You are on page 1of 10

Chatper 3: Rings and Modules of Fractions 3.

The set S0 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. Show that every minimal prime ideal of A is contained in D. The ring S1 A is called the total ring of fractions of A. Prove that 0 i) S0 is the largest multiplicatively closed subset of A for which the homomorphism A S1 A is injective. 0 ii) Every element in S1 A is either a zero-divisor or a unit. 0 iii) Every ring in which every non-unit is a zero-divisor is equal to its total ring of fractions (i.e. A S1 A is a bijection). 0

First we show any minimal prime ideal of A is contained in D. Let p be a minimal prime ideal of A and assume to the contrary that p D contains an element x. Let Ap be the localization of A at p; since 0Ap is the intersection of all minimal prime ideals over (0) we get 0Ap = pAp , which is the maximal ideal of Ap , because p is minimal and localization cuts out any prime ideal not contained in p. But then x/1 pAp = 0Ap so x/1 is nilpotent. By Corollary 3.12, x is in the nilradical of A which means x is a zero-divisor, so x D. Contradiction! i) Let S be a multiplicatively closed subset of A. The kernel of A S1 A is {a A | a/1 = 0} = {a A | as = 0 for some s S}. This is only trivial when S has no zero-divisors which implies S S0 ; hence S0 is the largest such set. ii) Let a/b S1 A. If a S0 , then a/b is a unit. If a0 S0 , then a is a zero-divisor by 0 denition of S0 , so su = 0 for some non-zero u A. Thus a/bu/1 = 0/b = 0. To see that u/1 = 0, we assume otherwise that u/1 = 0 which implies us = 0 for some s S0 ; but this means that s is zero or a zero-divisor which is a contradiction. Thus every unit of S1 A is either a unit or a zero-divisor. 0 iii) Assume that A is a ring in which every element is either a unit or a zero-divisor. By i), the map A S1 A is injective. Lets show that it is surjective making it a bijection. 0 Pick any a/b S1 . Since b S0 it is not a zero-divisor and is thus a unit. Thus we 0 have (a ab1 b) = a a = 0, so a/b = (ab1 )/1. Hence a/b is the image of ab1 , meaning A S1 A is a bijection, so A = S1 . 0 0

Chapter 4: Primary Decomposition

This section concerns the decomposition of an ideal into primary idealswhatever that means. It is a generalizaton of the prime factorization of integers and is also a powerful tool in algebraic geometry. Although the modern treatment of algebraic geometry relies on localization and less on primary decomposition, it is still of interest in itself. In unique factorization domains, we have unique factorization of elements. There is also a unique factorization of ideals in Noetherian rings (knowledge of Noetherian rings is not needed for this chapter). A prime ideal is, in a sense, a generalization of prime numbers. Further, a primary ideal is a corresponding generalization of prime numbers. It is the next best thing to have if we cannot have primeness. An ideal q is called primary if q = A and xy q = either x q or yn q for some n > 0. Equivalently, q is primary if A/q = 0 and every zero-divisor in A/q is nilpotent. We have every prime ideal is a primary ideal and the contraction of a primary ideal is a primary ideal, because if f : A B and q is primary then A/f1 (q) = B/q. Proposition (4.1). Let q be a primary ideal in a ring A. Then q is the smallest prime ideal containing q. Proof. We know q is the intersection of the prime ideals containing qby 1.14. Thus it is enough to show that q is prime. Let xy q; then (xy)m q for some m > 0, and therefore either xm q or ymn q for some n > 0. Thus x q or y q. If p = q, then q is said to be p-primary. Example. 1) The primary ideals in Z are (0) and (pn ) where p is prime, for these are the only ideals with a prime radicaland we can easily check they are primary. 2) Let A = k[x, y], q = (x, y2 ). Then A/q = k[y]/(y2 ), in which the zero-divisors are multiples of y. Hence the zero-divisors are nilpotent so q is primary and its radical is p = (x, y). We have p2 q q = p where the inclusions are strict, so that a primary ideal is not necessarily a prime-power. 3) Conversely, a prime power pn is not necessarily primary, although its radical is the prime ideal p. However, we have the following result: Proposition (4.2). If a is maximal, then a is primary. In particular, the powers of a maximal ideal m are m-primary. Proof. Let m := a. The image of m = ap p, where each p Spec(A), in A/a is the ap (p + a) which is nilradical of A/a Since the nilradical is the intersection of all prime ideals and the image of m is a maximal ideal (projection is surjective), we have that A/a has only one prime ideal, its maximal ideal. Hence every element of A/a is either a unit when it is not in the image of m or nilpotent when it is in the image since the image is a maximal 2

ideal since the A/a is local. It follows every zero-divisor of A/a is nilpotent. It follows from this because if xy a, then xy = 0 in A/a and x and y are nilpotent so xn , ym a for some n, m > 0. Next we study the presentation of an ideal as an intersection of primary ideals. Lemma (4.3). If qi for 1 i n are p-primary, then q = n qi is p-primary. i=1 Proof. q = n qi = n i=1 qi = p. Let xy inq, then xy qi for some i, so q inherits i=1 being primary. Lemma (4.4). Let q be a p-primary ideal, x an element of A. Then i) if x q, then (q : x) = (1); ii) if x q (QUESTION: x p), then (q : x) is p-primary, and therefore iii) if x p, then (q : x) = q (QUESTION: I think this should be = p). Proof. i) is trivial from denition. For iii), pick a (q : x) so ax q. Since x p we must have that an q for some n > 0. Thus a q = p. For a p, we have an q for some n > 0. Thus a(an1 x) = an x q since q is an ideal, and since an1 x (x) we get that a (q : x). For ii), let y (q : x). Then xy q, so y p since x p. Hence q (q : x) p; taking radicals, we get (q : x) = p. Let yz (q : x) with y p; then xyz q, so xz q, hence z (q : x). A primary decomposition of an ideal a in A is an expression of a as a nite intersection of primary ideals, say

(q : x) = p;


qi .


In general such a primary decomposition need not exist; in this chapter we shall restrict our attention to ideals which have a primary decomposition. If, moreover, (i) the radicals of each qi are all distinct, and (ii) we have qi j=i qj (1 i n) the primary decomposition (1) is said to be minimal. By collecting terms of the intersection and using (4.3) we can achieve (i), and by removing superuous terms we can achieve (ii); thus any primary decomposition can be reduced to a minimal primary decomposition. We shall say a is decomposable if it has a primary decomposition. Theorem (4.5). (First Uniqueness Theorem). Let a be a decomposable ideal and let a = n qi be a minimal primary decomposition of a. Let pi = qi . Then the pi are precisely the i=1 prime ideals which occur in the set of ideals { (a : x) | x A}, and hence are independant of the particular decomposition of a. Example. Let a = (x2 , xy) in A = k[x, y]. Then a = p1 p2 (QUESTION :(x, y)2 = (x2 , y2 , xy)?) 2 where p1 = (x), p2 = (x, y). The ideal p2 is primary by Theorem 4.2, so the prime ideals are 2 p1 , p2 . In this example, p1 p2 ; we have a = p1 p2 . 3

The prime ideals pi in Theorem 4.5 are said to belong to a, or to be associated with a. The ideal a is primary if and only if it has only one associated prime ideal. The minimal elements of the set {p1 , . . . , pn } are called the minimal or isolated prime ideals belonging to a. The others are called embedded prime ideals. In the previous example, p2 = (x, y) is embedded. Proposition (4.6). Let a be a decomposable ideal. Then any prime ideal p a contains a minimal prime ideal belonging to a, and thus the minimal prime ideals of a are precisely the minimal elements in the set of all prime ideals containing a. n Proof. If p a = n qi , then p = p n i=1 i=1 qi = i=1 pi . Hence by Theorem 1.11, we have p pi for some i. Thus p contains a minimal prime ideal of a. Remark. 1) The names isolated and embedded come from geometry. If A = k[x1 , . . . , xn ] where k is a eld, the ideal a gives rise to a variety X kn . The minimal primes correspond to the irreducible components of X, and the embedded primes correspond to sub-varieties of these, i.e. varieties embedded in the irreducible components.Thus in the example before, the variety dened by a is the line x = 0, and the embedded ideal p2 = (x, y) corresponds to the origin (0, 0). 2) It is not true that all the primary components are independent of the decomposition. For example, (x2 , xy) = (x) (x, y)2 = (x) (x2 , y) are two distinct minimal primary decompositions. However, there are some uniqueness properties which we will see later. Proposition (4.7). Let a be a decomposable ideal, let a = n qi be a minimal primary i=1 decomposition, and let qi = pi . Then

pi = {x A | (a : x) = a}.

In particular, if he zero ideal is decomposable, the set D of zero-divisors of A is the union of the prime ideals belonging to 0. Proof. If a is decomposable, then 0 is decomposable in A/a. Say 0 = qi where qi is the image of qi in A/a, and is primary. Hence it is enough to prove the last statement in the Proposition. We have D = x=0 (0 : x) by (1.15). From the proof of (4.5), we have (0 : x) = xqj pj p for some j. Hence D n pi . But also from (4.5) each pi is of i=1 the form (0 : x) for some x A; hence pi s D. Thus if the zero ideal is decomposable, D= and the nilradical is R= . Next we study the behavior of primary ideals under localization. 4 of all minimal primes belonging to 0. of all prime ideals belonging to 0,

Proposition (4.8). Let S be a multiplicatively closed subset of A, and let q be a p-primary ideal. i) If S p = , then S1 q = S1 A. ii) If S p = , then S1 q is S1 p-primary and its contraction in A i sq. Hence primary ideals correspond to primary ideals in the correspondence (3.11) between ideals in S1 A and contracted ideals in A. For any ideal a and any multiplicatively closed subset S in A, the contraction in A of the ideal S1 a shall be denoted by S(a). Proposition (4.9). Let S be a multiplicatively closed subset of A and let a be a decomposable ideal. Let a = n qi be a minimal primary decomposition of a. Let pi = qi and suppose i=1 the qi are numbered so that S meets pm+1 , . . . , pn but not p1 , . . . , pm . Then
m m

S1 a =

S1 qi ,

S(a) =

qi ,

and these are minimal primary decompositions. A set of prime ideals belonging to a is said to be isolated if it satises the following condition: if p is a prime ideal belonging to a and p p for some p , then p . (QUESTION: Wouldnt any ideal in the set be isolated so any set of ideals is isolated?) Let be an isolated set of prime ideals belonging to a, and let S = A p p. Then S is multiplicatively closed and, for any prime ideal p belonging to a, we have that p = p S = p = p p p (by (1.11)) = p S = . Thus we deduce from (4.9) Theorem (4.10). (Second Uniqueness Theorem). Let a be a decomposable ideal, let a = n qi be a minimal primary decomposition of a, and let {pi1 , . . . , pim } be an isolated set of i=1 prime ideals of a. Then qi1 qim is independent of the decomposition. Proof. We have qi1 qim = S(a) where S = A pi1 pim , so it depends only on a since the pi depend only on a. In particular: Corollary (4.11). The isolated primary components ( i.e. the primary components q corresponding to minimal prime ideals pi ) are uniquely determined by a. Remark. On the other hand, the embedded primary components are in general not uniquely determined by a. If A is a Noetherian ring, there are in fact innitely many choices for each embedded component. 5

In the polynomial ring Z[t], the ideal m = (2, t) is maximal and the ideal q = (4, t) is m-primary, but is not a power of m.

Since t = 2 = 0 in Z[t]/(2, t) any element is of the form x + (2, t) where x {0, 1}. Thus it must be Z2 which is a eld. Hence m is maximal. Note m2 is generated by elements of the form (2a + tb)(2c + td) = 4ac + t2ad + t2bc + ttbd = 4ac) + t(2ad + 2bc + tbd) (4, t) so m2 q m. Thus m = m2 q m = m which shows that q = m. Since m is maximal, (4.2) implies that q is primary. The inclusion along with q = m2 since t m also shows that q cannot be a power of m.

In the polynomial ring K[x, y, z] where K is a eld and x, y, z are independent indeterminates, let p1 = (x, y), p2 = (x, z), m = (x, y, z); p1 and p2 are prime, and m is maximal. Let a = p1 p2 . Show that a = p1 p2 m2 is a reduced primary decomposition of a. Which components are isolated and which are embedded?

We have p1 = p1 , p2 = p2 , and m2 = m which are all dierent. We have p1 p2 = (x) m2 , p1 m2 (z) p2 , and p2 m2 (y) p1 , so the primary decomposition is minimal. We have m p1 p2

so the isolated components are p1 , p2 and the embedded component is m.

Let A be a ring and let A[x] denote the ring of polynomials in one indeterminate over A. For each ideal a of A, let a[x] denote the set of all polynomials in A[x] with coecients in a. i) a[x] is the extension of a to A[x]. ii) If p is a prime ideal in A, then p[x] is a prime ideal in A[x]. iii) If q is a p-primary ideal in A, then q[x] is a p[x]-primary ideal in A[x].

iv) If a = n qi is a minimal primary decomposition in A, then a[x] = n qi [x] is a i=1 i=1 minimal primary decomposition in A[x]. v) If p is a minimal prime ideal of a, then p[x] is a minimal prime ideal of a[x].

i) Since a is closed under action by A and 1 A, this follow ii) Since p is prime, A/p is an integral domain. Thus A[x]/p[x] = (A/p)[x] is an integral domain. iii) By hypothesis, q = p. Since q[x] q = p we get q[x] p[x]; also q p, so p[c] q[x]. iv) Using the one-to-one correspondence from 1.17 with i) and that p1 [x] = p2 [x] when p1 = p2 with iii) we get the prime ideals qi [x] belonging to a[x] are all dierent. By contracting and using the minimal primary decomposition in A of a we easily see that qi [x] j=i qj [x]. v) Suppose there is a prime ideal q[x] p[x]. By i), qe pe ; hence i1 (qe ) = q i1 (pe ) = p.

In a ring A, let D(A) denote the set of prime ideals p which satisfy the following condition: there exists a A such that p is minimal in the set of prime ideals containing (0 : a). Show that x A is a zero divisor if and only if x p for some p D(A). Let S be a multiplicatively closed subset of A, and identify Spec(S1 A) with its image in Spec(A) (Chapter 3, Exercise 21). Show that D(S1 A) = D(A) Spec(S1 A). If the zero ideal has a primary decomposition, show that D(A) is the set of associated prime ideals of 0.

For the rst part, let x A be a zero divisor.

For any prime ideal p in a ring, let Sp (0) denote the kernel of the homomorphism A Ap . Prove that i) Sp (0) p. ii) Sp (0) = p if and only if p is a minimal prime ideal of A. 7

iii) If p p , then Sp (0) Sp (0). iv) pD(A) Sp (0) = 0, where D(A) is as in Exercise 9.

i) Let x Sp (0). Then x/1 = 0 so xu = 0 for some u A p. Then xu p, so x p. ii) Suppose Sp (0) = p. Let m be the maximal ideal of Ap , so we have m = ( Sp (0))p (Sp (0))p = 0Ap , so m is the only prime ideal of Ap . By the correspondence in (3.13), p is a minimal prime ideal of A. iii) Let x Sp (0). As in i), xu = 0 for some u A p Then u A p be hypothesis, so x/1 = 0 in Ap meaning x Sp (0). iv) Let x pD(A) Sp (0). Then xu = 0 for some u A pD(A) p, so u (0 : x). Let q be a minimal prime ideal over (0 : x); hence u q D(A), a contradiction.

Let A be a ring and p a prime ideal of A. The nth symbolic power of p is dened to be the ideal (in the notation of Exercise 12) p(n) = Sp (pn ) where Sp = A p. Show that i) p(n) is a p-primary ideal; ii) if pn has a primary decomposition, then p(n) is its p-primary component; iii) if p(m) p(n) has a primary decomposition, then p(m+n) is its p-primary component; iv) p(n) = pn if and only if pn is p-primary.

i) We have S1 p is the maximal ideal of S1 A, so S1 pn = S1 pn = S1 p. By (4.2) it p p p p p 1 n 1 follows then that Sp p is Sp p-primary. By (4.8) i) follows. ii) Let n qi be the minimal primary decomposition of pn , and let pi = qi . Consider i=1 n A/pn . Since p pn and p = p is the smallest prime containing pn we have that p/pn is the smallest prime of A/pn . Thus by Exercise 11, p(n) /pn is the smallest p/pn -primary ideal. It follows p(n) is the smallest p-primary ideal containing pn . Hence it must be in the decomposition and thus is the p-primary component. iii) Note that pm+n = pm pn , so by 1.18 we have p(m+n) = Sp (pm+n ) = Sp (pm pn ) Sp (pm )Sp (pn ) = p(m) p(n) . The rest follows like ii). iv) () is clear from i).

() Suppose pn is p-primary. As we showed in ii), p(n) is the smallest p-primary ideal containing p by looking at quotients and using Exercise 11. Thus we must have p(n) = pn given the hypothesis.

Let M be a xed A-module, N a submodule of M. The radical of N in M is dened to be rM (N) = {x A | xq M N for some q > 0}. Show that rM (N) = N : M = Ann(M/N). In particular, rM (N) is an ideal. State and prove the formulas for rM analogous to (1.13).

This seems like it follows immediately.

An element x A dened an endomorphism x of M, namely m xm. The element x is said to be a zero-divisor (resp. nilpotent) in M if x is not injective (resp. nilpotent). A submodule Q of M is primary in M if Q = M and every zero-divisor in M/Q is nilpotent. Show that if Q is primary in M, then (Q : M) is a primary ideal and hence rM (Q) is a prime ideal p. We say that Q is p-primary (in M). Prove the analogues of (4.3) and (4.4).

Let Q be primary in M.Then Q = M and every zero-divisor in M/Q is nilpotent. Suppose xy (Q : M) with y (Q : M). Hence xyM Q. If y is a zero-divisor in M/Q, then it is nilpotent by hypothesis, so yn M/Q = 0 whence yn (Q : M). If y is not a zero-divisor, ym = 0 if and only if m = 0 so it must be that xM Q whence x is a zero-divisor implying xn M/Q = 0 or xn (Q : M). Thus (Q : M) is a primary ideal.

A primary decomposition of N in M is a representation of N as an intersection N = Q1 Q n of primary submodules of M; it is a minimal primary decomposition if the ideals pi = rM (Qi ) are all distinct and if non of the components Qi can be omitted from the intersection. 9

Prove the analogue of (4.5), that the prime ideals pi depend only on N (and M). They are called the prime ideals belonging to N in M. Show that they are also the prime ideals belonging to 0 in M/N.


State and prove the analogues of (4.6)-(4.11) inclusive. (There is no loss of generality in taking N = 0.)