18.07.2014 Views

Seidenberg's theorems about Krull dimension of polynomial rings ...

Seidenberg's theorems about Krull dimension of polynomial rings ...

Seidenberg's theorems about Krull dimension of polynomial rings ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Seidenberg’s <strong>theorems</strong> <strong>about</strong> <strong>Krull</strong> <strong>dimension</strong> <strong>of</strong> <strong>polynomial</strong> <strong>rings</strong><br />

Darij Grinberg<br />

written back in 2009 if not earlier; translated into English in 30 minutes in 2011.<br />

Rough writing and mistakes guaranteed.<br />

—————————————————————————————————-<br />

In the following, ”ring” always means ”commutative ring with 1”.<br />

By dim R, we always denote the <strong>Krull</strong> <strong>dimension</strong> <strong>of</strong> a ring R.<br />

If we have a ring R and speak <strong>of</strong> R [X] or R [X 1 , X 2 , ..., X n ] without having defined<br />

the variables X resp. X 1 , X 2 , ..., X n beforehand, it is to be understood that X resp.<br />

X 1 , X 2 , ..., X n mean free variables (for <strong>polynomial</strong> <strong>rings</strong>).<br />

The free coefficient <strong>of</strong> a <strong>polynomial</strong> always means its coefficient before X 0 (or, if<br />

we work with several variables, the coefficient before X 0 1X 0 2...X 0 n).<br />

A ring is called nontrivial if its 0 is non-equal to its 1.<br />

Not only the zero element <strong>of</strong> a ring, but also the zero ideal <strong>of</strong> a ring will be denoted<br />

by 0.<br />

The following <strong>theorems</strong> are results by A. Seidenberg, quoted after<br />

http://planetmath.org/?op=getobj&from=objects&id=7196<br />

Theorem 1. Let R be a nontrivial ring. We are going to show two ways<br />

to obtain an ideal <strong>of</strong> R [X] from an ideal <strong>of</strong> R.<br />

a) For every ideal a ⊆ R, the set<br />

aR [X] = {P ∈ R [X] | all coefficients <strong>of</strong> P lie in a}<br />

is an ideal <strong>of</strong> R [X]. If a ⊆ R is a prime ideal, then aR [X] ⊆ R [X] is a<br />

prime ideal, too.<br />

b) For every ideal a ⊆ R, the set<br />

aR [X] + XR [X] = {P ∈ R [X] | the free coefficient <strong>of</strong> P lies in a}<br />

is an ideal <strong>of</strong> R [X]. If a ⊆ R is a prime ideal, then aR [X]+XR [X] ⊆ R [X]<br />

is a prime ideal, too.<br />

Using a) and b), we can show the following property <strong>of</strong> <strong>polynomial</strong> <strong>rings</strong>:<br />

c) We have dim (R [X]) ≥ dim R + 1.<br />

Pro<strong>of</strong> <strong>of</strong> Theorem 1 (sketched). a) The only nontrivial part here is to show that<br />

if a ⊆ R is a prime ideal, then aR [X] ⊆ R [X] is a prime ideal. The easiest way to<br />

show this is by noticing that (R [X]) (aR [X]) ∼ = (Ra) [X], and that (Ra) [X] is<br />

an integral domain if Ra is one. For an alternative pro<strong>of</strong>, consider two <strong>polynomial</strong>s<br />

P, Q ∈ R [X] \ (aR [X]); each <strong>of</strong> these has at least one coefficient not lying in a; now<br />

take the highest such coefficients and use these to show that P Q /∈ aR [X] . Proceeding<br />

in this way one should not forget proving that aR [X] is not the whole R [X].<br />

1


) The only nontrivial part here is to show that if a ⊆ R is a prime ideal, then<br />

aR [X] + XR [X] ⊆ R [X] is a prime ideal. This can be shown by proving that<br />

(R [X]) (aR [X] + XR [X]) ∼ = Ra (in fact, the map R [X] → Ra defined by<br />

n∑<br />

a i X i ↦→ (a 0 ) a<br />

is surjective, and its kernel is aR [X] + XR [X]), or again by hand.<br />

i=0<br />

c) Each chain p 0 p 1 ... p n−1 p n <strong>of</strong> prime ideals <strong>of</strong> R yields (due to a) and<br />

b)) a chain p 0 R [X] p 1 R [X] ... p n−1 R [X] p n R [X] p n R [X] + XR [X] <strong>of</strong><br />

prime ideals <strong>of</strong> R [X]. This latter chain is one element longer than the former. This<br />

yields Theorem 1 c).<br />

Theorem 2. Let R be a ring. Theorem 1 a) and b) gave two ways to<br />

construct ideals <strong>of</strong> R [X] from ideals <strong>of</strong> R; the converse direction is simpler:<br />

a) For every ideal a ⊆ R [X], the set a ∩ R is an ideal <strong>of</strong> R. If a ⊆ R [X] is<br />

a prime ideal, then a ∩ R ⊆ R is a prime ideal.<br />

Next let us show how ideals <strong>of</strong> R [X] can be lifted to ideals <strong>of</strong> (Quot R) [X]<br />

when R is an integral domain:<br />

b) Let R be an integral domain, and let Quot R be the quotient field <strong>of</strong> R.<br />

Let us canonically identify R with a subring <strong>of</strong> Quot R. Each element <strong>of</strong><br />

(Quot R) [X] has the form P for some P ∈ R [X] and some r ∈ R \ {0} .<br />

r<br />

b 1 ) For every ideal a ⊆ R [X], the set<br />

{ }<br />

P<br />

a ((Quot R) [X]) =<br />

r | P ∈ a; r ∈ R \ {0}<br />

is an ideal <strong>of</strong> (Quot R) [X] .<br />

b 2 ) If a ⊆ R [X] is a prime ideal such that a ∩ R = 0, then a ((Quot R) [X])<br />

is a prime ideal as well.<br />

b 3 ) If a is a prime ideal <strong>of</strong> R [X] and b is an ideal <strong>of</strong> R [X] with a b and<br />

a ∩ R = 0, then a ((Quot R) [X]) b ((Quot R) [X]) .<br />

c) Let R be an integral domain. If a and b are two prime ideals <strong>of</strong> R [X]<br />

with 0 a ⊆ b and b ∩ R = 0, then a = b.<br />

d) Let R be an arbitrary ring again. Let p, q and r be three prime ideals<br />

<strong>of</strong> R [X] with p q r. Then, p ∩ R = q ∩ R = r ∩ R cannot hold. (In<br />

other words, at least one <strong>of</strong> the two inclusions in p ∩ R ⊆ q ∩ R ⊆ r ∩ R is<br />

a strict inclusion.)<br />

e) Let R be an arbitrary ring. Then, dim (R [X]) ≤ 2 dim R + 1.<br />

Pro<strong>of</strong> <strong>of</strong> Theorem 2 (sketched).<br />

a) is completely trivial.<br />

b 1 ) Proving a ((Quot R) [X]) =<br />

{ P<br />

r | P ∈ a; r ∈ R \ {0} }<br />

is easy (just notice<br />

that every element <strong>of</strong> (Quot R) [X] has the form P for some P ∈ R [X] and some<br />

r<br />

r ∈ R \ {0}). That it is an ideal is more than obvious.<br />

2


2 ) Let P r<br />

and Q with P, Q ∈ R [X] and r, s ∈ R \ {0} be two elements <strong>of</strong><br />

s<br />

(Quot R) [X] which satisfy P r · Q<br />

s ∈ a ((Quot R) [X]). Then, P<br />

r · Q<br />

s = T for some<br />

{ t }<br />

P<br />

T ∈ a and some t ∈ R \ {0} (due to a ((Quot R) [X]) =<br />

r | P ∈ a; r ∈ R \ {0} ).<br />

Thus, tP Q = rsT. Since T ∈ a, this leads to tP Q ∈ a, and thus t ∈ a or P ∈ a or<br />

Q ∈ a. But t ∈ a is impossible, since t ∈ R \ {0} and R ∩ a = 0. Thus, P ∈ a or<br />

Q ∈ a. Therefore, P r ∈ a ((Quot R) [X]) or Q ∈ a ((Quot R) [X]) . It now remains to<br />

s<br />

prove that a ((Quot R) [X]) is not the whole ring (Quot R) [X]. But this is easy again<br />

(assume the contrary; then, 1 ∈ a ((Quot R) [X]) ; thus there exist some P ∈ a and<br />

some r ∈ R \ {0} such that 1 = P , so that P = r, so that r ∈ a, which is impossible<br />

r<br />

because <strong>of</strong> r ∈ R \ {0} and R ∩ a = 0).<br />

b 3 ) From a ⊆ b we immediately obtain a ((Quot R) [X]) ⊆ b ((Quot R) [X]). The<br />

only thing we must show is therefore a ((Quot R) [X]) ≠ b ((Quot R) [X]) . For the sake<br />

<strong>of</strong> contradiction, assume that a ((Quot R) [X]) = b ((Quot R) [X]) . Since a b, there<br />

exists some element P ∈ b such that P /∈ a. Now, P ∈ b yields P ∈ b ((Quot R) [X])<br />

(we identify P with P ), so that P ∈ a ((Quot R) [X]) . Therefore there exists an S ∈ a<br />

1<br />

and an r ∈ R \ {0} such that P = S . Hence, rP = S. Since S ∈ a, this yields rP ∈ a.<br />

r<br />

Since a is a prime ideal, this entails r ∈ a or P ∈ a. But r ∈ a is impossible (since<br />

r ∈ R \ {0} and a ∩ R = 0), and P ∈ a is impossible as well (remember P /∈ a). This<br />

is the contradiction we wanted.<br />

c) Assume that a ≠ b. Then, 0 a ⊆ b becomes 0 a b.<br />

Let Quot R be the quotient field <strong>of</strong> R. Since a ∩ R = 0 (because b ∩ R = 0 and<br />

a ⊆ b) and b ∩ R = 0, Theorem 2 b 2 ) yields that the ideals a ((Quot R) [X]) and<br />

b ((Quot R) [X]) are prime ideals <strong>of</strong> (Quot R) [X] . Moreover, 0 a b yields (according<br />

to Theorem 2 b 3 )) that 0 ((Quot R) [X]) a ((Quot R) [X]) b ((Quot R) [X])<br />

(since 0 and a are prime ideals, a and b are ideals, and we have 0∩R = 0 and a∩R = 0).<br />

Since 0 ((Quot R) [X]) = 0, this becomes 0 a ((Quot R) [X]) b ((Quot R) [X]) .<br />

We thus have two prime ideals a ((Quot R) [X]) and b ((Quot R) [X]) in the ring<br />

(Quot R) [X] satisfying the inclusion 0 a ((Quot R) [X]) b ((Quot R) [X]). But<br />

(Quot R) [X] is a principal ideal domain (since Quot R is a field), and thus any prime<br />

ideal <strong>of</strong> Quot R distinct from 0 is a maximal ideal. This yields that no prime ideal<br />

<strong>of</strong> Quot R distinct from 0 is contained in another prime ideal <strong>of</strong> Quot R. This contradicts<br />

0 a ((Quot R) [X]) b ((Quot R) [X]) (since both a ((Quot R) [X]) and<br />

b ((Quot R) [X]) are prime ideals <strong>of</strong> (Quot R) [X]). Contradiction! This proves Theorem<br />

2 c).<br />

d) Assume that p ∩ R = q ∩ R = r ∩ R. Set p ∩ R = q ∩ R = r ∩ R = n. Then,<br />

n ⊆ R is a prime ideal (due to Theorem 2 a), since p ⊆ R [X] is a prime ideal). The<br />

ring Rn thus is an integral domain.<br />

The canonical surjection ( R → Rn induces a canonical surjection ν : R [X] →<br />

n<br />

)<br />

∑ ∑<br />

(Rn) [X] , given by ν a i X i = n (a i ) n<br />

X i . It is easy to see that Ker ν = nR [X],<br />

i=0<br />

i=0<br />

and thus the images ν (p) , ν (q) and ν (r) <strong>of</strong> the prime ideals p, q and r <strong>of</strong> R [X] under<br />

this surjection ν are prime ideals <strong>of</strong> (Rn) [X] (this follows from the trivial fact that if<br />

3


f : A → B is a surjective ring homomorphism between two <strong>rings</strong> A and B, and p ⊆ A<br />

is a prime ideal with Ker f ⊆ p, then f (p) ⊆ B is a prime ideal as well). Furthermore,<br />

p q r yields ν (p) ν (q) ν (r) (here we use the following fact: if f : A → B is a<br />

surjective ring homomorphism between two <strong>rings</strong> A and B, and if a b are two ideals<br />

<strong>of</strong> A such that Ker f ⊆ a, then f (a) f (b)).<br />

Since ν (p) ν (q), we have 0 ν (q) (because 0 ⊆ ν (p)). Thus, 0 ν (q) ν (r) .<br />

Now we will show ν (r) ∩ (Rn) = 0. In fact, every element <strong>of</strong> Rn ( has the form<br />

n<br />

)<br />

∑<br />

a n for some a ∈ R. If such an element a n lies in ν (r), then a n = ν a i X i for<br />

i=0<br />

n∑<br />

∑<br />

some <strong>polynomial</strong> a i X i ∈ r, so that a n = n (a i ) n<br />

X i . By comparing coefficients, we<br />

i=0<br />

therefore conclude that a n = (a 0 ) n<br />

and (a i ) n<br />

= 0 for all i > 0. Thus, a − a 0 ∈ n and<br />

a i ∈ n for all i > 0. Since n = r ∩ R, this yields a i ∈ r for all i > 0 (since a i ∈ n for<br />

∑<br />

all i > 0). Combined with n a i X i ∈ r, this results in a 0 X 0 ∈ r, so that a 0 ∈ r. Since<br />

i=0<br />

a 0 ∈ R, this yields a 0 ∈ r ∩ R = n, so that a n = 0. We thus have shown that if an<br />

element a n <strong>of</strong> Rn lies in ν (r), then a n = 0. In other words, ν (r) ∩ (Rn) = 0.<br />

We summarize: The ring Rn is an integral domain. Furthermore, ν (q) and ν (r)<br />

are two prime ideals <strong>of</strong> (Rn) [X] satisfying 0 ν (q) ν (r) and ν (r) ∩ (Rn) =<br />

0. Thus, Theorem 2 c) yields ν (q) = ν (r) , contradicting to ν (q) ν (r). This<br />

contradiction proves Theorem 2 d).<br />

e) Every chain p 0 p 1 ... p n−1 p n <strong>of</strong> prime ideals <strong>of</strong> R [X] gives rise to a<br />

chain p 0 ∩ R ⊆ p 1 ∩ R ⊆ ... ⊆ p n−1 ∩ R ⊆ p n ∩ R <strong>of</strong> prime ideals <strong>of</strong> R (according to<br />

Theorem 2 a)). According to Theorem 2 d), no two subsequent ⊆ signs in this chains<br />

can simultaneously be = signs. Thus, if we remove repeated terms from this chain, it<br />

will be shortened by a factor <strong>of</strong> no more than 2. This easily yields Theorem 2 e).<br />

i=0<br />

4

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!