Professional Documents
Culture Documents
J.T.KNIGHT
Commutative Algebra
WG3V;
Published in the United States of America by Cambridge University Press, New York
James T. Knight
Barry R. Tennison
October 1970
111
J. T. Khigiit
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.
IV
Riemann surfaces of field extensions, Proc. Camb. Phil. Soc.
65 (1969) 635-650.
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
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
Bibliography 124
Indices 125
vm
1. Preliminaries
1.1 INTRODUCTION
0 0
0 0
(i) a c A
(ii) i i a
(iii) a n tt (A) = 0
(i) 1€S
(ii) x, y e S implies xy e S;
(i) A\ 9 is multiplicative
(ii) A/ p is integral.
Procf. Let
X = {b an A-ideal: b2 a and b n S = 0} .
n (A) = n p
9
taken over all prime ideals p of A.
m
a b = i 2 a.b.: a. e a and b. e b }.
i l l 1 l 1
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.
1.3 MODULES
(A + A')m = Am + A'm
(AAf)m = A(ATm)
lm = m
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 .
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
f + W i— I f(*,m) g(Z,m) .
I eL
m €M
8
L 8>A M - N such that L x M - L ® A M
A ,A
L 8> ( © M ) •* © (L 8) M ) given 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 ))
x = . 2 MA x ) (1 ® mAj .D
A eA
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
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.
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
A sequence
f
n-l
M
12
of A-module morphisms is called exact iff
0 - k e r f - M - f [M] - 0
0 - N/f[M] - L - L/g[N] - 0
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.)
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
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.
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
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.
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.
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
h( 1 L x . ) = IL m . .
A A A A
AeA XeA
hoi / ,
Then y^ |g commutes, D
17
g:M — P are such that g o f = 1 then P is projective.
2. 2 FLAT MODULES
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.
may define:
p
E ®A M/ — • E ®A N
A A
/ Im(l 8) f)
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 ^ ^^ _________________»
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
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
E jg) jj ^ E jg) M .
A A
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.
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-
E ®A N
E®, inj
N ®A(M/N)- —0
21
(ii) First let A = {1,2} and M^ ^ We have the
exact sequences
A l A A M
A
A 2
0 0 N-
E ®A (M n N) •E ® A N •
A 1
inj
inj
—E
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
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
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
24
E &>A N be exact. We have
onto
E ® A (L/Ker(gof))
E®Af[L] Ker g
In E ® A M we have:
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
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.
P - 0
-N/p[LQ]
X - l ( x , y ) : x e L i ; y e M; p(«(x)) = £(y) }
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
is exact.
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
29
T'(LQ)-
onto
so that M= N.
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)
31
3. Fractions
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.
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.
X = { p e spec(A): pnS=jZi)
33
•* T" X B such that
m + m/ _ sTm + smT
s sf ~ ss f
m = >m
s s £
a m am
s t st
34
Proof. Let us define a map
g: S M - S A®AM by - ,_±(g) m .
A S S
s sT
= 0.
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.
m mT mmT
s sT ssT
3.2 LOCALISATION
Ap = (A\ 9 )'lA
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 ).
(A )= { £ :p e p and s e A\ p }
= 9 Ap
= (A\ p )" X p
K (© A ^ ) c © n A ^ c ©A^ since
= m ( A n ) c A n .
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.
inj
bij bij
E N
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
39
Exercise (hard). Let A be a ring. Show that the following
statements are equivalent:
(We call such a ring absolutely flat. See Bourbaki, Alg. Comm.
I, §2, Ex 17.)
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.
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 .
41
3. 3 PROJECTIVE MODULES AND LOCALISATION
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).
43
Let us define by transfinite recursion the function
0i—^1 taking On t o ^ ( I ) as follows:
(ii) \,— f)
0
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 ;
p E
t suc(/S)] ® ( 1 " p)[Esuc(/3)^ ; and
<P uttin S 1 " P in place of p)
Therefore
p[E ] = U p[E ] = u e Q = © Q ;
p y
yep yep hey ° 6 e^S °
y 6
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
(iii) P = Q0 R.
m! = m. + a.m. if i e J and i * j
= m. if i = j or i e I \ J
i € J: so if we define
46
m!T = y. if i e J
= m. if i e I \ J
= 2 Ay
i €J l
R = P n ( I Am.) .
l
i€l\J
Thus
(i) x €Q
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
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.
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 .
n
so that Z |.x. = 0 and f is injective. D
1 X
49
(i) M is free
(ii) M is projective
(iii) M is flat
(iv) m
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
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
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 •—^
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
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
iso
ISO
(A
53
JL / JL f x
fn fm • " gfn
54
If P is a projective A-module of finite type and p
is constant, we call its constant value the rank of P.
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
56
Exercise. Construct such a Pic (A).
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)
(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
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
(i) M(AT:M) = AT
(ii) jeM(A':M)
58
inverse (AT:M) of M.
(i) M is invertible
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
= 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
o iC N c . . . c Nn = M
61
4.1.1 (Proposition). (i) An A-module is of finite length iff it
is both Noetherian and Artinian.
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
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).
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.
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.
65
4. 3 PRIME IDEALS ASSOCIATED TO A MODULE
ann(m) = | H A : Am = 0} .
66
Therefore ann(m) e spec(A) and ann(m) € ass(M).
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).
(^-) € ass(M p )
(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:
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.
by 4. 2. 1 (i). D
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
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
70
ISO 2
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
71
5. Integers
(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:
72
Proof. Trivial, a
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
73
n
.- 2 a x )
= 0
Proof. 1. 3. 2. 3. D
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.
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
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
Proof. 3. 2.1. 3. D
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.
n
f(X) = II (X - b.)
l
m
g(X) = II (X - d.)
J
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.
= Gm(X)
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
80
there is no $ e spec(B) lying over p i f f A n p B = > p .
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
81
that ip lies over p in A. a
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
83
However y = ( IT a(x)) x € © ; and y e k n B = A; so
y e (fi n A = || £ $ , a contradiction, a
84
6. Some geometrical results
85
N
i
belongs to k*. Let z. = y. - y for i = 1, . . . , m-1. Then:
N N
l-l
= o
86
APPLICATION TO ALGEBRAIC CLOSURES OF FIELDS
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.)
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
88
7. Valuation rings
for all x, y, z e G.
We say that G is replete iff for all n e o> with n > 0 we
have:
89
x u (A) < y u (A) iff xy" 1 £ A .
7. 2 VALUATION RINGS
Proof. Immediate, o
90
if x, y € A, then (say) |x| < |y | and Ax + Ay = Ay.
A n m (B) = m (A)
k* n u (B) = u (A)
A/ m (A) - B / m (B)
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 | .
inj e(B|A)
k*/ tt (A) -K*/ tt (B) - ^ = ^ - k * / tt (A) .
inj
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:
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
94
(i) (A, p ) < (A', p')
Proof. Immediate, o
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:
Thus the valuation rings in k are the local rings which are
maximal for the orders both of domination and of strong domination.
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.
a x n + . .. + a = 0 .
n o
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
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. 4 AN APPLICATION
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
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
g(X) = n (X - y ) = X r + a X1*"1 + . .. +
1 1
Bibliography.
Schilling, Valuations (dull).
Ax and Kochen, Amer. J. Math. 87(1965), 605-730 (beautiful
but undigested).
101
EXAMPLES ON GENERAL VALUATIONS
(Af)(a) = Af(a)
(fg)(a) = I f(T)g(T-lo) ,
T eG
102
4. If G is a totally ordered abelian group, F[G] is an
integral domain.
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
K:k K:k K* : k
104
8. PrUfer and Dedekind rings
Proof. Let (m.), < . < generate M for some least n e u>.
n
Suppose J £.m. = 0 for |. € A with (say) £ € A* and
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
Proof. Both (iii) implies (v) and (v) implies (i) are already con-
tained in the proof that (iii) implies (i). Q
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.
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
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
n
o r d m ( a ) = inf o r d m (x.) = 0
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
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. •
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
112
9. General exercises
113
[Hint: use 1 (vi). ] Use this to construct an epimorphism not of
the form (i) above. [Hint: take A = Z, a * 0, 1. ]
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?
115
Appendix 1
116
EXAMPLES
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.
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).
(i)* F [ [ A , B ^ ] C [F(B),F(A)]^ ;
EXAMPLES
118
= {x € A:f (x) e p }
spec(/ta) = spec(a)spec(/3)
X(A)
F(A)
G(a)
EXAMPLES
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
120
Appendix 2
T = K(B):B an A-algebra} .
na
A lim B. »lim C® A B. .
A - 1 iso - A 1
iel iel
121
Theorem, (i) = ?(B) n
(iii) B)= n __
U I
id
(iv) ( = n
HA X eA
Proof. Looking at
•B
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)
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
Cohen 14 , faithfully 26
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
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