Beruflich Dokumente
Kultur Dokumente
;"
. DAVID
M.~BURTON
.'
'Y 'Y
A First Course in
;"
. DAVID
M.~BURTON
.'
'Y 'Y
.~
'{
I
1
'1
'.1
l
1
i
.~
'{
I
1
'1
'.1
l
1
i
----------------------~~
----,------------ .----_.-
..
_.-
----
----:-----.---------- --"----- _ . _ - - - - _ . ,
PREFACE
:1'
.' ".
~'
el
.~
----------------------~~
----,------------ .----_.-
..
_.-
----
----:-----.---------- --"----- _ . _ - - - - _ . ,
PREFACE
:1'
.' ".
~'
el
.~
vi
PREFACE
C;rl: 4 585
of topics not treated in the body of the text, as well as impart additional
information about material covered earlier; sorne, especially in the later
chapters, provide substantial extensions of the theory. We have, on the
whole, resisted the temptation to use the exercises to develop results that
will subsequently be needed (although this is not hard and fast). Those
problems whose solutions do not appear straightforward are often accompanied by hints.
The text is not intended to be encyc10pedic in nature; many fascinating
aspects of this subject vie for inc1usion and sorne choice is imperative. To
this end, we merely followed our own tastes, condensing or omitting entirely
a number of topics that might have been encompassed by a book of the
same tltle. Despite sorne notable omissions, the coverage should provide
a firm foundation on which to build.
A great deal of valuable criticism was received in the preparation of this
work and ourmoments of complacence have admitted many improvements.
Of those students who helped, consciously or otherwise, we should like
particularly to mention Francis Chevarley, Delmon Grapes, Cynthia
Kennett, Kenneth Lidman, Roy Morell, Brenda Phalen, David Smith, and
John Sundstrom; we valued their critical reading of sections of the manuscript and incorporated a number of their suggestions into the texto It is a
pleasure, likewise, to record our indebtedness to Professor James Clay of
the University of Arizoria, who reviewed the final draft and offered helpful
comments leading to its correction and improvement. We also profited
from many conversations with our colleagues at the University of New
Hampshire, especial1y Professors Edward Batho, Homer Bechtell, Robb
Jacoby, and Richard Johnson. In this regard, special thanks are due Professor William Witthft, who was kind enough to read portions of the
galleys; his eagle-eyed attention saved us from embarrassment more than
once. We enjoyed the'luxury of unusually good secretarial help and take
this occasion to express our appreciation to Nancy Buchanan and Sola'nge
Larochelle for their joint labors on the typescript. To my wife must go tbe
largest .debt of gratitud e, not only for generous assistance with the text at
all stages of development, but for her patience and understanding on those
occasions when nOtlling would go as we wished.
Finally, we should like to acknowledge the fine cooperation of the staff
of Addison-Wesley and the usual high quality of their work. The author,
needless tq say, must accept the full responsibility for any shortcomings and
errors which remain.
Durham, New Hampshire
J anuary 1970
o'
7
I
CONTENTS
Chapter 1 Introductory Concepts .
Chapter 2 Ideals and Their Operations
16
39
52
71
90
112 ...
157
180
Chapter lQ
204
Chapter 11
217
234
262
287
AppendixA. Relations.
AppendixB.
Zorn's Lernma
296
300
, Bibliography
Index oC Special Symbols
303
Index .
305
D.M.B.
vii
vi
PREFACE
C;rl: 4 585
of topics not treated in the body of the text, as well as impart additional
information about material covered earlier; sorne, especially in the later
chapters, provide substantial extensions of the theory. We have, on the
whole, resisted the temptation to use the exercises to develop results that
will subsequently be needed (although this is not hard and fast). Those
problems whose solutions do not appear straightforward are often accompanied by hints.
The text is not intended to be encyc10pedic in nature; many fascinating
aspects of this subject vie for inc1usion and sorne choice is imperative. To
this end, we merely followed our own tastes, condensing or omitting entirely
a number of topics that might have been encompassed by a book of the
same tltle. Despite sorne notable omissions, the coverage should provide
a firm foundation on which to build.
A great deal of valuable criticism was received in the preparation of this
work and ourmoments of complacence have admitted many improvements.
Of those students who helped, consciously or otherwise, we should like
particularly to mention Francis Chevarley, Delmon Grapes, Cynthia
Kennett, Kenneth Lidman, Roy Morell, Brenda Phalen, David Smith, and
John Sundstrom; we valued their critical reading of sections of the manuscript and incorporated a number of their suggestions into the texto It is a
pleasure, likewise, to record our indebtedness to Professor James Clay of
the University of Arizoria, who reviewed the final draft and offered helpful
comments leading to its correction and improvement. We also profited
from many conversations with our colleagues at the University of New
Hampshire, especial1y Professors Edward Batho, Homer Bechtell, Robb
Jacoby, and Richard Johnson. In this regard, special thanks are due Professor William Witthft, who was kind enough to read portions of the
galleys; his eagle-eyed attention saved us from embarrassment more than
once. We enjoyed the'luxury of unusually good secretarial help and take
this occasion to express our appreciation to Nancy Buchanan and Sola'nge
Larochelle for their joint labors on the typescript. To my wife must go tbe
largest .debt of gratitud e, not only for generous assistance with the text at
all stages of development, but for her patience and understanding on those
occasions when nOtlling would go as we wished.
Finally, we should like to acknowledge the fine cooperation of the staff
of Addison-Wesley and the usual high quality of their work. The author,
needless tq say, must accept the full responsibility for any shortcomings and
errors which remain.
Durham, New Hampshire
J anuary 1970
o'
7
I
CONTENTS
Chapter 1 Introductory Concepts .
Chapter 2 Ideals and Their Operations
16
39
52
71
90
112 ...
157
180
Chapter lQ
204
Chapter 11
217
234
262
287
AppendixA. Relations.
AppendixB.
Zorn's Lernma
296
300
, Bibliography
Index oC Special Symbols
303
Index .
305
D.M.B.
vii
.ONE
INTRODUCTORY CONCEPTS
.,'
CONVENTIONS
Rere we sha11 set forth certain conventions in notatio'i(,~nd terminology
used throughout. the text: the standard sym bols of se" theory will be
e~ployed, namely, E, u, n, -, and 0 for the empty set. In particular,
A - B = {xix E A and x! B}. As regards inclusion, the symbols ~ ~~d
;;2 mean ordinary inclusion between sets (they do not exclude the posslbIllty
of equality), whereas e and ::J indicate proper inclusion. When we ~eal
with an indexed co11ection of sets, say {Ali E IJ, the cumbersome notatlOns
U {AliEI} and n {AliEI} will genera11y. be ~bbreviated to. u A and
n A; it being uIiderstood that the operabons are always over the fu11
domain on which the index is defined. Fo11owing custom, {a} denotes the
set whose only member is a. Provided that there is no dsk of confusion, a
one-element set will be identified with the element itself.
.
A function f (synonymous with mapping) is indicated by a strai?ht
arrow going from domain to range, as in the case f: X .-+ Y, and the no.tatl~n
always signifies thatfhas domain X. Under these cl~cumstan.ces,fls sald
to be a function on X, or from X, into. Y. In representmg functlOnal values,
we adopt the convention of writing the .function on the left, so that/~x), or
occasiona11y fx, denotes the image of an element x E X. The restnctlOn of
f to a subset A of X is the function flA from.1 into Y defined. by
(fIA)(x) = f(x) for a11 x in A: F~r the compo.sltlOn of two func~lOns
f: X -+ Yand g: Y -+ Z, we wIlI wnte g o f; that lS, g o f: X ~ Z .satlsfies
(g o f)(x) = g(J(x)) for each x E X. (It is important to bear m mmd that
our policy is to apply the functions from right to left.)
Sorne 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.
Fina11y, let us reserve the symbol Zfor th~ set of all integers, Z~ for the
set of positive integers, Q fo! the set of ratlOnal numbers, and R for the
set of real numbers.
viii
The present chapter sets the stag'~for much that fo11ows, by reviewing sorne
of the basic elements of ring theory: I t al so serves as an appropriate vehicle
for codifying certain notation and, technical vocabulary used throughout
the text . With an eye to the b~,r~ning student (as well as .to minimize a
sense of vagueness), we have also'l!1duded a ~umbel of pertinenrexamples
of rings. The mathematica11y matre reader who finds thepace'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 inten!st in this book, the notion of a ringo
A ring is an ordered triple (R, +,.) consisting of a
nonempty set R and two binary operations + and . defined on R such
that
Definition 1-1.
1)
+.
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 ri.p.g and of (R, .) as the multiplicative semigroup of the. ringo
By analogy with the integers, the unique identity element for addition
is caBed 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
befo re addition. Accordingly, the expression a'b + e stand s for (a'b) + e
and not for a'(b + e). Because ofthe general associative law, parentheses
.ONE
INTRODUCTORY CONCEPTS
.,'
CONVENTIONS
Rere we sha11 set forth certain conventions in notatio'i(,~nd terminology
used throughout. the text: the standard sym bols of se" theory will be
e~ployed, namely, E, u, n, -, and 0 for the empty set. In particular,
A - B = {xix E A and x! B}. As regards inclusion, the symbols ~ ~~d
;;2 mean ordinary inclusion between sets (they do not exclude the posslbIllty
of equality), whereas e and ::J indicate proper inclusion. When we ~eal
with an indexed co11ection of sets, say {Ali E IJ, the cumbersome notatlOns
U {AliEI} and n {AliEI} will genera11y. be ~bbreviated to. u A and
n A; it being uIiderstood that the operabons are always over the fu11
domain on which the index is defined. Fo11owing custom, {a} denotes the
set whose only member is a. Provided that there is no dsk of confusion, a
one-element set will be identified with the element itself.
.
A function f (synonymous with mapping) is indicated by a strai?ht
arrow going from domain to range, as in the case f: X .-+ Y, and the no.tatl~n
always signifies thatfhas domain X. Under these cl~cumstan.ces,fls sald
to be a function on X, or from X, into. Y. In representmg functlOnal values,
we adopt the convention of writing the .function on the left, so that/~x), or
occasiona11y fx, denotes the image of an element x E X. The restnctlOn of
f to a subset A of X is the function flA from.1 into Y defined. by
(fIA)(x) = f(x) for a11 x in A: F~r the compo.sltlOn of two func~lOns
f: X -+ Yand g: Y -+ Z, we wIlI wnte g o f; that lS, g o f: X ~ Z .satlsfies
(g o f)(x) = g(J(x)) for each x E X. (It is important to bear m mmd that
our policy is to apply the functions from right to left.)
Sorne 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.
Fina11y, let us reserve the symbol Zfor th~ set of all integers, Z~ for the
set of positive integers, Q fo! the set of ratlOnal numbers, and R for the
set of real numbers.
viii
The present chapter sets the stag'~for much that fo11ows, by reviewing sorne
of the basic elements of ring theory: I t al so serves as an appropriate vehicle
for codifying certain notation and, technical vocabulary used throughout
the text . With an eye to the b~,r~ning student (as well as .to minimize a
sense of vagueness), we have also'l!1duded a ~umbel of pertinenrexamples
of rings. The mathematica11y matre reader who finds thepace'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 inten!st in this book, the notion of a ringo
A ring is an ordered triple (R, +,.) consisting of a
nonempty set R and two binary operations + and . defined on R such
that
Definition 1-1.
1)
+.
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 ri.p.g and of (R, .) as the multiplicative semigroup of the. ringo
By analogy with the integers, the unique identity element for addition
is caBed 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
befo re addition. Accordingly, the expression a'b + e stand s for (a'b) + e
and not for a'(b + e). Because ofthe general associative law, parentheses
INTRODUCTOR Y CONCEPTS
Example 1-2 Let X be a. given set and P(X) be the collection oC all subsets
of X. The symmetric difference of two subsets A, B ~ X is the set A I:l B,
where
A I:l B =. (A
B) u (B - A).
can also be otntted 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 ringo A ring (R, +, . ) consists of a nonempty set R together with two
binary operations + and . of addition and multiplcation on R for which
the following conditions are satisfied :
1)
2)
3)
4)
5)
6)
+ b = b + a,
(a +' b) + C = a + (b +
c),
(Q,+,'),
A I:l B,
A B = A n B,
then the system (P(X), +, .) forros a commutative ring with identity. The
empty set0 serves as the zero element, whereas the multiplicative identity
is X. Furthermore, each set in P(X) is its own additive in verse. It is
interesting to note that if X is nonempty, then neither (P(X), u, n) nor
(P(X), n, u) constitutes a ringo
.
Example 1-3. Given a ring (R, +, '), we may consider the set M,,(R) oC
n x n matrices over R. If 1"
{1,2, .. " n}, a typical member oC M,,(R)
is a function 1: In X 1" --, R. In practice, one identifies such a Cunction
with its values aij I(,}), which are displayed as the n x n rectangular
array
~: 11
(
...
~: 1" )
(aij E R).
Il"l , .. a""
L" aik'b'j'
k=l
(R#,+,')
are all examples of rings (here, + and . are taken to be ordinary addition
and multiplication). In each oC these cases, the ring is commutative and
has the integer 1 for an identity elemento
oij
J1
iC i
1,0 ifi
= j
=1=
INTRODUCTOR Y CONCEPTS
Example 1-2 Let X be a. given set and P(X) be the collection oC all subsets
of X. The symmetric difference of two subsets A, B ~ X is the set A I:l B,
where
A I:l B =. (A
B) u (B - A).
can also be otntted 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 ringo A ring (R, +, . ) consists of a nonempty set R together with two
binary operations + and . of addition and multiplcation on R for which
the following conditions are satisfied :
1)
2)
3)
4)
5)
6)
+ b = b + a,
(a +' b) + C = a + (b +
c),
(Q,+,'),
A I:l B,
A B = A n B,
then the system (P(X), +, .) forros a commutative ring with identity. The
empty set0 serves as the zero element, whereas the multiplicative identity
is X. Furthermore, each set in P(X) is its own additive in verse. It is
interesting to note that if X is nonempty, then neither (P(X), u, n) nor
(P(X), n, u) constitutes a ringo
.
Example 1-3. Given a ring (R, +, '), we may consider the set M,,(R) oC
n x n matrices over R. If 1"
{1,2, .. " n}, a typical member oC M,,(R)
is a function 1: In X 1" --, R. In practice, one identifies such a Cunction
with its values aij I(,}), which are displayed as the n x n rectangular
array
~: 11
(
...
~: 1" )
(aij E R).
Il"l , .. a""
L" aik'b'j'
k=l
(R#,+,')
are all examples of rings (here, + and . are taken to be ordinary addition
and multiplication). In each oC these cases, the ring is commutative and
has the integer 1 for an identity elemento
oij
J1
iC i
1,0 ifi
= j
=1=
FIRST
COURSE IN RINGS AND IDEALS
,
INTRODUCTORY CONCEPTS
t ,
Example 1-4. To develop our next example, let X be an arbitrary (non- '
enipty) set and(R, +, .) be a ringo We adoptthe notation map(X, R} for
the set consisting f a11 mappings from ,X into R; in symbols,
g)(X) = f(x)
+ g(x),
(f'g)(x) = f(x)'g(x),(x
,1
{a + knlkeZ}.
Of course, the same congrtience class may very well arise from another
integer; any integer a' Cor which [a/J = [a J is said to be a representative
of [a]. Qne final, purely notational, remark : the collection of a11 congruence
classes oCintegers modulo n will be designated by Zn'
It can be shown tbat the congruence cIasses [OJ, [lJ, ... , [n - 1J
exhaust the elements of Z.. Given 'an arbitrary integer a, the division
algorithm asserts tbat there exist uniqueq, re Z, with O :s;; r < n, such
that a = qn + r. By the definition of congruence, a == r (rnod n), or
{'
- 1J}.
its
Zn
X).
Example 1-5.. Our final example i8 that of the ring oC integers modulo n,
wbere n is a fixed positive integer. In order to describe tbis system, we
first introduce th notion of congruence: two integers a and b are said to
be eongruent 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 nd 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 c1asses of congruent
e1ements, caBed eongruenee classes. For each integer a, let the congnence
class to which a belongs be denoted by [aJ:
[aJ +n [bJ
[a
+ bJ,
. [ab].
:.... fr.
.\.
a'b
= (a +
kn)(b + jn)
ab
+ (aj + bk + kjn)n .
Hence, a'b' == ab (mod n) and so [a'b'] = [abJ, as desired. The proof that
addition is unambiguously defined proceeds similarIy.
We omit the detailed verification of the faet tbat (Z., +n' 'n) is a commutatve ring with identity (tradtioIiaIly known as the ring of integers
moaulo n), remarking only that the various ring axiom8 hold in Zn simply
because they hold in Z. The distributive law, for instance, fo11ows in Zn
from its validity in Z:
'
[aL([bJ
+" [eJ)
= [aL[b
[ab + aeJ
eJ = [a(b
+ e)]
[abJ +n [aeJ
= [aL[bJ +11 [aL [e].
=
Notice, too, that the congruence classes [OJ and [lJ serve as the zero element
and multiplieative identity, respectively, whereas [-aJ i8 the additive
in verse of [aJ in Zn' When no eonfusion is likely, we sha1l1eave off the
brackets from the elements of Zn, thereby making no genuine distnctiori
FIRST
COURSE IN RINGS AND IDEALS
,
INTRODUCTORY CONCEPTS
t ,
Example 1-4. To develop our next example, let X be an arbitrary (non- '
enipty) set and(R, +, .) be a ringo We adoptthe notation map(X, R} for
the set consisting f a11 mappings from ,X into R; in symbols,
g)(X) = f(x)
+ g(x),
(f'g)(x) = f(x)'g(x),(x
,1
{a + knlkeZ}.
Of course, the same congrtience class may very well arise from another
integer; any integer a' Cor which [a/J = [a J is said to be a representative
of [a]. Qne final, purely notational, remark : the collection of a11 congruence
classes oCintegers modulo n will be designated by Zn'
It can be shown tbat the congruence cIasses [OJ, [lJ, ... , [n - 1J
exhaust the elements of Z.. Given 'an arbitrary integer a, the division
algorithm asserts tbat there exist uniqueq, re Z, with O :s;; r < n, such
that a = qn + r. By the definition of congruence, a == r (rnod n), or
{'
- 1J}.
its
Zn
X).
Example 1-5.. Our final example i8 that of the ring oC integers modulo n,
wbere n is a fixed positive integer. In order to describe tbis system, we
first introduce th notion of congruence: two integers a and b are said to
be eongruent 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 nd 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 c1asses of congruent
e1ements, caBed eongruenee classes. For each integer a, let the congnence
class to which a belongs be denoted by [aJ:
[aJ +n [bJ
[a
+ bJ,
. [ab].
:.... fr.
.\.
a'b
= (a +
kn)(b + jn)
ab
+ (aj + bk + kjn)n .
Hence, a'b' == ab (mod n) and so [a'b'] = [abJ, as desired. The proof that
addition is unambiguously defined proceeds similarIy.
We omit the detailed verification of the faet tbat (Z., +n' 'n) is a commutatve ring with identity (tradtioIiaIly known as the ring of integers
moaulo n), remarking only that the various ring axiom8 hold in Zn simply
because they hold in Z. The distributive law, for instance, fo11ows in Zn
from its validity in Z:
'
[aL([bJ
+" [eJ)
= [aL[b
[ab + aeJ
eJ = [a(b
+ e)]
[abJ +n [aeJ
= [aL[bJ +11 [aL [e].
=
Notice, too, that the congruence classes [OJ and [lJ serve as the zero element
and multiplieative identity, respectively, whereas [-aJ i8 the additive
in verse of [aJ in Zn' When no eonfusion is likely, we sha1l1eave off the
brackets from the elements of Zn, thereby making no genuine distnctiori
INTRODUCTOR Y CONCEPTS
Proof. These turn out, in,t.he m~in, to be simple consequences of the distributive laws. For instari2e, Irom O + O = O, it follows that
Oa=;(O
+ O)a =
Oa
Da.
Thus, by the cancellation law:ror the additive group (R, +), we have Oa = O.
In a like manner, one obtains aO = O. The proof of (2) requires the fact
that each element of R h~SI a unique additive inverse (Problem 1). Since
b + (-b) = O,
'"
ab
a("':'~) = a(b
(-b)) = aO = O,
which then implies that -(ab) = a( -b). The argument that (-a)b is also
the additive inverse of ab proceeds similarly. Tbis leads immediately'to (3):
abo
Proof. Since R =1= {O}, there exists sorne nonzero element a E R. If Oand
1 were equal, it would follow that a = al.. = aO = O, an obvious contradiction.
CONVENTION: Let us assume, once and for all, that any ring with identity
contains more than oneelement. This will rule out the possibility that O
and 1 coincide.
We now make several remarksabout the concept of zero divisors (the
term "divisors of zero" is also in common use): .
Defution 1-3. If R is a ring and O =1= a E R, then a is called a left
(right) zero divisor in R if there exists sorne b =1= O in R such that
ab = O (ba = O). A zero divisor is any element of R that is either a
left or right zero divisor.
According to this definition, O 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 riIig
with zero divisors is Z., where the integer n > 1 is composite; if n = n 1 n 2
in Z (O < n 1 , n 2 < n), then the product n 1 .n 2 = O in Z .
For the most part, we shall be studying rings without zero divisors.
In such rings it is possible to conc1ude from the equation ab = O that
either a = O 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, e E R,
ab = ae and ba = ca, where a =1= O, implies b = e.
Proof. Suppose that R is without zero divisors and let ab = ae, a =1= O.
Then, the product a(b - e) = O, which means that b - e = O and b = e.
The argument is the same for the equation ba = ca. Conversely, let R
satisfy the cancellation laws and assume that ab = O, with a =1= O. We then'
have ab = aO, whence by cancellation b = O. Similarly, b =1= O implies
a = O, 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 ofan integral domain
is the ring ofintegers; hence the choice ofterminology. 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
INTRODUCTOR Y CONCEPTS
Proof. These turn out, in,t.he m~in, to be simple consequences of the distributive laws. For instari2e, Irom O + O = O, it follows that
Oa=;(O
+ O)a =
Oa
Da.
Thus, by the cancellation law:ror the additive group (R, +), we have Oa = O.
In a like manner, one obtains aO = O. The proof of (2) requires the fact
that each element of R h~SI a unique additive inverse (Problem 1). Since
b + (-b) = O,
'"
ab
a("':'~) = a(b
(-b)) = aO = O,
which then implies that -(ab) = a( -b). The argument that (-a)b is also
the additive inverse of ab proceeds similarly. Tbis leads immediately'to (3):
abo
Proof. Since R =1= {O}, there exists sorne nonzero element a E R. If Oand
1 were equal, it would follow that a = al.. = aO = O, an obvious contradiction.
CONVENTION: Let us assume, once and for all, that any ring with identity
contains more than oneelement. This will rule out the possibility that O
and 1 coincide.
We now make several remarksabout the concept of zero divisors (the
term "divisors of zero" is also in common use): .
Defution 1-3. If R is a ring and O =1= a E R, then a is called a left
(right) zero divisor in R if there exists sorne b =1= O in R such that
ab = O (ba = O). A zero divisor is any element of R that is either a
left or right zero divisor.
According to this definition, O 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 riIig
with zero divisors is Z., where the integer n > 1 is composite; if n = n 1 n 2
in Z (O < n 1 , n 2 < n), then the product n 1 .n 2 = O in Z .
For the most part, we shall be studying rings without zero divisors.
In such rings it is possible to conc1ude from the equation ab = O that
either a = O 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, e E R,
ab = ae and ba = ca, where a =1= O, implies b = e.
Proof. Suppose that R is without zero divisors and let ab = ae, a =1= O.
Then, the product a(b - e) = O, which means that b - e = O and b = e.
The argument is the same for the equation ba = ca. Conversely, let R
satisfy the cancellation laws and assume that ab = O, with a =1= O. We then'
have ab = aO, whence by cancellation b = O. Similarly, b =1= O implies
a = O, 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 ofan integral domain
is the ring ofintegers; hence the choice ofterminology. 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 zeto divisors.
We change direction somewhat to deal with the situation where a subset
of a ring again constitutes a ringo Formally speaking,
INTRODUCTORY CONCEPTS
This odefinition is adequate, but unwieldy, siIice all the aspects of the
definition of a ring must be checked in deteimining whether a given subset
is a subring. In seeking a simpler criterion, noticethat (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 automaticaJly 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 cuefrom these remarks, a subring could just as well be
defined as follows. The system (S, +, o) forms a subring of the ring (R, +, .)
if and only if
.
To add th~ 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, bES. By these
observations we are led toa set of c10sure conditions wruch make it somewhat easier to verify tbat a particular subset is actually a subring.
Theorem 1-3. Let R be a ring and 0 =1= S 5; R. Theri, S i8 a subring
of R if and only if
1) a, b E S imply a - b E S .(closure under differences),
(closure under multplication).
2) a, b E S imply ab E S
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.
2(2nm)
Z.,
Z .
Prior to stating our next theorem, let us define the center of a ring R,
denoted by cent R, to be th set
1'"
cent R=' {a
RJar
ar - br = ra - rb
r(a - b),
1) Sorne subfing has a multiplitative identity, but the entire ring does noto
2) Boththe ring and one ofits 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 ringo To justify this claim, let l' =1= O denote the
identity element of the subririg S; we assume further that l' does not act as
an identity for the whole ring R. Accordingly, there exists sorne element
a E R for which al'. =1= a. It is dear that
(al')l'
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.
- m)
1) S is a nonempty subset of R,
2) (S, +) is a subgroup of(R, +), and
3) the set S is closed under multiplication.
(2n) (2m)
= 2(n
2n -:- 2m
o
ev~n
a(l'l')
al',
or (al' - a)l' = O. Since rieither al' -' a nor l' is zero, the ring R has
zero divisors, and in particular l' is a zerp divisor.
Example 1-8. to present a simple illustration of a ring in which the seoond .
of the aforementioned possibilities occurs, consider the set R = Z x Z,
8.
in this case the term "integral domain" would merely indicate a commutative
ring without zeto divisors.
We change direction somewhat to deal with the situation where a subset
of a ring again constitutes a ringo Formally speaking,
INTRODUCTORY CONCEPTS
This odefinition is adequate, but unwieldy, siIice all the aspects of the
definition of a ring must be checked in deteimining whether a given subset
is a subring. In seeking a simpler criterion, noticethat (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 automaticaJly 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 cuefrom these remarks, a subring could just as well be
defined as follows. The system (S, +, o) forms a subring of the ring (R, +, .)
if and only if
.
To add th~ 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, bES. By these
observations we are led toa set of c10sure conditions wruch make it somewhat easier to verify tbat a particular subset is actually a subring.
Theorem 1-3. Let R be a ring and 0 =1= S 5; R. Theri, S i8 a subring
of R if and only if
1) a, b E S imply a - b E S .(closure under differences),
(closure under multplication).
2) a, b E S imply ab E S
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.
2(2nm)
Z.,
Z .
Prior to stating our next theorem, let us define the center of a ring R,
denoted by cent R, to be th set
1'"
cent R=' {a
RJar
ar - br = ra - rb
r(a - b),
1) Sorne subfing has a multiplitative identity, but the entire ring does noto
2) Boththe ring and one ofits 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 ringo To justify this claim, let l' =1= O denote the
identity element of the subririg S; we assume further that l' does not act as
an identity for the whole ring R. Accordingly, there exists sorne element
a E R for which al'. =1= a. It is dear that
(al')l'
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.
- m)
1) S is a nonempty subset of R,
2) (S, +) is a subgroup of(R, +), and
3) the set S is closed under multiplication.
(2n) (2m)
= 2(n
2n -:- 2m
o
ev~n
a(l'l')
al',
or (al' - a)l' = O. Since rieither al' -' a nor l' is zero, the ring R has
zero divisors, and in particular l' is a zerp divisor.
Example 1-8. to present a simple illustration of a ring in which the seoond .
of the aforementioned possibilities occurs, consider the set R = Z x Z,
10
INTROOUCTORY CONCEPTS
(e, d) = (a
+ e, b +
d),
A routne calculation will show that Z x {O} = {(a, O)la E Z} forms a subring with .identty element (1, O). This obviously differs from the identity
of the entire ring R, which turns out to be the ordered pair (1, 1). By our
prevous rernarks, (1, O)rnustbeazerodivisorinR;infact,(l, 0)(0,1) = (O, O),
where (O, O) serves as the zero element of R.
If R is an arbitrary ring and n a positive integer, then the nth power a"
of an element a E R is defined by the inductve condtions al = a and
a" = a"-la. Frorn tbis the usuallaws of exponents follow at once:
a"a m = an+ m, (a"t = a"m
(n, m E Z+).
+ a, when n >
1.
(n
+ m)a
na
+ ma,
(nm)a = n(ma),
n(a
+ b)
= /la
+ nb,
and
(na)(mb) = (nm)(ab).
11
rp
= m(la) = (m1)a = Oa
10
INTROOUCTORY CONCEPTS
(e, d) = (a
+ e, b +
d),
A routne calculation will show that Z x {O} = {(a, O)la E Z} forms a subring with .identty element (1, O). This obviously differs from the identity
of the entire ring R, which turns out to be the ordered pair (1, 1). By our
prevous rernarks, (1, O)rnustbeazerodivisorinR;infact,(l, 0)(0,1) = (O, O),
where (O, O) serves as the zero element of R.
If R is an arbitrary ring and n a positive integer, then the nth power a"
of an element a E R is defined by the inductve condtions al = a and
a" = a"-la. Frorn tbis the usuallaws of exponents follow at once:
a"a m = an+ m, (a"t = a"m
(n, m E Z+).
+ a, when n >
1.
(n
+ m)a
na
+ ma,
(nm)a = n(ma),
n(a
+ b)
= /la
+ nb,
and
(na)(mb) = (nm)(ab).
11
rp
= m(la) = (m1)a = Oa
12
PROBLEMS
a
Oofthe group (R, +) to have order m mean s that ma = Oand ka
O
irO < k < m.) But the retation O == ma = (ml)a mplies that mI = O, Cor
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 argumen't can be employed when char R
O.
The equaqon ma = O would lead, as before, to m1 = O or m
O. In this
case every nonzero eIement a E R mus be oC infinite order.
The last result seives to bring
on record as .
.
OUt
;',.:PROBLEMS
Continuing in thls veln, let us next show tbat not any commutative
group can serve as the additive group of an integral domain.
::.' ). Verify that the zero elernent of a ring R is unique, as s the additive inverse of each.:
element a E R.
.'
3. Prove that any ring R in which the two operations are equal (that is, a
for all a, b e R) must be the trivial ring R = {O}.
.
b = ab
5. 'a) Ifthe set X eontains more than one elernent, prove that every nonempty proper
subset of X is a zer divisor in the ring P(X).
b) Show that, if n > 1, the matrix ring Mn(R) has zero divison even though the
'_
ring R may not.
6. Suppose !hat R is a ring with identity 1'itnd having no divisor s ofzero. For a, bE R,
verify that
a) ab
1 if and only if ha = 1,
b) if a2
1, then either a 1 or a = -1.
= {nlln E Z}.
Zl
13
(nl)(ml) = (nm)l
binomial expansion
one can easily infer that ZI itself [orms a (commutative) ring with identity.
The order of the additive cyclic group (Z1, +) is simply the characteristic
oC the given ring R.
When R happens to be an integral domain, then Zl is a subdomain of
R (that s, Z1 is also an integral domain with respect to the operations in
R). In fact, ZI is the smallest subdomain oC R, in the sense that it is contained in every other subdomain of R. If R is a domain of characteristic p,
(a
+ bY'
an + ('i.)an-1b
(n~l)Qb-l
+ b",
where
k)!
..
12
PROBLEMS
a
Oofthe group (R, +) to have order m mean s that ma = Oand ka
O
irO < k < m.) But the retation O == ma = (ml)a mplies that mI = O, Cor
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 argumen't can be employed when char R
O.
The equaqon ma = O would lead, as before, to m1 = O or m
O. In this
case every nonzero eIement a E R mus be oC infinite order.
The last result seives to bring
on record as .
.
OUt
;',.:PROBLEMS
Continuing in thls veln, let us next show tbat not any commutative
group can serve as the additive group of an integral domain.
::.' ). Verify that the zero elernent of a ring R is unique, as s the additive inverse of each.:
element a E R.
.'
3. Prove that any ring R in which the two operations are equal (that is, a
for all a, b e R) must be the trivial ring R = {O}.
.
b = ab
5. 'a) Ifthe set X eontains more than one elernent, prove that every nonempty proper
subset of X is a zer divisor in the ring P(X).
b) Show that, if n > 1, the matrix ring Mn(R) has zero divison even though the
'_
ring R may not.
6. Suppose !hat R is a ring with identity 1'itnd having no divisor s ofzero. For a, bE R,
verify that
a) ab
1 if and only if ha = 1,
b) if a2
1, then either a 1 or a = -1.
= {nlln E Z}.
Zl
13
(nl)(ml) = (nm)l
binomial expansion
one can easily infer that ZI itself [orms a (commutative) ring with identity.
The order of the additive cyclic group (Z1, +) is simply the characteristic
oC the given ring R.
When R happens to be an integral domain, then Zl is a subdomain of
R (that s, Z1 is also an integral domain with respect to the operations in
R). In fact, ZI is the smallest subdomain oC R, in the sense that it is contained in every other subdomain of R. If R is a domain of characteristic p,
(a
+ bY'
an + ('i.)an-1b
(n~l)Qb-l
+ b",
where
k)!
..
14
11. A Boo/ean ring is a ring with identity every element of which is idempotent. Prov~
that any Boolean ring R is cornmutative. [Hin!: First show that a = -a for
every aE 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 c10sed under both the operations of addition and multiplication.
14. Assume that R is a ring l/lld a E R. If C(a) denotes the set of all elements which
commute with a,
{r E Rlar
= ra},
n.on C(a).
S;
C(a)
S; S is a subring of R}
.is the smaHest (in the sense of inc1usion) 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 rt S, the
subring generated by the set S u {a} is represented by (S, a). If a E cent R,
establish tha t
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 rillg which has no nonzero nilpotent elements, deduce that aH the idempotent elements of R belong to cent R. [Hint: If a2 = a, then (ara - arV =
(ara - ra)2 = O for aH 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 cornmutative ringo [Hint: Utilize the expression
(a + W + (a + b) to show first that ab + ba lies in the center for aH a, b E R.]
23. Let (G, +) be a commutative group and R be the set ofall (group) homomorphisms
of G into itself. The pointwise sum f + g and composition f o g of two functions
f, g E R are defined by the usual rules
(f + g)(x) = f(x)
+ g(x),
= {ro +
rla
+ ... +
r.a"ln E Z+; r
S}.
(f o g)(x) = f(g(x)
(XE G).
"
x"'
(rER).
rx i
i=l
Two such expressions are regarded as equal if they have the same coefficients ..
Addition and multiplication can be defined in R(G) by taking
rx + sx = (r + s)x
(. rx) [. SX) = . tx
i=l
i=l
and
i=l
j,
,=1
l=1
(S, a)
15
PROBLEMS
1=1
where
ti
rjsk
17. Let R be an arbitrary ring apd n E Z+. Ifthe set Sft is defined by
XJXk=Xf
= {a E Rlnka =
(The meaning of the last-written sum is that the surnmation is to be extended over
aH subscripts j and k for which xjxk = x.) Prove that, with respect to these
operations, R(G) constitutes a ring, the so-called group ring of G over R.
S.
14
11. A Boo/ean ring is a ring with identity every element of which is idempotent. Prov~
that any Boolean ring R is cornmutative. [Hin!: First show that a = -a for
every aE 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 c10sed under both the operations of addition and multiplication.
14. Assume that R is a ring l/lld a E R. If C(a) denotes the set of all elements which
commute with a,
{r E Rlar
= ra},
n.on C(a).
S;
C(a)
S; S is a subring of R}
.is the smaHest (in the sense of inc1usion) 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 rt S, the
subring generated by the set S u {a} is represented by (S, a). If a E cent R,
establish tha t
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 rillg which has no nonzero nilpotent elements, deduce that aH the idempotent elements of R belong to cent R. [Hint: If a2 = a, then (ara - arV =
(ara - ra)2 = O for aH 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 cornmutative ringo [Hint: Utilize the expression
(a + W + (a + b) to show first that ab + ba lies in the center for aH a, b E R.]
23. Let (G, +) be a commutative group and R be the set ofall (group) homomorphisms
of G into itself. The pointwise sum f + g and composition f o g of two functions
f, g E R are defined by the usual rules
(f + g)(x) = f(x)
+ g(x),
= {ro +
rla
+ ... +
r.a"ln E Z+; r
S}.
(f o g)(x) = f(g(x)
(XE G).
"
x"'
(rER).
rx i
i=l
Two such expressions are regarded as equal if they have the same coefficients ..
Addition and multiplication can be defined in R(G) by taking
rx + sx = (r + s)x
(. rx) [. SX) = . tx
i=l
i=l
and
i=l
j,
,=1
l=1
(S, a)
15
PROBLEMS
1=1
where
ti
rjsk
17. Let R be an arbitrary ring apd n E Z+. Ifthe set Sft is defined by
XJXk=Xf
= {a E Rlnka =
(The meaning of the last-written sum is that the surnmation is to be extended over
aH subscripts j and k for which xjxk = x.) Prove that, with respect to these
operations, R(G) constitutes a ring, the so-called group ring of G over R.
S.
'
TWO
17
, Example 2-1. For each integer a E Z"let (a) r.epresent the set consisting
of all integral muItiples of a; that is,
(a)
{naln E Z};
na
ma
(n - m)a,
m(mi) = (mn)a,
n, n EZ.
In particular, since (2) = Ze,the ring of even integers forms an {deal of Z. "
Notice, too, that (O) = {O} and (1) = Z.,
;'.,'
Example 2-2. Another illustration is furnished by map (X, R), th~ ring of .'
For a fixed ,l.,~
, element x E X, we denote by Ix the set of all mappings which take on the ...
,value
O at . x:
,,',;
..
)~?ppings froIn the set X into'the ring R (see Example 1-4).
~'.-
Ix
{f E map(X, R)/f(x)
= O}.
Now, choose J, g E Ix and hE map(X, R). From the dfinHion of the ring
operations in map(X, R),
(f -g)(x)
= f(x)-g(x) = O-O = O,
while
(fh)(x) = f(x)h(x)
Oh (x)
O,
= O for all x E S}
Before presenting our next example, we derive a fact which, despite its
apparent simplicity, will be frequently applied in the sequel.
In what follows, let us agree that theterm "ideal", unmodified, wilJ always mean two-sided ideal.
Proa! Let I be an ideal of R and suppose that there is sorne member a =1= O
of I such that a-lexists in R. (The theorem is trivial when I = {O}.) Since'
CONVENTION
'
TWO
17
, Example 2-1. For each integer a E Z"let (a) r.epresent the set consisting
of all integral muItiples of a; that is,
(a)
{naln E Z};
na
ma
(n - m)a,
m(mi) = (mn)a,
n, n EZ.
In particular, since (2) = Ze,the ring of even integers forms an {deal of Z. "
Notice, too, that (O) = {O} and (1) = Z.,
;'.,'
Example 2-2. Another illustration is furnished by map (X, R), th~ ring of .'
For a fixed ,l.,~
, element x E X, we denote by Ix the set of all mappings which take on the ...
,value
O at . x:
,,',;
..
)~?ppings froIn the set X into'the ring R (see Example 1-4).
~'.-
Ix
{f E map(X, R)/f(x)
= O}.
Now, choose J, g E Ix and hE map(X, R). From the dfinHion of the ring
operations in map(X, R),
(f -g)(x)
= f(x)-g(x) = O-O = O,
while
(fh)(x) = f(x)h(x)
Oh (x)
O,
= O for all x E S}
Before presenting our next example, we derive a fact which, despite its
apparent simplicity, will be frequently applied in the sequel.
In what follows, let us agree that theterm "ideal", unmodified, wilJ always mean two-sided ideal.
Proa! Let I be an ideal of R and suppose that there is sorne member a =1= O
of I such that a-lexists in R. (The theorem is trivial when I = {O}.) Since'
CONVENTION
18
that is, R S;; J, whence the equality J = R. This contradicts the hypothesis
that J is a proper subset of R.
Notice that, en route, we have also established
Corollary. In a ring with identty, no proper (right, left, two-sided) ideal
contans the identity elemento
Example 2...3. Ths 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 notational device, let us define Eij to be the n x n matrix having 1 as its ijth
entry and zeroes elsewhere. Now, suppose that J =1= {O} is any ideal of the
ring Mn(R#). Then J must contain sorne nonzero matrix (aij), with, say,
rsth entry a,s =1= O. Since J is a two-sided ideal, the product
Err(b) (a)Ess
is a member of J, where the matrix (b;) is chosen to have the element a;. 1
down its main diagonal and zeroes everywhere else. As a result of al! the
zero entries in the various factors, it is easy to verify that this product is
.equal to E,s' Knowing ths, the relation
Eu
= EE,sE.l
= Ell + E 22 + ... + E nn ,
which leads to the conelusion that (Oij) E J and, appealing to the above
corollary, thatl = Mn(R#). In other words, Mn(R#) possesses no nonzero
proper ideals, as assertt;d...
As a matter of definhon, let us call a ring R =1= {O} simple if R has no
two-sided ideals other tbAn {O} and R. In the light of Example 2-4, the
matrix ring Mn(R#) is a simple ringo
We now take up some of the standard methods for constructing new
ideals from given ones. To begin with simpler things:
Theorem 2-2. Le {J} be an arbitrary collection of (right, left, twos.ided) ideals of the ring R, where i ranges over sorne index set Then
n Ji is also a (right, left, two-sided) ideal of R.
Proof. We give the proof for the case in whch the ideals are two-sided.
First, observe that the intersection n Ji is nonempty, for each of the ideals
Ji must contain the zero element of the ringo Suppose that the elements
a, b E n Ji and r E R. Then a and b are members of Ji' where i varies over
the indexing set. Inasmuch as JI is assumed to be an ideal of R, it follows
that a - b, ar and ra alllie in the set Ji' But this is true for every value of
19
= n {JI S
S;;
J; J is an ideal of R}.
The collection of all ideal s which contain S is not ernpty, since the entire
ring itself is an ideal containing any subset of R; thus, the set (S) exists and
satisfies the inclusion S S;; (S). By virtue ofTheorem 2-2, (S) forms an ideal
of R, lrnown as the ideal generated by the set S. It is noteworthy that whenever J is any ideal of R with S s;; J, then necessarily (S) S;; 1. For tbis rcason,
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 o'he-sided
ideals generated by S.
If S consists of a finite number of elements, sayal'
a2 , , an , then, the
l.,.
idcal'which t4ey generate is customarily denoted by (al' a2, ... , aJo Such an
ideal is said to be finitely generated with the given eIernents al as its
generators. An ideal (a) generated by just one ring eIement 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,
saya, of an arbitrary ring R. The right ideal generated by a is caBed a
principal right ideal and is denoted by (a),. Being closed with respect to
multiplication on the right, (a), necessarily contains al! products ar (r E R),
as well as the elements na (n an integer), and, hence, ineludes 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 iGeal of R. Observe, too, that the elernent a is a member
of the ideal, .since a = aO + la. These remarks make it clear that
(a),
{ar
+ nalrER; nE Z}.
ar
+ na
= ar
+ a(nl) =
a(r
nI)
ar',
where r ' = r + nI is some ring elemento Thus, the set (a), consists of all
right multiples of a byelements of R. If R is a ring with identity, we shall
frequently employ the more suggestive notation aR in place of (a),; that is,
(a)r = aR
Similar remarks
bya.
{arlr E R}.
18
that is, R S;; J, whence the equality J = R. This contradicts the hypothesis
that J is a proper subset of R.
Notice that, en route, we have also established
Corollary. In a ring with identty, no proper (right, left, two-sided) ideal
contans the identity elemento
Example 2...3. Ths 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 notational device, let us define Eij to be the n x n matrix having 1 as its ijth
entry and zeroes elsewhere. Now, suppose that J =1= {O} is any ideal of the
ring Mn(R#). Then J must contain sorne nonzero matrix (aij), with, say,
rsth entry a,s =1= O. Since J is a two-sided ideal, the product
Err(b) (a)Ess
is a member of J, where the matrix (b;) is chosen to have the element a;. 1
down its main diagonal and zeroes everywhere else. As a result of al! the
zero entries in the various factors, it is easy to verify that this product is
.equal to E,s' Knowing ths, the relation
Eu
= EE,sE.l
= Ell + E 22 + ... + E nn ,
which leads to the conelusion that (Oij) E J and, appealing to the above
corollary, thatl = Mn(R#). In other words, Mn(R#) possesses no nonzero
proper ideals, as assertt;d...
As a matter of definhon, let us call a ring R =1= {O} simple if R has no
two-sided ideals other tbAn {O} and R. In the light of Example 2-4, the
matrix ring Mn(R#) is a simple ringo
We now take up some of the standard methods for constructing new
ideals from given ones. To begin with simpler things:
Theorem 2-2. Le {J} be an arbitrary collection of (right, left, twos.ided) ideals of the ring R, where i ranges over sorne index set Then
n Ji is also a (right, left, two-sided) ideal of R.
Proof. We give the proof for the case in whch the ideals are two-sided.
First, observe that the intersection n Ji is nonempty, for each of the ideals
Ji must contain the zero element of the ringo Suppose that the elements
a, b E n Ji and r E R. Then a and b are members of Ji' where i varies over
the indexing set. Inasmuch as JI is assumed to be an ideal of R, it follows
that a - b, ar and ra alllie in the set Ji' But this is true for every value of
19
= n {JI S
S;;
J; J is an ideal of R}.
The collection of all ideal s which contain S is not ernpty, since the entire
ring itself is an ideal containing any subset of R; thus, the set (S) exists and
satisfies the inclusion S S;; (S). By virtue ofTheorem 2-2, (S) forms an ideal
of R, lrnown as the ideal generated by the set S. It is noteworthy that whenever J is any ideal of R with S s;; J, then necessarily (S) S;; 1. For tbis rcason,
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 o'he-sided
ideals generated by S.
If S consists of a finite number of elements, sayal'
a2 , , an , then, the
l.,.
idcal'which t4ey generate is customarily denoted by (al' a2, ... , aJo Such an
ideal is said to be finitely generated with the given eIernents al as its
generators. An ideal (a) generated by just one ring eIement 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,
saya, of an arbitrary ring R. The right ideal generated by a is caBed a
principal right ideal and is denoted by (a),. Being closed with respect to
multiplication on the right, (a), necessarily contains al! products ar (r E R),
as well as the elements na (n an integer), and, hence, ineludes 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 iGeal of R. Observe, too, that the elernent a is a member
of the ideal, .since a = aO + la. These remarks make it clear that
(a),
{ar
+ nalrER; nE Z}.
ar
+ na
= ar
+ a(nl) =
a(r
nI)
ar',
where r ' = r + nI is some ring elemento Thus, the set (a), consists of all
right multiples of a byelements of R. If R is a ring with identity, we shall
frequently employ the more suggestive notation aR in place of (a),; that is,
(a)r = aR
Similar remarks
bya.
{arlr E R}.
20
{na
+ ra + as +
r,as;!r,s,r"s,ER; nEZ}.,'
finite
+ a2 + ... + a"la,EIJ,
I i ; aH but a
Proof If 1 = {O}, the theorem is triviaHy true, since' the zero ideal {O} is
the principal ideal generated by O. Suppose then tbat 1 does not consist
ofthe zero element alone. Now, ifm E 1, -m also Iles in 1, so that the set 1
contains positive integers. Let n designate the least positive integer in 1.
As 1 forms a~ ideal of Z, each integral multiple of n must belong to 1, whence
{al
'=
{a,la,
+ ... + In
..
finte~.llnber
ofthe a ~~e O},
"':.;:
where it s understood that represents an arbitrar y sum whh orie or more
I,
11 +1 2
finite
The reader will take careto remember that, although {J,} may be an infinite
family of ideals, only finte sums of elemnts of R are involved in the
definition above. An alternative descriptio~o'f 1, could be 'given by
~~L
'
To establish the inc1usion 1 ~ (n), let k be an arbitrary element of 1.
By the di:vision aigorithm there existintegers q and r for which k = qn + r,
with O ~ r < n. Since k and qn are both members of 1, it foHows that
r = k - qn E 1. Ifr > O, we would have a contradiction to the assumption
that n is the smaHest positive integer in 1. Accordingly, r ~ O and
k = qn E (n). Thus, only multiples of n belong to 1, implying that 1 ~ (n).
The two inc1usions show that 1 = (n) and the argument is complete.
+ J = {a + b/aEI; bEJ}.
(a)
21
a) R = 11
b) Ii n (JI
Then the
al
x = al
+ b1 = a2 +
Q2
(a i EI 1, bEI 2 ).
20
{na
+ ra + as +
r,as;!r,s,r"s,ER; nEZ}.,'
finite
+ a2 + ... + a"la,EIJ,
I i ; aH but a
Proof If 1 = {O}, the theorem is triviaHy true, since' the zero ideal {O} is
the principal ideal generated by O. Suppose then tbat 1 does not consist
ofthe zero element alone. Now, ifm E 1, -m also Iles in 1, so that the set 1
contains positive integers. Let n designate the least positive integer in 1.
As 1 forms a~ ideal of Z, each integral multiple of n must belong to 1, whence
{al
'=
{a,la,
+ ... + In
..
finte~.llnber
ofthe a ~~e O},
"':.;:
where it s understood that represents an arbitrar y sum whh orie or more
I,
11 +1 2
finite
The reader will take careto remember that, although {J,} may be an infinite
family of ideals, only finte sums of elemnts of R are involved in the
definition above. An alternative descriptio~o'f 1, could be 'given by
~~L
'
To establish the inc1usion 1 ~ (n), let k be an arbitrary element of 1.
By the di:vision aigorithm there existintegers q and r for which k = qn + r,
with O ~ r < n. Since k and qn are both members of 1, it foHows that
r = k - qn E 1. Ifr > O, we would have a contradiction to the assumption
that n is the smaHest positive integer in 1. Accordingly, r ~ O and
k = qn E (n). Thus, only multiples of n belong to 1, implying that 1 ~ (n).
The two inc1usions show that 1 = (n) and the argument is complete.
+ J = {a + b/aEI; bEJ}.
(a)
21
a) R = 11
b) Ii n (JI
Then the
al
x = al
+ b1 = a2 +
Q2
(a i EI 1, bEI 2 ).
22
Then al - a 2 = b2 - bl' But the left-hand side of this last equation lies
in 1 1 , while the right-hand side is in 12 , so that both sides be long to
11 n 12 = {O}. Itfollowsthata 1 - a 2 = b2 - b1 = 0,ora 1 = a2 ,b 1 = b2
In other words, x is uniquely representable as a sum a + b, a E 1 l' 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 + O (here x E 1 1
and OE 12 ) and x = O + x (here OE 1 1 and x E 12 ), The uniqueness
assumption of(2) implies that x = O, in'consequence ofwhich 1 1 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 1 and J are two ideals of
the ring R. To be consistent with our earlier definition of the sum 1 + J,
we should define fue product 'IJ to be the collection of all simple products
ab, where a E 1 and b EJ. 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 = O=ablaEl; bEJ}.
finile
23
::2
12
::2
13
::2 '"
::2
1"
::2
Remark. If 1 is a right ideal and S a nonempty subset of the ring R, then '
SI
{L arla E S; r El}
finite
al
{arlr E I}.
Definition 2-5. Let 1 and J be two ideal s of the ring R. The right (left)
quotient of 1 by J, denoted by the symbol 1 :r J (I :1 J), consists of all
elements a E R such that aJ S;; 1 (:la S;; 1). In the event R is a commutative ring, we simply writel: J.
It is by no means obvious that the set
where the a and a; are in 1, and the b and b; are in J. From this we obtain
x - y
rx
(-a~)b~,
1"
= {L..,
"
finite
1:.1 = {aERlaJ
S;;
I}
22
Then al - a 2 = b2 - bl' But the left-hand side of this last equation lies
in 1 1 , while the right-hand side is in 12 , so that both sides be long to
11 n 12 = {O}. Itfollowsthata 1 - a 2 = b2 - b1 = 0,ora 1 = a2 ,b 1 = b2
In other words, x is uniquely representable as a sum a + b, a E 1 l' 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 + O (here x E 1 1
and OE 12 ) and x = O + x (here OE 1 1 and x E 12 ), The uniqueness
assumption of(2) implies that x = O, in'consequence ofwhich 1 1 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 1 and J are two ideals of
the ring R. To be consistent with our earlier definition of the sum 1 + J,
we should define fue product 'IJ to be the collection of all simple products
ab, where a E 1 and b EJ. 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 = O=ablaEl; bEJ}.
finile
23
::2
12
::2
13
::2 '"
::2
1"
::2
Remark. If 1 is a right ideal and S a nonempty subset of the ring R, then '
SI
{L arla E S; r El}
finite
al
{arlr E I}.
Definition 2-5. Let 1 and J be two ideal s of the ring R. The right (left)
quotient of 1 by J, denoted by the symbol 1 :r J (I :1 J), consists of all
elements a E R such that aJ S;; 1 (:la S;; 1). In the event R is a commutative ring, we simply writel: J.
It is by no means obvious that the set
where the a and a; are in 1, and the b and b; are in J. From this we obtain
x - y
rx
(-a~)b~,
1"
= {L..,
"
finite
1:.1 = {aERlaJ
S;;
I}
24
=
=
{a
RlaJ ~ n 1}
{a
i}
n{aERlaJ ~ 1} c= n(l:r J )
25
. ,
Remark. Simila~ results,:h6Id for left quotients; the sole difference being
that, instead of (3),'one riD\V has 1 : (J K) = (1 : J) :1 K.
This may be a good pl~ce to observe that if 1 is an ide~l of. the ring R
and J is an ideal of 1, then J need not be 'an ideal of the enhre rmg. R. For
an illustration, we tum to the ring map R # and l~t R be the ~ubnng consisting of all continuous functions from, R # into ltself. Conslder the sets
l: r ~,J
= fa E Rla(~ J;)~ 1}
= {q:E RlaJ' ~ Ifor all i}
n (1 :r J).
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 1,which, in turn, forms an
2
2
ideal ofR. However, J fails to be an ideal of R, since i E J, while ti .~ J.
(The symbol .1 is used in this setting to represent the constant funch.on
2
whose value a~ each real number is l) We as sume that ti E J and denve
a contradiction. Then,
ti 2 = fi 2 + ni 2
for a suitable choice of f E R and n ~ Z, with feO) = O. In consequence,
fi2 = (t - n)i 2, implying that f(x) = t - n.=I= O for ;very O =1= x E R~;
in other words, f is a nonzero constant functlOn on R - {O}. But tbis
obviously violates the continuity off 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 sorne a' E R such that aa' a = a.
. ~
' I
such that
f(a+ b) = fea)
+ f(b), f(ab)
= f(a)f(b)
24
=
=
{a
RlaJ ~ n 1}
{a
i}
n{aERlaJ ~ 1} c= n(l:r J )
25
. ,
Remark. Simila~ results,:h6Id for left quotients; the sole difference being
that, instead of (3),'one riD\V has 1 : (J K) = (1 : J) :1 K.
This may be a good pl~ce to observe that if 1 is an ide~l of. the ring R
and J is an ideal of 1, then J need not be 'an ideal of the enhre rmg. R. For
an illustration, we tum to the ring map R # and l~t R be the ~ubnng consisting of all continuous functions from, R # into ltself. Conslder the sets
l: r ~,J
= fa E Rla(~ J;)~ 1}
= {q:E RlaJ' ~ Ifor all i}
n (1 :r J).
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 1,which, in turn, forms an
2
2
ideal ofR. However, J fails to be an ideal of R, since i E J, while ti .~ J.
(The symbol .1 is used in this setting to represent the constant funch.on
2
whose value a~ each real number is l) We as sume that ti E J and denve
a contradiction. Then,
ti 2 = fi 2 + ni 2
for a suitable choice of f E R and n ~ Z, with feO) = O. In consequence,
fi2 = (t - n)i 2, implying that f(x) = t - n.=I= O for ;very O =1= x E R~;
in other words, f is a nonzero constant functlOn on R - {O}. But tbis
obviously violates the continuity off 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 sorne a' E R such that aa' a = a.
. ~
' I
such that
f(a+ b) = fea)
+ f(b), f(ab)
= f(a)f(b)
26
For future use, we shalllabel the set of all hornornorphisrns frorn the
ring R into the ring R' by the syrnbol horn(R, R'). In the event that R = R',
the sirnpler notation horn R will be used in place of horn(R, R). (Sorne
authors prefer to write end R, for endornorphisrn, in place of horn R; both
notations have a certain suggestive power and it reduces to a rnatter of
personal preference.)
A knowledge of a few sirnple-rninded exarnples will help to fix ideas.
Example 2-4. Let R and R' be arbitrary rings andf: R --> R' be the function
which sends each elernent of R to the zero elernent of R'. Then,
f(a
+ b) =
f(ab)
O = O + O = f(a)
= O = OO = f(a)
+ f(b),
(a, b E R),
f(b)
2-7.
Example 2-5. Consider the ring Z ofintegers and the ring Zn of integers
modulo n. Definef: Z --> Zn by takingf(a) = [a]; that is, rnap each integer
into the congruence class containing it. Thatfis a hornornorphisrn follows
directly frorn the definition of the operations i Zn:
f(a + b) = [a + b]
f(ab)
[ab]
[a] +n [b]
[aln[b]
f(a) +nf(b),
= f(a)nf(b).
Example 2-6. In the ring rnap(X, R), define La to be the function which
assigns to each fE rnap(X, R) its value at a fixed elernent a E X; in other
words, La(J) =:= f(a): Then La is a hornornorphisrn frorn rnap(X, R) into R,
known as the evaluation homomorphism at a. We need only observe that
Two cornrnents regarding part (3) of the aboye theorern are in order:
First, it is evident that
f(a)l
= (Jg)(a) = f(a)g(a)
f(a - b)
f(a)
f(a1) = f(a)f(l)
f(a)
+ f( -
b)
f(a) - f(b).
= Ta(J)~a(g).
l.
Proo! 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 th~ set S, as do a -:- b and
ab (S being a subring of R). Hence,
f(a) - f(b) = f(a - b) Ef(S)
and
If, in addition, R and R' are both rings with identity and f(R) = R',
then
3) f(l) = 1,
4) f(a- 1) = f(a)-l for each in vertible elernent a
27
R.
26
For future use, we shalllabel the set of all hornornorphisrns frorn the
ring R into the ring R' by the syrnbol horn(R, R'). In the event that R = R',
the sirnpler notation horn R will be used in place of horn(R, R). (Sorne
authors prefer to write end R, for endornorphisrn, in place of horn R; both
notations have a certain suggestive power and it reduces to a rnatter of
personal preference.)
A knowledge of a few sirnple-rninded exarnples will help to fix ideas.
Example 2-4. Let R and R' be arbitrary rings andf: R --> R' be the function
which sends each elernent of R to the zero elernent of R'. Then,
f(a
+ b) =
f(ab)
O = O + O = f(a)
= O = OO = f(a)
+ f(b),
(a, b E R),
f(b)
2-7.
Example 2-5. Consider the ring Z ofintegers and the ring Zn of integers
modulo n. Definef: Z --> Zn by takingf(a) = [a]; that is, rnap each integer
into the congruence class containing it. Thatfis a hornornorphisrn follows
directly frorn the definition of the operations i Zn:
f(a + b) = [a + b]
f(ab)
[ab]
[a] +n [b]
[aln[b]
f(a) +nf(b),
= f(a)nf(b).
Example 2-6. In the ring rnap(X, R), define La to be the function which
assigns to each fE rnap(X, R) its value at a fixed elernent a E X; in other
words, La(J) =:= f(a): Then La is a hornornorphisrn frorn rnap(X, R) into R,
known as the evaluation homomorphism at a. We need only observe that
Two cornrnents regarding part (3) of the aboye theorern are in order:
First, it is evident that
f(a)l
= (Jg)(a) = f(a)g(a)
f(a - b)
f(a)
f(a1) = f(a)f(l)
f(a)
+ f( -
b)
f(a) - f(b).
= Ta(J)~a(g).
l.
Proo! 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 th~ set S, as do a -:- b and
ab (S being a subring of R). Hence,
f(a) - f(b) = f(a - b) Ef(S)
and
If, in addition, R and R' are both rings with identity and f(R) = R',
then
3) f(l) = 1,
4) f(a- 1) = f(a)-l for each in vertible elernent a
27
R.
28
f(b) must be members of S'. Since S' is asubring of R', it follows at once
that
fea - b) = fea) - f(b) E S'
and
f(ab) = f(a)f(b) E S'.
. , This means that a - b and ab le in f-l(S'), from which we conc1ude.that
f-l(S') forms a subring of R.
Left unresolved is the matter ofreplacing the term "subring" in Theorem
2-8 by "ideal". It is not difficult to show that part (2) of the th~orem
remains true under such a substitution. More precisely: if /' is an ideal of
R', then the subringf-l(!') is an ideal of R. For instance, suppo~ethat
a Ef-l(!'), so thatf(a) E /', and let r be an arbitrary element of R. >.Then,
f(ra) = f(r)f(a) E/';.in other words, the product ra is inf-l(/,). Likwise,
ar E f-l(/,), which helps to make f-l(!') an ideal of R .
.:.:..
. Without further restriction, it cannot be inferred that the image'f(l)
will be an ideal of R', whenever 1 is an ideal of R. One would need to know
that r'f(a) Ef(l) for all r' E R' and a E l. In general, there isno way of
replacing r' by som~f(r) in order to exploit the fact that l is an ideal. The
answer is o bvious : j ust take f to be an onto mapping.
Surnmarizing these remarks, we may now state:
CoroUary. 1) For each ideal /' of R', the subring f-l(/,) is an ideal
of R.
2) Ifj(R) = R', then for each ideal l of R, the subringf(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 j,consistsof those elements in
R which are mapped by f onto the zero element of the ring R':
.
= {a E Elf(a) = O}.
indicates that ker f is a nonempty
ker f
Theorem 2-7
subset of R, since, if
nothing else, O E ker f. Except for the case .of the trivial homomorphism,
the kernelwill alwa,ys 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 Iof a homomorphism f from a ring R
into a ring R' is an ideal'of R.
Proof. We already know that the trivjal subring {O} forms an ideal of R'.
Since ker f = f-l(O), the conclusion follows from the last corollary.
29
= fea) - f(b) =
O,';
Zlnl = O} = (P)
28
f(b) must be members of S'. Since S' is asubring of R', it follows at once
that
fea - b) = fea) - f(b) E S'
and
f(ab) = f(a)f(b) E S'.
. , This means that a - b and ab le in f-l(S'), from which we conc1ude.that
f-l(S') forms a subring of R.
Left unresolved is the matter ofreplacing the term "subring" in Theorem
2-8 by "ideal". It is not difficult to show that part (2) of the th~orem
remains true under such a substitution. More precisely: if /' is an ideal of
R', then the subringf-l(!') is an ideal of R. For instance, suppo~ethat
a Ef-l(!'), so thatf(a) E /', and let r be an arbitrary element of R. >.Then,
f(ra) = f(r)f(a) E/';.in other words, the product ra is inf-l(/,). Likwise,
ar E f-l(/,), which helps to make f-l(!') an ideal of R .
.:.:..
. Without further restriction, it cannot be inferred that the image'f(l)
will be an ideal of R', whenever 1 is an ideal of R. One would need to know
that r'f(a) Ef(l) for all r' E R' and a E l. In general, there isno way of
replacing r' by som~f(r) in order to exploit the fact that l is an ideal. The
answer is o bvious : j ust take f to be an onto mapping.
Surnmarizing these remarks, we may now state:
CoroUary. 1) For each ideal /' of R', the subring f-l(/,) is an ideal
of R.
2) Ifj(R) = R', then for each ideal l of R, the subringf(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 j,consistsof those elements in
R which are mapped by f onto the zero element of the ring R':
.
= {a E Elf(a) = O}.
indicates that ker f is a nonempty
ker f
Theorem 2-7
subset of R, since, if
nothing else, O E ker f. Except for the case .of the trivial homomorphism,
the kernelwill alwa,ys 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 Iof a homomorphism f from a ring R
into a ring R' is an ideal'of R.
Proof. We already know that the trivjal subring {O} forms an ideal of R'.
Since ker f = f-l(O), the conclusion follows from the last corollary.
29
= fea) - f(b) =
O,';
Zlnl = O} = (P)
I
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 l of the ring R determines an
ideal f(I) of the ring R'. It goes without saying that ring theory would be
considerably simplified ifthe 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 refiected in the fact that if l and J are two ideals of R
with l ~ J ~ l + kerJ, thenf(I) = f(J). The quickest way to see this is
to notice tbat
f(l) ~ f(J) ~ f(l
kerf) = f(I)
+ f(kerf)
31
= f(l),
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 kerf = {O} or else narrowing our view to coilsider only ideals l with
ker f ~ l. In either event, it follows that l ~ J ~ l + ker f = l and, in
consequence, l = J. The first of the restrictions just cited has the effect of
making the function f one-to-one, in which case R and 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 tum aside briefiy to establish a preliminary lemma which will
pro vide the key to later success.
Lemma. Letfbe a homomorphism froro the ring R onto the ring R'.
If l is any ideal of R such that ker f ~ 1, then l = f - 1 (J(l) ).
Proof. Suppose that the element a E r1(j{l)), so that f{a} E f{l). Then
f(a) = f(r) for sorne choice ofr in l. As a tesult, we will havef(a - r) = 0,
or, what amounts to the same thing, a - rE ker f ~ l. This implies that
a E l, yielding the inclusion f-1(J(I))~.-I. Since the reverse inclusion
l = f-1(J(I)) = f-1(J(J)) = J.
actually maps onto the set of all ideals of R'. In other words, starting with
an ideal /' of R', we must produce sorne ideal l of the ring R, with ker f ~ l,
oH-
:,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),
f- 1(/,).
la
and, similarly,
(O, l)(a, n) = (a, n).
I
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 l of the ring R determines an
ideal f(I) of the ring R'. It goes without saying that ring theory would be
considerably simplified ifthe 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 refiected in the fact that if l and J are two ideals of R
with l ~ J ~ l + kerJ, thenf(I) = f(J). The quickest way to see this is
to notice tbat
f(l) ~ f(J) ~ f(l
kerf) = f(I)
+ f(kerf)
31
= f(l),
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 kerf = {O} or else narrowing our view to coilsider only ideals l with
ker f ~ l. In either event, it follows that l ~ J ~ l + ker f = l and, in
consequence, l = J. The first of the restrictions just cited has the effect of
making the function f one-to-one, in which case R and 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 tum aside briefiy to establish a preliminary lemma which will
pro vide the key to later success.
Lemma. Letfbe a homomorphism froro the ring R onto the ring R'.
If l is any ideal of R such that ker f ~ 1, then l = f - 1 (J(l) ).
Proof. Suppose that the element a E r1(j{l)), so that f{a} E f{l). Then
f(a) = f(r) for sorne choice ofr in l. As a tesult, we will havef(a - r) = 0,
or, what amounts to the same thing, a - rE ker f ~ l. This implies that
a E l, yielding the inclusion f-1(J(I))~.-I. Since the reverse inclusion
l = f-1(J(I)) = f-1(J(J)) = J.
actually maps onto the set of all ideals of R'. In other words, starting with
an ideal /' of R', we must produce sorne ideal l of the ring R, with ker f ~ l,
oH-
:,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),
f- 1(/,).
la
and, similarly,
(O, l)(a, n) = (a, n).
.~
;1
;1
32
(a - b, O),
(ab, O),
f(au), = f(a)f(u)
f(a)1
fea),
+ b)
= f(a
=
f(au)
g(a)
33
g(ab)
f(abu)
= f(abu Z)
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 offto the set R'i9all it h. Sincefan'd h must agree
on 1 and, more specifically, at the elemept u, h(u) = f(u) = -1. With this
in mind, it follows that
_,,:', '
h(a)
for all a E R and so h and g are the sa~~ 'fl;J,nction. Hence, tl1ere.i:; one and
only one way of extendingfhomomorphiclly from the ideallto'the whole
.
ring R.Before closing the present chapter, there is another type of direct sum
whichdeserves mention. To this purpose, let R I, R z' ... , R" be a finite
number of rings (not necessarily subrings of a common ring) and consider
x R consisting of all ordered n-tuples
their Cartesian product R
(al' az' Oo., a"), with al E R' One can easily convert R into a ring by
performing the ring operations componentwise; in other words, if
(al' az' : .. , a") and (b l , bz' ... , b,,) are two elements of R, simply define
(al' az, : .. , all )
(al
b,,)
and
(al' az, ... , all)(bl,b z' ... , b,,) = (a l b l , azb z, ... , a"b").
The ring so obtained is caBed the external direct sum of R l , R z, , .. , R"
and is convenientIy written R = RI
Rz
R". (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 interna! 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 c1ear in the next paragraph.
If R is the external direct sum of the rings R (i = 1, 2, .,. , n), then the
individual R need not be subrngs, or even subsets, of R. However, there is
an ideal of R which is the isomorphic image of R' A straightforward
calculation will convillce the reader that the set
+ + ... +
I1
.~
;1
;1
32
(a - b, O),
(ab, O),
f(au), = f(a)f(u)
f(a)1
fea),
+ b)
= f(a
=
f(au)
g(a)
33
g(ab)
f(abu)
= f(abu Z)
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 offto the set R'i9all it h. Sincefan'd h must agree
on 1 and, more specifically, at the elemept u, h(u) = f(u) = -1. With this
in mind, it follows that
_,,:', '
h(a)
for all a E R and so h and g are the sa~~ 'fl;J,nction. Hence, tl1ere.i:; one and
only one way of extendingfhomomorphiclly from the ideallto'the whole
.
ring R.Before closing the present chapter, there is another type of direct sum
whichdeserves mention. To this purpose, let R I, R z' ... , R" be a finite
number of rings (not necessarily subrings of a common ring) and consider
x R consisting of all ordered n-tuples
their Cartesian product R
(al' az' Oo., a"), with al E R' One can easily convert R into a ring by
performing the ring operations componentwise; in other words, if
(al' az' : .. , a") and (b l , bz' ... , b,,) are two elements of R, simply define
(al' az, : .. , all )
(al
b,,)
and
(al' az, ... , all)(bl,b z' ... , b,,) = (a l b l , azb z, ... , a"b").
The ring so obtained is caBed the external direct sum of R l , R z, , .. , R"
and is convenientIy written R = RI
Rz
R". (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 interna! 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 c1ear in the next paragraph.
If R is the external direct sum of the rings R (i = 1, 2, .,. , n), then the
individual R need not be subrngs, or even subsets, of R. However, there is
an ideal of R which is the isomorphic image of R' A straightforward
calculation will convillce the reader that the set
+ + ... +
I1
34
PROBLEMS
therefore regard R as being an ideal of the ring R'. Our hypothesis now
comes into play and asserts that R' = R E!1 J fr a suitable ideal J ~f R'.
It is thus possible to choose an element (e, n) in J so that (O, -1) = (r, O) +
(e, n), for sorne r E R. The last-written equation tells us that e = - r and
n = -1; what is important is the resulting conc1usion that (e, -1) E J.
For arbitrary rE R, the product (r, O){e, -1) = (re - r, O) will consequently
be in both R and J (each being an ideal of R'). The fact that R n J = {O}
forces (re - r, O) = (O, O); hence, re = r. In a like fashion, we obtain
er = r, proving that R admits the element e as an identity.
(that is, the set consisting of aH n-tuples with zeroes in all places but the
ith) forms an ideal of R naturally isomorphic to R under the mapping
which sends (O, ... , O, a, O, ... ,O) to the element a. Since
(al' a 2 ,
a n ) = (al' O, O, ... , O)
+ '" +
(R ~ 1J
35
PROBLEMS
1. If 1 is a right ideal and J a left ideal of the ring R such that 1 n J
that ab = O for aH a E 1, b E J.
= {O}, prove
QE
R}.
I~
3. a) Show by example that if 1 and J are both ideals of the ring R, then 1 u J need
not bean ideal oC R.
b) If {1} (i = 1, 2, ...) is a colIection of ideal s ofthe ring R such thatI 1 ~ 12 ~ ...
~ In ~ ... , prove that u li is also an ideal of R.
4. Consider the ring Mn(R) oC n x n matrices over R, a ring with identity. A square
matrix (alj) is said to be upper triqngular if aij = O for i > j and strictly upper
triangular if a] = O Cor i ~ j. Let T,,(R) and T~(R) denote the sets of aH upper
triangular.and strictIy 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).
e) A matrix (aij) E T" (R) is invertible in T" (R) if and only if ai is invertible in R
Cor i = 1,2, ... , n. [Hint: Induct on the order n.]
d) Any matrix (aj) E T~(R) is nilpotent; in particular, (a])n = O.
5. Let 1 b: an ideal of R, a commutative ring with identity. For an element a E R,
the ideal generated by the set 1 u {a} is denoted by (1, a). Assurning that a rt 1,
show that
(1, a)
= {i +
rali
1, r
R}.
6. In the ring Z of integers consider the principal ideals (n) and (m) generated by
the integers n and m. Using the notation of the previous probJem, verify that
((nj, m)
((m), n)
= (n) +
(m) = (n, m)
= (d),
34
PROBLEMS
therefore regard R as being an ideal of the ring R'. Our hypothesis now
comes into play and asserts that R' = R E!1 J fr a suitable ideal J ~f R'.
It is thus possible to choose an element (e, n) in J so that (O, -1) = (r, O) +
(e, n), for sorne r E R. The last-written equation tells us that e = - r and
n = -1; what is important is the resulting conc1usion that (e, -1) E J.
For arbitrary rE R, the product (r, O){e, -1) = (re - r, O) will consequently
be in both R and J (each being an ideal of R'). The fact that R n J = {O}
forces (re - r, O) = (O, O); hence, re = r. In a like fashion, we obtain
er = r, proving that R admits the element e as an identity.
(that is, the set consisting of aH n-tuples with zeroes in all places but the
ith) forms an ideal of R naturally isomorphic to R under the mapping
which sends (O, ... , O, a, O, ... ,O) to the element a. Since
(al' a 2 ,
a n ) = (al' O, O, ... , O)
+ '" +
(R ~ 1J
35
PROBLEMS
1. If 1 is a right ideal and J a left ideal of the ring R such that 1 n J
that ab = O for aH a E 1, b E J.
= {O}, prove
QE
R}.
I~
3. a) Show by example that if 1 and J are both ideals of the ring R, then 1 u J need
not bean ideal oC R.
b) If {1} (i = 1, 2, ...) is a colIection of ideal s ofthe ring R such thatI 1 ~ 12 ~ ...
~ In ~ ... , prove that u li is also an ideal of R.
4. Consider the ring Mn(R) oC n x n matrices over R, a ring with identity. A square
matrix (alj) is said to be upper triqngular if aij = O for i > j and strictly upper
triangular if a] = O Cor i ~ j. Let T,,(R) and T~(R) denote the sets of aH upper
triangular.and strictIy 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).
e) A matrix (aij) E T" (R) is invertible in T" (R) if and only if ai is invertible in R
Cor i = 1,2, ... , n. [Hint: Induct on the order n.]
d) Any matrix (aj) E T~(R) is nilpotent; in particular, (a])n = O.
5. Let 1 b: an ideal of R, a commutative ring with identity. For an element a E R,
the ideal generated by the set 1 u {a} is denoted by (1, a). Assurning that a rt 1,
show that
(1, a)
= {i +
rali
1, r
R}.
6. In the ring Z of integers consider the principal ideals (n) and (m) generated by
the integers n and m. Using the notation of the previous probJem, verify that
((nj, m)
((m), n)
= (n) +
(m) = (n, m)
= (d),
---------~~--
--- -
-~-~-~--~----~-~~-
---------~
,1
"
36
I.
PROBLEMS
,37
7. Suppose that 1 is ~ left ideal and J a right ideal of the ring R. Consider the set
1J =
where I: represents a finite sum of one or more terms. Establish that 1J is a twosided ideal of R and, whenever 1 and J are themselves two-sided, tbat IJ.;; 1 n J.
8. If S is any
giv~n
R IRa s; 1}
O for aIl a E S}
17. Given that f is a homomorphism from the ring R onto the ring R', prove that
a) f(eent R) S; cent R'.
.
.
b) If R 18 a principal ideal rng, then the same is true of R'. [Hint: For any a E R,
f(a)
(f(a.]
,..,
18. Let R be a ring witl:J identity. For eaeh invrtible element a E R, show that the
funetionf.: R ..... R defined by f.(x) = axa~'11s an automorphism of R.
19. Let f be a homomorphism from the rng R j1to itself and S be Jhe set of elements
that are left fixe;d by f; in symbols,
9. Le! 11,1 2 , ... ,1. be ideals ofthe ring R with R = 11 + 12 + ... + In' Show
tbat this sum is direct if and only if al + a2 + ... + ah = O, with a E 1, implies
tbat each a = O;
10. lf P(X) is the ring of all subsets of a given set X, prove tllat
a) the collection of all finite subsets of X forms an ideal of P(X);
b) for each subset Y ;; X, P(Y) nd P(X ~ Y) are both principal ideals of P(X),
with P(X)
P(Y) Ea P(X ~ Y).
11. Suppose tbat R is a commutative ring with identity and that the element a E R
is an idempotent dilferellt from Oor 1. Prove that R is the direct Sum of the principal
a).
ideals (a) and (1
+J
= (3),
1n J
(841
IJ
= (21), then
(252),
l:J
= a}.
21. Letfbe a homomorphism from the cornmutative rng R onto the rng R'. If 1 and
, Jare deals of R, verfy each ofthe following:
a)
b)
e)
d)
22. Sho'l'{ that the relation R ~ R' is a equivalenee relation on any set f rings.
{a E Rlf(a)
= (4),
J:1
= (7).
23. Let R be an arbitrary ringo For each fixed element a E R, define the left-multiplication function T,,: R ..... R by taking T,,(x)
ax. If TR denotes th!; set of all such
funetions, prove the following:
a) T" is a (group) homomorphism of the additive group of R into itself;
b) TI! forms a rng, where multiplicaton is taken to be funetional eomposition;
e) He mappingf(a) = T" determines a homomorphism of R onto the ring TR ;
d) the kernel offis the ideal annlR;
.
e) iUor each O =1= a E R, there exists so me b E R such that ab =1= O, then R ~ TR
(In particular, part(e) holds whenever R has an identity eIement.)
24. Let R be an arbitrary ring and R x Z be the exten8ion rng eonstructed in Theorern
2-12. Establish that
a) R)( {O} ii an ideal of R x Z;
b) Z~ {O}xZ;
e) if a is an idempotent element of R, then the pair (-a, 1) is idempotent in R x Z,
while (a, O) is a zero divisor.
~---
---------~~--
--- -
-~-~-~--~----~-~~-
---------~
,1
"
36
I.
PROBLEMS
,37
7. Suppose that 1 is ~ left ideal and J a right ideal of the ring R. Consider the set
1J =
where I: represents a finite sum of one or more terms. Establish that 1J is a twosided ideal of R and, whenever 1 and J are themselves two-sided, tbat IJ.;; 1 n J.
8. If S is any
giv~n
R IRa s; 1}
O for aIl a E S}
17. Given that f is a homomorphism from the ring R onto the ring R', prove that
a) f(eent R) S; cent R'.
.
.
b) If R 18 a principal ideal rng, then the same is true of R'. [Hint: For any a E R,
f(a)
(f(a.]
,..,
18. Let R be a ring witl:J identity. For eaeh invrtible element a E R, show that the
funetionf.: R ..... R defined by f.(x) = axa~'11s an automorphism of R.
19. Let f be a homomorphism from the rng R j1to itself and S be Jhe set of elements
that are left fixe;d by f; in symbols,
9. Le! 11,1 2 , ... ,1. be ideals ofthe ring R with R = 11 + 12 + ... + In' Show
tbat this sum is direct if and only if al + a2 + ... + ah = O, with a E 1, implies
tbat each a = O;
10. lf P(X) is the ring of all subsets of a given set X, prove tllat
a) the collection of all finite subsets of X forms an ideal of P(X);
b) for each subset Y ;; X, P(Y) nd P(X ~ Y) are both principal ideals of P(X),
with P(X)
P(Y) Ea P(X ~ Y).
11. Suppose tbat R is a commutative ring with identity and that the element a E R
is an idempotent dilferellt from Oor 1. Prove that R is the direct Sum of the principal
a).
ideals (a) and (1
+J
= (3),
1n J
(841
IJ
= (21), then
(252),
l:J
= a}.
21. Letfbe a homomorphism from the cornmutative rng R onto the rng R'. If 1 and
, Jare deals of R, verfy each ofthe following:
a)
b)
e)
d)
22. Sho'l'{ that the relation R ~ R' is a equivalenee relation on any set f rings.
{a E Rlf(a)
= (4),
J:1
= (7).
23. Let R be an arbitrary ringo For each fixed element a E R, define the left-multiplication function T,,: R ..... R by taking T,,(x)
ax. If TR denotes th!; set of all such
funetions, prove the following:
a) T" is a (group) homomorphism of the additive group of R into itself;
b) TI! forms a rng, where multiplicaton is taken to be funetional eomposition;
e) He mappingf(a) = T" determines a homomorphism of R onto the ring TR ;
d) the kernel offis the ideal annlR;
.
e) iUor each O =1= a E R, there exists so me b E R such that ab =1= O, then R ~ TR
(In particular, part(e) holds whenever R has an identity eIement.)
24. Let R be an arbitrary ring and R x Z be the exten8ion rng eonstructed in Theorern
2-12. Establish that
a) R)( {O} ii an ideal of R x Z;
b) Z~ {O}xZ;
e) if a is an idempotent element of R, then the pair (-a, 1) is idempotent in R x Z,
while (a, O) is a zero divisor.
~---
38
+ (y,h) =
(x, a)(y, h)
THREE
lis
29. Let R be a ring with the prbperty that every subring of R is necessarily an ideal of
R. (The ring Z, for instan ce, enjoys this property.) If R contains no divisors of zero,
prove that muItiplication is commutative. [Hint: Given O =1= a e R, consider the
subring S generated bya. For arbitrary h eR, ab = r e S, so that ar = ra.]
+ il'ie l}.
If a = b + i is any member of b +
{b
38
+ (y,h) =
(x, a)(y, h)
THREE
lis
29. Let R be a ring with the prbperty that every subring of R is necessarily an ideal of
R. (The ring Z, for instan ce, enjoys this property.) If R contains no divisors of zero,
prove that muItiplication is commutative. [Hint: Given O =1= a e R, consider the
subring S generated bya. For arbitrary h eR, ab = r e S, so that ar = ra.]
+ il'ie l}.
If a = b + i is any member of b +
{b
40
Given an ideal 1 of the ring R, let us employ the symbol Rjl to denote
.
the collection of all cosets of 1 in R; that is,
R.jl
{a -: IlaER}.
(a
+ 1)
(a + l)(b + 1)
1)
(b
= (a
= db
+ b) +
+ 1.
1,
One is faced with the usual problem of showing that these operatio~s are
actually well-defined, so that thesum and product of the two cosets a + 1
and b + 1 do not depend on their particular representatives a and b. To
tbis end, suppose that
. ';'
o"
ci+l=.a'+l
"
.. '
and
.b
+ 1 = b' + l.
ab - a'b'
sincro both the products ai2and tI b' must be i~ 1. The implication, of course,
is that ab + 1 = a'b' + 1; hence, our definition of multiplication in Rj1 is
meaningful.
The verifiGatiqn that Rjl, under the operations defined abo ve, forms a
ring is easy and the details are left to the reader. To assure completeness,
we simply state
.
Theorem 3-3. If1 is an ideal of the ring R, then Rj1 is also a ring, known
as the quotient ring (or factor ring) of R by l.
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 1 be an ideal of the ring R. Then the mapping
nat f : R
--.
Proo! The fact that nat is a homomorphism follows directly from the
manner in which the operations aredefined in the quotient ring:
nat(a
The set Rjl can be endowed with the structure of a ring in a natural way;
all we need do is define addition and multiplication as follows:
+1
41
+ b)
+ b + 1 = (a + 1) + (b +
+ nat(b);
ab + 1 = (a + l)(b + 1)
a
= nat'(a)
nat(ab) =
=
1)
'""
nat(a) nat(b).
That nat carries R ont~ Rjl is all but obvious; indeed, every element of
Rjl is a coset a + 1, with a E R, and so by definition nat(a) = a + 1.
Inasmch as the coset 1 =0 + 1 serves as the zero element for the
ring Rj1, we necessarily have
ker (nat) ~. {a
.
= . {a
E
E
Rlnat(a) = l}
Rla+ 1 = l}
1.
The
last eq;~ality
was achieved by invoking Theorem 3-2.__" ":
.
.. :.
It is customary to speak of the function nat, which:~ssigns to each
element of R the coset in Rjl of which it is the representativ,as the natural,
or canonical, mapping of R onto the quotient ring Rj1. When there is no
danger of confusion, we shall omit the subscript 1 in writing this mapping.
There are two standard techniques for investigating the structure of a
particular ringo 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 deternrine all homomorphisms from the given ring
hito a simpler ring; the idea here is that the homomorphi'c images will tend
to reflect sorne of the algebraic properties of the original ringo (The reader
is warned to proceed with sorne 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 the'se .
are the same; every ideal determines a homomorphism, and every homomorphism determines an ideal.
+ (n)
{a
+ lenlle 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 congruence classes in Z" can now be vie:-ved as coset operations in Zj(n):
(a + (n)) + (b + (n))
(a + (n))(b + (n))
=
=
a + b + (n), _
ab + (n).'
40
Given an ideal 1 of the ring R, let us employ the symbol Rjl to denote
.
the collection of all cosets of 1 in R; that is,
R.jl
{a -: IlaER}.
(a
+ 1)
(a + l)(b + 1)
1)
(b
= (a
= db
+ b) +
+ 1.
1,
One is faced with the usual problem of showing that these operatio~s are
actually well-defined, so that thesum and product of the two cosets a + 1
and b + 1 do not depend on their particular representatives a and b. To
tbis end, suppose that
. ';'
o"
ci+l=.a'+l
"
.. '
and
.b
+ 1 = b' + l.
ab - a'b'
sincro both the products ai2and tI b' must be i~ 1. The implication, of course,
is that ab + 1 = a'b' + 1; hence, our definition of multiplication in Rj1 is
meaningful.
The verifiGatiqn that Rjl, under the operations defined abo ve, forms a
ring is easy and the details are left to the reader. To assure completeness,
we simply state
.
Theorem 3-3. If1 is an ideal of the ring R, then Rj1 is also a ring, known
as the quotient ring (or factor ring) of R by l.
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 1 be an ideal of the ring R. Then the mapping
nat f : R
--.
Proo! The fact that nat is a homomorphism follows directly from the
manner in which the operations aredefined in the quotient ring:
nat(a
The set Rjl can be endowed with the structure of a ring in a natural way;
all we need do is define addition and multiplication as follows:
+1
41
+ b)
+ b + 1 = (a + 1) + (b +
+ nat(b);
ab + 1 = (a + l)(b + 1)
a
= nat'(a)
nat(ab) =
=
1)
'""
nat(a) nat(b).
That nat carries R ont~ Rjl is all but obvious; indeed, every element of
Rjl is a coset a + 1, with a E R, and so by definition nat(a) = a + 1.
Inasmch as the coset 1 =0 + 1 serves as the zero element for the
ring Rj1, we necessarily have
ker (nat) ~. {a
.
= . {a
E
E
Rlnat(a) = l}
Rla+ 1 = l}
1.
The
last eq;~ality
was achieved by invoking Theorem 3-2.__" ":
.
.. :.
It is customary to speak of the function nat, which:~ssigns to each
element of R the coset in Rjl of which it is the representativ,as the natural,
or canonical, mapping of R onto the quotient ring Rj1. When there is no
danger of confusion, we shall omit the subscript 1 in writing this mapping.
There are two standard techniques for investigating the structure of a
particular ringo 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 deternrine all homomorphisms from the given ring
hito a simpler ring; the idea here is that the homomorphi'c images will tend
to reflect sorne of the algebraic properties of the original ringo (The reader
is warned to proceed with sorne 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 the'se .
are the same; every ideal determines a homomorphism, and every homomorphism determines an ideal.
+ (n)
{a
+ lenlle 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 congruence classes in Z" can now be vie:-ved as coset operations in Zj(n):
(a + (n)) + (b + (n))
(a + (n))(b + (n))
=
=
a + b + (n), _
ab + (n).'
42
--+
by taking
l(a
l)
f(a)
(a E R).
f(a - b
b)
f(a - b)
+ f(b) =
f(b)
and, by the manner in which 1 was defined, that l(a + l) = J(b + l).
Hence, the functionlis constant on the cosets of l, as we wished to demonstrate.
43
l)
(b
l)) = J(a
= f(a)
+ b + l) ~ f(a + b) _
+ f(b) = f(a + l) + f(b +
l);
and, likewse,
l((a
l)(b
l))
= J(ab +
1) ~f(ab) _
= f(a)f(b)
= f(a +
l)f(b
l).
whence the equality f = lo nat. It only remans to show that ths factorzation s unque. Suppose also thatf = g o nat for sorne other function
g: Rjl --+ R'. But then
l(a
1)
f(a)
= (g o nat Ha) =
g(a
l)
for all a in R, and so g = J. The nduced mappingls thus the only functon
from the quotient rng Rjl into R' satisfyng the equaton f = Jo nat.
Corollary. The induced mapping 1 s an isomorphsm f and only f
kerf f; l.
Proa! What is requred here s an explicit description of the kernel of J,
to wit
+ l~(a + l) = O}
= {a + l f(a) = O}
= {a .+ l a E ker f} = nat(ker f).
ker 1 = {a
1 to
-E~tl\
11
Rjl
42
--+
by taking
l(a
l)
f(a)
(a E R).
f(a - b
b)
f(a - b)
+ f(b) =
f(b)
and, by the manner in which 1 was defined, that l(a + l) = J(b + l).
Hence, the functionlis constant on the cosets of l, as we wished to demonstrate.
43
l)
(b
l)) = J(a
= f(a)
+ b + l) ~ f(a + b) _
+ f(b) = f(a + l) + f(b +
l);
and, likewse,
l((a
l)(b
l))
= J(ab +
1) ~f(ab) _
= f(a)f(b)
= f(a +
l)f(b
l).
whence the equality f = lo nat. It only remans to show that ths factorzation s unque. Suppose also thatf = g o nat for sorne other function
g: Rjl --+ R'. But then
l(a
1)
f(a)
= (g o nat Ha) =
g(a
l)
for all a in R, and so g = J. The nduced mappingls thus the only functon
from the quotient rng Rjl into R' satisfyng the equaton f = Jo nat.
Corollary. The induced mapping 1 s an isomorphsm f and only f
kerf f; l.
Proa! What is requred here s an explicit description of the kernel of J,
to wit
+ l~(a + l) = O}
= {a + l f(a) = O}
= {a .+ l a E ker f} = nat(ker f).
ker 1 = {a
1 to
-E~tl\
11
Rjl
,"1""
......
'.
j
44
-f(-n)
-(-n)f(1)
nf(1).
Proo.f Suppose that the rings R and Z are somorphic under two functions
f, g: R -> Z. Then the composition .f o g - 1 is a homomorphic mapping
from the ring Z onto itself. Knowing this, the emma just pro ved implies
thatfog-l = iz , orf= g.
We now ha ve the necessary information to prove the following result.
Theorem 3-8. Any homomorphism from an arbitrary ring R onto the
ring Z ofintegers is uniquely determined by its kernel.
45
another, the composition of mappings along these paths produces the same
function.)
A rather simple observation, with far-reaching implications, is that
whenever 1 = ker f, so that both the Factorization Theorem and its Corollary are applicable, f induces a mapping J under which Rjl and R' are
isomorphic rings. We surnmarize all this in the following theorem, a result
which ~ll be invokedon many occasions in the sequel.
f(n) =f(-(-n))
l'
g =
g o natkerg
. .~ r
The next two theorems are somewhat deeper result.s than usual and
require the full force of our accumulated machinery. They c0mprise what
are often called the First and Second Isomorphism Theorems and have
important applications in the sequel. (The reader is cationed' that there
seems to be no universally accepted numbering for these,thecirems.)
Theorem 3-9. Letfbe ahomomorphism of the rih~ R'cinto th~ ring
R' and let 1 be an ideal of R. If ker f S;; 1, then Rjl ~ R' j f(1).
Proof. Before becoming in volved 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'j f(1).
Let us now define the function g: R -> R'j f(1) by g = natJ(l) of, where
nat J(l): R' -> R'j f(1) is the usual natural mapping. Thus, g merely assigns
to each element a E R the coset fea) + f(1) in R'j f(1). Since the functions
f and natJ(l) are both onto homomorphisms, their composition carries R
homomorphically onto the quotient ring R' j f(1).
The crux of the argument is to show thatker g = 1, for then the desired
conclusion would be an immediate consequence of the Fundamental
Homomorphism Theorem. Sin ce the zero element of R' j f(I) is just the
coset f(I), the kernel of g consists of those members of R which are mapped
.
by g onto f(1):
ker g = {a E Rlg(a)
=
=
{a
{a
E
E
= f(l)}
Rlf(a) + f(1) = f(1)}
Rlf(a) Ef(/)} = f-l(f(1)).
,"1""
......
'.
j
44
-f(-n)
-(-n)f(1)
nf(1).
Proo.f Suppose that the rings R and Z are somorphic under two functions
f, g: R -> Z. Then the composition .f o g - 1 is a homomorphic mapping
from the ring Z onto itself. Knowing this, the emma just pro ved implies
thatfog-l = iz , orf= g.
We now ha ve the necessary information to prove the following result.
Theorem 3-8. Any homomorphism from an arbitrary ring R onto the
ring Z ofintegers is uniquely determined by its kernel.
45
another, the composition of mappings along these paths produces the same
function.)
A rather simple observation, with far-reaching implications, is that
whenever 1 = ker f, so that both the Factorization Theorem and its Corollary are applicable, f induces a mapping J under which Rjl and R' are
isomorphic rings. We surnmarize all this in the following theorem, a result
which ~ll be invokedon many occasions in the sequel.
f(n) =f(-(-n))
l'
g =
g o natkerg
. .~ r
The next two theorems are somewhat deeper result.s than usual and
require the full force of our accumulated machinery. They c0mprise what
are often called the First and Second Isomorphism Theorems and have
important applications in the sequel. (The reader is cationed' that there
seems to be no universally accepted numbering for these,thecirems.)
Theorem 3-9. Letfbe ahomomorphism of the rih~ R'cinto th~ ring
R' and let 1 be an ideal of R. If ker f S;; 1, then Rjl ~ R' j f(1).
Proof. Before becoming in volved 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'j f(1).
Let us now define the function g: R -> R'j f(1) by g = natJ(l) of, where
nat J(l): R' -> R'j f(1) is the usual natural mapping. Thus, g merely assigns
to each element a E R the coset fea) + f(1) in R'j f(1). Since the functions
f and natJ(l) are both onto homomorphisms, their composition carries R
homomorphically onto the quotient ring R' j f(1).
The crux of the argument is to show thatker g = 1, for then the desired
conclusion would be an immediate consequence of the Fundamental
Homomorphism Theorem. Sin ce the zero element of R' j f(I) is just the
coset f(I), the kernel of g consists of those members of R which are mapped
.
by g onto f(1):
ker g = {a E Rlg(a)
=
=
{a
{a
E
E
= f(l)}
Rlf(a) + f(1) = f(1)}
Rlf(a) Ef(/)} = f-l(f(1)).
46
R'/1'.
S;
l. Then
~R/l
~~R/J
1 ~ (R/J)/(l/J)
1'
nat'/l
natjJ o natJ'
1,
I~
(I
,1
+J
/nat,
J)jJ
=
=
{a
{a
ll f (a) = J}
1a
J}
{a
lla E J}
1 \ J.
47
We conclude tbis chapter with a brief excursion into the theory of nil
and nilpotent ideals: a (right, left, two-sided) ideal 1 of the ring R is said to
be a nil ideal if each element x in 1 is nilpotent; that is to say, if there exists
a positive integer n for which x!' = O, where n depends upon the particular
element x. As one might expect, the ideal 1 will be termed nilpotent provided
1" = {O} for sorne positive integer n. By definition, In denotes the set of all
finite sums of products of n elements taken from 1, so that 1" = {O} is
equivalent too requiring that for every choice of n elements al' a 2 , ... , a n E 1
(distinct or not), the product a 1a Z '" an = O; in particular, a n = O for all
a in 1, whence every nilpotent ideal is autornatically a nil ideal. We speak
of the ring R as being nil (ni!potent) if it is ni! (ni!potent) 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 le,?mas.
Lemma. 1) If R is a ni! (nilpotent) ring, then every subring and every
homomorphic image of R is ni! (nilpotent).
2) If R contains an ideal 1 such that 1 and R/l are both ni! (ni!potent),
then R is a pil (ni!potent) ringo
Proof The proof of assertion (1) follows irnmediately from the definitions
and Problem i-17.' To verify(2), assume that 1 and R/l are nil rings and
that a E R. Then there exists sorne positive integer n for which the coset
(a
l)n = a n
+1
= 1,
signifying that the lement a n E l. Inasrnuch as lis a nil ideal, (an)m = a nnl = O
for sorne m E Z+. This implies that a is ni!potent as a member of R and, in
consequence, R is a ni! ringo The remainder of the proof is left to the
reader's careo
Lemma. If NI and N 2 are two nil (nilpotent) ideals of the ring R, then
their sum NI + N z is likewise a ni! (nilpotent) ideal.
Proof With reference to Theorem 3-10, we have (NI + Nz)/N 1 ~
N z/(N 1 \ N 2)' The right-hand side (hence, the left-hand side) of this
equation is a nil ring, being the hornomorphic image of the nil ideal N 2'
46
R'/1'.
S;
l. Then
~R/l
~~R/J
1 ~ (R/J)/(l/J)
1'
nat'/l
natjJ o natJ'
1,
I~
(I
,1
+J
/nat,
J)jJ
=
=
{a
{a
ll f (a) = J}
1a
J}
{a
lla E J}
1 \ J.
47
We conclude tbis chapter with a brief excursion into the theory of nil
and nilpotent ideals: a (right, left, two-sided) ideal 1 of the ring R is said to
be a nil ideal if each element x in 1 is nilpotent; that is to say, if there exists
a positive integer n for which x!' = O, where n depends upon the particular
element x. As one might expect, the ideal 1 will be termed nilpotent provided
1" = {O} for sorne positive integer n. By definition, In denotes the set of all
finite sums of products of n elements taken from 1, so that 1" = {O} is
equivalent too requiring that for every choice of n elements al' a 2 , ... , a n E 1
(distinct or not), the product a 1a Z '" an = O; in particular, a n = O for all
a in 1, whence every nilpotent ideal is autornatically a nil ideal. We speak
of the ring R as being nil (ni!potent) if it is ni! (ni!potent) 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 le,?mas.
Lemma. 1) If R is a ni! (nilpotent) ring, then every subring and every
homomorphic image of R is ni! (nilpotent).
2) If R contains an ideal 1 such that 1 and R/l are both ni! (ni!potent),
then R is a pil (ni!potent) ringo
Proof The proof of assertion (1) follows irnmediately from the definitions
and Problem i-17.' To verify(2), assume that 1 and R/l are nil rings and
that a E R. Then there exists sorne positive integer n for which the coset
(a
l)n = a n
+1
= 1,
signifying that the lement a n E l. Inasrnuch as lis a nil ideal, (an)m = a nnl = O
for sorne m E Z+. This implies that a is ni!potent as a member of R and, in
consequence, R is a ni! ringo The remainder of the proof is left to the
reader's careo
Lemma. If NI and N 2 are two nil (nilpotent) ideals of the ring R, then
their sum NI + N z is likewise a ni! (nilpotent) ideal.
Proof With reference to Theorem 3-10, we have (NI + Nz)/N 1 ~
N z/(N 1 \ N 2)' The right-hand side (hence, the left-hand side) of this
equation is a nil ring, being the hornomorphic image of the nil ideal N 2'
PROBLEMS
48
49
Since (N I +, N 2)/N I and NI are both nil, it follows from the previous lemma
that NI + N 2 is necessarily a nil ideal. Similar reasoning applies to the
nilpotent case.
Corollary. The sum of any finite number of ni] (nilpotent) ideals of the
.
ring R is again nil (nilpotent).
.after a certain, but not fixed, point. One may easily check that R constitut
a subring ?f
ringo S (in fact, R is not only a subring, but actually an
of S): It IS III the nng R that we propose to construct our example of
non-nilpotent nil ideal.
'
a
Le~ u~ den.ote b~ J the set o(sequences in R whose nth term belo'ngs to
the pnnclpal Ideal III Zpn generated by p; in other words the sequence
a E J if and only if it is of the form
.
,
~he
ide:~
Proa! If the element a E I Ni' then, by definition; a lies in some finite sum
of nil ideal s of R; say, a E NI + N 2 + .... +. N n , wh<::re each N k is nil.
By virtue of the last Gorollary, the sum N f' + N 2 + ... + N n must be a
nil ideal; hence, the element a is nilpotent. This argument shows that I Ni
is a nil ideal.
It is possible to deduce somewhat mor; na~mely,
Corollary. The sum of all the nilpotent ideals of the ring R is a nil
ideal.
Proa!. Since each nilpotent ideal is a nil ideal, thesum N of all nilpotent
i is itse1f
ideals of R is contained in I Ni' the sum of all nil ideals. But
a nil ideal, making N nil.
IN
(r k E Zpk).
O = (O, O, O, ... )
for n large enough, .maki~g J a nil, ideal. (This also depends on the fact
that a has only a fimte number of nonzero terms.)
."
. At the present stage, it is still conceivable that 1 mighLbe a nilpotent
Id~al of R. I:Iowever, we:can show that for each posit:ive.iiteger n there
eXIst elements (sequen ces) a E 1 for which' an f O. For instance defi
b
k'
.
'
ne
a = .{ ak } Y ta lllg ak = P If k = 1, 2, .. , ,n + 1 and ak = O if k > n + l'
~~
(p, ... , p, p, O, : .. )
with
+ 1 p's.
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 exacdy one ideal for each positive divisor of pn and no other
ideals; these are simply the principal ideal s (l) = lz pn (O :::; k :::; n). For
O < k :::; n, we ha ve
'
where all the terms are zero except the (n + l)st, which is pn. Sin ce pn is a
nnonzero elemento of the ring Zpn+l, the sequence an f O, implying that
J. f {O}. As thls argument holds for any n E Z+, the ideal J carinot be
mlpotent.
W,e sh.all retu.rn. to these ide~s at the appropriate place in the sequel,
at whlch tIme thelr Importance wI1l become dear.
{a,,} + {b,,}
{a n + bn},
{a,,}{b,,}
{a"b,,}.
The reader will find that the zero e1ement of this ring is just the sequence
formed by the zero elements af the various Zpn and the negative of {a n}
is { - a,,}. Now, consider the set R of all sequences in S which become zero
PROBLEMS
PROBLEMS
48
49
Since (N I +, N 2)/N I and NI are both nil, it follows from the previous lemma
that NI + N 2 is necessarily a nil ideal. Similar reasoning applies to the
nilpotent case.
Corollary. The sum of any finite number of ni] (nilpotent) ideals of the
.
ring R is again nil (nilpotent).
.after a certain, but not fixed, point. One may easily check that R constitut
a subring ?f
ringo S (in fact, R is not only a subring, but actually an
of S): It IS III the nng R that we propose to construct our example of
non-nilpotent nil ideal.
'
a
Le~ u~ den.ote b~ J the set o(sequences in R whose nth term belo'ngs to
the pnnclpal Ideal III Zpn generated by p; in other words the sequence
a E J if and only if it is of the form
.
,
~he
ide:~
Proa! If the element a E I Ni' then, by definition; a lies in some finite sum
of nil ideal s of R; say, a E NI + N 2 + .... +. N n , wh<::re each N k is nil.
By virtue of the last Gorollary, the sum N f' + N 2 + ... + N n must be a
nil ideal; hence, the element a is nilpotent. This argument shows that I Ni
is a nil ideal.
It is possible to deduce somewhat mor; na~mely,
Corollary. The sum of all the nilpotent ideals of the ring R is a nil
ideal.
Proa!. Since each nilpotent ideal is a nil ideal, thesum N of all nilpotent
i is itse1f
ideals of R is contained in I Ni' the sum of all nil ideals. But
a nil ideal, making N nil.
IN
(r k E Zpk).
O = (O, O, O, ... )
for n large enough, .maki~g J a nil, ideal. (This also depends on the fact
that a has only a fimte number of nonzero terms.)
."
. At the present stage, it is still conceivable that 1 mighLbe a nilpotent
Id~al of R. I:Iowever, we:can show that for each posit:ive.iiteger n there
eXIst elements (sequen ces) a E 1 for which' an f O. For instance defi
b
k'
.
'
ne
a = .{ ak } Y ta lllg ak = P If k = 1, 2, .. , ,n + 1 and ak = O if k > n + l'
~~
(p, ... , p, p, O, : .. )
with
+ 1 p's.
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 exacdy one ideal for each positive divisor of pn and no other
ideals; these are simply the principal ideal s (l) = lz pn (O :::; k :::; n). For
O < k :::; n, we ha ve
'
where all the terms are zero except the (n + l)st, which is pn. Sin ce pn is a
nnonzero elemento of the ring Zpn+l, the sequence an f O, implying that
J. f {O}. As thls argument holds for any n E Z+, the ideal J carinot be
mlpotent.
W,e sh.all retu.rn. to these ide~s at the appropriate place in the sequel,
at whlch tIme thelr Importance wI1l become dear.
{a,,} + {b,,}
{a n + bn},
{a,,}{b,,}
{a"b,,}.
The reader will find that the zero e1ement of this ring is just the sequence
formed by the zero elements af the various Zpn and the negative of {a n}
is { - a,,}. Now, consider the set R of all sequences in S which become zero
PROBLEMS
50
R/l.
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 d' = bm = O for integers n and In,
consider (a - b)"h.J
b) The quotient ring R/N has no nonzero nilpotent elements.
PROBLEMS
51
a) [R, RJ S; kerf;
b) f = lo nat(R RJ, where]is the induced mapping;
c) if kerf s; [R, RJ, then R/[R, RJ "" R'/[R', R'J.
12. a) Suppose that l1 and 12 are ideills of the ring R for which R = 1 1 EB 1 2 Prove
that R/l1 "" 12, and R/12 "" 1 1.
b) Let R be the direct sum of the rings R (i = 1,2, ... , n). If 1 is an ideal of R
and 1 = 1 1 EB 12 EB ... EB 1", show that
f(a)
+ f(l).
R
nat1
LR'=f(R)
1
R/l
nat f(I)
R'If(l)
Zm EB Zn'
15. If 1 is an ideal of the ring R, prove that the matrix ring M n (R/l) is isomorphic to
Mn(R)fMn(I). [Hint: Consider the mapping f: Mn(R) -+ M n(R/1) defined by
f(a)) = (a) + l).J
.
16. Let R be a ring without divisors of zero. Imbed E. 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 does not.) Assuming that 1 denotes the left annihilator
of R in R',
1 = {a ER'\ar = O for all r ER},
verify that
a) l forms an ideal of R'. [Hint: R is an ideal of R'.J
b) R'/l is a ring with identity which has no divisors of zero.
c) R'/l contains a subring isomorphic to R. [Hint: Utilize Problem
9.J
50
R/l.
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 d' = bm = O for integers n and In,
consider (a - b)"h.J
b) The quotient ring R/N has no nonzero nilpotent elements.
PROBLEMS
51
a) [R, RJ S; kerf;
b) f = lo nat(R RJ, where]is the induced mapping;
c) if kerf s; [R, RJ, then R/[R, RJ "" R'/[R', R'J.
12. a) Suppose that l1 and 12 are ideills of the ring R for which R = 1 1 EB 1 2 Prove
that R/l1 "" 12, and R/12 "" 1 1.
b) Let R be the direct sum of the rings R (i = 1,2, ... , n). If 1 is an ideal of R
and 1 = 1 1 EB 12 EB ... EB 1", show that
f(a)
+ f(l).
R
nat1
LR'=f(R)
1
R/l
nat f(I)
R'If(l)
Zm EB Zn'
15. If 1 is an ideal of the ring R, prove that the matrix ring M n (R/l) is isomorphic to
Mn(R)fMn(I). [Hint: Consider the mapping f: Mn(R) -+ M n(R/1) defined by
f(a)) = (a) + l).J
.
16. Let R be a ring without divisors of zero. Imbed E. 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 does not.) Assuming that 1 denotes the left annihilator
of R in R',
1 = {a ER'\ar = O for all r ER},
verify that
a) l forms an ideal of R'. [Hint: R is an ideal of R'.J
b) R'/l is a ring with identity which has no divisors of zero.
c) R'/l contains a subring isomorphic to R. [Hint: Utilize Problem
9.J
53
FOUR
+ (e, d) = (a + e, b . + d),
(a, b)(~, d)' = (ae - bd, ad + be).
(a, b)
~he re~der may verify without difficulty that e, together with these operabons, IS a commutative ring with identity. In this setting, the pair (1 O)
se~ves as the multiplicative identity, and (0, O) is the zero element ofthe ri~g.
Glven any nonzero element (a, b) of e, either a =1= O or b =1= 0, so that
a 2 + b 2 > O; thus,
.
exists in
(a b){
i'
!
:".
. i
, \a
b2 '
.
2
\a
This shows that each nonzero member of e has an inverse 'under multi- .
plication, thereby proving the system e to be a field.
It is worth pointing out that the field e contains a subring isomorphic
to the field ofreal numbers .. For, if
.
R# x {O} = {Ca, O)la E R#},
i
1
!
11
,1
it follows that R # .~. R ~ x {O} via the mapping f defIned by fea) = (a, O).
Inasmuch as the dlsbnctlOn between these systems is only one of notation
we customarily identify the real lUmber a with the corresponding ordered
pair (a, O); in this sense, R# may be regarded as a subring of C..
Now, the definitin of the operations in e enables us to express an
. arbitrary element (a, b) E e as
. . . . ......
(a, b) = (a, O)
. where the pair (O, 1) is such that (O, 1)2 = (O, 1)(0, 1)
the symbol i as an abbreviation for (O, 1), we have
(a, b)
(a, O)
(b, O).
Finally, if it is agreed to replace pairs of the form (a, O) by the first component a ~this is justified by the precedng paragraph), the dsplayed
representatlOn becomes
(a, b) = a
bi,
with
2 = -1.
I~ ot~er wor~s, the field e as defined initiaIly is' nothing more than a
. dlsgUlsed verSlOn ofthe fmilar complex number system.
53
FOUR
+ (e, d) = (a + e, b . + d),
(a, b)(~, d)' = (ae - bd, ad + be).
(a, b)
~he re~der may verify without difficulty that e, together with these operabons, IS a commutative ring with identity. In this setting, the pair (1 O)
se~ves as the multiplicative identity, and (0, O) is the zero element ofthe ri~g.
Glven any nonzero element (a, b) of e, either a =1= O or b =1= 0, so that
a 2 + b 2 > O; thus,
.
exists in
(a b){
i'
!
:".
. i
, \a
b2 '
.
2
\a
This shows that each nonzero member of e has an inverse 'under multi- .
plication, thereby proving the system e to be a field.
It is worth pointing out that the field e contains a subring isomorphic
to the field ofreal numbers .. For, if
.
R# x {O} = {Ca, O)la E R#},
i
1
!
11
,1
it follows that R # .~. R ~ x {O} via the mapping f defIned by fea) = (a, O).
Inasmuch as the dlsbnctlOn between these systems is only one of notation
we customarily identify the real lUmber a with the corresponding ordered
pair (a, O); in this sense, R# may be regarded as a subring of C..
Now, the definitin of the operations in e enables us to express an
. arbitrary element (a, b) E e as
. . . . ......
(a, b) = (a, O)
. where the pair (O, 1) is such that (O, 1)2 = (O, 1)(0, 1)
the symbol i as an abbreviation for (O, 1), we have
(a, b)
(a, O)
(b, O).
Finally, if it is agreed to replace pairs of the form (a, O) by the first component a ~this is justified by the precedng paragraph), the dsplayed
representatlOn becomes
(a, b) = a
bi,
with
2 = -1.
I~ ot~er wor~s, the field e as defined initiaIly is' nothing more than a
. dlsgUlsed verSlOn ofthe fmilar complex number system.
54
= (0,0,0,1).
i2 =
ij = le, jk = i, k = j, ji
k2
-1,
= - k,
lej = - i,
ik
al - (bi
cj
+ d/c)2 =
a2
+ b2 +
e2
+ d2 =t=
0,
+ b2 +
c2
+ d2)-lq.
Proof. Since every field is a commutative ring with identity, we need only
prove that F .contains no zero divisors. To thls purpose, suppose a, b E F,
with ab
O. If the element a =t= 0, then it must possess a multiplicative
inverse a -1 E F. But then the hypothesis that ab = yields
q-1 = (a 2
= (0,0,0, 1, O),
ijq
55
a- 10 = a-1(ab) = lb
= b,
as desired.
There obviously exist integral domains which are not fields; a prime
example is the ring Z ofintegers. liowever, an integral dornain having only
a finite ilUmber of elements must necessarily be a field.
TheO'em 4-2. Any integral domain R with only a finite number of ideals
is a field.
(a n)
{r d'lr E R}.
54
= (0,0,0,1).
i2 =
ij = le, jk = i, k = j, ji
k2
-1,
= - k,
lej = - i,
ik
al - (bi
cj
+ d/c)2 =
a2
+ b2 +
e2
+ d2 =t=
0,
+ b2 +
c2
+ d2)-lq.
Proof. Since every field is a commutative ring with identity, we need only
prove that F .contains no zero divisors. To thls purpose, suppose a, b E F,
with ab
O. If the element a =t= 0, then it must possess a multiplicative
inverse a -1 E F. But then the hypothesis that ab = yields
q-1 = (a 2
= (0,0,0, 1, O),
ijq
55
a- 10 = a-1(ab) = lb
= b,
as desired.
There obviously exist integral domains which are not fields; a prime
example is the ring Z ofintegers. liowever, an integral dornain having only
a finite ilUmber of elements must necessarily be a field.
TheO'em 4-2. Any integral domain R with only a finite number of ideals
is a field.
(a n)
{r d'lr E R}.
56
[lJ
[al'
+ nsJ =
[arJ
+"
= [arJ
+"
57
l'
and s such
[nsJ
[OJ = [aJ ." [rJ,
[OJ,
n},
This leads at once to a c1assical result of Euler concerning the phifunction; the simplicity of the argument illustrates the advantage of the
algebraic approach to number theory.
56
[lJ
[al'
+ nsJ =
[arJ
+"
= [arJ
+"
57
l'
and s such
[nsJ
[OJ = [aJ ." [rJ,
[OJ,
n},
This leads at once to a c1assical result of Euler concerning the phifunction; the simplicity of the argument illustrates the advantage of the
algebraic approach to number theory.
58
Proof. The congruence c1ass [a] can be viewed as an eIernent of the rnultiplicative group (G m ',,). Since this group has order </len), it foIlows that
[a]"'(II) = [1] or, equivalentIy, a"'(II)
1 (mod. n). (Recall tIlat if Gis a finite
group of order le, then Xk = 1 for aIl x E G.)
{O} and R are its only ideals. Let us suppose to the cOritrary that there
exists sorne nontrivll ideal 1 of R. By our assurnption, the subset 1 is such
that 1 =1= {O} and 1 =1= R. This mean s that there exists sorne nonzero elernent
a E l. Since R is taken to be a fieId, a has a rnultiplicative in verse a- 1 present
in R. By the definition ofideal, we thus obtain 1 = a- l a E 1, which in turn
implies that 1 = R, contradicting our choice of l.
Conversely, suppose that the ring R has no nontrivial ideals. Given a
nonzero eIernent a E R, consider the principal ideal (a) generated by a:
(a)
{ralr
R}.
Now, (a) cnnot be the zero ideal, inasrnuch as a = a 1 E (a), with a =1= O.
It foIlows frorn the hypothesis that the only other possibility is that (a) = R.
In particular, since 1 E (a),there exists an elernent r E:R for whichr a = l.
MuItiplication is cornrnutative, so that r = a-l. T~e;~fore, each nonzero
elernent of R is rnultiplicativeIy in vertible and we are done.
In view of this last result, the ring Z of integers faiIs to be a field sin ce
it contains the nontrivial ideal Ze"
'
Theorern 4-6 is useful in revealing the nature of hornornorphisrns
between fieIds. We exploit it to prove
i
'1
!I i
,
1
II!
11
Proof. The proof consists of noticing that since ker fis an idea! of the field
F, either ker f = {O} or else ker f = F. The condition ker f = {O} irnplies
that f is a one-to-one function. On the other hand, if it happens that
kerf = F, then each elernent of F is carried onto O; that is to say,Jis the
trivial hornornorphisrn.
59
{ab-lla,bER;b =1=
O},
58
Proof. The congruence c1ass [a] can be viewed as an eIernent of the rnultiplicative group (G m ',,). Since this group has order </len), it foIlows that
[a]"'(II) = [1] or, equivalentIy, a"'(II)
1 (mod. n). (Recall tIlat if Gis a finite
group of order le, then Xk = 1 for aIl x E G.)
{O} and R are its only ideals. Let us suppose to the cOritrary that there
exists sorne nontrivll ideal 1 of R. By our assurnption, the subset 1 is such
that 1 =1= {O} and 1 =1= R. This mean s that there exists sorne nonzero elernent
a E l. Since R is taken to be a fieId, a has a rnultiplicative in verse a- 1 present
in R. By the definition ofideal, we thus obtain 1 = a- l a E 1, which in turn
implies that 1 = R, contradicting our choice of l.
Conversely, suppose that the ring R has no nontrivial ideals. Given a
nonzero eIernent a E R, consider the principal ideal (a) generated by a:
(a)
{ralr
R}.
Now, (a) cnnot be the zero ideal, inasrnuch as a = a 1 E (a), with a =1= O.
It foIlows frorn the hypothesis that the only other possibility is that (a) = R.
In particular, since 1 E (a),there exists an elernent r E:R for whichr a = l.
MuItiplication is cornrnutative, so that r = a-l. T~e;~fore, each nonzero
elernent of R is rnultiplicativeIy in vertible and we are done.
In view of this last result, the ring Z of integers faiIs to be a field sin ce
it contains the nontrivial ideal Ze"
'
Theorern 4-6 is useful in revealing the nature of hornornorphisrns
between fieIds. We exploit it to prove
i
'1
!I i
,
1
II!
11
Proof. The proof consists of noticing that since ker fis an idea! of the field
F, either ker f = {O} or else ker f = F. The condition ker f = {O} irnplies
that f is a one-to-one function. On the other hand, if it happens that
kerf = F, then each elernent of F is carried onto O; that is to say,Jis the
trivial hornornorphisrn.
59
{ab-lla,bER;b =1=
O},
60
y = (ad
xy-l
i=
O, d
i=
O.
A simple
i=
= (ad)(eb)-l E F'.
~~"
a = al,
al- 1 E F '
for each a in!R, implying that R F ' . FinaHy; any subfield of F whiCh
contruns R necssarily'includes all products ab-l, with a, O i= bE R, and,
hence, contais<F' .
Theorem':4l:-S began wth an integral domain airead y imbedded in a:
field. In the geheral 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 i= O. Our thinking is that (a, b) will play
a role analogous to ah- 1 in fue foregoing theorem.
ActuaIly, the proposed constr.uction ls not just confined to integral
domains, but wilI apply to a much wider c1ass oC ring~; it will imbed any
commutative ring R that contaius a (nonempty) set of elements that are
not zero divisors in a ring Qcl(R), which may be described as folIows.
Definition 4-3. Let R be a ring with at least one non-zero-divisor. A
RxS
'
be)(bd)-lEF'.
61
= be.}'
ar
sb,
bt
re.
'o-'!.'
Now multiply the first equation by t and the second by s '1.0 get
art
',;: -;-
= sbt,
sbt
sre.
a/s
= {(b, r) ar
The coIlection of all equivaIence cIasses a/s relative to '" will be denoted
'by Qc(R):
Q.,(R). = {a/sla
R; s E S}.
a/s +, b/r
(a/s)(b/r)
=
=
(ar + sb)/sr,
ab/sr.
60
y = (ad
xy-l
i=
O, d
i=
O.
A simple
i=
= (ad)(eb)-l E F'.
~~"
a = al,
al- 1 E F '
for each a in!R, implying that R F ' . FinaHy; any subfield of F whiCh
contruns R necssarily'includes all products ab-l, with a, O i= bE R, and,
hence, contais<F' .
Theorem':4l:-S began wth an integral domain airead y imbedded in a:
field. In the geheral 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 i= O. Our thinking is that (a, b) will play
a role analogous to ah- 1 in fue foregoing theorem.
ActuaIly, the proposed constr.uction ls not just confined to integral
domains, but wilI apply to a much wider c1ass oC ring~; it will imbed any
commutative ring R that contaius a (nonempty) set of elements that are
not zero divisors in a ring Qcl(R), which may be described as folIows.
Definition 4-3. Let R be a ring with at least one non-zero-divisor. A
RxS
'
be)(bd)-lEF'.
61
= be.}'
ar
sb,
bt
re.
'o-'!.'
Now multiply the first equation by t and the second by s '1.0 get
art
',;: -;-
= sbt,
sbt
sre.
a/s
= {(b, r) ar
The coIlection of all equivaIence cIasses a/s relative to '" will be denoted
'by Qc(R):
Q.,(R). = {a/sla
R; s E S}.
a/s +, b/r
(a/s)(b/r)
=
=
(ar + sb)/sr,
ab/sr.
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
(dr' + sb')/s'r'.
as'
= sa',
br' = rb'.
=
=
(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 Qc(R) under these
operations.
Lemma. The syst~m Qc(R) forms a commutative ring with identity.
for arbitrary a/s in Qc(R), since (as')s = (ss')a. Loosely speaking, common
factors belonging to S may be .canceIled in a fracti<?n as/ss.
This prov.es part of the theorem below.
Theorem 4-9. Any comrnutative ring R with at least one non-zerodivisor possesses a classical ring of quotients.
63
The reader can easily check that K is a subring of Qc(R). An obvious (onto)
mappingf: R -+ K is defined by takingf(a) = aso/so. Since the condition
aso/so = 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 preserving both addition and multiplication:
(sso/so)(so/sso)
ss~/ss~
= so/so.
Since so/so plays the role of the identity element for Qc(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 Q.(R) can be written as as-l. It should be clear that
a/s
(aso/so)(so/sso)
(aso/so)(ssO/SO)-l.
(so
S).
Identifying sso/so with s and aso/so with a, we conclude from this that the
equation sx = a always possesses a solution in Qc(R), namely, x = a/s =
as-l. Second, notice that in Qc(R) multiplicative inverses exist not only for
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
(dr' + sb')/s'r'.
as'
= sa',
br' = rb'.
=
=
(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 Qc(R) under these
operations.
Lemma. The syst~m Qc(R) forms a commutative ring with identity.
for arbitrary a/s in Qc(R), since (as')s = (ss')a. Loosely speaking, common
factors belonging to S may be .canceIled in a fracti<?n as/ss.
This prov.es part of the theorem below.
Theorem 4-9. Any comrnutative ring R with at least one non-zerodivisor possesses a classical ring of quotients.
63
The reader can easily check that K is a subring of Qc(R). An obvious (onto)
mappingf: R -+ K is defined by takingf(a) = aso/so. Since the condition
aso/so = 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 preserving both addition and multiplication:
(sso/so)(so/sso)
ss~/ss~
= so/so.
Since so/so plays the role of the identity element for Qc(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 Q.(R) can be written as as-l. It should be clear that
a/s
(aso/so)(so/sso)
(aso/so)(ssO/SO)-l.
(so
S).
Identifying sso/so with s and aso/so with a, we conclude from this that the
equation sx = a always possesses a solution in Qc(R), namely, x = a/s =
as-l. Second, notice that in Qc(R) multiplicative inverses exist not only for
l'
.'
64
members of S but for all elements of Qcl(R) which can be repiesented in the
form r/s, where r, s are both non-zero-divisors; in fact,
(r/s)(s/r) = rs/sr = so/so
When thering R is an integral domain, we may take the set S ofnon-zerodivisors as consisting of a1l the elements of R which are not Zefo. The last
remark of the preceding paragraph then leads to the following:important
theorem.'
Theorem 4-10. For any integral domain R, the system Qc(R) forms a
field, customari\y known as the field of quotients of R.
Since an integral domain is (isomorphic to) a subring of lts field .of
....
quotients, we also obtain
Corollary. A ring is an integral domain if and only if it isasubring of
a field:'
. '.
It should be pointed out that the hypothesis of commutativify:is essential
to this last theorem; indeed, there exist noncommutative rings without
divisors ofzero that cannot be imbedded in any division ringo
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 ca~ be
imbedded, in the sense that any field in which R is imbeddable contams a
subfield isomorphic to QcI(R) (Problem 20).
. The existence theorem for the classical ring of quotients can be supplemented by the following result, which shows that it is essentia11y unique.
Theorem 4-11. Let R and R' be two commutative rings, each containing
at least'one non-zero-devisor. Then, any isomorphism of R onto R ' has a
/
nique extension to an isomorphism of QCI(R) onto Qcl(R );
Proo! To begin with, each member of Qcl(R) may be written in the form
ab -1 where a b E R and bis a non-zero-divisor in R. Given an isomorphism
1
cp: R'-4 R' , the element cp(b) will be a no~-zero-divis~r of R' , s~ that cp(bt
is present in Qcl(R /). Suppose that cp admlts an extenslOn to an Isomorphlsm
1
<1>: Qcl(R) -4 Qcl(R/). Sin ce a = (ab- )b, we would then ha ve
.
cp(a)
= <I>(a)
= CD(ab- 1)<I>(b)
<I>(ab- 1)cp(b),
65
= cp(ab)ep(b)-1
= cp(a)~(b)cp(b)-1
.',
cp(a).
":
l'
.'
64
members of S but for all elements of Qcl(R) which can be repiesented in the
form r/s, where r, s are both non-zero-divisors; in fact,
(r/s)(s/r) = rs/sr = so/so
When thering R is an integral domain, we may take the set S ofnon-zerodivisors as consisting of a1l the elements of R which are not Zefo. The last
remark of the preceding paragraph then leads to the following:important
theorem.'
Theorem 4-10. For any integral domain R, the system Qc(R) forms a
field, customari\y known as the field of quotients of R.
Since an integral domain is (isomorphic to) a subring of lts field .of
....
quotients, we also obtain
Corollary. A ring is an integral domain if and only if it isasubring of
a field:'
. '.
It should be pointed out that the hypothesis of commutativify:is essential
to this last theorem; indeed, there exist noncommutative rings without
divisors ofzero that cannot be imbedded in any division ringo
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 ca~ be
imbedded, in the sense that any field in which R is imbeddable contams a
subfield isomorphic to QcI(R) (Problem 20).
. The existence theorem for the classical ring of quotients can be supplemented by the following result, which shows that it is essentia11y unique.
Theorem 4-11. Let R and R' be two commutative rings, each containing
at least'one non-zero-devisor. Then, any isomorphism of R onto R ' has a
/
nique extension to an isomorphism of QCI(R) onto Qcl(R );
Proo! To begin with, each member of Qcl(R) may be written in the form
ab -1 where a b E R and bis a non-zero-divisor in R. Given an isomorphism
1
cp: R'-4 R' , the element cp(b) will be a no~-zero-divis~r of R' , s~ that cp(bt
is present in Qcl(R /). Suppose that cp admlts an extenslOn to an Isomorphlsm
1
<1>: Qcl(R) -4 Qcl(R/). Sin ce a = (ab- )b, we would then ha ve
.
cp(a)
= <I>(a)
= CD(ab- 1)<I>(b)
<I>(ab- 1)cp(b),
65
= cp(ab)ep(b)-1
= cp(a)~(b)cp(b)-1
.',
cp(a).
":
66
PROBLEMS
67
Since every field contains a unique prime subfield, the following subsidiary result is of interest.
CoroUary l. Every field contains a subfield which is isomorphic either
to the field Q or to one of the fields Z p'
Theorem 4-12 also provides sorne information regarding field automorphisms.
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).
=1=
O}
or
F2
{n1ln
according as the characteristic of F is O or a prime p. Since any automorphism of a field carries the identity 1 onto itself, the result should be c1ear.
PROBLEMS
1. a) Assuming that R is a division ring, show that cent R forms a field.
b) Prove that every subring, with identity, ofa field is an integral domain.
2. Let R be an integral domain and consider the set Z1 of aH integral multiples of
the identity eleme~t :
Z1 = {n1ln e Z}.
Establish that Z1 is a field if and onlyif R has positive characteristic.
3. In the field e, define a mapping f: e .:.. e by sending each complex number to its
conjugate; that is,f(a + bi) = a --:-: ~i. Verify thatfis an automorphism of C.
4. FiiId the center of the quaternion rifig H.
5. Le R be the subring of M 2( C) consisting of aH matrices of the form
(a,b,e,deR#).
Prove that R is a division ring isomorphic to the division ring of real quaternions.
=1=
O}.
It is now a purely routine matter to verify that the fields F and Q are isomorphic under the mapping g(n/m) = (n1)(m1)-1; we leave the details as
an exercise.
66
PROBLEMS
67
Since every field contains a unique prime subfield, the following subsidiary result is of interest.
CoroUary l. Every field contains a subfield which is isomorphic either
to the field Q or to one of the fields Z p'
Theorem 4-12 also provides sorne information regarding field automorphisms.
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).
=1=
O}
or
F2
{n1ln
according as the characteristic of F is O or a prime p. Since any automorphism of a field carries the identity 1 onto itself, the result should be c1ear.
PROBLEMS
1. a) Assuming that R is a division ring, show that cent R forms a field.
b) Prove that every subring, with identity, ofa field is an integral domain.
2. Let R be an integral domain and consider the set Z1 of aH integral multiples of
the identity eleme~t :
Z1 = {n1ln e Z}.
Establish that Z1 is a field if and onlyif R has positive characteristic.
3. In the field e, define a mapping f: e .:.. e by sending each complex number to its
conjugate; that is,f(a + bi) = a --:-: ~i. Verify thatfis an automorphism of C.
4. FiiId the center of the quaternion rifig H.
5. Le R be the subring of M 2( C) consisting of aH matrices of the form
(a,b,e,deR#).
Prove that R is a division ring isomorphic to the division ring of real quaternions.
=1=
O}.
It is now a purely routine matter to verify that the fields F and Q are isomorphic under the mapping g(n/m) = (n1)(m1)-1; we leave the details as
an exercise.
68
PROBLEMS
'.
8. Let 1(~) denote the number of (dstinet) positive divisors of an integer n > 1.
Prove that
. , '
a) If n has the prime faetonzaton n = pilpi' ... P'k" where the Pi are distinet
primes and ni E Z+. then 1(n)'= (n + 1)(n2 + 1) ... (n, + 1),
b) The number of ideals of Z,i~t(n).
e) "t"{n).p(n);;:: n, [Hint: II(nt1)II(1
- l/Pi) ;;:: 2'n(1/2)k.]
,;
.
Th'iJfem:
di~lg'r of R.
b) IfR-'has an identity and every non-zero-divisor of R s!"ertible iriR. then
Rf=<.Qcl(R); in particular, F
QcI(F) for any field F.
e) Q.lni.l(R)) = Qc(R)~
.
..
d) Ir R is finite, then R Qcl(R). [Hint: For any non-zero-divsor a E R, there
is sorne bE R such that a2 b = a; ah is idempotent; thus, R has an identity
element 1 and ab = 1 by Problem 12, Chapter 1.]
19. Utilize part (d) of the preeeding problern to give another proof that any finite
integral domain ls a fieId.
20. Show that any field eontaining the integral domain R as a subrilg contains the
field of quotients Q.I(R); in this sense, Q.(R) i8 the smallest field containing R.
12. Let f be a homomorphism from the ring R into the ring R' and suppose that R
has a subring F whieh is a field. Establish that either F ~ ker f or else R' eontams
a subring isomorphie to F. .
22. Let R be an arbitrary ring (not necessarily commutative) with al least one non-zerodivisor. Prove that R possesses a c1assicaI ring of quotients if and only if it satisfies
the so-ealled Ore eondition: for aU a, b E R. b being a non-zero-divisor, there exst
elements e, d E R, with d a non-zero-divisor such that ad
be.
23. Prove tbat any automorpbism of an integral domain R admits a unique extension .
to the field of quotients Qcl(R).
~ ~(
18. Let R he a commutative ring eontaining at least one non-zero-divisor. Prove that
a) Al'lement ab- 1 is a non-zero-divisor of Q'I(R) if and only if a is a non-zero-
'
9. Given that the set H~= {[aJ Z,I[a] is not a zero divisor of Z.}, prove that
(H" 'h) forms a finite gioup 'qforder .p(r).
69
24. Let F be a field and 21 the set of integral multiples of the identity. Verify that
the prime subfield of F coincides with Q,(Zl). [Hint: Problem 20.J
25. Establish the following assertion, thereby completing the proof of Theorem 4-12:
If F is a field of eharaeteristie zero and
15. Prove that if the field F is of eharaeteristie p > O, then every subfield of F has
eharacteristie p.
K
"
= {(n1)(m1)-1In. m E Z; m =1= O)
C!!
(nl)(m1)-.
./..
68
PROBLEMS
'.
8. Let 1(~) denote the number of (dstinet) positive divisors of an integer n > 1.
Prove that
. , '
a) If n has the prime faetonzaton n = pilpi' ... P'k" where the Pi are distinet
primes and ni E Z+. then 1(n)'= (n + 1)(n2 + 1) ... (n, + 1),
b) The number of ideals of Z,i~t(n).
e) "t"{n).p(n);;:: n, [Hint: II(nt1)II(1
- l/Pi) ;;:: 2'n(1/2)k.]
,;
.
Th'iJfem:
di~lg'r of R.
b) IfR-'has an identity and every non-zero-divisor of R s!"ertible iriR. then
Rf=<.Qcl(R); in particular, F
QcI(F) for any field F.
e) Q.lni.l(R)) = Qc(R)~
.
..
d) Ir R is finite, then R Qcl(R). [Hint: For any non-zero-divsor a E R, there
is sorne bE R such that a2 b = a; ah is idempotent; thus, R has an identity
element 1 and ab = 1 by Problem 12, Chapter 1.]
19. Utilize part (d) of the preeeding problern to give another proof that any finite
integral domain ls a fieId.
20. Show that any field eontaining the integral domain R as a subrilg contains the
field of quotients Q.I(R); in this sense, Q.(R) i8 the smallest field containing R.
12. Let f be a homomorphism from the ring R into the ring R' and suppose that R
has a subring F whieh is a field. Establish that either F ~ ker f or else R' eontams
a subring isomorphie to F. .
22. Let R be an arbitrary ring (not necessarily commutative) with al least one non-zerodivisor. Prove that R possesses a c1assicaI ring of quotients if and only if it satisfies
the so-ealled Ore eondition: for aU a, b E R. b being a non-zero-divisor, there exst
elements e, d E R, with d a non-zero-divisor such that ad
be.
23. Prove tbat any automorpbism of an integral domain R admits a unique extension .
to the field of quotients Qcl(R).
~ ~(
18. Let R he a commutative ring eontaining at least one non-zero-divisor. Prove that
a) Al'lement ab- 1 is a non-zero-divisor of Q'I(R) if and only if a is a non-zero-
'
9. Given that the set H~= {[aJ Z,I[a] is not a zero divisor of Z.}, prove that
(H" 'h) forms a finite gioup 'qforder .p(r).
69
24. Let F be a field and 21 the set of integral multiples of the identity. Verify that
the prime subfield of F coincides with Q,(Zl). [Hint: Problem 20.J
25. Establish the following assertion, thereby completing the proof of Theorem 4-12:
If F is a field of eharaeteristie zero and
15. Prove that if the field F is of eharaeteristie p > O, then every subfield of F has
eharacteristie p.
K
"
= {(n1)(m1)-1In. m E Z; m =1= O)
C!!
(nl)(m1)-.
./..
70
FIVE
{ab- 1 EQc(R)laER,bES},
prove that Rs is a subring of the ring of quotients Qc(R), known as the ring of
qUQtients of R relative to S.
27. a) Show that the set S = {n E Zlp { n; p a' fixed prime} is multiplieatively c10sed
and.determine Zs, the ring of quotients of Z relative to S.
b) If R is any ring satisfying Z ~ R ~ Q, prove that R = Zs for a suitable
multiplieatively c10sed subset S ~ Z. [Hint: Cnsider the set S = {m E Zlfor
sorne n E Z, nlm E ,R; ged (n, m) = 1}.]
= Rs.
29. Let S be a multiplieatively c10sed subset of the ring R whieh eontains no zero
divisor of R nor zero.
a) If 1 is an ideal of R. verify that the set IS-1 = {ab- 1 E Qc{R)la El, b E S} is
anideal,ofQc(R). Conversely,eaehidealJ ofQc(R)isoftheformJ = (J n R)S-I.
b) For ideals l. J of R. establish the identities
71
70
FIVE
{ab- 1 EQc(R)laER,bES},
prove that Rs is a subring of the ring of quotients Qc(R), known as the ring of
qUQtients of R relative to S.
27. a) Show that the set S = {n E Zlp { n; p a' fixed prime} is multiplieatively c10sed
and.determine Zs, the ring of quotients of Z relative to S.
b) If R is any ring satisfying Z ~ R ~ Q, prove that R = Zs for a suitable
multiplieatively c10sed subset S ~ Z. [Hint: Cnsider the set S = {m E Zlfor
sorne n E Z, nlm E ,R; ged (n, m) = 1}.]
= Rs.
29. Let S be a multiplieatively c10sed subset of the ring R whieh eontains no zero
divisor of R nor zero.
a) If 1 is an ideal of R. verify that the set IS-1 = {ab- 1 E Qc{R)la El, b E S} is
anideal,ofQc(R). Conversely,eaehidealJ ofQc(R)isoftheformJ = (J n R)S-I.
b) For ideals l. J of R. establish the identities
71
72
.,
Example 5-1. We propose to show that in. the ring Z of integers the
maximal ideIs correspond to the prime numbers; more precisely: the
principal ideal (n), n > 1, is maximal if and only- if n is a prime.
Suppose tbat (n) is a maximal ideal of Z. If the integer n is not prime,
then n = n l l1 2, where 1 < nl < n2 < n. This implis that the ideals (nJ
~~~~~'
(n) c: (n l ) c:; Z,
..- ' .
73
'!I
(n) c: (n 2 ) c: Z,
,;
simply be their set-theoretic union. For this particular setting, Zorn 's
Lemma may be formulated as follows:
Let.91 be a nonempty family of subsets of some fixed nonempty' set with'
the property tbat for each chain C(j in .91, the union u C(j also belongs
. to.9l. Then.91 contains a set which is maximal in the sense that it is
not properIy contained in any member of .91. i
Because this may be the reader's first contact with Zorn's Lemma, we
proceed'in somewhat leisurely fashion to establish
,'o,'
Theorem 5-2. lf the ring R is finitely
of R is contained in a maximal ideal.
proper;W~~l of R}.", :
1:"(
ClearIy, some partial orderings are more useful than others in applications of Zorn's Lemma. In our later investigations we shall frequently take
S to be a fami1y of subsets of a given set and tbe partal ordering to be the
usual inclusion relation; an upper bound of any chain of elements would
The significant point, of course, is that this theorem asserts the existence
of certain maximal ideals, but gives no clue as to how actualIyto find them.
The chiefvirtue ofTheorem 5-2 is that it Ieads immediately to the following "
celebrated result.
. '
M = {fERlf(O)
= O}.
72
.,
Example 5-1. We propose to show that in. the ring Z of integers the
maximal ideIs correspond to the prime numbers; more precisely: the
principal ideal (n), n > 1, is maximal if and only- if n is a prime.
Suppose tbat (n) is a maximal ideal of Z. If the integer n is not prime,
then n = n l l1 2, where 1 < nl < n2 < n. This implis that the ideals (nJ
~~~~~'
(n) c: (n l ) c:; Z,
..- ' .
73
'!I
(n) c: (n 2 ) c: Z,
,;
simply be their set-theoretic union. For this particular setting, Zorn 's
Lemma may be formulated as follows:
Let.91 be a nonempty family of subsets of some fixed nonempty' set with'
the property tbat for each chain C(j in .91, the union u C(j also belongs
. to.9l. Then.91 contains a set which is maximal in the sense that it is
not properIy contained in any member of .91. i
Because this may be the reader's first contact with Zorn's Lemma, we
proceed'in somewhat leisurely fashion to establish
,'o,'
Theorem 5-2. lf the ring R is finitely
of R is contained in a maximal ideal.
proper;W~~l of R}.", :
1:"(
ClearIy, some partial orderings are more useful than others in applications of Zorn's Lemma. In our later investigations we shall frequently take
S to be a fami1y of subsets of a given set and tbe partal ordering to be the
usual inclusion relation; an upper bound of any chain of elements would
The significant point, of course, is that this theorem asserts the existence
of certain maximal ideals, but gives no clue as to how actualIyto find them.
The chiefvirtue ofTheorem 5-2 is that it Ieads immediately to the following "
celebrated result.
. '
M = {fERlf(O)
= O}.
74
Corollary. An element of a cornmutative ring R with identity is invertible if and only if it belongs to no maximal ideal of R.
o 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 identityo A ring of this kind
is, of course, much easier to handle than one which is not commutativeo
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 assume that al! g(ven rings are commutative with identity,eve~ when
this is not explicitly men,tionedo To be sure, a good deal of the subsequent
material could be presented without this additional restriotiono
o The Krull-Zorn Theorem has many important applications throughout
Ideal theoryo 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 O and 1.
Proof. Assume that the theorem is false; tha t is, su ppose tha t there exists
an idempotent a E R with a =1= O, 1. The relation a 2 = a implies a(l - a) = O,
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 ilE. But this means that the
principal ideals (a) and (1 - a) are both prop~:~i~eals of the ring R. As
such, they must be contained in M, the sole nj~;xi~al of R. Accordingly,
the elements a and 1 - a lie in M, whence
1= a
(1 - a)
Mo" .
'~::R.
75
to pro ve that Rjl is a field, it suffices to show that each nonzero element of
Rjl has a multiplicative inverse. Now, if the coset a + l =1= l, then a f# l.
By virtue of the fact that l is a maxirnal ideal, the ideal (l, a) generated by
l and a must be the whole ring R:
R
(1, a)
+ rali E l, rE R}o
{i
ra
+ l
(r
+ lHa + l),
+ lHb + l) = ab + l = 1 + l,
+ nE((4),n),
4 + (4)
(4).
74
Corollary. An element of a cornmutative ring R with identity is invertible if and only if it belongs to no maximal ideal of R.
o 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 identityo A ring of this kind
is, of course, much easier to handle than one which is not commutativeo
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 assume that al! g(ven rings are commutative with identity,eve~ when
this is not explicitly men,tionedo To be sure, a good deal of the subsequent
material could be presented without this additional restriotiono
o The Krull-Zorn Theorem has many important applications throughout
Ideal theoryo 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 O and 1.
Proof. Assume that the theorem is false; tha t is, su ppose tha t there exists
an idempotent a E R with a =1= O, 1. The relation a 2 = a implies a(l - a) = O,
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 ilE. But this means that the
principal ideals (a) and (1 - a) are both prop~:~i~eals of the ring R. As
such, they must be contained in M, the sole nj~;xi~al of R. Accordingly,
the elements a and 1 - a lie in M, whence
1= a
(1 - a)
Mo" .
'~::R.
75
to pro ve that Rjl is a field, it suffices to show that each nonzero element of
Rjl has a multiplicative inverse. Now, if the coset a + l =1= l, then a f# l.
By virtue of the fact that l is a maxirnal ideal, the ideal (l, a) generated by
l and a must be the whole ring R:
R
(1, a)
+ rali E l, rE R}o
{i
ra
+ l
(r
+ lHa + l),
+ lHb + l) = ab + l = 1 + l,
+ nE((4),n),
4 + (4)
(4).
76
We no.w shift o.ur attenti~n [ro.m maximal ideals to. prime ideals. Befo.re
fo.rmally defining this no.tio.n, let us turn to. the ring Z o.f integers fo.r
mo.tivatio.n. Specifically, co.nsider 'the principal' ideal (p) generated by a
prime number p. If the pro.duct ab E (p), where a, bE Z, then p divides ab.
But if a prime divides a pro.duct, it necessarily divides o.ne o.f the facto.rs.
This being the case, either aE (p) o.r bE (p). The ideal (p) thus has the
interesting property that, whenever (p) co.ntains a pro.duct, at least o.ne o.f
the facto.rs must belo.ng to. (P). This o.bservatio.n serves to. suggest and
partly to. illustrate the next defJ.nitio.n.
.,\
Definition 5-2. An ideal Tbf the ring R is a prime ide~l if, fo.r all a, b in
R, ab E 1 implies that eit?e~ a E 1 o.r b E l.
By inductio.n, Defir;itio.rii :S-2 can easi1y be extended to. finitely many
elements: an ideal 1 o.R i~,iprime if, whenever a pro.duct a 1a 2 an o.f
e1ements o.f R belo.ngs to. 1, then'at least o.ne o.f the a; E l. In this co.nnectio.n,
we should cautio.n the reader::iat many autho.rs insist that the ter~ "prime
ideal" always ineans a proper ideal.
'
Example 5-4. A co.mmutative ring R with identity is an integral do.main if
and o.nly if the zero. ideal {O} is a prime ideal o.f R.
Example5-5. The prime ideals of the ring Z are precisely the ideals (n),.
where n is a prime number, to.gether with the two. trivial ideals {O} and Z.
Fro.m abo.ve, we already kno.w that if n is a prime, then the principal ideal
(n) is a prime ideal o.f Z. On the o.ther hand, co.nsider any ideal (n) withn
co.mpo.site (n =f. O, 1); say, n = n 1n2 , where 1 < n 1 , n 2 < n. Certainly the
product n 1 n2 = n E (n). Ho.wever, sin ce neither n 1 no.r n 2 is an integral
multiple o.f n, n 1 ~ (n) and n2 ~ (n). Hence,when n is co.mpo.site, the ideal
(n) canno.t be prime. No.tice also that altho.ugh {O} is prime, it is no.t a
maximal ideal o.f Z.
Example 5-6. Fo.r an illustration o.f a ring po.ssessing allo.ntrivial prime
ideal which is no.t maximal, take R = Z x Z, where the o.peratio.ns are perfo.rmed co.mpo.nentwise. One may readily verify that Z x {O} is a prime
ideal o.f R. Since
Zx{O} e ZxZe e R,
with Z x Ze an ideal o.f R, Z x {O} fails to. be maximal.
By analo.gy with Theo.rem 5-5, the prime ideals o.f a ring may be characterizedi,n the fo.llo.wing manner.
Theorem 5-6. Let 1 be a pro.per ideal o.f the ring R. Then 1 is a prime
ideal if ~nd o.nly if the quotient ring Rjl is an integral do.main.
77
,ring with identity, so. is the quo.tient ring Rji. It remains therefo.re o.nly
to. verify that Rj1 is free o.f zero. diviso.rs. Fo.r this, assume that
\
(a
+ l)(b +
1) = l.
In o.ther wo.rds, the pro.duct o.f these two. co.sets is the zero. element o.f the
ring RjI: The fo.rego.ing equatio.n is plainly equivalent to. requiring that
ab + lj= l;o.r what amo.unts to. the same thing, ab E l. Since 1 is assumed
to. be a',prime ideal, o.ne o.f the facto.rs a o.r b must be in l. But this means
that ei~~rr the. co.set a + 1 = i o.r else b + 1 = 1; hence, Rjl is witho.ut
'
zero. diviso.rs.
To.:p:~o.ve the co.nverse, we simply reverse the argumento 'Acco.rdingly,
suppo.s'e:that Rjl is an integral do.main and the pro.duct ab E 1. In terms o.f
co.sets;~J4is mean s that
'
h!,
(a
+ I)(b +
1) = ab
1 = l.
By h;;;rhesis Rjl co.ntains no. diviso.rs o.f zero., so. that a + 1 = 1 o.r
b + 1 =;"1. In any event, o.ne o.f a o.rb belo.ngs to. 1, fo.rcing"t.to. be a prime
ideal o.f R.
There is an impo.rtant c1ass o.f ideals which are always prime, namely,
the maximal ideals. Fro.m the several ways o.fpro.ving this result, we cho.o.se
the argument given belo.w; anothei' appro.ach invo.lves the use o.f Theo.rems
5-5 and 5-6~
Theorem 5-7. In a co.mmutative ring with identity, every maximal ideal
is a prime ideal.
Proo! Assume that 1 is a maximal ideal o.f the ring R, a co.mmutative ring
with identity, and the pro.duct ab El with a ~ l. We pro.po.se to. sho.wthat
b E l. The maximality o.f 1 implies that the ideal generated by 1 and a must
be the who.le ring: R = (1, a). Hence, there exist elements i E 1, r E R such
that 1 = i + ra. Since bo.th ab and i belong to. 1, we co.nc1ude that
b = lb = (i
76
We no.w shift o.ur attenti~n [ro.m maximal ideals to. prime ideals. Befo.re
fo.rmally defining this no.tio.n, let us turn to. the ring Z o.f integers fo.r
mo.tivatio.n. Specifically, co.nsider 'the principal' ideal (p) generated by a
prime number p. If the pro.duct ab E (p), where a, bE Z, then p divides ab.
But if a prime divides a pro.duct, it necessarily divides o.ne o.f the facto.rs.
This being the case, either aE (p) o.r bE (p). The ideal (p) thus has the
interesting property that, whenever (p) co.ntains a pro.duct, at least o.ne o.f
the facto.rs must belo.ng to. (P). This o.bservatio.n serves to. suggest and
partly to. illustrate the next defJ.nitio.n.
.,\
Definition 5-2. An ideal Tbf the ring R is a prime ide~l if, fo.r all a, b in
R, ab E 1 implies that eit?e~ a E 1 o.r b E l.
By inductio.n, Defir;itio.rii :S-2 can easi1y be extended to. finitely many
elements: an ideal 1 o.R i~,iprime if, whenever a pro.duct a 1a 2 an o.f
e1ements o.f R belo.ngs to. 1, then'at least o.ne o.f the a; E l. In this co.nnectio.n,
we should cautio.n the reader::iat many autho.rs insist that the ter~ "prime
ideal" always ineans a proper ideal.
'
Example 5-4. A co.mmutative ring R with identity is an integral do.main if
and o.nly if the zero. ideal {O} is a prime ideal o.f R.
Example5-5. The prime ideals of the ring Z are precisely the ideals (n),.
where n is a prime number, to.gether with the two. trivial ideals {O} and Z.
Fro.m abo.ve, we already kno.w that if n is a prime, then the principal ideal
(n) is a prime ideal o.f Z. On the o.ther hand, co.nsider any ideal (n) withn
co.mpo.site (n =f. O, 1); say, n = n 1n2 , where 1 < n 1 , n 2 < n. Certainly the
product n 1 n2 = n E (n). Ho.wever, sin ce neither n 1 no.r n 2 is an integral
multiple o.f n, n 1 ~ (n) and n2 ~ (n). Hence,when n is co.mpo.site, the ideal
(n) canno.t be prime. No.tice also that altho.ugh {O} is prime, it is no.t a
maximal ideal o.f Z.
Example 5-6. Fo.r an illustration o.f a ring po.ssessing allo.ntrivial prime
ideal which is no.t maximal, take R = Z x Z, where the o.peratio.ns are perfo.rmed co.mpo.nentwise. One may readily verify that Z x {O} is a prime
ideal o.f R. Since
Zx{O} e ZxZe e R,
with Z x Ze an ideal o.f R, Z x {O} fails to. be maximal.
By analo.gy with Theo.rem 5-5, the prime ideals o.f a ring may be characterizedi,n the fo.llo.wing manner.
Theorem 5-6. Let 1 be a pro.per ideal o.f the ring R. Then 1 is a prime
ideal if ~nd o.nly if the quotient ring Rjl is an integral do.main.
77
,ring with identity, so. is the quo.tient ring Rji. It remains therefo.re o.nly
to. verify that Rj1 is free o.f zero. diviso.rs. Fo.r this, assume that
\
(a
+ l)(b +
1) = l.
In o.ther wo.rds, the pro.duct o.f these two. co.sets is the zero. element o.f the
ring RjI: The fo.rego.ing equatio.n is plainly equivalent to. requiring that
ab + lj= l;o.r what amo.unts to. the same thing, ab E l. Since 1 is assumed
to. be a',prime ideal, o.ne o.f the facto.rs a o.r b must be in l. But this means
that ei~~rr the. co.set a + 1 = i o.r else b + 1 = 1; hence, Rjl is witho.ut
'
zero. diviso.rs.
To.:p:~o.ve the co.nverse, we simply reverse the argumento 'Acco.rdingly,
suppo.s'e:that Rjl is an integral do.main and the pro.duct ab E 1. In terms o.f
co.sets;~J4is mean s that
'
h!,
(a
+ I)(b +
1) = ab
1 = l.
By h;;;rhesis Rjl co.ntains no. diviso.rs o.f zero., so. that a + 1 = 1 o.r
b + 1 =;"1. In any event, o.ne o.f a o.rb belo.ngs to. 1, fo.rcing"t.to. be a prime
ideal o.f R.
There is an impo.rtant c1ass o.f ideals which are always prime, namely,
the maximal ideals. Fro.m the several ways o.fpro.ving this result, we cho.o.se
the argument given belo.w; anothei' appro.ach invo.lves the use o.f Theo.rems
5-5 and 5-6~
Theorem 5-7. In a co.mmutative ring with identity, every maximal ideal
is a prime ideal.
Proo! Assume that 1 is a maximal ideal o.f the ring R, a co.mmutative ring
with identity, and the pro.duct ab El with a ~ l. We pro.po.se to. sho.wthat
b E l. The maximality o.f 1 implies that the ideal generated by 1 and a must
be the who.le ring: R = (1, a). Hence, there exist elements i E 1, r E R such
that 1 = i + ra. Since bo.th ab and i belong to. 1, we co.nc1ude that
b = lb = (i
78
(1 - a) EJ.
Proo! Assume that (a) is a prime ideal and let 1 be any ideal of R satisfying
(a) f '1 s;;; R. Because R is a principal ideal ring, there exists an element
bE R,for whichl = (b). Now, a E (a) e (b); bence, a = rb for sorne choice
ofr iii R. By supposition, (al is a prime ideal, so that either r E (a) or b E (a).
The possibility that b E (a) leads immediately to the contradiction (b) s;;; (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 =f Oand R is an integral domain, we
must have 1 = sb. Thi8, of course, means that the identity element
1 E (b) = 1, whenc 1 = R, making (a) a maximal ideal of R. Theorem
5-7 takes care of the converse.
79
.J{2 2 3} =
3} l k
4)
: = .Ji.
S;;;
ji, and
78
(1 - a) EJ.
Proo! Assume that (a) is a prime ideal and let 1 be any ideal of R satisfying
(a) f '1 s;;; R. Because R is a principal ideal ring, there exists an element
bE R,for whichl = (b). Now, a E (a) e (b); bence, a = rb for sorne choice
ofr iii R. By supposition, (al is a prime ideal, so that either r E (a) or b E (a).
The possibility that b E (a) leads immediately to the contradiction (b) s;;; (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 =f Oand R is an integral domain, we
must have 1 = sb. Thi8, of course, means that the identity element
1 E (b) = 1, whenc 1 = R, making (a) a maximal ideal of R. Theorem
5-7 takes care of the converse.
79
.J{2 2 3} =
3} l k
4)
: = .Ji.
S;;;
ji, and
80
Proof. Sinc~ .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 1J, then anE 1 n J, and so anEl, anEJ. We thus concIude
s.JI n Ji. On the other hand, if it happens that
that jTJ S
a E -Ji n ,.J, there must exist positive integers n, m, for which anE 1 and
am E J . . This implies that the element an+ m = anam E IJ; hence, a E
Accordingly, -Ji n
S .jYJ and the desired equality follows.
.:i.
J"Tn7
.,
'..
JI1.
Ji
nM
n(p) = (p)
=1=
{O}
k'
Jn (pk).
"1"
Ji
PRI~ARY
IDEALS
81
Proof. Because P is prime, RjP possesses no zero divisors and, in" particular,
no nonzerb nilpotent elements.
.
.
This corollary pro vides another good reason why a semi prime 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 b.t said about semi prime
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 ideal s o( a rather wide class of
rings (to. be quite explicit, the Noetherian rings),:9bey factorization laws
which are roughly similar to the prime factorizati9"P- laws for the positive
integers. It wiI1 turn out that the primary ideaIs;,:;YIhich we are about to
introduce, playa role analogous to the powers ofp:p.e numbers in ordinary
.arithmetic.
. ,
.JI;
.h.'
,.
Proof. Suppose that ab E.jQ, with a ~ .JQ. Then (ab)n = a"b" E Q for
sorne positive integer n. But a" ~ 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 m E Z;.;.and so bE.jQ. This is simply the statement that .JQ 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
ofintegers where, for any.n E Z +, (p) is the prime ideal associated with each
of the primary ideals (p").
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
80
Proof. Sinc~ .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 1J, then anE 1 n J, and so anEl, anEJ. We thus concIude
s.JI n Ji. On the other hand, if it happens that
that jTJ S
a E -Ji n ,.J, there must exist positive integers n, m, for which anE 1 and
am E J . . This implies that the element an+ m = anam E IJ; hence, a E
Accordingly, -Ji n
S .jYJ and the desired equality follows.
.:i.
J"Tn7
.,
'..
JI1.
Ji
nM
n(p) = (p)
=1=
{O}
k'
Jn (pk).
"1"
Ji
PRI~ARY
IDEALS
81
Proof. Because P is prime, RjP possesses no zero divisors and, in" particular,
no nonzerb nilpotent elements.
.
.
This corollary pro vides another good reason why a semi prime 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 b.t said about semi prime
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 ideal s o( a rather wide class of
rings (to. be quite explicit, the Noetherian rings),:9bey factorization laws
which are roughly similar to the prime factorizati9"P- laws for the positive
integers. It wiI1 turn out that the primary ideaIs;,:;YIhich we are about to
introduce, playa role analogous to the powers ofp:p.e numbers in ordinary
.arithmetic.
. ,
.JI;
.h.'
,.
Proof. Suppose that ab E.jQ, with a ~ .JQ. Then (ab)n = a"b" E Q for
sorne positive integer n. But a" ~ 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 m E Z;.;.and so bE.jQ. This is simply the statement that .JQ 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
ofintegers where, for any.n E Z +, (p) is the prime ideal associated with each
of the primary ideals (p").
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
Pis any prime ideal containing Qand let a E JQ. Then there exists a suitable
positive integer n such that a" E Q S;; P. Being prime, the ideal P must
contain the element a itself, which yields the inclusion .JQ S;; P.
The primary ideals of R may be characterized in the folIowing way.
Theorem 5-13. Let 1 be an ideal of the ring R. Then 1 is a primary
ideal ifand only if every zero divisor ofthe quotient ring RIl is nilpotent.
Proa.! . First, suppose that 1 is a primary ideal of R and take a + 1 to be
a zero divisor of RIl. Then there exists .sorne coset b + 1 =1= l, the zero
element of RIl, for which (a + l)(b + l) = l; that is, ab + 1 = l. Therefore ab El and, since b + 1 =1= l, we al so have b rt l. Now, 1 is assumed to
be primary, so that a" E 1 for some positive integer n. This being the case,
+ 1)" = a" + 1 =
which shows that the coset a + 1 is nilpotent.
l,
(a
Going in the other direction, we assume that any zero divisor of RI1
is nilpotent and let ab El, with b rt l. It then folIows that (a + I)(b + 1) = l,
while b + 1 =1= l; if a + 1 =1= l, this amounts to saying that a + 1 is a zero
divisor in RIl. By hypothesis, there must exist some n E Z+ such that
(a + l)" = l, which forces the element a" to be in l. Thus, 1 is a primary
ideal of R.
Theorem 5-13 serves to emphasize the' point that primary ideal s 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 folIowing somewhat special result..will be needed later, so we pause
,.,. \
to establish it before proceeding.
Corollary. If Ql' Q2' ... , Q" are a frn:ite set of primary ideals of the
ring R, alI of them having the sam&associated prime ideal P, then
Q = ni=1 Qi is also primary, with JQ; = P.
Proo.! Before we del ve into the detail:of the proof, observe that, by
Theorem 5-10,
JQ = .J n
Q = n
= () p
= P.
(a
+ Q)(b + Q)
= Q.
Since b rt Q = h Q, there exists some index i for which b rt Q. Furthermore, ab E Q with Q primary, so that the elernent a E.JQ = P = JQ.
This implies a" E Q for sorne integer n; in consequence,
(a
Q)" = a"
= Q,
83
r'(ab)
JI
82
Pis any prime ideal containing Qand let a E JQ. Then there exists a suitable
positive integer n such that a" E Q S;; P. Being prime, the ideal P must
contain the element a itself, which yields the inclusion .JQ S;; P.
The primary ideals of R may be characterized in the folIowing way.
Theorem 5-13. Let 1 be an ideal of the ring R. Then 1 is a primary
ideal ifand only if every zero divisor ofthe quotient ring RIl is nilpotent.
Proa.! . First, suppose that 1 is a primary ideal of R and take a + 1 to be
a zero divisor of RIl. Then there exists .sorne coset b + 1 =1= l, the zero
element of RIl, for which (a + l)(b + l) = l; that is, ab + 1 = l. Therefore ab El and, since b + 1 =1= l, we al so have b rt l. Now, 1 is assumed to
be primary, so that a" E 1 for some positive integer n. This being the case,
+ 1)" = a" + 1 =
which shows that the coset a + 1 is nilpotent.
l,
(a
Going in the other direction, we assume that any zero divisor of RI1
is nilpotent and let ab El, with b rt l. It then folIows that (a + I)(b + 1) = l,
while b + 1 =1= l; if a + 1 =1= l, this amounts to saying that a + 1 is a zero
divisor in RIl. By hypothesis, there must exist some n E Z+ such that
(a + l)" = l, which forces the element a" to be in l. Thus, 1 is a primary
ideal of R.
Theorem 5-13 serves to emphasize the' point that primary ideal s 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 folIowing somewhat special result..will be needed later, so we pause
,.,. \
to establish it before proceeding.
Corollary. If Ql' Q2' ... , Q" are a frn:ite set of primary ideals of the
ring R, alI of them having the sam&associated prime ideal P, then
Q = ni=1 Qi is also primary, with JQ; = P.
Proo.! Before we del ve into the detail:of the proof, observe that, by
Theorem 5-10,
JQ = .J n
Q = n
= () p
= P.
(a
+ Q)(b + Q)
= Q.
Since b rt Q = h Q, there exists some index i for which b rt Q. Furthermore, ab E Q with Q primary, so that the elernent a E.JQ = P = JQ.
This implies a" E Q for sorne integer n; in consequence,
(a
Q)" = a"
= Q,
83
r'(ab)
JI
84
{O}
S;
K.
Corollary 2. If Pis a prime ideal of the ring R, then P contains a mini mal
prime ideal of R. '
85
13
in
84
{O}
S;
K.
Corollary 2. If Pis a prime ideal of the ring R, then P contains a mini mal
prime ideal of R. '
85
13
in
86
PROBLEMS
for sorne j =1= i, which is impossible by our original choice of r.. On the
other hand, if j =1= i, the element aj necessarily lies in Pi (r i being la factor of
aj). For the final stage of the proof, put a = L ajo We first note that,
because each of al' a2 , , an is in l, the element a E 1. From the relation
ai = a - L!'Fiaj, with j'fi aj E P, it follows that a ~ Pi; otherwise, we
would obtam ai E Pi' an obvious .contradiction. Our construction thus
ensures the existence of an element a = L aj which belongs to the ideal l
and not to any Pi' thereby proving the theorern.
CoroUary. Let l be an arbitrary ideal ofthe ring R and PI' P 2 ,
be prime ideals of R. If l S;; U Pi' then l S;; Pi for some i.
87
9. a) With the aid oC Theorem 5-5 and Example 5-1, obtain another prooC oC the Caet
that Zp is a field iC and only iC p is a prime number.
b) Prove that in Zn the maximal ideals are the principal ideals (p)
p is a prime dividing n.
= pZ",
where
10. Given thatfis a homomorphism Crom the ring R onto the ring R', veriCy that
a) R' is a field iC and only iC ker f is a maximal ideal oC R,
b) R' is an integral domain iC and only iC ker f is a prime ideal oC R.
'*
11. a) Show that iC PI and P 2 are two ideals oC the ring R sueh that PI
P 2 and
P 2 $ PI' then the ideal PI n P 2 is not pririle.
b) Let {PJ be a ehain oC prime ideals oCthe ring R. Prove that u Pi and n Pi are
both prime ideals oC R.
'
Pn
12. Prove that iC 1 is an ideal oC the ring R and P is a prime ideal oC 1, then P is an
ideal oC the. whore ring R.
PROBLEMS
13. Let R denote the set oC all infinite sequenees {a n} oC rational numbers (that is,
an E Q Cor every n). R becomes a eornmutative ring with identity iC the ring
operations are defined termwise:
In the Collowing set oC problems, all rings are assumed to be eornmutative with identity.
1. a) Prove that Z Etl Z. is a maximal ideal oC the external direet sum Z Etl Z.
b) Show that the ring R is a field iC and only iC {O} is a maximal ideal
oC R.
{a n}
2. Prove that a proper ideal M oC the ring R is maximal.iC and only iC, Cor every
element r ~ M, there exists some a E R sueh that 1 + ra E M.
{b n}
{a n + bn},
{an}'{bn}
{anb n}
re MI and M 2 are distinet maximal ideals oC the ring R, estabJish the equ~t~
M I M2 = MI nM2
.
; ';
5. Let M be a proper ideal of the ring R. Prove that M is a maximal ideal if ;~d
only iC, Cor eaeh ideal 1 oC R, either 1 ~ Mor else 1 + M = R.
't',
14. Assume that P is a proper prime ideal oC the ring R ~i.t~< the property that the
quotient ring R/P is finite. Show that P must be a ma~iinal ideal oC R.
3. Letfbe a homomorphism Crom the ring R onto the ring R'. Pro ve that
a) iC M is a maximal ideal oC R with M ;2 ker J, thenf(M) is a maximal ideal oC R',
b) iCM' is a fl1aKimal ideal oC R', thenf-~(M') is a maximal ideal oC R,
e) the mapping M --+ f(M) defines a one-to-one eorrespondenee between the set
oC maximal ideals oC R whieh contain ker f and the set oC all maximal ideals
oCR'.
4.
6. An ideal 1 oC the ring R is said to be minimal iC 1 =1= {O} and there exists no id~l:l)
J oC R sueh that {O} e J e 1.
t'
a) Prove that a nonzero ideal 1 oC R is a minima1 ideal iC and only iC (a) = 1 for
eaeh nomero element a e 1.
b) VeriCy that the ring Z oCintegers has no minimal ideals.
'~l'
I
1,
7. ~et 1 be a proper ideal oC the ring R. Show that 1 is a prime ideal iC and oIily
IC the eomplement (lC 1 is a multiplieatively closed subset oC R.
8. In the ring R
15. Let R = R Etl R 2 Etl ... Etl Rn be the direet sum oC a finite number oC rings Ri'
Establish that a proper ideal 1 oC R is a maximal idealjf and only iC, Cor some i
(1 :5: i :5: n); 1 is oC the Corm
re 1 $
17. Letfbe a homomorphism Crom the ring R onto the ring R'. Prove that
I
\
;2
86
PROBLEMS
for sorne j =1= i, which is impossible by our original choice of r.. On the
other hand, if j =1= i, the element aj necessarily lies in Pi (r i being la factor of
aj). For the final stage of the proof, put a = L ajo We first note that,
because each of al' a2 , , an is in l, the element a E 1. From the relation
ai = a - L!'Fiaj, with j'fi aj E P, it follows that a ~ Pi; otherwise, we
would obtam ai E Pi' an obvious .contradiction. Our construction thus
ensures the existence of an element a = L aj which belongs to the ideal l
and not to any Pi' thereby proving the theorern.
CoroUary. Let l be an arbitrary ideal ofthe ring R and PI' P 2 ,
be prime ideals of R. If l S;; U Pi' then l S;; Pi for some i.
87
9. a) With the aid oC Theorem 5-5 and Example 5-1, obtain another prooC oC the Caet
that Zp is a field iC and only iC p is a prime number.
b) Prove that in Zn the maximal ideals are the principal ideals (p)
p is a prime dividing n.
= pZ",
where
10. Given thatfis a homomorphism Crom the ring R onto the ring R', veriCy that
a) R' is a field iC and only iC ker f is a maximal ideal oC R,
b) R' is an integral domain iC and only iC ker f is a prime ideal oC R.
'*
11. a) Show that iC PI and P 2 are two ideals oC the ring R sueh that PI
P 2 and
P 2 $ PI' then the ideal PI n P 2 is not pririle.
b) Let {PJ be a ehain oC prime ideals oCthe ring R. Prove that u Pi and n Pi are
both prime ideals oC R.
'
Pn
12. Prove that iC 1 is an ideal oC the ring R and P is a prime ideal oC 1, then P is an
ideal oC the. whore ring R.
PROBLEMS
13. Let R denote the set oC all infinite sequenees {a n} oC rational numbers (that is,
an E Q Cor every n). R becomes a eornmutative ring with identity iC the ring
operations are defined termwise:
In the Collowing set oC problems, all rings are assumed to be eornmutative with identity.
1. a) Prove that Z Etl Z. is a maximal ideal oC the external direet sum Z Etl Z.
b) Show that the ring R is a field iC and only iC {O} is a maximal ideal
oC R.
{a n}
2. Prove that a proper ideal M oC the ring R is maximal.iC and only iC, Cor every
element r ~ M, there exists some a E R sueh that 1 + ra E M.
{b n}
{a n + bn},
{an}'{bn}
{anb n}
re MI and M 2 are distinet maximal ideals oC the ring R, estabJish the equ~t~
M I M2 = MI nM2
.
; ';
5. Let M be a proper ideal of the ring R. Prove that M is a maximal ideal if ;~d
only iC, Cor eaeh ideal 1 oC R, either 1 ~ Mor else 1 + M = R.
't',
14. Assume that P is a proper prime ideal oC the ring R ~i.t~< the property that the
quotient ring R/P is finite. Show that P must be a ma~iinal ideal oC R.
3. Letfbe a homomorphism Crom the ring R onto the ring R'. Pro ve that
a) iC M is a maximal ideal oC R with M ;2 ker J, thenf(M) is a maximal ideal oC R',
b) iCM' is a fl1aKimal ideal oC R', thenf-~(M') is a maximal ideal oC R,
e) the mapping M --+ f(M) defines a one-to-one eorrespondenee between the set
oC maximal ideals oC R whieh contain ker f and the set oC all maximal ideals
oCR'.
4.
6. An ideal 1 oC the ring R is said to be minimal iC 1 =1= {O} and there exists no id~l:l)
J oC R sueh that {O} e J e 1.
t'
a) Prove that a nonzero ideal 1 oC R is a minima1 ideal iC and only iC (a) = 1 for
eaeh nomero element a e 1.
b) VeriCy that the ring Z oCintegers has no minimal ideals.
'~l'
I
1,
7. ~et 1 be a proper ideal oC the ring R. Show that 1 is a prime ideal iC and oIily
IC the eomplement (lC 1 is a multiplieatively closed subset oC R.
8. In the ring R
15. Let R = R Etl R 2 Etl ... Etl Rn be the direet sum oC a finite number oC rings Ri'
Establish that a proper ideal 1 oC R is a maximal idealjf and only iC, Cor some i
(1 :5: i :5: n); 1 is oC the Corm
re 1 $
17. Letfbe a homomorphism Crom the ring R onto the ring R'. Prove that
I
\
;2
88
.JI
.JI.
20. Letfb,~ a homomorphism from the ring R onto the ri;g R'. Prove that
, a) ifJjs an ideal of R with 1 ;2 ker f, then
= f(.JI),
b) ift';is an ideal of R', theri
1(1') =
\Jf)..
.Jr
.JJ[i)
21. Verlt'y;that the intersection of semiprime ideals of the ring R is again a semiprime
ide~Cf R.
'.
. . .
22. If l:~:~ ideal ofthe ring R, pro ve that .JYis the smallest (in the set-theoretie sen se)
semipdme ideal of R whieh eontains l.
.
23. Establish that every divisor of zero in the ring Z p" (p a prime, n > O) is nilpotent.
~, J, and Q be ideals of the ring R, with Q primary. Prove the foHowing
statements:
a) if 1 $ JQ, then the quotient Q:l ,;. Q;
b) if IJ s;: Q and 1 $ JQ, thenJ s:; Q;
el if IJ s;: Q and the ideal J is finitely generated, then either 1 s:; Q orelse 1" s;: Q
for sorne n E Z+. [Hnt: If 1 $ Q, eaeh generator of J is in .JQ.J
24. Let
If.JI
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 corriplement of P:
E
Vp = {a/b
Show that
. a) Vp is a valuation ring of Q;
b) the tinique rnaidmal ideal of v" is Mp = {a/bE Qlp1b, butpla};
.
e) the field VJMp ~ Zp. [Hint: Let the homomorphismf: Vp -> Zpq~ defined
byf(a/b)
[a] [b]-l.] .
"
30. a) Let 11 , 12 ,
,In be arbitrary ideals of the ring R and P be a primeicl~al of R.
If 1112
In' s:; P, establish that li s:; P for at least one value of i:, :;[Hint: If
li $ P for aH i, ehoose a i E li - P and eonsider the element a = ai;a~
an .]
b) Assume that M is a maximal ideal of R. Prove that, for eaeh integer n E Z+,
the only prime idea,l eontaining M n is M.
oo.
oo.
.oo
;:~t;
.';'.
31. Let R be an integral domain with the property that every proper ideal ist}:t~.:produet
of maxirnalideals. Prove that
. '
.
~:.: ..:
a) If M is a rnaximal ideal of R, then there exists an element ,a E.R' ilid ideal
K =1= {O} suehthat MK = (a). [Hint: If M =1= {O}, piekO'=I= aEM. Then
M 1M 2 oo M n = (a) s:; M for suitable rnaximal ideals Mi; henee, M = Mi for
sorne i.]
b) If 1, J, M are ideals of R, with M maximal, then 1M = JM implies 1 = J.
32. a) Ir 1 is an ideal of the ring R sueh that 1 s:; (a), show that there exists an jdeal
J of R for which aJ = l. [Hint: Take J = (1: (a)).]
b) Prove that if a principal ideal (a) of the ring R properly eontains a prime ideal
P, then P s:;
(a").
n=1
Rp .= {ab- 1
89
PROBLEMS
Qcl(R)laER; b'P}.
Prove that the ring Rp (whieh is known as the localizaton of R at the prime ideal P)
has exaetly one maximal ideal, namely, 1 = {ab- 1 E QclR)la E P; b, P}.
a,
27. A ring R is said to be a local ring if it has a unique maximal ideal. If R is a local
rlng with M as its maximal ideal, show that any element
M is in vertible in R.,
33. Let 1 be a primary ideal of the ring R. Prove that 1 has exaetly one minimal prime
ideal, namely,
[Hint: Problem 19.]
88
.JI
.JI.
20. Letfb,~ a homomorphism from the ring R onto the ri;g R'. Prove that
, a) ifJjs an ideal of R with 1 ;2 ker f, then
= f(.JI),
b) ift';is an ideal of R', theri
1(1') =
\Jf)..
.Jr
.JJ[i)
21. Verlt'y;that the intersection of semiprime ideals of the ring R is again a semiprime
ide~Cf R.
'.
. . .
22. If l:~:~ ideal ofthe ring R, pro ve that .JYis the smallest (in the set-theoretie sen se)
semipdme ideal of R whieh eontains l.
.
23. Establish that every divisor of zero in the ring Z p" (p a prime, n > O) is nilpotent.
~, J, and Q be ideals of the ring R, with Q primary. Prove the foHowing
statements:
a) if 1 $ JQ, then the quotient Q:l ,;. Q;
b) if IJ s;: Q and 1 $ JQ, thenJ s:; Q;
el if IJ s;: Q and the ideal J is finitely generated, then either 1 s:; Q orelse 1" s;: Q
for sorne n E Z+. [Hnt: If 1 $ Q, eaeh generator of J is in .JQ.J
24. Let
If.JI
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 corriplement of P:
E
Vp = {a/b
Show that
. a) Vp is a valuation ring of Q;
b) the tinique rnaidmal ideal of v" is Mp = {a/bE Qlp1b, butpla};
.
e) the field VJMp ~ Zp. [Hint: Let the homomorphismf: Vp -> Zpq~ defined
byf(a/b)
[a] [b]-l.] .
"
30. a) Let 11 , 12 ,
,In be arbitrary ideals of the ring R and P be a primeicl~al of R.
If 1112
In' s:; P, establish that li s:; P for at least one value of i:, :;[Hint: If
li $ P for aH i, ehoose a i E li - P and eonsider the element a = ai;a~
an .]
b) Assume that M is a maximal ideal of R. Prove that, for eaeh integer n E Z+,
the only prime idea,l eontaining M n is M.
oo.
oo.
.oo
;:~t;
.';'.
31. Let R be an integral domain with the property that every proper ideal ist}:t~.:produet
of maxirnalideals. Prove that
. '
.
~:.: ..:
a) If M is a rnaximal ideal of R, then there exists an element ,a E.R' ilid ideal
K =1= {O} suehthat MK = (a). [Hint: If M =1= {O}, piekO'=I= aEM. Then
M 1M 2 oo M n = (a) s:; M for suitable rnaximal ideals Mi; henee, M = Mi for
sorne i.]
b) If 1, J, M are ideals of R, with M maximal, then 1M = JM implies 1 = J.
32. a) Ir 1 is an ideal of the ring R sueh that 1 s:; (a), show that there exists an jdeal
J of R for which aJ = l. [Hint: Take J = (1: (a)).]
b) Prove that if a principal ideal (a) of the ring R properly eontains a prime ideal
P, then P s:;
(a").
n=1
Rp .= {ab- 1
89
PROBLEMS
Qcl(R)laER; b'P}.
Prove that the ring Rp (whieh is known as the localizaton of R at the prime ideal P)
has exaetly one maximal ideal, namely, 1 = {ab- 1 E QclR)la E P; b, P}.
a,
27. A ring R is said to be a local ring if it has a unique maximal ideal. If R is a local
rlng with M as its maximal ideal, show that any element
M is in vertible in R.,
33. Let 1 be a primary ideal of the ring R. Prove that 1 has exaetly one minimal prime
ideal, namely,
[Hint: Problem 19.]
SIX
91
3) ifalb,thenaelbe;
4) if alb and ble, then ale;
5) if ela and elb, then el(ax
.j ..,
90
f;
(a).
Indeed, al b means that b = ae for sorne e E R; thus, b E (a), so that (b) f; (a).
Conversely, if(b) f; (a), then there exists an e1ement e in R for which b = ae,
implying that alb.
Questions concerning divisibility are complicated somewhat by the
presence of in vertible 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 a11.
In order to overcome the difficulty that is produced by invertible elements, we introduce the following definition.
Definition 6-2. Two elements a, bE R are said to be associated elements
or simply assoeiates if a = bu, where u is an in vertible 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 c1asses 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, b E Z; i2 = -1}.
+ bi E Z(i)
+ di) = 1,
1 = (a
=
SIX
91
3) ifalb,thenaelbe;
4) if alb and ble, then ale;
5) if ela and elb, then el(ax
.j ..,
90
f;
(a).
Indeed, al b means that b = ae for sorne e E R; thus, b E (a), so that (b) f; (a).
Conversely, if(b) f; (a), then there exists an e1ement e in R for which b = ae,
implying that alb.
Questions concerning divisibility are complicated somewhat by the
presence of in vertible 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 a11.
In order to overcome the difficulty that is produced by invertible elements, we introduce the following definition.
Definition 6-2. Two elements a, bE R are said to be associated elements
or simply assoeiates if a = bu, where u is an in vertible 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 c1asses 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, b E Z; i2 = -1}.
+ bi E Z(i)
+ di) = 1,
1 = (a
=
92
+ bi
aL
integral domain R.
a;
3).(a)
(b).
yla l
Proof To prove the ~\livalence of (1) an (2), suppose that a = bu, where
u is an invertible element; then, also, b. au-l,s~rtbat both alb and bla.
Going in the opposite direction, ir al b, we can writeq;~= ax for sorne x E R;
while, from bla, t follows that a
by wth y E R. 1h'erefore, b :::. (by)x
b(yx). Since b =1= ,O, 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 oI (2) and (3) stems from our earlier remarks
~:":::~l.,
(d E R).
which is to say tbat cid. Thus, dis a greatest common divisor of al' a 2 ,
and has tbe desired representation.
... ,
an
Tbis fact shows that (al' a 2 , ",:,!:.'qn) s;:; (d), and equality follows.
For the converse, let (al' 1l~?. , a.) be a principal ideal of R:
2) ela for i
93
gcd (al' a2 ,
... ,
an)
I
,1
... ,
r1a l
rn E R.
When. (al' a 2, ... , an ) = R, the elements al' a2, ... ,an must have a
common divisor which is an invertible element of R; in this case, we say
tbat al' al' ... , are relatively prime and sball wrte gcd (al' a2, ... ,an ) = 1.
Ir al' a2 , '" , a" are nonzero elements of a principal ideal ring R, then
the corollary to Theorem 6-3 tells us that al' a2 , '" , an are re1atively prime
fand only iftbere exist r l , r2 , .;" r" E R such tbat
a.
rla l
= 1
(Bezout's Identity).
92
+ bi
aL
integral domain R.
a;
3).(a)
(b).
yla l
Proof To prove the ~\livalence of (1) an (2), suppose that a = bu, where
u is an invertible element; then, also, b. au-l,s~rtbat both alb and bla.
Going in the opposite direction, ir al b, we can writeq;~= ax for sorne x E R;
while, from bla, t follows that a
by wth y E R. 1h'erefore, b :::. (by)x
b(yx). Since b =1= ,O, 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 oI (2) and (3) stems from our earlier remarks
~:":::~l.,
(d E R).
which is to say tbat cid. Thus, dis a greatest common divisor of al' a 2 ,
and has tbe desired representation.
... ,
an
Tbis fact shows that (al' a 2 , ",:,!:.'qn) s;:; (d), and equality follows.
For the converse, let (al' 1l~?. , a.) be a principal ideal of R:
2) ela for i
93
gcd (al' a2 ,
... ,
an)
I
,1
... ,
r1a l
rn E R.
When. (al' a 2, ... , an ) = R, the elements al' a2, ... ,an must have a
common divisor which is an invertible element of R; in this case, we say
tbat al' al' ... , are relatively prime and sball wrte gcd (al' a2, ... ,an ) = 1.
Ir al' a2 , '" , a" are nonzero elements of a principal ideal ring R, then
the corollary to Theorem 6-3 tells us that al' a2 , '" , an are re1atively prime
fand only iftbere exist r l , r2 , .;" r" E R such tbat
a.
rla l
= 1
(Bezout's Identity).
94
+ sebo
1) ald for i
95
... ,
... ,
... ,
... ,
ra n)
r lcm (al' a2 ,
gcd (ra l , ra 2 ,
...
,ran )
... ,
= ~
... ,
... ,
gcd (al' a2 ,
a,,).
... ,
an ).
Proof. First, as sume that d = lcm (al'~ 'q. ... ,a,,) exists. Then ajd for
each value of i, whence ra;jrd. Now, let d' be any common multiple of
ra 2 , ra 2 , ... ,ran Then rld', say d' = rs, w~ere s E R. It follows that als
for every i and so dls. As a result, rdk.if:or rdld'. But this means that
lcm (ra l , ra 2 , ... , ran ) exists and equals r{'';' r lcm (al' a2 , ... , an ).
As regards the second assertion, suPP<i~ that e = gcd (ra l , ra 2 , ... , ran )
exists. Then rle; hence, e = rt for suitable;'tE 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 , ... , a~. Then rt'lra for i = 1,2, ... , n
and therefore rt'le. But e = rt, so that rt'lrt or t'lt. The implication is
that gcd (al' a2 , ... , an ) exists and equals t. This pro ves what we wanted:
gcd (ra l , ra 2 ,
...
,ran)
rt
r gcd (al' a2 ,
... ,
an ).
Remark. It is entirely possible for gcd (al' a2 , ... , an ) to exist without the
existence of gcd (ra l , ra 2 , ... , ran ); tbis accounts for the lack of symmetry
in the statement of the aboye lemma. (See Example 6-4.)
94
+ sebo
1) ald for i
95
... ,
... ,
... ,
... ,
ra n)
r lcm (al' a2 ,
gcd (ra l , ra 2 ,
...
,ran )
... ,
= ~
... ,
... ,
gcd (al' a2 ,
a,,).
... ,
an ).
Proof. First, as sume that d = lcm (al'~ 'q. ... ,a,,) exists. Then ajd for
each value of i, whence ra;jrd. Now, let d' be any common multiple of
ra 2 , ra 2 , ... ,ran Then rld', say d' = rs, w~ere s E R. It follows that als
for every i and so dls. As a result, rdk.if:or rdld'. But this means that
lcm (ra l , ra 2 , ... , ran ) exists and equals r{'';' r lcm (al' a2 , ... , an ).
As regards the second assertion, suPP<i~ that e = gcd (ra l , ra 2 , ... , ran )
exists. Then rle; hence, e = rt for suitable;'tE 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 , ... , a~. Then rt'lra for i = 1,2, ... , n
and therefore rt'le. But e = rt, so that rt'lrt or t'lt. The implication is
that gcd (al' a2 , ... , an ) exists and equals t. This pro ves what we wanted:
gcd (ra l , ra 2 ,
...
,ran)
rt
r gcd (al' a2 ,
... ,
an ).
Remark. It is entirely possible for gcd (al' a2 , ... , an ) to exist without the
existence of gcd (ra l , ra 2 , ... , ran ); tbis accounts for the lack of symmetry
in the statement of the aboye lemma. (See Example 6-4.)
j,l
' I
96
rE
,a.) lcm (b l , b 2 :
'" ,
b.)
... ,
... ,
... ,
ro
x.
a.) gcd (b l , b2 ,
i'
:: i
Proo! 'For a":p;oof of statment (1), set a = lcm (al' a2 , , a.). Then
ada for i = 1/2,~ ... , n,'say a = ra. Frorn the relation x
a;b, we see
tbat ab = (rJ:iiJli = riJa and so xlab;. On the other band, consider any
divisor y of th.~''izb. Then yal(ah)a, or yalxa, making xa a commOn
rnultiple of yal;Y9~, ... , ya. According to the lemrna, lcm (ya l' ya 2 , '" , ya.)
exists and equals ya. Tbus, by tbe definition of least COn;lmon multiple, we
conclude tbat yalxa, whence ylx. To recapitulate, we have shown that
x/ab for each i and wbenever y/ah, then ylx. This simply asserts that
'x
97
;:<
b.),
Although prime elements are irreducible in integral domains, the converse is not always true, as we shall see later on: In thecontext of principal
ideal domains (our primary interest in this chapter), the no'tions of an
irreducible element and a prime element coincide. This is brought out in
the theorem below.
.
Dur next result is rather striking in that it tells us tbat, at leas't.ror integral
domains, tbe gcd-property implies tbe lcm-property, and conversely.
Theorem 6-7. An integral dornain R has tbe gcd-property if and only
if R has tbe lcrn-property.
'
.1
j,l
' I
96
rE
,a.) lcm (b l , b 2 :
'" ,
b.)
... ,
... ,
... ,
ro
x.
a.) gcd (b l , b2 ,
i'
:: i
Proo! 'For a":p;oof of statment (1), set a = lcm (al' a2 , , a.). Then
ada for i = 1/2,~ ... , n,'say a = ra. Frorn the relation x
a;b, we see
tbat ab = (rJ:iiJli = riJa and so xlab;. On the other band, consider any
divisor y of th.~''izb. Then yal(ah)a, or yalxa, making xa a commOn
rnultiple of yal;Y9~, ... , ya. According to the lemrna, lcm (ya l' ya 2 , '" , ya.)
exists and equals ya. Tbus, by tbe definition of least COn;lmon multiple, we
conclude tbat yalxa, whence ylx. To recapitulate, we have shown that
x/ab for each i and wbenever y/ah, then ylx. This simply asserts that
'x
97
;:<
b.),
Although prime elements are irreducible in integral domains, the converse is not always true, as we shall see later on: In thecontext of principal
ideal domains (our primary interest in this chapter), the no'tions of an
irreducible element and a prime element coincide. This is brought out in
the theorem below.
.
Dur next result is rather striking in that it tells us tbat, at leas't.ror integral
domains, tbe gcd-property implies tbe lcm-property, and conversely.
Theorem 6-7. An integral dornain R has tbe gcd-property if and only
if R has tbe lcrn-property.
'
.1
98
bta = bsp
+ pct
= p(bs
ct),
99
inverse, then a = r-1p E (P), from wbich it follows that (a) s: (p), an obvious
contradiction. Accordingly, the element a is invertible, whence (a) = R.
This argument shows that no principal ideallies 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 elemento Then
p admits a factorization p = ab where a, b E R and neither a nor b is invertible (the alternative possibility that p has an inverse implies (p) = R,
so may be ruled out). Now, if the element a were in (p), then a = rp ~or
sorne 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) e (a).
Next, observe that if (a) = R, then a will possess an inverse, contrary to
assumption. We thus conclude that (p) e (a) e R, which denies that (p)
is a maximal principal ideal. Our original supposition is false and a must
be an irreducible element of R.
With regard to the second assertion ofthe 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 thre exists an element rE R for
wbich ab = rp; hence, plab. By hypothesis, p is a prime element, so th~t
either pi a 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 factthat (p) is a prime ideal, it follows
that one of a or b lies in (p). This means that either pla or else plb, and makes
'p a prime element of R.
" For principal ide~l domains, aH of this may be summarized by the;;'
, following theorem.
,'
Theorem 6-10. Let R be a principal ideal domain. The non trivial ideal
(p) is a maximal {prime) ideal of R if and only if pis an irreducible (prime),
element of R.
An immediate consequence of this theorem is that every nonzero noninvertible element of R is divisible by sorne prime.
CoroUary. Let a =F O be a noninvertible element of the principal ideal
domain R. Then there exists a prime p E R such that pla.
Proa! Since a is not invertible, the principal ideal (a) =F R. Thus, by
Theorem 5-2, there exists a maximal ideal M of R such that (a) s: M. But
the preceding result teHs us that every maximal ideal is ofthe 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) s: (p), wbich is to say
that pla.
'
98
bta = bsp
+ pct
= p(bs
ct),
99
inverse, then a = r-1p E (P), from wbich it follows that (a) s: (p), an obvious
contradiction. Accordingly, the element a is invertible, whence (a) = R.
This argument shows that no principal ideallies 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 elemento Then
p admits a factorization p = ab where a, b E R and neither a nor b is invertible (the alternative possibility that p has an inverse implies (p) = R,
so may be ruled out). Now, if the element a were in (p), then a = rp ~or
sorne 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) e (a).
Next, observe that if (a) = R, then a will possess an inverse, contrary to
assumption. We thus conclude that (p) e (a) e R, which denies that (p)
is a maximal principal ideal. Our original supposition is false and a must
be an irreducible element of R.
With regard to the second assertion ofthe 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 thre exists an element rE R for
wbich ab = rp; hence, plab. By hypothesis, p is a prime element, so th~t
either pi a 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 factthat (p) is a prime ideal, it follows
that one of a or b lies in (p). This means that either pla or else plb, and makes
'p a prime element of R.
" For principal ide~l domains, aH of this may be summarized by the;;'
, following theorem.
,'
Theorem 6-10. Let R be a principal ideal domain. The non trivial ideal
(p) is a maximal {prime) ideal of R if and only if pis an irreducible (prime),
element of R.
An immediate consequence of this theorem is that every nonzero noninvertible element of R is divisible by sorne prime.
CoroUary. Let a =F O be a noninvertible element of the principal ideal
domain R. Then there exists a prime p E R such that pla.
Proa! Since a is not invertible, the principal ideal (a) =F R. Thus, by
Theorem 5-2, there exists a maximal ideal M of R such that (a) s: M. But
the preceding result teHs us that every maximal ideal is ofthe 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) s: (p), wbich is to say
that pla.
'
100
"''.,
"
where p~
1.
None of tbe listed primes p divides a. If a were divisible by Pi' for instance,
we would tben have pl(a - PlP2 ... Pm), by Tbeorem 6-1 (S), or pJ1; but
this is impossible by part (2) of the same tbeorem. Since a > 1, Theorem
6-11' asserts that it must llave a prime factor. Accordingly, a isdivisible
by a prime whicb is not among those enumerated. Tbis arguments shows
tbat tbere is no finite listing of tbe prime numbers.
with all-l = p"a" for sorne prime Pn E R. This process goes on as long as
all is not an invertible element of R.. But Theorem 6-9 asserts that the
a = PIP2'" Pn-ll!n'
':'
R.
We conclude from tbis that the element a is expressible as the finite product
oC primes
101
"
(n 5: m)"
100
"''.,
"
where p~
1.
None of tbe listed primes p divides a. If a were divisible by Pi' for instance,
we would tben have pl(a - PlP2 ... Pm), by Tbeorem 6-1 (S), or pJ1; but
this is impossible by part (2) of the same tbeorem. Since a > 1, Theorem
6-11' asserts that it must llave a prime factor. Accordingly, a isdivisible
by a prime whicb is not among those enumerated. Tbis arguments shows
tbat tbere is no finite listing of tbe prime numbers.
with all-l = p"a" for sorne prime Pn E R. This process goes on as long as
all is not an invertible element of R.. But Theorem 6-9 asserts that the
a = PIP2'" Pn-ll!n'
':'
R.
We conclude from tbis that the element a is expressible as the finite product
oC primes
101
"
(n 5: m)"
102
whe~e. the pand q are aH primes. Since Pll(qlq2 '" qm), it foHows that
PI dIVIdes sorne q (1 :::;; i :::;; m); renumbering, if necessary, we may suppose
that Pllql' Now, P~ and qf are both prime elements of R, with pllql' so
they m~st be assoClates: ql = PIU 1 for sorne invertible element U E R.
l
Cancelhng the cornmon factor PI' we are left with
P2"'Pn = ul q2"'qm'
Continuilg this argument, we arrive (after n steps) at
1=
U I U2
103
bi
Since the q are not invertible, tbis forces m = n. It has also been shown
th~t every p! has sorne qj as an associate and conversely. Thus, the two
pnme factonzations are identical, apart from the order in which the factors
appear and from replac~ment of factors by associates.
Attention is called to' the fact that the converse of Theorem 6-13 is not
true; ~n the next chapter, we shall give an exarnple of a unique factorization
domalll which is not a principal ideal domain.
A useful fact to bear in mind is that in a unique factorization domain
~ ~y irreducible element P E R is necessarily prime. For, suppose that P
dIvIdes the prodct ab, say pe = abo Let
#i.e
am~
j ".,
~t
r5(a
O(a
b)
l(a
b) - a,
b) >
102
whe~e. the pand q are aH primes. Since Pll(qlq2 '" qm), it foHows that
PI dIVIdes sorne q (1 :::;; i :::;; m); renumbering, if necessary, we may suppose
that Pllql' Now, P~ and qf are both prime elements of R, with pllql' so
they m~st be assoClates: ql = PIU 1 for sorne invertible element U E R.
l
Cancelhng the cornmon factor PI' we are left with
P2"'Pn = ul q2"'qm'
Continuilg this argument, we arrive (after n steps) at
1=
U I U2
103
bi
Since the q are not invertible, tbis forces m = n. It has also been shown
th~t every p! has sorne qj as an associate and conversely. Thus, the two
pnme factonzations are identical, apart from the order in which the factors
appear and from replac~ment of factors by associates.
Attention is called to' the fact that the converse of Theorem 6-13 is not
true; ~n the next chapter, we shall give an exarnple of a unique factorization
domalll which is not a principal ideal domain.
A useful fact to bear in mind is that in a unique factorization domain
~ ~y irreducible element P E R is necessarily prime. For, suppose that P
dIvIdes the prodct ab, say pe = abo Let
#i.e
am~
j ".,
~t
r5(a
O(a
b)
l(a
b) - a,
b) >
104
with both ( -a) = (a) < (a + b) and (b) < (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
q'b + r'
=1=. q'.
(r
(T'
=
=
with n =1= 1 a square-free integer (that is, an integer not divisible by the
square of any positive integer > 1). When n < O, we may view Q(.J) as
a subdomairt ofthe complex number system e and represent its elements in
the standar4form a + b.J i. It is not difficult to show that if nI' n 2 are
square-freeintegers, then Q(.Jn l ) = Q(Jn 2 ) if and only if nI = n2
Each ele~ent ex ~ a + bJ E Q(J) gives rise to, another element
.a = a- bjJn of Q(.Jn), which we shall call the conjugate of ex (for n < O,
ais the usual: complex conjugate of ex). A simple argument establishes that
the mappi~g;j: Q("-n) 4 Q(J) defined by f(ex) = a is anisomorphism.
To studydivisibility properties of Q(J), it is convenient to make use
of the concept of the norm of an element (an analog of the absolute value
notion inZY;:.
.
'.'
Then we:have
(b) ~ ((q - q')b) = (rt. r') < max {(r), ( -r')} < (b).
This is only possible if one of r Tr' or q - q' is zero. Since each of these
conditions implies the other, uniq'iieness
follows ..
.,:'
CoroJlary. (Division'AlgoritJiiffor Z). If a, b E Z, with b
exist unique integers q and ~}~uch that
a = qb;;;
Proof Utilize the valuation
=1=
O, then there
O~ r <
la.
la, for all nonzero a E Z.
gi~:~ by (a) =
Definition 6-9. For each element ex = a + b.J in Q(.J), the norm N(ex)
of ex is simply the product of ex and its conjugate a:
Unique factorization in Z folJows ultimately from the Division Algorithm. It is not surprising that in rings where there is an analog of division
with remainder, we can also prove uniqueness of factrization. The rnain
line of argument consists of showing that every EucIidean domain is a
principal ideal domain. (One need only consider the ring Ze to see that the
converse of tbis does not hold.)
N(ex) = exa = (a
= a2
b 2 n.
O},
b.J) (a - bJ)
{(a)laE1; a
Sorne properties of the norm function which follow easily from the
definition are listed below.
s=
105
= N(1 2 ) = N(l)N(l)
N(W,
whence N(l) = L
AJthough Q(.jn) has been labeled as a field, we actua1Jy have not proved
this to be the case; it is high time to remedy this situation.
104
with both ( -a) = (a) < (a + b) and (b) < (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
q'b + r'
=1=. q'.
(r
(T'
=
=
with n =1= 1 a square-free integer (that is, an integer not divisible by the
square of any positive integer > 1). When n < O, we may view Q(.J) as
a subdomairt ofthe complex number system e and represent its elements in
the standar4form a + b.J i. It is not difficult to show that if nI' n 2 are
square-freeintegers, then Q(.Jn l ) = Q(Jn 2 ) if and only if nI = n2
Each ele~ent ex ~ a + bJ E Q(J) gives rise to, another element
.a = a- bjJn of Q(.Jn), which we shall call the conjugate of ex (for n < O,
ais the usual: complex conjugate of ex). A simple argument establishes that
the mappi~g;j: Q("-n) 4 Q(J) defined by f(ex) = a is anisomorphism.
To studydivisibility properties of Q(J), it is convenient to make use
of the concept of the norm of an element (an analog of the absolute value
notion inZY;:.
.
'.'
Then we:have
(b) ~ ((q - q')b) = (rt. r') < max {(r), ( -r')} < (b).
This is only possible if one of r Tr' or q - q' is zero. Since each of these
conditions implies the other, uniq'iieness
follows ..
.,:'
CoroJlary. (Division'AlgoritJiiffor Z). If a, b E Z, with b
exist unique integers q and ~}~uch that
a = qb;;;
Proof Utilize the valuation
=1=
O, then there
O~ r <
la.
la, for all nonzero a E Z.
gi~:~ by (a) =
Definition 6-9. For each element ex = a + b.J in Q(.J), the norm N(ex)
of ex is simply the product of ex and its conjugate a:
Unique factorization in Z folJows ultimately from the Division Algorithm. It is not surprising that in rings where there is an analog of division
with remainder, we can also prove uniqueness of factrization. The rnain
line of argument consists of showing that every EucIidean domain is a
principal ideal domain. (One need only consider the ring Ze to see that the
converse of tbis does not hold.)
N(ex) = exa = (a
= a2
b 2 n.
O},
b.J) (a - bJ)
{(a)laE1; a
Sorne properties of the norm function which follow easily from the
definition are listed below.
s=
105
= N(1 2 ) = N(l)N(l)
N(W,
whence N(l) = L
AJthough Q(.jn) has been labeled as a field, we actua1Jy 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(.j) forms
,
a field; in fact, Q(.j) is a subfield of C.
Proo! The reader may easly verify that Q(.j) is a cornmutative ring with
identity. It remains only to establish that each nonzero element of Q(J)
has a multiplicative inverse in Q(J). Now, if O =1= a E Q(J), then the
element fJ = a/N(a) evidently les in Q(.j); furthermore, the product
afJ
= a (fi./N(a) ) = N(a)/N(a) =
1,
{a + b.fla, b E Z}.
la
Now, set
1.
O'
= x
+ y~.
xl
S;
Then
jb -
1/2,
O' E
yl
S;
1/2
Since N(a) an,d'N(fJ) are both integers, this implies that N(a) =' 1.
Next, suppse that a has the property that N(a) = p, where p is a
prime num~~r, As N(a) =1= O, 1, the element a is ileither Onor invertible in
Z(~). If a '? fJy is a factorization of a in Z(J), then
N(fJ)N(y) = N(a) =
-1, -2,2,3, is
Proo! The strategy ernployed in the proof is to show that the function c5
defined on Z(.f) by c5(a) = IN(a)1 is a Euclidean valuation for n =
1,
2, 3. We c1early ha ve c5(a) = Oif and only ir a = O, so tbat c5(a) ;;::: 1 ror all
a =1= O. Since both the norm and its absolute value are multiplicative,
condition (2) of Definition 6-8 is always satisfied:
N(a)N(fJ)
1.
Perhaps the most obvious approach to the question of unique factorization in the quadratic dornains zCJ) is to try to show that they are Euclidean
domains (a natural candidate for the Euclidean valuation is c5(a) = IN~).
In the coming theorern, we shall do precisely this for the dornains Z(~ -1),
Z(F'2), Z(J'i), and Z(J3). Although there are other Euclidean quadratic
domains, our attention is restricted to these few for which the division
algorithrn is easi1y established.
1'07
IN(ap-l -
0')1
=
=
jN(a - 2X ) + (b ,- y1J) I
(a - x) - n(b- y)!I
...
p,
from which it follows that' one of N(fJ) or N(y) must have the value 1.
From the first part of the lemma, we may thus conclude that either fJ or y
is invertible in Z(J), whlle the other is an associate of IX. Accordingly, a
is an irreducible element of Z(J).
I(a
Example 6-3. Let us find all invertible elements in Z(Z) = 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, bE Z. This
c5(P)
.1
106
Theorem 6-16. For each square-free integer n, the system Q(.j) forms
,
a field; in fact, Q(.j) is a subfield of C.
Proo! The reader may easly verify that Q(.j) is a cornmutative ring with
identity. It remains only to establish that each nonzero element of Q(J)
has a multiplicative inverse in Q(J). Now, if O =1= a E Q(J), then the
element fJ = a/N(a) evidently les in Q(.j); furthermore, the product
afJ
= a (fi./N(a) ) = N(a)/N(a) =
1,
{a + b.fla, b E Z}.
la
Now, set
1.
O'
= x
+ y~.
xl
S;
Then
jb -
1/2,
O' E
yl
S;
1/2
Since N(a) an,d'N(fJ) are both integers, this implies that N(a) =' 1.
Next, suppse that a has the property that N(a) = p, where p is a
prime num~~r, As N(a) =1= O, 1, the element a is ileither Onor invertible in
Z(~). If a '? fJy is a factorization of a in Z(J), then
N(fJ)N(y) = N(a) =
-1, -2,2,3, is
Proo! The strategy ernployed in the proof is to show that the function c5
defined on Z(.f) by c5(a) = IN(a)1 is a Euclidean valuation for n =
1,
2, 3. We c1early ha ve c5(a) = Oif and only ir a = O, so tbat c5(a) ;;::: 1 ror all
a =1= O. Since both the norm and its absolute value are multiplicative,
condition (2) of Definition 6-8 is always satisfied:
N(a)N(fJ)
1.
Perhaps the most obvious approach to the question of unique factorization in the quadratic dornains zCJ) is to try to show that they are Euclidean
domains (a natural candidate for the Euclidean valuation is c5(a) = IN~).
In the coming theorern, we shall do precisely this for the dornains Z(~ -1),
Z(F'2), Z(J'i), and Z(J3). Although there are other Euclidean quadratic
domains, our attention is restricted to these few for which the division
algorithrn is easi1y established.
1'07
IN(ap-l -
0')1
=
=
jN(a - 2X ) + (b ,- y1J) I
(a - x) - n(b- y)!I
...
p,
from which it follows that' one of N(fJ) or N(y) must have the value 1.
From the first part of the lemma, we may thus conclude that either fJ or y
is invertible in Z(J), whlle the other is an associate of IX. Accordingly, a
is an irreducible element of Z(J).
I(a
Example 6-3. Let us find all invertible elements in Z(Z) = 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, bE Z. This
c5(P)
.1
l
108
PROBLEMS
Definition 6-8 is therefore satisfied in its entirety when n = the corresponding quadratic domains Z(.J') are Euc1idean.
t,' -
an
of
irreducible
" de
.
.
. elem;nt and of a nonzero .prime do not al ways COInCl
In an arbItrary Integral domain .. Specifically, we have (2' + .J -5)1 3 ' 3, but
(2 + F)t 3, so that 2 + F5 cannot be a prime element of Z(~.
2, 2, 3 and
PROBLEMS
Example 6-'4. The various integral domains studied in this chapter might
suggest that unique factorization of elements always holds. To round o~t
the picture with an example of the failure of unique factorization, let:us
consider the quadratic domain Z(.'=5); Observe that the element 9 'has
two factorizations in Z(F"S), namely,'
.
.
9
= 33 = (2 + .J -5)(2 -:-
2.
R).
c) 1
3,
a2
= 2 + .J - 5,
a3
N(a}
N(fJ)N(y),
N(fJ), N(y)
p,
Z+,
brs;
3 gcd (3, 2
.. ;
= R
6. !n a princi~al ideal dornain R, establish that the primary ideals are the two trivial
Ideals and Ideals of the forrn (p"), where p is a prime element of R d
Z
.
h
n
an n E +.
[H'Int : If JIS'pnrnary,
t en -VI = (ji) for sorne prime elernent p. Choose 11 E Z
such thatI ~ '(pO), but 1 $ (p"+l), and show that 1 = (p").]
.
+
7. If R is a pr~nci?a.l idea! doman, we define the Length A(a) for each nonzero a E R
a IS.
Invertlble,
.
. (as follows: If.
, . then A(a) = O" otherwise A(a) is the numb er of pnmes
not necessanly dls.tlnct) 1ll any factorization of a. Prove the following assertions'
a) the length of a IS well-defined'
' .
b) if alb, then A(a) 5: A(b);
,
.
c) ~fa/b al1d,A(a) = A(b), then bla;
'.
d) If a { b and b { a, then there exist nonzero p, q E R such that
.=s
+ .J-:::S)) ".;
.1
5. If R i.s an.inte~ral domain having ~he gcd~property, show that a nonzero element
of R IS pnme If and only if it is irreducible.
R.
.'
K) = (1
J) n (l
K).:" .
K) = (l n J) + (l n K).
F5
= 2-
+ (J n
d) 1 n (J t
e) lJ = 1 n J if and only if 1 + J
b) l(J n K) = IJ n lK.
109
A)
is defined in Z(.J - 5), thereby illustrating the remark 011 page 95.
This example has the additional feature of showing that the concepts
A(pa
I
!
..
__. _ - - - - - - - - - - - - - - -
------------------_._----~-.- ~---_.------
l
108
PROBLEMS
Definition 6-8 is therefore satisfied in its entirety when n = the corresponding quadratic domains Z(.J') are Euc1idean.
t,' -
an
of
irreducible
" de
.
.
. elem;nt and of a nonzero .prime do not al ways COInCl
In an arbItrary Integral domain .. Specifically, we have (2' + .J -5)1 3 ' 3, but
(2 + F)t 3, so that 2 + F5 cannot be a prime element of Z(~.
2, 2, 3 and
PROBLEMS
Example 6-'4. The various integral domains studied in this chapter might
suggest that unique factorization of elements always holds. To round o~t
the picture with an example of the failure of unique factorization, let:us
consider the quadratic domain Z(.'=5); Observe that the element 9 'has
two factorizations in Z(F"S), namely,'
.
.
9
= 33 = (2 + .J -5)(2 -:-
2.
R).
c) 1
3,
a2
= 2 + .J - 5,
a3
N(a}
N(fJ)N(y),
N(fJ), N(y)
p,
Z+,
brs;
3 gcd (3, 2
.. ;
= R
6. !n a princi~al ideal dornain R, establish that the primary ideals are the two trivial
Ideals and Ideals of the forrn (p"), where p is a prime element of R d
Z
.
h
n
an n E +.
[H'Int : If JIS'pnrnary,
t en -VI = (ji) for sorne prime elernent p. Choose 11 E Z
such thatI ~ '(pO), but 1 $ (p"+l), and show that 1 = (p").]
.
+
7. If R is a pr~nci?a.l idea! doman, we define the Length A(a) for each nonzero a E R
a IS.
Invertlble,
.
. (as follows: If.
, . then A(a) = O" otherwise A(a) is the numb er of pnmes
not necessanly dls.tlnct) 1ll any factorization of a. Prove the following assertions'
a) the length of a IS well-defined'
' .
b) if alb, then A(a) 5: A(b);
,
.
c) ~fa/b al1d,A(a) = A(b), then bla;
'.
d) If a { b and b { a, then there exist nonzero p, q E R such that
.=s
+ .J-:::S)) ".;
.1
5. If R i.s an.inte~ral domain having ~he gcd~property, show that a nonzero element
of R IS pnme If and only if it is irreducible.
R.
.'
K) = (1
J) n (l
K).:" .
K) = (l n J) + (l n K).
F5
= 2-
+ (J n
d) 1 n (J t
e) lJ = 1 n J if and only if 1 + J
b) l(J n K) = IJ n lK.
109
A)
is defined in Z(.J - 5), thereby illustrating the remark 011 page 95.
This example has the additional feature of showing that the concepts
A(pa
I
!
..
__. _ - - - - - - - - - - - - - - -
------------------_._----~-.- ~---_.------
110
9. Verify that any two nonzero elements of a unique factorization domain possess a
greatest cornmon divisor. [Hint: Ir a = P~'P~' ... P~' and b = pl(p~ ... p~ (Pi
irreducible), then gcd (a, b) = P{'rl' ... d, whereji = min (k i , IJ]
19. Show that the quadratic domain Z(.J - 5) is not a principal ideal domain. [Hint:
Consider the ideal (3,2
free integer.
below:
. a) For nonzero a, bE R, if alb and t5(a) = t5(b), then a and b are associates. [Hi~t:
Show that bla.]
b) Fr nonzero a, bE R,.t5(ab) > t5(a) if and only if b is not an invertible element.
[Hint: Use the division algorithm to wri: a = q(ab) + r.]
c) Ir n is any integer such that 15(1) + n ~ O, then the function 15': R - {O} -+ Z
defined by t5'(a) = oCa) + n is also a Euclidean valuation on R.
13. For each ideal 1 in Z(i), the domain of Gaussian integers, establish that the quotient
ring Z(i)fI is finite. [Hint: Write 1 = (a) and use the division algorithm on ex and
any PE Z(i).]
14. Let R be a Euclidean domain with valuation o.
a) Determine whether the set 1 = {a E Rlo(a) > o(l)} u {O} is an ideal of R.
b) Assuming that the set F = {a ~ Rlo(a) = 1} u {O} is c10sed under addition,
verify that F forms a field.
15. a) Prove that if ni and n 2 are square-free integers !!-nd ni -1' n 2 , then the quadratic
field Q(.jn;,) is not isomorphic to Q(.jn;).
b) For each square-free integer n, determine all~he subfields of the quadratic
field Q(.J).
1.
1 ::;; u(l
.J2.
+ .J2)-n <
4- RinZ(R).
+ .j=5).]
20. Describe the field of quotients of the quadratic domain Z(.J) where n'is a square-
12. Assuming that R is a Euclidean domain with valuation 15, prove the statements
b) For n > 1, Z(.J) has infinitely many invertible' elements. [Hint: Ir al' b l
is a solution of the equation a2 - nb2 = 1, conc1ude that ak , bk is also a
solution, where ak + bJn = (al + bl.J)k, k r~Z+,]
c) The invertible elements of Z(.J2) are precisely the elements of the form
(1 + .J2)n, n E Z+. [Hint: Ir u is any positive invertible element of Z(.J2),
then (1 + J2)n ::;; U < (1 + .J2)n+ 1 for sorne n E Z + ; hence,
32 does
18. Prove that the domain Z(R) is not a unique factorization domain by discovering
two distinct factorizatioris of the element 10. Do the same for element 9 in the
domain Z(J7f).
10. Let R be an integral domain. Prove that R is a unique factorization domain if and
only if ,every nontrivial principal ideal f R is the product of a finite number of
maximal principal ideals and these ideals are uniquc up to a permutation of order.
11. Show that t5(a) =
= (.J6)2 =
111
I
I
110
9. Verify that any two nonzero elements of a unique factorization domain possess a
greatest cornmon divisor. [Hint: Ir a = P~'P~' ... P~' and b = pl(p~ ... p~ (Pi
irreducible), then gcd (a, b) = P{'rl' ... d, whereji = min (k i , IJ]
19. Show that the quadratic domain Z(.J - 5) is not a principal ideal domain. [Hint:
Consider the ideal (3,2
free integer.
below:
. a) For nonzero a, bE R, if alb and t5(a) = t5(b), then a and b are associates. [Hi~t:
Show that bla.]
b) Fr nonzero a, bE R,.t5(ab) > t5(a) if and only if b is not an invertible element.
[Hint: Use the division algorithm to wri: a = q(ab) + r.]
c) Ir n is any integer such that 15(1) + n ~ O, then the function 15': R - {O} -+ Z
defined by t5'(a) = oCa) + n is also a Euclidean valuation on R.
13. For each ideal 1 in Z(i), the domain of Gaussian integers, establish that the quotient
ring Z(i)fI is finite. [Hint: Write 1 = (a) and use the division algorithm on ex and
any PE Z(i).]
14. Let R be a Euclidean domain with valuation o.
a) Determine whether the set 1 = {a E Rlo(a) > o(l)} u {O} is an ideal of R.
b) Assuming that the set F = {a ~ Rlo(a) = 1} u {O} is c10sed under addition,
verify that F forms a field.
15. a) Prove that if ni and n 2 are square-free integers !!-nd ni -1' n 2 , then the quadratic
field Q(.jn;,) is not isomorphic to Q(.jn;).
b) For each square-free integer n, determine all~he subfields of the quadratic
field Q(.J).
1.
1 ::;; u(l
.J2.
+ .J2)-n <
4- RinZ(R).
+ .j=5).]
20. Describe the field of quotients of the quadratic domain Z(.J) where n'is a square-
12. Assuming that R is a Euclidean domain with valuation 15, prove the statements
b) For n > 1, Z(.J) has infinitely many invertible' elements. [Hint: Ir al' b l
is a solution of the equation a2 - nb2 = 1, conc1ude that ak , bk is also a
solution, where ak + bJn = (al + bl.J)k, k r~Z+,]
c) The invertible elements of Z(.J2) are precisely the elements of the form
(1 + .J2)n, n E Z+. [Hint: Ir u is any positive invertible element of Z(.J2),
then (1 + J2)n ::;; U < (1 + .J2)n+ 1 for sorne n E Z + ; hence,
32 does
18. Prove that the domain Z(R) is not a unique factorization domain by discovering
two distinct factorizatioris of the element 10. Do the same for element 9 in the
domain Z(J7f).
10. Let R be an integral domain. Prove that R is a unique factorization domain if and
only if ,every nontrivial principal ideal f R is the product of a finite number of
maximal principal ideals and these ideals are uniquc up to a permutation of order.
11. Show that t5(a) =
= (.J6)2 =
111
I
I
POLYNOMIAL RINGS
SE VEN
k=
POLYNOMIAL RINGS
i+ j=k
aibj
0k
is given by
aObk +alb k - l
+ ... +
ak-.lb l
113
akb o'
(It i.s understood t?a~ the aboye s'ummation runs over alI integers i, j ~
subJect to the restnctlOn that i + j = k.)
. A routine ~heck e.sta?lis~es that with these two definitions seq R becomes
a nng. To.venfy a dlstnbutlve law, for instance, take
.,
f= (a o, al' ... ),
g = (b o, b l , .. ),
h = (c o, C l ).
""","
The next step in our program is: to apply some of the previously developed
f.(g:+ h)
";",i_
f
of elements a
(a o' al' a2 ,
ak, ... )
f =
(a o, al' a 2 ,
.,.)
and
g = (b o, b l , b2 ,
are considered to be equal if and only if they are equal term by. term:
f =
+ g = (a o + bo, al + b l , ...),
fg
where
ca' bl
cl '
.1:,'
... )
(do. d l , ... ),
dk =
.,4:
.+ J=k
ai(bj
+ c)
i+ j=k
(aibj
aic.)
J
=t
aibj + I ai c "
i+ j=k
i+ j=k
J
Theorem 7-1.
(form~l)
tlve wlth ldentlty lf and only if the given ring R has these properties .
If S represents the subset of all sequences having for every term beyond
the first, that is, the set
. .
S = {(a, 0, 0, ... )Ia E R},
then it is .not. p~ticularl~ difficult to show that S constitutes a subring of'
seq R WhlCh IS Isomorphlc to R; one need only consider the mapping that
. sends ~~e seq~ence (a, 0, 0, ... ) to the element a. In this sense, seq R contains
the ongmal rmg R as a subring.
Having reached this stage, we shaIl no longer distinguish bet'ween 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 represen t power
senes the way we would like them to look. As a first step in this direction
we let ax designate the sequence
'
(O, a, 0, 0, ... ).
POLYNOMIAL RINGS
SE VEN
k=
POLYNOMIAL RINGS
i+ j=k
aibj
0k
is given by
aObk +alb k - l
+ ... +
ak-.lb l
113
akb o'
(It i.s understood t?a~ the aboye s'ummation runs over alI integers i, j ~
subJect to the restnctlOn that i + j = k.)
. A routine ~heck e.sta?lis~es that with these two definitions seq R becomes
a nng. To.venfy a dlstnbutlve law, for instance, take
.,
f= (a o, al' ... ),
g = (b o, b l , .. ),
h = (c o, C l ).
""","
The next step in our program is: to apply some of the previously developed
f.(g:+ h)
";",i_
f
of elements a
(a o' al' a2 ,
ak, ... )
f =
(a o, al' a 2 ,
.,.)
and
g = (b o, b l , b2 ,
are considered to be equal if and only if they are equal term by. term:
f =
+ g = (a o + bo, al + b l , ...),
fg
where
ca' bl
cl '
.1:,'
... )
(do. d l , ... ),
dk =
.,4:
.+ J=k
ai(bj
+ c)
i+ j=k
(aibj
aic.)
J
=t
aibj + I ai c "
i+ j=k
i+ j=k
J
Theorem 7-1.
(form~l)
tlve wlth ldentlty lf and only if the given ring R has these properties .
If S represents the subset of all sequences having for every term beyond
the first, that is, the set
. .
S = {(a, 0, 0, ... )Ia E R},
then it is .not. p~ticularl~ difficult to show that S constitutes a subring of'
seq R WhlCh IS Isomorphlc to R; one need only consider the mapping that
. sends ~~e seq~ence (a, 0, 0, ... ) to the element a. In this sense, seq R contains
the ongmal rmg R as a subring.
Having reached this stage, we shaIl no longer distinguish bet'ween 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 represen t power
senes the way we would like them to look. As a first step in this direction
we let ax designate the sequence
'
(O, a, 0, 0, ... ).
r
114
That is, ax is the specific member of seq R which has the element ti for its
second term and for all other terms. More generally, the symbol ax",
n ~ 1, will denote the sequence
ax
ax 2
and
ax 3
(0,0, a, 0, ... )
(0,0,0, a, 0, ...),
(ao, al' a2 ,
...
1+
= (a o, 0,0, ... )
x4
+ ... +
x 2
+ ... E Z[[xJ]
with tbe obvious identification oC a o witb the sequence (ao, O, O, ... ). Thus,
tbere is no loss in regarding tbe power series ring seq R as consisting of all
formal expressions
= m, so
that
f(x) = q.x" + an + lX'+ 1 + ...
g(x) = bmx"' + bm+ lX m+ 1 + ...
From the ~efinition ofmultiplication in R[[x]], the reader may easily check
that all.coefficients oC f(x)g(x) up to the (n + m)th are zero, whence
f(x)g(x) = anbmx+ m + (a + b + a.b + )x"+m+l + ....
n
1ll
wbere
not all the ak = O) in R[[x]], then the smallest integer n such that
a. =1= is called the arder off(x) and denoted by ordf(x).
f= a o
X2
,a., ...)
115
'ord ((x)g(x)
=1=
1 ; ~
Theorem 7-2. Iff(x) and g(x) are nonzero power series in R[[x]], 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 (x) + g(x) = or
ord ((x)
+ g(x)
r
114
That is, ax is the specific member of seq R which has the element ti for its
second term and for all other terms. More generally, the symbol ax",
n ~ 1, will denote the sequence
ax
ax 2
and
ax 3
(0,0, a, 0, ... )
(0,0,0, a, 0, ...),
(ao, al' a2 ,
...
1+
= (a o, 0,0, ... )
x4
+ ... +
x 2
+ ... E Z[[xJ]
with tbe obvious identification oC a o witb the sequence (ao, O, O, ... ). Thus,
tbere is no loss in regarding tbe power series ring seq R as consisting of all
formal expressions
= m, so
that
f(x) = q.x" + an + lX'+ 1 + ...
g(x) = bmx"' + bm+ lX m+ 1 + ...
From the ~efinition ofmultiplication in R[[x]], the reader may easily check
that all.coefficients oC f(x)g(x) up to the (n + m)th are zero, whence
f(x)g(x) = anbmx+ m + (a + b + a.b + )x"+m+l + ....
n
1ll
wbere
not all the ak = O) in R[[x]], then the smallest integer n such that
a. =1= is called the arder off(x) and denoted by ordf(x).
f= a o
X2
,a., ...)
115
'ord ((x)g(x)
=1=
1 ; ~
Theorem 7-2. Iff(x) and g(x) are nonzero power series in R[[x]], 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 (x) + g(x) = or
ord ((x)
+ g(x)
POLYNOMIAL RINGS
116
117
ordf(x)
J,!;
(/': Although arbitrary pow~r series rings are of sorne interest, the most
Xk
=1=
O.
f(x)g(X)-1 El,
xk(bnxn-k
(Xk) r611;ws.
~orollary l. The ring F[[ x]] is a local ring with (:i)~s its maximal
Ideal.
Proa! Inasmuch as the ideal s of F[[ x J] form a chain
F[[xJ]
(x)
(X2)
'~
...
{O},
=
k
corresp;~~g
+ rf(x)
(1
POLYNOMIAL RINGS
116
117
ordf(x)
J,!;
(/': Although arbitrary pow~r series rings are of sorne interest, the most
Xk
=1=
O.
f(x)g(X)-1 El,
xk(bnxn-k
(Xk) r611;ws.
~orollary l. The ring F[[ x]] is a local ring with (:i)~s its maximal
Ideal.
Proa! Inasmuch as the ideal s of F[[ x J] form a chain
F[[xJ]
(x)
(X2)
'~
...
{O},
=
k
corresp;~~g
+ rf(x)
(1
118
POLYNOMIAL RINGS
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 the~e would exist a power series bn:x" in MI with constant term
bo = t By the last lemma, bnxn would then be an invertible element, so
that MI = R[[x]], whichis impossible. Owingto the inc1usion M' ~ (M,x)
and the fact that M' is maximal in R[[x]], it now follows that MI = (M, x).
To verify that the corresponden ce 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 r E M be arbitrary.
Givenf(x)E R[[x]], the. sum r.+ f(x)x E (M,x) = (M, x), so that
+ f(x)x
r} = ord (g(x)
ord(r
f(x)
f(x)
(g(x)
f(x) )x.
ord x 2: 1,
alx
deg(J(x)
ak
f(x) = 1
we obtainf(x)g(x) = 4
ab j
O fork
~ m
+...n,
.
2.x,
g(x)
+ x + 4x 2 ,
+ x + 6x 2 , so that
deg (J(x)g(x)
2 < 1
degf(x)
+ deg g(x).
bk = O fork ~ max{m,n},
I+J=k
Example 7-1. Aslan illustration ofwhat might happen if R has zero divisors,
consider Zg, the ring of integers modulo 8. Taking
+ g(x)
O)
Theorem 7-5. If f(x) and g(x) are nonzero poIynomiaIs in R[x], then
1) either f(x)g(x) = 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) = Oor
Definition 7-2. Let R[x] denote, t~e set of all power series in R[[xJ]
whose coefficients are zero from s,o,!llF index onward (the particular index
varies from series to series):
..
=1=
Power series have so far received all the attention, but our primary
concern is with polynomials.
{ao
(a"
R[xJ
in R[x], we call a" the leading coefficient of f(x); and the integer n, the
degree of the polynomial.
.
r + g(x)x
for appropriate r E M and g(x) E R[[x]]' Hence, r If g(x) - f(x) =1= 0, then, upon taking orders,
119
:1:
118
POLYNOMIAL RINGS
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 the~e would exist a power series bn:x" in MI with constant term
bo = t By the last lemma, bnxn would then be an invertible element, so
that MI = R[[x]], whichis impossible. Owingto the inc1usion M' ~ (M,x)
and the fact that M' is maximal in R[[x]], it now follows that MI = (M, x).
To verify that the corresponden ce 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 r E M be arbitrary.
Givenf(x)E R[[x]], the. sum r.+ f(x)x E (M,x) = (M, x), so that
+ f(x)x
r} = ord (g(x)
ord(r
f(x)
f(x)
(g(x)
f(x) )x.
ord x 2: 1,
alx
deg(J(x)
ak
f(x) = 1
we obtainf(x)g(x) = 4
ab j
O fork
~ m
+...n,
.
2.x,
g(x)
+ x + 4x 2 ,
+ x + 6x 2 , so that
deg (J(x)g(x)
2 < 1
degf(x)
+ deg g(x).
bk = O fork ~ max{m,n},
I+J=k
Example 7-1. Aslan illustration ofwhat might happen if R has zero divisors,
consider Zg, the ring of integers modulo 8. Taking
+ g(x)
O)
Theorem 7-5. If f(x) and g(x) are nonzero poIynomiaIs in R[x], then
1) either f(x)g(x) = 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) = Oor
Definition 7-2. Let R[x] denote, t~e set of all power series in R[[xJ]
whose coefficients are zero from s,o,!llF index onward (the particular index
varies from series to series):
..
=1=
Power series have so far received all the attention, but our primary
concern is with polynomials.
{ao
(a"
R[xJ
in R[x], we call a" the leading coefficient of f(x); and the integer n, the
degree of the polynomial.
.
r + g(x)x
for appropriate r E M and g(x) E R[[x]]' Hence, r If g(x) - f(x) =1= 0, then, upon taking orders,
119
:1:
120
POLYNOMIAL RINGS
rnultiplicative in verse. For, suppose that f(x) E R[x J, with degf(x) > O;
ifj(x)g(x) = 1 for sorne g(x) in R[xJ, we could obtain the contradiction '
O = deg 1 = deg (J(x)g(x) = degf(x)
+ deg g(x)
=/=.
f(x) = ao + alx
in R[xJ, we may define f(r)
+ g(x),
+ ... + a.r.
k(x) = f(x)g(x),
k(r) = f(r)g(r).
This being so, it may be conc1uded that the mapping rpr: R[xJ --+ R' which
sendsf(x) tof(r) is a homomorphism of R[ x Jinto R'. Such a homomorphism
, will be called the substtution homomorphsm determined by r and its range
denoted by the symbol R[r]:
R[r]
{J(r)/f(x) E R[xJ}
{ao +
al/'
~'.
= ao +,~ ~J r
+ ...
1l
+ r(a.)r(x)
n
+ anr = f(r) = rpr (J (x) ).
=,
(x - il)(x - b) =
h(r) = r2
then,
h(r) = f(r) +g(r),
r(J(x) = r(a o)
then
Proof We need only verify that 4>r is unque. Suppose, then, tht there is
another horrtomorphism r: R[xJ --+ R satisfy in the indicated conditions
and con~ider any polynornialf(x) = ao..+ alx + '" + a.xER[xJ. By
,assu.mptlOn, r(ak) = ak fon each coefficlent ak , while r(x k) = r(x)k = rk.
Takmg stock of the fact that r is a homomorphism,
R' by taking
f(r) = ao + alr
O. '
121
(a
X2 -
(a
+ b)x
;'h
ab;
+ b)r + ab.
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) = O, we call the element r a root or zero ofthe polynomial
f(x). Of cou!se, a given polynomial f(x) E R[x J may not ha ve a root in R;
we shall see later that when R is a field, there alwaysexists an extension
field R' of R in whichf(x) possesses a root. It is perhaps appropriate to point
out at t~is time that the problem of obtaining all roots of a polynornial
f(x) E R[ x] is equivalent to that of finding all elements r E R' for which
f(x) E ker rpr'
'
After this brief digression, let us now state and prove the division algorithm for polynomials.
= O or
q(x)g(x)
+ r(x),
Proof The proof is by induction on the degree off(x)~ First, notice that if
f(x) = Oorf(x) =1= Oand degf(x) < deg g(x), a representation meeting the
120
POLYNOMIAL RINGS
rnultiplicative in verse. For, suppose that f(x) E R[x J, with degf(x) > O;
ifj(x)g(x) = 1 for sorne g(x) in R[xJ, we could obtain the contradiction '
O = deg 1 = deg (J(x)g(x) = degf(x)
+ deg g(x)
=/=.
f(x) = ao + alx
in R[xJ, we may define f(r)
+ g(x),
+ ... + a.r.
k(x) = f(x)g(x),
k(r) = f(r)g(r).
This being so, it may be conc1uded that the mapping rpr: R[xJ --+ R' which
sendsf(x) tof(r) is a homomorphism of R[ x Jinto R'. Such a homomorphism
, will be called the substtution homomorphsm determined by r and its range
denoted by the symbol R[r]:
R[r]
{J(r)/f(x) E R[xJ}
{ao +
al/'
~'.
= ao +,~ ~J r
+ ...
1l
+ r(a.)r(x)
n
+ anr = f(r) = rpr (J (x) ).
=,
(x - il)(x - b) =
h(r) = r2
then,
h(r) = f(r) +g(r),
r(J(x) = r(a o)
then
Proof We need only verify that 4>r is unque. Suppose, then, tht there is
another horrtomorphism r: R[xJ --+ R satisfy in the indicated conditions
and con~ider any polynornialf(x) = ao..+ alx + '" + a.xER[xJ. By
,assu.mptlOn, r(ak) = ak fon each coefficlent ak , while r(x k) = r(x)k = rk.
Takmg stock of the fact that r is a homomorphism,
R' by taking
f(r) = ao + alr
O. '
121
(a
X2 -
(a
+ b)x
;'h
ab;
+ b)r + ab.
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) = O, we call the element r a root or zero ofthe polynomial
f(x). Of cou!se, a given polynomial f(x) E R[x J may not ha ve a root in R;
we shall see later that when R is a field, there alwaysexists an extension
field R' of R in whichf(x) possesses a root. It is perhaps appropriate to point
out at t~is time that the problem of obtaining all roots of a polynornial
f(x) E R[ x] is equivalent to that of finding all elements r E R' for which
f(x) E ker rpr'
'
After this brief digression, let us now state and prove the division algorithm for polynomials.
= O or
q(x)g(x)
+ r(x),
Proof The proof is by induction on the degree off(x)~ First, notice that if
f(x) = Oorf(x) =1= Oand degf(x) < deg g(x), a representation meeting the
122
POLYNOMIAL RINGS
123
g(x) = bo
+ blx + ... +
bmx
nt
,
bm =f= O
(n
m).
f(x) = (x - a)q(x)
= f(x)
- (anb~l):x:,-mg(x)
where r(x)
fea)
(ql(X)
r(x),
q(x)g(x)
(anb;;;-l)xn-"')g(x)
f(x)
= q(x)g(x) +
r(x),
r(x)
q'(x)g(x)
r'(x),
E R[x]
is divisible by x - a if and
real = r,
Theorem 7-9. Let R be an integral domain andf(x) E R[x] be a nonzero polynomial of degree n. Then f(x) can have at most n distinct
roots in R.
n.
where r(x) and r'(x) satisf)' the requirements of ,the theorem. Subtn:~pting,
we obtain
, ,
Let us next show that a polynomial cannot have more roots in an integral
domain than its degree.
r(x)
which shows that the desired representation also exists when degf(x)
As for uniqueness, suppose that
(a - a)q(a)
= 1.
as desired.
+ r(x),
a contradiction; the last inequality relies on the fact that the degr~es 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.
O = fea') = (a - al)q(a)
and, since R has no zero divisors, q(a')
I
q(x) has at most n - 1 distinct roots. As the oniy roots of f(x) are a and
those of q(x),J(x) cannot possess more than n distinct roots in R.
122
POLYNOMIAL RINGS
123
g(x) = bo
+ blx + ... +
bmx
nt
,
bm =f= O
(n
m).
f(x) = (x - a)q(x)
= f(x)
- (anb~l):x:,-mg(x)
where r(x)
fea)
(ql(X)
r(x),
q(x)g(x)
(anb;;;-l)xn-"')g(x)
f(x)
= q(x)g(x) +
r(x),
r(x)
q'(x)g(x)
r'(x),
E R[x]
is divisible by x - a if and
real = r,
Theorem 7-9. Let R be an integral domain andf(x) E R[x] be a nonzero polynomial of degree n. Then f(x) can have at most n distinct
roots in R.
n.
where r(x) and r'(x) satisf)' the requirements of ,the theorem. Subtn:~pting,
we obtain
, ,
Let us next show that a polynomial cannot have more roots in an integral
domain than its degree.
r(x)
which shows that the desired representation also exists when degf(x)
As for uniqueness, suppose that
(a - a)q(a)
= 1.
as desired.
+ r(x),
a contradiction; the last inequality relies on the fact that the degr~es 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.
O = fea') = (a - al)q(a)
and, since R has no zero divisors, q(a')
I
q(x) has at most n - 1 distinct roots. As the oniy roots of f(x) are a and
those of q(x),J(x) cannot possess more than n distinct roots in R.
.,'
124
p,roof. The p,olynomial h(x) = f(x) - g(x) is such ,that deg h(x) ::;; n and,
by supposition, has at least n + 1 distinct roots in R., This is ,impossible
unless h(x) = 0, whencef(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. Uf(a) = O for al1 a E S, thenf(x) is the zero
poly:J.omial.
.. . ~
Proof. Suppose that R[x] is not a unique factorization dornain and let S
be the set of al1 nonconstant polynornials in R[x] which do not have a
unique factorization into irreducible elements. Select f(x) E S to be of
minimal degree. We may assurne that
'
f(x)
J:
...::'
~:.
+ deg g(x)
'
=;=
'lo.
125
~ degf(x) = (!(x) ,
satisfies
the requisiteproperties of
Now, either g(x) = 0, which forces aql(x) = bPI(X)Xn- m, or else deg g(x) <
degf(x). In the latter event, g(x) must possess a unique factorization into
irreducibles, sorne of which are qix), ... , q.(x) and PI (x). The net resuIt of
this is that PI(X)/g(x), but PI(X) f q(x) for i > 1, so that
PI(x)/(aql(x)- bpI(X)Xn- m ),
C I C 2 ...
-----_._----
ckql(X) = PI (x)h(x),
.,'
124
p,roof. The p,olynomial h(x) = f(x) - g(x) is such ,that deg h(x) ::;; n and,
by supposition, has at least n + 1 distinct roots in R., This is ,impossible
unless h(x) = 0, whencef(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. Uf(a) = O for al1 a E S, thenf(x) is the zero
poly:J.omial.
.. . ~
Proof. Suppose that R[x] is not a unique factorization dornain and let S
be the set of al1 nonconstant polynornials in R[x] which do not have a
unique factorization into irreducible elements. Select f(x) E S to be of
minimal degree. We may assurne that
'
f(x)
J:
...::'
~:.
+ deg g(x)
'
=;=
'lo.
125
~ degf(x) = (!(x) ,
satisfies
the requisiteproperties of
Now, either g(x) = 0, which forces aql(x) = bPI(X)Xn- m, or else deg g(x) <
degf(x). In the latter event, g(x) must possess a unique factorization into
irreducibles, sorne of which are qix), ... , q.(x) and PI (x). The net resuIt of
this is that PI(X)/g(x), but PI(X) f q(x) for i > 1, so that
PI(x)/(aql(x)- bpI(X)Xn- m ),
C I C 2 ...
-----_._----
ckql(X) = PI (x)h(x),
126
aq(x) = p(x)ah(x)
for sorne h(x) in R[x] or, upon canceling, q(x) = p(x)h(x); in other
words, p(x)lq(x). Using the irreducibility of q(x) as a member of R[x], '
p(x) must be an associate of q(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, Qn the positive side, that one can prove that the ring
of formal power series oyer a principal ideal domain does in fact comprise
a unique factorization domain (a not altogether trivial task).
--fo
(j(x)),
and
= h(x)f(x)
for suitable g(x), h(x) in R[~J The first of these relations signifies that
degf(x) = O, say f(x) = ao;'ilrid as a result deg h(x) = 1, say h(x) = bo +
bx (b =1= O). We thus obt~in' x = ao(b o + bx). But this means that the
product aob = 1, therebY,making ao (or, equivalently, f(x) an invertible
element of R. The implicatin 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:b: xk(x)
ak 2 (x),
with the two polynomials k(x), kix) E R[x]. This can only happen if
Co =1= O is th constant term of kix).
In consequence, the
element a has a multiplicative inverse in R, which settles the whole affair.
At the heart of a11 the interesting questions on factorization in R[x]
lies the idea of an irreducible polynomial, which we formulate in a rather
general way as fo11ows:
ac o = 1, where
Definition 7-4. Let R be an integral domain. A nonconstant polynomialf(x) E R[x] is said to be irreducible over R, or is an irreducible
POLYNOMIAL RINGS
127
In the case of the principal ideal domain F[x], where F is a freId, the
irreducible polynomials are precisely the irreducible elements of F[ x] (reca11
that the in vertible elements of the polynomial ring F[x] are just the nonzero 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 polynomials 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 we11 happen that a given polynomial is irreducible
when viewedas an element of one domain, yet reducible in another. One
such example is the polynomial X2 + 1; it is irreducible in R #[x], but
'~ducible in both C[x], where X2 + 1 = (x + i)(x and Z2[X], where
x + 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 ringo If F is a finite field, say one
of the fields of integers modulo a prime, we may actually examine a11 of the
possible roots. To cite a simple illustration, the polynomial f(x) = x 3 +
X + 1 is'irreducible in Z2[X]. Ifthere are any factors ofthis polynomial,
at least:one must be linear. But the only possible roots for f(x) are O and 1,
yetf(O) .~ f(l) = 1 =1= O, showing that no roots exist in Z2'
n,
ax
+ b = g(x)h(x),
g(x), h(x) E R[x],
1 s deg h(x) (S impbssible. This signifies
that every
(ad
bc)x
bd,
.;
:~:
126
aq(x) = p(x)ah(x)
for sorne h(x) in R[x] or, upon canceling, q(x) = p(x)h(x); in other
words, p(x)lq(x). Using the irreducibility of q(x) as a member of R[x], '
p(x) must be an associate of q(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, Qn the positive side, that one can prove that the ring
of formal power series oyer a principal ideal domain does in fact comprise
a unique factorization domain (a not altogether trivial task).
--fo
(j(x)),
and
= h(x)f(x)
for suitable g(x), h(x) in R[~J The first of these relations signifies that
degf(x) = O, say f(x) = ao;'ilrid as a result deg h(x) = 1, say h(x) = bo +
bx (b =1= O). We thus obt~in' x = ao(b o + bx). But this means that the
product aob = 1, therebY,making ao (or, equivalently, f(x) an invertible
element of R. The implicatin 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:b: xk(x)
ak 2 (x),
with the two polynomials k(x), kix) E R[x]. This can only happen if
Co =1= O is th constant term of kix).
In consequence, the
element a has a multiplicative inverse in R, which settles the whole affair.
At the heart of a11 the interesting questions on factorization in R[x]
lies the idea of an irreducible polynomial, which we formulate in a rather
general way as fo11ows:
ac o = 1, where
Definition 7-4. Let R be an integral domain. A nonconstant polynomialf(x) E R[x] is said to be irreducible over R, or is an irreducible
POLYNOMIAL RINGS
127
In the case of the principal ideal domain F[x], where F is a freId, the
irreducible polynomials are precisely the irreducible elements of F[ x] (reca11
that the in vertible elements of the polynomial ring F[x] are just the nonzero 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 polynomials 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 we11 happen that a given polynomial is irreducible
when viewedas an element of one domain, yet reducible in another. One
such example is the polynomial X2 + 1; it is irreducible in R #[x], but
'~ducible in both C[x], where X2 + 1 = (x + i)(x and Z2[X], where
x + 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 ringo If F is a finite field, say one
of the fields of integers modulo a prime, we may actually examine a11 of the
possible roots. To cite a simple illustration, the polynomial f(x) = x 3 +
X + 1 is'irreducible in Z2[X]. Ifthere are any factors ofthis polynomial,
at least:one must be linear. But the only possible roots for f(x) are O and 1,
yetf(O) .~ f(l) = 1 =1= O, showing that no roots exist in Z2'
n,
ax
+ b = g(x)h(x),
g(x), h(x) E R[x],
1 s deg h(x) (S impbssible. This signifies
that every
(ad
bc)x
bd,
.;
:~:
128
POL'NOMIA~ RINGS
ac
whence c = l/a, d
obtain
ad + bc
1,
=-
o=
now to the real field, we can obtain the form of the prime factorization in
bd
0,
- 2,
+ b/a
( - 2a
R#[x] (bear in mind that polynomials with coeflicients from R# are polynomials in C[x] and therefore have..rootsin C).
'1
bc
0, we
bt)/ab.
129
,1
J'2)
For ease of reference more than to present rie!y concepts, let us summarize in the next theorem some of the results of prvious chapters (specifically, Theorems 5-5 and ~7) as applied to the pr~P,9pal ideal domain F[x]'
a;
(x -ck ) (x - Ck)";~
X2 -
qg:~ratic
+:
2ax
+ (a 2 +
b 2 )ER!I![x]
',.r
Theorem 7-13.
We cal1.f(x;)
. ' "
128
POL'NOMIA~ RINGS
ac
whence c = l/a, d
obtain
ad + bc
1,
=-
o=
now to the real field, we can obtain the form of the prime factorization in
bd
0,
- 2,
+ b/a
( - 2a
R#[x] (bear in mind that polynomials with coeflicients from R# are polynomials in C[x] and therefore have..rootsin C).
'1
bc
0, we
bt)/ab.
129
,1
J'2)
For ease of reference more than to present rie!y concepts, let us summarize in the next theorem some of the results of prvious chapters (specifically, Theorems 5-5 and ~7) as applied to the pr~P,9pal ideal domain F[x]'
a;
(x -ck ) (x - Ck)";~
X2 -
qg:~ratic
+:
2ax
+ (a 2 +
b 2 )ER!I![x]
',.r
Theorem 7-13.
We cal1.f(x;)
. ' "
130
POLYNOMIAL RINGS
131
(a o + (p))
+ (al +
(p))x
+ ... +
(a n + (p))xn.
=1=
O.
. The assertion of the theorem is now a direct consequence of our last resulto
bcfz{x).
Since fl(X) and f2(X) are both primitive, the corollary to Gauss's L~mma
implies that we must have ad = ubc for sorne invertible element u E R.
In consequence, fl(X) = Uf2(X), showing that fl(X) is unique to within
invertible factors in R.
130
POLYNOMIAL RINGS
131
(a o + (p))
+ (al +
(p))x
+ ... +
(a n + (p))xn.
=1=
O.
. The assertion of the theorem is now a direct consequence of our last resulto
bcfz{x).
Since fl(X) and f2(X) are both primitive, the corollary to Gauss's L~mma
implies that we must have ad = ubc for sorne invertible element u E R.
In consequence, fl(X) = Uf2(X), showing that fl(X) is unique to within
invertible factors in R.
132 .
POLYNOMIAL RINGS
133
Now consider the reduction off(x) modulo the ideal P. Invoking hypothesis
(2), it can be inferred that
Since tlfe polynomial ring (R/P)[.ic] comprises an integral domain, the only
possible factorizations of (a n + P)x" are into linear factors. This being so,
a moment's refiection shows that
with a, b, e, d E R
ea- 1h 1(x),
h(x)
Thus,
v(g(x)
(br ,+ P)x',
v(h(x))
(e.
+ P)x".
f)x.
TIis mearts that the constant terms of these reductions are zero; that is,
= aeg1(x)h1(x).
1) an f/: P"
v(g(x)v(h(x) = v(J(x) = (a n
. '-~, I
bo
+P =
eo
+P
= P.
l.\lt9gether we have proved that both bo, eo E P, reveaHng at the s.ame time ,.j:;:
ihf ao = boco E p2, which is untenable by (3). Accordingly, no such
f;].9t~rization oC f(x) can occur, and f(x) is indeed irreduci,ple in R[ x J.
Theorem '7-18 leads almost immediately to the Eisenstein test for
. irredtcibility.
CoroUary. (The Eisenstein Criterion).. Let R be a unique factorization
domain and K be its field .of quotients. Letf(x) = ao + a1 x + '" +
anx" be a nonconstant polynomial in R[xJ. Suppose further ,that for
plak Cor O :;;; k < n, and p2 ari. Then,j(x) is
sorne prime j E R, p
irreducible in K[xJ.
tan'
aEZ[x]
(n >'1),
(r
+s
= n; r, s
> O).
polynomial
132 .
POLYNOMIAL RINGS
133
Now consider the reduction off(x) modulo the ideal P. Invoking hypothesis
(2), it can be inferred that
Since tlfe polynomial ring (R/P)[.ic] comprises an integral domain, the only
possible factorizations of (a n + P)x" are into linear factors. This being so,
a moment's refiection shows that
with a, b, e, d E R
ea- 1h 1(x),
h(x)
Thus,
v(g(x)
(br ,+ P)x',
v(h(x))
(e.
+ P)x".
f)x.
TIis mearts that the constant terms of these reductions are zero; that is,
= aeg1(x)h1(x).
1) an f/: P"
v(g(x)v(h(x) = v(J(x) = (a n
. '-~, I
bo
+P =
eo
+P
= P.
l.\lt9gether we have proved that both bo, eo E P, reveaHng at the s.ame time ,.j:;:
ihf ao = boco E p2, which is untenable by (3). Accordingly, no such
f;].9t~rization oC f(x) can occur, and f(x) is indeed irreduci,ple in R[ x J.
Theorem '7-18 leads almost immediately to the Eisenstein test for
. irredtcibility.
CoroUary. (The Eisenstein Criterion).. Let R be a unique factorization
domain and K be its field .of quotients. Letf(x) = ao + a1 x + '" +
anx" be a nonconstant polynomial in R[xJ. Suppose further ,that for
plak Cor O :;;; k < n, and p2 ari. Then,j(x) is
sorne prime j E R, p
irreducible in K[xJ.
tan'
aEZ[x]
(n >'1),
(r
+s
= n; r, s
> O).
polynomial
134
<lJ(X)
xl' -J
=X -
p
X - 1
XP-~
+ ... +
+ 1 EZ[X]
(p prime),
With these operations, the set R [x, y] beco mes a rng contaning R (or rather
an isomorphic copy of R) as a subring.
The {total) degree of a nonzero polynomial
m
f(x, y) =
aljxiyi
;=0 j=O
<lJ(x, + 1) = (x
(x
If the Eisenstein criterion is now applied, it ls easy to see that all the requirements for the irreducibility of <lJ(x + 1) in Z[x] are satis:fied (in the binomial
coefficient (fJ = p!/k!(P - k)!, the numerator is divisible by p for k < p,
but not the denominator). dne finds in tbis way that the original cyclotonic
polynomal <lJ(x) must Be irreducible in Z[x]; hence, al so a~ a polynomial
.
in Q[x].
Starting with a ring R we can first form the polynomal ring R[x], with
indeterminant x,and then the polynomial nng (R[x])[y] in another indeterminant y. As the notation indicates, the elements of (R[x ])[Y] are
simply polynomials
g = fo(x)
+ ... + amk~mk
From this rule, one can subsequentJy 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 ourselveswith looking at two examples.
Example7-7. To illustrate that the ideal structure of the ring F[x,
(F a field) is more complicated than that of F[x], let us show that F[x,
] = {f(x, y)x
(aij
R).
g(x, y) =
m " aijxiyi,
i=O
y]
y]
aux
is the largest of the integers i + j for which the coefficient aij :f= and is
depoted, as before, by degf(x, y). Without going into details here, let us
simply state that it is possible to obtain inequalities involying degrees
analogous to those ofTheorem 7-5; in particular, if R is an integral doman,
we still have
.
135
POLYNOMIAL RINOS
j=O
with m, n nonnegati;~ integers and aij elements of R (one makes the obvous
conventions that aooxOyO = a oo , a;oxiyO = aiox i and aOjxyi = aOjyi). In
accordance with ttadition, we shall hereafter denote (R[x])[y] by R[x, y]
and refer to the members o ths set as polynomials over R in two indeterminants x and y. ,,'
Two such polynomials with coefficients ai and bij are by de:finition
equal if al} = bij for all i and j. Addit'ion of polynomials is performed
termwise, while multiplication is given by the rule:
y]} .
Notice tbat the elements of this ideal are just the polynomials in F[x, y]
having zero constant t e r m . ,
Suppose tbat ] was actually principai, say ] = (h(x, y), where"
deg h{x, y) ~ 1. Since both x, y E ], there would exist polynomials f(x, y);,
g(x, y) i11 F[x, y] satisfying
x = f{x, y)h{x, y),
y = bh(x, y)
(a,
bE F).
= Co + C1X + CzY
(C E F).
:f= 0, then x cannot be a multiple of h(x, y),
h(x, y)
where
134
<lJ(X)
xl' -J
=X -
p
X - 1
XP-~
+ ... +
+ 1 EZ[X]
(p prime),
With these operations, the set R [x, y] beco mes a rng contaning R (or rather
an isomorphic copy of R) as a subring.
The {total) degree of a nonzero polynomial
m
f(x, y) =
aljxiyi
;=0 j=O
<lJ(x, + 1) = (x
(x
If the Eisenstein criterion is now applied, it ls easy to see that all the requirements for the irreducibility of <lJ(x + 1) in Z[x] are satis:fied (in the binomial
coefficient (fJ = p!/k!(P - k)!, the numerator is divisible by p for k < p,
but not the denominator). dne finds in tbis way that the original cyclotonic
polynomal <lJ(x) must Be irreducible in Z[x]; hence, al so a~ a polynomial
.
in Q[x].
Starting with a ring R we can first form the polynomal ring R[x], with
indeterminant x,and then the polynomial nng (R[x])[y] in another indeterminant y. As the notation indicates, the elements of (R[x ])[Y] are
simply polynomials
g = fo(x)
+ ... + amk~mk
From this rule, one can subsequentJy 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 ourselveswith looking at two examples.
Example7-7. To illustrate that the ideal structure of the ring F[x,
(F a field) is more complicated than that of F[x], let us show that F[x,
] = {f(x, y)x
(aij
R).
g(x, y) =
m " aijxiyi,
i=O
y]
y]
aux
is the largest of the integers i + j for which the coefficient aij :f= and is
depoted, as before, by degf(x, y). Without going into details here, let us
simply state that it is possible to obtain inequalities involying degrees
analogous to those ofTheorem 7-5; in particular, if R is an integral doman,
we still have
.
135
POLYNOMIAL RINOS
j=O
with m, n nonnegati;~ integers and aij elements of R (one makes the obvous
conventions that aooxOyO = a oo , a;oxiyO = aiox i and aOjxyi = aOjyi). In
accordance with ttadition, we shall hereafter denote (R[x])[y] by R[x, y]
and refer to the members o ths set as polynomials over R in two indeterminants x and y. ,,'
Two such polynomials with coefficients ai and bij are by de:finition
equal if al} = bij for all i and j. Addit'ion of polynomials is performed
termwise, while multiplication is given by the rule:
y]} .
Notice tbat the elements of this ideal are just the polynomials in F[x, y]
having zero constant t e r m . ,
Suppose tbat ] was actually principai, say ] = (h(x, y), where"
deg h{x, y) ~ 1. Since both x, y E ], there would exist polynomials f(x, y);,
g(x, y) i11 F[x, y] satisfying
x = f{x, y)h{x, y),
y = bh(x, y)
(a,
bE F).
= Co + C1X + CzY
(C E F).
:f= 0, then x cannot be a multiple of h(x, y),
h(x, y)
where
136
POLYNOMIAL RINGS
IZ
= (X Z, xy, yZ)
(X Z, y) S l.
Z
Inasmuch as -J(X Z, y) = 1, Problem 25, Chapter 5, g~ara~tees tbitt (X , y)
is primary. A straightforward argument shows that (x , y) IS not the power
bf any prime ideal ofF[x, yJ. For, in the contrary case, (X Z, y) = pn, where
Pis a prime ideal and n ~ 1. Since pn S l,with J prime, we may appeal
to Problem 30, Chapter 5, to concIde that P S; l. By the same token, the
incIusions IZ S pn S P, coupled with the fact that Pis a prime ideal, yield .
1 S P. Hence, J = P, so that 1" = (x 2 , y). Now, the element x E 1, while
x ~ (x 2 , y), implying that n ~ 1. On the other hand~ y E (~Z, y), bu! y ~ z = '
(x 2 , xy, l), which mean s z e (x 2 , y) e l. These mcIuslOn relatIons show
that it is impossible to have In == (x 2 , y) for any n ~ L
:'.,.
...
,x~]
137
136
POLYNOMIAL RINGS
IZ
= (X Z, xy, yZ)
(X Z, y) S l.
Z
Inasmuch as -J(X Z, y) = 1, Problem 25, Chapter 5, g~ara~tees tbitt (X , y)
is primary. A straightforward argument shows that (x , y) IS not the power
bf any prime ideal ofF[x, yJ. For, in the contrary case, (X Z, y) = pn, where
Pis a prime ideal and n ~ 1. Since pn S l,with J prime, we may appeal
to Problem 30, Chapter 5, to concIde that P S; l. By the same token, the
incIusions IZ S pn S P, coupled with the fact that Pis a prime ideal, yield .
1 S P. Hence, J = P, so that 1" = (x 2 , y). Now, the element x E 1, while
x ~ (x 2 , y), implying that n ~ 1. On the other hand~ y E (~Z, y), bu! y ~ z = '
(x 2 , xy, l), which mean s z e (x 2 , y) e l. These mcIuslOn relatIons show
that it is impossible to have In == (x 2 , y) for any n ~ L
:'.,.
...
,x~]
137
138
POLYNOMIAL RINGS
139
F(r) ~ QCI(F[x]).
This theorem completely determines the structure of simple transcendental extensions over F', they are all isomorphic . to the field of rational
../2
functions over F, and, hence, to each other. Thus, for lDstance, Q(n) ~ Q(2 ).
As regards simple algebraic extensions, we have
If the kernel of 4Jr is nonzero, then ker 4Jr = (J(x)) for sorne 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 b,e true of its isomorphic image F[r]. But F(r) is the
smallest field to contain 60th F and r, from which it folIows that F[r] = F(r);
this leads to the isomorphism
F(r) ~ F[x ]/(J(x)).
+;
J2
But (Jn.)2
simplyas
aJn.
ai.Jp)2
+ ... +
ak(Jn.)klaEQ; k
~ O}.
bJn.la, bE Q}.
138
POLYNOMIAL RINGS
139
F(r) ~ QCI(F[x]).
This theorem completely determines the structure of simple transcendental extensions over F', they are all isomorphic . to the field of rational
../2
functions over F, and, hence, to each other. Thus, for lDstance, Q(n) ~ Q(2 ).
As regards simple algebraic extensions, we have
If the kernel of 4Jr is nonzero, then ker 4Jr = (J(x)) for sorne 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 b,e true of its isomorphic image F[r]. But F(r) is the
smallest field to contain 60th F and r, from which it folIows that F[r] = F(r);
this leads to the isomorphism
F(r) ~ F[x ]/(J(x)).
+;
J2
But (Jn.)2
simplyas
aJn.
ai.Jp)2
+ ... +
ak(Jn.)klaEQ; k
~ O}.
bJn.la, bE Q}.
140
I
I
POLYNOMIAL RINGS
141
The implication is that, not only .J, but every member of Q(J) is algebraic
over Q; our next-proved theorem demonstrates that tbis is no accident.
Let us establish a simple, but nonetheless effective, result about successive extensions.
fi~~~
Proof. Let r E: F and co~ider th<::.,~!@ll1ents 1, r, r'-, '" , 1"', where n = [F' :Fl
These n + 1 powers of r are all in!.F"and, hence, must be linearIy dependent
OVer F (silice F' is a space of dirtension n). Thus, there exist eIements
bo,b 1, .. : ,bnEF, which arenotallzro,suchthat bol + b1r + ... + bnr" = O.
But then,f(x) = bo + b1x + ;.. + bnxn is a nonzero polynomaI in F[x]
and J(r)
0, implying that r is algebraic over F.
I
Proof. To begin with, suppose that F' i8 an algebraic extension and let R
be a subring of F ' wruch contains F; F s;:; R s;:; F I For any nonzero element
rE R, the inclusion F[r] S R certainly holds; Sin ce r is algeb'raic over F,
. we know from what has been estabIished earlier that F[r] coincides with
thefieId F(r), But then, r- 1 E F[rJ s;:; R, making thering R a field.
As regards the converse, assume that every subring of F' which contains
F forms a field. Given an eIement O f rE F', F[r] is a subrng of F' containing F and so must be a field ; in particular, r - 1 E F[r]. K.nowing this,
we may inferthe existence ofa polynomialJ(x) in F[x]such thatJ(r) = r- 1.
The element r thus becomes a root of thepolynomial g(x) = xJ(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 condtion for a finite extension
to be simple: Ir F I is a finite extension of the field F, then F' 1S . simple
extension if and only if there are only a finite number of subfields of F'
containng F.
'[F":F]
[F":F'][F' :FJ.
Proof. An 'ibbreviated proof runs as, follows. Suppose that [F' :F] = n
and [F": F'J;= m. If {al' a 2 , :, a n } is a basis ror F' as a vector space over
F and {bl,ill~, "', bm } is a basis for F" over FI, then the set of mn elements
ofthe formqb j constitutes a basis for F" over F. This implies that
,.tl
..
[F" :F]
mn = [F":F'][F' :F].
We stiil' have a few locis~ ends to tie together, inc1uding a more precise
descripti1.~8f F(r), when r is algebraic over F.
i
.F~
Proof. Let a be any element of F(r) = F[r]' Then there exists a polynomial
= g(r). Applying the division algorithm to g(x) and
the mnimum polynomial J(x) of r, we can find q(x) and s(x) in F[x]
satisfying
.
g(x) E F[x] such that a
g(x) = q(x)J(x)
s(x),
"h(x)
= Co +
clx
+oo. +
Cn_1Xn-1EF[x]
obtain
= degJ(x)
n,
140
I
I
POLYNOMIAL RINGS
141
The implication is that, not only .J, but every member of Q(J) is algebraic
over Q; our next-proved theorem demonstrates that tbis is no accident.
Let us establish a simple, but nonetheless effective, result about successive extensions.
fi~~~
Proof. Let r E: F and co~ider th<::.,~!@ll1ents 1, r, r'-, '" , 1"', where n = [F' :Fl
These n + 1 powers of r are all in!.F"and, hence, must be linearIy dependent
OVer F (silice F' is a space of dirtension n). Thus, there exist eIements
bo,b 1, .. : ,bnEF, which arenotallzro,suchthat bol + b1r + ... + bnr" = O.
But then,f(x) = bo + b1x + ;.. + bnxn is a nonzero polynomaI in F[x]
and J(r)
0, implying that r is algebraic over F.
I
Proof. To begin with, suppose that F' i8 an algebraic extension and let R
be a subring of F ' wruch contains F; F s;:; R s;:; F I For any nonzero element
rE R, the inclusion F[r] S R certainly holds; Sin ce r is algeb'raic over F,
. we know from what has been estabIished earlier that F[r] coincides with
thefieId F(r), But then, r- 1 E F[rJ s;:; R, making thering R a field.
As regards the converse, assume that every subring of F' which contains
F forms a field. Given an eIement O f rE F', F[r] is a subrng of F' containing F and so must be a field ; in particular, r - 1 E F[r]. K.nowing this,
we may inferthe existence ofa polynomialJ(x) in F[x]such thatJ(r) = r- 1.
The element r thus becomes a root of thepolynomial g(x) = xJ(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 condtion for a finite extension
to be simple: Ir F I is a finite extension of the field F, then F' 1S . simple
extension if and only if there are only a finite number of subfields of F'
containng F.
'[F":F]
[F":F'][F' :FJ.
Proof. An 'ibbreviated proof runs as, follows. Suppose that [F' :F] = n
and [F": F'J;= m. If {al' a 2 , :, a n } is a basis ror F' as a vector space over
F and {bl,ill~, "', bm } is a basis for F" over FI, then the set of mn elements
ofthe formqb j constitutes a basis for F" over F. This implies that
,.tl
..
[F" :F]
mn = [F":F'][F' :F].
We stiil' have a few locis~ ends to tie together, inc1uding a more precise
descripti1.~8f F(r), when r is algebraic over F.
i
.F~
Proof. Let a be any element of F(r) = F[r]' Then there exists a polynomial
= g(r). Applying the division algorithm to g(x) and
the mnimum polynomial J(x) of r, we can find q(x) and s(x) in F[x]
satisfying
.
g(x) E F[x] such that a
g(x) = q(x)J(x)
s(x),
"h(x)
= Co +
clx
+oo. +
Cn_1Xn-1EF[x]
obtain
= degJ(x)
n,
142
{ao
alr
+ ... +
~_lr"-llak e F}.
Then
Proof By the last-written theorem, F(r) has a finite degree [F(r) :F]. Now,
F(r) is a subspace (over F) ofthe vector space F'. This corollary is equivalent
to asserting that a subspace is the entire space ir and only if the dimensions
of the two are equal.
Example 7-10. Consider the eIement r = .Ji + i E e ;;2 Q, e as usual
so that (r 2 - 1)2 =
being the complex number fieId. Then r 2 = 1 +
4
2
- 8 or r - 2r + 9 = O. Thus, r is a root of the polynomiaI
2J2i,
2:x;2
9 e Q[xJ
and, hence, l~ an algebraic eIement over Q. Now, f(x) has the irreducible
factorizatin, ;over
e,
:;"
~
f(x)
J2 -
i)(x
+ .Ji +
i)(x
.Ji - i),
which indi~~tes that f(x) has no linear or quadratic factors in QIx]. Therefore,J(x) i$)rreducible asa member of Q[ x J and serves as the minimum
polynomialJof r over 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,
r2 = 1
2.Ji,
r3 = -
J2 + 5i.
At the same time r is"a root of the polynomiaI Xl - z.Jix + 3 e R"' [xJ,
with X2
2J2x + 3 irreducible over R#; thus, r is of degree 2 over R#.
Example 7-11. For a second illustration, we turo to the extension field
Q(J2, 13) The eIements .Ji and
are c1earIy aIgebraic over Q, being roots
ofthe polynomials xl - 2, X2 ~ 3 E Q[xJ, respectively. Our contention is
13
J2:.
13)
2.j2
(hefi(:~,
f(x) = x 4
143
= (.j2 + .J3)3
- 9(.J2
+ .J3)
is a rnember of Q(.j2
13
= Q(.j2)(13)
by a single eIernent.
= Q(.j2,
"
-' '),
','
Until now, we have always begun by assuming the existence fan extension field F' of F and then studied the structure of sii:nple exten~!E>ns F(r)
within F'. The subject can be approached from a somewhat diffe~eIlt standpoint. Given a field F and an irreducible polynomialf{x) e F[xJ~ pne 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'[xJ, has a root. (If degf(x) = j~; then, in
a trivial sense, F is itself the required extension).
To answerthis question, we take our cue from Theorem 7-19. For if
such an extension of F can be found at aH, 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[xJ and such that
F(r) ~ F[xJ/(g(x). This suggests that, when starting with a prescribed
irreducible polynomial f(x) e F[ x J, the natural object of interest should be
the associated quotient ring F[ x J/(J(x).
. After this preamble, let us proceed to some pertinent details.
142
{ao
alr
+ ... +
~_lr"-llak e F}.
Then
Proof By the last-written theorem, F(r) has a finite degree [F(r) :F]. Now,
F(r) is a subspace (over F) ofthe vector space F'. This corollary is equivalent
to asserting that a subspace is the entire space ir and only if the dimensions
of the two are equal.
Example 7-10. Consider the eIement r = .Ji + i E e ;;2 Q, e as usual
so that (r 2 - 1)2 =
being the complex number fieId. Then r 2 = 1 +
4
2
- 8 or r - 2r + 9 = O. Thus, r is a root of the polynomiaI
2J2i,
2:x;2
9 e Q[xJ
and, hence, l~ an algebraic eIement over Q. Now, f(x) has the irreducible
factorizatin, ;over
e,
:;"
~
f(x)
J2 -
i)(x
+ .Ji +
i)(x
.Ji - i),
which indi~~tes that f(x) has no linear or quadratic factors in QIx]. Therefore,J(x) i$)rreducible asa member of Q[ x J and serves as the minimum
polynomialJof r over 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,
r2 = 1
2.Ji,
r3 = -
J2 + 5i.
At the same time r is"a root of the polynomiaI Xl - z.Jix + 3 e R"' [xJ,
with X2
2J2x + 3 irreducible over R#; thus, r is of degree 2 over R#.
Example 7-11. For a second illustration, we turo to the extension field
Q(J2, 13) The eIements .Ji and
are c1earIy aIgebraic over Q, being roots
ofthe polynomials xl - 2, X2 ~ 3 E Q[xJ, respectively. Our contention is
13
J2:.
13)
2.j2
(hefi(:~,
f(x) = x 4
143
= (.j2 + .J3)3
- 9(.J2
+ .J3)
is a rnember of Q(.j2
13
= Q(.j2)(13)
by a single eIernent.
= Q(.j2,
"
-' '),
','
Until now, we have always begun by assuming the existence fan extension field F' of F and then studied the structure of sii:nple exten~!E>ns F(r)
within F'. The subject can be approached from a somewhat diffe~eIlt standpoint. Given a field F and an irreducible polynomialf{x) e F[xJ~ pne 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'[xJ, has a root. (If degf(x) = j~; then, in
a trivial sense, F is itself the required extension).
To answerthis question, we take our cue from Theorem 7-19. For if
such an extension of F can be found at aH, 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[xJ and such that
F(r) ~ F[xJ/(g(x). This suggests that, when starting with a prescribed
irreducible polynomial f(x) e F[ x J, the natural object of interest should be
the associated quotient ring F[ x J/(J(x).
. After this preamble, let us proceed to some pertinent details.
144
= 1 + 1 1= 1,
POLYNOMIAL RINGS
has degree less than that off(x) or el se is the zero polynomiaL In ract, the
cosets of l are uniquely determined by remainders on division by f(x) in
the sense that g(x) + 1 = h(x)+ l if and only if g(x) and h(x) leave the
same remainder when divided by f(x).
Thus, ir degf(x)
n > 1 (for instance,f(x) = ao + a.x + '" + anxn),
then the extension field F' may be described by
F'
F}.
l)
out with the aid ofthe relation a o + alA + :.. '::p. a.A" '70.
The last paragraph serves to bring out the point' tba! F' 18 a finite
extension of F with basis {1, A, .1. 2 , ... , A- 1 }; in particular, we infer that
(a o + 1)
+ a 1(x + 1) + ... +
b 1(x
+ ... + bn - . + 1)'-1;
As a final simplification, let us replace x + 1 by:some new symbol A, so that
bo
llb k
ao
145
1)" = O,
. [F':F] = n = degf(x) .
. To recapitulate: iff(x) E F[x] is an irreducible polynomial over F, then
there exists a finite extension P of F, such that [P: F] = degf(x), in which
f(x) has a root. Moreover, F' is a simple algebraic extenSon generated by
a root of f(x). (Admittedly, sorne work could be .saved by an appeal to
Theorems 7-21 and 7;-25, bjlt our object here is to present an alternative
approach to the slibject.)
We pause now to examine two concrete examples of the ideas just
presented.
. Example 7-12. Consider Z2: the fild of integers modulo 2, and the polynomial (x) = x 3 + X + 1 E Z2[ x]. Since neither of the elements O and 1
is a root of x 3 + x + l,f(x) must be irreducible in Z2[XJ.
Theorem 7-27 thus guarantees the existence of an extension of Z2'
. specifical1y, the field Z2[X ]/(J(x), in which the given polynomial has a root.
Denoting this root by A, the discussion aboye tells us that
{a + bA + c,.l2/a,b, e E Z2}
Z2[X]/(J(x)
.1.2 , 1
+ A + A2 },
.1. 3
-(A
1)
= A + 1,
..14
..1 2
144
= 1 + 1 1= 1,
POLYNOMIAL RINGS
has degree less than that off(x) or el se is the zero polynomiaL In ract, the
cosets of l are uniquely determined by remainders on division by f(x) in
the sense that g(x) + 1 = h(x)+ l if and only if g(x) and h(x) leave the
same remainder when divided by f(x).
Thus, ir degf(x)
n > 1 (for instance,f(x) = ao + a.x + '" + anxn),
then the extension field F' may be described by
F'
F}.
l)
out with the aid ofthe relation a o + alA + :.. '::p. a.A" '70.
The last paragraph serves to bring out the point' tba! F' 18 a finite
extension of F with basis {1, A, .1. 2 , ... , A- 1 }; in particular, we infer that
(a o + 1)
+ a 1(x + 1) + ... +
b 1(x
+ ... + bn - . + 1)'-1;
As a final simplification, let us replace x + 1 by:some new symbol A, so that
bo
llb k
ao
145
1)" = O,
. [F':F] = n = degf(x) .
. To recapitulate: iff(x) E F[x] is an irreducible polynomial over F, then
there exists a finite extension P of F, such that [P: F] = degf(x), in which
f(x) has a root. Moreover, F' is a simple algebraic extenSon generated by
a root of f(x). (Admittedly, sorne work could be .saved by an appeal to
Theorems 7-21 and 7;-25, bjlt our object here is to present an alternative
approach to the slibject.)
We pause now to examine two concrete examples of the ideas just
presented.
. Example 7-12. Consider Z2: the fild of integers modulo 2, and the polynomial (x) = x 3 + X + 1 E Z2[ x]. Since neither of the elements O and 1
is a root of x 3 + x + l,f(x) must be irreducible in Z2[XJ.
Theorem 7-27 thus guarantees the existence of an extension of Z2'
. specifical1y, the field Z2[X ]/(J(x), in which the given polynomial has a root.
Denoting this root by A, the discussion aboye tells us that
{a + bA + c,.l2/a,b, e E Z2}
Z2[X]/(J(x)
.1.2 , 1
+ A + A2 },
.1. 3
-(A
1)
= A + 1,
..14
..1 2
146
POLYNOMIAL RINGS
+b+
e)
aA
(a
+ b)A 2 =
1.
a+b
+e=
1,
0,
+ b = 0,
+ A + A2)-1
a
withsolutiona
b. O,e = l;therefore,{l
= A2
3
It is worth noting that x + x + 1 factors completely into linear
factors in Z2[X]/(J(x) and has the three roots A., A2, and A + A2:
x 3 + x + 1 = (x - A)(X - A2)(X - (A + A2).
Example 7-13. The quadratic polynomial X2 + 1 is irreducible in R"'[x].
F or, if X2 + 1 were reducible, it would be of the form
X2
+ 1 = (ax + b)(ex + d)
2
= aex + (ad + be)x +
bd,
be
O.
1) is described by
:1 ~t'"
R"'[x]/(x2
1)
(a
+ bA) + (e + dA) =
(a
bA)(e
+ dA)
= (ae
=
bd)
(ae - bd)
(a
~t~) + (b + d)A
+ (ad + be)A +
+ (ad + bd)A.
bd(A 2
147
1)
The similarity of these formulas to the usual rules for addition and multiplication of complex numbers should be apparent. As a matter of fact,
R#[x]/(x2 + 1) is isomorphic to the field e of complex numbers under the
a + bi. Tbis
mapping 11>: R#[x]/(x2 + 1) ~ e given by q)(a + bA)
provides an elegant way of construoting e from R#.
= Q[x]/(x 2
2)
= {a + bAla, b E Q; A2
= (x
A)(X 2
.[2) (x +
3)
.[2)(x
.[2.)
3).
2 ;;:: O};
146
POLYNOMIAL RINGS
+b+
e)
aA
(a
+ b)A 2 =
1.
a+b
+e=
1,
0,
+ b = 0,
+ A + A2)-1
a
withsolutiona
b. O,e = l;therefore,{l
= A2
3
It is worth noting that x + x + 1 factors completely into linear
factors in Z2[X]/(J(x) and has the three roots A., A2, and A + A2:
x 3 + x + 1 = (x - A)(X - A2)(X - (A + A2).
Example 7-13. The quadratic polynomial X2 + 1 is irreducible in R"'[x].
F or, if X2 + 1 were reducible, it would be of the form
X2
+ 1 = (ax + b)(ex + d)
2
= aex + (ad + be)x +
bd,
be
O.
1) is described by
:1 ~t'"
R"'[x]/(x2
1)
(a
+ bA) + (e + dA) =
(a
bA)(e
+ dA)
= (ae
=
bd)
(ae - bd)
(a
~t~) + (b + d)A
+ (ad + be)A +
+ (ad + bd)A.
bd(A 2
147
1)
The similarity of these formulas to the usual rules for addition and multiplication of complex numbers should be apparent. As a matter of fact,
R#[x]/(x2 + 1) is isomorphic to the field e of complex numbers under the
a + bi. Tbis
mapping 11>: R#[x]/(x2 + 1) ~ e given by q)(a + bA)
provides an elegant way of construoting e from R#.
= Q[x]/(x 2
2)
= {a + bAla, b E Q; A2
= (x
A)(X 2
.[2) (x +
3)
.[2)(x
.[2.)
3).
2 ;;:: O};
148
2d2
3)
2cdJ2 = O,
+ 2d2
3 == O::
cd = O.
.'
The lalter equation implies that either ck. O or d = O. But neither c nor d
can be zero, since thi~ would mean that #?" = 3/2 or c2 = 3, which is clearIy
impossibl. According1y, X2 - 3 does;~'t;t split in F 1 [x].
In order to factorf(x) into linear fagtQrs, it becomes necessary to extend
the coefficient field further. We therefqr~' construct a second extension F 2'
where
;,. '.'
;2P~la,PEF1;fl2 - 3 =
The elementsof F 2 can be expressed ~it~tnatively in the form
F2
(a
+ bJ2)
+(e
+ d.J2).J3
= a
+ b.J2 +
O}.
cJ3 + d.j6,
(x - A)(X
= (x -
149
POLYNOMIAL RINGS
!-~..
BeCore presenting the main theorem, two preparatory results of a somewhat technial nature are needed:
Lemma. Letf(x) be an irreducible poIynomiaI in F[x] and r be a root
oC f(x) in sorne extension field K oC F. Then F(r) ~ F[x]/(f(x) under
an isomorphism whereby the element r corresponds to the coset
x'+ (f(x).
Proof. Since the eIement r is algebraic over F, it folIows directly. from
Theorem 7-19 tha~; F(r) ~ F[xJ/(f(x)) via an isomorphism B with the
property that </Ir ';;(ll o nat(f(x))' (As usual, </Ir: F[x] ~ K is the substitution homomorphi~m induced by r.) Regarding the las! statement oC the
Iemma, we necesslij:ly have
.
+ (f(x))).
The chief vale:~~ this lemma ls that it leads almost immediately to the
Collowing theorem~: .'~
.
- ~ .;>:
Theorem7-29. -r(Isomoiphism Extension Theorem): Let (J be an isomorphismfrom thefieldF onto thefieldF'. Afilo,letf(x)
a o + al x +
... + a"xn be an irreducible polynomial in F[x] and f'(y) = (J(a o) +
(J(aJy + ... + (J(anly" he the corresponding polynomial in F'EyJ.
Then, f'(y) is likewise irreducible. Furthermore, if r is a root of f(x}
in sorne extension field oC F and s a rootof f'(y) in some extenson
field of F', then (J can be extended to an isomorphism (J> of F(r) onto
F'(r') with <D(r)
r '.
iig(x) = ii(b o
b1x
for any
to the reader the task of supplying the necessary details that ii is an isomorphism of F[ x ] onto F,[yJ. 1t is important to notice that for any polynomial g(x) in F[x], an element a E F is a root of g(x) iC and only if (J(a) is a
rootoC iig(x). Indeed, if, as before, g(x) = bo + b 1 x + ... + bnx", then,
upon evaluating iig(x) at (J(a),
(g(x))((J(a))
= (J(g(a)),
. from which our assertion foIlows. In particular, we infer that the poIynomiaIs g(x) and iig(x) are simultaneously reducible or irreducible in F[ x]
and F'[y], respectiveIy. ,This being SO,J'(y) = f(x) is irreducible in F'[y J.
148
2d2
3)
2cdJ2 = O,
+ 2d2
3 == O::
cd = O.
.'
The lalter equation implies that either ck. O or d = O. But neither c nor d
can be zero, since thi~ would mean that #?" = 3/2 or c2 = 3, which is clearIy
impossibl. According1y, X2 - 3 does;~'t;t split in F 1 [x].
In order to factorf(x) into linear fagtQrs, it becomes necessary to extend
the coefficient field further. We therefqr~' construct a second extension F 2'
where
;,. '.'
;2P~la,PEF1;fl2 - 3 =
The elementsof F 2 can be expressed ~it~tnatively in the form
F2
(a
+ bJ2)
+(e
+ d.J2).J3
= a
+ b.J2 +
O}.
cJ3 + d.j6,
(x - A)(X
= (x -
149
POLYNOMIAL RINGS
!-~..
BeCore presenting the main theorem, two preparatory results of a somewhat technial nature are needed:
Lemma. Letf(x) be an irreducible poIynomiaI in F[x] and r be a root
oC f(x) in sorne extension field K oC F. Then F(r) ~ F[x]/(f(x) under
an isomorphism whereby the element r corresponds to the coset
x'+ (f(x).
Proof. Since the eIement r is algebraic over F, it folIows directly. from
Theorem 7-19 tha~; F(r) ~ F[xJ/(f(x)) via an isomorphism B with the
property that </Ir ';;(ll o nat(f(x))' (As usual, </Ir: F[x] ~ K is the substitution homomorphi~m induced by r.) Regarding the las! statement oC the
Iemma, we necesslij:ly have
.
+ (f(x))).
The chief vale:~~ this lemma ls that it leads almost immediately to the
Collowing theorem~: .'~
.
- ~ .;>:
Theorem7-29. -r(Isomoiphism Extension Theorem): Let (J be an isomorphismfrom thefieldF onto thefieldF'. Afilo,letf(x)
a o + al x +
... + a"xn be an irreducible polynomial in F[x] and f'(y) = (J(a o) +
(J(aJy + ... + (J(anly" he the corresponding polynomial in F'EyJ.
Then, f'(y) is likewise irreducible. Furthermore, if r is a root of f(x}
in sorne extension field oC F and s a rootof f'(y) in some extenson
field of F', then (J can be extended to an isomorphism (J> of F(r) onto
F'(r') with <D(r)
r '.
iig(x) = ii(b o
b1x
for any
to the reader the task of supplying the necessary details that ii is an isomorphism of F[ x ] onto F,[yJ. 1t is important to notice that for any polynomial g(x) in F[x], an element a E F is a root of g(x) iC and only if (J(a) is a
rootoC iig(x). Indeed, if, as before, g(x) = bo + b 1 x + ... + bnx", then,
upon evaluating iig(x) at (J(a),
(g(x))((J(a))
= (J(g(a)),
. from which our assertion foIlows. In particular, we infer that the poIynomiaIs g(x) and iig(x) are simultaneously reducible or irreducible in F[ x]
and F'[y], respectiveIy. ,This being SO,J'(y) = f(x) is irreducible in F'[y J.
150
a(r) = x
(J(x)),
(J(r') = y
(J'(y)).
r(g(x)
+ (f(x))) =
ag(x)
(J'(y))
(g(x)
F[x]).
(f'(y)).
where <IJ: F(r) --+ F'(r'); this situation is portrayed in the diagram below:
<IJ
:I
~~,F(l
F[x ]/(J(x))
-----?
F'ey]/{J'(y))
Certainly, <IJ is an isomorphism of F(r) onto P(r'), for the individual mappings
a, r, p- 1 are themselves isomorphisms. If a is an arbitrary element of F, then
<lJ(a)
(f(x)))
-..""
+ ...
151
PROBLEMS
f(x) belong to F and F is itself the splitting field of f(x); that is, K = F.
This in turn induces a splitting of the polynomial f'(y) into a product of
linear factors in F'[y], so that K', = F'. Thus, when it happens that n = 0,
the isomorphism a 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 f(x) and f'(y) over isomorphc fields E
and E', provided that the number of roots of roots of f(x) outside of E is
less than n (n ~ 1).
If f(x) E F[x] is a polynomial having n roots outside of F, then not all
of the irreducible factors of f(x) can be linear in F[x] ; for, otherwise,f(x)
would split cbmpletely in F, contrary to assumption. Accordingly, f(x)
must have sorne factor g(x) of degree m > 1 which is irreducible in F[x].
Let g'(y) denote the corresponding irreducible factor of f'(y). Since K is
a splitting field of f(x) over F, g(x) in particular must have a root in K;
call it r. Similar1y, one of the roots of the polynomialf'(y), say 1", is a root
of g'(y) in K'. By Theorem 7-29, a can be extended to an isomorphism a'
between the fields F(r) and F'(r'). Now, K is a splitting field off(x), viewed
as a polynomial with coefficients from F(r); in a like manner, K' can be
regan;led as a splitting field off'(y) over the field F'(r'). Because the number
of roots off(x) lying outside of F(r) is les s than n,the induction hypothesis
permits us to extend a' (itse1f an extension of a) to an is''fuorphism <IJ of
K' onto K. This completes the induction step and the probfofthe theorem
as well, for a has been suitably extended.
With the corollary below, we achieve our objective.'
CoroUary. Any two splitting fields of a nonconstant polynomial
f(x) E F[x] are isomorphic via an isomorphism l> such that the restriction <lJIF is the identity mapping.
PROBLEMS
1. If R is a commutative ring with identity, prove that
a) The set 1 = {(x) e R[[x]]lordf(x) > O} u {O} forros an ideal of the ring
R[[x]]; in fact, 1 = (x).
150
a(r) = x
(J(x)),
(J(r') = y
(J'(y)).
r(g(x)
+ (f(x))) =
ag(x)
(J'(y))
(g(x)
F[x]).
(f'(y)).
where <IJ: F(r) --+ F'(r'); this situation is portrayed in the diagram below:
<IJ
:I
~~,F(l
F[x ]/(J(x))
-----?
F'ey]/{J'(y))
Certainly, <IJ is an isomorphism of F(r) onto P(r'), for the individual mappings
a, r, p- 1 are themselves isomorphisms. If a is an arbitrary element of F, then
<lJ(a)
(f(x)))
-..""
+ ...
151
PROBLEMS
f(x) belong to F and F is itself the splitting field of f(x); that is, K = F.
This in turn induces a splitting of the polynomial f'(y) into a product of
linear factors in F'[y], so that K', = F'. Thus, when it happens that n = 0,
the isomorphism a 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 f(x) and f'(y) over isomorphc fields E
and E', provided that the number of roots of roots of f(x) outside of E is
less than n (n ~ 1).
If f(x) E F[x] is a polynomial having n roots outside of F, then not all
of the irreducible factors of f(x) can be linear in F[x] ; for, otherwise,f(x)
would split cbmpletely in F, contrary to assumption. Accordingly, f(x)
must have sorne factor g(x) of degree m > 1 which is irreducible in F[x].
Let g'(y) denote the corresponding irreducible factor of f'(y). Since K is
a splitting field of f(x) over F, g(x) in particular must have a root in K;
call it r. Similar1y, one of the roots of the polynomialf'(y), say 1", is a root
of g'(y) in K'. By Theorem 7-29, a can be extended to an isomorphism a'
between the fields F(r) and F'(r'). Now, K is a splitting field off(x), viewed
as a polynomial with coefficients from F(r); in a like manner, K' can be
regan;led as a splitting field off'(y) over the field F'(r'). Because the number
of roots off(x) lying outside of F(r) is les s than n,the induction hypothesis
permits us to extend a' (itse1f an extension of a) to an is''fuorphism <IJ of
K' onto K. This completes the induction step and the probfofthe theorem
as well, for a has been suitably extended.
With the corollary below, we achieve our objective.'
CoroUary. Any two splitting fields of a nonconstant polynomial
f(x) E F[x] are isomorphic via an isomorphism l> such that the restriction <lJIF is the identity mapping.
PROBLEMS
1. If R is a commutative ring with identity, prove that
a) The set 1 = {(x) e R[[x]]lordf(x) > O} u {O} forros an ideal of the ring
R[[x]]; in fact, 1 = (x).
152
,l
n..
~n,
together with O,
9. Consider the polynomial domain F[x], where F is a field, and a fixed element
rE F. Show that t,he set of al1 polynomials having r as a' root,
2. For any field F, eonsider the set F(x> eonsisting of all expressions of the form
ak xk = IL.x-'
+ a_.+1x-+l +
oo'
k=-n
. If addition and multiplieation are defined in the obvious way, F(x> becomes a
"i\,ring, known as the ring af extended (formal) .pawer series aver F. Show that F(x>
/('is in faet the field of quotients of the domain F[[x]J. [Hint: Given nEZ+,
\YQc(F[[xJ]) must eontain x-'.J
.3/''t,et R be a eornmutative ring with identity. If R is a local ring, prove that the
~':''''power series ring R[[x]] is also local.
then there exists ari element O -+ rE R sueh that rf(x) = O. [Hint: Assume that
f(x)g(x) = O. Use the polynomials akg(x) to obtain O =1= h(x) E R[x], with
.
deg h(x) < deg f(x), satisfying h(x)f(x) = O.J
153
PROBLEMS
M,
= ker<p" where
<p, : F[x J
10. Regarding thering ofExample 8, Chapter 1, show thatthe polynomal (a, 0)x 2 E R[x]
has infinitely many roots in R[x]..
.
11. Given f(x)
= ao + a1x +
oO,
+ a.x',
iik denotes the usual eomplex eonjugate of dk Verify t,ii'
](x) = iio
+ ii1x +
'oO
where
a) rE C is a root of/(x) if and only if r is a r90t ofJ(x). [Hint':]fr) = 1(r).J
b) If f(x) E R#[xJ ~ C[x] and r is a eomplex root of f(x), th,en'r is also a root
of f(x),,~.';';c
'::::':: . :;.
12. Let R be a eommutative ring with identity and let f(x) E R[~J: The fundion
1: R -+ R defined by taking l(r) = f(r) for every rE R is eal1ea .the palynamial
jimetian indueed by f(x). Assuming that P R denotes the set of all polynomial
. funetions indueed by elements of R[x], prove that
a) PR forms a subring of map R, known as the ring of polynomial funetions on R;
b) the mapping a: R[xJ -+ PR given by a(!(x)) = 1 is a homomorphism of R[xJ
onto PR ;
.
e) if the element rE R is fixed and 1, = {lE P Rll<r) = O}, then Ir is an ideal of
PR'
13. a) When R is an integral domain, show that <;Iistinet polynomials in R[x] induce
distinet polynomial funetions (in other words, the mapping a: R[x J -+ P R is
one-to-one) if ando only if R has an infinite number of elements.
b) Give an example oftwo distinet polynomials whieh induce the same polynomial
funetion.
14. Let R be a eommutative ring with identity anddefine the funetion 15: R[xJ
the so-ealled derivative fonctian, as fo1!ows:
If
f(x) = ao + a1x +
+ a.X' E R[x],
(x)
{a1x
+ a2 r +
oO,
a.X'lak E Z; n ~ 1}
is a prime ideal of Z[x], but not a maximal ideaL Ineidentally, (x) is maximal
. in F[x], where F is a field.
. .
b) Z[x] is not a principal ideal domain. [Hint: Conslder (x, 2), the (maxlmal)
ideal of polynomials with even eonstant terros.J
e) The primary ideal (x, 4) is not the power of any prime ideal of Z[x]. [Hint:
(x, 2) is the only prime ideal eontaining (x,4).J
8. Let P be a prime ideal of R, a eommutative ring with identity. Prove that P[xJ
is a prime ideal of the polynomial ring R[x]. If M is a maximal ideal of R, is
M[xJ a maximal ideal of R[x]?
-+
R[x],
.oO
then
For any f(x), g(x) E R[x] and any rE R, establish that
+ 15g(x).
e) 15(!(x)g(x))
off(x).J
15. Suppose that R is a eommutative ring with identity and let r E R be a root of the
nonzero polynomial f(x) E R[x]. We eall r a multiple raat of f(x) provided that
f(x)
(x - r)'g(x)
(n > 1),
152
,l
n..
~n,
together with O,
9. Consider the polynomial domain F[x], where F is a field, and a fixed element
rE F. Show that t,he set of al1 polynomials having r as a' root,
2. For any field F, eonsider the set F(x> eonsisting of all expressions of the form
ak xk = IL.x-'
+ a_.+1x-+l +
oo'
k=-n
. If addition and multiplieation are defined in the obvious way, F(x> becomes a
"i\,ring, known as the ring af extended (formal) .pawer series aver F. Show that F(x>
/('is in faet the field of quotients of the domain F[[x]J. [Hint: Given nEZ+,
\YQc(F[[xJ]) must eontain x-'.J
.3/''t,et R be a eornmutative ring with identity. If R is a local ring, prove that the
~':''''power series ring R[[x]] is also local.
then there exists ari element O -+ rE R sueh that rf(x) = O. [Hint: Assume that
f(x)g(x) = O. Use the polynomials akg(x) to obtain O =1= h(x) E R[x], with
.
deg h(x) < deg f(x), satisfying h(x)f(x) = O.J
153
PROBLEMS
M,
= ker<p" where
<p, : F[x J
10. Regarding thering ofExample 8, Chapter 1, show thatthe polynomal (a, 0)x 2 E R[x]
has infinitely many roots in R[x]..
.
11. Given f(x)
= ao + a1x +
oO,
+ a.x',
iik denotes the usual eomplex eonjugate of dk Verify t,ii'
](x) = iio
+ ii1x +
'oO
where
a) rE C is a root of/(x) if and only if r is a r90t ofJ(x). [Hint':]fr) = 1(r).J
b) If f(x) E R#[xJ ~ C[x] and r is a eomplex root of f(x), th,en'r is also a root
of f(x),,~.';';c
'::::':: . :;.
12. Let R be a eommutative ring with identity and let f(x) E R[~J: The fundion
1: R -+ R defined by taking l(r) = f(r) for every rE R is eal1ea .the palynamial
jimetian indueed by f(x). Assuming that P R denotes the set of all polynomial
. funetions indueed by elements of R[x], prove that
a) PR forms a subring of map R, known as the ring of polynomial funetions on R;
b) the mapping a: R[xJ -+ PR given by a(!(x)) = 1 is a homomorphism of R[xJ
onto PR ;
.
e) if the element rE R is fixed and 1, = {lE P Rll<r) = O}, then Ir is an ideal of
PR'
13. a) When R is an integral domain, show that <;Iistinet polynomials in R[x] induce
distinet polynomial funetions (in other words, the mapping a: R[x J -+ P R is
one-to-one) if ando only if R has an infinite number of elements.
b) Give an example oftwo distinet polynomials whieh induce the same polynomial
funetion.
14. Let R be a eommutative ring with identity anddefine the funetion 15: R[xJ
the so-ealled derivative fonctian, as fo1!ows:
If
f(x) = ao + a1x +
+ a.X' E R[x],
(x)
{a1x
+ a2 r +
oO,
a.X'lak E Z; n ~ 1}
is a prime ideal of Z[x], but not a maximal ideaL Ineidentally, (x) is maximal
. in F[x], where F is a field.
. .
b) Z[x] is not a principal ideal domain. [Hint: Conslder (x, 2), the (maxlmal)
ideal of polynomials with even eonstant terros.J
e) The primary ideal (x, 4) is not the power of any prime ideal of Z[x]. [Hint:
(x, 2) is the only prime ideal eontaining (x,4).J
8. Let P be a prime ideal of R, a eommutative ring with identity. Prove that P[xJ
is a prime ideal of the polynomial ring R[x]. If M is a maximal ideal of R, is
M[xJ a maximal ideal of R[x]?
-+
R[x],
.oO
then
For any f(x), g(x) E R[x] and any rE R, establish that
+ 15g(x).
e) 15(!(x)g(x))
off(x).J
15. Suppose that R is a eommutative ring with identity and let r E R be a root of the
nonzero polynomial f(x) E R[x]. We eall r a multiple raat of f(x) provided that
f(x)
(x - r)'g(x)
(n > 1),
154
0-+
21. Apply the Eisenstein Criterion to establish that the following polynomials are
irreducibleinQ[xJ:f(x) = X2 + 1,g(x) = X2 - X + 1,andh(x) = 2x s - 6x 3 +
1), g(-x).J
22. Let R be a unique factorization doniih.t and K its field of quotients. Assume that
ab- l E K (where a and b are relativi)(p'rime) is a nonzero root of the polynomial
f(x) = a o + alx + ... + a.X' E R[x;J" Verify that aJa o and bJa.
23. Prove ,the following assertions concr~,ng the polynomial ring Z[x, y]:
a) The ideals (x), (x, y) and (2, x, y) i''e aH prime in Z[x, yJ, but only the last is
maximal.
. :. :'
b) (x, y) = .J(x2, y) = .J(x2, xy, y2):;::;
c) The ideal (xl', xy, y2) is primary iifZ[x, yJ for any integer k E Z+.
d) If l = (x 2, xy), then.JI is a prime i?eal, but lis not primary. [Hint:.JI = (x).J
24. Consider the polynomial domain F[J, yJ, where F forms a field.
a) Show that (x 2, xy, y2) is not a prin'cipal ideal of F[ x, y J.
'
b) Establish the isomorphlsm F[ x, y J/(x + y) ~ F[ xJ.
25. Let the element r be algebraic over the field F and let f(x) E F[ x J be a monic
polynomial such thatf(r) = O. Prove thatf(x) is the mnimum polynomial of r
over F if and only iff(x) is irreducible in F[xJ.
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 prilI!e, F' is a simple extension of F; in fact, F' = F(I') for every
element rE F' - F.
b) Iff(x) E F[xJ is an irreducible polynomial whose degree is relatively prime to
[F':FJ, thenf(x) has no roots in F'.
c) If r E F' is algebraic of degree n, then each element of F(r) has as its degree an
integer dividing n.
d) Given fields Ki (i = 1,2) such that F' 2 Ki 2 F, with [Kl:F] and [K2:FJ
relatively prime integers, necessarily Kl n K 2 = F.
27. Show that the following extension fields of Q are simple extensions and determine
their respective degrees: Q(.j3, .j7), Q(.j3, 0, Q(.ji, ~),
28. a) Prove that the extension field F' = F(r l , r 2, ... , r.), where each element r i is
algebraic over F, forms a finite extension of F. [Hint: If Fi = Fi_l(r;), then
F. = F' and [F':FJ = IT[Fi+1:FiJ.]
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 r E FU is a root of sorne
polynomial f(x) = a o + alx + ... + a.x' E F'exJ; consider the extension
fields K ~ F(a o, al' ... , 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 aH elements in F' which
are algebraic over F constitute a s,ubfield 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):FJ is finite; hence, F(r, s) is an algebraic extension of F.]
Q(J) = {a
+ bJJa, b E Q}.
"'e
+1
of 1 - 2r
3;'2 in Q(r).
34. Letf(x) E F[xJ be an irreducible polynomial and r, s be two roots off(x) in sorne
splitting field. ' Show that F[rJ ~ F[sJ, by a unique isomorphlsm 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[xJ; say
(ri E F', a
-+
O).
154
0-+
21. Apply the Eisenstein Criterion to establish that the following polynomials are
irreducibleinQ[xJ:f(x) = X2 + 1,g(x) = X2 - X + 1,andh(x) = 2x s - 6x 3 +
1), g(-x).J
22. Let R be a unique factorization doniih.t and K its field of quotients. Assume that
ab- l E K (where a and b are relativi)(p'rime) is a nonzero root of the polynomial
f(x) = a o + alx + ... + a.X' E R[x;J" Verify that aJa o and bJa.
23. Prove ,the following assertions concr~,ng the polynomial ring Z[x, y]:
a) The ideals (x), (x, y) and (2, x, y) i''e aH prime in Z[x, yJ, but only the last is
maximal.
. :. :'
b) (x, y) = .J(x2, y) = .J(x2, xy, y2):;::;
c) The ideal (xl', xy, y2) is primary iifZ[x, yJ for any integer k E Z+.
d) If l = (x 2, xy), then.JI is a prime i?eal, but lis not primary. [Hint:.JI = (x).J
24. Consider the polynomial domain F[J, yJ, where F forms a field.
a) Show that (x 2, xy, y2) is not a prin'cipal ideal of F[ x, y J.
'
b) Establish the isomorphlsm F[ x, y J/(x + y) ~ F[ xJ.
25. Let the element r be algebraic over the field F and let f(x) E F[ x J be a monic
polynomial such thatf(r) = O. Prove thatf(x) is the mnimum polynomial of r
over F if and only iff(x) is irreducible in F[xJ.
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 prilI!e, F' is a simple extension of F; in fact, F' = F(I') for every
element rE F' - F.
b) Iff(x) E F[xJ is an irreducible polynomial whose degree is relatively prime to
[F':FJ, thenf(x) has no roots in F'.
c) If r E F' is algebraic of degree n, then each element of F(r) has as its degree an
integer dividing n.
d) Given fields Ki (i = 1,2) such that F' 2 Ki 2 F, with [Kl:F] and [K2:FJ
relatively prime integers, necessarily Kl n K 2 = F.
27. Show that the following extension fields of Q are simple extensions and determine
their respective degrees: Q(.j3, .j7), Q(.j3, 0, Q(.ji, ~),
28. a) Prove that the extension field F' = F(r l , r 2, ... , r.), where each element r i is
algebraic over F, forms a finite extension of F. [Hint: If Fi = Fi_l(r;), then
F. = F' and [F':FJ = IT[Fi+1:FiJ.]
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 r E FU is a root of sorne
polynomial f(x) = a o + alx + ... + a.x' E F'exJ; consider the extension
fields K ~ F(a o, al' ... , 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 aH elements in F' which
are algebraic over F constitute a s,ubfield 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):FJ is finite; hence, F(r, s) is an algebraic extension of F.]
Q(J) = {a
+ bJJa, b E Q}.
"'e
+1
of 1 - 2r
3;'2 in Q(r).
34. Letf(x) E F[xJ be an irreducible polynomial and r, s be two roots off(x) in sorne
splitting field. ' Show that F[rJ ~ F[sJ, by a unique isomorphlsm 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[xJ; say
(ri E F', a
-+
O).
156
EIGHT
Prove that F' = F(r l' r2' ... , r.). As a particular illustration, establish' that
Q(J2,J3) is the splitting field of (x 2 - 2)(x 2 - 3) E Q[x].
'..
: . ~.
~~ ~ ...
";j,
"'.:
rad R =
156
EIGHT
Prove that F' = F(r l' r2' ... , r.). As a particular illustration, establish' that
Q(J2,J3) is the splitting field of (x 2 - 2)(x 2 - 3) E Q[x].
'..
: . ~.
~~ ~ ...
";j,
"'.:
rad R =
158
to Example 5-1, the maximal ideals of Z are precisely the principal ideals
generated by the prime numbers; thus,
rad R
O for all x E X}
{O}, it follows
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 maxirnal ideal s M' of R[[xJ] 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[[xJ]
(n M, x)
(rad R, x).
159
(x),;Jhe
:
i~;';;~d R is O.
1,
158
to Example 5-1, the maximal ideals of Z are precisely the principal ideals
generated by the prime numbers; thus,
rad R
O for all x E X}
{O}, it follows
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 maxirnal ideal s M' of R[[xJ] 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[[xJ]
(n M, x)
(rad R, x).
159
(x),;Jhe
:
i~;';;~d R is O.
1,
160
that, if every ideal of R is finitely generated, then rad R is not only ni! but
nilpotent.
This is a cpnvenient place to also point out that a homomorphic image
of a semisimple rng need not be semisimpleo An explicit example of tbis
situation can easily be obtained from the ring Z of integerso While Z forms
a ring without a Jacobson radical, its homomorphic image Zpn (p a prime;
n > 1) eqntains the nil ideal (p); appealing to Corollary 3 aboye, we see
that Zp. cannot be semisimpleo
Example 8-4.;iConsider F[[x ]], the rng of formal power series over a field
F. From th6';lemma 011 page 116, it is known that an element f(x) =
ao + a 1x + '.o:.~ + a.x + .0. has an inverse in F[[x]] if and only if the
constant ter~~ap =1= 00 This observatin (in conjunction with Theorem 8-2)
implies that:if g(x) = biJ + b1x + o.. + bllx" + ... , then
, rad
F[[~:iJ~'~
{j(x)!
= {j(x)!a o =
O} = (x).
+ 1)
- (r+ l)(a
1) = 1 - ra
+ l)(b +
b - rab
= 1 -
1(1 - b
+ rab)
1,
.,
'0'.
Continuing this theme, let us expre~s the Jacobson radical ofthe qiJotient
rng Rj1 as a function of the radical of R.
;'::
,' ~
5;
+1
"
1) = 1
lo
;,';'
:..
,'
and,
Thus, we have a second proof ofthe fact that the Jacobson radical of F[[xJ]
is the principal ideal generatedby Xo
,
(1
161
rad (R/I)
2 nat (rad R
+ 1), =
R with 1
5;
M}
rad R+ 1
1
'
which is the first part of our theorem (the crucial step requires the inclusion
nJ<;MM 2 1 + rad R)o
'
'
With an eye to proving (2), notice that whenever 1 5; rad R, then
'radR+l
rad (Rjl) 2
Thus, we need only to show the inclusion (rad R)jl 2 rad (Rjl)o To this
purpose, choose the coset a + 1 Erad (Rjl) and let M be an arbitrary
maximal ideal of R. Since 1 5; rad R 5; M, the image natM = Mjl must
be a maximal ideal of the qUbtient rng Rjl (Problem 3, Chapter 5)0 But
'
then,
+ 1 Erad (R/I)
5;
Mjl,
foreing the element ato lie in Mo As this holds for every maxirnal ideal of
R, it follows that a Erad R and so a + 1 E (rad R)/lo All in all, we have
proved that rad (Rjl) 5; (rad R)jl, which, combined with our earlier
inclusion, leads to (2).
Armed with Theorem 8-4, we are in a position to establish:
160
that, if every ideal of R is finitely generated, then rad R is not only ni! but
nilpotent.
This is a cpnvenient place to also point out that a homomorphic image
of a semisimple rng need not be semisimpleo An explicit example of tbis
situation can easily be obtained from the ring Z of integerso While Z forms
a ring without a Jacobson radical, its homomorphic image Zpn (p a prime;
n > 1) eqntains the nil ideal (p); appealing to Corollary 3 aboye, we see
that Zp. cannot be semisimpleo
Example 8-4.;iConsider F[[x ]], the rng of formal power series over a field
F. From th6';lemma 011 page 116, it is known that an element f(x) =
ao + a 1x + '.o:.~ + a.x + .0. has an inverse in F[[x]] if and only if the
constant ter~~ap =1= 00 This observatin (in conjunction with Theorem 8-2)
implies that:if g(x) = biJ + b1x + o.. + bllx" + ... , then
, rad
F[[~:iJ~'~
{j(x)!
= {j(x)!a o =
O} = (x).
+ 1)
- (r+ l)(a
1) = 1 - ra
+ l)(b +
b - rab
= 1 -
1(1 - b
+ rab)
1,
.,
'0'.
Continuing this theme, let us expre~s the Jacobson radical ofthe qiJotient
rng Rj1 as a function of the radical of R.
;'::
,' ~
5;
+1
"
1) = 1
lo
;,';'
:..
,'
and,
Thus, we have a second proof ofthe fact that the Jacobson radical of F[[xJ]
is the principal ideal generatedby Xo
,
(1
161
rad (R/I)
2 nat (rad R
+ 1), =
R with 1
5;
M}
rad R+ 1
1
'
which is the first part of our theorem (the crucial step requires the inclusion
nJ<;MM 2 1 + rad R)o
'
'
With an eye to proving (2), notice that whenever 1 5; rad R, then
'radR+l
rad (Rjl) 2
Thus, we need only to show the inclusion (rad R)jl 2 rad (Rjl)o To this
purpose, choose the coset a + 1 Erad (Rjl) and let M be an arbitrary
maximal ideal of R. Since 1 5; rad R 5; M, the image natM = Mjl must
be a maximal ideal of the qUbtient rng Rjl (Problem 3, Chapter 5)0 But
'
then,
+ 1 Erad (R/I)
5;
Mjl,
foreing the element ato lie in Mo As this holds for every maxirnal ideal of
R, it follows that a Erad R and so a + 1 E (rad R)/lo All in all, we have
proved that rad (Rjl) 5; (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
163
Theorem 8-5. For any ring R, rad R is the smallest ideal l of R such
that the quotient ring Rll is semisimple (in other words, if Rll is a
semisimple ring, then rad R S;;; l).
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 J must be finite.
i=1
so that R cannot be semisimple.. Finally, observe that if the set {Pi} is empty,
then each nonzero element oJ Ris invertible and R is a field (in which case
rad R = { O } ) . ' ( !
CoroUary. The ring Z of,integers is semisimple.
Theorem 8-7. Let {M;},i' F J, be the set of maximal ideals of the ring
R. If, for each i, there !e?Cists an element ai e Mi such that 1 - a i e
rad R - Mi' then {Ma i~ a finite set.
Proof. Suppose that the indx set J is infinite. Then there exists a wellordering ~ of J under which J has no last el~ent. (See Appendix A for
terminology.) For each ieJ, we define li ~"I( li<jMj' Then {l;} forms
a chain of proper ideals of R. 'By hypothesis, we can select an element
ai e Mi such that 1 -, ai e li - Mi' Now the ideal l = u li is also a proper
ideal of R, since 1 ~ 1. By our choice of the li' l is not contained in any
maximal ideal of R. Indeed, suppose that there does exist n index i for
which l S;;; Mi; then,
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 S;;; 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[ [xJ]
of formal power series over a field F has zero prime radical but, as we already
know, a non trivial Jacobson radical.
Perhaps the most striking result of the present chapter is that the prime
radical, although seemingly quite different, is actually equal to the nil
radical of a ringo The lemma below provides the key to establishing this
assertion.
Lemma. Let l be an ideal ofthe ring R. Further, assume that the subset
. S 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 l
an '40 not meet s; any such ideal is necessarily prime.
Proof.Consider the family $' of all ideals J of R such that l S;;; J and
J n S == ifJ. This family is not empty since l itself satisfies the indicated
conditions. Our immediate aim is to show that for any chain of ideals {J;}
in $', their union u Ji also belongs to $'. It has already be'en established
in Theorem 5-2 that the union of a chain of ideals is again an ideal; moreover, since l S;;; Ji for each i, we certainly have l S;;; u Ji' Finally, observe
that '
(u J) n S = u (Ji n S) = u ifJ = ifJ.
The crux of the matter is that Zorn 's Lemma can now be applied to inter
that $' has a maximal element P; this is the ideal that we want.
By definition, P is maximal in the set of ideal s which contain l 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 strictIy larger than P, the ideal (P, a) must
.)
.'
162
163
Theorem 8-5. For any ring R, rad R is the smallest ideal l of R such
that the quotient ring Rll is semisimple (in other words, if Rll is a
semisimple ring, then rad R S;;; l).
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 J must be finite.
i=1
so that R cannot be semisimple.. Finally, observe that if the set {Pi} is empty,
then each nonzero element oJ Ris invertible and R is a field (in which case
rad R = { O } ) . ' ( !
CoroUary. The ring Z of,integers is semisimple.
Theorem 8-7. Let {M;},i' F J, be the set of maximal ideals of the ring
R. If, for each i, there !e?Cists an element ai e Mi such that 1 - a i e
rad R - Mi' then {Ma i~ a finite set.
Proof. Suppose that the indx set J is infinite. Then there exists a wellordering ~ of J under which J has no last el~ent. (See Appendix A for
terminology.) For each ieJ, we define li ~"I( li<jMj' Then {l;} forms
a chain of proper ideals of R. 'By hypothesis, we can select an element
ai e Mi such that 1 -, ai e li - Mi' Now the ideal l = u li is also a proper
ideal of R, since 1 ~ 1. By our choice of the li' l is not contained in any
maximal ideal of R. Indeed, suppose that there does exist n index i for
which l S;;; Mi; then,
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 S;;; 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[ [xJ]
of formal power series over a field F has zero prime radical but, as we already
know, a non trivial Jacobson radical.
Perhaps the most striking result of the present chapter is that the prime
radical, although seemingly quite different, is actually equal to the nil
radical of a ringo The lemma below provides the key to establishing this
assertion.
Lemma. Let l be an ideal ofthe ring R. Further, assume that the subset
. S 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 l
an '40 not meet s; any such ideal is necessarily prime.
Proof.Consider the family $' of all ideals J of R such that l S;;; J and
J n S == ifJ. This family is not empty since l itself satisfies the indicated
conditions. Our immediate aim is to show that for any chain of ideals {J;}
in $', their union u Ji also belongs to $'. It has already be'en established
in Theorem 5-2 that the union of a chain of ideals is again an ideal; moreover, since l S;;; Ji for each i, we certainly have l S;;; u Ji' Finally, observe
that '
(u J) n S = u (Ji n S) = u ifJ = ifJ.
The crux of the matter is that Zorn 's Lemma can now be applied to inter
that $' has a maximal element P; this is the ideal that we want.
By definition, P is maximal in the set of ideal s which contain l 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 strictIy larger than P, the ideal (P, a) must
.)
.'
164
(P, ab)
Po
Two special cases of, this general setting are particularly noteworthy:
S = {1} and l = {OJo In the event S = {l}, the i~1iI P mentioned in the
lemma is actual1y a maximal ideal (in the usuCl<:1eal-theoretic sense);
consequentIy, we have a somewhat different proof onhe facts that (i) every
proper ideal is contained in a maximal ideal and (ii) each maximal ideal is
primeo
The case where l is the zero ideal is the subject ofthe following corollary,
a result which will be utilized on several occasions in thesequel.
Corollary. Let S be a subset of the ring R which is closed under multiplication and does not contain 00 Then there exists an ideal maximal
in the set of ideals disjoint froro 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 l is precisely the nilradical of l :
)7 = n {plp
2, l; P
is a prime ideal}o
Ji,
2 l; P
is a prime ideal} ~
.JI.
The reverse inclusion folIows readily upon noting that ifthere exists a prime
ideal which contains l but not a, then a ~ )7,' since no power of a belongs
toPo
165
As with the case of the Jaco bson radical, the prime radical may be
characterized by its elements; this is brought out by a result promised
earliero
'
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.
Proa! The assertion is aIl b~t obvious upon taking
8-80
'
r= {O} in Theorem
J[n)
n(Pi)o
pil
+ l is
164
(P, ab)
Po
Two special cases of, this general setting are particularly noteworthy:
S = {1} and l = {OJo In the event S = {l}, the i~1iI P mentioned in the
lemma is actual1y a maximal ideal (in the usuCl<:1eal-theoretic sense);
consequentIy, we have a somewhat different proof onhe facts that (i) every
proper ideal is contained in a maximal ideal and (ii) each maximal ideal is
primeo
The case where l is the zero ideal is the subject ofthe following corollary,
a result which will be utilized on several occasions in thesequel.
Corollary. Let S be a subset of the ring R which is closed under multiplication and does not contain 00 Then there exists an ideal maximal
in the set of ideals disjoint froro 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 l is precisely the nilradical of l :
)7 = n {plp
2, l; P
is a prime ideal}o
Ji,
2 l; P
is a prime ideal} ~
.JI.
The reverse inclusion folIows readily upon noting that ifthere exists a prime
ideal which contains l but not a, then a ~ )7,' since no power of a belongs
toPo
165
As with the case of the Jaco bson radical, the prime radical may be
characterized by its elements; this is brought out by a result promised
earliero
'
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.
Proa! The assertion is aIl b~t obvious upon taking
8-80
'
r= {O} in Theorem
J[n)
n(Pi)o
pil
+ l is
166
tells us that
1
+ 1, and,
= (Rad R)/l.
Proof. If aohas an inversein R and al' al' ... , a" are all nilpotent, then the
polynomial f(x) = a o + alx + ... + anxn is tbe sum of irt ,invertible
element ando a nilpotent elemento Hence, f(x) must tself be a'n"invertible
element of R[x J (Problem 5, Chapter 7). ,
k,' ,
Going in the other direction, assume that the polyno~~l f(x) =
a o + alx + ... + an~ER[xJ possesses an ,nverse. That'o is then
in;ert?le in R should be obvous. Forany prime ideal P of R,P[xJ is a
pnme Ideal of R[xJ and the quotient ring R[xJ/P[xJ ~ (R/P}[xJ. Thus,
the homomorpbic image off(x) in (R/P)[x J,
'1
for any field F. That is to say, the polynomal ring F[x J constitutes a semi"
simple ringo
Suppose for the moment that 1 is an ideal of the ring R with 1 ~ Rad R.
Given a idempotent element e =1= O in.,\R, we know that the coset e + 1
will be idempotent in R/I. What is not SQobvious lS that e + 1 =1= 1; tbis
follows from the fact that Rad R containsIi'11onzero idempotents (Corollary
to Theorem 8-2). We are mainly conceined with the converse here: If
u + 1 is a nonzero idempotent of the quotient ring R/I, does there exist an
idempotent e E R for which e + 1 = u + I?
Before becoming involved in this ,discussion, let us give a general
' "
definition.
' ' ' -,
+ P = al + P = ... = al! + P = P;
, I
hence, the elements al' al' ... , an all He in P. As this statement holds for
every prime ideal of R, it follows that al' al' ... , an ERad R. By the corollary
to The'orem 8-8, the elements al' al' ... , an must therefore be nilpotent.
= {O}
P)~
The assertion ofTheorem 8-13 can be improved upon. For the reader
will have little difficulty in now convincing bimself that
Theorem 8-12. For uny ring R, Rad R is the smallest ideal 1 of R such
that the quotient ring R/I is without prime radical.
(a o + P)
+ xf(x) =
2) whenever 1
= Rad R[x].
To round out the picture, two theorems are stated without proof; it will
be o bserved that these take the same form as the corresponding result
established for the Jacobson radical (Theorems 8-4 and 8-5).
1) Rad (R/I);:2 Rad ~
167
Definition 8-3 means just this: the idempotents of R/I can be lifted if
for each element u E R such that ul - U El there exists some element
el
e E R with e - u E l. Although t is surely too much to expect the
lifting ofidempotents to take place for every 1, we shalI see that tbis situation
does occur whenever 1 is a nil ideal (or, equvalently, whenever 1 ~ Rad R).
Let us begin with a lemma, important in itself.
166
tells us that
1
+ 1, and,
= (Rad R)/l.
Proof. If aohas an inversein R and al' al' ... , a" are all nilpotent, then the
polynomial f(x) = a o + alx + ... + anxn is tbe sum of irt ,invertible
element ando a nilpotent elemento Hence, f(x) must tself be a'n"invertible
element of R[x J (Problem 5, Chapter 7). ,
k,' ,
Going in the other direction, assume that the polyno~~l f(x) =
a o + alx + ... + an~ER[xJ possesses an ,nverse. That'o is then
in;ert?le in R should be obvous. Forany prime ideal P of R,P[xJ is a
pnme Ideal of R[xJ and the quotient ring R[xJ/P[xJ ~ (R/P}[xJ. Thus,
the homomorpbic image off(x) in (R/P)[x J,
'1
for any field F. That is to say, the polynomal ring F[x J constitutes a semi"
simple ringo
Suppose for the moment that 1 is an ideal of the ring R with 1 ~ Rad R.
Given a idempotent element e =1= O in.,\R, we know that the coset e + 1
will be idempotent in R/I. What is not SQobvious lS that e + 1 =1= 1; tbis
follows from the fact that Rad R containsIi'11onzero idempotents (Corollary
to Theorem 8-2). We are mainly conceined with the converse here: If
u + 1 is a nonzero idempotent of the quotient ring R/I, does there exist an
idempotent e E R for which e + 1 = u + I?
Before becoming involved in this ,discussion, let us give a general
' "
definition.
' ' ' -,
+ P = al + P = ... = al! + P = P;
, I
hence, the elements al' al' ... , an all He in P. As this statement holds for
every prime ideal of R, it follows that al' al' ... , an ERad R. By the corollary
to The'orem 8-8, the elements al' al' ... , an must therefore be nilpotent.
= {O}
P)~
The assertion ofTheorem 8-13 can be improved upon. For the reader
will have little difficulty in now convincing bimself that
Theorem 8-12. For uny ring R, Rad R is the smallest ideal 1 of R such
that the quotient ring R/I is without prime radical.
(a o + P)
+ xf(x) =
2) whenever 1
= Rad R[x].
To round out the picture, two theorems are stated without proof; it will
be o bserved that these take the same form as the corresponding result
established for the Jacobson radical (Theorems 8-4 and 8-5).
1) Rad (R/I);:2 Rad ~
167
Definition 8-3 means just this: the idempotents of R/I can be lifted if
for each element u E R such that ul - U El there exists some element
el
e E R with e - u E l. Although t is surely too much to expect the
lifting ofidempotents to take place for every 1, we shalI see that tbis situation
does occur whenever 1 is a nil ideal (or, equvalently, whenever 1 ~ Rad R).
Let us begin with a lemma, important in itself.
1.68
Lemma. If e and e' are two idempotent elements ofthe ring R such that
e - el ERad R, then e = el.
el){l - (e + el)
O, it lS enough
to show that 1 - (e + el) is an invertible element of R. Now, one may
write 1 - (e + e') in the form
1 - (e',;4- e1 = (1 - 2e)
+ (e
e),
R/l~
r~t.i:'i
'
Proof Suppose that u ~;l}s any idempotent of RjI. Since 1 S;; J, it follows
U E 1 S;; J. By assumpthat u + J is an idempoterit element of RfJ; u2
tion, there must exist somee2 = e in R such that e + J
u + J, whence
e - u E J. But then
(e
+ 1)
- (u
+ 1)=
e - u
+ 1 E J/l
S;;;
Applying the lemma to the quotient ring R/l, we conclude that the coset
u + 1 = e + land so the idempotents of R/l can be lifted.
The key to showing that the idempotents of RJRad R are liftable is the
circumsta~'ce that certain quadrati~ equations have a solution in the prime
radical of R.
x)(l
4r)
r = O.
x =
Hl . 1
1/2(2r
@r2
(~)r3
... ).
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 ERad R.
CoroUary. For any nil ideal 1 of R, the idempotents of R/l can be lfted.
Proof Because 1
169
S;;
is legitimate.
\f:,et us define a ring R to be primary whenever the zero ideal is a primary
ideljbf R. This readily translates into a state'ment involving the elements
of]{ :.R is a primary ring if and only if every zero divisor of R is nilpotent.
Int~.g.al domains are examples of primary pngs. In general, primary rings
cai'bt:, obtained by constructing quotient rings R/Q, where Q is a primary
ideaLof
R.
- ';-;:'N,s anapplication of the preceding ideas; we shall characterize such
rings;in terms ,of minimal prime ideaIs. (A prime ideal is said to be a mini mal
prime ideal if it is minimal in the set of prime idel:!.ls; in a cornmutative ring
with identity, such ideals are necessarily proper.) .The crucial step in the
proof is the corollary on page 164.
~'}-
Proof F or the first half of the proof, let R' be a primary ringo 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 lS to say, N is a minimal
prime ideal.
The converse is less obvious; in fact, it is easiest to prove the contrapostive form of the converse. Suppose, then, thatR ha~ a mnimal. prime
ideal P which contalns all zero divisors and let a E R be any nonnilpotent
element. We define the set S by
= {r~nlr~P; n ~
O}.
S is easily seen to be closed under multiplication and 1 E S, Notice, particularly, that the Zero element does not le in S, for, otherwise, we would have
ran = O with an =1= O; this mplies that r lS 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 Pi. Since pi S;; R - S S;;; P,
S = P. But a E S,
with P being a mnimal prime ideal, it follows that R
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 al so be
;~
1.68
Lemma. If e and e' are two idempotent elements ofthe ring R such that
e - el ERad R, then e = el.
el){l - (e + el)
O, it lS enough
to show that 1 - (e + el) is an invertible element of R. Now, one may
write 1 - (e + e') in the form
1 - (e',;4- e1 = (1 - 2e)
+ (e
e),
R/l~
r~t.i:'i
'
Proof Suppose that u ~;l}s any idempotent of RjI. Since 1 S;; J, it follows
U E 1 S;; J. By assumpthat u + J is an idempoterit element of RfJ; u2
tion, there must exist somee2 = e in R such that e + J
u + J, whence
e - u E J. But then
(e
+ 1)
- (u
+ 1)=
e - u
+ 1 E J/l
S;;;
Applying the lemma to the quotient ring R/l, we conclude that the coset
u + 1 = e + land so the idempotents of R/l can be lifted.
The key to showing that the idempotents of RJRad R are liftable is the
circumsta~'ce that certain quadrati~ equations have a solution in the prime
radical of R.
x)(l
4r)
r = O.
x =
Hl . 1
1/2(2r
@r2
(~)r3
... ).
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 ERad R.
CoroUary. For any nil ideal 1 of R, the idempotents of R/l can be lfted.
Proof Because 1
169
S;;
is legitimate.
\f:,et us define a ring R to be primary whenever the zero ideal is a primary
ideljbf R. This readily translates into a state'ment involving the elements
of]{ :.R is a primary ring if and only if every zero divisor of R is nilpotent.
Int~.g.al domains are examples of primary pngs. In general, primary rings
cai'bt:, obtained by constructing quotient rings R/Q, where Q is a primary
ideaLof
R.
- ';-;:'N,s anapplication of the preceding ideas; we shall characterize such
rings;in terms ,of minimal prime ideaIs. (A prime ideal is said to be a mini mal
prime ideal if it is minimal in the set of prime idel:!.ls; in a cornmutative ring
with identity, such ideals are necessarily proper.) .The crucial step in the
proof is the corollary on page 164.
~'}-
Proof F or the first half of the proof, let R' be a primary ringo 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 lS to say, N is a minimal
prime ideal.
The converse is less obvious; in fact, it is easiest to prove the contrapostive form of the converse. Suppose, then, thatR ha~ a mnimal. prime
ideal P which contalns all zero divisors and let a E R be any nonnilpotent
element. We define the set S by
= {r~nlr~P; n ~
O}.
S is easily seen to be closed under multiplication and 1 E S, Notice, particularly, that the Zero element does not le in S, for, otherwise, we would have
ran = O with an =1= O; this mplies that r lS 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 Pi. Since pi S;; R - S S;;; P,
S = P. But a E S,
with P being a mnimal prime ideal, it follows that R
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 al so be
;~
170
abandoned, but this seerns unnecessarily elaborate for the purposes in rnind).
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 ringo
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 rnaximal ideals, c1early
requires the presence of an identity). A more useful clue is provided by
Theorem 8-2, wbich asserts that an element a Erad R if and only if 1 - ra
1S invertible for every choice of r in R, or, to put it somewhat differentIy,
the principal ideal (1 ~, ra) = R for all r E R. This latter condition can be
\Hitten as {x - raxlx E.,R} = R for each r in R, and is meaningful in the
absence of a rnultiplicative identity. It thus would appear that Theorem 8-2
constitutes a hopeful starting point to the solution of the problem before
uso 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 elernent a E R the set
la
b - ab = O.
la' or rather R
+ ab - ab = a(a + b - ab) = O.
This means R
a2
With tbis notation, Theorem 8~16 may be rephrased so asto assert that an
element a E R is quasi-regular if and only if there exists sorne second element
b E R for which a o b = O.
It is a simple matter to verify that the pair (R, o) is a semigroup with
identity element O; in particular, one infers from tbis that quasi-inverses
are unique, whenever they exist. An even stronger result is that the quasiregular elernents of R form a group with respect to the circ1e operation.
LastIy, let us 'call attention to the fact that if R possesses a multiplicative
identity 1, then
'
(1 - a)(1 - b) = 1 - a o b.
171
, 1
x(e - b)
(e' - a)(e - b)
=e
- aob
Here are sorne consequences: Every nilpotent elernent of R is quasiregular. Indeed, if an = O, a stiaightforward calculation will establish that
.t.
170
abandoned, but this seerns unnecessarily elaborate for the purposes in rnind).
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 ringo
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 rnaximal ideals, c1early
requires the presence of an identity). A more useful clue is provided by
Theorem 8-2, wbich asserts that an element a Erad R if and only if 1 - ra
1S invertible for every choice of r in R, or, to put it somewhat differentIy,
the principal ideal (1 ~, ra) = R for all r E R. This latter condition can be
\Hitten as {x - raxlx E.,R} = R for each r in R, and is meaningful in the
absence of a rnultiplicative identity. It thus would appear that Theorem 8-2
constitutes a hopeful starting point to the solution of the problem before
uso 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 elernent a E R the set
la
b - ab = O.
la' or rather R
+ ab - ab = a(a + b - ab) = O.
This means R
a2
With tbis notation, Theorem 8~16 may be rephrased so asto assert that an
element a E R is quasi-regular if and only if there exists sorne second element
b E R for which a o b = O.
It is a simple matter to verify that the pair (R, o) is a semigroup with
identity element O; in particular, one infers from tbis that quasi-inverses
are unique, whenever they exist. An even stronger result is that the quasiregular elernents of R form a group with respect to the circ1e operation.
LastIy, let us 'call attention to the fact that if R possesses a multiplicative
identity 1, then
'
(1 - a)(1 - b) = 1 - a o b.
171
, 1
x(e - b)
(e' - a)(e - b)
=e
- aob
Here are sorne consequences: Every nilpotent elernent of R is quasiregular. Indeed, if an = O, a stiaightforward calculation will establish that
.t.
172
which lar = R for every r E R; recasting this in terms of the notion of quasiregularity : .
Definition 8-4. The J-radical J(R) of a rng R, with or without an
identity, is the set
J(R)
If J(R)
.'~~'
+x -
aa' x
= O.
Multiplying this equation by a and using the fact that a2 a' = O, we deduce
tbat a = O, whence J(R) = {O}.
Example 8-9. Consider the ring F[[x]] of formal power series over the
field F. As we know, F[[x]] is a commutative rng with identity in which
an element f(x) = akX' is invertible if and only if ao =1= o. If f(x) belongs
to the principal ideal (x), thenf(x) has zero constant term; hence, (1 - f(x)) - 1
exists in F[[x]]' Takingf(x)' = (1 - f(X)tI, we see thatf(x) o f(x)' = O.
Thus,everymemberof(x)isquasi-regular, whichimpliesthat(x) S J(F[[x]]).
On the other hand, any element not in (x) is invertible and therefore cannot
be in the J-radical. (In general, ir 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 = O, a contradiction.) The implication is that J(F[[ x]]) S (x)
and equality follows:
J(F[[x]]) = (x) = rad F[[x]]'
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 folIows that a2 in particular wiII be quasi-regular. Therefore, we can
obtain an element bE R for which a2 o b = O. But a simple computation
shows that
ao((-a)ob) = (ao(-a))ob = a20 b = O
173
One fincls in .this way that the element a is quasi-regular with quasi-inverse
(-a)
b.
(u
v)
a(x - ux),o.v:
u)v.
Taking stock of the fact that a; b belop,g;to J(R), we can select an element u
such that -(bx) o u = b( -x) o u =:0 'ahd a second element v for which
a(x- ax) o V.= O; in consequeIi.ce, (~' b)x o (u o v) =0. This being the
case, (a - b)x is qua si-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 aIl of these results, the
following deserves to be carried through.
Theorem 8-18. For any rng R, tbe quotient ring RjJ(R) is J-semisimple.
Pro~f. Take a
J(R)) (x
J(R))
ax
J(R)
J(R))
(y
+ J(R)) =
J(R).
But this implies that the element ax o y lies in J(R), and, hence, is quasiregular as a member of R ; say (ax o y) o Z = O, where z, E R. It foIlows from
the associativity. of o that ax is itself quasi-regular in R, with quasi-inverse
y o z. Since tbis 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 tums out that the c1ass of ideals which must replace the maximal
ideals are precisely those ideals whose quotient rings possess a multiplicative
identity.
172
which lar = R for every r E R; recasting this in terms of the notion of quasiregularity : .
Definition 8-4. The J-radical J(R) of a rng R, with or without an
identity, is the set
J(R)
If J(R)
.'~~'
+x -
aa' x
= O.
Multiplying this equation by a and using the fact that a2 a' = O, we deduce
tbat a = O, whence J(R) = {O}.
Example 8-9. Consider the ring F[[x]] of formal power series over the
field F. As we know, F[[x]] is a commutative rng with identity in which
an element f(x) = akX' is invertible if and only if ao =1= o. If f(x) belongs
to the principal ideal (x), thenf(x) has zero constant term; hence, (1 - f(x)) - 1
exists in F[[x]]' Takingf(x)' = (1 - f(X)tI, we see thatf(x) o f(x)' = O.
Thus,everymemberof(x)isquasi-regular, whichimpliesthat(x) S J(F[[x]]).
On the other hand, any element not in (x) is invertible and therefore cannot
be in the J-radical. (In general, ir 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 = O, a contradiction.) The implication is that J(F[[ x]]) S (x)
and equality follows:
J(F[[x]]) = (x) = rad F[[x]]'
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 folIows that a2 in particular wiII be quasi-regular. Therefore, we can
obtain an element bE R for which a2 o b = O. But a simple computation
shows that
ao((-a)ob) = (ao(-a))ob = a20 b = O
173
One fincls in .this way that the element a is quasi-regular with quasi-inverse
(-a)
b.
(u
v)
a(x - ux),o.v:
u)v.
Taking stock of the fact that a; b belop,g;to J(R), we can select an element u
such that -(bx) o u = b( -x) o u =:0 'ahd a second element v for which
a(x- ax) o V.= O; in consequeIi.ce, (~' b)x o (u o v) =0. This being the
case, (a - b)x is qua si-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 aIl of these results, the
following deserves to be carried through.
Theorem 8-18. For any rng R, tbe quotient ring RjJ(R) is J-semisimple.
Pro~f. Take a
J(R)) (x
J(R))
ax
J(R)
J(R))
(y
+ J(R)) =
J(R).
But this implies that the element ax o y lies in J(R), and, hence, is quasiregular as a member of R ; say (ax o y) o Z = O, where z, E R. It foIlows from
the associativity. of o that ax is itself quasi-regular in R, with quasi-inverse
y o z. Since tbis 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 tums out that the c1ass of ideals which must replace the maximal
ideals are precisely those ideals whose quotient rings possess a multiplicative
identity.
174
175
.,.}.
Proof. If the element a.rj J(R), thtm there is sorne x E R such that ax is not
quasi-regular. Corollary 1 asserts the existence of a modular maximal ideal
of R which exc1udes ax and, in consequence, does not contain a.
rl
a = aob
(ab - b) E M.
= n
=f
R, then
ra
Now, let e be an identity element for R modulo M. Then there exist suitable
i E M, rE R and an integer n for which
+ na.
an ideal of R, the sum ra + na E J(R), so that e
e=
As J(R) fonns
ra
- i
J(R).
174
175
.,.}.
Proof. If the element a.rj J(R), thtm there is sorne x E R such that ax is not
quasi-regular. Corollary 1 asserts the existence of a modular maximal ideal
of R which exc1udes ax and, in consequence, does not contain a.
rl
a = aob
(ab - b) E M.
= n
=f
R, then
ra
Now, let e be an identity element for R modulo M. Then there exist suitable
i E M, rE R and an integer n for which
+ na.
an ideal of R, the sum ra + na E J(R), so that e
e=
As J(R) fonns
ra
- i
J(R).
176
+ (xe
e = i - ix
x) E M.
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 f/: J(~h in fact, if 1 + b - lb = O for some b
in R, we would have 1 == O, a cttradictio;o.. When an identity element is
available, all ideals of R are automatically' modular. In tbis situation, the
J-radical will coincide ?Vith the Ja'cpbson radical of R: J{R) = rad R.
If J{R) = R, then th.e rng R. n!ly contain maximal ideals, but no such
ideal can be modular. ndeed,:~~ppose that l is any modular ideal of R,
R, with e acting as an identity foi.,R,.modulo 1. By suppsition, the element
e E J(R), so that e has a quasi-invrse e'. The modularity of l then yields
e == e'e - e' E l, .which implies that l = 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 ideal s 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 Theoren
2-12 for details). Then R, or more precisely, its isomorphic image R x {O},
is an ideal of R' and R'IR !:::! Z; thus, R'IR is semi simple. 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' 5 R,
which implies that J(R) = rad R' .
177
3. Prove that rad R is the largest (in the set-theoretie sense) ideal 1 of R sueh that
1 + a is invertibJe for all a e: l.
4. a) Let the ring R have the property that al! zero divisors le in rad R. If (a) = (b),
show that the elements a and b must be associates.
.
b) Verify that ifthe element a e: rad R and ax
x for some x e: R, then x = O..
5. ~rovethata.P?werseriesf(x) = ao + a1x+ .,. +a.,x" + ... belongstoradR[[x]]
lf and only .,!f lts constant term ao belongs to cad R.
6. Prove the .f,ollowing assertions eonceming semisimple rings:
a) A ring !i.tis semisimple if and only if a :f= O implies that there exists some element
. r e R fdt'\.vhich 1 - ra is not invertible.
b) Every :&:Aunutative regular ring is semisimple.
e) Suppo,s~:that {J,} is a family of ideals of R sueh that R1 , is semisimple for
eaeh i,a~~ n 1i == {O}. ThenRitselfisasemisimplering. [Hint:Theorem 8-5.J
7. 1: R
(1
"
,./,
,i~r.(
..,
cad R
J..etf be a homomorphism from the ring R onto the ring R'. Show that f(rad R) ;;
rad R ' and, whenever kerf ;; rad R, then rad R = f-1(rad R'); do the same for
.RadR..
1 implies that a e: 1.
12. Establish that an ideal 1 of R contains a prime ideal f and only if fo! eaeh 11
a1a2 ... a" = implies that ak E 1 for some k. [Hint: The set
'
PROBLEMS
Unless indicated to thecontrary,allrings are assumed to beeommutativewith identity.
1. Describe the Jacobson radical ofthe ring Zn ofintegers modulo n. [Hint: Consider
the prime factorization of I1.J In particular, show that Zn is semisimple if and
only if 11 is a square-free integer.
2. Prove that F(xJ, the ring of polynomials in x over a field F, is semisimple.
S= {b1b2 .. b"lbkEi1;n~ 1}
is cJosed under multiplication and
a~ S.]
13. Show that the prime radical of a ring R contains the sum of all nilpotent ideals of R.
14. Establish the equivalence ofthe statements below:
a) {a} is the only nilpotent ideal of R;
b) R is without prime radical; that is, Rad R = {O};
e) for any ideals 1 and J of R, 1J = {O} implies that 1 n J
= {O}.
176
+ (xe
e = i - ix
x) E M.
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 f/: J(~h in fact, if 1 + b - lb = O for some b
in R, we would have 1 == O, a cttradictio;o.. When an identity element is
available, all ideals of R are automatically' modular. In tbis situation, the
J-radical will coincide ?Vith the Ja'cpbson radical of R: J{R) = rad R.
If J{R) = R, then th.e rng R. n!ly contain maximal ideals, but no such
ideal can be modular. ndeed,:~~ppose that l is any modular ideal of R,
R, with e acting as an identity foi.,R,.modulo 1. By suppsition, the element
e E J(R), so that e has a quasi-invrse e'. The modularity of l then yields
e == e'e - e' E l, .which implies that l = 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 ideal s 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 Theoren
2-12 for details). Then R, or more precisely, its isomorphic image R x {O},
is an ideal of R' and R'IR !:::! Z; thus, R'IR is semi simple. 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' 5 R,
which implies that J(R) = rad R' .
177
3. Prove that rad R is the largest (in the set-theoretie sense) ideal 1 of R sueh that
1 + a is invertibJe for all a e: l.
4. a) Let the ring R have the property that al! zero divisors le in rad R. If (a) = (b),
show that the elements a and b must be associates.
.
b) Verify that ifthe element a e: rad R and ax
x for some x e: R, then x = O..
5. ~rovethata.P?werseriesf(x) = ao + a1x+ .,. +a.,x" + ... belongstoradR[[x]]
lf and only .,!f lts constant term ao belongs to cad R.
6. Prove the .f,ollowing assertions eonceming semisimple rings:
a) A ring !i.tis semisimple if and only if a :f= O implies that there exists some element
. r e R fdt'\.vhich 1 - ra is not invertible.
b) Every :&:Aunutative regular ring is semisimple.
e) Suppo,s~:that {J,} is a family of ideals of R sueh that R1 , is semisimple for
eaeh i,a~~ n 1i == {O}. ThenRitselfisasemisimplering. [Hint:Theorem 8-5.J
7. 1: R
(1
"
,./,
,i~r.(
..,
cad R
J..etf be a homomorphism from the ring R onto the ring R'. Show that f(rad R) ;;
rad R ' and, whenever kerf ;; rad R, then rad R = f-1(rad R'); do the same for
.RadR..
1 implies that a e: 1.
12. Establish that an ideal 1 of R contains a prime ideal f and only if fo! eaeh 11
a1a2 ... a" = implies that ak E 1 for some k. [Hint: The set
'
PROBLEMS
Unless indicated to thecontrary,allrings are assumed to beeommutativewith identity.
1. Describe the Jacobson radical ofthe ring Zn ofintegers modulo n. [Hint: Consider
the prime factorization of I1.J In particular, show that Zn is semisimple if and
only if 11 is a square-free integer.
2. Prove that F(xJ, the ring of polynomials in x over a field F, is semisimple.
S= {b1b2 .. b"lbkEi1;n~ 1}
is cJosed under multiplication and
a~ S.]
13. Show that the prime radical of a ring R contains the sum of all nilpotent ideals of R.
14. Establish the equivalence ofthe statements below:
a) {a} is the only nilpotent ideal of R;
b) R is without prime radical; that is, Rad R = {O};
e) for any ideals 1 and J of R, 1J = {O} implies that 1 n J
= {O}.
178
PROBLEMS
23. If f is a homomorphism froID the ring R onto the ring R', establish the inelusion
f(J(R) S;;; J(R'); also show that ir kerf S;;; J(R), then J(R) = rl(J(R'o
24. Prove eaeh of the statements below:
a) J(R) eontains every nil ideal of the ring R.
b) J(R) is a semiprime ideal of R. [Hint: Theorem 5-11.]
e) For any ring R, Rad R S;;; J(R)o
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 semisimpleo
e) Let ehar R = n > 00 Prove that if R is without prime radical, then n is a
square-free integero [Hint: Assume that n = p2q for some prime p; then there
exists an element a E R sueh that pqa =1 O, but(pqaV = 00]
25. Ifwe define R = {2n/(2m + 1)ln, m E Z}, then R forms a eommutative ring under
ordinary addition and multiplieationo Show that J(R) = R, whle Rad R = {O}o
[Hint:(~)o(
2m + 1 2( -
)= O]
'I
lo
28. Assume that the ideal I of R consists of elements whieh are quasi-regular modulo
J(R)o (J-Ie say that a is quasi-regular modulo J(R) provided that there exists an
e!ement b E R sueh that a o b E J(R)o Establish that I S;;; J(R)o
19. A ring R is termed a Hilbert ring if eaeh proper prime ideal of R is an interseetion
of maximal idealso Prove that:
a) R is a Hilbert ring if and op,ly if for every proper ideal I of R,
-2n
+ m)
27_ Prove that ifthe element a fE J(R) is idempotent modulo the ideal I (in other words,
(a + I? = a + 1), then a E 1.
26. LetI be an ideal of the ring R. Regarding I as a ring, deduce thatJ(l) = J(R)
179
j'
30. We shall eall an ide'al I of R regular if the quotient ring R/I is a regular ring; in
other words, if for eaeh a E R, there exists an element b E R sueh that a2 b - a E lo
Prove that
o
'~,' :;
a) Every modular maximal ideal of R is regular.
."-,:;~
b) If I I and 12 are both regular ideals of R, then so also is I I 'l1 2 o [Hint: Given
a E R, there exist b, e E R sueh that a2 b - a E I I and (a 2 b - a)2c -'''(a 2 b - a)
belongs to 12 ; rewrite the last expressiono]
,',
e) J(R) = 'I {III is a regular ideal of R}o
""5
[Hint: Assume that a E J(R), but not the right-hand side, so that a rtUor some
regular ideal lo If S = I 'I J(R), then a1b - a E S for some b in',Ro Take
e = ab E J(R)o Then, el - e E So Show that e E S, whieh leads to the
eontradietion that a E lo]
178
PROBLEMS
23. If f is a homomorphism froID the ring R onto the ring R', establish the inelusion
f(J(R) S;;; J(R'); also show that ir kerf S;;; J(R), then J(R) = rl(J(R'o
24. Prove eaeh of the statements below:
a) J(R) eontains every nil ideal of the ring R.
b) J(R) is a semiprime ideal of R. [Hint: Theorem 5-11.]
e) For any ring R, Rad R S;;; J(R)o
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 semisimpleo
e) Let ehar R = n > 00 Prove that if R is without prime radical, then n is a
square-free integero [Hint: Assume that n = p2q for some prime p; then there
exists an element a E R sueh that pqa =1 O, but(pqaV = 00]
25. Ifwe define R = {2n/(2m + 1)ln, m E Z}, then R forms a eommutative ring under
ordinary addition and multiplieationo Show that J(R) = R, whle Rad R = {O}o
[Hint:(~)o(
2m + 1 2( -
)= O]
'I
lo
28. Assume that the ideal I of R consists of elements whieh are quasi-regular modulo
J(R)o (J-Ie say that a is quasi-regular modulo J(R) provided that there exists an
e!ement b E R sueh that a o b E J(R)o Establish that I S;;; J(R)o
19. A ring R is termed a Hilbert ring if eaeh proper prime ideal of R is an interseetion
of maximal idealso Prove that:
a) R is a Hilbert ring if and op,ly if for every proper ideal I of R,
-2n
+ m)
27_ Prove that ifthe element a fE J(R) is idempotent modulo the ideal I (in other words,
(a + I? = a + 1), then a E 1.
26. LetI be an ideal of the ring R. Regarding I as a ring, deduce thatJ(l) = J(R)
179
j'
30. We shall eall an ide'al I of R regular if the quotient ring R/I is a regular ring; in
other words, if for eaeh a E R, there exists an element b E R sueh that a2 b - a E lo
Prove that
o
'~,' :;
a) Every modular maximal ideal of R is regular.
."-,:;~
b) If I I and 12 are both regular ideals of R, then so also is I I 'l1 2 o [Hint: Given
a E R, there exist b, e E R sueh that a2 b - a E I I and (a 2 b - a)2c -'''(a 2 b - a)
belongs to 12 ; rewrite the last expressiono]
,',
e) J(R) = 'I {III is a regular ideal of R}o
""5
[Hint: Assume that a E J(R), but not the right-hand side, so that a rtUor some
regular ideal lo If S = I 'I J(R), then a1b - a E S for some b in',Ro Take
e = ab E J(R)o Then, el - e E So Show that e E S, whieh leads to the
eontradietion that a E lo]
NINE.
a Boole~n ring,
f(x)
O,then ,
181
1, then
whereas if f(x)
1.
; ~:~
The two theorems whichwe shall be primarily eoneemed with are the Stone
Representation Theorem for Boolean rings and Wedderburn's Theorembn
finite division rings. O[ the two, the proof of Wedderburn's Theorem..is
more troublesome 'and occupies the major portion of our effort in '"ti~
chapter. Before embarkirg on this latter task, it is necessary to assemble'~
number of results pertinent to the strueture of finite, fields. OUT aim har
been to spell out all the important details and the reacier may find some of
the arguments ra thei complex.'
.
We first take up the celebrated result of Stone, which asserts that eaeh
.Boolean ring can be represented by a ring of sets. At the outset, let us recall
that by a Boolean ring is meant a ring with identity every element of which
is idempotent. ~t may be well to emphasize that the existen ce of an identity
is frequently omitted in the definition of a Boolean ring; for the applications
which we have in mind, the presence of such an e1ement will be convenient.
(One can show that if the number of elements of a Boolean ring is finite,
then a multiplicative identity always exists.)
Let us pause long enough to indicate several standard examples of
Boolean rings.
'
Example 9-1. The ring Z2 of integers modul 2.
Example 9-2. The ring (P(X), A, n) of subsets of a nonempty set X, with
the usual interpretation of A and n.
Example 9-3. For a less obvious illustration, let R = map(X, Z2), where .
X is an arbitrary nonempty .set. As is eustomary in this setting, the ring
operations are define4 pointwise; that 1S, ifj and g ,are in R, then
(f + g)(x) = f(x)
(fg) (x)
f(x)
'2
g(x)
+2 g(x),
(x E X).
X; hence,j2;
ba == O. Setting a
2a
+ 'a =
a2
a2
=. ,
The relation ab + ba
-ba
-a for any a in R,
ba.
NINE.
a Boole~n ring,
f(x)
O,then ,
181
1, then
whereas if f(x)
1.
; ~:~
The two theorems whichwe shall be primarily eoneemed with are the Stone
Representation Theorem for Boolean rings and Wedderburn's Theorembn
finite division rings. O[ the two, the proof of Wedderburn's Theorem..is
more troublesome 'and occupies the major portion of our effort in '"ti~
chapter. Before embarkirg on this latter task, it is necessary to assemble'~
number of results pertinent to the strueture of finite, fields. OUT aim har
been to spell out all the important details and the reacier may find some of
the arguments ra thei complex.'
.
We first take up the celebrated result of Stone, which asserts that eaeh
.Boolean ring can be represented by a ring of sets. At the outset, let us recall
that by a Boolean ring is meant a ring with identity every element of which
is idempotent. ~t may be well to emphasize that the existen ce of an identity
is frequently omitted in the definition of a Boolean ring; for the applications
which we have in mind, the presence of such an e1ement will be convenient.
(One can show that if the number of elements of a Boolean ring is finite,
then a multiplicative identity always exists.)
Let us pause long enough to indicate several standard examples of
Boolean rings.
'
Example 9-1. The ring Z2 of integers modul 2.
Example 9-2. The ring (P(X), A, n) of subsets of a nonempty set X, with
the usual interpretation of A and n.
Example 9-3. For a less obvious illustration, let R = map(X, Z2), where .
X is an arbitrary nonempty .set. As is eustomary in this setting, the ring
operations are define4 pointwise; that 1S, ifj and g ,are in R, then
(f + g)(x) = f(x)
(fg) (x)
f(x)
'2
g(x)
+2 g(x),
(x E X).
X; hence,j2;
ba == O. Setting a
2a
+ 'a =
a2
a2
=. ,
The relation ab + ba
-ba
-a for any a in R,
ba.
182
el~ment
1 +2 f(a)
= f(l)
+ a El
+zf(a)
2 2,
a E R, we then
have
a = al = a(aa- 1 ) = a2a- 1 = aa- 1 = 1.
This reasoning shows that the only nonzero element of R is the multiplicative
identity; in other words, R = {O, 1}. But any two-element field is isomorphic to 2 2 , The Converse of the theorem is fairly obvious.
This gives rise to another characterization of maximal (equivalently,
prime) ideals in Boolean"rings.
CoroUary. A proper ideal l of a Boolean ring R is a maximal ideal if
and onIy ifR.jl ~ 2 2 ,
Proof. Since the quotient ring Rjl inherits the idempotency condition, Rjl
is itself a Boolean ringo By Theorem 5-5, l is a maximal ideal if and only
if Rjl is a (Boo.1ean) field. An appeal to the aboye theorem now completes
the p~oof.
S;
M, the coset 1
= f(l +
= g(l +
183
M = M,
a)
a
+ M) =
g(M) = O.
We now set our~elves to the prinCipal task, that of showing that each
Boolean ring is essentially a ring of sets.
Theorem 9-3. (Stone Representation Theorem). Any Boolean ring R
is isornorphic to a ring of subsets of sorne fixed seto
Proof. To begin the attac~, let H denote the collection of all hornornorphisrns
of R onto the field 2 2 , ~ext, define a function h: R -+ P(H) by assigning
to each element a E R thQsemembers of H which as sume the value 1 at a;
in other words,
..
"
':
1}
182
el~ment
1 +2 f(a)
= f(l)
+ a El
+zf(a)
2 2,
a E R, we then
have
a = al = a(aa- 1 ) = a2a- 1 = aa- 1 = 1.
This reasoning shows that the only nonzero element of R is the multiplicative
identity; in other words, R = {O, 1}. But any two-element field is isomorphic to 2 2 , The Converse of the theorem is fairly obvious.
This gives rise to another characterization of maximal (equivalently,
prime) ideals in Boolean"rings.
CoroUary. A proper ideal l of a Boolean ring R is a maximal ideal if
and onIy ifR.jl ~ 2 2 ,
Proof. Since the quotient ring Rjl inherits the idempotency condition, Rjl
is itself a Boolean ringo By Theorem 5-5, l is a maximal ideal if and only
if Rjl is a (Boo.1ean) field. An appeal to the aboye theorem now completes
the p~oof.
S;
M, the coset 1
= f(l +
= g(l +
183
M = M,
a)
a
+ M) =
g(M) = O.
We now set our~elves to the prinCipal task, that of showing that each
Boolean ring is essentially a ring of sets.
Theorem 9-3. (Stone Representation Theorem). Any Boolean ring R
is isornorphic to a ring of subsets of sorne fixed seto
Proof. To begin the attac~, let H denote the collection of all hornornorphisrns
of R onto the field 2 2 , ~ext, define a function h: R -+ P(H) by assigning
to each element a E R thQsemembers of H which as sume the value 1 at a;
in other words,
..
"
':
1}
184
185
= {aERlh(~);:~::
0} = {O}.
AH the pieces faH into place and we s.~:~that the ring R i8 isomorphic to a
subring (namely, the sll,bring h(R) oqI:~JH), 8, tl).
"/-;.'l
Proof. Let 1 be a rigbt ideal and J a left ideal of the regular ring R. Since
the inclusion IJ S;; 1 () J always obtains, we have only to show that any
element a of 1 tl J is in U. By the regularity of R, aa' a = a for some a' E R.
Rere a'a is a me~ber ofthe left ideal J, so that the product a(a'a) E IJ, from
which one infers that a E U.
For the converse, we assume that the indicated condition holds and
]2roceed to establish that R is regular. Let a be an arbitrary elemenf of R.
Then (in the absence of an identity) the right ideal generate9 by a is the set
aR + Za
{ar + nalr E R, n E Z}. By virtue oC our hypothesis,
+ Za
;: (aR
+ Za)
() R = (aR
+ Za)R
a E aR
tl
Ra = aR2a.
'
0l
2.fdl.
.
"c:;', (ab) = (a)
,-? '.
lS eqUlvalent t.q,,tegularity.
~
tl
(b),
a, bE R
:,~~,!
.l::,
aR
~hus; there exists an element a' E R such that a = aa'a, making R a regular
~S.
= aR,
Proof.. S~ppos~ that R is regular and aR ii> a principal right ideal.. If the
element a E R lS such that aa'a = a, then
(aa')2 = (aa'a)a'
= aa',
= 00' is !dempot~nt ..
so that e
From a .= ea E eR, we deduce that aR S;; eR.
Rowever~ e ;: aa E aR, Yleldtng eR S;; aR and the subsequent equ:ality.
~ Turmng matte~s around, let us now assume that every principal. right
Ideal of R has an ldempotent generator: Given an element a E R 'ch
'd
' oose
~ I empotent e such that ~R = eR: Then, for suitable r, s in R, the equatlons a = er and e = as wIIl be sattsfied. But this implies that
asa'= ea = e(er) = er = a,
whence R forms a regular ringo
!he elementary lemma just proved enables us to throw new light on
theldeaI structure of regular rings.
.
184
185
= {aERlh(~);:~::
0} = {O}.
AH the pieces faH into place and we s.~:~that the ring R i8 isomorphic to a
subring (namely, the sll,bring h(R) oqI:~JH), 8, tl).
"/-;.'l
Proof. Let 1 be a rigbt ideal and J a left ideal of the regular ring R. Since
the inclusion IJ S;; 1 () J always obtains, we have only to show that any
element a of 1 tl J is in U. By the regularity of R, aa' a = a for some a' E R.
Rere a'a is a me~ber ofthe left ideal J, so that the product a(a'a) E IJ, from
which one infers that a E U.
For the converse, we assume that the indicated condition holds and
]2roceed to establish that R is regular. Let a be an arbitrary elemenf of R.
Then (in the absence of an identity) the right ideal generate9 by a is the set
aR + Za
{ar + nalr E R, n E Z}. By virtue oC our hypothesis,
+ Za
;: (aR
+ Za)
() R = (aR
+ Za)R
a E aR
tl
Ra = aR2a.
'
0l
2.fdl.
.
"c:;', (ab) = (a)
,-? '.
lS eqUlvalent t.q,,tegularity.
~
tl
(b),
a, bE R
:,~~,!
.l::,
aR
~hus; there exists an element a' E R such that a = aa'a, making R a regular
~S.
= aR,
Proof.. S~ppos~ that R is regular and aR ii> a principal right ideal.. If the
element a E R lS such that aa'a = a, then
(aa')2 = (aa'a)a'
= aa',
= 00' is !dempot~nt ..
so that e
From a .= ea E eR, we deduce that aR S;; eR.
Rowever~ e ;: aa E aR, Yleldtng eR S;; aR and the subsequent equ:ality.
~ Turmng matte~s around, let us now assume that every principal. right
Ideal of R has an ldempotent generator: Given an element a E R 'ch
'd
' oose
~ I empotent e such that ~R = eR: Then, for suitable r, s in R, the equatlons a = er and e = as wIIl be sattsfied. But this implies that
asa'= ea = e(er) = er = a,
whence R forms a regular ringo
!he elementary lemma just proved enables us to throw new light on
theldeaI structure of regular rings.
.
186
Theorem 9,5. The sum of two principal right (left) ideal s of a regular
ring R is itself a principal right (left) ideal.
.JI,
Proo! Consider the right ideal aR + bR. As reference to the lemma will
confirm, aR is of the form aR = eR, with e an idempotent. We also ha ve
eR + bR = {ex + bylx,YER}
=
=
+ eby + (1
{e(x + by) + (1
{ex
= {ez
- e)bylx, y
R}
= eR + (1 - e)bR.
eg
ge
Theorem 9-7. If F is a finite field, then F has exactly p" elements for
sorne prime p and n E Z+.
= ef(1 - e) = 0,
= f(1 - e)e = O.
As bothf = gfe gR and g = f(1 - e) EfR, the ideal fR = gR. One finds
in this way that the sum aR + bR = eR + gR.
. , ,Our remaining objective is to demonstrate that eR + gR = (e + g)R.
Tle reasoning proceeds as follows. Since e + g e eR + gR, the inclusion . ..1
(e.
g)R S;; eR + gR evidentIy holds. On the other hand, the element
.~ e2 + ge = (e + g)e E (e + g)R,and,atthesametime,g = eg + g2 =
(~.;.+ g)g E (e + g)R, which forces eR + gR S;; (e + g)R. Thisleads to the i.
t:gp.ality eR + gR = (e + g)R and in turn to aR + bR = (e + g)R, t
concluding the proof.
.JI,
.JI
- e)bylx, y E R}
(1 - e)bylz, y E R}
187
'~'.
.JI.)
186
Theorem 9,5. The sum of two principal right (left) ideal s of a regular
ring R is itself a principal right (left) ideal.
.JI,
Proo! Consider the right ideal aR + bR. As reference to the lemma will
confirm, aR is of the form aR = eR, with e an idempotent. We also ha ve
eR + bR = {ex + bylx,YER}
=
=
+ eby + (1
{e(x + by) + (1
{ex
= {ez
- e)bylx, y
R}
= eR + (1 - e)bR.
eg
ge
Theorem 9-7. If F is a finite field, then F has exactly p" elements for
sorne prime p and n E Z+.
= ef(1 - e) = 0,
= f(1 - e)e = O.
As bothf = gfe gR and g = f(1 - e) EfR, the ideal fR = gR. One finds
in this way that the sum aR + bR = eR + gR.
. , ,Our remaining objective is to demonstrate that eR + gR = (e + g)R.
Tle reasoning proceeds as follows. Since e + g e eR + gR, the inclusion . ..1
(e.
g)R S;; eR + gR evidentIy holds. On the other hand, the element
.~ e2 + ge = (e + g)e E (e + g)R,and,atthesametime,g = eg + g2 =
(~.;.+ g)g E (e + g)R, which forces eR + gR S;; (e + g)R. Thisleads to the i.
t:gp.ality eR + gR = (e + g)R and in turn to aR + bR = (e + g)R, t
concluding the proof.
.JI,
.JI
- e)bylx, y E R}
(1 - e)bylz, y E R}
187
'~'.
.JI.)
188
Proo! The n.onzer.o elements .of F f.orm a muItiplicative gr.oup .of .order
p" .....: 1, s.o tht f.or each such a E F, we must have a P'- l = 1. Since
aP'
a ==
a(aP"-l -
Proo! Consider the splitting fieId F' of tbe p.oIynomial f(x) ,= xpn - x in
Zp [x]. Since F' contains Zp' it has finite cbaracteristic p. Now, the
derivative .off(x) is
1),
of(x) = p"xP''-
it f.olJ.ows tbat every eIement a E F (zer.o .or n.onzer.o) satisfies t):e equati.on
aP" - a = O.
'
189
1.
1 =
By virtue of ProbIem 15, Cbapter 7, this mean s that th~ poIynomiaI has n.o
repeated r.o.ots in F ' . Let the subset F'!;;; F' c.onsist orihe p' distinct roots
off(x)inF':
- ,
CoroUary 1. Any finite field F with p. elements is the splitting field .oC ,
p
= x ' - XE Zp[X].i,
the'p.olyn.omiaIf(x)
.
~
,~_Proo! We begin by recalIing that f(x) can have at m.ost aegf(x) = p~,:;;!
.~:.:.distinct r.ots in F. But p' r.o.ots are already kn.own, namely, the elements;\
}\;of F; whence, the p.olyn.omiaI f(x} splits c.ompIetely int.o linear fact.ors in,,':;}
\:;'F[x]. Needless t.o say, it cann.ot split in any pr.oper subfield .of F, f.or n?".;
;li,,~"pr.oper subfieId c.ontairi~ p' eIements. Thus, we c.onclude that F lS th~ 'l.
;:;t'~plitting field .ofj(x) .over Zp and
'. ',;
~I,l\.;,',r ',,'
~:<"'~~~
p
'!:5Z
x ' - x == TI (x - r),
\~;i~'
, F = {a
risF
aP"(bp"r = ab-
'.
{.S;:b,
Corollary2. Any two finite fields having the: same number of elements
Corollary. F.or any finite field Fand positive integer n, Jbere exists an
extension fieId .of F of degree n.
are is.om.orphic.
Before going on, let us digress to p.oint .out that this Iemma has an
interesting appHcati.on in the the.ory .of numbers. We wish to esta.bli~h w?at,
in the lterature, goes by the name of Wilson's Theorem; t.o Wlt, If P IS a
prime number, then
(p - 1)! == -1 (modp).
Since this holds trivialIy f.or p = 2, .one may assume that p > 2 ;thus, p is
an .odd prime. We talce f.or F the field Zp' Then, from ab.ove, every nonzero
1 e Z p [x] :
'
eIement of Z p is a ro.ot of the p.olynomial xP- 1
Thus, each eIement .of F is also a r.o.ot .of the polynomial x p"'
t.o say that F !;;; F'. FinalIy, .observe that
mn
x, which is
Theorem 9-8. F.or any prime number p and positive integer n, there
exists a fieId (unique up t.o isomorphism) with exactly p" eIements.
'
Putting x
'Si
t;'
(ab-IV"
XP-l
bV" = a P"
.'f
":"
"! .
188
Proo! The n.onzer.o elements .of F f.orm a muItiplicative gr.oup .of .order
p" .....: 1, s.o tht f.or each such a E F, we must have a P'- l = 1. Since
aP'
a ==
a(aP"-l -
Proo! Consider the splitting fieId F' of tbe p.oIynomial f(x) ,= xpn - x in
Zp [x]. Since F' contains Zp' it has finite cbaracteristic p. Now, the
derivative .off(x) is
1),
of(x) = p"xP''-
it f.olJ.ows tbat every eIement a E F (zer.o .or n.onzer.o) satisfies t):e equati.on
aP" - a = O.
'
189
1.
1 =
By virtue of ProbIem 15, Cbapter 7, this mean s that th~ poIynomiaI has n.o
repeated r.o.ots in F ' . Let the subset F'!;;; F' c.onsist orihe p' distinct roots
off(x)inF':
- ,
CoroUary 1. Any finite field F with p. elements is the splitting field .oC ,
p
= x ' - XE Zp[X].i,
the'p.olyn.omiaIf(x)
.
~
,~_Proo! We begin by recalIing that f(x) can have at m.ost aegf(x) = p~,:;;!
.~:.:.distinct r.ots in F. But p' r.o.ots are already kn.own, namely, the elements;\
}\;of F; whence, the p.olyn.omiaI f(x} splits c.ompIetely int.o linear fact.ors in,,':;}
\:;'F[x]. Needless t.o say, it cann.ot split in any pr.oper subfield .of F, f.or n?".;
;li,,~"pr.oper subfieId c.ontairi~ p' eIements. Thus, we c.onclude that F lS th~ 'l.
;:;t'~plitting field .ofj(x) .over Zp and
'. ',;
~I,l\.;,',r ',,'
~:<"'~~~
p
'!:5Z
x ' - x == TI (x - r),
\~;i~'
, F = {a
risF
aP"(bp"r = ab-
'.
{.S;:b,
Corollary2. Any two finite fields having the: same number of elements
Corollary. F.or any finite field Fand positive integer n, Jbere exists an
extension fieId .of F of degree n.
are is.om.orphic.
Before going on, let us digress to p.oint .out that this Iemma has an
interesting appHcati.on in the the.ory .of numbers. We wish to esta.bli~h w?at,
in the lterature, goes by the name of Wilson's Theorem; t.o Wlt, If P IS a
prime number, then
(p - 1)! == -1 (modp).
Since this holds trivialIy f.or p = 2, .one may assume that p > 2 ;thus, p is
an .odd prime. We talce f.or F the field Zp' Then, from ab.ove, every nonzero
1 e Z p [x] :
'
eIement of Z p is a ro.ot of the p.olynomial xP- 1
Thus, each eIement .of F is also a r.o.ot .of the polynomial x p"'
t.o say that F !;;; F'. FinalIy, .observe that
mn
x, which is
Theorem 9-8. F.or any prime number p and positive integer n, there
exists a fieId (unique up t.o isomorphism) with exactly p" eIements.
'
Putting x
'Si
t;'
(ab-IV"
XP-l
bV" = a P"
.'f
":"
"! .
190
Proof. Let E be a finite field with p" elements and E* be its multiplicative
group of nonzero elements; this group has order p" - 1. The argument
about to be presented hinges on finding an element in E* of order
h = p" - 1. To this end, we first consider the prime factorization of h:
h =
q'q~'
.,. if",m,
where th~ q are distinct primes and r E Z+. For i = 1,2, ... , m, set
h = h/q. Now, there exists a nonzero element a E E which is not a root
of the polynomial X" - 1 E E[x] ; for this polynomial has at most h distinct
roots in E and h < h, the number of nonzero elements of E.
Next, take
b -and define b = b1 b2
a"I.,
,
wh ere
.. b m
-- qr, (. := .'"
1 2' .... m)
We certainly have
bf'
a7
= 1,
bh1
=;,:.bi'b~'
"
... b~l.
191
= [E':
E]
n.
190
Proof. Let E be a finite field with p" elements and E* be its multiplicative
group of nonzero elements; this group has order p" - 1. The argument
about to be presented hinges on finding an element in E* of order
h = p" - 1. To this end, we first consider the prime factorization of h:
h =
q'q~'
.,. if",m,
where th~ q are distinct primes and r E Z+. For i = 1,2, ... , m, set
h = h/q. Now, there exists a nonzero element a E E which is not a root
of the polynomial X" - 1 E E[x] ; for this polynomial has at most h distinct
roots in E and h < h, the number of nonzero elements of E.
Next, take
b -and define b = b1 b2
a"I.,
,
wh ere
.. b m
-- qr, (. := .'"
1 2' .... m)
We certainly have
bf'
a7
= 1,
bh1
=;,:.bi'b~'
"
... b~l.
191
= [E':
E]
n.
I
I
192
fk(x) =
(-lima i xa k- i
(k
.~
i=O
To cope with the' problem at hand, we shall also need the following:
Lemma 2. If P:'i~ finite field and O 9= a E F, then there exist elements
a, b E F such tpi..at a = a2 + b2
yp" - y
I1
'
(y - r),
reZp(a)
y = y
I1
This formal identity requires only that y cornmute with all elements rE Zp(a).
Taking stock of the fact thatf o T,. = T,. o J, as well as the relation fP" = J,
we thereby obtain
o = fP - f = foIl (f - 'T,.).
O'i"reZp(a)
epi
~'
kerf= {x
(y - r).
o 'i"reZp(a)
Since r
Proo! We first di#;nse with the case where char F = 2. In this special
situation, F. has 2~Jiements and any element of F satisfies the equation
, x 2 " = X. Thus, every nonzero member a of F is a square and, in particular,
a = a 2" = (a 2"-1)2.
sorne integer k E Z +.
k
r.:, . = a for sorne k, with 2 :::;; k :::;; s,- 1. '
r.
9= a for
In consequence, xai- 1
ak
Proo! Since aP"-l = 1, the element a has finite order as a member of the
multiplicative group R*. Let sbe the/order of a. Then, in the field Zp(a),
each of the s elements 1, a; a2, .:. , a -1 is a root of the polynomial yS y E Zp[xJ. This polynomial can possess at most s I"oots in Zp(a) and
1, a, ... , as - 1 are all'distinct. But xax- 1 E Zia) and c1early
: (xax- 1)S = .xas x- 1 = xx- 1 = 1.
1).
193
Since char F 9= 2, 1 and -1 are necessarily distinct., This implies that, for
eachp Ef(F*), thereexistexactlytwoelementsa1,a 2 in F* withai = a~ = P;
in fact, a 2 = -al' To put it another way,for each pair ofelements al and
-':a'in F*, we get one element which is a square. Hence, halfthe elements
of F* will be squares, call these P1' P2' ... ,Pk' where the integer k =
(p" - 1)/2. Given O 9= a E F, assume that a is not a square and consider
the set
s = {a - p;ji = 1,2, ... , k}.
If it turns out that a - Pi is not a square for any value of i, the set S (which
contains k distinct elements) mustcoincide with the k nonsquares of F*.
But then a willlie in S, yielding a = a - Pi for sorne choice of i; whence
Pi = O, an obvious conttadiction. This being the case, we conc1ude that
a - Pi = Pj for suitable integers i and j, or a = Pi + pj ' Thus, a is the
sum oftwo squares in R and the requisite equation holds.
Corollary. If F is a finite field and O 9= a E F, then there exist elements
a, b in F such that 1 + a2 - ab 2 = O.
I
I
192
fk(x) =
(-lima i xa k- i
(k
.~
i=O
To cope with the' problem at hand, we shall also need the following:
Lemma 2. If P:'i~ finite field and O 9= a E F, then there exist elements
a, b E F such tpi..at a = a2 + b2
yp" - y
I1
'
(y - r),
reZp(a)
y = y
I1
This formal identity requires only that y cornmute with all elements rE Zp(a).
Taking stock of the fact thatf o T,. = T,. o J, as well as the relation fP" = J,
we thereby obtain
o = fP - f = foIl (f - 'T,.).
O'i"reZp(a)
epi
~'
kerf= {x
(y - r).
o 'i"reZp(a)
Since r
Proo! We first di#;nse with the case where char F = 2. In this special
situation, F. has 2~Jiements and any element of F satisfies the equation
, x 2 " = X. Thus, every nonzero member a of F is a square and, in particular,
a = a 2" = (a 2"-1)2.
sorne integer k E Z +.
k
r.:, . = a for sorne k, with 2 :::;; k :::;; s,- 1. '
r.
9= a for
In consequence, xai- 1
ak
Proo! Since aP"-l = 1, the element a has finite order as a member of the
multiplicative group R*. Let sbe the/order of a. Then, in the field Zp(a),
each of the s elements 1, a; a2, .:. , a -1 is a root of the polynomial yS y E Zp[xJ. This polynomial can possess at most s I"oots in Zp(a) and
1, a, ... , as - 1 are all'distinct. But xax- 1 E Zia) and c1early
: (xax- 1)S = .xas x- 1 = xx- 1 = 1.
1).
193
Since char F 9= 2, 1 and -1 are necessarily distinct., This implies that, for
eachp Ef(F*), thereexistexactlytwoelementsa1,a 2 in F* withai = a~ = P;
in fact, a 2 = -al' To put it another way,for each pair ofelements al and
-':a'in F*, we get one element which is a square. Hence, halfthe elements
of F* will be squares, call these P1' P2' ... ,Pk' where the integer k =
(p" - 1)/2. Given O 9= a E F, assume that a is not a square and consider
the set
s = {a - p;ji = 1,2, ... , k}.
If it turns out that a - Pi is not a square for any value of i, the set S (which
contains k distinct elements) mustcoincide with the k nonsquares of F*.
But then a willlie in S, yielding a = a - Pi for sorne choice of i; whence
Pi = O, an obvious conttadiction. This being the case, we conc1ude that
a - Pi = Pj for suitable integers i and j, or a = Pi + pj ' Thus, a is the
sum oftwo squares in R and the requisite equation holds.
Corollary. If F is a finite field and O 9= a E F, then there exist elements
a, b in F such that 1 + a2 - ab 2 = O.
194
195
After this preparation, we now undertake the task of proving the theorem
which serves as the focal point ofthe present chapter.
Proa! Suppose, for purposes of contradiction, that the theorem is not true
for all finite division rings. Let R have minimal order among the set of
noncomrp.utative division rings, so that any division ring with fewer elements
tban R will be commutative.
Before becoming involved in the technical argument of the proof, let us
note that if the elements a, b E R satisfy abk = bka, but ab =1= ba, then
bk E cent R. For, consider the centralizer of bk in R:
CW) = {x E Rlxb k = bkx}.
from which we derive abo = bOa. Again in the light of our opening remarks,
since a commutes with b" but not with b, necessarily b" E cent R.
We now assert that whenever an element y of R satisfies y" = 1, then it
must be of the form y = r , where O ~ i ~ n - 1. Indeed, the extension
field (cent R)(y) (although awkward, the notation conveys the point) contains
at most n roots of the polynomial z" - 1. But, since r is of prime order n,
the elements 1, r, r2 , , r"-l comprise n distinct roots of z" - 1 in tbis
field. These remarks should make it plain that y = r for some i. In passing,
we might also observe that, because y E cent R, (cent R)(y) = centRo
With reference to Theorem 9-9, inasmuch as cent R constitutes a finite
field, its multiplicative group of nonzero elements must be cyc1ic; say with
generator S. Accordingly, a" = si, b" = Si for suitable integers j and l.
Furthermore,. n divides neither j nor l. To see this, suppose that j = nk;
then, a" = si = s"k, whence a"(s-~" = 1. As the element s lies in cent R,
we would have (as-k)" = 1. But the preceding paragraph then yields
as- k = r for some integer i, or a = rsk E cent R, which is impossible. In
a similar fashion, one is able to estabHsh tbat n does not divide l. We now
set e = d,.d = bi . Then,
cm
cent R}
so, by induction, x"-l ax -(n-1) = al<!'-1. Since we know that n is prim; 'the
Little Fermat Theorem (Problem 10, Chapter 4) tells us that there exisfs an
integer u satisfying k"-l = 1 + un. Therefore,
al<!'-1 = a 1 + u" = aaun = ra,
where r = (a")U E cent R. Setting b = x"-l, one gets bab -1 = ra. Now,
the element x fj cent R, as xax- 1 =1= a, so that b cannot be long to cent R
by the minimal nature of n. From the observation at the beginning of the
theorem, we thus conc1ude that ba =1= abo The implication of aIl this is that
r =1= 1. On the other hand, since r and a" both He in the center of R,
(r- j/)"
(r")-jl
Pausing for a moment to tidy up, let us point-out tbat the reasoning so
far has succeeded in producing two elements e, dE R with the following
properties :
1) e" = d" = ex E cent R,
2) ed = tde, with t E cent R,
3) t =1= 1, but t" = 1.
From these relations, we may easily compute (e- 1 d)". In this connection,
notice that
194
195
After this preparation, we now undertake the task of proving the theorem
which serves as the focal point ofthe present chapter.
Proa! Suppose, for purposes of contradiction, that the theorem is not true
for all finite division rings. Let R have minimal order among the set of
noncomrp.utative division rings, so that any division ring with fewer elements
tban R will be commutative.
Before becoming involved in the technical argument of the proof, let us
note that if the elements a, b E R satisfy abk = bka, but ab =1= ba, then
bk E cent R. For, consider the centralizer of bk in R:
CW) = {x E Rlxb k = bkx}.
from which we derive abo = bOa. Again in the light of our opening remarks,
since a commutes with b" but not with b, necessarily b" E cent R.
We now assert that whenever an element y of R satisfies y" = 1, then it
must be of the form y = r , where O ~ i ~ n - 1. Indeed, the extension
field (cent R)(y) (although awkward, the notation conveys the point) contains
at most n roots of the polynomial z" - 1. But, since r is of prime order n,
the elements 1, r, r2 , , r"-l comprise n distinct roots of z" - 1 in tbis
field. These remarks should make it plain that y = r for some i. In passing,
we might also observe that, because y E cent R, (cent R)(y) = centRo
With reference to Theorem 9-9, inasmuch as cent R constitutes a finite
field, its multiplicative group of nonzero elements must be cyc1ic; say with
generator S. Accordingly, a" = si, b" = Si for suitable integers j and l.
Furthermore,. n divides neither j nor l. To see this, suppose that j = nk;
then, a" = si = s"k, whence a"(s-~" = 1. As the element s lies in cent R,
we would have (as-k)" = 1. But the preceding paragraph then yields
as- k = r for some integer i, or a = rsk E cent R, which is impossible. In
a similar fashion, one is able to estabHsh tbat n does not divide l. We now
set e = d,.d = bi . Then,
cm
cent R}
so, by induction, x"-l ax -(n-1) = al<!'-1. Since we know that n is prim; 'the
Little Fermat Theorem (Problem 10, Chapter 4) tells us that there exisfs an
integer u satisfying k"-l = 1 + un. Therefore,
al<!'-1 = a 1 + u" = aaun = ra,
where r = (a")U E cent R. Setting b = x"-l, one gets bab -1 = ra. Now,
the element x fj cent R, as xax- 1 =1= a, so that b cannot be long to cent R
by the minimal nature of n. From the observation at the beginning of the
theorem, we thus conc1ude that ba =1= abo The implication of aIl this is that
r =1= 1. On the other hand, since r and a" both He in the center of R,
(r- j/)"
(r")-jl
Pausing for a moment to tidy up, let us point-out tbat the reasoning so
far has succeeded in producing two elements e, dE R with the following
properties :
1) e" = d" = ex E cent R,
2) ed = tde, with t E cent R,
3) t =1= 1, but t" = 1.
From these relations, we may easily compute (e- 1 d)". In this connection,
notice that
196
t 1 + 2 e- 3 d3
"'....:
.,.,
~-Ite
,.:,.]
= (dc- 1d- 1 )e =
d(d- 1 e}e- 1
+ dX 1 +
cdx::Y
e2(1
+ xi -
O =1= 2e 2 = e(e
+ dx 1
O,
!Xxi)
dX 1
edx2 =
O. To
O,
197
Proo! To begin with, we assert that R can possess no, nonzero nilpotent
elements. Indeed, if x =1= O, the condition xn(x) = x necessarily implies that
xm =1= O for all m ~ 1. Now, suppose that e is any idempotent elemeI1t of
R; then, for any x E R,
, ,
.
(xe - exe)2
=:
(ex
exe)2
O,
196
t 1 + 2 e- 3 d3
"'....:
.,.,
~-Ite
,.:,.]
= (dc- 1d- 1 )e =
d(d- 1 e}e- 1
+ dX 1 +
cdx::Y
e2(1
+ xi -
O =1= 2e 2 = e(e
+ dx 1
O,
!Xxi)
dX 1
edx2 =
O. To
O,
197
Proo! To begin with, we assert that R can possess no, nonzero nilpotent
elements. Indeed, if x =1= O, the condition xn(x) = x necessarily implies that
xm =1= O for all m ~ 1. Now, suppose that e is any idempotent elemeI1t of
R; then, for any x E R,
, ,
.
(xe - exe)2
=:
(ex
exe)2
O,
198
PROBLEMS
so that xe - exe = O = ex- exe. Therefore, ex = exe = xe, in consequence of which e E cent R. It follows that every idempotent of R must
be in the center.
Given that a E 1, with an = a (n > 1), it is easy to show that e = an - 1
is an idempotent element of R:
where r' = an - 2 ra. Since ar' e 1, this shows that ra el also, making 1 a
two-sided ideal of R.
Lemma 2. Let R be,a J-ring. For all a, b e R, the element ab
in radR.
ba les
199
Corollary. Let R be a ring with the pr.operty that every nonzero subring
of R forms a division ringo Then R is a field.
Proo! Observe first that the ring R has prime characteristic. Indeed, iC R
were of characteristic zero, it would contain a proper subfield isomorphic
to Q and, hence, a proper subring isomorphic to Z. Since the ring Z of
integers is not a division ring, we obtain a contradiction.
Now, let S be the subring of R generated by any nonzero element a e R.
Then S oonsists of all polynomials in a over the prime subfield of R; that
is to say, S = Zp[aJ, for some prime p. Since the e1ement a- 1 e S, a- 1
must be a polynomial in a, which implies that a is a root oC some polynomial
with coefficients from Zp. In consequence, S forms a simple algebraic
extension (field) of Zp. By Theorem 7-26, we also know that S is a finite
field. This being the case, anCa) = a, where n(a) is the number oC elements
in S. From Jacobson's result, it follows that R is necessarily commutative;
hence, a field.
There are a number of other fairIy general assumptions which at a
glance seem quite far removed from commutativity, bui when imposed on
a given ring render it commutative. In this connection, we might mention
without proof
Theorem 9-14. (Herstein). Let R be a ring with the property that, for
each x E R, there exists an integer n(x) > 1 dep.ending on x such that
xn(x) x E cent R. Then R is commutative.
We have noted that in a J-ring some positive power of every element
lies in the center. This provides another path along which to proceed to
comntutativity.
Theo,rem 9-15. (Herstein). Let R be a ring with the property that for
each x E R there exists a positive integer n(x) depending on x such that
xn(x) E cent R. If R contains no nonzeronil ideals, then R is commutative.
PROBLEMS
1. a) If the BooIean ring R has at least three elements, prove that every nonzero
eIement except the identity is a zero divisor oC R.
b) Verify that the idempotent eIements oC any commutative rlng with identity of
characteristic 2 Corm a BooIean subring.
2. Show that any ring R (not necessarily with identity) in which each element is
idempotent can be imbedded in a BooIean ringo [Hint: Let R'
R x Zz and
mimic the argument oC Theorem 2-U.]
3. a) Let R be a commutative ring with identity. and S the set oC all idempotent
elements of R. Define a new sum of a and b in S by taking
+' b = a + b -
2ab.
198
PROBLEMS
so that xe - exe = O = ex- exe. Therefore, ex = exe = xe, in consequence of which e E cent R. It follows that every idempotent of R must
be in the center.
Given that a E 1, with an = a (n > 1), it is easy to show that e = an - 1
is an idempotent element of R:
where r' = an - 2 ra. Since ar' e 1, this shows that ra el also, making 1 a
two-sided ideal of R.
Lemma 2. Let R be,a J-ring. For all a, b e R, the element ab
in radR.
ba les
199
Corollary. Let R be a ring with the pr.operty that every nonzero subring
of R forms a division ringo Then R is a field.
Proo! Observe first that the ring R has prime characteristic. Indeed, iC R
were of characteristic zero, it would contain a proper subfield isomorphic
to Q and, hence, a proper subring isomorphic to Z. Since the ring Z of
integers is not a division ring, we obtain a contradiction.
Now, let S be the subring of R generated by any nonzero element a e R.
Then S oonsists of all polynomials in a over the prime subfield of R; that
is to say, S = Zp[aJ, for some prime p. Since the e1ement a- 1 e S, a- 1
must be a polynomial in a, which implies that a is a root oC some polynomial
with coefficients from Zp. In consequence, S forms a simple algebraic
extension (field) of Zp. By Theorem 7-26, we also know that S is a finite
field. This being the case, anCa) = a, where n(a) is the number oC elements
in S. From Jacobson's result, it follows that R is necessarily commutative;
hence, a field.
There are a number of other fairIy general assumptions which at a
glance seem quite far removed from commutativity, bui when imposed on
a given ring render it commutative. In this connection, we might mention
without proof
Theorem 9-14. (Herstein). Let R be a ring with the property that, for
each x E R, there exists an integer n(x) > 1 dep.ending on x such that
xn(x) x E cent R. Then R is commutative.
We have noted that in a J-ring some positive power of every element
lies in the center. This provides another path along which to proceed to
comntutativity.
Theo,rem 9-15. (Herstein). Let R be a ring with the property that for
each x E R there exists a positive integer n(x) depending on x such that
xn(x) E cent R. If R contains no nonzeronil ideals, then R is commutative.
PROBLEMS
1. a) If the BooIean ring R has at least three elements, prove that every nonzero
eIement except the identity is a zero divisor oC R.
b) Verify that the idempotent eIements oC any commutative rlng with identity of
characteristic 2 Corm a BooIean subring.
2. Show that any ring R (not necessarily with identity) in which each element is
idempotent can be imbedded in a BooIean ringo [Hint: Let R'
R x Zz and
mimic the argument oC Theorem 2-U.]
3. a) Let R be a commutative ring with identity. and S the set oC all idempotent
elements of R. Define a new sum of a and b in S by taking
+' b = a + b -
2ab.
200
PROBLEMS
With
as the addition, pro ve that S eonstitutes a Boolean ring, known as
the idempotent Boolean ring of Ro
b) In particular, obtain the idempotent Boolean ring of Z120
,
e) Show that the idempotent Boolean ring of any integral domain is isomorphie
to Z20
201
then the ring R is isom~rphie to a subring of the ring (P(X), d, n)o [Hint: Consider
the mappingf: R -+ P(X) defined by fea) = Sao]
11. Assume that a and b are elements of the Boolean ring R with a i bo Deduce the
existenee of a maximal ideal M qf R sueh that a ~ M, but bE Mo
12. Pro ve that any proper ideal of a Boolean ring R is a semiprime ideal o
8. Let Rbe a Boolean ringo For any nonzero element a E R, show that there exists a
maximal ideal M of R such that a ~ M from this, deduce that R is semisimpleo
[Hint: Apply Zorn's Lemrna to the family of all ideals of R which eontain 1 - a,
l'.i, {1 -
alaEI}o
Show that 1 u ]' is the smallest s~~ring of R' in which 1 is a maximal ideal.
14. Silppose that S is a subring of theBoolean ring R. Prove that any homomorphism
f from S onto the field Z2 can be:e~tended to all of R. [Hint: Use Theorern 2-6;
ker f is contained i a maximal ideal M, where R/M ~ Z2o]
.
17. Show that if R is a regular ring, then eent R is also regular. [Hint: Given that
a E centR, then aa' a = a for sorne a' E R; show that axa :::: a, where x = 'a(a')2
belongs to eent Ro]
7. Byan atom of a Boolean ring R is meant an element a =1= Osueh that r ::; a implies
either r = a or r = 00 Prove that
a) the ideal la s' maximal if and only if 1 - a is an atom of R (see Problem 6 for
the definition of la);
b) any maximal ideal eontains all the atoms of R, except at most oneo [Hint: Use
Theorem 9-1.]
W~~
13. Let 1 be a proper ideal of a Boolean ring R and define the set l' by
18. Assuming that R is a regular ring with identity, prove the statements below:
a) if O and 1 are the only idempotent elements of R, then R is a division ring (this
holds, in particular, if R has no divisors of zero);
.
b) ir R is of positive eharaeteristic n, then n is a square-free integer;
e) R has nO'nonzero nilpotent (two-sided) ideals;
d) for every right ideal 1 of R, 1 = ann,(annl)o [Hint: If 1 = eR, where e is
idempotent, thenannl = R(l - e).}
e) If R has no nonzero nilpotent elements, then aR = Ra for every a E Ro [Hint:
Choose a' E R such that aa'a = a; sinee thiddempotent a'a E eentR, ar = ar(a'a)
= sao for any r E R.]
19. If R = R 1 Ei3l R 2 Ei3l ooo Ei3l Rn is the direet sum of a finite number of regular rings
R (i = 1, 2, o, o, n), show that R is also regularo
20;' Verify that the ring L(V) of linear mappings of an n-dimensional vector spaee V
into itself forms a regular ring; in this setting, ring multiplication is taken to be
funetional compositiono [Hint: Starting with O =1= fE L(V), a basis {Xl' 000, x k }
fOI kerfand a basis {Xl> 000, Xk> 000' xn} for V, extend the linearly independent set
{j(Xk+ ), 000 J(xn)} to a basis {y, ooo, Yk,!(Xk+ ), ooo,f(Xn )} for Vo Given any k
elements Z1, 000, Zk E V, define j' E L(V) by j'(y) = Z for 1 ::; i ::; k,!'{J(x) = X
for k + 1 :S i ::; n,]
21. Prove that an integer n > 1 is prime if and only if (n - 1)!
1 is divisible by no
200
PROBLEMS
With
as the addition, pro ve that S eonstitutes a Boolean ring, known as
the idempotent Boolean ring of Ro
b) In particular, obtain the idempotent Boolean ring of Z120
,
e) Show that the idempotent Boolean ring of any integral domain is isomorphie
to Z20
201
then the ring R is isom~rphie to a subring of the ring (P(X), d, n)o [Hint: Consider
the mappingf: R -+ P(X) defined by fea) = Sao]
11. Assume that a and b are elements of the Boolean ring R with a i bo Deduce the
existenee of a maximal ideal M qf R sueh that a ~ M, but bE Mo
12. Pro ve that any proper ideal of a Boolean ring R is a semiprime ideal o
8. Let Rbe a Boolean ringo For any nonzero element a E R, show that there exists a
maximal ideal M of R such that a ~ M from this, deduce that R is semisimpleo
[Hint: Apply Zorn's Lemrna to the family of all ideals of R which eontain 1 - a,
l'.i, {1 -
alaEI}o
Show that 1 u ]' is the smallest s~~ring of R' in which 1 is a maximal ideal.
14. Silppose that S is a subring of theBoolean ring R. Prove that any homomorphism
f from S onto the field Z2 can be:e~tended to all of R. [Hint: Use Theorern 2-6;
ker f is contained i a maximal ideal M, where R/M ~ Z2o]
.
17. Show that if R is a regular ring, then eent R is also regular. [Hint: Given that
a E centR, then aa' a = a for sorne a' E R; show that axa :::: a, where x = 'a(a')2
belongs to eent Ro]
7. Byan atom of a Boolean ring R is meant an element a =1= Osueh that r ::; a implies
either r = a or r = 00 Prove that
a) the ideal la s' maximal if and only if 1 - a is an atom of R (see Problem 6 for
the definition of la);
b) any maximal ideal eontains all the atoms of R, except at most oneo [Hint: Use
Theorem 9-1.]
W~~
13. Let 1 be a proper ideal of a Boolean ring R and define the set l' by
18. Assuming that R is a regular ring with identity, prove the statements below:
a) if O and 1 are the only idempotent elements of R, then R is a division ring (this
holds, in particular, if R has no divisors of zero);
.
b) ir R is of positive eharaeteristic n, then n is a square-free integer;
e) R has nO'nonzero nilpotent (two-sided) ideals;
d) for every right ideal 1 of R, 1 = ann,(annl)o [Hint: If 1 = eR, where e is
idempotent, thenannl = R(l - e).}
e) If R has no nonzero nilpotent elements, then aR = Ra for every a E Ro [Hint:
Choose a' E R such that aa'a = a; sinee thiddempotent a'a E eentR, ar = ar(a'a)
= sao for any r E R.]
19. If R = R 1 Ei3l R 2 Ei3l ooo Ei3l Rn is the direet sum of a finite number of regular rings
R (i = 1, 2, o, o, n), show that R is also regularo
20;' Verify that the ring L(V) of linear mappings of an n-dimensional vector spaee V
into itself forms a regular ring; in this setting, ring multiplication is taken to be
funetional compositiono [Hint: Starting with O =1= fE L(V), a basis {Xl' 000, x k }
fOI kerfand a basis {Xl> 000, Xk> 000' xn} for V, extend the linearly independent set
{j(Xk+ ), 000 J(xn)} to a basis {y, ooo, Yk,!(Xk+ ), ooo,f(Xn )} for Vo Given any k
elements Z1, 000, Zk E V, define j' E L(V) by j'(y) = Z for 1 ::; i ::; k,!'{J(x) = X
for k + 1 :S i ::; n,]
21. Prove that an integer n > 1 is prime if and only if (n - 1)!
1 is divisible by no
202
PROBLEMS
22. Show that GF(p"') is (isomorphie to) a subfield of GF(p") if and only if mln; in
faet, ifmln,then there is exaetly one subfield with p" elements. [Hint: In case mln,
use the faet that a - llak - 1 for k > 1 t6 eonclude that xP'" - xlx P" - x.]
23. Establish the following assertions:
a) given that an irreducible polynomial f(x) E Zp[X], then f(x)lx P " - x if and
.
only if degf(x)ln;
bl if an irreducible polynomialf(x) E Zp[X] has a root in GF(p"), thenf(x) splits
eompletely in GF(p");
e) x}'" - x is the product of all the irreducible monic polynomialsf(x) E Zp[x]
sueh tbat degf(x)ln.
24. If pisan odd prime, prove that the Galois field GF(p") eontains an lement which
is not a square.
25. Let P be a prime ideal of R, a commutative ring with identity. If the quotient
ring ,R./P has only a finite number of eIements, verify that R/P is a Galois field.
26. Prove tbat if F is a finite field and K is a subfield of F, then F forms a simple
extension field of K. [Hint: Any generator of F* will generate F as a vector space
over K.]
34. Show that the assumption of an identity element ls unneeessary in proving that
J-rings are eommutative; in other words, if R is a ring with the property that
for every a E R there is an integer n(a) > 1 for whleh an(a) = a, then R must be
commutative. [Hint: Sineeany idempotenteisin cent R, thesubringS ... eR ... Re
has e for an identity and, hence, is eommutative by Theorem 9-13; then
"-1
(xy - yx)e = O. for all x, y E R; now use the faet that e ( xy - yx)
lS
idempotent.]
35. A ring R is ealled an H-ring if for every x E R there exists an integer n(x) ~ 1
such that x"(x)
x E cent R. Assuming that R ls an H-ring, prove the followmg
assertions:
a) Any homomorphic image of R is againan H-ring.
b) For eaeh x E R, there exist arbitrarily large n for which x"
x E eent R.
e) Al! the idempotentand nilpotent elements of R lie in eent R.
d) If a E R is a zero divisor, then there exists sorne nonzero e E cent R sueh that
ac= O. [Hint: If ab = O, with b O, then e
b"
bE cent R and ac
O;
if e = O, look at the idempotent d = b"-l.]
27. Let F be a finite field with p" elements. Prove that the mapping O' p: F -+ F defined
by taking O'p(a) = aP is an automorphism, the so-eaUed Frobenius automorphism
of F; furthermore, ~ = ir
28. al Suppose that R is a ring with identity (not necessarily commutative). If R has
..
..
{r E Rlra = arlo
l'
203
202
PROBLEMS
22. Show that GF(p"') is (isomorphie to) a subfield of GF(p") if and only if mln; in
faet, ifmln,then there is exaetly one subfield with p" elements. [Hint: In case mln,
use the faet that a - llak - 1 for k > 1 t6 eonclude that xP'" - xlx P" - x.]
23. Establish the following assertions:
a) given that an irreducible polynomial f(x) E Zp[X], then f(x)lx P " - x if and
.
only if degf(x)ln;
bl if an irreducible polynomialf(x) E Zp[X] has a root in GF(p"), thenf(x) splits
eompletely in GF(p");
e) x}'" - x is the product of all the irreducible monic polynomialsf(x) E Zp[x]
sueh tbat degf(x)ln.
24. If pisan odd prime, prove that the Galois field GF(p") eontains an lement which
is not a square.
25. Let P be a prime ideal of R, a commutative ring with identity. If the quotient
ring ,R./P has only a finite number of eIements, verify that R/P is a Galois field.
26. Prove tbat if F is a finite field and K is a subfield of F, then F forms a simple
extension field of K. [Hint: Any generator of F* will generate F as a vector space
over K.]
34. Show that the assumption of an identity element ls unneeessary in proving that
J-rings are eommutative; in other words, if R is a ring with the property that
for every a E R there is an integer n(a) > 1 for whleh an(a) = a, then R must be
commutative. [Hint: Sineeany idempotenteisin cent R, thesubringS ... eR ... Re
has e for an identity and, hence, is eommutative by Theorem 9-13; then
"-1
(xy - yx)e = O. for all x, y E R; now use the faet that e ( xy - yx)
lS
idempotent.]
35. A ring R is ealled an H-ring if for every x E R there exists an integer n(x) ~ 1
such that x"(x)
x E cent R. Assuming that R ls an H-ring, prove the followmg
assertions:
a) Any homomorphic image of R is againan H-ring.
b) For eaeh x E R, there exist arbitrarily large n for which x"
x E eent R.
e) Al! the idempotentand nilpotent elements of R lie in eent R.
d) If a E R is a zero divisor, then there exists sorne nonzero e E cent R sueh that
ac= O. [Hint: If ab = O, with b O, then e
b"
bE cent R and ac
O;
if e = O, look at the idempotent d = b"-l.]
27. Let F be a finite field with p" elements. Prove that the mapping O' p: F -+ F defined
by taking O'p(a) = aP is an automorphism, the so-eaUed Frobenius automorphism
of F; furthermore, ~ = ir
28. al Suppose that R is a ring with identity (not necessarily commutative). If R has
..
..
{r E Rlra = arlo
l'
203
TEN
(- a)()
+ b)(i) = a(i) +
b(z),
(ab)(i)
= a(i)b(}
204
a.)(b l , b 2 ,
.:.,
b~)
(al
bl , a2
(a1b 1, a2 b2 ,
...
b 2, ... ,an
bn ),
,anbn ).
(a E R).
EB R and,
It follows without difficulty from the ring axioms in each component that
the rsulting system comprises a ringo The zero element of EB R is the
+ (b l , b2, ... , bn )
Now, let us define 11 to be the set of all n-tuples (al' a~, ... , a.) E R with the
property that ak = O for k =f i. It is easily 9hecked that 11 constitutes an
'ideal of R, which is isomorphic to the ring R under the assignment
RJ
a(i).
..:,:;;1
Definition 10-1. Let {R} be a family of rings indexed by some set .!l.
The complete direct sum of the rings R , denoted by EB R, consists
of all functions a defined on the index set .!I subject to the condition
tha!. for each element E .!I the functional value a(i) lies inRi;
{ala:.!I -> v R; a(i) E
205
l.
I;
Tbis feature throws us back into the situation described in Chapter 2 (see
page 21). If we invoke Theorem 2-4, it follows that the. ring R is the
direct sum (in the sen se ofOefinitlon 2-4) ofthe ideals li' The point which
we wish to make is that the concept of complete direct sum extends our
previously defined direct sum; in the finite case, the two notions coincide
up to isomorphism of components. The particular ring $0 obtained is
customarily denoted by either 7= 1 EB R or Rl EB Rz EB ... EB R". We
might also mention in passing that if.!l is the positive integers, then EB R
may be viewedas the set of alI infinite sequences (al' az, ... , an , . ) such
that a E R for each i E .!l.
Since the generality of the com plete direct sum confronts the imagination
with sucn a hurdle, we shall seldom have occasion to use it. Certain subrings
ofthe complete direct sum are more manageable and more interesting, 'For
instance, the discrete direct sum of the rings R is the subring of EB R
consisting of those functions which are zero for almost all i; here the phrase
"for almost all i" is short for "for all i with at most a finite number of
TEN
(- a)()
+ b)(i) = a(i) +
b(z),
(ab)(i)
= a(i)b(}
204
a.)(b l , b 2 ,
.:.,
b~)
(al
bl , a2
(a1b 1, a2 b2 ,
...
b 2, ... ,an
bn ),
,anbn ).
(a E R).
EB R and,
It follows without difficulty from the ring axioms in each component that
the rsulting system comprises a ringo The zero element of EB R is the
+ (b l , b2, ... , bn )
Now, let us define 11 to be the set of all n-tuples (al' a~, ... , a.) E R with the
property that ak = O for k =f i. It is easily 9hecked that 11 constitutes an
'ideal of R, which is isomorphic to the ring R under the assignment
RJ
a(i).
..:,:;;1
Definition 10-1. Let {R} be a family of rings indexed by some set .!l.
The complete direct sum of the rings R , denoted by EB R, consists
of all functions a defined on the index set .!I subject to the condition
tha!. for each element E .!I the functional value a(i) lies inRi;
{ala:.!I -> v R; a(i) E
205
l.
I;
Tbis feature throws us back into the situation described in Chapter 2 (see
page 21). If we invoke Theorem 2-4, it follows that the. ring R is the
direct sum (in the sen se ofOefinitlon 2-4) ofthe ideals li' The point which
we wish to make is that the concept of complete direct sum extends our
previously defined direct sum; in the finite case, the two notions coincide
up to isomorphism of components. The particular ring $0 obtained is
customarily denoted by either 7= 1 EB R or Rl EB Rz EB ... EB R". We
might also mention in passing that if.!l is the positive integers, then EB R
may be viewedas the set of alI infinite sequences (al' az, ... , an , . ) such
that a E R for each i E .!l.
Since the generality of the com plete direct sum confronts the imagination
with sucn a hurdle, we shall seldom have occasion to use it. Certain subrings
ofthe complete direct sum are more manageable and more interesting, 'For
instance, the discrete direct sum of the rings R is the subring of EB R
consisting of those functions which are zero for almost all i; here the phrase
"for almost all i" is short for "for all i with at most a finite number of
206
d $
{a E $ R;\a()
Again, if the index set J is taken to be finite, say J = {l, 2, ... , n}, then
the stipulation "for almost all i" is redundant and may be dropped from the
descriptin of d (fl R; in this latter setting,
d $
1I:(a) = a(i).
One can verify that 11:1 is a homomorphism of EB RI onto the ring R,
called the ith component projection. Ir S is any subring of $ R, the
restriction 1I:/IS defines a homomorpbism of S into R and, hence, onto a
subring 11:1(S) of R' The case ofprincipal interest is that in wbich 1I:(S) = R
for each index i; in this event, we call S a subdirect sum of the rings RI'
Let us record these remarks as a formal definition.
Definition 10-2. A subring S of the complete direct sum EB R is
said to be a subdirect sum of the rings R , written S = s $ R, if the
induced projectiol'l 1I:ls: S --? R is an onto mapping for each i. The
subdirect sUrii;:1S nontrivial if none of the mappings 1I:j$ is one-to-one
(hence, S is nqt isomorphic to any R i ).
In effect, a suprlng S 5; $ Ri is a subdirect sum of tl1e rings R if
and only ir, for ea.~h index i, every element of RI appears as the functional
'
'
value at i of sorne ;function ln S.
Definition 10~2,raises a rather significant question: What necessary and
sufficient conditiQI),S upon a ring R will enable us to write it (up to isomorpbism) as the'subdirect sum of more tractable rings R? Up to this
point, everything' has been a matter oC definition and observation; with the
needed preliminar,ies finally compiled, let us make a start at providing an
answer to the a:bove problem.
Lemma. A ring R is isomorphic to a subdirect sum of the rings R if
and only if there exists an isomorphism f: R --? $ R such that, for ,
each i, 11:1o f is a homomorphism of R onto R.
207
' (fl R.
Proo! To start, we assume that R :::=: B $ R. Then there exists an isomorphism f: R --? s EB R such that the "natuml" homomorphisms
11:1 0 f: R --? R are all onto mappings. Using the Fundamental Homomorphism T.heorem, this implies that RJl :::=: R, where 11 = ker(1I: 01).
Note Curther that
ker f = {r E Rl!(r)
= O}
nI.
11 = {O}.
l for all i}
{a E Rla + li =
JI
)'.
for all i}
= nI, = {O}.
We leave the checking oCthe remaining details as an exercise.
Most applications depend more directly on the following version of
Theorem 10-L
CoroDary. A ring R is isomorphic to Si subdirect sum of the quotient
rings RJ1i if and only if R contains a collection oC ideals {11} such that
n li = {O}. Furthermore, the subdirect sum is nontrivial if and only
ir 11 {O} for all i.
206
d $
{a E $ R;\a()
Again, if the index set J is taken to be finite, say J = {l, 2, ... , n}, then
the stipulation "for almost all i" is redundant and may be dropped from the
descriptin of d (fl R; in this latter setting,
d $
1I:(a) = a(i).
One can verify that 11:1 is a homomorphism of EB RI onto the ring R,
called the ith component projection. Ir S is any subring of $ R, the
restriction 1I:/IS defines a homomorpbism of S into R and, hence, onto a
subring 11:1(S) of R' The case ofprincipal interest is that in wbich 1I:(S) = R
for each index i; in this event, we call S a subdirect sum of the rings RI'
Let us record these remarks as a formal definition.
Definition 10-2. A subring S of the complete direct sum EB R is
said to be a subdirect sum of the rings R , written S = s $ R, if the
induced projectiol'l 1I:ls: S --? R is an onto mapping for each i. The
subdirect sUrii;:1S nontrivial if none of the mappings 1I:j$ is one-to-one
(hence, S is nqt isomorphic to any R i ).
In effect, a suprlng S 5; $ Ri is a subdirect sum of tl1e rings R if
and only ir, for ea.~h index i, every element of RI appears as the functional
'
'
value at i of sorne ;function ln S.
Definition 10~2,raises a rather significant question: What necessary and
sufficient conditiQI),S upon a ring R will enable us to write it (up to isomorpbism) as the'subdirect sum of more tractable rings R? Up to this
point, everything' has been a matter oC definition and observation; with the
needed preliminar,ies finally compiled, let us make a start at providing an
answer to the a:bove problem.
Lemma. A ring R is isomorphic to a subdirect sum of the rings R if
and only if there exists an isomorphism f: R --? $ R such that, for ,
each i, 11:1o f is a homomorphism of R onto R.
207
' (fl R.
Proo! To start, we assume that R :::=: B $ R. Then there exists an isomorphism f: R --? s EB R such that the "natuml" homomorphisms
11:1 0 f: R --? R are all onto mappings. Using the Fundamental Homomorphism T.heorem, this implies that RJl :::=: R, where 11 = ker(1I: 01).
Note Curther that
ker f = {r E Rl!(r)
= O}
nI.
11 = {O}.
l for all i}
{a E Rla + li =
JI
)'.
for all i}
= nI, = {O}.
We leave the checking oCthe remaining details as an exercise.
Most applications depend more directly on the following version of
Theorem 10-L
CoroDary. A ring R is isomorphic to Si subdirect sum of the quotient
rings RJ1i if and only if R contains a collection oC ideals {11} such that
n li = {O}. Furthermore, the subdirect sum is nontrivial if and only
ir 11 {O} for all i.
20a
'
'
it being understood that the summation runs over al! primes. At the same
time, Z can be represented as a subdirect sum of the rings ZpI, since the
intersection of the ideals (p2) isalso the zero ideal:
.!
xER#
$ Zp.
p prime
Z ~
209
Sa
$ ZpI'
pprime
. AH the component rings in the ftrst representation are fields, while none is
a field in the second. This shows that a given ring may be representable
as a subdirect sum oC rings having quite differerit properties.
{a,a 2 ,
... ,
a", ... }
is cIosed under multiplication and does not contain O. Thus, there exists a
. prime ideal Po oC R, with Po n So = 0 (corollary on page 164). We assert
that R ~
$ (RIPa ), where the summation ranges over aH the nonnilpotent elements oC R.
20a
'
'
it being understood that the summation runs over al! primes. At the same
time, Z can be represented as a subdirect sum of the rings ZpI, since the
intersection of the ideals (p2) isalso the zero ideal:
.!
xER#
$ Zp.
p prime
Z ~
209
Sa
$ ZpI'
pprime
. AH the component rings in the ftrst representation are fields, while none is
a field in the second. This shows that a given ring may be representable
as a subdirect sum oC rings having quite differerit properties.
{a,a 2 ,
... ,
a", ... }
is cIosed under multiplication and does not contain O. Thus, there exists a
. prime ideal Po oC R, with Po n So = 0 (corollary on page 164). We assert
that R ~
$ (RIPa ), where the summation ranges over aH the nonnilpotent elements oC R.
210
$ (R/l;) by
x+
... ,
U.
= a1a 2
...
aj - 1a j + 1 ... an
nI;.
; 'fj '"
where x
= r;x;.
x + lj
= {Xl + 11 , x 2 + 12 ,
... ,
xn
+':ln ),
j
lj as
;'fj
Et> (R/IJ
I
I
211
Corollary. Let 1 1,1 i, ... , 1" be a finite set of ideals of the ring R with
the property that 1; + lj = R whenever i 1= j. Given any n elements
Xl' x 2 , , Xn E R, there exists sorne x E R such that x - x; E 1; for
i = 1,2, ... , n.
This corollary may be applied to the ring Z of integers and to the
principal ideals (m 1), (m 2 ), ' , (~), where the integers m; are relatively prime
in pairs. One then obtains an old and famous theorem about congruences
which goes by the name of the Chinese Remainder Theorem (the result
being known to Chinese mathematicians as early as A.D. 250):
Theorem 10-7. (Chinese Remainder Theorem). Let m1' m 2 , .... , m n be
positive integers such that gcd (m;, mj) = 1 for i 1= j. If al' a 2 , ... , an
are any n integers, then the system of congruences
x == al (mod m 1 ),
admits a simultaneous solution. Furthermore, tbis solution is unique
modulo m' = m 1m 2 ... mn ,
The hypothesis in Theorem 10-6 is conveniently expressed in terms of
the following: a finite set of ideals 1 1,12 , ... , In of a ring R is said to be
pairwise comaximal (or pairwise relative1y prime, in the older terminology)
if 1; 1= R and 1; + lj = R for i 1= j; when n = 2, we simply term 1 1 and
12 comaximal. Thus, the condition on the ideal s in Theorem 10-6 is that
they be pairwise comaximal. EvidentIy, the definition ofpairwise comaximal
implies that 1; 1= lj for i 1= j, as well as 1; 1= {O} for all i.
If, in the representation of a ring R as a subdirect sum of the rings R,
the "natural" homomorphism ofi:onto R; happens to be an isomorphism
for sorne i, then the representatio~ ls, termed trivial; in the contrary case it
is non trivial. (A nontrivial repres~ntation does not rule out the possibility
that R ~ R; by way of sorne mapping other than the "natural" homomorphism of R onto R;.) A ring.,R is called subdirectly irreducible if there
is no nontrivial representation of R as a subdirect sumo Let us summarize
these remarks in a definition.
Definition 10-3. A ring R is said to be subdirectly irreducible if, in any
representation of R as a subdirect sum ofthe rings R;, at least one ofthe
associated homomorphisms of R onto R; is actual1y an isomorphism;
otherwise, .R is subdirectly reducible.
The corollary to Theorem 10-1 may be taken as asserting that R is
subdirectIy reducible if and only if there exists in R a set of nonzero ideal s
with zero intersection. An equivalent and often handier formulation is the
following: a ring R is subdirectIy irreducible if and only if the intersection
of all the nonzero ideals of R is different from the zero ideal.
210
$ (R/l;) by
x+
... ,
U.
= a1a 2
...
aj - 1a j + 1 ... an
nI;.
; 'fj '"
where x
= r;x;.
x + lj
= {Xl + 11 , x 2 + 12 ,
... ,
xn
+':ln ),
j
lj as
;'fj
Et> (R/IJ
I
I
211
Corollary. Let 1 1,1 i, ... , 1" be a finite set of ideals of the ring R with
the property that 1; + lj = R whenever i 1= j. Given any n elements
Xl' x 2 , , Xn E R, there exists sorne x E R such that x - x; E 1; for
i = 1,2, ... , n.
This corollary may be applied to the ring Z of integers and to the
principal ideals (m 1), (m 2 ), ' , (~), where the integers m; are relatively prime
in pairs. One then obtains an old and famous theorem about congruences
which goes by the name of the Chinese Remainder Theorem (the result
being known to Chinese mathematicians as early as A.D. 250):
Theorem 10-7. (Chinese Remainder Theorem). Let m1' m 2 , .... , m n be
positive integers such that gcd (m;, mj) = 1 for i 1= j. If al' a 2 , ... , an
are any n integers, then the system of congruences
x == al (mod m 1 ),
admits a simultaneous solution. Furthermore, tbis solution is unique
modulo m' = m 1m 2 ... mn ,
The hypothesis in Theorem 10-6 is conveniently expressed in terms of
the following: a finite set of ideals 1 1,12 , ... , In of a ring R is said to be
pairwise comaximal (or pairwise relative1y prime, in the older terminology)
if 1; 1= R and 1; + lj = R for i 1= j; when n = 2, we simply term 1 1 and
12 comaximal. Thus, the condition on the ideal s in Theorem 10-6 is that
they be pairwise comaximal. EvidentIy, the definition ofpairwise comaximal
implies that 1; 1= lj for i 1= j, as well as 1; 1= {O} for all i.
If, in the representation of a ring R as a subdirect sum of the rings R,
the "natural" homomorphism ofi:onto R; happens to be an isomorphism
for sorne i, then the representatio~ ls, termed trivial; in the contrary case it
is non trivial. (A nontrivial repres~ntation does not rule out the possibility
that R ~ R; by way of sorne mapping other than the "natural" homomorphism of R onto R;.) A ring.,R is called subdirectly irreducible if there
is no nontrivial representation of R as a subdirect sumo Let us summarize
these remarks in a definition.
Definition 10-3. A ring R is said to be subdirectly irreducible if, in any
representation of R as a subdirect sum ofthe rings R;, at least one ofthe
associated homomorphisms of R onto R; is actual1y an isomorphism;
otherwise, .R is subdirectly reducible.
The corollary to Theorem 10-1 may be taken as asserting that R is
subdirectIy reducible if and only if there exists in R a set of nonzero ideal s
with zero intersection. An equivalent and often handier formulation is the
following: a ring R is subdirectIy irreducible if and only if the intersection
of all the nonzero ideals of R is different from the zero ideal.
212
Proo! First, suppose that the element r <t ann R v Then ra =f O for sorne
choice of a in R v. Since ra lies in R v, it will serve as a generator for R v ;
that is, R v = (ra). Thus, we can find an eIement s E R satisfying a = ras,
whence the product (1 - rs}a =, O. It follows that 1 - rs E ann R v,' in
consequence of which ann R v is a maximal ideal of R (Problem 2, Chapter
5).
'I
As a special case of part (3) aboye, we might point out that any subdirectly irreducible Boolean ring must be a field, which is c1early isomorphic
to Z2 (Theorem 9-2).
There is a corollary to Theorem 10-9 that wIl be usefullater on.
Corollary. If R V =f {a}, then the annihilator ofthe set ofzero divisors
of R is precisely R v.
Proo! With reference to the theorem, it is enough to prove that
ann (ann R V) = R v Since one always has R v S;; ann (ann R V), let us
concentrate on the reverse inclusion. If a is any nonzero element of
ann (ann R V), then R v 5; (a) and, hence, O =f 'ar E R v for sorne choice of
, r ~ ann R v (in other words, r is not a zero divisor of R). As in the proof
of Theorem 10-9, we can find an element s E R for which 1 - rs E ann R v
This means that a(1 - rs) = O and so a = (ar)s E R v. It follows that
aun (ann R V) 5; R v, which completes the argumen1.
'
213
212
Proo! First, suppose that the element r <t ann R v Then ra =f O for sorne
choice of a in R v. Since ra lies in R v, it will serve as a generator for R v ;
that is, R v = (ra). Thus, we can find an eIement s E R satisfying a = ras,
whence the product (1 - rs}a =, O. It follows that 1 - rs E ann R v,' in
consequence of which ann R v is a maximal ideal of R (Problem 2, Chapter
5).
'I
As a special case of part (3) aboye, we might point out that any subdirectly irreducible Boolean ring must be a field, which is c1early isomorphic
to Z2 (Theorem 9-2).
There is a corollary to Theorem 10-9 that wIl be usefullater on.
Corollary. If R V =f {a}, then the annihilator ofthe set ofzero divisors
of R is precisely R v.
Proo! With reference to the theorem, it is enough to prove that
ann (ann R V) = R v Since one always has R v S;; ann (ann R V), let us
concentrate on the reverse inclusion. If a is any nonzero element of
ann (ann R V), then R v 5; (a) and, hence, O =f 'ar E R v for sorne choice of
, r ~ ann R v (in other words, r is not a zero divisor of R). As in the proof
of Theorem 10-9, we can find an element s E R for which 1 - rs E ann R v
This means that a(1 - rs) = O and so a = (ar)s E R v. It follows that
aun (ann R V) 5; R v, which completes the argumen1.
'
213
214
PROBLEMS
8. Pro ve that an irredundant subdireet sum of a finite number of simple rings is their
direet Sunl.
In the set ofproblems below, aH rings are assumed to be eonmlUtative with identity.
1. Prove each ofthe foIJowing assertions regarding the complete direet sum El:> R:
a) If a and b ar~ elements of El:> R sueh that n(a) = n(b) for eaeh index i,
then a = b.
b) If an element r E R is given for eaeh i, then there exists a unique a E El:> R
satisfying Aa) = r.
e) 'If R =1= 0 for aIJ i, then the ith projection n maps El:> R onto R.
.[
2. Prove that an arbitrary funetion f from a ring R into the complete direet Sunl
El:> R of the rings R is a homomorphism if and only if the eomposition
n o f: R -+ R is itself a homomorphism for each value of i.
3. Consider the complete direct sum
1;
= Oforj
=1=
12. a) Let al> a2' ... ,anbe a finite set ofnonzero elements ofthe principal ideal domain
R sueh that a and aj are relatively prime for i =1= j. If a = lem (al> a2, ... ,' an),
show that RI(a) ~ EB (RI(a.
b) Prove that if the integer n > 1 has the prime faetorization n = p~'~2 ... p~',
i},
El:> R =
then Z" ~
1 El:> J.
1112
In = 11
Il
12
Il ... Il
JI are
In'
= Rn = n(In + l i) S; In + (Il l i) S;
Il .:;; Il
l n-l>
',,1',0',,'
Z;i".R
ni"')
",
5. Establish that each of the given rings has a representation as a subdireet sum of
the ri~gs R{i = 1,2, 3, ... ):
a) zi R = Zpl, where P is a fixed prime.
b) Z; R = Zp, where PI is an odd prime.
,
e) Z.;){ = Z~/(2). (In the situation eonsidered, (2) = {2r + 21nlr E Z.; n E Z}
= 2Z.)
d)
= Z.I(Pi)' where p is an odd prime.
Supps~ that R is isomorphie to a subdireet sum of the rings R under the homomorphismf We say that the subdireet sum is irredundant ir, for eaeh indexj, the
mapping h): R -+ "'j EB Ri defined by hk) = f(r)I'fj El:> R is not one-to-one
(that is, ker h) =1= {O}.
Prove the equivalenee ofthe foIlowing statements:
a) the subdirect sum ' EB R is irredundant;
b) ker (nj o f) ;j2 ni"') ker (n e f) for eaeh indexj;
e) thereexists a eollection ofideals {I} of R sueh that (1) RI ~ Rll, (2) 1 = {O},
and (3)
1 =1= {O} for eaeh index j.
EB Zp,"
13. Let 1 l> 12 , ... , In be a finite set of ideals of the ring R. Prove that
a) the ideals 1 are pairwise eomaximal if and only if their ni! radical s
pairwise eomaximal;
b) if the ideals 11 are pairwise eomaximal, then their produet
4. Prove that a ring R is isomorphic to a subdireet sum of rings R if and only if, for
eaeh i" there exists a homomorphism gl of R onto R sueh that if =1= rE R, then
gl(r) =1= Ofor at least one value ofi. [Hint: Assume that the stated eondition holds.
For fixed rE R, define f,. E L El:> R by f,.(i) = g(r). Now, eonsider the mapping
f: R -+ L El:> 'R in whieh f(r) = f,..]
6.
10. a) Prove that a ring Ris isomorphie to the complete direet Sunl of a finite number
of fields if and only if (i) R eontains only a finite number of ideals and (ii)
rad R = {O}.
b) Prove that a finite ring R is a direet sum offields if and only ifit has no nonzero
nilpotent elements.
11. Demonstrate that the conc1usion ofTheorem 10-6 is false if an infinite numbr of
ideals li are allowed. [Hint: Consider the ring Z and the ideals li = (p), where
215
PROBLEMS
15. Prove that any subdireetly irreducible ring has eharacteristie zero or a pdwer of a
prime. In partir;ular, eonclude that Zn is subdirectly irreducible if and 'only if n
is a power of a prime.
, 16. If R is a subdireetly irreducible ring, show that O and 1 are the only idempotents
of R. [Hint: For an idempotent e E R, consider the principal ideals (e) and
(1 - e).]
17. a) Verify that any subdirectly irreducible Boolean ring is a field.
b) Prove that a ring R is a Boolean ring if and only if R is isomorphie to a subdireet sum offields Z2' [Hint: Theorenl 10-9 and part (a).]
214
PROBLEMS
8. Pro ve that an irredundant subdireet sum of a finite number of simple rings is their
direet Sunl.
In the set ofproblems below, aH rings are assumed to be eonmlUtative with identity.
1. Prove each ofthe foIJowing assertions regarding the complete direet sum El:> R:
a) If a and b ar~ elements of El:> R sueh that n(a) = n(b) for eaeh index i,
then a = b.
b) If an element r E R is given for eaeh i, then there exists a unique a E El:> R
satisfying Aa) = r.
e) 'If R =1= 0 for aIJ i, then the ith projection n maps El:> R onto R.
.[
2. Prove that an arbitrary funetion f from a ring R into the complete direet Sunl
El:> R of the rings R is a homomorphism if and only if the eomposition
n o f: R -+ R is itself a homomorphism for each value of i.
3. Consider the complete direct sum
1;
= Oforj
=1=
12. a) Let al> a2' ... ,anbe a finite set ofnonzero elements ofthe principal ideal domain
R sueh that a and aj are relatively prime for i =1= j. If a = lem (al> a2, ... ,' an),
show that RI(a) ~ EB (RI(a.
b) Prove that if the integer n > 1 has the prime faetorization n = p~'~2 ... p~',
i},
El:> R =
then Z" ~
1 El:> J.
1112
In = 11
Il
12
Il ... Il
JI are
In'
= Rn = n(In + l i) S; In + (Il l i) S;
Il .:;; Il
l n-l>
',,1',0',,'
Z;i".R
ni"')
",
5. Establish that each of the given rings has a representation as a subdireet sum of
the ri~gs R{i = 1,2, 3, ... ):
a) zi R = Zpl, where P is a fixed prime.
b) Z; R = Zp, where PI is an odd prime.
,
e) Z.;){ = Z~/(2). (In the situation eonsidered, (2) = {2r + 21nlr E Z.; n E Z}
= 2Z.)
d)
= Z.I(Pi)' where p is an odd prime.
Supps~ that R is isomorphie to a subdireet sum of the rings R under the homomorphismf We say that the subdireet sum is irredundant ir, for eaeh indexj, the
mapping h): R -+ "'j EB Ri defined by hk) = f(r)I'fj El:> R is not one-to-one
(that is, ker h) =1= {O}.
Prove the equivalenee ofthe foIlowing statements:
a) the subdirect sum ' EB R is irredundant;
b) ker (nj o f) ;j2 ni"') ker (n e f) for eaeh indexj;
e) thereexists a eollection ofideals {I} of R sueh that (1) RI ~ Rll, (2) 1 = {O},
and (3)
1 =1= {O} for eaeh index j.
EB Zp,"
13. Let 1 l> 12 , ... , In be a finite set of ideals of the ring R. Prove that
a) the ideals 1 are pairwise eomaximal if and only if their ni! radical s
pairwise eomaximal;
b) if the ideals 11 are pairwise eomaximal, then their produet
4. Prove that a ring R is isomorphic to a subdireet sum of rings R if and only if, for
eaeh i" there exists a homomorphism gl of R onto R sueh that if =1= rE R, then
gl(r) =1= Ofor at least one value ofi. [Hint: Assume that the stated eondition holds.
For fixed rE R, define f,. E L El:> R by f,.(i) = g(r). Now, eonsider the mapping
f: R -+ L El:> 'R in whieh f(r) = f,..]
6.
10. a) Prove that a ring Ris isomorphie to the complete direet Sunl of a finite number
of fields if and only if (i) R eontains only a finite number of ideals and (ii)
rad R = {O}.
b) Prove that a finite ring R is a direet sum offields if and only ifit has no nonzero
nilpotent elements.
11. Demonstrate that the conc1usion ofTheorem 10-6 is false if an infinite numbr of
ideals li are allowed. [Hint: Consider the ring Z and the ideals li = (p), where
215
PROBLEMS
15. Prove that any subdireetly irreducible ring has eharacteristie zero or a pdwer of a
prime. In partir;ular, eonclude that Zn is subdirectly irreducible if and 'only if n
is a power of a prime.
, 16. If R is a subdireetly irreducible ring, show that O and 1 are the only idempotents
of R. [Hint: For an idempotent e E R, consider the principal ideals (e) and
(1 - e).]
17. a) Verify that any subdirectly irreducible Boolean ring is a field.
b) Prove that a ring R is a Boolean ring if and only if R is isomorphie to a subdireet sum offields Z2' [Hint: Theorenl 10-9 and part (a).]
T
216
ELEVEN
18. Prove that a ring R is subdirectIy irreducible ir and only ir R contains an element
r V with the Collowing two properties:
i) the principal ideal (r V ) has nonzero intersection with every nopzero ideal oC R;
ii) ann (r V) is a maximal ideal of R.
.
[Hint: Assume the conditions and let a 1= O; from (a) () (r V ) 1= {O}; deduce that
.
(r V ) !;; (a).]
19. Prove that the idempotent Boolean ring oC Z. is isomorphic to the Boolean ring
of 2/1 elements, where k is the number of distinct prime divsors oC n. [Hint: Shpw .
tllat Z. has exactly 2k idempotents or that X2 == x (med n) has 2& solutions;Jor
k > 1 use the Chin~se Remainder Theorell.]
,~.l ,
,ir..
(~:,
. ~;:':i ~:,
'(:1 :
..... ',.
In
T
216
ELEVEN
18. Prove that a ring R is subdirectIy irreducible ir and only ir R contains an element
r V with the Collowing two properties:
i) the principal ideal (r V ) has nonzero intersection with every nopzero ideal oC R;
ii) ann (r V) is a maximal ideal of R.
.
[Hint: Assume the conditions and let a 1= O; from (a) () (r V ) 1= {O}; deduce that
.
(r V ) !;; (a).]
19. Prove that the idempotent Boolean ring oC Z. is isomorphic to the Boolean ring
of 2/1 elements, where k is the number of distinct prime divsors oC n. [Hint: Shpw .
tllat Z. has exactly 2k idempotents or that X2 == x (med n) has 2& solutions;Jor
k > 1 use the Chin~se Remainder Theorell.]
,~.l ,
,ir..
(~:,
. ~;:':i ~:,
'(:1 :
..... ',.
In
218
Example 11-2. In the ring of integers, the inclusion (n) S;; (m) implies that
m divides n. Since a nonzero integer can have only a finite number of
distinct divisors, the ring Z evidently satisfies Definition 11-l.
Example 11-3. As a more interesting example, let us show that the
ascending chain condition is satisfied by any principal ideal ring R. For
this purpose, consider an increasing sequence of ideal s of R,
11
S;;
12
S;; S;;
In
S;; .
whence 1m
S;;
In
S;;
1m
S;;
219
1,
In' as desired.
/L'" ~~ \
1 = (~i, a2 ,
. ,
ar )
S;;
In
S;;
1m
S;;
1;
hence,lm = In' Our argurnent shows that every ascending chain of ideals
R terminates at sorne point.
Rings satisfying any one of the three equivalent conditions of Theorern
11-1 (hence, a11 three cohditions) are ca11ed Noetherian rings, in honor of
Ernrny Noether, who first initiated their study. The fact that, when dealing
with Noetherian rings, we can restrict our attention to finitely generated
ideals is of great advantage; the next two results should arnply illustrate
this.
Theorem 11-2. If 1 is an ideal of the Noetherian ring R, then 1 contains
sorne power of its nil radical; that is, (Ji)n S;; 1 for sorne n E Z +.
218
Example 11-2. In the ring of integers, the inclusion (n) S;; (m) implies that
m divides n. Since a nonzero integer can have only a finite number of
distinct divisors, the ring Z evidently satisfies Definition 11-l.
Example 11-3. As a more interesting example, let us show that the
ascending chain condition is satisfied by any principal ideal ring R. For
this purpose, consider an increasing sequence of ideal s of R,
11
S;;
12
S;; S;;
In
S;; .
whence 1m
S;;
In
S;;
1m
S;;
219
1,
In' as desired.
/L'" ~~ \
1 = (~i, a2 ,
. ,
ar )
S;;
In
S;;
1m
S;;
1;
hence,lm = In' Our argurnent shows that every ascending chain of ideals
R terminates at sorne point.
Rings satisfying any one of the three equivalent conditions of Theorern
11-1 (hence, a11 three cohditions) are ca11ed Noetherian rings, in honor of
Ernrny Noether, who first initiated their study. The fact that, when dealing
with Noetherian rings, we can restrict our attention to finitely generated
ideals is of great advantage; the next two results should arnply illustrate
this.
Theorem 11-2. If 1 is an ideal of the Noetherian ring R, then 1 contains
sorne power of its nil radical; that is, (Ji)n S;; 1 for sorne n E Z +.
220
r
I
... + km
n1
n2
+ ... +
nm,
then we must have k ~ n for sorne index, i (i = 1, 2, ... ,m). This implies
that a~' El, hence that the dement a1 1 a~2 :/:, a~m E l. Since aH the generators
of (.JI)n lie in 1, it foHows that (.JI)n S; l.":'
.
~
'.-
b.o +
b 1x
+' ... +
br _ 1x r - 1
bx'.
~~k~
l'
The Hilbert Basis Theorem asserts that if R is a Noetherian ring (comrnutative with identity), then the polynomial ring R[ x] inherits this property.
Since any principal ideal domain and, in particular, any fie1d, is Noetherian,
Hilbert's Theorem pro vides us with a rather extensive c1a.ss of Noetherian
rings. The proof is somewhat demanding, but the result so elegant, that we
hope aH readers will work through the details.
1~
~,r
:, :
When
221
k
I'X E
I}
c2fn2(X)
+ ... +
cmnf"m.(x))
belongs to 1 and has degree les S than r; indeed, the coefficient of x' in this
pplynomial is
.
.
mn
b - L ca., = O.
=1
{O}.
I t is easily checked that 1k forms an ideal of the ring R with 1k S; 1k+ l'
(The second assertion follows from the fact that if r E 1k' then r occurs as
the leading coefficient of ~+ 1 when the corresponding polynomial is multiplied by x; hence, r E 1k+1') Since we are assuming that the ascending chain
'
220
r
I
... + km
n1
n2
+ ... +
nm,
then we must have k ~ n for sorne index, i (i = 1, 2, ... ,m). This implies
that a~' El, hence that the dement a1 1 a~2 :/:, a~m E l. Since aH the generators
of (.JI)n lie in 1, it foHows that (.JI)n S; l.":'
.
~
'.-
b.o +
b 1x
+' ... +
br _ 1x r - 1
bx'.
~~k~
l'
The Hilbert Basis Theorem asserts that if R is a Noetherian ring (comrnutative with identity), then the polynomial ring R[ x] inherits this property.
Since any principal ideal domain and, in particular, any fie1d, is Noetherian,
Hilbert's Theorem pro vides us with a rather extensive c1a.ss of Noetherian
rings. The proof is somewhat demanding, but the result so elegant, that we
hope aH readers will work through the details.
1~
~,r
:, :
When
221
k
I'X E
I}
c2fn2(X)
+ ... +
cmnf"m.(x))
belongs to 1 and has degree les S than r; indeed, the coefficient of x' in this
pplynomial is
.
.
mn
b - L ca., = O.
=1
{O}.
I t is easily checked that 1k forms an ideal of the ring R with 1k S; 1k+ l'
(The second assertion follows from the fact that if r E 1k' then r occurs as
the leading coefficient of ~+ 1 when the corresponding polynomial is multiplied by x; hence, r E 1k+1') Since we are assuming that the ascending chain
'
222
223
will serve as a simple illustration: in a power series ring F[[x]] over a field
F, rad F[[x]] = (x), but
(X)" = (x")
:/= {O}
Levitsky proved that for Noetherian rings the converse also holds: nil ideals
are nilptent. This fact is brought out as a corolIary to our next theorem.
Proof. .At the outset, observe that since R is Noetherian, we can use the
maximum condition to select an ideal N of R which is maximal with respect
to being nilpotent Our contention is that N is the largest nilpotent ideal
of R (in the sense of containing a11 other nilpotent ideals). To set this in
evidence, let N 1 be an arbitrary nilpotent ideal of R, say N~ = {O}; assume
further that Ni = {O}. Tben (N + N ly+k {O}, so that the ideal N + N 1
is nilpotent. From the inclusion N S;;; N + N 1 and the maximal property
of N, it follows that N = N + NI' One is then left with Ni S;;; N, which
settles the point. Now every nilpotent ideal must also be nil and thus
N S;;; Rad R by the corollary to Theorem 8-8. To derive the reverse
inc1usion, assume that a + N is any nilpotent element of the quotient ring
RfN. Then a" + N = (a + N)" = N for some n E Z +, implying that
a" E N. Becau$e N is a nil ideal, there exists a positive integer m for which
(a"'j" ,,;. 0, and so a is nilpotent as an element of R. This being the case, we
conc1ude that the principal ideal (a) is nilpotent; hence, (a) S;;; N, by the
maximality of N. Tbe rest should be c1ear: since a E (a) S;;; N, the coset
d':+-
= N.
" Our reasoning shows that the quotient ring R/N contains no nonzero
nilpotent elements, which is to say that R/N has zero prime radical. But it
is aIready known that Rad R is the smallest ideal of R possessing a quotient
ring without prime radical (Theorem 8-12). Tberefore, Rad R S;;; N, which
yields the desired equality N = Rad R; the theorem is now established.
As corollaries we have
Corollary 1. In a Noetherian ring, any nil ideal is nilpotent.
Proo! The proof amounts to the observation that any nil ideal is contained
in the prime radical of a ringo
Corollary 2. A semisimple N oetherian ring contains no nonzero
nilpotent ideals.
The breakdown of Levitsky's Theorem is rather dramatic when one
replaces the prime radical by the Jacobson radical. The foIlowing example
111+1
= 1.+2 =
Thus, the
222
223
will serve as a simple illustration: in a power series ring F[[x]] over a field
F, rad F[[x]] = (x), but
(X)" = (x")
:/= {O}
Levitsky proved that for Noetherian rings the converse also holds: nil ideals
are nilptent. This fact is brought out as a corolIary to our next theorem.
Proof. .At the outset, observe that since R is Noetherian, we can use the
maximum condition to select an ideal N of R which is maximal with respect
to being nilpotent Our contention is that N is the largest nilpotent ideal
of R (in the sense of containing a11 other nilpotent ideals). To set this in
evidence, let N 1 be an arbitrary nilpotent ideal of R, say N~ = {O}; assume
further that Ni = {O}. Tben (N + N ly+k {O}, so that the ideal N + N 1
is nilpotent. From the inclusion N S;;; N + N 1 and the maximal property
of N, it follows that N = N + NI' One is then left with Ni S;;; N, which
settles the point. Now every nilpotent ideal must also be nil and thus
N S;;; Rad R by the corollary to Theorem 8-8. To derive the reverse
inc1usion, assume that a + N is any nilpotent element of the quotient ring
RfN. Then a" + N = (a + N)" = N for some n E Z +, implying that
a" E N. Becau$e N is a nil ideal, there exists a positive integer m for which
(a"'j" ,,;. 0, and so a is nilpotent as an element of R. This being the case, we
conc1ude that the principal ideal (a) is nilpotent; hence, (a) S;;; N, by the
maximality of N. Tbe rest should be c1ear: since a E (a) S;;; N, the coset
d':+-
= N.
" Our reasoning shows that the quotient ring R/N contains no nonzero
nilpotent elements, which is to say that R/N has zero prime radical. But it
is aIready known that Rad R is the smallest ideal of R possessing a quotient
ring without prime radical (Theorem 8-12). Tberefore, Rad R S;;; N, which
yields the desired equality N = Rad R; the theorem is now established.
As corollaries we have
Corollary 1. In a Noetherian ring, any nil ideal is nilpotent.
Proo! The proof amounts to the observation that any nil ideal is contained
in the prime radical of a ringo
Corollary 2. A semisimple N oetherian ring contains no nonzero
nilpotent ideals.
The breakdown of Levitsky's Theorem is rather dramatic when one
replaces the prime radical by the Jacobson radical. The foIlowing example
111+1
= 1.+2 =
Thus, the
224
= {fE RI!(x)
O for -r
:5; X :5;
r}.
13 c:: 12
c:
I1
C;
225
In the sequel, ther~ occur eertain results which hold for both Noetherian
and Artinian rings. Where the proofs are virtually the same, our policy
will be to establish the theoremin question only in the Noetherian case.
Let us first show that the chain conditions are not deStroyed by homomorphisms.
'
Theorem 11-6. If R is a Noetherian (Artinian) ring, then any homomorphic image of R is also Noetherian (Artinian) .. ,!:
I 1/ 2 c: I 1 / 3 c: ....
)rhe iQlplication lS that R eontains ascending and descending chains that do'
:)ot becomestationary, whenee R is neither Artinian nor Noetherian. It~'i
's perhaps appropriate to call attention to the ,fact that eaeh of the ideals";:
'Jr is properly contained in the maximal ideal M = {fE Rlf(O) = O}.
""
not':
co~61Iary:
and
He in I, it follows that (rb + Sp)jpk
1jpk also belongs to 1, contradicting
therninimaJity of k. Thus, the ideal 1 is finite and is given by
i = k - j
and, in consequence, k
+j
1n K
J n I,
1.
{O} e 11 c: 12
c: ... e
1k c: ... e Z(paJ).
11 n 1
12 n 1
'" s;;; ln nI
S;;; ... ,
224
= {fE RI!(x)
O for -r
:5; X :5;
r}.
13 c:: 12
c:
I1
C;
225
In the sequel, ther~ occur eertain results which hold for both Noetherian
and Artinian rings. Where the proofs are virtually the same, our policy
will be to establish the theoremin question only in the Noetherian case.
Let us first show that the chain conditions are not deStroyed by homomorphisms.
'
Theorem 11-6. If R is a Noetherian (Artinian) ring, then any homomorphic image of R is also Noetherian (Artinian) .. ,!:
I 1/ 2 c: I 1 / 3 c: ....
)rhe iQlplication lS that R eontains ascending and descending chains that do'
:)ot becomestationary, whenee R is neither Artinian nor Noetherian. It~'i
's perhaps appropriate to call attention to the ,fact that eaeh of the ideals";:
'Jr is properly contained in the maximal ideal M = {fE Rlf(O) = O}.
""
not':
co~61Iary:
and
He in I, it follows that (rb + Sp)jpk
1jpk also belongs to 1, contradicting
therninimaJity of k. Thus, the ideal 1 is finite and is given by
i = k - j
and, in consequence, k
+j
1n K
J n I,
1.
{O} e 11 c: 12
c: ... e
1k c: ... e Z(paJ).
11 n 1
12 n 1
'" s;;; ln nI
S;;; ... ,
226
227
One can say considerably more about the ideal structure of an Artinian
ring R: R has only a finite number of prime (hence, .maximal) ideals. For,
suppose that there exists an infinite sequence {P} of distinct proper prime
ideals of R. We would then be able to form a descending chain of ideals
PI
P IP 2
P I P 2P 3
~ ....
and
for alI m ~ n. This being the case, an invocation ofthe lemma is permissible;
it folIows that J m = J n whenever m ~ n, whence R comprises a Noetherian
ringo
Artinian rings are generalIy more restrictive than Noetherian rings; for
instance, the only integral domains which satisfy the descending chain
condition are fields (this 1S not to suggest, however, that Artinian rings are
without interest).
Theorem 11-8. Any Artinian domain R (integral domain and Artinian
ring) is a field.
PIP2
Pn = P I P 2 ... PnP n+ l
It folIows from this that P I P 2 '" Pn S; Pn+ l , whence P k S; P n+ 1 for sorne
k ::; n. But P k is a maximal ideal of R, so that we must have P k = Pn+l'
...
contrary to the fact that the Pi are distinct. These observations are summarized as
Theorem 11-10. Every Artinian ring has only a finite number of proper
prime ideals, each of which is maximal.
We now come to the interesting part ofthe theory; namely, the extension
of Levitsky's' Theorem to Artinian rings.
Theorem 11-11. If R is an Artinian ring, then rad R forms a nilpotent
:~
rad R
(a n + 2)
= ....
.~'
Then there exists an element r E Rs~ch that an = ra n + l ; using the cancelIation law, it folIows that 1 = ra, whjch, pro ves our assertion.
Proof. Suppose that 1 is a proper prime ideal of R. Then, the quotient ring
Rll forms an integral domain which satisfies the descending chain condition
because R does. It folIows from Theorem 11-8 that Rll must be a field,
whence 1 is a maximal ideal of R.
....
By the descending chain condition, this chain must be of finite length, say
(a) ~ (a 2) ~ ... ~ (anfe:: (a n + l )
ideal.
Rad R.
(rad R)2
(rad R)3
...
shows that there exists an integer n for which (rad R)" = (rad R)n+r"
If we put 1 = (rad R)", then 1 S; rad R and 12 = 1. Our contention is that
{O}.
Assume-{qr. the moment that 1 =1= {O} and consider the family of alI
ideal s J of R such that (i) J S; 1 and (ii) JI =f {O}. This colIection is not
empty since jt "contains 1 and, hence, it admits a minimal member K. By
(ii), Kl =f {O},~so that al =f {O} for sorne nonzeroelement a E K. Thus,
(al)l = aJ2 = al
=f {O},
With littIe additional effort we can learn a good de al more about nilpotent ideals in rings with the descending chain condition.
226
227
One can say considerably more about the ideal structure of an Artinian
ring R: R has only a finite number of prime (hence, .maximal) ideals. For,
suppose that there exists an infinite sequence {P} of distinct proper prime
ideals of R. We would then be able to form a descending chain of ideals
PI
P IP 2
P I P 2P 3
~ ....
and
for alI m ~ n. This being the case, an invocation ofthe lemma is permissible;
it folIows that J m = J n whenever m ~ n, whence R comprises a Noetherian
ringo
Artinian rings are generalIy more restrictive than Noetherian rings; for
instance, the only integral domains which satisfy the descending chain
condition are fields (this 1S not to suggest, however, that Artinian rings are
without interest).
Theorem 11-8. Any Artinian domain R (integral domain and Artinian
ring) is a field.
PIP2
Pn = P I P 2 ... PnP n+ l
It folIows from this that P I P 2 '" Pn S; Pn+ l , whence P k S; P n+ 1 for sorne
k ::; n. But P k is a maximal ideal of R, so that we must have P k = Pn+l'
...
contrary to the fact that the Pi are distinct. These observations are summarized as
Theorem 11-10. Every Artinian ring has only a finite number of proper
prime ideals, each of which is maximal.
We now come to the interesting part ofthe theory; namely, the extension
of Levitsky's' Theorem to Artinian rings.
Theorem 11-11. If R is an Artinian ring, then rad R forms a nilpotent
:~
rad R
(a n + 2)
= ....
.~'
Then there exists an element r E Rs~ch that an = ra n + l ; using the cancelIation law, it folIows that 1 = ra, whjch, pro ves our assertion.
Proof. Suppose that 1 is a proper prime ideal of R. Then, the quotient ring
Rll forms an integral domain which satisfies the descending chain condition
because R does. It folIows from Theorem 11-8 that Rll must be a field,
whence 1 is a maximal ideal of R.
....
By the descending chain condition, this chain must be of finite length, say
(a) ~ (a 2) ~ ... ~ (anfe:: (a n + l )
ideal.
Rad R.
(rad R)2
(rad R)3
...
shows that there exists an integer n for which (rad R)" = (rad R)n+r"
If we put 1 = (rad R)", then 1 S; rad R and 12 = 1. Our contention is that
{O}.
Assume-{qr. the moment that 1 =1= {O} and consider the family of alI
ideal s J of R such that (i) J S; 1 and (ii) JI =f {O}. This colIection is not
empty since jt "contains 1 and, hence, it admits a minimal member K. By
(ii), Kl =f {O},~so that al =f {O} for sorne nonzeroelement a E K. Thus,
(al)l = aJ2 = al
=f {O},
With littIe additional effort we can learn a good de al more about nilpotent ideals in rings with the descending chain condition.
T
228
229
",
Now, x" is not in In' but a(q"x n}' = ar v = 0, since a s a zero divisor of R.
Therefore, x n lies in ln+1 and the lk form a properly ascending ch,ain. This
contradicts the ascending ch~n condition and no such element a exists.
, I
Remark. Over the course oC th,e next several pages, we shall often simply
say "the set ofzero divisors':of R form anideal" when what is really meant
is "the set of zero divisors,()gether with zero, form an ideal".
(a - b)xu
).
,'.,
........
By assumption, none of these is the zero ideal and, because of the minimum
cOridition, we must have (a n) = (a n + 1 ) for sorne n E Z+. This being the case,
a n = ra n+ 1 or a"(1 - ra) = O, with rE R. Inasmuch as a" =F O, the expression in parentheses is a zero divisor of R and, hence, lies in ann R v by
Theoreml0-9, Thus,foranynonzero elementx E R V, wehavex(l - ra) = O.
But xa =- O, since a also belongs to ann R v , and SO x = O. This contradiction forces the element a to be nilpotent, as desired.
We next extend the stated result to rings with the ascending chain
condition. As in the previous paragraph, suppose that the element a is a
zero divisor of R which is oot nilpotent. Then all the powers a 2 , a 3 , .. , , an, .,.
are zero divisors and, of course, none is zero. Thus, for every power a",
there exists an element XII such that
aX 1
a2x 2
= ... =
anx"
= .,. =
r V =F O.
'In the lght ofthe corolIary aboye, it would appear natural to study rings
whose zero divisors form an ideal which is contained in the Jacobson radical
(we point out that this condition holds trivially in any integral domain).
Our next two results pre,sent criteria for these rings to become local rings,
Theorem 11-13. Let 1 be an ideal of the ring R with 1
R is a local ring if and only if RIl is a local ringo
S;
rad R.
Then-
T
228
229
",
Now, x" is not in In' but a(q"x n}' = ar v = 0, since a s a zero divisor of R.
Therefore, x n lies in ln+1 and the lk form a properly ascending ch,ain. This
contradicts the ascending ch~n condition and no such element a exists.
, I
Remark. Over the course oC th,e next several pages, we shall often simply
say "the set ofzero divisors':of R form anideal" when what is really meant
is "the set of zero divisors,()gether with zero, form an ideal".
(a - b)xu
).
,'.,
........
By assumption, none of these is the zero ideal and, because of the minimum
cOridition, we must have (a n) = (a n + 1 ) for sorne n E Z+. This being the case,
a n = ra n+ 1 or a"(1 - ra) = O, with rE R. Inasmuch as a" =F O, the expression in parentheses is a zero divisor of R and, hence, lies in ann R v by
Theoreml0-9, Thus,foranynonzero elementx E R V, wehavex(l - ra) = O.
But xa =- O, since a also belongs to ann R v , and SO x = O. This contradiction forces the element a to be nilpotent, as desired.
We next extend the stated result to rings with the ascending chain
condition. As in the previous paragraph, suppose that the element a is a
zero divisor of R which is oot nilpotent. Then all the powers a 2 , a 3 , .. , , an, .,.
are zero divisors and, of course, none is zero. Thus, for every power a",
there exists an element XII such that
aX 1
a2x 2
= ... =
anx"
= .,. =
r V =F O.
'In the lght ofthe corolIary aboye, it would appear natural to study rings
whose zero divisors form an ideal which is contained in the Jacobson radical
(we point out that this condition holds trivially in any integral domain).
Our next two results pre,sent criteria for these rings to become local rings,
Theorem 11-13. Let 1 be an ideal of the ring R with 1
R is a local ring if and only if RIl is a local ringo
S;
rad R.
Then-
230
Xi
Xi E1 i , YiEMi
Yi'
(i
1, 2, ooo, n)o
231
Then R
.~~ ! .
Although the hypotheses of Theorem 11-14 appear somewht formidable, it is worth remarking that the power series ring F[[xJ] (F a fi~l~)
satisfies the requisite conditionso
' . \Our next goal is to describe semisimple Artinian rings; crucialtd tli'e
discussion is the fact that such rings have only a finite number of max~i:nal
ideal s with zero intersectiono The theorembelow is the cornmutative veision
of Wedderburn's fundamental result (Theorem 13-3)0
since a - x k
i= 1
Xi = (a - Xk) -
Yk E M k, while Xi E 1i
a -
S;
Xi E n
Xi E M k,
i'f'k
M k for i
=1=
ko Thus,
M k = {O}
i'f'k
1i ) ~ 1k n M k = {O}o
Seeond proof ofTheorem 11-150 Let MI' M 2' ooo';!lfn be the ~axin:"al ideals
of R (there is no harm in assuming that {O} i~ ~o~ a ~axlmal lde~l, for
otherwse the theorem follows trivially)o The m~IJ(lmahty of these ldea~s
implies that Mi + Mj = R whenever i =1= j. :r;hus, by Theorem 10-5, lt
follows that
R = R/n Mi ~
EB> (R/M)
where each ofthe quotient rings R/Mi is a fiel~ (i = 1,2, o.. ,n): But, in
the finite case, the complete direct sum coincides with the usual dlfect sumo
Notice that, in carrying out the aboye argument, we have proved a
o
subresult which is interesting in its own right: If a ring R has a fimte number
of maximal ideals M. with zero intersection, then R ~ (f) (R/M).
There is a coroll~ry to Theorem 11-15 which is worthy of emphasiso
Coro]]ary. Any semisimple Artinian ring is Noetheriano
230
Xi
Xi E1 i , YiEMi
Yi'
(i
1, 2, ooo, n)o
231
Then R
.~~ ! .
Although the hypotheses of Theorem 11-14 appear somewht formidable, it is worth remarking that the power series ring F[[xJ] (F a fi~l~)
satisfies the requisite conditionso
' . \Our next goal is to describe semisimple Artinian rings; crucialtd tli'e
discussion is the fact that such rings have only a finite number of max~i:nal
ideal s with zero intersectiono The theorembelow is the cornmutative veision
of Wedderburn's fundamental result (Theorem 13-3)0
since a - x k
i= 1
Xi = (a - Xk) -
Yk E M k, while Xi E 1i
a -
S;
Xi E n
Xi E M k,
i'f'k
M k for i
=1=
ko Thus,
M k = {O}
i'f'k
1i ) ~ 1k n M k = {O}o
Seeond proof ofTheorem 11-150 Let MI' M 2' ooo';!lfn be the ~axin:"al ideals
of R (there is no harm in assuming that {O} i~ ~o~ a ~axlmal lde~l, for
otherwse the theorem follows trivially)o The m~IJ(lmahty of these ldea~s
implies that Mi + Mj = R whenever i =1= j. :r;hus, by Theorem 10-5, lt
follows that
R = R/n Mi ~
EB> (R/M)
where each ofthe quotient rings R/Mi is a fiel~ (i = 1,2, o.. ,n): But, in
the finite case, the complete direct sum coincides with the usual dlfect sumo
Notice that, in carrying out the aboye argument, we have proved a
o
subresult which is interesting in its own right: If a ring R has a fimte number
of maximal ideals M. with zero intersection, then R ~ (f) (R/M).
There is a coroll~ry to Theorem 11-15 which is worthy of emphasiso
Coro]]ary. Any semisimple Artinian ring is Noetheriano
232
PROBLEMS
233
finite direet suin ofNoetherian rings (in this case, fields) is again Noetherian.
We shalI see later that the imposed semisiinplicity condition is unnecessarly stringent; indeed, the foregoing result can be sharpened to read that
everycommutative Artinian ring with identity lS Noetherian. The ring of
integers shows that the converse need not hold.
In Qp, define addition lo be ordinary addition of ralional numbers and mllltiplication to be the trivial multiplication (i.e. ab = O for alI a, b E Qp). Establish that
a) Z forms an ideal ofthe resulting ring Qp;
b) the quotient ring Qp/Z is isomorphic to 2(p<X;
PROBLEMS
In all problems, R is, ~conm1Utative ring with identity.
l. Le! 1 be a nonzi.ideal of !he principal ideal doman R Prove that the quotlerit
ring R/l satisfie~' both cPain conditions.
" 11. In an Artinan ring R, prove that the zero ideal is a prodllct of maximal ideIs.:
, [Hint: rad R = M 1 n M 2 n ... n'M., where each Mi s maxhllal; use Theor~;'
11-11 and Problern 13, Chapter 10, to conclude {O} (rad R)k MfM~ .. , M~"
. .
b~~~~
.JI/l
12. Establish that an Artinian ring R is isomorphic to a finite diiect sum of Artinian
local rings. [Hint: From Problem 11 and the fact!hat the ideals M} arecomaximal,
we have M~ n ... n M~ ... M~ '" M! = {O}. By Theorem 10-1, R c:: RIM~.
4. Prove that every id~al of a NO,etherian ringR contains a product of prinle ideals.
[Hint: If not, let S be the set of those ideals of R which do not contain a product
of prime idealsand apply ,the maximllm condition.]
.
13. Prove that any semisinlple Artinian ring possesses only a finite number of ideals.
[Hlnt: Assume that R admits that decomposition R = F1 F2 ... F., F
a fie1d; if 1 is an idei, of R, then 1 = 11 ... In witb li an ideal of Fd
14. a) Let 1 be a nontrvial mnimal ideal of the Artinian ring R Show that the
annihilator ai:m 1 forms a prime and, hence, maxinlal, ideal of R [Hint: If
a ~ ann 1, al ~ l.]
b) Assume tbat 1 is a nonzero ideal of the ring R (no chain conditions). If Pis a
maxinlal member of the eollection {ann (x)IO -+ x E 1), d~uc~ thatP is a prime
ideal. [Hlnt: Let abEP= ann(r), with a~P; then, P 5;; (P,b) ~ ann(ar).]
.JI
<;
Assuming
'
that R is a divisible ring prove the following:
a) R is a local ring if and only jf the set D of a1l zero divisors (together with zero)
is included in a proper ideal of R; in this case, D itself becomes an ideal.
b) If 11 n 12
{O} for any two nonzero ideals of R, then R s a local ringo [Hint:
Show !hat the sum pf two noninvertible elenlents of R is again noninvertible.]
+.
16. Let R be a principal ideal ring which is not an integral domain. If!he set of al!
zero divisors D = rad R, verify that R is a local ring,
17. a) If R is a finite Boolean ring, prove that R is isomorphic to the direct sum of a
finite number of fields 2 2 , [Hint: See the remark following the second proof
ofTheorem 11-15.]
b) Prove that a finite Boolean ring has 2" elements for some n E Z+.
232
PROBLEMS
233
finite direet suin ofNoetherian rings (in this case, fields) is again Noetherian.
We shalI see later that the imposed semisiinplicity condition is unnecessarly stringent; indeed, the foregoing result can be sharpened to read that
everycommutative Artinian ring with identity lS Noetherian. The ring of
integers shows that the converse need not hold.
In Qp, define addition lo be ordinary addition of ralional numbers and mllltiplication to be the trivial multiplication (i.e. ab = O for alI a, b E Qp). Establish that
a) Z forms an ideal ofthe resulting ring Qp;
b) the quotient ring Qp/Z is isomorphic to 2(p<X;
PROBLEMS
In all problems, R is, ~conm1Utative ring with identity.
l. Le! 1 be a nonzi.ideal of !he principal ideal doman R Prove that the quotlerit
ring R/l satisfie~' both cPain conditions.
" 11. In an Artinan ring R, prove that the zero ideal is a prodllct of maximal ideIs.:
, [Hint: rad R = M 1 n M 2 n ... n'M., where each Mi s maxhllal; use Theor~;'
11-11 and Problern 13, Chapter 10, to conclude {O} (rad R)k MfM~ .. , M~"
. .
b~~~~
.JI/l
12. Establish that an Artinian ring R is isomorphic to a finite diiect sum of Artinian
local rings. [Hint: From Problem 11 and the fact!hat the ideals M} arecomaximal,
we have M~ n ... n M~ ... M~ '" M! = {O}. By Theorem 10-1, R c:: RIM~.
4. Prove that every id~al of a NO,etherian ringR contains a product of prinle ideals.
[Hint: If not, let S be the set of those ideals of R which do not contain a product
of prime idealsand apply ,the maximllm condition.]
.
13. Prove that any semisinlple Artinian ring possesses only a finite number of ideals.
[Hlnt: Assume that R admits that decomposition R = F1 F2 ... F., F
a fie1d; if 1 is an idei, of R, then 1 = 11 ... In witb li an ideal of Fd
14. a) Let 1 be a nontrvial mnimal ideal of the Artinian ring R Show that the
annihilator ai:m 1 forms a prime and, hence, maxinlal, ideal of R [Hint: If
a ~ ann 1, al ~ l.]
b) Assume tbat 1 is a nonzero ideal of the ring R (no chain conditions). If Pis a
maxinlal member of the eollection {ann (x)IO -+ x E 1), d~uc~ thatP is a prime
ideal. [Hlnt: Let abEP= ann(r), with a~P; then, P 5;; (P,b) ~ ann(ar).]
.JI
<;
Assuming
'
that R is a divisible ring prove the following:
a) R is a local ring if and only jf the set D of a1l zero divisors (together with zero)
is included in a proper ideal of R; in this case, D itself becomes an ideal.
b) If 11 n 12
{O} for any two nonzero ideals of R, then R s a local ringo [Hint:
Show !hat the sum pf two noninvertible elenlents of R is again noninvertible.]
+.
16. Let R be a principal ideal ring which is not an integral domain. If!he set of al!
zero divisors D = rad R, verify that R is a local ring,
17. a) If R is a finite Boolean ring, prove that R is isomorphic to the direct sum of a
finite number of fields 2 2 , [Hint: See the remark following the second proof
ofTheorem 11-15.]
b) Prove that a finite Boolean ring has 2" elements for some n E Z+.
TWELVE
235
... ~ ]:
(a") ~ ...
TWELVE
235
... ~ ]:
(a") ~ ...
236
237
x = i
+ rak =
i'
r' b
for suitably chosen elements i, i' El and r, r' E R. Consecju~ntIy, the product
raHl = (i'
- i)a
r'(ab)
J,
.'--.
These results may now be put into the form of a ~ecomposition theorem,
the so-called Primary Decomposition Theorem of Noether.
Theorem 12-1. (Noether).
Every ideal of a Noetherian ring can be
represented as a finite intersection of primary ideals.
Let us call a representation of an ideal 1 in the form J- ~ n Q, where
eacp Q is a primary ideal, a primary representation of J; the individual Q
are said to be the primary components of the representation, while JQ are
the prime ideals associated witb l. What was just pro~ed is that, in a
Noetherian ring, every ideal admits a finite primary representation.
Before announcing the next result concerning primary representations,
we wish to introduce a new idea.
Definition 12-2. A primary representation 1 = n~= 1 Q will be termed
irredundant if it satisfies the following two conditions.
1) No Q contains the intersection of the other primary components;
that is to say, n;=j Q =1= n Q for any j = 1,2, ... , n.
2) .jQ
=1=
.jQj for i
=1=
j.
wi,th al ~ P and so, by the definition of prime ideal, (a~' ... a~n) E P. Repeating
thlS argument, we finally obtain a~n E P, whence an E P, which is impossible.
RecaH that a prime idal of R is said to be a mini mal prime of the ideal
1 if it.is minimal in t,he set of prime ideals containing l. Keeping the same
notatlOn, the foregomg lemma asserts that the minimal prime ideals of J
ar~ the ~i~im~l elements ofthe family {.jQ}, partiaHy ordered by inc1usion.
WIth thls m mmd, we can now formulate
.
,
!
I
Theore~
"\'
:/
236
237
x = i
+ rak =
i'
r' b
for suitably chosen elements i, i' El and r, r' E R. Consecju~ntIy, the product
raHl = (i'
- i)a
r'(ab)
J,
.'--.
These results may now be put into the form of a ~ecomposition theorem,
the so-called Primary Decomposition Theorem of Noether.
Theorem 12-1. (Noether).
Every ideal of a Noetherian ring can be
represented as a finite intersection of primary ideals.
Let us call a representation of an ideal 1 in the form J- ~ n Q, where
eacp Q is a primary ideal, a primary representation of J; the individual Q
are said to be the primary components of the representation, while JQ are
the prime ideals associated witb l. What was just pro~ed is that, in a
Noetherian ring, every ideal admits a finite primary representation.
Before announcing the next result concerning primary representations,
we wish to introduce a new idea.
Definition 12-2. A primary representation 1 = n~= 1 Q will be termed
irredundant if it satisfies the following two conditions.
1) No Q contains the intersection of the other primary components;
that is to say, n;=j Q =1= n Q for any j = 1,2, ... , n.
2) .jQ
=1=
.jQj for i
=1=
j.
wi,th al ~ P and so, by the definition of prime ideal, (a~' ... a~n) E P. Repeating
thlS argument, we finally obtain a~n E P, whence an E P, which is impossible.
RecaH that a prime idal of R is said to be a mini mal prime of the ideal
1 if it.is minimal in t,he set of prime ideals containing l. Keeping the same
notatlOn, the foregomg lemma asserts that the minimal prime ideals of J
ar~ the ~i~im~l elements ofthe family {.jQ}, partiaHy ordered by inc1usion.
WIth thls m mmd, we can now formulate
.
,
!
I
Theore~
"\'
:/
238
JQ are unique and are the same in all irredundant primary representations
of I; thus, it is the number of primary components that will be unique. To
veriCy this, it is enough to show that the associated prime ideals can be
characterized in terrns of the properties oC I alone, independent oC any
particular primary representation considered. Before proceeding to the
prooC, let us illustrate the fact that an ideal in a Noetherian ring need not
have a unique irredundant primary representation.
{
Example 12-1. In the polynomial ring F[x, yJ, where F is any field, consider
the ideal (x2, xy). It is easy to see that (x2; xy) consists of those polynomials
which have x as a factor and which do not possess linear terms. Now, the
nonzero elements of(x2, xy,
are precisely the polynomials each ofwhose
terms are of degree 2; hence the intersection (x2, xy, y2) n (x) contains
the zero polynomial together with aH polynomials of degree 2 which
bave x as a factor. Th~s, we have
r)
(x 2, xy)
nk
I: (a) :;;;;
.JQ, :; ;
JI: (a).
The first inclusion is justified by the fact that, since a(I: (a) :;;;; I :;;;; Q with
a ~ Q, necessari1y I: (a) :;;;;.J(j. To see 'tb,e second inc1usion, simply note
,that aQ :;;;; I, whence Q :;;;; I: (a).
239
Enk
I: (a)
= JQ;. n JO:;. n
... n
JQ:..
Knowing this, the proof is easi1y completed; for, by Problern 30(a), Chapter
5, P must contain one of the ideals ..JQ;. and is obviously contained by
it, whence P = ..jQik'
What we are realIy after is the corollary below.
Corollary. Let I be an ideal of the Noetherian ring R. Suppose that
I = Ql n ... n Qn = Q~ n ... n Q~ are two finite irredundant primary representations of 1. Then, n = m and the assoclated prime ideals
of these two representations are equal (t,bat is, with a slltable renumbering of the indices .JQ. =
for 1 .$ , :s; n = m).
.Jffi
regardless of the choice of n > 1. The corresponding nil radicals (that is,
the associated prime ideals) are (x, y) and (x). Thus, in any irredundant
represen ta tion
(x 2, xy) = Ql n Q2
by primary ideals Ql and Q2' we must have
.JQ;
(x, y),
JQ;
= (x).
238
JQ are unique and are the same in all irredundant primary representations
of I; thus, it is the number of primary components that will be unique. To
veriCy this, it is enough to show that the associated prime ideals can be
characterized in terrns of the properties oC I alone, independent oC any
particular primary representation considered. Before proceeding to the
prooC, let us illustrate the fact that an ideal in a Noetherian ring need not
have a unique irredundant primary representation.
{
Example 12-1. In the polynomial ring F[x, yJ, where F is any field, consider
the ideal (x2, xy). It is easy to see that (x2; xy) consists of those polynomials
which have x as a factor and which do not possess linear terms. Now, the
nonzero elements of(x2, xy,
are precisely the polynomials each ofwhose
terms are of degree 2; hence the intersection (x2, xy, y2) n (x) contains
the zero polynomial together with aH polynomials of degree 2 which
bave x as a factor. Th~s, we have
r)
(x 2, xy)
nk
I: (a) :;;;;
.JQ, :; ;
JI: (a).
The first inclusion is justified by the fact that, since a(I: (a) :;;;; I :;;;; Q with
a ~ Q, necessari1y I: (a) :;;;;.J(j. To see 'tb,e second inc1usion, simply note
,that aQ :;;;; I, whence Q :;;;; I: (a).
239
Enk
I: (a)
= JQ;. n JO:;. n
... n
JQ:..
Knowing this, the proof is easi1y completed; for, by Problern 30(a), Chapter
5, P must contain one of the ideals ..JQ;. and is obviously contained by
it, whence P = ..jQik'
What we are realIy after is the corollary below.
Corollary. Let I be an ideal of the Noetherian ring R. Suppose that
I = Ql n ... n Qn = Q~ n ... n Q~ are two finite irredundant primary representations of 1. Then, n = m and the assoclated prime ideals
of these two representations are equal (t,bat is, with a slltable renumbering of the indices .JQ. =
for 1 .$ , :s; n = m).
.Jffi
regardless of the choice of n > 1. The corresponding nil radicals (that is,
the associated prime ideals) are (x, y) and (x). Thus, in any irredundant
represen ta tion
(x 2, xy) = Ql n Q2
by primary ideals Ql and Q2' we must have
.JQ;
(x, y),
JQ;
= (x).
240
-J(x 2 , xy)
(x).
Incidentally, our example has the added advantage of showing that -JI can
'
be prime without the ideal l being primary.
'}
<::,,':,,1. (1, b)
;-?:.:!::;
T~~p; C = a
(el) c2 ,
... ,
... ,
dm ).
a", b) .
n-JQi nQi
;2
i'f'i
1: (b) == (d 1 , d2 ,
and
en)
l =
... ,
(1, b) = (al' a 2 ,
.JI
241
;2
(Y, Y E R);
1,
iti
which means that l = ni'f'j Qi' This, however, contradicts the hypothesis
that the given representation of l is irredundant.
,
Next, fix the integer j and let a be any element of the ideal~. Since
nitj J[ =1= n -JQ, we can find sorne b E n~j.JQ; with b ~ 'VQj' Then
the product ab E J[ = l ~ Qj' whence a lS a member of Qi: .The~e
fore, .J(E S; Qi' yielding the subsequent equality Qi = .J(E. ThlS lmpbes
that Qi is a prime ideal, which was what had to be proved.
y = dlz l
Let us change direction now. The reader will no doubt recall that a ring
R is Noetherian if and only if every ideal of R is finitely generated (Theorem
11-1). Actually, it is enough to consider just the prime ideals of R, the proof
being due to l. S. Cohen.
d2 z 2
+ ... + dmz m,
leading directly to
x = aly
=
The equality l = J now follows and so one concludes that l itself is finitely
generated, an impossibility since l E ff. This contradiction completes the
proof.
Scrutiny of the preceding argument reveals a fact which is important
enough to be stated independently: Let l be an ideal of the ring R and b
an element of R; if the ideals (1, b) and 1: (b) are both finitely generated,
,/,;(,:
{.;~,f
l ','.
240
-J(x 2 , xy)
(x).
Incidentally, our example has the added advantage of showing that -JI can
'
be prime without the ideal l being primary.
'}
<::,,':,,1. (1, b)
;-?:.:!::;
T~~p; C = a
(el) c2 ,
... ,
... ,
dm ).
a", b) .
n-JQi nQi
;2
i'f'i
1: (b) == (d 1 , d2 ,
and
en)
l =
... ,
(1, b) = (al' a 2 ,
.JI
241
;2
(Y, Y E R);
1,
iti
which means that l = ni'f'j Qi' This, however, contradicts the hypothesis
that the given representation of l is irredundant.
,
Next, fix the integer j and let a be any element of the ideal~. Since
nitj J[ =1= n -JQ, we can find sorne b E n~j.JQ; with b ~ 'VQj' Then
the product ab E J[ = l ~ Qj' whence a lS a member of Qi: .The~e
fore, .J(E S; Qi' yielding the subsequent equality Qi = .J(E. ThlS lmpbes
that Qi is a prime ideal, which was what had to be proved.
y = dlz l
Let us change direction now. The reader will no doubt recall that a ring
R is Noetherian if and only if every ideal of R is finitely generated (Theorem
11-1). Actually, it is enough to consider just the prime ideals of R, the proof
being due to l. S. Cohen.
d2 z 2
+ ... + dmz m,
leading directly to
x = aly
=
The equality l = J now follows and so one concludes that l itself is finitely
generated, an impossibility since l E ff. This contradiction completes the
proof.
Scrutiny of the preceding argument reveals a fact which is important
enough to be stated independently: Let l be an ideal of the ring R and b
an element of R; if the ideals (1, b) and 1: (b) are both finitely generated,
,/,;(,:
{.;~,f
l ','.
242
(1 :-- e)n E 1
000'
000'
(1 - rJl
!;;
(1 - r;)IJ
lJo
(1 - r+l)l
In consequence,
k=i
bkak
(b ik E 1)0
(1 - r;)(1 - r - bJl
!;;
(1 - r i
bu )l
n In
n=l
IX)
. Proa! For ease ofnotation, let S denote the right-hand side ofthe indicate,d,
equationo If the element r E S, so that (1 - a)r = O for suitable a E 1,
would necessarily have
we.
r.
243
= ar = a2 r =
o o o
a"r
nn
nn
IJ =
n Qi'
i
(Qi primary)o
n"
242
(1 :-- e)n E 1
000'
000'
(1 - rJl
!;;
(1 - r;)IJ
lJo
(1 - r+l)l
In consequence,
k=i
bkak
(b ik E 1)0
(1 - r;)(1 - r - bJl
!;;
(1 - r i
bu )l
n In
n=l
IX)
. Proa! For ease ofnotation, let S denote the right-hand side ofthe indicate,d,
equationo If the element r E S, so that (1 - a)r = O for suitable a E 1,
would necessarily have
we.
r.
243
= ar = a2 r =
o o o
a"r
nn
nn
IJ =
n Qi'
i
(Qi primary)o
n"
244
In any 'event, J S; Q for each value of i, whence J S; 1J and eq:u:ality follows. . From the lemma just proved, there exists an elementd 'i; 1 such that
(1 - a)J = {O}. But this amounts to asserting that J = n~.t~ is contained
in the set S and thereby completes the proof.
")
n:= 1
n:=1
Proo! Take l ~ rad R in Krull's Theorem. By Theorem 8-2, every elem.ent of 1 - radR is invertible and thus cannot be a zero divisor.
COfolIary 3. In a Noetherian domain R, any prime principal ideal (a)
is a minimal prime idealof R.
aP
a2 p = ...
S;
n(a") =
11
(a)" = {O},
"
n:=l
245
n:=l
Theorem 12-10.
{O} = Q1 n Q2
of {O}. Assume
but not for 1 ~
n 1"
"=1
co
a)r
<,
..
n:=l
244
In any 'event, J S; Q for each value of i, whence J S; 1J and eq:u:ality follows. . From the lemma just proved, there exists an elementd 'i; 1 such that
(1 - a)J = {O}. But this amounts to asserting that J = n~.t~ is contained
in the set S and thereby completes the proof.
")
n:= 1
n:=1
Proo! Take l ~ rad R in Krull's Theorem. By Theorem 8-2, every elem.ent of 1 - radR is invertible and thus cannot be a zero divisor.
COfolIary 3. In a Noetherian domain R, any prime principal ideal (a)
is a minimal prime idealof R.
aP
a2 p = ...
S;
n(a") =
11
(a)" = {O},
"
n:=l
245
n:=l
Theorem 12-10.
{O} = Q1 n Q2
of {O}. Assume
but not for 1 ~
n 1"
"=1
co
a)r
<,
..
n:=l
246
to M, which signifies that they are both invertible. The conve~se should' be
obvious.
Concerning (2), let R = ab. If the element b is not in vertible, then part
(1) forces M = (p) e (a). The inaximality of M then ensures that (a) = R,
whence a has an inverse in R.
To see the final assertion, assume that a = qnu = qmv, with m > n.
Then qnu(l - qm-nvu- 1 ) = O. By the I:).rgument of the first paragraph, this
relation implies that a = O.
Theorem 12-11. Let R be a local ring with principal maximal ideal
M = (p). Then every element O =1= a E M has a factorization in the form
a = pmu, where u is invertible, if and only if n~= 1 (pn) = {O}.
nn
n"
S;;;
(a)
S;;;
2) (rs)a
(pk)
S;;;
(pm),
whence the equality (pm) = (pk). The lemma now tells us that pm = pkV,
where v is an invertible element of R.. This mean s that a = pkVU, with vu
invertible, contradicting the last assertion of the lemma.
Corollary 1. LetR bea local ring with principal maximal ideal M = (p).
Assume further that n:'=l M" = {O}. If 1 is any nontrivial ideal of R,
then 1 = M k for sorne k E Z + (hence, R is Noetherian).
Proo! Clearly, 1 S;;; M, so that each nonzero element of 1 can be written
. as pnu, with u invertible. Take k to be the least integer such that pku E l.
It then follows that 1 S;;; (pk). On the other hand, since pku E 1, so does
pk = (pkU)U- 1 ; this implies that (pk) S;;; 1 and equality follows.
247
3) r(a
s)a
r(sa),
ra
b) = ra
sa,
rb,
4) la = a,
246
to M, which signifies that they are both invertible. The conve~se should' be
obvious.
Concerning (2), let R = ab. If the element b is not in vertible, then part
(1) forces M = (p) e (a). The inaximality of M then ensures that (a) = R,
whence a has an inverse in R.
To see the final assertion, assume that a = qnu = qmv, with m > n.
Then qnu(l - qm-nvu- 1 ) = O. By the I:).rgument of the first paragraph, this
relation implies that a = O.
Theorem 12-11. Let R be a local ring with principal maximal ideal
M = (p). Then every element O =1= a E M has a factorization in the form
a = pmu, where u is invertible, if and only if n~= 1 (pn) = {O}.
nn
n"
S;;;
(a)
S;;;
2) (rs)a
(pk)
S;;;
(pm),
whence the equality (pm) = (pk). The lemma now tells us that pm = pkV,
where v is an invertible element of R.. This mean s that a = pkVU, with vu
invertible, contradicting the last assertion of the lemma.
Corollary 1. LetR bea local ring with principal maximal ideal M = (p).
Assume further that n:'=l M" = {O}. If 1 is any nontrivial ideal of R,
then 1 = M k for sorne k E Z + (hence, R is Noetherian).
Proo! Clearly, 1 S;;; M, so that each nonzero element of 1 can be written
. as pnu, with u invertible. Take k to be the least integer such that pku E l.
It then follows that 1 S;;; (pk). On the other hand, since pku E 1, so does
pk = (pkU)U- 1 ; this implies that (pk) S;;; 1 and equality follows.
247
3) r(a
s)a
r(sa),
ra
b) = ra
sa,
rb,
4) la = a,
248
Example 12-3.
a+ ...
Example 12-5. If 1 is a left ideal of a ring R with identity, then the underlying additive group (1, +) ofjforms a left R~module. Indeed, the definition
of left ideal insures that thedng product ra E 1 for every rE R and a El.
As a special case, any ring R'with identity I::an be viewed as a left (or right)
R-module over itself.
:' '
,
Example 12-6. Consider th~set hom G of all homomorphisms of a com-
mutative group (G, + )""into 'itsdf (that is, the set of endomorphisms of G).
It is aIread y known that (hpIp.~, +, o) constitutes a ring with identity, where
o indica tes the operation of ftiri.tional composition. To provide G with the
structure of a left module over hom G, we define the module product fa
by putting
(f E hom G, a E G).
fa = f(a)
Condition (3) of Definition 12-3 is satisfied by virtue of the fact that f is a
homomorphism.
.
.
To avoid a proliferation of symbols, O will be used to designate the
additive identity element of the group (M, +) as well as the zero element
of R. This convention should lead to no ambiguity if the reader attends .
c10sely to the context in which the notation is employed: As with vector
spaces, we have the laws (i) Oa = rO = O, (ii) r(-a) = (-r)a = -(ra),
for all r E R and a E M.
Oile can introduce the notions of submodule, quotient module, and
module homomorphisms, all by natural definitions. These are of fundamental importance for our theory and from them our ultimate goal,
Hopkin's Theorem, will follow easily. In the remainder of this discussion,
we shall drop the prefix "left", so that the ter m "R-module" will always
mean "left R-module"; it should be apparent that the entire discussion
applies equally well to right R-modules. Of course, when R is acommutative
ring, any left R-module can be turned into a right R-module simply by
putting ar = ra. Modules over commuta:tive rings are essentially two-sided
and all distinction between left and right disappears (it is merely a matter
of personal preference whether one writes the ring elements on the left or
on the right).
A natural starting point is, perhaps, to .examine the concept of a submodule. Suppose then that M is an arbitrary module over the ring R. By
249
(a
N)
(b
N) = a
N.
r(a
N)
ra
N.
248
Example 12-3.
a+ ...
Example 12-5. If 1 is a left ideal of a ring R with identity, then the underlying additive group (1, +) ofjforms a left R~module. Indeed, the definition
of left ideal insures that thedng product ra E 1 for every rE R and a El.
As a special case, any ring R'with identity I::an be viewed as a left (or right)
R-module over itself.
:' '
,
Example 12-6. Consider th~set hom G of all homomorphisms of a com-
mutative group (G, + )""into 'itsdf (that is, the set of endomorphisms of G).
It is aIread y known that (hpIp.~, +, o) constitutes a ring with identity, where
o indica tes the operation of ftiri.tional composition. To provide G with the
structure of a left module over hom G, we define the module product fa
by putting
(f E hom G, a E G).
fa = f(a)
Condition (3) of Definition 12-3 is satisfied by virtue of the fact that f is a
homomorphism.
.
.
To avoid a proliferation of symbols, O will be used to designate the
additive identity element of the group (M, +) as well as the zero element
of R. This convention should lead to no ambiguity if the reader attends .
c10sely to the context in which the notation is employed: As with vector
spaces, we have the laws (i) Oa = rO = O, (ii) r(-a) = (-r)a = -(ra),
for all r E R and a E M.
Oile can introduce the notions of submodule, quotient module, and
module homomorphisms, all by natural definitions. These are of fundamental importance for our theory and from them our ultimate goal,
Hopkin's Theorem, will follow easily. In the remainder of this discussion,
we shall drop the prefix "left", so that the ter m "R-module" will always
mean "left R-module"; it should be apparent that the entire discussion
applies equally well to right R-modules. Of course, when R is acommutative
ring, any left R-module can be turned into a right R-module simply by
putting ar = ra. Modules over commuta:tive rings are essentially two-sided
and all distinction between left and right disappears (it is merely a matter
of personal preference whether one writes the ring elements on the left or
on the right).
A natural starting point is, perhaps, to .examine the concept of a submodule. Suppose then that M is an arbitrary module over the ring R. By
249
(a
N)
(b
N) = a
N.
r(a
N)
ra
N.
250
condition can also be applied to R-modules, the sole difference being that,
in our earlier definitions, the term"ideal" must now be replaced by the word
"submodule". Adapting the argument of Theorem 11-2, it is a simple
matter to show that an R-niodule M satisfies the ascending (descending)
chain condition on submodules if and only if M satisfies the maximum
(minimum) condition on submodles; we leave the verification of this to
the reader.
The corning theorem indicates how the chain conditions on submodules
are affected by certain operations.
in?n
Theorem 12-14. 1) If the R-module M satisfies the ascending (descending) chain condition, then so does every homomorphic image of M.
2) Let N be a submodule of the R-module M. Then M satisfies the
ascending (descending) chain condition if and only if N and M/N both
satisfy it.
For the most part, the stated results are merely a translation ofTheorems
11-6 and 11-7 into the language of modules. What is new in the present
situation is that any submodule N of M inherits the ascending (descending)
chain condition. This follows from the fact that any submodule of N is
itself a submodule of M (a marked contrast to the behavior of ideals).
Before the reader collapses under a burden of definitions, let us turn our
attention to the matter of normal and composition series.
By a normal series for an R-module M is meant a (finite) chain of Rsubmodules running from M to {O}:
Example 12-7. If R = F, where F is an arbitrary field, the R-homomorphisms are just the linear mappings (linear transformations) from M to N.
Example 12-8. Let N be a submodule of the R-module M. The function
nat N : M -> M/N which assigns to each element a E M its coset a + N is
an R-homomorphism; for, by definition,
nat N (ra)
ra
+N =
r(a
N) = r nat N (a).
\\, )
,----'
~I,'.
'"/
,
f(M).
M = Mo
;;2
, ...
o',
M1
M n- I
;;2
Mn =
{O}.
is .said to be a refinement of
M = Mo ::::>
M1
-
::::>
... ::::>
-
M n-1
::::>
-
Mn
= {O}
provided that there exists a one-to-one function f from {O, 1, ... , n} into
{O, 1, ... , m} such that Mi = N f(i)' This amounts to saying that every Mi
must appear as one of the N j A refinement of a normal series is termed
proper if the refinement contains a submodule not in the original series. A
normal series which adrnits no proper refinement is called a composition
series. We summarize this in the following definition.
::::>
MI
::::> '"
:;:>
Mn-I
::::>
M n = {O}
./
250
condition can also be applied to R-modules, the sole difference being that,
in our earlier definitions, the term"ideal" must now be replaced by the word
"submodule". Adapting the argument of Theorem 11-2, it is a simple
matter to show that an R-niodule M satisfies the ascending (descending)
chain condition on submodules if and only if M satisfies the maximum
(minimum) condition on submodles; we leave the verification of this to
the reader.
The corning theorem indicates how the chain conditions on submodules
are affected by certain operations.
in?n
Theorem 12-14. 1) If the R-module M satisfies the ascending (descending) chain condition, then so does every homomorphic image of M.
2) Let N be a submodule of the R-module M. Then M satisfies the
ascending (descending) chain condition if and only if N and M/N both
satisfy it.
For the most part, the stated results are merely a translation ofTheorems
11-6 and 11-7 into the language of modules. What is new in the present
situation is that any submodule N of M inherits the ascending (descending)
chain condition. This follows from the fact that any submodule of N is
itself a submodule of M (a marked contrast to the behavior of ideals).
Before the reader collapses under a burden of definitions, let us turn our
attention to the matter of normal and composition series.
By a normal series for an R-module M is meant a (finite) chain of Rsubmodules running from M to {O}:
Example 12-7. If R = F, where F is an arbitrary field, the R-homomorphisms are just the linear mappings (linear transformations) from M to N.
Example 12-8. Let N be a submodule of the R-module M. The function
nat N : M -> M/N which assigns to each element a E M its coset a + N is
an R-homomorphism; for, by definition,
nat N (ra)
ra
+N =
r(a
N) = r nat N (a).
\\, )
,----'
~I,'.
'"/
,
f(M).
M = Mo
;;2
, ...
o',
M1
M n- I
;;2
Mn =
{O}.
is .said to be a refinement of
M = Mo ::::>
M1
-
::::>
... ::::>
-
M n-1
::::>
-
Mn
= {O}
provided that there exists a one-to-one function f from {O, 1, ... , n} into
{O, 1, ... , m} such that Mi = N f(i)' This amounts to saying that every Mi
must appear as one of the N j A refinement of a normal series is termed
proper if the refinement contains a submodule not in the original series. A
normal series which adrnits no proper refinement is called a composition
series. We summarize this in the following definition.
::::>
MI
::::> '"
:;:>
Mn-I
::::>
M n = {O}
./
252
Proa! Suppose that b9th chain conditions and, hence, the maximum and
minimum conditions, hold in M. Applying the maximum condition to the
set of submodules different from M, we can select a maximal submodule
MI e M. Now, either MI = {O} and the proof halts, or there exists a
submodule M 2 of MI which is maximal with respect to being proper.
Continuing in this way, we get a strict1y decreasing chain of R-submodules
ofM:
such that the quotient modules M/M+ I are all simple; in other words,
the inclusions M ;2 N ;2 M+I' where N is a submodule of M,imply
that either N = Mi or N = M+I' The number. of submodules in a
composition series is called the length of the series.
Two normal series for the R-module M,
M = Mo :::J MI :::J .,. :::J M n - I ,:::J' M n = {O}
and'
M
= No :::J NI
::::l.
:::J N m -
:::J N m
= {O}
th~';~,
.J.,!,
\.
1) M
Z24
---------_..
{O},
{O}
This implies that the module NjN+1 is simple and so the chain (2) actualIy
comprises a composition series for M/Mn - I. As a result, we are able to
conc1ude that I(M/Mn _ l ) = n - 1. By our induction assumption, the
quotient module M/Mn - I must satisfy both the ascending ,and descending
chain conditions. Since M n - I is a simple R-module, an appeal to Theorem
12-14 is legitimate; we thus deduce that M itself satisfies both chain
conditions for submodules.
and
both form composition series for Z24' One way to verify this is to check
the orders of the subgroups involved. For instance, to inserf a submodule
between (2) and (4) there wquld have to exist a subgroup of Z24 of order n,
6 < n < 12, such that n divides 12 and is itself divisible by 6; clearly, no
such subgroup exists.
:::::J
will form a normal chain for the quotient module N = M/M n _ l . Using
the first isomorphism theorem for modules (Problem 26, Chapter 12),
Z24
Mo
253
_----~--,,-----------------------
---
------~
--
- - -
252
Proa! Suppose that b9th chain conditions and, hence, the maximum and
minimum conditions, hold in M. Applying the maximum condition to the
set of submodules different from M, we can select a maximal submodule
MI e M. Now, either MI = {O} and the proof halts, or there exists a
submodule M 2 of MI which is maximal with respect to being proper.
Continuing in this way, we get a strict1y decreasing chain of R-submodules
ofM:
such that the quotient modules M/M+ I are all simple; in other words,
the inclusions M ;2 N ;2 M+I' where N is a submodule of M,imply
that either N = Mi or N = M+I' The number. of submodules in a
composition series is called the length of the series.
Two normal series for the R-module M,
M = Mo :::J MI :::J .,. :::J M n - I ,:::J' M n = {O}
and'
M
= No :::J NI
::::l.
:::J N m -
:::J N m
= {O}
th~';~,
.J.,!,
\.
1) M
Z24
---------_..
{O},
{O}
This implies that the module NjN+1 is simple and so the chain (2) actualIy
comprises a composition series for M/Mn - I. As a result, we are able to
conc1ude that I(M/Mn _ l ) = n - 1. By our induction assumption, the
quotient module M/Mn - I must satisfy both the ascending ,and descending
chain conditions. Since M n - I is a simple R-module, an appeal to Theorem
12-14 is legitimate; we thus deduce that M itself satisfies both chain
conditions for submodules.
and
both form composition series for Z24' One way to verify this is to check
the orders of the subgroups involved. For instance, to inserf a submodule
between (2) and (4) there wquld have to exist a subgroup of Z24 of order n,
6 < n < 12, such that n divides 12 and is itself divisible by 6; clearly, no
such subgroup exists.
:::::J
will form a normal chain for the quotient module N = M/M n _ l . Using
the first isomorphism theorem for modules (Problem 26, Chapter 12),
Z24
Mo
253
_----~--,,-----------------------
---
------~
--
- - -
r
254
Lemma l.. Suppose that in the ring R the zero ideal is a product of
maximal ideals, say {O} = M 1M 2 Mil' If
(M 1M 2
M-1)/(M 1M 2
M)
N)(r
M) = xr
(r
R).
x(r - r')
+ (x
- x')r' E Ni-1M i
Ni S Ni'
No
;;2
Nl
;;2
N2
;;2 ;;2
N.
= {O}
Here now is the main result of this chapter; our proof follows the lines
of [36].
255
{JIJ is an ideal of R; 1: J
=1=
R}.
That is to say, 1: J' e l :,J'a =1= R, whence J'a e J'. Since this contradicts
the minimal nature of J':'in $', P must indeed be a prime ideal. Now, the
quotient ideal 1: P ;;2 1; inasmuch as 1: P ;;2 J, which is not contained in
1, it follows that l e 1: Pand our assertion is proved. (IncidentaUy, this
argument proves the existtmce of prime ideals in R.)
For the final stage of,the proof, let K be minimal in the set of those ideal s
of R which are products 0fiprime ideals. If K =1= {O}, then the ideal l = O: K
is different from R, for 1 fft l. By the last paragraph, there is a prime ideal
P of R such that l el: P; in other words, O: K e O: KP. This implies
that the ideal KP e K and contradicts the minimality of K. In consequence,
K = {O}, so that {O} is a product of prime (maximal) ideals. Lemnia 2
now completes our task.
.Corollary. Any commutative Artinian ring with identity is Noetherian.
Having come this far, it might be useful to pro ve Fitting's Lemma, a
result which requires both chain conditions on submodules. First, we pause
to establish a fact of independent interest.
.
r
254
Lemma l.. Suppose that in the ring R the zero ideal is a product of
maximal ideals, say {O} = M 1M 2 Mil' If
(M 1M 2
M-1)/(M 1M 2
M)
N)(r
M) = xr
(r
R).
x(r - r')
+ (x
- x')r' E Ni-1M i
Ni S Ni'
No
;;2
Nl
;;2
N2
;;2 ;;2
N.
= {O}
Here now is the main result of this chapter; our proof follows the lines
of [36].
255
{JIJ is an ideal of R; 1: J
=1=
R}.
That is to say, 1: J' e l :,J'a =1= R, whence J'a e J'. Since this contradicts
the minimal nature of J':'in $', P must indeed be a prime ideal. Now, the
quotient ideal 1: P ;;2 1; inasmuch as 1: P ;;2 J, which is not contained in
1, it follows that l e 1: Pand our assertion is proved. (IncidentaUy, this
argument proves the existtmce of prime ideals in R.)
For the final stage of,the proof, let K be minimal in the set of those ideal s
of R which are products 0fiprime ideals. If K =1= {O}, then the ideal l = O: K
is different from R, for 1 fft l. By the last paragraph, there is a prime ideal
P of R such that l el: P; in other words, O: K e O: KP. This implies
that the ideal KP e K and contradicts the minimality of K. In consequence,
K = {O}, so that {O} is a product of prime (maximal) ideals. Lemnia 2
now completes our task.
.Corollary. Any commutative Artinian ring with identity is Noetherian.
Having come this far, it might be useful to pro ve Fitting's Lemma, a
result which requires both chain conditions on submodules. First, we pause
to establish a fact of independent interest.
.
256
PROBLEMS
f"(y)
257
O, or, equiva-
x = f"(y)
R-submodules
M
I
2.
f(M)
2.
f2(M)
2. ....
PROBLEMS
Unles8 indicated to the contrary, all rings considered are assumed to be commutative
.
with identity,
-JT:2
= n(M) (El
2. a) In the polynomial ring F[xJ, where F is a field, show thatthe ideal (x 2 , 2x, 4)
Is primary, but reducible. [Hint: (X2, 2x, 4) = (X2, 2) n (x, 4).]
b) Express the ideal (x 2 , xy, 3) as an intersection ofprimary ideals in F[x, y].
ker no
n7=
M2. f(M)
2.
P(M)
2. ... ,
= ni=
S;;; 1. If 1
1 Q
is an irredundant primary representation of 1, establish that
.
a) 1/1 =
(QJJ) is an irredundant primary representation of the ideal 1/1
in R/J;
b) .J(QJJ) =JQJJ are the associated prime ideals of J/J.
I~
,'
1:
256
PROBLEMS
f"(y)
257
O, or, equiva-
x = f"(y)
R-submodules
M
I
2.
f(M)
2.
f2(M)
2. ....
PROBLEMS
Unles8 indicated to the contrary, all rings considered are assumed to be commutative
.
with identity,
-JT:2
= n(M) (El
2. a) In the polynomial ring F[xJ, where F is a field, show thatthe ideal (x 2 , 2x, 4)
Is primary, but reducible. [Hint: (X2, 2x, 4) = (X2, 2) n (x, 4).]
b) Express the ideal (x 2 , xy, 3) as an intersection ofprimary ideals in F[x, y].
ker no
n7=
M2. f(M)
2.
P(M)
2. ... ,
= ni=
S;;; 1. If 1
1 Q
is an irredundant primary representation of 1, establish that
.
a) 1/1 =
(QJJ) is an irredundant primary representation of the ideal 1/1
in R/J;
b) .J(QJJ) =JQJJ are the associated prime ideals of J/J.
I~
,'
1:
258
PROBLEMS
.JI
e)
is a prime ideal if and only if 1 has a single minimal prime.
d) If Pis a minimal prinle ideal of 1, then the primary eomponent eorresponding
to P is the same for all irredundant primary representations of l.
6. Let 1 be an ideal of a Noetherian ring R in whieh every nontrivial prime ideal is
maxima!. Show that 1 is a produet of primary ideals. [Hint: If 1 = ni Qi' then
the ideals .jQ are pairwise eomaximal when non trivial ; now use Problem 13,
Chapter 1O.J
7. Let R be a Noetherian ring, and 1 and J two ideals of R with 1 f K Prove that
l:J = 1 if and only if J is eontained in no assoeiated prime ideal of l. [Hint:
Assume that 1 = ni Q, Qi prinlary. If J
.jQ; for all i, then, by Problem 24(a),
Chapter 5, Q:J = Q. Conversely, let l:J = l. If J s .jQ;., then Jn s (.jQ,Jn S
Qk for some n; whenee 1 = 1:J" = ni (Qi :J") = ni'/'k (Qi :J") 2 ni'/'k Q 2 J.J
.nr=
10. Assume that R is a principal ideal ring with zero prinle radica!. Deduce that the
.. zero ideal is the interseetion of a finite number of prinle ideals.
11. Given that 1 is an ideal of the Noetherian ring R, establish the following:
a) If 1 S rad R, then
1" = {O}.
b)
(1 + (rad R)n) = 1. [Hint: Apply part (a) to R/l.J
e) If 1 + radR = R, then 1 = K [Hint: R = Rn =
(1 + radRf S
{l + (rad R)n) = 1.J
n:=l
n:..
n:=
n:..
(I + Mn) = 1.J
n:..
n:=
259
13. a) Derive the Krull Interseetion Theorem from Theorem 12-10. [Hint: Problem
8(b).J
.
.
b) Show that the set-theoretie eondition .jQ; n (1 - l) = 0 appearmg In
Theorem 12-10 is equivalent to requiring that .jQi + 1 f K
14. Let 1 be a proper ideal of the integral domain K Assume further that, for any
ideal J of R, there exists an integer k for whieh lk n J S IJ (when R is a
Noetherian domain, every ideal! has this property [33J). Prove that the interseetion
1" = {O}. [Hint: Takea E
In and eonsider the prineipalidealJ = (a).J
n:..
n:=
15. Suppose that R is a local ring whose maxinlal ideal M is principal, say M = (p).
If M is a ni! ideal of R, pro ve that
a) M is a nilpotent ideal of K
b) For any proper ideal 1 of R, 1 = ann (ann 1). [Hint: By Theorem 12-11,
1 = (p~ for sorne integer k.J
16. Let R be a ring possessing an ideal M whieh is both maximill and ni!. Verify that
R is a local ring with unique maximal ideal M. [Hint: Ifthe element a ~ M, show
that a is invertible by expanding (ab - l)n.J
In Problems 17-26, the term R-module means left R-module.
U::
. .,~. ~
258
PROBLEMS
.JI
e)
is a prime ideal if and only if 1 has a single minimal prime.
d) If Pis a minimal prinle ideal of 1, then the primary eomponent eorresponding
to P is the same for all irredundant primary representations of l.
6. Let 1 be an ideal of a Noetherian ring R in whieh every nontrivial prime ideal is
maxima!. Show that 1 is a produet of primary ideals. [Hint: If 1 = ni Qi' then
the ideals .jQ are pairwise eomaximal when non trivial ; now use Problem 13,
Chapter 1O.J
7. Let R be a Noetherian ring, and 1 and J two ideals of R with 1 f K Prove that
l:J = 1 if and only if J is eontained in no assoeiated prime ideal of l. [Hint:
Assume that 1 = ni Q, Qi prinlary. If J
.jQ; for all i, then, by Problem 24(a),
Chapter 5, Q:J = Q. Conversely, let l:J = l. If J s .jQ;., then Jn s (.jQ,Jn S
Qk for some n; whenee 1 = 1:J" = ni (Qi :J") = ni'/'k (Qi :J") 2 ni'/'k Q 2 J.J
.nr=
10. Assume that R is a principal ideal ring with zero prinle radica!. Deduce that the
.. zero ideal is the interseetion of a finite number of prinle ideals.
11. Given that 1 is an ideal of the Noetherian ring R, establish the following:
a) If 1 S rad R, then
1" = {O}.
b)
(1 + (rad R)n) = 1. [Hint: Apply part (a) to R/l.J
e) If 1 + radR = R, then 1 = K [Hint: R = Rn =
(1 + radRf S
{l + (rad R)n) = 1.J
n:=l
n:..
n:=
n:..
(I + Mn) = 1.J
n:..
n:=
259
13. a) Derive the Krull Interseetion Theorem from Theorem 12-10. [Hint: Problem
8(b).J
.
.
b) Show that the set-theoretie eondition .jQ; n (1 - l) = 0 appearmg In
Theorem 12-10 is equivalent to requiring that .jQi + 1 f K
14. Let 1 be a proper ideal of the integral domain K Assume further that, for any
ideal J of R, there exists an integer k for whieh lk n J S IJ (when R is a
Noetherian domain, every ideal! has this property [33J). Prove that the interseetion
1" = {O}. [Hint: Takea E
In and eonsider the prineipalidealJ = (a).J
n:..
n:=
15. Suppose that R is a local ring whose maxinlal ideal M is principal, say M = (p).
If M is a ni! ideal of R, pro ve that
a) M is a nilpotent ideal of K
b) For any proper ideal 1 of R, 1 = ann (ann 1). [Hint: By Theorem 12-11,
1 = (p~ for sorne integer k.J
16. Let R be a ring possessing an ideal M whieh is both maximill and ni!. Verify that
R is a local ring with unique maximal ideal M. [Hint: Ifthe element a ~ M, show
that a is invertible by expanding (ab - l)n.J
In Problems 17-26, the term R-module means left R-module.
U::
. .,~. ~
260
PROBLEMS
x=
XI
x2
b) TR is an ideal of homRR;
e) ,the mapping f(a) =' T;. determines a (ring) homomorphism of R onto Tn ;
,d) if for each O 1= a E R, there exists an element b E R such that ab 1= O, then
. R z TR (hence, R can be imbedded as an ideal in a ring with identity);
. e) whenever R has a multiplicative identity, then TR = homRR.
+ ... + x.
,:~.1;. ,,\
'::23.
X E
Rx = {rxlrER};
"
b) if N, N 2 are submodules of M, with N simple, and if N n N 2 1= {O}, then
N ~ N 2 ,
25. Derive tbe Second Isomorphism Theorem for Modules: If N and N 2 are two
submodules of the R-module M, then N/(N n N;) '" (N-j- N 2 )/N 2 . [Hint:
Mimic the argument ofTheorem 3-10.]
26. Derive the First lsomorphism Theorem for Modules: lf NI and N 2 are two
submodules of an R-module' M with N ~ N 2 , then N 2 /Nl' is a submodule of
M/N and
(M/N)/(Nz/N) z M/N 2
261
.;1.(
',:'
,"
260
PROBLEMS
x=
XI
x2
b) TR is an ideal of homRR;
e) ,the mapping f(a) =' T;. determines a (ring) homomorphism of R onto Tn ;
,d) if for each O 1= a E R, there exists an element b E R such that ab 1= O, then
. R z TR (hence, R can be imbedded as an ideal in a ring with identity);
. e) whenever R has a multiplicative identity, then TR = homRR.
+ ... + x.
,:~.1;. ,,\
'::23.
X E
Rx = {rxlrER};
"
b) if N, N 2 are submodules of M, with N simple, and if N n N 2 1= {O}, then
N ~ N 2 ,
25. Derive tbe Second Isomorphism Theorem for Modules: If N and N 2 are two
submodules of the R-module M, then N/(N n N;) '" (N-j- N 2 )/N 2 . [Hint:
Mimic the argument ofTheorem 3-10.]
26. Derive the First lsomorphism Theorem for Modules: lf NI and N 2 are two
submodules of an R-module' M with N ~ N 2 , then N 2 /Nl' is a submodule of
M/N and
(M/N)/(Nz/N) z M/N 2
261
.;1.(
',:'
,"
THIRTEEN
This, our concluding chapter, is designed primarily for the reader who
wishes to know somet,rnng about noncommutative ideal theory. It is not
our intention to treat this subject in any exhaustive manner; rather, we have
concentrated on those major results which could be fitted into a concise
development and which do not require many specialized preliminaries (even
with this restraint, some of the theorems are fairly sophisticated). Particular
effort is devoted to provingthe far-reaching Wedderburn Structure Theorems
for nil-semisimple rings satisfying the descending chain condition on right
ideals. These and other related results make intriguing use of the g~neral
theory of idempotents, as developed in the present chapter.
From this point onward, R will denote a ring with identity element,
not necessarily commutative (for most of our work the assumption of an
identity is not i:eally essential).
In the previous chapters, consider,able progress was made after imposing
a chain condition on the ideals oftll.e, ring; this was an entirely natural
procedure and it is equally expedie~~:~o do so here. To have a concise
statement, we shall can a ring R rig~t\ {1rtinian if it satisfies the descending
chain condition on right ideals. This chain condition adniits the usual
equivalent formulation: every nonempty set of right ideals of R possesses a
minimal member. An important t~~rem of Brauer, which requires only
the hypothesis that R be right ArtiniJ;l. is that each nonnilpotent right ideal
of R contains an idempotent element' We choose to begin our discussion
with a proof of this result.
Theorem 13-1. (Brauer). In a right Artinian ring R, every nonnilpotent
right ideal I contains a nonzero idempotent elemento
Proof. The collection of nonnilpotent right ideals of R which are contained
in 1 is not empty, for 1 itself is such an ideaL By the minimum condition on
right ideals (equivalent to the assumed chain condition), there exists a
mnimal member 11 of this collection. In particular, any right ideal of R
properly included in 1 1 must be nilpotent. Since li forros a nonnilpotent
right ideal contained in 1 1, it follows that
= 1 l'
lt
262
263
Now, consider the family of all right ideals J of R with the proper~ies
(i) Jl 1 =fo {O} and (H) J 5 1 1 , Such ideals certainly exist, for we have Just
seen that 1 1 satisfies the indicated conditions. From among these ideals, a
mnimal one can be obtained, call it J l' Using (i), there exists an element
u =fo O in J 1 such that ul1 =fo {O}. Since u/ 1 is a right ideal of R contained
in 1 1, with
(ul 1 )/ 1 = ulf = u/ 1 =fo {O},
the minimality of JI inplies that ul 1 = J l' As a result, it is possible to
find an element a E 1 1 5 I for which ua = u. Hence,
u=ua=ua 2
or u = ua" for all n. The conclusion is that a is not nilpotent and, in
consequence, 1 cannot be a nil right ideal.
The key to constructing the required idempotent is to consider the
right annihilator of u in / 1 , defined by
A(u) = {r E 11 1ur
O}.
Proof. Necessity follows from the definition of nilpotent ideal. That the
stated condition is also sufficient is a direct consequence of the theorem and
the observation that a nonzero idempotent cannot be nilpotent.
THIRTEEN
This, our concluding chapter, is designed primarily for the reader who
wishes to know somet,rnng about noncommutative ideal theory. It is not
our intention to treat this subject in any exhaustive manner; rather, we have
concentrated on those major results which could be fitted into a concise
development and which do not require many specialized preliminaries (even
with this restraint, some of the theorems are fairly sophisticated). Particular
effort is devoted to provingthe far-reaching Wedderburn Structure Theorems
for nil-semisimple rings satisfying the descending chain condition on right
ideals. These and other related results make intriguing use of the g~neral
theory of idempotents, as developed in the present chapter.
From this point onward, R will denote a ring with identity element,
not necessarily commutative (for most of our work the assumption of an
identity is not i:eally essential).
In the previous chapters, consider,able progress was made after imposing
a chain condition on the ideals oftll.e, ring; this was an entirely natural
procedure and it is equally expedie~~:~o do so here. To have a concise
statement, we shall can a ring R rig~t\ {1rtinian if it satisfies the descending
chain condition on right ideals. This chain condition adniits the usual
equivalent formulation: every nonempty set of right ideals of R possesses a
minimal member. An important t~~rem of Brauer, which requires only
the hypothesis that R be right ArtiniJ;l. is that each nonnilpotent right ideal
of R contains an idempotent element' We choose to begin our discussion
with a proof of this result.
Theorem 13-1. (Brauer). In a right Artinian ring R, every nonnilpotent
right ideal I contains a nonzero idempotent elemento
Proof. The collection of nonnilpotent right ideals of R which are contained
in 1 is not empty, for 1 itself is such an ideaL By the minimum condition on
right ideals (equivalent to the assumed chain condition), there exists a
mnimal member 11 of this collection. In particular, any right ideal of R
properly included in 1 1 must be nilpotent. Since li forros a nonnilpotent
right ideal contained in 1 1, it follows that
= 1 l'
lt
262
263
Now, consider the family of all right ideals J of R with the proper~ies
(i) Jl 1 =fo {O} and (H) J 5 1 1 , Such ideals certainly exist, for we have Just
seen that 1 1 satisfies the indicated conditions. From among these ideals, a
mnimal one can be obtained, call it J l' Using (i), there exists an element
u =fo O in J 1 such that ul1 =fo {O}. Since u/ 1 is a right ideal of R contained
in 1 1, with
(ul 1 )/ 1 = ulf = u/ 1 =fo {O},
the minimality of JI inplies that ul 1 = J l' As a result, it is possible to
find an element a E 1 1 5 I for which ua = u. Hence,
u=ua=ua 2
or u = ua" for all n. The conclusion is that a is not nilpotent and, in
consequence, 1 cannot be a nil right ideal.
The key to constructing the required idempotent is to consider the
right annihilator of u in / 1 , defined by
A(u) = {r E 11 1ur
O}.
Proof. Necessity follows from the definition of nilpotent ideal. That the
stated condition is also sufficient is a direct consequence of the theorem and
the observation that a nonzero idempotent cannot be nilpotent.
264
+ ei -
e 2 e l = e1
',"1
{O},
llex = O}
265
el(ee l )
= ei =
el =1= O;
l = el
eR
l,
ideal l
ex
e2 y
= ey = X.
erre)
(er)e
= ero
e'.
264
+ ei -
e 2 e l = e1
',"1
{O},
llex = O}
265
el(ee l )
= ei =
el =1= O;
l = el
eR
l,
ideal l
ex
e2 y
= ey = X.
erre)
(er)e
= ero
e'.
266
a= re = (re)e = ae EJe,
= (Je)R =
J(eR) = JI
5;
J,
267
R = 1 1 EB 12 EB o.. EB 1.0
To complete the proof, it remains only to show that the 1i include aIl
the minimal two-sided ideal s of R. Pursuing this aim, let 1 =1= {O} be any
minimal ideal of R. Since R admits the direct sum decomposition
R = 1 1 Ei112 EB oo. EB l., we thus have
1 = R1 = 1 11 EB 121 EB ...
1.1.
266
a= re = (re)e = ae EJe,
= (Je)R =
J(eR) = JI
5;
J,
267
R = 1 1 EB 12 EB o.. EB 1.0
To complete the proof, it remains only to show that the 1i include aIl
the minimal two-sided ideal s of R. Pursuing this aim, let 1 =1= {O} be any
minimal ideal of R. Since R admits the direct sum decomposition
R = 1 1 Ei112 EB oo. EB l., we thus have
1 = R1 = 1 11 EB 121 EB ...
1.1.
268
lJ
+ lz + ... +
lk'
where '{i 1,":!2' ... , ik} is a set of distinct integers between 1 and n. It follows
that ther~:are exactIy 2n ideal s in R, namely, the ideals 1! + 12 + .;. + lk'
,"i,
,t.
'
(r.E R).
=
+ e2 + ... + en'
+ e~r2 +
e 1 r 1 + e2 r2 +
eir 1
Proo! Before embarking on the proof proper, we note that the set eRe
forms a nonzero subring of R with identity element e. Suppose first that
eR is a minimal right ideal of R. To show that eRe comprises a division
ring, it is enough to find an inverse for each nonzero elemerit. If O 9= ere E eRe,
holds, making e a left identity for R. On the other hand, consider the left
ideal 1 = {r - reir E R}. Because R = eR, we have
(1' - re)R = (1' - re)eR = r(e - e2)R = {O},
---
-_._------------
----------------
In are.generated
' ,.
.... ,
where
1 = el
269
------
-----
---- - - -
268
lJ
+ lz + ... +
lk'
where '{i 1,":!2' ... , ik} is a set of distinct integers between 1 and n. It follows
that ther~:are exactIy 2n ideal s in R, namely, the ideals 1! + 12 + .;. + lk'
,"i,
,t.
'
(r.E R).
=
+ e2 + ... + en'
+ e~r2 +
e 1 r 1 + e2 r2 +
eir 1
Proo! Before embarking on the proof proper, we note that the set eRe
forms a nonzero subring of R with identity element e. Suppose first that
eR is a minimal right ideal of R. To show that eRe comprises a division
ring, it is enough to find an inverse for each nonzero elemerit. If O 9= ere E eRe,
holds, making e a left identity for R. On the other hand, consider the left
ideal 1 = {r - reir E R}. Because R = eR, we have
(1' - re)R = (1' - re)eR = r(e - e2)R = {O},
---
-_._------------
----------------
In are.generated
' ,.
.... ,
where
1 = el
269
------
-----
---- - - -
270
x[(ere)x](ere)y
This enables us to conc1ude .that each right inverse is also a left inverse,
yielding the desired outcome.
As regards the con~t'(rse, as sume that eRe constitutes a division ring and
let 1 =1= {O} be any right ideal of R contained in eRo This gives el = l.
Notice also that le =1= {O}; in the contrary case, 12 ~ leR = {O}, which
conflicts with our hypothesis that R has no nonzero nilpotent ideals.
Accordingly, there exists an element r E 1 such that re =1= and, :;ince 1 = el,
we must have ere = re =1= O. Because eRe is taken to be a division ring, ere
possesses an inverse s E eRe. But, ere El; hence, e = (ere)s E l. This forces
eR ~ 1 and the equality 1 = eR follows.
CoroIlary. Let R be a nil-semisimple ring andlet =1= e E R be idempotent. Then.eR is a minimal right ideal if and only if Re is a minimal
l.: '~,,'
left ideal.
It is reasonable to ask ~h~ther the statement of Theorem 1.3-5 could be
improved upon by the stipuliltion of a chain condition. In pursuit of an
answer, we make the follQwing definition. An idempotent =1= e E R is
called primitive if e is not t4~'sum of two orthogonal nonzero idempotents
of R; that is, it is not posjble to write e = u + v, where u 2 = U =1= 0,
2
v = v =1= and uv = vu ~(O.
We can characterize wlen an idempotent element of R is primitive in
terms of the idempotents of the ring eRe. To be precise:
Proof. Let e be primitive and assume that ere is idempotent for sorne r E R.
Then the element e - ere is also idempotent:
(e - ere)2 = e2 - e2re - ere 2 + (ere)2
= -e - ere - ere
+ ere
271
e - ere.
w2
vw
=
=
wv
e2 - eue - ue 2 + (ue)2
ue(e - ue) = ue 2 - (ue)2 ,;,. ue - ue
(e - ue)2
(e - ue)ue
eue - (ue)2
ue - ue
e - ue - ue
+ ue
= 0,
=
O.
were)
= w,
Indeed,
270
x[(ere)x](ere)y
This enables us to conc1ude .that each right inverse is also a left inverse,
yielding the desired outcome.
As regards the con~t'(rse, as sume that eRe constitutes a division ring and
let 1 =1= {O} be any right ideal of R contained in eRo This gives el = l.
Notice also that le =1= {O}; in the contrary case, 12 ~ leR = {O}, which
conflicts with our hypothesis that R has no nonzero nilpotent ideals.
Accordingly, there exists an element r E 1 such that re =1= and, :;ince 1 = el,
we must have ere = re =1= O. Because eRe is taken to be a division ring, ere
possesses an inverse s E eRe. But, ere El; hence, e = (ere)s E l. This forces
eR ~ 1 and the equality 1 = eR follows.
CoroIlary. Let R be a nil-semisimple ring andlet =1= e E R be idempotent. Then.eR is a minimal right ideal if and only if Re is a minimal
l.: '~,,'
left ideal.
It is reasonable to ask ~h~ther the statement of Theorem 1.3-5 could be
improved upon by the stipuliltion of a chain condition. In pursuit of an
answer, we make the follQwing definition. An idempotent =1= e E R is
called primitive if e is not t4~'sum of two orthogonal nonzero idempotents
of R; that is, it is not posjble to write e = u + v, where u 2 = U =1= 0,
2
v = v =1= and uv = vu ~(O.
We can characterize wlen an idempotent element of R is primitive in
terms of the idempotents of the ring eRe. To be precise:
Proof. Let e be primitive and assume that ere is idempotent for sorne r E R.
Then the element e - ere is also idempotent:
(e - ere)2 = e2 - e2re - ere 2 + (ere)2
= -e - ere - ere
+ ere
271
e - ere.
w2
vw
=
=
wv
e2 - eue - ue 2 + (ue)2
ue(e - ue) = ue 2 - (ue)2 ,;,. ue - ue
(e - ue)2
(e - ue)ue
eue - (ue)2
ue - ue
e - ue - ue
+ ue
= 0,
=
O.
were)
= w,
Indeed,
272
Tbeorem 13-7. In a nil-semisimple rigbt Artinian ring R, an idempotent e' =1= O is primitive if and only if eRe forms a division ringo
We continue our development after a few preparatory remarks about
modules. Let M, N be nonzero rigbt R-modules, wbere R is any ringo
(The reader is reniinded that, wbenever tbere is reference to a' ring R, it is
tacitly assumed tbat R possesses an identity elem~t.) We shall bereafter use
the notation bomR(M, N) to denote the set of all R-bomomorpbisms from
M into N; in other words, tbe set of all mappin~s, f: M -+ N sucb tbat
~
f{x
y) = f(x)
+ f(y),
f(xr)
..
~.
= f(x)r'::,;;
= (j{x)
wbence f
g(x)r
- g(x)r .;" (f
g)(x)r,
(f o g)(xr)
= f(g(xr)
= f(g(x)r)
(f o g)(x)r,
(T,. o f)(x)
(x E M),
f(x)r
(x EM).
= f(xs)r
f(x)sr
= f(x)rs
(fr)(x)s.
= f(g(x)r
273
eR,
Thus,
(a, bE 1; rE R).
(f + g)(e)
fe)
+ g(e) =
cp(f)
CP(g).
cp(f o g) = (f o g)(e)
=
j(g(e) = f(ete)
= CP(f)CP(g).
272
Tbeorem 13-7. In a nil-semisimple rigbt Artinian ring R, an idempotent e' =1= O is primitive if and only if eRe forms a division ringo
We continue our development after a few preparatory remarks about
modules. Let M, N be nonzero rigbt R-modules, wbere R is any ringo
(The reader is reniinded that, wbenever tbere is reference to a' ring R, it is
tacitly assumed tbat R possesses an identity elem~t.) We shall bereafter use
the notation bomR(M, N) to denote the set of all R-bomomorpbisms from
M into N; in other words, tbe set of all mappin~s, f: M -+ N sucb tbat
~
f{x
y) = f(x)
+ f(y),
f(xr)
..
~.
= f(x)r'::,;;
= (j{x)
wbence f
g(x)r
- g(x)r .;" (f
g)(x)r,
(f o g)(xr)
= f(g(xr)
= f(g(x)r)
(f o g)(x)r,
(T,. o f)(x)
(x E M),
f(x)r
(x EM).
= f(xs)r
f(x)sr
= f(x)rs
(fr)(x)s.
= f(g(x)r
273
eR,
Thus,
(a, bE 1; rE R).
(f + g)(e)
fe)
+ g(e) =
cp(f)
CP(g).
cp(f o g) = (f o g)(e)
=
j(g(e) = f(ete)
= CP(f)CP(g).
274
fea) = f(er)
f(e)r
=O
ere
x,
so that cp maps onto eRe. The proof that cp serves as a ring isomorphlsm
between homR(l, 1) and eRe is now complete.
In combination with'Theorem 13-5, one obtains as a corollary
CoroJlary. If R is a simple ring with minimal right ideal 1, then
homR(I,1) forms a division ringo
The foregoing corollary is a special case of a much wider theorem due
to Schur. In accordance with the terminology of Section 12, we shaJl call
an R-module simple provided it is nonzero and has no nontrivial submodules.
Tbeorem 13-9. (Schur's Lemma). If M is a simple R-module, then
homR(M, M) forms a division ringo
+ 1) +
(y
+ 1)
= x
+ y + l.
+ l)r = xr +
(r E R),
R/l carries a well-defined structure as a right R-module. Indeed, the verification tbat the module moms are satisfied presents no problem and will
275
therefore be omitted. The reader may also conftrm that the submodules of
R/l are of the form J/l, where J is a right ideal of R and J ;;;2 l. These
remarks lead us to a precise identification of aH simple R-modules.
Theorem 13-10. A right R-module M is simple if and only if it is Risomorphic to a quotient module R/l for some maximal right ideal 1
of R.
A(M) = {r E RIMr
{O}}
is called the annihilator of the module M.As is easily verified, A(M) forms
a two-sided ideal of R. We shall refer to Mas afaithful R-module or speak
of Ras acting faithful1y on M whenever A(M)
{O}; tbat is to say, whenever Mr = {O} implies that r = O. Since.the ring R is assumed to pqssess
an identity element, R is evidently Caithfulas a module over itself.
To round out our studies, let us introdce a noton of radical which is
meaningful for general rings and which agrees with the previously defined
. Jacobson radical in the event that the ringJs conuilUtative. The definition
we give is phrased in terms of annihilators oC modules.
Definition 13-1. The radical of the. ring R is the set J(R) = fl A(M),
where the intersection varies over all simple right R-modules M; if
there are no simple right R-modules, put J(R) = R.
In short, the radical of a ring R may be described as the annihilator of
all simple right R-modules. Notice, too, tbat J(R), being the intersection of
two-sided ideals oC R, itself is an ideal of R. A final point to which attention
should be drawn is that if R happens to have a faithful simple right R-
274
fea) = f(er)
f(e)r
=O
ere
x,
so that cp maps onto eRe. The proof that cp serves as a ring isomorphlsm
between homR(l, 1) and eRe is now complete.
In combination with'Theorem 13-5, one obtains as a corollary
CoroJlary. If R is a simple ring with minimal right ideal 1, then
homR(I,1) forms a division ringo
The foregoing corollary is a special case of a much wider theorem due
to Schur. In accordance with the terminology of Section 12, we shaJl call
an R-module simple provided it is nonzero and has no nontrivial submodules.
Tbeorem 13-9. (Schur's Lemma). If M is a simple R-module, then
homR(M, M) forms a division ringo
+ 1) +
(y
+ 1)
= x
+ y + l.
+ l)r = xr +
(r E R),
R/l carries a well-defined structure as a right R-module. Indeed, the verification tbat the module moms are satisfied presents no problem and will
275
therefore be omitted. The reader may also conftrm that the submodules of
R/l are of the form J/l, where J is a right ideal of R and J ;;;2 l. These
remarks lead us to a precise identification of aH simple R-modules.
Theorem 13-10. A right R-module M is simple if and only if it is Risomorphic to a quotient module R/l for some maximal right ideal 1
of R.
A(M) = {r E RIMr
{O}}
is called the annihilator of the module M.As is easily verified, A(M) forms
a two-sided ideal of R. We shall refer to Mas afaithful R-module or speak
of Ras acting faithful1y on M whenever A(M)
{O}; tbat is to say, whenever Mr = {O} implies that r = O. Since.the ring R is assumed to pqssess
an identity element, R is evidently Caithfulas a module over itself.
To round out our studies, let us introdce a noton of radical which is
meaningful for general rings and which agrees with the previously defined
. Jacobson radical in the event that the ringJs conuilUtative. The definition
we give is phrased in terms of annihilators oC modules.
Definition 13-1. The radical of the. ring R is the set J(R) = fl A(M),
where the intersection varies over all simple right R-modules M; if
there are no simple right R-modules, put J(R) = R.
In short, the radical of a ring R may be described as the annihilator of
all simple right R-modules. Notice, too, tbat J(R), being the intersection of
two-sided ideals oC R, itself is an ideal of R. A final point to which attention
should be drawn is that if R happens to have a faithful simple right R-
276
:00 x(s
t - st) = xs
xt - xst = x
xt -:- xt = x,
Remark, For the reader's guidance, it needs to be' statedthat, when R lacks
an identity element, the analog of Theorem 13..,.11 asserts that J(R) is the
intersection of the modular maximal right ideals of R. (Of course, in a
ring with identity, a1l ideals aretriviaHy modular.) Furthermore, the intersection of the modular maximal right ideals of R always coincides with the
intersection of aH modular maxirnalleft ideals [10].
In order to extend Schur's Lemma, we introduce a further concept,
which, to begin with, may appear unrelated. Let (G, +) be a commutative
group having more than one element and let E(G) denote the collection of
endomorphisms of G. (Remember that by an endomorphism of G is meant
a homomorphisrn of G into itself.) Dnder the operations of pointwise
addition and functional composition, E(G) forms a ring with identity. Any
277
'.'
,-,o'
f(y)
= f(h(x 1
x 2 )
h(J(x 1
h) = h(O)
O,.
o,,:}:
276
:00 x(s
t - st) = xs
xt - xst = x
xt -:- xt = x,
Remark, For the reader's guidance, it needs to be' statedthat, when R lacks
an identity element, the analog of Theorem 13..,.11 asserts that J(R) is the
intersection of the modular maximal right ideals of R. (Of course, in a
ring with identity, a1l ideals aretriviaHy modular.) Furthermore, the intersection of the modular maximal right ideals of R always coincides with the
intersection of aH modular maxirnalleft ideals [10].
In order to extend Schur's Lemma, we introduce a further concept,
which, to begin with, may appear unrelated. Let (G, +) be a commutative
group having more than one element and let E(G) denote the collection of
endomorphisms of G. (Remember that by an endomorphism of G is meant
a homomorphisrn of G into itself.) Dnder the operations of pointwise
addition and functional composition, E(G) forms a ring with identity. Any
277
'.'
,-,o'
f(y)
= f(h(x 1
x 2 )
h(J(x 1
h) = h(O)
O,.
o,,:}:
r
278
(a
1
'1
of nght mUltiplicatioris T,., rE R, forms a primitive ring of (group) endomorphisms of M. Furthermore, the supposition that M is faithful implies
that T,. = O if and only if r = O. The mapP!ng p: r ~
thus determines
a ring-isomorphism of R onto a primitive ring of endomorphisms of M,
givin~ rise to the conc1usion that R is a primitive ringo
On the other hand, suppose that R is isomorphic by means of the
mapping r ~ r' to a primitive ring R' of endomorphisms of a commutative
group M. Define a module structure on M by setting xr = r'(x), where
x E M, rE R. It follows at once that Mr = {O} if and only if r'(x) = O for
all x E M; that is to say, if and only if the endomorphism r' = O, or,
equivalently, r = O. Thus, R acts faithfully on M as a right R-module.
To see that the module in question is simple, notice that the primitivity of
R' implies that M;= {r'(x)lr' E R'} for each nonzero x E M; this in its turn
tells us that M '= '):R for' any O =1= x E M, which makes M a simple Rmodule.
r..
= xr.
+ 1 = 1,
A(Rjl)
A(M)
= {O}.
Since {O} is a maximal ideal of the ring R, it follows that R must be a field.
In this connection, we should observe that every simple ring R (hence,
any field) is primitive. For, if 1 is a maximal right ideal of R, then Rjl is a
simple right R-module. Its annihilator A(RjI) forms a two-sided ideal of
R, which is necessarily proper since 1 rt A(Rjl). By hypothesis, we thus
have A(Rjl) = {O}, so that Rjl is a faithful simple R-module; this assures
that R is a primitive ringo
The theory of primitive rings is extensive but somewhat specialized
from our present point of view. We shall therefore break offthe discussion
in order to turn to more important matters.
Lemma. For any finite set Xl' x 2 , ... , XII in M, linearly independent
with respect to D, there exists an element a E R, such that
= x(ab)r
ar
Let M =1= {O} be a simple right R-module, where R is a given ring with
identity, and consider the cJivision ring D = homR(M, M) of all R-endomorphisms of M. For any fE D, we can define a (left) module product
fx by fx = f(x), where x E M. With this definition M beco mes in a natural
way a left D-module (that is, a vector space over the division ring D); the
verification is routine.
Bearing this in mind, we next present a technical lemma, the value o
. ,~which will become c1ear as we proceed.
xr(ab)
+ l)r =
Proo! Let M be any faithful simple R-module. Since Mis simp'le, we have
M =
. xR for each O =1=., x E M. The force of this o bservation is that the ring
279
xla
= ... = xII-la = O,
xlla
=1=
O.
'"
I
,.'
N={XEMlxR= {Ol}
{aERlxla = ...
x ll _ 2 a = O}.
r
278
(a
1
'1
of nght mUltiplicatioris T,., rE R, forms a primitive ring of (group) endomorphisms of M. Furthermore, the supposition that M is faithful implies
that T,. = O if and only if r = O. The mapP!ng p: r ~
thus determines
a ring-isomorphism of R onto a primitive ring of endomorphisms of M,
givin~ rise to the conc1usion that R is a primitive ringo
On the other hand, suppose that R is isomorphic by means of the
mapping r ~ r' to a primitive ring R' of endomorphisms of a commutative
group M. Define a module structure on M by setting xr = r'(x), where
x E M, rE R. It follows at once that Mr = {O} if and only if r'(x) = O for
all x E M; that is to say, if and only if the endomorphism r' = O, or,
equivalently, r = O. Thus, R acts faithfully on M as a right R-module.
To see that the module in question is simple, notice that the primitivity of
R' implies that M;= {r'(x)lr' E R'} for each nonzero x E M; this in its turn
tells us that M '= '):R for' any O =1= x E M, which makes M a simple Rmodule.
r..
= xr.
+ 1 = 1,
A(Rjl)
A(M)
= {O}.
Since {O} is a maximal ideal of the ring R, it follows that R must be a field.
In this connection, we should observe that every simple ring R (hence,
any field) is primitive. For, if 1 is a maximal right ideal of R, then Rjl is a
simple right R-module. Its annihilator A(RjI) forms a two-sided ideal of
R, which is necessarily proper since 1 rt A(Rjl). By hypothesis, we thus
have A(Rjl) = {O}, so that Rjl is a faithful simple R-module; this assures
that R is a primitive ringo
The theory of primitive rings is extensive but somewhat specialized
from our present point of view. We shall therefore break offthe discussion
in order to turn to more important matters.
Lemma. For any finite set Xl' x 2 , ... , XII in M, linearly independent
with respect to D, there exists an element a E R, such that
= x(ab)r
ar
Let M =1= {O} be a simple right R-module, where R is a given ring with
identity, and consider the cJivision ring D = homR(M, M) of all R-endomorphisms of M. For any fE D, we can define a (left) module product
fx by fx = f(x), where x E M. With this definition M beco mes in a natural
way a left D-module (that is, a vector space over the division ring D); the
verification is routine.
Bearing this in mind, we next present a technical lemma, the value o
. ,~which will become c1ear as we proceed.
xr(ab)
+ l)r =
Proo! Let M be any faithful simple R-module. Since Mis simp'le, we have
M =
. xR for each O =1=., x E M. The force of this o bservation is that the ring
279
xla
= ... = xII-la = O,
xlla
=1=
O.
'"
I
,.'
N={XEMlxR= {Ol}
{aERlxla = ...
x ll _ 2 a = O}.
280
--)o
(a
f(X"_la) = x"a
1)
',.
la
ror
281
R rt = {T.;I a E R;
There is no special difficulty in verifying tht::R. rt is a subring of the endomorphism ring homD(I, 1). At the same tile~ the correspondence a
Ya
sets up in a .natural way a nonzero homomoi'phism of R onto R rl ,. whose
kernel is a two-sided ideal of R. Inasmuch as R is assumed to be simple,
this kernel is {O}, whence the isomorphism R ,::: Rrl follows.
The main contention of our 'proof is that R rt
homD(I,
To settle
this point, note first that the set
--)o
f(x,.-la
x"_lb)
f(x,,-l(a
= x,,(a
=
xna
+ b)
+ xnb
b)
f(x n- 1a)
+ f(X"-lb).
Using the fact that 1 is a right ideal of R, we note further that if a E Iand
rE R, then
f(xn-1ar)
xn(ar)
(x"a)r = f(X"_la)r.
xla
...
= x,,_za
(fxn':' l
x.)a
O.
-+ fE D)
. ReR =
X2,'" , XII
Armed with this rather intricate machinery, we can now derive the
fundamental structure theorem for simple right Artinian rings (the so-called
Second Wedderburn Structure Theorem).. From the many ways of proving
this result, we select a module-theoretic approach essentially due to
Henderson [40].
fl:: aebla, b E R}
f(x)
Again, from our induction supposition that the emma holds for n - 1
elements, it follows that
(O
f(erl)
f(er
aeb) =
f(I eraeb)
= If(eraeb,)
= I eraef(eb;)
= er I aef(eb)
(since eraje
D)
xI aef(eb').
Fromthis formula, it appears thatf(x) = T.x, where the element s (which
does not depend on x) is given by s = I aef(eb}. Therefore, fE R rt ,
confirming that Rrl = hom D(I, 1). Puttirig our remarks together, we obtain
the isomorphism
280
--)o
(a
f(X"_la) = x"a
1)
',.
la
ror
281
R rt = {T.;I a E R;
There is no special difficulty in verifying tht::R. rt is a subring of the endomorphism ring homD(I, 1). At the same tile~ the correspondence a
Ya
sets up in a .natural way a nonzero homomoi'phism of R onto R rl ,. whose
kernel is a two-sided ideal of R. Inasmuch as R is assumed to be simple,
this kernel is {O}, whence the isomorphism R ,::: Rrl follows.
The main contention of our 'proof is that R rt
homD(I,
To settle
this point, note first that the set
--)o
f(x,.-la
x"_lb)
f(x,,-l(a
= x,,(a
=
xna
+ b)
+ xnb
b)
f(x n- 1a)
+ f(X"-lb).
Using the fact that 1 is a right ideal of R, we note further that if a E Iand
rE R, then
f(xn-1ar)
xn(ar)
(x"a)r = f(X"_la)r.
xla
...
= x,,_za
(fxn':' l
x.)a
O.
-+ fE D)
. ReR =
X2,'" , XII
Armed with this rather intricate machinery, we can now derive the
fundamental structure theorem for simple right Artinian rings (the so-called
Second Wedderburn Structure Theorem).. From the many ways of proving
this result, we select a module-theoretic approach essentially due to
Henderson [40].
fl:: aebla, b E R}
f(x)
Again, from our induction supposition that the emma holds for n - 1
elements, it follows that
(O
f(erl)
f(er
aeb) =
f(I eraeb)
= If(eraeb,)
= I eraef(eb;)
= er I aef(eb)
(since eraje
D)
xI aef(eb').
Fromthis formula, it appears thatf(x) = T.x, where the element s (which
does not depend on x) is given by s = I aef(eb}. Therefore, fE R rt ,
confirming that Rrl = hom D(I, 1). Puttirig our remarks together, we obtain
the isomorphism
282
283
1rr = {a E R/xla
T~en 11
= ... =
= O}
X' I
a J.
jo g
If Vis an n-dimensional vector space over the division ring D, then the
ring homD(J~ V) is well-known to be isomorphic to the ring M,,(D) of n x n
matrices over D. To spell out somedetails, let Vhave the basis {XI' X , '" , x,,}.
2
If j E homD(v, V), each of thse basis elements will be mapped by
j into a
uniquely determined linear combination of XI' x 2 , ... , x". In other words,
there exist elements ai E D, uniquely defined by j, such that j(x ) can be
expressed in the form
.
i
~-".
aijx
= I
~.
(Observe that the summation is over the first i:ndex.) Thus, to each endomorphism jE.homD(v, V) there corresponds #.i;unique n x n matrix (aij)
with entries from D.
: ,
There is no problem in showing that the ~flPj -+ (aij) yields a one-toone function from homD(v, V) onto the matrixring M,,(D). Indeed, starting
withanarbitrary(a) E M,,(D),onedefinesaD-endomorphismj E homD(V, V)
by first setting j(x) = = I aUx and then extending linearly to all of V;
it is evident that (a) is precisely the matrix identified with the resulting
endomorphism. If (a u ) and (bu) are the matrix representations of two
elements of homD(v, V), say j and g, then
(f + g)(x) = j(xi )
and
+ g(x)
=I
aijx
(aij)
+ (bu),
-+(
11
aikbkj) = (aj)(bJ
k=1
leading to the conc1usion that 1" is proper1y contained in In-l' The point
which we wish tp make is tha t 1 I ::J 12 ::J ::J 1" ::J ... is a strictly
descending chain, in violation of the assumption that R is a right Artinian.
Accordingly, the dimension dim D l < co and this finally ends the proof of
Theorem 13-14.
= "
-+ (aj + bu) =
j(xi )
+g
+ "
i=1
bjx =
"
i=1
(aij
b)x
. .
.
h' ch associates with eachj E hom D (V, V)
The conc1usIOn IS that the mappmg w .1
th fixed basis) induces a ring
its matrix representation {a u) (relatlve to e
isomorphism
homD(V, V) ~ M,,(D) ..
.
..
Its may be collected
?n thedstre~gtt~
to glve
a escnp IOnof ~7.e:.I_:::;~~feurri~~7~:tf~::n rings in terms of
matrix rings.
..
.
Th
Theorem 13-15. Let R be a nil-sernisimple right Artmta.n nng.
e~
there exist division rings Di and suitable integers ni (1 = 1, 2, ... , r
such that
R ~ M",(D I ) Etl M",(D 2 ) E!7 ... E!7 M,,/D r ).
PROBLEMS
Unless spedlied otherwise, R always (jenotes an arbitrary ring with identity.
282
283
1rr = {a E R/xla
T~en 11
= ... =
= O}
X' I
a J.
jo g
If Vis an n-dimensional vector space over the division ring D, then the
ring homD(J~ V) is well-known to be isomorphic to the ring M,,(D) of n x n
matrices over D. To spell out somedetails, let Vhave the basis {XI' X , '" , x,,}.
2
If j E homD(v, V), each of thse basis elements will be mapped by
j into a
uniquely determined linear combination of XI' x 2 , ... , x". In other words,
there exist elements ai E D, uniquely defined by j, such that j(x ) can be
expressed in the form
.
i
~-".
aijx
= I
~.
(Observe that the summation is over the first i:ndex.) Thus, to each endomorphism jE.homD(v, V) there corresponds #.i;unique n x n matrix (aij)
with entries from D.
: ,
There is no problem in showing that the ~flPj -+ (aij) yields a one-toone function from homD(v, V) onto the matrixring M,,(D). Indeed, starting
withanarbitrary(a) E M,,(D),onedefinesaD-endomorphismj E homD(V, V)
by first setting j(x) = = I aUx and then extending linearly to all of V;
it is evident that (a) is precisely the matrix identified with the resulting
endomorphism. If (a u ) and (bu) are the matrix representations of two
elements of homD(v, V), say j and g, then
(f + g)(x) = j(xi )
and
+ g(x)
=I
aijx
(aij)
+ (bu),
-+(
11
aikbkj) = (aj)(bJ
k=1
leading to the conc1usion that 1" is proper1y contained in In-l' The point
which we wish tp make is tha t 1 I ::J 12 ::J ::J 1" ::J ... is a strictly
descending chain, in violation of the assumption that R is a right Artinian.
Accordingly, the dimension dim D l < co and this finally ends the proof of
Theorem 13-14.
= "
-+ (aj + bu) =
j(xi )
+g
+ "
i=1
bjx =
"
i=1
(aij
b)x
. .
.
h' ch associates with eachj E hom D (V, V)
The conc1usIOn IS that the mappmg w .1
th fixed basis) induces a ring
its matrix representation {a u) (relatlve to e
isomorphism
homD(V, V) ~ M,,(D) ..
.
..
Its may be collected
?n thedstre~gtt~
to glve
a escnp IOnof ~7.e:.I_:::;~~feurri~~7~:tf~::n rings in terms of
matrix rings.
..
.
Th
Theorem 13-15. Let R be a nil-sernisimple right Artmta.n nng.
e~
there exist division rings Di and suitable integers ni (1 = 1, 2, ... , r
such that
R ~ M",(D I ) Etl M",(D 2 ) E!7 ... E!7 M,,/D r ).
PROBLEMS
Unless spedlied otherwise, R always (jenotes an arbitrary ring with identity.
284
m",
= {O} or 12 = 1;
in th second case, deduce' that J = eR = el for some idempotent e =1= O in l.
[Hint: See the proofs ofTheorems 13-1 and 13-2; first, establish the existence of
an element a e J ror whie~.l = J.]
1,
5. Assume !hat 1. and J are/t~o minimal right ideals of the ring R which are iso1, prove that
morphie as right R-modul~8: Ir J2
a) Any R-isomorphismj)~ J is orthe fornlf(x) = ax for some a ej.
b) The produet JJ
J.. [Hint: By part (a),. J
al, where a E J.]
6. For nonzero idempotents';'u of the ring R, prove that
a) eR = uR ir and onlylf 4 = u and ue = e.
b) eR and uR are isomorphi as R-modules if and only ifthere.exist x, y E R sueh
tbat xy = u and yx = e. [Hint: Sinee xe
ux and yu "" ey, the funetion
f(er) = x{er) = u(xr) defines an R-isomorphism f: eR ..... uR, with f-I(US) =
y(us) = e(ys).]
e) eR ~ uR, as R-modules, if and only if Re ~ Ru.
7. Let eR and uR be two minimal right ideals of the ring R, where e and u are nonzero idempotents of R. Show tbat eR and uR are R-isomorphe ifand only iftheir
produet uReR =1= {O}. [Hint: Ifure =1= Ofor some r E R, define the R-isomorphisnl
f: eR ..... uR by f(es) = (ure)s.]
.
8. Let R bea nil-semisimple ring without identity. Ifthe element a E Rand aR {O},
establish that a O.
[Hint: The ideal A(R) = {r E RlrR = {O}} satisfies
A(R)2 ~ A(R)R = {O}.]
9. Establish the statements below:
a) A right ideal J of the ring R is a direct summand of R if and oruy ir 1 = eR for
some idempotent e e R ;
b) a mininlal right ideal J of the ring R is a dreet summand of R if and only ir
12 =1= {O}.
10. Prove that a right Artinian fing R is nil-semisinlple ir and only ir 12
each mnimal rigbt ideal 1 of R.
285
PROBLEMS
=1=
{O} ror
11. Assuming tbat R is a nil-semisimple right Artinian ring, veriry the following
assertions:
a) R is rigbt Noetherian; that is, R satisfies the ascending ehain condition on right
ideals. [Hint: By Theorem 13-2, the right ideals of R are finiteIy generated.]
b) The mapping e ..... eR defines a one-to-one eorrespondence between the set of
all idempotent elements e e cent R and the set of two-sided ideaIs of R.
c) For any two-sided ideal 1 of R, ann,1 = ann1 and so R = 1 EB annl.
d) Every right ideal J of R is a direct sumnand of any right ideal containing it.
12. a) Suppose that the ring R is a finite direct sum of right ideals J =1= {O}, say
R = 1 fE) J 2 EB ... EB Jn Ir 1
e + e2 + ... + e", where eel" prove
that the elements e fornl a set of orthogonaI idempotents and that 11 = eR
(i = 1, 2, .. , , n).
b) If the ideals 1 of part (a) are aH two-sided, show that e E cent R an4 so serves
as an identity element for 1.
13. a)' Prove that an idempotent e =1= O of the ring R is prinlitive if and only if R
contains no idenlpotent u =1= e sueh that eu = ue = u.
bl,'Establish that any idempotent elenlent e =1= O of a nil-semisimple right Artinian
nlng R is the sum of a finite number of orthogonal primitive idempotents.
;.CHint: There exists a minimal rigbt ideal 1 S;; eRo Write eR = 1 fE) J, where
:~ .the right ideal J
1. Now, either J = {O}, or else e ;,. el + ez, with e e J
.:/aprimitive idempotent. If ez J is not primitive, repeat this proeess as applied
J
e2R.]
/:ro
14. ~)f M
=1= {O} is a right R-modul, verify that M becomes a left homR(M, M)"'!:.,rllodule on defining the module productfx by
fx
='
f(x)
b) Let M and N be right Rmodules which are R-isomorphic under the mapping
a: M -> N. Show that homR(M, M) ~ homR(N, N), as rings, by means of the
funetion that ames fe homR(M, M) toa. fo a-l.
15. Let F be a field and M 2(F) denote the ring of 2 x 2 matrices over F. Prove tht
a) The matrices el =
1 1 EB lz;
16; Use Theorem 13-14 to deduce that any commutative semisinlple (in the usual
sense) Artinian ring is a finite direct sum of fields.
17. Prove tbat a right Artinian ring R is a regular ring if and only if R is nil-semisinlple.
[Hint: Problems 19 and 20, Chapter 9.J
18. Let M
=1= {O} be a simple right R-module and so, by Theorem 13-9, a vector
space over the division ring D
homR(M, M). Prove ihe following version of
the Jacobson Density Theorem. Given any x, X2' ... , x. e M which are lineady
independent' with respect to D ald arbitrary y, Yz, ... , Y. e M, there exists some
element a E R (equivalently, some D-endomorphism Ya e Rr) such that Xka = Yk
for k = 1,2, ... , n. [Hint: From the lenlllla preeedngTheorem 13-14, it is possble
to choose elements aj e R for which
= Ofor j =1= i
xja {=1= Ofor j = i .
Let r E R be an elenlent such that xar = Yi' Now, consider a
= 2::i= ar.]
284
m",
= {O} or 12 = 1;
in th second case, deduce' that J = eR = el for some idempotent e =1= O in l.
[Hint: See the proofs ofTheorems 13-1 and 13-2; first, establish the existence of
an element a e J ror whie~.l = J.]
1,
5. Assume !hat 1. and J are/t~o minimal right ideals of the ring R which are iso1, prove that
morphie as right R-modul~8: Ir J2
a) Any R-isomorphismj)~ J is orthe fornlf(x) = ax for some a ej.
b) The produet JJ
J.. [Hint: By part (a),. J
al, where a E J.]
6. For nonzero idempotents';'u of the ring R, prove that
a) eR = uR ir and onlylf 4 = u and ue = e.
b) eR and uR are isomorphi as R-modules if and only ifthere.exist x, y E R sueh
tbat xy = u and yx = e. [Hint: Sinee xe
ux and yu "" ey, the funetion
f(er) = x{er) = u(xr) defines an R-isomorphism f: eR ..... uR, with f-I(US) =
y(us) = e(ys).]
e) eR ~ uR, as R-modules, if and only if Re ~ Ru.
7. Let eR and uR be two minimal right ideals of the ring R, where e and u are nonzero idempotents of R. Show tbat eR and uR are R-isomorphe ifand only iftheir
produet uReR =1= {O}. [Hint: Ifure =1= Ofor some r E R, define the R-isomorphisnl
f: eR ..... uR by f(es) = (ure)s.]
.
8. Let R bea nil-semisimple ring without identity. Ifthe element a E Rand aR {O},
establish that a O.
[Hint: The ideal A(R) = {r E RlrR = {O}} satisfies
A(R)2 ~ A(R)R = {O}.]
9. Establish the statements below:
a) A right ideal J of the ring R is a direct summand of R if and oruy ir 1 = eR for
some idempotent e e R ;
b) a mininlal right ideal J of the ring R is a dreet summand of R if and only ir
12 =1= {O}.
10. Prove that a right Artinian fing R is nil-semisinlple ir and only ir 12
each mnimal rigbt ideal 1 of R.
285
PROBLEMS
=1=
{O} ror
11. Assuming tbat R is a nil-semisimple right Artinian ring, veriry the following
assertions:
a) R is rigbt Noetherian; that is, R satisfies the ascending ehain condition on right
ideals. [Hint: By Theorem 13-2, the right ideals of R are finiteIy generated.]
b) The mapping e ..... eR defines a one-to-one eorrespondence between the set of
all idempotent elements e e cent R and the set of two-sided ideaIs of R.
c) For any two-sided ideal 1 of R, ann,1 = ann1 and so R = 1 EB annl.
d) Every right ideal J of R is a direct sumnand of any right ideal containing it.
12. a) Suppose that the ring R is a finite direct sum of right ideals J =1= {O}, say
R = 1 fE) J 2 EB ... EB Jn Ir 1
e + e2 + ... + e", where eel" prove
that the elements e fornl a set of orthogonaI idempotents and that 11 = eR
(i = 1, 2, .. , , n).
b) If the ideals 1 of part (a) are aH two-sided, show that e E cent R an4 so serves
as an identity element for 1.
13. a)' Prove that an idempotent e =1= O of the ring R is prinlitive if and only if R
contains no idenlpotent u =1= e sueh that eu = ue = u.
bl,'Establish that any idempotent elenlent e =1= O of a nil-semisimple right Artinian
nlng R is the sum of a finite number of orthogonal primitive idempotents.
;.CHint: There exists a minimal rigbt ideal 1 S;; eRo Write eR = 1 fE) J, where
:~ .the right ideal J
1. Now, either J = {O}, or else e ;,. el + ez, with e e J
.:/aprimitive idempotent. If ez J is not primitive, repeat this proeess as applied
J
e2R.]
/:ro
14. ~)f M
=1= {O} is a right R-modul, verify that M becomes a left homR(M, M)"'!:.,rllodule on defining the module productfx by
fx
='
f(x)
b) Let M and N be right Rmodules which are R-isomorphic under the mapping
a: M -> N. Show that homR(M, M) ~ homR(N, N), as rings, by means of the
funetion that ames fe homR(M, M) toa. fo a-l.
15. Let F be a field and M 2(F) denote the ring of 2 x 2 matrices over F. Prove tht
a) The matrices el =
1 1 EB lz;
16; Use Theorem 13-14 to deduce that any commutative semisinlple (in the usual
sense) Artinian ring is a finite direct sum of fields.
17. Prove tbat a right Artinian ring R is a regular ring if and only if R is nil-semisinlple.
[Hint: Problems 19 and 20, Chapter 9.J
18. Let M
=1= {O} be a simple right R-module and so, by Theorem 13-9, a vector
space over the division ring D
homR(M, M). Prove ihe following version of
the Jacobson Density Theorem. Given any x, X2' ... , x. e M which are lineady
independent' with respect to D ald arbitrary y, Yz, ... , Y. e M, there exists some
element a E R (equivalently, some D-endomorphism Ya e Rr) such that Xka = Yk
for k = 1,2, ... , n. [Hint: From the lenlllla preeedngTheorem 13-14, it is possble
to choose elements aj e R for which
= Ofor j =1= i
xja {=1= Ofor j = i .
Let r E R be an elenlent such that xar = Yi' Now, consider a
= 2::i= ar.]
286
19. Given a right R-module M, set M* = homR(M, R). Prove the statements below:
a) M* caD. be made into a left R-module (known as the dual module of M) by
defining the module product rf as
(rf)(x)
= rf(x)
(rER,x
APPENDIX A
M).
J1 for
for i = j
-lO
i 0/= j
ji *(Xj ) -_ bij _
RELATIONS
24. An ideal 1 of the ring R is said to be a primitive ideal if R/l is a primitive ringo
Establish that the radical J(R) can be represented as J(R) =
P, where the
intersection is taken over a1l primitive ideals P of R. [Hint: If P = A(M), where
M is a simple R-module, then M is a faithful simple module over the ring R/P.]
286
19. Given a right R-module M, set M* = homR(M, R). Prove the statements below:
a) M* caD. be made into a left R-module (known as the dual module of M) by
defining the module product rf as
(rf)(x)
= rf(x)
(rER,x
APPENDIX A
M).
J1 for
for i = j
-lO
i 0/= j
ji *(Xj ) -_ bij _
RELATIONS
24. An ideal 1 of the ring R is said to be a primitive ideal if R/l is a primitive ringo
Establish that the radical J(R) can be represented as J(R) =
P, where the
intersection is taken over a1l primitive ideals P of R. [Hint: If P = A(M), where
M is a simple R-module, then M is a faithful simple module over the ring R/P.]
288
~xample/A-1.
RELATIONS
289
as the c1ass [al] for many elements al E S.) As a notational device, let us
henceforth use the symbol SI ~ to represent the set of al1 equivalence c1asses
of the relation ~; that is,
=1=
0 ;i?,
.
;
3) for all a,
bE S, [a]
a~
br"}
,,'~.
5) U [a] = s.
:~"~'~'.'.
aeS
Proof. Clearly, the element a E [a], for a ~ a. Tci:'prove the' second
assertion, let b E [a], so that b ~ a. Now, suppse that x E [a], which
mean s 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 inc1usion [a] S;; I b]. A similar argument yields the reverse
inc1usion and equality folIows. As regards (3), first assume that [a] = [b];
then .aE [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, e E [a] n [b]. Statement (2) then informs us that [a] = [e] = [b].
In brief, if [a] n [b] =1= 0, then we must have [a] = [b]. Fin ally, since
each c1ass [a] S;; S, the inc1usion u {[a]la E S} S;; Scertainly holds. For
the cpposite inc1usion, it is enough to show that each element a E S belongs
to sorne equivalence c1ass; but this s no problem, for a E [a].
As evidenced by Example A-4, any mapping determines an eq uivalence
re1ation in its domain. The foIlowing corolIary indica tes that every
equivalence relation arises in this manner; that is to say, each eq uivalence
relation is the associated equivalen ce relation of sorne function.
Corol1ary. 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
q(a)
g(b).
288
~xample/A-1.
RELATIONS
289
as the c1ass [al] for many elements al E S.) As a notational device, let us
henceforth use the symbol SI ~ to represent the set of al1 equivalence c1asses
of the relation ~; that is,
=1=
0 ;i?,
.
;
3) for all a,
bE S, [a]
a~
br"}
,,'~.
5) U [a] = s.
:~"~'~'.'.
aeS
Proof. Clearly, the element a E [a], for a ~ a. Tci:'prove the' second
assertion, let b E [a], so that b ~ a. Now, suppse that x E [a], which
mean s 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 inc1usion [a] S;; I b]. A similar argument yields the reverse
inc1usion and equality folIows. As regards (3), first assume that [a] = [b];
then .aE [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, e E [a] n [b]. Statement (2) then informs us that [a] = [e] = [b].
In brief, if [a] n [b] =1= 0, then we must have [a] = [b]. Fin ally, since
each c1ass [a] S;; S, the inc1usion u {[a]la E S} S;; Scertainly holds. For
the cpposite inc1usion, it is enough to show that each element a E S belongs
to sorne equivalence c1ass; but this s no problem, for a E [a].
As evidenced by Example A-4, any mapping determines an eq uivalence
re1ation in its domain. The foIlowing corolIary indica tes that every
equivalence relation arises in this manner; that is to say, each eq uivalence
relation is the associated equivalen ce relation of sorne function.
Corol1ary. 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
q(a)
g(b).
290
Ex~mple -5. This example is given to illustrate that any mapping can be
wntten 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 with! If the element a E X, then we have
[a]
!neffect" the equivalence c1asses for the relation ~ are just the inverse
lmages.off-1(y), where y Ef(X) 5; y.
Now, ~efine the f~nctionl: X/~_ -+ Yby the ruleJ([a]) = f(a). Since
[a] = [b]lfand only lfj(a) = f(b),fis well-defined. Observe that whereas
the original ~unction L may no.! have been one-to-one, 1 happens to be
one-to-one; mdeed, f([a]) = f([b]) implies that f(a) = f(b), wherice
[a].= [b]. At this point, we intro~uce the ~nto function g;)( -+ X/~ by
sett!ng g(a) = [a]. Tht(nf(a) =J([a]) = f(g(a)) = (Jo g)(a) for all a in
X, m consequence of which f = f o g. This achieves our stated aim.
We next connect the notion of an equivalence relation in S with that
.
of a partition of S.
Definition -3. By a partition ofthe set S is meant a family
of S with the properties
1) 0 ~ [l},
2) for any A, B
3) u [l} = S.
E [l},
either A = B or A
11
=0
{!J
of subsets
(pairwise disjoint),
291
but not under the other; say a ~ b, but not a ~' b. By Theorem A-1,
there is an equivalence c1ass in SI ~ containing both a and b, while no such
c1ass appeats in SI ~'. Accordingly, SI ~ and SI ~., differ.
Theorem -2. If [l} is a paitition of the set S, then there is a unique
equivalence relation in S whose equivalel}ce c1asses are precisely the
members of [l}.
Proo! Given a, b E S, we write a ~ b if and only if a and b both belong to
the same subset in [l}, (The fact that [l} partitions S guarantees that each
element of S lies in exactly one member of [l}.) 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 [l} has the form SI ~. If the subset P E [l},
then a E P for .some a inS. Now, the element bE P if and only if b ~ a,
or, what amounts to the same thing, if and on!:y if b E [a]. This demonstrates
the equality P = [a] E SI ~. Since this holds for each P in [l}, it follows
that [l} 5; Sl~. On the other hand, let fa] be an arbitrary equivalence
class and P be the partition set in !?J> to \vhich the element a belongs. By
similar reasoning, we conc1ude that [a] = P; hence, S/~ 5; [l}. Thus, the
set of equivalence c1asses for ~ coincides with the partition {!J. The
uniqueness assertion is an immediate consequence of the lernma.
To surnmarize, 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 h~~ been considered from two different points of view.
Another type of'relation which occurs in various branches of mathematics is the so-called partial order relation. Just as equivalence generalizes
equality, this relation'(as we define it below) generalizes the idea of "Iess
than or equal to" on the realline.
Definition -4. 'A. relation R in a nonempty set S is called a partial
order in S if the (ollowing three conditions are satisfied;
1) aRa (reflexiveproperty),
2) if aRb and bRa, then a = b (antisymmetric property),
3) if aRb and bRe, then aRe (transitive property),
where a, b, e denote arbitrary elements of S.
Proo! If ~ and ~' are the same, then surely SI ~ = SI ~'. 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 .qne of the relations,
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 foregoing axioms then read; (1) a ::;; a, (2) if a ::;; b and b ::;; a, then a = b,
and {3) if a ::;; b and b ::;; e, then a ::;; e. As a linguistic convention, let us
also agree to say (depending on the circumstance) that "a precedes b" or
290
Ex~mple -5. This example is given to illustrate that any mapping can be
wntten 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 with! If the element a E X, then we have
[a]
!neffect" the equivalence c1asses for the relation ~ are just the inverse
lmages.off-1(y), where y Ef(X) 5; y.
Now, ~efine the f~nctionl: X/~_ -+ Yby the ruleJ([a]) = f(a). Since
[a] = [b]lfand only lfj(a) = f(b),fis well-defined. Observe that whereas
the original ~unction L may no.! have been one-to-one, 1 happens to be
one-to-one; mdeed, f([a]) = f([b]) implies that f(a) = f(b), wherice
[a].= [b]. At this point, we intro~uce the ~nto function g;)( -+ X/~ by
sett!ng g(a) = [a]. Tht(nf(a) =J([a]) = f(g(a)) = (Jo g)(a) for all a in
X, m consequence of which f = f o g. This achieves our stated aim.
We next connect the notion of an equivalence relation in S with that
.
of a partition of S.
Definition -3. By a partition ofthe set S is meant a family
of S with the properties
1) 0 ~ [l},
2) for any A, B
3) u [l} = S.
E [l},
either A = B or A
11
=0
{!J
of subsets
(pairwise disjoint),
291
but not under the other; say a ~ b, but not a ~' b. By Theorem A-1,
there is an equivalence c1ass in SI ~ containing both a and b, while no such
c1ass appeats in SI ~'. Accordingly, SI ~ and SI ~., differ.
Theorem -2. If [l} is a paitition of the set S, then there is a unique
equivalence relation in S whose equivalel}ce c1asses are precisely the
members of [l}.
Proo! Given a, b E S, we write a ~ b if and only if a and b both belong to
the same subset in [l}, (The fact that [l} partitions S guarantees that each
element of S lies in exactly one member of [l}.) 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 [l} has the form SI ~. If the subset P E [l},
then a E P for .some a inS. Now, the element bE P if and only if b ~ a,
or, what amounts to the same thing, if and on!:y if b E [a]. This demonstrates
the equality P = [a] E SI ~. Since this holds for each P in [l}, it follows
that [l} 5; Sl~. On the other hand, let fa] be an arbitrary equivalence
class and P be the partition set in !?J> to \vhich the element a belongs. By
similar reasoning, we conc1ude that [a] = P; hence, S/~ 5; [l}. Thus, the
set of equivalence c1asses for ~ coincides with the partition {!J. The
uniqueness assertion is an immediate consequence of the lernma.
To surnmarize, 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 h~~ been considered from two different points of view.
Another type of'relation which occurs in various branches of mathematics is the so-called partial order relation. Just as equivalence generalizes
equality, this relation'(as we define it below) generalizes the idea of "Iess
than or equal to" on the realline.
Definition -4. 'A. relation R in a nonempty set S is called a partial
order in S if the (ollowing three conditions are satisfied;
1) aRa (reflexiveproperty),
2) if aRb and bRa, then a = b (antisymmetric property),
3) if aRb and bRe, then aRe (transitive property),
where a, b, e denote arbitrary elements of S.
Proo! If ~ and ~' are the same, then surely SI ~ = SI ~'. 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 .qne of the relations,
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 foregoing axioms then read; (1) a ::;; a, (2) if a ::;; b and b ::;; a, then a = b,
and {3) if a ::;; b and b ::;; e, then a ::;; e. As a linguistic convention, let us
also agree to say (depending on the circumstance) that "a precedes b" or
292
RELATIONS
293
292
RELATIONS
293
294
RELATlONS
Example A-12. Consi.der 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
mnimal (maximal) element and that of a first (last) element.
295
, ~" :
294
RELATlONS
Example A-12. Consi.der 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
mnimal (maximal) element and that of a first (last) element.
295
, ~" :
ZORN'S LEMMA
APPENDIX B
297
ZORN'S LEMMA:t::
'/
~'
"
;'::\1n this Appendix, we give a brief ccount of sorne of the axioms of set theory!l"i,
,}with the primary purpose of introducing Zom's Lemma. Our presentatioQ::
l
descriptive and most of the facts are merely stated. The reader who is
not content with tbis 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, ifit does, sorne subset could very well fal to possess one. This prompts
the foIlowing definition: ti partially ordered set (S, ~) is said to be wellordered if every nonempty subset A f; S has afirst element ("with respect
to ~"being understood). The set Z + is well-ordered by the usual ~ ; each
nonempty subset has a first element, iJamely, 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} f; 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 weIl-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).
ts
fl({1,2})
1,
fl({l})
f2({1,2})
2,
f2({1}) = 1,
1,
fl({2})
2,
f2({2}) = 2.
l'
~~;
Since this general principie 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 {X} is a family
of nonempty sets indexed by the nonempty set .1, then the Cartesian product
X i.1 Xi is nonempty (it should be clcar that the e1ements of X XI are precisely the choice functions for {Xi})' For another common phrasing, which
again expresses the idea of se1ection, let ce be a collectionof disjoint, nonempty sets. The axiom of choice, as we have stated it, is equivalent to
ZORN'S LEMMA
APPENDIX B
297
ZORN'S LEMMA:t::
'/
~'
"
;'::\1n this Appendix, we give a brief ccount of sorne of the axioms of set theory!l"i,
,}with the primary purpose of introducing Zom's Lemma. Our presentatioQ::
l
descriptive and most of the facts are merely stated. The reader who is
not content with tbis 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, ifit does, sorne subset could very well fal to possess one. This prompts
the foIlowing definition: ti partially ordered set (S, ~) is said to be wellordered if every nonempty subset A f; S has afirst element ("with respect
to ~"being understood). The set Z + is well-ordered by the usual ~ ; each
nonempty subset has a first element, iJamely, 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} f; 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 weIl-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).
ts
fl({1,2})
1,
fl({l})
f2({1,2})
2,
f2({1}) = 1,
1,
fl({2})
2,
f2({2}) = 2.
l'
~~;
Since this general principie 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 {X} is a family
of nonempty sets indexed by the nonempty set .1, then the Cartesian product
X i.1 Xi is nonempty (it should be clcar that the e1ements of X XI are precisely the choice functions for {Xi})' For another common phrasing, which
again expresses the idea of se1ection, let ce be a collectionof disjoint, nonempty sets. The axiom of choice, as we have stated it, is equivalent to
298
ZORN'S LEMMA
299
298
ZORN'S LEMMA
299
BIBLIOGRAPHY
18. lACOBSON, N., SlrUClure of Rings, Rev. Ed. Providence: American Mathematica1
Society, 1964.
19. JANS, J., Rings and Homology. New York: Hol!, 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.: B1aisdell, 1966.
23. MCCARTHY, P., Algebraic Extensions of Fields. Waltham; Mass.: BlaisdeIl, 1966.
24. McCoy, N., Rings andldeals. (Carus Monographs). Menascha, Wis.: Mathematical
Association of America, 1948.'
25. McCoy, N., TheoryofRings. NewYork: Macmillan, 1964.
26. NAGATA, M., Local Rigs. NewYork: lnterscience, 1962 ..
27. NORTHCOTT, D. G., Ideal Theory. Cambridge, Eng1and: Cambridge UniversityPress,
1953.
. .
.
BIBLIOGRAPHY
,.--!'
GENERAL REFERENCES
.
Our purpose bere is to present a list Of sugg~sons for collateral reading and
further study. Tbe specialized sources wiUtarry tbe reader considerably
beyond tbe point attained in tbe final pages'f tbis work.
1. ADAMSON, T., Introduction 10 Field Theory. New York: lnterscience, 1964.
2. ARTIN, E., Galois Theory, 2nd Ed. NotreDame, lnd.: University ofNotre Dame Press,
1955.
3. AATIN, E., C. NESBITT, and R. THRALL~ Rings wilh Mnimum Condtion. Ann Aibor,
Mich,: University of Michigan Press, 1944.
4. AUSLANDER, M., Rings, Modules and Homology, Chapters l and n. Waltham, Mass.:
Department of Mathematics, Brandeis University (lecture notes), 1960.
5. BARNES, W., Introduction 10 Abstract Algebra. Boston: Heath, 1963.
6. BOURllAKI, N., Algebra, Chapter 8. Paris: Hermann, 1958.
7. BOURBAKI, N., Algebra Commutative, Chapters 2, 4 and 5. Paris: Hermann, 1961.
8. BURTON, D. ,'Introducton to M odern Abstract Algebra. Reading, Mass : Addison-Wes1ey,
1967.
.
9. CURTIS, C. and 1. REINER, Representalon Theory of Finte Groups and Associatve
Algebras. NewYork: lnterscience, 1962.
lO. DIVINSKY, N., Rings and Radicals. Toranto: University ofToronto Press, 1965.
11. GOLDIE, A., Rings with Maximum.Condition. New Haven: Department of Mathematics, Yale University.(lecture notes), 1961.
12 .. HALMOS, P., Naive Set Theory. Princeton: Van Nostrand, 1960.
13. HERSTEIN, 1. N., Topies in Algebra.. New York: Blaisdell, 1964.
14. HERSTEIN, l. N., Theory of Rings. Chicago: Department of Mathematics, University
of Chicago (lecture notes), 1961.
,15. HERSTEIN, 1. N., Noncommutative Rings. (Carus Monographs). Menascha, Wis.;
Mathematical Association of America, 1968.
16. HEWITT, E. and K. STROMBERG, Real and Abstract Analysis. New York: SpringerVerlag, 1965.
.
17. JACOBSON, N., Lectures in AbstractAlgebra, Vol. l, Basic Concepts. Princeton: Van
Nostrand, 1951.
300
301
JOURNAL ARTICLES
1;
',1
'f'
1:
34. BROWN, B. and N. McCoy, "Radicals and Subdirect Sums," Am. J. Math. 69, 46-58
(1947).
.
35. BUCK, R. c., "Extensions ofHomorphisms and Regular ldeals," J. Indian Math. Soco
14, 156-158 (1950).
36. COHEN, 1. S:, "Commutative Rings with Restricted Minimum Condition," DukeMath.
J. 17, 27-42 (1950).
37. DIYINSKY, N., "Cornmutative Subdirectly Irreducible Rings," Proc. Am. Math Soco
8, 642~648 (1957).
.
38. FELLER, E., "A Type of Quasi-FrobeniusRings," Canad. Math. Bull. lO, 19-27 (1967).
39. GIFFEN, c., "Unique Factorization of Polynomials," Proc. Am. Math. Soco 14, 366
(1963).
40. HENDERSON, D., HA Short Praof ofWedderburn's Theorem," Am. Math. Monthly 72,
385-386 (1965).
41. HERSTEIN, 1. N., "A Generalzation of a Theorem of Jacobson, l," Am. J. Math. 73,
756-762 (1951).
42. HERSTEIN, 1. N., "An Elementary Proof of a Theorem of Jacobson," Duke Math. J.
21,45-48 (1954).
43. HERSTEIN, 1. N., ~'Wedderburn's Theorem and a Theorem cif 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 ofPrime ldeals of a Ring," Fund. Math. 45, 17-27 (1957).
1:
l'
u
-
..... ....
__..
_.
--~-
-------~
- - - - - - - - - - - - _.
. ~~-.-~----------.-._._-----
BIBLIOGRAPHY
18. lACOBSON, N., SlrUClure of Rings, Rev. Ed. Providence: American Mathematica1
Society, 1964.
19. JANS, J., Rings and Homology. New York: Hol!, 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.: B1aisdell, 1966.
23. MCCARTHY, P., Algebraic Extensions of Fields. Waltham; Mass.: BlaisdeIl, 1966.
24. McCoy, N., Rings andldeals. (Carus Monographs). Menascha, Wis.: Mathematical
Association of America, 1948.'
25. McCoy, N., TheoryofRings. NewYork: Macmillan, 1964.
26. NAGATA, M., Local Rigs. NewYork: lnterscience, 1962 ..
27. NORTHCOTT, D. G., Ideal Theory. Cambridge, Eng1and: Cambridge UniversityPress,
1953.
. .
.
BIBLIOGRAPHY
,.--!'
GENERAL REFERENCES
.
Our purpose bere is to present a list Of sugg~sons for collateral reading and
further study. Tbe specialized sources wiUtarry tbe reader considerably
beyond tbe point attained in tbe final pages'f tbis work.
1. ADAMSON, T., Introduction 10 Field Theory. New York: lnterscience, 1964.
2. ARTIN, E., Galois Theory, 2nd Ed. NotreDame, lnd.: University ofNotre Dame Press,
1955.
3. AATIN, E., C. NESBITT, and R. THRALL~ Rings wilh Mnimum Condtion. Ann Aibor,
Mich,: University of Michigan Press, 1944.
4. AUSLANDER, M., Rings, Modules and Homology, Chapters l and n. Waltham, Mass.:
Department of Mathematics, Brandeis University (lecture notes), 1960.
5. BARNES, W., Introduction 10 Abstract Algebra. Boston: Heath, 1963.
6. BOURllAKI, N., Algebra, Chapter 8. Paris: Hermann, 1958.
7. BOURBAKI, N., Algebra Commutative, Chapters 2, 4 and 5. Paris: Hermann, 1961.
8. BURTON, D. ,'Introducton to M odern Abstract Algebra. Reading, Mass : Addison-Wes1ey,
1967.
.
9. CURTIS, C. and 1. REINER, Representalon Theory of Finte Groups and Associatve
Algebras. NewYork: lnterscience, 1962.
lO. DIVINSKY, N., Rings and Radicals. Toranto: University ofToronto Press, 1965.
11. GOLDIE, A., Rings with Maximum.Condition. New Haven: Department of Mathematics, Yale University.(lecture notes), 1961.
12 .. HALMOS, P., Naive Set Theory. Princeton: Van Nostrand, 1960.
13. HERSTEIN, 1. N., Topies in Algebra.. New York: Blaisdell, 1964.
14. HERSTEIN, l. N., Theory of Rings. Chicago: Department of Mathematics, University
of Chicago (lecture notes), 1961.
,15. HERSTEIN, 1. N., Noncommutative Rings. (Carus Monographs). Menascha, Wis.;
Mathematical Association of America, 1968.
16. HEWITT, E. and K. STROMBERG, Real and Abstract Analysis. New York: SpringerVerlag, 1965.
.
17. JACOBSON, N., Lectures in AbstractAlgebra, Vol. l, Basic Concepts. Princeton: Van
Nostrand, 1951.
300
301
JOURNAL ARTICLES
1;
',1
'f'
1:
34. BROWN, B. and N. McCoy, "Radicals and Subdirect Sums," Am. J. Math. 69, 46-58
(1947).
.
35. BUCK, R. c., "Extensions ofHomorphisms and Regular ldeals," J. Indian Math. Soco
14, 156-158 (1950).
36. COHEN, 1. S:, "Commutative Rings with Restricted Minimum Condition," DukeMath.
J. 17, 27-42 (1950).
37. DIYINSKY, N., "Cornmutative Subdirectly Irreducible Rings," Proc. Am. Math Soco
8, 642~648 (1957).
.
38. FELLER, E., "A Type of Quasi-FrobeniusRings," Canad. Math. Bull. lO, 19-27 (1967).
39. GIFFEN, c., "Unique Factorization of Polynomials," Proc. Am. Math. Soco 14, 366
(1963).
40. HENDERSON, D., HA Short Praof ofWedderburn's Theorem," Am. Math. Monthly 72,
385-386 (1965).
41. HERSTEIN, 1. N., "A Generalzation of a Theorem of Jacobson, l," Am. J. Math. 73,
756-762 (1951).
42. HERSTEIN, 1. N., "An Elementary Proof of a Theorem of Jacobson," Duke Math. J.
21,45-48 (1954).
43. HERSTEIN, 1. N., ~'Wedderburn's Theorem and a Theorem cif 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 ofPrime ldeals of a Ring," Fund. Math. 45, 17-27 (1957).
1:
l'
u
-
..... ....
__..
_.
--~-
-------~
- - - - - - - - - - - - _.
. ~~-.-~----------.-._._-----
1;
302
BIBLIOGRAPHY
46. KOVACS, L., ;'A Note on Regular Rings," Publ. Math. Debrecen 4,465-468 (1956).
47. LUH, J., "On the Cornrnutativity of J-Rings," Canad. J. Math. 19, 1289-1292 (1967).
48. McCov, N., "Subdirectly Irreducible Cornrnutative Rings," Duke Math. J. 12, 3lH-387
(1945).
49. McCov, N.; "Subdirect Sums ofRings," Bull. Am. Math. Soco 53, 856-877 (1947).
50. McCoy, N., HA Note on Finite Unions of Ideals and Subgroups," Proc. Am. Math.
Soco 8, 633-637 (1957).
51. NAGATA, M., "On the Theory of Radicals in a Ring," J. Math. Soco Japan 3, 330-344
(1951).
52. VON, NEUMANN, J., "On Regular Rings," Proc. Nat!. Acad. Sci. U.S. 22, 707-713 (1936).
53. NORTHCOIT, D., HA Note on the Intersection Theorern for Ideals," Proc. Cambridge
Phil. Soco 48, 366-367 (1952).
54. PERL1S, S. HA Characterization of the Radical of an Algebra," Bul!. Am. Math. Soco
48, 128-132.(1942).
.
55. SAMUEL, P., "On Unique F~ctorization Dornains," Illinois J. Math, S, 1-17 (1961).
56. SATYANARAVANA, M., "Rings with Prirnary Ideals as Maximal Ideals," Math. Scand:
20; 52-54 (1967).
51. SATYANARAYANA, M.,. "Characterization ofLocal Rings," TohokuMath.J.19,411-416
(1967).
58. SNAPPER, E., "Cornpletely Prirnary Rings, 1," Ann. Math. 52, 666-693 (1950).
59. STONE, M. H., "The Theory of Representations of Boolean AIgebras," Trans. Am.
Math. Soco 40, 37-Ill (1936).
'.
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}
[a]
a+l
(a)
aR
annS
alb, atb
a == b (mod n)
alb
aob
AAB
AxB
A(M)
C(a)
centR
char R
contf(x)
degf(x)
f(A)
(A)
F(a)
F[aJ
[F':F]
gcd (a,b)
GF(p")
hom(R,R')
303
1;
302
BIBLIOGRAPHY
46. KOVACS, L., ;'A Note on Regular Rings," Publ. Math. Debrecen 4,465-468 (1956).
47. LUH, J., "On the Cornrnutativity of J-Rings," Canad. J. Math. 19, 1289-1292 (1967).
48. McCov, N., "Subdirectly Irreducible Cornrnutative Rings," Duke Math. J. 12, 3lH-387
(1945).
49. McCov, N.; "Subdirect Sums ofRings," Bull. Am. Math. Soco 53, 856-877 (1947).
50. McCoy, N., HA Note on Finite Unions of Ideals and Subgroups," Proc. Am. Math.
Soco 8, 633-637 (1957).
51. NAGATA, M., "On the Theory of Radicals in a Ring," J. Math. Soco Japan 3, 330-344
(1951).
52. VON, NEUMANN, J., "On Regular Rings," Proc. Nat!. Acad. Sci. U.S. 22, 707-713 (1936).
53. NORTHCOIT, D., HA Note on the Intersection Theorern for Ideals," Proc. Cambridge
Phil. Soco 48, 366-367 (1952).
54. PERL1S, S. HA Characterization of the Radical of an Algebra," Bul!. Am. Math. Soco
48, 128-132.(1942).
.
55. SAMUEL, P., "On Unique F~ctorization Dornains," Illinois J. Math, S, 1-17 (1961).
56. SATYANARAVANA, M., "Rings with Prirnary Ideals as Maximal Ideals," Math. Scand:
20; 52-54 (1967).
51. SATYANARAYANA, M.,. "Characterization ofLocal Rings," TohokuMath.J.19,411-416
(1967).
58. SNAPPER, E., "Cornpletely Prirnary Rings, 1," Ann. Math. 52, 666-693 (1950).
59. STONE, M. H., "The Theory of Representations of Boolean AIgebras," Trans. Am.
Math. Soco 40, 37-Ill (1936).
'.
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}
[a]
a+l
(a)
aR
annS
alb, atb
a == b (mod n)
alb
aob
AAB
AxB
A(M)
C(a)
centR
char R
contf(x)
degf(x)
f(A)
(A)
F(a)
F[aJ
[F':F]
gcd (a,b)
GF(p")
hom(R,R')
303
304
hornR(M,M')
IJ
J+J
JrJ;JJ
(J ;J)
Ll i
.Jl
J(R)
kerf
I(M)
lcrn(a,b) ;
Mn(R)
rnap(X,R) ,
nat
ordf(x)! :'<
,",'f:.
4J(n)
4Jr
P(X)
Q
Q(.j)
Qc(R)
R#
R*
R[x]
R[x,y]
R[[x]]
R/J
rad R
Rad R
Etl R
IrJ;JR i
Z'Ze
Z+
Is
Zl
Z(i)
Zn '
Z(.j)
+",
*n
INDEX
o',:.,
'. ",'
305
304
hornR(M,M')
IJ
J+J
JrJ;JJ
(J ;J)
Ll i
.Jl
J(R)
kerf
I(M)
lcrn(a,b) ;
Mn(R)
rnap(X,R) ,
nat
ordf(x)! :'<
,",'f:.
4J(n)
4Jr
P(X)
Q
Q(.j)
Qc(R)
R#
R*
R[x]
R[x,y]
R[[x]]
R/J
rad R
Rad R
Etl R
IrJ;JR i
Z'Ze
Z+
Is
Zl
Z(i)
Zn '
Z(.j)
+",
*n
INDEX
o',:.,
'. ",'
305
lNDEX
element(s),
algebraic, 138
associa te, 91
conjugate, 105
idempotent, 14
identity, 2
; invertible, 2
irreducible, 97
nilpotent, 14
prime, 97
quasi-regular, 170
related to an ideal, 258
relatively prime, 93
transcendental:, 138
torsion, 259
zero, 1
Eisenstein irreducibility criterion, 133
endomorpbism of a module, 272
ofa ring, 25
evaluation homorphism, 26
equivalence class, 288
relation, 287
Euler pbi-function, 57
Euclidean domain, 102
valuation, 102
extension, algebraic, 140
simple, 137
extension ring, 31 ;
faithful module, 275,,;,,';
field, 52
algebraically c10sedj ;, 156
extension, 136 "
Galois, 191
,
obtained by adjoining an element,
of algebraic numbers; 155
of complex numbers, 53
of quadratic numbers, 105
of ratiolal functions, 138
skew, 52
splitting, 148
finite division ring, 194
integral domain, 56
field, 187
ring, 2
finitely generated, 19
first element, 294
137
307
306
semiprime, 80
fixed field, 69
formal fraction, 61
formal po.wer series, 112
Frobenius automorphism, 202
Fundamental Homorphism Theorem, 44
Fundamental Theorem of Algebra, 128
INDEX
J-radical, 172
J-ring, , 196
Jacobson radical,
236
157
kernel of a homomorphism, 28
lNDEX
element(s),
algebraic, 138
associa te, 91
conjugate, 105
idempotent, 14
identity, 2
; invertible, 2
irreducible, 97
nilpotent, 14
prime, 97
quasi-regular, 170
related to an ideal, 258
relatively prime, 93
transcendental:, 138
torsion, 259
zero, 1
Eisenstein irreducibility criterion, 133
endomorpbism of a module, 272
ofa ring, 25
evaluation homorphism, 26
equivalence class, 288
relation, 287
Euler pbi-function, 57
Euclidean domain, 102
valuation, 102
extension, algebraic, 140
simple, 137
extension ring, 31 ;
faithful module, 275,,;,,';
field, 52
algebraically c10sedj ;, 156
extension, 136 "
Galois, 191
,
obtained by adjoining an element,
of algebraic numbers; 155
of complex numbers, 53
of quadratic numbers, 105
of ratiolal functions, 138
skew, 52
splitting, 148
finite division ring, 194
integral domain, 56
field, 187
ring, 2
finitely generated, 19
first element, 294
137
307
306
semiprime, 80
fixed field, 69
formal fraction, 61
formal po.wer series, 112
Frobenius automorphism, 202
Fundamental Homorphism Theorem, 44
Fundamental Theorem of Algebra, 128
INDEX
J-radical, 172
J-ring, , 196
Jacobson radical,
236
157
kernel of a homomorphism, 28
~ I~
,I
INDEX
105
308
157
309
INDEX
138
88
ABCDE79876543210
~ I~
,I
INDEX
105
308
157
309
INDEX
138
88
ABCDE79876543210