Anda di halaman 1dari 42

Ring Theory (MA4H8)

Charudatta Hajarnavis
Notes by Florian Bouyer
Copyright (C) Bouyer 2011.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3
or any later version published by the Free Software Foundation;
with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.
A copy of the license can be found at http://www.gnu.org/licenses/fdl.html

Contents
1 Chapter 1: Rings

1.1

Rings

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

Properties of Addition and Multiplication

. . . . . . . . . . . . . . . . . . . . . . . . . .

1.3

Subrings and Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.4

Cosets and Homomorphism

1.5

The Isomorphism Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.6

Direct Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.7

Division Rings

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.8

Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.9

Factor Modules and Homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.10 The Isomorphism Theorem


1.11 Direct Sums of Modules

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8
9

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.12 Products of subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.13 A construction

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10

1.14 Zorn's Lemma, Well-ordering Principle, The Axiom of Choice . . . . . . . . . . . . . . .

10

2 Chapter 2: The Jacobson Radical

11

2.1

Quasi-regularity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11

2.2

Commutative Local Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

12

3 Chapter 3: Chain conditions

13

3.1

Finitely Generated Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

13

3.2

Finiteness Assumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

13

3.3

Nil and Nilpotent Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

14

3.4

Nakayama's Lemma and an Application

16

. . . . . . . . . . . . . . . . . . . . . . . . . . .

4 Commutative Noetherian Rings


4.1

Primary Decomposition

4.2

Decomposition of

17

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

17

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18

4.3

Localisation [Commutative Algebra Section 3] . . . . . . . . . . . . . . . . . . . . . . . .

20

4.4

Localisation of a Module [Commutative Algebra 3.1] . . . . . . . . . . . . . . . . . . . .

22

4.5

Symbolic Powers

4.6

The Rank of a Prime Ideal

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22

4.7

Regular Local Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

24

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22

5 Projective Modules

25

5.1

Free Modules

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5.2

Exact Sequences

5.3

Projective Resolutions and Projective Dimension

. . . . . . . . . . . . . . . . . . . . . .

29

5.4

Localization and Global Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

34

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6 Global Dimension of Regular Local Rings

25
25

35

6.1

Change of Rings Theorems

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

35

6.2

Regular Local Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

37

7 Unique Factorization

39

7.1

Unique Factorization Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

39

7.2

Stably Free Modules

40

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Any reference to Commutative Algebra refer to the 2011-2012 Commutative Algebra Lecture notes.
Rings studied will be mostly commutative. We aim to prove:

Theorem (Auslander - Buschsbaum 1959).

A regular local ring is a unique factorization domain.

Reason for selecting this theorem as our destination:


1. It requires sophisticated results from the theory of commutative Noetherian rings.
2. It requires methods from homological algebra. All known proofs require this.
3. At a crucial stage it helps to think in terms of non-commutative rings.
Prerequisite: MA3G6 Commutative Algebra
Topics assumed:
1. Basic properties of Noetherian rings and modules.
2. Primary decomposition
3. Technicality of localization

Denition.

Let R be a commutative Noetherian local ring with 1 and unique maximal ideal M . Let
M = a1 R + + an R (ai M ) be chosen such that n is as minimal as possible. Construct a chain
of prime ideals M ) P1 ) ) Pr (Pi prime) such that r is greatest possible. Then R is regular if
r = n (note that r n always in a Noetherian ring)
Local rings arise naturally in geometry. In algebraic geometry points correspond to local rings.
Existence of an identity is not part of our denition of a ring. For us a right, left or (two sided)
ideal is a subring (Note that in a non-commutative ring, by ideal we will mean a two sided ideal). So
for a right

R-module M , m 1 = m m M

is not a part of our denition.

we shall assume this.

But whenever R has 1,

Chapter 1: Rings

1.1

Rings

Denition 1.1.

Let

be a non-empty set which has tow law of composition dened on it. (we call

these law addition and multiplication respectively and use the familiar notation). We say that

is a ring if the following hold:


1.

a+bR

2.

a + b = b + a a, b R

3.

a + (b + c) = (a + b) + c a, b, c R

and

ab R a, b R
(Commutativity of addition)

4. There exists an element


5. Given

aR

0R

(Associativity of addition)

such that

a R

there exists an element

6.

a(bc) = (ab)c

7.

a(b + c) = ab + ac

a, b, c R

for all

and

a+0=a

for all

such that

aR

a + (a) = 0

(Associativity of multiplication)

(a + b)c = ac + bc

(Distributive Laws)

Thus a ring is an additive Abelian group on which an operation of multiplication is dened; this
operation being associative and distributive with respect to the addition.

is called a commutative ring if it satises in addition

ab = ba

for all

a, b R

The term

non-commutative ring usually stands for a not necessarily commutative ring

1.2

Properties of Addition and Multiplication

The following can be deduced from the axioms for a ring:


1. The element
2. Given

is unique

a R, a

3.

(a) = a

4.

a+b=a+c

5. Given

is uniquely

for all

aR

if and only if

a, b R,

b=c

the equation

Notation. We write

ab

for

x+a=b

to mean

6.

(a + b) = a b

for all

a, b R

7.

(a b) = a + b

for all

a, b R

8.

a0=0a=0

for all

aR

9.

a(b) = (a)b = ab

for all

10.

(a)(b) = ab

a, b R

11.

a(b c) = sb ac

Notation.

Mn (R),

Z,

for all

for all

the integers.

the ring of

nn

Q,

a, b, c R
has a unique solution

x = b + (a)

a + (b)

a, b R

a, b, c R
the rational numbers.

R,

the real numbers.

matrices whose entries are from the ring

R.

C,

the complex numbers.

1.3

Subrings and Ideals

Denition 1.2.

Proposition 1.3.
ab S

whenever

Proof. If

A subset

of a ring

is called a subring of

if

itself is a ring with respect to the

laws of composition of

A non-empty subset

of a ring

is a subring of

if and only if

ab S

and

a, b S

is a subring then obviously the given condition is satised. Conversely, suppose that the

a S . We have a a S hence 0 S . Hence for any x S we have


x S . Finally, if a, b S then by the above b S . Therefore a (b) S , i.e.,
So S is closed with respect to both addition and multiplication. Thus S is a subring since

condition holds. Take any

0x S
a + b S.

so

all the other axioms are automatically satised.

Denition 1.4.
1.

is an ideal of

ra I

of a ring

a I, r R ar I

Proposition 1.5.
and

and

ra I

we denote this fact by

A non-empty subset

a, b I

whenever

is called an ideal if

is a subring of

2. For all
If

A subset

and

I C R.

of a ring

is an ideal of

if and only if

a b I, ar I

rR

Proof. Exercise

1.4

Cosets and Homomorphism

Denition 1.6.

Let

in

called the coset of

xR

be an ideal of a ring

It is denoted by

with respect to

I.

and

. Then the set of elements

{x + i : i I}

is

x+I

When dealing with cosets, it is more important to realise that, in general, a given coset can be
represented in more than one way. The next lemma shows how the coset representatives are related.

Lemma 1.7.

Let

be a ring with an ideal

and

x, y R.

Then

x + I = y + I x y I

Proof. Exercise

R with respect to I by R/I . We can give R/I the structure of a


(x + I) + (y + I) = (x + y) + I and (x + I)(y + I) = xy + I for x, y R.
here is that the sum and the product of R/I are well-dened, that is, they are

We denote the set of all cosets of


ring as follows: Dene
The key point

independent of the coset representatives chosen. Check this and make sure that you understand why
the fact that

is an ideal is crucial to the proof.

Denition 1.8.

R/I

is called the residue class ring of

R/I is 0 + I = i + I
{s + I : s S} of R/I .

The zero element of


by

S/I

the subset

Proposition 1.9.

Let

for any

be an ideal of a ring

R/I
K C R, K I K/I C R/I

1. Every ideal of the ring

R.

is of the form

iI

with respect to

. If

is a subset of

with

SI

we denote

Then

K/I

where

KCR

2. There is a one to one correspondence between ideals of the ring

and

R/I

K I.

Also conversely,

and the ideals of

R containing

I
Proof.

1. If

K C R/I ,

dene

K]{x R : x + I K }

2. The correspondence is given by

K K/I

where

. Then

K C R, K I

K C R, K I

and

K/I = K

Denition 1.10. A mapping of a ring R into a ring S is said to be a (ring) homomorphism if


(x + y) = (x) + (y) and (xy) = (x)(y) for all x, y R.
dened by (r) = 0 for all r R is a homomorphism. It is called the zero homomorphism.
dened by (r) = r for all r R is also a homomorphism. It is called the identity homomorphism
Let I C R. Then : R R/I dened by (x) = x + I for all x R is a homomorphism of R onto
R/I . This is called the natural (or canonical ) homomorphism.
Proposition 1.11.

R, S

Let

be rings and

1.

(0R ) = 0S

2.

(r) = q(r)

3.

K = {x R : q(x) = 0S }

4.

R = {(r) : r R}

for all

:RS

a homomorphism. Then :

rR
is an ideal of

is subring of

Proof. Exercise

is called the kernel of

Denition 1.12.
if

Let

and

is called the (homomorphic) image of

R.

The ideal

is

ker .

sometimes denoted by

be a homomorphism of a ring R into a ring S . Then is called an isomorphism


R and S are isomorphic rings and denote this by R
= S.

is a one to one and onto map. We say that

1.5

The Isomorphism Theorems

R,

Question: Given a ring

what rings can occur as its homomorphic images?

The importance of the rst isomorphism theorem lies in the fact that it shows the answer to lie

R itself. It tells us that if we know all the ideals of R then we know all
R.Only the rst isomorphism theorem contains new information. The

with

the homomorphic images

of

other two are simply its

application.

Theorem 1.13.

Let

be a homomorphism of a ring

S.

into a ring

Then

R
= R/I

where

I = ker

: R/I R by (x + I) = (x) for all x R. The map is well dened since for
x, y R, x + I = y + I x y I = ker (x y) = 0 (x) = (y). is easily seen to be the

Proof. Dened

required isomorphism.

Theorem 1.14.

Let

be an ideal and

Proof. Let

(L + I)/I .

The kernel of

for all

a subring of a ring

be the natural homomorphism

Theorem 1.15.
Proof.

Let

K/I C R/I
x R. The

I, K

and so
map

restricted to

is

be ideals of a ring

(R/I)/(K/I)

ker = K/I .

Then

L/(L I)
= (L + I)/I

R R/I . Restrict to the ring L.


L I . Now apply previous theorem.
R

such that

I K.

is dened. Dene a map

Then

We have

: R/I R/K

x+K =K

xK

x + I K/I

Now apply the rst isomorphism theorem.

L =

(R/I)/(K/I)
= R/K
by

(x + I) = x + K
R/K . Further,

is easily seen to be well dened and a homomorphism onto

(x + I) = K

Therefore

R.

1.6

Direct Sums

Denition 1.16.

The internal direct sum: Let

{I }

be a collection of ideals of a ring

I = {x R : x = x1 + +
sum is the collection of nite sums of elements of the I 's.

dene their sum to be

I 's

We say that the sum of the

x1 + + xk

with

xi Ii .

Proposition 1.17.

xk , xi Ii , k = 1, 2, 3, . . . }.

is direct if each element of

In this case we denote this sum as

The sum

I is direct if and only if

R.

We

That is the

I is uniquely expressible as
I or I1 In if is nite.

P
I ( ,6= I ) = 0

for all

Proof. Exercise

Denition 1.18.

R1 , . . . , Rn be rings. We dene the external direct


n-tuples {(r1 , . . . , rn ) : ri Ri }. On S we dene addition and multiplication
component wise. This makes S a ring. We write S = R1 Rn .
The set (0, . . . , 0, Rj , 0, . . . , 0) is an ideal of S . Clearly S is the internal direct sum of these ideals.
But (0, . . . , Rj , . . . 0)
= Rj . Because of this S can be considered as a ring in which the Rj are ideals
and S is their internal direct sum. Also in internal direct sum we can consider I1 In to be the
external direct sum of the rings Ij . Hence, in practice, we do not need to distinguish between external
sum

The external direct sum : Let

to be the set of all

and internal direct sums.

1.7

Division Rings

Denition 1.19.

Let R be a ring with 1. An element u R is said to be a unit if there exists an


v R such that uv = vu = 1. The element v is called the inverse of u and is denoted by u1
A ring D with at least two elements is called a division ring (or a skew eld ) if D has an identity
and every non-zero element of D has an inverse in D

element

A division ring in which the multiplication is commutative is called a eld-discriminant

Example.

The Quaternions: Let

a0 + a1 i + a2 j + a3 k where ai R.
ai = bi for i = 0, 1, 2, 3. We make the

be the set of all symbols

Two such symbols are considered to be equal if and only if


ring as follows: Addition is component-wise.
the relations

i2 = j 2 = k 2 = 1

non-commutative ring with zero and


Then not all the

ai

Two such symbols are multiplied term by term using

ij = jk = k, jk = kj = i, ki = ik = j .
identity. Let a0 + a1 i + a2 j + a3 k be a non-zero

and

Then

is a

element of

D.

are zero. We have

(a0 + a1 i + a2 j + a3 k)(a0 a1 i a2 j a3 k) = a20 + a21 + a22 + a23 6= 0


n = a20 + a21 + a22 + a23 ,
a0 + a1 i + a2 j + a3 k . Thus D is

(a0 /n) + (a1 /n)i + (a2 /n)j + (a3 /n)k

. So letting

the element

of

a division ring. It is called the division ring of real quaternions.

Rational quaternions can be dened similarly where the coecients are from

1.8

Let

be a ring. A set

R-module

is called a right

if:

is an additive abelian group

2. A law of composition

M RM

3.

(x + y)r1 = xr1 + yr1

4.

x(r1 + r2 ) = xr1 + xr2

5.

x(r1 r2 ) = (xr1 )r2

A left

Q.

Modules

Denition 1.20.
1.

is the inverse

R-module

Example.
2. Let

1.

is dened, which satises for

is dened analogously. Here the product of

and

{0}

are left

R-modules.

be a vector space over a eld

F.

mM

They are also right

Then

of the vector space axioms

is a left

x, y M

and

rR

and

r1 , r2 R

is denoted by

rm.

R-modules.

F -module.

The module axioms are part

3. Any abelian group can be considered a left


Let

g A

and

k Z.

We dened

Z-module:

kg = g + + g
| {z }

if

k > 0, 0Z g = 0A

and

kg = [(k)g]

if

k times

k < 0.
4. Let

be a ring. Then

Mn (R)

r
0

rX = 0
..
.
0
Clearly, we can also make
The symbol

R-module.

MR

will denote

Mn (R)
M

R-module

becomes a left

a right

is a right

0
r
0

0
0
r

.
.
.

.
.
.

..

if we dene for

rR

and

X Mn (R)

0
0

0
X

R-module.

R-module,

while the symbol R M will denote

M is a left
R-modules while dealing with
We say simply say that M is a

For technical reason it is sometimes easier to work with right

non-commutative rings (when we choose to write maps on the left).


module if the other details are clear from the context.

Proposition 1.21.

Let

R-module.

be a right

1.

0M r = 0M

for all

rR

2.

m0R = 0M

for all

m M.

3.

(m)r = m(r) = mr

for all

mM

Then:

and

rR

Proof. Exercise

Denition 1.22.

Let

(or just submodule) if

Proposition 1.23.
and

xr K

for all

K be a subset of a right R-module M . Then K is called a right R-submodule


K is also a right R-module under the laws of composition dened on M .

Let K be a non-empty subset of MR . Then K


x, y K and r R

is a submodule of

M xy K

Proof. Exercise

Denition 1.24.
ideals of

1.9

Submodules of

RR

are called right ideals of

and submodules of R R are called left

R.

Factor Modules and Homomorphisms

R-module M . Consider the facto group M/K . Elements of M/K


m + K with m M . We can make M/K a right R-module by dening
[m + K]r = mr + K for all m M and r R. Check that this action is well dened and the module
axioms are satised to make M/K a right R-module.
Let

be a submodule of a right

are cosets of the form

Proposition 1.25.

Let

1. every submodule of

MR .

Then

A/K

where

be a submodule of

M/K

has the form

is a submodule of

2. There is a one to one correspondence between the submodules of


containing

M/K

and

A K.

and the submodules of

Denition 1.26.

Let

M0

be right

for all

x, y M

and

R-modules.

homomorphism if:
1.

(x + y) = (x) + (y)

2.

(xr) = (x)r

for all

xM

and

rR

A mapping

: M M0

is called an

R-

If
is an

K is a submodule
R-homomorphism

Proposition 1.27.

of
of

Let

MR then the map : M M/K dened by (m) = m + K


M onto M/K . It is called the canonical R-homomorphism

: MR MR0

1.

(0M ) = 0M 0

2.

K = {x M : (x) = 0M 0 }

3.

M = {(m) : m M }

be an

R-homomorphism.

is a submodule of

is a submodule of

for all

mM

Then:

M0

Proof. Exercise

is called the kernel of

if and only if

and

is called the image of

is a one to one correspondence map

ker = 0

Denition 1.28.

Let

: MR MR0

be an

R-homomorphism.

Then

1.10

R-isomorphism
M
= M0

is called an

it is in addition a one to one correspondence and onto map. In this case we write

if

The Isomorphism Theorem

There are similar to those for rings

Theorem 1.29.
M
= M/K

where

Theorem 1.30.

M and M 0
K = ker

Let

Let

L, K

Theorem 1.31. If K, L
(M/K)/(L/K)
= M/L.

be right

R-modules

be submodules of

are submodules of

MR .
MR

and

Then
and

: M M0

and

R-homomorphism.

Then

(L + K)/K
= L/(L K)

K L

then

L/K

is a submodule of

M/K

and

The proofs of these theorems are similar to those for rings

1.11

Direct Sums of Modules

M1 , . . . , Mn be right R-modules. The set of n-tuples {(m1 , . . . , mn ) : mi Mi } becomes a right


+ m0n ) and (m1 , . . . , mn )r =
R-modules if we dene (m1 , . . . , mn ) + (m01 , . . . , m0n ) = (m1 + m01 , . . . , mnP
n
(m1 r, . . . , mn r). This is the external direct sum of the Mi and is denoted i=1 Mi or M1 P
Mn .
Let {M } be a collection of submodules of a right R-modules M . We dene their sum
M
P
to be {m1 + + mk : mi Mi for all possible subsets {1 , . . . , k } of }. Thus
M
is the

set of all nite sums of elements of the M 's. It is easy to see that this is a submodule of M .
P
P
m1 + + mk for
M is said to be direct if each

PM has a unique expression asP


some mi Mi . As in 1.6 we can show that
M
is
direct

P
,6= M } = {0}
P
for all . If
M
is
direct,
we
denote
it
by
M
or
M

M
if is a nite set.

1
n

Let

As explained for rings in 1.6, there is no real dierence between (nite) external and internal direct
sums of modules.

Denition 1.32.

Let

be a ring with 1. A module

MR

is said to be unital if

We shall assume that all modules considered are unital whenever

1.12

for all

mM

is a ring with identity.

Products of subsets

R-module.
Let K, S be non-empty subsets of M and R respectively. We dened their
Pn
{ i=1 ki si |ki K, si S; i = 1, 2, . . . }. Thus KS consists of all possible nite
sums of elements of the type ks with k K and s S . If K is a non-empty subset of M and S is a
right ideal of R then KS is a submodule of M . (Check that we require nite sums in our denition to
Let

m1 = m

be a right

products

KS

to be

make this work)

M = R. Thus if S is a non-empty subset of R


Pn
S 2 = { i=1 si ti : si , ti S; n = 1, 2, . . . }. Extending inductively, S n consist of all nite sums of
elements of the type x1 x2 . . . xn with xi S .
n
Note that if S is a right ideal of R then so is S
The above denition applies, in particular, when

then

1.13

A construction

and M a right R-module. In general, M need not be a right R/I M a right R/I -module structure if M I = 0. In this case we dene
mr = m[r + I] for all m R and r R. It can be checked that this is well-dened right R/I -module
action. Further, under this action the R and R/I submodules of M coincide.
2
In particular, I/I is naturally a right (and left) R-module. This fact will be used repeatedly. In
n
n+1
general same for I /I
.
Let

be a ring with an ideal

module.

1.14

However, we can give

Zorn's Lemma, Well-ordering Principle, The Axiom of Choice

Denition 1.33.
relation

in

1. A non-empty set

(a)

aa

(b)

a b, b c a c

(c)

a b, b a a = b

2. Let
pair

is said to be partially ordered if there exists a binary

which is dened for certain pairs of elements in

S be a partially
a, b we have

ordered set. A non-empty subset


either

ab

or

S be a partially ordered set. An elements x S


y S x = y . Similarly, for a minimal element

in

cS

Zorn's Lemma (Axiom).


subset of

such that

xc

for all

If a partially ordered set

has an upper bound in

A non-empty set
of

is said to be totally ordered if for every

is called a maximal element if

be a totally ordered subset of a partially ordered set


if there exists

S,

and satises:

ba

3. Let

4. Let

then

S.

We say that

xy

with

has an upper bound

x .
S

has the property that every totally ordered

contains a maximal element.

is said to be well-ordered if it is totally ordered and every non-empty subset

has a minimal element.

The Well ordering Principle.


Axiom (The Axiom of Choice).

Any non-empty set can be well-ordered.


Given a class of sets, there exists a choice function, i.e., a function

which assigns to each of these sets one of its elements.


It can be shown that Axiom of Choice is logically equivalent to Zorn's Lemma which is logically
equivalent to the Well-ordering Principle.

10

Chapter 2: The Jacobson Radical

All rings considered in this chapter are assumed to have an identity.

2.1

Quasi-regularity

Denition 2.1.
M0 ) M

Let M be a right ideal


M 0 Cr R M 0 = R .

with

of

R. M

is said to be a maximal right ideal if

M 6= R

and

Similar denition is applied for a maximal two-sided ideal, and maximal left ideal.

Proposition 2.2.
of

I 6= R

Let

be a right ideal of a ring

R.

Then there exists a maximal right ideal

M I.

such that

S be the set of all proper right ideals


R containing I . Partially order S by inclusion. Let {T } be a totally ordered subset of S . Let
T = T . Then T Cr R and T I . Moreover T is proper since T = R 1 T 1 T for
some T = R. Thus T 6= R and so T S . Thus T 6= R and so T S . Now T T for
all . Hence Zorn's Lemma applies and S contains a maximal element, say M . Clearly M is a
maximal right ideal and M I .
c.f. Commutative Algebra, Theorem 1.4. By Zorn's Lemma. Let
of

Corollary 2.3.
Proof. Take

A ring with identity contains a maximal right ideal.

I=0

in the above theorem.

Remark. This is not true for rings without 1

Denition 2.4.

The intersection of all maximal right ideals of a ring

It is usually denoted by

J(R)

(or simply

is called its Jacobson radical.

J)

Remark. Strictly speaking the above denition was for the right Jacobson radical. However we shall
show that this is the same as the left Jacobson radical.

Theorem 2.5

(Crucial Lemma)

K = {r R : ar M }.

Then

1. if

aM

then

K=R

2. if

a
/M

then

Let

K Cr R

be a maximal right ideal of a ring

and let

a R.

Dene

and:

is also a maximal right ideal.

K Cr R, Now assume that a


/ M so that M + aR = R (). Dene an R-module
: R R/M by r 7 ar + M r R. Check that this is a homomorphism of right
R-modules. By (), is an onto map. So by the isomorphism theorem for modules: R/M
= R/ ker =
R/K . It follows that K is a maximal right ideal.
Proof. Clear that
homomorphism

Theorem 2.6.

J CR

J Cr R. Now let j J and a R and suppose aj


/ J . Then by denition there exists
M such that aj
/ M . Dene K = {r R : ar M }. By the previous theorem K is a
maximal right ideal. But j
/ K since aj
/ M hence j
/ J . This is a contradiction. Hence aj J for
all j J and r R. Thus J C R.
Proof. Clearly

a right ideal

Denition 2.7.

Let

A subset

of

R. We say
(1 x)y = 1

be an element of a ring

has a right inverse, i.e., if

y R

such that

that

is a right quasi-regular (rqr) if

is called right quasi-regular if every elements of

1x

is rqr

Left quasi-regular (lqr) is dened analogously


We call an element or set quasi-regular if it is both lqr and rqr.

Lemma 2.8.

Let

be a rqr right ideal of

R.

Then

M be a maximal right ideal of R. If I * M then I + M = R, so 1 = x + m for some x I


m M . Hence 1 x M , now there exits y R such that (1 x)y = 1, so 1 M hence M = R.
contradiction, thus I J as required.

Proof. Let
and
A

IJ

11

Lemma 2.9.
Proof. Let

Let

j J.

be a ring,

J(R)

Suppose that

is a right quasi-regular ideal.

1 j has
M such

there exists a maximal right ideal

1 = 1 j + j M,
So J is a rqr.

Lemma 2.10.

hence

Let

M = R.

(1 j)R M .

This is a contradiction, hence

be an ideal of a ring

(1 j)R 6= R so by Theorem 2.2


j M by denition of J(R) so
1 j has a right inverse for all j J .

no right inverse. Then


that

R.

Then

But

rqr if and only if

lqr.

Proof. Suppose that I is rqr. Let x I , then there exists a R such that (1 x)(1 a) = 1. So
a = xa x I since I Cr R. Hence there exists t R such that (1 a)(1 t) = 1, so 1 x = (1 x)1 =
(1 x)(1 a)(1 t) = 1(1 t) = 1 t. Hence (1 a)(1 x) = 1, thus x is lqr. By symmetry we can
run the converse argument.

Theorem 2.11.

The (right) Jacobson radical is a qr ideal and contains all the rqr right ideals.

Proof. This is what we have proved above.

Corollary 2.12.
Jl

The Jacobson radical of a ring

is equal to the right Jacobson radical

Jl
Jr = Jl .
Proof.

is left right symmetric, i.e., left Jacobson radical

is a qr ideal by the left hand version of the theorem, so

Theorem 2.13.

Let

be a ring with Jacobson radical

Proof. The maximal right ideals of the right

Jr

is a maximal right ideal of

R/J

J.

Then

Jl Jr .

Similarly

Jr Jl ,

hence

M/J

where

J(R/J) = 0

are precisely the right ideals of the form

Remark. The theory can be adjusted to deal with rings without an identity.

2.2

Commutative Local Rings

Denition 2.14.

Let

R be a commutative ring, R is said to be a local

ring if

R has a unique maximal

ideal

Note. This terminology is slightly dierent from Kaplansky's


Let

be a commutative local ring with 1. Let

1.

2.

R/M

is the Jacobson radical of

be the maximal ideal of

R,

then:

is a eld

x R, x
/ M then x is a unit of R.


Example. Let R = ab |a, b Z, bodd
Check that R is a local ring. Find its unique
localised at the prime ideal 2Z
3. If

maximal ideal.

In fact

R = Z(2) ,

i.e., the ring

Remark. There exists a non-commutative ring with unique maximal ideal (in fact the only proper
non-zero ideal) which is not its Jacobson radical.

12

Chapter 3: Chain conditions


1

Rings need not have

3.1

in this chapter

Finitely Generated Modules

Denition 3.1.
submodule of

Let

T be

MR . The smallest submodule of M containing T is called


T , i.e., it is the intersection of all submodules of M containing T .
{0} to be the submodule generated by the empty set .

a subset of

the

generated by

By convention we take

T consists of a singles element a M . In general the


{ar + a|r R, Z}. This equals aR when R has 1 and M is unital.

Of particular importance is the case when


submodule generated by

is

Denition 3.2.

If MR is generated by a single element then we say that M is a cyclic module


R-module M is said to be nitely generated (f.g.) if it is the module generated by a
nite subset. If R has 1 and M is a nitely generated module then a1 , . . . , an M such that
M = a1 R + + an R.
Cyclic submodules of RR [R R] are called principle right (left ) ideals.
A right

3.2

Finiteness Assumption

Denition 3.3.
1. An element

Let

be a non-empty collection of submodules of a right

K S is

said to be maximal in

Similarly for a minimal element of


2.

if

@K 0 S

such that

K0 ) K.

is said to have the ascending chain condition (ACC) for submodules in

submodules
3.

R-module M .

A1 A2 . . .

with

Ai S

is said to have the maximum condition on submodules in

submodules in

if every chain of

has equal terms after a nite number of terms.

if every non-empty collection of

has a submodules maximal in this collection.

The descending chain condition (DCC) and minimum condition are dened analogously.

Proposition 3.4.

Let

be a non-empty collection of submodules of

MR

then the following are

equivalent:

1.

has ACC [DCC] on submodules in

2.

has the maximum [minimum] condition on submodules in

Proof. Exercise
Particularly important is the case when
 M has ACC will mean that

consists of all submodules in

has ACC on the set of all submodules of

M.

MR .

The abbreviation

Similarly for the other

conditions.

Proposition 3.5.

The following are equivalent for a right

1.

has ACC

2.

has the maximal condition

3. Every submodule of

R-module M .

is nitely generated.

Proof. This is Commutative Algebra Proposition 5.1

Example.

ZZ

Remark.

1. ACC does not imply the existence of an integer

has ACC since every ideal is principle (this follows from the Euclidean Algorithm)

This is easily checked on

n such that all chains stop after n steps.

2. Similarly with DCC. Examples are harder but they do exists.

13

MR

3. However if

has both ACC and DCC then such an integer does exists. This follows from the

theory of composition series.

Lemma 3.6

(Dedekind Modular Law)

Let

A, B, C

be submodules of

MR

such that

A B.

Then

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

Proposition 3.7

(Commutative Algebra 5.4)

ACC [DCC] if and only if both

and

M/K

Suppose that

is a submodule of

MR .

Then

has

have ACC [DCC]

: Straightforward
: Let M1 M2 . . . be an ascending chain of submodules of M . Consider the chains M1
K M2 K . . . and M1 + K M2 + K . . . . The rst chain stops since it consists of
submodules of K . So there exists k 1 such that Mk K = Mk+i K for all i 1. The second
chain stops since it consists of submodules of M which are in 1 to 1 correspondence with those of
M/K . So there exists an l such that Ml + K = Ml+i + K for all i 1. Let n = max{k, l}. Then
Mn+i = Mn+i (Mn+i + K) = Mn+i (Mn + K) = Mn + (Mn+i K) by the Modular Law (since
Mn+i Mn ). And Mn + (Mn+i + K) = Mn + Mn K = Mn , and this is true i 1. So MR has
Proof.

ACC
Similarly for DCC
This important proposition has many consequences

Corollary 3.8 (Commutative Algebra 5.5 ).


Mi

If each

Let

M1 , . . . , Mn be submodules
M 1 + + Mn = K .

of a right

R-modules M .

has ACC [DDC] then so does their sum

M1 +M2
K1
2
Proof. Take K1 = M1 + M2 . We have K1 /M1 =
. So
= M1MM
M1
M1 has ACC [DCC] since
2
M2
M1 M2 is a factor modules of M2 and M2 has ACC. Also M1 has by assumption ACC [DCC]. So by
the proposition 3.7, K1 has ACC [DCC].
This can easily be extended by induction.

Corollary 3.9.

Let

1. Suppose that R
R-module. Then MR has

be a ring with

a nitely generated unital right

has ACC [DCC] on right ideals. Let

MR

be

ACC [DCC] on submodules.

MR is nitely generated and unital, there exists m1 , . . . , mk such that M = m1 R +


. . . mk R. So by Corollary 3.8 it is enough to show that each mi R has ACC [DCC]. The map r mi r
for all r R is an R-homomorphism of RR onto mi R. So mi R is isomorphic to a factor of RR . So
mi R has ACC [DCC] on submodules.
Proof. Since

Remark. If

1, the ACC part of the corollary still holds but the


(mi ) = {mi r + mi |r R, Z} and Z has ACC but not DCC

does not have

This is because

Denition 3.10.

DCC part is false!

A modules with ACC on submodules is called a Noetherian module.

A modules

with DCC on submodules is called an Artinian module


A ring with ACC on right ideals is called a right Noetherian ring. A ring with ACC on left ideals
is called a left Noetherian ring.
A ring with

and DCC on right ideals is called a right Artinian ring. A ring with

and DCC on

left ideals is called a left Artinian ring.

3.3

Nil and Nilpotent Ideals

Denition 3.11.
exists an integer
an integer
If

k1

Let

k1

be non-empty subset of a ring

(which depends on

such that

s)

such that

R. S is said to be nil if given any s S there


sk = 0. S is said to be nilpotent if there exists

Sk = 0

consists of a single element, there is no dierence between nil and nilpotent and we normally

say that the element is nilpotent.

Proposition 3.12.

Let

be a ring with

1.

Every nil one sided ideal of

14

is inside

J(R).

I be a nil right ideal and x I . Then xk = 0


) = 1 so x is r.q.r. so x J(R). Thus I J(R).

Proof. Let
k1

+ x

Remark. This is also true without

Lemma 3.13.
I

1. If

and

Let

k 1.

for some

We have

(1 x)(1 + x +

1.

be a ring:

I +K

are nilpotent right ideals then so are

and

RI

2. Every nilpotent right ideal lies inside a nilpotent ideal.

I k = 0 and K l = 0, k, l 1. Then (I + K)k+l1 = 0 since every term in the


k
l
k
expansion lies in either I or K and hence is zero. So I + K is nilpotent. (RI) = (RI)(RI) . . . (RI)
k1
k
R(IR)
I RI = 0. So RI is nilpotent.
Suppose that I is a nilpotent right ideal. Then I I + RI . Now I + RI C R and is nilpotent by

Proof. Suppose that

the rst part.

Denition 3.14.

The sum of all nilpotent ideals of

radical). It is usually denoted by

Note.

N (R) J(R)

(or the Wedderburn

always.

It follows from Lemma 3.13 that

N (R)

R is called the Nilpotent radical

N (R).

N (R) =

nilpotent right ideals

nilpotent left ideals. Clearly

is a nil ideal. It is in general not itself nilpotent.

Example

. Let F be a eld, I the open interval (0, 1) and R a vector space


{x
|i

I}
. Dene a multiplication on F by extending the following product of basis
i
(
xi+j if i + j < 1
elements xi xj =
. Thus every element of R can be written uniquely in the form
0
if i + j 1
P
iI ai xi where ai F and ai = 0 for all except a nite number of i. Check that N (R) = R but R is
over

(Zassenhaus's Example)

with basis

not nilpotent.

Proposition 3.15.
of

Let

be a commutative ring. Then

N (R)

equals the set of all nilpotent elements

R.

Proof. Let

be a nilpotent element. This implies that the principle ideal generated by

is nilpotent.

(Prove!)

Example.
Q.

Then

has only two ideals

Denition 3.16.
AP

or

An ideal

B P.

We exclude

and

R.

of a ring

So

N (R) = 0

is said to be a prime ideal if

Let R be a commutative ring


ab P a P or b P .

we have

Proof. Trivial if

has

1.

P C R.

and

Not so trivial but still true if

Proposition 3.18 (Commutative Algebra 1.10 ).


of

but


0
0

ring of

22

matrices over

AB P , A, B C R

implies

itself from the set of prime ideals.

Proposition 3.17.
(a, b R)

R be the
2
1
= 0.
0

The above is false for non-commutative rings. e.g, let

Let

Then

is a prime ideal if and only if

does not have

1.

be a ring. The intersection of all prime ideals

is a nil ideal.

x R is not nilpotent then there exists a prime ideal excluding it.


x R is not nilpotent. Let S be the set of ideals which contains no power of x. S 6= 0
since {0} S . Check that Zorn's lemma applies. So S contains a maximal element, say P . Claim:
P is a prime ideal. If not then ideals A and B of R such that AB P but A * P and B P .
k
l
Then A + P ) P and B + P ) P . So x A + P and x B + P for some integers k, l. But
k+l
then x
(A + P )(B + P ) P which is a contradiction. Thus P is a prime ideal proving the
Proof. We shall show that if
Suppose that

proposition.

Corollary 3.19.

In a commutative ring

N (R)

equals the intersection of all prime ideals of

15

R.

Proof. This follows from Theorem 3.15 and the previous theorem.

Corollary 3.20.
when

Proof. Let

In a commutative ring with

N (R)

is Noetherian

a nitely generated nil ideal is nilpotent. In particular

is nilpotent.

R.

be a nitely generated ideal of

Let

K = k1 R + + ks R with ki K . Each ki is
R is Noetherian N (R) is nitely

nilpotent hence so is the ideal. The result follows by 3.13. When


generated and so nilpotent by above.

3.4

Nakayama's Lemma and an Application

Denition 3.21.
1.

a1 , . . . , an

Let

I Cr R.

generate

We say that

{a1 , . . . , an }

Nakayama's Lemma.
J(R)

Let R be a
M I = M M = 0.

Then

is minimal generated set for

if:

2. No proper subset of

of

a1 , . . . , an

generates

ring with

I.

and

MR

a nitely generated module. Let

be a subset

M I = M . Then we have M J = M . Suppose that M 6= 0. Let a1 , . . . , an be a minimal


M . We have M = a1 R+ +an R so that M J = a1 J + +an J . Now a1 M = M J
so a1 = a1 x + + an xn for some xi J . Now a1 (1 x) = a2 x2 + + an xn (a1 (1 x1 ) = 0 if n = 1).
1
So a1 = a2 x2 (1 x1 )
+ + an xn (1 x1 )1 (a1 = 0 if n = 1). This contradicts the minimality of
n. Hence M = 0

Proof. Let

generated set for

Remark. This is also valid for rings without

Let

J/J 2 is
x + J 2.

be a commutative local ring with

R/J -module,
2
So x R/J .

an

Lemma 3.22
(as an

J/J 2

with unique maximal ideal

is a vector space over the eld

R/J .

If

J . Then R/J is a eld. So


x R let x denote the coset

. Let R be a commutative local ring with 1. Let J be the


J is nitely generated and x1 , . . . , xk J . Then x1 , . . . , xk generate
x1 , . . . , xk is a set which spans the vector space J/J 2 (over the eld R/J )

(Commutative Algebra 2.17)

maximal ideal of

i.e.,

1.
1

R.

Suppose that

R-module)

J/J 2 as an R-module so x1 , . . . , xk generate J/J 2 as an R/J -module,


J/J 2 .
) Let I = x1 R + + xk R. Then I J , x1 , . . . , xk generates J/J 2 as an R-module, hence
I + J 2 = J . This implies that (J/I)J = J/I where J/I is considered as an R-module. So J/I = 0 by
Nakayama's lemma, so J I . Hence J = x1 R + + xk R.
Proof.

) x1 , . . . , xk

generate

i.e., they span the vector space

Corollary 3.23.

In the above ring x1 , . . . , , xk is a minimal generated set for


J/J 2 over R/J .

J x1 , . . . , xk

is a

basis for the vector space

Proof. Follows from above

Theorem 3.24.
R.

Let

be a commutative Noetherian local ring with

Then any two minimal generating set of

1.

Let

be the maximal ideal of

contain the same number of elements.

Proof. This is a direct consequence of the corollary


Notation. We shall denote this common number by
over the eld

R/J .

16

V (R).

Thus

V (R) = dim J/J 2

as a vector space

Commutative Noetherian Rings

All rings considered in this chapter are assumed to be commutative rings

4.1

1.

Primary Decomposition

Denition 4.1.
for some integer

An ideal

is said to be primary if

ab Q (a, b R)

implies that

aQ

or

bn Q

n.

Clearly a prime ideal is primary.

Denition 4.2.

is called a primary ring if

Clearly an ideal

Denition 4.3.

is primary if and only if

We say that

is a primary ideal.

R/Q

is a primary ring.

has primary decomposition if every ideal of

is expressible as a nite

intersection of primary ideals.

Denition 4.4.

An ideal is said to be meet-irreducible if

Note. The two dierent denitions:

I =AB

is meet-irreducible if

Lemma 4.5

I = A B , A, B C R implies I = A or I = B .

MR is irreducible if {0}
I = A or I = B

and

are the only submodules.

I CR

implies

(Commutative Algebra 6.18)

Let

be a Noetherian ring.

Then every ideal of

is

expressible as a nite intersection of meet-irreducible ideals.

A C R be a maximal counterexample. Then A is not meet-irreducible. So


A = B C , B, C C R, B ) A, C ) A. By maximality of A, both B and C are nite intersection of
meet-irreducible ideals. Hence so is A. Contradiction hence the result holds.

Proof. Suppose not. Let

Notation. Let

MR . The annihilator
ann(S) Cr R. If S is a submodule

be a subset of

is non-commutative

of

in

is

ann(S) = {r R|Sr = 0}.


S is a subset of R.

For

then typically

Theorem 4.6 ((Noether) Commutative Algebra 6.20).

A Noetherian ring has primary decomposition

Proof. By the previous lemma it is enough to show that a meet-irreducible ideal is primary. Without

ab = 0, a, b R.
n 1 such that bn R ann(bn ) = 0.
2
n
2n
Since the chain ann(b) ann(b ) . . . stops there is an integer n 1 such that ann(b ) = ann(b ).
n
n
n
z
2n
n
Now z b R ann(b ) x = b t for some t R and b = 0. So b t = 0 b t = 0 z = 0. Since 0
n
n
n
is meet-irreducible either b R = 0 or ann(b ) = 0. Thus b = 0 or a = 0 and 0 is a primary ideal
loss of generality assume

to be meet-irreducible. Suppose that

Claim: There exists an integer

Denition 4.7.

Q be
Q and

Let

called the radical of

P/Q be
Q is P -primary.

Q by Q.

a primary ideal. Let

Notation. We denote the radical of

Recall that for a commutative ring

Proposition 4.8.
1.

R/Q. P

is

R, N (R) =set

be a primary ideal and let

of all nilpotent elements of

P =

Q.

R.

Then:

is a prime ideal

2. If further
Proof.

Let

the nilpotent radical of the ring

we say that

1. Let

/Q
a

so

is Noetherian, then

Pk Q

for some

k 1.

ab P with a, b R. Then (ab)n Q for some n 1 so an bn Q. If a


/ P then
(bn )s Q for some s 1 by denition of primary. Hence b P . Thus P is a prime

ideal/
2.

P/Q

is a nil ideal of

R/Q.

reference maybe)). Hence

If R/Q is Noetherian, P/Q


P k Q for some k 1.

Theorem 4.9 (Commutative Algebra 6.24).


0

where

Let

is nilpotent (by Proposition 3.13 ?(check

R be a commutative Noetherian ring.

J = J(R).
17

Then

n=1 J =

n
X =
n=1 J . Let XJ = Q1 Qn be a primary decomposition for X . Fix i and let
Pi = Qi , if X * Qi then J Pi . So J ki Qi for some ki 1 by the previous proposition. Thus
X Qi or J ki Qi . So X Qi for all i = 1 . . . , n in any case. Hence X XJ . So X = XJ hence
by Nakayama's lemma X = 0.

Proof. Let

This is a surprisingly useful result.

Remark. For a right Noetherian ring this is false (Proven by Herstein in 1965). While for left and right
Noetherian rings the result is still an open problem.

Denition 4.10.

A ring is called an integral domain if the product of any two non-zero elements of

the ring is non-zero.

Theorem 4.11.

Let

be a commutative, local, Noetherian ring. Suppose that

ideal. Then every non-zero ideal of

is a power of

J.

In particular,

J = J(R)

is a principle

is a principal ideal ring.

n
0 6= I C R, I 6= R. Then I J . Since
n=1 J = 0 there exists an integer k 1 such that
k+1
m
I J but I * J
. Let J = aR (a J ), then J
= am R m 1. Now there exists an element x
k+1
k
such that x I but x
/a
R (). Since x a R we have x = ak t for some t R. Now t
/ J = aR
k
1
by (). So t must be a unit of R. So a = xt
I . Hence J k = ak R I . It follows that I = J k

Proof. Let
k

proving the theorem.

Corollary 4.12.

Let

be a commutative, local, Noetherian ring.

1. If

is not nilpotent then

2. If

is nilpotent then

is Artinian and

Proof. Exercise. (Note that in 2.

4.2

Decomposition of

Denition 4.13.
are

Pi -primary.

1. No
2.

Qi

Let

is an integral domain and

and

are the only prime ideals of

is the only prime ideal of

R.

R.

J s = 0 for some s 1 so R, J, J 2 , . . . , J s = 0 are the only ideals.

I = Q1 Qn

be a primary decomposition for an ideal

I.

Suppose that

Qi

We say the decomposition is normal [Commutative Algebra: minimal] if

is superuous

Pi 6= Pj

for all

Given that

i 6= j

has a primary decomposition, we can arrange a normal decomposition for

by:

1. Removing any superuous primary ideals and


2. By applying the following:

Lemma 4.14.

If

Q1

and

Q2

are

P -primary

ideals then so is

Q1 Q2

ab Q1 Q2 , a, b R. If a
/ Q1 Q2 then a
/ Q1 say. Then bn Q1 for some n 1. So
s
b P . Hence b Q2 for some s 1 since Q2 is P -primary. Let k =
max(n, s) then bk Q1 Q2 .
t
Q1 Q2 . But Q1 Q2 Q1
Now p P implies p Q1 Q2 for suciently large t 1. Hence P

so
Q1 Q2 Q1 = P , thus P = Q1 Q2 .
Proof. Let

Thus whenever necessary we shall assume that the primary decomposition being considered is normal.

Remark. We may still have

Qi )

Qj

with a normal primary decomposition [Commutative Algebra,

example before 6.8]

Denition 4.15.
QP

with

Lemma 4.16.

Let

be a ring.

prime implies

We say that a prime ideal

is a minimal prime ideal of

if

R be a commutative
Noetherian ring. Suppose that 0 = Q1 Qn be a primary

0. Let Pi = Qi and suppose (after possible renumbering) that P1 , . . . , Pk are minimal


{P1 , . . . , Pn }. Then P1 , . . . , Pk are precisely the minimal primes of R.
Let

decomposition of
in the set

Q = P.

18

R then P Pj for some 1 j k . By


ki 1 such that Piki Qi for i = 1, . . . , n. Then
P1k1 P2k2 . . . Pnkn Q1 Qn = 0. In particular, P1k1 . . . Pnkn P hence Pm P for some m with
1 m n. But since P1 , . . . , Pk are minimal in the set {P1 , . . . , Pn } we have Pj Pm for some j ,
1 j m. Thus P Pj with 1 j m as required.

Proof. It is enough to show that if

is a prime ideal of

Theorem 4.6 (? check reference) there exists integers

Denition 4.17.

Let

c R,

we say that

is regular if

cx = 0, x R x = 0

An element which is not regular is called a zero-divisor.

Notation. Let

I C R.

Write

C (I) = {x R|x + I

Clearly C (0) = {regular


C (P ) = R \ P .

Proposition
4.18.
Let

Qi

Pi =

Let

R}.

If

R/I }

is a prime ideal, in a commutative ring then

R be a Noetherian ring and 0 = Q1 Qn a normal primary decomposition.


P1 , . . . , Pk are the minimal primes of R. Then:

and suppose that

1.

N (R) = P1 Pk .

2.

C (0) = R \ ni=1 Pi

3.

C (N ) = R \ ki=1 Pi

Proof.

elements of

is regular in the ring

1. Clearly

N P1 Pk .

Now P1 Pk Pj for all 1 j n. By Proposition 4.8


(P1 Pk )ti Qi . Let t = max{ti }, then (P1 Pk )t
P1 Pk N and so P! Pk = N .

there exists an integer ti such that

Q1 Qn = 0.

Thus

c R \ ni=1 Pi . Then cx = 0, x R x Qi
Hence x Q1 Qn = 0, so c C (0).

2. Let

for all

i 1 i n,

since

Pi .

belong to no

Pini Qi for some ni . So Pini [Q1 Qi1 Qi+1 Qn ] Q1 Qn = 0.


Now Q1 Qi1 Qi+1 Qn 6= 0 since our decomposition is normal. So Pi is does
n
not contain a regular elements and hence i=1 Pi does not contain a regular element. Hence
n
C (0) = R \ i=1 Pi
Now

3. Exercise

Lemma 4.19.
prime. Let

Let

Remark. Note that

P1 , . . . , Pn be ideals of R, at least n 2 of which


S ni=1 Pi , then S Pk for some k , 1 k n.

be a commutative ring. Let

be a subring of

R.

Suppose that

does not (necessarily) contain

1,

are

since our denition of rings did not include

n. For n = 1, result is trivial.


n = 2 if S * P1 and S * P2 then choose x1 , x2 S such that x1
/ P2 and x2
/ P1 . Then
x1 + x2 S but x1 + x2
/ Pi , i = 1, 2.
Now assume n > 2 and that the result holds for values < n.
n
Clearly any selection of n 1 of the Pi at most 2 will be non-prime. Suppose that S i=1 Pi but
S
/ Pi for any i (i = 1, 2, . . . , n). Then S * P1 Pk1 Pk+1 Pn by induction hypothesis
(as k varies). Now choose xk S such that xk
/ P1 Pk1 Pk+1 Pn . Thus xk Pk .
Since n > 2 at least of the Pi must be prime, say P1 . Let y = x1 + x2 . . . xn , then y
/ Pi for any
i = 1, . . . , n. This is a contradiction. This completes the induction.
Proof. Proof by induction on
For

Proposition 4.20.

R be a commutative
ann I = 0.

Let

element if and only if

Noetherian ring. Let

I C R,

then

contains a regular

: Trivial
: Suppose that every element of I is a zero divisor. Then by the Proposition 4.18 part 2)
I ni=1 Pi (where the Pi are as in Proposition 4.18. So I Pj , for some j , 1 j n. We have
ann I ann Pj 6= 0. This completes the proof.
Proof.

Proposition 4.21.

R be a commutative Noetherian
I = c1 R + + cn R where each ci

Let

regular element. Then

19

ring and

I C R.

is regular.

Suppose that

contains a

I . So I \ K is either empty or
P1 , . . . , Pn be the primes associated with a primary decomposition of 0
(as in Proposition 4.18). So I \ K P1 Pn by Proposition 4.18 part 2, so I K P1 Pn .
Hence I K or I Pi for some i (by Lemma 4.19). But I * Pi for any i since I contains a regular
element but all Pi contains zero-divisors. Hence I K and so I = K . Since R is Noetherian it follows
Proof. Let

be the right ideal generated by the regular elements in

consists of zero divisors. Let

that we can nd a nite generating set consisting of regular elements.

4.3

Localisation [Commutative Algebra Section 3]

Denition 4.22.
if

Let S
s1 , s2 S s1 s2 S .

be a non-empty subset of a ring

R.

We say that

is multiplicatively closed

C (P ) = R \ P where P is a prime ideal in a commutative ring. We shall always


1 S.
0
Dene an equivalence relation on R S as follows: (a, s) (b, t) if there exists s S such that
0
(at bs)s = 0 (where (a, s), (b, t) R S )
a
Let
c be the equivalence class of (a, b) and let RS denote the set of all such equivalence classes.
a b
a
b
at+bs
a
b
ab
For
,
and
s t RS dene s + t =
st
s t = st .
Check that this is well-dened and that RS is a ring. We have a natural ring homomorphism
: R RS given by (r) = 1r for all r R
Typical example:

assume

0
/S

and

Denition 4.23.

RS constructed above is called a localizations of R at S


A, B be rings with 1 and : A B a homomorphism of rings. In this
assume (1A ) = 1B
Let

context we shall always

The Universal Mapping Property.


A


Bo

Let

A, B

AS

S a multiplicatively closed subset of A. Suppose that : A B is a ring homo(s) is a unit in B for all s S . Then there exists a unique ring homomorphism
that =

be rings and

morphism such that

: AS B

such

Proof. See Commutative Algebra 3.2-point


The ring homomorphism
1.

sS

2. Given

implies

(s)

is a unit in

a R, (a) = 0

3. Every element of

: R RS

RS

Let

A, B

RS

if and only if

as = 0

is expressible as

These three properties determine

Theorem 4.24.

has the following properties:

RS

for some

(a)[(s)]1

sS

for some

a R, s S .

to within isomorphism.

be rings and

a multiplicatively closed subset of

A.

Suppose that

:AB

is a ring homomorphism such that:


1.

sS

2.

(a) = 0

implies

(s)

implies

3. Every element of

is a unit of

as = 0
B

for some

sS

is expressible as

Then there exists a unique isomorphism

(a)[(s)]1

: AS B

for some

such that

A AS .
A


Bo

20

AS

a A, s S .
= ,

where

is the natural map

: AS B

Proof. By the universal mapping property there is a unique homomorphism


1
1

= ,

where

check that

(as

is given by

) = (a)[(s)]

such that

(used property 1.) Then use property 2 and 3 to

is an isomorphism.

R at S . Also since as = a1 1s we usually write as1


a
rather than
s for elements of RS .
Particularly important is the case when elements of S are regular, in this case the natural map
In view of this we speak of the localization of

R RS
and

RS , we write r instead of
S = R \ {0} then RS is just the

Lemma 4.25.
1. if
2.

R with its image in RS . Thus we may assume that R is a


r
for elements of R. In particular when R is an integral domain
1
eld of fractions of R.

is a monomorphism. We identity

subring of

Let

be a ring and

I C R IRS C RS

a multiplicatively closed subset such that

and every element of

K C RS K R C R

and

IRS

is expressible as

xd1

S C (0).

for some

xI

Then:
and

d S.

(K R)RS = K .

1
Proof. We are assuming that R is a subring of RS . So a typical element of IRS is x1 r1 c1 + +
1
xn rn cn for some xi I, ri R and ci S . Let d = c1 c2 . . . cn and di = c1 c2 . . . ci1 ci+1 . . . cn then
1
1
x1 r1 c1
= xd1 where x = x1 r1 d1 + + xn rn dn I .
1 + + xn rn cn = (x1 r1 d1 + + xn rn dn )d
The rest is an exercise.

Remark. If

ICR

we have

IRS R I

but we do not have equality in general.

E.g.

R = Z

and

R S = Q.
However, see Lemma 4.27 part 2 below.

Corollary 4.26.

If

is a Noetherian ring then so is the ring

RS .

Proof. Clear from the previous lemma (part 2)

Lemma 4.27.

Let

be a ring and

a multiplicatively closes subset. Suppose that the elements of

are regular. Then


1. If

is a prime ideal of

RS

2. If

is a prime ideal of

Proof.

then

and

is a prime ideal of

P S =

then

P RS

is a prime ideal of

RS

and

P RS R = P

1. Easy

P RS R = P . Clearly P RS R P . Let z P RS R, then


z = ps1 for some p P and s S Lemma 4.25 part 1. So zs = p P with z, s R. Now z P
since s
/ P and P is prime. Thus P RS R = P . Now let P RS with , RS . Then
= ac1 and = bd1 where a, b R, c, d S . So abc1 d1 P RS hence ab P RS R = P .
So P RS or P RS , hence P RS is a prime ideal of RS . (Note: P RS 6= RS since P 6= R)

2. We shall rst need to show that

Theorem 4.28.

Let

R, S

between the prime ideals of

be as above. Then there is a one to one order preserving correspondence

which do not intersect

and the prime ideals of

Proof. This follows from the previous lemma. The correspondence is

RS

P P RS .

Remark. Theorems analogous to the above hold even when the elements of

are not assumed to be

regular.

Notation. Of special importance is the case when


case it is customary to write

Proposition 4.29.
regular. Then

P RP

Let

RP

instead of

RC (P )

is a prime ideal and

or

RR\P .

be a prime ideal of a ring

is the unique maximal ideal of

RP

S = R \ P = C (P ).

and suppose that the elements of

and thus

RP

In this

C (P )

are

is a local ring.

I C RP , I 6= RP . Then I does not contain a unit of RP . [I R] C (P ) = , i.e., I R P .


I = (I R)RP P RP , since P C (P ) = , P RP 6= RP . It follows that P RP is the unique
maximal ideal of RP .
Proof. Let

So

Remark. Hence the name localization

Example.

R = Z, P = 2Z,

then

Z(2) =

a

b |a, b


Z, b odd
21

4.4

Localisation of a Module [Commutative Algebra 3.1]

Let

R-module and S a multiplicatively closed subset of R such that 0


/ S , 1 S . Dene an
M S as follows: (m, s) (m0 , s0 ) if there exists t S such that (ms0 m0 s)t =
that is an equivalence relation. Denote equivalence class of (m, s) by m/s. Let MS be the

be an

equivalence relation on

0.

Check

collection of all such equivalence classes. Dene

ms0 + m0 s m r
mr
m m0
,
+
=
=
, m, m0 M, s, s0 , t S, r R
s
s
ss0
s t
st
MS into an RS -module. Uniqueness corresponding to Theorem 4.24 can also be
MS the localization of M at S .
Note that if A is an RS -module then A can be considered an R-module via the action a r =
a 1r a A, r R. In this case A
= AS as RS -module [Check that ac a 1c is an isomorphism
AS S ]
Check that this turns
proved. We call

4.5

Symbolic Powers

Let

be a prime ideal. Then the powers of

need not be

P -primary [Commutative Algebra Example

after 6.3]

P (n) = {x R|xc P n

Denition 4.30.
Clearly

P (n)

P (1) = P

Lemma 4.31.

P (n)

is called the

and

is

for some

P (n) P

c C (P }.
nnt

Check that

symbolic power of

for all

P (n) C R.
P

n.

P -primary

ab P (n) , a, b R. Then abc P n for some c C (P ). If no power of b lies in P (n) then


b
/ P , i.e., b C
(P ), We have a(bc) P n with bc C (P ). Hence a P (n) and P (n) is primary. It is
P (n) = P
easy to see that
Proof. Let

Lemma 4.32.

Let

be a prime ideal and suppose that elements of

C (P )

are regular. Then fro every

n 1:
1.

(P RP )n = P n RP

2.

P n RP R = P (n)

3.

P (n) RP = P n RP

Proof.
2.

1.

n
= P n RP
(P RP )n = P n RP

x P (n) xc P n for some c C (P ).


RP . Hence x P n RP R.

So

xcRP P n RP xRP P n RP

q P n RP R q = pc1 with p P n and c C (P ).


(n)
and noting that q R, we haveP
= P n RP R

Conversely:

qP

(n)

since

Hence

is a unit of

qc = p P n ,

so

3. Exercise

4.6

The Rank of a Prime Ideal

Denition 4.33.

A prime ideal P is said to have rank r (or height r ) if there exists a chain of prime
P1 ( P2 ( ( Pr ( P but none longer. If there does not exists a maximal nite chain of
primes then we say rk P = . If P contains no other primes, we dene rk P = 0
ideals

Note that

rk P = 0

Denition 4.34.
+ an R)

Let

if and only if

a1 , . . . , a n R ,

is a minimal prime.

P is minimal
R/(a1 R + + an R).

we say that prime

is a minimal prime of the ring

22

over

a1 , . . . , an

if

P/(a1 R +

Lemma 4.35.

R be a
n 1.

Let

is Artinian for all

Noetherian ring,

A C R.

Suppose that

R/A

is an Artinian ring. Then

R/An

R/A2
2
A/A2 (by the third isomorphism theorem). Note A/A is nitely generated as an R/A2
2
module, so by Corollary 3.9 A/A is Artinian. Since R/A and A/A are Artinian, it follows from
2
Proposition 3.7 that R/A is Artinian. The proof then extends by induction.

R/A
=

Proof.

Krull's Principal Ideal Theorem.


that

is a prime ideal minimal over

Let

a.

be a Noetherian Ring. Let

Then

Proof. We shall rst deal with the case when


local ring with Jacobson radical

P.

be a non-unit, suppose

R,

is the unique maximal ideal of

Suppose we have

Q1 Q ( P .

i.e., when

Factoring out by

Q1

is a

we may

R/aR, P/aR is both the


P/aR = N (R/aR).
n
By Proposition 3.20(Check this reference) there exists an integer n 1n such that P aR.
n
Now R/P is a eld so by Lemma 4.35 R/P
is Artinian. Hence R/aR is an Artinian ring. Hence
(k)
there exists k 1 such that Q
+ aR = Q(k+1) + aR. So Q(k) Q(k+1) + aR. Let x Q(k) , then
x = y + at for some y Q(k+1) , t R. Hence at = x y Q(k) . Now a
/ Q since P is minimal over a.
(k)
(k)
(k)
(k)
(k+1)
(k)
So t Q
, thus Q
hQ(k+1)i+ aQ
. Hence Q
=
Q
+
aQ
(since the other containment
h (k) i
Q
Q(k)
=
aR
where
[
]
is
viewed
as
an
R-module.
is true trivially). Hence
Q(k+1)
Q(k+1)

without loss of generality assume that

aR

rk P 1.

is an integral domain. In the ring

unique maximal ideal and a minimal prime. Hence by Proposition 4.18 we have

Q(k)
= 0 by Nakayama's Lemma since aR J(R), so Q(k) = Q(k+1) . Now localize at Q.
Q(k+1)
(k)
So Q
RQ = Q(k+1) RQ and Qk RQ = Qk+1 RQ by Lemma 4.32 part 3. So (QRQ )k = (QRQ )k+1 by
k
k
Lemma 4.32 part 1. So (QRQ ) = 0 by Nakayama's Lemma since QRQ = J(RQ ). Hence Q = 0 and
So

hence

Q=0

since

is a domain.

Now in the general case again suppose that


integral domain. Now localize at

P.

Q1 Q ( P . Factor out Q1 and assume that R


Q1 and assume that R is an integral domain.

Factor out

is an
Now

P , by Theorem 4.28, there exists an inclusion preserving one to one correspondence between
primes of R lying inside P and primes of the ring RP . Use this and the rst part of the proof applied
to the ring RP to nish the proof.
localise at

The Generalised Principal Ideal Theorem.


that

R be a commutative Noetherian
x1 , . . . , xr R. Then rk P r.

Let

is a prime ideal minimal over the elements

ring. Suppose

Proof. We prove this by induction


For

r=1

we use Krull's Principal Ideal Theorem.

r 1 elements. Suppose that P is minimal


x1 , . . . , xr and suppose that we can construct a chain of primes P = P0 ) P1 ) ) Pr+1 . If
x1 Pr then in the ring R/x1 R we have a chain of primes P0 /x1 R ) P1 /x1 R ) ) Pr /x1 R () But
P0 /x1 R is minimal over the images of x2 , . . . , xr in the ring R/x1 R. So () contradicts the induction.
So x1
/ Pr .
+x1 R
Let k be such that x1 Pk but x1
/ Pk+1 . So we have Pk /Pk+2 Pk+2
) Pk+2 /Pk+2 . By
Pk+2
Krull's Principal Ideal Theorem Pk /Pk+2 can not be minimal over [x1 + Pk+2 ] (since otherwise we
0
0
have Pk /Pk+2 ) Pk+1 /Pk+2 ) Pk+2 /Pk+2 ). So there exists a prime ideal Pk+1 such that Pk ) Pk+1
0
Pk+2 + x1 R ) Pk+2 . Proceeding this way we can build a new chain P = P0 ) P1 ) ) Pk ) Pk+1
)
0
0
) Pr ) Pr+1 . Now we have x1 Pr and this leads to a contradiction as in ().
Now assume the result is true for primes minimal over

over

Denition 4.36.

Let

be a commutative ring. We dene the Krull dimension of

by

K dim(R) =

supP prime rk P .
Note.

K dim

can be innite in a Noetherian ring even thought the rank of each prime ideal is nite.

Proposition 4.37.

Let

be a commutative Noetherian local ring with Jacobson radical

J.

Then

K dim(R) = rk J < .
Proof. Since

is local,

K dim(R) = rk J ,

and

rk J <

by the Generalised Principal Ideal Theorem

as it is minimal over its generators.

Lemma 4.38.
n 1.

R be a commutative Noetherian local ring with K dim(R) = n.


c is regular then equality holds.

Let

Further, if

23

Then

K dim(R/cR)

J be the maximal ideal of R. Then rk J = n, so there exists a chain of primes J = P0 )


P1 ) ) Pn . As in the Generalised Principal Ideal Theorem we can construct a new chain of primes,
J = Q0 ) Q1 ) ) Qn1 with c Qn1 . Hence rk(J/cR) n 1 ().
Now assume that c is regular. If J/cR = T0 /cR ) ) Tk /cR is a chain of primes in R/cR then
J = T0 ) T1 ) ) Tk is a chain of primes in R. Since c is regular by Proposition 4.18 Tk can not be
a minimal prime of R since c Tk . So n = rk J rk J/cr + 1. Hence rk J/cR = n 1 from () when
c is regular.

Proof. Let

4.7

Regular Local Ring

R be a Noetherian local ring with Jacobson radical J . We have V (R) = dim J/J 2 as a vector space
over the eld R/J . So V (R) =the number of elements in a minimal generator set for J by Corollary
3.23. By The Generalised Principal Ideal Theorem we have rk J V (R)

Let

Denition 4.39.

A Noetherian local ring is called a regular local ring if

rk(J) = V (R).

A local principal ideal domain is regular by Theorem 4.12

Lemma 4.40.

x J \ J 2,

let

Let R be a Noetherian local ring with Jacobson


R = R/xR. Then V (R ) = V (R) 1.

J (R

not a eld). Suppose that

J = J/xR. Let y1 , . . . , yk

be a minimal generating set for J . Choose y1 , . . . , yk J such that yi 7 yi under the natural
homomorphism R R/xR. Claim x, y1 , . . . , yk is a minimal generating set for J . We shall now
2
show that the homomorphic images of x, y1 , . . . , yk in the vector space J/J are linearly independent.
2


2

Suppose that xr+y1 r1 + +yk rk J (). So y1 r1 + +yk rk (J ) where ri are the homomorphic

images of ri under R R/xR. It follows that ri J since y1 , . . . , yk is a minimal generating set for
J and dim J /(J )2 = k . So ri J for all i. It follows from () that xr J 2 since ri , yi J . So
r J since x
/ J 2 . (Note that J 2 is J -primary check!) This completes the proof.
Proof. Note that

radical

is a Noetherian local ring with Jacobson radical

Theorem 4.41.
the ring

Let

R = R/xR

be a regular local ring with Jacobson radical

J.

Suppose that

x J \ J 2.

Then

is also regular local.

Proof.

V (R) 1 = V (R )

by the previous lemma

rk J

where

J = J/xR

rk J 1

by Theorem 4.38

by the General Principal Ideal Theorem

= V (R) 1
So

V (R ) = rk J .

Thus

is a regular local ring

Remark. We have also shown that

Lemma 4.42.

Let

rk J = rk J 1.

be a Noetherian local ring which is not an integral domain. Let

be a prime ideal. Then

P = pR (p P )

rk P = 0.

Q ( P where Q is a prime ideal. Then p


/ Q. Now q Q implies q = pt for
t R. Hence pt Q t Q since p
/ Q. So q pQ P 2 p2 R. Preceding this way we have

n
n

Q P n for all n 1, so Q (
n=1 P n=1 J where J = J(R). But by Theorem 4.9 n=1 J = 0,
so Q = 0 which is a contradiction since R is not a domain. Hence rk P = 0
Proof. Suppose that
some

Theorem 4.43.

A regular local ring is an integral domain.

K dim R = rk J . If rk J = 0 then R must be a eld.


rk J = n > 0 and assume result for rings of K dim < n. Since J 6= J 2
2

by Nakayama's lemma choose x J \ J . By Theorem 4.41, R = R/xR is regular local. Also

K dim R = K dim R 1. By induction hypothesis R is an integral domain, i.e., xR is a prime ideal.


Suppose that R is not an integral domain, then by Lemma 4.42 xR is a minimal prime. Let P1 , . . . , Pk
2
2
be the minimal primes of R. We have show that J \ J P1 Pk . So J J P1 P2 . So
J Pj for some j by Lemma 4.19 hence J = Pj . So rk J = 0, which is a contradiction. So R is an
Proof. By induction on
Suppose now that

integral domain.

24

Projective Modules

All rings in this chapter are assumed to have


Suppose

5.1

is regular local and

but need not be commutative.

prime. How about the ring

RP ?

Free Modules

Denition 5.1.
1.
2.

M
P
S

R-module M

is generated by a subset

ai ri = 0

finite

Then

A right

1.

RR

SM

if and only if

ri = 0ri R, ai S .

M.

is called a free basis for

Remark.

is said to be free if:

is free with free basis

1
2 2 matrices

2. In a free module not every minimal generating set is a free basis. e.g: in the ring of


over

Q,


0
0

1
0

3. By convention,

Lemma 5.2.

Let


0
0

and

0
1


is a minimal generating set but not a free basis.

is considered to be a free module on the empty free basis.

be a commutative ring, then any two free basis of a free

R-module

have the same

cardinality.
contains a maximal ideal, M say. Then R/M is a eld. Let A be a free
R R
{x } . We claim: xxM
= M (as R and hence as R/M -modules). To see
x R
this, dene : R
by (r) = x r + x M . Then is an R-homomorphism and ker M . But
x M
M is maximal, so ker() = M , proving our claim.
x R

Write B =
= R/M each B is a 1-dimensional vector space over the eld R/M .
x M , since BP
From the external direct sum
B . Now A/AM is an R/M -module. (see Section 1.11). We
P
have A/AM
=
B (as R-modules and hence also as R/M -modules). Hence dimension of
A/AM as a vector space is ||. The dimension of A/AM is invariant by vector space theory, hence

Proof. By Theorem 2.2,

R-module

with a free basis

the result.

Remark. Over a non-commutative ring it is possible to have

R
=RR

as right

R-modules.

The Free Module

FA . Let A be a set indexed by . We dene FA to be the set of all symbols


a A, r P
R, , P
where all but a nite number of r are zero. We further require
these expression to satisfy
a
r
=
a s rP
= s .PWe can make FA a right R-module

P
P
P
by dening
a r + a s =
a (r + s ) and ( a r ) r =
a (r r) (for all r , s , r R)
A is a free basis for FA (identifying a A with a 1 FA )
P

a r

with

Proposition 5.3.

Every right

R-module

is a homomorphism image of a free right

R-module

M be a right R-module. Index the elements


of M and form the free right R-module
P
F
. Elements of FM are formal sums of the form
(m
)r
M
i
i , mi M, ri R. Dene FM M by
P
P
(mi )r 7
mi ri M . This map is well-dened and is an R-homomorphism by the denition of
FM .

Proof. Let

5.2

Exact Sequences

Mi

Let

R-modules

be right

nite or innite)
for all

fi+2

and

/ Mi+1

fi R-homomorphism
fi+1

/ Mi

fi

/ Mi1

of

Mi

fi1

into

Mi1 .

is said to be exact if

. Note that since


we have

im fi+1 = ker fi

i.

A short exact sequence (s.e.s.) is an exact sequence of the form

i.e.,

The sequence (which maybe

/M

00

/ M0
/0

/M

is exact we have

ker(f ) = 0, i.e., f

/ M0

/M

25

/ M 00

/0

is a monomorphism. Similarly

M = im(g), i,e, g is an epimorphism. We


00
of M . Also M
= M/ ker(g) = M/f (M 0 ).

is exact so

is isomorphic to a submodule

00

have

M0
= f (M 0 ),

B A, we can construct the short exact sequence 0


inclusion map and the canonical homomorphism.

Given modules
where

is the

Proposition 5.4 (c.f.

Graduate Algebra Theorem 5.3)

/B

/A

Given a short exact sequence

/ A/B

/Ao

/0

Bo

the following conditions are equivalent.


1.

im

is a direct summand of

2. There exists a homomorphism

:CB

such that

= 1C

3. There exists a homomorphism

:BA

such that

= 1A

Proof. 1. 2.) Let B = im() + B1 = ker + B1 . Let 1 be the restriction of to B1 . We have


B = 1 B1 = C , so 1 is an epimorphism. Also ker 1 im B1 = 0. Hence 1 is an isomorphism
and C
= B1 . Dene : C B to be the inverse of 1 . It follows that
2. 1.) We shall show that B = (A)+(B) = ker +(B). Let b B , then b = (bb)+b.
Now b b ker since (b b) = b b = b 1C b = b b = 0. If z ker B
means z = b for some b B and (z) = 0. This means 0 = (x) = b = b x = 0. Thus
B = ker() (B)
Similarly we can show 1 3.

Denition 5.5.

We say that the short exact sequence split if any (and hence all) of the above condition

holds.
Note that if the above short exact sequence split then we have

B = im B1
= AC

(external

direct sum)

Denition 5.6.

A right

R-module P

is said to be projective if every diagram of the from


/B

/0

exact

can be embedded in he diagram

A
in such a way that

Lemma 5.7.
Proof. Let

= .


/B

/0

(the diagram commutes)

A free module is projective.


be a free right module with a free basis

{e }.

Consider


/B

/0

exact

Let b = e . As is an epimorphism, we can choose a A such that a = b . Now dene :


P
P
P
aP
R. Then is anPR-homomorphism F A and ( e r ) =
FP
A by ( Pe r ) =
r , r P
( a r ) = (a )r =
b r =
(e )r = ( e r ). Therefore = .
A projective module need not be free. To be shown later.

Lemma 5.8.

Let

P ( )

be right

R-modules.

Then

are projective

26

is projective if and only if all

/0,

Proof. Let i be the injection map

and let

be the projection map

Consider the diagram

P
f


/B

/0

f to P , f |P = f say. Then f = f i . Since


each P is projective, there exists
P
P
f
:
P

A
such
that
f
=
f
.
Dene
f
=

f p . Then f =
f p =
P
P
f
p
=
f
i
p
=
f
.
So
P
is
projective.

Restrict
maps

For any

consider

P
f


/B

/0

This gives rise to

A
f :
p A.

So there exists

f i

maps

Proposition 5.9.

P A

is a projective right

2.

is a direct summand of a free module

such that

/0
f = f p .

Hence

f i = f p i = f

and

R-module

3. Every short exact sequence

32

f p


/B

The following conditions are equivalent:

1.

Proof.

/ M0

/M

Consider the short exact sequence

kernel of the map

FP P .

/P

/0

/ KP

Since this sequence splits,

21

Follows from Lemma 5.7 and Lemma 5.8

13

Consider

splits.

/ Fp
/P
FP
= P KP

/0

where

KP

is the

0
Since

is projective, there exists

/ M0

/M

:P M


/P

1P

such that

/0
g = 1P .

Thus the short exact

sequence splits.

Example.

Projective does not imply Free. Let R = Z/6Z, A = 2Z/6Z and B = 3Z/6Z, then A, B C R
R = AB . A being a direct summand of R is projective, but is not free since it has fewer elements
than R

and

Theorem 5.10.

Over a commutative local ring, nitely generated projective modules are free.

27

R be a commutative local ring with unique maximal ideal J . Let M be a nitely genR-module. Let {a1 , . . . , an } be a minimal set of generators for M . Then there exists a free
onto
module with a free basis {x1 , . . . , xn } and an R-homomorphism : F M such that (xi ) = ai
Proof. Let
erated

(See note on page 25, Question 1 on Exercise sheet 6 or Commutative Algebra).

/K

Thus we have

/F
/M
/ 0 where K = ker().
K F J . If not there exists an element k = x1 r1 + + xn rn (ri R) of F such that
k K but ri
/ J for some i. Say r1
/ J . Since R is local, r1 must be a unit. Since k ker ,
a1 r1 + + an rn = 0. So a1 = r11 (a2 r2 + + an rn ) contradiction the fact that {a1 , . . . , an } was a
minimal generating set. Thus K F J .
0
0
Now since M is projective, the above short exact sequence split. So F = K M where M
= M.
0
0
0
Hence F J = KJ M J . So K = F J K = K (KJ M J) = KJ (K M J) by the modular law.
0
0
But K M J K M = 0, so K = KJ . Now K is nitely generated (check this!). By Nakayama's
0
Lemma K = 0, thus M and hence M is free.
0

Claim:

Remark. Kaplansky has shown that the result is true even without the nitely generated assumption.

The Dual Basis Lemma


Let R be a commutative integral domain with a eld
A = {k K : kA R}. Then A is an R-module.

Lemma 5.11.
q A

of fraction

R,K, A be as above. Let : A R


(x) = qx for all x A.

Let

such that

be an

K.

Let

0 6= A C R

R-homomorphism.

and dene

Then there exists

AK = K . So a typical element of K is expressible as ac1 with a, c R, c 6= 0. Now can be

1
extended to a K -homomorphism, : K K by (ac
) = (a)c1 . Check that is well dened

and K -homomorphism. Let (1) = q K . Then for x A, (x) = (x) = (1x) = (1)x = qx.

Clearly q A .

Proof.

Proposition 5.12

(The Dual Basis Lemma - Special Case)

projective if and only if


Proof.

Let

1 = x1 q 1 + + xn q n

for some

R-homomorphism : F  A.
R-homomorphism : A F such that = 1A

be a free module with an

there exists an

F o
Let

With the notation as above:

(Or equivalently A A

xi A, qi A .

Since

AR is
= R)

is projective,

A.

{f } be a free basis for F . Then for each y A, we have (y) = f1 r1 + +fn rn uniquely
fi {f } and ri R. So for each i, y ri is an R-homomorphism A R. So by

previous lemma, there exists qi A such that (y) = f1 q1 y + + fn qn y . So

for some
the

y = (y)
= (f1 q1 y + + fn qn y)
= (f1 )q1 y + + (fn )qn y since qi y R
So

1 = (f1 )q1 + + (fn )qn = x1 q1 + + xn qn ,

Dene

: A R R
|
{z
}

by

where

(x) = (q1 x, . . . , qn x)

xi = (fi ) A.

for all

x A.

ntimes

Ao
Note that

qi x R

since

+ xn rn , ri RThen

qi A .

R R

Dene

: R R A
|
{z
}

ntimes
is an R-homomorphism and for any

(y)

= (q1 y, . . . , qn y)
= x1 q1 y + + xn qn y
= y
28

by

(r1 , . . . , rn ) = x1 r1 +

yA

So

= 1A ,

AR

hence

is projective.

Proposition 5.13.
is a projective
Proof.

I=0

Let R be a commutative Noetherian integral domain and I CR.


RM -module for each maximal ideal M of R. Then IR is projective.

IRM

I 6= 0.

is trivial so assume

Suppose that

R. Then F is also the eld of fractions of each RM (check!).


IRM is RM -projective by the Dual Basis Lemma, there exists some
x0i IRM and qi F such that 1 = x01 q1 + + x0n qn and qi I RM . Now qi I is a nitely generated
R-module. So qi I = z1 R + + zk R with zi RM . Let a R be a common denominator of the x0i ,
let b R be a common denominator of the zj . Let d = ab, then d C (M ), d = x1 (q1 b) + + xn (qn b)
0
where xi = xi a I and qi bI R ().

Now I I C R, by () I I C (M ) 6= . This is true for all maximal ideal M . Hence I I = R. Thus

1 I I and so IR is projective by the dual basis lemma.


Proof. Let

be the eld of fractions of

M.

Consider a maximal ideal

Since

Remark. This is a special case of a standard result. If


mutative Noetherian ring

M.

maximal ideal

then

AR

is a nitely generated module over a com-

is projective if and only if

Marsumura: Commutative ring Theory Theorem 7.12

Rotman: Intro to homological algebra Exercise 9.22 p258

5.3

AM

is a projective

RM -module

for all

See:

Projective Resolutions and Projective Dimension

Denition 5.14.

If

is a right

/ Pn

...
Pi

where each

R-module,

and exact sequence

/ Pn1 n1 / . . .

/ P0

/A

is projective is called a projective resolution for

A.

/0

(This sequence may be nite or

innite)

Construction of a Projective Resolution


Let

be a right

R-module.

A is a homomorphic image of a free module, say

F0  A

the exact sequence

and we have

/ K1
/ F1 / K0
ker = K0 = im = im .
0

Here

F1

/ F0

/A

F0

(by Proposition

/0,

and

Even if

F1 F0

/ K0

where is the homomorphism


K0 = ker and i =inclusion map. If K0 is projective the above is a projective resolution.
K0 is not projective it is still a homomorphic image of a free module, say F1 . So we have

5.3). So we have the exact sequence

and

F0

/ K1

/0

/ F1

are free and hence projective. If

where

K1 = ker .

Let

i = .

Thus

maps

So we have the exact sequence

/ F0
K1

/A

/0

is not projective the procedure can be repeated.

It may happen that after a nite number of steps we get an exact sequence

/ Kn

0
where the

Kn

/ Fn

are projective and all the

Denition 5.15.

A right

R-module A

exact sequence

0
where each

Pi

/ Fn1

is projective.

/ Pk
k

Fi

/ ...

/ F1

/ F0

/A

/0

are free.

is said to have nite projective dimension if there exists an

/ Pk1

/ ...

/ P1

/ P0

is called the length of this sequence.

29

/A

/0

Further, we say that

has projective dimension

if

is the least integer for which there exists a

projective resolution

/ Pn1

/ Pn

/ ...

/ P1

/ P0

/A

/0

A by pdR (A) (or simply pd(A)) If A does not


pd A = . If A = 0 we take pd A = 1 conventionally.

We denote the projective dimension of


projective dimension we write

pd A = 0

It is clear that

Schanuel's Lemma.
0

if and only if

Let

/K

be two short exact sequence. If


Proof. Dene

is projective.

R-module

be a right

/A

/M

and

L = {(x, y)|x X, y Y

have nite

and let

/0

/ K0

are projective then

such that

/Y

/M

/0

X K0
= Y K.

f (x) = g(y)}.

Then

is a submodule of

X Y.


/M

/0

X is projective there exists an R homomorphism : X Y such that f = g. Dene


: X K 0 X Y by (x, k 0 ) = (x, (x)+g(k 0 ) with x X, k 0 K 0 . is clearly an R-homomorphism,
0
0
also g((x) + g(k)) = g(x) + gg(k ) = f (x) + 0. Thus is an R-homomorphism X K L. Now
0
0
0
(x, k ) = 0 x = 0 and g(k ) = 0 x = 0 and k = 0. Thus is a monomorphism.
Finally if (x, y) L then f (x) = g(y), so g(x) = g(y). So g[(x) + y] = 0. Hence (x) + y
ker g = im(g) = g(K 0 ). Hence there exists k10 K 0 such that g(k10 ) = (x) + y . Thus (x, k 0 ) = (x, y)
and is an epimorphism.
0
So we have X K
= L and Y K
= L and we are done.
Since

Corollary 5.16.

In the above situation

K0

is projective if and only if

is projective.

Remark. For free modules the result corresponding to Schanuel's Lemma does not work.

Generalised Schanuel's Lemma.


quences of

with

Pj , Pj0

is a right

R-module

/ Kn

/ Pn

/ Pn1

/ ...

/ P1

/ P0

/ Kn0

/ Pn0

0
/ Pn1

/ ...

/ P10

/ P00

projective for

(
0
Pn1

Suppose that

and we have two exact se-

R-modules

P00
P0

n
n

j = 1, 2, . . . , n.

odd
even

/0

/ A0
(
P0
Kn Pn0 Pn1
P00

/0
n
n

Pt Pt1

and

even

= Kn0 Pn

If

So assume the result for


map

odd

n = 0 this is just Schanuel's lemma.


0
0
n = j 1, i.e., Kj1 Pj1
...
Pj1 . . .
= Kj1
0
0
0
Kt = ker of map Pt Pt1 . So we have the exact sequences

Proof. By induction on

n.

Then

/A

/ Kj

/ Pj

/ Kj1

/0

/ K0
j

/ P0
j

/ K0

/0

j1

where

we obtain

/ Kj

/ Pj P 0 Pj2 . . .
j1
30

/ Kj1 P 0

j1

Pj2 . . .

/0

Kt = ker

of

/ K0

/ K0

/ P 0 Pj1 P 0 . . .
j2
j

j1

/0

0
...
Pj1 Pj2

In both these sequences the middle terms are projective and the right hand side terms are isomorphic

0
Kj Pj0 Pj1 . . .
....
= Kj0 Pj Pj1

by induction assumption. So by Schanuel's lemma

This

completes the proof.

Corollary 5.17.

With the above notation we have

Corollary 5.18.

If

pd AR = m
/K

Example.

projective if and only if

Kn0

is projective.

and

/ Pm

is an exact short sequence with

Kn

/ Pm1

Pj 's

/ ...

projective. Then

/ P1

/ P0

/A

/0

is projective.

A module with innite projective dimension.

Consider

Z/2Z

as a module over the ring

[x + 2Z][a + 4Z] = [xa + 2Z], x, a Z.

dened by

Z/4Z

Look at

/ Z/4Z

...

/ Z/4Z
9

d2

%
2Z/4Z
8

2Z/4Z
8
0
where

/ Z/4Z
9

d1

&

 : [a + 4Z] [a + 2Z]

and

di : [a + 4Z] [2a + 4Z]

/ Z/2Z

/0

for all i. The kernel at each stage is

2Z/4Z

and thus cannot be projective (why?).

Proposition 5.19.

Let

{A }

be a family of right

R-modules.

Then

pd


A = sup pd A

Proof. We shall do this for the direct sum of two modules, the general case just involves more notation.
Let

/ Qn

...
be projective resolution for

...

/ Pn1 n1 / . . .

/ P1

/ P0

/ Qn1 n1 / . . .

/ Q1

/ Q0

/ Pn

...

/ Pn Qn

and

B.

/A

/0

/B

/0

Consider

/ Pn1 Qn1

/ ...

/ P1 Q1

n (pn , qn ) = (n pn , n qn ), pn Pn , qn Qn .
pd(A B) sup(pd A, pd B)
Suppose that pd(A B) = m < . Consider

where

/ P0 Q0

/ AB

This is an exact sequence and each

/0
Pi Qi

is

projective. It follows

/ Tm

m1
/ Pm1 Qm1

/ ...

/ P0 Q0

/ A / B

1 are the maps dened above, since pd(AB)


= m. But Tm = ker m1
= ker m1 ker m1 .
This implies pd A pd(A B) and pd(B) pd(A B).
The above argument shows that if either pd A or pd B = then pd(A B) = and conversely.

where

This completes the proof.

Lemma 5.20.

Suppose that

0
is an exact sequence with
and we have in this case

/K

P projective and A
1 + pd K = pd A.

/P

/A

/0

not projective. Then

pd K <

if and only if

Proof. Follows from denition of projective dimension and generalised Schanuel's Lemma.

31

pd A <

Mk

Recall how build our projective resolution for

=0

0
!
/ Pn

/ ...

= K1

!
/ P1

/ P2
=

< K2
0

Theorem 5.21.

Let

/ P0
=

/0

= K0
!

0BAC0

/M

be a short exact sequence. If the projective dimension of

any two module is the short exact sequence is nite then so is the third. Furthermore we have
1. if

pd A > pd B

then

pd C = pd A

2. if

pd A < pd B

then

pd C = pd B + 1

3. if

pd A = pd B

then

pd C pd A + 1.

Proof. To prove the rst part we induct on

the sum of the nite projective dimension.

then both modules must be projective. If one of these is

If

n = 0

then the short exact sequence splits. So

A or B is projective then so is the other. On the other hand if A and B are


pd C 1.
Now suppose that n > 0 and the result is true when the sum of the two projective dimension
is < n. We may also assume that neither A nor C is projective. Now there exists a projective P
such that 0 D P A 0 is exact (). So A
= P/D. Hence there exists a submodule E
with P E D such that B
= E/D, moreover C
= A/B
= (P/D)/(E/D)
= P/D (by the third
by Lemma 5.8 if one of
projective then

isomorphism theorem). Thus we have short exact sequences

Now

()

and

()

give

/E

/P

/C

/0

()

/D

/E

/B

/0

()

pd D = pd A 1

pd A < and pd E = pd C 1 if pd C < (by the


() gives that if any two of D, E, B have nite projective
the same is true for A, B and C .
if

previous lemma). So by induction hypothesis


dimension then so does the third. Hence

Now assume that all the projective dimension are nite. We prove the second part by induction
on the sum of all three projective dimension. If

n = 0,

nothing to prove (see the base case of the rst

part of the proof )


Let

n > 0.

If either

A or C

is projective, then the result holds. So assume that neither is projective.

Induction hypothesis applied to

()

gives:

i If

pd E > pd D

then

pd B = pd E

ii if

pd E < pd D

then

pd B = pd D + 1

iii if

pd E = pd D

then

pd B pd D + 1

In terms of

A, B

and

these gives

a If

pd C > pd A

then

pd B = pd C 1

b If

pd C < pd A

then

pd B = pd A

c If

pd C = pd A

then

pd B A.

It can be seen (check!) that a. b. and c. are logically equivalent to 1. 2. and 3. of the theorem.

Theorem 5.22 (Auslander).

Let

be a right

R-module, I

a family of submodules such that:

32

a non-empty well-ordered set and

{Mi }iI

1.

Mi M j

2.

M = iI Mi

3.

pd(Mi /Mi0 ) n

then

ij

if

where

Mi0 = j<i Mj

pd M n

n. If n = 0 then for all i I , pd(Mi /Mi0 ) 0 so Mi /Mi0 is projective. So each


0
0
short exact sequence 0 Mi Mi Mi /Mi 0 splits. So there exists submodules Ci of Mi such
0
0

that Mi = Mi Ci where Ci = Mi /Mi . So each Ci is projective.


P
We claim that M =
iI C . The sum is direct for suppose ci1 + ci2 + + cim = 0 where
cij Cij and i1 < i2 < < im , then cim = ci1 + + cim1P Mi0m Cm = 0. So cim = 0 and
similarly ci1 = ci2 = = cim1 = 0. Suppose now that M 6=
so there exists i I such
iI Ci ,P
P
that Mi *
C
.
Suppose
that
j
is
the
least
index
such
that
M
*
i
j
iI
iI ci . So there exists
P
m P
Mj such that m
/ iI Ci . Now Mj = Mj0 Cj , so m = b +P
c for some b Mj0 , c Cj . But
b P
iI Ci by the minimality of j (b Mk some k < j ). So m
iI Ci a contradiction. Thus
M = iI Ci as required. Hence pd M 0 since M is a direct sum of projective modules.
0
Now assume the result for n 1. We are given that pd(Mi /Mi ) n for all i I . Let F (= FM )
0
be the free module with free basis M , let Fi be the free module with free basis Mi and let Fi be
0
0
the free module with free basis Mi . We have F Fi Fi so we have the short exact sequence
0 K F M 0. Dene Ki = Fi K and Ki0 = Fi0 K . From the relations Mi Mi0 , Fi Fi0
and the short exact sequences 0 Ki Fi Mi 0, it follows that the sequences
Proof. By induction on

/ Ki /K 0
i

/ Fi /F 0
i

/ Mi /M 0
i

/0

(Ki + Fi )/Fi0
= Ki /(Ki Fi0 ) by the third isomorphism theorem. But this is
0
0
Ki /(Ki Fi Fi ) = Ki /Ki . ] Each Fi /Fi0 is free since Fi has a set of generators, a subset of which
0
0
0
generates Fi . Hence Fi /Fi is projective so by Lemma 5.20 pd Ki /Ki n 1. It can be checked that:

are exact. [Note that

i < j , i, j I

ii

K = iI Ki

implies
and

Ki Kj

Ki0 = j<i Kj .

So by Lemma 5.20, we have

Denition 5.23.
modules of

Let

R. D(R)

Lemma 5.24.

Let

pd M 1 + pd K n.

be a ring. We dene

This completes our proof.

D(R) = sup{M } pd M
R.

where

ranges over all right

is called the right global dimension of

be a cyclic module over a ring

R.

Then

M
= R/I

where

is a right ideal of

R.

Proof. Exercise sheet 2. Q4 i)

Theorem 5.25.

Let

be a ring. We have

1.

D(R) = sup{B} pd B

2.

D(R) = sup{I} pd R/I

3. Further if

D(R) 6= 0

where

where

then

runs over all cyclic right

R-modules

runs over all right ideals of

D(R) = 1 + sup{I} pd I

where

runs over all right ideals of

R.

Proof. The equivalence of 1 and 2 follows from the previous lemma. The equivalence of 2 and 3 is

0 I R R/I 0. So we prove 1.
R-module. Well order the elements xi of M (i I ) and denote by Mi [respectively
0
0
by Mi ] the submodule of M generated by all xj , j i [respectively j < i]. Then Mi /Mi is either 0
0
or generated by a single element xi . So pd(Mi /Mi ) n where n = sup{B} pd B where B ranges over
all cyclic right R-modules. Since the family {Mi }iI satises the hypothesis of Theorem 5.22, we have
pd M n, hence D(R) n. But by denition D(R) n, hence D(R) = n = sup{B} pd B .
clear from Lemma 5.20 using the short exact sequence
Let

be a right

Remark. Auslander has shown that for a (left and right) Noetherian ring

is the same as the right global dimension of

33

R,

left global dimension of

5.4

Localization and Global Dimension

All rings are commutative in this section.

R, 0
/ S, 1 S.

multiplicative subset of

M, K be R-modules and : M K andRRS -homomorphism : MS KS by m


s =

Commutative Algebra). If is an epimorphism, so is .


Let

homomorphism. Then we can dene a corresponding

(m)
with
s

m M, s S .

Lemma 5.26.
0

/ AS

/A

If

/ BS

(Check details, c.f.

/ CS

/B

/0

/C

/0

is an exact sequence of

is an exact sequence of

R-modules

then

R -modules.

Proof. See Commutative Algebra 3.3

Lemma 5.27.

If

is a projective

R-module,

then

PS

is a projective

RS -module.

Proof. Routine from rst principle

Lemma 5.28.
Proof. If

D(RS ) D(R)

D(R) =

So assumeD(R)

there is nothing to prove.

< .

Let

section 4.4) using Lemma 5.26 and 5.27

Example.

RS -module. View A as an R-module. Since AS


= A
we get pdR A pdR A. It follows that D(RS ) D(R)
S

be an

D(Z) = 1, D(Z/4Z) = . D(Z(2) ) = 1, D(Z(2) /4Z(2) ) =

34

(see

Global Dimension of Regular Local Rings

6.1

Change of Rings Theorems

Theorem 6.1.
the ring

R/xR

R be a commutative ring and suppose that x is a regular element of R. Denote


R . Let M be a non-zero R -module with pdR M = n < . Then pdR M = n + 1

Let
by

n.
n = 0, i.e., M is R -projective, so there exists a free module F such that F = M M 0
0

(for some submodule M of F ). Now 0 xR R R 0 is exact as R-modules. xR


= RR , so

xR is R-projective. Hence pdR (R ) 1. By Proposition 5.19, it follows that


Proof. By induction on
Suppose that

pdR F 1 ()
So

pdR M 1.

Now

that

MR

cannot be

R. So x does not annihilate any


R-module. But M x = 0, so it follows

does not annihilate any non-zero elements of

R-module and hence


projective. Thus pd M = 1.

non-zero elements of a free

of a projective

n > 0 and assume the result for integers less than n. Now there exists a free R -module
G such that 0 K G M 0 is exact. Since M is not R -projective, pdR (K) = n 1. Hence
pdR (K) = n by induction hypothesis. Also pdR (G) 1 as in (). So by Theorem5.21pdR M = n + 1
if n 6= 1, and pdR M 2 if n = 1.
In the rst case we are done, so now we deal with the case n = 1 and we must rule out the possibility
that pdR M 1 when pdR M = 1. So assume that pdR M 1 and pdR M = 1. So there exists a
free R-module H such that
0 T H M 0 ()
So now let

is exact. So

is projective since

pdR M 1.

exact sequence

Also

/ T /Hx

Hx T

/ H/Hx

since

M x = 0.

/M

/0

Therefore

()

induces the

(check!) and pdR M = 1. Thus T /Hx is R -projective. But by the third


T /T x

isomorphism theorem
Hx/T x = T /Hx as R -modules. Hence Hx/T x is a direct summand of T /T x.

Since T is R-projective, T /T x is R -projective. [If


F = T K then F/F x = T /T x K/Kx].
Rfree
R free

Hence Hx/T x is R -projective. But Hx/T x


= H/T since x is regular. But H/T
= M , so M is
Now

H/Hx

is

R-projective,

R -free

contradiction. So we have proved that

Corollary 6.2.

In the above situation if

Theorem 6.3.

Let

R be
R such
pdR (M/M x) pdR M .

regular element of

D(R ) = n < ,

a commutative ring.
that

pdR M = 1

Let

then

implies

pdR M = 2

D(R) n + 1

be a right

R-module. Suppose that x is a


M . Write R = R/xR. Then

annihilates no non-zero elements of

pd MR = then nothing to prove. So assume pdR M = n < . We prove the result by


n.

Suppose n = 0. If F is R-free then F/F x is R -free. Hence if M is a direct summand of an R-free

module, then M/M x is a direct summand of R -free module. (This argument was used before). Thus
M/M x is R -projective, as required.
Now suppose that n > 0 and the result holds for integers smaller than n. There exists a R-module
F such that
/K
/F
/M
/ 0 ()
0
Proof. If

induction on

is exact, so

pdR (K) = n 1.

Hence

pdR (K/Kx) n 1

by induction hypothesis. From

the exact sequence:

0
so we have

/
/

K+F x
Fx

/ F/F x

/ M/M x

/0

F
KF x

/ F/F x

/ M/M x

/0

35

()

we get

is exact. We claim

f F.

But

K F x = Kx, clearly Kx K F x. Suppose that k = f x K F x, where k K ,


F/K
= M . Thus we have the exact sequence of R -modules

is not a zero divisor on

Since

pdR (K/Kx) n 1,

We get equality if

/ F/F x

/ K/Kx

it follows that

pdR (M/M x) n.

is Noetherian and

/ M/M x

/0

This completes the proof

lies in the Jacobson Radical of

R.

Lemma 6.4.

Let R be a commutative Noetherian ring. Let M be a nitely generated module and


x is a regular element lying in J(R). Suppose that x does not annihilate any non-zero

of M . Write R = R/xR.

M/M x is R -projective implies that M is R-projective.

suppose that
elements
Then

M/M x is R -free. Let v1 , . . . , vn be a free basis of M/M x. Let u1 , . . . , un


be elements of M mapping onto v1 , . . . , vn under the natural homomorphism M M/M x.
Claim: M is R-free with basis u1 , . . . , un .
Let C be the submodule of M generated by u1 , . . . , un . Then clearly, C + M x = M . This gives
[M/C]Rx = [M/C], so M/C = 0 by Nakayama's lemma. Thus M = C and u1 , . . . , un generate M .
Suppose that u1 , . . . , un is not a free basis for M . Then (after possible renumbering) there exists
non-zero r1 , . . . , rk R such that u1 r1 + + uk rk = 0, k n (). Thus v1 r1 + + vk rk M x.

Hence ri xR for all i since v1 , . . . , vk is part of a free basis of an R -module. Say ri = xsi for si R.
We claim rk R ( sk R. Clearly rk R sk R and rk R = sk R would imply sk = rk tk for some tk R, i.e.,
sk = xsk tk and so sk (1 xtk ) = 0. Hence xk = 0 since 1 xtk is a unit since x J(R). But is sk = 0
then rk = 0 contrary to our assumption. Now cancelling out x, () gives u1 s1 + + uk sk = 0 with


sk 6= 0 since rk 6= 0. We can write this symbolically as u1 rx1 + . . . un rxk = 0. Repeating the above
Proof. First suppose that

process we get an ascending chain of ideals

rk R (
This is a contradiction since
So

is

r 
k

R(

r 
k

x2

R ( ...

is a Noetherian ring. Hence

u1 , . . . , un

is a free basis for

as claimed.

R-free.

Next suppose that

M/M x

is

R -projective.
0

/K

Then there exists a free module

/F

is exact. As before this induces the exact sequence of

/ K/Kx

/ F/F x

/M

such that

/0

R -modules
/ M/M x

/0

()

B = M K ( )(external direct sum). Then Bx = M x Kx. This gives B/Bx =


M/M x K/Kx. Since M/M x is R -projective, () splits so F/F x
= M/M x K/Kx
= B/Bx.

Therefore B/Bx is R -free and by earlier part of the proof B is R-free. Hence from ( ) we have
that M is R-projective.
Now write

Theorem 6.5.

Let R be a commutative Noetherian ring, MR a nitely generated module. Suppose


is a regular element such that x J(R). Suppose also that x does not annihilate any

non-zero elements of M . Write R = R/xR. Then pdR (M/M x) = pdR (M )

that

x R

pdR (M/M x) = n.
pdR (M/M x) = then pdR (M ) = by Theorem 6.3
So assume that n < . We induct on n. For n = 0 the result
Assume that n > 0 and the result for values smaller than n.

Proof. Let
If

that the sequence

/K

/F

/M

is proved by previous Lemma.


There exists a free module

/0

is exact. As before this induces the short exact sequence

/ K/Kx

/ F/F x

36

/ M/M x

/0

()

such

R -free we have that pdR (K/Kx) = n 1. Since R is Noetherian and M is nitely


generated we have K is nitely generated. Clearly x annihilates no non-zero elements of K . Now
pdR (K) = n 1 by induction hypothesis. So () gives pdR M = n (unless pdR (M ) = 0 but in this
case pdR (M/M x) = 0 by Theorem 6.3) This completes the proof.
Since

F/F x

is

Corollary 6.6.

Let

Proof. We have

D(R) n + 1

R be a commutative
D(R ) = n < then D(R) = n + 1.

Noetherian ring. Let

x J(R)

be regular and let

R /xR.

If

pdR M = k .
R-projective.

by Corollary 6.2. Now let

k n + 1. This is
R-module F such that

We shall not show that


So there exists a free

/K

/F

be a nitely generated

k = 0

clear if

/M

R-module.
M is

, so assume that

Let
not

/0

pdR K = k = 1. Since R is Noetherian and F nitely generated, we have K is


K F , x does not annihilate any non-zero elements of K . So by the
previous theorem pdR K = pdR (K/Kx) n. So pdR M = 1 + pdR K n + 1 . But by Theorem
5.25 D(R) = sup{M f.g} pd M . Hence D(R) n + 1. Thus D(R) = n + 1.
R
is exact.

We have

nitely generated. Also since

6.2

Regular Local Ring

Lemma 6.7.

Let

be a regular local ring of Krull dimension

n.

Proof. By induction on

Let

be the Jacobson radical of

n.

R.

Then
If

D(R) = n.

n=0

we have

J = 0,

i.e.,

is a

eld and the result is true.

n > 0 and assume the result holds for regular local ring of K dim n 1. Since n > 0, J 6= 0
J 6= J 2 by Nakayama's lemma. Let x1 , . . . , xn be a minimal generating set for J . Then there
exists xi such that xi
/ J 2 . Write xi = x. Since R is an integral domain, x is regular. Let R = R/xR.

By Lemma 4.38 K dim R = n 1. Clearly the images of x1 , x2 , . . . , xi1 , xi+1 , . . . , xn are a minimal

generating set for J/xR. Thus R is a regular local ring, hence D(R ) = n1 by induction hypothesis.
Therefore D(R) = n by Corollary 6.6. This completes the proof.
Let

and so

Lemma 6.8.
J(R)).

Then

Let R be a Noetherian
pd M = 0 or .

pd M 6= 0

Proof. If

or

then there exists a module

0
where

is free and

commutative local ring. Suppose that

K FJ

/K

/F

(as in Theorem5.10). So

such that

/B

pd B = 1.

Ann J 6= 0

(where

J =

Now consider

/0

Ann K 6= 0.

But since

pd B = 1, K

is projective

and hence free. This is a contradiction since a free module cannot have a non-zero annihilator.

Lemma 6.9.
xJ

but

Let

x
/ J 2.

be a regular local ring with Jacobson radical

Then

J/xR

J.

is isomorphic to a direct summand of

Let x R
J/xJ .

be regular such that

x
/ J2

we can choose a minimal generating set x, y1 , . . . , yr of J . Write S = xJ + y1 R +


+ yr R. Then clearly S + xR = J . We claim that S xR = xJ , clearly xJ S xR. Let z S xR.
2
Then z = xj + u1 s1 + + yr sr = xt for some h J, si R, t R. So xt y1 s1 yr sr J ,
since x, y1 , . . . , yr is a minimal generating set for J , we have t J , proving the claim.
J/xJ
Hence we have J/xJ
= S/xJ xR/xJ (check!). Now J/xR
= xR/xJ
= S/xJ which is a direct
summand of J/xJ .

Proof. Since

Proposition 6.10.
R

Let

be a Noetherian local ring with Jacobian radical

is a regular local ring of Krull dimension

Proof. If

J =0

then

5.10). So

pd J = 1 and K dim R = 0, so the


m = 0. We can assume J 6= 0. Since J

is a eld,

We now deal with the case

J.

If

pd J = m <

then

m+1
result is true.
is projective it is free (Theorem

is a principal ideal generated by a regular element, so by Theorem 4.12,

and the result holds.


We now prove the result by induction on

k,

the Krull dimension of

37

R.

rk J = K dim R = 1

J is the unique minimal prime of R. Hence ann J 6= 0 (see Proposition 4.18). Then
pd J = 0 and this is dealt with above (we get J = 0)
So suppose that k > 0 and that the result holds for rings of smaller Krull dimension. Clearly we
may also assume m > 0. We have 0 < m < . So by 6.8 ann J = 0. So by Proposition 4.20, J
contains a regular element, say x. By Proposition 4.21, we may choose x such that x
/ J 2 . Write

R = R/xR, J = R/xR. Since x is regular by Lemma 4.38 we have K dim R = k 1.

Claim: pdR J = m 1. We have pdR (J/xJ) pdR J by Theorem 6.3, but by Lemma 6.9 J is

a direct summand of J/xJ , so pd J < . Since m 1, applying Theorem 5.21 to


If

k=0

then

by Lemma 6.8

0
we have

pdR J = pdR J = m,

So by induction hypothesis

R is regular
rk J = m + 1)

local. (J

and

/ xR

/J

so by Theorem 6.1

/ J

/0

pdR J r = m 1.

is a regular local ring of Krull dimension

is generating by

elements so

m.

K dim R = m+1
m + 1 elements. But

Hence

is generated by

Collecting these results together we have

Theorem 6.11

(Serre)

of Krull dimension of

Corollary 6.12.
Proof.

RP

If

Let

In fact, if

R
S

is regular local ring

then the ring

RP

is also regular local

D(R) < .

Hence

D(R) = n.

is a prime ideal of a regular local ring

is a Noetherian local ring, by the previous theorem

Lemma 5.28.

be a commutative Noetherian local ring. Then

if and only if

D(RP ) <

is regular local by the previous Theorem


is a multiplicatively closed subset of

38

and

D(R) <

then

D(RS ) D(R) < .

by

Unique Factorization

All rings are commutative with

7.1

Unique Factorization Domain

Denition 7.1.
Note. If

0 6= p R

An element

is a prime element, then so is

Denition 7.2.

The ring

and every non-zero element

is said to be a prime element if

up

where

pR

is a prime ideal.

is a unit.

is called a unique factorisation domain (UFD) if

aR

is expressible as

a = up1 . . . pn

where

is an integral domain

is a unit and the

pi

are

prime elements.

Proposition 7.3.

p1 . . . pn where the pi are


and multiplication by a unit.

If an element of an integral domain is expressible as

primes, then this expression is unique up to a permutation of the

pi 's

Proof. Algebra II course. (Or Hartley and Hawkes: Rings, Modules and Linear Algebra; Theorem
4.10)

Denition 7.4.
there exists

Let

cR

Proposition 7.5.
if every rank
Proof.

be an integral domain and

such that

Let

a, b R.

We say that

divides

and write

be a commutative Noetherian integral domain. Then

prime ideal of

Let

be the set of all elements of

pi

if

is a UFD if and only

is principal.

Let P be a rank 1 prime ideal of R. Let a P . Then a must be


a = up1 . . . pn where u is a unit and the pj are primes. Hence pi P for
P = pi R since P is a rank 1 prime ideal and pi R is a non-zero prime ideal.

unit and each

a|b

b = ac.

which are expressible in the form

a non-unit, so
some

up1 . . . pn

and so

with

is prime.

a
/ S then aR S = . Suppose not. Let b R such that
ab = up1 . . . pn and n is the least possible, where u is a unit and the pj are primes. (Note:
ab cannot be a unit since a is not a unit). Now pi - b for any i since if pi |b b = pi ti for
some ti R. Hence ati pi = up1 . . . pn ati = up1 . . . pi1 pi+1 . . . pn which contradicts the
choice of n. Now p1 |ab so p1 |a. Let a = p1 a1 where a1 R. Then p1 a1 b = up1 . . . pn and so
a1 b = up2 . . . pn . Again p2 |a1 since p2 - b. Proceeding this way we obtain that b is a unit of
R. Therefore a = b1 up1 . . . pn , a contradiction since a
/ S.

We shall rst show that if

R is not a UFD. Then there exists a non-zero element a R such that


a
/ S . By the above aR S = . Choose P aR to be an ideal maximal with respect to
P S = . Then P is a prime ideal (check!). However, P will contain a rank 1 prime ideal
and hence, by assumption, a prime element. This is a contradiction since P S = . Thus
R must be a UFD.

Now suppose that

Lemma 7.6.
ideal with

Let

s
/ A.

be a non-zero prime element of a Noetherian local domain R. Let A be a prime


S = {sn }. If ARS is a principal ideal of RS then A is a principal ideal of R

Let

n
ARS = bRS . We may assume that b A (why?). By Lemma 4.9
n=1 s R = 0. So there
k
k+1
k
exists k 0 such that b s R but b
/ s
R. Let b = s a where a R. Then a
/ sR. We have
ARS = bRS = ask RS = aRS . Also ask A gives a A since s
/ A and A is prime
Claim: A = aR
m
m
Let x A. Then x aRS . So x = ars
for some m, suppose m 1. Hence xs
= ar.
m
Since a
/ sR, r sR since sR is prime. So r = sr1 for some r1 R. Hence xs = asr1 and so
xsm1 = ar1 sR if m 1 > 0. Proceeding as above we nally obtain x aR. Thus A = aR as

Proof. Let

required.

39

7.2
Let

Stably Free Modules

A, B

be

nn

matrices over a commutative integral domain. Then

|AB| = |A| |B|

a (non-zero) ideal of

where

denotes the determinant of the matrix

Notation. Let

be a ring. We write

Rn

R(n) )

(or sometimes

for

R R
{z
}
|
n times

Theorem 7.7
such that

(Kaplansky )

A Rn1
= Rn

as

Let R be a
R-modules.

commutative integral domain and

Then

is a principal ideal of

R.

A Rn1 has a free basis consisting of n elements, say 1 , . . . , n .


j = (1j , 2j , . . . , nj ) where 1j A and ij R. Let

11 12 . . . 1n
21 22
2n

= .

.
..
..

n1 n2
nn

Proof. The isomorphism shows that


Each

Then

Mn (R),

is an

n-tuple,

so let

|| A.

note that

Now consider

I
R

X =.
..
R
X Cr MN (R).

Then

a1j A

and

a12
b22

sij R

since

a11
b21

..
.
bn1
Thus

..

I
R

a1n
b2n

bnn

...
..

bn2

bij R for 2 i n. Writing the elements of A R R


a1j
11
12
1n
bij 21
22
2n

.. = .. s1j + .. s2j + + .. snj


. .
.
.
bnj
n1
n2
nn
=1

with

...

Let

a11
b21

..
.
bn1
where

I
R

1 , . . . , n
a12
b22

...
..

bn2

=2

is a free basis for

=n

AR

a1n
11
21
b2n

= ..
.
bnn
n1

12
22

. In the matrix from these can be written

...
..

n2

1n
s11
s21
2n

..
.
nn
sn1

s12
s22

B Mn (R)

such that

1
1
..

0
40

= B

...
..

sn2

X Mn (R), but Mn (R) X since X C R. Hence X = Mn (R).

1
0

..

.
0
1

so by above there exists

as columns we have

Now let

s1n
s2n

snn
x A and consider

Take determinants, we have


and

x = || |B|.

Thus

A ||R,

but

||R A

since

A C R.

Thus

A = ||R

is principal.

Denition 7.8.

MR is said to have a nite free resolution


Fn1 F0 M 0 with each Fi is free.

if there exists an exact sequence

0 Fn

Clearly, over a regular local ring each nitely generated module has a nite free resolution

Lemma 7.9.

Let

be a multiplicatively closed subset of a commutative ring

resolution then so does the

R.

If

MR

has nite free

RS -module MS

Proof. Exercise

Denition 7.10.
F

and

R-module M
GM
= F.

An

such that

is called stably free if there exists nitely generated free modules

Clearly a stably free module is projective. A stably free module is a nitely generated projective
module with a nitely generated free complement

Lemma 7.11.

Let

be a commutative ring. A projective

R-module

with nite free resolution is stably

free
Proof. We prove this by induction on the length of the nite free resolution. Let
resolution module.
For

n=1

we have

0 F1 F0 M 0. M

is projective. So this splits, so

be a nite free

F0
= F1 M

and

is stably free.
Now suppose we have

/ Fn

/ ...

/ F1

/
; F0

/M

/0

: K0
$

0
We have

F0
= K0 M

since

is projective.

hypothesis there exists a nitely free module


with both

F0 G

and

K0 G

K0 has nite free resolution of length n1 . By induction


G such that K0 G is free. Hence F0 G
= K0 G M

free.

R is a Noetherian domain and 0 6= A C R


m = n 1 (Q4 on exercise sheet 7)

If
case

such that

Theorem 7.12 (Auslander - Buchsbaum 1959).

is stably free then

A Rm
= Rn .

In this

A regular local ring is a UFD.

R be a regular local ring of dimension n. We prove the theorem by induction on the (Krull)
n.
If n = 0 then R is a eld and there is nothing to prove.
2
Assume result for regular local rings of dimension less than n. Let J = J(R), choose p J \ J . By
Theorem 4.41 R/pR is regular local. By Theorem 4.43 pR is a prime ideal and p is a prime element.
n
Let S = {p }, then clearly K dim RS < K dim R.
Now let T be a rank 1 prime of RS . Let M be a maximal ideal of RS . Then either T (RS )M = T RS
or T (RS )M is a rank 1 prime ideal of (RS )M . By induction hypothesis (RS )M is a UFD. So by
Proposition 7.5 T (RS ) is principal and hence a projective (free) (RS )M -module. So by Proposition
5.13 T is a projective RS -module. Now let A be a rank 1 prime of R. By above ARS is a projective
RS -module. Since every nitely generated module over RS has nite free resolution by the previous
lemma, ARS is stably free. So by Theorem 7.7 ARS is free. Thus ARS is a principal ideal. So by
Lemma 7.6 A is a principal ideal if p
/ A. However if p A then pR = A since rank A is 1. So by
Proposition 7.5 R is a UFD
Proof. Let
dimension

Key point.

RS

is not local.

41

Beyond the Course


Theorem 7.13.

Let

1. Every ideal of
2.

RM

3.

be a commutative Noetherian integral domain. The following are equivalent:

is a product of prime ideals

is a PID for each maximal ideal

is integrally closed and

K dim R = 1

(There are various other characterisation) Such a ring is called Dedekind Domain.

R is a commutative integral domain, I C R, F the eld of fraction, then I = {q

Then I I R, I I C R.

to be invertible if I I = R. By the dual basis lemma I invertible is the same as IR

Recall that if

F |qI R} .
I is said

projective. So we can add:


4. Every non-zero ideal of
5. Every ideal of

is invertible

is projective.

2), MR projective implies M RM projective. So M RM is free by Theorem 5.10. Thus


M RM is principal, hence by Theorem 4.11 RM is a PID.
2) 5). Let I C R, then IRM is principal. So for each maximal ideal M of R. So each IRM is
RM -projective. Hence by Proposition 5.13 IR is projective.

Proof. 5)

Thus a Dedekind domain is a Noetherian domain

42

with

D(R) = 1.