Beruflich Dokumente
Kultur Dokumente
M. F. ATIYAH, FRS
UNIVERSITY OF OXFORD
l. G. ~1ACDONALD
Chnptcr 2 Modules . • • • • . . .. . • • • • • • 17
Modules and module homomorphisms . • • • • • • • 17
Submodulcs and quotient modules . • • • • • • • 18
Operations on submodules . . . . • • • • • • 19
Dircct sum and product . . . • • • • • • • 20
Finitely gencrated modules . • • • • • • • • • 21
Exact seq uences . . . . . . . • • • • • • • • 22
Tensor product of modules . . . • • • • • • • • 24
Restriction and extension of scalars . • • • • • • • • 27
Exact ncss properdes of the tensor product . • • • • • • 28
Algebras . . . . . . . . . • • • • • • • 29
Tensor product of algebras . . . . • • • • • • • 30
Exercises . . . . . . . . . • • • • • • • 31
Copyright ~ 1969 by Add '
ISOn-Wesley p bl ' h' Chapter 3 Rings nnd Modules of Fractions 36
u IS mg Company, lnc. • • • • • • • • • •
Britain l•'bra rw•se, WJthout pr'•or w . any means, electronic, mechanical photocopying, so
· ry of e0 · ruten p · · . •. . Cbapter 4 Primary Decompositioo • • • • • • • • • • • •
• n&ress Catalog Card N erm•ss•on of thc pubhsher. Pnnted in Great
rs11~ 0-2n 1 "" 0 · 72-79530
55
- "' 3~ 1 -9 . Exercises. • • • • • • • • • • • •
'fNOilnp_,~A-R9A • • • •
V
vi coNTENTS
D ndence nnd Vnluntions • •
•
Chapter S Integral epe •
59
1nregral dependence . . . . . . . . •
59
he oing-up theorem . . . . . . . . . . . . . •
T g Josed integral domains. The g01ng-down theorem 61
11
Integra Y e · • 62
Valuation rings · · · · · · · • 65
• • • • • • •
Exerciscs . • • • • 67
ix
1
•
~e shall ~egi~ b.y r~viewing rapidly the definition and eleme~ry properties of
nngs. Thts wtll mdtcate how much we are going to assume of the reader and it
will also serve to fix notation and conventions. After this review we pass on
to a discussion of prime and maximal ideals. The remainder of the chapter is
devoted to explaining the various elementary operations which can be performed
on ideals. The Grothendieck language of schemes is dealt with in the exercises
at the end.
A ring A is a set with two binary operations· (addition and multiplication) such
that
1) A is an abelian g roup with respect to addition (so that A has a zero element,
denoted by O, and every x E A has an (additive) inverse, - x).
2) Multiplication is associative ((xy)z = x(yz)) and distributive over addition
(x(y + z) = xy + xz, (y + z)x = yx + zx).
We shall consider only rings which are commutative:
3) xy = yx for all x, y E A,
"'' . . . ,~~
and have an identity element (denoted by 1):
and so A has only one element, O. In this case A is the zero ring, denoted by O
(by abuse of notation).
1
•
2 RINGS ANO IDEALS
PRIME IDEALS ANO MAXIMAL IDEALS 3
A ring homomorplzism is a mapping/ of a ring A into a ring B such that
Thc multiplcs ax of an element x E A forma pr' · ¡ 'd 1 d
i) f(x + y) = f(x) + f(y) (so that f is a homomorphism of abelian groups, or Ax. x is a unit <=> (x) = A = (1). Thc zero ide:;c(Ó)~ I ea¡'¡ ednoted by (x)
and therefore alsof(x -y) = f(x) - f(y),f( -x) = -f(x),f(O) = 0), A fi Id . . . lS usua y enoted by O
re .Is a ~mg A m whi~h 1 =/: O and every non-zero element is a unit~
ii) f(xy) = f(x)f(y), Every field ts an mtegral domam (but not converscly: z is nota field).
iii) f(l) = l. Proposition • 1
1.2. Let A be a ring =/: O· Then the fiollow 1•ng are equwa
. ent:
In other words, f respects addition, multiplication and the identity element. 1) A is a field;
ii) the only idea/s in A are O and (1); ·
A subset S of a ring A is a subring of A if S is closed under addition and
multiplication and contains the identity element of A. The identity mappíng of iii) every homomorphism of A into a non-zero ring Bis injective.
S into A is then a ring homomorphism. Proof i) => ii). Let a =/: O be an ideal in A . Then a contains a non-zero
If/: A --* B, g: B --* C are ring homomorphisms then so is their composition element x; x is a unit, hence a 2 (x) = (1), hence a = (l).
g o f : A--* C. . ii) => iii!. Let cp: A --* B be a ring homomorphism. Then Ker ~e/>) is an
1deal =/: (1) m A , hence Ker (cp) = O, hence cp is injective.
IDEALS. QUOTIENT RINGS iii) => i). Let ~ be an element ?fA which is not a unit. Then (x) =~: (1),
hence B = A j (x) IS not the zero nng. Let cp: A --* B be the natural homo-
An ideal a of a ríng A is a subset of A which is an additive subgroup and is such morphism of A onto B, with kernel (x). By hypothesis, 4> is injective, hence
that Aa s; a (i.e., x E A and y E a imply xy E a). The quotient group A j a (x) = O, hence x = O. •
inheríts a uniquely defined multiplication from A which makes it into a ring,
called the quotient ring (or residue-class ring) Aja. The elements of Aja are the PRIME IDEALS AND MAXIMAL IDEALS
cosets of a in A, and the mapping cf>: A --* A j a which maps each x E A to its
coset x + a is a surjective ring homomorphism. An ideal lJ in A is prime if ~ =/: (1) and if xy E ~ => x E\:> or y E p.
We shall frequently use the following fact: An ideal m in A is maximal if m =/: (1) and if there is no ideal a such that
m e a e (1) (strict inclusions). Equivalently:
Proposition 1.1. There is a one-to-one order-preserving correspondence
p is prime <=> Afp is an integral domain;
between the idea/s ó of A which contain a, and the ideals b of A j a, given by
m is maximal <=> Aj m is a field (by (1.1) and (1.2)).
() = cp -l(b). •
Hence a maximal ideal is prime (but not conversely, in general). The zero ideal
If f: A --* Bis any ring homomorphism, the kernel off(=¡- (0)) is an ideal
1
is prime <=> A is an integral domain.
a of A, and the image off(=f(A)) is a subring C of B ; andfinduces a ring
lf f: A --* B is a ring homomorphism and q is a prime ideal in B, then
isomorphism Aj a :::' C.
¡ - 1 ( q) is a prime ideal in A, for A//- 1( q) is isomorphic to a subring of Bfq and
We shall sometimes use the notation x =y (mod a); this means that hence has no zero-divisor =/: O. But if n is a maximal ideal of B it is not neces-
x-y ea.
sarily tr.ye thatf- 1 (n) is maximal in A; all we can say for sure is that it is prime.
(Example: A = Z, B = Q, n = 0.)
ZERO-DIVISORS. NILPOTENT ELEMENTS. UNITS Prime ideals are fundamental to the whole of commutative algebra. The
following theorem and its corollaries ensure that there is always a sufficient
A ~ero-divisor in a ring A is an element x which "divides O", i.e., for which there
supply of them.
CXJ~ts Y :F O i~ A such that xy = O. A ring with no zero-divisors # O (and in
whtch 1 :F O~ Is calle~ an integral domain. For example, z and k[x1 , ••• , Xn] Theorem 1.3. Every ring A =1: Ohas at /east one maximal ideal. (Remember
(k a field, x, mdetermmates) are integral domains. that "ring" means commutative ring witb l.)
An ~lement x E A is nilpotent if xn = O for sorne n > O. A nilpotent Proof This is a standard application of Zorn's lemma. • Let :E be the set of all
element 1.s ~ zero:d'JVIsor
· (unless A = 0), but not conversely (in general). ideals =1: (1) in A. Order :E by inclusion. :E is not empty, since O E :E. To apply
A Uhlt m A 1s an element x which "divides 1", i.e., an element x such that
xy. = 1 for sorne Y E A. The element y is then uniquely determined by x 1 and is • Let S be a non-empty partially ordered set (i.e., we are given a relation x ~ y on S
wntten x - 1 · Th e umts· ·m A form a (multiplicative) abelian group. which is reflexive and transitive and such that x ~ y and y ~ x together imply
4 RINGS ANO IOEALS
NILRADICAL ANO JACOBSON RADiCAL 5
Zorn 's lemma we must show that every chain in ~ has an upper bound in ¿:; it is the kernel of the homomorphism A --7- k which maps / EA t 1''0)) B ·
Iet then (aa) be a cbain ofideals in L, so that for each pair of índices a, f3 we have . . . 1 'd . o \ . ut tf
11 > 1, m ts nota pnnctpa t eal : m fact it requires at Ieast n generators.
either aa s; a6 or a8 s; aa. Let a = Ua aa. Then a is an ideal (verify this) and
1 f# a because 1 f# aa for all a. Hence a E ~' and a is an upper bound of the chain . . 3). A principal ideal
. domain is an integral domain 1'n which every 1'deal ·ts
pnnct ~al. .rn such a nng every non-zero prime ideal is maximal. For if (x) i= o
Hence by Zorn's Jemma ~ has a maximal element. • is a pnme tdeal and (y) ::;, (x), we have x E (y), say x = yz, so that yz E (x) and
Coro//ary 1.4. Jf a # (1) is an ideal of A, rhere exists a maximal ideal of A y f# (x), hence z E (x): say z = tx. Then x = yz = ytx, so tbat yt = 1 and
• • therefore (y) = (1).
contammK a.
Proof Apply (1.3) to Aj a, bearing in mind (1.1). Alternatively, modify the
'"NILRADICAL ANO JACOBSON RADICAL
proof of (1.3). •
Corollary 1.5. Every non-unit of A is contained in a max imal ideal. • Proposition 1.7. The set m of al! nilpotent elements in a ring A is an ideal
ancl Ajm has 110 11ilpotent element :1= O. '
Remarks. 1) If A is Noetherian (Chapter 7) we can avoid the use of Zorn's
Jemma: the set of all ideals # (1 ) has a maximal ele men t. Proof If X E m, clearly ax E m for all a E A. Let X' y E m: say xm = O, yn = o.
By the binomial theorem (which is valid in any commutative ring), (x + y)m+n-1
2) There exist rings with exactly one maximal ideal, for example fields.
is a sum of integer multiples of products x'y', where r + s = m + 11 - 1; we
A ring A with exactly one maximal ideal m is called a local ring. The field
cannot ha ve both r < m and s < 11, hence each of these products vanishes and
k = A/m is called the residue field of A. therefore (x + y)m+n - l = O. Hence X + y E m aod therefore misan ideal.
•
Proposition 1.6. i) Let A be a ring and m :/= (l) an ideal of A suclz ~that Let x E Afm be represented by x E A. Then X' is represented by xn, so that
every x E A - m is a unit in A . Then A is a local ring and m its maximal ideal. _xn = o => xn E w => (xn)k = o for sorne k > o => X E m => -~ = O. •
ii) Let A be a ring and m a maximal ideal of A, suclz that every element of The ideal W is called the ni/radical of A. The following proposition gives an
1 + m (i.e., every 1 + x, where x E m) is a unit in A . Tlzen A is a local ring. alternative definition of W:
Proof i) Every ideal # (1) consists of non-units, hence is contained in m. Proposition 1.8. The ni/radical of A is the intersection of al/ the prime ideals
Hence m is the only maximal ideal of A. of A.
ii) Let x E A - m. Sin ce m is maximal, the ideal genera ted by x and m is Proof Let 91' denoté the intersection of all the prime ideals of A. If fE A is
(1), hence there existy E A and t E m such that xy + t = 1; hence xy = 1 - t nilpotent and if .p is a prime ideal, then /" = O E .p for sorne n > O, hence
belongs to 1 + m and therefore is a unit. Now use i). • f e p (because .pis prime). Hence/ E W'.
Conversely, suppose that f is not nilpotent. Let l: be the set of ideals a
A ring with only a finite number of maximal ideals is called semi-local.
with the property
Exa~ples. 1) A = k[x¡, . .. , Xn], k a field. Let f E A be an irreducible poly- · n > O => ¡n rt a.
nom1al. By unique factorizatjon, the ideal (/) is prime.
Then L is not empty because O El:. As in (1.3) Zorn's lemma can be applied
~) A. = Z. Every ideal in Z is of the form (m) for sorne m ~ O. The ideal
to the set l:, ordered by inclusion, and therefore l: has a maximal element. Let
(m) 1s pnme <> m. = Oora prime number. All the ideals (p), where pis a prime
.p be a maximal element of l:. We shall show that .P is a prime ideal. Let
number, are max1mal: Z/(p) is the field of p elements.
x, y rt .p. Then the ideals .p + (x), .p + (y) strictly contain .P and therefore do
The sa~e h~lds in Example 1) for n = 1, but not for n > J. The ideal m of
all polynom1als m A = k[x¡, . .. , Xn] with zero constant term is maximal (since not belong to L; hence
¡ mE .p + (x), /" E .P + (y)
x =.y). A subset T of S is a chain if either x ~ y or y ~ x for every pair of elements . for sorne m, n. It follows thatjm +n E p + (xy), hence the ideal t.> + (xy) is not
~,Y 10 T. Th~n Zo~n's ~emma may be stated as follows: if every chain T of S has an in l: and therefore xy rt .p. Hence we have a prime ideal~ such thatf rt ~.so that
pper bound m ~ (t.e., tf there exists x E S such that r ~ x for all r e T) then S has
at 1east one maxtmal element. frt W'. •
well~~~:ri~roof .or ~he equivalence of Zorn's lemma with the axiom of choice, the . The Jacobson radical mof A is defined to be the intersection of all the maxi-
1 prmciple, etc., see for example P R Halmos Nafve Set Theory mal ideals of A . It can be characterized as follows:
Van Nostrand (1 960). · · ' '
l
6 RJ NGS ANO IDEALS O PERATIONS ON lDEALS 7
Proposition 1.9. X E m <=> 1 - xy is a unit in A foral/ y E A . Two. ide~ls a, b are said to be coprime (or comaximal) if a + ó = (1). Thus
Proof => : Suppose 1 - xy is nota unit. By (1. 5) it belongs to sorne maximal for copn me
. tdeals we . have a f1 b = aó. Clearly two 1·deals a, 1'.u are copnme
• •
1f
ideal m ; but x E m s m, hence xy E m and therefore 1 E m, which is absurd. and only tf therc cx1st x E a and y E b such that x + y = 1.
<= : Suppose x rF m for sorne maximal ideal m. Then m and x genera te the Let A" . . . , A n be rings. Their direct product
unit ideal {1 ), so that we bave u + xy = 1 for sorne u E m and sorne y E A . n
1e 1
ideals a1•
The intersection o[ aoy family (a1) 1e 1 of ideals is an ideal. Thus the ideals of A by the rule cp(x) = (x + a 11 . . • , x + an).
form a complete lattice with respect to inclusion. Proposition 1.10. i) If a" ajare coprime whenever i ::p j, then na, = n ~·
Theproduct of two ideals a, ó in A is the ideal aó generated by all products xy,
ii) cp is surjective <=> o 11 aj are coprime wlzenever i =1= j.
where x E a and y E ó. lt is the set of all finite sums 2: x 1y 1 where each x 1 E a and
each y, E f>. Similarly we define the product of any finite farnily of ideals. In iii) cp is injective <=> n
a, = (0).
particular the powers an (n > O) of an ideal a are defined; conventionally, Proof i) by induction on n. The case n = 2 is dealt with above. Suppose
a0 = (1). Thus an (n > O) is the ideal generated by all products x 1x 2 • · · Xn n > 2and theresulttruefora 11 .•• , an_ 11 andletf> = f1f;} a,= nf;l a,.Since
in whjch each factor x 1 belongs to a. a1 + an = (1) (1 ~ i ~ n 1) we have equations X¡ + y, = 1 (X¡ E a¡, Yt E nn)
and therefore
Examples. 1) If A = Z, a = (m), ó = (n) then a + ó is the ideal generated by
rr x, n (1 -
n-1 n-1
the h.c.f. ofm and n ; a n bis the ideal generated by their l.c.m.; and aó = (mn) . y,) = l (mod an).
1= 1 1= 1
Thus (in this case) af> = a n ó <=> m, n are coprime.
2) A = k[x¡, . .. , Xn], a = (x 11 .•• , Xn) = ideal generated by x 11 •.. , Xn· Hence an + b = (1) and so
Then am is the set of all polynomials with no terms of degree < m.
The three operations so far defined (sum, intersection, product) are all
rr
1
n
..
1
O¡ = Úan = b fl an =
n
n a¡.
fol
commutative and associative. AJso there is the distributive law ii) => : Let us show for example that a 17 a2 are coprime. There exists x E A
a(b + e) = aó + ae.
such that cf>(x) = ( I, O, ... , O); hence x 1 (mod a 1) and x= O(mod a:l), so =
that
In the ring Z, n and + are distributive over each other. This is not the case 1 = (1 - x) + x E a1 + a2.
in general, and the best we have in this direction is the modular law
<= : It is enough to show, for example, that there is an elcment x E A such that
a n (ó + e) = a n ó + a n e if a 2 'ó ora 2 c. c/>(x) = (1 ,0, . .. , O). Since a 1 + o1 = (1) (i > 1) we bave equations u, + v, = 1
(u1 E a 1 , v1 E a 1). Take x = Tif,.. 2 v11 then x = TI(l - u,) = 1 (mod a¡), and
Again, in Z, we have (a + ó)(a n b) = a'ó; but in general we have only x = O (mod o1), i > 1. Hence cf>(x) = (1, O, ... , O) as required.
(a + b)(a n b) 5; al> (since (a + ó)(a n b) = a(a no) + ó(a n b) s; aó).
Clearly aó 5; a n b, hence iii) Clear, since n a is the kernel of e/>.
1 •
The union a u ú of ideals is not in general an ideal.
a n b = aó provided a + ó = ( 1) .
8 RINGS ANO IDEA L$
EXTENSION ANO CONTRACTION 9
Proposition 1.1 J. i) Let p¡, ... , Pn be prime . ideals ami let a be an ideal Exercise 1.13. i) r(a) 2 a
contained in U~-1 p,. Then o ~ v ,.for some l. ii) r(r(a)) = r(a)
··) Let a ¡ , · • . , 0 n be ideals and /et p be a prime ideal containing
11 •
n~ 1 a,. .. iii) r(aú) = r(a n ú) = r(a) fl r (ó)
Then V 2 O¡ for some i. 1/ p = n
a, then p = a, for som e l . iv) r(a) = (1) <:> a = (1)
Proof i) is proved by induction on n in the form v) r(a + ó) = r(r(a) + r(ó))
11 vi) if +> is prime, r(+> n) = p foral/ n > O.
a $ p1 (1 ~ i ~ n) => o $ ,U
.. 1
p ,.
~ropositi~n 1.1 4• .The radical of an ideal a is the intersection of ·the rime
It ¡5 certainly true for n = 1. If n > 1 and thc result is true for n - 1, then for 1deals wluclz contarn a. p
each ¡ there exists x, E o such that x1 <t .l-11 whenevcr j i= i. If for so me i we ha ve · Proof Apply ( 1. 8) to A fa. •
x, r¡: p" we are through. If not, then x 1 E p, fo r all i. Consider the element
n More generally, we may define the radical r(E) of any subset E of A in the
y= 2 X¡X2 · · · X¡ - ¡X¡+l.XI +2' · ·Xn; same way. lt is not an ideal in general. We have r(Ua Ea) = U r(Ea), for any
1"" 1
family of subsets Ea of A .
we ha ve y E a and y rj: l:J¡ (l ~ i ~ n). Hence o $ 1 Vt· ur. .
ii) Suppose lJ $ a1 for all i. Then there exist x, E O¡, x, rj: v (1 ~ i ~ n), and
Pt·oposition 1.15. D = set of zero-divisors of A = U x ti o r{Ann (x)) .
therefore Ilx, E ITa, ~ n a,; but Ilx, rj: lJ (since p is prime). Hence p ? a,. n Proo.f D = r( D ) = r(Uxt:o Ann (x)) = Ux~o r(Ann (x)). •
Finally, if p = n a" then -l-1 ~ a, and hence p = a, for sorne i. • Examplc. If A = Z , a = (m), Iet p¡ (1 ~ i ~ r) be the distinct prime divisors
If a, 6 are ideals in a ring A, their ideal quotient is ofm . Then r(o) = (pl ··· Pr) = n t=l (p,).
(a:ó) = {x E A :xó ~ a} Proposition 1.16. Let a, ó be ideals in a ring A such rhar r (a), r(&) are
coprime. Then a, ó are coprime.
which is an ideal. In particular, (O: ó) is called the annihilator of ó a nd is al so
denoted by Ann (ó): it is the set of aH x E A such that xó = O. In this notation Proof r(a + ó) = r(r(a) + r(b)) = r( l) = (1), hence o+ ú = (l) by (1.13).
the set of all zero-divisors in A is
D = U Ann (x). EXTENSION ANO CONTRACfiON
x~O
If 6 is a principal idea) (x), we shall write (a: x) in place of (a : (x)). Letf: A - >- B be a ring homomorphism . If a is an ideal in A, the setf(a) is not
necessarily an ideal in B (e.g., let f be the embedding of Z in Q, the field of
Example. If A = Z, a = (m), ó = (n), whcre say m = Op pu,, n = OP pv, , rationals, and take a to be any non-zero ideal in Z.) We define the extension
then (a:b) = (q) Where q = p Yp and np ae of a to be the ideal Bf(a) generated by f(a) in B : explicitly, ae is the set of all
y, = max {p., - vp, O) = f.l..p - min (J-Lp, vP) . sums 2: y,f(x1) where x 1 E a, y, E B: ·
Hence q = m/(m, n), where (m, n) is the h.c.f. of m and n. If ú is an ideal of B, then ¡ - 1 (ó) is always an ideal of A, called the contrac-
tion be of ú. If ó is prime, then óc is prime. If o is prime, at need not be prime
Exercise 1.12. i) a ~ (a:&) 1 (for example, f: Z -+ Q , a # O; then oe = Q, which is nota prime ideal).
ii) (a:b)b ~ a We can factorize fas follows:
~ii) ((a:b):c) = (a:flc) = ((a:c):&)
IV) en,a,:b) = n,
(a1:b) A ~f(A) ~ B
v) (a:¿, b,) = n,
(a:b,). where p is surjective and j is injective. For p the situation is very simple (1.1):
If a is any ideal of A, the radical of a is there is a one-to-one correspondence between ideaJs of f(A) and ideals of A
which contain Ker (j), and prime ideals correspond to prime ideals. For j, on
r(a) = {x E A :x" E a for sornen > O}. the other hand, the general situation is very complicated. The classical exarnple
lf;: A -+ A/a is the standard homomorphism then r(a) = 4>- 1 (W ) and hence is from algebraic number theory.
r(a) is an ideal by {1.7). ' A"'
•
JO Rl NGS ANO IOEALS EXERCISES 11
Example. Consider Z ->- Z [i], where i = -v=l. A prime ideal (p) of Z mayor i) f is a unit in A[x ] <=> ao is a unit in A and a
h .
m •
.
1, ••• • an are ru 1poten t. (lf
may not stay prime when extended to Z[i]. In fact Z [i] is a principal ideal ho + b 1 x + · · · + bm X IS t e mverse of [, prove by ind t '
,+1 _ H
1
. . ' uc ton on r that
domain (beca use it has a Euclidean algorithm) and the situation is as follows: a~ h.'" _, - . 0 cncc s 1ow that an .•s mlpotent, and then use Ex. l .]
ii) f ts mlpotent <=> ao, a1, ... , a" are ntlpotcnt.
i) (2)e = ((1 + i)2), the square of a prime ideal in Z(i] ; ii i) ¡ is a zcro-divisor <=> thcre exists a :f:. O in A such that arf == 0 [Ch
· 1 b b · oose a
ii) Jf p = 1 (mod 4) then (py is the product of two d istinct prime idcals polynomta g = o + 1x + · · · + bmx'" of lcast degree m such that fg = o.
{for example, (5)e = (2 + i)(2 - i)); Then anbm = O, hence ang = O (because ang annihilates ¡ and h d
h · d · as egree
~ m: ) N b
ow s .o~ . y .m uct10n that an _,g == O (O ~ r ~ n).]
üi) If p = 3 (mod 4) then (p)e is prime in Z(i]. iv)/tssatdtobe prumtwe Jf (ao,al, ... ,an) = {1 ). Provethat if" e A[] h
r. ·
g ts pn· · ·
mJtJve ¡ an d g are pnmthvc.
· .. " g x , t en
Of these, ii) is not a trivial result. It is effec tively equiva len t to a t heorcm of 1 <=>
Fermat which says that a prime p = 1 (mod 4) can be cxpressed , csscn tially 3 • Gencralize the results of Exercise 2 toa polynomial ring A[x 1, ••• ,
x , ] 1·n 5evera1
uniquely, as a sum of two in teger squares (thus 5 = 22 + 12 , 97 = 92 + 4 2 , indcterminates.
etc.). 4. In the ring A[x], thc Jacobson radical is cqual to the nilradical.
In fac t the behavior of prime ideals under extensions of this sort is o ne of the
5. Let A be a ring and Jet A[[x ]] be thc ring of formal power series¡= 2:. 0 anxn
central problems of algebraic number theory. · with coefficients in A . Show that
i) fis a unit in A[[x ]] <=> ao is a unit in A .
Letf: A -+ B, a and ó be as befare. Then
ii) If f is nilpotent, then O n is nilpotcnt for all n ~ O. Is the converse true ?
Proposition 1.17. i) a ~ acc, ó 2 f> Ct; •
(Sec Chapter 7, Exercise 2.)
iii) f belongs to the Jacobson radical of A [[x]] <=> a0 belongs to the Jacobson
ii) OC = DCtC a e = a tCC,
, ' radical of A .
iii) Jje is tlze set of contracted idea/s in A ancl if E is the set of extended icleals · iv) Thc contraction of a maximal ideal m of A[[x]] is a maximal ideal of A and
lll ÍS generated by me and X.
. '
in B, then e = {al aec = a}, E = {új (> C!! = ú}, anda 1-+ ae is a bijective map
v) Every prime ideal of A is the contraction of a prime ideal of A[[x]].
of e onto E, whose inverse is ú ~--+ úc.
Proof i) is trivial, and ii) follows from i). 6. A ringA is such that every ideal not contained in the nilradical contains a non-
zero idempotent (that is, an element e such that e2 = e '# O). Prove that the
iii) If a E e, then a = óc. = ócec = ate; co nverse! y if a = aec then a is the nilradical and Jacobson radical of A are equal.
contraction of a 6 • Similarly for E. • 7. Let A be a ring in which every elernent x satisfies x n = x for sorne 11 > 1
· (depending on x). Show that every prime ideal in A is maximal.
Exercise 1.18. If a 1 , a 2 are idea/s of A and if ó1 , ó2 are iclea/s of B, then
8. Let A be a rii1g ':/; O. Show that the set of prime ideals of A has minimal ele-
(al + a2)e = a~ + a~, (ó 1 + ú2Y 2 ú f + ú~, ments with respect to inclusion.
(al n a2 )c ~ a! n a~, (ó1 n ó2Y = ói n ó2, 9. Let a be an ideal ':/; (1) in a ring A . Show that a = r(a) <=>a is an intersection
(a 1 a 2 )e = a~a~, Có1 ó2Y 2 óió2, of prime ideals.
(a 1 : a 2 )e ~ (a!: a~), (ó1:ú2)c ~ (ó i:ú~),
10. Let A be a ring, ~ its nilradical. Show that the followi ng are equivalent :
r(aY ~ r(ae), r(óY = r(óc). i) A has exactly one prime ideal ;
Tlze set of idea/s E is e/osee/ under sum and product, and C is etosed under ii) every element of A is either a uni t or nilpotent ;
the other tlzree operations. iii) A/W is a field.
11. A ringA is Boo/ean if x 2 = x for all x e A. In a Boolean ringA, show that
i) 2x = O for all x E A;
EXERCISES ii) eve·r y prime ideal l:' is maxi mal, and A/'V is a field with two elements;
iii) every fin itely generated ideal in A is principal.
1 • Let x be a nilpotent element of a ringA. Show th at 1 + x is a unit of A . D educe 12. A local ring contains no idempotent ':/; O, l .
that the sum of a niJpotent element a nd a unit is a unit.
Construction of an algebraic closure of a fteld (E. Artin).
2. L~t A be a ~ing and let A[x ] be the ring of polynomials in a n indeterminate x, 13
Let K be a field and let ~ be the set of a ll irreducible monic polynomials 1 in onc
WJth coefficJents in A. Let f = ao + a1x + · · · + anxn E A[x]. Preve that '
•
12 RINGS ANO IDEALS
EXERCISES 13
indeterminate with coefficients in K . Let A be the polynomial ring over K 18. For psychological reasons it is sometimes conveoient to d .
. . enote a Pnme ideal
generated by indeterminates x" one for each f e I:. Let a be t~e ideal of A of A by a letter such as x or y when thinkmg ofitas a po·1 t r X _ S
geoerated by the polynomialsf(x1) for alife :E. Show that a ~ (1). When thinking of x as a prime ideal of A, we denote it by p ~ ° 'call- fpec (A).
10
Let m be a maximal ideal of A containing a, and Jet K1 = A /m . Then K 1 is it is the same thing). Show that " g¡ y, 0 course,
ao extension field of K in which each f e :E has a root. Repeat the construction i) the ~et {x} is closed (we say that x is a "closed point") in Spec (A) ~ .
with K 1 in place of K, obtaining a field K2, and so on. Let L = U;'.. t Kn. Then mruomal; P" IS
Lis a field in which each[e I: splits compJetely into linear factors . Let K be the ÍÍ) {x} = V(tJx) ;
set of all elements of L which are algebraic over K. Then K is an algebraic iii)
- <::- Px S P11;
y E {x}
closure of K .
14. In a riog A, let :E be the set of all ideals in which every element is a zero-divisor.
i:
iv) X is a. To-sp~cehb(thihs means that x, y are distinct points of X, theo either
there JS a nc1g or ood of x whtch does not contain y or else th . ·
Show that the set I: has maximal eJements and that every maximal element or::E is . hborh oo d of y wh'te h d oes not contain x).
netg , ere ts a
a prime ideal. Heoce the set of zero-divisors in A is a union of prime ideaJs.
19. A topological space ~ is ~ai d to be irreducible if X ~ 0 and if every pair of
The prime spectrum o[ a ring non-emp.ty open scts m X mtersect, or equivalently if every non-empty open set
15. Let A be a ring and Jet X be the set .of all prime ideals of A. For each subset is dense m X . Show that Spec (A) is irreducible if and only if the nilradical of
E of A, Jet V(E) denote the set of all prime ideaJs of A which contain E . Prove A is a prime ideal.
tbat .
20. Let X be a topological space.
i) if a is the ideal generated by E, then V(E) = V( a) = V(r(a)) .
ii) V(O) = X, V(l) = 0 . i) If Y is an irreducible (Exercise 19) subspace of X, then the closure y of y
iií) íf (E,),er ís any family of subsets of A, then in X is irreducible. ·
ii) Every irreducible subspace of X is contained in a maximal irreducible
V( U
IGI
E,) = n
te/
V(E1) . subspace.
iv) V(a n b) = V(ab) = V(a) u V(ó) for any ideals o, 6 of A . iii) The maximal irreducible subspaces of X are closed and cover X. They are
These results show that the sets V(E) satisfy the axioms for closed sets called the irreducible components of X. What are the irreducible components
in a topolog~cal space. ~he resulting topology is called the Zariski topology. of a Hausdorff space?
The topolog¡caJ space X 1s called the prime spectrum of A, and is written Spec (A). iv) If A is a ring and X = Spec (A), then the irreducible components of X are
the closed sets V(p), where p is a minimal prime ideal of A (Exercise 8).
16. Draw pictures of Spec (Z), Spec (R), Spec (C [x]), Spec (R[x]), Spec (Z[x]).
17. For each fe A, let Xr denote the complement of V([) in X = Spec (A). The 21. Let t/>: A ~ B be a ring homomorphlsm. Let X = Spec (A) and Y = Spec (B).
. sets X, are open. Show that they form a basis of open sets for the Zariski If q e Y, then t/> - l(q) is a prime ideal of A, i.e., a point of X. Hence tf> induces a
topology, aod that mapping tf>* : Y ~ X. Show that
i)x,nxg = 1x g; i) If 1 e A then cp* - 1 (X1) = Y41c1, and hence that tf>* is continuous.
=
ii) Xr 0 ~ f is nilpotent · ii) If a is an ideal of A, then tf>*- 1 (V(a)) = V(a41).
iii) Xr = X~ lis a unit · ' iii) If ó is an ideal of B, then cp*(V(ó)) = V(bc).
i~) ~ . = Xg ~ r((f)) = 'r((g)); · iv) If t/> is surjective, then tf>* is a homeomorphism of Y onto the closed subset
)
15
~uast-compact (that ts, every open covering of X has a finite sub· V(Ker (t/>)) of X. (In particular, Spec (A) and Spec (A/91) (where 91 is the
covenng). \ nilradical of A) are naturaUy homeomorphic.)
~~) More generally, each Xr is quasi-compact. .. v) lf t/> is ínjective, then tf>*( Y) is dense in X. More precisely, ;•( Y) is dense in
vu) An open subset of X is u · · · X~ Ker (t/>) s; VC.
sets X 1• q as1-compact 1f and only if it is .a finite union of
vi) Let r/J: B - >- C be another ring homomorphism. Then (1/J o ; ) • = tt>• o ,p•.
The sets Xr are caiJed basic open sets of X = Spec (A) vii) Let A be an integral domain with just one non-zero prime ideal p, and let K
[T0 prove (v), remark that it is h . ·
•
open sets X (i e /) Sh h enoug to consJder a covering of X by basic be the field of fractions of A. Let B = (A/p) x K. Define;: A--+ B by
,, . ow t at thejj1 generate th 't 'd 1 . t/>(x) = (x, x), where x is the image of x in A jp . Show that ;• is bijective
is an equation of the form e um I ea and hence that therc
but not a homcomorphism.
r
1 L g,Jj
= te/ (g, E A) 22 · Let A = flf.. 1 A, be the direct product of riogs A,. Show that Spec (A) is the
disjoint union of open (and closed) subspaces X, where X, is canonically
where J is some finile subset of 1 Then the X (. J)
· r 1 1 e cover X.] homeomorphic with Spec (A 1).
•
-
14 RINGS ANO IDEALS EXERCISES 15
Conversely, Jet A be any ring. Show that the following statcmcnts are and multiplyi ng thcir valucs). For each x E X 1 l b
equivalent: sueh that f (x) = O. Thc ideal m_,. ¡5 maxim;l cb mx e ~h: set of all f E C (X)
i) X = Spec (A) is disconnected. . . (surjective) homomorphism C(X) -+ R which' ta~~u;et~t 15 the kernel of the
ií) A ~ A1 x A 2 where ncither of thc rings A1, A2 JS the zcro nng. Mnx ( C(X)), we ha ve therefore dcfincd a mappin . X J<¡>· lf X denotes
iii) A contains an idempotent ~ O, l. Wc shall show that p. is a homeomorphism 0 ; ; · -+ 0 • namely x H mx.
In particular, the spectrum of a local ring is nlways connected (Exercise 1
') Le t m b e any max1ma
. l 'd
1 cal of C( X), and let v =onto A .
V(m)
12). mon zeros of the functions in m: that is, be the sct of com-
23. Let A be a Boolean ring (Excrcise 11), and let X = Spec (A). . V = {x E X :f(x) = O for all f E m }.
i) For each/E A, the set X 1 (Exercise 17) is both open and closed m X.
") Let ¡;lt ·· · t¡, E A · Show that X 11 u· ·· u X 1" = X 1 for sorne / E A. Supposc that Vi s cmpty. Thcn for each x Ex the · f
. . . re ex1sts _,. E m such that
JI n fx(x) ~ O. S1nce fx JS contmuous, there is an open ne·1ghb h d U ·
iii) The sets x, are the only subsets of X which are both open and closed. . X 1. h r d . or oo
m . o n w 11c ; x oes .not vamsh. By compactncss a fim1'te num ber ·of the
:e of x
[Let y ~ X be both open and closed. Since Y is open, it is a union of basic
ne1ghborhoods, say U x 1 , • • • , Ux"' cover X. Let
open sets X 1• Since Y is closed and X is quasi-compact (Exercise 17), Y is
quasi-compact. Hence Y is a finite union of basic open sets; now use (ii) f = ¡; + .. . +
1
f2
X" •
above.]
•
iv) X is a compact Hausdorff space.
Then f ~oes not vanish at any point of X, hence is a unit in C(X). But this
contrad1cts / Em, hence Vis not empty.
24. Let L be a Iattice, in which the sup and inf of two elements a, b are denoted by Let x be a point of V. Then m s;;: m_,.., hence m = mx because m is
a v b and a A b respectively. L is a Boo/ean /attice (or Boo/ean a/gebra) if maximal. Hence p. is surjective.
i) L has a Ieast element and a greatest element (denoted by O, 1 respectively). ii) By Urysohn's l.emma (this i.s thc only non-trivial fact required in the argu-
ii) Each of v, A is distributive over the other. ment) the contmuous funct10ns separate the points of X. Hence x ':f. y =>
iii) Each a EL has a unique "complement, a' E L such that a v a' = 1 and m.:c # tn 11 , and therefore p. is injective.
a A a' = O. iii) Let f E C( X); let
(For example, the set of aH subsets of a set, ordered by inclusion, is a Boolean U1 = {x E X:f(x) ':f. 0}
lattice.)
and let
Let L be a Boolean Jattice. Define addition and multiplication in L by the
rules Ü 1 = {m E X:frt m}
a + b = (a A b' ) v (a' A b), ab = a A b. Show that p.( U1 ) = Ü1• The open sets U1 (res p. Ü1) forma basis of the top-
Verify that in this way L becomes a Boolean ring, say A(L). ology of X (resp. X) and therefore 1-' is a homeomorphism.
Conversely, starting from a Boolean ring A, define an ordering on A as Thus X can be reconstructed from the ring of functions C(X).
follows: a ~ b means that a = ab. Show that, with respect to this ordering, A Affine a/gebraic varieties
is a Boolean lattice. [The sup and inf are given by a v b = a + b + ab and
27. Let k be an algebraically closed field and let
a A b = ab, and the complement by a' = 1 - a.] In this way we obtain a
one-to-one correspondence between (isomorpbism classes of) Boolean rings and /a(ll, ... , In) = O
(isomorphism classes of) Boolean lattices.
be a set of polynomial equations in n variables with coefficients in k. The set X
lS. From the last two exercises deduce Stone's theorem, that every Boolean lattice of all points x = (x1 , ••• , Xn) E k" which satisfy these equations is an affine
is isomorphic to the lattice of open-and-closed subsets of sorne compact Haus- a/gebraic variety.
dorff topological space. Consider the set of all polynomials g E k[t 1 , • • • , In] with the. pr~perty th~t
g(x) = O for all x E X . This set is an ideal I(X) in the polynom•al rmg, and IS
16. Let A be a riog. The subspace of Spec (A) consisting of the maximal ideals of A,
with the induced topology, is called the maximal spectrum of A and is denoted by called the ideal of tite variety X. The quotient ring
Max (A). For arbitrary commutative rings it does not have the nice functorial P(X) = k[t11 .. . , l n]/l(X)
properties of Spec (A) (see Exercise 21), because the inverse image of a maximal
ideal under a ring homomorphism need not be maximal. is the ring of polynomial functions on X, beca use two polynomials g, h d~fine the
Let X be a compact Hausdorff space and let C(X) denote the ring of all same polynomial function on X if and only if g - Ir vanishes at every pomt of X,
real-valued continuous functions on X (add and multiply functions by adding that is, if and only if g - hE J(X ).
--~ . ' ----,
. ......
SUBMODULES AND QUOTIENT MODULES defined A-module homomorphism of L/N onto L/ M, and its kernel is M/N ;
hence (i).
A submodule M' of Mis a subgroup of M which is el o sed under multiplication by ii) The composite homomorphism M 2 --+ M 1 + /vf 2 ~ (M1 + M2)/ Nf1 is
elements of A. The abelian group M /M ' then inherits anA-module structure surjective, and its kernel is M 1 () lv/2 ; hence (ii). •
from M, defined by a(x + M ') = ax + M '. The A-module lvf/M' is the
quotient of M by M'. The natural map of Monto M / M' is anA-module homo- We cannot in general define the product of two submodules, but we can
morphism. There is a one-to-one order-preserving correspondence between define the product aM, where a is an ideal and ManA-module; it is the set ofall
submodules of M which contain M', and submodules of M " Uust as for ideals ; finite sums 2: a1x 1 with a1 E a, x, E M, and is a submodule of M.
the statement for ideals is a special case). If N, Pare submodules of M, we define (N:P) to be the set ofall a E A such
Iff: M -r Nis anA-module homomorphism, the kernel offis the set that aP S N; it is an ideal of A. In particular, (0: M) is the set of all a E A such
that a M = O; this ideal is called the annihilator of M and is also denoted by
Ker (f) = {x E M :f(x) = O} ~nn (M). If a s Ann (M), we may regard M as an A/a-modu~e,. a~ follows:
and is a submodule of M. The image off is the set Jf x E Aja is represented by x E A, define .~Ym to be xm(m E lv/): thts ts mdepend-
ent of the choice of the representative x of i, since aM = O.
Im (f) = f(M)
20 MODULES
FINITELY GENERATED MODULES
21
An A-module is faitliful if Ann (M) = O. If Ann (M) = a, thcn M is FJNITELY GENERATED MODULES
faithful as an Aj a-module.
A free A-module is one which is isomorphic to an A-mod
Exercise 2.2. i) Ann (M + N) = Ann (M) 1\ Ann (N) . Ef1tor 1~,;¡, where each Mt ~ A (as an A-module). The notation ~~> ?f the ~orm
15
ii) (N:P) = Ann ((N+ P)/N ). uscd. A finitely generated free A-module is therefore isom h" somettmes
h ~ A
· }1 1·s denote d by A n . (Conventionall OArpo . IC hto A (i) ... w
( 11 sumrnan d s) , w 1c
If x is an element of M, the set of all multiplcs ax(a E A) is a submodule of y, IS t e zero module
deno ted bY · O) •
Al, denoted by Ax or (x). If M = Í:ter Ax" the X¡ are snid to be a set of ge11•
erators of M; tbis means that every element of NI can be exprcsscd (not ncces. Proposition 2.3. M is a finitely generated A-module <:> M is iso .
sarily uniquely) as a finite linear combination of the X¡ with cocfficien ts in A. a quotient of A" for some integer n > o. morphrc to
An A-module Mis said to be fin itely generated if it has a finite set of genera. Proof => : Let X 1, • . . , X n genera te M . Define <P : A n -+ M b -l.(
tors. TI -J. • A d 1 y 't' a lt • . . , an) =
x + · · · L
-r
'
a"x". len 't' IS an -mo u e homomorphism 0 1 M d
n ° •an there-
01 1
fore M ~ A"/ Ker (4>).
More generally, if (M1) 1G1 is any family of A-modules, we can define their direct
sum Ef),&1 M 1; its elements are fami lies (x1) 1Gr su eh that x, E M 1 fo r ea eh i E l and wlzere tlze a 1 are in a.
almost all x1 are O. If we drop the restriction on the number of non-zcro x's wc Proof Let X¡, ••. , x" be a set of generators of M. Then each cp(x1) E aM, so that
have the direct product Titer M 1• Direct sum and di rect product are therefore the we ha ve say <f>(x1) = 27 . . 1 a11x 1 (l ~ i ~ n; a11 E a), i.e.,
same if the index set 1 is finite, but not otherwise, in general.
Suppose that the ringA is·a direct product Of.. 1 A 1 (Chapter 1). Then the n
of A as a direct sum of ideals, we have Proof. Take 4> = identity, x = 1 + a 1 + · · · + a" in (2.4). •
Proposition 2.6. (Nakayama's /emma). Let M be a finitely generated
A-module anda an ideal of A contained in the Jacobson radical mofA . Then
aM = M imp/ies M = O.
where b, • (±), .. ,a,. Each ideal a, is a ring (isomorphic to A/ b1). The identity Fi~st.Proof By (2.5) we have xM = O for sorne x = 1 (mod m). By (1 .9) x is a
element e, of a, is an idempotent in A, and a, = (e1). unit In A, hence M = x- 1x M = O. •
>
-
22 MODULES
EXACT SEQUENCES
23
Second Proof. Suppose M :F O, and Iet u11 ••• , u~ be a minimal set of gene . ii) Ler
ators of M. Then un e aM, bence we ha ve an equatwn of the form un == a1u1 ~ O -r N ' ~ N~ N "
· · · + anun, with the a, E a. Hence (S)
{J - Gn)Un = a1U1 + ' ' ' + Gn - llln - 1; be a sequence of A -modules and homomorphisms. TI zen tlze sequence (S) is
exact <:>foral/ A-modu1es M, the sequence
since an E ffl, it folJows from (1.9) that 1 - an is a unit in ~ ..Hence Un belongs lo
the submodule of M generated by u1 , •.• , un-1: contrad1ctwn. • O - >- Hom (M, N ') ~ Hom (M, N)~ Hom (M, N ")
(5')
Corollllry 2.7. Let M be a.finitely generated A-module, N a submodule of M is exact.
a s; man ideal. Then M = aM + N => M = N. ' All four parts of this proposition are easy exercises. For exam le
Proof. Apply (2.6) to M /N, observing tha t a(lvf/N) = (a ¡\tf + N)/N. • that (4') 1·s exac t 10r
r
a11 N . F'trst o f a 11 , smce
· · · P ' suppose
iJ 1s mjective for all N it foil tha
. N t l - - O h . ows t
Let. A be a local ring, m its maximal id.eal, k .~ A/ m its residue field. Let M vis surjcctive. e':, we 1ave u o v =:= , ~ at ts v o u o¡= o for allj: M "-+ N.
be a .fimtely generated A -module. M / mlvf JS anmh1lated by m, hence is natural! , Taking N to be M and f to be the 1dent1ty mapping, it follows that v o u = 0
an A/ rn-module, i.e., a k-vector space, and as such is fini te-dimensional. } hence Im (u) s; Ker (v). Next take N = M flm (u) and 1et e/> : M -r N b th'
projection. Then e/> E Ker (ü), hence there exists t/J: M " -r N such that e/> = ~ o :
Proposition 2.8. L et x, ( 1 ~ i ~ n) be elements of M whose images in Consequently Im (u) = Ker (~) 2 Ker (v) . • ·
Mfmlvf forma basis of this vector space. Then the x, genera te M .
Proposition 2.10. Let
Proof. Let N be the submodule of M generated b;r the x,. Then the composite
map N -r M -r M f mM maps N onto M fmM, hence N + mM = M, hence 0-+ M ' ~M ~ M "-rO
N= M by (2.7). • r t q. {,r
O -r N ' u' >- N ;¡;+ N" -r O
EXACf SEQUENCES be a commutative diagram of A-modules and lzomomorphisms, with the rows
A sequence of A-mod ules and A-homomorphisms exact. Then there exists an exact sequence
··· Mt- 1
1
' M,
1
'•
1
M, +1 · • · (O) O-r Ker (/') ~ Ker (f) ~ Ker (f) ~
a· o·
Coker (/')-+ Coker (f)-+ Coker (j")-+ O (6)
is said to be exact at M , if Im (¡;) = Ker u;+ 1) . The sequen ce is exact if it is •
exact at each M,. In particular : in which ü, o are restrictions of u, v, and ü', ü' are induced by u', v'.
O -r M ' 4 M is exact <:> f is injective; (1) The boundary hom omorphism d is defined as follows: if x" E Ker (j"), we have
M !.., M " -r O is exact <:> g is surjective; (2) x" = v(x) for sorne x E M, and v'(f(x) ) = f "(v(x)) = O, hencef(x) E Ker (v') =
Im (u'), so thatf(x) = u'(y') for sorne y' E N '. Then d(x") is defined to be the
?-r M ' 4 . M 9
? M " -r O is exact <:> f is injective, g is surjective and g
image of y' in Coker (/'). The verification that d is well-defined, and that the
mduces an Jsomorphism of Coker (/) = M /f(M ') onto M ". (3)
sequence (6) is exact, is a straightforward exercise in diagram-chasing which we
A sequence of type (3) is caBed a slzort exact sequence. Any long exact
leave to the reader. •
sequence (O) can be split up into short exact sequences: if N, = Im UD==
Ker if. u ), we have short exact sequences O-+ N,-+ M , -+ N, +1 -+ o for each i. Remark . (2.1O) is a special case of the exact homology sequence of homological
Proposition 2.9. i) Let
algebra.
Let C be a class of A-modules and let .\be a function on C with values in~
M ' ">- M 2+ M " -r O (4) t
(or, more generally, with values in an abelian group G). The function Á ts
be a sequence of A-modules and lzomomorplzisms. Then the sequence (4) is additive if, for each short exact sequence (3) in which all the terms belong to C,
' exact <:> foral/ A-modules N, tlze sequence we have A(M' ) - A(M) + .\(M") = O.
O-+ Hom (M ", N) .!r Hom (M, N ) a? Hom (M ', N ) (4') ' Example. Let A be a field k and let C be the class of all finite-dimensional
•
1s exact. k-vector spaces V. Then V r-:.
dim Vis an additive function on C.
2+ I.C.A•
•
24 MODULES
TENSOR PROOUcr OF MODULES
25
Proposition 2.1J. Let O ~ M o - >- M1 - >- · · · - >- lvln -+- O be an exacr Let T = e¡ D. For each basis element (x, y) of e, let x ®Y denot .
quence of A-modules in whiclz al/ the modules A1, ancl the kerne/s 0¡ al/ ~e. . e in T. Then T is generated by the elements of the form x 10. d e tts
homomorphisms be long to c. Then for any additive fimction ,\ on e We ha:: unag
our definitions we have
'CI Y, an from
n
L (- 1)1A(M1) = O. (X + x') @ y = X @ y
(ax ) ® y
+ x'
=
@ y' X @(y + y ' ) =
x ® (ay) = a(x ® y)
X ®y + X ® '
Y'
f "' o
Proof. SpJit up the sequence into short exact sequences Equivalently, the mapping g: M x N ~ T defined by g(x, y) = x ®Y is
O ~ N, ->- A1, ~ N, +1 -+ O A-bilincar. .
Any map f of M_ x ~ mto an A-module. P extends by linearity to an A-
(No = Nn+I = 0). Then we have ,\(M,) = ,\(N,) + ,\(N,+¡). Now takc the module homomorphtsm j: C - >- P. Suppose tn particular that J is A-bilinea
alternating sum of the ,\(M,), and everything cancels out. • Then, from the definitions, J ~anishes on all the geoerators of D, hence 00 t:;
whole of D, and therefore mduces a well-defined A-homomorphism f ' of
TENSOR PRODUCI' OF MODULES
T = C/ D int? P su7~ that f '(x ® y) = f(x, ~). The m~ppiog f' is uniquely
defincd by thts condtt!On, and therefore the pa1r (T, g) sahsfy the conditions of
Let M, N, P be three A-modules. A mapping f : A1 x N ->- P is said to be the proposition. •
A-hilinear if for each x E M the mapping y 1->- f(x, y) of N into Pis A-linear, and Remarks. i) Thc module T constructed above is called the tensor product of M
for each y E N tl1e mapping x H- f(x, y) of M into Pis A-linear. and N, and is denoted by M ® A N , or just M ® N if there is no ambiguity
We shall construct an A-module T, called the tensor product of lvf and N about the ringA . It is generated as an A-module by the "products" x ®y. If
wíth the property that the A-bilinear mappings Jrf x N~ P are in a natural' (x,),e¡, (y1) 1e 1 are families of generators of M, N rcspectively, then the elements
one-to-one correspondence with the A-linear mappings T ~ P, for all A- x, ® y, generate M ® N . In particular, if A1 and N are finitely generated, sois
modules P. More precise!y: M ® N.
Proposition 2.12. Let M, N be A-modules. Tiren there exists a pair (T, g) ii) The notation x ® y is inherently ambiguous unless we specify the tensor
consisting of anA-module T andan A-bilinear mapping g: M x N -+ T, with ' product to which it belongs. Let M', N' be submodules of M, N respectively,
tite fo/lowing property: and let x E M ' and y E N ' . Then it can happen that x ®y as an element of
Given any A-module P and any A-bilinear mapping f: M x N~ P, M ® N is zero whilst x ® y as an e/ement of M' ® N' is non-zero. For
there exists a unique A-linear mapping f ' : T ->- P sueh that f = f ' o g (in example, take A = Z , M = Z, N = Z f2Z, and let M ' be the submodule 2Z of Z,
other words, every bilinear function on M x N factors tlzrouglz T) . - · whilst N ' = N . Let x be the non-zero element of N and consider 2 ® x. Asan
Moreover, if(T, g) and (T', g') are two pairs witlz this property, titen there , element of NI ® N, it is zero because 2 ® x = 1 ® 2x = 1 ® O = O. But as
exists a unique isomorphism j: T ~ T ' such tlzat j o g = g'. an element of M' ® N' it is non-zero. See the example after (2.18).
Proof. i) Uniqueness. Replacing (P,f) by (T', g') we get a unique j: T ~ T' However, there is the following result :
such that g' = j o g. Interchanging the roles of Tand T ', we getj': T ' -+ T such
Corollary 2.13. Let x , E A1, y 1 E N he such that :L x, ® y, = O in Af ® N.
thatg =j' og'. Each ofthe compositionsj oj ',j' oj must be the identity, and
therefore j is an isomorphism. Tiren there exist .finitely generated submodules Mo of M and No of N such that
L x, ® y, = O in M 0 ® No.
ii) Existence. Let e denote the free A-module A <M N>. The elements of e
X
are formal linear combinations of elements of M x N with coefficients in A, Proof If ¿ x, ® y, = O in M ® N, then in the notation of the proof of (2.11)
i.e. they are expressions of the forro 2:~- 1 a,· (x 11 y,)(a, E A, x, E M, y 1 E N). we have ¿ (x11 y,) E D, and therefore 2: (x11 y,) is a finite sum of generators of
0~ ~
Di
. Let D be the submodule of e generated by all elements of e of the follow-
Let Mo be the submodule of M generated by the x, and all the elements
mg types: which occur as first coordinates in these generators of D, and define No suru-
(x + x', y) - (x, y) - (x', y) larly. Then 2: x, ® y, = O as an element of Mo ® No. •
(x, y + y') - (x, y) - (x, y') m) We shall never again need to use the construction of the tens~r prod~ctl
(ax, y) - a·(x, y) · m
glven . (2.12), and the reader may safely forget 1t. 1·r he prec.ers
Jl •
Wnat 1s essent1a
(x, ay) - a · (x, y). to keep in mjnd is the defining property of the tensor product.
•
p
a) X ® y H- y ® X
such that
b) (x ® y) ® z H- x ® (y ® z) H- x ®y ® z
e) (x, y) ® z H- (x ® z, y ® z) (/ ® g)(x ® y) = f(x) ® g(y) (x E M, y E N).
d) a® x H- ax. Let f ': M ' -+M" and g ' : N '-+ N" be homomorphisms of A-modules.
Proof. In each case the point is to show that the mappings so described are well Then clearly the homomorphisms (J' o f) ® (g' o g) and (/' ® g') o (f ® g)
defined. The technique is to construct suitable bilinear or multilinear mappings, agree on all elements of the form x ® y in M ® N. Since these elements
and use the defining property (2.12) or (2.12*) to infer tbe existence of bomo- generate M ® N, it follows that
morphisms of tensor products. We shall pro ve half of ii) asan example of the
method, and leave the rest to the reader. (J' o f) ® (g' o g) = (f' ® g') o (J ® g).
We shall construct homomorphisms
such thatf((x ®y)® z) = x ®y® z and g(x ® y ® z) = (x ®y) ® z for Let f: A -+ B be a homomorphism of rings and let N be a E-module. Th:n N
all x E M, y E N, z E P. has an A-module structure defined as follows: if a E A and x E N, then ax ~s ~e
To construct f, fix the element z E P. The mapping (x, y) r+ x ®y ® z fined to be f(a)x. This A-module is said to be obtained from N by restnctwn
(x E M, Y E N) is bilinear in x and y and therefore induces a homomorphism of scalars. In particular, f defines in this way an A-module structure on B.
28 MODULES ALGEBRAS 29
Proposition 2.16. Suppose N is jinitely ge~zera~ed as a B-module and tizar Bis .. ft is nol in general true that, if M ' -+ M -+ M" is an exact sequ
jinitely generated asan A-module. Then N ¡sfimtely generated asan A -module. 11) and l ·
homomorp11tsms, th e sequence M ' ® N-+ M 10. N M" ence of
d 1eS
A- . d by tensoring with an arb'ttrary A-module N is exact '<:::1 ~
rno u ® N
Proof. Let Y1t .. . , Yn genera te N over B, and let X1 , . . . , X m genera te B as an obtatnc ·
A-module. Then the mn products x 1y 1 generate N ovcr A. • Take A = Z and consider the exact sequence o-+ z 4 z h
exaDIP1c. . , w ere
Now Jet M be an A-module. Since, as we ha ve j ust secn, B can be rcgarded (~) ::::: 2x for all x E z. If we tensor w1th N = Z/2Z, the sequence 0 -+ z ®
asan A-module, we can form the A-module A1n = B ® A At/. In fact M 8 carries
-1 ~ 1 _.,.. z ® N ¡5 not exact, because for any x ® y E z ® N we have
a E-module structure such that b(b' ® x) = bb' ® x for all b, b' E B and aU
x E M . The E-module Mn is said to be obtained from M by extension of scalars. (f ® l)(x ® y) = 2x ® y = x ® 2y = x ® O = o,
Proposition 2.17. Jf M is finitely generated as an A-module, then M n is
50
that/ ® 1 is the zero mapping, whereas Z ® N =F o.
finitely generated as a E-module.
The functor T N: M 1-+ M ® A N on the category of A-modules and homo-
Proof. Jf x 17 •• • , Xm genera te Mover A, then the l ® X¡ genera te /vi n over B. •
morphisms is therefore not in general exact. 1~ TN is exact, that is to say if
tensoring with N transforms all exact sequences mto exact sequences, then N is
EXACfNESS PROPERTIES OF THE TENSOR PRODUCT said to be a jlat A-module.
Let f: M x N~ P be an A-bilinear mapping. For each x E A1 the mapping Proposition 2.19. The following are equivalen/, for an A-module N:
y r-r f(x, y) of N into P is A-linear, hence f gives rise to a mapping M --+ .
i) N is flat.
Hom (N, P) whjch is A-linear beca use f is linear in the variable x. Converse}y
any A-homomorphism 4>: M - >- HomA (N, P) defines a bilinear map, namely ii) Jf O -+ M '-+ M-+ M H-+ O is any exact sequence of A-modules, the
(x, y) r+ 4>(x)(y). Hence the set S of A-bilinear mappings M x N -+ P is in tensored sequence O -+ M' ® N-+ M ® N~ M" ®N~ O is exact.
natural one-to-one correspondence with Hom (M, Hom (N, P )) . On the other iii) Jf f: M ' -+ M is injective, then f ® 1: M ' ® N~ M ® N is injective.
hand S is in one-to-one correspondence with Hom (M ® N , P), by the de- iv) Jf f : M ' ~ Mis injective and M, M ' arefinilely generated, then
fining property of the tensor product. Hence we ha ve a canonical isomorphism f ® 1: M ' ® N -+ M ® N is injective.
Hom (M ® N, P) ~ Hom (M, Hom (N, P) ). (l ) Proof i) <=> ii) by splitting up a long exact sequence into short exact sequences.
Proposition 2.18. Let ii) <=> iii) by (2. 18).
M' 4 A1 4 M "-+ O iii) => iv): clear.
(2)
be an exact sequence of A-modules and lzomomorphisms, and /et N be any iv) => iii). Letf: M ' -+ M be injective and let u = 2 x, ® y 1 E Ker (f ® 1),
A-module. Tlzen tlze sequence so that 2./(x;) ® y 1 = Oin M ® N. Let M~ be the submodule of M' generated
by the x; and let u0 denote¿ x; ® y, asan element of M~ ® N. By (2.14) there
1 1
M' ® N ® ) M ® N 9
®
1
:.- M " ® N -+ O (3) exists a fi nitely generated submodule M 0 of M containiogf(M~) aod such that
(where 1 denotes the identity mapping on N) is exact. 2. /(x;) ® y, = O asan element of M 0 ®N. If f 0 : M~~ M 0 is the restriction
off, this means that Uo Q9 1)(u0) = O. Since M 0 and M~ are finitely generated,
Proof. Let E denote the sequence (2), and let E ® N denote the sequence (3).
fo ® 1 is injective and therefore u 0 = O, hence u = O. •
~t P be any A-module. Since (2) is exact, the sequence Hom (E, Hom (N, P))
15 e~c~ by {2.9); hence by (1) the sequence Hom (E ® N, P) is exact. By (2.9)
Exercise 2.20. Jff· A -+ B is a ring homomorphism and Mis aflat A-module,
agam, 1t follows that E ® N is exact. • . then Ms = B ®~t Mis aflat E-module. (Use the canonical isomorphisms
Remarks. i) Let T(M) = Af ® N and let U(P) = Hom (N, P). Then (1) (2.14), (2.15).)
takes the form Hom (T(M), P) = Hom (M, U(P)) for all A-modules Jvf and P.
~n the l~nguage of abstract nonsense, the functor T is the left adjoint of U and U ALGEBRAS
Is ftthed.r1ght adjoint of T Th f f( '. .
1 8 . . . · e proo o 2.18) shows that any functor whtch IS a
Letf: A ->- B be a ring homomorphism. lf a E A and b E B, define a product
e JOmt IS nght exact. Likewise any functor which is a right adJ'oint is Jeft
exact.
ab = f(a)b.
30 MODULES EXERCISES 31
This definition of scalar multiplication makes the ring B into an A-module Ct . • • turn by (2.1 1) corresponds to an A-bilinear mapping
1 18 and thlS 111
a particular example ofrestriction ofscalars). Thus B has anA-module struct p.: D X D ~ D
as well as a nng . structure, and t hese two struct ures are compatt'b le .111 a seUre
which the reader wiJI be a~Ie t~ formula te for himself. The ring B, equipped w~:~ which is such thnt
this A-module structure, rs sard to be an A-algebra. Thus an A-algebra is b p.(b ® e, b' ® e') = bb' ®ce'.
de.finition, a ring B together with a ring homomorphism f; A -)- B. , Y
orcou rse, we could have written
. down
h
this formula directly' but wt'thout some
Remarks. i) In particular, if A is a field K (and B :f: O) then f is injective b such argument as we have g1ven t ere would be no guarantee. that
· fL w · MWe- 11
(1.2) and therefore K can be canonically identified with its image in B. Thus Y dcfined. . . .
K-algebra (K a field) is effectively a ring containing K as a subring. a We have therefore defined a multrp~t~atr~n on the tensor product D =
ii) Let A be any ring. Since A has an identity element there is a unique B ®AC: for elements of the form b ® e 1t IS g1vcn by
homomorphism of the ring of integers Z into A, namely n t-7 n.l. Thus every (b ® c)(b' ®e') = bb' ® ce' ,
ring is automaticaily a Z-algebra.
and in general by
Let f: A~ B, g: A ~ e be two ring homomorphisms. An A-algebra
homomorphism h : B ~ e is a ring homomorphism which is also an A-module
homomorphism. The reader should verify that Jz is an A-algebra homomor-
(2 (b, ® C¡)) ( 2 (bj ® cí)) = 2: (b bj ® c cj).
1 J f,J
1 1
phism if and only if lz o f = g. The reade~ should check that with this multiplic~tion D is a commutative ring,
A ring homomorphismf: A-)- B isfinite, and Bis afinite A-algebra, if B with identity clement 1 ® l . Furthermore, D 1s an A-algebra: the mapping
is finitely generated as an A-module. The homomorphism f is offinite type, and a r+ f(a) ® g(a) is a ring homomorphism A --+ D.
Bis afinitely-generated A-algebra, if there exists a finite set of elements x 11 ••• xn In fact there is a commutative diagram of ring homomorphisms
in B such that every element of B can be written as a polynomial in x 11 •• • , xn B
with coefficients inf(A); or equivalently if there is an A-algebra homomorphism
/1/ "'".u
from a polynomial ring A[t11 • • • , tn] onto B.
A ring A is said to be finitely generated if it is finitely generated as a z- A/ ""' n
algebra. Thls means that there exist finitely many elements x 1 , .. • , Xn in A such "'g~
that every element of A can be written as a polynomial in the x 1 with rational e
integer coefficients.
in which u, for example, is defined by u(b) = b ® l.
If ,¡,: Am -+ An is surjective, then m ~ n. In particular, any A-module is the d irect limit of its finitely generated sub-
If ,¡,: A m -+ An is injective, is it always the case that m ~ n? modules.
12. Let M be a finitely generated A-module and 4>: M - )o- An a surjectivc homo- 18. Let M = (M., ¡.t11), N = (N, v,
be direct systems of A-modules over the same
1)
morphism. Show that Ker (cp) is finitely generated. directed set. Let M, N be the direct limits and JLt: M, -1- M, v,: N, .- N the
[Let e¡, .. . , en oc a basis of An and eh o ose u, E M sueh that if>(u1) = e1 associated homomorphisms.
(1 ~ i ~ n). Show that M is the direct sum of Ker (if>) and the submodule A homomorpliism <P : M _ ,.. N is by definition a fa mil y of A-module homo-
generated by u11 ••• , Un.] morphisms ,¡,,: M 1 -+ N, sueh that ,¡,, o JLtJ = v 11 o ,P1 wbenever i ~ f. Show that
<P defines a unique homomorphism 4> = lim) ,¡,,: M -+ N su eh that 1> o llt =
13. Letf: A -1- B be a ring homomorphism, and let N be a B-module. Regarding N
as an A-module by restriction of scalars, form the B-module N 8 = B ®..t N.
v, o ,p, for all i E / .
Show that the homomorphism g: N-+ N 8 which inaps y to 1 ® y is injective 19. A sequence of direct systems and bomomorphisms
and that g(N) is a direct summand of N 8 •
M -+ N -+ P
[Definep: Ns -+ N by p(b ® y) = by, and show that N 8 = Im (g) Ef) Ker (p). ]
is exact if the corresponding sequence of modules and module homomorphisms
Direct limits
is exact for each i E / . Show that the sequence M N - >- P of direct limits is
14. A partially ordered set 1 is said to be a directed set if for ea eh pair í, j in 1 thcre then exact. [Use Exercise 15.]
exists k E 1 such that 1 ~ k and j ~ k.
Le~ A be a ring, let 1 be a directed set and let (M1)1f:I be a fa mil y of A-modules 1 Tensor products commute witlt direct limits
indexed by /. For each pair ;, j in 1 su eh that ¡ ~ j, let ,.,.,, : M 1 -+ M 1 be an 20· Keeping the same notation as in Exercise 14, let N be any A-module. :h~n
A-homomorphism, and suppose tbat the following axioms are satisfied: (M, ® N, JLtJ ® 1) is a direct system; let P = lin;
(M, ® N) be its direct hmtt.
EXERCISES 35
34 MODULES
25. Let O -+ N' -+ N-+ N" ~ O be an exact sequence with N" fiat Then N
1 •
IS
ftat ~ N is fiat. [Use Bxercise 24 and the Tor exac~ sequence.] .
RJ NGS ANO MODULES OF FRACflONS
37
3 .
ExcrclS •
e Veri'Y
'J •
that rhese definitions
d .
are
_1
independent
.
of rhe choices if
o rep-
(a( ives (a s) anci (b, t), an t1zat S A sa11sjies tlze axioms 0 r a
resell ' . . 'J com-
mutative ring with zdenllty.
We also have a ri.n~ h~momorphism f: A -+ S - A defined by /(x) = xf l.
1
in general mJecttve.
Rings and Modules of Fractions . • 1101
ThtS 15
denote the equivalence class of (a, s), and let S - 1A denote the set of equivaleoce morphism. Precisely:
classes. We puta ring structure on S - lA by defining addition and multiplication
of these "fractions" ajs in the same way as in elementary algebra: that is, Corollary· 3.2. If g: A - > Bis a ring Jwmomorphism sueh tlzat
i) sE S ~ g(s) is a unir in B;
(afs) + (b/t) = (at + bs)fst,
(afs}(bft) = abfst. ii) g(a) = O => as = Ofor sorne sE S;
36 1
RINGS AND MODULES Of FRACTIONS RlNGS ANO MODULES OF FRACTIONS 39
38
1
is exact, i.e., if M ' ~ M~ M" .
•
iii) Every element of B is of tite form g(a)g(s) - ,· then there IS a .. The operation S - 1
Ullique 3
propOSIIIOn 3• • s- l f s -lg lS
isomorpltism h: S -1 A -+ B suclz tltat g = lt o f. Jvf then s - M '-
1 ). s - M1 1
). s - M" is exact at s-1M.
1 1 '
Proof. By (3.1) we ha ve to show that h : S - A -+ B, defined by exact a
\Ve ha ve g o¡ = O, hencc
S - 1go .
S - lf
:= S - l(O)
= O, hence Im (S -1[)
h(afs) = g(a)g(s) - 1 proof S -lg). To prove the rever.se mclus10n, let mfs E Ker (S -1g), then
(this defioition uses i)) is an isomo.rphism. By iii), h is surjective. To show h . ~ J(:r ~ o in s -1M ", hence there extsts t E S such ~hat tg(m) = o in M". But
inJ'ective look at the kernel of h: tf h(afs) = O, then g(a) = O, hence by 11..) rs g(m)/ _ . (tn) since g is an A-module homomorphtsm, hcnce tm e Ker (g) =
have at ,=o for sorne t E S, hence (a, s ) -= (O, 1) , t.e.,
' f
as= · s -1A. • We
O 10 g 1therefore tm = f( m ') rtOr sorne m ' E M ' . H ence m
tg(m) - and · s - 1M we have
Jm CJ_}f( ')/st = (S - 1f)(m' fst) E Im (S - 1[). Hencc Ker (S - 1g) s; Im (S -1f).
Examples. 1) Let p be a prime ideal of A. Then S = A - pis multiplicar m/s - 111
. u1 . 1' . 1 1 d . . )
closed (in fact A - p ts m ttp tcattve y e ose <=> P ts pnme . We write A r
·h r 'd 1 .
1ve!y
ll tOr
•
1 wbt/t .a E p 1?r.mAan 1 e~ m m A¡¡. If bft ~ nt particular, it follows from (3.3) that if M ' is a submodule of M, the map-
1
s - lA in this case. The elements as
tben b r1 p, hence bE S and therefore . t 1s a umt . 1~ - 1M ' -+ s - 1M is injective and therefore S - l M ' can be regarded as a
. md . ¡¡. h t tollows that if a IS
· an'
ideal in Ap and a $ m, then a contams a umt an ts t erefore the whole ri ~~~~odule of S - 1 M . With this conven~ion,
Hence mis the only maximal ideal in A¡¡; in other words, A¡¡ is a local ring. ng.
Corollary 3.4. Formation of fractions commutes with formation of finite
The process of passing from A to Av is called localization at p. sums, finite intersections and quotients. Precise/y, if N, P are submodules
2) S - 1 A is the zero ring <=>O E S. of anA-module M, then
3) Let/E A and let S= {f"}n~o· We write A1 for s - A in this case.
1
i) s - 1(N + P) = s - 1 (N) + s - (P)
1
4) Let a be any ideal in A, and let S = 1 + a = set of all 1 + x where ii) s- (N n P) = s - (N} () s- (P)
1 1 1
Proposition 3.9. Let 4>: M -)o N be an A-module homomorp/Jism. Tiren rhe Proposition 3.11. i) Every ideal in S - lA is an extended ideal.
following are equivalen!: íi) !f a is an ideal in A, then acc = Uses (a:s). Hence ae = (1) ifandonly
i) 4> is injective; if a meets S.
ii) c/>v: M., -> Np is injective for each prime ideal -\'; e
iii) n E <=> no element of S is a zero-divisor in Aja.
iii) c/>m: Mm -> Nm is injective for each maximal ideal m. iv! Tlze prime ideals of S-lA are in one-to-one correspondence (.P ...-. S -lp)
Similar/y with "injective" replaced by "surjective" througltout. Wll/z the prime ideals of A whz'ch don't meet S.
RINGS ANO MODULES OF FRACTIONS EXERCISES 43
42
v) The operar ion S commutes witlz formar ion of finire sums .fi Id called thc residue field at P. which can be obtaincd eithcr as
. .
-t
' produc w1th a 1C • ' . f
inlersectrons and radrcals. ~~. end up of fractions of the integral domam A P or as the residuc field of the
the field
Proof i) Lct b be an ideal in S-tA, and 1et xfs E ú. Thcn x / 1 E b 1 local ring A IJ ·
x e b e and therefore x fs E fi c 41 • Since ó => óce in any case (1.17), it follo' tcnee osition 3.14. Let M be afinitely genera red A-module, S a multiplicative/y
b = f>C41.
=
ws that
x fl = afs for sorne a E a, sE S .:;:. (xs _ a)t ~
::a~d subset of A.
Then S - l (Ann (!vi)) = Ann (S- 1 M).
ii) x E 0 tc (S - la)c <:> this ¡5 truc for two A-modules, 1\1, N, it is true for M + N :
for sorne r E S<=> xst E a<=> x E UsGs (a:s).
0 Proof lf
s - l(Ann(M +N)) = s - 1 (Ann(M)nAnn(N))by(2.2)
iii) 0 E e<=> ate s; a <=> (sx E a for sorne s E S ==> x E o) ·=> no sE S is
. . . A/a. a · = s - 1 (Ann (M)) n S - 1 (Ann (N)) by (3.4)
zero-dIVJSor m = Ann (S - 1 M) n Ann (S - 1 N) by hypothesis
iv) If q is. a prime ideal in. S - lA, then qc is a. pri~e ide~l in .A (tltis much is = Ann (S - 1 M+ s - 1 N) = Ann (S- 1 (M +N)).
true for aoy nng h omo~o~ph2~m). ~onvcrsely, .1f .P IS a pnrnc Ideal in A, thcn
Aj p is an integral domam; 1f S 1s the 1mage of S m A/v, wc ha ve S - 1A/ s - 1v "' f-1ence
.t ¡5 cnough to provc (3.14) for M generated by a single element · then
J ·
M~ A/n (as A-modu 1c), whcrc:_ a = A~~ (M); s - /v/ ~ (S - A)/(S-ln) by
1
S - 1 (A/'fJ) which is either Oor else is contained in the field of fractions of A/p and 1
1
is therefore an integral domain, and therefore S - 1 .P is eithcr prime or is the un·t1 (J.4), so that Ann (S - M) = S o = S (Ann (M)). •
1
Coro/lary 3.12. lfW is the ni/radical of A, the ni/radical of S - lA is S - 1m. • l. Let S be a multiplicatively closed subset of a ring A, and let M be a finitely
generated A-module. Pro ve that S - 1 M = Oif and only if there exists sE S such
Coro/lary 3.13. lf 'p is a prime ideal of A, the prime ideals of the local ring that s/vf = O.
Apare in one-to-one correspondence wit/z tlze prime ideals of A contained in p. 2• Let o be an ideal of a ringA, and Jet S = 1 + a. Show that s - 1o is contni ned
Proof Take S = A - 'p in (3. 11) (iv). • in the Jacobson radical of s- 1 A.
Use this result and Nakayama's lemma to give a proof of (2.5) which does
Remark. Thus the passage from A to A p cuts out all prime ideals except those
not depend on determinants. [If M = aM, then S - t M = (S - 1a)(S - t M),
contained in 'p. In the other direction, the passage from A to A/'fJ cuts out all hence by Nakayama we ha ve S - 1 M = O. Now use Exercise l.]
prime ideals except those containing p. Hence if v, q a re prime ideals such that
3· Let A be a ring, let S and T be two multiplicatively closed subsets of A, and Jet
V 2 q, then by localizing with respect to p and taking the quotient mod q
(in either order : these two operations commute, by (3.4)), we restrict our atten· .u be the image of T in S - 1 A. Show that the rings (ST) - 1A and U - I(S - t A) are
Jsomorphic.
tion to thosc prime ideals which lie between .p and q. In particular, if 'fJ == q we
EXERClSES 45
44 RINGS AND MODULES OF FRACTIONS
. ce (i.e. cvery subset consisting of a single point is closed).
4 Lct !" A -+ B be n homomorphism of rings and Jet S be a multiplicnti v 1 (;1) 1s n T1-spa '
¡¡i) spcc ) 1·s Hnusdorff. .
• closed subset of A. Let T = f(S). Show that S -1 B and r - 1 B are isomorp~¡~ •v) Spcc (;1 d' . ns are satisfied, show that Spcc (A) 1s compact and totally
1
as S - 1 A-modules. If t hese con . ttiO • conncctcd subsets o f Spec ( A ) are those conststmg
h , only . . of a
nn cctcd (1.e. t e
S. Let A be a ring. Suppose that, for each prim~ ideal IJ, thc local ring A¡¡ has diSCO
110
. t t elemcnt ::fi o. Show that A has no mlpotent clemcn t ;t O. If cach 51·ngle point).. .
, 1domain und lvJ an A-module. An element x E M 1s a rorslon
m1po en . 1 d · ? "1n
• egra 1 domn 1
· an m
IS 'n, is A necessarily an mtcgra omnm A be nn tntegra . . . . db
1 12· Lct ' M if Ann (x) ,¡; O, that JS 1f x JS kJlle Y sorne non-zero element of A.
Let A be a ring # o and Jet :E be the set of all. multiplicativcly closcd subscts S element of s'aon elcments of M form a submodule of M . This submodule
6 thnt t 11e 1or .
· O f A SUC11 tllnt
<
o¿ s Show that
'F • •
:E • has
•
max1mal
• •
elemcnts, and that sE t .IS Sho\V d thc torsion submodulc of lvf and lS dcnoted by T (lvl ). If T(M) = o,
maximal if and only if A - SISa mJmmal pnmc 1dcal of A . is calle M ·s snid to be torsio n-free. Show that
1 1
thc modu. e A-module thcn M /T(lvl) is torsion-fre.e.
. A multiplicatively closed subset S of n ring A is snid to be saturated if ·) If M ts any ' . ( )
1 .~ . lvf - )- N is a module homomorph1sm, thcn 1 T(M) s T(N).
xy E S <=> x E S and y E S. u) Jf 1· M ' - :.- M _ ,.. Jo.;!" is an exact scqucnce, then the sequcnce O - T(M' )
'") If 0 -
111
Prove ti1at . . _. T(M) _,.. T(M") is cxact. . .
') s is saturated =- A - S is a un ion of pnme 1dcals. ·v) If Mi s any A-module, then T(M) ts the kernel . of the mappmg x H 1 0 x
.~) If S ¡5 any multiplicatively closed subset of A, there is a uniquc smallcst
11 1
1 f lvf into K ®A M, whcre K is the field of fracttons of A .
saturated multiplicatively closed subset S contatntng
. . S , an d that S ¡5 thc
f
0
. ) show that K may be regarded as the direct limit of its submodules
~~ ~/~ K); using Chaptcr 1.' Excrcise 15 and Exercise 20, show that if 1 0 x ~ O
0
complement in A of the union of the prime ideals which do not mcct S.
(S is caBed tbe saturatlon of S.) in K ® ft1then 1 0 x = Om A~ 0 Mfor sorne~ =fi O. Deduce that ~- 1 x = O.]
lf S = 1 + a, where a is an ideal of A, find S.
• Let s be a multiplicatively closcd subset of an integral domain A . In the n~tation
8. Let s, Tbe multiplicatively closcd subsets of A, such that S s T. Lct cp: s - 1A 13
of Excrcise 12, show that T(S - l M) = S - l(TM). Deduce that the followmg are
-+ r -1 A be the homomorphism which maps each a/s E S - tA to afs considercd
equivalent:
asan elemcnt of r - 1 A. Show that the following statcments are cquivalent :
i) Mi s torsion-free.
i) 4> is bijective. ii) Mv is torsion-(rec for all prime ideals l>·
ii) For each 1 E T, t/ 1 is a unit in S - 1 A. iii) Mm is torsion-free for all maximal ideals m.
iii) Por each 1 E T there exists x E A such that xt E S .
iv) T is contained in the saturation of S (Exercise 7). 14. Lct M be an A-module and a an ideal of A. Suppose that lv/m = O for all
v) Every prime ideal which meets T also meets S . rnaximal ideals m 2 n. Prove that !vi = o/11. [Pass to the Afa-module M /oM
and use (3.8).]
9. The set So of all non-zero-divisors in A is a saturated multiplicatively closed
subset of A. Hence the set D of zero-divisors in A is a union of prime ideals (sce 15. Lct A be a ring, and lct F be the A-module A". Show that every set of n gen-
Chaptcr 1, Exercisc 14). Show tbat every mini mal prime ideal of A is contained erators ofF is a basis ofF. [Let x 1 , ••• , Xn be a set of generators and e1, ... , en
in D . [Use Exercise 6.] the canonical basis of F. Define 4>: F - )- F by 4>Ce1) = x 1• Then 4> is surjective
The ring So 1 A is called the total ring offractions of A . Prove that and we have to prove that it is an isornorphism. By (3.9) we may assumc that A
i) So is the Iargest rnultiplicatively closed subset of A for which the hamo· is a local ring. Let N be the kernel of 4> and let k = A/m be the residue field of A.
morphism A -+ So 1 A is injective. Sincc F is a flat A-module, the exact sequence O-+ N-+ F-+ F-+ O gives an
ii) Every element in So 1A is either a zero-divisor or a u ni t. exact sequence O ;:.- k 0 N ;;. k ® F 10Q >k ® F ;:.- O. Now k ® F = k'"
iii) Every ring in which every non-unit is a zero-divisor is equal to its total ring •
1s an n-dimensional vector space o ver k; 1 ® 4> is surjective, hence bijective,
of fractions (i.e., A - )- So 1 A is bijective). hcnce k ® N = o.
1
10. Let A be a ring. Also N is finitely generated, by Chapter 2, Exercise 12, hcnce N = O by
i) U A is absolutely flat (Chapter 2, Exercise 27) and S is any multiplicativeiY 1 Nakayama's lemma. Hencc 4> is an isomorphism.]
closed subset of A, then s- 1 A is absolutely flat. Deduce that every set of generators ofF has at least n elemcnts.
ii) A is absolutely flat <=> Am is a field for each maximal ideal m. 16
· L~t B be a flat A-algebra. Then the following conditions are equivalent :
11. Let A be a ring. Prove that the following are equivalent: .~) a cc = a for all ideals a of A.
i) A/~ is absolutely flat (in being the nilradical of A). .~~) Spec (B) -- Spec (A) is surjective.
ii) Every prime ideal of A is maximal. 111) For every maximal ideal m of A wc ha ve me ,¡; (1).
..
46 RlNGS AND MODULES OF FRACTIONS EXERCl SES 47
iv) If Mis any non-zcro A-module, then M n ::¡:. O. .A _. 8 be a ring homomhorphis~. Let ~ =dSp~ch(/.A) and y =
v) For ev~ry A-module J.!• the. ~a.pping x 1-+ 1 ® x of M into M . . . ¡i) Lel l · d t ¡• : y -~> X be t e mappmg assoctate w1t . Identifying
1
[For i) ~ u), use (3.16). u) ~ m) IS clear. 18
n 10Jcctive spec (B)~ 1an c' th its canonical image S - l X in X, and Spec (S -1 B)
iii) ~ iv): Let x be a non-zero element of M and let M ' = Ax s· · Spec (S (;~S;-\ B)) with ¡ts canonical image S - 1Y in show that s -1¡•:
Y,
, · tncc 13 ·
over A it is enough to show t h at Mo ::¡:. O. We ha ve M ' ~ A j a f
, j N or som .
ts fiat · es-
( == Sp;~ 1B ) _ ,. . Spcc 1 A) is the rcstriction of ¡• to s- Y, and that
1
a ~ (1), hence Mn ~ B a•. ow a s; m for sorne maximal ideal e Idea) Spec ( - ¡• -1(S -1 X).
a' s; m' ::¡:. (1 ). Hence M~ ~ O. lll, hcnee S -l y - .de al of A and let b = a e be its extension in B . Let J: A ja ~ B[b
iv) ~ v): Let M ' be the kernel of M ~ M n. Sincc Bis flat ovcr A th iii) Let a bcha~omorphism induced by f If Spec (A ja) is identified with its
O -+ M~~ M s -~> (M o)o is exact. But (Chaptcr 2, Excrcise 13 ,~itl e~equcnce be th~ ~ ·mage V( a) in X, and Spec (B[ó) with its image V(b ) in Y, show
the mapping M 8 ~ (Mn)o is injective, hence M~ = O and thcr~forc ~' : M,) canontca 1
t ¡• is thc restriction o
r¡• t vf1~)
o \u •
v) ~ i): Take M = A j a.] - O. . th~ be a prime ideal of A. Take S = A - ~ in ii) and then reduce
Bis said to be fait!tfully flat over A. 1v) Le d\Js - 1+> as in iii). Deduce that the subspace¡• - 1 (~) of Y is naturally
mo omorphic to Spec (Bp/lJBv) = Spec (k(~) ® ..t B), where k(~) is the
17. Let A f.... B ~ C be ring homomorphisms. If g o 1 is flat and g is faithfull fi bomc .
Y at, 'due ficld of thc local nng Av.
then 1 is fiat. resl fi
Spec (k(tJ) ®..t B) is called the 1ber o
r¡• over f>.
18. Let/: A ~ B be a ftat homomorphism of rings, let q be a prime ideal of B a d A be a ring and lJ a prime ideal of A . Then the canonical image of Spec (A~)
lJ = qc. Thenj• : Spec (Bq) -+ Spec (A~) is surjective. [For B p is flat over ~ ~~
1
22• ~e~pec (A) ¡ equal to the intersection of all the open neighborhoods of ~ in
5
(3.10), and Bq is a local ring of Bv, hence is ftat over Bp. H ence Bq is flat ov/}
and satisfies condition (3) of Exercise 16.] r P Spec (A).
Let A be a ring, let X = Spec (A) and let U be a basic open set in X (i.e., U = X 1
19. Let A be a rin~, M_ an A-module. The support of M is dcfined to be the sct 23
· for so me fE A: Chapter 1, Exercise 17).
Supp (M) of pnme tdeals lJ of A such that M~ ~ O. Pro ve the following results· i) If u = x,, show that the ring A ( U) = At depends only on U and not on f
i) M ~ O ~ Supp (M) ~ 0. .
ii) Let U' = X 9 be another basic open set such tha t U ' S U. Show that there
ii) V(a) = Supp (A j a). ¡5 an cquation of the form gn = uf for sorne integer n > O and sorne u E A,
iii) If O-+ M ' ~ M -+ M " ~ O is an exact sequence, then Supp (M) = 1
and use this to define a homomorphism p: A( U) -~o A( U' ) (i.e., A1 ~ A 9 ) by
Supp (M') u Supp (M"). mapping aff"' to au"'fg"'n. Show tha t p depends only on U and U'. This
iv) If M = L: M, then Supp (M ) = U Supp (M ,). homomorphism is caUed the restriction homomorphism.
v) If Mis finitely generated, then Supp (M) = V(Ann (M)) (and is therefore iii) If U = U' , then p is the identity map.
a closed subset of Spec (A)). iv) lf U 2 U' 2 u~~ a re basic open sets in X, show that the diagram
vi) If M, N are finitely generated, then Supp (M ®" N) = Supp (M) n
Supp (N). [Use Chapter 2, Exercise 3.] A(U) A(U")
~ /
vii) lf Mis finitely generated and a is an ideal of A, then Supp (M/aM ) = A( U ')
V(a + Ann (M)).
viii) If /: A ~ B is a ring homomorphism and M is a finitely generated A· (in which the arrows are restriction homomorphisms) is commutative.
module, then Supp (B ®A M) = f* - 1 (Supp (M>). v) Let x ( = tJ) be a point of X. Show that
lim A( U) ~ A.,.
lO. Let /: A -+ B be a ring homomorphism, ¡•: Spec (B) -+ Spec (A) the associated )
Uax
mapping. Show that
i) Every prime ideal of A is a contracted ideal <:> f* is surjective. The assignment of the ring A( U) to each basic open set U of X, and the
ii) Every prime ideal of B is an extended ideal ~ f* is injective. restri~tion homomorphisms p, satisfying the conditions iii) and iv) above,
ls the converse of ii) true? conshtutes a presheaf of rings on the basis of open sets (X1) 1e..t· v) says that the
stalk of this presheaf at x e X is the corresponding local ring Av.
i) Let A be a ring, S a multiplicatively closed subset of A, and cp: A ~ S~ A
1
21.
24• Show that the presheaf of Exercise 23 has the following property. Let ( U,),.r be a
the canonical homomorphism. Show that cp*: Spec (S - A) ~ Spec (A) ~
1 1
51
homeomorphism of Spec (S - A) onto its image in X = Spec (A}. Let t
1 covering of X by basic open sets. For each i E 1 let se E A( Uc) be such that, for
image be denoted by S - t X. X1 each pair of in dices i, j, the images of s, and s1 in A( U, " U1) are equal. Then
In particular, if /E A, the image of Spec (A,) in X is the basic open set tbe~e exists a unique sE A ( = A( X)) whose image in A(Uc) is s,, for all i E/.
(Chapter 1, Exercise 17). (Th•s essent·ta11 Y ·•mplies
· that the presheaf ts
· a sheaf.)
EXERCISES 49
48 RJ NGS ANO MOD ULES OF FRACTIONS
. homomorphism. Show that ¡• : Spec ( 8 ) - Spec (A) is
)>
.A B g · A _. e be ring homomorphisms nnd Jet lt : A - ¡.. n J?. C : A _.,. B be 3 nn~ ing (i.e., maps closed sets to closed sets) for the con-
25 Let f - t ' ) • "-" A be 29· [...el / . us closed maPP
• fi ·d b h( ) - f(x) ® g(x) Let X, Y. Z, T be t 1c pnmc spcctra of "' lJ continuo
de ne Y x - · • •(Z) ' • C' a tibie topology.
B ®..e e respectively. Thcn ¡,•(T) = f .y rl C · . • struc . k 1' t0 pology and the constructible topology on Spec (A) are
[Let ~ e X, and Jet k = k(lJ) be the restduc fi~d at l-J. ~Y Excrctsc 21, the fiber ShoW tha t the Zansl .r Af~Jl ¡5 absolutely flat (wherc 91 is the nilradical of A).
¡,• -x(p) is the spectrum of (B ®.;~e) ® ,.k = (B 0 ,. k) ®~e (e ® ... k) . Hcncc JO. nlc if and on Y 1
~eh•(T) ~ (B ®,.k) ®~e (C ®,.k) :1: O ~ B ® ,.k ~ O nnd C ® ;~k :;é 0 ~ the sa ]
[Use Exercisc 11 .
~ e f• ( Y) f"\ c •(Z). ]
26. Let (B11 , C~~H) be a direct system of ~ings nnd B thc dircct li mit. For cach a, lct
lc . A -+ B11 be a ring homomorphtsm such thnt Can o l a = In whencvcr a ~ fl
(J.C.
~· the B11 form a direct systcm of A-algebras). Thc la induce 1: A - >- B. Sl;oW
that
¡•(Spcc ( B)) = n /a*(S pcc (Ba)).
a
'
[Let ~ E Spec (A). Then ¡• - l(p) is the spectrum of
B ® ,.k(~) ~ Iim (Ba ® ,. k(l-!))
)
(since tensor products commute with direct li mits : Chaptcr 2, Excrcisc 20). ,
By Exercise 21 of Chapter 2 it follows that 1* - l (~) = 0 if and only if ¡
0 11 ® ... k(~) = O for some a, i.e., if and only if l a* - 1 (p) = 0 .]
1
!
27. i) Let fa : A - >- Ba be any family of A-algebras and lct / : A - >- B be thcir j
tensor product over A (Chapter 2, Exercise 23). Then
•
PRIMARY OECOMPOSITION PRIMAR y DECOMPOSITION 53
52
all distinct and (ii) we ha ve q, ~ n,~ 1 q, (1 ~ i ~ 11) .e varietics embedded in the irreducible e
over (i) the r(q,) ar.e. n (') is sa,id to be minbnal (or irrcdundant, or red: lhe of the~~· ~~f~rc (4.6) thc varicty dcfined by a is the 1ine :m!o;e:~ Thus in the
· ary decomposJtiO . . Ccd 0 exnrnP _ (x y) corrcsponds to thc origin (O O) ' the embedded
pnm By ( .3) we can achieve (1) and then w~ .can om1t any superfiu~ r
4 ¡den1 P:l - ' . ' ·
normal, ··. ).. (") · thus any primary decompos1t10n can be rcduccd Us ) lt is not true that all the pnmary componcnts are inde
. . to achJevc
terms W ub 11 , say that a is decomposable 1·r 1t
. has a pnmary. dccon to .1a 2 osition. For cxamplc (x2 , xy) = (x) n (x y) 2 = pe~dent of tbe
OllmmaJ onc. es a 1Pos •
d~c~rnf rninimal primary decompositions. Howev~r there ~) n (x , y) .are two
tion. dJsttnC . (4 1O) ' re sorne umqueness
_¿ (l st uniqueness theorem). Let o be a decomposable ideal propcrtJCS : scc . .
Theorem •·5• d .. .r and
/et a -
_ ()" q be a minimal primary ecomposzttoll o1 o. Lct \J :::::: (
1• 1 1 • . 'd ¡. ¡ . ' r q,) proposition 4.7. Let o b~ .a decomposable ideal, ler a = nr
. ~ ) ""hen the .p 1 are precrsely the prmze 1 ea .s w ueh occur in th mínima/ primary decomposttiOil, and let r(q 1) = Vt· Then
1 1
• q be a
( 1 ~ z ---= n . .l 1 • d e~ •r e set
o1.r zuea
•.1 ls r(a·· ,..v) e~· E A) • and lzence are l/1 epellCICll f O; the particular e1e. , n
composition of o. U .p,
1• 1
= {x E A : (a: x) -1-
r
a}.
Proof. For any xEA we have (a:x) = ((') q,:x) ~ (') ~q,:x), hcncc r(n: .\") :::
nr-1 IS
r(q,:x) = nx~Q¡VJ by (4.4). Suppos~ r(a_:x) pnmc ; thcn by (1.11) We 1 particular, if tlze zero
111 •
ideal
•
is decomposab/e
• '
rhe ser D 0'Jrzero• d'IVISOTS
. o1.r A
ha ve r(a :x) = p, for somej. Hence every.pnmc 1dcal of the form r(o.: x) is one of is tite 1mion of rlze pnme rdeals belongmg to o.
tho ~,. Converse!y, for each i there exJsts x, fj: Q¡, x, E n," 1 q,, smcc the de. 1 Proof _I~ 0 is ?ecomposa?Ic, thcn O i.s dc~omposablc in. ~fa: namely 0 = n c¡1
composition is minimal ; and we have r(a:x1) = .p,. • where q, JS the Jmage of <11m Af o, and ts pnmary. Hence tt.ts enough to prove the
Remarks. 1) Tbe abo ve proof, coupled with. the la~t part of (4.4), shows that Iast statement of(4.7). By(l.lS)wehaveD = Uxlio r(O:x); fromtheproofof
for each i there exists x, in A sueh that (a: x,) IS .P1-pnmary. (4.5), we ha ve r(O: x) .= rtx~q1 .p, ~ l-J, for sorne j, hence D ~ U r. 1 .\>,. But
also from (4.5) each p11s ofthe form r(O :x) for sorne x E A, hence U .p, ~ D. •
2) Considering Afa asan A-module, (4.5) is equivalent to saying that the p,
are precisely the prime ideals which occur as radicals of annihilators of clements Thus (the zero ideal being dccomposable)
D = set of zcro-divisors
of Afa.
= U of all prime ideals belonging to O;
Example. Let a = (x 2 , xy) in A = k[x, y] . Then a = l'1 n l-J~ wherc p1 = (x),
9~ = set of nilpotent elements
p = (x, y). The ideal lJ~ is primary by (4.2). So the prime ideals are p11 h
2
In this example .p 1 e .p 2 ; we have r(a) = .P 1 n P2 = P1, but o is not a primary
= n of all minimal primes belonging to o.
54 PRIMARY D ECOMPOSITION
EXERCISES
Proposition 4.9. Let S be a multiplicarively closed subset of A and 55
1
decomposable ideal. Let a = nf
•1 q, be a mínima/ primary deconzln er. ~ be a
rOSlllOt¡
eXERCISES
a. Let .p 1 = r (q 1) and suppose tlze q, numbered so that S meets .p o¡ 'deal 0 has a primary decomposition then S (
'T'I m + 1, . . P J. Jf an 1 'blc components. , pec A/a) has only fin.itely
b ut not p¡, ... , .Pm· .1 rzen ·, 11 irreduct many
z. If 0 == r(a), then a has no emb~dded prime ideals.
s - lo = n
f- 1
m
s - lq., S(a) - nm
f- 1
qt. If A is absolutcly flat, every pnmary ideal ¡5 maximal.
3
' I the polynomial ring Z[t], the ideal m == c2 t ) . .
and tlzese are minimal primary decomposirions. 4. n== (4, 1) is m-pnma. ry, b ut 1s
· not a powcr of m.· ts maxtmal and the .tdeal
Proof ~ -l a = nr~1 s - tq, by (3. 1~ ) = nr-1
s - tq,. b~ (4.8), a!1d s -1q, is 1
5
11
In the polynomial ring K[x, y , z] where K is a ficld and .
S - 1.p 1-pnmary for 1 = 1, ... , m. S mee the .p, are dtstmct, so are thc S _1 . indeterminates, Jet.\:)¡ = (x , y), .\.'2 == (x, z), m = (x Y x): ~· z are mdependent
(1 ~ ¡ ~ m), hence we have a minimal primary decomposition. Contracti:' , and m is maxtma. 1 L t h
. e a = ,..t.P::~. h , , z , ,..1 and P2 are .
S ow that 0 == h n h • Pnme,
both sides, we get g .
primary decompos1·rton o f o. Wh'tch components are ,..1 . ,..2 n m2 IS a red uced
embeddcd ? 1
ISO ated and which are
S(a) = n
m
f •l
es -lq,y = m
n q,
f• l 6. Let x
be an infinite compact Hausdorff space C(X) th .
f . X (CI , e nng of real-valued
continuous uncttons on 1apter 1, Exercise 26). Is th .
by (4.8) again. • composable in this ring ? e zero 1deal de-
A set :E of prime ideals belonging to a is said to be isolated if it satisfies the 7 Let A be a ring and let A[x ] denote the ring of polynomials 1·n 0 . d .
· · ne m etermmate
ovcr A. For each Ideal a of A, let a [x] denote the set of all pol .1.
following condition: if p' is a prime ideal belonging to a and p' s; .p for sorne with coefficients in a. ynom~a s m A[x]
1
.p E :E, then .p' E :E. i) a[x] is the extension of a to A[x].
Let :E be an isolated set of prime ideals belonging to a, and let S = A - ii) 1f pis a prime ideal in A, thcn p[x] is a prime ideal in A[x].
U*'&D .p. Then S is multiplicatively closed and, for any prime ideal .p' belonging ! iii) If q is a .\J-primary ideal in A , then q[x] is a p[x]-primary ideal in A[ l
toa, we have [Use Chapter 1, Exercise 2.] x·
iv) Ifna = nr~ 1 q, i~ .a mini~al primary de~~mp~sition in A, then a[x] =
.p' E :E => .p' (\ S = 0 ;
n,..1 q,[x] IS a mtntmal pnmary decompOSltlOn m A[x].
.p' !1 :E => .p' $ U .P (by (I.ll)) => .p' n S :1= 0 . v) lf .P is a mini mal prime ideal of a, then p[x] is a mini mal prime ideal of a[x].
.,Et
8. Let k be a fiel d. Show that in the polynornial ring k [x1, •• • , Xn ] the ideals
Hence, from (4.9), we deduce ~~ = (x1, .. . , x,) (1 ~ i ~ n) are prime and all their powers are primary.
[Use Exercise 7.]
Theorem 4.10. (2nd uniqueness theorem). Let a be a decomposable ideal, let
a = nr.l q, be a mínima[ primary decomposition of a, and let {.p,l, ... 'tJ, ..} 9. In a ringA, let D(A) denote the set of prime ideals p which satisfy the following
be an isolated set of prime ideals of a. Then q11 n · · · n q,"' is independent of condition: there exists a E A su eh that .p is minimal in the set of prime ideals
containing (O: a). Show that x E A is a zero divisor <:> x E ~ for sorne~ e D(A).
the decomposition . ,
Let S be a multiplicatively closed subset of A, and identify Spec (S - 1A)
In particular: with its image in Spec (A) (Chapter 3, Exercise 21). Show that
Corollt1ry 4.11. The isolated primary components (i.e., the primary com· D(S - 1 A) = D(A) n Spec (S - 1 A).
ponents q, corresponding to mínima/ prime ideals .p,) are uniquely determined If the zero ideal has a primary decomposition, show that D(A) is the set of
by a. associated prime ideals of O.
10
Proof of (4.10). We have q11 n · · · n q,. = S(a) where S = A - .p,l u· ·· u tJ,,., · For any prime ideal .p in a ring A let Sp(O) denote the kernel of the homo-
hence depends only on a (since the .p, depend only on a). • ~orphism A ~ A.,. Prove that '
· neral .~) S4>(0) s; .p.
Remark. On the other hand, the embedded primary components are m ~e f: .~~) r(Sp(O)) = .p -<:> .p is a mínima! prime ideal of A .
1
not uniquely determined by a. lf A is a Noetherian ring, there are ~ ~~ 1 1
.u) lf P 2 p', then S4J(O) s; Sp, (O).
infinitely many choices for each embedded component (see Chapter 8, Exerctse · 3 IV) nP.D(A) s.,(O) = o, where D(A) is defined in Exercise 9.
+I.C.A.
56 PRfMAR y DECOMPOSlTION •
EXERCISES
57
. . . al prime ideal of a ring A, show that Sv(O) (Excrcise lO) . ction start ing with O¡, and so on. At the nth sta
J t. If .p JS a m1mm 1s the nstru · 'd ge we ha ve a == Q¡ " · ..
smallcst v-primary ideal. . e0 n 0 wherc the q 1 are pnmary 1 eals, a" ¡5 maximal
r _ b the 1·ntersection of the tdeals Sv(O) as V runs through thc min· r. q ~ " == On n q 11 such that a = q 1 f\ .. ·f\ q f\ b among the ideals b con-
. J.Nt'da 1e fA Show that a is contame . dm t e m'lra d 1' ca 1 of A.
. h una¡ rainJnSOnl~~vc o,.= ( 1), the proccss stops and"a 1.5• anfid~" $ Pn. lfatany
pnme 1 ea s o · . bl p h · '. we'dcals. If not, contm
stase · ue by transfinite
'
inducr a bmte .inte rsechon· of
Suppose that the zero ideal JS decomposa c . rovc t ato = O lf and onJy if
pnmnrY 1 • • ] Jon, o servmg that each
every prime ideal of O is isolated. . tly contmns On- t·
stnc 0 11
toa, we have f e p <=> p rf: :E, and let S 1 be the set of all powers off Show that such that ú $ ll n· Show that a = q 1 n · · · n qn has the required properties.]
qt = S,(a) = (a :f") for all large 11.
Primary decomposition of modules
16. lf A is a ring in which every ideal has a primary decomposition, show that every Practically the whole of this chapter can be transposed to the context of
ring of fractions S -t A has the same property. modules over a ring A. The foJlowing exercises indicatc how this is done.
17. Let A be a ring with the following property. h ZO. Let M be a fixed A-module, N a submodule of M . The radical of N in M is
(Lt) For every ideal a -:/; (1) in A and every prime ideal .p, there exists x rf: l' suc defincd to be
that Sll(a) = (a: x), where Sv = A - .p. . ,
Then every ideal in A is an intersection of (possibly infinitely many) pnman r,.,(N) = {x e A :xqM s; N for some q > O}.
ideals. ·me Show that r"'(N) = r(N: M) = r(Ann (M/ N)). In particular, '~'t(N) is an
[Let a be an ideal # (1) in A, and let p 1 be a minimal element of the set of pn _ ideal.
ideals containing a. Then q1 = Sp 1 (a) is p 1 -primary (by Exercise 11), and q1 -
State and prove the formulas for rM analogous to ( 1.13).
(a: x) for sorne x rt V1· Sbow that a = q 1 n (a + (x)) . h that
Now let a 1 be a maximal element of the set of ideals b 2 a suc t the li. An element x E A defines an endomorphism 4>x of M, namely m ~-+ xm: ~he
ql n () = a, and choose a1 so that x e ah and therefore a1 $ .P1 · Repea element x is said to be a zero-divisor (resp. nilpotent) in M if 4>..: is not inJectwe
58 PRIMARY DECOMPOSITION
21.
A primary decomposition o[ N in M is a rcprescntation of N as an interscctlon
. Integral Dependence and
N = Q¡ il•••tl Q n
INTEGRAL DEPENDENCE
Let B be a ring, A a subring of B (so that 1 E A). An clement x of B is said to
be integral over A if x is a root of a monic polynomial with coefficients in A, that
is if x satisfies an equation of the form
(1)
59
E ANO VALVA TIONS THE GOING· U p THEOR
INTEGRAL DEPENDENC . EM 61
60 ated A-module by (2.16) and thercfore . .
. . rfi 1 A[x]-module M which is finitely generared asan ·telY
l
fi n ) •
gener x ts tntegral over A by ...)
iv) There exzsts a f01th u 111
A-module. of (S.l · ar S.5. Let A s;; B be rings and let e be the
..) F m (1) we have
Proof. i) => u · ro ') ~o;lt rtenC is integrally closed in B. 111
tegra/ closure of A
x"+r = - ( a 1X n+r - 1 + ... + a"x zn ~et x E B be integral over C. By (5.4) x is inte ral
. d t. all positive powers of x lie in the A-modul proof. g over A, hence
O
for all r ~ ; le '
1 nce by m uc wn, . d (
"_1 Hence A [x ] is genera te as an A-module) by
e \'E:
·
c. • xt proposJtton
. . s11ows tl1at .mtegral dependence is p
generated by 1, x, · · ·' x · Th~ ne d to rings of fractions: reserved on passing
l,x, . . . ,xn -1 · to qUouents an
e
ii) => iü). Take = A[x]. .
propoSition 5.6. Let A S B be rings, B integral over A ·
. _
iii) => IV). Take M - ' w
e hich is a faithful A[x]-module (smce y C = O :::.
i) JjfJ is an ideal of Banda = úc = A f\ b, then Bf b is integral over A/a.
y·! = O). JI ti m (2 4) · take .J. to be multiplication by x and ..) ¡r¡ S is a multiplicatively closed subset of A then s - 1B 1·s · 1
. ) 1·) Th's1 fo ows ro · · ~ ' 11 ' ll1 egra1 over
s - lA.
bave x M S M SI·nce Misan A[x)-module); since Mis faithful , we
=> •
- IVA (we
a - "_1 + ... + a = o for suitable a, E A. • .r i) If x EB we have, say, x" + alxn-1 +···+ a" = O, with a.EA.
have x" + a1x " ProOJ· . d r~ •
Reduce this equatwn mo . u .
Coro/huy 5.2. Le1 .x,, (1 ~ ~
i~
~
n) be elements of B, each
d ¡ integral over A. ii) Let xfs E S - 1 B(x E B, sE S). Then the equation above gives
. A[X¡, . · ·, x n ] is afinitely-generated A-mo u e.
Then the rmg
B · d
Proof. y m uc Jon t. 0011 . The casen . = 1 is part of (5. 1). Assume n > 1, let
. . fi . 1 (xfs)n + (a¡Js )(xfs )" - 1 + ... + anfs" = o
_ [ ] . then by the inducttve hypothes1s An- 1 1s a m te y generated
A - A X¡, •.. , x, ' d 1 (b h which shows that x fs is integral over s- 1 A. •
A:module. A"= An_ 1[xn] is a finitely generated An)-1-m.ofiu~ Y t e case •
·
n = 1, SJDCC · 1·ntegral
Xn IS 1
. over A n - 1 ) · Hence by (2.16 A n JS m te y generatcd
as an A-module. • THE GOING·UP THEOREM
Corolltuy 5.3. The set e of elements of B which are integral over A is a Proposition 5.7. Let A s; B be integral domains, B inregra/ over A. Then B
subring of B containing A. is a field if and only if A is a fiel d.
Proof. If x, y E C then A[x, y] is a finitely generated A-module by (5.2). Hence Proof Suppose A is a field; let y E B, y # O. Let
x ± y and xy are integral over A, by iii) of (5.1). •
y 11
+ a 1 y" - 1 + · · · + an = 0 (a, E A)
The ring e in (5.3) is called the integral c/osure of A in B. If e = A, then A
be an equation of integral dependence for y of smallest possible degree. Since B
i~ said to be integral/y closed in B. If e = B, the ring B is said to be integral
isanintegral domain we have an =fi O, hencey - 1 = -a; 1(y11 - 1 + a1y"- 2 + · · ·
over A.
. +an- 1) E B. Hence B is a field .
Remark. Let f: A ~ B be a ring homomorphism, so that B is an A-al~ebr~. Conversely, suppose B is a field; let x E A, x # O. Then x- 1 E B, hence is
Then f is said to be integral, and B is said to be an integral A-algebra, tf B Js integral over A, so that we have an equation
integral over its subringf(A). In this terminology, the above results show that
. x - m + a~x - m+ 1 + ··· + a~= O (a;EA) .
finite type + integral = finite. lt follows that x- 1 = _(a~ + a;x + . .. + a~xm -1) E A, hence A is a field. •
Coro/huy 5.4. If A 5; B 5; e are rings and if Bis integral over A, and C is
integral over B, then C is integral over A (transitivity of integral dependence). Corollary S.S. Let A s B be rings, B integral over A: let q be a ~rime ~deal
ofB and let .p = qc = q n A . Tlzen q is maximal if and only if'V zs maxmzal.
Proof. Let x e e, then we havean equation
Proof.. By (5.6), Bfq is integral over A/~, and both these rings are integral
+ b1x"- + · · · + bn domams. ~ow use (5.7). •
1
x" = O (b, E B).
The ring B' = A[blt ... , b"] is a finitely generated A-module by (5.2), and B~[x] C l''-·· A . ¡ 1 1 q' be prime
.doro ""Y 5.9. Let A 5; B be rings, B integral over ' e ':
Íl a finitely generated B'-module (since x is integral o ver B'). Hence B' [x 115 a 1
eals of B suclz that q s q' and qc = q'c = V say. Tiren q = q ·
NTEORALLY CLOSED INTEGRAL DOMAlNS G
62 INTEGRAL DEPENDENCE ANO VALVA TIONS 1 • OING-DOWN T
HEOREM 63
.r 8 ( 5 6) B is integral over Ap. Lct m be the extension of pinA., and 1 ed (see (5.0)). Thc same argu_ment shows that any un·
PfOOJ. y . , p • • TI . 1 . Cl cloS . ·s intcgrally closed. In particular a polynom· 1 . •que factorization
n n' be the cxtcnsioos of q, q' respcctJvcly m !lP·
1en m JS t 1c maximal idea) doJlllllO J 1 d
d is intcgrally e ose .
, Ja nng k(x
1• • • • , Xn1 over
' A . e , and 11o = n'c = m. By (5.8) 1t follows that n, n' are maximal
of p, n - tt, . , , a fiel ral closurc is a local property ·
hence n = n', beoce by (3. 11)(Jv) q = q · • Integ ·
Tlu!orem 5.10. Let A s; B be rings, B integral over A, and let p be a prime proposition 5.13. Let A be an integral domain. Then the follow ·
ideal of A. Then there exists a prime ideal q of B such that q f1 A = p. equivalen/: mg are
Proof By (5.6), Bp is integral over A~, and thc diagram i) A is integral/y closed;
A--7-B ii) Av is integrally closed, for eaeh prime ideallJ;
at f p iii) An, is integral/y closed, for eaeh maxima/ ideal m.
A., --7- Bp
Let K be the field of fra ctions of A, Jet C be thc intcg l l
Prooif.· 'd . . ra e osure of A · K
(in which the horizontal arrows are injection~) is ~ommutative. Le~ n be a m_axi- d Jet f: A --7- e be the 1 CDtJty mappmg of A into c. Th . . m 1
mal ideal of Bp; then m = . n ti A., IS max1m;(t )byh(5.8), . hen~e JS the untquc an . . . d b (5 12) en A rs mtegrally
closcd <=> f lS s~rJe.cttve, an y . A P (resp. A m) is integraUy closed <>
maximal ideal of the local ringA.,. If q = {3- n , t en q 1s pr11ne and we ha ve (resp.J;n) is surjecttve. Now use (3.9). • f~
q ti A = a- 1 (m) = p. •
Let A s; B be rings and let a be an ideal of A . An element of B · 'd
Theorem 5.11. ("Going-up theorem"). Let A s; B be rings, B integral ·f · · fi · . . JS sa1 to be
integral over a 1 1t satiS es an equat10n of Integral dependence over A . h" h
over A,· /el p1 s; .. · s; Pn be a chain of prime ideals of A and q s; · · · s qm . ¡· . Th . 1 1 m w •e
all the coefficten_ts te m a. e mtegra e osure of a in Bis the set of all elements
(m < n) a chain of prime ideals of B suclz tlzat q, fl A = p, (1 ~ i ~ m). of B which are mtegral over a.
Tiren the clrain q 1 s; · · · s; CJm can be extended lo a chain ql e · · · s; Cln such
tlrat q1 t1 A = p1 for 1 ~ i ~ n. Lemma 5.14. Let C be the integral c/osure of A in B and let ae denote the
Proof By induction we reduce immedi~ tely to the case_m = l , n = 2. Let extension of a in C. Tlzen tlze integral c/osure of a in B is the radical of ae
A = A/th, Jj = Bjq 1 ; then A s; 13, and Bis integral ove_:_ A by (5.6). Hence, by (and is t/zerefore c/osed under addition and multiplication).
(5.10), there exists a prime ideal q2 of B such that q2 f1 A = ~2, the imagc of P2 Proof If x E Bis integral over a, we havean equation of the form
in A. Lift back q2 to B and we have a prime ideal q2 with the required pro-
perties. •
with ah ... , an in a. Hence X E e and x" E ae, that is X E r(ae). Conversely, if
INTEGRALLY CLOSED INTEGRAL DOMAINS. x e r(a') then x" = .2 a1x 1 for sornen > O, wbere tbe a1 are elements of a and the
TIIE GOING-DOWN THEOREM x, are elements of C. Since each x1 is integral over A it follows from (5.2) that
Proposition (5.6)(ii) can be sharpened: M= A[x¡, ... , Xn ] is a finitely generated A-module, and we have xnM S aM.
Hence by (2.4) (taking 4> there to be multiplication by x") we see that xn is
Proposition 5.12. Let A s; B be rings, C the integral closure of A in B. Let S
integral over a, hence x is integral over a. •
be a multiplicalively closed subset of A . Then S -le ;s the integral closure of •
S-lA in S - 1B. Proposition 5.15. Let A s; B be integral domains, A integral/y closed, and
Proof By (5.6), S - le is integral o ver S - lA. Conversely, if b/s E S -l B is let x E B be integral over an ideal a of A. Then x is algebraic over the field of
1
integral over S - 1 A, then we havean equation of the form fractions K of A, and if its minimal polynomial over K is t" + a1t" - + · · ·
(bfs )" + (at /S¡)(bjs )" - 1
+ · ·· + an/Sn = 0
+ an, then al, ... 'an /ie in r(a). •
y is integral ove~ >¡J2 , and hence by 5.1 tts mm1mo equat10n ovcr K, thc fie}d of • _ 1 == a· (ax) -1 e B. Next let x, y be non-zcro elements of m
8 .and thercfore
fractions of A, JS of the form X _ 1 e B. If xy - 1 e B then x +y== (1 + xy - t) · Then etther xy - te B
or x y . 'd l d t hercfore B ¡ a loY E Bm e m a d . .
) 1' + u y' - 1
+ ... + u = o _ 1 E B Hcnce m JS an 1 ea an 5 1 . - n sunllarly 1
say,
s' + v1 s'- 1 + ··· + v, =O (2) witb the b, E B. If x E B there is nothing to prove. If not1 then x- 1
x == -(b 1 + b2 x - 1 + · · · + b,.x 1 -") E 8 . E Bl hence
where v1 = u1/x'. Consequently
Let K be a field, n an algebraically closed field Let ¿ be h
1
x v1 = u1 E p 2 (1 ~ i ~ r ). (3) . b . fK df . . t e set of all pairs
(A.f), where A ts a su nng o an ts a homomorphism of A into n \ .
But s is integral over A, hence by (5.15) (with a - (1)) each v, is in A. order the set :E as follows: · "''e parttally
Suppose x rt p2 • Then (3) shows that each v, E P2, hence (2) shows that s' E Bp2 ( A, f) ~ ( A' ,/') <=> A s; A' and /' lA = f.
s; Bp 1 s; q11 and therefore s E q11 which is a contradiction. Hence x E p2 and
therefore Bq1 p2 n A = l3 2 as required. • The condition.s of Zorn's lemma are clearly satisfied and therefore the set E has at
Jeast one max1mal element.
The proof of the next proposition assumes sorne standard facts from field
theory. . Let (B, g) be a maximal
. element of
. L . We want to prove that B ·15· a val ua t'ton
nng of K. The first step m the proof 1s
Proposition 5.17. Let A be an integral/y elosee/ domain, K its field offractions,
L afinite separable algebraic extension of K, B the integral closure of A in L. Lemma 5.19. Bis a local ring and m == Ker (g) is its maximal ideal.
Then there exists a basis v1 , • •• , Vn ofL over K sueh that B ~ 21 ... 1 Av1• Proof. Since g(B) is a subring of a field and therefore an integral domain the ideal
~~ = Ker (g) is prime. We can extend g toa homomorphism g: Bm n by putting -)o
Proof. If vis any element of L, then vis algebraic over K and thereforc satisfies g(b/s) = g(b)fg(s ) for all b E B and all s E B - m, since g(s) will not be zero. Since
an equation of the form the pair (B, g) is maximal it follows that B = Bm, hence B is a local ring and nt is its
a0 v' + a 1 v' - 1 + · · · + an = O (a1 E A). maximal ideal. •
Multiplying this equation by a~- 1 , we see that a0 v = u is integral over A, and Lemma 5.20. Let x be a non-zero element of K . Let B[x] be the subring of K
hence is in B. Thus, given any basis of L over K we may multiply the basis ele· generated by x over B, and let m[x] be tire extension of m in B[x]. Then either
ments by suitable elements of A to get a basis u11 • •• , un such that each u, E B. m[x] ~ B[x]orm[x - 1 ] -:1= B[x- 1 ] .
Let T denote trace (from L to K). Since L/ K is separable, the bilinear form Proof: Suppose that m[x] = B[x] and m[x - 1 ] = B[x- 1 ] . Then we shall ha ve
(x, y) H- T(xy) on L (considered as a vector space over K) is non-degenerate, equations
and hence we ha ve a dual basis v1 , • •. , vn of L o ver K, defined by T (u,v¡) = 811. Uo + U1X + · · · + UmX" = 1 (u, E m) (l)
Let x E B, say x = ¿, x1v1 (x1 E K). We have xu1 E B (since u1 E B) and therefore . Vo + v1x- 1 +···+ Vnx-" = 1 (v1 E m) (2)
T(xu¡) e A by (5.15) (for the trace of an element is a multiple of one of the co· tn Whicb We · ¡ S
"' ~ n may assume that the degrees m, n are as small as posstb e. uppose that
efficients in the minimal polynomial). But T(xu,) = ¿1 T(x1u1v1) = L-1 X¡T(u,vJ)
, and multiply (2) through by xn:
• ¿, x 1 8,1 = X¡, hence x, E A. Consequently B s ¿1 Av1. •
( 1 - Vo)x n = v1 x" - 1 + · · · + Vn• (J)
66 JNTEGRAL DEPENDENCE AND VALUATIONS
EXERCISE.s
Since v0 E m, it follows from (5. J9) that 1 - vo is a unit in B, and (3) may h ' and let f: A - n be su eh that /(u) ~ O Th 67
be written in thc form t ercrore - ooao.
~,.Ct" - morphism 1:
¡; A [11 - 1] - )- 0 ( Wtthf1(u-
. 1) == ·/( ) en¡
-t)
ca be
n extended fi .
+ •• • + Wn (w1 E tll). tO11 bOlll~r hislll lt : e - 0, whcrc e is a valuation rinu , a.n~ thcn by (S 2llrst
A
11 horno~nt:gral over [u -1 ], hencc by (5.22) x E e sog t~ontammg A[u -11.. F:to
lat e contains D ando~
Hcnce wc can replacc x"' in (1 ) by WtX"' - + · · · + w"x"' - \ and this contract· 1
t (1). X ls E e On thc othcr han d, from (2) V- 1 is ¡'nt
the minimaHty of Lhe exponcnt m. • lcts ticulnr v . . . C Th r . ' cgra over A [ - 1 • In
pnr (5.22) again JS JO • ere¡ ore v ts a unit in e, and he u ], and thcre-
Tlreorem 5.11. Let (B, g) be a maximal element o[ :E. Titen B ¡ forc bY 1 rcstriction of lz to B . • ncc h(v) -:¡; O. Now tak
s a Valuation K to be t 1e e
ring of tite field K .
l' arl} 5 24. Let k be a fteld ami B a /inite/y "'e
Proof. We havc to show that if x ::/; O is an e1em~nt of K, th~n cithcr x e B '
Coro ,, " • . 1ge b ratc
. o nerateci k-a/o b
en il is a fimte a extension of k (>e ra. Jf Bis a
x - 1 En. By (5.20) we mayas well assume that m[x ] ts not the untt ideal of ti .or jiek1t1t ·
. contame . d m. a maxtma . 1 1' dea 1 m of B' , and wc1C hnng. .r Takc A := k, V == 1 and n = algcbraic closurc of k
B' = B[x]. Then m[x ] JS 1
PrOOJ· • • •
m ' n n = m (because m n B· ts
1
. a prope: 1'd ea 1 o f B an d contams
· m). Hcncc <he thc . nc form of Hilbcrt's Nullstcllensatz. For anoth
(5.24) tS o cr proof, sce (7 .9).
embedding of B in B ' induces an embeddmg of the fie1d k = B/m in thc ficld k' _
B'jm'; al so k ' = k [x ] wherc x is the image of x in k ' , hence ..vis algebraico ver k -
therefore k' is a fimte . a1geb ratc
. extenston
. of k . • and
EXERCISES
Now thc homomor~hism g_ induces ?n embedding_if of k in n, sincc by (S.! )
9
m is thc kernel of g . Smce n JS algebratcally closed, g can be extended to an e . l. Let/: A -:->- B be an integ~al h~momo~phism of rings. Show that¡t : S ec
bedding ¡¡' of k' into n. Composi ng g' with the natural homomorphism B ' ~ k' : Spcc.(A) 1s a closed mappmg, J.c. that tt maps closcd scts 10 closed s P (~).-
ha ve, say, g ' : B ' - >- n which extends g. Sin ce thc pair (B, g) is maximal, it foll~w: 'geometrical equivalcnt of (5. 10).) ets. (Thts ts a
that B' = B and therefore x E B . •
2. Let A be a subri ng
.
of a riog B such that B is integral over A d
f A .
•
, an et f A - n
1
Corol/ary 5.22. Let A be a subring of a fie/d K . Then the integral c/osure ;¡ 0¡ A be a homomorphtsm o mto an algcbraically closcd ficld n Sh h.
. . · ow t at f can
in K is tite intersection of al/tite valuation rings of K which contain A . be extended to a homomorp 1sm of B mto n. [Use (5.10).]
h
Proof Let B be a valuation riog of K such tha t A s; B . Since Bis integrally closed, J. Let f: B - > B ' be a homomorphism of A-algebras, and let e be an A-algcbra
by (5.18) iii), it follows that As; B. . Ufis integral, prove thatf ® 1: B ® _. C-. B' ® ..t e is integral (Tb' · l d .
•
Conversely, let x rt A. Then x is not in the ringA ' = A[x- 1 ] . Hence x-1 is a . · Js me u es
(5.6) ii) as a specJal case.)
non-unit in A' and is therefore contained in a maximal ideal m ' of A'. Let n be an
4. Lct A be a subring of a ring B such that Bis integral over A. Let n be a maximal
algebraic closure of the field k' = A'/m'. Then the restrictioo to A of the natural
ideal of B and let m = n n A be the corrcspond ing maximal ideal of A. Is B
homomorphism A' -- k' defines a homomorphism of A into n. By (5.21 ) this can necessarily integral over A m? n
be extended to sorne valuation ring B 2 A . Since x - 1 maps to zero, it follows that
[Consider the subring k[x 2 - 1] of k [x ], where k is a field, and let n = (x _ 1).
x;B. • Can the clement 1/(x + 1) be integral ?]
Proposltlon 5.23. Let A s; B be integral domains, B finitely generaled over A.
5. Lct A s B be rings, B integral over A .
Let v be a non-zero e/ement of B . Tiren there exists u ::f: O in A wilh the fol/owing
i) If x E A is a unit in B then it is a unit io A .
property: tmy lrom omorphism fofA in lo an algebraically c/osed fie/d .O suclz that
ii) The Jacobson radical of A is the contraction of the Jacobson radical of B.
f (u) ::f: O can be extended to a lwmomorphism g of B inlo n suclz that g(v) ::! O.
Proof By induction on the number of generators of B over A we reduce immediately 6. Let B¡, ... , Bn be in tegral A-algebras. Show that n~. 1 B1 is an integral A-
algebra.
to the case where Bis generated over A by a single element x .
i) Suppose x is transcendental ovcr A, i.e., that no non-zero polyoomial with 7. Let A be a subring of a ring B, such that the set B - A is closed under multi-
coefficients in A has x as a roo t. Let v = a0 x" + a 1 x" - l + · · · + an, and take plication. Show that A is integrally closed in B.
e
u = U o. Then if ! =A -- n is sueh that f (u) ::f: o, there exists E n su eh that /(~o)f' 8· i) Let A be a subring of an integral domain B, and let C be the integral closure -
+ /(a1){" - 1 + · · · + f(an) ::f: O, because íl is infinite. Define g; B --+ n extendmg/ of A in B. Let /, g be monic polynomials in B[x ] such that fg E C[x). Thcn
by putting g(x) = { . Then g(v) ::f: O, as required. Th l. g are in C [x] . [Take a field containing B in which thc polynomials l. g
jj) Now suppose x is algebraic o ver A (i.e. o ver the field of fractions of A). en split into linear factors: say ¡ = rr (x - ~ 1 ), g = II (x - rJJ). Ea~h l1and
sois v- 1 , because vis a polynomial in x. Hence we ha ve equations of the form eacb "11 is a root of fg, hence is integral over C. Hcnce the cocffic•ents of 1
(1)
+ a 1 x"' - l + · · · + am = O
a 0 x"' (a, E A ) .. and g are integral over C.] .
aóv-n + aív 1 - " + · · · + dn = O (ajE A).
(2) u) Prove the s~me result without assuming that B (or A) is an integral domam.
•
•
68 INTEGRAL DEPENDENCE ANO VALUATIONS
9. Let A be a .su brin~ of a ri ng B and Jet C be. thc integral closure or . EXERCISES 69
that C[x] ts thc mtegral closure of A[x] m B[x ] . [Tf / E B[x ] isA· 10 B. Prov~ :3 e p and let X E ~1 · Then n o o(x) E l>l n A G - e::
A[x], then tntcgra) over (Let P1V u E G. Deduce that ~ 1 is contained in U- ~ - P2, hence o(x)ep
{ot somed (5.9). ] aao o(p2), and then ap 1:~
(g¡ E A[x]). (t.tn nn . PY
A be an intcgrally closed domam, K its field of f .
Let r be an integer Iarger than m and the dcgrecs of g 1 J4. ¡,et 1 separable extension of K . Let G be the Galois racttons and L a finite
f - x", so that ' · · · ' g ,., and let ¡1 "' orma fA . L group of L ov K
n h integral closure o m . Show that o(B) = 8 f er and let
(ft + xr)m + l!t(f + xr)'"- 1 + · · · + Cm = O B be t oe or aH o e G• and that
or say A ::::: B .
/;" + lrtflm - 1
+ ' •· + hm = O, t A, K be as in Exercise ~ 4, let L be any finite extension field of
tS. Le 'ntegral closurc of A m L . Show that if ~ ¡5 any P . 'd K, and let B
be t 1te 1 f h' h ' nme 1 eal of A th
where lrm = (xr)m + C1Cxr)m- t + · · · + Cm E A [x ]. Now apply Exercisc of prime ideals q o B :' te contract to p ¡5 finitc (in oth • · en the
8 10 sct (B) -+ Spec (A) has fimte fibers). cr words, that
polynomials - / 1 and Jr - l + htfr - 2 + · · · + Ir m _ 1 . ] the
Spec
R duce to the two cases a ( ) L separable ovcr K and (b) L pu 1 .
10. A ring homomorphism !=A -~ Bis said to ha ve thc going-up property (
·
gomg-down properly) 1·r t he conc 1usJon
· o f t 11c gomg-up
· rcsp. the
theorcm (5.11) (res
[ e K In case (a), embed L in a finite normal separable cxtcnsi~c Y fmKseparable
over . d 14 1 (b) ·r . . . no ,anduse
l
going-down theorem (5. 16)) holds for B and its subringf(A). p. the.
Exercises 13 an . n case ' J q ts a pnme Ideal of B such that q n A - . {
"that q is the set of all x E B such that x P"' e .p for sorne m >- 0 h .- P. .
Let/* = Spec (B ) -~ Spec (A) be the mapping associatcd wi th f slto\-. . ,. . . • w ere p ts the '
characteristtc of K, and hence that Spec ( B) - >- Spec (A) is bijective in this case.] ,
i) Consider the following three statements :
(a) ¡• is a closed mapping.
(b) f has the going-up property.
Noether's normaliza/ion lemma -
16• Let k
be a field and let A '#• O be a fini tely. generated
.
k-algebra · Then there eXJst
.
(e) Let q be any prime ideal of B and let ~ = qc. Then ¡• : Spec (Bfq) _ elements y¡, . .. , y, E A whtch are algebraJcally mdependent over k and such that
Spec (A/~) is surjective. A is integral over k [y¡, ... , y.,] .
Prove that (a) => (b) ~ (e). (See also Chapter 6. Excrcise 11 .) We shall assume that k is infinite. (The result is still true if k is finite, but a
ii) Consider the following three statements : different proof is needed.) Let X¡, ... , Xn generate A as a k-algebra. \Ve can
(a') ¡• is an open mapping. renumber the x, so that x1o . .. , x , are algebraically indcpendent over k and each
(b') f has the going-down .property. of Xr+ 1 , ••• , Xn is algebraic over k[x1, ... • x .,]. Now proceed by induction on n.
(e') For any prime ideal q of B, if ~ = qc, then ¡• : Spec (Bq) _ , Spec (A¡¡) is If 11 = r there is nothing to do, so suppose n > r and the result true for 11 - 1
surjective. generators. The generator Xn is algebraic over k[x¡, ... , X 11 - 1], hence there
Prove that (a') => (b') ~(e'). (See also Chapter 7, Exercise 23.) exists a polynomial f '# O in n variables sueh that /(X¡, . .. , X 11 - 1, X11) = O. Let
[To prove that (a') = (e'), observe that Bq is the direct Iimit of the rings B, F be the homogeneous part of highest degree in f. Since k is infinite, therc exist
where t E B - q; hence, by Chapter 3, Exercise 26, we ha ve /*(Spec (Bq)) = .\1, •• • , An - l E k such that F(A 11 ••• , An - 1, 1) ':/; O. Put x~ = x1 - A¡X11
(J,f*(Spec (B,)) = (1 1 /*( Y,), Since yt is an open neighborhood of q in Y, and (1 ~ i ~ n - 1). Show that Xn is integral over the ringA ' = k[xí, ... , x~ - tl.
since ¡• is open, it follows that f*( Y,) is an o pen neighborhood of .P in X and and hence that A is integral over A' . Then apply the inductivc hypothesis toA'
thercfore contains Spec (Ap).] to complete the proof.
f has the going-down From the proof it follows that y 11 ••• , y, may be chosen to be linear com-
ll. Let f: A - >- B be a flat homomorphism of rings. Then
binations of x1o ... , x 11 • This has the following geometrical interpretation: if k is
property. [Chapter 3, Exercise 18.]
algebraically closed and X is an affine algebraic varicty in kn with coordinate
0
12. Let G be a finite group of automorphisms of a ring A, and let A denote the ringA # O, then there exists a linear subspace L of dimension r in kn anda linear
subring of G-invariants, that is of all x E A such that o(x) = x for all u E G. mapping of k" onto L which maps X onto L. [Use Exercise 2.]
Prove that A is integral over A 0 • [lf x E A, observe that x is a root of the poly- •
Deduce that every maximal ideal in thc ring k [l t, . .. , t,.] ¡5 of thc fo ~ be a ring. Show that thc following are . 71
_ alt • • . , 1n - a11 ) wherc a1 E k. Lct /1 'd 1 · A · · equ¡valcnt ·
( 11 rlll zJ. . EvcrY prime 1 ca JO • 1~ an lntcrscction of max· ·.
18. Let k be a field and let 1! be a bfini.tcly gcn:rntedf kk-al(gT ebl:a·. Supposc that lJ
.u)~> 1o every homomorph1c Jmagc of A the nilr d' un~l Jdcals.
a tea\ lS cq 1
· ficld . Thcn B is a fimte algc ra1c cxtcns10n
JS a . of . ~. 11s 1s anothc r vcrs¡ radical. . . . ua to the Jacobson
of HilbcrCs NuJlstellcnsatz. The followmg proo 1s duc to Zariski. For Othon ... Every prime 1deal JO A wh1ch is not maximat .
proofs, sce (5.24), (7.9).) . cr
JJI) of the prime idca~s ~~ich ~.ontain it strictly. . IS equal to the intersection
only hard part JS m ) => u). Supposc ii) fal h
Let X Jo . . . , X " geocrate B as a k-algebra. Thc proof 1s by induct 1·0 n on ['fJlC • • f . se, t en the . .
lf n = 1 thc result is clearly true, so nssume n > l . Lct A == k [x¡) and ¡"· wu } 'cll is not an mtersechon o max¡mal idcals p .
ay nssume that A
.
1s an
.
mtegral domain
· assmg to th
whos
re ts a Pnmc id 1
e
.
quottent
ea
ring
K - k(x ) be the field of fractions of A . By thc inductivc hypothcsis n . Cl wc rn e 1acobson d' •
-
finite 1
algebrtuc • •
extcnsJon of K,, hence cae h o f X2 , •.• , x•,. satis • IS a
. fi1es a monic Lctfbe a non-zero e cment of m. Thcn A -J. O
1 ra tea\ mis not
. m . n., Poly. zero. . . A . . . 1 .,... , hcnce A h . .
nomial equation with coe m 1C1cnts v .
J.c. coc ffi.c1cn
. t f 1 f
s o t 1c orm afb whcrc a 'd aJ whosc contractJOn m.
1 e ,
Js a pnme 1dcal ,.,h sueh t11at ¡ ~ 1h as a maximal ·
and b are in A. If [is t~lc. product of thc denommators of al! thcsc cotfficicnts, ...... ¡mal with respcct to thJs property. Thcn h is not . ,.,, and which is
mw- . f h . . ,., maximal and .
then each of x:h . •. , x ,. 1s mtcgral over A1 • Hence B and thereforc K ¡5 integral to the intersectl~n o t e pnme Jd~als strictly containing V ] 1s not equal
A ring A wtth the three equ¡ valent propcrties ab · .
over A 1• •
ove JS called a J acobson
Suppose x 1 is transcendental over k . Then A is intcgrally closcd, beca use it . TlfiK·
a unique factorization domain. Hcncc A 1 is integrally closcd (5. 12), and thcr~~ .. .4 Let A be a Jacobson ring (Exercise 23) and B an A-alg b
rore A, = K, wbich is clearly absurd. Hence x1 is algcbraic over k , hcncc K Jlf· •
either (i) mtegra 1 over A. or ( 11" ) fi mtely
. e ra.
generated as an Show that ·r
A-al b J B .ts
(and thcrefore B) is a finito extension of k. Jacobson. [Use Exercise 22 for (ii).] ge ra, then B is
19. Deduce the result of Exercise 17 from Exercise 18. In particular, cvery finitely gcnerated ring and eve fi .
a]gebra over a field, 1.s a Jacobson nng. . ' ry mte1Y generae t d
20. Let A be a subring of an integral domain B such th at Bis finitcly gcncratcd over
A. Show that therc exists s :f; O in A and clements Y1, .. . , y,. in B, algebraically 15. Let A be a ring. Show that the following are equivalent:
independent over A and such that B, is integral ovcr B;, whcrc B' = i) A is a Jacobson ring ;
A[y11 • • • , Ynl· [Let S = A - {O} and let K = S - lA, the field of fractions of A. ii) Every finitely generated A-algcbra B which is a field is finite over A.
Then S - t Bis a finitcly gencrated K-algcbra and thereforc by thc normalization [i) => ii). Reduce to the case wherc A is a subring of B and use E · 1.
. . . • xerc1se 2
Jemma (Exercise 16) there exist x 11 • •• , Xn in S - lB, algebraicaUy indepcndent rr.s: A ts as m Exerctse 21 , th.cn there exists a maximal ideal m of A not con-
over K and such tbat S - l B is integral over K[x1, ... , x ,. ]. Lct z1 , . • • , zn tamJDg s, and the homomorphtsm A - >- A/m = k extends to a homomorph'
genera te B as an A-algebra. Then each z1 (regarded as an elcmcnt of S - lB) is . h J b . ISffi
g of B mto t e a ge raJe closure of k. Sincc B is a field, g is injective, and g{B)
integral over K[x 1 , ••• , Xn]. By writing an equation of integral dependence is algebraic over k, hence finite algebraic over k.
for each z¡, show that there cxists s E S such that x, = y tfs (l ~ i ~ 11) with ii) => i). Use criterion iii) of Exercise 23. Let ~ be a prime ideal of A whicb ¡5
y, E B, and such that each sz1 is integral over B'. D educe th at this s satisfies the no! maximaJ, and let B = Af-p. Let f be a non-zero clement of B. Tben B1 is a
conditions stated.] fimtely generated A-algebra. If it is a field it is fi.nite over B, hence integral
21. Let A, B be as in Exercise 20. Show that therc exists s :¡; O in A such that, over B and therefore Bis a field by (5.7). Hcnce B1 is nota field and therefore
if O is an algebraically closed field and / : A __, n is a homomorphism for which has a non-zero prime idea], whose contraction in B is a non-zero ideal \)' such
tbatf~ lJ'.]
f(s) :f; O, then f can be extended to a ho momorph ism B -- n. [With the
notation of Exercíse 20, f can be extended first of all to B' , for examplc by 26. Let X be a topological space. A subset of X is local/y closed if it is the inter-
mapping each y, to O; then to n; (because f(s ) '# O), and finally to B, (by section of an open set and a closed set, or equivalently if it is open in its closure.
Exercise 2, because B, is integral over B;).] The following conditions on a subset Xo of X are equivalent:
22. Let A, B be as in Exercise 20. If the Jacobson radical of A is zero, then sois thc {1) Every non-empty locally closed subset of X meets Xo;
Jacobson radical of B. (2) For every closed set E in X we ha ve En Xo = E;
[Ley v :¡;. O be an element of B. Wc havc to show that there is a maximal ideal (J) The mapping U t-t U fl Xo of the collection of open sets of X onto the col-
of B which does not contain v. By applying Exercise 21 to the ring Bv and its lection of open sets of X 0 is bijective.
subring A, we obtain an element s ::f:. O in A. Let m be a maximal ideal of A A subset Xo satisfying these conditions is said to be very dense in X.
such that s; m, and let k = A/m. Then the canonical mapping A - >- k extends . lf A is a ringr show that the following are equivalent:
0
toa homomorphism g of Bu into an algebraic closure n of k . Show that c(v) -f 1
• .> A is a Jacobson ring;
and that Ker (g) n 8 is a maximal ideal of B.] u) The set of maximal. ideals of A is very dense in Spec (A);
72 INTEGRAL DEPENDENCE ANO VALVATIONS
EX!!RCISts
.. ") Evcry Jocally closcd subset of Spcc (A) consisting of a single point is 1 '' [l""'] be the group ulgebra of L' over k . By d •fi . . 73
'r'·~) d ···)are gcometrical formulations of conditions ii) a nd iii) of Exc ~ osed. A :::: 1\ r space by e 1emcn ts Xa (ce e r) such the lnltton• A ts . frecly
u an m rctsc 2).] k-vcc to . at x ..xa _ &enerated a
a ·ntcgrnl domtun. - xH a. Show th s
Va/uation rings and ualuations is an a
Jf 11
\
::: "tXcrl
+ . . . + ..\ x is a
n "n ny non-zcro elcm
at A
27• Let A, B be two local rings. Bis ~aid t~ dominate. A if.A is a subring of .8 a
0 and al < .. . < « n, define Vo(u) to be ent of A, Wherc lh >.
all ~ _ {O} _ ,.. r satisfies conditions ( 1) and (2) ~1 E Sh?w that the ~a i ~re
the maxima1 ideal m of A is contamed m the maxJmal 1deal n of B (or, cquiv:d
let ~ be the set of all local sub . · va: A t K be thc ficld of fractions of A. Show th
0 xerctse 31. PPtng
1en tly, if nt = n fl A). Let K be a. field f and
d . . 1 h nngs Le tuation u o f K ' an d tha t the valuc group ofat Vo can be ·
of X. If 1: is ordered by ~he rel~tton ·or odmtn at J?fnA , s .10w t at I:. has maxirna¡ a . . unlquely extend d
lO a V • • • V IS precasely r . e
elements and that A E ~ ts maxtmal 1 a n on1Y 1 tS a valua tJ on ri ng or k.
A be a valuataon nng a nd K Jts field of fracr
(Use (5.21).] 34. Let morphism such that ¡• : Spec ( B ) - ,. Spcc (Aa)o.ns. Let / : A - B be a ri
J¡omo b h 1s a closed . ng
l8. Let A be an integral domain, K its field of fractions. Show that thc following are _.. K ¡5 any A-a 1gc ra omomorphism c·1 e 1·r 0 . mapptng. Then 1·r
g: .B K ) we have g(B ) = A . . ' ' g f IS the ernbeddang . of
equivalent: 10
(1) A is a valuation ring of K ; . A
[Let e == g(B); o b VIOUS. 1y C 2 A . Let 11 be a maximal .d
1
(2) If a, b are any two ideals o~ A, t~en e1ther. a s; ~ or .ó s; o. ed m = n fl A is the maximal ideal of A when A cal of C. Since ¡- ¡5
Deduce tbat if A is a valuat1on rmg a nd ~ .IS a pnme tdeal of A , thcn A P and e1os , • ce m = A Als
·ng Cn dominntes A m. Hence by Exercise 27 we hav C · o the local
A/P are valuation rings of their fields o f f ract10ns. rl A ]
e ~ . e n = A and therefore
29 • Let A be a valuation ring of a field K . Show that every subring of K which From Exercises 1 and 3 it follows that, if /: A ~ B is inte .
35
contains A is a local ring of A . · A-nlgebra, then the mapping (/ ~ 1)* : Spec (B 0 C) _,. S gral(Can)~ C ts any
,. pec lS a closed
30. Let A be a valuation ring of a field K. The group U of units of A is a subgroup map.
of tbe multiplicative group K* of K. Conversely, suppose that / : A - >- B has this property d h . .
. Th f . . l . an t at B IS an tn-
Let r = K*/ U. If ~. 7] E r are represen ted by x, y E K, define ~ ~ r¡ to h
tegral domam. A en B lS ~~~~ra [~~pla.cmg A by its image in B, reduce to
mean xy - 1 E A. Show that this defines a tota l orderi ng on r which is compatible the case w~e~e ~ t .an . JS tf ~ lnJcct~o~. Let K be the field of fractions of
e
with the group structure (i.e., ~ 7] => ~w ~ 7]W for all w E r) . In other words, B and Jet A ~a va ua 10 0 rmg o . contammg A. By (5.22) it is enough to show
r is a totally ordered abelian group. It is called the ualue group of A . that A' contams B. By hypoth:s1s Spec (B 0 ... A ' ) -+ Spec (A') is a closed
Let u: K* _. r be the canonical homomorphism. Show that v(x + y) ~ map. Apply the result of Exerc1se 34 to the homomorprusm B ® A' _ K
mio (u(x), v( y)) for all x, y E K* . defined by b ~ a' ._. ba'. It follows that ba' E A' for all b E B anda~ a' e ~'·
taking a' = 1, we have wha t we want.] '
31. Conversely, let r be a totally ordered abelian group (written additively), and Jet . Show that t.h: result. just. proved re~ains .valid if B is a ring with only
K be a field. A valuation of K witlr values in r is a mapping u: K* - ). r such that firutely many mm1mal pnme 1deals (e.g., 1f B 1s Noetherian). [Let \Jt be the
(1) v(xy) = v(x) + u(y), minimal prime ideals. Then each composite homomorphism A -+ B -+ Bjb1
(2) u(x + y) ~ mio (v(x), v( y)), is integral, hence A - ). TI (B/p 1) is integral, hence A _,.. BjSJl is integral (where
for aU x, y E K*. Show that the set of elements x E K* su eh that v(x ) ~ O is a !Jl is the nilradical of B), hence finally A - B is integral.]
valuation ring of K. This ring is called the ualuation ring of v, a nd the subgroup
v(K*) of r is the value group of v.
Thus the concepts of valuation ring and valuation are essentially equivalent.
32. Let r r
be a totally ordered abelian group. A subgroup A of is isolated in r if,
whenever O ~ f3 ~ a and a E A, we ha ve f3 E A. Let A be a valuation ring of a
field K, with value group r (Exercise 31). If lJ is a prime ideal of A, show that
v(A - .p) is the set of elements ~ O in an isola ted subgroup A of r , a nd that the •
mapping so defined of Spec (A) into the set of isolated subgroups of r is bi·
jective.
lf .P is a prime ideal of A, what are the value groups of the valuation rings
A/.p, Ap 1
JJ. Let r be a totally ordered abelian gro u p. We shaU show how to coostruct a .field
K and a valuation v of K with r as value group. Let k be any field and let
•
CHAIN CON
Dll10Ns 75
" for cach ll ~ O, and Go e Gl e ... e G
6 orde~does not satisfy the a.c.c. On the other hand ~· .. (strict inclusion )
thnt the G so that G does satisfy d e e e only proper subg s so
0 f 0 are "' · · · roups
) The group H of all rational numbers of th r
4 . d' . F e IOrm mfp" (
. fi s neithcr cham con ttJOn . or we ha vean exa t m, n E Z, n > O)
¡atJS ¡e 1 d 't t.1S fy d.c.c. beeausc Z e sequence O ...
so that }; oesn sa d ~ Z ~ H
Chain Conditions G-" O,
. f a e e beca use
Gd
oesn •t. oesn't ·, an d H doesn't ~
satts Y · • •
S) Thc ring k [x ] (k a field~ x an indeterminatc) satisfi
es a.e.c. but not d
on ;deals. .c.c.
So far we have considered quite arbitrary commutative ri ngs (with idcntity). 6) The polynomiaJ.ring k d' [x¡, x2, . .. ] in an infinitc b .
To go further, however, and obtain de~per theo~en~s we need to i~posc some . num er ofmdct .
. sntisfies nelther el1am con 1t10n on ideals: for the ermmatcs
finiteness conditions. The most convement way 1s 111 thc form of chain con- Xn ' •
1 . • d h sequence (x 1) e (
e ... ¡5 stnct. y mcreasmg, an t e sequence (x1) ::::> (x2 ) ::::> 3 xlt x~)
ditions". These apply both to rings and modules, and in this chaptcr we strictly decreasmg. 1
(x t) ::::> • • • 1s
consider the case ofmodules. Most of the argumcnts are of a rather formal kind
1) We shall see later that a ring which satisfies de e 'd
and beca use of this there is a symmetry between the ascending and desccnding 'd 1 (Th · · · · · on 1 eals must al
chains-a symmetry which disappears in the case of rings as we shall see in satisfy a.c.c. on 1 ea s. 1s 1s no/ truc in general for mod J. • so
2, 3 above. ) u es. see Examples
subsequent chapters.
Let 1: be a set partially ordered by a relation ~ (i.e., ~ is reflexive and Proposition 6.2. Mis a Noetherian A-module <=> euery submodule .
EXERCISES
•
80
PRIMARY DECOMPOSITION IN
82 NOETHERIAN RI NGS NEOTHERIAN RINGs
. 83
In a Noetherwn ring A every ideal . .
Let B 0 be thc algebra genera ted ovcr A by thc b11 and thc b s·
tncc ,no 7· ·
11 zs a fimte ;
Noetherian, so is B0 by (7.7), and A S B o S B.
flk ·
A is Lt l Jb/e ;dea/s. ntersection of
·rref.IIC se not ; thcn t he se t o f 1'deals in A for wh· h h
1
Any elcment of C is a polynomial in the x, with cocfficicnts in .f suppO . 1 l IC t e lcmm .
stituting (1) and making repeated use of (2) shows that cach clcmcnt 0Af S~b. proO;·
mptY 1lC
nce has a maxtma
d
e ement a . Since . .
a lS reductbl
a ts false is
Jinear combination of the y, with cocfficicnts in Bo, and hcncc e . c .Is a not Cll• , here f> :::> a an e ::::> a. Hencc ea eh of b e is fi . . e, we have
{¡ n e w r . , a nttc tnte .
generated as a B0 -modulc. Since Bo is Nocthcrian, and B is a submo~s finttely 0 :::
. eductble 1
• 'deals and there,ore so ts a: contradiction. • rsectton of
1 0
it foJiows (by (6.5) and (6.2)) tbat Bis finitely generated as a Bo-modulu e ~ C, ,rr ]2. fn a N oelhenan
. rmg . every irreducible id . .
B0 is finitely generated as an A-algcbra, it follows that B is finitcly gen e. Stncc Lernrna 7• . . . . ea1 rs Prtmary.
- crated B passing to the quot1ent nng, 1t ts enough to show th t 1·r h
an A-algebra . • as froof. ~ble then it is primary. Let xy = O with y =1= 0 and a .~e zero ideal
is irreduc• ( ) e Ann(x2) ~ ... . By thc a.c e this h, . . con~l er the chain
Proposition 7.9. Let k be a field, E a finitely generated k-algebra. lf E .. 'd als Ann x - · ., e am 1s stahonar ·
fie/d then it is a finile algebraic extension of k. 's a ofJ e
Ann ( x") -- Ann(xn + 1) = . .. . for sorne n· lt 1'"oll ows that (x") n y,( l.e., we
) _ .
Proof. Let E = k[x¡, . .. , Xn ]. If E is not algebraic over k thcn we e
haV~ (y) then ax = O, and 1f a E (x") then a = bxn hence b n+1 _ Y - O,
foflf O E ( n) h bx n = 0 ,. t hat1s,
· ' X - 0, hence
number the x 1 so that Xtt ••• , .Xr are algcbraically indepcndcnt over k anhrc. be Ann ( xn+l) = Ann x ,
r
ence
h .n
a
.
= O. Since(O) ' · d .
lS 1rre uctble
. . ' w ere _J. owe must there,ore 0
ave x = , and th1s shows that 'O) · .
r ~ 1, and ea eh of Xr + 1 , • •• , x" JS algebrmc over the ficld F = k(x . and ( y -r
) · \ ts pnmary. •
Hence E is a finite algebraic extension ofF and thcrcfore finitely gcncr~,t~d -~ .\,). From these two lemmas we have at once
F-mod ule. Applying (7.8) to k ~ F ~ E, it follows that F is a finitcly gcner:ta~
k-algebra, say F = k[y1 , ..• , y3J. Each y 1 is of the form /¡fg1, where ~ ande "''heorem 7.13. In a Noetherian ringA every ideal has a primary decomposr-.
.1,
. ] . J g,
are po1ynom1a s m X¡, .. . , Xr · tion. •
•
Now there are infinitely many irred ucible polynomials in the ring Hence all the results of Chapter 4 apply to Noethcrian rings. •
k[x 11 • •• , Xr ] (adapt Euclid's proof of the existence of infinitcly many prime
numbers). Hence there is an irreducible poJynomial h which is prime to each Proposition 7.14. In a Noetherian ring A, every ideal a contains a power of
ofthe g1 (for example, h = K1K2· · ·Ks + 1 would do) and the element lz - 1 ofF ils radical.
couJd not be a polynomial in the y 1. This is a contradiction . Hence E is alge. Proof. Let x 1 , ••• , xk genera te r(a): say xf' E a (l ~ i ~ k). Let m =
braic over k, and therefore finite algebraic. • ¿~. 1 ( 111 - 1) + l. Then r(a)m is generated by the products xít. . ·Xkk with
¿ ,1 = m; from the definition of m we must have r, ~ n1 for at least one index ;,
Coro/lary 7.10. Let k be afie/d, A afinitely generated k-algebra. Let m be a hence each such monomiallies in a, and therefore r(a)m s a. •
maximal ideal of A. Titen the fie/d A j m is a finite algebraic extension of k.
In particular, if k is algebraical/y closed tlzen A / m ~ k. Corol/ary 7.15. In a Noetherian ring the ni/radical is nilpotent.
Proof. Take E = Af m in (7.9). • Proof. Take a = (O) in (7 .14). •
(7.10) is the so-called "weak" version of Hilbert's Nullstellensatz Corollary 7.16. Let A be a Noetherian ring, m a maximal ideal of A, q any
( = theorem of the zeros). The proof given here is due to Artin and Tate. For ideal of A. Then the following are equivalent :
its geometrical meaning, and the "strong" form of the theorem, see the Exercises i) q is m-primary;
at the end of this chapter. ii) r(q) = m;
iii) m" S q ~ mfor some n > O.
Proof. i) => ii) is clear; ii) => i) from (4.2); ii) => iii) from (7 .14); üi) => ii) by
PRIMARY DECOMPOSJTION IN NOETHERIAN RINGS taking radicals: m = r(m") s; r(q) s; r(m) = m. •
Propositio11 7.17. Let a # (1) be an ideal in a Noetherian ring. The~ the
The next two lemmas show that cvcry ideal -:¡: (1) in n Noetherian ring has u
prime idea/s which belong to a are precise/y the prime ideals which occur rn the
primary decomposition.
set of idea/s (a: x) (x e A).
An ideal a is said to be irreducible if Proor B . O
. ?· Y passmg to A f a we may assume that a = . e 1• ' . 0 fL t 1 nft
q = O be a
111111101al primary decomposition of the zero ideal, and let \'1 be the radtcal
n = b n e ::::- (a = b or a = e). q,.
84 NOETIIERJA N RINGS EXERCISES &
d Je
""'O'" apply (7.8) to obtain a contradiction H
d · ence K ¡ f
S
Let a1 = n 1 ~ 1 q1 :f O. Then frorn the proof of (4.5) we hu ve r(Ann(x)) ""' rno ~ Ílcnce is finitely genera te ns a Z/(p)-algcbra. Use (7 9; o characteristic
J .,
[Let a be a maximal element of E, and supposc that therc exist x, y E A su eh that Yh ich contain X o. Smee m ,+ It . . . , m, +• do not contain o there ex·tst 1 ea
\ ' ) S' X¡E asuch
x rt a and y~ a and xy E a. Show that there cxists a finitcly gcncratcd ideal rhat XJ ~ lltr + J. (1 ~ J .~ s : mee each A m, (J ~ i ~ r) is Noctherian, the ex-
o0 s a such that n 0 + (x) = n + (x). and that n = Clo + x ·(a :x). Sincc tension of a m• A 111 1 ts fimtely gencrated. • Hencc there exist x , . 1t . . . . x, .m a
(a: x) strictly contains a. it is finitely gcnerated and thcrcforc so is n.] whose images m Am1 genera te Am,n for 1 = 1• ... , ' · Let a = (x )
Hence a ring in which every prime ideal is finitcly gcncratcd is Nocthcrian Show that o0 anda ha ve the same cxtcnsion in Am for every maxtmal id~Lm . ·', an
x,d.
(l. S. Cohen). deduce by (3.9) tha t Oo = a .]
A-automorphisms of n, and no the set of all elements of B which are left fixed Ut ts clear that r(a) s /(V). Conversely, Jet/~ r(o), then there is a pnmc Ideal!
containing a such that f ~ p. Let f be the image off in B = Ajp, let e = B, -
by every element of G. Show that no is a finitely generated A-algebra.
B[I ffJ, and let m be a maximal ideal of C. Since C is a finitely generated k·
6. If a finitely generated ring K is a field, it is a finite field . algebra we have C/m ~ k, by (7.9). The images X¡ in C/m of t~e generato~::
(lf K has characteristic o. we have z e Q s; K. Sincc K is finitely generated of A thus define a point x = {X¡, •.. ' Xn) E kn, and the constructJOn shows
over Z it is finitely generatcd o ver Q, hcncc by (7 .9) is a finitely generated Q· xe Vandf(x) ~ 0.]
EXE~CISES 87
86 NOETHERIAN RlNGS
topological space and lct .<JF be the smallest coll .
Lct X be a . s all open subsets of X and is closed with ectJon of subsets of X
15 Let A be a Noetherian local ring, m its maximal ideal an~ k its residue field, nnct zo. whtC••
. h contatn
.
1
sections and comp ements.
rcspect to th
e formation /
• let M be a finitely generated A-module. Then the followmg are equivalent : r. 'te Jnter fXb 1
of ,JOI 1ata subsct E o e ongs to §:" if and onty if E . .
i) Mis free; i) ShOW tfl 01 u n e, whcre V is open and e is closed IS a finnc Un ion of sets
ii) Mis flat ; . . . . f the or . d . bl .
iii) the mapping of m ® M into A ® M 1s mJcctJve; o ose that X is trre uct e and let E e :F. Show that .
ii) SuPP '=' = X) if and only if E co ntains a non-empty E .15 dense in X
iv) Tort (k, M) = O. . . (i.e., that L 0 Pen set m x.
[fo show that iv) => i), Jet X¡, • .. , Xn be elements of M whosc Irnages m lvlf lllM
form a k-basis of this vector s,pace. By (2.8), the x, genera te M . Let F be a free be a Noethcrian topolo~ical space (Chapter 6, Excrcise 5) and l
A-module with basis e¡, ... , en and define</> : F ~ M b~ </>(e,) = x,. Let E == Kcr zt. Let X that E E !F if and only tf, for ea eh irreducible closed set X et E ~ X.
(rf). Then the exact sequence O ~ E ~ F _ ,.. M ~ O g1ves us an exact sequcnce
ShoW
- Xo :¡; Xo or clse E r: x; ·
o contams a non-empty open subsct of X
o S X, e1ther
En ar Thcn the collectton of closed sets X ' s x such that E '""' x?· [Su~pose
o )o k ®A E )o k ® ,. F l®~) k ®A M )o o. E~ .:r • d therefore has a mmtma· · 1 e1emcnt X • Show that X .' '. ~ fF
0 . 1s not
emPtYan . b o ts trreductble and
Since k ® p and k ® M are vector spaces of the same dimension over k, it h 1 each of the alternattvcs a ove leads to the conclusion th. E
thcn t a or 11 d 1 at n Xo e§: ]
follows that 1 ® 4> is an isomorphism, hence k ~. E = O, hence E = o by The scts belonging to .:r are ca e t 1e constructib/e subsets of X. .
Nakayama's Lemma (E is finitely generated beca use 1t ISa submodule ofF, andA
is Noetherian).] L t X be a Noetherian topological space and let E be a subset of X Sh h E
2
2 • . e pen in x if and only if, for cach irreducible closed subset x' ·10 °Xw t .atht
16. Let A be a Noetherian ring, M a finitely generated A-module. Then the following JS o . o , et . er
E r. Xo = 0 or else E fl X o c~ntams a non-empty open subsct of Xo. [The
are equivalent : proof is similar to that of Exerc1se 21.]
i) Mis a fiat A-module;
ii) M-p is a free Ap-module, for all prime ideals +>;
23• Let A be a Noet~erian ring, f: A ~ B-a ring homomorphism of finite type (so
iii) Mm is a free Am-module, for all maximal ideals m. that B is Noethenan). Lct X = Spec (A), Y = Spec (B) and Iet ¡ •: y __.. X be
In other words, flat = locally free. [Use Exercise 15.] the mapping associated with f Then the image under ¡• of a constructiblc
11. Let A be a ring and M a Noetherian A-module. Show (by imitating the proofs subset E of Y is a constructible subsct of X.
of (7.1 1) and (7.12)) that every submodule N of M has a primary decomposition [By Exercise 20 it is enough to take E = V n C where U is open and C is closed
(Chapter 4, Exercises 2Q-23). in Y; then, replacing B by a homomorphic image, we reduce to the case where E
is opcn in Y. Since Y is Noetherian, E is quasi-compact and thercfore a finite
18. Let A be a Noetherian ring, p a prime ideal of A, and M a finitely generatcd union of open sets of the form Spec (8 0 ). Hence reduce to the case E = Y. To
A-module. Show that the following are equivalent : show thatf•( Y) is constructible, use the criterion of Exercise 21. Let Xo be an
i) p belongs toO in M; irreducible closed subset of X such that [•( Y) n X0 is dense in Xo. We ha ve
ü) there exists x e M such that Ano (x) = p;
¡t(Y) fl Xo = f*(f* - 1(X0 )), and/* - 1 (X0 ) = Spec ((A/lJ) ® A B), where Xo =
iii) there exists a submodule of M isomorphic to A/p.
Spec (A/p). Hence reduce to the case where A is an integral domain and/is injec-
Deduce that there exists a chain of submodules
tive. If Y¡, ... , Y 11 are the irreducible components of Y, it is enough to show that
O = Mo e M 1 e · · · e M, = M somef•( Y,) contains a non-empty open set in X. So finally we are brought do~n
to the situation in which A, B are integral domains and f is injective (and sull
such that each quotient Me/M 1_ 1 is of the form A/p, where p, is a prime ideal
of A. of finite type); now use Chaptcr 5, Exercise 21 to complete the proof.]
•
19. Let a be an ideal in a Noetherian ring A. Let 24. With the notation and hypotheses of Exercise 23, ¡• is an open mapping .;:>
f has the going-down property (Chapter 5, Exercise 10). [Suppose 1 has tl:e
. that E -- ¡•e Y) ts
, 1
a = n b, = fl
/. i
C¡
.
gotng-down property As in Exercise 23 reduce to provmg
· · ' ·
lJ E E and l->' e- lJ 'dthen
C•l
open m X. The going-down property asserts that 1f X
be two minimal decompositions of a as intersections of irreducible ideals. Prov~ p' e E: in other words that if X.0 is an irreducible closed subset of x.anX] 0
that r = s and that (possible after re-indexing the c1) r(6 1) = r(c,) for all 1• meets E, then E n X 0 is' dense in X 0 • By Exerctses
. 20 and 2-, · 0 pen m .
., E 15
[Show that for each i = 1, ... , r there exists j such that
2S· Let A be· Noetherian f· A ~ B of finite type and flat ('t.e., B is ftat. as anand
A·
a= blrt···rtb 1 _ 1 rt c1 rtb 1+ 1 rt···rt b,.] mod ' · · [Exerc1se 24
C ule). Then ¡•: Spec (B) ~ Spec (A) is an open mappmg.
State and prove an analogous result for modules. hapter S, Exercise 11.]
4+I.C.A,
88 NOETHERIAN RINGS
Grolltendieck groups 8
26. Let A be 3 Noctherian ring and Jet F (A) denote the set of all isomorphis .
e 1 . m ~lass
of finitely gcnerated A-modules. ctL b e t 1e r~c a e tan group gcncratcd ·es
f b l
F(A). With each short cxact sequcnce O -).. M - > l vf - > tvr -~ o of fi . by
1
· h 1 ( " '
f ' )
generated A-modules we assoc1ate t e e cmcnt ¡ v, - (!vi) + (/'ot/") 0itcly .
01
where (M) is t11c isomorphism class of !<.11, cte. Lct D be thc suboro
b up or C,
e ¡\.rti n Ri11g·s
generated by these clements, for nll s 110rt cxnc t scqucnccs. Thc quoticnt .
e¡ D is callcd the Grothendieck g roup of A , and is dcnot~d by .K (A). 1f /\~ ~u~
1
finitely generated A-module, Jet y(i\lf), or YA(NI) , denote the 1magc of (M) i
K(A). n
i) Show that K (A ) has the following universal propcrty: fo r cach a·•d· . . ring is onc which satisfics thc d.c.c. (or cqu· 1
• • u lllvc An Artm . .1 tva cnt1y thc minimal
function .\ on thc class of firutcly gcneratcd A-modules, wi th valucs ·
d .t,·on) on tde~• s.
abelian group G, thcre ex1.sts a u m.que 110momorpl11.sm A0 : K(A) - > G1n an
1
eon apparent symme t ry W .Jt 11 N oct l1cnan
· nngs
· is howcv .. .
suc 1\ TIl e , A . . . cr mts 1cadmg 1
that .\(M) = Ao(r(M)) for all M . " C will show tha t an rtm nng 1s nccessarily Noethcria d f · n
ii) Show that K (A) is generated by thc clcmcnts y(A/V), whcrc p is a prime id t
fac '" A · · . . n an o a very
1 . 1kind rn a sen se a n rt111 n ng ts thc s1 m plest kind of ring r· fi
of A. [Use Exercise 18.] ca specHl • f . . a ter a 1cld, and
udy them not beca use o thc1r gencraltty but beca use of the' . ..
iii) If A is a ficld, or more generally if A is a principal ideal domain, thcn WC St Ir Slmp1ICJty.
K(A) ~ Z. Proposition 8.1. In an Artin ring A erery prime ideal is maxima/.
iv) Let J; A -> B be a finite ring homomorph ism. Show that restriction of
Proof. Lct V be a prime ideal of A. T hen 8 = A/p is an Artinian integral
scalars givcs rise to a homomorphism Ji: K ( B ). - > K (A ) su eh that fi(y 1j(N))
domain. Let xE B, x # O. By the .d .c.c. we have (x 11 ) = (xn+l) forsomen
= YA(N) for a E-module N. lf g: B -+ e I S a nother fi ni tc ring homo-
morphism, show that ( g o J) 1 = Ji o g 1• hcnce xn = x n+ly rt Or sorne Y E B . Smee B '1s an mtegral· domain and x ~ O, it'
follows that we may cancel x", hencc xy = l. Hcncc x has an inverse ¡11 B and
27. Lct A be a Noethcrian ring and Jet Ft (A) be the sct of a ll isomorphism classes of therefore Bis a field , so tha t V is a maximal ideal. • '
finitcly generatcd /!al A-modules. Repcating thc construction of Excrcisc 26 we
obtain a group Kt( A). Let Y1 (.Nf) denote the imagc of {M ) in K 1(A). Corol/ary 8.2. In an Artin ring the ni/radical is equal to tite Jacobson
i) Show that tensor product of modules over A induces a commutativc ring radical. •
structure on K1(A), such that Y1(M ) · Y1(N ) = Yt( M 0 N). Thc identity
elcment of this ring is y 1{A). Proposition 8.3. An Artin ring has only a finitc numbcr of maximal idcals.
ii) Show that tensor product induces a K1 (A)-module structure on the group Proof. Considcr the set of all fin ite intersections m 1 n · · · n mn whcre thc m1
K(A), such that y 1 (M) · y( N) = y( M ® N) . are maximal ideals. This set has a mi ni mal elcmcnt, say m1 fl· · ·flm 11 ; hencc
iii) lf A is a (Nocthcrian) local ring, thcn K 1(A) ~ Z . for any maximal idea l m wc have m n 111 1 fl · · ·fl mn = m1 fl· · · fl lll 11, and
iv) Let J: A - > B be a ring homomorphism, B bcing Noetherian. Show that therefore m 2 m1 n · · · n mn. By ( l . ll ) m 2 m1 for sorne i, l~ence m = m,
extension of scalars gives rise to a ring homomorphism J ': K 1 (A ) -+ K 1(B) since m1 is maximal. •
such that f!(rt(M)) = Yt(B ® A M). [lf M is ftat and finitely generated
over A, then B ®,, Mis flat and finitely generated over B .] If g: B _..,. C is Proposition 8.4. In an Arrin ring tire ni/radical ~¿ is ni/poten!.
another ring homomorphism (with e Noetherian), then ( f o g ) l = [ 1 o g 1• Proof By d.c.c. we ha ve mk = mk + 1 = ... = o say, for so me k > O. Suppose
v) If J; A -,. Bis a finite ring homomorphism then 0
f: O, and let :E denote the set of all ideals ú such that oó =/= O. Then ~ is not
empty, since o E :E. Let e be a minimal element of L; then there exists x E e such
fi(J 1(x)y) = xjj(y )
that xo :1: O; we ha ve (x ) s; e, hence (x) = e by the minimality ?r.c. .But
for x E Kt(A), Y E K(B). In other words, regarding K (B) as a K 1 (A)-module (xa)a = xa 2 = xa '::/: O, and x a s (x) , hence xo = (x) (again by rmmmahty).
by restriction of scalars, the homomorphism J ' is a K 1 (A)-module homo·
morphism. ~ence x = xy fvr sorne y E a, and therefore x = xy = xy2 = · · · = xy" =~~:15·
ut Y E 0 == 91,.. 2 ~1, hence y is nilpotent and thercfore x = xyn = 0·
Remark. Since Ft(A) is a subset of F(A) we have a group homomorphism
': Kt(A) ~ K(A), given by E(Yt{M)) = y(M). If the ringA is finite-d imensional contradicts the choice of x, therefore o = O. •
•
and re~ular•. i.e., if all its local rings A v are regular (Chapter 11) it can be shown By a chain of prime ideals of a ring A we mean a finite strictly incfireastmhg
that ( 1s nn 1somorphism. ' sequence · 1·5 11 We de ne e
~lo e P1 e • • • e .\Jn; the /ength of the eh a m •
89
90 ARTJN RINGS
EXERClSES
dimension of A to be the supremum of the lcngths of all chains of prin . A ring with only one prime ideal necd not b N 91
. . . . . A O) fi . 1C tdca) mplc. L t A - k[ · e oethe ·
m A: Jt IS an mteger ~O, or + oo (assummg . :¡:. . A 1cld has dtmension O. ths f;%11 ¡\rtin ring). e . - . X¡, x 2, ... ] be the polynomial . n~n (and hence
ring Z has dimension l . e
1
not ~n set of indetermmates x" over a field k, and let a be n~g tn a countabl
¡nftnJt~ The ring B = A j a h~s only one prime ideal ( the tdeal (x¡, ~~ .. :
Tlreorem 8.5. A ringA is Artin -=· A is Noetherian and di m A = O. -"~' · · · · x . .. ) ), hence B 1s a local ring of di m _namely the image or'
Proor. =>: By (8.1) wc ha ve dim A = O. Let m, ( 1 ~ i ~ n) be thc d . . t . .. , "' d 'ffi l ens10n O B
(XJ• ' 2 ' • 0 for it is not 1 1cu t to see that its prim .d . · ut B is not
'J.
maximal ideals of A (8.3) . Th en ri", .. 1 m," s; Cn", ... 1 m,)" = 9l " = O. Hcncc tsttnct
b ,.roetherta
l"'
' e 1 eal ts not fi ntte
. l
y gen~
(6.11) A is Noethcrian. Y erated.
is a local ring, m its maximal ideal, k = A/ . .
<=: Since the zero idepl has a primary decomposition (7 . 13) A has oni . 1 Ir le rn/nt2 is annt'h t'lated bY m and therefore has th m lts restdue field, the
A
finite number of minimaJ prime ideals, and these are all maximal sin ce di m A !_ .t A-modu . d( . . e structure of a k ,
Hence 91 = nr. 1 m , say; we have V'l" = O by (7. 15), hence . 01.... 1 tn~ = o (a-s 1?· Jflll is fimtely generate e.g., 1f A 1s Noetherian) th . . ~\ector
.
the previous part of the proof. Hence by (6. J 1) A ts an Arttn nng. •
n spa~· of generators of tn will span mfm2 as a vecto~ se tmages m mfm2 of
a. se (tn/nt2) is finite. (See (2.8).) pace, and therefore
Jf A is an Artin local ring with maximal ideal m, thcn m is thc only pri dunk
ideal of A and therefore m is the nilradical of A. Hcncc every elcmcnt of 11 ~c Prornosition 8.8. Let A be an Artin local ring. ,.,..,
1
J ,zen the following are
nilpotent, and m itself is nilpotent. Every element of A is either a unit or .s equivalen!:
. . j (p IS
nilpotent. An example of such a ~mg ts Z "), wherc p is prime and n ~ l. i) every ideal in A is principal;
Proposition 8.6. Let A be a Noetherian local ring, m its maximal ideal. Then ii) the maximal ideal m is principal;
~ l.
2
exactly one of tlze following two statements is true: iii) dimk (m/m )
i) m" '/; m" +1 for al/ n,· Proof. i) => ii) => iii) is clear.
2
ii) m" = Ofor some n, in which case A is an Artin local ring. iii) => i): If dimk (m/tn ) .= O, then m = m2, hence m = oby Nakayama's
Proof Suppose m" = m"+ 1 for sornen. By Nakayama's lemma (2.6) we havc lemma (2.6), and therefore A IS a field and there is nothing to prove.
m" = O. Let ~ be any prime ideal of A. Then m" s; ~, hence (taking radicals) If dimk (m/m2 ) = 1, then m is a principal ideal by (2.8) (take M = m
m = ~. Hence m is the only prime ideal of A and thercfore A is Artin ian. • there), say m = (x). Let a be an ideal of A, other than (O) or (1). We have
m = 91, hence m is nilpotent by (8.4) and therefore there exists an integer r
Th~or~m 8.7. (structure theorem for Artin rings). An Artin ring A is such that a ~ m', a $ n1' + 1; hence there exists y E a sueh that y = ax',
uniquely (up to isomorphism) a finite direct product of Artin local rings. y~ (x'+l); consequently a f# (x) anda is a unit in A . Hence x' e a, therefore
Proof Let m, (1 ~ i ~ n) be the distinct maximal ideals of A. From the proof m' = (x') ~ a and hence a = rn' = (x'). Hence a is principal. •
of (8.5) we ha ve 0f. 1 m~ = O for sorne k > O. By ( 1.16) the ideals mf are
coprime in pairs, nence n m~ = 0 m~ by (1.10). Consequently by (l. lO) again Example. The rings Zf(p") (p prime), k[x]f(f") (f irreducible) satisfy the
2 3 4
tbe natural mapping A~ Df- 1 (A/ mn is an isomorphism . Each A/mf is an conditions of (8.7). On the other hand, the Artin local ring k[x x ]/(x ) does
,
Artin local ring, hence A is a direct product of Artin local rings. not: here m is generated by x 2 and r (mod x 4), so that m2 = O and
Conversely, suppose A ~ Dr. . 1 A" where the A, are Artin local rings. dim (m/m2) = 2.
Then for each i we have a natural surjective homomorphism (projection on thc
ith factor) t/>,: A ~ A1• Let a1 = Ker (t/>1). By (1.10) the a1 are pairwisc coprime,
n
and a, = O. Let q, be the unique prime ideal of A" and let ~~be its contracti~n EXERCISES
•
;,-l(q,). The ideal ~~ is prime aod therefore maximal by (8.1). Since q, is ntl·
potent i.t .follows that a"' is .P 1-primary, and hence n
a, = (O) is a primary de· l. ~t q1 (') • • • f\ q" = O be a mini mal primary decomposition of the ze~o ideal
composttton of the zero ideal in A. Since the a 1 are pairwise coprime, so are the m a Noetherian ring and let q1 be ~ -primary. Let ~i'> be the rth symbollc.power
of "' (Ch ' n there extsts an
apter 4, Exercise 13). Show that for ea eh z = 1' · · ·,
' •
.p., and they are therefore isolated prime ideals of (0). Hence all the primary . "''
co~ponents a, are isolated, and therefore uniquely determined by A, by the_2n~ lnteger r, such that h<,,> e q1
S .,.,, - '
· 1 1
• an Artm oca
umqueness theorem (4.11). Hence the rings A ~ Aj a are uniqueJy determme . uppose q, is an isolated primary component. Then Av, tsffi . tly large r
~ 1 - 1 n ng hence ·r . . . . , - o 1~'or all su . cien ,
bYn, • h • I m, ts tts maxtmaltdeal we have m, - ' ·
uniquely in the form pay, where a E Z and both numerator and denomina ter of . ;~: · (2 8) h d' ( ¡ 2 x - (x).
iii) => lV). By . we . ave lm" m m ) ~ 1, and by remark (B) mfm2 =/: O.
are prime to p . Define vp(x) to be a. The valuation ring of Vp is the local rin~ iv) => v). Lct a be an tdeal :F (0), (l). By remark (A) we have a :::> "f¡
Zcp>· some n; from (8.8) (applied to A/m") it follows that a is a power of m.- m or
2) K = k(x), where k is a field and x an indeterminate. Take a fixcd v) => vi). By rema.r k (B), m ::f- m 2 , hence therc exists x E m, x ~ m2. But
irreducible polynomial/E k [x] and defi ne v1 justas in 1). The valuation ring of (x) = m' by hypothes1s, hence r = 1, (x) = m, (xk) = m".
v1 is then the local ring of k [x ] with respect to the prime ideal (/).
vi) => i). Clearly (x) = m, hence (x") ::f- (x" + 1) by remark (B). Hence if a
An integral domain A is a discrete valuation ring if there is a discrete valua- ¡5 any non-zero element of A, we have (a) = (xk) for exactly one value of k.
tion v of its field of fractions K such that A is the valuation ring of v. By (5.18), Define v(a) = k and extend v to K* by defining v(ab - 1 ) = v(a) - ~;{b). Check
A is a local ring, ancl its maximal ideal mis the set of all x E K sucp that v(x) > O. that vis well-defined and is a discrete valuation, and that A is the valuation ring
If two elements x , y of A have the same value, that is if v(x) = v(y), then of v. •
v(xy - 1) = O and therefore u = xy - 1 is a unit in A. Hence (x) = (y).
If a :¡:: O is an ideal in A, there is a least integer k such that v(x) = k for
DEDEKJND DOMAINS
sorne x E a. It follows that a contains every y E A with v(y) ~ k, and therefore
the only ideals :¡:: O in A are the ideals m k = {y E A: v(y) ~ k}. These form a Theorem 9.3. Let A be a Noetherian domain of dimension one. Then the
single chain m ::::> m2 ::::> m3 ::::> • • · , and therefore A is Noetherian. fo/lowing are equivalent:
Moreover, since v: K* ~ Z is surjective, there exists x E m such that i) A is integral/y c/osed;
v(x) = 1, and then m = (x), and mk = (xk) (k ~ 1). Hence. m is .the o~ly ii) Every primary ideal in A is a prime power;
non-zero prime ideal of A, andA is thus a N oetherian local domam of dunenswn iii) Every local ring A" (p :¡:: O) is a discrete valuation ring.
•
one in which every non-zero ideal is a power of the maximal ideal. In fact many Proof. i) <=> iii) by (9.2) and (5. 13). .
of these properties are characteristic of discrete valuation rings. iQ <=> iii). Use (9.2) and the fact that primary ideals and powers of ldeals •
. . ifd ' ·on one tn its behave weU under localization : (4.8), (3.11). •
Propo1itio11 9.2. Let A be a Noethenan local domam o 1mensl .' .
1
maximal ideal, k = Af m its residue field. Then the f ollowing are equzva ent. A ring satisfying the conditions of (9.3) is called a Dedekind domain. '
••
•
i) A is a discrete valuation ring; CoroUary 9.4. In a Dedekind domain every non-zero ideal has a unique
ii) A is integral/y closed,· factorization as a product of prime ideals.
iii) m is a principal ideal; Proof. (9.1) and (9.3). •
96 DISCRETE VALUATION RINGS ANO DEDEKIND DOMAlNS
FRACTIONAL lDEAts
Exam¡)Jcs. 1) Any principal ideal domnin A. For 1 1• No ti . 97
'd 1 · fi · / 5 e 1crnn ( ·
r ea rs nrtely gencrated) and of dimcnsion onc (Examplc . f :. 511 H:c cvcry t lbility is a local propcrty :
rnvcr
3
cvery local ring Av(lJ =/: O) is a principal ideal domain hcncc l.t tc(r (1. 6)). 1\lso . osition 9.6. For a fractional ideal M thefc .
)y 9·-?) .•1 d'rscrctc
va 1uatron
. .
rrng; hencc A rs.
a Dedekmd . '
dorna in by (9.3). proP . . , o11owmg are equil'a/ .
i) tvf is ínverllble, . em.
2) Lct K be an algcbraic number ficlcl (a finite algebraic . .
Its ring ofintegers A is the integral closurc of z in K. (For exam ~~~c~~"''? 11 of Q). ") ¡.,l ís finite/y generated al/(/, jor each prime ideal h M . .
JI • ,.,, P ts mvertible .
then A = Z[i], the ring of Gaussian integcrs.) Then A is a D~d \ ~ 1\ ::::: Q(i), ·'') M is fimtely generated al/(/, for each max· ·[ ·¡
Jll • una u ea[ lll Al
.
.
. . . e rnd dornain · · vert1'b/e•·
111 • m IS
Tlleorcm 9.5. The rmg of mtcgcrs 111 an algC'braic 111 m1hcr }iC'/d , . ·
Dedekind domain. 1\. IJ u .r i) => ii): Av = (M ·(A:Nl))v = Mv·(Av:Mv) by (3 11)
Pro.o,.fi itely gcncrated, bccause invcrtiblc). · and (3. 15) (for
Proof K is a separable extcnsion of Q (beca use the charactcri ~r , . . M!S Jll
hence by.(5.1~) thcrc is a basis ut> .. . , un of K ovcr Q such that ;t= ~s ;e~~_}· ii) => iii) as usual.
Hence A JS finrtely gencrated as a Z-mod ulc and thercforc Nocthcrian \j )· ...1 => i): Lct a = NI · (A: !vf), which is an integral ideal F . h .
JI ( ( • or cae maxtmal
is integrally closed by (5.5). To complete the proof we must show that e~/ , so ,..J 'd ,11 lit wc have Clm = lvfm· Am:Mm) by (3.11) and (315)) _A b
1 Ce • r1- . • - m ecause
· 'd 1
zero pnme 1 ea .P o f A · · 1 · r) 11011·
1s maxrma , a n? th rs f~llow~ from (5.8) and (S. ): Mmis invcrtJble. Hence a ::;:: m. Consequently a = A and therefore M is
9
(5.9) shows that p n Z =/: O, hence \.> n Z 1s a maxrmalrdcal of z and thcr·r invcrtiblc. •
· maxtma
.p rs · 1 111
· A by (5. 8) . • Ca ore
Proposition 9.7. Let A be a local domain. Then A is a discrete raluation
R emark. The uniquc factorization theorern (9.4) was originally preved for ring <=> every JlOII-zcro fractiona/ ideal of A is invertible.
rings of intcgers in algebraic numbcr ficlds. The un iqucncss thcorcms of Chapter Proof =>. Lct x be a generator of thc maximal ideal m of A, and Jet M ::¡:. 0
4 m ay be rcgarded as gencralizations of this result: prime powcrs ha ve to be be a fractional ideal. Then there exists y E A such that y M s A: thus y Misan
replaced by primary ideals, and products by intersections. integral ideal, say (xr), and therefore M = (xr -s) where s = v(y).
FRACfiONAL IDEALS <= : Every non-zcro integral ideal is invertible and therefore finitely gen-
erated, so that A is Noetherian. It is therefore enough to preve that every non-
Lct A be an integral domain, K its field of fractions. An A-submodulc ¡\/ of K zero integral ideal is a power of m. Suppose this is false; let L be the set of non-
is a fractional ideal of A if xA1 S A for sorne x =/: O in A. 1n particular, the zero ideals which are not powers of m, and let a be a maximal element of E.
"ordinary" ideals (now called integral ideals) are fractional ideals (takc x = 1). Then a =fi m, hence a e m; hence m -l a e m - 1m = A is a propcr (integral)
Any element u E K generates a fractional ideal, denoted by (u) or Au, and called ideal, and m - 1 n 2 n. If m -lo = n, then a = ma and therefore a = O by
principal. If Mis a fractional ideal, the set of all x E K such that x 1\l s; A is dc- Nakayama 's lcmma (2.6); hence m -la ~ a and hence m- 1a is a powcr of m
noted by (A: M). (by the maximality of a). Hence a is a power of m: contradiction. •
Every finitely generated A-submodule M of K is a fractional ideal. For if M
is generated by X¡, •. . , Xn E K, wc can writc x 1 = y 1z - 1 (1 ~ i ~ n) wh~re J'¡ The "global" counterpart of (9.7) is
and z are in A, and then zM s A. Convcrsely, if A is Noetherian, cvery fractronal Theorem 9.8. L et A be an integral domain. Then A is a Dedekind domain <=>
ideal is fini tely generaled, for it is of the form x- 1 n for sorne integral ideal 11 •
every non-zero fractional ideal of A is invertible. .
An A-submodule M of K is an invertible ideal if there exists a submodulc
Proof. =>: Let M =1= O be a fractional ideal. Since A is N~etheri.an, ft~ 15
N of K such that MN = A. The module N is then unique and cqual to (A: /11~, ·
firnrtcly . . 1
generated. For each pnme 1dea lJ =1= ' P 0 1~~ 1·s a fractronal rdea r 0
H
1
\/ ¡5
for we havc N s (A:M) = (A:M)MN sAN =· N. Jt follows that M 15
A1 · (A·· M) == ce
A ~f the discrete valuation ring Av, hence is invertible by (9.7). cnce 1
fi nitcly generated, and therefore a fractwnal · ·tdeal: 10r
r ·
smce
rnvertible, by (9. 6). .
-
thcre exist x1 E M and y 1 E (A: M) (1 ~ i ~ n) such that L: Xt.Yt ~ 1, and hdcnby . . . 'bl hence finitely gencrated,
for any x E Al/ we have x = L. "' (y 1x )x1: eac l1 y,x E A , so that NI JS oocneratc -:::: Every non-zero integral tdeal ts mvertr e, . discretc valun-
~cnce A is Noctherian. We shall show that each Av~~ .¡. ~ ¡~:al =F o in Av is
0 1
x1, . .. , x". . .·
'bl its inversc
Clearly every non-zero principal fractional ideal (u) ts mvertr e,, . . ation ~ron r~ng. For this it is enough to show that ea~h 1 :te~;aideal in JÍp, and Jet
. .
being (u- 1). T he mvert1ble . h resp ect to mu trp1re
.tdeals form a group w1t , rnvertrble, and then use (9. 7). Let ó =/= O be an (mteo:a • t'bl by (9.7). •
n -- ur.c == ú n A. Then n •ts ,mvert•b • } l
e, 1ence r.u -- oP JS mver 1 e
whose identity element is A = (J ).
98 OISCRETE VALU ATJON RI NGS ANO DEDEK INO DOMAINS
•
Corol/ary 9.9. lf A is a D edekind domain tite 11011_ ,.,,.0 r., . 99
' .... J' act1o,1 1
form a g roup with respect to multiplicarion. • a idea/so¡ A ~cJSES . . ..
Tbis gro up is callcd the g roup of idea/s of A; we denote 't b A be a Dcdckmd do.mam, S a. multiphcativcly closcd subs
1 l· Lct A ·s either a Dedckmd domam or the ficld of fract' et of A. Show that
1
tenninoJogy (9.4) says that is a free (abclian) gro up, gencra tcd b 1· In this {h S ~ •Sulpposc that S ::/:= A - {O}, and lct H , H ' be the i~:~s ~fA.
prime ideals of A . Y e non-zcro
~ •A respectively. Show thnt cxtcnsion of ideals inda e ass gr~ups of A and
Let K* den o te thc multiplicative group of thc ficld of fract' _, S 1 • m H _ ,.. H ' . uces a surJective homo-
morP 115
Each u E K* defines a fractio nal idea l (u), and th e mapping ll l->- ( ·~·~s .1\ of .{
be a Dedekind domain. lf 1 = a 0 + a 1x + ... + .
morphism e/>: K* -4- l. The image P of e/> is th e group of princ/' / sr ~ h~rno. ., Let A . A f . llnx" IS a poly .
'" with coefficients 10 , t 11c content o 1 1s the ideal c{f) == (a )n~m1al
ideals: the quotient group H = 1/P is callcd the ideal class gror~a /act~ona) prove Gauss's /emma ~hat c.(d/g) =] c( f} c(g).
1 1
0
• • .. , a" 1n A.
kernel Uof c/> is the set ofall u E K* such that (u)= (1 ) so tha t it is t~ A. Thc
units of A . We ha ve a n exact sequence ' le group o¡ [Localize at each max1 ma 1 ea .
al uation ring (other than a field) is Noetherian if and onl ·r. . .
J, A v . Y 1 1t 1s a dtscrete
1 -4- U - >- K* - >- 1 - >- H ~ l . vaJuation nng.
t A be a local domain which is not a field and in which the maxim 'd .
Remark. For the Dedekind do ma ins th a t a rise in numbe r theory tllc
. . ' re are
L
4. e n <O n - o p h A. .
principal and n. 1 m - · ro ve t at · 1s a d1screte valuation ring.
a11 ea1m IS
class1cal theorems .r ela!mg to .the gro ups f! a~d U. Let K_ be an a lgcbraic numbcr
fie.ld and let A be 1ts rmg of mtegers, wh1ch JS a Dedekmd domain by (9.5). In S. Let M be a finitelt generated mod ule over a Dedekind doma in. Preve that Mis
th1s case: flat <=> Mi s tors10n-free.
1) H is a finite gro u p.' Its o rder h is th e class number of the field K. Thc foi- [Use Chapter 3, Exercise 13 and Chapter 7, Exercise 16.]
Jowing are equivalent: (i) h = 1; (ii) 1 = P ; (iii) A is a pri ncipa l ideal domain· 6 Let M be a finitely~generated torsion module (T(M ) = M) over a Dedekind
(iv) A is a unique factorization do main. ' · domain A. Pro ve that M is uniquely representable as a finite direct sum of mod-
ules A/p~•, where P1 are non-zero prime ideals of A. [For each p '1= o, Mp is a
2) U is a finitely -generated abelia n gro up. M o re prcciscly, wc can specify torsion Av-module ; use the structu re theorem for modules over a principal ideal
the number of generators of U. First, the elemcnts of.fi nite o rdc r in U are just doma in.]
the roots of unity which lie in K, and they fo rma fi ni te cyclic group W; U/ W is
1. Let A be a Dedekind doma in and a :f:. O an ideal in A. Show that every ideal in
torsion-free. The number of generators of Uf W is give n as follows: if (K : Q) = n
Afa is principal.
there are n distjnct embeddings K -+ e (the field of complex nu mbers). Of Deduce that every ideal in A can be generated by at most 2 elements.
these, say r 1 map K into R, and the rest pa ir off (if a is one, then w o a is another,
where w is the automorphism of e defined by w(z) = z) into say pairs: thus '2 8. Let a, ó, e be three ideals in a Dedekind domain. Prove that
r 1 + 2r2 = n. The number of generators of Uf W is then r 1 + r 2 - l. a r. (ú + e) = (a l l ó) + (a n e)
The proofs of these results belong to algebraic number theory and not to a + (6 n e) = (a + ó) n (a + e).
commutative algebra: they require techniques of a different nature from those [Localize.]
used in this book. 9. (Chinese Remainder Theorem). Let alt ... , On be ideals and let X1, • • · ' x" be
elements in a Dedekind domain A. Then thc system of congruences x e
Examples. 1) . K = Q(v=l); n = 2, r 1 = O, r2 = 1, r 1 + r2 - 1 = O. The x, (mod O¡) (1 ~ i ~ n) has a solution x in A <=> x, = x, (mod a, + a,) when-
only units in Z[i] = A are the four roots of unity ± 1, ±i. ever i :¡:. j .
[This is equivalent to saying that the sequence of A-modules
2) K= Q(v2); n = 2, r 1 = 2, r 2 = O, r 1 + r 2 - 1 = l. W = {± l},
and U/ W is infinite cyclic. In fact the units in A = Z[v'2] are ± (1 + Ví)", A ~ Ef)
n "
Afa, -+ ffi A/(a, + 01
) -
•
where n is any rational integer. l • l 1<1
analogous to tbe successive approximations given by the terms of a Taylo: very definite geometrical interpretation. For example, if A is the local ring of a
expansion and, justas it is convenient to introduce formal power series, so it is point p on a variety V with a as maximal ideal, then Ga(A) corresponds to the
convenien t to introduce the p-adic numbers, these being the limit in a certain projective tangent cone at P, i.e. all the lincs through P which are tangent to v
sen se of Z/p"Z as n -+ oo. In one respect, however, the p-adic numbers are more at P. This geometrical picture should help to explain the significance of G11(A)
complica ted than form al power series (in, say, one variable x). Whcreas the in connection with the properties . of V near P and in particular in connection
~
polynornials of degree n are naturally embedded in the power series, the group with the study of the completton A .
Zfp 11Z cann ot be embedded in Z. Although a p-adic integer can be thought
of as a power series ¿ OnP" (O ~ a" < p) this representa ti on docs not beha\'e
well undcr the ring operations. TOPOLOGIES AND COMPLETIONS
In this chapter we shaJl describe t he general process of "adic" completion- Lct G be a topological abelian group (written additively), not necessarily
the pri me p being replaced by a general ideal. It is most conveniently expressed Hausdorff: thus Gis both a topological space and an abelian group, and the two
in topological terms but the reader should beware of using thc topology of the structures on G are compatible in the sense that the mappings G x G ~ G and
real numbers asan intuitive guide. Instead he should think of the power series G-+ G, defined by (x, y) H- x + y and x ~ - x respectively, ~re contt.nuous.
topology in which a power series is "small" if it has only terms of high orde~. If {O} is closed in G then the diagonal is closed in G x G (bemg the mver~e
Alternatively he ca n thi nk of the p -adic topology on Z, in which an integer 15 ' mapping (x, y)~ x-y) and so G ts
image of {O} under the · H ausd?rff· lf a ts
"small" if it is divisible by a high power of p . a fixed elemen t of G the translation Ta defined by Ta(x) == x +.a 15 a \hohmeo-
. · T oJ. ence
Completion, like localization, is a met hod of simplifying things by co~· morphism of G onto G (for Ta is continuous, and tts mverse ts -. ' d
centrating attention nea r a point (or p rime). It is, however, a more drtstl~ if U is any neighborhood of Oin G thcn U + a is a neighborhood of a 111 G,,an
' · h. f¡ Thus the topo ogy
simplificati on than localization. For example, in algebraic geometry th~ 0 ~~ conversely every neighborhood of a appears m t IS o~m.
ring of a non-singular point on a va riety of dimension n always has . ~r ~: of G is uniquely determined by the neighborhoods of O111 G.
completion the ring of formal power series in n variables (this will essent•a ~ ts Lemma 10.1. Let H be tlze intersection of al/ neighborlloods oJO in G. Then
. . ft such polO
pro ved 111 Cha pter 11 ). On the other hand the local nngs o wo . . aliY i) H is a subgroup .
cannot be isomorphic unless th e va rieties o n which they lie are b•ratwn ·
lOO
J 02 COMPLETIONS
TOPOLOGIES AN
. D COMPLETtONs
ii) H is tlze c/osure of {0}.
+ Gn ¡5 a netghborhood of g; sincc g + G e: . 103
iii) G/ H is Haus dorff. then g J-Ience for any ¡, the coset h + Gn is open and " - G" thts shows G .
iv) Gis Hausdorff <=> H = O. open. . since this is the complement of Gn in G it ~ollthcreforc Uhto (h + "ats
·s open, . . b ows that G . ~ ,J
Proof i) follows from the continuity of the group operat ions F .. 1 f r topologJes gtven Y scq uenccs of subgrou ps th . " ts closed
. or 11) wc h o .. f 1 1 . . ere t s a n l . .
, ·e dcfi mtwn o t 1e comp etton whtch is oft . a ternattVe pu 1
x E H <> OE x - U fo r aJI neighborh oods u of ave : algebrat . G TI 1 . en convemcnt S re y
- 0 e uehYseq uencc m . len t le unagc of x in G/G . . uppose (x )
<> X E {0}. is a a f f . -v " ts ultimat 1 '
l saY to ~~~ · 1 we pass rom 11 + l to " tt is clcar that ~ e y constant
cqua ~ " +1 ~ t: '
ii) implies that the cosets of H are a ll closed; thus points are 1 . · ction ~ .. Under the
proJe o,. ~ 1
and so G/ H is Hausdorff. Thus H = O :::- G is Hausdorff an~ tlsed 10 GfH
0
•rs tnVIa
• • J. • , lC CO!lVC
G/Gn+ 1 · >- G/ G11 •
rsc
ThuS a Cauch y sequcnce (xv) in G defines a coherem sequence ( .
Assume for simplicity that O E G has a countable fu nda ment 1 en) tn the sense
neighborhoods. Then the completion G of G may be defi ned in the ~su~s~; of that
on+l~n +l = ~" for all n.
means of Cauchy sequences. A Cauclzy sequen ce in G is defined to be Yby
. a scquencc
(xv) of elements of G such that, for any nerghborhood U of o there . Moreover it is clear.
that equivalent Cauchy sequences defi 1
h nc t 1e same seq
· s(U) wrt· h t he property t h at ' cxtsts an
mteger (!>')
n •
Finally, gtven any co erent sequcnce (g ) we ca
• • • • b k. n,
uence
n construct a Caueh
sequence (x n) gtvmg rrse to tt y ta mg Xn to be any clement . th y
x 11 - xv E U for all ¡.t, v ;;;¡: s( U).
that x 1 - Xn E Gn)· T hus (J can equally well be defined as tthn e cofset gn (so
- n+ • h b . e set o coherent
Tw~ Cauchy sequences are equivalen/ if ~v Yv ~ O in G. The set of all
- sequences (~11) wtth t ~ o vtous gro~p structu re.
eqmvalence classes of• Cauchy sequences •
rs denoted
• .A
by é. If (x v), (y)v are , We have now arnved at a spectal case of inverse fimit< "·· More generally, 1
Cauchy sequences, so rs (xv + Yv), and 1ts class m v depends only on the classes consider any sequence of groups {An} and homomorphisms
of (xv) and (Yv). Hence we ha vean addition in G with respect to which 6 ¡5 an
On+ l: An +l-+ An.
~belian group. For each x E G the class of the constant sequence (x) is an
element cf>(x) of G, and 4>: G --)- G is a homomorphism of abelian groups. Note We call this an inverse system, a?d the group of all coherent sequences (an)
that 4> is not in general injective. In fact we have (i.e., a11 E A,.. and 0,.. +1an +1 = an) 1s called the inverse limit of the system. It is
Ker e/> = U n usually written ~m A n, the homomorphisms O" being understood. With this
where U runs through all neighborhoods of O in G, a nd so by ( 10. 1) ~ is injective notation we bave
if and only if G is Hausdorff. G ~ lim G/Gn.
(
We shaJJ then say that we have an exact sequence of inverse .r Apply (l0.2) to the cxact sequences lOS
díagram certainly induces homomorphisms systems. 'fhe
proo; · G' G G"
O - >- Ji m A n -+ Hm B" - >- lim
o( o( <
e" _,. O O -+ G' n Gn ~ Gn ~ pG"-+ O. a
but this sequence is not always exact. Howevcr, wc have . ticular we can apply (10.3) with G' == G then G" _
In par ology so that G" = G" . Hence we dedu~ - G/G,.. has the
Proposition 10.2. Jf O - >- {A n} -+ {Bn} - >- {en} - >- O is an e>:acr . d'screte
1 top
inverse systems tlzen · .sequence oJ.r Corol',ary
' JO •4. <in is a subgroup of (; and
o( o( ~
. inverse limits in (10.4) wc deduce
Taking
is always exact. Jf, moreover, {An} is a surjective system then
proposition 10.5. (; ~ Ó. •
O-+ limAn -+ lim Bn - >-limen -+ O
< < ~--- J.. G -+ (j is an isomorphism we shall say that G is complete. Th
is exact. If 'f'. . f G. l us (lO.S)
that the completJOn o ts compete. Note that our definition of e -
Proof Let A = f1:.
1 An and define dA: A -+ A by el 11 (an) = an - O (a ) asser~ ludes Hausdorff (by (10.1 )). om
plete me f 1 ·
Then Ker dA ~ o(Jim An. Define B, e and d 8 , d e similarly. The exactn+l n+l.
sequcncc The most importan~ class o :xamp es of topologtcal groups of the kind
consideríng are gtven by takmg G = A, Gn = a", whcre a is an ideal in a
of inverse systems then defines a cornrnutative diagram of exact sequences
'~e a~ The topology so defined on A is called the n-adie topology, or just the
0--rA--rB -+C-+0 nng 1·ogy· Since the a" arehideals,
a-topo h it. is not hard
. to check that
. with this topology
dA~ dn-J, de_¡, . topological ring, i.e. t at t e nng operat10ns are contmuous. By (10.1) thc
0-+A-+B -+ C -+0 A IS al gy is Hausdorff
topo o
~n... .
a" = (0). The completion A of A is again a topo-
. . h h' h k .
logical ring; 4>: A -+ A ts a contmuous nng omomorp tsm, w ose ernel 15
and hence by (2.10) an exact sequence
O -+ Ker dA-+ Ker dB -+ Ker d e--+ Coker dA ->- Coker d 8 -+ Coker d e -7 O. n•¿ikewise for an A-module M: take _G = M! G, = a"M_· Thjs defines ~he
To complete the proof we ha ve only to prove that a-topo1ogy 011 M ' and the completion M of M
. ts a topologtcal A-module .(t.e.
A x M--7 k! is continuous). If f: M -+ N ts any A-module homomorphtsm,
{An} surjective => dA surjective, then f(a"M) = a"f(M) s a" N, and therefore f !s con,_tinuous (with respect to
the a-topologies on M and N) and so defines/: M --7 N.
but this is clear because to show dA surjective we have only to solve inductively ,
the equations Examples. I) A = ·k [x ], where k is a field a~d x an indeterminatc; a = (x).
Then Á = k[[x]], the ring of formal power senes. . .
Xn - On+l(Xn+l) = On 2) A = z a = (p) p prime. Then A is the ring of p-ad1c mtegers. lts
for Xn E An, given an E An. • '
elements are infinite '
series L.:... o anp", o ~ an ~ p - 1. We have p" --7 Oas
n -7 oo.
Remark. The group Coker dAis usually denoted by lím 1 An, smce · 1't IS
· a derived
(
functor in the sense of homological algebra.
FlLTRATIONS
Coro/ltuy 10.3. Let O -+ G' -+ G!!...:,. G" --+ O be an exact sequence a:~
groups. Let G have the topo/ogy defined by a sequence {Gn} o~ subgrouptG }. The a-~opology of an A-module M was defined by ta
ki
:gthe submodules a" M
of defining the same
as bas1c neíghborhoods of O, but there are other way =' ... where the M,.
give G', G" the induced topo/ogies, i,e. by the sequences {Gn n Gn}, n
Then topology. An (infinite) chain M= M~ 2 M1 · · · ~e!ted by (M ). lt is an
2 1
11
are submodules of M, is called ajiltratlon of M, and . n if aAf :::: Mn+t for
«-filtration if nM e M c-
10
r all n and a stable a-.filtratiO "
n - n+l ' ' ' n
ls exact. all sufficiently large n. Thus (n" M) is a stable a-filtrauo · •
J06 COMPLETlONS
G RADED RINGS ANO
Lemma 10.6. /f (M")' (M~) are stab/e a-filtrations of tv/ Mooull:s 107
bounded di.fference: that is, there e.xists m 1 ínteger 11 suc·l • lhcn lhey hav A be a ring (not graded), a an ideal of A Th
, o · 1 11wt 111 e .
Lct s· ·¡ 1 ·f M l.S an A-moduleand
• :::: Et>:-o a". 1~1 ar y, 1
· en we ca t
. n orm a &raded
and Mn+no S Mnfor al/n ~ O. Hence al/ stable a-:filtrar · " +no S M'
tons derer11u,e nnS ;1 M* =' EBn Nln ts a gradcd A*-module s· M.,.,ts an a-filtration f
same topo/ogy on Af, namely the a-topology. ·
lhe~ J,f thcn . . fi . 1 ' mee amM e M o
, If A is Noethcnan, a ts. n1tc Y gencratcd, say b x n - '~~+r.·
Proof. Enough to take Al~ = a"NI. Sincc ald e M ro ~] and is Noethcnan by (7 .6). y 1> • • • • xr; then A' ::::
r " - n+1 r a 11 11 w 1
a"M S M n; a so aAfn = Mn + 1 10r all n ~ n0 say, hencc /vi
1 ' e lave A[x1' • • • ' ' ' •
S n"M. • n+no ::::: O"Af ,ma 10.8. Let A be a Noetherian ring M afi .,
~o L en . .r ·t ' m e1Y-generated A
J ,( ) an o--filtra/1011 O; M. Tllen the following arn equiVa
. 1ent: ·module,
( Jt'J n . <;
i) M* fs a finitely-generated A*-module,·
GRADED RINGS AND MODULES
ii) The filtration (Mn) is stable.
A graded ring is a ring A together with a family (A")"> 0 of subgrou s .r Each Mn is finitely gcnerated, hcncc so is cach Q = r.:L\n ..
additive group of A, such that A = '+':.. 0 An and AmA e A r0 r all p of the PrOO;· • b ut no t ('m genera l) an A *-submodule. n. Howe
W r•o Mr: th1s IS a
w 1\ - m + JI m 11 > o
1\
b Oli'P of M .
Thus A 0 is a subring of A, and each An is an A0 -module. ' ... · su g ver, It generates
one,' namely
Examplc. A = k[x11 • •• , x,], An = set of all homogeneous polynomials M: = M o EE>···EE> Mn Et) aMn Ef> a2 M 11 Ei) ... ffi a'M11 EiJ ...
degree n. of
Since Qn is finitely *generated as an A-module,
d. .
M:
is finitely generated
as an
.rr A is a graded ring, a graded A-module is an A-module kf togcther with a r
A*-module. The M n 10rrn an aseen mg cham, whose union is M•. Since A* is
family(Mn)n >o ofsubgroupsof ~such thatM = E8:-o AlnandArnlvln s M m +n Noetherian, M* is finitely generated as an .A*-module <> the chain stops, i.e.,
for all m, n ~ O. Thus each Mn IS an Ao-rnodule. An elemcnt x of N[ is ¡, 01110 _ M* = #f't0 for sorne no <=> M no + r = o'M no for all r ~ O<> the filtration ¡5
geneous if x E Mn for sorne n (n = degree of x). Any clcmcnt y E Nf can be stable. •
written uniquely as a finite surn 2n Yn. where Yn E Ñln for all n ;;?: O, and all but a
finite number of the Yn are O. The non-zero cornponents Yn are callcd the Proposition 10.9. (Artin- Rees lemma). Let A be a Noetherian ring, a
lwmogeneous components of y. an ideal in A, J\!1 a finitely-generated A-module, (M11) a stable a-filtration
If M, N are graded A-modules, a homomorphism ofgraded A-modules is an of .M. If A1 ' is a submodule of M, then (M' n Mn) is a stable a-filtration
A-module homomorphism f: J\!1 ->- N such thatf(Mn) S Nn for all n ~ O. of M '.
If A is a graded riog, Jet A+ = EF>n>o An. A + is an ideal of A. Proof. We have a(M ' fllvfn) S aM ' n oM" S M' n .Mn+ 11 hence(M' fl M.,.,)
is an a-filtration. Hence it defines a graded A*-module which is a submodule of
Proposition 10.7. The following are equivalen!, for a graded ring A: M* aod therefore finitely generated (since A* is Noetherian). Now use (10.8). •
i) A is a Noetlzerian ring ,- Taking M" = an M we obtain what is usually known as the Artin-Rees
ii) A 0 is Noetherian and A is finitely generated as an A 0 -algebra. lemma:
Proof. i) => ii). A 0 ~ A / A+' hence is Noetherian. A + is an ideal in A, hence is Corollary 10.10. There exists an integer k suclz rhat
finitely generated, say by X¡, . .. , Xs, which we may take to be homogeneous
elements of A, of degrees k¡, ... , k, say (all > 0). Let A' be the subring of A (a" M ) n NI ' = o"-k ((okM) n M ')
generated by x 11 •• • , x, over A 0 • We shall show that An f;; A' for all n ~ O, foral! n ~ k . •
by induction on n. This is certainly true for n = O. Let n > O and let Y E An· . h t he e1ementary lemma (10.6)
Since y E A+, y is a linear combination of the X¡, say y = 2:f .. 1 a,x" wh~re On the other hand, combining (10.9) w1t
a, E An-k, (conventionally Am = O if m < 0). Since each k 1 > O, th.e ind~c~ve we obtain the really significant version:
hypothesis shows that each a, is a polynomial in the x's with coeffictents tn o· 'd 1 M afinitely-generated
1
Theorem 10.11. Let A be a Noetherian ring, a an ea' . "M' and
Hence the same is true of y, and therefore y E A'. Hence An f;; A' and therefore A-module and M ' a submodule of M. ,.,., the filtratwns a
J. ,zen ¡ of M'
A =A'. (a" M) n M' Jzave bounded difference. In particular the a-topo ogy
ii) => i): by Hilbert's basis theorem (7.6). • coincides wit/z tire topology induced by the a-topology of M. •
s•
J08 COMPLETIONS G RADED RINGs
ANO MODULEs •
Remark. In this chapter we sball apply the las t part of ( 10 11) 'tions (1 0. 12) and ( 10. 13) together a 109
propo
Sl ssert that h
. · concc ·
topologies. However, m the next chapter the strongcr result abou t b rnJng M js exact on the. category of finitely-generated A t e functor M ~
ditferences wíll be needed. ounded ) ®.. t ·an). As shown tn Chapter 2 this proves . -modules (when A .
r-~oetherJ . · 1s
osition 10.14. lf A zs a Noetherian ring a 'd
As a first application of ( 10.11) ~ve combine it with ( 10.3) to gct the .
A
ts not exact:
O -? M ' -? Jvf ->- Atf " - > O the g 't ly-generated modules. two functors coincide
on finl e
be an exact sequence offinitely-generated modules o ver a Noetheria11 rmg
• We Procecd now to study thc ri ng Á in more detat.1· F'1rst sorne el
A 'tions · ementary
Let a be an ideal of A, then tlze sequence of a-adic completions o propost ·
,.. 1 """' ,..
Pro'Position 10.15. If A is Noetherian, Á its a-adic 1 .
O - >- M -? M --+ M " --+ O .. .. comp etiOn, tiren
i) a == A a ~ A ® Á a ;
is exact. •
ii) (a")"' == (á)";
Since we bave a natural homomorphism A --+ Á we can regard Á as ..Ul') a"fa" + l = an¡an + l '.
/"V
In general, for arbitrary A and M, this is neither injective nor surjective, but we whose image is Aa, is an isomorphism. This proves i). Now apply i) to a11 and
do have: we deduce that
• (a")"' = Aa" = (Áa)" by (1.18)
Proposition 10.13. For any ringA, if Jvf isfinitely -generated, Á ®A M -r M
= (á)" by i).
is surjective. IJ, moreover, A is Noetherian tlzen Á ®,1 M --+ M is an iso-
morphism. Applying (10.4) we now deduce
Proof Using (10.3) or otherwise it is clear that a-adic completion commutes A fa" ~ Á/o"
witb finite direct sums. Hence if F ~ A" we ha ve A ® .. F ~ f . Now assume M
is fin itely generated so that we have an exact sequence from which iii) follows by taking quotients. By ii) and (10.5) we see that Á
is complete for its a-topology. Hence for any X E a
O --+ N-? F -? M -+ O.
(1 - x ) - 1 = l + x + x2 + · · ·
This gives rise to the commutative diagram o
Proof. By (10.15) iii) we have Aj{it ~ Afm hence Á/ lit is a field and n! a 50 15
~n which ~he to~ li~e is exact (by (2.18)). By (10.3) 8 is surjective. Since fi is an
150 ~?rphJsm thts tmplies that a is surjective, proving the first part of the pro- ma~imal ideal. By (10.15) iv) it foll;vs tha~ 111 is the Jacobson radical of A and
posJtton. Assume now that A is Noetherian then N is also finitely generated so 15 the unique maximal idea l. Thus Á is a local ring. •
50 t~at 'Y is surjective and, by (10.12), the bot;om line is exact. A little diag.ram Th · 1 t1·0 n is answered by
e tmportant question of how much we lose on comp e
chasmg now proves that a is injective and so an isomorphism. • Krull's Theorem:
110 COMPLETlONS
T HE ASSOClATEO O
RADEo lUNa
Theorem 10.17. Let A be a Noetlterian ring, a an ideal M fi . BY (1.9) every element of l + a is a u 't 111
~ · • a mue/y
A-modt1e and M the a-comp/etzon of M . Then the kernel E == ~generated proof. . . m· •
M~ M consists of t/zose X E M annihilmed by soma elen lellt ofn~~-.1
l +
Q'fiA{ or
'J
A
nS
a particularly ¡mportant spcc~al case of (10 19)
· we have:
. Proof Since E is the intersection of all neighborhoods of o E Nf n. o/lary 10.20. L et A be a Noetherian local . .
1
induced on it is trivial, i.e., E is the only neighborhood of o E E 'othe topology eo'te/y-genera ted A -moci u1e. Then the m-to rmg,
r
1
m lis .
rnaxtrnal ideal M
induced topology on E coincides with its a-topology. Since a E is · ~ (lO. lt) the '"
fiparticular 1 .r . 'PO ogy of M • H
tlze m -topo ogy OJ A ts Hausdorjf. • ts
. , a
ausdorff. In
in the a-topology it follows that aE = E. Since Al is finitel y-gc a nctghborhooct
. . fi . . nerated a11 d . We can restate (10.20) slightly differently if
Noethenan, E rs aJso mtely-generated and so we can apply (2.5) and' A ts fA is just any ideal containcd between m wde recaU that an m-primary
from aE = E that (I - a) E = O for sorne a E a. The converse ¡ _deduce idea 1 o ( 1O.20) 1mp
. 1' an sornef power ln~'~ (use (4.2)
(1 - a)x = O, then s obvtous: if !4)). Thus 1es that the intersectio
an d (7. f A . N l . . n o a1l m-pri .
of A is zero. 1 now . rs any . oet 1cna~ nng, P a prime ideal mary ldcals
X = ax = a 2 x = o o o En a"/YJ =
nal
a)
E. •
ver
sion of (10.20) to the local rmg A p. L1fting back toA d '.we can apply this
(4 S) b t . an usmg the 0 t
rrespondence . e ween ~-pn mary ideals of A and . . ne- o-one
co \'A ) we deduce · m-pnmary ldeals of A
Remarks. 1) If S is the multiplicatively closed set 1 + a, then (IO.l 7) asserts (where 111 = ,., P • ll
that
Corollary 10.21. LelA be a Noerherian ring, p a prime 1'd 10.r
A -+ A and A -+ s- 1 A · OJ.r a 11 'p-pnmary
intersectzon · r'dea¡s of A is the kernel oif Aea 1 A. Then the
. -+ All. •
have the same kernel. Moreover for any a E a
(1 - a) - l = 1 +a + cc2 + ... THE ASSOCIATED GRADED RING
. ~
converges In A,
_
so that
.
every element of •S becomes a unit in A· By tl1e u ntversa
·
1
Let A be a ring and a an ideal of A . Define
property of ~ ~ thrs mea~s ~ha_t ~he~e rs a natural homomorphism s - IA ~Á
1
a)
and (10.17) 1mplres that thrs rs mJective. Thus s- 1A can be identified with G (A)(= Gn(A)) = EB a"f an +l (a0 = A).
n .. o
subring of A. a
2) Krull's Theorem (10.17) may be false if A is no t Noetherian. Let A be the This is a graded ring, in which the multiplication is defined as follows:
ring of aU C«J functions on the realline, and let a be the ideal of all f which van- For each Xn E a", let Xn denote the image of X 11 in a"fa" +1; define .Xm.X" to be
ish at the origin (a is maximal since A j a ~ R). In fact a is generated by the XmX"' i.e., the image of XmXn in a m+ n¡ a m+n+l; check that XmX11 does not depend
identity function X, and n:. .
1 a" is the set of all 1 E A, all of whose derivatives on the particular representatives chosen.
Similarly, if Mis an A-module and (Mn) is an a-filtration of M, define
vanish at the origin. On the other hand 1 is annihilated by sorne element 1 + a
(a E a) if and only iflvanishes identically in sorne neighborhood ofO. The well- a)
known function e- 1/x'J, which is not identically zero near O, hut has vanishing G(M) = EB
n=O
Mn/Mn +l
derivatives at O, then shows that the kernels of
.... which is a graded G(A)-module in a natural way. Let G11(M) denote Mn/M11+l·
A-+ A andA - >- s- 1A (S = 1 + a)
Proposition 10.22. Let A be a Noetherian ring, a an ideal of A. Then
do not coincide. Thus A is not Noetherian . i) Ga(A) is Noetherian;
Krull's Theorem has many corolJaries: ii) Ga(A) and Ga(A) are isomorphic as graded rings;
Corollllry 10.18. Let A be a Noetherian doma in, n ::/; ( l) all ideal of A. iii) if Mis afinitely-generated A-module and (Mn) is a stable a-filtration of M,
Then n
an =O. then G(M) is a finitely-generated graded Ga(A)-module.
Proof. 1 + a contains no zero-divisors . ~roof. i) Since A is Noetherian a is finitely generated, say by X¡, ·. · · ' x,. ~t
• x, be the image of x, in afa2 'then G(A) = (A/a)[i1, ... , i,]. Smce Ala IS
Coro/huy 10.19. Let A be a Noetherian ring, a an ideal olA contained in the Noethenan,
· G(A) is Noetherian' by the Hilbert bas1s · theorem.
Jacobson radical and /et Al be a finitely-generated A-module. Then the
a-topo/ogy ol M is Hausdorjf, i.e. a" M = O. n ii) o."fa" + 1 ~ ñ"fa" + 1 by (10.15) iii).
EXER.ClSES
112 COMPLETIONS
cd groups, and G(cp) : G( F ) _, G(M) is a h 113
iii) There cxists no such that M no+ , = a'J\tf no for all r ~ O, hcnce G(M ) is of filttr BY construction it is surjective. Hence b tmom~~phism of G( )
genera ted by
ú:\ G (M ). Each Gn(A1) = JWn/ A1n+ 1 ts Nocthcrian
w n <no n ' and
mod~;:; 00w the diagram Y · ) u) ~ is surjecti~
10 23
~
'h'l t d by 0 hence is a finitely-generated A/ n-module, hcnce \I/n(no GnH/
(l ) coost o
F --r M
.
3flnt tae J
t GnC~) t an+ 1 t an
We can now deduce the result we are after :
•
O-r Bn/Bn+l ~ B/Bn+l -r B/Bn - >- O.
Theorem 10.26. Jf A is a Noerherian ring, a 011 ideal of A, then the a-
This gives the exact sequence completion Á of A is Noetherian.
O~ Ker Gn(4>) - >- Ker an +1 -r Ker an ~ Coker Gn(4>) -r Coker an+l Proof. By (10.22) we know that
- >- Coker an ~ O.
Ga(A) == Ga(Á)
From ~.his we see, by induction on n, that Ker an = O (case i)) or Coker a" = O is Noetherian. N ow apply (10.25) to the complete ring Á, taking lv/ = A
(case u)}. Moreover in case ii) we also ha ve Ker ex 1 - >- Ker ex surjective.
Taking the inverse limit ofthe homomorphisms o:" and ;;plying (10.2) the lemma (filtered by an, and so Hausdorff). •
follows. • Corol/a,.y 10.27. Jf A is a Noerlzerian ring, tire power series ri11g B =
A[[x¡, . . . , Xn]] in n variables is Noetherian. In particular k[[X¡, · · ., x,J]
. We ca~ now f~rm a resu!t which is a partial converse of (10.22) iii) and is the
mam step 10 showmg that A is Noetherian. (k afiele/) is Noetlzerian. . .
. 'lb b · tl1eorem and B 1s 1ts
p roposl~lon
. . JO .24. Let A be a ring, a an ideal of A, ManA-module, (MJ an Proof. A[x 11 • •• , xn] is Noethenan by the Ht ert asts •
completion for the (x 1 , . . . , xn)-adic topology. •
a-fi/tratwn .of.M. Suppose that A is complete in the o-topology and that Mis
~a;,;do:! m 11~ filtration topology (i.e. that nn
M" = O). Suppose also that
( ) a fimtely-generated G(A)-module. Then M is a .finitely-generated EXERCISES
A-module. tl-1
1 Le · ps givcn by ctn(l) == P '
Phroof. Pick a finite set of generators of G(M) and split them up into their · t a": Z /pZ -~ Z /p"Z be the injection of abehan grou A .1 countable direct
and let a: A -+ B be the direct sum of all the etn (whcre ~ !;) ShoWthat the
therefore th ·compo nen tS, say ~~
omogeneous e (1 ~ i ~ v) where
' e, ·
has degree say n(i), and lS 1
sum of copies of Z /pZ and B is the direct sum of th~ Pf A ·for the topology
a-filtration e. •ma~ of say x, E Mn<t>· Let F 1 be the module A with the stable °
P·adic completion of A' is just A but that thc complettOn f the Z/pZ. Deduce
generator 1g~;:: : :.~ = ak +nm and put F = EB¡-1F'. Then mapping the . d . h d' ect producto
tn uced from the p-adic topology on B JS t e tr tl catcgorYof a
11 z-
e to x, defines a homomorphism th · f ctor on l C
at P·adtc completion is not a right-exact un
cp: F ~M modules.
114 COMPLETJONS
l lS
. A" is flat over A, it is enough to show th t
·se 1 Jet A - a- 1 (p"B), and consider thc exact scq uencc [S•nce .a
l. 1n E xerc1 , " -
O~ An A ~ A/An -+O. M ~ !Vf injcctive for all finitely generated .
-> M ~ A IS Za . .
1 e (1 0.19) and Excrcise 6.] nskt;
Show that Hm is not right exact, and compute jim A n. noW uS
~
A be the local ring of the origin in en (i.e. th .
J. Let A be a Noetherian ring, o a~ ideal and M a finitely-generated A-modu¡ s. L; G C(zh ... , Zn) with g(O) -:¡:. 0), let B be thc' rin~ r~~g of all rational functio
1
Using Krull's Tbeorem and Exerc1se 14 of Chapter 3, prove that e. flh'cll converge in sorne neighborhood of the 0 • . powcr series in z1 · ns
\'1 1 • • ngtn, and let e b ' • • . ' Zr¡
n a"M
CID
n Ker (M ~
e rmal powcr senes m Z¡, ••. , Zn, so that A e B e e e the ring of
· Sh~w that Bis a local
= Mm),
n•l m2n ?ng and that its completion for tbe maximal ideal t 1
rl . l . Opo ogy IS e As . .
where m runs over aH maximal ideals containing o. B is Noethenan, pro ve t 1at B ts A-flat. [Use Cha ter . ~ummg that
Exercise 7 above.] P 3, Exerctse 17, and
Deduce that
M = o <=> Supp (M) n V(a) = 0 (in Spec (A)). Let A be a local ring, m its maximal ideal. Assume that A 15
. .
9· For any polynomial/(x) E A[x ], let](x) E (A/m)[x] denot .ntt-adtcal~y complete.
[fbe reader should tbink of M" as the "Taylor expansion" of M transversal t0 ' 1
Prove Hensel s emma : 1
·r f( ) . . e l s reductton m0 d
x ts mo me of degree n and ·r th . ·.m.
the subscheme V(a): the abo ve result then shows that M is determined · monic polynomials g(x), fi(x) E (A /m)[x] of degrees / , ~re ex~st copnme
neighborhood of V(a) by its Taylor expansion.] m a l 'f - ( ) fi( ) b . ' r Wtlh /(x) =
g(x)fi(x), then we can 1 t g x , 1 x ack to monte polynomials (x) 11( ·
sueh that f(x) = g(x)lz(x). g ' x) e A[x]
4. Let A be a Noetherian ring, a an ideal in A, and A the n-adie completion. For
any x E A, let x be the image of x in A. Show that [Assume inductively ;hat we ha ve constructed g"(x), hk(x) E A[x] such that
gk(x).hk(x) - f(x) E m _A [x]. Then use the fact that since g(x) and fi(x) are
x nota zero-divisor in A => ..'( not a zero-divisor in A. copnme we can find ap(x), Óp(x), of degrees ~ 11 - r, r respectively, such that
xP = a (x)gk(x) + Óp(x)lik(x), where p is any integer such that 1 ~ P ~
Does this imply tbat 9 """ """ n.
Finally, use the completeness of A to show that the sequences g~c(x), /~¡,(x)
A is an integral domain => A is an integral domain ? converge to the required g(x), h(x).]
[Apply the exactness of completion to thc sequence O --r A ~> A.] 10. i) With the notation of Exercise 9, deduce from Hensel's lemma that if /(x)
has a simple root a E A/m, then f(x) has a simple root a e A such that
S. Let A be a Noetherian ring and Jet a, ó be ideals in A . If Mi s anv A-module Jet
a= a mod m.
Ma, M & denote its n-adie and ó-adic completions respectively. ·If Mis fini~ely
ii) Show that 2 is a square in the ring of 7-adic integers.
generated, prove that (Ma)b ~ Ma+f.l ,
iii) Let f(x, y) E k [x, y], where k is a field, and assume that /(0, y) has y = ao
[fake the n-adie eompletion of the exact sequence
as a simple root. Prove that there exists a formal power series y(x) =
0 -+ ómM-+ M/ómM -+ O L.:.. o a"x" such thatf(x, y(x)) = O.
(This gives the "analytic branch" of the curve f = O through tbe point (0, ao).)
and apply (10.13). Then use the isomorphism
11. Show tbat the converse of (1 0.26) is false, even if we assume that A is local and
~~ (~irn M/(a"M + fJmM)) ~ lim M /(a"M + ó"M) that A is a finitcly-generated A-module.
"' " +<-
n [Take A to be the ring of germs of Ca> functions of x at x = O, and use Borel's
and the inclusions (a + ó)2" S a" + Ó" s (a + ó)".] Theorem that every power series occurs as the Taylor expansion of sorne C"'
6· Let A be a Noetherian rin funct.ion.]
d · .
Jacobson radical fA ·r g an a a~ Ideal m A. Prove that a is contained in the 12· If A is Noetherian then A [[x . • . xnll is a faithfully ftat A-algebra. [~xpsre(ss)
1
a-topology. (A N~th . and only ~ eve~y maximal ideal of A is closed for the ' 1' ' · d se Exerctse v
an ideal contained . ~nan topologtcal nng in which the topology is defined by A ~ A[[x1, ... , Xn]] as a composition of flat extens10ns, an u
local rings and {b m(l~ ~SJXa~ob)son r~dical is called a Zariski ring. Examples are of Chapter l.]
Y · Iv) a-ad1c completions.)
1• Lct A be a Noeth · ·
enan rmg 'd ...
that Á is faithfully flat ' a an 1 eal of A, and A the n-adie completion. Prove
rin¡ (for the a-topolog~;~r A (Chapter 3, Exercise 16) if and only if A is a Zariski
HiLBERt FUNCT
IONs ll?
rem JJ.l. (Hilbert,Serrc). P (M, t)is arationa .
ThtO!D' ( 1 - t "•), where f(t) E Z(t) 1functlon in rof rhe r
11 f(t) , .. 1 • JOrm
BY induction on s, the number of generators of A
proof. tl ·s ...,eans tha tAn = O for all n > O so that A Aover Ao. Start With
..... O; ll •·-- ' === oand M '1 ·
s -- tedA o module, hencc Mn = Ofor alllarge n. Thus P(M ) . S a fimtely.
genera , t ts a polynomial
Dimension Theory . this case.
10 "'r v suppose s > O and thc theorcm true for s _ M . . .
r ,.O\ h' 1· u1t1phcat1o b
module homomorp tsm of M 11 into M he . . n Y x,
is an A- n+k,, nce tt glVes an exact
scquencc, say x.
O -r Kn --7 Mn ~ Afn+ k __,. L - >O
• • ll+.k , • (l)
One of the b
asic notions
.
in algebraic. geometry is tha t of the dimension
.
f
o a Let K == Ef)n Kn, L = EBnLn, these ~re both finttely-generated A-modules
. t This ¡5 essentlally a local not10n, and, as we shall show m this cha use K is a submodule and La quottent module of M) and both .
vane y. f d. . ti Pter, (beca h A [ , are anm-
. 3 very satisfactory theory o tmensiOn
there ts . or general
. Noethcrian 1oca 1 hilated by x, hence t ey are o x11 ... , x, _d-modules. Applying .\ to (1) we
.
nngs. The main theorem asserts. .the eqUJvalence
. of three
. ddierent dcfinitio, 1s of have by (2.11)
dimension. Two of these defi.mtJOns hav~ a ~atr1y obvwus geometrical contcnt, ' )\(Kn) - 'A(Mn) + )\(Mn+k,) - .\(Ln+k,) ===O;
but the third involving the Hllbert functJOn ts Iess conceptual. It has, howcver multiplying by t n+k, and summing with respect to n we get
many technical advantages and the whole theory becomes more strcamlined if
ooe brings it in at an early stage. (1 - t k•)P(M, t) = P(L, t) - t k•P(K, t) + g(t) (2)
After dealing with dimen.sion we give. a brie~ ac~ount of r~gula r local rings, where g(t) is a polynomial. Applying the inductive hypothesis the result now
which correspond to the not10n of non-smgulanty m algebrarc geomctry. We follows. •
establish the equivalence of three definitions of regularity. The order of 'the pole of P(M, t) at t = 1 we shall denote by d(M). It
Finally we indicate how, in the case of algebraic varieties over a field , thc provides a measure of the "size" ?f M ~relati~e to .\). In particular d(A) is
local dimensions we ha ve defined coincide with the transcendence degree of thc defined. The case when all k, = 1 1s spectally sunple:
function field.
Coro//ary 11.2. If each k 1 = 1, then foral/ sufficiently large n, ~(MJ is a
polynomial in n (with rational coefficients) of degree* d - l.
lfiLBERT FUNCfiONS Proof By (11.1) we have >..(Mn) = coefficient of 111 inf(t)·(l- t)-'. Cancel-
ing powers of (1 - t) we may assume s = d and f(l) :F O. Suppose f(t) =
Let A = Ef):O.o A" be a Noetherian graded ring. By (10.7) A 0 is a Noetherian
<r;;N k •
ring, andA is generated (asan A0-algebra) by say X¡, • •• , x$, which we may take L.k•o akt ; smce
the form "f.¡JJ.x)m¡, where jj(x) E A is homogeneous of degree 11 - r1 (and d+n-k-1 for alln ~ N.
A(Nln) =
therefore zero if n < r,). It follows that M is finitely generated as an Ao· d-l
module n 1 . . n . .
o . . . l in n with leading term
' ame Ytt JS generated by all glx)m, where gJCx) is a monomtal m the k ..
x, of total degree n _ r,. and the sum on the right-hand stde ts a polynomta •
11 1
Let ,\ be an additive function (with values in Z) on the class of all finitely· ("¿ak)n - /(d- 1)! -:1: O. • . f¡ r all
h t f(n) is an mteger o
~e:rated Áo·.modules (Chapter 2). The Poincaré series of M (with respect toA) Remarks. 1) For a polynomial f(x) to be such t a ffi . ts · e g. l.x(x + 1).
. . t r coe cten . . . , l
e generatmg function of t\(M,.), i.e., it is the power series tntegers n, it is not necessary for f to ha ve ege
111 'al is -1: also
•W 0 f the zero po}ynonu
CIO
e adopt the convention here that the degree _ _ l.
P(M, t) = 2: >..(M,.)t n E Z[[t ]]. that the binomial coefficient ( _n ) = O for n ~ O• and == 1 for n -
1
11•0
116
• •••
Propos1•1¡on
•
>~( ) . on q M) 1s denoted by
d(Af) - 1. • Xq " • XqM(n) = /( M 1q" M) (for all large 11).
We shall use (11.1 ) in the case where A.a is an A rtin ring (in particular, a
field) and >.(Af) ¡5 the /engtlt I(Nf ) of a fimtely-genera tcd Aa-module M . By If 11¡f = A, we write Xq(n) for x~(n) and call it thc character· f
of the rn-primary ideal q. 1n this case ( 11 .4) gives ts le po1ynomial
(6.9) /(M) is additive.
Example. Let A = A0 [x 1 , x,], whe:e Aa is an Artin ring and the x, are
•• • , Corollary 11.5. For al/ large n, tite lengrh f(A /q") is a 1 .
1 · 1 1 po ynonua/ X (n)
independent indeterminates. Thcn An ts a free Aa-module generatcd by the of degree ~ s, w zere s rs t re east number of generators 0¡ q. • q
Noethenan A-r~10dule annihilated by q, hence a Noetherian A/ q-module, and Let o(A) = least number of generators of an m-primary i~eal fs ~·
0
n
first link in this chain:
1" = l(M/Mn) = L l(M, _¡JM,). (1)
r •l
Proposition 11.7. o(A) ~ d(A). • th.
ii) If xh · · ·, x, generate q, the images x, of the x in qfq2 genera te G(A) . 0
f (l 1.3). Note that JS
: ~~n~/;~alg:bra, and e~ch x, has degree l. Hence by (1 /.2) we ha ve /(Mn/Mn+l)
Next we shall prove the analogue for local nngs . t the topological
Proof uses the strong version of the Artin- Rees lemma (not JUS
y, here /(n) JS a polynomial in n of degree ~ s - 1 for all large n. Part).
DJMENSION 11-IEORY
OF NOETHE!UA
120 DJMENSION THEORY • N lOCAL lUNas
o/lary 11.12. In a Noethenan ring e 121
.. Let A, m, q be as before. Let M be a .finitely-generated eorere rore tlze ser o
if .
pnme
'd
z ea/s in a
very prime id 1
Noetl . ea has .finite 1
Proposlllolf 118
Á.a non-zero-divisor in M and M' = M jxM. Then ti1 ;¡ • •
hain cond/twn. •
1enan ring . leighr d
satls.fies the d , an
· A-module, x e · e escendin
deg xr· ~ deg x~' - 1.
crllark. Likewise we may define the depth o[ P b e . g
_ xM· then N ~ M as A-modules, by virtue of the assumption ~ Js which start at "': clearly depth p = dim 'A¡y onstde~ing chains f .
=
Proof. Let N N~ q"M. Then we have exact scqucnces
on x. Let Nn -
¡dea . N h . .
'd 1 even m a oet cnan nng, may be i fi .
1 ea , .
p. But th
e depth of a .
n mte (unless th . . Pnme
o Prune
Also, ifm' is the maximal ideal of A', A'/m'" is a homomorphic image of A/m", iii) tlze least number of generators of an m-primary ideal of A.
hence I(Afm") ~ I(A'fm'") and therefore d(A) ~ d(A'). Consequently Proof (11.7), (11.10), (11.13). •
d(A'/(x')) ~ d(A) - 1 = d - l. ~xample. Let A be the polynomial ring k[xh .. . , xnl localized at the maximal
tdeal m = (xh .. . , xn)· Then Gm(A) is a polynomial ring in n indeterminates
H~n'7, .by the inductive hypothesis, the length of any chain of prime ideals in ~~d. so its Poincaré series is ( 1 - t)-". Hence, using the equivalence of (i) and
~ f(x) 15 ~ d - l. But the images of p11 • •• , p, in A'J(x') form a chain of (n) tn (11.14), we deduce that dim Am = n.
~ngth ' - 1, hence r - 1 ~ d - 1 and consequently r ~ d. Hence
drm A ~ d. • Coro/lary 11.15. dim A ~ dimk (m/m 2). .
Proof lf x, E m (1 ~ i ~ s) are such that their images in mfm forro a basts
2
Corollluy 11•11• If A is a Noetherian local ring, dim A is finite. • of this vector space then the x genera te m by (2.8); hence dimt (m/m ) =
2
¿ H 1 Let x y- denote t e
onfun~hes th.at](tl, · . . , td) (the reduction of fmod q) is in the kernel of a .
· '
IDtegers r S ~ 0 SUCh that X E a' X ¿ ar+l Y E as, Y~ a ' •' --
• , ~ , 'F ' -
tmages of x, y in G,(A), Gs(A) respectively. Then x ::f: , Y
o - ::¡: O hence xy =
'
Assume ifposslble that sorne coefficient offis a unit then/is nota zero-divisor
(cf. Chapter 1, Exercise 3). Then we have ' X· y -::/: O, hence xy # O. • .
. . 1 re precisely the dts-
d(Gq(A)} ~ d((A/q}[tll ·· .,Id]/(])) becausefe Ker {ce} Hence by (9.2) the regular local rings of dtmenston a
= d((Af q)[th · · ·, ldl) - 1 by (11.3) crete valuation rings. . . integrally closed
= d - 1 by the example following {11.3). It can also be shown that if A is a local nng and Gm(J1) lS an ular local ring
Btraudt"dt(Gq(A)) = d by the main theorem (1 1.14) This gives the required con- integral domain then A is integrally closed. It follows that a ~eg of dun ' ension
· •
18 , · 1 1domams ·
le 1on. • · tntegrally closed · but there are integrally closed oca
>1 ,
Which are not regular .
•
•
EXER.ClSES
124 DIMENSION TtfEORY , rna ]},16. Let B S A be integral doma · . 125
Le b . ms wu}z B ·
• • ,¿ L A be a Noetherian local ring. Tlzen A is regular if and . tegral over B . Let m e a max zma/ ideal orA 1 Integral/y closed d
111 d d' A d' , anc le t n an A
Propos111011 11.2.,. e1 'J
zs. n1aximal an tm m = tm Bn. = tn " B· -rh
1 , en n
if Á is regular. ~· N h ·
1
on Y' ) d (ll 19) we know that A 1s a oet cnan local ri ng if. This is an easy consequence of the result5 f
Proof. By (10..16), (~ · ~ ~ and. with 1ñ as maximal ideal. Now use (10.22)
02
5 proo. · al'="Y (5.8). Next if
rnaxltn l:l
Chapter 5. F'
trst n is
°
of the same d¡menswn) ,..., G· (A} and the result follows. •
which asserts that Gm(A = m ... . • m :::::~ q 1 :::::~ q2 ::::;, • • • => qd
.vhat we have said above that A ts also an mtegral
ks l) It foUows firom ' . strict chain of primes in A, its intersection with B 15
. b (1)
Remar · . a1l aking this means that (loca11Y) JS ~ Y(5.9) a strict chain of
domain. Geometnc . y spe pnrnes
non-singularity => analytic irreducibility n ::::> lJ 1 ::::;, P2 ::::> • • • => p
• • d• (2)
. 1 point there is only one analytic "branch,.
or that, at a non-smgu ar ' . . . This proves d1m Bn ~ dtm A m. Conversely given the strict h .
(S.16), lift this to a chain (1) (necessarily strict): thus di m ~ mat~""" (d ~1mweBn.
2
· fi Id k mapping Isomorphtcally onto A / m (the geometnc can, by
2) If A(col ta)L~S a r:s
0
22
1
that Á ¡5 a formal power series ring over k in d in-
case) then 1 · IIDP 1
• f · 1 · We can now proceed to :
determinates. Thus the completions .of local ~mgs o non-smgu ar pomts on
d-dimensional varieties over k are all tsomorphtc. Proor of (11 .25). By the Normalization Lemma (Chapter 5 Exerc· l6)
'J • • ' tse , we can
find a polynomt.al. nng B = k[x¡, . : . , x 4 ] contained in A(V) such that d =
· Let A - k[" "" ] (k any fieid x 1 independent indeterminates);
Examp1e. - "'~ h . . . ' .~ " . ' " . . dim Vand A(V) ts mtegral o ver B. St?ce B is integrally closed (remark following
let m = (x1 , • • • , x"). Then Am (the local nn~ of .affi~e spac~ k at the ongm) (5.12)) we can apply (11.26) and thts reduces our task to proving (11.25) for
¡5 a regular local ring: for Gm(A) is a polynomml nng m n vanables. the ring B, i.e. for affine space. But any point of affine space can be taken as
the origin of coordina tes and, as we have already seen, k[x¡, ... , xd] Iocalized
TRANSCENDENTAL DIMENSION at the maximal ideal (x h . . . , xd) is a local ring of dimension d. •
We shall conclude this brief treatment of dimension theory by showing how the Corollary 11.27. For every max imal ideal m of A( V) we have
dimension of local rings connects up with the dimension of varieties defined
dim A(V) = dim A(V)m.
classically in terms of the function field.
Assume for simplicity that.k is an algebraically closed field and let V be an Proof. By definition we have dim A( V) = supmdim A(V)sn. But by (11 .25)'
irreducible affine variety over k. Thus ~he coordinate ring A(V) is of the form all A(V)m have the same dimension. •
A( V) = k[x11 ••• , Xn]/p
where ~ is a prime ideal. The field of fractions of the integral domain A( V) is EXERCISES
called the field of rational functions on Vand is denoted by k( V) . It is a finitely·
generated extension of k and so has a finite transcendence degree over k-the l. Let fE k[x1, . . . , xn ] be an irreducible polynomial over an aJgebraically dos~~
maximum number of aJgebraically independent elements. This number is field k. A point P on the variety f(x) = O is non-singular <=> not all the partia
de~ned to be the dimension of V. Now recall that, by the Nullstellensatz, the derivatives offox, vanish at P. Let A = k[xiJ . .. , Xn ]f(f), and let m .be ~e
pomts of V correspond bijectively with the maximal ideals of A( V). lf P is a maximal ideal of A corresponding to the point P. Prove that Pis non-smgu ar
point with maximal ideal m we shall call dim A(V)m the local dimension of V Am is a regular local ring.
<=>
at P. We propose to prove [By (11.18) we have dim Am = n - l. Now
nt/m ~ (X¡, . . . , Xn)/(X¡, . · ·, Xn)
2
2 + (/)
TMorem 11.25. For any irreducible variety V over k the local dimension of V
at any point is equal to dim v. and has dimension " - 1 if and only if / rF (Xtt · · ·, xnr~.]
2 I the homomorphism
Remark. .we already know by (11.21) that dim V ~ dim A for all m. The ' n 01.21) assume that A is complete. Prove that . d that A is a
proble~ 15 to prove the opposite inequality and for this mpurpose the rnain k[[~ 1 , · · ·, td]] ~ A given by t, ~ x, (1 ~ i ~ d) is injectlvc an
l
emma .1s: ' fimtely-generated module over k[[t 1 , ••• , Id]]. [Use <10·24).]
DJWENSION THEOR y
126
"e) t 0 non-aJgebraically-closed fields. [If k is thc algcbraic 1
3• Extend (1 J,¿,J
r.[ mtegra 1 over k [x1, ... , Xn 11
. .
x 115 •
e asure
of k, thcn " Xh • • ·' • n
f Noetherian domain of infinite dimension (Nagata). Lct k b
1
4. An cxamdJpteAo a k[x x" . , Xn, • •• ] be a polynomial ringo ver k in a cou t be a
&Id an e = 1t "' • • • n a · !y
'ndeterminates. Let m1, m 2, . . . be an mcrcasing seque
¡nfinitc sct of J r . nce or
. i t gers such that m, + 1 - m, > m, - m, - t .or all ' > l. Lct h
pos1.tivc ne · f 1 . Index
and Jet s be the comp ement m A o t 1c umon of thc ideal ,
't-'t :::::
) 1
(X•+l•···•XIIIt+l d h r h S ' )' . Spt·
1
Each .t.lr is 3 prime ideal an t ere. ore t e se~ •s m u hplJcativcly c1oscd.
. S -1 A is Noetherian by Chapter 7, Exerc1se 9. Eaeh S - 1 p1 has he'
Thc nng . _1 1g11t
equal to mr+ t - m., hence dJm S A = co.
S. Reformulate (1 1. 1) in terms of the Grothendieck group K(A 0 ) (Chaptcr 7• Additive function, 23 Gcnerators of a module 20
Adie topology, 105 Grothendieck group, 88•
Excrcisc 25).
Algebra, 30
6• Let A be a ring (not necessarily Noetherian). Prove that homomorphism, 30 Height, 120
1 + dim A ~ dim A[x] ~ 1 + 2 dim A .
Annihilator ideal, 8, 19 Hcnsel's lcmma, 115
Artinian, module, 74 Hilbert basis theorem 81
[Let/: A- A(x] be the embedding and consider the fiber off* : Spec (A[x]) -.. ring, 76 Hilbert function, 118 '
Spec (A) over a prime ideal p of A. This fiber can be identified with the spectrum Artin-Rees Jemma, 107 Hilbert Nullstellensatz, 67, 69, 82, 85
of k ®.. A[x] ~ k[x], where k is the residue field at ~ (Chaptcr 3, Exercise 21),
and dim k(x] = l. Now use Exercise 7(ii) of Chapter 4.] Boolean ring, 11 Ideal(s), 2
coprime, 7
7. Let A be a Noethcrian ring. Then Chain conditions, 74 decomposable, 52
Chain of submodules, 76 direct product of, 7
dim A[x] = 1 + dim A, Cokernel, 19 fractional, 96
and hencc, by induction on n, Complete, 105 generated by, 5
Completion, 102 intersection of, 6
dim A[x1, . . . , Xn] = n + di m A. Composition series, 76 invertible, 96
[Let 4' be a prime ideal of height m in A. Then there exist a11 ••• , am e p sueh that Constructible topology, 48 maximal, 3
~ is a minimal prime ideal belonging to the ideal a = (a 1 , ••• , am). By Exercisc 7
Contraction, 9 primary, 50
ofChapter 4, ~[x] is a minimal prime ideal of a[x] and therefore height ~[x] ~m. Dedekind domain, 95 prime, 3
On the other hand, a chain of prime ideals ~o e ~ 1 e . . . e ~m = lJ givcs rise Dimension, 90 principal, 11
to a chain 4'o[x] e ··· e ~m[x] = ~[x], hence height ~[x] ~ m. Hence height Discrete valuation ring, 94 product of, 6
~[x] = height ~. Now use the argument of Exercise 6.] Domain, integral, 2 quotient, 8
principal ideal, 5 sum of, 6
Image, 2
Exact sequence, 22 Integral, A-algebra, 60
Extension of an ideal, 9 closure, 60
element, 59
Field, 3
1ntegrally closed domain, 60
residue, 4, 43
Filtration, 105 Kernel, 2, 18
Finite, A-algebra, 30
type, 30 Length, 76
Finitely-generated 30 Limit, direct, 32
Flat, 29 ' inverse, 103
faithfully, 29 Local ring, 4
Fractions, ring of, 36 Localization, 38
127
J28 JNDEX
Ring, 1
Modular Jaw, 6
absolute1y flat, 35
Module(s), l 1 Boolean, 11
direct product of, 20
discretc va1uatio n, 94
direct sum of, 20
graded, 106
faithfuJ. 20
finiteJy.generated, 20 homomorphism, 2
local, 4
flat~ 29
of fractions, 36
free, 21 quotient, 2
graded, 106
homomorphism, 18 semi-local, 4
sub, 2
product of, 19
valuation, 65
quotient, 18
sub, 18
sum of, 19 Satu ratcd , 44
Multiplicatively closed set. 36 Scalars, extcnsion of, 27
restriction of, 27
Nakayama's Jemma, 21 Spectrum, maximal, 14
Nilpotent, 2 prime, 12
Noetherian, module, 74 Support, 46
ring, 76 Symbolic powcr, 56
Normalization lemma, 69
Tensor product, of algebras, 30
Parameters, system of, 122 of mod ules, 24
Poincaré series, 116 Torsion, element, 45
Prirnary, decomposition, 51 submodule, 45
ideal, 50
Prime ideal(s), 3 UniL, 2
associated, 52
embcdded, 52 Valuation ring, 65
isolated, 52 Yarieties, a ffinc algebraic, 15