Bachelorarbeit
Zur Erlangung des akademischen Grades Bachelor of Science
Dimension theory for commutative rings
Eingereicht von: Daniel Heiß Eingereicht bei: Prof. Dr. Niko Naumann
Universität Regensburg Fakultät für Mathematik
Ausgabetermin: 22.08.2013 (Semester 5) Abgabetermin: 25.09.2013 Contents
Contents
I Introduction 1
II Primary Decomposition 2 2.1 Primary ideals ...... 2 2.2 The Lasker-Noether decomposition theorem ...... 4
III Krull’s principal ideal theorem 6 3.1 The n-th symbolic powers ...... 6 3.2 Krull’s principal ideal theorem ...... 7
IV A first boundary for dim(R[X]) 11
V Dimension theory of K[X1,...,Xn] 14
VI Dimension theory of Noetherian rings 16 6.1 Dimension under polynomial extensions ...... 16 6.2 Non-transcendental extensions ...... 17 Summary of non-transcendental extensions ...... 21
VII Prüfer domains and a note on Seidenberg’s F -rings 21 7.1 Dimension theory of Prüfer domains ...... 21 7.2 A note on F -rings ...... 22 Summary of polynomial extensions ...... 27
VIII A journey to the power series ring 27 Summary and comparison with polynomial extensions ...... 29
References iii
ii I Introduction
I Introduction
In mathematics there are several concepts of dimension. Some of them are quite intuitive, like the Hamel dimension: The Hamel dimension of a vector space is the minimal number of n vectors that are needed to generate the vector space. For the euclidean space R this coin- cides with the “natural” term of dimension which is the minimal number of coordinates that are necessary to specify each point of the space. On the other hand there are less intuitive concepts of dimension: Studying fractals one usually comes across the Hausdorff dimension which associates to any metric space a certain non-negative real number, which (for fractals) can be non-integer. For example the Sierpinski triangle (a famous fractal) has the Hausdorff log(3) dimension log(2) ∈/ Q. In algebraic geometry one of the basic objects of study are affine algebraic varieties. The dimension of an affine algebraic variety Y is defined as the supremum of all natural numbers n ∈ N such that there exists a chain Z0 ( Z1 ( ... ( Zn of irreducible closed subsets of Y . This idea of dimension is clearly the generalization of the intuitive Hamel dimension; and it can be translated into notion of commutative algebra: For this let k be an algebraically closed field and m ≥ 1 an integer. For an affine algebraic subvariety Y of km the chain of irre- ducible closed subsets of Y correspond by Hilbert’s Nullstellensatz to a chain of prime ideals p0 ( p1 ( ... ( pn of k[X1,...,Xm] containing the vanishing ideal I(Y ) of Y . This chain in turn corresponds to a chain of prime ideals p¯0 ( ... ( p¯n in the ring k[X1,...,Xm]/I(Y ). So the dimension of a variety Y can be interpreted as the supremum of the lengths of chains of prime ideals in the ring k[X1,...,Xm]/I(Y ). So we translated the concept of geometric dimension into notion of pure algebra. There this concept of dimension is called the Krull dimension of a ring and throughout this thesis we will call it just “the dimension”. Clearly the Krull dimension of any field k is zero and any principal ideal domain that is not a field is one-dimensional. The main interest of this thesis will be the behavior of the dimension of a commutative unitary ring R under certain extensions R −→ S. Mainly we study polynomial extensions and see that in several cases the dimension rises by the number of adjoint indeterminates.
This thesis is organized as follows: We start with a brief introduction to primary de- composition in order to proof Krull’s principal ideal theorem and similar results in section III. Equipped with this we are able to give a lower and an upper bound for the dimension dim(R[X]) of the polynomial ring R[X] as it is provided by [Sei53]. Then we follow princi- pally [BMRH73] and we will give an elementary proof that dim k[X1,...,Xn] = n for any field k or more generally any Artinian ring. Next we pass from Artinian rings to Noethe- rian rings and we will see – still following [BMRH73] – that we have a similar behavior: For any Noetherian ring we have dim(R[X]) = dim(R) + 1 fitting the previous result for zero-dimensional Noetherian rings. Still in the Noetherian case and now following [Cla65] we will further study the Krull dimension under ring extensions R −→ R[x] where x is not an indeterminate. It will turn out that the dimension either decreases by one or does not
1 II Primary Decomposition change at all and we will give some necessary and sufficient conditions for the dimension to decrease. In section VII we will find out that the beautiful property of the dimension under polynomial extensions in the case of Noetherian rings is also true for semi-hereditary rings, especially it is true for Prüfer domains. Further we will discuss rings that do not behave like Noetherian rings or Prüfer domains under polynomial extentions. Especially we will have a look at F -rings provided by Seidenberg in [Sei53, Sei54]. Eventually we will make a short visit to the formal power series rings in one and several indeterminates observing that in the Noetherian case the behavior is just like in the polynomial case while in the non-Noetherian case various weird things can happen. In this last section we only want to give a rough overview of Arnold’s work in [Arn73a,Arn73b,Arn82] concerning the large theory about the Krull dimension under power series extensions.
Notation. Throughout the whole thesis R denotes a commutative unitary non-zero ring of finite Krull dimension. Further the set of natural numbers N is supposed to contain zero. If a ring does not contain any proper zero-divisors we call it a domain. Moreover if not
otherwise specified X,X1,...,Xn denote indeterminates and p ⊆ R is assumed to be a prime ideal. The set of all prime ideals of R is denoted by Spec(R). When passing to a residue class ring R/a for some ideal a we denote with x¯ the residue class of x for any element x ∈ R and similar by b¯ we mean the image of the ideal b.
II Primary Decomposition
In this section we will briefly develop the theory of primary decomposition which is a basic tool studying ideals. Mainly we follow [ZS75], but also [AM69].
2.1 Primary ideals
We start with some basic facts about the radical of an ideal. √ n o n Definition 2.1. Let a ⊆ R be an ideal. Then we call a := x ∈ R ∃n ∈ N: x ∈ a the radical of a. √ √ Lemma 2.2. (i) If a ⊆ b for some ideals a, b ⊆ R, then a ⊆ b. √ (ii) For any prime ideal p ∈ Spec(R) the equation pn = p holds for all n ≥ 1.
(iii) Let I be a finite set and for all i ∈ I let qi ⊆ R be an ideal. Then we have pT T √ i∈I qi = i∈I qi. √ (iv) Let a ⊆ R be an ideal. Then a is the intersection of all prime ideals p ∈ Spec(R) containing a. Proof. Those are well-known properties. See [AM69, 1.13f.] among others. 2
2 II Primary Decomposition
p T Example 2.3. (i) For the nilradical of R we have N(R) := (0) = p∈Spec(R) p. (ii) For a prime number p ∈ we see pT (pn) = p(0) = (0) 6= (p) = T p(pn). Z n∈N n∈N This shows that we cannot omit the finiteness of I in Lemma 2.2 (iii).
In some sence the concept of prime ideals corresponds to those of prime elements. Now we introduce ideals that – in a similar way – correspond to powers of prime elements.
Definition 2.4. Let q ( R be an ideal such that √ for x, y ∈ R with xy ∈ q and x∈ / q we have y ∈ q.
√ Then q is called primary ideal of R. In this case one says that q is p-primary if p = q. √ Remark: In Proposition 2.8 we will see that for a primary ideal q the radical q ⊆ R is a prime ideal.
Remark 2.5. An ideal q ⊆ R is primary if and only if every zero-divisor of R/q is nilpotent.
Example 2.6. (i) Every prime ideal p is p-primary.
n (ii) The primary ideals of Z are (0) and (p ) where p is a prime number and n ≥ 1. (iii) Consider the ideal q := (X,Y 2) ⊆ R := K[X,Y ] where K denotes a field. Since we have R/q =∼ K[Y ]/(Y 2) all zero-divisors of R/q are multiples of Y hence nilpotent. √ 2 So q ⊆ R is primary. But we have p := q = (X,Y ) and p ( q ( p. This shows that a primary ideal need not be a power of a prime ideal.
(iv) Conversely not every power of a prime ideal is primary: Let R := K[X,Y,Z]/(XY −Z2) and p := (X,¯ Z¯). Then p is prime since R/p =∼ K[Y ], but p 2.2 X¯ · Y¯ = Z¯2 ∈ p2 and X/¯ ∈ p2, Y/¯ ∈ p2 = p.
So p2 is not primary although it is a power of a prime ideal.
We have seen that in general the powers of prime ideals need not be primary. However: √ Lemma 2.7. If p := q ∈ MaxSpec(R) then q is p-primary. In particular the powers of maximal ideals are primary. Proof. The second assertion follows from the first since for any n ≥ 1 and any maximal √ ideal m ∈ MaxSpec(R) we have mn 2.2= m. For the first assertion observe that p is the intersection of all prime ideals containing q, so the image of p in R/q is the intersection of all primes of R/q, so it is the nilradical N := N(R/q). As p ⊆ R is maximal, the nilradical N = p¯ ⊆ R/q is maximal. Now for any prime ideal a ⊆ R/q we have N ⊆ a and since N is maximal it follows that N = a yielding N is the only maximal ideal. An arbitrary zero-divisor x ∈ R/q is contained in some maximal ideal hence in N. Therefore it is nilpotent and the proof is complete. 2
3 II Primary Decomposition
In Example 2.6 we have seen that for an ideal q ⊆ R to be primary it is not sufficient that √ q ⊆ R is a prime ideal, but the following proposition states that it is necessary.
Proposition 2.8. Let q be p-primary. Then p is the smallest prime ideal containing q. √ Proof. By the fact that p = q we only have to show that p is prime. Let x, y ∈ R such that xy ∈ p, x∈ / p. So there is a some n ≥ 1 such that (xy)n = xnyn ∈ q. Since x∈ / p we have xn ∈/ q so – since q is primary – there is m ≥ 1 such that (yn)m = √ ynm ∈ q. Thus y ∈ q = p. 2
2.2 The Lasker-Noether decomposition theorem
Definition 2.9. (i) An ideal a ⊆ R is said to be irreducible if a = b ∩ c for some ideals b, c ⊆ R implies a = b or a = c.
(ii) For two ideals a, b ⊆ R the ideal
n o
(a : b) := x ∈ R xb ⊆ a ⊆ R
is called the ideal quotient. Remark: It is clear that the ideal quotient is indeed an ideal of R.
Example 2.10. Every prime ideal p ⊆ R is irreducible while a primary ideal need not be. Proof. The first assertion is clear (see [AM69, 1.11]). Now for a field K let m := (X,Y ) ∈ MaxSpec(K[X,Y ]). Then by Lemma 2.7 the ideal m2 = (X2,XY,Y 2) is primary, but m2 = m2 + (X) ∩ m2 + (Y ). 2
T Definition 2.11. A representation a = i qi of an ideal a ⊆ R as a finite intersection of primary ideals qi ⊆ R is called a primary decomposition. It is said to be irredundant or reduced if the following conditions hold: T (i) qi 6⊆ qj ∀j. i6=j √ √ (ii) qj 6= qi ∀i 6= j.
The main purpose of this section is to prove that for any ideal of a Noetherian ring there is an irredundant primary decomposition. First we notice:
Lemma 2.12. Let R be Noetherian and a ⊆ R an ideal. Then there exist finitely many T irreducible ideals qi ⊆ R such that a = i qi. Proof. Suppose the set F of all ideals failing the condition is non-empty. Therefore F contains a maximal element m. In particular m is not irreducible, so there are ideals b, c ⊆ R such that m = b ∩ c and m ( b, c. But since m is maximal in F it follows that b, c are finite intersections of irreducible ideals thus m is. 2
4 II Primary Decomposition
Lemma 2.13. In a Noetherian ring R every irreducible ideal is primary. Proof. Let q be non-primary. Then there exist elements a, b ∈ R such that ab ∈ q, a∈ / q √ and b∈ / p := q. Since R is Noetherian the increasing sequence q:(bs) becomes s∈N stationary. So there is some n ≥ 1 such that
(∗) q:(bn) = q:(bn+1).
Claim: q = q + (bn) ∩ q + (a) So the lemma follows since bn, a∈ / q, hence q is not irreducible. Proof of the claim: Since “⊆” is clear, it suffices to prove the reverse: Let x ∈ q + (bn) ∩ q + (a). Then there are u, v ∈ q, y, z ∈ R such that
(†) x = u + ybn = v + za.
(†) (†) So we have: bx = bv + zab ∈ q =⇒ q 3 bx = bu + ybn+1. (∗) =⇒ ybn+1 ∈ q =⇒ y ∈ q:(bn+1) = q:(bn) =⇒ ybn ∈ q. =⇒ x = u + ybn ∈ q and the claim follows. 2
Now we are able to get to the main result of this section:
Theorem 2.14 (Lasker-Noether decomposition theorem). Let R be Noetherian. Then every ideal a ⊆ R admits an irredundant finite primary decomposition. T Proof. By Lemma 2.12 and Lemma 2.13 we have a = i qi for finitely many primary ideals qi. So we just have to make sure that this representation is irredundant or that we can make it irredundant. T First if there is some ideal qj such that i6=j qi ⊆ qj we omit qj in the representation. It T is clear that we still have a = i6=j qi. If on the other hand there are i 6= j such that √ √ qi = qj then we replace qi and qj by ˜qij := qi ∩ qj. So we just have to make sure that
˜qij is primary. Tn More generally we show: If q1,..., qn are p-primary then q := i=1 qn is p-primary. pTn 2.2 Tn √ T For this we first observe that i=1 qi = i=1 qi = i p = p. Now let x, y ∈ R such that xy ∈ q, x∈ / q. So there is some 1 ≤ j ≤ n such that x∈ / qj but qj is p-primary and √ xy ∈ qj, so it follows y ∈ qj = p. 2
T Definition 2.15. Let a ⊆ R be an ideal admitting a primary decomposition a = qi. √ i The radicals pi := qi are called the associated primes or the prime ideals belonging to a
or simply the primes of a. If pi is minimal in the set of associated primes, then pi is said to be an isolated prime of a.
Remark. In general the primary decomposition is not unique. However throughout this thesis we are only interested in the isolated primes of some ideals and it can be shown that these are uniquely determined. See [ZS75, Cap. IV, §5] for a proof.
5 III Krull’s principal ideal theorem
The following theorem will be used repeatedly:
Theorem 2.16. Let {pi}i be the collection of associated primes of an ideal a ⊆ R. Then
a prime ideal p ∈ Spec(R) contains a if and only if it contains some pi. In particular the isolated primes of a are precisely the prime ideals that are minimal among all prime ideals that contain a.
Proof. (⇒) Clear since a ⊆ pi ⊆ p. T (⇐) We have p ⊇ a = qi. Then by [AM69, 1.11] there exists i such that qi ⊆ p and it √ √ i follows that pi = qi ⊆ p = p by Lemma 2.2. 2
Corollary 2.17. Let a ⊆ R be an ideal and p ∈ Spec(R).
(i) Let furthermore b ⊆ a be an ideal and denote by a¯, p¯ the image of a, p respectively in R/b. Then p¯ is an isolated prime of a¯ iff p is an isolated prime of a.
(ii) Let S ⊆ R be a multiplicatively closed subset of R such that p does not meet S. Then S−1p is an isolated prime of S−1a if and only if p is an isolated prime of a. Proof. Follows straight from the definitions, Theorem 2.16 and the ideal theory of the localization and residue class rings (see [AM69, 1.1, 3.11]). 2
III Krull’s principal ideal theorem
Equipped with the theory of primary decomposition and isolated primes we now formulate Krull’s principal ideal theorem together with some generalizations. But first we need a bit preparatory work.
3.1 The n-th symbolic powers
Notation. Let S ⊆ R be multiplicatively closed and ϕ: R −→ S−1R.
−1 −1 (i) For an ideal a ⊆ R let S a := ϕ∗(a) denote the image of a in S R.
(ii) For an ideal b ⊆ S−1R we write b ∩ R := ϕ∗(b) for the pre-image of b in R.
−1 (iii) For a prime ideal p ∈ Spec(R) we write aRp := (R \ p) a.
Proposition 3.1. Let (R, m) be a Noetherian local ring and a ( R an ideal admitting m as an isolated prime. Then a is m-primary and m is the only prime ideal containing a. Proof. Let p ∈ Spec(R) be any prime ideal containing a. Since R is local we have p ⊆ m. But m is an isolated prime of a yielding p = m (by Theorem 2.16). Now the radical of a is the intersection of all prime ideals containing a, hence m. So by Lemma 2.7 a is m-primary and we are through. 2
6 III Krull’s principal ideal theorem
We have seen in Example 2.6 that – in general – powers of prime ideals are not necessarily primary. However in the following we construct a certain primary ideal which we can associate with the power of a given prime ideal.
Definition 3.2. For a prime ideal p ∈ Spec(R) we define
(n) n p := p Rp ∩ R
to be the n-th symbolic power of p.
Lemma 3.3. The ideal p(n) ⊆ R is p-primary. n n Proof. First pRp is maximal in Rp, so by Lemma 2.7 we have p Rp = (pRp) is pRp- primary. Then more generally: For any ring homomorphism ϕ: R −→ S the pre-image ϕ∗(q) ⊆ R is primary for any primary ideal q ⊆ S. This is clear since R/ϕ∗(q) is isomorphic to some subring of B/q (now see Remark 2.5). p Finally one immediately verifies p(n) = p. 2
(n+1) (n) Corollary 3.4. If R is a Noetherian domain and p 6= (0) then p ( p ∀n ≥ 1. (j) j (n+1) (n) Proof. Ideal theory provides p Rp = p Rp for all j ≥ 1, so if p = p for some n ≥ 1, n+1 n then p Rp = p Rp in the local Noetherian ring (Rp, pRp) yielding that Rp is an Artinian ([AM69, 8.6]) domain, hence it is a field. Contradiction to p 6= (0). 2
Example 3.5. In the situation of 3.4 all conditions are necessary:
2 n (i) Let R := Z/4Z and (0) 6= (2) =: p. Then clearly p = (0) = p for all n ≥ 2.
(ii) Consider the integral ring extension Z ,−→ R := Z¯C and some (0) 6= p ∈ Spec(R) (then p is necessarily maximal). Observe the local ring (Rp, pRp): For any α ∈ pRp √ √ we evidently have α ∈ Rp. Now α is not a unit in Rp otherwise α would be. This √ √ √ 2 2 states α ∈ pRp and it follows pRp 3 α = α α ∈ (pRp) hence pRp = (pRp) .
As Rp is a domain, this additionally shows that Rp is not Noetherian (nor is R).
3.2 Krull’s principal ideal theorem
Definition 3.6. Let p ∈ Spec(R). By ht(p) := dim(Rp) we denote the height of p and dp(p) := dim(R/p) is called the depth of p.
Remark 3.7. (i) We have that ht(p) is the supremum of all n ∈ N such that there exists a chain of prime ideals p0 ( p1 ( ... ( pn = p ending with p and analogously dp(p) is the supremum of the lengths of chains of prime ideals starting with p.
(ii) We have ht(p) + dp(p) ≤ dim(R), but in general there is no equation: Take any ring having two maximal ideals of different height. An example is given in 6.3.
7 III Krull’s principal ideal theorem
(iii) Whenever we want to show an assertion about the height of a prime ideal p ⊆ R then we may assume that R is local and p its unique maximal ideal. This follows from the
ideal theory of the localization R −→ Rp.
Now we can prove the first version of Krull’s principal ideal theorem:
Theorem 3.8 (Krull’s principal ideal theorem). Let R be a Noetherian domain and (0) 6= (a) ( R a principal ideal. Then every isolated prime of (a) is of height one. Proof. Let p be an isolated prime of (a). By Remark 3.7 we may assume p is the only maximal ideal. Now suppose there is some prime ideal (0) ( v ( p. (n) Set qn := v and obtain by Lemma 3.3 and Corollary 3.4 an infinitely decreasing chain of primary ideals (∗) q1 ) ... ) qn ) qn+1 ) ... So we get a decreasing chain qn + (a) n of ideals containing (a). Since p is maximal and an isolated prime of (a) Proposition 3.1 provides that (a) is p-primary and p is the only prime ideal containing (a). (♦) It follows that R/(a) has only one prime ideal – namely the image of p in R/(a). Therefore we have dim R/(a) = 0, so R/(a) is Artinian and thus fulfills the descending chain condition.
So the sequence qn + (a) and therefore also qn + (a) become stationary. So there is n ≥ 1
such that qn + (a) = qn+1 + (a) yielding qn ⊆ qn+1 + (a).
Thus for an arbitrary chosen x ∈ qn there exist y ∈ qn+1 ⊆ qn, z ∈ R such that x =
y + za =⇒ za = x − y ∈ qn (+). Evidently a∈ / v, otherwise we get (a) ⊆ v ( p in contradiction to (♦). √ So since qn is primary and a∈ / qn (+) implies z ∈ qn.
We get: qn ⊆ qn+1 + aqn and since the other inclusion is obvious, it follows
(†) qn = qn+1 + aqn.
0 Now pass to R := R/qn+1 in which we denote by ¯q the image of qn. The equation (†) implies: ¯q =a ¯ · ¯q =⇒ ¯q = (¯a) · ¯q. Now ¯q is finitely generated as R0 is Noetherian. Furthermore we have Jac(R0) = p¯ since p is the only prime ideal of R. This gives (¯a) ⊆ p¯ = Jac(R0), so Nakayama’s lemma [AM69, 2.6]
supplies ¯q = (0)¯ . So qn = qn+1. to (∗). 2
Having proved the original principal ideal theorem we proceed with some generalizations.
Proposition 3.9. A prime ideal p ⊆ R of a Noetherian ring R has height ht(p) = 0 iff it is an isolated prime of (0). Proof. Clear by Theorem 2.16. 2
Corollary 3.10. Let R be a Noetherian ring and p an isolated prime of a principal ideal (a) ( R. Then ht(p) ≤ 1. Note that R needs not be a domain.
Proof. Let p ⊇ p1 ⊇ p2 be a chain of prime ideals of R. Then pass to the Noetherian
8 III Krull’s principal ideal theorem
domain R/p2 and get p¯ ⊇ p¯1 ⊇ (0)¯ . Corollary 2.17 gives that p¯ is an isolated prime of (¯a).
If (¯a) = (0)¯ Proposition 3.9 provides ht(p¯) = 0 and thus p¯ = (0)¯ (so p = p2).
If (¯a) 6= (0)¯ it follows by Theorem 3.8 that ht(p¯) = 1 yielding p¯1 = (0)¯ or p¯1 = p¯. 2
A further generalization can be made by varying the number of generators of the “principal” ideal. For this we need the following preparatory lemma:
Lemma 3.11. Let R be Noetherian, p0 ) ... ) pm a chain of prime ideals and let qi (for i ∈ I) be finitely many prime ideals not containing p0. Then there are prime ideals 0 0 0 p0 ) p1 ) ... ) pm−1 ) pm such that none of the pj are contained in any of the qi. Proof. It suffices to prove the claim for m = 2. S Since p0 6⊆ qi ∀i ∈ I we have: p0 6⊆ i qi ∪ p2 (see [AM69, 1.11]). S So there exists an element x ∈ p0 : x∈ / i qi ∪ p2.
Set a := p2 + (x) and denote by F := {p˜i} the set of associated primes of a (existence by 0 Theorem 2.14). Since a ⊆ p0 there is by Theorem 2.16 an isolated prime p1 ∈ F such that 0 0 0 0 p1 ⊆ p0. Further as x ∈ a ⊆ p1 we have by construction: p1 6⊆ qi ∀i ∈ I and p1 ) p2. 0 So the only thing left to show is that p0 ) p1:
Observe the Noetherian domain R/p2 in which a¯ is obviously a principal ideal, generated 0 0 by x¯. By Corollary 2.17 p1 is an isolated prime of a¯. Thus if we had p0 = p1 it would follow that p¯0 is an isolated prime of a¯ hence by Corollary 3.10 we have ht(p¯0) ≤ 1, but we have ¯ by assumption p0 ) p1 ) p2 yielding p¯0 ) p¯1 ) p¯2 = (0). 2
Eventually we obtain the final version of Krull’s principal ideal theorem:
Theorem 3.12 (Krull’s principal ideal theorem, generalized). Let R be a Noetherian ring, a ⊆ R an ideal generated by r elements and p an isolated prime of a. Then ht(p) ≤ r. Proof. Proof by induction on r. (r = 0) The only ideal generated by ∅ is (0) and thus ht(p) = 0 by Proposition 3.9.
(r − 1 r) Let a = (x1, . . . , xr) and b := (x1, . . . , xr−1). If p is among the isolated primes {vi} of b we are through by induction hypothesis.
So assume p is not contained in any vi. (In fact: b ⊆ a ⊆ p and vi are the minimal primes containing b.) Now let m := ht(p) and assume m 6= 0. So by definition there is a chain of prime ideals
(∗) p =: p0 ) p1 ) ... ) pm−1 ) pm.
Since p0 is not contained in any of the vi we can by Lemma 3.11 assume that pm−1 6⊆ vi ∀i. Now pass to the ring R/b in which a¯ is a principal ideal, generated by the residue class of
xr. By Corollary 2.17 p¯0 is an isolated prime of a¯ hence Corollary 3.10 supplies ht(p¯0) ≤ 1.
Recall that pm−1 6⊆ vi ∀i.
Thus pm−1 + b 6⊆ v¯i while the latter are the isolated primes of (0)¯ ( ♦).
9 III Krull’s principal ideal theorem
Claim: p¯0 is an isolated prime of pm−1 + b.
Proof : Obviously we have pm−1 + b ⊆ p¯0. Then let ˜q be an isolated prime of pm−1 + b
contained in p¯0 (Theorem 2.16). It follows ht(˜q) ≥ 1 since otherwise ˜q would be an isolated prime ideal of (0)¯ (Proposition 3.9) contradicting (♦).
So we have ˜q ⊆ p¯0 while ht(˜q) ≥ 1 and ht(p¯0) ≤ 1, so ˜q = p¯0. This proves the claim.
It follows that p0 is an isolated prime of pm−1 + b (Corollary 2.17) and therefore p0/pm−1 is
an isolated prime of (pm−1 + b)/pm−1 which is generated by r − 1 elements in R/pm−1.
The induction hypothesis implies that ht(p0/pm−1) ≤ r − 1. (†) Moreover the order-preserving one-to-one correspondence of ideals (see [AM69, 1.1]) supplies
ht(p0/pm−1) = m − 1 (see (∗)). Therefore by (†) m − 1 ≤ r − 1 hence m ≤ r. 2
Corollary 3.13. (i) Let (R, m) be a local Noetherian ring and n := dim(R). Then m cannot be generated by less than n elements. Remark: If it can be generated by exactly n elements the ring is called regular.
(ii) We have ht(p) < ∞ for any prime ideal p ∈ Spec(R) and thus Noetherian rings satisfies the descending chain condition for prime ideals. Note that dp(p) may be infinite, even in the Noetherian case! An example is provided in [Nag62, Appendix].
The following theorem is in some sence the converse of the principal ideal theorem 3.12:
Theorem 3.14. Let R be Noetherian and p ∈ Spec(R) with ht(p) = h. Then there is an ideal generated by h elements admitting p as an isolated prime. Proof. Proof by induction on i < h:
Let i < h and a1, . . . , ai ∈ p such that every isolated prime pj of (a1, . . . , ai) is of height i.
Since ht(pj) = i < h = ht(p) we have p 6⊆ pj ∀j. It follows that there is an element S ai+1 ∈ p\ j pj. Now an arbitrary chosen isolated prime v of (a1, . . . , ai+1) contains some pj by Theorem 2.16 and this inclusion is proper since ai+1 ∈ v \ pj. Remember that ht(pj) = i,
so we have ht(v) ≥ i + 1, but v is an isolated prime of (a1, . . . , ai+1) hence by Theorem 3.12 we get ht(v) = i + 1.
So successively this procedure supplies elements a1, . . . , ah ∈ p such that the isolated primes
vj of (a1, . . . , ah) are of height h. Again Theorem 2.16 yields that p contains some vj and since both ideals are of height h we have equality and the claim follows. 2
2 3 ∼ 2 3 2 3 Example 3.15. Consider R := C[X,Y ]/(Y − X ) = C[t , t ] and m := (t , t ). By 2 3 Theorem 3.8 ht (Y − X ) = 1 and as we will see in section V dim C[X,Y ] = 2, so dim(R) = 1 yielding ht(m) = 1. Thus ht(mRp) = 1, but mRm cannot be generated by less
than two elements. Especially Rm is not regular. Cf. Corollary 3.13. Furthermore we see: In order that a local Noetherian ring (R, m) is at most one-dimensional, it is sufficient but not necessary that m is principal.
10 IV A first boundary for dim(R[X])
IV A first boundary for dim(R[X])
Now the preparatory work is done and we can start studying the Krull dimension under certain ring extensions. We start with R,−→ R[X] where X denotes an indeterminate. The purpose of this section is to establish a first boundary for dim(R[X]) in terms of dim(R) as it is provided by [Sei53]. But first we need to take a look on the prime spectrum of R[X].
Notation. Throughout this section let ι: R,−→ R[X], p ∈ Spec(R) and P ∈ Spec(R[X]). ∗ Further we write pR[X] := ι∗(p) for the image of p under ι and analogously P ∩ R := ι (P) denotes the pre-image of P under ι. Finally we define
nX i o p[X] := aiX ∈ R[X] ai ∈ p ∀i ⊆ R[X]. i
Lemma 4.1. (i) We have p[X] = pR[X] and there is an isomorphism of rings
R[X]/pR[X] −→∼ (R/p)[X].
(ii) For a multiplicative system S ⊆ R we get (S−1R)[X] =∼ S−1(R[X]). Proof. (i) In the first assertion p[X] ⊆ pR[X] is clear and so we just have to show “⊇”: P P j Let α ∈ pR[X]. Then α = i αiβi with αi ∈ p, βi ∈ R[X], thus βi = j γijX , so P j α = i,j αiγij X ∈ p[X]. | {z∈p } For the second assertion observe
η : R[X] − (R/p)[X] X i X i aiX 7−→ a¯iX i i
Apply the fundamental theorem on homomorphisms observing ker(η) = p[X] = pR[X]. (ii) Observe
R u with
−1 v ( S R R[X] −1 −1 _ η1 : S R −→ S (R[X]), α 7−→ α η1 η2 X X ai η : R[X] −→ (S−1R)[X], a Xi 7−→ Xi 2 i 1 i i
u g ) −1 o −1 (S R)[X] / S (R[X]) f
X We set f(X) := 1 and by universal properties the existences are clear and it is easily verified that g ◦ f = id, f ◦ g = id. 2
An immediate consequence of Lemma 4.1 (ii) is:
11 IV A first boundary for dim(R[X])
Corollary 4.2. Let R be a domain. Then Lemma 4.1 (ii) in particular implies that