A First Course in

RINGS AND IDEALS

David Burton
CONTENTS

Chapter 1 Introductory Concepts .

Chapter 2 Ideals and Their Operations 16
Chapter 3 The Classical Isomorphism Theorems. 39
Chapter 4 Integral Domains and Fields 52
Chapter 5 Maximal, Prime, and Primary Ideals 71
Chapter 6 Divisibility Theory in Integral Domains 90
Chapter 7 Polynomial Rings 112
Chapter 8 Certain Radicals of a Ring . 157

Appendix A. Relations . 287

Appendix B. Zom'sLemma 296
Bibliography 300
Index of Special Symbols 303
Index. 305

Vll
CONVENTIONS

Here we shall set forth certain conventions in notation and terminology

used throughout the text: the standard symbols of set theory will be
employed, namely, E, U, n, -, and 0 for the empty set. In particular,
A - B = {XIXEA and x¢B}. As regards inclusion, the symbols £; and
:2 mean ordinary inclusion between sets (they do not exclude the possibility
of equality), whereas c and ::::> indicate proper inclusion. When we deal
with an indexed collection of sets, say {Aili E I}, the cumbersome notations
U {AiliEI} and n {AiliEI} will generally be abbreviated to U Ai and
n Ai; it being understood that the operations are always over the full
domain on which the index is defined. Following custom, {a} denotes the
set whose only member is a. Provided that there is no risk of confusion, a
one-element set will be identified with the element itself.
A function f (synonymous with mapping) is indicated by a straight
arrow going from domain to range, as in the case f: X ~ Y, and the notation
always signifies thatfhas domain X. Under these circumstances,fis said
to be a function on X, or from X, into Y. In representing functional values,
we adopt the convention of writing the function on the left, so that f(x), or
occasionally fx, denotes the image of an element x E X. The restriction of
f to a subset A of X is the function flA from A into Y defined by
(fIA)(x) = f(x) for all x in A. For the composition of two functions
f: X ~ Yand g: Y ~ Z, we will write g f; that is, g f: X ~ Z satisfies
0 0

(g f)(x) = g(J(x)) for each x E X. (It is important to bear in mind that

0

our policy is to apply the functions from right to left.)

Some knowledge of elementary number theory is assumed. We simply
remark that the term "prime number" is taken to mean a positive prime;
in other words, an integer n > 1 whose only divisors are ± 1 and ± n.
Finally, let us reserve the symbol Z for the set of all integers, Z + for the
set of positive integers, Q for the set of rational numbers, and R # for the
set of real numbers.
viii
ONE

INTRODUCTORY CONCEPTS

The present chapter sets the stage for much that follows, by reviewing some
of the basic elements of ring theory. It also serves as an appropriate vehicle
for codifying certain notation and technical vocabulary used throughout
the text. With an eye to the beginning student (as well as to minimize a
sense of vagueness), we have also included a number of pertinent examples
of rings. The mathematically mature reader who finds the pace somewhat
tedious may prefer to bypass this section, referring to it for terminology
when necessary.
As a starting point, it would seem appropriate formally to define the
principal object of interest in this book, the notion of a ring.
Definition 1-1. A ring is an ordered triple (R, +,.) consisting of a
non empty set R and two binary operations + and . defined on R such
that
1) (R, +) is a commutative group,
2) (R,·) is a semigroup, and
3) the operation· is distributive (on both sides) over the operation +.
The reader should understand clearly that + and· represent abstract,
unspecified, operations and not ordinary addition and multiplication. For
convenience, however, one invariably refers to the operation + as addition
and to the operation· as multiplication. In the light of this terminology, it
is natural. then to speak of the commutative group (R, +) as the additive
group of the ring and of (R, .) as the multiplicative semigroup of the ring.
By analogy with the integers, the unique identity element for addition
is called the zero element of the ring and is denoted by the usual symbol O.
The unique additive inverse of an element a E R will hereafter be written
as - a. (See Problem 1 for justification of the adjective "unique".)
In order to minimize the use of parentheses in expressions involving
both operations, we shall stipulate that multiplication is to be performed
before addition. Accordingly, the expression a·b + c stands for (a·b) + c
and not for a·(b + c). Because of the general associative law, parentheses
2

can also be omitted when writing out sums and products of more than two
elements_
With these remarks in mind, we can now give a more elaborate definition
of a ring_ A ring (R, +, -) consists of a nonempty set R together with two
binary operations + and - of addition and multiplication on R for which
the following conditions are satisfied:
1) a + b = b + a,
2) (a + b) + c = a + (b + c),
3)
4)
° °
there exists an element in R such that a + = a for every a E R,
for each a E R, there exists an element - a E R such that a + (- a) = 0,
5) (a-b)-c = a-(b-c), and
6) a:(b + c) = a-b + a-c and (b + c)-a = boa + c-a,
where it is understood that a, b, c represent arbitrary elements of R.
A ring (R, +, -) is said to be a finite ring if, naturally enough, the set R
of its elements is a finite set. By placing restrictions on the multiplication
operation, several other specialized types of rings are obtained_
Definition 1-2. 1) A commutative ring is a ring (R, +, -) in which
multiplication is a commutative operation, a- b = b -a for all a, b E R.
(In case a-b = boa for a particular pair a, b, we express this fact by saying
that a and b commute_)
2) A ring with identity is a ring (R, +, -) in which there exists an
identity element for the operation of multiplication, normally represented
by the symbol 1, so that a-I = I-a = a for all a E R.
Given a ring (R, +, -) with identity 1, an element a E R is said to be
invertible, or to be a unit, whenever a possesses a (two-sided) inverse with
respect to multiplication_ The multiplicative inverse of a is unique, when-
ever it exists, and will be denoted by a- l , so that a-a- l = a-loa = 1_ In the
future, the set of all invertible elements of the ring will be designated by the
symbol R*_ It follows easily that the system (R*, -) forms a group, known
as the group of invertible elements_ In this connection, notice that R* is
certainly nonempty, for, if nothing else, 1 and -1 belong to R*_ (One must
not assume, however, that 1 and -1 are necessarily distinct.)
A consideration of several concrete examples will serve to bring these
ideas into focus_
Example 1-1. If Z, Q, R# denote the sets of integers, rational, and real
numbers, respectively, then the systems
(Z, +,-), (Q, +,-), (R#, +,-)
are all examples of rings (here, + and - are taken to be ordinary addition
and multiplication)_ In each of these cases, the ring is commutative and
has the integer 1 for an identity element.
INTRODUCTORY CONCEPTS 3

Example 1-2 Let X be a given set and P(X) be the collection of all subsets
of X. The symmetric difference of two subsets A, B s; X is the set A Ll B,
where
A Ll B = (A - B) u (B - A).
If we define addition and multiplication in P(X) by
A + B = All B, A·B = A II B,
then the system (P(X), +, .) forms a commutative ring with identity. The
empty set 0 serves as the zero element, whereas the multiplicative identity
is X. Furthermore, each set in P(X) is its own additive inverse. It is
interesting to note that if X is nonempty, then neither (P(X), u, II) nor
(P(X), II, u) constitutes a ring.
Example 1-3. Given a ring (R, +, .), we may consider the set Mn(R) of
n x n matrices over R. If In = {1,2, ... , n}, a typical member of Mn(R)
is a function f: In X In --'f R. In practice, one identifies such a function
with its values aij = f(i, j), which are displayed as the n x n rectangular
array

! ~. 11 •••

a n1 ••• ann
~. 1n ) (a ij E R).

For the sake of simplicity, let us hereafter abbreviate the n x n matrix

whose (i, j) entry is aij to (a i ).
The operations required to make (Mn(R), +, .) a ring are provided by
the familiar formulas
and
where
n
cij = I aik • bkj •
k=1

(We shall often indulge in this harmless duplication of symbols whereby

+ and· are used with two different meanings.) The zero element of the
resulting ring is the n x n matrix all of whose entries are 0; and
-(aij) = (-a i ) . The ring (Mn(R), +, .) fails to be commutative for n > 1.
It is equally easy to show that if (R, +, .) has an identity element 1,
then the matrix with 1's down the main diagonal (that is, aii = 1) and O's
elsewhere will act as identity for matrix multiplication. In terms of the
Kronecker delta symbol J ij , which is defined by

Jij =
f if i
1ifi =j
LO =1= j (i, j = 1, 2, ... , n),
the identity matrix can be written concisely as (Jij).
4

Example 1-4. To develop our next example, let X be an arbitrary (non-

empty) set and (R, +,.) be a ring. We adopt the notation map(X, R) for
the set consisting of all mappings from X into R; in symbols,
map(X, R) = {tlf: X --+ R}.
(For ease of notation, let us also agree to write map R in place ofmap(R, R).)
Now, the elements of map (X, R) can be combined by performing algebraic
operations on their functional values. More specifically, the pointwise sum
and product of f and g, denoted by f + g and 1" g, respectively, are the
functions which satisfy
(f + g)(x) = f(x) + g(x), (f·g)(x) = f(x)·g(x), (x EX).
It is readily verified that the above definitions provide map (X, R) with
the structure of a ring. We simply point out that the zero element of this
ring is the constant function whose sole value is 0, and the additive inverse
- f of f is characterized by the rule ( - f)(x) = - f(x).
Notice that the algebraic properties of map(X, R) are determined by
what happens in the ring (R, +, .) (the set X furnishes only the points for
the pointwise operations). For instance, if (R, +, .) has a multiplicative
identity 1, then the ring (map(X, R), +, .) likewise possesses an identity
element; namely, the constant function defined by 1(x) = 1 for all x E X.
Example 1-5. Our final example is that of the ring of integers modulo n,
where n is a fixed positive integer. In order to describe this system, we
first introduce the notion of congruence: two integers a and b are said to
be congruent modulo n, written a == b (mod n), if and only if the difference
a - b is divisible by n; in other words, a == b (mod n) if and only if
a - b = kn for some k E Z. We leave the reader to convince himself that
the relation "congruent modulo n" defines an equivalence relation on the
set Z of integers. As such, it partitions Z into disjoint classes of congruent
elements, called congruence classes. For each integer a, let the congruence
class to which a belongs be denoted by [a] :
[a] = {x E Zlx == a (mod n)}
= {a + knlkEZ}.

Of course, the same congruence class may very well arise from another
integer; any integer at for which [at] = [a] is said to be a representative
of [a J. One final, purely notational, remark: the collection of all congruence
classes of integers modulo n will be designated by Zn.
It can be shown that the congruence classes [0], [1], ... , [n - 1]
exhaust the elements of Zn. Given an arbitrary integer a, the division
algorithm asserts that there exist unique q, r E Z, with 0 ::::;; r < n, such
that a = qn + r. By the definition of congruence, a == r (mod n), or
INTRODUCTOR Y CONCEPTS 5

equivalently, [a] = [r]. Thus, there are at most n different congruence

classes in Zn' namely, [0], [1], ... , [n - 1]. But these n classes are them-
selves distinct. For if 0 ::;; b < a < n, then 0 < a - b < n and so a - b
cannot be divisible by n, whence [a] =1= [b]. Accordingly, Zn consists of
exactly n elements:
Zn = {[OJ, [1], ... , [n - 1]}.
The reader should keep in mind that each congruence class listed above is
determined by anyone of its members; all we have done is to represent
the class by its smallest nonnegative representative.
Our next step is to define the manner in which the members of Zn are
to be added and multiplied, so that the resulting system will form a ring.
The definitions are as follows: for each [a], [b] E Zn'
[a] +n [b] = [a + b], [aL[b] = [ab].
In other words, the sum and product of two congruence classes [a] and [b]
are the unique members of Zn which contain the ordinary sum a + band
ordinary product ab, respectively. Before considering the algebraic properties
of these operations, it is necessary to make certain that they are well-defined
and do not depend upon which representatives of the congruence classes
are chosen. In regard to multiplication, for instance, we want to satisfy
ourselves that if [a'] = [a] and [b'] = [b], then [aiL [b'] = [aL [b], or,
rather, that [a'b '] = [abJ. Now, a' E [a'] = [a] and b' E [b'] = [b],
which signifies that a' = a + kn and b' = b + jn for some k, j E Z. But
then
a'b = (a + kn)(b + jn) = ab + (aj + bk + kjn)n.
I

Hence, a'b' == ab (mod n) and so [a'b '] = [ab], as desired. The proof that
addition is unambiguously defined proceeds similarly.
We omit the detailed verification of the fact that (Zn' +n' ·n) is a com-
mutative ring with identity (traditionally known as the ring of integers
modulo n), remarking only that the various ring axioms hold in Zn simply
because they hold in Z. The distributive law, for instance, follows in Zn
from its validity in Z:
[aL([b] +n [e]) = [aL[b + e] = [a(b + e)]
= [ab + ae] = [ab] +n [ae]
= [aL[b] +n [aln[e].

Notice, too, that the congruence classes [0] and [1] serve as the zero element
and multiplicative identity, respectively, whereas [- a] is the additive
inverse of [a] in Zn. When no confusion is likely, we shall leave off the
brackets from the elements of Zn' thereby making no genuine distinction
6

between a congruence class and its smallest nonnegative representative;

under this convention, Zn = {a, 1, ... , n - 1}. It is perhaps worth com-
menting that, since Z 1 = Z, a number of texts specifically exclude the value
1 for n.
Although it is logically correct (and often convenient) to speak of a
ring as an ordered triple, the notation becomes unwieldy as one progresses
further into the theory. We shall therefore adopt the usual convention of
designating a ring, say (R, +, .), simply by the set symbol R and assume
that + and, are known. The reader should realize, however, that a given
set may perfectly well be the underlying set of several different rings. Let
us also agree to abbreviate a + (- b) as a - b and subsequently refer to
this expression as the difference between a and b. As a final concession to
brevity, juxtaposition without a median dot will be used to denote the
product of two ring elements.
With these conventions on record, let us begin our formal development
of ring theory. The material covered in the next several pages will probably
be familiar to most readers and is included more to assure completeness
than to present new ideas.
Theorem 1-1. If R is a ring, then for any a, b, c E R
1) Oa = aO = 0,
2) a(-b) = (-a)b = -Cab),
3) (-a)( -b) = ab, and
4) a(b - c) = ab - ac, (b - c)a = ba - ca.
Proof These turn out, in the main, to be simple consequences of the dis-
° °
tributive laws. For instance, from + = 0, it follows that
Oa = (0 + O)a = Oa + Oa.
Thus, by the cancellation law for the additive group (R, +), we have Oa = 0.
In a like manner, one obtains aO = 0. The proof of (2) requires the fact
that each element of R has a unique additive inverse (Problem 1). Since
b + (-b) = 0,
ab + a(-b) = a(b + (-b)) = aO = 0,
which then implies that -Cab) = a( -b). The argument that (-a)b is also
the additive inverse of ab proceeds similarly. This leads immediately to (3):
(-a)( -b) = -( -a)b = - (-(ab)) = abo
The last assertion is all but obvious.
There is one very simple ring that consists only of the additive identity
° °
0, with addition and multiplication given by + = 0, 00 = 0; this ring
is usually called the trivial ring.
INTRODUCTORY CONCEPTS 7

Corollary. Let R be a ring with identity 1. If R is not the trivial ring,

then the elements 0 and 1 are distinct.
Proof Since R =1= {O}, there exists some nonzero element a E R. If 0 and
1 were equal, it would follow that a == al = aO = 0, an obvious contradic-
tion.
CONVENTION: Let us assume, once and for all, that any ring with identity
contains more than one element. This will rule out the possibility that 0
and 1 coincide.
We now make several remarks about the concept of zero divisors (the
term "divisors of zero" is also in common use):
Definition 1-3. If R is a ring and 0 =1= a E R, then a is called a left
(right) zero divisor in R if there exists some b =1= 0 in R such that
ab = 0 (ba = 0). A zero divisor is any element of R that is either a
left or right zero divisor.
According to this definition, 0 is not a zero divisor, and if R contains
an identity 1, then 1 is not a zero divisor nor is any element of R which
happens to possess a multiplicative inverse. An obvious example of a ring
with zero divisors is Zn' where the integer n > 1 is composite; ifn = n 1 n2
in Z (0 < n 1 , n2 < n), then the product n 1 ·nn 2 = 0 in Zn.
For the most part, we shall be studying rings without zero divisors.
In such rings it is possible to conclude from the equation ab = 0 that
either a = 0 or b = O. '
One can express the property of being with or without zero divisors in
the following useful way.
Theorem 1-2. A ring R is without zero divisors if and only if it satisfies
the cancellation laws for multiplication; that is, for all a, b, c E R,
ab = ac and ba = ca, where a =1= 0, implies b = c.
Proof Suppose that R is without zero divisors and let ab = ac, a =1= o.
Then, the product a(b - c) = 0, which means that b - c = 0 and b = c.
The argument is the same for the equation ba = ca. Conversely, let R
satisfy the cancellation laws and assume that ab = 0, with a =1= o. We then
have ab = aO,. whence by cancellation b = O. Similarly, b =1= 0 implies
a = 0, proving that there are no zero divisors in R.
By an integral domain is meant a commutative ring with identity which
has no zero divisors. Perhaps the best-known example of an integral domain
is the ring of integers; hence the choice of terminology. Theorem 1-2 shows
that the cancellation laws for multiplication hold in any integral domain.
The reader should be warned that many authors do not insist on the
presence of a multiplicative identity when defining integral domains; and
8

in this case the term "integral domain" would merely indicate a commutative
ring without zero divisors.
We change direction somewhat to deal with the situation where a subset
of a ring again constitutes a ring. Formally speaking,
Definition 1-4. Let (R, +, .) be a ring and S ~ R be a nonempty subset
of R. If the system (S, +, .) is itself a ring (using the induced operations),
then (S, +, .) is said to be a subring of (R, +, .).
This definition is adequate, but unwieldy, since all the aspects of the
definition of a ring must be checked in determining whether a given subset
is a subring. In seeking a simpler criterion, notice that (S, +, .) is a subring
of (R, +, .) provided that (S, +) is a subgroup of (R, +), (S, .) is a subsemi-
group of (R, .), and the two distributive laws are satisfied in S. But the
distributive and associative laws hold automatically for elements of S as a
consequence of their validity in R. Since these laws are inherited from R,
there is no necessity of requiring them in the definition of a subring.
Taking our cue from these remarks, a subring could just as well be
defined as follows. The system (S, +, .) forms a subring of the ring (R, +, .)
if and only if
1) S is a nonempty subset of R,
2) (S, +) is a subgroup of (R, +), and
3) the set S is closed under multiplication.
To add the final touch, even this definition can be improved upon; for
the reader versed in group theory will recall that (S, +) is a subgroup of
the group (R, +) provided that a - b E S whenever a, b E S. By these
observations we are led to a set of closure conditions which make it some-
what easier to verify that a particular subset is actually a subring.
Theorem 1-3. Let R be a ring and 0 =1= S ~ R. Then, S is a subring
of R if and only if
1) a, b E S imply a - bE S (closure under differences),
2) a, b E Simply ab E S (closure under multiplication).
If S is a subring of the ring R, then the zero element of S is that of R
and, moreover, the additive inverse of an element of the subring S is the
same as its inverse as a member of R. Verification of these assertions is left
as an exercise.
Example 1-6. Every ring R has two obvious subrings, namely, the set {O},
consisting only of the zero element,· and R itself. These two subrings are
usually referred to as the trivial subrings of R; all other subrings (if any
exist) are called nontrivial. We shall use the term proper subring to mean
a subring which is different from R.
INTRODUCTORY CONCEPTS 9

Example 1-7. The set Ze of even integers forms a subring of the ring Z
of integers, for
2n - 2m = 2(n - m) E Ze'
(2n)(2m) = 2(2nm) E Ze'
This example also illustrates a fact worth bearing in mind: in a ring with
identity, a subring need not contain the identity element.
Prior to stating our next theorem, let us define the center of a ring R,
denoted by cent R, to be the set
cent R = {a E Rlar = ra for all r E R}.
Phrased otherwise, cent R consists of those elements which commute with
every member of R. It should be apparent that a ring R is commutative if
and only if cent R = R.
Theorem 1-4. For any ring R, cent R is a subring of R.
Proof To be conscientious about details, first observe that cent R is non-
empty; for, at the very least, the zero element 0 E R. Now pick any two
elements a, b in cent R. By the definition of center, we know that ar = ra
and br = rb for every choice of r E R. Thus, for arbitrary r E R,
(a - b)r = ar - br = ra - rb = r(a - b),
which implies that a - b E cent R. A similar argument affirms that the
product ab also lies in cent R. In the light of Theorem 1-3, these are
sufficient conditions for the center to be a subring of R.
It has already been remarked that, when a ring has an identity, this
need not be true of its subrings. Other interesting situations may arise.
1) Some subring has a multiplicative identity, but the entire ring does not.
2) Both the ring and one of its subrings possess identity elements, but they
are distinct.
In each of the cited cases the identity for the subring is necessarily a divisor
of zero in the larger ring. To justify this claim, let l' =1= 0 denote the
identity element of the subring S; we assume further that l' does not act as
an identity for the whole ring R. Accordingly, there exists some element
a E R for which aI' =1= a. It is clear that
(al')l' = a(l'l') = aI',
or (aI' - all' = O. Since neither aI' - a nor l' is zero, the ring R has
zero divisors, and in particular l' is a zero divisor.
Example 1-8. To present a simple illustration of a ring in which the second
of the aforementioned possibilities occurs, consider the set R = Z x Z,
10

consisting of ordered pairs of integers. One converts R into a ring by

defining addition and multiplication componentwise:
(a, b) + (e, d) = (a + c, b + d),
(a, b)(e, d) = (ae, bd).
A routine calculation will show that Z x {O} = {(a, O)la E Z} forms a sub-
ring with identity element (1,0). This obviously differs from the identity
of the entire ring R, which turns out to be the ordered pair (1, 1). By our
previous remarks, (1, 0)mustbeazerodivisorinR;infact,(1, 0)(0,1) = (0,0),
where (0, 0) serves as the zero element of R.
If R is an arbitrary ring and n a positive integer, then the nth power an
of an element a E R is defined by the inductive conditions a l = a and
an = an-lao From this the usual laws of exponents follow at once:
anam = an+m , (an)m = anm (n, mE Z+).
To establish these rules, fix m and proceed by induction on n. Observe also
that if two elements a, b E R happen to commute, so do all powers of a and
b, whence (ab)n = anb n for each positive integer n.
In the event that R possesses an identity element 1 and a- l exists,
negative powers of a can be introduced by interpreting a- n as (a- l )", where
n > O. With the definition aD = 1, the symbol an now has a well-defined
meaning for every integer n (at least when attention is restricted to invertible
elements).
Paralleling the exponent notation for powers, there is the notation of
integral multiples of an element a E R. For each positive integer n, we define
the nth natural multiple na recursively as follows:
1a = a and na = (n - 1)a + a, when n > l.
If it is also agreed to let Oa = 0 and ( - n)a = - (na), then the definition of
na can be extended to all integers. Integral multiples satisfy several
identities which are easy to establish:
(n + m)a = na + ma,
(nm)a = n(ma),
n(a + b) = na + nb,
for a, b E R and arbitrary integers nand m. In addition to these rules, there
are two further properties resulting from the distributive law, namely,
n(ab) = (na)b = a(nb), and (na)(mb) = (nm)(ab).
Experience impels us to emphasize that the expression na should not
be regarded as a ring product (indeed, the integer n may not even be a
member of R); the entire symbol na is just a convenient way of indicating
INTRODUCTORY CONCEPTS 11

a certain sum of elements of R. However, when there is an identity for

multiplication, it is possible to represent na as a product of two ring elements,
namely, na = (n1)a.
To proceed further with our limited program, we must first frame a
definition.
Definition 1-5. Let Rbe an arbitrary ring. If there exists a positive
integer n such that na = 0 for all a E R, then the smallest positive integer
with this property is called the characteristic of the ring. If no such
positive integer exists (that is, n = 0 is the only integer for which na = 0
for all a in R), then R is said to be of characteristic zero. We shall write
char R for the characteristic of R.
The rings of integers, rational numbers, and real numbers are all
standard examples of systems having characteristic zero (some writers
prefer the expression "characteristic infinity"). On the other hand, the
ring P(X) of subsets of a fixed set X is of characteristic 2, since
2A = A ~ A = (A - A) u (A - A) = ¢
for every subset A £ X.
Although the definition of characteristic makes an assertion about every
element of the ring, in rings with identity the characteristic is completely
determined by the identity element. We reach this conclusion below.
Theorem 1-5. If R is any ring with identity 1, then R has characteristic
n > 0 if and only if n is the least positive integer for which n1 = o.
Proof If char R = n > 0, then na = 0 for every a E R and so, in particular,
n1 = O. Were ml = 0, where 0 < m < n, it would necessarily follow that
ma = m(1a) = (ml)a = Oa = 0
for every element a E R. The implication is that char R < n, which is
impossible. One establishes the converse in much the same way.
As we have seen, multiplication exerts a strong influence on the additive
structure of a ring through the distributive law. The following corollary to
Theorem 1-5 shows that by sufficiently restricting the multiplication in a
ring R it is possible to reach some interesting conclusions regarding the
characteristic of R.
Corollary 1. In an integral domain R all the nonzero elements have
the same additive order; this order is the characteristic of the domain
when char R > 0 and infinite when char R = o.
Proof To verify this assertion, suppose first that char R = n > o. Accord-
ing to the definition of characteristic, each element 0 1- a E R will then
possess a finite additive order m, with m ::;; n. (Recall that for an element
12

a+-O of the group (R, +) to have order m means that ma = 0 and ka +- 0

if 0 < k < m.) But the relation 0 = rna = (m1)a implies that m1 = 0, for
R is assumed to be free of zero divisors. We therefore conclude from the
theorem that n ::;; m, whence m and n are equal. In consequence, every
nonzero element of R has additive order n.
A somewhat similar argument can be employed when char R = O.
The equation ma = 0 would lead, as before, to m1 = 0 or m = O. In this
case every nonzero element a E R must be of infinite order.
The last result serves to bring out another useful point, which we place
on record as
Corollary 2. An integral domain R has positive characteristic if and
only if na = 0 for some 0 +- a E R and some integer n E Z +.
Continuing in this vein, let us next show that not any commutative
group can serve as the additive group of an integral domain.
Theorem 1-6. The characteristic of an integral domain is either zero
or a prime number.
Proof Let R be of positive characteristic n and assume that n is not a prime.
Then, n has a nontrivial factorization n = n 1 n 2 , with 1 < n 1 , n 2 < n. It
follows that
o = n1 = (n 1n 2 )1 = (n 1 n 2 )12 = (n 1 1)(n 2 1).
By supposition, R is without zero divisors, so that either n 1 1 = 0 or n2 1 = O.
Since both n 1 and n 2 are less than n, this contradicts the choice of n as the
least positive integer for which n1 = O. We therefore conclude that char R
must be prime.
Corollary. If R is a finite integral domain, then char R = p, a prime.
Turning again to the general theory, let R be any ring with identity and
consider the set Zl of integral multiples of the identity; stated symbolically,
Zl = {nlln E Z}.
From the relations
nl - ml = (n - m)l, (nl)(m1) = (nm)1
one can easily infer that ZI itself forms a (commutative) ring with identity.
The order of the additive cyclic group (ZI, +) is simply the characteristic
of the given ring R.
When R happens to be an integral domain, then Zl is a subdomain of
R (that is, ZI is also an integral domain with respect to the operations in
R). In fact, ZI is the smallest subdomain of R, in the sense that it is con-
tained in every other subdomain of R. If R is a domain of characteristic p,
PROBLEMS 13

where p is a prime, then we are able to deduce considerably more: each

nonzero element of ZI is invertible. Before establishing this, first observe
that the set ZI, regarded as an additive cyclic group of order p, consists of
p distinct elements, namely, the p sums nl, where n = 0, 1, ... ,p - 1. Now
let nl be any nonzero element of ZI (0 < n < p). Since nand p are relatively
prime, there exist integers rand s for which rp + sn = 1. But then
1 = (rp + sn)1 = r(pl) + (sl)(nl).
As pI = 0, we obtain the equation 1 = (sl)(nl), so that sl serves as the
multiplicative inverse of nl in Z1. The value of these remarks will have to
await further developments (in particular, see Chapter 4).

PROBLEMS
1. Verify that the zero element of a ring R is unique, as is the additive inverse of each
element a E R.
2. Let R be an additive commutative group. If the product of every pair of elements
is defined to be zero, show that the resulting system forms a commutative ring
(this is sometimes called the zero ring).
3. Prove that any ring R in which the two operations are equal (that is, a +b = ab
for all a, b E R) must be the trivial ring R = {O}.
4. In a ring R with identity, establish each of the following:
a) the identity element for multiplication is unique,
b) if a E R has a multiplicative inverse, then a-I is unique,
c) if the element a is invertible, then so also is -a,
d) no divisor of zero can possess a multiplicative inverse in R.
5. a) If the set X contains more than one element, prove that every nonempty proper
subset of X is a zero divisor in the riIig P(X).
b) Show that, if n > 1, the matrix ring M.(R) has zero divisors even though the
ring R may not.
6. Suppose that R is a ring with identity 1 and having no divisors of zero. For a, b E R,
verify that
a) ab = 1 if and only if ba = 1,
b) if a2 = 1, then either a = 1 or a = -1.
7. Let a, b be two elements of the ring R. If n E Z + and a and b commute, derive the
binomial expansion

where
(kn) = k!(n n!
- k)!
is the usual binomial coefficient.
14

8. An element a of a ring R is said to be idempotent if a 2 = a and nilpotent if an = 0

for some n E Z +. Show that
a) a nonzero idempotent element cannot be nilpotent,
b) every nonzero nilpotent element is a zero divisor in R.
9. Given that R is an integral domain, prove that
a) the only nilpotent element is the zero element of R,
b) the multiplicative identity is the only nonzero idempotent element.
10. If a is a nilpotent element of R, a ring with identity, establish that 1 + a is
invertible in R. [Hint: (1 + a)-l = 1 - a + a 2 + ... + (_1)n- 1an- 1 , where
an = 0.]

11. A Boolean ring is a ring with identity every element of which is idempotent. Prove
that any Boolean ring R is commutative. [H int: First show that a = - a for
every a E R.]
12. Suppose the ring R contains an element a such that (1) a is idempotent and (2) a
is not a zero divisor of R. Deduce that a serves as a multiplicative identity for R.
13. Let S be a nonempty subset of the finite ring R. Prove that S is a subring of R
if and only if S is closed under both the operations of addition and multiplication.

14. Assume that R is a ring and a E R. If C(a) denotes the set of all elements which
commute with a,
C(a) = {r E Rlar = raJ,

show that C(a) is a subring of R. Also, verify the equality cent R = naeR C(a).
15. Given a ring R, prove that
a) if S; is an arbitrary (indexed) collection of subrings of R, then their intersection
n S; is also a subring of R;
b) for a nonempty subset T of R, the set

(T> = n {SiT ~ S; S is a subring of R}

is the smallest (in the sense of inclusion) subring of R to contain T; (T> is

called the subring generated by T.
16. Let S be a subring of R, a ring with identity. For an arbitrary element a ¢ S, the
subring generated by the set S u {a} is represented by (S, a>. If a E cent R,
establish that

I
17. Let R be an arbitrary ring and n E Z+. If the set Sn is defined by
Sn = {a E Rlnka = 0 for some k > OJ,
determine whether Sn is a subring of R.

18. Establish the following assertions concerning the characteristic of a ring R:

PROBLEMS 15

a) if there exists an integer k such that ka = 0 for all a E R, then k is divisible by

char R;
b) if char R > 0, then char S :0;; char R for any subring S of R;
c) if R is an integral domain and S is a subdomain of R, then char S = char R.
19. Let R be a ring with a finite number of elements, say al, a2, ... , an' and let ni be
the order of ai regarded as a member of the additive group of R. Prove that the
characteristic of R is the least common mUltiple of the integers ni (i = 1, 2, ... , n).
20. Suppose that R is a ring with identity such that char R = n > O. If n is not prime,
show that R has divisors of zero.
21. If R is a ring which has no nonzero nilpotent elements, deduce that all the idem-
potent elements of R belong to cent R. [Hint: If a 2 = a, then (ara - ar)2 =
(ara - ra)2 = 0 for all r E R.]
22. Assume that R is a ring with the property that a2 + a E cent R for every element
a in R. Prove that R is necessarily a commutative ring. [Hint: Utilize the expression
(a + b)2 + (a + b) to show first that ab + ba lies in the center for all a, b E R.]
23. Let (G, +) be a commutative group and R be the set of all (group) homomorphisms
of G into itself The pointwise sum f + g and composition fog of two functions
J, g E R are defined by the usual rules
(f + g)(x) = f(x) + g(x), (f 0 g)(x) = f(g(x)) (x E G).

Show that the resulting system (R, +,0) forms a ring. At the same time determine
the invertible elements of R.
24. Let (G, .) be a finite group (written multiplicatively), say with elements Xl' x 2' ••• , xn ,
and let R be an arbitrary ring. Consider the set R(G) of all formal sums

(ri E R).

Two such expressions are regarded as equal if they have the same coefficients.
Addition and multiplication can be defined in R(G) by taking
n n n
I rixi + i=l
i=l
I SiXi = I (ri + S;)Xi
i=l
and

(.f rixi) (.f SiXi) .f tixi,

1=1 1=1
=
,=1
where
ti = I
XjXk=Xi
rjSk'

(The meaning of the last-written sum is that the summation is to be extended over
all subscripts j and k for which XjXk = Xi') Prove that, with respect to these
operations, R(G) constitutes a ring, the so-called group ring of Gover R.
TWO

IDEALS AND THEIR OPERATIONS

Although it is possible to obtain some interesting conclusions concerning

subrings, this concept, if unrestricted, is too general for most purposes. To
derive certain highly desirable results (for instance, the fundamental iso-
morphism theorems), additional assumptions that go beyond Definition 1-4
must be imposed. Thus, in the present chapter we narrow the field and focus
attention on a class of subrings with a stronger type of multiplicative closure,
namely, closure under multiplication by an arbitrary ring element.
Definition 2-1. A subring [ of the ring R is said to be a two-sided ideal
of R if and only if r E R and a E [ imply both ra E [ and ar E 1.
Viewed otherwise, Definition 2-1 asserts that whenever one of the
factors in a product belongs to [, then the product itself must be in [. (This
may be roughly summarized by saying that the set [ "captures" products.)
Taking stock of Theorem 1-3, which gives a minimal set of conditions
to be a subring, our current definition of a two-sided ideal may be reformu-
lated as follows.
Definition 2-2. Let [ be a nonempty subset of a ring R. Then [ is a
two-sided ideal of R if and only if
1) a, b E [ imply a - bE [, and
2) r E R and a E [ imply both products ra, ar E [.
If condition (2) of the above definition is weakened so as to require
only that the product ra belongs to [ for every choice of r E R and a E [,
we are led to the notion of a left ideal; right ideals are defined in a sym-
metric way. Needless to say, if the ring R happens to be commutative (the
most important case so far as we shall be concerned), then there is no
distinction between left, right, and two-sided ideals.
CONVENTION In what follows, let us agree that the term "ideal", un-
modified, will always mean two-sided ideal.
Before proceeding further, we pause to examine this concept by means
of several specific examples.
16
IDEALS AND THEIR OPERATIONS 17

Example 2-1. For each integer a E Z, let (a) represent the set consisting
of all integral multiples of a; that is,
(a) = {naln E Z}.
The following relations confirm (a) to be an ideal of the ring of integers:
na - rna = (n - rn)a,
rn(na) = (rnn)a, "n, nEZ.

In particular, since (2) = Ze' the ring of even integers forms an ideal of Z.
Notice, too, that (0) = {O} and (1) = Z •."
Example 2-2. Another illustration is furnished by map (X, R), the ring of
mappings from the set X into the ring R (see Example 1---4). For a fixed
element x E X, we denote by Ix the set of all mappings which take on the
value 0 at x:
Ix = {f E map (X, R)lf(x) = O}.
Now, choose f, g E Ix and hE map (X, R). From the definition of the ring
operations in map(X, R),
(f -g)(x) = f(x)-g(x) = 0-0 = 0,
while
(fh)(x) = f(x)h(x) = Oh(x) = 0,

and, in a similar manner, (hf)(x) = O. Thus, f - g, fh and hf all belong to

Ix, which implies that Ix is an ideal.
More generally, if S is any nonempty subset of X, then
I = {f E map (X, R)lf(x) = 0 for all XES}
comprises an ideal of map (X, R). Since I = nXEslx, we have a situation
where the intersection of ideals is once again an ideal. (Theorem 2-2 shows
that this is no accident.)
Before presenting our next example, we derive a fact which, despite its
apparent simplicity, will be frequently applied in the sequel.
Theorem 2-1. If I is a proper (right, left, two-sided) ideal of a ring R
with identity, then no element of I possesses a multiplicative inverse;
that is, I n R* = 0.
Proof. Let I be an ideal of R and suppose that there is some member a =1= 0
of I such that a-I exists in R. (The theorem is trivial when I = {O}.) Since
I is closed under multiplication by arbitrary ring elements, it follows that
1 = a-Ia E 1. By the same reasoning, I contains r = r1 for every r in R;
18

that is, R ~ I, whence the equality I = R. This contradicts the hypothesis

that I is a proper subset of R.
Notice that, en route, we have also established
Corollary. In a ring with identity, no proper (right, left, two-sided) ideal
contains the identity element.
Example 2-3. This example is given to show that the ring Mn(R#) of
n x n matrices over the real numbers has no nontrivial ideals. As a nota-
tional device, let us define Eij to be the n x n matrix having 1 as its ijth
entry and zeroes elsewhere. Now, suppose that I =1= {O} is any ideal of the
ring MiR#). Then I must contain some nonzero matrix (a i ;), with, say,
rsth entry ars =1= O. Since I is a two-sided ideal, the product
Err (bij)(aij)Ess
is a member of I, where the matrix (bij) is chosen to have the element a;/
down its main diagonal and zeroes everywhere else. As a result of all the
zero entries in the various factors, it is easy to verify that this product is
equal to Ers . Knowing this, the relation
(i, j = 1, 2, ... , n)
implies that all n2 of the matrices Eij are contained in 1. The clinching
point is that the identity matrix (C>i) can be written as
(C>i) = Ell + E22 + ... + E nn ,
which leads to the conclusion that (C>i) E I and, appealing to the above
corollary, that I = Mn(R#). In other words, Mn(R#) possesses no nonzero
proper ideals, as asserted.
As a matter of definition, let us call a ring R =1= {O} simple if R has no
two-sided ideals other than {O} and R. In the light of Example 2-4, the
matrix ring Mn(R#) is a simple ring.
We now take up some of the standard methods for constructing new
ideals from given ones. To begin with simpler things:
Theorem 2-2. Let {I;} be an arbitrary collection of (right, left, two-
sided) ideals of the ring R, where i ranges over some index set. Then
n Ii is also a (right, left, two-sided) ideal of R.

Proof. We give the proof for the case in which the ideals are two-sided.
First, observe that the intersection n Ii is nonempty, for each of the ideals
Ii must contain the zero element of the ring. Suppose that the elements
a, bEn Ii and r E R. Then a and b are members of Ii' where i varies over
the indexing set. Inasmuch as Ii is assumed to be an ideal of R, it follows
that a - b, ar and ra all lie in the set Ii. But this is true for every value of
IDEALS AND THEIR OPERATIONS 19

i, whence the elements a - b, ar and ra belong to n Ii' making n Ii an

ideal of R.
Consider, for the moment, an arbitrary ring R and a nonempty subset
S of R. By the symbol (S) we shall mean the set
(S) = n {II S ~ I; I is an ipeal of R}.
The collection of all ideals which contain S is not empty, since the entire
ring itself is an ideal containing any subset of R; thus, the set (S) exists and
satisfies the inclusion S ~ (S). By virtue of Theorem 2-2, (S) forms an ideal
of R, known as the ideal generated by the set S. It is noteworthy that when-
ever I is any ideal of R with S ~ I, then necessarily (S) ~ 1. For this reason,
one often speaks of (S) as being the smallest ideal of R to contain the set S.
It should be apparent that corresponding remarks apply to the one-sided
ideals generated by S.
If S consists of a finite number of elements, say a 1 , a2 , ••• , an' then the
ideal which they generate is customarily denoted by (a 1 , a2 , ••• , an). Such an
ideal is said to be finitely generated with the given elements ai as its
generators. An ideal (a) generated by just one ring element is termed a
principal ideal.
A natural undertaking is to determine the precise form of the members
of the various ideals (right, left, two-sided) generated by a single element,
say a, of an arbitr~ry ring R. The- right ideal generated by a is called a
principal righJ ideal and is denoted by (a)r. Being closed with respect to
multiplication on the right, (a)r necessarily contains all products ar (r E R),
as well as the elements na (n an integer), and, hence, includes their sum
ar + na. (As usual, the notation na represents the n-fold sum of a.) It is a
fairly simple matter to check that the set of elements of the form ar + na
constitutes a right ideal of R. Observe, too, that the element a is a member
of the ideal, .since a = aO + la. These remarks make it clear that

(a)r = {ar + nair E R; n E Z}.

When there is an identity element present, the term na becomes superfluous,
for, in this setting, we may write the expression ar + na more simply as
ar + na = ar + a(nl) = a(r + nl) = ar',
where r' = r + nl is some ring element. Thus, the set (a)r consists of all
right multiples of a by elements of R. If R is a ring with identity, we shall
frequently employ the more suggestive notation aR in place of (a)r; that is,
(a)r = aR = {arlr E R}.
Similar remarks apply, of course, to the principal left ideal (a) I generated
bya.
20

As a general comment, observe that the products ar (r E R) comprise the

set of elements of a right ideal of R even when the ring does not possess an
identity. The difficulty, however, is that this ideal need not contain a itself.
With regard to the two-sided ideal (a) generated by a, the situation is
more complicated. Certainly the elements ras, ra, as and na must all belong
to the ideal (a) for every choice of r, s E Rand n E Z. In general, the sum
of two elements ras and r' as' is no longer of the same form, so that, in order
to have closure under addition, any finite sum L rias i , where ri , Si E R, is
also required to be in (a). The reader will experience no difficulty in showing
that the principal ideal generated by a is given by
(a) = {na + ra + as + L riasilr, s, r;, S; E R; n E Z}.
finite
In case R happens to have an identity, this description of (a) reduces to the
set of all finite sums L rias i .
A particularly important type of ring is a principal ideal ring, which
we now define.
Definition 2-3. A ring R is said to be a principal ideal ring if every ideal
I of R is of the form I = (a) for some a E R.
The following theorem furnishes an example of such rings.
Theorem 2-3. The ring Z of integers is a principal ideal ring; in fact,
if I is an ideal of Z, then I = (n) for some nonnegative integer n.
Proof If I = {O}, the theorem is trivially true, since the zero ideal {O} is
the principal ideal generated by O. Suppose then that I does not consist
of the zero element alone. Now, if mEl, - m also lies in I, so that the set I
contains positive integers. Let n designate the least positive integer in I.
As I forms an ideal of Z, each integral multiple of n must belong to I, whence
(n) 5; 1.
To establish the inclusion I 5; (n), let k be an arbitrary element of 1.
By the division algorithm there exist integers q and r for which k = qn + r,
with 0 :::; r < n. Since k and qn are both members of I, it follows that
r = k - qn E I. If r > 0, we would have a contradiction to the assumption
that n is the smallest positive integer in I. Accordingly, r = 0 and
k = qn E (n). Thus, only multiples of n belong to I, implying that I 5; (n).
The two inclusions show that I = (n) and the argument is complete.
Let us now describe certain binary operations on the set of all ideals
of R. (Similar considerations apply to the sets of right and left ideals, but
for economy of effort we concentrate on two-sided ideals.) Given a finite
number of ideals 11 , 12 , ••• , In of the ring R, one defines their sum in the
natural way:
11 + 12 + ... + In = {a 1 + a2 + ... + anla; E IJ
IDEALS AND THEIR OPERATIONS 21

Then 11 + 12 + ... + In is likewise an ideal of R and is the smallest ideal

of R which contains every Ii; phrased in another way, 11 + 12 + ... + In
is the ideal generated by the union 11 U 12 U ... u In. In the special case
of two ideals I and J, our defirtition reduces to
I + J = {a + b Ia E I; b E J}.
More generally, let {Ii} be an arbitrary indexed collection of ideals of
R. The sum of this collection may be denoted by I Ii and is the ideal of
R whose members are all possible finite sums of elements from the various
ideals Ii:
IIi = {Ia;jai E I;}.
finite
The reader will take care to remember that, although {Ii} may be an infinite
family of ideals, only finite sums of elements of R are involved in the
definition above. An alternative description of I Ii could be given by
I Ii = {I ai la i E Ii; all but a finite number of the ai are O},
where it is understood that I represents an arbitrary sum with one or more
terms. Just as n Ii can be interpreted as the largest ideal of R contained
in every Ii' the sum I Ii supplies the dual notion of the smallest ideal
containing every Ii.
If R = 11 + 12 + ... + In' then each element x E R can be expressed
in the form x = a 1 + a 2 + ... + an' where a i lies in Ii. There is no
guarantee, however, that this representation of x is unique. To ensure that
every member of R is uniquely expressible as a sum of elements from the
ideals Ii' an auxiliary definition is required.
Definition 2-4. Let 1 1 , 12 , ••• , In be ideals of the ring R. We call R the
internal direct sum of 11,12, ... , In' and write R = 11 Ee 12 Ee ... Ee In'
provided that
a) R = 11 + 12 + ... + In' and
b) Ii n (Il + ... + I i- 1 + I i+ 1 + ... + In) = {O} for each i.
As was heralded by our remarks, we are now in a position to prove
Theorem 2-4. Let 11 , 12 , ••• , In be ideals of the ring R. Then the
following statements are equivalent:
1) R is the internal direct sum of I l ' 12 , ••. , In.
2) Each element x of R is uniquely expressible in the form
x = a 1 + a 2 + ... + an' where aiEl i ·
Proof There is no loss in confining ourselves to the case n= 2; the general
argurpent proceeds along similar lines. We begin by assuming that
R = 11 Ee 12. Suppose further that an element x E R has two representations
22

Then a 1 - a 2 = b 2 - b 1 • But the left-hand side of this last equation lies

in 11 , while the right-hand side is. in 12 , so that both sides belong to
11 n 12 = {O}. Itfollowsthata 1 - a 2 = b 2 - b 1 = 0,ora 1 = a 2,b 1 = b2·
In other words, x is uniquely representable as a sum a + b, a E 11 , b E 12 •
Conversely, assume that assertion (2) holds and that the element
x E 11 n 12. We may then express x in two different ways as the sum of
an element in 11 and an element in 12 ; namely, x = x + 0 (here x E 11
and 0 E 12) and x = 0 + x (here 0 E 11 and x E 12). The uniqueness
assumption of (2) implies that x = 0, in consequence of which 11 n 12 = {O};
hence, R = 11 EB 12 • This completes the proof of the theorem.
We now come to a less elementary, but extremely useful, notion; namely,
the product of ideals. Once again, assume that I and J are two ideals of
the ring R. To be consistent with our earlier definition of the sum I + J,
we should define the product IJ to be the collection of all simple products
ab, where a E I and bE J. Unfortunately, the resulting set fails to form an
ideal. (Why?) To counter this difficulty, we instead take the elements of
IJ to be all possible finite sums of simple products; stated explicitly,

IJ = {L aibilai E I; bi E J}.
finite

With this definition IJ indeed becomes an ideal of R. For, suppose that

x, Y E I J and r E R; then,

where the ai and a; are in I, and the bi and b; are in J. From this we obtain
x - Y = a 1b 1 + ... + anbn + (-a'l)b'l + ... + (-a~)b~,

rx = (ra 1 )b 1 + (ra 2 )b 2 + ... + (ran)b n.

Now, the elements -a; and rai necessarily lie in I, so that x - y and
rx E IJ; likewise, xr E IJ, making IJ an ideal of R. In point of fact, IJ is
just the ideal generated by the set of all products ab, a E I, b E J.
There is no difficulty in extending the above remarks to any finite
number of ideals 11 , 12 , ••• , In of the ring R. A moment's thought shows
that the product 11 I 2 .•• In is the ideal consisting of finite sums of terms of
the form a 1 a 2 ••• an' with a i in Ii. (It is perhaps appropriate to point out
that, because of the associative law for multiplication in R, the notation
11 I 2 ••• In is unambiguous.) A special case immediately presents itself:
namely, the situation where all the ideals are alike, say equal to the ideal
I. Here, we see that JR is the set of finite sums of products of n elements
from I:
IDEALS AND THEIR OPERATIONS 23

In this connection, it is important to observe that

I ::2 12 ::2 1 3 ::2 ... ::2 r ::2

forms a decreasing chain of ideals.

Remark. If I is a right ideal and S a nonempty subset of the ring R, then

SI = 0:: airilai E S; ri El}

finite

forms a right ideal of R. In particular, if S = {a}, then aI (a notation we

prefer to {a}I) is given by
aI = {arlr E I}.
Analogous statements can be made when I is a left ideal of R, but not, of
course, a two-sided ideal.
The last ideal-theoretic operation which we wish to consider is that of
the quotient (or residual), defined below.
Definition 2-5. Let I and J be two ideals of the ring R. The right (left)
quotient of I by J, denoted by the symbol I :r J (I :z J), consists of all
elements a E R such that aJ ~ I (Ja ~ I). In the event R is a com-
mutative ring, 'we simply write I: J.
It is by no means obvious that the set
I:J = {aERlaJ ~ I}
actually forms an ideal of R, whenever I and J are ideals. To verify this,
suppose that the elements a, bE I :r J and r E R. For any x E J, we clearly
have (a - b)x = ax - bx E I, since ax and bx both belong to I by
definition. This establishes the inclusion (a - b)J ~ I, which in turn
signifies that a - bE I :r J. Likewise, the relations raj ~ rI ~ I and
arJ ~ aJ ~ I imply that ra, ar E I :r J. In consequence, I :r J comprises
an ideal of R in its own right, and that I :z J is also an ideal follows similarly.
The purpose of the coming theorem is to point out the connection
between the quotient ideal and the operations defined previously. This
result, although it might seem to be quite special, will serve us in good
stead when we develop the theory of Noetherian rings.

Theorem 2-5. The following relations hold for ideals in a ring R (capital
letters indicate ideals of R):
1) (n Ii) :rJ = n (Ii :rJ),
2) I :r L J i = n (I :r J;),
3) I:r (JK) = (I :r K) :r J.
24

Proof Concerning (1), we have

(n Ii):,J = {a E RlaJ ~ n IJ = {a E RlaJ ~ Ii for all i}
= n{aERlaJ ~ Ii} = n(Ii:,J)·
With an eye to proving (2), notice that the inclusion J i ~ ~ J i implies
a(~ J;) ~ I if and only if aJi ~ I for all i; thus,

I:r~ J i = {aERla(~ J;) ~ I}

= {a E RlaJi ~ I for all i} = n (I:r J;).
Confirmation of the final assertion follows from
I:r(JK) = {aERla(JK) ~ I} = {aERI(aJ)K ~ I}
= {aERlaJ ~ I:rK} = (I:rK):,J.
Remark. Similar results hold for left quotients; the sole difference being
that, instead of (3), one now has I :1 (J K) = (I :1 J) :1 K.
This may be a good place to observe that if I is an ideal of the ring R
and J is an ideal of I, then J need not be an ideal of the entire ring R. For
an illustration, we turn to the ring map R # and let R be the subring con-
sisting of all continuous functions from R # into itself. Consider the sets
I = {fiIJE R;f(O) = O},
J = {ji 2 + ni 2 1JE R;f(O) = 0; n E Z},
where i denotes the identity function on R# (that is, i(x) = x for all x E R#).
A routine calculation verifies that J is an ideal of I, which, in turn, forms an
ideal of R. However, J fails to be an ideal of R, since i2 E J, while !i 2 ¢ J.
(The symbol ! is used in this setting to represent the constant function
whose value at each real number is !.) We assume that !i2 E J and derive
a contradiction. Then,

for a suitable choice of JE Rand nEZ, with J(O) = O. In consequence,

Ji 2= (! - n)i 2 , implying that J(x) = ! - n =1= 0 for every 0 =1= x E R # ;
in other words,Jis a nonzero constant function on R# - {O}. But this
obviously violates the continuity of J at O.
A condition which will ensure that J is also an ideal of R is to take R
to be a regular ring, a notion introduced by Von Neumann [52].
Definition 2-6. A ring R is said to be regular if for each element a E R
there exists some at E R such that aa' a = a.
If the element a happens to have a multiplicative inverse, then the
regularity condition is satisfied by setting a' = a- 1 ; in view of this, a' is
IDEALS AND THEIR OPERATIONS 25

often referred to as the pseudo-inverse of a. In the commutative case, the

equation aa'a = a may, of course, be written as a 2 a' = a.
The result which we have in mind now follows.
Theorem 2-6. Let I be an ideal of the regular ring R. Then any ideal
J of I is likewise an ideal of R.
Proof To start, notice that I itself may be regarded as a regular ring.
Indeed, if a E I, then aa'a = a for some a' in R. Setting b = a'aa', the
element b belongs to I and has the property that
aba = a(a'aa')a = (aa'a)a'a = aa'a = a.
Our aim is to show that whenever a E J ~ I and r E R, then both ar
and ra lie in J. We already know that ar E I; hence, by the above, there
exists an element x in I for which arxar = ar. Since rxar is a member
v of I and J is assumed to be an ideal of I, it follows that the product a(rxar)
must belong to J, or, equivalently, ar E J. A symmetric argument confirms
that ra E J.
Although Definition 2-6 appears to have a somewhat artificial air, we
might remark that the set of all linear transformations on a finite dimensional
vector space over a field forms a regular ring (Problem 20, Chapter 9).
This in itself would amply justify the study of such rings.
We now turn our attention to functions between rings and, more
specifically, to functions which preserve both the ring operations.
Definition 2-7. Let Rand R' be two rings. By a (ring) homomorphism,
or homomorphic mapping, from R into R' is meant a function f: R ~ R'
such that
f(a + b) = f(a) + f(b), f(ab) = f(a)f(b)
for every pair of elements a, b E R. A homomorphism which is also
one-to-one as a map on the underlying sets is called an isomorphism.
We emphasize that the + and· occurring on the left-hand sides of the
equations in Definition 2-7 are those of R, whereas the + and· occurring
on the right-hand sides are those of R'. This use of the same symbols for
the operations of addition and multiplication in two different rings should
cause no ambiguity if the reader attends closely to the context in which
the notation is employed.
If f is a homomorphism of R into R', then the image f(R) of R under f
will be called the homomorphic image of R. When R = R', so that the two
rings are the same, we say that f is a homomorphism of R into itself. In
this connection, a homomorphism of R into itself is frequently referred to
as an endomorphism of the ring R or, if an isomorphism onto R, an auto-
morphism of R.
26

For future use, we shall label the set of all homomorphisms from the
ring R into the ring R' by the symbol hom(R, R'). In the event that R = R',
the simpler notation hom R will be used in place of hom(R, R). (Some
authors prefer to write end R, for endomorphism, in place of hom R; both
notations have a certain suggestive power and it reduces to a matter of
personal preference.)
A knowledge of a few simple-minded examples will help to fix ideas.
Example 2-4. Let Rand R' be arbitrary rings and f: R ~ R' be the function
which sends each element of R to the zero element of R'. Then,
f(a + b) = 0 = 0 + 0 = f(a) + f(b),
(a, bE R),
f(ab) = 0 = 00 = f(a) f(b)
so thatfis a homomorphic mapping. This particular mapping, the so-called
trivial homomorphism, is the only constant function which satisfies Definition
2-7.
Example 2-5. Consider the ring Z of integers and the ring Zn of integers
modulo n. Definef: Z ~ Zn by takingf(a) = [a]; that is, map each integer
into the congruence class containing it. That f is a homomorphism follows
directly from the definition of the operations in Zn:
f(a + b) = [a + b] = [a] +n [b] = f(a) +nf(b),
f(ab) = Cab] = CaLeb] = f(a)·nf(b).

Example 2-6. In the ring map(X, R), define 'a to be the function which
assigns to each f E map (X, R) its value at a fixed element a EX; in other
words, 'a(f) = f(a). Then 'a is a homomorphism from map (X, R) into R,
known as the evaluation homomorphism at a. We need only observe that
'a(f+ g) = (f + g)(a) = f(a) + g(a) = 'a(f) + 'a (g),
'a(fg) = (fg)(a) = f(a)g(a) = 'a (f)'a(g)·
We now list some of the structural features preserved under homo-
morphisms.
Theorem 2-7. Letfbe a homomorphism from the ring R into the ring
R'. Then the following hold:
1) f(O) = 0,
2) f( -a) = -f(a) for all a E R.
If, in addition, Rand R' are both rings with identity and f(R) = R',
then
3) f(1) = 1,
4) f(a -1) = f(a) - 1 for each invertible element a E R.
IDEALS AND THEIR OPERA TIONS 27

Proof From f(O) = f(O + 0) = f(O) + f(O), we obtain f(O) = O. The

fact that f(a) + f( - a) = f(a + (- a)) = f(O) = 0 yields f( - a) = - f(a).
As regards (3), let the element a E R satisfy f(a) = 1; then,f(l) = f(a)f(1) =
f(a1) = f(a) = 1. Finally, the equationf(a)f(a- 1 ) = f(aa- 1) = f(1) = 1
shows that f(a)-l = f(a- 1), whenever a E R has a multiplicative inverse.
Two comments regarding part (3) of the above theorem are in order.
First, it is evident that
f(a)1 = f(a) = f(a1) = f(a)f(1)
for any a in R. Knowing this, one might be tempted to appeal (incorrectly)
to the cancellation law to conclude thatf(1) = 1; what is actually required
is the fact that multiplicative identities are unique. Second, if the hypothesis
thatfmap onto the set R' is omitted, then it can only be inferred thatf(1)
is the identity for the homomorphic image f(R). The element f(1) need not
serve as an identity for the entire ring R' and, indeed, it may very well happen
that f(1) =1= 1.
We also observe, in passing, that, by virtue of statement (2),
f(a - b) = f(a) + f( -b) = f(a) - f(b).
In short, any ring homomorphism preserves differences as well as sums and
products.
The next theorem indicates the algebraic nature of direct and inverse
images of subrings under homomorphisms. Among other things, we shall
see that iffis a homomorphism from the ring R into the ring R', thenf(R)
forms a subring of R'. The complete story is told below.
Theorem 2-8. Letfbe a homomorphism from the ring R into the ring
R'. Then,
1) for each subring S of R,j(S) is a subring of R'; and
2) for each subring S' of R',j-l(S') is a subring of R.
Proof To obtain the first part of the theorem, recall that, by definition,
the imagef(S) = {f(a)la E S}. Now, suppose thatf(a) andf(b) are arbitrary
elements of f(S). Then both a and b belong to the set S, as do a - band
ab (S being a subring of R). Hence,
f(a) - f(b) = f(a - b) Ef(S)
and
f(a)f(b) = f(ab) Ef(S).
According to Theorem 1-3, these are sufficient conditions for f(S) to be a
subring of R'.
The proof of the second assertion proceeds similarly. First, remember
thatf-1(S') = {a E Rlf(a) E S'}. Thus, if a, b Ef-l(S'), the imagesf(a) and
28

f(b) must be members of S'. Since S' is a subring of R', it follows at once
that
f(a - b) = f(a) - f(b) E S'
and
f(ab) = f(a)f(b) E S'.
This means that a - band ab lie inf- 1(S'), from which we conclude that
f- 1(S') forms a subring of R.
Left unresolved is the matter of replacing the term "subring" in Theorem
2-8 by "ideal". It is not difficult to show that part (2) of the theorem
remains true under such a substitution. More precisely: if l' is an ideal of
R', then the subring f- 1(1') is an ideal of R. For instance, suppose that
a Ef-1(1'), so thatf(a) E 1', and let r be an arbitrary element of R. Then,
f(ra) = f(r)f(a) E 1'; in other words, the product ra is inf-1(1'). Likewise,
ar E f- 1(1'), which helps to make f-1(1') an ideal of R.
Without further restriction, it cannot be inferred that the image f(l)
will be an ideal of R', whenever I is an ideal of R. One would need to know
that r'f(a) Ef(l) for all r' E R' and a E I. In general, there is no way of
replacing r' by some f(r) in order to exploit the fact that I is an ideal. The
answer is obvious: just take f to be an onto mapping.
Summarizing these remarks, we may now state:
Corollary. 1) For each ideal l' of R', the subring f- 1(1') is an ideal
of R.
2) If f(R) = R', then for each ideal I of R, the subring f(I) is an ideal
ofR'.
To go still further, we need to introduce a new idea.
Definition 2-8. Let f be a homomorphism from the ring R into the
ring R'. The kernel off, denoted by ker f, consists of those elements in
R which are mapped by f onto the zero element of the ring R':
ker f = {a E R!f(a) = O}.
Theorem 2-7 indicates that ker f is a nonempty subset of R, since, if
nothing else, 0 E ker f. Except for the case of the trivial homomorphism,
the kernel will always turn out to be a proper subset of R.
As one might suspect, the kernel of a ring homomorphism forms an
ideal.
Theorem 2-9. The kernel ker f of a homomorphism f from a ring R
into a ring R' is an ideal of R.
"
Proof. We already know that the trivial subring {O} forms an ideal of R'.
I'
"I Since ker f = f- 1(0), the conclusion follows from the last corollary.
IDEALS AND THEIR OPERATIONS 29

The kernel of a homomorphism may be viewed as a measure of the

extent to which the homomorphism fails to be one-to-one (hence, fails to
be an isomorphism). In more concrete terms, we have
Theorem 2-10. A homomorphismffrom a ring R into a ring R' is an
isomorphism if and only if ker f = {O}.
Proof First, iffis a one-to-one function andf(a) = 0 = f(O), then a = 0,
whence kerf = {O}. On the other hand, suppose that the kernel consists
exactly ofO. Iff(a) = f(b), then
f(a - b) = f(a) - f(b) = 0,
which means that a - bEkerf Since kerf = {O},wemusthavea - b = 0,
or a = b, making f a one-to-one function.
Two rings Rand R' are said to be isomorphic, denoted by R ~ R', if
there exists an isomorphism from the ring R onto the ring R'. Although
this definition is unsymmetric in that it makes mention of a function from
one particular ring to another, let us remark that iff: R -+ R' is a one-to-one,
onto, homomorphic mapping, the function f- 1 : R' -+ R also enjoys these
properties. We may therefore ignore the apparent lack of symmetry and
merely speak of two rings Rand R' as being isomorphic without specifying
one ring as isomorphic to the other; notationally, this situation is recognized
by writing either R ~ R' or R' ~ R.
Isomorphic rings are indistinguishable from the structural point of view,
even though they may differ in the notation for and nature of their elements
and operations. Two such rings, although not in general formally identical,
are the same for all purposes; the underlying feature is the existence of a
mapping which transports the algebraic structure of one ring to the other.
In practice, we shall often identify isomorphic rings without explicit mention.
This seems to be a natural place to insert an example.
Example 2-7. Consider an arbitrary ring R with identity and the mapping
f: Z -+ R given by f(n) = n1. (At the\ risk of being repetitious, let us again
emphasize that nl means the n-fold sum of 1.) A simple computation shows
that J, so defined, is a homomorphism from the ring Z of integers into the
ring R:
f(n + m) = (n + m)l = nl + ml = f(n) + f(m)
and
f(nm) = (nm)l = n(ml) = (nl) (ml) = f(n)f(m).
Since ker f constitutes an ideal of Z, a principal ideal ring, it follows that
kerf = {n E Zlnl = O} = (P)
for some nonnegative integer p. A moment's reflection should convince
the reader that the integer p is just the characteristic of R. In particular,
30

any ring R with identity which is of characteristic zero will contain a subring
isomorphic to the integers; more specifically, Z ~ Zl, where 1 is the
identity of R.
Suppose that f is a homomorphism from the ring R onto the ring R'.
We have already observed that each ideal 1 of the ring R determines an
ideal f(1) of the ring R'. It goes without saying that ring theory would be
considerably simplified if the ideals of R were in a one-to-one correspondence
with those of R' in this manner. Unfortunately, this need not be the case.
The difficulty is reflected in the fact that if 1 and J are two ideals of R
with 1 ~ J ~ 1 + ker f, then f(1) = f(J). The quickest way to see this is
to notice that

f(1) ~ f(J) ~ f(1 + kerf) = f(1) + f(ker f) = f(1),

from which we conclude that all the inclusions are actually equalities. In
brief, distinct ideals of R may have the same image in R'.
This disconcerting situation could be remedied by either demanding
that ker f = {o} or else narrowing our view to consider only ideals 1 with
ker f ~ 1. In either event, it follows that 1 ~ J ~ 1 + kerf = 1 and, in
consequence, 1 = J. The first of the restrictions just cited has the effect of
making the function f one-to-one, in which case Rand R' are isomorphic
rings (and it then comes as no surprise to find their ideals in one-to-one
correspondence). The second possibility is the subject of our next theorem.
We turn aside briefly to establish a preliminary lemma which will
provide the key to later success.

Lemma. Letfbe a homomorphism from the ring R onto the ring R'.
If 1 is any ideal of R such that kerf ~ 1, then 1 = f-l(J(1)).
Proof Suppose that the element a E jl(t{1)), so that j{a) E j{1). Then
f(a) = f(r) for some choice of r in 1. As a result, we will have f(a - r) = 0,
or, what amounts to the same thing, a - r E ker f ~ 1. This implies that
a E 1, yielding the inclusion f-l(J(l}) ~ 1. Since the reverse inclusion
always holds, the desired equality follows.
Here now is one of the main results of this section.

Theorem 2-11. (Correspondence Theorem). Letfbe a homomorphism

from the ring R onto the ring R'. Then there is a one-to-one correspon-
dence between those ideals 1 of R such that ker f ~ 1 and the set of all
ideals l' of R' ; specifically, l' is given by l' = f(1).
Proof Our first concern is to show that the indicated correspondence
actually maps onto the set of all ideals of R'. In other words, starting with
an ideal l' of R', we must produce some ideal 1 of the ring R, with ker f ~ 1,
IDEALS AND THEIR OPERATIONS 31

such thatf(I) = 1'. To accomplish this, it is sufficient to take I = f-1(1').

By the corollary to Theorem 2-8, f - 1(1') certainly forms an ideal of Rand,
°
since E 1',
kerf = f- 1(0) 5; f-1(1').
Inasmuch as the function f is assumed to be an onto map, it also follows
thatf(I) = f(l-1(1')) = 1'.
Next, we argue that this correspondence is one-to-one. To make things
more specific, let ideals I and J of R be given, where ker f 5; I, kerf 5; J, .
and satisfying f(I) = f(J). From the elementary lemma just established,
we see that
I =f- 1 (1(I») =f- 1(1(J») = J.
One finds in this way that the correspondence I +-+ f(I), where kerf 5; I, is
indeed one-to-one, completing the proof.
Before announcing our next result, another definition is necessary.
Definition 2-9. A ring R is said to be imbedded in a ring R' if there
exists some subring S' of R' such that R ~ S'.
In general, if a ring R is imbedded in a ring R', then R' is referred to as
an extension of R and we say that R can be extended to R'. The most
important cases are those in which one passes from a given ring R to an
extension possessing some property not present in R. As a simple applica-
tion, let us prove that an arbitrary ring can be imbedded in an extension
ring with identity.
Theorem 2-12. (Dorroh Extension Theorem). Any ring R can be im-
bedded in a ring with identity.
Proof. Consider the Cartesian product R x Z, where
RxZ= {(r,n)lrER;nEZ}.

If addition and multiplication are defined by

(a, n) + (b, m) = (a + b, n + m),
(a, n)(b, m) = (ab + ma + nb, nm),
then it is a simple matter to verify that R x Z forms a ring; we leave the
actual details as an exercise. Notice that this system has a multiplicative
identity, namely, the pair (0, 1); for
(a, n)(O, 1) = (aO + la + nO, nl) = (a, n),
and, similarly,
(0, l)(a, n) = (a, n).
32

Next, consider the subset R x {O} of R x Z consisting of all pairs of the

form (a, 0). Since
(a, 0) - (b 0) = (a - b, 0),
(a, O)(b, 0) = (ab, 0),
it is evident that R x {O} constitutes a subring of R x Z. A straightforward
calculation, which we omit, shows that R x {O} is isomorphic to the given
ring R under the mapping f: R --t R x {O} defined by f(a) = (a, 0). This
process of extension therefore imbeds R in R x Z, a ring with identity.
A point to be made in connection with the preceding theorem is that
the imbedding process may be carried out even if the given ring has an
identity to start with. Of course, in this case the construction has no
particular merit; indeed, the original identity element only serves to introduce
divisors of zero into the extended ring.
Although Theorem 2-12 shows that we could confine our study to rings
with identity, it is nonetheless desirable to develop as much of the theory
as possible without the assumption of such an element. Thus, unless an
explicit statement is made to the contrary, the subsequent discussions will
not presuppose the existence of a multiplicative identity.
We now take a brief look at a different problem, namely, the problem
of extending a function from a subring to the entire ring. In practice, one
is usually concerned with extensions which retain the characteristic features
of the given function. The theorem below, for instance, presents a situation
in which it is possible to extend a homomorphism in such a way that the
extended function also preserves both ring operations.
Theorem 2-13. Let I be an ideal of the ring Rand f a homomorphism
from I onto R', a ring with identity. If I ~ cent R, then there is a
unique homomorphic extension of f to all of R.
Proof As a start, we choose the element u E I so that f(u) = 1. Since I
constitutes an ideal of R, the product au will lie in the set I for each choice
of a E R. It is therefore possible to define a new function g: R --t R' by
setting g(a) = f(au) for all a in R. If the element a happens to belong to
I, then
g(a) = f(au) = f(a)f(u) = f(a)1 = f(a),
showing that g actually extends the original functionf
The next thing to confirm is that both ring operations are preserved
by g. The case of addition is fairly obvious: if a, b E R, then

g(a + b) = f((a + b)u) = f(au + bu)

= f(au) + f(bu) = g(a) + g(b).
IDEALS AND THEIR OPERATIONS 33

As a preliminary step to demonstrating that g also preserves multiplication,

notice that
f(ab)u 2 ) = f(abu)f(u) = f(abu).
From this we are able to conclude that
g(ab) = f(abu) = f(abu 2 ) = f(au)(bu))
= f(au)f(bu) = g(a)g(b).

The crucial third equality is justified by the fact that u E cent R, hence, -
commutes with b.
As regards the uniqueness assertion, let us assume that there is another
homomorphic extension off to the set R; call it h. Sincefand h must agree
on I and, more specifically, at the element u, h(u) = f(u) = 1. With this
in mind, it follows that
h(a) = h(a)h(u) = h(au) = f(au) = g(a)
for all a E R and so hand g are the same function. Hence, there is one and
only one way of extendingfhomomorphically from the ideal I to the whole
ring R.
Before closing the present chapter, there is another type of direct sum
which deserves mention. To this purpose, let R I , R 2 , ... ,Rn be a finite
number of rings (not necessarily subrings of a common ring) and consider
their Cartesian product R = X Ri consisting of all ordered n-tuples
(aI' a2 , ... ,an)' with ai E R i . One can easily convert R into a ring by
performing the ring operations componentwise; in other words, if
(aI' a2 , •.. , an) and (b l , b2 , ••• , bn) are two elements of R, simply define
(aI' a 2 , •.• ,an) + (b l , b2 , .•• , bn) = (a l + b l , a2 + b2 , .•• ,an + bn)
and
(aI' a2 , ••• , an)(b l , b2 , .•• , bn) = (alb l , a2 b2 , .•• , anbn)·
The ring so obtained is called the external direct sum of R I , R 2 , ••. , Rn
and is conveniently written R = RI R2 +
Rn. (Let us caution + ... +
that the notation is not standard in this matter.) In brief, the situation is
this: An external direct sum is a new ring constructed from a given set of
rings, and an internal direct sum is a representation of a given ring as a sum
of certain of its ideals. The connection between these two types of direct
sums will be made clear in the next paragraph.
If R is the external direct sum of the rings Ri (i = 1,2, ... , n), then the
individual Ri need not be subrings, or even subsets, of R. However, there is
an ideal of R which is the isomorphic image of R i • A straightforward
calculation will convince the reader that the set
Ii = {(O, ... , 0, ai' 0, ... , O)iai E R i }
34

(that is, the set consisting of all n-tuples with zeroes in all places but the
ith) forms an ideal of R naturally isomorphic to Ri under the mapping
which sends (0, ... , 0, ai' 0, ... , 0) to the element ai. Since
(ai' a2 , .•• ,an) = (ai' 0, 0, ... ,0) + (0, a 2 , 0, ... ,0) + ... + (0, 0, ... , 0, an),
it should also be clear that every member of R is uniquely representable as
a sum of elements from the ideals Ii. Taking note of Theorem 2-4, this
means that R is the internal direct sum of the ideals Ii and so
Rl + R2 + ... + Rn = Ii EEl 12 EEl ... EEl In (Ri ~ Ii)·
In summary, the external direct sum R of the rings R 1 , R 2 , ••• , Rn is also
the internal direct sum of the ideals Ii' 12 , ••• , In and, for each i, Ri and
Ii are isomorphic.
In view of the isomorphism just explained, we shall henceforth refer to
the ring R as being a direct sum, not qualifying it with the adjective
"internal" or "external", and rely exclusively on the EEl-notation. The term
"internal" merely reflects the fact that the individual summands, and not
isomorphic copies of them, lie in R.
We take this opportunity to introduce the simple, but nonetheless useful,
notion of a direct summand of a ring. In formal terms, an ideal I of the
ring R is said to be a direct summand of R if there exists another ideal J
of R such that R = I EEl J. For future use, let us note that should the ideal
I happen to have an identity element, say the element e E I, then it will
automatically be a direct summand of R. The argument proceeds as follows.
For any choice of r E R, the product re E I. The assumption that e serves
as an identity for I then ensures that e(re) = reo At the same time (and for
thesamereasons),(er)e = er. Combining these pieces, wegetre = ere = er,
which makes it plain that the element e lies in the center of R. This is the
key point in showing that the set J = {r - relr E R} forms an ideal of R;
the details are left to the reader. We contend that the ring R is actually the
direct sum of I and J. Certainly, each element r of R may be written as
r = re + (r - re), where re E I and r - re E J. Since I II J = {O}, this is
the only way r can be expressed as a sum of elements of I and J. (A moment's
thought shows that if a E I II J, say a = r - re, then a = ae = (r - re)e =
r(e - e2 ) = 0.) It is also true that the ideal I = eR = Re, but we did not
need this fact here.
As a further application of the idea of a direct summand, let us record
Theorem 2-14. If the ring R is a direct summand in every extension
ring containing it as an ideal, then R has an identity.
Proof To set this result in evidence, we first imbed R in the extension ring
R' = R x Z in the standard way (see Theorem 2-12). Then, R ~ R x {O},
where, as is easily verified, R x {O} constitutes an ideal of R'. We may
PROBLEMS 35

therefore regard R as being an ideal of the ring R'. Our hypothesis now
comes into play and asserts that R' = R EB J for a suitable ideal J of R'.
It is thus possible to choose an element (e, n) in J so that (0, -1) = (r,O) +
(e, n), for some r E R. The last-written equation tells us that e = - rand
n = -1; what is important is the resulting conclusion that (e, -1) E J.
for arbitrary r E R, the product (r, O)(e, -1) = (re - r, 0) will consequently
be in both Rand J (each being an ideal of R'). The fact that R n J = {O}
forces (re - r, 0) = (0, 0); hence, re = r. In a like fashion, we obtain
er = r, proving that R admits the element e as an identity.

PROBLEMS
1. If I is a right ideal and J a left ideal of the ring R such that I n J = {O}, prove
that ab = 0 for all a E I, b E J.
2. Given an ideal I of the ring R, define the set C(I) by
C(I) = {r E Rlra - ar E I for all a E R}.
Verify that C(I) forms a subring of R.
3. a) Show by example that if I and J are both ideals of the ring R, then I u J need
not be an ideal of R.
b) If {Ii} (i = 1,2, ... ) is a collection of ideals of the ring R such that! 1 ~ I 2 ~ ...
~ In ~ "', prove that u Ii is also an ideal of R.

4. Consider the ring Mn(R) ofn x n matrices over R, a ring with identity. A square
matrix (au) is said to be upper triangular if au = 0 for i > j and strictly upper
triangular if aij = 0 for i ~ j. Let T,,(R) and T!(R) denote the sets of all upper
triangular and strictly upper triangular matrices in Mn(R), respectively. Prove
each of the following:
a) T,,(R) and T!(R) are both subrings of Mn(R).
b) T!(R) is an ideal of the ring T,,(R).
c) A matrix (au) E T,,(R) is invertible in T,,(R) if and only if aii is invertible in R
for i = 1,2, ... , n. [Hint: Induct on the order n.]
d) Any matrix (a ij ) E T!(R) is nilpotent; in particular, (ai)n = O.
5. Let I be an ideal of R, a commutative ring with identity. For an element a E R,
the ideal generated by the set I u {a} is denoted by (I, a). Assuming that a ¢ I,
show that
(I, a) = {i + rali E I, r E R}.
6. In the ring Z of integers consider the principal ideals (n) and (m) generated by
the integers nand m. Using the notation of the previous problem, verify that

«n), m) = «m), n) = (n) + (m) = (n, m) = (d),

where d is the greatest common divisor of nand m.
36

7. Suppose that I is a left ideal and J a right ideal of the ring R. Consider the set

where 1: represents a finite sum of one or more terms. Establish that lJ is a two-
sided ideal of R and, whenever I and J are themselves two-sided, that lJ s;;; I II J.

8. If S is any given nonempty subset of the ring R, then

annrS = {r E Rlar = 0 for all a E S}

is called the right annihilator of S (in R); similarly,

ann/S = {r E Rlra = 0 for all a E S}

is the left annihilator of S. When R is a commutative ring, we simply speak of the

annihilator of S and use the notation ann S. Prove the assertions below:
a) annrS (ann/ S) is a right (left) ideal of R.
b) If S is a right (left) ideal of R, then annrS (ann S) is an ideal of R.
c) If S is an ideal of R, then annrS and ann/S are both ideals of R.
d) When R has an identity element, annrR = ann/R = {O}.

9. Let 11,1 2, ... , In be ideals of the ring R with R = 11 + 12 + ... + In. Show
that this sum is direct if and only if a 1 + a2 + ... + an = 0, with ai E Ii' implies
that each ai = O.
10. If P(X) is the ring of all subsets of a given set X, prove that
a) the collection of all finite subsets of X forms an ideal of P(X);
b) for each subset Y S;;; X, P(Y) and P(X - Y) are both principal ideals of P(X),
with P(X) = P(Y) Et> P(X - Y).
11. Suppose that R is a commutative ring with identity and that the element a E R
is an idempotent different from 0 or 1. Prove that R is the direct sum of the principal
ideals (a) and (1 - a).

12. Let I, J and K be ideals of the ring R. Prove that

a) I(J + K) = lJ + IK, (I + J)K = IK + JK;
b) if I :2 J, then I II (J + K) = J + (I 11K).

13. Establish that in the ring Z, if I = (12) and J = (21), then

1+ J = (3), I II J = (84), lJ = (252), I:J = (4), J:I = (7).

[Hint: In general, (a):(b) = (c), where c = a/gcd (a, b).]

14. Given ideals I and J of the ring R, verify that
a) O:rI = ann/I, and 0:/1 = ann.] (notation as in Problem 8);
b) 1:.1 (I:/J) is the largest ideal of R with the property that (I:.1)J s;;; I
(J(I:/ J) s;;; I).

15. Let I, J and K be ideals of R, a commutative ring with identity. Prove the following
assertions:
PR08LEMS 37

a) If I~ J, then I:K ~ J:K and K:I ;2 K:J.

b) 1:1"+1 = (I:1"):J = (I:J):1"foranynEZ+.
c) I:J = R if and onlyifJ ~ 1.
d) I:J = 1:(1 + J).

16. If I is a right ideal of R, a ring with identity, show that I:/R = {a E RIRa ~ I}
is the largest two-sided ideal of R contained in I.
17. Given that f is a homomorphism from the ring R onto the ring R', prove that
a) f(cent R) ~ cent R'.
b) If R is a principal ideal ring, then the same is true of R'. [Hint: For any a E R,
f(a)) = (J(a)).]
c) If the element a E R is nilpotent, then its image f(a) is nilpotent in R'.
18. Let R be a ring with identity. For each invertible element a E R, show that the
function!,.: R --+ R defined by fa (x) = axa- 1 is an automorphism of R.
19. Letfbe a homomorphism from the ring R into itself and S be the set of elements
that are left fixed by f; in symbols,
S = {a E Rlf(a) = a}.
Establish that S forms a subring of R.
20. If f is a homomorphism from the ring R into the ring R', where R has positive
characteristic, verify that char f(R) :::;; char R.
21. Letfbe a homomorphism from the commutative ring R onto the ring R'. If I and
J are ideals of R, verify each of the following:
a) f(I + J) = f(1) + f(1);
b) f(I1) = f(I)f(1);
c) f(I n 1) ~ f(I) n f(1), with equality if either I ;2 kerf or J ;2 kerf;
d) f(I: 1) ~ f(1) :f(1), with equality if I ;2 kerf
22. Show that the relation R ~ R' is an equivalence relation on any set of rings.
23. Let R be an arbitrary ring. For each fixed element a E R, define the left-multiplica-
tion function T.: R --+ R by taking T.(x) = ax. If TR denotes the set of all such
functions, prove the following:
a) T. is a (group) homomorphism of the additive group of R into itself;
b) TR forms a ring, where multiplication is taken to be functional composition;
c) the mappingf(a) = T. determines a homomorphism of R onto the ring TR ;
d) the kernel offis the ideal ann,R;
e) if for each 0 =1= a E R, there exists some bE R such that ab =1= 0, then R ~ TR •
(In particular, part(e) holds whenever R has an identity element.)
24. Let R be an arbitrary ring and R x Z be the extension ring constructed in Theorem
2-12. Establish that
a) Rx {O} is an ideal of RxZ;
b) Z ~ {O}xZ;
c) if a is an idempotent element of R, then the pair ( - a, 1) is idempotent in R x Z,
while (a, 0) is a zero divisor.
38

25. Suppose that R is a ring of characteristic n. If addition and multiplication are

defined in R x Zn = {(x, a)lx E R; a E Zn} by

(x, a) + (y, b) = (x + y, a +n b),

(x, a)(y, b) = (xy + ay + bx, a·nb),

prove that R x Zn is an extension ring of R of characteristic n. Also show that

R x Zn has an identity element.
26. Let R = Rl $ R2 $ ... $ Rn be the (external) direct sum of a finite number of
rings Ri with identity (i = 1, 2, ... , n).
a) For fixed i, define the mapping 7I: i : R ..... Ri as follows: if a = (a l , a 2, ... , an),
where a j E R j , then 7I: i(a) = ai . Prove that 7I: i is a homomorphism from the
ring R onto R i •
b) Show that every ideal I of R is of the form I = 11 $ 12 $ ... $ In' with Ii
an ideal of R i . [Hint: Take Ii = 7I: i(I). If b i E Ii' then there exists some (b l , ••. ,
bi' ... , bn) E I. It follows that (b 1, ... , bi' ... , bnHO, ... , 1, ... ,0) = (0, ... , bi'
... ,0) E I.]
27. A nonempty subset A of a ring R is termed an adeal of R if
(i) a, b E A imply a + b E A,
(ii) r E R and a E A imply both ar E A and ra E A.
Prove that
a) An adeal A of R is an ideal of R if for each a E A there is an integer n +- 0,
depending upon a, such that na E aR + Ra. (This condition is satisfied, in
particular, if R has a multiplicative identity.)
b) Whenever R is a commutative ring, the condition in part (a) is a necessary as
well as sufficient condition for an adeal to be an ideal. [Hint: For any a E R,
the set A = {naln E Z+} + aR is an adeal of R; hence, an ideal of R.]
28. Let R be a ring with identity and Mn(R) be the ring of n x n matrices over R.
Prove the following:
a) If I is an ideal of the ring R, then Mn(I) is an ideal of the matrix ring Mn(R).
b) Every ideal of Mn(R) is of the form Mn(I), where I is an ideal of R. [Hint: Let
Fij(a) denote the matrix in Mn(R) having a as its ijth entry and zeroes elsewhere.
For any ideal .It in Mn(R), let 1 be the set of elements in R which appear as
entries for the matrices in.lt. Given any a E I, say a is the rsth entry of a
matrix A E .It, it follows that Fij(a) = Fir(1)AFsi1) E .It.]
c) If R is a simple ring, then so is the matrix ring Mn(R).
29. Let R be a ring with the property that every subring of R is necessarily an ideal of
R. (The ring Z, for instance, enjoys this property.) If R contains no divisors of zero,
prove that multiplication is commutative. [Hint: Given 0 +- a E R, consider the
subring S generated by a. For arbitrary bE R, ab = rES, so that ar = ra.]
THREE

THE CLASSICAL ISOMORPHISM THEOREMS

In this chapter we shall discuss a number of significant results having to

do with the relationship between homomorphisms and quotient rings (which
we shall shortly define). Of these results perhaps the most crucial is Theorem
3-7, commonly known as the Fundamental Homomorphism Theorem for
Rings. The importance of this result would be difficult to overemphasize,
for it stands as the cornerstone upon which much of the succeeding theory
rests.
The notion of an ideal carries with it a natural equivalence relation.
For, given an ideal I of the ring R, it is a routine matter to check that the
relation defined by a == b if and only if a - bEl is actually an equivalence
relation on R. As such, this relation induces a partition of R into equivalence
classes, the exact nature of which is determined below.
Theorem 3-1. If I is an ideal of the ring R, then the equivalence class
of b E R for the relation == is the set
b + I = {b + iii E I}.
Proof Let [b] = {x E Rlx == b}. If a = b + i is any member of b + I,
then a - b = i E I. By definition of ==, this implies that a E [b], and so
b + I ~ [b J. On the other hand, if x E [b], we must have x - b = i for
some i in I, whence x = b + i E b + I. Thus, the inclusion [b] ~ b + I
also holds and equality follows.
The usual practice is to speak of any set of the form b + I as a coset
of I in R, and to refer to the element b as a representative of b + 1. For
future reference we next list some of the basic properties of co sets ; these
are well-known facts about equivalence classes (see Appendix A) translated
into the present notation.
Theorem 3-2. If I is an ideal of the ring R and a, b E R, then each of
the following is true:
1) a + I = I if and only if a E 1.
2) a + I = b + I if and only if a - bEl.
3) Either a + I = b + I or else a + I and b + I are disjoint.
39
40

Given an ideal I of the ring R, let us employ the symbol R/I to denote
the collection of all cosets of I in R; that is,
R/I = {a + Iia E R}.
The set R/I can be endowed with the structure of a ring in a natural way;
all we need do is define addition and multiplication as follows:
(a + I) + (b + I) = (a + b) + I,
(a + I)(b + I) = ab + 1.
One is faced with the usual problem of showing that these operations are
actually well-defined, so that the sum and product of the two co sets a + I
and b + I do not depend on their particular representatives a and b. To
this end, suppose that
a+I=a'+I and b +I= b' + I.
Then a - a' = il and b - b' = iz for some ii' iz E 1. From this we conclude
that
(a + b) - (a' + b') = (a - a') + (b - b')
= il + iz E I,
which, by Theorem 3-2, indicates that (a + b) + I = (a' + b') + I. The
net result is that (a + I) + (b + I) = (a' + I) + (b' + I). With regard
to the multiplication of cosets, we observe that
ab - a'b' = a(b - b') + (a - a')b'
= aiz + i1b' E I,
since both the products aiz and i l b' must be in 1. The implication, of course,
is that ab + 1 = a'b' + I; hence, our definition of multiplication in RjI is
meaningful.
The verification that R/I, under the operations defined above, forms a
ring is easy and the details are left to the reader. To assure completeness,
we simply state
Theorem 3-3. If I is an ideal ofthe ring R, then R/I is also a ring, known
as the quotient ring (or factor ring) of R by 1.
In Theorem 2-9 we saw that certain ideals occur as kernels of homo-
morphisms. Let us now demonstrate that every ideal does indeed arise in
this manner.
Theorem 3-4. Let I be an ideal of the ring R. Then the mapping
nat I : R -+ R/I defined by natI(a) = a +I
is a homomorphism of R onto the quotient ring R/I; the kernel of natI
is precisely the set I.
THE CLASSICAL ISOMORPHISM THEOREMS 41

Proof The fact that nat[ is a homomorphism follows directly from the
manner in which the operations are defined in the quotient ring:
nat[(a + b) = a + b + J = (a + J) + (b + J)
= nat[(a) + nat[(b);
nat[(ab) = ab + J = (a + J)(b + J)
= nat[(a) nat[(b).
That nat[ carries R onto R/J is all but obvious; indeed, every element of
R/J is a coset a + J, with a E R, and so by definition nat[(a) = a + 1.
Inasmuch as the coset J = 0 + J serves as the zero element for the
ring R/J, we necessarily have

ker (nat[) = {a E Rlnat[(a) = I}

= {a E Ria + J = J} = J.
The last equality was achieved by invoking Theorem 3-2.
It is customary to speak of the function nat[, which assigns to each
element of R the coset in R/J of which it is the representative, as the natural,
or canonical, mapping of R onto the quotient ring R/1. When there is no
danger of confusion, we shall omit the subscript J in writing this mapping.
There are two standard techniques for investigating the structure of a
particular ring. One method calls for finding all the ideals of the ring, in
the hope of gaining information about the ring through its local structure.
The other approach is to determine all homomorphisms from the given ring
into a simpler ring; the idea here is that the homomorphic images will tend
to reflect some of the algebraic properties of the original ring. (The reader
is warned to proceed with some care, since, for example, it is quite possible
for multiplication to be commutative in the image ring, without the given
ring being commutative.) Although these lines of attack aim in different
directions, Theorems 2-9 and 3-4 show that for all practical purposes these
are the same; every ideal determines a homomorphism, and every homo-
morphism determines an ideal.
Example 3-1. A simple illustration to keep in mind when working with
quotient rings is provided by the ring Z of integers and the principal ideal
(n), where n is a positive integer. The co sets of (n) in Z take the form

a + (n) = {a + knlk E Z},

from which it is clear that the cosets of (n) are precisely the congruence
classes modulo n. What we earlier described as the operations for con-
gruence classes in Zn can now be viewed as coset operations in Z/(n):
(a + (n)) + (b + (n)) = a + b + (n),
(a + (n))(b + (n)) = ab + (n).
42

In short, the ring Zn of integers modulo n could just as well be interpreted

as the quotient ring of Z by (n).
As regards the incidence of ideals in a quotient ring, it should be noted
that the Correspondence Theorem applies, in particular, to the case in
which we start with an ideal 1 of the ring R and take the homomorphism
f to be the natural mapping natI ; R -+ R/1. Since ker (natI ) = 1, the
conclusion of the Correspondence Theorem is modified slightly.
Theorem 3-5. Let 1 be an ideal of the ring R. Then there is a one-to-one
correspondence between those ideals J of R such that 1 £ J and the
set of all ideals J' of the quotient ring R/1; specifically, J' is given by
J' = natIJ.
Viewed otherwise, Theorem 3-5 asserts that the ideals of R/1 have the
form J/1, where J is an ideal of R containing 1. In this context, J/1 and
natIJ are both used to designate the set {a + 11a E J}.
By way of an application of these ideas, consider the following statement;
The ring Zn of integers modulo n has exactly one ideal for each positive
divisor m of n, and no other ideals. In the first place, since Zn = Z/(n),
Theorem 3-5 tells us that there is a one-to-one correspondence between
those ideals of the ring Z which contain (n) and the set of ideals of Zn' But
the ideals of Z are just the principal ideals (m), where m is a nonnegative
integer. The outcome is that there is a one-to-one correspondence between
the ideals of Zn and those ideals (m) of Z such that (m) :2 (n); this last
inclusion occurs if and only if m divides n.
. Theorem 3-6. (Factorization of Homomorphisms). Let f be a homo-
morphism of the ring R onto the ring R', and 1 be an ideal of R such
that 1 £ kerf. Then there exists a unique homomorphismJ: R/1 -+ R'
with the property thatf = J natI .
0

Proof. To start, we define a functionJ; R/1 -+ R', called the induced mapping,
by taking
J(a + 1) = f(a) (a E R).
The first question to be raised is whether or not J is actually well-defined.
That is to say, we must establish that this function has values which depend
only upon the co sets of 1 and in no way on their particular representatives.
In order to see this, let us assume a + 1 = b + 1. Then a - b E 1 £ ker f.
This means that
f(a) = f(a - b + b) = f(a - b) + f(b) = f(b)
and, by the manner in which J was defined, that J(a + 1) = J(b + 1).
Hence, the functionJis constant on the cosets of 1, as we wished to demon-
strate.
THE CLASSICAL ISOMORPHISM THEOREMS 43

A routine computation, involving the definition of the operations in

R/f, confirms thatJis indeed a homomorphism:
J((a + f) + (b + f)) = na + b + f) = f(a + b)
= f(a) + f(b) = na + f) + nb + f);
and, likewise,
J((a + f)(b + f)) = nab + f) = f(ab)
= f(a)f(b) = na + f)nb + f).
In this connection, notice that for each element a E R,
f(a) = na + f) = J(natI(a)) = (J natI )(a) 0

whence the equality f = J natI . It only remains to show that this fac-
0

torization is unique. Suppose also thatf = go natI for some other function
g: R/f --+ R'. But then
na + f) = f(a) = (g 0 natI )(a) = g(a + f)
for all a in R, and so g = J. The induced mappingJis thus the only function
from the quotient ring R/f into R' satisfying the equation f = J natI . 0

Corollary. The induced mapping J is an isomorphism if and only if

kerf ~ f.
Proof What is required here is an explicit description of the kernel of J,
to wit
kerJ = {a + f17(a + f) = O}
= {a + flf(a) = O}
= {a + flaEkerj} = natI (kerf).
With reference to Theorem 2-10, a necessary and sufficient condition for
J to be an isomorphism is that ker J = f. In the present setting, this
amounts to the demand that natI(kerf) = f, which in turn is equivalent
to the inclusion ker f ~ f.
In view of the equality f = J natI , the conclusion of Theorem 3-6 is
0

sometimes expressed by saying that the homomorphism f can be factored

through the quotient ring R/f or, alternatively, that f can be factored by
natI . What we have proved, in a technical sense, is that there exists one
and only one function J which makes the following diagram of maps
commutative:
R~R'
~atI\ IJ
R/f
(Speaking informally, a "mapping diagram" is commutative if, whenever
there are two sequences of arrows in the diagram leading from one ring to
44

another, the composition of mappings along these paths produces the same
function.)
A rather simple observation, with far-reaching implications, is that
whenever J = ker f, so that both the Factorization Theorem and its Corol-
lary are applicable, f induces a mapping J under which RIJ and R' are
isomorphic rings. We summarize all this in the following theorem, a result
which will be invoked on many occasions in the sequel.
Theorem 3-7. (Fundamental Homomorphism Theorem). If f is a
homomorphism from the ring R onto the ring R', then Riker f ~ R'.
Theorem 3-7 states that the images of R under homomorphisms can
be duplicated (up to isomorphism) by quotient rings of R ; to put it another
way, every homomorphism of R is "essentially" a natural mapping. Thus,
the problem of determination of all homomorphic images of a ring has
been reduced to the determination of its quotient rings.
Let us use Theorem 3-7 to prove that any homomorphism onto the
ring of integers is uniquely determined by its kernel. As the starting point
of our endeavor, we establish a lemma which is of independent interest.
Lemma. The only nontrivial homomorphism from the ring Z of
integers into itself is the identity map iz .
Proof Because each positive integer n may be written as n = 1 + 1 + ...
+ 1 (n summands), the operation-preserving nature of f implies that
f(n) = nf(l). On the other hand, if n is an arbitrary negative integer, then
-n E Z+ and so
f(n) = f( -( -n)) = -f( -n) = -( -n)f(l) = nf(l).
Plainly,f(O) = 0 = Of(l). The upshot is thatf(n) = nf(l) for every n in Z.
Because f is not identically zero, we must have f(l) = 1; to see that this
is so, simply apply the cancellation law to the relation f(m) = f(m1) =
f(m)f(l), where f(m) =1= o. One finds in this way that f(n) = n = iz(n) for
all nEZ, making f the identity map on Z.
Corollary. There is at most one homomorphism under which an
arbitrary ring R is isomorphic to the ring z.
Proof Suppose that the rings Rand Z are isomorphic under two functions
f, g: R -4 Z. Then the composition .r
g - 1 is a homomorphic mapping
0

from the ring Z onto itself. Knowing this, the lemma just proved implies
thatfo g-l = i z , or f = g.
We now have the necessary information to prove the following result.
Theorem 3-8. Any homomorphism from an arbitrary ring R onto the
ring Z of integers is uniquely determined by its kernel.
THE CLASSICAL ISOMORPHISM THEOREMS 45

Proof Let f and g be two homomorphisms from the ring R onto Z with
the property that ker f = ker g. Our aim, of course, is to show that f and g
must be the same function. Now, by Theorem 3-7, the quotient rings
Riker f and Riker g are both isomorphic to the ring of integers via the
induced mappings J and g, respectively. The assumption that f and g have
a common kernel, when combined with the preceding corollary, forces
J = g. It follows at once from the factorizations
f = Jo nat kerf , g = go nat kerg
that the functionsfand g are themselves identical.
The next two theorems are somewhat deeper results than usual and
require the full force of our accumulated machinery. They comprise what
are often called the First and Second Isomorphism Theorems, and have
important applications in the sequel. (The reader is cautioned that there
seems to be no universally accepted numbering for these theorems.)
Theorem 3-9. Let f be a homomorphism of the ring R onto the ring
R' and let I be an ideal of Rlf ker f ~ I, then RII ~ R' I f(1).
Proof Before becoming involved in the details of the proof, let us remark
that 'the corollary to Theorem 2-8 implies that f(1) is an ideal of the ring
R'; thus, it is meaningful to speak of the quotient ring R'I f(1).
Let us now define the function g: R -+ R'I f(l) by g = natf(l) f, where
0

nat1(1): R' -+ R'I f(1) is the usual natural mapping. Thus, g merely assigns
to each element a E R the cosetf(a) + f(1) in R'If(1). Since the functions
f and natf(l) are both onto homomorphisms, their composition carries R
homomorphically onto the quotient ring R' I f(1).
The crux of the argument is to show that ker g = I, for then the desired
conclusion would be an immediate consequence of the Fundamental
Homomorphism Theorem. Since the zero element of R' I f(l) is just the
coset f(1), the kernel of g consists of those members of R which are mapped
by g onto f(1):
ker g = {a E Rlg(a) = f(1)}
= {a E R f(a) + f(l) = f(1)}
= {a E Rlf(a) Ef(l)} = f-1(1(I)).

The hypothesis that ker f ~ I allows us to appeal to the lemma preceding

Theorem 2-11, from which it may be concluded that I = f-1(1(1)). But
then I = ker g, completing the argument.
When applying this result, it is sometimes preferable to start with an
arbitrary ideal in R' and utilize inverse images rather than direct images.
The theorem can then be reformulated in the following way.
46

Corollary 1. Let f be a homomorphism from the ring R onto the ring

R'. If l' is any ideal of R ', then RIf-1(1') ~ R'/I'.
Proof In compliance with the corollary to Theorem 2-8,J-1(I') forms an
ideal of R. Furthermore, ker f S; f- 1 (I'), so that Theorem 3-9 leads directly
to the isomorphism

Another special case, itself of interest, is the following.

Corollary 2. Let 1 and J be two ideals of the ring R, with J S; 1. Then
I/J is an ideal of RjJ and (R/J)/(I/J) ~ R/1.
Proof As we know, if 1 is an ideal of the ring R andfis any homomorphism
of R, thenf(l) constitutes an ideal of the imagef(R). In the setting at hand,
take f to be the natural mapping natJ : R ~ R/1; then 1/1 = nat J 1 forms
an ideal ofthe quotient ring R' = RjJ. Since ker (nat J ) = J S; 1, Theorem
3-9 implies that R/1 is isomorphic to (RjJ)/(I/J) under the induced mapping
g where 9 = natllJ 0 natJ •
The diagram displayed below may be of some help in visualizing the
situation described by the last corollary:
R ~R/1
natJ

RjJ
1 ----+
1~
(R/J)/(1jJ)
natll }
By virtue of our assumptions, there exists a (necessarily unique) isomorphism
g: R/I ~ (RjJ)/(IjJ) such that
g natl 0 = nat llJ 0 natJ .
Let us now take up the second of our general isomorphism theorems.
Theorem 3-10. If I and J are ideals of the ring R, then
1/(1 n J) ~ (1 + J)/J.
Proof Reasoning as in Theorem 3-9, we seek a homomorphism f from 1
(regarded a,<; a ring) onto the quotient ring (1 + J)jJ such that ker f = 1 n J.
Our candidate for this functionfis defined by declaring thatf(a) = a + J,
a E 1. A trivial, but useful, observation is that 1 S; 1 + J, whence f can
be obtained by composing the injection map il : 1 ~ I + J with the natural
mapping nat J : 1 + J ~ (1 + J)jJ. To be quite explicit, f = natJ i l or, 0

in diagrammatic language,
THE CLASSICAL ISOMORPHISM THEOREMS 47

From this factorization, it is easy to see that J is a homomorphism with

J(I) = (I + J)/1. To confirm that the kernel ofJis precisely the set I 11 J,
notice that the coset J serves as the zero element of (I + J)/1, and so
ker J = {a E IIf(a) = J}
= {aElla + J = J} = {aEllaEJ} = III J.
The asserted isomorphism should now be evident from the Fundamental
Homomorphism Theorem.
We conclude this chapter with a brief excursion into the theory of nil
and nilpotent ideals: a (right, left, two-sided) ideal I of the ring R is said to
be a nil ideal if each element x in I is nilpotent; that is to say, if there exists
a positive integer n for which xn = 0, where n depends upon the particular
element x. As one might expect, the ideal I will be termed nilpotent provided
r = {O} for some positive integer n. By definition, r denotes the set of all
finite sums of products of n elements taken from I, so that r = {O} is
equivalent to requiring that for every choice of n elements at, a2 , ••• , an E I
(distinct or not), the product a l a2 •.. an = 0; in particular, an = 0 for all
a in I, whence every nilpotent ideal is automatically a nil ideal. We speak
of the ring R as being nil (nilpotent) if it is nil (nilpotent) when regarded as
an ideal. Notice, too, that any ideal containing a nonzero idempotent element
cannot be nilpotent.
With these definitions at our disposal, we can now prove two lemmas.
Lemma. 1) If R is a nil (nilpotent) ring, then every subring and every
homomorphic image of R is nil (nilpotent).
2) If R contains an ideal I such that I and RjI are both nil (nilpotent),
then R is a nil (nilpotent) ring.
Proof The proof of assertion (1) follows immediately from the definitions
and Problem 2-17. To verify (2), assume that I and RjI are nil rings and
that a E R. Then there exists some positive integer n for which the coset
(a + I)n = an +I = I,
signifying thatthe element an E 1. Inasmuchaslisanilideal,(anr = anm = 0
for some m E Z +. This implies that a is nilpotent as a member of R and, in
consequence, R is a nil ring. The remainder of the proof is left to the
reader's care.
Lemma. If N land N 2 are two nil (nilpotent) ideals ofthe ring R, then
their sum N 1 + N 2 is likewise a nil (nilpotent) ideal.
Proof With reference to Theorem 3-10, we have (Nl + N 2)jN 1 ~
N 2j(N 1 11 N 2). The right-hand side (hence, the left-hand side) of this
equation is a nil ring, being the homomorphic image of the nil ideal N 2.
48

Since (N 1 + N 2)/N land N 1 are both nil, it follows from the previous lemma
that N 1 + N 2 is necessarily a nil ideal. Similar reasoning applies to the
nilpotent case.
Corollary. The sum of any finite number of nil (nilpotent) ideals of the
ring R is again nil (nilpotent).
Having completed the necessary preliminaries, let us now establish
Theorem 3-11. The sum L Ni of all the nil ideals Ni of the ring R is a
nil ideal.
Proof If the element a E L N i , then, by definition, a lies in some finite sum
of nil ideals of R; say, aEN l + N2 + ... + N n, where each Nk is nil.
By virtue of the last corollary, the sum Nl + N2 + ... + N n must be a
nil ideal; hence, the element a is nilpotent. This argument shows that L Ni
is a nil ideal.
It is possible to deduce somewhat more, namely,
Corollary. The sum of all the nilpotent ideals of the ring R is a nil
ideal.
Proof Since each nilpotent ideal is a nil ideal, the sum N of all nilpotent
ideals of R is contained in L N i , the sum of all nil ideals. But L Ni is itself
a nil ideal, making N nil.
Example 3-2. For examples of nilpotent ideals, let us turn to the rings
Zpn, where p is a fixed prime and n > 1. By virtue of the remarks on page
42, Zpn has exactly one ideal for each positive divisor of pn and no other
ideals; these are simply the principal ideals (pk) = pkZpn (0 ~ k ~ n). For
o < k ~ n, we have

so that each proper ideal of Z pn is nilpotent.

Before leaving this chapter, we should present an example to show that,
in general, nil and nilpotent are different concepts.
Example 3-3. For a fixed prime p, let S be the collection of sequences
a = {an} with the property that the nth term an E Zpn (n ~ 1). S can be
made into a ring by performing the operations of addition and multiplication
term by term:

The reader will find that the zero element of this ring is just the sequence
formed by the zero elements of the various Zpn and the negative of {an}
is {-an}. Now, consider the set R of all sequences in S which become zero
PROBLEMS 49

after a certain, but not fixed, point. One may easily check that R constitutes
a subring of the ring S (in fact, R is not only a subring, but actually an ideal
of S). It is in the ring R that we propose to construct our example of a
non-nilpotent nil ideal.
Let us denote by I the set of sequences in R whose nth term belongs to
the principal ideal in Z pn generated by p; in other words, the sequence
a E I if and only if it is of the form
a = (prl' pr2' ... , prn' 0, 0, .... )
A routine calculation confirms that I is an ideal of R. Since each term of
a is nilpotent in the appropriate ring, it follows that
an = °= (0, 0, 0, ...)
for n large enough, making I a nil ideal. (This also depends on the fact
that a has only a finite number of nonzero terms.)
At the present stage, it is still conceivable that I might be a nilpotent
ideal of R. However, we can show that for each positive integer n there
exist elements (sequences) a E I for which an =1= 0. For instance, define
°
a = {a k} by taking ak = P if k = 1, 2, ... ,n + 1 and ak = if k > n + 1;
that is,
a = (p, ... , p, p, 0, ... ) with n + 1 p's.
One then obtains
an = (0, ... , 0, pn, 0, ... ),
where all the terms are zero except the (n + 1)st, which is pn. Since pn is a
nonzero element of the ring Z pn + 1, the sequence an =1= 0, implying that
1" =1= {OJ. As this argument holds for any n E Z+, the ideal I cannot be
nilpotent.
We shall return to these ideas at the appropriate place in the sequel,
at which time their importance will become clear.

PROBLEMS

1. Let == be an equivalence relation on the ring R. We say that == is compatible

(with the ring operations) if and only if a == b implies a + c == b + c, ae == be,
ea == eb for all a, b, e E R. Prove that there is a one-to-one correspondence between
the ideals of R and the set of compatible equivalence relations on R.
2. If R is an arbitrary ring and n E Z+, prove that
a) the sets In = {nala E R} and I n = {a E RJna = O} are both ideals of R;
b) char (RlIn) divides n;
c) if char R =1= 0, then char R divides n char (R/ln).
50

3. Let I be an ideal of the ring R. Establish each of the following:

a) RjI has no divisors of zero if and only if ab E I implies that either a or b belongs
to I.
b) Rjl is commutative if and only if ab - ba E I for all a, b in R.
c) Rjl has an identity element if and only if there is some e E R such that
ae - a E I and ea - a E I for all a in R.
d) Whenever R is a commutative ring with identity, then so is the quotient ring
Rjl.
4. Let R be a commutative ring with identity and let N denote the set of all nilpotent
elements in R. Verify that
a) The set N forms an ideal of R. [Hint: If an = bm = 0 for integers n and m,
consider (a - b)n+m.J
b) The quotient ring RjN has no nonzero nilpotent elements.
5. Prove the following generalization of the Factorization Theorem: Let f1 and f2
be homomorphisms from the ring R onto the rings R1 and R 2, respectively. If
kerf1 ~ kerf2' then there exists a unique homomorphismJ: R1 -> R2 satisfying
f2 = J o f1· [Hint: Mimic the argument of Theorem 3-6; that is, for any element
f1(a) E R 1, defineJ(J1(a») = f2(a).J
6. Let I be an ideal of the ring R. Assume further that J and K are two subrings of
R with I ~ J, I ~ K. Show that
a) J ~ K if and only if natrJ ~ natrK
b) natr(J n K) = natrJ n natrK.
7. If I is an ideal of the ring R, prove that
a) Rjl is a simple ring if and only if there is no ideal J of R satisfying I c J c R;
b) if R is a principal ideal ring, then so is the quotient ring Rjl; in particular, Zn
is a principal ideal ring for each n E Z+. [Hint: Problem 17, Chapter 2.J
8. a) Given a homomorphism f from the ring R onto the ring R', show that
{f-1(b)lb E R'} constitutes a partition of R into the co sets of the ideal kerf
[Hint: If b = f(a), then the coset a + kerf = r1(b).J
b) Verify that (up to isomorphism) the only homomorphic images of the ring Z
of integers are the rings Zn' n > 0, and {O}.
9. Suppose that S is a subring and I an ideal of the ring R. If S n I = {O}, prove
that S is isomorphic to a subring of the quotient ring Rjl. [Hint: Utilize the
mappingf(a) = a + I, where a E S.J
10. A commutator in a ring R is defined to be any element of the form [a, bJ = ab - ba.
The commutator ideal of R, denoted by [R, RJ, is the ideal of R generated by the
set of all commutators. Prove that
a) R is a commutative ring if and only if [R, RJ = {O} (in a sense, the size of [R, RJ
provides a measure of the noncommutativity of R);
b) for an ideal I of R, the quotient ring Rjl is commutative if and only if
[R, RJ ~ I.
11. Assuming that f is a homomorphism from the ring R onto the commutative ring
R', establish the assertions below:
PROBLEMS 51

a) [R, R] £; kerf;
b) f = Jo nat[R RI' whereJis the induced mapping;
c) if ker f £; [R, R], then R/[R, R] ~ R'/[R', R'J.
12. a) Suppose that 11 and 12 are ideals of the ring R for which R = 11 EB 12. Prove
that R/I1 ~ 12, and R/I2 ~ 1 1.
b) Let R be the direct sum of the rings Ri (i = 1,2, ... , n). If Ii is an ideal of Ri
and I = 11 EB 12 EB ... EB In' show that

R/I ~ (R 1/I 1) EB (R 2/I 2)EB ... EB (Rn/In)·

[Hint: Find the kernel of the homomorphismJ: R -> 1: EB (RjIJ that sends
a = (a 1,a2, ... ,an) tof(a) = (a 1 + I 1,a2 + 12, ... ,an + In)·]
13. For a proof of Theorem 3-9 that does not depend on the Fundamental Homo-
morphism Theorem, define the function h: R/I -> R'/f(I) by taking h(a + I) =
f(a) + f(I).
a) Show that h is a well-defined isomorphism onto R'/f(I); hence, R/I ~ R'/f(I).
b) Establish that h is the unique mapping that makes the diagram below
commutative:

R L R ' =f(R)

natil 1natf(Il
R/I h R'/ f(I)

14. Given integers m, n E Z +, establish that

a) if m divides n, then Znj(m)!(n) ~ Zm;
b) if m and n are relatively prime, then Zmn ~ Zm EB Zn·
15. If I is an ideal of the ring R, prove that the matrix ring Mn(R/I) is isomorphic to
Mn(R)!Mn(I). [Hint: Consider the mapping f: Mn(R) -> Mn(R/I) defined by
f((a i)) = (aij + I).]
16. Let R be a ring without divisors of zero. Imbed R in the ring R' = R x Z, as
described in Theorem 2-12. (The case R = Ze illustrates that R' may contain
zero divisors even though R d~es not.) Assuming that I denotes the left annihilator
of R in R',
1= {aER'lar = 0 for all rE R},

verify that
a) I forms an ideal of R'. [Hint: R is an ideal of R'.]
b) R'/I is a ring with identity which has no divisors of zero.
c) R'/I contains a subring isomorphic to R. [Hint: Utilize Problem 9.]
FOUR

INTEGRAL DOMAINS AND FIELDS

In the preceding chapters a hierarchy of special rings has been established

by imposing more and more restrictions on the multiplicative semigroup
of a ring. At first glance, one might be tempted to require that the multi-
plicative semigroup actually be a group; such an assumption would be far
too demanding in that this situation can only take place in the trivial ring
consisting of zero alone. A less stringent condition would be the following:
the nonzero elements comprise a group under multiplication. This leads
to the notion of a field.

Definition 4-1. A ring F is said to be a field provided that the set

F - {O} is a commutative group under the multiplication of F (the
identity of this group will be written as 1).
Definition 4-1 implicitly assumes that any field F contains at least one
element different from zero, for F - {O} must be nonempty, serving as the
set of elements of a group. It is also to be remarked that, since aO = 0 = Oa
for any a E F, all the members of F commute under multiplication and not
merely the nonzero elements. Similarly, the relation 10 = 0 = 01 implies
that 1 is the identity for the entire ring F. Viewed otherwise: a field is a
commutative ring with identity in which each nonzero element possesses
an inverse under multiplication.
Occasionally, we shall find it convenient to drop the requirement of
commutativity in the consideration of a field, in which case the resulting
system is called a division ring or skew field. That is to say, a ring is a
division ring if its nonzero elements form a group (not necessarily com-
mutative) with respect to multiplication.
After this preamble, let us look at several examples.

Example 4-1. Here are some of the more standard illustrations of fields:
the set Q of all rational numbers, the set F = {a + bJ2la, b E Q}, and
the set R # of all real numbers. In each case the operations are ordinary
addition and multiplication.
52
INTEGRAL DOMAINS AND FIELDS 53

Example 4-2. Consider the set C = R # X R # of ordered pairs of real

numbers. To turn C into a field, we define addition and multiplication by
(a, b) + (e, d) = (a + e, b + d),
(a, b)(e, d) = (ae - bd, ad + be).
The reader may verify without difficulty that C, together with these opera-
tions, is a commutative ring with identity. In this setting, the pair (1, 0)
serves as the multiplicative identity, and (0, 0) is the zero element ofthe ring.
Given any nonzero element (a, b) of C, either a =1= 0 or b =1= 0, so that
a 2 + b2 > 0; thus,

exists in C and has the property that

This shows that each nonzero member of C has an inverse under multi-
plication, thereby proving the system C to be a field.
It is worth pointing out that the field C contains a subring isomorphic
to the field of real numbers. For, if
R# x {O} = {(a, O)la E R#},
it follows that R# ~ R# X {O} via the mapping f defined by f(a) = (a, 0).
Inasmuch as the distinction between these systems is only one of notation,
we customarily identify the real number a with the corresponding ordered
pair (a, 0); in this sense, R # may be regarded as a subring of C.
Now, the definition of the operations in C enables us to express an
arbitrary element (a, b) E C as
(a, b) = (a, 0) '+ (b,O)(O, 1),
where the pair (0, 1) is such that (0, 1)2 = (0, 1)(0, 1) = (-1,0). Introducing
the symbol i as an abbreviation for (0, 1), we have
(a, b) = (a, 0) + (b,O)i.
Finally, if it is agreed to replace pairs of the form (a, 0) by the first com-
ponent a (this is justified by the preceding paragraph), the displayed
representation becomes
(a, b) = a + bi, with
In other words, the field C as defined initially is nothing more than a
disguised version of the familiar complex number system.
54

Example 4-3. For an illustration of a division ring which is not a field, we

turn to the ring of (Hamilton's) real quaternions. To introduce this ring, let
the set H consist of all ordered 4-tuples of real numbers:
H = {(a, b, c, d)la, b, c, dE R#}.
Addition and multiplication of the elements of H are defined by the rules
(a, b, c, d) + (a', b', c', d') = (a + a', b + b', c + c', d + d'),
(a, b, c, d)(a', b', c', d') = (aa' - bb' - cc' - dd', ab' + ba' + cd' - dc',
ac' - bd' + ca' + db', ad' + be' - cb' + da').
A certain amount of tedious, but nonetheless straightforward, calculation
shows that the resulting system is a ring (known as the ring of real quater-
nions) in which (0, 0, 0, 0) and (1, 0, 0, 0) act as the zero and identity elements,
respectively.
Let us next introduce some special symbols by putting
1 = (1, 0, 0, 0), i = (0, 1,0, 0), j = (0, 0, 0, 1, 0), k = (0, 0, 0, 1).
The elements 1, i, j, k have a number of distinctive properties; specifically,
1 is the mUltiplicative identity of Hand
i2 = j2 = k2 = -1,
ij = k, jk = i, ki = j, ji = - k, kj = - i, ik = - j.
These relations demonstrate that the commutative law for multiplication
fails to hold in H, so that H definitely falls short of being a field.
As in Example 4-2, the definition of the algebraic operations in H
permits us to write each quaternion in the form
(a, b, c, d) = (a, 0, 0, 0)1 + (b, 0, 0, O)i + (c, 0, 0, O)j + (d, 0, 0, O)k
Since the subring {(r, 0, 0, O)lr E R#} is isomorphic to R#, the notation can
be further simplified on replacing (r, 0, 0, 0) by the element r itself; adopting
these conventions, the real quaternions may henceforth be regarded as the
set
H = {a + bi + cj + dkla, b, c, dE R#},
with addition and multiplication performed as for polynomials (subject to
the rules of the last paragraph). The reader versed in linear algebra should
recognize that H comprises a four-dimensional vector space over R# having
{I, i, j, k} as a basis.
The main point in our investigation is that any nonzero quaternion
q = a + bi + cj + dk (in other words, one of a, b, c, d must be different
from zero) is a multiplicatively invertible element. By analogy with the
complex numbers, each quaternion has a conjugate, defined as follows:
q = a - bi - cj - dk.
INTEGRAL DOMAINS AND FIELDS 55

It is easily verified that the product

qq = qq = a2 - (bi + cj + dk)2 = a2 + b2 + c 2 + d2 =f 0,
thus exhibiting that q has the multiplicative inverse
q-1 = (a 2 + b2 + c2 + d2)-lq.
Incidentally, the totality of all members of H of the form (a, b, 0, 0) =
a + bi, the special quaternions, constitute a subring isomorphic to C; as
substitutes, one might also consider the set of all elements (a, 0, b, 0) or all
elements (a, 0, 0, b). In this light, the real quaternions may be viewed as a
suitable generalization of the complex numbers.
The following theorem shows that any field is without divisors of zero,
and consequently a system in which the cancellation law for mUltiplication
holds.
Theorem 4-1. Every field F is an integral domain.
Proof Since every field is a commutative ring with identity, we need only
prove that F contains no zero divisors. To this purpose, suppose a, bE F,
with ab = 0. If the element a =f 0, then it must possess a multiplicative
inverse a -1 E F. But then the hypothesis that ab = yields °
as desired.
There obviously exist integral domains which are not fields; a prime
example is the ring Z of integers. However, an integral domain having only
a finite number of elements must necessarily be a field.
Theorem 4-2. Any integral domain R with only a finite number of ideals
is a field.
Proof Let a be any nonzero element of R. Consider the set of principal
ideals (an), where n E Z + :
(an) = {r anlr E R}.
Since R has only a finite number of distinct ideals, it follows that (am) = (an)
for certain positive integers m, n with m < n. Now, am, as an element of
(am), must lie in (an). This being so, there exists some r E R for which
am = r an. By use of the cancellation law,
1= r an- m = (1' an- m- 1)a.
Because multiplication is commutative, we therefore have a- 1 = r an - m - 1 .
This argument shows that every nonzero element of R is invertible; hence,
R forms a field.
56

Corollary. Any finite integral domain is a field.

Because the aforementioned corollary is so basic a result, we offer a
second proof. The "counting argument" involved in this latter proof adapts
to a variety of situations in which the underlying ring is finite.
The reasoning proceeds as follows. Suppose that aI' a2 , ... , an are the
members of the integral domain R. For a fixed nonzero element a E R, we
consider the n products aa l , aa 2 , ... , aan. These products are all distinct,
for if aa i = aaj , the cancellation law (valid in any integral domain) would
yield ai = aj . It follows that each element of R must be of the form aa i
for some choice of i. In particular, there exists some ai E R such that aai = 1.
From the commutativity of multiplication, we infer that a-I = ai' whence
every nonzero element of R possesses a multiplicative inverse.
There are no finite division rings which are not fields. To put it another
way, in a finite system in which all the field properties except the commuta-
tivity of multiplication are assumed, the multiplication must also be com-
mutative. Proving this renowned result is far from being as elementary as
the case of a finite integral domain and is deferred until Chapter 9.
For the moment, let us take a closer look at the multiplication structure
of Zn" It has been previously shown that, for each positive integer n, Zn
comprises a commutative ring with identity. A reasonable question is: For
precisely what values of n, if any at all, will this ring turn out to be a field?
For a quick answer: n must be a prime number. (What could be simpler
or more natural?) This fact is brought out by the coming theorem.
Theorem 4-3. A nonzero element [a] E Zn is invertible in the ring
Zn if and only if a and n are relatively prime integers (in the sense that
gcd(a, n) = 1).
Proof If a and n are relatively prime, then there exist integers rand s such
that ar + ns = 1. This implies that

[1] = Car + ns] = Car] +n ens]

= Car] +n [0] = [a] ·n [r],

showing the congruence class [r] to be the multiplicative inverse of [a J.

Now to the "only if" part. Assume [a] to be a multiplicatively invertible
element of Zn; say, with inverse [bJ. We thus have [ab] = [a]·n [b] = [1],
so that there exists an integer k for which ab - 1 = kn. But then
ab + n( - k) = 1; hence, a and n are relatively prime integers.
Corollary. The zero divisors of Zn are precisely the nonzero elements
of Zn which are not invertible.
Proof Naturally, no zero divisor of Zn can possess a multiplicative inverse.
On the other hand, suppose that [a] 9= [0] is not invertible in Zn' so that
INTEGRAL DOMAINS AND FIELDS 57

gcd(a, n) = d, where 1 < d < n. Then, a = rd and n = sd for suitable

nonzero integers rand s. This leads to
[a] ·n [s] = [as] = [rds] = Ern] = [OJ.
Since the defining properties of s rule out the possibility that [s] = [0],
it follows that [a] is a zero divisor of Zn.
These results may be conveniently summarized in the following state-
ment.
Theorem 4-4. The ring Zn of integers modulo n is a field if and only if
n is a prime number. If n is composite, then Zn is not an integral domain
and the zero divisors of Zn are those nonzero elements [a] for which
gcd(a, n) 9= l.
Every field necessarily has at least two elements (1 being different from
0); Theorem 4-4 indicates that there is a field having this minimum number
as its number of elements, viz. Z2.
As an interesting application of these ideas, consider the following
assertion: If there exists a homomorphism f: Z -+ F of the ring Z of integers
onto a field F, then F is necessarily a finite field with a prime number of
elements. For, by the Fundamental Homomorphism Theorem, Z/ker f ~ F.
But ker f = (n) for some positive integer n, since Z is a principal ideal domain.
(In this connection, observe that n 9= 0, for otherwise Z would be isomorphic
to a field, an impossibility.) Taking stock of the fact that Z/(n) = Zn' we are
thus able to conclude that Zn ~ F, in consequence of which F has n
elements. At this point Theorem 4-4 comes to our aid; since F, and in turn
its isomorphic image Zn' forms a field, n must be a prime number.
A useful counting function is the so-called Euler phi-function (totient),
defined as follows: 4>(1) = 1 and, for each integer n > 1, 4>(n) is the number
of invertible elements in the ring Zn. By virtue of Theorem 4-3, 4>(n) may
also be characterized as the number of positive integers < n which are
relatively prime to n. For instance, 4>(6) = 2, 4>(9) = 6, and 4>(12) = 4; it
should be equally clear that whenever p is a prime number, then 4>(p) = p - l.
Lemma. If Gn is the subset of Zn defined by
Gn = {[a] E Znla is relatively prime to n},
then (G n, ·n) forms a finite group of order 4>(n).
Proof In the light of the preceding remarks, (G n , ·n) is simply the group of
invertible elements of Zn.
This leads at once to a classical result of Euler concerning the phi-
function; the simplicity of the argument illustrates the advantage of the
algebraic approach to number theory.
58

Theorem 4-5. (Euler-Fermat). Ifn is a positive integer and a is relatively

prime to n, then a</>(n) == 1 (mod n).
Proof The congruence class [a] can be viewed as an element of the multi-
plicative group (G n, ·n). Since this group has order 4>(n), it follows that
[a ] </>(n) = [1] or, equivalently, a</>(n) == 1 (mod. n). (Recall that if G is a finite
group of order k, then Xk = 1 for all x E G.)
There is an interesting relationship between fields and the lack of ideals;
what we shall show is that fields have as trivial an ideal structure as possible.
Theorem 4-6. Let R be a commutative ring with identity. Then R is a
field if and only if R has no nontrivial ideals.
Proof Assume first that R is a field. We wish to show that the trivial ideals
{O} and R are its only ideals. Let us suppose to the contrary that there
exists some nontrivial ideal I of R. By our assumption, the subset I is such
that I =1= {O} and I =1= R. This means that there exists some nonzero element
a E 1. Since R is taken to be a field, a has a multiplicative inverse a- 1 present
in R. By the definition of ideal, we thus obtain 1 = a -1 a E I, which in turn
implies that I = R, contradicting our choice of 1.
Conversely, suppose that the ring R has no nontrivial ideals. Given a
nonzero element a E R, consider the principal ideal (a) generated by a:
(a) = {ralr E R}.
Now, (a) cannot be the zero ideal, inasmuch as a = alE (a), with a =1= O.
It follows from the hypothesis that the only other possibility is that (a) = R.
In particular, since 1 E (a), there exists an element r E R for which r a = 1.
Multiplication is commutative, so that r = a- 1 . Therefore, each nonzero
element of R is multiplicatively invertible and we are done.
In view of this last result, the ring Z of integers fails to be a field since
it contains the nontrivial ideal Ze'
Theorem ~ is useful in revealing the nature of homomorphisms
between fields. We exploit it to prove
Theorem 4-7. Let f be a homomorphism from the field F into the
field F'. Then eitherfis the trivial homomorphism or elsefis one-to-one:
Proof The proof consists of noticing that since ker f is an ideal of the field
F, either ker f
= {O} or else ker f = F. The condition ker f = {O} implies
that f is a one-to-one function. On the other hand, if it happens that
ker f = F, then each element of F is carried onto 0; that is to say, f is the
trivial homomorphism.
Corollary. Any homomorphism of a field F onto itself is an auto-
morphism of F.
INTEGRAL DOMAINS AND FIELDS 59

Any ring with identity which is a subring of a field must of necessity be

an integral domain. Turning the situation around, one might ask whether
each integral domain can be considered (apart from isomorphism) as a
subring of some field. More formally: Can a given integral domain be
imbedded in a field? In the finite case there is plainly no difficulty, since
any finite integral domain already forms a field.
Our concern with this question arises from the natural desire to solve
the linear equation ax = b, where a 1= o. A major drawback to the notion
of an integral domain is that it does not always furnish a solution within
the system. (Of course, any such solution would have to be unique, since
aX l = b = aX 2 implies that Xl = X 2 by the cancellation law.) It hardly
seems necessary to point out that when the integral domain happens to be
a field, the equation ax = b (a 1= 0) is always solvable, for one need only
take x = a-lb.
We begin our discussion of this problem with an obvious definition.

Definition 4-2. By a subfield of a field F is meant any subring F' of F

which is itself a field.

For example, the ring Q of rational numbers is a subfield of the real

field R # ; the same is true of the field F = {a + bJ2/ a, b E Q}.
Surely, the set F' will be a subfield of the field F provided that (1) F' is a
subgroup of the additive group of F and that (2) F' - {O} is a subgroup of
the multiplicative group F - {O}. Recalling our minimal set of conditions
for determining subgroups (see page 8), it follows that F' will be a subfield
of F if and only if the following requirements are met:
1) F' is a nonempty subset of F containing at least one nonzero element,
2) a, b E F' imply a - b E F', and
3) a, b E F', with b 1= 0, imply ab- l E F'.
The coming theorem furnishes a clue to the nature of the field in which
we wish to imbed a given integral domain.

Theorem 4-8. Let the integral domain R be a subring of the field F.

If the set F' is defined by

F' = {ab-l/a, b E R; b 1= O},

then F' forms a subfield of F with R £ F'; in fact, F' is the smallest
(in the sense of inclusion) subfield of F containing R.
Proof Notice first that 1 = 11 -1 E F', so that F' 1= {O}. Now consider two
arbitrary elements x, y of F'. With reference to the definition of F', we then
have
60

for a suitable choice of a, b, c, d E R, where b 1- 0, d 1- o. A simple

computation shows that
x - y = (ad - bc)(bd)-1 E F'.

Also, if y is nonzero (that is, whenever c 1- 0), we conclude that

xy-1 = (ad)(cb)-1 E F'.

By virtue of the remarks following Definition 4-2, this is sufficient to

establish that the set F' is a subfield of F. Furthermore,
a = al = al - 1 E F'

for each a in R, implying that R £ F'. Finally, any subfield of F which

contains R necessarily includes all products ab -1, with a, 0 1- b E R, and,
hence, contains F'.
Theorem 4-8 began with an integral domain already imbedded in a
field. In the general case it becomes necessary to construct the imbedding
field. Since the expression ab -1 may not always exist, one must now work
with ordered pairs (a, b), where b 1- O. Our thinking is that (a, b) will play
a role analogous to ab - 1 in the foregoing theorem.
Actually, the proposed construction is not just confined to integral
domains, but will apply to a much wider class of rings; it will imbed any
commutative ring R that contains a (nonempty) set of elements that are
not zero divisors in a ring Qcl(R), which may be described as follows.

Definition 4-3. Let R be a ring with at least one non-zero-divisor. A

classical ring of quotients of R is any ring Qcl(R) satisfying the conditions
l) R £ Qcl(R),
2) every element of Qcl(R) has the form ab -1, where a, b E Rand b is
a non-zero-divisor of R, and
3) every non-zero-divisor of R is invertible in Qcl(R).

Convention. An element a E R is termed a non-zero-divisor if ar 1- 0, and

ra 1- 0 for all 0 1- r E R; in particular, the phrase "non-zero-divisor"
excludes the zero element.
As a starting point, let S denote the set of all elements of R, a commutative
ring, which are non-zero-divisors; we will assume hereafter that S 1- 0.
Needless to say, if there happens to be an identity element 1 available, then
1 E S. Notice, too, that the set S is closed under multiplication. For,
suppose that the elements S1' S2 E Sand (s1s2)a = O. Then S1(S2a) = 0
and, since S1 is not a divisor of zero, it follows that S2a = 0; this in turn
implies that a = O. Therefore, the product S1S2 is not a zero divisor of R,
whence S1S2 E S.
INTEGRAL DOMAINS AND FIELDS 61

Now consider the set of ordered pairs

R x S = {(a, s)la E R, s E S}.
A relation '" may be introduced in R x S by taking
(a, s) '" (b, r) if and only if ar = bs.
(We have in mind the previous theorem, where ab- 1 = cd- 1 if and only
if ad = bc.)
It is not difficult to verify that the relation "', thus defined, is an
equivalence relation in R x S. The transitive property is perhaps the least
obvious. To see this, assume that (a, s) '" (b, r) and (b, r) '" (c, t), so that
ar = sb, bt = rc.
Now multiply the first equation by t and the second by s to get
art = sbt, sbt = src.
Putting these relations together, we obtain atr = scr. Since r is not a zero
divisor, the cancellation law gives us at = sc, which is exactly the condition
that (a, s) '" (c, t).
Next, we label those elements of R x S which are equivalent to the pair
(a, s) by the symbol a/s; in other words,
a/s = reb, r)l(a, s) '" (b, r)}
= {(b, r) ar = sb}.

The collection of all equivalence classes a/s relative to '" will be denoted
by Qc)(R):
Qc)(R) = {a/sla E R; s E S}.
From Theorem A-1, we know that the elements of Qc)(R) constitute a
partition of the set R x S. That is, the ordered pairs in R x S fall into
disjoint classes (calledformalfractions), with each class consisting ofequiva-
lent pairs, and nonequivalent pairs belonging to different classes. Further,
two such classes a/s and b/r are identical if and only if ar = sb; in particular,
all fractions of the form as/s, with s E S, are equal.
With these remarks in mind, let us introduce the operations of addition
and multiplication required to make Qc)(R) into a ring. We do this by
means of the formulas
a/s + b/r = (ar + sb)/sr,
(a/s)(b/r) = ab/sr.
Notice, incidentally, that since the set S is closed under multiplication,
the right-hand sides of the defining equations are meaningful.
As usual, our first task is to justify that these operations are well-defined;
62

that is to say, it is necessary to show that the sum and product are independent
of the particular elements of R used in their definition. Let us present the
argument for addition in detail. Suppose, then, that a/s = a'/s' and
b/r = b'/r'; we must show that
(ar + sb)/sr = (a'r' + s'b')/s'r'.
From what is given, it follows at once that
as' = sa', br' = rb'.
These equations imply
(ar + sb)(s'r') - (a'r' + s'b')(sr) = (as' - sa')(rr') + (br' - rb')(ss')
= O(rr') + O(ss') = O.
By the definition of equality of equivalence classes, this amounts to saying
that
(ar + sb)/sr = (a'r' + s'b')/s'r',

which proves addition to be well-defined. In much the same way, one can
establish that
ab/sr = a'b'/s'r'.

The next lemma reveals the algebraic nature of Qc1(R) under these
operations.
,
Lemma. The system Qcl(R) forms a commutative ring with identity.
Proof It is an entirely straightforward matter to confirm that Qcl(R) is a
commutative ring. We leave the reader to make the necessary verifications
at his leisure, and merely point out that O/s serves as the zero element, while
-a/s is the negative of a/so
That the equivalence class s'/s', where s' is any fixed non-zero-divisor
of R, constitutes the multiplicative identity is evidenced by the following
computation:
(a/s)(s'/s') = as'/ss' = a/s
for arbitrary a/s in Qcl(R), since (as')s = (ss')a. Loosely speaking, common
factors belonging to S may be cancelled in a fraction as'/ss'.
This proves part of the theorem below.
Theorem 4-9. Any commutative ring R with at least one non-zero-
divisor possesses a classical ring of quotients.
Proof We begin by establishing that the ring Qcl(R) contains a subring
isomorphic to R. For this, consider the subset K of Qcl(R) consisting of all
INTEGRAL DOMAINS AND FIELDS 63

elements of the form aso/s o, where So is a fixed non-zero-divisor of R (recall

that the equivalence class aso/s o depends only upon a, not upon the choice
of so):

The reader can easily check that K is a subring of Qcl(R). An obvious (onto)
mappingf: R ~ K is defined by takingf(a) = aso/s o. Since the condition
aso/s o = bso/so implies that as~ = bs~ or, after cancelling, that a = b, f
will be a one-to-one function. Furthermore, it has the property of pre-
serving both addition and multiplication:

f(a + b) = (a + b)so/so = aso/s o + bso/so = f(a) + f(b)

f(ab) = (ab)so/so = (ab)sUs~ = (aso/so)(bso/s o) = f(a)f(b).

In this way, R can be isomorphically embedded in ~l(R).

By identifying R with K, we may henceforth regard R as actually being
contained in Qcl(R). In practice, one simply replaces the fraction aso/so E K
by the corresponding element a E R.
We proceed to show that all the elements of S are invertible in Qcl(R).
Any non-zero-divisor s E S has, after identification, the form sSo/so. Now,
the equivalence class so/sso is also a member of Qcl(R) (note the crucial
use of the closure of S under multiplication) and satisfies the equation
(sso/so)(so/sso) = ss~/ss~ = so/so.
Since so/so plays the role of the identity element for Qcl(R), we see at once
that (ssO/SO)-l = so/sso.
All that remains to complete the proof is to verify that each member
a/s of Qcl(R) can be written as as-i. It should be clear that
a/s = (aso/so)(so/sso) = (as o/s o)(sso/so)-i.
Replacing aso/s o by a and sSo/so by s, the displayed equation assumes the
·more familiar form a/s = as-i. The point is this: the set Qc1(R) may now be
interpreted as consisting of all quotients as- \ where a E R, s E S.
Thus, Qcl(R) satisfies Definition 4--3 in its entirety, thereby becoming a
classical ring of quotients of R.
Two comments are in order. In the first place, given any element s E R
which is a non-zero-divisor, it follows that
(so E S).
Identifying sSo/so with sand aso/s o with a, we conclude from this that the
equation sx = a always possesses a solution in Qcl(R), namely, x = a/s =
as-i. Second, notice that in Qcl(R) multiplicative inverses exist not only for
64

members of S but for all elements of Qcl(R) which can be represented in the
form r/s, where r, s are both non-zero-divisors; in fact,
=
rs/sr = so/so.
(r/s)(s/r)
When the ring R is an integral domain, we may take the set S of non-zero-
divisors as consisting of all the elements of R which are not zero. The last
remark of the preceding paragraph then leads to the following important
theorem.
Theorem 4-10. For any integral domain R, the system Qcl(R) forms a
field, customarily known as the field of quotients of R.
Since an integral domain is (isomorphic to) a subring of its field of
quotients, we also obtain
Corollary. A ring is an integral domain if and only if it is a subring of
a field.
It should be pointed out that the hypothesis of commutativity is essential
to this last theorem; indeed, there exist noncommutative rings without
divisors of zero that cannot be imbedded in any division ring.
The field of quotients constructed from the integral domain Z is, of
course, the rational number field Q. Another fact of interest is that the field
of quotients is the smallest field in which an integral domain R can be
imbedded, in the sense that any field in which R is imbeddable contains a
subfield isomorphic to Qc](R) (Problem 20).
The existence theorem for the classical ring of quotients can be supple-
mented by the following result, which shows that it is essentially unique.
Theorem 4-11. Let Rand R' be two commutative rings, each containing
at least one non-zero-devisor. Then, any isomorphism of R onto R' has a
unique extension to an isomorphism of Qc](R) onto Qc1(R').
Proof To begin with, each member of Qc](R) may be written in the form
ab -1, where a, b E Rand b is a non-zero-divisor in R. Given an isomorphism
¢: R -+ R', the element ¢(b) will be a non-zero-divisor of R', so that ¢(b)-1
is present in Qc](R'). Suppose that ¢ admits an extension to an isomorphism
«1>: Qcl(R) -+ Qc1(R'). Since a = (ab- 1)b, we would then have
¢(a) = «I>(a) = «I>(ab- 1)«I>(b) = «I>(ab- 1)¢(b),
which, as a result, yields «I>(ab- 1) = ¢(a)¢(b)-1. Thus, «I>
is completely
determined by the effect of ¢ on R and so determined uniquely, if it exists
at all.
These remarks suggest that, in attempting to extend ¢, we should consider
the assignment:
for all
INTEGRAL DOMAINS AND FIELDS 65

For a verification that <I> is a well-defined function, let ab -1 = ed- 1 in

Qcl(R); that is to say, ad = be in R. Then, the equation ¢(a)¢(d) = ¢(b)¢(e)
holds in R' £ Qcl(R') or, viewed otherwise, ¢(a)¢(b)-l = ¢(e)¢(d)-l. But
this means <I>(ab- 1 ) = <I>(ed- 1 ), so that <I> does not depend on how an
element in Qcl(R) is expressed as a quotient.
One verifies routinely that <1>, as defined above, is a homomorphism of
Qcl(R) into Qcl(R'). This homomorphism certainly extends ¢; indeed, if a
is an arbitrary element of Rand b is a non-zero-divisor of R, <I> maps
a = (db)b-1.irito
<I>(a) = ¢(ab)¢(b)-l = ¢(a)¢(b)¢(b)-l = ¢(a).
To see that <I> is a one-to-one function, we examine its kernel. Now, if
<I>(ab -1) = 0, then ¢(a) = 0. But, ¢ being an isomorphism, this implies
that a = 0, whence ab- 1 = 0. Accordingly, ker<l> = {O}, which forces <I>
to be one-to-one. Without going into the details, we also point out that <I>
carries Qcl(R) onto Qcl(R') (this stems from the fact that ¢ maps onto R').
Therefore, <I> is the desired extension of ¢.
A special case of particular importance occurs when Rand R' are the
same ring and ¢ is taken to be the identity isomorphism on R.
Corollary. Any two quotient rings of a commutative ring R with at
least one non-zero-divisor are isomorphic by a unique mapping fixing
all the elements of R.
At this point we leave the theory of quotients and turn to prime fields.
Clearly, any field F has at least one subfield, namely, F itself; a field which
does not possess any proper subfields is called a prime field.
Example 4-4. The field Q of rational numbers is the simplest example of
a prime field. To see this, suppose that F is any subfield of Q and let a E F
be any nonzero element. Since F is a subfield of Q, it must contain the
product aa -1 ~ 1. In turn, n = nl E F for any n in Z; in other words,
F contains all the integers. It then follows that every rational number
n/m = nm- 1 (m =1= 0) also lies in F, so that F = Q.
Example 4-5. For each prime p, the field Zp of integers modulo p is a prime
field. The reasoning here depends on the fact that the additive group
(Zp, +p) is a finite group of prime order and therefore by Lagrange's theorem
has no non-trivial subgroups.
An observation which will not detain us long is that each field F contains
a unique prime subfield. To make things more specific, let {Fi} be the
collection of all subfields of F. Then the intersection n Fi is also a subfield
of F. Now, if F' is any subfield of the field n F i, then F' E {F i }, whence
n Fi £ F'; the implication is that F' = n F i , forcing n Fi to be a prime
66

field. As regards the uniqueness assertion, suppose that K 1 and K2 are

both prime subfields of F. Then Kl n K2 is a subfield of F as well as K 1 ,
with Kl ;2 Kl n K 2. But Kl can possess no proper subfields, which
signifiesthatK l = Kl n K 2· Likewise,K 2 = Kl n K 2,whenceK l = K 2.
We conclude this chapter by showing that, to within isomorphism, the
rational number field and the fields Zp are the only prime fields.
Theorem 4-12. Any prime field F is isomorphic either to Q, the field
of rational numbers, or to one of the fields Zp of integers modulo a
prime p.
Proof To begin, let 1 be the multiplicative identity of F and define the
mapping f: Z -+ F by fin) = n1 for any integer n. Then f is a homo-
morphism from Z onto the subring Zl of integral multiples of 1. In com-
pliance with Theorem 3-7, we therefore have Z/ker f ~ Z1. But kerf is
an ideal of Z, a principal ideal domain, whence kerf = (n) for some
nonnegative integer n. The possibility that n = 1 can be ruled out, for
otherwise 1 = f(l) = 0 or, what amounts to the same thing, F = {O}.
Notice further that if n =1= 0, then n must in fact be a prime number.
Suppose to the contrary that n = n 1n 2, where 1 < n i < n. Since n E ker f,
it follows that
(nl1)(n21) = (n 1n2)1 = n1 = 0,
yielding the contradiction that the field F has divisors of zero. (This result
is not entirely unexpected, because the integer n is the characteristic of F
and as such must be a prime, whenever n =1= 0.)
The preceding discussion indicates that two possibilities arise: either
1) Zl ~ Z/(P) = Zp for some prime p, or
2) Zl ~ Z/(O) = Z.
Turning to a closer analysis of these cases, assume first that Zl ~ Zp,
with p prime. Inasmuch as the ring of integers modulo a prime forms a
field, the subring Zl must itself be a field. But F, being a prime field,
contains no proper subfields. Accordingly, Zl = F, which leads to the
isomorphism F ~ Z p'
For the final stage of the proof, consider the situation where Zl ~ Z.
Under these circumstances, the subring Zl is an integral domain, but not a
field. Taking stock of Theorem 4-8, as well as the hypothesis that F is a
prime field, we conclude that
F = {ab-1Ia, b E Zl; b =1= O}
= {(n1)(ml)-1In,mEZ; m =1= O}.
It is now a purely routine matter to verify that the fields F and Q are iso-
morphic under the mapping g(n/m) = (n1)(m1)-1; we leave the details as
an exercise.
PROBLEMS 67

Since every field contains a unique prime subfield, the following sub-
sidiary result is of interest.
Corollary 1. Every field contains a subfield which is isomorphic either
to the field Q or to one of the fields Zp.
Theorem 4-12 also provides some information regarding field auto-
morphisms.
Corollary 2. If f is an automorphism of the field F, then f(a) = a for
each element a in the prime subfield .of F (hence, a prime field has no
automorphism except the identity).
Proof The prime subfield of F is either
Fl = {(n1)(m1)-1In,mEZ; m =1= O}
or
F2 = {n1ln = 0,1, ... ,p - 1},
according as the characteristic of F is 0 or a prime p. Since any automor-
phism of a field carries the identity 1 onto itself, the result should be clear.

PROBLEMS
1. a) Assuming that R is a division ring, show that cent R forms a field.
b) Prove that every subring, with identity, of a field is an integral domain.
2. Let R be an integral domain and consider the set Z1 of all integral multiples of
the identity element:
Z1 = {n11 n E Z}.

Establish that Z1 is a field if and only if R has positive characteristic.

3. In the field C, define a mapping f: C -+ C by sending each complex number to its
conjugate; that is, f(a + bi) = a - bi. Verify that f is an automorphism of C.
4. Find the center of the quatemion ring H.
5. Let R be the subring of M 2 (C) consisting of all matrices of the form

(-7Ja P)
a = (a + bi
- c + di
c + di)
a - bi
(a, b, c, dE R#).

Prove that R is a division ring isomorphic to the division ring of real quatemions.
6. By the quaternions over a field F is meant the set of all q = a + bi + cj + dk,
where a, b, c, d E F and where addition and mUltiplication are carried out as with
real quatemions. Given that F is a field in which a 2 + b2 + c2 + d2 = 0 if and
only if a = b = c = d = 0, establish that the quatemions over F form a division
ring.
68

7. Establish the following facts concerning the Euler phi-function:

a) If nand m are relatively prime integers, then </>(nm) = </>(n)</>(m).
b) For any prime p and n > 0, </>(pn) = pn(1 - lip) = pn - pn-I. [Hint: The
integers k such that 0 < k < pn and gcd (k, pn) =1= 1 are p, 2p, ... , pn-Ip.]
c) If PI' P2' ... ,Pk are the distinct prime divisors of an integer n > 1, then
</>(n) = n(1 - Ilpd(1 - Ilp2) ... (1 - Ilpd.
d) n = Lain </>(a).

8. Let -r(n) denote the number of (distinct) positive divisors of an integer n > 1.
Prove that
a) If n has the prime factorization n = p~1P22 ... pZ", where the p; are distinct
primes and n; E Z+, then -r(n) = (nl + l)(n2 + 1) ... (nk + 1).
b) The number of ideals of Zn is -r(n).
c) -r(n)</>(n) ~ n. [Hint: II(n; + I)II(1 - lip;) ~ 2kn(1/2t.]

9. Given that the set H n = {[a] E Zn I[a] is not a zero divisor of Zn}, prove that
(Hn' 'n) forms a finite group of order </>(n).

"*
10. a) Derive Fermat's Little Theorem: If p is a prime number and a 0 (mod p),
then aP-1 == 1 (mod p).
b) If gcd (a, n) = 1, show that the equation ax == b (mod n) has a unique solution
modulo n. [Hint: All solutions are given by x = ba<!>(m-I)+ kn.]

11. a) Prove that every field is a principal ideal domain.

b) Show that the ring R = {a + b.J2Ia, b E Z} is not a field by exhibiting a
nontrivial ideal of R.

12. Let f be a homomorphism from the ring R into the ring R' and suppose that R
has a subring F which is a field. Establish that either F ~ ker f or else R' contains
a subring isomorphic to F.

13. Derive the following results:

a) The identity element of a subfield is the same as that of the field.
b) If {F;} is an index collection of subfields of the field F, then r. F; is also a
subfield of F.
c) A subring F' of a field F is a subfield of F if and only if F' contains at least one
nonzero element and a-I E F' for every nonzero a E F'.
d) A subset F' of a finite field F is a subfield of F if and only if F' contains more
than one element and is closed under addition and multiplication.

14. a) Consider the subset S of R # defined by

S = {a + b.Jpla, b E Q; p a fixed prime}.
Show that S is a subfie1d of R #.
b) Prove that any subfield of the field R# must contain the rational numbers.

15. Prove that if the field F is of characteristic p > 0, then every subfield of F has
characteristic p.
PROBLEMS 69

16. Let F be a field of characteristic p > O. Show that for fixed n E Z +}

F' = {aEFla pn = a
is a subfield of F.
17. Let F be a field, F' a subfield of F, and fan automorphism of F. We say thatf
fixes an element a E F in case f(a) = a. Prove the following assertions:
a) The set of all automorphisms of F form a group (in which the binary operation
is composition of functions).
b) The set of all automorphisms of F which fix each element of F' comprise a group.
c) If G is a group of automorphisms of F, then the set of all elements of F that
are fixed by G (that is, the set F(G) = {a E Flf(a) = a for allfE G}) is a sub-
field of F, known as the fixed field of G.
18, Let R be a commutative ring containing at least one non-zero-divisor. Prove that
a) An element ab- 1 is a non-zero-divisor of Qcl(R) if and only if a is a non-zero-
divisor of R.
b) If R has an identity and every non-zero-divisor of R is invertible in R, then
R = Qcl(R); in particular, F = Qcl(F) for any field F.
c) Qcl(Qcl(R)) = Qcl(R). .
d) If R is finite, then R = QCl(R). [Hint: For any non-zero-divisor a E R, there
is some bE R such that a2 b = a; ab is idempotent; thus, R has an identity
element 1 and ab = 1 by Problem 12, Chapter 1.]
19. Utilize part (d) of the preceding problem to give another proof that any finite
integral domain is a field.
20. Show that any field containing the integral domain R as a sUbring contains the
field of quotients Qcl(R); in this sense, QclR) is the smallest field containing R.
21. a) If R = {a + b.J2Ia, bE Z}, then R forms an integral domain under ordinary
addition and multiplication, but not a field. Obtain the field of quotients of R.
Do the same for the domain Ze'
b) If K is a field of,guotients of an integral domain R, prove that K is also a field
of quotients of e}o'ery subdomain of K containing R.
I .
22. Let R be an arbitrary ring'(not necessarily commutative) with at least one non-zero-
divisor. Prove that R possesses a classical ring of quotients if and only if it satisfies
the so-called Ore condition: for all a, b E R, b being a non-zero-divisor, there exist
elements c, d E R, with d a non-zero-divisor such that ad = bc.
23. Prove that any automorphism of an integral domain R admits a unique extension
to the field of quotients Qcl(R).
24. Let F be a field and ZI the set of integral multiples of the identity. Verify that
the prime subfield of F coincides with QclZ1). [Hint: Problem 20.]
25. Establish the following ass~rtion, thereby completing the proof of Theorem 4-12:
If F is a field of characteristic zero and
K = {(nl)(ml)-lln, mE Z; m =1= O}
is the prime subfield of F, then K ~ Q via the mappingf(n/m) = (nl)(ml)-l.
70
In Problems 26-29, R is assumed to be a commutative ring.
26. Let S be any multiplicatively closed subset of the ring R (that is, the product of
any two elements of S again lies in S) which contains no zero divisor of Rand
o¢ S. If the set Rs is defined by
Rs = {ab- 1 E Qcl(R)la E R, bE S},
prove that Rs is a subring of the ring of quotients Qc.(R), known as the ring of
quotients of R relative to S.
27. a) Show that the set S = {n E Zip .t n; p a fixed prime} is multiplicatively closed
and determine Zs, the ring of quotients of Z relative to S..
b) If R is any ring satisfying Z s;; R S;; Q, prove that R = Zs for a suitable
multiplicatively closed subset S s;; Z. [Hint: Consider the set S = {m E Zifor
some nEZ, n/m E R; gcd (n, m) = I}.]
28. Let S be a multiplicatively closed subset of the ring R with 0 ¢ S. Prove the
statements below:
a) The set I = {a E Rlas = 0 for some s E S} is an ideal of R.
b) S/I = natIS is a multiplicatively closed subset of the quotient ring R/I.
c) No element of S/I is.a zero divisor of R/I. (Thus, one can form the ring of
quotients of R/I relative to S/I; the result is called the generalized ring of
quotients of R relative to S.)
d) If S contains no zero divisor of R, then (R/I)s/l = Rs.
29. Let S be a multiplicatively closed subset of the ring R which contains no zero
divisor of R nor zero.
a) If I is an ideal of R, verify that the set IS- 1 = {ab- 1 E Qcl(R)la E I, bE S} is
anidealofQc.(R). Conversely, each ideal J ofQcl(R) is ofthe formJ = (J n R)S-1.
b) For ideals I, J of R, establish the identities
FIVE

MAXIMAL, PRIME, AND PRIMARY IDEALS

The present chapter is devoted to a study of certain special types of ideals,

most notably maximal, prime, and primary ideals. On the whole, our
hypothesis will restrict us to commutative rings with identity. The require-
ment is motivated to some extent by the fact that many of the standard
examples of ring theory have this property. Another reason, which is
perhaps more important from the conceptual point of view, is that the most
satisfactory and complete results occur here. We begin our discussion with
the following definition.
Definition 5-1. An ideal I of the ring R is said to be a maximal ideal
provided that I =1= R and whenever J is an ideal of R with I c J ~ R,
then J = R.
Expressed somewhat loosely, an ideal is maximal if it is not the whole
ring and is not properly contained in any larger proper ideal; the only ideal
to contain a maximal ideal properly is the ring itself.
It is usually quite awkward to prove that an ideal is maximal directly
from Definition 5-1. We therefore need several theorems which will help
to determine whether or not a given'jdeal is actually maximal, but which
are, in general, easier to apply than Definition 5-1. One such result is
presented below.
Theorem 5-1. Let / be a proper ideal of the ring R. Then / is a maximal
ideal if and only if (I, a) = R for any element a ¢ /. Here (I, a) denotes
the ideal generated by / u {a}.
Proof First, notice that (/, a) satisfies I c (/, a) ~ R. These inclusions
imply that if / were a maximal ideal of R, we would necessarily have
(I, a) = R. On the other hand, assume that J is an ideal of the ring R with
the property that / c J ~ R. If a is any element of J which does not lie
in I, then / c (/, a) ~ J. The requirement that (/, a) =.R would thus
force J to be all of R, and we could conclude that / is a maximal ideal.
A knowledge of several moderately simple examples will provide some
basis for understanding these ideas.
71
72

Example 5-1. We propose to show that in the ring Z of integers the

maximal ideals correspond to the prime numbers; more precisely: the
principal ideal (n), n > 1, is maximal if and only if n is a prime.
Suppose that (n) is a maximal ideal of Z. If the integer n is not prime,
then n = n 1 n2 , where 1 < n 1 < n 2 .< n. This implies that the ideals (n 1 )
and (n 2 ) are such that
(n) c (n 1 ) c Z, (n) c (n 2 ) c Z,

contrary to the maximality of (n).

For the opposite direction, assume now that the integer n is prime. If
the principal ideal (n) is not maximal in Z, then either (n) = Z or else there
exists some proper ideal (m) satisfying (n) c (m) c Z. The first case is
immediately ruled out by the fact that 1 is not a multiple of any prime
number. The alternative possibility, (n) c (m), means that n = km for some
integer k > 1; this is equally untenable, since n is prime, not composite.
At any rate, we conclude that (n) must be a maximal ideal.
Example 5-2. For an illustration of the practicality of Theorem 5-1, we
take R = map R #, a commutative ring with identity (Example 4, Chapter
1). Consider the set M of all functions which vanish at 0:
M = {IE Rlf(O) = O}.
Evidently, M forms an ideal of the ring R; we contend that it is actually
a maximal ideal. Indeed, iff 1= M and i is the identity map on R # (that is,
i(x) = x), one may easily check that (i 2 + j2)(x) =f. 0 for each x E R #.
Hence, the function i 2 + f2 is an invertible element of R. Since
R :2 (M, f) :2 (i,.I), with i 2 + F E (i,.n. this implies that (M, f) = R; in
consequence, M is a maximal ideal of R. (Here (i, f) denotes the ideal
generated by i and f; that is, (i, f) = {ri + sflr, s E R}.)
Our immediate goal is to obtain a general result assuring the existence
of suitably many maximal ideals. As will be seen presently, the crucial
step in the proof depends on Zorn's Lemma (see Appendix B), an exceedingly
powerful tool which is almost indispensable in modern mathematics.
Zorn's Lemma (traditionally called a lemma, but in fact an equivalent form
of the Axiom of Choice) asserts:
Zorn's lemma. If (S, ~) is a partially ordered set with the property
that every chain in S has an upper bound in S, then S possesses at least
one maximal element.
Clearly, some partial orderings are more useful than others in applica-
tions of Zorn's Lemma. In our later investigations we shall frequently take
S to be a family of subsets of a given set and the partial ordering to be the
usual inclusion relation; an upper bound of any chain of elements would
MAXIMAL, PRIME, AND PRIMARY IDEALS 73

simply be their set-theoretic union. For this particular setting, Zorn's

Lemma may be formulated as follows:
Let d be a non empty family of subsets of some fixed nonempty set with
the property that for each chain ~ in d, the union u ~ also belongs
to d. Then d contains a set which is maximal in the sense that it is
not properly contained in any member of d.
Because this may be the reader's first contact with Zorn's Lemma, we
proceed in somewhat leisurely fashion to establish
Theorem 5-2. If the ring R is finitely generated, then each proper ideal
of R is contained in a maximal ideal.
Proof Let I be any proper ideal of R, a finitely generated ring; say,
R = (a 1 , a z, ... ,an). We define a family of ideals of R by taking

d = {JII ~ J; J is a proper ideal of R}.

This family is obviously nonempty, for I itself belongs to d.
Now, consider an arbitrary chain {IJ of ideals in d. Our aim, of course,
is to establish that u Ii is again a member of d. To this purpose, let the
elements a, bE u Ii and r E R. Then there exist indices i and j for which
a E Ii' bE I j • As the collection {Ii} forms a chain, either Ii ~ I j or else
I j ~ Ii. For definiteness, suppose that Ii ~ I j , so that both a, bE I j • But
I j is an ideal of R; hence, the difference a - b E I j , ~ u Ii. Also, the
products ar and ra E Ii ~ U Ii. All of this shows u Ii/'to be an ideal of R.
Next, we must verify that u Ii is a proper ideal of R. Suppose, to the
contrary, that u Ii = R = (a 1 , a z, ... ,an). Then, each generator ak would
belong to some ideal Iik of the chain {I;}. There being only finitely many
I ik , one contains all others, call it Ii'. Thus, a 1 , az , ... , an all lie in this one
Ii'. In consequence, Ii' = R, which is clearly impossible. Finally, notice
that I ~ u Ii' whence the union u Ii Ed.
Therefore, on the basis of Zorn's Lemma, the family d contains a
maximal element M. It follows directly from the definition of d that M
is a proper ideal of the ring R with I ~ M. We assert that M is in fact a
maximal ideal. To see this, suppose that J is any ideal of R for which
M c J ~ R. Since M is a maximal element of the family d, J cannot
belong to d. Accordingly, the ideal J must be improper, which is to say that
J = R. We thus conclude that M is a maximal ideal of R, completing the
proof.
The significant point, of course, is that this theorem asserts the existence
of certain maximal ideals, but gives no clue as to how actually to find them.
The chief virtue of Theorem 5-2 is that it leads immediately to the following
celebrated result.
74

Theorem 5-3. (Krull-Zorn). In a ring R with identity each proper ideal

is contained in a maximal ideal.
Proof An appeal to Theorem 5-2 is legitimate, since R = (1).
Corollary. An element of a commutative ring R with identity is in-
vertible if and only if it belongs to no maximal ideal of R.
Although maximal ideals were defined for arbitrary rings, we shall
abandon a degree of generality and for the time being limit our discussion
almost exclusively to commutative rings with identity. A ring of this kind
is, of course, much easier to handle than one which is not commutative.
Another advantage stems from the fact that each ideal, other than the ring
itself, will be contained in a maximal ideal. Thus, until further notice, we
shall assJ,J):l1e that all given rings are commutative with identity, even when
this is not explicitly mentioned. To be sure, a good deal of the subsequent
material could be presented without this additional restriction.
The Krull-Zorn Theorem has many important applications throughout
ideal theory. For the moment, we content ourselves with giving an ele-
mentary proof of a somewhat special result; although the fact involved is
rather interesting, there will be no occasion to make use of it.
Theorem 5-4. In a ring R having exactly one maximal ideal M, the
only idempotents are 0 and 1.
Proof Assume that the theorem is false; that is, suppose that there exists
an idempotent q, E R with a =1= 0,1. The relation a 2 = a implies a(l - a) = 0,
so that a and 1 - a are both zero divisors. Hence, by Problem 4(d), Chapter
1, neither the element a nor 1 - a is invertible in R. But this means that the
principal ideals (a) and (1 - a) are both proper ideals of the ring R. As
such, they must be contained in M, the sole maximal of R. Accordingly,
the elements a and 1 - a lie in M, whence
1 = a + (1 - a) E M.
This leads at once to the contradiction that M = R.
Although more elementary proofs are possible, Theorem 5-4 can be
used to show that a field has no idempotents except 0 and 1. A full
justification of this statement consists of first establishing that the zero ideal
is the only maximal ideal in a field.
We now come to a characterization of maximal ideals in terms of their
quotient rings.
Theorem 5-5. Let J be a proper ideal of the ring R. Then J is a maximal
ideal if and only if the quotient ring RjJ is a field.
Proof To begin, let J be a maximal ideal of R. Since R is a commutative
ring with identity, the quotient ring RjJ also has these properties. Thus,--
MAXIMAL, PRIME, AND PRIMARY IDEALS 75

to prove that R/I is a field, it suffices to show that each nonzero element of
R/I has a multiplicative inverse. Now, if the coset a + I =I- I, then a ¢ 1.
By virtue of the fact that I is a maximal ideal, the ideal (I, a) generated by
I and a must be the whole ring R:
R = (l,a) = {i + raliEI,rER}.
That is to say, every element of R is expressible in the form i + ra, where
i E I and r E R. The identity element 1, in particular, may be written as
1 = T + ra for suitable choice of TEl, r E R. But then, the difference
1 - ra E I. This obviously implies that
1 + I = ra + I = (1' + I)(a + I),
which asserts that r + I = (a + 1) - 1. Hence, R/I is a field.
For the opposite direction, we suppose that R/I is a field and J is any
ideal of R for which I c J 5; R. The argument consists of showing that
J = R, for then I will be a maximal ideal. Since I is a proper subset of J,
there exists an element a E J with a ¢ I. Consequently, the coset a + I =l-
I, the zero element of R/1. Since R/I is assumed to be a field, a + I must
have an inverse under multiplication,
(a + I)(b + I) = ab + I = 1 + I,
for some coset b + IE R/1. It then follows that 1 - ab E I c J. But the
product ab also lies in J (recall that a is an element of the ideal J), imply~ng
that the identity 1 = (1 - ab) + ab E J. This in turn yields J = R, as
desired.
Example 5-3. Consider the ring Ze of even integers, a commutative ring
without identity. In this ring, the principal ideal (4) generated by the integer
4 is a maximal ideal, where
(4) = {4(2j) + 4kjj, k E Z} = 4Z.
The argument might be expressed as follows. If n is any element not in (4),
then n is an even integer not divisible by 4; consequently, n can be expressed
in the form n = 4m + 2 for some integer m. We then have
2 = 4(-m) + nE((4),n),
so that Ze = (2) = ((4), n). By virtue of Theorem 5-1, this is sufficient to
demonstrate the maximality of the ideal (4).
Now, note that in the quotient ring Ze/(4),
(2 + (4))(2 + (4)) = 4 + (4) = (4).
The ring Ze/(4) therefore possesses divisors of zero and cannot be a field.
The point which we wish to make is that the assumption of a multiplicative
identity is essential to Theorem 5-5.
76

We now shift our attention from maximal ideals to prime ideals. Before
formally defining this notion, let us turn to the ring Z of integers for
motivation. Specifically, consider the principal ideal (p) generated by a
prime number p. If the product ab E (P), where a, b E Z, then p divides abo
But if a prime divides a product, it necessarily divides one of the factors.
This being the case, either a E (p) or bE (P). The ideal (P) thus has the
interesting property that, whenever (P) contains a product, at least one of
the factors must belong to (P). This observation serves to suggest and
partly to illustrate the next definition.
Definition 5-2. An ideal I of the ring R is a prime ideal if, for all a, b in
R, ab E I implies that either a E I or b E 1.
By induction, Definition 5-2 can easily be extended to finitely many
elements: an ideal I of R is prime if, whenever a product a 1 a 2 ••• an of
elements of R belongs to I, then at least one of the a i E I. In this connection,
we should caution the reader that many authors insist that the term "prime
ideal" always means a proper ideal.
Example 5-4. A commutative ring R with identity is an integral domain if
and only if the zero ideal {O} is a prime ideal of R.
Example 5-5. The prime ideals of the ring Z are precisely the ideals (n),
where n is a prime number, together with the two trivial ideals {O} and Z.
From above, we already know that if n is a prime, then the principal ideal
(n) is a prime ideal of Z. On the other hand, consider any ideal (n) with n
composite (n =F 0, 1); say, n = n 1 n 2 , where 1 < n 1, n 2 < n. Certainly the
product n 1 n 2 = n E (n). However, since neither n 1 nor n 2 is an integral
multiple of n, n1 ¢: (n) and n 2 ¢: (n). Hence, when n is composite, the ideal
(n) cannot be prime. Notice also that although {O} is prime, it is not a
maximal ideal of Z.
Example 5-6. For an illustration of a ring possessing a nontrivial prime
ideal which is not maximal, take R = Z x Z, where the operations are per-
formed componentwise. One may readily verify that Z x {O} is a prime
ideal of R. Since
Zx {O} C ZxZe c R,
with Z x Ze an ideal of R, Z x {O} fails to be maximal.
By analogy with Theorem 5-5, the prime ideals of a ring may be charac-
terized in the following manner.
Theorem 5-6. Let I be a proper ideal of the ring R. Then I is a prime
ideal if and only if the quotient ring RjI is an integral domain.
Proof First, take I to be a prime ideal of R. Since R is a commutative
MAXIMAL, PRIME, AND PRIMARY IDEALS 77

ring with identity, so is the quotient ring Rjl. It remains therefore only
to verify that Rjl is free of zero divisors. For this, assume that
(a + I)(b + I) = I.
In other words, the product of these two cosets is the zero element of the
ring Rjl. The foregoing equation is plainly equivalent to requiring that
ab + I = I, or what amounts to the same thing, ab E I. Since I is assumed
to be a prime ideal; one of the factors a or b must be in 1. But this means
that either the coset a + I = I or else b + I = I; hence, Rjl is without
zero divisors.
To prove the converse, we simply reverse the argument. Accordingly,
suppose that Rjl is an integral domain and the product ab E I. In terms of
cosets, this means that
(a + I)(b + I) = ab + I = I.
By hypothesis Rjl contains no divisors of zero, so that a + I = I or
b + I = 1. In any event, one of a or b belongs to I, forcing I to be a prime
ideal of R.
There is an important class of ideals which are always prime, namely,
the maximal ideals. From the several ways of proving this result, we choose
the argument given below; another approach involves the use of Theorems
5-5 and 5-6.
Theorem 5-7. In a commutative ring with identity, every maximal ideal
is a prime ideal.
Proof Assume that I is a maximal ideal of the ring R, a commutative ring
with identity, and the product ab E I with a ¢ I. We propose to show that
bEl. The maximality of I implies that the ideal generated by I and a must
be the whole ring: R = (I, a). Hence, there exist elements i E I, r E R such
that 1 = i + ra. Since both ab and i belong to I, we conclude that
b = Ib = (i + ra)b = ib + r(ab) E I,
from which it follows that I is a prime ideal of R.
We should point out that without the assumption of an identity element
this last result does not remain valid; a specific illustration is the ring Ze
of even integers, where (4) forms a maximal ideal which is not prime. More
generally, one can prove the following: if R is a commutative ring without
identity, but having a single generator, then R contains a nonprime maximal
ideal. To establish this, suppose that R = (a). First observe that the
principal ideal (a 2 ) is a proper ideal of R, since the generator a ¢ (a 2 ).
Indeed, were a in (a 2 ), we could write a = ra 2 + na 2 for some r E Rand
n E Z; it is a simple matter to check that the element e = ra + na would
78

then serve as a multiplicative identity for R, violating our hypothesis.

+
Since (a 2 ) R, Theorem 5-2 guarantees the existence of a maximal ideal
M of R with (a 2 ) ~ M. However, M is not a prime ideal, as can be seen
by considering the product of elements in the complement of M (given
r, s ¢ M, the product rs E (a 2 ) ~ M).
Of course, the converse of Theorem 5-7 does not hold; Example 5-6
shows that there exist nontrivial prime ideals which fail to be maximal
ideals. The special properties of Boolean rings and principal ideal domains
guarantee that the notions of primeness and maximality are equivalent for
these important classes of rings. Let us look at the details.
Theorem 5-8. Let R be a Boolean ring. A nontrivial ideal I of R is
prime if and only if it is a maximal ideal.
Proof It is sufficient to show that if the ideal I is prime, then I is also
maximal. To see this, suppose that J is an ideal of R with the property that
I c: J ~ R; what we must prove is that J = R. If a is any element of J
not in I, then a(1 - a) = 0 E I. Using the fact that I is a prime ideal with
a ¢ I, we infer that 1 - a E I c: J. As both the elements a and 1 - a lie
in J, it follows that
1 = a + (1 - a) E J.
The ideal J thus contains the identity and, consequently, J = R. Since no
proper ideal lies between I and the whole ring R, we conclude that I is a
maximal ideal.
Remark. Since every integral domain contains the two trivial prime ideals,
the use of the term "prime ideal" in a principal ideal domain customarily
excludes these from consideration.
Theorem 5-9. Let R be a principal ideal domain. A nontrivial ideal
(a) of R is prime if and only if it is a maximal ideal.
Proof Assume that (a) is a prime ideal and let I be any ideal of R satisfying
(a) c: I ~ R. Because R is a principal ideal ring, there exists an element
bE R for which I = (b). Now, a E (a) c: (b); hence, a = rb for some choice
ofr in R. By supposition, (a) is a prime ideal, so that either r E (a) or b E (a).
The possibility that bE (a) leads immediately to the contradiction (b) ~ (a).
Therefore, the element r E (a), which implies that r = sa for suitable choice
of s in R, or a = rb = (sa)b. Since a +0 and R is an integral domain, we
must have 1 = sb. This, of course, means that the identity element
1 E (b) = I, whence I = R, making (a) a maximal ideal of R. Theorem
5-7 takes care of the converse.
Corollary. A nontrivial ideal of the ring Z is prime if and only if it is
maximal.
MAXIMAL, PRIME, AND PRIMARY IDEALS 79

Before taking up the matter of primary ideals, let us detour briefly to

introduce a concept which plays an important role in many aspects of ideal
theory.
Definition 5-3. Let I be an ideal of the ring R. The nil radical of I,
designated by Fis the set
.JT= {r E Rlrn E I for some n E Z+ (n varies with r)}.
We observe that the nil radical of I may equally well be characterized
as the set of elements r E R whose image r + I in the quotient ring R/I is
nilpotent. The nil radical of the zero ideal is sometimes referred to as the
nil radical of the ring R; this set consists of all nilpotent elements of Rand
accounts for the use of the term.
Example 5-7. In the ring Z, let us show that if n = p~J p~2 ... p~r is a
factorization of the positive integer n =1= 1 into distinct primes Pj' then
J(ii) = (PlP2 ... Pr)·
Indeed, if the integer a = PlP2 ... Pr and k = max {kl' k2' ... , k,.}, then we
have ak E (n); this makes it clear that (PlP2 ... Pr) 5;; J(ii). On the other
hand, if some positive integral power of the integer m is divisible by n (that
is, if mE .,j(n»), then m itself must be divisible by each of the primes
Pl' Pz, ... ,Pr' and, hence, a member of the ideal
(Pl) II (P2) II ... II (Pr) = (PlP2 ... Pr)·
As concrete illustrations ofthis situation, observe that J(12) = J(2 2 3) =
(6) and J(8) = (2).
Although it is not obvious from the definition, /i
is actually an ideal
of the ring R which contains I. In the first place, if a and b are elements of
/i, then there exist suitably chosen integers n, m E Z + such that an E I and
bmE 1. Now, every term in the binomial expansion of (a - b)n+m contains
either an or bmas a factor. This implies that (a - b)n+m lies in I and therefore
the difference a - bE/i. Next, ifr is any element of R, then (ra)n = r"an E I,
so that ra E /i; thus, /i is indeed an ideal of R. That I 5;; /i should be
clear from the definition.
Some of the basic properties of the nil radical of an ideal are assembled
in the theorem below.
Theorem 5-10. If I and J are two ideals of the ring R, then
1) .,jIJ = .,jI II J = /i II .,jJ,
2) .,jI + J = JJ I + .,jJ ;2 /i + .,j1,
3) Ik 5;; J for some k E Z + implies that /i JJ, and
5;;

4) !JI /i. =
80

Proof. Since property (1) is the only fact that will be explicitly required
in the body of the text, we shall content ourselves with its derivation; the
proofs of the remaining assertions are quite elementary and are left as an
exercise.
Now, if an E IJ, then an E I n J, and so an E I, an E J. We thus conclude
that JIJ s;; .jI n J s;; JT n.jJ. On the other hand, if it happens that
a E JI n .jJ, there must exist positive integers n, m, for which an E I and
am E J. This implies that the element an+ m = anamE IJ; hence, a E .jIJ.
Accordingly, JI n .jJ S;; JfJ and the desired equality follows.

In passing, we might point out that although property (1) easily

generalizes to finite intersections, it is false if infinite intersections are
allowed. This is best brought out by once again considering the ring Z and
the collection of principal ideals (pk), where p is a fixed prime and k ;;::: 1;
it follows readily that
nk J(pk) nk(p)'
= = (p) =1= {O} = J nk (p~.
A problem of central interest is that of determining conditions under
which a given ideal coincides with its nil radical; in this connection, the
following definition will be useful (the reason for our choice of terminology
appears shortly).
Definition 5-4. An ideal I of the ring R is said to be a semiprime ideal
if and only if I = ji
In effect, Definition 5-4 states that an ideal I is semiprime if and only
if an E I for some n E Z + implies that a itself lies in I. Our next result
characterizes semiprime ideals by the quotient rings which they determine.
Theorem 5-11. An ideal I of the ring R is a semiprime ideal if and only
if the quotient ring RjI has no nonzero nilpotent elements.
Proof. Suppose that a + JI
is a nilpotent element of Rj JI.
Then there
exists some n E Z+ such that (a + ..)1)" = an + ..)1 = .jJ; that is to say,
the element an E..)1. l!ence, (ant = anm E I for some positive Jnteger m.
This implies that a E .jI and, consequently, that a + .JI
= ")1, the zero
element of Rj.j1.
As regards the converse, assume that RjI has no nonzero nilpotent
elements and let a E..)1. Then, for some positive integer n, an E I. Passing
to the quotient ring RjI, this simply means that (a + I)n = I; in other
words, a + I is nilpotent in RjI. By supposition, we must have a + I = I
and, in consequence, the element a E I. Our argument shows that JI s;; I;
since the reversed inclusion always holds, I = JI,
so that I is semiprime.
This being proved, it is not hard to establish
MAXIMAL, PRIME, AND PRIMARY IDEALS 81

Corollary. If P is a prime ideal of the ring R, then P is semiprime.

Proof Because P is prime, RIP possesses no zero divisors and, in particular,
no nonzero nilpotent elements.
This corollary provides another good reason why a semiprime ideal was
termed as it was; being a semiprime ideal in a ring is a bit weaker than
being prime. There is much more that could be said about semiprime
ideals, and more will be said later in the text, but let us now turn our
attention to primary ideals.
In Chapter 11 we shall show that the ideals of a rather wide class of
rings (to be quite explicit, the Noetherian rings) obey factorization laws
which are roughly similar to the prime factorization laws for the positive
integers. It will turn out that the primary ideals, which we are about to
introduce, playa role analogous to the powers of prime numbers in ordinary
arithmetic.
Definition 5-5. An ideal I of the ring R is called primary if the conditions
ab E I and a ¢ I together imply b n E I for some positive integer n.
Clearly, any prime ideal satisfies this definition with n = 1, and thus,
the concept of a primary ideal may be viewed as a natural generalization
of that of a prime ideal. Lest the reader jump to false conclusions, we hasten
to point out that a primary ideal is not necessarily a power of a prime ideal
(see Example 8, Chapter 7). Notice too that Definition 5-5 may be stated
in another way: an ideal I is primary if ab E I and a ¢ I imply b E JT;this
formulation in terms of the nil radical is frequently useful.
In the ring Z, the primary ideals are precisely the ideals (pn), where p
is a prime number and n ~ 1, together with the two trivial ideals.
Our first theorem on primary ideals is simple enough; it shows that to
every primary ideal there corresponds a specific prime ideal.
Theorem 5-12. If Q is a primary ideal of the ring R, then its nil radical
.J"Q is a prime ideal, known as the associated prime ideal of Q.
Proof Suppose that ab E .J"Q, with a ¢ J"Q. Then (abt = anbn E Q for
some positive integer n. But an ¢ Q, for otherwise a would lie in.JQ. Since
Q is assumed to be primary, we must therefore have (bn)m E Q for suitable
choice of mE Z +, and so bE .J"Q. This is simply the statement that J"Q is a
prime ideal of R.
It may very well happen that different primary ideals will have the same
associated prime ideal. This is demonstrated rather strikingly in the ring
of integers where, for any n E Z +, (p) is the prime ideal associated with each
of the primary ideals (pn).
It might also be of interest to mention that the nil radical .JQ is the
smallest prime ideal to contain a given primary ideal Q. For, suppose that
82

P is any prime ideal containing Q and let a E JQ. Then there exists a suitable
positive integer n such that an E Q C;; P. Being prime, the ideal P must
contain the element a itself, which yields the inclusion JQ C;; P.
The primary ideals of R may be characterized in the following way.
Theorem 5-13. Let I be an ideal of the ring R. Then I is a primary
ideal if and only if every zero divisor of the quotient ring R/I is nilpotent.
Proof First, suppose that I is a primary ideal of R and take a + I to be
a zero divisor of R/I. Then there exists some coset b + I =F I, the zero
element of R/I, for which (a + I)(b + I) = I; that is, ab + I = I. There-
fore ab E I and, since b + I =F I, we also have b ¢ I. Now, I is assumed to
be primary, so that an E I for some positive integer n. This being the case,
+ I)n = an + I =
(a I,
which shows that the coset a + I is nilpotent.
Going in the other direction, we assume that any zero divisor of R/I
is nilpotent and let ab E I, with b ¢ 1. It then follows that (a + I)(b + I) = I,
while b + I =F I; if a + I =F I, this amounts to saying that a + I is a zero
divisor in R/I. By hypothesis, there must exist some n E Z + such that
(a + I)" = I, which forces the element an to be in 1. Thus, I is a primary
ideal of R.
Theorem 5-13 serves to emphasize the point that primary ideals are a
modification of the notion of a prime ideal; for, in the quotient ring of a
prime ideal, there are no zero divisors (hence, in a vacuous sense, every zero
divisor is nilpotent).
The following somewhat special result will be needed later, so we pause
to establish it before proceeding.
Corollary. If Ql' Q2' ... , Qn are a finite set of primary ideals of the
ring R, all of them having the same associated prime ideal P, then
Q = n7=1 Qi is also primary, with JQ = P.
Proof Before we delve into the details of the proof, observe that, by
Theorem 5-10,
.JQ = J n Qi = n JQ; = n P = P.
Now, suppose that a + Q is a zero divisor of the quotient ring R/Q. In
this event, we can find a coset b + Q =F Q such that
ab + Q = (a + Q)(b + Q) = Q.
Since b ¢ Q = n Qi' there exists some index i for which b ¢ Qi' Further-
more, ab E Qi with Qi primary, so that the element a E JQi = P = JQ.
This implies an E Q for some integer n; in consequence,
(a + Q)n = an + Q = Q,
MAXIMAL, PRIME, AND PRIMARY IDEALS 83

which is to say that a + Q is nilpotent. As every zero divisor of the quotient

ring R/Q is nilpotent, an appeal to Theorem 5-13 is in order and we may
conclude that Q is a primary ideal of R.
There is another, frequently useful, criterion for deciding whether a
given ideal is actually primary.
Theorem 5-14. Let P and Q be ideals of the ring R such that
1) Q ~ P ~ ~Q,
2) if ab E Q with a ¢ P, then b E Q.
Under these conditions, Q is a primary ideal of R with P = JQ..
Proof To see that Q is a primary ideal, suppose that the product ab E Q
but b ¢ Q. Using (2), we may conclude that a E P ~ ~, whence an E Q
for some positive integer n; this shows that Q is primary.
In order to prove that P = JQ, we need only establish the inclusion
~ ~ P, since equality would then follow from (1). For this, let the element
b E JQ, so that there exists some n E Z + for which bn E Q; assume that n'
is the smallest positive integer with this property. If n' = 1, we would have
bE Q ~ P, from condition (1). If n' > 1, it follows that bn' = bn' - l
b E Q, with bn' - 1 ¢ Q; hence, b E P by (2). In any event, we have shown
that b E ~Q implies b E P, as required.
A relationship between maximal ideals and primary ideals is brought
out in the following corollary to Theorem 5-14.
Corollary. If M is a maximal ideal of the ring R, then all its powers
Mn (n ;;::: 1) are primary ideals.
Proof Since Mn ~ M = ~ Mn, we need only verify condition (2) of the
foregoing theorem. Suppose, then, that ab E Mn with a ¢ M. Because M is
maximal, the ideal (M, a), generated by M and a must be the whole ring R.
Hence, the identity element 1 E (M, a), so, for some m E M and r E R, we
must have 1 = m + ra. Now, mn lies in Mn. Raising the equation
1 = m + ra to the nth power and using the binomial theorem, it follows that
1 = mn + r' a, where r' E R. But then
b = bmn + r'(ab)
is an element of Mn, and Mn is primary.
Another result which has this same general flavor, but which we leave
as an exercise, is the following: If the nil radical JI of an ideal I is a maximal
ideal, then I itself is primary.
Before closing this chapter, we present two additional theorems regard-
ing prime ideals. The first of these involves the notion of a minimal prime
ideal of an ideal.
84

Definition 5-6. Let I be an ideal of the ring R. A prime ideal P of R

is said to be a minimal prime ideal of I (sometimes, an isolated prime
ideal of I) if I s P and there exists no prime ideal P' of R such that
I S P' c P.
By abuse of language, we shall refer to the minimal prime ideals of the
zero ideal {O} as the minimal prime ideals of the ring R; that is, a prime
ideal is a minimal prime ideal (of R) if it does not properly contain any other
prime ideal.
Let us observe that, in the ring Z of integers, the minimal prime ideals
of a nonzero ideal (n) are precisely the prime ideals (p), where p is a prime
dividing n. In particular, one infers that every ideal in Z possesses only a
finite number of minimal prime ideals (this result is generalized in Theorem
12-3).
It is not immediately clear that any (proper) ideal admits minimal prime
ideals, although this is indeed the case. To dispose of the question requires
an appeal to Zorn's Lemma; the details are set out below.
Theorem 5-15. Let I and P be ideals of the ring R, with P prime. If
I s P, then P contains a minimal prime ideal of I.
Proof Denote by ff the family of all prime ideals of R which contain I
and are contained in P:
ff = {pllp l is a prime ideal of R; I s P'S Pl.
We point out that ff is not empty, since P itself belongs to ff. Next,
introduce a partial order ::;; in ff which is opposite to the usual inclusion
relation; that is to say, if P', pI! E ff, interpret P' ::;; pI! to mean pI! S P'.
Consider any nonempty subset {PJ of ff which is totally ordered by ::;;
(more simply, {Pi} is a chain in ff). Put P = n Pi' Then P is a prime
ideal of R (Problem 11, Chapter 5) containing I and contained in P; hence,
P E ff. Since P s Pi for every value of i, it follows that Pi ::;; P, making
P an upper bound for {PJ. All the hypotheses of Zorn's Lemma, as applied
to (ff, ::;;) are satisfied, so that ff has a maximal element, say p* (this
means that if P' E ff and p* ::;; pI, then p* = PI). Inasmuch as p* E ff,
it is a prime ideal of R with I s P* s P. There remains the task of showing
that p* is necessarily a minimal prime ideal of I. For this, we suppose that
P' is any prime ideal of R satisfying I S P' s P*. Then P' E ff and
p* ::;; P'. By the maximal nature of P*, we thus have p* = pI, signifying
that p* is a minimal prime ideal of I.
Corollary 1. Every proper ideal of the ring R possesses at least one
minimal prime ideal.
Proof Since any proper ideal of R is contained in a maximal (hence, prime)
ideal of R, the theorem can be applied.
MAXIMAL, PRIME, AND PRIMARY IDEALS 85

The existence of minimal prime ideals (of R) is assured by taking I = {O}

in the statement of Theorem 5-15.
Corollary 2. If P is a prime ideal ofthe ring R, then P contains a minimal
prime ideal of R.
Our final theorem concerns the (set-theoretic) union of a finite number
of prime ideals. In this connection, we first observe that if an ideal I of the
ring R is contained in the union J u K of two arbitrary ideals of R, then
I must be contained in one of them. For, suppose that I £; J u K with
I $ J. It is therefore possible to choose an element a E I (') K such that
a ¢ J. If bEl (') J, then the sum a + b ¢ J (otherwise, a = (a + b) - b
is in J) and so a + b E K, whence b E K. The implication is that I (') J £; K;
consequently,
I = I (') (J u K) = (I (') J) u (I (') K) £; K.
The next point to which attention should be drawn is that the above
fact about the union of two ideals is no longer true when we pass to the
union of three or more ideals. For a simple example, let R = Z2 X Z2:
R = {(O, 0), (0, 1), (1, 0), (1, 1)}.
We turn R into a ring by taking the addition to be componentwise addition
modulo 2 and defining all products to be zero. Then,
11 = {(O, 0), (0, 1)}, 12 = {(O, 0), (1, O)}, 13 = {(O, 0), (1, 1)}
are all ideals in R and R = 11 U 12 U 13 • It is clear, however, that R
(regarded as an ideal) is not contained in anyone of the Ii.
The situation just described can be countered by imposing the demand
that each of the ideals involved in the union be prime. The theorem we
have in mind asserts that if an ideal is contained in a finite union of prime
ideals, then it is entirely contained in one of them. Actually, it is easier to
prove the contrapositive of this statement, viz. :
Theorem 5-16. LetI be an arbitrary ideal ofthe ring RandP t , P 2 , ••• , P n
be prime ideals of R. If I $ Pi for all i, then there exists an element
a E I such that a ¢ u Pi; hence, I $ u Pi.
Proof The argument will be by induction on the number n of prime ideals.
Assume that the theorem has already been established when there are only
n - 1 ideals (when n = 1, the result is trivial). Then, for each i (1 ~ i ~ n),
there exists an element r i E I with r i ¢ Ui'1'i
Pj. If, for some value of i, it
happens that ri ¢ Pi' then ri ¢ u P j and there is nothing to be proved. Thus,
we may restrict our attention to the case where ri E Pi for all i.
In what follows, let ai = r t ... r i- t ri+ 1 ••• rn. We assert that ai ¢ Pi.
Since Pi is prime, the contrary assumption a i E Pi would imply that rj E Pi
86

for some j =1= i, which is impossible by our original choice of r j • On the

other hand, if j =1= i, the element aj necessarily lies in Pi (ri being a factor of
aj). For the final stage of the proof, put a = L aj. We first note that,
because each of ai' a2 , .•• , an is in I, the element a E I. From the relation
ai = a - Li1iaj, with Lj1i aj E Pi' it follows that a ¢ Pi; otherwise, we
would obtain a i E Pi' an obvious contradiction. Our construction thus
ensures the existence of an element a = L aj which belongs to the ideal I
and not to any Pi' thereby proving the theorem.
Corollary. Let I be an arbitrary ideal of the ring R and Pi' P 2 , ••• , Pn
be prime ideals of R. If I £ U Pi' then I £ Pi for some i.

PROBLEMS
In the following set of problems, all rings are assumed to be commutative with identity.
1. a) Prove that Z EB Ze is a maximal ideal of the external direct sum Z EB z.
b) Show that the ring R is a field if and only if {O} is a maximal ideal of R.
2. Prove that a proper ideal M of the ring R is maximal if and only if, for every
element r ¢ M, there exists some a E R such that 1 + ra E M.
3. Letfbe a homomorphism from the ring R onto the ring R'. Prove that
a) if M is a maximal ideal of R with M ;2 ker J, thenf(M) is a maximal ideal of R',
b) if M' is a maximal ideal of R', thenf-l(M') is a maximal ideal of R,
c) the mapping M ..... f(M) defines a one-to-one correspondence between the set
of maximal ideals of R which contain kerf and the set of all maximal ideals
of R'.
4. If M 1 and M 2 are distinct maximal ideals of the ring R, establish the equality
M1M2 = Ml n M 2 •
5. Let M be a proper ideal of the ring R. Prove that M is a maximal ideal if and
only if, for each ideal I of R, either I ~ M or else I + M = R.
6. An ideal I of the ring R is said to be minimal if I +- {O} and there exists no ideal
J of R such that {O} c J c I.
a) Prove that a nonzero ideal I of R is a minimal ideal if and only if (a) = I for
each nonzero element a E I.
b) Verify that the ring Z of integers has no minimal ideals.
7. Let I be a proper ideal of the ring R. Show that I is a prime ideal if and only
if the complement of I is a multiplicatively closed subset of R.
8. In the ring R = map R #, define the set I by

I = {IE RI!(I) = f( -1) = O}.

Establish that I is an ideal of R, but not a prime ideal.
PROBLEMS 87

9. a) With the aid of Theorem 5-5 and Example 5-1, obtain another proof of the fact
that Zp is a field if and only if p is a prime number.
b) Prove that in Zn the maximal ideals are the principal ideals (p) = pZn' where
p is a prime dividing n.

10. Given thatfis a homomorphism from the ring R onto the ring R', verify that
a) R' is a field if and only if ker f is a maximal ideal of R,
b) R' is an integral domain if and only if ker f is a prime ideal of R.
11. a) Show that if P I and P 2 are two ideals of the ring R such that PI '*
P 2 and
P2 '*P l' then the ideal Pin P 2 is not prime.
b) Let {Pi} be a chain of prime ideals of the ring R. Prove that v Pi and n Pi are
both prime ideals of R.
12. Prove that if I is an ideal of the ring Rand P is a prime ideal of I, then P is an
ideal of the whole ring R.
13. Let R denote the set of all infinite sequences {an} of rational numbers (that is,
an E Q for every n). R becomes a commutative ring with identity if the ring
operations are defined termwise:

{an} + {bn} = {an + bn}, {an}'{b n} = {anb n}·

Verify each of the following statements:
a) the set B of bounded sequences is a subring (with identity) of R,
b) the set C of convergent sequences is a subring (with identity) of B,
c) the set Co of sequences which converge to zero is a subring of C,
d) Co is an ideal of B, but not a prime ideal,
e) Co is a maximal ideal of C,
f) the set D of Cauchy sequences is a subring (with identity) of B,
g) Co is a maximal ideal of D.
Remark. Since the field DIC o is isomorphic to R#, this provides an alternative
procedure for constructing the real numbers [16].
14. Assume that P is a proper prime ideal of the ring R with the property that the
quotient ring RIP is finite. Show that P must be a maximal ideal of R.
15. Let R = Rl EB R2 EB ... EB Rn be the direct sum of a finite number of rings R i .
Establish that a proper ideal I of R is a maximal ideal if and only if, for some i
(1 :::::; i :::::; n), I is of the form

I = Rl EB ... EB R i- 1 EB Mi EB RHI EB ... EB R n,

where Mi is a maximal ideal of R i . [Hint: Problem 26, Chapter 2.]

16. Let P and I be ideals of the ring R, with P prime. If I

ideal P:I = P.
'* P, prove that the quotient
17. Letfbe a homomorphism from the ring R onto the ring R'. Prove that
a) if P is a prime (primary) ideal of R with P ~ kerf, thenf(P) is a prime (primary)
ideal of R';
88

b) if P' is a prime (primary) ideal of R', thenf -I(P') is a prime (primary) ideal of R;
c) the mapping P ---+ f(P) defines a one-to-one correspondence between the set
of prime (primary) ideals of R which contain ker f and the set of all prime
(primary) ideals of R'.
18. If M is a maximal ideal of the ring Rand n E Z+, show that the quotient ring R/Mft
has exactly one proper prime ideal. [Hint: Problem 17(c).]
19. a) For any ideal I of R, prove that I and .J1 are contained in precisely the same
prime ideals of R.
b) Using part (a), deduce that whenever I is a prime ideal of R, then I = .J1.
20. Letfbe a homomorphism from the ring R onto the ring R'. Prove that
a) if I is an ideal of R with I ;2 kerf, then .Jl(i) = f(.J1),
b) if l' is an ideal of R', then .Jr 1(1') = r I(.J r).
21. Verify that the intersection of semiprime ideals of the ring R is again a semiprime
ideal of R.
22. If I is an ideal ofthe ring R, prove that.J1 is the smallest (in the set-theoretic sense)
semiprime ideal of R which contains 1.
23. Establish that every divisor of zero in the ring Zpn (p a prime, n > 0) is nilpotent.

24. Let" J, and Q be ideals of the ring R, with Q primary. Prove the following
statements:
a) if I <t .jQ, then the q~otient Q:I = Q;
b) ifIJ ~ QandI <t .JQ,thenJ ~ Q;
c) if IJ ~ Q and the ideal J is finitely generated, then either I ~ Q or else r ~ Q
for some n E Z+. [Hint: If I $ Q, each generator of J is in.jQ.]
25. Assume that I is an ideal of the ring R. If.J1 is a maximal ideal of R, show that
I is primary. [Hint: Mimic the argument of the corollary to Theorem 5-14.]
26. Let R be an integral domain and P be a prime ideal of R. Consider Rp , the ring
of quotients of R relative to the complement of P:
Rp = {ab- I E Qc)(R)la E R; b ¢ Pl.
Prove that the ring Rp (which is known as the localization of R at the prime ideal P)
has exactly one maximal ideal, namely, I = {ab- I E Qc)(R)la E P; b ¢ Pl.
27. A ring R is said to be a local ring if it has a unique maximal ideal. If R is a local
ring with M as its maximal ideal, show that any element a ¢ M is invertible in R.
28. A subring R of a field F is said to be a valuation ring of F if for each nonzero
element a E F at least one of a or a-I belongs to R. Assuming that R is a valuation
ring of F, prove the following:
a) R contains all the idempotent elements of F.
b) R is a local ring, with unique maximal ideal M = {a E Rla- I ¢ R}.
[Remark. a-I denotes the inverse of a in F.]
c) For any two elements a, bE R, either aR ~ bR or bR ;2 aR. [Hint: Either
ab- I E R or ba- I E R.]
PROBLEMS 89

d) If I is a proper ideal of Rand b E R - I, then I £ bR.

29. For a fixed prime p, consider the subset of rational numbers defined by

v;, = {alb E Qlp f b}.

Show that
a) v;, is a valuation ring of Q;
b) the unique maximal ideal of v;, is Mp = {alb E Qlp{b, but pia};
c) the field VJM p ~ Zp. [Hint: Let the homomorphismf: v;, ..... Zp be defined
by f(alb) = [a][b]-I.]
30. a) Let I l' 12 , ••• , I" be arbitrary ideals of the ring Rand P be a prime ideal of R.
If 11 I 2 ••• I" £ P, establish that Ii £ P for at least one value of i. [Hint: If
Ii $ P for all i, choose at E Ii - P and consider the element a = a1 a2 ••• a".]
b) Assume that M is a maximal ideal of R. Prove that, for each integer n E Z+'
the only prime ideal containing M" is M.
31. Let R be an integral domain with the property that every proper ideal is the product
of maximal ideals. Prove that
a) If M is a maximal ideal of R, then there exists an element a E R and ideal
K =1= {O} such that MK = (a). [Hint: If M =1= {O}, pick 0 =1= a E M. Then
MIM2 ... M" = (a) £ M for suitable maximal ideals M i ; hence, M = Mi for
some i.]
b) If I, J, M are ideals of R, with M maximal, then 1M = JM implies I = J.
32. a) If I is an ideal of the ring R such that I £ (a), show that there exists an ideal
J of R for which aJ = l. [Hint: Take J = (I: (a»).]
b) Prove that if a principal ideal (a) of the ring R properly contains a prime ideal
P, then P £ n
n=l
(a").
33. Let I be a primary ideal of the ring R. Prove that I has exactly one minimal prime
ideal, namely, JI. [Hint: Problem 19.]
SIX

DIVISIBILITY THEORY IN INTEGRAL

DOMAINS

As the title suggests, this chapter is concerned with the problem of factoring
elements of an integral domain as products of irreducible elements. The
particular impetus is furnished by the ring of integers, where the Funda-
mental Theorem of Arithmetic states that every integer n > 1 can be written,
in an essentially unique way, as a product of prime numbers; for example,
the integer 360 = 2·2·2·3·3· 5. We are interested here in the possibility
of extending the factorization theory of the ring Z and, in particular, the
aforementioned Fundamental Theorem of Arithmetic to a more general
setting. Needless to say, any reasonable abstraction of these number-
theoretic ideas depends on a suitable interpretation of prime elements (the
building blocks for the study of divisibility questions in Z) in integral
domains. Except for certain definitions, which we prefer to have available
for arbitrary rings, our hypothesis will, for the most part, restrict us to
integral domains. The plan is to proceed from the most general results
about divisibility, prime elements, and uniqueness offactorization to stronger
results concerning specific classes of integral domains.
Throughout this chapter, the rings considered are assumed to be com-
mutative; and it is supposed that each possesses an identity element.
°
Definition 6-1. If a =1= and b are elements of the ring R, then a is
said to divide b, in symbols alb, provided that there exists some c E R
such that b = ac. In case a does not divide b, we shall write a { b.
Other language for the divisibility property a Ib is that a is a factor of
b, that b is divisible by a, and that b is a multiple of a. Whenever the notation
alb is employed, it is to be understood (even if not explicitly mentioned)
that the element a =1= 0; on the other hand, not only may b = 0, but in
such instances we always have divisibility.
Some immediate consequences of this definition are listed below; the
reader should convince himself of each of them.
Theorem 6-1. Let the elements a, b, c E R. Then,
1) alO, 11a, ala;
90
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 91

2) all if and only if a is invertible;

3) if alb, then aelbe;
4) if alb and ble, then ale;
5) if cia and elb, then el(ax + by) for every x, y E R.
Division of elements in a ring R is closely related to ideal inclusion:
alb if and only if (b) s; (a).
Indeed, alb means that b = ae for some e E R; thus, b E (a), so that (b) s; (a).
Conversely, if (b) S; (a), then there exists an element e in R for which b = ae,
implying that alb.
Questions concerning divisibility are complicated somewhat by the
presence of invertible elements. For, if u has a multiplicative inverse, any
element of a E R can be expressed in the form a = a(uu- 1 ), so that both ula
and u - 11 a. An extreme situation occurs in the case of fields, where every
nonzero element divides every other element. On the other hand, in the
ring Ze of even integers, the element 2 has no divisors at all.
In order to overcome the difficulty that is produced by invertible ele-
ments, we introduce the following definition.
Definition 6-2. Two elements a, b E R are said to be associated elements
or simply associates if a = bu, where u is an invertible element of R.
A simple argument shows that the relation "', defined on R by taking
a '" b if and only if a is an associate of b, is an equivalence relation with
equivalence classes which are sets of associated elements. The associates of
the identity are just the invertible elements of R.
Example 6-1. In the case of the ring Z, the only associates of an integer
n E Z are ± n, since ± 1 are the only invertible elements.
Example 6-2. Consider the domain Z(i) of Gaussian integers, a subdomain
of the complex number field, whose elements form the set
Z(i) = {a + bila,bEZ; i2 = -I}.
Here, the only invertible elements are ± 1 and ± i. For, suppose a + bi E Z(i)
hasamultiplicativeinversee + di. Then, we must have (a + bi)(e + di) = 1,
so that (a - bi)(e - di) = 1. Therefore,
1 = (a + bi)(e + di)(a - bi)(e - di)
= (a 2 + b2 )(e 2 + d2 ).
From the fact that a, b, e, d are all integers, it follows that a 2 + b2 = 1.
The only solutions of this last equation are a = ± 1, b = 0 or a = 0,
b = ± 1. This leads to the four invertible elements ± 1, ± i. In con-
92

sequence, the class of associates determined by any Gaussian integer a + bi

consists of exactly four members:
a + bi, - a - bi, - b + ai, b - ai.

Since associated elements are rather closely related, it is not surprising

that they have similar properties; for instance:
Theorem 6-2. Let a, b be nonzero elements of an integral domain R.
Then the following statements are equivalent:
1) a and b are associates,
2) both alb and bla,
3) (a) = (b).

Proof. To prove the equivalence of (1) and (2), suppose that a = bu, where
u is an invertible element; then, also, b = au-I, so that both alb and bla.
Going in the opposite direction, if alb, we can write b = ax for some x E R;
while, from bla, it follows that a = by with y E R. Therefore, b = (by)x =
b(yx). Since b =1= 0, the cancellation law implies that 1 = yx. Hence, y is
an invertible element of R, with a = by, proving that a and b must be
associates. The equivalence of (2) and (3) stems from our earlier remarks
relating division of ring elements to ideal inclusion.
We next examine the notion of a greatest common divisor.
Definition 6-3. Let aI' a 2 , ... ,an be nonzero elements of the ring R.
An element dE R is a greatest common divisor of aI' a 2 , •• , ,an if it
possesses the properties
1) dla; for i = 1, 2, ... , n (d is a common divisor),
2) cia; for i = 1, 2, ... , n implies that cld.
The use of the superlative adjective "greatest" in this definition does
not imply that d has greater magnitude than any other common divisor c,
but only that d is a multiple of any such c.
A natural question to ask is whether the elements aI' a 2 , ••• , an E R
can possess two different greatest common divisors. For an answer, suppose
that there are two elements d and d' in R satisfying the conditions of
Definition 6-3. Then, by (2), we must have did' as well as d'ld; according
to Theorem 6-2, this implies that d and d' are associates. Thus, the greatest
common divisor of aI' a2 , ••• , an is unique, whenever it exists, up to arbitrary
invertible factors. We shall find it convenient to denote any greatest com-
mon divisor of aI' a 2 , ... , an by gcd (aI' a 2 , ... , an)'
The next theorem will prove, at least for principal ideal rings, that any
finite set of nonzero elements actually does have a greatest common divisor.
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 93

Theorem 6-3. Let a l , a2, ... ,an be nonzero elements of the ring R.
Then a l , a2, ... , an have a greatest common divisor d, expressible in the
form
(ri E R),
if and only if the ideal (a l , a2, ... , an) is principal.
Proof Suppose that d = gcd (a l , a2, ... , an) exists and can be written in
the form d = rla l + r 2a2 + ... + rnan, with ri E R. Then the element d
lies in the ideal (a l , a2, ... ,an)' which implies that (d) £ (a l , a2, ... , an).
To obtain the reverse inclusion, observe that since d = gcd (a l , a2, ... , an),
each ai is a multiple of d; say, ai = xid, where Xi E R. Thus, for an arbitrary
member yla l + Y2a2 + ... + Ynan of the ideal (a l , a2, ... , an), we must have
yla l + Y2 a 2 + ... + ynan = (YIX l + Y2 X2 + ... + Ynxn)d E (d).
This fact shows that (a l , a2, ... ,an) £ (d), and equality follows.
For the converse, let (a l , a2, ... , an) be a principal ideal of R:
(d E R).
Our aim, of course, is to prove that d = gcd (a l , a2, ... ,an)' Since each
ai E (d), there exist elements bi in R for which ai = b;d, whence dla i for
i = 1, 2, ... ,n. It remains only to establish that any common divisor C of
the ai also divides d. Now, ai = SiC for suitable Si E R. As an element of
(a l , a2, ... , an), d must have the form d = rla l + r2a2 + ... + rnan, with
r i in R. This means that
d = (rls l + r 2s2 + ... + rnsn)c,
which is to say that cld. Thus, d is a greatest common divisor of a l , a2, ... , an
and has the desired representation.
Corollary. Any finite set of nonzero elements a l , a2 , ... , anofaprincipal
ideal ring R has a greatest common divisor; in fact,
gcd (a l , a2, ... , an) = rla l + r2a2 + ... + rnan
for suitable choice of r l , r2, ... , rn E R.
When (at, a 2, ... , an) = R, the elements a l , a2, ... ,an must have a
common divisor which is an invertible element of R; in this case, we say
that a l , a2, ... , an are relatively prime and shall write gcd (a l , a2, ... ,an) = 1.
If a l , a2, ... , an are nonzero elements of a principal ideal ring R, then
the corollary to Theorem 6-3 tells us that a l , a2, ... , an are relatively prime
if and only if there exist r l' r2' ... , rn E R such that
(Bezout's Identity).
One ofthe most useful applications of Bezout's Identity is the following
(it also serves to motivate our coming definition of a prime element).
94

Theorem 6-4. Let a, b, c be elements of the principal ideal ring R. If

clab, with a and c relatively prime, then clb.
Proof Since a and c are relatively prime, so that gcd (a, c) = 1, there exist
elements r, s E R satisfying 1 = ra + sc; hence,

b = 1b = rab + scb.

As clab and clc, Theorem 6-1(5) guarantees that cl(rab + scb), or rather,
clb.
Dual to the notion of greatest common divisor there is the idea of a
least common multiple, defined below.

Definition 6-4. Let ai' a z, ... , an be nonzero elements of a ring R. An

element d E R is a least common multiple of ai' az , ... , an if
1) aild for i = 1,2, ... ,n (d is a common multiple),
2) aile for i = 1,2, ... , n implies die.
In brief, an element d E R is a least common multiple of ai' az , ... , an
if it is a common multiple of ai' a z, ... , an which divides any other common
multiple. The reader should note that a least common multiple, in case
it exists, is unique apart from the distinction between associates; indeed, if
d and d' are both least common multiples of ai' a z, ... ,an' then did' and
d'i d ; hence, d and d' are associates. We hereafter adopt the standard notation
lcm (ai' az , ... , an) to represent any least common multiple of ai' a2 , ... , an'
The next result is a useful companion to Theorem 6-3.

Theorem 6-5. Let ai' a2 , ... , an be nonzero elements of the ring R.

Then ai' a2 , ... , an have a least common multiple if and only if the ideal
n (a i ) is principal.

Proof We begin by assuming that d = lcm (ai' a z, ... ,an) exists. Then
the element d lies in each of the principal ideals (a;), for i = 1,2, ... , n,
whence in the intersection n (a i ). This means that (d) S; n (aJ On the
other hand, any element r En (a;) is a common multiple of each of the ai .
But d is a least common multiple, so that dlr, or, equivalently, r E (d). This
leads to the inclusion n (a;) S; (d) and the subsequent equality.
Going in the opposite direction, suppose that the intersection n (a i ) is
a principal ideal of R, say n (a;) = (a). Since (a) S; (a;), it follows that
aila for every i, making a a common multiple of ai' a2 , ... ,an' Given any
other common multiple b of ai' a2 , ... , an' the condition ailb implies that
(b) S; (a;) for each value of i. As a result, (b) S; n (a i ) = (a) and so alb.
Our argument establishes that a = lcm (ai' a2 , ... ,an)' completing the
proof.
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 95

At this point, we introduce two additional definitions. These will help

to describe, in a fairly concise manner, certain situations which will occur
in the sequel.
Definition 6-5. A ring R is said to have the ged-property (lem-property)
provided that any finite number of nonzero elements of R admit a
greatest common divisor (least common multiple).
The content of Theorem 6-3 is that a ring R has the gcd-property if
and only if every finitely generated ideal of R is principal. Likewise,
Theorem 6-5 tells us that R possesses the km-property if and only if the
intersection of any finite number of principal ideals of R is again principal.
Suffice it to say, every principal ideal ring satisfies both these properties.
iThe immediate task is to prove that any integral domain has the gcd-
property if and only if it has the km-property: In the process, we shall
acquire certain other facts which have significance for our subsequent
investigation. So as to avoid becoming submerged in minor details at a
critical stage of the discussion, let us first establish a lemma.
Lemma. Let at, a2 , ... , an and r be nonzero elements of an integral
domain R.
1} Iflcm (a l , a2 , ••• , an) exists, then km (ra l , ra 2 , ••• , ran) also exists and
km (ra l , ra 2 , •.. ,ran) = r km (a l , a2 , ... , an)·

2} If gcd (ra l , ra 2 , ••• , ran) exists, then gcd (at, a2 , ..• , an) also exists and

Proof First, assume that d = km (a l , a2 , ••• ,an) exists. Then a;ld for
each value of i, whence radrd. Now, let d' be any common multiple of
ra 2 , ra 2 , ... ,ran. Then rid', say d' = rs, where s E R. It follows that ads
for every i and so dis. As a result, rdlrs or rdld'. But this means that
km (ra l , ra 2 , •.• , ran) exists and equals rd = r km (a l , a2 , ... , an).
As regards the second assertion, suppose that e = gcd (ra l , ra 2 , ••• , ran)
exists. Then rle; hence, e = rt for suitable t E R. Since elra;, we have tla; for
every i, signifying that t is a common divisor of the a;. Now, consider an
arbitrary common divisor t' of al' a2 , •.• , an. Then rt'lra; for i = 1,2, ... , n
and therefore rt'l e. But e = rt, so that rt'l rt or t' Jt. The implication is
that gcd (a l , a2 , ... , an) exists and equals t. This proves what we wanted:

gcd (ra l , ra 2 , •.• ,ran) = e = rt = r gcd (a l , a2 , ••. , an).

Remark. It is entirely possible for gcd (a l , a2 , ••• , an) to exist without the
existence of gcd (ra l , ra 2 , ... , ran); this accounts for the lack of symmetry
in the statement of the above lemma. (See Example 6-4.)
96

Although the coming theorem is somewhat specialized in character, the

information it contains is frequently useful.
Theorem 6-6. Let ai' a z, ... , an and b l , b z, ... , bn be nonzero elements
of an integral domain R such that alb l = azb z = ... = anb n = x.
1) If km (ai' a z, ... , an) exists, then gcd (b l , b z, ... , bn) also exists and
satisfies
km (ai' az, ... , an) gcd (b l , b z, ... ,bn) = x.
2} If gcd (ra l , ra z, ... , ran) exists for all 0 =1= r E R, then km (b l , b z, ... ,
bn ) also exists and satisfies
gcd (ai' a z, ... , an) km (bl,b z, ... , bn) = x.
Proal For a proof of statement (1), set a = km (ai' az, ... , aJ Then
ada for i = 1, 2, ... , n, say a = ria i . From the relation x = aib i, we see
that ab i = (r;ai)b i = rix and so xlab;. On the other hand, consider any
divisor y of the ab;. Then ya;l(ab;}a;, or ya;jxa, making xa a common
multipleofya l , ya z, ... , yan· According to the lemma, km (ya l , ya z, ... , yan)
exists and equals ya. Thus, by the definition of least common multiple, we
conclude that yalxa, whence Ylx. To recapitulate, we have shown that
xlab; for each i and whenever Ylab;, then Ylx. This simply asserts that
x = gcd (ab l , ab z, ... , ab n)
= a gcd (b l , b z, ... ,bn) = km (ai' a z, ... , an) gcd (b l , bz, ... , bn),
where, once again, the lemma has been invoked.
We omit the proof of the other half of the theorem, which follows by
much the same reasoning. In order to apply the lemma, it is now necessary
to assume n. only that gcd (ai' a z, ... , an) exists but, more generally, the
L

existence of g\.:c (ra l , ra z, ... , ran) for all r =1= o.

Our next result is rather striking in that it tells us that, at least for integral
domains, the gcd-property implie~ the km-property, and conversely.
Theorem 6-7. An integral domain R has the gcd-property if and only
if R has the km-property.
Proal Let b l , bz, ... , bn be nonzero elements of R and suppose that R
possesses the km-property. Taking x = b l b z ... bn and ak = b l ... bk - l
bk + 1 ... bn for k = 1, 2, ... , n, we may appeal to the first part of Theorem
6-6 to conclude that gcd (b l , bz, ... ,bn) exists; hence, the gcd-property
holds in R. Conversely, if it is hypothesized that any finite number of non-
zero elements of R admit a greatest common divisor, then the existence of
km (b l , bz, ... , bn) can be inferred in the same way.
We now have quite a bit of information about divisibility in integral
domains, but the basic question remains unanswered: when does a ring
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 97

possess a factorization theory in which the analog of the Fundamental

Theorem of Arithmetic holds? To this end, let us introduce two new classes
of elements, prime and irreducible elements; when the ring is specialized
to the ring of integers, these concepts are equivalent and yield the usual
notion of a prime number.
Definition 6-6. 1) A nonzero element pER is called a prime if and only
if p is not invertible and plab implies that either pia or else plb.
2) A nonzero element q E R is said to be irreducible (or nonJactorizable)
if and only if q is not invertible and in every factorization q = bc with
b, c E R, either b or c is invertible.
Briefly, an irreducible element q is an element which cannot be factored
in R in a nontrh:,ial way; the only factors of q are its associates and the
invertible elements of R. In such rings as division rings and fields, where
each nonzero element possesses a multiplicative inverse, the concept of an
irreducible element is of no significance.
Observe also that every element which is an associate of an irreducible
(prime) element is itself irreducible (prime). It follows by an easy induction
argument that if a product a 1 a2 ••• an is divisible by a prime p, then p must
divide at least one of the factors a i (ii = 1, 2, ... , n).
Lemma. In an integral domain R, any prime element p is irreducible.
Proof Suppose that p = ab for some a, bE R. Since p is prime, either pia
or plb; say p divides b, so that there exists some element c in R for which
b = pc. We then have abc = pc = b. It follows from the cancellation
law that ac = 1; hence, a is invertible. This allows us to conclude that p
must be an irreducible element of R.
Although prime elements are irreducible in integral domains, the con-
verse is not always true, as we shall see later on. In the context of principal
ideal domains (our primary interest in this chapter), the notions of an
irreducible element and a prime element coincide. This is brought out in
the theorem below.
Theorem 6-8. Let R be a principal ideal domain. A nonzero element
pER is irreducible if and only if it is prime.
Proof By what we have just proved, p prime always implies p irreducible.
So, assume that p is an irreducible element and that p divides the product
ab, say pc = ab, with c E R. As R is a principal ideal ring, the ideal
generated by p and a, .
(p, a) = (d)
for some choice of d in R; hence, p = rd, for suitable r E R. But p is
irreducible by hypothesis, so that either r or d must be an invertible element.
98

If d happened to possess an inverse, we would have (p, a) = R. Thus,

there would exist elements s, t E R for which 1 = sp + tao Then,

b = bl = bsp + bta = bsp + pet = p(bs + et),

which implies that plb.
On the other hand, if r is invertible in R, then d = r -1 p E (P), whence
(d) ~ (P). It follows that the element a E (p) and, in consequence, pia. At
any rate, if plab, then p must divide one of the factors, making p a prime
element of R.
We next take up two theorems having to do with the ideal structure of
a principal ideal domain; the first result has considerable theoretical
importance and will, in particular, serve as our basic tool for this section.
Theorem 6-9. Let R be a principal ideal domain. If {In}, n E Z+, is
any infinite sequence of ideals of R satisfying

then there exists an integer m such that In = 1m for all n > m.

Proof. It is an easy matter to verify that I = u In is an ideal of R (see
the argument of Theorem 5-2). Being an ideal of a principal ideal ring,
I = (a) for suitable choice of a E R. Now, the element a must lie in one
of the ideals of the union, say the ideal 1m. For n > m, it then follows that
I = (a) ~ 1m ~ In ~ I;
hence, In = 1m, as asserted.
In asserting the equivalence of maximal and prime ideals in principal
ideal domains, Theorem 5-9 failed to identify these ideals; this situation is
taken care of by our next theorem. First, let us define a principal ideal of
the ring R to be a maximal principal ideal if it is maximal (with respect to
inclusion) in the set of proper principal ideals of R.
Lemma. Let R be an integral domain. For a nonzero element pER,
the following hold:
a) p is an irreducible element of R if and only if (p) is a maximal
principal ideal;
b) p is a prime element of R if and only if the principal ideal (P) =1= R
is prime.
Proof. To begin, we suppose that p is an irreducible element of R and that
(a) is any principal ideal for which (p) c (a) ~ R. As p e (a), we must have
p = ra for some r E R. The fact that p is an irreducible element implies
that either r or a is invertible. Were r allowed to possess a multiplicative
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 99

inverse, then a = r-1p E (P), from which it follows that (a) £ (p), an obvious
contradiction. Accordingly, the elemerit a is invertible, whence (a) = R.
This argument shows that no principal ideal lies between (p) and the whole
ring R, so that (p) is a maximal principal ideal.
On the other hand, let (p) be a maximal principal ideal of R. For a
proof by contradiction, assume that p is not an irreducible element. Then
p admits a factorization p = ab where a, b E R and neither a nor b is in-
vertible (the alterpative possibility that p has an inverse implies (p) = R,
so may be ruled but). Now, if the element a were in (p), then a = rp for
some choice of r E R; hence, p = ab = (rp)b. Using the cancellation law,
we could deduce that 1 = rb; but this results in the contradiction that b
is invertible. Therefore, a ¢ (p), yielding the proper inclusion (p) c (a).
Next, observe that if (a) = R, then a will possess an inverse, contrary to
assumption. We thus conclude that (p) c (a) c R, which denies that (p)
is a maximal principal ideal. Our original supposition is false and must /V
be an irreducible element of R.
With regard to the second assertion of the lemma, suppose that p is any
prime element of R. To see that the principal ideal (p) is in fact a prime
ideal, we let the product ab E (p). Then there exists an element r E R for
which ab = rp; hence, plab. By hypothesis, p is a prime element, so that
either p Ia or pi b. Translating this into ideals, either a E (p) or b E (p); in
consequence, (p) is a prime ideal of R.
The converse is proved in much the same way. Let (p) be a prime ideal
and plab. Then ab E (p). Using the fact that (p) is a prime ideal, it follows
that one of a or b lies in (p). This means that either pia or else plb, and makes
p a prime element of R.
For principal ideal domains, all of this may be summarized by the
following theorem.
Theorem 6-10. Let R be a principal ideal domain. The nontrivial ideal
(p) is a maximal (prime) ideal of R if and only if p is an irreducible (prime)
element of R.
An immediate consequence of this theorem is that every nonzero non-
invertible element of R is divisible by some prime.
Corollary. Let a =1= 0 be a noninvertible element of the principal ideal
domain R. Then there exists a prime pER such that pia.
Proof Since a is not invertible, the principal ideal (a) =1= R. Thus, by
Theorem 5-2, there exists a maximal ideal M of R such that (a) £ M. But
the preceding result tells us that every maximal ideal is of the form M = (p),
where p is a prime element of R (in this setting, there is no distinction
between prime and irreducible elements). Thus, (a) £ (p), which is to say
that pia.
100

Many authors do not insist that an integral domain possess an identity

element; for this reason, let us sketch a second proof of the foregoing
corollary which avoids the use of Theorem 5-2. First, put (a) = 11. If 11
is not already a maximal ideal, then there exists an ideal 12 of R such that
11 C 12. By the same reasoning, if 12 is not maximal, then 11 C 12 C 13
for some ideal 13 • Appealing to Theorem 6-9, this process must terminate
after a finite number of steps; in other words, we can eventually, find a
maximal ideal M of R containing 11 = (a). As before, M = (p), with P a
prime in R; the remainder of the proof is like that above.
If R is an integral domain with the property that every noninvertible
element of R can be expressed uniquely (up to invertible elements as factors
and the order of factors) as a product of irreducible elements, then we say
that R is a unique factorization domain; for a more formal definition:
Definition 6-7. An integral domain R is a unique factorization domain
in case the following two conditions hold:
I

1) every element a E R, which is neither zero nor invertible, can be

factored into a finite product of irreducible elements;
2) if a = P1P2 ... Pn = qlq2 ... qm are two factorizations of a into
irreducible elements, then n = m and there is a permutation 'It of the
indices such that Pi and q"(i) are associates (i = 1,2, ... , n).
In short, an integral domain is a unique factorization domain if it
possesses a factorization theory in which the analog of the Fundamental
Theorem of Arithmetic holds. We intend to show that any principal ideal
domain is a unique factorization domain; towards this goal, let us first
prove:
Theorem 6-11. If R is a principal ideal domain, then every element of
R which is neither zero nor invertible has a factorization into a finite
product of primes.
Proof Consider any nonzero noninvertible element a E R. By the last
corollary, there exists a prime Pl in R with Plla. Then a = Plal for some
(nonzero) a 1 E R, whence (a) ~ (a 1). Were (a) = (a 1), we would have
a 1 = raforsuitablerER;itwouldfollowthata = Plal = Plra, or 1 = Plr,
resulting in the contradiction that Pl is invertible. Consequently, we have
the proper inclusion (a) c (a 1 ).
Repeat the procedure, now starting with a 1 , to obtain an increasing
chain of principal ideals
(a) c (a 1) c (a 2) c ... c (an) c ... ,
with an- 1 = pnan for some prime Pn E R. This process goes on as long as
an is not an invertible element of R. But Theorem 6-9 asserts that the
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 101

displayed chain of ideals eventually terminates; in other words, an must

possess an inverse for some n, and

We conclude from this that the element ais expressible as the finite product
of primes )
)_ I
a - PlP2 ... Pn-lPn'
where p~ = Pnan, being an associate of a prime, is itself prime.
Corollary. In a principal ideal domain R; every nontrivial ideal is the
product of a finite number of prime (maximal) ideals.
Proof If the element 0 =1= a E R is not invertible, then a has a representation
as a finite product of primes; say a = PlP2 ... Pn' where each Pi is a prime
element of R. It then follows that
(a) = (PlP2 ... Pn) = (Pl)(P2) ... (Pn)'
with (Pi) a prime ideal of R.
Specializing to the ring of integers, we obtain a celebrated result.
Theorem 6-12. (Euclid). There are an infinite number of primes in z.
Proof Assume that the assertion is false; that is, suppose that there are
only a finite number of primes, say Pl' P2' ... ,Pn. Consider the positive
integer
a = (PlP2 ... Pn) + 1.
None of the listed primes Pi divides a. If a were divisible by Pi' for instance,
we would then have Pil(a - PlP2 ... Pm), by Theorem 6-1 (5), or Pill; but
this is impossible by part (2) of the same theorem. Since a > 1, Theoren:t
6-11 asserts that it must have a prime factor. Accordingly, a is divisible
by a prime which is not among those enumerated. This arguments shows
that there is no finite listing of the prime numbers.
Having proved the existence of a prime factorization in principal ideal
domains, one is naturally led to the question of uniqueness. Our next
theorem is the focal point of this chapter.
Theorem 6-13. Every principal ideal domain R is a unique factorization
domain.
Proof Theorem 6-11 shows that each noninvertible element 0 =1= a E R,
has a prime factorization. To establish uniqueness, let us suppose that a
can be represented as a product of primes in two ways, say
a = PlP2 ... Pn = qlq2 ... qm (n.::; m),
102

where the Pi and qi are all primes. Since Pll(qlq2 ... qm), it follows that
PI divides some qi (1 .::; i .::; m); renumbering, if necessary, we may suppose
that Pllql. Now, PI and ql are both prime elements of R, with Pllql' so
they must be associates: ql = PI u l for some invertible element u l E R.
Cancelling the common factor PI' we are left with
P2 ... Pn = u l q2 ... qm·
Continuing this argument, we arrive (after n steps) at
1 = U I U 2 ••• unqn+l ... qm.
Since the qi are not invertible, this forces m = n. It has also been shown
that every Pi has some qj as an associate and conversely. Thus, the two
prime factorizations are identical, apart from the order in which the factors
appear and from replacement of factors by associates.
Attention is called to the fact that the converse of Theorem 6-13 is not
true; in the next chapter, we shall give an example of a unique factorization
domain which is not a principal ideal domain.
A useful fact to bear in mind is that in a unique factorization domain
R, any irreducible element pER is necessarily prime. For, suppose that P
divides the product ab, say pc = abo Let
a = PI ... Pn' b = ql ... qm' and c = tl ... ts

be the unique factorizations of a, b, and c into irreducible factors. We then

have

Since the factorization of ab into irreducibles is unique, the element P must

be an associate of one of the Pi or qi' and, consequently, P divides either
a or b.
Another interesting class of integral domains, which we propose to look
at now, is provided by the so-called Euclidean domains; these arose out of
attempts to generalize the familiar Division Algorithm for ordinary integers
to arbitrary rings. The precise definition follows.
Definition 6-8. An integral domain R is said to be Euclidean if there
exists a function J (the Euclidean valuation) such that the following
conditions are satisfied:
1) J (a) is a nonnegative integer for every 0 1- a E R;
2) for any a, b E R, both nonzero, J(ab) ?: J(a);
3) for any a, b E R, with b 1- 0, there exist elements q, r E R (the quotient
and remainder) such that a = qb + r, where either r = 0 or else
J(r) < J(b).
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 103

Although property (2) seems unsymmetric, R is a commutative ring;

hence, (2) also asserts that J(ab) 2 J(b) as well as J(ab) 2 J(a).
As simple examples of Euclidean domains, we may take
1) any field F, with valuation defined by i(a) = 1 for all nonzero a E F;
2) the ring Z, with valuation defined by J(a) = lain
for all nonzero a E Z
(fixed n E Z+);
3) the Gaussian integers Z(i), with valuation defined by J(a + bi) = a2 + b2
for all nonzero a + bi E Z(i) (see Theorem 6-17).
Several rudimentary properties of Euclidean domains appear in the
lemma below.
Lemma. Let R be a Euclidean domain with valuation J. Then,
1) for each nonzero a E R, J(a) 2 J(I);
2) if two nonzero elements a, b E R are associates, then J(a) = J(b);
3) an element 0 =1= a E R is invertible if and only if J(a) = J(I).
Proof Assertion (1) follows from the fact that a = aI, whence
J(a) = J(al) 2 J(I).
If a and b are associates, then a = bu, with u an invertible element of
R ; then, also, b = au - 1. This means that
J(a) = J(bu) 2 J(b),
which says that J(a) = J(b).
To prove (3), suppose that a =1= 0 has an inverse in R, so that ab = 1 for
some choice of bE R. Then, using (1),
J(a) :s; J(ab) = J(I) :s; J(a),
or J(a) = 1. Conversely, suppose that the element 0 =1= a E R is such that
J(a) = 1. Applying the division algorithm to 1 and a, there exist q, r E R
for which
1 = qa + r,
where r = 0 or J(r) < J(a). The latter alternative implies that J(r) < 1,
which is impossible, so that r = 0; in other words, 1 = qa and a is invertible
in R.
Theorem 6-14. The quotient and remainder in condition (3) of the
definition of a Euclidean domain are unique if and only if
J(a + b):S; max{J(a),J(b)}.
Proof Suppose that there exist nonzero a, bE R such that J(a + b) >
max {J(a), J(b)}; then,
b = O(a + b) + b = l(a + b) - a,
104

with both <5( -a) = <5(a) < <5(a + b) and <5(b) < <5(a + b). This exhibits the
lack of uniqueness of quotient and remainder in condition (3).
Conversely, assume that the indicated inequality holds and that the
element a E R has two representations;
a = qb + r (r = 0 or c5(r) < <5(b»,
a = q'b + r' (r' = 0 or <5(r') < c5(b»)
with r =1= r' and q =1= qt. Then we have

c5(b) ~ <5((q - q')b) = <5(r - r') < max {<5(r), <5( - r')} < c5(b).

This is only possible if one of r - r' or q - q' is zero. Since each of these
conditions implies the other, uniqueness follows.
Corollary. (Division Algorithm for Z). If a, bE Z, with b =1= 0, then there
exist unique integers q and r such that
a = qb + r, o~ r < lal.
Proof Utilize the valuation <5 given by <5(a) = 14 for all nonzero a E Z.
Unique factorization in Z follows ultimately from the Division Algo-
rithm. It is not surprising that in rings where there is an analog of division
with remainder, we can also prove uniqueness of factorization. The main
line of argument consists of showing that every Euclidean domain is a
principal ideal domain. (One need only consider the ring Ze to see that the
converse of this does not hold.)
Theorem 6-15. Every Euclidean domain is a principal ideal domain.
Proof Let R be a Euclidean domain with valuation <5 and I be an ideal of
R; ignoring trivial cases, we may suppose that I =1= {O}. Consider the set
S defined by
S = {<5(a)la E I; a =1= O}.
Since S is a nonempty subset of nonnegative integers, it has a least element
by the Well-Ordering Principal. Pick bEl, so that <5(b) is minimal in S.
Our contention is that I = (b).
Let a be an arbitrary element of I. By the definition of Euclidean
domain, there exist elements q, r E R for which a = qb + r, where either
r = 0 or <5(r) < <5(b). Now, r = a - qb E I, since I is an ideal containing
both a and b. The alternative <5(r) < <5(b) would therefore contradict the
minimality of c5(b). Consequently, we must have r = 0, and a = qb E (b);
this implies that I £; (b). The reverse inclusion clearly holds, since bEl,
thereby completing the proof.
Corollary. Every Euclidean domain is a unique factorization domain.
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 105

It would seem inappropriate to conclude this chapter without some

mention of the quadratic number fields; the elements of these domains form
the sets
Q(Jn) = {a + bJnla, b E Q},
with n =1= 1 a square-free integer (that is, an integer not divisible by the
square of any positive integer > 1). When n < 0, we may view Q(Jn) as
a subdomain of the complex number system C and represent its elements in
the standard form a + bJn i. It is not difficult to show that if nl, n 2 are
square-free integers, then Q(Jn l ) = Q(.Jn 2 ) if and only if n l = n2 .
Each element a = a + bJn E Q(Jn) gives rise to another element
a = a - bJn of Q(Jn), which we shall call the conjugate of a (for n < 0,
a is the usual complex conjugate of a). A simple argument establishes that
the mappingf: Q(Jn) -+ Q(Jn) defined by f(a) = a is an isomorphism.
To study divisibility properties of Q(Jn), it is convenient to make use
of the concept of the norm of an element (an analog of the absolute value
notion in Z):
Definition 6-9. For each element a = a + bJn in Q(Jn), the norm N(a)
of a is simply the product of a and its conjugate a:
N(a) = aa = (a + bJn)(a - bJn) = a 2 - b2n.
Some properties of the norm function which follow easily from the
definition are listed below.
Lemma. For all a, p E Q(Jn), the following hold:
°
1) N(a) = if and only if a = 0;
2) N(ap) = N(a)N(fJ);
3) N(I) = 1;
Proof Given a = a + bJn in Q(Jn), N(a) = a2 - b2n = if and only °
°
if both a = b = (that is, a = 0); otherwise, we would contradict the
choice of n as a square-free integer.
Since the mapping f(a) = a is an isomorphism, N is a multiplicative
function in the sense that
N(afJ) = apap = apaJj = aapJj = N(a)N(fJ)
for all a, p E Q(Jn).
The proof of assertion (3) follows from the fact that
N(I) = N(12) = N(I)N(I) = N(I)2,
whence N(I) = 1.
Although Q(J1i.) has been labeled as a field, we actually have not proved
this to be the case; it is high time to remedy this situation.
106

Theorem 6-16. For each square-free integer n, the system Q(.jn) forms
a field; in fact, Q(J"ii) is a subfield of C.
Proof The reader may easily verify that Q(-jn) is a commutative ring with
identity. It remains only to establish that each nonzero element of Q(.jn)
°
has a multiplicative inverse in Q(.jn). Now, if =1= a E Q(.jn), then the
element P = a/N(a) evidently lies in Q(.jn); furthermore, the product
ap = a(a/N(a) ) = N(a)jN(a) = 1,
so that Pserves as the inverse of a.
Contained in each quadratic field Q(.jn) is the integral domain

Z(.jn) = {a + b-jnla, b E Z}.

Since Z(.jn) is closed under conjugation, the norm function enables us to
get a clear view of the sets of invertible and irreducible elements in these
domains. The multiplicative property of the norm, for instance, transfers
any factorization a = PI' of an element a E Z(.jn) into a factorization
N(a) = N(P)N(y) of the integer N(a). This is particularly helpful in proving
Lemma. For any a E Z(-jn), the following hold: _
1) N(a) = ± 1 if and only if a is invertible in Z(.jn);
2) if N(a) = ±p, where p is a prime number, then a is an irreducible
element of Z(-jn).
Proof As regards (1), observe that if N(a) = ± 1, then aa = ± 1; thus all,
which is to say that a is invertible. To prove the converse, let a be an
invertible element of Z(-jn), so that ap = 1 for some P in Z(.jn). Then,

N(ex)N(p) = N(ap) = N(l) = 1.

Since N(a) and N(P) are both integers, this implies that N(a) = ± 1.
Next, suppose that a has the property that N(a) = ±p, where p is a
°
prime number. As N(a) =1= 0, 1, the element ex is neither nor invertible in
Z(-jn). If a = PI' is a factorization of a in Z(.jn), then
N(P)N(y) = N(a) = ±p,

from which it follows that one of N(P) or N(y) must have the value ± 1.
From the first part of the lemma, we may thus conclude that either P or y
is invertible in Z(.jn), while the other is an associate of a. Accordingly, a
is an irreducible element of Z(.jn).
Example 6-3. Let us find all invertible elements in Z(i) = Z(R), the
domain of Gausian integers, by finding those members a of Z(i) for which
N(a) = 1 (in this setting, the norm assumes nonnegative values). If
a = a + bi E Z(i) and N(a) = 1, then a2 + b2 = 1, with a, b E Z. This
DIVISIBILITY THEORY IN INTEGRAL DOMAINS 107

equation is possible only if a = ± 1 and b = 0, or a = and b = °

Hence, the only choice for invertible elements in Z(i) are ± 1 and ± i.
± l.

Perhaps the most obvious approach to the question of unique factoriza-

tion in the quadratic domains Z(-Jn) is to try to sllow that they are Euclidean
domains (a natural candidate for the Euclidean valuation is J(a) = IN(a)I).
In the coming theorem, w~ shall do precisely this for the domains Z(-J -1),
Z(-J -2), Z(.j2), and Z(-J3). Although there are other Euclidean quadratic
domains, our attention is restricted to these few for which the division
algorithm is easily established.
Theorem 6-17. Each of the domains Z(-Jn), where n = -1, -2,2,3, is
Euclidean; hence, is a unique factorization domain.
Proof The strategy employed in the proof is to show that the function J
defined on Z(-J"ii) by J(a) = IN(a) Iis a Euclidean valuation for n = -1, -2,
°
2,3. We clearly have J(a) = if and only if a = 0, so that J(a) 2 1 for all
a =f 0. Since both the norm and its absolute value are multiplicative,
condition (2) of Definition 6-8 is always satisfied:

J(afJ) = J(a)J(fJ) 2 J(a)·1 = J(a)

°
whenever a, P =f are in Z(-J"ii).
Since P =f 0, the product ap-l E Q(-J"ii) and so may be written in the
form ap- 1 = a + b-J"ii, with a, b E Q. Select integers x and y (the nearest
integers to a and b) such that
la - xl :s; 1/2, Ib - yl :s; 1/2.
Now, set (1 = x + y-Jn. Then (1 E Z(-Jn) and norm formula (valid also in
Q(-Jn)) shows that
IN(ap-l - (1)1 = IN((a - x) + (b - y)-Jn) I
= I(a - X)2 - n(b _ y)21.
But the manner in which x and y were chosen imply that
-n14 :s; (a - X)2 - n(b - y)2 :s; 1/4, if n > 0,
°:s; (a - X)2 - n(b - y)2 :s; 1/4 + (- n)1/4, if n < 0.
In terms of the function J this means
J(ap- 1 - (1) = I(a - X)2 - n(b - yfl < 1

for n = -1, - 2, 2, 3. Putting p = p(ap- 1 - (1), we have a = (1P + p.

Since a and (1P are in Z(-Jn), this equation implies that p is in Z(-Jn) also.
Moreover, for the previously indicated values of n,
J(p) = J(p(ap- 1 - (1)) = J(fJ)J(ap-l - (1) < J(p).
108

Definition 6-8 is therefore satisfied in its entirety when n = -1, - 2,2, 3 and
the corresponding quadratic domains Z(Jn) are Euclidean.
Corollary. The domain Z(i) of Gaussian integers is a unique factoriza-
tion domain.
Remark. By investigating further the divisibility properties of Z(i), one can
prove the classic "two squares theorem" of Fermat: every prime number
ofthe form 4n + 1 is the sum oftwo squares; the interested reader is advised
to consult [13] for the details.
Example 6-4. The various integral domains studied in this chapter might
suggest that unique factorization of elements always holds. To round out
the picture with an example of the failure of unique factorization, let us
consider the quadratic domain Z(.j - 5). Observe that the element 9 has
two factorizations in Z(.j - 5), namely,

9 = 3·3 = (2 + J - 5)(2 - .j - 5).

Clearly, no two of the factors 3, 2 + J - 5, and 2 ~j=5 are associates of
each other (the only invertible elements in Z(.j - 5) are ± 1), and it can
easily be shown that they are all irreducible. Indeed, were anyone of

a1 = 3, a2 = 2 + J - 5, a3 = 2 - J - 5
reducible, we could write a i = py, where neither of the elements p,
y E Z(.j - 5) is invertible. Taking norms, it follows that
9 = N(ai) = N(P)N(y), N(P), N(y) E Z +,
which in turn yields N(P) = N(y) = 3. Hence, if p = a + we find bR,
that we must solve the equation a2 + 5b 2 = 3 for integers a and b; but
°
this equation obviously has no solutions in Z (b 1= implies that a 2 + 5b 2 ~ 5,
and if b = 0, then a 2 = 3). Thus, we have exhibited two genuinely different
factorizations of the element 9 into irreducibles, so that unique factorization
does not hold in Z(.j - 5).
Notice further that the common divisors of 9 and 3(2 + j=5) are
1, 3, and 2 + .j - 5. N one of these latter elements is divisible by the others,
so that gcd (9, 3(2 + J - 5)) fails to exist (in particular, Z(j=5) does not
have the gcd-property). On the other hand, the greatest common divisor
of 3 and 2 + j=5 does exist and, in fact, gcd (3, 2 + j=5) = 1. It follows
that only the right-hand side of the formula

gcd (3·3, 3(2 + .j -5)) = 3 gcd (3, 2 + J -5)

is defined in Z(J - 5), thereby illustrating the remark on page 95.
This example has the additional feature of showing that the concepts
PROBLEMS 109

of an irreducible element and of a nonzero prime do not always coincide

in an arbitrary integral domain. Specifically, we have (2 + .J-=S)J3· 3, but
(2 + j=5)i3, so that 2 + j=5 cannot be a prime element of Z(j=5).

PROBLEMS

1. Let R be a commutative ring with identity and the elements a, e E R, with e2 = e.

Prove that
a) If (a) = (e), then a and e are associates. [Hint: a = (1 - e + a)e.]
b) If for some n E Z + the elements a' and e are associates, then am and e are
associates for all m ~ n.
2. Given that 1, J, and K are ideals of a principal ideal domain R, derive the following
relations:
a) If 1 = (a) and J = (b), then 1J = (ab); in particular, 1" = (a').
b) I(J II K) = 1J II IK.
c) 1 +(J II K) = (1 + J) II (1 + K).
d) 1 (J + K) = (1 II J) + (1 11K).
II
e) 1J = 1 II J if and only if 1 + J = R for all nonzero 1, J.
3. Suppose that R = Rl EB R2 EB ... EB R., where each Ri is a principal ideal ring.
Verify that R is also a principal ideal ring.
4. Let R be an integral domain having the gcd-property. Assuming that equality
holds to within associates, prove that, for nonzero a, b, e E R,
a) gcd (a, gcd (b, e») = gcd (gcd (a, b), c).
b) gcd (a, 1) = 1.
c) gcd (ea, eb) = e gcd (a, b); in particular, gcd (e, eb) = 1.
d) if gcd (a, b) = 1 and gcd (a, c) = 1, then gcd (a, be) = 1.
e) if gcd (a, b) = 1, ale and ble, then able.
f) gcd (a, b) Icm (a, b) = abo [Hint: Theorem 6-6.]
5. If R is an integral domain having the gcd-property, show that a nonzero element
of R is prime if and only if it is irreducible.
6. In a principal ideal domain R, establish that the primary ideals are the two trivial
ideals and ideals of the form (p'), where p is a prime element of Rand n E Z+.
[Hint: If 1 is primary, then .Ji= (p) for some prime element p. Choose n E Z+
such thatl s; (p'), butl $ (p.+l), and show thatl = (p').]
7. If R is a principal ideal doman, we define the length A(a) for each nonzero a E R
as follows: if a is invertible, then A(a) = 0; otherwise, A(a) is the number of primes
(not necessarily distinct) in any factorization of a. Prove the following assertions:
a) the length of a is well-defined;
b) if alb, then A(a) :;:; A(b);
c) if alb and A(a) = A(b), then bla;
d) if a .t band b .t a, then there exist nonzero p, q E R such that

A(pa + qb) :;:; min {A(a), A(b)}.

110

8. Let a be a nonzero element of the principal ideal domain R. If a has length n,

prove that there are at most 2" ideals containing a.
9. Verify that any two nonzero elements of a unique factorization domain possess a
greatest common divisor. [Hint: If a = p~lp~2 ... p~r and b = plt'p~ ... p~ (Pi
irreducible), then gcd (a, b) = P{lpil ... ptr, where ji = min (k i , [J]
10. Let R be an integral domain. Prove that R is a unique factorization domain if and
only if every nontrivial principal ideal of R is the product of a finite number of
maximal principal ideals and these ideals are unique up to a permutation of order.
11. Show that o(a) = lal is not a Euclidean valuation on the domain Q.
12. Assuming that R is a Euclidean domain with valuation 0, prove the statements
below:
a) For nonzero a, bE R, if alb and o(a) = o(b), then a and b are associates. [Hint:
Show that bla.]
b) For nonzero a, bE R, o(ab) > o(a) if and only if b is not an invertible element.
[Hint: Use the division algorithm to write a = q(ab) + r.]
c) If n is any integer such that 0(1) + n ~ 0, then the function 0': R - {O} -+ Z
defined by o'(a) = o(a) + n is also a Euclidean valuation on R.
13. For each ideal I in Z(i), the domain of Gaussian integers, establish that the quotient
ring Z(i)/I is finite. [Hint: Write I = (a) and use the division algorithm on a and
any PE Z(i).]
14. Let R be a Euclidean domain with valuation o.
a) Determine whether the set I = {a E Rlo(a) > 0(1)} v {O} is an ideal of R.
b) Assuming that the set F = {a ElRlo(a) = 1} v {O} is closed under addition,
verify that F forms a field.
15. a) Prove that if n i and n 2 are square-free integers and n i ~ n 2 , then the quadratic
field Q(.jn I) is not isomorphic to Q(.jn;).
b) For each square-free integer n, determine all the subfields of the quadratic
field Q(.jn).
16. Establish the following assertions (where n is a square-free integer):
a) For n < -1, the only invertible elements of the quadratic domain Z(.jn) are
±l.
b) For n > 1, Z(.jn) has infinitely many invertible elements. [Hint: If aI' b i
is a solution of the equation a2 - nb 2 = ± 1, conclude that ak, bk is also a
solution, where ak + bk.jn = (a l + bI.jn)k, k E Z+.]
c) The invertible elements of Z(.j2) are precisely the elements of the form
±(1 + .j2)", n E Z+. [Hint: If u is any positive invertible element of Z(.j2),
then (1 + .j2)" ::;; u < (1 + .j2)"+ 1 for some n E Z + ; hence,
+ .j2)-" < 1 + .j2.
1 ::;; u(1
Assuming that u(1 + .j2)-" = a + b.j2, show that a = 1, b = 0.]
17. a) Factor each of the following into primes: 11 + 7i in Z(i); 4 + 7.j2 in Z(.j2);
4- RinZ(R).
PROBLEMS 111

b) Show that in the quadratic domain Z(.)6), the relation 6 = (.)6)2 = 3·2 does
not violate unique factorization.
18. Prove that the domain Z(0) is not a unique factorization domain by discovering
two distinct factorizations of the element 10. Do the same for element 9 in the
domain Z(.J - 7). ~
19. Show that the quadratic domain Z(.J -5) is not a principal ideal domain. [Hint:
Consider the ideal (3, 2 + .J - 5).]
20. Describe the field of quotients of the quadratic domain Z(.jYt) where n is a square-
free integer.
SEVEN

POLYNOMIAL RINGS

The next step in our program is to apply some of the previously developed
theory to a particular class of rings, the so-called polynomial rings. For
the moment, we shall merely remark that these are rings whose elements
consist of "polynomials" with coefficients from a fixed, but otherwise
arbitrary, ring. (The most interesting results occur when the coefficients
are specialized to a field.)
As a first order of business, we seek to formalize the intuitive idea of
what is meant by a polynomial. This involves an excursion around the
fringes of the more general question of rings of formal power series. Out
of the veritable multitude of results concerning polynomials, we have
attempted to assemble those facets of the theory whose discussion reinforces
the concepts and theorems expounded earlier; it is hoped thereby to convey
a rough idea of how the classical arithmetic of polynomials fits into ideal
theory. Our investigation concludes with a brief survey of some of the
rudimentary facts relating roots of polynomials to field extensions.
To begin with simpler things, given an arbitrary ring R, let seq R denote
the totality of all infinite sequences
f = (aD, aI' a 2, ... , ak, ... )
of elements ak E R. Such sequences are called formal power series, or merely
power series, over R. (Our choice of terminology will be justified shortly.)
We intend to introduce suitable operations in the set seq R so that the
resulting system forms a ring containing R as a subring. At the outset, it
should be made perfectly clear that two power series
f = (aD, aI' a2, ... ) and g = (b o, bl , b2, ... )
are considered to be equal if and only if they are equal term by term:
f = g if and only if ak = bk for all k ;:::.: O.
Now, power series may themselves be added and multiplied as follows:
f + g = (aD + bo, a l + bl , ... ),

fg = (CO, C I ' C 2 ' ••• ),

112
POLYNOMIAL RINGS 113

where, for each k ~ 0, Ck is given by

Ck= L
i+j=k
aibj = aOb k + aIb k- 1 + ... + ak-Ib l + akb o·

(It is understood that the above summation runs over all integers i, j ~ 0
subject to the restriction that i + j = k.)
A routine check establishes that with these two definitions seq R becomes
a ring. To verify a distributive law, for instance, take

One finds quickly that

f(g + h) = (ao, aI' ... )(b o + Co' b l + C 1 , ... ) = (do, d l , ... ),

where
dk =
i+j=k
L ai(bj + c) =
i+j=k
L (aib j + aicj )

A similar calculation of fg + fh leads to the same general term, so that

f(g + h) = fg + fh. The rest of the details are left to the reader's care.
We simply point out that the sequence (0,0,0, ... ) serves as the zero element
of this ring, while the additive inverse of an arbitrary member (a o, aI' a2 , ... )
of seq R is, of course, (-a o, -aI' -a 2 , ... ). To summarize what we know
so far:
Theorem 7-1. The system seq R forms a ring, known as the ring of
(formal) power series over R. Furthermore, the ring seq R is commuta-
tive with identity if and only if the given ring R has these properties.
If S represents the subset of all sequences having 0 for every term beyond
the first, that is, the set
S = {(a, 0, 0, .. .)ia E R},
then it is not particularly difficult to show that S constitutes a subring of
seq R which is isomorphic to R; one need only consider the mapping that
sends the sequence (a, 0, 0, ... ) to the element a. In this sense, seq R contains
the original ring R as a subring.
Having reached this stage, we shall no longer distinguish between an
element a E R and the special sequence (a, 0, 0, ... ) of seq R. The elements
of R, regarded as power series, are hereafter called constant series, or just
constants.
With the aid of some additional notation, it is possible to represent power
series the way we would like them to look. As a first step in this direction,
we let ax designate the sequence
(0, a, 0, 0, ... ).
114

That is, ax is the specific member of seq R which has the element a for its
°
second term and for all other terms. More generally, the symbol ax n ,
n ~ 1, will denote the sequence
(0, ... , 0, a, 0, ... ),
where the element a appears as the (n + l)st term in this sequence; for
example, we have
ax 2 = (0,0, a, 0, ... )
and
ax 3 = (0, 0, 0, a, 0, ... ).
By use of these definitions, each power series

may be uniquely expressed in the form

f = (a o, 0, 0, ... ) + (0, aI' 0, ... ) + ... + (0, ... ,0, an' 0, ... ) + ...
= ao + alx + a 2 x 2 + ... + anxn + ...
with the obvious identification of ao with the sequence (a o, 0, 0, ... ). Thus,
there is no loss in regarding the power series ring seq R as consisting of all
formal expressions
f = ao + alx + a2 x 2 + ... + anxn + ...,
where the elements ao, aI' ... , an' ... (the coefficients of f) lie in R. As a
notational device, we shall often write this as f = L akxk (the summation
symbol is not an actual sum and convergence is not at issue here).
Using sigma notation, the definitions of addition and multiplication of
power series assume the form
L akxk + L bkxk = L (a k + bk)x\
(L akxk)(L bkxk) = L ckx\
where

We should emphasize that, according to our definition, x is simply a

new symbol, or indeterminant, totally unrelated to the ring R and in no
sense represents an element of R. To indicate the indeterminant x, it is
common practice to write R[[xJJ for the set seq R, andf(x) for any member
of the same. From now on, we shall make exclusive use of this notation.
Remark. If the ring R happens to have a multiplicative identity 1, many
°
authors will identify the power series + 1x + OX2 + OX3 + ... with x,
thereby treating x itself as a special member of R[[x JJ ; namely, the sequence
POLYNOMIAL RINGS 115

x = (0,1,0,0, ... ). From this view, ax becomes an actual product of

members of R[[ x J] :
ax = (a, 0, 0, ... )(0, 1,0,0, ... ).

Concerning the notation of power series, it is customary to omit terms

with zero coefficients and to replace (-ak)x k by -akxk • Although x is not
to be considered as an element of R[[xJ], we shall nonetheless take the
liberty of writing the term 1Xk as Xk (k ~ 1). With these conventions, one
should view, for example, the power series
1 + x2 + X4 + ... + x2n + ... E Z[[ X J]
as representing the sequence (1, 0, 1, 0, ... ).
An important definition in connection with power series is that of order,
given below.
Definition 7-1. If f(x) = L
akx k is a nonzero power series (that is, if
not all the ak = 0) in R[[x J], then the smallest integer n such that
°
an =1= is called the order of f(x) and denoted by ord f(x).
Suppose f(x), g(x) E R[[xJ], with ordf(x) = nand ord g(x) = m, so
that
f(x) = an~ + an+1 x n+1 + .. . (an =1= 0),
g(x) = bmxm + bm+1 x m+1 + .. . (b m =1= 0).
From the definition of multiplication in R[[x J], the reader may easily check
that all coefficients of f(x)g(x) up to the (n + m)th are zero, whence
f(x)g(x) = anbmx n+m + (a n+1 bm + anbm+l)Xn+m+l + ....
If we assume that one of an or bmis not a divisor of zero in R, then anb m =1=
and
°
ord (j(x)g(x)) = n +m = ordf(x) + ord g(x).

This certainly holds if R is taken to be an integral domain, or again if R has

an identity and one of an or bm is the identity element.
The foregoing argument serves to establish the first part of the next
theorem; the proof of the second assertion is left as an exercise.
Theorem 7-2. Iff(x) and g(x) are nonzero power series in R[[xJ], then
°
1) either f(x)g(x) = or ord (j(x)g(x)) ~ ord f(x) + ord g(x),
with equality if R is an integral domain;
2) either f(x) + g(x) = or °
ord (j(x) + g(x)) ~ min {ord f(x), ord g(x)}.
The notation of order can be used to prove the following corollary.
116

Corollary. If the ring R is an integral domain, then so also is its power

series ring R[[ x JJ.
Proof We observed earlier that whenever R is a commutative ring with
identity, these properties carryover to R[[ x JJ. To see that R[[ x JJ has no
zero divisors, select f(x) =1= 0, g(x) =1= 0 in R[[ x JJ. Then,
ord (J(x)g(x») = ordf(x) + ord g(x) > 0;
hence, the product f(x)g(x) cannot be the zero series.
Although arbitrary power series rings are of some interest, the most
important consequences arise on specializing the discussion to power series
whose coefficients are taken from a field. These will be seen to form principal
ideal domains and, in consequence, unique factorization domains. The
following intermediate result is directed towards establishing this fact.
Lemma. Let R be a commutative ring with identity. A formal power
series f(x) = I akxk is invertible in R[[xJJ if and only if the constant
term ao has an inverse in R.
Proof If f(x)g(x) = 1, where g(x) = I bkxk, then the definition of multi-
plication in R[[ x JJ shows that aob o = 1; hence, a o is invertible as an
element of R.
For the converse, suppose that the element ao has an inverse in R. We
proceed inductively to define the coefficients of a power series I bkxk in
R[[xJJ which is the inverse off(x). To do this, simply take bo = ao l and,
assuming b l , bz, ... , bk- I have already been defined, let
bk = -aol(albk_1 + aZb k- Z + ... + akb o)·
Then aob o = 1, while, for k ~ 1,
ck = I a;b j = aOb k + aIb k- 1 + ... + akb o = O.
;+j=k
By our choice of the bk'S, we evidently must have (I akxk)(I bkxk) = 1,
and so I akxk possesses an inverse in R[[xJJ.
Corollary. A power series f(x) = I akxk E F[[ x JJ, where F is a field,
has an inverse in F[[ x JJ if and only if its constant term ao =1= O.
Having dealt with these preliminaries, we are now ready to proceed to
describe the ideal structure of F[[ x JJ.
Theorem 7-3. For any field F, the power series ring F[[xJJ is a principal
ideal domain; in fact, the nontrivial ideals of F[[ x JJ are of the form
(Xk), where k E Z +.
Proof Let I be any proper ideal of F[[xJJ. Either I = {O}, in which case
I is just the principal ideal (0), or else I contains nonzero elements. In the
POLYNOMIAL RINGS 117

latter event, choose a nonzero power series f(x) E I of minimal order.

Suppose thatf(x) is of order k, so that
f(x) = akxk + ak+1xk+l + ... = xk(a k + ak+l x + ... ).
Since the coefficient ak =1= 0, the previous lemma insures that the power
series ak + ak + 1 X + ... is an invertible element of F[[ x]] ; in other words,
f(x) = xkg(X), where g(x) has an inverse in F[[ x]J. But, then,
Xk = f(x)g(X)-l E I,
which leads to the inclusion (Xk) ~ 1.
On the other hand, take h(x) to be any nonzero power series in I, say of
order n. Since f(x) is assumed to have least order among all members of I,
it is clear that k ~ n; thus, h(x) can be written in the form
h(x) = xk(bnxn- k + bn+1Xn-k+l + ... ) E (x k).
This implies that I ~ (x k), and the equality I = (Xk) follows.
Corollary 1. The ring F[[x]] is a local ring with (x) as its maximal
ideal.
Proof. Inasmuch as the ideals of F[[ x]] form a chain
F[[x]] ~ (x) ~ (x 2) ~ ... ~ {O},
the conclusion is obvious.
Corollary 2. Any nonzero element f(x) E F[[ x]] can be written in the
formf(x) = g(x)x\ where g(x) is invertible and k ~ 0.
To this we add, for future reference, the following assertion regarding
the maximal ideals of a power series ring over a commutative ring with
identity.
Theorem 7-4 Let R be a commutative ring with identity. There is a
one-to-one correspondence between the maximal ideals M of the ring
R and the maximal ideals M' of R[[x]] in such a way that M' corresponds
to M if and only if M' is generated by M and x; that is, M' = (M, x).
Proof. Assume that M is a maximal ideal of R. To see that M' = (M, x)
forms a maximal ideal of the ring R[[ x]], we need only show that for any
power series f(x) = L akxk ¢ M', the element 1 + g(x)f(x) E M' for some
g(x) in R[[x]] (Problem 2, Chapter 5). Since the series f(x) does not lie in
M', its constant term ao ¢ M; hence, there exists an element r E R such that
1 + rao E M. This implies that
1 + rf(x) = (1 + ra o) + r(a 1 + a 2x + ... + anx n- 1 + ···)x E (M, x),
and so M' is a maximal ideal, as required.
118

Next, take M' to be any maximal ideal of R[[x]] and define the set M
to consist of the constant terms of power series in M' :
M = {aoERILakxkEM'}.
The reader can painlessly supply a proof that M forms a maximal ideal of
the ring R. Notice incidentally that M must be a proper ideal. Were M = R,
then there would exist a power series L bnxn in M' with constant term
bo = 1. By the last lemma, L bnxn would then be an invertible element, so
that M' = R[[ x]], which is impossible. Owing to the inclusion M' ~ (M, x)
and the fact that M' is maximal in R[[ x]], it now follows that M' = (M, x).
To verify that the correspondence in question is indeed one-to-one,
suppose that (M, x) = (M, x), where M, M are both maximal ideals of the
ring R; what we want to prove is that M = M. Let rEM be arbitrary.
Given f(x) E R[[x]], the sum r + f(x)x E (M, x) = (M, x), so that
r + f(x)x = r + g(x)x
for appropriate r E M and g(x) E R[[x]J. Hence, r - r= (g(x) - f(x) )x.
If g(x) - f(x) =1= 0, then, upon taking orders,
° = ord(r - r) = ord (g(x) - f(x») + ord x ~ 1,
an absurdity. In consequence, we must have g(x) - f(x) = which, in its
turn, forces r = rEM. The implication is that M ~ M and, since M is
°
maximal, we end up with M = M. This completes the proof of the theorem.
Power series have so far received all the attention, but our primary
concern is with polynomials.
Definition 7-2. Let R[x] denote the set of all power series in R[[x]]
whose coefficients are zero from some index onward (the particular index
varies from series to series) :
R[x] = {a o + a 1 x + ... + anXnlak E R; n ~ o}.
An element of R[ x] is called a polynomial (in x) over the ring R.
In essence, we are defining a polynomial to be a finitely nonzero sequence
of elements of R. Thus, the sequence (1, 1, 1, 0, 0, ... ) would be a polynomial
over Z2' but (1, 0, 1,0, ... , 1,0, ... ) would not.
H is easily verified that R[x] constitutes a subring of R[[x]], the so-
called ring of polynomials over R (in an indeterminant x); indeed, if
f(x) = L akxk, g(x) = L bkxk are in R[x], with ak = °
for all k ~ nand
bk = °for all k ~ m, then
ak + bk = ° for k ~ max {m, n},
L
i+j=k
aib j = ° for k ~ m +-n,
POLYNOMIAL RINGS 119

so that both the sum f(x) + g(x) and product f(x)g(x) belong to R[ x].
Running parallel to the idea of the order of a power series is that of the
degree of a polynomial, which we introduce at this time.
Definition 7-3. Given a nonzero polynomial

in R[ x], we call an the leading coefficient of f(x); and the integer n, the
degree of the polynomial.
The degree of any nonzero polynomial is therefore a nonnegative integer;
no degree is assigned to the zero polynomial. Notice that the polynomials
of degree 0 are precisely the nonzero constant polynomials. If R is a ring
with identity, a polynomial whose leading coefficient is 1 is said to be a
monic polynomial.
As a matter of notation, we shall hereafter write degf(x) for the degree
of any nonzero polynomialf(x) E R[ x].
The result below is similar to that given for power series and its proof
is left for the reader to provide; the only change of consequence is that we
now use the notion of degree rather than order.
Theorem 7-5. If f(x) and g(x) are nonzero polynomials in R[ x], then
1) either f(x)g(x) = 0 or deg (J(x)g(x)) ::;; degf(x) + deg g(x), with
equality whenever R is an integral domain;
2) either f(x) + g(x) = 0 or

deg (J(x) + g(x)) ::;; max {degf(x), deg g(x)}.

Knowing this, one could proceed along the lines of the corollary to
Theorem 7-2 to establish
Corollary. If the ring R is an integral domain, then so is its polynomial
ring R[x].
Example 7-1. As an illustration of what might happen if R has zero divisors,
consider Z 8' the ring of integers modulo 8. Taking
f(x) = 1 + 2x, g(x) = 4 + x + 4x 2 ,
we obtainf(x)g(x) = 4 + x + 6x 2 , so that

deg (J(x)g(x)) = 2 < 1 + 2 = degf(x) + deg g(x).

Although many properties of the ring R carryover to the associated

polynomial ring R[ x], it should be pointed out that for no ring R does
R[x] form a field. In fact, when R is a field (or, for that matter, an integral
domain), no element of R[ x] which has positive degree can possess a
120

multiplicative inverse. For, suppose that f(x) E R[ x], with degf(x) > 0;
if f(x)g(x) = 1 for some g(x) in R[ x], we could obtain the contradiction
o= deg 1 = deg (J(x)g(x)) = degf(x) + deg g(x) 1- o.
The degree of a polynomial is used in the factorization theory of R[ x]
in much the same way as the absolute value is employed in Z. For, it is
through the degree concept that induction can be utilized in R[ x] to develop
a polynomial counterpart of the familiar division algorithm. One can
subsequently establish that the ring F[x] with coefficients in a field forms a
Euclidean domain in which the degree function is taken to be the Euclidean
valuation.
Before embarking on this program, we wish to introduce several new
ideas. To this purpose, let R be a ring with identity; assume further that
R' is any ring containing R as a subring (that is, R' is an extension of R)
and let r be an arbitrary element of R'. For each polynomial

f(x) = ao + alx + ... + anxn

in R[ x], we may define f(r) E R' by taking
f(r) = ao + alr + ... + anrn.

The elementf(r) is said to be the result of substituting r for x inf(x). Suffice

it to say, the addition and multiplication used in defining f(r) are those of
the ring R', not those of R[x].
Now, suppose thatf(x), g(x) are polynomials in R[x] and rEcent R'.
We leave the reader to prove that if

h(x) = f(x) + g(x), k(x) = f(x)g(x),

then
h(r) = f(r) + g(r), k(r) = f(r)g(r).
This being so, it may be concluded that the mapping cPr: R[x] --+ R' which
sendsf(x) tof(r) is a homomorphism of R[x] into R'. Such a homomorphism
will be called the substitution homomorphism determined by r and its range
denoted by the symbol R[r] :

R[r] = {f(r)lf(x) E R[x]}

= {aD + alr + ... + anrnlakER; n ;;::: a}.
It is a simple matter to show that R[r] constitutes a subring of R' ; in fact,
R[r] is the subring of R' generated by the set R u {r}. (Since R has an
identity element 1, 1x = X E R[x], and so r E R[r].) Notice also that
R[r] = R if and only if r E R.
The foregoing remarks justify part of the next theorem.
POLYNOMIAL RINGS 121

Theorem 7-6. Let R be a ring with identity, R' an extension ring of R,

and the element rEcent R'. Then there is a unique homomorphism
<Pr: R[ x] --+ R' such that <Pr(x) = r, <Pr(a) = a for all a E R.
Proof We need only verify that <Pr is unique. Suppose, then, that there is
another homomorphism ,: R[x] --+ R satisfy in the indicated conditions
and consider any polynomial f(x) = ao + a1x + ... + anxn E R[x]. By
assumption, ,(ak ) = ak for each coefficient ak , while ,(xk ) = ,(X)k = ~.
Taking stock of the fact that, is a homomorphism,
,(j(x») = ,(ao) + ,(a1),(x)+ ... + ,(an),(x)n
= ao + a1r + ... + anrn = f(r) = <Pr(j(x»).
This proves that, = <p" yielding the uniqueness conclusion.
Without some commutativity assumption, the above remarks need not
hold. For, if we let
h(x) = (x - a)(x - b) = x 2 - (a + b)x + ab,
then
h(r) = r2 - (a + b)r + abo
Lacking the hypothesis that rEcent R', it cannot be concluded that
(r - a)(r - b) = r2 - ar - rb + ab
will equal h(r); in other words, h(x) = f(x)g(x) does not always imply
h(r) = f(r)g(r).
Wheneverf(r) = 0, we call the element r a root or zero of the polynomial
f(x). Of course, a given polynomial f(x) E R[x] may not have a root in R;
we shall see later that when R is a field, there always exists an extension
field R' of R in whichf(x) possesses a root. It is perhaps appropriate to point
out at this time that the problem of obtaining all roots of a polynomial
f(x) E R[x] is equivalent to that of finding all elements r E R' for which
f(x) E ker <Pr.
After this brief digression, let us now state and prove the division algo-
rithm for polynomials.
Theorem 7-7. (Division Algorithm). Let R be a commutative ring with
identity and f(x), g(x) =/= 0 be polynomials in R[ x], with the leading
coefficient of g(x) an invertible element. Then there exist unique poly-
nomials q(x), r(x) E R[ x] such that
f(x) = q(x)g(x) + r(x),
where either r(x) = 0 or deg r(x) < deg g(x).
Proof The proof is by induction on the degree of f(x). First, notice that if
f(x) = 0 or f(x) =/= 0 and degf(x) < deg g(x), a representation meeting the
122

requirements of the theorem exists on taking q(x) = 0, r(x) = f(x). Further-

more, if degf(x) = deg g(x) = 0, f(x) and g(x) are both elements of the
ring R, and it suffices to let q(x) = f(x)g(x) -1, r(x) = 0.
This being so, assume that the theorem is true for polynomials of degree
less than n (the induction hypothesis) and let degf(x) = n, deg g(x) = m,
where n ~ m ~ 1; that is,
f(x) = ao + a1x + ... + anx n, an =/= 0,
g(x) = bo + b1x + ... + bmxm , bm =/= ° (n ~ m).
Now, the polynomial
f1(X) = f(x) - (a nb,;;-l)xn- mg(x)
lies in R[x] and, since the coefficient of xn is an - (an b';;- 1)bm = 0, has degree
less than n. By supposition, there are polynomials q1(x), r(x) E R[ x] such
that
f1(X) = q1(X)g(X) + r(x),
where r(x) = °or deg r(x) < deg g(x). Substituting, we obtain the equation
f(x) = (q1(X) + (anb,;;-l)xn-m)g(x) + r(x)
= q(x)g(x) + r(x),
which shows that the desired representation also exists when degf(x) = n.
As for uniqueness, suppose that
f(x) = q(x)g(x) + r(x) = q'(x)g(x) + r'(x),
where r(x) and r'(x) satisfy the requirements of the theorem. Subtracting,
we obtain
r(x) - r'(x) = (q'(x) - q(x»)g(x).
Sincetheleadingcoefficientofg(x)isinvertible,itfollowsthatq'(x) - q(x) =
if and only if r(x) - r'(x) = 0. With this in mind, let q'(x) - q(x) =/= 0.
°
Knowing that bm is not a zero divisor of R,
deg (q'(x) - q(x»)g(x) = deg (q'(x) - q(x») + deg g(x)
~ deg g(x) > deg (r(x) - r'(x»),

a contradiction; the last inequality relies on the fact that the degrees of
r(x) and r'(x) are both less than the degree of g(x). Thus, q'(x) = q(x), which
in turn implies that r'(x) = r(x).
The polynomials q(x) and r(x) appearing in the division algorithm are
called, respectively, the quotient and remainder on dividing f(x) by g(x).
In this connection, it is important to observe that if g(x) is a monic poly-
nomial, or if R is taken to be a field, one need not assume that the leading
coefficient of g(x) is invertible.
POL YNOMIAL RINGS 123

We now come to a series of theorems concerning the factorization

properties of R[ x].
Theorem 7-8. (Remainder Theorem). Let R be a commutative ring
with identity. If f(x) E R[x] and a E R, then there exists a unique
polynomial q(x) in R[x] such thatf(x) = (x - a)q(x) + f(a).
Proof All this is scarcely more than an application of the division algorithm
to the polynomialsf(x) and x-a. We then obtain
f(x) = (x - a)q(x) + r(x),
where r(x) = 0 or deg r(x) < deg (x - a) = 1. It follows in either case
that r(x) is a constant polynomial, say r(x) = r E R. Substitution of a for
x leads to
f(a) = (a - a)q(a) + r(a) = r,
as desired.
Corollary. The polynomial f(x) E R[x] is divisible by x - a if and
only if a is a root off(x).
Let us next show that a polynomial cannot have more roots in an integral
domain than its degree.
Theorem 7-9. Let R be an integral domain andf(x) E R[x] be a non-
zero polynomial of degree n. Then f(x) can have at most n distinct
roots in R.
Proof The argument proceeds by induction on the degree of f(x). When
degf(x) = 0, the result is trivial, since f(x) cannot have any roots. If
degf(x) = 1, for instance, f(x) = ax + b (a =1= 0), then f(x) has at most
one root; indeed, if a is an invertible element, it follows that -a-1b is the
only root off(x). Now, suppose that the theorem is true for all polynomials
of degree n - 1 and let degf(x) = n. If f(x) has a root a, the preceding
corollary gives f(x) = (x - a)q(x), where the polynomial q(x) has degree
n - 1. Any root a' of f(x) distinct from a must necessarily be a root of
q(x) for, by substitution,
o= f(a') = (a - a')q(a')
and, since R has no zero divisors, q(a') = O. From our induction hypothesis,
q(x) has at most n - 1 distinct roots. As the only roots of f(x) are a and
those of q(x),f(x) cannot possess more than n distinct roots in R.
With this step forward we can establish
Corollary 1. Let f(x) and g(x) be two nonzero polynomials of degree n
over the integral domain R. If there exist n + 1 distinct elements
ak E R (k = 1, 2, ... , n + 1) such that f(a k ) = g(ak ), then f(x) = g(x).
124

Proof The polynomial h(x) = f(x) - g(x) is such that deg h(x) ~ nand,
by supposition, has at least n + 1 distinct roots in R. This is imposs~ble
unless h(x) = 0, whence f(x) = g(x).
Corollary 2. Let f(x) E R[ x], where R is an integral domain, and let S
be any infinite subset of R. Iff(a) = 0 for all a E S, thenf(x) is the zero
polynomial.
Example 7-2. Consider the polynomial x P - x E Zp[x], where p is a prime
number. Now, the nonzero elements of Zp form a commutative group under
mUltiplication of order p - 1. Hence, we have aP-l = 1, or aP = a for
every a =1= O. This is equally true if a = O. Our example shows that it may
very well happen that every element of the underlying ring is a root of a
polynomial, yet the polynomial is not zero.
With the Division Algorithm at our disposal, we can prove that the
ring F[ x] is rich in structure.
Theorem 7-10. The polynomial ring F[x], where F is a field, forms a
Euclidean domain.
Proof As has been noted, F[ x] is an integral domain. Moreover, the
function t5 defined by t5(j(x») = degf(x) for any nonzero f(x) E F[x] is a
suitable Euclidean valuation. Only condition (2) of Definition 6-6 fails to
be immediate. But iff(x) and g(x) are two nonzero polynomials in F[x],
Theorem 7-5 implies that

c5(j(x)g(x») = deg (j(x)g(x»)

= degf(x) + deg g(x) ;:::: degf(x) = t5(j(x») ,

since deg g(x) ;:::: O. Thus, the function t5 satisfies the requisite properties of
a Euclidean valuation.
The reader is no doubt anticipating the corollary below.
Corollary. F[ x] is a principal ideal domain; hence, a unique factoriza-
tion domain.
For a less existential proof of the fact that F[x] is a principal ideal
domain and a considerably more precise description of its ideals, one can
repeat (with appropriate modifications) the pedestrian argument used in
Theorem 2-3. It will appear that any nontrivial ideal I of F[ x] is of the
form I = (j(x»), where f(x) is a nonzero polynomial of minimal degree in 1.
Since a field is trivially a unique factorization domain, part of the last
corollary could be regarded as a special case of the coming theorem.
Theorem 7-11. If R is a unique factoiization domain, then so is R[x].
POL YNOMIAL RINGS 125

Proof Suppose that R[x] is not a unique factorization domain and let S
be the set of all nonconstant polynomials in R[xJ which do not have a
unique factorization into irreducible elements. Select f(x) E S to be of
minimal degree. We may assume that
f(x) = Pl(X)P2(X) ... Pr(x) = ql(X)q2(X) ... qs(x),
where the Pi(X) and qlx) are all irreducible and

m = deg Pl(X) ;;::: deg P2(X) ;;::: ;;::: deg Pr(x),

n = deg ql(X) ;;::: deg q2(X) ;;::: ;;::: deg qs(x),
with n ;;::: m > 0; it is further evident that no Pi(X) = uqj(x) for any-..
invertible element u (otherwise, the polynomial obtained on dividing f(x)
by qlx) will have unique factorization; this implies that f(x) can also be
factored uniquely). Let a, b be the leading coefficients of Pl(X), ql(X), respec-
tively, and define
g(x) = af(x) - bpl(X)Xn - m q2(x) ... qs(x).
On one hand, we have
g(x) = apl(x)P2(x) ... pAx) - bpl(x)xn - m q2(x) ... qs(x)
= Pl(x)(ap2(x) ... Pr(X) - bxn - m q2(x) ... qs(x)),
and, on the other hand,
g(x) = aql(x)q2(x) ... qs(x) - bpl(x)x n - m q2(x) ... qs(x)
= (aql(x) - bpl(X)Xn - m )q2(X) ... qs(x).

Now, either g(x) = 0, which forces aql(x) = bpl(X)Xn - m , or else deg g(x) <
degf(x). In the latter event, g(x) must possess a unique factorization into
irreducibles, some of which are q2(X), ... , qs(x) and Pl(X). The net result of
this is that Pl(x)lg(x), but Pl(X) t qJx) for i > 1, so that

Pl(x)l(aql(x) - bpl(X)X n - m ),

and therefore Pl(x)laql(x). In either of the two cases 'COnsidered, we are

able to conclude that Pl(X) divides the product aql(x); this being so,
aql(x) = Pl(x)h(x) for some polynomial h(x) E R[x]. Since R is taken to be
a unique factorization domain, a has a unique factorization as a product of
irreducible elements of R-hence, of R[xJ-say, a = C 1C2 ••• Ck , where each
C i is irreducible in R[x]. (The only factorizations of a as an element of
R[xJ are those it had as an element of R.) Arguing from the representation
C 1 C2 ••• ckql(X) = Pl(x)h(x),
126

with PI(X) an irreducible, it follows that each Ci and, in consequence, the

element a divides h(x). But, then,
aql(x) = PI(x)ahl(x)
for some hl(x) in R[x] or, upon canceling, ql(X) = PI(x)hl(x); in other
words, PI(X)!ql(X). Using the irreducibility of ql(X) as a member of R[x],
PI(X) must be an associate of ql(X). However, this conflicts with our original
assumptions. Thus, we see that R[x] is indeed a unique factorization
domain.
Remark. For many years, it was an open question as to whether a power
series ring over a unique factorization domain is again a unique factorization
domain; a negative answer was established not long ago by Samuel [55].
To this we might add, on the positive side, that one can prove that the ring
of formal power series over a principal ideal domain does in fact comprise
a unique factorization domain (a not altogether trivial task).
Coming back to the corollary to Theorem 7-10, there is an interesting
converse which deserves mention: namely, if R is an integral domain such
that the polynomial ring R[x] forms a principal ideal domain, then R is
necessarily a field. In verifying this, the main point to be proved is that any
nonzero element a E R is invertible in R. By virtue of our hypothesis, the
ideal generated by x and a must be principal; for instance,
(x, a) = (j(x)), °f f(x) E R[x].
Since both x, a E (j(x)), it follows that
a = g(x)f(x), and x = h(x)f(x)
for suitable g(x), h(x) in R[ x]. The first of these relations signifies that
degf(x) = 0, say f(x) = aD, and as a result deg h(x) = 1, say h(x) = bo +
b1x (b l f 0). We thus obtain x = ao(b o + b1x). But this means that the
product aOb l = 1, thereby making aD (or, equivalently, f(x)) an invertible
element of R. The implication is that the ideal (x, a) is the entire ring R[x].
It is therefore possible to write the identity element in the form
1 = xkl(x) + ak 2(x),
with the two polynomials kl(x), k2(X) E R[x]. This can only happen if
aco = 1, where Co f 0 is the constant term of k2(X). In consequence, the
element a has a multiplicative inverse in R, which settles the whole affair.
At the heart of all the interesting questions on factorization in R[ x]
lies the idea of an irreducible polynomial, which we formulate in a rather
general way as follows:
Definition 7-4. Let R be an integral domain. A nonconstant poly-
nomial f(x) E R[x] is said to be irreducible over R, or is an irreducible
POLYNOMIAL RINGS 127

polynomial in R[ x], if f(x) cannot be expressed as the product of two

polynomials (in R[x]) of positive degree. Otherwise, f(x) is termed
reducible in R[x].
In the case of the principal ideal domain F[ x], where F is a field, the
irreducible polynomials are precisely the irreducible elements of F[ x] (recall
that the invertible elements of the polynomial ring F[ x] are just the non-
zero constant polynomials); by Theorem 5-9, these coincide with the prime
elements of F[ x]. Of the equivalent notions, irreducible polynomial,
irreducible element, and prime element, the term "irreducible polynomial"
is the one customarily preferred for F[ x ].
Perhaps we should emphasize that Definition 7-4 applies only to poly-
nomials of positive degree; the constant polynomials are neither reducible
nor irreducible. Thus, the factorization theory of F[ x] concerns only
polynomials of degree ;;::: 1.
The dependence of Definition 7-4 upon the polynomial domain R[x]
is essential. It may very well happen that a given polynomial is irreducible
when viewed as an element of one domain, yet reducible in another. One
such example is the polynomial x 2 + 1; it is irreducible in R # [x], but
reducible in both C[x], where x 2 + 1 = (x + i)(x - i), and Z2[X], where
x 2 + 1 = (x + l)(x + 1). Thus, to ask merely whether a polynomial is
irreducible, without specifying the coefficient ring involved, is incomplete
and meaningless.
More often than not, it is a formidable task to decide when a given
polynomial is irreducible over a specific ring. If F is a finite field, say one
of the fields of integers modulo a prime, we may actually examine all of the
possible roots. To cite a simple illustration, the polynomial f(x) = x 3 +
X + 1 is irreducible in Z2[X]. If there are any factors of tlJ.is polynomial,
°
at least one must be linear. But the only possible roots for ftx) are and 1,
yetf(O) = f(l) = 1 =1= 0, showing that no roots exist in Z2.
Example 7-3. Any linear polynomial ax + b, a =1= 0, is irreducible in R[x],
where R is an integral domain. Indeed, since the degree of a product of two
polynomials is the sum of the degree of the factors, it follows that a
representation
ax + b = g(x)h(x), g(x), h(x) E R[ x],
with 1 ::; deg g(x), 1 ::; deg h(x) is impossible. This signifies that every
reducible polynomial has degree at least 2.
Example 7-4. The polynomial x 2 - 2 is irreducible in Q[ x], where Q as
usual is the field of rational numbers. Otherwise, we would have
x 2 - 2 = (ax + b) (ex + d)
= (ae)x2 + (ad + be)x + bd,
128

with the coefficients a, b, e, d E Q. Accordingly,

ae = 1, ad + be = 0, bd = -2,
whence e = 1/a, d = -2/b. Substituting in the relation ad + be = 0, we
obtain
°
= -2a/b + b/a = (-2a 2 + b 2)/ab.

Thus, -2a 2 + b 2 = 0, or (b/a)2 = 2, which is impossible because )2 is

not a rational number. Although irreducible in Q[ x], the polynomial x 2 - 2
isnonethelessreducibleinR#[x];inthiscase,x2 - 2 = (x - .jfJ(x + .j2)
and both factors are in R # [x].
For ease of reference more than to present new concepts, let us sum-
marize in the next theorem some of the results of previous chapters (specifi-
cally, Theorems 5-5 and 6-7) as applied to the principal ideal domain F[x].
Theorem 7-12. If F is a field, the following statements are equivalent:
1) f(x) is an irreducible polynomial in F[x];
2) the principal ideal (j(x)) is a maximal (prime) ideal of F[x];
3) the quotient ring F[ x ]/(j(x)) forms a field.
The theorem on prime factorization of polynomials is stated now.
Theorem 7-13. (Unique Factorization in F[x]). Each polynomial
f(x) E F[ x] of positive degree is the product of a nonzero element of
F and irreducible monic polynomials of F[x]. Apart from the order
of the factors, this factorization is unique.
Suffice it to say, Theorem 7-13 can be made more explicit for particular
polynomial domains. When we deal with polynomials over the complex
numbers, the crucial tool is the Fundamental Theorem of Algebra.
Theorem 7-14. (The Fundamental Theorem of Algebra). Let C be the
field of complex numbers. If f(x) E C[x] is a polynomial of positive
degree, thenf(x) has at least one root in C.
Although many proofs ofthe result are available, none is strictly algebraic
in nature; thus, we shall assume the validity of Theorem 7-14 without proof.
The reader will experience little difficulty, however, in establishing the
following corollary.
E C[ x] is a polynomial of degree n > 0, then f(x)
Corollary 1. If f(x)
can be expressed in C[x] as a product of n (not necessarily distinct)
linear factors.
Another way of stating the corollary above is that the only irreducible
polynomials in C[x] are the linear polynomials. Directing our attention
POLYNOMIAL RINGS 129

now to the real field, we can obtain the form of the prime factorization in
R # [x] (bear in mind that polynomials with coefficients from R # are poly-
nomials in C[x] and therefore have roots in C).
Corollary 2. If f(x) E R # [ x] is of positive degree, then f(x) can be
factored into linear and irreducible quadratic factors.
Proof Since f(x) also belongs to C[ x ],j(x) factors in C[ x] into a product
of linear polynomials x - Ck, CkEC. If CkER#, then x - CkER#[xJ.
Otherwise, Ck = a + bi, where a, b E R # and b =f O. But the complex roots
of real polynomials occur in conjugate pairs (Problem 7-11), so that
ck = a - bi is also a root off(x). Thus,
(x - ck) (x - ck) = x2 - 2ax + (a 2 + b2 ) E R#[x]
is a factor of f(x). The quadratic polynomial x 2 - 2ax + (a 2 + b2 ) is
irreducible in R # [x], since any factorization in R # [x] is also valid in C[ x}
and (x - ck ) (x - ck ) is its unique factorization in C[xJ.
An interesting remark, to be recorded without proof, is that if F is a
finite field, the polynomial ring F[x] contains irreducible polynomials of
every degree (see Theorem 9-10).
This may be a convenient place to introduce the notion of a primitive
polynomial.
Definition 7-5. Let R be a unique factorization domain. The content
of a nonconstant polynomial f(x) = ao + alx + ... + anxn E R[x],
denoted by the symbol contf(x), is defined to be the greatest common
divisor of its coefficients:
contf(x) = gcd (a o, aI' ... , an)·
We callf(x) a primitive polynomial if contf(x) = 1.
Viewed otherwise, Definition 7-5 asserts that a polynomialf(x) E R[x]
is primitive if and only if there is no irreducible element of R which divides
all of its coefficients. In this connection, it may be noted that in the domain
F[ x] of polynomials with coefficients from a field F, every nonconstant
polynomial is primitive (indeed, there are no primes in F). The reader should
also take care to remember that the notion of greatest common divisor and,
in consequence, the content of a polynomial is not determined uniquely,
but determined only to within associates.
Given a polynomialf(x) E R[x] of positive degree, it is possible to write
f(x) = Cfl(X), where C E Rand fl(X) is primitive; simply let C = contf(x).
To a certain extent this reduces the question of factorization in R[x] (at
least, when R is a unique factorization domain) to that of primitive poly-
nomials. By way of specific illustrations, we observe that f(x) = 3x 3 -
130

4x + 35 is a primitive polynomial in Z[ x], while g(x) = 12x 2 + 6x - 3 =

3(4x 2 + 2x - 1) is not a primitive element of the same, since g(x) has
content 3.
Here is another new concept: Suppose that I is a (proper) ideal of R, a
commutative ring with identity. There is an obvious mapping v: R[ x] --+
(R/I)[x]; for any polynomialf(x) E R[x] simply apply nat l to the coefficients
ofj(x), so that
v(J(x)) = (a o + I) + (a 1 + I)x + ... + (an + I)x n,

or, more briefly, v(J(x)) = L (natIak)x k. The reader will encounter no

difficulty in verifying that v, defined in this way, is a homomorphism of
R[x] onto (R/1)[x], the so-called reduction homomorphism modulo 1. The
polynomial v(J(x)) is said to be the reduction of f(x) modulo I.
Although it might seem to be rather special, the reduction homomorphism
will serve us in good stead on several occasions. We make immediate use
of it to characterize primitive polynomials.

Theorem 7-15. Let R be a unique factorization domain and let f(x) =

ao + a 1 x + ... + anxn E R[x], with degf(x) > o. Then f(x) is a
primitive polynomial in R[x] if and only if, for each prime element
pER, the reduction of f(x) modulo the principal ideal (p) is nonzero.
Proof By definition, the reduction off(x) modulo (p) is

v(J(x)) = (a o + (p)) + (a 1 + (p))x + ... + (an + (p))xn.

Thus, to say that v(J(x)) = 0 for some prime pER is equivalent to asserting
that ak E (p), or rather, plak for all k. But the latter condition signifies that
contf(x) +- 1; hence,f(x) is not primitive.
One of the most crucial facts concerning primitive polynomials is
Gauss's Lemma, which we prove next.

Theorem 7-16. (Gauss's Lemma). Let R be a unique factorization

domain. Iff(x), g(x) are both primitive polynomials in R[x], then their
productf(x)g(x) is also primitive in R[x].
Proof Given a prime element pER, (p) is a prime ideal of R, whence the
quotient ring R' = R/(p) forms an integral domain. We next consider the
reduction homomorphism v modulo the principal ideal (p). Since R'[ x] is
an integral domain, it follows that the reduction of f(x)g(x) cannot be the
zero polynomial:
v(J(x)g(x)) = v(J(x) )v(g(x)) +- o.
The assertion of the theorem is now a direct consequence of our last result.
POLYNOMIAL RINGS 131

Corollary. If R is a unique factorization domain andf(x), g(x) E R[x],

then
cont (j(x)g(x)) = contf(x) cont g(x).
Proof As noted earlier, we can write f{x) = afl(x), g(x) = bg1(x), where
a = contf(x), b = cont g(x) and where fl(X), gl(X) are primitive in R[xJ.
Therefore,f(x)g(x) = abfl(x)gl(X). According to the theorem, the product
fl(X)gl(X) is a primitive polynomial of R[x]. This entails that the content
off(x)g(x) is simply ab, or, what amounts to the same thing, contf(x)cont g(x).
Any unique factorization domain R, being an integral domain, possesses
a field of quotients K = Qc,(R) and we may consider the ring of polynomials
R[ x] as imbedded in the polynomial ring K[ xJ. The next theorem deals
with the relation between the irreducibility of a polynomial in R[ x] as
compared to its irreducibility when considered as an element of the larger
ring K[ xJ. (The classic example of this situation is, of course, the poly-
nomial domain Z[x] s;;; Q[xJ.) Before concentrating our efforts on this
relationship, we require a preliminary lemma.
Lemma. Let R be a unique factorization domain, with field of quotients
K. Given a nonconstant polynomialf(x) E K[x], there exist (nonzero)
elements a, bE R and a primitive polynomial fl(X) in R[x] such that
f(x) = ab- 1fl(X).
Furthermore, fl(X) is unique up to invertible elements of R as factors.
Proof Inasmuch as K is the field of quotients of R,f(x) can be written in the
form
f(x) = (a obi)l) + (a 1 b1 1)x + ... + (a nb;;l)xn,
where ai' bi E Rand bi =1= o. Take b to be any common m6ltiple of the bi ;
for instance, b = bob1 ••• bn. Then b =1= 0 and, since the coefficients of
bf(x) all lie in R, we have bf(x) = g(x) E R[xJ. Accordingly,
f(x) = b- 1 g(x) = ab- 1fl(X),
where fl(X) E R[x] is a primitive polynomial and a = cont g(x). We
emphasize that fl(X) is of the same degree as f(x), so cannot be invertible
in R[xJ.
As for uniqueness, suppose thatf(x) = ab- 1fl(X) = cd- 1f2(x) are two
representations that satisfy the conditions of the theorem. Then,
adfl(x) = bcf2(X).
Since fl(X) and fix) are both primitive, the corollary to Gauss's Lemma
implies that we must have ad = ubc for some invertible element u E R.
In consequence, fl(X) = Uf2(X), showing that fl(X) is unique to within
invertible factors in R.
132

Theorem 7-17. Let R be a unique factorization domain, with field of

quotients K. Iff(x) E R[x] is an irreducible primitive polynomial, then
it is also irreducible as an element of K[x].
Proof Assume to the contrary that f(x) is reducible over K. Then,
f(x) = g(x)h(x), where the polynomials g(x), h(x) are in K[x] and are of
positive degree. By virtue of the lemma just proven,

with a, b, c, dE Rand gl(X), h1(x) primitive in R[x]. Thus,

bdf(x) = acg1(x)h1(x).

Now, Gauss's Lemma asserts that the product gl(x)h1(x) is a primitive

polynomial in R[x], whencef(x) and gl(x)h1(x) differ by an invertible element
of R:
f(x) = ugl(x)h1(x).
Since deg gl(X) = deg g(x) > 0, deg h1(x) = deg h(x) > 0, the outcome is
a nontrivial factorization off(x) in R[x], contrary to hypothesis.
There is an obvious converse to Theorem 7-17, viz.: if the primitive
polynomial f(x) E R[x] is irreducible as an element of K[ x], it is also
irreducible in R[x]. This is justified by the fact that R[x] (or an isomorphic
copy thereof) appears naturally as a subring of K[ x]; thus, if f(x) were
reducible in R[ x], it would obviously be reducible in the larger ring K[ x].
Our remarks lead to the following conclusion: Given a primitive poly-
nomial f(x) E R[ x], R a unique factorization domain, f(x) is irreducible in
R[x] if and only iff(x) is irreducible in K[x].
Our next concern is a generalization of a famous theorem of Eisenstein
dealing with the problem of irreducibility (this result is of fundamental
importance in the classical theory of polynomials with integral coefficients).
The generalization which we have in mind is formulated below.
Theorem 7-18. Let R be an integral domain and the nonconstant
polynomialf(x) = ao + a1x + ... + anxn E R[x]. Suppose that there
exists a prime ideal P of R such that
1) an ¢ P, 2) ak E P for 0 ::; k < n,
Thenf(x) is irreducible in R[x].
Proof Assume that, contrary to assertion, f(x) is reducible in R[x]; say,
f(x) = g(x)h(x) for polynomials g(x), h(x) E R[x], where

g(x) = bo + b1x + ... + b,x',

h(x) = Co + c1x + ... + c.x' (r +s= n; r, S > 0).
POLYNOMIAL RINGS 133

Now consider the reduction off(x) modulo the ideal P. Invoking hypothesis
(2), it can be inferred that
v(g(x))v(h(x)) = v(J(x)) = (an + p)xn.
Since the polynomial ring (R/P)[x] comprises an integral domain, the only
possible factorizations of (an + P)xn are into linear factors. This being so,
a moment's reflection shows that
v(g(x)) = (b r + P)xr ,
v(h(x)) = (c s + P)xs•
This means that the constant terms of these reductions are zero; that is,
bo + P = Co + P = P.
Altogether we have proved that both bo, Co E P, revealing at the same time
that a o = boc o E p2, which is untenable by (3). Accordingly, no such
factorization of f(x) can occur, and f(x) is indeed irreducible in R[ x J.
Theorem 7-18 leads almost immediately to the Eisenstein test for
irreducibility.
)
Corollary. (The Eisenstein Criterion). Let R be a unique factorization
domain and K be its field of quotients. Letf(x) = a o + a1x + ... +
an xn be a nonconstant polynomial in R[x J. Suppose further that for
some prime pER, P fan' plak for 0 :::;; k < n, and p2 {a o. Then,f(x) is
irreducible in K[xJ.
Proof We already know that (p) is a prime ideal of R. Taking stock of
the theorem, f(x) is an irreducible polynomial of R[ x] ; hence, of K[ x] (at
this point a direct appeal is made to Theorem 7-17).
This is probably a good time at which to examine some examples.
Example 7-5. Consider the monic polynomial
f(x) = xn + aEZ[x] (n > 1),
where a =/= ± 1 is a nonzero square-free integer. For any prime p dividing
a, p is certainly a factor of all the coefficients except the leading one, and
our hypothesis ensures that p2 {a. Thus, f(x) fulfils Eisenstein's criterion,
and so is irreducible over Q. Incidentally, this example shows that there
are irreducible polynomials in Q[x] of every degree.
Ontheotherhand,noticethatx4 + 4 = (x 2 + 2x + 2)(x 2 - 2x + 2);
one should not expect Theorem 7-18 to lead to a decision in this case, since,
of course, 4 fails to be a square-free integer.
Example 7-6. Eisenstein's test is not directly applicable to the cyclotonic
polynomial
134

cll(x) = x P - 1 = XP-l + xP- 2 + ... + X + 1 E Z[ x] (p prime),

x-I
because no suitable prime is available. This problem is resolved by the
observation that cll(x) is irreducible in Z[ x] if and only if cll(x + 1) is
irreducible. A simple computation yields

cll(x + 1) = (x + 1)P - 1 = (xp + (f)XP-l + ... + px)/x

(x + 1) - 1
= XP-l + (nXP-2 + ... + P E Z[xJ.
If the Eisenstein criterion is now applied, it is easy to see that all the require-
ments for the irreducibility of cll(x + 1) in Z[x] are satisfied (in the binomial
coefficient (&) = p!jk!(P - k)!, the numerator is divisible by p for k < p,
but not the denominator). One finds in this way that the original cyclotonic
polynomial cll(x) must be irreducible in Z[x]; hence, also as a polynomial
in Q[xJ.

Starting with a ring R we can first form the polynomial ring R[x], with
indeterminant x, and then the polynomial ring (R[ x ])[y] in another in-
determinant y. As the notation indicates, the elements of (R[ x ])[y] are
simply polynomials .
g = fo(x) + fl(X)Y + ... + J,,(x)yn,
where each coefficientf,.(x) E R[x], so that
f,.(x) = a Ok + a 1kx + ... + amkkXmk (aij E R).
Consequently, g can be rewritten as a polynomial in x and y,
m n
g = g(x, y) = L L aijxiy,
i=O j=O

with m, n nonnegative integers and aij elements of R (one makes the obvious
conventions that aooxOyO = aoo, aiOxiyo = aiOxi and aOjxOy = aOjY). In
accordance with tradition, we shall hereafter denote (R[ x ])[y] by R[ x, y]
and refer to the members of this set as polynomials over R in two indeter-
minants x and y.
Two such polynomials with coefficients aij and bij are by definition
equal if aij = bij for all i and j. Addition of polynomials is performed
termwise, while multiplication is given by the rule:

where
C ij = L
k+k'=i
ak1bk'I'·
1+1'=j
POLYNOMIAL RINGS 135

With these operations, the set R[x, y] becomes a ring containing R (or rather
an isomorphic copy of R) as a subring.
The (total) degree of a nonzero polynomial
m n
f(x, y) = L L aijxiyj
i=O j=O

is the largest of the integers i + j for which the coefficient aij +- and is
denoted, as before, by degf(x, y). Without going into details here, let us
°
simply state that it is possible to obtain inequalities involving degrees
analogous to those of Theorem 7-5; in particular, if R is an integral domain,
we still have
deg (j(x, y)g(x, y») = degf(x, y) + deg g(x, y).

From this rule, one can subsequently establish that whenever R forms an
integral domain, then so does the polynomial ring R[ x, y].
Rather than get involved in an elaborate discussion of these matters,
we content ourselves with looking at two examples. :>

Example 7-7. To illustrate that the ideal structure of the ring F[x, y]
(F a field) is more complicated than that of F[ x], let us show that F[ x, y]
is not a principal ideal domain. This is accomplished by establishing that
the ideal I = (x, y) is not principal, where
I = U(x, y)x + g(x, y)ylf(x, y), g(x, y) E F[x, y]}.
Notice that the elements of this ideal are just the p01ynomials in F[x, y]
having zero constant term.
Suppose that I was actually principal, say L; = (h(x, y»), where
deg h(x, y) 2 1. Since both x, y E I, there would exist polynomials f(x, y),
g(x, y) ip F[ x, y] satisfying
x = f(x, y)h(x, y), y = g(x, y)h(x, y).
Now, deg x = deg y = 1, which implies that deg h(x, y) = 1, and degf(x, y) =
deg g(x, y) = 0; what amounts to the same thing,
x = ah(x, y), y = bh(x, y) (a, b E F).
Moreover, h(x, y) must be a linear polynomial; for instance,
(C i E F).
But if the coefficient C2 +- 0, then x cannot be a multiple of h(x, y), and if
C 1 +- 0, y is not a multiple of h(x, y). This being the case, we conclude
that C 1 = c2 = 0, a contradiction to the linearity of h(x, y), and so I does
not form a principal ideal.
136

Another point worth mentioning is that since F[ x] constitutes a unique

factorization domain, so does (F[x])[y] = F[x, y] (Theorem 7-11). The
present situation thus furnishes us with an illustration of a unique factoriza-
tion domain which is not a principal ideal domain.
Example 7-8. This example is given to substantiate a claim made earlier
that a primary ideal need not be a power of a prime ideal. Once again,
consider the ideal 1 = (x, y) of the ring F[ x, y], where F is a field. If the
polynomial f(x, y) ¢ I, f(x, y) necessarily has a nonzero constant term a o.
But a o lies in the ideal (1,J(x, y)) and so 1 = ai) lao E (1,J(x, y)), forcing
this ideal to be the entire ring. In consequence, 1 = (x, y) is a maximal
(hence, prime) ideal of the ring F[ x, yJ. (The maximality of 1 could other-
wise be deduced from the fact that it is the kernel of the substitution homo-
morphismf(x, y) -+ f(O, 0).)
Next, let us look at the ideal (x 2 , y) of F[ x, y J. As the reader may verify
12 = (x 2, xy, y2) S (x 2, y) s I.
Inasmuch as -J(x 2 , y) = I, Problem 25, Chapter 5, guarantees that (x 2 , y)
is primary. A straightforward argument shows that (X2, y) is not the power
of any prime ideal of F[x, yJ. For, in the contrary case, (x 2, y) = pn, where
p is a prime ideal and n ;;::: 1. Since pn s I, with 1 prime, we may appeal
to Problem 30, Chapter 5, to conclude that PsI. By the same token, the
inclusions 12 S pn S P, coupled with the fact that P is a prime ideal, yield
1 S P. Hence, 1 = P, so that 1" = (X2, y). Now, the element x E I, while
x ¢ (X2, y), implying that n ;;::: 1. On the other hand, y E (x 2, y), but y ¢ 12 =
(x 2, xy, y2), which means 12 c (X2, y) C 1. These inclusion relations show
that it is impossible to have 1" = (x 2 , y) for any n ;;::: 1.
Let us close this phase of our investigation by saying that there is no
difficulty in extending the above remarks to polynomials in a finite set of
indeterminants. For any ring R, just define recursively

It would not be out of place to devote the remainder of this chapter to

the matter of field extensions (most notably, algebraic extensions) and'
splitting fields. The concepts are presented here partly for their own interest
and partly to lay a foundation for a proof of the celebrated Wedderburn
theorem on finite division rings (Theorem 9-11). We shall have neither
occasion nor space for more than a passing study, and certain topics are
touched upon lightly.
By an extension F' of a field F, we simply mean any field which contains
F as a subfield. For instance, the field of real numbers is an extension of
Q, the rational number field. In view of Theorem 4-12, it may be remarked
POL YNOMIAL RINGS 137

that every field F is an extension of a field isomorphic to Q or to Zp'

according as the characteristic of F is 0 or a prime p.
Assume that F' is an extension field of a field F and let the elements
r l , r2' ... ,rn all lie in F'. The subfield of F' generated by the subset
F u {rl' r2, ... , rn} is customarily denoted by F(rl' r 2, .. ·, rn):
F(rl' r2, ... , rn) = n {KIK is a subfield of F'; F £; K; r i E K}.
Thus, F(rl' r2, ... , rn) is an extension field of F containing the elements r i
(clearly it is the smallest such extension) and one speaks of F(rl' r2, ... , rn)
as being obtained by adjoining the r i to F, or by adjunction of the elements
r i to F. The purpose of the coming theorem is to determine, up to iso-
morphism, the structure of all simple extension fields, that is, extension fields
F(r) arising from a field F by the adjunction of a single element r.
Now, for each element r E F', we have at our disposal the substitution
homomorphism <Pr: F[ x] --+ F'; the reader will recall that this is defint\d
by taking <Prf(x) = f(r). As before, the image of F[x] under <Pr is repre-
sented by the symbol F[r] :
F[r] = {f(r)lf(x) E F[x]}.
The set F[r] forms an integral domain (being a subring of the field F')
and therefore has a field of quotients K = QCl(F[r]) in F'. It is apparent
that F u {r} £; F[r] £; K. But F(r) is the smallest subfield of F' to contain
both F and r, whence F(r) £; K. On the other hand, any subring of F'
which contains F and r will necessarily contain the elements of F[r]; in
particular, F[r] £; F(r). Since F(r) is a subfield of F', it must also contain
all the quotients of elements in F[rJ. Thus, K £; F(r) and equality follows.
This leads to the more constructive description of F(r) as F(r) = QCl(F[r]).
The key to classifying simple extensions is the natupe of the kernel of the
substitution homomorphism (bear in mind that ker <Pr consists of all
polynomials in F[x] having r as a root).
Theorem 7-19. Let F' be an extension field of the field F and let r
E F'.
Then either
1) F(r) ~ QCl(F[x]), or else
2) F(r) ~ F[x ]/(J(x)) for some monic irreducible polynomialf(x) EF[ x]
such that f(r) = 0; this polynomial is uniquely determined.
Proof By the Fundamental Homomorphism Theorem, we know that

F[r] ~ F[x]/ker <Pr'

As F[r] is a subring of a field, it must be an integral domain. Hence, ker <Pr
constitutes a prime ideal of the principal ideal domain F[ xJ. One observa-
tion is quite pertinent: ker <Pr cannot be all of F[ x], since the identity element
138

of F[ x] is not mapped onto zero. Now two possibilities arise, either

ker <Pr = {O} or ker <Pr =1= {O}.
Suppose first that ker <Pr = {O}. Then the homomorphism <Pr is
actually an isomorphism and F[r] ~ F[ xJ. In the situation at hand, F[r]
is not a field, since F[ x] fails to be one. However, from Theorem 4--11, we
know that the quotient fields of F[r] and F[ x] are isomorphic under a
mapping induced by the substitution homomorphism. Since F(r) is the
quotient field of F[r], one thus obtains
F(r) ~ Qc)(F[x]).
If the kernel of <Pr is nonzero, then ker <Pr = (j(x)) for some irreducible
polynomial (prime element) of F[x], where f(x) can be taken to be monic.
Because every nonzero prime ideal of F[ x] is maximal, F[ x ]/(j(x)) forms a
field and the same will be true of its isomorphic image F[r J. But F(r) is the
smallest field to contain both F and r, from which it follows that F[r] = F(r);
this leads to the isomorphism
F(r) ~ F[ x ]/(j(x)).

As ker <Pr = (j(x)), a polynomial g(x) E F[ x] has r as a root if and only

if g(x) is divisible by f(x). Accordingly, if g(x) is any monic irreducible
polynomial in F[x] having r as a root, each of f(x) and g(x) divides the
other; since both of these polynomials are monic, this is possible only if
f(x) = g(x). Thus,f(x) is unique, as asserted in the theorem.
Another virtue of the substitution homomorphism is that it permits us
to put the elements of an extension field into one of two essentially different
categories:
Definition 7-6. Let F' be an extension field of a field F. An element
rEF' is said to be algebraic over F if ker <Pr =1= {O}; otherwise, r is
termed transcendental over F.
Definition 7-6 in effect asserts that r is algebraic over F if there exists a
nonzero polynomialf(x) E F[x] such thatf(r) = 0; on the other hand, r is
a transcendental element over F in case f(r) =1= 0 for every nonzero poly-
nomial f(x) in F[ x J. As illustrations of these notions take F' = R # and
F = Q; then, J2 + --/3 is algebraic over Q, being a root of the polynomial
X4 - 10x 2 + 1 E Q[x], while nand 2.fi are both transcendental over Q.
Every element of the field F is trivially algebraic over F, for the element
rEF is a root of the linear polynomial x - r E F[ xJ.
The proof of Theorem 7-19 furnishes us with more detailed information
concerning simple extension fields, which we now state as two separate
theorems. First, however, let us remark that the field of quotients of the
polynomial domain F[ x] is traditionally called the field of rational functions
POL YNOMIAL RINGS 139

over F. Thus, a rational function (in an indeterminant x) over F can be

written as a quotient f(x)/g(x), where f(x) and g(x) +- 0 are both poly-
nomials in F[ x J.
Theorem 7-20. (Simple Transcendental Field Extensions). If rEF' ~ F
is transcendental over F, then F(r) is isomorphic to the field of rational
functions over F. In fact, there is an isomorphism lXr of F(r) into
Qcl(F[x]) such that lXr(r) = x and lXr(a) = a for every a ER.
This theorem completely determines the structure of simple transcen-
dental extensions over F; they are all isomorphic to the field of rational
functions over F, and, hence, to each other. Thus, for instance, Q(n) ~ Q(2..J2).
As regards simple a]gebraic extensions, we have
Theorem 7-21. (Simple Algebraic Field Extensions). If rEF' 2 F is
algebraic over F, then there exists a unique monic irreducible Iloly-
nomial f(x) E F[ x] such that f(r) = o. Furthermore, if g(x) is a poly-
nomial in F[x] for which g(r) = 0, necessarily f(x)ig(x).
The unique polynomial f(x) of Theorem 7-21 is referred to as the
minimum polynomial of (or belonging to) rover F; as the name suggests, the
minimum polynomial of r is the monic polynomial in F[ x] of minimal
degree having r as a root. The degree of an algebraic element rEF' ~ F
is just the degree of its minimum polynomial (this degree is 1 if and only
ifrEF).
In the course of proving Theorem 7-19, we established the surprising
fact that, whenever r is algebraic over F, the integral domain F[r] becomes
a field, so that F(r) = F[rJ. This amounts to saying that every element of
F(r) is of the formf(r), wheref(x) is a polynomial in F[xJ.
Example 7-9. If n +- 1 is any square-free integer, t~ element J"ii E R if ~ Q
is a root of the quadratic polynomial x 2 - n E Q[ x] and, hence, is algebraic
over Q. From the preceding paragraph we know that Q[.Jn] is a field and
so Q[.jn] = Q(.jn); in other words, every member of Q(J"ii) is of the form
f(.jn), where f(x) is a polynomial in Q[ x] :
Q(.jn) = {aD + a 1 J"ii + a2(J"ii)2 + ... + ak(J"ii)kia;EQ;k ~ OJ.
But (.jn)2 = n, (.jn)3 = n.jn, ... , so that Q(.jn) can be described more
simply as
Q(.jn) = {a + bJ"iiia, bE Q}.
That is to say, the simple field extension Q(.jn) is what we referred to as a
quadratic field in Chapter 6.
Notice also that an arbitrary element a + b.jn E Q(J"ii) satisfies the
polynomial
140

The implication is that, not only In,

but every member of Q(Jn)
is algebraic
over Q; our next-proved theorem demonstrates that this is no accident.
Before pressing on, some additional terminology is required. If F' is an
extension field of a field F, F' may be regarded as a vector space over F.
We shall call F' a finite extension of F, if the vector space F' is finite dimen-
sional over F. For example, the complex field C is a finite extension of R#,
with {I, i} serving as a basis for C (over R#). The dimension of F' as a
vector space over F is called the degree of the extension and written [F' : FJ.
The reader who is versed in linear algebra will have no difficulty with
the next few theorems.
Theorem 7-22. If F' is a finite extension of the field F, then every
element of F' is algebraic over F.
Proof Let rEF' and consider the elements 1, r, r2, ... , r", where n = [F': FJ.
These n + 1 powers of r are all in F' and, hence, must be linearly dependent
over F (since F' is a space of dimension n). Thus, there exist elements
bo, b i , ... , bnE F, which are not all zero, such that bol + b i r + ... + bnr" = O.
But then,f(x) = bo + b i x + ... + bnxn is a nonzero polynomial in F[x]
and f(r) = 0, implying that r is algebraic over F.
This leads us to the following concept: An extension field F' of F is.
said to be algebraic if every element of F' is algebraic over F. The content
of Theorem 7-22 is that a finite extension is an algebraic extension.
Theorem 7-23. Let F' be an extension of the field F. Then F' is an
algebraic extension of F if and only if every subring of F' containing F
is a field.
Proof To begin with, suppose that F' is an algebraic extension and let R
be a subring of F' which contains F; F ~ R ~ F'. For any nonzero element
r E R, the inclusion F[r] ~ R certainly holds. Since r is algebraic over F,
we know from what has been established earlier that F[r] coincides with
the field F(r). But then, r- i E F[r] ~ R, making the ring R a field.
As regards the converse, assume that every subring of F' which contains
F forms a field. Given an element 0 =1= rEF', F[r] is a subring of F' con-
taining F and so must be a field; in particular, r- i E F[rJ. Knowing this,
we may infer the existence ofa polynomialf(x) in F[x] such thatf(r) = r- i .
The element r thus becomes a root of the polynomial g(x) = xf(x) - 1
and, hence, is algebraic over F.
We take this opportunity to mention an interesting theorem due to
Steinitz which gives a necessary and sufficient condition for a finite extension
to be simple: If F' is a finite extension of the field F, then F' is a simple
extension if and only if there are only a finite number of subfields of F'
containing F.
POL YNOMIAL RINGS 141

Let us establish a simple, but nonetheless effective, result about succes-

sive extensions.
Theorem 7-24. If F' is a finite extension of F and F" is a finite extension
of F'; then F" is a finite extension of F. Furthermore,
[F":F] = [F":F'][F':F].
Proof An abbreviated proof runs as follows. Suppose that [F': F] = n
and [F": F'] = m. If {a l' a2 , ... , an} is a basis for F' as a vector space over
F and {b 1 , b2 , ... , bm} is a basis for F" over F', then the set of mn elements
of the form ajb j constitutes a basis for F" over F. This implies that
[F":F] = mn = [F":F'][F':F].
We still have a few loose ends to tie together, including a more precise
description of F(r), when r is algebraic over F.
Theorem 7-25. Let F' be an extension field of F and rEF' be algebraic
over F of degree n. Then the elements 1, r, ... , rn - 1 form a basis for
F(r) (considered as a vector space over F).
Proof Let a be any element of F(r) = F[r]. Then there exists a polynomial
g(x) E F[x] such that a = g(r). Applying the division algorithm to g(x) and
the minimum polynomial f(x) of r, we can find q(x) and sex) in F[x]
satisfying
g(x) = q(x)f(x) + sex), (
where either sex) = ° or deg sex) < n. Since fer) = 0, it follows that
s(r) = g(r) = a. If sex) = 0, necessarily a = 0, while if sex) = bo + b 1 x +
... + bmxm is a nonzero polynomial of degree m < n, then a = bo + b1r +
... + bmrm. Therefore, the elements 1, r, ... , rn - 1 generate F(r) as a vector
space over F.
It remains to show that the set {1, r, ... , r"-1} is linearly independent.
Pursuing this aim, let us assume that col + c1 r + ... + Cn _ 1r"-1 = 0,
where the Ck E F. Then the polynomial
hex) = Co + C1 x + ... + Cn _ 1 X n - 1 E F[x]
and clearly her) = 0, so that hex) E ker CPr = (j(x)). This being the case,
hex) = f(x)k(x) for some polynomial k(x) in F[ x]. But if hex) =1= 0, we
obtain
n > deg hex) = degf(x) + deg k(x) ;;:: degf(x) = n,
a manifestly false conclusion. Thlis, the polynomial hex) = 0, which forces
the coefficients Co = C1 = ... = Cn - 1 = 0. The proof that the n elements
1, r, ... , r"-1 constitute a basis for F(r) over F is now complete. '
The statement of Theorem 7-25 can be rephrased in several ways.
142

Corollary 1. If rEF' ;2 F is algebraic of degree n, then every element

of the simple extension F(r) is of the form f(r), where f(x) E F[ x] is a
polynomial of degree less than n:
F(r) = {aD + air + ... + an_irn-ilakEF}.
Corollary 2. If rEF' ;2 F is algebraic of degree n, then F(r) is a finite
(hence, algebraic) extension with [F(r): F] = n.
We include the next theorem for completeness; it is an immediate
consequence of Theorem 7-22 and Corollary 2 above.
Theorem 7-26. Let F' be an extension of the field F. An element rEF'
is algebraic over F if and only if F(r) is a finite extension of F.
From this, it is a short step to
Corollary. Let rEF' ;2 F be algebraic and [F' :F] finite. Then
F' = F(r) if and only if [F(r) : F] = [F': FJ.
Proof By the last-written theorem, F(r) has a finite degree [F(r):FJ. Now,
F(r) is a subspace (over F) of the vector space F'. This corollary is equivalent
to asserting that a subspace is the entire space if and only if the dimensions
of the two are equal.
Example 7-10. Consider the element r = .J2
+ i E C ;2 Q, C as usual
being the complex number field. Then r2 = 1 + 2.J2i, so that (r2 - If =
- 8 or r4 - 2r2 + 9 = o. Thus, r is a root of the polynomial
f(x) = X4 - 2X2 + 9 E Q[x]
and, hence, is an algebraic element over Q. Now, f(x) has the irreducible
factorization over C,
f(x) = (x - )2 + i)(x - .J2 - i)(x + )2 + i)(x + .J2 - i),

which indicates thatf(x) has no linear or quadratic factors in Q[xJ. There-

fore, f(x) is irreducible as a member of Q[ x] and serves as the minimum
polynomial of rover Q; in particular, the element r has degree 4. By
Theorem 7-25, the simple extension Q(r) is a four-dimensional vector space
over Q, with basis
1, r = .J2 + i, r3 = -.J2 + 5i.
At the same time r is a root of the polynomial x 2 - 2.J2x + 3 E R # [x],
with x 2 - 2.J2x + 3 irreducible over R # ; thus, r is of degree 2 over R # .
Example 7-11. For a second illustration, we turn to the extension field
Q(.J2, )3). The elements.J2 and .J3
are clearly algebraic over Q, being roots
of the polynomials x 2 - 2, x 2 - 3 E Q[x], respectively. Our contention is
POLYNOMIAL RINGS 143

that the field Q(.j2, .J3) is actually a simple algebraic extension of Q; in

fact, Q(.j2,.J3) = Q(.j2 + .J3), with.j2 + .J3 algebraic over Q.
Since t!!.e element .J2 + .J3 belongs to Q(.j2, .J3), we certainly have
Q(.j2 + ../3) S; Q(.j2, .J3). As regards the reverse inclusion, a simple
computation shows that
2../2 = (.j2 + .J3)3 - 9(.J2 + ../3)
is a member of Q(../2 + .J3), and therefore so is .J2. But then,
.J3 = (.j2 + .J3) - .j2
also _lies i!!. Q(.J2 + .J3). This leads to the inclusion Q(../2,.J3) S;
Q(.J2 + .J3) and the asserted equality.
To see that r = ../2 + .J3 is an algebraic element over Q, notice that
r2 = 5 + 2.j6, (r2 - 5)2 = 24, and, hence, the polynomial f(x) = X4 -
lOx 2 + 1 has r as a root. One may verify that f(x) is irreducible in Q[ x],
making it the minimum polynomial of r. Perhaps the quickest way to see
this is as follows. Let F' = Q(.J2); then [F':Q] = 2, with basis {l, ../2},
and [F'(.J3):F'] = 2, with basis {I, J3}. From Theorem 7-24, it follows
that [F'(../3): Q] = 4 and a basis for F'(../3) over Q is given by {I, ../2,.J3, ../6}.
But
F'(.J3) = Q(.j2)(.J3) = Q(.j2, ../3) = Q(.j2 + .J3),
and we know that the dimension of Q(.j2_+ .J3) is equal to the degree of
the minimum polynomial of r = .j2 + .J3. ~
Incidentally, there are five fields between Q and Q(.j2, .J3), namely,
Q, Q(.J2), Q(.J3), Q(.j6), and Q(.j2, ../3). Taking stock of Steinitz's theorem
(page 140), it should come as no surprise that Q(ft, .J3) can be generated
by a single element.
Until now, we have always begun by assuming the existence of an exten-
sion field F' of F and then studied the structure of simple extensions F(r)
within F'. The subject can be approached from a somewhat different stand-
point. Given a field F and an irreducible polynomialf(x) E F[x], one may
ask whether it is possible to construct a simple extension F' of F in which
f(x), thought of as a member of F'[ x], has a root. (If degf(x) = 1, then, in
a trivial sense, F is itself the required extension).
To answer this question, we take our cue from Theorem 7-19. For if
such an extension of F can be found at all, it must be of the form F(r), with
r algebraic over F. As pointed out in our earlier discussion, r will possess
a minimum polynomial g(x) which is irreducible in F[x] and such that
F(r) ~ F[x]/(g(x)). This suggests that, when starting with a prescribed
irreducible polynomial f(x) E F[ x], the natural object of interest should be
the associated quotient ring F[ x ]/(j(x)).
After this preamble, let us proceed to some pertinent details.
144

Theorem 7-27. (Kronecker). If f(x) is an irreducible polynomial in

F[x], then there is an extension field of Fin whichf(x) has a root.
Proof For brevity, we shall write I in place of the principal ideal of F[x]
generated by polynomial f(x); that is to say, I = (j(x)). Since f(x) is
assumed to be irreducible, the associated quotient ring F' = F[x]/I is a
field. To see that F' constitutes an extension of F, consider the natural
mapping nat I : F[ x] --+ F'. According to Theorem 4-7, either the restriction
natIIF is the trivial homomorphism or else nat] (F) forms a field isomorphic
to F, where as usual
nat](F) = {a + Iia E F}.
The first possibility is immediately excluded by the fact that
natI(I) = I + I =1= I,
which is the zero element of F'. Therefore, F is imbeddable in the (quotient)
field F' and, in this sense, F' becomes an extension of F.
It remains to be established that the polynomialf(x) actually has a root
in F'. Assuming that f(x) = a o + a 1 x + ... + anxn, then, from the
definitions of coset addition and multiplication,
(a o + I) +
+ I)(x + I) + ... + (an + I)(x + It
(a 1
= ao + a 1x + ... + anxn + I = f(x) + I = 0 + I.
Ifwe now identify an element ak E F with the coset ak + I which it determines
in F' (the fact that F is isomorphic to natI(F) permits this), we obtain
ao + a 1(x + I) + ... + an(x + It = 0,
which is equivalent to asserting that f(x + I) = O. In other words, the
coset x + I = Ix + I is the root off(x) sought in F'.
Since each polynomial of positive degree has an irreducible factor
(Theorem 7-13), we may drop the restriction thatf(x) be irreducible.
Corollary. If the polynomialf(x) E F[x] is of positive degree, then there
exists an extension field of F containing a root off(x).
To go back to Theorem 7-27 for a moment, let us take a closer look at
the nature of the cosets of I = (j(x)) in F[ x], with the aim of expressing
the extension field F' = F[x]/I in a more convenient way. As usual, these
cosets are of the form g(x) + I, with g(x) E F[xJ. Invoking the division
algorithm, for each such g(x) there is a unique polynomial r(x) in F[ x]
satisfying g(x) = q(x)f(x) + r(x), where r(x) = 0 or deg r(x) < degf(x).
Now g(x) - r(x) = q(x)f(x) E I, so that g(x) and r(x) determine the same
coset; g(x) + I = r(x) + I. From this, it is possible to draw the following
conclusion: each coset of I in F[ x] contains exactly one polynomial which
POLYNOMIAL RINGS 145

has degree less than that off(x) or else is the zero polynomial. In fact, the
cosets of I are uniquely determined by remainders on division by f(x) in
the sense that g(x) + I = h(x) + I if and only if g(x) and h(x) leave the
same remainder when divided by f(x).
Thus, if degf(x) = n > 1 (for instance,f(x) = ao + a1x + ... + anx n),
then the extension field F' may be described by
F' = {b o + blx + ... + bn_lxn- 1 + IlbkEF}.
Identifying bk + I with the element bk , we see as before that a typical coset
can be uniquely represented in the form
bo + bl(x + + ... + bn-1(x + I)n-l.
I)
As a final simplification, let us replace x + I by some new symbol A., so that
the elements of F' become polynomials in A:
F' = {b o + blA + ... + bn_lAn-llbkEF}.
Observe that since A = x + I is a root off(x) in F', calculations are carried
out with the aid of the relation ao + alA + ... + anAn = o.
The last paragraph serves to bring out the point that F' is a finite7\
extension of F with basis {l, A, A2 , ••• , An - l }; in particular, we infer that
[F':F] = n = degf(x).
To recapitulate: if f(x) E F[ x] is an irreducible polynomial over F, then
there exists a finite extension F' of F, such that [F' :F] = degf(x), in which
f(x) has a root. Moreover, F' is a simple algebraic extension generated by
a root of f(x). (Admittedly, some work could be saved by an appeal to
Theorems 7-21 and 7-25, but our object here is to present an alternative
approach to the subject.)
We pause now to examine two concrete examples of the ideas just
presented.
Example 7-12. Consider Z2' the field of integers modulo 2, and the poly-
nomial f(x) = x 3 + X + 1 E Z2[X]. Since neither of the elements 0 and 1
is a root of x 3 + x + 1,f(x) must be irreducible in Z2[X].
Theorem 7-27 thus guarantees the existence of an extension of Z2'
specifically, the field Z2[ x ]/(j(x)), in which the given polynomial has a root.
Denoting this root by A, the discussion above tells us that
Z2[X]/(j(x)) = {a + bA + cA2la, b, CE Z2}
= {O, 1, A, 1 + A, A2, 1 + A2, A + A2, 1 + A + A2},
where, of course, A3 + A + 1 = o.
As an example of operating in this field, let us calculate the inverse of
1 + A + A2 • Before starting, observe that by using the relations
A3 = -(A + 1) = A + 1, A4 = A2 + A
146

(our coefficients come from Z2> where -1 = 1), the degree of any product
can be kept less than 3. Now, the problem is to determine elements a, b,
e E Z2 for which

Carrying out the multiplication and substituting for ,13, ,14 in terms of 1,
A, and ,12, we obtain
(a + b + e) + aA + (a + b)A2 = l.
This yields the system of linear equations
a + b + e = 1, a = 0, a + b = 0,
with solution a = b = 0, e = 1; therefore, (1 + A + ,12)-1 = ,12.
It is worth noting that x 3 + x + 1 factors completely into linear
factors in Z2[X]/(j(x)) and has the three roots A, ,12, and A + ,12:
x3 + x + 1 = (x - A)(x - A2)(x - (A + ,P)).
Example 7-13. The quadratic polynomial x 2 + 1 is irreducible in R#[x].
For, if x 2 + 1 were reducible, it would be of the form
x2 + 1 = (ax + b)(ex + d)
= aex2 + (ad + be)x + bd,
where a, b, e, d E R #. It follows at once that ae = bd = 1 and ad + be = 0.
Therefore, be = -(ad), and
1 = (ae)(bd) = (ad)(be) = -(adf,
or rather, (ad)2 = -1, which is impossible.
In this instance, the extension field R#[x]/(x2 + 1) is described by

R#[x]/(x 2 + 1) = {a + bAla, b E R#; ,12 + 1 = O}.

Performing the usual operations for polynomials, we see that

(a + bA) + (e + dA) = (a + c) + (b + d)A
and
(a + bA)(e + dA) = (ae - bd) + (ad + be)A + bd(A2 + 1)
= (ae - bd) + (ad + bd)A.
The similarity of these formulas to the usual rules for addition and multi-
plication of complex numbers should be apparent. As a matter of fact,
R#[x]/(x 2 + 1) is isomorphic to the field C of complex numbers under the
mapping $: R#[x]/(x2 + 1) --* C given by $(a + bA) = a + bi. This
provides an elegant way of constructing C from R #.
POLYNOMIAL RINGS 147

Before proceeding further, two comments are in order. First, Example

7-12 shows that there exist finite fields other than the fields Zp of integers
modulo a prime p. The fact that the field of this example has 2 3 = 8 ele-
ments is typical of the general situation: if F is a finite field, then F contains
pH elements, where the prime p is the characteristic of F (Theorem 9-7).
In the second place, the construction of Theorem 7-27 yields an exten-
sion of the field F in which a given (nonconstant) polynomial f(x) E F[x]
splits off one linear factor. By repeated application of this procedure, we
can build up an extension F' of Fin whichf(x), thought of as a member of
F'[ x], factors into a product of linear factors; that is, the field F' is large
enough to contain all the roots off(x) (technically speaking, the polynomial
splits completely in F'[ x]). We present this result in the form of an existence
theorem.
Theorem 7-28. If f(x) E F[x] is a polynomial of positive degree, then
there exists an extension field F' of F in which f(x) factors completely
into linear polynomials.
Proof The proof is by induction on n = degf(x). If n = 1, then f(x) is
already linear and F itself is the required extension. Therefore, assume
that n > 1 and that the theorem is true for all fields and for all polynomials
of degree less than n. Now, the polynomial f(x) must have some irreducible
factor g(x). By Theorem 7-27, there is an extension field K of F in which
g(x) and, hence, f(x), has a root r 1; specifically, the field K = F[ x ]/(g(x)).
Thus,J(x) can be written in K[ x] as f(x) = (x - r 1)h(x), where deg h(x) =
n - 1. By our induction assumption, there is an extension F' of K in which
h(x) splits completely; say h(x) = a(x - r2 )(x - r3) ... (x - rH)' with r i E F',
a +- O. From this, we see that f(x) can be factored into linear factors in
F'[xJ.
Corollary. Let f(x) E F[x], degf(x) = n > o. Then there exists an
extension of Fin whichf(x) has n (not necessarily distinct) roots.
Example 7-14. To illustrate this situation, let us look at the polynomial
f(x) = (x 2 - 2)(x 2 - 3) over the field Q of rational numbers. From
Example 7-4, x 2 - 2 (and by similar reasoning, x 2 - 3) is already known
to be irreducible in Q[x J. So we begin by extending Q to the field F l' where
Fl = Q[x]/(x 2 - 2) = {a + bAla, b E Q; ,1.2 - 2 = O};
and obtain the factorization
f(x) = (x - A)(x + A)(x2 - 3)
= (x - J2)(x + J2)(x 2 - 3).

(As ,1.2 = 2, one customarily identifies A with J2.)

148

However, f(x) does not split completely, since the polynomial x 2 - 3

remains irreducible in F l[x]. For, suppose to the contrary that x 2 - 3 has
a root in F 1; say c + dJ2, with c, dE Q. Substituting, we find that
(c 2 + 2d2 - 3) + 2cdJ2 = 0,
or, what amounts to the same thing,
cd = o.
The latter equation implies that either c = 0 or d = O. But neither c nor d
can be zero, since this would mean that d2 = 3/2 or c2 = 3, which is clearly
impossible. Accordingly, x 2 - 3 does not split in Fl[X].
In order to factor f(x) into linear factors, it becomes necessary to extend
the coefficient field further. We therefore construct a second extension F 2 ,
where
F2 = Fl[X]/(X 2 - 3) = {ex + PJ1.lex,PEF 1 ;J1. 2 - 3 = O}.
The elements of F 2 can be expressed alternatively in the form
(a + bJ2) + (c + dJ2)J3 = a + bJ2 + cJ3 + dj6,
where, of course, the coefficients a, b, c, d all lie in Q. It follows without
difficulty that the original polynomial now factors in F 2[x] as
f(x) = (x - A)(X + A)(X - J1.)(x + J1.)
= (x - J2)(x + J2)(x - J3)(x + J3).
Let a field F be given and consider a nonconstant polynomialf(x) E F[ x].
An extension field F' of F is said to be a splitting field for f(x) over F provided
that f(x) can be factored completely into linear factors in F'[ x], but not so
factored over any proper subfield of F' containing F (this minimum nature
of the splitting field is not required by all authors). Loosely speaking, a
splitting field is the smallest extension field F' in which the prescribed
polynomial factors linearly:
f(x) = a(x - rl)(x - r 2) ... (x - rn) (ri E F').
To obtain a splitting field for f(x), we need only consider the family {Fi}
of all extension fields Fi in whichf(x) can be decomposed as a product of
linear factors (Theorem 7-27 guarantees the existence of such extensions);
then n Fi serves as a splitting field for f(x) over F.
Having thus indicated the existence of a splitting field for an arbitrary
polynomial in F[ x], it is natural to follow this up with a query as to unique-
ness. For a final topic, we shall prove that any two splitting fields of the
same (nonconstant) polynomial are isomorphic; this being so, one is justified
in using the definite article and speaking of the splitting field of a given
polynomial.
POLYNOMIAL RINGS 149

Before presenting the main theorem, two preparatory results of a some-

what technical nature are needed.
Lemma. Letf(x) be an irreducible polynomial in F[x] and r be a root
of f(x) in some extension field K of F. Then F(r) ~ F[x ]/(j(x») under
an isomorphism whereby the element r corresponds to the coset
x + (j(x»).
Proof. Since the element r is algebraic over F, it follows directly from
Theorem 7-19 that F(r) ~ F[x]/(j(x») via an isomorphism () with the
property that 4Jr = () nat(J(x». (As usual, 4Jr: F[x] --+ K is the substitu-
0

tion homomorphism induced by r.) Regarding the last statement of the

lemma, we necessarily have
r = 4Jr(x) = «() 0 nat(J(x»)(x) = ()(x + (f(x»).
The chief value of this' lemma is that it leads almost immediately to the
following theorem.
Theorem 7-29. (Isomorphism Extension Theorem). Let (J be an iso-
morphism from the field F onto the field F'. Also,letf(x) = a o + a1x +
... + an xn be an irreducible polynomial in F[x] and f'(y) = u(a o) +
u(a1)y + ... + u(an)yn be the corresponding polynomial in F'[y J.
Then, f'(y) is likewise irreducible. Furthermore, if r is a root of f(x)
in some extension field of F and r' is a root of f'(y) in some extension
field of F', then u can be extended to an isomorphism (l> of F(r) onto
F'(r') with (l>(r) = r'.
Proof. Let us first extend u to a mapping ii between the polynomial rings
F[x] and F'[y] by setting
iig(x) = ii(b o + b1x + ... + bnxn) = u(b o) + u(b1)y + ... + u(bn)yn
for any polynomial g(x) = bo + b1x + ... + bnxn E F[xJ. We bequeath
to the reader the task of supplying the necessary details that ii is an iso-
morphism of F[ x] onto F'[y]. It is important to notice that for any poly-
nomial g(x) in F[ x], an element a E F is a root of g(x) if and only if u(a) is a
root of iig(x). Indeed, if, as before, g(x) = bo + b1x + ... + bnxn, then,
upon evaluating iig(x) at u(a),
(iig(x»)(u(a») = u(b o) + u(b1)u(a) + ... + u(bn)u(a)n
= u(b o + b1a + ... + bnan)
= u(g(a»),
from which our assertion follows. In particular, we infer that the poly-
nomials g(x) and ijg(x) are simultaneously reducible or irreducible in F[x]
and F'[y], respectively. This being so, f'(y) = iif(x) is irreducible in F'[y J.
150

By the foregoing lemma, we know that there exist isomorphisms

a: F(r) ~ F[x]/(j(x)) and p: F'(r') ~ F,[y]/(j'(y)), with
a(r) = x + (j(x)), p(r') = y + (j'(y)).
Moreover, it is an easy matter to show that there is also an isomorphism r of
F[ x ]/(j(x)) onto F'[y ]/(f'(y)) defined by
r(g(x) + (f(x))) = ifg(x) + (j'(y)) (g(x) E F[x ]).
Observe particularly that r carries the coset x + (j(x)) onto y + (j'(y)).
We contend that F(r) ~ F'(r') under the composition of maps
<I> = p-l 0 r 0 a,
where <1>: F(r) ~ F'(r'); this situation is portrayed in the diagram below:
<I>
F(r) ---.-+) F'(r')

aj P-' j
F[ x ]/(j(x)) ~ F,[y]/(j'(y))
Certainly, <I> is an isomorphism of F(r) onto F'(r'), for the individual mappings
a, r, p-l are themselves isomorphisms. If a is an arbitrary element of F, then
<I>(a) = (P-l r) (a(a)) = (P-l r)(a + (f(x)))
0 0

= p-l(u(a) + (f'(y))) = u(a),

whence <I> is actually an extension of u to all of F(r). Finally, we point out
that
<I>(r) = (P-l 0 r)(a(r)) = (P-l 0 r)(x + (f(x))
= P-l(y + (f'(y))) = r',
as required, and the theorem is proved in its entirety.
For a simple, but nonetheless satisfying, illustration of this last result,
take both F and F' to be the real number field R#; letf(x) E R#[x] be the
irreducible polynomial f(x) = x 2 + 1, so that f'(y) = y2 + 1 (recall that
the identity map is the only isomorphism of R # onto itself). Finally, choose
r = i and r' = -i. Theorem 7-29 then asserts that R#(i) ~ R#( -i) under
anisomorphismwhichcarriesionto -i. InasmuchasR#(i) = R#(-i) = C,
the isomorphism in question is just the correspondence between a complex
number and its conjugate.
We now have the mathematical machinery to show the uniqueness (to
within isomorphism) of splitting fields. Actually, we shall prove a some-
what more general result.
Theorem 7-30. Let u be an isomorphism of the field F onto the field
F'. Let f(x) = ao + a 1 x + ... + anx n E F[x] and f'(y) = u(a o) +
PROBLEMS 151

0"(a 1 )y + ... + O"(an)yn be the corresponding polynomial in F'[y]. If

K is a splitting field of J(x) and K' a splitting field of J'(y), then 0" can
be extended to an isomorphism <I> of K onto K'.
Proof Our argument will be by induction on the number n of roots ofJ(x)
that lie outside F, but (needless to say) in K. When n = 0, all the roots of
J(x) belong to F and F is itself the splitting field of J(x); that is, K = F.
This in turn induces a splitting of the polynomial J'(y) into a product of
linear factors in F'[y], so that K' = F'. Thus, when it happens that n = 0,
the isomorphism 0" is, in a trivial sense, the desired extension to the splitting
fields.
Let us next assume, inductively, that the theorem holds true for any
pair of corresponding polynomials J(x) and J'(y) over isomorphic fields E
and E', provided that the number of roots of roots of J(x) outside of E is
less than n (n ;:=: 1).
IfJ(x) E F[x] is a polynomial having n roots outside of F, then not all
of the irreducible factors of J(x) can be linear in F[ x]; for, otherwise,f(x)
would split completely in F, contrary to assumption. Accordingly, J(x)
must have some factor g(x) of degree m > 1 which is irreducible in F[xJ.
Let g'(y) denote the corresponding irreducible factor of J'(y). Since K is
a splitting field of J(x) over F, g(x) in particular must have a root in K;
call it r. Similarly, one of the roots of the polynomialJ'(y), say r', is a root
of g'(y) in K'. By Theorem 7-29,0" can be extended to an isomorphism (1'
between the fields F(r) and F'(r'). Now, K is a splitting field ofJ(x), viewed
as a polynomial with coefficients from F(r); in a like manner, K' can be
regarded as a splitting field ofJ'(y) over the field F'(r'). Because the number
of roots ofJ(x) lying outside of F(r) is less than n, the induction hypothesis
permits us to extend 0"' (itself an extension of 0") to an isomorphism <I> of
K' onto K. This completes the induction step and the proof of the theorem
as well, for 0" has been suitably extended.
With the corollary below, we achieve our objective.
Corollary. Any two splitting fields of a nonconstant polynomial
J(x) E F[x] are isomorphic via an isomorphism <I> such that the restric-
tion <l>JF is the identity mapping.
Proof This is an immediate consequence of the theorem on taking F = F'
and 0" to be the identity isomorphism iF.

PROBLEMS

1. If R is a commutative ring with identity, prove that

a) The set 1= {f(x) E R[[x]]lordf(x) > O} U {O} forms an ideal of the ring
R[[x]]; in fact, I = (x).
152

b) The ideal I" consists of all power series having order 2.n, together with O.
c) nnez+1" = {O}.
2. For any field F, consider the set F<x) consisting of all expressions of the form

where all the ak E F and n 2. 0 varies.

If addition and multiplication are defined in the obvious way, F<x) becomes a
ring, known as the ring of extended (formal) power series over F. Show that F<x)
is in fact the field of quotients of the domain F[[x]J. [Hint: Given nEZ+,
Qcl(F[[x]]) must contain x-n.]

3. Let R be a commutative ring with identity. If R is a local ring, prove that the
power series ring R[[x]] is also local.

4. Given that R is a commutative ring with identity, deduce that

a) No monic polynomial is R[x] is a zero divisor.
b) If the polynomial f(x) = ao + a1x + ... + an~ is a zero divisor in R[x],
then there exists an element 0 =1= r E R such that rf(x) = O. [Hint: Assume that
f(x)g(x) = O. Use the polynomials akg(x) to obtain 0 =1= h(x) E R[x], with
deg h(x) < deg f(x), satisfying h(x)f(x) = 0.]
5. If R is a commutative ring with identity, verify that the polynomial 1 + ax is
invertible in R[x] if and only ifthe element a is nilpotent in R. [Hint: Problem 10,
Chapter 1.] "

6. For an arbitrary ring R, prove that

a) If I is an ideal of R, then I[ x] forms an ideal of the polynomial ring R[x].
b) If Rand R' are isomorphic rings, then R[x] is isomorphic to RT x ].
c) char R = char R[x] = char R[[x]J.
d) If I is a nil ideal of R, then I[x] is a nil ideal of R[x]. [Hint: Induct on the
degree of polynomials in I[ x].]

7. Establish the following assertions concerning the polynomial ring Z[x]:

a) The ideal
(x) = {a1x + a2 x 2 + ... + anxnlakEZ;n 2.1}
is a prime ideal of Z[x], but not a maximal ideal. Incidentally, (x) is maximal
in F[ x], where F is a field.
b) Z[x] is not a principal ideal domain. [Hint: Consider (x, 2), the (maximal)
ideal of polynomials with eve!l constant terms.]
c) The primary ideal (x, 4) is not the power of any prime ideal of Z[x]. [Hint:
(x,2) is the only prime ideal containing (x, 4).]

8. Let P be a prime ideal of R, a commutative ring with identity. Prove that P[x]
is a prime ideal of the polynomial ring R[x]. If M is a maximal ideal of R, is
M[x] a maximal ideal of R[x]?
PROBLEMS 153

9. Consider the polynomial domain F[ x], where F is a field, and a fixed element
rEF. Show that the set of all polynomials having r as a root,
Mr = {f(x) E F[x]lf(r) = O},

forms a maximal ideal of F[x], with F[x]/Mr ~ F. [Hint: Mr = kerQJ" where

QJr: F[ x] ...... F is the substitution homomorphism induced by r.]
10. Regarding the ring of Example 8, Chapter 1, show that the polynomial (a, O)x 2 E R[ x]
has infinitely many roots in R[ x].
11. Given f(x) = ao + a 1x + ... + a.x· E C[x], define the polynomial J(x) by

J(x) = ao + a1x + ... + ii.x·,

where ak denotes the usual complex conjugate of ak • Verify that
a) r E C is a root of f(x) if and only if l' is a root of J(x). [Hint: f(r) = 1(1').]
b) Iff(x) E R#[x] ~ C[x] and r is a complex root off(x), then r is also a root
off(x).
12. Let R be a commutative ring with identity and let f(x) E R[x]. The function
1: R ...... R defined by taking l(r) = f(r) for every r E R is called the polynomial
junction induced by f(x). Assuming that PH denotes the set of all polynomial
functions induced by elements of R[x], prove that
a) PH forms a subring of map R, known as the ring of polynomial functions on R;
b) the mapping ex: R[x] ...... PH given by ex(j(x») = 1 is a homomorphism of R[x]
onto PH;
c) if the element r E R is fixed and Ir = {l E PHIl(r) = O}, then Ir is an ideal of
PH'
13. a) When R is an integral domain, show that distinct polynomials in R[x] induce
distinct polynomial functions (in other words, the mapping ex: R[x] ...... PH is
one-to-one) if and only if R has an infinite number of elements.
b) Give an example of two distinct polynomials which induce the same polynomial
function.
14. Let R be a commutative ring with identity and· define the function 0: R[x] ...... R[x],
the so-called derivative function, as follows:
If f(x) = ao + a1x + ... + a.x· E R[x],

then
For any f(x), g(x) E R[x] and any r E R, establish that
a) o(j(x) + g(x») = of(x) + og(x).
b) o(rf(x» = rof(x).
c) o(j(x)g(x») = of(x)'g(x) + f(x)·og(x). [Hint: Induct on the number of terms
off(x).]
15. Suppose that R is a commutative ring with identity and let r E R be a root of the
nonzero polynomial f(x) E R[ x]. We call r a multiple root of f(x) provided that
f(x) = (x - r)·g(x) (n > 1),
154

where g(x) E R[x] is a polynomial such that g(r) 0/= O. Prove that an element
r E R is a multiple root of f(x) if and only if r is a root of both f(x) and t5f(x).
16. Let F be a field and f(x) E F[x] be a polynomial of degree 2 or 3. Deduce that
f(x) is irreducible in F[x] if and only ifit has no root in F. Give an example which
shows that this result need not hold if degf(x) ~ 4.
17. Prove that if a polynomial f(x) E F[x] (F a field of characteristic 0) is irreducible,
then all of its roots in any field containing F must be distinct. [Hint: First show
that gcd (J(x), t5f(x)) = 1.]
18. Given that I is a proper ideal of R, a commutative ring with identity, establish the
assertions below:
a) If v: R[x] ---> (R/I)[x] is the reduction homomorphism modulo the ideal I,
then ker v = I[x]; hence, R[x]/I[x] ~ (R/I)[x].
b) If the polynomial f(x) E R[x] is such that v(J(x)) is irreducible in (R/I)[x],
then f(x) is irreducible in R[ x].
c) The polynomial f(x) = x 3 - x 2 + 1 is irreducible in Z[x]. [Hint: Reduce
the coefficients modulo 2.]
19. Let R be a unique factorization domain. Show that any nonconstant divisor of a
primitive polynomial in R[x] is again primitive.
20. Utilize Gauss's Lemma to give an alternative proof of the fact that if R is a unique
factorization domain, then so is the polynomial ring R[x]. [Hint: For
00/= f(x) E R[x], induct on degf(x); if degf(x) > 0, write f(x) = Cfl(X), where
c E Rand fl(X) is primitive; if fl(X) is reducible, apply induction to its factors.]
21. Apply the Eisenstein Criterion to establish that the following polynomials are
irreducibleinQ[x] :f(x) = x 2 + l,g(x) = x 2 - X + l,andh(x) = 2x 5 - 6x 3 +
9x2 - 15. [Hint: Considerf(x + 1), g(-x).]
22. Let R be a unique factorization domain and K its field of quotients. Assume that
ab- 1 E K (where a and b are relatively prime) is a nonzero root of the polynomial
f(x) = ao + a1x + ... + anx" E R[x]. Verify that ala o and blan.
23. Prove the following assertions concerning the polynomial ring Z[x, y]:
a) The ideals (x), (x, y) and (2, x, y) are all prime in Z[x, y], but only the last is
maximal.
b) (x, y) = .J(x2, y) = .J(x2, xy, y2).
c) The ideal (x\ xy, y2) is primary in Z[x, y] for any integer k E Z +.
d) If I = (x 2, xy), then .Jlis a prime ideal, but I is not primary. [Hint:.JI = (x).]
24. Consider the polynomial domain F[ x, y], where F forms a field.
a) Show that (x 2, xy, y2) is not a principal ideal of F[ x, y].
b) Establish the isomorphism F[x, y]/(x + y) ~ F[x].
25. Let the element r be algebraic over the field F and let f(x) E F[ x] be a monic
polynomial such thatf(r) = O. Prove thatf(x) is the minimum polynomial of r
over F if and only if f(x) is irreducible in F[ x].
26. Assuming that F' is a finite extension of the field F, verify each of the statements
below:
PROBLEMS 155

a) When [F: F] is prime, F is a simple extension of F; in fact, F' = F(r) for every
element rEF' - F.
b) If f(x) E F[ x] is an irreducible polynomial whose degree is relatively prime to
[F':F], thenf(x) has no roots in F.
c) If rEF is algebraic of degree n, then each element of F(r) has as its degree an
integer dividing n.
d) Given fields K; (i = 1,2) such that F:2 K;:2 F, with [KI:F] and [K2:F]
relatively prime integers, necessarily KI ( l K2 = F.
27. Show that the following extension fields of Q are simple extensions and determine
their respective degrees: Q(.J3, .j7), Q(.J3, i), Q(.J2, ~5).
28. a) Prove that the extension field F = F(rl' r2, ... , r.), where each element r; is
algebraic over F, forms a finite extension of F. [Hint: If F; = F;_l(r;), then
F. = F' and [F':F] = TI;[F;+1: FJ]
b) If F" is an algebraic extension of F' and F' is an algebraic extension of F, show
that F" is an algebraic extension of F. [Hint: Each rEF" is a root of some
polynomial f(x) = ao + a1x + ... + a.x" E FT x]; consider the extension
fields K = F(a o, aI' ... , a.) and K' = K(r); r is algebraic over K'.]
29. Let F be an extension of the field F. Prove that the set of all elements in F' which
are algebraic over F constitute a subfield of F' ; applied to the case where F' = C
and F = Q, this yields the field of algebraic numbers. [Hint: If r, s are algebraic
over F, [F(r, s):F] is finite; hence, F(r, s) is an algebraic extension of F.]
30. a) Granting that f(x) = x 2 + X + 2 is an irreducible polynomial in Z3[X],
construct the multiplication table for the field Z3[X]/(j(x»).
b) Show that the polynomial f(x) = x 3 + x 2 + 1 E Z2[X] factors into linear
factors in Z2[ x ]/(j(x») by actually finding the factorization.
31. If n+- 1 is a (nonzero) square-free integer, verify that Q[x]/(x2 - n) forms a field
isomorphic to the quadratic field

Q(.Jn) = {~ + b.JnJa, b E Q}.

32. Describe the splitting fields of the following polynomials:
a) x 3 - 3 E Q[ x],
b) x 2 + X + 1 E Zs[x],
c) X4 + 2X2 + 1 E R#[x],
d) (x 2 - 2)(x2 + 1) E Q[xJ.

33. Let r be a root of the polynomial f(x) = x3 - X + 1 E Q[ x J. Find the inverse

of 1 - 2r + 3r2 in Q(r).
34. Letf(x) E F[x] be an irreducible polynomial and r, s be two roots off(x) in some
splitting field. Show that F[r] ~ F[s], by a unique isomorphism that leaves
every element of F fixed and takes r into s.

35. Suppose that F' is the splitting field for the polynomialf(x) E F[x]; say

f(x) = a(x - rl)(x - r2) ... (x - r.) (r; E F', a +- 0).

156

Prove that F' = F(rl' r 2 , ••• , r.). As a particular illustration, establish that
Q(J2, .J3)is the splitting field of (x 2 - 2)(x 2 - 3) E Q[x].
36. Letf(x) E Zp[x] be an irreducible polynomial of degree n, p a prime. Verify that
the field F = Zp[x]/(J(x») contains p. elements.
37. If F' is a splitting field of a polynomial of degree n over F, show that [F':F] ::::; n!
38. A field F' is said to be algebraically closed if F' has no proper algebraic extensions.
Assuming that F' is an algebraic extension of F, prove the equivalence of the
following statements:
a) F' is algebraically closed.
b) Every irreducible polynomial in F'[x] is linear.
c) Every polynomial in F[x] splits in F'.
(For a proof that every field has an algebraic extension which is algebraically
closed, the reader is referred to [23].)
EIGHT

CERTAIN RADICALS OF A RING

We touched earlier on the radical concept by briefly considering the notion

of the nil radical of an ideal. There are a number of other radicals in
circulation; several of the more prominent ones are introduced in this
chapter. These various formulations are not, in general, equivalent to one
another and this has given rise to certain confusion and ambiguity in the
use of the term. (Indeed, whenever the reader encounters the word "radical"
by itself, he should take some pains to discover just what is meant by it.)
By way of removing some of this confusion, qualifying adjectives are given
to the different types of radicals which appear here. In addition to indicating
the importance of these new radicals in the structure theory, we will be
concerned with the nature of the inclusion relations between them and the
circumstances under which various radicals coincide. The reader is again
reminded that, in the absence of any statement to the contrary, the term
"ring" will always mean a commutative ring with identity.
It appears in order to define one of the radicals around which our
interest centers.

Definition 8-1. The Jacobson radical of a ring R, denoted by rad R, is

the set
rad R = n {MJM is a maximal ideal of R}.

If rad R = {O}, then R is said to be a ring without Jacobson radical or,

more briefly, R is a semisimple ring.

The Jacobson radical always exists, since we know by Theorem 5-2 that
any commutative ring with identity contains at least one maximal ideal.
It is also immediately obvious from the definition that rad R forms an
ideal of R which is contained in each maximal ideal.
To fix ideas, let us determine the Jacobson radical in several concrete
rings.

Example 8-1. The ring Z of integers is a semisimple ring. For, according

157
158

to Example 5-1, the maximal ideals of Z are precisely the principal ideals
generated by the prime numbers; thus,
rad R = n {(p)lp is a prime number}.

Since no nonzero integer is divisible by every prime, we see at once that

rad R = {O}.
Example 8-2. A more penetrating illustration is furnished by the ring
R = map (X, F), where X is an arbitrary set and F a field. For any element
x E X, consider the function 7: x f = f(x) which assigns to each function in
R its value at x. It is easily checked that 7: x is a homomorphism of R into F;
since R contains all the constant functions, this homomorphism actually
maps onto the field F. Thus, by Problem 10, Chapter 5, its kernel is the
maximal ideal
Mx = {fE Rlf(x) = O}.

Because rad R £ n Mx = {f E Rlf(x) = 0 for all x E X} = {O}, it follows

that R must be a semisimple ring.
Example 8-3. For a final example, we turn to the ring R[[x]] of formal
power series. Here, there is a one-to-one correspondence between the
maximal ideals M of R and maximal ideals M' of R[[ x]] in such a way
that M' corresponds to M if and only if M' is generated by M and x
(Theorem 7-4). Thus,

rad R[[x]] n {M'IM' is a maximal ideal of R[[x]]}

n {(M, x)IM is a maximal ideal of R}
= (n M, x) = (rad R, x).
In particular, if R is taken to be a field F, we have rad F[[ x]] = (x), the
principal ideal generated by x.
Our first theorem establishes a basic connection between the Jacobson
radical and invertibility of ring elements.

Theorem 8-1. Let I be an ideal of the ring R. Then I £ rad R if and

only if each element of the coset 1 + I has an inverse in R.

Proof We begin by assuming that I £ rad R and that there is some

element a E I for which 1 + a is not invertible. Our object, of course, is
to derive a contradiction. By the corollary to Theorem 5-3, the element
1 + a must belong to some maximal ideal M of the ring R. Since a E rad R,
a is also contained in M, and therefore 1 = (1 + a) - a lies in M. But this
means that M = R, which is clearly impossible.
CER TAIN RADICALS OF A RING 159

To prove the converse, suppose that each member of 1 + I has a

multiplicative inverse in R, but I $ rad R. By definition of the Jacobson
radical, there will exist a maximal ideal M of R with I $ M. Now, if a is
any element of I which is not in M, the maximality of M implies that the
ideal (M, a) = R. Knowing this, the identity element 1 can be expressed
in the form 1 = m + ra for suitable choice of m E M and r E R. But then,
m = 1 - ra E 1 + I, so that m possesses an inverse. The conclusion is
untenable, since no proper ideal contains an invertible element.
The form which this result takes when I is the principal ideal generated
by a E rad R furnishes a characterization of the Jacobson radical in terms
of elements rather than ideals. Although actually a corollary to the theorem
just proved, it is important enough to be singled out as a theorem.
Theorem 8-2. In any ring R, an element a E rad R if and only if 1 - ra
is invertible for each r E R.
This theorem adapts itself to many uses. Three fairly short and
instructive applications are presented below.
Corollary 1. An element a is invertible in the ring R if and only if the
coset a + rad R is invertible in the quotient ring R/rad R.
Proof Assume that the coset a + rad R has an inverse in R/rad R, so that
(a + rad R)(b + rad R) = 1 + rad R
for some b E R. Then 1 - ab lies in rad R. We now appeal to Theorem
8-2, with r = 1, to conclude that the product ab = 1 - 1(1 - ab) is
invertible; this, in turn, forces the element a to have an inverse in R. The
other direction of the corollary is all but obvious.
Corollary 2. The only idempotent element in rad R is o.
Proof Let the element a E rad R with a2 = a. Taking r = 1 in the pre-
ceding theorem, we see that 1 - a has an inverse in R; say (1 - a)b = 1,
where b E R. This leads immediately to
a = a(1 - a)b = (a - a2 )b = 0,
which completes the proof.
Corollary 3. Every nil ideal of R is contained in rad R.
Proof Let N be a nil ideal of R and suppose that a E N. For every r E R,
we then have ra E N, so that the product ra is nilpotent. Problem 10,
Chapter 1, therefore implies that 1 - ra is invertible in R. This shows that
the element a lies in rad R, from which it follows that N s; rad R.
Although the Jacobson radical of a ring R is not necessarily a nil ideal,
very little restriction on R forces it to be nil. We shall see subsequently
160

that, if every ideal of R is finitely generated, then rad R is not only nil but
nilpotent.
This is a convenient place to also point out that a homomorphic image
of a semisimple ring need not be semisimple. An explicit example of this
situation can easily be obtained from the ring Z of integers. While Z forms
a ring without a Jacobson radical, its homomorphic image Zp" (p a prime;
n > 1) contains the nil ideal (p); appealing to Corollary 3 above, we see
that Z pn cannot be semisimple.
Example 8-4. Consider F[[ x]], the ring of formal power series over a field
F. From the lemma on page 116, it is known that an element f(x) =
ao + a1x + ... + anxn + ... has an inverse in F[[x]] if and only if the
constant term ao =1= O. This observation (in conjunction with Theorem 8-2)
implies that if g(x) = b o + b1x + ... + bnxn + ... , then

rad F[[ x]] = {f(x)ll - f(x)g(x) is invertible for all g(x) E F[[ x]]}
= {f(x) 11 - aob o =1= 0 for all bo E F}
= {f(x)la o = O} = (x).
Thus, we have a second proof of the fact that the Jacobson radical of F[[x]]
is the principal ideal generated by x.
We next prove several results bearing on the Jacobson radical of quotient
rings. The first of these provides a convenient method for manufacturing
semisimple rings; its proof utilizes both implications of the last theorem.
Theorem 8-3. For any ring R, the quotient ring Rlrad R is semisimple;
that is, rad(R/rad R) = {O}.
Proof Before becoming involved in details, let us remark that since rad R
constitutes an ideal of R, we may certainly form the quotient ring Rlrad R.
To simplify notation somewhat, we will temporarily denote rad R by 1.
Suppose that the coset a + J E rad (RIJ). Our strategy is to show that
the element a E J, for then a + J = J, which would imply that rad (RIJ)
consists of only the zero element of RI1. Since a + J is a member of rad (RI J),
Theorem 8-2 asserts that
(1 + J) - (r + J)(a + J) = 1 - ra + J
is invertible in RIJ for each choice of r E R. Accordingly, there exists a
coset b + J (depending, of course, on both r and a) such that
(1 - ra + J)(b + J) = 1 + 1.
This is plainly equivalent to requiring
1 - (b - rab) E J = rad R.
CERTAIN RADICALS OF A RING 161

Again appealing to Theorem 8-2, we conclude that the element

b - rab = 1 - 1(1 - b + rab)
has an inverse c in R. But then
(1 - ra)(bc) = (b - rab)c = 1,
so that 1 - ra possesses a multiplicative inverse in R. As this argument
holds for every r E R, it follows that a E rad R = I, as desired.
Continuing this theme, let us expre!:s the Jacobson radical ofthe quotient
ring RjI as a function of the radical of R.
Theorem 8-4. If I is an ideal of the ring R, then

1) rad (Rjl) 2
rad R
I
+ I
' and,

2) whenever I ~ rad R, rad (Rjl) = (rad R)jI.

Proof Perhaps the quickest way to establish the first assertion is by means
of the Correspondence Theorem; using this, one has
rad (Rjl) = n {M'IM' is a maximal ideal of Rjl}
= n {natIMIM is a maximal ideal of R with I ~ M}
radR + I
2 nat I (rad R + I) = I '

which is the first part of our theorem (the crucial step requires the inclusion
nI!:;MM 2 I + radR).
With an eye to proving (2), notice that whenever I ~ rad R, then

rad (Rjl) 2
rad R
I
+ I
2 (rad R)jI.

Thus, we need only to show the inclusion (rad R)jl 2 rad (Rjl). To this
purpose, choose the coset a + IE rad (Rjl) and let M be an arbitrary
maximal ideal of R. Since I ~ rad R ~ M, the image natIM = Mjl must
be a maximal ideal of the quotient ring Rjl (Problem 3, Chapter 5). But
then,
a + IE rad (Rjl) ~ Mjl,

forcing the element a to lie in M. As this holds for every maximal ideal of
R, it follows that a E rad R and so a + IE (rad R)jI. All in all, we have
proved that rad (Rjl) ~ (rad R)jl, which, combined with our earlier
inclusion, leads to (2).
Armed with Theorem 8-4, we are in a position to establish:
162

Theorem 8-5. For any ring R, rad R is the smallest ideal I of R such
that the quotient ring R/I is semisimple (in other words, if R/I is a
semisimple ring, then rad R ~ I). .
Proof From Theorem 8-3, it is already known that R/rad R is without
Jacobson radical. Now, assume that I is any ideal of R for which the
associated quotient ring R/I is semisimple. Using part (1) of the preceding
theorem, we can then deduce the equality (I + rad R)/I = I. This in turn
leads to the inclusion rad R ~ I, which is what we sought to prove.
This may be a good place to mention two theorems concerning the
number of maximal ideals in a ring; these are of a rather special character,
but typify the results that can be obtained.
Theorem 8-6. Let R be a principal ideal domain. Then, R is semi-
simple if and only if R is either a field or has an infinite number of
maximal ideals.
Proof Let {Pi} be the set of prime elements of R. According to Theorem
6--10, the maximal ideals of R are simply the principal ideals (Pi). It follows
that an element a E rad R if and only if a is divisible by each prime Pi. If
R has an infinite set of maximal ideals, then a = 0, since every nonzero
noninvertible element of R is uniquely representable as a finite product of
primes. On the other hand, if R contains only a finite number of primes
Ph P2' ... , Pn' we have
rad R = n (Pi) = (P1P2 ... Pn) +- {O},
n

i= 1

so that R cannot be semisimple. Finally, observe that if the set {Pi} is empty,
then each nonzero element of R is invertible and R is a field (in which case
rad R = {O}).
Corollary. The ring Z of integers is semisimple.
Theorem 8-7. Let {Mi }, i E oF, be the set of maximal ideals of the ring
R. If, for each i, there exists an element a i E Mi such that 1 - a i E
rad R - M i , then {Mi} is a finite set.
Proof Suppose that the index set oF is infinite. Then there exists a well-
ordering ::5: of oF under which oF has no last el~ent. (See Appendix A for
terminology.) For each i E oF, we define Ii ~. . ( )i <j Mj . Then {I;} forms
a chain of proper ideals of R. By hypothesis, we can select an element
ai E Mi such that 1 - ai E Ii - Mi. Now the ideal I = u Ii is also a proper
ideal of R, since 1 ¢ I. By our choice of the Ii' I is not contained in any
maximal ideal of R. Indeed, suppose that there does exist an index i for
which I ~ M i ; then,
CERTAIN RADICALS OF A RING 163

yielding the contradiction 1 E Mi' But it is known that every proper ideal
of R is contained in a maximal ideal of R (Theorem 5-2). From this
contradiction we conclude that § must be finite.
Let us now turn to a consideration of another radical which plays an
essential role in ring theory, to wit, the prime radical. Its definition may
also be framed in terms of the intersection of certain ideals.
Definition 8-2. The prime radical of a ring R, denoted by Rad R (in
contrast with rad R), is the set
Rad R = n {pip is a prime ideal of R}.
If Rad R = {O}, we say that the ring R is without prime radical or has
zero prime radical.
Theorem 5-7, together with Definition 8-1, shows that the prime radical
exists, forms an ideal of R, and satisfies the inclusion Rad R ~ rad R. It
is useful to keep in mind that, for any integral domain, the zero ideal is a
prime ideal; for these rings, Rad R = {O}. In particular, the ring F[[ x]]
of formal power series over a field F has zero prime radical but, as we already
know, a nontrivial Jacobson radical.
Perhaps the most striking result of the present chapter is that th~ prime
radical, although seemingly quite different, is actually equal to the nil
radical of a ring. The lemma below provides the key to establishing this
assertion.
Lemma. Let I be an ideal of the ring R. Further, assume that the subset
S ~ R is closed under multiplication and disjoint from 1. Then there
exists an ideal P which is maximal in the set of ideals which contain I
and do not meet S; any such ideal is necessarily prime.
Proof. Consider the family ffi of all ideals J of R such that I ~ J and
J n S = cp. This family is not empty since I itself satisfies the indicated
conditions. Our immediate aim is to show that for any chain of ideals {Ji}
in ffi, their union u J i also belongs to ffi. It has already been established
in Theorem 5-2 that the union of a chain of ideals is again an ideal; more-
over, since I ~ J i for each i, we certainly have I ~ u J i • Finally, observe
that
(u J i ) n S = u (J i n S) = u cp = cp.
The crux of the matter is that Zorn's Lemma can now be applied to infer
that ffi has a maximal element P; this is the ideal that we want.
By definition, P is maximal in the set of ideals which contain I but do
not meet S. To settle the whole affair there remains simply to show that
P is a prime ideal. For this purpose, assume that the product ab E P but
that a ¢ P and b ¢ P. Since it is strictly larger than P, the ideal (P, a) must
164

contain some element r of S; similarly, we can find an element SE S such

that S E (P, b). This means that

rs E (P, a)(P, b) ~ (P, ab) ~ P.

As S is hypothesized to be closed under multiplication, the product rs also
lies in S. But this obviously contradicts the fact that P II S = 0. Our
argument therefore shows that either a or b is a member of P, which proves
that P is a prime ideal.
Remark. The ideal P need not be a maximal ideal of R, in the usual meaning
of the term, but only maximal with respect to exclusion of the set S. To put
it another way, if J is any ideal of the ring R which properly contains P,
then J must contain elements of S.
Two special cases of this general setting are particularly noteworthy:
S = {I} and I = {o}. In the event S = {I}, the ideal P mentioned in the
lemma is actually a maximal ideal (in the usual ideal-theoretic sense);
consequently, we have a somewhat different proof of the facts that (i) every
proper ideal is contained in a maximal ideal and (ii) each maximal ideal is
prime.
The case where I is the zero ideal is the subject of the following corollary,
a result which will be utilized on several occasions in the sequel.
Corollary. Let S be a subset of the ring R which is closed under multi-
plication and does not contain o. Then there exists an ideal maximal
in the set of ideals disjoint from S; any such ideal is prime.
As it stands, the preceding lemma is just the opening wedge; we can
exploit it rather effectively by now proving
Theorem 8-8. The intersection of all prime ideals of R which contain
a given ideal I is precisely the nil radical of I:

.JI = II {pip :2 I; P is a prime ideal}.

Proof If the element a ¢.ji, then the set S = {anln E Z+} does not intersect
I. Since S is closed under multiplication, the preceding lemma insures the
existence of some prime ideal P which contains I, but not a; that is, a does
not belong to the intersection of prime ideals containing I. This establishes
the inclusion
II {pip :2 I; P is a prime ideal} ~ .JT.
The reverse inclusion follows readily upon noting that if there exists a prime
ideal which contains I but not a, then a ¢ .JI,
since no power of a belongs
to P.
CERTAIN RADICALS OF A RING 165

As with the case of the Jacobson radical, the prime radical may be
characterized by its elements; this is brought out by a result promised
earlier.
Corollary. The prime radical of a ring R coincides with the nil radical
of R; that is, Rad R is simply the ideal of all nilpotent elements of R.
Proof The assertion is all but obvious upon taking I'= {O} in Theorem
8-8.
An immediate consequence of this last corollary is the potentially
powerful statement: every nil ideal of R is contained in the prime radical,
not simply contained in the larger Jacobson radical (Corollary 3 to Theorem
8-2).
Example 8-5. For an illustration of Theorem 8-8, let us fall back on the
ring Z of integers. In this setting, the nontrivial prime ideals are the principal
ideals (p), where P is a prime number. Given n > 1, the ideal (n) ~ (p) if
and only if P divides n; this being so,
.J(nj = n (pJ
pdn
Thus, if we assume that n has the prime power factorization
n = p~'p~2 ... p~r (k; E Z+),
it follows that
.J(nj = (Pi) n (P2) n ... n (Pr) = (PiP2 ... Pr)·
Let us go back to Theorem 8-8 for a moment. Another of its advantages
is that it permits a rather simple characterization of semiprime ideals.
(The reader is reminded that we defined an ideal I to be semiprime provided
that I = J1).
Theorem 8-9. An ideal I of the ring R is a semiprime ideal if and only
if I is an intersection of prime ideals of R.
Proof The proof is left to the reader; it should offer no difficulties.
Corollary. The prime radical Rad R is a semiprime ideal which is
contained in every semi prime ideal of R.
Before pressing on, we should also prove the prime radical counterpart
of Theorem 8-3.
Theorem 8-10. For any ring R, the quotient ring R/Rad R is without
prime radical.
Proof For clarity of exposition, set I = Rad R. Suppose that a + I is
any nilpotent element of R/I. Then, for some positive integer n,
(a + I)n = an + I = I,
166

so that an E I. But I consists of all nilpotent elements of R. Thus, we must

have (an)m = 0 for suitably chosen m E Z +; this is simply the statement
that a E I, and, hence, a + I is the zero element of R/I. Our argument
implies that the quotient ring R/I has no nonzero nilpotent elements, which
is to say that Rad (R/I) = {O}.
To round out the picture, two theorems are stated without proof; it will
be observed that these take the same form as the corresponding result
established for the Jacobson radical (Theorems 8-4 and 8-5).
Theorem 8-11. If I is an ideal of the ring R, then
1) Rad (R/I) 2 Rad ~ + I, and,

2) whenever I s; Rad R, Rad (R/I) = (Rad R)/I.

Theorem 8-12. For any ring R, Rad R is the smallest ideal I of R such
that the quotient ring R/I is without prime radical.
A problem exerting a natural appeal is that of describing the prime
radical of the polynomial ring R[x] in terms of the prime radical of R.
As a starting point, let us first prove a lemma which is of interest for various
parts of ring theory.
Lemma. A polynomialf(x) = ao + alx + ... + anxn is invertible in
R[ x] if and only if ao is invertible in R and all the other coefficients
aI' a 2, ... , an are nilpotent elements of R.
Proof If ao has an inverse in R and aI' a2, ... , an are all nilpotent, then the
polynomial f(x) = ao + alx + ... + anxn is the sum of an invertible
element and a nilpotent element. Hence, f(x) must itself be an invertible
element of R[x] (Problem 5, Chapter 7).
Going in the other direction, assume that the polynomial f(x) =
ao + alx + ... + anxn E R[x] possesses an inverse. That ao is then
invertible in R should be obvious. For any prime ideal P of R, P[x] is a
prime ideal of R[x] and the quotient ring R[x]/P[x] ~ (R/P)[xJ. Thus,
the homomorphic image off(x) in (R/P)[x],
(a o + P) + (a l + P)x + ... + (an + P)xn

must have an inverse. Since R/P is an integral domain, the invertible

elements in (R/P) [ x] are nonzero constant polynomials. This implies that
al +P= a2 + P = ... = an +P= P;
hence, the elements aI' a2, ... , an all lie in P. As this statement holds for
every prime ideal of R, it follows that aI' a2, ... , an E Rad R. By the corollary
to Theorem 8-8, the elements aI' a 2, ... , an must therefore be nilpotent.
CER TAIN RADICALS OF A RING 167

Having dealt with these preliminaries we are now ready to prove

Theorem 8-13. For any ring R, rad R[x] = Rad R[x].
Proof. It is enough to establish the inclusion rad R[ x] £; Rad R[ x J. If
the polynomial f(x) = a o + a1x + ... + anxn E rad R[x], Theorem 8-2
tells us that
1 + xf(x) = 1 + aox + ... + anxn+ 1
must be invertible in R[ x J. Hence, by the above lemma, the coefficients
ao, a 1, ... , an are all nilpotent elements of R. For a sufficiently large power,
f(x) will then be nilpotent in R[ x] and thus be in Rad R[x J.
The assertion of Theorem 8-13 can be improved upon. For the reader
will have little difficulty in now convincing himself that
Rad R[x] = (Rad R)[x],
where (Rad R)[x] denotes the ring of polynomials in x with coefficients
from Rad R. In fact, the inclusion Rad R[ x] £; (Rad R)[x] is implicit in
the foregoing proof; the opposite inclusion requires the corollary to Theorem
8-8.
By virtue of the displayed equation, we have
radF[x] = (RadF)[x] = {O}
for any field F. That is to say, the polynomial ring F[ x] constitutes a semi-
simple ring.
Suppose for the moment that I is an ideal of the ring R with I £; Rad R.
Given an idempotent element e =1= 0 in R, we know that the coset e + I
will be idempotent in RjI. What is not so obvious is that e + I =1= I; this
follows from the fact that Rad R contains no nonzero idempotents (Corollary
to Theorem 8-2). We are mainly concerned with the converse here: If
u + I is a nonzero idempotent of the quotient ring RjI, does there exist an
idempotent e E R for which e + I = u + I?
Before becoming involved in this discussion, let us give a general
definition.
Definition 8-3. Let I be an arbitrary ideal of the ring R. We say that
the idempotents of RjI can be raised or lifted into R in case every idem-
potent element of RjI is of the form e + I, where e is idempotent in R.
Definition 8-3 means just this: the idempotents of RjI can be lifted if
for each element U E R such that u 2 - U E I there exists some element
e 2 = e E R with e - U E 1. Although it is surely too much to expect the
lifting of idem po tents to take place for every I, we shall see that this situation
does occur whenever I is a nil ideal (or, equivalently, whenever I £; Rad R).
Let us begin with a lemma, important in itself.
168

Lemma. If e and e' are two idempotent elements of the ring R such that
e - e' E Rad R, then e = e'.
Proof Inasmuch as the product (e - e')(1 - (e + e'») = 0, it is enough
to show that 1 - (e + e') is an invertible element of R. Now, one may
write 1 - (e + e') in the form
1 - (e + e') = (1 - 2e) + (e - e'),
where(1 - 2e)2 = 1 - 4e + 4e 2 = 1. The implication is that 1
- (e + e'),
being the sum of a nilpotent element and an invertible element, is necessarily
invertible in R (Problem 10, Chapter 1).
Corollary. Let I, J be ideals of the ring R with I ~ J ~ Rad R. If the
idempotents of RjJ can be lifted into R, then so can the idempotents of
RjI.
Proof Suppose that u + I is any idempotent of RjI. Since I ~ J, it follows
that u + J is an idempotent element of RjJ; u 2 - U E I ~ J. By assump-
tion, there must exist some e2 = e in R such that e + J = u + J, whence
e - u E J. But then
(e + I) - (u + I) = e - u + IE JjI ~ (Rad R)jI = Rad (RjI).
Applying the lemma to the quotient ring RjI, we conclude that the coset
u + I = e + I and so the idempotents of RjI can be lifted.
The key to showing that the idempotents of RjRad R are liftable is the
circumstance that certain quadratic equations have a solution in the prime
radical of R.
Theorem 8-14. For any ring R, the idempotents of RjRad R can be
lifted into R.
Proof Let u + Rad R be an idempotent element of RjRad R, so that
u2 - u = r E Rad R. The problem is to find an idempotent e E R with
e - u E Rad R or, putting it another way, to obtain a solution a of the
equation (u + a)2 = u + a, with a E Rad R.
We first set a = x(1 - 2u), where x is yet to be determined. Now, the
requirement that the element u + a = u + x(1 - 2u) be idempotent is
equivalent to the equation
(x 2 - x)(1 + 4r) + r = O.
By the quadratic formula, this has a formal solution

X=Hl-~)
= Ij2(2r - (i)r 2 + (~)r3 - ... ).
CER TAIN RADICALS OF A RING 169

Since r is nilpotent (being a member of Rad R), the displayed series will
terminate in a finite number of steps; the result is a perfectly meaningful
polynomial in r with integral coefficients. Thus, the desired idempotent is
e = u + x(1 - 2u), where x E Rad R
Corollary. For any nil ideal I of R, the idempotents of RII can be lifted.
Proof Because I ~ Rad R, an appeal to the last lemma (with J = Rad R)
is legitimate.
Let us define a ring R to be primary whenever the zero ideal is a primary
ideal of R. This readily translates into a statement involving the elements
of R: R is a primary ring if and only if every zero divisor of R is nilpotent.
Integral domains are examples of primary rings. In general, primary rings
can be obtained by constructing quotient rings RIQ, where Q is a primary
ideal of R.
As an application of the preceding ideas, we shall characterize such
rings in terms of minimal prime ideals. (A prime ideal is said to be a minimal
prime ideal if it is minimal in the set of prime ideals; in a commutative ring
with identity, such ideals are necessarily proper.) The crucial step in the
proof is the corollary on page 164.
Theorem 8-15. A ring R is a primary ring if and only if R has a minimal
prime ideal which contains all zero divisors.
Proof For the first half of the proof, let R be a primary ring. Then the set
of zero divisors of R, along with zero, coincides with the ideal N of nilpotent
elements, and N will be prime. Being equal to the prime radical of R, N is
necessarily contained in every prime ideal of R; that is to say, N is a minimal
prime ideal.
The converse is less obvious; in fact, it is easiest to prove the contra-
positive form of the converse. Suppose, then, that R has a minimal prime
ideal P which contains all zero divisors and let a E R be any nonnilpotent
element. We define the set S by
S = {ranlr¢P; n ~ O}.
S is easily seen to be closed under multiplication and 1 E S. Notice, particu-
larly, that the zero element does not lie in S, for, otherwise, we would have
ran = 0 with an =1= 0; this implies that r is a zero divisor and therefore a
member of P. We now appeal to the corollary on page 164 to infer that
the complement of S contains a prime ideal P'. Since pi ~ R - S ~ P,
with P being a minimal prime ideal, it follows that R - S = P. But a E S,
whence a ¢ P, so that a cannot be a zero divisor of R. In other words, every
zero divisor of R is nilpotent, which completes the proof.
For the sake of refinement, let us temporarily drop the assumption that
all rings must have a multiplicative identity (commutativity could also be
170

abandoned, but this seems unnecessarily elaborate for the purposes in mind).
To make things more specific, we pose the problem of constructing a radical
which will agree with the Jacobson radical when an identity element is
available. Of course, it is always possible to imbed a ring in a ring with
identity, but the imbedding is often unnatural and distorts essential features
of the given ring.
The direct approach of considering the intersection of all maximal ideals
is not very effective, because one no longer knows that such ideals exist
(Theorem 5-2, our basic existence theorem for maximal ideals, clearly
requires the presence of an identity). A more useful clue is provided by
Theorem 8-2, which asserts that an element a E rad R if and only if 1 - ra
is invertible for every choice of r in R, or, to put it somewhat differently,
the principal ideal (1 - ra) = R for all r E R. This latter condition can be
written as {x - raxlx E R} = R for each r in R, and is meaningful in the
absence of a multiplicative identity. It thus would appear that Theorem 8-2
constitutes a hopeful starting point to the solution of the problem before
us. Needless to say, it will be necessary to introduce concepts capable of
replacing the notions of an invertible element and maximal ideal which
were so essential to our earlier work.
One begins by associating with each element a E R the set
la = {ax - XIXER}.
A moment's thought shows la to be an ideal of R. Now, it may very well
happen that la = R; in this event, we shall say that a is a quasi-regular
element. There is another way of looking at quasi-regularity:
Theorem 8-16. An element a of the ring R is quasi-regular if and only
if there exists some b E R such that
a + b - ab = 0.
The element b satisfying this equation is called a quasi-inverse of a.
Proof Suppose that a is quasi-regular, so that la = R. Since a E la, we
must have a = ab - b for suitable b in R, whence a + b - ab = o.
On the other hand, if there exists some element b E R satisfying
a + b - ab = 0, then a E la. Thus, for any r in R, ar E la. By virtue of the
definition of la' we also have ar - r E la' which implies that
r = ar - (ar - r) E la.
This means R ~ la' or rather R = la' and a is quasi-regular.
Corollary. An element a E R is quasi-regular if and only if a E la.
Here are some consequences: Every nilpotent element of R is quasi-
regular. Indeed, if an = 0, a straightforward calculation will establish that
CER T AIN RADICALS OF A RING 171

b = - a - a2 - ... - an - 1 is a quasi-inverse of a. Notice also that zero

is the only idempotent which is quasi-regular. For, if a2 = a and a + b -
ab = 0 for some b in R, then we have
a = a2 + ab - ab = a(a + b - ab) = O.
One of the most useful tools in handling the concept of quasi-regularity
is the so-called "circle operation" of Perlis [54]. Given a, bE R, we define
a b by
0

a b = a + b - abo
0

With this notation, Theorem 8-16 may be rephrased so as to assert that an

element a E R is quasi-regular if and only if there exists some second element
b E R for which a b = O.
0

It is a simple matter to verify that the pair (R, 0) is a semigroup with

identity element 0; in particular, one infers from this that quasi-inverses
are unique, whenever they exist. An even stronger result is that the quasi-
regular elements of R form a group with respect to the circle operation.
Lastly, let us call attention to the fact that if R possesses a multiplicative
identity 1, then
(1 - a)(1 - b) = 1 - a 0 b.
Accordingly, a is quasi-regular if and only if 1 - a is an invertible element
of R (Tying this idea more closely to Theorem 8-2, we see that the product
ra is quasi-regular for every r E R if and only if 1 - ra is invertible for all
r in R)

Example 8-6. Consider the assertion: if every element of a commutative

ring R is quasi-regular, with exactly one exception, then R must be a field.
To see this, let us take the element e to be the one exception; certainly
e =1= 0, since 0 is quasi-regular.
Now, e2 0 a = eo (-e 0 a) =1= 0 for each a E R, from which we infer
that e 2 = e. Observe also that if e a =1= e, then there would exist some
0

element bE R such that (e a) b = O. Associating, we obtain eo (a b) = 0

0 0 0

and so e is a quasi-regular element, a contradiction. Accordingly, we must

have e a = e for every choice of a E R or, upon expanding, ae = a for all
0

in R; this implies that e acts as a multiplicative identity for R

Finally, given an element x =1= 0, we write x = e - a, with a E R
Then, since a =1= e,
x(e - b) = (e - a)(e - b) = e - a 0 b = e
for suitable b ERIn other words, every nonzero element of R is invertible,
confirming R to be a field.
As heralded by our earlier remarks, we now define the l-radical (for
Jacobson, naturally enough) of a ring R to consist of those elements a for
172

which lar = R for every r E R; recasting this in terms of the notion of quasi-
regularity:
Definition 8-4. The J-radical J(R) of a ring R, with or without an
identity, is the set
J(R) = {a E Rlar is quasi-regular for all r E R}.
If J(R) = {O}, then R is said to be a J-semisimple ring.
To reinforce these ideas, let us consider several examples.
Example 8-7. The ring Ze of even integers is J-semisimple. For, suppose
that the integer n E J(Ze). Then, in particular, n2 is quasi-regular. But the
equation
n2 + x - n2 x = 0,
or, equivalently, n2 = (n 2 - l)x, has no solution among the even integers
unless n = O. This implies that J(Ze) = {O}.
Example 8-8. Any commutative regular ring R is J-semisimple. Indeed,
given a E J(R), there is some a' in R such that a2 a' = a. Now, aa' must be
quasi-regular, so we can find an element x E R satisfying
aa' +x- aa'x = O.
Multiplying this equation by a and using the fact that a2 a' = 0, we deduce
that a = 0, whence J(R) = {O}.
Example 8-9. Consider the ring F[[xJ] of formal power series over the
field F. As we know, F[[ xJ] is a commutative ring with identity in which
an element f(x) = L akxk is invertible if and only if ao =1= o. If f(x) belongs
to the principal ideal (x), thenf(x) has zero constantterm; hence, (1 - f(X))-l
exists in F[[xJJ. Takingf(x), = (1 - f(x)t 1, we see thatf(x) f(x)' = O.
0

Thus, every member of(x) is quasi-regular, which implies that (x) s;; J(F[[ xJ]).
On the other hand, any element not in (x) is invertible and therefore cannot
be in the J-radical. (In general, if a E J(R) has a multiplicative inverse, then
1 = aa -1 is quasi-regular; but zero is the only quasi-regular idempotent,
so that 1 = 0, a contradiction.) The implication is that J(F[[ xJ]) s;; (x)
and equality follows:
J(F[[ xJ]) = (x) = rad F[[ xJJ.
Turning once again to generalities, let us show that any element a E J(R)
is itself quasi-regular. Since ar is quasi-regular for each choice of r in R,
it follows that a2 in particular will be quasi-regular. Therefore, we can
obtain an element b E R for which a2 b = O. But a simple computation
0

shows that
ao((-a)ob) = (ao(-a))ob = a2 0b = 0
CERTAIN RADICALS OF A RING 173

One finds in this way that the element a is quasi-regular with quasi-inverse
(-a)ob.
It is by no means apparent from Definition 8-3 that J(R) forms an ideal
of R; our next concern is to establish that this is actually the case.
Theorem 8-17. For any ring R, the J-radical J(R) is an ideal of R.
Proof Suppose that the element a E J(R), so that for any choice of x E R,
ax is quasi-regular. If r E R, then certainly (ar)y = a(ry) must be quasi-
regular for all y in R, and therefore ar E J(R).
It remains to show that whenever a, b E J(R), then the difference a - b
lies in J(R). Given x, u, v E R, a fairly routine calculation establishes the
identity
(a - b)x 0 (u 0 v) = a(x - ux) 0 v + (-bx 0 u) - (-bx 0 u)v.
Taking stock of the fact that a, b belong to J(R), we can select an element u
such that -(bx) u = b( -x) u = 0 and a second element v for which
0 0

a(x - ax) v = 0; in consequence, (a - b)x (u v) = O. This being the

0 0 0

case, (a - b)x is quasi-regular for every x E R, whence a - bE J(R). Thus,

the J-radical satisfies the defining conditions for an ideal of R.
As one would expect, there are many theorems concerning the J-radical
which are completely analagous to theorems stated in terms ofthe Jacobson
radical. Although it would be tedious to prove all of these results, the
following deserves to be carried through.
Theorem 8-18. For any ring R, the quotient ring RjJ(R) is J-semisimple.
Proof Take a + J(R) to be an arbitrary element of J(RjJ(R)). Then
(a + J(R))(x + J(R)) = ax + J(R)
is quasi-regular for each x E R. Accordingly, there exists some coset
y + J(R) in RjJ(R), depending on both a and x, for which
(ax + J(R)) 0 (y + J(R)) = J(R).
But this implies that the element ax y lies in J(R), and, hence, is quasi-
0

regular as a member of R; say (ax 0 y) 0 z = 0, where Z E R. It follows from

the associativity of 0 that ax is itself quasi-regular in R, with quasi-inverse
y 0 z. Since this holds for every x E R, the element a belongs to J(R), and
we have a + J(R) = J(R), the zero of the quotient ring RjJ(R).
It turns out that the class of ideals which must replace the maximal
ideals are precisely those ideals whose quotient rings possess a multiplicative
identity.
Definition 8-5. An ideal I of the ring R is called modular (or regular, in
the older terminology) if and only if there exists an element e E R such
174

that ae - a E I for every a in R. Such an element e is said to be an

identity for R relative to I, or modulo I.
In passing, we should remark that whenever R has an identity element
1, then 1 can be taken as the element e of Definition 8-5, and all ideals of
R are modular. Notice, too, that if e is an identity for R relative to I, then
the same is true for the elements e + i, where i E I, and en (n E Z).
By a modular maximal ideal, we shall mean a maximal (hence, proper)
ideal which is also modular. Paralleling the proof of Theorem 5-5, it can
be shown without too much difficulty that a proper ideal M of R is a modular
maximal ideal if and only if the quotient ring R/M forms a field.
The existence of suitably many modular maximal ideals is assured by
the following:
Theorem 8-19. Each proper modular ideal of the ring R is contained
in a modular maximal ideal of R.
Proof Let I be a proper modular ideal of Rand e be an identity element
for R relative to 1. We consider the family d of all proper ideals of R which
contain I; because I itself is such an ideal, d is certainly nonempty.
It is important to observe that the element e lies outside each ideal J
of d. Indeed, if e did belong to J, we would then have ae E J for all a in R.
By virtue of the fact that I is modular, ae - a E I s; J, from which it follows
that
a = ae - (ae - a) E J.
One finds in this way that J = R, a flat contradiction, inasmuch as J is a
proper ideal of R by definition of d.
Now, let {Ji} be any chain of ideals from d. Then the set-theoretic
union u J i forms an ideal of R containing I. Since e ¢ u J i , this ideal is
proper, whence u J i E d. Thus, Zorn's Lemma asserts the existence of a
maximal ideal M of R with I S; M. Any such ideal will be modular, because
ae - a E I S; M for each element a E R.
This theorem has a number of important consequences (which we list
as corollaries) having to do with quasi-regularity.
Corollary 1. If the element a E R is not quasi-regular, then there exists
a modular maximal ideal M of R such that a ¢ M.
Proof Since a is not quasi-regular, la = {ra - rlr E R} forms a proper
ideal of R. Moreover, la is modular, with the element a as an identity for
R modulo la. Knowing this, it follows from the theorem that there exists a
modular maximal ideal M of R containing la and excluding a.
Corollary 2. If J(R) =1= R, then the ring R contains modular maximal
ideals; in fact, for any a ¢ J(R), there exists a modular maximal ideal
M with a¢M.
CERTAIN RADICALS OF A RING 175

Proof If the element a ¢ J(R), then there is some x E R such that ax is not
quasi-regular. Corollary 1 asserts the existence of a modular maximal ideal
of R which excludes ax and, in consequence, does not contain a.
Corollary 3. An element a E R is qlJasi-regular if and only if, for each
modular maximal ideal M of R, there exists an element b such that
a 0 bE M.
Proof The indicated condition is clearly necessary, for it suffices to take
the quasi-inverse of a as the element b. Suppose now that the condition is
satisfied, but that a does not possess a quasi-inverse. Then the modular
idealla = {ar - rlr E R} will be contained in some modular maximal ideal
M of R. By assumption, we can find an element b in R for which a 0 b E M.
But ab - bEla' whence
a = a0 b + (ab - b) E M.

It follows that ar E M for arbitrary r in R, and, consequently, that

r = ar - (ar - r) E M. Therefore, M = R, which is impossible.
In the presence of an identity element, the Jacobson radical rad R is
the intersection of all the maximal ideals of a ring R. One would rightly
suspect that there is a similar characterization of the J-radical in terms of
modular maximal ideals (the sole difference being that, in the present
setting, we must impose the demand that J(R) does not exhaust the ring R).
Theorem 8-20. If R is a ring such that J(R) =1= R, then

J(R) = 1\ {MIM is a modular maximal ideal of R}.

Proof As so often happens, one inclusion will be quite straightforward and
easy, and the other will be deeper and more complicated. In the first place,
suppose that the element a lies in every modular maximal ideal of R, but
that a ¢ J(R). Using Corollary 2, we could then find a modular maximal
ideal M for which a ¢ M, a contradiction; consequently, a E J(R).
Going in the other direction, take a in J(R). We wish to show that
a E M, where M is any modular maximal ideal of R. Assume for the moment
that a ¢ M. Owing to the fact that M is maximal, the ideal generated by M
and a must be the whole ring R; therefore (in the absence of an identity),

R = {i + ra + naliEM, rER, nEZ}.

Now, let e be an identity element for R modulo M. Then there exist suitable
i E M, r E R and an integer n for which
e = i + ra + na.
As J(R) forms an ideal of R, the sum ra + na E J(R), so that e - i E J(R).
176

This implies that e - i is quasi-regular, say with quasi-inverse x. From the

equation (e - i) a X = 0, together with the modularity of M, we obtain
e = i - ix + (xe - x) E M.
In consequence, M = R, and an obvious contradiction ensues. Thus,
a E M, from which one concludes that J(R) is contained in the intersection
of the modular maximal ideals of R, completing the proof.
The hypothesis that J(R) 1= R is certainly fulfilled whenever the ring
R possesses a multiplicative identity 1. Specifically, the element 1 itself is
not quasi-regular, whence 1 ¢ J(R); in fact, if 1 + b - Ib = for some b °
in R, we would have 1 = 0, a contradiction. When an identity element is
available, all ideals of R are automatically modular. In this situation, the
J-radical will coincide with the Jacobson radical of R: J(R) = rad R.
If J(R) = R, then the ring R may contain maximal ideals, but no such
ideal can be modular. Indeed, suppose that I is any modular ideal of R,
R, with e acting as an identity for R modulo I. By supposition, the element
e E J(R), so that e has a quasi-inverse e'. The modularity of I then yields
e = e' e - e' E I, which implies that I = R. Accordingly, the ring R
possesses no proper modular ideals and, in particular, no modular maximal
ideals. However, the possibility of the existence of maximal ideals in R is
not excluded.
The following theorem provides a convenient result with which to close
this chapter.
Theorem 8.21. A ring R can be imbedded in a ring R' with identity
such that J(R) = rad R'.
Proof If R already has an identity, we simply take R' = R. Otherwise,
we imbed R in the ring R' = R x Z in the standard way (see Theorem
2-12 for details). Then R, or more precisely, its isomorphic image R x {O},
is an ideal of R' and R'/R ~ Z; thus, R'/R is semisimple. This being so, it
follows from Theorem 8-5 that rad R' s R. Since R is an ideal of R', we
also have J(R) = J(R') n R = rad R' n R (Problem 26). But rad R' s R,
which implies that J(R) = rad R'.

PROBLEMS
Unless indicated to the contrary, all rings are assumed to be commutative with identity .
1. Describe the Jacobson radical of the ring Z. of integers modulo n. [Hint: Consider
the prime factorization of n.] In particular, show that Z. is semisimple if and
only if n is a square-free integer.
2. Prove that F[ x], the ring of polynomials in x over a field F, is semisimple.
PROBLEMS 177

3. Prove that rad R is the largest (in the set-theoretic sense) ideal 1 of R such that
1 + a is invertible for all a E 1.
4. a) Let the ring R have the property that all zero divisors lie in fad R. If(a) = (b),
show that the elements a and b must be associates.
b) Verify that if the element a E rad R and ax = x for some x E R, then x = o.
5. Provethatapowerseriesf(x) = ao + a1x + ... +a"xn + ... belongstoradR[[x]]
if and only if its constant term ao belongs to rad R.
6. Prove the following assertions concerning semisimple rings:
a) A ring R is semisimple if and only if a =1= 0 implies that there exists some element
r E R for which 1 - ra is not invertible.
b) Every commutative regular ring is semisimple.
c) Suppose that {Ii} is a family of ideals of R such that R/li is semisimple for
each i and Illi = {a}. Then R itself is a semisimplering. [Hint: Theorem 8-5.]
7. If R = Rl (9 R2 (9 ... (9 Rn is the direct sum of a finite number of rings Ri
(i = 1, 2, ... , n), prove that

rad R = rad Rl (9 rad R2 (9 ... (9 rad R n.

8. Establish that the conditions below are equivalent:

a) the ring R has exactly one maximal ideal (that is, R is a local ring);
b) rad R is a maximal ideal of R;
c) the set of noninvertible elements of R coincides with rad R;
d) the set of noninvertible elements of R form an ideal;
e) the sum of two noninvertible elements of R is again noninvertible;
f) for each element r E R, either r or 1 - r is invertible.
9. Let R be a principal ideal domain. If the element a E R has the prime factorization
a = p~lp~2 ... p~r, prove that the Jacobson radical of the quotient ring R/(a) is
(PIP2 ... Pr)/(a). [Hint: The maximal ideals ofR containing (a) are (PI)' (P2)' ... , (Pr)·]
10. Letfbe a homomorphism from the ring R onto the ring R'. Show thatf(rad R) S;
rad R' and, whenever kerf S; rad R, then rad R = f-l(rad R'); do the same for
RadR.
11. Prove: An ideal 1 of R is semiprime if and only if a2 E 1 implies that a E 1.

12. Establish that an ideal 1 of R contains a prime ideal if and only if for each n,
a1a2 ... an = 0 implies that ak E 1 for some k. [Hint: The set

S = {b 1 b2 ••• bnlbk¢J; n ~ 1}
is closed under multiplication and 0 ¢ S.]
13. Show that the prime radical of a ring R contains the sum of all nilpotent ideals of R.
14. Establish the equivalence of the statements below:
a) {O} is the only nilpotent ideal of R;
b) R is without prime radical; that is, Rad R = {O};
c) for any ideals 1 and J of R, lJ = {OJ implies that 1 Il J = {OJ.
178

15. A multiplicatively closed subset S of the ring R is said to be saturated if ab E S

implies that both a E Sand b E S. Prove that
a) S is a saturated multiplicatively closed subset of R if and only if its complement
R - S is a union of prime ideals;
b) the set of non-zero-divisors of R is a saturated multiplicatively closed subset
(hence, the set of zero divisors of R, along with zero, is a union of prime ideals).

16. a) Prove that Rad R is the maximal nil ideal of R (maximal among the set of nil
ideals); this property is often taken as the definition of the prime radical of R.
b) If R has no nonzero nil ideals, deduce that the polynomial ring R[x] is semi-
simple.
c) Let char R = n > 0. Prove that if R is without prime radical, then n is a
square-free integer. [H int: Assume that n = p2 q for some prime p; then there
exists an element a E R such that pqa =I 0, but (pqa)2 = 0.]
17. Prove that the following statements are equivalent:
a) R has a unique proper prime ideal;
b) R is a local ring with rad R = Rad R;
c) every noninvertible element of R is nilpotent;
d) R is a primary ring and every noninvertible element of R is either a zero divisor
or zero.
18. Supply a proof of Theorems 8-11 and 8-12.
19. A ring R is termed a Hilbert ring if each proper prime ideal of R is an intersection
of maximal ideals. Prove that:
a) R is a Hilbert ring if and only if for every proper ideal I of R,
rad (R/I) = Rad (R/I).
b) Any homomorphic image of a Hilbert ring is again a Hilbert ring.
c) If the polynomial ring R[x] is a Hilbert ring, then R is one also. [Hint: Utilize
(b) and the fact that R[ x ]/(x) ~ R.]
20. If I is an ideal of the ring R, prove each of the following statements:
a) The nil radical of I is the intersection of all the minimal prime ideals of I.
b) Rad R is the intersection of all the minimal prime ideals of R.
c) The union of all the minimal prime ideals of R is the set

S = {r E Rlra E Rad R, for some a ¢ Rad R}.

d) If I is a primary ideal of R, then ji is the only minimal prime ideal of I.

[Hint: Problem 19, Chapter 5.]
In Problems 21-30, the ring R need not possess an identity element.
21. Consider the ring P(X) of subsets of some (nonempty) set X. Show that in this
setting the circle operation reduces to the union operation and determine the
quasi-regular elements.
22. a) Prove that if the element a E R has the property that an is quasi-regular for
some n E Z+, then a itself must be quasi-regular. [Hint: an = a Lk;;~ ak ).]
0 (-
PROBLEMS 179

b) Verify that the annihilator of a J -semisimple ring R is zero; in other words,

ann R = {o}.
23. IfJis a homomorphism from the ring R onto the ring R', establish the inclusion
J(J(R») ~ J(R'); also show that ifkerf~ J(R), then J(R) = r1(J(R'»).

24. Prove each of the statements below:

a) J(R) contains every nil ideal of the ring R.
b) J(R) is a semiprime ideal of R. [Hint: Theorem 5-11.]
c) For any ring R, Rad R ~ J(R).
25. If we define R = {2nj(2m + 1)ln, mE Z}, then R forms a commutative ring under
ordinary addition and multiplication. Show that J(R) = R, while Rad R = {o}.
[Hint:(~)o(
2m + 1 2( - n
-2n
+ m) +
)= 0.]
1
26. Let I be an ideal ofthe ring R. Regarding I as a ring, deduce that J(I) = J(R) n I.
27. Prove that if the element a E J(R) is idempotent modulo the ideal I (in other words,
(a + W = a + I), then a E 1.
28. Assume that the ideal I of R consists of elements which are quasi-regular modulo
J(R). (yVe say that a is quasi-regular modulo J(R) provided that there exists an
element bE R such that a bE J(R»). Establish that I ~ J(R).
0

29. a) Prove that an element a E R fails to be quasi-regular if and only if a is an identity

for R relative to some proper modular ideal of R.
b) If I is an ideal of the ring Rand K is a modular ideal of I, show that K is also an
ideal of R.
30. We shall call an ideal I of R regular if the quotient ring RjI is a regular ring; in
other words, if for each a E R, there exists an element bE R such that a2b - a E I.
Prove that
a) Every modular maximal ideal of R is regular.
b) If 11 and 12 are both regular ideals of R, then so also is 11 n 12 , [Hint: Given
a E R, there exist b, C E R such that a2b - a E 11 and (a 2b - a)2c - (a 2b - a)
belongs to 12 ; rewrite the last expression.]
c) J(R) = n {III is a regular ideal of R}.
[Hint: Assume that a E J(R), but not the right-hand side, so that a ¢ I for some
regular ideal 1. If S = I n J(R), then a2 b - a E S for some b in R. Take
e = ab E J(R). Then, e2 - e E S. Show that e E S, which leads to the
contradiction that a E I.]
APPENDIX A

RELATIONS

We herein append a few definitions and general results concerning certain

types of relations that can be imposed on a set. For the most part, our
attention is confined to two relations of particular utility, namely, equivalence
relations and order relations.
Intuitively, a (binary) relation on a set S provides a criterion such that
for each ordered pair (a, b) of elements of S, one can determine whether the
statement "a is related to b" is meaningful (in the sense of being true or
false according to the choice of elements a and b). The relation is completely
characterized once we know the set of all those pairs for which the first
component stands in that relation to the second. This idea can best be
formulated in set-theoretic language as
Definition A-t. A (binary) relation R in a nonempty set S is any subset
of the Cartesian product S x S.
If R is a relation, we express the fact that the pair (a, b) E R by saying
that a is related to b with respect to the relation R, and we write aRb. For
instance, the relation < in R can be represented by all points in the plane
lying above the diagonal line y = x; it is customary to write 3 < 4, rather
than the awkward (3, 4) E <.
Our immediate concern is with equivalence relations. In practice, these
arise whenever it is desirable to identify, as a single entity, all elements of a
set that have some preassigned characteristic.
Definition A-2. A relation R in a set S is an equivalence relation in S
provided that it satisfies the three properties,
1) aRa for all a E S (reflexive property),
2) if aRb, then bRa (symmetric property),
3) if aRb and bRc, then aRc (transitive property).
Equivalence relations are usually denoted by the symbol '" (pronounced
"tilda") rather than by R as heretofore. With this change in notation, the
conditions of the above definition may be recast in a more familiar form:
288

for every a, b, c E S, (1) a '" a, (2) a '" b implies b '" a, (3) a '" band b '" c
together imply a '" c. We say that a is equivalent to b if and only if a '" b
holds.
In the following set of miscellaneous examples, we leave to the reader
the task of verifying that each relation described actually is an equivalence
relation.
Example A-I. Let S be a nonempty set and define, for a, b E S, a '" b if
and only if a = b; that is, a and b are the same element. (Technically
speaking, '" is the subset {(a, a)la E S} of S x S.) Then '" satisfies the
requirements of Definition A-2, showing that the equivalence concept is a
generalization of equality.
Example A-2. In the Cartesian product R# x R#, let (a, b) '" (c, d) signify
that a - c and b - d are both integers. A simple calculation reveals that
"', so defined, is an equivalence relation in R # X R # .
Example A-3. As another illustration, consider the set L of all lines in a
plane. Then the phrase "is parallel to" is meaningful when applied to the
elements of L and may be used to define a relation in L. If we agree that
any line is parallel to itself, this yields an equivalence relation.
Example A-4. Let f: X --. Y be a given mapping. Take for '" the relation
a '" b if and only if f(a) = f(b); then '" is an equivalence relation in X,
called the equivalence relation associated with the mapping/. More generally,
if fF is an arbitrary family offunctions from X into Y, an equivalence relation
can be introduced in X by interpreting a '" b to meanf(a) = f(b) for every
f E fF. (The underlying feature in the latter case is that any intersection
of equivalence relations in X is again an equivalence relation, for
'" = nfe, {(a, b)lf(a) = f(b)}.)
One is often led to conclude, incorrectly, that the reflexive property is
redundant in Definition A-2. The argument proceeds like this: if a '" b,
then the symmetric property implies that b '" a; from a '" band b '" a,
together with the transitivity of "', it follows that a '" a. Thus, there appears
to be no necessity for the reflexive condition at all. The flaw in this reasoning
lies, of course, in the fact that, for some element a E S, there may not exist
any b in S such that a '" b. As a result, we would not have a '" a for every
member of S, as the reflexive property requires.
Any equivalence relation '" determines a separation of the set S into
a collection of subsets of a kind which we now describe. For each a E S,
let [a] denote the subset of S consisting of all elements which are equivalent
to a:
[a] = {b E Sib", a}.
This set [a] is referred to as the equivalence class determined by a. (The
reader should realize that, in general, the equivalence class [a] is the same
RELA TIONS 289

as the class [a'] for many elements a' E S.) As a notational device, let us
henceforth use the symbol Sj'" to represent the set of all equivalence classes
of the relation '" ; that is,
Sj", = {[a]ja E S}.
Some of the basic properties of equivalence classes are listed in the
theorem below.
Theorem A-I. Let", be an equivalence relation in the set S. Then,
1) for each a E S, the class [a] =1= 0;
2) if b E [a], then [a] = [b]; in other words, any element of an
equivalence class determines that class;
3) for all a, bE S, [a] = [b] if and only if a '" b;
4) for all a, bE S, either [a] n [b] = 0 or [a] = [b];
5) U [a] = S.
aeS
Proof. Clearly, the element a E [a], for a '" a. To prove the second
assertion, let bE [a], so that b '" a. Now, suppose that x E [a], which
means x '" a. Using the symmetric and transitive properties of "', we thus
obtain x '" b, whence x E [b]. Since x is an arbitrary member of [a], this
establishes the inclusion [a] £; [b]. A similar argument yields the reverse
inclusion and equality follows. As regards (3), first assume that [a] = [b];
then a E [a] = [b] and so a '" b. Conversley, if we let a '" b, then the
element a E [b]; hence, [a] = [b] from (2).
To derive (3), suppose that [a] and [b] have an element in common,
say, C E [a] n [b]. Statement (2) then informs us that [a] = [c] = [b].
In brief, if [a] n [b] =1= 0, then we must have [a] = [b]. Finally, since
each class [a] £; S, the inclusion u {[a]la E S} £; S certainly holds. For
the opposite inclusion, it is enough to show that each element a E S belongs
to some equivalence class; but this is no problem, for a E [a].
As evidenced by Example A-4, any mapping determines an equivalence
relation in its domain. The following corollary indicates that every
equivalence relation arises in this manner; that is to say, each equivalence
relation is the associated equivalence relation of some function.
Corollary. Let", be an equivalence relation in the set S. Then there
exists a set T and a mapping g: S ~ T such that a '" b if and only if
g(a) = g(b).
Proof. Simply take T = Sj'" and g : S ~ T to be the mapping defined by
g(a) = [a]; in other words, send each element of S onto the (necessarily
unique) equivalence class to which it belongs. By the foregoing theorem,
a '" b if and only if [a] = [b], or equivalently, g(a) = g(b).
290

Example A-5. This example is given to illustrate that any mapping can be
written as the composition of a one-to-one function and an onto function.
Letf: X - Ybe an arbitrary mapping and consider the equivalence relation
- associated withf If the element a e X, then we have
[a] = {b e Xif(b) = f(a)} = f-l(J(a».
In effect, the equivalence classes for the relation - are just the inverse
images off-l(y), where yef(X) £; Y.
Now, define the function!: Xj- - Yby the ruleJ([a]) = f(a). Since
[a] = [b] ifand only iff(a) = f(b),Jis well-defined. Observe that whereas
the original function f may not have been one-to-one, J happens to be
one-to-one; indeed, J([ a]) = J([b]) implies that f(a) = feb), whence
[a] = [b]. At this point, we introduce the onto function g: X - Xj- by
setting g(a) = [a]. Then f(a) = 1([a]) = J(g(a») = (J 0 g)(a) for all a in
X, in consequence of which f = log. This achieves our stated aim.
We next connect the notion of an equivalence relation in S with that
of a partition of S.
Definition A-3. By a partition of the set S is meant a family r!J of subsets
of S with the properties
1) 0 ¢ r!J,
2) for any A, B e r!J, either A = B or A n B = 0 (pairwise disjoint),
3) u r!J = S.
Expressed otherwise, a partition of S is a collection r!J of nonempty
subsets of S such that every element of S belongs to one and only one member
of r!J. The set Z of integers, for instance, can be partitioned into the subsets
of odd and even integers; another partition of Z might consist of the sets
of positive integers, negative integers and {O}.
Theorem A-I may be viewed as asserting that each equivalence relation
- on a set S yields a partition of S, namely, the partition Sj - into the
equivalence classes for -. (In this connection, notice that, for the equiva-
lence relation of equality, the corresponding classes contain only one element
each; hence, the resulting partition is the finest possible.) We now reverse
the situation and show that a given partition of S induces an equivalence
relation in S. But first a preliminary lemma is required.
Lemma. Two equivalence relations - and -' in the set S are the same
if and only if Sj - and Sj -' are the same.
Proof If - and -' are the same, then surely Sj- = Sj-'. So, suppose
that - and -' are distinct equivalence relations in S. Then there exists
a pair of elements a, b e S which are equivalent under one of the relations,
RELA TIONS 291

but not under the other; say a '" b, but not a ",' b. By Theorem A-I,
there is an equivalence class in S/ '" containing both a and b, while no such
class appears in S/"". Accordingly, S/'" and S/",' differ.
Theorem A-2. If &J is a partition of the set S, then there is a unique
equivalence relation in S whose equivalence classes are precisely the
members of &J.
Proof Given a, b E S, we write a '" b if and only if a and b both belong to
the same subset in &J. (The fact that &J partitions S guarantees that each
element of S lies in exactly one member of &J.) The reader may easily check
that the relation "', defined in this way, is indeed an equivalence relation
in S.
Let us prove that the partition &J has the form S/ "'. If the subset P E &J,
then a E P for some a in S. Now, the element b E P if and only if b '" a,
or, what amounts to the same thing, if and only if b E [a]. This demonstrates
the equality P = [a] E S/ "'. Since this holds for each P in &J, it follows
that &J s;;; S/ "'. On the other hand, let [a] be an arbitrary equivalence
class and P be the partition set in &J to which the element a belongs. By
similar reasoning, we conclude that [a] = P; hence, S/'" s;;; &J. Thus, the
set of equivalence classes for '" coincides with the partition &J. The
uniqueness assertion is an immediate consequence of the lemma.
To summarize, there is a natural one-to-one correspondence between
the equivalence relations in a set and the partitions of that set; every
equivalence relation gives rise to a partition and vice versa. We have a
single idea, which has been considered from two different points of view.
Another type of relation which occurs in various branches of mathe-
matics is the so-called partial order relation. Just as equivalence generalizes
equality, this relation (as we define it below) generalizes the idea of "less
than or equal to" on the real line.
Definition A-4. A relation R in a nonempty set S is called a partial
order in S if the following three conditions are satisfied:
1) aRa (reflexive property),
2) if aRb and bRa, then a = b (antisymmetric property),
3) if aRb and bRc, then aRc (transitive property),
where a, b, c denote arbitrary elements of S.
From now on, we shall follow custom and adopt the symbol ::;; to
represent a partial order relation, writing a ::;; b in place of aRb; the fore-
going axioms then read: (1) a ::;; a, (2) if a ::;; band b ::;; a, then a = b,
and (3) if a ::;; b and b ::;; c, then a ::;; c. As a linguistic convention, let us
also agree to say (depending on the circumstance) that "a precedes b" or
292

"a is a predecessor of b", or "b succeeds a", or "b is a successor of a" if

a ::;; b and a =1= b. By a partially ordered set is meant a pair (S, ::;; ) consisting
of a set S and a partial order relation ::;; in S. In practice, one tends to
ignore the second component and simply speak of the partially ordered set
S, or, when more precision is required, say that S is partially ordered by ::;;.
If A is a subset of a partially ordered set S, then the ordering of S
restricted to A is a partial ordering of A, called the induced partial order;
in this sense, any subset of a partially ordered set becomes a partially ordered
set in its own right. When considering subsets of a partially ordered set as
partially ordered sets, it is always the induced order that we have in mind.
Let S be a set partially ordered by the relation ::;;. Two elements a
and b of S such that either a ::;; b or b ::;; a are said to be comparable. In
view of the reflexivity of a partial order, each element of S is comparable
to itself. There is nothing, however, in Definition A-3 that ensures the
comparability of every two elements of S. Indeed, the qualifying adverb
"partially" in the phrase "partially ordered set" is intended to emphasize
that there may exist pairs of elements in the set which are not comparable.
Definition A-5. A partial order::;; in a set S is termed total (sometimes
simple, or linear) if any two elements of S are comparable; that is,
a ::;; b or b ::;; a for any two elements a and b of S.
A partially ordered set (S,::;; ) whose relation ::;; constitutes a total
order in S is called a totally ordered set or, for short, a chain.
Let us pause to illustrate some of the preceding remarks.
Example A-6. In the set R # of real numbers, the relation ::;; (taken with
the usual meaning) is the most natural example of a total ordering.
Example A-7. Given the set Z+ of positive integers, define a ::;; b if and
only if a divides b. This affords a partial ordering of Z +, which is not total;
for instance, the integers 4 and 6 are not comparable, since neither divides
the other.
Example A-S. Let S be the collection of all real-valued functions defined
on a nonempty set X. Iff::;; g is interpreted to mean f (x) ::;; g(x) for all
x E X, then::;; partially, but not totally, orders S.
Example A-9. For a final illustration, consider the set P(X) of all subsets
of a set X. The relation A ::;; B if and only if A s;;; B is a partial ordering of
P(X), but not a total ordering provided that X has at least two elements.
For example, if X = {1,2, 3}, and A = {1,2}, B = {2,3}, then neither
A ::;; B nor B ::;; A holds. As regards terminology, any family of sets ordered
in this manner will be said to be ordered by inclusion.
Let (A, ::;;) and (B, ::;;) be two partially ordered sets (when there is no
danger of confusion, we write ::;; for the partial orders in both A and B).
RELATIONS 293

A mappingf: A --+ B is said to be order-preserving or an order-homomorphism

if for all a, b E A, a :s; b implies f(a) :s; f(b) in B. A one-to-one order-
homomorphismf of the set A onto B whose inverse is also an order-homo-
morphism (from B onto A) is an order-isomorphism. If such a function
exists, we say that the two partially ordered sets (A,:S; ) and (B,:S; ) are
order-isomorphic. When the partial order is the primary object of interest
and the nature of the elements plays no essential role, order-isomorphic sets
can be regarded as identical.
The coming theorem emphasizes the fundamental role of our last
example on partially ordered sets (Example A-9), for it allows us to represent
any partially ordered set by a family of sets.
Theorem A-3. Let A be a set partially ordered by the relation :s; .
Then A is order-isomorphic to a family of subsets of A, partially ordered
by i~clusion.
Proof For each a E A, let la = {x E Alx :s; a}. It is not hard to verify
that the mapping f: A --+ P(A) defined by f(a) = la is an order-homo-
morphism of A into P(A). Indeed, if a :s; b, then the condition x :s; a
implies x :s; b and therefore la ~ lb' or, equivalently,J(a) ~ f(b). To see
that f is one-to-one, suppose a, b E A are such that f(a) = f(b). Then the
element a E la = lb' and, hence, a :s; b by definition of lb; likewise, b :s; a,
from which it follows that a = b. Finally, the inverse f - 1 is also order-
preserving. For, if the inclusion la ~ lb holds, then a E lb and so a :s; b.
These calculations make it clear that A is order-isomorphic to a certain
set of subsets of P(A).
Corollary. For no set A is A order-isomorphic to P(A).
Proof We argue that iff: A --+ P(A) is any order-homomorphism from A
into P(A), then f cannot map onto P(A). For purposes of contradiction,
assume that f does carry A onto P(A). Define B = {a E Ala ¢f(a)} and
B* = {c E Alc :s; a for some a E B}.
By supposition, the set B* = f(b) for some element bE A. If b ¢ B*,
then, according to the definition of B, bE B ~ B*, a contradiction. Hence,
b E B* and so b :s; a for some a in B. From the order-preserving character
of J, B* = f(b) ~ f(a). But then, a E B ~ B* ~ f(a). The implication is
that a ¢ B, which is again a contradiction. -

In an ordered set, there are sometimes elements with special properties

that are worth mentioning.
Definition A-6. Let S be a set partially ordered by the relation :s;. An
element XES is said to be a minimal (maximal) element of S if a E S
and a :s; x (x :s; a) imply a = x.
294

In other words, x is a minimal (maximal) element of S if no element of

S precedes (exceeds) x. It is not always the case that a partially ordered
set possesses a minimal (maximal) element and, when such an element exists,
there is no guarantee that it will be unique.
Example A-tO. The simplest illustration of a partially ordered set without
minimal or maximal elements is furnished by the set R #- with the ordering
~ in the usual sense.

Example A-H. In the collection P(X) - {0} of all nonempty subsets of

a nonempty set S (partially ordered by set-theoretic inclusion), the minimal
elements are those subsets consisting of a single element.
Example A-12. Consider the set S of all integers greater than 1 and the
partial order ~ defined by a ~ b if and only if a divides b. In this setting,
the prime numbers serve as minimal elements.
It is technically convenient to distinguish between the notion of a
minimal (maximal) element and that of a first (last) element.
Definition A-7. Let S be a set partially ordered by the relation ~.
An element xeS is called the first (last) element of S if x ~ a (a ~ x)
for all a E S.
Let us point out immediately the important distinction between first
(last) elements and minimal (maximal) elements. Definition A-7 asserts that
the first (last) element of a partially ordered set S must be comparable to
every element of S. On the other hand, as Definition A-6 implies, it is not
required that a minimal (maximal) element be comparable to every element
of S, only that there be no element in S which precedes (exceeds) it. A
minimal (maximal) element has no predecessors (successors), whereas a first
(last) element precedes (succeeds) every element. Clearly, any first (last)
element is a minimal (maximal) element, but not conversely.
First (last) elements of partially ordered sets are unique, if they exist
at all. Indeed, suppose that the partially ordered set (S, ~ ) has two first
elements, say x and y; then, x ~ y and y ~ x, so that x = y by the anti-
symmetric property. Thus, x is unique and we are justified in using the
definite article when referring to the first element of S. A similar argument
holds for last elements.
Let us introduce some additional terminology pertaining to partially
ordered sets.

Definition A-8. Let S be a set partially ordered by the relation ~ and

let A be a subset of S. An element XES is said to be a lower (upper)
bound for A if x ~ a (a ~ x) for all a E A.
RELA TIONS 295

We emphasize that a lower (upper) bound for a subset A of a partially

ordered set is not required to belong to A itself. If A happens to have a
first (last) element, then the same element is a lower (upper) bound for A;
conversely, if a lower (upper) bound for A is contained in the set A, then it
serves as the first (last) element for A. Notice, too, that a lower (upper)
bound for A is a lower (upper) bound for any subset of A.
A subset of a partially ordered set need not have upper or lower bounds
(just consider Z S R# with respect to ~ ) or it may have many. For an
example of this latter situation, one may turn to the family P(X) of all
subsets of a set X, with the order being given by the inclusion relation; an
upper bound for a subfamily .91 S P(X) is any set containing u .91, while
a lower bound is any set contained in n d.
APPENDIX B

ZORN'S LEMMA

In this Appendix, we give a brief account of some ofthe axioms of set theory,
with the primary purpose of introducing Zorn's Lemma. Our presentation
is descriptive and most of the facts are merely stated. The reader who is
not content with this bird's-eye view should consult [12J for the details.
As we know, a given partially ordered set need not have a first element
and, if it does, some subset could very well fail to possess one. This prompts
the following definition: a partially ordered set (S, :::;;) is said to be well-
ordered if every nonempty subset A !;;; S has a first element ("with respect
to :::;; .. being understood). The set Z + is well-ordered by the usual :::;;; each
nonempty subset has a first element, namely, the integer of smallest
magnitude in the set.
Notice that any well-ordered set (S, :::;; ) is in fact totally ordered. For,
each subset {a, b} !;;; S must have a first element. According as the first
element is a or b, we see that a :::;; b or b :::;; a, whence the two elements
a and b are comparable. Going in the other direction, any total ordering
of a finite set is a well-ordering of that set. Let it also be remarked that a
subset of a well-ordered set is again well-ordered (by the restriction of the
ordering).

ExampleD-I. Consider the Cartesian product S = Z+ xZ+. We par-

tially order S as follows: if (a, b) and (a', b') are ordered pairs of positive
integers, (a, b) :::;; (a', b') means that (1) a < a' (in the usual sense) or (2)
a = a' and b :::;; b'. (This is called the lexicographic order of Z+ x Z+,
because of its resemblance to the way words are arranged in a dictionary.)
For instance, (4,8) :::;; (5,2), while (3,5) :::;; (3,9). To confirm that :::;; is a
well-ordering of S, let 0 +- A !;;; S and define B = {a E Z+ I(a, b) E A}.
Since A is a nonempty subset of Z +, it has a first element, call it ao. Now,
let C = {b E Z+I(ao, b) E A}. Again, the well-ordering of Z+ under :::;;
guarantees that C has a first element, say boo We leave it to the reader to
convince himself that the pair (a o, bo) serves as the first element of A, thereby
making S a well-ordered set relative to :::;;.
?QI'O
ZORN'S LEMMA 297

A fundamental axiom of set theory, which has a surprising variety of

logically equivalent formulations, is the so-called Well-Ordering Theorem
of Zermelo (1904). The designation "theorem" notwithstanding, we take
this to be an axiom (assumed and unproven) of our system. We state:
Zermelo's Theorem. Any set S can be well-ordered; that is, there is a
partial ordering ::::;; for S such that (S, ::::;;) is a well-ordered set.
Accepting the existence of such orderi~gs, we do not pretend at all to
be able to specify them. Indeed, nobody has ever "constructed" an explicit
function that well-orders an uncountable set. Moreover, the promised
well-ordering may bear no relation to any other ordering that the given set
may already possess; the well-ordering of R #- , for instance, cannot coincide
with its customary ordering.
Zermelo based the "proof" of his classical Well-Ordering Theorem on
a seemingly innocent property whose validity had never been questioned
and which has since become known as the axiom of choice. To state this
axiom, we first need the definition of a choice function.
Definition B-1. Let ~ be a (nonempty) collection of nonempty sets. A
function f: ~ -+ u ~ is called a choice function for ~ if f(A) E A for
every set A in ~.
Informally, a choice function f can be thought of as "selecting" from
each set A E ~ a certain representative elementf(A) of that set. As a simple
illustration, there are two distinct choice functions f1 and f2 for the family
of nonempty supsets of {I, 2}:

f1({1,2}) = 1, f1({1}) = 1, f1({2}) = 2,

f2({1,2}) = 2, f2({I}) = 1, f2({2}) = 2.
The question arises whether this selection process can actually be carried
out when ~ has infinitely many members. The possibility of making such
choices is handled by the axiom mentioned above:
Axiom of Choice. Every collection ~ of nonempty sets has at least one
choice function.
Since this general principle of-choice has a way of slipping into proofs
unnoticed, the reader should become familiar with its disguised forms. For
instance, one often encounters the following wording: if {XJ is a family
of nonempty sets indexed by the nonempty set J, then the Cartesian product
X ieJ Xi is nonempty (it should be clear that the elements of x Xi are pre-
cisely the choice functions for {Xi})' For another common phrasing, which
again expresses the idea of selection, let ~ be a collection of disjoint, non-
empty sets. The axiom of choice, as we have stated it, is equivalent to
298

asserting the existence of a set S with the property that A (') S contains
exactly one element, for each A in ~.
Granting Zermelo's Theorem, it is clear that a choice function f can
be defined for any collection ~ of nonempty sets: having well-ordered u ~,
simply take f to be the function which assigns to each set A in ~ its first
element. As indicated earlier, Zermelo's Theorem was originally derived
from the axiom of choice, so that these are in reality two equivalent principles
(although seemingly quite different).
Although the axiom of choice may strike the reader as being intuitively
obvious, the sou:ldness of this principle has aroused more philosophical
discussion than any other single question in the foundations of mathematics.
At the heart of the controversy is the ancient problem of existence. Some
mathematicians believe that a set exists only if each of its elements can be
designated specifically, or at the very least if there is a rule by which each
of its members can be constructed. A more liberal school of thought is that
an axiom about existence of sets may be used if it does not lead to a contra-
diction. In 1938 Godel demonstrated that the axiom of choice is not in
contradiction with the other generally accepted axioms of set theory
(assuming that the latter are consistent with one another). It was sub-
sequently established by Cohen (1963) that the denial of this axiom is also
consistent with the rest of set theory. Thus, the axiom of choice is in fact
an independent axiom, whose use or rejection is a matter of personal
inclination. The feeling among most mathematicians today is that the axiom
of choice is harmless in principle and indispensable in practice (provided
that one calls attention to the occasions of its use). It is also valuable as
an heuristic tool, since every proof by means of this assumption represents
a result for which we can then seek proofs along other lines.
A non-constructive criterion for the existence of maximal elements is
given by the so-called "maximality principle", which generally is cited in
the literature under the name Zorn's Lemma. (From the point of view of
priority, this principle goes back to Hausdorff and Kuratowski, but Zorn
gave a formulation of it which is particularly suitable to algebra; he was also
the first to state, without proof, that a maximality principle implies the
axiom of choice.)
Zorn's Lemma. Let S be a nonempty set partially ordered by :::;; .
Suppose that every subset A S;; S which is totally ordered by :::;; has an
upper bound (in S). Then S possesses at least one maximal element.
Zorn's Lemma is a particularly handy tool when the underlying set is
partially ordered and the required object of interest is characterized by
maximality. To demonstrate how it is used in practit:e, let us prove what is
sometimes known as Hausdorff's Theorem (recall that by a chain is meant
a totally ordered set):
ZORN'S LEMMA 299

Theorem B-1. Every partially ordered set contains a maximal chain;

that is, a chain which is not a proper subset of any other chain.
Proof Consider the collection ~ of all chains of a partially ordered set
(S, :::;; ); ~ is nonempty, since it contains the chains consisting of single
elements of S. Partially order ~ by inclusion and let d be any chain of ~
(for the ordering £:;). We maintain that the union u d belongs to~. Given
elements a, b E U d, we have a E A E d and b E BEd, for some A, B.
As d is a chain, either A £:; B or B £:; A; suppose, for convenience, that
A £:; B. Then a, b both lie in B and, since B is itself a chain (in S), it follows
that a :::;; b or b :::;; a. In consequence, any two elements of u d are com-
parable, making u d a chain in S. Since u d is clearly an upper bound
for d in ~, Zorn's Lemma implies that (~, £:;) has a maximal member.
As another brief application of Zorn's Lemma, consider the following
assertion: if (S, :::;;) is a partially ordered set every chain of which has an
upper bound, then for each a E S there exists a maximal element XES with
the property that a :::;; x; in other words, there exists a maximal element
larger than the given element. For a proof of this, first observe that the set
J a = {y E Sia :::;; y} satisfies the hypotheses of Zorn's Lemma (under the
restriction of :::;;); hence, has a maximal element x. But x is maximal in S,
not merely in J a • For, suppose that S E S with x :::;; s. Then a :::;; s (since
both a :::;; x and x :::;; s) and so s E J a • From the maximality of x in J a , it
then follows that s = x, completing the argument.
Needless to say, we could just as well have phrased Zorn's Lemma in
terms of lower bounds and minimal elements. The assertion in this case is
that there exists at least one minimal element in S.
Before concluding, let us state
Theorem B-2. Zermelo's Well-Ordering Theorem, the Axiom of Choice
and Zorn's Lemma are all equivalent.
The deduction of these equivalences is somewhat involved and the
argument is not presented here; the interested reader can find the proofs
in any number of texts on set theory.
BIBLIOGRAPHY

GENERAL REFERENCES

Our purpose here is to present a list of suggestions for collateral reading and
further study. The specialized sources will carry the reader considerably
beyond the point attained in the final pages of this work.

1. ADAMSON, I., Introduction to Field Theory. New York: Interscience, 1964.

2. ARTIN, E., Galois Theory, 2nd Ed. Notre Dame, Ind. : University of Notre Dame Press,
1955.
3. ARTIN, E., C. NESBITT, and R. THRALL, Rings with Minimum Condition. Ann Arbor,
Mich. : University of Michigan Press, 1944.
4. AUSLANDER, M., Rings, Modules and Homology, Chapters I and II. Waltham, Mass.:
Department of Mathematics, Brandeis University (lecture notes), 1960.
5. BARNES, W., Introduction to Abstract Algebra. Boston: Heath, 1963.
6. BOURBAKI, N., Algebra, Chapter 8. Paris: Hermann, 1958.
7. BOURBAKI, N., Algebra Commutative, Chapters 2, 4 and 5. Paris: Hermann, 1961.
8. BURTON,D., Introtiuction to Modern Abstract Algebra. Reading, Mass: Addison-Wesley,
1967.
9. CURTIS, C. and I. REINER, Representation Theory of Finite Groups and Associative
Algebras. New York: Interscience, 1962.
10. DIVINSKY, N., Rings and Radicals. Toronto: University of Toronto Press, 1965.
II. GOLDIE, A., Rings with Maximum Condition. New Haven: Department of Mathe-
matics, Yale University (lecture notes), 1961.
12. HALMOS, P., Naive Set Theory. Princeton: Van Nostrand, 1960.
13. HERSTEIN, I. N., Topics in Algebra. New York: Blaisdell, 1964.
14. HERSTEIN, I. N., Theory of Rings. Chicago: Department of Mathematics, University
of Chicago (lecture notes), 1961.
15. HERSTEIN, I. N., Noncommutative Rings. (Carns Monographs). Menascha, Wis.;
Mathematical Association of America, 1968.
16. HEWITT, E. and K. STROMBERG, Real and Abstract Analysis. New York: Springer-
Verlag, 1965.
17. JACOBSON, N., Lectures in Abstract Algebra, Vol. I, Basic Concepts. Princeton: Van
Nostrand, 1951.
300
BIBLIOGRAPHY 301

18. JACOBSON, N., Structure of Rings, Rev. Ed. Providence: American Mathematical
Society, 1964.
19. JANS, J., Rings and Homology. New York: Holt, 1964.
20. KUROSH, A., General Algebra. New York: Chelsea, 1963.
21. LANG, S., Algebra. Reading, Mass.: Addison-Wesley, 1965.
22. LAMBEK, J., Lectures on Rings and Modules. Waltham, Mass.: Blaisdell, 1966.
23. MCCARTIlY, P., Algebraic Extensions of Fields. Waltham, Mass.: Blaisdell, 1966.
24. McCoy, N., Rings and Ideals. (Carus Monographs). Menascha, Wis.: Mathematical
Association of America, 1948.
25. McCoy, N., Theory of Rings. New York: Macmillan, 1964.
26. NAGATA, M., Local Rings. New York: Interscience, 1962.
27. NORTIlCOTT, D. G., Ideal Theory. Cambridge, England: Cambridge University Press,
1953.
28. NORTIlCOTT, D. G., Lessons on Rings, ModulesandMultiplicities. Cambridge, England:
Cambridge University Press, 1968.
29. REDEl, L., Algebra, Vo!.1. Oxford, England: Pergamon, 1967.
30. SAH, C.-H., Abstract Algebra. New York: Academic Press, 1967.
31. WARNER, S., Modern Algebra, 2 Vols. Englewood Cliffs, New Jersey: Prentice-Hall,
1965.
32. WEISS, E., Algebraic Number Theory. New York: McGraw-Hill, 1963.
33. ZARISKI, O. and P. SAMUEL, Commutative Algebra, Vo!' I. Princeton: Van Nostrand,
1958.

JOURNAL ARTICLES

34. BROWN, B. and N. McCoY, "Radicals and Subdirect Sums," Am. J. Math. 69, 46-58
(1947).
35. BUCK, R. C., "Extensions of Homorphisms and Regular Ideals," J. Indian Math. Soc.
14, 156-158 (1950).
36. COHEN, I. S., "Commutative Rings with Restricted Minimum Condition," Duke Math.
J. 17,27-42 (1950).
37. DIVINSKY, N., "Commutative Subdirectly Irreducible Rings," Proc. Am. Math Soc.
8,642-648 (1957).
38. FELLER, E., "A Type of Quasi-Frobenius Rings," Canad. Math. Bull. 10, 19-27 (1967).
39. GIFFEN, c., "Unique Factorization of Polynomials," Proc. Am. Math. Soc. 14, 366
(1963).
40. HENDERSON, D., "A Short Proof of Wedderburn's Theorem," Am. Math. Monthly 72,
385-386 (1965).
41. HERSTEIN, I. N., "A Generalization of a Theorem of Jacobson, I," Am. J. Math. 73,
756-762 (1951).
42. HERSTEIN, I. N., "An Elementary Proof of a Theorem of Jacobson," Duke Math. J.
21,45-48 (1954).
43. HERSTEIN, I. N., "Wedderburn's Theorem and a Theorem of Jacobson," Am. Math.
Monthly 68, 249-251 (1961).
44. JACOBSON, N., "The Radical and Semi-Simplicity for Arbitrary Rings," Am. J. Math.
67, 300-320 (1945).
45. KOHLS, C., "The Space of Prime Ideals of a Ring," Fund. Math. 45,17-27 (1957).
302 BIBLIOGRAPHY

46. KOVACS, L., "A Note on Regular Rings," Publ. Math. Debrecen 4, 465-468 (l956).
47. LUH, J., "On the Commutativity of J-Rings," Canad. J. Math. 19, 1289-1292 (l967).
48. McCoy, N., "Subdirectly Irreducible Commutative Rings," Duke Math. J.12, 381-387
(1945).
49. McCoy, N., "Subdirect Sums of Rings," Bull. Am. Math. Soc. 53, 856-877 (1947).
50. McCoy, N., "A Note on Finite Unions of Ideals and Subgroups," Proc. Am. Math.
Soc. 8, 633--637 (1957).
51. NAGATA, M., "On the Theory of Radicals in a Ring," J. Math. Soc. Japan 3,330-344
(l951).
52. VON, NEUMANN, J., "On Regular Rings," Proc. Natl. Acad. Sci. U.S. 22, 707-713 (1936).
53. NORTHCOTT, D., "A Note on the Intersection Theorem for Ideals," Proc. Cambridge
Phil. Soc. 48,.366-367 (1952).
54. PERLIS, S. "A Characterization of the Radical of an Algebra," Bull. Am. Math. Soc.
48, 128-132 (1942).
55. SAMUEL, P., "On Unique Factorization Domains," Illinois J. Math. 5, 1-17 (1961).
56. SATYANARAYANA, M., "Rings with Primary Ideals as Maximal Ideals," Math. Scand.
20, 52-54 (1967).
57. SATYANARAYANA, M., "Characterization of Local Rings," Tohoku Math. J.19, 411-416
(l967).
58. SNAPPER, E., "Completely Primary Rings, I," Ann. Math. 52, 666-693 (1950).
59. STONE, M. R., "The Theory of Representations of Boolean Algebras," Trans. Am.
Math. Soc. 40, 37-111 (l936).
INDEX OF SPECIAL SYMBOLS

The following is by no means a complete list of all the symbols used in the
text, but is rather a listing of certain symbols which occur frequently.
Numbers refer to the page where the symbol in question is first found.

{a} set consisting of the element a, 8

[a] congruence class determined by the element a, 4
a+I coset of the ideal I, 39
(a) smallest (two-sided) ideal containing the element a, 19
aR smallest right ideal of R containing the element a, 19
annS annihilator of the set S, 36
alb, atb a divides (does not divide) the element b, 90
a == b (mod n) integer a is a congruent to integer b modulo n, 4
alb formal fraction of elements a and b, 61
aob circle-product of elements a and b, 171
AAB symmetric difference of sets A and B, 3
AxB Cartesian product of sets A and B, 9
A(M) annihilator of the module M, 275
C field of complex numbers, 53
C(a) centralizer of the element a, 14
cent R center of the ring R, 9
char R characteristic of the ring R, 11
contf(x) content of the polynomialj(x), 129
degf(x) degree of the polynomial j(x), 119
f(A) direct image of the set A under J, 27
jl(A) inverse image of the set A under J, 27
F(a) field generated by the element a over F, 137
F[a] set of polynomials in the element a, 120
[F':F] degree of the field F' over the subfield F, 140
gcd(a,b) greatest common divisor of the elements a and b, 92
GF(p") Galois field with p" elements, 191
hom(R,R') set of ring homomorphisms from R into R', 26
303
304 INDEX OF SPECIAL SYMBOLS

homR(M,M') set of R-module homomorphisms from Minto M', 272

IJ product of the ideals 1 and J, 22
1+J sum of the ideals 1 and J, 21
1ffiJ internal direct sum of the ideals 1 and J, 21
(1:J) quotient of the ideal 1 by the ideal J, 23
'1:.1i sum of a set of ideals 1i' 21
Jl nil radical of the ideal 1, 79
J(R) J-radical of the ring R, 172
kerf kernel of the homomorphism j, 28
I(M) length of the module M, 252
lcm(a,b) least common multiple of the elements a and b, 94
Mn(R) ring of n x n matrices over R, 3
map(X,R) ring of mappings from X into R, 4
natI natural mapping determined by the ideal 1, 40
ordj{x) order of the power seriesj{x), 115
o the empty set, 3
cP(n) Euler phi-function, 57
cPr substitution homomorphism induced by the element r, 120
P(X) power set ofthe set X, 3
Q field of rational numbers, 2
Q(.jn) quadratic number field, 105
Qc)(R) classical ring of quotients of R, 60
R# field of real numbers, 2
R* set of invertible elements of R, 2
RV heart of the ring R, 212
R[x] polynomial ring in one indeterminant over R, 118
R[x,y] polynomial ring in two indeterminants over R, 134
R[[x]] power series ring in one indeterminant over R, 114
R/1 quotient ring of R by the ideal 1, 40
radR Jacobson radical of R, 157
RadR prime radical of R, 163
I" ffi Ri subdirect sum of a set of rings Rio 206
I ffi Ri complete direct sum of a set of rings R i , 204
Z,Ze ring of integers (even integers), 2, 9
Z+ set of positive integers, 12
Zl ring of multiples of the identity element, 12
Z(i) domain of Gaussian integers, 91
Zn ring of integers modulo n, 4
Z(.jn) a domain of quadratic integers, 106
+n' On addition (multiplication) modulo n, 5
is isomorphic to, 29
INDEX

additive group of a ring, comparable elements, 292

adeal, 38 component rings (in a direct sum), 204
adjunction (of an element to a field), 137 component projection, 206
algebraic element, 138 commutative diagram, 43
extension field, 140 commutative ring, 2
number field, 155 complete direct sum, 204
algebraically closed field, 156 composition series, 251
annihilator of a subset, 36 congruence modulo n, 4
Artinian ring, 223 congruence class, 4
ascending chain condition, 217 representation of, 4
associated elements, 91 conjugate of an element, 105
prime ideal of a primary ideal, 81 content of a polynomial, 129
prime ideal in a Noetherian ring, 236 Correspondence Theorem, 30
atom in a Boolean ring, 200 coset of an ideal, 39
automorphism, 25
axiom of choice, 297 degree, of an extension field, 140
of a polynomial, 119
derivative function, 153
Bezout identity, 93 descending chain condition, 223
binomial equation, 13 direct sum, complete, 204
Boolean ring, 14 discrete, 205
external, 33
cancellation law, 7 internal, 21
center of a ring, 9 of modules, 259
centralizer, of an element, 194 direct summand, 34
of a set of endomorphisms, 277 divides (divisor), 90
chain (in a partially ordered set), 292 division ring, 52
chain conditions, 217, 223 finite, 194
characteristic of a ring, 11 divisor of zero, 7
choice function, 297 domain, Euclidean, 102
classical ring of quotients, 60 integral, 7
coefficients of a power series, 114 principal ideal, 20
comaximal ideals, 211 unique factorization, 100
305
INDEX 306

element(s), fixed field, 69

algebraic, 138 formal fraction, 61
associate, 91 formal power series, 112
conjugate, 105 Frobenius automorphism, 202
idempotent, 14 Fundamental Homorphism Theorem, 44
identity, 2 Fundamental Theorem of Algebra, 128
invertible, 2
Galois field, 191
irreducible, 97 Gaussian integers, 91
nilpotent, 14
goo-property, 95
prime, 97
generators (of an ideal), 19
quasi-regular, 170
greatest common divisor, 92
related to an ideal, 258
group of invertible elements (of a ring), 2
relatively prime, 93
group ring, 15
transcendental, 138
torsion, 259 H-ring, 203
zero, I heart of a ring, 212
Eisenstein irreducibility criterion, 133 Hilbert ring, 178
endomorphism of a module, 272 homomorphism, 25
ofa ring, 25 evaluation, 26
evaluation homorphism, 26 kernel of, 28
equivalence class, 288 of modules, 250
relation, 287 of partially ordered sets, 293
Euler phi-function, 57 of rings, 25
Euclidean domain, 102 reduction, 130
valuation, 102 substitution, 120
extension, algebraic, 140 trivial, 26
simple, 137 homomorphic image, 25
extension ring, 31 associated prime, 236
comaximal, 211
faithful module, 275 commutator, 50
field, 52 finitely generated, 19
algebraically closed, 156 ideal, 16
extension, 136 irreducible, 235
Galois, 191 left (right), 16
obtained by adjoining an element, 137 maximal, 71
of algebraic numbers, 155 minimal, 86
of complex numbers, 53 minimal prime, 84
of quadratic numbers, 105 modular, 173
of rational functions, 138 modular maximal, 174
skew, 52 nil, 47
splitting, 148 nilpotent, 47
finite division ring, 194 primary, 81
integral domain, 56 prime, 76
field, 187 product of, 22
ring, 2 quotient, 23
finitely generated, 19 regular, 179
first element, 294 sum of, 20
307 INDEX

semiprime, 80 prime ideal of a ring, 84

minimum condition, 223
idempotent Boolean ring, 200 polynomial, 139
element, 14 modular ideal, 173
orthogonal, 268 module, 247
primitive, 270 annihilator of, 275
imbedding, 31 centralizer of, 272
induced partial order, 292 direct sum, 259
irreducible element, 97 dual, 286
ideal, 235 endomorphism of, 272
polynomial, 126 faithful, 275
irredundant primary representaion, 236 homomorphism of, 250
subdirect sum, 214 indecomposable, 260
isomorphism of modules, 250 isomorphism, 250
of partially ordered sets, 293 quotient, 249
of rings, 25 simple, 249
J-radical, 172 submodule, 249
J-ring, 196 torsion-free, 259
Jacobson radical, 157 monic polynomial, 119
multiplicatively closed set, 70
kernel of a homomorphism, 28 multiplicative semigroup of a ring,
last element, 294
natural mapping, 41
lern-property, 95
nil ideal, 47
least common multiple, 94
nilpotent element, 14
left annihilator, 36
ideal, 47
ideal, 16
nil radical of an ideal, 79
lemma,
ofa ring, 79
Fitting's, 256
nil-semisimple ring, 264
Gauss', 130
Noetherian ring, 219
Nakayama's, 243
nontrivial subring, 8
Schur's, 274
subdirect sum, 206
length of an element, 109
non-zero-divisor, 60
of a normal series, 252
norm, 105
of a module, 252
normal series, 251
lexicographic order, 296
lifting idempotents, 167
order homomorphism, 293
local ring, 88
order isomorphic, 293
localization, 88
order of a power series, 115
lower bound (for a partially ordered set),
Ore condition, 69
294
orthogonal idempotents, 268
maximal element, 293
ideal, 71 partial order, 291
maximum condition, 218 partition, 290
minimal element, 293 polynomial, 118
ideal, 86 content, 129
prime ideal of an ideal, 84 cyclotomic, 133
INDEX 308

degree of, 119 defined by a partition, 291

function, 153 equivalence, 287
in two indetenninants, 134 partial order, 291
irreducible, 126 reflexive, 287
leading coefficient, 119 symmetric, 287
minimum, 139 transitive, 287
monic, 119 relatively prime elements, 93
primitive, 129 Remainder Theorem, 123
root of, 121 refinement of a normal series, 251
primary component, 23f. ring, I
ideal, 81 Artinian, 223
representation, 236 Boolean, 14
ring, 169 commutative, 2
prime element, 97 divisible, 233
field, 65 division, 52
ideal, 76 finite, 2
radical, 163 H-, 203
primitive idempotent, 270 Hilbert, 178
ideal, 286 J-, 196
polynomial, 129 local, 88
ring, 278 nil-semisimple, 264
principal ideal, 19 Noetherian, 219
ideal ring, 20 of endomorphisms of a module, 272
proper subring, 8 of extended power series over R, 152
pseudo-inverse, 25 of formal power series over R, 113
of functions between a set and ring, 4
quadratic number field, 105 of integers modulo n, 5
quasi-inverse, 170 of polynomials over R, 118
quasi-regular, 170 of matrices over R, 3
quaternions, 54 primary, 169
quotient ideal, 23 quotient, 40
field of, 60 regular, 24
module, 249 right Artinian, 262
ring, 40 semisimple, 157
simple, 18
radical, J-, 172 subdirectly irreducible, 211
Jacobson, 157 with identity, 2
nil, 79 without radical, 157, 163
prime, 163 zero, 13
rational function, 138 ring of quotients, classical, 60
reduction homomorphism, 130 generalized, 70
regular ring, 24 relative to a set, 70
relation (binary), 287 root of a polynomial, 121
antisymmetric, 291 multiple, 153
associated with a function, 288
compatible equivalence, 49 saturated, 178
congruence modulo n, 4 semisimple ring, 157
309 INDEX

series, composition, 251 Hilbert Basis, 220

equivalent, 252 Jordan-Holder, 252
normal, 251 Jacobson's. 198
length of, 252 Jacobson Density, 285
refinement of, 251 Kronecker's, 144
simple extension field, 137 Krull Intersection, 244
simple module, 249 Krull-Zorn, 74
ring, 18 Levitski's, 222
skew field, 52 McCoy's, 212
splitting field, 148 Noether's, 236
square-free integer, 105 Stone Representation, 183
subdirectiy irreducible, 211 Wilson's, 188
subdirect sum, 206 Wedderburn's, 194, 266, 230
subfield, 59 Wedderburn-Artin, 281
submodule, 249 Zermelo's, 297
subring, 8 total ordering, 292
generated by a set, 14 torsion element, 259
proper, 8 torsion-free module, 259
trivial, 8 transcendental element (over a field), 138
substitution homomorphism, 120 trivial subring, 8
in a polynomial, 120 homomorphism, 26
symmetric difference, 3
unique factorization domain, 100
Theorem, upper bound (for a partially ordered set),
Akizuki-Hopkins, 255 294
Birkhoff's, 212
Brauer's, 262 valuation ring, 88
Chinese Remainder, 211
Cohen's, 241 well-ordered set, 296
Dorroh Extension, 31 without radical, 157, 163
Euler-Fermat, 58
Euclid's, 101 zero divisor, 7
Fermat's little, 68 zero element of a ring,
Hausdorff's, 299 zero ring, 13
Herstein's, 199 Zorn's lemma, 298

ABCDE79876S43210