Академический Документы
Профессиональный Документы
Культура Документы
INTRO-DUCTION TO
MOO'ERN ABSTRACT ALGEBRA
This book has been written with the intention of providing an introduction to
model'll aht:!t.ract algebra for mathematics major8. The r('.ader is not presumed
at thc Outllct to possess any previout:! knowledge of the concepts of modem
algebra. Accordingly, our beginuing is somewhat elementary, with the exposition
in the earlier sections proceeding at a leisurely pace; much of this early material
may be covered rapidly on a first reading. An attempt. has been made to keep
the book as self-contained as possible. To smooth the path for the unexperienced
reader, the fir8t chapter is devoted to a review of the basic facts concerning
sets, functions and number theory; it also serves as a suitable vehicle for intro-
ducing some of the notation and terminology used subsequently.
A cursory examination of the table of contents will reveal few surprises; the
topics chosen for discussion in COUr8es at this level are fairly standard. However,
our aim has been to give a presentation which is logically developed, precise,
and in keeping with the spirit of the times. Thus, set notation is employed
throughout, and the distinction is maintained between algebraic systems as
ordered pair8 or triples and their underlying sets of elements. Guided by the
principII! thnt 0. st(~l"ly diet. of definitions and cxullIples SOOIl 1)(~c()llIel'l unpalat-
able, our eiTOl-ts are directed towards establishing the most important and
fruitful results of the subject in a formal, rigorous fashion. The chapter on
groups, fOl' example, culminates in a proof of the classic Sylow Theorems, while
ring and ideal theory are developed to the point of obtaining the Stoile Repre-
sentation Theorem for Boolean rings. Ell route, it is hoped that the, reader will
gain an appreciation of precise mathematical thought and t.he current standards
of rigor.
At the eud of each section, there will be found a collection of problems of
varying degrees of difficulty; these constitute an integral part of the hook. They
introduce a variety of topics not treated in the main t.('xt, as well as impart
much additional detail ahout material covered earlier. Home, especially in the
latter seet.ions, pl'Ovide slIbl'ltantial extensionl'l of t.he'theory. We have, on the
whol(', resist.('d t.h(' t.('mptation to lise the exercises to develop results that will
be ncedl'd HllbsequC'nt.ly; aN a 1(,~lIlt., the reader need not work all the problems
in order t.o read the reHt of the hook. Problems whose solut.ions do not appear
particularly straiJ!;htforward arc accompanied by hints, Besides the general
index, a glossary of !!pecial Hymhol!! iN also included.
v
vi PREFACE
PRELIMINARY NOTIONS
This ehaptcr hrieAy summarizes IIOme of the bll.llic notions concerning sets,
functiollH, and number theory; it alllO serves Itl! It vehicle for establishing COIl-
ventions in notation and terminology used throughout the text. Inasmuch as
this material is intended to serve primarily for background purposes, the
reader who is already acquainted with the ideas in this chapter may prefer to
cmbark directly 011 the next.
Within the confines of one section, it is obviously impossible to give complete
coverage to set theory or, for that matter, to achieve a logically coherent exposi-
tion of such a formalil!tic diseipline. The subsequent presentation should thus
be regarded simply as a summary of the fundamental aspects of the subject,
and not as a sYRtematic development.
Rather than attempt to list the undefined terms of set theory and the various
axioms relnting them, we shall take an informal or naive approach to the sub-
jeet. To thil! cnd, the term set will be intuitively understood to mean II. collection
of objects having some common characteristic. The objects that make up a
given sct arc called its elements or members. Sets will generally be designated
by capital letters and their elements by small letters. In particular, we shall
employ the following notations: Z is the set of integers, Q the set of rational
numbers, and R' the set of real numbers. The symbols Z+, Q+, and R~ will
stand for the positive elements of these sets.
If x is an element of the set A, it is customary to use the notation x E A
and to read the symbol E as "belongs to." On the other hand, when x fails to
be an element of the set A, we shall denote this by writing x I;l A.
There arc two common methods of specifying a particular set. First, we may
list all of its elements within braces, as with the set {-I, 0,1, 2}, or merely
list some of its elements and use three dots to indicate the fact that certain
obvious clements have been omitted, as with the set {I, 2, 3,4, ... }. When
such a liHting il! not practical, we may indicate instead a characteristic property
whereby we can determine whether or not a given object is an element of the
set. IHore specifically, if P(x) is a statement concerning x, then the set of all
elementR x for whieh the J;tatcmcnt P(x) is true is denoted by {x I P(x)}. For
example, we might have {x I x is an odd integer greater than 2I}. Clearly,
1
vi PREFACE
Index 305
vii
2 PRELIMINARY NOTIONS I-I
{I, 2, 3} = {3,I,2,2},
since each set contains only the integers 1, 2 and 3. Indeed, the order in which
the elements are listed in a set is immaterial, and repetition conveys no addi-
tional information shout the ad.
An empty set or null set, represented by the symbol 0, is any Bet which has
no elements. For instance,
AnB= {xlxeAandxEB}.
The difference of A and B (sometimes called the relative complement of B in
A), denoted by A - B, is the subset of U defined by
A - B = {x I x E A but x ~ B}.
In the definition of union, the word "or" is used in the "and/or" sense. Thus
the statement "x E A or x E B" allows t.he possibility that x is in both A and B.
It might also he Jlot(ld par(mthetiml.\ly that, utilizing this new notion, we could
define A to be a proper subset of B provided A ~ B with B - A ~ 0.
The particular difference U - B is called the (absolute) complement of B
and designated simply by -B. If A and B are two nonempty sets whose inter-
section is empty, that is, A n B = 0, then they are said to be disjoint. We
shall illustrate these concepts with an example.
Examp.e 1-1. Let the universe be U = {O, 1,2,3,4,5, 6}, the set A = {I, 2, 4},
and the set B = {2, 3, 5}. Then
A uB = {1,2,a,4,5}, A n JJ = {2}, A - B = {t,4},
and
B - A = {3,5}.
Also,
-A = {O, 3, 5, 6}, -B = {O, 1,4, 6}.
We shall verify the first equality of (d), since its proof illustrates a tech-
nique mentioned previously. Suppose that x E Au (B n C). Then, either
x E A or x E B n C. Now, if x E A, then clearly both x E Au Band
x E A U C, so that x E (A u B) n (A u 0). On the other hand, if x E B n C,
then x E B and therefore x E A u B; also x E C and therefore x E A u C.
The two conditions together imply that
x E (A U B) n (A u C).
This establishes the inclusion,
A u (B n 0) ~ (A u B) n (A u 0).
Conversely, suppose x E (A U B) n (A u 0). Thcll both x E A U Band
x E AU C. Since x E A U B, either x E A or x E Bi at the same time, since
x E A U C, either x E A or x E C. Together, theMel conditions mean that
x E A or x E B n C; that is, x E A u (8 n C). This proves the opposite
inclusion,
(A U B) n (A u 0) ~ A u (B n C).
By part (e) of Theorem 1-1, the two inclusions are sufficient to establish the
equality,
A u (B n C) = (A u 8) n (A u C).
xE-(AUB).
The first two parts of the above theorem are commonly known as DeMorgan'8
rules.
There will be occasions when we wish to consider sets whose elements them-
selves are sets; in order to avoid the awkward repetition "set of sets," we shall
frequently refer to these as families of sets. One family which will prove to be
of considerable importance is the so-called power set of a given set.
Definition 1-4. If A iii I1.n nrhit.mry l'Ict, thcn t1w !let WhOHC clements are
nl\ the HuhRI't.H of A ill known nil the power 8(~t of A allli dmlOtcu by P(A):
peA) = {B I B ~ A}.
P(A) = {0, {a}, {b}, {c}, {a, b}, {a, c}, {b, c}, A}.
6 PRELIMINARY NOTIONS 1-1
It is both desirable and possible to extend our definitions of union and inter-
section from two sets to any number of sets. Assume to this end that a is a
nonempty family of subsets of the universe U. The union and intersection of
this arbitrary family are defined by,
a= {Ad i E I},
is then said to be indexed by the elements of I, and the set 1 is called an index
Bet for a. Wh(ln t.hifl notation is employed, it is customary to denote the union
and intcrHCction of the fnmily a by
and
Example 1-3. If A" = {x E R' / -lin ~ x ~ lin} for n E Z+, then
Note that according to this definition, a and b are not elements of (a, b) but
rather components. The actual elements of the set (a, b) are {a, b}, the un-
ordered pair involved, and {a}, that member of the unordered pair which has
been selected to be first. This agrees with our intuition that an ordered pair
should be an entity representing two elements in a given order.
For a ~ b, the sets {{a, b}, {a}} and {{b, a}, {b}} are unequal, having
different elements, so that (a, b) ~ (b, a). Hence, if a and b are distinct, there
are two distinct ordered pairs whose components are a and b, namely, the
pairs (a, b) and (b, a). Ordered pairs thus provide a way of handling two things
as one while losing track of neither.
In the next th(!Orem, a useful criterion for the equality of ordered pairs is
obtained; the proof ill subtle, but simple, relying mainly on Definitions 1-1
and 1-5.
Theorem 1-4. Two ordered pairs (a, b) and (e, d) are equal if and only if
a = e and b = (l.
Proof, If a = e and b = d, then it is clear from Definition 1-1 that
This in tum implies {{a, b}, {a}} = {{e, d}, {e}}, whence (a, b) = (e, d).
As for the converse, suppose that {{a, b}, {a}} = {{e, d}, {en. We dis-
tinguish two possible cases:
1) a = b. In this case, the ordered pair (a, b) reduces to a singleton, since
Some of the baKil: properties of equivalence classes are listed in the next
theorem.
Theorem 1-5. J.A:~t - he an equivalence relation in the set A. Then,
1) for each a E A, [a] 0,
2) if b E [a], then raj = [I)]; that is, any clement of the equivalence class
[aJ determines the cllUlS,
3) for any a, b E A, with faJ rbI, [a] n [b] = 0,
4) U{[a]laE A} = A.
Proof. Clearly, a E [a], since a-a. To prove (2), let bE [a], RO that b - a.
Now, SIIPPOKI' ;r E [a). whil'" impIiI'R.1" - a. UKinK t.he KYlllmctril~ and transitive
propcrt.il's of -, it follmvH that. :c - b, h('nl'c x E fbI. This estnhlislwH the in-
clusion [a] s;; [b]. A Himilllr argument. yil'lds the opposite in('lusion and thus the
equality [a] = fbI.
We derive (3) hy assuming, to the contrary, that there is some element
C E [a] n fbI. Then by statement (2), which has just been verified, [a] =
[e] = rbI, an obvious contradietion. Finally, since each clll.SS [a] s;; A, the
inclusion U{[a] I a E A} s;; A ill apparent. To obtain the reverse inclusion one
need only demonstrate that each clement a in A belongs to some equivalence
c1aSllj but this is evident: if a E A, then a E [a].
We next conned. the idea of an equivalence relation in A with the concept
of a partition of A.
Definition 1-8. A partition of a set A iH a family {Ai} of nonempty subsets
of A with the properties
1) if Ai Aj, then Ai n Aj = 0 (pairwise disjoint),
2) UAi = A.
Briefly, a partition of A is a family {Ai} of nonempty subsets of A such that
every element of A belongs to one and only one member of {Ai}. The integers,
for instance, have a partition ('onsisting of the sets of odd and even integers:
Z = Z. u Zo, z.
n Zo = 0. Another partition of Z might be the sets Z+
(positive integ<'rH), Z_ (nl'Kntive integers), and {O}.
Theorem 1-5 may be viewed as asserting that if - is an equivalence relation
in A, then the fnmily of nil l'(Juivalent'e classes (with respect to the relation -)
forms a partition of A. We now reverse the situation and show that a given
partition of A indu('l's a natural equivalence relation in A.
Theorem 1-6. If {Ai} is a partition of the set A, then there is an equivalence
relation in A whose cquivalen('e classes are precisely the sets A j.
Proof. For clements a, b E A, we take a - b if and only if a and b belong to
the same subset A j . Th(' reader may check that the relation -, so defined, is
1-1 THE ALGEBRA OF SETS 11
This set may be represented geometrically as a straight line with slope 1 passing
through the point (a, b). Therefore, the relation"", partitions A into a family
of parallellinel!.
PROBLEMS
7. The notion of set indw;ion may be expressed either in terms of union or inter-
section. To see this, prove that
a) AS:;; B if and only if .1 U B = B,
b) ..l s:;; R if and only if .1 n R = :1.
8. a) If .1 s:;; /I nllli .1 ~ -II, prove I.hat :I .., tl.
b) If A !; Hand -.1 !;; H, prove that B - ll.
9. Establish the two absorption laws:
A U (A () B) = A, A n (A U B) = A.
AUB=,IUC and
Using the functional value not.ation, onll would then wriw f(x) = x 2 + 2 for
each x E R'. Needles!! to say, there are funetions whose ordered pairs would
be difficult-if IIOt impoliliible-to express in terms of a formula. The diseerning
reader is advised to keep in mind the distinction between a function and its
values or, as the case may be, its formula; although the notation sometimes
leads to confusion, these concepts are obviously quite different.
Definition 1-10. COII;;ider a funetion J ~ X X Y. If /), = X, then we
say t.hat J iii a fUflct.inn fmlll X int() }', or t.hat f mall'~ X into Y; thiH Kitua.tion
is expressed symboIieally by writing f: X ----> 1".
14 PRELIMINARY NOTIONS 1-2
+
(f gHx) = I(x) + ,,(x), )
(f - g)(x) = f(x) - g(x), where xeD, n Dg
(I g)(x) = f(x)(j(x),
(f/g)(x) = f(x)/g(x) , where x e (D, n Dg) - {x e Dill g(x) = OJ.
We term f + (I, f - g, f g and fig, the pointwise sum, difference, product and
quotient of f and g.
Examp'e 1-11. SUppOtlC
v4-x 2 x_~
(f/g)(.r) = 2/x = 2 v4 - ;1:2.
This last notation serves to explain the order of symbols in f 0; the letter
g is written directly beside x, since the functional value O(x) is obtained first.
It is apparent from the definition that, so long as RII n D, 'F- 0,/. g is meaning-
ful. Also, Dlo ll ~ DII and Rlo ll ~ RI .
Theorem 1-7. If /, g and h are functions for which the following operations
are defined, then .
1) (f. g) Ii - 1 (y It),
0
= [(f h) . (g h)](x),
0 0
To state this result a little more concisely, let us introduce some special
terminology .
Definition 1-13. Given a nonempty set X, the function ix: X -+ X defined
by ix(x) = x for each x E X is called the identity [unction on Xj that is to
say, ix merely maps caeh element of X onto itself.
ExpreRRed in terms of the identity funetion, what was just seen is that for
any function j: X -+ Y which is both one-to-one and onto Y,
{(3x-2,x)lxER'} = {(x,!(x+2lxER'},
we choose to write
r 1 = {(x,!(x + 2) I x E R'}.
On the ot.her hand, if B !;; Y, then the inverse image of B, denoted by f-l(B},
is the subset of X defined by
about notlltion nmy fcelllOnll'what UIICIJ.'W about this double usc of the symbol
f- 1 The abulIC of notation Ilhould not cauRe any confusion, howev<lr, for in
any given eontext it should be perfectly clear how f- I ill to be interpreted.
Certain properties of the function f may be conveniently characterized in
terms of inverR<l illlllg(~R. To he more (~xJllidl., J iR 1\ onc-t.o-QII(! fun<~tion if and
only if the invcrHC imagc! of each clement of III ill a Ringleton. Whereas j maps
onto Y if and only if the inverse image of eaeh nonempty subset of Y is non-
empty.
We shall now preRC'nt two short theorems which will establish the re\.ationship
between direct and inverse images; these results are rather trivial but will be
essential for later study. In the two theorems, supposc f to be an arbitrary
function from the sct X into the sct Y.
Theorem 1-8. For each subset B !;; Y,
J(rl(B) !;; B.
ProoJ. If b eJ(j-I(R, then b = f(a) f~r some element a in f-t(B). From
this, it follows that f(a) e B, and conscquclntly b e B.
Corollory. If, in addition, j maps onto the sct Y, then
Proof. In view of the indusion proved in the theorem, we need only establish
that, under the existing hypothesis, B !;;f(j-I(B). For this, let be B; then,
as j is by supposition an onto function, b = f(a) for some choice of a in X.
Since a ej-l(B),
80 that
1-2 FUNCTIONS AND ELEMENTARY NUMBEU THEORY 19
Proof. H Jllninly HllfIlI'I'H 1.0 I'HI.nhliflh t.hn 0111' il",hlllicm , - I <t(A) !;;; A. To
II Ef I (j(A). We then hlWt! f(a) ef(A) ILUU thUl"f(a) = f(a')
HLILI'L with, 11'1.
for some a' in A. Since' is assumed to be one-to-one, a = a' and 80 a e A.
Before terminating this section, it may be well to review, quickly, some of
the facts from number theory which we shall require later. Most of these results
depend on the HC)-(~nlll!(l Wcll-Ordnring Prirwiplc:
Well.Orderlng Principle. Every nonempty subset B of nonnegative integers
contains a smallest element; that is, there exist..~ some (unique) element
a e S with a ::; b for all b E S.
Let us start with the following result.
Theorem 1-10. If a, be Z, with b > 0, then there exist unique integers
q and r such that
a= qb+ r, 0::; r < b.
r = a - qb, r ~ O.
We now show that r < b. In the contrary case, r ~ band
o S Iq - q'l < 1.
Since Iq - q'l is a nonnegative integer, it follows that q = q', which in tum
gives T = r'.
Corollary. (Division Algorithm). If 0, b E Z, with b 0, then there exist
unique integers q and r such that
4) if a I band b I c, then a I c,
5) a I band b I a if and only if a = b,
6) if c I a and c I b, then c I (ax + by) for every x, y E Z.
Definition 1-16. Let a and b be integers, not both of which are zero. The
{JreairHtr()mmon tiiviR01' of a I\lId b, dell()u,d by Ked (a, b), is t.he positive in-
teger tl Ilueh t.hnt
1) d I a and d I b,
2) if c I a and e I b, then e I d.
Briefly, ged (a, b) is the largest integl1r in the set of nil common divisors of
a alld b.
A naturu.l qU(!lltion to ask is whether the integers a Ilnd b can posseS!! two
different greatest eommon divisors. For an an!!wer, suppose there are two
positive integers d and d' whieh satisfy the conditions of Definition 1-16. Then
by (2), we must have did' as well as d' I d, whence d = d' [Theorem 1-11(5)].
Since tl and d' are hoth positive integers, it follows that d = d'. Thus, the
greatest COlIllllOII diviHor of a and b is unique, when it exists. The following
theorem will prove that any two integenl, which are not both zero, actually
do have a greatest common divisor.
Theorem 1-12. If a, b are integers, not both of which are zero, then
ged (a, b) exists; in fnet, t.here ('xiRt integerR x II.l\d y such that
From the Division Algorithm, one can obtain integers q and r such that
a = qd
+ r, ~ r < d. But then r will be of the form
r = a - qd =a- q(ax + by)
= a(1 - qx) + b( -qy).
Were r> 0, this representation would imply rES, and contradict the fact
that d is the least integer in S. Thus r = 0, so that d I a. A similar argument
establishes that d I b, making d a common divisor of a and b. On the other
hand, if e I a and e I d, then by Theorem 1-11(6), e I (ax + by), or rather,
c I d. From these two statements, we conclude that d is the greateflt common
divisor of a and b.
It may be well to record the fact that the integers x a.nd y in the representa-
tion gcd (a, b) = ax + by are by no means unique. More concretely, if a = 90
and b = 252, then
ged (90, 252) = 18 = (3)90 + (-1)252.
Among other possibilities, we also have
18 = (3 + 252)90 + (-1 - 90)252 = (255)90 + (-91)252.
There is a special case of Theorem 1-12 which will play an important role
in the future; while it is, in effect, a corollary of the foregoing result, we shall
single it out as a theorem. But first, a definition: two integers a and b, not
both of which are zero, are said to be relatively prime (or prime to each other)
if and only if gcd (a, b) = 1. For instance, the integers 8 and 15 are relatively
prime, although neither is itself a prime.
The.Nm 1-13. Let a, ,) E Z, lIot both zero. Th(m a Ilnd b are relatively
prime if and only if there exiflt integers x and 11 such that
1 = ax +by.)
Proof. If a and b are relatively prime, 80 that gcd (a, b) = 1, Theorem 1-12
guarantees the existence of x and y satisfying 1 = ax + by. Conversely,
suppose 1 = ax + by for suitable x, y E Z and that d = gcd (a, b). Since
d I a, d I b, Theorem 1-11(6) implies d I (ax + by), or rather d 11. Because d is
positive, this forces d = 1 [Theorem 1-11(2), as desired.
In light of Theorem 1-13, one may easily prove
Theorem 1-14. (Euclid's Lemma). If a I be, with a and b relatively prime,
then a I e. .
Proof. Since gcd (a, b) = I, there exist integers x and y for which 1 = ax + by.
MUltiplying bye, wc obtain
e = (ax + by)e = a(cx) + (be)y.
1-2 FUNL'TIONS ANI) ELI!:MENTARY NUMB.m TIIEOItY 23
where the Pi and qi are primes. The argument proceeds by induction on the
integer n. In the case n = 1, we have a = PI = ql(q2' . q...). Since PI is
prime, it possesses no proper factorization, so that m = 1 and PI = ql.
Next, assume n > 1 and that whenever a can be expressed as a product of less
than n factors, this representation is unique, except for the order of the factors.
From the equality PIP2 ... Pn = qlq2' . q... , it follows that PI I (qlq2' q...).
Thus, by the preceding corollary, there is some prime qk, 1 :::; k :::; m, for which
PI I qki relabeling, if Iwccssary, we may suppose, PI I ql. But then PI = qt,
24 PRELIMINARY NOTIONS 1-2
for ql has no divisors other than 1 and itself. Canceling this common factor,
we conclude 1'2 .. PA = q2 .. qm' According to the induction hypothesis, a
product of n - 1 primes can be factored in essentially onc way. Therefore,
the primes Q2, . , qm are simply a rearrangement of the primes 1'2,'" , PA'
The two prime factorization!'! of a are thus identical, (~ompleting the induction
step.
An immediate consequence of this theorem is the following:
Proof. Assume the Rtatement is false; that is, assume there are only a finite
number of primes PIr 112, ... ,1'". Consider the positive integer
a = (1'11'2' .. PA) + 1.
None of the primeR Pi divides a. If a were divisible by PIr for instance, we would
then have p, 1(0 - P,1'a'" 1'A) by Theorem 1-11(6), or PI 11; this is im-
possible by part (2) of the same theorem. But, since a > 1, the Fundamental
Theorem asserts it must have a prime factor. Accordingly, a is divisible by a
prime which ifol not amonK our list of all primes. ThiR lugument shows there
is no finite listing of till! prillle int<~gerH.
This comJlI(!t<~H our Hurvc~y of HOmel of till! fUluJu.numtlLl notion!!! c:onceming
sets, functions, and arithml'tie in thll int,egers. Although the treatment was
purposely sketchy, it is hoped that the reader did not find it too superficial. In
the subsequent chapters of the text, we shall utilize the foregoing concepts by
applying these ideas to certain specific! situations.
PROBLEMS
a) I = {(x, x 2 +
x) I x E R'}, g = {(x, ax +
4) I x E R'},
b)/= {(x,x/(x 2 +
1)) IxER'}, g= {(x,l/x)lxER'-{O}}.
5. Let I, g and h be runction!; such that Rio ~ D. and R. ~ D,. Verify the law
1 0 (g h) = (I. g) 0 h.
1-2 FUNCTlO!IIS AND ELEMENTARY NUMnER THEORY 25
6. Let I, g and h hE' fundionl! from R' into R' with Rio ~ D, n /)/1' PrO'!e that
7. Determine which of the function14 below are one-to-one. In thollC cases in which
the fund inn I ill not (llll'-tn-Onl', "xhihit. two pl1irH (:tl, YI), (X2, Y2) EI such that
Xl ;o! X2 but YI = Y2.
a) I = {(x,:t 2 + 1) I :t E R'},
b)/= {{x,lx-Jl)I-2:5:t:52},
c) I I
= {(x, 11x) x E R~}.
8. For every pair of real numbers a and b, define a function I",,: R' -+ R' by the
formula/ob{x) = a:t +
b for each x E R'.
a) Show t.hat. /I. /'.0 = lab.
0
b) For a ;o! 0, prove that I"" ill both one-to-one and onto.
c) For a ;o! 0, obtain 1.. 1
9. Using fllnl'tions/: R' -- R', give an example of a function which ill
a) one-to-one but not onto,
b) onto but not one-to-one.
10. For functionR g: X -+ }' and I: Y -+ Z, flhow that the following statements are
true:
a) If I iH all nnto funl!tioll, then I ill 8\1111.
0 (/
GROUP THEORY
In this I!hapter, nlld t.hroughout the renllLind(~r of thc text" we shall deal with
mathemntielll HYHtemK whidl ar(~ d(~fined hy a preserihcd list of properties.
Emphasis will be on deriving theoremll that follow logically from t.he postulates
and which, at the same time, help to describe the algebraic structure of the
particular system under consideration. This axiomatic approal!h not only
penn its us to concentrate on essential ideas, but also unifies the prescntation
by showing the basic similarities of many diverse and apparent.1y unrelated
examples.
We first confine our attention to systems involving just one operation, since
they are amenable to the simplest fornlal description. Despite this simplidty,
the axioms permit the construction of a profuse and elegant theory in which
one encounters many of the fundamental notions common to all algebraic
systems.
Before beginning, however, it iH necessary to arrive at some understanding
conceming the UKe of the equivalClwc relation =. We will henceforth take
the equality Hign t.o mean, intuitively, "is the same as." In other words, the
Hymhol = aHK(~rt.K thnt the two pllrtieular expreHllions involved arc merely
differcnt nameH for, or descriptions of, one and the HaOle objcct; jUHt onc objed
is being conllidered, and it ill named twice. To indicate that a and b arc not
the same object we shall, naturally enough, write a F- b.
As a fimt steJl in our program, we introduce the concept of a binary operation.
This idea is the cornerstone of all that follows.
Definition 2-1. Given a nonempty set S, any function from the Cartesian
produet S X S int.o S is "ailed 1\ binary operation on S.
A biliary op('rntion Oil S thuH 1l..'i."IigIlK to ('ad I ordl'J'('(\ pair of clements of S
a uniquely determined third element of the same sct S. For instllnee, if P(A)
denotes the power set of a fixed set A, then hoth U Ilnd n are binllry operations
on P(A). In praetice, we shall generally use the symbol * to represent a
binllry openlt.ioll lind write a * II, inst.ead of *(a, IJ)), for its vlLlue nt the
ordered pair (a, b) E S X S. While this eonvcntion is at varimlee with the
functional notation developed in the previous ehapter, its use in the prescnt
27
28 W({JUl' TH~;OHY 2-J
2 3
1 1 2 3
2 3 1 2
3 2 3 1
2-1 In:FINITION AND ~:XAMJ>I..:I'\ OF <moups 29
According to the t.able, the product 2 3, for inst.ance, iR equal to the clement 2,
located at the interRection of the row marked 2 and t.he column marked 3.
Given an arbitrary binary operation ., there iH ('ertninly no ren.son to expect
that a b will be the Rllllle IlS 11 a for nil a Ilnd b. In faet, it can be seen in the
above example that 1 2 = 2, whereas 2. 1 = a. One must consequently
take care to refer to a b as the product. of a and b and to b a as the product
of b and a; the distinction is quite importallt. We should also point out that it is
obviously possible to combine an element with itself. Thnt is to say, a a
can be defined.
Example 2-3. The pair (8, .), where the set S = {I, -1, i, -i} and the
operation is that of mdinary mult.iplication, is a mltth('nmtieall'!ystem provided
one defines i 2 = -1.
Example 2-4. If Ze and Zo denote the even and odd int<'gers, respectively,
then (Z., +, .) ('onstit.utes a mathematical system, while (Z., .+, .) cloes not..
In the latter caRe, Ow set Z. ill not. ('Ios(>d undN addition, sin('l' t.he sum of two
odd integerl'! is lweessarily even.
The systemH to be studied Hlloscqtwntly arc dall8ificd aecording to the proper-
ties they possess or, to put it anot.her way, accordillg to the axioms they satisfy.
Our object will be to present a sequent.ial development of the principal mathe-
matical Hystems of modern algebra, beginning with t.hose involving relatively
few axioms and progressing to systellls sat.isfying more detailed hypotheses.
The axioms whieh form the starting point of the abstraet theory can be, by
nature, ratlwr varied. The growing t.endeney of modern mathematies is to
isolate almost. any convenient set of propl'rtie~ from its original context, to
define a parti(,ular system, llnd to develop the eorrcspollding Ilbstract theory
through logical d(~dul'tion. ROllle of tlH'se forlllal :txiolllat ie theorie~, sueh as
the notion of It group, have It fundamentlll irhportalH'(' to the whole of mathe-
matics and have been instrullll'ntal in IInifyillg various apPllrently unrelated
branches; other I heories, while sat.isfying the esHwt i(' and inquisitive needs of
the mathemati('ian, are limited in the extent of their applicability. We do not
mean to create the impression that it, is the usmil praetiec for one to define a
new system by arbitrarily (apart from logical (,oll~iderntions) writing down
axioms. Although thpr(' is no part.icular np('('ssity for the model to precede the
thcorct~al development, in most cases the axioms nrc the abstract realization
30 GlWUP THEORY 2-1
tivity of the operation, for one must compute all possible threefold products.
On the other hand, it may be fur easier to show thltt the operation is not
a.'lROciativc, as all we need do in thi::; ea.'lC is find three pnrticular clements from
the underlying Ret fOf whieh the nS!lO(~iative law fails.
Example 2-7. Conl'lid(!f the I)JI(!MLI.iuli t11!fiw,t1 on the set S = {l, 2,' 3} by the
operntion tlLllle:
1 2 3
1 1 2 3
2 3 1 2
a 2 :J
The associative law thus fails to hold in the system (8, .).
The mathematical system which we shall use to build up more complieated
nlgehmil! llt.rudufCS ill known 11."1 a semigroup.
Definition 2-4. A Bemi(JTOUl1 ill a paif (8, .) (!onllillting of a nOnCml)ty scI. 8
together with an associative (binary) operation. defined on S.
Let us stress t.hat it is an abuse of lnnguage to say a certain sct alone is a
semigroup without also specifying the operation involved, as it may be quite
possible to equip the same set with several associative operations. For this
reason, wc have utilized the ordered pair notation to indicate both the operation
and the underlying set of elements.
Observe that since any three elements from the set of a semigroup always
associate, there is no particular reason for parentheses. Consequently, when
dealing with sueh It system, the I!ymbol a ~ c has meaning in the sense that
we are frL'C to interpret it either as a (b cj or as (a * b) c. More generally,
the notlttion al a2 .. am is unambiguous, for it can be shown that all
ways of inserting parentheses 80 as to give this expression a value yield the
same felmlt, (Theorem 2-4). An opemtion whieh il! not as80ciative has the
dedded disadvantage thtLt the notation for multiple-factored products can
be(!ome quite unwieldy 11." a result of the eonstltnt need for parentheses.
In order to ROlidify the notion of a semigroup, we present several examples.
Example 2-8. There Itrl! severnl semigroups wit.h whi('h t,he reader is already
familiar. If, for insta.nce, Z+ denotes t.he set of ull positive integers, then both
the pairs (Z +, +) and (Z +, .) form semigroups. Similar statements hold for
the sets Z, Q, and R'.
(mOUI' 'l'llI':Olty 2-1
Example 2-9. Define the operation * on the real numbers by the rule
a*e=e*a=a
2-1 UJo.WINITION ANU EXAMPLES OF GROUPS 33
S = {a + bv'2 I a, b E Z},
a * a' = a' *a = e.
An element a' having thiH property iH called an inverse of a and is customarily
denoted hy a-I.
An inverse hall the effect of reducing a given clement, under the operation,
to the identity element. In particular, since e * e = e, we may infer that
e- I = e.
It will be established shortly that, for a semigroup with identity, each element
has at most one inverse relative to the unique identity clement. (The reader
might try to work out the proof for himself.) Thus, when dealing with such a
system, there is no ambiguity of meaning in the symbol a-I and, if it exists,
we are free to speak of "the invefHC of an element. "
Example 2-15. Let S be the set of all ordered pairs of nonzero-real numbers
and * the binary operation defined by
(a, b) * (e, d) = (ae, bd).
Then the system (S, *) forms a (commutative) semigroup with identity, with
the pair (1,1) serving as its identity element. For (a, b) E S, we evidently
have (a, b)-I = (l/a, lib), since
and the operation defined by Table 2-1. In this group, the identity element
is the pair (0,0), and every clement is its own inverse. Here the verification
of the associative law becometi a process of detailed enumeration of all possible
cases that could arise.
Table 2-1
---_.
(0,0) (0, I) (1,0)
.
(I, I)
- - .--
(0, II) (II, II) (0, I) (1,0) (I, I)
(0, I) (0, I) (0,0) (I, I) U,O)
(1,0) (I, 0) (I, I) (0,0) (0, I)
(I, I) (I, 1) (1,0) (0, 1) (0,0)
Since the entire table is symmetric about the main diagonal (upper left to
lower right), the group operation. is commutative. Note that each element
of G appears once and only once in each row and column of the table. Indeed,
any multiplication table for a group has this feature.
Example 2-20. Let P(X) be the power set of somc fixed nonempty set X. As
we have SI.'Cn, the systcms (P(X), u) and (P(X), n) possess identity elements
o and X, respectively, but neither system has invel'SCs for any clements other
than their respective identities. Consequently, P(X) does not constitute a
group with regard to the formulation of either unions or intersections. It is
possible, however, to remedy this deficiency by defining another operation on
P(X) in terms of union and intersection which will fCsult in a group structure.
More specifically, conHi<i<'r the op<'ratioll ~ (thl' HYlll-
hoi is traditional) given hy the formula
A ~ B = (A - B) U (B - A),
for
A, B E P(X).
This operation, known as the symmetric difference of A and B, yields the set
which is represcnted by the shaded area in Fig. 2-1. We shall leave as an
exercise the v(>rificatioll t hat the symnlC'tric differenc(> operation is commutative
and associative.
It is easy to H(,(, that fot any Hl't A s;;; X (that iH, fot' any clement of P(X,
A ~ 0= (A - 0) u (0 - A) = A U 0 = A,
which proves that the empty 8('t 0 serves ns un identity element for~. Moreover,
A ~ A = (A - A) U (A _.- A) = 0 u 0 = 0.
2-1 1>";)o'lNITION ANIl l~XAMPL"~'i OF GUOUPS 37
This implicH t.hllt, e:wh dl'lIWllt, of J>(X) ill it.!l own invc1'IIC. Conscquently, the
mllthclllatil'nl HYHtl'l1l (f'(X), a) is n I:OJlllllut,nt.ive gnlUp.
Examp'e 2-21. As 1\ silllple exampl(! of a noneollllllutative group, let the set
G consist of all ordered pail'R of real numbers with nonzero first component:
G= {(a,b)la,bER',a~O}.
Define the operation on G by the formula
J. Determine which of the following binary operations on the set Q are associative
and which are commutative.
a) a * b = 0 b) a * b = l(a b) +
c) a * b = b d) a * b = a b- 1 +
2. Rllppose Lhe system (S, *) haR an identity element; show that if the equation
(a * h) * (c * d) = (a * e) * (h * d)
holds for all pOH"ible choices of element!i a, h, c and d of S, then the operation *
i!! both aH:-Iociative and commutative.
3. Prove that the set of ordered pairs of real numbers together with the operation *
defined on 8 by
(a, h) * (e, d) = (a +
e, b d + +
2bd)
* a h e
a abe
h h a e
c e c e
Verify that the pair (8, *) is a commutative semigroup with identity, but not a
group.
6. In the following illlltanceR, determine whether the systemR (G, *) described are
(~ommul.ative groups. For tholie systems failing to be so, indicate which axioms
are not satisfied.
a) (} - Z+, (I * II - max {(I,ll:,
h) a = Z, tl * b = mill {fl, bl (the Hlllall(!r of a and b),
e) (} = n' , a * b = a +
h - ab,
d) (J = Z+, a * b = Illax {a, b} - min {a, b},
e) (J = Z X Z, (a, b) * (e, d) = (a +
e, b d), +
f) G = R' X R', (a, b) * (e, d) = (ae +
bd, ad +. bd),
g) (J = R' X R' - (0,0), (a, b) * (e, d) = (ae - bd, ad be). +
7. RUJlPol'l(l that a E R' - {O, I} Ilnd conRider the set G of integral powers of
a: G -= {a k IkE Z}. If denotes ordinary multiplication, prove that (G,') is a
group.
40 GROUP THEORY 2-1
8. u-t G = {I, (-1 + iV3)/2, (-1 - iV3)/2}, where i 2 = -1. Show that the
lIytltem (a, .) is a group.
9. Con!!idl'r the set a consisting of the fOllr functions/J,i2,Ja,/4:
1
/I(x) = x, ! 2(X) = -x' hex) = -x,
x
with x E R' - {O}. Prove that (G,o) is a commutative group, where 0 denotes
functional composition.
10. Prove that the symmetric difference opt'ration
:1 aB = ( ..1 - B) U (B - A)
12. If G = {a E R'I -1 < a < I}, verify that (G, >II) form!! a commutative group,
with the operation'" givl'n by
a+b
a>llb = - - for a, bEG.
I fib +
13. Let (S, >II) Ill' a sl'migroup wit.hout illl'nt.ity and e hl\ any ('It~ment not in 8. Define
the olll'ration on the t't"" = S U {e) by ml'SnH of I.he ruleH
0
for all a, b E 8,
aoe=a=eoa for all a E 8'.
Show that the pair (8',0) is a semigroup with the identity element e; (8',0) is
said to be obtained by adjoining an identity.
14. Prove that the following weakened Het of axioms are actually equivalent to the
classical axioms for a group as given in Definition 2-11: A group is a mathe-
matical system (G, >II) for which
1) >II is an aN80ciative oper&tion,
2) there exists an element e in G such that a >II e = a for all a E a (existence of a
right identity),
3) for ('.sch a E G, thero ('xists an ('lement a-I in G such that a'" a-I = e (exist-
ence of right inverse.'1).
2-2 CERTAIN ELEMENTAltY TUEOItEMS ON GROUPS 41
AI! wc rmllllrke<i I'lLrlil'r, ollr apprO/wh t.o t.he suhjed, mnt.ter in this text is
dmrnl'l.l'rized hy tlH' ILhsll'lld. Ilxifllnalil: dpvdopnll'llt of llIodpl'n mu.themo.ties.
Thus PlLeh sYHI,(,1lI 10 hI' illvl'stiglLt('d (~onsists of It sd. of elcllIcntH and onc or
more opernt ions whi('h arc \II\(jelilll'd except that they ILre lUi8umed to obey
certain rules known as axioms or postulates. Once the axioms are granted, all
else is obtained by careful logical argument. Each new theorem is deduced
from the definitions, the axioms, and the t111'orenu! prcvioutlly proved. The
great advantage of the axiomatic method is that any particular example we
encounter which satislil's the axioms of a givpn mathl'nmti(!al system will also
satisfy nil Ow theorems which are true for that system. While the uninitiated
render llIay filld this approach uncomfortably abstraet at first, the feeling will
disappear as he progrl'ssl'H furt.her into the tcxt. III till! prO(!ess, wc hope that
he will gain an apprc('iatioll of llIathematics as !til exaeting, logical discipline.
In proving t.he following theorems on groups, it is essential to understand
that we know nothing whatever about the actual nature of either the set G
or the operation * ; both are completely abstract and unspecified. Our knowledge
of (J and il.s asso(,inted olwrat,ion is strietly eonfined to the information eon-
tairH'<i ill t Iw definitioll of grollP givl'lr (urli('r. Tlw muder would be welllLdvit!cd
at Uris poillt t,o r('view Definition ~-1I, Seet.ion ~-l, Iwfof(! proceeding. We
shall begin with a simple, yet highly important result.
Theorem 2-2. The identity clement of a group (G, *) is unique, and each
element of a group has precisely one inverse element.
Proof. Theorem 2-1 may be used to establish that the identity is unique. To
show Omf. an element has exaetly one inverse in G, we proceed as in Theorem
2-1 by showing that two sUPJlosedly distinct inverses are actually equal. To
this end, lu!sumc thc clement a E a has two invtrscs, a; and a~. Then according
to the definition of inverse,
a * a~ = a~ *a = P, a * a~ = a~ *a = e.
But, the identity element is the same in both eases (there is only one identity
as we have Illr('lldy 8('l'n), so that
a * a~ = a * a~.
!\I uitiply both I!ides of this equation on the left by a; (or by a~) to get
a~ * (a * aD = a~ * (a * a~).
Using the associative law, we have (a; * a) * u; = (a~ * a) * a~, and so
(' * a~ = e * a~ or a~ = a~,
An inspection of the proof Hhows that we have established a little more than
is indicated by the statement of the theorem. We have, in fact, shown that if
an element of a semigroup with identity has an inverse, then it must be unique.
Corollary. Each clement of a semigroup with identity has at most one
inverse.
A further useful conclusion to be drawn from Theorem 2-2 is that
(a-I)-I = a. This stems from the observation that (a- 1)-1 is an element of
G for whidl
Since a itself 1m.'! t.his propcrty, and sin(!c invcl'8CH have jUtit been scen to he
uni<lue, the ('1(,lIIcnt a lIIust he the inverRC of a-I. For the proof of the next
theorem, we shall require a prl'liminary lemma.
Lemma. If a, b, c, dE G and (G, *) is a 8('migroup, then
(a * b) * (c * d) = a * b * c) * d).
Proof. L('t UII t('lllpomrily dpn()t.(! t.lw produet c * d by x. Then, Killce the
opemtion * is I\.'!.'l(!iat.iv(!, WI! hnvn
a * b * c) * d) = a * (b * (c * d
= a * (b * x)
(a * b) * x
= (a * b) * (c * d).
Theorem 2-3. If (G, *) is a group and a, bEG, then (a * b)-J = b- I * a-I.
That is, the inv('rRC of a prodmt of group clements is the product of their
inverses in reverse ordlr.
Proof. A(!cording to the dpfinition of inverflC, nil we nc(ld to show is that
(a * b) * (11- 1 * a-I) = (b- I * a-I) * (a * b) = e,
how(wl'r, in t.!1(' ahM<!lw(! of t.hiH hypothl'His tl1l'n1 is 110 gUI\rn.ntee t.hnt the inverse
of a produd is e(]uul to the produet of the respective inverses. The next example
should help to make t.hese not.ions c1enrer.
Example 2-23. Lpt. r: dC'lI11l.ll tim HI1. (If nil urdtln~1 pnil"!! of renl 1lI11nbers
with nonzero first. ClolllpOIwnt. If t.he hinary operation is defined on the
sct G by the rule
(a, b) * (e, cl) = (ae, be d), +
t.hen (a, *) is n nOllc'omlllutntive ~roup [Example 2-21). The ici<'ntity clement
of t.he group iK t.Iw pnir (1,0); the inverKll of lUI eleml'lIt, (a, b) EGis
(1Ia, -l)la). A din'lt c'()JlIIHllat.ioll KIIOWI'I t.hat,
we obtain
(1, -:I) (l, -2) = (1, -i),
so that
1,3) (2, 4-1 ~ (1,3)-1. (2,4)-1,
I,'or this ~rollP thl'lI, t.he inverl'lc of a pmchl<'t of I'lements if! not equal to the
procllld of t.hC'il rC'sJlc'diVtl illVClrKC'H ill clirC'1't ordm', ThiH Hhoulcl lint. be plLrtic-
ularly surprising inasmueh as the group (G, *) is noncommutativc.
The previous Il'mma is Iwtually a special case of a result mentioned carlier
whil'h in C'fTC't'1. aSKI'rts t.hat, pan'ntheRCs nrc superfluous in 1\ product of group
e\C'IIl('nt 1'1, and C'OIlH('qU(lIt.1y t.hat t.Iwil' olllisHioll ('lUI Il'ILd t.0 110 misunderstanding.
TI\(' exa(t story is told ill thl' npxt t.hcort'lll.
First., howevl'r, let us int.n>duI'" Homc ll.lIxiliary notut.ion: Impposc (8, .) is a
Hcmigrollp I1l1cl \.lie ,,11'IlIPlltH a I t a2t ' , , a,. h('\ollg 1.0 S (whcre 11 > 1), Their
44 mWlll' TIIJ<:OIlY 2-2
Whl'n n = 'I, for imll.iuH'l', u. (12 a:1 a4 will t.urn 0111, t.o he
Next, assume that the assertion holds for all products of m: factors, where
a :5 m: < n, and let x be any product involving n factors.
Sinccx is obtained by slleceflsive applications of the operation -, there must
be a final multiplication between two expreflsions having less than n factors.
By the induction hypotheses, each of t.hese two expressions equals the standard .
product, flO that x may be written as
l:5k<n.
whence
This completes the induction stcp and therefore the proof of the theorem.
Given four elements aJ, a2, aa, a4 E S, parentheses can be legally inserted
and the elements mult.iplied (in the given order) five different ways, Theorem
2-4 Jl(~rmjt.H \If'! to cO/u'lude that all theHC produnts nrc cqual:
(a * c) * c:- I = (b * c) * c-'.
This last theorem allows us to cancel, from the same side, in equations in-
volving group clements. Wc cannot conclude, however, that a * c = c * b
implies a = b, unless the group is known to be commutative.
An arbitrary binary operation need not satisfy the cancellation law. To see
this, we consider the set G = {I, 2, 3} under the following multiplication table:
* I 2 3
1 I 2 3
2 2 I 2
3 3 2 I
2 * (2 * 3) ~ (2 * 2) * 3.
Theorem 2-6. In a group (G, *), the equations a * x = band y *a = b
have unique solutions.
Proof. First, x = a-I * b satisfies thc group equation a * x = b, since_
a * Xl = b and a * XII = b.
2-2 CERTAIN ELEMENTARY THEOREMS ON GROUPS 47
However, this situation is plainly incompatible with the above theorem which
IlSserts that there is one and only one solution of the equation a * x = b. A
proof for columns can be obtained by imitating the argument for rows.
It can be shown that all groups with fewer than six elements are commu-
tative; thus a noncom mutative group must necessarily contain at least six
clements. The proof of thill faet ill ROmewhat, lengthy, although the actual
details are not by themselves particularly difficult. The subsequent lemma
will serve to isolatc the most tedious aspect of the theorem.
Lemma. If a and bare noncom muting elements of a group (G, *)-that is,
a * b b * a-then the clements of the set,
{e, a, b, a * b, b * a},
are all distinct.
Proof. The basic idea of the proof is to examine the members of the set
{e, a, b, a * b, b * a} two at a time, and show that each of the ten possible
equalities leads to a contradiction of the hypothesis
a*bb*a.
On several occasions, the cancellation law is used without explicit reference.
The argument runs as follows:
1) e = a implies a * b = e * b = b = b * e = b * a,
2) e = b implies a * b = a * e = a = e * a = b * a,
3) e = a * b implies a * e = e * a = (a * b) * a = a * (b * a), so that
e = b * a or a * b = b * a,
4) e = b * a implies e * a = a * e =.a * (b * a) = (a * b) a, so that
e = a. b or b * a = a * b,
5) a = b implies a * b = a * a = b * a,
6) a = a * b implies e = b, reducing to case (2),
7) a = b * a implies e = b, reducing to case (2),
8) b = a * b implies e = a, reducing to case (1),
9) b = b a impli(,H e = a, reducing to case (1),
10) a *br b * a contradicts the hypothesis.
The proof of the lemma is now complete.
Theorem 2-7. Any Jloneommutative group has at least six clements.
Proof. If (G, *) is a noneommutative group, it must have a pair of noncom-
muting c1emcnts a and b. According to the lemma, the set fe, a, b, a * h, b a}
then consist.s of distinet member.;. We now proceed to establish that one of
tho group ('I('men/s a * a or a * h * a is distinct from these five; however, it is
not possible to lSJlceify ablStraeily whether it is a * a or a * b * a.
48 GROUP THEORY 2-2
With the aid of the lemma, we first show that a '" a is different from each
member of {a, b, a '" b, b '" a}. To start with, observe that
a) a '" a = a implies a = e, reducing to case (1) of the lemma,
b) a '" a = b implies a '" b = a'" (a'" a) = (a'" a) '" a = b '" a,
c) a '" a = a'" b implies a = b, reducing to case (5),
d) a'" a = b '" a implies a = b, reducing to case (5).
Thus, either a '" a ~ e, in which case a '" a is the sixth distinct element of 0
or else a '" a = e.
In this latter instance, we ('an show a '" b '" a to be distirwt from each of
e, a, b, a '" b, b '" a and (~onS(CJ\wl\t.Iy to be th(~ required sixth clement. The
reasoning IlI'/'/' ,I"I)('nclH on the fud that
c) a '" b '" a = (' implies b '" a = a '" (a '" b '" a) = a '" e = a, reducing to
case (7),
f) a'" b '" a = a implies a '" b = e, reducing to case (3),
g) a '" b '" a = b implies a '" b = a '" (a '" b '" a) = b '" a,
h) a'" b '" a = a'" b implies a = e, reducing to case (1),
i) a'" b '" a = b '" a implies a = r, reducing t.o ease (1),
whi('h cstllblishcH till' n~HIlIt. Anltlt.crnative proof will be presented in Acetion 2-fi.
The generalized associative law insures that the product a '" a"" .. '" a has
I\n unarnhigtlouR mel\ning irreHpectivc of how the factors arc parenthesized.
Designating the foregoing product hy the symbol ale (assuming there arc k
factors), we may int rodUl~e the notion of th(~ pOHitive powers of a. The next
definition extmlllH t.hiH i<iI'IL t.o ILrhit.mry int.(grnl 1)('w/~rH.
Definition 2-12. In any group (0, "'), the integ'ral powers of an element
a E 0 are defined by
where k EZ+.
With t.hese convent.ions, t.he customary laws of exponents have their counter-
part in group theory.
Theorem 2-8. Let (G, "') be a group, a E G, and m, n E Z. The powers
of a obey the following laws of exponents:
2) (an)'" = a nm = (am)",
4) e" = e.
2-2 CEHTAIN Jo;Lt:MENTARY THEOHEMS ON GROUPS 49
Table 2-3
The ('olllpll'le llIultipli('alion talll!' for the op!'ratioll * is shown ill Table 2-a,
Note t.hut /lallo 8('rVCH liS the idcntity elCIIH'lIt ami cl1l'h of R ISI )' /l3so, H, V,
D" alHI [)2 il:l itl:! own illv!'rse, whereas Roo lind U 270 ar<~ inverses of mteh other.
50 GROUP THEOUV 2-2
The assoc~iativc law also holds, but this is not immediately obvious. We shall
sec later that thc symmctric!! of the !!quare are equivalent to a group of perm uta-
tiom:! of the Het {I, 2, :l, 4} (obtICrve that a HymmetrY ill complctely dcscribc(~
by its effed on the vcrticlcs) and asso(~iativity therefore follows from the associa-
tivity of funetional composition. Granting this for the moment, the proof that
(G, *) constitutes a group is complete.
Similar groups may be defined for other geometric figures; in fact, for any
regular n-sided polygon. Problem 14 at the end of this section den.ls with the
group of symmetriel!l of the equilatcro.l trianglc.
Example 2-25. Let (G, *) bc an arbitrary group. For a fixed clement a E G,
defin<l the left-multiplication function fa: G -> a by
PROBLEMS
1. Given that a, b, c, Illltl d arp l'\cmcnts of thc scmi!!:roup (G, *), prove that
a * b) * c) * d = a * (b * (c * d.
2-2 CERTAIN ELEMENTARY THEOREMS ON GROUPS 51
then a 4 = e = b4
9. Prove that if (G,.) is a group having more than two elements, then there exist
a, bEG, with a "" b, a "" e, b "" e, such that a. b = b. a.
10. If (8,.) is a semigroup with identity and G the set of all elements of 8 having
inverseK wit,h rC!lpect to the operation *, verify that the pair (G, *) ill a group.
II. Prove that a group may alternativl~ly he defined as a semigroup (G,.) in whieh,
for all a, bEG, ea(~h of the equations a x = band y. a = b has a KOlution in G.
[Hint: Use the eharacteriza~ion of a group given in Problem 14, Section 2-1.)
12. Hhow that any semigraul) (."I,.) with a finite number of element!! possesses an
iil,mJlot.ent e\em('nt. II/int: For 11 > I, proceed hy induction OIl n, the number
of elements of ."Ii givon a E S, lot o't = {ak I k = 2,3, ...}, and argue according
to whether or not .tI containll the clement a.)
13. For allY syKtem (B, .), define the Het
15. Let (0,.) ancl (11,0) he two difltin~t groups. Define a hinary operation' on the
Cartellian product
ax II = {(g, II) Ig E (I, II E II},
all followK : for
Prove that (0 X H,') ill a group, called the direct product group of (0,.) and
(H, o)j show further that this group is commutative whenever the original groups
are ('ommutativc.
This SC(t ion is tI('vot~,tI 10 an eXllIllilHlI ion of I,wn important and frequently
m,d grllllpH: til(' gl'ClllJl of illlC'g(,r!I Illoclul" ft allcl Ihe 1!;11111P of l><'rmutntionH of
the I'lellwllts of a 1'1'1 (tllC' s(J-I'alled HynlllH'1 ri(' group). Whilc thcl!C group!! nrc
of ROIlW int('rC',,1 pl'l' 1;(', our mnin purpo"C' ill illt.rochlC'inp; them ill to provide two
more ('oncrete examplelS 10 iIIustml.e eonccplK which will be developed subse-
quently.
W(! hegin wit h all illvl'KI igal.ioll of t.he not.ioll of ('ongrtwne(', in t.erms of
whieh the group of illt.cWr!I modulo n will hI' forlllulat(!d.
Definition 2-13. Let n be a fixed positive integer. Two integers a and b are
said to be congrent modulo n, written
a == b (mod n),
if and only if the difference a - b is divisible by n. That is, a == b (mod n)
if and only if a - b = 1m for some integer k.
For instance, if n = 7, we hnve
== 24 (mod 7),
:J
-!i == 2 (mod 7),
-s == -50 (mod 7), etc
If a - b is not divisiblc by n, we Ray that a is incongruent to b modulo nand,
in this cast', write a of h (mod n).
It ill noh'wort hy I hat, C'vC'ry pair of int.I'p;('fII nre ('onp;ru<mt. mociulo 1, while II.
1':1 il' of illl C'g(,l"~ al'" ('OIlI(I"IIC'II1. lI\o(hllo '! pnlViclt,t! I.ht,y an' hot.h (WI'II or hot,h odd.
Our firHt thc'orl'lI\ pnlvicll's a useful .. Iuund.('rizltt,ion of (~ongrll('JI('(l modulo n
in tl'rms of rCIlHliIHlerl! 011 division hy n.
Theorem 2-9. I..(It 11 Iw a fixed J>o~itive intl'gl'r nnd a, b be arbitrary integers.
Then a == b (mod n) if alld only if a and b have the same remainder when
divided by n,
2-a TWO IMI'OItTANT GIlOUPS 53
Proof. SUppOIIC firMt a IE b (mod n), 110 that a = II + kn for HOme integer k.
On division by n, b lelwes a certain remainder r:
II = qn I r, whef(l o :5 r < n.
ThUI!, a = b '1- kn = ('1 1- k)n 1 1', whi"h HhowH a has t.he Maille remainder
I\!! b.
On the other hand, let a = qln +
l' and b = q2n +
r, with the same re-
mainder r (0 :5 r < n). Then
h a -. ( k)n,
a - b = kn and b-c=hn
AIHO
[I) = {X EZ IX = 1 + ak for som<' k E Z}
{... , -8, -5, -2, 1,4,7,10, ... }.
Similarly,
[2) = [ ... , -7, -4, -1,2,5, X, 11, ... }.
Obs<'l"Ve that every integer lies in olle of these three classcs. Int.egers in the
sanle eonp;rucn(~c dass af(~ eongl"lll'nt Il\odulo 3, while integers in different
c-Iassl's an' iIH" 0 IIJ?; r \1('11 I lIIodulo :t
A particular ('()ngru('n('(~ (,lass Illay be d('sigllal.(~d in a variety of ways by
merely changing its reprcRentativ(-. I II the above illustration, for instance,
It sulli(~cs 1.0 rl'lIIllrk t.hat the eharaeteristic feat.nrc of t.lwsc variou!! represcnta-
tives is that., in I his I'as(', tlH'Y all difTer from clwh other by multiples of 3, and
in p;(,IH'ml, difTer by multiples of n. For ('onveni(mee, onc often selects the
slllllll('st nOIlIlI'p;at.ive int.cp;pr from ('tuh congruence dass t.o represcnt it; in
pr:wt.i(I', WI' shall :ulh('f(- 1.0 t.his notat.ional (onv('nl.ioll.
To return 10 t.11e general ease of congruence modulo n, let
t.his i:> not quite in accnr'll with our intuition, we bow to long-standing custom
I1IHI Hindi eOllt,illll!' 1.0 n~fl'r 1.0 Z" aH t.h(, illt,l'lI;l'rloi IIIOdliio n.
Bya partition of t1w set S, we mean a family of nonempt,y subsets of S which
are pairwise disjoint and whose union is all of S. It, follows from Theorem 2-12
that, for eaeh n E Z+, the integers modulo n {~onst.itute a partition of the set Z.
Definition 2-15. A hillary operation +" may be defined on Zn 11.'1 follow:>:
for ('ad, [al. [II] E Z", 1'1, [a] -f .. [II] ~' [alii].
D{'finition 2-1;; ass!'!'ls I,hat. Hw lIlodular sum of t.wo {~ongruellcc (,lasses
[a] and [b] is tl\(' Uni'll\(' mellliwr of Z" whil'h ('olltaills the ordinary sum a II. +
Ilow(w(~r, th!'r!! is It ImbUe prohlelll (olllll'et.l'd wit.h t.his delillit.ioll. IlIlllollllud,
as addit iOIl of ('ollll;rUl'n('e c:lItsses in Z,. is detin{'d in teruls of reprel:lCntatives
from th(,sl' dusses, we must show that the operation +n doeM not depend Oil
the two l'I'P!'l's('lItativ{'s dloS<1JI. It mllst be prov('d formally, that if [a'l = [a)
alld W) = [/1), tlH'II [a'] +n [b'] = [a] +n [b], or rather, [a' +
b') = [a b]. +
;\Iowa' E [a'] ~ [a] alld 1/ E W] = [IJ], whil'h impii!'s
While these results appear superficially different, both congruence classes [9]
and [-!)] may be !'xpressed more simply as (2). Thus, although written in
terms of different repreRellt.atives, either modular addition gives the same sum,
[:2). Other pORsihl1' ('hoi(~es [-4] -h [-X] = [-121. [17) +7 [0] = (23), f3] +7
[13) = [Hi], also yil'ld the sum (2).
We nt( 1I0W in a posit ion t.o prove one of UI(' prindpnl theorems of this section.
Theorem 2-13. For elwh posit.ive inl!'ger n, t.l1(' mathematical system
(Z", t-,,) forms n ('omlllutative group, known as t.he (/1'OUp of inieyel's modulo n.
J'l'()of. TllP ass()('iat ivily and ('ommut.ativity of t.\w opl'ralioll -1-,. nre 1\ direct
consequenee of til(' sanl(' properties of the intl'geriS under ordinary addition.
2-:1 TWO IMI'OHTANT GROUPS 57
+4 0 1 2 3
0 0 2 :~
1 1 2 3 0
2 2 ;~ 0 I
3 3 0 1 2
For the seeollu of our I.wo exnmpleH, let us turn to the study of permutation
groups. To this clld, supposc Umt N is n finite set having n clements which,
for simplieit.y, we tnkl' to be t.he first n natural numbers; that is to SI\Y,
N= {1,2, ... ,n}.
58 GROUP THEORY 2
where till' (~nrrl'Rpollding imag(,R uppear 1)('low eaeh integer. Clearly, the ord,,
of the ('ICIIII'lIts ill th(' t.op row of thiR lIymhol ill inulIllwriul, for th(l columll:
may lw rl'arrnllged without affectillg t.he natufC of the function. Prccil:ll'l)
speaking, if (J iR Illl arbitrary p(!rmutation of the integers, 1, 2, ... ,n, thcn .'
could equally well b(' given by
From thill, WI' illf('r UIIlt. (,Iwh of the n! {>I'rmutntions in' 8" may be writ,tA'I'
in n! ditTl'fCnt way!!. For illlltarU'e, the following two symbols both reprcsclIl
the same clement of 8.:
(1 2 3 4)
2 4 :1 I '
( 214:i).
4 2 1 3
(1 2 (1 2 ...
fog = f(t) f(2)
n)
f(n) 0 g(1) g(2) ... g~)
g~)
( (1(\) g(2) g(n) ) ( 1 2
= l(g(1 f(g(2) ) f(g(n 0 g(1) g(2)
f(g~
2
= C(g;l) l(g(2 )-
2-3 TWO IMPORTANT GROUPS 59
What we have done is to rearrange the columns of the first (left) permutation
until its top row is thc same as the bottom row of the second (right) permuta-
tion; thp product /0 g is then the permutation whose top row is the top row
of the se(!OII(I factor and whose bottom row il:l the bottom row of the first factor.
With a little pmet.iee one can evaluate produet.l:I without luwillg to w.rite out.
this intermediat!' preparation. }lany authors prefer to carry out the multiplica-
tion of P("llllll.ations in the oppo:-;it,(! onl"r (that ill, they apply the factors in It
prouud frolll )(ft. to right.), alld t.he read('r :-;hould h, parl,il'tllllrly wlltehful for
this,
Before stating a theorelll whieh indieutell the algebmic nature of Sn under
I his 1lI,,1 h()(1 of (!Ollll)()sit ion, we hop(~ to dnrify sOllie of till' foregoing "ointH
Wit.ll IlII ('xILlllple,
Example 2-27. If 1.111' s(I. N ('onsiHls of t.he illt.(/.t(rs I, 2, a, tlllln t.hcn! nrc
a! Ii Jll'rlllllf.nl.ions ill Sa, 1I1L1l1dy,
II =C :~) , 2
2 :~
/2 =C :;) , .ra=Ca :J) ,
2
2
:l
2
2
f4 =C :) , fs =C :a)J , .f6 =C ;) .
2 2 2
:l 2 3 2
.f4 f6=C
0
2
a
a)'Ca D
2
2
2
=G
2
:~ :)G 2
2
:J)= C
I 2
2
:l
:)=12,
16 f4 =Ca 2
2
:;) c 2
a
:~)
2
The proof of f his fad. is OIllil.t.('(1 in:l.'lIl1ll('h as !l mom gllllCral version of the
thcorem will be l!;iv('11 shortly. III Jlus:sing, it is only 11I..'Cc8I!ary to note that the
(iO I:llOl'1' 'I'll EOHY 2-:\
2 n),
2 n
whil<' thc 1lI111t.iJllieativ(~ inverse of any permutation f E ,sn is described by
for 1 $ i < k,
f( 11k) - nI , alld
fen) = It for n fl {nl, n2, ... , nl:},
(2 5 3) = (5 3 2) = (3 2 5).
2-:l TWO IMPOHTANT GROUPS 61
(2 5 a) 0 (t 2 4 a) = C 2:l45)
4 135
since
f(I) = fl, r(I) = /(3) = 1,
and
f(2) = 4, /2(2) = f(4) = 2.
The simplest of pennut.ations arc undoubtably the 2-cycles, for they just
int.erehnnJl:(J t,W() dl'flll'nt.s and Il'nve all ol.llI'rs fixed. It is cust.omary to refer
to sl.wh eydeH as Imnxposili()ns.
Corollary. Every permutation may be expressed as the product of trans-
positions.
Proof. In light. of the preceding result, it suffices to write any k-cycle as a
product of transpositions. A direct computation shows that this can be done
rather simply in the following manner:
As the above trnlll'ipOAitiollS havl' all illt,(~jI;('r in ('om mOil, t.hey do not, ill
general, commute. Furtlwl"more, the decomposition is by no means unique.
For instance,
(1 2 3) = (1 3). (1 2) = (2 1). (2 3).
While n jl;ivl'n pl'rlJlut at ion may hI' fUl:tored into It produet of tmnspo8itions ill
a variety of ways, tl\(' IIl1mlwr of \,mllsp08itioml involved will alwaY8 have till'
same parit.y. That is, if 0111' fadol"i7.lIt,ion has lin ev('n (odd) number of tmnsposi-
tions, then cvrry faI'tOl"i7.:t\.ion mllst have an cvrn (odd) numher of transpositions.
The notion of permutation as presented in Definition 2-16 is not. sufficiently
general for all ollr pllrpos('s, since it restriets liS to finite Rets only. It would
sl'em Illore lIat uml to dl'finp this I'OIWI'P\' for arhitrary Het.H, fillite or infinitl',
as follows.
Definition 2-18. If G is any nOlll'mpty set, then n permutation of G is a
one-to-one function from the Ret G Ollto itself.
We rcprl's('nt tIll' spt of all pNmutatiolls of G hy the symhol, sym G (tlll'
reason for this ehoi('e will he clear in a moment). To he sure, if G is a finitl'
set having n ('11'1111'11\ s, tlwn sym G = 8,..
I n this genrral fmml'\\'ork, it, is possihle to prove a st.rueture theorem from
which Thrort'1II 2- 14 will follow as a speeial ense.
Theorem 2-16. For a 1I0011'mpty set G, the pair (sym G, ~ constitutes a
/!:roup, ('alled Ihe symmetric IImUJI of G.
['roof. Tlw firs\. thill/!: to til' dlllll' is to show that. is adually a binary operation
on I:lylll U. (:ivplI nrhit rary pC'rIJlII\.atiom; I, f/ E sym fl, t.he ('omposition f f/
if! ohviollsly It fUlII'tioll frolll (J illto a. If t.he e1l'llll'llt C E a, thell t.here exists
sOllie /) E (J SIl"'1 tim!. 1(11) .. c; silllilarly, thl'm is lUi dellll1llt a in (J for whieh
g(a) = b. HC'Il!:I',
(f. y)(a) = I(f/(a) = I(b) = c,
1'11(' as.'<llI'ial ivily of fUlldiollal ('Olliposit iOIl follows frolll Tll('lIrCIll )7.
Plllinly, till' irhlIlit.y fllllr:tioll ia iH a J1prnllltat.ioll of r; lWei is :-nwh thnt.
f io = f = io I,
2-3 TWO IMPORTANT GROUPS 63
rOl'lLny I E sym (f, showing; i(; to Ill' lUI identity elemcnt for the liystem (sym G, 0).
If I E sym (;, the inverse fUlwtion I-I exists, is a one-to-one function, and
maps the sl'!. (; Ollto itsdr. l\Iorl'ovPI', I-I is the inverse of I with rClipeet to
C~olllp()sitioll. All of whi .. h justifies th .. st.atpmpnt that (sym G, .) is 1\ group.
PROBLEMS
and
11. Express the following permut.ations as (a) productll of cycles having no elements
in common, (h) prudul'l.!S of transpositions.
2 :J 4 5 6) (1 234567)
(j 4 1 2 5' Ii 7 (j 2 1 3 4
12. Hhow t hat, the e,\'C'le (1 2 ;~ 4 5) may he written as a product of 3-cycle.<l.
1a. Verify the relutioll (1 2 :J .. 11)-1 = (II 11 - 1 ... 2 1); in particular deduce
t.hnt. C'vl'r,\' tl'lIl1spo"it.ion is it." own inversc'.
14. Thn sYIIIIII('tril's of thl' S(pIII.rll (Exulllple 2-24) muy be interpreted &8 permuta-
lions of t.he. vl'rti(,C>!. HCI'fl'sent (,ach of thCI!IC lIymmctricli by a corresponding
permutation.
64 GROlTP 'fHEOIlY 2-4
! _ (1 2 3 4 5 6),
2 :J 4 a 6 1
2-4 SUBGROUPS
There nrc two Ktnnd:ml t.eehniqul's in at.ta(king t.lw problem of t.he strueture
of a parti('ular j!;roup. One method ('ulls for finding all the subgroups of the
group, with the hopl' of gaining information about the parent group through
its local structure. The other approach is to determine all homomorphisms
from the given group into a more familiar group; the idea here is that the
images will reflect some of the algebraic properties of the original group. On
dOKl'r Kcrutiny. WI' KhaH KI'C that while HICKe lill('to! (If inveto!tigation aim in
difT('rent dirpdions, t.hey nrc not entirely unrelnt(!d, hut rather'll.spects of th(!
Kiln\(' probl(m. For the moment, however, our attcntion ito! focused 011 analyzing
It group by means of its subgroups; the question of structure-preserving map-
pings is a morl' sub tic matter and will be deferred to n lat.er section.
From variolls examples nnd exercises, the reader may have noticed that
"I'rlnill suhsl't.s of 1.1 ... 1'II'III1'II1s of n grouJI Il'llcl 1.0 III'W I!;rlluJls wlll'lI UlIt! n!Htri(l.s
tlw group olIPratioll to these subsets. It iH this situation in which we shall be
primarily interested.
Definition 2-19. Let (G, *) he n group and II ~ G be a lIonempty subsef
of G. The pair (II,.) is snid to be 1\ subgroup of (0, *) if (II,.) is itself n
group.
Eaeh group (0, *) hUH two oLvioUl~ subgroupH. For, if e EGis the idcntity
clement of th(! group (0, *), t.hen both ({e}, *) and (0, *) arc subgroups of
(G, *). These two sUhgfOUJlIl arc often referred to as the trivial subgroups of
(G, *); all suhgroups hdw(,pn thelle two extrernefol are calIed nontrivial sub-
groups. Any subgrollp different from (0, *) is termcd lIroller.
Example 2-28. If Z. and Zo dcnote the setH of even and odd integers, rcspe(~
tively, then (Z., +) is a subgroup of the group (Z, +), while (Zo, +) is not.
Example 2-29. Consid('r (Zo, +6), the group of integers modulo 6. If
II ~ {O, 2, 4},
2-4 SUBGROUPS 65
then (H, +6), whose operation table is given below, is a subgroup of (Zo, +0)'
+0 0 2 4
o 024
2 2 4 0
4 4 0 2
Example 2-30. Let (G, *) he the group of symmetl'ies of the square (Ree Ex-
ILlIlP'" 2 21), wlll'I'I' a 0_:U uo , H IH ." H 2711 , H:lflfl , II, I', /)It 1)2] IUIII t.11I' OPI'I'n-
tion * l'onKiKt.!oI of following olle mot.ion by anot.her. Thi!ol group I~ollt,ltin!ol eight.
nont.rivial subgroups. We leave it to the reader to V('rify timt the following
sets comprise the ei(!ment.s of these subgroups:
Suppose (H, *) is a subgroup of the group (G, *). Since the identity element
of (H, *) satisfies the equation x * x = x, it mURt be the sn.me as the identit.y
of UI(' pnnmt. group (a, *), for otherwi!olc we would hlWI~ two idl~mJlotcnt ele-
nll'lIt.R in a, l~olltrlLry to Theorem 2-li. The identit.y demellt of a group thUl:l
also serves 3.'> the identity element for any of its subgroups. Moreover, the
uniqueness of the inverse elements in a group implies the inverse of nn element
hE H in the subgroup (H, *) is the same as its inverse in the whole group (G, *).
I'll l'fltnhliHII tlml. n l(ivl'lI !oIlIhH(I. 1/ flf a, nlnlll( wil,h I,h,' illllul!l'Il IIJlIrntiml
of (a, *), eonHt.it.utl'!oI n subgroup, W(! IIlU!:!t verify tlmt nil the eonditions of
Definition 2-11 arc sl\tiRfied. However, the asROdat.ivit.y of the operation * in
II is an immediate consequence of its 3.'>sociativity in G, since H !; G. It is
neeeRsnry then to show only the following:
1) a, bEll implies a * b E H (closure),
2) e E II, where e iR the identity element of (0, *),
3) a E II implies a-I E II.
Needless to say, the saving in not having to eheck the associative law can
prove t.o btl cOIl!olidl'lable.
A theorem whieh e!olt.ablishcs a single eonvf'nient criterion for determining
subgroups i!ol given below.
Theorem 2-17. Let (G, *) be a group and 0 "t; H !; G. Then (H, *) is a
subgroup of (a, *) if and only if a, /J E II implies a * II-I E H.
Pl'Oof. If (II, *) i!ol a !oIuhgl'oup and a, bEll, tlWII II-I Ell, 1UIIl!olO a * b- I Ell
by the dosurl' "oJl(lit ion. COllvlrMlly, suppose 1/ is tL 1101l(!Jllpty suhset of G
66 GROUP THEORY 2-4
Thus cent. (J ('onsists of those clements which commute with every element
of O. For example, ill the group of symmetries of the square,
The reader may already have deduccU that a group (G, *) is commutative if
and only if c('nt. a = O.
As ilIustratiolJs of the usc of Theorem 2-17 in determining when a subset
of the clements of a group is the set of elements of a subgroup,..we present the
following two theorems.
Theorem 2-18. The pair (cent 0, *) is a subgroup of each group (G, *).
PmoJ. We finlt ohserve that cent Gis nonempty, for at the very least e E cent G.
Now consider any two elements a, bE cent G. By the definition of center, we
know that a * x = x * a and II * x = x * b for every clement x of G. Thus,
if x E G,
(a * b- ' ) * x = a * (b- I * x)
= a * (x- 1 * b)-I
= a * (b * X-I)-l
= a * (x * b- 1 )
(a * x) * b- I
= (x * a) * b- I = X * (a * b- I ),
which implies a * b- I E cellt G. According to Theorem 2-17, this is a sufficient
('ondition for (,pnt a, *) to he a subgroup of (a, *).
Theorem 2-19. If (Hi, *) is IUJ arhit.rary indexed collection of subgroups of
the group (a, *), then (nil;, *) is alHO It subgroup.
Proof. SilH'e tht' Sl't,; IIi all ('ontain t.Iw identity (lement. of (a, *), the inter-
scl'tion nil. ~~. 1\('xt" supposc a !\luI II Ilrc any t.wo clcmcnt8 of nH,; thcn
2-4 SUBGROUPS 67
a, bElli, where i rangeM over the index Met. The pair (Hi, *) being a subgroup,
it follows that the product a * b- I 11.\:;0 belongs to Hi. As this is true for every
index i, a * b- I E nIl.-, which implies (nHi , *) is a subgroup of (G, *).
In regard to the group of symmetries of the square, we could take
Proof. By now the pattern of proof should be clear. We assume that a and b
are arbitrary elemcnt.s of UIIi and Rhow that a * b- I E UHi. If a, bE UHi,
then there exist subsets II i, II j cont.aining a and b. rel'lpectivcly. Aecording to
our hypothesis, HiS;;;; H k and H j S;;;; Ih for some choice of lIt in {Hi}. Since
(H k , *) is a subgroup and both a, bE H k , it follows that the product a * b- I
belongs to H k Accordingly, a * b- I E UHi as was daimed at the beginning.
Asa particular case of the foregoing result, (~onsider just two subgroups
(H I, *) and (H 2, *). Theorem 2-20 may be interpreted as asserting that
(H I U H 2*) will again be a subgroup of (G, *) provided either HIS;;;; H 2 or
H2 S;;;; HI' 'What is rather interesting is that thil:! condition is I1eeessary, as well
as sufficient. The next theorem gives the details.
Theorem 2-21. Let (HI, *) and (H 2 , *) be subgroups of t.hc group (G, *).
The pair (H I U H 2,) is also a subgroup if and only if HIS;;;; H 2 or /J 2 S;;;; HI.
Proof. I n view of the prececiing remarks, it il'l enough to I'lhow thut if (II, U 11 2 *)
is l~ tiubgl'Oup, then olle of the I'lets H I or 112 must be contained in the other.
Suppose to the contrary that this assertion were false: that is, HI g; 112 and
112 g; II,. Then there would exist elementli a E III -- J/ 2 and bEll 2 - Ill'
!i8 WUltI!' TIH:OItY 2-,1
Now, if the procille! a * IJ were n member of the set Il" wc (~ould infcr that
b = a-I * (a * b) E Ill,
which it! clpurly not. t.rul'. On the other hand, the possibility a * bEll 2 yields
the equally false conclut!ion
The set (8) dearly exists, for G itself is a member of thc family appcaring
on the right; t.hat is, (G, *) is a (trivial) subgroup of (G, *) and 8 ~ G. In
addition, sinee S is ('olltnillcd in eaeh of the sets being intersect&d, we always
have thc inclusion S ~ (8).
Theorem 2-22. The pair S), *) is n subgroup of (G, *), known cithcr as
tl\(' f'nveilJpiny .~ulJyl'Oul' for 8 or the RUIJI/roup yenerated by the set S.
P/'(}of. The proof is !In imnwdiutl' consequl'lwe of Theorem 2-19.
Definition 2-21 implies thnt whenever (Il, *) is a subgroup of the group (G, *)
with s ~ II, then (8) ~ II. For this reuson, one speaks informally of 8), *)
us being the smallest subgroup which contains thc set 8. Of course, it ma:y
well happen that (8) = G, and ill such a situation, the group (G, *) is said
to bc (Jenel'ate(i by the subset S. For example, it is easy to sec that the group
(Z, +) is generated by Zo, thc set of odd integers.
We shall givl' an a\1ernative description of the sublict (8) which is frequently
elllii('l' to work wit.h than Definition 2-21. III what follow:;, the symbol 8- 1 is
u&-'<I to indicat~ t.he ('ollection of inverses of f'l('mlmts in 8: S- I = {a-I I a E 8} .
The result. we propose to obtaill i:ol that
All abhlpviaj.pd proof of tIll' ass!'rtion ahov!' llIi!1:ht filii as folloWll: TIll'
"yst(,11I (181, *) is a subg;roup or Ih(' g;I'OUp (a, *) wilh tlIP prnppl'ty '" ~ [81.
A" (8), *) is tll(, slll:dlpst. s\IIh sUbg;l'OlIP, it follows t.hat (8) ~ IS). Tlw rever~('
inl'\usion ill just ifi!'d by tlw fad that any l'HIi>gI'OUP whieh (~ontllins the set 8
Illust J1('(~!'sslll'ily ('olltain nil the l'I!'llI!'nts of [S).
A ('asl' of spp(~i:t.I illlportlllH'!' aris!'s \\'11('11 S ('onsists of It singl( plpllJ('nt a. In
I.his sitllatioJl, it is US \I: II to \Hitp (It) im;tpad of (fn:) alld n,f!'1' t.o t.ht' Ilsso(~iatpd
sUhg;I'Olljl (a), *) as the cydic SUIJUI'IJIlJl (/('naaled IllI a. The subs('1. (a) is rather
!'asy to d!'s('I'il)('; as all ils produds involve the cI('llIent a or its inverse, (a)
silllply I'('dw'('s to t he integral powel's of a:
It. is ('ntirdy possihl(' that. the gl'oup (a, *) is <'qllal to OIIP of it.s (yelie suh-
groups, that. is, for Rom!' (hoiee of a E a, (a) =, a. lInd!'I' t.heRe rircllmstane!'s,
the group (G, *) is I'ef('rrl'<i to as a cyclic YTOUP wit.h !It'lI/TatoT a. Thus, to say
a group is ('ydie llH'I\IlS tlmt. ('Itch of its members ('(UI Iw ('xpr!'Flsed ItS lUi intpgml
po\V<'1' of Home fixed clement of the group.
A (yelie group may POSSCHS severnl different gmwrators; indepd, one always
has (a) = (a-I). l'\otiee in addition that, as a rOnS('qllC'IlCe of the law of
el\pOllPlltH, any (yelie !1:I'OUP must 1)(' eommutativ!'.
To muk!' t hpse notions dearer, let UH pUllse to examine s('vNal examples.
Example 2-31. First eonsider the group of int.eg!'1'S (Z, +). In this case,
silH'P t.h!' 1I;1'01IP opl'mt iOIl is t.hllt, of addit ion, the ahst.md pro<iud a * a * ... * a
is rephu'ed hy t.he ordinary SIIIII a -I- a -I- ... + a. A(:eor<iingly, the subset
g(merat.!'<i hy till' ('Iement a tak!'s the form
(a) = {na I n E Z] .
Using this notation, [O} = (0), Z = (1), while Ze = (2). We may thus eon-
dude that both thc groups (Z, -f ) ntHl (Zc, -I ) are (ydie, with the integers 1
and 2 as their reHpcC't.iv(' gelwratOI'S.
Example 2-32. Another illustration is furnished by (Z 12, +12), the group of
integers modulo 12. Not.!' that we now write a +12 a +12" +12 a for
a * a * ..... a. The ('y<,li(' subgroup gellerated hy, say a, is ([0, :l, (i, 9}, +12),
for herc,
(a) = {an (mod 12) In E Z} = 10, :l, n, !l}.
+
As (1) = Z 12, til(' group (Z 12, 12) is its!'lf cyelie; other possible generators
of (ZI2,t 12) are;}, 7ltnd 11. It is lIot difficult, t.o SPC' that., in general, the group
of int!'g;!'!'s Illodulo 11 iH eyelie with 1 as a g(lwmi.or.
Example 2-33. 'Th!' group of SylllllH't!'il's of tI .. , s<ruan~ is not ('ydi!', for
('ydie gr'()lIpH al'l' tl('('('ss:trily eomnlllial ivC'. Havl' for t.he i<\l'lItity, t.he distind
70 (;Itol''' Tln:OllY 2-4
Proof. As the sl't (a) is finih', not all the powen; of the generator a are distinct,
There must be 801ll{' rl'JlI't.ition at = ai with i < j, On multiplying this equation
by a- i = (ai)-I, it follows that a i - i = e, Thus the sct of positive integers k
for which ak = e is noncmpty, SUppORe m is the smallest positive integer with
this prop<>rty; that is, am = e, while a k e for 0 < k < m.
The sct. 8 = :1', n, a 2, ' , , ,am-I} con8ists of distinct clements of (a). For
a' = a', with 0 ~ r < 8 ~ m - 1, implies that a'-' = e, contrary to the
minimalit.y of m. To I'olllplete the pl'Oof, it remains to show each member ak
of the group a),.) i;; {'qual to IlIl clem('nt of S. Now, by the division algo-
rit hill, we Illay writ(' k ~ 1J1It ~I l' for SOIllI' integers q !lnd ,. with 0 ~ r < m.
II CII(,(, ,
ak = (am)q. a' = e * a' = a' E S,
This means 011' 1-<1'1 (a) C;; S, yielding (a) = ["~, a, a:.!, . . , ,am-I) and the Bub~
sel)uent ellul\lity 111 = n.
If a is an el(,lllent of t.h ... group (G, *), we define the order of a to be the order
of the eydic! subgroup a), *) generntec.l by a, The last result permits an
altcrnativ(' vil'wpoint: the orch'r of a is the 1(,!tHt )lositivc integ{!r n, provided it
('xiHtl'l, HIII'1r t Iral (I." ('. Of 1'0111'1'1(', if 110 sud\ integ!'r (~Xilltll, a ill of infinite
orll('r. AM 1111 illusl ntl ion, I'ollside'r I II(' gl'OlIp (Z4' 14); Il('re, the elemcnt 0 has
order 1, 11m:; orcll'l' .. , :! hal'l 0('(\t'1' :!, whill' a hitS order 4.
III certai'l (,ltSI'S (unfol'ftIlIlL(('ly, far 100 f('w) , it. is pOl!sihle to characterize
compll'll'ly tIl(' suhgroups of a give'" gl'OUP, To cit.e 0111' instance, ill the
aclditivi' group of illll'gl'rs (Z,I), 1I11' I'IlIhj.!;IOIlJlH 11I'I' all of the form n), +)
for H()IIII' 110111 tIj.!;:t t ivl' illll'gl''' II, Adllally, 1he I-<ilu!tt ion is sonwwhat, mono
general, for it. I':\n hI' showlI that allY suhgroup of It c'yelic group is again cyclic;
we shall diH('uHS this III'Xt.
2-4 KUBGHOlTPS 71
I'/'III/f. LI't. (a), *) 1)(, n ('y('li(' Jl;roup gl'lIl'rah'lI by Uw 1'11'1IWIlt. a llllll let, (1/, *)
be OIW of its liubgl'Oups. If /I '''' :(':, LlII' t.!WOI1' II , is t.rivially l,rllI' , for (teJ , *J
is the l'y(lie subgroup gl'lIcmtcd by the idcntity e1emcnt. We may tlll.IS supposc
the set H {e}. If am E H, whcre m 0, then a- m is allio IlII clemcllt of H;
hence, II must. contain positive powers of a. Let n he thl' smalll'l'It. llOsit.ive illtl'l!;l'l'
sueh that an E /I. We propose to show Il = (a").
To ('stablish till' inl'lusioll Il <;: (a"), h,t a k bl~ all lU'bit.I':tI'Y I'lellll'nt in till'
KeL 1/. The divisioll ulJl;orithlll illlplil's 1.llI're I'xist illtegl'J'M q nnd r fill' whieh
II *K = {h * k I It E H, k E K}.
A brief ('Olllml'lIt 011 IIot.at.ion that should Ill' lIuule is that the usual custom
is to writ.e Hie produl't II /I 1I11'1'('ly 11.8 1/ 2 :lnd, if olle of the sets eOllsist.s of a
sillgle delllt'llt a, silllplify tal * II to a * II.
At first sight, th(~ readl'r JIlight real-lollahly ('ollj('('ture that whenever (H, *)
!lud (K, *) an' both subgroups of (0, *), tllI'lI (II K, *) will also bl' a subgroup.
The group of l-Iymmdril's of the square, howev('r, showH that such It simple
outeOIlHl is 1101, to bl' exp(,(,t.('d. IIl'n', it is ('Ilough to (~ow'iider UII' subgroups
hltvillg I'hnlt'lIt.s /I = {N: wo , /),: nlld K =- {U: wo , V:.
A lJui(~k dw('k ('stah-
lishes that there is 110 subgrouJl whose melllbers comprise th(' produl'l. set
In fad., Hl(, set II K isn't even dmled under the group opemtion. One ne('d
not b( dismaYl'd by this state of affairs, for an additional assumption on til\'
subsets II and K readily overcomes the difficulty.
Theorem 2-25. If (II, *) and (K, *) are subgroups of the group (G, *)
such that II * K = K * II, then the pair (II * K, *) is also a subgroup.
Proof. Il1nol!uolIS as the equality /I * K = K * /I appears, it is nonetheless th(!
source of sOllie difficulty. This notation docs not mean Cll.!!h element of /I
eommutes with ea('h element of K; all it signifiC/!! is Umt whenever hand t.-
Itrc arhitrary lIIelllhers of 11 and K, then ther(! exist elements h' E 11, k' E K
for which It k = k' * h'. Bearing this in mind, let us proceed with the proof
proper.
Plainly, HI(' produd. set 11 * K iii noncmpty, for e = e * e E II * K. Now,
I('t a and b he allY pair of clements in II K. Then a = h * k and b = h, * k,
for suitable .. hoice of h, h, E II and k, k, E K. As usual, our aim in what
follows is to show that the produet a b-' lies in II * K. This is achieved
through first noting
a */)-' = (h. k). (h, * k.)-' = h * k * ki') * hi').
Since K is elOS('<i under *, the clement k * ki' belongs to K and consequently
for the closure of t.he set II insures II * 112 also is a member of it. To complete
the proof, it suffices to invoke Theorem 2-17.
Corollary. If (II, *) and (K, *) arc Imbgroupli of the (!ommutative group
(G, *) UH'1l (II * K, *) is llJl;aill Il :,;ubgrollp.
Tlw III ility of Th"l/fI'lIl 2 2!i IiI'S in the flld tlmt it pcrmitH another character-
izat ion of the suhgroup Jl;l'rwml.ed by a union of sets. What happens is thi:,;:
if th(' pair (II * K, *) forrw:I a suhJ!;roup of (G, *), it mllst in fact be the sub-
group gl'nerat('d by II UK; in symbols,
Let us briefly outline the argunwnt involvt'd. To start with, the two inclusions
Hence, the subgroup of (Z 12, + 12) gel1emtl'd by the union {0,6} u {O, 4, S} is
just ({O, 2, 4, 6, H, 1O), + 12)'
PROBLEMS
t. In c'lwh of t.hl' r"lIowinl( elL~C'S, c'stahliHh tlULt (//,) iii IL rmhl(fUUp of tho p;roup
(,', .) :
a) II = {I,-I}, (} - {1,-I,i,-iJ, whnrci 2 - -I;
b) II = {2n 111 E Z}, G = Q - {O};
e) II = Q -" {OJ, G = n' - {O) ;
d)/I = {(1 + 211)/(1 + 2m) 111, /1/ EZ}, G = Q - {O).
2. Prove that ({O, 4, X, I~; ,I-Jn) is It subgroup of (Ztn, -hn), t.he group of integers
lIlodulo IIi.
:1. III CIt(, SYllllllc'tri .. f,\rc>UJl (S,.. uJ, ld If tlt'noh, t.he 1:ICL of (l1'rIlIlILlttiuns 1l'ILvinl( the
integer n fixed:
II = {f E S. I/(n) = 11).
Ohsl'rvl' that in tIll' additive group of intl'gl'rs (Z, +), where the subgroups an'
of thl' forlll n), +), 11 a nonnl'gatin' intl'gl'r, Ihis fl'latiun reducl's to eongrul'nl'l'
modulo 11.
6. SIIPPOSl' thnt. (0, *) is a p;roup 11.1111 aE a. Let Cia) denol.e the Sl't of all elements
of () whieh ('ollllllute with a:
Cia) = {x E G I a * x = x * a}.
Prove that. thl' pair (C(a), *) is a Ruhp;roup of (U, *), known /1..'1 the centralizer
of a in a. ,\Iso vl'rify the equality, l~cnt G = n.'EG C(a).
7. <livl'lI (0, *) is n finill' p;rnup, prove t.hat
IL) t.Il1'rl I,isls Ii posit.ivl! inlc'p;(!r 11 Mud. t.llILt a" = I: for all ( E (1,
h) if II is a nOIll'llIpty Huhset of U which is closed under the operation *, then
(II, *) is a subp;rollp of (0, *).
R. In t.hl' ('olllmlllativl' p;roll(l (a, *), define the Sl't II by
17. Consider the group of symmetries of the square. Use Theorem 2-25 to obtain
the subgroup J.!;cnprll.ted by II U K, where II = f RplO, R:wo) , K = {Raoo, Dd.
IS. Ll't (0, *) hI' It J.!;rollp of ord('r n, W}IPf(' n is odd. Prove thllt ('Iu:h "Ielnpnt. of G
is Ii sqIlIL),l' (i.('., if rEO, thl'1l .x = y2 for ~Olll(, y ill G).
Deftnition 2-23. Let (H, *) be a subgroup of the group (G, *) and let a E G.
The set
a * II = {a * It I It E II}
a=a*eEa*lI,
71i 2-1i
t.hat is, every c1('m('nt a of G beJon~8 to some left coset of H, and more specific-
ally, to tIl!' ('081'1. (l II. We shnll nmkl' uS(' of t.his fad in 1\ Iitt.J(, while,
We nolp further that t.here is It onl'-to-on(' ('orreHpondence between thl'
('1<'111('11 Is of II llIlIl t.hoS(' of any "OHet of II. I n<iel.'d, if a II ill 11 I(~ft coset of II,
WI' may d('fill(' It mappinJl; f: II --+ a /I by f(lt) = a h. This function maps
01110 a /I, sill('p ('vl'ry I'll'mcnt of a II is of t.he form a h for some choi(~('
of h E /I. III addition f is a one-to-one funetion, for if a. hi = a. h2' where
hi, h2 E /I, th(~ "arwella1ion law for groUJl!I yieldH hi = h 2 That is,j(hd = f(h 2 )
impJi('s hI = },2' If tIll' group (G, .) has n finit.e numlwr of c1emcnts, we may
t.herdor!' ('olwhlll<~ that. nny two Idt {~()Imt.s of /I have the Harne numher of
"'l'llll'lIls, nallll'ly, I.hl' IlIl1l1hl'r of 1'll'llll'lIl.s ill /I.
Example 2-35, Heturning Olll'e again to the group of symmetries of the
square, let us sl'icd. the subgroup (8, .), where S = {R360, V}. The tn.sk of
computing the I('ft eosets of 8 is .straightforward, sinee we have the operation
tahle for this group at. our disposal (HeC Example 2-24).
From n quick inspeetion, the render will observe that there are only four
distinct (~osets,
{Ilon, J)21, { /lillO, II}, f R270' J).}, and {fl36o, V} = S.
Thl'sl' '~OH('tH aI'<' disjoillt alld their union is the un<il'rlying set of elements of
t.he whole group. As WI' shull see, this is always the ease. Also, for this sub-
J,!;roup thl' not ions of IpH allli right ('OSl't~ do not IlgJ'l'I', since
since the set H, being the set of elements of a subgroup, is closed under the
gnlllp op{'\"IL\.ion *. The oPpoHite inclusion is ohtained by noting that ench
elcment It E If may be w\'itten as
h = a * (a-I * h).
Here, a-I * hEll, since both a, hEll and (II, *) is a subgroup of (0, *).
This implics that It E a * If and consequently II ~ a * H.
Our next theorem provides a simple criterion for the equality of two left
(:osets, whcn It rcpn~lICntativc of each is known.
Theorem 2-27. If (//, *) is a Huhgl'Uul' of the gnllll' (a, *), thCIl
a*H = b*H,
Remark. When working with right cOIICts, the requirement a-I - b E H must
be rcplaced by a- b- I E II; that is, Ii * a = Ii - b if and only if a * b- I E H.
As an imnwdiatc (~onsequence of the 1I11;t theorem, we tICC that any element al
of the ldt em;ct a *11 detcrmincs this eosct. For if al E a - II, then al = a * hI
for suitable hI E II. Thus a-I - al E /I, so that by th(! theorem, a * H = al - H.
This Illl'tUlH tim\. (,twh pJ('nwnt of 11 (:oset ean be thought of ns It represcntative
of that. ('oset. In IL (~ert.ain HI'IlHI! we ure heing prejutiic('(i whenever we denote
n ,'oSt'\, hy a * II, for SOI\l('()I\(' !'lSI! lIIi!!:ht dlool!(~ to "all it. ,) - II wlwre b ;t! a,
but a-I -" Ell.
78 G1WUP THf:ORY 2-5
or
RilleI' (II, *) is a slIhl(l'OlIP, t.lw pl'Odud hI * h'2l, alld t.I11IH a-I * b, 1II11tlt lie
ill the HI'f 1/. 0111' IIl'l'd ollly apply T(lI'orl'l1l 2 -27 to eOlwhul1'
a * 1/ = b * 1/.
We saw earlier thllt eneh clement a EGis a member of /:lome left CQI,lct of
H in 0, namely. the C'Ol'lPt a * H; that is,
is exhausted by its left ('OsetR.
Theorelll 2-2S illdil'llt.('l'I that nn clement can b!'lollg to one and only one left.
C'O!wt of 1/. Thlll'l t ht' ~('t a is partit i()J}('d hy 1/ into disjoint sets, each of whi('h
has eXlwtly as mllny elements as H. For case of future reference let us summarize
these rellJllrks ill tIll' followiul!; theorem.
Theorem 2-29. If (II, *) is n ~lIbjl;roup of the I/;l'Olip (0, *), the left (right)
('osetR of II ill (J forlll a partit.ioll of till' set a.
Example 2-36. ConsiJer (Z 12, + 12), the group of intcgefl:! moJulo 12. If we
take {O, 4, R} for the set H, then ({O, 4, S}, +12) is evidently a subgroup of
(ZI2, +12)' The left {,ORets of II in ZI2 arc
In this ('aRp, thp I'oset dl'('omposition of Z12 relative to the subset H is just
Z 12 = [0,4, Rl U {I, 5, 9} U {2, 6, to} U {3, 7, Il} .
RUJlpOHI' 1I0W Ihal (r:. *) is a lillit .. jl;rollP, Hay of order n, nllll (H, *) iR n
l'Il\hjl;roup of (0, *) of orll'l' k. WI' (:tI1 thell deeomposc the set (l iuto 1\ union
of Il liuitt, 1l1l1lllwr of diHjoillt left ('osets of 11:
Theorem 2:-31. If (U, *) is a finite group of composite order, then (G, *) has
nontrivial subgroups.
Proof. If the group (G, *) ill not cyclic, any element a E G with a ~ e generates
a nontrivial cyclic subgroup (a), *). Thus, it. sufficeH to consider cyclic groups
of compoHite order. To t.his end, l-iuppose G = (a) when' the g<'lIera\.or a has
order nm (n, 111 ~ J). Tll('n (a")m = e, while (an)m' ~ e for 0 < 171' < m.
From t.his, it. iH obviolll-i thut (an), *) is a lIontrivial cyclic l-illbgrollJl of (U, *)
with order m.
2 :J 4)
C 2 :~ C C 4 2 3 '
G G
2 3
4 3
G: 3
1 2 3 2 3 2 3
(
:3 4 G 2 4 G 1 2
C
2 3
C~ 3 (: 2 3
3 2
Since a is in the IcCt coset a * H, this means that a is also II. member of the
right ('Ol'll't If * II. The COK<'ts 11 * b I\nd 1/* a have t.he clement a in common;
so the Ilnalog of Theorem 2-28 for right cosets implies that
11* b = 1/* a.
2-5 NOItMAL KUI!G1W\1I'1'! ANI> QUOTJIo:NT GROUI'S 81
a. h = h' * a.
It would be gratifying to have a less cumbersome procedure than to compute
cosets for determining whether a given subgroup is in fact a nonnal subgroup.
Just such a criterion is given in the next theorem, and we shall have frequent
occasion to make use of it.
Theorem 2-32. The subgroup (II, *) is a normal subgroup of the group
(], *) if and only if for each clement a E G,
a.H. a-I ~H.
a. h = (a. h a-I) a = hI a.
The product hI a lies in the right coset II a, so we conclude that
h2 E /I such that
Consequently,
.- fl f2fa i4. is f6
fl It 12 f3 f4 hi f6
f2 f2 It f6 fs f4 fa
f3 fa f4 It f2 f6 f5
f. f4 f3 fs f612 It
f5 fro f6 f4 f3It 12
fa f6 fs !2 It fa 14
If we take as H t.he subset {fl.J .J6}, thcn it is easily verified that the pair
(H,o) is a normal suhgroup of (G, 0). The I:osct breakdown for the subgroup
in question is
fk 0 H = {!t.hf6} = H of,. for k = 1, 4, 6,
fk 0 II = {h,Ja,fs} = IIf,. for k = 2, 3, 5.
On the other hand, the subgroup ({It,!2},.) is not normal; a short com-
putation indi{'atcs why the criterion of Theorem 2-32 fails to be satisfied:
in a natural way with the original group. More specifically, we shall show that
the set of eoscts of a nonnal subgroup is itself the set of clements of a group.
If (H, *) is a lIornmll:!ubgroup of the group (G, *), then we shall denote the
collectioll of distinct (!oscts of H in G by GIH:
Gill = {a*ll/aE(l}.
Thesc are also right coscts, since the definition of a normal subgroup guarantees
that a * /I = /I * a for every a E a.
A rule of (!ompositioll may be defined on (JIll by the formula
a*H = al *H and
t.hen also
(a ~ b)-I * (al * bl )
is a member of H. Now, a * H = at * H and boll H = b l * H imply both
a-I * at. b- l * b1 E H. Silll.:e (H, oil) is normal in (a, oil), we know that
Theorem 2-33. If (11, *) iR a normal Rubgroup of the group (G, *), then
the system (a/II, ) formR a group, known as the quotient group of G by II.
Prool. First, lelt 11M ohHCrve t,hat, UIC! ulIIIO!lintivity or the Opl!ratioll is a direct
consequence of the associativity of * in G:
The coset If = e.11 is the identity clement for the operation , since
It is equally easy to sec thllt the inverse of the coset a * If is a-I * II, where
a-I denotE's the inverse of a in (0, *). This is evident from the computation
Hence all the group postula,t.es are fulfilled and the proof is complete.
Example 2-38. Once n.p;ain we fnll bl~ck on th(l group of symmetries of the.
square for an illustration. Here, the subgroup
(S, *) = ({Rilla, R 36o}, *)
is normal, being the center of the group. Its distinct cosets, that is, the elements
of G/S, are I
G/S = I {R I8o , R 36o }, {R go, Rna}, {V,N}, {Dh ~2}1.
A typical coset multiplieation proceeds as follows:
To lIluJt.iply two (:Ol'!ds under , nil we really need to do is select all arbitrary
rnprcflCmtnl,ive from ('/wlt 1'()l'!HI" Illultiply tlwlic c1emcnts under tho group
uJlI'/ILI.ioli ... ILlIII tIHt.l'rlllillC! to whith eOIi(I. the rtliIlIt.illg prothwt. bolollgll.
The operation table for the quotient group (G/S, ) il:l shown in Table 2-4.
Table 2-4
Example 2-39. A simple, hut. lIscful example to kl'l'p ill milld when working
wit.h IJllot.i(!lIt. grouJlIi iii furnislutJ hy till' atltJit.iv(! gJ't)up of integers (Z, +).
11. hu.s heell prl'viollsly ('stnhlislll'd that tlw (IIOl'mal) suhgroups of (Z, +) are
the eydie suhgroupli n), I), n a lIollllcgatiVl' illll'ger. The (~osets of (n) ill Z
t.ake t.he form
a + (n) = {a + kn IkE Z} = [a].
In other words, the cosets of (n) are merely the congruence classes modulo n.
Coset multiplication in Z/(n), moreover, is given by
(a + (n (b + (n = a + b + (n),
or, with a judicious change of notation,
[a] [b] = [a +- b] .
. Wc thus (!Iduce that. the quot.ient ~roup of Z by (n) is none ot.her than the
~roup of int,I'I!;('1'lI modulo n,
Am()n~ ot.hl'r t.hings, t.ltis illllieates that had we so desired, t.he study of the
integcr!:! modulo n eouid havc been subsumed under t.he more general theory of
quotient groups.
It is a simple matter to sec that any quotient group of a commutative group
is ne(:('sl'iarily eommutativc. A natural qucstion is whether a noncommutative
group ('1U1 poss('ss eommlltnt.iv(' qllot,i('nt grollps ILnd, morc pointedly, what
eondit.iow; (if allY at. all) would insure their exishmee. As a concluding topic
in this seetion, we investigate t.hil'i particular liituatioll. Our analysis begins
wit.h a basic definition.
Definition 2-26. <:ivcn a group (U,"') ILnd c1cments a, bE 0, the com-
lIlutator of a ILnd b is defined to be the product a * b'" a-I ... b- I .
86 GHOUl' TlU:OItY 2-5
a *b = [a, bl * b * a,
one may view the commutator of a and h as a measure of the extent to which
a * b differs from b * a. Indeed, the elements a and b commute if and only if
[a,b) = e.
In general, the commutators do not by themselves form the elements of a
subgroup, !linee thC'y fail to be closed under multiplieation. The mmal pro-
cedure for bypassing this difficulty is to work imltead with the suhgroup gener-
ated by all the commutators [a, h), a, bEG. The resulting subgroup is known
either as the dcril'cd s1iligroup or commutator subgroup of (G, *) and may be
denoted simply by (lG, G), *).
Now, the illver~c of a eommutator is again a commutator: [a, bl- I = [b, al.
There if! no nC'ccl'l'ity then of explicitly conKidering inverses in the definition of
the set [a, G]; ilK clements consist merely of products of finitely many com-
mutators of G. That is,
But, Theorem 2-26 tells us a necessary and sufficient condition for the last
equality to hold is that
la, bJ = a * b * a-I * b- I E H.
In other words, commutativity of the quotient group (GIH, ) is equivalent
to requiring that the subgroup (H, *) contain all the commutators of G. As
([0, GJ, *) is by definition the smallest subgroup with this property, the latter
eOlldition may be replaced by [G, GJ S;;; II.
A special case, but itself of interest, occurs on taking H = [G, 0):
Corollary. For any group (G, *), the commutator quotient group (Gil G, 0], )
is commutative.
The foregoing theorem says, in effect, that the commutator group is the
smallest (again, in the sense of inclusion) normal subgroup whose associated
quot.ient group is commutative. The transition from a group to its commutator.
quotient group is mfermd to II-S the abelizalion of the group and provides a
cOllvenient melUUI of mallufacturing commutative groupt! from noncommutative
ones.
PROBLEMS
J. If 11 = {O,6, 12, Ig}, show that (II, -t2.) is a cyclic subgroup of (Z24, +24).
Also, lillt t.he elementl! of eaeh coset of 1/ in Z24.
2. In the symmetric group (84,0), let the set 1/ consist of the four permutations
( 1 2 3 4),
I 2 3 4
(21 2I 34 4),
3
(1 2 3 4).
432 1
4. Det('rmin(' till' l('ft ('OH('t dc('omposition of ttl(' jl;roup of !:!ymmetriefl of the squarc
with rcsl'ed to the slih/l:roup ([ Raun, /)I!, *).
5. In the group of symmetries of the elillilateral trian/l:lc, fin(l:
a) all Rllhjl;roU(lR,
b) all normal IIU hgrou pR,
I') the "('1\ t('r of t.hl' /l:flllIl'.
6. Hhow that if the cyelie jl;roup a), *) is infinite, then a and a-I are its only gener-
ators, and all subgroups except ({e}, *) are infinite.
7. Lilt (II, *) he a Ruh/l:fOllP of index 2 in t1l1ll(rOllp (0, *). Prove that (II, *) i!i a
normal Sllhl(fOUp. [lfint: /I U (a * /I) = 0 = /I U (II * a) for any a E G - /1.1
~. Given t.hat (III, *) Imd (11 2 , *) af(~ hoth normal HlIhl(fOlIPH of the 1(ff1IlP (G, *),
prove t.hat, tlH' SlIhl(rflllp (1/1 n 112, *) is also normal.
9. Let (1/, *) be a Huhl!:rollp of t.he I(roup (G, *) and the set N(l/) be defined by
a) Prove that the pair (N(ll), *) is a subgroup of (G, *), called the normalizer
of IT in G.
b) Prove that (1/, *) ill normal if and only if NUl) - G.
10. Rllpposc that (II, *) an,1 (K, *) are normal Rllhl(roUpR of the group (G, *), with
/I n K = ':('1. By ('onsidl'rinl!: l'il'lIH'nt.s of thl' form" * k * II-I * k- I , !lhow that.
h * k = k * It for all It E /I, k E K.
II. Given (If, *) and (K, *) arc subgroupR of the group (G, *) and one of these sub-
groups is normal, prove that the pair (ll * K, *) is a subgroup of (G, *) i when both
are normal subgroups, show the group (11* K, *) is also normal.
12. Find an example of a I(rollp (G, *) having a subgroup (//, *) for which the produet
of two left CORllts of // in 0 need not be a left coset of ll. '
13. Describe the quotillnt group of
a) (Z.,+) in (Z,+), b) ({O,2, 4, 6, 8},+1O) in (ZIO,+lO),
c) (Z,rt-) in (Q,+), d) ({I, -I},) in ({I, -1, i, -i}, .).
14. Let (G, *) be a cyclic group with generator a and (II, *) be any subgroup of
(G, *). Prove that the quotient group (G/II, ) is also cyclic with the coset a * II
as a generator.
15. Given (II, *) is a normal subgroup of the group (G, *), prove that the quotient
group (G/II, ) is I'ommutative whenever (G, *) is commutative.
16. For any group (G, *), describe the quotient grouJlS of the trivial normal sub-
groups ({e},*) and (G,*).
17. Consider the eyclie I!;roup a), *) of or(ler 15 and the subgroup a 3 ), *). List
the elements of eaeh coset of (a 3 ) and construe!. the multiplicat.ion table for the
quotient grouJl a)/(a 3 ), ).
18. In the commutat,ive group (G, *), let the set II consist of all elements of G with
finite order. Prove that
a) (11, *) is a normal subgroup of (G, *), called the torsion 8ubgroup,
b) the quot.ient group (G/II, ) is torsion-free; that is, none of its elements
other than the identity are of finitl~ order.
2-{i 1I0MOMORl'HI8MS 8U
19. Rhow that a group (G, *) is commut.at.ive if ancl only if (G, GJ = {e}.
20. For any group (a, *), h-t (II, *) he the l"Iuhgroupgcnerated hy the Kct of Ilquares
of clmnenl.!I of a. Jo:stnhlish t.he following:
a) (II, *) is a normalllllhgfoup of (G, *).
h) The qllotil'nt p;rollp (G/II,0) is commutative. (llint: [G, GJ ~ II, since
[a, II) = (a * II)".! * (/}-I * a-I * /})2 * /}-2.)
21. u-t (IIi. *) he a ('olll-I,tion of nontrivial norllIal ~ubgroup>! of the group (G. *) such
that G = U/h Assume further that II, n IIi = {e} for i ;>C j, Prove that the
parl-nt group (G, *) ill n"""I'<lInrily I'ornlllutative, [/lint: Let a E II. and bElli' For
i ;>C j, w.,e Prohlem 10. In (~a.~e i = j. choose any clement, c E G - II;. Then
c aud c * a "OIlIllIUIl- wit,h IWI~ry ('lllment of IIi; in partkular, e = [e * a. b * al =-
[a, bl.]
22. Prove that if th!' qUoti"nt group (G/cent G, 0) is I~y('lil'. then (G, *) ill a I:om-
mutative group,
2-6 HOMOMORPHISMS
Up t.o thiH point, in the t(~xt, w(~ hll.vI~ not eonHidcred ml\(lpinll;f1 from one group
to Ilnothel'; in(l<~('d. any knowledge of funetions was irr<,l(lvallt to most of
the t.opi(s (~onsid('r('(1. TllI'Y now ent.er in an eHslmt,illl way. for we wish to
inl.rodu<:e It ('0111'<'(11. IIII' idm of algl'iH'uienlly indistinguishable sYHI,ems-whieh
will be of fundamental importance throughout thc remainder of the book and
whieh ill, in fat't. one of t.he most important notions in mathemlltics,
This section begilll'l, however, with an analysis of a class of functions which
preserve alj!;ehmic Rtrueture.
Deftnition 2-27. Let (G, *) and (G', .) be two groups and f a function from
G into a'. f: a -+ G', Tlwn J is said to be a homomorphism (or operation-
preserving fundion) from (G, *) into (G', .) if and only if
f(a * b) = f(a) 0 feb)
for every pair of clements a. bEG.
A few remarks are in order before considering any examples. First, notice
that on the left.-hand side of the above equation, the product a * b is computed
in G, while on the right side the product f(a) feb) is that of elements of G'.0
thnt till' dinJ.(mlll of mappings on page 8!J is ('ollllllutal.iv(', For this eCllulit.ioll
aHsertl'! thaI. if WI' I'!tal't with dl'nwlltH a, b of a and move them to 0' by eit.IH'1'
of the two rouh'H indieateu by the arrowli -hy fil'lit forming the product a * /,
and t.lWII applying f t.o it or hy firlit. oht.aining UII' imagl'li f(a) and f(b) and
then taking I heir produl'l. --the rCHult will be the Hllme,
It may niSI) st.rike t.he reauer that the langlmge ill whil'h Definition 2-2i i"
coudwd is 01)('11 to I~ritieilim. To speak of IL hOlllolllorphiHm from a group (G, *)
illt.o It group' (0', .) iii S<lmewhll.t imrlrl~ciSl', Hinee the lIIapping cOlleerned is
aetually IlI'l.wl'l'lI t.h('ir underlying 1i('ls of ('1('n1('nls. ThiH linguiHt.ic eonvelltioll
haH til(' d(~eidl'd advanlage, however, of illllicatillg Ihe operations preserved as
well ali the domaill and rangl' of the funct.ion.
Example 2-40. FOI' lin arhitmry group (G, *), d('filll' the funetion f: G -+ (/
hy taking f = io , the ident.it.y map on G. It is It trivilLlit.y to check that f is
!l homomorphililll from t.he group (G, *) into itself, a8
For fulure liSP, W(' shall label the set of all hOlllolllorphiHtns from th(, group
(0, *) inlo ilsdf (thp so-calh'd ('1u/olllurphis1ns) hy t.he Hymbol hom Gj a fr('-
!(llI'ntly IIs!'ll altl'rnal iv(' notal ion iH to write end G. Both notations hllv(' a
el'rlain HUJl;gestive pow('r, and it rctIuces to 11 matter of pel'l:lOnal preference.
2-6 IIOMOMOltPlIlHMS 91
IlIterCl:it ill)!;ly ('lIoll/!:h, the set hom a cnll be enuowl,d with an nlgebraic
stmdur!' :
Theorem 2-36. Tilt' pail' (hOIlI 0, 0), wh('l"(' 0 dl'lIot.ml flllwtionnl compOl~i
tiOIl, forms a semigroup wit.h idcnt.ity.
PJ"()of. For' III!' proof, whidl is quite dCllwnt.nry, it, IIlUHt fif/lt. he Hhown that
t.h(l ('olllposil ion f (I of two flllwLionH f, (I E hom G Itl(uin pr('HerVCS tho group
operation. ThiH is pltsily Itcl'ompli!lhed by noting that whenever a, bEG,
(f 0 (f)(a * b) = f(y(a * 1))
= f(r/(a) * f/(/I)
= f(a(a) * f(I/(II) = (f 0 g)(a) * (f 0 g)(b).
aH does the ident.ity map i a. It remainH ollly to verify here that whenever a
funetion f E A (G), it.s inver!le f- I (whieh clearly is a member of sym G) is a
homomorphi!llll. If if, 1) E G, the onto dmmeter of f implies if = f(a), lj = feb)
for some choice of a, b in G. Therefore,
f- I (if * b) = f-I (f(a) * f(ll
= f-I(J(a * b)
= a*b = f-I(if) * f-I(b),
and t II(' proof is ('olllplete.
There are many important and inlN('sl illg fad.s ('olwerning homomorphic
mappings. I n I h(' :-;u('(~eedin)!; theorellls, we shall examine some of tJwse results
in detail.
92 G1WUP THEOHY 2-6
In ttw second part of the t.1worem, it ill first necessary to show that
We can then conclude from the uniqueness of the inverse of f(a) in (G' , .) that
f(a)-I= j(a- I ). To obtain this result, we make use of part (1) t.o get
f(a) 0 f(a- 1 ) = f(a * a-I) = f(e) = e'
Similarly,
j(a- I ) f(a) = e/.
The erlJ(~inl st.ep, f(-n) =-: (I( -n), is justified by til(! fact f lmd g arc already
known t.o agree on the positive intl'ge,'s, By the firnt. plLrt of Theorem 2-38,
f(O) = 1 = y(O), so t.hat fen) = yen) for every integer n; therefore, f = y,
The next result indicates the algebraic nature of direct and inverse images
of subgroups under homomorphisms, Among other things, we shall see that if
f is a homomorphism from the group (G, *) into the group (G', .), then (f(G), 0)
forms a subgroup of (G', 0), The complete story is told below:
Theorem 2-39. Let f be a homomorphism from the group (G, *) into the
group (G', 0), Then
1) for each subgroup (H, *) of (G, *), the pllir (j(II),.) is a subgroup of
(0',0),
2) for each subgroup (H',o) of (G', 0), the pair (I-I(H'),.) is a subgroup
of (G, *),
Proof. To obtain the first part of the theorem, recall the definition of the
image set f(H) :
f(H) = {f(lt) I h e H},
Now, suppose f(h) and f(k) are arbitrary clements of f(H) , Then both hand k
belong to the set H, as does the product h * k- 1 , Hence,
Our argument shows that whellever f(h),f(k) ef(H), thenf(h) of(k)-l lies in
feB); this is a sufficient condition for (j(H), 0) to be a subgroup of (G', 0).
The proof of the second statement proceeds in a similar manner. First,
rememher thut
j-I(H') = {aeGlf(a)eH'},
Thus, if a, b ef-'(H'), t.he images f(a] and feb) must be elements of H', It
follows at once that
for all a' E G' and It E H. In general, there is no way of replacing the element
a' by someJ(a) ill order to (!xploit the normality of (H, *). A slight strengthen-
ing of the hypot.hellis overcomes this difficulty; simply tak!l f to be a.n onto
ma.pping. (Rc<:all that the word "onto" r<-'quircs every member of (J' to be the
image of at. I<'l\~t one clement of G.)
Summarizing tlH'sC r('marks, we mny now state:
Corollary. I) For !lll.dl normal lmhgroup (H',o) of (G',o), the subgroup
(f-I(II'), *) is norlllal in (0, *).
2) If f(G) --=- a', Uwn for ca<:h normal Imbgroup (ll, *) of(G, *), the HUh-
group (j(l/), 0) it! normal in (a', 0).
In much of our subsequent work, the object of interest will be the kernel
of a homomorphism.
Definition 2-28. Let J be a homomorphism from the group (G, *) into the.
group (G', 0) and let e' be the idenhity element of (G', 0). The kernel of f,
denot,ed by ker (f), is the set
Thus ker (f) consists of those clements in G which are mapped by f onto
the identity eicment of the group (G', 0). Theorem 2-38 indicates that ker (f)
Is a nonempty subset of a, since e E ker (f). It may well happen, as Example
2-41 shows, that kl'r (f) = G. Except for the trivial function indicated there,
the kernel ill IlIW:1YII a proper subset of G.
Our definition of It homomorphism did 1I0t require that it be a one-to-one
function, and iIHl('cd, we hllve preRCntcd several examples where it failed to be
so. There ill, howcV!'r, 11 simple charudcrization of a on<-'-to-one homomorphic
mapping in tenns of the kenlel.
2-(i HOMOMOUPHISMS 95
Theorem 2-40. l.,(~t f he a homomorphi8m from the group (G, *) into the
group (G', .). ThenJ is one-to-one if and only if ker (f) = {e}.
Proof. SuppoS(! the function f is one-to-one. We already know that e E ker (/).
Our aim is to show that this is the only element in the kernel. If there existed
another clement a E ker (f), a e, then we would havef(a) = e' = I(e). That
is, f(a) = fee) but a e. This would contradict the hypothesis that I is one-
to-Gne.
On the oUler hand, !lUppOH(! thaI, ker (f) = {e}. Let a, bEG and f(a) = feb).
To prove f is one-to-one, we must 8how t1mt a = b. But if f(a) = feb), then
f(a * b- I ) = lea) of(b-I) = f(a) of(b)-I
~-= lea) f(a) - I '" e',
0
Theorem 2-42. Let (H, *) be a normal subgroup of the group (G, *). Then
the mapping natH: G -+ G/ H defined by
is II. homomorphism from (0, *) onto the quotient group (G/H, ); the
kernel of natll is prc("iscly the set H.
Proof. Tlw fal"!, t.hat. t.he mapping natH is homolllorphie follows directly from
till' mlUlIIl'r in whi!"h 1I\11lt,iplic~llt.iClII i>l clC'linml ill thc~ ,,1IC1t,icmt KrOllI':
To show thnt, nat/( is an onto funet ion is almost trivial, sinee every element
of G/If is Il c'ost't a * H wh('\'( a E G find Ilntll(a) = a * H.
IlIIuml\lch lUI the ("0>1('1, /I serV!'s IlS the identity clement for (O/H, ), we
must have
ker (natll) = {a E G I natH(a) = H}
{a E G I a * /I = /I} = II.
Definition 2-29. Two groups (G, *) and (G',.) are said to be i8omorphic,
denoted (G, *) ~ (G', .), if there exi~ts a one-to-one homomorphism f
of (G, .) onto (G', .), that is, f (G) = G'. Such a homomorphism f is called
lUI i8()mOTp"i.~m, or isomorphic mapping, of (G, *) onto (G', .).
and i 2 = -1. The operation tables for these two systems arc
+. 0 I 2 3 1 -1 -1
0 0 2 3 1 1 -1 i -i
1 1 2 3 0 -1 -1 1 -1
2 2 :3 0 1 i -i -1 1
3 :3 0 2 -I -~ i 1 -1
Wc wish to provc that the groups (Z., +.) and (G, .) are abstractly "equal." To
do so, we must produce a one-to-one homomorphism f from Z. onto G.
Since the preservat.ion of identity clcments is !\ general feature of any homo-
morphism, f llIust be such that f(O) = 1. Let us suppose for the moment that
we were to define f( 1) = -1. The image of an inverse clement must equal
98 GROUP THEORY 2-6
Clearly this fundion is 1\ ono-t.()-()no mappinll; of tho HCt Z. onto the set G.
Furthermore, 1 uetuully pn~HCrve8 tho operations of the groups. Merely to
verify one instanco, we ObH(~rv(l that
Apart from the particular symbols used, this group table is identical to that of
(Z .. , +.. ), for corr<'sponding elements appear at the same place in each table.
Both groups are simply disguises for the same abstract system.
In passing, we might note that (Z .. , +.) is also isomorphic to (G,.) under
the function g, whereby g(O) = 1, g(l) = -i, g(2) = -1, g(3) = i.
Example 2-47. Let G = {e, a, b, c} and the operation * e a b c
be defined by the table at the right. The reader may verify
*
that the pair (G, *) is IL group, known as Klein's four- e e a b c
group. The two groups (Z., +.. )
and (0, *) are not a a e c b
abstractly equal, however, for every one-to-one function b b c e a
I from the S('t. Z. onto G fails to be operation-preserving. c c b a e
2-6 HOMOMORPHISMS 99
From t.his we conclude t.hat. there arl' at least. two distinct algebraic structures
for groups with fOllr clements.
To illustrate this point, we shall check several possibilities for the function f.
Consider the mapping defined on the set Z4 by
for all a, bE Z. If x donotes the clement of Z such that f(x) = -1, then
According to Theorem 2-40, the identity element of (Z, +) is the unique clement
of Z corresponding t.o the identity of (Q - [O},), so that 2x = 0 or x = O.
Consequently, both f(O) = 1, and f(O) = -1, contradicting the fact that the
functionf is one-to-one. This argument shows that (Z, +) cannot be isomorphic
to (Q - {O},), for no fUlH'tion satisfying Definition 2-29 can exist.
Example 2-49. For an instructive example in connection with the additive
group of integers ('ollsidcr the following ll..'!SCrtio:l: the only functions under
which (Z, +) is isomorphie to it.'lClf are the idcntit.y mapping and its negative.
A fairly suceiru:t dl'seription of all this is that A(Z) = {iz, -iz}.
100 Gnoup THEOny 2-6
Perhaps the quiekest proof of til(! above WI.'!crtion consists of ahowing that if
J E A (Z), then the eY('lie subgroup (j(l)), +) generated by J(I) is the group
(Z, +) itself. Since the illdusion (j(l ~ Z trivially holdR, our aim would bc
achieved by establishing Z ~ (/(1. But this is a l:!traightforward matter. If
n is an arbitrary integer, n = J(m) for some mE Z-recall J is a. ma.pping
onto Z-so that
n = J(1 + + ... +
1 1) (m summands)
= J(I) + J(I) + ... + J(I)
= mJ(I),
whence n E (J(l.
Since land -1 are the only generators of (Z, +), either J(I) = lor J(I) = -1.
However, the preceding computation indicates fen) = nf(l) for each n E Z;
fromthis,itisclearthatJ= izorJ= -iz according asf(I) = 10rJ(I) =-1.
Let us return to general considerations by showing that the groups (Z", +,,)
and (Z, +) are the prototypes of all finite and infinite cyclic groups, respec-
tively.
Theorem 2-44. Every finite cyclic group of order n is isomorphic to
(Z", +,,) and every infinite cyclic group is isomorphic to (Z, +).
Proof. First, suppose the cyclic group a),.) is of finite ordeF- n. In this
case, we know from Theorem 2-23 that
It scems nutural tht'll t,o illvel:!tigate thc mapping J: (a) --. Z,. givcn by thc
rule f(ak) = [k], 0 ::; k < n. This function plainly carries (a) onto the sct Z,..
Next observe that f is Ollc-to-one: if J(a k) = f(a i ), then k == j (mod n) so that
a k = a i . Finally, for any clemcnts a\ a i in (a), we have
This shows t.hc function J prescrves the respective group operations and eom-
plctes the proof of the iRomorphism C(a), .) ~ (Z,., +.. ).
For t.he seeond purt of the theorem, thc cyelic group C(a),.) is assumed to
he of infinite order. HI're, the choice of a mapping I betwccn (a) and Z is
obvious: simply take I(a fr ) = k. It is immediate that J, so defined, is an onto
mapping. Further, this function is one-to-one, for all the powers of the generator
must be distinct; if two different powers of a were equal, the argument of
Theorem 2-23 could be employed to obtain the contradiction that (a) is a
finite set. The rest is routine:
Corollary. Any t,wo "Yl'lic grouP!! of the same order arc isomorphic.
Trivially, llny group (0, *) is isomorphi(~ to itself under the identity mapping
i(}. A reasonlLhl" query is whether (0, *) is isomorphic to any group other than
itself. The concluding th(!orem in this HCetion, a classical result due to Cayley,
answers the question in the affirmative.
We begin, however, by reenlling HOIlt(! definitions and notation. For an
arbitrary clement a in 0, the left-multiplication function fa: G -+ G was defined
by taking fa (x) = a * x for every x E G. The collection of all functions obtained
ill this way is labeled by Fo: Fa = {fa I a E G}. Example 2-25 established
the stnlctural nature of the pair (Fa, )-in the present context indicates
0 0
a = a *e = b *e = b,
f .. (x) = a + x, x E R'.
That iR, the function fa merely has the effect of translating or shifting elements
by lln llmount a. Cayley's Theorem MHerts that the group (R', +) and the
I/:rollP (Fa, .) of transl:tt.iollR of t.he real line nre indiRtinguishablc a8 far as their
nlgl'hmie propl'rt.i(s an! (!OIw(!rlled.
PROBLEMS
II = {a E G I/(a) = a).
Prove that (ll,.) is a subgroup of (G, .).
6. Let (G,.) he 11 gruul' and tho (,)omont a E a be (ixlle\. Provo that (G,.) ill iso-
morphic to itllclf--that is, (0, .) ~ (G, . ) - under the mapping I defined by
xEG.
II. If
Rhow that. if t.l1I1 group (/',0) is ,'ommutative, then h ill also a homomorphism.
13. COIl~idcr the following two groups: (GI, *), where
various tl\('or('m~, WI' shall frequently 1I0t troll!>l!! to spedfy this familiar open-
ing phm&. Ae('ol'dingly, unl('S!; Uwre is ('I('ar indicat.ion to the eontmry, any
rIf(~rel\(e to til!' fUIH'tionj i~ undel'Mtood implieitly t.o involve the aforementioned
hypoth('~ill.
W(~ h(gin wil II It proof of the Fnctor Theorem, stated here in a form beij!.
Huitl'd to 0111' immedinte needs.
Theorem 2-46. (Far/or Theorem). LC't (//,.) be a normal subgroup of
the group (0, .) l>uch that H ~ ker (f). Then there exiijts a unique homo-
morphism7: Gill -> G' with the property
1= 7 nntl/.
Prool. Before becoming explicit in the details of the proof, let us remind the
r('ad('r thnt the symbol nat/l designates the natural mapping of G onto GIH;
that is, natu: G -+ Gill with natH(a) = a. II.
To start. with, we ddine a function 1: GIH -+ G', called the induced mapping,
by taking
l(a. II) = I(a), aE G.
for nil a ill fl, Ji()] 1/. '1'111' irllitweci nlllppillg ] iK tllIr..ror() 1i('(~11 to hI' thl! Ollly
functioll from a/ /I to (J' lSutilSfying the equation f = ] natH' 0
where, of CO\lI'8(', e' denotes the identity of the group (G' , 0). Another way of
saying the same thing is
natll (ker U = H,
. H.
whieh is equivalent to the inclusion ker (f) S;
often described by saying that the function f can be factored through the
quotient, group (G/H, ) or, alternatively, that f can he factored by natH.
The followillg dillgmm muy help to elurify the relations umong the various
function!! :
G/H
What. we have just proved, ill effect, is that there' ('xi!lts one and only one
function 7 whieh makcs thilS triangle of maps commutative.
Example 2-50. Lct us eit.(~ a SI)(!cific instance of the Factor Theorem. Our
content.ion is that whenever the group (G', 0) is commutative, the function f
can nhvays he fadored t.hrough the commutator quotient group (G/[G, G), );
ot.herwise stated,
f = 7 natlo,o).
lOG (;)(WP TIIEOHY 2-i
For this, f'ollsidl'r allY ('Ollllllutator fa, 11J = a * b * a-I * /,-1, with a, II E G.
As a ('OIlS('q\l('IH'1' of till' ('OIltllllltativity of (0', 0),
so that [a, b] E k('r (f). 1'\ow, the ('omnlllt.at.or subgroup (0, G], *) is, so to
speak, the smallest subgroup of (G, *) to contain :\11 the commutators. Thus
we must ('OIl1'lude that [0, OJ C;;; ker (f). Having t.his inclusion, it is only neces-
sary to apply Theor!'1ll 2-1(i to obtain the dl'sired factorization.
Although WI' did Ilot I'xprl'ssly IWluire the information in {hp proof of Theorem
2-W, it Illight II(' poilltl'd out thaI til(' ilulw'l'd nlappinll:7 ('lllTil'M (;/11 ont~) till'
set (I'. 'fhi,.; is I'mdily ohtainable 1'1'0111 til(' defillition of 7 and t.hc fact f itself
is un onto fundion. To hp more Sl)('('ific, if a' E (;', thcn a' = f(a) for some
c1ellll'llt a in 0; h('III'I',
a' = f(a) = 7(a * II).
The FlIIuianunl.al '1'1\('01'('111 l.ill'n I!;lIarnntel~s thnt ( (Z., Zo}, ) and ( {I, -I}, ,)
Itrl! il'lIIll1rphi(! Jl:rllllps illlll'l'ci, t.hiH is flLirly I'viill'lIt fnull 1m inspection of
their multipliclLtioll tables;
1 -1
Z. Z. Zo 1 1-1
Zo Zo Z. -1 -1 1
Further, the proof of the theorem indicates that the function which actually 7
establil:lhm; thil:! iHOmorphil:lm (the induced mappinp;) is given by
a), *) ~ (Z .. , -III)' Tlwsc fads arc already familiar, of (~ourse, but thc
argument involved furnishes an alternative approach to Theorem 2-44.
The next theorem not only provides further evidence of the power of the
Fundamental Theorem, but is of independent interest sinee it gives additional
insight into the stmeturc of the quotient group (G/cent G, ). To prepare
the wny, it is 1\(!(~(,SS:U'y t,o <iigr('sl:! for Il. moment.
<:iv<!n It fixed d('nll'lIt. II of the group (a, *), define t.he nlllJlping CT tI : (J -+ a
by taking
for every x in G. I.('t, us obtain a few of the special properties of this (unction.
Finlt, CT" turns out to he a homomorphiHm: if XI, X2 E 0, then
CT" E A(G).
Funetiow; of t.lw form CT tI, with a E G, are usually (:aIlcd inner automorphism8
of the groUJl (1, *); to he more preeise, CT a is the inner automorphism induced
by the elelllent a. For hrevity, we label the set of fum:tions arising in this
way by I(G):
I(a) = fCT" I a E G}.
In the ease of a commutative group, I(G) reduces to just the identity mapping
io = CT.. Thus, it is only when (G, *) is noneommutative that the notion
becomes meaningful.
Lemma. The pair (I(G),.) constitutes a group, known as the group of
inner aulomo/'phisms of (G, *); in fad., (I(G),.) is a normal subgroup of
(A(G), .).
Proof. The proof that, (I(G),.) is a group presents no difficulties; it consists
of nothinl!: more than noticing that whellever CT a, CTb E I(G),
In addition to disposinJ!; of the closure condition, thili relation also indicates CT.
is the iden! ity elemellt for the system and CT;I = CTa-l.
2-7 THE FUNDAMENTAL THEOREMS 109
The reasons for earh of these steps are reasonably KClf-cvident, and the reader
should mnke Hure he ullderstlUuls them. What is Hignifimmt is that
f 0 U" 0 I-I = (l'/(n) E leG),
as was to be proved.
We are now in a position to obtain the result which has been our goal. Namely,
that the groUpH (G/I!ent G, ) and (I(G), .) refICmblll Clwh other in all essential
aspects.
Theorem 2-41. For each group (G, *), (G/cent G, ) ~ (I(G), .).
Proof. To begin, consider the mapping 0: G -+ /(G) whereby g(a) = u o That
this function maps G onto the set /(G) is quite plain. It is equally easy to see
that 9 is a homomorphism, since
.
g(a b) = (l'a.b = (l'a 0 Ub = yea) g(b), a,bE G.
The crudul IUipect of the proof is now to IdlOW that ker (g) = cent Gj once
this is done, the desired conclusion will follow immediately from the Funda-
mental Theorem.
III f,he clll;e at hand, t.he inlier automorphiHm U c serves as the identity clement
for the group (I(G), .). A('cordingly, the kernel of the function 9 is defined by
RI'ferring t.o the ddinition of p<)lInJif,y of funetions, we eonclu<ie that a E ker (g)
if and only if
a. x. a-I = x
from ",hii'll Wl' infpr Ihat. all thc inclusions are actuuIly equalities. In essence,
we llre observing that diRt.inet RubRet.s of G may have the same image set in G'.
TIl(' diffieult.y in thc last paragraph could be remedied by either requiring
that kcr (I) = {e} or clse by narrowing our view to considcr only subgroups
(II, *) with ker (f) ~ II. In eith('r event., it followH I,hat
Lemma. If II is llny suhset of G such that ker (I) ~ II, then II = f-I (f(II).
Proof. Suppose the cl!'ll1!'nt. a is ill f-'(j(II), so that f(a) Ef(II). Then
f(a) = f(h) for HlHne choi('(~ of II. E 11. As the equation f(a) = f(lt) is equivalent
to f(a * It-I) = e', we have a * It-I E ker (f) ~ II. This implies a also belongs
to the set II ulld yield:-; the in('lu:-;ion f-I (f(II) ~ II. The opposite illdusion'
always holds (Theorem 1-7), whenl'e II = f-I(j(II.
Tim n'latinllHhip "I'IW('('II t.11I~ HII"I(I'OIl(l1l of (0, *) lun! Un~ suhgrou(l of (0', 0)
may he stall1I "" follow ... :
Proof. Let us first dJ('('k t hal the indj(at.ed correspondencc is onto. In other
words, if (II', .,) is any :-<lIbgroup of (U', .), we must. produce some subgroup
(II, *) of (a, *) with ker (f) ~ /I for whieh f(lI) = 11'. To !l.(!(omplish thill, it
is sufficient to take II = f- I (II'). lly Theorem :!-;m, the pair (j-I (II'), *) is
2-7 III
a subgroup of (a, Oo) and, sincc e' Ell',
l\Ioreover, t.hl' fund-ioll f hcing Itll OlltO IIlltppillg, I,he corollary to Thcorem 1-6
indicatcs thatf(j-l(II') = Il'.
Next, we vcrify that this corrcspondence is also one-to-one. To this end,
supposc (Ill, Oo) and (H2' *) are both subgroups of (G, Oo) with ker (I) ;;; H.,
kcr (I) ;;; H 2 and such that f(1I 1 ) = f(1I 2 ). According to the preceding lemma,
we then must. have
Tlw t.Il1'orl'lll llpplil!s, ill pllrt.il!ular, to I hp Ias(' ill whidl we st.art with It
normal :;uhgrollp (//, *) of (a, *) alii I lake f \'0 be the lIatural mapping
natu: G -> a/II of U onto a/ Il. Hinee ker (Ilntu) = Il, the conclusion is
modified slightly.
Corollary. Let (II, *) be a normal subgroup of the group (G, Oo). There
is a onc-to-one eorrC'spondenee b('t\V('CIl those subgJ"Oups (K, *) of (G, Oo)
Hu(~h that. II ~ K Ilnd thc set of all Hubgroups of the quotimlt group (a/II, ).
propl'r suhgroups. Alt.hough t.\lI'rl' is nothing vt'ry HllrpriHing about LhiH asser-
tion, our aim is to demonstrate it.s validity hy IIHing the corollary to the
Corresponuenec Theorem.
First observe that, sincc the groups (a, *) and (Zn, +n) are isomorphic, there
is np losH in generalit.y in working with (Zn, +n). Furthermore, as we have
poirill'd out 011 :several occasions, (Zn. -In) = (Z/(n), ). By the corollary,
we learn there is a one-to-one eorrespondelwe between those subgroups of the
group (Z, I) whi("h eontain the spt (n) and IIII' subgroups of (Zn, -hi).
Bul., the suhgroups of (Z, +) are jm;t the ('ydie subgroups m), +), where
m is It lIonnl'galive intpger. Comhining these rpsult,s, we arrive at the COIl-
c1u:sion then' is a one-to-one eorrcsponucnce bctween the subgroups of (Zn' +n)
112 GROUP THEORY 2-7
and those subgroups m), +) of (Z, +) such that (m) 2 (n). This last in-
clusion occurs if and only if m divides n.
Tlw two ':olwluiling tlwOr'!IllH of thiH Kedion arc KOmewhat dl.'cper results
than U811111 IUIII rl''1l1ire the full forI".' of ollr necllmlllutcd machinery; thl!y
eOlllpriSl.' what un' oftl'n <:ulled the FirKt allli Second lliomorphilim Theorelllli
for Groups and are of great importanee in the study of group structure.
Theorem 2-50. If (II, *) is allY normal subgroup of the group (G, *) such
that ker (I) ~ II, then (G/II, ) ~ (G'/f(ll), /).
Proof. As a prefatory remark, we might point out that the corollary to
Theorem 2-30 implil's the pair (J(II), 0) is a normal subgroup of (G',o), so
thnt it. is I'l'rt.llinly J1l'rmiHMihlc to form f he quotient group (0' /f(II) , ').
Let us now define thl' function 7: (; --> 0' /f(J1) by
1= nat/Ol) f,0
where natf(ll,: 0 --> O'/f(lI) deRiglllLtes the naturnl mapping. The following
diagram helps to visualize the situation:
G'/I{H)
Observe that 1 nwr(ly ILssigns to (ach element a EO the coSet f(a) of(H) of
0' /f(II). Since both the functions f and nat/(II) are onto and operation-pre-
serving, the composition of these gives us a homomorphic mapping from the
group (G, *) onto the group (G'/f(ll), ').
The main line of the argument is to show that ker 0) = H, for then the
desired reRu!t. would be a simple eonl!l'quence of the Fundamental Theorem.
Now till! ililmtity dlmll'nt of (a' /f(lI) , ') iM jUHt the (:OKl.'t f(lI) = e' f(II).
This means the kernel of 1 ('onsists of those members of G which are mapped
by ] onto f(lI); that is,
As we Ilrt givI'n f.Imt, kl'f (f) ~ II, the lelllllll\ prm'eding Theorem 2-49 may b('
invoked t.o (,OIl1'lud(' II = f-I(J(ll. Hence, k('r (]) = H, which completes
the proof.
In applicat.ions of this theorem, we frmluently Htart with an arbitrary noml1l1
subgroup of (a', .) lind utilize inVl'rsc images rather than direct images.
2-7 1I:J
Corollary. If (II',.) is any normal suhgroup of the group (G', .), then
(G//- 1 (1I'), ) ~ (G'/H', ').
I'mo/. By the ('llrollul'y to Tlwllrmll 2--:m, nUl pair (f-I(lI'), *) iK I~ Ilornml
f:!lIhl(rIllip of (a, *). :'Itofl'llVl'r, kl'r (j) r;;/-I(II'), SIl 1.111' hYIlOUlC'siK of th(,
UIC'()J'('III is ('()/IIJ1h,tdy sut isfil'd. This leads to till' iSOIllOl'phiKIIl
Since / is a mnpping onto a', 1/' = /(j-I (1/', and we arc done.
Our final theorem, a rather technical result, will be crucial to the proof of
the ,Jor(lnll-Htildl'r Theorem.
Theorem 2-51. If (//, *) lUlU (K, *) an! KuhgroliPH of 1.1)(' KrouP (a, *) with
(K, *) llorllllLl,l.hen (II/II n K, ) ~ (11* K/K, ').
Proof. NeedlesH to Hay, it should be che(,ked Uu~t the quotient groupII appearing
in the stat.emf'llt of the theorem nrc actually definc'd. We leave to t.he reuder
the routine t.usk of verifying thltt (11 n K, *) ill U Ilormal Huugroup of (II, *),
I.hnt.-(II * K, *) is 1\ group, lUui that (K, *) h~ llorlllal ill (1/ * K, *).
Our proof is putt.erneu on t.hat of Theorem 2-flO. H(!I't~, the problem is
to construct a homomorphism (I from the group (Il, *) ont.o t.he quotient group
(/1* K/K, ') for whieh ker (I) = II n K, To n(,hil've this, cOllsili('r t.he
function (I(h) = It * K, It E /I. Note, II = II * e ~ /I * K, so that a can be
obtained by composing the inclusion map ill: /I .-. II * K with the natural
mapping natK: H * K .-./1 * ~/K. In other words,
H oK/K
Example 2-54. As an illustration of this last result, let us return again to the
group of integers (Z, +) anrl consider the cyclic subgroups 3), +) and
4), +). Both these Imbgroups are normal, since (Z, +) is a commutative
group. Morc!Ovcr, it is fllirly obviolls tlmt
PROBLEMS
1. Let (G, *) he the group of symmetries of the square and (G', oLbe the Klein
four-grouJl (see Examples 2-24 and 2-47). The following mapping defines a
homomorphism from (G, *) onto (G', 0):
G------------~--------~ ~
[Uint:
I,(a)
{St
E (/" !ldint'] fly lU,(a = h(a).1
/
G,
""arly the same argument a.~ for Theorem 2-46; that is, for any element
2-7 115
(I 2 4),
123 4
!l
( I 2 3 4),
2 I 4 3
(I 2 3 4)
3 4 1 2 '
(I 2 3 4).
432 1
20. Illustrate the result of Problem 19 wit.h (Zu, -h), the group of integers modulo 6,
and the two subgroups ({O, 3}, +6) and ({O, 2, 4}, +6).
21. Prove that the Klein four-grouJl is isomorphic to (Zz X Z2, .).
22. Hhow that if 711 and n are relatively prime, then t.he direct product (Z .. X Z., .)
is a cyclic group of order mn.
AM the !.it.!!! ill(lil~atn!!, the main PIlIlIOH(~ of thi!! brief Hedion ill to estn.hlish the
Higllifil'l1l1t. I1nt! hil>wrie .Jordan-Holder Theorem. For the sake of simplicity,
we Hhall lilllit Olll'lielV!'H tll 1,111' I'II..'~ of finite gJ'()\IJlH; 1.111' inl.l~J'(~Ht.ed fI~n.der ill
referred to n lIIore gmwrtll lreutnwllt in [WJ. BI!('nuSt~ the theormn is ruthel'
involved, it will be eOllvellicllt. to begin by introdueing t!Ollle special terminology.
Definition 2-30. By n. dlain for a group (G, *) il> 1II('l\nt any finite sequence
of SUhsetH of a,
(proper inclusions),
descending from G to {e} with the property that all the pairs (Hi, *) are
subgroups of (0, *). The integer n is called the length of the chain.
What we are really interested in, and henceforth IIhall confine our attention
to, are the so-called normal chains. These are chains in which each group
(Hi, *) is a normal subgroup of its immediate predecessor (Hi_\, *). For grouP!!
(a, *) with two or more clemellts there is always one normal chain, namely
the trivial e1uLin a:J (e) ; however, this may v(~ry W('lI be the only such chain.
Example 2-55. In UII' group (ZI2, +12) of intl'gl'l'li modllio 12, the following
('hnilU~
ure normal dmillH:
Bl'forl' prol'l'l~d ing t.o dl'VI'lop SOITlI! of till' properties of composition chainN,
I('t us pausl' to I'oll:;id('r l<('v{'ral examples.
arl' hot It (lIlllpllsit.ion 1'lminN for (Z24, I 24)' One way of verifying this is to
dll'l'k t Iw ordefH of the l'Iuhgroups eOIl('crncd, For instalw(!, to inscrt, a tmbset
b('t.w('t'n (2) and (4) 111('1'(' would have to exil'lt a suhgroup of (Z24' +24) of order
n, Ij < n < I:!, s\II'h that n divides 12 and is it.self divisihle by (i; clearly no
slIeh l<uhp;l'OlIp ('xistl<.
Example 2-57. For':l ~('('OI\(1 iIIust.ration, let (G, *) he the group of symmetries
of till' sIl"an'. III fhil'l ('nl<!', the following S<'C\II(,IWC of sets serves as n composition
.. hain:
(J J : I( I so. I(alll), fl, V} J : Hallllt Il} J .[ H:IIH1}.
It is worth lIutill/( that tIll' slIhjl;l'OllP (:/(:1110. II:, *) is no\. 11111'1111\1 ill till' pnrl'llt,
P;l'OllJl (1, *), bllt only ill its illulH'diaLt' PI'('III'('(I'ISiIl'.
2-8 'rIlE JOnDAN-I10LD~;lt THEOHEM lin
A normal ('IUlin for (G, *) which fails 1.0 Ill' a ('oIllJlosit.ioll ehain is
This ('huill admils allY olle of Ht'vt'ral I,t'finemellts; among others, tlwre is
Example 2-58. To see that not every group possesses u composition chain,
merely ('onsider the additive group of integers (Z, +), We have previously
observed that til(' normal subgroups of (Z, +) are the c:yclic :subgroups (n), +),
II It nOlllll'gal iv(' intl'gPl', Rinf't' til(' i!lf'lUHiCIII (kn) <;;;; (n) holdH for nil k E Z+,
th('n' always ('xisls It PI'OJl('I' suhgroup of any J!:ivell J!:roup. Aecordillgly, each
dillin for (Z, I) Illay 1)(, rdilled ilult'finitt'ly,
Condilion (3) of Definition 2-:n merits e\OS('f attcntion, Looscly speaking,
it prohibits liS from sql\('ezing U normal suhl/:roup hl'tween (l/j_l, *) and (H j , *),
Th<'l'c is a sl)('cial tC'rminoloJ!:y for this situat ion.
Definition 2-32. A norIllal suhgroup (1/, *) is cl1lled It maximal nurmal
SUbyJ'OllJl of tl\(' group (G, .) if H ~ G and ti1l'rc exists no normal subgroup
(K, *) of (G, .) such that II eKe G,
III terms of I.his IWW notion, a chain
is n composition ('hnin, and we are through. When (H, *) iSllot already maximal
normal, a Illrger normal subgroup (K, *) with 1I eKe G exists. Now, if
(H, *) is maximal in (K, *) and (K, *) is maximal in (G, *), a::J K::J H ::J {e]
is the <i('sin'lI I'ompositioll chain. Cont.inue to argue along these lines. Sinc'
U:, *) is finitl', t1l1're ar!' only a finit.P lIumber of subgroups and thit! procel:l..'i
must ultimatl'ly tNminat!'. The rCHulting ehain of sets is obviously a com-
position (hain for (0, *).
Definition 2-33. Two composition chains for the group (G, *),
are t('rmcd cquil'aicnt if th'y II/we the same length (n = m) and there exists
1\permutation f E Sn Hueh t.Imt
In oth('r wOrill', two ('huills ILre equivl1l('nt if t1wir assoeiat!'d quot.ient groups
are isolJlorphi(~ in Snlll!! order.
We are now in a posit.ion, hnvinj!; I1ssemhled the n('C(!HHury machinery, to
att ack the .J()fIlan-IIoldl'r Theon'lll. To simplify our prcsentation, part of the
proof iH scparated I1H u preliminary lemma.
Lemma. If (1/, *) alld (K, *) art' distilwt. maximal normal subgroups of the
p;rollP (a, *), thl'\I
1) t.lU' pair (II n K, *) iH II nmxinl:ll 1I()J"llIul ~lIh~r()lIp of bo\.h (II, *) ami
(K, *),
2) (a/II, ) '" (K/II n K, ) and (O/K, ) ~ (ll/ll n K, ).
2-8 THE JOIU):\:\I-1I0LUER THEOREM 121
J>1"OIIf. Since the suhgroups (11, *) and (K, *) arc both normal, the pair
(II * K, *) is ilHl'lf a normal suhgroup of (G, *) [Problem 11, Section 2-5].
Plainly, we have the ill('lusion II ~ H * K ~ G. By hypothesis, the subgroup
(II, *) is maximal norlllal in (G, *), which implies either H * K = H or
H * K = G. Our eontelltion is that H * K = G. Were this not the case, the
equality }[ * K = H would mean K ~ H (with K ~ II), contradicting the
m:tximlllity of (K, *); therefore, the only possibility left us is that G = H * K.
A stmightforwunl applieation of Theorem 2-51 now establishes the iso-
morphislll (G/K, 0) ~ (II/II n K, 0). But the quotient group (G/K, 0) is
asilllpll' gmup (ThporPIIl 2-!i2),!lO the same must also he true of (H/H n K, 0).
Fmlll I,his, w(~ condudp the group (/1 n K, *) is maximal normal in (H, *),
again ut.ilizillg Tlwol'l'lJI 2--!i2.
If the ldlns /I and K arc interc1umged, the symmetry of the hypothesis
yields the r('/uuining parts of the lemma.
Theorem 2-54. (Jordan-Holder). In a finite group (G, *) with more than
one element, any two composition chains arc equivalent.
Proof. The proof wiII proceed by induction on the order of the group. If (G, *)
is simple, the only pOllsible composition chain is the trivial chain G::) {e}, so
the t1wol1'm is cert,ninly true. This estahlishes the result for groups of order 2,
the I:!mullest admissible order. III general, let
be any two composition (~hnins for the group (G, *). Assume further that the
theorem holds for all groups having order less than that of (G, *). We dis-
tinguil:!h two easel:!:
('A~m I. III = K I. In this ('asc, if the Sl't. G is deleted from (1) and (2), the
rcsuItin/l: ehllins
anel
III :>K 2 :> ::)K m_ 1 ::)Km = {e}
hoth r('pre~ent (~()/IIJ1()silioll ehains fOI' the I:!uh/l:roup (Il" *). As the theorem
is as!';lIllll'd true for (// t, *), whose order is l(~ss than that of (G, *), these two
chains are necessarily equivalent. But (G/llt, 0) = (G/K l , 0)j hence, the
given I'Omposit.ion ehains (1) and (2) lIlust, also be equivnlent.
(' AHE 2. II. ~ K I' Eilhpr II, n KI = k!, 0" hy Th('Orp/ll 2-53, t.here exists
a ('ompositioll chain
Il, n K, ::) 1-, :> ... ::) 1-r _, :> l"r = {e}
122 mlOt1P TIII';()(lY 2-S
for the tmbgroup (Ill nK 1 , . ) . Now, a(~(~rding to the lemma just estab-
lished, (H InK 10 .) is a maximal normal subgroup of both (H 10 .) and (K 10 .).
It then follows that the two chains .
nJlli
(4)
are actually composition chains for (G, .). ApJX'aling to the lemma once more,
we see that
and
Hence, the eomposition quotient groups ohtained from (:l) are isomorphie in
pail"!! to those ohlllilwd frolll Uw e1min (4); let Ul! ngrcll to ubbrcviate tlii"
situation by writing (;) ~ (4).
Next, consider the following two composition ehains for the group (G, .):
and
Since these ehains have their first two terms in common, we can concludp
from case 1 that they must be equivalent. In a like manner,
and
are equivalent dmins for (G, .). Combining our results, we observe that
Proof. Let
G = Ho:::> HI:::>' .. :::> H n_ 1 :::> Hn = {e}
be a fixed solvable chain for (G, .). Given a subgroup (K, .) of (G, .), define
Ki = K n Hi Ii = 0, 1, ... ,nJ. We intend to prove that the chain
K = Ko:::> K 1 :::> :::> K n_ 1 :::> Kn = {e}
is a Rolvahl{' chain for th~ subgroup (K, .). First, observe that the pair (K i , .)
is Illlorlllul ~lIbJ(I'OIlP of (K i _ h .); in fact., for each a E K;-h
(f. Ki a- I = (a. K i a-I) n K
S;;; (a. IIi. a-I) n K ~ IIi n K = Kj.
Also,
K; = K n IIi = K n l1 i_ 1 n IIi = K'_ 1 n Hi,
so Umt
Finally, because fell;) = 1I~, the function /; maps onto the set H~_dH~. From
th('Hll faetH, W(l ILre ahle to eondu<ie tlllLt
nnd, Hitu!I' (II j . ./11 j, 0) iH n 1()llllllut.nt.iv(' gnlllJl, HlI niHIl iH t.l1Il (Iuot.il'llt. group
(II~-I/lli, 0').
TIl(' following I Iwort'llI giv('s fUl'tiwl' illsight. illlo t.he lIatuJ'(~ of t.hc c'omposit.ion
('hains for 11 fillile solvahle group.
Theorem 2-57. Ld (0, *) he a finite solvable group. Then the quotiellt.
groups of nny (omposition chnin for (a, *) nrc ('yclie groups of p"ime or<1c'l'.
Proof. \Vc pro('cl'cI by induetion on the order of (G, *). If order G is a prime,
there is nothing to provc), sinee G ::::> :e~ is the only c~omposit.ion ('hain. (This
remark takes carC' of I he basis for the incluetion when order (] = 2,) In the
contrary <'a~C, (G, *) has a nontrivial, normal ~ubgroup (II, *), [Why?] By thl'
induct.ion as!-mmption, (II, *) ancl (0/11, 0), being solvable, have composition
chains whose quot.iPllt groups are of prime orciN: say, the dmins
If=lIo::::>lI'::::>'''::::>lIn_I::::>lIn= {e}
ancI
PROBLEMS
1. Check that the following chains represent composition chains for the indicated
group.
a) For (Zafi, +:m), the group of intcgerH modulo 36:
2. Find a composition chain for the symmetric group (84,0). [Hint: Consider the
Imt of pcrmut,ations on p. 80.)
3. Show that no infinite cyclic group possesses a composition chain.
4. Prove that if (G, *) is a finite cyclic group of order 2", then there is exactly one
('om position ('hain for (G, *).
5. D!'rive the result: if the quotient group (G/ll, ) is of prime order, then (H,.)
is a maximal normal subgroup of (G, *). [Hint: Use the Correspondence Theorem.)
6. Prove that the <:Y('!ic :mbgroup n), +) is a maximal normal subgroup of (Z, +) if
and (lnly if n hi a prime number.
7. Establish that the followinl!; two composition ehains for (Zu, +24) are equivalent:
Z24 J (3) J (6) J ~12) J {O},
Z2.' :) (2) :) (4) J (12) J {O}.
8. Find all COmI)()Rition chains for (a) the group (Zao, +:10) of integers modulo 30,
(h) tIll' group of llymml'tril'" of the llquare, and verify t.he Jordan-Hold!'r Theorem
in both (:ases. [/lint: There are four composition chains for each group.)
128 GROUP Tln:OIlY 2-9
9. Apply the .Jordan-Holder Theorem to the group of integerR modulo n to prove that
a positive integer n may be factored uniquely (apart from the order of the factors)
into. positive primes.
10. Ll't (n, .) be t.he dirl'd prodlll"t nf three nont,rivial groups (nk , *k) Ik = 1,2,3);
that ill to Hlty, G = GI X G z X (fa wit.h t1w biliary oIK,mt.ioll . (Idillcd by
(a, h, c) (a', b', e') = (a *1 a', b *2 h', C *3 e').
a) Verify that
If (H, *) is a subgroup of the finite group (G, *), the illdex of H in G was
d~fined to be the number of distinet left cosets of H (ill light of Problem 3(e),
Section 2-1), the numher of right eoscts of H in G is the same as the number of
Icft C'Oscls, so I hc'r!' is no 1I('('d to dist inJl;uish b!'tw('cn It right IUlclleft index). To
silllplify t.1I" Wl'il illl!; lall'l' 011, WI' sllllll IlI'rI'llfll'l' n~pl'l'l'Il'lIl. til" illdex ur II in a
hy [a: HI. Sin!!!! the ord(!r of the group (G, *) may be interpreted as the index
of the trivial subgroup (re~, *), one might, ill keeping with the foregoing
notation, write order 0 = [G: e). We shall not employ this convention, how-
ever, hut instead conform to the praetiee of denoting the order of (G, *) by the
symbol 0(0).
For any I'IlIhJ!:l'OuJI (IT, *) of tIll' /l;roU(I (0, *) IlllclllllY ('IC'lIl1'lIl, a E a, consider
the followilll!; I'IlIhsl'l. of (,:
whprc', sin!!(' (IT, .) is a subgroup, the p ro<i II ('f. hi * It:; I E II. But this mellns
X*y-I Ea*lI*a- l , so that (a*ll*a- I ,*) is its('lf a subgroup of (G,*).
We are thus led to the next theorem.
Theorem 2-59. If (H, *) is a subgroup of the /l;roup (G, *) and if a E G,
t.Il1'lI t.Il1' pair (n * IT * a-I, *) is al::;o a sllbgl'OllJl of (a, *) c~alled the conjuyate
sub(Jroup of (II, *) itll/w'l'Il by the element fl.
If a * II * a-I = II for sonw a E a, Wl' say that the :-mhgroup (II, *) i::;
invariant, or sdJ-conjullatf:, under a. AllY sllbgl'Oll(l i::;, of COUfsc, illvlu-illnt
under each of it::; own clements. From this definition, it is immediate that the
eoneept of It nornml ::;ubgroup can be arrived at in an alternative way: a sub-
group (/I, *) i::; normal in (G, *) if and only if (H, *) is invariant under each
clement of G. \
Observe also that conjugate subgroups are i::;omorphic. If (a * H * a-I, *)
is It conjugate subgroup of (II, *), define J: II --> a * H * a-I by taking
J(h) = a * h * a-I; then J is all i::;olllorphism.
Now, if (K, *) is any other subgroup of (G, *), we may cert.ainly c~onsider
those element::; of K under which (IT, .) is invariant; in ot.IH'r words, the spt
N K(I/) = {k E K I k * H * k- I = l/}.
When K = G, one customarily writc13 N (H) for N a(H) and refers to th(
suhgroup (N(H), *) simply as the normalizer of H. Although the subgroup
(H, *) i:-< not necessarily normal in (a, *), a simple argument shows it to 1)('
normal in it s normalizer; in fact, (N (H), *) is the largest subgroup of (G, ,.)
in whieh (1/, *) is nomml.
Wit.h t.1t is not.ation ill rn illd, W(~ can now "(lOU!lt" til(! eOlljugate subgroups
(a" 1/ * a-I, *) of (II, *).
Theorem 2-61. The number of distinct conjugates of a subgroup (H, *) of
(G, *) induced by the elements of a subgroup (K, *) is equal to [K: N K(H) J.
the index of N K(ll) in K.
l'I'ol/!, For "'1, 1>2 E K, WP haV<' kl * 1/ * k,1 -= "'2 * /I * k;-I if and only if
/I (I> I I .. k~) * 11 * (k i I * I.''.!) -I, whidl lIol<ll'! if ml(l (Jnly if kil * k2 E N K(1I);
t,his is tme, in turn, if lind only if k\ * N K(H) = k2 * N K(H). Hence, the num-
hl'r of distin<'1, ('onjugate subgroups of (II, *) by clements of K is equal to the
!lumber of distind coseis of N K(H) in K, namely, [K: N K(H)].
Corollary. "Pl. (II, *) and (K, *) he t.wo subgroups of t.he group (G, *),
If (II, ,.) is illvariallt ullder n e1ellJ('nts of K, then (1/, *) has o(K)jn conjugat(~
suhgroups by elements of K.
PI'oof. TIIP hypot.hesis asserts o(N K(lI) = n. But, by Lagrange's Theorem
tI(K) = [K: N K(H)] o(N K(lI),
from whi('h t.I11' ('orollnry foHoWR.
I.('t us I'ont inuI' our dis('w~sion hy introducing It llew elMS of groups.
/
Definition 2-35. A group (a, *) is said to hI! 1\ p-group if the order of
('Iwh 1'1('IlII'IIt. of (l is Jl powl'r of It fixed prinw p.
Example 2-60. Any group of order pn (11 It prime) is It 11-grouP, since till'
(mh'r of I'lli'll 1'11'111('111, IJIlISt. divide t.he or<i(H' of t.he group. In partieular, UII'
group of syntnll't.J'iI'H of a slJuare is It p-group, where p = 2.
Example 2-61. Let (U, *) h(~ It eommutative group and the set H consist of
thosc e1l'l1ll'lIts whosl' orders ar(' powers of a fixed prime p (quite possibly,
/I = :(~1). TllI'n (II, *) fOl'llls a Hubgroup of (a, *) which, by its definition, is
Ilutoillat i(,lllly a 11-l!;l'oup.
0111' llIil!;lIt IIsk wlll't \11'1' till' ('OIIVI'rH(' of EX!ullpl(~ 2-liO holtlt!. More pnwitlely,
if (0, *) is a fI-I!;J'OuP. is I/(U) SOI1W power of p'! The following lemmu will
SI'I'V(' liS Il start illl!; point ill ohtailling til(' :tIlSWl'r,
identit.y has order 7); this take::! care of the basis for the in<iuetioll when
o(G) = 2. Now, let a E G, with a ~ e. Since (G,.*) is a tinite group, the
element a hUH It finite of(ler n. Either 1) I n 01" fI tn. III case pin, say n = kp,
then (ak)P = an = e. The order of a k therefore divides p, which, because p
is prime, impli(!s /I(ak) = p. (Recall thlLt. ;r m = e if Iwd only if o(x) 1m.)
Next, Il't us SUJlPOS(~ fJ tn. lleclLul'!e (0, *) is commut.ative, !.Iw eydie sub-
group a), .) is surely norlllal, and WI! can form the ~omlllutILtive) quotient.
group (O/(a), ). Allio, sinee (a) ~ {e], o(O/(a < n(O). From
and till' fnd. 1/ tn, it, follows tlmt. o( a/(al) is divisihle hy fl. lTsinJl; our induction
hypoUH'sis, U/(a) JIlUHt t11('1'('fol'(~ eOlltain l'!onw dl'nwllt. Ii * (Il) IIf llI'd,,!' p. But
the equation
1/' * (a) = (b * (aP = c * (a)
implies bP E (a), 80 t.hat (bn)P = (bP)n = c. Were I)n = Il, then (b * (an =
e * (a) and we would eOlwllld" 1/ I 1t, It I'ontmdi(~t.ion. Thill'! the dement. bn has
order p.
Corollary. Let (G, *) be a finite eOJlllllutative group and ]I It prime dividing
0(G). Then (G, *) hItS IL subgroup of order p.
Our next ohjl'et.ive iH to ('siabliHh tlJ(' preceding lelllllla without. the hypothesis
of commutativity. We puu::!e, hOW(lVer, to develop tllll lIe(lCSKary mllthemt~tical
tools.
Definition 2-36. The centralizer oj an element a of the group (a, *), denoted
hy C(a), is t.Iw s('t. of nil ,,1!IIWII!.t1 in (/ whi/h (Ollllllu1.n wit.h a:
Definition 2-37. L(,t, (a, *) hI! a j./;101l1> alld u., II E (r'. The dl'lIl1'ut. b il'!
Haid t.o be It ('{Inju(!a/I! of a (ill (l) if j.lwre exi~ts ~()IIW E (J for which
b == x * a * X-I.
The reader shoukl now prove th( following unalog of Thporelll 2-(il which
clarities the rel:ttiollship hetween the last two delillitiollH.
GHOUP TJn:OIty 2-9
As pach elonwnt of G lies ill one and only one ('onjugaey class, we may deter-
mine the lIumber of (lements in G by eounting t.hem class by class. To do so,
first observe that. there arc exactly o(cent G) dasses which contain one element.
If we add to O(I:OIlt. G) the number of clements in the nonsingleton conjugacy
eiasl:!es, the c1emcnts of G will have all been eoullted. This leads to the im-
portant (conjugacy) clas8 equation:
0(0) = o(cent G) + [G: C(R 90) + [0: C(D 1)] + [0: C(H)].
2-!)
Using the dass equation of a group, we are able to ('x tend the last lemma
on commutative groups to a more general setting.
Theorem 2-64. (Cauchy). If (G, *) is a finite group whose order is divisible
hy a prim' 71, Own 0 ('onl.nills an element of order p.
JlrtltlJ. I1t>!C' illdlldioll 011 IIII' III'dl'r of (U, .), If tI(U) :.!, IIII'll (G,.) is It
"Ollllllllinliv,' ).(I'III1P nltd II", 11"'01'('111 I'..dlll'('t>! t.o IhC' 1"III11I1L, HIIJlJlIII!, ill-
dlldivC'ly, 1.11111. till' ,111'01'1'111 holds 11'111' rill' /1.11 ~I'OIIJlt>! of III'dl'l' 11't>!t>! Umn 0(0).
Now ,'ollsid"r t.he dnss "'Illation 1'01' (U, .),
Hilll'C' t.Ite t.1,,'or,'1lI it>! IlSSllIIlI't! t.o hold for t.hl! HuhJ,!;nlllp (('(It), *), if Jllo(e(a,
then e'(a) has all element of order p, and wc are donc. On tltp ot.her hand, if
1d o(C(a for evpry a fl eentdl, 11 must divide [a; C(a); for JI is prime and
divides 0(0) = [G; C(a)) o(C(a). But then, in the dass equation, 11 divides
each term of t.he summation alld also divides o(G), so that p I o(cent (J). As
(cent G, *) is n comlllutat.ive subgroup of (G, *), it follows (again, from the
II!mom) t.hat. ,~ellt. (J ('ollt.aiIlH :lII 1'1('IIII'II!. of ordl'r fl.
Corollary. (0, *) is:t finit.1' p-J,!;roup if :tll(i only if o( a) = 11k, for some k > o.
Proof. Rupposc (U, *) is a /I-group, but. q I o(G) for some prime q p. Then
by Cauchy's Theorl'lll, 0 has lUi element of order q, I!ont.radil!tin)!; the fact that
(0, *) is a {I-group. Thus, JI is the only prime divisor of o(J), which implies
o(G) = 11k (k > 0). Conversdy, if o(a) = pk, then from Lagrange's Theorem
each clement of G has a power of 11 as its order.
As all appli'at.ioll of C:ulI'hy's Tll<'ol'l'l1I, it. lIlight. Ill! of intl'rl'st. t.o mention
t.he r('sult. hl'lo\\'.
Theorem 2-65. If (a, *) is It finit.!! p-grollJl wit.h 11101'1' Ihall OIW elcl1I('nt., t.hen
(:ellt a f/:.
Proof. As a start in)!; point, consider the class equat.ioll for (G, *),
[G:C(a) = o(U)/o(C(a
llIu:;t 1)(' divi:.:ihl(~ hy p. ~ow, /I divide:; clwh term of the summation, as well
a:; divicll':; o(l), :;0 that. JlI n(ccnt. G). Thi:; implies o(ecnt G) > 1.
WI' II0\\' :;hifl our I'IlIJ1ha:;i:; frolll nrhit l'ary JI-lI;ronps t.o Sylow p-subgroups.
Definition 2-38. LC'I (G,.) be n fi II if,(' group IUld p It prime. A subgroup
(1', .) of (0, .) is said to be a Sylow p-subymup if (P, .) is a p-group and is
not properly ('()J\jaill!'d ill nny olh!'r p-subgroup of (G, .) for the same prime p.
Example 2-63. Thp :;Ylllllwl.ric group (.'01:1,.) hILS thrpc (eolljuglltc) Sylow
~-snhgl'llup:;, slllpili(ally, t.he Mnhgrnup:.: who:.:c! sf't.s of dmmmts nrc!
Utilizing the Hylow Tlworl'ms, we c!tn obt!tin some rat.her general results con-
('('rnill~
IIII' ('xil'<tplH'p of I'<lIh~rollpl'< of rillit.e grollpl'<; for t.his, olle IIl'l'd ollly
kllow sOllldhilig ahout t.l1I~ oJ'(ll'r of tlw grollp.
The next theorelll~ are major ones and require t.wo preparatory lemmas.
Lemma. Let (II, *) h(' a normall'ubgroup of (G, *). If (H, *) and (GIH, )
nrc bot h p-groupH, 1.111'11 (a, *) it.self iH It l'-group.
Proof. For I'llph a ill a, the ('Ol'<ct a *11 E alII has order some power of p,
Ray order pk. Then, apk * II = (a * H)Pk = e * H, so that apk E H. Thus,
the' elelllent. apk abo hm; p-power order, for inst.allce, order pi. This implies
(a,k),,/ = a ,Hi = I', WIWIII'(' o(a) I Jlk+ j Bllt, silwc fI i;; 1\ prinw, it follows
that. 0(0) is II power of /1.
Hill('(' allY ('olljllgat!' of (1'1, *) is, at. thc SIllIU' tinll', a ('olljugate of (P, *),
we lirst. ('ollsidl'l' (onjugat.!'s of (1'1, *) illlhll'ed hy dl'nwllt.s fl'lllll P. Aeeording
I :~{j (;I{()l'i' Tln;o!!y 2-9
10 till! last, 1('111111:1, Ih" 1'1"lIlent.H of I' 1I1ulel' whidl (1'1'.) IS illvariant an'
prpeisl'iy t.hos(' of 1'1 n I':
Hilll'(' (/'1 n /',.) is II p-sIlIoJ(l'lll/p of (f:, .l, it. follows t.hat. 11(1'1 n I') is a
powl'r of p, ~ow, from Theof'(,111 :!-lil, the number of diHtinct eonjugn\,cs
of (/' 1, *) by d(!llIl'lItlol of P is (!quul to
ThiH "IIII\\';; Ihali/': N1,(I'll] nll/st. III' >11111' (low('r of p, "Ily plr.", Furtill'rlllOre,
Itl > 0, fof' if Itl = 0, Ih('n 0(1') = 0(1'1 n I'), HO l' = PI n 1', or rather
l' ~ 1'\. Bllt I hI' fae! that (1', .) iH a maximal p-subgroup would imply l' = PI,
whieh iH impol'l'ihl('. l'\ote also that (1', .) will not be found among the con-
jlll!;at(' SlIhl!;l'OIlJlH of (1'1, .) by ell'n1l'nls of P; weI'(' P = a. PI a-I for some
a E P, the'n 1'1 = a-I * p. a ~ P, again h':u\inl!; to a eontradietion,
l\OW, if t.hl' ('OIijlll!;ah' ~iUhl!;roups of (1', .) are no\' exhausted by (I', .) and
the pk, eonjugates of (1'1,.) indul'et\ hy nl('mbers of 1', choose ano\'her con-
jugate (1'2, .) distini't frolll any yet ('numerat.cd. As above, we can obtain the
pk2, "'2> 0, di"linl'l eonjugat.(,H of (1'2,.) indll('ct\ by P. No ('onjugate of
(1'2, .) hy an ,,11'1lI1'1I1 of I' will alHo b(, 11I1 (~onjllgal.(! of (1) ... ) by IlII clement
of 1', for if
a, iJ E P,
then (b- I a) '" PI (iJ- 1 a)-I = 1'2, with b- I a E 1', contrary to the
choice of P 2 If nce(,HHary, n'p('at thiH argument onee mor(', Since (G, .) is a
finilc grollp, all the eonjllgates of (I',.) are found af\'er a finite number of
HI('pH, Hay n sl('ps. Tlw ('onjllgate subgroups of (1', .) Ilrc these: (1', .) itself,
the pk, eonjllgal('s of (/'), .) by elements of P, \'he 11 k eonjuglltes of (1'2,.)
by eleml'Tlts of P, ('Ie'. 'I'll(' t.otal numlwr of distinet (!onjugntcs is therefore
I I 1/" I 1/" I , .. I pk. = I 1 mp == I (mod p), k. > 0,
At this st.age, an ohvious qucstion arises: are then! allY Sylow p-subgroups
whieh nrc not. ('onjllgate suh!l;roups of (1', .)? The unswer will be sccn to be
no. Let us llSSlIllW slwh It subgroup (R, .) does exist, anel arrive at a contra-
did.ion. As h('fOl"(, , we eount the eonjugates of (R, .). This is done by first
lilldillJ,!; tI)(' lolljllJ.!:at('~ of (H,.) indlJ('l'c\ hy ('11'1lH'nt.s of P; t.here nre
(I(J/o(N n 1')-= pi,
of them, whel'(> j. > 0 (if .il = 0, then R = 1', n eontradiction). Note that
(R, .) i" its('lf jlll'huh-d in this eOllnt, sinee R = e R e-t, eEl', Following
Ihc proc'('dllrl' ahov(', II\(' I.olnl III1111bN of (~onjugat(' subgroups of (R,.) is
found to be
j; > O.
2-9 SYLOW THEOREMS 137
Thc ilJlport.:Illt. point 1l('I'(' is t.hat no l('rll1 of this SIIIII /'('(ilw(,1-l to 1; if some
j. = 0, t.Il(~n (ll, *) wOllld I)(~ a (~()njllgat.(! of (I', .), whidl is impoH!>iblc, But
we havl' Jll'('viollsly dl'tl'rJIlilH'd Ihat. I hI' nllllll)('1' of ('I)njllgah's of any Sylow
P-HIlItJ,!;I'OIiP i" (,OIlJ,!;I'tIl/tt. to I IIlo<iulo /I, Hi/wI' no illtl'gl'/' I':tll lit. 0111'1' hI'
"'" 0 (Illod /I) IIlId -- 1 (Illod II), WI' havl' t.ll(' dl'sirl'd ('olltl'luli(,t.ion,
Filially, t.Iw lIumher of distinct eonjugal.('s of till' Sylow p-subgroup (/', .)
is, by Theorem 2-61, [G: No(P)]; sinee (Nu(P),.) is Il. subgroup of (G, tit),
[G: NGW)] divides o(G), This proves the theorem in its entirety,
nre prime to p, Now, [a: N(P)] is equal to the number of eonjugntc suhgroups
of (1', tit) IUIII, hy our last. r('su It. , this valuc must. be ((lIIgrUl'nt. to 1 modulo p;
thereforI', HI(' possibility 1'1 [a:N(p)) eannotoeeur,
To :see that o(N(P)/1') ha:s no p-power divisor, let us assume the coset
a P E N(P)/ P is of order 1', Then, as before, a P I' = (a'" P)P = e tit P,
whieh 1l1l'IlnS a P E 1'. As (I', tit) is a p-group, the element a P has order a
pow('r of 1', HO t1mt a itself has [I-power o rill' r, Because a E N (I'), the Inst
lemma illlpliml a E J'; hp/wl', a'" J' -= e 1', or o(a. 1') = 1. From thiR
eontradidion and Cmll'hy'H Theorem, we illfl'r t.hat p t o(N(I')/I'), Since p
is a prime and divides neither of the factors of [G: 1'], it cannot divide [G: PJ
either, thereby eompleting the argument,
,
Example 2-64. There call be 110 simple group (a, *) with o(G) = 42, Sillee
42 = 2, 3, 7, the preceding theorem implies the group ('olltains a :subgroup
(II, .) of order 7, By Throrem 2-()7, the lIumber of (!onjugate subgroups of
mWlJl' TJlEOHY 2-9
(//, *) ii'! of til(' fOl'lll J I 7k, k E Z, anti tlivid('"42; k = 0 i~ the only pO!l8ibilit.y.
11 .. ,11''' (1/, *) il'l (.Fli( lind 'L 1l01'1I11t1 Hllill(I'Oll\l (i .... , HIM-I~onjlll(lLt(l), "0 t.JIILt.
(1, *) ('allllOl 1)(' ~illlpl('.
From Thl'orem 2-(i)';, WI' can obtain a partial converse to Lagrange's Theorem:
If (0, *) is a fillite group whose order is a produet of powers of distinct primes,
Ul('n (G, *) ha~ subgroup~ of ordl'rs p~i (i = 1,2, ... ,n). As a matter of fact,
it enn 1)(' ~howlI t.hat, for I'l\('h prime Pi, (G, *) eontains subgroups of orders
{Ii, {I;, /1;1, ... ,{I~" Thlls, for im,t.ane(, ILIlY p;rollJl of ord..r ]2 = 22. :J will
hav(' Huhgl'llups of ol'deri'! 2, 2~ alld :1; nothing, how(wer, can be concluded ill
regard to IL subgroup of order n.
Example 2-65. Our work up t.o now hus given us sufficient tools to show that
any group (G, *) of order 1.') is cycliC'. Since 15 = 3 5, we first note that (0, *)
has at least, one Sylow :l-suhgroup and at least one Sylow 5-subgroup. The
number of Sylow 3-suhgrollps must, by Theorem 2-67, divide 15 and yet be
== 1 (mod 3). As the only such integer is 1, there is a unique Sylow 3-subgroup,
call it (SYa, *); in other words, (SY3, *) is cyclic (being of prime order) and
normal in (G, *). By the same reasoning, there is precisely one Sylow 5-subgroup,
label it (SYIi, *), whieh is also cyC'lic and normal. We propose m show that
(G, *) ~ (Sy:! X BY5, .).
Detouring slightly, let us record several preliminary observations: Because
(SY3 n SYs, *) is n xuhgroup of both (SY3, *) and (SYI;, *), it follows that
o(Sy:! n By ... ) divid('s hoth :land !i. The only possibility is for o(SYa n SYa) = 1,
whelll'e BYa n SY5 = {e}. In particular, this menns each element of (SYa, *)
rommutes wit.h eReh clement of (SYs, *) [problem 10, Seetion 2-.')].
Next, we considl'r ('osds a * SY5, where the clement a E SY3' These coscts
arc evidently nil distilll't,; for, if a * RY5 = 1) * RY5 (a, b E SY3), then
It .,ydi(~ group of or<il'r I:; (t.he ('qllipnwnt, nl'('<i(~d is Prohlem 22, Section 2-7).
2-9 RYLOW THEOREMR 139
'fhii'! filliKhl'i'! our Kt.lldy of KrullP f.1l1'ory. WI' h/\VI' lint nt,\,l'lllpl.l'd to rnlll/:"
live!" till! whol(~ UU'o!"Y 1101' tu cXlUuilW ill uepth allY pluticular Wlpet:t of it.
Instead, we have hwrely seratched the surface, introducing the reader to a
few of the high points. Needless to say, current research in this branch of
mathemnties is both vigorous and extensive. A variety of dassical problems
still remain unsettled, while in some directions the research has only recently
begun.
PROBLEMS
RING THEORY
while
(a b) (a c)
0 0 = (2ab) (2ac) = 2ab + 2(2ac)
= 2ab + 4ac.
Because is a commutative operation, we (!onclude that it is also right distribu-
0
tive over .
Definition 3-2. A rina iH a mathematiell.l system (ll,., 0) eonsisting of a
1J0nempty !let R aud two hinary operntions and. ddined on R such that
1) (H,.) is a commutat ive group,
2) (R, 0) is a scmigroup,
:J) t.he !I(,JIIigrollp olll'ratioll iR distrihutive over the group olwrll.tion .
It is convenient and customary to usc + for the group operation and for
the semigroup operation, rather t.han the symbols. and o. This convention is
long-standing and is parti('ularly helpful in emphasizing the analogy between
result.'i obtained for rings and those of the familiar numher systems. The above
definition then takes t.he following foml.
Definition 3-3. A ring is an ordered triple (R, +, .)
consisting of a non-
empty set R and two binary operations + and on R such that
1) (R, +) is a commutative group,
2) (R,') is a semigroup,
3) the two olwrations arc related by the distributive laws
a (b + c) = + (a c),
(a b)
(b + c) . a = (b a) + (c a)
for all a, b, c E R.
The reader should clt'arly understand that + and . represent abstract un-
specified operations 1\111\ not ordinary addition and multiplication. For con-
venience, however, we shaII refer to the operation + all addition and the operation
as multiplication. In the light of this terminology, it is natural then to speak
of the commut!Ltive group (R, +) as being the additive group of the ring and
(R, .) th(' multiplicative semiyroup of the ring.
The unique identity elem('nt for addition is calIed the zero element of the ring
!Lnd is denoted by the usual symbol for zero, O. The unique additive inverse
of an clement a E H !lhalI he written as -a.
In order to minimiz{' the use of parentheses in expressions involving both
operations, W{' make the Rtipulation that multiplication is to be performed
h('fore addit.ion. Thlls th(' ('xpression a b + c IItands for (a b) + c and not
fOI' a . (b I c).
With this notation in mind, we {'nn now amplify our eurrcnt definition of a
ring. A ring (R, +, .) ('onsists of a nOllempty set R and two operations, called
addit.ion llnd Illult ipli('at iOIl and denoted hy + and " respectively, satisfying
a-I UEHNITION AND ELEMENTAUY PItOPEltTn.;S OF lUNGS
the requirements:
1) R iH closed under audition,
2) a + b = b + a,
3) (a + II) + C = a + (b + c),
4) t.here existl:! an element 0 in R sueh that a + 0 = a for every a E R,
Ii) for <!twh a E R, there cxil:!ts an element -a E R I:!uch t,hat a + (-a) = 0,
(j) R iH dOHed IInder Il1l1ltiplieation,
7) (a b) c = a (/). c),
~) a (II+ +
c) = a b a c and (b + c) . a = b a + t: a,
wlJl'rl' it, is ululerHtood t.hat a, I}, t: represent arbitrary elements of R.
By pl:willl!; further rl'sl.ri(:tionH Oil the opemLion of Illultiplieation, !!everal
special types of rings arc obtained.
Detlnition 3-4. 1) A commutative ring is a ring in which (R,) is a com-
mutative scmigroup; that is, the operation of multiplication is commu-
tative.
2) A rina with identity is It ring in whieh (R,) ill a semigroup with identity;
that is, there exists an identity element for the operation of multiplica-
tion, customarily denoted by the symboll.
In a ring (R, +,.) with identity, we sayan element a E R is invertible if it
possesses an inverse relative to multiplication. Multiplicative inverses are
unique, when they exist, and shall be denoted as before by a-I. The symbol
R* wiII be used subsequently to represent the set of all inyertible clements of
the ring.
Theorem 3-1. Let (R, +,.) be a ring with identity. Then the Jlllir (R*,)
forms a group, known aI:! the group of invertible elements.
Proof. The set R* is nonempty, for at thc very least the multiplicative identity
1 E R*; moreovcr, 1 serves as the identity clement for the system (R*, .). If
a, bE R*, the equation!!
(a b) . (b- I . a-I) = a 1 a-I = 1,
(b- I . a-I) . (a II) = 1>-11 b = 1
The first three of t.he!!C rings have an identity element, the integer 1, for multi-
plication.
Exampl. 3-3. Anotlll'r ('xlunple of a rillK is provided hy the sct
R = {a + bval a, b E Z},
and the operations of onlinary addition and multiplication. The set R is ob-
viously closed under these operations, for
A n (B ~ C) = A n [(B - C) u (0 - B)
= IA n (B - C) u [A n (0 - B)]
= ((A n B) - (A nO) u [(A n C) - (A n B)
= (A n B) ~ (A nO).
Example 3-5. Let R denote the tlCt of all functions I: R' -+ R'. The sum
I +g and product I . (f of two functions I, g E R are defined as usual, by the
equations
(f + g)(a) = I(a) + !/(a),
(f. g)(a) = I(a) . g(a), a E R'.
In other words, we specify the functional value of these combinations at each
point in tlwir domain. That (R, +,.) is a (~ommutat.ive ring with identity
J)";~'INITION AND ELJo:MENTAUY PROPEIlTlJo:S OF RINGS 145
follows from the fact that the real numbers with ordinary addition and multi-
pliention comprisc 8u(!h a system. In particular, the multiplicative identity
d('lIwnt ill the NIIIHtUllt fUllctioll whoHe value at each real number is 1.
It iH int('rpsting to nof.(! that the triple (R, /-,0), where 0 indicates the opera-
tion of fUllctional eompositioll, fails to be a ring. The left distributive law
With the possible exception of the closure condition, it is fairly evident that
the pnir (hom G, +) is a commutative group; the trivial homomorphism acts
as the zero element, while the negative of each function I E hom G is obtained
hy taking (-- I)(a) = f(a)-l fOr nil a E G. To estnhlish (llmlllm, (!hooBrlnrbitmry
elt!lIlcnts a, b E (J and fUllctiollti I, y ill hom G. Theil
Example 3-7. In Section 2-3, we considered the group (Z.. , +..) of integers
modulo n. This group was obtained upon defining in the set Z.. the notion of
addition of congruence classes:
We first observe that a' E [a'] = [a] and b' E [b'] = [b] imply that
Given a ring CR. +.. ), all the results of the previous chapter on groups apply
to the system (R, +). For instance, we know that the zero element a.nd additive
inverses are unique, that the cancellation law holds for addition, and so on. In
the theorems to follow, we shall establish ;;Dme of the fundamental properties
which depelld 011 hot.h ring operations. As usual, we may aSl"lUme nothing about
t.he Hpm:ific IIUI.III"" of t.he KYHt."1lI (R, I,) x:ept. t.hut. it. Hat.iHfies the pOld,ulates
presented in Defillition :J-a.
Theorem 3-2. In any ring (R, +, .), if a E R, then a 0 = o a = O.
Proof. First note that an application of the left distributive law (and the fact
that zero is t.he idl'llt,ity element for addition) yil'lds
a = a 1 = a 0 = 0,
Here, we have 2'42 == 0, the product of nonzero elements being zero. Note
also that 2 . 4 1 = 2 4 3, yet it is clearly not true that 1 == 3. The multiplicative
semigroup (Z4, '4) does not satisfy the cancellation law.
Deftnhlon ~5. A ring (R, +, .)
is said to have diviBor, oj Z6TO if there
exist nonzero elements a, b e R such that the product 0 b == O.
We exhibited two rings in the above example which possess dWisors of zero.
The second ring, in particular, suggests a relationship between the existence
of divisors of zero and the failure of the cancellation law for multiplication.
We shall see shortly that this is indeed the case.
First, several preliminary results concerning additive inverses are required.
+
We shall adopt the usual convention of writing a (-b) as a - b and refer
to this expression as the difference between a and b.
Th.....m 3-4. Let (R, +, .) be a ring a.nd a, b e R. Then
a (b - c) = a b - a c, (b - c) . a = b. a - c a.
a (b - c) = a (b + (-c) = a b + a (-c)
= a b+ (-(a c)
=ab-ac.
In a like manner, a ring distributive law for differences is obtained.
Having thus laid the groundwork, we are now able to establish the following
result.
Theorem 3-5. A ring (R, +, .) is without divisors of zero if and only if
the cancellation law holds for multiplication.
Proof. First, we assume the ring (R, +, .) contains no divisors of zero. Let
a, b, c be elements of R s~ch that a 0 and a b = a c. Then,
a (b - c) = a b - a c = o.
Since a 0 and (R, +, .) has no zero divisors, this last equation implies
b - c = 0 or b = c. In a similar manner, we can prove that b (I = c (I,
~ith a 0, yields b = c.
Conversely, suppose that the cancellation law holds in the semigroup (R,)
and the product (I. b = O. If the element (I is nonzero, then it may be can-
\eeled from the equation (I. b - a 0 to conclude b = O. On the other hand,
b 0 implies (I = 0 by the same argument. This shows (R, +, .) is free of
divisors of zero.
Corollary. Let (R,+, .) be a ring with identity which has no zero divisors.
Then the only solutions of the equation a2 = a arc a = 0 and a = 1.
Proof. The proof is easy. If (12 = a = a 1, with (I 0, then by virtue of
the cancellation law a = 1.
Definition 3-6. An integral domain is a commutative ring with identity
which does not have divisors of zero.
We have just shown that the cancellation law for multiplication is satisfied
in any integral domain. .
150 RING THEORY 3-1
Deftnitlon 3-7. Let (Il, +,.) be a ring and S ~ R be II. nonempty subset
of R. If th(> triple (8, +, .) is itHClf a ring, then (8, +, .) is said to be a 8ubriny
of (R,l ,.).
An examination of the definition of ring as given in Definition 3-2 shows that
(8, +, .)is a subring of (R, +, .)
provid(>d (8, +) is II. subgroup of (R, +),
(8, .) is a suhsemigroup of (R, .), and til(' two dist.ributive laws hold for ele-
mellts of S. But hoth t,he distributive alld assoeiativ(> laws hold automatically
ill S as a ('Ollsequl'll('(> of their validity in R. Sinee these laws arc inherited from
H, there is 110 partirular n('('(>ssity of requiring them ill the definition of II. subring.
In view of this observation, II. subring eould alternatively be defined as
follows: the t.ripl!' (8, -1-, .) is It suhring of the ring (R, +, .) whenever
1) S iF; II nOIl(,llIpty HuhHet of Il,
2) (8, +) is a subgroup of (R, +),
3) S is closed undcr multiplication.
Even this definition may be improved upon, for the reader may recall that
if 0 ~ II ~ G, then the pair (II, *) is a subgroup of the group (G, *) provided
that a, bEll implies a * b- 1 E H. Adjusting the notation to our present
situation, we obtain a minimal set of conditions for detennining subrings.
Example 3-9. Evpry ring (R, +, .) has two trivial subrings; for, if 0 denotes
the zero element of the ring (R, -1-, .), then both ({OJ, +, .) and (R, +, .) are
sub rings of (R, +, .).
Example 3-10. In thl' ring of integers (Z, +, .), the triple (Z" 1-,') is U Hub-
rin!!:. while (Zo, +, .) i~ not. In particular, we infer that in It ring with identity,
a suuring does not n('('11 to contain the iden-
tit,y element.
+6 0 2 4 '0 0 2 4
Example 3-11. COllsi(\l'r (Zo, I 0, '6), the - - - - - -
ring of int('gl'rs modulo (i. IfS= ~O, 2, 4) , 0 0 2 4 0 0 0 0
then (8, +6, '6), whost, op<'mt.ioll tables arc 2 2 4 0 2 0 4 2
given at the right, is a subring of (Zo, -h, '6)' 4 4 0 2 4 0 2 4
3-1 DEFINITION AND ELEMF.NTAUY PltOl'lmTIES OF RINGS 151
Since neither a I' - a nor I' is zero, the ring (R, +,.) has zero divisors; in
particular, I' is a zero divisor of (R, +, .).
To give a simple illustration of a ring in which possibility (2) occurs, we may
refer to Example 3-8. There, the system (R ' X R', +, .) was shown to be a
ring provided addition and multiplication are defined by
(n + m)a = na + ma,
(nm)a = n(ma),
n(a + b) = na + nb.
152 RING THEORY ~1
In addition to these rules, there are two further properties resulting (rom the
distributive law, namely
n(a b) = (na) b = a (nb),
(na) . (mb) = (nm)(a b).
2A = A !:l A = (A - A) U (A - A) =0
from every subset A of X.
Theorem 3-6. Let (R, +,.) be a ring with identity. Then (R, +,.) has
characteristic n > 0 if and only if n is the least positive integer for which
ni == O.
Proof. If the ring (R, +, .) is of characteristic n > 0, it follows trivially that
nl = O. Were mi = 0, where 0 < m < n, then
ma = m(I . a) = (mI) . a = o a = 0
(or every element a E R. This would mean the characteristic of (R, is +, .)
Jeaa than n, an obvious contradiction. The converse is established in much
the same way.
Corollary. In an integral domain, all the nonzero elements have the same
additive order, which is the characteristic of the domain.
Proof. To verify this assertion, suppose the integral domain (R, +, .) has
positive characteristic n. According to the definition o( characteristic, any
a E R (a F- 0) will then poBse88 a finite additive order m, with m S n. But the
equation
o= ma = (mI) a
implies ml = 0, since (R, +,.) is free of zero divisors. We may therefore con-
3-1 DEFINITION AND ELEMENTARY PROPERTIES OF RINGS 153
clude from the theorem that n ::; m. Hence m = n and every nonzero element
of R has additive order n.
A somewhat similar argument can be employed when (R, +, .) is of character-
istic zero. The equation 100 = 0 would lead, as before, to ml = 0 and con-
sequently m = O. In this case, each nonzero element of R must be of infinite
order.
The last theorem serves to bring out another point.
Corollary. The characteristic of an integral domain (R, +, .) is either zero
or a prime number.
Proof. Let (R, +,.) be of positive characteristic n and assume that n is not a
prime. Then n can be written as n = nln2 with 1 < ni < n (i = 1,2). We
therefore have
PROBLEMS
a*b - a+ b+2,
a b - ab
0 + 2a + 2b + 2.
Show that i~ dhltributive over *.
0
2. Let (R, +) b(' any commutative group. Determine whl'th('r (R, +,.) forms a
ring if muitipiieation is dt'finl'd by
a) a' b = a.
h) a' b = O. wh('re 0 is the identity element of the group (R, +).
3. Let (R, + .. ) be an arbitrary ring. In R define a new binary operation by the 0
rule
aob = ab+ba
fllr all a. bE fl. Establish that (fl, +,0) ill a commutative ring.
4. Obtain the group of invertible elements (ZT2, '12) for the ring (ZI2, +12, '12). In
addition, show that (Z!2, '12) is isomorphic to Klein's four-group.
5. Given that a, b, c, d are clements of a ring (R, prove that +, .),
a) (a +
b) . (c +
d) = a' c b.c a' d + + + b . d,
b) -(abc) = (-a) (-b) (-c),
(.) if a . b = b . a = 0, then (a b)" = an + + b" for n E Z+.
(As usual, a" = a' a ... a, n times.)
6. Define two binary op('rations * and 0 on the set Z of integers as follows:
a*b=a+b-l,
a .. b = a + b - abo
(a + b)2 = a 2 + 2(a . b) + b2
(a +
b)1' = a" b" +
for all a and b.
20. Suppose (R, ., 0) and (ll', .', .') urn t.wo rinll;H. J)('lim' hi nary operations + and
lin the Cart.c>sian procluet Il X Il' as follow!!:
The following relations show the triple 0). +.. ) to be an ideal of the ring of
integers (Z. +, .):
no - rna = (n - m)o.
m(no) = (mn) 0, n, m eZ.
In particular, since (2) = Z the ring of even integers (Z +..) is an ideal of
(Z, + .. ).
3-2 IDEALS AND QUOTIENT RINGS 157
id<>al (R,+, .)
clearly belongs to it; thus, the set (8) exists and is such that
8 ~ (8). Theorem :l-S leads directly to the following result.
Theorem 3-9. The triple 8), +,.) if! an ideal of the ring (R, +, .), known as
the ideal ymeraled by Ihe set S.
It is not.eworthy that whenever (I, +, .) is any ideal of (R, +, .) for which
S ~ I, t.hell (8) ~ I. III view of this, olle frequently speaks of S), +,.) ILS
being the smallesl ideal to contain the set S. An ideal generated by a single
ring clement, say a, is called a principal ideal and is designated by a), +,.).
A natural undt'rtaking is to determine the precise form of the members of
(S). If we impot) the requirement that (R, +, .) he a commutativo ring, it is
a fnirly Hilllpl!' matter to e/wek Umt (8) is givell by
where the symbol 1: ill<iieaies a fillite sum with one or more terms. The proof
is left to the rend<>r as all exprcil:!C.
In the eal:!C of the prineipal ideal a), +, .),
this description of (8) reduces to
III actual fnet, the elements 1" a (r E R) compril:!C the set of elements of an
ideal of (R, +,.) evell whell the ring doesllot possess an identity; the difficulty,
however, iH that thil:! i<ieuilleed not l'ontnin a itself. The next theorem formalizes
these idells.
Theorem 3-10. If (R, -f ,.) is 1\ ('ommutative ring with identity and a E Il,
then thc prineipal ideal a), +, .) gencrated by a is such that
(a) = {r a 11' E R}.
The prirll'ipnl ilil'all:! are the only il\l'als of the ring of integers, as the following
theorem will :;how.
Theorem 3-11. If (/, t-,.) is an idl'al of the ring (Z, +, .), then 1= (n)
for some Ilollll('gative integer n.
3-2 IDEALS ANO QUOTIENT RINGS 159
Proof. If I = {O}, Ill(' th{'orpm is trivially tmc, for till' zpro ideal ({O}, +, .)
il:! the print"ipal idml g('ncrateu by O. Suppose then t hal I uoel:! not (~()nsist of
the zero clement alone. :\'ow, if mEl, -m also belong;; to I, so that I contains
pOliit.iv(' illtq(crli. Let It deHignlLte the lellst pmdtiv(' intlger in I. As (I, +,.)
il:! lUi ideal, eaeh intel!;rul multiple of n must. be in I, UlIlt is, (n) ~ I.
On the otIl('r hand, any integer k E I may be exprel-lscd as k = qn + r,
whpre q, 1" E Z and () ~ r < n. Hinee k and qn are nwmbers of I, it follows
that k - qn = TEl. Our definition of the integer n implies T = 0, and
consequently k = qn. Thus every member of I is a multiple of n implying that
I ~ (n). The two inclu!!ions !!how I = (n), eompleting the argument.
Actually, It !lllll'h shortl'r proof of thl' fOll'l(oilll( 'IHUIt. ('lIuld hI' lIiltnilll'(\
using Tlworclll :!-:!4; till' proof, as given, hILS till' advantage of hl'ing HIlf
('ontainI'd.
By It principal ideal Tina is meant a commutative ring with identity in which
every ideal is principal. It is apparent from Theorem ;~-11 that the ring of
integers constitutes IL principal ideal ring.
Now, I:!Uppose that ai, a2, ... , an arc nonzero c1ement.s of (R, +, .),
a com-
mutative ring with identity. An clement a E R il:! said to be a common multiple
of aI, a2, ... , an provided a is a ring multiple of Clwh of t.hese. For instance,
the products al . a2 ... an and - (al . a2 ... an) are both common multiples
of all a2, . . , an. We shalL cull the element a It least common multiple of
all a2, .. ,an if (1) a is It common multiple of these clements and (2) any
other common multiple of a" a2, ... , an is a multjp!l~ of a liS well.
We make immediate use of this terminology to prove the next theorem.
Theorem 3-12. Let aI, a2, ... , an be nonzero elements of It principal ideal
ring (R, I,). Theil
(n(ai),+,) = a),+,),
which shows (b) ~ (aj) for eaeh value of i. TheJ"{fo!"1' (b) ~ n(aj) = (a) aAd
accordingly b must be a multiple of a. Our argllntellt. establi!;hes that a is a
least common mult.iple of ai, a2, ... , a".
To illustrate thi!; theorem, eonsider the prinl"ipal ideals 4), and +, .)
6), +,.) generated by the integer!; 4 and 6 in the ring (Z, +, .). The reader
160 RING THEORY 3-2
We omit the details of the proof and merely point out that the zero element
of (R/I, +,.) is the coset 0 + I = I, while -(a + I) = (-a) + I.
Example 3-17. In the ring (Z, +,.) of integers, consider the principal ideal
n), +, .), where n is a nonnegative integer. The cosets of (n) in Z take the
form
a+ (n) = {a+ 1m I k eZ} = [a].
That is, the cosets are precisely the congruence classes modulo n. It follows
from the definition of coset addition and multiplication that the quotient ring
of Z by (n) is merely the ring of integers modulo n:
(Z,., +,., .,.) = (Z/(n), +, .).
A homomorphism between two rings (R, +,.) and (R', +', .'), as one might
expect, is a function f: R -+ R' which preserves both ring operations. This
amounts to applying the familiar homomorphism concept to the additive
groups (R, +) and (R', +'), and to the multiplicative semigroups (R,) and
(R', .'). The precise definition follows.
Definition 3-12. Let (R, +,.) and (R', +',.') be two rings and f a function
from R intO R'; in symbols, f: R -+ R'. Thenf is said to be a (ring) h0mo-
morphism from (R, +,.) ~to (R', +',.') if and only if
la(x) = a X a-I
Theorem 3-14. Let I be a homomorphism from the ring (R, +,.) into the
ring (R', -1-', .'). Then the following hold: .
1) 1(0) = 0', where 0' iR the zero element of (R', +', .').
2) I( -a) = -f(a) for all a E R.
3) The triple (f(R), +', .,) is a subring of (R', +', ").
If, in addit.ion, (R, +,.) !\nd (R', -+', .') a.rc rings with identity elements 1
and 1', respectively, and I( R) = R', then
4) 1(1) = 1',
Ii) J(a- I ) ~ J(U)-I fll/' I'llI'h invt'rtihll,I'II'llI!'n\. 11 Ell.
Two COlllllwntll regarding pnrt (4) of the theorem are in order. First, it is
evident that
J(a) .' l' = J(a) = J{a 1) = J(a) .' J(I)
for all a in R. From this, one might be tempted to (incorrectly) invoke the
cancellation law to mme\llIle that I( 1) = 1'. What iR uctunlly required is the
fact thut Il\ul\.iplicntive idl'ntit.ies lUI' uni1lue.
Secondly, if thl' hypo\.h!'sis tlmt. I mups ont.o t.he 1I('t. Il' ill omitted, then it
can only be inferred t.lmt J(1) is t.he idl'ntity of the subring (I(R), f-', "). Tltl'
:J-2 IDEALS AND QUOTIENT RINGS 163
clement fell need not serve as an identity for the entire ring (R', +', .'); in
faet, it may very well happen thatf(I) ~ 1'.
We also observe, in passing, that by statement (2),
I(a - b) = I(a) - feb), a, b E R.
where, as usual, 0' designates the zero clement of (R', Ignoring the+', ").
multiplication operations in the rings, this is just the usual definition of the
kernel of a homomorphism between the additive groups (R, +) and (R', +').
As before, f is a one-to-one mapping if and only if ker (f) = {O}.
For our analogy between ideals and normal subgroups to be meaningful, one
would anticipate that the kernel of a homomorphism is an ideal. This is indeed
the content of the following theorem.
Theorem 3-15. If , is a homomorphism from the ring (R, +,.) into the
ring (R', +', .'), then the "triple (ker (f), +, .)
is an ideal of (R, +, .).
:~-14, 0 E ker (f), liO that the kernel is Ilonempty.
Proof. By part (I) of Thcor(!m
Now, any two elements a and b in ker (f); by definition, f(a) = 0' =
COli Hider
feb). Since any homomorphic mapping between rings preserves differences, it
follows that
f(a - b) = f(a) - feb) = 0' - 0' = 0',
Since (ker (f),J , ,) is then an ideal of (Z, +, ,), it follows at once from
164 RING THEORY 3-2
for some nonnegative integer m. A moment's reflection will convince the reader
that the integer m must be the characteristic of the ring (R, +, .). In other
words, the ideal (ker (J), +, .) is nothing more than the principal ideal gener-
ated by the characteristic of (R, +, .).
In agreement with our previous use of the term, two rings (R, +, .) and
(R', +', .') are said to be iBOmOrpilic if there exists a one-to-one homomorphism
from the ring (R, +,.) onto the ring (R', +', .'). We indicate this by writing
(R, +,.) ~ (R', +', .'). The last example, for instance, implies that any ring
(R', +',.') with identity which is of characteristic zero contains a subring
isomorphic to the integers; more specifically, (Z, +,.) ~ (Zl, +', .'), where 1
is the identity element of (R', +', .').
As we have seen, many aspects of ring theory are considerably simplified
when a multiplicative identity exists. The next theorem shows that there is
no real loss in generality in assuming the presence of such an element, for
every ring is isomorphic to a subring of a ring with identity. Because this
result is often phrased differently, we require another definition.
DeflnlHon 3-13. A ring (R, +,.) is imbtdded in a ring (R', +', .') if there
exists some subring (S, +', .') of (R', +', .') such that (R, +,.) ~(S, +', .').
Theorem 3-16. Any ring can be imbedded in a ring with identity.
Proof. Let (R, +, .) be an arbitrary ring and consider the Cartesian product
RXZ = {(r, n) IrE R, n E Z},
natI(a) = a+1
for all a E R. Paralleling our work in group theory, it is possible to prove:
Theorem 3-17. The natural mapping nati is a homomorphism from
(R, +,.) onto (RII, +,.) with kernel I.
Proof. By the corresponding theorem for groups, Theorem 2-42, natl is known
to be a group homomorphisni of (R, +) onto (RI I, +) such that ker (natI) = I.
Thus, we need simply show that products are preserved by nat/. This follows
immediately from the fact that, for any a, b E R,
Theorem 3-18. If f is a homomorphism from the ring (R, +,.) onto the
ring (R', +', .'), then (Tl/ker (f), +,.) ~ (R', +', .').
Proof. Just as in the group case, we define a function 1: R/ker (f) --+ R', the
induced mapping, by taking 7(a + ker (f) = f(a); it is this function which
establishes the desired isomorphism. From the proof of Theorem 2-47, we
already know that (R/ker (1), +) ~ (R', +') by]. To complete the argument,
it remains only to settle the question of whether 1 preserves the multiplication
operation in (R/ker (f), +, .); but this is straightforward:
Example 3-23. For a simple, but useful, example illustrating some of these
ideas, eonRider the rin~R (Z., +., .) and (Z2, -h, '2)' We define the function
I: Z. --+ Z2 as follows:
The operation tables for the quotient ring (Z./ker (f), +, .) are as shown:
Theorem 3-18 asserts that the Ryst('m (Z./ker (f), +, .) must be isomorphic
to the ring of integers modulo 2; indeed, this is reasonably evident from the
nature of the foregoing tables.
As an applicntion of Theorem 3-JR, we propose to show that each homo-
morphism onto the ring of integers (Z, +, .) is entirely determined by the set
a-2 IDEALS AND QUOTIENT RINGS 167
where it 1l1lSIIInes t.he value O. Contrll.!lt this with the case of groups: both
iz and -iz nrc homomorphisms from the additive group of integers (Z, +)
onto itself. The kernel of each of these mappings is {O}, but clearly iz "'" -iz.
We begin hy establishing a lemma which is of independent interest.
Lemma. The only nontrivial homomorphism from the ring of integers
(Z, +,.) into itself is the identity map iz.
Proof. Consider any homomorphism f having the asserted properties. Because
each positive int.eger n may be written II.!! I + + ... +
I 1 (n summands), the
operation-preserving nature of f implies fen) = nf(l) for each n E Z+. On
the ot.her hand, if n ill lUI arhitmry negative integer, -n E Z+, and accordingly,
fen) = nf(l)
g = g natker(Q)
that the funetiolls f and g arc themselves identical.
168 RING THEORY 3-2
g(a + b) = Ia + b) . u) = I(a u + b . u)
= I(a u) +' I(b u) = g(a) +' g(b).
As a preliminary step to demonstrating that g also preserves multiplication,
note that
Ja' b) u 2 ) = Ia b u) . u) == Ia' b) u) .' I(u) == Ia b) . u).
In licht of this, we are able to conclude
g(a b) = Ia' b) u) = Ia b) u2 )
= Ia u) (b u = I(a u) .' I(b u) = g(a) .' g(b).
The third equality is justified by the fact that u E cent R, hence commutes
with b.
3-2 IDEALS AND QUOTIENT RINGS
PROBLEMS
1. Determine all ideals of (Z12. +12. '12). the ring of integers modulo 12.
2. Show by example that if (11,+.') and (12,+,) are both ideals of the ring
(R.+ .). then the triple (11 U 12,+.') is not necessarily an ideal.
3. For any ideal (1.+,,) of the ring (R,+, .). define C(l) to be the set
C(I) - {r e R I r . a' - a . reI for all a e R}.
Determine whether (C(l),+.) forms a subring of (R,+ .).
4. If (11,+,') and (12,+,') are ideala of the ring (R,+,') such that 11 n 12'" {O},
prove a b = 0 for every a E 11, bE 12.
5. a) Verify thanlienng--orreal numbers (R', +, .). is a simple ring.
b) Prove that for each n e Z+, tho rin, (Z .. ,+..,'..) of intepre modulo" It a
principal ideal ring.
6. Let (1,+,,) be an ideal of the ring (R,+.) and define
ann 1 :- {r E R ! r a - 0 for all a E l}.
Prove that the triple (ann 1,+.,) constitutes an ideal of (R,+ ). called the
annihilator ideal of 1.
7. Let (1,+.') be an ideal of (R.+, .), a commutative ring with identity. For an
arbitrary element a in R, the ideal generated by 1 U {a} is denoted by
(I, a), +, .).
Assuming a Ii! I. show that
(I,a) - {i+ra!iEI.rER}.
8. Suppose (II, +,.) and (12,+,') are ideals of the ring (R.+, .). Define
/11+12 - {a+blaElt,bE I 2}.
Show that (11 + 12,+.) is an ideal of the ring (R.+ .); in Tact, (11 + 12.+,)
is the ideal generated by 11 U 12.
170 RING THEORY 3-2
9. In the ring of integers, consider the principal ideals n),+,) and m),+,)
generated by nonnegative integers nand m. Using the notation of the previous
two problems, verify that
n), rn) = rn), n) = (n) + (m) = ({m, n}) - (d),
fG(x) = a x, xE R.
If F H denotes the set of all sllch functions, prove the ring analog of Cayley's
theorem:
a) The triple WH, +..) +
forms a ring. where denotes the usual pointwise addition
of funct.ions and. denotes functional composition.
b) (R,+,) ~ (FH,+, .). [Hint: Consider the mappingf(a) = fG.]
15. Let CR, +, .)be an arbitrary ring and (R X Z, +', .')
be the ring constructed in
Theorem 3-16. Establish that
a) (R X 0, +',.') is an ideal of (R X Z, +', .'),
b) (Z, +, .)
~ (0 X Z, +', .'),
e) if a is an idempotent element of R, thl'n the pair (-a, 1) is idempotent in the
ring (R X Z, +', .')
and (a, 0) .' (-a, I) "" (0,0).
3-2 IDEALS AND QUOTIENT RINGS 171
16. Suppose (/,,+,.) and (/2,+,') are both ideals of the ring (R,+, .). Define the
set 1,12 by
where:E denote!! a finitl~ Rum with olle IIr IlIOfe term",. Prove that (h, 12, +, .)
is an ideal of (R, -+, ').
17. GiVl'II that. (/. -I, .) iH all idllal of the ring (R, -/-, .), Khow that
a) whenevllr (ll, -I,,) ill Cllllllllutative with idellt.ity, then 1111 ill the quotiont ring
(Rl1, +, .),
h) the ring (Rl1, +,.) lIIay have ,livitlorll of Z(,fO, I,vell though (R, -t,.) does
not hav" a.ny,
c) if (R, +,.) is a principal ineal rinJ!;, then so is the quotient ring (RIT,+, .).
IH. IAlt (ll, I , .) till " (!IIlIIlIIutu.tive rillK wit.h hlt'lltity u.nl! I"', N ,It'notl' the 81~t of
nilpotent clements of R. Verify that
a) the triple (N, +,.) is an ideal of (R, +, .). [lJint: If an = b'" = 0, consider
(a - b)n+"'.)
b) the quotient ring (RIN, +,.) has no nonzero nilpotent elements.
19. Assume (R, +, .)
is a ring with the property that a 2 +
a E cent R for every
element a in R. Show that (R,+, .) is a commutative ring. [Hint: Make use
of the expression (a b)2 + + +
(a +
b) to prove, first, that a' b b . a lies in the
center.)
20. Illustrate Theorem 3-18 by considering the rings (Zo, +6, '0), (Za, +a, 'a), and
the homomorphismf: Zo -> Za defined by
f<O) = f(3) = 0, 1(1) = f(4) = 1, f(2) = f(5) = 2.
21. Let (S, +, .) be a subring and (1, +, .) an ideal of the ring (R, +, .). Assuming
S n 1 = {O}, prove that (S, +, .) is isomorphic to a subring of the quotient ring
(Rl1, +, .). [llint: Use the mapping f(a) = a + I, where a E S, in conjunction
with Theorem 3-18.)
22. a) Let f be a homomorphism from the ring (R, +,.) into the ring (R', +', .').
Given that a E R is nilpotent, show its image f(a) is nilpotent in R'.
b) Suppose (R, +, .)
is a ring which has no nonzero nilpotent elements. Deduce
that all the idempotent elements of R bcJong to the center. [llint: If a2 = a,
then (a r . a - a r)2 = (a r . a - r' a)2 = 0 for all r E R.J
23. A ring (R, +, .)
ill Raid to be the direct BUm of the two ideal!! (/ .. +, .)
and (12, +, .),
indicated by writing R = 11 ~ 12, if R = 11 + 12 with h n 12 = {O}. Given
R = I, ~ 12, show that
a) It 12 = 12' I, = {O},
b) every element a E R can be uniquely expressed as a sum
where
24. Let (R, +,.) hI' a commut.ative ring with identity and a E R he an idempotent
whil'h ill dilT('rl'nl from II IIr I. Prove that (R,I , .) is thl' direct 1111111 of the principal
ideals a),+,') and (I - a),+,.): R = (a) ~ (I - a). [llint: Utilize the fact
a' (I - a) = 0.)
170 RING THEORY 3-2
9. In the ring of integers, consider the principal ideals n),+,') and m),+,')
generated by nonnegative integers nand m. Using the notation of the previous
two problems, verify that
n), 711) = m), n) = (n) + (m) ... ({m, n}) - (d),
where d ill the grl'atl'Rt common rlivisor of nand m.
10. ConMilier t1w rinJl; (1'(:\), ~, n) of Example 3-4. For a fixed subset S ~ X, definn
the function I: I'(X) -+ P(X) by
I(A) == An 8.
Show that I is a homomorphism and determine its kernel.
II. Utilize Prohll'll1 10, HI'(:tion 3-1, to Hhnw that in an integral domain (H, +, .) or
characteristic p > 0, the mapping
I(a) = a", aEH,
is a homomorphism of (R, +, .)
into itself.
12. Given that f is a homomorphism from the ring (R, +, .)
onto the ring (R', +', .'),
prove that
a) if (1, +, .)
is an ideal of (R, +, .),
then the triple (f(l), ,I) is an ideal of+',
(R', +', .'),
b) if (I', +', .') is an ideal of (R', +', ,I), then the triple (f-I(1'), +,.) is an ideal
of (H, +,.) with ker (f) ~rl(1'),
c) if (H,+,') is a principal ideal ring, then the same is true of (R',+', ,f). [Hint:
For a E H, f( (a = (f(a) ).J
13. Let f be a homomorphism from the ring (H, +, .)
into itself and 8 be the set of
elements that are left fixed by f:
8 = {a E R I f(a) = a}.
Establish that (8, +,.)
is a subring of the ring (R, +, .).
14. For a fixed element a of (R, +, .), a ring with identity, define the lelt-multiplication
functionf,,: H -+ R by taking
f,,(z) = a z, zE R.
If Fit denotes the set of all such functions, prove the ring analog of Cayley's
theorem:
a) The triple Wit, +, 0) forlns a ring, where + denotes the usual pointwise addition
of funct.ions and denotes functional composition.
0
wh('re :E llenott~1'I a finitl~ slim with 0111' IIr 1II0fll terlllH. Prove that (Ii 12, +, .)
is an ideal of (R,-t, ').
17. Give'll that. (/,1,') iH nil ideal of the ring (H,+, .), Khow that
a) whcnevI~r (Ii, -I,,) ill commutative with idclltity, then 1111 ill the quotiont ring
(HI I, +, .),
h) the ring (RI I, +,.) may hav{, divisors of Z(,fO, nVCJl though (H, +,.) dOCB
not have, II.ny,
c) if (R, +, .) is a principal icieal rin!!:, then so is the quotient ring (RII, +, .).
I~. (A,t (Ii, I , .) he a e'"lIIlIlutat,ive rinK with hit'ntity and II!\. N ,Ie'uotl' tho set of
nilpotent clements of H. Verify that
a) the triple (N,+,') is an idell.l of (H,+, .). [/lint: If an = b'" = 0, consider
(a - b)n+ .... ]
b) the quotient ring (HIN, +, .)
has no nonzero nilpotent elements.
19. Assume (R, +, .) +
is a ring with the property that a2 a E cent R for every
element a in H. Show that (H, +,.) is a commutative ring. [Hint: Make use
of the expression (a b)2 + + +
(a +
b) to prove, first, that a' b b . a lies in the
center.]
20. II1ustrate Theorem 3-18 by considering the rings (Z(l, +6, '6), (Za, +a, '3), and
the homomorphism f: Z6 --+ Za defined by
f<O) = f(3) = 0, f(l) = f(4) = 1, f(2) = f(5) = 2.
3-3 FIELDS
In the preceding two sections, a hierarchy of special rings has been obtained
by imposing more and more restrictions on the mUltiplicative semigroup of a
ring. One might be tempted to require that the multiplicative semigroup
actually be a group. Such an assumption would be far too demanding, for this
situation can only take place in the trivial ring consisting of the zero element
alone. It turns out, however, that there do exist rings in which the nonzero
elements form a group under mUltiplication. This leads us to the notion of a
field.
DefInition 3-15. A ring (F,+,.) is said to be a field provided the pair
(F - {O},) forms a commutative group (the identity of this group will be
written as 1).
It should. be evident that any field (F, +, .) must contain at least one non-
zero element, for F - {O} is nonempty, being the set of elements of a group.
Moreover, since a 0 = 0 = O a for every a e F, all the elements of F com-
mute under multiplication and not merely the nonzero elements. In brief, a
field is a commutative ring with identity in which each nonzero element has
an inverse under multiplication. Observe also that if the zero element were
allowed to possess a multiplicative inverse, then 0 = a 0 = 1 for some element
a e Fj this would imply F =' {O}, contrary to our convention that any ring
with identity contains more than one member. -
As before, to distinguish the two inverses of a nonzero element a in F, we
shall denote the mUltiplicative inverse by a-I and use the notation -a for its
inverse relative to addition.
Example 3-24. Both the systems (R', +,.) and (Q, +, .), where + and
indicate ordinary addition and multiplication, are examples of fields.
Example 3-25. Let F be the set of real numbers of the form a + by'3, with
a and b rational: F = {a + by'3 I a, b e Q} . It is straightforward to check
that the triple (F, +,.) is a commutative ring with identity (see Example 3-3).
The additive and multiplicative identity elements in this case are
0= O+oy'3,
To show that (F, +,.) is a field, we must verify that each nonzero element of
F has an inverse belonging to F. Suppose then that a + by'3 e F, where
a and b are not both zero. Under these circumstances, a 2 - 3b2 ,.s 0, for
otherwise y'3 would be rational. This means that
a-bV'3
1
a+bV3a-bV3
,
a -b _ r.;
::= a2 _ 3b 2 + a2 _ 3b 2 v 3 e F.
3-3 J'IBLD8 173
Since a/(a' - 3b') and -b/(a' - 3b') are both rational numbers, the resulting
inverse docs have the required fonn to be a member of F.
Note that if a and b were restricted simply to the set of integers, then (F, +, .)
would no longer be a field, for then the element
a -b
a' - 3b' + a' - 3b'
va3
would not necessarily lie in F.
Example 3-26. Consider the set 0 = R' X R' of ordered pairs of real num-
bers. To endow C with the structure of a field, we define addition and multi-
plication by
(a, b) + (e, d) = (a + e, b + d),
(a, b) (e, d) = (ae - bd, ad + be).
The reader may verify without difficulty that the triple (0, +, .) is a com-
mutative ring with identity. Here the pair (1,0) serves as the mUltiplicative
identity and (0,0) is the zero element of the ring. Now, suppose (a, b) is any
nonzero member of C. Sioce (a, b) ~ (0,0), either a ~ 0 or b ~ 0, 80 that
a2 + b' > OJ thus
(a, b) ( a 2
a
+b2 ' a2
-b)
+ = (a +
b2 + b -00
a +
+ 00) =
a2
2 2
b" l b2 (1,0).
This shows that the nonzero elements of C have inverses under multiplication,
proving the system (C, +, .) to be a field.
The field (C, +, .) contains a subring which is isomorphic to the ring of real
numbers. For if
R' X 0 = {(a, 0) I a e R'},
it follows that (R', +,.) ~ (R' X 0, +,.) via the mapping I defined by
I(a) = (a, 0), aeR'.
(Verify this!) As the distinction between these systems is one only of notation,
we customarily identify the real number a with the corresponding ordered pair
(a, 0) j in this sense, (R', +, .) may be regarded as a subring of (C, +, .).
The definition of the operations + and enables us to express any element
(a, b) E C a s '
(a, b) = (a,O) + (b,O) . (0, I),
where the pair (0,1) is such that (0,1)' = (0,1) (0,1) = (-1,0). lntto-
174 RING THEORY 3-3
with i 2 = - 1. In otlH'r WOrthl, the field (e, +, .) lUi defined above is nothing
more than the familiar complex number system.
The following th('ormn shows that a field is wit.hout divisors of zero, and
consequently it! a system in which the cancellation law for multiplication holds
(see Theorem 3-5).
Theorem 3-21. If (P,! , .) is a field ILnd a, b E F with a' b = 0, then cithcr
a = 0 or b = O.
Prooj. If a = 0, the th('on'm is alreudy est.ublish(!d. So let us suppose that
a ;r6 0 and prove that b = O. By the definition of a field, the element a, being
nonzero, must have a multiplicative inverse a-I E F. The hypothesis a b = 0
then yields
o= a-I. 0 = a-I. (a b) = (a-I. a) . b = 1 . b =-b,
as desired.
Since a field is a commutative ring with identity, and we have just proved
that it contains no divisors of zero, we conclude that any field is an integral
domain. There obviously are integral domains which are not fields; for instance,
the ring of integers. However, an integral domain having a finite number of
clements must necessarily be a field.
Theorem 3-22. AllY finite integral domain (R, +, .) is a field.
Pl'Ooj. Suppo!!C at, a2, ... , an are the members of the set R. For a fixed non-
zero clement a E R, consider the n products a alt a a2, ... , a an. These
products are all distinct, for if a ai = a aj, then aj = aj by the cancellation
law. It follows that each clement of R is of the form a ai. In particular,
there exililts some aj E R stwh t.hat a aj = 1; since multiplication is com-
mutative, we thus have ai = a-I. This shows that every nonzero element of
R is invertible, so (R, +, .) is a field.
It was previously seen that for each positive integer n the system (Zn' +.. , '..)
is a commutative ring with identit.y. Our next result indicates for precisely
what values of n this ring is a field.
Theorem 3-23. The ring (Z", 1-", 'n) of integers modulo n is a ficld if and
only if n is 1\ prime number.
3-3 FIELDS 175
Prooj. We first show that if n is not prime, then (Zn, +n, 'n) is not a field. Thus
assume n = a b, where 0 < a < nand 0 < b < n. It follows at once that
although hoth [a) .,& (0), [b) .,& [0]. This means that the system (Zn, +n, 'n) is
1I0t lUI intl'p;ral domain, and hcnce not a fidd.
On Ow ol,hN hand, SIlPJlOSP that n is It prime number. To show that
(Z", -I n, ',,) iH a fidd, it. sufliccs to pl'ove here that each 1l0llZt'ro e1emcnt of Z ..
has a multiplicative inverse in Zn. To this end, let raj E Zn, where 0 < a < n.
Aeeording to Theorem 1-1:~, Rinee a and n have no common factors, there exist
intcgerH r and II lmdl tlmt
ar+n8=1.
Thill implies tlmt.
Now a), +,.) cannot be the zero ideal, since a = a 1 E (a), with a"& O.
It follows from the hypothesis that the only other pORsibility is a), = +, .)
(R, +, .);
that is, (a) = R. In particular, since 1 E (a), there exists an element
,. E R for whieh i'. a =, 1. Multiplication is eommulative, so that,. = a-I.
Hence e:wh nonzero {'Iement of R hn" a lIlult.iplientive inverse in R.
176 RING THEORY 3-3
F' = {a b- 1 I a, b E Rj b ~ O},
then the triple (F', +,.) fonns a subfield of (F, +,.) such that R ~ F'. In
fact, (F', +,.) is the smallest subfield containing R.
Proof. Note first that the definition of the set F' is meaningful; indeed, if
a, b E H with b ~ 0, the product a b- 1 must be in F by virtue of the fact
(F, +,.) is a field. Since 1 = 1.1- 1 E F', F' ~ {O}. Now consider two
arbitrary elements x, y of F'. We then have
a d = h e, e f = d e.
a . d . f = h . e . f = b d e,
[a, b] = [a', b/] and [e, d] = [e ' , d/l. From the equations
The next lemma estabHs}les the algebraic nature of the triple (F, +', ./).
Lemma. The system (F, +',.') is a field, generally known as the field of
quotients of the integral domain (R, +, .).
Proof. It is an entirely straightforward matter to establish that the triple
(F, +', .')
is a commutative ring. We leave the reader to make the necessary
verifications at his leisure, and merely point out that [0, bl serves as the zero
element while [-a, b] is the negative of [a, b].
That the equivalence class [a, a], where a is any nonzero element, constitutes
the multiplicative identity is evidenced by the following:
Proof. Consider the subset F' of F consisting of all elements of the form [a, 1],
where 1 is the multiplicative identity of (R, +, .):
F'- {[a,I]laeR}.
It is readily checked that the triple (F', +', .') is a subring of (F, +', .') and,
in actual fact, is an integral domain. Now, let /: R -+ F' be the onto mapping
defined by
lea) - [a, 1]
Accordingly, (R, +,.) ~ (F', +', .') under f, and the proof is complete.
Several remarks are in order. First, note that any member [a, b) of F can be
written in the form
[a, b) = [a, 1] .' [1, b) = [a, 1] .' [b, 1]-1.
Since the systems (R, +, .) and (F', +',.') are isomorphic, one customarily
identifies the element [a, 1] E F' with the element a of R. The above equation
then becomes [a, b] = a' b- 1 The point is this: we may now regard the set
F as consisting of all quotients a.' b- 1 , with a and b '" 0 in R. It should also
be observed that for any a '" 0,
[a, 1] .' [b, a] = [a b, a] = [b, 1].
Again writing [a, 1] simply as a, we infer that the equation a' s == b always
has a solution in F, namely s = [b, a) = b' a-I.
A final fact of interest is that the field of quotients (F, +" .') is the smallest
field in which thc integral domain (R, +, .) can be imbedded, in the sense that
any field in which (R, +, .) is imbeddable contains a subfield isomorphic to
(F, +', .') (Problem 14 of this section).
The field of quoticnts constructed from the integral domain (Z, +, .) is, of
course, the rational number field (Q, +, .).
Definition 3-17. A field which does not have any proper subfields is called
a prime field.
Example 3-27. The field of rational numbers, (Q, +, .),
is a prime field. To
see this, suppose (F, +, .) is a subfield of (Q, +, .) and let a e F be any non-
zero element. Since (F, +, .) is a subfield, it must contain the product
3-3 FIICLDS 181
yielding the contradiction that the field (F, +, .) has divisors of zero. (This
result is not entirely unexpected, because n is the characteristic of (F, +, .)
and as such must be prime.)
The preceding discussion indicates that two possibilities arise: either
\ 1) (f(Z), +, .) ~ (Z/(p), +, .) == (Zp, +p, .p) for some prime p, or
2) (f(Z), +, .) ~ (Z/(O), +, .)
= (Z, +, .).
. Turning to a closer analysis of these C8.ses, suppose first that (f(Z), ~ +, .)
(Zp, +p, .p), with p prime. Inasmuch as the ring of integers modulo a prime
forms a field, the subring (f(Z), +, .) must itself be a field. But (F, +, .) con-
tains no proper subfields. Accordingly, feZ) = F and (F, +, .)
~ (Zp, +", .,,).
182 JUNll TIfI<:OltV 3-3
Next, consider th(~ sit.uution (f(Z), +, .) ~ (Z, +, .). Under these circum-
stances, the subring (f(Z), +, .) is an integral domain, but not a field. Theorem
3-26, in conjunction with the hypothesis (F, +, .) is a prime field, then implies
F= {ab-1Ia,be/(Z);b;><!O}
[(nl) (ml)-t In, me Z; m ;><! O}.
PROBLEMS
I. If, + and denote ordinary addition and multiplication, for which of thp. following
sets It' is (F, +, .)
1.1. field?
1.1.) F = {a - bv'21 a, b e Z}
b) F = {a+ b01 a, beQ}
+ +
c) F = {a b0 c~ I a, b, ceQ}
+, .),
2. In a fi!'ld (F, Hhow t.hat 1IH1 !''Iuation (,2 = a impli('!! eithl'r a - 0 or a -
3. Define two binary operations * and on the set Z of integers by letting
0
a* b = a +b - 1, a,bei.
. vated to some extent by the fact that many of the standard examples of ring
theory have this property. A,further reason, which is perhaps more important
from the conceptual point of view, is that the moat satisfactory and complete
results occur here. We begin by making the following definition.
DeflnlHon 3-18. An ideal (I, +,.) of the ring (R, +,.) is a maximal ideal
provided I jI6 R and whenever (J, +,.) is an ideal of (R, +,.) with
I C J ~ R, then J = R.
Expressed rather loosely, an ideal is maximal if it is not the whole ring and
is not properly contained in any larger nontrivial ideal. The only ideal to
contain a maximal ideal properly is the ring itself.
Assume, (or the moment, that (1, +, .) is a proper ideal of the ring (R, +, .)
and a is an element which does not belong to 1. Then the ideal I, a), +, .),
the ideal generated by the set I u {a}, is such that I C (I, a) ~ R. These
inclusions imply that if (I, +, .) were a maximal ideal, the set (I, a) must be
all of R, (I, a) = R. On the other hand, suppose (J. +,.) is an ideal of
(R, +, .) for which I C J ~ R. If a is an element of J not in I, then
I C (I, a) ~ J. The condition (I, a) = R would therefore force J = R and we
could conclude that (I, + .. ) is a maximal ideal. Summing up, the ideal (I, +, .)
is a maximal ideal of (R, +, .) if and only if I jI6 R and (I, a) = R for every
element a ~ I. This fact will prove quite helpful a little later.
To illustrate the concept, let us show that in the ring of intege.. the maximal
ideals correspond to the prime numbers.
Theorem 3-29. Let (Z, +, .)be the ring of integers and n > 1. Then the
principal ideal n). +,.) is maximal if and only if n is a prime number.
Prool. First, suppose n), +,.) is a maximal ideal of (Z, +, .). If the integer
n is not prime, then n = nln2, where 1 < nl ~ n2 < n. This implies the
ideals n1), +, .)and n2), +, .)
are such that
Then (R, +, .) is a commutative ring with identity (see Example 3-5). Consider
the set M of functions in R which vanish at 0:
(M, f) ;2 (i, f) = R,
and therefore (M, f) = R for every Ie M [here (i, f) designates the elements
of the ideal generated by i and I; that is, (i, f) = {r. i + 8 I 1 r, 8 E R} J.
Our immediate goal is to obtain a general result assuring the existence of
suitably many maximal ideals. As will be seen presently, the crucial step in the
proof depends upon the maximal element principle, or what is commonly termed
Zorn's Lemma. It would take us somewhat far afield to do much more than
merely formulate thislcmma as an axiom; the reader who wishes to pursue the
topic further is directed to the comprehensive discussion in [38].
For ease of reference, let us recall that by a chain is meant a collection e
of sets such that A, BEe implies either As;;; B or B S;;; A. We now state
Zorn's Lemma in a form best suited to our present needs.
Zorn '. Lemma. Let (1 be a nonempty family of subsets of some fixed set
with the property that for each chain e in (1, the union ue also belongs to (1.
Then (1 contains a set which is maximal in the sense that it is not properly
contained in any member of (1.
The significant point, needless to say, is that this lemma asserts the existence
of a certain maximal element without actually giving a constructive process for
findjng it.
Theorem ~30. (Krull-Zorn). In a commutative ring with identity, each
proper ideal is contained in a maximal ideal.
Proof. Let (I, +,.) be any proper ideal of (R, +, .), a commutative ring with
identity. Ddino n family of subsets of R by taking
(1 = {J 1 IS;;; J; (J, +,.) is a proper ideal of (R, +, .)}.
This family is obviously nonempty for I itself belongs to (1.
Now, consider an arbitrary chain {l;} in (1. Our aim, of course, is to establish
that ul; is again a member of (1. Notice first that UI; ."" R, since 1 e I, for
any i. Next, let the elements a, bE UIi and r E R. Then there exist indices
i ~nd j for which a Eli, bE Ii' As the collection {Ii} forms a chain, either
Ii S;;; Ii or else Ii S;;; Ii; say, for definiteness, Ii S;;; Ii' But (Ij, +,.) is an ideal,
ISH It! Nil 'I'll EOilY :J-4
even when t.his if< 1101. (xplicitly ml'nlioned. To he Rum, some of the s\lb~q\lent
material coulc! he prcsC'lItl'd without this additional restridion.
In a striet sl'n s<, , it. i:-; inC'orrect to speak of an ideal being eontained in a
maximal iilpnl. for til!' irwlusion aet.ultlly rl'fl'rs to t.he underlying sets of ele-
nH'nts. "Nevl'rtlH'lpss, the phra.sing is eonvenient and we ('ommit this inaccuracy
frcPly.
As an applieation of t.lw Krull-Zorn TheorC'm, we next I!;ive an elementary
proof of a somewhat slweial result; while the fad invoivpd is rather interCt!ting,
tll('rl' will be no oec'asion to nlake lise of it..
Theorem 3-31. In It ring; (fl, +, .) havinl!; eXIlC't.Iy one maximal ideal
(.11. ".), HIl' only idC'lIIpolC'nt. 1'1(,1II1'1I1H are 0 and I.
Proof. ASSUlll(' til(' theorl'lll iH false; t.hat. is, HUppOse t.here exists an idempotent
a E R with a 0, 1. The rplatioll a 2 = a implies a (I - a) = 0, so that
a and 1 -- a af(' zl'ro divisors. Hem'c, hy Problem S(d), Redion :~-l, neither
thl' elemenl a nor 1 - a iH illvl'rtible in R. But. this means the prin('ipal ideals
Ua), +,.) and (1 -
a), 1-,) are bot.h proppr icipalH of the ring (R, +,.).
As su(h, t1wy mllst. he ('ontained in (III, I,'), t.hc solt' maximal ideal of
3-4 CEH1'AIN SPECIAL WEALS 187
1=~+f"a
for suitable choice of i E I. f' E R. But then the difference 1 - f" a E I. This
is obviously equivalent to
1+ I = f'. a +I
+ I) . (a + I),
= (f'
Example 3-29. Consider the ring of even integers, (Z., +, .), a commutative
ring without identity. In this'ring, the principal ideal 4), +,.) generated by
the integer 4 is a maximal ideal.
The argument might be expressed as follows: If n is any element not in (4),
then n is an even integer not divisible by 4; consequently, the greatest common
divisor of nand 4 must he 2. By Problem 9, Section 3-2, we then have
4), n) = (2) = Z.,
so that the ideal generated by (4) and n coincides with the whole ring. This
reasoning shows that there is no ideal of (Z., +, .)
contained strictly between
4), +, .)and (Z., +, .).
Now note that in the associated quotient ring (Z./(4), +, .),
(2 + (4. (2 + (4 = 0 + (4).
The ring (Z./(4), +, .) therefore has divison; of zero and cannot possibly be a
field. The point we willh to make is that the asllumption of an identity element
is essential to Theorem 3-32.
Wo now shift our attention from maximal to prime ideals. Before formally
defining this notion, let us tum to the ring of integers (Z, +, .) for motivation;
specifically, consider the principal ideal p), +, .) generated by a prime p. If
ab E (p), where a, be Z, then ab = np for some integer n. Since the product
ab is divisible by p, either p divides a or else p divides b (corollary to Theo-
rem 1-4). This being so, it follows that either
for suitable choice of n., n2 E Z. The ideal p), +,.) thus haa the agreeable
property that whenever (p) contains a product at least one of the factors must
belong to (p). This observation serves to illustrate and partly to suggest the.
next definition.
Definition 3-19. An ideal (1,+,,) of the ring (R,+,') is a prime ideal if
for all a, b E R, a bEl implies either a E I or bEl.
+, .) are precisely the ideals
Example 3-30. The prime ideals of the ring (Z,
p), +, .), where p is a prime number, ,together with the trivial ideals
({O},+,') and (Z,+,).
Example 3-31. A commutative ring with identity (R, +,.) is an integral
domain if and only if the zero ideal ({O} , +, .)
is a prime ideal.
The prime ideals of a ring may be characterized in the following manner.
Theorem 3-33. Let (1, +,.) be a proper ideal of the ring (R, +, .). Then
(I, +,.) is a prime ideal if and only if the quotient ring (RIl, +,.) is an
integral domain.
CERTAIN SPECIAL IDEALS 189
Proof. First, take (1, +, .) to be a prime ideal of (R, +, .). Since (R, +, .) is a
commutative ring with identity, so is the quotient ring (R/1, +, .). It remains
therefore only to verify (R/1, +, .) is free of zero divisors. For this, assume that
(a + 1) . (b + 1) = I.
In other words, thll product of these (lOsets equals the zero element of the ring
+
(R/ I, +, .). The foregoing equation is plainly equivalent to a b I = 1, or
what amounts to the same thing, a bel. Since (1, +, .) is a prime ideal,
one of the factors a or b must lie in I. But this means either the coset a + 1 = 1
or else b + I = I, hencc (R/ I, +, .)
is without zero divisors.
To prove the convcl'!lC, we just reverse the argument. Accordingly, suppose
(R/ I, +, .)
is an integral domain and the product a bel. We then have
+ J) . (b + I) = a b + I = I.
(a
l=i+ra.
+ r a) b ~ i b + r (a b) e 1,
b = (i
Let us remark that the converse of Theorem 3-34 need not hold; there exist
examples of prime ideals which are not maximal ideals. The special properties
of the principal ideal domains, however, guarantee that the notions of primeness
and maximality are equivalent for this important class of rings. Sinee every
integral domnin contains the two trivial prime ideals, the use of the term prime
ideal in a principnl ideal domnin customarily excludes these from consideration.
Th(' rluli('al always ('xist.s, sinee we know by Tlwofl'm 3-30 that any ring con-
tains at, h'ast one maximal ideal. It is abo imnwdiale from the definition that
the t.riple (rad Il, f-,.) forms :tn i<iml of (Il, -/-, .).
(I + r a) . (b c) = (b + a r b) . c = 1,
IiOthat 1 -I- r a iii inwrtible in R. Bt'Cl\use this argument holdH for every
r E R, it follows that a E rad R, as desired.
PROBLEMS
In thl' fullowill~ ~t of prohll'm>!, all rings arll aSflllmlld to bl' (~ommutativll with
idl'ntity.
J. Show that the only maximal ideal in a field <p,+,.) is the zero ideal ({O},+, .).
2. Givl'n I is a hnmomnrphi~m from the ring (Ii, -1-,') nnto the ring (R', +', .'),
prove that
a) if (1,+,,) is a maximal ideal of (R,+,'), then the triple (f(l),+',.') is a
maximal idl'al of (R', +', .'),
b) if ([',+',J) is a maximal ideal of (R',+',.'), then the triple (f-l(1'),+,')
is a muxiulltl i,lplLl uf (Ii, 1-, .).
a. If I i!l IL hOlllolllurphi"m from the ring (R, -1-,') onto the fidd W, show +', .'),
that (k('r (fl, -L .) is a maximal ideal of (R, +, .).
4. With the aid of Thl'urems 3-29 and 3-32, obtain another proof of the fact that
(Z", +", ',,) is a field if and only if p is a prime number.
n. An idl'al (1, +, .)
of Ihe ring (R, -1-, .) is said to be minimal if I # {OJ and there
exists no idl'al (J, +,.) of (Ii, +,.) such t.hat {OJ C J C I.
a) Prove the idl'al (/, +, .)
is a minimal ideal if and only (a) = I for every non-
zero elenll'nt II E I.
h) Verify thlLI. Ill<' ring of intl'gers (:I., +, .)
ha.~ no minimal ideals.
fl. I?ur any i<l"nl (/, I,') of tlil' ring (fl, I .. ). Ildin" Hl<' ",It \l'i hy
\1'1 = {rElllrnE/fnrRomenEZ+}.
Estahlish I.hl' following assl'rtion!l:
a) The triplt (\1'1, -1-,') is an i,II'al of (R, +,.) such that I r;; Vi.'
b} If (J, +,'J i" an ill('al uf (Ii, -1-,') fur which I r;; J r;; vI, then vJ = Vi.
c) If (II, +, .) 0.011 (12, -1-. ) arl' both ideals of (R, then +, .),
vII n 12 = VYin vI2'
7. If f is IL homomorphism from the ring (R, +, .)
onto the ring (R', -1-', .'), show t.he
following:
a) If (/,-1-,') is a prime ideal of (R,+,') with 12 ker(j), then the triple
(fU), +',
J) is a prime idl'al or (R', +', .').
:l-4 CJ<:HTAIN SPI<;CIAL IDEALS 195
14. Describe the radiea! of (Z., +., ',,). II/int: Considl'r the prime factorization of n.)
15. Given (I, +,.) is all ideal of the ring (R, +,.) such that the quotient ring
(RI I, +, .)
is semisimp!e, prove that rad R ~ I. [Him: If a E rad R, show that
a +
IE rad (Ill I) by usc of Theorem 3-:J8.)
16. Prove that if (I, +, .)is an ideal of t.he ring (R, +, .),
thell rad I = I n rad R.
17. a) Hhow that the annihilator of a semisimple ring (R, +,.) is zero; in other words,
ann R = {O}.
b) Let (R,+,') h(, a ring with the property that I = aIm (ann 1) for every ideal
(/, +,.) IIf (/i, I, .). (In gl'lIl'rlLl, WI' IInly IlILvl' I.h. illl'itl"ioll I!; unn (ILnn 1).)
Pro v,' I,hat. if (.II, I,) is a maximal ideal of tl", ring (Ii, I, .), then (ann .11, +,.)
iH a minimal idl!al.
IH. Let (1,+,,) hi' an ideal of ti\(' ring (R,+,') 8.nd S~ Il he a Het that is closed
under multiplication and is disjoint from I.
a) Using Zorn's Lemma, show that t111're t'xiHts all icleal (/', +,.) of (R, +, .)
which is maximal with respect to the,;e prop('rtiell:
(i) I ~ P, (ii) P nS = ~.
b) Prove that (/', +,.) is n('cl'Hsarily a prime ideal. [/lint: If a fl. P, b fl. 1', then
(I', a) and (I', b) intersect S, whieh implies a . b fl. P.)
19. Prove: 8.11 ideal (I, +,.) of (R' +,.) i~ the intersection of prime ideals if and
only if a 2 E I implies a E 1. [/lint: Foreadl a fl. I, there is a prirneideal (P,+,')
196 RINO THEORY 3-5
of (R, +,.) which is maximal with respect to"disjointedness from the set
\a,
," a,
2 ... , a"
, ...
}
The next step in our program is to apply some of the previousty developed
theory to a certain class of rings, the so-called polynomial rings. For the moment,
we shall merely remark that these are" rings whose elements consist of poly-
nomials with coefficients from a fixed, but otherwise ar\>itrary, ring. (The
most interesting results occur when the coefficients are specialized to a field.)
The matter of formalizing the intuitive idea of what is meant by a poly-
nomial always presents a serious problem. We propose first to put this notion
on a sound basis, then to discuss the properties of polynomials and polynomial
rings. Our investigation will culminate with a survey of the basic facts relating.
roots of polynomials and field extensions. En route, the reader will see that
many of the familiar results of elementary algebra are but special cases of
general theorems to be obtained.
For an arbitrary ring (R, +, .), let poly R designate the set of all infinite
sequences
for the polynomial with last nonzero term an; when n = 0, we permit ao = 0
in order to include the zero polynamial (0, 0, 0, ... ) each of whose terms is
zero. With this agreement, the set of polynomials over R may be regarded as
the IKlt
Thus, the sequence (1, 1, 1,0,0, ... ) would be a polynomial over (Z2, +2, '2),
but (0, 1,0,1, ... ,0,1, ... ) would not.
We next introduce operations of addition and multiplication in the set
poly R, 80 that the resulting system is a ring containing (R, +, .) as a subring.
Let us first, however, make clear that two polynomials
and
arc equal if and only if their corresponding terms are equal: / = g if and only
if ak = bk for every integer k ~ O. Now, define the sum / + (J by the rule
If 8 rl!prcHcmtH Ule HPj, of all constant TJolyrtomials,' that is, the set
(0, a, 0, 0, ... ).
That is, ax is the spcdfie nu'mber of poly R which has the element a for. its
8('cond term nnel 0 for nil other terms. :\[ore genemlly, the symbol ax", n ~ 1,
will dC'lIot.e tlw polynomial
(0, ... , 0, a, 0, ... ),
wll(~re t.he ('Iem('nt a iH t.Iw (n + I )Ht. tc'rm in this polynomial; for example, we
have
ax 2 = (0, 0, a, 0, ... )
and
ax:! = (0,0,0, a, 0, ... ).
f = (an, 0, 0, ... ) -I- (0, alt 0, 0, ... ) -I- ... + (0, ... ,0, a", 0, ... )
= an -I- alx -1 a2x2 + ... + a"x",
with, of course, ao repl:wing (an, 0, 0, ... ). Thus, there is no harm in regarding
t.he polynomial rillj!; (poly n,l, .) IlS consisting of all formal expressions
From the definition of multipliention, the reader mny ensily check t.hat all
coefficients of I(x) . g(x) beyond the (n + m)th arc zero, whence
The foregoing argument serves to establish the first part of the next theorem;
the proof of the second auertion is left as an exercise.
Theorem 3-41. Let (R, +,.) be an integral domain and I(x), g(x) be two
nonzero elements of (R[x), +, .). Then
I} deg (f(x) . g(x = deg/(x) + deg g(x), and
2) either I(x)+ g(x) = 0 or deg (f(x)+ g(x ~ max {deg/(x), deg g(x)}.
The notion of degree may be used to prove the following corollary.
Corollary. If the ring (R, +, .) is an integral domain, 80 also is its polynomial
ring (Rlx), +, .).
Proo/. We observed earlier that whenever (R, +, .) is a commutative ring with
identity, these properties carry over to (Rlx], +, .).
To see that (R[x], +,.)
has no zero divi80rs, choose/(x) p4 0, g(x) p4 0 in R[x]. Then deg (f(x) . g(x =
+
deg!(x) deg g(x) ~ 0, hence the product I(x), g(x) cannot be the zero
polynomial.
Example 3-33. As an illU8tration of what might happen if (R, +, .)
baa aero
divi80rs, considcr (Zs, +s, 's), the ring of integers modulo 8. Taking
I(x) = 1 + 2%,
g(x) = 4 + x + 4x 2 ,
While many of the properties of the ring (R, +, .) carry over to the poly-
nomial ring (B[x], +, .), it should be observed that for no ring (R, +, .) is
(B[x], +, .) a field. In fact, no element of B[x] which has positive degree can
have a multiplicative iuverse. For suppose I(x) E B[x] with deg I(x) > 0; if
I(x} . g(x) = 1 for 80me o(x) E B[x], we would obtain thc contradiction
o = deg 1 = deg (j(:f) . ,(x = deg/(x) + deg g(x} p4 O.
Proof. The proof proceeds by induction on the degree of f(x). First note that
if f(x) = 0 or degf(x) < deg g(x), then a representation meeting the require-
ments of the theorem exists with q(x) = 0, rex) = f(x). Furthermore, if
degf(x) = dcg g(x) = 0, f(x) and g(x) are both elements of R, and it su1ficea
to take q(x) = f(x) . g(x)-t, rex) = O.
So, suppose the theorem is true for polynomials f(x) of degree less than n
(the induction hypothesis), and let deg f(x) = n, deg g(x) = m, where n ;;:: mj
that is,
f(x) = ao + alX + ... + a..x", a.. 'F 0,
g(x) = bo + b1x + ... + b".x"', b", 'F 0, (n;;:: m).
The polynomial
lies in R[x] and, since the coefficient of x" is a.. - (a..' b;l) . b", = 0, has
degree less than n. By induction, there are polynomials ql(X), rex) E R[x]
such that
. !I (x) = ql(X) . g(x) + rex),
where rex) = 0 or deg rex) < deg g(x). Substituting, we obtain the equation
f(x) +
= (91(X) (a..' b;I)X"-IR) g(x) + rex)
= q(x) . g(x) +
rex),
which shows that the desired representation also exists when deg f(x) = n.
As for uniqueness~ suppoSe
On the other hand, since the degrees of rex) and r'(x) are both less than that
of g(x), it follows that
deg (r(x) - r'(x < deg g(x).
This contradiction implies rex) = r'(x), which in tum yields q(x) = (x).
202 ItINO TIU;OIlY :J-5
givl'n hy the Divillioll Algorithm arc (~alled, rcllpectively, the quotient and
remainder on dividing f(x) hy g(x). Obscrve also that if g(x) is a monic poly-
nomial, or if (R, +, .) is a field, it is not necessary to assume the leading co-
cffident of fI(X) t.o he invl'rt.ihlE'.
Bl'forc~ dilll!lIll~illg HII' l'OII~I'qllmwclI of Theorem :~-42, wr~ paulii' to intl'odlle(l
I"\CIIl\(' nddit.illlllli (IIII1"I'ptS. To this I'lId, II!L (Ii, I,) h(~ n eonullut.ntivc ring
with idellt ity allll,' he nil nrhitrnry d!'llJ('lIt of N. For e:wh polYllominl
In partirmlar, it. llIay hI' C'OIlI"llIded t.hat t.he mapping whieh III'nds f(x) to J(r)
is a hOlllomorphisllI from (lllr), +, .) into (Ii, +, .);
the range of this homo-
morphism will he (It'noted hy R(r).
Some diffi('uIt.y arisl'M whell t.h(' ring (R, +, .)
fails t.o he (!ommutative; in
this case, the above result lIeed not hold. Observe that if we let
hex) = (x - a) . (x - b) = x2 - (a + b)x + a b,
UlI'n
"(I') .,.- ,.2 (I I I,) 1'1 a . II.
will (~qual her); ill ot.hpr wordR, hex) = J(x) fI(X) docs not always imply
her) = J(r) . (1(1').
0111' IIIC1fI' r1 .. finit iOIl Hhollirl I", illfl"llrilll'l'ri: H J(r) mill I/e,) rtf () Ilre ill Rlx),
W(~ Hay that fI(X) iH a Jactor of J(x) [or !ICc) dil'irle.~ f(x)] provid(~d there ('xists
some IlOlynominl h(.r) E ll[x) for whi('h J(x) = hex) . g(x).
:J-5 J'Ol,vNOMIAL RINGS
J(x) = (x - a) . q(x)
if and only if J(a) = O.
Let us next show t.hat t.he number of root.s of a polynomial over an integral
domai II call not ('xcccd i ts d~gf(~e.
Theorem 3-44. Let (fl, 1-,') be lUI integrnl elomnin and J(x) E nIx] be a
nOllzero polynomial of degl'ee n. Theil I(x) has at mO!:lt n dilltinet roots in n.
ProoJ. We proC!l!Cd by indu(~t.ioll on til(! degree of f(J:). When degJ(x) = 0,
the result is trivial, silll'e J(x) eallllot have any root.s. If degf(x) = 1, say
J(x) = ax + b, a ~ 0, t.hen J(x) Ims at most olle 1'00t; indeed, if a is invertible,
_a-I. b ill the only root, of J(x). Now, suppose the theorem is t.rue for all
polYllolllinls of dl'grcl' n - 1 ~ 1, and let dcgJ(x) = n. If f(x) has a root r,
the prl!eeding ('())'olhu'y I!;iv!'s
where the polynomial q(x) haR degree n - 1. Any root 1'1 of J(x) distinct from
r must be It root. of q(x), for, by substitut.ion,
allll, sill(!(~ (ll, I,') has 110 zero divisors, Q(I'I) =- n. FI'OIII our illlilwtioll
hypothesis, q(x) has at. most. n - 1 distinet roots. AI! the only root!! of f(x)
are r anu those of q(x), J(x) cannot have more than n distinct roots in R.
204 RING THEORY 3-5
with 0 < deg g(x) < 1,0 < deg hex) < 1 is impossible. Thus, every reducible
polynomial has degree at least 2.
Example 3-36. The polynomial x 2 - 2 is irreducible in Q[xl, where (Q, +, .)
is the field of rational numbers. Otherwise, we would have
x2 - 2 = (ax + b) . (cx + d)
= (ac)X2 + (ad + be)x + bd,
where the coefficients a, b, c, d E Q. Accordingly,
ac = 1, ad + be = 0, bd = -2,
:.!fl(j :l-5
Theorem 3-47. If (F, I,,) is IL lipid, t.he following st.af,ements are (ljuivalcnt.:
1) f(x) is an irreducible polynomittl in F[x).
:!) TIH' prin(~i"al idl'al f(x, +,.) ill:t maximal (prime) ideal of (F[x], +,.).
a) The quot.ient ring (F[xl/(f(x, +.. )
is It field.
A further faet whieh we Hhall require shortly is that every nonconstallt
polynomial ('1111 hl' faetor('d into irredueible polynomials.
Theorem 3-48. (Unique Factorization Theorem). Eaeh polynomial f(x) E F[x]
of positive degree is the produet of a nonzero clement of F and irreducible
monic' polynomials of F[x]. Apart from the order of the factors; this faetor-
izat ion is unique.
Proof. We prove our thoorem by induction on the degree n of f(x). If n = 1,
Ihl'n f(x) = ax +- b = a' (x +- a-I. b), where by Example 3-35, x +- a-I. b is_
a monic irrl'dwihle polynomial. Next, suppose f(x) has degree n > 1 and that
the thoorem holds for all polynomials of degree less than n. If f(x) is irreducible
in F[x], WI' are through Ilfh'r factoring out. its IC'uding (oeffi(ient. Otherwise,
.r(x) ill rt'.hll'ihl(~ and it. ill pORllihl(l t.o writ(~ f(x) = !I(x), hex) with y(x),
hex) E F[x], 0 < d('g y(x) < 11, 0 < d('g hex) < n. Therefore, by the induction'
hypothesis,
(/(.r:) = flJ OI(X)' 02(X)' .. O,(;!:),
hex) = a2 . h, (.1') h2(.r) ... h.(x),
whC're a" a2 are nonzero elemen!.s of F (in fact, the leading (:oefficients of o(x)
and h(x) al\ll (I.(:c), ".(x) art' irr('(hwihle monic polynomial!:!. Thus,
/(./,) (11'1' (I:.d . f/.(./,) .. Yr(./') . h,(.c) .. 1I.{.r)
Given a field (F, +, .) and an irreducible polynomial J(z) e FIz], we may ask
whether one can construct an 'extension field (F' , +,.) of (F, +,.) in which
J(z), thought of as a member of F/[z), has a root. (If degJ(z) = 1, then, in a
trivial sense, (F, +,.) is itself the required extension.) Our n~t theorem
anawen this question in the aftirmative.
Theorem 1-49. (Kronecker). If J(z) is an irreducible polynomial in F[z],
then there is an extension'field of (F, +, .)
in which J(z) has a root.
ProoJ. Let (I, +,.) denote the principal ideal of (F[z), +, .) generated by the
polynomial/(z); that is to say, I = (f(z. SinceJ(z) is assumed to be irreduc-
ible, the corresponding quotient ring (FIz]/ I, +, .) is a field. To see that
(FIz]/ I, +, .) constitutes an extension of (F, +, .), consider the natural map-
ping natl: F[z] - F[x)/l. According to Theorem 3-25, either the restriction
nat, I F is the trivial homomorphism or else the triple (natl (F), +, .) is a
field isomorphic to (F, +, .), where
nat, (F) = {a + I I a E F}.
But the fint possibility is immediately excluded by the fact that
natl (1) = 1+ I I,
the zero element of F[z]/ I. Therefore, (F, +, .) is imbeddable in the quotient
field (F[z]/ I, +, .) and, in this sense, (F[z]/ I, +, .) becomes an extension of
(F,+,')'
Next, we need to show that the polynomiaIJ(x) actually has a root in F[x1/ I.
If
J(z) = ao + alx + ... + a,.x",
Identifying b" + I with the element b", we see that a typical coset can be
(uniquely) represented in the form
(Z2[X]/(J(X), +, .),
in which the given polynomial has a root. Denoting this root by X, the dis-
cussion above tells us that
(our ('o('fIi('i('nts ("ollie from Z2, whence -1 = 1), the degf(~ of any product
can he kept h-ss than:t Xow, the problem at hand is to determine elen\('lIt.s
a, /1, f E :x ~ f(II' whi(~h
C:U'ryill~ out. t.he lJIult.ipli('atioll I\nd I'mhsLituting for ),. 3 , ),. 4 in terms of 1, ),.
allll ),. 2 , we ohtain
(II 1 /1 1 c) 1 a),. 1 (a 1 11,.2 =, I.
TIll' Milllilarity of theMe formulas to the rules for addition and multiplication
of eompl('x numbers should be obvious, As a mattcr of fact, the field
(R'[x]!(x 2 + 1), +,') is adual\y i";olllorphi(, to (C, +, ,)
under the mapping
4>: Jl'[.cI/(x 2 1 1) --> (,' givell hy
Corollary. Lt't, f(x) E F[x] with deg f(x) = n > O. Then there exists an ex-
tension of (F, +, .)
in which f(x) has n roots.
I(z) = (z - + X) (x' - 3)
X) (x
= (x - V2) . (x + V2) . (x 2 - 3).
(el +U 2 - 3) + 2cdV2 == 0,
hence
el +U' - 3 == 0, cd== O.
I(x) = +
(x - X) (x X) (x - 1') . (x 1') +
== (x - 0) . (x + 0) . (x - v'!) . (x + v'!).
(F', +, .)in which f(x) factors completely (the theorem guarantees the existence
of such extensions); then (nF" +, .)
serves as a splitting field for /(x) over F.
Having thus indicated the existence of splitting fields, it is natural to inquire
about their uniqueness. As a final topic for this section, we shall prove that
any two splitting ficldll of the I!&mc (nonconstant) polynomial are isomorphic;
this being 110, one ill justified ill using the definite article and speaking of the
splitting field of a given polynomial.
Before presenting the main theorem, two preparatory results of a somewhat
technical nature are needed. As previously noted, if r is a fixed element of a
field (K, +,.) and F!',;;; K, we write F(r) for the set of finite sums:
Proo/. The mapping II; F[x] - K defined by setting IIf(x) = /(r) is easily
checked to be a homomorphism of (F[x], +, .) onto the ring (F(r), +, .). It
follows at once that
f(x) E ker (II) = {g(x) E F[x] I g(r) = O},
whence (/(x !; ker (II). One observation is quite pertinent: the possibility
that ker (II) = F(x].does not arise, since the identity element of (F[x], +,.) is
not mapped onto zero. Asf(x) is assumed to be irreducible in F[x], (f(x , +, .)
is a maximal ideal of (F[x], +, .), so that (f(x = ker (II); in other words,
(f(x simply consists of all polynomials in F[x] having the element r as a
root. Thus, by the fundamental homomorphism theorem for rings (Theorem
3-18), there exists an isomorphism 8 of (F[x]/(I(x, +,.) onto (F(r), +,.)
such that II = 8 nat(/(:<)l: As regards the last assertion of the theorem. we
0
evidently have
r = II(X) = (8 nat(/(:<)(x) = e(x
0 (I(x). +
Although of some interest in its own right. the value of this lemma is that it
leads almost immediately to the following theorem:
Theorem 3-51. Let t/> be an isomorphism from the field (F, +..)
onto the
field (F', +', .'). Also let f(x) = al + alX + ... + ~x" be an irreducible
polynomial in F[x] and 1'(y) = t/>(ao) + t/>(al)Y + ... + t/>(a..)y" be the
corresponding polynomial in F'[y]. Then1'(Y) is irreducible in F'[y]. Further-
more, if r is a root of f(x) in some extension field of (F, +.. ) and r'is a root
of f'(y) in some extension field of (F', +', .'), then t/> can be extended to an
isomorphism 4> of (F(r), +, .)
onto (F'(r'), +'.. ')
with 4>(r) = r'.
214 UlNH '1'ln;OTty
Proof. Let us first extend 4> t.o a mapping f, between the polynomial rings
(F[x), +, .)
and (F'[y], +', .')
by taking
f,g(x) = f,(b o -I- blx + ... + II"x") = 4>(b o) + I/>(bdy + ... + 4>(b,,)y"
for every polynomial g(x) = bo + blx + ... + b"x" E F[x). Using the fact
that I/> ill an isomorphism, it is an easy matter to verify t.hat f, is an isomorphism
of (F[x), +,.) onto (F'[y], +',
.')j we leave the render to supply t.he necessary
dctni\s. Not i('(! that. for !lny polynomilll Vex) ill FIx}, an clement 0 E F is a
root. of g(x) if and only if 4>(0) iN a root of 4iu(x). Indeed, if (I(x) = bo + blx +
... + b..x", we hnvc
(f,(J(;rJ)(r/>(fl </,(1'11) 1 r/>(/'.)r/>(ll) 1 .,. 14>(I'")'4>(tt)"
"-c 1/>(11 0 -I b l a -I 1- b,. a")
~: 4>(u(a,
from which the nr-;s('rtion follows. In pnrti('ular, the givl'n polynomial f(x) is
irrcdueiblc in FIx} if and only if f'(y) = f,f(x) is irredueible in F'[y].
Now, by !.Ill' fort'going lemma, we know that there exist isomorphisms
a: F(I') -+ F[.rl/(J(x and (3: F'(r') -+ F'[y]/ (J'(y.
OblS('rV<' , parti('ularly, that T ('luries the ('Of;(~t, x + (j(x onto 11 -I- (j'(1J.
WI! (!Ollt(,(ld that (F(I'), I,) ~ (F'(r'), I', .') vin tIl(! ('ompoHition
':
where 1>: F(l') -+ F'(l") j thiR situation is portrayed in the diagram below:
."
-r----~----.
Flx]/(I(x))---..;T---_. F'r,]/(r(y))
O'rtninly I> is an i8()morphism of (F(r), -I,,) onto (F'(r'), +', .'), for th(l
individual mnppinjl;s a, T, and (3-1 nrc themselves isomorphisms. If a is an
nrhitrnry ('I('nlt'lI!. of F, then
I>(a) = uri 0 T) (a(a)) = (rioT) (a + (j(x)
= fj-I(r/>(a) -I (f'(1J) = 4>(a) ,
I'OI.YNOMIAL IUNGR 215
when('e <I> is actually an extension of tJ> to F(r), Finally, we point out that
PROBLEMS
4. Show that the relation - defined by taking f(x) - g(x) if and only if
deg fex) = deg g(x)
ill an equivalence relation in the set of nonzero polynomials of R(x].
5. a) Let P be the set of all polynomials in Z(x] with constant term 0:
P = {a\x + a2x2 + ... + a"x" I a. E Zj ft ~ I}.
Establillh that the triple (P, +, .)is a prime ideal of (Z[x1, +, .).
b) Show the principal ideal 12:),+,') is a prime ideal of the polynomial ring
(Z[x], +, .),but not a maximal ideal.
6. If (Il, +,.) iM a commutative ring with identity, prove that the polynomial I +
ax
is invertible in R[x] if and only if the element a is nilpotent.
7. Let (R, +,.) be a commutative ring with identity and the element r E R be
fixed.
a) If R(r) denotes the set
R(r) = U<r) I f(x) E R[x]},
wh('re g(x) E Rlx) is a polynomial such that g(r) ~ O. Prove that an element
r E R i.. 0. multiple root of f(x) if and only if r iii 0. root of both f(x) and of(x).
16. Determine
a) 0.1\ irreducibl{' Jlolynomials of degree 2 in Za[x], and
b) all irreducible polynomials of degree 3 in Z2[X).
17. Let (f', +, .)
b(' a fie>l.l anll/(x) E F[x) be a Jlolynomial of degree 2 or 3. Establish
that f(x) is irre>ducihle in FIx) if and only if f(x) ha.'i no root in F. Give an l'xample
which >!howH that this fI'sult IU'C't\ not he true if t\eg fIx) ~ 4.
IS. Prov!' tluLt if f(x) is all irn'dllcible polynomial in Z[xl. then f(x) i>! also irredllcibl('
whf'n rf'gardf'd 1\." all 1'1C'lIll'lIt of Qlx).
19. Df'riw Iht following analog of Euclid's LI'llIma: L.. t (R, +,.) be a commutative
ring with iell'nlhy and f(x) be all irrC'ducihlt' polynomial ill Rlx). If f(x) divides
the prulhH't g}(x) . U2(:r) . g,,(x), then !(x) divi.l(s gk(X) for some k, 1 ::; k ::; n.
20. In regan 1 to ProblC'm 7, show that the ring (R(r), +, .)
is a field if and only if
ker (I/ = (f(x where I(x) is an irreducible polynomial in R[x).
219
Under these operations, the triple (R, +,.) is a commutative ring with identity;
the proof is straightforward and will not be given in detail. To establish the
idempotency condition, we proceed as follows: If the function fER is such
that f(x) = 0, then
(J2)(X) = I(x) '"/(x) = 0. 2 0 = O.
While if lex) = 1, then
(J")(x) = lex) ,/(x) = 1,1 = 1.
a +h = + h)2 = a2 + a h + h a + h2 = a + a h + b a + b,
(a
2a = a + a = a 2 + a 2 = 0
The ideal (J, +,.) thus contains the identity, and consequently J == H.
A natural undertaking is to determine which Boolean rings are also fields. We
may dispose of this question rather easily: up to isomorphism, the only Boolean
field is the ring of integers modulo 2.
Theorem 3-56. A Boolean ring (H, +, .) is a field if and only if (H, +, .) ~
(Z2' +2, '2)'
Proof. Let (H, +, .) be a Boolean field. For any nonzero element a E H, we
then have
This argument shows that the only nonzero element of H is the identity; in
other WOrdR, R = {O, I}. But any two-element field is isomorphic to (ZI' +1, '1)'
The opposite direction of the theorem is fairly obvious.
222 lUNG TIIF..olty 3-6
1 = r (I + a)2 = (r (1 + a) . (1 + a) = 1 (1 + a),
from which it folloWH that a = 0, contrary to hypothesis.
BCCIUlHe (/, ~I . ) is a proppr iil(,lLl, Theorem a-ao aHHurefl the ()xistence of 1\
mnximal ideal (111, +,.) of (fl. + .. ) sud, that I ~ M. In light of the result
just proved, the associated quotient ring (RI 1II, +, .)
will be isomorphic to
(Z2, +2. '2) vin KOIlW homomorphism g.
W(~ may thl'refom define I~ funet.ion f: U --+ Z2 hy t.aking f = !I natM, where0
nlLtM ill Himply the lIatural mapping of U Ollto Illlll. The situation ill con-
veniently depicted by the following diagram of maps:
'~/Z',~
RIM
._
The rcmllinil('r of till' proof amounts t.o showing that t.he function I, flO
i1efillC'i1, hUH t hI' 11I'/)pI'rlil's ILH.'!I'rted in till! statement of the tlwormn. Plainly,
f is hoth 1\1\ unto mup and a homomorphism I)('illg the composition of two such
:~-6 223
1 = a + (1 - a) EM.
This leads at once to M = R, the desired contradiction.
Having a!i.'iembled the necessary machinery, we now Ret ourselves the principal
taHk, that of showing that each Boolean ring is (~8sentially a ring of sets. This
theorem, now considered a landmark, was first proved in 1936 by the American
mathematician, l'larshall Stone.
Theorem 3-57. (Stone Representation Theorem). Every Boolean ring (R, +, .)
ill isomor(lhie t.o a ring of Ilubllets of !rome fixed 8<!t.
Proof. To begin the attack, let H denote the collection of all homomorphisms
from (R, +, .) ont.o the field (Z2, +2, '2)' Next define It flllWtioll h: R -. P(H)
hy assigning to caeb element a E R t.hose members of II which assume the value
1 at a; in other wordll,
h(a) = {fEHlf(a) = I}.
a V (b 1\ c) = (a V b) 1\ (a V c),
a 1\ (b V c) = (a " b) V (a 1\ c) for all a, b, c e B.
(P.) There exist distinct identity elements 0 and 1 relative to the operations
V and 1\, respectively; that is,
V 1 2 5 10 A 1 2 5 10
1 1 2 5 10 1 1 1 1 1
2 2. 2 10 10 2 1 2 1 2
5 5 10 5 10 5 1 1 5 5
10 10 10 10 10 10 1 2 5 10
The formal verification that the triple (B, V, A) constitutes a Boolean algebra
is left as an exercise. (We should caution the reader that in this example the
integer 1 plays the role of the identity element for the operation V, while the
integer 10 serves as the identity element for the operation A.) A quick in-
spection of the foregoing tables will reveal the various complements to be
We ca.ll attention to the fact that a' is simply the quotient when 10 is divided
bya.
The first thing one notices on inspection of the axiom system for a Boolean
algebra is the perfect symmetry or duality between the properties of the two
operations V and A. That is to say, if V and A are interchanged in the axioms
and if, at the same time, 0 and 1 are also interchanged, then the properties are
merely permuted amongst themselves. This principle of duality permits us to
state all theoreIns in dual pairs (unless, of course, a statement happens to be
its own dual) and guarantees that the proof of one of the pair of statements
will be sufficient for the establishment of both; the proof of the dual theorem
is obtained by maki~g the appropriate interchange of symbols in the proof of
the original theorem.
We now proceed to deduce from the postulated properties of the operations
in a Boolean algebra a series of further properties, including, for instance, the
226 lUXG THEORY 3-fi
aSHOciative laws for V and 1\. Dual statements are placed side by side; ill
view of the principle of duality, only one statement from each dual pair need
be proved. In order to condense the demonstrations, we shall arrange the steps
80 far as possible one under another, citing to the right those I>ropositions used
in passing a('ross succcssive equality signs.
Theorem 3-58. III !LilY Boolc!L1I ILlgchra (8, V, 1\), the following prolK!rtil'H
hold:
1) The c1cmellts 0 and 1 are unique.
2) For each clement a E B,
a V a = a, al\a=a.
:J) For l'll.{'h dement a E B,
a V 1 = 1, a 1\ 0 = o.
4) For eaeh pair of elements a, b E B,
a V (a 1\ b) = a, a 1\ (a V b) = a.
Proof. To establish (1), we need only appeal to Theorem 2-1. The proof of
(2) is indicated below:
a=aVO (by P a)
= a V (a 1\ a / ) (by 1>4)
= (a V a) 1\ (a Va') (by P2)
= (a V a) 1\ 1 (by P 4 )
=aVa (by Pal.
We obtain (3) as foIlows:
l=aVa' (by P4)
= a V (a' 1\ 1) (by )la)
= (a Va') 1\ (a V 1) (by P2)
= 1 1\ (a V 1) (by P4)
=aVI (by Pal.
The proof of (4) requires the use of (3):
a=al\l (by Pa)
= a 1\ (b V 1) (by 3)
= (a 1\ b) V (a 1\ 1) (by P 2 )
= (a 1\ b) V a (by P a)
= a V (a 1\ b) (by PI)'
3-6 BOOLEAN RINGS AND BOOLEAN ALGEBRAS 227
We did not include the 8II8Ocint.ive laws for V and 1\ among our axioms for
a Boolean algebra, as is frequently done, since they arc logical consequences of
the properties listed. This is demonstrated in our next theorem.
Theorem 3-59. In any Boolean algebra (B, V, 1\), each of the operations
V IUlIl 1\ is I\IIHOdat.iv(; thl~t. is, for pVl'ry t.ripll of I'll'tnc'nts a, b, c e B,
a V (b V c) = (a V b) V c, a 1\ (b 1\ c) = (a 1\ b) 1\ c.
a 1\ Y = [a 1\ (a V b)] V (a 1\ c) (by P 2 )
= a V (a 1\ c) [by Theorem 3-58(4)
=a [by Theorem 3-58(4)].
Therefore a 1\ x = a 1\ y. Now,
and also
(a 1\ x) V (a' 1\ x) = (a 1\ y) V (a' 1\ y)
(a 1\ a') V x = (a 1\ a') V y (by Ph P 2)
IVx=IVy (by P,,)
x=y (by Pa),
proving the associative law for the operation V; that 1\ is also associative
follows by a dual argument.
228 BING TBJlORY 3-6
Proof. For (1), asaume there are two elements x and y of B such that
a V x = 1, a 1\ x = 0,
a V 11 = 1, al\y=O.
We then have
(I," = (a')' = a.
o VI = 1, 01\1=0,
Finally, we prove the first statement of (4). Note that, since complements
are unique, it is enough to establish
(a V b) V (a' A b') = 1, (a V b) A (a' A b') =0
Now,
(a V b) V (a' A b')
= [(a V b) Va'] A [(a V b) Vb'] (by P 2 )
= [(a Va') V b) A [a V (b Vb')] [by Ph Theorem 3-59]
= (1 V b) A (a V 1) (by p.)
=lAI [by Theorem 3-58(3)]
= 1 (by Pa).
Furthermore,
(a V b) A (a' A b')
= (a' A b') A (a V b) (by PI)
= [(a' A- b') A a] A [(a' A b') A b) (by P 2)
= [(a' A a) A b'l
A [a' A (b' A b)] [by Ph Theorem 3-59]
= (0 A b') A (a' A 0) (by Pl. p.)
=OAO [by Theorem 3-58(3)
=0 (by Pa).
Furthl!rmore,
a +0 = (a A 0') V (a' A 0)
= (a A 1) V (a' A 0)
=aVO=a
and
a +a = (,I A a') V (a' A a) = 0 V 0 = O.
This shows that the element 0 acts as the identity for the system (B, +), while
each element is its own (additive) inverse.
To establish the associativity of addition, let us first perform a preparatory
calculation:
(a + b)' = [(a A b') V (a' A b)]'
= (a A b')' A (a' A bY
= (a' V b) A (a Vb')
= [(a' V b) A a) V [(a' V h) A b']
= (a A b) V (a' A h').
Utilizing this relation, we then have
wher('IUI
a b + a e = (a /\ b) + (a /\ e)
= [(a /\ b) /\ (a /\ e)'] V [(a /\ b)' /\ (a /\ e)]
= [(a /\ b) /\ (a' V c')l V [(a' Vb') /\ (a /\ e)]
= (a /\ b /\ a') V (a /\ b /\ e') V (a' /\ a /\ e) V (b' /\ a /\ e)
= (a /\ b /\ e') V (a /\ b' /\ e).
The proof of the theorem is therefore complete.
We now reverse this process; in other words, we start with a Boolean ring
and transform it into a Boolean algebra by suitably defining the opera.tions
V and /\.
Theorem 3-62. Every Boolean ring (B, +,.) becomes a Boolean algebra
(B, V, /\) on defining
Proof. That V and /\ are both commutative followli immediately from the
commutativity of the operations in (B, +, .).
A simple calculation will show
that /\ is distributive over V :
a /\ (b V e) = a . (b + e + b . e)
= a b + a . e + (a b) . (a e)
= (a b) V (a e) = (a /\ b) V (a /\ e).
The verification of the other distributive law relics on the fact tha.t, since
2x = 0 for every clement of the ring (R, +, .), it is ullnecessary to distinguish
betwccn addition and subtraction:
(a V b) /\ (a V e) = (a + b + a ' b) . (a + e + a . e)
= a+ab+ab+ae+be+abc+ac
+abc+abc
= a+bc+abc
=a V (b c) = a V (b /\ e).
If 0 and 1 are the additive and mUltiplicative identities of (B, +, .), then
a V 0= a+O+a'O= a, a/\l=al=a
for every a E B. Filllllly, we hllve
a V (1 + a) = a + (1 + a) + a . (1 + a) = 1 + 4a = 1,
a /\ (1 + a) = a' (1 + a) = a + a 2 = 2a = 0,
232 RING THEORY 3-6
PROBLEMS
1. Prove that in a Boolean ring (R, +, .), every triple of elements a, h, c E R satisfies
the identity
(a + h) (h + c) (c + a) = O.
2. If a Boolean ring (R,+,) has at least three elements, show that every nonzero
element except the identity is a divisor of zero. [Hint: For a, hER, consider the
+
product (a h) a h.]
3. Prove that any ring (R,+,) in which each element is idempotent can be imbedded
in a Boolean ring. [Hint: Let R' - R X Z2 and mimic the argument of Theorem
3-16.]
4. a) Let (R, +,.) be a commutative ring with identity and S the set of idempotents
of R. For a, h E S, define the operation * by taking
a * h .. a +h- 2(a . h).
Prove that the triple (8, *, .) forms a Boolean ring, known &8 the idempoUnt.
Boolean rifl(J of (R, +, .).
b) In particular, obtain the idempotent Boolean ring of (Z12,+12, '12)'
5. Suppose (1,+,) is an ideal of the Boolean ring (R,+, .). Show that (1,+,.) is
a proper prime (maximal) ideal if and only if for each element a in R, either a E I
or 1 - a E I, but not both.
6. Given (R,+,) is a Boolean ring. For an element a E R, define the set I(a) by
I(a) >= {I 1(1,+,,) is a maximal ideal of (R,+, .); a E I}.
Verify that the sets I(a) have the following properties:
a) I (a) ,. ., whenever a ,. O.
+
b) l(a h) - l(a) ~ l(h).
c) l(a' h) - I(a) n I(h).
d) l(a) - I(b) if and only if a-h.
[Hint: a E 1 if and only if 1 - a E 1.]
BOOLEAN RINGS AND BOOLEAN ALGEBRAS 233
16. By means of Theorem 3-61, convert the Boolean algebra (B, V, A), as defined
below, into a Boolean ring.
V abc d A abc d
a abc d a a a a a
b b b b b b abc d B == {a, b, c, d}
c c b c b cae c a
d d b b d dad a d
17. Suppose a Boolean algebra (B, V, A) is made into a Boolean ring (B, +,.) via
Th('orem 3-61, and then (B, +,.) is convertt"d baek to a Boolean algebra
(R, V \,1\ \) via Tlworem 3-62. Verify that (R, V, 1\) ... (B, VI, 1\ .).
18. Suppose (B, V, 1\) is a Boolean ring and 0 ;o!i I\; B. The triple (I, V, 1\) is
said to be a (Boolean) ideal of (B, V, 1\) if and only if
i) a, bEl implies a V bEl,
ii) aE I and bE B imply a 1\ bE I.
a) Prove that every Boolean algebra (B, V, 1\) has two trivial ideals, namely,
({O}, V, I\) and (B, V, 1\).
b) If (Ii, V, 1\) is a collection of ideals of (B, V, 1\), show that (nIi, V, 1\) is
also an ideal.
e) Prove that if (I, V, 1\) ill an ideal of (B, V, 1\) and 1 e 1, then 1 - B.
19. Let (P(X), U, n) be the Boolean algebra of subsets of a nonemptlt set X and Xo
be any element of X. Prove that
a) if I is the family of all subsets A \; X such that Xo ~ A, then the triple (1, U, n)
is an ideal of (P(X), U, n),
b) if X is infinite and I is the family of all finite subsets A \; X, then the triple
(1, U, n) is an ideal of (P(X), U, n).
20. Let (B, V, 1\) and (B\, V \,1\ \) he two Boolean algebras and I a mapping from
B into B.. Then I is Haid to be a Boolean homomorphiam from (B, V, 1\) into
(Bl' VI, 1\.) provided
I(a V b) = I(a) V. I(b),
I(a 1\ b) = I(a) 1\. I(b) ,
I(a') "" I(a)'
for all elements a, bE B. (The formation of complements may be regarded as a
unitary operation.) Show that such a function has the following properties:
a) 1(0) ... 0.,/(1) = 1\.
b) (f(B), V., 1\.) is a suhalgebra of (B., VI, 1\ .).
c) If a :S b is taken to mClI.n a 1\ b .. a, then I(a) :S I(b) whenever a :S b.
d) The triple (ker(!), V, 1\) is an ideal of (B, V, 1\), where
VECTOR SPACES
The theory of matrices has long occupied a strategic position in various branches
of mathematics, physics, and enginccring. Only in reccnt years has its im-
portance in the social and biological sciences as well hecome apparent. Thc
subject t.oday has become an indispensable tool ill :mcll new field!! as game
theory, linear programming, and statistical decision theory. Part of the reason
for the widening applicability of matrix theory is no douht the role it plays in
the analysis of di!!eretc'obHcrvations and the case with whieh matric operations
may bl~ progmmmed for modern highapeed computen;.
We do not intend to give a complete nnd systcmatie aeeount of the problems
of matrix theory and its diverse applications. Rather, the operations and the
basie properties of veetors and matrices arc approached from an algebraic point
of view with the aim of illustrating some of the conccpts of the previous chap-
ters. One result of sueh a study will be t.hc formulat.ion of a mathematical
sy!!tcm, somewhat more complicated than tho!IC st.udied earlier, known as a
v~tor spa(!e.
The basic definition whieh atl1rts U!! ofT ill that of II. vector, the fundamental
object in our study.
Deftnitlon 4-1. An n-componcllt or n-dimensional vector over a field (F, +,.)
is un ordered n-tuple (aI, a2, .. ,an) of clements ak E F.
The clements ak E F ure called the comTllInents of the vector and we say n is
its dimension. Clearly, the set of all one-dimensional veetors over (F, +, .) can
be identified with F itself. To have to write out the whole vector is somewhat
awkward; hereafter, we will condense our notation and designate the vector with
(~OmpollentR ak by (a,,). It is hardly necessary to point out t.hat two n-component
V(,(,t.()rtI (ak) and (Ilk) Iln~ (~qlllll, ill whidl ellS(. w/' writ./' (nk) ~ (lJd, if Illld only if
t.hcir eorrcspolldillg eompOltelltll arc equal: (ak) = (I)k) if and only if ak = bk
for k = 1, 2, ... , n.
Definition 4-2. n) The sum of two n-component vectors (ak) and (b k) , denoted
by (at) + (b,,), is the veetor obtained by adding their corresponding com-
ponents. Thus, (at) + (b,,) = (a" + bk).
235
236 VECTOR SPACES .4-1
Here, we conform with the standard practice of using the plus sign in two
different contexts, for vectors and for their components. It should be perfectly
clear in any given situation whether we are adding vectors or elements of F.
Note, incidentally, that the difference of two vectors may be expressed in
terms of the operations already given:
For a simple illustration of these ideas, consider vectors over (R', +, .); in this
case, we have
(1,2,3) - 2(1,0, -1) = (1,2,3) + (-2,0,2) = (-1,2,5).
Definition 4-3. Any vector whose components are all zero is called a zero
vector and is represented by the symbol O.
Let V,,(F) denote the set of all n-component vectors over an arbitrary field
(F, +, .). Inasmuch as vector addition enjoys the basic additive properties of
its components, the following theorem concerning the algebraic nature of
the pair (V,,(F), +) is obvious. -
Theorem 4-1. The system (V,,(F), +) is a commutative group, having the
zero vector of dimension n as its identity element and (-a,,) as the inverse
of a vector (a,,) E V,,(F).
The operation of multiplication of vectors by elements of F, as defined above,
has the following properties: if r, E F and (a,,), (b,,) are vectors in V,,(F), then
1) (r + .)(a,,) - r(a.) +.(aAl),
2) r[(a,,) +(b,,)] = rea,,) +
r(b.),
3) r[.(a,,)] = (r 8) (a.,) ; l(a.) = (a,,); O(a.) = o.
Verification of these facts is not particularly difficult and is left to the reader.
Vectors may also be combined under a rule of composition known as inner
product multiplication.
Definition 4-4. The inner product of two vectors (a,,), (b,,) E V,,(F), denoted
by (a,,) (b.), is defined to be
0
(aAl) 0 (b,,) =
.
L al . bl;.
1-1
However, one should not jump to hasty conclusions concerning divisors of zero,
for on the right-hand side we have the real number zero and not a 3-eomponent
zero vector.
While failing to be even a binary operation, inner product multiplication
nonetheless enjoys some interesting properties, several of which are listed in
the next theorem.
Theorem 4-2. If rEF and (at), (bt ), (Ct) are vectors in V .. (F), then
1) (a.,) 0 (bAo) = (bAo) 0 (at),
2) 0 (a,,) = 0 = (a,,) 0,
0
4) (a.,). [(b,,) +
(Ct)] = (aAo) (bAo) (al;) (CI;).+
Proof. Let us illustrate the type of argument involved by establishing the last
statement; the remainjng parts of the theorem are left as an exercise. We
proceed as follows:
..
(a.,) 0 [(b.,) + (CI;)] = (al;) (bl; + CI;) = L aAo' [bl; + Ct]
.. .. "-1
= E a., . b., + E a., . c.,
"-1 "-1
= (a.,) (b.,) + (a.,) (c.,).
Definition 4-5. By an m X n matriz (plural: matrice.) over the field (F, +,.)
we mean a function from {I, 2, ... ,m} X {I, 2, ... I n} into F.
In the case of matrices, one customarily arranges the functional values block
fashion in a table made up of m rows and n columns. S~cifically, if the value
of the matrix at the ordered pair (i,j) is denoted by aij, where 1 ~ i ~ m,
1 ~ j ~ n, then the matrix is indicated by the following rectangular array:
all a12 ... a l .. )
(
~21 ~22 ... ~2. . .
Note that clements are located ill the matrix by the use of double subscripts.
the first subscript indicating ttHl row. and the second subscript the column in
which the element is found. For instance. the element a23 is in the second row
and third column.
To avoid cumbersome notations. it is convenient to abbreviate a matrix aH
(ai;)",X", to be read "the matrix of dimension m X n whose elements are the
ai/s." When the numbers of rows and columns are clearly understood. we may
instead simply write (aij). If m = n. the mntrix is snid to be square of order n.
Definition 4-6. Two m X n matrices (ai;) and (b i ;) are equal, for which we
write (ai;) = (b i ;) , if and only if their corresponding elements are equal;
Lhllt iH, flij l)ij fill' ILII i ILIIII j.
Since the rowl'! of all m X n mlttrix may be regarded as clements of the vector
space V,,(F) , it iH 1I0t HlIl'priHing that the operations defined in V,,(F) have
natural generalizations to matrix operations.
Definition 4-1. a) The Hum of two m X n matriceH (ai;) and (bij). denoted
by (aii) + (b i ,), is the matrix obtained by addillg their corresponding ele-
ments. Thus, (aij) + (b i;) = (ai; + bi ;).
b) The product of the matrix (ai;) and the element rEF. denoted by r(ai;),
is the matrix obtaitwd by mult.iplying each element of (aij) by r. Thus,
r(ai;) = (r ail)'
Observe that by its definition, addition is a binary operation on the set of
all matrices of a given size; that is, the sum of two n X m matrices is again an
n X m matrix.
Example 4-1. Taking (ll', +, .) as the base field, let
~) , :).
-6 4
A=e B= (:
3 0 -1
Then
2A + = (:
B
-12
0
:)+(: 4
-1 ~)
= (:
-8
-1
:).
A matrix each of whose elements is zero is called a zero matrix and is denoted
by O. Accordingly, a zero matrix need not be square. For the zero matrix
whose dimensions are those of (ai;), we have
(ai;) +0 = = 0 + (a;j).
(ai;)
(ai;) -I- (-I)(aij) = 0 = (-I)(a;j) + (aij).
4-1 THE ALGEBRA OF MATRICES 239
Let. us denote the set of all m X n matrices over the field (F, by the +, .)
symbol M ",n(F). The following theorem establisheH the algebraic properties of
(M",,,(F), +) under matrix addition.
Theorem 4-3. The system (JIf ",,,(F), +) is a commutative group, with the
m X n zero matrix as the identity element and (-aij) as the inverse of a
matrix (aij) E M ",,,(F).
Proof. Definition 4-7 indicates that each property of matrix addition is derived
from the correHpondillg ndditive property in the field (F, +, .). For instanoe,
to establish the commutative law, let (aij), (bij) EM",,,(F). Then
The rest of t.he proof proceeds along similar lines and is left to the reader.
Although multiplieation of matrices by a field element is not a binary opera-
tion on M ".,,(F) lunless, of couI"s(~, m = n = 1), this operation has several
int.eresting features. Specifieally, if (aij), (b ii ) E JIf ".n(<') and r, 8 E /I', then
r[(aij) + (bij)] = r(aij) + r(bij),
(r 8)(aij) = r[8(aij)],
(r + 8)(ai;) = rea,;) + 8(a,;),
l(a,;) = (a,j), O(ai;) = O.
Our main purpose for introducing inner product multiplication for veetors
becomes apparent with the following definition.
Definition 4-8. If (aij) is an m X n mntrix and (b ij) is an n X r matrix, then
their product (Cii) = (aii)' (bi') is an m X r matrix whose elements are given
hy the formula
.
Cij = ~ ail, . bk ; for i = 1,2, ... ,mi j = 1,2, ... , r.
k-1
As the SUbscripts indicate, the ijth entry c'l in the product matrix (ai;) . (b i ;)
is obtained by taking the inner product of the ith row of (aii) and the jth
column of (h ij):
b"J
As we observed previously, the inner product of two vectors is defined only
if the vectors involved have the same number of components. Thus for the
matrix produet (aij) . (bij) to exist, the number of columns in the matrix (ai;)
240 VECTOR SPACES 4-1
A = (: 2 -11) and
Then
2X3
2, 3 + O (-1)
3X2
+ 1 0 21+00+12 )
- (6 4 )-
=
(
33+2(-1)+(-1)0
31 + 2 + (-1)2 - 7
1
2X2
On the other hand,
1)
2 ~1
3X2 2X3
3.2+1.3 30+1-2
3.1+1.(-1>]
= 1-12+0-3 -10+02 -1 . 1 + 0 (-1)
02+23 0-0+2-2 01 +2 (-1)
3X3
We next dispose of one natural question that arises here, namely, the question
of commutativity of matrix mUltiplication. First of all, given an m X n matrix
(ao/), the matrix products (a'/)' (bu) and (b'/)' (O(/) are both defined if and
only if (b ii ) is an n X m matrix. When the latter condition holds and it is a.t
least possible to form these two products, (aii) (bi/) and (bii) (4;i) will be of
different dimensions unless m = n. Even if this is the case, where it is mean-
ingful to ask whether (ail) . (bi i ) a.nd (b ii ) . (aii) are equal, matrix mUltiplication
will not as a rule be commutative. One need only consider the computation
Due to the asymmetric way in which two matrices combine in a product, such
an outeome is not totally unsuspected. It is quite possible, of course, that a
particular pair of matrices may commute.
I'or the zero .matrices of appropriate dimensions, (aii)' 0 = 0 and
O (ail) = O. In particular, if both (aii) and 0 are square matrices of the
same order, then (aii) . 0 = O (aii) = O.
If (a;j) is any square matrix, then that part of the matrix consisting of the
the elements ali is called the (main) diagonal of the matrix.
DefInition 4-9. The identity matrix oj order n, designated by I .. , or simply I
when there is no chance of confusion, is the square n X n matrix having ones
down its diagonal and zeros elsewhere.
It is helpful to have some notation for the elements of the identity matrix.
Consequently, we will denote the element in the ith row and jth column of I ..
by the symbol o;/t where
I for i = j,
'il ={ (the Kronecker delta)
o for i ~ i,
and thus write Ifl = (O;i)' To illustrate,
For each positive integer n, the set of all square matrices of order n over the
field (P, +, .) will be represented by M fI(F), rather than M .... (F). The identity
matrix I .. serves as an identity element for the operation of matrix multiplica-
242 VI<lCTOIt I'IPAC!EFI 4-1
tion in the set M,.(F). Indeed, if (a;i) E M,.(F), then
o if j -F 8.
4-1 THE AWEBIlA OF MATRICES 243
where til!' nllltrix (lJij) IIml the value a;jl down itt! llIain diltgonnl lUll I zeros
elllwhN(~. Due to t.he presnrwc of nil the 1:ero elll.rieH in the variout! fnetors,
thit! produet is equal t.o H. t l\Iorcover, Hilwe (S,I-, .) is an ideal, the matrix
g., belongH t.o S. The relation
shows further that all n 2 of the matrices Eij arc contained in S. The set S,
being closed under addition, then has the identity mlLtrix In as a member,
from which we conelude that S = 111n(F), In other words, if S ~ {O}, then
S = M,.(F).
In passing, let. us remark that while the ring (llf,,(p), +,.) fails to have
nontrivial (two-sided) ideals, it may very well possess one-sided ideals. For
example, the set of all matrices of the form
a, b EF,
a,beF.
We propose to show that the system (S, +,.) is a field. If the matrices
Now if
(-: :) ~ (~~),
either a ~ 0 or b ~ 0, 80 that a 2 + b2 ~ o. Accordingly, (as + b')-l exists
and, as a direct computation will show,
( a b)-l = (a2
-b a
+ bS)-l(a
b
-b) S.
a
e
f( b)-b
a
a
= (a, b).
= (a+ e, b + d)
= (a, b) + (e, d)
=f(-b b)+1"'\-d d).
a
a
e
e
Moreover, if
I( b) .f( Cd),
a
, -b a -d e
then from the definition of equality of the vectors (a, b) and (e, d), we must
have a = e, b = d. This implies that
(: :), neZ,
246 VECTOR Sl'ACES 4-1
in T will hc~ cll'llot{~d Hilllilly hy (n); while thit! iH 11I1 incorll'c~t notation, it ill much
leHl! ullwiddly than the I~orrect olle.
WllIlt we willh to show here is that the mapping J: T - Z. defined by
Jn = 2n yield!:! an isomorphism between (T, +,.) arid (Z., +, .), the ring
of even integers. The demonstration that J preserves addition is straight-
forward, so we shall consider only multiplication: for n, m E Z,
PROBLEMS
1. Determine the values of a and b for which the following matrix equation holds:
[30]
1 I . (4 7) + 2 0
[a -3] [20 15]
5 10 25
5 2 6 8 4 b 40 25
3. Show that ('ach of the following matrices from J\'h(R') is a solution of the matrix
+
t'qllation .\'2 - 5.\' 41 = 0:
and ( 3-2).
-I 2
(3-4).
1 -I
a) Show that the pair (G,) is a group, where indicatcs matrix multiplication.
b) Ruppose II and K denote the sets of matrices of the type
nnd
n"IH{!i\,!ly. I'ru\'t'I.hnl (II,J i.. "l<uhl('""uI' "f Hit, I(rulll' (I, .J, wllil!' (1\,,)
i.. II lI"rltlnl .. llh~,..HI" .. f (tI, .J.
6. Uivt'n til(' matrix (a,j) E .Hm.(ft') and r, 8 E 10', prove t.hat
a) (r+8)1 .. = rl .. +81., b) (r8)/. = (rl.)(81 .. ),
c) I", (ao) - (a;j)' I .. "" (aiS).
7. Let thl' Ret 0 conKiHt. of the six 3 X 3 matrices
~] ~]
0 0
[: 0 [:
0
[:
0
0 :]
[~
0 0
0
0
;] [:
0
:] [; 0 :]
EHtahliKh t.hat thl' pair (G, .) formll a group illmorl.hic to (Sa, .), the symmetric
grllu p on a HymholM.
:
8. ConHidt'r setH Ganci /I conHiHting of all matrices rrom M a(/") of the form
[: ::] and
[~ ;]
reHIK'(:tivl!iy. Vt'rify that I.htl pair (G,') is a group and that (II,') formH a normal
I!ubgroup of (0, .).
9. In quantum mechanics, the Pauli theory of electron spin utilizes the following
complex matrices:
248 VECTOlt SPACES 4-1
where, of course, i 2 = -1. Prove that the Pauli matrices, together with matrix
multiplication, consititute a group.
10. A square matrix (aii) is said to be diagonal if au - 0 for i pi. j; in other words,
(41/) - (a,8i/). Assuming diag M ,,(F) denotes the set of all diagonal matrices of
order n, prove that the triple (diag M ..(F),+,) forms a commutative subring
of (M ..(F),+, .).
11. By a ICalar matrix is meant any diagonal matrix (4i/) having equal diagonal
elements: (aii) - (diJ). If S ..(F) is the set of scalar matrices of order n over the
field (F,+, .), show that
a) the triple (S.(F),+,) is a field isomorphic to (F,+, .),
b) S .. (F) - cent M .(F). [Hint: Consider the products (41/) E., - E., (41/)']
12. If M:(F) designates the set of nonsingular matrices of order n, verify that the
triple (M:CF), +, .) does not form a subring of the ring (M ..(F), +,.).
13. A square matrix (aii) is upper triangular if ail - 0 for i > j and atricll" upper
triangulm- if ail = 0 for i ~ j. Let T,,(F) and n(F) denote the sets of all upper
triangular and strictly upper triangular matrices of order n, respectively. Prove
the following:
a) A matrix (ail) e T .. (F) is nonsingular if and only if 4i1 '" 0 for i - I , 2, ... , n.
b) The triple (T,,(F),+,.) isasubringofthering (M ..(F),+,).
c) Each matrix (aii) e T:(F) is nilpotent; in particular, (41/)" - o.
d) (T:(F),+,') is an ideal of the ring (T ..(F),+, .).
14. The trampoae of a matrix (41/), designated by (ail)', is the matrix whose ijth
entry is the jith entry of (a,;), that is, (aii)' .. (alf)' Given matrices (ail), (h#) e
M ..(F), verify that
a) (41/)" - (41/), b) [r(ai/) +
.(hi/)]' ., r(ai/)' +
.(hi/)', r,. e F,
c) [(4i/)' (bi /' - (bi/)' (4ii)',
d) whenever (a#) is nonsingular, 80 also ill (a#)', with [(44/)'1- 1 - (4#)-11'.
15. Show that the field (S, +,.) of Example 4-3 is isomorphic to the complex numbers
(C, +, .)
under the mapping
j
I\_b 4 h) =
a
(a, b).
16. In the ring (M2(C),+, .), let D be the set of all matrices having the form
where 4 is the complex conjugate of a. Prove that the triple (D, is a division +, .)
ring, but not a field.
17. Show that for any element a e F, the following matrices are both idempotents in
(Al2(F),+,'):
and
4-2 ELEMENTARY PROPERTIES OF VECTOR SPACES 249
18. A matrix (a.;) E M,,(F') is said to be aymmetric if (a.;)' - (atj) and ,kew-.ym-
metric if (ai;)' ... -(ail)' Establish the following assertions:
+
a) If (a.;) and (bij) are symmetric matrices, so also is r(ai;) .(bo;).
b) The products (ai;) . (ail)' and (ai;)' . (aij) are both symmetrio.
c) If (aij) and (b ij) are symmetric, then (aij) . (bij) is a symmetrio matrix if and
only if (ail) . (bii) - (bi;) . (ail)'
d) The diagonal elements of a skew-llymmetric matrix are all sero.
e) Every (square) matrix can be written as the sum of a symmetrio and skew-
symmetrio matrix.
19. Let the set 0 be comprised of the following matrices:
y = (0 -1).
-1 0
a) Prove that the pair (0, .) forms a group.
b) If II = {I, 8, T, U}, show that (H, .) is a normal subgroup of (0,) and find
the cosets of H in O.
While the addition symhol has been used in two contexts in the above defini-
tion, to d!'signate the operation of the group and one of the operations of the
fi!'ld, no eOllfllsioll should arise from thiN pmdil'e. It will alwaYII he dear in
any giv('n :-;it.uatioll wlwtlwr v('(t.ors or sealars are being added. When both
veetor additioll and sealar multipliention are involved in an expression, we
follow our usual understanding in omitting parentheses: multiplication takes
precedence over addition.
In the Sl'qu!'l, a vector RP:WC over the fi!'ld (F, +,.) will be denoted merely
by l"(F), rather than the corred but cumbersome notation (V, +), (F, +, .),0).
The ('onveniellC'e rmmlt.ing from this convention more than outweighs its lack
of pl'('('i8ion. For fu.-Owr simplicity, we shall hel'('aftcr drop the and write cx
0
In view of th(' r('sults of the lnst. seetion, we thus obtain Il vector space, which
will hl'nc~pfOl'f.h hI' denoted simply hy 1" "(F).
Example 4-6. If /tf", .. (F) rcpre'spnts t.hl' c'oll('dioll of m X n matriccI-I over
(P, +, .)
and I is t.he op('rntion of matrix additioll, thl'n (At mn(P), +) is a
4-2 ELEMENTAItV PROPERTIES OF VECTOR SPACES 251
ceF.
The particulur v('(:tor lipaee whieh arilles when m = 7~ will, in the future, be
indicated by !Ifn(F).
Example 4-7. Given a field (F, +, .), take V to he all functions from an
arbitrary llet X into F. For f, g e V and c e F, define the functions f + g and
cf by spe(!ifying their vnlues at each point of X:
and
where m ~ n, then
p(x) + q(x) = + bo) + (a1 + b 1)x + ... + (a... + b...}x"',
(ao
cp(x) = (c ao) + (c a1)x + ... + (c a,.)x".
(Needless to say, ill the sum p(x) + q(x), we sct ak = 0 for n + 1 :S k :S m.)
Let us retain the symbol F[x) for this vector space, in preference to the correct
but awkward (F[x])(F).
Example 4-9. For any field (F, +, .), Example 4-5 may be generalized by
using infinitdy many elenumts of F: jUtit. take V .. (ft') to be the totality of all
illfiniw 1IC<lIlClJ(!(lS from F. The members of If.. W) may be eonvcniently identified
wit.1I row V('(:\.Ofti
ak e F,
If (alt a2, aa) and (b l , b2, ba) are arbitrary elements of W, then their sum
(al + bit a2 + b2, aa + ba) is such that
(al + bl) + (a2 + b2) + (aa + ba) = (al + a2 + aa) + (bl + b2 + hi)
= 0+0 = O.
254 .
VECTOR SPACES 4-2
The two conditions of Definition 4-12 may be combined into a single easily
applied criterion.
Theorem 4-8. W(F) is a subspace of the vector space V(F) if and only if
o W ;;; V and ex +
dy E W whenever x, YEW, e, d E F.
Proof. If W(F) is a suhspace, t.hen hy definition W is nonempty and contains
ex + fly for all x, YEW, e, dE F. Conversely, if thi!! condition holds, W must
contain Ix +- ty = x + y and ex + Oy = ex for every x, YEW, C E F. Accord-
ingly, W is c1osl'd wit.h rl"slled In t.lm vl'dor HIl'U~!l ClI)(!rUUoIIH.
We next consider operations on the subspaces of a vector space that produce
other subspaccs, the mOlolt important of which are sum and intersection. If U
and W are nonempt.y suhsets of thc veetor space V(F), their (linear) sum is
defined to be the set .
IT +W = {u + W I U Ell, W E W}.
The following simpl<', hut quite uS<'ful, fact. will be needed on several occasions.
Theorem 4-9. If P(F) and W(F) arc suhspace!! of the vector space V(F),
then (If +W)(Ji') is also a subspace.
Proof. The sum l! + W is plainly not empty, for U and W each contain the
zero vector, hence 0 = 0 + 0 E U + W. Suppose x and y are arbitrary vectors
in U + W. Th!'n x = UI + WI and y = U2 + 102, where Ui E U, Wi E W. For
scalars c, d E F,
ex + dy = c(?lJ+ WI) + d(U2 + W2)
= (CUi + dU2) + (ewJ + dW2).
4-2 ELEMENTARY PROPERTIES OF VECTOR SPACES 255
alHl WW) are hoth to!UhHPIlCI!H of V(ft'), it followH Umt CUI + dU2 E l!
Hj'l!'l~ (l(ft')
and CWI + dW2 E W. Thus cx + dy is again in U + W, implying that
(U + W)(F) is a subspace.
From our earlier work with other subsystems, one would expect the inter-
section of subspaces to be also a subspace and, indeed, this is the content of
the next theorem. As it would be repetitious to present the details again, tpe
proof is left to the reader.
Theorem 4-10. If Wi(F) ito! an indexell eollcction of to!uhsplwes of the vector
Kpuce V(I<'), then (nWi)(I<') is a :mbspace of V (1<').
In analogy with the corresponding ideas for groups and rings, this result
provides an effective means for generating subspaces: Given a nonempty subset
8 of the vector space V(F), we define
There is at least one subspace containing 8, namely the whole space V(F), 80
that [8] c(~rtainly exists and is unique. We infer at once from Theorem 4-10 that
[S](F) is a subspace of V(F), called the subspace generatell or spanned by the
sct, S. More important still, [SJ(F) is the smalIl'st :mbHpttee containing S, in the
scnHe of being ineluded in every subspace which contaim; S.
1'he coming theorem gives an alternative and more constructive d('J~criptioIl
of rSJ(/i'); before obtaining this, one more definition is required.
Definition 4-13. Huppose V(F) is It vector spaee nnd X" X2, ,X" E V.
AllY !inite MUIlI of thc form
[SI={ECkXkICkEF,XkEs,nEz+}.
256 VECTOR SPACES 4-2
Proof. For the moment, let us denote the set on the right by lin S. Evidently
a linear combination of lin~llOmbinations of elements from S can again be
written as a linear combination of members of S, so (lin S)(F) is itself a sub-
space of V(F). Moreover, for each % e S, we have % = be lin S, hence
S !; lin S. As (8)(F) is the smallest subspace of V(F) to contain S, it follows
that (8) s;;; lin S. On the other hand, since [SJ(F) is a subspace containing S, it
must contain all linear combinations of elements of S. Accordingly, the in-
clusion lin S ~ (8) also holds and the proof is complete.
Exampl.4-14. Let S be the subset of V,,(F) whose elements are the n vectors
~l =
(l,O, .. , 0), e2 =
(O, 1, ... , 0), ... , en = CO, 0, .. , 1); in general, e. is
the vector with 1 in the kth component position and 0 elsewhere:
k = 1,2, ... , n.
Here [S](F), the subspace spanned by these vectors, is all of V,,(F). Indeed,
for any n-tuple (alo a2, .. , a,.) over F, we have
(alo a2, . ,a,,) = + a2(O, 1, ... ,0) + ... + a,.(0, 0, ... , 1)
al(l,O, .. ,0)
= aiel + a~2 + ... + a,.e" e (8).
Let us lee what happens when we generalize this situation to the eequenee
space V.(F). In this context, e. now denotes the infinite aequence whose kth
term is I, while all other terms are 0:
k = 1,2, ...
As before, let S be the set of all elc. Recall that in forming the linear span [8],
only finite linear combinations of elements of S are utilized. It follows therefore
that [8](F) is not V .. (F), but rather the (proper) subspace consisting of those
vectors having only a finite number of nonzero entries:
aeF,
4-2 ELEMENTARY PROPERTIES OF VECTOR SPACES 267
beF.
That U(F) and W(F) are both subspaccs of M 2(F) is easily verified. Note,
however, that while the matrices
and
( 1 0) + ( 0 1) _ ( 1 1)
o 1 -1 0 -1 1
and
respectively, we have
258 VECTOR SPACES 4-2
To see thill, one 11('1'<\ only observe that each of the matrices on the right can be
written 111'1
Ui E U, Wi E W.
Then UI - U2 = 11)2 - W\. But the left side is in II and the righ,t; side is in W,
so both sides belong to l! n W. It follows therefore that
or
UjE U, wiEW.
Subtracting, we obtain
4-2 ELEMENTARY PROPERTIES m' VECTOIt SPACES 259
It is not difficult to establish that the subspaces W(F) and W'(F) are both
tOmplementary to U(F).
What is quite signifieant is that every subspace ha.~ eomplementary sub-
spaces, which, to be sure, are not necessarily unique. For the proof of this
assertion, we resort to the powerful tool of Zom's Lemma (sec p. 185).
Theorem 4-15. Every subspace W(F) of the vector space V(F) has a com-
plementary subspace. -
Proof. Define a family of suhsets of V by taking
This family is nonempty for, trivially, the set {OJ satisfies the defining proper-
ties.
Now, consider any chain {Si} iu 5'. Our object, of course, is to show that
US; is again a member of 5'. To achieve this, let x, y E USi and e, d E F. Then
there exist indices i and j for which x E Si, Y E Sj. Because the collection {S;}
forms a chain, either S; ~ Sj or else Sj ~ Si; say, Si ~ Sj, so that both x, y E Sj.
Since Sj(F) is a subspa('c of V(F), we must then have ex +
dy E Sj ~ uS;.
This proves that (uS;)(F') forms a subspaee of Yep). Moreover,
Thus, on the basis of Zom's Lemma, the family fF contains a maximal element
U. Our contention is that the subspace U(F) is complementary to W(F). In
+
order to establish this, it suffices merely to show that U W = V. To the
contrary, suppose there is some vector x e V with x ft U + W. But then
[U, z)(F), the subspace generated by U and x, will be disjoint from W(F), in
the sense that [U, xl n W = {O}. (Verify this!) Since U is a proper subset of
[U, x), we have arrived at the contradiction to the maximality of U in If.
Accordingly, V = U + W, and the proof is complete.
The foregoing ideas are easily generalized to finitely many subspa.ces:
If WI (F), W 2(F), ... , W,,(F) are each subspaces of V(F), their sum is defined
exactly as for two llubllpMClIl,
x = WI + W2 + ... + WIt
with Wi e W (i = 1,2, ... , n).
The concept of a quotient structure carries over to vector spaces as expected.
In the present context, we encounter one slight, but highly important difference.
When fonning quotient groups and quotient rings, it was necessary to introduce
speoial lubaYltcm" (namely, nonnal aub/iCroupa, and ideals) in order to enlure
that the operations of the quotient structure were well-defined. For vector
spaces, no such distinguished subsystem need be defined.
To be more concrete, let W(F) be an arbitrary subspace of the 'vector space
V(F). Since (W, +) is a subgroup of the commutative group (V, +), we may
fonn the quotient group (V /W, +). The elements of this group are just the
cosets x + W, with coset addition given by
(x + W) + (y + W) = x + y + W.
To equip V /W with the structure of a vector space, a notion of multiplication
by scalars is introduced by taking
e(x + W) = ex + W, ceF.
As usual, we must first satisfy ourselves that scalar multiplication is un-
ambiguously defined, depending only on the coset x + Wand scalar e e F.
4-2 ELEMENTARY PROPERTIES OF VEeroR SPACIIlS 261
then (V /W)(F) is itself a vector spaoe, known u.s the quotient trpace of V by W.
Example 4-15. For a concrete example, consider the quotient space
(V 3/W )(R'), where
W = {(w, 0, 0) IW E R'}.
PROBLEMS
2. Prove that in any vector space V(P), if :1:, 'Y E V and c E P, then
a) c(x - y) = cx - cy,
b) ex = 0 implies eithl'r c ... 0 or:ll "" O.
3. In any vector spac,' 11(/<'), Hhow that the following cancellation laws hold:
a) If x E l" wit.h .r ~ 0, t.hen CIX "" C2X implies CI '" C2.
b) If x, y art' nonzl'm I'It'ments of V, tht'n cx = cy with c ~ 0 implies z == y.
4. Prove that W(/<') is a subspace of the vector space V(P) if and only if ~ " JV!;;; V
anll ax + /1 E \I" for 1\11 x, 11 E W, a E !-'.
ri. I,'or I'lI.l'h or till' rollowiug ,mlll \1', deu!rminc wlwl.llI'r I\' Ut') ill a 8ubllpacc of the
vector KpacO V nUt').
a) W = {(aI, a2, ... ,an) I al + + ... -/-
a2 a" ~ O}
b) Jr = {(ai, a2, ... ,an) I al = a2 = ... = an}
c) Jr = {(aI, a2, .. , an) I ala2 = O}
d) II' = {(ai, a2, .. ,an) I ak E Z for all k}
6. Let U(F) and II'(/") he slIhl'lPI1Ct'H of the veel,or "pace V(F). Prove that (U U Jr)(Ji')
forms a Hubspace of V(F) ir and only if U!;;; \I' or II' ~ U.
7. Determine whether II' (1<') ill a subspace of the indicated vector space:
a) V 3 (F); for fixt'd scalars aI, a2, a3 E F,
b) l'",(F); II' consists of sequt'nces where all but a finite number of the terms are
equal to zero, that is,
V(>rify that VaW) is Ow dircet slim of the lIuhspacl'f! U(/") and lI"(F), that is,
Va=UEf>W.
14. Let V(R') be the vector space of all functions from R' into itself and
Perhaps the most far reaching notion in the study of vector spa(!es and the
prindpal t1wme of this 1I(,dion ill thnt of a hllllis. As will 1:10011 be evident, this
concept is preeisely the tool needed to formalize the more or less intuitive idea
of what is meant by the dimension of a space. A convenient starting point is
the following definition.
Deftnition 4-16. Let V(F) be II. vector space over the field F. A finite set
{X., X2, , Xn} of vectors from V is said to be linearly dependent (over F)
if the zero vector is a nontrivial linear combination of x.,
X2, , X,,; other-
wise, the set {XI, X2, , xn} is termed linearly independent.
Aecording to our dnfinition, {xt. X2, ,xn } ill a linearly uepenuent set if
and only if there exist scalars c., C2, , r.n E /<', lIot all of whi('h are zero, such
that
CI = C2 = ... = Cn = O.
264 . VIICl'OR SPACES
1 + x+ 2x 2 , 2 - X+XlI, -4 + 5z + xli
of degree two. Since
tiel + C2tll + ... + cne" = (CII 0, ... , 0) + (0, C2, , 0) + ... +iO, 0, . , c,.)
= (CII C2,"" cn);
hence the condition I:Z-I cj:ej: = (0,0, ... ,0) implies Cl = ClI = ... = c,. = O.
The essence of this concept lies in the fact that if {XII XI, ... ,z..} is a linearly
independent set, then any vector X e [XII XI, ... , z,,) is uniquely expressible as
a linear combination of the xj:'s. Indecd, suppose there are two such repreeenta-
tiona,
X = bixi + +
bllZlI + ... + bnz.. = CIXI ClIZlI + ... + c,.z..i
The hypothesized linear independence of the vectors XI, Xli, .. ,z.. then (orces
or (or all k.
where not all the c's are zero. Let k be the largest integer for which CAl O. If
k = 1, we would have C1Xl = 0, hence Xl = 0, contrary to assumption. Thus
k> 1 and
wiwrI' 1 ~ it < i2 < ... < im ~ n, of the original set of n vectors, still having
lilwnr spall V aJl(1 su(,h thnt no Xi. is a linear combination of the preceding vec-
t.ors. That the set {Xi" Ii 2 , , ,Xi.. } is linearly independent follows immediately
from Theorem 4-17,
Example 4-18. As a simple illustration of some of t1wsc ideas, let us observe
that
l' 3 = [(1, 1,0), (1,0, 1), (0, 1,0), (1, 1, 1)].
The rml.soninJl; h('re is jmltifif!d by t.hc fa('t t.hat. e:wh of t1w unit, vectors et. e2, ea
iM It lillcllr ('(lIllhillul.ioll of I.JIIMI~ Vl'dOrM:
Since V 3(P) is spanned by the ullit vedors, we infer that every clement of V 3
I\lUSt. 1)(':\ lilwar ('omhinatioll of (I, 1,0), (1,0,1), (0,1,0), (1,1,1),
The lilJ('ar dl'l)('JI(\I'III'(~ or iIlC\I'JlC'tHII'II('c of t.his SC't. of ve(~t.ors is equivalent
III Ilw exist.ellc'(! or 1I00I1'xisll'lI('(' of S('lIl:trl; (~I, 1'2, ra, (:4 (1I0t. nil zero) sudt thnt
whi('h, of eounw., impli!'s that t.he given set of gem'rut,or!! is linearly dependent.
To obtain a linearly independent subset of the aforementioned vectors, we
need only removc the element (1,1,1). In fact, (1, 1,1) is the only vector
that can be written as a linear combination of those preceding it:
wh('nec
Va = [(1,1,0), (1,0,1), (0, 1,0)1.
The next, somewhat technical, result is the key to all that follows.
theorem 4-19. (Steinitz Replacement Theorem). Let W(F) be 8. finitely
gellerau~d subsplwe of t.Il(, vc(:tor I'lpnce V(F), W = Ix., :1:2, ,xnl, and let
{Y1o Y2, .. , Ym} be any linearly independent subset of W. Then m of the
Xt'S, say X10 X2, , X m, may be replaced by Y .. Y2, . ,Ym, so that
W = [y., . .. ,Ym, Xm+1o'" ,xnl, in particular, m ~ n.
Proof. Since {X., X2, , x,.} spans W(F), the vector Yl E W can be expressed
as 8. linear combination of the Xk'S:
be zero, for this would imply that b l r& 0 and, in tum, that
bl 1l1 + (-1)111 + Oys + ... + 011", = 0,
contrary to the independence hypothesis. Hence, one of the coefficients
bl, ... , b" is nonzero; let us, for simplicity, take this to be bl. As before, we
can solve for X2 in terms of the vectors YIo Y2, Xa, . , x" to obtain
W = [1Il> YI, Xa, ... , x,.1.
Continue in this manner: At each stage a y-vector can be introduced and an
x-vector deleted so that the new set still spans W(F). If m were larger than n,
after n steps all the Xk'S could be removed and the set {1Il> 112, . , 11..} would
span W(F). Accordingly, the vector 11"+1 E W could be written as a linear
combination
11"+1 = CIYI + C2112 + ... + c,.1I"
with not all the c" being zero, since 11,,+1 po! O. Once again a contradiction to
the independence of 111, Y2, . ,11.. would arise. It follows that the 1I'S must be
exhausted before the x's (that is, n ~ m) and W = [YIo , Y"., x..+1, , x ..1.
Of course, the possibility that n = m is not excluded; in this situation, the set
{x..+1, , x,.} is empty and the vectors 111, YI, .. ,1/" themselves span W(F).
Corollary. If a vector space V(F) is spanned by n vectors, thea any set of
n +1 vectors from V is linearly dependent; in particular, any n 1 vectors +
of the n-tuple space V,,(F) are dependent.
Proof. A set of n + 1 linearly independent vectors from V would be impossible,
since the theorem would imply n ~ n + 1, an obvious contradiction.
The Replacement Theorem has several notable consequences, but these will
have to await further developments.
DefInition 4-17. A bam for a vector space V(F) is a linearly independent
subset of V that spans the entire space V(F).
Example 4-19. The familiar unit vectors e" e2, ... , ell of V,,(F), the space of
n-tuples of scalars, form a basis (see Examples 4-14 and 4-17). Hereafter, we
shall refer to this particular ba.sis as the natural or standard ba8i8 for V,,(F).
Example 4-20. For a more general example, consider the vector space V.(F)
of infinite sequences of elements from F. If
k = 1,2, ... ,
then the set 8 = {e" e2, ...} is linearly independent. These vectors do not
constitute a basis for V.. (F), however, since it was shown earlier that the linear
span [81 is a proper subset of V.
Example 4-21. Let M,,(F) be the vector space of n X n matrices over a field
F. This space has a basis consisting of the n 2 matrices EiJr where EiJ is the
4-3 BASES AND DIMENSION 269
square matrix of order n having 1 as its ijth entry and zeros elsewhere. Any
matrix (aij) E M .. (F) can obviously be written as
the relation
constitute a second basis for Va(F). In this case, an arbitrary vector (alJ all, 4a)
of Va can be written as
exists in which not all the cocffieicnts are zero. Were a = 0, we would contradict
th~ lill<'ar indl'II('II<i!'nc'e of S; h(,llee a "" 0, whi(~h le/HIR to
or x E [.'I). WI' eOlll'hlll(' from this that t.he lin('ar spnll of 8 is the whole folpac('
V(F), finiHhing the proof.
The ItrgllJnf'nt lIS<'d in Th('orem 4-:W slIgg('st,s t.he following ehll.rnderizntion
of a hasis: S is :l hasis ror 1,111, Vf'C'lor Spill'" reF) if llllli only if S i~ IL nlllxilllal
linearly in<iI'pl'IIIIPnt suhs('t of V (maxilllal, in the sense t.hat any foluhset of V
which properly ('ollt.ains S is itself linearly dl'p('ndent). One direction having
alrcady h('('n ('stahlislw<i, let. liS ohs('rv(' t,hat if S is a basis for V(F), thcn any
set. S U {x}, wherl' x E V -- S, is linearly d('pendent.. In fact, since V = IS),
4-3 DASES AND DIMENSION 271
for some finite set {XI, X2, ,Xn } of vectors from S and suitable scalars
CIcE F. But then
The proof of til!' t.heon'nl is rOlltin<', and we 1<'lLve til(' det.ails to the reader.
WI' arl' 1I0W pnpan1I III ('iiI ahlilih tllP Illain t.hl'lIl'('nl of I his iiI,('tion: all hlU!<!s
for It givl'lI vl~etor spal'(! have 1.111' SILIIIP nlllllher of dl'nU'lIt.s (for the IIstute
reader', t.he sam!' eurdilllllity). As the proof ill full g(ncmlit.y is somcwhut
sophh;tieated, we shall deal with the finitl' easc only.
Theorem 4-22. If V(P) is u. finitely g<'lH'rtlted nOllzero vee tor spaee, then
any two bases for V(F) have precisely the same number of clements.
Proof. Let {XI, :r2, ... , xn) he one basis for thl' spaee V(F) and {YI, Y2,' .. ,Ym}
be unother. Beeause the vedors YI, !/2, ... , 11m are linearly independent and
V(F) is spannell hy :rl, X2, .. , Xn , it follows from t.h(~ Rephwcmcnt Theorem
t.hat t.he integer m is no larger than n, tlmt. is, m ~ n. Simply revertling the
roles of {XI, X2, , .. ,xn } and {Yb Y2, .. , Ym) in the argument, we also con-
dude n ~ In, whClII'e n = m.
Theorem 4-22 ean be used to define dimenllion. For if II vedor space V(F)
has a finite basis, then every ballill for the IlJlace wiII have the sallle finite number
of clemen til. This unique integer is called t.he dimension of V(F) anti dellignatcd
by dim V. Although UII' notion of hmlis iH IlOt ltpplimhlc to the z(~\'O space, it
is (~IIKt.omary t.o t.n'llt. :0) (P) lUI finit.e-dinwllliiollal, wil h Zf'ro dimellsion. A
(nollzcro) v('ctor sJllwe ill ~tid to be injinite-dill/.fIt.~il}Ttal if it. ill not spanned by
any fillite sllhiil~t., that. is, if it iH not finite-dimellsiollal. Not every spnce is of
finil.t, dilll('IISioll, II..'! ('videlll(d hy F[.rj, t.I\(' vpdo\' KpIU'(' of' polynomials ill x.
III lIIost of what. follows, we lihall for 1.1", sakI' of I(llll'ppl.llal HilllJlli"ily eOllfille
ollr alll'lItioll t.o fillitl!-dilll('IIHiolllll VI'I'I.OI liJlIU~I'S.
'1'111' IIl'xl I hl'fl 1'('11 I illdi":lII'K t.Iml al II':llit r'JI' lillill'-,liIlII'IIHioll:l1 HJlIU'I'H, Oil'
"lUI f01'1II haiil'" without. 100 I11I1I'h dilli,ulty.
Proof. Suppose V = [Zl, Z2, .. ,Zt&J. According to Theorem 4-18, some sub-
eet of {Zh Z2, ... ,Z,.} is a basis for V(F). But the previous result implies that
this subeet must contain n elements and accordingly is the entire eet
{Zh Z2, ... , z,.}.
For a proof of the second assertion, let Zh Z2, ... ,Z.. be n linearly inde-
pendent vectors of V. If Z E V is arbitrary, the set {z, Zh ... , z,.} is dependent,
since the mwcimalllumoor of linearly independent vectors in V is n. Thus some
nontrivial dependence relation exists among theee vectors:
making the eet {Zit Z~h ... ,z,.} a ; ; . for V(F). ...
Theorem 4-18 told us that it is ble to chooee a basis for a vector space
V(F) from any set of generators of V. In the opposite direction, the coming
theorem asserts that any linearly independent subset of V is either a basis or
else can be extended to a basis for V(F), in the sense that vectors may be added
to it to fonn a basis.
Theorem 4-24. If V(F) is a finite-dimensional vector space and {ZltZ" ... ,Z,.}
is a linearly independent subset of V, then there exist vectors 1I,.+h ... ,II...
such that {Zl, ... ,.z,., 1I.. +h ... ,II..} forms a basis for V(F).
Proof. The proof is short: Since V(F) is finite-dimensional, it has a finite basis
IIh IIllt .. ,II.. As V = [lIh 112, ... , II..], we may apply the Steinits Theorem
to replace n of the II's by z's and obtain a eet {ZI, . , z,., 1I"+l, . , II..} whose
linear span is atill V. But any generatin, set with m elements in an m-dimen-
sional space is a basis.
Corollary. Every basis for a subspace of a finite-dimensional vector space
can be extended to a basis for the entire space. .
Example 4-23. As a particular case of this last point, consider the vector
space M,(F) and the three matrices
4-3 BABES AND DDlJ!lNBION 273
Cs - 0,
which implies CI = C2 = Ca = O.
Theorem 4-24 now tells us that the given matrices will form at least part of
a basis for M 2(F). We shall leave the details to the reader and content ourselves
with this one comment: the addition of any fourth matrix having a nonzero
element in the (2, 2)-position will yield a linearly independent set. For instance,
among the many possibilities, the matrices
are linearly independent and, being four in number, must therefore be a basis
for the 4-dimensional space M 2(F). This illustrates the wide latitude of choice
for bases of M 2(F).
We terminate the present section by giving some useful results concerning
the dimensions of subspaces of a given vector space. The reader should first
prove the following theorem.
Theorem 4-25. If W(n is a subspace of a finite-dimensional vector space
V(F), then
1) W(F) is also finite-dimensional with dim W S dim V.
2) dim W = dim V if and only if W = V.
The next theorem relate~ the dimensions of the sum and intersection of two
subspaces.
Theorem 4-26. If U(F) and W(n are 8ubspaces of a finite-dimensional
vector I!plwe V(F), then
ing to Theorem 4-24, there will exist vectors u., U2, , u'" such that
{X., .. , Xn, u., ... , u"'} is a basis for the subspace U(F) and vectors
1/'., 11'2, , //', ~\wh that {Xb' , X"' WI, , w,} i~ a bu.sis for the subspace
W(F). Combine these two bases into a single set
Becausc the finlt m + n vectors of the foregoing set are a basis for U(F) alld
th(' last TI -I r vect.ors are a basis for W(F), any c1f1ment of the sum (! + W
may I)(~ cxprN;sc!\ ItS It lincar eomhirmtion of the vectors of thi" sct, that is
so that
But the set {XI, , X"' till, ... , til,} is linearly independent, being a basis for
the subspace lV(F), hell(~e d l = ... = dn = CI = ... = Cr = O. In con-
junction wit-h the IinC'ar indl'pcIHI!'n('e of fx., ... , Xn, U" , u m }, this fon~es
al = ... = am = II. = ... = b" = O. We havl' thus sueeccded in provinj!;
that t.I\I~ vcclol"ll I l l , " " Itn" .r" ... , .r", w" ... ,
!t., arc linearly independent,
!l.S required.
Proof. Let {Xl. X2, ,XII} be a basis for W(F). By adding vectors
Yt. Y2, .. , y"" we may extend this set to a baliis {x!, ... , X,,, Yh . ,y",} for
the whole !!pacc V(/t'). Any vector x E V can then be written in the form
x = atXl + ... + a"x" +
b1Yl + ... +
b",y", for appropriate choice of co-
efficients. Since x - (b'Yl + ... +
b",y",) E W, the coset x + W is expressible
as
In othn!" words, i1w demcnl.li Yl + W, 712 + W, ... ,!1m + W span the quotient
Rpuee (V /W)(P).
The remainder of t.he proof amounts to showing th('HC coset!! to be linearly
independent and heIl(~e a basi!! for (V /W)(F). To see thill, we suppose
But the lincar indepcndence of the set {x!", . , x"' YI, .. , y",} then implies
1:, ~ . = em = (I, = ... = till = O.
The foregoing urgulllcnt. indicates that the quotient space (V /W)(F) has a
basis eonsisting of the m cosets Yl + W, Y2 + W, ... , y", + W. The conclusion
of Uw U\('orem followH imml'diutcly from this, sirwil
W = {(a, 0, 0) I a E F}.
In this case, the quotient space (Va/W)(F) has dimension 2, for the equation
of Theorem 4-27 reads
then serve as a basis for (Va/W)(F). Indeed, it is easy to show that any element
+
(CllI Cl2, CIa) W of Va/W m&y be written in the form
PROIUMS
10. Assume the space V(n is finite-dimensional with basis {~l, ~2, ... ,z,,}. If
W.(n is the subspace generated by the vector Zk (Ie - I, 2, ... , n), verify that
V - WIE9 W2E9 . E9 W".
11. Let U(n and W(n be subspaces of V,,(F) such that dim U > n/2, dim lV > ,,/2.
Show Un w pi {OJ.
12. Suppose {~l, ~2, ... ,~,,} is a basis for the subspace U(n of V(F) and
{Ill, 112, ,II.. } is a basis of the subspace W(F). Given that the set
{~l, ... , X", Ill, .. ,II..} forms' a basis for the entire apace V(F), prove that
V - U E9 W.
13. Let (F[x),+, .) be the ring of polynomials in X over F and p(~) E F[~) be a poly-
nomial of degree n. If p(~, +, .) is the principal ideal generated by p(x),
establish that (F[~1I(p(~)(F) is a veotor space of dimension n. [Hint: Consider
the cosets 1 + (p(~, Z + (p(x, ,X,,-1 + (p(x) ).]
14. Determine the dimension of the quotient space (Va/W)(F), where the set W is
defined by W - {(a, h, a+ h) I a, h E F}.
15. In the vector spaCe M,,(n, let T ..(F) be the subspace of upper triangular matrices
[matrices (a.,) such that a., - 0 for i > j) and T~(/") be the subspace of lower
triangular matrices [matrices (ai,) such that a.i == 0 for i < j), Find dim T",
dim T~, dim (T" n n), dim (T.. + T~), and verify the truth of Theorem 4-26
in this particular case.
for all x, Y E V and a, b E F. Suffiee it to say, the above definition makes sense
only when both vector spaces are taken over the same field.
Before proceeding to the theory of linear mappings, let us illustrate to some
extent the great variety of possible examples.
Example 4-25. Let M m,,(F) be the vector space of all m X n matrices over a
field F, and let (ai;) he!\ fixed m X m matrix (again, over F). Define a function
I: 1If rn" -+ 111 mn by
The reader may easily verify that I is a linear mapping from VlI (F) to V..,(F).
In fact, any power of I is again linear, Hinee
Example 4-27. Next consider F[x], the space of polynomials in the indeter-
minant x with coefficient,s from F. A lillear mapping 011 F[x] is given by means
of Uw Ho-{:nlInd dijJerentiat1:on functilln. That. iH, for lUI arhitrary polynomial
p(x) = ao --j- alx + a2x2 + ... -f- a"x" ill ~'[x], let
pletely described hy specifying its values on a basis for V(F). For suppose
{Xli X2 ,;r n } is It basis of the space V(F) j then Mch vector X in V has the
form
Thus,! is completdy determined onee its cITed on the bu.'lis vectors Xl, X2, , Xn
is known. The next. theorem indicates that this effeeL may be prcserihed
nrhil.rllrily.
Theorem 4-28. 1,,'1. :.r I, .r~, ... ;r,,: be! IL bMiH fOl t,lIn fillit~,-clillloIlHi()tll\l
V("'IIII'HIII\I'" 1"(/<') 111111 :111, II~, ,/I,,: 1111 nl"l,il.nLl'Y Hill. Ill' " VC,d.IIIH frllIll
w(/,'). '1'111'11 I.\wn. iH lxlIl!t.ly (lI1() Iillellr II IllppiIIg f E LO', W) sllch thnt
Proof. To prove that there d(){.s exist such a lllapping f, we proceed as follows:
Since {Xb X2, , xn) is u basis for V(F), each vector x E V is uniquely express-
ible in the form
Then
We therefore conclude that the linear mapping I for which I(z,,) - 11" is the
only one possible.
To see how closely linear mappings arc associated with matrices, let us now
suppose the space W(F) is also finite-dimensional, say, with basis {Yh Y2, .. , y",}.
Then each vector I(Xi), being in W, is a unique linear combination of these
basis vectors:
I(Xi)
'" ai;Yi,
= :E (j = 1, 2, ... , n).
i-I
(Observe that, in contrast with many texts, the summation is on the first index
of 4;j.) In this fashion, we produce an m X n matrix (aii) called the matrix
representatiun of I relative to the bases {Xl, X2, ,XII} and {Yt. Y2, .. , y",}.
Notice particularly that the matrix (aii) depends on the pair of bases used for
V(F) and W(F) as well as oni; any change in the basis elements, even in their
order, would lead to a different matrix. Accordingly, the same linear mapping
(whose definition, after all, does not depend on any basis) might be represented
by several matrices.
Now, any vector X E V can be expressed as a linear combination of the basis
{zt, X2, , XII} : . ,
Hence,
Once the bases are picked, the effect of I on each vector of V is therefore com-
pletely determined by the scalars Ci and the representing matrix (4;j).
On the other hand, two bases {Xl, X2, , X..} and {Yt. Y2, ... , y",} for V(F)
and W(F), respectively, along with an m X n matrix (4;/) of elements from F,
give rise to a unique linear transformation IE L(V, W). All we need do is
define the proposed mapping on the basis of V(F) by specifying
Recall also that 1(0) = 0, whence ker (f) oF- 0, and f( -x) = -/(z}.
By this stage, the first two parts of the following theorem should come as
no surprise; their demonstration is not difficult a.nd we ask the reader to fill
in the details.
Theorem 4-29. Let V(F) and W(F) be vector spaces over the field F and
the mapping I e L(V, W). Then
I} (ker (J)(F) i~ a subspace of V(F);
2} (f(V(F) is a subspace of W(F);
3) dim V = dim ker (f) + dimf(V) [Sylvester's Law].
Proof. ,We establish assertion (3) only in the case where dim V is finite. Suppose
first that ker (f) oF- {O}, so that a basis {Xlt X2, , z,.} may be selected for
the subspace {ker (f(F). Using Theorem 4-24, this set can now be extended
to a basis {Xh ,z,., Yl, ... , Y... } for the entire space by adding new vectors
Yl, Y2,"" Y.... Given any element Y e/(V), there will exist some x in V such
that Y = f(x). In terms of the basis for V(F), the vector x is representable as
We maintain further that the vectors f(Yl), I(Y2), .. ,/(y... ) are linearly in-
dependent and consequently a basis for (f(V (F). Assume some linear com-
bination of them is t~e zero vector, say
'.,; .
Since XI, XZ, ,X" are a basis for (ker (f)) (F), there must exist scalars
dJ, ... , d" such that
Were any of th!,H(~ N)('f1i(~i!'ntH nonzero, Il. eontradiction to the linear indepen-
denee of t.he lief, {x I, . , X n, Y I, , Ym} would result. Hencc CI = ... = C m = 0
and we have Slle('('eded in proving that the m vectors f(Yl), f(Y2), ... , f(Ym)
form a basis for the image space (J(V)(F). This means
The kllrncl of J iH el~Hily 1'41l1'l1 to lin thn flCt of VClltOrH ()f tho t,ypc (a, a, a, a),
while itH range is all of V 3' Hcncc,
The terms isomorphil' alHl isomorphism have the obvious meanings for
V(Itor KIIIUC'S. Two V<'dor spaees V(F) and W(F) ILre isomorphic, written
V(F) ~ lV(F), if and only if there exists It one-to-onc mappingf: V -+ Wof V
onto lV whidl preS('rv('s the basic vector space operations; that is to say, we
rcqllin' f E 1.( r. lV). AI'I wit h our earlier Kysiems, Sill'll It funetion f is eltlled un
isomo/'ph iSIll.
The ('ollling t.heorem il'l of ('oJll'li<ierahle interest in that it r('veals the n-tuple
t'pu('(' F ,,(f') to 1)(' the prototype of all n-dimensiotlltl Ve(ltor spu.ces.
4-4 1,INt:AIt MAI'I'INUH
Since the n-tuple of scalars (a" a2, .. , an) is uniquely determined by X (relative
to the given basis), we may define a function f: V - V n by taking
f(x) = (a" a2, ... , an).
A routine (lnl(mlntioll eHtahlhdw8 that this mapping dTe(lts the required iso-
morphism.
Corollary. Two fillil.n-dimnnHionnl vC(!tor HplweH V(/<') unci W(F) arc 180-
morphic if and only if dim V = dim W.
Proof. If dim V = dim W = n, then each spaee is isomorphic to Vn(F), hence
isomorphic t.o one another. Conversely, if V(F) ~ W(F) under the linear
mapping f, Sylvester's Law yields
A remark in passing: The isomorphism between V(F) and V.. (F) was obtained
by arbitrarily choosing one particular basis for V(F) in preference to all others.
If we were to confine further study of finite-dimellsional vector spaces ex-
clusively to the n-tuple spaces V.. (F), as the above theorem suggests, we would
always be restricted to a prescribed basis. Since this would be contrary to our
policy of giving, so far as possible, a basis-free treatment of vector spaces, we
shnll for thn most part ignore Thl.'Orem 4-30.
The corollnry 1.0 Thnorem 4-:JO ennblm~ 11M to establish I~ relationship between
qllotilllij. HPIU!()H nllIl (!omplmnelltnry HuhHpaC()H.
Theorem 4-31. Let U(P) and WW) be c:omplemnntnry Hubspaces relative
to th() finite-dimensional ve(,tor spac:e V(F); that ill, V = II 9 W. Then
(V /W)(F) ~ U(F).
Proof. We nlretuiy know from Theorem 4-26 that dim V = dim U + dim W.
Appealing to Theorem 4-27,
!ii(X,,) = ai"Yi = {o .
if i '" k,
.
Yi If ,== k.
The contention is that these nm functions serve as a basis for the space L(V, W).
4-4 LINEAR MA.PPINGS
First, let f be an arbitrary linear transfonnation from V(F) into W(F). For
each index i (1 ~ i ~ n), fez,) lies in W, hence is a linear combination of the
basis vectors 11" 112, , 11"" say
G'I eF.
In this fashion, we produce a set of nm scalars Gil; the problem is now to IIhow
that f can be represented as
Evaluating the right-hand side of this equation at the basis elements Zll leads to
'" GlliYi =
=~ I(XII).
i-I
Since the linear mappings I and Ei.l (E:.. l t!.tj/ii) both have the same effect
on the basis of V(F), they must be identical. From this, we conclude that the
functions hi span the space T.I(V, W)(F). .
To establish the linear independence of the hi> suppose some linear oombina-
tion of them is the zero mapping: .
(Cij E F).
HI riding the imllJl;l' "'PIWI', we arc ahle to introduce stilll~nother hinnry operation:
composition of mappings. Indeed, if j, g E L(V, V), it is ensy to see that
n n n
= 2: ajp:j + 2: bjjxj = 2: (ajj + bjj)xj.
i ...... l i-I i==1
(f g)(Xi)
0 = f(g(xj) = f ( t bkiXk)
k-I
= t
k-1
bki!(Xk) = t
k_l
bkj(t aikXi).
i_I
By reversing the order of summation, this can bc expr<lsliCd ill the form
Thus, the ijth entry of the matrix t\Ssociated with f (J is IiCCIl to be I:t=1 aitb"iI
which is alllo the ljt.h ent.ry of the product (a'j) . (bij ). Ac(~ordingly, we conclude
that
<I>(f. g) = (t k=1
a,kbk j ) = (a'i) . (b i ;) = <I>(f) . <I>(g),
fAV, V)(F) mul AI .. (F) IIIUSt. also I}(' isolllorphi(' liS vedOl' splwell; in fact, a
stmightJorw/1J"(1 ('alcuilltion shows the funet.ioll <I> outaill(d in Theorem ..-a!) to
be a Vl'ctor space isomorphism, not only a ring iHolllorphism.
The structure of L(V, V) can be approached f!"Om olle more direction. For
this, let liS define what. is meallt hy all (assoeiativ(~) aJI(Jwa over a field.
Deftnition 4-19. A vee tor SPIU'!' V(F) i:,llittid t.o he lUI alyeill"a ovel' tltefielrl F
if its elements can be multiplied in Iml'h It way that I'(F) becomes a ring in
VEC'rOa SPACES
Our immediate goal is to establish that, for any vector space V{F), the ring
(L(V, V), +,0) has the structure of an algebra over F. In light of earlier results,
all that really need be demonstrated here is the mixed assoeiative law. To
obtain this, we let e e F,f e L(V, V), and % E V; a straightforward calculation
then yields
y- _-(0o 0) , 1
4-4 LINEAR MAPPINGS 289
each element of T,(R') may be expressed in the form
(: :) = al + bX + (c - a)Y.
Now, if I is any nontrivial homomorphism from the algebra (T 2 (R'), +,.) into
the algebra (R', +.. ), then
Since I(T) = I(T) . l{l) for all T E T,(R'), and I is not identically zero, we
must have 1(1) = 1. What can be said about the functional value I(Y)? Ob-
serve first that the matrix Y is idempotent with respect to multiplication; this
means
I(Y) = I(Y') = I(Y)',
1(: :) = a or c.
This suggests consideration of mappings /;: T 2 (R') --. R' (i = 1,2) defined
by
sum decomposition
v= ker (f) C!> f(V).
Proof. We start hy showing that the function fw, as defined in the statement
of the theorem, is actually in [,(V, V). Suppose x, Y E V, so that x = UI + Wit
Y = u21 '/1'2, with 14" E P, Wt E W (k = 1,2). Theil for scalars a, b,
The reader may verify that t.he corresponding sllbspaces are actually comple-
mentary.
Throughout the remainder of this section, we restrict our attention to linear
mappings on V (F) which assume values in the associated scalar field F [if
scalar multiplication is defined to be the multiplication in F, then F may be
viewed as /I. ono-dim(lnsional Ve(ltor space over itself]. It is customary to use a
special terminology for such scalar-valued functions and to refer to them 8.B
linear functionals, or merely functionals, on V(F).
Definition 4-20. The vector spaee L( V, F)(F) eonsisting of all linear fune-
tionals on V(F) is called the dual space (often called the conjugate space)
of V(F) and is denoted by V*(F).
Example 4-31. Let (a;j) be an n X n matrix. The sum of all elements on the
main diagonal is known as the trace of the matrix and is represented by tr (aij):
n
tr (aij) = 1: au
k_1
It is not difficult to verify that the function tr: M n -+ F, defined in this way,
is a linenr functional 011 the spnce 111n(F); for, if (aij) , (b ij) are two matrices of
order n alld r, 8 E F, then
n
t.r [r(aij) +- s(b ij) J = tr (raij + sb ij) = 1: (rau + sbu )
k_1
= r t
k-l
au +s t
k_1
bu = r tr (ai;) + 8 tr (b i ;).
292 VECTOR SPACES
From Theorem 4-33 we already know something about the space V*(F) j
namely, that if V(F) is finite-:dimensional, then dim V* = dim V. Moreover,
the proof of this theorem provides an explicit way of constructing a basis of
V*(F): Given a particular basis {Xl. X2,' ,x..} for V(F), a unique linear
functional h is defined on V by prescribing
Theorem 4-33 then tells us that the n functions It, f 2, , f" form a basis for
V*(F), the so-called dual bC&8ie to XIo X2, , x.. or dual basis of V*(F).
The definition of h is somewhat opaque, but it may be interpreted in a way
that is more illuminating. Since any vector x E V is of the form x = alXl +
aJX2 + ... + a"x", it is clear that
. .
hex) = L a;!;(x;) = ;-1
;-1
L a;6i; = ai
From this, we see that the functional h has the effect of assigning to each
vector x the coefficient of Xi in the representation of x as a linear combination
of the basis vectors. A suggestive name for the mappings f, might be to call
them coordinat.e functional8 (with resPect to a fixed basis, of course).
An interesting observation which we shall use almost immec!jately is the
following.
Lemma. If V(F) is a finite-dimensional vector space and x is any nonzero
vector of V, then there is some linear functional f e V* for which f(x) po!! O.
Proof. Since x P&!! 0, there exists a basis {Xl, X2, , x..} of V(F) with Xl = X
(Theorem 4-24). tf {fir f2, .. ,f.. } is the corresponding dual basis for V*(F),
we then have
Since the dual space V*(F) is a vector space, V*(F) has a dual of its own;
we shall denote this latter space by V**(F) and refer to it as the second dual
of V(F). If V(F) is of finite dimension n, we know that V*(F) has the same
dimension, hence in tum dim V** = n. By the corollary to Theorem 4-30,
this equality of dimensions implies that V(F) and V**(F) are isomorphic.
There is one perfectly natural and useful isomorphism between these spaces,
the so-called canonical imbedding j the full story is told below.
Given a vector x e V and functional f e V*, f(x) is a scalar. Although we
have grown accustomed to thinking of this as a function of x for fixed f, let us
now invert this usual practice and allow f to range over V*, while holding x
fixed [many authors would emphasize this by writing x(f) in place of f(x)].
Specifically, define the function T",: V* --+ F by
T",(f) = f(x), fe V*.
LINEAR MAPPINGS 293
PROBLEMS
Unless indicatl'd otherwise, VW) and Jl"W) denote vector spaees over an arbitrary
field P.
1. Using nothing other than Definition 4-18 and Theorem 4-7, show that if the
mapping/E ,,(V, l\"), then/CO) .. 0 andf(-x) .. -/(x) for each x E V.
2. Determine whieh of the following functions are linear mappings of V s(R') into
itself:
a) I(al, a2, a3) = (a2, -aI, a3),
b) I(al, a2, a3) = (ai, 0, as),
+ +
c) I(al, 42, as) = (ai, al + a2, al a2 as),
d) I{al, a2, as) = (al +
2a2 - a3, -al+ a2, 2al a2). +
3. Assume that IE l.. (lr, Jr) and U(F) is a vector subspace of lr(F). Prove that
1-1(U)(F) ill a subspace of V(I"), where, as usual,
7. Prove that if f E L( V. Jr). then I is a one-to-one function if and only if the vectors
l(xl). I(X2) .. !(x ure linearly independent whenever XI. X2 x.. are
A )
linearly indl'pendent.
R. Show hy exn.mple that the concluHion of the corollary to Theorem 4-29 is false
if V (P) iH infinite-dimensional.
9. Suppose the mapping f E /,( V.1\"). with dim V > dim It". Show that there exists
a nonzero vector Xo E V for which I(xo) = o.
10. Obtain the Fundaml"ntal Homomorphism Theorem for Vector Spacl's: If I is a
linear mapping from the vector space V(F) onto the vector space W(F). then
(V /ker (f(F) ~ W(F).
) 1. Let V(F) be finite-dimensional with basis {XI. X2 . x ..}. and let VI. Y2 . y ..
be any n pll'IIl1'nt,H of V. If thtl function!: V -+ V i~ defined by taking
!(nIXI 1 ... -I a.x ..) - alYI + ... I a.y ...
prove that f is ~ linear mapping; determine when f will be an isomorphism.
12. Prove that if the mapping f E 1.( V. V) is such that ker (j) = ker (p). then
V = ker (j) E9 f(V).
13. For a fixed element a E F. define the scalar tran8formation f.: V --+ V by
f ..(x) = ax, xE V.
x =
.
~ 1.(X)Xi,
k_l
296 VECTOR SPACES
Our purpose here is to present a list of suggestioDil for collateral reading and further
study. Those works classified under General References roughly parallel the content
of this book; the specialized sources more fully develop topics mentioned in the text
and will carry the reader considerably beyond his present knowledge.
Gen_IR......nc .
1. A. A. ALBERT, Fundamental Concepti of Higher Algebra. Chicago: The University
of Chicago Preas, 1956.
2. W. BARNES, Introduction to Abstract Algebra. Boston: Heath, 1963.
3. R. BEAUMONT and R. BALL, Introduction to Modern Algebra and Matrix ThMry.
New York: Holt, Rinehart and WiDilton, 1961.
4. G. BIRltH()FF and S. MACLANE, A Survey of Modern Algebra, 3rd ed. New York:
Macmillan, 1965.
5. R. DEAN, Elements of Abstract Algebra. New York: Wiley, 1965.
6. R. DUBlscH, Introduction. to AbBtract Algebra. New York: Wiley, 1965.
7. I. N. HERSTEIN, Topic' in Algebra. New York: Blaisdell, 1964.
8. N. JACOBSON, Lectu.ret in Abstract Algebra, Vol. I, Basic Concepts. Princeton:
Van Nostrand, 1951.
9. S. LANG, Algebra. Reading, Mass.: Addison-Wesley, 1965.
10. D. J. LEWIS, Introduction to Abstract Algebra. New York: Harper and Row,
1965.
11. N. McCoy, Introduction to lIlodern Algebra. Boston: Allyn and Bacon, 1962.
12. G. MosTOw, J. SAMPSON, and J. P. MEYER, Fundamental StrUcturet of AlgebrlJ.
New York: McGraw-Hill, 1963.
13. H. PALEY and P. WEICHSEL, A First Cou.ru in Abstract Algebra. New York: Holt,
Rinehart and Winston, 1966.
14. R. E. JOHNSON, University Algebra. Englewood Cliffs, N. J.: Prentice-Hall, 1966.
15. S. WARNER, Modern .llgebra, 2 vols. Englewood Cliffs, N. J.: Prentice-Hall, 1965.
16. J. E. WHITESITT, Principles of Modern AlgebrlJ. Reading, Mass.: Addison-Wesley,
1964.
17. O. ZARISKI and P. SAMUEL, Commutative Algebra, Vol. I. Princeton: Van Nostrand,
1958.
297
298 SELEcrED REFERENCES
Group Th.ory
18. C. CURTIS and I. REINJo:R, Repre8entation Theory oJ Finite GrOUp8 and Associative
Algebras. New York: Interscience, 1962.
19. M. HALL, The Theory 0/ Groups. New York: Macmillan, 1959.
20. A. KUROSH, The Theory oj Groups, 2nd ed. New York: Chelsea, 1960.
21. W. LED.:RMANN, Introduction to the Theory 0/ Finite Groups, 5th ed. New York:
lnterscience, 1964.
22. J. ROTMAN, The Theory o/Groups: An Introduction. Boston: Allyn and Bacon, 1965.
23. K SCHENKMAN, Group Theory. Princeton: Van Nostrand, 1966.
24. W. !;COTT, Group Theory. l<:nglewood Cliffs, N. J.: Prentice-Hall, 1964.
25. H. ZAS8ENHAUfI, The Theory oJ Groups, 2nd ed. New York: Chelsea, 1958.
Ring. ond FI.ld.
26. I. ADAMSON, Introduction to Field Theory. New York: Interscience, 1964.
27. E. ARTIN, Galois Theory, 2nd ed. Notre Dame, Ind.: University of Notre Dame
Press, 1955.
28. N. McCoy, Rings and Ideals. Buffalo: Mathematical Association of America, 1948.
29. N. McCoy, The Theory 0/ Ring. New York: Macmillan, 1964.
30. D. NORTHCOTT, Ideal Theory. Cambridge, England: Cambridge University Press,
1953.
31. A. ROBINSON, Numbers and Ideal. San Francisco: Holden-Day, 19.5.
UnoarAlg.bra
32. C. CURTlli, Linear Algebra: An Introductory Approach. Boston: Allyn and Bacon,
1963.
33. P. HALMOS, Finite Dimensional Vector Spaces, 2nd ed. Princeton: Van Nostrand,
lUSH.
34. K. HOFFMAN and R. KUNZE, Linear Algebra. Englewood Clifftl, N. J.: Prentice-
Hall, 1961.
35. S. LANG, Linear .-1lgebra. Reading, Mass.: Addison-Wesley, 1966.
36. D. RAIKOV, Vector Spaces. Groningen: P. Noordhoff, Ltd., 1965.
37. F. STF.WART, Introduction to Unear Algebra. Princeton: Van Nostrand, 1963.
s.. Th.ory and Function.
38. P. HALMOS, Naive S/'t Theory. Prim'cton: Van Nostrand, 1960.
39. N. HAMILTON and J. LANDIN, Set Theory: The Structure oJ Arithmetic. Boston:
Allyn and Bacon, 196!.
40. R. STOLL, Set Theory and Logic. San Francisco: Freeman, 1963.
41. P. SUPPES, ,txiomatic Set Theory. Princeton: Van Nostrand, 1960.
Index of Special Symbol. and Notation.
In the list below, numbers refer to the page on which the symbol is first defined or used.
A ~ B, :t C B Sl't ..t is im'luoeo (properly ineluoerl) in 8t't R, 2
.1 U /I IInilln .. f 111'1>1 .1 RIIII n, :J
..t () B intersection of sets A and B, 3
A - B relative complement of set B in set A, 3
.-1 + B sum of !lets .1 and B, 161
.-1 A B symmetric difference of sets 1t and B, 36
A X B Cartesian product of sets A and B, 8
-A complement of the set A, 3
{a} singleton a (set consisting of the element a), 2
(a] equivalence class determined by the element a, 9
a E A, a ~ .1 element a belongs (does not belong) to the set A, 1
a- b element a is equivalent to element b, 9
alb, atb a divides (does not divide) b, where a, b are integers, 20
a == b (moo n) integer a ill congruent to integer b modulo n, 52
gcd (a, b) greateKt common divisor of the inttlgers a and b, 21
lcm (a, b) least common multiple of the integers a and b, 26
(a, b) ordered pair of a and b, 6
[)" n, domain ami range of the fun(ltioll I, 13
I( A) direct image of the set A under the function I, 17
I-I (A) invf'rl!f1 image of the set A under the function I, 17
r l inverse of the function I, 16
II A restriction of the function I to the set A, 17
/. g composition of the functions I and g, 15
iA ident,ity funetion on the ACt .-\, 17
"
P( A) power set of the fI(.t .-1, 5
I'mpty Ill.t, 2
..
Q set of rational number:!, 1
n', R'-+ set of real (positive real) numbers, 1
U universal set, 2
Z ACt of in teg('rs, 1
Z., Z., Z+ sets of even, odd, and positive integers, 10
Z.. . set of integers modulo n, 55
+., ',. addition (multiplication) modulo n, 56, 146
Notation. from Group Theory
[a, bJ a. b a-I. b- I , 85
a.1/ I'IISI'\. flf til(' >III hp;roll II (/[, .), 7Ii
2tJ9
300 INDEX OF SPECIAL SYMBOLS AND NOTATIONS
The following chart should help the reader to visualize the interrelations among the
various algebraic IlYfltcmll considered in the text.
Commutative _i
groups
Vector spaces
Division rings
Algebras
INDEX
INDEX