Introduction To Abstract Algebra - Dubisch
Introduction To Abstract Algebra - Dubisch
Introduction To Abstract Algebra - Dubisch
. Abstract Algebra
INTRODUCTION TO
Abstract Algebra
BY ROY DUBISCH
Intermediate Algebra
with Vernon E. Howe: and Steven J. Bryant
ROY DUBISCH
PROFESSOR OF MATHEMATICS
UNIVERSITY OF WASHINGTON
by
John Wiley 8; Sons, Inc.
The aim of this text is to provide a gradual introduction to the basic concepts
of abstract algebra. It begins very slowly with the development of a set of
postulates for the natural numbers and, throughout, alternates between dis-
cussions of familiar systems and their more abstract generalizations. Thus the
discussion of natural numbers and the integers is followed by a discussion
of integral domains; the discussion of rational numbers is followed by a dis-
cussion of groups and fields, etc.; and in the last two chapters the concept of a
vector is applied to polynomials, and standard results on the theory of equa-
tions are presented. Following the same principle of gradual development,
the text presents induction in Chapter 2 in its simplest form and then returns
to a more sophisticated treatment in Chapter 5; the simple concept of a group
is presented in Chapter 7 with further work on this topic given in Chapter 9.
In addition to a large number of problems of varying degrees of difficulty,
the text includes, at the end of each chapter, a list of problems for further
investigation. These are, in general, open-ended questions designed to stimu-
late active investigation by the student and to lead him to further reading not
only in the standard texts on the subject but also in such journals as the
American Mathematical Monthly.
For a student with a background of four years of conventional secondary
school mathematics, the material in this text should serve for approximately
two-thirds of a year’s course in abstract algebra to be followed by material
in linear algebra. Students with a background of more modern secondary
school mathematics should be able to cover the material in less time.
Within any chapter, theorems and definitions are numbered sequentially.
A reference to a theorem or definition in another chapter will include the
chapter number and the number of the theorem or definition. Thus, for
example, Theorem 7.2 refers to Theorem 2 of Chapter 7. An asterisk pre-
ceding a problem indicates a more difficult problem; a notation such as [Bl]
refers to the books listed in the bibliography; a list of symbols used appears
just before the text.
Rov Dusxscn
Seattle, Washington
February, 1965
Contents
vii
Chapter 6 (continued)
4. The Integers as a Subset of the Rational Numbers 64
S. Additive and Multiplicative Inverses in R 65
6. The Ordering of the Rational Numbers 66
Chapter 7. Groups and Fields 70
Definitions and Examples 70
PPNPwéfifiP
set membership 2
bficglnm
n
set inclusion 2
null set 1
Union for sets 3
intersection for sets 3
complement of a set 4-
for sets 4
lzww
i‘s
ix
Sets
1. Definition
Modern algebra, as we shall see, does not deal only with the properties of
numbers and the solution of equations. One of the simplest examples of a
nonarithmetical nature is elementary set algebra; we will also find it useful
to use some of the language of sets throughout this book.f
A set is simply a collection of objects, finite or infinite, either exhibited
explicitly or described in some definite fashion.
Often the so-called set-builder notation is used. Thus {xlx is a number and
x + 3 = 5} is read, “The set of all x such that x is a number and x + 3 = 5."
Clearly {xlx is a number and x + 3 = 5} = {2}. Often phrases such as “x
is a number” are left out and simply implied by the context. Thus
{x|x + 3 = S} = {x]x is a number and x + 3 = 5}. Similarly the set of all
numbers greater than 3 may be written {xlx > 3}.
A set is calledfinite if it can be put into one-to-one (1 — 1) correspondence
with the set of positive integers {1, 2, . . . , n} for some positive integer n;
that is, if we can match up each element of the given set with each element
of the set {1, 2, . . . , n}. Thus the sets in Examples 1 and 2 are finite but
the sets in Examples 3 and 4 are not finite and are called infinite.
It is convenient to introduce the convention of the null set, denoted by z ,
with no elements. (Do not confuse (a with the set {0}, which has the one
element, 0, in it. For example, {xlx + 3 = 3} = {0} but {xlx > 0 and
x + 3 = 3} = 121.)
T In fact, many of the newer texts on elementary and secondary mathematics make
extensive use of the concept of a set.
2 Introduction to Abstract Algebra
Exercises 1.2
1. Decide which of the following statements are true and which are false.
(8) 1 60.2} (b){1)E{1,2}
(C) 1 C {1, 2) (d) {1} C {1, 2}
(e) {1.2) EU. 2.1. 2. 1} (f) (1. 2} 2 {1. 2. 1. 2.1}
(g) deladall (h) {1, 2} = {1,2, 1,2,1}
(0 £3 EMJflD (j) {Z C{fl.w))
(k) {1} €{{1}.{2}} (1) {1) C {(11.9)}
(mu. 1,2} = (1.2} (n) {1. 2} 3 {1. 1.2}
(o) {xix is a number and x” — 5x + 6 = 0} = {2, 3}.
(p) {xlx is an integer and x is divisible by 3} = {yly is an integer and y is
divisible by 3).
If A and B are two sets we define their union, A U B, to be the set of ele-
ments which are either in A or in B (or in both). That is, x e A U B if and only
if x E A or x e B.
We are now in a position to prove many simple theorems in set theory such
as:
l. A n B = B n A; A U B = B U A (commutative laws).
2. A n A = A; A U A = A (idempotent laws).
3. An(Bn C) =(AnB)n C;AU (BU C) = (AUB)U C(associative
laws).
4. An(BUC)=(AnB)u(An C);AU(Bn C)=(AU B)n (AU C)
(distributive laws).
5.AnBEA;AUBQA.
Exercises 1.3
4. Ordered Pairs
The set {a, b} is equal to the set {b, a}. For an orderedpair, (a, b), however,
we wish to make the requirement that (a, b) = (c, d) if and only if a = c and
b = (1. Thus, for example, {1, 2} = {2, 1} but (1, 2) ye (2, 1).
We can do this in an elegant fashion by utilizing the concept of a set and
defining (a, b) = {{a}, {a, b}}. To show, then, that (a, b) = (c, J) implies
a = c and b = J we first prove that
(l) {x, y} = {x, 2} implies y = 2.
For, if {x, y} = {x, 2}, we have y e {x, z} so that either y = x or y = 3.
But, also,ze{x,y} so thatz = xorz =y. Thuswehavey = z or x = y = 2.
Now suppose that (a, b) = {{a}, {a, b}} = (c,d) = {{c}, {a, d}}. Then
{a} e {{c}, {5, d}} and hence {a} = {c} or {a} = {c, d}. In either case 0 e {a} and
hence c = a.
With 1: = a we have {{a}, {a, b}} = {{a}, {a, J}}. By (1) it follows that
{a, b} = {a, d} and, again by (1), that b = J. On the other hand, if a = c and
b = J it is obvious that (a, b) = (c, J).
This shows that if we define (a, b) = {{a}, {a, b}}, then (a, b) = (c, d) if
and only ifa = c and b = d.
sets 5
5. Binary Operations
If we have given a set S such that every ordered pair, (at, y), of elements
x, y e S determines uniquely another element, x a y in S, we say that we have
defined a binary operation on S. For example, if S is the set of positive integers
we might define as t y as x + y and observe that, since the sum of two positive
integers is a positive integer, x I y e S. On the other hand, if S is the set of
odd positive integers and x, y e S, x + y is even so that x + y ¢ S. Hence
in this case we have not defined a binary operation on S. But if we define
x a y as xy we have a binary operation on the set of odd positive integers
since the product of two odd integers is again an odd integer.
Exercises 1.5
1. The Problem
2. Closure
We are given a set S with two binary operations defined on it. Now since
we know that we are going to end up with the natural numbers and the ordi-
nary operations of addition and multiplication, it is natural to use the ordinary
symbols for these operations. To do so, however, at this stage of the game is
to prejudge the question. So let us hedge by using 9 and ® to desig-
nate the two operations with the understanding that, at the moment, a ® b
might turn out to mean, for example, a + b for a and b nonzero rational
numbers or that a e b might mean a U b for a and b sets. Since we are
requiring that 69 and ® be binary operations on S we may list as our first
postulate the requirement of closure with respect to these operations.
Pastulate 1.1faeS and hes, then aebeSand a®beS.
6
THE NATURAL NUMBERS 7
Exercises 2.2
Determine whether or not the following sets with the given operations satisfy
Postulate l.
l. S = rational numbers; a (ab = ab, a ® b = a + b.
S = rational numbers; a (-Bb =a — b, a ®b =b —a.
99““9999P
S = setofsubsetsofagivensetA;a ®b = an b,a ®b = aU b.
S =setofodd integers; a @b =a +b, a ®b =ab.
S =setofevenintegers;a (-Bb =11 +b,a ®b =ab.
S = setofrationalnumbers; a @b = a — b, a 8 b = a +b.
S = set of nonzero rational numbers; a G) b = a — b, a ® b = a + b.
o
By this we mean that the table is used as an ordinary addition table would
be used: a 6 a = a, a 6 b = c and so on. Finally, let x ®y = a for all
x, y e S. It is easy to see that Postulates 1 and 2 are satisfied by this system,
but
(aQb)®c=c®r=a
while
a®(b®c)=a®c=b
Note, however, that the rational numbers under ordinary addition and
multiplication still satisfy all of our postulates, so that these three postulates
still do not uniquely characterize the natural numbers.
Before going on to the next postulate we remark that it is possible to give
examples of systems in which Postulate 1 holds and the first half of Postu-
late 2 (a 6 b = b $ a) but not the second ((1 ® b = b (8 a). Similarly,
we can give examples of systems in which Postulates 1 and 2 and the first
half of Postulate 3 hold, but not the second. Or, we can give examples of
systems where Postulates l and 3 hold, but not 2, and so on. A great multi-
plicity of examples, however, may be confusing at this stage and if the student
will simply have faith that such examples exist, nothing will have been lost
by omitting discussion of them at this time. (See also the exercises below.)
Exercises 2.3
1. For the following sets and operations determine whether or not the opera-
tions are commutative and/or associative (both for (9 and for ®).
(a) S = rational numbers; a e b = (lb/2, a 8) b = (a + b)/2.
(b) S = rational numbers; a (43 b = b, a ® b = a.
(c) S =set ofsubsets ofa given set A; a (-Bb =an b, a ®b =aU b.
(d) S = set ofintegers; a @b = a — b,a ‘8 b = ab.
(e) S = setofintegers;a 69b = 2a + b, a ® b = 2ab.
2®(3®4)=2+(3x4)=2+12=14
but
(2-®3)$(2®4)=(2+3)x(2+4)=5x6=30.
We note, however, that the rational numbers under ordinary addition and
multiplication also satisfy Postulates 1 to 4 so these four postulates still do
not uniquely characterize the natural numbers.
Exercises 2.4
1. Determine whether or not Postulate 4 is satisfied for the sets and operations
given in problem 1 of Exercises 2.3.
2. Find a set S closed under the operations (49 and ® such that
(a) Postulates 2 and 4 are satisfied but one or both parts of Postulate 3
do not hold.
(b) Postulates 3 and 4 are satisfied but one or both parts of Postulate 2
do not hold.
On the other hand, the positive rational numbers under ordinary addition
and multiplication still satisfy Postulate 6, as well as Postulates 1 to 5.
We have seen that the set of positive rational numbers under ordinary
addition and multiplication has been difficult to eliminate as a “candidate"
for the title of a set of natural numbers in that it continues to satisfy all of
the various postulates that we have laid down. We now outlaw it and almost
all others by requiring that the principle of finite (or mathematical) induction
hold.
{1,1%,2,2§,3,3é,-..,n+;,---}
1$l=l®1=1
To check Postulates 1 to 6 for this system is then very trivial since in every
such postulate we must take a = b = c = l, and every possible computation
yields the answer of 1. Furthermore the only nonempty subset of S is S
itself so that Postulate 7 is trivially satisfied.
Exercises 2.5
6. Isomorphism
3 = {1,2, 3,4, ...} and s' = {1, II, III, IV, ...}
Clearly, however, S and S’ differ only in the symbols used in designating the
elements of the set. More specifically, by the usual rules in regard to Roman
numerals versus Arabic numerals, there is a one-to-one (1—1) correspondence
between the symbols used to describe S and the symbols used to describe S'.
Thus, for example, 1H I, 24—» II, 194—) XIX. (It is true that the Roman
numeral scheme tends to break down for larger numbers, but we can easily
7 Up to an “isomorphism.” This term will be defined precisely in the next section.
3 See, for example, [B4].
§Again, up to an isomorphism.
12 Introduction to Abstract Algebra
21—)II 3<—>III
Then
2+3=5 II+III=V
2x3=6 IIxIII=VI
and 5 4—) V, 6 H VI.
Let us contrast this with the situation where our two sets are: S =
{1, 2, 3, 4, . . . , n, . . .} and T = {2,4, 6, 8, . . . , 2n, . . .). It is still possible
to set up a 1—1 correspondence between the elements of the two sets; for
example, 1 (—> 2, 2 H 4, 3 <—> 6 etc. Then n 4—» 2n, rm—v 2m and, if the corres-
pondence is preserved under addition, we must have both 1: + m<—->
2 (n + m) and n + rn<—>2n + 2m. Since 2n + 2»: = 2(n + m) we conclude
that the correspondence is indeed preserved under addition. But, for example
2H4, 3H6
with
2x3=6, 4x6=24
Definition 1. Let S and S' be two sets. On S are defined the binary
operations +, x , . . . while on S’ are defined the binary
operation: +’, x’, . Then we say that S and S’
are isomorphic (with respect to the operations defined
on them) if
(1) There is a 1—1 correspondence between the elements of
S and S’.
THE NATURAL NUMBERS l3
Since we must have a 1—1 correspondence it is clear that two finite sets
with a different number of elements cannot possibly be isomorphic.
We will consider the topic of isomorphisms again in greater detail in Section
7.6.
Exercises 2.6
Let us return now to the principle of finite induction as a tool for the proof
of propositions involving natural numbers. For example, consider the
statement:
If n is a natural number, then
n(n + l)
1+2+...+n
2
To prove this we let S be the set of natural numbers, n, for which
n(n + 1)
1+2+...+n=
2
14 Introduction to Abstract Algebra
That is,
n(n + 1)
S = {n In a natural number and 1 + 2 + +n = T}
[:02 + l)
(1) 1+2+...+k _ 2 ,
Exercises 2.7
();+;+;+ ++—M_
3 1-2-3 2-3-4 3.45 n(n+l)(n+2)_4(n+1)(n+2)
(h)1 +2 +2: +... +2"-1=2"—1
(i) 3 +3‘ +35 + +3""1 =i(9" —1)
At the beginning of this chapter we commented that there is more than one
way to characterize the natural numbers. A very neat way is to replace our
eight postulates by the following (due to G. Peano, [1858—1932]).
Our undefined terms are natural number, successor, and 1.
Exercises 2.8
1. Prove that Peano’s postulates for the natural numbers are implied by the set
of eight postulates previously given.
'2. Formulate a suitable definition of the sum, in + n, of two natural numbers
m and n in terms of the successor.
*3. Formulate a suitable definition of the product, run, of two natural numbers
m and n in terms of the successor.
1. Equivalence
Let us go back to our reason behind the introduction of ordered pairs
of natural numbers and suppose that we are working with equations whose
solutions are natural numbers. In particular, suppose that the equation
x + b = a has a solution in natural numbers that is the same asthe solution
in natural numbers of the equation x + d = c. More precisely, suppose that
a, b, c, and d are natural numbers, A = {xlx E N and x + b = a} = B =
{xlx EN and x + d = c}, and A aé 21. We claim that A = B if and only
if a + d = c + b.
17
18 Introduction to Abstract Algebra
s + (d + b) = c + b
Continuing, we have
a+d=c+b sinces+b=a
a = b for two natural numbers a and b we mean that a and b are identical
natural numbers. Thus, for example 1 + 2 = 3 means that “1 + 2” and
“3” name the same number. When, however, we define (a, b) ~ (1:, d) if
and only if a + d = c + b we are defining “~” in terms 0 “=” and we
have no guarantee that “~ ” has any of the properties of “=”. Thus, for
example, if we were to define (a, b) ~ (c, d) if and only if a + b = c we would
not even have (2, 1) ~ (2, 1) since 2 + l yé 2.
We note that if (a, b) = (c, :1) then (a, b) ~ (c, (I). For if (a, b) = (c, d)
thena=candb=dandhencea+d=c+bbecomesa+d=a+d.
Furthermore, we may prove that “ ~ ” does indeed possess some of the most
significant properties 0 “=”.
Theorem 1.
(1) (a, b) ~ (a, b) (~ is reflexive);
(2) If (a, b) ~ (c, d), then (c, (1') ~ (11, b) (~ is rymetrtl‘);
(3) If (a, b) ~ (c, d) and (c, d) ~ (2, f), then (a, b) ~ (2, f) (~ is transitive).
Proof.
(1) (a, b) ~ (a, b) since a + b = a + b.
(2) If (a, b) ~ (c, 4), then (c, d) ~ (a, b) since if a + d = c + b we have
c + b = a + d.
(3) If (a,b) ~ (c,d) and (0,4) ~ (e,f), then a + d = c + b and l.‘ +f
= e + d. Hence
(a+d)+e=(c+b)+e
a + (d + e) = (c + b) + e by the associative law of addition in N
a + (e + d) = (c + b) + e by the commutative law of addition in N
a+(£+f) =(c+b)+e sincee+d=c+f
a + (f + c) = (c + b) + e by the commutative law of addition in N
(a + f) + c = c + (b + e) by the associative law of addition in N
(a + f) + t‘ = (e + b) + c by the commutative law of addition in N
a +f= e + b by the cancellation law of addition in N
But ifa +f = e + b, then (a, b) ~ (e,f) as desired.
We have written this proof out in great detail with reasons for each step.
From now on we will condense steps and frequently omit reasons. Until the
student is thoroughly familiar with this type of proof he should supply the
omitted steps and give a reason for each step.
Exercises 3.1
1. Let S be the set of all triangles in the plane. If a, b E S we write a g b
if and only if a and b are congruent triangles. Is “ g" reflexiveP; symme-
tric?; transitive?
2. Let S be the set of all integers. If a, b e S we write a ~ b, if and only if
a — b is divisible by 7. Is “~” reflexive?; symmetric? ; transitive?
20 Introduction to Abstract Algebra
(x+y)+(b+d)=a+c
(The student should supply the details.) Thus, since (a, b) is the symbolic
solution of x + b = a and (c, d) the symbolic solution of y + d = c we
see that (a + c, b + d) is the symbolic solution of z + (b + d) = a + c
where a = x + y. Hence it is natural to make the following definition.
(3) a+d=e+bande+v=u+f.
Now (a, b) (8 (e,f) = (ae + bf, af + be) and (c, d) ® (u, v) = (cu + do,
w + 1114). Thus (2) becomes (are + bf, af + be) ~ (cu + do, w + du) so
that, by definition of ~ , (2) is equivalent to
That is, to establish (2) we need to derive (4) from (3). We proceed.
From (3) we obtain, by the distributive law in N, ae + de = cc + be and
C: + a! = cu + cf, which then yield, by addition,
22 Introduction to Abstract Algebra
(6) (ae+de+w)+ce=(be+cf+cu)+c¢.
(7) ae+de+w=be+cf+cu
(8) cf+bf=af+dfanddu+df=de+dv.
(9) (ae+de+w)+(6f+bf)+(du+df)
=(be+tf+cu)+(af+df)+(de+dv)
We apply the associative and commutative laws for addition in N once again
to obtain, from (9),
[(ae+bf)+(w+du)]+(de+tf+df)=
[(cu+dv)+(af+be)]+(de+€f+df)
Exercises 3.2
We thus see that the commutative laws of addition and multiplication and
the associative law of addition for these pairs of natural numbers may be
proved on the basis of our definitions and the postulates for the natural
numbers.
The proofs for the associative law of multiplication and the distributive law
for multiplication with respect to addition are entirely similar. But since
they reflect no new principles and involve somewhat tedious computations
24 Introduction to Abstract Algebra
Theorem 6. (1) If (a, b) @ (c, d) ~ (e,f) G) (c, d), then (a, b) ~ (e,f).
(2) If (a, b) (8 (c, d) ~ (e,f) (X) (c, d) and c 9E d, then (a, b) ~ (e,f).
Proof of(1) Since (a, b) e (c, d) = (a + c, b + d) and (e,f) e (c, d)
= (e + c, f + d) we see that our hypothesis implies
(a+c,b+d)~(e+c,f+d)
But then, by Definition 1,
(a+c)+(f+d)=(e+c)+(b+d)
We now use the associative and commutative properties for the natural
numbers to obtain
(a+f)+(c+d)=(e+b)+(c+d)
and hence
But then
(a+f)c+d(b+e)=(b+e)c+d(a+f)
EQUIVALENT PAIRS OF NATURAL NUMBERS 25
(a+f)(d+x)+d(b+e)=(b+e)(d+x)+d(a+f)
(ad+ax+fd+fx)+(db+de)=(bd+bx+ed+ex)+(da+df)
or
(ax+fx)+(ad+fd+db+de)=(ex+bx)+(ad+fd+db + (1:)
ax+fx=ex+bx
or
(a +f)x = (e + b)x
Now, finally, we apply the cancellation law of multiplication for natural
numbers to obtain a + f = e + b and, hence, (a, 11) ~ (e,f).
The case when there exists a natural number y such that c + y = :1
presents no new difliculties and we leave this case to the student. (Note that
the case of t = d corresponds to x -0 = y -0 even when x aé y.)
Using only single ordered pairs we have come about as far as we can in
our efforts to develop the integers from the natural numbers. Thus if we
examine our set of pairs of natural numbers for an analogy to Postulate 6 we
find that (2,1), (3, 2), (4, 3), and, in general, (11 + 1, n) all serve as unit
elements for multiplication in the sense that, for example, (2, 1) ® (a, b) =
(241 + 17, 2b + a) ~ (a, b).
We turn now, in the next chapter, to a consideration of equivalence classes
and will construct the integers as classes of equivalent pairs of natural
numbers.
Exercises 3.3
3. Prove the distributive law for multiplication with respect to addition for
these pairs of natural numbers.
*4. Prove that if (x, y) ® (a, b) N (a, b) for all pairs (a, b), then (x, y) ~
(2, 1).
l. Suppose that S is again the set of all ordered pairs of natural numbers,
(a, b), where now (a, 17) ~ (6, d) if and only if a = c and b = d, (a, b) e
(c, d) = (a + c, b + d), and (a, b) ® (c,d) = (ac, bd). Provide an ana-
logous treatment for S and its operations to the treatment just given.
Suppose that Mr. X dislikes working with “signed” numbers. He replaces
5"
1. Relations
the natural numbers is not an equivalence relation for, while we do have the
transitive property (a < b and b < c implies a < c), we do not have the
reflexive (a < a) or symmetric (a < b implies b < a) properties.
Exercises 4.1
1. Determine, for the following sets and relations, whether or not the relation
is (i) reflexive, (ii) symmetric, and (iii) transitive.
(a) S = N; a R b if and only if a divides b.
(b) S = integers; a R b if and only if a — b is divisible by 3.
(c) S = integers; a Rb if and only if a S lbl.
(d) S = integers; a R b if and only if a2 = b”.
(e) S = lines in a plane; a R b if and only if a is perpendicular to b.
(f) S = lines in a plane; a R b if and only if a is parallel to b.
(g) S = set of all people; a R b if and only if a is a sibling of b.
(h) S = set of all people; a R b if and only if a is a spouse of b.
3 Let S = ((a, b)|a, b EN} and define (a, b) ~ (6, d) if and only if ad = bc.
Prove that ~ is an equivalence relation on S.
6. What is wrong with the following argument showing that a relation which is
both symmetric and transitive is necessarily reflexive? a R b implies b R a
by symmetry; by transitivity, a R b and b R (1 implies a R a.
(2) Suppose that we have two partitioning classes Pa and Pb such that
that P,, n P, 9e {3 . We wish to show that Pa = P,J (condition (2) of Defini-
tion 3). Since Pa n P, aé 125 there exists an x e S such that x 6 Pa and x e Pb.
Now consider any element y 6 Pa. Then y R (1. Now as R a and x R b be-
cause at 6 Pa and x e Pb. But x R a implies a R a: (symmetric property)
and yRa and aRx implies n (transitive property). But also, xRb,
and again applying the transitive property (now to y R x and x R b) we obtain
y R b, so that y e Pb. Thus y E Pa implies y e P, and P,z E Pb. Similarly,
we obtain P, 9 Pa and hence Pa = P, as desired.
II. Now let P1, P2, . . . be a partition of S and define for x,y E S, x Ry if
and only if x and y are in the same partitioning class P,. It is easy to verify
that R is then an equivalence relation on S and this is left as an exercise for the
30 Introduction to Abstract Algebra
Example 3. Let S = N and let a R b if and only if a and b are both odd
or both even. ThusZ R 6 and 3 R 5 but 2 R 3 and 3 R 6. Since all natural
numbers are either even or odd we have two partitioning classes,
P1={x[xeNand1}andP2={xlxeNand2}
Example 5. Let S be the set of all triangles in a plane and let R be the
relation of similarity. Then each partitioning class of S is a set of similar
triangles. There are an infinite number of partitioning classes and each
partitioning class contains an infinite number of elements.
Exercises 4.2
1. Describe the partitioning classes of the set of integers under the equivalence
relation a R b if and only if a — b is divisible by S.
2. Consider the set S and the equivalence relation on it as given in problem 3
of Exercises 4.1. Describe the partitioning classes under this equivalence
relation.
3. Consider a set S and a relation R on it. Define, as before, P. = {xlx e S
and x Ra} for any a e S. Show, by examples, that the set of all such P.
is not a partition of S if (a) R is reflexive and transitive but not symmetric;
(b) R is symmetric and transitive but not reflexive; (c) R is reflexive and
symmetric but not transitive.
4. Give the details of the verification that, in the proof of the second part of
Theorem 1, R is an equivalence relation on S.
EQUIVALENCE CLASSES AND 'n-n: 1mm 31
—2={(1,3),(2,4), ,(a,a+2),...}
—1 ={(1,2),(2,3), ,(a,a+ 1),...}
0={(1,1),(2,2),...,(a,u),...}
+1 ={(2,1),(3,2),...,(a+1,a),...}
+2={(3,1),(4,2), ,(a+2,a),...}
In general, for k a natural number, we define
Proof; If I (a, b) = I (c, 11) then, since (a, b) e I (a, b), we have (a, b)e
I(c, (1). Hence (a, b) ~ (1:, d) and thus a + d = c + b.
On the other hand, if a + d = c + b, then (a, b) ~ (c, d). If (x, y) e
1(a, b), (x, y) ~ (11, b). By the transitive property of ~ we have (x, y) ~
(1:, d) and hence (x, y) eI(c, d). Thus I (a, b) S I (c, d ). Similarly, we
obtain [(c, d) E 1(a, b) and hence [(a, b) = 1(c, d).
32 Introduction to Abstract Algebra
Note that we are using the sign of equality in the proper sense of identity;
the two sets I (a, b) and I (c, d) are identical if a + d = c + b.
Exercises 4.3
1. Perform the indicated operations and express your answer in the form
[(1, 1), [(12, 1), or [(1, k).
(a)I<2,3>+I<4,7> (b)1<3,1>+1<5.8>
(C)1<1.1> +I<5.2> (d) I<6,2> +I<8,3>
(e) [(2,3) x [(4,7) (f) [(3,1) x [(5,8)
(910.1) >< [(5,2) (h) 1(6.2> X [(8,3)
2. Verify Theorem 3 in the following cases.
(a) 19.3) =I<5,6>,I(4,7> =1<1.4)
0910.1) =I<514>11<5,8> =I(2.5>
(01(1)!) =I<3,3),I<5,2> =I(7,4>
(d) [(6,2) =[<8,4),[<8,3> =[<6,1>
laws and that the cancellation laws hold as modified in Theorem 3.6. That is,
ifI(a, b) + [(0, d) = I<e,f> + I(c, d), then I<a, b) = I<e,f> and if
I<a,b> x I<c,d) = I(e,f) x I<c, d) and: ye d,thenI(a,b> = I<e,f).
We now consider some additional properties of the integers.
Theorem 4. There is a unique multiplicative identity, +1, for the
integers.
Proof: We recall that +1 = I(2,1>.ThenI(2,1) x I(a, b) = I<2a+ b,
a + 2b) = [(a, b). Furthermore, if I<r, s) x I(a, b) = I<a, b) for all
natural numbers a and b, we have I(r, s) x I(a, b) = I(2, l) x I(a, b)
for all natural numbers a and b. Taking (1 ye b we may use the cancellation
law for multiplication to obtain I (r, s) = I (2, 1 >.
The student should compare the situation here very carefully with the
situation in Chapter 3. There we pointed out that, for example, both (2, 1)
and (3,2) were multiplicative identities and, while (2,1) ~ (3, 2), (2, 1) ye
(3, 2). Now we have I(2,1> = [(3, 2) = {(2,1), (3,2), (4, 3), . . .,
(k + 1, lg), . . . }. Thus Postulate 6 for the natural numbers continues to hold
for the integers.
Our next theorem is a modification of Postulate 8 for the natural numbers.
In the proof of this theorem and elsewhere the following lemma is useful.
Lemma. I(x + 5,9: + d) = I<c,d>.
Proof: By Theorem 2, I<x + c, x + d) = I(c, d) if and only if
(x+c)+d=c+(x+d). But
(at + c) + d = x + (e + d) Associative law of addition in N
= x + (d + c) Commutative law of addition in N
= (x + d) + c Associative law of addition in N
= c + (x + d) Commutative law of addition in N
Theorem 5. If I (a, b > and I (c, d ) are two integers, there exists a unique
integer, I(x, y) such that [(a, b) + I<x, y) = I(c, d). We have 1(x, y)
=10! + c,a +41).
Prov/2
I(a,b> +I<b+c,a+d) =I<a+(b+c),b+(a+d)>
Definition of addition
=I<(a+b)+c,(b+a)+d>
Associative law of addition in N
=I<(a+b) +c,(a+b) +11)
Commutative law of addition in N
= I (c, :1) Lemma
34 Introduction to Abstract Algebra
Exercises 4.4
Exercises 4.5
From these definitions, the usual rules of operation with integers follow
quickly. We state and prove two such properties here and list others as
exercises for the student.
Theorem 9. —(I<a, b) - 1(c, 11)) = —I(a, b) + 10:, d).
Exercises 4.6
1. In the set of all people, the relation “is an uncle of” is clearly connected
with the relations “is a brother of” and “is a parent of.” Can you state
any similar general rule for constructing a new relation out of two given
ones? [T1]
2. If R and T are any two relations on a set S can you gize reasonable inter-
pretations to R S S, R = S, R U S, and R n S, P R (See problem 6,
Exercises 1.3.) How far can you go in constructing an algebra of relations?
[T1]
Exercises 5.1
In problems 1—10 determine whether or not the given sets are integral
domains with respect to the usual definitions of addition and multiplication.
1. (b Vin; e R}
2. {3mlm e I}
3. {a + bx/Z—laJeI)
4. (a + bxfz‘la,b em
5. {a + bila,b ER)
6. {a + Ina/61a, b 512}
40 Introduction to Abstract Algebra
7. The set of all polynomials f(x) with integral coefficients such that
f( -x) = f(x)
3. {a + Ina/5 + mama, 12,: 6R}
9. The set of even integers
10. {1}
11. Construct addition and multiplication tables for J3, 1., and Is
’12. Show that the word "unique” may be omitted in Postulates 3, 4-, and 6
in Definition 1 ; that is, prove the uniqueness from the assumption of the
existence of at least one such element as described.
Exercises 5.2
1. Determine which of the sets in problems 1—10 of Exercises 5.1 are rings.
2. Prove that J... is a ring for every m e N.
3. Consider the set S of pairs of rational numbers (a, b) with equality, addition,
and multiplication defined as follows: (a, b) = (c, d) if and only if a = cand
b = (1; (a,b) + (c,d) = (a + c,b + :1); (a,b) x (c,d) = (ac,bd). Prove
that S is a ring but not an integral domain.
42 Introduction to Abstract Algebra
N(at) = a2 — 5b2 61
If, now, [3 = e + d x/g e D, direct computation shows that N(aq3) =
N(a)N(fl). Now if at = a + b \/§ is a unit, then there exists ,3 = c + d 1/5
e D such that at! = 1 and hence N(afi) = N(1) = 1. Thus we must have
Exercises 5.3
1. What are the units of 15?
2. Prove that the only units of {a + bila, b e I) are 1, —1,i, and —i.
T A formal proof of this will be given in Section 7.
INTEGRAL DOMAINS 45
and bla.
. In Example 8, prove that N(¢fl) = N(a)N(B).
“60!
Thus we see that conditions (1) and (2) of Definition 11 are satisfied. Finally,
we observe that either
a — bw/i = 0(whencea = b= 0),a — bx/Z—EP,
or —(a—b\/2_) = —a+b\/§eP
and that only one of these alternatives may hold by our assumption that P
is a set of positive elements in R“ D D. Hence
a+b‘\/§=0,a+b\/§.EP2,or—a—b\/§= —(a+bx/§)eP,
by our definition of a + b V2 6 P2. Thus condition (3) of Definition 11 is
satisfied.
Definition 12. Let D be an ordered integral domain and P a set ofpositive
elements of D. Then by
(1) a > OwemeanaeP
(2) a > bwemeana—b > 0
(3) a < bwemeanb > a.
where a, b e D. '
(NOTE: Technically, we should write a > PO etc., since the meaning of “> ”
depends on P. Thus, in Example 5, 1 + V: >150 but 1 + Vi :1» “0
since 1 - fl ¢ P. We shall understand, however, that we are always re-
ferring to a fixed P whenever we use the symbol “ > " in connection with a
given integral domain.)
Theorem 7. Let D be an ordered integral domain. Then, for all a, b, c e D
(1) If a > b and b > c, then a > c (transitive property)
(2) Ifa>b,thena+c>b+c
(3) For any a and b, one and only one of the relations a > b, a = b, or
a < b hold
(4) Ifa > band: > 0, then ac > be.
Proofof(1). Ifa > b and b > cwe have a —beP and b—ceP by
Definition 12. Hence (a — b) + (b — c) = a — c e P by Definition 11.
Thus a > c by Definition 12.
The proofs of the other properties are left as exercises.
Definition”. azbtfandonlyifa>bora=bm sbifandonlyg'f
a <bora=b.
Theorem 8. Let D be an ordered integral domain. Then, for all a, b,
ceD
(1)1fa abandb 2 c,thena 2c
(2) Ifo 2bthena+c 2b+c
(3) Ifa Zbandc 20,thenac 2b:
(4) Ifa Zbandc Zd,thena+c 2b+d.
The proof of Theorem 8 is left as an exercise for the student.
48 Introduction to Abstract Algebra
Exercises 5.4
D is a positive element of D.
. Complete the proof of Theorem 7.
#HN
. Prove Theorem 8.
. Let D be an ordered integral domain with a, b, c e D. Prove that
(a) Ifa Zbandc <0,thenae _<_bc
(b)1fa 2b 2. Oandc ZdZ 0,thenac Zbd.
5. Prove (1) of Theorem 9.
INTEGRAL DOMAINS 49
Now that we have available the integers and concepts of inequality and
divisibility for the integers we can formulate useful variants of the principle
of mathematical induction. Our most important variant is the well-ordering
principle which we will establish by first proving three lemmas.
Lemma 1. n 2 1 for allneN.
(1) k3+1>k2+k.
0n the other hand, k + 1 e S if and only if
(2) (k+1)3+1 >(k+1)2+(k+1).
To establish (2) we add 3]:2 + 3k + 1 to both sides of (l) and obtain
(k3+3k’+3k+1)+1=(k+1)3+1>4k2+4k+1=
=(k+1)2+(k+1)+(3k2+k—1)
Now 3k2+k=k(3k+1)>1 since k>1 and 3k+l> 1. Thus
3k2+k— 1 >0and(k+ 1)2+(k+1)+(3k2+k— 1) >(k+1)2
+ (k + 1). Hence (I: +1)3 + 1 > (k + 1)2 + (k + 1) which is (2).
Thus I: + 1 e S and, by Theorem 12,S contains the setof all natural numbers
2 2. Since1¢S, S = {nlneNandn 22}.
Example 3. Prove that the maximum number of lines determined by'n
points in a plane is n(n — l)/2.
Let S = {nln is a natural number and the maximum number of lines
determined by n points is n(n —- 1)/2}. Now 1 e S since, if n = 1, n(n — l)/2
= 0 and one point does not determine a line. (Since this is a rather exceptional
case we had better see if 2 e S. Now 2 e S since 2 points do determine exactly
one line and, ifn = 2, 2(2 — 1)/2 = 1.)
Now suppose I: e S. That is, we suppose that k points determine at most
120: — 1)/2 lines. Now suppose that we have a (k + 1)st point. We can draw
at most k new lines joining the new point with the previous 12 points. (Some
of these lines may coincide with previous lines as shown below for k = 2.)
o—o——O
I 2 3
52 Introduction to Abstract Algebra
Exercises 5.5
The simple fact that if we have two natural numbers a and b with a > b
we can divide a by b and obtain a non-negative remainder less than I: is a
matter of considerable significance. We generalize this fact slightly in our
next theorem.
INTEGRAL DOMAINS 53
Theorem 13. The Division Algorithm. Given two integers a and b with
b > 0, there exists a unique pair of integers q and r such that a = bq + r
where 0 S r < b.
Let us first note that the algorithm can be made plausible by considering
the multiples, bq, of b displayed on a line as shown below:
I I l I I I I
"‘ I l I l I I I "'
b(-3) b(-2) b(-l) b°0 b'l b'2 b'3
Clearly, the point representing a must fall in some one of the invervals
determined by these points corresponding to the multiples bq. Suppose that
it falls in the interval between bq and b(q + 1) (exclusive of the right hand
point). Then a — bq = r, where r represents a length shorter than the whole
length of the interval and, hence, 0 S r < b as desired.
Now while the above argument makes Theorem 13 plausible, it is not a
legitimate proof of the theorem since it rests on some unproved statements
concerning the correspondence between numbers and points on a line. It
does, however, motivate the proof which follows.
Consider the set S = {a — bx|x eI and a — bx 2 0}. We note first that,
since b > 0, it follows that b 2 1 (Lemma 1) and hence -— |a|b S —|a| S a.
Now consider y = a — (—[a|)b. Then, since —[a|b S a, it follows that
y 2 0. Hence y e S and S aé Q . Thus either 0 e S, in which case there
existsaqsuchthata - bq = 0,a = bq + 0; orelse0¢SbutScontains
a smallest positive integer, r = a — bq (Theorem 10). Now 1 > 0 and if
rzb, then OSr—b=a—bq—b=a-(q+l)b<rcontrarytothe
hypothesis that 0¢S and r is the smallest positive integer in S. Hence
7 < b as desired.
To prove the uniqueness of q and r we assume a second pair of integers,
q’ and r’, such that a = bq’ + r’ where 0 S r’ < b. But then bq’ + r' = bq
+ r and r’ — r = (q — q')b so that b[(r’ — r). But |r' — r[ < b and hence
r’—r=0and,sinceb9é0,q’—q=0.
The integer r is called the remainder and the integer q the quotient.
The division algorithm, so basic in our further development, is capable
of considerable generalization. It cannot, however, be generalized to an
arbitrary integral domain even if the integral domain is ordered. Thus, for
example, if we consider the integral domain R and the elements f and i- in
R we can write
i=§-§+0whereOS0<§
i=§-%+}where0Si<§
etc. so that there is not a unique quotient or remainder satisfying the condi-
tions of Theorem 13.
54 Introduction to Abstract Algebra
'1=’2‘13+'3 0<r3<r2
Since the 7, form a decreasing set of non-negative integers, there must exist
a natural number n such that rn+1 = 0. Thus our sequence of equalities
ends with
’n—1 = 7n9n+1
But then (a, b) = (b, r) = (r, 71) = (71, 72) = . . . = (7nd, r”) = in.
Theorem 15. There exist integers m and n such that (a, b) = ma + ub.
Proof. From the above series of equalities in the proof of Theorem 14
we have
Exercises 5.6
1. Find (595, 252) and express it in the form 595m + 2521: in two ways.
2. Do as in problem 1 for (294, 273).
3. Do as in problem 1 for (163, 34-).
4. Do as in problem 1 for (6432, 132).
5. Find (3456, 7234).
6. Prove that if m > 0, (ma, mb) = m (a, b).
7. Prove that ((a, b),c) = (a,(b, 0)) = ((a, c),b).
8. Prove that (a, m) = (b, m) = 1 if and only if (ab, 1») = 1.
9. Prove that if there exist integral solutions, x, y, of the equation ax + by = l
where a and b are integers, then (a, b) = 1.
10. Prove that if the equation ax + by = «1,4 and b integers, (a, b) = d,
hasasolutionx =mandy =n,thenx =m +bt,y=n +atisalso
a solution for every integer t.
*11. Let k be a binary operation on Nsuch that (a) a a b = b a a, (b) a . a = a,
and (c)a w(a + b) = a «bforalla, b EN. (1)Provethata :b = (a,b);
(2) show that if any one of the conditions (a), (b), or (c) are omitted that
a 1: b need not be equal to (a, b).
7. Unique Factorization in I
and have proved that the exponents <11, «2, . . ., or, as well as the primes
131, p2, . . . , p, are uniquely detemiined.
Let us return now to the topic of residue classes to prove the following
theorem.
Theorem 18. I", is an integral domain if and only if m is a prime.
Praofl It is obvious that the condition that m be a prime is necessary since
if»: = m1m2,l < m1 < m, 1 < m2 < m, then lCma = Co as, for example,
in J6 we have C203 = C0. Furthermore, if m = 1, we have only one element,
Co, contrary to our requirement that an integral domain have at least two
elements.
Whether or not fit is a prime, properties 1, 2, and 5 of Definition 1 follow
easily from Definition 3 and the properties of integers. For property 3 we
observe that C, + Co = C‘; for property 4 that C, + C,,,_I = Co and
58 Introduction to Abstract Algebra
Co + C0 = Col; and for property 6 that CIC, = 0,01 = C}. Now for
property 7 we observe that if C,C,- = C0, then 0 E ij (mod m). Hence mlij
so that if m is a prime, mli or mlj by Theorem 16. But i,j < m and hence
i = 0 or 1' = 0.
Exercises 5.7
l. Prove that if d = (a, b), a = ad, and b = bld, then (a1, bl) = l.
2. Prove that if (a, b) = 1, ale, and bit, then able.
3. Prove that if (a, c) = 1 and club, then db.
4. Prove the corollary to Theorem 16.
5. Prove that if (c, d) = 1, then (c", d) = 1 for all 1: EN.
6. Prove that if (a, b) = 1, then (a + b, a — b) = l or 2.
7. If (a, c) E d, prove that alb and cld if and only if aclbd.
‘8. Prove that the number of primes is infinite. (Hint: Assume that the number
of primes is finite and form their product p11); . . . p... Now consider plpg
p, + 1.)
l. Equivalence
We let
59
60 Introduction to Abstract Algebra
Exercises 6.2
1. Perform the following computations and write your answers in the form
R( a, b ) with a and b relatively prime.
(a)R(3,4) +R(—1,2) (b)R(5,12) +R(3,—8)
(c)R(0.1> +R(—7.9> (d)R(—3,4> ><R(2.—5)
(e)R(6.5) xR<—3,—4) (f)R(1.1) xR(-7.9)
2. Verify Theorem 2 for the following cases:
(a)R(2.3) =R(4.6>,R( -5,2 > =R(15, -6>
(b)R(-1. -1) =R<3,3).R(1.7) =R(2.14)
(C)R(0,4) =R( 0. -1).R( -3.4> =R( -6,8)
3. Complete the proof of Theorem 2.
It is convenient to define
(a, b) ® (6, d) = (ad + bc, bd) and (a, b) (8 (t, d) = (at, bd)
and to observe that
R<a, b) + R(c, d) = R((a, b) e (c, d)) and
R(a, b) x R<c, d) = R<(a, b) 8) (c, 11))
Theorem 3. The operations of addition and multiplication on the rational
numbers are commutative and associative, and multiplication is distributive
with respect to addition.
62 Introduction to Abstract Algebra
Proof; We prove only the commutative law of addition and the associative
law of multiplication, leaving the others as exercises for the student.
From our previous remarks it follows that we need only to show that
Exercises 6.3
. Prove the commutative law of multiplication for rational numbers.
un—
-(§-§) lid-3)]
l
= —[5 +“:1
_ [ad 23—0)]
66 Introduction to Abstract Algebra
—[ad + b(—c)]
=T
= —(ad) — [b(-c)1
bd
bc — ad
= bd
by the definitions for rational numbers and the properties of the integers.
On the other hand
1: —a c
—+ _ = L2 + _
a
1; d b d
(—a)d + be b4: — ad
— bd _ bd
Note that (5) can also be established as a consequence of the fact that R
is an integral domain (problem 8 of Exercises 6.3) and the fact that if x andy
are elements of an integral domain, —(x — y) = —x + y (problem 4d of
Exercises 5.2).
We pointed out in Theorem 5 that R<l, 1) = 1/1 is the multiplicative
identity for the rational numbers. We now establish the existence of multi-
plicative inverses in R.
Theorem 10. If a aé 0, the unique multiplicative inverse of a/b is b/a.
Proofi (a/b)(b/a) = ab/ba = 1/1 with uniqueness following from the
cancellation law of multiplication for rational numbers.
Exercises 6.5
l. Prove that (—a)/( —b) = a/b.
2. Prove that —[( —a)/b] = a/b.
3. Prove that —[—(a/b)] = a/b.
4. Prove that [—(a/b)][ —(c/d)] = (a/b)(c/d).
Our next theorem has to do with the extension (Definition 5.14) of the
unique (Theorem 5.6) ordering of I to an ordering of R and states that the
extension is both possible and unique.
Theorem 11. The only ordering of R is that defined by a/b >0 if and only
if ab is a positive integer (a, b e I, b 9e 0).
THE RATIONAL NUMBERS 67
Prov/Z We may first observe that if the relation a/b > 0 defined by ab > 0
is an ordering, it is an extension of the ordering of I since, if a e I, a = all
> 0 if and only if a - = a > 0. (Here, of course, we should really write
a/l >30 and a - l >p0 to indicate that, in the first case, we are talking
about our newly defined ordering in R and, in the second case, the previously
defined ordering in I. Which ordering we are considering is, however,
always clear from the context.)
Next we show that if R has any ordering at all it must be the one given.
For, given any b 9E 0 61, we have b > 0 or -—b > 0 by condition (3) of
Definition 5.11 and hence, in either case, b3 = b - b = (—b)(—b) > 0 by
condition (2) of Definition 5.11. Thus if a/b > 0 we must have (a/b)b2
= ab > 0.
It remains to show that the condition a/b > 0 if and only if ab > 0 actually
defines an ordering. That is, by Definitions 5.11 and 5.12, we must show that:
(1) If a/b > 0 and c/d > 0, then (a/b) + (all!) > 0;
(2) If a/b > 0 and c/d > 0, then (a/b)(c/d) > 0;
(3) One and only one of the alternatives a/b > 0, —(a/b) > 0, and a/b = 0
holds.
a c ad+bc
b+ d M
Then (a/b) + (c/d) > 0 if and only if (ad + bc)(bd) = 42(ab) + b2(cd) > 0.
Since 11", b2, ab, and cd are all positive in I we have our desired result.
Similarly, for (2) we have (a/b)(c/d) = ac/bd and ac/bd > 0 if and only if
(ac)(bd) = (ab)(cd) > 0. But (ab)(cd) > 0 since ab > 0 and cd > 0 in I.
For (3) we know that one and only one of the alternatives ab > 0, —(ab)
> 0, and ab = 0 holds. If ab = 0 we have a = 0 since b aé 0 and hence
a/b = 0. If ab > 0 then 11/]; > 0 while if —(ab) = (—a)b > 0 we have
(—a)/b = —(a/b) > 0.
One further point needs to be considered since there is nothing in our
definition of all) > 0 which says that we could not have, for example, 2/3 > 0
but 4/6 < 0 even though 2/3 = 4/6. Now it is easy to check that 4/6 > 0
but we now ask, in general: if a/b > 0 and c/d = a/b, is c/d > 0? To see
that we have an affirmative answer we simply observe that c/d = a/b means
ad = 12:. Then (bc)(ad) = (bc)2 > 0 and since ab > 0 it follows that Cd > 0.
Proofl By Definition 6, a/b > c/d if and only if (a/b) — (c/d) > 0. Now
a/b - c/d = (ad — bc)/bd. Hence, by Theorem 11, a/b > c/d if and only if
(ad — bc)(bd) > 0. Now by Theorem 9, a/(-b) = (—a)/b, so that we know
that we may assume b > 0 and d > 0 and have bd > 0. Hence by the pro-
perties of ordering for the integers we conclude that a/b > old if and only
ifad—bc > Oorad> bc.
Theorem 13. (The Archimedean law) Let x = a/b and y = c/d be
positive rational numbers. Then there exists n e N such that 11.7: > y.
Prov/Z We know by Theorems 9 and 11 that we may assume a, b, c, and d
to be positive integers. We wish to find 11 E N so that
na M c
— > .—
l b b d
By Theorem 12 this will happen if and only if
(1) (1111):! = n(ad) > be.
Now if ad > be we take 11 = 1 and if ad = bc we take n = 2 for the desired
result. If ad < bc we use the division algorithm to obtain
bc=q(ad)+r (OSI<ad)
Thus we have ad > r so that
q(ad) + ad > q(ad) + r = be
Hence
(4 + 1)(ad) > bc
and (1) is satisfied by taking 11 = q + l.
Exercises 6.6
70
GROUPS AND FIELDS 71
Example 3. 13 is a group of order 3 with respect to addition. Co is the
identity for addition; Co is the additive inverse of Co; C1 is the additive inverse
of C2; and 02 is the additive inverse of Cl. Note that this is a finite group
whereas the groups of Examples 1 and 2 are infinite groups.
Exercises 7.1
1. Do the even integers form a group with respect to addition? Do the odd
integers?
2. Do the positive irrational numbers form a group with respect to multi-
plication?
3. In J, which of the following sets form a group with respect to multiplica-
tion: (8) {Ch Ca, Ca); (b) {C0, 01, C2, Cs, 04, Cs, cs); (C) {Ch Ca);
(‘3') (Ch C3, C4, Cs); (e) {Ch Ca: Ca. C4; 05: Cs}?
* It is not sufficient just to write 1/(a + bt/2_). Rather, we must explicitly exhibit
a number of the form r + dfiwhere c, de R such that (a + bl/Z—Xc + tit/E) = 1.
GROUPS AND FIELDS 73
4. Let G = {CkICk 61", [2 ye 0, and (k, m) = 1). Prove that G is a group
under multiplication.
5. Which of the integral domains of problems 1—7 of Exercises 5.1 are fields?
'6. Let P be the set of all positive real numbers and q a fixed positive real
number with q # 1. For a, b E P define
aC-Bb =ab and a ®b =a‘°h".
Show that P is a field with respect to the operations (+3 and 8.
l 2
4 1
which, except for the labels on the vertices, is the original square. We may
indicate R symbolically by giving its effect on the vertices thus:
1234
R=( 2341 )
That is, the vertex 1 is where the vertex 2 was in the original square; the vertex
2 is where the vertex 3 was, and so on. Similarly, we have
2 1 1 4 l 2
3 4 2 3 4 3
2 1 3 2 1 4
v=< 1
2
2 3 4
1 4- 3 > D=< 1
3
2
2
3
1
4
4 > v=<
Then R’ corresponds to a rotation through 180°; R” to a rotation through
l
1
2 3 4
4 3 2 >
270°; H to a reflection about a horizontal line through the middle of the
square; Vto a reflection about a vertical line through the middle of the square;
and D and D’ to reflections about the diagonals. Finally, we may consider
the original position of the square as having been obtained by the identity
transformation,
1 2 3 4
’=( 1 2 3 4- )
Let us now consider the result of applying first R and then H to the square.
We have
4 3 3 2 4 1
R H
—> —>
l 2 4 1 3 2
4 3 1 2 2 3
H R
——> —>
l 2 4 3 1 4
I R R’ R” H V D D’
I I R R’ R” H V D D’
R R I D
R’ R’ I
R” RI! I
H H D’ I
D D I
D’ D’ I
If the student will now compute the remaining products he will see that
we certainly have closure, a right-hand (and also left-hand) identity, and right-
hand (and also left-hand) inverses for every element. Thus only the verifica-
tion of the associative property is needed to proclaim the fact that we have a
(noncommutative) group, called the gmup of symmetries of the square. Un-
fortunately, however, the fact that our operation is indeed associative cannot
be readily seen from the table and to check the associative property case by
case is much too lengthy a procedure. Later we will show that the multiplica-
tion of permutations is always associative but for the moment we will simply
assume this fact.
1 In honor of A. Cayley (1821—1895) who was one of the first mathematicians to con-
duct extensive research in the theory of groups.
76 Introduction to Abstract Algebra
Exercises 7.2
1. Complete the Cayley array for the group of symmetries of the square.
2. Construct the group of symmetries of an equilateral triangle.
3. Construct the group of symmetries of a rectangle that is not a square.
*4. Construct the group of symmetries of a regular tetrahedron.
Exercises 7.3
1. If a, b, and c are elements of a group, prove that the equation bxaxba = bx:
has a unique solution.
2. Prove that, if x is an element of a group and xx = x, then x = e, the identity
element of the group.
3 Prove that (ab)(ab) = (aa)(bb) for all elements a and b of a group if and
only if the group is Abelian.
4. Prove that if G is a group with identity 2 such that a e G implies a2 = e,
then G is an Abelian group.
5. In a group with an even number of elements, show that there is an element
besides the identity that is its own inverse.
6. Let S be a set of elements, a, b, c, . . . and o a binary operation on S
satisfying (3) of Definition 1, and in addition, having the property that all
equations x0 a = b and a o y = b have solutions x and y in S. Prove
that S is a group.
‘7. If x is an element of a group G and there exists y e G such that y’ = x we
will say that x has a square root. Prove (1) if G is a finite group and each
a e G has a square root, then the root is unique and (2) each element of a
finite group has a square root if and only if the order of the group is odd.
*8. Let a1, a2, .. . , an be 11, not necessarily distinct, elements of a group G
of order n. Show that there exist integers p and q, 1 S p S q S n such that
a, a, u . . . a, = e, the identity of G.
*9. Prove that if G is a group with identity e and having more than two ele-
ments, then there exist a, b e G with a # b, a aé e, b 9E e such that
ab = ba.
4. Permutations
p=(. . 12
1112
.)
n
In
78 Introduction to Abstract Algebra
we mean the permutation which sends 1 into jl, 2 into jg, . . . , it into jn
wherejl, jg, . . . , j,, is any arrangement of the integers l, 2, . . . , 11. Note that
the order of the columns in the symbol is immaterial. Thus, for example,
( 1 2 3 ) ( 2 1 3 )
3 1 2 _ 1 3 2
Likewise, our definition of multiplication is a generalization of the multi-
plication used in our example: If
q=(j1ja in)
12112212,,
we have
1 2 ... n
P! =
[21 k2 ... k,l
We have observed that multiplication of permutations is not commutative.
We now prove that, however, it is associative. For if
(12, k2 kn)
7':
m1 m2 m”
wehave
1 2 ... n
p<qr>=(..
h 12 11.
)( jm1 j"'22 ... j
")
mu
(12 ...n)
1711 ma ...mn
Furthermore, we have the identity permutation
( 1 2 ... n )
l 2 ... n
for any 1: symbols and, relative to this identity permutation, we have
_1= his in
P (12 ...n)
Now let us consider the set, S", of all n! permutations on n symbols.
Our remarks above give us the following theorem.
Theorem 3. S,l is a group with respect to permutation multiplication.
Definition 5. Sn 1': called the symmetric group on n symbols.
cnoups AND FIELDS 79
It is often convenient to use the cyclic notation for permutations in which,
for example, the permutation
( l 2 3 4 )
3 1 4 2
is written as (l 3 4 2). We read the new symbol, called a cycle, as follows:
1 is replaced by 3, 3 is replaced by 4, 4 is replaced by 2, and 2 is replaced by 1.
Note that (l 3 4 2) = (3 4 2 1) = (4 2 1 3) = (2 1 3 4). Similarly,
the permutation
( 1 2 3 4 5 )
2 3 l S 4
is written as (l 2 3)(4 5) and the permutation
( 1 2 3 )
3 2 1
as (1 3)(2) or simply as (1 3) where it is understood that when a symbol is
omitted it remains fixed in the permutation.
Definition 6. Two cycles are said to be disjoint if they have no symbol:
in common.
Thus (1 3 4) and (2 5) are disjoint but (1 3 4) and (2 3 5) are not disjoint.
From the procedure used in writing permutations in terms of cycles, the
following result is now evident.
Theorem 4. Every permutation can be written as the product of disjoint
cycles.
Example. Express (1 3 2 5)(1 4 3)(2 5 1) as the product of disjoint
cycles.
Under the first cycle, 1 goes to 3 but then, under the second cycle 3 goes to
1. Finally, under the third cycle 1 goes to 2. Thus we have (1 2 . . .). Now
under the first cycle 2 goes to 5 and then, under the third cycle, 5 goes to 1.
Thus we have (1 2) . . . . Now we look at what happens to 3. First 3 goes to 2
(first cycle). Then 2 goes to 5 (third cycle) so we have (1 2)(3 5 .
Now 5 goes to 1 (first cycle) and 1 goes to 4 (second cycle). Thus we have
(1 2)(3 S 4 . . . . Now 4 goes to 3 (second cycle) so we have (1 2)(3 5 4).
That is, (1 3 2 5)(1 4 3)(2 5 1) = (1 2)(3 5 4).
Definition 7. A permutation that can be written a: a Lycle on just two
symbols, (1'j), is called a transposition.
Theorem 5. Any permutation can be written as a product of transpositions.
80 Introduction to Abstract Algebra
Proof: We first observe that a cycle of n symbols can be written as a pro-
duct of transpositions since
(1 2 3 4...n)=(1 2)(1 3)(1 4)...(1 n)
Our result then follows from Theorem 4.
We note that the expression of a permutation as a product of transpositions
is by no means unique. For example, (1 2 3) = (1 2)(1 3) = (1 3)(l 2)
(1 3)(l 2). In both cases, however, we observe that the number of trans-
positions is even.
Theorem 6. If p is a product of s transpositions and also a product of r
transpositions, then r E 5 (mod 2); i.e., r and s are either both odd or both
even.’r
Prnofi Our proof uses the alternating polynomial, A, in the 11 symbols
x1, x2, . . . , at”. It is defined as the product of the n(n - 1)/2 factors at, — x,
where i <j. Thus ifn = 4
1y . ,)
1|
A = x — x
(xn—l _ xn)
(7) ("1 - x.)(x1 — x1), (x2 — ”0(‘2 " x1)» -- - :(x1-1 — xlxxi-l — “1)
(8) (x: —‘ xt+l)(xl+1 — x1), (x1 — mgxxm — x1), - - - , (x: — x,_1)(x,_1 — ”1)
and the n — j factors in (3) may be paired with the n — j factors in (6)
as follows:
(9) (xi "' mx; " ”#1), (x: — xi+2)(x] " ”1”): - - - . (“t —' ”71X": — 9‘»)-
B‘" in (7), (”k — mm — xj)‘ = (xk " x;)(xk — xi) = (“k " mm - ”1)
fork = 1,2, . . . ,t' -— 1;in (8), (xi — xk)(xk — x,)t = (x, — xk)(xk — x1)
= (x, — .1t,',c)(.1c,c — x,) for k =i+ 1, ,j —- l; and in (9), (x, — xk)
(x, — xk)‘ = (x, — xxx": — xk) = (x, — xk)(x; — Ma) f°Tk=j+ 1: - - - : "~
(The student should check that there are just three special cases in which
the tabulation of factors given in (1) to (6) breaks down. These cases are:
(a) When i = l, in which case factors of the form (1) and (4) do not appear;
(b) when j = n, in which case factors of the form (3) and (6) do not appear;
and (c) when j = i + 1, in which case factors of the form (2) and (5) do not
appear. Because of the pairing of factors in (1) and (4-), (3) and (6), and (2)
and (5), however, these exceptional cases do not invalidate our result.)
Thus the total effect of p on A is simply to change at. — x, to x, — x, and
hence At = —A.
Now suppose p is a product of r transpositions. Then Ap = (—1)’A.
Likewise, if p is the product of 3 transpositions, Ap = (—1)’A. Hence
(—1)' = (—l)‘, r E 5 (mod 2).
Exercises 7.4
5. Mappings
Roughly speaking, a mapping (or function) from one set, S, into another,
T, is a rule that associates with each element 5 e S a unique element t e T.
For example, if S is the set of real numbers and T the set of non-negative
real numbers we may consider the mapping that associates x3 e T with x e S;
x e S —> x2 e T.
For a more precise definition, recall that (Section 1.4) we have defined
the Cartesian product of two sets, S x T, by S x T = {(3, 1)]: e S, t e T}.
GROUPS AND FIELDS 83
Definition 9. If S and T are nonempty sets, then a mapping from S
into T is a subset, M, ofS x T such thatfor every 3 e S
there is a unique 16 T such that the ordered pair (5', t)
i: in M.
In terms of this definition, then, the mapping defined above is the subset
M of S X T. Nevertheless, we find it more useful and convenient (as do
most mathematicians) to continue to think (somewhat loosely) of a mapping
as a rule which associates with any element in S some element in T, the
rule being to associate s e S with t in T if and only if (s, t) e M. We call t the
image of 3 under the mapping.
Now we need some notation. Let a be a mapping from S into T; we write
a : S —> T or S L T
If t e T is the image of s e S under 6 we may also write
a : s —> t or s cr—> 1
Still other notations are, for s E S and t e T the image of s,
t = so or t = 6(5)
The notation t = so is the one preferred by most algebraists; the notation
t = 0(3) by most other mathematicians. Since the latter notation is more
familiar in terms of the customary functional notation we shall utilize it some-
times but will also use the notation t = so when convenient.
Example 1. Consider any set S 4: a and define I. : S —> S by L(:) = u
= s. The mapping 1. is called the identity mapping of S into S.
Example 2. Let S =f= {a and T :4: :5 be sets and t, a fixed element of T.
Then a mapping a of S into T is defined by a z s e S —> t, e T.
Example 3. Let S = N and T = I. Then a mapping 1' of S into T is
defined by -r : h e S —> +k e T (see Section 4.5).
Example 4. Let S = I and T = R. Then a mapping a of S into T is
defined by a z a e S ——> all e T (see Section 6.4).
Example 5. Let S = I x I, T = I, and define a- : S—> T by (m, 11)::
= m+ nform,neI.
Example 6. Let S = {x1, x2, x3} and define -r : S —> S by -r(x1) = x2,
1-(x2) = “a, 7(x3) = x1.
Example 7. Let S = I and T = {00, CI} = 12. Define a : S —> T by
«(211): Co and o-(Zn + 1) = C1 for n 61.
There are two special kinds of mappings that are of great importance.
84 Introduction to Abstract Algebra
Definition 10. A mapping 0' of S into T is said to be onto T iffor every
t e T, then exists an 5 e S such that t = 6(5).
The mapping of Example 1 is clearly onto; the mapping of Example 2 is
onto if and only if T = {ta}; the mapping of Example 3 is not onto since no
element of S maps, for example, into —1. It is left as an exercise for the
student to show that the mappings of Examples 5, 6, and 7 are onto but that
the mapping of Example 4 is not onto.
Definition 11. A mapping, a, of S into T i: said to be a ane-to-one
(1—1) mapping if whenever 51 ye :3 for 31,52 6 S, then
:10- ;é 526' 01, equivalently, if 516 = 320' implies :1 = .52.
The mapping of Example 1 is clearly 1—1 but, unless S has only element,
the mapping of Example 2 is not 1-1. The mapping of Example 3 is 1—1
since if 5,, 52 e N, and 51 aé 52 we have 510' = +51 aé :20- = +32; the mapping
of Example 5 is not 1—1 since, for example, (1, 3) 9e (2,2) but (1, 3)o'
= l + 3 = 4 = (2, 2)o = 2 + 2. It is left as an exercise to the student to
show that the mappings of Examples 4 and 6 are 1—1 but that the mapping of
Example 7 is not 1—1.
Exercises 7.5
5. If S and T are two nonempty sets prove that there exists a 1—1 map of
S X Tonto T x S.
GROUPS AND FIELDS 85
6. Isomorphisms and Automorphisms of Groups
x 1 —1 + l Co Cr
1 1 —1 Co Co Cl
-1 —1 1 Cr C! Co
e e o b c
a a e c b
b b c e a
c c b a e
Then G and G’ are not isomorphic.
Before we investigate the reason for the non-isomorphism let us prove the
following theorem:
Theorem 9. If G g G’ under the mapping a : a e G —>a’ e G’, e is the
identity of G, and e’ the identity of G’, then 0(2) = e’ and if 0(a) = a’,
then 6(a‘ 1) = (a’)‘1.
Proof; (1) Suppose o-(e) = a’ e G’. We let b’ be any element of G’ and
choose beG so that a(b) = b’. Then 6(eb) = c(b) = b’ and also o(eb)
= o(e)e(b) = a’b’. Hence a’b’ = b’ for all b’ eG’ and thus a’ = e' by
uniqueness of the identity.
(2) Suppose o-(a) = a’ and 012—1) = b’. Then o(oa‘ 1) = 6(a) = e’ by
(1). But also 6(aa'1) = tr(a)c(a'1) = a’b’. Hence a’b’ = e’ so that
b’ = ((1’)‘1 by uniqueness of inverses.
If we apply this theorem to Example 3 we may first conclude that if G
is isomorphic to G' under the mapping a, then 0(1) = e. Then we have the
following three possibilities:
(I) «(1') = a; (2) «(1') = b; (3) so) = c
Now if 6(1) = a, then 6(i'1) = a'1 by Theorem 9. However, i'1 = —i and
a‘1 = a so that a(—i) = 1. But this contradicts o-(i) = a since our mapping
must be 1—1. Since b"1 = b and c'1 = e we may similarly show that the
two remaining possibilities are not permissible and hence G ¥ G’.
It was fairly easy to show non-isomorphism in the example just discussed.
In general, however, to decide whether or not two finite groups (of the same
order) are isomorphic is not an easy task. It may require considerable in-
genuity to avoid considering too long a list of possibilities.
Definition 13. An isomorphism of a group G with itself is called an
automorphism of G. That is, an automorphism of G
is a 1—1 mapping of G onto G that is preserved under
the group operation.
88 Introduction to Abstract Algebra
Since an automorphism of a group is an isomorphism we must have e (-) e
and, ifaHb, a‘le—v b‘l.
Example 4. Every group possesses at least one automorphism, called the
identity automorphism, in which a e G 4—» a e G.
Example 5. In G = {1, —1,i, —-i} under multiplication, define a by
6(1) = 1, o-(—1)= —l, 6(1') = —i, and o(—i) = i. This mapping defines
an automorphism of G since it is clearly 1—1 and onto and, for example,
6[(—1)fl = 6(-4) = i = (—1)(-i) = °(-1)6(i);°[(—1)(-i)l = c*(i) = -i
= (—1)i= c(—1)a(—-i), etc.
Example 6. For G as in Example 5, the mapping defined by -r(l) = 1,
-r(—1) = i, 1(i) = —1, and -r(—i) = —i does not define anautomorphism
since 1-[(—1)'1]= 1-(—1)=i;é [-r(—1)]'1 = i"1 = —i.
An important special class of automorphisms is described in the following
theorem.
Theorem 10. Let G be a group and g any element of G. The mapping
6,, : a e G —>g ag‘l is an automorphism of G.
I’IRIR'IR"|”IVlD|D'
a! I R R’ R” H V D D’
an I R R’ R” V H D’ D
on I R” R’ R H V D’ D
an I R” R' R V H D D’
Thus, for example, the second line of the table means that
The student should check that a, = 63', “R = 5n”: a” = av, and on = 00'.
Exercises 7.6
7. Automorphisms of a Field
Then a o 1- is defined by
(a + Ni +ci+ dfii)(ao-r) = [(a + In/E + ci+d\/§i)o]r
=(a—bx/2+ci—dx/§i)-r=a—b\/§—ci+d\/§i
and we again leave it to the student to show that a o -r is an automorphisml
of F.
Proof of Lemma.
(1) p is 1—1 and onto by Lemma 1.
(2) If x,y e F, then (x + y)p = (x +y)(o o -r) = [(x + y)a]-r = (xa + ya)-r
= (xo)r+ (ya):- = x(ao-r) + y(ao-r) = xp + yp since 0’ and 1- are
automorphisms of F.
(3) If x, y e F. then (xy)p = (xc o r) = “may = [(ayonr = [(xam
[(yc)-r] = [x(o- o -r)]|'y(o- o -r)] = (xp)(yP) since a and 1- are automorphisms
of F.
Theorem 13. The set of all the automorphisms of a field F forms a group.
This group is called the automorphism group of the field.
Proof: This follows directly from Lemmas l, 2, 3, and 4 and the observa-
tion that the identity mapping of F onto F is an automorphism of F.
It is obvious that if we considered only mappings that are preserved under
addition we would also obtain a group of such correspondences. Since the
elements of F form an Abelian group under addition, this means that the set
of all automorphisms of an Abelian group themselves form a group. However,
the student may check that nothing in the proof of Theorem 13 involves
the use of commutativity of addition in F. Hence we have the following result.
T In fact, it may be shown that the only automorphisms of F are the identity, 0, r,
and a o r.
94 Introduction to Abstract Algebra
Thearem 14. The set of all the automorphisms of a group forms a group
called the automorphism group of the group.
The concept of an automorphism of a field may be used to prove, in a very
neat fashion, an important theorem of elementary algebra.
Theorem 15. If p(x) = aox" + alx"‘1 + . . . + an is a polynomial with
real coefficients such that p(a + bi) = 0, where a and b are real numbers,
then we also have p(a — bi) = 0.
Now the usual proof of this theorem is rather clumsy, involving a division
of p(x) by [x - (a + bi)][x — (a — bi)] and an argument to show that the
remainder upon this division is zero*.
On the other hand, if we first check that, if (a + bi)c = a — bi is an auto-
morphism of C, we have, from
ao(a + bi)" + a1(a + bi)"’1 + + an = 0
that [ao(a + bi)’| + a1(a + bi)"‘1 + . . . + «do = 06 = 0
(aoa)[(a + bi)a]" + (a,a)[(a + bi)a]"‘1 + + ano' = o
a0(a — bi)" + a1(a — bi)"‘1 + . .. + an = 0.
Exercises 7.7
8. Cyclic Groups
Theorem 16. Let G be a group and a e G. Then (1) amo" = 11"“HI and
(2) (am)" = a""‘ for any integers m and n.
Exercises 7.8
9. Subgroups
G= (I,R.R'.I?'.H, V,D.D'l
\
/mlm
(I,D.D’R’l (I. H, v.1?)
Note that every group G with identity e has at least two subgroups: G and
{9}. (Of course if G is of order 1, G = {e}). By a proper subgroup of a group
G we mean a subgroup other than G or {e}.
Example 1. Consider the group {5-1, 32, :3, :4} with Cayley array
I :1 :a sa :4
n n 5: Sa 34
3: S: 31 $4 33
Sa 58 $4 $1 3!
n :4 :3 s: 31
Then
51 52 53 54 ‘1 ‘2 53 54
5131 5251 3351 3431 31 ’2 33 ’4
S ( 51 ‘2 33 54 ) ( ‘1 ‘2 53 ‘4 )
1' H = =
2 2 5152 5252 5332 5452 32 ‘1 34 ’3
31 52 ‘3 54 51 ‘2 53 3;
3133 3233 3353 3433 53 ‘4 51 52
31 52 ‘3 5; ) (’1 52 53 5';
31:4 52:4 sas4 :434 :4 :3 :2 :1
modulo 12.
2. Find all of the subgroups of the cyclic group of order 12.
3. Let S and T be two subgroups of a group G. Prove that S n T is a sub-
group of G.
4. Let G be a group and S = {xix e Gand ax = am for all a EG}. Prove
that S is a subgroup of G. (It is called the center of G.)
5. Find the regular permutation group isomorphic to the cyclic group of
order 5.
6. Do as in problem 5 for the cyclic group of order 6.
7. Do as in problem 5 for the attic group defined as {51, :3, . . . , s.) where
51 = (1)(2)(3)(4), :2 = (1234). In = (”X“). s. = (1432), $5 = (13).
:5 = (24), 5-, = (12)(34), and rs = (14)(23).
8. Prove the corollary to Theorem 19.
*9. Prove that if a and b are elements of a group G much that ab 9e ba, then
the five elements a, b, ab, ba, and e (the identity of G) are distinct. More-
over, prove that one of a2 and aba is distinct from all these five. Use these
results to prove that a noncommutative group has at least six elements.
*10. Prove that a group that has only finitely many subgroups must be a finite
group.
The motivation for our treatment of the real numbers comes from the fact
that all rational numbers may he expressed as certain kinds of decimals.
102
THE REAL NUMBERS 103
Thus some rational numbers can be expressed as terminating decimals:
2 = 2.0, i = 0.25, l» = 0.125
and so on and all other rational numbers can be expressed as repeating
decimals:
% = 0.3333 . . ., % = 0285714285714 . . ., 3% = 3.83333 . . ..
The reason for this is quite simple. In expressing a/b (a, b e N) as a decimal
by dividing b into a, our successive remainders in the usual division process
are always less than b. Thus the only possibilities for remainders are 0,
l, . . . , b —— l and hence if the division is not exact the remainders must
eventually repeat.
Conversely, if we have given any terminating decimal it is obvious that
we have a rational number: 0.25 = 25/100, 0.125 = 125/1000 etc. Further-
more, as we will prove later, every repeating decimal also represents a rational
number. Since we can consider any terminating decimal as a repeating deci-
mal with 0's (e.g. 2 = 2.0000 . . .) we need only consider repeating decimals
when we are discussing the rational numbers.
The situation then is this: we have a set of repeating decimals that we
associate with the rational numbers. Nonrepeating decimals, then, such as
0.1010010001 . . .
(where we add, after each 1, an additional 0 as a digit) should represent the
irrational numbers.
All of this, however, is very imprecise since we have no formal definition
of a decimal, equality of decimals, and addition and multiplication of decimals.
The rest of this chapter is devoted to making precise the intuitive notions that
we have just sketched.
Exercises 8.1
2. Sequences
The student has met many examples of sequences in his previous work
in mathematics; for example, arithmetic sequences such as 2, 5, 8, 11,
104 Introduction to Abstract Algebra
. . . , 3k — 1, . . . ; geometric sequences such as 2, 6, 18, . . . , 2‘3“”, . . t 7
and the harmonic sequence 1, i, g, . . . , l/k, . . . . We will consider here
sequences a1, a2, . . . ,ak, . . . of rational numbers; that is, al, a2, . . . ,
ak, . . . will be rational numbers. For brevity we write [ak] for the
sequence a1, a2, . . . , am. . ..
We leave as an exercise for the student to prove, using Definitions 1—5 and
the properties of the rational numbers, the following theorem.
3. Cauchy Sequences
Our way of insuring that a sequence will not fluctuate unduly or be un-
bounded is to demand that it satisfy the Cauchy condition.
Definition 6. Let R + be the set ofpositive rational numbers. Iffor any
a ER+ there exists a natural number 4)(ot) such that
when n 2 95(ot) and m 2 95(1), then [an —— am] < at,
we call [ak] a Cauchy sequence. The condition that
la,l — am| < a: is called the Cauchy condition.
Example 1. Suppose ak = l — 1/2". Then consider |an — a,,,| where,
since we are using absolute values, we may take n 2 m so that 1/2"I 2 1/2".
We have
Il
Ian—aml=|(1—%)_(1_2_lfi) =__ ._1 1
<___
2"!
2m 2"
Thus to have Ia,I — aml < at we need only take m such that 1/2’” S at or,
equivalently, choose m such that 2'" 2 1 /a. Clearly such a choice of m e N
is always possible and hence [ak] is a Cauchy sequence. (A formal proof,
using the Archimedean law (Theorem 6.13), that this choice is possible is
left as an exercise for the student.)
Example 2. Suppose that our sequence is 1, 0, l, 0, l, . . . where ak = 1
if k is odd and ak = 0 if k is even. Then no matter how large n is we have
la"+1 — an! = l and hence [ak] is not a Cauchy sequence.
Example 3. Suppose our sequence is l, 2, . . . , k, . . . . Again [an+1 — a]
= l and [ak] is not a Cauchy sequence.
It is intuitively plausible that the terms of a Cauchy sequence cannot get
arbitrarily large since the later terms differ by an arbitrarily small amount.
We phrase this fact more formally in our next theorem.
Theorem 2. If [ak] is a Cauchy sequence, then there exists a B e R+
such that |ak| < B for all k e N. (Such a number B is called an upper bound
for the sequence.)
Proofi By the properties of absolute value (Theorem 5.9) we have
lakl = lar + (”k " ar)l S larl + lak _ ail
106 Introduction in Abstract Algebra
Now since [11k] is a Cauchy sequence there exists ¢(1) e N such that when
k, r 2 95(1), |ak — (1,] < 1. Hence for all k, r 2 ¢(1) we have
laul < larl + 1
Now pick any 7 2 95(1); say 7 = 95(1). Then, for k 2 ¢(1) we have
lak| < |a,| + 1 = c where ceR+ is a constant. Thus |ak| < c for all
k 2 56(1). To take care of |ak| for k < ¢(1) we consider the finite set of
positive rational numbers {Ia1|, |a2|, . . . , |a¢(1)|, c} and let B be any rational
number such that B > |ai| for 1': 1, 2, . . . , 95(1) and also B > c (e.g.,
take B =|a1| + |a2| + . . . + la¢(1)| + 6). Then lakl < B for all k EN.
NOTE. The choice of a = 1 in ¢(a) above is purely arbitrary. Any at e R+
would do.
Example 4. Supposed1 = 20, a2 = —3,a3 = 4-,a5 = 1 + Lac = 1 + i,
. , «1,, =1 +1/2"‘4 for k > 4. Then for 12,1 2 5, |ak — (1,] <1.
Thus for all k 2 5, Iak| < Ia5| + 1 = 5/2. Now let B = [20] + [—3l
+ |4| + 5/2 = 29%. Clearly Ink] < B for all 1: EN. (It is also clear that
smaller upper bounds exist; e.g. 20.)
Theorem 3. The sum of two Cauchy sequences is a Cauchy sequence
and the product of two Cauchy sequences is a Cauchy sequence.
Proof: Let [ak] and [bk] be Cauchy sequences. Then [ak] + [bk] = [ah
+ bk]. Consider
Exercises 8.3
l. Prove, using the Achimedean law, that for any a: e R+, there exism m 6N such
that 2" z 1/ a.
2. Prove that S = ([ak]l[ak] is a Cauchy sequence} is not an integral domain
by exhibiting divisors of zero.
3. Prove that if B is an upper bound for a sequence [ax] and A > B, then A
is also an upper bound for the sequence [an].
4. A sequence [ax] is said to be bounded if there exists B e R" such that for
all k e N, Iakl < B. Prove that the sum and product of bounded sequences
are bounded sequences.
4. Null Sequences
Let us recall the motivation behind our discussion to date. We are trying
to associate sequences with our intuitive notion of a real number. Our prob-
lem is that there is a natural association of more than one sequence with any
real number. Thus, for example, the sequences
[l—l/k], [1—1/2k], and [1—1/k3]
are all naturally associated with the number 1. Our basic device for handling
this problem is that of a null sequence.
Definition 7. [ak] is a null sequence If and only if for every a: e R+
there exists (flat) 6 N such that n 2 95(11) implies
|an| < at.
Example I. [l/k], [1 [2"], and [1 /k"] are all null sequences.
We leave to the student the proof of the following two theorems.
108 Introduction to Abstract Algebra
Theorem 5. Every null sequence is a Cauchy sequence.
Theorem 6. Let [ak] and [bk] be null sequences and [ck] be any Cauchy
sequence. Then
Exercises 8.4
1. Prove Theorem 5.
2. Prove Theorem 6.
3. Prove that the set of all null sequences is a commutative ring without a
multiplicative identity.
{[ak]llak] ~ [1]}
THE REAL NUMBm 109
and the class
{laklllakl ~ [ZED
Definition 9. The set of all equivalence classes of {[ak]|[uk] a Cauchy
sequence} is the set of real numbers, R*.
We now need to define addition and multiplication for these classes.
A X B = A3 = {[zk]i[zk] ~ [ak][bk]}
Here, of course, we should write A G B and A ® B to distinguish the
newly defined addition of classes from the previously defined addition and
multiplication of sequences. At this point, however, the student can certainly
keep this distinction in mind without the need of a special notation.
As usual, we need to prove that our definitions are independent of our choice
of representative of our equivalence classes.
Exercises 8.5
1. Discuss the difi'erence between the natural number 2, the integer 2, the
rational number 2, and the real number 2.
2. Give the details of the proof of Theorem 9.
3. Consider the example that you previously used to show that the set of all
Cauchy sequences was not an integral domain. What “happens” to this
example here? (i.e., why cannot it be used to show that R“ is not an integral
domain?)
Ultimately we will show that R“ is an ordered field but let us first consider
how the rational numbers appear as real numbers. Among the elements of
R“ are those classes {[xk]|[xk] ~ [7]} where r e R. Let us, for brevity
write {[xflllxk] ~ [7]} = {M}-
Theorem 10. The rational numbers are isomorphic to the set of elements
{[r]} of R* via the correspondence
1' eR<—>{[r]} eR*
Proof. This follows immediately from the fact that if r1, r2 ER, then
{[71]} + {[72]} = {[71] + [72]} = {[71 + r2]} and {D11} X {['2]} = {[71]['2]}
= {Dial}.
At this point let us return to the statement made in Section 1 that every
repeating decimal represents a rational number. First of all, let us note that
every repeating decimal can be represented as a rational number plus a sum
of terms in a geometric progression. Thus, for example,
0.3333 = 0.3 + (0.3)(1‘7) + (0.3)(i—‘o)2 + - -- + (0.3)(fi)"‘1 + ---,
7.4393939 - - - = 7.4 + 0.039 + (0-039t) + - - - + (0.039)(—,.;3)"‘2 + . . .,
1.237237 = 1 + 0.237 + (0.237)(”;,) + + (0.237)(,—;fi)n-2+ -
In general we have
’2
[ale] ~ ['1 + m] = ['1
and thus that
r2 .3 3 9 l
r=rl+1—-_—@;=O+1 _%6 =1—0+1—0=3
12 0.039 74 39 99
”Hm“ +mh (@7136)
37 13 12,275 491
5 +330" 1650 _ 66
Now from the theory of geometric progressionst we know that if
S = b + bc + be” + + bcn‘1,then
c" - l
S= b
c -— 1
Applying this formula to ak — 71 with b = 72, c = 1/10‘, and n = k we have
ak = r1
+ r2 (1%)"
—
—1
(96):; __ 1
=[(—:1._.)f W]
f See any college algebra textbook.
112 Introduction to Abstract Algebra
is a null sequence. That is, given any at E R+ we want to be able to choose
flat) 6 N such that, when n 2 Mat),
12 ( 1 )8»
Thus we want
r - (3-0)” = —- —
1 — (1‘5)‘ 10
<a
(+3)" < 0t —
1 — (a)s
’2
01'
Exercises 8.6
72
00')" >——m[1 _ (1'16")‘1
Since one of out motivations for introducing real numbers was the non-
existence of a rational number r such that r2 = 2, it is in order here to show
that there is indeed an element U 6 R* such that U2 = {[2]}. To do this we
define the sequence [14k] as follows:
_ “A: + (2/“Ic)
k+1 2 (keN)
THE REAL NUMBERS 113
(This type of definition is known as a recursive definition. It yields :11 = 1,
u2 = 1.5, 143 = 17/12 etc.) We shall prove that [uk] is a Cauchy sequence
such that if U = {[xk]|[xk] ~ [uk]} eR*, then U x U = U2 = {[2]}
(H 2 e R).
We proceed to the proof of U2 = {[2]} in several stages, the details of which
we leave to the student.
(1) For all 12 EN, uk > 0 and u: 96 2 since each uk 6 R.
2 a 2_ a
a)nnmkeMuLI—2=(”+2)—2=(” 2)
2a,, 2a,,
(3) By (1) and (2), for all k e N such that k 2 2, u: > 2.
(4) For all keN, uk 21 since 141 = 1 and if uk <1 for any It > 1, it
would follow that a; < 1 contrary to (3).
(5) For allkeN, 0 < 14:” — 2 51/2"+1.
Proof of(5). Now a:+1 - 2 > 0 by (3) and we proceed to prove the
second part of our inequality by induction. First, if k = 1 our inequality
becomes
fi—zss=l
But It; — 2 = (3/2)2 — 2 = 1/4- so that our inequality holds when k = 1.
Now assume that the inequality holds for k = n so that
fifl—zsurfl
Now
fl —22 2 —22
("7H s (1H
n+1
by (2) and (4-). But then, by our induction hypothesis,
2 l 2 1 2 l 2 1
“Wm—25 (E) (2n+1) =(2n+2) < 2n+2
Thus our inequality holds for k = 1 and, if it holds for k = n, it holds for
k = n + 1. By induction it then follows that it holds for all k e N.
(6) If a: €R+, then there exists a t eNsuch that 1/2”1 < at.
Proof of(6). We take x = 1 and y = 1/0: in Theorem 6.13 (the Archi-
medean law) and have that there exists 5 eN such that s - l = s > Ila.
Now let S = {nln e N and there exists t e N such that 2‘+1 > n}. We wish
to show that S = N and thus prove that, no matter how large n is, we can
find a t EN such that 2H1 > 11. Now 1 e S since we may take t = 1 and
have 21+1 > 1. Suppose now that k e S. Then there exists 1 e N such that
2”1 > k. But then 2-2”1 > 2k so that
2U+1l+1>k+kzk+1
114 Introduction to Abstract Algebra
and since t + 1 e N we have shown that, if k e S, k + l e S. Thus S = N.
Since S = N, for s > l/ot as chosen above, we may find a t e N such that
2H1 > s > Us:
Hence 1/2”1 < a as desired.
(7) The sequence [ufi] — [2] = [14%, — 2] is a null sequence.
Proof of(7). By (5) we have
ht:+1 — 2| S 1/2"+1
Now for any at e R+ we take 9501) = t for t as determined in (6) and have,
for n 2 flat) = t,
|uf|+1 — 2] S 1/2"+1 < at
since, ifn 2 t, 0/2"”) S (1/2‘”).
(8) [1:3] is a Cauchy sequence.
Proaf0f(8). By Theorem 5, [14: — 2] is a Cauchy sequence since it is
a null sequence. Hence for any at e R+ there exists ¢(u) e N such that when
n, m 2 <l>(a).
[(ufi — 2) - (ufn — 2)| = l”: — 143'] < at
Thus [ufi] is a Cauchy sequence.
(9) [uk] is a Cauchy sequence.
Praofof(9)- lug; _ “ii = lun "‘ umllun + “ml = lun _ uml(un + um)
since u,c > 0 for all k e N. Now since [1412‘] is a Cauchy sequence, for any
a e R+ we may obtain 45(24) 6 N such that for n, m 2 ¢(2m), lug| — ufn|
< 20:. Hence [an — u,,,[(un + u,,,) < 2:: so that for n, m 2 ¢(2oc),
I I 20: < 25:
u — u < ——~ _ — =
n m u“ + um 2 on
since 14,, 2 1 for all k e N by (4). Thus [uk] is a Cauchy sequence.
(10) We now consider {[uk]} and have {[195]}2 = {[155]} = {[2]} by (7)
and our definition of equality for real numbers.
Exercises 8.7
2. Generalize the method used in establishing the existence of 1/5 and 1/3; e R“
to establish the existence of (a for any a e R‘.
THE REAL NUMBERS 115
3. Use the sequence u; = l, u“; = [1/(uk + 1)] + 1 to prove the existence
of V2 eR'.
(1) [01:] > [0] (2) [‘11:] ~ [0] (3) [‘11.] < [0]
Proof; We proceed by showing in turn that
(a) If (2) holds, neither (1) nor (3) can hold;
(b) (1) and (3) cannot both hold;
(c) If neither (1) nor (3) holds, then (2) holds.
From (a) and (b) we will then conclude that at most one of (1), (2), and (3)
116 Introduction to Abstract Algebra
holds, and from (c) that at least one of (1), (2), and (3) holds. Thus the
establishment of (a), (b), and (c) will provide a proof of the theorem.
We now tum to the proofs of (a), (b), and (c).
(a) Suppose [ak] ~ [0] and also [ak] > [0]. We then let (m, 1) be a test
pair for [ak]. Since [ak] ~ [0], there exists «fir/2) 6N such that when
n 2 950/2), an < r/2 eR". Now choose 71 2 max {01, ¢(r/2)} and have
an < 7/2 because [ak] ~ [0] and also an 2 1' because [11k] > [0]. From this
contradiction we conclude that if [ak] ~ [0], then we cannot have [ak] > [0].
On the other hand, if [11k] ~ [0] and also [ak] < [0] we observe that then
[-ak] ~ [0] and [—ak] > [0] so that we again have a contradiction.
(b) If [ah] > [0] and also [ak] < [0] it follows that [ah] > [0] and also
[—ak] > [0]. But then, by Theorem 11, [ak] + [—ak] > [0] and since [ak]
+ [—ak] = [0], we have [0] > [0], contrary to (a).
(c) Suppose [ak] :1» [0] and [ak] 4: [0]. Then [—ak] :1» [0] for if [—ak]
> [0] it would follow by Definition 12 that [—(—ak)] = [11k] < [0], a con-
tradiction. Since [ak] is a Cauchy sequence we know that for any 7 e R+
there exists ¢(r/2) EN such that when n, m 2 (fir/2), we have Ian —— aml
< 7/2. But since [ak] :l> [0] it has no test pair and, in particular, (fir/2),
1/2) is not a test pair. Hence there exists a natural number m 2 95(1/2) such
that (11,,l < 1/2. We now write
an=am+(an_am)sam+lan—am|
and have, for n 2 “1/2)
a" < (7/2) + (7/2) = r
for any 1 eR”. But, by the same argument, we may use the fact that
[—ak] 11’ [0] to obtain -a,l < r and hence that, for '1 2 “7/2)’ Ia“l
= Ian .. 0| < r. Thus [an] ~ [01‘
Theorem 13. If [ak] > [0] and [bk] ~ [0], then [ak] + [bk] > [0].
Similarly if [ak] < [0] and [bk] ~ [0], then [ak] + [bk] < [0].
Proof; Suppose [uk] > [0] and consider [ck] = [11k] + [bk]. By the tricho-
tomy law (Theorem 12), [ck] > [0], [ck] ~ [0], or [ck] < [0]. But if [ck]
~ [0], then (Theorem 6), [ak] = [ck] — [bk] ~ [0] which is a contradiction
to the trichotomy law. Similarly, if [ck] < [0] we have [—ck] > [0] and
hence, by Theorem 11, [—ck] + [ak] > [0]. But then [bk] = —([-ck]
+ [ak]) < [0] which is again a contradiction to the trichotomy law. Hence
[ck] > [0].
The proof of the second part of the theorem and the following corollary
are left as exercises for the student.
Corollary. If [ak] > [0] and [bk] ~ [ak], then [bk] > [0]. Similarly, if
[ak] < [0] and [by] ~ [ax], then [bk] < [0].
THE REAL NUMBERS 117
We are now in a position to define the concept of positiveness for real
numbers.
Theorem 14. IfA, B eR* and A, B > 5, then A + B > U and AB > 0.
Proof Let A—
= {[ak]} and B—
= {[bk]} where [ak] > [0] and [bk] > [0].
Then A + B = {[ak] + [bk]} and AB = {[ak][b,,]} where [ak] + [bk] > [0]
and [ak][bk] > [0] by Theorem 11. Hence A + B > G and AB > 0.
We leave the trichotomy law for real numbers as an exercise for the student
to prove using Theorem 12.
Theorem 15. If A e R*, then one and only one of the following alterna-
tives hold: (I) A > (7, (2) A = 6, or (3) —A > 0.
Exercises 8.8
The student will recall (Theorem 9) that we have shown that R“ is a com-
mutative ring with a multiplicative identity 7 = {[1]}. Thus to show that R*
is a field we need only to show that every nonzero element of R* has a
multiplicative inverse. (Cf. Definitions 5.4 and 7.3.)
The basic result we need to establish this result is embodied in the follow-
ing lemma.
Lemma. Suppose [ak] is a Cauchy sequence and [uh] > [0] with test pair
(t, r). Then if [bk] is defined by bk = 0 for k < t and bk = ak‘l if k 2 t,
we have (1) [bk] is a Cauchy sequence and (2) [ak][bk] e {[1]}.
Proof; (1) Since [ak] is a Cauchy sequence, for any x e R+ we know that
there exists (“512) e N such that if n, m 6N and n, m 2 #312) we have
]an — am] < 3r“. Now let 5(5) = max {t, ¢(512)}. Then if n, m 2 11(5) we
have
a —-a lam — anl "2
lbn "‘ bml = Ian—l _ “In-” = il—n = —— —a = s
anam and,” r
Hence [bk] is a Cauchy sequence.
(2) Consider the sequence [ak][bk] — [1]. Now akbk — 1 = 0 for all
k 2 t and hence it is obvious that [ak][bk] -— [1] ~ [0] and thus that [ak]
[bk] 6 {[1]}-
Thearem 17. R* is a field.
Proofl Suppose A = {[ak]} eR* and A 95 0. Then [ak] w [0]. If [ak]
> [0] we invoke our lemma to conclude that there exists a Cauchy sequence
[bk] such that [ak][bk] ~ [1]. Now we let B = {[bk]} and have AB = {[ak]
[bk]} = {[1]} as desired.
If [ah] < [0], then [—ak] > [0] and we again invoke our lemma to obtain
a Cauchy sequence [—bk] such that [—ak][—bk] ~ [1]. Finally, since
[—aklt—bkl = [aknbu we have AB = {[ak][bk]} = {[11} as before.
Example. Let A = {[ak]} where a1 = —1000, a2 = 0, a3 = 17/5, a4 = 1,
and
ak + 2/11,. (1,,2 + 2
ak+1 =
2 " Zak
THE REAL NUMBERS 119
for k 2 4. Then if]:1 = b2 = 0, 123 = 5/17, I;4 =1,
b _ Zak
n+1 — for k 2 4-,
“k 2
and B = {[bkl}! we have AB = {[ak][bk]} = {[1]}-
Exercises 8.9
1 1-1 1_1 1 1 1
1 ,2 ,3 ,4 ,5 ,6’,7 a ""'
l / / I I / /
1’ 1’ 1’ 1’ 3.’ 1’ 1 l .....
1 ,2 ,3 ,4 ,5 ,s 7 B
I I / / I
1’ 1’ 1’ 1’ 1’ 1 1 1
1 ,2 ,3 ,4 ,5 s 7 s
l/ l / I
4/ —
— 4/ —4/ —4/ —4 a
— 4
— 4
— .....
1 /2 ,3 M 5 6 7 a
/ l ’
1’ 1’ 1’ 1 1 1 1 1 .....
I ,2 ,3 4 5 s 7 s
I / /
1’ 1’ 1 1 1 1 1 1 .....
1 ,2 3 a 5 s 7 a
1/1
1 2
13 14 15 16 77 7
s ""'
and proceed as before along the dotted line to determine the matching.
Thus, for example, 3 e N —> (11, 73) e R+ x R“.
Because of the fact that I is equinumerous with N it is certainly plausible
that R is equinumerous with R+ and that the set of ordered pairs of elements
of R is equinumerous with the set of ordered pairs of elements of R+; we
leave the proofs to the student. Hence if it were possible to represent the
real numbers as pairs of rational numbers, it would follow that R“ would be
equinumerous with N. We will show that this is not the case and, indeed,
122 Introduction to Abstract Algebra
that even the set of real numbers r with 0 S r S 1 is not equinumerous with
N.
We recall that a real number, r, is a set of equivalent Cauchy sequences:
r = {[xk]|[xk] ~ [ak]}. For our purposes here, however, it will do no harm
to regard a real number as a single sequence as long as we take care not to
list two equivalent sequences. In particular, let us consider the sequence
[1k] where 71 = 0411, r2 = 0.ala2, . . . , rk = 0.ala2 . . . ak where a1, a2, a3, . . .
are in the set {nln e N and 0 S n S 9}.
We leave it to the student to prove that such sequences are Cauchy
sequences and hence are representatives of the class of sequences that are our
real numbers. We also leave to the student to prove that the only such
sequences that are equivalent (but not equal) are those in which we have
eventually a repetition of 9’s on the one hand and arepetition of 0’s on the
other; e.g., 0.32000 . . . and 0.31999 . . . . To avoid such situations we will
agree not to use as our representatives such sequences that end in a repetition
of 9’s.
Now let P be the set of all such sequences [rk] as just described and suppose
that we do have a 1—1 correspondence of P with N as indicated below:
leNH0.a1u2a3... =aeP
26N<—>O.b1b.3b3 =beP
36NH0.c1c3c3 =reP
4eN4—>0.d1d2d3... =deP
SENH0.e1e293 =eeP
Exercises 8.10
Exercises 9.1
(x+1)(x+2)=(x+l)(x+2) x2+3x+2=fi+3x+2
=——1+3x+2
=(—l)+3x+2=3x+1
In the following chapter we will consider this particular set of residue
classes in a little more detail since it is obvious that it “looks” like the set of
complex numbers. We continue now, however, with the general theory.
Proof. We first need to show that (2) and (3) are adequate definitions of
addition and multiplication. That is, we need to show that if 5 =1? and
i :7, then
1 Up to this point in the proof we have used only the fact that R is a subring. Now
we need to use the fact that R is an ideal since it does not follow that s? = s—’t7if S is
a subring of S but not an ideal of S. See problem 6 of Exercises 9.2.
RINGS, IDEALS, AND HOMOMORPHISMS 131
Once our operations are shown to be well defined it is very easy to verify
the ring properties for S — R. Thus, for example, t'+ :- = E + I since
i+§=T¥1§+i=s+—‘t, and t+s=s+t. Observe that O is the
identity for addition and the additive inverse of E is 3.
A suggestive notation for 5 = {xlx e S and s — x eR} is s + R. Then
s'+t'=(s+R)+(t+R) = (s+t)+R and §i=(s+R)(t+R)=:t
+ R. Furthermore, if a ER, (5 + a) + R = s + R. In this notation, our
first example of 13 has the elements 5 = 0 + [3], 1 = 1 + [3], 2 = 2 + [3]
and our second example has the elements ax + b + [x2 + 1] where a, b e R*.
Then, for example, (2 + [3]) + (2 + [3]) = 4 + [3] = (1 + 3) + [3] = 1
+ [3] since 3 e[3] and (x + [x2 + 1])(x + [:c2 + 1]) = x2 + [x2 + 1]
= —1+(x2+1)+[x3+1]= —1+[x2+1]sincex2+le[x2+l].
Exercises 9.2
3. Homomorphisms
Consider I and I — [2]. The mapping that sends even integers into 5 and
odd integers into 1 is certainly not an isomorphism of I and I — [2]. Never-
theless the correspondence is preserved under addition and multiplication
since
odd+odd =even—>T+1=0
odd+even=odd—>1+U=T
even+even=even—>0+0=6
oddxodd =odd—>I><l'=T
oddxeven=even—>Tx0=5
even x even = even—>6 x O = 5.
132 Introduction to Abstract Algebra
Definition 4. A homomorphism of a ring S into a ring S’ is a mapping
a : S —> S’ with the property that the correspondence is
preserved under addition and multiplication; that it, if
s, t E S, then (s + t)o = so + to and (“)6 = (so)(tc). If
for every 5’ e S’ there exists an s e S such that so- = s’ we
say that the homomorphism is onto.
(n1+n.)c=2:*+7;=rz+n:=nla+naa and
(“1"2) 5: "1—"; = "—177; = ("16)("213'l
An isomorphism is a homomorphism that is 1—1 and onto. In general,
a homomorphism is a many-one mapping and is not onto. Thus in the
homomorphism of 1 into I -— [2] we have —2 —>(_), 0 —-> 6, 2 -—> 5 and so on,
although, in this case the mapping is obviously onto.
On the other hand, if S = I[x] and S’ = {a + bflla, b e R}, the mapping
a defined by f(x)o = flfl) for f(x) 6 I[x] is a homomorphism that is neither
1—1 nor onto. Thus, for example, (2x)o = 2V2 = x% so that the mapping
is not 1—1. Furthermore, if we consider 1} + V2 6 S’ it is clear that there
is no f(x) eI[x] such that f(x)c = 4; + fl. However, the mapping is a
homomorphism since for f(x), g(x) e I[x], we have
(5+i)r=(3+_t)r=(s+t)’=(s+t)o=sc+to=s’+t'=§r+t‘r
and
(ii):- = (3)1 = (st)’ = (st)o = (so)(to) = s't’ = (ma?)
where we are using the definitions for f + t' and if given following Theorem 3.
(2) The mapping described is an onto mapping since every .6 in S — R is
equal to so for s e S. Furthermore, (s + r)o = s + r = E + i = so + ro
and (sr)o = $7 = if = (so)(ro) so that the mapping is a homomorphism. The
kernel of the homomorphism consists of those elements 3 ES such that
so = 3 = 6; i.e., those elements 5 e S such that r e R.
As a second example of our theorem, consider the homomorphism o
defined by f(x)o =f(fl) where f(x) 6 S = I[x] and f(\/§) e S' = {a + b
Vila, b 61}. If f(x)o = 0’ = 0 + 04/2 6 S’, thenffi/f) = 0. The ideal
K, = {f(x)]f(x) EI[x] and f(\/§) = 0} of I[x] is the kernel of the homo-
morphism. We have (problem 4 of Exercises 9.1) that K, = [x2 — 2].
RINGS, IDEALS, AND HOMOMORPHISMS 135
Exercises 9.3
4. Homomorphisms of Groups
ea = [(1)(2)(3)(4)]c = e' = «3
[(12)(34)]<r = e'
[(13)(24)]cr = e’
[(14)<23)]c = e'
(123)0' = a (132)0' = 112
(243).: = a (234»: = a2
(l42)o‘ = a (124)a = a2
(134)0' = a (143)0' = a2
Proof: Let e be the identity for G. Then for any gh egH we have
gh = (gh)e e(gh)H so that gH E (gh)H. On the other hand, for any
(gh)h1 e (gh)H we have (gh)h1 = g(hh1) = gh3 egH since H is a subgroup of
G. Hence (gh)H g gH so that gH = (gh)H.
iS=bS={i,b}=Si=Sb;
aS=cS={a,c}=Sa=Sc;
dS=eS={d,e}=Sd=Se;
fS=gS={f:£}= Sf=Sg
RINGS, IDEALS, AND HOMOMORPHISMS 139
I iS aS dS fS
is is (IS (IS f8
as as iS fS dS
as as fS iS 113
f8 f5 as aS iS
I e x y z
e e x y z
x x e z y
y y z e x
z z y x e
(2) Using the results of Example 6, with G as the octic group and H =
{3} b}, we have 1', b—>iH; a, c—>aH; d, e—)dH; f, g—>fH defining our
homomorphism of G onto G/H. The elements 1' and b map into if! so that
the kernel of the homomorphism is {i, b} = H.
Theorem 11 tells us, in effect, that the problem of finding all of the homo-
morphisms of a group is equivalent to the problem of finding all of the
normal subgroups of a group. In trying to determine the subgroups of a
finite group, a powerful tool is the fact that the order of a subgroup of a
group must divide the order of the group. We will establish this fact in the
next section; in the meantime, however, it may be used to advantage in
some of the problems given in the following set of exercises.
RINGS, IDEALS, AND HOMOMORPHISMS 141
Exercises 9.4
1. Let G be the octic group (Example 3) and let S = (i, e}. List the right
cosets of S in G.
2. Let G be the symmetric group on four symbols and S be the octic group.
List the left cosets of S in G.
3. Let G be the alternating group on four symbols and S = {(1)(2)(3)(4),
(12)(34), (13)(24), (14)(23)}. List the left cosets of S in G.
4. Determine all of the normal subgroups of the group of symmetries of
the square (Section 7.9) and determine the quotient group for each of
these normal subgroups.
5. Determine all of the normal subgroups of the symmetric group on three
symbols and determine the quotient group for each of these normal
subgroups.
6. Set up a homomorphism of the additive group of integers onto the
multiplicative group {1, —1, i, —i) (i = 1/__1). What is the kernel of
the homomorphism?
7. Which of the following mappings map the multiplicative group of all
nonzero real numbers homomorphically on a part of itself? If the mapping
is a homomorphism, identify the homomorphic image and the kernel of
the homomorphism.
(a) x —> -x; (b) x -> lxl; (C) 9: ->3x; (d) x 9x“;
(e) x ->x“; (f) x ->1lx; (g) x —> —1lx; (h) x ->{/§-
8. Set up a homomorphism of the octic group onto the group ((1)(2)(3)(4-),
(12)(34), (13x24), (14)(23)). What is the kernel of the homomorphism?
(Hint: See Example 7.)
9. Set up a homomorphism of the octic group onto the cyclic group of order
two. What is the kernel of the homomorphism? (Hint: Find a normal
subgroup of order four of the octic group.)
10. In problem 4, find the homomorphisms determined by the normal
subgroups.
11. Do as in problem 10 for problem 5.
12. If H is a subgroup of a group G, prove that for g e G, gH = H if and only
if g e H.
13. Prove Theorem 7.
l4. Prove Theorem 8.
15. Prove the corollary to Theorem 11.
16. Prove that the homomorphic image of a cyclic group is cyclic.
17. Prove that the center S (problem 4, Exercises 7.9) of a group G is a normal
subgroup of G and that GIS is isomorphic to the group of inner auto-
morphisms of G.
I’18. If S, and S: are two subsets of a group G, define their pioduct, S1 *Sg,
as S, #82 = ($133151 6 S1, 52 6 S3}. Then (1) prove that if H is a normal
subgroup of G, then (aH) II (11H) = (ab)H for all a, b e G; (2) if (aH) # (bH)
is a left coset of S in G for all a, b e G, then H is a normal subgroup of G.
‘19. Let G be a group. For a, b e G we define the commutator of a and b
as a“b"ab. Let S be the set of all products of such commutators; prove
that S is a normal subgroup of G.
’20. In problem 17, prove that G/S is Abelian.
*2]. Two subgroups S and T of a group G are called conjugate if there exists a
g e G such that g'1Sg = T. Given any subgroup S of G prove that the
142 Introduction to Abstract Algebra
intersection of S with all of its conjugates is a normal subgroup of G.
1‘22. In problem 4 of Exercises 7.9 the center of a group G was defined. Now
define the rim, R(G) of a group as R(G) = {a E Glab = ba implies there
exists c e G such that a = :1, b = c" for integers j and k} and the anti-
center of G, AC(G), as AC(G) = {axag . . . anlal E R(G)). Prove:
(a) If e is the identity of G, then e e R(G);
(b) a e R(G) implies a—1 e R(G);
(c) a e R(G) and b e G implies b“ab E R(G);
(d) AC(G) is a normal subgroup of G;
(e) If G is cyclic, then R(G) = AC(G) = G;
(f) If G is finite and Abelian, then R(G) = AC(G) = G if and only if
G is cyclic;
(g) G g 0‘ implies AC(G) ; AC(G‘);
(h) AC(G) = AC(AC(G)).
5. Lagrange’s Theorem
In this section we continue to work with left cosets but the student should
note that our results, here and in the previous section, could equally well
be phrased in terms of right cosets. For normal subgroups, of course, the
concepts are identical since if H is normal, then gH = Hg.
Definition 9. If S is a subgroup of a group G we will say that a e G is
left congruent to b e G if and only if a ebS. We write
a E b (mod S).
If we let G be the group of integers under addition and S = {3n[n e I},
the cosets of S are 0+S=S,1+S={3n+l|nel}, and 2+S=
{3n + 2|n 61} (using, of course, the additive notation). Then to say that
aEb(mod S) means thataeb+ Sorthata =b+s=b+ 311 or that
o — b = 3n. Thus a E b (mod S) means the same as our previous a E b
(mod 3) and this fact helps to explain the language used in Definition 9.
Lemma 3. The relation of left congruence is an equivalence relation on G.
Proof: (1) a E a (mod S) since a = ae eaS; (2) If a E b (mod S) then
a = brebS. Thus b = as“ eaS so that b E a(mod S); (3) If a -=—b
(mod S) and b E 1: (mod S), then a = bs1 ebS and b = cs2 6 LS. Hence
a = (“2);-1 = 4:251) = (:33 e S and a E I: (mod S).
We may now apply Theorem 4.1 to conclude that a group G is partitioned
by the classes Pb = {ala e G and a ebS} = bS. In particular, then, every
element of G is in one and only one coset of S. Thus if aS and bS are two
cosets, either aS = 115 or aS n bS = (a . Furthermore, if G is finite, the
number of elements in any two cosets is the same and is equal to the order
of S.
RINGS, IDEALS, AND HOMOMORPHISMS 143
Theorem 12. If (15 and bS (So and Sb) are two left (right) cosets, then
either aS = bS(Sa = Sb) or aSn bS = $(San Sb = o).
Proof: Let n be the order of G and s the order of S. Separate the elements
of G into left cosets. By our previous remarks, every element of G is in one
and only one left coset and each left coset contains the same number of
elements, say k. Hence 1: = rk and sin.
Exercises 9.5
The results of Section 9.2 suggest that we might define the set of complex
numbers as the ring C = R*[x] — [ac2 + 1]. Computations in C then pro-
ceed as in R*[x] except that, whenever 52 appears, we replace it by :1.
Therein I. The ring C = R*[x] — [x2 + 1] is a field that contains a sub-
field R g R*.
Prob/2 We know in general (Theorem 9.4) that the difference ring of a
commutative ring is a commutative ring. It is obvious that T is a multipli-
plicative identity for C. Furthermore, if m 9e 5 l then one, at least,
of a and b is not zero and we observe that
Exercises 10.1
l The notation a + bx used here is, of course, the notation adopted for residue
classes in Section 9.2 and is not be be confused with the notation a + bi = a — bi
for the so-called complex conjugate of a + bi.
145
146 Introduction to Abstract Algebra
7. Find the multiplicative inverse of
(a)2—3x (b) —2+3x
(c)4+3x (d)7+x
8. Perform the following divisions. (Him: (a + bx) + (c + dx) = (a + bx)
[1/(6 + (195)]-
(a)(2 +3x)—:-(2—3x) (b) (—1 +x)-:—(2+5x)
(c)(—l —x)+(3—Sx) (d) (—4 +3x)+(—2-5x).
Exercises 10.2
1. Prove Theorem 2.
2. Prove that the mapping a defined by (a + bi)o = a — bi is an auto-
morphism of C.
COMPLEX NUMBERS AND QUATERNIONS 147
4. De Moivre’s Theorem
While it is not easy to show that the complex numbers form an algebraically
closed field, it is easy to show that the equation ax" + b = 0 (a, b e C, a aé 0,
n e N) always has a complex root. We will, in fact, describe a method for
finding all of the roots of such equations. To do this we rewrite the equation
ax" + b = 0 as x" = —b/a and turn our attention to finding the nth roots
of any complex number.
148 Introduction to Abstract Algebra
We begin by representing any complex number a + bi as a point with co-
ordinates (a, b) in the Cartesian plane as shown in the figure below. The hori-
zontal axis is called the real axis (R) and the vertical axis the imaginary axis
I
(a, b)
where {/— is the ordinary (real) nth root of r and 9 is the degree measure of 9.
Proof: By Theorem 4
_ 9 360 " _ _
[crs(—+k-——)] =c1s(9+k-360)=c159
n n
for k = 0, 1, . . . , n -— 1. Thus the complex numbers given by (1) are nth
roots of p. That they are all distinct follows from the fact that
9 360 . 5 _ 360
cis(—+1'-—) #us (~+1-————)
n n n n
foriaéj,i,j=0, 1,2, ,n— 1.
Solution. -—1 = cis 180°. Thus the three cube roots of —1 are given by
180° 360°
cis( +k- 3), k=0,1,2
Hence they are cis 60° = 1}; + i» 1/31', cis 180° = —1, and cis 300° = % —— i
fix:
Example 6. Find the two square roots of i.
150 Introduction to Abstract Algebra
Solution. i = cis 90°. Thus the two square roots of i are given by
90° 360°
cis( +k- 2), k=0,l
Hence they are cis 45° = fl/Z + (4/272)i and cis 225°: —1/2—/2 — (V272)i.
Example 7. Find the four fourth roots of 1 + 1‘.
Solution. 1 + i = fl cis 45°. Thus the four fourth roots of 1 + 1' are
given by
45° 360°
Vici5(T+k. 4 ), k=0,1,2,3
Hence they are ‘8/2- cis “.25", W cis 101.25°, 18/5 cis 191.25”, and
18/5 cis 28125". (It is possible, of course, to calculate cis 11.25°, cis 101.25°
etc. in terms of radicals by using the half-angle formulas on cos 45° and sin
45° to get cos 224E”, sin 22%O etc., but we prefer to leave the answer as is l)
The successful use of De Moivre’s theorem depends strongly on the exis-
tence of a simple polar form of the complex number involved. Thus, for
example, if we seek to find (—2 + 31')3 by the use of De Moivre’s theorem,
the best we can do is to write
—2 + 31' g 1/13 cis 123° 40’
or as
_ 3
—2 + 31' = 4/13 cxs [arctan (~- 5) + 180°].
Arm/1‘3 9./13
=13¢fi( ' =46 93'.
169 + 1691) +
The direct algebraic computation is definitely simpler in this case.
COMPLEX NUMBERS AND QUATERNIONS 151
Exercises 10.4
1. Plot the following complex numbers: 2 + 31', —1 — 2i, —3, 4i, (1 + i)/
(1 — 1), 2/1'.
2. Find the amplitude and absolute value of each of the following complex
numbers:
4 (cos 40° + isin 40°), —2 (cos 15° +isin15°),
3 (sin 15° + icos 15°), 3 (cos 40° —- isin 40°),
3 (—cos 30° + isin 30°)
Find the polar form of AB, A/B, Aa/B, and HA if A = 3 cis 60° and
B = 4 cis 40°.
. Prove that the absolute value of the product of two complex numbers is
the product of their absolute values, and that the amplitude of the product
is the sum of their amplitudes.
. What is the geometrical locus of the set of complex numbers of (a) fixed
absolute value, (b) of fixed amplitude?
WritethenumbersA =1 — Vii-LB = —1 — 1/31‘, andC == —1 + fii
in polar form. Find the amplitude and absolute value of A2 and of BalAC.
. Use De Moivre's theorem to calculate:
(a) (1 +13“ (M (1 - 1fii)‘
(C) (V? - 1')" (d) (1 - 0"
8. Use De Moivre's theorem and the binomial theorem to find expressions
for cos 30 and sin 39 in terms of cos 6‘ and sin 0
. Find the three cube roots of 8.
10. Find the three cube roots of 81'.
11. Find the four fourth roots of 1.
12. Find the two square roots of (I + i)/\/§. (Leave your answer in trigono-
metric form.)
13. Find the three cube roots of 1 — Vii. (Leave your answer in trigo-
nometric form.)
14. Prove that there exists a complex number R such that the n nth roots of 1
are given by R, R2 . . . , R".
15. A complex number 2 is called a primitive nth root of 1 if z” = 1 but
z"¢ lwhenO <m <n.
(a) Find the primitive sixth roots of 1.
*(b) Prove that a necessary and sufficient condition that R" (problem 14)
be a primitive nth root of 1 is that (k, n) = 1.
*16. Use De Moivre’s theorem to find
(a) (1 - 25)‘ (b) (3 + 21')a
without the use of trigonometric tables.
5. Quaternions
A system which has all of the properties of a field except that of com-
mutativity of multiplication is called a skew-field. The first example of a skew-
field was given by Hamilton in 1843, and it served as a stimulus to the con-
struction later of many other abstract systems. The system that Hamilton
developed is called the algebra of quaterniom over the field R*.
152 Introduction to Abstract Algebra
More generally, however, we may consider a quaterm'on algebra over any
field F as the set
a0+a1i+a2j+a3k=bo+b1i+b2j+b3k
if and only if a0 = b0, a1 = b], a2 = [12, and a3 = b3. The additive identity
for Q is 0 + 01' + Oj + 0k and the additive inverse of a0 + 4111‘ + a2j + ask
is -—a0 + (—a1)i+ (—a2)j+ (—a3)k = —a0 - ali — azj -— ask. The
scalar product a(a° + ali + (12j + ask) = aao + aali + aazj + aaak for
any a e F.
Multiplication of elements in Q is performed just as for polynomials
except that we replace i”, F, and k2 by —1 whenever the occur and if by k,
jiby —k. ik by «in = i2; = -i, kiby (W = i(ji) = —i(ij) = _.-2j =jetc.
(See multiplication table for the “units” below.)
Example. (1+i—k)(2+4j—3k)=2+4j—3k+2i+4ij—3ik
—2k—4kj+3k2=2+4j—3k+2i+4k+3j—2k+4i—3=—1
+ 61' + 71' — k.
The student should check that if at = a0 + ali + (121 + ask where
d = a?) + af + a; + a§ 9e 0, then the multiplicative inverse of a is
I
a‘1 = a6 — ali— a’zj— aék
where
a’1 = a,/d (1' = 0,1,2, 3)
Exercises 10.5
1. Definition
/
/
’-a= (-1)a
It is easy to verify that ordinary vectors in two or three dimensions satisfy
the requirements for a vector space in the sense of the following definition.
Definition 1. A vector (or linear) space V over a field F is a set of
elements, called vectors, such that a e V and fl 6 V define
a (unique) at + )3 e V and for any a e V and scalar
c e F there is defined a scalar product Cot e V such that
(1) V is an Abelion group under addition
(2) c(a + )9) = ta + :3, (c + c’)a = cot + c’a
154
VECTOR spaces 155
(3) (ed): = e(e’at), lat = at
for all e, e’ e F, a, fl 6 V, and l the multiplicative identity
of F.
To avoid confusion we will designate the identity for addition in F by 0
and the identity for addition in V by 6.
The student should check that the following are examples of vector spaces.
Example 1. For any fixed 1; EN, V,,(F) = {(41, . . . , an)|a‘ 61" for
i=1, ,n}where(a1, ,a,,) + (b1, ,bn) = (a1 + b1, ,a,,+ bu)
and e(a1, . . . , an) = (cal, . . . , can) is a vector space over F. (This space is
called the vector space of n-tuples.)
S={Zb,x2‘[meN,b,eR,i= l, ,m}
i=1
is a subspace of the vector space of all polynomials with rational coefficients.
Example 6. Any vector space is a subspace of itself.
Theorem I. Let V be a vector space over a field F. Then
(1)0a = 5 for all as V;
(2) L0 = afor alleeF;
(3) The additive inverse, — a, of an is equal to (—l)ot.
156 Introduction to Abstract Algebra
Proof. (1) We have (6 + 0)0l = cot = ta + 5 = w + 00!. By the cancel-
lation property for groups we have 6 = Oct.
(2) c(ot+6)=ca=ca+5=ca+c§sothat6=d1
(3) a +(—1)(1 =10: + (—1)a =[1+(——1)]a = 00: = 5 so that (—1)a.
= —¢.
Exercises 11.1
at = 2 cf,
i=1
vscron smcm 157
Example I. In V3(F) the vectors £1 = (1, 0, 0), f2 = (0, 1, 0), and
£3 = (0, 0, 1) span V3(F) since (:11, a2, a3) = 11151 + azfz + aafa. Also
51: E2, E3, and £4 = (1,1,1)span V3 (F) since (“hazy “3) = ”151 + “252 + “353
+ 054, but 51 and £2 alone do not span V3 (F).
2 m, = 0
i=1
Example 3. In Example 1 the vectors {1, £2, and {3 are clearly linearly
independent since £151 + £252 + c353 = (cl, :2, £3) = (0, 0, 0) implies c1
= 2 = as = 0. But £1, £2, £3, and {4 are linearly dependent since 15 + 152
+1E3 + (_1)£4 = (0! 0r 0)'
Exercises 11.2
1. Prove that the vectors (a;, a2) and (b1, by) in V2(F) are linearly independent
if and only if 0162 — dab; sé 0.
2. In V3(R), examine each of the following sets of vectors for linear dependence.
(3) (1,3,2)and(1,4,1) (b) (6.4. 12) and (3.2. 6)
(0) (1, 2. 3) and (1. 4, 9) (d) (1, 2, 3). (1, 0.1), and (0. 1. 0)
(e) (1. 2. 0), (0, 3,1), and (-1, 0,1) (f) (—1, 2,1).(3. 0. -1), and (-5.4, 3)
3. Prove that a and fl sé 0 are linearly dependent in a vector space V over F
if and only if a = of! for c e F.
4. Do the results of Example 5 hold over any field F?
5. How many elements are there in each subspace spanned by three linearly
independent elements of V503). Generalize your result.
7 We shall show later that every vector space spanned by a finite number of vectors
is isomorphic to a space of n-tuples.
VECTOR smces 159
6. If clot + cgfi + £37 = 6 where clcs 9!: 0, show that a and B generate the
same subspace as do [3 and y.
7. Exhibit three different bases for V3(R).
Our next sequence of theorems will lead us to the concept of the dimension
of a vector space spanned by a finite number of elements.
Theorem 2. If {51, . . . , {a} is a basis of V over F, then so are
E = 25451,
i=1
then
6
+ tug"
f = CIEI + -.- + ‘1-1‘51-1 + 7301:.) + ‘nt-u + . ..
and hence £1, . . . , §,_1,k§i, gm, . . . , é" span V. Thus {51, . . . , {1—1
k5,, 5H1, . . . , {n} is a basis of V.
(2) Ifclgl + .. - + ‘i—1EI—1 + ‘1(§I_ k5) + 51+151+1 + - - - + tugn = 0,
we have 5151 + + ”(—154 + 0151 + 51+1§1+1 + + 51—15-1 + (‘1
— kc‘)§, + 9,15,” + . . . + an = 0 and, hence, since £1, . . . , in are
linearly independent we have cl = . . . = c, . . . = c,_1 = c, — kc, = a,”
= = in = 0. Thusc, = 0alsoand§1, ,£4_1,E; — Mpg“, ,5
are linearly independent.
Next, if
g = 25151:
i=1
then£= £15 + + c‘(§, — k5,) + + (c, + c,k)§, + + anand
hence £1, .. . , f, — k§,, . . . , En span V. Thus {$1, . . . , E,_1,§‘ — kg},
£4”, . . . , §,,} is a basis of V.
160 Introduction to Abstract Algebra
Example. If {£1) £2) £3) £4} l5 2. baSis 0f V4(R)! then 50 are {£11 £2, 3:3»
£4} and {$1. E2 - 2a £3, {4}-
Theorem 3. A set of nonzero vectors, {51, ,gm}, of V is linearly
dependent if and only if some one of the 5k is a linear combination of the
preceding vectors £1, . . . , §k_1.
2 (1,5, = 6
i=1
with dk 9E 0, (1“1 = . . . = dm = O (i.e. dk is the last nonzero d). Then if
k aé 1, we solve for £1: and are done. But if k = l, dlfl = U, d1 9e 0 and
£1 = 6, a contradiction.
Theorem 4. A set S = {51, . . . , §m} of vectors of V is linearly dependent
if and only if it contains a proper subset R such that S and R generate the
same subspace of V.
Proofl (1) If S is a linearly dependent set we form R by deleting from
S all zero vectors and those which are (by Theorem 3) linear combinations
of the vectors preceding it in the list. By Theorem 3 at least one vector will
be deleted so that R C S.
(2) Suppose now that S contains a proper subset, R, which generates the
same subspace of V as S. By renumbering our vectors if necessary we may
assume R = {61, . . . , 5‘} where k < m. Since S and R generate the same
subspace, {“1 = 5151 + + ckék and hence clgl + . . . + ckfk +
(—1)£,‘,+1 + 05¢” + . . . + 05,. = 6 so that {51, . . . , §,,I} is a dependent
set of vectors.
Theorem 5. If «1, . . . , at" span V and £1, . . . , g, are linearly independent
vectors of V, then 11 Z 7'.
Proof: For A and B span V and are linearly independent; hence we may
apply Theorem 5 to get n 2 m and m 2 n so that m = n.
Definition 8. Two vector spaces V and V’ over the same field F are said
to be isomorphic if there is a 1—1 correspondence between
the elements of V and V’ that is preserved under addition
and scalar multiplication. That is, if a e V<—> ot’ e V’ and
BeVHfl’eV’, then ot+fleV<—>oc’+f3’eV’ and
can 6 V<—> cat’ 6 V’for every c eF.
Theorem 8. A vector space V of dimension n over F is isomorphic to
Vn(F)'
Proof: Let {11, . . . , an} be a basis for V and {$1, . . . , fin} be a basis for
Vn(F). Then if
n
3: Z‘IWGV
=1
162 Introduction to Abstract Algebra
we let
Exercises 11.3
1. Find a basis for V;(R) which includes the vectors (1, — 1, 1) and (2, l, 1).
2. Prove that no basis of V3(R) can include both of the vectors (1, —1, 1) and
(2, —2, 2).
3. Prove that the vectors (1, 2, 3, 4), (2, 3, l, 5), (3, 1, 2, 7), and (l, 1, -2, 1)
in V.(R) are linearly dependent. Which of these four vectors is not in the
subspace spanned by the other three?
4. Prove that for n vectors 6:, . . . , f. of an n-dimensional vector space V to be
a basis, it is sufficient that they span V or that they be linearly independent.
5. If two subspaces S and Tof a vector space V over F have the same dimension,
prove that S E T implies S = T.
l. The dot, scalar, or inner product of two vectors § = (x1, .. . ,x,.) and
fl
1. Definitions
aibk
1+ Ic=t—1
Finally, 1] a = (a0, a1, . . . , a”, 0, 0, . . .) and ceD,
we define the value, etc, of at at c to be ac = no + alc +
. . . + ant" e D.
Proof: We prove only the first part of the theorem leaving the other two
parts as exercises for the student. We have, for in Z n,
Exercises 12.1
It is easy to show that, for any integral domain D, D[x] = {a0 + alx + . . .
+ anx"|n e N, a, e D, i = 1, . . . , n} is also an integral domain whose units
are the units of D. Thus two polynomials f(x) and g(x) of D[x] are associates
if and only iff(x) = ug(x) where u is a unit of D. (See Section 5.3.)
In the remaining part of this chapter we will consider only the case when
D is a field although some of the results we will state will be true when D
is simply an integral domain. (Recall, in this connection, that every non-
zero element of a field is a unit of the field so that any nonzero multiple of a
polynomial over a field is an associate of the polynomial.)
168 Introduction to Abstract Algebra
Theorem 3. (the division algorithm) Let F be a field. If f(x), g(x) e F[x]
and g(x) aé 0. then there exist unique q(x), r(x) e F[x] such that
f(x) = o - :00 + no
so that our result holds if degf(x) = n = 0. We proceed with an inductive
proof by assuming that the theorem is true when degf(x) < n and showing
that then it is also true when deg f(x) = n. (See Theorem 5.11.1 To this
end take it 2 m and n > 0 and let
f0?) = (x - 6M") + 7
= (’ " 050) + (be " ‘51)” + (1’1 — 51’2l +
+ (bu-2 — ‘bn—1)xn_1 + bin—1""
When we equate coeflicients in the two representations of f(x) we have
a,I = bu-” an_1 = bn—2 — ebn_1, . . . , a1 = b0 — obi, a0 = r - ebo
170 Introduction 10 Abstract Algebm
For computational purposes this work may be arranged as follows:
an ah1 ... a1 do If
cbn_1 ... cb1 cbo
4 —8 13 —27 56
Exercises 12.2
Proofi The proof follows the same pattern as that of the corresponding
theorem for the integers (Section 5.6). We give only an outline of the proof
here, leaving the details to be filled in by the student.
By the division algorithm we obtain the sequence of equalities
Theorem 5. Let d(x) be the g.c.d. of two polynomials f(x) and g(x) e F[3:].
Then there exist polynomials m(x) and n(x) e F[x] such that
10x+10=(x3—3x—2)—(x—4)(x2+4x+3)
—30x—30= —3(x3—3x—2)—(x—4)(—3x2— 12x—9)
—30x—30=—3(x3—3x—2)—(x—4)[(x5—x‘—6x3—2x2—x—3)
—(x2—x—3)(x3—3x—2)]
=[(x—4~)(x2—x—3)—3](x3—3x—2)
+(4—x)(x5—x4—6x3—2x3—x—3)
=(x3—5x2+x+9)(x3—3x—2)
+(4—x)(x5—x4—6x3—2x2—x—3)
POLYNOMIALS 173
and
x+l=(—-.‘sx3+%x2—-.+.x+fir)(x3—3x—2)
+(.—F.-x--,'—.)(x5—x‘-6x3—2x2—x—3)
(Notice how we again avoided the use of fractions until the last step.)
Corresponding to the notion of a prime integer we have the notion of a
prime or irreducible polynomial.
Exercises 12.3
1. Find (f(x), g(x)) over the indicated field, F, of coefficients and express it in
the form m(x)g(x) + n(x)f(x) for m(x), n(x) e F[x]:
(a) f(x) =x5 —x‘ —6Jt:a ~-2x2 +Sx +3,g(:c)=.1c3 —3x—2;F=R
(b)f(x)=x‘—5x° +6x‘ +4x—8,g(x) =x“—x'—4x +4;F=R
(c) f(x) =x‘ —4ix +3,g(x) =3:3 -i;F = C
2. In R[x] find (f(x), g(x)) when
(a)f(x) =x‘ --x3 —3x2+x +2,g(x) =x3 —x' —5x +2
(b)f(x) =x‘+x3+2x’+x +1,g(x) =x" —1
3. In 1219:] find (f(x), g(x)) when
(a) f(x) = x5 + 2x3 + x’ + 2x, g(x) = x‘ + ac“ + x2
(b) f(x) = x3 + 2x2 + 2x +1,g(x) = x2 + 2
4. In R[x] determine 0 so that (f(x), g(x)) is linear and then find (f(x), g(x)) if
(a)f(x) =x“ +cx’ - x +2c,g(x) =x’ +cx —-2
(b)f(x) =ac2 +(c —6)x +21: —1,g(x) =x’ +(c +2)x +2:
. Prove Theorem 5.
“Q60!
. Prove Theorem 6.
. Prove the corollary to Theorem 6.
. Prove Theorem 7.
4. Zeros of a Polynomial
Theorem 9. If f(x) eF[x] and degf(x) = n > 1, then f(x) has at most 11
zeros in F.
Proof. If n = 1, f(x) = a0 + alas (a1 yé 0) and f(x) has the one zero,
-a0/a]. Thus the theorem is true for n = 1. Now suppose degf(x) > 1.
If f(x) has a zero, r1, the factor theorem gives us f(x) = (x — r1)q(x) where
q(x) is of degree n — 1. Clearly, every zero of q(x) is a zero of f(x). On the
other hand,f(x) has no other zeros than r1 and the zeros of q(x). For iff(a) = 0,
then (a — rl)ql(a) = 0. Hence a = r1 or q1(a) = 0 since a field has no
divisors of zero.
If we now make the induction hypothesis that the theorem is true for poly-
nomials of degree less than n, we see that q(x) has at most 71 — 1 zeros in F
so that, by the above argument, flat) has at most 1 + (n — 1) = n zeros in F.
The proof of the following theorem closely parallels the proof of Theorem 8
and is left as an exercise for the student.
Theorem 12. If f(x) 6 R*[x] has the zero a + bi (a, b eR*, b 9!: 0), then
it also has the conjugate zero a — bi.
Theorem 14. Let c/d where (c, d) = 1 be a rational zero off(x) = a0 + (11::
+ . . . + aux" e I[x]. Then clao and dlan.
3 5 5 2
9 15 15 6 0
3 5 5 2 - 2/3
1+fli)
2
f(x) = 9(x — 1/3)(x — 2/3)(x _ ’1_J;V_§‘) (x +
We turn now to a brief consideration of the problem of finding zeros of
polynomials over finite fields, 1,. Here, in contrast to the situation for in-
finite fields, we may actually examine all of the possibilities for zeros. For
example, in J2 the only possible zeros for any polynomial are 0 and 1. In
particular then, x2 + x + 1 is irreducible in 12 since 02 + 0 + 1 aé 0 and
12 + l + 1 = 1 96 0.
We may exploit this situation in seeking zeros or factors of polynomials
in I[x] or f] since the equality of two integers in I, a = b, certainly implies
their equality in 1,, a a b (mod p) (although, of course, not conversely).
Thus, forexample,2 + 3 =5,2 x 3 = 6andac2 + 5x + 6 = (x + 2)(x + 3)
are identities modulo 7;: for any positive integer m.
For example, suppose we wish to know whether or not 27x2 + 159x — 21557
is irreducible over R. In 12 this polynomial is equal to x2 + x + 1 since
27 E l, 159 E 1, —21557 E 1 (mod 2). But we have already observed
that x2 + x + 1 has no zeros in 12. Hence at2 + x + 1 is irreducible over
12 and thus also irreducible over R.
178 Introduction to Abstract Algebra
As a more complicated example let us ask whether or not 27x5 + 40x3
— 16x3 — 19::2 + 100:: + 45 can be factored over R. In J2 this poly-
nomial is equal to x5 + x2 + 1. Now obviously, if x5 + ac2 + 1 factors, one
of the factors must be linear or quadratic since a cubic factor will produce
another factor that is quadratic, and a quartic factor will produce another
factor that is linear. As before we observe that the only possible linear
factorsinJ2 arexandx +1. Butx5 + x3 +1 = .7c(:c4 + x)+1=(x +1)
(36‘ + x3 + x3) + l in 12 so that x5 + x2 + 1 is not divisible by x or x + 1 in
13. Furthermore, the only possible quadratic factors in 12 are x2, x2 + x
= x(x+ I), .9:2 +120: +1)(x +1), and x2 + a: +1. We do not need
to consider the first three possibilities, however, since, for example, if
x5 + x2 + l is divisible by x‘2 in 12 it is certainly divisible by x in J:—
contrary to what we have just shown. Finally, since :55 + x2 + l = (ac2
+ x + 1)(ac3 + x2) + 1 in 12, it follows that x5 + .:c2 + 1 is not divisible
by x2 + x + l in 12. (In fact, if we observe that x5 + x2 + 1 = x2(x + 1)
(x2 + x + 1) + 1 in 12 we may eliminate all possible factors immediately.)
Thus 3:5 + x2 + 1 cannot be tactored over 12 and hence 27x5 + 40.7:4
— 16xa — 19.1:2 + 100:: + 45 cannot be factored over R.
While irreducibility over J, implies irreducibility over the rational numbers
the converse is not true. Thus 3:” + 1 is irreducible over the rational numbers
but reducible over Ja since x2 + 1 = (x + 1)(x + 1) in 12. Similarly,
f(x) = 25964 + 153::3 + 21.7:2 — 65 is equal to a." + x3 + x2 + 1 = (x + 1)
(x3 + x + l) in 12. Hencef(x) is reducible over 12 but we still do not know
if it is reducible or irreducible over R. However, we may conclude that if
f(x) is reducible over R, f(x) has a linear factor. For if f(x) = g(x)h(x) over
R wheref(x) and g(x) are both quadratic, f(x) will also factor into two quad-
ratic factors over 1,. But since f(x) = (x + 1)(:c3 + x + l) in 12, f(x) can
factor into quadratic factors in J2 if and only if x3 + x + 1 is reducible in
12. Butx3+x+l=x(:¢2+l)+1=x(x+1)(x+1)+linsothat
x3 + x + 1 is irreducible in 12 since it has a remainder of 1 when divided by
x or x + l (the only two linear polynomials in 12). Having established that,
iff(x) has a factor, it must have a linear factor, we go to 13 and observe that,
in J3, the only possible linear factors are x, x + 1, and x + 2 (= x — 1).
Butf(x)=x4—2=x4+ l inla andhencef(x)=x-x3+1=(x+1)
(xs—x2+x—1)+2=(x—-1)(x3+x3+x+l)+2in13. Thusf(x)
is not divisible by x, x + 1, or x —- 1 and is irreducible over 13. Hence f(x)
is irreducible over R.
The last example serves to bring out another point. Our analysis in 12
showed that iff(x) had any factors at all it had to have a linear factor. Hence
even if our results in 13 had not been conclusive we would still have known
that if we tried all possible linear factors of f(x) over R, (namely, 25:: i 1,
25x 1 13, 51: i 1, 5x 1 13, x i 65, x i 5, x i 13) and none of them were
POLYNOMIALS 179
actually factors, there would be no point in looking for a factorization of
flat) into irreducible quadratic factors over R. On the other hand, a test
over [2 easily shows that g(x) = x‘ + 2x3 + 3x2 + 2x + 1 has no linear
factors over J2 and hence none over R. Thus our search for factors of g(x)
over the rational numbers may be confined to a search for quadratic factors
and we note that, actually, g(x) = (ac2 + x + I)”.
Exercises 12.4
(1) 4x3+bx2+cx+d=0(a,b,c,deC,aaé0)
Then (2) is an equation whose roots differ from those of (1) by 12/31:.
Now we make the substitution y = z — p/Sz which, as the student should
show, gives us
8 —8
x1 = 2.1 — 2/21 = {/3 — 2/‘a/T4 = {/4 + = {/4 + {/2
{/4
x2 = 32 — 2/33 = We) — 2/{/—-—_4w
—8w3
= —4w + = flu) + {/27422
—4w
(since we = 1) and
We leave it to the student to show that the other solutions for 2 yield no
new values for x.
For the quartic equation
(6) ux4+bx3+cx2+dx+e=0(a,b,c,d,eeC,a;é0)
we again divide through by a and obtain
(7) x4+px3+qx2+rx+s=0.
Our procedure now is to seek to determine numbers at, [3, and 7 such that
(8) x‘+px3+qx2+rx+s+(ax+fi)2=[x2+(1>I2)x+y]*
for then, by (7), we will have
(10) a” + q = 2y + pal-t.
(11) Zufi + r = yp,
and
so that the four roots of (6) can be found by solving the quadratic equations
(14) and (15).
Solution. The equations (10), (11), (12), and (13) become, respectively,
(a)aL2—12=2'y+1,
(b) 2013 + 10 = —27,
(c) #2 + 3 = ,2,
and
(d)y3+6y2—8y—32=0
Our resolvent cubic, (d), has (fortunately!) the rational root —-2. This
gives us, by (a) and (b), at = 3 and B = —1. Then (9) becomes (3x - 1)2
=(x2—x—2)zsothatx2—x—2=3x— lorxz—x—2= —(3x—1).
Solving these two quadratic equations we obtain as our solution set for the
given equation, {1, —3, 2 + x/g, 2 - V3}.
It is obvious that, except for very special values of the coefficients, the
solution of cubic and quartic equations by the methods just presented is
very difficult. In practice, when the roots cannot be found by simple methods
(such as rational roots by the use of Theorem 14 or by De Moivre’s theorem
in the case of equations of the form x" —— a = 0), we are interested mainly
in finding the roots approximately to a prescribed degree of accuracy. Many
such methods of approximate solution for both real and complex, nonreal,
roots can be found in texts on the theory of equations and numerical methods
of mathematics.
From both the theoretical and historical point of view, however, the
existence of a solution “by radicals” for the linear, quadratic, cubic, and
quartic equations is of considerable interest. After the solution of the quartic
was obtained it was naturally conjectured that a similar solution existed for
the quintic and, even more generally, for polynomial equations of any
POLYNOMIALS 183
degree. It was one of the great achievements of nineteenth-century mathe-
matics to discover that no such solution exists for polynomial equations with
coeflicients complex numbers and of degree _>_ 5. Many of the concepts
concerning groups and fields that we have considered here were developed
in connection with work on this problem.
It is beyond the scope of this book to give the details concerning this prob-
lem here but we would like at least to state the problem and the conclusions
precisely. First let us define what we mean by “a solution by radicals.”
Definition 9. The polynomial equation f(x) = anx" + an_1x”'1 + . . .
+a0=0 (a,eCfori=0, 1, ,n; anyéO) issaid
to be solvable by radicals in C if the roots off(x) = 0 can
be expressed in terms of the coefi‘icients of a0, a], . . . , an
using a finite number of additions, subtractions, multiplica-
tions, divisions, and extraction of roots.
We have seen that polynomial equations of degree S 4 are solvable by
radicals; the fundamental result of Galois (1811—1832) was that equations
of degree 2 5 are not solvable by radicals. It is important to note the following
points:
(1) We know by Theorem 10.3 that a polynomial equation of any degree
with coefficients in C has a solution in C; the result of Galois simply says that
the solution cannot, in general, be obtained in a certain form if the degree of
the equation is Z 5.
(2) Many special equations of higher degree are certainly solvable by radi-
cals (e.g. x5 — 1 = 0); what the result of Galois asserts is that there exists
no solution in terms of radicals for the general polynomial equation of degree
2 5.
(3) There are standard methods for finding approximate solutions to any
polynomial equation; the result of Galois has reference to exact solutions of
a certain form.
(4) We are restricting ourselves to a finite number of arithmetic operations.
Thus, for example, the use of infinite series is not allowed.
Exercises 12.5
1. Solve the following cubic equatiOns by the method outlined in this section.
(a)x3—18x+35=0 (b)x°+6x+2=0
(c) x’ + 6.1::8 + 3x + 18 = 0 (d) 283:” + 9x8 — 1 = 0. (Hint: First
let x = 1/y.)
2. Solve the following quartic equations by the method outlined in this section.
(a) x‘ +2x3—12x’ —10x +3 =0 (b) x4—8x3 +9:x:a +8x-10=0
(c)x‘—3x’+6x—2=0 (d)x‘—10x’—20x—16=0
3. Verify the results given in equations (2) and (3).
4. Verify that (10), (11), and (12) are the conditions for (8) to hold and that
(13) follows from (10), (11), and (12).
184 Introduction to Abxtmct Algebra
PROBLEMS FOR FURTHER INVESTIGATION
B1 Barnett, R. A. and Fujii, J. N. Vectors. New York: John Wiley and Sons, 1963.
132 Beaumont, R. A. “Equivalent Properties of a Ring." American Mathematical
Monthly, 57 (1950): 183.
B3 Beaumont, R. A. and Pierce, R. S. The Algebraic Foundations of Mathematics.
Reading, Mass.: Addison—Wesley Publishing Co., 1963.
B4 Birkhoff, G. and MacLane, S. A Survey of Modern Algebra (revised edition).
New York: The Macmillan Co., 1953.
BS Brand, L. “The Roots of a Quaternion." American Mathematical Monthly.
49 (1942): 519—520.
01 Ore, 0. Number Theory and It: History. New York: McGraw—Hill Book Co.
1948.
Pl Pollard, H. The Theory of Algebraic Number: (Carus Monograph No. 9). New
York: John Wiley and Sons, 1950.
V1 Van det Waerden, B. L. Modern Algebra, Vol. 1 (second edition). New York:
F. Ungar Publishing Co., 1953.
W Weiss, M. J. Higher Algebra for the Undergraduate (second edition). New York:
h—h