London Mathematical Society Lecture Note Series.

J.T.KNIGHT

Commutative Algebra

WG3V;

CAMBRIDGE AT THE UNIVERSITY PRESS 1971

CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo

Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK

Published in the United States of America by Cambridge University Press, New York

www. Cambridge. org

Information on this title: www.cambridge.org/9780521081931
© Cambridge University Press 1971

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.

First published 1971

Re-issued in this digitally printed version 2007

A catalogue recordfor this publication is available from the British Library

Library of Congress Catalogue Card Number: 76-152625

ISBN 978-0-521-08193-1 paperback

 These notes were put together with scissors and paste
from my manuscript, with as many errors corrected as my
audience and I could find. I should like to thank everyone who
helped, and especially Barry Tennison.

James T. Knight

Dr J. T. Knight died in April 1970 from injuries

received in a motor accident. In preparing the notes for
publication, I have made some slight amendments; I hope that
the original flavour has been preserved. My thanks are due
to the London Mathematical Society, and particularly to
Professors J. W. S. Cassels and P. M. Cohn for their help.

Barry R. Tennison

October 1970

111
J. T. Khigiit

James Thomson Knight was born in Glasgow in 1942. It

was in his home city that he began his university education, after
he had gained first place in the Glasgow University Entrance
Bursary Competition in 1960. Four years later he graduated with
first class honours in Mathematics, and, for achieving first place
in the final examination, was awarded the Bryce Fellowship.
He then went to Trinity College, Cambridge, where he
worked in number theory and commutative algebra under the super-
vision of Dr. (now Professor) J. W. S. Cassels, leading in 1967
to the degree of Ph. D., and election to a Junior Research Fellow-
ship at Churchill College, Cambridge.

After spending the next year as Lecturer in Mathematics at

the United College, Chinese University of Hong Kong, he returned
to take up the Fellowship in Cambridge. His research interests
were in commutative algebra and ring theory. The list below
summarises his published work, which, apart from the first item,
some work in physical chemistry done with ICI while he was an
undergraduate at Glasgow University, has all appeared in the
Proceedings of the Cambridge Philosophical Society.
Dr. Knight died in a car accident in April 1970 while visiting
a friend in Ireland.

J. T. Knight's Publications
A kinetic theory of droplet coalescence with application to emulsion
stability (with R. A. W. Hill), Trans. Faraday Soc. 61(196*5)
170-181.

Quadratic forms over R(t), Proc. Camb. Phil. Soc. 62 (1966)

197-205.
Binary integral quadratic forms over R(t), Proc. Camb. Phil.
Soc. 62 (1966) 433-440.

IV
Riemann surfaces of field extensions, Proc. Camb. Phil. Soc.
65 (1969) 635-650.

Some rings of interest in the study of places, Proc. Camb. Phil.

Soc. 68 (1970) 255-264.

On epimorphisms of non-commutative rings, Proc. Camb. Phil.

Soc. 68 (1970) 589-600.
A note on residually finite groups, Proc. Camb. Phil. Soc. 69
(1971).

B. R. T.
Contents

Page
1. Preliminaries 1
1.1 Introduction 1
1. 2 Definitions and recapitulations 1
1. 3 Modules 6

2. Flatness 16
2.1 Projective modules 16
2. 2 Flat modules 18
2. 3 Faithfully flat modules 24

3. Fractions 32
3. 1 Rings, modules and algebras of fractions 32
3. 2 Localisation 36
3. 3 Projective modules and localisation 42
3. 4 Submodules of fraction algebras 55

4. Supporting and associated prime ideals 61

4. 1 Lengths and ranks of modules 61
4. 2 The support of a module 64
4. 3 Prime ideals associated to a module 66

5. Integers 72
5. 1 Definition of integers 72
5. 2 Integers and prime ideals 80

Vll
6. Some geometrical results 85

7. Valuation rings 89
7.1 Ordered groups 89
7. 2 Valuation rings 90
7. 3 Extension theorems 94
7. 4 An application 99

8. Prufer and Dedekind rings 105

9. General Exercises 113

Appendix 1. Categories 116

Appendix 2. The constructible topology 121

Bibliography 124

Indices 125

vm
1. Preliminaries

1.1 INTRODUCTION

The subject matter of commutative algebra is the common

ground of geometry and arithmetic. As these subjects have grown
more abstract, more and more common ground has been found, and
commutative algebra has grown very large.
In chapters 1 to 5 we shall cover some general material;
in chapter 6 there will be some results useful in geometry; and
in chapters 7 and 8 some arithmetical results.
An undergraduate algebra course should be all you need
to follow this course.

1. 2 DEFINITIONS AND RECAPITULATIONS

By Tringf we mean 'commutative ring with a oneT; for

example, the zero ring {0 }:

0 0

0 0

By Tring morphismT we mean a one-preserving homo-

morphism.
Let A and B be rings. We say that A is a subring
of B iff A Q B and the inclusion map: A -* B is a ring morphism.
For example {0 } is not a subring of the integers Z ; but Z is
a subring of the rationals Q.
We say that a ring A is integral iff
(i) A is non-zero
(ii) for x, y e A, if xy = 0, then x = 0 or y = 0.
We say that a ring A is a field iff
(i) A is non-zero
(ii) for all x e A with x =£ 0 there exists y e A such that
xy=l.
(We write x for this unique y.)
Let us write A* for the set of all non-zero elements of
a ring A. Then A is integral (resp. a field) iff A* is a
semigroup (resp. a group) under multiplication. Thus every field
is integral.
For a non-zero ring A we define 11 (A) = {x e A; xy = 1
for some y e A }; It (A) is a group under multiplication and is
called the group of units of A. For a field tt(A) = A*.
Let a be an ideal of A. The following statements are
equivalent:

(i) a c A

(ii) i i a
(iii) a n tt (A) = 0

and we call such an ideal proper. Note that for x € A we have:

x i u (A) iff xA is proper.

A subset S is of a ring A is called multiplicative iff

(i) 1€S
(ii) x, y e S implies xy e S;

for example: {1}; tt(A); A itself.

 Let p be an ideal of A. The following conditions are
equivalent:

(i) A\ 9 is multiplicative
(ii) A/ p is integral.

We call such an ideal prime. Note that it must be proper. For

example, A is integral iff {0 } is a prime ideal.
Let m be an ideal of A. The following conditions are
equivalent:

(i) ttt is a maximal element of the set of all proper ideals

ordered by inclusion
(ii) A/ m is a field.

We call such an ideal maximal. Every maximal ideal is prime.

1. 2.1 (Proposition ). Let A be a ring; S Q A be multiplicative;

and a be an ideal of A with a n S = 0. Then there is an ideal
p of A, maximal among those ideals b with b2 a and
b n S = 0; and any such p is prime.

Procf. Let

X = {b an A-ideal: b2 a and b n S = 0} .

Then a e X * 0. Let Y c X be non-empty and totally ordered

by inclusion. Then u (Y) e X and by ZornTs lemma X has
maximal elements. If p is such an element, 1 ft p ; and
if x, y ft 9 and x y e )) , then p + xA, p + yA D p so
that 9 + xA, p + yA ft X. Thus there exist s, t e S; p, q e p ;
a, b e A such that s = p + xa and t = q + yb. Then
st = pq + xaq + ybp + abxy e p n S, a contradiction. Thus

A\ p is multiplicative and p is prime, a

1.2.1,1 (Corollary). A non-zero ring has a maximal ideal.

Proof. Take S = {1} and a = { 0 }. D

1. 2.1. 2 (Corollary). Let A be a non-zero ring. Then
tt(A) = A\u ttl where the union is taken over all maximal
m
ideals m of A.

Proof. One way is immediate. Conversely let x € A\ u (A).

Then x A c A and there is a maximal ideal VX of A with
xA c m : so that x e m . a

We say that x € A is nilpotent iff x11 = 0 for some

n e w . We write tt (A) for the ideal of all nilpotent elements of
A, and call n (A) the nilradical of A.

1. 2.1. 3 (Corollary). Let A be a non-zero ring. Then

n (A) = n p
9
taken over all prime ideals p of A.

Proof. If x11 = 0, then x11 € p so x e p for all prime ideals

p . Conversely if x ± 0 for all n e co, then S — {x : n e co}
is multiplicative and S n {0} = 0; thus S n p = 0 for some
prime ideal p of A, and in particular x ft p . a
 It may be that a non-zero ring A has only one maximal
ideal III (A). In this case we call A local; HI (A) is its
greatest proper ideal; and tt (A) = A\ til (A). We call
A/ m (A) the residual field K(A) of A. For example, a field
is local and is its own residual field.
If a and b are ideals of a ring A we define the ideal
a b as follows:

m
a b = i 2 a.b.: a. e a and b. e b }.
i l l 1 l 1

If 9 is prime and 9 2 ab , then plainly 9 2 a or

9 2 b . Thus 2 is like 'divides'.
a is
E ( J\ A a family of ideals of A, we define
AA 6 A
Z a A
to be the ideal of all Z a. for families (aA A
X eA X €AX XX €A
in II a with a = 0 for all but a finite humber of X e A.
x A X X

X eA
We call the set spec(A) of all prime ideals of a ring A
the spectrum of A. It is non-empty iff A is non-zero. For
ideals a of A we define

V( a ) = { 9 espec(A): 9 2 a } so that

spec(A) = V({0})

0 = V(A)

v( a ) u v( b ) = v( ab )

n V( a J = V( 2 a )
A A
X eA X eA
and the V( a ) are thus the closed sets for a topology on spec(A),
called the Zariski topology. It is rarely Hausdorff as we shall
see.

Exercise. spec(A) is compact.

If f:A •* B is a ring morphism we define F:spec(B) —

spec(A) by F( q ) = f~1[ q ]. Then F is continuous because
F"X[V( a )] = V(Bf[ a ]) for any ideal a of A. Thus spec
is a contravariant functor from the category of rings to the
category of topological spaces (see Appendix 1). If f is onto,
the homomorphism theorems for rings show that F is an embedding.

1.3 MODULES

Let A be a ring. An A-module M is an additive group

(M, +) together with a multiplication: A x M -* M such that

A(m + mT) = Am + Am'

(A + A')m = Am + A'm

(AAf)m = A(ATm)

lm = m

for all A, Af e A and m,m T e M. For example, if A is afield,

an A-module is just a vector space over A.
Most definitions and some theorems for modules are just
like those for vector spaces. For example, if (M ). . is a
family of A-modules, the direct sum © M consists of all
XcA A
families (mA . e n M. of finite support (that is with
X eA
m. = 0 for all but a finite number of A € A), and with component-
A
wise addition and multiplication.
A ring is a module over itself; and its submodules are its
ideals.
Let L and M be A-modules and N be an additive group
(resp. an A-module). Let g:L x M -* N. We say that g is
bilinear iff
(i) g(Z + I ',m) = g(Z,m) + g ( r , m )
(ii) g(l, m + m') - g(l, m) + g(Z, mT)
(iii) g(AZ,m) - g(Z,Xm).
. g(U, m) = g(Z, Am) = Ag(Z, m))

for I, ZT e L ; m , m ' e M; and A e A.

On the set Z^ L X M^ of all functions f:L x M - Z with
f(l, m) =. 0 for all but a finite number of (I, m) e L x M, define
addition thus:

(f +f')(Z,m) = f(l,m) +f T (Z,m)

Thus Z becomes an additive group. Let W be the sub-

group generated by

f
(Z + Z ' , m ) ~ f ( Z , m ) - f ( r , m )

m')"f(J,m)"f(Z,m')

f
(AZ,m)
for all I, IT € L; m, mT e M; and X e A:
where f is the function such that

fa((3) = 1 if j3 = a

= 0 if p * a .

We call Z ^ L X M ^ / W the tensor product L ® A M of L and M,

and write I ® m for f n , + W. Thus if x e L ^ . M then
(£,m) A
r
x = Z I. ® m. for L e L and m. e M; and (g) :L x M - L ® A M
._1 1 1 1 1 ' A

is bilinear.
L ^ . M has the following universal property: if N is an
additive group and g:L x M •* N is bilinear, then there is one and
only one group morphism: L &. M -* N such that the diagram

L x M

commutes: namely the function:

f + W i— I f(*,m) g(Z,m) .
I eL
m €M

We use this universal property first of all to make

L ® A M into an A-module: multiplication by H A is the unique
group morphism: L ® * M -* L ®A M such that I ® m «-•
(XI) ® m. With this structure, if N is an A-module and
g: L x M -* N is bilinear, the unique group morphism:

8
 L 8>A M - N such that L x M - L ® A M
A ,A

commutes is also a module morphism (that is, a linear map).

Next if g:L -* LT and h:M -* MT are A-module morphisms,
we define g 8) h to be the unique group morphism L ^ > . M ^
LT ®A M! such that ^ m ^ g ( ^ h(m). Of course g 8) h
is also an A-module morphism.
If (M.) A is a family of A-modules, there is a natural
A X € A
module isomorphism:

L 8> ( © M ) •* © (L 8) M ) given by

I 8>(m K » !-• (I 8) m ) . (its inverse comes from putting

together the natural maps: L 8> M -* L 8>A( © M ) for each
A IX A x » X
^ X eA
Let M be an A-module and (m ) A be a family in M.
A A 6A
We say that (m ) . generates (resp. bases) M iff for all
m e M there is a family (resp. a unique family) ( l . k A
of finite support and such that m = 2 £^ m v
X €A X X
If M has a base we say that M is free: for example
vector spaces are free.
If some finite family generates M, we say that M is
of finite type.
If f :A — B is a ring morphism we make B into an
A-module by defining Xb = f(X)b for X € A and b e B. If also
M is an A-module we make B 8). M into a B-module by defining
]Lt(b 8) m) = (jub) 8) m for \i, b e B and m e M. (We call B 8)A M
the B-ification of M.)
1, 3.1 (Lemma). Let M be an A-module and f :A - B be a
ring morphism. Suppose (m ) A generates (resp. bases) M.
Then (1 ® m.K A generates (resp. bases) the B-module
A A €A
B ®A M.

Proof. The generates part is immediate.

Suppose therefore that ( m j . A bases M. For \x e A
A A €. A

define the A-module morphism p :M - A by

p( Z Lm)= «„
and the B-module morphism h :B ®. M - B by h (b 8) m)
bf(p (m)). Then

h ( 2 ij (1 ® m )) = Z ??, f(p,,(m ))

for any family (77.) A in B of finite support. Thus if

A A €A
x e B &>. M, there is a unique family (h (x)) . in B of finite
support such that

x = . 2 MA x ) (1 ® mAj .D
A eA

1. 3.1.1 (Corollary). Let A be a non-zero ring and M be a

free A-module of finite type. Then there exists n e w such that
if (m ) A bases M, then A has n elements.

10
Proof. A has a maximal ideal lit and A/m &. M is a vector
space of finite type. Let n be the dimension of A/ m ® A M.
m
Since (1 ® > )> A bases A/m ®A M, A has n elements.D

We call n the rank of M.

Strange to say 1. 3.1.1 breaks down for non-commutative
rings.
There is a proof of 1. 3.1. 1 which does not use the axiom
of choice.

Exercise. Let (m ) (resp. (n ) ) generate (resp. base)

A A € A jJL \1 fc JVI
t h e A - m o d u l e X ( r e s p . Y ) . S h o w t h a t ( m . 8> n ) , . v A Y ,,
A fJL \A, \X) € A /\ JVI
generates (resp. bases) the A-module X ®A Y.
A.
An A-module M is called Noetherian iff it satisfies the
three equivalent conditions:

(i) every submodule of M is of finite type

(ii) every non-empty set of sub-modules of M has a maximal
element
(iii) every increasing sequence N Q N c . . . of sub-modules
of M is eventually constant.

For example, a vector space of finite type is Noetherian.

If a ring A is a Noetherian A-module it is called a
Noetherian ring.

1. 3. 2 (Proposition). Let M be an A-module and N be a

submodule. Then M is Noetherian iff N and M/N are
Noetherian.

11
Proof. Suppose N and M/N are Noetherian and let
L Q L Q . . . be an increasing sequence of submodules of M.
Then there exists lew* such that L. + N = L. + N and
L. n N = L. n N for all j > i. From the commutative diagram

ISO
N/L. n N (L. + N)/L.

inclusion

iso
N/L. n N (L. + N)/L.

we see that L. = L..

The converse is immediate, D

1. 3. 2.1 (Corollary).
K M = I M. is a finite sum of
i=l 1
Noetherian submodules M., then M is Noetherian. o

1, 3. 2. 2 (Corollary). If M is an A-module and A is Noetherian,

M is Noetherian iff it is of finite type, a

1. 3. 2. 3 (Corollary). A submodule of an A-module of finite type

over a Noetherian ring A is itself of finite type, o

Exercise. If M and N are Noetherian A-modules, so is

M S>A N.

A sequence
f
n-l
M

12
of A-module morphisms is called exact iff

f.[M.] - Ker(f. +1 ) for i = 0, . .. , n - 2 .

For example 0 -* M — N is exact iff M — N is injective;

f g
M — N — 0 is exact iff M — N is onto; and O—M—N—L—O
is exact iff f is injective and N/f[M] -* L is a well-defined
isomorphism.

Exercise. A functor T: A-modules -* additive groups preserves

all exact sequences iff it preserves all exact sequences of three
modules iff it preserves all exact sequences of the form
0 - M - N - L - O . (Break up M - N ^ L as follows:

0 - k e r f - M - f [M] - 0

0 - f [M] - N - N/f [M] - 0

0 - N/f[M] - L - L/g[N] - 0

and apply T.) We call such a functor T exact.

An A-algebra M is an A-module M together with a

bilinear multiplication: M x M -* M which makes M into a
ring. Thus the map: A - M given by X !-• XI is a ring morphism.
Conversely any ring morphism: A -* B makes B into an A-algebra.
If M and N are A-algebras there is one and only one
bilinear multiplication on M ® A N such that

(m ® n) (m1 8> n1) = (mmT) ® (nnT)

for m, mT e M and n, nT e N. We call M 8). N with this algebra

13
structure the Kronecker product of M and N. (In fact it is
the sum of M and N in the category of A-algebras.)

1.3.3 (Theorem). (I. S. Cohen) Let A be a ring. Then A is

Noetherian iff every prime ideal is of finite type.

Proof. If A is Noetherian, every ideal is of finite type.

Conversely, suppose A is not Noetherian; then the set X of
ideals not of finite type is non-empty; and if Y Q X is totally
ordered by C, u Y e X. By Zorn, X has a maximal element p
We show p prime.
First, 1 i p . Suppose p is not prime, and let
a,b e A\ p be such that ab e p . Then p + aA D p so
rn + aA I X. Hence rp + aA = x A + . .. + x A for some
l r

x. = p. + ay. e p + aA. Now consider

b = {y € A: ay € p } .
Then p Q b and b e b ,sob 2 p+bADp . Hence
b ft X, and so

= z A + . . . + z A for some z , . .. , z e h
1 S IS

Clearly

A + ... + p A + az A + . . . +azA
l r l s

and if p e p then

p = I a.(p. + ay,) (*)

14
for some a. e A. Hence

r r
a (2 ay) = p - 2 a p e p
il l l t l l l
r r s
and so 2 a.y. e b . Therefore 2 a.y. = 2 b.z., say.

Hence from (*)

p- 2 a.p. + 2 b. (az )
fc=l l l j=l J J

Thus p = p A + . . . + p A + az A + . . , + a z A , but
l r l s
p e X, a contradiction, Q

15
2. Flatness

2.1 PROTECTIVE MODULES

If f :M -* N is a morphism of A-modules and P is an

A-module, we can define an A-module morphism:
Hom A (P, M) - Hom A (P, N) by 0 i—^f ° 0 ; and M i-^ Hom A (P, M)
becomes a covariant functor, written Hom A (P, ).

f g
2.1,1 (Proposition), (i) Let 0 - L - M * N be an exact
sequence of A-modules. Then for any A-module P the sequence
0 - Hom A (P, L) - Hom A (P, M) - Hom A (P, N) is exact.

(ii) The following conditions on an A-module P are equivalent:

(a) Hom A (P, ) is exact,

(b) given a diagram P

M-N- 0
of A-modules where the row is exact, there is a morphism:
.P
P -* M such that / \ commutes,
M- N
(c) there is an A-module Q such that P 0 Q is free.

Proof, (i) If 0:P - L is such that f ° 0 = 0, then 0=0

because f is injective.
If i//:P - M is such that g o ty = 0, then for all x eV
there is a unique </>(x) e L such that f(0(x)) = i//(x): thus
0 € Hom A (P,L) and 0 h-*-t// e Hom A (P,M). Conversely if

16
0:P - L, then 0 i - ^ g o f o i / / = O e Hom A (P, N).

(ii) (b) says that we can put 0 on the ends of the sequence in
(i). Thus (see last exercise) (a) and (b) are equivalent.
Suppose (b) holds. Let F be a free module (for example
v(P)
A 0 and f :F -* P be onto. Then there is a morphism g:P - F

such that s \ P commutes. Let us map F •* P © Ker f

F -P
f
by x H-(f(x), x-g(f(x))); and P © Ker f - F by (x,y) i-*g(x)+y;
these morphisms are inverse to each other and thus P © Ker f
is isomorphic to F.
Conversely suppose (c) holds and let (x ) A base
i p A A €A
P © Q and let P - P © Q ** P be the natural maps. Consider
P
° ; for A e A choose m € M such that f(m ) =
A A
M- N- 0
g(p(x )) and define h:P © Q -* M by
A

h( 1 L x . ) = IL m . .
A A A A
AeA XeA

hoi / ,
Then y^ |g commutes, D

We call an A-module P which satisfies (a), (b) and (c)

projective. For example, all modules over fields are free and
therefore projective; Z/2Z is not a projective Z-module.
An easy modification of 2.1. 1 shows that P is projective
and of finite type iff there is a module Q such that P © Q is
free and of finite type. Note that if M is free and f :P — M and

17
g:M — P are such that g o f = 1 then P is projective.

2. 2 FLAT MODULES

(These are rather like projective modules, only ® * is

used instead of Horn..)
Let f:M — N be a morphism of A-modules and E be an
A-module. Thus 1 £ ^ f : E ^ A M - E ® A N, and M H—E ® A M
becomes a covariant functor, which we write E 8)A.

f g
2.2.1 (Proposition). Let L — M •* N -> 0 be an exact sequence
of A-modules and E be an A-module. Then E 8>. L — E 8>. M -
E ® A N - 0 is exact.

Proof. Let x e E ®A N. Then x = 2 e. 8) n. for some

A
i=l x x

e. e E and n. € N. Moreover n. = g(m.) for some m. e M;

r
thus x = (1E 8> g)( Z e.x ® m.).
x
i=l
Plainly (1_ «) g) o (1 8 f) = 1 8) (g o f) = 0. Thus we
Hi Hi HJ

may define:

p
E ®A M/ — • E ®A N
A A
/ Im(l 8) f)

and it remains to show that this is an isomorphism. To do this we

construct its inverse: we map e ® n i—• e 8> m + Im(l 8) f) for
E
any m e M with g(m) = n. This is in order, for if g(m) = g(mT)
= n, then m - mT = i(l) for some I e L and:

T7 8) f) . a
Hi

18
2. 2.1.1 (Corollary). Let E be an A-module. The following
conditions are equivalent:

(i) E ® A is exact,
(ii) if M is an A-module and N is a submodule of finite type,
then E & N - E & . M is injective.
______ j ^ ^^ _________________»

Proof. Suppose (ii) holds and let f:L -~ M be an injective

morphism of A-modules. Let x € E & A L be such that
A
r
x •-•O 6 E ® A M. Then x = 2 e. ® I. for e. e E and
A i_=1 I i i
I. e L. Let N be the submodule of M generated by
f^rM
(f (I.))-, < . < • We have L f-i T commuting; thus

T inj commutes; so that _ e. 8) f(Z.)

1 X
' i-1
A

0 e E <8> A M a n d t h u s Z e. ® f ( i . ) = 0 and:

x = ( l - ® f " 1 ) (I e. ® f ( 2 . ) ) = 0.
E
i l x x

Thus E 8>. L - E ®A M is injective and E ® A i s exact.

The converse is immediate, a

If E satisfies the conditions of 2. 2. 1.1 we say that E

is flat.

19
Exercises. If (E ) is a family of flat modules, 0 E.
X e A
is flat.
If E and F are flat, so is E &>. F.
K A -* B -* C are ring morphisms, B is a flat A-module,
and C is a flat B-module, then C is a flat A-module.

iso
Examples. Since A ® A M • M by X ® m i—• Xm, A is a
flat A-module. Thus (see exercise) every free module is flat.
If P is a projective A-module, then P 0 Q is free for
some module Q; so if M -* N is an injective morphism of
A-modules we have the commutative diagram:

inj
(P © Q) ® M +* (P © Q) 8> A N

(P ®A M) 0 (Q ® M)
iso
I-
I iso
..(P 8)A N) 0 (Q $
A 7 A A A
inj inj

®AN

Thus P ® A M -*P ® A N is injective and P is flat.

We say that an A-module E is flat for an A-module M

iff for every submodule N of M of finite type we have:

E jg) jj ^ E jg) M .
A A

Thus E is flat iff E is flat for every M.

20
2.2.2 (Lemma), (i) If E is flat for M and N is a submodule
of M, then E is flat for M/N.

(ii) If E is flat for M for each A e A, then E is flat for

>M
A "

Proof, (i)
The argument of 2. 2.1.1 shows that for every sub-
mi
module L of M we have E ^ . L ^E 18) M. Thus if L/N is
a submodule of M/N (of finite type), we have the commutative
diagram

N- -L/N-
L
N

N-

with exact rows. Therefore by 2. 2.1

E ®A N

E®, inj

N ®A(M/N)- —0

c o m m u t e s and has exact rows. Let x e E ^ ) . (L/N) and

x h ^ O e E ®A(M/N). T h e r e exists y e E ® ' L such that
y h-^x. Suppose y h ^ z e E ^ . M , Then z i e E 8) (M/N)
so w i-^z for s o m e w e E ^ . N . L e t w i-^u e E ^ . L , then
u , y h-^z e E ^ . M so u = y and w H ^ X along the top. Thus
x=0: that i s , E is flat for M/N.

21
(ii) First let A = {1,2} and M^ ^ We have the
exact sequences

A l A A M
A
A 2

since E ® A M is isomorphic to (E (E <8> A M ).

> A M,)
A A 1 A 2
Let N be a submodule of M (of finite type): we have
the commutative diagram

0 0 N-

with exact rows: thus the diagram

E ®A (M n N) •E ® A N •
A 1

inj

inj
—E

commutes and has exact rows. Let x £ E & A N and

x i—•O eE ®AM. Thus x i-^0 eE ® A p [ N ] so y i - ^ x for
some y e E ^ A (M n N). But y i-^0 e E ® A M; thus y = 0
and x = 0. Therefore E is flat for M.
Next by induction we extend the result to any finite A.

22
 Finally let N be a submodule of finite type of
M = ©ML. Then ML for some finite A c A.
A eA A X l

Let M = ML and M = 1VL so that M = M © M

1
A eAi X 2
A A X 2

inj
As before E E > . M *-E ®» M; and we have shown that
E ®AN-^E ®AA M
M ll ; so E ® A N-^LE ^ M and E is
flat for M. D

2.2.3 (Theorem). E is aflat A-module iff E ® A a ^E

by x ® A •-•Ax for any ideal a of A of finite type.

Proof. We must show that if E is flat for A (the assertion on

the right) then E is flat for any A-module M. Firstly E is flat
for © A by 2.2. 2; and (£ ) -.!-•£ ij m maps
J
,, ' mmeM ., m
m eM m eM
© A onto M; thus E is flat for M by 2. 2. 2. a
m €M
2. 2. 3.1 (Corollary). Let A be a Bezout ring (i. e. an integral
ring in which every ideal of finite type is principal) and E be an
A-module. Then E is flat iff E is torsion-free (i. e. Am = 0
for A e A and m e E implies A = 0 or m = 0).

Proof. Suppose E is torsion-free. If A e A we have the

commutative diagram
X &> \1
E A
®A

onto x A xA

AA •

23
If x € E ® . AA and x ( - • O e E , there exists y e E ® . A such
that y i—•x; and y n-^OeE; so y = 0 and x = 0. Thus

E ® A AA ^E. If A = 0 there is nothing to prove.

Conversely let A e A*. Then A •A by JU i-^jtiA;
so E ® A A ^E ®A A. But E ® A A is isomorphic to E;
A A
A x A
inj
so E ^E: that is, E is torsion-free, a
X A
Examples. If A is afield, every A-module is flat. (Choice is
not used here.)
A Z-module is flat iff it is torsion-free; thus Q is a flat
Z-module but Z/7Z is not.

2. 3 FAITHFULLY FLAT MODULES

2. 3.1 (Theorem). Let E be an A-module. The following

three statements are equivalent:
f g
(i) L - M — N is an exact sequence of A-modules iff
E ®A L -* E ®A M •- E ®A N is an exact sequence of A-modules;

(ii) E is flat; and for any A-module M, if E ® A M = 0, then

M = 0;

(iii) E is flat and m E c E for every maximal ideal m of A.

Proof. Suppose (i) holds. Then 0 -* E ® A M - 0 is exact if

E ®A M = 0, so 0 - M - 0 is exact and M = 0. Thus (ii) holds.
Conversely suppose (ii) holds and let E ®A L -* E ®. M —

24
E &>A N be exact. We have

onto

E ® A (L/Ker(gof))

commuting; thus E ® > (L/Ker(gof)) = 0 and g o f = 0.

Let X = Im(E ® A L - E ® A M)
= Ker(E ® A M - E N)

and consider the commutative diagram:

E®Af[L] Ker g

In E ® A M we have:

X c lm(E ® A f[L]) c lm(E ^ A Ker g) c X

ISO
so that E ®A f[L] Ker g. But

25
 0 - E 8>A f [L] - E <8>A Ker g - E ® A Ker g/f[L] - 0

is exact: so E ® A Ker g/f[L] = 0 and f[L] = Ker g. That is,

(i) holds.
Suppose (ii) holds and let ttl be a maximal ideal of A.
Then A/ ttl * 0 and thus E ® A A/ ttl * 0. But
iso
E®AA/ttl ^E/Itl E b y x ® ( X + t t l ) i - » - A x + IKE
(whose inverse is x + ttl E i-^x 8) (1 + m')); so E / ttl E * 0:
that is, ttl E c E and (iii) holds.
Conversely let (iii) hold and let M 4- 0: say m e M with
m * 0. Let a be the kernel of the map: A — M by A
and let ttl be a maximal ideal of A with ttl 2 a . Then
iso
aEc m E c E so that E 8>A A/ a ^ E / a E * 0.
mj
However A/ a » M; so that E ^ . M ^ O since E is flat.
Thus (ii) holds, o

If E satisfies the equivalent conditions of 2. 3.1 we call E

faithfully flat. For example, A is a faithfully flat A-module; and
a vector space is faithfully flat iff it is non-zero. However, Q is
not^ a faithfully flat Z-module since (pZ)Q = Q if p is prime.
Let M be an A-module. An exact sequence
L •* L -* M -* 0 of A-modules is called a finite presentation of
M iff L and L are free and of finite type. For example, if M
is projective and of finite type, M 0 L is free and of finite type for
some L, so N •* L — 0 is exact for some free N of finite type
and N - * M ® L - M - * 0 is a finite presentation of M. Again, if
A is Noetherian and M is of finite type, L — M -* 0 is exact for
some free L of finite type: and if N = Ker(L — M), N is of
finite type by 1. 3. 2. 3, so that Q — N — 0 is exact for some free
Q of finite type and Q ^ L - * M - ~ 0 is a finite presentation of M.

26
 f g
2. 3. 2 (Lemma). Let O - * M ^ N - P - * O be an exact sequence
of A-modules. Suppose that N is of finite type and P is of finite
presentation. Then M is of finite type.

Proof. Let L x ^ LQ - P - 0 be a finite presentation of P. We

have the commutative diagram with exact rows and columns:

P - 0

-N/p[LQ]

where p is defined as follows: let ( e . ) 1 < . < base L , choose

n. 6 N such that g(n.) = *//(e.), and define p(Z | . e.) = Z | . n.; and
X and F are the only maps which make the diagram commute,
namely

X - l ( x , y ) : x e L i ; y e M; p(«(x)) = £(y) }

F = {(x + X [ L J , f (x) + p[LQ]):x e M } .

Then F is an isomorphism, for F~ : N/p[L ] -~ M/x[L ] consists

of all pairs (y + P[ L J> z +
x[L 1 ]) such that y e N and z e M

27
and g(y) = i//(u) and f(z) = y - p(u) for some u e L . Thus
M/x[L ] is of finite type; and L is of finite type; so M is of
finite type, o

f g
2. 3. 3 (Lemma). Let L -> M — N — 0 be an exact sequence of

A-modules and Q be an A-module. Then

0 - HomA(N, Q) ^Hom A (M, Q) • HomA(L, Q)

is exact.

Proof. Just like 2.1. l(i). a

This lemma leads to the definition of injective modules as
modules Q such that HomA( ,Q) is exact, e.g. Q/Z is injective
but not flat; Z is flat but not injective (each as Z-modules).

2.3.4 (Proposition). Let B be a flat A-algebra; M be an

A-module of finite type (resp. of finite presentation); and N be
an A-module. Then the natural B-module morphism:
B ®A HomA(M, N) •* Horn (B ®AM, B ® A N) given by b 8> f H-^0,
A A JD A A
where 0:c ® m i-^bc ® f(m), is injective (resp. bijective).

Proof. Let T (resp. TT) be the functor B ® A H o m . ( ,N) (resp.

Hom B (P ®A , B ® A N)). Note that if (M.). ^ is a finite family of
A-modules, there are natural B-module morphisms:

T( 0 M.) - © T(M.) and TT( © M.) - © T!(M.)

x l l l

such that

28
 T( © M.) ( © M.)
i €l l i el 1

0 T(M.) 0 TT(M.)
x l
i el i el

commutes.
Suppose first that M = A. The diagram

T(A) T'(A)

ISO ISO

B ®A N
A.
—B

commutes, where T(A) -> B ®A N is given by b ® f i—•b ® f(l)

(and its inverse by b ® n »—•b ® 0 where $:a ^—^ an) and
TT(A) — B ®. N is given by f i—^f(1 8) 1) (and its inverse by
b 8) n i-^f where f(c ® a) = be 8> an). Thus T(A) - TT(A) is
bijective. Therefore (taking account of the first paragraph)
T(M) •* TT(M) is bijective if M is free and of finite type.
(Note: we have not yet used the flatness of B.)
Now let L - L — M -* 0 be exact and L (resp. L
and L ) be free and of finite type. By 2. 3. 3 and since B is flat
we have the commutative diagram

29
 T'(LQ)-

with exact rows; and u (resp. u and v) is (resp. are) bijective.

It is immediate that T(M) - T'(M) is injective (resp. bijective). a

2, 3. 5 (Proposition). Let B be a faithfully flat (resp. flat)

A-algebra and M be an A-module. Then M is of finite type (resp.
of finite presentation; projective of finite type) if (resp. only if)
B M is so too.

Proof. Suppose B ® * M is of finite type; then (1 ® m.)-, <. <

generates B A M for some m. € M. Let (m.), < . < generate
the submodule N of M; then

onto
so that M= N.

Suppose B ®A M is of finite presentation; then M is of

finite type so L — M — 0 is exact for some free L of finite type.
Let N = Ker(L - M) so that 0 - B ^ N - B M- 0
AL - B A
is exact. By 2. 3. 2 B ®A N is of finite type so that N is of finite
type and R -~ N -* 0 is exact for some free R of finite type.. Thus
R - * L - M - ~ 0 is a finite presentation of M.
Suppose B ®AA M is projective and of finite type. Then M
onto
is of finite presentation. Let L •N be a morphism of
A-modules and consider the commutative diagram

30
 iso
B ® A Hom A (M,L) — Hom B (B M, B 8> A L)

iso
B ® A Hom A (M, N) — Hom B (B M, B 8>A N)

in which the horizontal isomorphisms are given in 2. 3. 4. The

onto
vertical arrow on the right is onto because B ^ . L •B ® N
and B ® AA M is projective; thus the vertical arrow on the left is
onto
onto. But B is faithfully flat: so Hom A (M,L) ^Hom A (M,N)
and M is projective (and of finite type).
The converse is immediate, o

31
3. Fractions

3.1 RINGS, MODULES AND ALGEBRAS OF FRACTIONS

Let A be a ring and S Q A be multiplicative. On

S x A define an equivalence ~ as follows:

(s,a) ~ (s\a T ) iff t(sTa - sa f ) = 0 for some t eS.

We write S A for S x A/~ and - for the equivalence class of

s
_2
(s, a). We make S A into an A-algebra by defining

s r
a , aT _ a + sa f
s sT ss T
a a' aaf
s " sF"

x sa ~"_ Xa
s
for A € A.

—1 3.

The natural map: A — S A is given by a H > T .

The A-algebra S~ A has the following universal property:
if f :A •* B is a ring morphism such that f[S] Q tt (B), there is
one and only one ring morphism: S" A -* B such that A -* I

a —1
commutes, namely - i-^ f(s) f(a). (In other words, there is a
unique A-algebra morphism: S" A -* B.)
The one of S" A is y and its zero is j . Thus - = 0
iff ta = 0 for some t e S; and S -1 A = 0 iff 0 € S.

32
 The map: A - S " 1 A is an isomorphism iff SQ tt (A).
Let T = {t € A: a e A* => ta € A* }, the multiplicative
subset of non-divisors of zero in A. We call T~1A = tot(A) the
total ring of fractions of A.
The kernel of the map: A -* S"1A is the ideal
( a e A : ta = 0 for some t e S ) ; thus A™3 S^A iff S c T.
inj
In particular A -* tot (A).
If A is integral T = A* and tot(A) is the field of fractions
of A. Conversely if tot(A) is integral it is a field and A is
integral.

Exercise. If A is Noetherian, so is S A.

3.1.1 (Proposition). Let A be a ring and S c A be multipli-

cative. Then spec(S" A) -> spec (A) is an increasing embedding
whose image is

X = { p e spec(A): pnS=jZi)

and whose inverse is also increasing.

Proof. The map: spec(S A) -* X Q spec(A) is given by

q i-Hp € A: j e q }; its inverse: X — spec(S~ A) is given
by p i - H - : p e p and s € S }.
s
The rest is immediate, o

Let f: A - B be a ring morphism and S c A and T c B

be multiplicative and such that f [S] Q T. Then there is one and
only one ring morphism:

33
 •* T" X B such that

commutes: namely ~ | -^f7~r-

For example let A be integral and k = tot(A) be its
field of fractions, so that we may identify A with the subring
{j: a €A of k. If we take f = 1 A and T = A* we can also
1
identify S" A with the subring {s"" 1 a: a e A and s e S ) of k,
for any multiplicative S Q A with 0 ft S.
Let M be an A-module and S Q A be multiplicative. We
define the A-module S"" M analogously:

5? = 25: iff t(sTm - smT) = 0 for some t e S

s s

m + m/ _ sTm + smT
s sf ~ ss f
m = >m
s s £

We can also make S" M into an S A module by defining

a m am
s t st

3.1. 2 (Lemma). The map f: given by

- ® m !-• is an isomorphism of S A modules.

34
Proof. Let us define a map

g: S M - S A®AM by - ,_±(g) m .
A S S

This is well-defined, for if ?? = H.', then t(sfm - smT) = 0

s s
for some t e S, and

s sT

= 0.

We have also f(g(5?)) = f(-®m) = ™ ; and

s s s

r a. r a.m.
g(f(I -^ m.)) g(2
i=l s i x
i=l s i
r
2 s x . . . s....s r a.m.

s
2 s,...8,..sa1m(
i- 8 r t

r a.

Thus g = f" . D

35
3.1. 3 (Proposition). Let A be a ring and S £ A be multiplica-
tive. Then S~ A is a flat A-module.

Proof. Let F: M -* N be an injective morphism of A-modules.

We must show that S A®AM • S A8).N is injective;

m f(m)
or by 3.1. 2 that S^M — ?-— S"XN is injective. Let ^L = 0;
s
then tf(m) = 0 for some t € S; so f(tm) = 0 and tm = 0 and
s ~~

For example Q = tot(Z) is a flat Z-module.

If M is an A-algebra and S c A is multiplicative, we make
S" M into an S A-algebra (and an A-algebra) by defining

m mT mmT
s sT ssT

This S~ A-algebra structure is the same as that got from the

natural one on S" A&.M using 3.1. 2.

3.2 LOCALISATION

Let A be a ring and p e spec (A). Then A\ p is

multiplicative and we define

Ap = (A\ 9 )'lA

called the localisation of A at p .

36
 Note that A D is a non-zero ring if A is.
(If A is integral we usually regard A „ as a subring of
the field of fractions of A.)
From 3.1. 1 we have an increasing map:

bij
X = { q espec(A): q c p } ^spec(A ).

But X has a greatest member, p ; so A has a greatest proper

ideal

(A )= { £ :p e p and s e A\ p }

= 9 Ap

= (A\ p )" X p

and is a local ring.

In a similar
similai way we define the A -module M n , iso-
morphic to A ®AM, for any A-module M.

3. 2.1 (Theorem). ®A (taken over all maximal ideals ttl

of A) is a faithfully flat A-module.

Proof. By 3. 1. 3 each A m is flat, so © A m is flat; and if n

is a maximal ideal of A, we have

K (© A ^ ) c © n A ^ c ©A^ since

= m ( A n ) c A n .

Thus by 2. 3.1 © A is a faithfully flat A-module. a

III "*"

37
3. 2.1.1 (Corollary). The sequence L — M — N of A-modules is
exact iff the sequence L •*
-* M
M -*-* N dt^K -modules is
exact for every maximal ideal m of A.

Proof. L •* M - N is exact iff (© A - (0 A m )I8> . »

m ' v
m m; A
(0 A m )® A N is exact iff 0 (AJ ®AL)-©(A
0 v(A m ® AN) is exact iff L m is exact for every
m in A m
maximal ideal m of A. a

3.2.1. 2 (Corollary) Let E be an A-module. Then E is flat

iff E m is a flat A -module for every maximal ideal m of A.

Proof. Suppose E is flat and let M — N be an injective morphism

of A m -modules. We have the commutative diagram

inj

bij bij

E N

Where the vertical map is e®m (and its inverse is

- m). Thus E M J N and
s ' m
m m
Em is a flat A m -module.

38
 Conversely suppose that E is flat for each m and let
M — N be an injective morphism of A-modules. Since A m is
aflat A-module, we have M • N m . Thus the diagram
m m
inj
E M J N
m m m m m

bij bij

m m
e m ._L®(e®m)
commutes, where the vertical map is - ® T-
s i
3 3.P1
(and its inverse is -®(e®m) H^ — ®m). Therefore
s s
- • (E® A N) m for every m , and by 3. 2.1.1
A m
inj
E ® . N and E is flat.a

3. 2.1. 3 (Corollary). Let A be an integral ring and k be its

field of fractions, so that we can identify A with a subring of
m taken over
k for any maximal ideal m of A. Then A = n
all maximal ideals m of A.

Proof. We can identify k m with k by the isomorphism:

a -l 1
A >Ak — k given by - ®x • - • s " ax (its inverse is y H^y
A S J-
m Let B = n A
Thus A c B C A c k for all
m m m
so A c B £ (A m ) t and
m m
c B c k by the above identification. Thus
m m m
= B m for all m and by 3. 2.1.1 the inclusion: A - B is
m
onto. Thus A = B. o

39
Exercise (hard). Let A be a ring. Show that the following
statements are equivalent:

(i) every principal ideal of A is generated by an idempotent

(i. e. an element e € A with e = e)

(ii) every A-module is flat

(iii) tt (A) = { 0 } and spec (A) is Hausdorff.

(We call such a ring absolutely flat. See Bourbaki, Alg. Comm.
I, §2, Ex 17.)

Let A be a ring and f € A. We write Af for S"1A where

S = {fn: n e (JL)}. The image of spec(Af) in spec (A) is the open
set D(f) = { p e spec(A): f ft p }; and the sets D(f) form a base
for the Zariski topology, for

spec(A)\V( a ) = u D(f)
f ea
for any ideal a of A.
Similarly we define Mf, etc.
If M is an A-module of finite type and M . = 0 for some
p e spec(A), then M. = 0 for some f e A\ p : for if
(m.)-, < • < generates M, we have s.m. = 0 for some s. e A \ p ,
and we let f = s . . . s . Conversely if ML = 0 and p e D(f),
Mp =0.

3. 2. 2 (Lemma). Let A be a ring and M — N be a morphism of

A-modules. Let p e spec(A).

40
 onto
(i) If M »-N and N is of finite type, then
onto
Mf • N for some f e A\ p .
bij
(ii) If M i-N
and M is of finite type and N is of
bij
finite presentation, then M • N for some g e A \ p .

Proof. Let 0 - R - M - N and M - N - Q - 0 be exact.

onto
(i) Then M • N - Q -0 is exact and Q = 0;
so Q> = 0 for some f e A\p since Q is of finite type; and
onto
Mf - Nf - Qf - 0 is exact: so Mf ^N f .

(ii) We have 0 -* Rf — Mf — N. — 0 exact; so (since Af is a

flat A-algebra, and by 2. 3. 5) N. is of finite presentation and Mf
is of finite type and by 2. 3. 2 Rf is of finite type. However
N
0 —RJJ -* MD ^ D is exact; so R . = 0 and
(R ) = 0. Thus T Rf = 0 for some h e A\ p ; and we take
1
bii*
g = fh to find that R = 0 and Q = 0: so that M - N .o

3. 2, 3 (Lemma). Let (f.). T be a finite family in a ring A which

generates the ideal A. Then II Af is a faithfully flat A-algebra.
i e I *i

Proof. Clearly II A« is flat. Let m be a maximal ideal of A.

i €l i
Then f. t ttl for some i e I, so that m A. c Af and H Af c
1
h i iel l
n A . Thus n A is faithfully flat by 2. 3.1 (iii). o
1 61 l i eI I

41
3. 3 PROJECTIVE MODULES AND LOCALISATION

We say that an A-module M is of countable type iff there

is a family (m ) which generates M. For example, modules
of finite type are of countable type.
If M is of finite (resp. countable) type and N is a sub-
module of M, then M/N is of finite (resp. countable) type.
Conversely if N and M/N are of finite (resp. countable) type, so
is M. Thus if M = P©Q, then M is of finite (resp. countable)
type iff both P and Q are so too. If (M ) is a family of
modules of countable type, © M is also of countable type.
n € a>
3. 3.1 (Lemma), (i) Let M be an A-module and P, Q, QT be
submodules such that M ^ P © Q = P © Q T . Then there is an iso-
morphism: Q — Q\

(ii) Let N be an A-module and let Q, L and M be submodules

with L c M and N = L ©Q. Then M = L © (Q n M).

(iii) Let f: M - M be an A-module morphism with f = f . Then

M = f[M] © ( 1 M - f) [M].

Proof, (i) It is {(x,xT): x e Q; x' eQ T ; x - x' e P } .

(ii) and (iii) are immediate, o

Let us recall some facts about ordinals.

A set a is called ordinal iff otQ&(a) and e\a is
a strict well order on a. For example 0 is an ordinal; and if
a is an ordinal and /3 e a, then /3 is an ordinal. If a is an
ordinal, a ft a and the successor suc(a?) = a u {a} of a is an
ordinal distinct from a. Not every ordinal is a successor: in fact
an ordinal a is not a successor iff a = u (a) iff suc(j8) e a

42
for all fi e a. Such an a we call a limit ordinal: for example
0 and u).
The class On of all ordinals is well ordered by inclusion.
If X c On, and a Q X implies a e X for all a e On, then
X = On. (Transfinite induction.)
If f is a function whose domain is the universe, there is
one and only one function g whose domain is On and for which
g(a) = f(g | a) for all a e On. (Transfinite recursion.)
The axiom of substitution guarantees that if A is a set, then
no map: On •* A can be injective (for otherwise, On would be a
set and On € On, a contradiction).

Exercise (for those unfamiliar with ordinals). Show that if a and

j3 are ordinals, then a c p or j3 Q a. Use this to prove all the
above statements.

3. 3. 2 (Lemma) (Kaplansky). Let P be a projective A-module.

Then P = © Q for a family (Q ) of submodules Q
Y Y Y r
y €a
of countable type.

Proof. Let P ©Q = M a free module and p: M — M be the

projection onto P, so that p = p 2 . Plainly M = © M. for
i €l l
some family (M.). , of submodules of countable type. Let
(X.: M - M.). -r be the associated family of projections.
Define a relation R = {(i, j) e l x I: X-(P(X)) * 0 for some
x e M . } on I, so that for all i e l the set R [ { i } ] = {j el:
(i, j) e R} is countable. For n e a> write R n for
R o R o . .. o R (n times).

43
 Let us define by transfinite recursion the function
0i—^1 taking On t o ^ ( I ) as follows:

(i) if p is a limit ordinal let Ifi = u I

(ii) if p — suc(y) for some y and 1 = 1 , let !„ — I

(iii) if P = suc(y) for some y and I c I, choose i e l \ l

and let lQ = I u ( U R n [{i}]).
r
^ new

Then (by transfinite induction where necessary) we have

(i) 1^ increases with p

(ii) \,— f)
0

(iii) I in\ \Io is countable for all p

(iv) if i € 1^ and (i, j) e R, then j e I

Suppose I c I for all a; then O n - J ^ ' (I) is strictly
increasing, contradicting the axiom of substitution. Thus I = 1
for some least limit ordinal a, and I = U Ifl.
P e a P
Define the family ( E J O of submodules of M as follows:
p p € 01
M.

so that M = U E o . Also p[E J £ E • for if x e M. and

i € I p , then (i, j) £R implies X- (p(x)) = 0, so that

p(x) e © M. c © M. by (iv)
J 3
(i, j) € R j € 1^

44
 We define too ( F l £ Q by:

F ft = © M ;

so that E //Qv = E,0F r for all j8 € a; and by (iii) FQ is of

SUC^p; p /. p
countable type for all pea.
For j8 € a we have
E
suc(/S) = E/S ® *>
= p[Ep] © (1 - p)[E/3] © F^ by 3. 3 . 1 (iii)

so that p [ E g u c ( / S ) ] = ptEp] © Q^ by 3. 3.1 (ii) where

p E
t suc(/S)] ® ( 1 " p)[Esuc(/3)^ ; and
<P uttin S 1 " P in place of p)

(1 " P)[ E s u c ( / 3 ) ] = 0- ~ P ) [ V © L^ for some L^ .

Therefore

and by 3. 3.1 (i) Q . 0 L and F fi are isomorphic. Thus Q^ is

of countable type for all pea.
Let us show by transf inite induction that p[E J = ® Q

for all /3 e Q?. If /S is a limit ordinal, then

p[E ] = U p[E ] = u e Q = © Q ;
p y
yep yep hey ° 6 e^S °

and if p — suc(y) then

y 6

Thus p[E J = © Q for a l l / S e a ; and

A y
y e p

P = p[M]= u P[E]= u © Q = © Q a
F
/Sea j8eaye/S7 y e a 7

45
3. 3. 3 (Lemma). Let A be a local ring; P be aprojective
A-module; and x € P with x * 0. Then there exist submodules
Q, R of P such that

(i) x eQ

(ii) Q is free and of finite type

(iii) P = Q0 R.

Proof. Let P 0 L = M, a free module. If (m.). x is a base of

M and x = Z |.m. we define J = U e I: £. * 0 }. Among all
i el x x
bases choose one for which the finite set J has the least possible
number of elements. Then if j e J we have £. t Z A£.: for
J
i*J *
if L= Z a I , define
J
i*j

m! = m. + a.m. if i e J and i * j

= m. if i = j or i e I \ J

so that (m!). - bases M and x = Z ^ m j , a contradiction.

Now for each i e I let m. = y. + z. for y. e P and
z.eL. Thus x = Z |.y. and Z £z. = 0. Let z.= Z V- m.
1
iej x l ieJ x x x
j e l 1J 3
for each i e j ; then Z ( Z ?7..£.) m. = 0 and Z V..^. = 0 for
j el i €J i €J
all j €l. Thus T?.. e m (A) for all i e J and j e l : for if
7j.. e it (A), then | . £ Z A|..
|. But y. = m.. - Z ^
^...m,
. , for

i € J: so if we define

46
 m!T = y. if i e J

= m. if i e I \ J

we see that (mT). j bases M; and we write

= 2 Ay
i €J l

R = P n ( I Am.) .
l
i€l\J

Thus

(i) x €Q

(ii) Q is free and of finite type, and

(iii) P = Q ® R by 3. 3.1 (ii) o

3. 3. 4 (Theorem). Let A be a local ring and P be a projective

A-module. Then P is free.

Proof. By 3. 3. 2 we may suppose that a family (wJ

generates P.
Let us define recursively sequences (Q ) , ( P ) n €Cl)>
Mfdlowfl:
bJnew
(i) Q o = O ; P o = P; p o ^ l p .

(ii) If pn(w ) = 0 for all y e u> we let P n + 1 = Q n + 1 = 0 and

n y

47
(iii) If p (w ) ^ 0 for some least y e w we use 3. 3. 3 to choose
submodules P , , and Q ,- of P such that
n+l n+1 n

(a) Pn(wy)€Qn+1

(b) Q ,, is free of finite type

and define p , , : P — P ,-, to be the projection.

Thus the sum 2 Q is direct; and if w ft 0 Q
n ew n r
new n
for some least y e w , we have w , . . . , w -, e © Q for
l y i
~ n ew n
n < N
s o m e least N e w ; thus PN(w ) * 0 so that w e © Q .
y y
n e w
n<N+l
Thus P = © Q and is free, o
n ew n

(Without the rather difficult 3. 3. 2 we see that projective

modules of countable type over local rings a r e free.)

3. 3. 5 (Lemma). Let A be a local ring and M be an A-module

of finite type. Let m = ttl (A). Suppose A / m &.M = 0. Then
M = 0.

iso
Proof. We have A / m ®AM •M/m M by (a + m
m
am + ttl M. Thus m M = M. Let ( i )i<i<n generate M for
n
some least n e w . If n > 0 we have m = 2 X.m. for some
n x l

n-1
X. e m , so that m = (1 - X ) 2 Am and (m.) 1 ; ^ i : ^
1=1
generates M, a contradiction. Thus n = 0 and M = 0. a

48
 (This is a special case of a famous lemma of Nakayama.)

onto
3. 3. 5a (Lemma). Let f: M • N be an A-module morphism;
and let (x.)-, < . < generate M and (y.)1 ^.. < base N, where
m < n. Then f is an isomorphism.

Proof. We may suppose that m — n.

There are matrices (a..)-i < • . < and (b, .)-j < t •<

n n
over A such that f(x.) = Z a..y. , y, = Z b, .f(x.) for

i,k = 1, . . . , n. Thus

y t = 2 (I b,.a^)y4

so that Z b, . a . . = 1 if k = j

= 0 if k ^ j .

That is, (b,A . is the inverse of (a..). .. Suppose

K l K, 1 1J 1 , J
n n n
f( Z |.x.) = 0 for L e A. Then Z ( Z |.a..)y. = 0 so that
i=l 1 l l
j=l t=l l 1J J
(bki)k t we obtain (|.). = 0,

n
so that Z |.x. = 0 and f is injective. D
1 X

(This proof was contributed by Barnard.)

3. 3. 6 (Theorem). Let A be a local ring and M be an A-module

of finite presentation. The following conditions are equivalent:

49
(i) M is free

(ii) M is projective

(iii) M is flat

(iv) m

Prod. Suppose (iv) holds and let m = tit (A) and k = A / m .

Then k®AM is a k-module of finite type: say (1 ® m.), €l bases

k® A M. Let A X - M by (£.). T h—2 |.m. and

A ii€i . € l 1 i

I ^ ^ M by ( ^ ^ H j ^ ^ m j . Then

I iso
commutes: so k® A A •k®AM.
A A is0
T I
Suppose A - M - Q - 0 is exact: then k ^ A A

k®AQ •* 0 is exact and k®AQ = 0. But Q is of finite type: so

T onto
Q = 0 by 3. 3. 5, and A1 • M.
Suppose 0 - * R - ~ A -* M -* 0 i s exact: w e have a commuta-
tive diagram
0
X
m ® A R • m ® AA • m ®AM

R • A1 • M
M

iso
AAR • k^.A -k«)A
A A A

50
with exact rows and columns. Let x € k® AR. Then
I I
x *—•() e k® AA . Let y »-• x for y e R, and y KzeA, Then
I I
z H-^0 e k® AA so w I—•z for some w e m ® A A . But
A A
z h-^0 e M so w l-^0 e M and w h-^ 0 e m ®AM. Thus
u i-^w for some u € m 8>AR; and u K z e A l ; so uh^yeR
and y h-> 0 = x € k® AR. That is, k®AR = 0. But by 2. 3. 2, R
I iso
is of finite type. Thus by 3. 3. 5, R = 0: that is, A • M and
(i) holds.
The other implications are immediate, a

3. 3. 7 (Theorem). Let A be a ring and P be an A-module. The

following conditions are equivalent:

(i) P is projective and of finite type

(ii) P is of finite presentation; and P m is a free A. -

module for every maximal ideal m of A

(iii) P is of finite type; P - is a free A -module for

every p € spec (A), of rank r say; and p i—•* r is
continuous

(iv) there is a finite family (f.). T in A, generating the ideal A,

and such that for all i e I the Af -module P f is free and of finite

type.

Proof. It is convenient to prove at the same time a fifth condition:

(v) for every maximal ideal m of A there exists f e A \ m

such that P f is a free Af-module of finite type.

51
 (i) =>(ii) Immediate from 3. 3. 4 and 2. 3. 5.
(ii)=5>(v) Let m be maximal; then (l®p.). T bases
y 1 1 cl
A m (8)AP, say. The map: A - P given by ( ^ i eI •—^

2 £.p. is such that (AT)m 1S

°» P m . Thus by 3. 2. 2 (ii),
T iso T iso T
(Ax)f - P f for some f e A\ m ; and (A )f - ( A f ) : that
is, P f is free and of finite type.
(v) =>(iv) Let X = {f e A: P is a free Af-module of
finite type}. Then X 4 W for every maximal m ; so X
generates the ideal A and 1 = 2 a.f. for some a. e A;
i el 1* l

f. e X; and I finite.
(iv)=>(iii) The II A -module II P is of finite
i e I: i i €I I
iso
type; and we have II P • ( n A. ) ® A P by:
i £l i i el ti A

P
< -
* 2 g . ^ p . where:
i el l 1

f n
i

Thus P is of finite type by 3. 2. 3 and 2. 3. 5.

For each p e spec(A) there exists i e I such that

p€D(f.); so that (Af ) A -A p by:

1 f
i
. m
f. a
a , s | ^ _i
r n -m ' ^ r n
f. f. sf.

52
for s e A\ p and a e A (the inverse is -»—•y / y ). Similarly
S JL -L
iso
(Pf ) A ^P so that P and P f are free A f l - and
i f. i
Af -modules of finite type and of the same rank by 1. 3.1. Further
i
r = n iff p e uD(f.) taken over i e l such that P f is of

rank n; so p*-^r is continuous.

F
i
>(v) Let m be maximal and (-y-) base P
(iii)
iel m
where I has n elements. Map AJ — P by (£.).
onto onto
Thus and (A )e for some
o
g e A\ m by 3. 2. 2 (i). By the continuity there is an h e A\ m
such that if p e D(h), then r n = n. Let f = gh. Then
I onto "
f e A\ m and (A ) • P f still, and r = n for all
T
1 onto
pcD(f). Thus (A ) -P for all p eD(f); and
I I
(A ) n (isomorphic
iso
to (A n ) ) and P n are free of rank n;
T
1
so (A ) P for all p e D(f) by 3. 3. 5a.
Let n be a maximal ideal of Af and q be its image in
spec(A), so that q e D(f). We have the commutative diagram

iso

ISO
(A

where the vertical map is given by

53
 JL / JL f x
fn fm • " gfn

for s e A\ q ; thus ((Aj)1) ——^(Pf^n f o r a11 m a x i m a l

ideals

It of Af, so that ( A f ) I - ^ - P f by 3. 2.1.1.

(iv) = > ( i ) By 3. 2. 3 and 2. 3. 5 it suffices to show that the
n A- -module of finite type II P f (isomorphic to ( II Af )®AP)
i € l fi i € l fi i e l *i A
is projective. Let (L.). T be a family of A -modules of finite
type such that each P f © L. is free and has the same rank. Let

</>.: P f -* P- © L. and i//.: P, © L. - P be the natural A -

i 1. t. I i t. I i. it

module morphisms. Then II (P © L.) is clearly a free

x
iel i
n A. -module and ( II \p.) <> ( II 0.) = 1, n p x so that
i el h i€l x i eI x \el f
i

n P f is projective (compare the note at the end of 2.1). o

i el i

3. 3. 7.1 (Corollary). A flat module of finite presentation is

projective.

Proof. 3. 3. 6 and 3. 3. 7 (ii). o

(This can be proved ad hoc even for non-commutative rings.)

3. 3. 7. 2 (Corollary). A flat module of finite type over a

Noetherian ring is projective. o

If spec (A) is connected (for example if A is integral or

local) and P is a projective A-module of finite type, the function
p h-•r is constant.

54
 If P is a projective A-module of finite type and p
is constant, we call its constant value the rank of P.

3. 4 SUBMODULES OF FRACTION ALGEBRAS

3. 4.1 (Theorem). Let M be an A-module of finite type. The

following two conditions are equivalent:

(i) M is projective of rank 1

(ii) M ® . N is isomorphic to A for some A-module N.

Moreover if (ii) holds then N is isomorphic to the dual

M* of M.

Proof. Suppose (i) holds and let M® M* - A be the natural map.

iso
Let m be a maximal ideal of A. Then (M®AM*) •
A Hi
M b
M
m ®A ( *)m y ^ ^ H^ — ®T (whose inverse is
III A HI S S -L

— g> z !_• m y ). However M is of finite presentation and A m

S t St HI
is a flat A-algebra; thus by 2. 3. 4 (M*) is isomorphic to the
dual ( M m )* of the A m -module M m . Thus M^M*-A
localises to the natural map: M m ®. m
(M m )* -* A m but
iso
by 3. 3. 7 (ii) M m is free of rank 1; so M m ® A (M m )* >.
iso
Am and by 3. 2.1.1 M® A M* • A and (ii) holds.
Suppose (ii) holds. By 3. 3. 7 (iii) it is enough to prove that
M is free of rank 1 under the assumption that A is local, with
Tso
m = m (A) and k = A / m , say. Tensoring M8>N

55
 iso
by k we obtain M/ ttt M &UST/ m N •k; thus M/ m M has
rank 1 as a k-module: say m + ttl M bases M/ m M. Let
Am - M - Q - 0 be exact: then k ^ A m - k ^ M - k®AQ - 0
is exact; but the diagram

k®AAm

iso ISO

onto
Am/ m m • M/ m M

commutes; and so k ® A Q = 0 . But Q is of finite type, and thus

onto
Q = 0 by 3. 3. 5 and M = Am. Thus A •M by X h-^ Am.
Suppose Xm = 0; and let li—• m® n under A -* M&.N; so
iso
X h-*-Xm®n=0 and X = 0. Thus A • M and M is free of
rank 1. Therefore (by 3. 3. 7 (iii) as already explained) (i) holds.
Finally the A-modules N, A® A N, M®AM* ® A N,
M*, A&.M* and M* are isomorphic. o

It is almost as if projective modules of rank 1 formed a

group under 8) A with identity A, and inverse M* of M. However
they form a proper class C and not a set. Thus define the Picard
group of A to be an abelian group Pic (A) and a function
[ ]: C - Pic (A) such that

(i) M is isomorphic to N-^> [M] = [N]

(ii) [M® A N]= [M] + [N]

(iii) Pic(A) is universal for (i) and (ii).

56
Exercise. Construct such a Pic (A).

For the rest of this section A is a ring; S £ A is multi-

plicative; B = S" A; and AT is the image of A in S~ A.
If M and N are sub-A-modules of B we define the sub-
A-modules

MN = { Z m.n.: I finite; m. e M; n. e N )
! 1 l 1
iel

(M:N) = I b e B : bN c M } (ratio)

so that A'M = M and N(M:N) c M.

3.4. 2 (Lemma). Let M be a sub-A-module of B. Then the

following conditions are equivalent:

(i) y e M for some s e S

(ii) BM = B
bii
(iii) B® A M • B by b®m h-^bm .

1 s
Proof. Suppose (i) holds and let b € B. Then b = ( b - ) T e BM:
S -L
that is, (ii) holds.
Suppose (ii) holds and let us prove (iii). Plainly
onto inj
B® A M • B . Since B is a flat A-module, B®AM -B&.B,
A ' A A '
and it remains to show that B®.B • B . Note first that

1® 1 !®a Ig (aj 1 a
s 1 s 1 s V 1 s '

57
so that

> y) = l®xy

in the A-algebra B® A B. Thus if 2 x.y. = 0 for x.,y. e B,

A. _ 1 I 11
we have

2 x ® y. = 2 1 ® x.y. = 1 ® 2 x.y. = 0

.. a.
Finally suppose (iii) holds. Then -=-= 2 — m. for
i e l S i x

some a., m., s.; so j—2 6 M: that is, (i) holds.D

We call a submodule M of B which satisfies the conditions

of 3. 4. 2 non-degenerate.
We have M(A':M) c A1 for any submodule M of B. The
following conditions are equivalent:

(i) M(AT:M) = AT

(ii) jeM(A':M)

(iii) MN = AT for some submodule N of B;

and if (iii) holds, N = (AT:M). We call such a submodule M

invertible. If M is invertible, we have

BM = BBM 2 BM(AT:M) = BAT = B

so that M is non-degenerate. Plainly the invertible submodules

form an abelian group under multiplication, with identity AT, and

58
inverse (AT:M) of M.

3.4. 3 (Theorem). Let M be a non-degenerate submodule of B.

The following conditions are equivalent:

(i) M is invertible

(ii) M is a projective A'-module

(iii) M is a projective A'-module of rank 1.

Proof. Suppose (i) holds. Then T e M(AT:M) and

= 2 m.n. T
1 L
iel 1 1
for some finite I and m. e M and n. € (AT:M). Let 0:M — AT
by m i — ( n m ) and i//:Af - M by (£.). T I *2 |.m..
I i€i nei . € l i i
Then i / / ° 0 = l M and M is a projective AT-module of finite type.
iso
In particular M is a flat AT-module; and M® A B •
by m ® b i—• m 8)b; so we have the commutative diagram

onto
• AT

ISO

ISO
B

ISO
and M^ A ,(A f :M) •A'. Thus M is projective of rank 1 by
3.4.1: that is (iii) holds.

59
 Suppose (ii) holds. Then (x ) A bases M © L for some
A A €A
AT-module L. Let ( x A * be the associated projections:
M © L — A! and let x. = m. + z for m. e M and z. e L and
A A A A A
X e A. Now y € M for s o m e seS, s o that y = 2 X>(?)m\
1 1 N A A 1 A
X € A
1 1 s t
and T = 2 - XxCy)111^ Let x e M. Then T x €
AT for some
1
AeAs H A l

t c S and we have - xx(f) x = -^ xx(f) y x

= Xx(x)

<EA'

e
for all X e A. Thus - XAT) (AT:M) for all X e A and M is
s xi
invertible: that is, (i) holds, a

60
4. Supporting and associated prime ideals

4.1 LENGTHS AND RANKS OF MODULES

An A-module M is called simple iff M =£ {0 } and its only

submodules are {0} and M. For example, if m is a maximal
ideal of A, then A/ in is a simple A-module. Conversely, every
simple A-module is isomorphic to A/ m for some maximal tn .
An A-module M is called Artinian iff it satisfies the
following two equivalent conditions:

(i) every non-empty set of submodules of M has a minimal

element;

(ii) every decreasing sequence N 2 N 2 . . . of submodules

of M is eventually constant.

For example, a simple module is both Noetherian and Artinian.

As in 1. 3. 2, if N is a submodule of M, then M is
Artinian iff both N and M/N are Artinian.
A sequence (N.) n < . < of submodules of an A-module M
is called a Jordan-Holder sequence iff

o iC N c . . . c Nn = M

and N. /N. 1 is simple for i = 1, . . . , n. If M has a Jordan-

Holder sequence we say that M is of finite length.

61
4.1.1 (Proposition). (i) An A-module is of finite length iff it
is both Noetherian and Artinian.

(ii) Let M be an A-module and and

be Jordan-Holder sequences in M. Then m = n and the

N./N. 1 are isomorphic in some order to the L./L. , .

Proof, (i) Suppose M is Noetherian and Artinian. As long as

possible, choose N = 0; N to be a minimal element of the set
of all submodules 0; N to be a minimal element of the set of
all submodules => N ; and so on. Let N be the last possible
choice. Then (N.) n < . < is a, Jordan-Holder sequence for M.
The converse follows at once from 1. 3. 2 and its Artinian
analogue.

(ii) Using (i) we may suppose that M is the least submodule

of M for which the desired result is false. Consider the commuta-
tive diagram

L nN .
"m-2 m-ln n-1 n-2

L
m-2nNn-l

62
 If L - * N .Tf then L . + N , = M and M/L 7 ,
m-1 n-l m-1 n-1 m-1
(resp. M/N .) is isomorphic t o N - / L , n N , (resp.
L T / L T fiN , ) . Moreover the result is true for L n
m-I m-l n-1 m-1
and for N n and for L , nN * thus it is true for M.
n-1 m-1 n-1
If L n = N , the result is true for L n and
m-1 n-1 m-1
therefore for M. (If we work more carefully we do not need to
use choice in (ii).) D

4. 1. 1 (ii) allows us to define the length long.(M) of a

module M of finite length to be new for any Jordan-Holder
sequence ( N . ) n < . < in M. If N is a submodule of M,
clearly M is of finite length iff both N and M/N are of finite
length; and then:

longA(M) = longA(N) + longA(M/N) .

For example, if A is afield, the A-modules of finite

length are the finite-dimensional ones; and their length is their
rank.
Let A be an integral ring and k be its field of fractions.
Let M be an A-module. We define the torsion submodule T\(M)
of M as follows:

T\(M) = {m € M: Xm = 0 for some X e A* }

so that (3.1. 2) the sequence 0 — TA(M) -» M •* k® A M is exact;

and (3. 1. 3) the functor: M i—"-kg)AM (taking A-modules to k-
modules) is exact. We say that M is of finite rank iff k®.M
is a k-module of finite type; and we call the rank of k®.M the
rank r»(M) of M. This definition of rank agrees (where appropri-
ate) with the other ones we have given.

63
 For example M has rank 0 iff M is torsion; and Q has
rank 1 as a Z-module (though Q is not even of finite type as a
Z-module).
If M is an A-module and N is a submodule, the exactness
of M K—k®AM shows that M is of finite rank iff both N and
M/N are of finite rank; and that then r A (M) = r A (N) + r A (M/N).

Exercise. If A is an integral ring and not a field, an A-module

of finite length is torsion.

4. 2 THE SUPPORT OF A MODULE

Let M be an A-module: we define

supp(M) = { p e spec(A): M * 0} ,

the support of M.
For example by 3. 2.1. 1 supp(M) = 0 iff M = 0.

4. 2.1 (Proposition). Let M be an A-module.

(i) If N is a submodule of M, then supp(M) = supp(N) u

supp(M/N).

(ii) If M = Z M. for a family (MA A of submodules,

— A A A A €A

then supp(M) = U supp(M ).

A
X €A
(iii) If M is of finite type, supp(M) is closed in spec(A).

64
Proof, (i) Since 0 - N - M - (M/N) - 0 is exact,
we have M = 0 iff N = 0 and (M/N) =0.

(ii) is similar.

(iii) Let a = { X e A : X m = 0 for all m e M }, and let

^ m ^ l < i < n S e n e r a t e M-
Let p e V( a ) and p i supp(M). Then for some
s € A\ p c A\ a we have sm. = 0 for i = 1, . . . , n, so that
sM = 0 and s e a , a contradiction.
Conversely let p e supp(M) and X e a \ p . Then
Xm ± 0 for some m e M, a contradiction.
Thus supp(M) = V( a ). •

In particular supp(A/ a ) = V( a ) for any ideal a of

A.

4.2.2 (Proposition). Let M and N be A-modules of finite

type. Then

supp(M®AN) = supp(M) D supp(N) .

Proof. As in the proof of 3. 4.1 (M®AN) and M ®A N

are isomorphic A D -modules. Thus it remains to show that
M ®. N = 0 implies M = 0 or N = 0 on the additional hypothesis
that A is local. From the natural isomorphism:
(k® A M)® k (k® A N) - k® A (M® A N), where k is the residual
field of A, we deduce that k® A M = 0 (say), so that M = 0
by 3. 3. 5. D

65
4. 3 PRIME IDEALS ASSOCIATED TO A MODULE

In this section (4. 3) A will denote a Noetherian ring.

A (*) denotes results where this assumption is not in fact needed.
If M is an A-module and m e M we define the annihilator
ann(m) of m as follows:

ann(m) = | H A : Am = 0} .

We then define ass(M), the set of prime ideals associated to M,

as follows:

ass(M) = {ann(m) e spec(A): m e M} .

For example, if p £ spec (A) and M is a non-zero submodule of

A/ p , then ass(M) = { p }.

4. 3.1 (Proposition). Let M be an A-module.

(*) (i) Every maximal element of the set X = {ann(m): m e M

and m * 0 } belongs to ass(M).

(ii) M = 0 iff ass(M) = 0 .

(*) (iii) If N is a submodule of M, then

ass(N) c ass(M) c ass(N) u ass(M/N) .

Proof, (i) Let ann(m) e X be maximal. Since m ± 0 we have

1 i ann(m). Let x,y e A be such that xy e ann(m) but
x i ann(m). Then xm 4- 0 and ann(xm) e X. but ann(xm) 2
ann(m); thus ann(xm) = ann(m) and y € ann(xm) = ann(m).

66
Therefore ann(m) e spec(A) and ann(m) € ass(M).

(ii) Suppose M * 0. Then X ^ 0 and has a maximal element.

Therefore ass(M) ^ 0 by (i). The converse is immediate.

(iii) Clearly ass(N) c ass(M). Let p = ann(m) e ass(M).

If Am n N = 0 we have p = ann(m + N) e ass(M/N). If
Am nN^O, the isomorphism: A/ p — Am given by X + p •—•Am
shows that Am n N is isomorphic to a non-zero submodule of
A/ p , and thus { p } = ass(Am n N) c ass(N).o

(*) Exercise. ass( © M ) = u ass(M ) .

X €A A X eA A

(*) 4..3, 2 (Theorem). Let M be an A-module and X c ass(M).

Then there is a submodule N of M such that ass(N) = X and
ass(M/N) = ass(M)\X.

Proof. Let Y = {P a submodule of M: ass(P) Q x } . By ZornTs

lemma Y has a maximal element N. By 4. 3.1 (iii) it suffices now
to show that ass(M/N) Q ass(M)\X. Suppose p € ass(M/N);
then p = ann(m + N) for some m £ N, and

ass(Am + N) c ass(N) u ass(Am + N/N)

= ass(N) u ass(A/ p )
- ass(N) U { p } ;

but Am + N 3 N; so p §f X and p e ass(Am + N) Cass(M) .D

67
4. 3. 3 (Proposition). Let M be an A-module and p € spec (A).
Then p € supp(M) iff p 2 q for some q e ass(M).

Proof. Suppose p 2 q for some q = ann(m) e ass(M). Then

(^-) € ass(M p )

so that by 4. 3.1 (ii) M * 0 and p e supp(M).

Conversely suppose p e supp(M); by 4. 3.1 (ii) there
exists $ = ann(^-) € ass(M ) and q = { q e A : ^ € ^ } =
ann(m) e ass(M) and p 2 q . o

Thus ass(M) c supp(M) and ass(M) and supp(M) have

the same minimal elements.

4. 3. 4 (Theorem). Let M be an A-module of finite type.

(i) There are sequences ( N . ) n < - < of submodules of M

and ( p l.)-,-L ^— i. ^— n in spec(A) such that
(a) 0 = N C N C . . . C N = M

(b) N./N. n a n d A / p . a r e i s o m o r p h i c f o r i = l , . . . , n.
1 1— X 1 —

(ii) K^ ( N . ) Q < . < and ( p • ) 1 < - < are such sequences, then:

ass(M) c { p ... , p } c supp(M) .

Proof, (i) Let X be the set of all submodules N of M for

which there exist sequences such that (a) and (b) hold with N in
place of M. By 1. 3. 2. 2 M is a Noetherian module; and
0 e X =£ 0; so there is a maximal element N of X. Suppose
N * M: then ass(M/N) ± <j) by 4. 3. 1 (ii) and there is a

68
 p = ann(m + N) e ass(M/N) for some m £ N. But Am + N/N
is isomorphic to A/ p and Am + N e X, a contradiction. Thus
MeX.

(ii) By 4. 3. 1 (iii) ass(M) c { p ^ . . . , p n }. Also

p . e V( p .) = supp(A/ p .) = supp(N./N._1) c supp(M)

by 4. 2. 1 (i). D

4. 3. 4,1 (Corollary). ass(M) is finite, a

4, 3. 4, 2 (Corollary). The following conditions are equivalent:

(i) M is of finite length;

(ii) if p e ass(M), then p is a maximal ideal of A;

(iii) if p € supp(M), then p is a maximal ideal of A.

Proof. Suppose (i) holds. Then M has a Jordan-Holder sequence;

and by 4. 3. 4 (ii) ass(M) c { p , . . . , p } for some
p . e spec(A) with A/ p . simple, so that p , . . . , p are
maximal and (ii) holds.
That (ii) implies (iii) follows at once from 4. 3. 3.
Suppose (iii) holds. By 4. 3. 4 (i), (ii) there is a sequence
(N-)n^-^
v such that 0 = N C N c . . . c N = M and there are
rO<i<n oi n
p . e. supp(M) such that A / p . and N./N. 1 are isomorphic for
i = 1, . . . , n. But p . is maximal and thus A/ p . is simple
for i = 1, . . . , n: so ( N . ) n < . ^ is a Jordan-Holder sequence
and (i) holds, Q

69
4. 3. 4. 3 (Corollary). If M is of finite length then ass(M) =
supp(M) and each has at most long,(M) elements, a

4. 3. 5 (Lemma). Let A be an integral (Noetherian) ring such

that if p € spec(A), then p = 0 or p is maximal.

(i) K^ M is a torsion A-module of finite type, then M is of

finite length.

(ii) If M is a torsion-free A-module of finite rank, and

a € A*, then M/aM and A/Aa are of finite length and

longA(M/aM) < r A (M) longA(A/Aa) .

Proof, (i) Since M is torsion, 0 £ ass(M), so that ass(M)

consists of maximal ideals. Thus M is of finite length by 4. 3. 4. 2.

(ii) Let N be a submodule of M of finite type. Then

r A (N) < r A (M) and N - ^ k ® A N , where k is the field of
fractions of A; so N has a free submodule L of rank r.(N);
and N/L is torsion and of finite type, thus of finite length by (i).
Also by (i) L / a 1 ^ and N/a n N are of finite length for n e w ;
and we have exact sequences

0 - L / a ^ n L - N/a n N - N/
/(n(N/L) - 0

L/a n L - L/a n N n L - 0

from which

long A (N/a n N) < long A (L/a n L) + long A (N/L)

70
 ISO 2

However N is torsion-free so N/aN • a N / a N by multiplica-

tion by a; so that long A (N/a n N) = n long.(N/aN) and

longA(N/aN) < long A (L/aL) + ± longA(N/L)

for n ^ 1. Taking n > longA(N/L) we have

longA(N/aN) < long A (L/aL) .

But L is free and long A (L/aL) = r A(N) longA(A/Aa) directly;

A. A A
so that

longA(N/aN) < r A (M) longA(A/Aa) .

Suppose now that (N.)n < . < is a sequence of submodules

of M such that

aM=N CN C...CN = M.
0 1 n

n
Let m. e N . \ N . for i = l , . . . , n . Let N = Z Am. and
x 1 1
\ " i=l x
M . = (( I A m . ) + a N ) / a N f o r i = 0 , . . . , n . Then
1 J

0 = M C MC . . . C M = N/aN
oi n

and n < rA(M) longA(A/Aa). Thus M/aM is of finite length

< rA(M) longA(A/Aa). D

71
5. Integers

5.1 DEFINITION OF INTEGERS

Let A be a ring and A[X] be the A-algebra of polynomials

f (X) over A in an indeterminate X. The A-algebra A[X] has
the following universal property: for every A-algebra B and
every x e B there is a unique A-algebra morphism: A[X] — B
such that X i—• x, namely a X n + . . . + a n-^a x11 + . . . +
a x + a 1. We write f(x) for the image of f(X) under this
morphism.
Similar remarks apply to the A-algebra A[X , . . . , X ]
of polynomials in n indeterminates over A.

Exercise. Solve the same universal problems in the category

of all rings, not necessarily commutative (but with ones).

A polynomial f (X) e A[X] is called unitary iff

f(X) = X n + a ^ " 1 + . . . + a n for some n > 0 and a^ . . . , a € A.

5.1,1 (Lemma). Let f (X) e A[X] be unitary.

(i) Let^ g(X) e A[X]. Then there exist unique q(X), r(X) e A[X]
such that deg(r(X)) < deg(f(X)) and:

g(X) - f(X)q(X) + r(X) .

(ii) Let a € A. Then X - a divides f (X) iff f (a) = 0.

72
Proof. Trivial, a

Let B be an A-algebra and x e B. We write A[x] for

the least subalgebra of B containing x: so that

A[x] = {f(x) e B : f (X) € A[X]} .

Similarly we define A[x, y], etc.

5.1. 2 (Theorem). Let B be an A-algebra and x e B. The

following statements are equivalent:

(i) f(x)= 0 for some unitary f(X) eA[X];

(ii) A[x] is an A-module of finite type;

(iii) there is a subalgebra C of B such that

(a) x e C,
(b) C is an A-module of finite type.

Proof. Suppose (i) holds and let f(X) have degree n. Then
(x 1 ^ < . ^ generates the A-module A[x].
Plainly (ii) implies (iii).
Finally suppose (iii) holds. Let (x.)-, < . < generate
the A-module C. Then there exist a.. eA with
n
xx. = Z a..x. for i = 1, . .. , n. In A define 6.. = 0 if i * j
i j = 1 13 3 13

and 6.. = 1. In C let (j3..) (resp. y) be the adjoint

11 l j J. — 1 > 3 — *1
(resp. determinant) of the matrix (6..x-a..1), . .< . Then
13 13 1 — I? 3 — *i

73
 n
.- 2 a x )

= 0

for k = 1, . . . , n. Thus yz = 0 for all z e C, and in particular

y = yl = 0. But y - det((5 i j x-a..l) 1 ^ . ^ ) SO that y = f(x) for
some unitary f (X) € A[X] and (i) holds, D

5.1. 2.1 (Corollary). Suppose also that A is Noetherian. We

have the further equivalent condition:

(iv) there is a submodule M of B of finite type with

A[x] c M.

Proof. 1. 3. 2. 3. D

If x e B satisfies the conditions of 5.1. 2 we say that x

is integral over A. If x is integral over A for all x e B, we
say that B is an entire A-algebra.
For example, the Gaussian integers comprise an entire
Z-algebra; but Q is not an entire Z-algebra. Again, if A and

74
B are fields, and B is an A-algebra, B is an entire A-algebra
iff B is an algebraic extension of A.
Note that if an A-algebra B is an A-module of finite type,
it is entire; for example, a finite field extension is algebraic.

5.1. 2. 2 (Corollary). Let x, y e B be integral over A and

a e A . Then x + y, xy and ax are integral over A.

Proof. Let (b.). -. (resp. (c ) J generate A[x] (resp. A[y])

1 161 J Jt J
where I and J are finite. Then (b.c.)/4 ^ T y T generates
A[x,y]; and x + y, xy, ax € A[x,y] are integral over A by
5. 1. 2 (iii). D
By 5.1. 2. 2 the set Af of all x e B integral over A is
a subalgebra of B. We call AT the integral closure of A in B.
If AT = Al we say that A is integrally closed in B. Note that

(i) Ar is the largest entire subalgebra of B;

(ii) B is entire iff B = A\

We say that a ring A is integrally closed iff A is

integrally closed in tot(A). For example Z is integrally closed;
and a field is integrally closed.

5.1. 2. 3 (Corollary). Suppose B is an entire A-algebra and C

is an entire B-algebra. Then C is an entire A-algebra.

Proof. Let x e C. Then x11 + t^x11"1 + .. . + t>nl = Q for

b eB A s in 5 l 2 2 A is an
some b , ..., n - -- - DV *"'' V
A-module of finite type; and A[b1, . . . , b j [x] is an A[bx, . . . ,b n ]-
module of finite type, being generated by ( x l ) o < i < n ' s o

75
A[b , . . . , b ] [x] is an A-module of finite type, and
x e A[b , . . . , b ] [x], so that x is integral over A by
5. 1. 2 (iii). D

It follows at once that (A?)T = Af, so that AT is the

smallest entire subalgebra of B such that AT is integrally closed
in B.

Exercises, (i) If B and C are entire A-algebras, so is

(ii) An integral ring with unique factorisation to irreducibles

is integrally closed.

5.1. 3 (Proposition). Let B be an A-algebra and AT be the

integral closure of A in_ B. Let S ^ A be multiplicative.
Then S~1AT is the integral closure of S"1A in S~1B.

Proof. Let - e S ^ B be integral over S"XA. Then y is

S -L
integral over S~ A and there exist a , . . . , a e A and r e S
such that
, n a , n-1 a .

Thus for some t e S we have

t(rb n + a b 11 " 1 + . . . + a 1) - 0
l n

in B; so that

76
 (trb)n + a^trb)11"1 + . . . + a ^ ' V l = 0

and trb e A'. Thus - e S^A' .

S n
r n 11
Conversely if b eA and s e S, we have b + ab " + . . . +
a
a 1 = 0 for some a , . . . , a e A; so that
n an , n - 1 a
V sV n1
s

and - is integral over S~ A. a

s

5.1. 3.1 (Corollary). Let A be an integral ring. Then A is

integrally closed iff A is integrally closed for every maximal
ideal ttt of A.

Proof. 3. 2.1. 3. D

5.1. 4 (Lemma). Let A be a ring.

(i) Let f(X) e A[X] be unitary. There is an A-algebra B with

A - ^ « - B and b e B such that f (b) = 0 (so that X - b divides
f(X) in B[X] by 5.1.1 (11)).

(ii) Let C be an A-algebra. Let f(X), g(X) e C[X] be unitary.

Then the coefficients of f(X) g(X) are integral over A iff those of
both f(X) and g(X) are integral over A.

Proof, (i) Let f(X) generate the ideal a of A[X] and let
B = A[X]/ a and b = X + a . Then f (b) = f (X) + a = 0 and
inj
it remains to show that A • B . Suppose a € A* is such that

77
a € a ; then a = f(X) h(X) for some non-zero h(X) eA[X] and
a has degree > 1 a s a polynomial in A[X], a contradiction.

(ii) Suppose that the coefficients of f(X) g(X) are integral

over A. Using (i) several times We construct a C-algebra B
with C • B and such that

n
f(X) = II (X - b.)
l

m
g(X) = II (X - d.)
J

for some b., d. e B. Thus the d. and b. are integral over A

by 5.1. 2. 3, and the coefficients of f(X) and g(X) (regarded as
polynomials in B[X]) are integral over A. But C • B ; so
the coefficients of f(X) and g(X) (regarded as polynomials over
C) are integral over A.
The converse follows at once from 5. 1. 2. 2. D

For example, let A be an integral ring; k be its field of

fractions; and K be an algebraic field extension of k. Thus if
x € K there is a unitary f (X) e k[X] of least degree such that
f(x) = 0; we call f(X) the minimal polynomial of x over k. If
now x is integral over A, then g(x) = 0 for some unitary
g(X) 6 A[X]; and by 5.1. 1 g(X) = f(X) h(X) for some h(X) € k[X],
Thus the coefficients of f(X) are integral over A by 5. 1. 4. In
particular, if A is integrally closed in k, then x e K is integral
over A iff f (X) e A[X].

78
5.1. 5 (Proposition). Let B be an A-algebra and f(X x , . . . , X n ) e
B[X , . . . , X ]. Then f is integral over A[X , . . . , X ] iff its
l n * J- n •
coefficients are integral over A.

Proof. The general case is an immediate induction from the case

n = 1.
Suppose then that f(X) e B[X] is integral over A[X]: say
H(f(X)) = 0 where

H(Y) = Y m + F ^ Y ™ " 1 + . .. + F m (X) e A[X][Y] .

Let n be strictly greater than the degrees of f(X) and of the

F (X) and let f^X) = f(X) - X n . Thus -f^X) is unitary and
Hl(f1(X)) = 0 where

H (Y) = H(Y + Xn) = Y m + G (X)Y m " 1 + . . . + G (X)

m

so that G (X) = H(Xn) is unitary. We have

m

= Gm(X)

and the coefficients of G (X) lie in A. Thus by 5.1. 4 (ii) the

m
coefficients of f (X), and therefore of f(X), are integral over A.
The converse is immediate from 5.1. 2. 2.o
5.1. 5.1 (Corollary). Let A be an integral ring. Then A is
integrally closed iff A[X , .. . , X ] is integrally closed.

Proof. Suppose first that A is a field. Then A is integrally

closed; and A[X , . .. , X ] has unique factorisation (see last

79
exercise) and is therefore integrally closed.
Suppose then that k is the field of fractions of A and that
A is integrally closed. Then k(X i , . . . , X ) is the field of
fractions of A[X , . . . , X ]. If x e k(X , . . . , X ) is integral
over A[XX, . . . , X j , then x e kfX^ . . . , X j by the first
paragraph, so that x e A[X , . . . , X ] by 5.1. 5. a

5. 2 INTEGERS AND PRIME IDEALS

Let f :A -* B be a ring morphism; p e spec(A); and

$5 e spec(B). We say that $J lies over p iff f'l[ | J ] = p :
in other words, iff p is the image of | J under the map:
spec(B) -* spec(A).
In particular, if A is a subring of B, $ e spec(B) lies
over p 6 spec (A) iff J n A = p .

5. 2.1 (Lemma). Let f :A — B be a ring morphism and

p € spec(A). Then there is a | J e spec(B) lying over p iff

Proof. Suppose | J € spec(B) and f" 1 [^P]= 9 . Then

f
[ 9 ] ^ $ s o t h a t p B C J , Thus 9 c f'l[ p B] c £
and f~1[ jpB] = p .
Conversely suppose f" 1 [ p B ] = p and let S = f[A\ p ].
If s € A\ p and f(s) e p B, then s e p , a contradiction. Thus
S n p B = 0 and by 1. 2.1 there exists | J e spec(B) with
J n S - j ^ and J 3 p B. Plainly f " 1 [ S ] = 9 • °

In particular, if A is a subring of B, and p e spec(A),

there is a ^ € spec(B) lying over p iff A n p B = p ; and

80
there is no $ e spec(B) lying over p i f f A n p B = > p .

5. 2. 2 (Proposition). Let B be an entire A-algebra; p e spec (A);

and <P e spec(B) lie over p . Then p is maximal iff 3$ is
maximal.

Proof. Since A / p HB/|J and B / | J is an entire A/ p -

algebra, we may suppose that p = 0 and JJ = 0 and A •B,
and must show that A is a field iff B is a field.
Suppose A is a field and let x e B * . Then x11 + a xn~ +
. . . + a 1 = 0 for some a , . . . , a e A and least n e w . But
n ,i' ' n
a * 0, for otherwise xn~ + a x11"2 + . . . + a - , 1 = 0 ; thus
there is an inverse

of x in B and B is a field.
Conversely suppose B is a field. Let x € A*. Then there
is an inverse x" 1 of x in B; and x" n + a x" n + ... + a 1 = 0
for some a , . . . , a e A; so x" = -(a 1 + . . . + a x11"1) e A,
1

and A is a field, D

5. 2. 3 (Theorem). (Cohen-Seidenberg). Let B be an entire

A-algebra and A — B be injective. Let p e spec (A). Then there
exists $5 e spec(B) lying over p .

Proof. By 5.1. 3 the A -algebra B is entire; and by

ini P P
3.1. 3 A — ^ B . Thus B * 0. Let m be a maximal ideal
P P P
of B . By 5. 2. 2 m lies over j A in A and therefore
over 9 in A. Let m lie over $3 in B. It follows from applying
spec to the commutative diagram

81
that ip lies over p in A. a

5. 2. 4 (Lemma). Let B be an entire A-algebra and A - B be

injective. Suppose that B is an integral ring. Let J5 e spec(B)
lie over p € spec(A). Then J = 0 iff j = 0.

Proof. Let x e $ and x * 0. Then x11 + a^x11""1 + . . . + a 1 = 0

for some a , . .. , a € A. and least n € w. Thus n > 1; and
n -| ———A
a * 0, for otherwise x " + a x + ... + a . , 1 = 0 ; and
a e p .
F
n
The converse is immediate, a

5. 2. 4.1 (Corollary). Let B be an entire A-algebra and

p e spec(A). Then the elements of spec(B) which lie over p
are pairwise incomparable by inclusion.

Proof. Let $ c <Jg in spec(B) lie over p . Then

inj
A/p **B/|J and <©/$ lies over 0 in A/p . Thus

5. 2. 5 (Lemma). Let A be a ring and p , . . . , p e spec(A).

Let q e spec(A) with q $. p . for i = 1, i .. . , n. Then
n there
Let q e spec(A) with q $ p . for i = 1
exists x e q such that x ft p . for all i.

82
Proof. We use induction on n. Thus we may suppose n > 1 and
that for each j there exists x. e q with x. £ p . for i ± j .
If x. £ p . for some i, there is nothing more to prove. Suppose
1 1 n
then that x. e p . for all i. Let x = x + II x.. Then x e q ,
1 l
x n i=21
and x £ p since II x. / p and x e p ; and x ft p .
X . Q 1 J. -L -L 1

~~ n
for i = 2, . . . , n, since x f. p . and II x. € p .. •
i i i 2 i i

Let A be an integral ring integrally closed in its field of

fractions k. Let K be a finite Galois extension field of k with
Galois group r . Let B be the integral closure of A in K. If
a e T, then a[B] Q B: and we have:

B = 1[B] = afcr'^B]] c a[B] c B

so a [ B ] = B . Let p e spec (A) and $ e spec(B) with

An f = p ; then a [ | } ] e spec(B) and A n a [ f ] = p too.
Moreover we have a converse result:

5. 2. 6 (Theorem). In the situation just described, T acts

transitively on the set of all ^ € spec(B) lying over p .

Proof. Let $ , © e spec(B) lie over p and suppose <x[ j j ] =

for all a € r . By 5. 2. 4. 1 <© 4 a [ ^ ] f o r a 1 1 a € r - T h u s bY
5. 2. 5 there exists x e 0 such that x £ a[ ^ ] for all a e T.
Therefore a(x) £ ^ for all a e T and n a(x) = y £ ^ .
a € r

83
However y = ( IT a(x)) x € © ; and y e k n B = A; so

y e (fi n A = || £ $ , a contradiction, a

Exercise. What is the analogue of 5. 2. 6 if K is a purely

inseparable field extension of k ?

84
6. Some geometrical results

We say that an A-algebra B is of finite type iff

B = A[x , . . . , x ] for some x , . . . , x e B . For example, an
entire A-algebra is of finite type iff it is of finite type as an A-
module.
Let k be a field and A be a k-algebra. We say that a
family (x.) . in A is algebraically free iff the unique k-algebra
A A til

morphism: k[X ] . - A such that X i—• x is injective (so

that k[X ] ^ and k[x ] A are isomorphic k-algebras).

6.1 (Theorem) (Noether's normalisation lemma). Let k be a

field and suppose A is a k-algebra of finite type. Then there is an
algebraically free family (x.)-, < . < in A such that A is an
entire k[x , . . . , x ]-algebra.

Proof. Let ( y ^ ^ ^ m in A be such that A = k[y 1 , . . . , y m ] .

We shall use induction on m. If m = 0 there is nothing to prove.
iso
Let m > 0. Then k[Y , . . . , Y ] / a • A for some
ideal a of k[Yx, . . . , Y m ]. K a = 0 we take ( x . ) 1 < i < n =
( y i H < i < m ' S u P P ° s e therefore that f(Yx, . . . , Y m ) e a \{0 }.
There is a sequence (N.)-, < . < , in w which increases so
rapidly that the leading term of the polynomial

85
 N
i
belongs to k*. Let z. = y. - y for i = 1, . . . , m-1. Then:

N N
l-l

= o

so that y is integral over k[z , . . . , z 1 ] , Therefore

y , .. . , y -, are also integral over k[z , . . . , z , ] and A
is an entire k[z , .. . , z , ]-algebra by 5.1. 2. 2. However
k[z , . . . , z ] is an entire k[x , . . . , x ]-algebra for some
algebraically free family (x.), < . < in k[z , . . . , z .. ] by the
inductive hypothesis; thus by 5. 1. 2. 3 A is an entire k[x , . . . , x ]
algebra, o

6.1.1 (Corollary), n < m. D

6.1. 2 (Corollary). (Weak Nullstellensatz). Let k be a field

and A = k[X , . . . , X ]. Let p e spec(A). Then p is maximal
iff A / p is a k-module of finite type.

Proof. Suppose p is maximal. Then by 6.1 A/p is an entire

k[y , . . . , y ]-algebra for some algebraically free family
(y.)-, < . < in A / j . Thus by 5. 2. 2 k[y , .. . , y ] is a field.
But k[y i , . . . , y m ] is isomorphic to k[Y i? . . . , Y m ], so
m = 0 and A/p is an entire k-algebra of finite type. Thus A/p
is a k-module of finite type, as we remarked above.
Conversely if A / p is a k-module of finite type it is a
field by 5.2.2.0

86
APPLICATION TO ALGEBRAIC CLOSURES OF FIELDS

If (M.) . is a family of A-modules, we define the

tensor product ® M as we should expect; but if (B ) A is
N A A A A €A
A € A
a family of A-algebras, we also define the restricted tensor product
res
® B to be the submodule of ® B generated by
A
A €A A eA X
® b for families (bA A e U B^ with bN = 1 for all but
N A A A A 6A . A A
A € A A € A

a finite number of X e A; and we give ® B, the obvious

X
X eA
res
A-algebra structure. (Clearly ® B with the natural maps:
A
X eA
BA •* ® Bx for each A is the sum of (B.). A in the category
A N » A A A t A

of A-algebras.)
Now let k be a field and define

A= © r e s k[x . . . , x ]/m .
n e to
m maximal
By 6. 1. 2 and 5. 1. 2. 2 A is an entire k-algebra; and if L is a
finite extension field of k, there is a k-algebra morphism: L -* A.
By flatness A is non- zero. Let J5 be any prime ideal of
A. Then by 5. 2. 2 k — A / | J is an algebraic closure of k.
(Compare any other construction of algebraic closures.)

6.1. 3 (Corollary) (HilbertTs Nullstellensatz). Let k be a

field; A = k[X , . . . , X ]; a be an ideal of A; and x e A.
l n
Suppose that for every finite extension field L of k and every
k-algebra morphism </>:A •* L with a Q Ker 0 we have </>(x) = 0.
Then x11 e a for some n e to.

87
Proof. We may suppose x ± 0.
In the A-algebra B = k[X]L, . . . , X n , X n + 1 ] let
b = Ba + B(l - xX +1 ) . Suppose b c B. Then b Q m
for some maximal ideal m of B; and B/tn is a finite extension
field of k by 6 . 1 . 2. Let </>:A -* B - B / m be the natural map.
Then a Q Ker 0 so 0(x) = 0 and 0(1) = 0(1 - x
0(x)0(X +1) = 0, a contradiction.
Thus b = B and

for some f. € a and g, g. € B. Now work in k(X , . . . , X , , )

-i l l x n > JL
and put - for X + , to obtain

so that x e a for some n eo).D

88
7. Valuation rings

7.1 ORDERED GROUPS

A partially ordered abelian group (G, +, <) is an abelian

group (G, +) with a partial order < on G such that:

x < y =-£> x + z < y + z

for all x, y, z e G.
We say that G is replete iff for all n e o> with n > 0 we
have:

x e G and nx > 0 => x > 0 .

For example, if G is totally ordered, it is replete.

7.1.1 (Lemma). Let G be a replete partially ordered torsion-

free abelian group; let H be a totally ordered subgroup; and G/H
be torsion. Then G is totally ordered.

Proof. Let x € G with x * 0. Then n x e H and nx * 0 for some

n > 0. Thus (say) nx > 0 and x > 0. o

Now let A be an integral ring and k be its field of fractions.

We make k*/ u (A) into a partially ordered abelian group under
multiplication by:

89
 x u (A) < y u (A) iff xy" 1 £ A .

7,1. 2 (Lemma). If A is integrally closed in k, then k*/tt (A)

is replete and torsion-free.

Proof. If x n e u (A), then x e n (A); if x11 € A \ u (A), then

X € A \ tt (A). D

7. 2 VALUATION RINGS

7. 2.1 (Proposition). Let A be an integral ring and k be its

field of fractions. The following statements are equivalent:

(i) k*/ tt (A) is totally ordered;

(ii) x €k\A ^ x " 1 € A;

(iii) the ideals of A are totally ordered by inclusion.

Proof. Immediate, o

We call such a ring A a valuation ring (in k, if we wish

to specify its field of fractions). For example, Q and Z „ for
p prime are valuation rings in Q. Note that (ii) already implies
that k is the field of fractions of A. By (ii) A is integrally
closed in k; and by (iii) A is a local ring.
We call k*/ tt (A) the value group of A and write | | for
the natural map: k* - k*/ tt (A) and define 10 | = 0. X h e n it
is plain that |x + y| ^ sup(|x|, |y|) for all x, y e k, so that if
|x| < |y|, then |x + y| = |y|. It follows that A is Bezout: for

90
if x, y € A, then (say) |x| < |y | and Ax + Ay = Ay.

Exercise. A local Bezout ring is a valuation ring.

Let K be afield extension of k and A (resp. B) be a

valuation ring in k (resp. K). We say that B extends A iff
B n k = A. (In any case it follows from 7. 2. 1 (ii) that B n k is
a valuation ring in k.) If B extends A, we find

A n m (B) = m (A)

k* n u (B) = u (A)

and have injections:

A/ m (A) - B / m (B)

k*/ tt (A) - K*/ tt (B)

preserving the field and ordered group structures. The dimension

of K(B) as a /c(A)-module is called the residual degree f(B|A)
of the extension; the index of |k* | as a subgroup of |K* [ is •
called the ramification e(B | A) of the extension.

7. 2. 2 (Proposition). If K is a finite (resp. algebraic) extension

field of k, so is K(B) of K(A); and | K * | / | k * | is a finite (resp.
torsion) group.

Proof. Suppose K is a finite extension of k. Note that

K*A* tt (B) and |K* |/|k* | are isomorphic. Let x , . . . , x r eK*
be in distinct cosets of k* tt (B). Then for any X , . .. , X ^ k *

91
 r
the values A.x. are all different: so = sup I X.xJ
1 1'
*0
1 1 x x
i-l i l
r
and I X.x. ± 0 . Thus r < | K:k | .

The other cases are similar, o

A valuation ring which is a field is called trivial. From

7. 2. 2, if K is an algebraic extension field of k and B in K
extends A in k, then B is trivial iff A is trivial.
If A is a valuation ring in k and k*/ tt (A) is infinite
cyclic we call A discrete. In this case there is a unique decreasing
isomorphism: k*/ tt (A) — Z, and we write ord for the induced
map: k* -* Z, defining also ord(O) = °°. Thus if a is a non-zero
ideal of A and ord(7r) = 1 and n = inf ord(x), then a = fl"nA.

Therefore A is a principal ring.

For example Z „ is a discrete valuation ring in Q and
p
an
ord(]5 P ) = n if a , b e Z and (a,p) = (b,p) = 1. Again, if
K = k(X i , . . . , Xn ) for a field k, we ^define the total degree
valuation ring A in K to consist of - for f , g e k[X , .. . , X ]
with g =£ 0 and deg(f) < deg(g). Clearly A is a discrete
valuation ring.
From 7. 2. 2, if K is a finite extension of k and B in K
extends A in k, then B is discrete iff A is discrete: for

inj e(B|A)
k*/ tt (A) -K*/ tt (B) - ^ = ^ - k * / tt (A) .
inj

Exercise. Let k be a finite field extension of Q. Show that

every non-trivial valuation ring in k is discrete, and that its
residual field is finite.

92
7. 2. 3 (Theorem). Let A be an integral ring; k be its field of
fractions; and A c k. The following conditions are equivalent:

(i) A is a discrete valuation ring;

(ii) A is a Noetherian local ring and m (A) is principal;

(iii) A is a Noetherian valuation ring;

(iv) A is a Noetherian; A is integrally closed in k; and

card(spec(A)) = 2.

Proof. We have seen that (i) implies (iii). Since a valuation ring
is Bezout, (iii) implies (ii). If (i) holds and p e spec (A) with
p ^ 0, then p = 77 A for some n > 0 with 77 € A with
ord(7r) = 1; and if n > 1, we have 7777 e p but 77, 77 ft p ;
so that n = 1 and (iv) holds.
Suppose (ii) holds, with m (A) = TTA say. Let y e n n A.
new
Then for all n e oo y = 77 x for some x e A. If y ^ 0 we have
J J
n n
x = 77x , n € A* for all n e w , so that Ax c Ax c Ax c . .
?
n n+1 0 1 2 '
a contradiction. Thus n 77 A = 0. Let x e A*. Then
n eo)
x £ 77 A for some least n e w and 77~nx e u (A). Thus 77 u (A)
generates k * / u (A); and 77 u (A) < 1; so that k * / u (A) is
totally ordered and infinite cyclic; that is, (i) holds.
Finally, suppose (iv) holds. Then A is local and
c
ttl (A) = m * 0. We have (A: m ) y" 1 A for any y e m n A*
(where (A: m ) is defined as in 3. 4 with B — k). Thus by
1. 3. 2. 3 (A: m ) is an A-module of finite type.
Let x € tn and x =£ 0. Suppose that n is a maximal ideal
of A . Then x ^ n so n n A = 0 . Let - e n ; then
x n
x

93
ye n n A = 0 and - ^ = 0. Thus n = 0 and A = k. Now
n x
x
xn
let z e m n A*. Then — e A for some n e w , and x c zA.
z
Thus (since in is an A-module of finite type, and using the multi-
N
nomial theorem) m £ zA for some least N € a>. Let
y € m N " 1 \ z A . Then ^ e (A: m )\A and (A: m ) => A.
z
We have m £ ttl (A: m ) Q A, so that m (A: m ) is
ttl or A. Suppose m (A: m ) = ttT and let x e (A: m ). Then
n
x m £ lit and x e (A: m ) for all n € a>. Let a. = A + Ax +
1
. . . + Ax for i e co. Then a c a c . . . c (A: m ) so that by
1. 3. 2. 3 a = a n for some n e a?. Thus x = b + b x + . . .
n n-1 oi
+ b ,x and x e A . That is, (A: m ) Q A, a contradiction.
Thus m (A: m ) = A: that is, m is invertible, so that m = 7rA
for some ir € A by 3. 4. 3 and 3. 3. 6. That is, (ii) holds, a

7. 3 EXTENSION THEOREMS

Among all pairs (A, p ) of subrings A of a field k and

p e spec(A) let us define the partial order < of domination:

(A, p ) < (AT, p f) iff A c AT and A n p T = p .

Thus the injection: A / p -- A T / p T

induces an injection: L — LT
of their fields of fractions and we define as follows the partial
order < * of strong domination:

(A, p ) <* (A', p ' ) iff:

94
(i) (A, p ) < (A', p')

(ii) Lf is an algebraic extension field of L.

In particular, if A and B are local rings in k, we define A < B

(resp. A < * B ) iff (A, m (A)) < (B, m (B)) (resp.
(A, m (A)) <* (B, ttl (B))).

7. 3.1 (Lemma). Let A and B be local subrings of a field k

and A < B. Suppose that A is a valuation ring in k. Then
A = B.

Proof. Immediate, o

7.3.2 (Lemma). Suppose (A, p ) < (AT, p T) in k. Then

(A, p ) < * (A", p ") for some (A", p ") in k.

Proof. Let f:Af — Lf be the natural map.

If Lf is an algebraic extension of L there is nothing to
prove. Thus we may suppose that there is a t e AT such that
f(t) e Lf is transcendental over L. Let g:L[f(t)] -*L be the
L-algebra morphism: f (t) i—•O. Then we take A" = Aft] and
p " = Kerfeof | A r t 0 , so that (A, p ) < (A", p ") and L" = L. a

Thus the maximal pairs in k with respect to < and to

<* are the same.

7. 3. 3 (Theorem) (Chevalley). Let (A, p ) be a pair in k. Then

there is a valuation ring B in k with (A, p ) <* (B, m (B)).

95
Proof. By Zorn's lemma, among all pairs >* (A, p ), ordered
by <*, there is a maximal one, (B, q ) say. By 5. 2. 3, B is
integrally closed in k. Since (B , q B ) >* (B, q ), B is
local and m (B) = q .
It remains to show that B is a valuation ring in k. Suppose
x € k\B. By 7. 3. 2 and 5. 2.1 ( q B[x]) n B D q so that
1 e q B[x]. Thus l = q + q x + . . . + q x for some q. e q ;
and:

(qQ - I)" 1 q^x- 1 ) 1 1 - 1 + . . . + (qo - 1 ) ^ = 0 .

That is, x" 1 is integral over B: so that x""1

Thus the valuation rings in k are the local rings which are
maximal for the orders both of domination and of strong domination.

7. 3. 3.1 (Corollary). If K is a field extension of k and A is

a valuation ring in k, there is a valuation ring B in K extending
A.

Proof, There is a valuation ring B in K with (B, m (B)) > *

(A, m (A)). Thus B n k = A by 7. 2. 1 (ii). D

7. 3. 3. 2 (Corollary). Let A be an integrally closed subring in a

field k. Then A = n | B ; B 2 A and B is a valuation ring in k } .

Proof. Suppose x e k\A. Then x j^Afx""1] so x" 1 ft tt (Afx"1])

and x ft p for some p € spec(A[x~ ]). There is a valuation
ring B in k with (B, m (B)) > (Afx"1], p ); thus x" 1 e m (B)
and x ft B. o

96
 From 7. 3. 3. 2 we see that the integral closure of a sub-
ring A in a field k is the intersection of all the valuation rings
B in k such that B 2 A.

7. 3. 3. 3 (Corollary). If in 7. 3. 3. 2 A is also local, we may

further restrict B to dominate A.

Proof. We let x ek\A and x c ttl for some maximal ideal

ttl of A. The composed map: A -* A[x" ] -* A[x" ] / ttl is onto;
so its kernel, namely A n ttl , is maximal in A. Thus
A n m = Itt (A) and B taken as before dominates A. a

7. 3. 4a (Lemma). Let K be an algebraic extension field of k

and A be a valuation ring in k. Let B be the integral closure
of A in K, C a valuation ring extending A, and } = w (C) n B.
Then C = B .

Proof. As in 7. 3. 3. 2 we see B ^ Q C. Conversely suppose

x e C ; then for some a , . .. , a € A, not all zero:

a x n + . .. + a = 0 .
n o

Choose s as the largest integer such that 0 < s ^ n and

| a | = max |a. |; then putting b. = a./a and dividing through
S r t ^ ^ . ^ - l I I S
. s ,
byJ a x we have
s

v(b x n " s + . . . + b , . + 1); + - (v b .. + . . . + b - i - y )7 = 0

n s+1 x s-1 o s-1
x

with b^, . . . , b o + n e ttl (A) c ni (C) (since | a e | > |a. | if

n' * * ' ' s+1 \ o i l l l ' v ' iu_

97
s < i < n) and b ,,..., b e A (since | a | > |a. |). Write
S — J- U S 1

Vn'S +
• • • + b s+l + 1
= y' b
s-l +
• • • + VVx 8 " 1 ) = z>
so that y + - = 0. We show z e B, y e B\;P . Suppose D is a
x
valuation ring in K such that D 2 B. If x € D, then y e D and
so z = -xy e D; if - e D, then z e D and so y = - z . — e D. Thus
x x
in any case both y, z e D. Hence by 7. 3. 3. 2 y, z e B. Further,
as b n , . . . , b g + 1 € m (C), y t m (C), so y ^ $ . a

7. 3. 4 (Theorem). Let K be an algebraic extension field of k,

and A be a valuation ring in k. Let B be the integral closure
of A in K. Then the valuation rings in K which extend A are
precisely the rings B as ^ runs through the maximal ideals
of B.

Proof. If C is a valuation ring in K extending A, we have by

5. 2. 2 that f = B 0 m (C) is maximal, and by 7. 3. 4a C = B ^ .
Conversely for ^ maximal in B by 7. 3. 3, the local ring
B is dominated by a valuation ring C, and by 7. 3. 4a C = B f
for 3$T = m (C) n B 2 $ , so that $ T = $ . Q

7. 3. 4.1 (Corollary). Suppose in addition that K is a finite

Galois extension of k with Galois group T. Then T acts
transitively on the valuation rings in K which extend A.

Proof. Immediate from 5. 2. 6. a

7. 3. 4. 2 (Corollary). Suppose instead that K is a purely

inseparable extension of k. Then there is one and only one
valuation ring in K which extends A, namely B.

98
Proof. By 7. 3.1 it is enough to show that B is a valuation ring
in K. Let p be the characteristic of k and define B ! = ( x e K :
Dn
3T € A for some n e w } Q B. By (ii) of 7. 2. 1 we see that Bf
pn
is a valuation ring in K; and if r e A and x e k\A, then
x" <EA and x = x ^ x " ) ^ " € A; so B ' n k = A . But BT is
1 1 1

integrally closed in K: so BT = B. a

7. 3. 4. 3 (Corollary). Suppose that K is a finite extension of k.

Then there are only finitely many valuation rings in K which
extend A.

Proof. Fit 7. 3. 4.1 and 7. 3. 4. 2 together with standard field theory, a

7. 4 AN APPLICATION

7, 4.1 (Lemma). Let A be a principal ring and M be a free

A-module. Let L be a submodule of M. Then L is free.

Proof. Let (m R ) R base M, where a is an ordinal, and

let (x^M - A)g be the associated projections. Let
M^ = 2 Am and L = L n M . for all /3 e a, so that

L= u hp. For each pea choose x^ e L^ such that

jS € a * P P

(i) if x^L^] = 0, then x^ = 0 ;

(ii) if x^L^] * 0, then A x ^ ) = x ^ ] .

99
Let N o = 2 Ax c i. c Mfl. We shall show by transfinite
p
y c p y P P
induction that N^ = L^ for all /3 e a. Suppose then that N = L
for all y € jS, and let x e L . If x J L J = 0, then x e L =
N £ Ng for some y e /3. If X^L^] * 0, then x ^ x ) = YX^x^)
for some y e A, so that Xn(x - yx J = 0 and x - yx^ € L =
N £ N~ for some y € p. Thus x e N • and N., = L« for all
j3 € a .

Therefore L = Z Ax..

r
Suppose 2 ^xp = 0 for some |~ eA*, non-zero

and j8 € . . . € p Then

r
o = 2 % x^ (x ) = | x^ (x^ ) * o
K
1=1 i ^r ^i *r ^r r

a contradiction.
Thus (x J n bases L. a
P X^Q ^ u

(We used choice when we supposed that a was an ordinal.)

7. 4.1.1 (Corollary). If M is of finite type, so is L. a

7. 4, 2 (Theorem) (Samuel). Let k be a field and A be an

integral k-algebra of finite type. Let K be the integral closure of
k in A (so that K is a field). Then under multiplication
U (A)/K* is a free Z-module of finite type.

Proof. By 6.1 A is an entire K[x , . . . , x ]-algebra for some

algebraically free family (x.)-, < • < in A. Let L be the field

100
of fractions of A, so that L is a finite extension field of
K(x , .. . , x ). Let M be a splitting field for L over
K(x , . . . , x ) with Galois group T. Let B be the total degree
valuation ring in K(x , . . . , x ) and let C , . . . , C be its
extensions to M. These are finitely many and T is transitive
on them by 7. 3. 4. 1, 2, 3; and they are discrete by 7. 2. 2. Let us
define ord and ord. to correspond.
Map u (A) - Z m by

u H— (ord^u), . . . , ord m (u)) .

Plainly K* - {(0, . . . , 0)}. Conversely let x e tt (A) and

ord.(x) = 0 for all i. Let y , . .. , y be the distinct members
of {cr(x): o e T }, so that

g(X) = n (X - y ) = X r + a X1*"1 + . .. +
1 1

is the minimal polynomial of x over K(x , . . . , x ). (A little

care is needed to see this, especially in characteristic p.) Thus
ord(a.) > 0 for all i, for ord.(y.) > 0 for all i, j . However
a. c K[x , . . . , x ] for each i by 5. 1. 5 and the remarks after
5.1. 4: thus a. € K for all i and x e K*.
Therefore tt (A)/K* -~ Z m is a monomorphism of Z-
modules and tt (A)/K* is a free Z-module of finite type by 7. 4.1
and 7. 4. 1. l.o

Bibliography.
Schilling, Valuations (dull).
Ax and Kochen, Amer. J. Math. 87(1965), 605-730 (beautiful
but undigested).

101
EXAMPLES ON GENERAL VALUATIONS

These examples get harder and harder. Numbers 3 to 5

should be done consecutively, as should numbers 6 to 11.

1. Find all valuation rings of the field Q of rational

numbers. [Hint: consider the intersections of their maximal
ideals with Z. ]

2. Considering |x - y | as the distance between x and

y, show that every triangle is isosceles, and that every point
inside a circle is a centre for the circle, with the same radius.
[Hint: if |x| < |y|, then |x + y | = | y | . ]

3. If G is a group and F is a field, define F[G] to be

su that f" 1 F* is finite. On F[G]
the set of all maps f :G --* F such
make the following definitions:

(f + g)(a) - f(cr) + g(a)

(Af)(a) = Af(a)

(fg)(a) = I f(T)g(T-lo) ,
T eG

for f, g € F[G], X € F and a e G.

Show that F[G] is an F-algebra with a 1. [Hint: it might
be easier to inject G -* F[G] as follows: en—** f , where
f (a) = 1 and f (T) = 0 if r±a; then an element of F[G] is a
finite linear combination of elements of G, with natural definitions
of operations. ]
(F[G] is called the group algebra of G over F.)

102
 4. If G is a totally ordered abelian group, F[G] is an
integral domain.

5. If G is a totally ordered abelian group and F is a

field, there is a field k and a valuation | | on k, for which
k* | = G, and whose residue class field is F. [Hint: let k be
the quotient field of F[G]. ]

6. Let K/k be a field extension. We say that a family

(a) . of elements of K is algebraically independent over k if
and only if the (larger) family of all finite products (including the
empty product) of elements of (a ) . is linearly independent
over k. In other words, if and only if the a satisfy no polynomial
A
equation with coefficients in k.
Use ZornTs lemma to show that there exist maximal
algebraically independent families in K. Such a family we call
a transcendence base of K/k.

7. If L is obtained from k by adjoining a transcendence

base of K/k, show that K/L is algebraic.

8. We write [K:k] for the number of elements in a

transcendence base (it may not be finite: we make no distinction
among infinite cardinals). Show that [K:k] is well-defined (i. e.
independent of the choice of transcendence base). [Hint: use 7. ]

9. For this definition, we need not restrict K to be a

field. All it need be is a commutative k-algebra with no divisors
of zero. Elucidate this statement, and show that if in this wider
situation, p is a prime ideal of K, then

[K:k] > [ p :k] + [K/ p :k] .

103
 10. If G is an abelian group, then the set {g , . . . , g } Q G
r
is called independent if and only if 2 n.g. = 0 for n. e Z implies
t=l 1 x l

that n. = 0 for all i. The supremum of the r for which there is

an independent set of r elements is called the rank of G. Show
that the rank of Q* is °° .

11. Now let K/k be a field extension, and | be a

valuation on K. Let KA be the corresponding residue class field
extension; show that p

[K:k] > rank(|K* |/|k* |) + [K:k] .

[Hint: in 9 take K to be the valuation ring of and to be

its maximal ideal. ] [Further hint: use 2. ]

12. If KA is an algebraic extension (i. e. [K:k] = 0 so

that KA is an algebraic extension), there is a curious analogue
of 11 (though right-minded people would say that 11 was a curious
analogue of 12). Define | K:k | to be the dimension of K as a
k-vector space (perhaps infinite), and show:

K:k K:k K* : k

More examples can be found in Bourbaki, Alg. Comm. , V.

104
8. PrUfer and Dedekind rings

Throughout this chapter A denotes an integral ring and k

its field of fractions.

8.1 (Lemma). Let A be a valuation ring and M be a torsion-

free A-module of finite type. Then M is free.

Proof. Let (m.), < . < generate M for some least n e u>.
n
Suppose J £.m. = 0 for |. € A with (say) £ € A* and

1| I > U . I for i = 1, .. . , n. Then f1^. e A for all i and

1s
n' 'sl' ' ' n s
i
n-1 n-1
(m + l ( m } = : S that m = Z (
*n n VV i ° ° n " . ^n V m i
and ( m . ) 1 < . < _, generates M, a contradiction. Thus £. = 0
for all i and (m.)-. < • <- bases M. D

8. 2 (Theorem). The following four conditions are equivalent:

(i) A is a valuation ring for every p € spec(A);

(ii) every torsion-free A-module of finite type is projective;

(iii) if a is a non-zero ideal of A of finite type, then a is an

invertible sub-A-module of k;

(iv) every sub-A-algebra of k is integrally closed in k.

105
Proof. Suppose (i) holds. Let M be a torsion-free A-module of
finite type. Then M is a torsion-free A -module of finite
type for every p e spec(A), so that by 8. 1 M is a free A -
module, of rank r say. However M « n | is isomorphic to
M ®A k for all p e spec(A), so by 1. 3. 1 r n = r , n i for
P Ap P 1U ]
all p e spec(A), and M is projective by 3. 3. 7 (iii). That is, (ii)
holds.
Suppose (ii) holds and let us prove (iii). By (ii) a is
projective, and thus invertible by 3. 4. 2 and 3. 4. 3.
Suppose (iii) holds and let p e spec(A) and x e k \ A n .
Then x = — v for y,z € A with z ± 0 and Ay 4- Az is invertible;
z
so A + Ax is invertible (for its inverse is z(A:Ay + Az)). Thus
1 e (A + Ax) (A:A 4- Ax) so that 1 = a + xb for some a, b e A
with ax, bx e A. Suppose a e n ( A ): then x = a" ax e A n ,
-1 -1
a contradiction. Thus a e nt (A n ) and x = (1 - a) b e A^ .
That is, (i) holds.
Suppose (i) holds and let B be a sub-A-algebra of k and
ttl be a maximal ideal of B. Then m n A = p e spec (A) and
A < B . Thus by 7. 3.1 A = B and B is a valuation
ring in k and is therefore integrally closed in k. But by 3. 2. 1. 3
B= n B and thus B is integrally closed in k: that is, (iv)
m
m
holds.
Finally suppose (iv) holds; let p € spec(A); and x e k\A~
7 7
Then A [x ] i s i n t e g r a l l y c l o s e d i n k, s o xeA [x ]. T h u s
2 2n "
x = b 4- b x 4- . . . + b x for s o m e b , . .. , b e A ^ with
n > 0; a n d
,. - i 2 n / u - i x 2 n - l , , 2 , ,. - i , 2 n - 2 , . ,2n,
( b x )N - ( b x ) + b b ( b x ) + . . . + b o - n0
o o o i o on

106
so that b x € A M . However b x e m (A ) for if not,

x = b (b x~ )~ e A ; and so

and x" 1 e A . Thus (i) holds, a

If A satisfies the conditions of 8. 2 we call A a Priifer

ring. For example Z is a Priifer ring.

8. 2. la (Corollary). We have a further equivalent condition:

(v) A + Ax is invertible for all x e k.

Proof. Both (iii) implies (v) and (v) implies (i) are already con-
tained in the proof that (iii) implies (i). Q

8.2.1 (Corollary). If A is a Priifer ring, every sub-A-algebra

B of k is a Priifer ring.

Proof. Let $ e spec(B). Then A nA < B ^ and B ^ is a

valuation ring in k by 7. 3. 1. •

8. 2. 2 (Corollary). Let A be a Priifer ring and K be an

algebraic extension of k. Let B be the integral closure of A in
K. Then B is a Priifer ring.

Proof. By 5. 1. 3 and 7. 3. 4. o

107
Exercises. Find (i) the rings A for which 8. 1 is true, (ii)
the subrings of Q, (iii) a Pruf er ring which is not Bezout.
We say that a ring A is a Dedekind ring iff it is a Noetherian
Priifer ring. For example Z is a Dedekind ring.

8.3 (Theorem). The following four conditions are equivalent:

(i) A is a Dedekind ring;

(ii) A is Noetherian and integrally closed in k; and if

p € spec(A), then p = 0 or p is maximal;

(iii) A is Noetherian and A is a field or a discrete valuation

ring for every maximal ideal m of A;

(iv) if a is a non-zero ideal of A, then a is an invertible

sub-A-module of k.

Proof. Suppose (i) holds. Then A is Noetherian and integrally

closed by 8. 2 (iv). Suppose p € spec(A) and 0 c p c xti for
some maximal ideal HI of A. Then A m c k and
card(spec(A )) > 3 and A is a Noetherian valuation ring: this
is a contradiction by 7. 2. 3 (iii), (iv). Thus (ii) holds.
Suppose (ii) holds. Then for any non-zero maximal ideal
HI of A, we have A m c k and A is a Noetherian ring,
integrally closed in k by 5.1. 3, and with card(spec(A m )) = 2.
Thus by 7. 2. 3 (iv) A is a discrete valuation ring in k: that
is, (iii) holds.
Suppose (iii) holds. Let p € spec(A) so that p c m
for some maximal m . Then A c A so that A is a
valuation ring, and A is a Priifer ring by 8. 2 (i). Thus (i) holds.

108
 Suppose (i) holds. Then (iv) holds by 8. 2 (iii).
Finally suppose (iv) holds. Then A is a Pruf er ring by
8. 2 (iii); and since an invertible sub-A-module of k is of finite
type, A is Noetherian. Thus (i) holds. Q

Suppose A c k is a Dedekind ring. Then A is a

discrete valuation ring in k for each maximal ideal nt of A, and
we write o r d m :k* — Z for the corresponding map.
Let a be an invertible sub-A-module of k. Then a is of
finite type, so that x a Q A for some x e A*, and we can define

ordm ( a ) = inf ord (x) .

0*x ea

Thus if G is the group of invertible sub-A-modules of k, then

o r d m :G — Z is a group morphism; and so we have a group
morphism: G -* II Z given by a •—•(ordm ( a )) m . In particular
if tl is a maximal ideal we have:

o r d m ( n ) = o if m * n
= 1 if m = n

8. 4 (Theorem). The
_ map: G -* mII Z just defined is a group
isomorphism: G •- © Z .
m
Proof. Let x c A * and ( nt.). be a sequence of distinct
maximal ideals of A with x e tn . for all i. Then
A => m => m i n m D . . . 2 xA, for if m n ... n m =
m i n .. . n m r - 1 we have n t r 2 m i . . . t n r - 1 and
m = nt. for some i < r. Thus A/xA is not Artinian and not
r I

109
of finite length; but A/xA is a torsion A-module of finite type: a
contradiction by 8. 3 (ii) and 4. 3. 5 (i). Thus x £ m and
ord m (x) = 0 for all but a finite number of m .
•H

Clearly this also holds for x e k*.

Let a e G. Then a = Ax + .. . + Ax for some
Let a e G. Ther l n
x , .. . , x e k*, so that:

n
o r d m ( a ) = inf o r d m (x.) = 0

for all but a finite number of in .

Thus G - © Z .
m
n mr
Moreover m •—•(**—
ni )—
in for any family ( r in
m )m
in
onto
of finite support in Z: so that G •© Z .
m
Finally let a € G and ord ( a ) = 0 for all m . Then
a
m ~ A m f o r e a c h m ' f o r A m i s a d i s c r e t e valuation ring;
so that a m = a m + A m = ( a + A) m = A m for all m ; and
by 3. 2.1. 1 applied to the inclusions: a , A -* a + A we have
a - A + a = A. That is, G - ^ L 0 Z . D
m
Thus every non-zero ideal of a Dedekind ring is uniquely
expressible as a product of maximal ideals. The converse is true
too:

8. 4.1 (Corollary). A c k is a Dedekind ring iff every non-zero

ideal of A is uniquely expressible as a product of prime ideals.

Proof. One way is 8. 4.

110
 Conversely let p € spec (A) and p =£ 0. Define o r d n (x)
for x f A * to be the exponent of p in the expression of xA as a
product of prime ideals. Clearly

ord (xy) = ord (x) + ord p (y)

ord (x + y) > inf(ord p (x), ord ? (y))

x e A\ p iff ord (x) = 0

for all x,y e A. We define ord :k* - Z by ord (-) =

ord n (x) - ord (y) for x, y e A*. Then we see that A . is a
discrete valuation ring and that ord M :k* •* Z is the associated
map.
By 8. 2 (i) it remains to show that A is Noetherian. Since

P 9
ITp 2 II p iff m <n for all p * 0,
9
*° P^° n
there are exactly n (n + 1) ideals between A and n p ;
9*0 p*0
so A satisfies the increasing sequence condition and is Noetherian. •

8. 5 (Theorem) (Krull-Akizuki). Let A c k be a Dedekind ring;

K be a finite field extension of k; and B be the integral closure of
A in K. Then B is a Dedekind ring.

Proof. From 8. 2. 2 B is a Prxifer ring: thus it remains to show

that B is Noetherian. For this it certainly suffices to show that
B / b is an A-module of finite length for any non-zero ideal b
of B.
Let y e b and y * 0. Then y11 + \Y + • • • + an = °
for some a , . . . , a € A and least n e w , so that a e A*. Thus
l n n

111
a B Q h and it is enough to show that B/a B is an A-module of
n n
finite length.
iso
We have B® k • K by x®yi—•xy; so B is a
torsion-free A-module of finite rank | K:k [. Thus by 4. 3. 5 (ii)
B/a B is an A-module of finite length, a

Exercises. (i) A Dedekind ring with only finitely many maximal

ideals is local.

(ii) If a is an ideal of a Dedekind ring A, there exist x, y e A

such that a = Ax + Ay.

112
9. General exercises

1. Recall that in any category, a morphism f :A -* B is

called an epimorphism iff given two morphisms g, h:B -* C for some
C with gf = hf, it follows that g = h; in other words a morphism
with domain B is determined by its composition with f.
Show that for f:A — B a ring morphism the following six
conditions are equivalent:
(i) f is an epimorphism
(ii) B® A B • B by b®bTi—*-bbT
onto
(iii) B • B ^ . B by bi—•btSl
, v onto
(iv) B • B 8 ) . B by bi—• 1 ® b
(v) for all b e B b ® l = l ® b in B® A B
iso
(vi) B® A B - B by multiplication.

2. Let A •* B be a ring morphism. Show the equivalence

of:
epi^
(ii) if M, N are B-modules and f :M -* N is an A-module
morphism, then it is also a B-module morphism
iso
(iii) if M, N are B-modules then M®AN "-M&> N by
A 13

3. (i) Let A be a ring and S a multiplicative subset

of A. Show that A - ^ - S ^ A .
(ii) Let A be a ring and a e A. Show that
epi x
A — ^ A / a A x A by xi—^ (x + aA, ~).
a J-

113
[Hint: use 1 (vi). ] Use this to construct an epimorphism not of
the form (i) above. [Hint: take A = Z, a * 0, 1. ]

4. Let f:A — B be aflat epimorphism of rings (that is

epi embed
A •B and B is a flat A-module). Show that spec(B) •
spec(A); indeed, for any ideal h of B f" [ b ]B = b .

5. Give a proof or counter-example for each of the

following statements:
(i) let A be an integral ring and make A &> A into
anA-moduleby A(x®y) = (Ax) ®y for A , x , y e A ; then A8>_A
is faithfully flat;
(ii) let A be a Priifer ring and B be the integral closure
of A in an algebraic extension K of the field of fractions k of
A; then B is a faithfully flat A-module.

6. Which of the following properties of a ring A remain

true for S" A for any multiplicative subset S of A? Give counter-
examples or proofs.
(i) A is Noetherian;
(ii) spec(A) is connected;
(iii) every A-module is flat;
(iv) A is a valuation ring or a zero ring;
(v) A has zero nilradical;
(vi) A is of finite type as a Z-algebra.

7. Let A be an integral ring of finite type as a Z-algebra.

Prove, or give a counter-example for each of the following state-
ments:

114
 (ii) if A is a valuation ring, A is a field;
(iii) if A is a local ring, A is a valuation ring.
What happens if we allow A to have divisors of zero?

8. Give proofs or counter-examples for the following

statements:
(i) if B is an entire A-algebra, and every A-module is
flat, and B has no nilpotent elements, then every B-module is
flat;
(ii) if A is integral and integrally closed in its field of
fractions, then A has unique factorisation;
(iii) let B be a ring integrally dependent on a local sub-
ring; then B is local.

115
Appendix 1

A NOTE ON CATEGORIES, FUNCTORS AND NATURAL

TRANSFORMATIONS

(The definitions here need a little set-theoretic polishing,

but this would make them much longer: see the introduction to
Freyd T s Abelian Categories.)

A category J*^ consists of

(i) a class of objects A, B , C , . . .

(ii) for every two objects A, B a set [A, B] ry/of morphisms

a,p,. . . from A to B (and we write a:A -* B to mean
ae[A,Bl

(iii) for every three objects A, B, C a multiplication taking

and must also satisfy the following conditions:

(i) (associativity) if a:A - B and j3:B -~ C and y:C •* D,

then y(jSa) = (yP)a

(ii) (existence of identities) for each object A there is a

morphism e.:A — A such that if a:A -* B (resp. iS:C •* A) we
have o e . = a (resp. e./3 = j8)

(iii) if [A, B] ^ meets [ C , D ] ^ ^ , then A = C and B = D.

116
EXAMPLES

1. The category Ring of (commutative) rings (with ones):

(i) the objects are rings;
(ii) if A and B are rings, [A, B] R . consists of the ordered
triples (A, B,f) where f is a function from A to B satisfying

f(l) = 1

and

f (x + y) = f (x) + f (y)
f (xy) = f (x)f(y)

for all x, y e A;
(iii) if (A,B,f):A-B and (B,C,g):B^C, then (B,C,g)(A,B,f) =
(A,C,g o f):A-C.

2. The category Top of topological spaces:

(i) the objects are topological spaces;
(ii) if S and T are topological spaces, then [S, T ] T o D
consists of the ordered triples (S, T,f) where f is a continuous
function from S to T;
(iii) the multiplication is just as for rings.

3. The dual category r>/* of any categoryS^ :

(i) the objects of j ^ * are those Sf
(ii) [A,
(iii) [ B , C ] ^ , x [A,B] - [A,C] is (fs, a)

(Thus (£/*)* = Sf.)

117
 Now let JS^ and .<^> be categories and F be a function from
the objects of S&to those of .^5? and from the morphisms of S&to
those of & . We say that F is a functor iromS^fto &0Mi:

(i)
F e
(ii) ( A^ = eF(A} for ever
y object A of
(iii) if a:A-B and p:B - C in J ^ then P(j8a) = F(j3)P(a).

Strictly speaking this is a covariant functor; a contravariant

functor from J^to.5Si?is just a covariant functor fromJ*^to
Thus F is a contravariant functor from.i^/to 0? Mi:

(i)* F [ [ A , B ^ ] C [F(B),F(A)]^ ;

(ii)* same as (ii) above;

(iii)* if a:A - B and £:B - C in J ^ f then F(j8a) = F(a)F(jS).

EXAMPLES

1. The identity function on a category ,

2. The contravariant functor spec from Ring to Top;

if (A, B,f):A-*B where A and B are rings, we define

spec(A, B, f) = (spec(B), spec(A),g):spec(B) -* spec(A)

where g is the (continuous)function from spec(B) to spec(A)

given by

118
 = {x € A:f (x) e p }

for p € spec(B); it is easy to check that

spec(/ta) = spec(a)spec(/3)

for a:A — B and 0:B — C.

Now let F and G be functors from S&to Q$. A
natural transformation x from F to G is a function taking
objects of J*^to morphisms of £•$such that

(i) x(A):F(A) - G(A) for all objects A of,

(ii) if Q?:A-*B in St/, then the diagram in £$

X(A)
F(A)

G(a)

commutes (i.e. G(o?)x(A) = x(B)F(o?)).

EXAMPLES

1. H F = G we can define x(A) = e_, A V

2. (The double dual.) Let Mod be the category of

modules over a fixed ring A. We have a contravariant functor F

119
from Mod to Mod: we define F(M) for each module M to be
[M, A] M , made into an A-module in the obvious way, and if
(M,N,f):M-N in Mod, we have F(M, N,f) = (F(N), F(M), g):
F(N) -* F(M) where g is a function from F(N) to F(M) given
by g(&) = a(M, N, f) for all a:N -* A. Then we have a natural
transformation x from 1,, , to F°F (which is a (covariant)
Mod
functor): namely

X(M) = (M,F(F(M)),h):M - F(F(M)) ,

where h(m) = (F(M), A, j):F(M) - A, where j(M,A,f) = f(m) for

all m e M and (M,A,f):M •* A.

120
Appendix 2

THE CONSTRUCTIBLE TOPOLOGY

(We describe topologies by their closed sets.)

Let A be a ring and a be the least topology on spec(A)

such that D(f) is both open and closed for all f e A. Then a is
Hausdorff and contains the Zariski topology.
If B is an A-algebra write £(B) for the image of spec(B)
in spec (A) and let

T = K(B):B an A-algebra} .

Recall that if (B.). T is a direct family of A-algebras with

maps f..:B. - B. for i < j we have the direct limit B = lim B.

defined as disjoint u B./~ where (x, i) ~ (y, j) iff

f (x) = f .,(y) for some I ^ i, j . We make B into an A-algebra in

an obvious way
way and
and let
let B.
B. ^-~ B by xi—•class of (x, i). If C is
another A-algebra we have:

na
A lim B. »lim C® A B. .
A - 1 iso - A 1
iel iel

121
Theorem, (i) = ?(B) n

(ii) ?(B x C) = ?(B) U C(C)

(iii) B)= n __
U I
id

(iv) ( = n
HA X eA

(v) r = a and is a compact Hausdorff topology on spec(A).

Proof. Looking at

•B

where k = A / p A for some p e spec(A) we see that

p € $(B) iff k® B * 0 . But (k^ A B)^ A (k8) A C) is isomorphic
eoi A A A
to k®A(B®AC) since A—^— k; and this gives (i). Similarly
we get (ii), (iii) and (iv) from the natural isomorphisms:

(B x c) ®Ak -• (B ®Ak) x (c ®Ak)

(lim - lim (B.® A k)

—*• 1 A .

lim ( BJ
XeA A
A cA
o
A finite
o

122
Thus r is a topology. If n £(B ) = ft, then k® A ( ® B ) = 0
X eA X eA
o
for some finite A c A, so n ?(B.) = 0 and r is compact.
0 X
X cA
o
We have D(f) = ?(Af) e r and X\D(f) = ?(A/fA) e r so that
r 2 a and by analysis r = a. a
We call r the constructible topology on spec(A).

123
Bibliography

For reference:
Bourbaki Algebre Commutative (Hermann)
Grothendieck Elements de Geometrie Algebrique
(IHES)
Zariski and Samuel Commutative Algebra (Van Nostrand
1958 and 1960)

For reading and exercises:

Atiyah and Macdonald Introduction to Commutative Algebra
(Addison- Wesley 1969)
Kaplansky Commutative Rings (QMC Notes)

For specialised topics:

Serre Corps Locaux (Hermann 1968)
Artin Theory of Algebraic Numbers
(Gottingen 1959)
Samuel (ed.) Les epimorphismes d'anneaux
(Paris 1968)
Cohen and Seidenberg Bull. Amer, Math. Soc. 52 (1946),
252-61
Nagata Local Rings (Interscience 1962)

124
Index of notation

On 42 M* 55
ord 92 supp(M) 64
90 ass(M) 66
94 ann(m) 66

<* 94 I ®m 8

A* 2 7
U (A) 2
M 87
n (A) 4 A
X € A
m(A) 5
87
K(A) 5
X e A
spec (A) 5
tot(A) 33 8

MN 57
v(a ) 5
(M:N) 57
at) 5
34
5
X €A * 37

r 51 Mf 40
p
longA(M) 63
D(f) 40 63
TA(M)
r A (M) 63
32
36
e(B|A) 91
40
f(B|A) 91
A[X] 72
[K:k] 103
A[x] 73
|K:k| 104

125
Index of terms

algebra 13 finite length 61

, entire 74 presentation 26
algebraic closure 87 rank 63
annihilator 66 support 7
Artinian 61 type algebra 85
module 9
base 9
flat 19
bilinear 7
for 20

category 116 , absolutely 40

Cohen 14 , faithfully 26

C oh en- Seidenberg 81 free 9

countable type 42 , algebraically 85
functor 118
direct sum 7
domination 94 generate 9

, strong 94 group algebra 102

dual 117 , partially or-

, double 119 dered abelian 89

, Picard 56
epimorphism 113 , value 90
exact sequence 13
functor 13 ification 9

extends 91 integral (element) 74

ly closed 75
field 2 closure 75
of fractions 33

126
invertible 58 object 116
ordinal 42
Jordan-HSlder 61
, limit 43

Kaplansky 43
, successor 42

Kronecker 14
prime ideal 3
Krull-Akizuki 111 associated 66

length 63
proper ideal 2

localisation 36
ramification 91
lying over 80
rank:
maximal ideal 3
free modules 11

minimal polynomial 78 projective

modules 55
module 6
, injective 28 finite rank
modules 63
, projective 17
morphism 116 abelian group 104
multiplicative 2 ratio 57
replete 89
Nakayama 49 residual degree 91
natural transfor- residue field 5
mation 119 ring 1
nilpotent 4 , Bezout 23
nilradical 4 , Dedekind 108
Noether 85 , integral 2
Noetherian 11 —— , local 5
non-degenerate 58 morphism 1
non-divisor of zero 33 , Prufer 107
Nullstellensatz: , valuation 90
HilbertTs 88 , , discrete 92
weak 86 , , trivial 92

127
Samuel 100
simple 61
spectrum 5
subring 1
substitution (axiom) 43
support 64

tensor product 8
, restricted 87
torsion module 64
submodule 63
torsion-free 23
total degree 92
total ring of fractions 33
transcendence base 103
transf inite induction 43
recursion 43

unitary polynomial 72
units 2

Zariski topology 6

128