Sie sind auf Seite 1von 57

Multiplicative Ideal Theory

Floris Ernst

Department of Mathematics and Statistics


University of Otago, Dunedin

27.09.2004
Contents

1 Preliminaries 3
1.1 Rings and Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Comaximality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4 Basic notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2 Quotient Rings 19
2.1 Inverting elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Minimal Primes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3 Localisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4 Fractional Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3 Extension Rings and Factorization of Ideals 32


3.1 Integral Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Dedekind Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3 Prufer Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

A Zorns Lemma 53

B The Chinese Remainder Theorem 54

C Notation 55

Bibliography 56

Background Literature 57

2
Chapter 1

Preliminaries

This chapter is intended to provide the reader with the necessary background for
understanding this project. The following will give a brief overview of the concepts
of fields, rings and ideals. Additionally, a good introduction can be found in [4].

Throughout the whole project, certain notational conventions are used. A short
summary can be found in Appendix C.

1.1 Rings and Fields


This section comprises a short reminder of the basic structure of rings and fields.

Definition 1.1 Ring

A ring (R, +, ) is a set of elements together with two operations +, called addition,
and , called multiplication. (R, +) is an additive abelian group, (R, ) is a multi-
plicative semigroup. Furthermore, the distributive laws hold.
If there is an element 1 R, 1 6= 0 (R, +), such that x 1 = 1 x = x for all x R,
R is said to be a ring with identity. If (R, ) is a commutative semigroup, the ring is
called commutative. For abbreviation purposes, a ring will be referred to as R and
a b will be denoted by ab.
A subset S of R is called a subring of R if (S, +|S , |S ) is a ring.

Examples:

(a) R is the ring of real numbers under the familiar operations.

(b) (Zp , +p , p ), p N, is the ring of integers modulo p where a +p b := (a +


b) mod p and a p b := (a b) mod p.

(c) M (n, R) is the ring of n n matrices with real entries, n 2.

3
1.1. RINGS AND FIELDS

(d) C (R, R) is the ring of continuous functions f : R R.


(e) F[X] is the ring of polynomials over a field F.

(f) 2Z is the ring of even integers.

(g) If R is a ring, R[X], the set of all polynomials over R, also is a ring.

Examples (a), (b), (d), (e), (g) are commutative rings with identity, (c) is non-com-
mutative with identity, (f) is commutative without identity.

Definition 1.2 Zero-divisors

A non-zero element r in a ring R is said to be a left zero-divisor if there exists a non-


zero element s R such that rs = 0. Right zero-divisors are defined analogously.
If an element is both a left and a right zero-divisor, it is simply called zero-divisor.
An element is called regular if it is neither a left nor right zero-divisor. A ring with
no zero-divisors is called an integral domain.
The set of all left zero-divisors of a ring R is denoted by zl (R). Similarly, zr (R)
denotes the set of all right zero-divisors of R and z(R) := zl (R) zr (R).

Examples:

Let R = M2 (Z). Then


" #
0 1
0 0

is a zero divisor since


" #" # " #" #
0 1 1 0 0 0 0 1
=0= .
0 0 0 0 0 1 0 0

Let R = C (R, R). Define two elements f, g R as follows:



R

R R

R
( (
f: 0 if x 0 and g : x if x 0
x

7 x

7
x if x > 0 0 if x > 0

Then, clearly, f g = 0 = g f , showing that both f and g are zero divisors.

Definition 1.3 Annihilator

Let R be a commutative ring, r R. Then the set

ann(r) := {s R : rs = 0}

4
CHAPTER 1. PRELIMINARIES

is called the annihilator of r. In a similar way, the annihilator of a subset S R is


defined as
\
ann(S) := ann(s).
sS

Definition 1.4 Fields

Let R be a ring with unity and no zero divisors. Then it is called a field if (R , ) :=
(R\ {0} , ) is an Abelian group.

Examples:

For any prime p, Zp is a field (see [4]). Q and R are fields.

1.2 Ideals
In the 19th century, while trying to extend the concept of unique factorisation to
rings beyond the integers, Kummer and Dedekind encountered several problems
(see the example following Theorem 3.8), leading them to examine a certain class
of subsets of rings, called ideals.

Definition 1.5 Ideals

A subset A of a ring R, A R, is called a left ideal of R if, for all x, y A, x y


lies in A and, for all r R and x A, rx lies in A. Similarly, it is called a right
ideal of R if, for all x, y A and r R, x y and xr lie in A. If A is both a left
and right ideal, A is called a two-sided ideal of R.

Note:

If R is commutative, all its ideals are two-sided.

Examples:

(a) R and 0 := {0} are always ideals of any ring R.

(b) The set of even integers, 2Z, is a two-sided ideal of Z.

(c) The set A := {f C (R, R) : f () = 0} is a two-sided ideal of C (R, R).


Note:

Although zl (R) is closed under left multiplication from R, it need not be closed
under right multiplication: consider the ring R := M (Z), the ring of all

5
1.2. IDEALS

matrices over the integers where every column and every row contains only finitely
many non-zero entries (this assures that standard matrix multiplication works).
Let

0 1 0 0

0 0 1 0
A := .

0 0 0 1
.. ..

. .

Then A is a left zero-divisor, since AB = 0 for



1 0
0 0 .

B :=
..

..
. .

Now, if zl (R) were closed under right multiplication, AAT would have to be a left
zero-divisor. But, quite obviously, AAT = 1, the identity matrix, and therefore
certainly not a zero-divisor. Since 0 = 0T = (AB)T = B T AT = BAT , AT is a right
zero-divisor and therefore this argument also shows that zr (R) is not closed under
left multiplication.

Definition 1.6 Factor Rings

Let R be a ring and let I be a two-sided ideal. Define two operations and on
the set of all cosets of I by setting

(x + I) (y + I) := (x + y) + I and (x + I) (y + I) := xy + I.

Then
  
R , , := ({x + I : x R} , , )
I

is a ring. It is called the factor ring of R with respect to I.

Example:

For any n N,

 Z
Zn = 0, 1, . . . , n 1 ' nZ.

Definition 1.7 Special ideals

Let R be a ring.

An ideal M of R is said to be a left maximal if, whenever N is a left ideal of


R such that M N R, then N = M or N = R.
If a ring R has only one maximal ideal, R is said to be quasi-local.

6
CHAPTER 1. PRELIMINARIES

A two-sided proper ideal P of R is said to be prime if, for any two elements
a, b R, aRb P implies that a P or b P .
It is called completely prime if ab P implies that a P or b P .

Example:

2Z is a maximal ideal of Z.

It turns out that the set of ideals of a ring R is closed under certain set-theoretic
operations. The following proposition gives the details.

Proposition 1.8 Binary operations on ideals

Given two left ideals I and J in a ring R. Then the following subsets of R are left
ideals:

(a) I J

(b) I + J := {i + j : i I, j J} (I + J is called the sum of I and J.)


( n )
X
(c) IJ := xi yj : n N, xi I, yj J, i, j = 1, . . . , n (IJ is called the
i=1
product of I with J.)

(d) I : J := {r R : rJ I} = {r R : r x I x J} (I : J is called the


divisional ideal of I and J.)

Proof:

(a) For x, y I J, xy I and xy J. Therefore, xy I J. Furthermore,


for r R and x I J, r x I and r x J. Hence, r x I J.

(b) I + J is an ideal since:


a1 , a2 I, b1 , b2 J : a1 + b1 (a2 + b2 ) = a1 a2 + b1 b2 I + J
| {z } | {z }
and I J

r R, x = a + b I + J, a I, b J : r x = r a + r b I + J

(c) Let
n
X m
X
x, y IJ, x = ai bi , y = cj dj
i=1 j=1

where ai , cj I, bi , dj J for i = 1, . . . , n and j = 1, . . . , m. Then

n+m
X
xy = ai bi where an+j := cj , bn+j := dj , j = 1, . . . , m
i=1

7
1.2. IDEALS

and so x y IJ. Now consider r x for some r R:

n
X n
X
rx=r ai bi = (r ai ) bi IJ
| {z }
i=1 i=1
I

(d) Let x, y I : J, i.e. x z, y z I for all z J


Then x y J, (x y) z = x z y z I and so x y I : J
Since x z J and x I x z I and thus x z z 0 I z 0 J. Hence
r x I : J. 

Note:
T
Part (a) of Proposition 1.8 can easily be extended to the intersection jJ Ij of an
arbitrary family {Ij : j J} of left ideals.
By induction, parts (b) and (c) can easily be extended to any finite family of left
ideals {I1 , . . . , In } to give their sum I1 + + In and product I1 I2 . . . In .
The product of two ideals introduced in part (c) can be extended to define powers
of ideals by setting, for any ideal M and any n N,

M n := M
| M {z
. . . M} .
n terms

Now the question arises how the ideals of a given ring can be characterised. First,
we look at a certain family of ideals, the so-called principal ideals. Then this con-
cept is extended to finitely generated ideals:

Definition 1.9 Principal and finitely generated ideals

Let R be a ring.

A left ideal I in R is called principal if there is an element x R such that


I = Rx := hxi := {rx : r R}. A commutative ring is called a principal ideal
domain, or PID, if it contains no zero-divisors and all its ideals are principal.

Let =
6 S = {si : i I} R. We say a left ideal A is generated by S if every
element a A can be expressed as

X
a= ri si , ri R,
iI

such that almost all ri = 0. We write I = hSi.


If I is finite, say I = {1, . . . , n} for some n N, we say that A is finitely
generated and we write A =: hs1 , . . . , sn i. It is straightforward to show that
Pn
in this case A is the sum i=1 hsi i of the principal left ideals hsi i, i = 1, . . . , n.

8
CHAPTER 1. PRELIMINARIES

Example:

Let R = Z. Then h2i = 2Z and h3i = 3Z are both principal ideals as is their sum
2Z + 3Z = 1Z = Z (since 1 = 1 2 + 1 3 h2, 3i).
Their intersection is the principal ideal h6i (since x h6i is equivalent to 2|x and
3|x, e.g. x h2i h3i).

Although the intersection of two ideals is again an ideal, we now show that the
intersection of two finitely generated ideals is not necessarily finitely generated.
Consider the ring R of polynomials over Q, R := Q[X], and the subring D of R,

D := Z + X 2 R = a0 + a2 x2 + + an xn : a0 Z, ai Q for i 2, n 0 .



The two ideals I := X 2 and J := X 3 of D are obviously finitely generated.
Now consider an element F A := I J. Then, for some a0 Z, a2 , . . . , an Q
and b0 Z, b2 , . . . , bm Q, we see:

F = X 2 a0 + a2 X 2 + + an X n = X 3 b0 + b2 X 2 + + bn X m
 

a0 X 2 + a2 X 4 + a3 X 5 + an X n+2 = b0 X 3 + b2 X 5 + + bn X m+3
a0 = a2 = b0 = 0 F X 5 R A X 5 R

Since obviously X 5 R A, we see that A = X 5 R. We will now show that A is


not finitely generated as an ideal of D. Assume it were finitely generated, by the n
elements F1 , . . . , Fn , say, where

Fi = a5i X 5 + a6i X 6 + , aki Q, k N, i = 1, . . . , n.

c5i
Say a5i = d5i where c5i , d5i Z, d5i 6= 0 and the fraction is in lowest form. Now
consider the element
n
!
1 Y
G := X 5 A where d := d5i + 1.
d i=1

Since for every element H hF1 , . . . , Fn i, H = h5 X 5 + , we know that h5


Qn
i=1 d5i Z, we see that G
/ hF1 , . . . , Fn i. This shows that A cannot be finitely
generated.

This bad example leads us to define a better class of rings, which includes all
PIDs, namely the coherent rings:

Definition 1.10 Coherent Rings

A commutative ring R is called coherent if

(a) the intersection of two finitely generated ideals is finitely generated and

(b) the annihilator of any element of R is finitely generated.

9
1.2. IDEALS

Note:

Similarly, the divisional ideal of two finitely generated ideals need not be finitely
generated. An example will be given in section 3.3.

We now can easily prove the following results.

Proposition 1.11 Basic results

Let R be a ring. Then:

(a) A proper two-sided ideal P of R is prime whenever I and J are left


ideals of R such that IJ P , either I P or J P .

(b) If R is commutative, an ideal P is prime whenever ab P , either a P


or b P . (E.g. if R is commutative, prime completely prime.)

(c) R is an integral domain the zero ideal is prime.

(d) If a, b R, then Ra Rb a = rb for some r R.

(e) Every two-sided maximal ideal is prime.

(f) A commutative ring R is a field the zero ideal is maximal.

(g) If R is commutative, an ideal P is prime R/P is an integral domain.

(h) Let A be a left ideal and let (Bi )i=1,...,n be a family of left ideals of R. Then
n
X n
\
A: Bi = (A : Bi ) .
i=1 i=1

(i) If I1 , . . . , Is R are ideals and P R is a prime ideal such that P I1 Is ,


then there exists k {1, . . . , s} such that Ik P .

Proof:

(a) Suppose P is prime and I, J are ideals such that IJ P but I 6 P


and J 6 P . Then x I, y J such that x 6 P, y 6 P but RxR RyR =
RxRyR RP R = P . Contradiction!

Suppose P is not prime. Then there are a, b R such that aRb P but
a 6 P and b 6 P . Let I = RaR and J = RbR. Then I, J are left ideals of R
with IJ = RaRbR RP R = P but I 6 P and J 6 P . Contradiction!

(b) Let P be prime, let a, b R such that ab P . Then abr = arb r R and
hence aRb P . Therefore, either a P or b P .
Conversely, let a, b R such that aRb P . Then ab = a 1 b P and
therefore either a P or b P .

10
CHAPTER 1. PRELIMINARIES

(c) R is an integral domain if and only if ab = 0 a = 0 or b = 0. This is


equivalent to the zero ideal being prime.

(d) Since Ra Rb, a Rb. Hence, r R such that a = r b.

If a Rb then for all r R, r a Rb. Hence, Ra Rb.

(e) Let M be maximal among the proper two-sided ideals. Assume M is not
prime. Then there are two-sided ideals A, B such that AB M and A 6 M ,
B 6 M . Thus A + M = R = B + M (A + M )(B + M ) = R but also
(A + M )(B + M ) = AB + AM + M B + M M ( R. Contradiction!
Now assume M is maximal among either the left or right ideals. Then M is
also maximal among the two-sided ideals and by above therefore is prime.

(f) Let 0 6= a R. Since 0 is maximal, Ra = R. Then b R such that


b a = a b = 1. Hence, a is invertible and therefore R is a field.

Let 0 6= M be an ideal in R. Then 0 6= a M . Since R is a field,


b R such that 1 = b a. Hence, 1 M and thus M = R.

(g) Let a + P, b + P R/P such that ab + P = 0 + P . Therefore, ab P .


Since P is prime, at least one of a, b P , say a P . Then a + P = 0 + P .
Therefore R/P is an integral domain.

Let a, b R such that ab P . Then (a + P )(b + P ) = ab + P = 0 + P .


Therefore, at least one of a + P, b + P are zero, say a + P = 0. This shows
that a P . Therefore, P is prime.
Pn Pn
(h) Since A : i=1 Bi = {r R : r ( i=1 Bi ) A}, we see that for all i = 1, . . . , n,
Pn Pn
A : i=1 Bi {r R : rBi A} = A : Bi . Therefore, A : i=1 Bi
Tn
i=1 (A : Bi ) .
Tn
Conversely, let r i=1 (A :: Bi ). Then for all i = 1, . . . , n, rBi A.
Pn Pn
This shows that i=1 rBi A and so s i=1 Bi A. This shows that
Pn
r A : i=1 Bi .

(i) Assume this is not true. Then there exist elements ak Ik \P , k = 1, . . . , s.


Then their product a1 as I1 Is P . Since P is prime, this shows that
at least one ak P . Contradiction! 

Note:

Two-sidedness in part (e) of Proposition 1.11 is necessary, not only because of the
definition of prime ideals: consider the ring R := M2 (Z) and the left ideal
" #
Z 2Z
I= .
Z 2Z
I is maximal:

11
1.2. IDEALS

Let J ) I be a left ideal in R. Let M J\I. Then


" #
a c
M=
b d
where at least one of c, d has to be odd.

Assume c odd, d even. Then


" # " # " #
a c+1 a c 0 1
= J.
b d b d 0 0

Since
" # " #" # " #
0 0 0 0 0 1 0 0
R, we see that = J.
1 0 1 0 0 0 0 1

Assume c even, d odd. Then


" # " # " #
a c a c 0 0
= J.
b d+1 b d 0 1

Assume both c and d are odd. Then


" # " # " #
a c+1 a c 0 1
= J.
b d+1 b d 0 1

Since
" # " #" # " #
0 0 0 0 0 1 0 0
R, we see that = J.
1 0 1 0 0 1 0 1

Therefore, in each case,


" # " #
1 0 0 0
1= + J,
0 0 0 1
and therefore J = R. This shows that I is maximal.

I is not two-sided:

Consider the following:


" # " # " #
1 0 1 1 1 1
=
/ I.
1 0 0 0 1 1
Hence, I is clearly not prime and not even contained in a prime ideal.

Note:

The converse of (e) in Proposition 1.8 is not true: consider the ring R := Z[X] and
the ideal hXi of R. hXi is prime since

Z[X]
hXi = {z + hXi : z Z[X]} = {z + hXi : z Z} ' Z

12
CHAPTER 1. PRELIMINARIES

is an integral domain. hXi is clearly not maximal since hXi ( h2, Xi ( Z[X].

Definition 1.12 Noetherian Rings

A ring R is called left (right) Noetherian if every left (right) ideal I in R is finitely
generated.

Note:

It is clear that all Noetherian rings are coherent.

Examples:

(a) Let R be the ring of functions from R to R and I be the set of those functions
f : R R such that there is nf N (depending on f!) for which f (x) = 0 if
x 6 [nf , nf ]. I is an ideal but it is not finitely generated:
Suppose there were a set of generators f1 , . . . , fk with associated interval
bounds n1 , . . . , nk . Then if ri R, ri fi is associated with ni . This shows
Pk
that i=1 ri fi corresponds to the interval bound n0 := max {n1 , . . . , nk }. Let
g R be non-zero on [n0 1, n0 + 1]. Then g 6 hf1 , . . . , fk i.

(b) Since the example following Definition 1.9 shows that D = Z + X 5 Q[X] is not
coherent, it cannot be Noetherian.

We now introduce a concept which will provide another way of characterising


Noetherian Domains.

Definition 1.13 Ascending Chain Condition

A collection C of ideals, C = {Is : s S}, is called a chain (with respect to inclu-


sion) if, given any s1 , s2 S, we have either Is1 Is2 or Is2 Is1 .

A ring R is said to satisfy the Ascending Chain Condition (ACC), if, given any
chain C of ideals, indexed by N, C = {In : n N}, where In In+1 n, there
exists k N such that Ik = Ik+i for all i 1. In this case we say that the chain
becomes stationary.

Note:

C
S
If = {Is : s S} is a chain, sS Is is an ideal.

We now show that there is a connection between the ACC and a ring being Noethe-
rian:

13
1.2. IDEALS

Theorem 1.14 Characterisation of Noetherian Rings

A ring R is left Noetherian if and only if R satisfies the ACC.

Proof:

Suppose that R is Noetherian and let (In )nN be an ascending chain of ideals in R.
S
Let I := nN In . Then I is an ideal of R. Furthermore, since R is Noetherian,
I is finitely generated, say by x1 , . . . , xk where xi Ini . Let n0 := max {ni }i .
Then x1 , . . . , xk In0 .
Thus Rx1 + Rx2 + . . . + Rxk In0 and so I In0 . Hence I = In0 (since In0
I ) and so R satisfies the ACC (the chain becomes stationary at In0 or earlier).

Suppose that R is not Noetherian. Then there is an ideal I of R which is not finitely
generated. Pick x1 I. Then I1 := Rx1 6= I. So there exists x2 I\I1 . Then
I2 := I1 + Rx2 6= I. So there exists x3 I\I2 . Continue in this way to find a
strictly ascending chain I1 ( I2 ( . This is a contradiction since R was required
to satisfy the ACC. 

Example:

It is a standard result in a first course in modern algebra that every ideal of Z is a


principal ideal. Hence Z is a Noetherian domain.

Note:

There are one-sided Noetherian rings as the following example shows.

Example:

Consider the ring


" # (" # )
Z Q z p
R := = : z Z, p, q Q (where the blank entry denotes 0).
Q q

Let L be a left ideal of R. Then


" # (" # )
S U s u
L= = : s S, u U, t T ,
T t

where S Z, U, T Q. It is obvious that S (Z, +), U, T (Q, +). Now let


" #
n q1
R.
q2

14
CHAPTER 1. PRELIMINARIES

Then
" #" # " # " #
n q1 S U nS nU + q1 T S U
= .
q2 T q2 T T

This shows that S must be an ideal of Z and that T must be an ideal of Q. This
leaves two cases, T = 0 and T = Q. We only look at the first case, which yields
" #
S U
L= .
0

Now consider the left ideal


" #
0 Q
L := .
0

If L were finitely generated as a left ideal of R, there would have to be elements


q1 , . . . , qn Q for some n N such that for all q Q there exist r1 , . . . , rn R
such that
" # " # " #
0 q 0 q1 0 qn
= r1 + + rn .
0 0 0

Say
" #
ni ki
ri =
li

so that
" # " #
0 qi 0 ni q i
ri = .
0 0

Since the Abelian group Q is not finitely generated (see [2]), L cannot be finitely
generated.
This shows that R is not left Noetherian.

Now let D be a right ideal of R, say


" #
S U
D= .
T

Again, for all n Z, q1 , q2 Q, we see


" #" # " # " #
S U n q1 nS q1 S + q2 U S U
= ,
T q2 q2 T T

yielding that S is an ideal of Z and that T is an ideal of Q. Consider these two


cases:

(a) S = 0: Then
*" #+
0 s
Ds/t =
t

15
1.3. COMAXIMALITY

where s/t Q. Quite obviously, for two different fractions s1 /t1 and s2 /t2
(both in lowest form), neither
*" #+ *" #+ *" #+ *" #+
0 s1 0 s2 0 s2 0 s1
nor
t1 t2 t2 t1

holds.

(b) S 6= 0: Then
" # " #
S Q S Q
D= or D = .
0 Q

Together, this yields the lattice of ideals shown in Figure 1.1.


N
  p p
0Q
Q r
@

J N

 J@
p p
   

 J@
00 r r r p p p Jr @r 0 Q

 J @
Q 0
@ JJ BB
@

@J B

@JB

r
JB

@
0
N
Figure 1.1: The lattice of ideals in R. denotes a lattice of ideals isomorphic to
the lattice of ideals of Z. The non-labelled s represent the Ds/t s.

Since Z is a principal ideal domain, it is Noetherian and therefore its lattice satisfies
the ACC (by Theorem 1.14). This shows that the lattice in Figure 1.1 also satisfies
the ACC. Therefore, R is right Noetherian.

1.3 Comaximality
In this section, we assume all rings to be commutative.

Definition 1.15 Comaximality

Two ideals I and J of a ring R are called comaximal if I + J = R. Similarly, a


family (I ) of ideals of a ring R is said to be pairwise comaximal if I + I =
R , ( 6= ).

Example:

If p and q are distinct primes, the principal ideals hpi and hqi of Z are comaximal
(since gcd(p, q) = 1, there are x, y Z such that px + qy = 1, e.g. 1 hpi + hqi).

16
CHAPTER 1. PRELIMINARIES

More generally, if P is the set of all primes, the family (hpi)pP is pairwise comax-
imal.

Proposition 1.16 Properties of comaximal ideals

(a) Let I and J be comaximal ideals. Then IJ = I J.

(b) If a ring R has more than one maximal ideal, the set M (R) of all maximal
ideals of R is pairwise comaximal.

(c) Let M1 , M2 be comaximal ideals of a ring R. For any k, l N, M1k and M2l
are comaximal.

(d) Let M1 , M2 , M3 be pairwise comaximal. Then the product M1 M2 and M3


are comaximal.

Proof:

(a) We always have IJ I J. Let x I J. Since I +J = R, there are elements


a I, b J such that a + b = 1. Therefore, x = x 1 = xa + xb IJ.

(b) If Mi and Mj are two distinct maximal ideals of R, Mi +Mj properly contains
Mi . Hence, Mi + Mj = R since Mi is maximal.

(c) There are elements a1 M1 and a2 M2 such that 1 = a1 + a2 . Then:

k+l  
k+l
X k+l
1k+l
= (a1 + a2 ) = ak+li
1 ai2
i=0
i

 
k+l
Since a1k+li M1k for i l, so is i ak+li
1 . When i < l, then ai2 M2l
 
and therefore k+li ai2 is.

Hence 1 M1k + M2l and so M1k + M2l = R.

(d) For some m1 M1 , m2 M2 , m3 , m3 M3 : 1 = m1 + m3 = m2 + m3 . Then:

1 = 1 1 = (m1 + m3 )(m2 + m3 ) = m1 m2 + m3 m2 + m3 m3 + m1 m3 =
= m1 m2 + (m1 + m3 )m3 + m3 m2 M1 M2 + M3
| {z } | {z }
M1 M2 M3

Thus R = M1 M2 + M3 . 

17
1.4. BASIC NOTIONS

1.4 Basic notions


Throughout the entire project, we will refer to a ring with the letter R. If not
stated otherwise, we will consider R to be not necessarily commutative and contain
a unity 1. Likewise, all ideals (usually denoted by I and J) will be assumed to be
two-sided unless stated otherwise. Quite often, though, certain proofs or definitions
are given for left or right ideals only. Most of the time the adaptation to the other
case is straightforward as is the two-sided version. Should this not be the case, it
will be clearly stated.
Further notational conventions are listed in Appendix C.

18
Chapter 2

Quotient Rings

2.1 Inverting elements


In a ring R, for any given element r R, there need not be an inverse s R such
that rs = 1. The question arises as to how the ring can be extended in a way such
that there is an inverse for every regular element.
Note that for a zero-divisor r there cannot be an inverse: assume that rs = 0, s 6= 0
and r0 is an inverse for r. Then:

0 = r0 0 = r0 rs = 1 s = s, a contradiction.

To be able to further investigate this, we require some definitions.

Definition 2.1 Multiplicatively closed set

Let R be a commutative ring. A subset S of R is called multiplicatively closed (a


multiplicatively closed set, or MCS ) if

(a) 1 S,

(b) 0
/ S,

(c) for all s, t S, we get st S.

Now, a multiplicatively closed set gives reason for the definition of the following
equivalence relation :

Proposition 2.2

Let R be a commutative ring and S R be an MCS. Then is an equivalence


relation on Q := R S by defining

(r1 , s1 ) (r2 , s2 ) : t S : t(r1 s2 r2 s1 ) = 0 for (r1 , s1 ), (r2 , s2 ) Q.

19
2.1. INVERTING ELEMENTS

Proof:

We need to show that is reflexive, transitive and symmetric. Let (r1 , s1 ), (r2 , s2 )
Q. Then:

(r1 , s1 ) (r1 , s1 ) by selecting t = 1.

If (r1 , s1 ) (r2 , s2 ) using t0 , then (r2 , s2 ) (r1 , s1 ), again using t0 (note that
this requires the ring R to be commutative).

Suppose (r1 , s1 ) (r2 , s2 ) and (r2 , s2 ) (r3 , s3 ) with

(a) t(r1 s2 ) = t(r2 s1 )

(b) u(r2 s3 ) = u(r3 s2 )

where u, t S. Then:

(a) (b)
uts2 (r1 s3 ) = us3 (tr1 s2 ) = us3 tr2 s1 = (ur2 s3 )ts1 = ur3 s2 ts1 =
= uts2 (r3 s1 )

and so

(r1 , s1 ) (r3 , s3 ) with t := uts2 .

This completes the proof. 

Note:

If R is an integral domain, t can always be cancelled and may hence be omitted


in the definition of .

Note that S may contain zero-divisors, but if s S is a zero-divisor, e.g.


rs = 0 for some r R, r 6= 0, r will not be in S.

Definition 2.3 Quotient Rings

For any commutative ring R and an MCS S R we define (r, s) to be the equiva-
lence class of (r, s) in Q, i.e.

(r, s) := {(x, y) Q : (r, s) (x, y)} .

Then we define
 n o
Q
RS := := (r, s) : (r, s) Q ,

20
CHAPTER 2. QUOTIENT RINGS

the set of all equivalence classes of in Q. The class (r, s) can also be denoted by
r
. Then it can easily be checked that (RS , +, ) is a commutative ring if we define
s
r1 r2 r1 s2 + r2 s1
+ := and
s1 s2 s1 s2
r1 r2 r1 r2
:= .
s1 s2 s1 s2
The ring RS is called the localisation of R at S.
If S is the set of all regular elements of R, RS is called the total quotient ring or
classical ring of quotients of R and is denoted by T .

Example:

Q is the total quotient ring of Z.

Note:

Obviously, if P is a prime ideal of R, R\P is an MCS. For abbreviation purposes,


it is common to denote RR\P by RP .

Although technically R is not contained in RS , R can be embedded into RS by


means of a homomorphism:

Definition 2.4 Embedding

Let R be a ring and S be an MCS. Define the following map:


(
R RS
: .
r 7 (r, 1)

Then is called the embedding homomorphism from R into RS and (R) is called
the embedding of R into RS .

Note:

Quite clearly, is a ring homomorphism: let r, s R. Then (r + s) =


(r + s, 1) = (r, 1) + (s, 1) = (r) + (s) and (rs) = (rs, 1) = (r, 1) (s, 1) =
(r) (s).

Note that this mapping is only injective if S contains no zero-divisors: let t S


be a zero-divisor, e.g. 0 6= r R such that tr = 0. Then (r + 1) = (1)
since t(r + 1 1) = tr = 0, although r + 1 6= 1 (since otherwise r = 0).
If S contains no zero-divisors, though, we see that is injective: let r, s R
such that (r) = (s). Then t S such that t(r s) = 0. Assume
r s 6= 0. Then, since 0
/ S and therefore t 6= 0, t and r s must be
zero-divisors. Contradiction!

21
2.2. MINIMAL PRIMES

Quite often we will not mention the embedding homomorphism but just
identify the elements r and (r, 1) to be able to define multiplication of elements
from R and RS as well as to simplify notation when dealing with fractional
ideals in section 2.4.

Theorem 2.5 Quotient Rings

Let R be a commutative integral domain and S R an MCS. Then RS is a field if


S is the set of all regular elements of R. RS is then called the total field of quotients
of R.

Proof:

1 r 1 r
Let 0 6= r R. Then r S and therefore r RS and 1 r = r = 11 , hence 1
r is the
inverse of r in RS . Therefore, RS is a field. 

2.2 Minimal Primes


The following two results are required for a generalisation of the concept of quotient
rings.

Definition 2.6 Minimal Primes

Let I be a proper left ideal of a ring R. A prime ideal P of R is said to be minimal


over I (or a minimal prime of I) if

(a) I P and

(b) if Q is a prime ideal of R such that I Q P , then Q = P .

We denote the set of all minimal primes of I by min(I).

Example:

In Z, hpi is a minimal prime over hni for all n Z such that p|n.

Proposition 2.7 Existence of maximal super-ideals

Let L be a proper left ideal of some ring R. Then there exists a maximal left ideal
M of R such that L M .

Proof:

We use Zorns Lemma (see Appendix A). Let P be the set of all proper left ideals

22
CHAPTER 2. QUOTIENT RINGS

of R which contain L.P 6= since L L and hence L P . Let be a partial or-


dering on P given by . Let C be a chain in (P , ), say C = {C : }. Then
C := C . Because C is a chain, C is a left ideal of R, clearly containing
S

L. Since 1 6 C for all , 1 cannot be in C . Hence, C is a proper ideal and


therefore an upper bound for C in P. This shows we can apply Zorns Lemma,
yielding the required left maximal ideal M . 

Proposition 2.8 Existence of min(I)

For each proper ideal I ( R of some commutative ring R, min(I) 6= .

Proof:

Let P be the set of all prime ideals of R which contain I. Since R is commutative,
every maximal ideal of R is prime (see Proposition 1.11) and hence P 6= (by
Proposition 2.7).
Define a partial ordering on P
by (reverse inclusion). Let C be a chain in
(P , ), say C
T
= {C : }. Let C := C I. Then C is an ideal of
R. Moreover, C C for all .
Suppose a, b R with ab C . Then , ab C . So, since C is prime, we
have either a C or b C . Suppose that a 6 C . Given , we have either
C C or C C . For C C , b C . Then, for C C we have a 6 C .
Hence, b C for all C C . Therefore, b C . Hence C is prime and so is
an upper bound for C. Thus we can apply Zorns Lemma. This gives a minimal
prime over I. 

Example:
D E
In Z, min(h6i) = {h2i , h3i}. In general, if I = pk11 pk22 . . . pknn for some pair-
wise distinct primes pi and integers ki , i = 1, . . . , n, then min(I) = {hp1 i , . . . , hpn i}.

Note:

If R is not commutative, min(I) may be empty! (see the note following Proposi-
tion 1.11)

2.3 Localisation
Now we will generalise the concept of a quotient ring. In this section we shall as-
sume every ring to be commutative.

Definition 2.9 Localisation

23
2.3. LOCALISATION

Let I be an ideal in a ring R and let S R be an MCS. Let be defined as in


Definition 2.3. Then the set

IS := I

is called the localisation of I with respect to S.

Definition 2.10 Extension and Contraction

Let A be an ideal in a ring R and let S R be an MCS. Then the localisation AS


of A is also called the extension of A with respect to S (or to RS ), denoted by Ae .
Vice versa, if B is an ideal of RS , the set of elements
 
c b
B := b R : B
1
is called the contraction of B.

Quite obviously, the contraction B c of an ideal B is an ideal itself. Again, if P is a


prime ideal we will denote IR\P by IP .

We will now state some straightforward properties of localisation:

Proposition 2.11 Properties of Localisation

Let A, B be ideals of a ring R, S R be an MCS and I be an ideal of RS . Then:

(a) AS BS = (AB)S

(b) AS BS = (A B)S

(c) AS + BS = (A + B)S

(d) If B is finitely generated, (A : B)S = AS : BS

(e) I = I ce

Proof:
(a)

a b
Each element in AS BS is a sum of products of the form s , t where a A,
b B, s, t S, hence it is a sum of elements of the form ab
st , which are in
(AB)S .

x
Suppose s (AB)S , x AB, s S. Hence, x is a sum of products of the
x ab
form ab, a A, b B. So s is a sum of products of the form s where a A,
ab ab
b B, s S, each of which is in AS BS , since s = s 1.

24
CHAPTER 2. QUOTIENT RINGS

(b)

d d d
Let s (A B)S , d A B, s S. Then s AS and s BS . Therefore,
d
s AS BS .

d d a b
Let s AS BS . Then s = u = t where a A, b B, u, t S. Then there
exist s1 , s2 S such that

s1 du = s1 as and s2 dt = s2 bs.

Then
s ass2 t

1
AS
d ds1 s2 at ss1 s2 ut d
= = and so AS BS .
s ss1 s2 ut s2 bss1 u s
BS
ss1 s2 ut
(c) This is obvious.

(d) The containment (A : B)S AS : BS always holds, even if B is not finitely


generated:
r b
Let s1 (A : B)S and let s2 BS . Then rB A and we note:

r b rb
= AS since rb A.
s1 s2 s1 s2
r
This shows that s1 AS : BS , establishing (A : B)S AS : BS .
We now prove the reverse inclusion. Assume that B = hbi is a principal ideal.
r r
Then if x AS : BS , x = s for some r R, s S. Furthermore, 1 AS : BS
rb c
as well. Therefore, 1 AS and so rb (AS ) . Assume that rb
/ A.
Then there exist a0 A, s0 S such that (rb, 1) (a0 , s0 ). This yields
rbs0 t = a0 t A for some t S. Therefore, there always exists an element
s1 S (possibly s1 = 1) such that rbs1 A. Consequently, rs1 A : hbi and
rs1
x= ss1 (A : B)S . Now, if B = hb1 , . . . , bk i is a finitely generated ideal of R,
we see:

k
! k
! k
X (c) X () \
AS : BS = AS : hbi i = AS : hbi iS = (AS : hbi iS ) =
i=1 S i=1 i=1
k k
!
\ (b) \
= (A : hbi i)S = A : hbi i = (A : B)S ,
i=1 i=1 S

completing the proof (note that () holds because of Proposition 1.11, part
(h)).
r
(e) It is clear that I ce I. Now consider x I, x = s for some r R, s S.
r c r
Then 1 = xs I and so r I and consequently x = s I ce .
Therefore, I ce I. 

25
2.4. FRACTIONAL IDEALS

We will now proceed to the most important result of localisation:

Theorem 2.12 The Localisation Principle

Let A, B be ideals of a ring R. If AP = BP for all maximal ideals P of R, then


A = B.

Proof:

a
Let a A. Then 1 BP for all maximal ideals P of R. Therefore, for each such
ideal P there exists an element cP R\P such that

cP a B.

Now consider the ideal C generated by

{cP : P maximal ideal of R} .

Assume that C 6= R. Then, by Proposition 2.7, there exists a maximal ideal M


such that C M . By definition of C this would require an element cM R\M to
be in C M , leading to a contradiction. Hence, C = R. Therefore, 1 C and
thus there are maximal ideals P1 , . . . , Pk for some k N such that

1 = x1 cP1 + + xk cPk

for some x1 , . . . , xk R. This yields

a = ax1 cP1 + + axk cPk B,

since each xi cPi B (by definition of the cP s). Hence A B. B A is shown


similarly. 

2.4 Fractional Ideals


Again, we assume every ring to be commutative. To be able to define fractional
ideals, we require some more algebraic background:

Definition 2.13 R-Submodule

Let R be a ring and let T be its total ring of quotients. An R-submodule of T is a


subset =
6 J T such that (J, +) (T, +) and rx J r R, x J.

Definition 2.14 Finitely Generated R-Module

Let M be an R-module over some ring R. Then a finite set of elements m1 , . . . , mt

26
CHAPTER 2. QUOTIENT RINGS

M for some t N is said to generate M if m M r1 , . . . , rt R such that


Pt
m = i=1 ri mi .
If such a set exists, M is said to be finitely generated.

r
We can think of R being contained in T by identifying r R with 1 (the embedding
homomorphism from Definition 2.4). Then R is a subring of T . Note that each
ideal of R is an R-submodule of R. Hence, if an R-submodule J of T is a subset of
R, J is an ideal of R.

Definition 2.15 Fractional Ideals

An R-submodule J of T is called a fractional ideal of R if there exists a regular


element d R such that dJ R.
If I is an ideal of R, it is also called an integral ideal.

The sum, product and intersection of two and hence any finite number of fractional
ideals are defined equivalently to the respective operations on integral ideals (see
Proposition 1.8). Only the divisional ideal of two fractional ideals is defined (some-
what) differently:

Definition 2.16 Division of fractional ideals

Let R be a commutative ring and let I and J be two fractional ideals of R with
respect to its total ring of quotients T . Then:

[I : J]T := {t T : tJ I} = {t T : t x I x J}

Whenever the context is clear we will write I : J instead of [I : J]T .

Proposition 2.17 Division of fractional ideals

Let R be a commutative ring and let I and J be two fractional ideals of R. Then
[I : J]T is a fractional ideal if there exists a regular element b J.

Proof:

It is obvious that [I : J]T is an R-module. Since I is a fractional ideal, there exists


a regular element d such that dA R. Since b J, it is clear that b [I : J]T A.
Since b is regular, db is regular and therefore db [I : J]T R. This shows that [I : J]T
is a fractional ideal of R. 

Examples:
Pn 1
For n N, non-zero ai Z, I := i=1 an Z is a fractional ideal of Z, since
dI Z taking d := a1 . . . an . The total quotient ring of Z is Q.

27
2.4. FRACTIONAL IDEALS

a

A := 2n , a Z, n N0 is not a fractional ideal of Z.

Any integral ideal I of R is a fractional ideal of R (since dI R with d = 1).

Note:

The set F (R) of all fractional ideals of a ring R is a commutative monoid under
multiplication, with the ideal R as its identity.

We will now inquire which elements are the units in F (R).

Definition 2.18 Invertible Ideals

If I is a fractional ideal of a ring R and there is a fractional ideal J of R such that


IJ = R, then I is called invertible and J is said to be its inverse.

Proposition 2.19 Uniqueness of inverses

If the inverse J of a fractional ideal I exists, it is unique and J = [R : I]T

Proof:

Uniqueness:

Assuming there were a second inverse, K, we see:

J = JR = JIK = RK = K

Form of J:

Let J := [R : I]T = {t T : tI R}. Since JI = R, we have J J . Also,


J = J R = J IJ RJ = J. Therefore, J = J . 

Note:

Suppose I is a fractional ideal of R and d is a non-zero element of R such that


dI R. Then dI is an integral ideal of R since

(a) if x, y dI then x = dx, y = dy for some x, y I and so x y = dx dy =


d (x y) dI.

(b) if x dI, r R, then x = dx for some x I and so rx = d (rx) dI.

Examples:

Let R := Z[X].

28
CHAPTER 2. QUOTIENT RINGS

1
1 1
1
(a) h2i and hXi are invertible where h2i = 2 and hXi = X :
Since
  ( n )
1 X 1
hXi = Xfi gi : fi , gi R for i = 1, . . . , n, n 0 R
X i=1
X

1

1
1
and 1 = X X hXi , we see that hXi = R. A similar argument

X1 X
proves that indeed h2i 2 = R.

(b) h2, Xi is not invertible:


Assume it were. By Proposition 2.19, its inverse must be [R : h2, Xi]T . Then
we see:

[R : h2, Xi]T = {t T : t h2, Xi R} = {t T : t(2f + Xg) R f, g R}

This yields (by (a)) that


   
1 1
[R : h2, Xi]T [R : hXi]T = and [R : h2, Xi]T [R : h2i]T = .
X 2

1
1 1 1
Let t 2 X . Then there are f, g R such that 2f = t = X g.
Multiplying this by 2X yields Xf = 2Xt = 2g. This shows that 2g(0) =
1
0f (0) = 0, therefore showing that g = Xg for some g R. Hence t = X Xg =

1
1
g R. Since obviously R 2 , R X and R [R : h2, Xi]T , we see that
[R : h2, Xi]T = R. But since 1
/ h2, Xi, we see that h2, Xi [R : h2, Xi]T =
h2, Xi R = h2, Xi =
6 R. Contradiction!

It turns out that the invertibility of an ideal can also be characterized by using
module theory. To see this, we first require three definitions.

Definition 2.20 Direct Sum

Let R be a commutative ring and let {Mi : i I} be a family of R-submodules of the


P P
R-module T . Then the sum iI Mi := iI mi : mi = 0 for almost all i I
L
also is an R-submodule of R. We say that this sum is direct and write it as iI Mi
P P
if every element m iI Mi has a unique representation in the form m = iI mi
where mi Mi and mi = 0 for almost all i I.

Definition 2.21 Free Module

Let R be a commutative ring and let M be an R-module. If M has a basis (i.e. if


there exists a linearly independent subset N of M such that M = hN i), M is called
free.

Definition 2.22 Projective Module

29
2.4. FRACTIONAL IDEALS

Let R be a commutative ring. An R-submodule P of R is called projective if there


exists a free module F such that F = P Q for some module Q.

Now we can state the aforementioned characterisation:

Theorem 2.23 Characterisation of invertible ideals

Let I be an ideal of some commutative integral domain R. Then I is invertible if


and only if I is projective as an R-module.

Proof:

: Let I be a non-zero ideal of R. Suppose that I is projective, i.e. that there


exist a module M and a free module F such that F = I M as above. Say that F
has the basis {e : }, so that
M
F = he i .

For each , let be defined as follows:


(
F R
: P .
r e 7 r

Quite obviously, is a module homomorphism for all .


For each , let e = a + m where a I, m M . Given x I, x =
P
(x)e , we see that
! !
X X X X
x= (x)(a + m ) = (x)a + (x)m = (x)a .

If 0 6= x, y I, then

(xy) = x (y) = y (x).

So in the total quotient field T we have


(x) (y)
b := = for all 0 6= x, y I.
x y
Note that since (x) = 0 for almost all , we have b = 0 for almost all . Also
(x)x
x I, b x = = (x) R,
x
so b I R. Now we see:
X X X
x I, x = (x)a = (xb )a = x b a

and so, since x 6= 0, we get


X
1= b a .

30
CHAPTER 2. QUOTIENT RINGS

Set I 0 := h{b : }i, then I 0 I = R and so I is invertible.


Pn
: Suppose I is invertible with II 0 = R. Let 1 = j=1 aj bj where aj I,
bj I 0 . Define
( Ln
I i=1 R
w:
x 7 (b1 x, b2 x, . . . , bn x)

and
( Ln
i=1 R I
: Pn .
(r1 , . . . , rn )
7 j=1 rj aj

Then both w and are module homomorphisms.


Now
n
X n
X
(w(x)) = (b1 x, . . . , bn x) = bj xaj = x aj bj = x.
j=1 j=1
Ln
Therefore, I is isomorphic to a direct summand of i=1 R. This shows that I is
projective. 

Example:

Since by the example following Proposition 2.19 h2, Xi is not an invertible ideal of
R := Z[X], Theorem 2.23 shows that h2, Xi is not projective as an R-module.

31
Chapter 3

Extension Rings and


Factorization of Ideals

3.1 Integral Extensions


In a similar way as algebraically extending a field (cf. [4, chap. 20]), we can extend
rings. We will call this integral extension, more rigidly explained in the following
definition.

Definition 3.1 Integral Extension

Let R be a ring. An element t T for some extension ring T R is called integral


over R if there exists a monic polynomial p R[X] such that p(t) = 0. If this
polynomial is of minimal degree, it is called the minimal polynomial of t over R. If
all elements of T are integral over R, T is said to be an integral ring extension of
R.

Proposition 3.2 Uniqueness of minimal polynomials

Let R be a ring, let a be integral over R. Then its minimal polynomial is unique.

Proof:

Let f R[X] be a minimal polynomial of a. Assume there exists a second minimal


polynomial g R[X] for a. Since they both are minimal, deg(f ) = deg(g). Now
consider

(f g)(a) = f (a) g(a) = 0 0 = 0.

Since both f and g are monic, deg(f g) < deg(f ). This is a contradiction to the
degree of f being minimal, therefore f g = 0, i.e. f = g. 

32
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

Before we can proceed to the next result, we require some definitions and a well-
known result from Linear Algebra:

Definition 3.3 Minors, Cofactors, Adjoint Matrix

Let A be an n n matrix over some commutative ring R. Then the matrix obtained
by dropping row i and column j from the matrix A is called the ij-th minor of A,
usually denoted by Mij . The determinant of Mij multiplied by (1)i+j is called
the ij-th cofactor of A, usually denoted by Aij :

Aij := (1)i+j |Mij |

The matrix
  T
adj(A) := Aij i=1,...,n
j=1,...,n

is called the adjoint matrix of A.

Theorem 3.4 Laplace Expansion

Let A be an n n matrix with entries aij over some commutative ring R. Denote
the ij-th cofactor by Aij . Then
n
!
X
A adj(A) = aik Akj = |A| 1.
k=1 i=1,...,n
j=1,...,n

Proof:

A proof shall not be given here. It can be found in [9, Theorem 23.10 on page 146]
and in many other texts on linear algebra. 

Theorem 3.5 Integral Extension

Let T be a ring extension of some commutative ring R. Then the following are
equivalent:

(a) t T is integral over R.

(b) The ring R[t] is finitely generated as an R-module, i.e.

R[t] = {r0 + r1 t + + rn tn : r0 , r1 , . . . , rn R} .

(c) R[t] is contained in a subring S T that is finitely generated as an R-module.

(d) There exists a finitely generated R-module S T such that

33
3.1. INTEGRAL EXTENSIONS

(i) tS S and

(ii) if u R[t] and uS = 0, then u = 0.

Proof:

(a) (b): Let t be integral over R with minimal polynomial p, deg(p) = m for
some m N. Then consider the finitely generated R-submodule N of T defined as
follows:

N := R + tR + . . . + tm1 R

Since p(t) = 0, we know that tm N . Now assume that for some integer k m,
Pm1
tk N . Hence, tk = i=0 ri0 ti for some ri0 R, i = 0, . . . , m 1. And therefore,
0
Pm1 0 i
in accordance with the above, tk+1 = ttk = rm1 tm + i=1 ri1 t has to lie in N .
This shows that for all k N, tk N and therefore N = R[t].

(b) (c): Obvious.

(c) (d): We take S to be the subring required to exist by condition (c). Then,
tS S since t S. Furthermore, if uS = 0, u = u 1 = 0, since 1 S.

(d) (a): (By [1, Lemma 1.3.2])


Let S be the finitely generated R-module with properties (i) and (ii), required to
exist by (d). By property (i), it is also finitely generated as an R[t]-module. Then,
by property (ii), ann(S) = {0}. Assume S is generated by n N elements from R,
say S = hs1 , . . . , sn i. Hence, for all k {1, . . . , n}:
n
X
tsk = akj sj for some aij R, i, j = 1, . . . , n
j=1
n
X
and so (kj t akj )sj = 0 k.
j=1

Now consider the matrix



A := ij t aij i=1,...,n .
j=1,...,n

Let D := det(A) = |A| and let Mij be the ij-th minor of A and

Aij := (1)i+j |Mij |,

the ij-th cofactor of A. Then for all i


n
X n
X n X
X n
0 = Aki (kj t akj )sj = Aki (kj t akj )sj =
k=1 j=1 k=1 j=1
n n
! n
3.4
X X X
= Aki (kj t akj ) sj = Dij sj = Dsi .
j=1 k=1 j=1

34
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

This shows that D ann(S), therefore D = 0. But we also know that for some
a0 , . . . , an1 R

D = |A| = tn + an1 tn1 + + a1 t + a0 .

Hence t is integral over R. 

Definition 3.6 Integral Closure

Let T be a ring extension of a ring R. Then the set R = {t T : t integral over R}


is called the integral closure of R in T .

Definition 3.7 Normal Ring

A ring R is called normal if it is integrally closed, i.e. if R = R, in its total ring of


fractions.

Theorem 3.8 Integral Closure

The integral closure R of some commutative ring R in some commutative superring


T of R is a ring.

Proof:

Consider two elements t1 , t2 R. Then, by Theorem 3.5, the ring R[t1 , t2 ] =


R[t1 ][t2 ] is a finitely generated R[t1 ]-module and hence a finitely generated R-
module. Since R[t1 t2 ], R[t1 + t2 ] and R[t1 t2 ] are all contained in R[t1 , t2 ], it
follows from Theorem 3.5 (c), that t1 t2 , t1 + t2 and t1 t2 are integral over R.
Hence R is a ring. 

Example:
 
Let R = Z, T = Q 5 . What is the integral closure of R in T ? We will proceed
in several steps:

(a) Let a, b R. Then a + 5b satisfies the monic polynomial p(x) = x2 2ax +
 
a2 + 5b2 = (x a)2 + 5b2 . Therefore, R Z 5 .

p p
(b) Consider q Q, p, q Z, q 6= 0, gcd(p, q) = 1. Now assume that q is integral
over R, i.e. there exists a monic polynomial f R[X] such that f (p/q) = 0,
f (x) = a0 + a1 x + . . . + an1 xn1 + xn , ai R, i = 0, . . . , n 1. Then:

n
 
p 1 p
f = n1 q n1 a0 + q n2 pa1 + . . . + pn1 an1 + = 0 R
q q | {z } q
R

35
3.1. INTEGRAL EXTENSIONS

Therefore, since R is a ring, pn /q R. Since gcd(p, q) = 1, this yields q = 1.



Hence, p/q R. Similarly, if 5p/q is integral over R, p/q R.

(c) Let z = a + 5b T , a, b Q, be integral over R such that

p1 p2
a= , b = , p1 , p2 , q1 , q2 Z, gcd(p1 , q1 ) = gcd(p2 , q2 ) = 1.
q1 q2

Then, by Theorem 3.8,

q 1 p2 q 1 p2
q 1 z p1 = 5 R and so, by (b), R , giving q2 |q1
q2 q2

and
q 2 p1 q 2 p1
q2 z 5p2 = R and so, by (b), R , giving q1 |q2 .
q1 q1

Therefore, q1 = q2 . Without loss of generality, assume that q := q1 = q2 .



Hence we can write z = rq where r := p1 + 5p2 R. Let g(x) = b0 + b1 x +
. . . + bn1 xn1 + xn , bi R, i = 0, . . . , n 1, be the minimal polynomial of
z. Then

1 n1 rn
g (z) = q b0 + q n2 rb1 + . . . + rn1 bn1 + = 0 R.
q n1 | {z } q
R

rn
This shows that, since R is a ring (by Theorem 3.8), q R. Consider
the prime factorisation of q, say q = s1 sk for some k N, si prime for
i = 1, . . . , k. Then si |r, yielding si |p1 and si |p2 for all i = 1, . . . , k. Since
gcd(q, p1 ) = gcd(q, p2 ) = 1, we see that si = 1 for all i = 1, . . . , k. Therefore,

q = 1. This shows that R Z[ 5].
 
Hence, R = Z 5 .

Furthermore, the ring R is an intriguing example: it shows that the property of


unique factorisation is lost in this simple extension ring:
 
1+ 5 1 5 = 6 = 2 3 (1 )

Even worse, in this ring prime numbers (i.e. the prime elements of Z) are not nec-
essarily irreducible:
 
6+ 5 6 5 = 41

We also see that 2, 3, 1 + 5 and 1 5 are irreducible. To show this, consider
the map
(
R N
N: .
z =a+ 5b 7 |z|2 = z z = a2 + 5b2

36
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

It is obvious that 2, 3
/ N (R) and that N preserves multiplication, i.e. N (xy) =
N (x)N (y) for any x, y R. Since N (2) = 4 = 2 2, N (3) = 9 = 3 3 and

N (1 + 5) = N (1 5) = 6 = 2 3, we see that the four aforementioned
elements are in fact irreducible and no two of them are associates.
This shows together with (1 ) that irreducible elements are not necessarily prime
(for a proper definition of prime and irreducible elements, refer to [4])! (see [5])

This phenomenon was examined in detail for the first time by Ernst Kummer in
1843 while trying to prove Fermats Last Theorem. He introduced so-called ideal
numbers, imaginary primes, to explain the equation (1 ) by setting

2 = 1 2 , 3 = 3 4 , 1 + 5 = 1 3 , 1 5 = 2 4 , (2 )

where N (1 ) = N (2 ) = 2 and N (3 ) = N (4 ) = 3. The achievement of Constantin


Dedekind then was to interpret these ideal numbers as ideals:





1 = 2, 1 + 5 , 2 = 2, 1 5 , 3 = 3, 1 + 5 , 4 = 3, 1 5

It can be shown that these ideals in fact are prime ideals in R and that the equations
(2 ) hold as equations of ideals:



h2i = 1 2 , h3i = 3 4 , 1 + 5 = 1 3 , 1 5 = 2 4

The following section will determine in what rings it is possible to decompose an


ideal into a product of prime ideals.

3.2 Dedekind Domains


Since as mentioned before not every ring supports the property of (unique) prime
factorisation, the question arises how much of this can be saved. The solution is
to look at classes of ring elements, ideals, and try to decompose these into products
prime ideals. We will now determine in which rings this kind of prime factorisation
is possible.

Throughout the remainder of this chapter, all rings will be considered to be com-
mutative unless stated otherwise.

Definition 3.9 Dedekind Domain

Let R be a commutative ring. Then R is called a Dedekind domain if

(a) R is a Noetherian domain,

(b) R is normal, and

(c) non-zero prime ideals of R are maximal.

37
3.2. DEDEKIND DOMAINS

Example:

The standard examples of Dedekind domains are the rings of algebraic integers

within any extension field of Q, for example the aforementioned ring Z[ 5]. That

Z[ 5] indeed is Dedekind can be shown using module theory and short exact se-
quences (see [1, p. 56]).

Definition 3.9 leads to the following interesting result:

Theorem 3.10 Invertibility of Prime Ideals

Let R be a Dedekind domain and P be a non-zero prime ideal of R. Then P is


invertible in T and its inverse P 1 = [R : P ]T .

Proof:

By Definition 2.16, [R : P ]T is a fractional ideal of R. This means that there exists


a regular element d R such that dI R. Since R is Noetherian, dI is finitely
generated by da1 , . . . , dan , say. Therefore, I is generated by a1 , . . . , an . We show
that [R : P ]T 6= R. Pick some 0 6= p P . Now we show that Rp contains a product
P1 Ps , s N, of non-zero prime ideals of R.
Suppose not. Then there exists a maximal ideal I which does not contain a product
of non-zero prime ideals, since R is Noetherian and hence satisfies the ACC (see
Theorem 1.14). Now, I cannot be prime. Hence (by Proposition 1.11), there exist
two ideals A, B of R such that AB I but A 6 I and B 6 I. Therefore, (I +
A)(I + B) = II + IA + IB + AB I. But I 6= I + A ) I and I 6= I + B ) I
and by definition of I there are non-zero prime ideals P1 , . . . , Pk and Q1 , . . . , Ql ,
k, l N, such that I + A P1 Pk and I + A Q1 Ql . However, this means
that I (I + A)(I + B) P1 Pk Q1 Ql . A contradiction!
Now, assume the number s to be minimal. Since P Rp P1 Ps , P contains
one of the Pi s, say P1 . Since R is Dedekind, P1 is maximal and hence P = P1 .
Next, assume s = 1. Then, P = Rp = P1 and [R : P ]T = R p1 is the unique inverse
of P . This fractional ideal is not R since p is not a unit (P is a prime ideal and
hence cannot contain units). Now assume s 6= 1. Now, P Rp P P2 Ps . By
the minimality of s we have Rp 6 P2 Ps . Therefore, there exists an element
r r
r P2 Ps such that r
/ Rp. This yields p T \R, i.e. p
/ R. Since

r 1 1 1
P = rP P P2 Ps Rp = R,
p p p p
r
we get p [R : P ]T and therefore [R : P ]T 6= R.
By definition of [R : P ]T we immediately see that P P [R : P ]T R. Since P
is maximal, to show that P [R : P ]T = R, it suffices to show that P [R : P ]T 6= P .
Suppose that P [R : P ]T = P . Then for every x [R : P ]T we have xP P . Since

38
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

R is Noetherian, P is finitely generated as an R-module. And since R is an integral


domain, uP = 0 requires u = 0 since P 6= 0. Therefore, by Theorem 3.5 (d), we
know that x is integral over R. Since R is normal, we establish x R, a contradic-
tion. Therefore, P [R : P ]T = R. 

Now we can prove that indeed the concept of unique factorisation which is not
necessarily possible in a Dedekind domain can be saved by generalising it to the
unique factorisation of ideals.

Theorem 3.11 Unique factorisation

Every non-zero ideal of a Dedekind domain has a unique (up to ordering) represen-
tation in the form of a product of prime ideals.

Proof:

Let R be a Dedekind domain. Denote by I the family of those non-zero ideals


in R which do not have a representation as a product of prime ideals. Suppose
that I 6= . Then since R is Noetherian there exists a maximal element I in
I. I cannot be maximal, since all maximal ideals are prime (Proposition 1.11).
Therefore it is properly contained in some maximal ideal P ) I. By Theorem 3.10,
P 1 := [R : P ]T is the unique inverse of P . From the condition I ( P we obtain
IP 1 P P 1 = R, while P 1 ) R yields IP 1 I. Furthermore we see that
I 6= IP 1 :
Assume they were equal. Then for each x P 1 we get xI I and therefore, by
Theorem 3.5 (d) and R being a Noetherian domain, x is integral over R. Since R
is normal, this yields P 1 R. This is not possible by the proof of Theorem 3.5.
Hence, IP 1 ) I and IP 1 is a proper ideal since the equality IP 1 = R would
imply that I = IP P 1 = RP = P , contradicting the assumption I I. Since
1 1 1
I ( IP ( R, IP must be a product of prime ideals say IP = P1 P2 Pk .
Then I = P P1 P2 Pk , showing that I 6= I .

It remains to show that this representation is unique. Assume that

() I = P1 Ps = Q1 Qr

for some proper non-zero primes Pi , Qj , i = 1, . . . , s, j = 1, . . . , r. Since P1


P1 Ps = Q1 Qr , Proposition 1.11 (i) shows that there exists j {1, . . . , s}
such that P1 Qj . In the same way it can be shown that Qj Pi for some i. Since
R is Dedekind, P1 is maximal. This shows that P1 = Qj = Pi . By Theorem 3.10,
P1 is invertible and therefore it can be cancelled from (), yielding

(0 ) P2 Ps = Q1 Qj1 Qj+1 Qr .

The same process can be continued to further reduce (0 ), thus showing that r = s

39
3.2. DEDEKIND DOMAINS

and that Q1 , . . . , Qr is just a permutation of P1 , . . . , Ps . 

From this theorem, the following corollary follows easily.

Corollary 3.12 Invertible Ideals

Every non-zero ideal of a Dedekind domain is invertible.

Proof:

Let I be a non-zero ideal of a Dedekind domain R. By Theorem 3.11, I =


P1 P2 Pn for some prime ideals P1 , . . . , Pn of R. By Theorem 3.10, each Pi is
invertible. Therefore, we see that
1
IPn1 Pn1 P11 = P1 P2 Pn Pn1 Pn1
1
P11 = R,

establishing that I is invertible. 

Example:

Although Z[X] is Noetherian (see [7, Thrm. 2.17, p. 44]), Z[X] is not Dedekind:
on the one hand side, not all prime ideals are maximal (see the second note after
Proposition 1.11), contradicting Definition 3.9. On the other hand, not all finitely
generated ideals are invertible (see the example after Proposition 2.19), contradict-
ing Corollary 3.12.

Again, there is a different approach to the definition of Dedekind domains by using


module theory:

Theorem 3.13 Dedekind Equivalences

Let R be a commutative integral domain. Then the following are equivalent:

(a) R is Dedekind.

(b) Every ideal of R is projective as an R-module.

(c) Every submodule of a projective R-module is projective (i.e. R is an hereditary


ring).

Proof:

The equivalency (a) (b) has been shown in Theorem 2.23. Due to the complex-
ity of the result (it uses the principle of well-ordering), a proof for the equivalence
(a) (c) shall not be given here. It can be found in [8, p. 124]. 

40
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

The concept of Dedekind domains can be further generalised. This shall be done in
the following section.

3.3 Prufer Domains


The well-known Chinese Remainder Theorem from Number Theory (Appendix B)
can be transferred to a situation where one does not deal with individual elements
of a ring but with ideals.

Definition 3.14 Chinese Remainder Theorem

Let R be a commutative ring, let A1 , . . . , Ak be a finite number of ideals of R and


let x1 , . . . , xk be elements in R. Consider the system of congruences

x x1 (mod A1 )


x x2 (mod A2 )

.. ()
.




x xk (mod Ak )

where x xi (mod Ai ) means that x xi Ai . If this system has a solution x R,


then for any i, j we have x xi Ai , x xj Aj , so xi xj Ai + Aj .
If this condition is also sufficient for all A1 , . . . , Ak ideals of R and for all x1 , . . . , xk
R, then R is said to satisfy the Chinese Remainder Theorem, or CRT.

Definition 3.15 Prufer Domains

An integral domain R is called a Prufer domain if it satisfies the CRT.

This definition immediately leads to some other characterisations of Prufer domains:

Theorem 3.16 Properties of the CRT

Let R be a ring. Then the following are equivalent:

(a) The CRT holds in R.

(b) If A, B, C are ideals in R then A (B + C) = A B + A C.

(c) If A, B, C are ideals in R then A + (B C) = (A + B) (A + C).

Proof:

(b) (c)

41
3.3. PRUFER DOMAINS

Assume (b) holds. Then for all A, B, C ideals of R we have

(A + B) (A + C) = ((A + B) A) + ((A + B) C) =
= A + (C A + C B) = A + (B C).

(c) (b)

Assume (c) holds. Then for all A, B, C ideals of R we have

(A B) + (A C) = (A B + A) (A B + C) = A (C + A B) ()

and

(A B) + (A C) = (A + A C) (B + A C) = A (B + A C).

This shows that C + A B = B + A C. Since for any ideal I, I = I + I, we notice


that

C + A B = C + A B + B + A C = B + C.

Together with (), this yields

A B + A C = A (B + C).

(a) (c)

In any ring, we always have

A + (B C) (A + B) (A + C)

since if r A + (B C), then r = a + b = a + c where a A, b B, c C and


b = c. Hence, r (A + B) (A + C).

The reverse inclusion is shown as follows: Let t (A + B) (A + C). We wish


to write t = x + y where x A, y B C. So we wish to solve the system of
congruences

x 0 (mod A)
x t (mod B)
x t (mod C).

Since t 0 = t A + B, t 0 = t A + C and t t = 0 B + C, x exists since


the CRT holds in R.

(c) (a)

Consider the system () of Definition 3.14 for which xi xj Ai +Aj for all i, j k.
We proceed by induction on k.
k=2

42
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

Then x1 x2 = a1 + a2 for some a1 A1 , a2 A2 . Let x = x1 a1 = x2 + a2 .


Then x x1 (mod A1 ) and x x2 (mod A2 ). Hence, k = 2 is true without using
(3).
k k+1
Assume any system of k congruences to be solvable. Consider the system of k + 1
congruences

x xi (mod Ai ), i = 1, . . . , k, k + 1

where xi xj Ai + Aj for all i, j k + 1.


By induction, there exists y R such that y xi (mod Ai ) for all i = 1, . . . , k.
Consider the system of two congruences
k
!
\
x y mod Ai
i=1
x xk+1 (mod Ak+1 ).

Then, from the k = 2 case, we know there exists a solution t R of this system,
provided
k
\
y xk+1 Ak+1 + Ai .
i=1

Then t y Ai for each i = 1, . . . , k and t xk+1 Ak+1 . But y xi Ai for


i = 1, . . . , k, so we also get

t xi = (t y) + (y xi ) Ai .

Hence, t will be a required solution for the system of k + 1 congruences.


Since (3) holds in R, we have
n
! n
\ \
A+ Bi = (A + Bi ) (by induction)
i=1 i=1

and hence
k
\ k
\
Ak+1 + Ai = (Ak+1 + Ai ) . (2 )
i=1 i=1

We also have:

y xk+1 = (y xi ) + (xi xk+1 ) Ai + Ak+1 i = 1, . . . , k


| {z } | {z }
Ai Ai +Ak+1
k
\
and so, by (2 ), y xk+1 Ak+1 + Ai
i=1

This completes the proof. 

Example:

43
3.3. PRUFER DOMAINS

Let R = Z and let A, B, C be ideals of R such that A = R pa1 1 pakk , B =


R pb11 pbkk , C = R pc11 pckk where pi prime and ai , bi , ci 0 for i = 1, . . . , k.
Then A B = R pd11 pdkk where di = max{ai , bi } and A C = R pe11 pekk
where ei = max{ai , ci }.
Hence, A B + A C = R pf11 pfkk where

fk = min{di , ei } = min{max{ai , bi }, max{ai , ci }} = max{ai , min{bi , ci }}.

Therefore, A B + A C = A + (B C).

Proposition 3.17

Let R be a commutative ring in which the ideals form a chain with respect to
inclusion. Then the CRT holds in R.

Proof:

Let A, B, C be ideals of R. Without loss of generality, we suppose that B C.


Then

(A + B) (A + C) = A + B = A + (B C).

Hence, by Theorem 3.16, the CRT holds in R . 

Theorem 3.18 Quotient Rings


Let R be a ring with ideals A1 , . . . , Ak for some k N. Put Ri := R
Ai for each
i k and define
( Qk
R i=1 Ri = R1 . . . Rk
f: .
r 7 (r + A1 , . . . , r + Ak )

Then f is an epimorphism if and only if the (Ai )i are pairwise comaximal.

Proof:

Suppose f is an epimorphism. We show that A1 + A2 = R, the argument for an


arbitrary pair is similar. Let e1 = (1 + A1 , 0, . . . , 0). Then, since f is onto, there
exists an x1 R such that f (x1 ) = e1 . But f (x1 ) = (x1 + A1 , x1 + A2 , . . . , x1 + Ak )
and hence, x1 + A1 = 1 + A1 and x1 + A2 = 1 + A2 . Therefore, x1 1 A1 and
x1 A2 .
Hence 1 = 1 x1 + x1 A1 + A2 A1 + A2 = R.

44
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

Suppose that Ai + Aj = R for all i 6= j. It is sufficient to show that if r R and


ri = (0, . . . , 0, r + Ai , 0, . . . , 0), then ri Im(f ).
T
For each i let Ci = j6=i Aj . Now, for each 1 < i k there exists an xi A1 and
yi Ai such that 1 = xi + yi , since A1 + Ai = R. Then y2 yk = (1 x2 ) . . . (1
xk ) 1 (mod A1 ) and y2 yk C1 . Therefore, A1 + C1 = R, 1 = a1 + c1 , say, for
some a1 A, c1 C. Then f (c1 ) = (c1 +A1 , c1 +A2 , . . . , c1 +Ak ) = (1+A1 , 0, . . . , 0).
Since f is a homomorphism, we find that f (r c1 ) = (r + A1 , 0, . . . , 0). A similar
argument for each Aj , j 6= 1, shows that f is an epimorphism. 

Example:

Let A1 , . . . , Ak , k N, be pairwise comaximal ideals of some ring R and let


x1 , . . . , xk R. Consider the system of congruences

() x xi(mod Ai ) (i = 1, . . . , k).

Then by Theorem 3.18 f : R
Qk R
i=1 Ai is onto. Hence r R such that
f (r) = (x1 + A1 , . . . , xk + Ak ). Therefore, r is a solution of ().

Definition 3.19 Chain Ring

A commutative ring R is called a chain ring or a valuation ring if the ideals of R


form a chain with respect to inclusion.
An integral domain R is called a chain domain or a valuation domain if it is a chain
ring.

Proposition 3.20 Chain Ring

A ring R is a chain ring if and only if for all a, b R either aR bR or bR aR.

Proof:

: obvious.

Let I, J be ideals of R. Suppose I 6 J and J 6 I. Then there exist a I\J,


b J\I, but either aR bR or bR aR, leading to a contradiction. 

Proposition 3.21 Units in a Chain Ring

Let R be a chain ring with a unique maximal ideal M . Then R\M is the set of
units in R.

45
3.3. PRUFER DOMAINS

Proof:

Obviously, since M 6= R, M cannot contain any units. Now, let s R\M . Assume
s is not a unit. Then Rs 6= R. Therefore, by Proposition 2.7, Rs M , yielding
s M . Contradiction! 

Example:

We will now give an example where the divisional ideal of two principal ideals is
not finitely generated, let alone principal (this was promised in section 1.2). See
also [6].
Let K be any field. Now we will consider the ring R of power series over K in the
indeterminate X. Define R as follows:
( )
X
i
R := K[[X]] := ai X : ai K
iN0

The units in R are exactly the elements where a0 6= 0: Assume f = iN0 ai X i R


P

is a unit, i.e. g = iN0 bi X i R such that f g = 1. This can only be the case if
P

a0 b0 = 1. Therefore b0 = a1
0 , i.e. a0 6= 0. The remaining coefficients of g can now
easily be calculated. Now if f, g R are non-zero non-units, there exist m, n N
and units k, h R such that

f = X mk and g = X n h.

Without loss of generality, we assume that m n. Then

g = X n h = (X nm hk 1 )X m k = (X nm hk 1 )f.

Therefore, hgi hf i. Proposition 3.20 then shows that R is a chain ring and this
yields (by Proposition 3.21) that the unique maximal ideal N = hxi.
This result can in a similar way be generalised to the ring R0 of generalised
power series over K,
( )
X
0 si +
R := ai X : ai K, si R0 where i < j si sj ,
iN0

yielding that R0 as well is a valuation domain with unique maximal ideal M. Then
M is idempotent in R0 , i.e. M2 = M: Consider X s k M, k a unit. Then
X sk = X 2 X 2 k M 2.
s s

Now we show that M cannot be finitely generated. Assume it were. Then since
R is a chain ring M = mR0 for some m R0 . Since M is idempotent, we have
0

m M 2 = m2 R0 , so m = m2 x, (x 6= 0), yielding mx = 1. This is a contradiction


since m cannot be a unit.
In fact, M is countably generated, namely

M = x0.1n : n N .

46
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

Now let 0 6= b M , I := bM , and consider the following factor ring



0
S := R
I.

It is easy to see that b + I is a non-zero element of S and that (b + I)2 = b2 + bM =


0 + M . Therefore, S is not an integral domain. Let b := b + I and consider the
divisional ideal



h0i : b = s S : s b h0i = s S : sb = 0 = ann(b) =
= {r + I S : (r + I)(b + I) = 0} = {r + I S : rb I} =
M
   
0
IrM =
= r+I R
I.

Now assume that M /I is finitely generated, say since S is a chain ring M /I =


hm + Ii. Then, given any y M , we have

y + I = rm + I for some r R0 .

So y rm bM y rm = bn for some n M . This yields

y = rm + bn and so y hm, bi ,

showing that

M = hm, bi ,


a contradiction. This establishes that h0i : b is not finitely generated.

Note:

The ring of generalised power series, R0 , defined in the above example is a Prufer
domain since it is a chain ring (by Proposition 3.17) but not a Dedekind domain
since it is not Noetherian (its unique maximal ideal M is not finitely generated).

Theorem 3.22 Characterisation of Prufer domains I

Let R be an integral domain. Then R is a Prufer domain if and only if RP is a


chain ring for all 0 6= P prime ideals of R.

Proof:

We will show this by proving equivalence to part (2) of Theorem 3.16. Assume that
for every proper non-zero prime ideal of R, RP is a chain ring. Now consider ideals
AP , BP , CP of R. Then without loss of generality BP CP . This shows:

AP (BP + CP ) = AP CP = AP BP + AP CP .

47
3.3. PRUFER DOMAINS

Now consider ideals A, B, C of R (we make frequent use of Proposition 2.11):

(A (B + C))P = AP (B + C)P = AP (BP + CP ) =


= AP BP + AP CP = (A B)P + (A C)P =
= (A B + A C)P .

By the Localisation Principle (Theorem 2.12), this shows that

A (B + C) = A B + A C.

Conversely, assume that (2) of Theorem 3.16 holds. Let P be a proper non-zero
prime ideal of R and let a, b R. Since a hbi + ha bi, we have

hai = hai (hbi + ha bi) = (hai hbi) + (hai ha bi) .

Then a = t + c(a b) where t hai hbi, c R and c(a b) hai. Then cb hai
and (1 c)a = t cb hbi. If c
/ P , then b aRP . If c P , then 1 c
/ P and
a bRP . Proposition 3.20 then shows that RP is a chain ring. 

This gives reason to a generalisation of Prufer domains, the so-called arithmetical


rings:

Definition 3.23 Arithmetical Ring

A ring R is arithmetical if RM is a chain ring for all maximal ideals M of R.

Theorem 3.24 Characterisation of Arithmetical Rings

In a commutative ring R, the following are equivalent:

(a) R is arithmetical.

(b) A : B + B : A = R for all finitely generated ideals A, B of R.

(c) Rx : Ry + Ry : Rx = R for all x, y R.

Proof:

Consider the set of maximal ideals of R, denoted by {M } . We will proceed in


several steps:

(1) Claim:
!c
\
U= UM for any ideal U of R.

48
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

Proof:


u
Since M is a maximal ideal, M 6= R and so 1
/ M . Thus 1 UM u
T c
U, and so U UM .

UM . Assume that v 6 (u, 1) u U . Let v = ur00 for


T
Let v
S T
some u0 U , r0 R\ M = R\M . But, by Proposition 2.7,
T
there exists 0 such that Rr0 M0 . Therefore, r0
/ R\M .
Contradiction!
T c
Hence U UM .

(2) By the Localisation Principle, Theorem 2.12, we know that for two ideals
U1 , U2 of R, U1 = U2 if and only if (U1 )M = (U2 )M .

(3) Consider the map

I I
(

: , ,
U 7 UM

where I are the non-zero ideals of R and I those of RM


.
Claim: For all , is surjective.
Proof:

Let I I . Then I c is an element of I and by Proposition 2.11, (I c ) =


(I c )S = I. Therefore, (I ) = I .

(4) By Proposition 2.11 we know that for A, B I

(A + B) = (A) + (B)
(AB) = (A) (B)
(A B) = (A) (B)
(A : B) = (A) : (B), if B is finitely generated.

Now suppose that an equation (E1 ) involving ideals and only the operations
+, , , : is true for every quasi-local ring. If (E1 ) holds for R, then, in partic-
ular, it is true for RM , . Furthermore, if (E1 ) (E2 ) for some other
equation (E2 ) involving ideals and only the operations +, , , :, we hence
note that (E2 ) holds for RM , . Then the Localisation Principle (Theo-
rem 2.12) shows that (E2 ) must also be true for R.
This shows that we can assume that R is quasi-local.

(a) = (b):

Now, if R is arithmetical and quasi-local, R is a chain ring, since, if M is the unique

49
3.3. PRUFER DOMAINS

maximal ideal of R, R = RM . Hence, if A, B are finitely generated ideals of R, we


note that either A B or B A, yielding that either A : B = R or B : A = R.
Therefore, A : B + B : A = R.

(b) = (c):

This is obvious.

(c) = (a):

Assume (c) holds. Let M be the unique maximal ideal of R. We know that not
both Rx : Ry and Ry : Rx can be contained in M . Without loss of generality,
Rx : Ry 6 M . Hence, Rx : Ry must contain an element m
/ M , which since
M is unique must be a unit, yielding y Rx. Therefore, Ry Rx. Hence, by
Proposition 3.20, R is a chain ring:

Let AM BM be two ideals of RM such that A and B are ideals of R. Let a A.


a
Then 1 BM and so m R\M such that am B. Consider the ideal generated
by m:

M 0 = Rm.

Assume M 0 6= R then, by Proposition 2.7, M 0 M . Contradiction! Hence,


M 0 = R.
Thus n R such that mn = 1 and so a = 1 a = mn a = n |{z}
ma B.
B

This shows that R is arithmetical. 

Quite obviously, if R also is required to be an integral domain, Theorem 3.24 yields


even more equivalent definitions for Prufer domains.

Theorem 3.25 Characterisation of Prufer domains II

In a commutative integral domain R, the following are equivalent:

(a) All finitely generated ideals of R are invertible.

(b) If AB = AC for a finitely generated, non-zero ideal A, B = C (the cancellation


property for finitely generated ideals).

(c) For all proper prime ideals P of R, RP is a chain ring.

(d) (A + B) : C = A : C + B : C for all ideals A, B, C of R, C finitely generated.

(e) Every non-zero ideal of R that is generated by two elements is invertible.

50
CHAPTER 3. EXTENSION RINGS AND FACTORIZATION OF IDEALS

Proof:

(a) (b):

If A is finitely generated, non-zero, and AB = AC, we see:

B = RB = A1 AB = A1 AC = RC = C.

(b) (c):

Suppose (b) holds. Now let A, B, C be ideals of R, A finitely generated, such that
AB AC. Then:

AC = AB + AC = A(B + C) and so, by (b), C = B + C which gives B C

Let P be a non-zero proper prime ideal of R. It is sufficient to show that for


a b

a
b
b
a
s , t RP , at least one of s t and t s holds (see Proposition 3.20).
1 1
Since s, t
/ P we see that s and t are units in RP and it therefore suffices to show
that either aRP bRP or bRP aRP . This obviously is true for a = 0 or b = 0.
Hence we can assume that both a 6= 0 and b 6= 0. Now consider the following:

habi ha, bi = habi hai + hbi = habi hai + habi hbi = a2 b, ab2


a2 b, ab2 , a3 , b3 = a3 + a2 b + ab2 + b3 =







2
2 
hai + hbi = a2 , b2 ha, bi


= a + b


Therefore, by the above, habi a2 , b2 . Then ab = xa2 + yb2 for some x, y R.
Now consider

yab, yb2 = y(xa2 + yb2 ), yb2 = xya2 + y 2 b2 , yb2 =






hybi ha, bi =
= xya2 , yb2 a2 , yb2 = a2 , xa2 + yb2 = a2 , ab = hai habi .




This yields by the above hybi hai. Say yb = au for some u R. Then

ab = xa2 + uab and so xa2 = ab(1 u). Thus xa = b(1 u).

/ P , then a = b uy bRP . If u P then 1 u


Now, if u / P and therefore
x
b = a 1u aRP .
(c) (d):

Assume (c) and let P be a non-zero proper prime ideal of R. Then (d) holds for
ideals of RP since, for AP , BP , CP ideals of RP , without loss of generality we can
assume that AP BP . Then (AP + BP ) : CP = BP : CP = AP : CP + BP : CP .
Now let A, B, C be ideals of R, C finitely generated (we make frequent use of
Proposition 2.11):

((A + B) : C)P = (A + B)P : CP = (AP + BP ) : CP = AP : CP + BP : CP =


= (A : C)P + (B : C)P = (A : C + B : C)P

Now, by the Localisation Principle (Theorem 2.12), (d) holds.

51
3.3. PRUFER DOMAINS

(d) (e):

Let 0 6= a, b R. If (d) holds, we see that

R = ha, bi : ha, bi = hai : ha, bi + hbi : ha, bi = hai : hbi + hbi : hai .

Let 1 = x + y where xb hai and ya hbi. Then hxbi b habi and hyai a habi.
Hence, ha, bi hbx, ayi habi. But ab = abx + aby, so habi = ha, bi hbx, ayi. Since
a, b 6= 0, the principal ideal habi is invertible. This shows that ha, bi is invertible.
(e) (a):

Let C = hc1 , . . . , cn i be a non-zero ideal of R. We will proceed by induction to show


that C is invertible. The claim obviously is true for n = 1 and by (e) for n = 2.
Now assume it is true for n 1, n > 2. Let

A := hc1 , . . . , cn1 i , B := hc2 , . . . , cn i ,


D := hc1 , cn i , E := c1 A1 D1 + cn B 1 D1 .

Then

CE = (A + hcn i) c1 A1 D1 + (hc1 i + B) cn B 1 D1 =
= c1 D1 + cn c1 A1 D1 + c1 cn B 1 D1 + cn D1 =
 ()
= c1 D1 R + cn B 1 + cn D1 R + c1 A1 = c1 D1 + cn D1 =


= hc1 , cn i D1 = DD1 = R.

This shows that C is invertible.


Note that the starred equality above holds because cn B 1 R and c1 A1 R. 

Note:

Because of Theorem 3.22, this shows that a Prufer domain is nothing more than a
generalisation of a Dedekind domain: not all ideals are necessarily invertible, just
those that are finitely generated. Furthermore, we note that in the case of R being
Noetherian, the two definitions coincide since then all ideals are finitely generated.

52
Appendix A

Zorns Lemma

A well-known yet fairly hard to grasp result is Zorns famous lemma. Since we used
it in chapter 3, it shall be stated here.

Definition A.1 Partial Ordering

Let X be some set. We say that X is partially ordered with respect to some binary
relation defined on X X if for any a, b, c X the following hold:

a a (reflexivity)

If a b and b a, then a = b (antisymmetry)

If a b and b c, then a c (transitivity)

Then (X, ) is called a partially odered set.

Definition A.2 Properties of partially ordered sets

Let (P, ) be a partially ordered set, also called poset. Given a subset X P , an
element u P is called an upper bound of X, if x u x X.
An element m X is called maximal if there exists no y P such that m < y.
A subset C P is called a chain in (P, ) if each pair of elements c, d C is
comparable, i.e. c d or d c

Theorem A.3 Zorns Lemma

If (P, ) is a poset such that every chain has an upper bound in P , then P has a
maximal element.

53
Appendix B

The Chinese Remainder


Theorem

The numbertheoretical version of the Chinese Remainder Theorem (the ring the-
oretic equivalent was stated in Definition 3.14) was first stated by the Chinese
mathematician Qin Jiushao in 1247.

Theorem B.1 Chinese Remainder Theorem

Suppose that n1 , . . . , nk are pairwise coprime integers for some k N. Then, for any
given integers a1 , . . . , ak , there exists a solution x Z to the system of simultaneous
congruences

x ai (mod ni ) for all i {1, . . . , k} .

Furthermore, all those solutions are congruent modulo n := n1 n2 nk .

The solution in question can be found using the extended Euclidian Algorithm. For
further information, see [3], [2] and [5].

54
Appendix C

Notation

Throughout the entire project, the following symbols were used:

Definition C.1 Notation

denotes containment, ( proper containment

:= is used as a defining operation: a := 1 means define a to be 1

: is used as a defining operation: (1) : (2) means define (1) to be


equivalent to (2)

N = {1, 2, 3, . . .} ( N0 = N {0} ( Z = {. . . , 2, 1, 0, 1, 2, . . .}

Z+ = N, Z+
0 = N0 , Z = Z\N0 , Z0 = Z\N
n o
Q = pq : p N0 , q N

Mn (R) is the ring of all n n matrices where the entries stem from R

In the context of groups, denotes a subgroup: (H, +) (G, +) means H


is an additive subgroup of the additive group G

For any sets S, T, U, V ,


" # (" # )
S T s t
:= : s S, t T, u U, v V
U V u v

denotes a set of 4 4 matrices.

Let S be some set and let x, y be some elements such that x s and s y are
defined for all s S. Then

xS := {x s : s S} and Sy := {s y : s S} .

If x s y is defined for all s S, we set

xSy := {x s y : s S} .

55
Bibliography

[1] Balcerzyk, Stanislaw and Jozefiak, Tadeusz (1989). Commutative Noetherian


and Krull Rings. New York: John Wiley & Sons 34, 38

[2] Clark, John (2004). Rings and Modules (Lecture Notes). Dunedin: University
of Otago 15, 54

[3] OConner, J. J. and Robertsen, E. F. (06.09.2004). Qin Jiushao. University of


St. Andrews: http://www-gap.dcs.st-and.ac.uk/history/Mathematicia
ns/Qin Jiushao.html 54

[4] Gallian, Joseph A. (2002). Contemporary Abstract Algebra (5th edition).


Boston, MA: Houghton Mifflin Company 3, 5, 32, 37

[5] Geyer, Wulf-Dieter (2003). Algebra (Lecture Notes). Erlangen: Friedrich-


Alexander-University 37, 54

[6] Gilmer, Robert (2002). A Note on Finitely Generated Ideals of an Arithmetical


Ring. International Journal of Commutative Rings, 1, 1-4. New York: Nova
Science Publishers, Inc. 46

[7] Larsen, Max D. and McCarthy, Paul J. (1973). Multiplicative Theory of Ideals.
New York: Academic Press 40

[8] Rotman, Joseph J. (1979). An Introduction to Homological Algebra. New York:


Academic Press 40

[9] Weidner, H.-G. (2001). Lineare Algebra (Lecture Notes). Erlangen: Friedrich-
Alexander-University 33

56
Background Literature

[10] Barshay, Jacob (1969). Topics in Ring Theory. New York: W. A. Benjamin

[11] Gilmer, Robert (1972). Multiplicative Ideal Theory. New York: Marcel Dekker

[12] McAdam, Stephen and Swan, Richard G. (2004). Unique comaximal factoriza-
tion. Journal of Algebra, 276, 180-192

[13] Sharpe, D. W. (1987). Rings and Factorization. New York: Cambridge Univer-
sity Press

[14] Sharpe, D. W. and Vamos, R. (1972). Injective Modules. London: Cambridge


University Press

[15] Zariski, Oscar and Pierre, Samuel (1958). Commutative Algebra (Vol. I).
Princeton, NJ: D. van Nostrand Company

57

Das könnte Ihnen auch gefallen