and only if, H;/H;_
1
is simple for each i.
Remark 22 Every finite group G possesses at least one composition series.
Proof: If G is simple, l e l k. G is a composition series. Suppose G is not
simple. Then G contains a proper normal subgroup (and therefore a proper
maximal normal subgroup), say Kr_
1
Now l e l C K
7
_
1
C G is a composition
series if K
7
_
1
is simple. If K
7
_
1
is not simple, then Kr-1 Kr-2 where Kr_
2
is a
maximal normal subgroup of K
7
_
1
l e l C K
7
_
2
C Kr_
1
C Gis a composition ser-
ies if K
7
_
2
is simple. If K
7
_
2
is not simple, we continue as above, arriving at a
composition series l e l C K
1
C K
2
C C K
7
_
1
C G. The finiteness of G im-
plies that there is only a finite number of the K;. (Each K;_
1
is a maximal nor-
mal subgroup of K;.)
Exercise 38 Prove that if G is an infinite cyclic group, G does not possess a
composition series.
Example 28 The cyclic group G of order 20 generated by a has the following
normal subgroups: \ e \, H
1
= \a
10
, e l,H
2
= \a
5
, a
10
, a
15
, e l,H
3
= la
4
,a
8
, a
12
,
a
16
, e\, H
4
= l a
2
, a
4
, ... , a
18
, e l, and G = fa, a
2
, ... , a
19
, e l. [ e l C H
3
C
H
4
C G, I e l CH
1
CH
4
C G, and lel CH
1
CH
2
C G are composition series
for G. The composition factors for [e l C H
3
CH
4
C G are G/H
4
= IH
4
, H
4
al,
H/H
3
=\H
3
, H
3
a
2
1, and H
3
/fel =H
3
. The factors of { e J CH
1
CH
2
C G are
G/H
2
=I H
2
, H
2
a, H
2
a
2
, H2a
3
, H2a
4
l, H2/H1, = [H1, H1a
5
\, and Htf!el =
Ht.
Now consider the following mappings: 7T1 : G/H4 -+ H2/H1 defined by
1r
1
(H
4
)=H
1
, 1r
1
(H
4
a)=H
1
a
5
; 1T2: H4jH3 -+Ht defined by.
1r
2
(H
3
a
2
)=a
10
; 1r
3
: H
3
-+G/H
2
defined by 1T3(e)=H2, 1T3(a')=H2a', i=
1, 2, 3, 4. 7T
1
, 7T
2
, and7T
3
are isomorphisms.
Thus it appears that a group G may have more than one composition series,
but when this is the case, they have the same length and the composition factors
of one are isomorphic to the composition factors of the other. This is the
Jordan-Holder theorem, which we prove after first considering the following:
Theorem 33 Suppose G is a group, H and K distinct maximal normal sub-
Solvable Groups
31
groups of G. Then G/H is isomorphic to K/H n K and G/K is isomorphic to
HjHnK.
Proof: We first note that H n K is normal in H and in K (Exercise 15).
Now HK is a normal subgroup of G (Exercise 18). Since Hand K are distinct,
we have H C HK i: G. The fact that His a maximal normal subgroup of G im-
plies that HK must be G.
Let D = H n K and decompose K according to cosets of D, obtaining K =
D + Dx
2
+ Dx
3
+ . Then HK = HD + (HD)x
2
+ (HD)x
3
+ . (For sup-
pose hk belongs to HK. Now k belongs to K, and so k belongs to Dxi for some i.
Thus hk is an element of (HD)xi, which implies that HK ~ HD + (HD)x
2
+ .
But HK = G implies HK -:2 HD + (HD)x
2
+ .) Since HD = H (H n K ~ H , we
have G = H + Hx
2
+ Hx
3
+ . This is a decomposition of G according to
cosets of H, since if Hxi = Hxj, then hxi = h'xj, implying that xixj-
1
= h-
1
h' is in
H. Now xi and Xj belong to K. Thus xixj-l belongs to K. Hence we have xixt
in D and so Dxi = Dxj; a contradiction. (D + Dx
2
+ Dx
3
+is a decomposi-
tion of K.)
Now G/H= \H,Hx
2
,Hx
3
, . ! and K/D= lD,Dx
2
,Dx
3
, .. ! . Let JT:
G/H-+K/D be defined by IT(Hxi)=Dxi, i=l, 2, .... (Hx
1
=H;Dx
1
=D.)
Clearly, 1T is 1-1 and onto. Since 1T (Hx#xj) = 1T (Hx#xj) = Dxixj = Dxpxj =
1T (f!xi) 1T (Hxj), 1T is an isomorphism. That G/K is isomorphic to H/D is left as
an exercise.
Note: In some of the material which follows we shall find it convenient
to change our notation for denoting a series. For example, we may write G =
H
0
-:J H
1
-:J -:J Hr = I e l instead of I e ! = H
0
C H 1 C C Hr = G.
Theorem 34 (Jordan-Holder) Let G =I= !e l be a finite group with two compo-
sition series
G=Ho -:JH1 -:J-:JHr= le!,
G=Ko -:JK1 -:J -:JKs= leL
Then r = s and the factor groups Hi.
1
/Hi are isomorphic with the factor groups
Kj_
1
/ Kj in some order.
Proof: Let G have order n = p
1
p
2
Pt (where the Pi are primes, Pot
necessarily distinct). The proof will be by induction on t. For t = 1, G is a group
of prime order, and thus has no proper subgroups. Therefore G -:J ! e! is the only
composition series. Suppose the theorem true for t = 1, 2, ... , k and let G be
a group whose order contains k + 1 prime factors. If G is simple, then G -::; I e!
is again the only composition series. If G is not simple, there is a composition
series of length greater than 1. (By "length" we mean the number of composi-
tion factors.) If this series is unique there is nothing to prove. If it is not unique
then there exists two distinct composition series of length greater than 1, say
32
Solvable Groups
(*) G=Ho -:JH1 -:J -:JHu = e , and
(**)G=Ko-:JK1-:J-:JKv= e
If H 1 = K 1 , the theorem follows by the induction hypothesis. (The order
of H 1 contains k or fewer prime factors.) If H
1
=I= K
1
, we let D = H
1
n K
1
and
consider
(i) G-:JH
1
-:JD-:JD
1
-:J-:JDw= e,
(ii) G-:JK
1
-:JD-:JD
1
-:J-:JDw= e,
where D -:J D
1
-:J -:J Dw is a composition series forD. Since H
1
is a maximal
normal subgroup in G, G/H
1
is simple. Since G/H
1
is isomorphic to K
1
/D
(Theorem 33), K
1
/D is simple, and thus Dis a maximal normal subgroup inK
1
.
Hence (i) and (ii) are two composition series. Since G/H
1
is isomorphic to K
1
/D
and H
1
/D is isomorphic to G/ K
1
, and the order of D contains fewer than k
prime factors, the theorem holds for the series (i) and (ii). Since G/H
1
is isomor-
phic to G/H
1
, and the order of H
1
contains k or fewer prime factors, the
theorem holds for the series ( *) and (i). Similarly, the theorem holds for the
series ( **) and (ii). It follows that the theorem is true for the composition
series ( *) and ( ** ). This completes the proof.
Definition 31 A group G is said to be solvable (soluble) if the sequence
G ::::2 G' ::::2 G" ::::2 . ,where each group c<i) is the derived group (see Exercise 20)
of the preceding, terminates in the identity subgroup in a finite number of
steps, say c(s) = lel.
Note: If c<i) = G(i+l), then c<i) = c<j) for all j > i.
Exercise 39 Prove that every subgroup of a solvable group is solvable.
Theorem 35 A group G of finite order is solvable if, and only if, the factor
groups in any composition series for G are cyclic of prime order.
Proof" Let G = H
0
-:J H
1
-:J -:J Hr = l e l be a composition series where
each Hi_IfHi is cyclic of some prime order. Since G/H
1
is Abelian, H
1
.::::2 G'
(Exercise 27). Since H
1
/H
2
is Abelian, and since ~ .::::2 G", we have
H2 .:2H
1
' .::::2 G". It follows that Hr.:2 c<r). Thus c<r) = lel, implying that Gis
solvable. Conversely, suppose G is a finite solvable group with G -:::; G' -:::; G" -:::;
-:J c<s) = l e \. Let H
1
be a maximal normal subgroup of G such that H
1
.::::2
G'. Now G/H
1
is simple (Remark 19). Since G/G' is Abelian and G/G' ;:2 G/H
1
,
G/H
1
is also Abelian. Thus G/H
1
is cyclic of prime order (Exercise 36). LetH
2
be a maximal normal subgroup inH
1
. SinceH
1
is solvable (Exercise 39),H
1
/H
2
is cyclic of prime order. Continuing, we have a composition series G = H
0
-:J
H
1
-:J H
2
-:J -:J Hr = I e l with each Hi_ If Hi cyclic of prime order. Thus the
factors in any composition series will be cyclic of prime order by the Jordan-
Holder theorem.
Solvable Groups
33
Theorem 36 G is solvable if, and only if, G has a finite normal series G =
H0 d.H1 d H2 d d Hs =! e! in which every factor Hi_t/Hi is Abelian.
Proof: This is clear if G is solvable. Suppose G = H
0
d H
1
d d Hs =
\ e ! is a normal series in which each Hi-t/Hi is Abelian. Then H
1
d G'. Sim-
ilarly, H2 normal in H1 and HdH
2
Abelian imply that H
2
dH/ (d G" since
H
1
::2 G'). Finally, we have \ e l =Hsd. c<s), implying that c<s) = \e!, which
in turn implies that G is solvable.
Example 29 Every Abelian group G is solvable, since in this case we have G'=
\e!.
Example 30 The symmetric groups Sn for n 4 are solvable. In S
3
, S
3
::J A
3
::J
\ e l is the only composition series. The factors S
3
/A
3
and A
3
/ l e l have order 2
and 3 respectively. le! CKCNCA
4
CS
4
whereN=\P
11
P
2
,P
3
,P4( andK=
!P
11
P2l, is a composition series for S
4
. Here, IS
4
/A
4
1 = 2, IA
4
/NI = 3, IN/KI =
2, and I K/ l ell = 2.
Remark 23 It can be shown (see Chapter 5) that the general nth degree equa-
tion is solvable by radicals (that is, by using rational operations and extraction
of roots) if, and only if, the symmetric group Sn is a solvable group. Thus there
exist formulas (such as the quadratic formula) that give the general solution to
equations of degree n 4. That no such formulas exist for equations of de-
gree n;;;. 5 (that is, such equations are not solvable by radicals) is evident when
one observes that Sn for n ;;;. 5 is not solvable. This is immediate from the fact
that the only composition series for such a symmetric group is Sn ::JAn ::J \ e l.
n!
Now Sn/An has prime order, but An/ I ej does not. IAn/ I ell =2, which is not a
prime for n > 4.
Exercise 40 Every quotient group of a solvable group is solvable.
Exercise 41 Prove that if a group G contains a subgroup H with the property
that Hand G/H are both solvable groups, then G must also be solvable.
2-4. DIRECT PRODUCTS
We conclude this chapter with some results on direct products of groups.
Definition 32 Let Hand K be groups. The setH X K = ! (h,k); h in H, kin Kl
is called the direct product of H and K. .
Suppose we define a binary operation* in H X K as follows: (h,k) *(h', k') =
(hh',kk'). Then H X K together with* is a group. For suppose (h
1
, k!},(h
2
, k
2
),
34
Solvable Groups
and (h
3
,k
3
) are elements of the direct product H X K. Then [(h
1
,kt)(h
2
,k
2
)] X
(h3,k3)=(h1 h2, k1 k2)(h3, k3)=((h1 h2)h3, (k1 k2)k3)=(h1(h2 h3)'
k
1
(k
2
k
3
)) = (h
1
,k t)(h
2
h
3
,k
2
k
3
) = (h
1
,k t) [(h
2
,k
2
)(h
3
,k
3
)], and * is associa-
tive. If e and e' denote the identities of Hand K respectively, then ( e ,e') is the
identity of H X K. Finally, (h,kr
1
= (K
1
,k-
1
).
Note: One can of course generalize Definition 32 to the direct product of
three or more groups. We shall, however, limit our discussion here to the direct
product of two groups.
Exercise 42 Suppose H is a cyclic group of order m and K a cyclic group of
order n, where m and n are relatively prime. Show that H X K is cyclic of order
mn.
Remark 24 H X K contains a subgroup isomorphic to H and a subgroup iso-
morphic to K.
Proof" Let H
1
= l (h,e'); h in H, e' the identity of K l and K
1
= \ (e,k); e
the identity of H, k in K l. Clearly, H
1
and K
1
are subgroups of H X K. Define
1r
1
: H
1
by 1T
1
((h,e'))=h. It is clear that 1r
1
is 1-1 and onto. Since
1r
1
((h,e')(h',e')) = 1T
1
((hh',e')) = hh' = 1T
1
((h,e'))1T
1
((h', e')), 1T
1
is an isomorphism.
Similarly, 7T
2
:K
1
~ K defined by 1r
2
((e,k)) = k, is an isomorphism of K
1
onto K.
We may write H
1
=H X l e' l and K
1
= \ e! X K. Since 7T
1
and 7T
2
are iso-
morphisms, we identify H with H X l e'! and K with l e l X K.
Exercise 43 Prove that (H X K)/H is isomorphic to K and (H X K)/K is iso-
morphic to H.
We have seen that given two arbitrary groups Hand K, we can form a group
H X K, the direct product of Hand K. We turn our attention now to the follow-
ing: suppose G is a group and Hand K subgroups of G. When will G be the di-
rect product of H and K? (More precisely, we want to know when G will be
isomorphic to the direct product H X K, since clearly G can never actually equal
this product.)
Theorem 37 Let G be a group,H and K subgroups of G. If
(i) every element of H commutes with every element of K, and
(ii) every element g in G can be uniquely expressed in the form g = hk,
h in H, k in K, then
(a) H n K = l e l, and
(b) G is isomorphic to H X K.
Proof" Suppose x is an e!ement of H n K. Then x-
1
belongs to H n K
(since H n K is a subgroup). Let g belong to G. Now g = hk and g = (hx) X
(
1
k) are two representations of g as a product of an element in H by an
Solvable Groups 35
element in K. Thus by (ii) we have hx = h and x-
1
k = k, which implies that
x = e. Therefore H II K = I e l .
To prove (b), let g belong to G. Then g = hk, h in H, kinK (and this repre-
sentation of g is unique). Consider rr:G---)- H X K defined by rr(g) = (h,k).
Clearly, rr is onto. Also (h,k) = (h' ,k') implies that h = h' and k = k', which im-
plies that hk=h'k', which implies that rr is 1-1. Supposeg
1
=h
1
k
1
andg
2
=
h2k2. Then rr(g
1
g
2
) = rr(h
1
k1h2k2) = rr(h
1
h2k
1
k
2
), (by (i)), = (h
1
h
2
,k
1
k
2
) =
(h
1
,k!)(h
2
,k
2
) = rr(g!)rr(g
2
). Therefore rr is an isomorphism and G is isomorphic
toHX K.
Thus (i) and (ii) above are enough to insure that G be isomorphic to the
direct product of its subgroups H and K. Conversely, if G = H X K, then (i)
and (ii) hold, as well as (a). We prove this now.
Theorem 38 Let G be a group, Hand K subgroups of G. If G is isomorphic to
H X K, then ( inHX K)
(i) every element of H commutes with every element of K,
(ii) every element g in G can be expressed uniquely in the form g = hk,
h in H, kinK, and
(iii) H II K = ! el.
Proof" We identify H with H X lel and K with lel X K. Since (h,e) X
(e,k) = (h,k) = (e,k)(h,e), (i) holds. Let g belong to G. Then g = (h,k) = (h, e) X
(e,k) is an element of (H X lel )( !e\ X K), which is isomorphic to HK. Thus
there is at least one representation of g of the required form. Suppose g =
(h,e)(e,k) and g = (h',e)(e,k'). Then (h,e)(e,k) = (h',e)(e,k'), which implies
that (h,k) = (h',k'). Thus h = h' and k = k', and there is at most one such repre-
sentation. Therefore (ii) holds. Finally, suppose x belongs to H n K. Then x is
in Hand x is inK. Thus x = (h,e) and x = (e,k). Therefore h = e and k = e, im-
plyingthatHIIK= l(e,e) \= !e\. Hence(iii)holds.
Remark 25 If H and K are normal subgroups of a group G such that H II K =
I e }, then every element of H commutes with every element of K.
Proof: Suppose h belongs to H and k belongs to K. Then h-
1
k-
1
hk =
h-
1
(k-
1
hk) belongs to H(h in H implies h-
1
in H; H b. G implies k-
1
hk is in H)
and h-
1
k-
1
hk = (h-
1
7(
1
h)k belongs to K (kinK implies k-
1
inK; K b. G implies
h-
1
k-
1
h is in K). Therefore h-
1
k-
1
hk belongs to H II K. Since H II K = I e l, we
have h-
1
k-
1
hk =e. Thus hk = kh.
It follows from Remark 25 that we may replace Theorem 37 by the
following:
Theorem 3 7' Let G be a group, Hand K subgroups of G. If
(i) His normal in G and K is normal in G,
36 Solvable Groups
(ii) every element g in G can be uniquely expressed in the form g = hk ,
h inH, kinK, and
(iii) H n K = e , then G is isomorphic to H X K.
From the proof of Exercise 43, we see that H and K are normal in H X K.
Thus Theorem 38 may be replaced by the following, which is the converse of
Theorem 37'.
Theorem 38' Let G be a group, Hand K subgroups of G. If G is isomorphic to
HX K, then
(i) Hand K are normal in G,
(ii) every element g in G can be uniquely expressed in the form g = hk,
h inH, kinK, and
(iii) H n K = l e J.
Example 31 Suppose His the multiplicative group of positive real numbers and
K the cyclic group of order 2 generated by -1 with respect to multiplication.
Then G, the multiplicative group of nonzero real numbers, is isomorphic to
H X K under the mapping 1T defined as follows: 1r(g) = (g,1), g positive, and
1T (g)= (lgl, -1 ), g negative.
Example 32 The group G of Example 11 is the direct product of its sub-
groupsH= 11,101 andK= \1,3,4,5,91.
Example 33 Consider the subgroups K
1
= II,Hl and K
2
= II,R,R',R"l of the
group G = II,R,R',R",H, V,D,D' I of symmetries of the square. Clearly,
K
1
n K
2
=I II. Also, G = K
1
K
2
, as is evident from Table 1. Thus (ii) and (iii)
of Theorem 37' hold. Now K
1
X K
2
= I(/,/), (I,R), (I,R'), (I,R"), (H,l), (H,R),
(H,R'), (H,R") j. Since G contains 5 elements of order 2, while K
1
X K
2
con-
tains only 3, there can be no isomorphism of G onto the direct product of its
subgroups K
1
and K
2
Therefore one of the subgroups,K
1
or K
2
, must not be
normal in G. (Which one?)
Exercise 44 Show that the group of symmetries of the square is not isomorphic
to a direct product of any of its proper subgroups.
Exercise 45 Prove or disprove: A direct product of solvable groups is solvable.
3
p-Groups
3-1. CONVERSE OF LAGRANGE'S THEOREM
DOUBLE COSETS
We have seen that the order of a subgroup of a finite group must be a di-
visor of the order of the group (Theorem 7). The converse of this, however, is
not generally true. For example, the reader can show that the alternating group
A
4
, although containing subgroups of order 2, 3, and 4, does not have any sub-
group of order 6.
Of interest to us now is determining when a group of order n will contain
a subgroup of order m, m a divisor of n. You will recall that we have already dis-
cussed this for cyclic groups (Remark 12).
Theorem 39 Suppose G is an Abelian group of order n. If n is divisible by p,
a prime, then G contains an element of order p (and thus a subgroup of order p).
Proof: Suppose n == mp and proceed by induction on m. For m == 1 the
theorem is clear. If m =t= 1, G contains proper subgroups (Theorem 10). Let H
be a proper subgroup having maximal order, say h. There are two cases:
(i) p divides h. By the induction hypothesis H contains an element x =t= e
such that xP ==e. But x belongs to G.
(ii) p does not divide h (that is, p and h are relatively prime). Since H C G,
there exists an element x not in H. Let the order of x be k, and K be
the cyclic subgroup generated by x. Consider HK. Now G is Abelian,
which implies that HK == KH. Thus HK is a subgroup of G (Exercise
16). Clearly, HK ::J H. Since H was assumed to be a proper subgroup
having maximal order, it follows that HK ==G. Now the order of HK is
hk
d, w!lere d is the order of H n K (Theorem 15). Thus nd == hk. Since
37
38 p-Groups
p divides nd, and since (p,h)=1, we havep divides k. Let k=ps.
Then the element y = xs has order p.
Theorem 39 is also true when G is non-Abelian, as we now show.
Theorem 40 Suppose G is a group of order n. If n is divisible by a prime p,
then G contains an element of order p.
Proof: As in the proof of Theorem 39, we let n = mp and proceed by in-
duction on m. The theorem is clear if m = 1. If m * 1, then G will contain at
least one proper subgroup. Again we distinguish two cases:
(i) G contains a proper subgroup H such that [G:H] is not divisible by p.
Then p divides the order of H, and by the induction hypothesis H (and
hence G) contains an element of order p.
(ii) If His a proper subgroup of G, then [G :H] is divisible by p. From
Exercise 28, n = n
1
+ n
2
+ + ns, where each ni is the number of con-
jugates in a class of elements of G. Ifni* 1, then ni must be the index
of a proper subgroup, and hence divisible by p. Now n
1
= 1 (the iden-
tity e is a class). Therefore, the number of ni = 1 is a multiple of p.
Since an element ai is a class in G if, and only if, it is in the center of G ,
the order of Z(G) is divisible by p. Since Z(G) is Abelian, Z(G) con-
tains an element x of order p (this is Theorem 39). Of course, x belongs
to G.
Let G be a group of order n. Now we may write n as a product of primes,
say n = p
1
p
2
Ps Of course, the Pi need not all be distinct. If this be the
case, then the order of G is divisible by some power of a prime, say pi. Theorem
40 guarantees the existence of a subgroup of G having order p, but not of order
pi. We shall show shortly, however, that when G has order divisible by a prime
power, then G has a subgroup of this prime power order.
In Sec. 1-2 the concept of coset was introduced. We shall now extend this
to the concept of double coset. (In making use of double cosets, our approach
is founded on the treatment given by Marshall Hall Jr. in his The Theory of
Groups (New York: Macmillan, 1959).)
Definition 33 Let G be a group, H and K subgroups of G (not necessarily dis-
tinct). The set HxK = l hxk; h in H, kin K, x a fixed element of G lis called a
double coset.
Exercise 46 Prove that two double cosets HxK and HyK are either disjoint or
identical.
Note: A double coset HxK contains all right cosets of H of the form Hxk,
kinK, and all left cosets of K of the form hxK, h in H.
Theorem 41 The number of left cosets of Kin HxK is the index of K n x-
1
Hx
p-Groups 39
in x-
1
Hx (that is, [x-
1
Hx: K II x-
1
Hx]) and the number of right co sets of H in
HxK is the index of K nx-
1
Hx inK (that is, [K: K II x-
1
Hx] ).
Proof: Let x-
1
Hx =A and K II A =D. We decompose A according to left
cosets of D such that A = D + u
2
D + + urD, where [A :D] = r. Then K, u
2
K,
... , urK are left cosets of K in AK. (The reader should show that these cosets
are distinct.) Now every left coset of K in AK has the form aK, a in A, where
a= uid, din D. But uidK = uiK Thus the number of left cosets of Kin AK is
equal to the index of Din A; that is, [x-
1
Hx: K II x-
1
Hx]. Since the 1-1 mapping
f:HxK...,. x-
1
HxK defined by f(hxk) = x-
1
hxk, is also a 1-1 mapping of the set
of left cosets of Kin HxK onto the set of left cosets of Kin x-
1
HxK, it follows
that [x-
1
Hx: K II x-
1
Hx] is the number of left cosets of K in HxK. Similarly
we may show that the number of right cosets of A in AK is equal to the index of
D in K; that is, [K :K II
1
Hx]. But this is the number of right cosets of H in
HxK since f is also a 1-1 mapping of the set of right cosets of H in HxK onto the
set of right cosets of x-
1
Hx in x-
1
HxK.
Example 34 In S4, let H= \P
1
, Ps, P
6
, P
16
, P22, P24l, K = l P
1
, P
2
, P
3
, P4,
P1s,P
19
,P2
2
,P2
3
l, andx = P
8
. Then HxK = S
4
,x-
1
Hx = \ P
1
, P
9
, P
10
, P
13
, P
21
,
P
22
I, and K II x-
1
Hx = \ P
1
, P
22
\. Clearly, there are 3 distinct left co sets of K
in HxK and 4 distinct right cosets of H in HxK. It is also clear that [x-
1
Hx: K II
x-
1
Hx] = 3 and [K:K nx-
1
Hx] = 4.
Theorem 42 Suppose G is a group of order n = pms, where pis a prime and
(p,s) = 1. Then G contains at least one subgroup of order pi, i = 1,2, ... , m.
Furthermore, each subgroup of order pi, i = 1, 2, ... , m- 1, is normal in at least
one subgroup of order pi+l.
Proof: Since G contains a subgroup of order p (Theorem 40) we may
proceed by induction on i. Let H be a subgroup of G having order pi, where
1 :;;;;; i < m, and decompose G according to double cosets of H obtaining G =
H + Hx
2
H + + HxrH. Let nj be the number of left cosets of H in HxjH.
Then nj = [xt Hx(H II Xj-! Hxj] by Theorem 41. Now [ G :H] = n
1
+ n
2
+
+ nr. Also, each nj is 1 or a power of p (note that n
1
= 1 since x 1 =e).
Since p divides [G:H], the number of n/s equal to 1 must be a multiple of p.
For nj = 1 we have xt Hxj = H, which implies that Xj belongs to Nc(H). And
for Xj in Nc(H), we have xt Hxj = H, and so nj = 1. Hence the number of
nj's equal to 1 is [N c(H) :H], and the order of N c(H)/H is divisible by p. Thus
N c(H)/H contains a subgroup K' of order p (Theorem 40). It follows from
Theorem 23 that K' = K/H where Hr;:;. K r;:;.Nc(H), and [K:H] = [K': \ e!] = p.
Hence, K is a subgroup of order pi+
1
and H K.
It is to be emphasized in Theorem 42 that the subgroups of G of order pm
40
p-Groups
need not necessarily be normal in G. S4 has order 24 = 2
3
3, and thus must con-
tain at least one subgroup of order 8. We have seen (Example 34) that K =
lP1, P2, P3, P4, P1s, P19, P
22
, P
23
j is such a subgroup. Yet, since P
7
-
1
P
18
P
7
=
P
17
is not an element of K, K is not a normal subgroup of S
4
.
Exercise 47 Show that no subgroup of order 8 in S
4
is normal.
3-2. P-GROUPS
Theorem 42 gives us a good deal of information about the subgroups of any
finite group G, since the order of any finite group can be expressed in the form
pm s, p a prime, (p,s) = 1. Of particular interest is the case where s = 1. Here, G
is a group of prime power order. These are the p-groups.
Definition 34 A group G is called a p-group if it has order a power of the prime
p.
Note: Clearly, if G is a p-group, then every element of G has order a
power of p. If G is a finite group such that every element of G has order a
power of p, then G itself has order a power of p (the order of G cannot be di-
visible by two different primes), and thus is a p-group.
Remark 26 All p-groups are solvable.
Proof: If G is a group of order pm, then G contains a normal subgroup
Hm_
1
of _order pm-
1
(Theorem 4;2). Hm_
1
contains a normal subgroup Hm_
2
of
order pm-
2
. By a continued application of Theorem 42 we have a composition
series I e l = H
0
C H
1
C C H m _
1
C G, where clearly, each composition factor
has prime order. Thus G is solvable by Theorem 35.
Theorem 43 If G is a p-group (G =F { e l), then Z(G) =F lei. (That is, the center
of a p-group =F l e l has order at least p.)
Proof: We have G = C
1
+ C
2
+ + Cr where each Ci is a class of conjugate
elements (Exercise 28). Let ni be the number of elements in Ci. From Theorem
16, ni is the index of a subgroup of G. Thus ni = I or is a power of p. If G has
order pm then pm = n
1
+ n
2
+ + nr. Since n
1
= 1 (we let C
1
= l e l), it fol-
lows that the remaining n/s are not all powers of p. Thus ni = 1 for at least one
i > 1, and so Z(G) contains elements other than e. (An element in Z(G) forms a
class by itself.)
Theorem 44 Suppose G is a p-group andN a normal subgroup of G of order p.
Then N s;:;Z(G).
Proof: Let a be a generator of N. Then N = l a, a
2
, ... , aP-
1
, e \. From
Theorem 16 we have that the number of conjugates of a is [G :N c(a)], and thus
p-Groups
41
is 1 or a power of p. Since a has at most p - 1 conjugates (the conjugates of a
are among the elements a, a
2
, , ap--
1
, since N is a normal subgroup), the only
conjugate of a is a itself. Thus a (and therefore N) is contained in Z(G).
Example 35 It has already been noted (Example 22) that the center of the
group G of symmetries of the square is K
4
= !I, R' L Thus, in this non-Abelian
p-group G, the order of the center is p. It follows from Theorem 44 that K
4
is
the only normal subgroup of G having order p. (See Example 24.)
It will be an immediate consequence of Theorem 46 that the center of any
non-Abelian p-group G of order p
3
is always of order p. The center can be the
only normal subgroup of G having order p by Theorem 44.
Theorem 45 Let G be a group andN a normal subgroup of G. If N _f Z(G) and
G/N is cyclic, then G is Abelian.
Proof: Suppose G/N is generated by the coset Na. Since every element of
G is in some coset of N, we may express the elements of G in the form g = nai,
where n is in the center of G. For g
1
= n
1
ai and g
2
= n
2
d, we have g
1
g
2
=
n
1
ain
2
d = n
1
n
2
ai"i = n
2
n
1
aj+i = n
2
dn
1
ai = g
2
g
1
Thus G is an Abelian group.
Remark 27 It follows from Theorem 45 that if G is a non-Abelian group, the
only possible cyclic quotient groups G/N of G are the quotient groups by normal
subgroups N such that N is not contained in the center of G. In particular,
G/Z(G) is not cyclic if G is not Abelian. (See comment following Example 23.)
Exercise 48 Find the cyclic quotient groups of the alternating group A
4
.
Note: Every homomorphic image of a cyclic group is cyclic (Exercise 25).
However, Exercise 48 shows that all homomorphic images of a noncyclic group
need not be noncyclic.
Theorem 46 Suppose G is a non-Abelian p-group. Then [ G: Z( G)] is divis-
ible by p
2
.
Proof: The quotient group G/Z(G) is not cyclic (Remark 27). Hence the
index of Z(G) in G (which is the order of G/Z(G)) is not equal to p. Since
[G:Z(G)] dividestheorderofG,ap-group,wehave [G:Z(G)] =pi,i:?2.
Remark 28 In the above proof we note that G has order pm, m:? 3, since the
order of G is the product of the order of Z(G) and the index of Z(G) in G; and
Z(G) has order P',j;? 1 by Theorem 43. Thus groups of order p
2
are Abelian.
Remark 29 Suppose G is a non-Abelian group of order p
3
. Then G/Z(G) has
order p
2
Therefore, by Exercise 27, the derived group G' of G is contained in
Z(G). Since IZ(G)I = p and G' =t=- l e l, (since G is non-Abelian) we see that G' =
Z(G). That is, in a non-Abelian p-group of order p
3
, the center and the derived
group are identical.
p-Groups
Exercise 49 Show the existence of two nonisomorphic groups of order p
2
and
prove that, to within isomorphism, these are the only groups having order p
2

3-3. SYLOW THEOREMS
We have seen that when the order of a group G is divisible by a power of a
prime p, then G has a subgroup H of this prime power order. Of course, His a
p-group (p-subgroup of G). Of interest to us now is the case where G is itself not
a p-group, but does contain p-subgroups. Of particular interest are the p-sub-
groups of G having maximal order.
Definition 35 Let G be a group and P a subgroup of G. If Pis a p-group, not
contained in any larger p-group which is also a subgroup of G, then we call P a
Sylow p-subgroup of G.
Note: It follows from Theorem 42 that every p-group which is a subgroup
of a group G, is contained in a Sylow p-subgroup of G.
Example 36 The Sylow 2-subgroups of s4 are HI= \PI' P2, P3,P4,Pis,PI9>
P
22
, P
23
J, H
2
= !P
1
, P
2
, P
3
, P4, P
13
, P14, P
15
, P
16
l, andH
3
= \P
1
,P
2
,P
3
,P
4
,
P17, P2o, P21, Pz
4
-f. For X1 =Ps, X2 = P
7
, and x
3
=P
9
, we have x
1
-
1
H
1
x
1
=
H
2
, x
2
-
1
H
1
x
2
=H
3
, and x
3
-
1
H
2
x
3
=H
3
. Thus the Sylow 2-subgroups of S
4
are conjugate. Note that the number of Sylow 2-subgroups is a divisor of the
order of s4.
The Sylow 3-subgroups of S
4
are K
1
= !P
1
, P
5
, P
6
l, K2 =! P1, Pu, P12l,
K
3
= !P
1
,P
9
,P
10
l, and K
4
= IP
1
,P
7
,Pal. Forx
4
=P
2
,x
5
=P
3
,x
6
=P
4
,x
7
=
P
16
,x
8
=P
6
,andx
9
=Pu,wehavex4-
1
K
1
x4 =K2,Xs-
1
K1xs =K3,X
6
-
1
K
1
x
6
=
K
4
, x
7
-
1
K
2
x
7
= K
3
, Xs-
1
K
2
xs = K4, and X9-
1
K
3
x
9
= K4. Thus the Sylow 3-
subgroups of S
4
are conjugate and the number of Sylow 3-subgroups is a divisor
of the order of s4.
We shall now prove that the above is true in any finite group G. That is,
the Sylow p-subgroups of a finite group G are conjugate and the number of
Sylow p-subgroups is a divisor of the order of G. In fact, we will show that the
number of Sylow p-subgroups has the form 1 + kp, k an integer.
Theorem 47 The Sylow p-subgroups of a finite group G are conjugate.
Proof: Let P
1
and P
2
be Sylow p-subgroups of G and decompose G ac-
cording to double cosets, obtaining
G=P
1
P
2
+P
1
x
2
P
2
+ +P1 XrPz
Let nj be the number of left cosets of P
2
in P
1
xjP
2
. Then nj = [xj-
1
P
1
x(P
2
n
xtP
1
xj]. Now nj = 1, or is a power of p. Since n1 + n2 + + nr = (G :P2 ]
p-Groups 43
is not a multiple of p (else P
2
would be contained in a larger p-group which is a
subgroup of G), it follows that ni = 1 for some j. Therefore xi-l P
1
xi = P
2
, im-
plying that P
1
and P
2
are conjugate.
Note: The Sylow p-subgroups of G are isomorphic, since they are con-
jugate.
Exercise 50 Suppose G is a group and P a Sylow p-subgroup of G. Prove that
the only Sylow p-subgroup of G that is contained in the normalizer Nc(P) of P,
isP.
Theorem 48 In a group G of order n, the number of Sylow p-subgroups is a
divisor of n of the form (1 + kp ), k an integer.
Proof: Let P
0
,P
1
,P
2
, ... , Ps be the Sylow p-subgroups of G. If s = 0,
the theorem is clear. Suppose s -=1= 0 and partition P
1
, P
2
, , Ps into disjoint
conjugate classes with respect to elements of p 0. That is, if there is an element X
in P
0
such that
1
Pix =Pi, then Pi and Pi are in the same class. Since Pi is the
only Sylow p-subgroup in its normalizer (Exercise 50), it follows that Np
0
(Pi),
i -=1= 0, is a proper subgroup of P
0
(Np
0
(PDt;;;, Nc(PD). Now the index of Np
0
(P;)
in P
0
is the number of conjugates of Pi by elements of P
0
(Theorem 16), and is a
power of p. Thus we may write s =pet + pe 2 + +pet = kp, which implies
that there are 1 + s = 1 + kp Sylow p-subgroups of G. Since the Pi are all con-
jugate (under G), the number of Sylow p-subgroups is [G :Nc(P
0
)]. Therefore
1 + kp is a divisor of n.
Note: Theorems 42, 47, and 48 are often referred to as the Sylow the-
orems.
Remark 30 Let G be a group and P a Sylow p-subgroup of G. P is the only
Sylow p-subgroup of G if, and only if, Pis normal in G.
Proof: Suppose P G. If P' is a Sylow p-subgroup of G, then P and P'
are conjugate. Thus, there exists an element x in G such that x-
1
Px = P'. But
x-
1
Px = P for all x in G since P is normal. Thus P' = P. Conversely, if P is the
only Sylow p-subgroup of G, then x-
1
Px = P for all x in G, implying that P is
normal in G.
Theorem 49 If the Sylow subgroups of a finite group G are normal, then G is
the direct product of its Sylow subgroups.
Proof: Suppose G has order n = p
1
et p
2
e
2
p/ s, Pi prime. There can be
only one Sylow Prsubgroup of G fori= I, 2, ... , s since the Sylow subgroups of
G are normal (Remark 30). Let Pi be the Sylow Prsubgroup of G. Now P
1
P
2
Ps is a subgroup of G (Exercise 18). In fact, P
1
P
2
Ps = G (since Pi has
order p{i andPinPi= !el, i-=Fj). Thus G=P
1
XP
2
X XPs (generaliza-
tion of Theorem 37').
44 p-Groups
Remark 31 If the Sylow subgroups of G are normal and cyclic, then G itself
must be cyclic (see Exercise 42). Thus we see that there can be but one Abelian
group of order p
1
p
2
Ps,Pi distinct primes; namely the cyclic group of this
order.
In Chapter 4 we shall discuss groups of order p
1
p
2
Ps, Pi prime, in more
detail. Of particular interest will be groups of order pq, p and q distinct primes.
It can be shown that there are no simple groups of order p
1
p
2
Ps, s > 1
(pi distinct primes). We do this here for a group of order 30.
Example 37 Suppose G is a group of order 30 = 235. Now if G is a simple
group, then the Sylow subgroups of G are not unique. Thus G has at least 3
Sylow 2-subgroups, 10 Sylow 3-subgroups, and 6 Sylow 5-subgroups. This im-
plies that G must contain at least 48 distinct elements. But G has order 30.
Therefore G is not simple.
Exercise 51 Suppose G is a group of order 56. Show that G cannot be a sim-
ple group.
Theorem SO Suppose G is a group, P a Sylow p-subgroup of G, and N
0
(P) the
normalizer of Pin G. If His a subgroup of G such that N
0
(P) ~ S: G, then
H=N
0
(H).
Proof" Let x be an element of G such that
1
Hx =H. We must show
that x belongs to H. Now x-
1
Px is a Sylow p-subgroup of H, say P'. Thus,
there exists an element h in H such that h-
1
P'h = P (Theorem 47). It follows
that h-
1
x-
1
Pxh = P, and so xh is an element of N
0
(P). Since N c(P) S: H, and
h is in H, we have x is in H.
Theorem 51 Let G be a finite group and Pap-subgroup of G. If Pis not a
Sylow p-subgroup then N c(P) properly contains P.
Proof: This is immediate from Theorem 42, since if P has order ps, then
P must be normal in some subgroup P' of order ps+l.
Exercise 52 Suppose P is a normal p-subgroup of a group G. Show that Pis
contained in every Sylow p-subgroup of G. (See Example 36.)
3-4. NILPOTENT GROUPS
In Sec. 2-3 the concepts of normal series and composition series of a group
were discussed. In this section of Chapter 3 we introduce the concept of a cen-
tral series for a group, and from this shall give a brief discussion of nilpotent
groups.
p-Groups
45
Definition36 LetGbeagroup. Aseries lel =H
0
{;,H
1

where Hi/Hi_
1
c:;Z(G/Hi-d for i = 1,2, ... , s, is called a central series. The
series le l = H
0
{;, H1 {;, H2

Hi+dHi = Z(G/Hi),
is called the ascending central series for G.
Example 38 The series I I l C ! l,R'I C I I,R,R',R''I C G, where G is the group
of symmetries of the square, is a central series for G. Since Z(G) = l!,R'l,we
have ll,R')/111 =Z(G/Ill). Since IG/II,R'II = 2
2
, G/II,R'l is Abelian. Thus
!I, R, R', R"ljii, R'l R'l). Clearly, G/ll, R, R', R"l =Z(G/
I!,R,R',R"l ). It is to be noted that \I,R, R',R "l/ll,R' l =F Z(G/ !I,R'l), and
therefore the series is not the ascending central series. However, the series
Ul C !I,R' I C G is the ascending central series for G.
Exercise 53 Find all central series for the group of symmetries of the square.
Definition 37 A group G is called nilpotent if it possesses a finite central series.
Since the group G of symmetries of the square possesses a finite central ser-
ies, G is nilpotent. We shall shortly prove that this is true for all finite p-groups.
The reader may recall that it has been shown that a group G is solvable if,
and only if, it possesses a finite normal series in which every factor is Abelian
(Theorem 36). Since the factors of a central series are clearly Abelian, we see
that every nilpotent group is also solvable.
Exercise 54 Give an example of a solvable group that is not nilpotent.
Remark 32 Clearly, if the ascending central series 1 e} =H
0





is such that G =Hi for some i, then G is nilpotent. Suppose l e \ = H
0
{;, H
1
<;;::_
H 2 = G (where c is the least positive integer such that He = G). The
number cis called the class of the nilpotent group G.
We may always form the series lei <:.Z(G)!:. G. If G is non-Abelian, then,
in order for this series to be a central series, we must have leI C Z( G) and
G/Z(G) Abelian. But G/Z(G) cannot be cyclic (Remark 27). If G/Z(G) is
cyclic, then G is Abelian, and therefore nilpotent of class c = 1. If G is nilpotent
of class c = 1, then G is Abelian.
Theorem 52 If G is a nilpotent p-group of class c > 1, then [ G :He- d is di-
visible by p
2
.
Proof: Suppose lei CZ(G)CH2 !:.He-1 <:.He=G. Since He-d
He_
2
= Z(G/He_
2
), the result follows from Theorem 46.
Example 39 Consider the alternating group A
4
(see page 27). Since Z(A
4
) =
46
p-Groups
I P
1
! , there is no central series for A
4
(why?). Thus A
4
is not nilpotent. (Are
any of the groups An nilpotent?)
Theorem 53 Every finite p-group G is nilpotent.
Proof: The ascending central series I e } = H
0
c;:;_ H
1
!;:;, H
2
c;:;_ .. , must ter-
minate with Hi= G for some i, since Z(G) =I= {e l (Theorem 43). In fact, if IGI =
pn, then Hn =G. (Of course, G could be equal to Hi for some i < n, as is the
case in Example 38.)
Exercise 55 Prove that a direct product of finite p-groups is nilpotent.
Remark 33 It is true that if His a proper subgroup of the nilpotent group G,
then His also a proper subgroup of its normalizer N c(H). We do not prove this
fact here, but shall use it to prove the following important theorem.
Theorem 54 Let G be a finite group. G is nilpotent if, and only if, G is the di-
rect product of its Sylow subgroups.
Proof: Suppose G is nilpotent. Let P be a Sylow p-subgroup of G and
N c(P) be the normalizer of P. Now the normalizer of N c(P) is N c(P), (Theorem
50). Thus, by the above remark, N c(P) = G. Therefore Pis a normal subgroup
of G and we have that G is the direct product of its Sylow subgroups by Theorem
49.
Conversely, if G is the direct product of its Sylow subgroups, then G is nil-
potent (Exercise 55).
It is immediate from Remark 33 that if His a maximal subgroup of the nil-
potent group G, then H must be normal. Now G/H contains no proper sub-
groups (since His maximal) and thus is cyclic of prime order (Theorem 10). We
also note that H contains the derived group of G (why?).
Theorem 55 If every maximal subgroup of a finite group G is normal, then G
must be nilpotent.
Proof: N c(P) is its own normalizer by Theorem 50, where Pis a Sylow p-
subgroup of G. Now N c(P) is maximal, since otherwise we would haveN c(P) C
H C G where H is maximal. That this is not possible follows from Theorem 50
(H is normal by hypothesis and N c(P) cannot be contained in a proper normal
subgroup of G). It follows that Nc(P) = G, and thus Pis normal in G. Hence G
is the direct product of its Sylow subgroups and is therefore nilpotent.
Exercise 56 Suppose G is a group and consider/( G), the group of inner auto-
morphisms of G (see Theorem 26). If/( G)= I e l, prove that G is nilpotent.
Also, consider /(/(G)), the group of inner automorphisms of /(G). If /(/(G))=
{ e}, show that G is nilpotent.
p-Groups 47
3-5. THE FRATTINI SUBGROUP
Before leaving this chapter we consider the group G of symmetries of the
regular octagon. Here G = I R r. R
2
, R
3
, , R
16
l where R
1
, R
2
, , R
8
are
rotations of 45, 90, ... , 360 clockwise about the center 0 of the octagon,
and R
9
, R
10
, , R
16
are reflections in the axes and diagonals as indicated in
Fig. 2.
R16 . Rg
Rto
Rn
Rt4 Rt2
R13
Rt3
R12 Rt4
Ru
Rts
Rto
Rg
Rt6
Fig. 2
The maximal subgroups of G are 1Rr.R
2
,R
3
,R
4
,R
5
,R
6
,R
1
,R
8
l, I R
2
,R
4
,
R
6
,R
8
,R
9
,R
11
,R
13
,R tsl, and IR2,R
4
,R
6
,R
8
,R
1
o,R
12
,R
14
,R
16
1, and the inter-
section of these subgroups is the subgroup !R
2
,R
4
,R
6
,R
8
1.
Now if G is any group, the intersection of its maximal subgroups is a sub-
group of G. It is called the Frattini subgroup of G and will be denoted by F( G).
Thus, for our group G above, we have F(G) = !R
2
,R
4
,R
6
,R
8
l. We note that
F( G) is nilpotent since it is a p-group (Theorem 53).
In general, if G is any finite group, then its Frattini subgroup is nilpotent.
It is also true that ifF( G) contains the derived group G' of a finite group G then
G will be nilpotent, as we shall prove in our next theorem. Note that in the
group of the octagon (which is nilpotent since it is a p-group) we have G' k
F( G). In fact, in this example we have G' = F( G) since the derived group of the
group of symmetries of the octagon is also I R
2
,R4 ,R
6
,R
8
1, as the reader should
show.
Exercise 57 In the group of the octagon, the maximal subgroups are normal
(why?) and thus so is the intersection of the maximal subgroups (Exercise 15).
Prove or disprove: The Frattini subgroup of a group is a normal subgroup.
48
p-Groups
Exercise 58 In a non-Abelian p-group G of order p
3
we have Z (G) = G' (Re-
mark 29). Show that this is not necessarily true for a non-Abelian p-group of
order p
4
by finding the center of the group of the octagon.
Theorem 56 If the Frattini subgroup of a finite group G contains the derived
group of G, then G is nilpotent.
Proof Suppose Pis a Sylow p-subgroup of G. Now N
0
(P) =G. For sup-
pose otherwise. Then N
0
(P) k.H where His a maximal subgroup of G. Since
F( G) k. H (Why?) and G' k. F( G), G' k. H. But since G/G' is Abelian, we have
H G (Theorem 29). That this is not possible follows from Theorem 50. Hence
Na(P) =G. G is then the direct product of its Sylow subgroups (Theorem 49)
and therefore nilpotent (Theorem 54).
Thus we know that if F(G) contains G' then G is nilpotent. But when does
F( G) contain G'?
Theorem 57 The Frattini subgroup of a group G contains G' if, and only if, all
maximal subgroups of G are normal.
Proof: Let H
1
, H
2
, be the maximal subgroups of G and suppose each
is normal. Now the quotient groups G/H; are Abelian since 'G/H; I is a prime
(Theorem 1 0). Thus G' r;,_ H; for each i (Exercise 27), and so G' r;,_ F( G).
Conversely, suppose G' r;,_ F(G). If His a maximal subgroup of G, then
G' k.H (since F(G) r;,_ H). Thus His normal in G (Theorem 29).
Theorem 58 The Frattini subgroup of a nilpotent group G contains G'.
Proof" This is immediate since the maximal subgroups of G are normal
(see comment following Theorem 54).
Exercise 59 In the group G of symmetries of the octagon, note that the quo-
tient group G/F(G) is solvable and cyclic (why?). Prove that if G is any finite
group then G is solvable if, and only if, G/F(G) is solvable and that G is cyclic
if, and only if, G/F(G) is cyclic.
4
Groups of Special Order
4-1. GROUPS OF ORDER 4 AND 6
In this chapter we shall apply results found earlier to groups of specific
orders, and develop some additional theory. It is realized that much of this ma-
terial could well have been presented in previous sections, but it is preferred to
discuss it now.
We have seen that there are two groups of order p
2
(Exercise 49), and that
both must be Abelian (Remark 28). One of them is cyclic generated by, say a,
and in the noncyclic group all elements other than the identity must have order
p. Tables 10 and 11 are the group tables for the two groups of order 4. The sec-
ond of these is called the Klein four-group.
TABLEIO
2 3
e a a a
e e a
a2 a3
a a
a2 a3
e
a2 a2
a
3
e a
a3 a3
e a
a2
TABLE 11
e a b ab
e e a b ab
a a e ab b
b b ab e a
ab ab b a e
4!)
50 Groups of Special Order
Exercise 60 Show that the Klein four-group is a direct product of two cyclic
groups, each having order two (see Sec. 2-4).
There is at least one group of order 6; the cyclic ~ r u p of that order. Since
there can be only one Abelian group of order 6 (Remark 31 ), any other group of
order 6 would have to be non-Abelian. Question: Do non-Abelian groups of
order 6 exist, and if so, how many are there? (That is, how many nonisomorphic
non-Abelian groups of order 6 are there?)
Now if such a group does exist, then all elements (other than e) cannot have
order 2 (Exercise 1). Thus there exists an element, say x, of order 3. (The
reader should, at this point, recall Exercise 8.) Let y be an element other than x,
x
2
, and x
3
= e, and consider the set G = \e,x,x
2
,y,xy,x
2
y !. One easily shows
that the elements in G are distinct (do this!). The problem now is to see how
Table 12 should be filled in so that G will be a non-Abelian group of order 6.
TABLE 12
e X y xy
Since closure must hold, y
2
must be an element in G. Clearly y
2
cannot be
equal to xy or x
2
y. Thus it must be the case that y
2
= e, y
2
=x, or y
2
=x
2
.
y
2
= x or y
2
= x
2
each implies that y is an element of order 3. That is,y
3
=e.
But then we would have xy =e or x
2
y =e. Since neither of these is possible, we
conclude thaty
2
=e.
Now yx must also belong to G. In fact, it must be the case that yx = x
2
y
(why?). This implies that (xy)
2
=e. Thus we have that x
3
= y
2
= (xy)
2
= e, and
the "multiplication" table for G must be as shown.
TABLE 13
2 2
e X X
y xy X y
e e X
x2
y xy
x2y
X X
x2
e xy
x2y
y
x2 x2
e X
x2y
y xy
y y
x2y
xy e
x2
X
xy xy y
x2y
X e
x2
x2y x2y
2
xy y X X e
Groups of Special Order
51
It is to be pointed out that the associative property was used to form some
of the above products. Of course, the associative property must hold in all cases
in order for G to be a group. It does. Therefore G is a non-Abelian group of
order 6. It is clear that any other non-Abelian group of order 6 must be isomor-
phic to G.
Exercise 61 Show that S
3
(see Example 4) is isomorphic to the group G above.
4-2. GROUPS OF ORDER pq AND p
2
q
A group of order 6 is a specific example of a group of order pq where p and
q are distinct primes. We shall now briefly discuss groups of this type.
We know that there is only one Abelian group G of order pq; the cyclic
group of this order. Question: Is there a non-Abelian group of order pq? The
answer is no if p is not a factor of q - 1; as we now see.
Theorem 59 Let G be a group of order pq where p and q are primes with p < q.
If p does not divide q - 1, then G is cyclic.
Proof: The number of Sylow p-subgroups of G is 1 + kp (Theorem 48).
Since this number is a divisor of pq, and p is not a divisor of q - 1, it follows
that G contains only one Sylow p-subgroup, say P. Since the number of Sylow
q-subgroups of G is a divisor of pq of the form 1 + kq, and q does not divide
p- 1 (p < q), G contains only one Sylow q-subgroup, say Q. Now P and Q are
normal in G (Remark 30). Thus G = P X Q (Theorem 49). That G is cyclic
follows from Exercise 42.
Note: Theorem 59 does not apply to groups of order 6. In fact, it does
not apply to any group of order pq where p = 2. Yet for groups of order say, 15,
33, 51, 35, 65, or 77, it does, implying the existence of just one group for each
of these orders.
One should now wonder what the answer to our question is if p does
divide q - 1. Here, if G is to be non-Abelian, G must contain q Sylow p-sub-
groups (since otherwise it will have only one, implying that G is cyclic as in
Theorem 59). That such a group does indeed exist is not shown here. But we do
observe the following:
G has only one Sylow q-subgroup, say Q, and q Sylow p-subgroups. Let P
be a Sylow p-subgroup. Suppose Q is generated by x and P is generated by y.
Then xq = e, yP =e. Now Pis not normal (Remark 30), but Q is. Thus y-
1
xy is
in Q, and so f
1
xy =xi for some i. One can show that iP must be congruent to
1 mod q while i is not congruent to 1 mod q. That is, we have iP = 1 (modq)
52
Groups of Special Order
and i * I (mod q). The relations xq = e,yP = e,y-
1
xy =xi where iP =I (mod q)
and i * I (mod q) define our non-Abelian group G of order pq where p divides
q- 1.
We return briefly to the case p = 2 and q = 3. Suppose a is an element of
order 3 and b is an element of order 2. Then a
3
= e, b
2
=e. We want b-
1
ab = ai
where i
2
= 1 (mod 3) and i $ 1 (mod 3). It follows that 3 must divide i + 1, and
so i = 2,5,8, .... Thus b-
1
ab = a
2
This relation leads us to Table 14. (Compare
with Table 13.)
TABLEI4
2
b ab a
2
b e a a
e e a
a2
b ab a
2
b
a a
a2
e ab a
2
b b
a2 a2
e a a
2
b b ab
b b a
2
b ab e
a2
a
ab ab b a
2
b a e
a2
a
2
b a
2
b ab b
a2
a e
Exercise 62 Construct the non-Abelian group of order 10.
Theorem 60 Let G be a group of order p
2
q where p and q are primes, p < q,
and p does not divide q - 1. Then G is Abelian.
Proof: The number of Sylow p-subgroups of G is 1 + kp and divides p
2
q.
Since 1 + kp and p are relatively prime it follows that 1 + kp = I, or 1 + kp = q.
But p does not divide q - 1, implying that G contains only one Sylow p-sub-
group, say P. The number of Sylow q-subgroups is 1 + kq. Since 1 + kq and q
are relatively prime and 1 + kq is a divisor of p
2
q, we have 1 + kq = 1,p, or p
2

Now 1 + kq = p
2
implies kq = p
2
- 1, which implies that q divides p- 1 (which
cannot be since p < q) or q divides p + 1 (which also cannot be, since then we
must have q = 3 and p = 2). Therefore G contains only one Sylow q-subgroup,
say Q. Thus G = P X Q, and since P and Q are Abelian, so is G.
Exercise 63 In Theorem 60, suppose q < p and q is not a divisor of p
2
- 1.
Prove G is Abelian.
Note: Groups of the type described in Theorems 59 and 60, and in Exer-
cise 63 are not simple groups. Clearly, they are solvable (Example 29).
As a closing remark in this section we note that the results obtained above
for groups of order p
2
q do not apply to groups of order 12. There do exist 5
distinct groups of order 12, and the reader is asked to describe these in the fol-
lowing exercise. (Recall that we have discussed one of these in some detail pre-
viously; namely, A
4
.)
Groups of Special Order 58
Exercise 64 Show the existence of 5 nonisomorphic groups of order 12. (Hint:
If G is a group of order 12 that is not isomorphic to A
4
, then G contains an ele-
ment of order 6.)
4-3. GROUPS OF ORDER p
3
We now discuss groups of order p
3
, p a prime.
It is clear that there exists 3 nonisomorphic Abelian groups of order p
3

For suppose H is cyclic of order p
2
generated by a and K is cyclic of order p
generated by b. Then G = H X K is Abelian of order p
3
. If His cyclic of order
p generated by a, K is cyclic of order p generated by b, and M is cyclic of order
p generated by c, then G = H X K X M is Abelian of order p
3
. The third Abelian
group of order p
3
is the cyclic group of that order.
For the special case p = 2, we have 3 Abelian groups of order 8. Tables 15
and 16 are the multiplication tables for two of these groups.
We now determine the non-Abelian groups of order 8. Such a group cannot
TABLE 15
2 3
b ab a
2
b a
3
b e a a a
e e a
a2 a3
b ab a
2
b a
3
b
2
a3
a a a e ab a
2
b a
3
b b
a2
2 3
a a e a a
2
b a
3
b b ab
a3
3
a e a
a2
a
3
b b ab a
2
b
b b ab a
2
b a
3
b e a
a2 a3
ab ab a
2
b a
3
b b a
a2 a3
e
a
2
b a
2
b a
3
b b ab
a2
3
a e a
a
3
b a
3
b b ab a
2
b
a3
e a
a2
TABLE 16
e a b c ab ac be abc
e e a b c ab ac be abc
a a e ab ac b c abc be
b b ab e be a abc c ac
c c ac be e abc a b ab
ab ab b a abc e be ac c
ac ac c abc a be e ab b
be be abc c b ac ab e a
abc abc be ac ab c b a e
54 Groups of Special Order
have an element of order 8, and all elements other than e cannot be of order 2
(why?). Thus there must be an element of order 4, say x. Suppose y is an ele-
ment other than e, x, x
2
, or x
3
, and consider the set G = l e,x,x
2
,x
3
,y,xy,x
2
y,
x
3
y l. Clearly,y
2
=I= y,xy,x
2
y or x
3
y. Ify
2
=x theny would be an element of
order 8. Likewise if y
2
= x
3
Thus we have the two possibilities: (i) y
2
= e or
(ii)y
2
=x
2

(i) Suppose y
2
=e. Now yx must be equal to x
2
y or x
3
y. (Show that if
yx=xy, we get the group of Table 15.) Foryx=x
2
y we obtainx=
y-
1
x
2
y, which implies that x
2
=e. Therefore, it must be the case that
yx = x
3
y (or that (xy)
2
=e). The relations x
4
= e,y
2
= e, and (xy)
2
= e
define a non-Abelian group of order 8. Table 17 is its multiplication
table.
TABLE 17
2 3 2 3
e X X X y xy X y X y
e e X
x2 x3
y xy
x2y x3y
2
x3
X X X e xy
x2y x3y
y
x2
2
X
x3
e X
x2y x3y
y xy
x3
3
X e X
x2 x3y
y xy
x2y
y y
x3y x2y
xy e
x3 x2
X
xy xy y
x3y x2y
X e
x3 x2
x2y x2y
xy y
x3y x2
X e
x3
x3y x3y x2y
xy y
x3 x2
X e
Exercise 65 Show that the group of symmetries of the square (Example 3) is
isomorphic to the above group.
(ii) Suppose y
2
= x
2
Now yx must be x
2
y or x
3
y. (If yx = xy, we again
obtain the group of Table 15.) For yx = x
2
y (where x
2
= y
2
) we have
yx = y
3
, which implies that x = y
2
= x
2
Therefore it must be the case
thatyx=x
3
y(or(xy)
2
=e). Therelationsx
4
=e,x
2
=y
2
,and(xy)
2
=
e define a second non-Abelian group of order 8. This group is called the
quaternion group and Table 18 is its multiplication table.
So far, we have seen that there are three Abelian groups of order p
3
where p
is any prime, and two non-Abelian groups when p = 2. It remains to see how
many non-Abelian groups G exist of order p
3
where p =I= 2. As in the case p = 2,
there are also two. We discuss one of these here.
In either of the groups, if G is to be non-Abelian it does not contain an ele-
ment of order p
3
Suppose all elements other than e have order p. Now Z(G)
has order p (Theorem 46). The quotient group G/Z(G), having order p
2
, must
be defined by the relations aP = e, bP = e, ab = ba. (G/Z(G) is not cyclic by
Groups of Special Order
55
TABLE18
2 3 2 3
e X X X y xy X y X y
e e X
x2 x3
y xy
x2y x3y
X X
x2 x3
e xy
x2y x3y
y
x2 x2
X
3
e X
x2y x3y
y xy
x3 x3
e X
x2
x3y
y xy
x2y
y y
x3y x2y
xy
x2
X e
x3
xy xy y
x3y x2y x3 x2
X e
x2y x2y
xy y
x3y
e
x3 x2
X
x3y x3y x2y
xy y X e
x3 x2
Remark 27. Also see Exercise 49.) If we consider the "natural homomorphism"
(see note following Theorem 21) <.p: G--+ G/Z(G), where <.p(x) =a and <.p(y) = b,
then we have that <.p(x
1
y-
1
xy) =
1
b-
1
ab =e. This implies that x-
1
y-
1
xy is in
the kernel of <.p, which is Z(G). Now x
1
y-
1
xy =Fe, since otherwise G would be
an Abelian group (why?) Therefore x
1
y-
1
xy is an element of order p and gener-
ates Z(G). If we set x
1
f
1
xy = z, then we have that xy = yxz, xz = zx, andyz =
zy. The relations xP = e, yP = e, zP = e, xy = yxz, xz = zx, and yz = zy define
our non-Abelian group of order p
3
, p =F 2.
Exercise 66 Construct a non-Abelian group of order 27.
Returning briefly to the quaternion group G defined by Table 18, we note
that the following are the subgroups of this group: \ e l, H = \ e, x
2
l , K
1
=
\e,x,x
2
,x
3
l, K
2
= \e,x
2
,y,x
2
yl, K
3
= \e,xy,x
2
,x
3
yl, and G. Now K
1
, K
2
,
and K
3
are normal in G since [ G: K;] = 2. H is normal in G since H = Z( G).
Thus we have our first example of a non-Abelian group with the property that
all of its subgroups are normal. Such a group is called a Hamiltonian group.
We remark that by a generalized quaternion group is meant a group of order
2n(n > 2) generated by elements x andy such that x2nl = e, yxy-
1
= x-
1
, and
x2n--2 = y2.
Exercise 67 Find an example of a Hamiltonian group other than the quaternion
group.
4-4. GROUPS OF ORDER 16
Thus far we have seen the existence of but one group of order 2, two groups
of order 4, and five groups of order 8. In this section we discuss the fourteen
(e, e)
(e,a)
(e,b)
(e,ab)
(x,e)
(x,a)
(x,b)
(x,ab)
(x
2
,e)
(x
2
,a)
(x
2
, b)
(x
2
,ab)
(x
3
, e)
(x
3
,a)
(x
3
,b)
(x
3
,ab)
TABLE 19
(e,e) (e,a) (e,b) (e,ab) (x,e) (x,a) (x,b) (x,ab) (x
2
,e) (x
2
,a) (x
2
,b) (x
2
,ab) (x
3
,e) (x
3
,a) (x
3
,b) (x
3
,ab)
(e,e) (e,a) (e,b) (e,ab) (x,e) (x,a) (x,b) (x,ab) (x
2
,e) (x
2
,a) (x
2
,b) (x
2
,ab) (x
3
,e) (x
3
,a) (x
3
,b) (x
3
,ab)
(e,a) (e,e) (e,ab) (e,b) (x,a) (x,e) (x,ab) (x,b) (x
2
,a) (x
2
,e) (x
2
,ab) (x
2
,b) (x
3
,a) (x
3
,e) (x
3
,ab) (x
3
,b)
(e,b) (e,ab) (e,e) (e,a) (x,b) (x,ab) (x,e) (x,a) (x
2
,b) (x
2
,ab) (x
2
,e) (x
2
,a) (x
3
,b) (x
3
,ab) (x
3
,e) (x
3
,a)
(e,ab) (e,b) (e,a) (e,e) (x,ab) (x,b) (x,a) (x,e) (x
2
,ab) (x
2
,b) (x
2
,a) (x
2
,e) (x
3
,ab) (x
3
,b) (x
3
,a) (x
3
,e)
(x,e) (x,a) (x,b) (x,ab) (x
2
,e) (x
2
,a) (x
2
,b) (x
2
,ab) (x
3
,e) (x
3
,a) (x
3
,b) (x
3
,ab) (e,e) (e,a) (e,b) (e,ab)
(x,a) (x,e) (x,ab) (x,b) (x
2
,a) (x
2
,e) (x
2
,ab) (x
2
,b) (x
3
,a) (x
3
,e) (x
3
,ab) (x
3
,b) (e,a) (e,e) (e,ab) (e,b)
(x,b) (x,ab) (x,e) (x,a) (x
2
,b) (x
2
,ab) (x
2
,e) (x
2
,a) (x
3
,b) (x
3
,ab) (x
3
,e) (x
3
,a) (e,b) (e,ab) (e,e) (e,a)
(x,ab) (x,b) (x,a) (x,e) (x
2
,ab) (x
2
,b) (x
2
,a) (x
2
,e) (x
3
,ab) (x
3
,b) (x
3
,a) (x
3
,e) (e,ab) (e,b) (e,a) (e,e)
(x
2
,e) (x
2
,a) (x
2
,b) (x
2
,ab) (x
3
,e) (x
3
,a) (x
3
,b) (x
3
,ab) (e,e) (e,a) (e,b) (e,ab) (x,e) (x,a) (x,b) (x,ab)
(x
2
,a) (x
2
,e) (x
2
,ab) (x
2
,b) (x
3
,a) (x
3
,e) (x
3
,ab) (x
3
,b) (e,a) (e,e) (e,ab) (e,b) (x,a) (x,e) (x,ab) (x,b)
(x
2
,b) (x
2
,ab) (x
2
,e) (x
2
,a) (x
3
,b) (x
3
,ab) (x
3
,e) (x
3
,a) (e,b) (e,ab) (e,e) (e,a) (x,b) (x,ab) (x,e) (x,a)
(x
2
,ab) (x
2
,b) (x
2
,a) (x
2
,e) (x
3
,ab) (x
3
,b) (x
3
,a) (x
3
,e) (e,ab) (e,b) (e,a) (e,e) (x,ab) (x,b) (x,a) (x,e)
(x
3
,e) (x
3
,a) (x
3
,b) (x
3
,ab) (e,e) (e,a) (e,b) (e,ab) (x,e) (x,a) (x,b) (x,ab) (x
2
,e) (x
2
,a) (x
2
,b) (x
2
,ab)
(x
3
,a) (x
3
,e) (x
3
,ab) (x
3
,b) (e,a) (e,e) (e,ab) (e,b) (x,a) (x,e) (x,ab) (x,b) (x
2
,a) (x
2
,e) (x
2
,ab) (x
2
,b)
(x
3
,b) (x
3
,ab) (x
3
,e) (x
3
,a) (e,b) (e,ab) (e,e) (e,a) (x,b) (x,ab) (x,e) (x,a) (x
2
,b) (x
2
,ab) (x
2
,e) (x
2
,a)
(x
3
,ab) (x
3
,b) (x
3
,a) (x
3
,e) (e,ab) (e,b) (e,a) (e,e) (x,ab) (x,b) (x,a) (x,e) (x
2
,ab) (x
2
,b) (x
2
,a) (x
2
,e)




.
"'

/J:)
"!;:!
ill
s
.....
c

Groups of Special Order 57
groups of order 16. (There are 51 groups of order 32, and 267 groups of order
64.)
Abelian Groups of Order 16
Let H 1 be the cyclic group of order 2, H
2
the cyclic group of order 4, H
3
the noncyclic group of order 4 (Table 11), and H
4
the cyclic group of order 8.
There are four distinct Abelian groups of order 16 besides the cyclic group of
that order. These may be described as follows: G
1
= H
1
X H
1
X H
1
X H
1
, G
2
=
H2 X H2, G3 =H
2
X H
3
, G
4
=H
1
X H
4
. Table 19 is the group table for G
3
.
Exercise 68 If H
5
and H
6
are the groups of Tables 17 and 18 respectively, and
Gs =H1 X Hs, G6 =H1 X H
6
, then G
5
and G
6
are Abelian groups of order 16.
Show that G
5
is isomorphic to one of the groups Gi above, as is G
6
.
Non-Abelian Groups of Order 16
In describing these groups we distinguish two cases: (i) the center has order
2, and (ii) the center has order 4. The center cannot have order 8 by Theorem
45. (See also Remark 27.)
(i) In the group G of symmetries of the regular octagon (Sec. 3-5) we
note that R
4
2
=R
8
, R
2
2
=R
4
, R
3
2
=R
2
-
1
, R
9
2
=R
8
, R
4
=R
2
-
1
R
9
-
1
R
2
R
9
,
and R
2
=R
3
-
1
R
9
-
1
R
3
R
9
These relations completely determine the group of
the octagon. (The reader should show this.) More generally, the relations a
2
=
e, b
2
=a, e
2
= b-
1
, d
2
= e, a= b-
1
a
1
bd, b =e-
1
a
1
ed, determine a non-Abelian
group of order 16. Its group table is given in Table 20.
The relations a
2
= e, b
2
=a, e
2
= !J
1
, d
2
=a, a= b-
1
a
1
bd, and b = e-
1
a
1
ed
also define a non-Abelian group of order 16 (whose center has order 2), as do the
relations a
2
=e, b
2
=a, e
2
=b, d
2
=e, a=b-
1
a
1
bd, and b=e-
1
a
1
ed. There
are three non-Abelian groups of order 16 whose centers have order 2.
(ii) We now discuss the non-Abelian groups of order 16 that have a center
of order 4. Here we again distinguish two cases: (1) the center is cyclic (Table
1 0) and (2) the center is noncyclic (Table 11 ).
1. There are two non-Abelian groups of order 16 having a center which is
the cyclic group of order 4. These may be described as follows: Let x be a
generator for the center. The relations a
2
= b
2
= e
2
= e, x
2
=a, and a= b-
1
c
1
be
(or (be)
2
=a) determine one of the groups; a
2
= b
2
= e, e
2
= x, x
2
=a, and a=
b-
1
e-
1
be determine the other. Table 21 is the group table for the first of these
groups.
2. There are four non-Abelian groups of order 16 having a center which is
noncyclic of order 4. The relations a
2
= e, b
2
= y, x = a
1
b-
1
ab, and the relations
a
b
c
d
ab
ac
ad
be
bd
cd
abc
abd
acd
bed
abed
e
TABLE 20
a b c d ab ac ad be bd cd abc abd acd bed abed e
e ab ac ad b c d abc abd acd be bd cd abed bed a
ab a be bd e abc abd ac ad bed c d abed acd cd b
ac be ab cd abc b acd e bed abd a abed bd d ad c
ad abd bed e bd abed a cd ab be acd b abc c ac d
b e abc abd a be bd c d abed ac ad bed cd acd ab
c abc b acd be ab cd a abed bd e bed abd ad d ac
d bd abed a abd bed e acd b abc cd ab be ac c ad
abc ac e bed c a abed b acd d ab cd ad bd abd be
abd d acd b ad cd ab bed e ac abed a c be abc bd
acd abed d c bed ad ac abd abc e bd be a ab b cd
be c a abed ac e bed ab cd ad b acd d abd bd abc
bd ad cd ab d acd b abed a c bed e ac abc be abd
cd bed ad ac abed d c bd be a abd abc e b ab acd
abed cd bd be acd abd abc d c b ad ac ab e a bed
bed acd abd abc cd bd be ad ac ab d c b a e abed
a b c d ab ac ad be bd cd abc abd acd bed abed e

(X;

....
0
.::
'!;:3
""

1:/)
'!;:3
"'
"' '""
..
0
....
R..
"' ....
e a b c X ab
e e a b c X ab
a a e ab ac ax b
b b ab e be bx a
c c ac abc e ex be
X X ax bx ex a abx
ab ab b a abc abx e
ac ac c be a a ex abc
ax ax X abx a ex e bx
be be abc ac b bcx c
bx bx abx X bcx ab ax
ex ex a ex abcx X ac bcx
abc abc be c ab abcx ac
abx abx bx ax abcx b X
a ex a ex ex bcx ax c abcx
bcx bcx abcx a ex bx abc ex
abcx abcx bcx ex abx be a ex
TABLE 21
ac ax be bx ex
ac ax be bx ex
c X abc abx a ex
abc abx c X bcx
a a ex ab abcx X
a ex e bcx ab ac
be bx ac ax abcx
e ex b bcx ax
ex a abcx b c
ab abcx a a ex bx
abcx b ex a abc
ax c abx be a
b bcx e ex abx
bcx ab a ex e be
X ac bx abc e
abx be ax c ab
bx abc X ac b
abc abx acx
abc abx a ex
be bx ex
ac ax abcx
b bcx ax
abcx b c
c X bcx
ab abcx X
bcx ab ac
e ex abx
a ex e be
bx abc e
a a ex bx
ex a abc
abx be a
X ac b
ax c ab
bcx
bcx
abcx
ex
abx
abc
a ex
bx
be
ax
ac
b
X
c
ab
e
a
abcx
abcx
bcx
a ex
bx
be
ex
abx
abc
X
c
ab
ax
ac
b
a
e
G"!
..,
0
;:::

""

VJ

"'
(":>
""'
..
a
..,

"'
..,
'"
(0
e
a
b
X
y
ab
ax
ay
bx
by
xy
abx
a by
axy
bxy
abxy
TABLE 22
e a b X y ab ax ay bx by xy abx a by axy bxy abxy
e a b X y ab ax ay bx by xy abx a by axy bxy abxy
a X ab ax ay bx e xy abx a by axy b bxy y abxy by
b abx X bx by a ab abxy e xy bxy ax ay a by y axy
X ax bx e xy abx a axy b bxy y ab abxy ay by a by
y ay by xy e a by axy a bxy b X abxy ab ax bx abx
ab b ax abx a by X bx by a axy abxy e xy bxy ay y
ax e abx a axy b X y ab abxy ay bx by xy a by bxy
ay xy a by axy a bxy y X abxy ab ax by bx e abx b
bx ab e b bxy ax abx a by X y by a axy abxy xy ay
by abxy xy bxy b ay a by abx y X bx axy a ab e ax
xy axy bxy y X abxy ay ax by bx e a by abx a b ab
abx bx a ab abxy e b bxy ax ay a by X y by axy xy
a by by axy abxy ab xy bxy b ay ax abx y X bx a e
axy y abxy ay ax by xy e a by abx a bxy b X ab bx
bxy a by y by bx axy abxy ab xy e b ay ax abx X a
abxy bxy ay a by abx y by bx axy a ab xy e b ax X

c
\;')
....
a
"" "l:::l
"'

V:;
"l:::l
"'
"' ""
12..
0
....
R..
"' ....
Groups of Special Order 61
a
2
= x, b
2
= y, x = a-
1
b-
1
ab determine two of these groups where x and y
are generators for the center (x
2
= y
2
=e). The remaining two groups are de-
termined by the relations a
2
= x, b
2
= e, x = a-
1
b-
1
ab, x
2
= y
2
= e, and the
relations a
2
= b
2
= x, x = a-
1
b-
1
ab, x
2
= y
2
= e. Table 22 is the group table for
the second of these groups.
Exercise 69 One of the groups defined by Tables 21 and 22 is Hamiltonian.
Which one? (See Exercise 67.)
5
Galois Theory
5-1. DEFINITION OF A FIELD
EXTENSION FIELDS
In Chapter 2 we mentioned that the general nth degree equation is not
solvable by radicals unless the symmetric group Sn is a solvable group (see
Remark 23). It is our purpose here in this final chapter to discuss this fact in
some detail. To do so will necessitate a discussion of a mathematical system
known as a field.
Although entire books have been written on field theory, we confine our
discussion here to a brief introduction, and mention only those facts needed to
relate the concepts of a field, a group, and solutions to polynomial equations.
Because we wish not to get too involved with field theory as such, several state-
ments are made without proof. For the interested reader, these proofs may be
found in most books dealing with the theory of fields.
Definition 38 Let K be a set containing at least two elements and suppose
that there are two binary operations which are denoted by + and defined on the
elements of K. We call K a field with respect to+ and if
1. a+ b = b +a for all a and bin K.
2. (a+ b)+ c =a+ (b +c) for all a, b, and c inK.
3. K contains an element denoted by 0 such that a+ 0 =a for all a inK.
(0 is called the zero element, or additive identity element of K.)
4. For each element a inK there is an element denoted by -a inK such that
a+ -a= 0. (-a is called the additive inverse of a.)
5. ab = ba for all a and bin K.
6. (ab)c = a(bc) for all a, b, and c inK.
64 Galois Theory
7. K contains an element denoted by 1 such that a 1 =a for all a in K.
(1 is called the unity element, or multiplicative identity of K.)
1
8. For each element a =I= 0 inK, there is an element denoted by a-
1
(or-)
a
inK such that a
1
= 1. (a-
1
is called the multiplicative inverse of a.)
9. a(b +c)= ab + ac for all a, b, and c inK. (Distributive property.)
Note: The elements of a field form a group with respect to+, and the ele-
ments =I= 0 form a group with respect to
Definition 39 A subset F of a field K is a sub field of Kif F is itself a field rela-
tive to the binary operations defined inK.
Remark 34 To show that F is a subfield of the field K, we need only show that
a+ b, a b, and -a are in F whenever a and bare in F, and that a-
1
is in F when-
ever a =I= 0 is in F.
Exercise 70 Show that if F
1
and F
2
are subfields of the field F then F
1
n F
2
is also a subfield of F.
Definition 40 Let F and K be fields. IfF is a subfield of K, we call K an exten-
sion field of F.
Example 40 The set C of complex numbers, the set R of real numbers, and the
set Q of rational numbers are all examples of fields with respect to the usual
operations of+ and . Cis an extension field of R and an extension field of Q.
R is an extension of Q.
Example 41 Let K = ! 0,1 ,2, ... , p- 1! where p is any prime and suppose +
and are defined on K as being ordinary addition and multiplication "reduced"
modulo p. (See Example 11.) Then K is a field. (If a field consists of a finite
number of elements it is called a Galois field.)
Example 42 Suppose Q is the field of rational numbers and let Q(v'2) =
fa+ bVl; a and b are in Q \. Q(Vl) is a subfield of the field R of real numbers
(the reader should show this) and is called the subfield of R obtained by adjoin-
ing V2 to the field Q.
Remark 35 Suppose K is an extension field ofF and let a be an element of K.
Let S be the collection of all subfields of K that contain F and a. The intersec-
tion of all elements of S shall be denoted by F(a). Now F(a) is a subfield of K
(Exercise 70). It is the smallest subfield of K that contains both F and a, and we
call it the sub field obtained by adjoining a to F.
Exercise 71 Show that (Q(y'2))(v'3) is a subfield of the field R of real num-
Galois Theory 65
bers. Here, (Q(y'2))(V3) = I c + dJ3; c and dare in Q(y'2) \. Is (Q(Vl))( v'3) =
Q(VI, v'3)?
Suppose F is a field and x an indeterminate. By F[x] we shall mean the set
consisting of all expressions of the form a
0
+a
1
x + a
2
x
2
+ + anxn, where the
elements a; are in F and n is any nonnegative integer. Such an expression is
called a polynomial in x over F. If we set a
0
+ a
1
x + + anxn = j(x), and an*
0, then the integer n is called the degree of the polynomial f(x ). (If a; = 0 for all
i = 1, 2, ... , n, we say the polynomial has degree 0.)
One can show that with "suitable" definitions for + and in the set F[x],
(for j(x) = a
0
+ a
1
x + a
2
x
2
+ + anxn and g(x) = b
0
+ b
1
x + b
2
x
2
+ +
bmxm, let j(x) + g(x) = c
0
+ c
1
x + c
2
x
2
+ + CyXr where c; =a;+ b;, and let
f(x) g(x) = d
0
+ d
1
x + d
2
x
2
+ + dsXs where d; = a;do + a;_
1
d
1
+ + a
0
d;),
that all properties except property (8) of Definition 38 are satisfield. Thus it is
to be remembered thatF[x] is not a field. (We remark here that if a setS is such
that properties (1 )-( 4), (6), and (9) of Definition 38, together with the fact that
(b + c)a = ba + ca for all a, b, and c inS hold, then Sis called a ring. Hence F[x]
is a ring. In fact, we can discuss S [x] when Sis not a field, but a ring, and still
S [x] will be a ring.)
As a generalization of the above, suppose F is a field with x
1
and x
2
indeter-
minates. We can construct (F[xt] )[x
2
] which is the set (ring) of the polyno-
mials b
0
+ b
1
x
2
+ + bmx
2
m where the b; are in F[x
1
]. That is, eachb; above
is an expression of the form a
0
+ a
1
x
1
+ + anx
1
n. A natural question to ask
here would be: is (F[xd )[x
2
] the same ring as the ring (F[x
2
]) [xd? Or more
precisely, are these rings isomorphic? The answer is yes, and we denote these
rings by F[x
1
,x
2
]. Generalizing further, we are led to the ring F[x
1
,x
2
, .. ,
Xn] of polynomials in x
1
, x
2
, , Xn over the field F.
In F[x], if we form the set F(x) of all g(x) * 0, then with
"suitable" definitions for + and , F(x) is a field; the field of quotients of F[x].
Generalizing, by F(x
1
,x
2
, , Xn) we shall mean the field of quotients of
F[x
1
,x
2
, , Xn]. F(x
1
,x
2
, ... , Xn) is called the field of rational functions in
X1,X2, ... ,Xn over F.
5-2. VECTOR SPACES
At this time we pause to give a brief discussion of vector spaces.
Definition 41 Let F be a field with respect to + and and suppose Vis a non-
66
Galois Theory
empty set having defined on its elements a binary operation * We call Va vec-
tor space over the field F if
1. Vis an Abelian group with respect to*
2. For a in F and x in V, there is an element which we denote by ax and ax
is in V.
3. For all a and b in F and for all x andy in V, we have a(X* y) =ax *ay,
(a+ b)x =ax* bx, a(bx) =(a b)x, and 1x = x. (1 is the unity of F.)
Note: The elements of V are usually called vectors and the elements ofF
are called scalars.
Example 43 Suppose R is the field of real numbers and let V = {(a
1
, a
2
, a
3
);
a; is in R l. We say that(a
1
,a
2
,a
3
) = (b
1
,b
2
,b
3
) if ai = bi for each i. We define
* on Vas follows: (a
1
,a
2
,aJ * (b1,b
2
,b
3
) = (a
1
+ bha
2
+ b
2
,a
3
+ b
3
) where
the " + " is ordinary addition for real numbers. If a is in R and x =(a
1
, a
2
, a
3
) is
in V, we define ax to be (a a
1
, a a
2
, a a
3
) where the " " is ordinary mul-
tiplication of real numbers. Note that ax is an element of V. Vis a vector space
over R.
Exercise 72 Show that ifF is a field, then F[x] may be regarded as a vector
space over F.
Definition 42 Suppose Vis a vector space over the field F and let x
1
, x 2 , .. ,
Xn beelementsofV. Any elementoftheforma
1
x
1
+a
2
x
2
++anXn,where
the ai are in F is called a linear combination of the xi. (We are using + in place of
*.)
The elements x
1
, x
2
, .. ,Xn are said to be linearly dependent if there exist
elements a
1
, a
2
, . , an in F, not all 0, such thata
1
x
1
+a
2
x
2
+ +anXn is
equal to the identity element of V. If the elements x
1
, x
2
, . , Xn are not
linearly de pendent we shall say that they are linearly independent.
Definition 43 A subset S of the vector space V is called a basis of V if the
elements of S are linearly independent and every element of V can be written as
a linear combination of the elements inS.
Definition 44 A vector space V is said to be finite-dimensional if it has a basis
S with a finite number of elements.
Remark 36 If V is a finite-dimensional vector space, then all bases of V have
the same number of elements. The number of elements in a basis is called the
dimension of V (over the field F).
Example 44 For the vector space of Example 43, the elements x
1
= (1,1,2),
x
2
= (1,2,1), and x
3
= (3,4,1) form a basis. That x
1
,x
2
, and x
3
are linearly in-
Galois Theory
67
dependent is as follows: Suppose a
1
, a
2
, and a
3
belong toR and consider a
1
x
1
+
a2x2 + a3x3. Now if a1x1 + a
2
x
2
+ a
3
x
3
is equal to the identity element of V
(which is the element (0,0,0)) then we must have (a
1
,a
1
,2a
1
) + (a
2
,2a
2
,a
2
) +
(3a
3
,4a
3
,a
3
) = (0,0,0). This leads to the system of equations
a1 + a
2
+ 3a
3
= 0
a 1 + 2a2 + 4a3 = 0
2ai + a2 + a3 = 0
which has a
1
= a
2
= a
3
= 0 for its only solution.
Also, every element in V can be written as a linear combination of the
elements x
1
, x 2, and x". For suppose x = (b
1
,b
2
,b
3
) belongs to V. We must
find elements a
1
, a
2
, and a
3
in R such that a
1
x
1
+ a
2
x
2
+ a
3
x
3
= x, or such that
(a1,a1,2ar) + (a2,2a2,a2) + (3a3,4a
3
,a3) = (b1,b
2
,b
3
). That a
1
, a
2
, and a
3
do
exist follows by solving the system of equations
a1 + a
2
+ 3a
3
= b
1
a1 + 2a2 + 4a
3
= b
2
2ai + az + a3 = b3
The elements y
1
= (1 ,0,0), y
2
= (0, 1 ,0), and y
3
= (0,0, 1) also form a basis
for V. The dimension of V (over R) is 3.
5-3. SPLITTING Fl ELDS
Returning now to our discussion of fields, it is clear that if K is an extension
field ofF, then we may regard K as a vector space over F. By the degree of K
over F, we shall mean the dimension of K as a vector space over the field F. We
denote the degree of Kover F by [K:F]. If [K:F] is finite (that is, if K has a
basis with a finite number of elements) then we call K a finite extension of F.
Exercise 73 Find [C:R], [C:Q], [R:Q], [Q(v/2):Q] ,[(Q(v/2))(YJ):Q(y'2)],
and [(Q(v/2))(YJ):Q].
After working with Exercise 73, the reader has perhaps guessed the content
of the following theorem.
Theorem 61 If K is a finite extension of E and E is a finite extension ofF, then
K is a finite extension ofF and we have [K:F] = [K:E] [E:F].
Proof: We shall give an outline of the proof; details are left to the reader.
Let lx
1
, ... ,Xn l be a basis forK over E and jy
1
, ... Ym\ a basis forE over
F. The elements xiYi (i = 1, ... , n, j = 1, ... , m) are linearly independent over
F, and every element in K is a linear combination of these elements. Thus the
nm elements XiJlj form a basis forK over F.
68
Galois Theory
Definition 45 Suppose K is an extension field ofF and f(x) is an element of
F[x]. Let b belong to K. b is called a root off(x) if/(b) = 0. (Iff(x) =a
0
+
a1x + + anxn, thenf(b) = a
0
+ a
1
b + + anbn .)
The reader is surely familiar with the following theorem from elementary
algebra, known as the Factor theorem.
Theorem 62 If b inK is a root of f(x), where f(x) is in F[x], then x- b
divides (is a factor of) f(x). (The converse also holds.)
Note: If b is a root of f(x) then x- b divides f(x). Suppose (x- b)':'
divides f(x), but (x- b)m+l does not. Then we say that b is a root of f(x)
having multiplicity m.
The polynomial f(x) = x
3
+ 2 belongs to Q [x], but no root of f(x) belongs
to Q. However, there are extensions of Q that contain at least one root of f(x).
In fact, there is an extension of Q containing all roots of f(x). In general, if f(x)
belongs to F[x], one can find an extension field ofF containing a root of f(x),
and one can find an extension field ofF containing all roots of f(x). This will
now be shown. First, we define what is meant by an irreducible polynomial
over a field F.
Definition 46 Letf(x) be an element of F[x] and suppose we may writef(x) =
g(x )h (x) where g(x) and h (x) also are in F [x] . If this factorization of f(x) is
only possible when either g(x) or h (x) has degree 0, then we say that f(x) is an
irreducible polynomial over the field F.
Example 45 f(x) = x
2
+ 1 is irreducible over R, the field of real numbers, but
not over the field C of complex numbers, since in C[x] we may write f(x) ==
(x + i)(x- i). The polynomial f(x) = x
3
+ 2 is irreducible over the field Q of
rational numbers, but not over the field R, since in R [x] we have (x
3
+ 2) =
(x +

- Here, \12 denotes the real cube root of 2.
Theorem 63 Suppose j(x) is an irreducible polynomial of degree n i= 0 in
F[x] . There is an extension field K ofF such that
1. f(x) has a root inK, and
2. [K:F] = n.
We omit the proof of this theorem, but shall prove the following:
Theorem 64 Suppose f(x) is in F[x] where the degree of f(x) is n i= 0. There
is an extension K ofF containing a root of f(x).
Proof: This is immediate from Theorem 63. In fact, this is Theorem 63
if f(x) is irreducible. If f(x) is not irreducible over F, then we have f(x) =
g(x)h(x), where we may assume that g(x) is irreducible (why?). Now by (1) of
Theorem 63,g(x) has a root inK. But then so doesf(x). Why?
Galois Theory 69
Exercise 74 In Theorem 64 show that the degree of K over F must be less
than, or equal to, the degree of f(x).
Theorem 65 Suppose f(x) is a polynomial of degree n -=!= 0 in F [x] . There is an
extension K ofF such that f(x) has n roots in K. (A root of multiplicity m is
counted as m roots.) Furthermore, [ K: F] <:;;; n!.
Proof: There is an extension K
1
ofF such thatf(x) has a root b
1
in K
1
(Theorem 64), and [K
1
:F] <:;;; n (Exercise 74). Thus we may write f(x) =
(x - b r)h
1
(x) in K
1
[x] , where h
1
(x) has degree n - 1. There is an extension K
2
of K
1
such that h
1
(x) (and hence f(x)) has a root b
2
in K
2
and [Kz :K d .:;;;; n- 1.
We may write h
1
(x) = (x- b
2
)h
2
(x) in K
2
[x1, where h
2
(x) has degree n- 2. It
follows that f(x) = (x- br)(x- b
2
)h
2
(x). By a continuation of the above, we
reach an extension field Kn-l of Kn_
2
such that f(x) has a root bn-l in Kn-l
and [Kn-l :Kn-
2
1 <:;;; 2. In Kn-l [x1 we have f(x) = (x- br)(x- b2)
(x- bn-dhn_
1
(x). Clearly, hn_
1
(x) =x- a, and so a is a root of f(x). Thus if
we set K=Kn_
1
, then f(x) has n roots inK, and [K:F] = [Kn-l :Kn-
2
1
[Kn-2:Kn-31 [KI:F1 <:;;;23 n=n!.
Note: The reader may be wondering whether or not a polynomialf(x) of
degree n in F [x 1 can have more than n roots in some extension field of F. This
is not possible, as one can show by induction on the number n.
Reviewing the results just mentioned, if f(x) is a polynomial of degree n
over the field F, there is a finite extension K ofF such that K contains the n
roots b1 , b2, ... , bn off(x). Now by Theorem 62, we may writef(x) = (x- b
1
)
(x-b
2
)(x-bn) in K[x1. That is, in K[x1, the polynomialf(x) splits
(factors) into a product of factors each having degree 1. (Such a factor is called
linear.)
It is clear that there exists an extension K ofF such that f (x) splits into
linear factors over K and [K:F] is minimal.
Definition 47 Let f(x) be an element of F[x 1. A finite extension K ofF is
called a splitting field over F of f(x) if f(x) can be written as a product of linear
factors in K[x] and [K:F1 is minimal. (That is,f(x) splits into linear factors
over K, but does not split into linear factors over any subfield of K.)
Note: K is a splitting field over F off(x), wheref(x) has degree n, if K is a
finite extension having minimal degree over F and such that K contains n roots
off(x).
Example 46 Consider again the polynomial f(x) = x
3
+ 2 in Q [x 1 . Now
Q(-\1'2) is an extension of Q containing a root of f(x), but Q(-\1'2) is not
a splitting field over Q of f(x). The factorization of f(x) in (Q(-\1'2)) [x] is
(x + ( x
2
- ~ + where clearly, x
2
- ~ + is irreducible over
70
Galois Theory
Q(- Now by Theorem 65, a splitting field K over Q of x
3
+ 2 is such that
[K: Q] 6. Since [Q(-\12): Q] = 3 (the reader should show this by exhibiting
a basis consisting of 3 elements for over Q), it follows that K must be an
extension field of having degree 2 over and containing the
roots of the polynomial x
2
- \Y2X + Such an extension is (Q(-
(To see this, solve x
2
- ...J'2x the quadratic formula.) Therefore, a split-
ting field of x
3
+ 2 over Q is K = Q(-..ifl,iy?;).
Exercise 75 Find a basis for Q(--ifi,i0) over Q.
Remark 37 In Example 46 we found a splitting field of x
3
+ 2. One should
wonder if a polynomial f(x) can have more than one splitting field over the field
F. Essentially the answer is no, since it can be shown that if K
1
and K
2
are two
splitting fields for the polynomial f(x) over the field F, then K
1
and K
2
are iso-
morphic. In fact, there is an isomorphism n from K
1
onto K
2
such that n(x) =
x for all x in F. (By an isomorphism of a field F onto a field F' we mean a 1-1
mapping n from F onto F' such that n(x + y) = n(x) + n(y) and n(xy) = n(x )n(y)
for all x andy in F.)
5-4. THE GALOIS GROUP OF A POLYNOMIAL
FUNDAMENTAL THEOREM OF GALOIS THEORY
Now that some of the necessary background in field theory has been dis-
cussed, we are ready to return to a discussion of groups and their relationship
to the roots of a polynomial over a field. In what follows, we assume that all
fields are of characteristic 0. That is, if n is an integer and a ::1= 0 belongs to the
field, na = 0 implies n = 0.
Definition 48 Suppose K is a field and a is a 1-1 mapping from K onto itself
such that a(x + y) = a(x) + a(y) and a(xy) = a(x)a(y) for all x andy inK. We
call a an automorphism of K.
Note: An automorphism of K is an isomorphism of K onto itself.
Exercise 76 Prove that the automorphisms of a field K form a group (with re-
spect to multiplication of mappings), which we denote by A(K).
Remark 38 Suppose G is a group of automorphisms of the field K (G is not
necessarily all of A (K)), and let Kc = ! x in K; a(x) = x for all a in G l. Kc is a
subfield of K, which we shall call the fixed field of G.
Proof: Let x andy be elements of Kc. Then for all a in G we have a(x) =
x and a(y) = y. Now a(x + y) = a(x) + a(y) = x + y and a(xy) = a(x)a(y) = xy.
Galois Theory
71
Thus x + y and xy are in K
0
. Since o:(-x) = -a(x) = -x, -x belongs to K
0
. Let
x=FO be an element of K
0
. Then o:(x-
1
)=o:(xr
1
=x-
1
, andx-
1
is in K
0
.
Therefore Kc is a subfield of K (Remark 34).
If K is an extension field of the field F, we shall denote by A (K,F) the set of
automorphisms of K that leave every element ofF fixed. That is, A(K,F) =
l a inA (K); a(x) = x for all x in F L
Theorem 66 A(K, F) is a subgroup of the group A(K).
Proof: Clearly, A(K,F) is a nonempty subset of A(K). (Why?) Let o:
1
and o:
2
belong to A(K,F) and suppose x is in F. Since (o:
1
o:
2
)(x) = o:
1
(o:
2
(x)) =
o: 1 (x) =x, o:
1
o: 2 belongs to A(K,F). Since o:(x) =x implies that o:-
1
(o:(x)) =
o:-
1
(x), or that x = a
1
(x), o:-
1
is in A(K,F). Thus, A(K,F) is a subgroup of
A(K) by Theorem 5.
Note:
The groups A(K,F) are of such importance that we emphasize them with
several examples.
Example 47 If R is the field of real numbers and C is the field of complex
numbers, then we have IA(C,R)l = 2. For suppose a belongs toA(C,R). Since
- 1 = a:( -1) = o:(i
2
) = [a (i)]
2
, it must be the case that o:(i) is i or - i. Now a leaves
every element of R fixed. Therefore the two mappings o:
1
and o:
2
defined by
o:
1
(a+ bi) =a+ bi and o:
2
(a + bi) =a- bi are the only automorphisms in A(C,R).
If a+ bi belongs to CA(C,R) the fixed field of A(C,R), then a(a + bi) =a+ bi
for all a in A(C,R). In particular, we have o:
2
(a + bi) =a+ bi, which implies that
a+ bi =a- bi or that b = 0. Therefore CA (C, R) = R.
Example 48 Let us find We may take as a basis for
Q(-V2,i.,ff) over Q the set f/2,i.J3 (See Ex-
ample 46 and Exercise 75.) The automorphisms of will be deter-
mined by the images of these basis elements; or particularly by the images of the
and iVf. Let a belong to Since
3
=
o:((-V'i)
3
) = o:(-2) = -2, must be a cube root of -2. Thus the possible
images are - ..f[, V2 (
1
+ V3) , and V2 (
1
-
Since [o:(i..{f)]
2
= o:((i..{f)
2
) = o:(-3) = -3, a(i..{f) must be a square root of
-3. Thus the possible images for ivG are ivG and -iyi It follows that there
are 6 automorphisms These are denoted by o:
1
, o:
2
, o:
3
, o:
4
, o:
5
,
and o:
6
, where
72
al =
az (- = -iff,
a
3
(-{/2) 1 :iYf),
=
as ( = e - '
(
1- iv'3)

2
'
al (iv'3) = iv'3
az (iy'3) = -iv'3
a3 {iv'3) = iv'3
Table 23 is the group table for A
TABLE 23
al az a3
04
as
<.%
al al az a3
04
as
<.%
az az al
<.%
as
04
a3
a3
a3 04
as
<.%
al az
04 04 a3.
az al
<.%
as
as
as <.%
al az a3
04
<.% <.% as 04
a3 az al
Galois Theory
Exercise 77 Show that A (Q(- Q) is isomorphic to the group of Table
13 (and hence is also isomorphic to S
3
by Exercise 61).
Exercise 78 Suppose K is an extension of Q, the field of rationals. Show that
any automorphism a of K must leave every element of Q fixed. [Hint: start
with the fact that a(l) = 1.]
Example 49 For the group A we must have a(-\o/5} =.:5 for any
automorphism a since there is only one cube root of 5 in Q(-\V5). Since every
element in is of the form a+ + where a, b, and c
are in Q, we have that a(a + bf/5 + = a(a) + a(b )a( f/5) + =
a+ b;y5 + Therefore there is only one automorphism in A
namely the identity map. Here, the fixed field of A(Q(f/S),Q) is Q(f/5).
Remark 39 One has perhaps observed from the previous examples that the
order of A (K, F) has been less than, or equal to, the degree of K over F. It is
Galois Theory 73
true that whenever K is a finite extension ofF, then L4 (K,F) ~ [K:F]. Equal-
ity holds when K is an extension of a particular type, called a normal extension.
Such an extension is now defined.
Definition 49 Suppose K is a finite extension of F. If there is an a: inA(K,.F)
such that a:(x) =l=x where x belongs to K, (but not to F), then K is said to be a
normal extension of F.
Note: K is a normal extension ofF if the fiXed field KA(K,F) of A (K,F) is
F.
The normal extensions K of a field F are characterized by the following
theorem.
Theorem 67 K is a normal extension ofF if, and only if, K is the splitting
field of some polynomial f(x) in F [x] .
At this time we shall define what is meant by the Galois group of a poly-
nomial f(x) over F and discuss the important theorem known as the funda-
mental theorem of Galois theory. This theorem will establish a 1-1 correspon-
dence between the subfields of the splitting field of f(x) that contain F and
subgroups of the Galois group of f(x).
Definition SO Let f(x) be an element of F[x] and suppose K is its splitting
field over F. The Galois group of f(x) is the group A(K,F).
Note: Henceforth, we shall use the notation G(K,F) to denote the Galois
group of f(x). We also note that G(K,F) may be regarded as a group of permuta-
tions of the roots of f(x), since roots of f(x) are mapped onto roots of[(x) by
automorphisms in G(K,F).
Theorem 68 Suppose f(x) belongs to F [x] and K is the splitting field of f(x ).
Let S denote the collection of subfields E of K such that F ~ E and letS' be the
collection of subgroups of G(K,F). The mappingg:S -+S' defined by g(E) =
G(K,E) is a 1-1 mapping fromS onto S'. Furthermore, we have
1. E = Ka(K,E)
2. If His a subgroup ofG(K,F), thenH= G(K,KH)
3. [K:E] = IG(K,E)i; [E:F] = IG(K,F)/G(K,E)I.
4. Eisa normal extension ofF if, and only if, G(K,E) is a normal subgroup
ofG(K,F).
5. If E is a normal extension ofF, then G(E,F) is isomorphic to the quo-
tient group G(K,F)/G(K,E).
Proof: K is the splitting field over E of f(x) since it is the splitting field
over F of f(x), and F E Thus K is a normal extension of E (Theorem 67) and
we have E is the fixed field of G(K,E). (See note following Definition 49.)
Therefore E = Ka(K,E)
74
Galois Theory
Now suppose g(EJ)=g(E
2
). That is, suppose G(K,EJ)=G(K,E
2
). By
what has just been proved, we have E1 = Kc (K,EI) = Kc (K,E
2
) =E
2
. This
says that g is a 1-1 mapping.
We do not prove (2) of the theorem, but note that this implies that g is a
mapping from S onto S'. For if His a subgroup ofG(K,F) andH = G(K,KH),
then KH is a subfield of K containingFwithg(KH) = G(K,KH).
Since K is a normal extension of Ewe have IG(K,E) I= [K :E] (Remark 39).
Also, IG(K,F)I = [K:F] = [K:E] [E:F] = IG(K,E)I [E:F], which implies that
the degree of E over F is the index of G(K,E) in G(K,F). This is (3).
Now E will be a normal extension ofF if, and only if, whenever a belongs to
G(K,F) and x is in E, then a(x) is also in E. (We do not prove this fact here.)
Thus, suppose a belongs to G(K,F), a' belongs to G(K,E), and x belongs toE.
Then Eisa normal extension ofF if, and only if, a' (a(x)) = a(x); or if, and only
if, a-
1
[a' (a(x))] = (a-
1
a' a)(x) = x. It follows that Eisa normal extension ofF
if, and only if, a-
1
a' a is in G(K,E) for all a in G(K,F) and for all a' in G(K,E).
Hence G(K,E) is a normal subgroup of G(K,F) by Remark 15.
It remains to show that (5) holds. If we can find a homomorphism <P from
G(K,F) onto G(E,F) such that the kernel of <Pis G(K,E), then we have it. (See
Theorem 22.) Let a belong to G(K,F) and let <P:G(K,F)--> G(E,F) be defined
by <P(a) = 11, where 11 is the mapping defined by 11(x) = a(x), x in E. We first
note that <P is well defined. Since E is a normal extension ofF, if a belongs to
G(K,F), then a(x) belongs toE whenever x is in E. Now if xis also in F then
clearly, a(x) =x (since a is in G(K,F)), and so 11 is an element ofG(E,F). It is
left to the reader to show <P is actually a homomorphism.
Now suppose a is in G(K,F) and <P(a) is the identity map of G(E,F). Then
it must be the case that a(x) = x for all x in E. Thus a belongs to G(K,E).
Clearly, if a belongs to G(K,E), then <P(a) is the identity map of G(E,F). There-
fore G(K,E) is the kernel of <P
Finally, the image under <P of G(K,F) in G(E,F) is isomorphic to the quo-
tient group G(K,F)/G(K,E). (Theorem 22 applies. How?) Now IG(K,F)/
G(K,E) I= [E :F] (why?)= IG(E,F) I. This implies that <P must be onto.
We now illustrate the fundamental theorem of Galois theory with the fol-
lowing example.
Example 50 The Galois group of x
3
+ 2 is the group A (Q(- :iff,i-J'f), Q) de-
fined by Table 23. (See Examples 46 and 48.) Now the subgroups of the Galois
group are: 1a d , H
1
= \a 1 ,a 2 }, H2 = 1a 1 ,a4 } , H 3 = \a 1 ,a
6
} , H 4 = l a 1 ,a3 ,
a
5
}, and !a
1
,a
2
,a
3
,a
4
,a
5
,a
6
}. These subgroups are in 1-1 correspondence
with the subfields of Q(-V'l,i-J'f) that contain Q as indicated by
Galois Theory
l e l <--->
H1 <---> Q(-ij2)
H2 <---> Q (V2 1-
H3 <---> Q ( V2 1 +
H 4 <---> Q (i Vf'J
75
Since H
4
is normal in l cx
1
, cx
2
, cx
3
, cx4, cx
5
, cx
6
l (why?) we have Q(iVf) is a
normal extension of Q. Thus Q(iVf) is the splitting field of some polynomial
f(x) in F[x] by Theorem 67. (Which polynomial?) Are any of the remaining
fields Q (-Y2(
1
+ ;Yf)), Q( -Y2(
1
-
2
iV3)), or normal extensions of Q?
Since Q(iV'f) is a normal extension of Q, then G(Q(iV'f),Q) is isomorphic
to the quotient group G(Q( --if[,iVf), Q)/G(Q(- <fi,iVJ), Q(iVf)). The
reader should find the two co sets of this quotient group, and the two automor-
phisms in G(Q(iV'f),Q) and establish the isomorphism.
5-5. SOLVABILITY BY RADICALS
We are about ready to show that if the general nth degree equation is solv-
able by radicals, then the symmetric group Sn is a solvable group. (The converse
of this is also true, but we shall omit the proof.) Our approach is founded on the
treatment given by I. N. Herstein in his Topics in Algebra (New York: Blaisdell,
1964).
First, we state what is meant by saying that the polynomial f(x) ofF [x] is
solvable by radicals over the field F. (The reader should recall Remark 35.)
Definition 51 Suppose F is a field and f(x) is a polynomial belonging to F [x] .
We say that f(x) is solvable by radicals over F if there exist fields F
1
, F
2
, .. ,Fr
such that
1. Fi = Fi_
1
(bi), where bi is a root of the polynomial xni- ai in Fi-t [x],
(here, F
0
=F) and
2. all roots of f(x) are in Fr.
76
Galois Theory
Remark 40 If bi is a root of xni- ai, then bi is an nth root of ai Also, if K is
the splitting field of f(x) over F and f(x) is solvable by radicals (over F), then
KCFr.
Example 51 Consider the polynomial f(x) = x
2
+ 2x - 4 in Q [x] , Q the field
of rationals. Now the roots of f(x) are -l VS. Letting Q
1
= Q
0
(b
1
), where
Qo = Q and b
1
= y'J, we see that f(x) is solvable by radicals over Q. (Fr = Q
1
and xni- ai = x
2
- 5.)
By the general polynomial of degree n over the field F we have meant the
polynomialf(x) =xn- a
1
xn-
1
+a
2
xn-
2
- + +(-Itan where the ai are inF
We note thatf(x) may be regarded as a particular polynomial over F(a
1
,a
2
, ,
an), the field of rational functions in a
1
, a
2
, . , an over F. (See last paragraph
of Sec. 5-l.) To say thatf(x) is solvable by radicals shall mean that it is solvable
by radicals over the field F(a
1
,a
2
, .. , an).
Remark 41 One can show that ifF is a field containing the nth roots of unity
for a particular choice of n, then the splitting field K of xn -a (a* 0 in F) is the
field F(b) where b is any root of xn - a. It is also true that the Galois group of
xn -a is an Abelian group. We give a proof of this.
Proof: Let o:
1
and o:
2
belong to the Galois group of xn - a. Then o:
1
and
o:
2
are automorphisms of K (where K=F(b)) that leave every element in F
fixed. (See Definition 50.) Now o:
1
(b) and o:
2
(b) are roots of xn- a (why?)
and so we have o:
1
(b)= wib and o:
2
(b) = c,)b for some i andj such that I <,. i <,.
n - 1, I <,. j <,. n - 1. ( w is an nth root of unity. Remember that F contains all
the nth roots of unity for our choice of n.) It follows that (o:
1
o:
2
)(b) = o:
1

(o:
2
(b)) = o:
1
(wj b) = wj o:
1
(b)= wj wi b = wj+ib, and that (o:
2
o:J)(b) = o:
2
(o:
1

(b))=o:
2
(wib)=wio:
2
(b)=wiwjb=wi+jb. Thuso:
1
o:
2
ando:
2
o:
1
have the same
effect on b. But since they have the same effect on any x in F, they have the
same effect on the elements of F(b) = K. Therefore, o:
1
o:
2
= 0:20:1 and the
Galois group is Abelian.
Theorem 69 Suppose F is a field containing the nth roots of unity for any
choice of n. Also suppose that f(x) in F [x] is solvable by radicals over F. Then
the Galois group of f(x) is a solvable group. (The reader should review the last
part of Sec. 2-3, beginning with Definition 31.)
Proof: Let K be the splitting field of f(x). Since f(x) is solvable by rad-
icals over F, there exist fields F 1 , F2 , .. , Fr such that Fi = Fi-1 (bi) where bi is
a root of xni- ai, a polynomial in Fi_
1
[x], and such that all roots of f(x) are in
F (Definition 51). Now we may assume that Fr is a normal extension of F.
(Try to prove this!) Thus Fr is a normal extension of each field Fi
(Theorem 67). Since Fi is normal over Fi_
1
(Remark 41), G(FnF;) is a normal
Galois Theory 77
subgroup of G(Fr,Fi_J). (By Theorem 68.) Therefore, the series I e l C G(Fn
Fr-d C G(Fr,Fr_
2
) C C G(Fr,Fd C G(Fr,F) is a normal series. Again by
Theorem 68, we have G(Fi,Fi-d is isomorphic to G(FnFi-1)/G(FnFi). But
G(Fi,Fi_
1
) is Abelian (Remark 41). Therefore each factor of the above normal
series is Abelian, which implies that G (Fn F) is a solvable group (Theorem 36) .
(Remember that we are trying to show that G(K, F) is solvable!)
Now K C Fr (Remark 40) and K is normal over F (why?). Thus, from
Theorem 68 (again!), G(FnK) is normal in G(FnF) and G(K,F) is isomorphic
to G(FnF)/G(Fr, K). Since G(Fr,F)/G(Fr, K) is a solvable group (Exercise 40),
it follows that G(K,F) is also solvable.
Note: Theorem 69 is true even when F does not contain roots of unity.
The end is finally in sight. If we recall what our goal is, and what is meant
by saying that the general nth degree equation f(x) is solvable by radicals
(comment following Example 51), all that remains to be shown is that the
Galois group of f(x) over F(a
1
,a2, ... , an) is Sn. This is the famous theorem
due to Abel.
Let abelongtoSn. Themappinga:F(x
1
,x
2
, ... ,Xn)-+F(x
1
,x
2
, ... ,Xn)
defined by a(r(x
1
,x
2
, ... , Xn)) = r(x
0
{
1
),Xcx(
2
), ... ,Xcx(n )) is an automorphism
of F(x
1
,x
2
, ... ,Xn). Now the fixed field F(x
1
,x
2
, ... ,Xn)sn of Sn is the set
of all elements r(x
1
,x
2
, ... , Xn) in F(x
1
,x
2
, ... , Xn) such that a(r(x
1
,x
2
, ... ,
Xn)) = r(x
1
,x
2
, ... , Xn) for all a in Sn (See Remark 38.) That is, all elements
r(x
1
,x
2
, ... ,Xn) in F(x
1
,x
2
, ... ,xn) such that r(xl,x2 , ... ,Xn)=r(xa(l),
Xa(
2
), .. ,Xa(n)) for all a in Sn. (In what follows we letS =F(x
1
,x
2
, ... ,
Xn)sn .)
Next let a1 =x1 +x
2
+ +xn, a2 =x
1
x
2
+x
1
x
3
+ +x
1
xn +x
2
x
3
+
X2X4 + +x2Xn + +xn-lXn, ... , an =x
1
x2 xn,andconsiderF(a
1
,a
2
,
... , an). F(a
1
,a
2
, ... , an) is a subfield of S. (Prove that the a; belong to S.)
Clearly, Sn is a subgroup of G(F(x
1
,x
2
, . .. , Xn),S). By Remark 39, I G(F(x
1
,
X2, ... , Xn),S)I.;;;; [F(x1 ,x2, ... ,Xn) :S]. Since ISn 1.;;;; IG(F(x
1
,X
2
, ... , Xn),
S)l, it follows that n! .;;;; [F(x
1
,x
2
, ... , Xn) :S].
Now note that the general polynomial f(x) =xn-a
1
xnl + a
2
xn-
2
- + +
(-ltan belongs toF(a
1
,a
2
, ... ,an)[x] and thatf(x)=(x-xJ)(x- x
2
)
(x- Xn) (Remember what the a; are.) It follows thatF(x
1
,x
2
, ... ,xn) is the
splitting field of f(x) over F(a
1
,a
2
, ... , an) where [F(x
1
,x
2
, ... , Xn): F(a
1
,a
2
,
... ,an)J .;;;;n!. (SeeTheorem65 andDefinition47.)
Finally, since F(a
1
,a
2
, ... ,

... , Xn), we have (from

what has already been shown above) thatn!.;;;; [F(x
1
,x
2
, ... ,xn):S] [S:F(a
1
,
a
2
, ... , an)] = [F(x
1
,x
2
, ... , Xn) :F(a1 ,a
2
, , an)] .;;;; n!; which implies that
[S:F(a
1
,a
2
, ... , an)]= 1. Therefore S=F(a
1
,a
2
, ... , an), and Sn =G(F(xb
x
2
, ... ,Xn), F(a
1
,a
2
, ... , an)j. (This also follows by what was shown above.)
78
Galois Theory
Final Remark In order for the general nth degree polynomial f(x) =
xn - a
1
xn-
1
+ a
2
xn-z - + + ( -1 tan to be solvable by radicals, the Galois
group of f(x) must be a solvable group (Theorem 69). We have just shown that
the Galois group of f(x) is Sn. Since Sn is not a solvable group for n:;;:. 5
(Remark 23), the general nth degree polynomial, where n :;;:. 5, is not solvable by
radicals.
Index
Abelian group, 3
Adjunction of an element to a field, 64
Alternating group, 27
Ascending central series, 45
Automorphism
of a field, 70
of a group, 21
Basis, 66
Binary operation,
Cayley's theorem, 13
Center of a group, 15
Central series, 44
Characteristic 0, 7 0
Class of a nilpotent group, 44
Commutator, 15
subgroup, 15
Complex, 14
Composition
factors, 29
series, 29
Conjugate
elements, 22
subgroups, 13
Coset, 7
Cycle, 25
Cyclic group, 8
Degree
of a polynomial, 65
of an extension, 67
Derived group, 15
Dimension, 66
Direct product, 33-34
Double coset, 38
79
Euler phi-function, 11
Even permutation, 26
Extension field, 64
Factor group, 19
Field, 63
of quotients, 65
of rational functions, 65
Finite
dimensional, 66
extension, 67
Fixed field, 7 0
Frattini subgroup, 47
Fundamental theorem of Galois theory, 73
Galois
field, 64
group, 73
General polynomial of degree n, 76
Group, definitions of, 1-3
Hamiltonian group, 56
Homomorphic image, 17
Homomorphism, 16-17
Index of a subgroup, 8
Inner automorphism, 22
Invariant subgroup, 14
Irreducible polynomial, 68
Isomorphic normal series, 29-30
Isomorphism
of a field, 70
of a group, 12
Jordan-Holder theorem, 31
80
Kernel, 17-18
Klein four-group, 49-50
Lagrange's theorem, 8
Linear combination, 66
Linearly dependent vectors, 66
Mapping, 1
Maximal normal subgroup, 27
Multiplicity, 68
Natural homomorphism, 20
Nilpotent group, 44-45
Normal
ex tension, 7 3
series, 29
subgroup, 13
Normalizer, 14
Odd permutation, 26
One-to-one mapping, 4-5
Onto mapping, 5
Order
of a group, 8
of an element, 9
Outer automorphism, 22
Permutation, 4
p-group, 40
Polynomial, 65
Quaternion group, 54-55
Quotient group, 19
Relatively prime, 11
Ring, 65
Root of a polynomial, 68
Simple group, 27
Solvable
by radicals, 75-76
group, 32
Splitting field, 69-70
Subfield, 64
Subgroup, 6
Sylow
p-subgroup, 42
theorems, 42-43
Symmetric group, 4
Transposition, 26
Vector space, 65
Index