Commutative Algebra I

1
Craig Huneke

June 27, 2012

1
A compilation of two sets of notes at the University of Kansas; one in the Spring
of 2002 by ?? and the other in the Spring of 2007 by Branden Stone. These notes
have been typed by Alessandro De Stefani and Branden Stone.
Contents

1 Rings, Ideals, and Maps 1

1 Notation and Examples . . . . . . . . . . . . . . . . . . . . . . . 1
2 Homomorphisms and Isomorphisms . . . . . . . . . . . . . . . . . 2
3 Ideals and Quotient Rings . . . . . . . . . . . . . . . . . . . . . . 3
4 Prime Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5 Unique Factorization Domain . . . . . . . . . . . . . . . . . . . . 13

2 Modules 19
1 Notation and Examples . . . . . . . . . . . . . . . . . . . . . . . 19
2 Submodules and Maps . . . . . . . . . . . . . . . . . . . . . . . . 20
3 Tensor Products . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4 Operations on Modules . . . . . . . . . . . . . . . . . . . . . . . 29

3 Localization 33
1 Notation and Examples . . . . . . . . . . . . . . . . . . . . . . . 33
2 Ideals and Localization . . . . . . . . . . . . . . . . . . . . . . . . 36
3 UFD’s and Localization . . . . . . . . . . . . . . . . . . . . . . . 40

4 Chain Conditions 44
1 Noetherian Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2 Noetherian Modules . . . . . . . . . . . . . . . . . . . . . . . . . 47
3 Artinian Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5 Primary Decomposition 54
1 Definitions and Examples . . . . . . . . . . . . . . . . . . . . . . 54
2 Primary Decomposition . . . . . . . . . . . . . . . . . . . . . . . 55

6 Integral Closure 62
1 Definitions and Notation . . . . . . . . . . . . . . . . . . . . . . . 62
2 Going-Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3 Normalization and Nullstellensatz . . . . . . . . . . . . . . . . . . 67
4 Going-Down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
CONTENTS iii

7 Krull’s Theorems and Dedekind Domains 77

1 Krull’s Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2 Dedekind Domains . . . . . . . . . . . . . . . . . . . . . . . . . . 80

8 Completions and Artin-Rees Lemma 83

1 Inverse Limits and Completions . . . . . . . . . . . . . . . . . . . 83
2 Artin-Rees Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3 Properties of Completions . . . . . . . . . . . . . . . . . . . . . . 87

Bibliography 94

Index 95
iv CONTENTS
Chapter 1

Rings, Ideals, and Maps

1 Notation and Examples

Through out these notes, a ring R is considered a commutative ring. That
is a set with two operations +, · such that under +, R is an abelian group
with additive identity 0. Multiplication is associative with identity 1 (or 1R ),
distributive: a(b + c) = a · b + a · c for all a, b, c ∈ R, and commutative: ab = ba.
Further, make note that there is no differentiation between the symbols ⊂
and ⊆. The symbol ( will be used to represent a proper subset.
A commutative ring is a field if for all non-zero elements r ∈ R, there exits
r0 ∈ R (denoted r−1 ) such that rr0 = 1.
Example 1. The integers, Z, and the fields Q, R, and C are rings.
Example 2. Let Mn (k) denote the n × n matrices over a field k. Choose
any set of commuting matrices A1 , . . . , Am ∈ Mn (k). Let R be the subring
generated by A1 , . . . , Am in Mn (k). I.e. all sums and products of them. Note
that dimk Mn (k) = n2 . How big can dimk R be?

Examples of building new rings form old ones

Example 3. Let R be a ring. Then R[x], the polynomial ring over R, is
{r0 + r1 x + · · · + rn xn |ri ∈ R, n > 0} with the usual addition and multiplication
of polynomials. Note that
r0 + r1 x + · · · + rn xn = s0 + s1 x + · · · + sn xn ⇐⇒ ri = si ∀i
Inductively we define R[x1 , . . . , xn ] to be R[x1 , . . . , xn−1 ][xn ].
Example 4 (Direct product). Let I be an index set and Ri be rings for i ∈ I.
Then the direct product is given by
Y
Ri = {(ri )i∈I |ri ∈ Ri }
i∈I

with component-wise operations.

2 Rings, Ideals, and Maps

Example 5. The integers with modulo n arithmetic; Z/nZ or Zn .

2 Homomorphisms and Isomorphisms

Let R, S be rings. A homomorphism f : R → S is a function from R to S
preserving the operations, i.e, for all r, r0 in R we have

f (r + r0 ) = f (r) + f (r0 )
f (rr0 ) = f (r)f (r0 )
f (1R ) = 1S

Note that f (0) = 0 is forced.

An isomorphism is an homomorphism f which is injective and surjective.
If there exists an isomorphism between R and S we say that R and S are
isomorphic and write R ' S.
Definition. If f : R → S is an homomorphism (or “map”) then the kernel of
f , denoted ker f , is {r ∈ R|f (r) = 0}.

Examples of kernels and homomorphisms

Example 6. Let R be a ring. Then there exists a unique homomorphism f
from Z to R given as follows: If n > 0 then
n times
z }| {
f (n) = f (1 + · · · + 1)
= f (1) + · · · + f (1)
= 1R + · · · + 1 R
= n · 1R

Similarly, if n < 0,
|n| times
z }| {
f (n) = f (−1R ) + · · · + f (−1R ) = n · 1R .

Example 7. Let R be a ring and choose n elements r1 , . . . , rn ∈ R. Then there

exists a unique homomorphism f : Z[x1 , . . . , xn ] → R such that f (xi ) = ri , i.e.
given by evaluating at (r1 , . . . , rn ). Namely if
X
p(x1 , . . . , xn ) = mv xv11 · · · xvnn
vi >0

where v = (v1 , . . . , vn ) ∈ Zn and mv ∈ Z. Then

X
f (p(x1 , . . . , xn )) = mv r1v1 · · · rnvn
vi >0
3 Ideals and Quotient Rings 3

3 Ideals and Quotient Rings

A subset I ⊆ R is said to be an ideal if I is a subgroup under addition and for
all r ∈ R, i ∈ I, we have ri ∈ I.

Example 8. If f : R → S is an homomorphism, then the kernel of f is an

ideal.

Example 9. Suppose {rλ }λ∈Λ ⊆ R. Then the ideal generated by this set,
denoted (rλ ), or (rλ )R, is the set of elements of the form
X
sλ rλ
λ∈Λ0

where Λ0 ⊆ Λ, |Λ0 | < ∞, and sλ ∈ R. This is the smallest ideal containing all
of rλ . We say I is finitely generated if I = (r1 , . . . , rn ), ri ∈ R and principal if
n = 1.

Example 10 (Direct Sum). The direct sum is an ideal Q of the direct product
(not necessarily a subring) and consists of
L all (ri ) ∈ Ri such that all but
finitely many ri are 0. We denote this by Ri where i ∈ I (see example 4).

Example 11. An arbitrary intersection of ideals is again an ideal. Note that

the union of ideals is not necessarily an ideal. For example consider 2Z and 3Z
in Z.

Example 12. Given two ideals I, J the product is an ideal and IJ = (ij) where
i ∈ I and j ∈ J. Note that in general IJ 6= {ij|i ∈ I, j ∈ J}

Example 13. The sum of two ideals I, J is an ideal and is given by I + J =

{i + j|i ∈ I, j ∈ J}

Example 14. Given two ideals I, J the colon is the ideal I : J = {r ∈ R : rJ ⊆

I}.

Example 15. If R = Z then every ideal is principal of the form (n) for some
n > 0.

Proof. Let I ⊆ Z be an ideal, I 6= (0). Note that n ∈ I if and only if −n ∈ I.

So without loosing generality, choose n ∈ I least such that n > 0. To show that
I = (n), consider m ∈ I. Then m = qn + r for 0 6 r < n. But this shows that
r = m − qn which is an element of I. Thus r = 0 and I = (n).

Example 16. Let k be a field, R = k[x]. Then every ideal of R is principal.

Example 17. Let k be a field, R = k[x, y]. If I = (x, y), that is the set of all
f ∈ R such that f (0, 0) = 0, then I is not principal. In fact (x, y)n requires
n + 1 generators.
4 Rings, Ideals, and Maps

Let I be an ideal in a ring R. Then the quotient ring, denoted R/I, is the
set {r + I|r ∈ R} of additive cosets of I. We make this into a ring by
(r + I) + (s + I) := (r + s) + I
(r + I) · (s + I) := rs + I
We need to show that the product is well defined. First recall that r + I =
{r + i|i ∈ I} and we have r + I = s + I if and only if there exists i, i0 ∈ I such
that r + i = s + i0 . This is equivalent to r − s ∈ I.
Now suppose r+I = r0 +I and s+I = s0 +I. We want to show rs+I = r0 s0 +I,
i.e. rs − r0 s0 ∈ I. So notice,
rs − r0 s0 = (r − r0 )s − r0 (s0 − s) ∈ I.
The zero element in R/I is 0 + I = I. The multiplicative identity is 1 + I. In
general we will write r = r+I. There is a surjective homomorphism π : R → R/I
defined by π(r) = r with ker(π) = I.
Example 18. Let R = Z. Then by example 15 we know that every ideal I is
principal, say I = (n). Then the map Z → Z/I is just Z → Z/(n) such that
m 7→ m for m ∈ Z.
Example 19. Let I = (x2 + 1) be an ideal in the polynomial ring R[x] for R
the real numbers. Then R[x] maps onto C via
R[x] → R[x]/(x2 + 1) ' C
Proposition 1 (Isomorphism Theorem). Let f : R → S be a ring homomor-
phism and set I = ker f . Then f factors as a surjection followed by an injection
as in the following commutative diagram:

R
f
/S
>
π
g

R/I

In particular, if f is onto, then g is an isomorphism.

Proof. Note if r ∈ R, then we must have (g(r) =) g(π(r)) = f (r) giving the
definition of g. To show g is well defined, suppose r = r0 . This implies that
r − r0 ∈ I. Hence f (r − r0 ) = 0 which gives f (r) = f (r0 ). So we have that
g(r) = g(r0 ). The fact that g is a homomorphism is clear. We need to show g is
injective. Suppose that g(r) = g(t). Then f (r) = f (t) which gives f (r − t) = 0.
Thus r − t ∈ I and r = t.
Example 20. Let R be a ring and I an ideal of R. Let IR[x] denote the ideal
in R[x] generated by {i|i ∈ I}. Then
 
R[x] R
' [x]
IR[x] I
3 Ideals and Quotient Rings 5

Proof. Apply proposition 1 with R := R[x] and S := (R/I)[x]. Define f : R → S

by f (r0 + r1 x + · · · + rn xn ) = r0 + r1 x + · · · + rn xn . This is onto and hence by
the proposition we have established the isomorphism.

Theorem 2. Let I ⊆ R be an ideal.

(1) Every ideal J in R := R/I is of the form J/I where J is an ideal in R

containing I and J/I = {j | j ∈ J}. Moreover, if J is an ideal in R
containing I, then J/I is an ideal in R/I.

(2) If J1 , J2 are ideals in R containing I, then J2 = J2 if and only if J1 = J2 .

Hence this correspondence is a one-to-one inclusion preserving correspon-
dence between ideals of R/I and ideals in R containing I.

(3) If J ⊇ I are ideals then R/J ' R/J

Proof. Let J ⊆ R be an ideal and define J = π −1 (J) under the projection map
π : R → R/I. Since 0 ∈ J, we must have I ⊆ J. It is easy to see that J is an
ideal and J/I = J by definition. Conversely, given an ideal J of R containing
I, it is also easy to see that J/I is an ideal.
To show the second statement, note the if J1 = J2 then we have J1 = J2 .
Conversely, suppose J1 = J2 and let ji ∈ J1 . Then there exists j2 ∈ J2 such
that j1 = j2 . This implies that j1 − j2 ∈ I. But I ⊆ J2 thus we have j1 =
(j1 − j2 ) + j2 ∈ J2 . Hence J1 ⊆ J2 . a similar argument shows that J2 ⊆ J1 .
For the third statement, use proposition 1 with S = R/J and the composi-
tion f of the surjections

R
π / R/I π / R/J.
=
f

Note that the ker(f ) = J.

Remark. Suppose I = (f ) is a principal ideal in R and g ∈ R = R/I. Suppose

we want to understand what R/(g) looks like. Set J := (g) ⊆ R and let J be
as in the theorem. What is J?
By definition, J = π −1 ((g)) where π : R → R/(f ). To understand the
structure of J, consider and element j ∈ J, i.e. j ∈ (g). Equivalently j has the
property that there exists an r ∈ R such that j = r · g. But this is the same as
saying j − rg ∈ (f ), that is for some s ∈ R, j − rg = sf . Thus j = rg + sf .
This forces J = (f, g). Therefore by the third part of theorem 2 we have that
R/(g) ' R/(f, g). By the same token, R/(f, g) ' R̃/(f˜) where R̃ = R/(g).

Example 21. Let i ∈ C, i2 = −1, and Z[i] = {a + bi | a, b ∈ Z}. This ring

is called the Gaussian integers. In order to analyze the quotient ring Z[x]/(5)
consider the unique homomorphism f : Z[x] → Z[i] defined by evaluating at i
6 Rings, Ideals, and Maps

(see example 7). From proposition 1 we have that Z[i] ' Z[x]/(x2 + 1). So by
the above remark we have that
Z[i] Z[x]
'
(5) (x2 + 1, 5)
Z[x]/5Z[x]
'
(x2 + 1)
(Z/5Z)[x]
'
(x2 + 1)
Z5 [x]
'
(x − 2)(x − 3)

4 Prime Ideals
Given a ring R, an ideal I is maximal if there does not exist a proper ideal J
such that I ( J. This is equivalent to R/I being a field. A proper ideal I is
prime if whenever ab ∈ I then a ∈ I or b ∈ I. A ring is called a domain if for
any elements a, b in the ring such that ab = 0 then a = 0 or b = 0. An ideal I in
a ring R is prime if and only if R/I is a domain. Hence we have that maximal
ideals are prime as fields are domains. On the other hand, the converse is not
true, consider the ring Z with ideal (0).

Topology on Spec(R) The set of prime ideals in a ring R is denoted Spec(R).

For an example consider the ring Z. Then Spec(Z) is all the ideals generated
by a prime element and the zero ideal. Another example is k[x], a polynomial
ring over a field. This is a PID so Spec(k[x]) is the set of ideals generated by
an irreducible element and the zero ideal.
In general, if f : R → S is an homomorphism and Q ∈ Spec(S) then we have
that f −1 (Q) is in Spec(R). So we get an induced map f ∗ : Spec(S) → Spec(R).
A special case of this is if I is an ideal in a ring R and π : R → R/I is
the projection map. If π ∗ : Spec(R/I) → Spec(R) is the induced map where
Spec(R/I) is the set of ideals in the form P/I, with P a prime ideal in R such
that P ⊇ I, then π ∗ (P/I) = P .
A topology on Spec(R) can be given by saying a set V ⊆ Spec(R) is closed
if and only if there exists and ideal I in R such that V = {p | p ⊇ I} (denoted
V (I)). Open sets are given by X −V (I) = U where X = Spec(R). This topology
on Spec(R) is called the Zariski topology. Note that under this topology the
induced map f ∗ is continuous. A basis of the open sets is {D(r)}r∈R where
D(r) = {p ∈ Spec(R) | r ∈ / p}. To see this notice that D(r) = X − V (r) and if
U = X − V (I) then U = ∪r∈I D(r).
For an example, consider the ring Z. Notice that (0) ∈ Spec(Z) but is not
closed. For if {(0)} = V (I) for some ideal I then I = (0) and thus V (I) =
Spec(Z), a contradiction. In fact, the closed points are maximal ideals.
Remark. Suppose that R is a domain. Then the intersection of any two nonempty
open sets is non-empty, i.e. Spec(R) is irreducible.
4 Prime Ideals 7

Proof. Let U1 , U2 be non-empty open sets, Ui = X − V (Ii ), i = 1, 2. Notice

that Ui = ∅ if and only if Ii = (0). This is true since (0) ∈ Spec(R). So
Ii 6= (0) iff (0) + Ii iff (0) ∈ X − V (Ii ). Conversely, if p ∈ Ui then p + Ii and
therefore (0) + Ii . This implies that Ii 6= 0. Therefore I1 6= (0), I2 6= (0), and
(0) ∈ U1 ∩ U2 6= ∅.

Nilradical and Jacobson Radical For r ∈ R a ring, we say r is a unit

if there exists an s ∈ R such that rs = 1. Equivalently,
√ (r) = R. Given an
ideal I in a ring R, the nilradical of I (denoted I) is defined as the set of all
elements of the ring R whose powers are in I. An element r of the ring R is
called nilpotent
√ if some power of r is zero. The set of all nilpotent elements can
be given by 0 and is denoted Nilrad(R). The Jacobson radical of R, denoted
Jac(R) is the intersection of all maximal ideals in R.
√
Remark. The radical of I is an ideal and V ( I) = V (I).
√
Proof. √If x ∈ I, r ∈ R then (xr)n = xn rn ∈ I for all n sufficiently large. If
x, y ∈ I, say xn , y m ∈ I then (x + y)m+n−1 is also in I. Notice that
m+n−1
X  
m+n−1 m + n − 1 i m+n−1−i
(x + y) = xy
i=0
i

is in I since either i > n or m + n − √1 − i > m. √

To see the second part, let p ∈ V (√ I). Then p ⊇ I ⊇ I and thus p ∈ V (I).
n
Conversely, let√p ∈ V (I) and x ∈ I. Then √ x ∈ I ⊆ p and hence x ∈ p.
Therefore p ⊇ I and we have that p ∈ V ( I).
Remark. The ring R is said to be reduced if there are no non-zero
p√ nilpotent
√ √
elements. For example, the ring R/ I is always reduced since I = I.
Lemma 3. If R is a ring and x is non-nilpotent then there exist ideals maximal
with respect to the exclusion of the set S = {xn }∞
n=0 and any such ideal is prime.

Proof. Let Σ be the set of ideals of R that do not meet S. Note that Σ is
non-empty since the zero ideal does not meet S. By Zorn’s lemma1 , we have
that Σ has at least one maximal element I. Now assume that ab ∈ I and that
a, b ∈
/ I. Hence we have that the ideals (I, a) and (I, b) have I as a proper subset
and hence neither are elements of Σ. That is, there exists natural numbers n, m
such that xn ∈ (I, a) and xm ∈ (I, b). So for some i1 , i2 ∈ I and r, s ∈ R we
have xn = i1 + ra and xn = i2 + sb. The product

xn+m = xn xm = (i1 + ra)(i2 + sb) = i1 i2 + i1 sb + i2 ra + abrs

is an element of I, a contradiction. So I is prime.

1 Let S be a non-empty partially ordered set (i.e. we are given a relation x 6 y on S which

is reflexive and transitive and such that x 6 y and y 6 x together imply x = y). A subset
T of S is a chain if either x 6 y or y 6 x for every pair of elements x, y in T . Then Zorn’s
Lemma may be stated as follows: if every chain T of S has an upper bound in S (i.e. if there
exists x ∈ S such that t 6 x for all t ∈ T ) then S has at least one maximal element.
8 Rings, Ideals, and Maps

Corollary 4. Every ideal is contained in a maximal ideal

Proof. Apply the lemma 3 to the ring S = R/I and x = 1 + I ∈ S. So there
exists an ideal J/I in S maximal with respect to not containing 1S . I.e. if J/I
is maximal in S then J ⊇ I and J is maximal in R.
Proposition 5. Let R be a ring and I and ideal of R. The following are
equivalent.
√
(1) x ∈ I
(2) For all ring homomorphisms φ : R → k where k is a field, φ(x) ∈ (φ(I)).
That is, the image of x in k is in the ideal generated by the image of I in
k.
(3) x ∈ p for all p ∈ Spec(R) such that p ⊇ I.
√
In particular, I is the intersection of all prime ideals containing I.
√
Proof. (1)⇒(2): Assume that x ∈ I and let k be a field. Consider a ring
homomorphism φ : R → k. Since (φ(I)) is an ideal of k it is either (1) or (0).
The former case is clear so assume the latter. For all large n, φ(xn ) = 0 so that
φ(x)n = 0. But we are in a field, in particular a domain, hence φ(x) = 0. (Note
that we could have restated (2) to say for all ring homomorphism φ : R → k
such that φ(I) = 0, φ(x) = 0.)
(2)⇒(3): Suppose that I ⊆ p for p a prime ideal. Consider the composition

R
π / / R/p   i / κ(p).
<
φ

where κ(p) is the field of fractions of R/p. Choose x ∈ R as in (2). Notice that
φ(I) = 0 since I ⊆ p. Therefore we have that φ(x) = 0. Since i is an injection,
we have that x + p = 0 in R/p and √ thus x ∈ p.
(3)⇒(1): Assume that x ∈ / I and pass to R/I (assume R = R/I). Thus √
we have reduced to the case in which x ∈ p for all p ∈ Spec(R) and x ∈ / 0.
But by lemma 3, since x is not nilpotent, there exists a prime ideal P maximal
with respect to the exclusion of {xn }∞ n=0 ; a contradiction.

Chinese Remainder Theorem

Definition. Two ideals I and J of a ring R are comaximal if I + J = R.
Example 22. If n, m ∈ Z are relatively prime, then (n) and (m) are comaximal.
Example 23. If k is a field, then for any two distinct elements α, β ∈ k, the
elements (x − α) and (x − β) are comaximal in the polynomial ring k[x]. To see
this notice
1 1
1= (x − α) − (x − β).
β−α β−α
More generally, if m1 , m2 are distinct maximal ideals then m1 and m2 are co-
maximal since m1 ( m1 + m2 implies m1 + m2 = R.
4 Prime Ideals 9

Theorem 6 (Chinese Remainder Theorem). Let R be a ring and I1 , . . . , In

ideals. If Ii and Ij are comaximal for all i 6= j, then

(1) R/I1 × · · · × R/In ' R/I1 ∩ · · · ∩ In

(2) I1 ∩ · · · ∩ In = I1 I2 · · · In .

Proof. First do the case n = 2. For simplicity, Let I = I1 and J = I2 . By

assumption there exists i ∈ I and j ∈ J such that i + j = 1. Since IJ ⊆ I ∩ J is
always true, consider an element r of I ∩ J. Then ri + rj = r and IJ ⊇ I ∩ J.
Hence the second part holds.
For the first part, consider a map f : R → R/I × R/J defined by f (r) =
(r+I, r+J). We need to prove that ker(f ) = I ∩J and f is onto. By proposition
1 these two facts give the desired result. First notice that

f (r) = 0 ⇔ r + I = I, r + J = J
⇔ r ∈ I ∩ J.

Thus ker(f ) = I ∩ J. Next let (a + I, b + J) be an element of R/I × R/J and

set r = aj + bi. Then

r+I =a+I ⇔ r−a∈I

⇔ aj + bi − a ∈ I
⇔ a(j − 1) + bi ∈ I
⇔ a(−i) + bi ∈ I.

A similar argument can be made for b and J. Hence f (r) = (a + I, b + J).

To complete the proof use induction on n. With out loosing any generality,
assume n > 2. If I1 ∩ · · · ∩ In−1 and In are comaximal, then by the induction

R/I1 × · · · × R/In−1 ' R/I1 ∩ · · · ∩ In−1

and
I1 ∩ · · · ∩ In−1 = I1 I2 · · · In−1 .
By the case of n = 2 we may conclude that

R/I1 ∩ · · · ∩ In−1 × R/In−1 ' R/I1 ∩ · · · ∩ In−1

and
(I1 ∩ · · · ∩ In−1 ) ∩ In = (I1 I2 · · · In−1 ) · In .
In order to ensure I1 ∩ · · · ∩ In−1 and In are comaximal suppose they are
not and argue by way of contradiction. That is assume I1 ∩ · · · ∩ In−1 and
In are not comaximal. Then (I1 ∩ · · · ∩ In−1 ) + In is a proper ideal and is
contained in some maximal ideal m. Since I1 · · · In−1 ⊆ I1 ∩ · · · ∩ In−1 we get
that I1 · · · In−1 ⊆ m. Since m is prime there exists 1 6 j 6 n − 1 such that
Ij ⊆ m. But then Ij + In ⊆ m, a contradiction.
10 Rings, Ideals, and Maps

Example 24. Recall from example 21 that

Z[i] (Z/5Z)[x] Z5 [x]

' 2
' .
(5) (x + 1) (x − 2)(x − 3)

Here the ideals (x − 2) and (x − 3) are comaximal.

Prime Avoidance

Theorem 7 (Prime Avoidance: Version 1). Let I1 , . . . , In be ideals in a ring

R, at most two are not prime. If J is another ideal and J ⊆ ∪ni=1 Ii then J ⊆ Ij
for some j.

Proof. First assume n = 2 and that J * I1 , J * I2 . Choose elements j1 ∈ J −I1 ,

j2 ∈ J − I2 and consider j = j1 + j2 . Since J ⊆ I1 ∪ I2 we have j1 ∈ I2 and
j2 ∈ I1 . But j ∈ J. So with out lose of generality, assume j ∈ I1 . Then
j1 = j − j2 ∈ I1 , a contradiction.
By induction on n, we will prove the general case. Let n > 2 and choose I1
to be prime. Assume that J * Ij for all j = 1, . . . , n. By the induction, for all
1 6 j 6 n, J * ∪l6=j Il , so choose aj ∈ J − ∪l6=j Il (note aj ∈ Ij ). Now consider
n
[
a = a1 + a2 a3 · · · an ∈ Ij .
j=1

Suppose that a ∈ Ij for some j > 2. Since aj ∈ Ij , a2 a3 · · · an ∈ Ij . Hence

a1 = a − a2 a3 · · · an ∈ Ij , a contradiction.
If a ∈ I1 , then a2 a3 · · · an = a − a1 ∈ I1 . Since I1 is prime, there exists an
l such that 2 6 l 6 n where al ∈ I1 , another contradiction. Therefore a ∈ J is
not contained in any Ij . In other words, J * ∪nj=1 Ij .

Theorem 8 (Prime Avoidance: Version 2). Let R be a ring and p1 , . . . , pn be

prime ideals of R. Suppose x ∈ R, I is an ideal of R, and

{rx + i | r ∈ R, i ∈ I} = (x) + I * pj

for 1 6 j 6 n. Then there exists an i ∈ I such that x + i ∈

/ p1 ∪ · · · ∪ pn .

Proof. We can assume pj 6⊆ pk for j 6= k. Assume by way of contradiction

that the coset x + I ⊆ p1 ∪ · · · ∪ pn and fix v such that x ∈ p1 ∩ · · · ∩ pv ,
x∈/ pv+1 ∪ · · · ∪ pn . We know by version 1 that I * p1 ∪ · · · ∪ pv . If so, I ⊆ pk
for some k 6 v and therefore (x) + I ⊆ pk , a contradiction.
Choose i0 ∈ I r (p1 ∪ · · · ∪ pv ) and r ∈ (pv+1 ∩ · · · ∩ pn ) r (p1 ∪ · · · ∪ pv )
(This is possible since pv+1 ∩ · · · ∩ pn * pi , 1 6 i 6 v). Then x + ri0 ∈ x + I
and x + ri0 ∈/ p1 ∪ · · · ∪ pn .

Example 25. Let R = Z2 [x, y]/(x, y)2 . Then (x, y)R ⊆ xR ∪ yR ∪ (x + y)R
but (x, y)R is not contained in any of them.
4 Prime Ideals 11

Proof. Every element in (x, y)R can be represented by an element of the form
αx + βy + (x, y)2 where α, β ∈ Z2 . Thus there are only 4 elements in (x, y)R;
namely x + y, x, y, and 0.

Example 26. Let k be a field and R = k[x] a polynomial ring p in one variable.
For an irreducible polynomial f (x), p = (f (x)) is prime and (f n (x)) = p.
p T
Example 27. Consider the integers Z. For n > 0, (n) = (p)⊇(n) (p) =
T Q
p|n (p) = p|n (p). (Chinese remainder theorem)

Example 28. Let k be a field and R = k[x] be a√polynomial ring in one variable.
Consider the ideal I = (x2 , xy, y 2 ) of R. Then I = (x, y).

Example 29. Let R = k[a, b, c, d] be a polynomial ring over in four variables

over the field k. Consider the matrix
    2 
a b a + bc ab + bd
A= A2 = .
c d ac + cd bc + d2
√
Let I = (a2 + bc, ab + bd, ac + cd, bc + d2 ). Then I = (det(A), trace(A)).

√
Proof. We will first show that (det(A), trace(A)) ⊆ I. By proposition 5 it
is enough to show that if homomorphism φ : R → k such that φ(I) = 0 then
φ(det(A)) = φ(trace(A)) = 0. Apply φ to the matrix A and let α, β, γ, δ be the
images of a, b, c, d respectively. Further let M = φ(A). Since φ(I) = 0 we have
that M 2 = 0. The characteristic polynomial is T 2 − trace(M )T + det(M ) · I.
So mM (T ) = T 2 and therefore Cm (T ) is also T 2 . This implies that trace(M ) =
det(M ) = 0. So trace(M ) = φ(trace(A)), det(M ) = φ(det(A)).
To see the reverse inclusion, it is enough to show I ⊆ (trace(A), det(A)) and
(trace(A), det(A)) is prime.

Example 30.

(1) (General set up) Let R = k[x1 , . . . , xn ], S = k[t1 , . . . , tn ] be polynomial

rings over the field k. In general, a homomorphism φ can be defined
from R to S by sending xi to fi (t1 , . . . , tm ) in S. This extends to a
surjective homomorphism φ0 from R to k[f1 , . . . , fn ] via φ(g(x1 , . . . , xn )) =
g(f1 , . . . , fn ). Notice that k[f1 , . . . , fn ] is a subring of S. So the kernel of
φ0 is a prime ideal in R. I.e. ker(φ) = {g(x1 , . . . , xn ) | g(f1 , . . . , fn ) = 0}.

(2) Let φ : k[x, y] → k[t] defined by x 7→ t2 and y 7→ t3 . The kernel of φ is

{g(x, y) | g(t2 , t3 ) = 0}. This is a prime ideal, say ker(φ) = p. Notice
x3 − y 2 ∈ ker φ.
Claim. ker(φ) = (x3 − y 2 )
12 Rings, Ideals, and Maps

Let R0 = k[x, y]/(x3 − y 2 ). Since (x3 − y 2 ) ⊆ p we have an induced

homomorphism φ0 : R0 → k[t2 , t3 ]. Consider the commutative diagram

φ0
k[x, y]/(x3 − y 2 ) / k[t2 , t3 ]
9

' f

k[x, y]/p

where f is an isomorphism by proposition 1. Since (x3 − y 2 ) ⊆ p, φ0 is

surjective.
Remark. If φ : V → W is a homomorphism of vector spaces, and {vi }
span V and if {φ(vi )} are a basis for W , then φ is an isomorphism.

P
Proof.
P Clearly φ is onto since {φ(vi )} are a basis. If φ( αi vi ) = 0 then
αi φ(vi ) = 0 and thus αi = 0 for all i. That is, φ is one-to-one.

Given the remark, consider the vector space basis for R0 and S. Notice
that
S = k + kt2 + kt3 + kt4 + kt5 + · · ·
and k[x, y] has a k-basis {xi y j }i,j>0 . In R0 , {xi y j }i,j>0 are a generat-
ing set. Since x3 = y 2 we can refine this generating set to {xi , xi y}i>0 .
Note that φ0 (xi ) = t2i and φ0 (xi y) = t2i+3 for i > 0. So under φ0 ,
{φ0 (xi ), φ0 (xi y)} is a k-basis of S. So the above remark shows φ0 is an
isomorphism. In particular, ker(φ0 ) = 0 + (x3 − y 2 ) and thus ker(φ) =
(x3 − y 2 ).

(3) Consider the map φ from k[x, y, z] to S = k[t3 , t4 , t5 ] defined by x 7→ t3 ,

y 7→ t4 , and z 7→ t5 . We have that φ is onto, but what is the kernel?
Notice that the expressions y 2 − xz, x3 − yz, z 2 − x2 y ∈ ker(φ).
Claim. The ideal I = (y 2 − xz, x3 − yz, z 2 − x2 y) is the kernel of φ.

Let R = k[x, y, z]/I. A k-basis for S is {ti }i>3,i=0 . A k-generating set of

R is {xi y j z k }i,j,k>0 . But we can refine this to {xi , xi y, xi z}i>0 . Apply φ
to get the set

{φ(xi ) = t3 i, φ(xi y) = t3i+4 , φ(xi z) = t3i+5 }.

This is a k-basis of S. So φ is an isomorphism and I = ker(φ).

(4) In general, if a, b, c ∈ Z+ such that (a, b, c) = 1 then if we map k[x, y, z]

into k[ta , tb , tc ] via the map φ, then the kernel of φ is always (minimally)
generated by the 2 or 3 elements which are exactly the least powers of
x, y, z expressible as a product of the other two.
5 Unique Factorization Domain 13

Going back to k[t3 , t4 , t5 ], notice that the three generators of the kernel of
φ are excactly the 2 × 2 minors of
 
x y z
y x2 z

up to sign.

5 Unique Factorization Domain

Definition. A (non-zero) non-unit r ∈ R is said to be irreducible if r 6= ab
where a and b are non-units. (Note that if u is a unit, r = u(u−1 r).)

Remark. In a domain R, a non-unit r is irreducible if and only if (r) is maximal

among principal (proper) ideals.

Proof. Suppose r is irreducible and (r) ⊆ (s). Hence s divides r and there exists
a t in R such that st = r. But then either t is a unit and (r) = (s), or s is a
unit and (s) = R.
Conversely, if (r) is maximal among proper principal ideals and r = ab, then
since (r) ⊆ (a) either (a) = R (a is a unit) or (r) = (a). In the latter case,
r divides a as well. So, rs = a for some s in R and thus rsb = r. Therefore
r(1 − sb) = 0. Since R is a domain and r 6= 0, then 1 = sb.

Definition. A ring R is a unique factorization domain (UFD, factorial in

french) if R is a domain and

(1) For all irreducible elements r ∈ R, (r) is prime, and

(2) Any non-zero, non-unit r is a product of Irreducible elements.

Theorem 9. Let R be an UFD. Then every element r ∈ R which is non-zero

and not a unit is uniquely a product

r = an1 1 an2 2 · · · ank k

where a1 , a2 , . . . , ak are irreducible elements, and (ai ) 6= (aj ) for i 6= j up to

rearrangements and units.

Proof. First show uniqueness: Suppose

r = an1 1 · · · ank k = bm ml
1 · · · bl
1

bi irreducible, (bi ) 6= (bj ) for i 6= j. By induction, it suffices to prove (bi ) = (aj )

for some i, j. Then bi = aj u for some u a unit. We cancel bi and aj up to units
and the induction follows.
But
b1 | an1 1 an2 2 · · · ank k
14 Rings, Ideals, and Maps

so,
an1 1 an2 2 · · · ank k ∈ (b1 ).
But (b1 ) is prime, so there exists and i, 1 6 i 6 k such that ai ∈ (b1 ). This
implies (ai ) ⊆ (b1 ). But ai is irreducible and thus (ai ) = (b1 ) by remark 5.
What is left is to prove existence: This is given by part (2) of the definition
of UFD.
Definition. An element e ∈ R is idempotent if e2 = e.
Remark (See Atiyah, p. 20). For an idempotent e in R, 1 = e + (1 − e), and
e2 = e if and only if e(1 − e) = 0. Further,

R ' Re × R(1 − e)

where Re has identity e and R(1 − e) has identity (1 − e).

Definition. A domain R is a principal ideal domain (PID) if every ideal is
principal.
Example 31. Z, k[x], k a field.
Theorem 10. If a ring R is a PID then it is also a UFD.
To prove this theorem, we need to prove the following
(A) If a is irreducible then (a) is prime.
(B) Every element factors into irreducible factors.
But first some lemmas.
Lemma 11. If R is a PID, then every ascending chain of ideals stabilizes.
(Such a ring is said to be Noetherian.)
Proof. Suppose
I1 ( I2 ( · · · ( Ii ( · · ·
is an infinite ascending chain. We can write Ii = (ai ). Let
∞
[
I= Ii .
i=1

This is an ideal. Therefore there exists a ∈ R such that (a) = I. But then there
exists an i such that a ∈ Ii . Then for all j > i

(ai ) ( (aj ) ⊆ I = (a) ⊆ (ai ),

a contradiction.
Lemma 12. Let R be a PID. The following are equivalent:
(1) The element a is irreducible.
5 Unique Factorization Domain 15

(2) The ideal (a) is maximal.

Proof. This is clear from remark 5.
Now we are ready to prove the theorem.

Proof of Theorem 10. (A) is a consequence of lemma 12, since maximal ideals
are prime. For (B) let a ∈ R, a is non-unit, non-zero. There exits a maximal
ideal m containing a. Since R is a PID, m = (a1 ), a1 is irreducible by lemma
12. So, a = a1 b1 . If b1 is a unit, done. If not, (a) ( (b1 ). Repeat with b1 in
place of a.
There exists an irreducible a2 such that b1 = a2 b2 . Therefore

(a) ( (b1 ) ( (b2 )

If b2 is a unit then a = a1 a2 b2 . If not, continue. By lemma 11, this chain stops

and we have our factorization.
16 Rings, Ideals, and Maps

Exercises
(1) Consider the natural injection f : Z → Z[x] and g : Z[x] → Z, the
evaluation at n ∈ Z. What is the kernel of g and g ◦ f ?
(2) Let x be a nilpotent element of a ring R. Show that 1 + x is a unit in R. [1]
Deduce that the sum of a nilpotent element and a unit is a unit.
(3) Let R be a ring and let R[x] be the ring of polynomials in an indeterminate [1]
x, with coefficients in R. Let f = r0 + r1 x + · · · + rn xn ∈ R[x]. Prove that
(a) f is a unit in R[x] iff r0 is a unit in R and r1 , . . . , rn are nilpotent.
(b) f is nilpotent iff r0 , r1 , . . . , rn are nilpotent.
(c) f is a zero-divisor iff there exists r 6= 0 in R such that rf = 0.
(d) f is said to be primitive if (r0 , r1 , . . . , rn ) = (1). Prove that if f, g ∈
R[x], then f g is primitive iff f and g are primitive.
(4) For a ring R, r ∈ Jac(R) iff 1 − rs is a unit for all s ∈ R.
(5) What is the Jacobson radical of the polynomial ring R[x]? [1]

(6) An ideal m of a ring R is maximal if and only if R/m is a field.

(7) An ideal p of a ring R is prime if and only if R/p is a domain.
(8) Let R be a ring in which every element x satisfies xm = x for some n > 1 [1]
(depending on x). Show that every prime ideal in R is maximal.
(9) Let k and k 0 be two fields, verify that the ring k × k 0 is not a field. [3]

(10) Show that a ring R is a field iff (0) is the unique proper ideal of R. [3]

(11) Let k be a field. Show that the composition homomorphism [3]

k[y]
i / k[x, y] π/
k[x, y]/(x)k[x, y]

is an isomorphism.
(12) If Ii , i = 1, . . . , n are ideals of a ring R and P a subset of R, show that [3]

(∩i Ii ) : P = ∩i (Ii : P ).

(13) Let k be a field and (a1 , . . . , an ) ∈ k n . Show that the set of all polynomials [3]
P ∈ k[x1 , . . . , xn ], such that P (a1 , . . . , an ) = 0, is a maximal ideal of
k[x1 , . . . , xn ] generated by x1 − a1 , . . . , xn − an .
(14) Show that all non-zero prime ideals of a principal ideal ring are maximal. [3]
(15) Show that if a ring R has only one prime ideal, then an element of R is [3]
invertible or nilpotent.
 5 Unique Factorization Domain 17

(16) Let ki , with i = 1, . . . , n, be fields. Show that the ring k1 × · · · × kn has [3]
only finitely many ideals.

(17) If R is a principal ideal ring and a ∈ R a non-zero element, show that the [3]
quotient ring R/aR has only finitely many ideals.

[3] (18) Let R be a UFD and a ∈ R a non-zero element. Show that the nilradical
of R/aR is the intersection of a finite number of prime ideals. If p1 , . . . , pn
are these prime ideals, show that for each prime ideal p of R/aR there
exists i such that pi ⊆ p.

[3] (19) Let R be a ring and a ∈ R. a nilpotent element. Show that 1 + a is

invertible. If an = 0 and an−1 6= 0 describe the inverse of a.

Zariski Topology

[1] (20) Let R be a ring, X = Spec(R) and V (E) denote the set of all prime ideals
of R which contain E. Prove that
p
(a) if a is the ideal generated by E, then V (E) = V (a) = V ( (a)).
(b) V (0) = X, V (1) = ∅.
(c) if (Ei )i∈I is any family of subsets of R, then

V (∪i∈I Ei ) = ∩i∈I V (Ei ).

(d) V (a ∩ b) = V (ab) = V (a) ∪ V (b) for any ideals a, b ∈ R.

These results show that the sets V (E) satisfy the axioms for closed sets in
a topological space. As mentioned before, the resulting topology is called
the Zariski tolology.

[1] (21) Draw the pictures of Spec(Z), Spec(R), Spec(C[x]), Spec(R[x]), Spec(Z[x]).

[1] (22) For R a ring, consider {D(r)}r∈R (the basis of open sets for the Zariski
topology). Prove that

(a) D(r) ∩ D(s) = D(rs);

(b) D(r) = ∅ iff r is nilpotent;
(c) D(r) = X iff r is a unit;
p p
(d) D(r) = D(s) iff (r) = (s).

[1] (23) It is sometimes convenient to denote a prime ideal of R by a letter such

as x or y when thinking of it as a point of X = Spec(R). When thinking
of x as a prime ideal of R, we denote it by px (logically this is the same
thing). Show that

(a) the set {x} is closed in Spec(R) iff px is maximal;

(b) {x} = V (px );
 18 Rings, Ideals, and Maps

(c) y ∈ {x} iff px ⊆ py ;

(d) X is a T0 space.
[1] (24) A topological space X is said to be irreducible if X 6= ∅ and if every pair
of non-empty open sets in X intersect, or equivalently if every non-empty
open set is dense in X. Show that Spec(R) is irreducible iff the nilradical
of R is a prime ideal.
Chapter 2

Modules

1 Notation and Examples

For a commutative ring R, an abelian group (M, +) is an R-module if there
exists a map R × M → M defined by (r, m) 7→ rm satisfying the following
properties for all r, s ∈ R and m, n ∈ M :
(i) 1 · m = m
(ii) (r + s)m = rm + sm
(iii) r(m + n) = rm + rn
(iv) (rs)m = r(sm)
Example 32. Let R = Z. A Z-module is an abelian group.
Proof. Let M be a Z-module and n ∈ Z, m ∈ M . Property (ii) and (iv) force
that
n−times

 z }| {
··· + m



 m + if n > 0
n·m= 0 if n = 0


 (−m) + · · · + (−m) if n < 0
 |
 {z }
|n|−times

This action forces (i) and (iii) to hold.

Example 33. If R = k a field, then modules are exactly vector spaces over k.
Example 34. If f : R → S is a ring homomorphism, then
(a) S is an R-module;
(b) any S-module is an R-module.
In particular, R is a module over itself with respect to the usual multiplication.
20 Modules

Proof. For (a), if r ∈ R and s ∈ S, define rs = f (r)s. Similarly for (b), if r ∈ R

and m ∈ M , for M an S-module, define r · m = f (r)m. The R-module M is
said to be obtained from the S-module M by restriction of scalars.
Example 35. Let k be a field and R = k[t] a polynomial ring with indetermi-
nate t. Let V be an R-module. By examples 33 and 34, V is a vector space
over k. But this vector space comes with an action of t on V . That is,
t:V →V
v 7→ t · v.
Note by example 33, for any v1 , v2 ∈ V ,
t(v1 + v2 ) = tv1 + tv2 .
By example 34, if α ∈ k and v ∈ V ,
t(αv) = (tα)v = α(tv).
Therefore, t is an endomorphism from V to V ; i.e. a linear transformation.
The definition of a module force that if p(t) = α0 + α1 t + · + αn tn , then
p(t) · v = α0 v + α1 (tv) + · + αn (tn v).
The converse also holds: given a vector space V and a linear transformation
T : V → V we can make V into a k[t]-module by tv = T (v) for all v ∈ V .
Example 36. If R = k[t1 , . . . , tn ] is a polynomial ring over a field with n
variables, then an R-module M is a vector space V with n linear transformations
T1 , . . . , Tn which commute.
Example 37. Suppose {Mi } are R-modules. We define the direct sum of a
module as
M
Mi = {(mi ) | mi ∈ Mi , all but finitely many mi = 0}.
i

This is an R-module. Note that the direct product

Y
Mi = {(mi ) | mi ∈ Mi }
i

is not necessarily an R-module.

2 Submodules and Maps

Definition. If M, N are R-modules, then an R-homomorphism φ : M → N is
an homomorphism of abelian groups such that φ(rm) = rφ(m) for all r ∈ R.
An R-homomorphism is an isomorphism if it is surjective and injective. As
with rings, the kernel of an R-homomorphism is the set of all elements that get
mapped to zero.
2 Submodules and Maps 21

Example 38. Let M, N be R-modules. Define HomR (M, N ) to be the set of

R-homomorphisms from M to N . This becomes an R-module with addition
and multiplication defined as follows: for any f, g in HomR (M, N ), r ∈ R,

(f + g)(m) := f (m) + g(m)

(r · f )(m) := rf (m)

Note. If k is a field, V ' k n , Homk (V, V ) ' Mn (k) where Mn (k) is the set of
n × n matrices with elements from k.
L
Definition. An R-module F is a free module if F ' i Ri where Ri ' R. In
other words, F is free if it has a basis.

L vector space V over a field k is free. Choose a basis {vi }

Example 39. Every
of V . Then V ' i kvi , kvi ' k.

Definition. Suppose that M is an R-module. We say that {xi }i∈I , xi ∈ M ,

generate M if for all x ∈ M there exists an equality (not necessarily unique)
X
x= ri xi
i

such that ri ∈ R and all but finitely many ri are zero. The R-module M is said
to be finitely generated if I is a finite set. I.e., there exists x1 , . . . , xk ∈ M such
that M = {r1 x1 + · · · + rk xk | ri ∈ R} = Rx1 + · · · + Rxk = hx1 , . . . , xk i.

L Let M be an R-module with a generating set {xi }i∈I . In this case, let
Remark.
F ' i Ri where Ri ' R. Consider the elements ei = (0, 0, . . . , 1, 0, . . .) ∈ F ,
where the one is is the ith position. There exists a homomorphism F → M
defined by ei 7→ xi . Thus
X
(r1 , r2 , . . .) 7→ ri xi ∈ M.
i

This is a surjective map and well defined since F is free.

Submodules and Quotient Modules If M is an R-module, a submodule

N of M is a subgroup N ⊆ M such that the restricted operations make N an
R-module. If N is a submodule of M , we can define a quotient module, denoted
M/N as follows: the group structure is as a quotient group and r(m + N ) =
rm + N . For instance, if M = R, then the submodules are exactly the ideals.

Example 40. Kernels and images of homomorphisms are submodules of their

respective modules, e.g. if N ⊆ M , the natural projection π : M → M/N
defined by x 7→ x+N is an homomorphism with ker(π) = N and im(π) = M/N .

The next three theorems are generalized versions of proposition 1 and theo-
rem 2. They are stated without proof.
22 Modules

Theorem 13 (Isomorphism Theorem). Let R be a ring and f : M → N an

R-homomorphism. Set K = ker(f ). Then f factors as a surjection followed by
an injection as in the following commutative diagram:

M
f
/N
<
π
g

M/K

In particular, if f is onto, then g is an isomorphism.

Theorem 14. Let R be a ring and N ⊆ M be R-modules. Then there exists
a one-to-one inclusion preserving correspondence between submodules of M/N
and submodules of K ⊆ M such that N ⊆ K:

K ←→ K/N.

Theorem 15. Let N ⊆ K ⊆ M be R-modules. Define M := M/N and K :=

K/N ⊆ M . Then
M/K ' M /K.
Definition. If N, K ⊆ M are R-submodules and I ⊆ R, then we define
 
X 
IM = ij mj | ij ∈ I, mj ∈ M ⊆ M
 
j

N + K = {n + k | n ∈ N, k ∈ K}

Remark. The objects IM and N + K in definition 2 are submodules of the

R-module M . Further, N + K is the smallest submodule of M containing N
and K.
Theorem 16 (Diamond Isomorphism). If N, K ⊆ M are R-submodules, then
there is an isomorphism theorem:
N +K K
' .
N K ∩N
Proof. Let f be the composition map from K to N + K/N
 / N +K π / N + K/N
K
9
f

defined by k 7→ k + N . Notice that f is surjective and by theorem 13 we have

that
K/ ker(f ) ' N + K/N.
3 Tensor Products 23

Finally
ker(f ) = {k ∈ K : f (k) = 0} = {k ∈ K : k + N = N } = K ∩ N.

Lemma 17. Let R be a ring and M an R-module. Then HomR (R, M ) ' M .
Proof. Define a homomorphism Φ : HomR (R, M ) → M by Φ(φ) = φ(1). Check
that Φ is an isomorphism.
Remark. From the lemma we know that HomZ (Z, Q) ' Q. Let φ ∈ HomZ (Q, Z)
be a non-zero Z-homomorphism, so that there exists α ∈ Q such that φ(α) 6= 0.
Without loosing generality, say φ(α) = n, with n > 0. So,
1 1
n = φ(α) = φ(m · · α) = m · φ( α) ∈ Z.
m m
Since m is arbitrary, we have arrived at a contradiction. Hence HomZ (Q, Z) = 0.

3 Tensor Products
Definition. Let M, N, P be R-modules. An R-bilinear map f : M × N → P
defined by (m, n) 7→ f (m, n) is a map such that
(i) If we fix x ∈ M and then define fx : N → P by x 7→ f (x, n), then fx is
an R-module homomorphism.
(ii) Similarly, fixing y ∈ N and defining fy : N → P by m 7→ f (m, y) is an
R-module homomorphism.
Theorem 18. Given R-modules M, N , there exists an R-module T and a bi-
linear map g : M × N → T such that
(i) Given any other R-module P and bilinear map f : M × N → P , there
exists a unique R-module homomorphism α : T → P such that

M ×N
g
/T

α
f
# 
P
is a commutative diagram.
(ii) Further, T and g are unique in the following sense: If T 0 , g 0 is another
pair satisfying (i), then there exists an isomorphism i such that

M ×N
g
/T

g0
#  i
T0
is a commutative diagram as well.
24 Modules

Definition. The R-module T in the above theorem is called the tensor product
of M and N (over R) and is written T = M ⊗R N .
L
Proof. For (i), let F be a free module R(m,n) with m ∈ M , n ∈ N and
R(m,n) ' R. Denote theL element with a 1 in the (m, n)th slot and 0 elsewhere
by [m, n]. Hence F = R[m, n].
Let C be the R-submodule of F generated by all elements of the following
form: mi ∈ M , ni ∈ N , r ∈ R.

(1) [m1 + m2 , n] − [m1 , n] − [m2 , n]

(2) [rm, n] − r[m, n]

(3) [m, n1 + n2 ] − [m, n1 ] − [m, n2 ]

(4) [m, rn] − r[m, n]

Set T = F/C and define the map g : M × N → T by (m, n) 7→ [m, n] + C. g is

clearly bilinear. Define [m, n] + C by m ⊗ n. Now let P be an R-module and
f : M × N → P be a bilinear map.

M ×N
g
/T

f
#
P

Note that T is generated as an R-module by elements m⊗n since F is generated

as an R-module by [m, n]. P
A typical element in T looks like i mi ⊗ ni , not m ⊗ n. (ri (mi ⊗ ni ) = 
ri mi ⊗ ni = m0i ⊗ ni )
We have no choice for α: α(m ⊗ n)P = f (m, n) is forced,
P and then to make
this a homomorphism, we must set α( mi ⊗ ni ) = f (mi , ni ). To see α is
Φ
well-defined, first observe the map F → P : [m, n] 7→ f (m, n) gives an R-module
homomorphism. To see α is well-defined, it suffices to prove Φ(C) = 0. Then Φ
induces a well-defined homomorphism

T /C → P : t + C 7→ Φ(t) (m ⊗ n 7→ f (m, n)).

It’s enough to prove Φ sends the specified generators of C to zero, and it does.
For (ii), if T , g and T 0 , g 0 are two such tensor products, then by

M ×N
g
/T
L
∃! α
g0
# 
β
T0

it is easy to check α ◦ β = 1T 0 and β ◦ α = 1T .

3 Tensor Products 25

Example 41. Let F and L be two field extensions of a field K, and let F L be
their compositum, so that we have a diagram of fields

FL

F L

There exists a K-bilinear map from F × L → F L and therefore there exists a

K-homomorphism F ⊗K L → F L. This is an isomorphism if and only if one of
the two equivalent conditions hold.

(1) Every set {xi } of elements in F which are linearly independent over K are
linearly independent over L.

(2) Every set {yj } of elements in L which are linearly independent over K are
linearly independent over F .

Exact Sequences

Definition. A sequence of modules and homomorphisms

φi+1
··· / M i1 / Mi φi
/ Mi−1 / ···

is said to be exact if for all i, ker(φi ) = im(φi+1 ). It is said to be a complex if

im(φi+1 ) ⊆ ker(φi ), equivalently, φi ◦ φi+1 = 0. A short exact sequence (s.e.s.)
is a sequence
0 /A α /B β
/C /0

which is exact. This means

(1) α is injective, i.e. ker(α) = 0.

(2) β is surjective, i.e. im(β) = C.

(3) ker(β) = im(α)

Example 42. Given R-modules N ⊆ M ,

0 /N i /M π / M/N /0

is a s.e.s.

Proposition 19. Let R be a ring and M, N, Q, Mi , Ni be R-modules. Also let

I be an ideal in R.
26 Modules

(1) M ⊗ N ' N ⊗ M

(2) (M ⊗ N ) ⊗ Q ' M ⊗ (N ⊗ Q)

(3) R/I ⊗ M ' M/IM

L L
(4) ( i Mi ) ⊗ N ' i (Mi ⊗ N )

(5) If K is a field, V , W vector spaces, {vi } a basis of V and {wj } a basis of

W , then {vi ⊗ wj } are a basis of V ⊗K W . In particular,

dimK (V ⊗K W ) = (dimK V )(dimK W ).

(6) If IM = IN = 0, then M and N are R/I modules and M ⊗R/I N '

M ⊗R N .

(7) IF φ : M → N is a homomorphism, then there exists an induced homo-

morphism φ ⊗ 1 : M ⊗ Q → N ⊗ Q by (φ ⊗ 1)(m ⊗ q) = φ(m) ⊗ q.
α β
(8) If M1 −→ M2 −→ M3 −→ 0 is exact, then

M1 ⊗ N
α⊗1
/ M2 ⊗ N β⊗1
/ M3 ⊗ N /0

is exact (right exactness).

N
(9) Hom, adjointness

HomR (M ⊗R N, Q) ' HomR (M, HomR (N, Q))

Proof. (1) Notice the bilinear map M × N → T is exactly the same as the
bilinear map N × M → T by “flipping”. So both M ⊗ N and N ⊗ M solve the
same universal problem. M ⊗ N ' N ⊗ M where m ⊗ n 7→ n ⊗ m.
(2) Exercise.
(3) We will show that M/IM has the correct universal property. Denote
cosets in R/I by and define the map R/I × M → M/IM by (r, m) 7→ rm.
This is well-defined: r = s ⇔ r − s ∈ I ⇒ (r − s)m ∈ IM ⇒ rm = sm in
M/IM .
It is bilinear, e.g. (r + s, m) 7→ (r + s)m = rm + sm.
Does there exist a unique R-module homomorphism h such that

R/I × M / M/IM
h
g
# |
Q

commutes? Define α : M → Q by α(m) = g(1, m). This is a R-module

homomorphism by the bilinearity of g.
3 Tensor Products 27

Remark. In general, if h : M → N is an R-module homomorphism and if

I · h(M ) = 0, for I an ideal in R, then there exists an induced homomorphism
h such that the following diagram commutes.

M
h /N
;
π
# h
M/IM

Claim. I · α(M ) = 0
To see this, let m ∈ M and α(m) = g(1, m). For i ∈ I,

i · α(m) = i · g(1, m) = g(i · 1, m) = g(i, m) = g(0, m) = 0.

Hence there exists an induced R-module homomorphism α : M/IM → Q.

Further we have that

g(r, m) = r · g(1, m) = r · α(m) = r · α(m) = α(rm).

So the diagram commutes for any bilinear map g, that is, M/IM ' R/I ⊗R M .
(4) Exercise.
P (5) Note, in general,
P that using (4) and
P induction, if F is the free module
Rvi and G = Rwi then F ⊗R G = i,j R(vi ⊗ wj ). Using (3) with I = 0,
we see that R ⊗R N ' N . e.g.

R2 ⊗ R2 = (R ⊕ R) ⊗ R2 ' R ⊗ R2 ⊕ R ⊗ R2 ' R2 ⊕ R2 ' R4 .

Now apply this to vector spaces which are free k-modules to get (5).
(6) Exercise.
(7) Exercise.
(8) As an R-module, M3 ⊗R N is generated by “decomposable” tensors,
m ⊗ n, m ∈ M3 , n ∈ N . But there exists an x ∈ M2 such that β(x) = m and so
(β ⊗ 1)(x ⊗ n) = m ⊗ n and thus β ⊗ 1 is onto. Note that ker(β ⊗ 1) ⊇ im(α ⊗ 1).
For let x ∈ M1 , n ∈ N ;

(β ⊗ 1)((α ⊗ 1)(x ⊗ n)) = βα(x) ⊗ n = 0 ⊗ n = 0.

Therefore, there exists an induced map

M2 ⊗ N β⊗1
−→ M3 ⊗ N −→ 0.
im(α ⊗ 1)

It is enough to show β ⊗ 1 is an isomorphism. We will construct h : M3 ⊗ N →

M2 ⊗ N/im(α ⊗ 1) such that (β ⊗ 1) ◦ h = h ◦ (β ⊗ 1) = id. Consider the
R-bilinear map
M2 ⊗ N
g : M3 × N −→ .
im(α ⊗ 1)
28 Modules

Let x ∈ M3 , n ∈ N and define g(x, n) = y ⊗ n in M2 ⊗ N where β(y) = x. We

want to prove that g is well defined, i.e. if β(y) = x = β(z), then for n ∈ N

y ⊗ n + im(α ⊗ 1) = z ⊗ n + im(α ⊗ 1).

This can be restated as (y − z) ⊗ n = y ⊗ n − z ⊗ n ∈ im(α ⊗ 1). However

β(y − z) = β(y) − β(z) = x − x = 0. So there exists u ∈ M1 such that
α(u) = y − z. Then (y − z) ⊗ n = α(u) ⊗ n = (α ⊗ 1)(u ⊗ n) ∈ im(α ⊗ 1), and
therefore g is well defined. Then g induces a R-homomorphism

h : M3 ⊗ N / M2 ⊗ N
im(α ⊗ 1)
 / g(x, n)
x⊗n

Now let x ∈ M3 , n ∈ N . If we compute (β ⊗ 1)(h(x ⊗ n)) where β(y) = x, we

have

(β ⊗ 1)(h(x ⊗ n)) = (β ⊗ 1)(y ⊗ n)

= β(y) ⊗ n
= x ⊗ n,

so that (β ⊗ 1)h = id. Conversely

h((β ⊗ 1)(y ⊗ n)) = h(β(y) ⊗ n)

= y ⊗ n,

and therefore β ⊗ 1 is an isomorphism, that is im(α ⊗ 1) = ker(β ⊗ 1) and the

sequence is exact.
Example 43. For x ∈ R, x is said to be a non-zero divisor if the map from R
to R defined by r 7→ rx is one-to-one. (i.e. rx = 0 ⇒ r = 0) Let x ∈ R be a
non-zero divisor. Then
x
0 −→ R −→ R −→ R/Rx −→ 0

is an exact sequence. Now let M be an R-module. So

R ⊗R M
x / R ⊗R M / R/xR ⊗R M / 0,

that is
M
x /M / M/xM / 0,

is exact. Notice that if x is also a non-zerodivisor on M , then

0 /M x /M / M/xM /0

is also exact.
4 Operations on Modules 29

x
Remark. The map M → M being one-to-one is equivalent to x being a non-zero
divisor of M . In fact, explicitly, this means xm = 0 ⇒ m = 0 for m ∈ M . For
example, if M = R/Rx then

R/xR
x / R/xR

is the zero map, hence not one-to-one.

Remark. Since M ⊗ N ' N ⊗ M , HomR (M ⊗ N, Q) ' HomR (N, HomR (M, Q)).
(9) Homomorphisms from M ⊗ N to Q correspond to bilinear maps from M × N
to Q. A bilinear map from M × N to Q is a linear map from M to HomR (N, Q).
That is exactly an element of HomR (M, HomR (N, Q)).

4 Operations on Modules
Definition. In analogy with the colon operation defined for ideals, given an
ideal I ⊆ R and R-modules N ⊆ M we define the R-submodule
N :M I = {m ∈ M : mI ⊆ N } ⊆ M.
Similarly, given N, L ⊆ M two R-submodules we can define the ideal
N :R L = {r ∈ R : rL ⊆ N } ⊆ R.
In particular, if N = 0 ⊆ M and L = M we define the annihilator of M as
ann(M ) = 0 :R M = {r ∈ R : rM = 0} ⊆ R.
Definition. If M is an R-module, M is said to be flat if whenever
α β
0 −→ N1 −→ N2 −→ N3 −→ 0
is a short exact sequence of R-modules, then
α⊗1 β⊗1
0 −→ N1 ⊗ M −→ N2 ⊗ M −→ N3 ⊗ M −→ 0
is also exact; equivalently, α ⊗ 1 is one-to-one. If f : A → B is a homomorphism
of rings, we say f is flat homomorphism if B is a is a flat A-module (B is an
A-module via f : a · b = f (a) · b).
Example 44. Any free module is flat.
Remark. M is finitely generated if and only if there exists a free module Rk
mapping onto M.
φ
Proof. If M is finitely generated and Rk −→ M −→ 0, φ(e P i ) = xi , then ker(φ)
is a submodule of Rk and ker(φ) = {(r1 , . . . , rk ) ∈ Rk | i ri xi = 0} where
φ
0 −→ ker(φ) −→ Rk −→ M −→ 0.
30 Modules

Note. Elements in the kernel of φ are called syzygies.

Theorem 20 (Nakayama’s Lemma, NAK). Let R be a ring, I ⊆ R and M
a finitely generated R-module. If IM = M , then there exists x ∈ I such that
(1 + x)M = 0.
Remark. If there exists an x ∈ I such that (1 + x)M = 0 then for all u ∈ M ,
u = −xu ∈ IM . So M = IM .
Proof. Let M = hm1 . . . mk i. For each 1 6 i 6 k, we can write
X
mi = xij uj uj ∈ M, xij ∈ I.
j
P
But since M is finitely generated, we have uj = l rjl ml where rjl ∈ R. There-
fore,
X
mi = xij uj
j
!
X X
= xij rjl ml
j l
 
X X
=  xij rjl  ml
l j
X
= yil ml
l
= yi1 m1 + · · · + yik mk
P
where yil = j xij rjl . Thus we have a system of linear equations
    
−y11 + 1 −y12 ··· −y1k m1 0
 −y21 −y 22 + 1 ··· −y2k    m2   0
   
= .
 
.. .. .. .
.. . ..
  ..  
  
 . . . . 
−yk1 −yk2 ··· −ykk + 1 mk 0
Recall that for a square matrix A, adj(A) · A = det(A) · I where I is the
identity matrix. Let A be the above matrix composed of the yik ’s. Hence
   
m1 0
adj(A) · A  ...  =  ... 
   

mk 0
or, by the recollection,
 
  
det(A) 0 m1 0
..   ..   ..  .
 .  = 

 . . 
0 det(A) mk 0
4 Operations on Modules 31

Thus for 1 6 i 6 k we have det(A) · mi = 0. This forces det(A) · M = 0. Now

modulo I the matrix A is the identity matrix. Hence det(A) ≡ 1 (mod I), i.e.
there exists an x ∈ I such that det(A) = 1 + x.
Corollary 21. If I ⊆ jac(R) and M is finitely generated with IM = M , then
M = 0.
Proof. By NAK, there exists an x ∈ I such that (1 + x)M = 0. So it is enough
to show that 1 + x is a \
unit. This is true if and only if 1 + x is not in any
maximal ideal. But x ∈ m, m maximal, implies 1 + x is not in any maximal
ideal.
Corollary 22. Suppose that I ⊆ jac(R). Assume M is an R-module, N ⊆ M
and M/N is finitely generated. If M = N + IM then M = N .
Proof. Notice that I(M/N ) = M/N . Let u ∈ M and consider
P u + N ∈ M/N .
Write u P= n + y for n ∈ N and y ∈ IM where y = ai yi . Then u + N =
y +N = ai (yi +N ) ∈ I(M/N ). Apply NAK to M/N , using the first corollary
we get that M/N = 0.
32 Modules

Exercises
Chapter 3

Localization

1 Notation and Examples

A subset W ⊆ R is said to me multiplicatively closed if 1 ∈ W and for any two
elements w1 , w2 ∈ W , w1 w2 ∈ W . Two main examples of this are as follows.

Example 45. If x ∈ R, not nilpotent, then W = {xn }∞

n=0 is multiplicatively
closed.

Example 46. If p ∈ SpecR the W = R r p is a multiplicatively closed set. In

particular, if R is a domain, then W = R r {0} is multiplicatively closed as well.

The idea in this chapter is to construct a ring in which all elements of W

become units, i.e., solving the following universal problem: If g : R → S is a
ring homomorphism such that g(w) are units in S for all w ∈ W , then there is a
unique ring L, and a unique homomorphism h such that the following diagram
commutes.
R /L

g
  h
S

Construction For a ring R, M an R-module and W a multiplicatively closed

set, consider elements in the set R×W (respectively M ×W ). Put an equivalence
relation ∼ on R × W by: (r1 , w1 ) ∼ (r2 , w2 ) if and only if there exists a w ∈
W such that (r1 w2 − r2 w1 )w = 0. This does define an equivalence relation:
Reflexivity and symmetry are clear. For transitivity, let (r1 , w1 ) ∼ (r2 , w2 ) and
(r2 , w2 ) ∼ (r3 , w3 ). Hence there exists a w, w0 ∈ W such that (r1 w2 −r2 w1 )w = 0
and (r2 w3 −r3 w2 )w0 = 0. Notice that ww0 [w3 (r1 w2 −r2 w1 )+w1 (r2 w3 −r3 w2 )] =
0. Simplifying we find (r1 w3 − r3 w1 )w2 ww0 = 0.
We denote the equivalence class or (r, w) by wr or w−1 r ∈ R × W/ ∼ (simi-
−1
larly, mw ). The set of equivalence classes in R × W/ ∼ we will denote W R(
34 Localization

or RW ) and similarly, M × W/ ∼ denoted W −1 M . We have to make W −1 R a

ring by defining
r1 r2 r1 w2 + r2 w1
+ :=
w1 w2 w1 w2

r1 r2 r1 r2
· := .
w1 w2 w1 w2
We make W −1 M a module over this ring W −1 R by defining
m1 m2 m1 w2 + m2 w1
+ :=
w1 w2 w1 w2

r m rm
· := .
w1 w2 w1 w2
Exercise 1. Check for the well-defined property then show that W −1 R is really
a commutative ring and W −1 M is really a module over W −1 R.
Remark. The identity of W −1 R is 11 or w
w for all w ∈ W . Also, there exists a
ring map (sometimes called the canonical map) R → W −1 R defined by r 7→ 1r .
Definition. Let R be a ring and M , R-module and W a multiplicatively closed
set. The localization of R with respect to W , is the ring W −1 R. Similarly, the
localization of M with respect to W is the W −1 R-module W −1 M .
Remark. If W = {xn }∞ n=0 (see example 45) then we denote W
−1
R by Rx .
Likewise if W is a complement of a prime p (see example 46) then we denote
W −1 R by Rp .
Proposition 23. The ring W −1 R has the following universal property:

R / W −1 R

∃!g
f
 |
S

Given a ring homomorphism such f (w) is a unit in S for all w ∈ W , then

there exists a unique ring homomorphism g : W −1 R → S making the diagram
commute.
Proof. Define g( wr ) = f (w)−1 f (r). This is forced since
w r r
g(r) = g( · ) = g(w) · g( ),
1 w w
but g(w) = f (w) and g(r) = f (r) are forced. Therefore f (w)−1 f (r) = g( wr ) is
forced.
The reader should check that g really is a ring homomorphism.
1 Notation and Examples 35

Remark. Let U , W be multiplicatively closed in R. The set U W = {uw | u ∈

U, w ∈ W } is also multiplicatively closed. (Caution, it could happen that 0 ∈
U W ). Then
(U W )−1 R ' U −1 (W −1 R)
This can be checked since the right hand side has the same universal property
as the left hand side.
Remark. If 0 ∈ W then W −1 R = 0
Theorem 24. Let M , N , L be R-modules and W a multiplicatively closed set.
α β W α βW
(1) If M −→ N −→ L is an exact sequence, then MW −→ NW −→ LW is
also an exact sequence of RW -modules.
(2) M ⊗R RW ' MW
x α
Proof. First given α : M → N , define αW : MW → NW by αW ( w )= w . This
n
is a well-defined RW -module homomorphism. Let w ∈ ker(βW ) where n ∈ N
and w ∈ W . Since
0  n  β(n)
= 0 = βW =
1 w w
there exists a u ∈ W such that u(0 · w − β(n)) = 0. Hence un ∈ ker(β) = im(α)
because uβ(un) = 0 implies β(un) = 0. So there exists m ∈ M such that
α(m) = un. Then
 m  α(m) un n
αW = = =
uw uw uw w
and therefore ker(βW ) ⊆ im(αW ).
n
Conversely, if w ∈ αW ( m
u ), then
n m β ◦ α(m) 0
βW = βW ◦ αW = = = 0.
w u u u
To show the second part, define θ : M × RW → Mw by θ(m, ur ) = rm u where
m ∈ M , u ∈ W , and r ∈ R. This is a bilinear map, so we get an induced map
φ
M ⊗ RW −→ MW
r rm
m ⊗ 7→ .
u u
Claim 1. The map φ is one-to-one and onto.
To show φ is onto, just note that φ(m ⊗ u1 ) = m
u . Now suppose that

n
!
X ri
φ mi ⊗ = 0.
i=1
ui

Fist note that with out any loss of generality all the ui ’s are the same, say u.
Then X ri X 1 X  1
mi ⊗ = mi ri ⊗ = mi ri ⊗
u u u
36 Localization

which is of the form m ⊗ u1 . Thus we have φ(m ⊗ u1 ) = 0 so m

u = 0. But this is
true if and only if there exists a w ∈ W such that wm = 0. Then
1 w 1 1
m⊗ =m⊗ = wm ⊗ =0⊗ = 0.
u wu wu wu

Corollary 25. The map R → RW is flat.

Proof. This follows from the canonical isomorphisms of the tensor product and
the definition of a flat homomorphism (see definition on page 29).

Corollary 26. If I is an ideal, M an R-module, and W a multiplicatively closed

set, then
W −1 (M/IM ) ' W −1 M/I(W −1 M ).
In particular,
W −1 (R/I) ' W −1 R/W −1 I.
In other words, localization commutes with quotients.

Proof. Apply the theorem to the left hand side and then use the canonical
isomorphisms of the tensor product to obtain the right hand side.

Example 47. Given the polynomial ring k[x, y] in two indeterminates over a
field k, the structure of the ring
 
k[x, y]
(xy) x

can be determined using the previous corollary. That is,

k[x, x−1 , y]
 
k[x, y]
' .
(xy) x (xy)k[x, x−1 , y]

Notice that (xy)k[x, x−1 , y] = (y)k[x, x−1 , y]. So substituting and applying the
corollary again we obtain
 
k[x, y]
' k[x]x ' k[x, x−1 ].
(xy) x

2 Ideals and Localization

Example 48. Let R be a domain with W a multiplicatively closed set in R,
/ W . Then W −1 R is still a domain.
0∈
0
Proof. Suppose wr · wr 0 = 0. This means there exists w00 ∈ W such that w00 rr0 = 0.
0
Since w00 6= 0, either r = 0 or r0 = 0. Thus either wr = 0 or wr 0 = 0.
2 Ideals and Localization 37

Definition. If R is a domain and W = R − {0}, then W −1 R is a field called

the field of fractions of R, e.g. if R = Z, then the field of fractions is Q.
Theorem 27. Let R be a ring, W a multiplicatively closed set of R.
(1) If I is an ideal of R then W −1 I is an ideal of W −1 R where
 
i
W −1 I = | i ∈ I, w ∈ W .
w

(2) If J is an ideal in W −1 R then

 
i
J ∩R= i∈R| ∈J .
1

(3) The previous two statements give a one-to-one inclusion preserving corre-
spondence between Spec(W −1 R) and primes p in R such that p ∩ W = ∅
Proof. For the first part, use the definition of addition and multiplication in the
ring W −1 R.
In general, if φ : R → S is a ring homomorphism, J ⊆ S, then φ−1 (J) is an
ideal in R. Apply this to the canonical map from R into W −1 R and we have
that φ−1 (J) = J ∩ R is an ideal. Further, if wi ∈ J then w wi = 1i implies that
i ∈ J ∩ R. Thus wi ∈ W −1 (J ∩ R) and the second statement follows.
For the third statement, consider q ∈ Spec(R) and q ∩ W = ∅. Then since

W −1 (R/q) ' W −1 R/W −1 q

we can apply the above general remark to the domain R/q. We see that
W −1 (R/q) is a domain and so W −1 q ∈ Spec(W −1 R). (W ∩ q = ∅ implies
that W −1 q = W −1 R)
Conversely, if Q ∈ Spec(W −1 R) then Q ∩ R = q is prime. (This is true for
general homomorphisms) By (2) we have that W −1 q = Q.
Remark. If I1 and I2 are ideals, it is possible that W −1 I1 = W −1 I2 without
I1 = I2 .
Example 49. Let R = k[x, y] be a polynomial ring in two variables over a
field and W = {y n }n>0 . Consider the ideals I1 = (x2 , xy) and I2 = (x). Then
W −1 I1 = W −1 I2 but W −1 I1 ∩ R = (x).
Remark. In general, if I ⊆ R is an ideal, then
[
W −1 I ∩ R = {r ∈ R | ∃ w ∈ W with w · r ∈ I} = I : w.
w∈W

This contains I if I is not prime. If I is prime,

[
I : w = I.
w∈W
38 Localization

Corollary 28. If p ∈ Spec(R) then Rp is local with maximal ideal pRp .

Proof. We know that Spec(Rp ) is the set of primes Q in R such that Q∩Rp = ∅,
i.e. primes Q such that Q ⊆ p.

Proposition 29. Let R be a ring, R[x] a polynomial ring over x. Suppose

Q1 ⊆ Q2 ⊆ Q3 for prime ideals of R[x]. If

Q1 ∩ R = Q2 ∩ R = Q3 ∩ R

then Q1 = Q2 or Q2 = Q3 .

Proof. Set q = Q1 ∩ R = Q2 ∩ R = Q3 ∩ R. Then

R/q[x] ' R[x]/qR[x].

Also set Q01 = Q1 /qR[x], Q02 = Q2 /qR[x], Q03 = Q3 /qR[x]. In R/q we have

Q01 ∩ R/q = Q02 ∩ R/q = Q03 ∩ R/q,

so without loss of generality we can assume R = R/q is a domain with Q1 = Q01 ,

Q2 = Q02 and Q3 = Q03 and

Q1 ∩ R = Q2 ∩ R = Q3 ∩ R = 0.

Let W = R r {0}. Then W ∩ Qi = ∅, therefore

W −1 Q1 ⊆ W −1 Q2 ⊆ W −1 Q3

is a chain of primes in W −1 (R[x]) = R[x] ⊗R W −1 R = (W −1 R)[x]. But W −1 R

is a field, so (W −1 R)[x] is a polynomial ring over a field, and this means:

Spec((W −1 R)[x]) = {0} ∪ {f (x)},

where f (x) 6= 0 is a irreducible polynomial. Hence the longest chain of primes

has length two (because (W −1 R)[x] has Krull dimension equal to one), therefore
W −1 Q1 = W −1 Q2 or W −1 Q2 = W −1 Q3 , which implies

Q1 = Q2 or Q2 = Q3 .

Theorem 30 (Local - Global Principle). Let R be a ring and M and R-module.

The following are equivalent:

(1) M = 0;

(2) Mp = 0 for all p ∈ Spec(R).

(3) Mm = 0 for all m maximal in Spec(R).

2 Ideals and Localization 39

Proof. Clearly (1) ⇒ (2) ⇒ (3). Now assume (3) and by way of contradiction
suppose M 6= 0, i.e. there exists x ∈ M , x 6= 0. This means ann(x) = {y ∈
6 R, therefore there exists a maximal ideal m ∈ Spec(R) such that
R : yx = 0} =
ann(x) ⊆ m.
x
Claim. 1 6= 0 in Mm .

Proof of the Claim. Assume x1 = 0, then there exists w ∈ R r m such that

xw = 0 in R. Therefore w ∈ ann(x) ⊆ m, which is a contradiction.

So the claim holds, but this is a contradiction since we assumed Mm = 0.

Corollary 31. Let f : M → N be an homomorphism of R-modules. Then f

is injective (respectively surjective, isomorphism) if and only if the homomor-
phisms fp : Mp → Np are injective (respectively surjective, isomorphism) for all
p in Spec(R).

f (m)
Proof. Note that fp m s = s for s ∈
/ p. Also

 f is injective ⇐⇒ ker f = 0.
f is surjective ⇐⇒ cokerf = 0.
f is isomorphism ⇐⇒ ker f = cokerf = 0.


But the Local-Global Principle says that it is enough to check (ker f )p and
(cokerf )p for all p ∈ Spec(R). Finally, since ⊗R Rp is flat, we get (ker f )p =
ker fp and (cokerf )p = cokerfp . Hence

 f is injective ⇐⇒ fp is injective for all p ∈ Spec(R).
f is surjective ⇐⇒ fp is surjective for all p ∈ Spec(R).
f is isomorphism ⇐⇒ fp is isomorphism for all p ∈ Spec(R).


Remark. If M is generated by x1 , . . . , xn , then for all multiplicatively closed

sets W , W −1 M is generated by x11 , . . . , x1n

Proof. Assume x1 , . . . , xn generate M , then there is a surjective R-homomorphism

ϕ : Rn → M
ei 7→ xi

ϕ is surjective, therefore ϕW is also surjective, i.e.

ϕW : W −1 Rn → W −1 M
ei xi
1 7−→ 1

e1 x1 en
is a presentation for W −1 M , which is generated by ϕW



1 = 1 , . . . , ϕW 1 =
xn
1 .
40 Localization

Remark. If R is local and M is finitely generated then all minimal generating

sets of M have the same number of elements; namely the dimension of M/mM
over R/m where m is the unique maximal ideal.
Proof. By Nakayama’s Lemma M = Rx1 + . . . Rxn if and only if M = Rx1 +
. . . + Rxn + mM . Therefore x1 , . . . , xn minimally generate M if and only if their
images x1 , . . . , xn form a k = R/m-basis for the vector space M/mM .
Example 50. Suppose R has nontrivial idempotents, say e. Then R itself is
generated by {1}, but it is also minimally generated by {e, 1 − e}
Definition. If M is an R-module, the support of M is the set of prime ideals
p in Spec(R) such that Mp 6= 0. This set is denoted Supp(M ).
Proposition 32. If M is finitely generated then

Supp(M ) = {p ∈ Spec(R) | p ⊇ ann(M )} = V (ann(M ))

Proof. Mp = 0 if and only if there exists s ∈/ p such that sM = 0 if and only if

there exists s ∈
/ p such that s ∈ ann(M ) if and only if ann(M ) 6⊆ p.

3 UFD’s and Localization

Theorem 33. Let R be a ring and let W be a multiplicatively closed set.
(1) If R is UFD then W −1 R is UFD.
(2) Suppose there exists a set of prime elements Λ = {xi } ⊆ W which are
NZD such that every element w ∈ W can be written as
Y
w= xai i .
i

If W −1 R is UFD and R satisfies ACC (Ascending Chain Condition) then

R is UFD.
To prove this theorem we need some further results.
Lemma 34. Let R, W and Λ be as above. Then every r ∈ R can be written in
the form r = wr0 for some w ∈ W and xi 6 | r0 for all xi ∈ Λ.
Proof. If xi 6 | r for all xi ∈ Λ then take r0 = r and w = 1. Otherwise r = r1 xi
for some r1 ∈ R, xi ∈ Λ. Note that (r) ( (r1 ). Repeat the process with r1 in
place of r. Inductively we get a chain

(r) ( (r1 ) ( (r2 ) ( . . .

which must stabilize at some point since R is Noetherian. If (rn ) = (rn+1 ) this
means that xi does not divide rn for all xi ∈ Λ. But then r = wrn for some
w ∈ W by construction.
3 UFD’s and Localization 41

Lemma 35. Suppose that z1 ⊆ W −1 R is a prime ideal. Write z = wz 0 as in

Lemma 34. Then

 z   z0   0
z
= and ∩ R = (z 0 ) is prime in R.
1 1 1

 0
Proof. Clearly z1 = z1 since w1 is invertible in W −1 R. Also (z 0 ) ⊆ R is a
 0
prime since z1 is a prime in W −1 R. Moreover we have:

z0
  [
∩R= (z 0 : w),
1
w∈W

therefore we are done if we prove that (z 0 : w) = (z 0 ) for all w ∈ W . Let

r ∈ (z 0 : w), then there exists s ∈ R such that wr = z 0 s. By
Q definition of Λ in
Theorem 33 there exists a1 , . . . , an integers such that w = i xai i . Therefore

xa1 1 · . . . · xann r = z 0 s.

Hence xi |z 0 s but xi 6 | z 0 for all i, and since the xi ’s are prime we get xi |s for
all i. So we can cancel the xi ’s one at a time to get r = z 0 t for some t ∈ R, i.e.
r ∈ (z 0 ).
Proof of Theorem 33. (1) Recall that to prove that a domain is UFD we only
need to show that every irreducible element is prime, provided we have ACC
(Ascending Chain Condition) on principal ideals. R is UFD, hence a domain,
therefore W −1
R is a domain. If x ∈ R is irreducible then x is prime, therefore
W −1 (x) = x1 is prime in W −1 R provided W ∩ (x) = ∅. Set

W 0 := {x ∈ R : ∃w ∈ W, x|w}

Claim. W −1 R = (W 0 )−1 R
Proof of the Claim. Let x ∈ W 0 and write xy = w ∈ W . in W −1 R w is a unit,
then
1 = w−1 xy = x(w−1 y) = y(w−1 x)
so x and y are also units. This means that W −1 R satisfies the same property
as (W 0 )−1 R and therefore they are isomorphic.
By the Claim we can assume without loss of generality that W = W 0 . Let
b
w ∈ W −1 R and write b = x1 · . . . · xn a product of primes (R is UFD). Then

b 1 x1 xn
= ... .
w w 1 1
Now x1i is either a unit (if xi ∈ W 0 ) or a prime element (if xi ∈
/ W 0 ).
(2) It suffices to show that every irreducible element r ∈ R is prime. There are
two cases:
42 Localization

(1) r
1is a unit in W −1 R, i.e. r ∈ W . Since every element in W is a product
of xi ’s in Λ, but also r is irreducible, we have that r = xj for some j, and
hence it is prime.
z 
(2) 1r is not a unit, then there exists a prime ideal ⊆ W −1 R such that
1
r  z 
∈ . Therefore there exist s ∈ R and w ∈ W such that
1 1
r z s
= ⇒ wr = zs.
1 1w
Without loss of generality we can replace z with z 0 of Lemma 35 to assume
xi 6 | z for all xi ∈ Λ. By Lemma 35 z is prime and z|wr but z 6 | w, therefore
z|r. Finally r is irreducible, hence (r) = (z) is prime.

Corollary 1. If R is a UFD, so is R[x1 , . . . , xn ].

Proof. By induction we can assume n = 1 (write x := x1 ). Let W := R r {0}.

Since R is UFD, W satisfies the conditions of Theorem 33 (2), and it satisfies
them not only for R but also for R[x]. Therefore R[x] is UFD if W −1 R[x] is
UFD. Finally W −1 R[x] = (W −1 R)[x] = K[x] where K = W −1 R is a field,
therefore it is a PID and hence a UFD.

Example 51. Consider

C[x1 , x2 , x3 , x4 ] C[u, v, s, t]
R := ' =: S
(x21 + x22 + x23 + x24 ) (uv − st)

via the isomorphism u = x1 + ix2 , v = x1 − ix2 , s = x3 + ix4 , t = x3 − ix4 . S

is clearly not UFD, and so is R.
Example 52. Consider n ≥ 5 and

C[x1 , . . . , xn ] C[u, v, x3 , . . . , xn ]
R := ' =: S.
(x21 + . . . + x2n ) (uv + x23 + . . . x2n )

Then u is clearly prime since S/uS ' C[v, x3 , . . . , xn ] is a domain. Also Su is a

UFD, hence S is a UFD by Theorem 33 (2), and so is R.
3 UFD’s and Localization 43

Exercises
(1) What is the cardinality of (Z200 )6 ?
(2) Atiyah ch 3: 1,2,5,12,13
Chapter 4

Chain Conditions

Definition. Let (S, ≤) be a partially ordered set. Then S satisfies:

(1) The ascending chain condition (ACC) if every ascending chain
s1 ≤ s2 ≤ . . .
of elements in S stabilizes, i.e. there exists n ∈ N such that sn = sn+1 =
sn+2 = . . ..
(2) The descending chain condition (DCC) if every descending chain
t1 ≥ t2 ≥ . . .
of elements in S stabilizes, i.e. there exists m ∈ N such that tm = tm+1 =
tm+2 = . . ..

1 Noetherian Rings
Definition. A ring R is Noetherian if the set of all ideals satisfies ACC with
respect to the inclusion. This means that every ascending chain of ideals has a
maximal element, i.e., if
I1 ⊆ I2 ⊆ · · · ⊆ Ii ⊆ Ii+1 ⊆ · · ·
is a chain of ascending ideals Ij , then there exists n sufficiently large such that
In = In+1 .
Proposition 36. The ring R is Noetherian if and only if every ideal in R is
finitely generated.
Proof. Assume R is Noetherian and let I be and ideal in R. Let f1 ∈ I. If
(f1 ) = I we are done. If not, choose f2 ∈ I \ (f1 ). If (f1 , f2 ) = I then stop.
Incuctively, we have a chain,
(f1 ) ⊆ (f1 , f2 ) ⊆ (f1 , f2 , f3 ) ⊆ · · · .
1 Noetherian Rings 45

Since R is Noetherian, this chain stops and it can only stop when I is generated
by these elements.
Conversely, if we have an ascending chain of ideals

I1 ⊆ I2 ⊆ · · · ⊆ Ii ⊆ Ii+1 ⊆ · · · ,

let J = ∪∞ i=1 Ii . This is an ideal, hence J = (f1 , . . . , fn ), and there exists an N

sufficiently large such that (f1 , . . . , fn ) ⊆ IN . Therefore

J ⊆ IN ⊆ IN +1 ⊆ · · · ⊆ J.

So IN = IN +1 .
Example 53. Examples of Noetherian Rings:
(1) The integers Z.
(2) Any field.
(3) If k is a field, then k[x] is Noetherian.
Theorem 37 (Hilbert Basis Theorem). If R is a Noetherian ring, then R[x1 , . . . , xn ]
is Noetherian.
Proof. By induction on n, it suffices to prove the case when n = 1 since
R[x1 , . . . , xn ] = R[x1 , . . . , xn−1 ][xn ]. We want to prove R[x] is Noetherian.
Let
f (x) = r0 + r1 x + r2 x2 + · · · + rn xn
be an element of R[x] such that rn 6= 0. Define in(f ) = rn . If I ⊆ R[x] is an
ideal, then
in(I)j = {in(f ) | f ∈ I, deg(f ) 6 j} ∪ {0}.
Notice that in(I)j is an ideal in R: take a ∈ in(I)j and r ∈ R, then if ra = 0
clearly ra ∈ in(I)j . If ra 6= 0 then we have

axi + lower degree terms

is an element of I, with i 6 j, therefore

(ra)xi + lower degree terms ∈ I,

which means ra ∈ in(I)j . To prove that ∈ Ij is closed under the sum pick
a, b ∈ in(I)j , then there exist

f (x) = axi + . . . g(x) = bxk + . . .

with k ≤ i ≤ j without loss of generality. Then a + b = in(f + xi−k g) which has

degree i ≤ j.
Now let I ⊆ R[x] be an ideal. Notice:

in(I)0 ⊆ in(I)1 ⊆ in(I)2 ⊆ . . .

46 Chain Conditions

is an ascending chain of ideals in R, which is Noetherian by assumption. There-

fore there exists N ∈ N such that

in(I)N = in(I)N +1 = . . .

Also in(I)j is finitely generated for all 0 ≤ j ≤ N , hence choose generators

rj1 , . . . , rjmj for ∈ Ij , for all 0 ≤ j ≤ N . Pick now fji ∈ I such that in(fji ) = rji .
Claim. I = (f01 , . . . , f0m0 , . . . , fN 1 , . . . , fN mN ) =: J.
Proof of the Claim. Clearly J ⊆ I. Conversely assume by way of contradiction
that I 6= J. Then choose f ∈ I of least degree such that f ∈
/ J. If deg f = k,
then in(f ) ∈ in(I)k . There are two cases:
• k ≥ N : under this assumption in(f ) ∈ in(I)N = (rN 1 , . . . , rN mN ). Write:
mN
X
in(f ) = sl rN l
l=1

with sl ∈ R and consider

mN
X
g := f − sl xk−N fN l .
l=1

This polynomial has coefficient zero in degree k, since

mN
X XmN
in(f ) = sl rN l = in( sl xk−N fN l ).
l=1 l=1

By minimality in the choice of f we have g ∈ J, therefore:

mN
X
f =g+ sl xk−N fN l ∈ J
l=1

which is a contradiction. Therefore I = J in this case.

• If k < N proceed as in the previous case: pick f of least degree such that
f ∈ I r J and write
mk
X
in(f ) = sl rkl .
l=1

Again cancel the leading term of f , which is in(f )xk , using fk1 , . . . , fkmk :
mk
X
g := f − sl fkl ∈ J
l=1

and therefore f inJ, contradiction. Hence again I = J.

2 Noetherian Modules 47

So I is finitely generated and R[x] is Noetherian.

Remark. If R is Noetherian and I ⊆ R is an ideal, then R/I is Noetherian.
Remark. If R is Noetherian and W is multiplicatively closed, then RW is Noethe-
rian.
Proof. Let J1 ⊆ J2 ⊆ . . . be an ascending chain of ideals in RW . Then there
exists N ∈ N such that JN ∩ R = JN +1 ∩ R = . . . which implies
JN = (JN ∩ R)RW = (JN +1 ∩ R)RW = JN +1 = . . . .

2 Noetherian Modules
Definition. Let R be a ring and let M be a R-module. The following are
equivalent:
(1) Every submodule of M is finitely generated.
(2) M satisfies ACC on submodules.
(3) Any ordered set of submodules has a maximal element with respect to
containment.
Such a module M is said to be Noetherian.
Proof. The proof is the same as the one given for ideals.
Proposition 38. (1) If N ⊆ M is a submodule and M is Noetherian, then
M/N is Noetherian.
(2) If N ⊆ M is a submodule and both N and M/N are Noetherian, then M
is Noetherian.
Proof. (1) Immediate from the definition and the 1-1 correspondence:
1−1
{K/N ⊆ M/N submodule} ←→ {N ⊆ K ⊆ M submodule}.
(2) Suppose we have an ascending chain of submodules of M :
M1 ⊆ M2 ⊆ . . .
Then consider:
M 1 ∩ N ⊆ M2 ∩ N ⊆ . . . ⊆ N
and
M1 + N M2 + N
⊆ ⊆ . . . ⊆ M/N.
N N
By assumption there exists n ∈ N such that
Mn + N Mn+1 + N
Mn ∩ N = Mn+1 ∩ N and = .
N N
48 Chain Conditions

Claim. Mn = Mn+1 .
Proof of the Claim. It is enough to show that if x ∈ Mn+1 , then x ∈ Mn . Notice
that x + N ∈ Mn+1N
+N
= MnN+N , hence there exists y ∈ Mn such that

x + N = y + N.

Therefore x − y ∈ N and x − y ∈ Mn+1 , and so:

x − y ∈ Mn+1 ∩ N = Mn ∩ N.

Finally, since y ∈ Mn and x − y ∈ Mn :

x = y + (x − y) ∈ Mn .

Proposition 39. Let R be a Noetherian ring and let M be a R-module. The

following are equivalent:
(1) M is Noetherian.
(2) M is finitely generated.
Proof. (1) ⇒ (2) Since M is Noetherian every submodule, in particular M itself,
is finitely generated.
(2) ⇒ (1) Let M =< x1 , . . . , xn > and consider the map:

f : Rn → M
ei 7→ xi

where < e1 , . . . , en > is the standard basis of Rn . R is a Noetherian, so is

a Noetherian R-module. Therefore Rn is Noetherian and Rn / ker f ' M is
Noetherian too.
Remark 1. If M is Noetherian, then R/ann(M ) is Noetherian.
Proposition 40. Let M be a Noetherian R-module and let f : M → M be a
surjective homomorphism. Then f is an isomorphism.
Proof. Let f n := f ◦ f ◦ . . . ◦ f the composition of f with itself n times. Note
that f n is surjective for all n and moreover:

ker f ⊆ ker f 2 ⊆ . . . ⊆ ker f n ⊆ . . . ⊆ M.

Since M is Noetherian there exists n ∈ N such that

ker f n = ker f n+1 .

Let now x ∈ ker f . Since f n is surjective there exists y ∈ M such that x = f n (y).
Apply f to get
0 = f (x) = f n+1 (y)
3 Artinian Rings 49

and hence y ∈ ker f n+1 = ker f n . This means:

0 = f n (y) = x,

i.e. f is injective and so it is an isomorphism.

Remark 2. This is no longer true if we switch surjective and injective in Propo-
sition 40. For instance consider:

Z /Z
 /2
1

which is injective but not surjective.

3 Artinian Rings
Definition. If the set of ideals in a ring R satisfies DCC then R is said to be
Artinian.
Example 54. (1) Fields are Artinian.
(2) Any finite ring, e.g. Z/nZ is Artinian.
(3) Let k be a field and
R = k[x1 , . . . , xn ]/I
be a quotient such that dimk ≤ ∞. Then R is Artinian.
(4) Let k be a field and k ⊆ R be a subring of Mn (k) the n × n matrices with
coefficients in k. Then R is Artinian since dimk R ≤ ∞.
Remark 3. If R is Artinian, then it is Noetherian. However the converse is not
true, for instance R = k[x] is Noetherian but

R ⊇ (x) ⊇ (x2 ) ⊇ . . .

is a descending chain which does not stabilize.

Remark 4. If R1 , R2 are Artinian then R1 × R2 is Artinian.
Proposition 41. if R is Artinian, then every prime is maximal and there are
only finitely many maximal ideals.
Proof. Let p ∈ Spec(R) and pass to R/p. Relabel it as R so that without loss
of generality we can assume that R is a domain and we have to prove that it is
in fact a field. Assume not and pick x ∈ R, x 6= 0 which is not a unit. Then
consider
R ⊇ (x) ⊇ (x2 ) ⊇ . . .
that has to stabilize since R is Artinian. So there is n ∈ N such that (xn ) =
(xn+1 ). So there exists a ∈ R such that xn = axn+1 and, since R is a domain,
50 Chain Conditions

we can cancel xn and get 1 = ax, i.e. x is a unit. This is a contradiction,

therefore R is a field and p is maximal.
Assume now that there are infinitely many distinct maximal ideals in R, say
{mi }∞
i=1 . Consider:

m1 ⊇ (m1 ∩ m2 ) ⊇ (m1 ∩ m2 ∩ m3 ) ⊇ . . .
Again since R is Artinian there exists k ∈ N such that
(m1 ∩ m2 ∩ . . . ∩ mk ) = (m1 ∩ m2 ∩ . . . ∩ mk ∩ mk+1 ).
But this means (m1 ∩ m2 ∩ . . . ∩ mk ) ⊆ mk+1 and therefore, since they are
maximal (prime was enough) there exists i ∈ {1, . . . , k} such that mi ( mk+1 ,
contradicting the maximality of mi . Hence there are just finitely many maximal
ideals in R.
Theorem 42. Let R be a ring. The following facts are equivalent:
(1) R is Artinian.
(2) R is Noetherian and there exist only finitely many prime ideals, and all of
them are maximal.
Proof. (1) ⇒ (2) By Proposition 41 we only need to show that R is Noetherian.
List all the maximal ideals m1 , . . . , mn .
Claim. There exists k ∈ N such that (m1 . . . mn )k = 0.
Proof of the Claim. Set I = m1 . . . mn , then
I ⊇ I 2 ⊇ I 3 ⊇ . . . ⊇ I k = I k+1
for some k since R is Artinian. Assume I k 6= 0 and consider:
Λ := {J ⊆ R : JI k 6= 0}.
Note that I ∈ Λ, so Λ 6= ∅. Therefore there exists a minimal element J, and
this has to be a principal ideal, otherwise there exists x ∈ J such that xI k 6= 0
(this is because JI k 6= 0) and so (x)I k 6= 0 and (x) ⊆ J. So set J = (x). Notice
that
(xI)I k = xI k+1 = xI k 6= 0
so xI ∈ Λ and xI ⊆ (x) = J. By minimality it has to be xI = (x) and since
I = m1 . . . mn ⊆ m1 ∩ . . . ∩ mn = Jac(R) it has to be x = 0 by NAK. This is a
contradiction, hence I k = (m1 . . . mn )k = 0.
By Chinese Remainder Theorem we have:
R R R
R= ' k × ... × k .
(m1 . . . mn )k m1 mn
R is Artinian, hence each R/mki is Artinian, and if each R/mki is Noetherian,
then so is R. So assume R = R/mk for some maximal ideal m, so that we
reduced to the case in which there is only one maximal ideal, and its k-th power
is zero. To prove that R is Noetherian induct on the least k such that mk = 0:
3 Artinian Rings 51

• If k = 1 then m = 0 and R is a field, and hence Noetherian.

• If k > 1 then by induction R/mk−1 is Noetherian. Note that ann(mk−1 ) =
m, so that mk−1 is a R/m-module, i.e. a vector space. Now any vector
subspace of mk−1 is an ideal in R, so that dimR/m mk−1 ≤ ∞, since R is
Artinian, so DCC is satisfied on ideals. Hence mk−1 is a Noetherian R-
module and since both R/mk−1 and mk−1 are Noetherian, R is Noetherian
too.
(2) ⇒ (1) Let m1 , . . . , mn be the maximal ideals in R. These are all the primes
in R, so p
(0) = m1 ∩ . . . mn ,
so, since R is Noetherian, there exists N >> 0 such that (m1 . . . mn )N = 0. By
Chinese Remainder Theorem
R R
R' × ... N .
mN
1 m n

A product of Artinian rings is artininan, so it is enough to show that R/mN

is Artinian for some ring R with a unique maximal ideal m such that mN = 0.
Induct on N :
• If N = 1 then R is a field, and hence Artinian.
• If N > 1 by induction R/mN −1 is Artinian. Also mN −1 is a R/m-vector
space of finite dimension, since R is Noetherian. Let

R ⊇ I1 ⊇ I2 ⊇ . . .

be a descending chain of ideals. Then going modulo mN −1 :

R I1 + mN −1 I2 + mN −1
⊇ ⊇ ⊇ ...
mN −1 mN −1 mN −1
must stabilize, and also

R ∩ mN −1 ⊇ I1 ∩ mN −1 ⊇ I2 ∩ mN −1 ⊇ . . .

must stabilize because mN −1 is a finite dimensional vector space. There-

fore there exists M >> 0 such that

IM + mN −1 = IM +1 + mN −1 and IM ∩ mN −1 = IM +1 ∩ mN −1 .

This implies IM ⊆ IM +1 + mN −1 and so:

IM ⊆ IM ∩(IM +1 +mN −1 ) = IM +1 +(IM ∩mN −1 ) = IM +1 +(IM +1 ∩mN −1 ) ⊆ IM +1 ,

which means IM = IM +1 and the chain stabilizes.

52 Chain Conditions

Definition. A R-module M is said to be Artinian if it satisfies DCC on sub-

modules.
Remark 5. R is Artinian as a ring if and only if R is Artinian as a module over
itself.
Remark 6. If N ⊆ M is a submodule, then M is Artinian if and only if N and
M/N are Artinian.
3 Artinian Rings 53

Exercises
Chapter 5

Primary Decomposition

1 Definitions and Examples

Definition. An ideal I ⊆ R is said to be irreducible if

I = J ∩ K ⇒ J = I or K = I.

Definition. An ideal q ⊆ R is said to be primary if whenever xy ∈ q and

√
/ q, it has to be y ∈ q.
x∈
Example 55. In Z we have (m) ∩ (n) = (lcm(m, n)), hence an ideal (k) is
irreducible if and only if k 6= lcm(m, n) unless k = m or k = n. This is
equivalent to require that k = pl for some p ∈ Z prime and some integer l.
Primary ideals in Z are also of this form. In fact if q = (pl ) and xy ∈ q, then
√
pl |xy. Assume x ∈/ q = (p), then p 6 | x and hence pl |y, that is y ∈ q.
Remark 7. If q ⊆ R = Z, then

q primary ⇐⇒ q irreducible ⇐⇒ q = (pl ) for some p prime and some l ∈ N.

Proposition 43 (Noether). If R is a Noetherian ring and q ⊆ R is irreducible,

then q is primary.
√
Proof. Suppose not, and let ab ∈ q such that a ∈ / q and b ∈
/ q. Notice that it
has to be an ∈/ q for all n > 0. For n ∈ N consider:

q : an = {r ∈ R : ran ∈ q}

and notice that

q ⊆ (q : a) ⊆ (q : a2 ) ⊆ . . .
Since R is Noetherian the chain stabilizes, i.e. there exists k ∈ N such that
(q : ak ) = (q : ak+1 ) = . . .
Claim. q = (q : ak ) ∩ (q + Rak ) and furthermore q 6= (q : ak ) and q 6= (q + Rak ).
2 Primary Decomposition 55

Proof of the Claim. Clearly q ⊆ (q : ak )∩(q+Rak ). Let r ∈ (q : ak )∩(q+Rak ),

then r = c + dak , with c ∈ q and d ∈ R. Also

ak (c + dak ) = rak ∈ q,

therefore da2k ∈ q, i.e. d ∈ (q : a2k ) = (q : ak ). Finally dak ∈ q and hence

r = c + dak ∈ q.
Furthermore q 6= q : ak since bak = (ba)ak−1 ∈ q, hence b ∈ ((q : ak ) r q) and
also ak ∈ (Rak r q) ⊆ ((q + Rak ) r q).

Remark 8. The converse of Proposition 43 is not true. For instance in k[x, y]

the ideal (x, y)2 = (x2 , xy, y 2 ) is primary (a justification is given later) but it is
not irreducible, in fact:

(x, y)2 = (x2 , y) ∩ (x, y 2 ).

Proposition 44. Let R be a ring.

√
(1) If q is primary, then q = p is prime (we often say that q is p-primary).
√
(2) If q = m is maximal,then q is m-primary.
√ √
Proof. (1) Suppose that ab ∈ q and a ∈ / q. There exists n >> 0 such that
an bn ∈ q, but am ∈/ q for all m ≥ 1. Since q is primary it has to be bn ∈ q, and
√
hence b ∈ q, which is prime.
√
(2) Suppose ab ∈ q and a ∈ / m = q. Then m + Ra = (1), therefore there exists
√
r ∈ R and m ∈ m such that 1 = ra + m. Since m ∈ m = q there exists n ∈ N
such that mn ∈ q, therefore:

1 = 1n = (m + ra)n = mn + sa

for some s ∈ R. Finally

b = mn b + sab ∈ q
and so q is m-primary.
√
Remark 9. The converse of Proposition 44 (1) is not true in general if q=p
is not maximal.

2 Primary Decomposition
Definition. Let R be a ring. An ideal I is said to have a primary decomposition
if we can write
I = q1 ∩ . . . ∩ qn ,
where each qi is primary.
In this section we prove that every Noetherian ring has a primary decompo-
sition. Such a decomposition is not unique:
56 Primary Decomposition

Example 56.

(x2 , xy) = (x) ∩ (x2 , xy, y n ) for all n ≥ 2.

However it has some degree of uniqueness.

Theorem 45 (Noether). Let R be a Noetherian ring. Then every ideal I ⊆ R

has a primary decomposition.

Proof. Since irreducible ideals are primary it suffices to prove that every ideal
is a finite intersection of irreducible ideals. Let

Λ := {J ⊆ R : J is not an intersection of finitely many irreducible ideals}.

Suppose Λ 6= ∅ and choose I ∈ Λ maximal (this is possible since R is Noethe-

rian). I itself cannot be irreducible (since I ∈ Λ), hence I = J ∩ K with I ( J
and I ( K. Since I is maximal in Λ we have J ∈ / Λ and K ∈ / Λ. Therefore we
can write J and K as finite intersections of irreducible ideals, say J = J1 ∩. . .∩Jj
and K = K1 ∩ . . . ∩ Kk . But then:

I = J ∩ K = J1 ∩ . . . ∩ Jj ∩ K1 ∩ . . . ∩ Kk

can be written as a finite intersection of irreducible ideals. This is a contradiction

since I ∈ Λ, therefore Λ = ∅ and every ideal I ⊆ R has a primary decomposition.

The main effort in this section from now on is to try to make the primary
decomposition as unique as possible.

Lemma 46. Let q1 and q2 be p-primary ideals. Then q1 ∩ q2 is p-primary.

√ √ √ √
Proof. Let ab ∈ q1 ∩q2 and assume a ∈ / q1 ∩ q2 . Since q1 ∩ q2 = q1 ∩ q2 =
√ √
p we have that a ∈ / q1 = q2 = p. But they both are primary, therefore b ∈ q1
and b ∈ q2 , that is b ∈ q1 ∩ q2 and q1 ∩ q2 is p-primary.

Definition. A primary decomposition I = q1 ∩ . . . ∩ qn is minimal if

(1) No one of the qi ’s can be deleted, that is

\
qj 6⊆ qi
j6=i

for all i = 1, . . . , n.
√ √
(2) qi 6= qj whenever i 6= j.

Remark 10. It follows immediately from Noether’s Theorem 45 and Lemma

46 that if R is Noetherian, then every ideal I ⊆ R has a minimal primary
decomposition.
2 Primary Decomposition 57

Theorem 47. Let R be a Noetherian ring, let I ⊆ R be an ideal and let I =

q1 ∩ . . . ∩ qn be a minimal primary decomposition of I. Then the following facts
are equivalent for a prime p ⊇ I:
√
(1) p = qi for some i.

(2) There exists x ∈ R such that I : x = p.

(3) There exists an embedding R/p ,→ R/I.

√
Proof. (1) ⇒ (2) Let p = qi and let m ≥ 1 be such that pm ⊆ qi . Such m
exists since R is Noetherian. Set J := q1 ∩ . . . ∩ qˆi ∩ . . . ∩ qn . We have:

Jpm ⊆ Jqi ⊆ J ∩ qi = I.

Choose N ≥ 1 such that JpN ⊆ I and JpN −1 6⊆ I. This N exists since the
decomposition is minimal, therefore J 6⊆ I, and it is finite since N 6 m. Choose
y ∈ JpN −1 r qi . Notice that
 
\ \
I : y =  qj  : y = (qj : y) = R ∩ R ∩ . . . ∩ (qi : y) ∩ . . . ∩ R = (qi : y)
j j

because y ∈ JpN −1 ⊆ J ⊆ qj for all j 6= i. Also notice that yp ⊆ JpN ⊆ I by

definition of y, hence
p ⊆ I : y = qi : y.
Finally, let z ∈ qi : y, then zy ∈ qi , with y ∈
/ qi . Since qi is primary there exists
√
n ≥ 1 such that z n ∈ qi , which means z ∈ qi = p. Therefore qi : y ⊆ p and
this forces p = I : y.
(2) ⇒ (3) Note that x ∈ / I. Define:

ϕ : R → R/I
r 7→ rx

Then ϕ is a R-homomorphism and ker ϕ = p, thus there is an induced monomor-

phism:
ϕ̂ : R/p ,→ R/I.
(3) ⇒ (1) Consider
ψ : R → R/I
such that ker ψ = p. Set x = ψ(1) modulo I, then x ∈
/ I. Note that

I : x = (q1 : x) ∩ . . . ∩ (qn : x) = p,

so that, using the fact that p is prime, there exists i such that

qi : x ⊆ p = (q1 : x) ∩ . . . ∩ (qn : x) ⊆ qi : x,
58 Primary Decomposition

and clearly
√ √
qi : x ⊆ p ⊆ qi : x.
√ √
This implies qi : x = p. As a consequence x ∈ / qi , otherwise p = qi : x = R,
√
which gives a contradiction. Then, let y ∈ qi : x, so that there exists n ≥ 1
such that xy n ∈ qi . But x ∈
/ q and qi is primary. This implies (y n )m ∈ qi for
√ i √ √
some m ≥ 1, that is y ∈ qi . Finally, one always has qi ⊆ qi : x, so that
√ √
qi : x = qi = p.
Corollary 2. Let R be a Noetherian ring and let I ⊆ R be an ideal. If I =
q1 ∩ . . . ∩ qn = q01 ∩ . . . ∩ q0m p
are minimal prime decompositions, then n = m
√
and, after re-indexing, qi = q0i for all i = 1, . . . , n.
Proof. Let a := {p : p is prime and p = I : x for some x ∈ R}. Then
√ √ p p
{ q1 , . . . , qn } = a = { q01 , . . . , q0m }
√ p
and they all are distinct. So it has to be n = m and qi = q0i after re-
indexing.
Definition. Let R be a Noetherian ring and let I ⊆ R be an ideal. Let
I = q1 ∩ . . . ∩ qn be a minimal primary decomposition. Then the prime ideals
√ √
q1 , . . . , qn are called associated primes to R/I and we denote
√ √
Ass(R/I) := { q1 , . . . , qn }.

Proposition 48. Let R be a Noetherian ring and let I ⊆ R be an ideal. Then

x ∈ R is a zero divisor modulo I if and only if there exists p ∈ Ass(R/I) such
that x ∈ p. In other words
[
p = {zero divisors modulo I}.
p∈Ass(R/I)

Proof. Suppose x ∈ p for some p ∈ Ass(R/I). Since p is associated there exists

r ∈ R such that p = I : r. If x ∈ I then it is zero modulo I. So suppose
x ∈ R r I. Since x ∈ p = I : r, we have rx ∈ I. Notice that r ∈ / I since
I : r = p 6= (1), therefore x is a zero divisor modulo I.
Conversely assume r is a zero divisor modulo I. So there exists y ∈ / I such that
ry ∈ I = q1 ∩ . . . ∩ qn . Since y ∈ / I there exists qi such that y ∈
/ qi . Since
√
ry ∈ I ⊆ qi it has to be r ∈ qi ∈ Ass(R/I).
Definition. Let R be a ring and let I ⊆ R be an ideal. A prime ideal p ⊇ I is
said to be a minimal prime of I if there is no prime ideal q such that

I ⊆ q ( p.

We denote the set of all minimal primes of I by Min(I).

Proposition 49. Let R be a Noetherian ring and let I ⊆ R be an ideal. Then
minimal primes of I exist and they are a finite number. In fact, every minimal
prime of I is an associated prime of R/I.
2 Primary Decomposition 59

√
Proof. Let I = q1 ∩. . .∩qn be a minimal primary decomposition and let pi = qi
for all i = 1, . . . , n be the associated primes. Take pi minimal among {p1 , . . . , pn }
(minimal in the sense that there is no pj such that pj ( pi ). Notice that in
general there might be more than one prime among p1 , . . . , pn satisfying this
condition. In that case just pick one. Assume that p is a prime such that
I ⊆ p ⊆ pi . Then
I = q1 ∩ . . . ∩ qn ⊆ p
and, since p is prime, we get qj ⊆ p for some j. Also, taking radicals, we get
pj ⊆ p. So pj ⊆ p ⊆ pi and, by minimality of pi , it has to be pj = q = pi . This
proves that pi is a minimal prime of I, and therefore minimal primes of I exist.
Now take p ⊇ I a minimal prime of I. With the same argument we can show
that I ⊆ pj ⊆ p for some j, and hence pj = p by definition of minimal prime.
Therefore every minimal prime of I is an associated prime of R/I. In particular
this means that there are only finitely many minimal primes of I.

Remark 11. Not every associated prime is a minimal prime. For instance:

I = (x2 , xy) = (x) ∩ (x2 , y) ⊆ k[x, y] = R.

In this case Min(I) = {(x)} ( Ass(R/I) = {(x), (x, y)}.

Proposition 50. Let R be a Noetherian ring and let I ⊆ R be an ideal. Let

√
I = q1 ∩ . . . ∩ qn be a minimal primary decomposition and suppose pi = qi is
a minimal prime of I. Then:
[
qi = (I : s).
s∈p
/ i

Proof. Let r ∈ (I : s) for some s ∈/ pi . Then rs ∈ I ⊆ qi , and therefore r ∈ qi

because it is primary. T
Conversely notice that j6=i qj 6⊆ pi , otherwise there would be qj ⊆ pi and hence
pj ⊆ pi . The decomposition is minimal, so the two primes cannot be equal, but
this T
would contradict the fact that pi is aTminimal prime of I. So we can pick
n
s ∈ j6=i qj r pi . Let x ∈ qi , then xs ∈ j=1 qj = I, which means x ∈ (I : s)
for s ∈
/ pi .

Corollary 3. Primary components whose radicals are minimal primes are in-
dependent of the primary decomposition.
√
Proof. If I = q1 ∩ . . . ∩ qn is a minimal primary decomposition and pi = qi is
a minimal prime of I, then by Proposition 50 we have
[
qi = (I : s).
s∈p
/ i

In particular the right hand side is independent of the chosen decomposition,

and so is qi .
60 Primary Decomposition

Theorem 51 (Yao). Let R be a Noetherian ring and let I ⊆ R be an ideal. Let

q1 ∩ . . . ∩ qn = I = q01 ∩ . . . ∩ q0n
√ p
be minimal primary decompositions, with qi = q0i . Then, for all 1 ≤ i ≤ n

I = q1 ∩ . . . ∩ qi−1 ∩ q0i ∩ qi+1 ∩ . . . ∩ qn

is still a minimal primary decomposition of I.

Remark 12. We know from Chapter 1 that
√ \ \
I= p= p.
p ⊇ I p∈Min(I)
p prime

√
Then this is the unique minimal primary decomposition of I.
2 Primary Decomposition 61

Exercises
Chapter 6

Integral Closure

1 Definitions and Notation

Definition. Let R ⊆ S be rings. An element s ∈ S is integral over R if there
exists a monic polynomial f (t) ∈ R[t] such that f (s) = 0. Trivially, if r ∈ R,
then r is integral over R. We say that S is integral over R if every element of S
is integral over R.
√
Example 57. The irrational number 2 is integral over Z because it satisfies
the polynomial t2 − 2.

Example 58. The rational number 1/2 is not integral over Z.

Example 59. If e is idempotent in S, then e is integral over R. That is, e is a

root of t2 − t.

Definition. We say R is integrally closed in S if s ∈ S integral over R implies

that s ∈ R. The set of all elements of S integral over R is called the integral
closure of R in S. If R is a domain with fraction field K, then the integral
closure of R is the set of all elements in K that are integral over R. We say
that R is integrally closed if R is integrally closed in K.

Remark. If K ⊆ L are fields, then L is integral over K if and only if L is

algebraic over K.

Definition. For rings R ⊆ S, S is said to be module-finite over R if S is finitely

generated as an R-module. S is said to be finite as an R-algebra if there exists
s1 , s2 , . . . , sn in S such that S = R[s1 , s2 , . . . , sn ].

Remark. If S is module-finite, then S is finite as an R-algebra, but not the 

converse.

Example 60. Given a field k, the polynomial ring k[x] is finite as an algebra,
but not as a k-module
1 Definitions and Notation 63

Lemma 52. Let R ⊆ S ⊆ T be rings. Suppose S is module-finite over R and

T is module-finite over S, then T is module-finite over R.
P
Proof. Let T = St1 + · · · + Stn and S = Rs1 + · · · + Rsm . Then T = i,j Rsi tj .
That is, the si tj ’s generate T.

Theorem 53. Let R ⊆ S be rings. The following are equivalent:

(1) S is module-finite over R;

(2) S = R[s1 , s2 , . . . , sn ], where si ∈ S and each one is integral over R;

(3) S is a finitely generated R-algebra and S is integral over R.

Proof. (3)⇒ (2) Easy exercise.

(2)⇒ (1) Induct on n. For n = 1, we have that S = R[s] and there exists a
monic polynomial p(t) = tm + r1 tm−1 + · · · + rm with ri ∈ R such that p(s) = 0.
Claim. S = R · 1 + R · s + · · · + R · sm−1

Proof of the Claim. Let the right hand side of the claim be defined as N . Clearly,
1, s, s2 , . . . , sm−1 ∈ S, so that N ⊆ S. Conversely, if w > m − 1, write

sm = −(r1 sm−1 + · · · + rm )

Hence if we multiply by sw−m we get

sw = −(r1 sw−1 + · · · + rm sw−m ).

Inductively we get sw ∈ N for all w ∈ N. This proves the claim.

For n > 1, let T = R[s1 , . . . , sn−1 ] and S = T [sn ]. Notice that by induction,
T is module-finite over R and that S is module finite over T . Hence by lemma
52 S is module-finite over R.
(1)⇒ (3): Write S = R · s1 + · · · + R · sn , then S = R[s1 , . . . , sn ], so S is
a finitely generated R-algebra. To show S is integral Pn over R, let u ∈ S and
notice usi ∈ S. So there exists equations usi = j=1 rij sj where rij ∈ R for
i = 1, 2, . . . , n. This implies
n
X
(u · δij − rij ) = 0 (6.1)
j=1

for all i = 1, 2, . . . , n. Let

 
u − r11 −rij
A=
 .. 
. 
−rij u − rnn
64 Integral Closure

Then 6.1 can be written as

  

s1 0
 s2  0
A  .  = . .
   
 ..   .. 
sn 0

Multiplication by the adjoint of A yields

   
  s1 0
det(A) 0  s2  0
 ..   ..  =  ..  .
   
 .  .  .
0 det(A)
sn 0

Hence det(A) · si = 0 for all i. Thus det(A)S = 0. In particular, det(A) · 1 = 0.

But if t is a variable, and we let we let
 
t − r11 −rij
à = 
 .. ,

.
−rij t − rnn

then det(Ã)(t) is a monic polynomial in t with coefficients in R and det(Ã)(u) =

0. Hence u is integral over R.

Corollary 54. Let R ⊆ S be rings and T be the integral closure of R in S.

Then R ⊆ T ⊆ S and T is a ring.

Proof. Let u, v ∈ T , then u, v are integral over R. By (2) ⇒ (3) of theorem 53,
R[u, v] is integral over R, that is R[u, v] ⊆ T . Hence u · v and u + v are integral
over R. Thus T is a ring.

2 Going-Up
Remark. If W is a multiplicatively closed set in R and R ⊆ S is an integral
extension, then so is RW ⊆ SW . (Just use the same equations for 1s ∈ SW .)
Moreover, in general, if s is integral over R and w ∈ R, then ws in integral over
R since R[s] is integral over R. (You could also multiply the integral equation
for s by wn .)

Theorem 55 (Lying Over). Let R ⊆ S be an integral extension of rings, then

i∗ : Spec(S) → Spec(R) is surjective. I.e. for all q ∈ Spec(R), there exists
Q ∈ Spec(S) such that Q ∩ R = q.

Proof. Let q ∈ Spec(R). Then Rq → Sq is integral. Suppose we prove that

there exists Q ∈ Spec(Sq ) such that Q ∩ Rq = qRq . Set Q0 = Q ∩ S ∈ Spec(S)
2 Going-Up 65

and compute

Q0 ∩ R = (Q ∩ S) ∩ R
= Q ∩ (S ∩ R)
=Q∩R
= (Q ∩ Sq ) ∩ R
= Q ∩ (Sq ∩ R).

But if s/t ∈ Sq ∩ R (for s ∈ S and t ∈ R r q), there exist r ∈ R and u ∈ R r q

such that
us = utr ∈ R.
Therefore
s us
= ∈ Rq ,
t ut
and this shows Sq ∩ R = Rq ∩ R. Hence

Q0 ∩ R = Q ∩ (Sq ∩ R)
= Q ∩ (Rq ∩ R)
= (Q ∩ Rq ) ∩ R
= qRq ∩ R
=q

Let us change the notation: without loss of generality, (R, m) is local and
q = m. R → S is integral. First suppose mS 6= S. Then there exists a prime
ideal Q in S such that mS ⊆ Q. But then m ⊆ mS ∩ R ⊆ Q ∩ R ⊆ R. Since m
is maximal, Q ∩ R = mPproving the theorem for this case.
n
If mS = S, write i=1 ri si = 1 for ri ∈ m and si ∈ S. By theorem 53,
B = R[s1 , . . . , sn ] is a finite R-module. Notice that we have mB = B. Hence
NAK implies that B = 0, a contradiction.
Remark. Suppose R ⊆ S is an integral extension and J ⊆ S is an ideal. Then
the injection
R S
,→
J ∩R J
is an integral extension.
Theorem 56 (Going-Up). Let R ⊆ S be an integral extension of rings. Let
q0 ⊆ q1 ⊆ · · · ⊆ qn be a chain of primes in R. Let Q0 ∈ Spec(S) such that
Q0 ∩ R = q0 . Then there exists a chain Q0 ⊆ Q1 ⊆ · · · ⊆ Qn of primes in S
such that Qi ∩ R = qi for i = 0, 1, . . . , n.
Proof. By induction, it is enough to show there exists Q1 such that Q0 ⊆ Q1 ,
Q1 ∩R = q1 , Q0 ∩R = q0 , and q0 ⊆ q1 . By the remark, R/q0 ,→ S/Q0 is integral
and q1 /q0 ∈ Spec(R/q0 ). Lying over gives a prime Q1 /Q0 ∈ Spec(S/Q0 ) such
that Q1 /Q0 ∩ R/q0 = q1 /q0 . Retracting back to S and R gives us the desired
result.
66 Integral Closure

Theorem 57 (Incomparable). Let R ⊆ S be an integral extension of rings.

Suppose that Q, Q0 ∈ Spec(S) and that Q ⊆ Q0 . If Q ∩ R = Q0 ∩ R then Q = Q0 .

Proof. Consider R/Q ∩ R ⊆ S/Q (still integral). Without loss of generality,

R ⊆ S are domains, Q0 ∈ Spec(S), Q0 ∩ R = 0 and we want to prove Q0 = 0.
Let W = R − {0}. Then RW ⊆ SW is integral and Q0 SW is a proper prime
(since Q0 ∩ W = ∅). We have reduced to: k ⊆ S is integral where k is a field
and S a domain, Q0 ∩ k = 0. This implies Q0 = 0.
Hence by the following lemma the theorem is proved.

Lemma 58. Let R ⊆ S be an integral extension of domains. Then S is a field

if and only if R is a field.

Proof. Assume that R is a field and let u ∈ S, u 6= 0. There exists an equation

of least degree un + α1 un−1 + · · · + αn = 0, αi ∈ R. By choice of n, αn 6= 0.
Hence u(un−1 + · · · + αn−1 ) = −αn ∈ R and u is a unit in S.
Now assume that S is a field and let u ∈ R, u 6= 0. Since S is a field,
u−1 ∈ S. Since S is integral over R, there exists an equation

(u−1 )n + r1 (u−1 )n−1 + · · · + rn = 0

for ri ∈ R. Multiplication by un and solving for 1 shows that the inverse of u is

an element of R.

Definition. The Krull dimension of a ring R, denoted dim R, is the supremum

on n of the set of chains
p0 ( p1 ( . . . ( pn
where pi ∈ Spec(R).

Example 61. (1) If k is a field, dim k = 0.

(2) dim Z = 1.

(3) If k is a field, dim k[x] = 1.

Theorem 59. Let R ⊆ S be an integral extension of rings. Then Q ∈ Spec(S)

is maximal if and only if Q ∩ R = q is maximal in R.

Proof. Notice that R/Q ∩ R ⊆ S/Q is an integral extension of domains. The

rest follows for lemma 58.

Example 62. The polynomial ring S = k[x, y, z] where k is a field is integral

over R = k[x3 , y 5 , z 9 ].

Proof. Consider the following, polynomials: t3 − x3 , t5 − y 5 , t9 − z 9 ∈ R[t]. Use

theorem 53 so say that S is integral over R.

Theorem 60. Let R ⊆ S be an integral extension of rings, then dim R = dim S

3 Normalization and Nullstellensatz 67

Proof. Let P0 ( P1 ( · · · ( Pn be a chain in R. By lying over, there exists Q0

over P0 and by going up, there exists

Q0 ( Q1 ( ··· ( Qn

  
P0 ( P1 ( ··· ( Pn

and all the Qi ’s are distinct. Hence dim R 6 dim S.

Conversely,

Q0 ( Q1 ( ··· ( Qn ∈ Spec(S)

  
Q0 ∩ R ( Q1 ∩ R ( ··· ( Qn ∩ R

where the Qi ∩ R are distinct by incomparability. Thus dim R > dim S.

Example 63. (1) If R is an Artinian ring, then dim R = 0 (since every prime
is maximal).

(2) Consider the extension Z ⊆ Z[i]. Since Z[i] is integral over Z, dim Z[i] = 1.

Remark. If R is Noetherian and dim R = 0, then R is Artinian.

3 Normalization and Nullstellensatz

Theorem 61 (Neother Normalization Lemma). Let k be a field and S a finitely
generated k-algebra. Then there exists y1 , y2 , . . . , yd in S which are algebraically
independent over k such that S is integral (even module-finite) over the subring
k[y1 , y2 , . . . , yd ] = R ⊆ S.

Definition. For k a field and S a ring such that k ⊆ S, y1 , y2 , . . . , yd ∈ S are al-

gebraically independent over k if for any non-zero polynomial p(T1 , T2 , . . . , Td ) ∈
k[T1 , T2 , . . . , Td ], p(y1 , y2 , . . . , yd ) 6= 0. Equivalently, k[y1 , y2 , . . . , yd ] is isomor-
phic to k[T1 , T2 , . . . , Td ].

Example 64. Let k be a field and S = k[t2 , t3 ] ' k[x, y]/(y 2 − x3 ). We have
that t2 is algebraically independent over k, so let R = k[t2 ], i.e. a polynomial
ring in one variable. Notice that (t3 )2 − (t2 )3 = 0. For p(T ) = T 2 − t6 ∈ R[T ],
p(t3 ) = 0. Thus S is integral over R.
68 Integral Closure

Lemma 62. Let k be a field and f (x1 , . . . , xn ) ∈ k[x1 , . . . , xn ]. Then for all N
sufficiently large, there exists a change of variables,

x0n = xn
x0n−1 = xn−1 − xN
n
2
x0n−2 = xn−2 − xN
n
..
.
n−1
x01 = x1 − xN
n

such that when we write f (x1 , . . . , xn ) = g(x01 , . . . , x0n ),

g(x01 , . . . , x0n ) = α(x0n )L + (x0n )L−1 g1 (x01 , . . . , x0n−1 ) + · · · + gL (x01 , . . . , x0n−1 )

where α 6= 0, α ∈ k and g1 , g2 , . . . , gL ∈ k[x01 , . . . , x0n−1 ].

Proof. Write f = α∈I λα xα where α = (α1 , . . . , αn ) ∈ Nn , |I| < ∞, λα ∈ k,
P
α1 α2
λα 6= 0, x = x1 x2 · · · xα
α
n . Choose N > maxα∈I {αi }. Rewrite f using the
n

change of variable (notice that k[x1 , . . . , xn ] = k[x01 , . . . , x0n ]). So

n−1
N n−2 α
X
f= λα (x01 + xN
n )α 0
1 (x2 + xn )2 · · · (x0n−1 + xN α αn
n )1 xn .
α∈I

We have to rewrite as a polynomial in xn . The highest degree contribution from

α1 N n−1 +αn N n−2 +···+αn
the α-monomial is xN . Since N > αi for all αi appearing,
if L = α1 N n−1 + αn N n−2 + · · · + αn , then this is the base N expansion of L.
This is unique. Hence f becomes monic up to a non-zero element of k as no
cancellation occurs.
Example 65. Let f (x1 , x2 , x3 , x4 ) = x1 x4 − x2 x3 and x4 = x04 + x1 . Keep
x1 , x2 , x3 fixed. Then

f (x1 , x2 , x3 , x4 ) = x1 (x04 + x1 ) − x2 x3 = x21 + x04 x1 − x2 x3

Proof of Noether Normalization Lemma; Theorem 61. We can write S = k[t1 , . . . , tn ].

Induct on n. If n = 0 there is nothing to prove. Assume that n > 0.
Case 1. t1 , . . . , tn are algebraically independent over k.
If this is the case, then take yi = ti , then R = S.
Case 2. There exists a non-zero polynomial f (x1 , . . . , xn ) ∈ k[x1 , . . . , xn ] such
that f (t1 , . . . , tn ) = 0. With out loss of generality, using the lemma, we can
assume that

f = tL L−1
n + tn g1 (t1 , . . . , tn−1 ) + · · · + gL (t1 , . . . , tn−1 ).

Then we have an integral extension

integral
k ⊆ k[t1 , . . . , tn−1 ] ⊆ k[t1 , . . . , tn ] = S.
3 Normalization and Nullstellensatz 69

By induction, there exists y1 , . . . , yd ∈ k[t1 , . . . , tn−1 ]. algebraically independent

over k such that
integral integral
k[y1 , . . . , yd ] ⊆ k[t1 , . . . , tn−1 ] ⊆ S.

Therefore S is integral over k[y1 , . . . , yd ].

Example 66. Let k be a field and consider the ring k[x, y, u, v] such that
     
x
xu xv xu xv
u v = and det = 0.
y yu yv yu yv

If S = k[xu, xv, yu, yv], find a polynomial subring over which S is integral.
Let f (x1 , x2 , x3 , x4 ) = x1 x4 − x2 x3 . Then f (xu, yu, xv, yv) = 0. If we let
x04 = x4 − x1 as in example 65, then

f (x1 , x2 , x3 , x4 ) = x21 + x04 x1 − x2 x3 = g(x1 , x2 , x3 , x04 )

and g(xu, yu, xv, yv) = 0. Since g is monic in x1 , xu is integral over the polyno-
mial ring k[yu, xv, yv−xu] ⊆ S and yu, xv, yv−xu are algebraically independent.

Corollary 63. Let K and L be fields, L a finitely generated K-algebra. Then

L is algebraic over K. In particular, if K = K (the algebraic closure of K),
then L = K.

Proof. By the normalization lemma, there exists y1 , . . . , yd algebraically inde-

pendent over K such that K[y1 , . . . , yd ] ⊆ L is integral. Since L is a field, we
must have K[y1 , . . . , yd ] is a field. Hence d = 0 and L is algebraic over K.

Theorem 64. Let S = k[t1 , . . . , tn ] be a finitely generated k-algebra, and let m

be a maximal ideal of S. Then there exists a canonical embedding k ,→ S/m = L
and L is algebraic over k.

Proof. There exists a map


k  /S / S/m.
>
φ

Since k is a field and φ 6= 0, ker(φ) = 0. But L = k[t1 , . . . , tn ], ti = ti + m.

Apply the corollary to finish the proof.

Corollary 65. Let k be an algebraically closed field, R = k[x1 , . . . , xn ]. Then

every maximal ideal m of R has the form:

m = (x1 − α1 , x2 − α2 , . . . , xn − αn )

where αi ∈ k. Conversely, all such ideals are maximal.

70 Integral Closure

Proof. Let α = (α1 , . . . , αn ) and φα : k[x1 , . . . , xn ] → k be the evaluation

map f (x1 , . . . , xn ) 7→ f (α1 , . . . , αn ). This is a surjective ring homomorphism
so R/ ker(φα ) ' k; i.e. mα = ker(φα ) is maximal. Clearly (x1 − α1 , x2 −
α2 , . . . , xn − αn ) ⊆ mα . By the Taylor expansion,
X ∂f
f (x1 , . . . , xn ) = f (α1 , . . . , αn ) + (xi − αi ) (α) + other terms in (xi − αi )
i
∂xi

Hence f (x) ≡ f (α) (mod (x1 − α1 , x2 − α2 , . . . , xn − αn )). So all said ideals are
maximal.
Let m be maximal in R. By theorem 64 the map k = k ,→ R/m implies that
k = R/m. Hence there exists αi ∈ k where αi 7→ xi + m for all i = 1, . . . , n. I.e.
αi + m = xi + m, thus xi − αi ∈ m. Therefore (x1 − α1 , x2 − α2 , . . . , xn − αn ) ⊆ m
and by above we have equality.
Theorem 66 (Hilbert’s Nullstellensatz). Let k be a field, R = k[x1 , . . . , xn ],
and I an ideal of R. Then √ \
I= m
m⊇I

where m are maximal ideals in R.

Proof. First reduce to the case k is algebraically closed. Let k be the alge-
braic closure of k. We now have an integral extension R = k[x1 , . . . , xn ] ⊆
k[x1 , . . . , xn ] = S. Suppose we prove the theorem for S. Then for all I ⊆ R,
√ \
IS = n
n⊇IS

where n is maximal in S. Hence

√ \ \ \
IS ∩ R = ( n) ∩ R = (n ∩ R) = m
n⊇IS n⊇IS m⊇I
√ T
where m is maximal in R. And since IS = p for p ∈ Spec(S) containing IS,
by going up √ \ \ √
IS ∩ R = p∩R= p = I.
p⊇IS p⊇I
T
Hence without loss of generality, k = k. Let f ∈ m⊇I m for m maximal in
√
R. Suppose f ∈ / I and consider R[y] = k[x1 , . . . , xn , y].
√
Claim. ( I, yf − 1) = R[y].
√
If not, then ( I, yf −1) ⊆ m for m maximal in R[y], i.e. m = (x1 −α, . . . , xn −
αn , y − β). Hence I ⊆ (x1 − α, . . . , xn − αn ) and βf (α1 , . . . , αn ) − 1 = 0. But
by assumption, f ∈ (x1 − α, . . . , xn − αn ), and thus f (α1 , . . . , αn ) = 0. A
contradiction.
Now write X
1= hi (x, y)gi (x) + l(x, y)(xf − 1)
i
4 Going-Down 71

√ P
where gi (x) ∈ I. Substitute y with 1/f to get 1 = hi (x, 1/f )gi (x). Hence
for N  0 we have that
X √
fN = f N hi (x, 1/f )gi (x) ∈ I.
i
√
That is, f ∈ I.

4 Going-Down
Proposition 67. Let A ⊆ B be rings and C the integral closure of A in B. If
S ⊆ A is multiplicatively closed, then CS is the integral closure of AS in BS .
Proof. Clearly , CS is in the integral closure of AS in BS . Conversely, for any
x/s ∈ BS which is integral over AS we have
 x n a  x n−1 an
1
+ + ··· + = 0.
s s s s
This gives the following equality,
xn + a1 xn−1 + · · · + an sn−1
= 0 ∈ BS .
sn
Thus there is a t ∈ S such that tn (xn + a1 xn−1 + · · · + an sn−1 ) = 0 in B. That
is, tx is integral over A, i.e. tx ∈ C. Hence xs = tx
ts ∈ CS .

Recall. As defined on page 62, an integral domain is called integrally closed if

it is integrally closed in the fraction field.
Proposition 68. Let A be a domain. The following are equivalent:
(1) A is integrally closed;
(2) Ap is integrally closed for all primes p in A;
(3) Am is integrally closed for all maximal m in A.
Proof. Let K be the fraction field of A, and let C be the integral closure of
A in K. Let f : A ,→ C be the inclusion map. But f is an isomorphism
iff fp is an isomorphism for all p ∈ Spec(A) iff fm is an isomorphism for all
m ∈ Spec(A).
Remark. Let A ⊆ B be rings and A ⊆ C ⊆ B. Then C is the integral closure
of A in B if and only if the statement is true locally.
Definition. Let A ⊆ B be rings and I an ideal of A. Then we say x ∈ B is
integral over I if x satisfies
xn + a1 xn−1 + · · · + an = 0
with ai ∈ I.
72 Integral Closure

Lemma 69. Let A ⊆ C ⊆ B, A, B be rings, C the integral closure

√ of A in B,
and I an ideal of A. Then the integral closure over I in B is IC.

Proof. Choose x ∈ B integral over I. Then xn + a1 xn−1 + · · · + a0 = 0 where

ai ∈ I and thus
xn = −(a1 xn−1 + · · · + a0 ) ∈ IC.
√
I.e. x ∈ IC. √ Pm
Now choose x ∈ IC. Then xn = i=1 ai xi with ai ∈ I, xi ∈ C. Let
M = A[x1 , . . . , xm ]. Then M is finite as an A-module, M is faithful, and
xn ∈ M ⊆ IM . Therefore xn (and even more so x) is integral over I by Cayley
Hamilton.

Proposition 70. Let A ⊆ B be domains, A integrally closed in its fraction field

k, I an ideal of A and x ∈ B integral over I. Let

f (t) = tn + an−1
t + · · · + an ∈ k[t]
√
be the minimal polynomial of x over k. Then ai ∈ I.

Proof. First, x satisfies g(x) = 0 where g(t) = tm + b1 tm−1 + · · · + bm ∈ A[t],

bi ∈ I. This implies that f (t)|g(t). Next choose a field such that B ⊆ L and
f (t) = (t − x1 )(t − x2 ) · · · (t − xn ). Since f divides g, g(xi ) = 0 for all i. Hence
√
each xi is integral over I and thus all ai ∈ k are integral over I. That is, ai ∈ I
by the previous lemma.

Lemma 71 ([1], prop 3.16). Let A → B be a ring homomorphism and let p be

a prime ideal of A. Then p is the contraction of a prime ideal of B if and only
if pec = p .

Proof. If p = qc then pec = p. Conversely, if pec = p, let S be the image of A r p

in B. Then pe does not meet S, therefore its extension in S −1 B is a proper ideal
and hence is contained in a maximal ideal m of S −1 B. If q is the contraction
of m in B, then q is prime, q ⊇ pe and q ∩ S = ∅ (since qS −1 B = m ( S −1 B).
Hence qc = p.

Theorem 72 (Going-Down). Let A ⊆ B be domains, A integrally closed, and

B integral over A. Given a chain of primes p1 ⊇ p2 ⊇ · · · ⊇ pn in A and a
chain of primes Q1 ⊇ Q2 ⊇ · · · ⊇ Qm in B such that m < n and Qi ∩ A = pi for
i = 1, 2, . . . , m, then the chain of primes in B can be extended to Q1 ⊇ . . . ⊇ Qn
such that Qi ∩ A = pi for i = 1, 2, . . . , n

Proof. It is enough to show when n = 2 and m = 1. Further, by lemma 71 we

only need to show p2 BQ1 ∩ A = p2 .
It is easy to see that p2 BQ1 ∩ A ⊇ p2 . Conversely, for x ∈ p2 BQ1 , x = ys
where y ∈ p2 B and s ∈ B − Q1 . By lemma 69 we have that y is integral over
p2 . Thus by prop 70 the minimal polynomial of y over K (f.f. of A) is

f (t) = tn + an−1
t + · · · + an ∈ K[t]
4 Going-Down 73

√
with ai ∈ p2 = p2 . Since y = sx, f (sx) = 0. Thus s satisfies g(t) = 0 where
g(t) = f (tx). Let

1
h(t) = g(t) = tn + v1 tn−1 + · · · + vn ;
xn
h(s) = 0 and s = x1 y. This implies that h(t) is the minimal polynomial of s
over K. Hence vi ∈ A by previous prop (I = (1)). But vi = ai /xi . Hence
xi vi√= ai ∈ p√
2 . If x ∈
/ p2 then vi ∈ p2 implies that s is integral over p2 . Hence
s ∈ p2 B ⊆ Q1 = Q1 . A contradiction. So x ∈ p2 and we have equality.

Theorem 73 (Dimension of Finitely Generated k-algebras). Let R be a finitely

generated k-algebra (k a field). Let R be integral over a subring k[y1 , . . . , yn ], yi
algebraically independent over k. Then

(1) every chain of primes in R has length less than or equal to n. In particular,
dim(R) 6 n;

(2) if R is a domain, then every saturated chain of primes has length n;

(3) in particular, dim(k[y1 , . . . , yn ]) = n. So dim(R) = n.

Recall. Polynomial rings are UFD’s; every polynomial f ∈ k[x1 , x2 , . . . , xn ] is

uniquely (up to order and units) a product f1a1 f2a2 · · · fkak , where fi are irre-
ducible polynomials. In particular, if f is irreducible and f |g · h then f |g or f |h.
I.e. (f ) is a prime ideal. Consequently, if R is a polynomial ring and Q is a
prime minimal over (0), but Q 6= (0), then Q = (f ).

Lemma 74. Suppose R = k[x1 , . . . , xn ], f ∈ R and f 6= 0. Then R/(f ) is

integral over a polynomial ring in n − 1 variables.

Proof. Use a change of variables (lemma 62) so that with out loss of generality,

f = xln + xl−1
n g1 (x1 , . . . , xn−1 ) + · · · + gl (x1 , . . . , xn−1 ).

Then R/(f ) is integral over k[x1 , . . . , xn−1 ] ⊆ R/(f ); xi = xi + (f ). To show

there are no relations on the xi , we need to show there does not exist a non
zero g ∈ R in n − 1 variables such that g(x1 , . . . , xn−1 ) 6= 0 in R/(f ). But
g(x1 , . . . , xn−1 ) = g(x1 , . . . , xn−1 ). So the above holds if and only if g 6= 0 such
that f |g in k[x1 , . . . , xn ]. But for all h ∈ R, f · h always has an xln term. Hence
such a g does not exist.

Proof of Theorem 73. (1): Induct on n. Consider a chain of primes

Q0 ( Q1 ( · · · ( Qm

in R. By incomparability (theorem 57), there is a chain of primes

qo ( q1 ( · · · ( qm
74 Integral Closure

in k[x1 , . . . , xn ] where qi = Qi ∩ k[x1 , . . . , xn ]. With out loss of generality,

we can assume q0 = 0. Further, there exists a non zero element g ∈ q1 . If
g = g1a1 · · · gtat , gj irreducible, then (0) 6= (gi ) ⊆ q1 for some i. Thus we may
also assume that q1 = (g) for some irreducible element g.
By the lemma, k[x1 , . . . , xn ]/(g) is integral over a polynomial ring in n − 1
variables. Since
0 = q1 /q1 ( q2 /q1 ( · · · ( qm /q1
is chain of primes in k[x1 , . . . , xn ]/(g), by induction m − 1 6 n − 1. Hence
m 6 n.
(2): Again, induct on n. Consider a saturated chain of primes

Q0 ( Q1 ( · · · ( Qm

in R. By (1), m 6 n. As before, if we contracting to A = k[x1 , . . . , xn ], we get

a chain of primes
qo ( q1 ( · · · ( qm
in k[x1 , . . . , xn ] where qi = Qi ∩ A. Because k[x1 , . . . , xn ] is integrally closed,
going-down holds on A ⊆ R.
Therefore there does not exist a prime p in A between (0) and q1 . (This
would not allow our original chain to be saturated.) As above, let q1 = (f )
for some irreducible f ∈ A. Then, by the above lemma, we have the following
integral extensions:

k[z1 , . . . , zn−1 ] ⊆ k[x1 , . . . , xn ]/(f ) ⊆ R/Q1

where zi are indeterminants over k. By induction, every saturated chain of

primes in R/Q1 has length n − 1. This applies to

0 = Q1 /Q1 ( Q2 /Q1 ( · · · ( Qm /Q1 .

Hence n − 1 = m − 1, that is, n = m.

(3): Combine (1) and (2).

5 Examples
Example 67. Let R = k[x, y, z]/(x) ∩ (y, z). The minimal primes of R are xR
and (y, z)R. Notice that

R/xR ' k[x, y, z]/(x) ' k[y, z].

So saturated chains of primes ending in xR will have length 2. Also,

R/(y, z)R ' k[x, y, z]/(y, z) ' k[x].

So saturated chains of primes ending (y, z)R have length 1.

5 Examples 75

Example 68. Let R = k[x3 , y 3 , z 3 , x2 y, x2 z, xyz]. Notice that R is integral

over k[x, y, z], so dim(R) = 3.
Example 69. Let R = k[x, y, z, xt, yt, zt]. Notice that the extension R ⊆
k[x, y, z, t] is not integral (there is not integral relation for t). By Noether’s nor-
malization lemma (theorem 61), there exists an integral extension k[y1 , . . . , yd ] ⊆
R where the yi ’s are algebraically independent over k. Let L = k(y1 , . . . , yd )
and Q(R) be the quotient field of R. Since Q(R) is algebraic over L (check
this), y1 , . . . , yd form a transcendence basis for Q(R). Since all transcendence
basis’ have the same cardinality, if we can find a transcendence basis for Q(R),
we can determine the dimension of R. (For more on transcendence basis’, see
appendix A1 in [2])
Notice that Q(R) = k(x, y, z, t). Since x, y, z, t are algebraically independent
over k, the elements for a transcendence basis and thus dim(R) = 4.
Example 70. Let R = k[xs, ys, zs, xt, yt, zt]. Again, k[x, y, z, s, t] is not inte-
gral over k. So consider Q(R) = k(xs, ys, zs, st ). So dim(R) = 4.

Example 71. Let R = k[x2 y, y 2 z, xz]. Notice that

(x2 y)2 (xz) (y 2 z)2 (x2 y) (xz)4 (y 2 z)

= x5 , = y5 , = z5.
y2 z (xz)2 (x2 y)2

Since these are algebraically independent elements of Q(R) and the dimension
is bounded by 3, dim(R) = 3.
Proposition 75. Let A ⊆ B be rings and b ∈ B. The following are equivalent:

(1) b is integral over A;

(2) the image of b in B/P is integral over A/(P ∩ A) for all P in Spec(B);
(3) the image of b in B/P is integral over A/(P ∩ A) for all P in Spec(B)
minimal over (0).

Proof. (1) ⇒ (2) ⇒ (3) is clear. For (3) ⇒ (1), assume b is not integral over A.
Let W = {f (b) | f ∈ A[T ], f monic}. W is multiplicatively closed and 0 is not
in W by assumption. Hence there exists a prime ideal Q ∈ Spec(B) such that
Q ∩ W = ∅. Hence there exists a minimal prime P ∈ Spec(B) such that P does
not meet W . Then in B/P , b is not integral over A/P ∩ A since f (b) 6= 0 for
all monic f ∈ (A/P ∩ A) [T ].
76 Integral Closure

Exercises
(1) Let k be a field and R = k[xa1 11 · · · xan1n , . . . , xa1 m1 · · · xanmn ] where xi are in-
determinants over k and aij are integers. Prove that dim(R) = rank(aij ).
Chapter 7

Krull’s Theorems and

Dedekind Domains

1 Krull’s Theorems
Definition. Let R be a ring, p ∈ Spec(R). Then the height of p, denoted ht(p),
is:
sup{n ∈ N : ∃ p0 ( p1 ( . . . ( pn = p}.
Note that ht(p) = dim R/p.

Theorem 76. Let k be a field and let R be a finitely generated k-algebra. As-
sume R is a domain, then

ht(p) + dim R/p = dim R

for all p ∈ Spec(R).

Proof. Set n = dim R. We proved that all saturated chains of primes have
length n. Set s = dim R/p and let
p p1 ps
( ( ... (
p p p

be a saturated chain in R/p. Set t = ht(p) and let

0 = q0 ( q1 ( . . . ( qt = p

be saturated. Then:

0 = q0 ( q1 ( . . . ( qt = p ( p1 ( . . . ( ps

is saturated of length s + t, and hence s + t = n.

78 Krull’s Theorems and Dedekind Domains

Example 1. Consider the map

ϕ : k[a, b, c, d]  k[x2 y 2 , y 3 u3 v 3 , xy 2 uv, x3 y 5 u2 v 2 ].

Then ker ϕ = p is prime. What is dim k[a, b, c, d]p ? By Theorem 76 we have

dim k[a, b, c, d]
dim k[a, b, c, d]p = ht(p) = dim k[a, b, c, d] − = 4 − 2 = 2.
p
Theorem 77 (Krull’s Intersection Theorem). Let R be a Noetherian ring and
let I ⊆ R be an ideal. Then there exists i ∈ I such that
\
(1 − i) I n = 0.
n≥1

In particular, if either I ⊆ Jac(R) or R is a domain,

\
I n = 0.
n≥1

Proof. Set J := n≥1 I n . We have that JI = J, in fact let JI = q1 ∩ . . . ∩ qn be

T
a primary decomposition. It is enough to show that J ⊆ qi for all i = 1, . . . , n.
√
There are two cases: if I ⊆ qi , then I N ⊆ qi for N >> 0 and hence J ⊆
√
I N ⊆ qi . Instead, if I 6⊆ qi , we have J ⊆ qi since qi is primary. Therefore we
conclude by NAK.
Definition. If p ∈ Spec(R) is prime, then

p(n) := pn Rp ∩ R

is the p-primary component of pn and it is called n-th symbolic power of p. In

terms of the primary decomposition it is

pn = p(n) ∩ qi1 ∩ . . . ∩ qin

√
where p ( q ij .
Theorem 78 (Krull’s Principal Ideal Theorem and Height Theorem). Let R
be a Noetherian ring. Then:
(1) If p is a minimal prime over a principal ideal (x), then ht(p) ≤ 1.
(2) If p is minimal over an ideal (x1 , . . . , xn ), then ht(p) ≤ n.
Proof. (1) Assume p is minimal over (x) but ht(p) ≥ 2. Replace R with Rp , so
that without loss of generality we can assume p = m the maximal ideal of a local
ring R. By minimality over (x) we have that R/xR is Artinian. Since ht(m) ≥ 2
there exists a prime q such that q ( m and dim Rq > 0. Let q(n) = qn Rq ∩ R be
the n-th symbolic power and note that q(n) is q-primary for every n. Also we
have
q = q(1) ⊇ q(2) ⊇ . . . ⊇ q(n) ⊇ q(n+1) ⊇ . . .
1 Krull’s Theorems 79

Going modulo (x) we get a descending chain

q q(1) q(2) + (x) q(n) + (x) q(n+1) + (x)
= ⊇ ⊇ ... ⊇ ⊇ ⊇ ...
(x) (x) (x) (x) (x)
that must stabilize. In other words there exists N >> 0 such that (q(N ) +(x)) =
(q(N +1) + (x)) = . . . and in particular q(N ) ⊆ q(N +1) + (x). Let s ∈ q(N ) and
write it as s =prx + t, where t ∈ q(N +1) . Then rx = s − t ∈ q(N ) and, using
that x ∈/ q = q(N ) because m is minimal over (x) and q(N ) is q-primary, we
get r ∈ q(N ) . This means s ∈ xq(N ) + q(N +1) and therefore
q(N ) = xq(N ) + q(N +1) .
By NAK we get q(N ) = q(N +1) . Now localize at q:
qN Rq = q(N ) Rq = q(N +1) Rq = qN +1 Rq .
But now in Rq apply NAK to q·qN Rq = qN Rq to get qN Rq = 0, i.e. dim Rq = 0.
This is a contradiction, hence ht(m) ≤ 1.
(2) Induct on n. If n = 1 it is just (1). Assume now n > 1 and let p be minimal
over (x1 , . . . , xn ). p
We can localize at p, hence without loss of generality R is
local, p = m and (x1 , . . . , xn ) = m. Choose any prime q ( m such that no
other prime is between q and m. Note that since R is Noetherian such prime
exists unless ht(m) = 0, and in this case the theorem is proved. By minimality
of m there exists xi ∈ / q. Without loss ofpgenerality we can assume xn ∈
/ q. By
choice m is minimal over q + (xn ), hence q + (xn ) = m. This means that there
exists N >> 0 such that
xN N
1 = r1 xn + y1 , . . . , xn−1 = rn−1 xn + yn−1 ,

where yi ∈ q. Notice that (y1 , . . . , yn−1 ) ⊆ q.

Claim. q is minimal over (y1 , . . . , yn−1 ).
Proof of the Claim. If not there exists a prime p such that
(y1 , . . . , yn−1 ) ⊆ p ( q ( m,
but going modulo (y1 , . . . , yn−1 ) we get ht(m/(y1 , . . . , yn−1 )) ≥ 2, while m is
minimal over (y1 , . . . , yn−1 , xn ) since
p p
m = (x1 , . . . , xn ) = (y1 , . . . , yn−1 , xn ) ⊆ m.
This picture contradicts (1), hence q is minimal over (y1 , . . . , yn−1 ).
By induction ht(q) ≤ n − 1. Since q was an arbitrary prime below m with
no other primes in between we have ht(m) ≤ n.
Corollary 79. Let R be a Noetherian ring and let p ∈ Spec(R). Then ht(p) < ∞.
Proof. Since R is Noetherian p is finitely generated, say p = (x1 , . . . , xn ). Then
ht(p) ≤ n.
Corollary 80. Let R be a Noetherian ring. Then Spec(R) satisfies DCC.
80 Krull’s Theorems and Dedekind Domains

2 Dedekind Domains
Definition. Let D be a Noetherian integrally closed domain with dim D = 1.
Then D is called a Dedekind domain.

Example 2. (1) Z is a Dedekind domain.

(2) k[x], where k is a field, is a Dedekind domain.

(3) Z[i] is a Dedekind domain.

Proposition 81. Let D be a Dedekind domain. Then every ideal I can be

written uniquely (up to order) as a product I = q1 . . . qn , where all qi ’s are
√ √
primary ideals and pi = qi 6= qj = pj if i 6= j.

Proof. Without loss of generality we can assume I 6= (0). Let I = q1 ∩ . . . ∩ qn

√
be a primary decomposition. Since dim D = 1 we have qi = pi are all maximal
ideals, hence qi + qj = D for all i 6= j. By Chinese Remainder Theorem we get
I = q1 . . . qn . Conversely assume I = q1 . . . qn , with qi primary and pi 6= pj if
i 6= j. Again by Chinese Remainder Theorem I = q1 ∩ . . . ∩ qn and by Corollary
3 we get qi = IRpi ∩ R. Hence they are unique.

Proposition 82 (Structure of local Dedekind domains). Let (D, m) be a local

Dedekind domain. Then:

(1) There exists t ∈ D such that m = (t).

(2) Every non zero ideal is of the form mn for some n ≥ 0 (m0 = D).

Proof. (1) Choose any x ∈ m, x 6= 0. Then Ass(D/xD) = {m}, therefore there

exists an embedding

0 → D/m → D/xD
1 7→ ȳ
y y
This means (x : y) = m and thus xm ⊆ D. But xm is an ideal, and there are
two cases:
y
• xm ⊆ m, in which case xy is integral over D by the determinant trick.
But D is integrally closed, hence xy = z ∈ D, i.e. y = zx. Therefore
1 ∈ (x : y) = (x : zx) = m, and this is a contradiction.
y
• xm = D. Then there exists t ∈ m such that xy t = 1, i.e. x = ty. This
means
m = (x : y) = (ty : y) = (t).

(2) Let I be a non zero ideal. By Krull’s Intersection Theorem

\
I 6⊆ mn = (0).
n≥0
2 Dedekind Domains 81

Therefore there exists n ≥ 0 such that I ⊆ mn = (tn ) but I 6⊆ mn+1 = (tn+1 ),

where t is as in (1). We can write I = tn J, where J is the ideal I : tn . If
J ⊆ m = (t), then I ⊆ tn (t) = (tn+1 ), and this is a contradiction. Hence J = D
and I = (tn ).

Remark. Let R be a ring and let q be a p-primary ideal. Then qRp ∩ R = q. In

fact suppose r ∈ qRp ∩ R, then there exists s ∈
/ p such that sr ∈ q. This implies
r ∈ q since q is primary.
Theorem 83. Let D be a Dedekind domain. Then every ideal can be written
uniquely (up to order) in the form

I = pn1 1 . . . pns s ,

where p1 , . . . , ps are distinct primes.

Proof. By Proposition 81 there exist unique q1 , . . . , qs primary ideals, with
√
qi = pi , such that I = q1 . . . qs . Furthermore

qi = IDpi ∩ D = pni i Dpi ∩ D,

where the last equality followsp from Proposition 82 (2). It is enough to show
that pni i = pni i Dpi ∩ D. But pni i = pi maximal, hence pni is primary and the
theorem follows by the previous Remark.
82 Krull’s Theorems and Dedekind Domains

Exercises
Chapter 8

Completions and
Artin-Rees Lemma

1 Inverse Limits and Completions

Example 3. Let R be a Noetherian ring, let I ⊆ R be an ideal and assume
n
T
n≥0 I = 0 (for instance if R is a domain, or I ⊆ Jac(R)). If x, y ∈ R set
v(x, y) = sup{n : x − y ∈ I n } and define
1
d(x, y) := .
2v(x,y)
Claim. (R, d) is a metric space.
Proof of the Claim. Clearly d(x, y) = d(y, x) since if J is any ideal, then x − y ∈
J
T if and only if y − x ∈ J. Also d(x, y) = 0 if and only if x − y = 0 since
n
n≥0 I = 0. Finally assume
1 1
= d(x, y) and = d(y, z).
2k 2l
This means x − y ∈ I k r I k+1 and y − z ∈ I l r I l+1 . Therefore
x − z = (x − y) + (y − z) ∈ I min{k,l} .
Therefore
1 1 1
d(x, z) ≤ ≤ k
+ l = d(x, y) + d(y, z).
2min{k,l} 2 2

Remark. (1) What are the open balls Bε (x) in this metric space? Without
loss of generality assume ε = 21k , then
1 1
B 1 (x) = {y ∈ R : d(x, y) < 2k
} = {y ∈ R : d(x, y) ≤ 2k+1
} =
2k

= {y ∈ R : x − y ∈ I k+1 } = x + I k+1 .
84 Completions and Artin-Rees Lemma

(2) If we put the product topology on R × R, then both

+:R×R→R and · : R × R → R

are continuous. In fact let x + I k be a basic open set around x ∈ R and

let (y, z) ∈ +−1 (x + I k ). This means y + z ∈ x + I k , and therefore

(y + I k ) + (z + I k ) ⊆ x + I k .

Hence +−1 (x + I k ) contains (y + I k ) × (z + I k ) around (y, z), i.e. + is

continuous. Similarly for the product.

Definition. Let R and I ⊆ R be as above. The metric d defines a topology

on R called I-adic topology. Also a null Cauchy sequence {xn } is a Cauchy
sequence (with respect to the metric d) such that for all ε > 0 there exists
k ∈ N such that for n ≥ k d(xn , 0) < ε. We can now form a new object

bI := {Cauchy sequences in R}/{null Cauchy sequences in R}.

R

bI the completion of R with respect to I and it is obtained from R by

We call R
bI is a ring since R itself is a ring.
formally adjoining the ”limit points”. Also R

Remark. More explicitly, how does a Cauchy sequence in (R, d) look like? For
a sequence {xn } in R to be Cauchy means that for all ε = 21l there exists k ∈ N
such that for all n, m ≥ k we get
1
d(xn , xm ) < ⇐⇒ xn − xm ∈ I l+1 ⇐⇒ xn + I l+1 = xm + I l+1 .
2l
So let yl+1 +I l+1 be the coset which is in the stable value of xn +I l+1 as n → ∞.
Note that under the natural map πl : R/I l+1  R/I l we have

πl+1 (yl+1 + I l+1 ) = yl + I l .

Algebraically, this is an inverse limit.

Definition. Let R be a ring and {Mn }∞

n=1 be a collection of R-modules with
maps ϕn+1 : Mn+1 → Mn . Then the inverse limit is
!
Y Y
lim Mn := ker D : Mn → Mn ,
←−
n n

where D is the R-homomorphism defined as follows:

D((m1 , m2 , . . .)) = (m1 − ϕ2 (m2 ), m2 − ϕ3 (m3 ), . . .).

Remark. Notice that

bI ' lim R/I n ,
R
←−
1 Inverse Limits and Completions 85

where πn+1 : R/I n+1 → R/I n is the natural projection map. In fact let {xn }
be a Cauchy sequence in the I-adic topology and recall that for large m the
coset xm + I n has a stable value. Choose a representative yn + I n , so that
yn + I n = yn+1 + I n for all n. Hence, corresponding to {xn } there is an element
in lim R/I n , i.e. a sequence of cosets
←−

πn+2 πn+1
... / yn+2 + I n+2 / yn+1 + I n+1 / yn + I n / ...

In this correspondence, null Cauchy sequences correspond to the zero element in

lim R/I n . Moreover, this correspondence preserves the ring operations. Finally,
←−
given an element y ∈ lim R/I n , we have y = {(yn + I n )n } ∈ n R/I n and also
Q
←−

yn+1 − yn ∈ I n for all n,

i.e. {yn } is a Cauchy sequence.

Definition. Let R and I ⊆ R be as above and let M be a R-module. We can

define a pseudo-metric on M using I n M instead of I n (it will be a metric if
n
T
n≥0 I M = 0). Then the completion of M with respect to I is

cI = lim M/I n M.
M
←−

In =
T
Proposition 84. Let R be a ring and let I ⊆ R be an ideal such that n≥0
0. Then
bI ⊆ Jac(R
IR bI ).

bI for all x ∈ I. Notice that

Proof. It is enough to show that 1 − x is a unit in R

1 X
= xi
1−x
i≥0

Pn
and define sn = i=0 xi . Then {sn } is a Cauchy sequence since sn+1 − sn =
xn+1 ∈ I n+1 , hence there exists s ∈ R
bI which is the limit of sn , i.e. s − sn ∈
n bI
I R . Therefore

(1 − x)s − (1 − x)sn = (1 − x)s − (1 − xn+1 ) ∈ I n R

bI .

Since xn+1 ∈ I n+1 we get

(1 − x)s − 1 ∈ I n R
bI for all n

and hence
(1 − x)s = 1 bI .
in R
86 Completions and Artin-Rees Lemma

2 Artin-Rees Lemma
L
Definition. A ring S is (non-negatively) graded if S = i≥0 Si as an abelian
group, and Si · Sj ⊆ Si+j for all i, j ≥ 0. In particular S0 is a ring and each
graded piece Sj isL
an S0 -module. An S-module M is graded if it can be written
in the form M = j Mj , with Si Mj ⊆ Mi+j .
Example 4. (1) S = A[x1 , . . . , xn ] is a graded ring with Sj the A-module
spanned by the homogeneous polynomials of degree j.
(2) If R is a ring and I ⊆ R is an ideal we define

R(I) := R ⊕ I ⊕ I 2 ⊕ · · ·

the Rees Ring of I. One can artificially put a variable t in to keep track
of the grading. In this way:

R(I) ' R ⊕ It ⊕ I 2 t2 ⊕ . . . = R[It] ⊆ R[t].

Remark. Let R be a ring and let I ⊆ R be an ideal. If R is Noetherian, so is

R(I).
Proof. By assumption I is finitely generated, say I = (x1 , . . . , xn ). Then

R(I) = R[x1 t, . . . , xn t] ⊆ R[t].

Hence there exists a surjection

R[T1 , . . . , Tn ]  R(I)
Ti 7→ xi t

By the Hilbert Basis Theorem 37 R[T1 , . . . , Tn ] is Noetherian, and so is R(I).

Definition. Let R be a ring and let M be a finitely generated R-module. Define

M(I) := M ⊕ IM ⊕ I 2 M ⊕ . . .

Then M(I) has the structure of a graded R(I)-module if, given itn ∈ I n tn ⊆
R(I) and m ∈ I j M , we set

(itn )m = im ∈ I n+j M.

Remark. Notice that if M = Rm1 + . . . + Rmk , then

M(I) = R(I)m1 + . . . + R(I)mk .

Theorem 85 (Artin-Rees Lemma). Let R be a Noetherian ring, let I ⊆ R be an

ideal and let N ⊆ M be finitely generated R-modules. Then there exists k ∈ N
such that for all n ≥ K

I n M ∩ N = I n−k (I k M ∩ N ).
3 Properties of Completions 87

I j M as above. Define N to
L
Proof. Consider R(I) = R[It] and M(I) = j≥0
be the following R(I) submodule of M(I):

N = N ⊕ (IM ∩ N ) ⊕ (I 2 M ∩ N ) ⊕ . . .

Notice that I(I n M ∩ N ) ⊆ I n+1 M ∩ N , so that M is a R(I)-module. Since

R(I) is Noetherian (by Remark 2) and M(I) is finitelyP generated, so is jN . Say
N = R(I)xi +. . .+R(I)xl , with xi ∈ N , and xP i = j xij , where xij ∈ I M ∩N ,
and all but finitely many are zero. So N = i,j R(I)xij , and without loss of
generality we can assume x1 , . . . , xl are homogeneous, say deg xi = ni , i.e.
xi ∈ I ni M ∩ N . Let k = max{n1 , . . . , nl }.
Claim. With this choice of k we have I n M ∩ N = I n−k (I k M ∩ N ) for all n ≥ k.
Proof of the Claim. Notice that I n−k (I k M ∩ N ) ⊆ I n M ∩ N for all n ≥ k.
Conversely let n ≥ k and let u ∈ I n M ∩ N . Write it as
l
X
u= ri xi ,
i=1

where ri ∈ R(I), and without loss of generality they can be chosen homogeneous.
Hence deg ri = n − ni , that is

ri ∈ R(I)n−ni = I n−ni tn−ni .

This means that

l
X l
X
u∈ I n−ni xi ⊆ I n−ni (Iin M ∩ N ) ⊆ I n−k (I k M ∩ N ).
i=1 i=1

3 Properties of Completions
Definition. Let R be a ring and let {An } and {Bn } be inverse limit systems
of R-modules. We say that α : {An } → {Bn } is a morphism of inverse limits
(α = {αn }) if:
αn+1
An+1 / Bn+1

 
An
αn
/ Bn

where each αi is a R-homomorphism. In this case α induces a homomorphism

α
lim An −→ lim Bn .
←− ←−
88 Completions and Artin-Rees Lemma

Lemma 86 (Snake Lemma). Let R be a ring and suppose we have the following
exact diagram of R-modules:

A
α /B β
/C /0

f g h
 0
 β0

0 / A0 α / B0 / C0

Then there exists a morphism δ : ker h → cokerf such that

α0 β0
ker f
α / ker g β
/ ker h δ / cokerf / coker g / coker h
α
is exact. Furthermore, if α is one-to one, then so is (ker f → ker g). Similarly,
β0
if β 0 is surjective so is (coker g → coker h).
Proof. Diagram chasing.
Lemma 87. Let {An }, {Bn } and {Cn } be inverse limit systems and assume
α β
that {An } → {Bn } → {Cn } are such that

0 / An αn
/ Bn βn
/ Cn /0

are all short exact sequences. Then

(1) The following sequence is exact

0 / lim An α / lim Bn β
/ lim Cn
←− ←− ←−

(2) If An+1 → An is surjective for all n, then β is surjective.

Q Q
Proof. Let dA : An → An be the map such that
lim An = ker dA .
←−

Similarly define dB and dC for {Bn } and {Cn }. We have the following row-exact
diagram:
/ αn Y
/ βn
/ /0
Y Y
0 An Bn Cn

dA dB dC

Y Y Y
0 / An
αn
/ Bn
βn
/ Cn /0

By Snake Lemma we get an exact sequence

0 / lim An α / lim Bn β
/ lim Cn / coker dA .
←− ←− ←−

This proves (1). Also, if An+1 → An is surjective for all n, we have coker dA = 0,
and hence (2) follows.
3 Properties of Completions 89

Definition. Let R be a ring and let I ⊆ R be an ideal. A sequence {In } of

ideals I1 ⊇ I2 ⊇ . . . is said to be cofinal with {I n } if for all n there exists k such
that Ik ⊆ I n , and conversely for all n there exists k such that I k ⊆ In .
Remark. If {In } is cofinal with {I n }, the I-adic topology {x + I n } is the same
as the topology determined by taking neighborhood basis of x to be {x + In }.
So the completions are isomorphic, i.e.
bI = lim R/I n ' lim R/In .
R
←− ←−

Likewise, for any chain M1 ⊇ M2 ⊇ . . . of R-modules cofinal with {I n M }, we

get
cI = lim M/I n M ' lim M/Mn .
M
←− ←−

Lemma 88. Let R → S be a ring homomorphism. Then S is flat over R

if and only if, given finitely generated R modules M and N and an injection
i i⊗1
0 → N → M , then 0 → N ⊗R S → M ⊗R S is exact.
i⊗1
Proof. If S is flat, then by definition 0 → N ⊗R S → M ⊗R S is exact for
i
all injections 0 → N → M , with M and N finitely generated R-modules.
i
P 0 → N 
Conversely, assume S is not flat. Then there exists an injection →M
i⊗1 l
such that 0 → N ⊗R S → M ⊗R S is not exact. Suppose i⊗1 i=1 ni ⊗ si = 0
i00
Plet N0 := Rn1 + . . . Rnl ⊆ N . Then 0 → N0 → M is exact and i ⊗
and
l
1 i=1 ni ⊗ si = 0. By definition, M ⊗R S = R[m, s : m ∈ M, s ∈ S]/Z,
where Z ⊆ R[m, s : m ∈ M, s ∈ S] is a submodule. Therefore
l
! l l
X X X
0 0
i ⊗1 ni ⊗ si = 0 ⇐⇒ i (ni ) ⊗ si = 0 ⇐⇒ [i0 (ni ), si ] ∈ Z.
i=1 i=1 i=1
Pl
Hence there exist rj ∈ R and zj ∈ Z such that i=1 [i0 (ni ), si ] = j rj zj . Let
P
M0 ⊆ M be the submodule generated over R by i0 (ni ) and all elements of M
appearing in the zj ’s. Then

0 / N0 / M0

is exact, andPN0 and M0 are finitely generated R-modules. By assumption,

l Pl
since i0 ⊗ 1 n
i=1 i ⊗ si = 0, we have n
i=1 i ⊗ si = 0 in N0 ⊗R S, and hence
it is zero in N ⊗R S.
Theorem 89. Let R be a Noetherian ring and let I ⊆ R be an ideal such that
n
T
n≥0 I = 0. Then
(1) If we have a short exact sequence of R-modules

0 /N α /M β
/L / 0,
90 Completions and Artin-Rees Lemma

the following sequence

0 /N
bI α /M
cI β
/L
bI /0

is also exact.
(2) If M is finitely generated, then
cI ' M ⊗R R
M bI .

bI is flat.
(3) The map R → R
Proof. (1) Tensor the short exact sequence with R/I n :

N αn
/ M βn
/ L / 0.
I nN I nM I nL
This sequence is exact since tensor product is right exact. Starting from this
sequence we get the following exact diagram (considering the kernels of the two
surjections)

/ N αn
/ M βn
/ L /0
0
I nM O∩N I nM
O I nO L
πn πn πn

/ N αn
/ M βn
/ L /0
0
I n+1 M ∩ N I n+1 M I nL
This gives a short exact sequence of inverse limits and, moreover, the natural
maps πn are all surjective, therefore, from Lemma 1, we get the following short
exact sequence

/ lim N /M
cI /L
bI / 0.
0
←− I nM ∩N
Clearly, for all n ∈ N, I n N ⊆ I n M ∩ N . Finally, by Artin-Rees Lemma, there
exists k such that I n M ∩ N ⊆ I n−k N for all n ≥ k. Therefore {I n M ∩ N } is
cofinal with {I n N } and hence
N N bI.
lim ' lim n = N
←− I nM ∩ N ←− I N

(2) Notice that we have always a map

bI → M
M ⊗R R cI ,

and if M is free it is clearly an isomorphism. Since M is finitely generated there

exists a short exact sequence

0 /K / Rn π /M / 0,
3 Properties of Completions 91

where K = ker π. Since R is Noetherian, K is finitely generated, therefore there

exists an exact sequence

Rm
α / Rn π /M / 0.

By (1) we get the following exact sequence

bI )m
(R
α / (R
b I )n π /M
cI /0
O
b b
O O
' '

R m ⊗R R
bI / R n ⊗R R
bI / M ⊗R R
bI /0

and (2) follows by Five Lemma.

(3) Assume 0 → N → M is an exact sequence of finitely generated R-modules.
Then, by (1) and (2), we get

0 /N
bI /M
cI

bI is flat by Lemma 88.

is exact, and therefore the map R → R
Theorem 90. Let R be a Noetherian ring and let I ⊆ R be an ideal such that
n bI is Noetherian.
T
n≥1 I = 0. Then R
Proof. One can prove that, if I = (a1 , . . . , al ), then

bI ' R[[x1 , . . . , xl ]]
R .
(x1 − a1 , . . . , xl − al )
Then it suffices to prove that R[[x1 , . . . , xl ]] is Noetherian. Induct on l, so that
it is enough to show that R[[x]] is Noetherian. We will show that every prime
p ∈ Spec(R[[x]]) is finitely generated. Define p0 ⊆ R as p0 := {g(0) : g ∈ p}
(the constant terms). Then p0 ⊆ R is finitely generated, say p0 = (a1 , . . . , an ).
Then there exists fi ∈ p such that fi (0) = ai for all i. Two cases are possible:
• If x ∈
/ p. In this case we claim that p = (f1 , . . . , fn ), so that p is finitely
generated. Let g ∈ p, so that g(0) ∈ p0 and hence
n
X
g(0) = ri0 ai ,
i=1

for some r10 , . . . , rn0 ∈ R. Then

n
X
g(x) − ri0 fi (x) = xg1 (x).
i=1

But p is prime and x ∈

/ p, hence g1 ∈ p. Repeat the process to get
n
X
g1 (x) − ri1 fi (x) = xg2 (x),
i=1
92 Completions and Artin-Rees Lemma

so that
n
X
g(x) − (ri0 + ri1 x)fi (x) = x2 g2 (x).
i=1

Inductively, there exist rij ∈ R such that

n
X
g(x) − (ri0 + ri1 x + . . . + xj rij )fi (x) = xj+1 gj+1 (x) for all j.
i=1

Therefore
n X
X ∞ \
g(x) − ( xj rij )fi (x) ∈ (j n ) = 0,
i=1 j=1 j≥0

that is g ∈ p.
• If x ∈ p, then for g(x) ∈ p write g(x) = g(0) + xh(x), for some h(x) ∈
R[[x]]. This implies g(0) ∈ p. But then

(p0 , x) ⊆ p ⊆ (p0 , x),

i.e. p = (p0 , x) = (a1 , . . . , an , x) is finitely generated.

3 Properties of Completions 93

Exercises
Bibliography

[1] M. F. Atiyah and I. G. Macdonald. Introduction to commutative algebra.

Addison-Wesley Publishing Co., Reading, Mass.-London-Don Mills, Ont.,
1969.
[2] David Eisenbud. Commutative algebra, volume 150 of Graduate Texts in
Mathematics. Springer-Verlag, New York, 1995. With a view toward alge-
braic geometry.

[3] Christian Peskine. An algebraic introduction to complex projective geometry.

1, volume 47 of Cambridge Studies in Advanced Mathematics. Cambridge
University Press, Cambridge, 1996. Commutative algebra.
Index

I-adic topology, 84 homomorphism, 2

HomR (M, N ), 21
Spec(R), 6 ideal, 3
colon, 3
algebraically independent, 67 finitely generated, 3
annihilator, 29 principal, 3
Artinian, 49 product, 3
ascending chain condition (ACC), 44 sum, 3
associated primes, 58 idempotent, 14
Incomparable, 65
bilinear map, 23 integral, 62
integral closure, 62
Cauchy sequence, 84
integral closure of R in S, 62
chain, 7
integral over I, 71
cofinal, 89
integrally closed, 62
comaximal, 8
integrally closed in S, 62
completion, 84
inverse limit, 84
complex, 25
irreducible, 13, 54
Dedekind domain, 80 isomorphic, 2
descending chain condition (DCC), 44 isomorphism, 2
Direct product, 1
Direct Sum, 3 Jacobson radical, 7
domain, 6 kernel, 2
exact, 25 Krull dimension, 66

field, 1 localization, 34
field of fractions, 37 Lying Over, 64
finite as an R-algebra, 62
maximal, 6
flat, 29
minimal, 56
flat homomorphism, 29
minimal prime, 58
free module, 21
module, 19
Gaussian integers, 5 direct product, 20
Going-Up, 65 direct sum, 20
graded, 86 finitely generated, 21
generate, 21
height, 77 homomorphism, 20
96 INDEX

isomorphism, 20
module-finite, 62
morphism of inverse limits, 87
multiplicatively closed, 33

Neother Normalization Lemma, 67

nilpotent, 7
nilradical, 7
Noetherian, 44
non-zero divisor, 28

primary, 54
primary decomposition, 55
prime, 6
prime avoidance, 10
primitive, 16
principal ideal domain, 14

quotient module, 21
quotient ring, 4

reduced, 7
Rees Ring, 86
restriction of scalars, 20
right exactness, 26

short exact sequence, 25

submodule, 21
support of M , 40
symbolic power, 78
syzygies, 30

tensor product, 24

unique factorization domain, 13

unit, 7

Zariski topology, 6
Zorn’s Lemma, 7